Role of ADAM-9 Disintegrin-Cysteine-rich Domains in Human Keratinocyte Migration*

ADAM-9 belongs to a family of transmembrane, disintegrin-containing metalloproteinases involved in protein ectodomain shedding and cell-cell and cell-matrix interactions. The aim of this study was to analyze the expression of ADAM-9 in skin and to assess the role of this proteolytic/adhesive protein in skin physiology. In normal skin, ADAM-9 expression was detected in both the epidermis and dermis and in vitro in keratinocytes and fibroblasts. Here we report that ADAM-9 functions as a cell adhesion molecule via its disintegrin-cysteine-rich domain. Using solid phase binding assays and antibody inhibition experiments, we demonstrated that the recombinant disintegrin-cysteine-rich domain of ADAM-9 specifically interacts with the β1 integrin subunit on keratinocytes. This was corroborated by co-immunoprecipitation. In addition, engagement of integrin receptors by the disintegrin-cysteine-rich domain resulted in ERK phosphorylation and increased MMP-9 synthesis. Treatment with the ERK inhibitor PD98059 inhibited MMP-9 induction. Furthermore, the presence of the soluble disintegrin-cysteine-rich domain did not interfere with cell migration on different substrates. However, keratinocytes adhering to the immobilized disintegrin-cysteine-rich domain showed increased motility, which was partially due to the induction of MMP-9 secretion. In summary, our results indicate that the ADAM-9 adhesive domain plays a role in regulating the motility of cells by interaction with β1 integrins and modulates MMP synthesis.

Degradation of the extracellular matrix is a prerequisite for tissue repair but also for cell migration and for release of bound factors and bioactive peptides. Different proteases have been implicated in these processes, such as the matrix metalloproteinase (MMP), 2 serine, cysteine, and aspartic protease families. In recent years, the family of proteases (a disintegrin and met-alloproteinase (ADAM)) has drawn attention because the manifold proteolytic and adhesive activities of the different ADAM family members were attributed a pivotal role in physiological and pathological situations.
The ADAM family includes ϳ30 members of proteins containing disintegrin-and metalloprotease-like domains. Most of the family members share a common well conserved domain structure, including a prodomain, metalloprotease, disintegrinlike, cysteine-rich, EGF-like, and a short cytoplasmic domain (reviewed in Refs. 1 and 2).
Structurally, the ADAMs are most closely related to the P-III snake venom metalloproteases. However, in contrast to snake venom metalloproteases, most ADAMs possess EGF-like, transmembrane, and cytoplasmic domains. Half of the ADAM proteins are predicted to be active metalloproteinases, although the identification of specific substrates is still lacking for most of them. Various cell surface proteins are shed by ADAMs, such as IL-6 receptor, FAS-ligand, transforming growth factor-␣, tumor necrosis factor-␣, heparin-binding EGF, and L-selectin. The release of soluble forms of these proteins might lead to autocrine and distal paracrine effects that are usually restricted to a limited environment when cell surface-bound (1).
The cell-adhesive function of ADAM proteins has been attributed to the presence of both the disintegrin and cysteinerich domains. These domains are involved in binding to integrins, the heterodimeric cell surface receptors involved not only in the interactions of cells with the surrounding matrix but also with neighboring cells (3). Most of the known ADAMs contain the integrin-binding amino acid sequence RGD (4) or instead ECD or DCD, which can compete with integrin-ligand interactions (4,5). ADAM-2 and ADAM-9 bind to -2 integrin through their disintegrin-like sequence ECD (6 -8), whereas ADAM-15 can interact with integrins on adjacent cells as observed with ␣␤3 and ␣5␤1 integrins on hematopoietic cells (9). Its activity seems to be mediated by the presence of an RGD motif in the disintegrin domain (10). However, recently, Takeda and coworkers (11) solved the crystal structure of a snake venom metalloproteinase, a homologue of mammalian ADAMs, and excluded this motif from binding activities due to its inaccessibility for protein binding. Thodeti et al. (12) have also shown association of ADAM-12 to syndecan-4, leading to cellular spreading, suggesting that additional receptors might be involved in the interaction with ADAM proteins. Recent studies suggest that the cytoplasmic domain of ADAMs may be involved in intracellular signaling leading to activation of proteolytic processes. The cytoplasmic domains of a significant number of ADAM proteins contain proline-rich SH3-ligand motifs and a consensus sequence for phosphorylation by protein kinase C that may transmit signals between the interior and exterior of the cells. In vitro binding assays have shown that ADAM-9 and ADAM-15 interact with two SH3-containing proteins, endophilin and SH3PX1, which may have a role in regulating the function of both proteases by influencing their intracellular processing, transport, and final localization (13). Similarly, PACSIN 2, another SH3-binding protein, was found to bind ADAM-13, thereby regulating its function during embryonic development (14). Further studies showed that protein kinase C␦ induces heparin-binding EGFlike growth factor shedding by binding to and phosphorylating the cytoplasmic domain of ADAM-9 (15).
ADAM-9 distribution is quite broad, and in human skin it is localized in epidermal keratinocytes, where it may be involved in the constitutive shedding of collagen XVII, thereby modulating migration of keratinocytes (16). However, it is still unclear whether ADAM-9 may also be involved in cell-cell interactions once exposed on the cell surface and, if so, which cellular receptor would be involved in these interactions. Whether its function is primarily enzymatic or adhesive is not certain. In the present work, we have analyzed the cell-adhesive function of ADAM-9 in keratinocytes and the signals elicited by these interactions.

EXPERIMENTAL PROCEDURES
Antibodies-For immunodetection analysis, goat polyclonal antibodies raised against human ADAM-9 were purchased from R&D Systems (Wiesbaden, Germany), and rabbit antifilaggrin antibodies were from Covance (Biozol, Eching, Germany). Actin was detected using a mouse monoclonal antibody (MP Biomedicals, Irvine, CA). Anti-His tag antibodies were from Qiagen (penta-His horseradish peroxidase-conjugated mouse monoclonal antibody, Qiagen, Hilden, Germany). Detection of phosphorylated and unphosphorylated p38, ERK, and c-Jun N-terminal kinase proteins was performed using antibodies that specifically recognize the phosphorylated and unphosphorylated forms (Santa Cruz Biotechnology, Heidelberg, Germany). The mouse anti-␤1 integrin antibodies used for Western blot analysis and immunofluorescence were from Biomol (Hamburg, Germany). The blocking monoclonal mouse antibody 4B4 directed against the human ␤1 integrin chain was obtained from Coulter Corp. (Hialeah, FL), whereas antibodies to the ␣ integrin subunits were from Chemicon (Beta1 Integrin Partners Kit; Chemicon, Hofheim, Germany). Mouse anti-MMP-9 antibodies were from Calbiochem (Merck). Purified control IgG was purchased from Dako (Hamburg, Germany).
Cells and Cell Culture-HaCaT cells were kindly provided by N. Fusenig (German Cancer Research Center, Heidelberg, Germany). Cells were routinely cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM glutamine, and 100 units/ml each of penicillin and streptomy-cin. Human epidermal keratinocytes were isolated from adult skin as previously described (17). Keratinocytes were cultured on collagen-coated dishes in FAD medium (DMEM/Ham's F-12 (3:1); Invitrogen) containing 100 units/ml penicillin, 100 g/ml streptomycin, 10% fetal calf serum, 5 g/ml insulin, 1 ng/ml epidermal growth factor, 10 Ϫ10 mol/liter cholera toxin, and 24 ng/ml adenine. Epidermal-dermal split skin was prepared by thermolysin treatment of human skin specimens overnight at 4°C (thermolysin bacillus type X, 1 mg/ml; Sigma). After washing twice with PBS, separated epidermis and dermis were directly processed for RNA preparation as described below.
To obtain undifferentiated/differentiated HaCaT cells, the cells were cultured for 2 weeks in low calcium medium (0.05 mM) and switched to high calcium medium (1.8 mM) for 4 days. After incubation, lysates were prepared as described below.
RNA Isolation and Reverse Transcription-PCR-Total RNA from human skin was prepared after fine mincing of the tissue using RNAzol according to the manufacturer's instructions (Wak-Chemie Medical GmbH, Bad Homburg, Germany). Reverse transcription-PCR was performed following the manufacturer's instructions (REDTaq TM ReadyMix TM PCR reaction mix; Sigma). Briefly, 1 g of RNA was reverse transcribed using oligo(dT) as primer in a total volume of 25 l. 5 l of the cDNA was used to amplify specific transcripts by PCR. The following primers were used: for amplification of human ADAM-9, 5Ј-CCTCGGGGACCCTTCGTGT and 5Ј-ATC-CCATAACTCGCATTCTCTAAA (18); for murine ADAM-9, 5Ј-TCTGACCATCCCAACGTACA and 5Ј-GCTGTTGTG-CAGAGGTTCAA; and for murine MMP-9, 5ЈAGTTTGGT-GTCGCGGAGCAC and 5Ј TACATGAGCGCTTCCGGCAC (19). Amplification of S26 was used for normalization (20). PCRs were performed on 1 l of cDNA for 35 cycles (within the linear range of amplification): denaturation (94°C, 1 min), annealing (60°C, 1 min), and extension (72°C, 1 min). The products were then analyzed on 2% agarose gels in TBE.
Expression and Purification of the Disintegrin-Cysteine-rich Domain of ADAM-9-Total RNA from human dermal fibroblasts was used as template for reverse transcription. Reverse transcription followed by PCR was carried out using the primers 5Ј-ttttgctagctagtgctccctcctgtggt-3Ј and 5Ј-ttttgcggccgcacagtcataattc-3Ј. The amplified DNA fragments were digested and cloned into the NheI/NotI-digested expression vector pCEP-pu BM40-cHis (21) This gave rise to a fusion protein with a His 6 tag placed in frame with the disintegrin-cysteine-rich coding regions of ADAM-9. After transfection of this plasmid into 293-EBNA cells by FuGENE (ratio 3:1; Roche Applied Science), the cells were subsequently selected for puromycin resistance (0.5 g/ml). Serum-free supernatants were tested for expression of the soluble disintegrin-cysteine rich domain (DC-9-his) by SDS-PAGE on a 10% polyacrylamide gel followed by immunoblotting using antibodies specific for the His tag. For purification, supernatants were loaded on an immobilized metal affinity chromatographic column (Talon metal affinity resin; Clontech) with a flow rate of 0.5 ml/min. After washing with 5 column volumes of a buffer containing 20 mM Hepes, 100 mM NaCl, pH 8.0, and 2.5 mM imidazol, the proteins were eluted with 250 mM imidazol and dialyzed

ADAM-9 in Keratinocyte Migration
against PBS overnight at 4°C. Purified proteins were stored at Ϫ80°C.
Transient transfections of HaCaT cells were performed using Lipofectamine (Invitrogen) on 70% confluent cell monolayers. After 6 h, medium was replaced, and expression was analyzed after a further 48 h of culture. The full-length cDNA for ADAM-9 was kindly provided by C. Blobel (Hospital for Special Surgery, New York).
Cell Adhesion Assays-Semiconfluent monolayer cultures of HaCaT cells or human primary keratinocytes were detached by incubation with 0.05% EDTA after three washes with 0.02% EDTA. The cells were then washed with PBS and resuspended in Hepes buffer containing 0.5% BSA and 1 mM each of CaCl 2 and MnCl 2 . Adhesion assays were performed as described before (22). Briefly, 96-well tissue culture plates were coated with recombinant DC-9-his (20 g/ml corresponding to ϳ0.6 M), His 6 peptides (0.6 M), bovine monomeric collagen type I (40 g/ml), and human plasma fibronectin (30 g/ml) at 4°C overnight. BSA coating and blocking of nonspecific binding sites were performed with heat-denatured BSA (1% BSA in Ca 2ϩ /Mg 2ϩ -free PBS) for 1 h at room temperature. After washing the wells twice, cells (2 ϫ 10 4 cells/well) were seeded and incubated for 1 h at 37°C. For competition assays, antibodies (10 g/ml) or peptides (0.6 M) were added to the cell suspension before plating. Nonadherent cells were removed by washing twice with PBS. Adherent cells were fixed with 3% formaldehyde in PBS, pH 7.6, and stained with 0.5% crystal violet in 20% (v/v) methanol. The dye was released from the cells by the addition of 0.1 M sodium citrate in 50% (v/v) ethanol. The optical density of the released dye solution was determined at 595 nm. Adhesion was either expressed directly as A 595 nm or in percentage relative to untreated controls, which were set arbitrarily as 100%. Statistical analysis was performed with Student's t test.
Zymographic Analysis-Serum-free conditioned media were analyzed by gelatin zymography as previously described (20). To analyze whether MMP-9 is secreted as active or latent form of the enzyme, samples were activated with 1 mM 4-aminophenylmercuric acetate) in substrate buffer (see below) for 1 h at 37°C and directly analyzed by gelatin zymography. Briefly, media were fractionated on 10% SDS-polyacrylamide gels containing 1 mg/ml gelatin (bovine; Sigma). After electrophoresis, the gels were washed in 2.5% Triton X-100 for 30 min before overnight incubation in metalloproteinase substrate buffer (50 mM Tris-HCl, pH 8.0, 5 mM CaCl 2 ). Thereafter, the gels were stained with Coomassie Blue R-250, and the bands corresponding to gelatinase activities appeared white against the blue background.
Immunoprecipitation and Western Blot Analysis-Lysates were prepared by washing twice the cells in PBS and directly scraping them off on ice in radioimmune precipitation buffer containing the protease inhibitors aprotinin (10 g/ml), pefabloc (0.25 mg/ml), and leupeptin (1 g/ml). For analysis of phosphorylated proteins sodium vanadate (50 mM) was additionally included. After overnight incubation at 4°C, lysates were clarified by centrifugation at 16,000 ϫ g and 4°C for 20 min, and the supernatant was collected and stored at Ϫ20°C until use. Protein concentration was determined using a commercial assay (Micro-BCA; Perbio Science, Bonn, Germany). For immunoprecipitations, equal amounts of lysates were precleared for 2 h on protein-G-Sepharose (Amersham Biosciences). After centrifugation at 1,000 ϫ g for 10 min, precleared lysates were either applied to mouse IgG or mouse anti-human ␤1 integrin antibodies bound to protein-G-Sepharose (2 g of antibodies preincubated for 2 h with protein-G-Sepharose at 4°C) and incubated overnight at 4°C. After incubation, protein-G-Sepharose-bound proteins were washed three times with radioimmune precipitation buffer, and bound proteins were eluted in sample buffer containing 0.7 M ␤-mercaptoethanol. Eluted protein was further analyzed by Western blotting. For Western blotting, proteins were fractionated by SDS-PAGE on 10% polyacrylamide gels under reducing conditions and transferred onto Hybond-C Super TM (Amersham Biosciences). After blockage of nonspecific binding sites with 5% skimmed milk (w/v) in PBS containing 0.5% Tween (v/v), for analysis of phosphorylated proteins, 5 mM sodium fluoride was included, and the blots were incubated with the primary antibodies overnight at 4°C. Bound primary antibodies were detected using a horseradish peroxidase-conjugated secondary antibody (1:2000; Dako, Hamburg, Germany) and visualized with the ECL TM system (Amersham Biosciences).
Cell Migration Assays-Cell migration assays were performed in 24-well tissue culture plates. Wells were coated with DC-9-his (20 g/ml) and human plasma fibronectin (30 g/ml) overnight at 4°C. BSA blockage of nonspecific binding sites was performed by a 1-h incubation with heat-denatured BSA (1% BSA in Ca 2ϩ /Mg 2ϩ -free PBS) at room temperature. HaCaT cells were treated with mitomycin-C (1.6 g/ml) for 2 h to arrest cell growth and then washed and detached with 0.05% EDTA. After washing twice with PBS, cells were resuspended in Hepes buffer containing 0.5% BSA and 1 mM of each CaCl 2 and MnCl 2. The cells (5 ϫ 10 4 cells/well) were seeded in cloning rings (0.5-mm diameter) and incubated for 1 h at 37°C. After removing the cloning rings and three washes with PBS to remove unbound cells, plates were placed on a microscope stage heated to 37°C in a humidified atmosphere. For inhibition experiments, cells were incubated, after washing, either in the presence of purified mouse IgG, used as control, or mouse anti-MMP-9 antibodies (10 g/ml). Images were collected every hour for 48 h. Areas covered by cells at these time points were calculated using CellR software (Olympus Biosystems, Munich, Germany).
Immunolocalization-To detect proteins in monolayer cultures, cells were cultured on tissue culture slides for 48 h and then fixed for 10 min with cold acetone. Stainings were performed, incubating the cell monolayers or tissue with the first antibodies described under "Antibodies" for 16 h at 4°C in PBS containing 2% BSA and 0.05% Tween. After extensive washes, primary bound antibodies were detected using donkey antigoat 594, goat anti-rabbit 594, and, for the colocalization studies, rabbit anti-mouse fluorescein isothiocyanate (all diluted in PBS with BSA/Tween) for 1 h at room temperature. Nuclei were counterstained with 1 g/ml 4Ј,6-diamidino-2-phenylindole (Roche Applied Science). Negative controls were performed using control IgG as primary antibodies.
For colocalization studies, fluorescence images were recorded using a Zeiss (Thornwood, NY) Axiovert M-200 inverted epifluorescence microscope with Apotome slider confocal attachment. Images captured at ϫ630 magnification were analyzed using the three-dimensional analysis imaging software of the Axiovision LE Rel. 4.5 and Adobe Photoshop version 7.0.

RESULTS
Expression of ADAM-9 in Human Skin-Analysis of protein expression by immunofluorescence showed ADAM-9 expression throughout the whole epidermis with a stronger staining in all suprabasal layers when compared with the basal layer. Expression was also observed in spindle-shaped cells (indicated by white arrows; Fig. 1a). To assess ADAM-9 mRNA expression in human skin, we performed reverse transcription-PCR analysis of RNA preparations from thermolysin-dissociated epidermis and dermis and from total skin (Fig. 1b). ADAM-9 transcripts were detected in both epidermis and dermis with a significantly higher expression in the epidermis. In agreement with the described expression pattern, specific ADAM-9 transcripts were detected by Northern blot analysis of total RNA preparations in cultured keratinocytes, endothelial cells, and fibroblasts but not in melanocytes (data not shown).
To confirm the protein expression pattern observed in the epidermis, we analyzed cellular extracts from undifferentiated (low Ca 2ϩ ) or differentiated (switched to high Ca 2ϩ ) HaCaT cells. The HaCaT cells are spontaneously immortalized keratinocytes, which closely resemble normal keratinocytes in their growth and differentiation characteristics. This has made the HaCaT cell line a widely used model of normal human keratinocytes (17,18).
Cells undergoing differentiation, as indicated by the expression of the late differentiation marker filaggrin (25), produced increased amounts of ADAM-9 protein. Both pro (115 kDa) and active forms (80 kDa) of ADAM-9 were detected by Western blot analysis of total cell lysates (Fig. 1c). By immunofluorescence staining of monolayer cultures, an increase in ADAM-9 expression was observed upon differentiation and paralleled the enhanced in filaggrin staining induced by the culture conditions (Fig. 1d). These data corroborate the expression pattern observed in human skin (Fig. 1a).
Recombinant Production of the Disintegrin-Cysteine-rich Domain of ADAM-9 and Analysis of Its Interaction with Epidermal Cells-The proteolytic functions of ADAM-9 have been analyzed in several studies (26,27). However, its binding activity and putative functional role as an adhesive cell receptor in human skin has not been investigated. Therefore, to elucidate whether ADAM-9 possesses an adhesive function on keratinocytes and to identify ADAM-9 ligands, we produced a recombinant protein consisting of the disintegrin-like cysteine-rich domain of ADAM-9 fused to a His 6 tag in a eukaryotic expression system. The protein was purified from serum-free supernatants of stably transfected 293-EBNA cells using immobilized metal affinity chromatography.
The purified recombinant disintegrin-like cysteine-rich Histagged domain of ADAM-9 (DC-9-his) was secreted as a 35-kDa protein as shown by SDS-PAGE and subsequent Coomassie staining and by immunoblotting with anti-His antibodies (Fig. 2). The identity of the protein was further confirmed by peptide mass fingerprint analysis of tryptic fragments of the recombinant protein (data not shown); the lower band was identified as a smaller degradation product of the disintegrincysteine-rich domain.
To examine whether the recombinant DC-9-his could support cell adhesion of human keratinocytes, we performed cell adhesion assays with immobilized recombinant protein.
HaCaT cells specifically adhered to the DC-9-his domain but very poorly to histidine peptides and BSA used as negative controls (Fig. 3A). The adhesion observed to DC-9-his was 60% of that measured on fibronectin and 35% of that measured on collagen type I, both used as positive controls. At low coating concentrations of DC-9-his, cells appeared rounded and dis- Counterstaining with 4Ј,6-diamidino-2-phenylindole was used to identify cell nuclei; swine anti-rabbit fluorescein isothiocyanate and donkey anti-goat Alexa 594 were used to detect bound primary antibodies (bar, 50 m). b, analysis of ADAM-9 transcripts (424 bp) by PCR in epidermal (E) and dermal (D) thermolysin split skin or in total skin preparations (S). S26 amplification (374 bp) was used as loading control. c, ADAM-9 and filaggrin protein expression was also analyzed by Western blotting/immunodecoration using the above indicated antibodies (1:1000) and horseradish peroxidase-conjugated secondary antibodies for detection. Actin immunodetection was used as loading control (1:10,000). The white arrows indicate spindle-shaped cells in the dermal compartment.

ADAM-9 in Keratinocyte Migration
played small filopodia projecting out of the cells. With increasing coating concentrations, HaCaT spread on the immobilized ligand (Fig. 3B). An adhesion pattern comparable with that of DC-9-his was observed using human primary keratinocytes (data not shown).

Interaction of Epidermal Cells with the Recombinant Disintegrin-Cysteine-rich Domain Requires ␤1 Integrin Receptors-To
identify the integrin subunits potentially responsible for cell adhesion to DC-9-his, inhibitory anti-integrin antibodies were used to analyze inhibition of keratinocyte adhesion to the DC-9-his protein. Antibodies directed against the ␤1 but not the ␤3 integrin subunit efficiently inhibited adhesion by 60% (Fig. 4A). Furthermore, antibodies against the ␣3 subunit reduced adhesion by 70%, whereas inhibitory antibodies directed against the ␣2, ␣5, and ␣6 subunits had no significant effect when compared with the IgG control antibodies. Interaction of integrin receptors or disintegrins with their substrates is mediated by RGD-containing amino acid sequences. In the case of ADAMs that do not contain such a motif (e.g. ADAM-9), the ECD sequence takes over the activity of the RGD amino acid sequence (28). Both RGD and ECD cyclic peptides as well as their respective control peptides containing the RAD and ECD motifs did not inhibit adhesion to immobilized DC-9-his (Fig.  4A). The association of the endogenous ␤1 integrin receptor subunit with ADAM-9 protein was further corroborated by coimmunoprecipitation studies.
Immunoprecipitation of ␤1 integrin from HaCaT cell lysates led to co-precipitation of ADAM-9, thus suggesting a direct interaction between these two proteins (Fig. 4B). This observation was further substantiated by immunolocalization of ADAM-9 and ␤1 integrin in cell monolayers, where both proteins co-localized at sites of cell-cell contacts, as indicated by the white arrow and shown in the detail of the optical section, whereas in cellular protrusions, only ␤1 integrin was detected (Fig. 4C).
The Recombinant Disintegrin-Cysteine-rich Domain Induces Migration of Keratinocytes-One of the cellular processes resulting from integrin receptor engagement by extracellular matrices is migration. In keratinocytes, this event is of particular importance during physiological wound repair, where different integrins are expressed at specific time points to facilitate wound re-epithelialization (29). To analyze whether the adhesive domain of ADAM-9 influences cellular migration, HaCaT cells were plated on plastic or on DC-9-his-coated surfaces, and their migration was monitored every hour by time lapse video microscopy (Fig. 5). The mean migration areas of HaCaT cells on plastic were 0.07 and 0.1 mm 2 at 24 and 48 h, respectively. In contrast, plating HaCaT cells on DC-9-his resulted in increased cellular migration of 0.2 and 0.33 mm 2 at 24 and 48 h, respectively (p Ͻ 0.0001).
Interaction of Cells with DC-9-his Leads to Changes in Cell Signaling-We have previously shown that activation of ␣2␤1 integrin receptors on cells by anti-integrin activating antibodies or by snake venom metalloproteinases (whose human homologues are ADAM proteins) is followed by enhanced MMP activities (22,24). To investigate the role of the adhesive property of ADAM-9 in inducing cellular signaling leading to alteration of proteolytic activities, either directly or indirectly by engagement of integrin receptors, we stimulated HaCaT cells with the soluble recombinant DC-9-his domain. The addition of DC-9-his to HaCaT cells resulted in a dosedependent increase of pro-MMP-9 secretion (Fig. 6A), whereas synthesis of other MMPs was unchanged (data not shown). Interestingly, transient transfection of full-length ADAM-9 cDNA in HaCaT cells leading to increased ADAM-9 mRNA levels was also associated with increased MMP-9 secretion (Fig. 6B).   OCTOBER 19, 2007 • VOLUME 282 • NUMBER 42

JOURNAL OF BIOLOGICAL CHEMISTRY 30789
Analysis of the intracellular signaling events that are activated upon HaCaT interaction with the DC-9-his domain indicates that activation/phosphorylation of ERK1/2 (5 min after stimulation) but not of c-Jun N-terminal kinase or p38 is involved in the regulation of MMP-9 expression, whereas no phosphorylation events were elicited by stimulation with histidine control peptides (Fig. 7A). In addition, a significant inhibition of MMP-9 secretion was observed only upon inhibition of ERK activation by PD98059 but not by other inhibitors, including the p38 inhibitor SB203580, the phosphatidylinositol 3-kinase inhibitor wortmannin, and the tyrosine kinase inhibitor genistein (Fig. 7B). No inducing effects were observed upon cell stimulation with either histidine peptides or Me 2 SO, which was used as solvent for both the SB203580 and PD98059 inhibitors.
To explore whether the increase in cell migration on DC-9his is dependent on MMP-9 secretion induced by this ADAM-9 domain, we performed migration experiments with HaCaT cells on immobilized DC-9-his in the absence or presence of anti-MMP-9 neutralizing antibodies. As shown in Fig. 8, migration of HaCaT cells on DC-9-his was inhibited in the presence of anti-MMP-9 antibodies at 24 and 48 h, thus suggesting that MMP-9 secretion induced by cell interaction with DC-9-his contributes to the increased migration capacity of keratinocytes on immobilized DC-9-his.

DISCUSSION
ADAMs present on the cell surface have emerged as key regulators of numerous cellular processes. They are localized at the cell surface or are themselves integral membrane proteins, which after activation can act in close proximity to the cell membrane. They influence cellcell interactions by shedding ligands and receptors involved in cell-cell contact and signaling but also modulate cell-matrix interactions by directly cleaving and remodeling ECM proteins and by interacting with matrix adhesion molecules (28).
In skin, we found ADAM-9 protein and transcripts in both the epidermal and dermal compartments, and its function in this tissue has been proposed to be the constitutive shedding of collagen XVII, which in turn may modulate keratinocyte migration (16). However, whereas proteolytic and adhesive functions of ADAM-9 in cancer have been investigated in many studies (30), little is known about the adhesive function in skin physiology.
In this report, we could demonstrate that keratinocytes interact with the recombinant disintegrin-cysteine-rich domain of ADAM-9. Adhesion required the presence of Mn 2ϩ but not Ca 2ϩ ions, suggestive of an integrinmediated ion-dependent interaction (data not shown). In addition, this interaction was found to be dependent on the ␤1 integrin receptor subunit but not on the ECD motif, present in the disintegrin domain of ADAM-9. In contrast, Nath et al. (7) as well as Mazzocca et al. (31) have shown that interaction of the ADAM-9 recombinant ectodomain with fibrosarcoma and liver stromal cells could be efficiently inhibited by ECD-containing peptides.
However, a recent analysis of the crystal structure of VAP (vascular apoptosis-inducing protein-1), a snake venom homologue of mammalian ADAMs, identified the high variable region of the cysteine-rich domain, present in this protein, as the responsible region for substrate interaction but not the disintegrin loop, which is packed and inaccessible for protein binding (11). This latter finding would agree with the lack of inhibition of DC-9-his-mediated cell adhesion by ECD peptides we observed in our experimental system.
Two main integrin receptors have been shown to interact with the ADAM-9 ectodomain, namely ␣6␤1 and ␣2␤1 (7, 31). In human skin and in vitro, keratinocytes express mainly ␣2 and . Adhesion of keratinocytes to DC-9-his is mediated by ␤1 integrin receptors. A, DC-9-his protein (0.6 M) was adsorbed onto microtiter plates. HaCaT cells were incubated in assay buffer with the addition of 10 g/ml inhibitory or stimulatory antibodies (as indicated), control IgG and 0.6 M RGD, RAD (control), ECD, or scrambled ECD (ECDscr.; control) peptides before plating. After incubation, wells were washed twice and the cells stained with crystal violet. Dye was eluted and quantified by A 595 nm . Adhesion is expressed as percentage versus cells plated on DC-9-his in the absence of any antibody. This experiment is one representative of three independent experiments performed in triplicate. B, ␤1 integrin was immunoprecipitated (IP) from lysates of confluent HaCaT monolayers. Immunoprecipitation with IgG was used as control. After immunoprecipitation, bound proteins were eluted from protein-G-Sepharose-coupled antibodies in sample buffer under reducing conditions and resolved on a 10% SDS-polyacrylamide gel. Western blot analysis was performed using goat anti-ADAM-9 (1:500) and rabbit anti-␤1 (1:1000) antibodies. C, HaCaT cells were plated onto glass coverslips and cultured for 24 h. Cell layers were washed once with PBS and fixed for with ice-cold acetone. Immunostainings were performed using mouse anti-␤1 integrin (1:100; green) and goat anti-ADAM-9 (1:300; red) antibodies. Secondary antibodies used were donkey anti-goat Alexa 954 (red) and rabbit anti-mouse fluorescein isothiocyanate (green) antibodies. ADAM-9 and ␤1 integrin were viewed using Apotome confocal immunofluorescence. An overlay of both proteins is shown on the right. The white arrow indicates an area of colocalization of the two proteins whose enlarged cross-section is shown at the side of the microphotograph.
␣3 integrin subunits. Additionally, in vitro keratinocytes also express the ␣5 integrin subunit in association with the ␤1 subunit, and the expression profile is influenced by the cellular differentiation status (32). Whereas expression of ␣2 and ␣6 integrins is limited to the basal layer, ␣3 integrin is expressed at low levels also in cells of the suprabasal layers. Furthermore, differentiation of keratinocytes by Ca 2ϩ switch leads to general reduced expression of ␣3 but more importantly, to relocation of the ␣3 subunit from the intracellular pool to the cell membrane (33).
We could show that binding of keratinocytes to the DC-9-his domain was primarily mediated by the ␣3␤1 integrin receptor, thereby suggesting that interaction of integrin receptors with ADAM-9 depends on the integrin repertoire expressed by different cell types. This is, for instance, the case for ADAM-12, where the main co-receptor in myogenic cells is represented by ␣9␤1 (34); however, carcinoma cells that do not express this integrin can utilize other receptors of the ␤1 integrin family (12).
We also demonstrate a direct interaction between the ␤1 subunit and ADAM-9, but we failed to co-precipitate the ␣ subunit (data not shown). One possibility to explain this finding is that an additional unidentified receptor located in the close vicinity of either the ADAM-9 or the integrin receptor might contribute to ADAM-9 binding. A similar hypothesis has been formulated for ADAM-10, whose association with heparinbinding EGF is followed by interaction with CD9, known to complex with the integrin ␣3␤1 (35,36). Interaction between ADAM-9 and ␤1 integrin might similarly involve an additional unknown interacting partner, an interesting question that will be explored in future work.
The reports dealing with the functional interplay between ADAMs and integrins indicate that ADAMs can modulate, by inhibiting or supporting, cell migration mediated by integrins independently of their metalloprotease activity (37). These interactions may play an important role in promoting cell migration in physiological processes, such as during embryogenesis, as well as in cancer cell invasion.
In the case of ADAM-9, Nath et al. (7) have shown in fibrosarcoma cells not only that the recombinant soluble ectodomain of ADAM-9 binds to ␣6␤1 but that this interaction leads to enhanced migration. Furthermore, an alternatively spliced form of ADAM-9, lacking the transmembrane and cytoplasmic domains, is secreted by hepatic stromal cells and binds to ␣6␤4 and ␣2␤1 integrins on colon carcinoma cells, thereby inducing an invasive phenotype (31). We also found that adhesion of keratinocytes to the disintegrin-cys- FIGURE 5. The DC-9-his domain of ADAM-9 induces cell migration. Top, time lapse video microscopy of HaCaT cells seeded onto plastic and recombinant DC-9-his. Cells were plated onto dishes coated with DC-9-his (20 g/ml) or without coating. Cell migration was monitored over 48 h, and the distance migrated was calculated by measuring the area covered by the cell sheet at the indicated time points. The area covered by the cell sheet is denoted as mm 2 and is representative of at least two experiments performed in sextuplicate. Bottom, representative microphotographs from the migration experiment are shown, and the white dotted line marks the migration front at the start of the experiment. FIGURE 6. ADAM-9 adhesive domains induce MMP-9 secretion. A, HaCaT cells were starved for 24 h in serum-free medium before stimulation. Cells were then incubated with DC-9-his protein (10, 50, 100 g/ml) or His 6 peptides (2.8 M) dissolved in serum-free medium. After stimulation, supernatants were collected, centrifuged, and analyzed by gelatin zymography (20 l/lane). Pro-MMP-2 (72 kDa) and the activated MMP-2 forms (62/59 kDa) as well as pro-MMP-9 (92 kDa) are indicated to the right. The activation status of MMP-9 was analyzed after 4-aminophenylmercuric acetate treatment of supernatants from cells stimulated with 100 g/ml DC-9-his, shown in the right microphotograph. B, HaCaT cells were transiently transfected with either the full-length ADAM-9 cDNA (A9) or the empty expression vector (Co). After 48 h, cells were washed and placed in serum-free medium for 24 h. Cleared supernatants were analyzed by gelatin zymography, and cells were processed for RNA purification and subsequent reverse transcription-PCR (RT-PCR) analysis of ADAM-9 transcripts. Amplification of S26 was used as loading control. M, molecular weight marker. OCTOBER 19, 2007 • VOLUME 282 • NUMBER 42 teine-rich domain of ADAM-9 leads to increased cell migration. In vivo migration of keratinocytes is important during reepithelialization of epithelial wounds, starting as early as 2 days after injury and continuing until the wound bed is completely covered (38). This process is known to be accompanied by cellular reprogramming, leading to an altered integrin expression pattern as well as to an increased secretion of proteolytic enzymes (39,40). For instance, pericellular proteolysis is thought to be essential for the detachment of keratinocytes from the basement membrane and for their migration into the wound bed (40). Increased expression of ADAM-9 transcript levels, which we have observed at the early time points of our wound healing study, might indeed be required not only for establishing cellular contacts to increase migration but also for the induction of proteolytic enzymes, such as MMP-9, which was found to be induced in the same phase of wound healing studies in mouse. 3 We have previously shown that jararhagin, a snake venom proteinase homologue of human ADAM proteins, can function as an agonist of collagen type I in fibroblast by inducing cellular signaling, leading to an up-regulation of MMP expression (24). However, the generation of cellular signals, leading to up-regulation of promigratory activities through the interaction of ADAM-9 and ␤1 integrin, has not yet been investigated. In this report, we could show that the adhesive domains of ADAM-9 indeed induce cellular signaling, leading to modulation of proteolytic enzymes.

ADAM-9 in Keratinocyte Migration
Stimulation of HaCaT cells with soluble DC-9-his enhanced MMP-9 secretion, which we also observed upon transient overexpression of the full-length ADAM-9 cDNA. In addition, this interaction involves phosphorylation of ERK kinases.
Importantly, Holvoet and co-workers (41) have shown that in HaCaT cells, induction of MMP-9 by tumor necrosis factor requires the mitogen-activated protein kinase pathway. However, in our studies, induction of MMP-9 by the DC-9-his domain did not occur via release of cytokines/growth factors, since we could not detect changes in these factors (e.g. tumor necrosis factor-␣) (data not shown). However, activation of integrins has also been shown to induce MMP-9 synthesis. Using antibodies raised against the ␤1 and ␣3 integrin subunits, expression of MMP-9 in cultured keratinocytes was induced in a dose-dependent manner (42). Thus, we speculate that upon binding of ADAM-9 to the integrin receptor, phosphorylation events take place, leading to up-regulation of MMP-9, which in turn contributes to the increased migratory activity.
Indeed, we could demonstrate partial inhibition of migration on DC-9-his by incubation of the cells with anti-MMP-9 neutralizing antibodies. This observation indicates that MMP-9 is directly involved in the change of the migratory capacity of keratinocytes in response to the disintegrin-cysteine-rich domain of ADAM-9. Future studies will be necessary to examine the role of ADAM-9 in vivo and the functional importance of its adhesive domains not only for modulation of protease expression but also for modulation of cell-cell interactions.
3 P. Zigrino, unpublished data. FIGURE 7. Induction of MMP-9 secretion by DC-9-his is mediated by ERK kinases. A, after stimulation with either DC-9-his (0.9 M ϭ 30 g/ml) or His 6 peptides (0.9 M), the cells were lysed at the indicated time points in radioimmune precipitation buffer containing a protease inhibitor mixture and sodium vanadate. 5 g of the protein extracts were resolved by SDS-PAGE. Western blot analysis was performed with antibodies specific for the phosphorylated and unphosphorylated kinases. B, after starvation for 24 h cell monolayers were stimulated with DC-9-his (0.9 M) in the absence or presence of various inhibitors (wortmannin (100 nM), genistein (100 nM), PD98059 (20 M), and SB203580 (20 M)) for 24 h. Stimulation with His 6 peptides (0.9 M) or Me 2 SO (DMSO; 0.1%) was used as control. Equal amounts of proteins (10 g/lane) from cleared supernatants were analyzed by gelatin zymography. JNK, c-Jun N-terminal kinase. FIGURE 8. MMP-9 contributes to keratinocyte migration on DC-9-his. Shown is time lapse video microscopy of HaCaT cells plated on the recombinant DC-9-his domain of ADAM-9 (20 g/ml). Cell migration was monitored over 48 h in the presence of purified mouse IgG (Ϫ) or of mouse anti-MMP-9 antibodies (10 g/ml). Migrated distance was calculated by measuring the area covered in time by the cell sheet and denoted in mm 2 . The experiment was repeated twice in triplicate.