Metargidin, a membrane-anchored metalloprotease-disintegrin protein with an RGD integrin binding sequence.

Cellular disintegrins are a family of membrane-anchored proteins with structural homology to snake venom metalloproteases and disintegrins. We report here the cDNA cloning and initial biochemical characterization of the first cellular disintegrin protein with an RGD sequence in its disintegrin domain, which we propose to name metargidin (for metalloprotease-RGD-disintegrin protein). The domain organization of metargidin is identical with that of previously reported members of the cellular disintegrin family, comprising (i) a pro- and a metalloprotease domain including a zinc-binding consensus motif, (ii) a disintegrin domain containing the RGD motif, (iii) a cysteine-rich domain, (iv) an epidermal growth factor-like domain, (v) a hydrophobic transmembrane domain, and (vi) a cytoplasmic tail with proline-rich sequences that could act as potential SH3 ligands. Antibodies raised against the cytoplasmic tail of metargidin recognize a glycoprotein of 110 kDa in MDA-MB-468 mammary epithelial carcinoma cells, which can be cell surface-biotinylated, indicating its localization in the plasma membrane. A second protein of 56 kDa co-immunoprecipitates with metargidin, suggesting that it is part of a protein complex. These features are consistent with a model in which metargidin is an integrin ligand which, as a transmembrane protein, might function in cell-cell adhesion and/or signaling.

short soluble protein components of Viperidae snake venoms that efficiently interfere with platelet aggregation by binding to the platelet integrin ␣ IIb ␤ 3 via an RGD sequence (1)(2)(3). In recent years, a family of membrane-anchored proteins containing a metalloprotease, a disintegrin, a cysteine-rich, and an epidermal growth factor-like domain followed by a transmembrane domain and a cytoplasmic tail has been discovered (4 -8). The best-characterized cellular disintegrins are the ␣ and ␤ subunit of the heterodimeric sperm-surface protein fertilin (formerly termed PH-30) (4,9). Fertilin has been implicated in gamete membrane binding and fusion (10,11), likely involving binding to the oocyte integrin ␣ 6 ␤ 1 in mouse (12). A recent study has demonstrated a potential role of another member of this protein family, meltrin ␣, in myoblast fusion (13). Based on the properties of snake venom disintegrins, as well as of fertilin and meltrin ␣, it is likely that other members of the membraneanchored cellular disintegrin protein family might also play a role in cell-cell interactions. We report here the identification of the first cellular disintegrin protein to display an RGDC motif in its putative integrin-binding site and discuss the possible implications of this finding.

MATERIALS AND METHODS
Animals and Reagents-Reagents, including Taq polymerase, were obtained from Boehringer Mannheim, unless indicated otherwise. Rabbits were purchased from Hazelton (Denver, PA). Radiolabeled nucleotides were supplied by Amersham. , and 2 min at 72°C, followed by an extension step of 10 min at 72°C). 30 subcloned PCR cDNA products (pCR-Script vector, Stratagene) were sequenced (Sequenase, U. S. Biochemical Corp.) and revealed cDNAs encoding (i) a novel disintegrin-coding protein, (ii) the MDC protein (14), and (iii) the MDC9 protein (15). A cDNA library was constructed from MDA-MB-468 mRNA in the Lambda ZAP II bacteriophage vector (Stratagene), as described earlier (16), and screened using the novel cDNA tag, labeled with [␣ 32 P]dCTP, as a probe. 130 primary candidate bacteriophage clones were analyzed for their insert sizes by PCR using the T3 primer and one 3Ј primer derived from the cDNA-tag sequence. Candidate clones showing the longest cDNA inserts were plaque-purified and subjected to in vivo excision. One fulllength clone thus identified was completely sequenced on both strands using internal primers. Sequence assembly and analysis were done using the AssemblyLign and MacVector programs (Kodak Scientific Imaging Systems, New Haven, CT).
Northern Blot Analysis-Human multiple tissue Northern blots I and II (Clontech), containing approximately 2 g of poly(A) RNA per lane, were probed with a [␣-32 P]dCTP-labeled metargidin cDNA probe under high stringency conditions using QuickHyb solution (Stratagene).
Production of Metargidin GST-Fusion Proteins-GST-fusion proteins with the 103 C-terminal amino acid residues within the cytoplasmic tail of human metargidin were generated by PCR using genespecific primers with added restriction sites for ligation into pGEX GST-fusion vectors (Pharmacia Biotech, Uppsala, Sweden). Constructs were sequenced for possible PCR mutations and then transfected into BL26 cells (Novagen, Madison, WI). Bacterial cells expressing the metargidin-cyto-GST-fusion protein were disrupted by French press * This work was supported in part by Cancer Center Support Grant NCI-P31-CA-08748. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Antibodies-Metargidin-cyto-GST fusion proteins in phosphatebuffered saline were mixed with Ribi adjuvant (RIBI Immunochemical Research Inc.) and used for immunizations according to established protocols (17). The antiserum raised against the metargidin-cyto-GST fusion protein was depleted of antibodies reacting with GST by incubation with GST coupled to CNBr-activated CL4B beads (Pharmacia). These antibodies are referred to as anti-metargidin IgG. As a negative control, an identical metargidin-cyto-GST serum sample was incubated with the metargidin-cyto-GST fusion protein coupled to CNBr-activated CL4B beads to remove all metargidin-cytotail specific antibodies. These antibodies are referred to as control IgG. Depletion of GST-specific, and of metargidin-cyto-GST-specific, antibodies was confirmed by Western blot analysis (not shown).
Cell Lysis, Enzymatic Deglycosylation, and Western Blot Analysis-MDA-MB-468 cells (American Type Culture Collection, ATTC) were grown to confluency in 10-cm tissue culture plates, washed in Trisbuffered saline, pH 7.4 (TBS), and lysed in 0.5% Nonidet P-40 in TBS, 2 mM Mg 2ϩ , 2 mM Ca 2ϩ , and protease inhibitors (9). Extracts were incubated with the lectin concanavalin A (Pharmacia), eluted in sample loading buffer, and reduced with 10 mM DTT as indicated. For Western blot analysis, eluted glycoproteins were separated by SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose, and processed as described (9,18). For removal of N-linked carbohydrate residues with endoglycosidase H f (New England BioLabs, Cambridge, MA), concanavalin A-enriched MDA-MB-468 extracts in sample loading buffer, 1% ␤-mercaptoethanol, were adjusted to pH 5.5 with sodium citrate and incubated at 37°C overnight with 2,000 NEB units of Endo H f .
Cell Surface Biotinylation and Immunoprecipitation-A confluent 10-cm dish of MDA-468 cells was washed with TBS and incubated on ice with a 1 mM solution of the non-membrane-permeable biotinylation reagent NHS-LC-biotin (Pierce) in TBS for 45 min. After washing and blocking with 50 mM glycine in TBS, cells were lysed as described above. Extracts were immunoprecipitated with anti-metargidin IgG, control IgG, or preimmune IgG, subjected to SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose, and probed with horseradish peroxidase-labeled streptavidin (Pierce, Rockford, IL). Bound horseradish peroxidase streptavidin was detected using a chemiluminescent detection system (ECL, Amersham).

RESULTS AND DISCUSSION
In order to identify novel cellular disintegrins, cDNA tags were generated by PCR on reverse-transcribed cDNA from the human mammary epithelial cell line MDA-MB-468 using degenerate primers. With this approach, a cDNA tag of a disintegrin protein harboring an RGD sequence was isolated. To isolate the full-length cDNA, we constructed a cDNA library from MDA-MB-468 cells, which was screened with the labeled cDNA tag. Sequencing of the clone with the longest insert thus isolated revealed a cDNA of 2740 base pairs, with an open reading frame encoding an 814-amino acid protein (Fig. 1A). This protein was named metargidin (metalloprotease-RGD-disintegrin; for reference purposes the metargidin cDNA was named ADAM 15, see also Ref. 8). Northern blot analysis revealed the metargidin mRNA to be close to 3 kilobases in length and to be present in all human tissues examined (Fig. 2).
Excluding the putative signal peptide (Fig. 1A), the mature metargidin protein has a predicted molecular mass of 85 kDa, and its extracellular moiety contains 5 potential sites of Nlinked glycosylation. The deduced metargidin protein sequence includes all the domains previously found in cellular disintegrin proteins (see above and Fig. 1A). The metalloprotease domain of metargidin has the HEXXHXLGXXHD zinc-binding motif, which suggests metalloprotease activity (4, 8, 19, 20). The putative boundary between the prodomain and the metalloprotease domain has four consecutive arginine residues, cre-ating a potential cleavage site for serine proteases such as the proprotein convertase furin (21)(22)(23). As is the case for other cellular disintegrin-metalloprotease domains that contain a zinc-binding consensus sequence, the prodomain contains an odd number of cysteine residues, one of which may regulate the activity of the metalloprotease through a cysteine switch mechanism (24).

Cellular Disintegrin Containing an RGD Integrin Binding Sequence 4594
RGD integrin ligand consensus motif in a position analogous to that found in snake venom disintegrins (see alignment in Fig.  1B). The RGD sequence is followed by an additional cysteine residue which so far has only been found in non-RGD-type snake venom disintegrins and cellular disintegrins, but not in RGD-type snake venom disintegrins. The structure of several snake venom disintegrins has been solved by NMR, revealing that the RGD sequence is present in a flexible hairpin loop (25,26). An RGD-containing loop structure has also been found in the otherwise unrelated Elapidae snake venom toxin dendroaspin (27,28), in the leech anticoagulant decorsin (29), and in the fibronectin type III repeat (30,31), and thus appears to be a preferred context for RGD motifs showing high affinity binding to an integrin. It is conceivable that the additional cysteine residue in the metargidin disintegrin domain may change its structure such that the RGDC sequence becomes cryptic. However, if the RGDC sequence is accessible to integrin binding, the cysteine residue may contribute to the specificity for binding to a certain integrin, since the amino acid residues following the RGD sequence are known to confer some specificity for different integrins (32)(33)(34). Data from phage display studies have shown that RGD sequences in the context of disulfide-bonded cysteine residues are high affinity integrin ligands (34 -36), and an RGDC sequence in a cyclic peptide with eight amino acid residues (CX 6 C) has high affinity for the platelet integrin ␣ IIb ␤ 3 (35). If metargidin is indeed capable of interacting with integrins through its RGD sequence, it would represent a new type of membrane-anchored integrin ligand containing an RGD sequence.
Two cellular disintegrin proteins, fertilin ␣ (4) and meltrin ␣ (13), are thought to be involved in membrane fusion. In both proteins, the cysteine-rich domain displays a short hydrophobic stretch that has been suggested to function as a fusion peptide (4,13). Comparison of a hydrophobicity plot of metargidin with fertilin ␣ and meltrin ␣ shows that metargidin lacks a similarly hydrophobic region in the cysteine-rich domain (data not shown). Metargidin is therefore not likely to play a direct role in membrane fusion.
Two short proline-rich sequences present in the cytoplasmic domain of metargidin (PPPPRKP and RPAPPPP, see Fig. 1A) show similarity to Src homology 3 (SH3) ligand domains (37)(38)(39). If active as SH3 ligands, these motifs in metargidin could mediate intracellular signaling or cytoskeletal attachment via proteins containing SH3 domains (40,41). Similar proline-rich sequences are also found in the cytoplasmic tail of some other cellular disintegrins, such as MDC9 (15), MS2 (42), and meltrin ␣ (13), and the putative proline-rich cytoplasmic SH3 ligand domains of MDC9 have been shown to interact with the SH3 domain of Src in a blot overlay assay (15).
To analyze the domain organization and potential processing of metargidin in MDA-MB-468 cells, antibodies were raised against a fusion protein between the metargidin cytoplasmic tail domain and GST. For use in Western blot analysis and immunoprecipitation, purified IgG were depleted of antibodies reactive to GST (referred to as anti-metargidin IgG), or depleted of GST-cytoplasmic tail-specific antibodies (referred to as control IgG, see "Materials and Methods"). MDA-MB-468 cell extracts enriched for glycoproteins through precipitation on concanavalin A were analyzed by immunoblotting with antimetargidin IgG. A band of 100 kDa was recognized under nonreducing conditions (Fig. 3A, lane 1), and a band of 110 kDa under reducing conditions (Fig. 3A, lane 2). The increase in apparent molecular mass upon reduction is consistent with the behavior of a cysteine-rich cellular disintegrin protein. When identical samples were probed with control IgG, three weaker nonspecific bands of 63, 73, and 90 kDa were visible under nonreducing (Fig. 3A, lane 5) or reducing conditions (Fig. 3A,  lane 6). After removal of N-linked carbohydrate residues with Endo H f (Fig. 3A, lane 3), reduced metargidin has an apparent molecular mass of 95 kDa, whereas mock Endo H f -treated metargidin has an apparent molecular mass of 110 kDa (Fig.  3A, lane 4). The apparent molecular mass of 95 kDa for Endo H f -treated metargidin is close to its predicted molecular mass of 85 kDa.
The size of metargidin on MDA-468-cells suggests that both the disintegrin and metalloprotease domains occur in a membrane-anchored form. The presence of both an integrin ligand domain and a metalloprotease domain in the metargidin protein suggests the possibility that both domains are active simultaneously. Alternatively, the RGD sequence may only become exposed after proteolytic processing, and the metalloprotease and disintegrin domains may thus function sequentially. The sperm protein fertilin ␤ is an example of a membrane-anchored cellular disintegrin that is processed at the N terminus of the disintegrin domain during sperm maturation, and processing correlates with the acquisition of fertilization competence (4,9). Furthermore, expression of a variant of the mouse cellular disintegrin meltrin ␣ lacking its metallo- Precipitated proteins were reduced with 10 mM DTT prior to electrophoresis on 7.5% SDS-poly-acrylamide gels, transferred to nitrocellulose and probed with horseradish peroxidase-coupled streptavidin.

Cellular Disintegrin Containing an RGD Integrin Binding Sequence 4595
protease domain in C2C12 myoblasts resulted in increased myoblast fusion, whereas expression of the full-length meltrin ␣ led to decreased fusion (13), indicating that certain cellular disintegrins may need to be processed in order to become active. Experiments are currently in progress to determine whether the RGD sequence within the metargidin-disintegrin domain indeed functions as an integrin ligand, and, if so, whether the prodomain or the metalloprotease domain needs to be removed for the RGD to become available for binding.
To determine whether metargidin, like fertilin, is also part of a protein complex on the cell surface, MDA-MB-468 cells were labeled with the water-soluble biotinylation reagent NHS-LCbiotin. Labeled extracts were immunoprecipitated with antimetargidin IgG or with control IgG. Anti-metargidin IgG immunoprecipitated a protein with an apparent molecular mass of 110 kDa under reducing conditions (Fig. 3B, lane 1) and 100 kDa under nonreducing conditions (not shown). This protein was not immunoprecipitated by control IgG (Fig. 3B, lane 2), or by preimmune IgG (Fig. 3B, lane 3). These results indicate that metargidin is present on the plasma membrane of MDA-MB-468 cells. In addition to the band of 110 kDa, a second band of 56 kDa was also immunoprecipitated by anti-metargidin IgG, but not by control antibodies. The 56-kDa protein also underwent a shift in apparent molecular mass upon reduction (not shown), indicating the existence of extracellular disulfide bonds. In a separate Western blot analysis using anti-metargidin IgG, the 56-kDa protein was not recognized (not shown), suggesting that it is distinct from metargidin. Since fertilin is a heterodimer consisting of two cellular disintegrin proteins, it will be interesting to determine whether the 56-kDa protein co-precipitating with metargidin is also a cellular disintegrin.
In conclusion, the presence of the RGD sequence in metargidin strongly suggests that metargidin could act as a membrane-anchored integrin ligand containing the RGD motif. This finding raises the possibility that, in addition to their role in cell matrix interactions, RGD-binding integrins might also mediate direct cell-cell interactions. Since metargidin contains potential cytoplasmic SH3 ligand domains, it is possible that it not only functions as a membrane-anchored integrin ligand, but also as a counter-receptor for integrins or other proteins.