Distinct Roles for the α and β Subunits in the Functions of Integrin αMβ2*

Integrin αMβ2 (Mac-1, CD11b/CD18) is a noncovalently linked heterodimer of αM and β2 subunits on the surface of leukocytes, where it plays a pivotal role in the adhesion and migration of these cells. Using HEK293 cells expressing αMβ2 or the individual constituent chains on their surface, we analyzed the contributions of the αM or β2 subunits to functional responses mediated by the integrin. In cells expressing only αM or β2, the individual subunits were not associated with the endogenous integrins of the cells, and other partners for the subunits were not detected by surface labeling and immunoprecipitation under a variety of conditions. The αM cells mediated adhesion and spreading on a series of αMβ2 ligands (fibrinogen, Factor X, iC3b, ICAM-1 (intercellular adhesion molecule-1), and denatured ovalbumin) but could not support cell migration to any of these. The spreading of the αM cells suggested an unanticipated linkage of this subunit to the cytoskeleton. The β2 cells supported migration and attachment but not spreading on a subset of the αMβ2 ligands. The heterodimeric receptor and its individual subunits were purified from the cells by affinity chromatography and recapitulated the ligand binding properties of the corresponding cell lines. These data indicate that each subunit of αMβ2 contributes distinct properties to αMβ2 and that, in most but not all cases, the response of the integrin is a composite of the functions of its individual subunits.

Integrins are a large family of heterodimeric cell adhesion receptors that mediate a wide spectrum of biological functions (reviewed in Ref. 1). The ␤ 2 subfamily, often referred to as the leukocyte integrins, is composed of four members that share a common ␤ 2 subunit that associates noncovalently with one of four distinct but structurally homologous ␣ subunits to form integrins ␣ M ␤ 2 (Mac-1, CD11b/CD18, CR3), ␣ L ␤ 2 (lymphocyte function-associated antigen-1, CD11a/CD18), p150/95 (␣ X ␤ 2 , CD11c/CD18), and ␣ D ␤ 2 (CD11d/CD18) (reviewed in Refs. 2 and 3). ␣ M ␤ 2 is expressed in monocytes, granulocytes, macrophages, and natural killer cells and has been implicated in diverse responses of these cells, including phagocytosis, cellmediated killing, chemotaxis, and cellular activation. These complex responses depend upon the capacity of ␣ M ␤ 2 to medi-ate leukocyte adhesion and migration; and consequently, ␣ M ␤ 2 plays a central role in inflammation. The characterization of ␣ M ␤ 2 -deficient mice has confirmed these findings by showing that a variety of leukocyte-dependent responses are compromised in these animals (e.g. Refs. 4 and 5).
Two segments, one within each of its constituent chains, have been implicated in the binding of protein ligands to ␣ M ␤ 2 : 1) the ␣ M I (or A) domain, an ϳ200-amino acid segment in the N-terminal third of the ␣ M subunit that is structurally very similar to the I domains within the ␣ subunits of the other leukocyte integrins (23,24), and 2) the ␤ 2 I-like domain, which is structurally similar to the ␣ subunit I domains as well as the I-like domains found in all ␤ subunits of the integrins (25)(26)(27). Each of these segments contains a metal ion-dependent adhesion site (MIDAS), which is critical to the ligand binding functions of ␣ M ␤ 2 and other integrins (23, 28 -30). One manifestation of the contribution of the MIDAS to ligand binding is that mutagenesis of the predicted cation coordination residues of the ␤ 2 or ␣ M MIDAS destroys the capacity of ␣ M ␤ 2 to bind multiple ligands (e.g. Refs. [31][32][33][34]. A second manifestation of the MIDAS contribution is that different divalent cations exert differential effects on ligand binding. Mn 2ϩ enhances ligand binding function, Mg 2ϩ supports it, and Ca 2ϩ can be suppressive (35)(36)(37). The crystal structures of the ␣ M I domain or other I domains show that different cations differentially influence structure (29). Direct evidence for the involvement of the ␣ M subunit in ligand binding has derived from multiple studies in which its I domain has been expressed and shown to engage ligand (11,38,39). Evidence for the involvement of the ␤ subunit is less direct. Certain mutations in the ␤ 2 subunit abolish the binding of multiple ligands (31,33), and certain ligands (iC3b and FX) are still capable of binding to recombinant ␣ M ␤ 2 lacking an ␣ M I domain (40). Rather than a direct involvement of the ␤ subunit, the ␤ 2 I domain may regulate the function of the ␣ M subunit. Consistent with such a regulatory role, epitopes of several monoclonal antibodies (mAbs) that activate the ␤ 2 integrins, KIM185, KIM127, MEM48, and CBR lymphocyte function-associated antigen-1/2, reside in or close to the ␤ 2 I domain (41)(42)(43), and cells expressing only the ␣ M subunit in the absence of ␤ 2 appear to bind ligands (such as Fg) that do not bind to ␣ M ␤ 2 in which residues in the ␤ 2 MIDAS motif have been mutated (22). Thus, the contribution of the ␤ 2 subunit to ligand recognition by ␣ M ␤ 2 remains uncertain.
In this study, we took advantage of the capacity of human epithelial kidney 293 (HEK293) cells to express the individual ␣ M or ␤ 2 subunits or heterodimeric ␣ M ␤ 2 on their surface (22,44). The adhesive and migratory properties of the cells were analyzed, and the cells also were used as a source to purify and characterize the individual subunits. These comparisons allowed us demonstrate that both subunits can contribute directly to ligand binding and to assign specific functions to each subunit of the receptor.
Development of ␣ M ␤ 2 , ␣ L ␤ 2 , ␣ M , and ␤ 2 Cell Lines-HEK293 cells were stably transfected using Lipofectamine Plus reagent (Invitrogen) with 0.5-5 g of pcDNA3.1 (Invitrogen) containing the full-length cDNAs for ␣ M and/or ␤ 2 or vector alone (mock-transfected) as a control as described previously (33,45). Transfected cells were selected using neomycin sulfate (Invitrogen), and cells expressing the receptors were detected and sorted by flow cytometry (FACS) using a FACStar instrument (BD Biosciences) and mAb 904, IB4 or TS1/18. The cell lines obtained were maintained in Dulbecco's modified Eagle's medium/nutrient mixture F-12 supplemented with 10% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, 2 M L-glutamine, and 2 mg/ml neomycin (all from Invitrogen) (33). For selected experiments, a HEK cell line expressing both ␣ M ␤ 2 and the urokinase-type plasminogen activator receptor was used; these cells have been described previously (46).
To verify that the specificity-determining loop was present and that ␤ 2 I domain was not mutated in the ␤ 2 cells, total RNA was isolated from 1.5 ϫ 10 5 ␤ 2 cells, and cDNA was prepared using the QIAamp DNA mini kit (Qiagen Inc., Santa Clara, CA) according to the manufacturer's protocol. The cDNA obtained was used as template for PCR performed with 5Ј-GATCCTGACTCCATTCGCTGCGACACCCGGCC-3Ј (␤ 2 238) as the 5Ј-primer and 5Ј-TCCATTGCTGCAGAAGGAGTCGTAGG-3Ј (␤ 2 1239) as the 3Ј-primer. After amplification, the PCR product was isolated by electrophoresis on 1% agarose gels, and only a band of ϳ1 kb was detected. This band was extracted from the gels using the QIAquick gel extraction Kit (Qiagen Inc.) according the manufacturer's protocol, and the sample was concentrated by ethanol precipitation and sequenced using ␤ 2 238 or 5Ј-GACCAGGCCAGGCAGCAGCGTTCAACG-TGACC-3Ј (␤ 2 395) as the sequencing primer.
FACS-Transfected HEK293 cells were harvested with cell dissociation buffer (Invitrogen); washed twice; suspended in staining medium consisting of Hanks' balanced salt solution (HBSS) containing 5 mM CaCl 2 , 5 mM MgCl 2 , 10 mM HEPES (pH 7.4), and 0.1% goat normal serum; and incubated at 4°C for 30 min with the selected primary mAb at 10 g/ml. In some experiments, the divalent cations were replaced with 1 mM EDTA. After incubation, the cells were washed twice by centrifugation and resuspended in 30 g/ml fluorescein-conjugated goat anti-mouse IgG (Zymed Laboratories Inc.) for 30 min at 4°C in the dark. The cells were washed three times with HBSS/HEPES, and cell-bound antibodies were detected using a FACScan. Data were analyzed using the LYSIS program (BD Biosciences). The mean fluorescence intensity of the ␣ M ␤ 2 , ␣ L ␤ 2 , ␣ M , and ␤ 2 cells used was 150 -400 units when staining with an appropriate mAb (see "Results") compared with ϳ5 units for mock-transfected cells and Ͻ10 units for a nonreactive mAb.
Immunoprecipitation of Surface-expressed Proteins-Transfected HEK293 cells were harvested with cell dissociation buffer as described above, washed twice with Dulbecco's phosphate-buffered saline (PBS), and resuspended to 10 8 cells/ml in Dulbecco's PBS. Surface-expressed proteins were labeled with EZ-Link TM sulfosuccinimidyl 6-(biotinamido)hexanoate (Pierce) according to the manufacturer's protocol. After 30 min, non-reacted sulfo-NHS-LC-biotin was removed by washing three times with HBSS/HEPES; the cells were resuspended in HBSS/ HEPES to 10 8 cells/ml; and mAb 44a (for the ␣ M cells) or IB4 (for all other cells) was added. After incubation for 60 min at 4°C, non-bound antibody was removed by washing three times with HBSS/HEPES, and 10 9 cells were pelleted by centrifugation and solubilized at 4°C with lysis buffer composed of Tris-buffered saline (TBS) (pH 7.4) containing 5 mM CaCl 2 , 5 mM MgCl 2 , protease inhibitor mixture for mammalian cells (Sigma) and one of the following detergents: 5% Triton X-100 (Fisher), 5% Tween 20 (Fisher), 1% CHAPS (Sigma), or 20 mM n-octyl ␤-D-glucopyranoside (Calbiochem). After mixing at 4°C for 20 min, the samples were clarified by centrifugation, and 15 l of protein G-agarose (Pierce) pretreated with non-transfected HEK293 cell lysate were added. After overnight incubation at 4°C, the beads were collected by centrifugation and washed four times with cold cell lysis buffer. After the last wash, 60 l of SDS electrophoresis endurance loading buffer (ICS BioExpress, Kaysville, UT) were added; samples were boiled for 2 min; and bound proteins were separated by electrophoresis in 4 -20% polyacrylamide gel plates (ISC BioExpress) using a Tris/SDS/HEPES buffer system. Separated proteins were transferred from the gel onto polyvinylidene difluoride membranes (Millipore Corp., Bedford, MA), and protein bands were developed using horseradish peroxidase-conjugated streptavidin (Pierce) and Opti-4CN-amplified substrate for horseradish peroxidase (Bio-Rad).
Receptor and Subunit Purification-␣ M ␤ 2 , ␣ M , and ␤ 2 were purified from transfected HEK293 cells using a modification of the method of Miller et al. (47). Briefly, 10 g of each cell line were harvested, washed, and lysed with 20 ml of 1% Triton X-100 in TBS containing 0.5 ml of protease inhibitor mixture for mammalian cells and 5 mM EDTA for 30 min at 4°C. The cell lysate was clarified by centrifugation, diluted 4-fold with TBS, and loaded onto a mAb column (1 ϫ 4 cm) at 4°C. For ␣ M purification, immobilized mAb LM2/1 (anti-␣ M ) was used; and for ␣ M ␤ 2 and ␤ 2 purification, immobilized mAb IB4 (anti-␤ 2 ) was employed. To prepare the immunoaffinity columns, purified mAbs were coupled to CNBr-activated Sepharose 4B (Amersham Biosciences) according to the manufacturer's protocol to final concentrations of 1.8 -2.2 mg of immobilized protein/1 ml of swollen gel. After washing with 100 ml of TBS containing 10 mM n-octyl ␤-D-glucopyranoside and 1 mM CaCl 2 , bound protein was eluted with 2 column volumes of 20 mM sodium acetate buffer (pH 4.2) containing 10 mM n-octyl ␤-D-glucopyranoside. Immediately after elution, the pH of the eluates was adjusted to pH 7.2 with 1 M Tris. In addition to SDS-PAGE (see "Results"), the purified ␤ 2 subunit from these cells was characterized by HPLC gel filtration chromatography on a 7.5 mm ϫ 60 cm UltroPac TSK G4000SW column (LKB Bromma, Uppsala, Sweden) in 0.15 M NaCl and 0.05 M Tris (pH 7.4).
Ligand Binding Assays-Na 125 I (specific activity of 15 mCi of 125 I/mg of iodine; Amersham Biosciences) was used to radioiodinate purified ␣ M ␤ 2 , ␣ M , and ␤ 2 by a modified chloramine-T procedure (48). Upon SDS-PAGE followed by autoradiography, radiolabeled ␣ M ␤ 2 showed only two bands with estimated molecular masses of ϳ150 and 90 kDa, and radiolabeled ␣ M and ␤ 2 showed only one band each with molecular masses of 150 and 90 kDa, respectively. The patterns and mobilities were indistinguishable from those of the unlabeled forms of the receptor and its subunits. The radiolabeled proteins were stored at Ϫ20°C and used within 1 month. Fg, FX, iC3b, ICAM-1, and denatured ovalbumin (dOva) were biotinylated using EZ-Link sulfosuccinimidyl 6-(biotinamido)hexanoate according to the manufacturer's protocol. Each biotinylated protein (2 nM) was mixed with 0.5 ml of Ultra-Link immobi-lized streptavidin (Pierce) and incubated for 1 h at 4°C. The nonreacted streptavidin was blocked with 1% biotin; and the beads were washed three times with 20 ml of TBS, stored at 4°C, and used within 1 week. To determine binding of the ligands to the radiolabeled receptor or its subunits, 20 l of the ligand beads were incubated with 10 g of radioiodinated ␣ M ␤ 2, ␣ M , or ␤ 2 for 30 min at 37°C in the presence of 2 mM MnCl 2 and then washed five times with TBS containing 2 mM MnCl 2 and 10 mM n-octyl ␤-D-glucopyranoside. The amount of radioactivity retained by the beads was measured using an Isotec ␥ counter (ICN Flow Titertec, ICN Biomedicals, Irvine, CA). Each point is the mean Ϯ S.E. of three independent experiments.
Cell AttachmentրAdhesion Assays-48-Well Costar tissue culture plates were coated with 200 l of different concentrations (0 -100 nM) of iC3b, Fg, FX, ICAM-1, or dOva overnight at 4°C and then post-coated with 0.5% polyvinylpyrrolidone (PVP) for 1 h at room temperature (33). Control wells were coated with PVP only. Prior to use, the plates were rinsed three times with PBS. Transfected HEK293 cells were harvested as described above, washed three times with 50 ml of divalent ion-free HBSS/HEPES (pH 7.4), and resuspended in divalent ion-free HBSS/ HEPES, and then the selected divalent cations were added. The cells were seeded at 1.5-2 ϫ 10 5 cells/well onto the assay plates and incubated at 37°C for 30 min. For inhibition experiments, the cells were pretreated with the selected mAbs or other reagents for 15 min at 37°C prior to addition to the coated wells.
To determine the extent of attachment/adhesion, the plates were washed three times with PBS, and the number of adherent cells in each well was quantified using the Cyquant cell proliferation assay kit (Molecular Probes, Inc., Eugene, OR) according to the manufacturer's instructions. Briefly, after washing, the plates were frozen at Ϫ70°C for 4 h and thawed in the presence of cell lysis buffer containing green fluorescent dye, which can be incorporated into DNA. After 30 min at room temperature in the dark, the fluorescence was measured using a CytoFluor II fluorescence multiwell plate reader (Molecular Devices, Inc., Sunnyvale, CA) using an excitation wavelength of 485 nm and an emission wavelength of 530 nm. The data from cell adhesion and migration assays (see below) are presented as mean fluorescence intensity Ϯ S.D. of three independent experiments. To distinguish attached and spread cells from those attached but not spread, after adhesion, the plates were rinsed three times with PBS, and the adherent cells were immediately photographed (magnification ϫ200). The spread cells were quantified as a percent of the attached cells, counting a total of 1000 cells.
Cell Migration Assays-Cell migration assays were performed in serum-free Dulbecco's modified Eagle's medium/nutrient mixture F-12 using Costar 24-transwell plates with tissue culture-treated 8-m pore polycarbonate filters (Corning Inc.) as described previously (20). The lower chambers contained 600 l of medium with the selected ligands, and the upper chambers contained final volumes of 200 l after addition of the cells. To begin the assay, 50 l of cell suspension (2 ϫ 10 5 cells/well) in medium were added to the upper chambers, and the plates were placed in a humidified incubator at 37°C and 5% CO 2 . Assays were stopped after 16 h by removing the upper wells and wiping the inside of the upper wells three times with a cotton swab to remove non-migrated cells. The migrated cells, present on the undersurface of the membrane as well as in the lower chambers, were quantitated using the Cyquant cell proliferation assay kit as described above.
Integrin Clustering and Immunofluorescence-Transfected cell lines were seeded onto CC2 TM -treated Lab-Tek® Chamber Slides TM (Nalge Nunc International, Naperville, IL) at 5 ϫ 10 4 cells/well in culture medium and incubated for 16 h at 37°C and 5% CO 2 . The cells were washed, and antibody against the integrin ␣ 4 or ␤ 1 subunit was added at 10 g/ml in HBSS (pH 7.3) containing 1 mM CaCl 2 , 1 mM MgCl 2 , and 1% bovine serum albumin and incubated for 30 min at 4°C. Cells were washed and incubated with Alexa 568-conjugated F(abЈ) 2 fragments of goat anti-mouse IgG (Molecular Probes, Inc.) at 10 g/ml for 30 min at 4°C. After washing, integrin cluster formation was allowed to occur at 37°C for 30 -45 min. Next, cells were fixed with 4% paraformaldehyde for 20 min at 22°C and stained with biotinylated mAb to the ␣ M subunit (44a) or the ␤ 2 subunit (IB4) for 30 min at 22°C, followed by incubation with Alexa 488-conjugated avidin (1:500 dilution; Molecular Probes, Inc.) for 30 min at 22°C. To analyze integrin localization and clustering on resting cells, the cells were first fixed and then stained to visualize the specific subunits as described above. The slides were mounted using Vectashield mounting medium containing 4Ј,6-diamidino-2-phenylindole (Vector Laboratories, Burlingame, CA) and observed under a fluorescence microscope (Leica Inc., Bannockburn, IL).

Cell-surface Expression and
Folding of the ␤ 2 Integrin Subunit-In a recent study, Takagi et al. (49) suggested that the expression and appropriate folding of the ␤ 2 subunit can occur only if its specificity-determining loop (SDL), a small segment within the ␤ 2 I-like domain, is deleted. Nevertheless, when the cDNA of the ␤ 2 subunit was transfected into HEK293 cells, clones exhibiting high expression of the ␤ 2 subunit were obtained. To verify that the SDL was present in the transfected cells, we isolated total RNA from the cells, amplified the entire region encoding the ␤ 2 I-like domain, and then sequenced the PCR product. The SDL was present, and no mutations were detected in the SDL as well as in the entire ␤ 2 I-like domain. The results from immunoprecipitation with mAb IB4 (anti-␤ 2 subunit) after surface labeling of one such cell line with biotin using n-octyl ␤-D-glycopyranoside for cell lysis are shown in Fig. 1A. Only one band with an estimated molecular mass of ϳ90 kDa was observed. The mobility of the band immunoprecipitated from the ␤ 2 cells was the same as that of the band immunoprecipitated from the ␣ M ␤ 2 cells, and its labeling with represents control immunoprecipitation from vector-alone cells using mAb IB4. The immunoprecipitates were analyzed by SDS-PAGE under reducing conditions, and transfers were probed with horseradish peroxidase-conjugated streptavidin. B, the ␣ M , ␤ 2 , ␣ M ␤ 2 , and mock-treated cells were stained with anti-␣ M mAb OKM1, anti-␤ 2 mAb IB4, and control normal mouse IgG (NMIgG) as described under "Experimental Procedures" and analyzed by FACS. Although the ␤ 2 cells exhibited substantial expression of the subunit, the level decreased over time, and periodic sorting of these cells was required. MFI, mean fluorescence intensity biotin is consistent with cell-surface expression. This single band was also observed using other detergents (Triton X-100, Tween 20, and CHAPS) as well as other concentrations of n-octyl ␤-D-glycopyranoside to solubilize the cells. No other protein band(s) co-immunoprecipitated with the ␤ 2 subunit in the molecular mass range of 3-400 kDa, even though the ␣ M subunit coprecipitated with the ␤ 2 subunit from the ␣ M ␤ 2 cells. As a positive control for other possible partners of the ␤ 2 subunit, HEK293 cells coexpressing the urokinase-type plasminogen activator receptor and ␣ M ␤ 2 (50) were used. After immunoprecipitation with mAb IB4, three protein bands (identified as ␣ M , ␤ 2 , and the urokinase-type plasminogen activator receptor) were detected in all of the detergents specified above.
The characterization of the ␤ 2 , ␣ M , and ␣ M ␤ 2 cells used in this study by FACS is summarized in Fig. 1B. Appropriate immunoreactivity patterns were observed. The ␤ 2 cells reacted with mAb to the ␤ 2 subunit (clone IB4; other clones are shown in Table I) but not with mAb to the ␣ M subunit, whereas the ␣ M cells reacted with mAb to the ␣ M subunit (clone 44a) but not with mAb to the ␤ 2 subunit. The ␣ M ␤ 2 cells reacted with both sets of mAbs. As noted by comparing the mean fluorescence intensity values of the ␤ 2 or ␣ M cells with those of the ␣ M ␤ 2 cells lines, cell lines could be obtained that expressed high and similar levels of the individual subunits compared with those found in the ␣ M ␤ 2 cells. Cell lines that were matched (Ͻ2-fold differences) in subunit expression levels based upon FACS were used in subsequent functional studies.
As we were unable to identify interaction of the individual subunits with other membrane proteins on the cells, we also considered whether they might associate weakly with the endogenous integrins of the cells. Integrin ␣ 4 ␤ 1 is abundantly expressed in HEK293 cells, and we clustered its subunits with monoclonal antibodies and Alexa 568-conjugated F(abЈ) 2 fragments of goat anti-mouse IgG and then determined whether the treatment altered the distribution of ␣ M and ␤ 2 subunits on the cells. As shown on Fig. 2A, in the absence of cross-linking antibodies, the ␤ 1 and ␣ M subunits were both distributed uniformly over the cell surface, and no clusters of these subunits were observed on the ␣ M cells. Clustering of integrin ␤ 1 with antibody did not trigger redistribution of ␣ M , and ␣ M did not co-localize with the clustered ␤ 1 subunits ( Fig. 2A). Similar results were observed when the endogenous ␤ 3 subunit expressed by the cells was clustered; no redistribution of the ␣ M subunit was observed (data not shown). When ␣ 4 and ␤ 2 subunits were stained on the ␤ 2 cells, some clusters of both subunits were observed on the periphery of the cells, and some of the clusters were co-localized (Fig. 2B). When the ␣ 4 subunit was cross-linked with antibody, it clustered and redistributed from the edges to the interior of the cells. However, this reorganization did not affect the distribution of the ␤ 2 subunit and did not lead to co-localization of the ␣ 4 and ␤ 2 subunits. Similarly, when the ␣ V subunit was clustered on the cells with antibody, the distribution of the ␤ 2 subunit was not affected (data not shown). In control experiments, we clustered ␣ M ␤ 2 on the ␣ M ␤ 2 cells. Antibody cross-linking of the ␣ M subunits coclustered the ␤ 2 subunits (Fig. 2C), and antibody cross-linking of the ␤ 2 subunits clustered the ␣ M subunits into the same structures. Thus, the endogenous integrins of the cells do not appear to serve as partners for the integrin ␣ M and ␤ 2 subunits, and the individual subunits distribute on the cell surface in a pattern similar to that of the heterodimeric integrins.
The properties of the ␤ 2 cells were further characterized with respect to expression of conformational and divalent ion-sensi- A, localization of the ␣ M subunit (green) before and after clustering of integrin ␤ 1 (red) on ␣ M cells. The ␤ 1 subunit was clustered on the ␣ M cells with anti-␤ 1 mAb and Alexa 568-conjugated F(abЈ) 2 fragments of goat anti-mouse IgG as described under "Experimental Procedures," or alternatively, the cells were not treated with these antibodies (resting cells). Next, the cells were fixed with 4% paraformaldehyde for 20 min at 22°C, stained with biotinylated mAb to the ␣ M subunit (clone 44a) and Alexa 488-conjugated avidin, and observed by fluorescence microscopy. Scale bars ϭ 8 m (upper panels) and 4 m (lower panels). B, localization of integrin ␤ 2 (green) before and after clustering of integrin ␣ 4 (red) on the ␤ 2 transfectants. The experiments were performed on the ␤ 2 cells as described for A using a mAb to the ␣ 4 subunit to induce clustering and biotinylated mAb to the ␤ 2 subunit for staining.  (44), and MEM148 (an activation-specific complex epitope dependent on the N-terminal and middle segments of the ␤ 2 subunit) (51). The binding of each of these mAbs was affected by divalent cations. mAb IB4 reacted well with ␤ 2 in the presence of Mg 2ϩ /Mn 2ϩ and less well and similarly in the presence of Ca 2ϩ or EDTA. The reactivity pattern of mAb TS1/18 with the ␤ 2 cells recapitulated the reactivity pattern observed with intact ␣ M ␤ 2 (44), viz. mAb TS1/18 recognized ␤ 2 in the presence of divalent cations and reacted poorly in the presence of EDTA, and mAb MEM148 showed a similar reactivity pattern. The similar reaction of the ␤ 2 cells with mAb 5A5 confirms that the cells could react similarly in the presence of either divalent cations or EDTA. Together, these data demonstrate surface expression of the ␤ 2 subunit and further suggest that the subunit folds and can transit between different divalent ion-dependent states on the cell surface in the absence of the ␣ M subunit.
Cell AttachmentրAdhesion-We tested the capacity of cell lines expressing ␣ M ␤ 2 , ␣ M , and ␤ 2 to adhere to a series of established ␣ M ␤ 2 ligands: Fg, FX, ICAM-1, iC3b, and dOva (Fig. 3A). The ␣ M ␤ 2 and ␣ M cells adhered to all of the ligands tested, whereas the ␤ 2 cells exhibited a more restricted recognition, attaching only to Fg, FX, and iC3b. During initial attachment (the first 10 -20 min), the extent of attachment of the ␤ 2 cells to the common substrates Fg, FX, and iC3b was ϳ30 -50% lower than observed with the ␣ M ␤ 2 and ␣ M cells. Shown in Fig. 3A is a 25-min time point, at which the differences in the adhesion of the ␤ 2 cells versus the ␣ M ␤ 2 and ␣ M cells to these substrates were still different. However, by 45 min, the extent of adhesion of all three cell lines to these substrates was similar, whereas background adhesion to PVP-coated wells remained low and similar for the three cell lines. As shown Fig.  3B, with iC3b used as a representative ligand, the reactions of the cell lines exhibited the appropriate specificity patterns: mAbs to the ␣ M I domain (clones 44, 44a, and 904) blocked the adhesion of the ␣ M ␤ 2 and ␣ M cells to all test substrates but did not affect substrate recognition by the ␤ 2 cells, and mAbs IB4 and TS1/18 blocked substrate recognition by the ␣ M ␤ 2 and ␤ 2 cells but not by the ␣ M cells. With all of the ␣ M ␤ 2 ligands tested in Fig. 3A, similar results were obtained, i.e. the appropriate inhibitory mAbs reduced adhesion by Ͼ85%. We also examined  Fig. 3C. The extent of adhesion of the ␤ 2 cells was slightly lower than that of the ␣ M ␤ 2 and ␣ M cells, but the dependence on ligand concentration was similar for all three cell lines and for all three ligands shown. Of note, the adhesion of the ␣ M ␤ 2 cells exhibited a bell-shaped dependence on Fg concentration as previously reported (22), and this pattern was also observed with the ␤ 2 cells but was less pronounced with the ␣ M cells. Hence, this unusual concentration dependence appears to depend upon the ␤ 2 subunit.
In addition to the specificity difference, a morphological distinction was observed in the way the ␤ 2 cells adhered to substrates compared with the ␣ M ␤ 2 and ␣ M cells. Whereas the ␣ M ␤ 2 and ␣ M cells attached and spread upon the ligands, the ␤ 2 cells attached but did not spread on ligand-coated surfaces. Even after 2 h, the ␤ 2 cells failed to spread on any of the ligands to which they attached, whereas the ␣ M and ␣ M ␤ 2 cells began spreading within 10 min. Consequently, the ␤ 2 cells could be detached from all ligands with rigorous washing. With gentle washing of the assay plates, they remained attached to wells coated with Fg, FX, and iC3b but not to wells coated with ICAM-1 and dOva or to control wells coated with PVP alone.
The difference in the spreading characteristics of the ␣ M ␤ 2 , ␣ M , and ␤ 2 cells is quantified in Fig. 4A using Fg, which is recognized by all three cell lines, as a substrate. After 30 min of incubation, most of the ␣ M ␤ 2 and ␣ M cells had spread on Fg, whereas Ͼ95% of the ␤ 2 cells remained unspread but still attached to the surface. Nevertheless, by 45 min, the total number of attached cells was similar: 85-90% of the ␤ 2 cells, 90 -95% of the ␣ M ␤ 2 cells, and 90 -95% of the ␣ M cells. The ␤ 2 cells did not spread upon attachment to ligand (Fig. 4B) and displayed no detectable filopodia.
We also further explored the cation dependence of ligand recognition by the three cell lines again using Fg as a common ligand. With all three cell lines, maximal adhesion was observed in the presence of Mn 2ϩ ; Mg 2ϩ supported adhesion; and adhesion in the presence of Ca 2ϩ was only slightly above the background level observed with mock-transfected cells (Fig.  5A). There are two known recognition sites for ␣ M ␤ 2 within the Fg ␥ chain: P1 (␥ chain amino acids 191-201) (21) and P2 (␥ chain amino acids 377-395) (22), of which P2-C (␥ chain amino acids 383-395) is particularly important. As shown in Fig. 5B, the ␣ M ␤ 2 cells adhered similarly to both P1 and P2-C. The extent of adhesion of the ␣ M cells to these peptides was the same as observed with the ␣ M ␤ 2 cells. 2 In contrast, the ␤ 2 cells attached only to P2-C and not to P1. Thus, there is a clear distinction in the recognition of the two Fg peptides by the ␣ M and ␤ 2 subunits.
Cell Migration-The capacity of the ␣ M ␤ 2 , ␣ M , and ␤ 2 cells to migrate to ␣ M ␤ 2 ligands was assessed in a modified Boyden chamber. Rather than counting the cells adhering to the substrate ligand coated onto the underside of the membrane separating the upper and lower chambers as we had done previously (20), the cells were quantified using the DNA-binding Cyquant dye, and the total number of migrated cells (those adhering to the underside of the filter and in the lower chamber) was measured. This procedure was particularly helpful with the ␤ 2 cells, which migrated but did not adhere well to the 2 D. A. Solovjov, E. Pluskota, and E. F. Plow, unpublished data.  substrates and consequently were recovered in the lower chamber. The migration of the various cell lines to Fg, FX, ICAM-1, iC3b, and dOva is shown in Fig. 6A. Although the extent of migration was dependent upon ligand concentration, the relative relationships between the various ligands and their capacity to support the migration of the cell lines was maintained at other ligand concentrations as well. 2 The ␣ M ␤ 2 cells migrated to all test ligands compared with the non-coated filters. The extent of migration varied from ϳ1.5-fold with dOva to ϳ5-7fold with the other ligands. None of the ligands supported migration of the ␣ M cells. To exclude the possibility that the ligand concentration selected (10 g/ml) may not be optimal to induce migration of the ␣ M cells, we performed the assay in the presence of a broad range of Fg concentrations (Fig. 6B). In contrast to the ␣ M ␤ 2 and ␤ 2 cells, the ␣ M cells did not migrate at any concentration tested. Thus, recognition of ligand by the ␣ M I domain is not sufficient to mount a migratory response. This contrasts with the behavior of the ␤ 2 cells: the ligands to which the ␤ 2 cells attached (Fg, FX, and iC3b) also supported migration of these cells. ICAM-1 and dOva did not support migration of the ␤ 2 cells. We also tested whether the capacity of the ␣ M cells to migrate to any ligand was impaired and used fibronectin, the ␤ 1 ligand, for this analysis. This ligand supported migration of the ␣ M cells to an extent similar to that observed with the mock-transfected cells as well as the ␣ M ␤ 2 and ␣ M cells. As quantified by the fluorescence of the migrated cells, all four cell types yielded a signal of 2200 -2500 units, whereas in the absence of fibronectin, the background migration was 250 -400 units. Thus, there was no intrinsic difference in the capacity of the cell lines to mount a migratory response.
To understand how metal ions influence cell migration, we tested the ability of the cell lines to migrate to Fg in the presence or absence of different divalent cations. As shown in Fig. 7A, neither ␣ M cells nor mock-transfected cells migrated to Fg under any of the divalent cation conditions used. In the presence of Ca 2ϩ , none of the cell lines migrated to Fg; the ␣ M ␤ 2 and ␤ 2 cells showed a slight tendency toward enhanced migration relative to the mock-transfected controls, but the differences were not statistically significant (p Ͼ 0.2). In the presence of Mg 2ϩ , the ␣ M ␤ 2 cells displayed a robust migratory response to Fg, whereas the ␤ 2 cells migrated weakly. Mn 2ϩ supported migration of both ␣ M ␤ 2 and ␤ 2 cells to the ligand, and the extent of migration of the two cell lines was similar. The Fg peptide P1 did not support migration of the ␣ M ␤ 2 and ␤ 2 cells, but P2-C supported migration of both cell types (Fig. 7B). This difference was observed in the presence 2 mM Mn 2ϩ and a 2 mM Ca 2ϩ /Mg 2ϩ mixture. Mg 2ϩ or Ca 2ϩ alone did not support migration to P2-C (data not shown).
Ligand Recognition by the Purified Integrin Subunits-The cell lines expressing ␣ M ␤ 2 , ␣ M , and ␤ 2 were used as a source for isolation of the individual subunits and the heterodimeric receptor. The ␣ M subunit was immunopurified on immobilized mAb LM2/1, and the ␤ 2 subunit and the ␣ M ␤ 2 complex on immobilized mAb IB4. After low pH elution followed by rapid neutralization and dialysis, SDS-PAGE was performed. The isolation yielded each subunit and the intact receptor in high purity (Fig. 8A). No other proteins in complex with the isolated ␣ M or ␤ 2 subunits were detected. Since dissociated ␤ subunits are known to aggregate on cell surfaces (52), gel filtration experiments were performed to assess the status of the purified ␤ 2 subunit. Upon HPLC gel filtration on a TSK G4000SW column, purified ␤ 2 eluted as a single major peak at the volume FIG. 6. The ␤ 2 subunit mediates cell migration to some but not all ␣ M ␤ 2 ligands. A, HEK293 cell lines expressing ␣ M ␤ 2 , ␤ 2 , or ␣ M were added (2 ϫ 10 5 cells/ well) to Costar 24-transwell plates in serum-free Dulbecco's modified Eagle's medium/nutrient mixture F-12 containing 2 mM Mn 2ϩ and incubated for 16 h in a humidified incubator at 37°C and 5% CO 2 . The lower chambers contained the indicated ␣ M ␤ 2 ligands (ϳ10 g/ml). The migrated cells were quantified as described under "Experimental Procedures." The data represent means Ϯ S.E. of duplicates from three experiments. B, the transfectants were allowed to migrate to various concentrations of Fg (0 -80 g/ml) present in the lower chamber. Experiments were performed as described for A. The background (migration in the absence of Fg) has been subtracted. The data are expressed as means Ϯ S.E. of triplicates from two independent experiments.
corresponding to a molecular mass of ϳ100 kDa in either the presence or absence of divalent cations (data not shown).
In Fig. 8B, we present the results obtained upon the direct binding of purified ␣ M ␤ 2 , ␣ M , and ␤ 2 to our series of ␣ M ␤ 2 ligands. The ligands Fg, FX, ICAM-1, iC3b, and dOva were biotin-labeled and immobilized on agarose beads with streptavidin, and the binding of the radiolabeled receptor or its subunits was evaluated after 20 min in a Mn 2ϩ -containing buffer. Both ␣ M ␤ 2 and ␣ M bound to all of the ligands tested. The reactivity of the ␣ M subunit with these ligands is consistent with various studies implicating the ␣ M I domain in their recognition by intact ␣ M ␤ 2 (reviewed in Ref. 24). The ␤ 2 subunit also recognized certain ligands (Fg, FX, and iC3b), consistent with the capacity of the ␤ 2 cells to attach and migrate to these substrates. In the absence of any divalent cations, ␣ M ␤ 2 or its subunits did not bind any of the ligands. Binding of the receptor or its subunits was low in the presence of Ca 2ϩ or Mg 2ϩ alone compared with Mn 2ϩ , but the combination of Mg 2ϩ and Ca 2ϩ (2 mM each) was almost as effective as Mn 2ϩ in supporting ligand binding. 2 The one distinction between the cells and the purified proteins was in the recognition of dOva; the ␣ M cells showed low adhesion to this ligand compared with the ␣ M ␤ 2 cells, whereas purified ␣ M and ␣ M ␤ 2 bound dOva similarly. DISCUSSION In this study, we have examined the contributions of the individual subunits to the functions of integrin ␣ M ␤ 2 . Our approach has been to interrogate the properties of cell lines expressing the individual subunits and to analyze the functions of the subunits purified from these cells. Our results indicate that the ␣ M subunit can mediate recognition of a broad spectrum of ␣ M ␤ 2 ligands independent of the ␤ 2 subunit. Interaction of ligand with the ␣ M subunit is sufficient to support firm adhesion and spreading of cells but cannot mediate cell migration. In contrast, the ␤ 2 subunit recognizes a more limited repertoire of ligands and cannot mediate firm cell adhesion and spreading but can support migration of cells on recognized ligands. Thus, it appears that many of the adhesive and migratory responses mediated by ␣ M ␤ 2 are reflections of the activities of the individual subunits. Only with a restricted group of ligands (represented by ICAM-1 and P1) are both subunits necessary, and this requirement is only necessary to mount a migratory but not an adhesive response. These observations appear to be counterintuitive to the picture deduced from the ␣ V ␤ 3 crystal structure, in which residues in both subunits simultaneously engage different residues within a bound RGD peptide ligand (53). This may represent a fundamental difference in the way in which integrins with I domains in their ␣ subunit engage ligands.
␣ M ␤ 2 expressed in HEK293 cells exhibited all of the functional responses attributed to the naturally occurring integrin in leukocytes (20,54). The transfected ␣ M ␤ 2 cells adhered to the same set of test ligands as recognized by ␣ M ␤ 2 in leukocytes and supported attachment, spreading, and migration on these ligands. The ligand recognition capacity was maintained with the purified receptor. The purified receptor also displayed the divalent ion requirements for ligand recognition similar to those of cells bearing the receptor, viz. divalent ions were required for ligand recognition, and maximal interactions were observed in the presence of Mn 2ϩ . Mn 2ϩ not only fulfills the divalent ion requirement for ligand binding but also activates this and other integrins (29,55), and these activities were observed with the ␣ M ␤ 2 cells and with the purified receptor. Mg 2ϩ also supported adhesion and migration of the ␣ M ␤ 2 cells, but Ca 2ϩ did not or did so poorly. A distinction between the purified receptor and the ␣ M ␤ 2 cells was found in the capacity of Mg 2ϩ to support responses: Mg 2ϩ supported adhesion and migration of the ␣ M ␤ 2 cells, but binding of the purified receptor to ligands was poor in the presence of this divalent cation. One can speculate that cellular signals elicited by ligand engagement help to stabilize the cellular interactions mediated by the receptor, but this cannot happen in the purified system. An analogous situation arises with integrin ␣ 2 ␤ 1 : cells expressing the receptor recognize collagen in the presence of Mg 2ϩ but not Ca 2ϩ , whereas the solubilized receptor is less cation-selective (56).
Cells that expressed ␣ M subunits alone were able to perform the adhesive functions of ␣ M ␤ 2 , i.e. they recognized the same set of ligands as the ␣ M ␤ 2 cells and mediated attachment and spreading over surfaces coated with them. The purified ␣ M subunit duplicated the ligand recognition specificity of the ␣ M cells. Ligand recognition either by the ␣ M cells or the purified ␣ M subunit was inhibited by mAbs to the ␣ M I domain. The contribution of the ␣ M subunit and its I domain to ligand recognition is consistent with numerous studies (e.g. Refs. 33, 38, and 57) demonstrating a prominent role for this region in function. Nevertheless, the ␣ M cells did not duplicate all of the properties of the ␣ M ␤ 2 cells; ␣ M cells were not able to migrate to any of the ␣ M ␤ 2 ligands tested. Thus, although ligand engagement may mediate firm adhesion of these cells, migration does not necessarily follow hand-in-hand as a functional response. Also surprising is the observation that the ␣ M subunit could support cell spreading. A spreading reaction reflects a link between the extracellular ligand and the cellular cytoskeleton via the integrin. In preliminary studies, we also found that cytochalasin D, a cytoskeletal disruptive agent, prevented adhesion and spreading of the ␣ M cells on the test ligands. Therefore, our data suggest a linkage, either direct or indirect, between the ␣ M subunit and the cytoskeleton and further imply that the ␣ M subunit has the ability to transmit an outside-in signal into the cells independent of the ␤ 2 subunit. Cytoskeletal connections have been ascribed primarily to the ␤ and not the ␣ subunit of integrins (58). Nevertheless, in unpublished studies, 2 we have found that deletion of the cytoplasmic tail of the ␣ M subunit resulted in activation of ␣ M ␤ 2 in a way similar to activation of the receptor by cytochalasins (59).
Although none of the ␣ M ␤ 2 ligands supported migration of the ␣ M cells, some of the ligands (iC3b, FX, Fg, and the Fg peptide P2-C) supported migration of the ␤ 2 cells. These same ligands also supported weak attachment of the ␤ 2 cells and bound the purified ␤ 2 subunit. Thus, recognition by the ␤ 2 subunit does equate with migration. However, weak adhesion, as defined by an inability to support spreading, does not equate with migration: ␣ M ␤ 2 cells spread and migrated on all of the test ligands, including ICAM-1 and P1, which were not recognized by the ␤ 2 cells. To account for migration to these latter ligands, one must suggest that ligand presentation by the ␣ M subunit allows for engagement by the ␤ 2 subunit that mediates a migratory function, or there is a distinction in the way that these ligands versus those that bind the ␤ 2 subunit induce migration. Migrated ␣ M ␤ 2 cells remained adherent to ligands, whereas the ␤ 2 cells did not but rather settled into the lower chamber after crossing the membrane. Whether this reflects a fundamental difference in the migratory response of the ␣ M ␤ 2 and ␤ 2 cells (e.g. chemotactic versus chemokinetic responses) will require detailed analyses. Our data on the function of the ␤ 2 cells and the purified ␤ 2 subunit under different divalent cation conditions establish that the subunit can fold independent of the ␣ M subunit. This conclusion is consistent with the data of Xiong and Zhang (44), who noted that the binding of iC3b and a mAb to their ␤ 2 cells was differentially affected by divalent cations.
Some of the specific observations made in this study merit special comment. First, we were able to express the individual ␣ M and␤ 2 subunits at levels readily sufficient to conduct functional analyses. The expressed subunits reacted with several different mAbs and appeared to undergo divalent ion-dependent conformational transitions consistent with those of the intact receptor. These observations, together with the ability of the subunits to recognize various ligands with appropriate specificities, would appear to refute claims that expression of functional subunits is only possible if the SDL is deleted (49). We verified that mRNA of the ␤ 2 cells did indeed code for the SDL. We also did not detect a partner for either the ␣ M or ␤ 2 subunit in the cells expressing the individual subunits, including usual partnering with the endogenous integrins of the cells. A partner may have gone undetected if it dissociated readily in the detergent used for extraction, if it did not label well with biotin or stain well with Coomassie Blue, or if it were of very low or very high molecular mass such that it would not be detected by SDS-PAGE analyses. All of these possibilities seem remote since the purified subunits were functional, and no additional component was detected in these preparations as well as in immunoprecipitates of the subunits from the cells in FIG. 8. Ligand recognition by purified integrin ␣ M ␤ 2 and its subunits. A, the ␣ M and ␤ 2 subunits and ␣ M ␤ 2 heterodimer were isolated from HEK293 cells expressing the receptor or its subunits on mAb columns and analyzed on 4 -12% acrylamide gradient gels under reducing conditions as described under "Experimental Procedures." The densitometry of the gels indicated that the proteins were Ͼ95% pure. B, shown is the direct binding of purified ␣ M ␤ 2 , ␣ M , and ␤ 2 to the indicated ␣ M ␤ 2 ligands. The ␣ M ␤ 2 ligands were biotinylated, immobilized on streptavidin-agarose, and allowed to interact with purified and radioiodinated ␣ M ␤ 2 , ␣ M , or ␤ 2 for 30 min at 37°C in the presence of 2 mM Mn 2ϩ . After washing, the amount of radioactivity retained on the beads was counted in a ␥ counter. The data are expressed as means Ϯ S.E. of quadruplets from three independent experiments. a variety of detergents extracts and under conditions in which the noncovalent complex of the urokinase-type plasminogen activator receptor with ␣ M ␤ 2 was detected. We do not exclude that the subunits may exist in multimers on the cell surface, but we could find no evidence of extensive clustering of the subunits on the ␣ M or ␤ 2 cells. Limited self-association could be undetected and remains an interesting possibility in view of recent studies suggesting that the transmembrane segments of integrin ␣ IIb and ␤ 3 subunits can undergo homo-oligomerization (60,61). However, by gel filtration, it appeared that the isolated subunits could function, even as monomers.
Second, the capacity of the ␤ 2 subunit to recognize certain ligands brings into question the ligand specificity differences between ␣ M ␤ 2 and ␣ L ␤ 2 ; iC3b and FX have been regarded as ligands of the former but not of the latter integrin (18,62). It should be noted that, with intact cells, the weak attachment and absence of spreading may have been registered as a negative response of ␣ L ␤ 2 to these ligands. Our results are consistent with those of Yalamanchili et al. (40), who detected interaction of their I domainless ␣ M ␤ 2 with iC3b and FX. A question for future studies is whether ␣ L ␤ 2 cells migrate and attach weakly to these ligands. It has already been reported that an I domainless ␣ L ␤ 2 failed to adhere to ligands (63). A difference does appear to exist in the capacity of the ␤ 2 cells and the ␣ M I domainless integrin to recognize Fg. This may reflect a masking of the ␤ 2 contact sites for Fg by the ␣ M subunit in the I domainless receptor. Alternatively, this may simply reflect a difference in assay conditions or in Fg ligands used. Adhesion of ␣ M ␤ 2 to Fg is very dependent on the ligand coating concentration (64). Also, the P2 region of Fg is not on the surface of soluble Fg, and the immobilization conditions may control its exposure (22).
Third, our data show that the P2-C sequence of Fg is recognized by both the ␣ M and ␤ 2 subunits, whereas P1 is recognized by only the ␣ M subunit. This difference provides a potential explanation for the differences in functional responses elicited by the two Fg peptides in cell migration assays (20), even though they compete with one another and bind to an overlapping site within the ␣ M I domain (64).
In summary, the individual subunits of ␣ M ␤ 2 are able to mediate specific and distinct functions of the heterodimeric integrin. Each subunit is able to undergo cation-dependent transitions independent of the other subunit. The ␣ M subunit appears to impart the firm adhesion mediated by the receptor, and the ␤ 2 subunit is critical for cell migration. Nevertheless, in the absence of detectable ␤ 2 recognition, the heterodimeric receptor can mediate migration to certain ligands, suggesting that, for certain functions, the ␤ 2 subunit may modulate responses dependent on the ␣ M subunit.