Identification of the Binding Site for Fibrinogen Recognition Peptide γ383–395 within the αMI-Domain of Integrin αMβ2 *

The leukocyte integrin αMβ2 (Mac-1, CD11b/CD18) is a cell surface adhesion receptor for fibrinogen. The interaction between fibrinogen and αMβ2 mediates a range of adhesive reactions during the immune-inflammatory response. The sequence γ383TMKIIPFNRLTIG395, P2-C, within the γ-module of the D-domain of fibrinogen, is a recognition site for αMβ2 and αXβ2. We have now identified the complementary sequences within the αMI-domain of the receptor responsible for recognition of P2-C. The strategy to localize the binding site for P2-C was based on distinct P2-C binding properties of the three structurally similar I-domains of αMβ2, αXβ2, and αLβ2,i.e. the αMI- and αXI-domains bind P2-C, and the αLI-domain did not bind this ligand. The Lys245-Arg261 sequence, which forms a loop βD-α5 and an adjacent helix α5 in the three-dimensional structure of the αMI-domain, was identified as the binding site for P2-C. This conclusion is supported by the following data: 1) mutant cell lines in which the αMI-domain segments 245KFG and Glu253-Arg261 were switched to the homologous αLI-domain segments failed to support adhesion to P2-C; 2) synthetic peptides duplicating the Lys245-Tyr252 and Glu253-Arg261 sequences directly bound the D fragment and P2-C derivative, γ384–402, and this interaction was blocked efficiently by the P2-C peptide; 3) mutation of three amino acid residues within the Lys245-Arg261 segment, Phe246, Asp254, and Pro257, resulted in the loss of the binding function of the recombinant αMI-domains; and 4) grafting the αM(Lys245-Arg261) segment into the αLI-domain converted it to a P2-C-binding protein. These results demonstrate that the αM(Lys245-Arg261) segment, a site of the major sequence and structure difference among αMI-, αXI-, and αLI-domains, is responsible for recognition of a small segment of fibrinogen, γThr383-Gly395, by serving as ligand binding site.

Integrin ␣ M ␤ 2 participates in the attachment of leukocytes to the endothelial lining of blood vessels and the subsequent transmigration of adherent cells during immune-inflammatory responses (1)(2)(3). The engagement of fibrinogen (Fg) 1 by ␣ M ␤ 2 on the surface of leukocytes and by intercellular adhesion molecule-1 (ICAM-1) on the endothelium may play a role in mediating the adhesion of leukocytes to the vessel wall (4,5) and in facilitating their subsequent extravasation across the endothelial monolayer (6). In addition, the binding of deposited fibrinogen or fibrin to ␣ M ␤ 2 may mediate leukocyte adhesion at extravascular sites of inflammation (7)(8)(9).
In previous studies, Altieri et al. (10,11) demonstrated that a peptide (designated P1), corresponding to residues 190 -202 of the ␥-chain of the D-domain of Fg, was recognized by ␣ M ␤ 2 . However, when residues key to the recognition of P1 by ␣ M ␤ 2bearing cells were mutated in the ␥-module, ␥148 -411, this recombinant fragment was as active as its wild-type counterpart in supporting ␣ M ␤ 2 -mediated adhesion (12). This observation led to the search for additional ␣ M ␤ 2 recognition sites within the ␥-chain, and ultimately the P2 peptide, corresponding to ␥377-395, was identified (12). Indeed, in comparative analyses, P2 was 10 -15-fold more potent than P1 in inhibiting adhesion of the ␣ M ␤ 2 -expressing cells to the D fragment of Fg. Further analyses of the adhesion-promoting activity of overlapping peptides showed that its COOH-terminal part, ␥ 383 TMKI-IPFNRLTIG 395 , designated P2-C, was the primary site of its biological activity (12). Recently, a second leukocyte integrin, ␣ X ␤ 2 , which is highly homologous to ␣ M ␤ 2 , was demonstrated to bind to the ␥-module and P2-C peptide (13), and soluble P2-C peptide efficiently blocked the ␣ X ␤ 2 -mediated adhesion.
Within the heterodimeric ␣ M ␤ 2 receptor, the I-domain, a region of ϳ200 amino acid residues, inserted in the ␣ M subunit, contributes broadly to the recognition of ligands by ␣ M ␤ 2 (14) and specifically to the binding of Fg to this integrin (14,15). In addition to Fg, this region also was implicated in the binding of ICAM-1 (15), iC3b (16), and neutrophil inhibitory factor, NIF (17,18). We have shown previously that P2 interacts with the recombinant ␣ M I-domain and that NIF partially blocked this interaction (12). Previous studies suggested that overlapping but not identical sites are involved in the recognition of iC3b, NIF, and Fg (19). However, although the binding sites for iC3b and NIF in the ␣ M I-domain were mapped extensively (20 -23), the recognition site for Fg has not been studied. Recently, sequences key to the binding of NIF and iC3b to the ␣ M Idomain were mapped using a homolog scanning mutagenesis strategy (22,24). This approach is based upon the structural similarity of the I-domains of ␣ M and ␣ L and the differences in their ligand recognition; i.e. the crystal structures of the I-domains of ␣ M and ␣ L are very similar (25)(26)(27)(28), but only the I-domain of ␣ M binds NIF with high affinity (17,18). Fg, together with NIF and iC3b, does not bind to ␣ L ␤ 2 , suggesting that differences in the structure of the ␣ M I-and ␣ L I-domains may be responsible for their distinction in ligand binding specificity. In this study, we have sought to localize the binding site for the P2-C sequence of Fg within the ␣ M I-domain. The strategy developed was based on the differences in the binding of P2-C to the ␣ M I-, ␣ X I-, and ␣ L I-domains and involved several independent approaches, including screening of mutant cells, synthetic peptides, site-directed mutagenesis, and the gain-infunction analyses. The binding site for P2-C was localized within the segment ␣ M (Lys 245 -Arg 261 ), a site of the major structure divergence between ␣ M I-, ␣ X I-, and ␣ L I-domains. The grafting of this segment into the ␣ L I-domain converted it to the P2-C-binding protein. Thus, a small amino acid sequence, P2-C, with a defined structure within a crystallized domain of fibrinogen (29) is shown to interact with a small segment that also has a defined structure within the ␣ M I-domain.

EXPERIMENTAL PROCEDURES
Proteins, Peptides, and Monoclonal Antibodies-Human kidney 293 cells expressing wild-type and the mutant forms of the ␣ M ␤ 2 receptor were described and characterized in detail previously (19,22). These cells were grown as adherent monolayers in Dulbecco's modified Eagle's medium/F-12 medium (BioWhittaker, Walkersville, MD), supplemented with 10% fetal bovine serum, 25 mM HEPES, and antibiotics. Human Fg was purified from fresh human blood by differential ethanol precipitation (30) or obtained from Enzyme Research Laboratories (South Bend, IN). The D 100 (M r 100,000) fragment was prepared by digestion of human Fg with plasmin (Enzyme Research Laboratories, South Bend, IN) and purified as described (31). The D 98 fragment (M r 98,000) was produced by digestion of the D 100 with plasmin. This fragment lacks 5-15 amino acid residues from the COOH terminus of the ␥-chain and will be described elsewhere. 2 D 98 was biotinylated with EZ-link Sulfo-NHS-LC-Biotin (Pierce) according to the manufacturer's instructions. P1 (␥190 -202), P2 (␥377-395), P2-C (␥383-395), H19 (␥340 -357), H2O (␥350 -374), L10 (␥402-411), and H12 (␥400 -411) peptides were synthesized and purified as described (12). In addition, an analog of P2-C, P2-Ce, a peptide with the extended COOH-terminal end, ␥384 -402, was also synthesized to test for direct binding to ␣ M Idomain peptides. The following peptides duplicating selected sequences within the ␣ M I-domain were synthesized: 147  Expression of Recombinant ␣ M I-, ␣ X I-, and ␣ L I-Domains and Sitedirected Mutagenesis-The I-domains were expressed as fusion proteins with glutathione S-transferase (GST) and purified from soluble fractions of Escherichia coli lysates by affinity chromatography. The coding regions for the ␣ M I-domain (residues Asp 132 -Ala 318 ) and ␣ L Idomain (Gly 153 -Ile 333 ) were amplified by polymerase chain reactions using as template plasmids pCIS2M-␣ M (19) and pCIS2M-␣ L , which contain the cDNA fragments coding for the full-length of ␣ M and ␣ L , respectively. The primers used for the ␣ M I-domain were 5Ј-TGTCCTG-GATCCGATAGTGACATTGCCTTCTTGA (forward) and 5Ј-TGAGTAC-CCGCGGCCGCCGCAAAGATCTTCTCCC (reverse). The primers used for the ␣ L I-domain were 5Ј-CAGGAAGGATCCAAGGGCAACGTAGAC-CTGGTATT (forward) and 5Ј-GTCCTGTTTGCGGCCGCCCTCAATGA-CATAGA (reverse). The underlined nucleotides are BamHI and NotI recognition sequences that were introduced in the primers. The fragments were digested with BamHI and NotI and cloned in the expression vector pGEX-4T-1 (Amersham Pharmacia Biotech). The accuracy of the DNA sequence was verified by sequencing. The plasmid was transformed in E. coli strain BL-21(DE3)pLysS competent cells, and expression was induced by adding 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 3-5 h at 37°C.
To express the ␣ X I-domain, the following primers were used to amplify a cDNA fragment encoding the ␣ X I-domain (residues Glu 148 -Ala 335 ) from a randomly primed cDNA library of U937 monocytoid cell line: 5Ј-AGGCTACCGGGATCCAGACAGGAGTGCCCAAGA (forward) and 5Ј-CATCTCCCAATTTGGCGGCCGCACTGCTTGTGGTC (reverse). The product was digested with BamHI and NotI and cloned into pGEX-4T-1. The plasmid was transformed in E. coli BL-21(DE3)pLysS cells, and the correctness of the ␣ X I-domain insertion was confirmed by sequencing. The ␣ X I-domain was expressed and purified from the cell lysates as a fusion protein with GST under the conditions used for the ␣ M I-and ␣ L I-domains.
Site-directed mutagenesis of the ␣ M I-domain was performed by using QuickChange TM mutagenesis kit (Stratagene, San Diego). The pGEX-4T-1 construct containing DNA encoding the ␣ M I-domain was modified by site-directed mutagenesis using two mutagenic primers containing the desired mutation. The mutations introduced in the ␣ M I-domain and the primers used are listed in Table I. The oligonucleotide primers, each complementary to opposite strands of the vector, were extended during temperature cycling by using PfuTurbo TM DNA polymerase. Following temperature cycling, the product was treated with DpnI endonuclease to digest the parental DNA template. The nicked vector DNA incorporating the desired mutations was then transformed into the Epicurian Coli ® XL1-Blue supercompetent cells, and cDNA from individual bacterial clones was analyzed by sequencing. The E. coli BL-21(DE3)pLysS host cells were then transformed with the mutant plasmids, and the mutant ␣ M I-domains were prepared as described above for the recombinant wild-type ␣ M I-domain. The immunoreactivity of wild-type and mutant recombinant I-domains was analyzed by enzyme-linked immunosorbent assay with mAbs 44a and 904 using our standard protocol (35).
Generation of the ␣ L (␣ M (Lys 245 -Arg 261 ))I-Domain Chimera-The segment corresponding to the sequence Ala 267 -Asp 278 within the ␣ L I was exchanged to the homologous segment of the ␣ M I-domain Lys 245 -Arg 261 . The segment switch was created by oligonucleotide-directed mutagenesis using polymerase chain reaction. The construction of the chimera was based on the observation that oligonucleotide sequence corresponding to ␣ M (Lys 245 -Arg 261 ) contains a unique restriction site for Eco81I (SauI), whereas the second site for this enzyme is present between 4760 and 4761 of the pGEX-4T-1 sequence. To switch the Lys 245 -Arg 261 from ␣ M to ␣ L , the mutagenic primers were designed to contain the Eco81I site within the ␣ M (Lys 245 -Arg 261 ) segment and additional unchanged ␣ L bases: 5Ј-CATCCCTGAGGCAGACAGAATCATCCGCTACATCATCG (forward), 5Ј-GCTGCCTC AGGGGATCACATCTTCATAACCCAAT-GGATCTCCAAACTTCTCCCCATCCGTGATGATGATAAG (reverse) (the ␣ M sequences are in bold, and the restriction site for Eco81I is underlined). The pGEX-4T-1 vector containing DNA encoding the ␣ L I- GTCATCCCTGAGGCAGccAGAGAAGGAGTCATTCG a The number indicates the position of amino acids in the full-length ␣ M subunit (32). b The lowercase letters indicate the mutagenic bases.
domain was modified by polymerase chain reaction using PfuTurbo DNA polymerase with the following cycling parameters: 95°C for 30 s, 55°C for 1 min, 68°C for 11.5 min. Following temperature cycling, the product was treated with DpnI to digest the parental DNA template. The linear product was purified and digested with Eco81I to produce the two cDNA fragments with cohesive ends. These fragments were ligated and transformed into Epicurian Coli XL1-Blue supercompetent cells. The accuracy of the DNA sequence and the correctness of the I-domain direction were verified by sequencing. The E. coli BL-21(DE3)pLysS cells were transformed with the mutated plasmid, and the chimeric molecule was prepared following the procedure described above for the wild-type and mutant I-domains. Flow Cytometry-FACS analyses were performed to assess the expression of ␣ M ␤ 2 on the surface of the cells transfected with wild-type and mutant forms of the receptor. The cells were harvested, and 10 6 cells were incubated with ␣ M -specific mAbs OKM1 or 44a at 15 g/200 l of cell suspension for 30 min at 4°C. The cells were then washed and incubated with fluorescein isothiocyanate-goat anti-mouse IgG (at a 1:1,000 dilution) for additional 30 min at 4°C. Finally, the cells were washed and analyzed in a FACS Star (Beckton Dickinson, Mountain View, CA). Populations of cells expressing a similar amount of the receptor were selected by FACS. A clonal population of each mutant was isolated by limiting dilution. After propagation in culture, the amount of the ␣ M ␤ 2 was again evaluated by FACS analysis with mAb OKM1.
Adhesion Assays-The wells of tissue culture plates (Costar, Cambridge, MA) were coated with 6.1, 12.5, 25, 50, and 100 g/ml P2-C peptide or 1, 5, and 25 g/ml of D 100 fragment for 3 h at 37°C. The amount of peptides immobilized onto the wells was measured by utilizing radiolabeled peptides (12). The coated wells were postcoated with 0.5% polyvinylpyrrolidone for 1 h at 22°C. The 293 cells expressing wild-type or mutated forms of ␣ M ␤ 2 were harvested with cell-dissociating buffer (Life Technologies, Inc.) for 1 min at 22°C and washed twice in HBSS/HEPES solution containing 10 mg/ml BSA, and resuspended at 5 ϫ 10 5 /ml in HBSS (without phenol red)/HEPES supplemented with 1 mM Ca 2ϩ and 1 mM Mg 2ϩ and 10 mg/ml BSA. 100-l aliquots of the cells were added to each well and incubated on the adhesive substrates for 25 min at 37°C in a 5% CO 2 humidified atmosphere. The nonadherent cells were removed by three washes with colorless HBSS. The adherent cells were frozen overnight at Ϫ20°C, then thawed and lysed by the addition of a buffer containing the CyQuant dye (Molecular Probes, Eugene, OR). The fluorescence was measured in a Cytofluorimeter (PerSeptive Biosystems, Framington, MA), and the number of adherent cells was determined from a reference standard curve prepared according to the manufacturer's protocol. Additionally, several experiments were performed using our established protocol with 51 Crlabeled cells (12), and the results were found to be identical with those obtained using the CyQuant reagent (see "Results").
Solid Phase Binding Assays-To test the interaction of the wild-type ␣ M , ␣ X , ␣ L , mutant ␣ M , and chimeric I-domains, 96-well plates (Immulon 4BX, Dynex Technologies Inc., Chantilly, VA) were coated with P2-C or P2-Ce at 50 g/ml overnight at 4°C and postcoated with 3% BSA or 0.5% polyvinyl alcohol for 2 h. The GST-I-domains in 20 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 0.05% Tween 20, and 5% glycerol were added to the wells and incubated for 3 h at 22°C. After washing, bound I-domains were detected with an anti-GST mAb (Upstate Biotechnology, Lake Placid, NY) at a 1:5,000 dilution. After washing, goat anti-mouse IgG conjugated to alkaline phosphatase was added for 1 h, and the binding of the I-domains was measured by reaction with p-nitrophenyl phosphate. As a control, the binding of GST to immobilized P2-C was typically ϳ5-10% that of wild-type ␣ M I-domain, and background binding to BSA or polyvinyl alcohol was subtracted.
To examine the interaction of the D fragment with the I-domain peptides, 96-well plates (Immulon 2 HB, Dynex Technologies Inc.) were coated with peptides Pro 147 -Arg 152 , Pro 201 -Lys 217 , Lys 245 -Tyr 252 , and Glu 253 -Arg 261 at 100 M (0.1 ml/well) overnight at 4°C and postcoated with 3% BSA for 1 h at 22°C. 10 g/ml biotinylated D 98 in 50 mM Tris-HCl buffer, pH 7.5, and 0.05% Tween 20 were added to the wells and incubated for 2.5 h at 37°C. In parallel, the same amount of the D 98 was mixed with different concentrations of P1, P2 peptides, or NIF and added to the wells with immobilized ␣ M I-domain peptides. After washing, streptavidin conjugated to alkaline phosphatase (Pierce) was added and incubated for 45 min at 37°C. D 98 binding was detected by reaction with p-nitrophenyl phosphate, measuring the absorbance at 405 nm.
To demonstrate the direct binding of the P2-C region (␥383-395) to the ␣ M I-domain peptides, experiments were performed as follows. Different concentrations of the P2-Ce peptide, ␥384 -402, in 50 mM Tris-HCl and 0.05% Tween 20 were added to wells with immobilized ␣ M Idomain peptides and incubated for 3 h at 37°C. After washing, the binding of the ␥384 -402 was detected with mAb 4-2. The mAb 4-2 recognizes an epitope within this peptide, although it reacts poorly with authentic P2-C. 3 The binding of this mAb to ␥384 -402 was measured by reaction with goat anti-mouse IgG, conjugated to alkaline-phosphatase (Pierce) and using p-nitrophenyl phosphate for detection.
Statistical Analyses-Values of adherent cells bound to substrates are given as means Ϯ S.E. and are based on two to six independent experiments with each ␣ M ␤ 2 cell line, performing triplicates at each experimental point.

RESULTS
Binding of Wild-type Recombinant ␣ M I-, ␣ X I-, and ␣ L I-Domains to the P2-C Peptide of Fg-In previous studies, we have demonstrated that the P2-C peptide binds directly to the recombinant ␣ M I-domain (12). ␣ X ␤ 2 also recognizes P2-C (13); but the role of the ␣ X I-domain in this interaction was not evaluated, and the recognition of P2-C by the ␣ L I-domain had not been tested. Therefore, the three I-domains were expressed as GST fusion proteins and tested for their ability to interact with the immobilized P2-C peptide. As shown in Fig. 1, the recombinant ␣ X I-domain exhibited a dose-dependent and saturable binding to the P2-C peptide similar to that of the recombinant ␣ M I-domain. In contrast, the recombinant ␣ L I-domain did not interact with P2-C even at the highest concentration of the ␣ L I-domain added (200 g/ml maximal testable concentration). The binding characteristics of the isolated I-domains parallel the binding properties of the corresponding intact receptors on cell surfaces, i.e. the ␣ M ␤ 2 -and ␣ X ␤ 2 -expressing cells adhere to Fg and P2-C, whereas the ␣ L ␤ 2 -bearing cells do not (see Figs. 2 and 3). Thus, these data confirm that binding of P2-C to two highly homologous integrins, ␣ M ␤ 2 and ␣ X ␤ 2 , is mediated by ␣ M I-and ␣ X I-domains, respectively, and further suggest that the lack of binding of P2-C to ␣ L ␤ 2 may be the result of sequence and/or structural differences.
Binding of P2-C to Mutant Cell Lines-As the first step to define the binding site for P2-C within the ␣ M I-domain, mutant cell lines, each expressing a mutant ␣ M ␤ 2 in which a short ␣ M I-domain sequence was replaced for the corresponding region of the ␣ L I-domain, were tested for their adhesion to im-3 V. P. Yakubenko and T. P. Ugarova, unpublished observation. containing 100 mM NaCl, 1 mM Mg 2ϩ , 1 mM Ca 2ϩ , 0.05% Tween 20, and 5% glycerol were added to the microtiter plates coated with 50 g/ml P2-C peptide and postcoated with 0.5% polyvinyl alcohol and incubated for 3 h at 22°C. After washing, anti-GST mAb (1:5,000) was added to the wells and incubated for an additional 1.5 h. The binding of the I-domains then was detected with a secondary goat anti-mouse IgG conjugated to alkaline phosphatase with subsequent development of the reaction with p-nitrophenyl phosphate. mobilized P2-C or D 100 fragment. These mutant cell lines have been used previously to examine the binding of NIF and iC3b to ␣ M ␤ 2 (22,24). For such experiments to be readily interpretable, cell lines expressing high and similar levels of wild-type and mutant receptors were selected by cell sorting using mAbs OKM1 and 44a followed by cloning by limiting dilution. The expression of the receptors as assessed by the mean fluorescence intensity in FACS analyses differed by less than 2-fold from the internal wild-type control.
The adhesion of each mutant was measured with increasing concentrations of immobilized P2-C and D 100 fragment to determine the maximal level of adhesion. In all cases where a cell line bearing a mutant receptor did adhere to either substrate, the adhesion was dependent upon the concentration of the immobilized ligand and reached a plateau. This maximal adhesion was compared with that of the cells expressing wild-type ␣ M ␤ 2 receptor and mock-transfected cells in the same experiment to allow normalization of results. The cell lines exhibiting adhesion similar to or greater than the wild-type ␣ M ␤ 2 cells were identified as "positive" mutants. The pattern of cell adhesion to P2-C obtained for wild-type ␣ M ␤ 2 and a representative mutant, ␣ M (K 231 NAF), is shown in Fig. 2. Adhesion of ␣ M (K 231 NAF) to P2-C was dose-dependent and saturable with ϳ70% of added cells adherent to P2-C at the plateau. Adhesion of wild-type and mutant cells to a control peptide, H19, was tested and was found to be less than 5% of added cells. Also, as the essential control, the ␣ L ␤ 2 -expressing cells adhered poorly to either adhesive substrates, P2-C and D 100 , (ϳ10 -15% of added cells) consistent with lack of interaction of ␣ L ␤ 2 with Fg. All negative mutants did not adhere to P2-C; adhesion of these cells was similar to that of mock-transfected cells.
The results of adhesion of 16 mutants to P2-C and the D 100 fragment are summarized in Fig. 3, A and B  ative mutants. Peptides corresponding to the wild-type ␣ M Idomain sequences were synthesized and tested for their ability to interact with the D fragment and P2-C. Because the sequences within two of the negative mutants, ␣ M (Pro 201 -Gly 207 ) and ␣ M (Arg 208 -Lys 217 ), were contiguous, one linear peptide spanning Pro 201 -Lys 217 was prepared. Also, the peptide, Lys 245 -Tyr 252 , containing the sequence of the negative mutant ␣ M (K 245 FG) was extended at its COOH terminus to include the D 248 PLGY 252 sequence. Thus, four peptides were synthesized: Pro 147 -Arg 152 , Pro 201 -Lys 217 , Lys 245 -Tyr 252 , and Glu 253 -Arg 261 .
In these experiments, we used a derivative of the D 100 fragment, D 98 , which has a higher affinity for ␣ M ␤ 2 . 2 D 98 contains the entire P2-C but differs from D 100 in the length of the constituent ␥-chain, which terminates at ␥397/405 in D 98 compared with the intact ␥411 in D 100 . These ␣ M I-domain peptides were immobilized onto microtiter plates, and the binding of biotinylated D 98 fragment to them was assessed. As shown in  (Fig. 4B; the effect of P1 on the binding to Glu 253 -Arg 261 is shown). Two control peptides duplicating fibrinogen sequences ␥340 -357 (H19) and ␥350 -374 (H2O) did not affect the interaction. P1 inhibited the interaction as efficiently as P2, consistent with ability of this peptide to compete with P2 for binding to ␣ M ␤ 2 (12). The interaction of the D 98 with immobilized ␣ M I-domain peptides was cation-independent; in fact, the higher level of binding was observed in the absence than in the presence of 1 mM Ca 2ϩ or Mg 2ϩ (Table II). Interestingly, NIF did not inhibit binding of the D 98 fragment to the immobilized Lys 245 -Tyr 252 and Glu 253 -Arg 261 peptides at concentrations as high as 100 g/ml (Table  II), although it efficiently inhibited adhesion of the ␣ M ␤ 2 -expressing cells to the D fragment and P2-C peptide at a concentration as low as 0.1 g/ml (not shown). These data suggest that NIF does not interact with the I-domain peptides, which are capable of binding P2-C and the D 98 fragment. Indeed, when direct binding of biotinylated NIF to the immobilized ␣ M I-domain peptides was tested, NIF did not bind to Lys 245 -Tyr 252 and Glu 253 -Arg 261 peptides (not shown).
In addition, we were able to detect direct binding of P2-C to the immobilized ␣ M I-domain peptides. Using a derivative peptide P2-Ce, ␥384 -402, which contains an epitope for the reporting mAb 4-2, we demonstrated that P2-Ce efficiently bound to immobilized Lys 245 -Tyr 252 and Glu 253 -Arg 261 in a dose-dependent and saturable manner (Fig. 5). The binding of P2-Ce to Pro 201 -Lys 217 was low, whereas Pro 147 -Arg 152 and control peptide Phe 223 -Asn 232 did not bind P2-Ce. Similar to the interaction of the whole D 98 fragment, the binding of P2-Ce to Lys 245 -Tyr 252 and Glu 253 -Arg 261 was cation-independent (Table II). To exclude further the possibility that the binding of P2-Ce to immobilized Lys 245 -Tyr 252 and Glu 253 -Arg 261 was nonspecific, we tested two control fibrinogen peptides, ␥402-411 (L10) and ␥400 -411 (H12). These peptides duplicate sequences that reside in close proximity to P2-Ce (␥384 -402) and interact with platelet integrin ␣ IIb ␤ 3 . These peptides contain an epitope at ␥405-411 recognized by the mAb 4A5 (34). L10 and H12 did not bind to the immobilized ␣ M I-domain peptides as judged by the lack of the mAb 4A5 immunoreactivity, providing further evidence for the specificity of the interaction between P2-Ce and two ␣ M I-domain peptides, Lys 245 -Tyr 252 and Glu 253 -Arg 261 . Thus, the D 98 and P2-C-derivative bound strongly to the same peptides from within the ␣ M I-domain, Lys 245 -Tyr 252 and Glu 253 -Arg 261 , and the interaction of both ligands with these peptides followed the same pattern. This is consistent with the binding of the D 98 fragment being mediated by P2-C. rental ␣ L I-domain was tested. Whereas the parental ␣ L I-domain does not bind to P2-C, the grafting of the ␣ M segment imparted the P2-C binding function to the chimeric I-domain. Although the affinity of the interaction could not be assessed accurately from the experimental format used, the concentration of the immobilized P2-C required to obtain 50% of the chimeric I-domain binding was similar to that for wild-type ␣ M I-domain, i.e. ϳ12 g/ml for the chimera compared with ϳ20 g/ml for the wild-type ␣ M I-domain. The interaction of the chimeric molecule with P2-C was blocked by soluble P2-C, thus confirming specificity (Fig. 6B).
Identification of Residues within the ␣ M (Lys 245 -Arg 261 ) Segment Critical for P2-C Binding-The residues within the ␣ M (Lys 245 -Arg 261 ) segment, responsible for the P2-C binding, were identified by site-directed mutagenesis. The selection of residues for mutational analyses was based on the following considerations: 1) the side chains of residues that participate in direct docking of P2-C should be exposed on the hydrated surface of the ␣ M I-domain; 2) given the similarities in the binding function between ␣ M and ␣ X I-domains, it is likely that the residues that interact with P2-C should be identical or conserved between two I-domains; 3) because the switch of the 248 DPLGY 252 segment did not affect adhesion of mutant cell line, this sequence was not included in mutational analyses. Nine residues, Lys 245 , Phe 246 , Gly 247 , Glu 253 , Asp 254 , Pro 257 , Glu 258 , Asp 260 , and Arg 261 are exposed on the surface of the ␣ M I-domain in the Lys 245 -Arg 261 stretch, and eight residues are identical in the ␣ M (Lys 245 -Arg 261 ) and ␣ X (Lys 262 -Ala 277 ) sequences (see Fig. 7A). Thus, only 5 of the 17 residues in Lys 245 -Arg 261 segment meet both criteria, being exposed and identical between ␣ M and ␣ X : Lys 245 , Gly 247 , Asp 254 , Pro 257 , and Asp 260 (Fig. 7A). To examine the contribution of these residues in the ligand binding function, single or multiple point mutations to alanine were introduced into the ␣ M I-domain (Fig. 7A), and the capability of mutant proteins to interact with the immobilized P2-C was tested (Fig. 7B). The binding of mutants containing TABLE II Effect of cations and NIF on the binding of D 98 fragment and the P2-C derivative, ␥384 -402, to immobilized Glu 253 -Arg 261 10 g/ml biotinylated D 98 fragment in 50 mM Tris-HCl buffer, pH 7.5, containing 1 mM Ca 2ϩ 1 mM Mg 2ϩ , 0.05% Tween 20 or in 50 mM Tris-HCl, pH 7.5, 0.05% Tween 20 without cations was added to the wells coated with 100 M Glu 253 -Arg 261 and incubated for 2.5 h at 37°C. After washing, bound D 98 was detected with streptavidin conjugated to alkaline phosphatase and nitrophenyl phosphate. The effect of 100 g/ml NIF on the binding of the D 98 fragment in the absence of cations was tested in parallel. The inhibitory effect of 10 g/ml each P2-C and P1 (both tested without cations) is shown for comparison. The binding of the ␥384 -402 was detected by using the mAb 4-2 as described in Fig. 5. The data shown are the mean values (ϮS.D.) of the absorbance at 405 nm of a representative experiment done in triplicate. , 100 mM NaCl, 1 mM Mg 2ϩ , 1 mM Ca 2ϩ , 0.05% Tween 20, and 5% glycerol were added to the wells coated with 50 g/ml P2-C and incubated for 3 h at 22°C. After washing, bound I-domains were detected by anti-GST mAb (1:5,000 dilution). After washing, goat anti-mouse IgG conjugated to alkaline phosphatase was added for 1 h, and the binding of the I-domains was measured by reaction with the p-nitrophenyl phosphate. Background binding to BSA or polyvinyl alcohol was subtracted. Panel B, inhibition of the chimeric I-domain binding by P2-C. 50 g/ml chimera was mixed with different concentrations of P2-C (q) or H19 (Ⅺ) and then added to the wells coated with P2-C. The binding of the I-domain was determined as in panel A. Results are presented as a percentage of maximal chimera binding in the absence of P2-C. single or double mutations of Asp 254 and Pro 257 (mutants 3, 4, 5, and 6) was significantly impaired compared with wild-type ␣ M I-domain. In contrast, I-domains with substitutions of Lys 245 , Gly 247 , and Asp 260 bound similarly to the wild-type ␣ M I-domain. Because the switch of 245 KFG impaired adhesion of the ␣ M ␤ 2 -expressing cells (see Fig. 3, A and B), Phe 246 might be critically involved in P2-C recognition. Accordingly, Phe 246 was mutated to Arg because previous data had indicated that substitution of Phe 246 to the positively charged residue abrogated binding of the ␣ M I-domain to iC3b, whereas mutation to Ala was without effect (23). As shown in Fig. 7B, the binding of the ␣ M I-domain with mutation F246R was reduced significantly. The reduced binding of the three mutants, F246R, D254A, and P257A, to P2-C was apparently not caused by perturbation in conformation based upon the following observations: 1) The reactivity of the mutant I-domains with anti-␣ M I-domain mAbs 44a and 904 was similar to that the wildtype ␣ M I-domain (not shown), consistent with what was reported previously for other mutations in this region (20). 2) Inspection of the three-dimensional structure of the ␣ M I-domain suggested that introduction of these substitutions would not produce conformational changes because side chains of the three mutated residues do not interact with neighboring residues and thus do not contribute into stabilization the overall structure; instead, alanine substitution of Pro 257 should improve the slightly imperfect ␣5 helix. 3) Typical proteins can tolerate major mutations within a single loop and remain properly folded (36). Thus, these data suggest that the three residues, Phe 246 , Asp 254 , and Pro 257 , are directly involved in the P2-C binding. DISCUSSION In this study, we have identified key elements of the binding site for a small amino acid sequence of Fg, ␥383-395 (P2-C), within the ␣ M I-domain of ␣ M ␤ 2 . The strategy to define the ligand binding site was based on the difference in the P2-C binding properties of the ␣ M I-, ␣ X I-, and ␣ L I-domains and entailed four complementary approaches. In the first approach, a series of homolog-scanning mutants, used previously to map the binding regions for NIF, iC3b, and Candida albicans (22,24,37), were screened for adhesion to P2-C and D 100 fragment of Fg. In these mutants, 16 segments from the ␣ M I-domain were replaced with the corresponding segments from the homologous ␣ L I-domain, which does not bind Fg. Because all of these swapped segments are located at the hydrated surface of the ␣ M I-domain, they should identify candidate sequences for interaction with the ligand. Five mutants lacked the ability to support adhesion to P2-C and D 100 : ␣ M (P 147 -R 152 ), ␣ M (P 201 -G 207 ), ␣ M (R 208 -K 217 ), ␣ M (K 245 FG), and ␣ M (E 253 -R 261 ). Therefore, these ␣ M I-domain segments may be critical for binding of these ligands. Alteration of three other regions, dE262G, Asp 273 -Lys 279 , and Phe 297 -Thr 307 , resulted in the partial loss of adhesive function. These segments may play an accessory role in ligand binding. Thus, the initial insight provided by these mutant receptors indicated that the P2-C binding interface within the ␣ M I-domain was composed of several nonlinear sequences. Based on the crystal structure of the ␣ M I-domain (25), the segments critical for recognition encompass a portion of helix 1, the loop between helix 3 and helix 4, the small 245 KFG segment in the loop between helix 5 and ␤-strand D, and the entire helix 5 (Fig. 8, left). These sequences form an almost continuous stretch on the upper face of the ␣ M I-domain (colored in different shades of green in Fig. 8, left).
The second approach entailed the use of synthetic peptides duplicating the sequences of the critical segments in the ␣ M I-domain. These analyses showed that two of four critical segments, ␣ M ( 245 KFG) and ␣ M (Glu 253 -Arg 261 ), may contain amino acid residues that participate directly in binding the P2-C sequence of Fg because the peptides that duplicated Lys 245 -Tyr 252 and Glu 253 -Arg 261 , bound P2-C. Although the peptide Pro 147 -Arg 152 did not interact with the Fg derivatives, and Pro 201 -Lys 217 interacted weakly, the negative results do not exclude a role for these peptides in binding function; the immobilized peptide may simply not assume the appropriate conformation for recognition by the ligand.
The Lys 245 -Tyr 252 and Glu 253 -Arg 261 sequences are contiguous, and the entire ␣ M (Lys 245 -Arg 261 ) sequence might serve as the primary binding site for P2-C. Of note, this segment is the most divergent between the ␣ M I-and ␣ L I-domains in terms of sequence homology and folding (Fig. 8, right). In fact, ␣ L lacks most of helix 5, which is formed by residues Tyr 252 -Arg 261 of ␣ M (27,28). In addition, the loop ␤D-␣5 is longer in ␣ M than in ␣ L and assumes a different conformation. Therefore, this difference between ␣ M and ␣ L could account for inability of ␣ L to bind Fg. To obtain direct evidence that the ␤D-␣5 loop-␣5 helix in  Fig. 6, this manipulation imparted P2-C binding capacity to the chimeric molecule, and the binding affinity of the chimeric receptor for P2-C was very similar to that of wild-type ␣ M I-domain. Thus, the role of the ␤D-␣5 loop-␣5 helix in P2-C binding, which initially was inferred from the loss-in-function experiments, was verified by the gain-in-function approach.
Finally, the implementation of the fourth method, site-directed mutagenesis, served the dual purpose. First, it provided the independent confirmation that the Lys 245 -Arg 261 segment is important for P2-C binding because mutations of the three residues resulted in the significant loss of P2-C binding. Second, it implicated Phe 246 , Asp 254 , and Pro 257 as contact residues. Taken together, the four approaches substantiated independently the role of the ␤D-␣5 loop-␣5 helix in the P2-C binding and provided evidence that three residues participate in ligand docking.
It is unclear why switches of the ␣ M (Pro 147 -Arg 152 ) and ␣ M (Pro 201 -Lys 217 ) resulted in the loss of function given that two other unsubstituted segments, ␣ M ( 245 KFG) and ␣ M (Glu 253 -Arg 261 ) could potentially support adhesion. The effect of the switches of ␣ M (Pro 147 -Arg 152 ) and ␣ M (Pro 201 -Lys 217 ) on P2-C recognition could conceivably arise from changes of conformation of the ligand binding region residing in the ␣ M (Lys 245 -Arg 261 ). In this regard, a subtle perturbation of the structure of ␣ M (P 201 -G 207 ) and ␣ M (P 147 -R 152 ) mutants was suggested previously by an altered reactivity with the conformation-dependent mAb 24 (22). In addition to the MIDAS motif (25), which is known to be required for the normal binding function (21,38), single point mutations in the regions outside MIDAS also may abrogate ligand binding by altering conformation (21,38,39). For example, alanine substitution of Asp 248 or Tyr 252 , the residues that are not exposed on the surface of the ␣ M I-domain, eliminated the binding of mutants to iC3b (21), suggesting that a structural alteration might have been involved in the loss-ofbinding function. Thus, even relatively small perturbations in the folding of the I-domain can lead to gross alterations in binding affinities.
Although ␣ M (Lys 245 -Arg 261 ) resides in close proximity to the cation binding MIDAS motif, none of the residues in this sequence is directly involved in coordination of the divalent cation (25). Our results indicate that the binding of Fg derivatives, the D fragment and P2-C, to immobilized peptides duplicating Lys 245 -Arg 261 region was cation-independent. This finding is consistent with previous data showing that EDTA or mutation of Asp 242 , a residue which coordinates to the bound metal, only partially impairs the binding of Fg to the ␣ M I-domain although it abolished the binding of other ligands, including NIF and iC3b to the recombinant fragment (25). Another report also suggests that ligands can bind to I-domains independent of cations. Peptides duplicating ␤D-␣5 loop in the ␣ M I-domain or immediately preceding it bound to iC3b in a cation-independent manner (16). Thus, at least in the case of P2-C, its binding to the ␣ M I-domain does not occur through direct interaction with the metal ion as was proposed (25). This conclusion is further supported by the fact that P2-C sequence in Fg does not contain a candidate acidic residue to provide a missing coordination to the metal. At the same time, P2-C does contain an arginine residue, Arg 391 , which could displace cation, a model suggested from the crystal structure of the ␣ 1 I-domain (40).
The sequence ␣ M (Lys 245 -Tyr 252 ), which overlaps with the identified Fg-binding region ␣ M (Lys 245 -Arg 261 ), was implicated previously in the binding of iC3b. Deletion of Phe 246 -Tyr 252 abolished rosetting of iC3b-coated erythrocytes with ␣ M ␤ 2 (21). In addition, mutation of Lys 245 to Ala (24) and Phe 246 to Lys (24) also significantly reduced iC3b binding. However, although the binding site for iC3b overlaps with the Fg binding site, the contribution of this region in recognition of two ligands appears to be distinct. For example, deletion of 248 DPLGY did not affect adhesion to P2-C or D 100 fragment in our experiments, but it reduced to some extent iC3b binding (19).
The overlapping nature of the NIF and Fg binding sites within the ␣ M I-domain was suggested previously based on the ability of NIF to inhibit interaction of the ␣ M ␤ 2 -bearing cells with Fg (18,19). The same segments which were identified as critical for Fg binding, ␣ M (Pro 147 -Arg 152 ), ␣ M (Pro 201 -Gly 207 ), ␣ M (Arg 208 -Lys 217 ), and ␣ M (Glu 253 -Arg 261 ), also have been shown to participate in NIF binding (22). The significant differences in the binding of ␣ M I-domain mutants to these two ligands were: 1) switch of 245 KFG, which completely abrogated adhesion to Fg peptides, was not critical for NIF binding; and 2) deletion of 248 DPLGY, which affected NIF binding, was not detrimental for adhesion to Fg derivatives. The binding site for NIF was verified previously by grafting the identified segments into the ␣ X I-domain because these swaps imparted NIF binding capacity to the chimeric receptor (22). However, because the identified segments were grafted simultaneously, it is un-