RGD-independent Binding of Integrin α9β1 to the ADAM-12 and -15 Disintegrin Domains Mediates Cell-Cell Interaction*

ADAMs (a disintegrinand metalloproteases) mediate several important processes (e.g. tumor necrosis factor-α release, fertilization, and myoblast fusion). The ADAM disintegrin domains generally lack RGD motifs, and their receptors are virtually unknown. Here we show that integrin α9β1specifically interacts with the recombinant ADAMs-12 and -15 disintegrin domains in an RGD-independent manner. We also show that interaction between ADAM-12 or -15 and α9β1supports cell-cell interaction. Interestingly, the cation requirement and integrin activation status required for α9β1/ADAM-mediated cell adhesion and cell-cell interaction is similar to those required for known integrin-extracellular matrix interaction. These results are quite different from recent reports that ADAM-2/α6β1 interaction during sperm/egg fusion requires an integrin activation status distinct from that for extracellular matrix interaction. These results suggest that α9β1 may be a major receptor for ADAMs that lack RGD motifs, and that, considering a wide distribution of ADAMs and α9β1, this interaction may be of potential biological and pathological significance.

The disintegrin domain of ADAMs is a potential integrin ligand, but generally lacks the RGD motif (unlike RGD-containing snake venom disintegrins): Human ADAM-15 is the only ADAM that has an RGD motif in its disintegrin-like domain (14). Integrin ␣ 6 ␤ 1 has been reported to interact with the disintegrin domains of fertilin ␣ and ␤ (ADAM-1 and -2) complex that has no RGD motif during fertilization (2). Integrin ␣ v ␤ 3 has been reported to specifically bind to the disintegrin domain of human ADAM-15 in an RGD-dependent manner (15,16). Interestingly, mouse ADAM-15 (mADAM-15) has a TDD sequence instead of RGD (17). This raised doubts about the role of ADAM-15 as a genuine integrin ligand. ADAM-15 (metargidin) is widely expressed in various tissues and cell types (14,17) and is implicated in atherosclerogenesis, since ADAM-15 is over-expressed in atherosclerotic legions (18). ADAM-12 has been implicated in myoblast fusion during myogenesis (5,6), and has a catalytically active metalloprotease domain and a non-RGD disintegrin domain. The cysteine-rich domain of ADAM-12 has a putative fusion peptide and a short hydrophobic stretch (19,20). The truncated mouse ADAM-12, which lacks the metalloprotease domain, enhances fusion of C2C12 myoblastic cells in vitro (5). These findings suggest that the disintegrin and/or cysteine-rich domains of ADAM-12 should be involved in cell-cell interaction during myoblast fusion. Since the ADAM-12 gene is activated in condensed mesenchymal cells that give rise to skeletal muscle, bones, and visceral organs (21), ADAM-12 may be involved in development of other organs as well.
A major question is whether the non-RGD disintegrin domains of ADAMs interact with integrins. In the present study, we designed experiments to address this question using recombinant disintegrin domain fragment and cells expressing recombinant ADAMs. We demonstrated a novel interaction between integrins and ADAMs that is RGD-independent and may play crucial roles in cell-cell interaction during development and in pathological conditions.

MATERIALS AND METHODS
Monoclonal Antibodies and Cell Lines-Hybridomas for antibodies TS2/16 (anti-␤ 1 , activating) and AIIB2 (anti-␤ 1 , function-blocking) were obtained from American Type Culture Collection. Chinese hamster ovary (CHO) cells expressing different integrins have been described (15). Ntera-2 human embryonal carcinoma cells were provided by Amos Baruch (Scripps Research Institute, La Jolla, CA). G361 human melanoma cells were obtained from American Type Culture Collection.
Preparation of K562 Cells Expressing Recombinant ␣ 9 ␤ 1 -␣ 9 cDNA in expression vector (22) was transfected into K562 cells. After selection with G-418 (1 mg/ml of medium), cells stably expressing human integrins were cloned by limited dilution and designated ␣ 9 -K562. ␣ 9 -K562 cells express ␣ 9 (as ␣ 9 ␤ 1 ) as detected by flow cytometry in FACScan (Becton Dickinson, San Jose, CA) with the anti-␣ 9 antibody Y9A2 (23) (Fig. 4). Protein of the Disintegrin-like Domain  of Mouse ADAM-15, and Human and Mouse ADAM-12-A cDNA fragment of about 270 nucleotides that encodes the disintegrin-like domain  of mouse ADAM-15 (Met-420 to Glu-510) (17) was amplified by polymerase chain reaction (PCR) with a mouse expressed sequence tag clone (AI317315) as a template using 5Ј-CGGGATCCATGGCTGCTTTCT-GCGG and 5Ј-CGGGATCCTTACTCGCCATCCCCTAGGCTG as primers. A cDNA fragment of 264 nucleotides that encodes the disintegrinlike domain of human ADAM-12 (Gly-423 to His-513) was amplified by PCR with a human placenta cDNA library (Invitrogen) as a template using 5Ј-CGGGATCCGGGGGCCAGAAGTGTGGAAACAG and 5Ј-CG-GAATTCTTAGTGCCCATCGTGCAGGTACACG as primers. A cDNA fragment of 270 nucleotides that encodes the disintegrin-like domain of mouse ADAM-12 (Gly-420 to His-510) was amplified by PCR with a full-length mouse ADAM-12 cDNA (5) as a template using 5Ј-CGG-GATCCGGGGGCCGGAAGTGTGGAAATG and 5Ј-CGGAATTCT-TAGTGGCCATCATGTAGGTACAC as primers. These cDNA fragments were subcloned into the BamHI site of the modified pGEX-2T vector (Amersham Pharmacia Biotech), in which a 6-His sequence was inserted between the thrombin cleavage site and the BamHI site. Synthesis of the GST fusion protein of the ADAM-15 disintegrin-like domain was induced in Escherichia coli DH5␣ by adding 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside in culture medium as described previously (24). Protein was extracted from the bacterial suspension by sonication and purified using glutathione-agarose (Sigma) affinity chromatography. The recombinant wild type fusion proteins used in this study are shown in Fig. 1A.

Preparation of the GST Fusion
Protein concentration was calculated from absorbance at 280 nm. We calculated the extinction coefficients based on the amino acid sequence using the ExPASy Protparam tool available from the ExPASy web site (25).
Removal of GST from GST Fusion Protein-The GST fusion protein has a 6-His sequence inserted between its thrombin cleavage site and disintegrin domain. The GST fusion protein was digested with thrombin (Sigma; 1 unit/1 mg of GST fusion protein) at room temperature for 6 h. The GST portion was removed by passing the digested materials through a GSH-agarose column. The disintegrin fragment was further purified by nickel-nitriloacetic acid affinity chromatography. The disintegrin domain was eluted with 250 mM imidazole. Imidazole was removed using a G10 column (Amersham Pharmacia Biotech).
Adhesion Assays-Adhesion assays were performed as described previously (15). Briefly, 96-well Immulon-2 microtiter plates (Dynatech Laboratories, Chantilly, VA) were coated with 100 l of PBS (10 mM phosphate buffer, 0.15 M NaCl, pH 7.4) containing substrates at a concentration of 20 g/ml and were incubated overnight at 4°C. Remaining protein binding sites were blocked by incubating with 1% bovine serum albumin (BSA; Calbiochem) for 1 h at room temperature. After washing with PBS, CHO cells (10 5 cells/well) in 100 l of Dulbecco's modified Eagle medium supplemented with 1% BSA were added to the wells and incubated at 37°C for 1 h. After unbound cells were removed by rinsing the wells with PBS, bound cells were quantified by measuring endogenous phosphatase activity (26). GST fusion protein of the 8 -11th type III repeats of fibronectin (GST-FN) (15) was used as a positive control, since untransfected CHO cells express endogenous hamster ␣ 5 ␤ 1 (27,28). Wild type GST (wt-GST) was used as a negative control. A function-blocking anti-human ␣ 9 monoclonal antibody, Y9A2, was used at a concentration of 10 g/ml. Hepes-Tyrode buffer (10 mM HEPES, 150 mM NaCl, 12 mM NaHCO 3 , 0.4 mM NaH 2 PO 4 , 2.5 mM KCl, 0.1% glucose, 0.02% BSA) supplemented with 1 mM EDTA, 2 mM Ca 2ϩ , 2 mM Mg 2ϩ , or 1 mM Mn 2ϩ was used instead of Dulbecco's modified Eagle medium for experiments to determine cation dependence.
Transient Expression of ADAMs Using the Bicistronic Expression System-We used the bicistronic expression system with an internal ribosome entry sequence (29,30), pIRES2-EGFP (CLONTECH, Palo Alto, CA) to express human ADAM-15 or mouse ADAM-12 on the cell surface. The expression of enhanced green fluorescent protein (EGFP) was used as a marker for ADAM protein expression. Full-length human ADAM-15 cDNA, and a truncated human ADAM-15 cDNA fragment encoding residues 420 -814 lacking pro-and metalloprotease domains (ADAM-15/PMϪ), were obtained by PCR amplification with a placenta cDNA library (Invitrogen, Carlsbad, CA) as a template. These cDNA contain a signal sequence (of ADAM-15) and a Myc tag sequence just after the signal sequence. These cDNAs were verified by DNA sequencing, and subcloned into the NheI site of the pIRES2-EGFP vector. Full-length mouse ADAM-12 cDNA (5) was also subcloned into the SalI/BamHI site of this vector.
Cell-Cell Binding Assay-The pIRES-EGFP vector constructs with or without ADAM cDNA (full-length or PMϪ) were transfected into CHO cells by electroporation as described (15). After 48 h, EGFP expression was determined by flow cytometry. 96-well plastic culture plates (Corning) were plated with CHO cells transiently expressing EGFP and ADAM protein (5 ϫ 10 4 /well), and cultured overnight. K562 or ␣ 9 -K562 cells were labeled with 2Ј,7Ј-bis-(2-carboxyethyl)-5-(and-6)-carboxyfluorescein, acetoxymethyl ester (Molecular Probes, Eugene, OR) according to the manufacturer's instructions. The labeled K562 cells (10 4 /well) were added to the CHO cell monolayer expressing EGFP and ADAM protein and incubated for 90 min at 37°C in RPMI 1640. After rinsing the wells with RPMI 1640 three times to remove unbound cells, bound cells were quantified by assaying fluorescence (excitation 485 nm, emission 530 nm) using a FL500 microplate fluorescence reader (Bio Tek Instruments, Winooski, VT). In soluble disintegrin competition experiments, soluble bacterial disintegrin domains were first added to labeled ␣ 9 -K562 cells in RPMI 1640 medium and incubated for 1 h at 37°C (the volume of the soluble disintegrin was 10% of total). ␣ 9 -K562 cells were then incubated with a monolayer of CHO cells expressing ADAM-12 or -15 for 90 min at 37°C. In some experiments, Hepes-Tyrode buffer containing 2 mM Ca 2ϩ , 2 mM Mg 2ϩ , or 1 mM Mn 2ϩ was used instead of RPMI 1640 to determine the ion dependence of the cell to cell interaction.
Other Methods-Site-directed mutagenesis was carried out using the unique site elimination method (31). The presence of mutations was verified by DNA sequencing. Flow cytometry was performed as described previously (27).

Adhesion of Integrin ␣ 9 ␤ 1 to the Mouse ADAM-15 Disintegrin Domain in an RGD-independent Manner-We have previously
shown that the recombinant disintegrin-like domain of human ADAM-15 binds to ␣ v ␤ 3 in an RGD-dependent manner (15). However, a mouse ADAM-15 homologue has recently been reported to have a TDD sequence instead of an RGD sequence in the putative integrin-binding site of its disintegrin domain (17). We generated a recombinant mouse ADAM-15 disintegrin fragment (Fig. 1A) and studied whether it supports ␣ v ␤ 3 -mediated cell adhesion. We did not detect significant adhesion of ␤ 3 -CHO cells that express hamster ␣ v /human ␤ 3 hybrid to the mouse ADAM-15 disintegrin domain (Fig. 1B).
Adhesion to the human and mouse ADAM-15 disintegrin domains was determined as a function of substrate concentration (Fig. 1C). ␣ 9 -CHO cells showed maximum adhesion to both human and mouse ADAM-15 disintegrin domains at the coating concentration of 20 g/ml protein. ␤ 3 -CHO cells showed maximal adhesion to human ADAM-15 disintegrin domain at 5 g/ml coating concentration, but never showed significant adhesion to mouse ADAM-15 disintegrin domain at the highest coating concentration used (50 g/ml). These results are consistent with the previous results that ␣ v ␤ 3 recognizes ADAM-15 in an RGD-dependent manner (15). We studied whether mutation of the RGD motif in the putative integrinbinding site of the ADAM-15 disintegrin domain affects adhesion to ␣ 9 ␤ 1 . ␣ 9 -CHO cells adhered to the two human ADAM-15 disintegrin domain mutants, human/SGA (in which the RGD sequence is mutated to SGA (Ref. 15)) and human/TDD (in which the RGD sequence is mutated to TDD), as well as to wild-type human and mouse ADAM-15 disintegrin domains. These results suggest that ␣ 9 ␤ 1 recognizes ADAM-15 in an RGD-independent manner. Adhesion of ␣ 9 -CHO cells to these recombinant proteins was completely blocked by Y9A2, an antihuman ␣ 9 integrin monoclonal antibody (23) (Fig. 1D), suggesting that this adhesion is specific to ␣ 9 ␤ 1 .
These results suggest that ␣ 9 ␤ 1 may bind to other non-RGD ADAMs. We studied whether another non-RGD ADAM, human and mouse ADAM-12, or meltrin-␣, which lacks the RGD motif, binds to ␣ 9 ␤ 1 using CHO cells expressing different recombinant integrins. We found that the human and mouse ADAM-12 disintegrin domains supported adhesion of ␣ 9 -CHO cells, but not of parent CHO cells or cells expressing the other integrins we tested, except for ␣ 4 ␤ 1 and ␣ v ␤ 3 , which weakly bound to this protein ( Fig. 2A). The ␣ 9 -CHO cells showed maximum adhesion to both the human and mouse ADAM-12 disintegrin domains at a coating concentration of 20 g of protein/ml (Fig. 2B). This interaction is also blocked by the anti-␣ 9 mAb Y9A2 (Fig. 2C), suggesting that ADAM-12 also specifically interacts with ␣ 9 ␤ 1 .
We studied whether non-recombinant ␣ 9 ␤ 1 binds to ADAM-12 and -15 using Ntera-2 human embryonic carcinoma cells that express ␣ 9 and ␤ 1 on the surface (Fig. 3A). Ntera-2 cells adhered to ADAM-12 and -15 in an ␣ 9 ␤ 1 -dependent manner, but required activation by activating anti-␤ 1 mAb TS2/16 (Fig. 3B). We obtained very similar results with G361 human melanoma cells that also express ␣ 9 ␤ 1 (32); G361 cells adhere to ADAM-12 and -15 in an ␣ 9 -dependent and activation-dependent manner (data not shown). These results suggest that . Cells homogeneously expressing different human integrins were added to wells and incubated at 37°C for 1 h. After rinsing the wells to remove unbound cells, bound cells were quantified using endogenous phosphatase activity. Data are shown as means Ϯ S.D. of triplicate experiments. The data suggest that ␣ 9 ␤ 1 adheres to the mouse ADAM-15 disintegrin domain, but the other integrins tested do not. C, adhesion of ␤ 3 -CHO and ␣ 9 -CHO cells to human and mouse ADAM-15 disintegrin domain as a function of substrate concentrations. Adhesion of ␤ 3 -CHO and ␣ 9 -CHO cells to human and mouse ADAM-15 disintegrin domains was determined as a function of the substrate concentrations. Data are shown as means Ϯ S.D. of triplicate experiments. The data suggest that the affinity of human ADAM-15 for ␣ 9 ␤ 1 (closed circle) is comparable to that of mouse ADAM-15, and that the affinity of the human ADAM-15 disintegrin domain for ␣ v ␤ 3 (open triangle) is comparable to or slightly higher than that for ␣ 9 ␤ 1 . Parent CHO cells (open square). D, effect of mutating the RGD motif of the ADAM-15 disintegrin domain on the adhesion of ␣ 9 -CHO cells. The RGD motif of the human ADAM-15 disintegrin domain was mutated to SGA or TDD by site-directed mutagenesis. GST fusion proteins of these mutants were used for adhesion assays with ␣ 9 -CHO cells. Wells of a 96-well titer plate were coated with proteins at 20 g/ml. An anti-␣ 9 function-blocking antibody (Y9A2) was used at 10 g/ml. Data are shown as means Ϯ S.D. of triplicate experiments. The data suggest that ␣ 9 ␤ 1 binding to ADAM-15 is not dependent on the RGD motif at the putative integrin-binding site.

FIG. 4. ␣ 9 ␤ 1 and ADAMs mediate cell-cell interaction.
A, expression of ␣ 9 ␤ 1 on K562 cells. Parent K562 cells or ␣ 9 -K562 cells were stained with control mouse IgG (dotted line) or Y9A2 (anti-␣ 9 mAb) (solid line), followed by fluorescein isothiocyanate-labeled goat anti-mouse IgG. Stained cells were analyzed by flow cytometry. The data suggest that ␣ 9 -K562 cells homogeneously express ␣ 9 ␤ 1 . B, expression of EGFP or EGFP containing ADAMs on CHO cells. The IRES2-EGFP bicistronic vector, in which both EGFP and ADAMs are encoded in a single mRNA with two separate translation starting sites, was used. EGFP vector alone, EGFP vector containing the full-length human ADAM-15 cDNA, the truncated human ADAM-15 cDNA fragment encoding residues 420 -814 lacking pro-and metalloprotease domains (ADAM-15/PMϪ), or the full-length mouse ADAM-12 cDNA (solid line) was transfected into CHO cells. These cells and mock-transfected (thin line) CHO cells were analyzed 48 h after transfection. Expression was analyzed by flow cytometry (excitation at 488 nm and emission at 507 nm). Transient EGFP expression was used as a marker of ADAM expression, and typically approximately 30% of transfected cells are EGFP-positive. C, ␣ 9 ␤ 1 and ADAM-15 or ADAM-12 mediate cell-cell interaction. CHO cells transiently expressing ADAMs were plated in wells of 96-well plastic culture plates, and grown overnight to near confluence. Fluorescence-labeled ␣ 9 -K562 or parent K562 cells were added to the wells and incubated with CHO cells in RPMI 1640 medium. After rinsing the wells to remove unbound cells, bound cells were quantified using a fluorescent plate reader at 485 nm for excitation and at 530 nm for emission. Cell-cell interaction was assayed in the absence (closed bar) or presence (open bar) of the anti-␣ 9 mAb, Y9A2 (10 g/ml). Note that only approximately 30% of the transfected CHO cells is EGFP positive. Thus, binding of 15-20% of added ␣ 9 -K562 cells is substantial. The data suggest that ADAM-15 (wt and truncated) and ADAM-12 on the cell-surface specifically interact with ␣ 9 ␤ 1 on the surface of apposing cells and mediate cell-cell interaction. D, binding of ␣ 9 -K562 cells to CHO cells expressing ADAM-12 or -15. A, a monolayer of CHO cells expressing ADAM-15 before adding ␣ 9 -K562 cells; B, ␣ 9 -K562 cells on CHO cells expressing EGFP only; C, ␣ 9 -K562 cells on CHO cells expressing ADAM-15; D, ␣ 9 -K562 cells bound to CHO cells expressing only EGFP after rinsing the well to remove unbound cells; E, ␣ 9 -K562 cells bound to CHO cells expressing ADAM-15; F, ␣ 9 -K562 cells bound to CHO cells expressing ADAM-15 in the presence of Y9A2. Note that only K562 cells bound to CHO cells after rinsing the wells to remove unbound cells are shown in D-F. CHO cells that express EGFP but no ADAM (Fig. 4C). ␣ 9 -K562 cells bound to cells expressing these ADAMs at significantly higher (approximately 3 times) levels than control parent K562 cells (Fig. 4, C and D). Levels of binding of ␣ 9 -K562 cells to CHO cells expressing ADAM-12 or ADAM-15 is low (15-20% of added cells) compared with those in other adhesion studies in this study. However, this binding is substantial considering that only approximately 30% of added cells express EGFP and ADAM-12 or -15. Parent K562 cells bound to CHO cells expressing ADAM-15 or -12 at the background levels. Binding of ␣ 9 -K562 cells to CHO cells expressing ADAM-15 (full-length and truncated) or ADAM-12 was completely blocked by anti-␣ 9 antibody. These results suggest that this interaction is specific to ␣ 9 ␤ 1 , and to ADAM-15 or ADAM-12.
We studied whether the soluble ADAM disintegrin domains affect ADAM/␣ 9 ␤ 1 -mediated cell-cell interaction. Binding of ␣ 9 -K562 cells to CHO cells transiently expressing ADAM-12 or -15 was determined in the presence of soluble ADAM-12 or -15 disintegrin domain in the assay medium (Fig. 5) These results suggest that ADAM disintegrin domains synthesized in bacteria effectively compete with those synthesized in mammalian cells for binding to ␣ 9 ␤ 1 , and that, consistent with the adhesion and cell-cell interaction results, ADAM-15 disintegrin domain has a higher binding affinity to ␣ 9 ␤ 1 than the ADAM-12 disintegrin domain.
We determined the cation requirement for adhesion of ␣ 9 -CHO cells to the recombinant disintegrin domain of human ADAM-15 (Fig. 6). We removed the GST portion of the fusion protein, since it generates high background binding to integrins in the presence of Mg 2ϩ . We found that Mg 2ϩ and Mn 2ϩ stimulated, but Ca 2ϩ suppressed, ␣ 9 ␤ 1 adhesion to the ADAM-15 disintegrin domain. Ca 2ϩ supported ␣ v ␤ 3 binding to ADAM-15. We also examined the cation requirement for ␣ 9 ␤ 1 / ADAM-15-mediated cell-cell interaction (Fig. 7). This interaction is higher in the presence of Mn 2ϩ and Mg 2ϩ than in the presence of Ca 2ϩ . We obtained essentially the same results with ␣ 9 ␤ 1 /ADAM-12-mediated cell-cell interaction (Fig. 7). These results suggest that the cation requirements for ␣ 9 ␤ 1 adhesion to the ADAM-15 disintegrin domain, and for ␣ 9 ␤ 1 / ADAM-15-or ADAM-12-mediated cell-cell interaction, are not different from known integrin-extracellular matrix interactions, but are different from the reported ADAM-2/␣ 6 ␤ 1 -mediated cell-cell interaction (4).
Effect of Activating and Blocking Antibodies on ␣ 9 ␤ 1 /ADAM-15, or ADAM-12-mediated Cell-Cell Interaction-Although the cation requirement for ␣ 9 ␤ 1 -ADAM protein-mediated cell-cell interaction is not different from that for known integrin extracellular matrix interactions, it is still possible that a different activation status of ␣ 9 ␤ 1 is required for this interaction. To address this question, we studied the effect of the activating anti-␤ 1 antibody TS2/16, and the inhibitory anti-␤ 1 antibody AIIB2, on this interaction. These mAbs stimulate or block ␤ 1 integrin binding to many extracellular matrix proteins (e.g. fibronectin) (reviewed in Ref. 33). We have previously reported that the epitopes for these antibodies overlap (residues 207-218 of ␤ 1 ) (33). These mAbs probably induce high or low affinity states by binding to the non-ligand binding site of ␤ 1 , thereby changing its conformation. We found that TS2/16 significantly increased the binding of ␣ 9 -K562 cells to CHO cells expressing ADAM-15, but AIIB2 completely blocked ADAM-15-mediated cell-cell interaction (Fig. 8A). We obtained essentially the same results with ADAM-12 (Fig. 8B). Binding of ␣ 9 -K562 cells to CHO cells expressing ADAM-15 (10 -20% of added cells) is substantial considering that only approximately 30% of added cells express EGFP and ADAM-15. These results again suggest that the activation status of ␤ 1 integrin that is required for ADAM-mediated cell-cell interaction is very similar to that required for adhesion of ␤ 1 integrins to extracellular matrix proteins. DISCUSSION A Novel Link between ADAMs and Integrin ␣ 9 ␤ 1 -It has not been established whether most of the ADAM disintegrin domains interact with integrins, since they generally lack the RGD motif in their putative integrin binding site. The present study establishes that ADAM disintegrin domains that lack an RGD motif (mouse ADAM-15, human ADAM-15 mutants, and human and mouse ADAM-12) support cell adhesion to ␣ 9 ␤ 1 in an RGD-independent manner. Thus ␣ 9 ␤ 1 may be a major receptor for ADAMs that lack RGD motifs. The human ADAM-15 disintegrin domain, which has the RGD motif, binds to ␣ v ␤ 3 in an RGD-dependent manner (15), and to ␣ 9 ␤ 1 in an RGD-independent manner. However, its interaction with ␣ 9 ␤ 1 appears to be physiologically more important, since this interaction is evolutionary conserved, while the RGD motif of ADAM-15 is not conserved.

FIG. 5. Inhibition of interaction between membrane-bound ADAM and
The present study for the first time demonstrated that ADAM-15 and ADAM-12 binding to ␣ 9 ␤ 1 mediates cell-cell interaction. ␣ 9 ␤ 1 is distributed in tissues including airway epithelia, the basal layer of squamous epithelia, smooth muscle, skeletal muscle, hepatocytes, neutrophils, and monocytes (34 -36). ␣ 9 ␤ 1 has been reported to recognize tenascin C type III repeat (22,37), vascular cell adhesion molecule-1 (36), and osteopontin (38) in an RGD-independent manner. It has been proposed that ␣ 9 ␤ 1 -mediated binding of neutrophils to endothelial cells may be involved in chemotaxis across activated endothelial monolayers by interacting with the endothelial ligand vascular cell adhesion molecule-1 (36) during inflammation. Thus, integrin ␣ 9 ␤ 1 may have broad ligand specificity similar to integrins of the ␣ v family. Considering a wide distribution of ADAMs and ␣ 9 ␤ 1 , ADAM/␣ 9 ␤ 1 -mediated cell-cell interaction may be involved in many developmental and pathological situations, including myoblast fusion, fertilization, and vascular and cartilage remodeling.
Human ADAM-15 potentially participates in vascular remodeling such as in atheroscrelosis, since its protein level is increased in the core of atherosclerotic lesions and in intimal The data suggest that the cation dependence of cell-cell interaction between ␣ 9 ␤ 1 and ADAM-15 or -12 is similar to that for known integrinextracellular matrix interactions.
FIG. 8. Effect of activating and blocking antibodies on ␣ 9 ␤ 1 / ADAM-15-mediated (A) or ␣ 9 ␤ 1 /ADAM-12-mediated (B) cell-cell interaction. CHO cells transiently expressing ADAMs (closed columns) were plated in wells of 96-well plastic culture plates, and grown overnight to near confluence. Fluorescence-labeled ␣ 9 -K562 or parent K562 cells were added to the wells and incubated with CHO cells in the absence or presence of the activating anti-␤ 1 antibody TS2/16 at 1.0 g/ml or the inhibiting anti-␤ 1 antibody AIIB2 (at 6.5g/ml). After rinsing the wells to remove unbound cells, bound cells were quantified using a fluorescent plate reader. Data are shown as means Ϯ S.D. of triplicate experiments. The data suggest that the activation status of ␣ 9 ␤ 1 that is required for cell-cell interaction with ADAM-15 or ADAM-12 is similar to that required for known integrin extracellular matrix interactions. Open columns, CHO cells expressing EGFP alone. cells close to the lumen, but not in normal vessel (18). In addition, the mRNA level of ADAM-15 is up-regulated in human osteoarthritic cartilage and neoplastic cartilage (chondrosarcoma), and therefore ADAM-15 has a potential role in cartilage remodeling (39). ADAM-15 has a catalytically active metalloprotease domain with a metalloprotease catalytic site consensus sequence (HEXXH)- (17). Since ␣ 9 ␤ 1 is highly and uniformly expressed in neutrophils, and weakly expressed in monocytes (34), ␣ 9 ␤ 1 /ADAM-15-mediated cell-cell interaction may be critically involved in recruitment of these cells to inflammatory sites, and subsequent vessel or tissue damage (40). ADAM-15/␣ 9 ␤ 1 and ADAM-15/␣ v ␤ 3 interaction (15) may play a crucial role during atherosclerosis and cartilage remodeling.
ADAM-12 has been implicated in myoblast fusion during myogenesis (5,6). The cysteine-rich domain of ADAM-12 has a putative fusion peptide and a short hydrophobic stretch (19,20). The metalloprotease domain-less mouse ADAM-12, not the full-length ADAM-12, induces fusion of C2C12 myoblastic cells in vitro (5). The myogenic activity of the processed ADAM-12 was also demonstrated in tumor cells expressing a secreted form of human ADAM-12 that has only disintegrin and cysteine-rich domains (6). The function of the metalloprotease domain during myoblast fusion is not established, although ADAM-12 has a catalytically active metalloprotease domain. Integrin ␣ 9 ␤ 1 is highly expressed in skeletal and smooth muscle cells (34), and therefore ADAM-12/␣ 9 ␤ 1 interaction in vivo through the disintegrin domain may be involved in myogenesis.
Recently Bigler et al. reports that GST-ADAM-2 disintegrin domain fusion protein expressed in bacteria effectively blocks sperm-egg binding and fusion (41). We have shown in the present paper that ␣ 9 ␤ 1 interacts both with GST-ADAM-12, or ADAM-15 disintegrin domain fusion protein that are expressed in bacteria and with ADAM-12 and -15 that are expressed on mammalian cells. Furthermore, we have shown that GST-ADAM disintegrin domains effectively competed with ADAMs on mammalian cells for binding to ␣ 9 ␤ 1 . These results suggest that the ADAM disintegrin domains expressed in bacteria are similar in integrin binding function to those expressed in mammalian cells, although the bacterial disintegrin domains lack glycosylation. Additionally, these results suggest that the disintegrin domains primarily mediate interaction with integrins, although we do not rule out the possibility that other domains (e.g. Cys-rich domain) may be involved in this interaction (42).
We have shown that Ntera-2 embryonic carcinoma cells and G361 melanoma cells expressing native ␣ 9 ␤ 1 adhere to ADAM-12 and -15 in an ␣ 9 ␤ 1 -dependent manner. Considering the wide distribution of ADAMs, it is possible that ␣ 9 ␤ 1 /ADAM interaction may be involved in cell-cell interaction during cancer metastasis. It is also possible that this interaction mediates transduction of proliferative signals through cell-cell interaction in tumor mass. Since ␣ 9 ␤ 1 on these cells requires activation for binding to ADAM disintegrin domains, it is likely that ␣ 9 ␤ 1 /ADAM interaction may be regulated by ␣ 9 ␤ 1 activation in these cells, unlike ␣ 9 ␤ 1 on CHO cells, which appears to be constitutively active.
Activation Status of Integrins during ADAM/Integrin-mediated Cell-Cell Interaction-Integrin ␣ 6 ␤ 1 , on mouse eggs and on ␣ 6 -transfected cells, has been reported to interact with the disintegrin domain of the sperm surface protein ADAM-2 (fertilin ␤) (2). Additionally, the activation status of ␣ 6 ␤ 1 for ADAM-2 binding has been reported to be different from that for laminin (4). Thus, a major question is whether ADAM-15-or -12-mediated cell-cell interaction is different from known integrin/extracellular matrix interactions. We have shown that ADAM-15-or ADAM-12/␣ 9 ␤ 1 -mediated cell-cell interaction is similar in cation requirement to ␣ 9 ␤ 1 adhesion to ADAM-15 or ADAM-12, and to their interaction with known integrin-extracellular matrix proteins (e.g. fibronectin). Consistently, both ADAM-15/␣ 9 ␤ 1 and ADAM-12/␣ 9 ␤ 1 interactions are stimulated by the activating anti-␤ 1 mAb TS2/16, and blocked by the function-blocking anti-␤ 1 mAb AIIB2. These results establish that the ␤ 1 integrin activation status that is required for ADAM-15 or ADAM-12 interaction with ␣ 9 ␤ 1 is similar to that required for the interaction between ␤ 1 integrins and extracellular matrix ligands. Thus, our results do not fit in well with the reported ADAM-2/egg interaction (4). Further studies will be required to resolve this apparent discrepancy.
Very recently, Miller et al. (43) reported that the ␣ 6 ␤ 1 is not essential for sperm-egg fusion using eggs from ␣ 6 null mice and proposed that ␤ 1 integrins other than ␣ 6 ␤ 1 might be involved in this process. In our preliminary experiments, we found that ␣ 9 ␤ 1 specifically binds to the disintegrin domain of ADAM-2 synthesized in bacteria, 2 suggesting that ␣ 9 ␤ 1 is a likely candidate integrin that interacts with ADAM-2 during sperm-egg binding and fusion. This is consistent with the idea that ␣ 9 ␤ 1 is a predominant receptor for non-RGD disintegrin domains of ADAMs. Further studies will be required to establish whether ␣ 9 ␤ 1 is really involved in sperm-egg binding and fusion.