Identification and Reconstruction of the Binding Site within αMβ2 for a Specific and High Affinity Ligand, NIF*

Engagement of the αMβ2 (CD11b/CD18, Mac-1) integrin on neutrophils supports adhesion and induces various cellular responses. These responses can be blocked by a specific ligand of αMβ2, neutrophil inhibitory factor (NIF). The molecular basis of αMβ2-NIF interactions was studied. The single chain αM subunit, expressed on the surface of human 293 cells, bound NIF with an affinity equivalent to that of αMβ2 heterodimer. This observation, coupled with previous data showing that the αMI domain alone supported high affinity NIF binding, indicated that the binding site for NIF is restricted to the I domain. Guided by the crystal structure of the αMI domain, 16 segments corresponding to the entire outer hydrated surface of αMI domain were switched to their counterparts sequences in αL, which does not bind NIF. Surface expression and heterodimer formation were achieved for all mutants, and correct folding was confirmed. Of the 16 switches, only 5 affected NIF binding substantially, reducing affinity by 8–300-fold. These data confined the NIF-binding site to a narrow region composed of Pro147-Arg152, Pro201-Lys217, and Asp248-Arg261 of αM. Verifying this localization, when these segments were introduced into the αXI-domain, the resulting chimeric receptor was converted into a high affinity NIF-binding protein.

NIF (neutrophil inhibitory factor), a novel glycoprotein isolated from canine hookworms, was originally identified as an inhibitor of a number of neutrophil functions, such as adhesion to endothelial cells and adhesion-dependent release of hydrogen peroxide (8). These functional effects resulted from its specific binding to ␣ M ␤ 2 but not to other ␤ 2 integrins. Subsequently, we and others have reported that NIF completely blocked ␣ M ␤ 2 -mediated binding of C3bi (9), ICAM-1 (10), and adhesion to protein-coated surfaces (11), and partially blocked fibrinogen binding (12) to ␣ M ␤ 2 -bearing 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 animal models (13,14).
The I domain, an inserted region of ϳ200 amino acids in the ␣ M subunit (␣ M I-domain), contributes to NIF binding to ␣ M ␤ 2 (10,12). I domains with highly conserved amino acid sequences are also found in several other integrin ␣-subunits as well as in other proteins, such as vWF, and mediate a variety of proteinprotein interactions, including ligand binding to integrins (15). Using blocking mAbs which map to the I domain or recombinant ␣ M I domain itself, this region has been implicated in the binding of ICAM-1, fibrinogen (16), and C3bi (9), as well as NIF, to ␣ M ␤ 2 . Collectively, these data suggest that the ␣ M I domain is an independent structural unit, capable of interacting with many different proteins. Nevertheless, the ligand binding functions of ␣ M ␤ 2 are not solely a property of its I domain. We (11) and Bajt et al. (17) have shown that mutations of Asp 134 , Ser 136 , or Ser 138 in ␤ 2 subunit abrogated the binding of ␣ M ␤ 2 to C3bi and adhesion to protein-coated surfaces, suggesting that, in addition to I domain, the ␤ subunit also is involved in the recognition of certain ligands by ␣ M ␤ 2 .
In this study, we have utilized NIF as a model ␣ M ␤ 2 ligand and have sought to delineate the molecular basis for its interaction with ␣ M ␤ 2 . Homolog-scanning mutagenesis (18) in which sequences within the ␣ M I domain have been switched to the homologous sequences within the ␣ L I domain has been used to map the NIF-binding site in ␣ M ␤ 2 . The NIF-binding site identified through the loss-of-function mutations was confirmed by a gain-in-function experiment, whereby the I domain of ␣ X was converted into a NIF-binding protein. Given the ability of NIF to inhibit the binding of many ligands to ␣ M ␤ 2 and the structural similarities among I domains, the molecular details of NIF-␣ M ␤ 2 interactions may also apply to other ligands of ␣ M ␤ 2 and to other I domain containing intergrins. Segment Switches by Site-directed Mutagenesis-Stretches of 7-12 amino acids within the ␣ M I domain were switched to the corresponding sequences in ␣ L . All such segment switches were created by oligonucleotide-directed mutagenesis using uracil-containing single stranded M13mp18 DNA (19). To facilitate the mutagenesis, two unique restriction sites, ClaI at position 535 (Ile 139 ), and NheI, at position 1141 (Ala 342 ), flanking the I domain of ␣ M , have been introduced into ␣ M previously (11). To switch a segment of 7-11 amino acids from ␣ M to ␣ L , the mutagenic primer was designed to contain the corresponding mutations and 15 additional unchanged bases at each end. The length of the primer was typically 54 bases. The site-directed mutagenesis was performed according to our published procedure using T7 DNA polymerase (19). The mutant was identified by DNA sequencing of 5 randomly picked plaques. DNA sequencing of the entire I domain was conducted, confirming the presence of the desired mutations and the correctness of the rest of the I domain. The mutated I domain was transferred back to the expression vector pCIS2M-␣ M using the unique ClaI and NheI restriction sites. The cDNA of ␣ M and ␤ 2 were inserted separately in the pCIS2M expression vector employing XbaI and XhoI sites; and expression of ␣ M or ␤ 2 was under the control of the human cytomegalovirus promoter and enhancer.

Materials-Human
To generate the chimeric molecule, ␣ M (I/␣ X )␤ 2 , where the ␣ M I domain was replaced with ␣ X I domain, the fragment of the ␣ X I-domain (Ile 137 to Val 340 ) containing ClaI and NheI restriction sites was prepared by reverse transcription and polymerase chain reaction using total RNA from polymorphonuclear cells and the following two primers: 5Ј-CC-CAAGACAGGAGCAGGACATTG-3Ј (forward) and 5Ј-TGTGAAGCT-AGCGCTGAAGCCCTC-3Ј (reverse). The reverse primer contains a NheI site (GCTAGC), and a naturally occurring ClaI site exists in the ␣ X cDNA at nucleotide position 530.
FACS Analysis-Approximately 10 6 cells in HBSS containing 1 mM Mg 2ϩ were incubated with 5 g of mAb for 30 min at 4°C. A subtypematched mouse IgG served as a control. After 3 washes with PBS (10 mM NaPO 4 , 150 mM NaCl, pH 7.4), cells were mixed with fluorescein isothiocyanate goat anti-mouse IgG(HϩL) F(abЈ) 2 fragment (1:20 dilution) (Zymed Laboratory, San Francisco, CA), kept at 4°C for another 30 min, washed with PBS, and then resuspended in 500 l of PBS. The FACS analysis was performed using FACScan, counting 10,000 events. For dual color FACS analysis, the cells (10 6 ) were stained with OKM1 (5 g) and biotinylated NIF (0.5 g) for 30 min at 4°C, followed by three washes with PBS. These cells were then mixed with fluorescein isothiocyanate-avidin conjugate and phycoerythrin goat anti-mouse IgG(HϩL) F(abЈ) 2 fragment (1:20 dilution) (Zymed Laboratory), kept at 4°C for another 30 min, washed with PBS, and then resuspended in 500 l of PBS. FACS analyses was performed as described above. The same procedure was used for mAb 24 staining, except that 1 mM Mn 2ϩ was substituted for the 1 mM Mg 2ϩ and incubations were at 37°C.
Analytical Procedures-The procedures used for NIF binding, and surface labeling and immunoprecipitation of cells have been described (11). Briefly, different concentration of NIF (0 -20 g) were incubated with 2 ϫ 10 6 ␣ M ␤ 2 -expressing cells at 4°C for 30 min. The bound NIF was separated from the free NIF by centrifugation through a 20% sucrose solution and counted with a ␥-counter. The NIF titration data were fitted to a single site binding model using the equation: where B max is the maximal NIF binding and K d is the dissociation constant, using a program in SigmaPlot (Jandel Co., San Rafael, CA).

RESULTS
The ␣ M Subunit Is Capable of High Affinity NIF Binding-When introduced into human kidney 293 cells in the absence of transfected ␤ 2 , the ␣ M subunit was expressed on the cell surface, albeit at low levels relative to the heterodimer. Immunoprecipitation of surface-labeled cells with OKM1 yielded a band of 165 kDa on polyacrylamide gels in SDS under both reducing and nonreducing conditions (Fig. 1A). No bands in the vicinity of 95 (␤ 2 ) or 120 (␤ 1 ) kDa were observed, suggesting that the surface-expressed ␣ M is not complexed with its natural partner, ␤ 2 , or with other typical integrin ␤ subunits. No bands were observed for mock transfection with either OKM1 or TS1/18 (Fig. 1B), verifying the specificity of the immunoprecipitation. The presence of ␣ M was confirmed by FACS analysis. The transfected cells were positive with OKM1, LM2/1, 2LPM19c, 44a, and 904, mAbs to the ␣ subunit, whereas mAbs to the ␤ 2 subunit, TS1/18, IB4, or MHM23 were unreactive. Bilsland et al. (21) also detected surface expression of the ␣ M subunit in the absence of ␤ 2 in COS cells. The ability of the ␣ M subunit to interact with soluble 125 I-NIF was assessed. As shown in Fig. 1C, the ␣ M -expressing cells bound NIF in a dose-dependent and saturable fashion. The specificity of this binding was verified by addition of a 20-fold excess of unlabeled NIF or 1 mM EDTA: in both cases, 125 I-NIF binding was reduced by more than 99%. Scatchard plots of the NIF binding isotherms were consistent with a single class of binding sites with respect to affinity and yielded a dissociation constant (K d ) of 2.1 nM. This value is very similar to the K d of 7 nM for NIF binding to heterodimeric ␣ M ␤ 2 . To determine the nature of the molecule on the cell surface that binds NIF, lysate of surfacelabeled ␣ M -transfected cells were incubated with biotinylated NIF. The NIF-receptor complex was captured with avidin-agarose resin and analyzed on 7% SDS-PAGE. As shown in Fig.  1B, only one band of 165 kDa was present. In contrast, two bands (165 and 95 kDa) were observed for the ␣ M ␤ 2 -expressing cells. The specificity of this assay was demonstrated by the absence of any band for the mock transfectant. These data indicate that, unlike other ␣ M ␤ 2 ligands (C3bi and adhesion (11)), all major contributing elements of the binding site for NIF reside in the ␣ subunit of ␣ M ␤ 2 . Consistent with this conclusion, we also found that: 1) mutations in the ␤ 2 subunit, which abrogated binding of other ligands, did not affect NIF binding of ␣ M ␤ 2 complex (11); and 2) replacement of the ␣ M with the ␣ X I domain in the context of the ␣ M ␤ 2 heterodimer completely abrogated NIF binding (see below, Fig. 5).
The I Domain Peptide N 232 AFKILVVITDGEK Is Not Required for NIF Binding-Rieu et al. (10) had previously used synthetic peptides and implicated two candidate ␣ M sequences in NIF binding: the A7 peptide, N 232 AFKILVVITDGEK, was the most potent inhibitor of NIF binding to the ␣ M I domain (10). To test the role of this sequence in NIF binding, we created two switch mutants in which the amino-and carboxylterminal portions of this peptide were changed individually to the corresponding sequences in ␣ L (␣ L ␤ 2 does not bind NIF (8): ␣ M (K 231 NAFKILVVITDGEK 245 FG) to ␣ L (PDATKVLIIIT-DGEATD). Thus, the first mutant, ␣ M (K 231 NAF), changed the amino-terminal non-conserved KNAF to PDAT, and the second mutant, ␣ M (K 245 FG), changed to non-conserved KFG to ATD. The hydrophobic cluster in the center of the peptide sequence is well conserved and was not altered. The mutant ␣ M vectors were transfected together with ␤ 2 into 293 cells. Both mutants were expressed well on the cell surface as judged by FACS analysis, and surface labeling and immunoprecipitation with OKM1 (data not shown). When stained with LM2/1 mAb, the mean fluorescence for the ␣ M (K 231 NAF)␤ 2 mutant was 392.0, compared with 398.4 for wild-type. A similar result was obtained for the ␣ M (K 245 FG)␤ 2 mutant. When stained with OKM1, the mean fluorescence was 377.2 for the mutant and 380.1 for the wild-type. Both mutants also bound NIF with high affinity (Fig. 2, and Table II), indicating that this peptide is not centrally involved in NIF binding when placed in the context of the intact receptor.
Homolog-scanning Mutagenesis-To systematically define the NIF-binding site in ␣ M , a homolog-scanning mutagenesis (18) strategy was implemented. Accordingly, guided by the crystal structure (15), the entire hydrated surface of the ␣ M I domain was replaced with sequences of the ␣ L I domain in segments of 7-11 amino acids. To apply this approach to the I domain of ␣ M (ϳ200 amino acids), 16 segments were switched. The primers used are listed in Table I. The efficiency of mutagenesis was typically 60%, with the 7-amino acid segment switches having substantially higher efficiency (Ͼ90%) than the longer 10-amino acid segment switches (ϳ30%). The appropriate DNA sequence of the entire I domain (from Ile 139 to Ala 332 ) was confirmed for each mutant before transferring back into the ␣ M subunit cDNA.
The functional consequences of these segment switches were initially investigated by transient expression in 293 cells. Forty-eight hours after transfection of ␣ M , together with ␤ 2 subunit, the cells were detached from tissue culture dish and double stained with OKM1 and biotinylated NIF. A representative dual-color FACS analysis is shown in Fig. 3. With wildtype ␣ M ␤ 2 transfectant, 1.83% of the cells were positive with both OKM1 and NIF, whereas less than 0.02% of mock transfected cells were positive. Two distinct patterns were observed for the mutants. One, represented by ␣ M (E 178 -T 185 ) in Fig. 3 The NIF-binding Site Is Composed of Segments P 147 -R 152 , P 201 -G 207 , R 208 -K 217 , D 248 -T 252 , and E 253 -R 261 -To quantitate the binding affinities of each mutant receptor for NIF, stable cell lines were established. All 16 ␣ M ␤ 2 mutants were cell surface expressed as heterodimers: immunoprecipitations of 125 I-surface-labeled cells with OKM1 yielded ␣ subunits of 165 kDa and ␤ subunits of 95 kDa on 7% SDS-PAGE (a representative gel is shown in Fig. 4A), similar to the patterns obtained with that of recombinant wild-type and naturally-occurring ␣ M ␤ 2 (11). Identical patterns were obtained when the immunoprecipitations were performed with TS1/18, a mAb to the ␤ 2 subunit (data not shown). The specificity of this assay was verified by the absence of any band when 125 I-surface-labeled mock transfected cells were immunoprecipitated with either OKM1 or TS1/18 (Fig. 1B), or when wild-type ␣ M ␤ 2 -expressing cells were immunoprecipitated with IV.3, an irrelevant mAb (11). NIF binding was measured as a function of ligand concentrations, and representative binding isotherms are presented in Fig. 4B. With all mutants, dose-dependent and saturable binding of 125 I-NIF was observed, which could be inhibited by unlabeled NIF and EDTA. The binding isotherms could be fitted to a one-site model as described under "Experimental Procedures." The K d values calculated from these binding data are summarized in Table II. The K d of 7 nM for wild-type ␣ M ␤ 2 was nearly identical to the K d of 5 nM reported for naturallyoccurring ␣ M ␤ 2 on neutrophils (8). The expression levels, reflected by the maximal NIF binding, were similar for all mutants, differing by less than 4-fold. The following mutants had similar

The Defective Mutants Possess Correct Conformations-
Given the similarities in the three-dimensional structures of ␣ M I and ␣ L I domains (15,22), it is unlikely that the defects in NIF binding of the five identified mutants arise from incorrect folding of their ␣ M I domain. This assertion was supported by a series of additional experiments. First, FACS analyses were performed for all 16 mutants with a panel of mAbs against ␣ M ␤ 2 , including OKM1, LM2/1, M1/70, TS1/18, and MHM23. The former three are ␣ M -specific, and the latter two are ␤ 2specific. In all cases, the mAbs reacted well with wild-type and all five mutant receptors. Second, the reactivity of the five mutants with a conformation-dependent antibody, mAb 24, which has been used to probe the cation-dependent conformational integrity of ␣ M I domain (23,24), was assessed. FACS analyses were performed on the five mutants, together with the wild-type receptor. As a control, FACS analyses were performed in parallel with OKM1, a conformation-independent mAb. The ratios of the mean fluorescence intensities of the various cell lines with mAb 24 and OKM1 are shown in Fig. 4C. All of the mutants defective in NIF binding exhibited the capacity to bind mAb 24. Mutant ␣ M (P 201 -G 207 ) showed a slightly attenuated reactivity with mAb 24, suggesting a subtle perturbation of its structure. Nevertheless, binding of mAb 24 to this and the other mutants remained cation-dependent; addition of FIG. 4. A, immunoprecipitation of 125 Isurface-labeled stable cell lines expressing wild-type and mutant ␣ M ␤ 2 . ␣ M ␤ 2 -expressing cells (1 ϫ 10 6 ) were surface labeled with Na 125 I and lactoperoxidase, and immunoprecipitated with 10 g of OKM1. After washing, the immunoprecipitates were subjected to SDS-PAGE (7% gels under nonreducing conditions) and exposed to Kodak XAR-5 film for 48 h.   (8), the expressed heterodimeric receptor, ␣ M (I/␣ X )␤ 2 , had no NIF binding capacity (Fig. 5), albeit expressed well on 293 cell surface. When stained with mAb OKM1, the mean fluorescence of the ␣ M (I/␣ X )␤ 2 chimera was 281.2, compared with 279.0 for wild-type ␣ M ␤ 2 receptor. This chimeric receptor continued to recognize the ␣ X ␤ 2 mAb, clone 3.9, and to rosette with C3bi (data not shown), indicating the retention of functional integrity. To provide direct evidence that the identified segments (Pro 147 -Arg 152 , Pro 201 -Gly 207 , Arg 208 -Lys 217 , Asp 248 -Thr 252 , and Glu 253 -Arg 261 ) constitute a functional NIF-binding site, we sought to impart NIF binding function to the chimeric receptor, ␣ M (I/␣ X )␤ 2 , by grafting the five segments from ␣ M into their counterparts' sequences in the ␣ X I domain. The chimeric molecule, ␣ M (IЈ/␣ X )␤ 2 , containing the reconstructed ␣ X I domain, was expressed in 293 cells, and NIF binding was assessed using different concentrations of 125 I-NIF (Fig. 5). While the chimeric receptor, ␣ M (I/␣ X )␤ 2 , containing the wild-type ␣ X I domain, did not bind NIF, upon grafting the five ␣ M segments into ␣ X I domain, full NIF binding function was imparted to the modified receptor. The reconstructed I domain has a K d of 2 nM for NIF binding, essentially identical to that of the original wild-type ␣ M I domain (7 nM). This interaction was blocked completely by EDTA (1 mM) or unlabeled NIF (10 g), verifying that NIF binding to the reconstructed ␣ X I domain remained cation-dependent and specific. DISCUSSION In this study, we have sought to map the binding site for NIF, a model ligand for the I domain of ␣ M ␤ 2 . First, we have demonstrated that the binding surface for NIF is located exclusively in the I domain of the receptor. This conclusion is supported by the observations that ␣ M alone (Fig. 1C) or the expressed I domain (10, 12) support NIF binding with an affinity similar to that of the intact ␣ M ␤ 2 heterodimer; that mutations in ␤ 2 subunit, known to abolish the binding of several other ligands to ␣ M ␤ 2 (11,17), have no effect on NIF binding; and that a swap of the ␣ M I domain for the ␣ X I domain in the context of the intact ␣ M ␤ 2 receptor completely abolishes the NIF binding activity of the resulting chimeric receptor (Fig.  5). These observations also distinguish NIF binding from C3bi binding and adhesion to protein substrates mediated by ␣ M ␤ 2 , interactions which require both subunits (11). Second, we have scanned the entire hydrated surface of I domain by substituting 16 ␣ L segments of 7-10 amino acids for their corresponding sequences in the I domain of ␣ M ␤ 2 . The approach of homologscanning mutagenesis had been previously employed to identify receptor and antibody epitopes in human growth hormone (18). Since the homologous segments of ␣ M and ␣ L adopted similar structures (15,22), and the structurally important residues involved in cation coordination (Asp 140 , Ser 142 , Ser 144 , Thr 209 , and Asp 242 ) are preserved, the structural integrity and the essential cation-binding site was maintained. Therefore, any functional changes resulting from the switches should reflect sequences that have a direct role in NIF binding. This strategy does not exclude a role for residues conserved among the two homologous proteins in NIF binding. Nevertheless, our study does identify segments that are required for a high af-  (15) using the Biosym software. The backbone of the ␣ M I domain is shown with oxygen atoms in red, hydrogen atoms in yellow, carbon atoms in green, and nitrogen atoms in blue. The NIF-binding site, as mapped in this study, is composed of three segments (P 147 -R 152 , P 201 -K 217 , and D 248 -R 261 ) and is shown as a red ribbon.
FIG. 5. The chimeric ␣ M (I/␣ X )/␤ 2 receptor, created by grafting of the identified NIF-binding site into ␣ X I domain, acquires the ability to interact with NIF with high affinity. Cells (2 ϫ 10 6 ) expressing the chimeric receptor, ␣ M (I/␣ X )/␤ 2 , containing the wild-type ␣ X I (E) or the five segments implicated in NIF binding (q) were incubated with different concentrations of 125 I-NIF in HBSS containing 2.5 mM Ca 2ϩ for 30 min at 22°C. Bound and free ligands were separated by centrifugation through 20% sucrose, and the cell associated radioactivity was measured. The data are representative of two independent experiments.