The Metalloprotease Disintegrin ADAM8

ADAMs (a disintegrinand metalloprotease domains) are metalloprotease and disintegrin domain-containing transmembrane glycoproteins with proteolytic, cell adhesion, cell fusion, and cell signaling properties. ADAM8 was originally cloned from monocytic cells, and its distinct expression pattern indicates possible roles in both immunology and neuropathology. Here we describe our analysis of its biochemical properties. In transfected COS-7 cells, ADAM8 is localized to the plasma membrane and processed into two forms derived either by prodomain removal or as remnant protein comprising the extracellular region with the disintegrin domain at the N terminus. Proteolytic removal of the ADAM8 propeptide was completely blocked in mutant ADAM8 with a Glu330 to Gln exchange (EQ-A8) in the Zn2+ binding motif (HE330LGHNLGMSHD), arguing for autocatalytic prodomain removal. In co-transfection experiments, the ectodomain but not the entire MP domain of ADAM8 was able to remove the prodomain from EQ-ADAM8. With cells expressing ADAM8, cell adhesion to a substrate-bound recombinant ADAM8 disintegrin/Cys-rich domain was observed in the absence of serum, blocked by an antibody directed against the ADAM8 disintegrin domain. Soluble ADAM8 protease, consisting of either the metalloprotease domain or the complete ectodomain, cleaved myelin basic protein and a fluorogenic peptide substrate, and was inhibited by batimastat (BB-94, IC50∼50 nm) but not by recombinant tissue inhibitor of matrix metalloproteinases 1, 2, 3, and 4. Our findings demonstrate that ADAM8 processing by autocatalysis leads to a potential sheddase and to a form of ADAM8 with a function in cell adhesion.

ADAM 1 (a disintegrin and metalloprotease domain) proteins constitute a family of transmembrane glycoproteins and serve essential physiological roles in fertilization, myogenesis, and neurogenesis. These functions are due to distinct protein domains involved in cell-cell fusion, cell-cell interaction, or proteolysis of membrane proteins, a process termed ectodomain shedding (1). To date, the family of ADAM proteinases comprises more than 30 members in different species (2,3), and 24 ADAM genes were found in the mouse genome. Fourteen of the murine ADAMs contain the catalytic consensus sequence HEXXHXXGXXHD in their metalloprotease domains and are therefore predicted to be proteolytically active (4). The cleavage of myelin basic protein (MBP) by ADAM10/MADM was the first demonstration of proteolysis by ADAMs (5). The tumor necrosis factor-␣ convertase (ADAM17) was purified on the basis of its ability to cleave tumor necrosis factor-␣ (6, 7) and a number of other peptide and protein substrates in vitro (1,8). Proteolysis of membrane-bound surface molecules was also demonstrated for heparin-binding epidermal growth factor (9) and amyloid precursor protein (10,11), which are cleaved by ADAMs 9 and 10, respectively.
Catalytically active ADAMs are usually activated by furincatalyzed removal of the prodomain or by other proprotein convertases. For cleavage by furin-like convertases or prohormone convertases, such ADAMs possess a consensus sequence RX(K/R)R localized between the pro-and metalloprotease domains. This type of activation has been demonstrated for a number of ADAM proteases, among them ADAMs 9 (12), 10 (13), 12 (14), 15 (15), and 17 (16). Two members of the ADAM family in the mouse, ADAM8 (17) and ADAM28 (18), do not contain the consensus sequence for activation by furin-like proteases. In ADAM28, a recently cloned ADAM, a point mutation from glutamate 343 to alanine in the catalytic domain prevents prodomain removal, suggesting an autocatalytic mechanism for the protease activation (18). Up to now, the mechanism of activation of ADAM8 is unknown.
In matrix metalloproteases and in the snake venom metalloprotease Adamalysin II, the free sulfhydryl of the cysteine switch residue in the prodomain is thought to bind to the Zn 2ϩ ion in the catalytic site. This interaction between the prodo-main and the active site appears to inhibit the catalytic site and acts as an intramolecular chaperone, allowing the pro-and the metalloprotease domain to fold properly during transport through the secretory pathway (19).
In addition to their proteolytic properties, ADAMs are able to mediate cell-cell interactions via their disintegrin/cysteine-rich domains. A role in cell adhesion has been demonstrated for ADAM9, ADAM12, and ADAM15 (20 -22). The recombinant disintegrin/cysteine-rich domain of ADAM12 mediates cell adhesion (23). Human ADAM15 is the only metalloprotease-disintegrin containing an RGD sequence within the integrin-binding loop of the disintegrin domain (24), and this has been shown to bind to the integrins ␣ v ␤ 3 and ␣ 5 ␤ 1 (21,25). However, although necessary for integrin binding, the tripeptide alone is not sufficient to determine specificity for cell adhesion; rather, the overall loop structure within the integrin binding loop forms the determinant for integrin binding (21). Two ADAMs, ADAM22 and 23, lacking the conserved catalytic motif are highly expressed in the CNS. Deficiency in ADAM23 leads to an embryonic lethality (26), whereas a knockout of ADAM22 causes ataxia (27), demonstrating that these ADAMs play essential roles in the CNS.
ADAM8 was originally cloned as MS2 or CD156 from mouse macrophages (17). It is expressed in macrophages, neurons, and oligodendrocytes and is up-regulated in the CNS following neurodegeneration and subsequent activation of glia cells, astrocytes, and microglia, suggesting that ADAM8 plays a role in neuron-glia interactions (28). In rats immunized with a recombinant ADAM8 disintegrin domain, experimental autoimmune encephalomyelitis induced by myelin basic protein fragments is significantly ameliorated (29). ADAM8 has also been implicated in the differentiation of osteoclasts, a process involving cell-cell fusion (30). These results suggest a role of ADAM8 in cell adhesion and cell fusion; however, the mechanisms of these activities are unknown. In the current study of ADAM8 we present the biochemical properties of ADAM8 with respect to catalytic activity, autocatalytic processing, and cell adhesion.
ADAM 8 Antibodies-The catalytic domain of murine ADAM8 (amino acids 190 -398) with a C-terminal histidine (HIS) tag was cloned using the Escherichia coli expression vector pRSETA. ADAM8catHIS was expressed in E. coli strain BL21 and purified from inclusion bodies using nickel-Sepharose according to manufacturer's instructions (Qiagen). After removal of urea by dialysis against PBS, the renatured protein was used to generate polyclonal rabbit antisera. Antibodies were affinity-purified by coupling the recombinant ADAM8catHIS to a HiTrap NHS column, passing 5 ml of serum over the column, washing, and eluting specific antibodies with 0.1 M glycine, pH 2.5. Eluted antibodies were neutralized with 1 M Tris-HCl, pH 8, and then dialyzed into PBS. The disintegrin domain (DD) of ADAM8 (amino acid residues 406 -482) was cloned using the E. coli expression vector PGEX-2T. The resulting GST-A8DD fusion protein was expressed in E. coli strain BL21 after induction with 1 mM isopropyl-1-thio-␤-D-galactopyranoside at 25°C for 6 h. E. coli lysates were prepared and the GST fusion protein was purified on a glutathione agarose column. Peak fractions were analyzed by SDS-PAGE and dialyzed against PBS. The resulting protein was used to generate a polyclonal rabbit antisera. Antibodies were affinity-purified by coupling the recombinant GSTA8DD protein to a HiTrap NHS column, passing 5 ml of serum over the column, washing, and eluting with 0.1 M glycine, pH 2.5. Eluted antibodies were neutralized with 1 M Tris-HCl and then dialyzed into PBS. To remove anti-GST antibodies and cross-reacting antibodies, the affinity-purified antibody was passed over an additional HiTrap NHS column to which a similarly expressed and purified ADAM15 GST-DD had been coupled. The flowthrough was collected and shown to be specific for GSTA8DD compared with the ADAM15 GST-DD in an enzyme-linked immunoabsorbance assay. Affinity-purified antibodies were stored in aliquots at Ϫ20°C. Antibody a-CTD was generated as described (28) and was purified by protein A-Sepharose chromatography.
Cell Culture-1321N1 astrocytoma cells were obtained from European Collection of Cell Cultures and were grown in Iscove's modified Dulbecco's medium (Invitrogen, Groningen, Netherlands) containing 5% FCS and 1% glutamine. COS-7 cells were grown in Dulbecco's modified Eagle's medium in the presence of 10% FCS and 1% glutamine. Transient transfections were performed with LipofectAMINE (Invitrogen) or by the calcium phosphate technique. Stable 1321N1 transfectants were selected with 400 g/ml G418.
For introducing the point mutation glutamate to glutamine (E330Q), the amplification product of the primer pair A8fw and EQA8 (5Ј-CCA GGT TGT GGC CCA GCT GAT G-3Ј) was used as a sense primer in conjunction with the A8rev primer for amplification of full-length EQ-A8cF or primer A8MP for the soluble EQ-A8MP construct. For generation of the construct A8-⌬DD, the amplification products of the primer pair A8fw and A8⌬DD, 5Ј-TCT AGA CAC GAA CCG GTT GAC ATC TGG-3Ј, as well as sA8⌬DD, 5Ј-TCT AGA TGC CCA GGG GGC TAC TGC TTT-3Ј, and A8rev were subcloned in the pCRII vector (Invitrogen), and the fragments were ligated together by the extra XbaI site, thereby introducing two additional residues, serine and arginine.
Immunofluorescence-Cells grown on collagen-coated coverslips were fixed with ice-cold methanol for 5 min at room temperature. To detect ADAM8, we used polyclonal antibodies against the cytoplasmic domain (a-CTD, 1:500 in PBS) (28), the disintegrin domain (a-DD, 1:200), or an antibody against the catalytic domain of ADAM8 (a-MP, 1:200) for 1 h at 37°C or overnight at 4°C. As secondary antibody, we used a goat anti-rabbit antibody coupled to Cy3 (Dianova, Hamburg, Germany). Fluorescent stainings were visualized using an Axiophot fluorescence microscope or by confocal laser microscopy (TCS SP2, Leica, Germany). Images were further processed with Adobe Photoshop 6.0.
Cell Surface Biotinylation and Immunoprecipitation-Transfected COS-7 cells were grown to confluency on 90-mm plates, washed with PBS at 4°C, and incubated for 45 min with the non-membrane-permeant biotinylation reagent NHS-LC biotin (Pierce, Bonn, Germany) at room temperature. After washing with 0.1 M glycine in PBS, the cells were lysed in RIPA buffer plus protease inhibitors. The lysates were subjected to immunoprecipitation with anti-Bi-Pro antibody and protein G-Sepharose (Sigma) (32). After elution with 2ϫ Laemmli buffer, the samples were applied on SDS-PAGE gels, and the ADAM8-specific bands were detected by Western blotting and staining with a-CTD (1:1000) or streptavidin-HRP (1:1000, Roche Molecular Biochemicals).

Expression of the Recombinant Disintegrin/Cys-rich/EGF-like (DC) Domain of ADAM8 -
The cDNA fragment encoding the A8 DC domain was generated by PCR using Platinum TM Pfx polymerase (Invitrogen) with the following primers: DCE-A8f, 5Ј-GGT GGC CCT GTG TGT GGA AAC-3Ј; DCE-A8r, 5Ј-TAC ACA GTT GGG TGG TGC CCA-3Ј. The resulting cDNA fragment was cloned into the bacterial expression vector pTrcHis2 (Invitrogen) containing a C-terminal Myc and His 6 tag. This vector was transformed into E. coli strain TOP10 (Invitrogen). Recombinant protein expression was induced with 1 mM isopropyl-1thio-␤-D-galactopyranoside for 5 h to overnight. Purification of the recombinant DCE domain was done using the Xpress TM system protein purification kit (Invitrogen) according to the manufacturer's instructions for native protein preparations.
Purification of Soluble Mouse ADAM8 Catalytic Domain-COS-7 cells were transiently transfected with constructs encoding various forms of C-terminally FLAG-tagged soluble catalytic domain lacking the transmembrane domain (see Fig. 1). The cells were grown to confluency, and 24 h before harvest, growth medium was exchanged by serum-free Dulbecco's modified Eagle's medium. The supernatants were collected and concentrated by centrifugation on Amicon YM-30 (Millipore) columns. The concentrated supernatants were further purified by affinity purification with ConA-Sepharose or as Fc fusion protein by protein A-Sepharose (Sigma). The purity of soluble proteases was confirmed by SDS-PAGE, and concentrations were determined using the BCA reagent. Further supplies of the ADAM8 catalytic domain expressed as an Fc fusion were generated essentially as described previously for ADAM17 (33). PCR-amplified DNA encoding the preprocatalytic domain was inserted upstream of an enterokinase cleavage site ([V 405 in ADAM8 V]DDDDK Ϫ ) followed by the human IgG1 heavy chain constant region, hinge, CH2, and CH3 in vector pEE12. Following stable transfection of NSO myeloma cells (31), conditioned media were incubated with protein A-Sepharose from which ADAM8 was eluted following washing and cleavage with recombinant enterokinase (33).
Protease Assays-Purified A8 proteins were incubated with 1-10 g Prior to electrophoresis, cell lysates were deglycosylated with EndoH (ϩ) or PNGaseF (ϩ) or mock-incubated for control (Ϫ). The primary antibody was a-FLAG. B, cell-surface biotinylation of ADAM8. COS-7 cells were transfected with the constructs pTarget (mock) or A8cF. After biotinylation and immunoprecipitation with a-F antibody, cell lysates were blotted and stained either with anti-Streptavidin-HRP (a-Str-av, 1:1000) or with a-A8-CTD (1:1000), as indicated below. the range 1 nM to 30 M. IC 50 values were calculated using GraphPad Prism. Data were fitted by using non-linear regression analysis.
N-terminal Sequence Analysis-For preparation of sequencing samples, cell lysates from ten 150-mm plates of COS-7 cells transfected with the cDNA encoding the complete A8 protein were lysed with RIPA buffer (see above). After ConA-Sepharose purification and immunoprecipitation with anti-FLAG antibody 4A6 (31), samples were run on an SDS-PAGE gel in borate buffer (50 mM H 3 BO 3 , pH 9.0, 20% ethanol, 1 mM dithiothreitol). After electrophoresis, samples were electroblotted onto polyvinylidene difluoride) membranes (Porablot, Macherey and Nagel, Dü ren, Germany) and stained with either Coomassie Brilliant Blue or anti-A8-CD. For N-terminal sequencing, ADAM8 bands were cut out and applied on a protein sequencer (Knauer, Germany). Usually, 5 cycles of automated Edman degradation were run to obtain sufficient sequence information.
Cell Adhesion Assays-Cell adhesion was tested essentially as described previously (23). Briefly, 96-well plates were covered with the indicated amount of recombinant protein in PBS at 4°C for 16 h. After blocking with bovine serum albumin for 1 h, 10 5 cells in PBS or medium (with 5% FCS) were seeded onto the plates. For blocking experiments, cells were incubated prior to seeding with a-DD (10 g/ml) for 15 min at room temperature. After 1-h incubation at 37°C, the wells were rinsed 3ϫ with PBS, and the remaining cells were quantified by counting 10 randomly chosen viewing fields (100-fold magnification) or by quantification through SDS-PAGE. The 100% value of cell adhesion was obtained by allowing 10 5 cells to adhere completely. Transiently transfected cells were co-transfected with a GFP vector, and the amount of GFP ϩ cells was counted in comparison to the total number of GFP ϩ cells seeded into control wells.

Expression and Maturation of Mouse ADAM8 in COS-7
Cells-The construct A8cF (Fig. 1) contains the complete cDNA sequence of mouse ADAM8 with 826 amino acids. For immunodetection and affinity purification, A8cF contains an additional FLAG sequence (F) of 10 amino acids derived from birch pollen profilin ("BiPro-flag") (31).
Upon expression of the full-length construct A8cF in COS-7 cells, three bands representing ADAM8 protein were observed in cell lysates: 1) a proform of M r 120,000; 2) a processed form with an M r of 90,000, which is consistent with propeptide removal; 3) a "remnant" form of ADAM8 protein with an apparent molecular weight of ϳ60,000 (Fig. 1B). The remnant ADAM8 protein has been detected even in the permanent presence of protease inhibitors during the preparation of the cell lysates and was found to be abundant in all tissues and cells expressing ADAM8 investigated to date (28). As expected, a-MP detected only pro-ADAM8 and the 90-kDa processed form (Fig. 1B), but not the remnant ADAM8 protein, which lacks the catalytic domain. The asterisk (weak band immediately below pro-A8), ADAM8 protein weakly processed at a furin-cleavage site in the prodomain (position, amino acids 39 -42) serves as internal control for dec-RVKR-CMK inhibition. B, expression of A8 and A8 mutant EQ-A8cF in COS-7 cells, analyzed in cell lysates (left half). Soluble A8 constructs were detected in COS-7 supernatants (right half), monitored by anti-FLAG antibody. Wild-type ADAM8 constructs give rise to proform (black arrowhead) and processed forms of ADAM8 (white arrowheads), whereas the EQ mutant is only detectable as proform. Soluble forms of A8, detected in COS-7 supernatants: A8EC, complete A8 ectodomain; A8MP, comprising Pro and MP domain. C, analysis of prodomain removal from EQ-A8cF with different ADAM8 variants by co-transfection or by A8-EC added to serum-free medium (ϩ). For cotransfection, DNA was transfected in equimolar ratios between EQ-A8cF and ADAM8 variants. Note that only A8-EC was able to remove the prodomain of EQA8cF, whereas ϳ100 g/ml recombinant A8-EC in the medium did not process EQ-ADAM8 synthesized by the cells.
To investigate the maturation of ADAM8, we performed Western blot analysis with an EndoH-treated sample of ADAM8 expressing COS-7 cells. Only pro-ADAM8 was sensitive to EndoH, whereas the mature and the remnant form of ADAM8 were EndoH-resistant. An identical sample cleaved with PNGaseF, which removes most or all carbohydrate moi-eties, showed a shift of all three ADAM8 bands, demonstrating N-glycosylation ( Fig. 2A). From surface-biotinylated COS-7 cells, only the processed forms of ADAM8 with molecular masses of 90 and 60 kDa were immunoprecipitated (Fig. 2B), indicating that prodomain removal occurs intracellularly. Taken together these results suggest that the observed processing of ADAM8 occurs by prodomain removal during passage through the secretory pathway.
Processing of ADAM8 -ADAM8 contains a non-perfect consensus cleavage sequence for furin (RETR in amino acid position 193). A specific inhibitor of furin-like proteases (decanoyl-RVKR-chloromethylketone) did not to inhibit the processing of A8cF in COS-7 cells, even in high concentrations of 50 M (Fig.  3A). This finding suggested that prodomain removal may be autocatalytic or may depend on other proteases. To test whether prodomain removal of ADAM8 is autocatalytic, we constructed an ADAM8 expression vector with an Glu 330 to Gln amino acid exchange in the Zn 2ϩ binding HE 330 XXHXXGXXH consensus motif of the catalytic domain (EQ-A8cF). When mutant EQ-A8cF was expressed in COS-7 cells, only the unprocessed proform of ADAM8 was detected (Fig. 3B), consistent with a failure of the prodomain to be removed. This is in contrast with the wild type ADAM8 construct, which undergoes processing, and supports the notion of an autocatalytic propeptide removal mechanism.
To define the requirements for autocatalytic processing in ADAM8, we co-transfected EQ-A8cF with several ADAM8 constructs (Fig. 3C). A8-EC comprised the ADAM8 ectodomain; A8-MP contains the pro-and catalytic domains of ADAM8. The A8-EC construct retains the disintegrin and cysteine-rich domains and is completely processed with a molecular mass (ϳ60 kDa) that is consistent with propeptide removal. A8-MP comprises only the pro-and catalytic domains and is partially processed (50% at most) to yield the prodomain (ϳ50 kDa) and catalytic domain (ϳ34 kDa, Fig. 3B). We performed co-trans- fections with EQ-A8cF and these ADAM8 constructs (Fig. 3C). In cell lysates, prodomain removal in EQ-A8cF was not detected after co-transfection of A8-MP, but with A8-EC. In addition, cells expressing EQ-A8cF were incubated with medium containing high concentrations of A8-EC (Fig. 3C, right lane).
Prodomain removal in EQ-A8cF was not observed when A8-EC was applied to the cells in a concentration of 100 g/ml (proteolytic activity was confirmed by simultaneous MBP cleavage, see Fig. 7). These experiments demonstrate that catalytically inactive ADAM8 (EQ-A8cF) can be processed only in cis and and non-reducing (Ϫ) conditions, detected with a-DD or a-myc antibody, directed against the disintegrin domain of ADAM8 or the myc-tag, respectively. B, COS-7 cells co-transfected with ADAM8 constructs (A8cF and EQ-A8cF) and GFP (ϳ50% of cells successfully transfected) were used for cell adhesion assays. Briefly, 10 5 cells in PBS were seeded into 96-well plates coated with various amounts of the recombinant DC domain: white, no recombinant protein; black, 5 g/ml; hatched, 20 g/ml; stippled, 50 g/ml. After 1 h, numbers of attached GFP-positive cells were determined by PAGE and subsequent Coomassie Blue staining and by counting 10 randomly chosen viewing fields under the microscope (ϫ100 magnification). Cell numbers were normalized to completely attached GFP ϩ cells (ϭ 100%). Note the reduction in cell adhesion for EQ-ADAM8, due to a lower amount of cell surface localized protein. C, parental 1321N1 human astrocytoma cells or a stable cell clone expressing A8cF (1321N1-A8) were seeded in the coated wells, and the number of cells was quantified by cell counting as described in B. Values were obtained by normalizing the cell number to the cell number obtained after complete attachment of the cells in control plates after 4 h. All data are shown as mean values of three independent samples. only by the complete ectodomain but not by the entire catalytic domain of ADAM8, as long as the two proteins are co-expressed in the same cell. The observation that the soluble ADAM8 MP domain was not able to process EQ-ADAM8 suggests that the disintegrin-cysteine-rich and EGF domain of ADAM8 might play a role in autocatalysis, e.g. by mediating protein-protein interactions between ADAM8 monomers.
Cellular Localization of ADAM8 Proteins-The localization of A8cF and EQ-A8cF was analyzed by cell surface biotinylation and subsequent immunoblotting (Fig. 4). Only weak unspecific signals were detected with mock vector pTarget, when streptavidin-horseradish peroxidase was used. For ADAM8 protein, a small amount of proform, some mature, and mostly the remnant form was detected. When expressing EQ-A8cF, only the proform is detected on the cell surface.
The ADAM8-specific antibody a-CTD (see Fig. 1A) was used in permeabilized transfected COS-7 cells for immunodetection of full-length wild-type ADAM8 (A8cF) and EQ-A8cF (Fig. 5). In cells with low expression levels of ADAM8 and EQ-ADAM8, a membrane and vesicular staining pattern could be observed (Fig. 5, A, C, D, and F). In cells expressing high levels of ADAM8 (Fig. 5, B and E), there is increased staining in components of the secretory pathway and the plasma membrane. For EQ-ADAM8, the staining pattern resembles that of wildtype ADAM8 (Fig. 5, D-F). In contrast, a mutant of ADAM8 lacking the disintegrin domain was retained in the ER (Fig. 5, G-I). From these observations, we conclude that processing is not required but facilitates membrane localization of ADAM8.
Activity of the ADAM8 Disintegrin/Cys-rich/EGF Domain in Cell Adhesion-The processing of ADAM8 yields a predominant remnant form of the truncated protein on the cell surface under physiological conditions. N-terminal sequence analysis confirmed that the processing event occurred within the sequence RFV 401 2GGP located between the catalytic domain and the disintegrin domain. Given that this truncated structure is comprised of the disintegrin/Cys-rich/EGF domains, it could be that these domains play a role in cell adhesion. To explore this idea further we synthesized a recombinant protein ADAM8-DC consisting of the disintegrin domain/cysteine-rich and EGF-like (DC) domain of ADAM8 from residues 404 to 642 as a fusion protein in E. coli. The recombinant protein is localized in the cytoplasma and was extracted under native conditions. The apparent molecular mass of the monomer is approx-imately 31 kDa under non-reducing conditions and slightly larger (ϳ32 kDa) under reducing conditions (Fig. 6A). This recombinant protein was used to coat culture dishes at concentrations between 2 and 50 g/ml. To demonstrate the correlation between processing of pro-ADAM8 and cell adhesion, COS-7 cells were co-transfected with A8cF and EQ-A8cF and a GFP vector (pCMV-GFP-N3). Successfully transfected COS-7 cells (as indicated by GFP expression), which adhered to wells covered with the recombinant DC domains, were counted and compared with the total amount of transfected cells. COS-7 cells transfected with A8cF showed a dose-dependent adhesion. Cell adhesion by COS-7 cells expressing EQ-A8cF was observed, although reduced (Fig. 6B). This is probably due to the lower amount of surface-localized ADAM8 protein (see Fig. 4).
In addition, the adhesive properties of the human astrocytoma cell line 1321N1 stably expressing mouse ADAM8 (1321N1-A8) was compared with those of the non-transfected 1321N1 cells. We observed a significant dose-dependent adhesion of 1321N1-A8 cells to plates covered with the ADAM8-DC in comparison to the non-transfected cells (Fig. 6C). This adhesion required non-reducing conditions and was blocked by 100 mM ␤-mercaptoethanol, indicating that the formation of cysteine bridges is essential for cell adhesion by the ADAM8-DC domains. In addition, preincubation of 1321N1-A8 cells with 50 g/ml recombinant DC or with an excess of antibody a-DD (ϳ100 ng/ml) abolished cell adhesion to the ADAM8-DC domains (Fig. 6C).
These results demonstrate that a homophilic interaction of the ADAM8-DC domains with surface-localized ADAM8 is suf- FIG. 8. Proposed model for ADAM8 processing and function. Following passage through the medial Golgi, pro-ADAM8 is processed by autocatalysis in the region between the prodomain and the metalloprotease domain. In this processed form, ADAM8 is transported to the cell surface where it was detected in a significant amount and where it can act as a sheddase (indicated by the question mark). A major proportion of ADAM8 is further processed by MP domain removal leading to a remnant form of ADAM8 mediating cell adhesion and probably fusion. Extracellular MP domain removal cannot be excluded (indicated by the question mark), although we did not detect a soluble MP domain in supernatants of cells transfected with A8cF. Despite the lack of processing by prodomain removal in EQ-ADAM8, a significant amount of unprocessed protein is transported to the cell surface. Thin arrows indicate processing steps of ADAM8, whereas thick arrows indicate ADAM8 functions. ficient for cell adhesion. Moreover, processing of ADAM8 by prodomain removal is not a prerequisite for ADAM8-mediated cell adhesion.
Proteolytic Activity of ADAM8 -The two forms of soluble ADAM8 protease (A8-EC and A8-MP) were isolated from supernatants of transfected COS-7 cells and purified by affinity chromatography with a-BiPro antibody or with protein A-Sepharose as Fc fusion protein. The protein preparations were used in equal amounts for cleavage assays with myelin basic protein (MBP) as a substrate (Fig. 7, A and B). Supernatants from vector pTarget-transfected cells served as controls for the possible presence of other proteases secreted by COS-7 cells. With control supernatant, cleavage of MBP was Ͻ5% of that obtained with ADAM8-transfected cells. When ADAM8 protease was incubated with MBP, the cleavage resulted in products with approximate molecular masses of 9 and 11 kDa. Mutant A8-MP (Glu to Gln exchange within the HEXXH motif) did not cleave MBP. We tested metalloproteinase inhibitors for their potential to inhibit ADAM8 protease activity. Broadrange inhibitors such as the Zn 2ϩ chelators, 1,10-ortho-phenanthroline (10 mM), the Ca 2ϩ chelator EGTA (10 mM), and BB94 (batimastat, 1 M) completely inhibited ADAM8 activity. In contrast, none of the recombinant tissue inhibitors of metalloproteases, TIMPs 1, 2, 3, and 4, did so (Fig. 7B), even at concentrations up to 500 nM. All the TIMPs were proven to be biologically active by their inhibition of other MMPs (not shown).
To define the major ADAM8 cleavage site in MBP, we analyzed the resulting MBP fragments by MALDI time-of-flight mass spectrometry after trypsin cleavage. ADAM cleavage of MBP was found to occur within the sequence: TTHY-GSLP2QKAQGQ. From this information a modified peptide substrate was derived. The resulting quenched fluorescent peptide: Suc-H-(Mcp)-GSLPQKSH-K(Dpa)-R-amide, was first used to confirm the catalytic activity of soluble A8-EC and A8-MP.
We then undertook a more detailed kinetic analysis using ADAM8 that had been affinity-purified from the conditioned media of stable transfectants of the mouse myeloma cell line NSO. This material is comprised of the ADAM8 catalytic domain alone and results from a concentration step on an affinity column, which first leads to the autocatalytic removal of the propeptide. In a second step, amino acids that comprise an affinity tag (located downstream of the catalytic domain from Val 405 ) are removed by cleavage with enterokinase. N-terminal amino acid sequencing was first used to determine that the prodomain removal had occurred at QPR 187 2NWL. This is immediately upstream of where sequence alignments place the start of the catalytic domain and is very close to two propeptide processing sites determined in the COS-7 cell generated material (LGP 176 2RAL for A8-MP and ALE 180 2IYR for A8-EC, determined by N-terminal sequencing). Use of the quenched fluorescent peptide with the purified ADAM8 catalytic domain confirmed the enzymatic properties first detected with the COS-7 cell-derived soluble forms of ADAM8 (Fig. 7C). Again, ADAM8 activity was blocked only by the metalloproteinase inhibitors 1,10-ortho-phenanthroline and BB-94 but not by TIMPs 1, 2, 3, and 4 (not shown). For BB-94, the IC 50 value was 51.3 Ϯ 0.7 nM (Fig. 7D). DISCUSSION We have demonstrated proteolytic activity of ADAM8 using two substrates, myelin basic protein (MBP) and ADAM8 itself. There was a difference observed for soluble ADAM8 proteases A8-MP versus A8-EC: although cleavage of MBP is equally efficient with both ADAM8 proteases, the proteolytic removal of the ADAM8 propeptide was more efficient in A8-EC compared with A8-MP, arguing for a role of the remaining domains in A8-EC (disintegrin/Cys-rich/EGF domains) for prodomain removal. Lack of inhibition of soluble ADAM8 proteases by any of the recombinant TIMPs 1, 2, 3, and 4 is in contrast to some ADAMs whose proteolytic activities have so far been characterized: they are inhibited by either TIMP1 (ADAM10) (33) or TIMP3 (ADAMs 10, 12, and 17) (33)(34)(35), whereas ADAM8 and ADAM9 are not inhibited (36). It is not known how proteolysis of those proteases is regulated in vivo. One possibility would be a hitherto unidentified TIMP. In the mouse genome data base, there are 59 expressed sequence tags (expressed sequence tags) with high homology to TIMP proteins (Unigene, NCBI, National Institutes of Health).
For several ADAMs (e.g. ADAM 10, 17, and 9), it has been demonstrated that furin-mediated cleavage leads to prodomain removal and activation of the protease. In ADAM9, mutation of Glu to Ala in the HEXXH motif blocked catalytic activity, but not processing (12), demonstrating that prodomain removal and catalytic activity are two distinct molecular events. Furin inhibitor studies indicated that prodomain removal in ADAM8 is mediated by a metalloproteinase and not catalyzed by furinlike proteases. We observed three different but very closely positioned cleavage sites within the propeptide, depending on precisely which form of soluble ADAM8 was analyzed. This indicates a susceptible region for propeptide processing, rather than a single specific site. Independent of the precise cleavage site, a cysteine-switch activation mechanism (37) can be proposed, as all three cleavage sites are downstream of the putative propeptide Cys 164 residue (by homology alignments with other ADAMs and MMPs). Given the facts that we observed processing in all soluble ADAM8 proteins and that in all cases the E330Q version failed to process, we conclude that the propeptide processing in ADAM8 is autocatalytic. Two arguments support a role for the disintegrin/Cys-rich domain in autocatalysis: 1) prodomain removal is more efficient in A8-EC compared with A8-MP; and 2) prodomain removal of EQ-ADAM8 works only with A8-EC, but not with A8-MP. An influence of the disintegrin domain on (furin-mediated) prodomain removal has been described for tumor necrosis factor-␣ convertase (38). Deletion of the disintegrin domain was not suitable to address this question, because the mutant A8-⌬DD protein was misfolded and retained in the ER.
Processing of ADAM8 into a remnant protein suggested a role for ADAM8 in cell adhesion. This remnant protein is abundant in all cell types and tissues expressing ADAM8 (28). The E330Q mutation also blocks the formation of this remnant protein, implying that, like propeptide removal, the generation of the remnant polypeptide could also be autocatalytic. However, we cannot distinguish whether this is because of a need for active enzyme or prodomain removal or both. In the case of the fertilins ␣ and ␤, processing to the truncated remnant form during spermiogenesis has been observed and is thought to be brought about by serine proteases (39): we cannot rule out a participation of serine proteases during ADAM8 processing.
Cell adhesion by ADAM8 is mainly mediated by remnant ADAM8 polypeptide, which is therefore the major determinant for cell adhesion and/or cell fusion (e.g. in osteoclasts) (30). The adhesion function of ADAM8 may be particularly important in neuropathology, e.g. in neurodegeneration or multiple sclerosis (MS). Studies on neurodegeneration in the Wobbler mutant of the mouse indicated that ADAM8 has a role in pathophysiological neuron-glia interactions (28).
With respect to MS, there was evidence for a role of ADAM8 in experimental autoimmune encephalitis (29), an animal model for MS. In a study, vaccination of rats against encephalitis was achieved by a recombinant ADAM8 disintegrin domain. We suspect from our experiments that this effect could be due to the cell adhesion function of ADAM8 rather than to proteolytic activity. There is recent experimental evidence for a sheddase function of ADAM8 by in vitro and in vivo cleavage of a cell adhesion molecule, demonstrating a role for ADAM8 in the nervous system. 2 From our data, we propose a model for processing and localization of ADAM8 (Fig. 8), by which ADAM8 is consecutively processed into two forms, both localized to the cell surface: a potential sheddase derived by prodomain removal and a remnant cell adhesion protein derived by further processing of ADAM8. Because the ADAM8 remnant form was never observed in EQ-ADAM8, we conclude that processing by prodomain removal precedes further processing of ADAM8, independent of the mechanism of this processing. As demonstrated by expression of EQ-ADAM8, cell adhesion is not dependent on processing of pro-ADAM8, but processing is expected to be necessary for proteolytic activity.