A Novel, Secreted Form of Human ADAM 12 (Meltrin α) Provokes Myogenesis in Vivo *

The ADAM (A DisintegrinAnd Metalloprotease) family of cell-surface proteins may have an important role in cellular interactions and in modulating cellular responses. In this report we describe a novel, secreted form of human ADAM 12 (meltrin α), designated ADAM 12-S (S for short), and a larger, membrane-bound form designated ADAM 12-L (L for long form). These two forms arise by alternative splicing of a single gene located on chromosome 10q26. Northern blotting demonstrated that mRNAs of both forms are abundant in human term placenta and are also present in some tumor cell lines. The ADAM 12-L transcript can also be detected in normal human adult skeletal, cardiac, and smooth muscle. Human A204 embryonal rhabdomyosarcoma cells that do not differentiate into muscle cells and do not express any form of ADAM 12 were stably transfected with an ADAM 12-S minigene encoding the disintegrin domain, the cysteine-rich domain, and the unique 34 amino acid carboxyl terminus. Nude mouse tumors derived from these transfected cells contained ectopic muscle cells of apparent mouse origin as shown by species-specific markers. These results may have potential applications in the development of muscle-directed gene and cell therapies.

ADAMs 1 are a recently discovered family of membrane-anchored cell-surface proteins. They are about 800 amino acids long and have a unique domain organization, containing pro-, metalloprotease, disintegrin, cysteine-rich, transmembrane, and cytoplasmic domains (1)(2)(3)(4). Because these domains are homologous to domains in proteins with established functions, ADAMs have been proposed as candidates for modulating pro-teolysis, cell adhesion, cell fusion, and signaling. The ADAMs have structural similarity and ϳ30% sequence identity to snake venom metalloproteases (SVMPs), which cause hemorrhage in snake bite victims (5,6). ADAMs and SVMPs are both members of the reprolysin subfamily of metalloproteases (6,7). Full-length SVMPs are processed to generate a metalloprotease, which is able to degrade proteins of the basement membrane such as type IV collagen and laminin (5), and a disintegrin domain, which can inhibit the function of platelets by interacting with platelet integrin GPIIb-IIIa (8). Within the past few years the number of identified ADAMs has expanded rapidly, and to date 18 different members have been reported in the literature (3, 4, 9 -14). For example, ADAM 10 (Kuz) plays a critical role in neurogenesis in Drosophila (9,10,15,16); ADAM 11 is a candidate tumor suppressor gene (17,18); and one of the most recently identified ADAMs, called TACE, is a tumor necrosis factor ␣-converting enzyme (11,13). The most studied ADAMs are fertilin ␣ and ␤ (ADAM 1 and 2) (19 -21) which interact with ␣6␤1 integrin during sperm-egg fusion in fertilization (22). In humans, however, ADAM 1 is a non-functional pseudogene (23), indicating that different species may utilize different sets of ADAM-integrin links.
The process of differentiation leading to cell fusion occurs in several different tissue types including placenta, bone, and muscle. During myoblast differentiation, the cells align and adhere to each other before their plasma membranes merge allowing the formation of multinucleated myotubes (24). On the assumption that myoblast fusion may be similar to spermegg fusion, Yagami-Hiromasa et al. (25) searched for homologs of ADAMs 1 and 2 in a mouse myogenic cell line and identified ADAM 12 (meltrin ␣). ADAM 12 showed strong expression in neonatal skeletal muscle and bone. In mouse C2 myoblast cultures, the expression of ADAM 12 became apparent upon muscle cell differentiation. Evidence for a role in muscle cell fusion was provided by studies showing that transfection of mouse C2 cells with a minigene of adam 12 lacking the pro-and metalloprotease domains accelerated cell fusion, whereas antisense constructs blocked myoblast fusion.
Here we describe the cloning of soluble and transmembrane forms of human ADAM 12 that arise by alternative splicing. We show the effects of the novel secreted form, designated ADAM 12-S, on recruitment and differentiation of ectopic muscle cells in a human tumor nude mouse model. These results may have potential applications in the development of a number of future muscle-directed gene and cell therapies.

EXPERIMENTAL PROCEDURES
Isolation and Sequencing of Human ADAM 12 cDNA Clones-A positive prey clone (S1) was isolated from a human yeast two-hybrid placental cDNA library (CLONTECH catalog number HL4025AH) us-ing a cDNA fragment of the laminin ␤2 chain as bait. 2 This clone was sequenced and found to be similar to mouse ADAM 12. A probe corresponding to the disintegrin domain (nt 1540 -1963) was used to screen a human placenta 5Ј-stretch plus gt11 cDNA library (CLONTECH catalog number HL50146). Twenty-six positive phage were plaquepurified, and the inserts from seven of the phage were subcloned into pBluescript SK(ϩ) (Stratagene). Sequencing was performed using either the Sequenase enzyme and reagents (Amersham Corp.) or the Vistra DNA Sequencer 725 (Amersham Corp.). DNA sequence analysis was performed using the software programs of the Wisconsin Package, version 9.1, of the Genetics Computer Group. Searches of expressed sequences data base (dBEST) were performed using Blast server (26). 3 Chromosomal Mapping-The chromosomal localization of the ADAM 12 gene was performed by fluorescence in situ hybridization. Metaphase spreads were prepared from phytohemagglutinin-stimulated human lymphocytes. The 5.1-kb cDNA insert of the ADAM 12-L phage (L1) in pBluescript was labeled with biotin-16-dUTP by nick translation and hybridized to the chromosome spreads as described previously (27), and the probe was detected by means of fluorescence isothiocyanateconjugated avidin (Vector laboratories). Chromosomes were counterstained and R-banded with propidium iodide as described (28).
Analysis of Alternatively Spliced Exons-To test the hypothesis that the two forms of ADAM 12 were splice variants, primers were designed to amplify the genomic DNA around the point of divergence. Human genomic DNA from HT1080 fibrosarcoma cells was amplified with the following primers: primer 322, a sense primer at nt 2378 -2400 common to both forms of ADAM 12 (5Ј-dGTTTGGCTTTGGAGGAAGCACAG); primer 323, an antisense primer at nt 2460 -2440 of ADAM 12-S (5Ј-dGCTCCCTGTTGGACTCTGCAG); primer 325, a sense primer at nt 3252-3274 of ADAM 12-S (5Ј-dCAATGTAAGAGCCTAACTCCATC); and primer 324, an antisense primer at nt 2498 -2476 of ADAM 12-L (5Ј-dGAGATAAACCACAAATCCGGCAG). The conditions for amplification were as follows: 1 cycle of 94°C for 2 min, 30 cycles of 94°C for 40 s, 55°C for 40 s, and 72°C for 3 min. The products were gel-purified, subcloned into the vector pCR 2.1 (Invitrogen), and sequenced.
Analysis of ADAM 12 mRNA Expression-Human multiple tissue Northern blots containing ϳ2 g per lane of poly(A) RNA were purchased (CLONTECH catalog numbers 7760-1 and 7765-1). Poly(A)enriched RNA was extracted from cultured cell lines using the Trizol reagent (Life Technologies, Inc.), ϳ15 g per lane was fractionated by formaldehyde-agarose gel electrophoresis and blotted onto nylon membranes using standard protocols (29). Blots were hybridized with [ 32 P]dCTP random primer labeled probes at 68°C for 1 h in 10 ml per blot of QuikHyb solution (Stratagene). Blots were washed three times at 50°C for 15 min in 2 ϫ SSC, 0.05% SDS, and twice for 20 min in 0.1 ϫ SSC and 0.1% SDS. Nucleotide probes used in this work were hybridized to the ADAM 12 pro-domain (nt 664 -1007), disintegrin domain (nt 1540 -1963), or were specific for ADAM 12-S (nt 2409 -3333), or specific for the coding region (nt 2476 -2717) or 3Ј-untranslated region (nt 4227-5067) of ADAM 12-L. A 467-bp probe specific for ADAM 12-L 8.6-kb band was also produced via PCR using primers (5Ј-dAC-CAGGGTGTTTTGTGGTTG and 5Ј-dTGCTGCTTTTGTGGTTTCTG) designed after examination of EST data bases (see below). Blots were were exposed to Kodak X-Omat AR film at Ϫ80°C with intensifying screens.
Cell Lines and Cell Culture-The cell lines used in this study included COS-7 (ATCC CRL 1651), HT1080 (ATCC CCL 121), A204 human embryonal rhabdomyosarcoma (ATCC HTB 82), RD human embryonal rhabdomyosarcoma (ATCC CCL 136), and HU-1 human lung adenocarcinoma (30). The cells were grown in DMEM with Glutamax I and 4500 mg/ml glucose, 50 units/ml penicillin, 50 g/ml streptomycin, and 10% fetal bovine serum (Life Technologies, Inc.) at 37°C in 5% CO 2 . Myogenic differentiation of RD cells was induced at confluence by replacing the growth medium with DMEM containing 1% fetal bovine serum as described (31). Under the same culture conditions no myogenic differentiation of A204 cells was observed, by morphological criteria and by lack of induction of expression of myogenin mRNA by RT-PCR (not shown).
Purification of Recombinant ADAM 12 and Production of Poly-and Monoclonal Antibodies-A plasmid for the production of recombinant ADAM 12 in Escherichia coli was constructed using the pQE32 His-tag expression vector (Qiagen). A 450-bp BamHI/XhoI fragment coding for nt 2000 -2433 of ADAM 12-S was inserted at the BamHI/SalI sites of pQE32 and transformed into E. coli strain M15[pREP4]. This plasmid (p1053) codes for a 17-kDa recombinant protein containing the cysteinerich domain and the first four amino acids (EARQ) of the unique carboxyl terminus of ADAM 12-S. Recombinant protein was purified as follows: 200 ml of E. coli culture was sedimented, lysed in 0.020 M Tris-HCl, pH 7.9, 0.5 M NaCl, 6 M guanidine HCl, centrifuged, and the supernatant applied to a column of TALON immobilized metal affinity resin (CLONTECH). After washing the resin with a urea buffer (0.020 M Tris-HCl, pH 7.9, 0.5 M NaCl, 6 M urea, 0.01 M imidazole), elution of bound material was performed with 0.05 M EDTA in 0.020 M Tris-HCl, pH 7.9, and 0.15 M NaCl. The yield from a typical preparation was ϳ2.5 mg.
Lewis female rats (Møllegaarden, Denmark) and female rabbits (Statens Seruminstitut, Copenhagen, Denmark) were immunized and boosted at monthly intervals with total E. coli fusion protein extract or with purified recombinant ADAM 12 derived from expression construct (p1053) emulsified in complete and incomplete Freund's adjuvant. Antisera (rb 104) were collected 10 -11 days after the second and all subsequent injections. To prepare monoclonal antibodies, rats immunized and boosted eight times were given a final boost intraperitoneally, and 5 days later hybridomas were prepared by fusing spleen cells from the rat with the nonsecreting mouse myeloma P3 ϫ 63Ag8.653 (ATCC TIB 18) as recently described (32). Supernatants of the resulting hybridomas were screened and characterized for their immunostaining of COS-7 cells transiently transfected with construct number 1095 (see below). The isotype of the rat 14E3 hybridoma was IgG2b as determined by Ouchterlony immunodiffusion using a series of anti-rat immunoglobulins purchased from Serotec and by the IsoStrip kit from Boehringer Mannheim. Hybridomas were grown in DMEM with Glutamax I and 4500 mg/ml glucose, 1 mM sodium pyruvate, 10 mM HEPES, OPI media supplement (0.15 g/ml oxaloacetate, 0.05 g/ml pyruvate, 0.0082 g/ml bovine insulin (Sigma)), 50 units/ml penicillin, and 50 g/ml streptomycin and 20% myoclone super plus fetal bovine serum (Life Technologies, Inc.) at 37°C in 10% CO 2 .
Transfection Assays, Immunostaining, and Immunoblotting-A plasmid for expression of an ADAM 12-S minigene was constructed using the pSecTagB vector (Invitrogen). A DNA fragment coding for the disintegrin domain, cysteine-rich domain, and the unique carboxyl terminus of ADAM 12-S was prepared by PCR amplification using the ADAM 12-S cDNA plasmid as a template and the following primers: 5Ј-dCCAAAGCTTGAAGTCAGGGAGTCTTTC and 5Ј-dCCATCTAGAT-CAGATGAGTGTCAGTGA. The 987-bp PCR product contained nt 1560 -2528 of ADAM 12-S, with HindIII and XbaI cloning sites. This fragment was inserted at the HindIII/XbaI sites of pSecTagB, yielding plasmid p1095, consisting of an ADAM 12 minigene driven by a cytomegalovirus promoter, fused to an Ig -chain leader sequence to allow secretion of the protein.
For transient transfections, COS-7 cells were electroporated as described previously (32) with a Bio-Rad Gene Pulser II, using 250 V and 1000 microfarads for 0.4 ml of cells and 10 g of plasmid in PBS/ HEPES, with an electrode gap of 0.4 cm. After electroporation, the cells were plated in Lab-Tek 8-well chambers (Nunc 177402). Cells were transfected with an expression plasmid for a human ADAM 12-S minigene (p1095) or the expression vector with no cDNA insert (pSecTagB). Two to three days later, immunostaining was performed as described previously (32). Briefly, the cells were rinsed with PBS, fixed with cold methanol, rinsed with PBS, and incubated with the primary antibodies diluted in 0.05% Tween 20, 0.05 M Tris-HCl, pH 7.2 (1:100 for the polyclonal antibodies and 1:2 with culture medium supernatant of the monoclonal antibody), for 1 h at room temperature. After rinsing, the sections were incubated with fluorescein-conjugated secondary antibodies for 1 h, washed, the slides mounted in buffered glycerol and examined under a Zeiss LSM-10 laser scan microscope.
A204 cells were stably transfected with an expression plasmid for a human ADAM 12-S minigene (p1095) or the expression vector with no cDNA insert (pSecTagB). Cells were transfected with LipofectAMINE from Life Technologies, Inc., using a ratio of 2 g of DNA to 20 l of LipofectAMINE. Two days' post-transfection, the cells were trypsinized and replated in DMEM containing 10% fetal bovine serum and 500 g/ml Zeocin (Invitrogen). Zeocin-resistant colonies were selected and grown in the presence of 200 g/ml Zeocin. Clones were assayed for the expression of ADAM 12 by Northern blot analysis of total cellular RNA using ADAM 12 common region cDNA as a probe and by immunoblotting using ADAM 12-specific antibodies. For detection of ADAM 12 in the medium of transfected cells, confluent cultures of cells were incubated in serum-free UltraDOMA-PF medium (BioWhittaker) for 2 days. The medium was concentrated 10-fold using an Amicon Centricon-10 filter. Samples were subjected to SDS-polyacrylamide gel electrophoresis on 10 -20% gradient gels (Novex) and transferred to nitrocellulose membranes. The membranes were incubated with medium from 14E3 hybridoma cells or rb 104 polyclonal antiserum and subsequently with peroxidase-conjugated rabbit anti-rat or swine anti-rabbit immunoglobulins (DAKO). Detection was performed using the enhanced chemiluminescence SuperSignal kit from Pierce.
Heterotransplantation into Nude Mice, Morphological Examination, Immunohistochemistry, and RT-PCR-A204 parental cells or transfectants were harvested by trypsin/EDTA, equilibrated in complete medium for 1 h in suspension at 37°C, rinsed in PBS, and finally resuspended at a concentration of 10 8 cells/ml. Tumor cells (10 7 per inoculum) were injected subcutaneously via a 26-gauge needle into the back of female (6 -8 week-old) nu/nu NMRI mice (Bomholtgaard, Denmark). Mice were killed by cervical dislocation after 6 -8 weeks. Tissue specimens were fixed in buffered formalin at room temperature or in cold 96% ethanol/glacial acetic acid (99:1 v/v) and processed for histological examination of hematoxylin and eosin-stained paraffin sections using standard techniques. For electron microscopy 2-to 3-mm 3 tissue specimens were promptly fixed at room temperature in Karnovsky's fixative for 2 h. The samples were dehydrated in graded ethanols, postfixed in 2% osmium tetroxide, and embedded in Epon. One-micrometer sections were cut on an LKB ultramicrotome and stained with toluidine blue. Ultrathin sections from selected areas were collected on copper grids, stained with uranyl acetate and lead citrate, and examined with a Phillips 201 electron microscope. Tissue specimens were also frozen in liquid nitrogen and stored at Ϫ70°C for later use in immunohistochemistry or RNA purification.
Immunostaining was performed using standard methods as described previously (30,32) or using the DAKO optimized staining system for automated slide processing according to the manufacturer's protocol (DAKO). For "manual" staining, frozen sections were cut, airdried, and fixed in precooled acetone. Formalin or ethanol/glacial acetic acid fixed paraffin sections were deparaffinized, and endogenous peroxidase was inhibited by treatment with 10% H 2 O 2 in methanol for 10 min at room temperature. Both frozen sections and paraffin sections were incubated with the mAb or rabbit antisera diluted as indicated and incubated for 1 h at room temperature. Following a thorough rinse, the sections were incubated with fluorescein isothiocyanate-or peroxidase-coupled swine anti-rabbit, rabbit anti-mouse, or rabbit anti-rat immunoglobulins (DAKO). On control sections, the specific antibodies were omitted or replaced with irrelevant mouse monoclonal antibodies of the same isotype or with non-immune mouse, rat, or rabbit serum. The slides were mounted in buffered glycerol and examined under a Zeiss LSM-10 laser scan confocal microscope.
RT-PCR was applied to examine for the presence of myf-5 transcript in cultured cells and nude mouse tumors using a Stratagene kit and primers specific for mouse myf-5 (5Ј-dCTCTCCCGATGATCACTCCT and 5ЈD-CCTGTAATGGATTCCAAGCTG), derived from GenBank number X56182 (35).

RESULTS
Cloning and Sequencing of Human ADAM 12 cDNAs-We screened a yeast two-hybrid cDNA library with laminin ␤2 cDNA as "bait," and one of the positive "prey" clones was homologous to mouse ADAM 12 (meltrin ␣) but had a divergent carboxyl terminus. To determine whether this represented an alternatively spliced form of ADAM 12, we set out to isolate full-length cDNA for human ADAM 12. We isolated cDNA clones that cover the full-length of the human homologue of mouse ADAM 12, designated ADAM 12-L (L for long), and partial clones of the smaller ADAM 12-S (S for short). A map of the clones is shown in Fig. 1A and the nucleotide and deduced amino acid sequence in Fig. 1B.
The full-length ADAM 12-L cDNA shown in Fig. 1B spans 5048 nt, including a 311-nt 5Ј-untranslated region, an open reading region frame of 2727 nt encoding 909 aa, a TGA stop codon, and a 3Ј-untranslated region of 2006 nt. The longest ADAM 12-S cDNA clone obtained began halfway through the pro-domain at nt 696. A full-length ADAM 12-S cDNA would have a 2214 nt open reading frame that is identical to ADAM 12-L up to nucleotide 2426, whereupon it diverges (Fig. 1, A  and B). The final 102 nt of the ADAM 12-S open reading frame encode a 34-aa carboxyl terminus, followed by a TGA stop codon, and a 3Ј-end untranslated region of 788 nt. The 3Јuntranslated regions were different in the two human ADAM 12 forms.
The open reading frame begins at the translation initiation codon ATG at nt 312. The first 28 aa (residues 1-28) encode a typical signal peptide, and the signal cleavage site is predicted to occur after the sequence CEN, in agreement with the "-3,-1" rule (36). The mature human ADAM 12-L contains 881 aa with a calculated M r of 96,917 and that of ADAM 12-S contains 718 aa with an M r of 77,775. Five potential N-linked glycosylation sites (NX(S/T)) are present. All five are also found at the same position in mouse ADAM 12, whereas three additional sites present in mouse are not found in human ADAM 12.
Analysis of the amino acid sequence of the human ADAM 12 revealed that it has a structural organization typical for the members of the ADAM family (1), shown schematically in Fig.  1A. Human ADAM-L and -S share a common region consisting of the prodomain (residues 29 -206), the metalloprotease domain (residues 207-417), the disintegrin domain (residues 417-512), and the cysteine-rich domain (residues 529 -614) that contains the putative fusion peptide. ADAM 12-L has a 21-aa transmembrane domain and a 179-aa cytoplasmic domain. ADAM 12-S has instead a shorter 34-aa carboxyl terminus with no apparent transmembrane domain. Comparison of human ADAM 12-L sequence with mouse ADAM 12 revealed an overall amino acid identity of 81%. Within the individual domains, the sequence similarity to mouse ADAM 12 was high in the cysteine-rich, metalloprotease, and disintegrin domains and lower in the pro-and cytoplasmic domains (Table I). We also compared the ADAM 12 amino acid sequences to all other known ADAMs, and the comparison with the four most similar ADAMs (the Xenopus ADAM 13, human ADAMs 8, 15, and 9) is shown in Table I. The most conserved sequences are in the metalloprotease and the disintegrin domains, and the least conserved regions are the prodomain and the cytoplasmic tail. The divergent carboxyl terminus of ADAM 12-S showed no similarity to any of the other known ADAM proteins nor to any other proteins in the data bases.
The human ADAM 12 metalloprotease domain contains the highly conserved zinc-binding motif HEXGHXXGXXHD regulated by a potential "cysteine switch" in the prodomain (3). This sequence is identical to mouse ADAM 12 and, as with other ADAMs containing this motif, is presumed to be catalytically active. The disintegrin domain contains a putative integrin binding loop, although like other ADAMs and the related P-III SVMPs, ADAM 12 does not have an RGD sequence (3). Both human and mouse ADAM 12 have the amino acids SNS at this position followed by an additional cysteine residue. The cysteine-rich domain of human ADAM 12 contains the putative fusion peptide-like sequence that can be modeled as a one-sided ␣-helix with one strongly hydrophobic face (2) and an epidermal growth factor-like repeat (37). ADAM 12-L contains a 21-aa, highly hydrophobic (18/21 aa) transmembrane domain which is consistent with the consensus sequence motif for type I membrane proteins (38). In addition, the flanking amino acid FIG. 1 sequence is consistent with the amino terminus being exposed to the cell exterior. The cytoplasmic domain of human ADAM 12-L is proline-rich (32 out of 179 aa) and contains at least three sites (RXXPXXP) that are potential ligands for the Src homology 3 domain (SH3) (39), as has been demonstrated for the proline-rich motifs in ADAM 9 (40).
Chromosomal Localization-By using fluorescent in situ hybridization, we mapped the chromosomal localization of the human ADAM 12 gene (ADAM12). 90% of the 30 metaphase cells analyzed showed specific fluorescent spots on the q26 band of the long arm of human chromosome 10 (Fig. 2). Subsequently we searched the data base of mapped STSs on the human genome (41) 5 and identified an STS (WI-17472) that is identical in sequence to part of the 3Ј-untranslated region of ADAM 12-L (nt 4044 -4145). WI-17472 was placed on the distal region of chromosome 10 by radiation hybrid mapping, in the interval between the Genethon markers D10S216 and D10S575 (158 -162 centimorgan), consistent with our cytogenetic localization of the gene. Thus the ADAM12 gene is located at 10q26.3.
ADAM 12-L and ADAM 12-S Arise by Differential Splicing-The finding of two cDNA forms of ADAM 12 sharing identical 5Ј-regions, but diverse 3Ј-ends, suggested that they were alternatively spliced versions of a single gene. This hypothesis was strengthened by the single chromosomal localization and the observation that a probe for the pro-domain common to both forms of ADAM 12 hybridized to a single band in human high molecular weight genomic DNA digested with seven different restriction enzymes (data not shown). We designed PCR primers to amplify the genomic DNA around the point of divergence in the ADAM 12 clones. Primers 322 and 324 amplified a 4-kb, 325 and 324 a 2-kb, and 322 and 323 a 1-kb DNA fragment (Fig.  3). Sequencing of these revealed that the ADAM 12 gene contains an intron at the point of divergence between the clones, followed by an exon encoding the ADAM 12-S-specific sequence, which does not appear to have any introns within it. At the end of the ADAM 12-S sequence, about 2 kb of intron DNA are present before the ADAM 12-L coding sequence. A consensus 5Ј-donor site was found at the point of divergence between the ADAM 12-L and -S sequences, and 3Ј-acceptor sites were present at the start of both the ADAM 12-L-and ADAM 12-Sspecific sequences.
Human ADAM 12 mRNA Expression-We examined the expression of ADAM 12 mRNA using probes common to both forms of ADAM 12 and probes specific for each form. Northern blot analysis with a probe for the disintegrin domain present in both forms of human ADAM 12 revealed three bands of 3.5, 5.4, and 8.6 kb in human full-term placenta RNA, expressed at relative levels of 2:1:1 (Fig. 4A, lane 1). Probes specific for the ADAM 12-S clone hybridized only to the smallest 3.5-kb band (Fig. 4A, lane 3), whereas probes specific for human ADAM 12-L hybridized to the two top bands only (Fig. 4A, lane 2). The 3.5-and 5.4-kb bands correspond to the sizes of the full-length ADAM 12-S and -L cDNAs, although we did not isolate a full-length cDNA clone equivalent in size to the 8.6-kb band observed on Northern blots. As the 8.6-kb band hybridized to the same probes as the 5.4-kb ADAM 12-L transcript, this suggests that the 8.6-kb transcript contained the same ADAM 12-L sequence as the 5.4-kb band but had an extended 3Јregion. To clarify this further, we searched the dbEST data base with 500 bp of the untranslated region at the 3Ј of the   2. Mapping of the human ADAM 12 gene by in situ hybridization. A, partial metaphase spreads observed after hybridization to a biotinylated human ADAM 12 cDNA probe. Arrows indicated specific hybridization signals detected using fluorescein isothiocyanate-conjugated avidin. B, the same partial metaphase spreads observed after R-banding and staining with propidium iodide to identify the chromosomes. C, ideogram of the human G-banded chromosome 10, showing the localization of the ADAM12 gene at 10q26. ADAM 12-L cDNA clone and were able to assemble a partial contig covering an additional kilobase of 3Ј-untranslated ADAM 12-L DNA (not shown). Primers specific for this ESTderived region were designed and used to amplify a 467-bp product from placenta cDNA. This probe hybridized only to the 8.6-kb band on a placenta RNA Northern blot (data not shown). Thus the 8.6-kb band observed on Northern blots appears to encode the identical sequence to the ADAM 12-L cDNA but has a longer 3Ј-untranslated region.
No ADAM 12 transcripts could be detected by Northern blot examination of mRNA from human brain, lung, liver, kidney or pancreas (data not shown). Under the same hybridization conditions expression of the 8.6-and 5.4-kb ADAM 12-L-specific bands was detected in mRNA isolated from heart, prostate, uterus (no endometrium), colon (no mucosa), small intestine, bladder, stomach, and skeletal muscle, but at levels at least 15-fold lower than in placenta (Fig. 4B). The 3.5-kb ADAM 12-S band was not observed in these blots. The source of ADAM 12-L  3). B, multiple muscle tissue Northern blot (CLONTECH catalog number 7765-1) hybridized with ADAM 12-L cytoplasmic specific probe. C, Northern blot of RNA from the human tumor cells lines HU-1 lung adenocarcinoma, undifferentiated RD rhabdomyosarcoma, and A204 rhabdomyosarcoma hybridized to an ADAM 12 disintegrin domain probe. In all three panels, the migration of molecular size markers is indicated on the left, and the sizes of the various ADAM 12 transcripts detected by the probes is indicated on the right. mRNA in the uterus, colon, small intestine, bladder, stomach, and prostate may be the smooth muscle cells, a hypothesis that is supported by our preliminary immunohistochemical analysis showing that these cells exhibited a positive immunostaining reaction with anti-ADAM 12 antisera. 6 Northern blot analysis of several cultured human cell lines demonstrated that the RD rhabdomyosarcoma and the HU-1 lung adenocarcinoma cells lines expressed all three ADAM 12 transcripts, although the A204 cell line did not express any (Fig. 4C). The ADAM 12 mRNA in these carcinoma cell lines appears to be expressed at a lower level than that observed in placenta but at a higher level than observed in normal tissue.
Some indication of gene expression can be obtained from EST data bases. ESTs specific for human ADAM 12-L have been isolated from cDNA libraries prepared from HeLa cell s3, fullterm placenta, and 20-week post-conception fetal liver and spleen. ESTs specific for ADAM 12-S have been isolated from cDNA libraries prepared from 6-week embryo, 8 -9-week postconception, and full-term placenta, and 20-week post-conception fetal liver and spleen. This may be taken as evidence that ADAM 12-S is expressed in normal tissues other than placenta.
Biological Function of ADAM 12-To begin analyzing the distribution and function of the ADAM 12 protein, we generated poly-and monoclonal antibodies to the 17-kDa cysteinerich domain of ADAM 12 produced in E. coli (Fig. 5A). These antibodies immunostained and reacted in Western blotting with COS-7 cells transiently transfected with an ADAM 12-S expression plasmid but not with cells transfected with a control plasmid lacking an ADAM 12 insert (Fig. 5, B and C). We then made an expression construct carrying an ADAM 12-S minigene coding for the disintegrin domain, the cysteine-rich domain, and the carboxyl terminus of ADAM 12-S. The rationale for using a minigene was based on previous studies showing that a mouse ADAM 12 minigene lacking the pro-and metalloprotease domains was biologically active, whereas the fulllength form was not (25). This plasmid containing the minigene (p1095) or the vector lacking a cDNA insert as a negative control was transfected into the human rhabdomyosarcoma cells A204 that do not express detectable amounts of ADAM 12 mRNA or protein. Three stably transfected clones were obtained that expressed ADAM 12-S minigene mRNA and secreted a 42-kDa ADAM 12-S polypeptide into the medium (Fig.  5D). Like the parental A204 cells, these three cell lines showed no apparent capacity to fuse in vitro (data not shown).
The parental A204 cells and A204 cells transfected with either the ADAM 12-S expression plasmid (three clones) or a control plasmid (three clones) were injected into nude mice and allowed to form subcutaneous tumors. No gross difference in tumor growth capacity was observed. However, morphological analysis revealed a striking difference in the stromal compartment (Figs. 6 and 7 and Table II). The tumors derived from the parental cells and from three control transfected A204 cell lines consisted of densely packed tumor cells with an appearance consistent with embryonal rhabdomyosarcoma. In contrast, in tumors generated by three ADAM 12-S minigene transfected A204 cell lines, a striking, bizarre pattern of muscle cell differentiation was observed (Fig. 6A). Irregular stellate and elongated myocyte-and myotube-like cells were scattered randomly in the stroma. The nuclei of the myotubes were either centrally or peripherally located, and cross-striation was seen in some of 6  the cells. Thus, the morphological pattern of these ectopic muscle cells was clearly distinct from that of normal adult muscle. We did not observe ectopic muscle cell differentiation in 50 nude mouse tumors generated by human breast carcinoma MDA-MD-435 cells transfected with full-length integrin ␤4 or a truncated form of it. 6 Confirmation that these cells in fact represented the muscle cell lineage was provided by electron microscopic demonstration of intracellular structures characteristic of developing myofibers and a pericellular basement membrane like structure. A further characterization of these ectopic muscle cells was obtained by immunohistochemistry (Fig. 7). Positive immunostaining was found using a rat mAb 1B11 specific to mouse tetranectin (Fig. 7A), a recently identified marker for myogenesis in mouse development. 4 Likewise an antibody to neural cell adhesion molecule that does not cross-react to human exhibited positive immunostaining of the ectopic muscle cells. Positive immunoreaction was found with antibodies to myogenin (Fig. 7B) and MyoD transcription factors that are both markers of muscle development. Positive immunoreaction was also seen with antibodies to other muscle markers including desmin, caveolin-3 (Fig. 7C), ␤-dystroglycan, and adhalin. Polyclonal antiserum to murine laminin immunostained a pericellular basement membrane-like structure (Fig. 7D), whereas stainings with monoclonal antibodies specific to several different human laminin chains (␣2,␤1,␤2) were negative (Fig. 7E). A small fraction of the tumor cells stained positively for p53, although no immunostaining was found of the ectopic muscle cells (Fig. 7F). Finally, myf-5, a muscle transcription factor detectable during embryogenesis only (42), was detected via RT-PCR in tumors from ADAM 12-S minigene-transfected cells but not in tumors from the A204 control transfected cells, using mouse-specific primers (not shown). Together these immunostaining and RT-PCR data strongly indicate that what we observe in this model system is formation of ectopic muscle cells of murine origin. DISCUSSION We have cloned the human ADAM 12 full-length cDNAs and discovered an alternatively spliced form, designated ADAM 12-S. This form of ADAM 12 has no transmembrane and cytoplasmic domains but has instead a short 34-aa carboxyl terminus. The resulting polypeptide becomes secreted, and transfection experiments indicated that this form provokes myogenesis.
Mouse ADAM 12 was first identified by Yagami-Hiromasa et al. (25) and was called meltrin ␣. In the same study they also isolated, via RT-PCR, two partial sequences that they desig-nated meltrin ␤ and meltrin ␥. Sequence comparison with other ADAM genes isolated since indicates that meltrin ␤ is more likely to be the murine equivalent of Xenopus adam 13 (Refs. 14 and 43; not shown), whereas meltrin ␥ shows 100% identity to nt 1289 -1738 of mouse adam 9 (40). Therefore, we prefer to use the ADAM nomenclature and refer to meltrin ␣ as ADAM 12.
In the present study we demonstrated that human ADAM 12 has two alternatively spliced forms designated ADAM 12-L and ADAM 12-S. Multiple transcripts of a single ADAM have been observed in monkey, mouse, and human (18,44,45). Mouse ADAM 1␣I and ADAM 1␣II are identical from nucleotides 702-2492 and contain the same domain structure but have distinct although related 5Ј-and 3Ј-ends (44). Monkey ADAM 6 has two isoforms, although these probably reflect two different genes (45). The only previously confirmed alternative splicing is of the human ADAM 11 gene (18). This candidate tumor suppressor gene for breast cancer is alternatively spliced to generate two different transcripts, MDC-769 and MDC-524. These transcripts differ at both the 5Ј-and 3Ј-ends. MDC-769 has full-length cysteine-rich, transmembrane, and cytoplasmic domains, whereas MDC-524 is a truncated protein that terminates in the cysteine-rich domain.
Human ADAM 12 appears to be encoded by a single copy gene that we mapped to chromosome 10q26 by in situ hybridization. Data base searches revealed that an EST that matches the ADAM 12-L cDNA sequence was localized to the same region of chromosome 10 by radiation hybrid mapping. Within this region genes for acyl-CoA dehydrogenase, fibroblast growth factor receptor, uroporphyrinogen-III synthase, and ornithine aminotransferase have been identified. With the exception of the ornithine aminotransferase (deficiency of this enzyme causes the eye disease gyrate atrophy of the choroid and retina), no disease loci have been mapped to this region. The ADAM 8 gene also maps to 10q26.3 (46), raising the possibility that these two genes may be clustered. The other ADAMs (ADAMs 1, -2, -4, -5, and -11) that have been assigned locations on the human genome are dispersed throughout different chromosomes (47)(48)(49).
By Northern blotting we observed three human ADAM 12 mRNAs of 3.5, 5.4, and 8.6 kb, whereas in the mouse only one transcript was reported (25). The 3.5-kb band was specific for ADAM 12-S, and the 5.4-and 8.6-kb bands were specific for ADAM 12-L. The 3.5-kb transcript arises from alternative use of an exon that encodes an ADAM 12-S-specific carboxyl terminus, 3Ј-untranslated region, and polyadenylation site. The  5.4-and 8.6-kb transcripts appear to be derived by alternative use of polyadenylation sites in the ADAM 12-L-specific 3Јuntranslated region. We have isolated the full-length cDNA representing the 5.4-kb transcript, but the 8.6-kb transcript that appears to contain a longer 3Ј-untranslated region has not been entirely isolated. The different 3Ј-untranslated regions of ADAM 12-L could affect the rates of translation or mRNA stability (50).
Analysis of the distribution patterns of the two ADAM 12 forms in normal human tissues revealed that the ADAM 12-S transcript was detected so far only in placenta, whereas the ADAM 12-L mRNAs were found in placenta and skeletal, cardiac, and smooth muscle. Splicing of the ADAM 12-S exon may be regulated by cell type-specific factors. The human ADAM 12-L transcript appears to have a more widespread expression than the mouse mRNA (25). Interestingly, both forms were detected in some tumor cell lines, indicating a possible association between ADAM 12-S and neoplasia.
Until now only two ADAMs that lack a transmembrane domain, and thus are assumed to be secreted, have been reported. These are ADAM 11/MDC-524 (18) and ADAMTS-1 (ADAM with thromobospondin motifs (12)). To this list we now add ADAM 12-S, and we have shown that it becomes secreted. The candidate tumor suppressor MDC-524 splice form was isolated from human cerebellar cDNA library and is expressed at very low levels compared with the cell membrane anchored form MDC-769 (18). ADAMTS-1, which is associated with cancer cachexia and inflammatory processes, lacks the cysteinerich, transmembrane, and cytoplasmic domains, having instead a thrombospondin homologous domain and type I thrombospondin motifs (12). Like ADAM 12-S, these cancerrelated, secreted ADAMs appear to have very restricted, low levels of expression in normal tissue. The best characterized soluble ADAM-like proteins are the snake venom metalloproteases (SVMPs) (7). The soluble ADAMs lacking the regulatory control of a transmembrane domain may be extremely potent like their SVMP counterparts. The highly restricted and low level of normal expression of these soluble ADAMs may reflect this potency, and continued comparison of ADAMs with SVMPs is warranted.
Mouse ADAM 12 has been implicated in cell fusion during C2C12 differentiation in vitro (25). In the present study we found that cells transfected with the shorter, secreted form of ADAM 12-S appear to be very potent in provoking myogenesis in vivo. We stably transfected the human embryonal rhabdomyosarcoma cell line A204 with the ADAM 12-S minigene composed of the disintegrin and cysteine-rich domains and the unique carboxyl terminus. Although a potential muscle precursor, the A204 rhabdomyosarcoma does not differentiate in vitro either spontaneously or after transfection with the ADAM 12-S minigene. However, nude mice tumors generated from these ADAM 12-S minigene transfected cells contained a striking pattern of ectopic muscle cell formation as compared with control tumors. A mixture of cells representing different stages of normal myogenesis was observed, including myoblasts and elongated multinucleated myotubes with cross-striation. These developing muscle cells were located in a disorganized pattern, as opposed to the normal adult skin muscle. Electron microscopy and immunostaining confirmed that these cells were in fact of the muscle cell lineage. Furthermore, based on combined immunostaining using mAbs specific for mouse and human antigens and RT-PCR using species-specific primers for myf-5, we conclude that these ectopic muscle cells are of an apparent murine origin rather than derived directly from the A204 human tumor cells.
What is the cell of origin for this myogenesis? There are at least two possibilities; one is the satellite cells, which are normally located in intimate relationship with existing myofibers beneath the basement membrane. Satellite cells are ubiquitous in normal adult muscle and represent the muscle progenitor cells during muscle regeneration (24). Another possible source is undifferentiated mesenchymal progenitor cells present in connective tissue. The mechanism by which ADAM 12-S may be involved in the recruitment and differentiation of muscle progenitor cells is not clear, and how it acts in the context of factors from the A204 rhabdomyosarcoma cells and/or the host stroma remains to be resolved.
In conclusion, we have characterized a novel form of secreted human ADAM 12, designated ADAM 12-S, and presented evidence that it provokes myogenesis in a nude mouse tumor model.