The Cysteine-Rich Domain of Human Adam 12 Supports Cell Adhesion through Syndecans and Triggers Signaling Events That Lead to β1 Integrin–Dependent Cell Spreading

Control of cell adhesion is important in several biological phenomena, including embryonal development and tumor cell invasion and metastasis (Damsky and Werb 1992; Yamada and Geiger 1997; Couchman and Woods 1999; Giancotti and Ruoslahti 1999). Understanding the molecular mechanisms of these processes requires information about a variety of extracellular ligands, their interaction with cell membrane receptors, and the subsequent downstream signal cascade. One class of cell adhesion receptors is the integrins that often work in cooperation with other cell surface molecules. Integrin α5β1–mediated cell adhesion of cultured fibroblastic cells to fibronectin requires syndecan-4 as a coreceptor (Saoncella et al. 1999), and the cell surface chondroitin sulfate proteoglycan NG2 cooperates with α4β1 in mediating melanoma cell adhesion and spreading (Eisenmann et al. 1999). Similarly, multimolecular receptor adhesion complexes have been described for tetraspan molecules and integrins (Hemler 1998).

The ADAMs (a disintegrin and metalloprotease) constitute a recently discovered family of cell adhesion receptors (Wolfsberg and White 1996; Blobel 1997; Black and White 1998; Schlöndorff and Blobel 1999; Stone et al. 1999; Primakoff and Myles 2000). Over 20 members have been described so far (see http://www.med.virginia.edu/~jag6n/whitelab.html). Most members are composed of pro-, metalloprotease, disintegrin-like, cysteine-rich, EGF-like repeat, transmembrane, and cytoplasmic tail domains. One important role of ADAMs relates to their metalloprotease activity (Black and White 1998; Schlöndorff and Blobel 1999). A well-studied example is ADAM 17 (TACE), which functions in cell surface shedding of a number of growth factors, including tumor necrosis factor-α (TNFα), transforming growth factor-α (TGFα), and Notch (Black et al. 1997; Moss et al. 1997; Peschon et al. 1998; Brou et al. 2000). Another important role of the ADAMs relates to cellular interactions, as shown for fertilin α and β (ADAM 1 and 2) in sperm–egg binding and fusion during fertilization (Almeida et al. 1995; Chen et al. 1999a,Chen et al. 1999b). However, the molecular mechanisms of the ADAMs in cellular interactions are not well-understood.

We have recently described the cloning of secreted and transmembrane forms of human ADAM 12 (Gilpin et al. 1998) that are implicated in myoblast fusion (Yagami-Hiromasa et al. 1995) and myogenesis (Gilpin et al. 1998). Human ADAM 12 is an active metalloprotease containing the highly conserved zinc-binding catalytic motif HEXGH- XXGXXHD (Loechel et al. 1998, Loechel et al. 1999). ADAM 12 expression is upregulated in cancer, and in several tumors ADAM 12 immunostaining is enriched along the tumor cell surfaces (Iba et al. 1999). Based on this intriguing tissue distribution pattern, we asked whether ADAM 12 was involved in cell adhesion. We found that the recombinant cysteine-rich domain of human ADAM 12 made in E. coli (rADAM 12-cys) supports carcinoma cell adhesion but fails to promote cell spreading (Iba et al. 1999). In this study, we explored the molecular mechanisms underlying this type of cell adhesion. We found that mesenchymal cells attach, spread, and form focal adhesions and organize stress fibers in response to ADAM 12, and that both syndecans and integrins are necessary to mediate these processes. Carcinoma cells, on the other hand, bind to syndecans but do not spread or engage β1 integrins in the ADAM 12–mediated adhesion process. However, they can be induced to spread by addition of either 1 mM Mn²⁺ or the β1 integrin–activating mAb 12G10.

Integrin β1 function-blocking mAbs rat IgG₁ AIIB2 and mouse IgG2b CSAT were obtained from the Developmental Studies Hybridoma Bank maintained by the University of Iowa, Department of Biological Sciences. Another integrin β1 function-blocking mAb (clone P5D2) was obtained from Chemicon. The β1 function-activating mAb 12G10 (Mould et al. 1995) was kindly provided by M. Humphries (University of Manchester, Manchester, UK) and an integrin β1 mAb (K20) with no reported effect on cell function was obtained from Immunotech. Rat IgG₁ mAb to mouse tetranectin (Wewer et al. 1998) was used as an isotype Ig control for the AIIB2. Function-blocking mAbs to integrin α1 (clone H22B6), α2 (clone Gi9), and α5 (clone SAM1) were obtained from Immunotech, and mAbs to integrin α3 (P1B5) was obtained from Life Technologies. Function-blocking mAb to integrin α6 (clone 135.13C) was kindly provided by A. Mercurio (Beth Israel Hospital, Harvard, Boston, MA). For immunostaining of focal adhesions, mAb to vinculin (kindly provided by M. Glukhova, Institut Curie, Paris, France) was used. Tetramethyl rhodamine isothiocyanate-phalloidin was obtained from Molecular Probes and used to stain actin. mAbs to syndecan-1 (281-2) (Jalkanen et al. 1985; Liebersbach and Sanderson 1994) and to glypican-1 (kindly provided by A.D. Lander, University of California Irvine, Irvine, CA) (Litwack et al. 1998) were used to stain transfected ARH-77 cells. For Western blotting, a pan-syndecan antiserum that reacts with all syndecans (kindly provided by A.C. Rapraeger, University of Wisconsin, Madison, WI) (Reinland et al. 1996; Ott and Rapraeger 1998) and an antiserum to human syndecan-4 (Kojima et al. 1996) were used. To confirm the identity of the recombinant ADAM 12 fragments by Western blotting, two rat mABs, 14E3 to the cysteine-rich domain (Gilpin et al. 1998) and a newly developed 2F7 to the disintegrin-like domain, were used. The 2F7 mAb was generated and characterized as described previously using recombinant ADAM 12 aa 412–557 polypeptide expressed in E. coli as an antigen (Gilpin et al. 1998). IgGs were purified using protein G–Sepharose as described by the manufacturer (Amersham Pharmacia Biotech). Fluorescein- and rhodamine-conjugated antibodies against rabbit, rat, and mouse Igs were purchased from DAKO.

Except when specified, cell lines were obtained from American Type Culture Collection. The following human cell lines were used: RKO colon carcinoma (kindly provided by A. Mercurio, Boston, MA), MDA-MB-231 breast carcinoma (HTB26), RD rhabdomyosarcoma (CCL 136), SV-HFO human fetal osteoblasts transformed with simian virus 40 virus (Chiba et al. 1993), MG-63 osteosarcoma (CRL 1427), Saos-2 osteogenic sarcoma (HTB 85), HOS osteogenic sarcoma (CRL 1543), MRC-5 lung fibroblast (CCL 171), IMR-90 lung fibroblast (CCL 186), WI-38 lung fibroblast (CCL 75), and CCD 1072SK foreskin fibroblast (CRL 2088). In addition, the murine osteoblastic MC3T3-E1 cell line was used (Sudo et al. 1983). These cell lines were grown in DME, supplemented with glutamax I and 4,500 mg/ml glucose and 10% FBS. The human ARH-77 B-lymphoid cell line and a series of ARH-77–transfected cell lines (Neo, Syn-1, -2, -4, Syn/279t, Syn/Glyp, Glyp/Syn, Glyp-1, and triple deletion mutant [TDM]) (Langford et al. 1998; Liu et al. 1998) were grown in suspension in RPMI 1640 medium with glutamax I supplemented with 5% FBS. All media used in this study were supplemented with 50 U/ml penicillin, 50 μg/ml streptomycin, and the cells were grown in Nunc tissue culture flasks (GIBCO BRL) at 37°C in 5% CO₂ in a humidified atmosphere.

NMRI mice were obtained from M & B and white Italian × Isabraum, F1 chicken embryos from Statens Serum Institut.

Primary mouse osteoblast-like cells were isolated from calvaria of newborn (0–3-d-old) NMRI mice using an enzyme mixture containing 0.32 mg/ml collagenase (Sigma C 5894) and 0.25% trypsin (GIBCO BRL) essentially as described (Bellows et al. 1986; Otto et al. 1996). The cells were grown in DME and 10% FBS. The osteogenic capacity of the cells was tested by their ability to deposit minerals. In brief, the cells were grown in above standard medium supplemented with 50 μg/ml ascorbic acid (Sigma A 2147) and 10 mM β-glycero-phosphate (Sigma G 9891). Mineralization was observed after 10–14 d, and mineralized foci were stained by Alizarin red S (Sigma A 5533) after 21 d in culture.

Primary mouse muscle cells were isolated from the hindlimb muscle of newborn (0–3-d-old) mice using an enzyme mixture of 0.15% trypsin and 0.1% dispase (both enzymes from GIBCO BRL) essentially as described (Allen et al. 1984; DiMario and Strohman 1988; Partridge 1997). The cells were cultured in DME with 20% FBS and 2% chick embryo extract. Subconfluent cell cultures were passaged and some cultures were changed to DME with 2% horse serum to confirm the ability of the myoblasts to differentiate and form contracting myotubes.

Primary chicken muscle cell cultures were prepared from 11-d-old chick embryo breast muscle using 0.25% trypsin essentially as described (Bischoff and Holtzer 1969; Knudsen and Horwitz 1977).

Chondroblast cell lines with a deletion of the β1 integrin gene were established using mice with a loxP-tagged (floxed) β1 integrin gene (Potocnik). Chondroblasts were isolated from 3-d-old pups of homozygously floxed mice and immortalized by infection with a retrovirus transducing a heat-labile variant of the simian virus 40 large T antigen (kindly provided by P. Sharp, Massachusetts Institute of Technology, Cambridge, MA). These +^c/+^c cells were subsequently infected with an adenovirus transducing the cre-recombinase (Anton and Graham 1995), resulting in excision of the β1 integrin gene in infected cells. β1 integrin–null cells (−/−) were purified by FACS and further cultured. FACS sorted β1 integrin–expressing cells, in which no or only one allele of the β1 integrin gene was deleted, served as control cells.

For FACS analysis, cells were briefly trypsinized, washed with growth medium, and resuspended in ice-cold PBS containing 1% BSA. Staining for β1 integrin was carried out in the same buffer using FITC-labeled mAb Ha2/5 (PharMingen). Cell sorting was done using a FACSvantage, and analytical stainings were analyzed with a FACSort (Beckton Dickinson).

To detect focal adhesions, cells were fixed at room temperature for 10 min in 4% paraformaldehyde in PBS containing 2 mM MgCl₂ and permeabilized with 0.1% Triton X-100 in DME with 15% FBS for 5 min at room temperature. After rinsing and blocking for 30 min in DME with 15% FBS, vinculin antibodies supplied as culture medium supernatant were diluted (1:10) in DME with 2% BSA and incubated with the cells for 30 min followed by a thorough rinse and incubation with secondary antibodies. The actin cytoskeleton was visualized by tetramethyl rhodamine isothiocyanate-phalloidin staining as recommended by the manufacturer. For immunostaining of cell surface syndecan-1 and glypican-1, ARH-77 cell suspensions were rinsed in culture medium and incubated with FITC-conjugated mAb 281 against syndecan-1 or anti–glypican 1 antiserum (10 μg/ml) followed by a second antibody. Cells were examined using an inverted microscope (Zeiss Axiovert) equipped with phase contrast optics and connected to a PentaMAX chilled charge-coupled device camera (Princeton Instruments, Inc.).

The attachment assay was performed as described previously (Iba et al. 1999). Substrates included purified rADAM 12-cys, disintegrin-like domain (rADAM 12-disintegrin), metalloprotease domain (rADAM 12-metallo) made in E. coli (Iba et al. 1999), purified recombinant mouse disintegrin-like and cysteine-rich domain (rADAM 12-DC) made in insect cells (Zolkiewska 1999), and for comparison, mouse laminin-1 from the EHS tumor, human laminin-10/11, and human fibronectin (GIBCO BRL). BSA was used as a negative control. Nunc-Immuno™ 96-well plates with MaxiSorp™ surface (Nunc) were coated with 1–50 μg/ml of the substrates in 0.1 M NaHCO₃ buffer, pH 9.5, overnight at 4°C, rinsed with PBS, and incubated with 10 mg/ml BSA in PBS for 1 h at 37°C. After a rinse with PBS, 100 μl of the cell suspension (0.6 × 10⁶/ml) was added to the wells for 1 h at 37°C in 5% CO₂ in a humidified atmosphere. The wells were rinsed twice in serum-free DME, fixed for 20 min in 2% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.2, rinsed in PBS and stained with 0.1% crystal violet in 10% methanol (vol/vol). Absorbance was measured with a Multiscan ELISA reader (Labsystems) at 590 nm. A blank value corresponding to an empty well was automatically subtracted. Each assay point was derived from three to six separate wells and repeated three times.

The following experimental conditions were tested: (a) The cells were grown in sulfate-free Fischer's medium containing 10% dialyzed FBS in the presence of 20 mM sodium chlorate (Sigma C 3171), or in both sodium chlorate and 10 mM sodium sulfate as a control (Baeuerle and Huttner 1986; Keller et al. 1989), for 24 h before the assay. (b) Plates were preincubated with heparin (Sigma H 2149), heparan sulfate (Sigma D 9808), heparin-albumin (Sigma H 0403), chondroitin sulfate A, B, and C, respectively (Sigma C 9819, C 0320, C 4384), or hyaluronic acid (Sigma H 1751) at the concentrations indicated. Alternatively, these compounds were added in solution during the incubation period. (c) The cells were pretreated for 30 min with 2 mU/ml heparitinase or 50 mU/ml protease-free chondroitinase ABC (#100703 and 100332, respectively; Seikagaku Corp.). (d) The cells were preincubated for 15 min at room temperature with 10 μg/ml of function-blocking mAbs to human integrin β1 (A2BII or P5D2), integrin α1, α2, α3, α5, α6 (H22B6, Gi9, PIB5, SAM 1, and 135.13C), or as an isotype control, with 10 μg/ml IgG₁ rat mAbs to tetranectin (1B11). For experiments with chicken muscle cells, the CSAT mAb was used. (e) Cycloheximide (Sigma C 7698) was added at 25 μg/ml for 3 h before the assay period to inhibit protein synthesis. (f) Cytochalasin B (Sigma C 6762) or D (Sigma C 8273) dissolved in DMSO was added at 20 and 0.1 μM, respectively. Unless otherwise indicated, the various reagents were present throughout the 1-h attachment assay period.

To test for induction of cell spreading, a modification of the above mentioned attachment assay was employed. The cells were allowed to attach for 30 min and then Mn²⁺ (0.1 or 1 mM MnCl₂) or 10 μg/ml of the β1 integrin mAbs (12G10 or K20) was added for another 30 min.

An expression construct for full-length human ADAM 12-S fused at the COOH terminus to green fluorescent protein (GFP) was constructed using the pEGFP-N1 from CLONTECH Laboratories, Inc. A DNA fragment containing the entire coding region of ADAM 12-S (pro-, metalloprotease, disintegrin-like, cysteine-rich, EGF-like repeat domains, and the 33–amino acid COOH tail; sequence data available from EMBL/GenBank/DDBJ under accession no. AF023477) was generated by PCR using Pfu DNA polymerase (Stratagene) and inserted at the EcoR1/SacII sites of pEGFP-N1, such that the COOH terminus of ADAM 12 was in-frame with GFP. A control plasmid for secretion of GFP alone was made by transferring the EcoR1/Not1 GFP cassette from pEGFP-N1 to the plasmid pSecTagA (Invitrogen Corp.), resulting in expression of a GFP polypeptide fused at the NH₂ terminus to an Igκ leader sequence. COS-7 cells were transiently transfected as described previously (Loechel et al. 1998, Loechel et al. 1999), the culture medium collected 2 d later, and any cellular debris removed by centrifugation. ARH-77 cells (5 × 10⁵) expressing syndecan-1 or -4 or control ARH-77 cells transfected with an empty vector (neo) were incubated with 1 ml of human ADAM 12-GFP or GFP-containing culture medium. After 1 h incubation at 37°C or at 4°C, the cells were examined by fluorescence microscopy. In other control experiments, heparin was added to the incubation medium or the cells were pretreated with heparitinase.

Total RNA was purified using the TRIzol reagent (GIBCO BRL). For the reverse transcriptase reaction, the cloned mouse Moloney leukemia virus reverse transcriptase was used as recommended by the manufacturer (Stratagene). Aliquots of cDNA were amplified using the following primers: human syndecan-1 (forward) 5′-GCTCTGGGGATGACTCTGAC-3′ and (backward) 5′-GTATTCTCCCCCGAGGTTTC-3′; human syndecan-2 (forward) 5′-ACATCTCCCCTTTGCTAACGGC-3′ and (backward) 5′-TAACTCCATCTCCTTCCCCAGG-3′; and human syndecan-4 (forward) 5′-GTCTGGCTCTGGAGATCTGG-3′ and (backward) 5′-TGGGGGCTTTCTTGTAGATG-3′.

After an initial denaturation at 95°C for 40 s, 35 cycles of denaturation at 94°C for 40 s, annealing at 55°C for 40 s, and extension at 72°C for 60 s were carried out. The reaction products were separated on agarose gels. The amplified DNA fragments (syndecan-1, 552 bp; syndecan-2, 539 bp; and syndecan-4, 397 bp) were cloned and sequenced to confirm the identity of the PCR products.

Extraction, isolation, and Western blotting of syndecans were performed as described previously (Lebakken and Rapraeger 1996). In brief, RD cells were grown to subconfluence, extracted in 8 M urea, 0.1% Triton X-100, 10 mM Tris-HCl, and 1 mM Na₂SO₄ and incubated with DEAE Sephacel in buffer overnight. The beads were washed and proteins eluted with 1 M NaCl in heparitinase buffer followed by digestion with heparitinase and chondroitinase. The samples were separated by SDS-PAGE and immunoblotting performed using the pan-syndecan antibody.

To identify which cell surface proteoglycans bind to rADAM 12-cys, an affinity chromatography protocol devised by Hoffman et al. 1998 was employed. In the Hoffman et al. protocol, urea was used in both extraction and running buffer, as the attempt to use either octyl glucoside or Triton X-100 were unsuccessful, apparently because syndecans are insoluble in these lysis buffers and precipitate with the cytoskeleton. RD cells were washed in PBS, incubated with Ez-Link sulfo NHS-LC-biotin (#21335; Pierce Chemical Co.) at a final concentration of 100 μg/ml for 1 h at room temperature, washed twice with 50 mM ammonium chloride to quench the free biotin, and washed once in PBS. Cell pellets were resuspended and lysed with a dounce homogenizer in 10 mM KCl, 20 mM Tris-HCl, pH 7.4, 0.1% β-mercaptoethanol, 1 mM EDTA (1 ml/10⁸ cells). After removal of the nuclei by centrifugation at 1,500 g for 5 min, NaCl was added to the supernatant at a final concentration of 0.15 M and the supernatant centrifuged at 50,000 g for 30 min at 4°C. The pellet was solubilized in 1 ml of 8 M urea, 2% Triton X-100, 2 mM PMSF in TBS, pH 7.4, and centrifuged at 14,000 g for 20 min at 4°C to remove insoluble matter. The resulting cell membrane protein solution was incubated with a rADAM 12-cys affinity matrix (rADAM 12-cys coupled to CnBr-activated Sepharose following the manufacturer's instructions) in 6 M urea, 1% Triton X-100, 2 mM PMSF in TBS, pH 7.4, (running buffer) overnight at 4°C under gentle rotation. After a thorough rinse in running buffer, elution was carried out in 20 mM EDTA. The eluted fractions were precipitated with acetone, washed with ethanol, and resuspended, and part of the material was digested with heparitinase (0.1 U/ml) and/or chondroitinase ABC (1 U/ml) at 37°C for 4 h in the presence of aprotinin (2.2 μg/ml), leupeptin (1 μg/ml), pepstatin A (2 μg/ml), and EDTA (5 mM). Undigested and digested material was separated on a 4–12% SDS-PAGE and transferred to Hybond-P (polyvinylidene difluoride) paper (Amersham Pharmacia Biotech). The paper was subsequently blocked in 50 M Tris, 0.5 M NaCl, 0.1% Tween 20, 5% nonfat milk powder, and incubated with peroxidase-conjugated streptavidin at a final concentration of 1 μg/ml or with a specific polyclonal antiserum to human syndecan-4 (Kojima et al. 1996), followed by a peroxidase-conjugated goat anti–rabbit IgG secondary antibody. The blots were developed using the chemiluminescence SuperSignal kit from Pierce Chemical Co.

Native human syndecan-4 ectodomain was purified from the culture medium supernatant of endothelium-like Eahy926 cells under nondenaturating conditions and iodinated as described (Kojima et al. 1996). The solid-phase binding assay was used as described previously (Kojima et al. 1996). In brief, microtiter wells were coated with 5 μg/ml of the basic fibroblast growth factor (bFGF) or 50 μg/ml rADAM 12-cys in 100 μl of 50 mM NaHCO₃, pH 9.2, overnight at 4°C, incubated with 1% BSA in PBS at room temperature to block nonspecific binding, and incubated with ¹²⁵I-labeled human syndecan-4 (800 cpm) in 100 μl of 0.1% BSA overnight at 4°C. After rinsing the wells, the amount of bound material was determined using an Aloka γ-counter (Aloka Co. Ltd.).

In a previous report, we showed that carcinoma cells attach but do not spread on rADAM 12-cys (Iba et al. 1999). Here we examine the molecular mechanism resulting in cell adhesion of a number of different cell types on human rADAM 12-cys made in E. coli and mouse rADAM 12-DC made in insect cells. In our assays, we tested cells of osteoblastic origin such as SV-HFO, HOS, Saos-2, MG-63, MC3T3-E1 cell lines, and primary murine osteoblasts; fibroblastic cells such as MRC-5, WI-38, IMR-90, and CCD1072SK; myoblastic cells such as RD cells and primary murine and chicken muscle cells. Each of these mesenchymal cell lines attached and >90% of the cells displayed a spread and flattened morphology on rADAM 12-cys (Fig. 1 A). A similar adhesion pattern was observed when the cells were plated on laminin or fibronectin (not shown). Cell attachment was specific and dose-dependent, and reached a maximum at ∼15–25 μg/ml (Fig. 1B and Fig. C). Cells attached efficiently to rADAM 12-DC (Fig. 1 C), but cell spreading appeared less efficient than on rADAM 12-cys at equimolar concentrations. No cell attachment was observed on either the disintegrin-like or the metalloprotease domains made in E. coli. (Fig. 1 C). The fully spread mesenchymal cells formed focal adhesions and contained an elaborate actin cytoskeleton when plated on rADAM 12-cys (Fig. 2). Cell spreading and stress fiber formation was decreased by treatment of the cells with cytochalasin B and D, inhibitors of actin polymerization (not shown). These results demonstrate that in contrast to carcinoma cells, which attach but fail to spread, mesenchymal cells attach, spread, form specialized adhesion structures, and reorganize the actin cytoskeleton in response to rADAM 12-cys.

We next determined which class of cell adhesion receptors mediates adhesion to rADAM 12-cys and rADAM 12-DC. Since we found previously that attachment of carcinoma cells was dependent on heparan sulfate chains (Iba et al. 1999) and that CHO cells deficient in heparan sulfate synthesis (CHO pgsD-677) were unable to attach to rADAM 12-cys (Iba, K., and U.M. Wewer, unpublished results), several compounds interfering with the function of cell surface heparan sulfates were tested in adhesion assays (Fig. 3 A). First, mesenchymal cells were depleted of functional cell surface proteoglycans by growing them in the presence of 20 mM sodium chlorate, which inhibits sulfation of cell surface glycosaminoglycans. Chlorate-treated mesenchymal cells were unable to attach to rADAM 12-cys (Fig. 3 A), but retained their ability to adhere to laminin (not shown). Thus, the lack of cell attachment on rADAM 12-cys was not due to chlorate-induced cytotoxicity. The inhibitory effect of sodium chlorate on cell attachment was restored by including 10 mM sodium sulfate in the chlorate-containing culture medium (not shown). Secondly, cell attachment was significantly inhibited when heparin or heparan sulfate was included in the assays. Similarly, cell attachment to rADAM 12-DC was inhibited by heparin (Fig. 3 B). In contrast, hyaluronic acid or chondroitin sulfate ABC did not interfere with cell adhesion (Fig. 3 A). Heparin was very potent in inhibiting cell attachment. For most of the cell lines tested, 1 μg/ml heparin reduced cell attachment to 50%, and 10–20 μg/ml led to a complete inhibition (not shown). The only exception was RD rhabdomyosarcoma cells, which exhibited only a partial inhibition of cell attachment when assays were performed in the presence of heparin but almost a complete inhibition (90%) when heparin was administered in the form of heparin-albumin (not shown). Thirdly, removal of cell surface heparan sulfate chains by pretreatment of the cells with heparitinase reduced cell attachment, whereas chondroitinase ABC had no effect (not shown). No effect on cell attachment was observed when cells were pretreated with β1 integrin function-blocking mAbs (Fig. 3 A). These results demonstrate that attachment of mesenchymal cells to rADAM 12-cys and rADAM 12-DC was critically dependent on cell surface heparan sulfate.

To identify the nature of the heparan sulfate proteoglycan, which provides the heparan sulfate chains for rADAM 12-cys– and ADAM 12-DC–mediated cell adhesion, we took advantage of the ARH-77 B lymphoid cell system (Langford et al. 1998; Liu et al. 1998). The ARH-77 cell line, which was derived from a human plasma cell leukemia, expresses very little heparan sulfate on the cell surface. However, when transfected with cDNAs encoding members of the syndecan family, these ARH-77 cells express functional cell surface syndecans that bear heparan sulfate chains (Langford et al. 1998; Liu et al. 1998). A series of stably transfected cell lines that express either syndecans-1, -2, or -4, or glypican-1 on their cell surface (Liu et al. 1998) were tested in a cell adhesion assay (Fig. 4 A). As expected, the parental ARH-77 cells transfected with vector only (ARH-77 neo) did not attach to rADAM 12-cys. Cells transfected with either syndecan-1,-2, or -4 attached to rADAM 12, but did not spread, whereas glypican-1–transfected cells did not attach. Similarly, ARH-77 cells expressing syndecan-1 also attached to rADAM 12-DC (Fig. 4 B). Adhesion of syndecan-1–transfected cells to rADAM 12-cys (Fig. 4 C) or rADAM 12-DC (not shown) is blocked by compounds that interfere with the function of heparan sulfate chains. Moreover, cells expressing syndecan-1 lacking all three heparan sulfate attachment sites (ARH-77–TDM) were unable to attach to rADAM 12-cys (Fig. 4 C). Some of the syndecan-1 TDM molecules do bear chondroitin sulfate chains (Langford et al. 1998), but apparently cannot mediate cell attachment to rADAM-cys. Together these data demonstrate that adhesion of the ARH-77 cells to ADAM 12 is mediated by heparan sulfate chains of syndecan.

The finding that glypican-1–expressing cells did not adhere to rADAM 12-cys indicated that cell surface heparan sulfate chains alone were not sufficient for adhesion to rADAM 12-cys and suggested a role for the syndecan core proteins. To examine this, we analyzed adhesion to rADAM 12-cys of ARH-77 cells expressing various mutant and chimeric forms of syndecan-1, including syn/279t (syndecan-1 lacking its cytoplasmic tail), syn/glyp (a chimeric proteoglycan composed of the syndecan-1 extracellular domain and a glycosyl phosphatidylinositol (GPI) anchor for attachment to the cell surfaces), and glyp/syn, (a chimeric proteoglycan composed of the glypican extracellular domain and the syndecan transmembrane and cytoplasmic domains). It has been shown previously that both the mutant and chimeric forms of syndecan are capable of binding ARH-77 cells to type I collagen and to fibronectin (Liu et al. 1998; this study; not shown). Representative syndecan-1 and glypican-1 cell surface immunostainings are shown as insets in Fig. 4 D. However, none of these mutant cell lines exhibited significant cell adhesion to rADAM 12-cys (Fig. 4 D). Taken together these results demonstrate that different syndecan family members can mediate cell adhesion to rADAM 12-cys. Moreover, in addition to the critical role of heparan sulfate chains, both the extracellular and cytoplasmic domains of the syndecan core protein are required for the ADAM 12–mediated cell adhesion process.

Having established the critical role of syndecans in rADAM 12-cys–mediated cell adhesion, we wanted to explore which syndecans can be functional in the mesenchymal cells. We examined RD rhabdomyosarcoma cells, which express syndecans-1 (80 kD), syndecan-2 (50 kD), and syndecan-4 (35 kD) based on RT-PCR and Western blotting (Fig. 5A and Fig. B). Cell surface biotinylated RD cells were extracted and applied on an affinity-matrix to which rADAM 12-cys had been immobilized. Elution of bound material and subsequent gel separation and Western blotting showed a smear of an approximate M_r of 200,000 (Fig. 5 C, lane 1). After digestion, the material with heparitinase, a sharp band of 35 kD, was seen (Fig. 5 C, lane 2). Syndecan-4 has a similar molecular mass and the protein was identified in Western blotting using antibodies specific for syndecan-4 (Fig. 5 C, lane 3). These results indicate syndecan-4 as one potential adhesion receptor for rADAM 12-cys. Since solubilization of syndecans required the use of urea in the extraction buffer (Hoffman et al. 1998), we do not know whether syndecan-4 is the sole syndecan used by the RD cells when they attach to rADAM 12-cys.

The interaction between native human syndecan-4 and rADAM 12-cys was confirmed in a solid-phase binding assay (Fig. 5D and Fig. E). As demonstrated in Fig. 5 E, rADAM 12-cys exhibited significant syndecan-4 binding, similar to that of bFGF. Thus, when 800 cpm of ¹²⁵I-labeled human syndecan-4 (Fig. 5 D) was added in the assay (Fig. 5 E), an average of 224 cpm (± 7.1) bound to bFGF, 105 cpm (± 15.1) bound to rADAM 12-cys, and 5 cpm (± 3.1) to BSA. Heparitinase treatment of syndecan-4 completely destroyed its ability to bind to both rADAM 12-cys and bFGF, demonstrating that the heparan sulfate glycosaminoglycan chains were crucial for binding of these two ligands.

The results presented so far were obtained using rADAM 12-cys made in E. coli or rADAM 12-DC made in insect cells. To confirm the interaction of full-length ADAM 12 with cell surface syndecans, we generated a GFP-tagged form of ADAM 12-S expressed in COS-7 cells. Culture medium from these cells was collected and mixed with ARH-77 cells expressing syndecans or control ARH-77 cells that do not express syndecans. After a 1-h incubation at 4°C or at 37°C, ARH-77 cells were examined by fluorescence microscopy. Distinct punctate cell surface fluorescence was observed on the syndecan-expressing cells but not on control cells (Fig. 6A, Fig. B, and Fig. D). No fluorescence was observed when the syndecan-expressing ARH-77 cells were pretreated with heparitinase or when heparin was included in the medium, or when the cells were incubated with medium containing GFP without ADAM 12-S (Fig. 6, C–E). These data support the model that ADAM 12 binds to cell surfaces primarily via syndecans.

When mesenchymal cells were treated with function-blocking mAbs to β1 integrin (AIIB2 and P5D2) the cells attached but no longer spread on rADAM 12-cys (Fig. 7, B–D). More than 90% of these cells had a rounded morphology with several fine 7–10-μm long cytoplasmic projections without organized focal adhesions and stress fibers. In parallel experiments, these antibodies inhibited cell attachment of RKO colon carcinoma cells to laminin-1 and -10 (not shown), confirming the function-blocking activity of the preparation of purified mAbs used. As a further control, we tested a control IgG and found that this mAb had no effect on cell attachment and spreading (not shown). The data shown represent experiments done with the SV-HFO cell line and are representative of the inhibitory effect of function-blocking β1 integrin mAbs on the spreading of the other mesenchymal cells used in this study. Finally, we examined cell spreading of SV-HFO, MRC-5, IMR-90, and RD cells that had been treated for 3 h before the experiment with 25 μg/ml cycloheximide to inhibit protein synthesis. Such treatment had no effect on cell attachment and spreading on rADAM 12-cys (not shown), indicating that cells bind and spread on the supplied rADAM 12-cys and not to substances secreted by the cells during the assay period.

To provide further evidence that integrin β1 is involved in rADAM 12–mediated spreading, we used genetically modified cells. Integrin β1–null mice are early embryonal lethal, precluding establishment of defined β1 integrin^−/− cell populations in culture (Fässler and Meyer 1995; Stephens et al. 1995). To overcome this problem, we established chondroblasts from conditional β1 integrin knock-out mice (Potocnik, A.J., C. Brakebusch, and R. Fässler, manuscript submitted for publication), immortalized them with simian virus 40 large T antigen, and subsequently deleted the β1 integrin by a cre recombinase transfection as described in Materials and Methods. Chondroblast+^c/+^c cells attached, and >90% of the cells spread on rADAM 12-cys and laminin-1, respectively (Fig. 8A and Fig. B). Chondroblast^−/− cells attached on rADAM 12-cys, but their spreading was significantly reduced (Fig. 8 C). As expected, β1 integrin–null chondroblasts did not attach on laminin-1 (Fig. 8 D). Moreover, we confirmed that both chondroblast+^c/+^c and −/− cell attachment to rADAM 12-cys was completely inhibited by heparin (Fig. 8 E). The lack of detectable β1 integrin expression on the surface of −/− chondroblasts was confirmed by FACS analysis (Fig. 8 F).

To determine which integrin α chain(s) were involved in cell spreading, we used a series of function-blocking mAbs in the attachment assays. No effect on cell attachment or cell spreading was observed with function-blocking antibodies to integrin α1, α2, α3, α5, and α6 when added individually (not shown). When administered in pairs of two mAbs, integrin α3 and α5, α3 and α6, and α5 and α6 mAbs had no effect on cell attachment, but a partial (∼30%) inhibitory effect on cell spreading (not shown). Other combinations were without effect on cell spreading. Finally, when integrin α3, α5, and α6 mAbs were combined, 90% inhibition of cell spreading was seen (Fig. 9). These results suggest that multiple integrin α subunits participate in cell spreading on rADAM 12-cys.

Our data suggested that mesenchymal cells use syndecans as the primary cell attachment receptor to rADAM 12-cys, and use β1 integrin to mediate cell spreading. However, in a previous report, we found that carcinoma cells had a rounded morphology and did not spread upon attaching to rADAM 12-cys (Iba et al. 1999). We therefore tested whether cell spreading could be induced (Fig. 10). Addition of either Mn²⁺ or β1 integrin–activating antibodies changes the conformation of the extracellular domain of β1 integrins to an active ligand-binding state by increasing the apparent affinity (Hynes 1992; Humphries 1996). Carcinoma cells were plated on laminin-1, fibronectin, rADAM 12-cys, or rADAM 12-DC for 30 min, followed by another 30-min incubation in the presence or absence of Mn²⁺. Carcinoma cells adhered tightly on laminin-1 and fibronectin and were well-spread. The addition of Mn²⁺ or β1 integrin–activating antibodies neither increased the number of adherent cells nor the extent of spreading (Fig. 10A, Fig. B, and Fig. I). In contrast, when cells were plated on rADAM 12-cys or rADAM 12-DC they adhered well but were not spread (Fig. 10C and Fig. G). Addition of 1 mM Mn²⁺ or β1 integrin–activating antibodies induced immediate spreading (Fig. 10D, Fig. F, and Fig. H), which was similar to cells plated on laminin or fibronectin. No effect was seen when the control K20 β1 integrin mAb was added (Fig. 10 E).

The data obtained in this study demonstrate that the rADAM 12-cys and rADAM 12-DC serve as potent adhesion substrates for a variety of mesenchymal cells, including lines of fibroblastic, osteoblastic, and myoblastic cell origin. Furthermore, rADAM 12-GFP binds to the cell surface. In the adhesion process, syndecan and integrin cell surface receptors act cooperatively to generate signals for cell adhesion and cell spreading, and for the assembly of focal adhesions and actin stress fibers. Syndecans are involved in cell attachment and β1 integrins in the subsequent cell spreading. Interestingly, carcinoma cells attach but fail to spread on rADAM 12-cys and rADAM 12-DC. Their spreading, however, can be efficiently induced by addition of either 1 mM Mn²⁺ or the β1 integrin–activating mAb 12G10. These data support a model in which the binding of cell surface syndecans with the ADAM 12 cysteine-rich domain leads to an integrin-dependent spreading of cells. Interestingly, in carcinoma cells, the binding to rADAM 12-cys by syndecans does not allow the integrin-mediated cell spreading. The signaling machinery involved in this syndecan–integrin cross-talk is unknown.

Syndecans are abundant cell surface heparan sulfate proteoglycans involved in multiple biological processes by virtue of the binding of their glycosaminoglycans to growth factors, extracellular matrix proteins, cell adhesion molecules, and proteases and protease inhibitors (Carey 1997; Rapraeger and Ott 1998; Woods et al. 1998; Bernfield et al. 1999; Zimmermann and David 1999). It is calculated that heparan sulfate proteoglycans occupy >25% of the cell surface of many cells (Lander 1999). The heparan sulfate chains, which are attached to the ectodomain of the core protein, have a strong negative charge and interact with proteins by forming multiple electrostatic contacts with basic amino acids. This interaction usually leads to relatively low affinity binding (dissociation constants in the range of 2 × 10⁻⁸ – 2 × 10⁻⁵ M) (Lander 1998). In this study, low concentrations of heparin (10–20 μg/ml) completely blocked the interaction of mesenchymal cells with immobilized rADAM 12-cys and rADAM 12-DC, and pretreatment of the cells with heparitinase likewise inhibited the interaction. Moreover, if cells were incubated with chlorate, an inhibitor of sulfation, no adhesion was observed. Finally, cells that do not contain cell surface heparan sulfates do not attach to rADAM 12-cys. This was shown with ARH-77 cells that express very little cell surface heparan sulfate and with several ARH-77 transfectants used as a model system to examine which cell surface heparan proteoglycan supports rADAM 12-cys–mediated cell adhesion. ARH-77 cell lines transfected with either syndecan-1, -2, or -4 attached to rADAM 12-cys, whereas cells transfected with glypican-1 did not, demonstrating that different syndecan members are able to mediate cell adhesion to rADAM 12-cys.

The syndecan family consists of four members, designated syndecans 1–4. Each syndecan shows a specific and distinct expression pattern during development and in adult tissue (Carey 1997; Rapraeger and Ott 1998; Woods et al. 1998; Bernfield et al. 1999; Zimmermann and David 1999). Syndecan-1 is primarily expressed on epithelial cells, syndecan-2 (fibroglycan) on fibroblastic, endothelial, and liver cells, and syndecan-3 (N-syndecan) on neuronal cells. Syndecan-4 (ryudocan, amphiglycan) shows a widespread expression pattern. In addition, syndecan-4 is specifically concentrated into focal adhesions with integrins, and may play an important role in the control of focal adhesion formation and the organization of the actin cytoskeleton (Woods and Couchman 1994; Baciu and Goetinck 1995; Echtermeyer et al. 1999; Longley et al. 1999). However, many cells express more than one syndecan (Kim et al. 1994). To test the idea that one syndecan might be preferentially used by specific cell types in the process of cell adhesion to rADAM 12-cys, we isolated proteins from biotinylated RD cell surfaces that bind to rADAM 12-cys by affinity chromatography. Indeed, syndecan-4 with a core protein of ∼35 kD bound detectably to immobilized rADAM 12-cys. A solid phase binding assay devised previously by Kojima et al. 1996 confirmed the interaction of purified human syndecan-4 with rADAM 12-cys. Thus, syndecan-4 represents a potentially important binding partner for ADAM 12 in the cell adhesion response described here and possibly also in modulating cell behavior in myogenesis (Yagami-Hiromasa et al. 1995; Gilpin et al. 1998) and cancer (Iba et al. 1999).

Syndecans often serve as coreceptors. The classical example is that syndecans supply the glycosaminoglycans necessary for bFGF to bind to its high affinity receptor (Rapraeger et al. 1991). More recently, several reports have pointed to a role for syndecans, in particular syndecan-4, as coreceptors for integrins (Woods et al. 1986; Saoncella et al. 1999). We also show in our study that both syndecans and integrins are involved in rADAM 12-cys–mediated cell attachment and spreading. The initial cell attachment is mediated by syndecans, and subsequent cell spreading by functional β1 integrins. This conclusion is based on the observation that cells treated with function-blocking antibodies to integrin β1 and cells in which the β1 integrin gene was deleted attach to rADAM 12-cys but do not spread. Mesenchymal cells express several β1 integrins. For example, SV-HFO osteoblastic cells express β1, α1–α3 and α5–α6, and αvβ3 (Iba et al. 1996). A combination of function-blocking antibodies reacting with α3, α5, and α6 integrins inhibited cell spreading. Whereas treatment of cells with antibodies in pairs showed partial inhibition of spreading, single antibody treatment had no effect. These data clearly suggest that syndecan and integrins cooperate during cell adhesion and spreading.

Integrins have been identified previously as receptors for the ADAMs and for the closely related snake venom metalloproteases (SVMPs). Several SVMPs interact with αIIbβ3 and α2β1 integrins, and through this binding compete with their natural ligands, causing severe hemorrhage in the bite victim (Jia et al. 1996). In the mouse, ADAM 2 (fertilin β) on the sperm cell surface interacts through the disintegrin-like domain with α6β1 integrin on eggs in the fertilization process (Almeida et al. 1995; Chen et al. 1999a,Chen et al. 1999b). The importance of ADAM 2 in this process is also supported by the finding that mice deficient of ADAM 2 are infertile (Cho et al. 1998). Further evidence for a role of the disintegrin-like domain in cellular interactions was provided by Zhang et al. 1998 and Nath et al. 1999. Thus, the Arg-Gly-Asp (RGD)-containing disintegrin-like domain of ADAM 15 made in E. coli interacts with the αvβ3 integrin (Zhang et al. 1998). ADAM 15 made in COS cells as a chimeric protein containing the extracellular domain of ADAM 15 fused to the Fc portion of human IgG interacts with αvβ3 and α5β1 integrins on different hemopoietic cells. We have demonstrated that rADAM 12-cys and rADAM 12-DC support cell adhesion (Iba et al. 1999; Zolkiewska 1999; this study). Furthermore, we show here that full-length ADAM 12 (rADAM 12-GFP) binds to cell surface syndecans. Based on these results, it appears that the cysteine-rich domain of ADAM 12 comprises most of the syndecan-mediated cell binding activity. Interestingly, the cysteine-rich domain of the hemorrhagic SVMP atrolysin A has been recently demonstrated to inhibit collagen-stimulated platelet aggregation, indicative of an interaction with integrin α2β1 (Jia et al. 2000). Thus, for ADAM 12, and possibly for other ADAMs and SVMPs, the cysteine-rich domain is involved in mediating cellular interactions via syndecans and integrins.

Integrins regulate many cellular events, including shape, adhesion, migration, and survival of cells. The execution of these functions depends on the activation state of integrins. In recent years, several laboratories have shown that integrins can be expressed on the cell surface but are unable to recognize and bind their ligands. Certain signals such as the activation of R-ras (Zhang et al. 1996) and the phosphorylation of integrin cytoplasmic domain–associated protein-1 (ICAP-1) (Bouvard and Block 1998) or the focal adhesion protein VASP (vasodilator-stimulated phosphoprotein) (Aszodi et al. 1999) may change the conformation of integrins and thereby modulate the ability to bind ligands. These changes are referred to as inside-out signaling. The data presented in this paper suggest a model in which the initial binding of ADAM 12 cysteine-rich domain by syndecan on mesenchymal cells triggers a conformational change of the ADAM 12-cys, which may allow integrins to bind ADAM 12-cys and subsequently the cells to spread. This hypothesis is supported by several findings. First, blocking antibodies to integrins inhibit spreading of mesenchymal cells on rADAM 12-cys. Secondly, chondroblasts which lack the expression of β1 integrin fail to spread on rADAM 12-cys. Thirdly, carcinoma cells can tightly bind to rADAM 12-cys, but in contrast to mesenchymal cells, are unable to spread on rADAM 12-cys. The inability to spread can be efficiently overcome by either adding activating antibodies or Mn²⁺, which both can change the conformation of the integrins and allow integrin binding. The inability of carcinoma cells to spread on rADAM 12-cys adds another twist to our findings. It suggests that in carcinoma cells, the binding of rADAM 12-cys to syndecan may lead to an inactivation of integrins through syndecan-mediated signals that consequently does not allow the integrins to bind to rADAM 12-cys. Alternatively, the binding of rADAM 12-cys to syndecan may fail to signal positively to the integrins in the carcinoma cells. Several questions need to be addressed in the future. It will be interesting to see whether the cross-talk between syndecans and integrins is only operating in transformed carcinoma cells or also in primary epithelial cells. Furthermore, we do not know why syndecan signaling to integrins is different between mesenchymal cells and carcinoma cells. Work addressing these interesting questions is currently in progress in our laboratory.

We thank Brit Valentin and Aase Valsted for technical assistance; Merton Bernfield, Guido David, Arthur D. Lander, and Alan Rapraeger for antibodies to syndecans and glypican; Martin Humphries and Arthur Mercurio for antibodies to integrins; Marina Glukhova for antibodies to vinculin; and Phillip Sharp for retrovirus reagent.

This work was supported by grants from the Danish Cancer Society, the Danish Medical Research Council, and by the Neye, VELUX, Novo-Nordisk, Haensch, Munksholm, Thaysen, Beckett, Hartmann, Fonden af 1870, Kjær, Frænkel, and Meyer Foundations. Ralph D. Sanderson is supported by the National Institutes of Health (CA68494). Ulla M. Wewer is a Professor of the Danish Medical Research Council and the Neye Foundation.