Edward F. Plow¹, Czeslaw S. Cierniewski², Zihui Xiao¹, Thomas A. Haas¹, Tatiana V. Byzova¹

¹Joseph J. Jacobs Center for Thrombosis and Vascular Biology, Department of Molecular Cardiology, Cleveland Clinic Foundation, Cleveland, OH, USA, ²Department of Biophysics, Medical University in Lodz, Lodz, Poland

Correspondence to: Dr. Edward F. Plow, Joseph J. Jacobs Center for Thrombosis and Vascular Biology, NB50, Cleveland Clinic Foundation, 9500 Euclid Avenue, Cleveland, OH 44195, USA — Tel.: 216-445-8200; Fax: 216-445-8204; E-mail: plowe@ccf.org

Key words

Anti-thrombotic therapy, integrin aIIbb3, adhesive proteins, blood coagulation proteins

Summary

Because of its major role in regulating platelet functions and its prominence on the cell surface, integrin aIIbb3 has been the subject of intensive investigations. Such studies have provided substantial insights into its structure-function relationships and have led to the devel-opment of anti-thrombotic drugs that target the receptor. Nevertheless, recent findings have indicated that our understanding of the structure and function of aIIbb3 remains inadequate. This article addresses two aspects of still evolving aIIbb3 function: 1) the interface between aIIbb3 and the blood coagulation system, resulting from interaction of prothrombin with the receptor; and 2) the molecular basis for recognition of the RGD and the fibrinogen g-chain peptide ligands by aIIbb3. As illustrated by these two examples, there is still much to be learned about aIIbb3 if we are to fully appreciate its functions and its potential as a therapeutic target.

Introduction

In early 1999, we reviewed the status of aIIbb3 antagonism as an anti-thrombotic therapy (1). The future of the aIIbb3 inhibitors appeared to be rosy. Three parenteral agents, abciximab (Reopro™), eptifibatide (Integrilin™) and tirofiban (Aggrastat™), had received approval by the US Food & Drug Administration for use in patients with acute coronary syndromes or undergoing coronary interventions. Altogether, more than 32,000 patients had been treated with these drugs; and, in each of ten clinical trials, the aIIbb3 blockers had shown a significant (p <0.000000009) beneficial effect compared to placebo. Moreover, no fewer than ten oral aIIbb3 antagonists were at various stages of devel-opment, including Phase II and Phase III trials in man. These agents were potent inhibitors of platelet aggregation, were specific for aIIbb3, and displayed excellent bioavailability and pharmacokinetic profiles. Thus, it seemed justified to anticipate a new era in the treatment of thrombotic syndromes was in the offing. Now, slightly over two years later, a vastly different picture must be painted. Clinical trials of several oral aIIbb3 antagonists were terminated due to adverse effects or failed to attain the anticipated efficacy. A recent meta-analysis of the clinical experience with the oral aIIbb3 antagonists had indicated a significant increase, not a decrease, in mortality (2). The development of many of the oral aIIbb3 antagonists has been halted. Even abciximab, which had consistently shown great efficacy in patients, had yielded disappointing results in a recent and large clinical trial (3). Thus, at this moment in time, rather than forecasting a bright future for aIIbb3 antagonism, with an expanding series of parenteral and oral drugs for an expanding range of clinical indications, the entire field is at a virtual standstill. Indeed, the entire future of aIIbb3 antagonism is most dubious.

As one reflects on these developments, several explanations can be considered. Perhaps, aIIbb3 is not an appropriate target. This conclusion is difficult to reconcile with the innumerable studies which concluded that occupancy of aIIbb3 is the final common step in platelet aggregation regardless of the agonist (4), that platelet aggregation is a central event in thrombus formation, and that weaker inhibitors of platelet aggregation do show efficacy. Perhaps, the margin between the safety and efficacy of aIIbb3 antagonism is simply too narrow; in the effort to emphasize safety, efficacy has been compromised. Perhaps, the current set of antagonists, or at least some of them, are ineffective in blocking functions of aIIbb3 distinct from its role in platelet aggregation; induce undesirable responses from their interaction with the receptor; or exert detrimental effects independent of aIIbb3 altogether. As one attempts to sift through these possibilities, the conclusion is inescapable: our knowl-edge base of aIIbb3 and its antagonism must simply be insufficient to reach an informed decision. This article considers two aspects of aIIbb3 Function that are under investigation in our laboratories. These are fundamental studies of aIIbb3, which are very much works in progress and which could ultimately bear upon the efficacy of aIIbb3 antagonists. Therefore, these studies support our premise that there is much more to be learned about aIIbb3 and its antagonism.

Prothrombin: The New Ligand on the Block

A characteristic of the integrin family of adhesion receptors is the capacity of each family to bind multiple ligands (5, 6). aIIbb3 is no exception. By the early 1980’s, it was already known that fibrinogen (7, 8), von Willebrand factor (vWF) (9, 10) and fibronectin (11) were adhesive ligands for aIIbb3. The relative importance of these ligands in terms of mediating platelet aggregation continues to evolve (12, 13). The recent observation that thrombi form at nearly normal rates in mice deficient in both fibrinogen and vWF (14) suggests that other ligands and/or aIIbb3 independent mechanisms of platelet aggregation may function in vivo. Recent additions to the list of aIIbb3 ligands include cyr61, Fisp12 and L1-Ig6 (15, 16). Prothrombin has been added to this list based upon studies from our laboratories (17).

When aIIbb3 was purified from platelets and immobilized on microtiter plates, specific and saturable binding of prothrombin to the receptor, demonstrated using either a radiolabeled ligand or monoclonalantibodies (mAbs) to prothrombin, was observed. Fifty percent saturation occurred well below the plasma concentration of prothrombin (100 mg/ml). The stoichiometry of prothrombin binding to the immobilized aIIbb3 was similar to that of fibrinogen, suggesting that the same set of receptors could bind both ligands. Also similar to fibrinogen, prothrombin binding required divalent cations and was inhibited by thesame set of RGD peptides and mAbs to the receptor, including abciximab. Prothrombin does contain a RGD sequence in close proximity to the catalytic site of thrombin, and the crystal structures of thrombin and prethrombin 2 (18, 19) indicate that the RGD sequence is more exposed in the thrombin precursor than in the active enzyme. While the role of the RGD sequence remains to be rigorously defined, it is clear that it interacts with aIIbb3 via an RGD recognition specificity (inhibited by RGD peptides). Since prothrombin has the capacity to interact with phospholipids, which could remain associated with aIIbb3, we compared prothrombin binding to preparations with relatively high and relatively low phospholipid content and found no differences in the sensitivity of the binding to RGD peptides and mAbs to aIIbb3. The inhibition of the interaction by these reagents also militates against a role for receptor-associated lipids in mediating the interaction.

The binding of prothrombin to aIIbb3 also is demonstrable with intact platelets. Originally demonstrated using ¹²⁵I-prothrombin, the results of the flow cytometry shown in Fig. 1 corroborate this conclusion (17).

Fig.1  Detection of prothrombin binding to aIIbb3 by flow cytometry. Washed human platelets were incubated with human prothrombin 100 mg/ml with mAb or intact 7E3 (10 mg/ml) present. Bound prothrombin was then detected with a mAb to prothrombin ollowed by a FITC-labeled goat anti-mouse IgG

 In this analysis, binding of prothrombin to platelets could be detected with a mAb to prothrombin. This interaction was suppressed by abciximab, implicating aIIbb3 in the binding. Notably, binding of prothrombin to aIIbb3 on platelets did not require activation of the cells. Platelets in a resting state, defined by negligible fibrinogen binding, bound similar amounts of prothrombin as platelets expressing activated aIIbb3 induced by ADP, epinephrine or PMA, which exhibited extensive fibrinogen binding. Consequently, fibrinogen inhibited prothrombin binding to stimulated platelets but not to non-stimulated platelets. Thrombin did induce a substantial increase in prothrombin binding compared to resting or ADP stimulated platelets, but this interaction was largely aIIbb3 independent, presumably reflecting exposure of prothrombin binding anionic phospholipids on the surface of the thrombin-stimulated cells. Consistent with these data, we demonstrated by flow cytometry that prothrombin bound to platelets, which were annexin V negative; i.e., had little phosphatidylserine on their surface.

Binding in the absence of function is of little consequence. Binding of prothrombin to aIIbb3 had a clear biological consequence, the acceler-ation of thrombin formation. The assay used to demonstrate this point is shown in Fig. 2.

Fig.2  Effect of aIIbb3 on prothrombin activation. The upper portion of the figure illustrates the chromagenic assay using S2244 as the thrombin substrate employed to measure prothrombin activation by Factor Xa in the presence of immobilized aIIbb3. The lower portion shows the effect of abciximab on thrombin generation in the assay

 Prothrombin was added to aIIbb3 immobilized onto the wells of microtiter plates. After an initial incubation in the presence or absence of abciximab to block or allow its binding to the receptor, an activator of prothrombin, either Factor Xa alone or in combination with Factor Va, was added. The results shown in Fig. 2 indicate that abciximab inhibited the rate of prothrombin activation. No effect was seen with irrelevant mAbs, with abciximab present but without receptor, or with receptor present but without activator. The extent of inhibition induced by abciximab was in the 25-30% range. This level of inhibition is significant in view of: 1) the data showing that mAbs and RGD peptides inhibit thrombin generation in the presence of platelets by a similar or even greater extent (20); and 2) thrombasthenic platelets are ~30% less effective in supporting prothrombin activation than normal platelets (20, 21). In the assay used in Fig. 2, only 1-2% of the total added prothrombin is actually bound to the immobilized aIIbb3. Therefore, the kinetic advantage for the activation of receptor-bound prothrombin is substantial. More recently, we have corroborated this conclusion using an active site labeling approach and with intact platelets (22, 23). Blockade of aIIbb3 produced a marked delay in prothrombin activation by Factor Xa in the presence of resting and activated platelets.

These data support the model shown in Fig. 3.

Fig.3  Model for prothrombin binding and activation on aIIbb3 on platelets. The details of the model are discussed in the text

As noted above, the capacity of prothrombin to interact with aIIbb3 in resting platelets is unique among the plasma ligands that interact with the receptor via a RGD recognition specificity. Consequently, platelets should circulate in blood with prothrombin bound to aIIbb3. The extent of occupancy should be determined by the affinity of aIIbb3 for prothrombin, and/or the existence of undefined competitors that also interact with aIIbb3 or with prothrombin. The prothrombin bound to aIIbb3 is poised for activation should Factor Xa be generated in the vicinity of the platelet surface. Since thrombin does not compete with prothrombin binding to aIIbb3, cleavage of the sessile bond in prothrombin by Factor Xa should release thrombin from the receptor. Therefore, one might speculate that the overall increase in thrombin generation of aIIbb3-bound prothrombin reflects the balance between cleavage of the sessile bond in prothrombin, release of thrombin from the receptor, and the rate at which substrates can access to the active site of the newly formed thrombin. Certainly, the thrombin formed locally would be in a favorable position to activate platelets via PAR-1 or could interact with GPV (24) or GPIb-IX (25, 26). Thrombin stimulation results in the activation of aIIbb3, and changes the relationship between aIIbb3, prothrombin and other ligands of the receptor. With activated aIIbb3, interaction with other ligands, such as fibrinogen or vWF, would be favored over ist recognition of prothrombin. Under these circumstances, little additional prothrombin would bind to the receptor, and bound prothrombin would be displaced. However, these circumstances should favor formation and assembly of the prothrombinase complex on the platelet surface, providing a new site for prothrombin activation. Prothrombin activation within the prothrombinase complex is markedly more efficient than on aIIbb3 (27), allowing for a burst of new thrombin generation and future amplification of the entire coagulation process. However, the initial spark of thrombin generation could very well be provided by activation of prothrombin bound to aIIbb3.

Does the interaction of prothrombin with aIIbb3 bear on the efficacy of aIIbb3 antagonists? Perhaps, markers of blood coagulation are reduced by aIIbb3 antagonists (14, 28, 29). The effects of two aIIbb3 antago-nists on prothrombin activation as measured with the chromagenic substrate assay described above (Fig. 2) are shown in Fig. 4. This assay was performed in platelet-rich plasma. With one of the drugs, abciximab, thrombin generation was substantially inhibited; whereas, with the second agent, an oral aIIbb3 antagonist, thrombin generation was increased relative to the control lacking an antagonist. Both drugs were used at concentrations that inhibited platelet aggregation by greater than 80%. These data are presented as the amount of thrombin formed at a single time point, 4 min but are reflective of the measurements made at several other time points during the course of thrombin gen-eration. Such differences also were observed with the platelets from several different but not all donors. While these observations were made with two particular aIIbb3 antagonists, other drugs of this class displayed effects ranging from an inhibitory activity similar to abciximab to a stimulatory activity similar to that induced by the drug shown in Fig. 4.

Fig.4  Effect of aIIbb3 antagonists on prothrombin activation. Recombinant tissue factor was added to platelet-rich plasma, and thrombin generation was measured after 4 min by a modification of the assay shown in Fig. 2. The two antagonists were used at concentrations which inhibited platelet aggregation induced by 10 mM ADP by greater than 80%

Thus, the effects of aIIbb3 inhibitors on prothrombin activation can be variable, and a class specific effect cannot be assumed or assigned.

The Ligand Binding Specificity of aIIbb3

The efforts to identify the recognition specificity of aIIbb3 led to the identification of two reactive sequences (30-32). The peptides thatcircumscribe these sequences are commonly referred to as the RGD and the g-chain peptides. RGD sequences reside at two sites in the Aa chain of fibrinogen; are present in multiple adhesive proteins that interaction with aIIbb3 ; and mediate the interaction of multiple ligands, including fibrinogen and vWF, with several other integrins. The minimal active sequence for g-chain peptide recognition is KQAGDV (30, 33). Mutation of this, but not the RGD sequences in fibrinogen, abolish recognition by aIIbb3 (34, 35). Thus, aIIbb3 is an integrin with a RGD recognition specificity, i.e., RGD peptides inhibit ligand binding to the receptor, but it is the g-chain peptide sequence that is necessary for a productive interaction. (The g-chain [or RGD sequences] may not be the sole sites for interaction of fibrin[ogen] with aIIbb3; evidence for additional interactive sites has emerged recently [36].) Both RGD and g-chain peptides interact directly with aIIbb3 (33, 37) and inhibit fibrinogen binding and platelet aggregation (reviewed in [38]), but the relationship between the binding sites for these peptide ligands is unresolved. Do the RGD and g-chain peptides bind to the same site within the receptor, or do they bind to distinct sites? In the latter case, these sites must be allosterically linked to explain the cross-inhibition between the two peptides for binding to aIIbb3 (37). The relatively low affinity of the g-chain and RGD peptides has been a major impediment in resolving this longstanding uncertainty.

To address this issue, we have synthesized two cyclic peptides as high affinity mimetics of the RGD and g-chain peptides. These two peptides are cyclo(S,S)KYGCHarGDWPC (cHarGD) and cyclo(S,S) KYGCRGDWPC (cRGD) (39). They differ in structure only with respect to an arginine vs a homonarginine, a difference of a single methylene group. cHarGD is a mimetic of the g-chain sequence within fibrinogen based upon its poor reactivity, as well as that of fibrinogen, with integrin avb3 in the presence of Ca2+ and the similar reactivity of these ligands with aIIbb3 regardless of the divalent ion conditions (40). It is also a competitive inhibitor of fibrinogen binding to aIIbb3. cRGD is a cyclic version of linear RGD peptides. Both cyclic peptides are potent inhibitors of fibrinogen binding to aIIbb3; having IC50 values in the nM to pM range depending on the divalent ion conditions (39).

Each cyclic peptide was labeled with a single fluorochrome on its NH2-terminus, cHarGD with rhodamine and cHarGD with fluorescein, and the derivatization did not affect their interaction with aIIbb3. Three independent approaches were used to distinguish between the oneshared site versus the two allosterically linked site models for the cHarGD and cRGD peptides (39). The first approach involved crossinhibition studies. The capacity of the nonlabeled peptides to inhibit the binding of each of the labeled peptides to purified aIIbb3 in detergent solution was assessed. Flu-cRGD binding was inhibited well by cRGD but cHarGD was a poor inhibitor. In contrast, Rho-cHarGD binding was inhibited well by HarGD but not by cRGD. This relative selectivity extended to the linear peptides: the g-chain peptide inhibited cHarGD better than the cRGD binding; and the linear RGD peptide was more effective than the g-chain peptide in inhibiting cRGD binding.

The second approach assessed the simultaneous binding of the two cyclic peptides to the same receptor using fluorescence resonance energy transfer (FRET). When fluorescein is excited at 494 nm, it emits light at 520 nm; rhodamine can accept light at this wavelength to emit at a maximum of 572 nm. Consequently, if the Flu-cRGD is excited at 494 nm and is bound sufficiently close to the Rho-cHarGD, an emission peak at 572 nm will be detected. This is the result that was obtained (Fig. 5) and suggests that cHarGD and cRGD bind in sufficiently close proximity within aIIbb3 for FRET to occur.

Fig.5  Fluorescence resonance energy transfer between cHarGD and cRGD bound to aIIbb3. Flu-cRGD (——); rho-cHarGD (....) or both peptides (———) were bound to aIIbb3 in detergent solution. Free peptides were removed from the receptor by rapid gel filtration, and the sample was excited at 494 nm and the emission spectrum was recorded. The details of the analyses and the data are adapted from ref (39)

To justify this interpretation, several control experiments were performed. The stoichiometry of the binding of each of the two cyclic peptides to aIIbb3 was determined to approach a value of unity when the FRET measurement was deter- mined, indicating that crosstalk between fluorophores involved major binding sites within the receptor. When the two cyclic peptides were bound and then dissociated from the receptor with EDTA but remained at the same concentration within solution, no FRET was observed. When the experiment was performed at aIIbb3 concentrations over a ten-fold range, the efficiency of FRET remained similar, indicating that the cyclic peptides were bound to the same receptor in contrast to communicating between two receptors. From the measured efficiency of FRET, the distance between the binding sites of the two cyclic peptides within aIIbb3 can be approximated. The spacing between the two fluorophores was estimated to be 6.1 nm, a distance at which the two peptides could not occupy the same space.

The third approach involved expression of a segment of aIIbb3 that bound one cyclic peptide but not the other. The piece of the receptor that displayed this differential binding was b₃95-373. This fragment of the b3 subunit was previously shown to contain the RGD crosslinking site (41); harbors the epitopes for several blocking mAbs (42, 43); exhibits the capacity to bind fibrinogen (44); and is predicted to contain a MIDAS cation binding motif, which may reside in an I-like domain fold (45, 46). This piece of b3 was expressed as a thioredoxin fusion protein in E. coli. After purification and a refolding, b₃95-373 was immobilized. As shown in Fig. 6, the immobilized fragment was capable of binding cRGD, but not cHarGD.

Fig.6  Binding of cHarGD and cRGD to b₃95-373. The recombinant fragment, expressed in E. coli, purified and renatured, was immobilized onto the wells of microtiter plates. The binding of Flu-cRGD and Rho-cHarGD was detected by the fluorescence of the peptides. The divalent cation condition is either 0.1 mM Ca2+ (–) or 0.1 mM Mn2+ (— — —). The details of the analyses and the data are adapted from ref. (39)

Thus, this fragment expressed a binding site for one of the cyclic peptides but not the other. These data should not be over-interpreted to indicate that this segment of b3 lacks a binding site for cHarGD; a cHarGD site may be present in this segment but the fragment may not be properly folded to express this function. Nevertheless, the segment is appropriately folded to express a cRGD binding site that does not binding cHarGD with high affinity.

Recently, we have begun to characterize the interaction of the two cyclic peptides with aIIbb3 in more detail by assessing the binding of the radiolabeled cyclic peptides to immobilized receptor. Binding isotherms were constructed with ¹²⁵I-cHarGD, and the Scatchard plots of the data were interpreted in either a 1 or 2 site binding model. The Kd values derived from the best fit of the data are summarized in Table 1.

Table1  Estimated Kd values for cHarGD binding to purified aIIbb3

The binding data for cHarGD peptide in the presence of Ca2+ suggested the presence of two affinity classes of binding sites with Kd values differing by about 10-fold, 3.9 and 35.5 nM. The fit of the data to a two-site model was significantly (p = 0.03) better than to a one-site model. In contrast, in the presence of Mn2+, only one class of binding sites could be discerned. This site had an estimated Kd of 4.5 nM, similar to the Kd of the high affinity site for the ligand in the presence of Ca2+. Thus, cHarGD may bind to two sets of sites within aIIbb3, and divalent cations may determine the competence of these two sites in the receptor. These data are consistent with the report that certain disintegrins can bind to more than one site in aIIbb3 (47). The data also are compatible with the model proposed by Hu et al. (48) in which the b3 integrins express two classes of Ca2+ binding sites, which exert functionally distinct effects on ligand binding.

Taken together, these data emphasize the complexity of ligand bind-ing to aIIbb3. The two peptide ligands bind to different but allosterically interactive sites in the receptor, and there may be more than one binding site for each peptide depending on the divalent ion conditions. The relationship between divalent cation and ligand binding sites is intimate and possibly overlapping. This close interrelationship has been the conclusion of many studies and has been re-enforced by the recent crystallographic structure of a ligand bound to an integrin I domain (49). The complexity of aIIbb3: peptide ligand interactions is likely to extend to the interactions of antagonists and macromolecular ligands with the receptor. Multiple contacts are likely to mediate binding of macromolecular ligands, and the antagonists may be more effective in perturbing specific sets of contacts more than others. Accordingly, certain aIIbb3 antagonists may be very effective in blocking platelet aggregation but may be less effective in inhibiting prothrombin binding and activation (or some other function) of the receptor. Thus, it may be premature to condemn the entire class of aIIbb3 antagonists. Rather, the members may show distinct differences in their safety and efficacy profiles. In conclusion, the interactions of macromolecular ligands, peptide ligand, and the divalent cations with aIIbb3 must be explored in greater depth to develop a further understanding of the functions of the receptor and to optimize its antagonism.