Red Cell ICAM-4 Is a Novel Ligand for Platelet-activated αIIbβ3 Integrin*

ICAM-4 (LW blood group glycoprotein) is an erythroid-specific membrane component that belongs to the family of intercellular adhesion molecules and interacts in vitrowith different members of the integrin family, suggesting a potential role in adhesion or cell interaction events, including hemostasis and thrombosis. To evaluate the capacity of ICAM-4 to interact with platelets, we have immobilized red blood cells (RBCs), platelets, and ICAM-Fc fusion proteins to a plastic surface and analyzed their interaction in cell adhesion assays with RBCs and platelets from normal individuals and patients, as well as with cell transfectants expressing the αIIbβ3 integrin. The platelet fibrinogen receptor αIIbβ3 (platelet GPIIb-IIIa) in a high affinity state following GRGDSP peptide activation was identified for the first time as the receptor for RBC ICAM-4. The specificity of the interaction was demonstrated by showing that: (i) activated platelets adhered less efficiently to immobilized ICAM-4-negative than to ICAM-4-positive RBCs, (ii) monoclonal antibodies specific for the β3-chain alone and for a complex-specific epitope of the αIIbβ3integrin, and specific for ICAM-4 to a lesser extent, inhibited platelet adhesion, whereas monoclonal antibodies to GPIb, CD36, and CD47 did not, (iii) activated platelets from two unrelated type-I glanzmann's thrombasthenia patients did not bind to coated ICAM-4. Further support to RBC-platelet interaction was provided by showing that dithiothreitol-activated αIIbβ3-Chinese hamster ovary transfectants strongly adhere to coated ICAM-4-Fc protein but not to ICAM-1-Fc and was inhibitable by specific antibodies. Deletion of individual Ig domains of ICAM-4 and inhibition by synthetic peptides showed that the αIIbβ3 integrin binding site encompassed the first and second Ig domains and that the G65-V74 sequence of domain D1 might play a role in this interaction. Although normal RBCs are considered passively entrapped in fibrin polymers during thrombus, these studies identify ICAM-4 as the first RBC protein ligand of platelets that may have relevant physiological significance.

The main physiological function of red blood cells (RBCs), 1 which encapsulate hemoglobin, is to ensure the respiratory gases transport throughout the human body. However, the recent demonstration that mature RBCs express a growing number of adhesion molecules, many of which exhibit blood group specificities (1)(2)(3), reinforces the necessity to revisit the functional interaction of RBCs with leukocytes, platelets, and vascular endothelium under normal and pathological conditions.
It is interesting that many RBC adhesion molecules contain protein domains characteristic of the immunoglobulin superfamily, suggesting some recognition function. These molecules might participate in the normal RBC physiology by playing a role during erythropoiesis (differentiation, maturation, enucleation, release), self-recognition mechanisms, red cell turnover, and cell aging through cellular interactions with counter receptors present on macrophages from bone marrow or reticuloendothelial system in spleen and liver (1, 4 -9). Along this process, some adhesion molecules are rapidly down-regulated and others are expressed at different stages and remain on RBCs (Refs. 10 and 11, and references therein). Finally, mature RBCs still express adhesion molecules which are usually associated with leukocytes (CD44, CD47, CD58) and others that have potential adhesion properties such as LW/ICAM-4 (CD242), Lu (CD239), Ok a (CD147), CD99/Xg, JMH (CD108), and DO (1)(2)(3). Nevertheless, normal RBCs do not adhere to circulating cells and vessel walls under normal circumstances, suggesting that the RBC adhesion molecules are inaccessible to their ligands. In contrast, the conversion of non-adherent RBCs to adherent state arises in several diseases. In such circumstances, adhesion molecules might be involved in the pathophysiology of malaria (12,13), sickle cell disease (14 -17), and diabetes (18,19), mainly through an abnormal adhesion to the vascular endothelium (1,20). Additionally, both phosphatidylserine exposure at the RBC surface and adhesion molecules on these cells might also play a role in hemostasis and thrombosis, for instance through interaction with cells expressing integrins, like activated leukocytes, monocytes, platelets, and endothelial cells (21,22). Interestingly also, RBCs have the necessary signal transduction pathways to mediate these functions (23).
Among RBC adhesion molecules, ICAM-4 (LW blood group glycoprotein, CD242) emerges from the others by its structural similarities to the ICAM family and its interaction characterized in vitro with different members of the ␤ integrin subfamilies (␣ L␤2 (LFA-1), ␣ M␤2 (Mac-1) (24 -26), ␣ 4␤1 (VLA-4), ␣ V integrins (␣ V␤1 and ␣ V␤5 ); Ref. 27). These two families of proteins are well known to play crucial role in cell-cell interactions and to be involved in a large range of biological functions (28 -31). For instance, ICAM-4/integrin interaction might play a role during erythroid maturation in bone-marrow or in the red cell turnover by spleen macrophages that express the ␣ d ␤ 2 integrin (25,27,32). Additionally, ICAM-4 as well as the Lu blood group protein might be involved in adhesion of sickle RBCs to TNF-␣-activated endothelial cells (HUVEC) (7) and to laminin (33,34), respectively. It is suspected that abnormal adhesion of sickle RBCs to endothelial cells and extracellular matrix proteins might be responsible for the painful crisis of the disease that result from vaso-occlusive episodes (35).
The purpose of this report was to examine the potential role of ICAM-4 in RBC-platelet interaction and to demonstrate that this protein interacts in vitro with the high affinity state of activated platelet ␣ IIb ␤ 3 integrin.
Platelet Adhesion Assays to Immobilized RBCs-RBCs were immobilized on microtiter plates through binding to coated anti-glycophorin A. Briefly, mAb E4 at 20 g/ml (50 l/well) in 25 mM Tris, pH 8, 150 mM NaCl, was adsorbed overnight at 4°C on flat-bottom 96-well microtiter plates (Nunc A/S, Roskilde, Denmark). After two washes of wells with the same buffer, RBCs (2.0 ϫ 10 6 /well in a final volume of 300 l) resuspended in modified Tyrode's buffer, pH 7.4 with or without cations (2 mM MgCl 2 and 2 mM CaCl 2 ) were added. After 1 h of incubation at 22°C, fixed GRGDSP-activated or unactivated platelets (5.0 ϫ 10 6 /well in a final volume of 100 l) in modified Tyrode's buffer, pH 7.4 with or without divalent cations, respectively, were added to RBC-coated wells. After 90 min at 22°C, non-adherent cells were removed by filling the wells with binding buffer, and the microplates were put to float upside down in a PBS solution. Cells that adhered to the plastic wells were recovered by vigorous shaking in 400 l of PBS and were counted by flow cytometric analysis using a FACSCalibur. Platelets and RBCs were distinguished by forward scatters and platelet staining with the fluorescein isothiocyanate (FITC)-anti-human CD61 mAb (clone VI-PL2, BD Biosciences).
RBCs Adhesion to Adherent Platelets-Following isolation, unactivated platelets (1 ϫ 10 7 /well in a final volume of 100 l) resuspended in RPMI 1640, 10 mM Hepes containing PGE 1 and apyrase were added to wells to adhere overnight at 37°C. After washing, adherent platelets were stimulated with thrombin (0.5 unit in 100 l/well) diluted in Hanks' Balanced Salts (HBSS) containing 2 mM CaCl 2 for 20 min at room temperature. After another washing, RBCs (3.3 ϫ 10 6 /well in a final volume of 300 l) resuspended in HBSS with 2 mM CaCl 2 , 1 mM MgCl 2 , were added to each well. After 90 min at 22°C, non-adherent RBCs were removed by filling the wells with binding buffer, and the microplates were put to float upside down in a PBS solution. Then RBCs numeration was done using a Nikon Eclipse TE300 microscope (Nikon, Paris, France) (ϫ10 objective) coupled to a Biocom informatic system of images integration (Biocom, les Ulis, France). For blocking experiments, RBCs and adherent platelets stimulated by thrombin were pretreated with specific mAbs (2.5 g/well) and ICAM-Fc protein (2.5 g/well), respectively, for 30 min at 22°C. ␣ IIb ␤ 3 -CHO Transfectants and DTT Activation-The Chinese hamster ovary cell line (CHO) was grown in Iscove's modified Dulbecco medium with Glutamax-1 (Invitrogen) supplemented with amphotericin-B-penicillin-streptamycin and 10% fetal calf serum. CD41 (␣ IIbchain) and CD61 (␤ 3 -chain) cDNAs subcloned into pcDNA3.1 vector (Invitrogen), kindly provided by Dr. P. J. Newman (Blood Center of Southeastern Wisconsin, Milwaukee, WI), were cotransfected into CHO cells using the lipofectin reagent according to the manufacturer's instructions (Invitrogen). Stable transformants resistant to G418 (0.6 mg/ml of geneticin) were selected for CD41 and CD61 expression by immuno-magnetic separation using mAb AP-2 and magnetic beads coated with anti-mouse IgG (Dynabeads-M-450, DYNAL, Oslo, Norway). CD41 and CD61 expression of stable clones was analyzed and quantified by flow cytometric analysis with Qifikit calibration beads, used according to the manufacturer's instructions (Dako, Denmark). One clone with the strongest expression of ␣ IIb ␤ 3 integrin was selected. For adhesion assays, ␣ IIb ␤ 3 -CHO transfectant and wild-type (parental) CHO cells were treated with or without 10 mM DTT in RPMI 1640, 10 mM Hepes, at 22°C for 20 min to activate the ␣ IIb ␤ 3 complex receptor (39).
Cell Adhesion Assays to Immobilized Proteins-Purified ICAM-Fc proteins diluted in 25 mM Tris, pH 8.0, 150 mM NaCl, 2 mM MgCl 2 , and 2 mM CaCl 2 , were absorbed to flat-bottom 96-well microtiter plates overnight at 4°C, at 2.5-20 g/ml (50 l/well in triplicate). The wells were then blocked for 2 h at 22°C with 1% nonfat milk in the same buffer. For adhesion assays, either fixed GRGDSP-activated or unactivated platelets (5 ϫ 10 6 /well in a final volume of 100 l) in modified Tyrode's buffer, pH 7.4, with or without divalent cations, respectively, wild-type CHO cells, DTT-activated or unactivated ␣ IIb ␤ 3 -CHO transfectants (1 ϫ 10 5 /well in a final volume of 100 l) resuspended in RPMI 1640, 10 mM Hepes containing 2 mM MgCl 2 and 2 mM CaCl 2 , were added to the coated wells and incubated for 90 min at 22°C. Non-adherent cells were removed by washings before microscopic observation and CHO cell numeration was done as indicated above. Platelets were counted by flow cytometric analysis as above. For blocking experiments, the cells were pretreated with specific peptides and their corresponding random counterpart (125 M final concentration) or with different mAbs (5 g for 5 ϫ 10 6 platelets or 1 ϫ 10 5 CHO cells/100 l) for 30 min at 22°C prior addition to protein-coated wells.

RBCs
Interact with Activated Platelets-To analyze molecular events occurring during RBC-platelet interaction, in vitro cell adhesion assays were developed using RBCs from donors of common and rare phenotypes immobilized to plastic surface via anti-GPA binding and platelets from normal healthy donors, pretreated or not with the synthetic GRGDSP peptide in the presence of inhibitors of platelet activation, thus resulting in specific ␣ IIb ␤ 3 integrin activation and the acquisition of high affinity Fg-binding state without addition of a cellular agonist (37). Accordingly, in addition to bind Fg, GRGDSP-treated platelets reacted strongly with the mAb PAC-1, which binds to the activated ␣ IIb ␤ 3 complex (40), but no reactivity with the mAb AK-4 (41), which binds to P-selectin normally contained in intracellular ␣-granules (not shown). As shown in Fig. 1, GRGDSP-activated platelets adhered more efficiently than unactivated platelets to immobilized ICAM-4-positive RBCs from control donors. The 100% relative binding was equivalent to 220 Ϯ 100 GRGDSP-activated platelets adhered to 1.0 ϫ 10 3 immobilized RBCs. When unactivated platelets were used as control, a 69% reduced adhesion was noted that corresponded to a mean background of 31 Ϯ 12%. As preliminary assays showed that similar results were obtained with fresh and unfrozen RBCs (not shown), the following studies were performed with unthawed RBCs since rare RBC variants lacking different membrane proteins were available from our frozen collection.
Activated platelets bind to coated RBCs lacking the blood group proteins Lu (CD239, laminin receptor of 78 -85 kDa), JMH (CD108, 80 kDa), and DO (ADP-ribosyltransferase 4 of 47-67 kDa) but expressing normal levels of ICAM-4, as efficiently as would normal ICAM-4-positive RBCs. Interestingly, when ICAM-4 negative (LW null ) RBCs lacking of the ICAM-4/LW glycoprotein (42 kDa) from three unrelated donors were coated to plastic wells, a 40% decrease binding of GRGDSPactivated platelets was observed after deduction of the mean background corresponding to the unactivated platelet adhesion to all types of RBCs (p Ͻ 0.001 versus unactivated platelets and p Ͻ 0.05 versus controls).
To confirm that ICAM-4 plays a role in RBC-platelet interactions, RBC adhesion on adherent platelets stimulated by thrombin, a more physiologically relevant platelet activator than the RGDS peptide, was also analyzed although in this assay platelets are more activated with ␣-granule release than GRGDSP-activated platelets (see above). Although ICAM-4positive RBCs did not bind to unstimulated adherent platelets in the presence of PGE1 and apyrase (not shown), they bind strongly to thrombin-stimulated platelets (Fig. 2). This binding was efficiently decrease to 50 Ϯ 9% and 11 Ϯ 1% by mAb BS56 and soluble ICAM-4-Fc protein, respectively, whereas the mAb AICD58 reacting with the erythroid membrane CD58 protein and the soluble ICAM-2-Fc protein had only a minor inhibitory effect (88 Ϯ 4 and 85 Ϯ 7%, respectively). Similarly, mAbs anti-RhD (LOR-15C9), anti-Fy 6 (BAM9917) and anti-MER2 (1D12 or 2F7) directed against various RBC surface membrane proteins did not exhibit any effect (not shown). Unfortunately, the nonspecific adherence of frozen RBCs in this assay made impossible the comparative analysis between the ICAM-4-positive and -negative RBCs. Altogether, these data suggests that ICAM-4 might take a significant part (about 50%) in the adhesion of RBCs to activated platelets. As the GRGDSP peptide is a trigger of a high affinity state of ␣ IIb ␤ 3 integrin, which mediates Fg binding and platelet aggregation (37), our data suggested that ICAM-4 might interact with ␣ IIb ␤ 3 integrin but also with other adhesive molecules.
RBC-Platelet Interaction Is Mediated via ICAM-4 -To obtain further evidence that ICAM-4 might interact with a high affinity state of ␣ IIb ␤ 3 integrin, type-I glanzmann's thrombastenia platelets from two unrelated patients who both exhibit a 6-bp deletion in exon 7 of the ␤ 3 gene (42), were used for cell adhesion assays to coated ICAM-4-Fc protein. Fig. 3A shows that unactivated platelets from normal control donors did not bind to immobilized ICAM-4-Fc, as expected from above data, whereas the same platelets activated by the GRGDSP peptide bound readily to coated ICAM-4-Fc, but not to immobilized ICAM-1. The 100% relative binding of GRGDSPactivated platelets to ICAM-4-Fc was equivalent to 12.5 Ϯ 3.0% of the total added platelets. Conversely, platelets from the thrombasthenic patients type 1 with a severe defect of ␣ IIb ␤ 3 integrin surface expression, either unactivated (not shown) or GRGDSP-activated, failed to bind to coated ICAM-4-Fc (Fig. 3A).
In order to determine the specificity of these interactions, the effect of different mAbs and synthetic peptides on the platelet adhesion to immobilized ICAM-4-Fc protein was investigated (Fig. 3B). Adhesion of activated platelets from normal control donors was efficiently blocked (approximately, 70 and 60%, respectively) by P2 and AP2 mAbs specific for the ␣ IIb -chain in the presence of the ␤ 3 -chain and the complex-specific epitope of the ␣ IIb ␤ 3 integrin, respectively. SZ21 and SZ22 mAbs that recognize the ␤ 3 -and ␣ IIb -chains alone, respectively, and the BS56 mAb specific for ICAM-4, partially but significantly inhibited the interaction between ICAM-4 and activated platelets, whereas the SZ2 mAb directed against the GPIb platelet glycoprotein and the control mouse IgG had no significant effect (Fig. 3B). In addition, mAbs FA6 and 3E12 directed against CD36 and CD47, respectively, did not inhibit the platelet-ICAM-4 interaction. Blocking experiments by synthetic peptides revealed that the RGD peptide that binds to ␣ IIb ␤ 3 integrin and inhibits Fg binding, strongly reduced by 75% the adhesion of activated platelets to ICAM-4, whereas the RGE peptide had no effect.
RBC-Platelet Interaction Is Mediated via ICAM-4/␣ IIb ␤ 3 Integrin-To provide further evidence that ICAM-4 may interact with the ␣ IIb ␤ 3 integrin, stable CHO transfectants expressing recombinant human ␣ IIb ␤ 3 were generated and used in cell adhesion assays (Fig. 4). Several ␣ IIb ␤ 3 -CHO transfectants were obtained, and one clone expressing a high level of ␣ IIb ␤ 3 integrin (␣ IIb , 18,600 molecules/cell and ␤ 3 , 67,000 molecules/ cell, as estimated by flow cytometric analysis with specific mAbs) was chosen for further studies. The ␣ IIb ␤ 3 integrin of these cells was activated by DTT treatment and the adhesion of DTT-activated and unactivated ␣ IIb ␤ 3 -CHO transfectants to immobilized ICAM-4-Fc was examined (Fig. 4). In a preliminary experiment we found that these cells also reacted with PCA-1 mAb that recognized the activated ␣ IIb ␤ 3 integrin complex (not shown). DTT-activated ␣ IIb ␤ 3 -CHO transfectants dosedependently bind to coated ICAM-4-Fc protein, whereas untreated ␣ IIb ␤ 3 -CHO transfectants as well as parental CHO cells, either or not treated with DTT, did not bind at all. About 32% of the total added DTT-activated ␣ IIb ␤ 3 -CHO transfectants adhered to coated ICAM-4-Fc, but there was no binding to immobilized ICAM-1-Fc protein used as control (Fig. 4B). Identical results were obtained when the ␣ IIb ␤ 3 -CHO transfectants were activated by the GRGDSP-peptide instead of DTT (not shown). The binding of DTT-activated ␣ IIb ␤ 3 -CHO transfectants to immobilized ICAM-4-Fc could be blocked by ϳ50% by mAbs specific for ICAM-4 (BS56) or for the complex-specific epitope of the ␣ IIb ␤ 3 integrin (AP2), but not with mAbs to the ␤ 3 -chain (SZ21) and ␣ IIb -chain (SZ22 and P2) of the ␣ IIb ␤ 3 integrin, as shown on Fig. 4B. The absence of inhibition noted with the mAb P2, which efficiently blocked activated platelet adhesion (see Fig. 3) might result from conformational changes and/or glycosylation variations of ␣ IIb ␤ 3 integrin in platelets and the CHO transfectants independently of the mode of integrin activation. As expected, control mouse IgG had no effect.
Putative Domains on ICAM-4 That Interact with the ␣ IIb ␤ 3 Integrin-As an attempt to localize the ␣ IIb ␤ 3 integrin binding site on ICAM-4, domain deletion mutants lacking either extracellular Ig-like domain D1 or domain D2 were produced and used in cell adhesion assays to chimeric Fc proteins. Fig. 5A showed that the binding of DTT-activated ␣ IIb ␤ 3 -CHO transfectants via the ␣ IIb ␤ 3 integrin required the presence of both domains D1 and D2, since a 50% decrease binding was observed in the absence of either domain D1 or D2. Similar effects with less amplitude were observed using GRGDSP-activated platelets (see Fig. 5B).
Further blocking experiments with synthetic peptides were performed. Adhesion of activated-platelet was efficiently inhibited to 14 and 58% by the FBI, Fg ␥-chain residues 400 -411) peptide and the ICAM-4 peptide Gly-Val (residues 65-74), respectively, two peptides exhibiting a QXXDV motif involved in the fibrinogen/␣ IIb ␤ 3 integrin interaction (Fig. 5B). When DTT-activated ␣ IIb ␤ 3 -CHO transfectants were used, a 78% decrease in binding was observed in the presence of the ICAM-4derived peptide Gly-Val whereas the peptide FBI failed to inhibit (Fig. 5A). The lack of inhibition by the peptide FBI, might result from some changes of ␣ IIb ␤ 3 integrin when expressed in CHO transfectants, as suspected for the reactivity of mAb P2 (see above). Neither the random peptides of FBI and Gly-65-Val-74 nor a control ICAM-4 peptide (Cys-Arg, residues 180 -192) had any inhibitory effect. Altogether, these results demonstrate that the ␣ IIb ␤ 3 binding site on ICAM-4 encompassed domains D1 and D2 and that it seems to reside at the tip of the E strand of domain D1, which is in contact with the loop CЈ-E of domain D2. DISCUSSION In this report in vitro cell adhesion assays have been developed to evaluate the capacity of red cell ICAM-4 to interact with platelets and to identify the molecular basis of the interaction. The ␣ IIb ␤ 3 integrin (platelet fibrinogen receptor GPIIb-IIIa) was identified as the receptor for RBC ICAM-4. However, we found that the ␣ IIb ␤ 3 integrin had to be in its high affinity state to bind ICAM-4, as the interaction occurred only after synthetic GRGDSP peptide activation, but not with untreated resting platelets. This was based on the following evidence: (i) activated platelets adhered less efficiently to immobilized ICAM-4-negative (LW null ) than to ICAM-4-positive RBCs, (ii) monoclonal antibodies specific for a complex-specific epitope of the ␣ IIb ␤ 3 integrin or to the ␤ 3 -chain alone and specific for ICAM-4 to a lesser extent, inhibited platelet adhesion, (iii) activated platelets from two unrelated type-I glanzmann's thrombasthenia patients that are deficient for the ␣ IIb ␤ 3 integrin (and vitronectin receptor ␣v␤3) did not bind to coated ICAM-4-Fc protein, and (iv) DTT-activated ␣ IIb ␤ 3 -CHO transfectants strongly adhere to coated ICAM-4-Fc protein but not to coated ICAM-1-Fc, and this was inhibitable by specific antibodies. It should be mentioned that ␣ IIb ␤ 3 integrin activation occurred in the absence of any signaling or secretion (37,39) and that antibodies specific for GPIb, the von Willebrand receptor of platelets, for CD36 (platelet GPIV) and CD47 (IAP, integrinassociated protein), two multifunctional membrane proteins acting as thrombospondin receptors (43,44), did not inhibit the platelet adhesion to immobilized ICAM-4-Fc protein. As thrombasthenic platelets that expressed normal levels of other platelet receptors like the ␣2␤1 integrin (collagen receptor), and the fibronectin and laminin receptors (GPIa-IIa and GPIcЈ-IIa, respectively), did not adhere to ICAM-4-Fc protein, it is assumed that these proteins do not play a significant role in RBCplatelet interaction under the experimental conditions used.
Further analysis with platelets, ␣ IIb ␤ 3 -CHO transfectants and ICAM-4 Fc mutant proteins have shown that the two Ig-like domains of ICAM-4 are required for ␣ IIb ␤ 3 integrin interaction, since domain deletion mutants lacking either the first (D1) or second (D2) Ig-domain exhibited significant reduced binding (see Fig. 5, A and B). A similar effect has been observed when ICAM-4 mutant proteins interact with the leukocyte ␣ M ␤ 2 (Mac-1) integrin, whereas interaction with the ␣ L ␤ 2 (LFA-1) integrin requires predominantly the first Ig domain D1 (26). The binding of adhesive proteins to ␣ IIb ␤ 3 integrin is predominantly mediated by the RGD peptide motif present on the respective adhesive ligands (45), but this peptide, which is absent from ICAM-4, blocks the ICAM-4/␣ IIb ␤ 3 interaction. The platelet ␣ IIb ␤ 3 integrin also binds to the carboxyl-terminal end of the Fg ␥-chain via a dodecapeptide sequence (peptide FBI, residues 400 -411) containing the motif QAGDV (46). Interestingly, ICAM-4 contains a similar motif at position 70 -74 (QLLDV) located in the first ICAM-4 Ig-like domain (26) and the ICAM-4 peptide Gly-Val (residues 65-74), including this motif, is a potent inhibitor of the ICAM-4/␣ IIb ␤ 3 interaction. Moreover, the peptide FBI also inhibited platelet binding to ICAM-4-Fc. These findings suggest that the platelet ␣ IIb ␤ 3 integrin might interact with the Gly-65-Val-74 sequence of ICAM-4 that includes a QXXDV motif known to be involved in the fibrinogen/␣ IIb ␤ 3 integrin interaction. The Gly-65-Val-74 sequence motif which forms the tip of the E strand of domain D1 and is in contact with the loop CЈ-E of domain D2 (26), most probably constitutes a major part of the ␣ IIb ␤ 3 integrin binding site on ICAM-4. Therefore, binding inhibition by RGD and FBI peptides, which bind to the ␤ 3 chain (GPIIIa) and ␣ IIb chain (GPIIb), respectively (47), suggest that ICAM-4 binds to the same or overlapping site(s) on the ␣ IIb ␤ 3 complex. It should be noticed that the G70R substitution responsible for the blood group LW a 3 LW b polymorphism (48), which corresponds to the first position of the QXXDV motif, had no effect on RBC platelet adhesion reported here, 2 suggesting that this polymorphism is neutral with regard to ICAM-4/␣ IIb ␤ 3 integrin interaction.
Our studies therefore indicate that adhesion of normal RBCs to activated platelets occur through a specific ligand/receptor interaction. Whether or not signaling events across the platelet and/or RBC membranes are triggered by the interaction of the ␣ IIb ␤ 3 integrin receptor with its RBC ICAM-4 ligand is currently unknown.
As LW null RBCs still adhere to activated platelets and RBCs adhere to adherent platelets stimulated by thrombin, which have release their ␣-granule contents, it is assumed that other factors critical for interaction may exist. ␣ IIb ␤ 3 integrin activation with either RGD-containing peptide or DTT induces Fg binding (37,39), suggesting that indirect RBC-platelet contacts via this adhesive macromolecule might occur. However, although ICAM-1 binds to Fg (49), interaction of the structurally related ICAM-4 protein with Fg has not yet been documented. If such interaction exists, Fg could form RBC-platelet and RBC-endothelium cross-bridges via RBC ICAM-4 and ␣ IIb ␤ 3 integrin on activated-platelet and ICAM-1 and/or ␣ v ␤ 3 integrin present on the stimulated vascular endothelium (50,51). Still other adhesion pathways mediating cross-bridges of RBCs with platelets and/or endothelial cells with a variety of adhesive proteins might also be operating, but this needs further investigation. Additionally, erythroid receptors for adhesive molecules, like the Lutheran (CD239, laminin receptor) (33,34) and sulfated glycolipids (receptors of laminin, TSP, and vWF) (52,53) are present on the RBCs and might also take part in the RBC-platelet and endothelial cell interactions. Moreover, direct RBC-endothelial cell interaction might also occur as ICAM-4 has been reported to bind to ␣ 4 ␤ 1 (VLA-4), and to ␣ v ␤ 1 and ␣ v ␤ 5 integrins present on hematopoietic cells and might also account for the binding of sickle RBCs to vascular endothelium (27).
All these findings indicate that ICAM-4 is an unusual adhesion molecule that has a broad ligand binding specificity, including at least some ␤ 1 , ␤ 2 , ␤ 3 (this report), and ␤ 5 integrins, but the binding affinity for these ligands may vary widely. Another example of receptor with a promiscuous specificity is the DARC protein (Duffy Antigen receptor for Chemokines), which binds to CC and CXC families of chemokines (54). Therefore, it is anticipated that ICAM-4 may have a potential role in a number of physiological processes, including hemostasis and thrombosis (21,22).
Although further investigation of RBC interaction with blood cells and vascular cells under various flow conditions should delineate more precisely the physiological relevance of these interactions, RBC interaction with activated platelets, is supported by several observations: (i) an active role in hemostasis and thrombosis (21,22) as interaction between metabolically active RBCs and platelets is known to enhance platelet reactivity (21), including the enhancement of ␣ IIb ␤ 3 integrin activation and P-selectin expression (55), (ii) the presence of RBCs as well as of leukocytes in a developing thrombus (56 -58) in which platelets are activated and may interact with RBCs, (iii) the presence of platelet-erythrocyte aggregates in patients with sickle cell anemia (59,60) and end-stage renal disease (61). However, the molecular target(s) responsible for RBCs-platelet interaction have not been characterized. Our results provide the first direct characterization of a molecular interaction between normal RBCs and platelets and together with the findings discussed above they strongly suggest that RBCs may play an active role in hemostasis and thrombosis.
After this article was submitted, adhesion of normal RBCs to fMLP (formyl-Met-Leu-Phe peptide)-activated neutrophils and collagen-activated platelets, as well as to fibrin, was shown under low shear rate conditions (62). Interestingly, the data suggested that adhesion of RBCs to neutrophils might be mediated through Mac-1 (CD11b/CD18) and ICAM-4, supporting recent findings indicating that ␤ 2 integrins and ICAM-4 interact with each other (26). Additionally, RBC-platelet interaction was strongly reduced by soluble fibrinogen and EDTA and was partially inhibited by antibodies to CD36 and GPIb, but no inhibition was noted with a single antibody against the ␣ IIb chain (CD41) of the ␣ IIb ␤ 3 integrin (62). Although no effect of monoclonal antibodies to GPIb and CD36 was found in static conditions of assay, our results indicate that the interaction of thrombin-activated platelets with intact RBCs (Fig. 2) and of GRGDSP-activated platelets with immobilized ICAM-4 ( Fig.  3B) could be inhibited by soluble ICAM-4 (by 89%) and to at least 50% by the monoclonal antibody P2 recognizing the ␣ IIbchain in the presence of the ␤ 3 -chain, or by the monoclonal antibody AP-2 specific for a complex-specific epitope of the ␣ IIb ␤ 3 integrin, respectively. Monoclonal antibodies SZ22 and SZ21 to the ␣ IIb -chain and ␤ 3 -chain alone, respectively, were weak inhibitors in the latter condition (Fig. 3B). Consistent with the above results, our studies have shown that ICAM-4 interaction with ␤ integrins is calcium-dependent (26). Obviously, distinct experimental conditions (platelet activation, flow conditions) and monoclonal antibodies used may explain the reported differences.
In conclusion, although passive entrapment of RBCs during coagulation or thrombosis is commonly accepted, these data provide independent evidence indicating that a physiological interaction between RBCs and activated platelets (and neutrophils) mediated by specific receptor/ligand interactions can occur in a variety of biological process, notably during normal hemostatic conditions (clot formation), pathological occlusion conditions (deep vein thrombosis, sickle cell disease) and possibly inflammation, particularly under low blood flow conditions, close to static, which may facilitate RBC adhesion events. Although ICAM-4 may play a significant role, clearly other receptor/ligand interactions are likely to occur which deserves further analysis.