Dual Functionality of the Anti-β1 Integrin Antibody, 12G10, Exemplifies Agonistic Signalling from the Ligand Binding Pocket of Integrin Adhesion Receptors*

Although integrins are known to mediate connections between extracellular adhesion molecules and the intracellular actin cytoskeleton, the mechanisms that are responsible for coupling ligand binding to intracellular signaling, for generating diversity in signaling, and for determining the efficacy of integrin signaling in response to ligand engagement are largely unknown. By characterizing the class of anti-integrin monoclonal antibodies (mAbs) that stimulate integrin activation and ligand binding, we have identified integrin-ligand-mAb complexes that exhibit differential signaling properties. Specifically, addition of 12G10 mAb to cells adhering via integrin α4β1 was found to trigger disruption of the actin cytoskeleton and prevent cell attachment and spreading, whereas mAb addition to cells adhering via α5β1 stimulated all of these processes. In contrast, soluble ligand binding to either α4β1 or α5β1 was augmented or unaffected by 12G10. The regions of the integrin responsible for differential signaling were then mapped using chimeras. Surprisingly, a chimeric α5 integrin containing the β-propeller domain from the ligand binding pocket of α4 exhibited the same signaling properties as the full-length α4 integrin, whereas exchanging or removing cytoplasmic domains had no effect. Thus the mAb 12G10 demonstrates dual functionality, inhibiting cell adhesion and spreading while augmenting soluble ligand binding, via a mechanism that is determined by the extracellular β-propeller domain of the associating α-subunit. These findings therefore demonstrate a direct and variable agonistic link between the ligand binding pocket of integrins and the cell interior that is independent of the α cytoplasmic domains. We propose that either ligand-specific transmembrane conformational changes or ligand-specific differences in the kinetics of transmembrane domain separation underlie integrin agonism.

domain is preceded by, and inserted into, a plexin-semaphorinintegrin (PSI) domain (9). The PSI domain is located below the hybrid domain and alongside the ␤-subunit leg, which consists of four tandem cysteine-rich epidermal growth factor-like domains and a C-terminal ␤-sheet domain termed the ␤-tail domain. Both subunits are linked to relatively short (typically Ͻ60 amino acid residues) cytoplasmic tail domains via a single transmembrane pass.
The extracellular matrix components and cell surface molecules that comprise the family of integrin ligands are many and varied. Individual integrin ligands have been shown to bind to several integrins, and reciprocally, individual integrins bind to multiple ligands (10). Ligand binding to integrins requires the formation of a ligand carboxyl-divalent cation coordination complex at the metal ion-dependent adhesion site in the integrin A domain (8,11). In non-␣A domain-containing integrins, residues from both the ␤-subunit A domain and the ␣-subunit ␤-propeller domain contribute to ligand binding, and ligand binding specificity has been shown to be determined by loop structures in these regions (12)(13)(14). Integrin-mediated adhesion has wide-ranging effects on cell survival, motility, differentiation, and proliferation (3). Unexpectedly, therefore, the signals generated by integrins and the composition of different adhesion signaling structures initiated by integrin-ligand engagement are highly diverse (2). However, the key molecular events determining this diversity and the mechanisms determining the variation in the signals transduced by different integrin heterodimers are largely unknown.
Distinctive cellular responses to integrin-ligand engagement have been reported on substrates recognized by the fibronectin (Fn)-binding integrin ␣4␤1. Engagement of this integrin results in an enhanced cell migratory phenotype coupled to a reduction in cell spreading and focal adhesion formation, compared with ␣2␤1 or ␣5␤1. These functional properties were demonstrated to be conferred by the ␣4 cytoplasmic domain (15)(16)(17). These studies suggest that the ␣4 cytoplasmic domain modulates the association of cytoskeletal and signaling molecules with its partner ␤1 cytoplasmic domain differently to that of ␣2 or ␣5. Alternatively they suggest that the ␣4 tail binds directly to cytoplasmic factors that modulate integrin signaling. In this regard, ␣4 has been shown to bind directly to paxillin, with the association reducing cell spreading and promoting cell migration (18). Thus, at present, the ␣-subunit cytoplasmic tails are believed to have a major effect on the generation of integrin signal diversity from different ␣␤ heterodimers. Recent evidence has indicated that the extracellular ␤A domain also plays a role in the generation of signal diversity. The ␤A domains of the ␤1and ␤3-subunits were found to display different abilities to activate the Rho family of GTPases, which are involved in cytoskeletal organization (19,20). Signal diversity from integrin-ligand engagement can also result from additional co-operating signaling partners. In this regard, ␣4␤1 and ␣5␤1 demonstrate a differential requirement for protein kinase C␣ signaling during cell migration and the formation of focal adhesion structures. These processes are mediated through the co-operative signaling of the proteoglycan co-receptor syndecan-4 with ␣5␤1, but not with ␣4␤1 (21).
To carry out its dynamic functions, integrin structure is specialized to be highly responsive and regulated (22,23). The receptors are able to switch from an inactive (low affinity) to active (high affinity) state, and vice versa, in response to binding events taking place at both the ligand binding pocket and the cytoplasmic domains. A large body of evidence exists to indicate that both priming of integrins, to promote ligand binding, and integrin activation, subsequent to ligand binding, involve conformational changes between and within the integrin subunits (22). A distinctive feature of integrin activation is the transition from a highly bent conformation, which represents the inactive form, to an extended, primed conformation. The bend in the molecule is located between the thigh and calf domains of ␣V and the epidermal growth factor-like 2/3 domains of ␤3 (5, 7).
The studies described above raise many questions relating to integrin structure and function. For example, how is the efficacy of signaling by different integrins determined? How is ligand binding converted into a signal? What is the signaling route taken through the integrin? Are the mechanisms of priming and activation different or the same? Some of these questions have already begun to be answered. For example, engagement of integrins with their ligands results in conformational changes within the integrin ␤A domains leading to the swing-out of the hybrid domain in relation to the ligand-binding head region and the propagation of integrin signaling in this manner (5,24,25). Furthermore, the binding of cytoplasmic factors such as talin leads to integrin priming via a route that most likely involves the separation of the cytoplasmic and transmembrane domains (6,26). The differences between these two processes has been suggested to be the basis of the discrimination between integrin priming and activation-induced signals (27).
We have attempted to address some of these issues by characterizing the class of anti-integrin monoclonal antibodies (mAbs) that stimulate integrin activation and ligand binding. Many function-modulating integrin mAbs act allosterically, displacing the conformational equilibrium of the active/inactive integrin to achieve their effects. Therefore anti-integrin mAbs are capable of acting as pseudo-agonists by stabilizing integrin signaling conformations (22,28). In this way we have identified integrin-ligand-mAb complexes that exhibit differential signaling properties and determined a hitherto unappreciated mechanism for controlling signal generation diversity that is dictated by the ␣-subunit region of the integrin ligand binding pocket.
All cell lines were from the European Collection of Animal Cell Cultures unless otherwise stated, passaged every 3-4 days and cultured at 37°C and 5% CO 2 in defined medium as follows. A375-SM human melanoma cells (provided by I. Fidler, University of Texas) were cultured in minimal essential medium with Earle's salts, supplemented with 10% (v/v) fetal calf serum, minimal essential medium vitamins, non-essential amino-acids, sodium pyruvate, and 1% (v/v) L-glutamine. COS-1 African green monkey kidney cells, IMR 32 human neuroblastoma cells, and HT1080 human fibrosarcoma cells were cultured in Dulbecco's minimal essential medium with 0.11 g/liter sodium pyruvate and pyridoxine supplemented with 10% (v/v) fetal calf serum, 1% (v/v) L-glutamine. Cells were detached from flasks by treatment with 1ϫ trypsin/EDTA for 5 min for routine passaging. The MOLT-4 human lymphoblastic leukemia suspension cell line, the K562 human chronic myelogenous leukemia suspension cell line, K562-␣3A cells (K562 cells stably transfected with the human ␣3A integrin subunit; provided by A. Sonnenberg, K562-␣4 cytoplasmic deletion mutant cells (36) (provided by M. Hemler, Dana-Farber Cancer Institute, Boston, MA) and the Jurkat human T cell lymphoblastic leukemia suspension cell line (provided by P. Shore, University of Manchester, UK) were cultured in RPMI 1640 supplemented with 10% (w/v) fetal calf serum and 1% (v/v) L-glutamine. An addition of 1 mg/ml G418 (Invitrogen) was made for cells transfected with the appropriate expression vector.
cDNA Plasmids and Construction of ␣4/␣5 Subunit Chimeras-The ␣6A cDNA in pRC-CMV was a gift from A. Sonnenberg. Site-directed mutagenesis of ␣4 and ␣5 cDNAs was performed using the method of Kunkel (37), or using the GeneEditor site-directed mutagenesis kit (Promega, Madison, WI) according to the manufacturer's instructions. For GeneEditor, 5Ј to 3Ј oligonucleotides (MWG-Biotech UK Ltd., Milton Keynes, UK) containing the desired mutations were GTAGTA-ATTGTTGACGCTAGCTTAAGCCACCCTGAGTCAG for ␣4 and ATCGTGTCCGCTAGTGCTAGCCTCACCATCTTCCCCGCC for ␣5. Briefly, to construct the ␣4/␣5 chimera composed of the extracellular and transmembrane region of ␣4 linked to the cytoplasmic domain of the ␣5-subunit, a HindIII site was introduced into ␣4 at the equivalent position to an existing HindIII site in ␣5 as described previously (16). To construct the ␣4P␣5L chimera, comprising the N-terminal ␤-propeller domain of ␣4 (Tyr 1 -Leu 440 ) with the ␣5 leg, transmembrane, and cytoplasmic domains (Thr 459 -Ala 1008 ), an alignment of the primary structure of ␣4 and ␣5 in the region of the ␤-propeller domain was performed (ClustalW, available at expasy.ch). The conserved ASL sequence (Ala 438 in ␣4 and Ala 456 in ␣5) was chosen to introduce a unique NheI restriction enzyme site into the equivalent position in the ␣4 and ␣5 cDNAs. After restriction enzyme digestion the required fragments were ligated into the mammalian expression vector pCDNA3 (Invitrogen) via XhoI/ SalI and XbaI sites to enable cell surface expression. All mutagenesis and chimera constructions were verified by DNA sequencing.
Transfection of Mammalian Cells-K562 cells were transfected with plasmid DNA either by electroporation or by using GeneJammer transfection reagent (Stratagene). Electroporation was performed with subconfluent cells resuspended to 6 -7 ϫ 10 6 cells/ml (K562 cells) or 7-10 ϫ 10 6 cells/ml (COS-1 cells). 750 l of cells was aliquoted into 0.4-cm electroporation cuvettes (Bio-Rad), and 10 -20 g of DNA in 50 l of water was added on ice for 20 min prior to electroporation at 230 V and 950 microfarads (K562 cells) or 250 V and 950 microfarads (COS-1 cells) with a resultant time constant of 18 -22 mS using a Bio-Rad Gene-Pulser II. Electroporated K562 cells were transferred to prewarmed medium and grown for 48 h before addition of the selection antibiotic G418 at 1 mg/ml. Alternatively, 6 l of GeneJammer transfection reagent (Stratagene) was diluted in serum-free medium for 5-10 min, 1 g of DNA was added, and the mixture was incubated for 5-10 min. Sub-confluent cells were resuspended at 1-2 ϫ 10 6 cells/ml, and 1 ml was seeded in 35-mm dishes. The DNA/GeneJammer mixture was added to the cells, and the mixture was incubated for 72 h before passaging into medium containing the selection antibiotic G418 at 1 mg/ml.
To enrich the population of transfected K562 cells, cells were subjected to immunomagnetic bead selection. After 2-3 weeks of growth in medium containing selection antibiotics, cells were resuspended to 1 ϫ 10 7 cells/ml in ice-cold RPMI 1640 supplemented with 1% (v/v) fetal calf serum (wash medium). Antibodies directed against epitopes of interest were added at 5 g/ml, and the mixture was incubated on ice for 30 min. Cells were washed with cold wash medium before addition of 2 ϫ 10 7 goat anti-mouse IgG-coated magnetic beads (Dynabeads M-450, Dynal, Bromborough, UK), and the mixture was incubated on ice for 20 min. Bead-cell complexes were isolated using a magnetic separation device (Dynal), and unbound cells were removed by washing in ice-cold wash medium. Bead-cell complexes were resuspended in growth medium plus selection antibiotic, and expression was assessed by fluorescence-activated cell scanning (FACS) analysis (FACscan, BD Biosciences). The bead-sorting procedure was repeated two to three times to obtain cells expressing high levels of the protein of interest. Cells were then cloned by limiting dilution. Identical results were obtained with mixed cell populations and multiple clones. Expression and folding of the ␣4subunit was verified by FACS using a panel of anti-␣4 and anti-␣5 mAbs, or GoH3 for K562-␣6 cells.
Cell Attachment Assays-96-well plates (Costar Corning) were coated for 60 -90 min at room temperature with protein ligands diluted in phosphate-buffered saline (Invitrogen). After aspiration, wells were incubated for 30 -60 min with 10 mg/ml filtered, heat-denatured bovine serum albumin (BSA). As controls, some wells were incubated with BSA only. Wells were washed with HEPES-buffered saline (HBS, 150 mM NaCl, 25 mM HEPES, pH7.4), before addition of 50 l of HBS containing 2ϫ final concentration of antibodies, cations, and/or inhibitors where appropriate. Subconfluent cells were washed with HBS, resuspended to 0.2 to 1 ϫ 10 6 cells/ml, and 50-l aliquots were added to wells. Plates were incubated for 30 min at 37°C and 5% (v/v) CO 2 . Unbound or loosely bound cells were removed by aspiration and gentle washing with HBS. Wells were fixed by addition of 5% (v/v) glutaraldehyde in HBS. To assess the total number of cells added, 100%, 75%, 50%, 25%, and 0% cells were added to wells and fixed by the addition of 1/10 volume of 50% (v/v) glutaraldehyde. Wells were aspirated and washed with HBS before addition of 0.1% (w/v) crystal violet in 200 mM MES pH 6 for 60 min. Wells were then aspirated and washed with water before addition of 10% (v/v) acetic acid, and the absorbance at 570 nm of each well was measured with a multiscan plate reader.
Cell Spreading Assays-Wells were coated and blocked as described for cell attachment assays. Adherent cells were detached with 0.05% (w/v) trypsin, 0.02% (w/v) EDTA, resuspended to 2 ϫ 10 5 /ml in Dulbecco's minimal essential medium/25 mM HEPES, and allowed to recover for 10 min at 37°C. For experiments examining the effects of mAbs on spreading, 50-l aliquots of the cell suspension were added to wells together with 50 l of mAbs diluted to 2ϫ the final concentration in Dulbecco's minimal essential medium. Plates were incubated at 37°C and 5% (v/v) CO 2 for various times ranging from 30 min to 2 h. The cells were fixed with 5% (w/v) glutaraldehyde for 30 min, and wells were washed with phosphate-buffered saline. The degree of cell spreading was assessed as a percentage of the total number of cells counted, using phase-contrast microscopy. At least 100 cells in four to six randomly chosen high powered fields were counted per treatment point. Axiovision version 4.2 software (Carl Zeiss, Hertfordshire, UK) was used to measure cell areas, and two-tailed t tests were performed to determine statistical significance.
Immunofluorescence-To assess the effect of mAbs on pre-spread A375-SM cells, immunofluorescence was performed essentially as described previously (21). Briefly, cells were incubated with 13-mm diameter coverslips that were precoated with 10 g/ml H/120 or FnIII (6 -10) and blocked with heat-denatured BSA. Cells were allowed to spread for 60 min before the addition of mAbs (10 g/ml) for a further 30 min. For cells spread on FnIII (6 -10) , the heparin-binding fragment of Fn (H/0) was added throughout to allow full development of focal adhesion structures (21). Cells were then fixed and processed to visualize vinculin (mouse anti-human hVIN-1 mAb, 1:400 dilution, Sigma) and actin (rhodamine-conjugated phalloidin, 1:1000 dilution, Sigma).
To study the effects of other divalent cations, this buffer was treated with Chelex beads (Bio-Rad) prior to the addition of any cations, and then MgCl 2 or CaCl 2 or MnCl 2 added. For solid phase assays, antiintegrin antibodies at 20 g/ml concentrations were added to wells in 50 l of TBS/Mn, and a range of VCAM-1-Fc, biotinylated H/120, or biotinylated FnIII (6 -10) concentrations in 50 l of TBS/Mn was also added to wells as indicated. For enzyme-linked immunosorbent assay, biotinylated mAbs were added at 0.3-10 g/ml. Plates were incubated at 37°C for 2-3 h before washing wells three times with TBS/Mn. VCAM-1-Fc binding was detected using horseradish peroxidase-conjugated anti-human-Fc antibody (Sigma), and biotinylated Fn fragments or mAbs were detected with horseradish peroxidase-conjugated ExtrAvidin (Sigma). Wells were washed with TBS/Mn before addition of ABTS substrate, and absorbance readings at 405 nm were measured using a multiscan plate reader. Each experimental condition was performed using at least three replicate wells, and background attachment to BSA-blocked wells was subtracted from all measurements.

RESULTS AND DISCUSSION
12G10 Selectively Disrupts ␣4␤1-mediated Cell Attachment, Cell Spreading, and Cytoskeletal Organization-The initial aim of these studies was to identify conditions under which individual integrins exhibited variable signaling. To this end, the effects of monoclonal antibodies, previously demonstrated to stimulate ligand binding, on the signaling properties of various integrin-ligand combinations were examined. Two integrin-ligand combinations (␣4␤1 binding to H/120, the IIICS region of fibronectin, or to VCAM-1, and ␣5␤1 binding to the central cell-binding domain of fibronectin, FnIII (6 -10) ), and two activating anti-␤1 mAbs (12G10 and TS2/16) were used. ␣4␤1 and ␣5␤1 were selected because they have been reported previously to transduce different effects on cell spreading and migration via different signals (15,16). Specifically, ␣4␤1 engagement promotes enhanced cell migratory activity while reducing spreading, in part due to the ability of ␣4 to bind paxillin (18). In addition, both receptors have a differential requirement for protein kinase C␣ activation when stimulating focal adhesion formation and migration (21). 12G10 and TS2/16 were compared because they represent the two main classes of anti-integrin mAbs that enhance ligand binding and cell adhesion (4,29,38). These classes are distinguished by the ability of divalent cations and ligand occupancy to modulate expression of their epitopes. Although both 12G10 and TS2/16 have been reported to stimulate ligand binding to purified integrins, 12G10 binding to the integrin is stimulated by ligand, Mn 2ϩ and Mg 2ϩ , and inhibited by Ca 2ϩ , but the expression of the TS2/16 epitope is unaltered by cation and ligand occupancy (31). These findings demonstrate that both 12G10 and TS2/16 are allosteric activators of integrins, but they imply that the mAbs function in different ways, possibly by stabilizing different conformations of the integrins (4,22).
To probe potential signaling differences in cells between the activation states of ␣4␤1 and ␣5␤1, the effect of 12G10 and TS2/16 on ␣4␤1and ␣5␤1-mediated cell spreading was studied. The ability of cells to spread on ligands has been extensively used to assess post-ligand binding events subsequent to integrin-mediated adhesion. The A375-SM cell line expresses both ␣4␤1 and ␣5␤1 and can be directed, via the use of specific integrin ligands, to utilize either of these receptors, facilitating cell spreading with the formation of distinct focal adhesions and actin stress fibers (21). As reported previously (32), 60 -70% of A375-SM cells spread via ␣4␤1 or ␣5␤1, forming phasedark, polyhedral morphologies. Surprisingly, addition of 12G10 inhibited cell spreading mediated by ␣4␤1, but not by ␣5␤1, resulting in phase-bright cells with a round morphology. This effect was not observed with TS2/16 (Fig. 1). A small number of cells were phase-dark on ␣4␤1 ligands in the presence of 12G10, but the surface area occupied by these cells was much less than control cells. Similar results were observed for 12G10 and TS2/16 treatment of HT1080 and MOLT-4 cells spreading via ␣4␤1. In other experiments, ␣2␤1and ␣6␤1-mediated HT1080 cell spreading was not inhibited by 12G10 or TS2/16 (data not shown). Thus, 12G10 modulates ␣4␤1 function differ-ently to other integrins such as ␣2␤1, ␣5␤1, and ␣6␤1 despite recognizing the common ␤1-subunit. Subsequent experiments were designed to test the working hypothesis that the effects of 12G10 on cells occur through the induction of different mAbinduced agonistic signaling states of ␤1 integrins.
is unlikely to be due to mAb-induced steric hindrance of the ligand binding site. Furthermore, TS2/16, and other activating anti-␤1 mAbs such as 15/7 and HUTS-4 (data not shown) stabilize different conformers of ␣4␤1 that do not induce this effect. We hypothesize that the 12G10 signaling conformer of ␣4␤1 leads to the disruption of the cytoskeleton, resulting in a less adhesive phenotype.
To test the hypothesis that the 12G10-stabilized conformers of ␣4␤1 and ␣5␤1 result in differential signaling, we focused on the effects of 12G10 on the preformed cytoskeleton. Untreated cells, or cells treated with a neutral anti-␤1 mAb, K20, formed prominent vinculin-containing focal adhesions and actin stress fibers when allowed to spread via ␣4␤1 or ␣5␤1. However, cells spread via ␣4␤1 and then 12G10-treated displayed a greatly reduced actin stress fiber network with a disrupted focal adhesion staining pattern (Fig. 3A). This effect preceded, and was independent of, the ability of 12G10 to reduce cell spreading, because the phenotype occurred in cells that had spread to the same extent as control cells and was distinctly different to the effects of mAb13, which perturbs integrin-ligand binding (Fig.  3A). In addition, quantification of A375-SM cell spreading demonstrated no significant difference in the mean cell area of cells spread for 60 min and then challenged with either 12G10 or K20 (see Fig. 3 legend). The effects of 12G10 on the cytoskeleton were not observed in cells spread via ␣5␤1 (Fig. 3B). These data therefore support a model of differential signaling between the 12G10-stabilized conformers of ␣4␤1 and ␣5␤1, with the former promoting disengagement from, and retraction of, the actin cytoskeleton, leading to reduced cell attachment and spreading.

Divalent Cation Regulation of 12G10 Binding and the Effects of 12G10 on Ligand Binding
Are Similar for Both ␣4␤1 and ␣5␤1-12G10 was initially characterized as an anti-␤1 mAb that recognized a ligand-induced binding site epitope and augmented ␣5␤1-Fn interactions (29). Subsequently, the epitope of 12G10 was found to overlap with those of other function-altering anti-␤1 mAbs such as mAb13 and TS2/16. Thus, the 12-amino acid epitope sequence (Lys 207 -Lys 218 ) for these mAbs is located in the ␣2 helix of the ␤1-subunit A domain, and it was suggested that this region was involved in conformational changes that regulate integrin-ligand binding (31,39). An interesting feature of many ligand-induced binding site mAbs is the divalent cation regulation of their epitopes. Because cations also regulate ligand binding, and in some cases the pattern of effects of different cations is the same for mAb and ligand binding, it appears that some activating mAbs recognize sites that are regulated by modulators of integrin function. One possibility is that cation-responsive, activating mAbs recognize naturally occurring conformers of integrins and that they are therefore able to displace a conformational equilibrium in favor of these forms. This displacement would lead to an increase in the proportion of ligand-competent integrin in the population. Other activating mAb epitopes are unaffected by either ligand or cation binding, and here the most likely mechanism of action is through inducing an activated conformation in the integrin rather than stabilizing a naturally occurring conformation. The effects of 12G10 on ligand binding to ␣5␤1 are well characterized (29,31). In agreement with the model described above, the binding of 12G10 to ␣5␤1 was regulated in a cation-dependent manner that correlated with the activation state of the receptor (31). The cation-regulated compo- nent of the 12G10 epitope was subsequently mapped to residues Arg 154 and Arg 155 in the ␤A domain ␣1 helix, indicating that this helix is also conformationally regulated by ligand binding (24). Therefore 12G10 is believed to report the activation state of the ␤1 A domain by detecting the position of the ␣1 helix (25). Reciprocally, 12G10 is thought to augment ligand binding by stabilizing the ␣1 and ␣2 helices of the ␤1A domain in a ligandcompetent, physiologically relevant conformation (4). The binding of TS2/16 to ␣5␤1 is not modulated by cations (31), and this mAb is therefore thought to activate the integrin by inducing a non-physiological conformer.
When the divalent cation-dependent binding of 12G10 to ␣4␤1 was examined and compared with the effects of cations on ligand binding, the observed pattern mirrored that seen for ␣5␤1 (31), i.e. Mn 2ϩ supported high levels, Mg 2ϩ moderate levels, and Ca 2ϩ did not support binding (Fig. 4, A and B). As observed for ␣5␤1, divalent cations did not modulate the expression of the TS2/16 epitope (data not shown). These data suggest that 12G10 binding to ␣4␤1 correlates with the activation state of the receptor and that the mAb recognizes similar conformational changes in ␣4␤1 and ␣5␤1.
FIG. 4. Expression of the 12G10 epitope correlates with the activation state of ␣4␤1. Divalent cation regulation of (A) 12G10 (0.3 g/ml) binding to purified ␣4␤1 or (B) VCAM-1-Fc (0.5 g/ml) binding to purified ␣4␤1 in the presence of Mn 2ϩ (circles), Mg 2ϩ (squares), or Ca 2ϩ (triangles). mediated by these integrins, indicating that the mechanism of action of 12G10 on cell adhesion does not involve a direct occlusion of the ligand binding pocket of ␣4␤1.
Solid phase assays allow analysis of receptor-ligand interactions in isolation, rather than in the context of cell membranebound proteins. To further investigate the effects of 12G10 on ligand binding, soluble ligand binding assays were employed with whole cells. Binding of soluble, fluorescently labeled VCAM-1 to K562-␣4 or A375-SM cells was inhibited by EDTA and anti-functional ␣4 and ␤1 mAbs, demonstrating the specificity of binding. In contrast to the effects of 12G10 on cell attachment, however, the mAb did not inhibit soluble VCAM-1-Fc binding to either K562-␣4 or A375-SM cells (Fig. 6 and data not shown). Taken together with the solid phase assay data described above, these findings provide further evidence that the mode of action of 12G10 to selectively inhibit ␣4␤1mediated cell spreading and attachment is not due to steric blocking of the ligand binding interface.
The Ligand Binding Pocket Determines Agonistic Differences between ␣4␤1 and ␣5␤1-Having identified integrin-ligand-antibody complexes with different signaling properties, we next aimed to identify the regions of the integrins that were responsible for these differences. In previous studies, the cytoplasmic tails of integrins have been shown to be responsible for some functional differences between receptors. Particularly pertinent to this study is the fact that the ␣4 tail has the property of reducing cell spreading and increasing migration due to its interaction with the cytoplasmic signaling adaptor paxillin. To test if the specific effects of 12G10 on ␣4␤1-mediated adhesion required the ␣4 tail, an integrin chimera comprising the ␣4 extracellular and transmembrane domains and the ␣5 cytoplasmic tail was generated (Fig. 7). This chimeric integrin was expressed on the surface of K562 cells and tested in cell adhesion assays. The chimera-expressing cells demonstrated ␣4␤1-mediated cell attachment to H/120 and VCAM-1, and 12G10 inhibited this attachment in an identical manner to that seen for full-length ␣4 (data not shown). Furthermore, K562 X4C0 cells expressing ␣4 with the tail truncated after the conserved GFFKR cytoplasmic motif (36) (Fig. 7) also displayed a 12G10 inhibition profile on ␣4␤1 ligands that was identical to K562-␣4 cells (Fig. 8, A and B). These data demonstrate that the effects of 12G10 on ␣4␤1 function are independent of the ␣ tail and cannot therefore be due to modulation of the ␣4-paxillin interaction.
To define further the region of the ␣4-subunit responsible for the effects of 12G10 on ␣4␤1 function, a second ␣4/␣5 chimera, comprising the N-terminal ␤-propeller domain of ␣4 (Tyr 1 -Leu 440 ) with the ␣5 leg, transmembrane, and cytoplasmic domains (Thr 459 -Ala 1008 ), was generated (Fig. 7) and again expressed in K562 cells (termed K562-␣4P␣5L). Cells expressing this chimera attached to ␣4␤1 ligands in a manner that was identical to that observed with wild type ␣4␤1 (data not shown). This finding confirms previous observations for integrins such as ␣5␤1 and ␣V␤3, where residues in the ␤-propeller domain have been shown to contribute to the ligand binding pocket and determine ligand-binding specificity. The K562-␣4P␣5L-expressing cells also demonstrated reduced cell attachment to ␣4␤1 ligands in the presence of 12G10 but not TS2/16 ( Fig. 8C and data not shown), as seen for wild type ␣4␤1. Again this contrasted with the effect of 12G10 on these cells to attach to the ␣5␤1 ligand FnIII (6 -10) (Fig. 8D). Therefore the ␣4-propeller domain of the ␣4-subunit is sufficient to confer the ability of 12G10 to inhibit cell attachment on the ␣5-subunit. Furthermore, the effects of 12G10 and TS2/16 on soluble VCAM-1-Fc binding to K562-X4C0 and K562-␣4P␣5L cells (performed as in Fig. 6) were identical to that observed for K562-␣4 cells, i.e. 12G10 and TS2/16 did not inhibit soluble ligand binding (data not shown). These results suggest that the ␤A domain conformation, at least in the region of the conformationally important 12G10 epitope, can be modulated by the ␤-propeller domain of the associating ␣-subunit. These data demonstrate that the different 12G10-induced agonistic states of ␣4␤1 and ␣5␤1 are determined by the extracellular ligand A series of previous studies have suggested that the cytoplasmic tails of integrins, or associating signaling molecules, are responsible for the differential signaling observed between integrins sharing a common ␤-subunit. There is some limited evidence, however, that the integrin extracellular domain makes a contribution (19,20). Specifically it was found that extracellular domains, in particular the ␤A-domain, differentially regulated Rho-family GTPase signaling from ␤1 and ␤3 integrins. The results described here support the concept of extracellular domain control of integrin signaling, but fundamentally differ to that of published data in that they describe a process by which the extracellular domain of the ␣-subunit influences the function of the associating ␤-subunit, thereby suggesting another possible mechanism by which integrin heterodimers can initiate specific signals and generate signal diversity from a common ␤-subunit.
Our findings raise the issue of how different conformations of the ligand binding pocket of integrins are transduced through the molecule to the cytoplasm. Initially, shape shifting in the ␤A domain needs to be propagated to the underlying hybrid/ PSI domains. A recent crystal structure analysis of ␣IIb␤3 complexed with different peptidomimetic ligands (9) suggests how this might take place. One possibility is that different ␣-subunits might influence the degree of hybrid domain swingout induced by ligand binding. Alternatively, the kinetics of leg separation might be different for different ␣-␤ combinations. The net result of either mechanism would be to alter the proportion of integrin molecules in an activated conformation. The question remains as to how the link between ␣4␤1 and the cytoskeleton is severed. Is this mediated by modulation of a direct ␤1-cytoskeletal link or via indirect signals to the cytoskeleton from the ␤-subunit? Treatment of cells with 12G10 perturbed the formation of stress fibers and resulted in the disassembly of existing stress fibers and focal adhesions. Because these effects are reminiscent of the effects of the Rhofamily of GTPases on cell morphology (41), we speculated that 12G10 might modulate the activity of this family of proteins. However, constitutively active or dominant negative Rho-family GTPases did not accentuate or rescue the effects of 12G10 observed in these assays, and the levels of GTP-bound Rho GTPases were also not affected by 12G10 (data not shown). It therefore seems unlikely that the mechanism of 12G10 action is a direct one on Rho GTPase signaling.
Recently cAMP-dependent protein kinase (PKA) was found to be required for the 12G10-induced cell-cell and cell-substrate adhesion of HT1080 cells (42). This study implied a role for PKA in ␤1 integrin activation-dependent signaling. We found that inhibition of PKA with myristoylated protein kinase inhibitor peptide did not modulate the effects of 12G10 on ␣4␤1or ␣5␤1-mediated cell attachment (data not shown). We can therefore discount PKA as a downstream target of the 12G10 signaling effects we observe. Inhibition of other signaling molecules such as MEK, phosphatidylinositol 3-kinase, and tyrosine kinase also had little or no effect on the 12G10 modulation of ␣4␤1. Tyrosine phosphorylation plays a central role in integrin signaling, and therefore its role in this phenomenon was assessed. 12G10 treatment of A375-SM cell spreading was found to reduce the overall tyrosine phosphorylation pattern for ␣4␤1 but not ␣5␤1 spread cells (data not shown). This tyrosine phosphorylation reduction, however, paralleled the reduction in A375-SM cell spreading and was therefore likely to be downstream of the effects on cell spreading. This close correlation of tyrosine phosphorylation with the 12G10 effect suggests that 12G10 is modulating a direct link to cytoskeleton.
An alternative signaling mechanism for ␣4␤1 could be via a signaling co-receptor, analogous to the way that syndecan-4 co-operates with ␣5␤1 to activate protein kinase C␣ (21,43). Previously, mutations in the extracellular ␤-propeller domain of the ␣4-subunit displayed defects in ␣4␤1-dependent static cell adhesion and adhesion under shear flow, but not to soluble ligand binding, similar to the data reported here (44). These ␤-propeller mutations were found to be defective in their interactions with CD81, a transmembrane-4 superfamily member protein that associates laterally with ␣4␤1 (45), and modulates co-operative signaling events with ␣4␤1 (46,47). The possibility that modulation of ␣4␤1-CD81 association by 12G10 could be responsible for the effects of the mAb on ␣4␤1 was discounted, however, because the ␣4P␣5L integrin chimera did not associate with CD81, unlike wild type ␣4␤1, as determined by co-immunoprecipitation experiments (data not shown).
In summary, the mAb 12G10 has been shown to stimulate ligand binding to both ␣4␤1 and ␣5␤1, reporting and stabilizing FIG. 7. Schematic diagram of the domain structure of an integrin and ␣4/␣5 chimeras. A, the domain structure of an integrin ␣ (gray)/␤ (black) heterodimer is shown as indicated by Refs. 7-9. The plasma membrane is depicted as a white rectangle. The ␤A domain contains the epitopes for a number of function modulating mAbs including 12G10 and TS2/16. In B-D schematic representations of the ␣4/␣5 chimeras expressed in K562 cells are shown in which the ␣4-subunit domains are gray and the ␣5-subunit domains are white. B, schematic of the chimera in which the ␣4 tail has been swapped for the ␣5 tail; C, schematic of the ␣4P␣5L chimera, which comprises the propeller domain of ␣4 (Tyr 1 -Leu 440 ) and the ␣5 leg, transmembrane, and cytoplasmic domains (Thr 459 -Ala 1008 ); D, schematic of the X4C0 chimera (36) in which the ␣4 tail was truncated after the GFFKR motif.
conformational changes within these integrins that correlate with their activation states. The 12G10-induced conformers, which likely reflect physiologically relevant agonistic states, differentially modulate ␣4␤1 and ␣5␤1 adhesive functions apparently via a severing of the connection with the cytoskeleton. Other activating mAbs, such as TS2/16, the epitope of which overlaps with that of 12G10, do not selectively modulate ␣4␤1 and ␣5␤1 function, illustrating the specific and novel nature of the 12G10-induced effects. Furthermore, these data indicate that the extracellular ␤-propeller domain of the ␣-subunit dictates the agonistic differences between ␣4␤1 and ␣5␤1. The 12G10-induced inhibition of cell adhesion is controlled independently of the ␣4 thigh, stalk, and tail domains suggesting that the conformation of the ␤1-subunit can be regulated by the associating ␣-subunit. Because the effects of 12G10 on ␣4␤1 are independent of the non-ligand-binding domains of the ␣4-subunit, these data indicate that signals can be propagated through the ␤-subunit to adjust cytoskeletal attachments independently from the ␣-subunit. Thus, these findings demon-strate a direct and variable agonistic link between the ligand binding pocket of integrins and the cell interior that is independent of the ␣ cytoplasmic domains and suggest that regulated signaling via changes in integrin extracellular domain conformation is more important than previously suspected.