Delineation of the Key Amino Acids Involved in Neutrophil Inhibitory Factor Binding to the I-domain Supports a Mosaic Model for the Capacity of Integrin (cid:1) M (cid:2) 2 to Recognize Multiple Ligands*

To gain insight into the mechanism by which the (cid:1) M I- domain of integrin (cid:1) M (cid:2) 2 interacts with multiple and unrelated ligands, the identity of the neutrophil inhibitory factor (NIF) recognition site was sought. A systematic strategy in which individual amino acid residues within three previously implicated segments were changed to those in the (cid:1) L I-domain, which is structur - ally very similar but does not bind NIF, was imple-mented. The capacity of the resulting mutants, expressed as glutathione S -transferase fusion proteins, to recognize NIF was assessed. These analyses ultimately identified Asp 149 , Arg 151 , Gly 207 , Tyr 252 , and Glu 258 as critical for NIF binding. Cation binding, a function of the metal ion-dependent adhesion site (MIDAS) motif, was assessed by terbium luminescence to evaluate conformational perturbations induced by the mutations. All five mutants bound terbium with unaltered affinities. When the five residues were inserted into the (cid:1) L I-do- main, the chimera bound NIF with high affinity. An-other ligand of (cid:1) M (cid:2) 2 , C3bi, which is known to use the same segments of the (cid:1) M I-domain in engaging the recep- tor, failed to bind to the chimeric (cid:1) L tions and energy minimizations, referred to as the molecular simulation procedure. Docking was performed manually with a distance cut-off of 15 Å using MSI’s docking program. Final van der Waals and electro- static energies of interaction were obtained using a distance cut-off of

I-domains (also known as A-domains), are inserted domains of ϳ200 amino acids, which are found in nine integrin ␣-subunits as well as other proteins (e.g. von Willebrand factor) and have been implicated in mediating a variety of protein-protein interactions, including ligand binding to integrins. Integrin ␤ subunits, including ␤ 2 , may also contain an I-like domain (40). The ␣ M I-domain and ␣ L I-domain were the first of several I-domains to have their three-dimensional structures solved by x-ray crystallography (41,42). They are all similar, being composed of six or seven ␣-helices and five ␤-sheets connected by flexible loops. A cation-binding site is formed at the vertex of the ␤-sheets, and the bound cation is coordinated in a MIDAS motif (41). Point mutations at residues involved in coordination of cation within the ␣ M I-domain destroy C3bi binding to the ␣ M ␤ 2 receptor (34). Using blocking monoclonal antibodies, which map to the I-domain or recombinant ␣ M I-domains, the MIDAS face of the ␣ M I-domain has been implicated in the binding of ICAM-1 and fibrinogen (36,39,43) as well as C3bi (38,44) and NIF (45,46). Taken together, these data suggest that the ␣ M I-domain can fold and function as an independent structural unit, capable of interacting with many different proteins. Nevertheless, the ligand binding function of ␣ M ␤ 2 is not merely a property of its I-domain. The ␤ 2 subunit and the EF-hand-like cation binding domain adjacent to the I-domain in the ␣ subunit also may contribute to ligand recognition (47).
NIF was originally identified as an activity within the extracts of canine hookworms that inhibited many neutrophil functions, such as adhesion to endothelial cells and the respiratory burst (21). These effects arose from its specific binding to ␣ M ␤ 2 but not to other ␤ 2 integrins. Subsequently, it was shown that NIF completely blocked ␣ M ␤ 2 -mediated binding of C3bi (38) and ICAM-1 (22) and adhesion to protein-coated surfaces (46) and partially blocked fibrinogen binding (37) to ␣ M ␤ 2bearing cells. As an antagonist, NIF has been shown effective in attenuating the deleterious effects of excessive neutrophil activation, such as tissue damage and ischemia-reperfusion injury in an animal model (48,49), and, hence, it is a potential anti-inflammatory drug. Three specific segments, Pro 147 -Arg 152 , Pro 201 -Lys 217 , and Asp 248 -Arg 261 , on the face of the ␣ M I-domain containing the MIDAS were implicated in NIF binding by a homologous scanning mutagenesis approach (50). In this approach, small structural units (helices, ␤-sheets, and loops) in the ␣ M I-domain were replaced by the homologous sequences from the ␣ L I-domain, and the segments noted above were identified as being necessary for NIF binding. In the present report, we have sought to precisely define the amino acids within these segments that are required for high affinity recognition of NIF by ␣ M I-domain, thereby defining the molecular basis for binding of this model ligand to ␣ M ␤ 2 . Ultimately, five individual amino acid residues are identified within the ␣ M I-domain as being critical to NIF binding. The placement of these five residues and our knowledge of the contact sites for other residues in the ␣ M I-domain support a new molecular explanation for the capacity of this integrin to recognize multiple ligands.

EXPERIMENTAL PROCEDURES
Materials-NIF was kindly provided by Corvas International (San Diego, CA). The NIF used in this study was a recombinant protein expressed in Pichia pastoris and was purified to homogeneity as described Muchowski et al. (37). Monoclonal antibodies 44, IB4, and CBL 189 were obtained from Sigma, the American Tissue Culture Collection (Manassas, VA), and Chemicon (Temecula, CA), respectively.
Site-directed Mutagenesis, Expression, and Purification of I-domain Fusion Proteins-The cDNAs of ␣ M I-domain (675 nucleotides, Arg 115 -Ser 340 ) and ␣ L I-domain (630 nucleotides, Pro 120 -Ser 330 ) were cloned and inserted into the pGEX-5X-3 expression vector (37). All wild-type and mutant I-domains were expressed as GST fusion proteins. Mutations were created in these I-domains by oligonucleotide-directed mutagenesis using the QuikChange site-directed mutagenesis kit (Stratagene, San Diego, CA). The selective introduction of the desired mutations into the I-domains was confirmed by DNA sequence analyses. I-domain expression plasmids were used to transform competent Escherichia coli cells (BL21(DE3)pLysS (Stratagene). One liter of L broth (Bio 101, Carlsbad, CA), containing 100 g/ml ampicillin (Sigma), was inoculated with 1 ml of an overnight culture of transformed E. coli, and cells were grown at 37°C with vigorous shaking until an absorbance of 0.6 -1 OD at 600 nm was attained. Expression was induced by the addition of 1 mM isopropyl-b-D-thiogalactopyranoside (Amersham Biosciences). Four hours after induction, cells were harvested by centrifugation and resuspended in 20 ml of sonication buffer (PBS containing 100 M 4-(2-aminoethyl)benzenesulfonyl fluoride (Calbiochem) and 50 g/ml leupeptin (Calbiochem)). Cells were disrupted with four 30-s cycles at 70 watts in a Branson Ultrasonics power sonicator. Insoluble cellular membranes were removed by centrifugation at 25,000 ϫ g for 15 min at 4°C. The glutathione S-transferase (GST)-I-domain fusion proteins were present in the supernatant and were purified by adsorption onto glutathione-Sepharose 4B (Amersham Biosciences) and elution with elution buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM glutathione, 2.5 mM CaCl 2 ). Such preparations of the fusion proteins were ϳ90% pure as assessed by SDS-PAGE. Concentrations of purified proteins were determined using the BCA Protein Assay (Pierce) and by spectrophotometry at 280 nm.
Radiolabeling and NIF Binding Assays-NIF (100 g) was radiolabeled with 0.7-1.0 mCi of Na 125 I (Amersham Biosciences) using the IODO-BEADS iodination reagent from Pierce. The reaction mix was incubated for 15 min at 22°C. Radiolabeled protein was separated from free iodine by gel filtration through a PD10-DG column (Bio-Rad) using PBS containing 1% BSA (Calbiochem) as the elution buffer. The specific activity of radiolabeled protein was ϳ12 Ci/g.
To assess NIF binding to the recombinant I-domains, 125 I-NIF (from 66 pM to 330 nM) was mixed with 1 g of ␣ M I-domain (or ␣ L I-domain) fusion proteins bound to glutathione-Sepharose 4B in TBS (20 mM Tris-HCl, pH 7.6, 250 mM NaCl) containing 1 mM Ca 2ϩ in a final volume of 50 l. The mixture was incubated overnight with agitation at 4°C. The resin then was washed three times with TBS, and bound 125 I-NIF was measured with a ␥ counter. Based upon the previous evidence that NIF binds to a single high affinity binding site in the ␣ M I-domain, NIF titration data were fit to a single site binding model using the following equation: where B max represents the maximal binding and K d is its dissociation constant, using the SigmaPlot program (Jandel Co., San Rafael, CA). K d values in Tables  I-III are presented as means Ϯ S.D.
For selected experiments, the I-domain was cleaved from its GST fusion partner by incubation of the recombinant protein adsorbed onto glutathione-Sepharose 4B with factor Xa (10 g/ml factor Xa per 200 g/ml I-domain) overnight at room temperature with agitation. The cleaved ␣ M I-domain was recovered by washing the resin with 3 column volumes of TBS, and the eluate was concentrated by centrifugation through a Millipore membrane. For solid phase binding assays with the isolated ␣ M I-domain, 96-well microtiter plates (Costar, Cambridge, MA) were coated with the ␣ M I-domain at 50 g/ml overnight at 4°C and postcoated with 3% BSA. 125 I-NIF was added, and binding was measured as described above.
C3bi Binding to the I-domains-C3bi binding was performed as described by Bilsland et al. (51) with certain modifications. Sheep erythrocytes (Colorado Serum Co., Denver, CO) at 7 ϫ 10 8 were washed twice in HBSS, containing 5 mM HEPES and 1 mM Mg 2ϩ , and treated with anti-sheep erythrocyte IgM antibody M1/87 (Accurate Chemical and Scientific Co., Westbury, NY) and human C5-deficient serum (Sigma). The coated erythrocytes were surface-labeled with biotin using 1 mg of sulfosuccinimidyl 6-(biotinamido)hexanoate (Pierce) at 37°C for 20 min. The biotinylated cells were resuspended in 0.9 ml of HBSS with 5 mM HEPES, 1 mM Ca 2ϩ , and 1 mM Mg 2ϩ ; mixed with 100 l of C5-deficient serum; and incubated at 37°C for 60 min. After washing twice, the resulting erythrocytes were resuspended in 2 ml of the above solution, and the cells were disrupted in a sonicator (Branson Ultrasonics, VWR Scientific) with one 3-s cycle at 70 watts. The insoluble cellular material recovered by centrifugation at 25,000 ϫ g for 15 min at 4°C was used as a C3bi source. To test the interaction of the wild-type ␣ M , ␣ L , and chimeric I-domains with C3bi, 96-well plates (Immulon 4BX; Dynex Technologies Inc., Chantilly, VA) were coated with the I-domains at 50 g/ml for 1 h at 37°C and postcoated with 2% BSA for 1 h at 37°C. After washing three times, the C3bi in 20 mM Tris-HCl, pH 7.6, containing 100 mM NaCl and 2 mM CaCl 2 , was added to the wells and incubated for 1 h at 37°C. After washing three times, the bound C3bi was detected using avidin-alkaline phosphatase and p-nitrophenyl phosphate. As a control, the background reaction on BSA-coated wells was subtracted.
Luminescence Emission Spectroscopy-Tb 3ϩ luminescence experiments were performed as previously described (52). The excitation wavelength used was 285 nm, and Tb 3ϩ emission was measured at 545 nm. Emission spectra were corrected for the blank contribution, and the instrument response and normalized to an absorbance of 0.100 at 285 nm in a quartz cell of 0.5-or 1-cm path length. All measurements were made in an FP-777 spectrofluorometer (Jaco, Hauppauge, NY).
SDS-PAGE-Gel electrophoresis was performed according to the method of Laemmli (53) in 8% acrylamide gels. Proteins were detected in the gels by staining with 0.04% Coomassie G-250 (Bio-Rad).
Molecular Modeling-The model building of the ␣ M I-domain based upon its crystal structure (54), Protein Data Bank code 1JLM, was carried out using InsightII (MSI, San Diego, CA) using the program defaults, unless otherwise stated, on a Silicon Graphics Indigo work station. The charges on the amino acids of ␣ M I-domain were adjusted to pH 7.2. The I-domain was then soaked with five layers of water, and the minimized energy conformation was obtained by gradually annealing the protein water as previously described (55). The resulting minimized structure was then subjected to a series of molecular dynamic simula-tions and energy minimizations, referred to as the molecular simulation procedure. Docking was performed manually with a distance cut-off of 15 Å using MSI's docking program. Final van der Waals and electrostatic energies of interaction were obtained using a distance cut-off of 100 Å.

Purification and NIF-binding Function of Recombinant ␣ M Idomains-Previous studies have established that the ␣ M I-do-
main is sufficient to account for the high affinity and specific binding of NIF to integrin ␣ M ␤ 2 (21,22,37). Furthermore, based upon a homolog mutagenesis strategy, in which homologous small structural elements, ␣-helixes, ␤-sheets, and connecting loops of the ␣ M I-domain were swapped into the ␣ L Idomain, three segments, corresponding to connecting loops on the cation-binding MIDAS face of the I-domain were specifically implicated in NIF recognition by ␣ M ␤ 2 (50). These segments were Pro 147 -Arg 152 , Pro 201 -Lys 217 , or Asp 248 -Arg 261 (numbering based on the amino acid sequence of intact ␣ M ; Pro 201 -Lys 217 actually consisted of two contiguous segments, Pro 201 -Gly 207 and Arg 208 -Lys 217 , but are combined for simplicity). Based upon these results, we sought to localize the specific amino acid residues within these segments of the ␣ M I-domain that are required for the high affinity recognition of NIF, thereby providing a molecular basis for recognition of this by To begin this analysis, wild-type ␣ M I-domain (Arg 115 -Ser 340 ), wild-type ␣ L I-domain (Pro 120 -Ser 330 ), and the three homolog scanning mutants were expressed as GST fusion proteins in E. coli. The recombinant proteins were purified from the bacterial lysates on glutathione-Sepharose. The purified I-domains migrated as major bands of ϳ48 and ϳ52 kDa for ␣ L I-domain and ␣ M I-domain, respectively, when analyzed by SDS-PAGE (Fig. 1, lanes 1 and 2), consistent with their predicted molecular weights. The three homolog scanning mutants also migrated as single bands on the gels (Fig. 1, lanes 3-5).
A facile binding assay for quantifying NIF binding to the recombinant I-domains was developed by measuring the binding of 125 I-NIF to the recombinant I-domains captured onto glutathione-Sepharose. 1 mM Ca 2ϩ was used as the cation. After incubation for 2 h at 37°C, the beads were washed by centrifugation and counted. As shown in Fig. 2A, 125 I-NIF exhibited minimal reaction with glutathione-Sepharose. Minimal reaction also was observed between the ␣ L I-domain fusion protein and radiolabeled NIF ( Fig. 2A). With the ␣ M I-domain, binding was observed, which was dependent on the concentration of 125 I-NIF added; 50% saturation was attained at 170 pM added. The K d of the interaction of 125 I-NIF with the ␣ M I-domain GST fusion protein was estimated from a Scatchard plot to be 5 nM, consistent with the reported K d of NIF for the ␣ M I-domain (22) and for intact ␣ M ␤ 2 (50). A similar interaction of 125 I-NIF with the ␣ M I-domain that was cleaved from the GST fusion protein and immobilized directly to the microtiter plates was also observed (see Fig. 2A). The K d of this interaction was estimated to be 4 nM, indicating that the GST fusion partner did not influence NIF binding to the ␣ M I-domain.
The specificity of the interaction of radiolabeled NIF with the ␣ M I-domain was indicated by the ability of nonlabeled NIF and EDTA to block binding (Fig. 2B). The inhibition produced by unlabeled NIF at 5 nM was 80% and did not increase further when the concentration of unlabeled NIF was increased by 20-fold, indicating that ϳ20% of the binding of the radiolabeled ligand was nonspecific. In separate experiments, we found that the capacity of unlabeled NIF to inhibit the binding of 125 I-NIF was dose-dependent. Half-maximal inhibition of the specific binding of 125 I-NIF added at 170 pM binding occurred at 150 pM  Fig. 3, the interaction of each of these mutants with 125 I-NIF was greatly diminished compared with the ␣ M I-domain and was similar to that of the ␣ L I-domain. Essentially no specific binding of NIF to these mutant I-domains was detected. To begin to identify the individual residues within these three segments that mediated NIF recognition, a series of 13 triple mutants were created. In each of these triple mutants, a set of three consecutive amino acids within the ␣ M I-domain was changed to the corresponding ␣ L residues. The specific substitutions made are identified in Fig.  4. If the ␣ M and ␣ L residues were the same, the amino acid was mutated to that present in the corresponding position in the ␣ X I-domain; or if the residues were identical in all three I-domains, the amino acid was replaced with an Ala or Gly (see Fig.  4). After the DNA sequence of each mutant I-domain was confirmed, it was expressed in E. coli and purified on glutathione-Sepharose. When analyzed by SDS-PAGE, each mutant migrated as a single band of ϳ52 kDa (not shown), similar to the GST fusion proteins characterized in Fig. 1. 125 I-NIF binding to each triple mutant was then assessed. The results in Fig.  5A show the binding of the 13 triple mutants to a concentration of 170 pM 125 I-NIF, a concentration below its K d for the ␣ M Idomain. Of the 13 triple mutants, five, P147S/H148D/D149E, F150A/R151Q/R152K, G207L/R208L/T209A, G251T/Y252D/ E253S, and P257D/E258A/A259G, showed a significant reduction in NIF binding. The K d values of these five triple mutants for NIF were estimated from binding isotherms and are summarized in Table I. The values were reduced by at least 10-fold compared with the wild-type ␣ M I-domain.
NIF Binding to Single Point Mutants of the ␣ M I-domain-Next, within the five triple mutants with defective NIF binding, each of the three amino acids was mutated individually to the corresponding residue in the ␣ L I-domain (see Fig. 4). Thus, a total of 15 single mutants were created. After again confirming the DNA sequence of these single mutants, each was expressed as a GST fusion protein in E. coli and purified. The capacity of the 15 single mutants to bind NIF is summarized in Fig. 5B. Within each of the five triple mutants with reduced NIF recognition, only one of the three single mutants exhibited reduced 125 I-NIF binding. These five single mutants were D149E, R151Q, G207L, Y252D, and E258A. The K d values of these single mutants for NIF were estimated and were increased by 10 -16 times compared with the wild-type ␣ M Idomain (see Table II).
Conformational Considerations-Loss of NIF binding function in these single mutants could reflect direct involvement of the specific residues in NIF binding or conformational perturbation of the resulting ␣ M I-domain due to substitutions at these positions. Two approaches were deployed to distinguish between these possibilities. First, we considered that alterations in the cation binding function of the MIDAS motif would provide a sensitive barometer of conformational alterations, since the three-dimensional fold of the ␣ M I-domain is necessary to bring the five cation-coordinating residues into the appropriate spatially proximity. Tb 3ϩ , a fluorescent trivalent cation that replaces Ca 2ϩ in most metal-binding proteins, including in I-domains (58), shows little luminescence when free but does so when it binds and can receive energy from an adjacent tryptophan residue (59,60). An appropriate tryptophan is not present in the ␣ M I-domain, but Phe 246 is in close proximity to the MIDAS and resides on the hydrated surface in the ␣ M I-domain crystal. Phe 246 is not involved in NIF binding (see Fig. 5). Computer modeling using the known crystal structure of the ␣ M I-domain as a template suggested that this residue could be mutated to tryptophan without perturbing conformation. The results shown in Fig. 6 indicate that the strategy of substituting Trp for Phe 246 was effective. The Tb 3ϩ luminescence of the ␣ M I-domain with the Trp was Ͼ50 times greater than that of the wild-type ␣ M I-domain with Phe at position 246. Therefore, this substitution was introduced into each of the five single point mutant ␣ M I-domains with altered NIF binding function. Each of the five single mutants exhibited a similar emission spectrum (i.e. similar Tb 3ϩ luminescent properties to the wildtype ␣ M I-domain with the Trp substitution) (Fig. 6). The affinities of all five mutants estimated from concentration curves were also similar. The estimated K d values of the mutants for  Table  III. The values for the five mutants varied by less than 26% and were not significantly different. These data suggest that five point mutations did not perturb the conformation of the ␣ M I-domain.   As a second approach, we tested the prediction that introduction of these five point mutations into the ␣ L I-domain should impart NIF binding function to the mutant protein. The chimeric I-domain was expressed as a GST fusion protein and purified, and its NIF binding properties were evaluated. This chimeric I-domain did bind NIF with an affinity substantially greater than the wild-type ␣ L I-domain (see Fig. 7). Indeed, its affinity for NIF was similar to that of the wild type of the ␣ M I-domain. The K d of the chimeric I-domain for NIF was 7 nM compared with 5 nM for the wild-type ␣ M I-domain. Thus, the insertion of these five amino acids was sufficient to impart high affinity NIF binding to the mutated ␣ L I-domain.
C3bi Binding-The five segments implicated in NIF binding were previously shown to be involved in C3bi recognition using the same set of homolog scanning mutants (61). To determine whether the same residues involved in NIF binding were also involved in C3bi engagement, the binding of C3bi to the chimeric ␣ L I-domain that bound NIF was assessed. As shown in Fig. 8, wild-type ␣ L I-domain and the chimeric ␣ L I-domain failed to bind C3bi, whereas C3bi showed a positive reaction with the wild-type ␣ M I-domain in the assay used. We verified that C3bi bound to the ␣ M I-domain with the appropriate specificity profile (i.e. the interaction was inhibited by NIF and by mAb CBL189 to C3bi). Thus, the five residues that implicated in NIF binding were not sufficient to impart C3bi binding to the ␣ L I-domain. In support of this conclusion, we tested C3bi binding to each of the five single mutants, D149E, R151Q, G207L, Y252D, and E258A, with impaired NIF binding. These were immobilized onto microtiter wells, and their binding of biotinylated C3bi was assessed as in Fig. 8. The C3bi titration curves were essentially the same as that obtained with wild-type ␣ M I-domain (data not shown). Half-maximal binding for all mutant and wild type ␣ M I-domains was obtained at a 1:6 dilution of C3bi (see Fig. 8). Thus, none of the five ␣ M I-domain residues involved in NIF binding was essential for C3bi recognition.

DISCUSSION
In this study, we have sought to identify the residues within the ␣ M I-domain that establish its high affinity and specificity for NIF, thereby serving as a model for the basis for ligand recognition by ␣ M ␤ 2 . The approach of homolog scanning mutagenesis had previously identified five loops, two adjacent and three other discontinuous in amino acid sequence on the MIDAS face of the ␣ M I-domain, as providing key contact sites for NIF recognition (50). The previous study had been performed in the context of the intact ␣ M ␤ 2 heterodimer, and the involvement of these segments has been confirmed in the present analysis by expressing these homolog scanning mutations in a different context (i.e. as I-domains). Whereas the ␣ M Idomain, either as a GST or as a free recombinant protein, bound NIF with the same high affinity as intact ␣ M ␤ 2 , the homolog scanning mutants showed negligible binding. It is also noteworthy that the ␣ L I-domain failed to bind NIF. This result is consistent with previous reports for both the ␣ L Idomain and ␣ L ␤ 2 (22,37,50), although other studies have provided evidence for an interaction of NIF (62) or NIF peptides with ␣ L ␤ 2 (63).
To further define the contact residues for NIF binding, a set of triple mutants was developed in which sets of three amino acid residues within the three implicated segments were changed to the corresponding ␣ L I-domain residues (or to ␣ x , Ala, or Gly residues if conserved in ␣ L ), and the capacity of the resulting mutants to bind NIF was assessed. Of the 13 triple mutants, five lost their ability to bind NIF with high affinity. The reduction in affinity of each of these triple mutants for NIF was at least 10-fold. Then, within these five triple mutants, the individual amino acids were mutated. The outcome of these data is the identification of residues Asp 149 , Arg 151 , Gly 207 , Tyr 252 , and Glu 258 within the ␣ M I-domain as being critical to NIF binding. Mutation at any one of these positions reduced affinity for NIF by at least 10-fold.  To consider whether mutation at these individual positions altered NIF binding by perturbing the conformation of the ␣ M I-domain, the effect of the substitutions on the affinity of the MIDAS motif for cation was assessed by Tb 3ϩ luminescence spectroscopy. Since the five amino acids of the MIDAS are brought into the appropriate spatial distance to coordinate with the bound cation by the three-dimensional fold of the I-domain, changes in the affinity of the MIDAS for cation should be a sensitive indicator of conformational change within the ␣ M I-domain. In a sense, this approach is similar to the use of the Ca 2ϩ monoclonal antibody 24, which has been utilized previously (64,65) to assess the conformational integrity of the ␣ M ␤ 2 , but the specific reporter is different. In order to implement this approach, Phe 246 , which resides on the hydrated surface and in close proximity to the MIDAS based on the crystal structure of the ␣ M I-domain, was mutated to a tryptophan to potentially enhance energy transfer to bound Tb 3ϩ . Substitution at this position did not alter NIF binding (see Fig.  5), was not predicted to alter the conformation of the ␣ M Idomain based upon computer modeling, and did, indeed, enhance the efficiency of energy transfer to bound Tb 3ϩ (see Fig.  6). Using this approach, all five point mutants were found to bind Tb 3ϩ with the same affinity as the wild-type ␣ M I-domain.
The involvement of the five residues identified through the loss-of-function resulting from their mutation was corroborated by a gain-in-function approach. When these five residues were inserted into the I-domain of ␣ L , the chimera acquired NIF binding function. The affinity of the chimeric I-domain was 7 nM compared with 5 nM for the authentic ␣ M I-domain. The slight difference in affinity may indicate that still other residues might contribute to NIF recognition. It is important not to overinterpret the result of this experiment to imply that these are the only residues involved in NIF binding. Residues that are conserved between the ␣ M I-and ␣ L I-domains may be necessary to mount a high affinity interaction of NIF with the ␣ M I-domain but may not by themselves support an interaction of sufficient affinity to detect NIF binding to the ␣ L I-domain. It is also noteworthy that two of the five residues implicated in NIF binding to the ␣ M I-domain, Gly 207 and Ala 258 , are conserved in the ␣ X I-domain, which itself does not bind NIF with high affinity (22). This suggests that the three remaining residues may play a particularly prominent role in defining the specificity of the ␣ M I-domain for NIF.
Based on the crystal structure (54), all of the five amino acids involved in NIF binding reside at the hydrated surface of the ␣ M I-domain; three of the five lie on the top of the MIDAS face, and two lie at its edges (see Fig. 9). Arg 151 , Gly 207 , and Tyr 252 are in relatively close proximity to the cation bound in the MIDAS motif, whereas Asp 149 and Glu 258 are more spatially distant. The capacity of NIF to contact several spatially distant residues over the ␣ M I-domain may account for the high affinity of its interaction and its capacity to inhibit the binding of multiple ligands that interact with the MIDAS face. The involvement of the MIDAS face is consistent with the role of cation in NIF binding (37), which is again corroborated in this study. In a previous study, Rieu et al. (45) mutated residues in the surface loops and helices of the MIDAS face of the ␣ M Idomain and implicated Gly 143 , Asp 149 , Glu 178 , Glu 179 , and Arg 208 in NIF binding. Only one of these residues, Asp 149 , coincides with our findings. The disparity regarding the role of Arg 208 is particularly difficult to explain, since the same mutation, R208L, was introduced by both groups. The segment containing the two glutamic acids, Glu 178 and Glu 179 , had been previously changed to Thr and Ser, respectively, in our original homolog scanning mutagenesis study without effect (50), as contrasted to the alanines by Rieu et al. (45). Gly 143 had not been altered in our original mutagenesis study but was absent in an ␣ M I-domain that bound NIF with an affinity comparable with the intact receptor. One potential explanation for these differences in results relates to the activation states of the ␣ M I-domains utilized. Relatively subtle differences in the length of ␣ M I-domain can influence its activation state (66 -68), which can, in turn, exert a profound influence on ligand recognition. Differences in activation state may also account for reports that integrin ␣ L ␤ 2 can recognize NIF (62,69).
To date, many lines of evidence have emphasized that many ligands share overlapping binding sites within ␣ M ␤ 2 , including the fact that NIF blocks the interaction of many ligands (C3bi, ICAM-1, fibrinogen, and several immobilized substrates) with the receptor (22,37,70). However, while the binding sites for these ligands may be overlapping, they need not be identical (46,61). Indeed, the ␣ M I-domain residues involved in NIF binding were not critical for C3bi binding (Fig. 8), although all segments originally implicated in NIF binding were also involved in C3bi binding (61). The nonidentity of these ligand binding sites is consistent with our previous data showing that mutation of Lys 245 to Ala significantly reduced C3bi but had no effect on NIF binding to ␣ M ␤ 2 (61). Thus, a model can now be proposed in which many of the same loops and helices of the ␣ M I-domain, on or near its MIDAS face, may engage ligands, but different amino acid residues within these structures may contact the ligands. Accordingly, the ␣ M I-domain may present a mosaic of different amino acids that are capable of contacting different ligands. Direct support for this mosaic model is seen within the Lys 245 -Arg 261 segment of the ␣ M I-domain, which has been implicated in both NIF and fibrinogen recognition (71). Key contact amino acids in this loop for fibrinogen binding are Phe 246 , Asp 254 , and Pro 257 , whereas Tyr 252 and Glu 258 are involved in NIF binding. This same ␣ M I-domain segment is also involved in C3bi recognition by ␣ M ␤ 2 (61). Although it is not known which residues in this segment serve as contact sites for C3bi, the residues involved in NIF binding, Tyr 252 and Glu 258 , do not appear to be critical. Pro 147 -Arg 152 and Pro 201 -Lys 217 are also involved in recognition of C3bi (61) and the fungal pathogen Candida albicans (although this latter ligand also interacts with several distinct regions of the ␣ M I-domain (70)). Whether different residues within these shared loops will be involved in contacting these ligands remains to be assessed. The systematic approach delineated in this study involving loss of function, gain in function, and conformational assessments provides a means for assessing whether this mosaic model provides a molecular basis for the capacity of ␣ M ␤ 2 and other integrins to recognize multiple and in many cases unrelated ligands.