ADAMTS-1 Is an Active Metalloproteinase Associated with the Extracellular Matrix*

Cellular disintegrin and metalloproteinases (ADAMs) are a family of genes with a sequence similar to the snake venom metalloproteinases and disintegrins. ADAMTS-1 is a unique ADAM family protein with respect to the presence of thrombospondin type I motifs and the capacity to bind to the extracellular matrix. Because ADAMTS-1 has a potential zinc-binding motif in the metalloproteinase domain, we examined in this study whether ADAMTS-1 is an active metalloproteinase by means of the proteinase trapping mechanism of α2-macroglobulin. We found that the soluble type of ADAMTS-1 protein is able to form a covalent-binding complex with α2-macroglobulin. Furthermore, the point mutation within the zinc-binding motif of ADAMTS-1 protein eliminates its capacity to bind to α2-macroglobulin. These data demonstrate that the metalloproteinase domain of ADAMTS-1 is catalytically active. In addition, we showed that the removal of the pro-domain from the ADAMTS-1 precursor is impaired in the furin-deficient cell line, LoVo, and that the processing ability of the cells is restored by the co-expression of the furin cDNA. These data provide evidence that the ADAMTS-1 precursor is processed in vivo by furin endopeptidase in the secretory pathway. Consequently, ADAMTS-1 is an active metalloprotease that is associated with the extracellular matrix. This study strongly suggests that ADAMTS-1 may play a role in the inflammatory process through its protease activity.

Disintegrin and metalloproteinases (ADAMs) 1 represent a new family of gene products that show a significant sequence similarity to snake venom metalloproteinase and disintegrin (1,2). Most of ADAMs are membrane-anchored glycoproteins that are comprised of a pro-domain, a metalloprotease-like domain, a disintegrin domain, a cysteine-rich region, an epidermal growth factor repeat, a transmembrane region, and a cytoplasmic domain. Fertilin, the first ADAM described, has been implicated in integrin-mediated sperm-egg binding (3,4). Subsequently, meltrin has been shown to be involved in muscle fusion (5). Therefore, ADAMs are thought to play a role in cell-cell interactions. In contrast, tumor necrosis factor ␣ (TNF-␣)-converting enzyme (TACE), which cleaves the membraneanchored precursor of TNF-␣ and releasing TNF-␣, also belongs to the ADAM family (6,7). In addition, TACE was shown to be involved in the processing of transforming growth factor ␣, TNF receptor, L-selectin, and the Alzheimer amyloid protein precursor (8,9). In Drosophila, an ADAM family gene, Kuzbanian, has been demonstrated to play a role in the lateral inhibition during neurogenesis by a cleavage of the extracellular domain of the transmembrane receptor Notch (10,11). These observations indicate that ADAMs also function as membraneanchored metalloproteinases that are involved in the ectodomain shedding of cell-surface proteins.
ADAMTS-1 was identified as a gene highly expressed in vivo in the colon 26 cachexigenic tumor (12). The mouse adamts-1 gene is mapped to chromosome 16, region C3-C5 (13). Like other typical ADAMs, the amino-terminal half-region of AD-AMTS-1 consists of a proprotein and a metalloproteinase domain and a disintegrin-like domain that shares sequence similarity to the snake venom metalloproteinases. In contrast, the domain organization of the carboxyl-terminal half of AD-AMTS-1 is completely different from other ADAMs. Instead of the transmembrane region, ADAMTS-1 has three thrombospondin (TSP) type I motifs that are found in both thrombospondins 1 and 2 (14). These TSP type I motifs of ADAMTS-1 are functional for binding to heparin. We have recently found that ADAMTS-1 is secreted and incorporated into the extracellular matrix (ECM) (15). Analyses of deletion mutants have revealed that the carboxyl-terminal spacing region as well as three TSP type I motifs are responsible for the anchoring to the ECM (15).
So far, approximately, 20 ADAM genes have been identified. These genes are classified into two groups. One is the metalloprotease active or potentially active group, which has the zincbinding consensus sequence in the metalloproteinase domain. In contrast, some ADAMs, such as fertilin ␤ (ADAM2), Crytestin (ADAM3), and MDC (ADAM11), have an inactive metalloproteinase domain apparently due to the absence of the zincbinding motif (16 -18). ADAMTS-1 possesses a potential zincbinding motif in its metalloproteinase domain, but this motif seems to be incomplete. In this study, to clarify the fundamental properties of ADAMTS-1, we examined whether the AD-AMTS-1 protein is proteolytically active based on its capacity to form a covalent complex with ␣ 2 -macroglobulin (␣2M).
Construction and Transfection of Expression Vectors-The FLAG epitope tag (Asp-Tyr-Lys-Asp-Asp-Asp-Asp-Lys) expression vectors for the wild type ADAMTS-1 protein or the carboxyl-terminal deletion mutants, X4 and X5, were constructed using pcDNA3 (Invitrogen) regulated under a cytomegalovirus promoter, as described previously (15). An amino acid-substituted mutant, E386Q, was constructed by mutagenesis based on a PCR technique. Briefly, the first PCR reaction was carried out between the upstream forward and reverse mutated primer within the zinc-binding motif using ADAMTS-1 cDNA as a template. The second PCR reaction was performed between the first PCR product and the downstream reverse primer. The resultant PCR fragments containing the mutated sequence were digested by XhoI and BamHI and were used for the replacement of the corresponding fragment of the X5 expression vector. The nucleotide sequences of the mutants were confirmed by direct sequencing. The DNA sequencing reaction was performed by a PCR employing fluorescent dideoxynucleotides and was analyzed by a model 373A automated sequencer (Applied Biosystem). The furin expression vector pCMVFur (19) was kindly provided by Dr. K. Nakayama (Tsukuba University). COS-7 cells (2 ϫ 10 5 per dish) were transfected with the ADAMTS-1 expression vectors (2 g) by means of the LipofectAMINE (6 l) (Life Technologies, Inc.). The liposome-DNA complex solution was spread on cells in 3.5-cm dishes and incubated for 6 h. Thereafter, 1 ml of the DMEM containing 5% fetal bovine serum was added into the medium. Eighteen hours after transfection, the cells were washed with DMEM and further grown in 1 ml of DMEM with or without serum. Similarly, LoVo cells (4 ϫ 10 5 per dish) were transfected with the ADAMTS-1 expression vector (1 g) and the furin expression vector (1 g) by means of the LipofectAMINE (6 l) and further cultured in 1 ml of Ham's F-12 medium without serum.
For the immunoprecipitation experiments, aliquots of culture supernatants from transfected COS-7 cells were pretreated with protein G-Sepharose beads (Amersham Pharmacia Biotech) and incubated with anti-bovine ␣2M antibodies (Yagai, Co. Ltd., Yamagata, Japan). After 2 h at 4°C, 20 l of protein G-Sepharose beads was added to the mixture and incubated for another hour. The precipitates were washed and separated by SDS-PAGE and analyzed by Western blotting.
Binding Assay of ␣2M-Recombinant FLAG epitope-tagged AD-AMTS-1 proteins were prepared by COS-7 cell transient expression system, as described above. The expression was carried out in the serum-free DMEM for 2 days after transfection. An aliquot of the resultant culture supernatant containing ADAMTS-1 proteins was concentrated with Microcon-10 (Amicon) by about 5-fold, and 40 l of concentrated sample was mixed with an equal volume of purified ␣2M solution (final concentration 0.25 units/ml) (Roche Molecular Biochemicals) and incubated at 37°C for 15 h. For some experiments, the culture supernatants were exchanged into the binding buffer (40 mM sodium phosphate (pH 7.5), 100 mM NaCl) by means of Microcon-10, and a binding reaction to ␣2M was carried out in the same binding buffer (␣2M final concentration, 0.25 units/ml). The reactions were terminated with addition of an equal volume of 2ϫ Laemmli sample buffer, and cross-linked products in ␣2M⅐ADAMTS-1 complexes were analyzed by Western blotting using anti-FLAG M2 monoclonal antibody.
Determination of the Amino-terminal Sequence of the Mature AD-AMTS-1-ADAMTS-1 deletion mutant, X6 protein, was transiently expressed in COS-7 cells, and the culture supernatant containing AD-AMTS-1 proteins was concentrated with Microcon-10. Thereafter, proteins were separated on 10% SDS-PAGE and transferred onto polyvinylidene difluoride membrane. After staining with Ponceau S, the band corresponding to the mature form of the X6 protein was excised, and its amino-terminal sequence was analyzed by Protein Sequencing System G1005A (Hewlett-Packard).

Formation of the High Molecular Weight Complexes of the ADAMTS-1 Protein in the Cell
Culture System-The plasma protease inhibitor, ␣2M, has been shown to be capable of forming complexes with a wide variety of proteinases (20 -23). Because the complex formation of proteinases with ␣2M is dependent on their proteolytic activities, the proteinase activity of the mouse ADAMTS-1 protein was determined based on its  3) and were then cultured in the DMEM containing 5% fetal bovine serum. The culture supernatants from these transfectants were immunoprecipitated (IP) with anti-bovine ␣2M antibodies (anti-␣2M) (lanes 3 and 4) or with control rabbit IgG (CR) (lanes 1 and 2). Coprecipitating ADAMTS-1 protein was detected by Western blotting with an anti-FLAG M2 antibody. capacity to bind to ␣2M. For this purpose, recombinant AD-AMTS-1 protein was prepared by means of the transient expression system in COS-7 cells. The FLAG epitope tag, consisting of nine amino acids, was added in the carboxyl terminus for the detection of the recombinant proteins. As shown in Fig. 1A, the full-length ADAMTS-1 protein was not secreted into the conditioned cell culture media (lanes 2 and 6), although the ADAMTS-1 is a secretory-type ADAM family protein. This is due to its binding property to the extracellular matrix (15). Therefore, to obtain the soluble recombinant protein of AD-AMTS-1, the carboxyl-terminal deletion mutants, X4 and X5 ( Fig. 2A), lacking most of the ECM binding domains, were also expressed in the COS-7 cells. These carboxyl-terminal deletion mutants were released into the cell culture medium (Fig. 1A,  lanes 3 and 4). The binding capacity of ADAMTS-1 to ␣2M was first addressed by the cell culture system in the presence of the serum containing ␣2M using soluble-type ADAMTS-1 proteins. Under serum-free conditions, ADAMTS-1 X4 and X5 proteins were present as both the precursor and processed forms in the culture supernatant (Fig. 1A, lanes 3 and 4). In contrast, when mutants X4 and X5 were expressed under serum-containing conditions, high molecular weight species of these ADAMTS-1 proteins (complexes I and II) were detected in addition to their precursor and processed forms (lanes 7 and 8). Since the acrylamide gel electrophoresis was carried out under denatured conditions, these complexes appeared to result from the covalent binding of the ADAMTS-1 protein to a certain serum protein. In addition, these high molecular complexes were very similar to ␣2M⅐proteinase complexes reported in other proteinases (20,21). Moreover, an immunoprecipitation experiment using anti-␣2M antibodies (Fig. 1B) revealed that high molecular weight complexes of ADAMTS-1 protein found in the COS-7 cell culture system resulted from the covalent binding of the ADAMTS-1 protein to ␣2M in serum.
ADAMTS-1 Forms Covalent Complexes with ␣ 2 -Macroglobulin-To determine directly the capacity of ADAMTS-1 to bind to ␣2M, the recombinant ADAMTS-1 X5 mutant protein was first transiently expressed in COS-7 cells under the serum-free conditions, and the resultant culture supernatants containing the ADAMTS-1 protein were further used for the binding assay to the purified ␣2M. As shown in Fig. 3A, the high molecular weight species of mutant X5 (complexes I, and II), similar to those observed under the serum-containing conditions, were generated after incubation with the purified ␣2M (lane 5). This result indicates that the amino-terminal half of the ADAMTS-1 protein containing the metalloproteinase domain is capable of covalently binding to ␣2M and forming the high molecular weight complexes.
Requirement of ADAMTS-1 Proteinase Activity for Complex Formation with ␣2M-The covalent binding of ␣2M to various proteinases has been shown to be initiated by the proteolytic cleavage of the bait region of ␣2M by proteinases (20,21). In order to investigate the involvement of the proteolysis of ␣2M, the effects of various protease inhibitors on the formation of the high molecular weight complexes were examined. As shown in Fig. 3B, the ADAMTS-1⅐␣2M complex formation was inhibited by the metalloproteinase inhibitors, EDTA and phenanthroline (lanes 2 and 3), whereas other protease inhibitors, leupeptin, aprotinin, phenylmethylsulfonyl fluoride, and pepstatin, did not (lane 4 -7). These data suggest that ADAMTS-1 has an active metalloproteinase domain and that the metalloproteinase activity of this domain is required for the complex formation with ␣2M.
Like other active ADAM metalloproteases such as TACE and KUZ, ADAMTS-1 possesses a potential zinc-binding motif (HEXXHXXXXXHD) in its metalloproteinase domain (Fig. 2B). To confirm further whether the proteinase activity of AD-AMTS-1 is required for the complex formation with ␣2M, an amino acid-substituted mutant was constructed within the zinc-binding motif. The mutant X5(E386Q), in which the essential Glu residue in the zinc-binding motif (HEXXH) was converted to Gln, can be expected to lose its zinc-binding capacity. When the X5(E386Q) mutant was expressed in COS-7 cells, the processed form of the protein was released into the culture supernatant at a level similar to that of X5 (Fig. 4A,  lanes 2 and 3). However, the high molecular weight species were not detected in the culture supernatant from the X5(E386Q)-transfected cells even under the serum-containing conditions (Fig. 4A, lane 3). Consistent with this finding, the X5(E386Q) mutant lost its binding capacity to ␣2M in the binding assay (Fig. 4B, lane 3). These results confirm that ADAMTS-1 is an active metalloproteinase and that a proteolytic cleavage of ␣2M is necessary for the ADAMTS-1⅐␣2M complex formation.
The ADAMTS-1 Protein Was Processed by Furin Endopeptidase in Vivo-As described above, the ADAMTS-1 proteins were produced from COS-7 cells in both the precursor and processed forms. In addition, the mature forms of ADAMTS-1 deletion mutants, the X4 and X5, migrated as a doublet of bands on SDS-PAGE (Fig. 1A). To investigate the processing mechanism of the ADAMTS-1 precursor, we first addressed how these two mature forms were produced. Because the X4 and X5 deletion mutants have an N-linked glycosylation motif ( Fig. 2A), the contribution of glycosylation to the generation of two different forms was examined by means of tunicamycin, an inhibitor of N-linked glycosylation. As shown in Fig. 5, the mature form of X4 proteins was detected as a doublet of bands under the normal growth conditions (lane 2). However, when X4 protein was expressed in the presence of tunicamycin, the upper band of the two mature forms completely disappeared, and only the lower band could be detected (Fig. 5, lane 3). Consistent with this finding, the mature form of another deletion mutant, X6, which lacks an N-linked glycosylation motif ( Fig. 2A), was detected as a single band (Fig. 5, lane 4). These results clearly demonstrate that two mature forms of the X4 protein result from differential glycosylation and that the AD-AMTS-1 precursor is processed at the single cleavage site.
To examine further the processing mechanism, we next determined the partial amino-terminal sequence for the mature form of the X6 protein. The amino-terminal sequence of the mature X6 protein was FVSSPRYV and was located just after the furin recognition site (RKKR) (Fig. 2A). These results strongly suggest that the precursor of the ADAMTS-1 protein is processed by a furin endopeptidase. To assess this possibility, we further examined the processing of the ADAMTS-1 protein using a human colon carcinoma cell line, LoVo, that does not produce functional furin (24). When the soluble type of the ADAMTS-1 protein, X5, was transiently expressed in LoVo cells, only the precursor form of the X5 protein was detected in the culture supernatant (Fig. 6, lane 2), suggesting that the processing of the ADAMTS-1 precursor is impaired in LoVo cells. Moreover, cotransfection of the mouse furin cDNA into LoVo cells resulted in the disappearance of the ADAMTS-1 X5 precursor and the appearance of its processed form (lane 3). In addition, a similar processing pattern was obtained when another deletion mutant of ADAMTS-1, X4, was expressed in LoVo cells (data not shown). These results indicate that the proprotein domain of the ADAMTS-1 precursor protein is proteolytically removed by furin endopeptidase in vivo.

DISCUSSION
The present study has demonstrated that the proteinase domain of ADAMTS-1 is capable of forming a covalent complex with ␣2M. ␣2M is a plasma proteolytic enzyme inhibitor that binds various types of proteinases including serine proteases, cysteine proteases, and metalloproteases (20 -23). The mechanism for the binding of proteinases to ␣2M is very unique. ␣2M contains a bait region in the middle of its peptide chain that provides a target for the proteinases (25). Bait region cleavage by proteinases triggers conformational changes in the ␣2M subunits that cause an encapsulation of the proteinases and the activation of internal thioesters of ␣2M, resulting in a covalent cross-linking of the active proteinases. Because the complex formation of proteinases with ␣2M is dependent on their proteolytic activities against the bait region, ␣2M is a useful tool for identifying unknown proteinases (22). Recently, Loechel et al. (23) have reported that an ADAM family protein, meltrin ␣, is an active metalloproteinase by means of this trapping mechanism of ␣2M. In this study, we have also revealed that ADAMTS-1 possesses an active metalloproteinase domain based on its ability to form a covalent complex with ␣2M. The finding that the mutation of the zinc-binding motif of ADAMTS-1 eliminated its capacity to bind to ␣2M confirmed this notion.
ADAM is a gene family that shows a sequence similarity to venom metalloproteinases such as hemorrhagic toxin (26). An x-ray crystal structure analysis of venom metalloproteinases, adamalysin II and atrolysin C (HT-d), has revealed that these two metalloproteinases consist of a large main molecular body and a small subdomain (27,28). These two domains are separated by a long active site cleft with the catalytic zinc ion at its basement. The catalytic zinc in the active site cleft of the venom metalloproteinases is coordinated by three histidines of the zinc-binding motif (HEXXHXXGXXH), whereas the Met turn forms the basement of the zinc locus. Recently, Maskos et al. (29) have demonstrated that a catalytic zinc environment of TACE (ADAM17) resembles that of the snake venom metalloproteinases. Since ADAMTS-1 possesses a potential zinc-binding motif and retains the subsequent methionine residue that may function as a Met turn (Fig. 2B), a similar zinc-binding environment can be expected for ADAMTS-1.
A sequence comparison of the active ADAM proteases with the snake venom metalloproteinases reveals a consensus sequence for the zinc-binding motif, HEXGHXXGXXHD. Furthermore, from analyses of a number of zinc endopeptidases including the snake venom metalloproteinases, astacin, serra- lysins, and matrix metalloproteinases, it appears that the zincbinding motif is more generalized in HEXXHXXGXXH (30 -32). However, in the potential zinc-binding motif of ADAMTS-1, the Gly residue of the consensus sequence (HEXXHXXGXXH) is not conserved (Fig. 2B). Since the present study has revealed that ADAMTS-1 is an active metalloproteinase, it is likely that the conserved Gly residue of the zinc-binding motif is functionally interchangeable with Asn.
Like matrix metalloproteinases and snake metalloproteinases, ADAMTS-1 possesses the conserved cysteine residue in the proprotein domain that may function to maintain the metalloproteinase domain in a latent state. Therefore, it is very important to investigate the processing mechanism of the AD-AMTS-1 precursor for an understanding the regulatory mechanism of ADAMTS-1 protease. In the COS-7 cells expression system, the ADAMTS-1 proteins were produced as processed forms as well as precursor forms. One possibility is that partial proteolytic cleavage of the proprotein domain by other proteases may result in an autolytic activation of ADAMTS-1. However, in this study, we found that the processed form of ADAMTS-1 was still present in the COS-7 cells culture supernatant even when its metalloproteinase domain was inactivated by point mutation within the zinc-binding motif. This finding negates the possibility that the precursor form of the ADAMTS-1 protein is processed by an autolysis mechanism under COS-7 cell culture conditions.
On the other hand, as described in our previous study (15), ADAMTS-1 possesses the typical furin recognition site (RX(K/ R)R) (33,34) in the boundary between the proprotein and the metalloproteinase domains, as estimated by sequence alignment with the venom metalloproteinases. In this study, we have shown that the amino-terminal sequence of the mature ADAMTS-1 protein is located just after this furin recognition site. Furthermore, we have shown that the processing of the ADAMTS-1 precursor is impaired in LoVo cells deficient in furin and that the processing capacity of the cells is restored by the transfection of furin cDNA. These results provide evidence that the ADAMTS-1 precursor is proteolytically processed by furin in vivo. It was shown that several ADAM family proteins contain a consensus furin cleavage site between their prodomain and metalloproteinase domains. Lum et al. (35) have reported that the pro-domain of MDC15 can be cleaved by a recombinant furin in vitro. But our study is the first report providing direct evidence that furin is involved in the processing of ADAM families in vivo. Furin is a serine endopeptidase in Golgi and cleaves a wide range of precursor proteins (33,34). Because furin is expressed in a wide variety of cells, our data implicate that the proprotein domain of the ADAMTS-1 protein is removed in the secretory pathway and that the proteolytically active form of the ADAMTS-1 protein is secreted from producing cells during inflammation.
Our previous study showed that ADAMTS-1 is associated with ECM through multiple ECM-binding domains in its carboxyl-terminal region, whereas its metalloproteinase domain is free (15). The present study has demonstrated that ADAMTS-1 has a functional proteinase domain. Taken collectively, ADAMTS-1 is an active ADAM protease that is associated with ECM. Further investigation is necessary to identify the in vivo substrate of ADAMTS-1 and to therefore provide understanding of the pathophysiological role of ADAMTS-1. But this study suggests that ADAMTS-1 may play a role in the inflammatory process through a processing of proteins that are present in the ECM.