Discovery and Characterization of a Novel, Widely Expressed Metalloprotease, ADAMTS10, and Its Proteolytic Activation*

We describe the discovery and characterization of ADAMTS10, a novel metalloprotease encoded by a locus on human chromosome 19 and mouse chromosome 17. ADAMTS10 has the typical modular organization of the ADAMTS family, with five thrombospondin type 1 repeats and a cysteine-rich PLAC (protease and lacunin) domain at the carboxyl terminus. Its domain organization and primary structure is similar to a novel long form of ADAMTS6. In contrast to many ADAMTS proteases, ADAMTS10 is widely expressed in adult tissues and throughout mouse embryo development. In situ hybridization analysis showed widespread expression of Adamts10 in the mouse embryo until 12.5 days of gestation, after which it is then expressed in a more restricted fashion, with especially strong expression in developing lung, bone, and craniofacial region. Mesenchymal, not epithelial, expression in the developing lung, kidney, gonad, salivary gland, and gastrointestinal tract is a consistent feature of Adamts10 regulation. N-terminal sequencing and treatment with decanoyl-Arg-Val-Lys-Arg-chloromethylketone indicate that the ADAMTS10 zymogen is processed by a subtilisin-like proprotein convertase at two sites (Arg64↓Gly and Arg233↓Ser). The widespread expression of ADAMTS10 suggests that furin, a ubiquitously expressed proprotein convertase, is the likely processing enzyme. ADAMTS10 expressed in HEK293F and COS-1 cells is N-glycosylated and is secreted into the medium, as well as sequestered at the cell surface and extracellular matrix, as demonstrated by cell surface biotinylation and immunolocalization in nonpermeabilized cells. ADAMTS10 is a functional metalloprotease as demonstrated by cleavage of α2-macroglobulin, although physiological substrates are presently unknown.

such as cytokines, growth factors, their binding proteins, receptors, and adhesion molecules has important biological consequences (1,2). Proteases that cleave such molecules thus play important roles in tissue remodeling, morphogenesis, inflammatory and degenerative diseases, and cancer. Zinc-metalloendopeptidases (metalloproteases) comprise an important superfamily of such enzymes. Specific roles for distinct metalloprotease families and their individual members have emerged over the past decade through genetic studies in humans and mice. Matrix metalloproteases are the major ECMdegrading enzymes, but they also have a role in proteolysis of other secreted molecules and cell surface proteins (1,2). AD-AMs are primarily "sheddases," proteases that process cell surface molecules and are thought to have little, if any, direct role in ECM catabolism (3,4). The ADAMTS (a disintegrin-like and metalloprotease domain with thrombospondin type 1 motifs) family was unknown until 1997 (5), but functions for some of these enzymes are beginning to emerge (5)(6)(7)(8)(9)(10)(11). Known AD-AMTS substrates include the proteoglycans aggrecan, versican, and brevican; the fibrillar procollagens I, II, and III; and von Willebrand factor (6,9,10,(12)(13)(14). The processing of von Willebrand factor and the fibrillar procollagens by ADAMTS13 and by the procollagen amino-propeptidases (ADAMTS2, -3, and -14), respectively, is essential for their maturation to fully functional molecules (6,7,9,15,16). These two processing activities appear to be highly specialized, and the enzymes responsible for them have distinct sequence and structural features not shared by the other ADAMTS proteases (17,18). On the other hand, a number of ADAMTS with disparate domain and sequence features (such as ADAMTS1, ADAMTS4, ADAMTS5, and ADAMTS9) are known to process large aggregating proteoglycans such as aggrecan and versican (10,12,20,21). Nevertheless, this is not a general property of all ADAMTS proteases, since we have shown recently that ADAMTS7 cannot process versican or aggrecan at sites cleaved by the other proteoglycan-degrading ADAMTS (22).
ADAMTS proteases are modular, consisting of a protease domain and an ancillary domain (23). The protease domain of these enzymes, like that of ADAMs, but not MMPs, is of the reprolysin (snake venom) type. The hallmark of the ADAMTS proteases is the presence of at least one thrombospondin type 1 repeat (TSR). Other highly conserved modules are arranged around this central TSR in a specific organization, and there are additional TSRs near the carboxyl terminus in all members of the ADAMTS family with the exception of ADAMTS4 (23). ADAMTS proteases are synthesized as zymogens that are targeted to the secretory pathway and activated by proprotein convertases. Zymogen processing leads to removal of a 200 -220-amino acid-long prodomain in the secretory pathway or at the cell surface. 19 mammalian ADAMTS proteases are known, and all except ADAMTS10, the subject of this article, have previously been described in the literature. Within the ADAMTS family, subsets of proteases have highly conserved domain organization, primary sequence and gene structure, suggestive of a close evolutionary and perhaps functional relationship (7,20,22,23). In this context, determination of the primary structure of ADAMTS10 led to realization of a putative long form of AD-AMTS6, whose domain organization and primary structure support the contention that it forms a phylogenetic subset with ADAMTS10. Unlike most other ADAMTS proteases, including ADAMTS6 (24), ADAMTS10 is widely expressed. We investigated the developmental regulation of the Adamts10 gene in mice and the activation mechanism and localization of the enzyme in cultured cells. ADAMTS10 is shown to be a functional metalloprotease, although its physiological substrates are presently unknown.

EXPERIMENTAL PROCEDURES
cDNA Cloning of Human and Mouse ADAMTS10 -Using the tBLASTn (Basic Local Alignment Search Tool) program at the National Center for Biotechnology Information, we searched the data base of expressed sequence tags (dBEST), using the protein sequences of a number of ADAMTS proteases, and identified similarities in a human EST (GenBank TM accession number AA588434) derived from the human prostate-derived IMAGE clone 1101403. The IMAGE clone was purchased (Research Genetics, Huntsville, AL), and the insert was sequenced in its entirety. Oligonucleotide primers based on the sequences at the ends of this clone were used with human fetal brain cDNA (Marathon cDNA, Clontech, Palo Alto, CA) as a template to perform iterative rapid amplification of cDNA ends by PCR at 5Ј-and 3Ј-ends, essentially as previously described (24,25). Nucleotide sequencing was done at the National Institutes of Health-supported Molecular Biotechnology Core of the Lerner Research Institute (Cleveland Clinic Foundation), and nucleotide sequence data were analyzed using Lasergene software (DNAStar Inc., Madison, WI). Integration of the overlapping sequences provided the complete ORF and primary sequence of human ADAMTS10.
To confirm that the overlapping human cDNAs were derived from a single transcript, we designed PCR primers incorporating the most 5Ј cloned human sequence and the stop codon of the ADAMTS10 ORF (forward primer, 5Ј-AAGAATTCAGAGACATGTGGACACGTGG-3Ј (EcoRI site underlined, start codon in boldface type); reverse primer, 5Ј-AAGTCGACCGAGTGGCCCTGGCAGGTTTTGC-3Ј (SalI site underlined, modified stop codon (to Ser) in boldface type)). PCR was done using human fetal lung cDNA or human lung cancer cell line A549 cDNA as templates (Clontech, Palo Alto, CA) and using the following conditions: 95°C for 1 min and then 30 cycles of 95°C for 30 s, 58°C for 30 s, and 68°C for 5 min. The resulting 3.2-kbp amplicon was gelpurified, ligated into pGEM-T Easy (Promega, Madison, WI), and sequenced.
The mouse IMAGE clone 1077653 (EST AA822090) was detected in GenBank TM as a presumptive Adamts10 clone, 2 purchased from Research Genetics (Huntsville, AL), and its 1.6-kbp insert was sequenced in its entirety. Additional 5Ј mouse cDNA sequence was deduced from mouse genomic sequences (available with GenBank TM accession numbers AC073802 and AC073766), using the GENSCAN program at the Massachusetts Institute of Technology (available on the World Wide Web at CCR-081.mit.edu/GENSCAN.html) to predict the exons in these sequences. The complete mouse ADAMTS10 ORF was amplified by PCR of mouse 17.5-day-old embryo cDNA in similar fashion to that described above for the human cDNA.
Northern Analysis-Mouse embryo Northern blots and multiple tissue Northern blots from adult human and mouse tissues and from human cancer cell lines (Clontech, Palo Alto, CA, and Seegene Inc.) were hybridized to the [␣-32 P]dCTP-labeled inserts of human and mouse ADAMTS10 IMAGE clones, as per the manufacturer's recommendations, followed by autoradiographic exposure for 4 days.
In Situ Hybridization-Adamts10 IMAGE clone 1077653 was digested with StuI and XhoI to delete 792 bp of the 1642-bp insert. The plasmid containing the remainder was blunt-ended with Klenow frag-ment of DNA polymerase I (New England Biolabs, Beverly, MA) and religated to obtain an 850-bp Adamts10 cDNA encoding part of the cysteine-rich domain, spacer domain, and first two TSRs (plasmid 1077653X1). This plasmid was used to transcribe sense and antisense cRNA probes continuously labeled with [ 35 S]UTP. Paraffin sections of formaldehyde-fixed mouse embryos of age 9. 5, 12.5, 14.5, 15.5, and 17.5 days were hybridized to the Adamts10 probes as previously described (20), followed by dipping in photographic emulsion for autoradiography. Adamts10 autoradiographic signal was visualized with dark field microscopy, whereas cell nuclei were stained with 4,6-diamidino-2-phenylindole (Hoechst 33258 dye; Sigma), which fluoresces blue under UV light.
ADAMTS10 Expression Plasmids-The preprocatalytic coding region of human ADAMTS10 (ADAMTS10-(1-463)) was amplified by PCR using the oligonucleotide primers 5Ј-AAGAATTCGGCCTCTATG-GCTCCCGCC-3Ј (forward primer) and 5Ј-AAGTCGACCACAAAGTC-CTGTCTGGG-3Ј (reverse primer; introduced SalI site is underlined) and Advantage 2 high fidelity polymerase (Clontech, Palo Alto, CA). The PCR products were gel-purified and ligated to the pGEM-T easy vector (Promega Corp., Madison, WI). The insert of a sequence-verified clone was then ligated into the EcoRI and XhoI site of pcDNAmyc His Aϩ (Invitrogen) for expression of ADAMTS10-(1-463) with a C-terminal tandem myc and His 6 tag. The full-length ADAMTS10 cDNA described above was cloned in frame with a C-terminal tandem myc and His 6

tag
Transfection and Selection of Stable Cell Lines-HEK293F cells (Invitrogen) at 80% confluence were transfected in 6-well plates with 100 ng of full-length ADAMTS10 or ADAMTS10-(1-463) expression plasmid DNA using Fugene 6 (Roche Applied Science) as per manufacturer's instructions. At the first medium change, it was supplemented with 1 mg/ml G418 (Mediatech, Herndon, VA). Discrete colonies were isolated using cloning discs (PGC Scientific, Frederick, MD) and expanded. Western blotting with anti-myc monoclonal antibody 9E10 (Invitrogen) was used to determine the level of protein expression in the media of these clones.
Expression and Characterization of ADAMTS10 and ADAMTS10-(1-463)-Stably transfected cells expressing full-length ADAMTS10 and ADAMTS10-(1-463) were cultured in three-tier flasks (Nunc, Rochester, NY) in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. When cultures were 80% confluent, the serumcontaining medium was replaced with serum-free 293CD medium (Invitrogen) followed by further culture at 37°C in the presence of 8% CO 2 for 5 days. Conditioned medium was collected, centrifuged briefly to remove cellular debris, and supplemented with NaCl to a final concentration of 0.5 M. ProBond resin (Invitrogen) was prepared by washing with 1 bed volume of binding buffer (0.5 M NaCl, 20 mM sodium phosphate, pH 7.8). The media and resin were mixed overnight at 4°C in a 1:1 (v/v) ratio in binding buffer. After this binding step, the resin was pelleted by centrifugation at 1000 ϫ g and then washed five times with 10 resin bed volumes of binding buffer. Bound proteins were eluted by sequential washes with binding buffer containing 50, 100, 150, 200, and 250 mM imidazole. The washes and eluted protein fractions were assayed for the presence and purity of desired proteins by Western blotting (using anti-myc monoclonal antibody) and by reducing SDS-PAGE with Coomassie Blue staining, respectively. Maximal yield was obtained on elution in 100 -250 mM imidazole.
Following purification of ADAMTS10-(1-463), major bands of ϳ52, ϳ50, and ϳ29 kDa were excised after electroblotting to a polyvinylidene difluoride membrane. N-terminal sequence was determined by Edman degradation at the National Institutes of Health-supported Biotechnology Core of the Lerner Research Institute.
Characterization of full-length ADAMTS10 was done using stably transfected HEK293F cells or transiently transfected COS cells or substantially purified protein. Western blotting was done with anti-myc antibody 9E10. Protein deglycosylation was done as previously described using purified protein (20,22). Processing of ␣ 2 -macroglobulin (␣ 2 -M) was tested by incubation with purified protein as described previously (22). Proteolysis of the aggrecan core protein using AD-AMTS10-and ADAMTS4-transfected cells was evaluated as described The key to the domains depicted in the schematic diagram is at the bottom. B, alignment of predicted amino acid sequences of mouse (Adamts10) and human ADAMTS10 with ADAMTS6B. Three potential furin cleavage sites (e.g. RQRR2), are indicated by brackets. The asterisk over two sites indicates that these were confirmed by N-terminal sequencing. The zinc-binding histidine triad of the active site and the "Met-turn" (LMA) are in boldface type and boxed. The disintegrin-like domain is in black type against a gray background, and the cysteine-rich domain is in white type against a gray background. The spacer domain is indicated by a dashed previously (20,22). Briefly, equal numbers of transfected cells were incubated with 20 g of aggrecan, and the presence of a cleaved peptide bond detected by the anti-AGEG neoepitope antibody (26) was sought by Western blotting of the aggrecan as previously described (20).
ADAMTS10 Localization in Transfected Cells-These studies examined the distribution of full-length ADAMTS10 in vitro, in regard to the cells expressing it. COS-1 cells (ATCC, Manassas, VA) were transiently transfected with 1 g of full-length ADAMTS10 prior to immunofluorescent localization of secreted protein in nonpermeabilized cells, essentially as previously described (27). ADAMTS10 was detected using antibody 9E10 and Alexa-488-conjugated goat anti-mouse secondary antibody (Molecular Probes, Inc., Eugene, OR) in an indirect immunofluorescence method that does not detect intracellular protein. Following staining for tagged ADAMTS10, cells were permeabilized and nuclei were stained with 4Ј,6-diamidino-2-phenylindole as previously described (27), followed by fluorescent microscopy. Medium, cell lysate, and extracellular matrix from these cultures and from stably transfected HEK293F cells were collected as previously described and subjected to immunoblotting with antibody 9E10 following reducing SDS-PAGE (27). Stably transfected HEK293F cells expressing ADAMTS10 in suspension were biotinylated as previously described (22). Isolation of biotinylated proteins from the cell surface and their analysis by electrophoresis was as previously described (22). As a control, cells were treated with trypsin to eliminate all cell surface proteins prior to biotinylation, essentially as previously described (22).

RESULTS
cDNA Cloning of Human and Mouse ADAMTS10 -Using the tBLASTn algorithm to scan dBEST for novel ESTs that were homologous to cognate ADAMTS proteins, we identified the EST 1101403. This EST was identified when the data base was screened with the sequence of ADAMTS6 but not with other ADAMTS proteins. Following extension of the EST by rapid amplification of cDNA ends in both directions, we generated an amplicon of 3.2 kbp from human fetal lung and the A549 cell line. The cDNA encoded an open reading frame of 1103 amino acids with the typical ADAMTS modular structure (Fig. 1A) and was designated as ADAMTS10.
Sequencing of mouse IMAGE clone 1077653 provided an ORF encoding the cysteine-rich domain through to the 3Јuntranslated region of mouse ADAMTS10. Mouse chromosome 17 genomic sequence in GenBank TM was used to predict the likely amino acid sequences upstream of this clone that were then validated by cDNA cloning (by PCR of the complete ORF using mouse embryo cDNA) and sequencing. The human and mouse nucleotide and predicted amino acid sequences have an overall identity of 86 and 91%, respectively (Fig. 1B), although mouse ADAMTS10 is one amino acid longer (1104 aa). The putative start codon of human ADAMTS10 is the N-terminalmost ATG codon in the predicted ORF and is within an appropriate Kozak consensus sequence (although there is not a purine nucleotide at position Ϫ3 relative to A (position 1) of ATG, there is a G at ϩ4 constituting an acceptable consensus start context) (28). The mouse ADAMTS10 start codon is also within a good Kozak consensus sequence.
Features of the Primary Structure of ADAMTS10 -Numerous sequence features in mouse and human ADAMTS10 are very highly conserved (Fig. 1B), and the discussion that follows pertains to human ADAMTS10, mentioning mouse AD-AMTS10 only where it differs. Overall, the domains of AD-AMTS10 are very similar to those of other ADAMTSs, and each of its domains is comparable in length and number of cysteine residues with those of the other ADAMTS proteases (24).
The start codon is followed by a signal peptide containing a region of 11 hydrophobic residues (Trp 10 -Phe 20 ), suggesting that ADAMTS10 is a secreted protein (Fig. 1B). According to consensus observations made for a number of proteins (the so-called Ϫ1, Ϫ3 rule) (29), it can be predicted that signal peptidase cleavage probably occurs following Ala 25 , and the secreted zymogen has the N terminus Phe 26 -Arg-Ser-Gln.
The prodomain, extending from Phe 26 to Arg 233 , by analogy with other ADAMTS proteases is somewhat unusual in the ADAMTS family in containing only one sequence motif, in complete agreement with the proproteinase convertase recognition sequence Arg-Xaa-Arg/Lys-Arg (i.e. Arg-Gln-Arg-Arg 66 ) ( Fig. 1B) (35)(36)(37). Most ADAMTS proteases have multiple proprotein convertase recognition sites, with the most C-terminal of these undergoing the processing that yields the final processed form (16,20,22,30). At the expected location corresponding to the final processing site (i.e. Arg 233 ), the sequence (Gly-Leu-Lys-Arg 233 ) does not match the optimal furin consensus (Fig. 1B). There is a dibasic motif encompassing the P1 and P2 residues, but the P4 residue is Gly, and there is not a compensating Arg residue at the P6 position (37). However, there is a Ser residue at the putative P1Ј position, which is found in over 50% of processed proproteins (37). The prodomain contains two N-linked glycosylation sites (Asn-Xaa-Ser/Thr, where Xaa can be any amino acid except Pro) (Fig. 1B).
The ADAMTS10 catalytic domain contains a typical zincbinding active site sequence (Fig. 1B) that is not, however, identical to any other ADAMTS protease. The catalytic domain, disintegrin-like domain, and cysteine-rich domain have the typical sequence layout and number of cysteine residues (8,8, and 10, respectively) seen in other ADAMTS and are predicted to be internally disulfide-bonded. The central TSR of AD-AMTS10 contains a possible sulfatide/glycosaminoglycan (GAG) binding motif (Trp 550 -Thr-Pro-Trp) at its N terminus and also contains a basic region (Arg 591 -Arg-Arg-His-Arg) that could mediate GAG binding (Fig. 1B). The four C-terminal TSRs do not have these motifs and do not resemble each other substantially, although each has the conserved N-terminal Trp residue and the signature six-cysteine arrangement typical of TSRs. At the carboxyl terminus of ADAMTS10, a cluster of six cysteine residues has the hallmark arrangement of a PLAC (protease and lacunin) domain (Fig. 1B). This domain was first described in an ADAMTS-like ECM protein, lacunin (31), and is also found toward the carboxyl terminus of some proprotein convertases and ADAMTS proteases.
In addition to glycosylation within the prodomain, other consensus N-linked glycosylation sites are present and conserved within mouse and human ADAMTS10, predicting that activated ADAMTS10 is likely to be a glycoprotein. Two such sites are located in the spacer and one within TSR3. An Nlinked glycosylation site in the catalytic domain of human ADAMTS10 is absent in the mouse. The predicted molecular mass of the human and mouse ADAMTS10 zymogen and fully processed forms are 118 and 95 kDa, respectively.
Homology to ADAMTS10 Uncovers a Longer Form of AD-AMTS6 (ADAMTS6B)-The close sequence similarity of AD-AMTS10 to ADAMTS6 led us to ask whether there existed a longer form of this protease with the same modular organization as ADAMTS10. Analysis of 3Ј genomic sequence of AD-AMTS6 revealed previously unknown exons that could splice to a putative splice donor site 150 bp upstream of the previously identified ADAMTS6 stop codon (24). The conceptual translation product that included the new exons added three additional TSRs and a C-terminal PLAC domain to the cognate ADAMTS6 protein, mirroring precisely the structure of AD-AMTS10. This conceptual product of ADAMTS6 is designated ADAMTS6B (Fig. 1B), and its existence is supported by numerous ESTs in GenBank TM .
ADAMTS10 and ADAMTS6B have an identical domain organization and amino acid identity and similarity (includes conserved amino acid substitutions) of 59 and 73%, respectively. The conservation extends to the positioning of two AD-AMTS6 proprotein convertase processing sites (Arg-Arg-Arg-Arg 65 and Arg-Gln-Lys-Arg 244 ), a highly similar zinc binding sequence within the active site (differing in one amino acid), and two N-linked glycosylation sites (Fig. 1B). A Gly-Ser 835 -Gly-Asp-Asn-Glu motif in the ADAMTS6 spacer (Fig. 1B) is relevant in light of the recent discovery of GAG attachment in ADAMTS7B (22). The two adjacent acidic residues (italicized above) may favor GAG attachment to the boldface Ser residue (32). This sequence motif is missing in ADAMTS10 (Fig. 1B). ADAMTS17 and ADAMTS19 (33) also have five TSRs and a C-terminal PLAC domain, but they have less homology to AD-AMTS10 (e.g. ADAMTS19 has 32% amino acid sequence identity and 44% similarity) and are thus less closely related to ADAMTS10 than is ADAMTS6.
ADAMTS6B and ADAMTS10 have identical gene structures, each having 25 exons with conserved splice boundaries (Fig.  1B). ADAMTS10 maps to human chromosome 19 and mouse chromosome 17. Gene location has been experimentally validated by interspecific backcross analysis in the mouse and radiation hybrid mapping in humans. 3 ADAMTS6 maps to human chromosome 6 and mouse chromosome 13 (24).
ADAMTS10 mRNA Is Widely Expressed-Adamts10 was expressed at all four mouse developmental ages examined. Maximal expression was seen in 15-and 17-day-old embryos ( Fig.  2A, left panel), and the lowest levels were present in 7-day-old embryos. A single mRNA of ϳ5 kb was detected in the mouse. A similar sized mRNA of comparable or greater intensity was seen in some adult mouse tissues such as the heart and lungs ( Fig. 2A, right panel). Fainter bands were seen in kidney, liver, spleen, brain, and testis, and no Adamts10 message was detectable in adult skeletal muscle. The ADAMTS10 (human) mRNA differed in that two species of 5 and 8 kb were detected on Northern blots from human organs and cell lines (Fig. 2B). Widespread expression was also seen in human tissues, and as in the mouse, skeletal muscle had the lowest levels of expression (Fig. 2B, left panel). Of the human cancer cell lines examined (Fig. 2B, right panel), the highest expression was seen in the A549 cell line, a lung-carcinoma-derived line with characteristics of a type II alveolar cell.
Adamts10 Is Dynamically Expressed during Mouse Embryogenesis-In situ hybridization of mouse embryos at different stages revealed a highly dynamic pattern of expression with widespread low level expression in 9.5-, 10.5-, and 12.5-day-old embryos and increasingly tissue-specific gene expression in 14.5-, 15.5- (Fig. 3A), and 17.5-day-old embryos (Fig. 3B). Expression was found in most tissues of the embryonic day 9.5-12.5 mouse embryo except ectoderm (Fig. 3A). Since organogenesis has not yet advanced substantially at this age, the broad tissue distribution was uninformative. At embryonic days 14.5 and 15.5, the expression pattern was identical, and the mRNA distribution was relatively organ-and cell typespecific (Fig. 3A). In the craniofacial region, strong expression was noted in the craniofacial mesenchyme (Figs. 3A and 4E), submandibular gland (Figs. 3A and 4A), mesenchyme sur-

FIG. 2. Expression pattern of Adamts10 (A) and ADAMTS10 (B) by Northern analysis.
In A, the blot on the left is from mouse embryo RNA (gestational age in days is indicated above each lane), and the blot on the right is made from adult mouse tissues. In B, the blot on the left is made from normal human tissues, and that on the right is from human cancer cell lines. RNA size markers (in kb) are shown on the left of each panel.
rounding the cochlear neuroepithelium (Fig. 4C), developing cerebral cortex (Figs. 3A and 4J), newly formed bone in the mandible (Fig. 3A), and tongue musculature (Fig. 3A). Strong expression was seen in perichondrium and periosteum but not in cartilage (Fig. 3A). Expression was essentially mesenchymal and was not seen in developing mandibular, tongue, or nasal epithelium, surface ectoderm, or cochlear neuroepithelium (Figs. 3A and 4, A, C, and E). Of the thoracic organs, the lungs had prominent expression restricted to the mesenchymal cells between the developing bronchial tree (Fig. 4K), and blood vessels expressed Adamts10 (Fig. 4F), but the heart was negative (Fig. 3A). In the abdomen, putative mesenchymal cells in the media of the stomach and duodenum and the pancreas were positive. The liver did not have levels of autoradiographic signal above background (signal seen in Fig. 3, A and B, is from red blood cells). In the dorsal regions, there was strong expression in the dorsal root ganglia and primary vertebral ossification centers of vertebrae (Fig. 4L). Unidentified cells in adrenal and renal cortex and the developing gonad were positive (Fig. 4, B, H, and I). High expression was noted in the connective tissue mesenchyme between the cartilaginous metacarpals of the hand and in the foot (data not shown). In addition, dense connective tissue such as joint capsule, tendons and ligaments (e.g. around the hip joint) (Fig. 4D) were strongly positive. Sense cRNA probe did not give signal above background levels, indicating that the antisense probe was specific (Fig. 4G).
In the 17.5-day-old embryo, the tissue-specific expression pattern (Fig. 3B) was essentially similar to that at embryonic days 14.5 and 15.5, with some notable differences. The expression levels in the lung mesenchyme, craniofacial mesenchyme, and the developing bone were of higher relative intensity than at preceding stages (Fig. 3B). The relative expression level was decreased in the developing brain, and for the first time, prominent expression was seen in chondrocytes in the developing cartilaginous skeleton (Fig. 4M). Rapid ossification of the skeleton at this developmental stage with strong expression in bone and cartilage and lung may explain why 17-day-old embryos had the highest expression levels on Northern blot. In addition, strong expression was seen in the walls of large arteries and around large vascular structures in the liver, which are precursors of the hepatic veins and the inferior vena cava.
Characterization of ADAMTS10 in Transfected HEK293F Cells-Western blotting of serum-containing conditioned medium from HEK293F cells stably transfected with myc-tagged human ADAMTS10 revealed a single band sized approximately at 130 kDa (Fig. 5A) or two closely approximated bands of nearly the same size (Fig. 5B). The ϳ130-kDa immunoreactive bands significantly exceeded the predicted size of the AD-AMTS10 zymogen (118 kDa) or mature enzyme (95 kDa). When stably transfected cells were cultured in serum-free medium, a number of immunoreactive bands were detectable by Western blotting using the anti c-myc antibody (Fig. 5A); this indicated that ADAMTS10 undergoes proteolysis in the absence of serum.
To determine whether the unexpected mass increase of AD-AMTS10 was a result of N-linked glycosylation, we deglycosylated it using PNGase F. This resulted in faster migration on SDS-PAGE, and the deglycosylated ADAMTS10 now had an apparent molecular mass of 120 kDa (Fig. 5B, left panel), which is close to the predicted size of the zymogen. A fainter band of ϳ100 kDa also emerged after deglycosylation, which may represent the mature ADAMTS10 enzyme (Fig. 5B). Other bands were smaller than expected and may be derived by proteolytic degradation of ADAMTS10. The secreted product of ADAMTS10-(1-463) transfected cells was also deglycosylated using PNGase F, but as predicted by the primary sequence, glycosylation was restricted to the prodomain, since migration of the 29-kDa fully processed catalytic domain was unaffected by PNGase F treatment (Fig. 5B, right panel). Four major myc-reactive bands were present in the medium of AD-AMTS10-(1-463)-transfected cells, namely, a pair of bands (52 and 54 kDa), possibly representing the zymogen and Arg 64processed zymogen, an uncharacterized intermediate of 37 kDa, and the 29-kDa mature derivative of ADAMTS10-(1-463). The failure of PNGase F to alter the relative size difference between the paired ϳ130-kDa (full-length ADAMTS10) and 52-54-kDa (ADAMTS10-(1-463)) bands suggested that they have different N termini. ADAMTS10 was able to cleave the broad spectrum protease substrate/inhibitor ␣ 2 -M (Fig. 5C). Cleavage of ␣ 2 -M within its bait region results in the entrapment of the cleaving protease in an irreversible complex with this protease inhibitor (34). Thus, an apparent size shift of myc-tagged ADAMTS10 is observed under nonreducing conditions, indicating cleavage of the inhibitor and entrapment of the enzyme (Fig. 5C). Pretreatment of the enzyme with 10 mM EDTA resulted in the abolition of the size shift, confirming the proteolytic activity of ADAMTS10 as a metalloprotease (Fig. 5C).
To assess whether ADAMTS10 was able to cut the large aggregating proteoglycan aggrecan, we used an antibody that detects a neoepitope generated by ADAMTS4/5/9 cleavage of the Glu 1771 -Ala 1772 peptide bond of the aggrecan core protein

FIG. 3. In situ hybridization of Ad-amts10 in 15.5-day-old (A) and 17.5day-old (B) mouse embryos.
Embryos are oriented with the cephalic (head) end facing up, and the sections were cut in the sagittal plane. Autoradiographic signal from the hybridized probe is red (pseudocolor imparted to signal from dark field microscopy), and the cell nuclei are blue (Hoechst 33258 visualized by UV light). Specific regions of interest are indicated as follows. cc, cerebral cortex; cfm, craniofacial mesenchyme; d, dermis; g, gonad; ge, ganglionic eminence over the caudate nucleus; gt, genital tubercle; h, heart; lu, lung; lv, liver; ma, mandible; pc, perichondrium; sc, spinal cord; smg, submandibular gland; k, kidney; ࡗ nasal epithelium. The asterisks in B indicate autofluorescence from coagulated blood. (26). ADAMTS10 digestion produced no detectable immunoreactivity of aggrecan to the neoepitope antibody on Western blots (Fig. 5D). In contrast, ADAMTS4 digestion did generate the neoepitope, as has been previously described (Fig. 5D) (20,26).
Characterization of ADAMTS10 Zymogen Processing-Because of its smaller size, which allows for easier discrimination of processed species from the zymogen, ADAMTS10-(1-463) was used for studies of zymogen processing. ADAMTS10-(1-463) was substantially purified from stably transfected HEK293F cells by chromatography on nickel-Sepharose (Fig.  6A, left panel). Bands visible on Coomassie Blue-stained gels (Fig. 6A, left panel) corresponded to bands prominent on Western blotting of the purified protein using anti-myc (Fig. 6A,  right panel), although the smaller amount of protein in Fig. 5B better permits distinction of the 52-and 54-kDa bands. Nterminal sequencing indicated the origin of the bands as follows: secreted zymogen (54 kDa; sequence NH 2 -Phe 26 -Arg-Ser-Gln-Asp), thus confirming the predicted signal peptidase cleavage site; partially processed zymogen (52 kDa; NH 2 -Gly 67 -Thr-Gly-Ala-Thr); and fully processed catalytic domain (29 kDa; NH 2 -Ser 234 -Val-Ser-Arg-Glu), respectively (Fig. 6A). Incubation of cells with the lipid-permeable furin inhibitor dec-RVKR-cmk showed that formation of the 29-kDa form was suppressed when cells were incubated in the presence of 10 -100 M inhibitor, but the levels of the larger forms or of the 37-kDa intermediate were affected only at correspondingly higher concentrations.
Some Secreted ADAMTS10 Localizes to Cell Surface and ECM-ADAMTS10 could be extracted from COS-1 or HEK FIG. 4. Organ-specific expression of Adamts10 in mouse development. In these paired panels, coupled dark field microscopy-fluorescent microscopy overlays are shown on the left, and the corresponding fluorescent image is shown on the right (except F and G) to provide morphological correlation. In the left panels, autoradiographic signal from the hybridized probe is red (pseudocolor imparted to signal from dark field microscopy), and the cell nuclei are blue (Hoechst 33258 visualized by UV light). Lack of hybridization to the sense probe is illustrated in G. A-L are from 15.5-day-old embryos, and M is from a 17.5-day-old embryo. A, submandibular gland. The asterisk indicates an epithelial lobule. B, developing kidney (kd) and adrenal (ad) show cortical signal. C, developing cochlea. A semicircular canal is shown in cross-section to illustrate lack of signal in lining neuroepithelium (ne) and surrounding cartilage (c). D, hip joint, showing the absence of signal in the developing femoral head cartilage (fe) and strong signal in overlying ligaments and tendons (l) and joint capsule. E, snout region, showing expression of Adamts10 in mesenchyme but not in hair follicle epithelium (hf). F, blood vessel (bv) showing expression in the wall. G, lack of sense probe hybridization to lung (compare with Figs. 3, A and B, and 4K). H, expression in gonad, most likely a testis (t), from the striped appearance, and in the wall of an adjacent artery (a). I, expression in developing gonad, likely to be a prospective ovary, showing that expression is absent in tubular epithelium (asterisk). J, expression in cerebral cortex is highest in the ventricular zone (vz). v, ventricle; cp, choroid plexus of lateral ventricle. K, in the lung, expression is excluded from bronchial epithelium (bronchial tubes are marked by the asterisk). L, section through the vertebrae, showing strong expression in dorsal root ganglion (drg) and bone (b). M, embryonic day 17.5 vertebral column shows strong expression in cartilage (c) and dorsal ligament (d), but not in intervertebral disc (ivd). Compare this with the lack of cartilage expression in 14.5-day embryos (C and D).
293F cells by incubation with 0.5 M NaCl (data not shown), suggesting that some ADAMTS10 is retained in the pericellular region by interaction with either cell surface or pericellular matrix components. Accordingly, nonpermeabilized COS-1 cells were immunostained with anti-myc antibody. Whereas untransfected cells showed no background fluorescence (Fig.  7A), transfected cells (Fig. 7B) showed significant levels of punctate fluorescence that was localized to both their surface (Fig. 7C) and their underlying matrix (Fig. 7, D and E). However, this does not seem to be the sole fate of secreted AD-AMTS10, since substantially more was detected in conditioned medium from these cells (Fig. 7F). Since 293 cells do not adhere well enough to the substratum to undertake immunostaining of live cells, we determined whether ADAMTS10 was cell-associated by biotinylating surface proteins. Suspensions of stably transfected HEK293F cells were treated either with trypsin (to remove all cell surface proteins) or with serum-free medium prior to biotinylation. Biotinylated proteins were recovered from cell lysate by streptavidin-Sepharose and analyzed by SDS-PAGE. The serum-free medium-treated cells contained a single c-myc immunoreactive band of 130 kDa that was trypsin-sensitive (Fig. 7G), indicating localization external to the plasma membrane. This result, taken together with the observation that HEK293F cells produce insignificant amounts of matrix, suggests that the secreted ADAMTS10 is held in proximity to the cell surface.

DISCUSSION
ADAMTS10 is probably the last mammalian ADAMTS that will be identified, taking the total number of mammalian AD-AMTS proteases to 19. Since over 99% of the human and mouse genomes have been annotated, it is very unlikely that additional ADAMTS proteases will be discovered in mammals. Although ADAMTS10 has a different domain organization from the cognate ADAMTS6, it preferentially clustered with AD-AMTS6 in sequence alignment and phylogenetic analysis (23). Since all ADAMTS enzymes within a clade have identical domain structures (23), we considered the possibility that there might exist an alternative form of ADAMTS6 having the domain structure of ADAMTS10. This appears to be the case. Although ADAMTS6B has not yet been cloned as a single contiguous mRNA, the similarity of the conceptual product to ADAMTS10 was felt to be of sufficient relevance to describe it here. The genesis of the alternative transcript and its tissuespecific expression are beyond the scope of the present work and will be described elsewhere. 4 Despite also having a domain 4 S. A. Oblander and S. S. Apte, unpublished data.  (1-463), and the medium was analyzed by Western blot using antibody 9E10. Note the reduced formation of the 29-kDa fully processed form (arrow) starting at 10 M that is nearly complete at 100 M. The N terminus of the 37-kDa intermediates is not known. structure similar to ADAMTS10, ADAMTS19 and ADAMTS17 are highly homologous to each other but not to ADAMTS6 or ADAMTS10 and constitute a separate subclass (23).
The primary structure of ADAMTS10 predicts many of the typical features of ADAMTS proteases. Like the majority of them, it is a secreted glycoprotein. The experimental data indicate that it is synthesized as a zymogen that is processed by proprotein convertases in the secretory pathway. However, the maturation process appears to be inefficient, since unprocessed zymogen is also secreted. In this respect, ADAMTS10 differs from other ADAMTS proteases. In HEK293F cells, both the ADAMTS10 zymogen and fully processed form are secreted from the cell, whereas ADAMTS9-transfected cells produce only the fully processed form (20). In ADAMTS7, where we previously described stepwise zymogen processing, some of the final processing to mature enzyme occurred at the cell surface and was substantially inhibited by 10 M dec-RVKR-cmk with complete inhibition achieved by 25 M (22). On the other hand, the final processing step in ADAMTS10 requires 100 M dec-RVKR-cmk for nearly complete inhibition. This suggests that this lipid-permeable inhibitor needs to penetrate into intracellular compartments to affect the final ADAMTS10 processing step. Curiously, the processing site for production of the mature ADAMTS10 enzyme lacks the typical proprotein convertase consensus sequence Arg-Xaa-Arg/Lys-Arg2 (35,36). Sequence comparison of a large number of furin substrates suggests an absolute requirement for the P1 Arg and requires that at least two of the three residues at P2, P4, and P6 must be basic for efficient cleavage (35,36). The cleaved site in ADAMTS10 does not fulfill the second requirement, although there is a Ser residue at the P1Ј position in ADAMTS10 that was noted in about 50% of furin substrates (37). Despite these discrepancies, however, both N-terminal sequencing and dec-RVKR-cmk inhibition suggest that the completely processed form is generated by a proprotein convertase, albeit inefficiently. Since ADAMTS10 is widely expressed and furin is the only ubiquitous proprotein convertase (35,37), it is likely to be the physiological processing enzyme of ADAMTS10. The origin of the 37-kDa intermediate found in purified ADAMTS10-(1-463) preparations has not been established; however, its formation appears to be inhibited by dec-RVKR-cmk, suggesting that it may result from processing at another atypical furin cleavage site Val-Tyr-Lys-Arg 182 2Ser that is conserved in mouse and human ADAMTS10. This putative site lacks both the P4 and P6 Arg residues required for optimal processing, but it has a P1Ј Ser. The production of unprocessed and partly processed zymogen might imply that the furin cleavage is inefficient because of the suboptimal recognition sequences at two of three processing sites. Partial activation could be an important physiological mechanism by which this highly expressed protease is regulated post-translationally.
Previous studies have demonstrated localization of AD-AMTS1, ADAMTS4, ADAMTS7, and ADAMTS9 to the cell surface (20, 22, 38 -40). The putative GAG-binding sequences in ADAMTS10 might mediate such localization, and as previously shown for ADAMTS1, ADAMTS4, and ADAMTS7, the cell surface may be a staging area for further proteolytic activation steps (22,38,39). ADAMTS10 in serum-free medium undergoes substantial proteolysis. Western blotting using monoclonal antibody 9E10 to the C-terminal myc epitope identified several myc-reactive fragments representing C-terminal processing events analogous to those reported for ADAMTS1 FIG. 7. Distribution of ADAMTS10 in transfected COS-1 cells. A-E, representative fluorescent microscopy images of ADAMTS10-transfected COS-1 cells stained with anti-myc antibody without permeabilization followed by a Alexa 488conjugated secondary antibody (green). Cell nuclei were visualized by blue fluorescence (4Ј,6-diamidino-2-phenylindole). A shows that untransfected cells lack fluorescence. B, an overview of the staining to show that only a fraction of the cells express ADAMTS10. C and D, higher power images of a cell in two planes of focus. The image in C is focused midway between the free and attached surfaces of the cell, whereas the image in D is focused in the plane of the cell substratum. These planes of photography and the cellular distribution are shown in the schematic diagram in E. F, Western blot analysis (antibody 9E10) of medium, lysate, and subcellular extracellular matrix of COS-1 cells transiently transfected with AD-AMTS10. A major 130-kDa band (arrow) is seen in all lanes, and a significant amount of a 40-kDa degradation product is present in the culture medium. Note that ADAMTS10 in the ECM is intact. The two bands in ECM could represent glycosylated and unglycosylated forms or zymogen and processed forms. G, AD-AMTS10 is biotinylated at the cell surface. A major 130-kDa band is biotinylated at the cell surface in transfected HEK293F cells not treated with trypsin prior to biotinylation. Prior treatment with trypsin (ϩ) eliminates cell surface ADAMTS10. and ADAMTS4. The removal of C-terminal modules in these proteases has been shown to have a profound effect on enzyme activity and specificity (19,38,40). Once specific ADAMTS10 substrates are identified, the regulatory role of C-terminal proteolysis can be studied in greater detail. The inhibition of proteolysis in the presence of serum, perhaps by broad spectrum protease inhibitors such as ␣ 2 -M suggests that the responsible enzymes originate in the 293 cells and that this processing of ADAMTS10 occurs extracellularly.
Few ADAMTS proteases other than ADAMTS9 have been shown to have such a broad expression profile, since constitutively active proteases such as the ADAMTS are likely to be highly regulated at the transcriptional and post-transcriptional level. An emerging theme in the ADAMTS family is that enzymes of a given subfamily appear to have very different expression patterns and levels and that biological roles for each enzyme may be determined by its nonredundant sites of expression. As a case in point, ADAMTS9 is widely expressed, but its homolog, ADAMTS20, is only sparingly expressed. Similarly, ADAMTS6 appears to be expressed primarily in the placenta (24), but ADAMTS10 is very widely expressed. Interestingly, the highest embryonic expression of ADAMTS10 is in the developing lung and among human tumor cell lines examined, in a lung carcinoma cell line. Since antibodies to AD-AMTS10 are not presently available, it is not known whether all of the expressed ADAMTS10 RNA is translated into protein.
Although ADAMTS10 may have a broad participation in mesenchymal and basement membrane remodeling in a variety of morphogenic processes, this does not imply that it has an essential role at all expression sites. Indeed, some widely expressed proteases have seemingly few nonredundant roles during development. The determination of specific developmental roles in genetic models as well as screens for substrates will provide more insight to follow on this initial characterization of ADAMTS10.