Dual Function for a Unique Site within the (cid:1) 2 I Domain of Integrin (cid:2) M (cid:1) 2 *

Integrin activation has been postulated to occur in part via conformational changes in the I domain of the (cid:1) subunit (the (cid:1) I domain), especially near the F- (cid:2) 7 loop, in response to “inside-out” signaling. However, direct evi-dence for a role of the F- (cid:2) 7 loop in ligand binding and activity modulation is still lacking. Here, we report our finding that the F- (cid:2) 7 loop (residues 344–358) within the (cid:1) 2 I domain has dual functions in ligand binding by (cid:2) M (cid:1) 2 . On the one hand, it supports intercellular adhesion molecule 1 (ICAM-1) binding to (cid:2) M (cid:1) 2 directly as part of a recognition interface formed by five noncontiguous segments (Pro 192 –Glu 197 , Asn 213 –Glu 220 , Leu 225 –Leu 230 , Ser 324 –Thr 329 , and Glu 344 –Asp 348 ) on the apex of the (cid:1) 2 I domain. On the other hand, it controls the open and closed conformation of the (cid:2) M (cid:1) 2 receptor, thereby indi- rectly affecting (cid:2) M (cid:1) 2 binding to other ligands. Switching the five constituent sequences of the ICAM-1-binding site within the (cid:1) 2 I domain to their (cid:1) 1 counterparts de- stroyed ICAM-1 binding but had no effect on the gross conformations of the receptor. Of the five ICAM-1 This resulted in (cid:4) reso- units of ICAM-1. A blank -ethyl-N and ethanolamine-blocked was used as a reference surface. Binding data were obtained by injecting various concentrations of the peptides over the ICAM-1-Fc surface at a flow rate of (cid:4) l/min. The experiments were conducted in a Tris buffer (20 Tris, M containing 1 m M MnCl 2 . The chip surface was regenerated with 0.1 M HCl. All of the data were analyzed using the BIAevaluation 3.0 program, subtracting binding to the blank flow cell to account for any nonspecific binding. To analyze steady state peptide binding, the peak response levels were plotted against the peptide concentrations, which were then fit to a single-site model (Langmuir isotherm) to determine the dissociation constants ( K d ). The specificity of peptide binding was further verified using the corresponding scrambled controls.

Integrins are heterodimeric surface receptors that play essential roles in cell-cell and cell-matrix interactions (1,2). One of the most prominent features of the integrin receptor is its ability to change conformation in response to intracellular activation signals ("inside-out" signaling). Based on the crystal structures of free and ligand-bound ␣ V ␤ 3 , several models have been proposed recently to explain the molecular mechanisms underlying receptor activation, including the "switchblade" model (3), the "bell-rope" model (4), and the "deadbolt" model (5). Despite the many differences, these models all predict that the F-␣ 7 loop (corresponding to residues 344 -358 of the ␤ 2 I domain) is critical to the transition of the ␤I domain from closed to open conformation. In particular, the deadbolt model hypothesizes that residues within the F-␣ 7 loop are in close contact with the residues in the extended CD loop of the ␤TD domain, such that the ␤TD domain acts as a deadbolt that freezes the movement of the F-␣ 7 loop. Movement of the ␤TD domain as a result of the inside-out signaling will disrupt its interaction with the F-␣ 7 loop and thus induce conformational changes within the ␤I domain and ultimately receptor activation (5). Although very attractive, the validity of this model needs to be tested experimentally.
Recently, using the homolog-scanning mutagenesis approach (22), we have identified a region within the ␤ 2 I domain (residues 125-385) that interacts directly with two ␣ M ␤ 2 ligands (Fg and C3bi) (23). Given that ICAM-1 behaves differently from C3bi in its interaction with ␣ M ␤ 2 (20), whether this same region is also involved in ICAM-1 binding is still unknown. In this study, we screened 17 ␣ M ␤ 2 receptors that contain individual substitutions within the ␤ 2 I domain for ICAM-1 binding. The results from this study demonstrate that ICAM-1 recognizes a novel region on top of the ␤ 2 I domain, which is well separated from the binding sites for C3bi and Fg. Most impor-tantly, the F-␣ 7 loop is located within this identified ligandbinding site. Perturbation of the F-␣ 7 loop could directly switch the conformation of the ␤ 2 I domain as well as other critical regions within ␣ M ␤ 2 from closed to open state, thus providing the first experimental support for a role of the F-␣ 7 loop of the ␤I domain in receptor activation. Altogether, our data suggest that inside-out signaling could activate the integrin receptor either directly, by changing the shape of the ligand-binding pocket, and/or indirectly, by affecting the global conformation of the heterodimeric receptor.

EXPERIMENTAL PROCEDURES
Materials-Human kidney 293 cells and the expression vector pCIS2M were gifts from Dr. F. J. Castellino (Notre Dame, IN). The cDNAs of CD11b and CD18 were obtained from Dr. B. Karan-Tamir (Amgen, Thousand Oaks, CA). mAb 24 was provided by Dr. Nancy Hogg (Imperial Cancer Research Fund, London, UK); mAb CBRM1/5 was obtained from eBioscience (San Diego, CA); IB4, OKM1, and TS1/18 were obtained from the ATCC (Manassas, VA); MEM148 was obtained from Serotec (Raleigh, NC); mAb 6.7 was from Pharmingen (San Diego, CA); and mAb MEM-48 was obtained from Biodesign (Kennebunk, ME). The recombinant ␥-module of Fg was provided by Dr. L. Medved (American Red Cross, Rockville, MD). All other reagents were the highest grade available from Sigma unless otherwise noted.
Site-directed Mutagenesis and Establishment of Stable Cell Lines-The detailed procedures for homolog-scanning mutagenesis and establishment of stable cell lines expressing wild type and the 17 ␣ M ␤ 2 mutants in human kidney 293 cells have been published (24). To create constitutively active ␣ M I domains, we changed residue Ile 316 to Gly, which enables the ␣ M I domain to exist in an open conformation with constitutive ligand binding activity (25), and alternatively, we changed both Phe 297 and Ala 304 to Cys (the resulting Cys 297 -Cys 304 linkage is compatible with an open conformation of the ␣ M I domain (26)), using mutagenic primers 5Ј-CGGGAGAAGGGCTTTGCGATCGAG-3Ј and 5Ј-GATCACGTGTGCCAGGTGAATAACTTTGAGTGTCTGAAGACCATT-3Ј, respectively. Human 293 cells expressing ␣ M ␤ 2 mutants that contain the active ␣ M subunits with either the wild type or different mutant ␤ 2 subunits were prepared as described previously. To obtain cell lines that express equivalent receptor numbers as wild type ␣ M ␤ 2 , each mutant cell line was subcloned by cell sorting using the ␣ M -specific mAb 2LPM19c. Up to 20 colonies were picked and analyzed for integrin expression by FACS analysis. Cells expressing levels of receptor similar to expressed wild type levels of ␣ M ␤ 2 were selected and subcloned. To exclude the possibility of subcloning artifacts, all of our studies have been repeated using the original pool for every mutant.
Preparation of ICAM-1-Fc Fusion Protein-The cDNAs of human IgG1 and human ICAM-1 were inserted separately into the pCIS2M expression vector (27), and the resultant ICAM-1-Fc fusion protein was then expressed in Chinese hamster ovary cells by co-transfection with pRSVneo using Lipofectamine (Invitrogen). Stable cells were established by selection with 800 g/ml G418. The ICAM-1-Fc fusion protein was purified from the conditioned media by affinity chromatography using protein G-Sepharose, and the purity of the purified protein was confirmed by a single band of 140 kDa on 10% SDS-PAGE under reduced conditions. FACS Analysis-A total of 10 6 cells expressing wild type or mutant ␣ M ␤ 2 in Hanks' balanced salt solution containing 1 mM Mg 2ϩ and 1 mM Ca 2ϩ were incubated with 1 g of mAb for 30 min at 4°C, except for mAb 24, which were carried out at 37°C. A subtype-matched mouse IgG served as a control. After washing with phosphate-buffered saline, the cells were mixed with fluorescein isothiocyanate goat anti-mouse IgG(HϩL) F(abЈ) 2 fragment (1:20 dilution) (Zymed Laboratory) and kept at 4°C for another 30 min. The cells were then washed with phosphate-buffered saline and resuspended in 500 l of DPBS. FACS analysis was performed using FACScan (Becton-Dickinson), counting 10,000 events. Mean fluorescence intensities were quantified using the FACScan program.
Adhesion Assays-Adhesion of the ␣ M ␤ 2 -expressing cells to ICAM-1 was conducted based on our published method (28), except that the cells were pretreated with an activating ␤ 2 -specific mAb (MEM48) (29). Briefly, a 24-well polystyrene plate was coated with ICAM-1-Fc (50 g/ml) and then blocked with 400 l of 0.05% polyvinylpyrrolidone in DPBS. A total of 2 ϫ 10 6 cells in Hanks' balanced salt solution containing 1 mM Ca 2ϩ and 1 mM Mg 2ϩ in the presence or absence of 2 mg/ml synthetic peptides or 50 nM NIF (specific for the ␣ M subunit) were added to each well and incubated at 37°C for 20 min. Unbound cells were removed by three washes with DPBS, and adherent cells were quantified by cell-associated acid phosphatase as described previously (28).
Solid Phase Binding Assays-To test the interaction between the identified sequences of the ␤ 2 I domain and ICAM-1, 96-well microtiter plates (Immulon 4BX; Dynex Technologies Inc., Chantilly, VA) were coated with different synthetic peptides at 2 mg/ml overnight at 4°C and postcoated with 3% BSA for 2 h at room temperature. ICAM-1-Fc (50 g/ml) in Tris-buffered saline (20 mM Tris, 150 mM NaCl, 1 mM Ca 2ϩ , 1 mM Mg 2ϩ , pH 7.4) was added to the wells and incubated for 2-3 h at 22°C. After washing with Tris-buffered saline, bound ICAM-1-Fc was detected using protein A conjugated to horseradish peroxidase (HRP), and the HRP substrate 3,3Ј,5,5Ј-tetramethylbenzidine (KPL, Gaithersburg, MD). Alternatively, the peptides were biotinylated via their N-terminal Cys using the EZ-link PEO-Maleimide activated biotin kit, or their N-terminal Lys using sulfosuccinimidyl 6-(biotinamido) hexanoate (Pierce), based on the product instructions. The plate was coated with ICAM-1-Fc (50 g/ml) and post-coated with BSA. Biotinylated peptides were added and incubated with ICAM-1-Fc for 2 h at room temperature. The bound peptides were detected using avidin conjugated to HRP.
To detect real time interactions between ICAM-1 and the ␤ 2 I domain peptides, a BIAcore 3000 SPR-based biosensor (BIAcore AB, Uppsala, Sweden) was used, and the protocols recommended by the manufacturer were followed. Briefly, the ICAM-1-Fc was covalently coupled to the CM5 sensor chips via primary amino groups, which was preactivated with N-ethyl-NЈ-(3-dimethylaminopropyl)carbodiimide hydrochloride/N-hydroxysuccinimide. This protocol resulted in ϳ1000 resonance units of immobilized ICAM-1. A blank flow cell that had been N-ethyl-NЈ-(3-dimethylaminopropyl)carbodiimide hydrochloride/N-hydroxysuccinimide-activated and ethanolamine-blocked was used as a reference surface. Binding data were obtained by injecting various concentrations of the peptides over the ICAM-1-Fc surface at a flow rate of 5 l/min. The experiments were conducted in a Tris buffer (20 mM Tris, 150 mM NaCl, pH 7.4) containing 1 mM MnCl 2 . The chip surface was regenerated with 0.1 M HCl. All of the data were analyzed using the BIAevaluation 3.0 program, subtracting binding to the blank flow cell to account for any nonspecific binding. To analyze steady state peptide binding, the peak response levels were plotted against the peptide concentrations, which were then fit to a single-site model (Langmuir isotherm) to determine the dissociation constants (K d ). The specificity of peptide binding was further verified using the corresponding scrambled controls.

RESULTS
Cell Adhesion to ICAM-1 by ␣ M ␤ 2 Is Activation-dependent-Ligand binding by the ␤ 2 integrins depends on receptor activation (2, 4), especially for ICAM-1 binding by ␣ M ␤ 2 , which is of low affinity (10). Indeed, when expressed on 293 cells, ␣ M ␤ 2 interacted with C3bi constitutively in the absence of receptor activation (28,29), whereas it failed to do so with ICAM-1, unless activated by the ␤ 2 -specific activating mAb MEM48 ( Fig. 1), which recognizes the mid-region of ␤ 2 (29). Addition of 10 g/ml of mAb MEM48 enhanced cell adhesion by 4 -8-fold, which could be blocked completely by the addition of an ␣ Mspecific antagonist (NIF) or EDTA, confirming the specificity and cation dependence of the adhesion assay. Moreover, mock transfected cells did not exhibit significant adhesion to ICAM-1. These results demonstrated the feasibility of using mAb MEM48 as an activator of ␣ M ␤ 2 for ICAM-1 binding. Therefore, all of the subsequent cell adhesion assays to ICAM-1 were conducted in the presence of MEM48.
A Unique Region on the Apex of the ␤ 2 I Domain Is Critical to ICAM-1 Recognition by ␣ M ␤ 2 -Previously, we have systematically substituted 17 segments that reside on the hydrated surface of the ␤ 2 I domain with their counterpart sequences of the ␤ 1 I domain and established stable cell lines for these 17 homolog-scanning ␣ M ␤ 2 mutants in 293 cells. We demonstrated that all 17 ␤ 2 I domain mutants, when transfected together with wild type ␣ M , were expressed at wild type levels on the cell surface. In addition, we have shown that these 17 ␤ 2 I domain mutants possessed a normal intact conformation, based on a number of different criteria, including surface labeling and immunoprecipitation, as well as reactivity toward a panel of conformation-dependant antibodies (24). Most importantly, we have studied these mutants using two conformation-and cationdependent mAbs and showed that all 17 mutants exhibited similar cation-dependent conformations (23,24). Thus, loss of function by these homolog-scanning mutants is most likely attributed to perturbations of the ligand-binding site by these mutations.
Goodman and Bajt (20) reported that a mutation within the ␤ 2 I domain had differential effects on ICAM-1 versus C3bi binding to ␣ M ␤ 2 , suggesting that the ␤ 2 I domain may possess different binding sites for ICAM-1 and C3bi. To locate critical sequences within the ␤ 2 I domain for ICAM-1 binding, we conducted cell adhesion assays on all 17 ␣ M ␤ 2 mutants using ICAM-1 as the substrate, based on the conditions established in Fig Fig. 3 shows the location of these critical residues within the ␤ 2 I domain, which was constructed earlier (23) based on its homology to ␤ 3 (30). Of the identified six sequences, segments Pro 192 -Glu 197 , Asn 213 -Glu 220 , Leu 225 -Leu 230 , Ser 324 -Thr 329 , and Glu 344 -Asp 348 reside proxi-mally in space and form a common ligand-binding site that is distinct from the C3bi-binding site we reported earlier (23). Therefore, we concluded that these five segments likely contributed directly to ICAM-1 binding. Segment Glu 162 -Gly 164 resided at the bottom of the ␤ 2 I domain that is far away from the identified ligand-binding site and thus less likely to contribute directly to ICAM-1 binding by ␣ M ␤ 2 .
A major ligand-binding site is located within the ␣ M I domain, which exists in an equilibrium between the open and closed conformations (31,32). Because activation of the ␣I domain depends critically on the ␤I domain (33) both adhered constitutively to ICAM-1 (ϳ7.4-and ϳ7.7-fold higher than that of wild type, respectively) in the absence of Mn 2ϩ , whereas wild type ␣ M ␤ 2expressing or mock transfected cells did not show significant adhesion, confirming that both ␣ M (I316G)␤ 2 and ␣ M (Cys 297 -Cys 304 )␤ 2 are constitutively active for ICAM-1 binding. As expected, when activated by Mn 2ϩ , the wild type receptor supported strong cell adhesion to ICAM-1 (Fig. 2C). Of interest, the ␣ M (Cys 297 -Cys 304 )␤ 2 receptor exhibited differential binding activities, where it supported constitutive adhesion to ICAM-1 ( Fig.  2D) but not to C3bi (26), which is consistent with the different conformational requirements for ␣ M ␤ 2 recognition of these two ligands (28,38). Most importantly, mutations within segment Leu 225 -Leu 230 , a component of the identified ICAM-1-binding site within the ␤ 2 I domain, abolished ligand binding by the constitutively active ␣ M (I316G)␤ 2 receptor (Fig. 2B) and the constitutively active ␣ M (Cys 297 -Cys 304 )␤ 2 receptor (Fig. 2D), regardless of Mn 2ϩ , suggesting that the ␤ 2 I domain mutations destroyed ligand binding independent of the conformation of the ␣ M I domain.
Synthetic Peptides of the ␤ 2 I Domain Blocked Cell Adhesion to ICAM-1 by ␣ M ␤ 2 -To see whether the residues within the identified ICAM-1-binding site could interact directly with ICAM-1, we prepared five synthetic peptides based on our model of the ␤ 2 I domain (Fig. 3) and the sequences that were mutated in the five defective ␤ 2 I domain mutants, including these soluble peptides to block wild type ␣ M ␤ 2 -mediated cell adhesion to ICAM-1 was tested, and the results are shown in Fig.  4. Among these five peptides, P7 and P12 showed the strongest inhibition, and addition of these two peptides reduced the number of adherent cells by 17-and 24-fold, respectively. Peptide P14 exhibited small but significant inhibition (3-fold), whereas peptides P5 and P17 had no significant effect on cell adhesion. To verify the specificity of peptide inhibition, we synthesized two scrambled control peptides sP7 (KQNLNNFVETTGSQ) and sP12 (CKTRTLVE-VESTWKY), corresponding to the two most active peptides P7 and P12, respectively. Both control peptides exhibited some, but significantly weaker, inhibition (ϳ1.5-fold) than their corresponding peptides (17-fold for P7 and 24-fold for P12). Furthermore, the addition of NIF blocked more than 95% cell adhesion, and no adhesion was observed for mock transfected cells, confirming the specificity of the adhesion assay. These data suggest that the ␤ 2 I domain peptides were effective soluble inhibitors of the ␣ M ␤ 2 -ICAM-1 interaction, most likely by binding directly to the ligand ICAM-1.
Direct Interactions between the ␤ 2 I Domain Peptides and ICAM-1-To see whether peptides P7, P12, and P14 inhibited cell adhesion by interacting with ICAM-1, we conducted direct binding assays using two different formats. First, we coated the synthetic peptides onto 96-well microtiter plates, and binding of soluble ICAM-1-Fc to the peptides was assessed. To make sure that these peptides had similar coating efficiencies, we measured the amount of immobilized peptides by labeling the free SH group (for peptides P12 and P14) or the NH 2 group (for peptides P5, P7, and P17) present on these peptides with biotin and then quantified the amount of immobilized biotin using an HRP conjugate of avidin. We found that similar amounts of peptides were present in each well (data not shown). The binding data are shown in Fig. 5A. Parallel to their inhibitory activity in the above cell adhesion assays, peptides P7 and P12 bound ICAM-1 effectively, which were significantly higher than their corresponding scrambled control peptides. The other three peptides, P5, P14, and P17, also exhibited significant binding to ICAM-1. As a negative control, we did not detect significant ICAM-1 binding to BSA-coated wells. In a reverse format, we coated 96-well microtiter plates with ICAM-1-Fc and then blocked with BSA. The biotinylated peptides were added, and bound peptides were detected using an HRP conjugate of avidin. In a separate experiment, we have verified that all five peptides had similar biotinylation efficiencies (data not shown). As shown in Fig. 5B, peptide P7 exhibited specific binding to the immobilized ICAM-1 that was significantly higher than its corresponding scrambled control. Peptide P12 did not exhibit specific binding in this format, and the other three peptides, P5, P14, and P17, exhibited modest binding. As negative controls, we found that none of these peptides bound specifically to BSA-coated surfaces. To study peptide binding to ICAM-1 in real time, we conducted additional experiments using surface plasmon resonance technology (BIAcore), where ICAM-1 was covalently coupled to the hydrophilic surface of the CM5 sensor chip, and peptide P5 or its scrambled control was added over the ICAM-1 surface at a flow rate of 5 l/min. As shown in Fig. 5C, P5 bound to ICAM-1 dose-dependently, and the calculated K d is 14.1 M. In comparison, sP5 bound ICAM-1 weakly with a K d of 5.7 mM (data not shown). Furthermore, analysis of steady state binding (i.e. the peak response level) achieved at different concentrations of peptide P5 supported a single-binding site model with a K d of 34 M (Fig. 5D), whereas no K d could be obtained for sP5. Collectively, these results demonstrated that the peptides derived from the identified ICAM-1-binding site interacted directly with ICAM-1 and therefore could provide a direct contact interface for ICAM-1 recognition.
A Dual Function for the F-␣ 7 Loop of the ␤ 2 I Domain in Ligand Binding-One of the functional segments within the identified ICAM-1-binding site is Glu 344 -Asp 348 , which resides in the F-␣ 7 loop (residues 344 -358) of the ␤ 2 I domain. The F-␣ 7 loop undergoes large movement upon ligand engagement to the ␤I domain and thus was proposed to play a key role in integrin activation (30). Specifically, it was hypothesized that residues within the F-␣ 7 loop interacted with residues within the extended loop of the ␤TD domain in an inactive receptor (the deadbolt model) (5), and disruption of this interaction leads to receptor activation. Hence, we speculated that mutations of this loop would directly break such interaction and thereby result in conformational changes of the ␤ 2 I domain and receptor activation. To test this hypothesis, we first evaluated the conformational states of mutant ␣ M ␤ 2 (Glu 344 -Asp 348 ), in which the sequence of the F-␣ 7 loop has been altered, using three different mAbs CBRM1/5, mAb 24, and mAb MEM148, which recognize activation-dependent neo-epitopes within the ␣ M I domain (34), the ␤ 2 I domain (35), and the ␤ 2 hybrid domain (36), respectively. Representative FACS analyses with mAb 24 were shown in Fig. 6A. Compared with the wild type receptor, the mutant bound mAb 24 much better, approaching its maximal binding activity, which was obtained in the presence of Mn 2ϩ . Verifying specificity, no mAb 24 binding was observed for either wild type or mutant in the presence of EDTA. To quantify the extent of receptor activation, we determined the mean fluorescence values for these three mAbs and then normalized surface expression based on mean fluorescence of two activation-independent mAbs, the ␣ M -specific OKM1 (for mAb CBRM1/5) and the ␤ 2 -specific 6.7 (for mAb 24 and mAb MEM148). As shown in Table I, little binding (for mAb 24 and MEM148), and weak binding (for mAb CBRM1/5) were observed for wild type ␣ M ␤ 2 . However, when the sequence within the F-␣ 7 loop was mutated, significant increases (2.5-, 2.9-, and 2.8-fold, respectively) in the mean fluorescence intensity was observed for mAbs CBRM1/5, 24, and MEM148, indicating that conformational changes within the ␤ 2 I domain had been propagated into at least two other regions within ␣ M ␤ 2 (the ␣ M I domain and the ␤ 2 hybrid domain). Thus, mutations of the F-␣7 loop disrupted potential interactions between the F-␣7 loop and the ␤TD domain, resulting in global conformational changes and receptor activation. In support of this notion, the ␣ M ␤ 2 (Glu 344 -Asp 348 ) mutant exhibited higher adhesive activity toward another ␣ M ␤ 2 ligand (the Fg ␥-module) (Fig. 6B). In this assay, a 293 cell clone that expresses wild type ␣ M ␤ 2 at a level equivalent to that of ␣ M ␤ 2 (Glu 344 -Asp 348 ) was used in parallel adhesion assays. As shown in Fig. 6B, cells expressing the mutant receptor exhibited stronger adhesion toward the ␥-module across a wide range of concentrations (5-40 g/ml), indicating that this mutant receptor is more active than wild type ␣ M ␤ 2 . The specificity of this assay was confirmed by the  (30). Molecular modeling was carried out using InsightII modules: Biopolymer, Homology, and Discover (Accelrys Inc.) (23). The backbone of the ␤ 2 I domain is shown with the ␤-sheets in light blue and ␣-helices in red, with the bound Ca 2ϩ cations as gray spheres. The Connolly surface of the ligand recognition sites generated using a probe size of 1.4 Å are highlighted in blue (residues 192-197), green (residues 213-220), dark orange (residues 225-230), pale green (residues 324 -329), and dark green (residues 344 -348). The F-␣ 7 loop is located within residues 344 -358 and contains part of the ligand-binding site (residues 344 -348 shown in dark green). This figure was prepared using the program MOLMOL (41).
FIG. 4. Inhibition of ␣ M ␤ 2 ligand binding by synthetic peptides derived from the ␤ 2 I domain. Inhibition of ␣ M ␤ 2 -mediated cell adhesion to ICAM-1 by synthetic peptides derived from the functional sequences identified within the ␤ 2 I domain was conducted similarly as in Fig. 1, except that 2 mg/ml peptides were added in the assay mixture. The number of the adherent cells in the absence of the peptides was taken as 1.0. Verifying specificity of the assay, cell adhesion to ICAM-1 could be blocked by the addition of NIF, and the scrambled peptides sP7 and sP12 did not significantly inhibit ICAM-1 binding by ␣ M ␤ 2 . The data shown are the means Ϯ S.D. of four to eight independent experiments. ability of NIF to completely abrogate cell adhesion (data not shown). DISCUSSION A well known feature of the integrin receptor is its ability to change conformation in response to intracellular activation signals (inside-out signaling). Although the detailed molecular mechanism underlying integrin inside-out signaling still remains elusive, several models have been recently proposed (3)(4)(5). A key element that is shared among these models is the assumption that the F-␣ 7 loop within the ␤I domain (residues 344 -358 of ␤ 2 ) is critical to the conformational transition from the closed to the open state of the integrin receptor. Yet, no study to date has directly mutated this loop and studied its impact on ligand binding. Therefore, this work provides the first experimental support for a critical role of this F-␣ 7 loop in controlling the open and closed conformations of the ␤ 2 I domain as well as the other two regions within the ␣ M ␤ 2 receptor. In addition, we found that the F-␣ 7 loop is also part of a novel binding interface for ICAM-1 within the ␤ 2 I domain. Thus, we conclude that the F-␣ 7 loop has dual functions in ligand binding by ␣ M ␤ 2 .
Previously, we reported our establishment and characterization of 17 ␣ M ␤ 2 homolog-scanning mutants that contain indi-vidual segment switches between the homologous ␤ 1 and ␤ 2 I domains, based on the premise that the ␤ 1 and ␤ 2 integrins are very similar in protein sequence (74% identify for the ␤I domains) yet recognize completely different sets of ligands. We showed that the homolog-scanning mutants, including those that were defective in C3bi and Fg binding, possessed intact conformations, judged by their reactivity toward a panel of conformation-dependent mAbs and by their ability to respond normally to Ca 2ϩ (24). Therefore, the loss of ligand binding function was most likely caused by direct perturbations of the ligand recognition site per se rather than indirectly by conformational changes within the mutant receptors. In this study, we screened these 17 ␣ M ␤ 2 mutants for ICAM-1 binding using our established adhesion method (28) with an additional activation step by a ␤ 2 -activating mAb (MEM48). Consistent with reports in the literature (20), we found that receptor activation was required for ICAM-1 binding by ␣ M ␤ 2 (Fig. 1). Based on the adhesion assays, we found five noncontiguous segments (Pro 192 -Glu 197 , Asn 213 -Glu 220 , Leu 225 -Leu 230 , Ser 324 -Thr 329 , and Glu 344 -Asp 348 ) on the apex of the ␤ 2 I domain that were critical to ICAM-1 recognition (Fig. 2). That these segments are involved directly in ligand binding was supported by the obser- A and B, enzyme-linked immunosorbent assay. 96-well microtiter plates were coated with 50 l of synthetic peptides (2 mg/ml) overnight (A) and then blocked with 3% BSA. ICAM-1-Fc (50 g/ml) was added to the wells and incubated for 2 h at 22°C. After washing, bound ICAM-1-Fc was detected with an HRP conjugate of protein A. The amount of bound ICAM-1 was measured by reaction with the HRP substrate 3,3Ј,5,5Ј-tetramethylbenzidine. Reciprocally, the microtiter plate was coated with ICAM-1-Fc (50 g/ml) (B) and then blocked with BSA. Biotinylated peptides were added, and bound peptides were determined using an HRP conjugate of avidin. C, overlay of sensorgrams of peptide P5 binding to the immobilized ICAM-1 measured by BIAcore. Peptide P5 or its scrambled control in different concentrations (0 -50 M) was injected into the ICAM-1 surface at a flow rate of 5 l/min. Specificity of the peptides-ICAM-1 interaction was verified by the weak binding observed for its scrambled peptide (sP5) (data not shown). D, analysis of peptide P5 binding to ICAM-1 under steady state conditions. The peak response levels achieved in the steady state region of each sensorgram were plotted against P5 concentration. The binding curve could be fit to a single-binding site model with a K d of 34 M. vation that four of the five defective mutants in ICAM-1 binding interacted with Fg and C3bi as well as the wild type receptor, and these five mutants all exhibited correct Ca 2ϩdepended conformations (23). In addition, using constitutively active ␣ M ␤ 2 , we demonstrated that inhibition of ligand binding by the ␤ 2 I domain mutations was independent of the conformational states of the ␣ M I domain, because mutating a functional segment (Leu 225 -Leu 230 ) abolished ligand binding by constitutively active ␣ M ␤ 2 receptors, which were generated by changing the critical Ile 316 residue or by creating a disulfide linkage between residues 297 and 304 within the ␣ M I domain (25) and by the addition of Mn 2ϩ (31) (Fig. 2). Furthermore, we showed that synthetic peptides derived from these sequences, but not scrambled control peptides, were able to inhibit the interaction between wild type ␣ M ␤ 2 and ICAM-1 when present in solution (Fig. 4) and to directly bind soluble ICAM-1 when coated on microtiter plates (Fig. 5A). Similar interactions between ICAM-1 and these ␤ 2 I domain-derived peptides were observed using a reverse format where ICAM-1 was immobilized on the plates, and the peptides were added in solution, providing further support that these sequences represent a direct ICAM-1-binding interface within the ␤ 2 I domain (Fig.  5B). Interestingly, several peptides supported ICAM-1 binding when immobilized on the microtiter plate but did not interact with ICAM-1 well in solution, suggesting that these peptides may adopt different conformations depending on the environment. Similar phenomena have been well documented in the literature for large proteins such as fibrinogen, vitronectin, and von Willebrand factor (37).
Although the newly identified ICAM-1-binding site, like those for Fg and C3bi, resides on the apex of the ␤ 2 I domain, two important differences separate ICAM-1 binding from C3bi and Fg binding to ␣ M ␤ 2 . First, we and others have shown that ␣ M ␤ 2 -expressing 293 cells interact with C3bi constitutively (28,38). In contrast, activation of ␣ M ␤ 2 by a ␤ 2 -specific activating mAb (MEM48) is required for ␣ M ␤ 2 -mediated cell adhesion to ICAM-1 (Fig. 1), suggesting that ICAM-1 recognizes a different conformation of ␣ M ␤ 2 than C3bi does. Second, mutations of the ␤ 2 I domain had differential effects on ␣ M ␤ 2 binding to ICAM-1 and C3bi. Among the 17 mutations within the ␤ 2 I domain, six of them, including Glu 162 -Gly 164 , Pro 192 -Glu 197 , Asn 213 -Glu 220 , Leu 225 -Leu 230 , Ser 324 -Thr 320 , and Glu 344 -Asp 348 , destroyed significant ICAM-1 binding by ␣ M ␤ 2 (Fig. 2). Yet most of these ␤ 2 I domain mutants, including ␣ M ␤ 2 (Glu 162 -Gly 164 ), ␣ M ␤ 2 (Pro 192 -Glu 197 ), ␣ M ␤ 2 (Asn 213 -Glu 220 ), ␣ M ␤ 2 (Ser 324 -Thr 320 ), and ␣ M ␤ 2 (Glu 344 -Asp 348 ), had normal or even better binding activity (1.5-5-fold) than the wild type receptor for C3bi and Fg (23). Only one mutation, Leu 225 -Leu 230 , destroyed ␣ M ␤ 2 binding activity for all three ligands (ICAM-1, C3bi, and Fg). Thus, the ICAM-1-binding site is distinct from those of C3bi and Fg. The only overlapping region between the ICAM-1 and the C3bi-and Fg-binding sites is the region surrounding sequence Leu 225 -Leu 230 , which is involved in the formation of the metal ion-dependent adhesion site motif. Consistent with our data, Goodman and Bajt (20) reported that the Glu 235 residue, which is located between segments Asn 213 -Glu 220 and Ser 324 -Thr 320 in space, is critical to ICAM-1 binding. Moreover, our mutagenesis results are consistent with the strong inhibitory activity of mAbs IB4 and MHM23 on ICAM-1 binding to ␣ M ␤ 2 (39). Both mAbs recognize segment Pro 192 -Glu 197 of the ␤ 2 I domain (24). The critical sequence Glu 162 -Gly 164 was not believed to be part of the binding site, because it resided at the bottom of the ␤ 2 I domain, far from the other five segments. The mechanism by which it abolishes ICAM-1 binding is unclear. However, considering its proximity to the ␤ 2 hybrid domain (30), it is likely that perturbation of this region may alter the relative orientation between the ␤ 2 I domain and the hybrid domain, thus affecting receptor activation by mAb MEM48. More studies will be needed to test this hypothesis.
Another important finding from this work is that sequence  Glu 344 -Asp 348 , located within the F-␣ 7 loop of the ␤ 2 I domain, exhibited dual activities in ligand binding; not only was it capable of controlling the open and closed conformations of the receptor, it also contributed directly to the formation of the ICAM-1-binding site. Our finding is consistent with the deadbolt model for integrin activation recently proposed by Xiong et al. (5). In this model, the authors proposed that the residues within the F-␣ 7 loop of the ␤I domain form a close contact with the residues in the elongated CD loop of the ␤TD domain, which then acts as a deadbolt to prevent the flip of the F-␣ 7 loop that is required for conversion of closed to open conformation of the ␤ 2 I domain. Interruption of this interaction releases the deadbolt from the ␤I domain, leading to conformational change and receptor activation. Our observation that mutation of residues within the F-␣ 7 loop activated ␣ M ␤ 2 is consistent with this hypothesis. Furthermore, our data demonstrated that conformational changes caused by perturbation of the F-␣ 7 loop could be transmitted throughout the heterodimeric receptor, causing multiple structural rearrangements, including the ligand-binding pocket within the ␣ M I domain that is recognized by mAb CBRM1/5 (34) and the hybrid domain within the ␤ 2 subunit that is recognized by mAb MEM148 (36). Thus, the conformations of different regions within the receptor are intimately connected. One of the potential mechanisms that could link together the ␣ M I and ␤ 2 I domains is likely mediated by interaction between Glu 320 of ␣ M , which residues in between the two I domains, and the metal ion-dependent adhesion site motif of the ␤ 2 I domain (33). Thus, mutation of Glu 320 could potentially disrupt the flow of conformational changes from ␤ 2 to ␣ M , as well as alter the relative orientation between these two ligand-binding sites, both resulting in defective ligand binding. In further support of our model, it was reported that the ␣ M I and ␤ 2 I domains exist in both open and closed conformations, and integrin activation may require activations of the individual I domains, as well as changes of the secondary and tertiary structures among these ligand binding domains (32,40). In summary, the results from this study strongly support our original model that the ␤ 2 subunit plays a direct role in ligand binding by ␣ M ␤ 2 (23). Specifically, our data demonstrated that there exist several discrete subdomains on the upper surface of the ␤ 2 I domain that allow it to interact with multiple unrelated ligands. Surprisingly, the ICAM-1 binding site is much broader than those for C3bi and Fg and is composed of five noncontiguous segments Pro 192 -Glu 197 , Asn 213 -Glu 220 , Leu 225 -Leu 230 , Ser 324 -Thr 320 and Glu 344 -Asp 348 . The ICAM-1 recognition site is well separated from the Fg-and C3bi-binding sites (23), although they did share the common metal ion-dependent adhesion site motif that is located within segment Leu 225 -Leu 230 . The three potential cation-binding sites within the ␤ 2 I domain are located inside this identified region, suggesting that these three sites may also contribute directly to ICAM-1 binding by ␣ M ␤ 2 . Most importantly, as the identified ligand-binding site contains a segment that is involved in modulating the open and closed conformations of the receptor in response to integrin inside-out signaling, our study raises an interesting possibility that inside-out signaling could directly modify the shape of the ligand-binding site, in addition to its ability to change the global conformation of the receptor. Given the similarity among all integrin ␤ subunits, our results may help us understand the underlying mechanism of integrin-ligand interactions in general.