Cooperation of the Metalloprotease, Disintegrin, and Cysteine-rich Domains of ADAM12 during Inhibition of Myogenic Differentiation*

The extracellular domain of the mature form of ADAM12 consists of the metalloprotease, disintegrin, cysteine-rich, and epidermal growth factor (EGF)-like domains. The disintegrin, cysteine-rich, and EGF-like fragments have been shown previously to support cell adhesion via activated integrins or proteoglycans. In this study, we report that the entire extracellular domain of mouse ADAM12 produced in Drosophila S2 cells supported efficient adhesion and spreading of C2C12 myoblasts even in the absence of exogenous integrin activators. This adhesion was not mediated by β1 integrins or proteoglycans, was myoblast-specific, and required the presence of both the metalloprotease and disintegrin/cysteine-rich domains of ADAM12. Analysis of the recombinant proteins by far-UV circular dichroism suggested that the secondary structures of the autonomously expressed metalloprotease domain and the disintegrin/cysteine-rich/EGF-like domains differ from the structures present in the intact extracellular domain. Furthermore, the intact extracellular domain (but not the metalloprotease domain or the disintegrin/cysteine-rich/EGF-like fragment alone) decreased the expression of the cell cycle inhibitor p21 and myogenin, two markers of differentiation, and inhibited C2C12 myoblast fusion. Thus, the novel protein-protein interaction reported here involving the extracellular domain of ADAM12 may have important biological consequences during myoblast differentiation.

Members of the ADAM family of proteins contain a region in their extracellular domains that bears similarity to the disintegrin and cysteine-rich region of class P-III snake venom metalloproteinases (1)(2)(3). The disintegrin (4,5) and cysteine-rich (6,7) domains of snake venom metalloproteinases bind to integrin receptors and interfere with the interactions between integrins and their extracellular matrix ligands. By analogy, ADAM proteins are thought to contain adhesion domains that interact with cell-surface integrins. Indeed, multiple studies have demonstrated that recombinant disintegrin and/or cysteine-rich fragments of many ADAM proteins are capable of supporting integrin-mediated cell adhesion (8 -19). In most cases, however, integrins need to be "activated," by either Mn 2ϩ ions or activating antibodies, to promote significant levels of cell adhesion (8 -15). Furthermore, although the disintegrin/ cysteine-rich domains contain multiple cysteine residues that form elaborate patterns of intramolecular disulfide bonds after passing through the secretory pathway of mammalian cells (4,20), recombinant disintegrins expressed in prokaryotic cells (in which the formation of disulfide bonds is hindered and in which the correct protein folding most likely is not achieved) often show the same ability to promote cell adhesion via activated integrins as recombinant disintegrins expressed in eukaryotic cells. The disintegrin domains of ADAM2, ADAM3, and ADAM9 (8,9), ADAM12 (10), and ADAM23 (11) produced in Escherichia coli; the disintegrin domains of ADAM7 (12), ADAM28 (12,13), and ADAM33 (12) produced in insect cells; the disintegrin/cysteine-rich domains of ADAM12 (14), ADAM13 (15), and ADAM28 (12,13) expressed in insect cells; and the entire extracellular domains of ADAM9 (16) and ADAM17 (17) expressed in mammalian cells require the presence of Mn 2ϩ or integrin-activating antibodies to support cell adhesion. The disintegrin domain of human ADAM15 is the only ADAM disintegrin that contains the RGD sequence, which is readily recognized by many integrins, including integrin ␣ v ␤ 3 . Consistently, the disintegrin domain of human ADAM15 expressed in E. coli (9,18) and the entire extracellular domain of human ADAM15 produced in mammalian cells (19) support substantial integrin ␣ v ␤ 3 -mediated cell adhesion even in the absence of Mn 2ϩ .
ADAM12 has been implicated in the development and/or regeneration of skeletal muscle (22)(23)(24)(25)(26)(27), but its mechanism of action is not clear. Although early studies utilizing clones of stably transfected C2C12 cells suggested that ADAM12 might be directly involved in cell-cell fusion during myogenic differentiation (22), recent generation of mice with targeted disruption of the Adam12 gene showed that ADAM12 is dispensable for myoblast fusion (28). Subsequently, it was demonstrated that transgenic expression of ADAM12 in myofibers of dystrophin-deficient mdx mice alleviates the symptoms of muscular dystrophy (25,26). However, endogenous ADAM12 does not appear to be highly expressed in myofibers of non-dystrophic animals, and its expression is confined mostly to activated satellite cells during muscle regeneration (23).
We have shown that, during myogenic differentiation of C2C12 cells in vitro, ADAM12 expression is down-regulated in differentiated myotubes, but is maintained in a pool of nondifferentiated "reserve" cells with a G 0 -like phenotype, characterized by increased expression of the cell cycle inhibitor p27 and the retinoblastoma-related protein p130 and down-regulation of MyoD (29). Recently, higher expression of ADAM12 in a pool of MyoD-negative cells compared with MyoD-positive cells has also been observed during differentiation of human rhabdomyosarcoma cells (30). Inhibition of ADAM12 expression by small interfering RNA is accompanied by lower expression levels of G 0 markers, whereas forced expression of ADAM12 induces a G 0 -like phenotype (29).
In this study, we re-evaluate the adhesive properties of the extracellular domain of ADAM12 and its effect on myoblast differentiation. We show that the recombinant extracellular domain of mouse ADAM12 produced in Drosophila S2 cells supports C2C12 cell adhesion in the absence of Mn 2ϩ . The presence of both the metalloprotease and disintegrin/cysteinerich domains was necessary for efficient cell adhesion. Structural analysis of the recombinant proteins by circular dichroism indicated that two complementary fragments of the extracellular domain of ADAM12, i.e. the metalloprotease domain and the disintegrin/cysteine-rich/epidermal growth factor (EGF) 1 -like region, assumed different folding when produced as autonomous proteins or as parts of the larger extracellular domain. The Mn 2ϩ -independent adhesion, which did not seem to be mediated by ␤ 1 integrins or proteoglycans, appeared to be muscle-specific and was not observed in non-myoblastic cell lines. Finally, a potential significance of the novel proteinprotein interaction reported here is validated by the fact that the entire extracellular domain (but not the metalloprotease domain or the disintegrin/cysteine-rich/EGF-like fragment alone) inhibited myoblast differentiation.

EXPERIMENTAL PROCEDURES
Expression Constructs-The cDNA fragments encoding the extracellular domain of ADAM12 (X, amino acids (aa) 32-707); the prodomain, metalloprotease domain, and disintegrin and cysteine-rich region (MDC, aa 32-657); the prodomain and metalloprotease domain (M, aa 32-421); the disintegrin, cysteine-rich, and EGF-like domains (DCE, aa 422-706); and the disintegrin and cysteine-rich region (DC, aa 422-657) were amplified by PCR using Pfu Turbo DNA polymerase and the mouse full-length Adam12 cDNA as a template. For expression in Drosophila S2 cells, the cDNA fragments were cloned into the pMT/ BiP/V5-HisA vector between the BglII and NotI sites in-frame with the BiP secretion signal present in the vector. The E349Q mutation, which abolishes the catalytic activity of the metalloprotease, was introduced into the X, MDC, and M constructs using the QuikChange site-directed mutagenesis method (Stratagene). In addition, the cDNA fragments encoding the DCE, DC, and C b E (the b subdomain of the cysteine-rich region and the EGF-like domain, aa 559 -706) constructs were cloned into pET20b vector between the NdeI and NotI sites for bacterial expression. All of the insect and bacterial expression constructs (see Fig. 1) contained a sequence encoding a His 6 tag at the C termini of the corresponding proteins.
For expression in mammalian cells, the following constructs were prepared: XT-wt (aa 1-734, corresponding to the extracellular domain, the transmembrane domain, and the first 7 amino acids of the cytoplasmic tail of ADAM12), XT (aa 1-734, containing the E349Q mutation), and ETCyt (aa 658 -903, corresponding to the EGF-like, transmembrane, and cytoplasmic domains, preceded by the Ig chain signal sequence). The cDNA fragments were cloned into the pEGFP-N3 vector (Clontech) between the KpnI and NotI sites, replacing the entire enhanced green fluorescent protein gene with the ADAM12 fragments.
Cell Culture-Drosophila S2 cells were cultured at 27°C in Schneider's Drosophila medium containing 10% heat-inactivated fetal bovine serum. C2C12 and NIH3T3 cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum. CHO-K1 cells were cultured in Ham's F-12K medium with 10% fetal bovine serum. C2C12, NIH3T3, and CHO-K1 cells were maintained at 37°C in the presence of 5% CO 2 under a humidified atmosphere.
Protein Expression and Purification-Drosophila S2 cells grown in 6-well plates were cotransfected with an expression vector and pCoblast, a selection vector, using the calcium phosphate precipitation method. Stable transfectants were selected with 20 g/ml blasticidin over a 2-week period. Protein expression was induced with 0.5 mM CuSO 4 in stably transfected S2 cells grown in four to five 175-cm 2 flasks (200 -250 ml of culture medium, 2-4 ϫ 10 6 cells/ml). Four days after induction, the culture medium was collected, filtered through a 0.22-m filter, and loaded onto a chelating Sepharose column (2-ml bed volume; Amersham Biosciences) at a rate of 5 ml/min. The column was washed first with 50 ml of Dulbecco's phosphate-buffered saline (DPBS), then with 50 ml of 0.5 M NaCl, and then with 5 ml of 50 mM Tris-HCl (pH 8.0) and 10 mM imidazole. Proteins were eluted with 7.5 ml of 50 mM Tris-HCl (pH 8.0) and 50 mM imidazole. The eluate was diluted 5-fold with 50 mM Tris-HCl (pH 8.0) and 300 mM NaCl and directly loaded onto a nickel-nitrilotriacetic acid-agarose column (0.4-ml bed volume; Qiagen Inc.) at a rate of 1.5 ml/min. After washing the column with 20 ml of 50 mM Tris-HCl (pH 8.0), 300 mM NaCl, and 10 mM imidazole, proteins were eluted with 1 ml of 50 mM Tris-HCl (pH 7.5), 300 mM NaCl, and 200 mM imidazole. The eluate was dialyzed against 0.1 M NaHCO 3 at pH 9.0 (for cell adhesion assays) or against DMEM (for cell differentiation assays). To determine the concentration of ADAM12 fragments, purified proteins were resolved by SDS-PAGE, stained with Coomassie Blue, and analyzed by densitometry using bovine serum albumin as a standard. Alternatively, protein concentration was determined by measuring absorbance at 280 nm using the molar absorptivity calculated as described (31). Protein concentrations obtained with the two methods differed by Ͻ10%.
Circular Dichroism Spectroscopy-The X, M, and DCE proteins purified from S2 cell culture medium were dialyzed against DPBS. CD spectra at equimolar concentrations of the X, M, and DCE proteins were obtained with a Jasco J-720 spectropolarimeter using a quartz cell with a 0.02-cm path length at 25°C. The spectra obtained for the M and DCE domains were added together, and the calculated spectrum was compared with the measured spectrum of the X domain to assess whether the secondary structures of the M and DCE domains were the same when the two domains were expressed autonomously and when they were parts of the larger X domain.
Cell Adhesion Assay-96-Well MaxiSorp plates (Nunc) were coated with recombinant proteins in 0.1 M NaHCO 3 (pH 9.0) for 12-16 h at 4°C and then blocked with 2.5% bovine serum albumin in DPBS at 37°C for 1-2 h. Cells were detached with 0.05% (w/v) trypsin and 1 mM EDTA in DPBS; sequentially washed with growth medium, DMEM, and Tyrode's buffer (5 mM HEPES-KOH (pH 7.4), 12 mM NaHCO 3 , 150 mM NaCl, 2.6 mM KCl, 5 mM glucose, 0.5 mM MgCl 2 , and 1 mM CaCl 2 ); and finally resuspended in Tyrode's buffer containing 1% bovine serum albumin at a density of 1 ϫ 10 6 cells/ml. One-hundred microliters of cell suspension was added to each well and incubated for 1 h at 37°C. In some experiments, cells were incubated with 1 mM MnCl 2 , integrin ␤ 1 functionblocking antibody Ha2/5 (Pharmingen), or integrin ␣ v ␤ 3 function-blocking antibody RMV-7 (Chemicon International, Inc.) at room temperature for 15 min prior to adding to wells. Unbound cells were removed by two washes with Tyrode's buffer; bound cells were fixed with 3% glutaraldehyde in DPBS for 20 min, washed twice with DPBS, and stained with 0.04% crystal violet in 20% ethanol for 5 min. Wells were then washed twice with water and treated with 100 l of 1% SDS, and absorbance at 595 nm was measured using a Bio-Rad 680 microplate reader. Each assay point was derived from 2 to 4 wells in at least two independent experiments.
Cell Survival Assay-The wells of a 96-well MaxiSorp plate were coated with the purified X domain of ADAM12, polylysine (Sigma), or Engelbreth-Holm-Swarm laminin (Invitrogen) in 0.1 M NaHCO 3 (pH 9.0) for 12 h at 4°C and then blocked with 2.5% bovine serum albumin in DPBS for 1 h at 37°C. C2C12 cells were detached with 0.05% (w/v) trypsin and 1 mM EDTA in DPBS and washed three time with DMEM. Cells were suspended in Tyrode's buffer and added to the wells (10 5 cells/well). After 1 h, unbound cells were removed, and serum-free DMEM was added to the wells. The number of viable cells was determined right after adding DMEM to the cells (time 0) or 24 h later using the CellTiter 96 AQ ueous One Solution cell proliferation assay (Promega).
Effect of Soluble Recombinant Proteins on Cell Differentiation-C2C12 cells cultured in 96-well plates were switched at ϳ90% confluence from growth medium to differentiation medium (DMEM supplemented with 2% horse serum) with or without a test protein (X, M, or DCE; each at 2 M). After 0, 1, 2, and 3 days in differentiation medium, cellular proteins were extracted for analysis of differentiation markers by Western blotting. The extent of cell fusion was analyzed by phasecontrast microscopy using an Axiovert 200 inverted microscope.
Immunofluorescence Analysis-Cells plated on coverslips placed in a 6-well plate were transfected with the XT-wt, XT, or ETCyt expression vector using FuGENE 6 transfection reagent (Roche Applied Science) and incubated for 12 h in growth medium. Then, 5 ϫ 10 5 untransfected cells were added to each well (to increase cell density and to achieve confluence), and cells were switched to differentiation medium. One day later, cells were fixed with 3.7% paraformaldehyde in DPBS for 20 min and permeabilized with 0.1% Triton X-100 in DPBS for 5 min. The coverslips were then incubated with rabbit anti-ADAM12 antibody (anti-disintegrin antibody for the XT-wt and XT constructs and anticytoplasmic peptide antibody for the ETCyt construct; each at 1:500 dilution) and mouse anti-myogenin antibody (1:50 dilution), followed by incubation with fluorescein isothiocyanate-conjugated anti-rabbit IgG antibody, rhodamine-conjugated anti-mouse IgG antibody, and 4Ј,6diamidino-2-phenylindole (0.01 g/ml; to visualize cell nuclei). After washing with DPBS, coverslips were mounted on glass slides and examined using an Axiovert 200 inverted fluorescence microscope.

RESULTS
We observed previously that the recombinant disintegrin and cysteine-rich (DC) domains of mouse ADAM12 expressed in Sf21 insect cells required the presence of Mn 2ϩ to support integrin ␤ 1 -mediated cell adhesion, suggesting low affinity interactions between the DC protein and integrins (14). Despite the presence of a signal peptide, the recombinant protein was not efficiently secreted from Sf21 cells and was purified from cell-associated material (14). This raised the possibility that the purified protein used in our previous adhesion assays might not have been properly folded. Here, we developed new ADAM12 expression constructs for production of secreted proteins in Drosophila S2 cells. In addition to the DC domain alone, the constructs included the entire extracellular domain of ADAM12 (X-wt and catalytically inactive X; see below), the extracellular domain lacking the EGF-like domain (MDC-wt and catalytically inactive MDC), the extracellular domain lacking the prodomain and metalloprotease domain (DCE), and the prodomain and metalloprotease domain alone (catalytically inactive M) (Fig. 1). All of the constructs contained an exogenous secretion signal and a C-terminal His 6 tag.
After induction of protein expression in stably transfected S2 cells, the recombinant proteins were efficiently (with a yield of at least 70%) secreted into the culture medium. The analysis of X-wt and MDC-wt expression by Western blotting using antibodies recognizing the C-terminal His 6 tag demonstrated that the apparent molecular masses of the secreted X-wt and MDC-wt proteins were ϳ30 kDa smaller than the molecular masses of the proteins detected in the total cell lysates (Fig. 2, left panel). This suggested that the X-wt and MDC-wt proteins present in the medium were fully processed and that the Nterminal prodomain was proteolytically removed in the secretory pathway (1,2). (The predicted molecular mass of the prodomain is ϳ20 kDa; multiple N-linked glycosylation sites are present.) Attempts to purify the intact X-wt and MDC-wt proteins from the cell culture medium were unsuccessful, as we consistently observed significant levels of protein degradation during the purification procedure, most likely mediated by the active metalloprotease domain of the X-wt and MDC-wt proteins (Fig. 2, left panel). No degradation was detected when the critical Glu 349 residue at the catalytic site was replaced with Gln ( Fig. 2, right panel) (32). The E349Q mutants of X-wt and MDC-wt (from now on referred to as X and MDC, respectively) were secreted into the medium and underwent the same intracellular processing as the wild-type proteins (Fig. 2, right  panel), suggesting that the E349Q mutation did not have a profound effect on protein folding. Because of the ease of purification and the unchanged intracellular processing, the X and MDC proteins (but not the X-wt and MDC-wt proteins) were used in this work. Fig. 3A shows a Coomassie Blue-stained reducing gel containing the purified proteins used in the rest of this study: X, MDC, M (containing the same E359Q mutation as the X and MDC proteins), DCE, and DC purified from S2 cell culture medium (Fig. 2, left panel) and DCE, DC, and C b E (20)  Fig. 1). Four days after induction of expression, recombinant proteins were purified from S2 cell culture medium. Samples of cell lysates (C; 0.002% of the total), cell culture medium (M; 0.001% of the total), and purified X and MDC proteins (P; 0.06% of the total) were analyzed by SDS-PAGE under reducing conditions and by Western blotting using anti-pentahistidine antibody.
purified from E. coli (right panel). Each of the recombinant proteins was at least 90% pure, with the exception of the M fragment, the purity of which was 70 -80%. A minor 30-kDa band detected in the preparations of the X, MDC, and M proteins most likely represented the prodomain that remained associated with the metalloprotease domain after proteolytic cleavage and that co-purified with the C-terminally tagged portions of the proteins. Western blot analysis under nonreducing conditions demonstrated that the S2 cell-expressed proteins contained much lower amounts of high molecular mass oligomers connected by intermolecular disulfide bonds compared with the bacterially expressed proteins (Fig. 3B), further indicating that they were most likely properly folded.
To obtain new insight into the role of ADAM12 in muscle cells, we examined the abilities of newly produced ADAM12 domains to support the adhesion of mouse C2C12 myoblasts. The DC protein purified from S2 cell culture medium supported significant cell adhesion only in the presence of Mn 2ϩ (Fig. 4, A  and B), which is similar to the results obtained previously for L6 rat myoblasts (14). Surprisingly, C2C12 cells adhered to the X protein in a dose-dependent manner even in the absence of Mn 2ϩ (Fig. 4, A and B). Although Mn 2ϩ was required to achieve the maximal (almost 100%) adhesion, we consistently observed that up to 60 -70% of the cells attached to X protein-coated plates when Mn 2ϩ was not included in the adhesion buffer. The higher cell adhesion to the X protein than to the DC protein in the absence of Mn 2ϩ was not caused by higher amounts of X protein present during the adhesion assays, as the efficiencies of coating of plastic wells with the two proteins were very similar (data not shown). After 1 h of attachment to X proteincoated wells in the absence of Mn 2ϩ , cells appeared flattened and well spread, whereas no spreading was observed when cells were incubated in DC protein-coated wells (Fig. 4C). As expected, in the presence of Mn 2ϩ , cell spreading was very efficient in both X and DC protein-coated wells (Fig. 4C).
Stimulation of cell adhesion by Mn 2ϩ to an immobilized ligand indicates that integrins are involved in the adhesion process, and a strong dependence of adhesion on Mn 2ϩ further suggests low affinity integrin-ligand interactions (33,34). We wondered whether C2C12 cells adhered to the X domain of ADAM12 via integrins and whether the reason why substantial adhesion was detected even in the absence of Mn 2ϩ was due to the fact that the X domain bound with higher affinity to integrin receptors compared with the smaller DC domain. First, we examined the effect of an integrin ␤ 1 function-blocking antibody on cell adhesion to the X and DC proteins. We observed that, although in the presence of Mn 2ϩ , adhesion to both the X and DC proteins was partially inhibited by monoclonal antibody Ha2/5 (Fig. 5A), in the absence of Mn 2ϩ , adhesion to the X protein not only was not blocked, but was somewhat stimulated by antibody Ha2/5. The complete lack of inhibition of cell adhesion by antibody Ha2/5 was observed even when a suboptimal coating concentration of X protein (1 g/ml) was used in Results are shown as the mean Ϯ S.E. from triplicate measurements; the experiment was repeated five times with similar results. C, C2C12 cells suspended in Tyrode's buffer with (lower panels) or without (upper panels) Mn 2ϩ were added to wells coated with 10 g/ml DC (left panels) or X (right panels) protein as indicated. After 1 h, total cells (without removing nonadherent cells) were examined by phase-contrast microscopy. the adhesion assays (Fig. 5B). The stimulatory effect of antibody Ha2/5 in the absence of Mn 2ϩ was more pronounced when cells were plated in DC protein-coated wells, where the basal level of cell adhesion was very low (Fig. 5B). This suggested that inhibition of ␤ 1 integrins might have activated another class of integrins, a phenomenon that is frequently observed in many other cell types and is refereed to as integrin cross-talk (35,36). Treatment of mammary carcinoma cells, for example, with integrin ␤ 1 function-blocking antibodies induces activation of integrin ␣ v ␤ 3 (37). Because myoblasts express integrin ␣ v ␤ 3 (38), we tested whether a similar ␤ 1 -to-␤ 3 cross-talk might take place in C2C12 cells. As shown in Fig. 5B, however, incubation of C2C12 cells with mouse-specific integrin ␣ v ␤ 3 function-blocking antibody RMV-7, either alone or in combination with antibody Ha2/5, did not have a significant effect on cell adhesion to the X protein. Thus, C2C12 cell adhesion to the X domain in the absence of Mn 2ϩ does not appear to be mediated by ␤ 1 integrins or integrin ␣ v ␤ 3 .
To further explore the nature of the interactions between cells and the X domain of ADAM12, we investigated whether the X domain can support the survival of C2C12 cells. As cell attachment to extracellular ligands via integrins usually promotes cell survival in the absence of exogenous growth factors and prevents cell death (39 -41), we asked whether adhesion to the X domain generates similar survival signals. C2C12 cells were allowed to adhere for 1 h to the X protein, laminin-1 (an integrin ligand and a positive control), or polylysine (a negative control). After removing unbound cells, the attached cells were incubated in serum-free medium, and the amount of viable cells was measured 24 h later. As shown in Fig. 5C, although 80 -90% of the cells attached to laminin-1 were viable after 1 day of incubation in the absence of serum, virtually no cells that were initially attached to the X protein survived the starvation period. Moreover, although cell attachment to the X protein initially led to efficient cell spreading (see Fig. 4C), after 24 h of incubation in serum-free medium, cells became rounded and detached (data not shown). Thus, adhesion to the X domain of ADAM12, unlike integrin-mediated adhesion to laminin, did not provide a survival signal. When cells were allowed to attach to wells coated with a mixture of laminin and the X protein, the survival rate was identical to the rate observed for laminin (Fig. 5C), suggesting that the X protein did not interfere with laminin-induced prosurvival signals. Collectively, the results in Fig. 5 suggest that, although Mn 2ϩ -activated integrins can clearly contribute to cell adhesion to the X domain of ADAM12, in the absence of exogenous integrin activators, the adhesion is most likely mediated by non-integrin proteins.
It was reported previously that part of the cysteine-rich domain of ADAM12 and the EGF-like domain (C b E) (see Fig. 1) produced in E. coli support adhesion of a number of cell types through syndecans (42,43). Because the C b E region of FIG. 6. Adhesion to the X domain in absence of Mn 2؉ does not involve proteoglycans. A, the wells of 96-well plates were coated with the bacterially expressed DC (‚), DCE (OE), or C b E (•) protein at the indicated coating concentrations. C2C12 cells were suspended in Tyrode's buffer without Mn 2ϩ , and cell adhesion was measured as described in the legend to Fig. 4. B, wells were coated with the X domain of ADAM12 expressed in Drosophila S2 cells or with the C b E domain expressed in E. coli (10 g/ml each). The wells were blocked with 2.5% bovine serum albumin for 1 h and then treated for 1 h with 2.5% bovine serum albumin (Control; black bars), with 20 g/ml heparin (gray bars), or with 20 g/ml heparan sulfate (white bars), followed by adding C2C12 cells and measuring cell adhesion. Results are shown as the mean Ϯ S.E. from three measurements.

FIG. 5. Adhesion to the X domain in absence of Mn 2؉ does not involve ␤ 1 integrins.
A, the effect of an integrin ␤ 1 function-blocking antibody on cell adhesion in the presence of Mn 2ϩ . The wells of 96-well plates were coated with the DC or X protein at 10 g/ml. C2C12 cells were suspended in Tyrode's buffer with 1 mM Mn 2ϩ and incubated for 15 min without any antibodies (Control; black bars) or with 20 g/ml antibody Ha2/5 (an integrin ␤ 1 function-blocking antibody; gray bars). Cells were then added to the wells (10 5 cells/well) and incubated for 1 h at 37°C. After removing unbound cells, bound cells were fixed and stained with 0.04% crystal violet. Results are shown as the mean Ϯ S.E. from three measurements. B, the effect of an integrin ␤ 1 function-blocking antibody on cell adhesion in the absence of Mn 2ϩ . The wells of 96-well plates were coated with the DC or X protein at 1 g/ml. Cells were incubated for 15 min in Tyrode's buffer without Mn 2ϩ and without antibodies (Control; black bars), with 20 g/ml antibody Ha2/5 (gray bars), with 20 g/ml antibody RMV-7 (an integrin ␣ v ␤ 3 function-blocking antibody; white bars), or with 20 g/ml antibody Ha2/5 and 20 g/ml antibody RMV-7 (cross-hatched bars) prior to adhesion assays. C, adhesion to the X domain does not support cell survival. The wells of 96-well plates were coated with 10 g/ml laminin (L), with 20 g/ml ADAM12 X domain, with 10 g/ml laminin and 20 g/ml X domain (LϩX), or with 20 g/ml polylysine (P). C2C12 cells were suspended in Tyrode's buffer and added to the wells (10 5 cells/well). After 1 h, unbound cells were removed, and serum-free DMEM was added to the wells. The number of viable cells was determined immediately after adding DMEM to the cells (time 0; white bars) or after 24 h (black bars). Results are shown as the mean Ϯ S.E. from four measurements.
ADAM12 was fully contained within the X (but not the DC) construct, we examined whether C2C12 cell adhesion to the X protein is mediated by syndecans. First, we compared the celladhesive properties of the bacterially expressed C b E fragment and its longer version, the DCE protein (Fig. 1). Although the C b E fragment clearly supported the adhesion of C2C12 cells in a dose-dependent manner and this adhesion did not require the presence of Mn 2ϩ , the DCE protein (and DC) induced much lower levels of cell attachment (Fig. 6A). Thus, similarly to the results of the previous report (42), the presence of the disintegrin domain and the C a subfragment of the cysteine-rich domain in the DCE protein appeared to block the adhesion. Although we were not able to directly compare the adhesive properties of the S2 cell-expressed C b E and DCE proteins (because we were not able to produce the C b E protein in S2 cells), the low level of cell adhesion to the S2 cell-expressed DCE protein (seen in Fig. 7 below) further indicates that the potential syndecan-binding site(s) in the C b E fragment is hindered by the N-terminal D and C a domains. Finally, although syndecan-mediated adhesion to the bacterially expressed C b E protein was effectively blocked by heparin or heparan sulfate (Fig.  6B) (42,43), cell adhesion to the S2 cell-expressed X protein was not inhibited by heparin or heparan sulfate (Fig. 6B). Furthermore, no inhibition was observed when heparin treatment was combined with treatment with antibodies Ha2/5 and RMV-7 (data not shown). Thus, we conclude that the interaction of C2C12 cells with the X domain of ADAM12 is not mediated by syndecans.
Next, we examined the role of the metalloprotease domain of ADAM12 in C2C12 cell adhesion. Although the MDC protein lacking the C-terminal E domain was as effective as the X construct in supporting cell adhesion in the absence of Mn 2ϩ , the DCE protein lacking the M domain failed to induce cell attachment (Fig. 7A). Adhesion to the DCE protein was similar to that to the DC protein, whereas adhesion to the MDC protein was virtually the same as that to the X protein (Fig. 7A). The M domain alone did not support efficient cell attachment (Fig.  7A). This suggests that the presence of both the M and DC domains is critical for C2C12 cell adhesion in the absence of integrin activation.
The requirement for the M and DC domains could mean that the binding site for a cell-surface protein responsible for cell adhesion is located at the interface between these two domains. Alternatively, the binding site could be confined to one of the two domains, but the presence of both the M and DC domains was needed for proper folding of each domain. To discriminate between these two possibilities, we analyzed the secondary structures of the purified X, M, and DCE proteins by CD. If the secondary structures of the M and DCE proteins are largely the same when expressed autonomously and when present as domains in the X protein, then the calculated sum of the CD spectra of the M and DCE proteins should correspond to the spectrum of the X protein (at equimolar concentrations of the X, M, and DCE proteins).
Typically, the far-UV CD spectra of polypeptides with extensive ␣-helical structure have two characteristic minima near 208 and 222 nm. ␤-Sheet structure yields a minimum at ϳ215 nm, and random coil is characterized by lack of a positive peak at ϳ195 nm, a negative peak in the vicinity of 200 nm, and low ellipticity at ϳ222 nm (44). The CD spectra of the X, M, and DCE proteins suggested that all three proteins have a high content of an ordered structure, with a large contribution of ␤-sheet, most likely stabilized by abundant disulfide bonds (Fig. 7B). Most important, a large deviation was observed between the measured X spectrum and the calculated sum of the M and DCE spectra (Fig. 7B). Although this result does not exclude the possibility that the M and DCE domains both participate in forming a protein-binding site required for cell adhesion, it strongly suggests that the autonomously expressed M and DCE fragments do not assume the same secondary structures that are present in the X protein. Thus, the inability of the autonomously expressed M, DCE, or DC domain to support cell adhesion may be due to incorrect protein folding.
A critical question is whether the structural features of the extracellular domain of ADAM12 that require the presence of both the metalloprotease and DC (or DCE) domains are biologically relevant. Because ADAM12 shows a rather narrow pattern of expression among normal adult tissues and is highly expressed in regenerating skeletal muscle (23,24) and because its function has been linked to muscle regeneration (25,26) and determination of reserve cells during myogenic differentiation in vitro (29), we first asked whether cell adhesion supported by the extracellular domain of ADAM12 is muscle-specific. We found that, although Chinese hamster ovary cells and mouse NIH3T3 fibroblasts in the presence of Mn 2ϩ adhered as well to the X domain of ADAM12 as did C2C12 myoblasts, they did not adhere to the X domain in the absence of integrin activation (Fig. 8).
We showed previously that overexpression of full-length ADAM12 in C2C12 cells leads to inhibition of myogenic differentiation and induction of a non-differentiated G 0 -like phenotype characteristic of reserve cells (29). Here, we studied the effect of the soluble extracellular domain of ADAM12 on C2C12 cell differentiation. When confluent C2C12 cells were transferred to differentiation medium containing 2% horse serum, cell differentiation was observed 1-2 days later, which was characterized by expression of the cell cycle inhibitor p21 and myogenin (Fig. 9A), well established differentiation markers,   FIG. 7. The metalloprotease, disintegrin, and cysteine-rich domains are required for C2C12 cell adhesion in the absence of Mn 2؉ . A, the wells of 96-well plates were coated with the X (E), MDC (OE), DCE (‚), DC (•), or M (Ⅺ) protein purified from S2 cell culture medium at the indicated coating concentrations. C2C12 cells were suspended in Tyrode's buffer without Mn 2ϩ , and cell adhesion was measured as described in Fig. 4. B, shown are the CD spectra of the recombinant ADAM12 fragments. The spectra of the X fragment (thick solid line), the DCE fragment (dotted line), and the M fragment (dashed line), each at 4 M in DPBS, were measured at 25°C. Each spectrum represents an average of five scans. The thin solid line represents the sum of the spectra for the DCE and M fragments. mdeg, millidegrees. as well as cell fusion (Fig. 9D). When the X domain of ADAM12 was included in the differentiation medium, differentiation was suppressed, and the expression of p21 and myogenin and the formation of myotubes were delayed (Fig. 9, A and D). Interestingly, the presence of the X domain did not have any effect on the expression of the cell cycle inhibitor p27 and the retinoblastoma-related protein p130, the two G 0 markers previously shown to be up-regulated by the full-length transmembrane form of ADAM12 (39). In contrast, addition of the M or DCE domain of ADAM12 to C2C12 cells at the same molar concentration as the X domain did not inhibit the expression of p21 or myogenin (Fig. 9, B and C). All three proteins were equally stable in the presence of C2C12 cells, and their concentrations in the medium did not change significantly during the time course of the experiment (data not shown). Thus, similar to the effect observed during cell adhesion, both the M and DCE domains are required for inhibition of myoblast differentiation.
The results presented so far suggested that the X domain of ADAM12 may interact with and inhibit a cell-surface protein that is critically involved in differentiation. To obtain further insight into the mechanism of this interaction, we asked whether a transmembrane ADAM12 protein, bearing the same X domain, can participate in similar protein binding events, leading to inhibition of differentiation, and whether the binding occurs in a cis-or trans-orientation. To this end, we transfected C2C12 cells with a membrane-anchored X domain of ADAM12 (XT-wt and catalytically inactive XT) (see Fig. 1). Twelve hours after transfection, cells were transferred to differentiation medium, incubated for an additional 24 h, and then fixed and stained with anti-ADAM12 antibody (to identify transfected cells) and anti-myogenin antibody (to identify differentiating cells). As a negative control, we used an N-terminally truncated form of ADAM12 (ETCyt) composed of the EGF-like, transmembrane, and cytoplasmic domains and lacking a major part of the extracellular domain.
The recombinant ADAM12 proteins expressed in C2C12 cells had correct molecular masses (Fig. 10A) and were located at the cell surface and in the secretory pathway (data not shown). Notably, we observed that expression of myogenin was inhibited in cells overexpressing the XT-wt or XT protein (Fig. 10B). In contrast, myogenin expression was not inhibited in cells overexpressing the ETCyt protein (Fig. 10B). This result demonstrates that the membrane-anchored X domain of ADAM12 is capable of inhibiting cell differentiation and suggests that protein-protein interactions involving the X domain and leading to inhibition of differentiation occur in a cis-orientation. DISCUSSION The disintegrin domains of several ADAM proteins (or larger fragments of ADAM extracellular domains containing the disintegrin region) support cell adhesion via integrin receptors (8 -19). With few exceptions, however, integrin-mediated adhesion requires pre-activation of cellular integrins with Mn 2ϩ ions or integrin-activating antibodies (8 -15), suggesting low affinity interactions between integrins and their ADAM ligands. Interestingly, such low affinity interactions with integrins are observed not only for recombinant ADAM domains expressed in E. coli, which most likely lack the correct arrangement of disulfide bonds, but also for ADAM domains produced in eukaryotic cells, where the formation of disulfide bonds is enabled.
We have recently shown that the DC domain of ADAM12 produced in insect cells interacts with integrin ␣ 7 ␤ 1 and that, in line with reports for other ADAM proteins, this interaction requires integrin activation (14). In the present work, we demonstrated that the entire extracellular domain of ADAM12 (the X domain) produced in Drosophila S2 cells supported efficient adhesion and spreading of C2C12 myoblasts in the absence of exogenous integrin activators. This adhesion appears to be a result of a novel interaction between the X domain and a cell-surface protein in C2C12 cells rather than increased affinity of ADAM12-integrin interaction for several reasons. First, C2C12 cell adhesion to the X domain in the absence of Mn 2ϩ was not inhibited by integrin ␤ 1 -blocking antibody, whereas cell adhesion in the presence Mn 2ϩ (when activated integrins participate in adhesion) was partially decreased by the same antibody. Second, adhesion to the X domain, in contrast to integrin-mediated adhesion (39 -41), did not represent a survival signal in C2C12 cells. Third, adhesion to the X domain seemed to be myoblast-specific and was not observed in CHO-K1 or NIH3T3 cells. Furthermore, although the shorter C b E fragment of ADAM12 supports cell adhesion via syndecans (42,43), adhesion to the X domain did not appear to be mediated by syndecans (Fig. 6). The identity of the receptor mediating interactions with the X fragment of ADAM12 is currently under investigation in our laboratory.
The X protein used in this study contained the E349Q mutation in the metalloprotease catalytic site to prevent autodegradation of the recombinant protein during purification (Fig.  2). A similar propensity for autodegradation was previously observed during attempts to purify catalytically active matrix metalloproteinases. Both matrix metalloproteinases and ADAM proteins belong to the metzincin superfamily of zinc-dependent endopeptidases (45). The active-site glutamic acid is conserved throughout the superfamily (with the exception of several enzymatically inactive ADAM proteins) and plays a critical role in the catalytic mechanism. X-ray crystal structure analysis demonstrated that the replacement of the catalytic Glu residue with Ala in MMP2 (46), Glu with Gln in MMP3/ stromelysin-1 (47), and Glu with Gln in MMP9 (48) does not have any major effects on the structure of the enzyme. Because the folding topology is very similar across the entire metzincin family (49 -51), the active-site environment of ADAM proteins should resemble the active site in matrix metalloproteinases. This notion is supported by crystal structure determination of the metalloprotease domain of ADAM17/tumor necrosis factor-␣-converting enzyme (52) and, more recently, of ADAM33 (53). Therefore, the substitution of Glu 349 with Gln in the extracellular domain of ADAM12 in our studies most likely did not have a significant effect on protein structure and did not generate new structural determinants that would account for the increased C2C12 cell adhesion. Consistently, the E349Q mutant was efficiently secreted into the culture medium, suggesting that it was properly folded and underwent correct intracellular processing in S2 cells. Finally, the inhibition of cell differentiation exerted by the soluble X domain containing the E349Q mutation was replicated by expression of either the E349Q mutant or the wild-type form of the transmembrane XT fragment of ADAM12 in C2C12 cells (Fig. 10). This strongly suggests that the wild-type and E349Q mutant forms of the extracellular domain of ADAM12 engaged in the same proteinprotein interactions at the cell surface.
The nascent X protein produced in S2 cells contained the region of ADAM12 extending from the prodomain to the EGFlike domain, preceded by a secretion signal. During the processing in the secretory pathway of S2 cells, the prodomain was cleaved off (Fig. 2), and the recombinant protein purified from S2 cell culture medium was largely devoid of the prodomain (Fig. 3A). We reasoned that the stronger adhesion of C2C12 cells to the X protein than to the DC domain must have been related to the presence of the metalloprotease or EGF-like domain. Based on the results shown in Fig. 7A, we further concluded that the joined presence of both the metalloprotease and DC domains is required for efficient cell adhesion. As high resolution structural data for the entire extracellular domain of ADAM12 (or any other ADAM protein) are not available, the roles of the M and DC domains in forming a more adhesive ligand are not clear. Analysis of the purified X, M, and DCE domains by far-UV CD suggests, however, that the secondary structure of the autonomously expressed M and DCE proteins may differ from the structures they assume in the context of the larger X domain (or full-length ADAM12). Thus, although the disintegrin domain of ADAM12 (and possibly of other ADAM proteins) represents a low affinity integrin ligand, the entire extracellular domain of ADAM12 possesses unique structural features that allow high affinity interactions with non-integrin receptors.
The novel protein-protein interactions reported here involving the extracellular domain of ADAM12 may have important biological consequences during myoblast differentiation. As shown previously, ADAM12 is expressed in proliferating C2C12 cells and, after induction of differentiation, in a pool of reserve cells that undergo reversible G 0 arrest, up-regulate the FIG. 11. Model for ADAM12-mediated inhibition of myoblast differentiation. In the absence of ADAM12, internal signals and external cues (transmitted into the myoblast interior by cell-surface protein(s), shown schematically as white ovals) promote myoblast differentiation (left). In the presence of ADAM12, differentiation is inhibited as a result of protein-protein interactions mediated by the extracellular domain of ADAM12 in a cis-configuration (right). The entire mature form of the extracellular domain of ADAM12, containing both the metalloprotease and DCE regions, is involved in this interaction. expression of the cell cycle inhibitor p27 and the retinoblastoma-related protein p130 (two markers of G 0 ), down-regulate the expression of MyoD, and remain undifferentiated (29). ADAM12 is not expressed in MyoD-positive, terminally differentiated, multinucleated myotubes (29). Recently, a similar inverse correlation between ADAM12 and MyoD expression was also observed during differentiation of human rhabdomyosarcoma cells (30). Moreover, we reported previously that overexpression of the full-length transmembrane form of ADAM12 in C2C12 cells is sufficient to induce the G 0 -arrested, non-differentiated phenotype of reserve cells (29). Because the formation of reserve cells and terminal differentiation are mutually exclusive (21,54,55), it is possible that the reduced differentiation of cells overexpressing transmembrane ADAM12 is a simple consequence of diverting these cells from differentiation into G 0 (29). Alternatively, ADAM12 may play a more active role during inhibition of differentiation that is independent of G 0 induction. The results presented here support the latter possibility, as the extracellular domain of ADAM12 inhibits C2C12 cell differentiation without up-regulation of p27 or p130 expression. This effect is most likely caused by binding of the extracellular domain of ADAM12 to and inhibition of a cell-surface protein that is critically involved in differentiation. Furthermore, our results suggest that the binding of the membrane-anchored extracellular domain of ADAM12 takes place in a cis-configuration (see model in Fig. 11).
In conclusion, we postulate that the full-length transmembrane ADAM12 protein plays a dual role in myoblasts: its extracellular domain acts to inhibit differentiation, whereas the intracellular domain is required to induce quiescence. Finally, the inability of the M and DCE domains alone to inhibit C2C12 cell differentiation mirror their inability to support cell adhesion in the absence of Mn 2ϩ , further supporting the notion that critical structural features of the extracellular domain are lost when it is divided into the M and DCE domains.