Specific Interaction of the Recombinant Disintegrin-like Domain of MDC-15 (Metargidin, ADAM-15) with Integrin a v b 3*

MDC-15 (ADAM-15, metargidin), a membrane-an-chored metalloprotease/disintegrin/cysteine-rich protein, is expressed on the surface of a wide range of cells and has an RGD tripeptide in its disintegrin-like domain. MDC-15 is potentially involved in cell-cell interactions through its interaction with integrins. We expressed a recombinant MDC-15 disintegrin-like domain as a fusion protein with glutathione S -transferase (des-ignated D-15) in bacteria and examined its binding function to integrins using mammalian cells expressing different recombinant integrins. We found that D-15 specifically interacts with a v b 3 but not with the other integrins tested ( a 2 b 1, a 3 b 1, a 4 b 1, a 5 b 1, a 6 b 1, a 6 b 4, a v b 1, a IIb b 3, and a L b 2). Mutation of the tripeptide RGD to SGA totally blocked binding of D-15 to a v b 3, suggesting that D-15- a v b 3 interaction is RGD-dependent. When the sequence RPTRGD is mutated to NWKRGD, D-15 is recognized by both a IIb b 3 and a v b 3, suggesting that the receptor binding specificity is mediated by the sequence flanking the RGD tripeptide, as in snake venom disintegrins. These results indicate that the disintegrin-like domain of Stockholm, Sweden) were labeled with fluorescein isothiocyanate (FITC) (33). Cells were incubated with mouse IgG or mAb PT25–2 at 10 m g/ml for 30 min at 4 °C in Dulbecco’s modified Eagle’s medium. Then, FITC-labeled fibrinogen was added at a final concentration of 100 m g/ml, and the mixture was further incubated for 30 min at room temperature. After washing the cells once with PBS to remove unbound labeled protein, bound protein was quantified by flow cytometry in FACSCan (Beckton-Dickinson). Other Methods— Site-directed mutagenesis was carried out using the unique site elimination method (34). The presence of mutations was verified by DNA sequencing.

Metalloprotease/disintegrin/cysteine-rich proteins (MDCs, also called ADAMs) 1 are membrane-anchored proteins with several domains including a metalloprotease domain, a disintegrin-like domain, a cysteine-rich sequence, an epidermal growth factor-like sequence, a transmembrane domain, and a short cytoplasmic domain (1). The biological functions of MDCs are not clear; however, we do know that fertilins (MDC-1 and -2) (2) are involved in sperm-egg binding and fusion (3), meltrins (MDC-12) (4) are involved in myoblast fusion during muscle development, and KUZ (a Drosophila MDC protein) (5) assists in neurogenesis. The MDC cytoplasmic domain has a proline-rich potential SH3 binding motif, suggesting that MDCcounter receptor interaction may induce signal transduction.
Integrins are a family of cell adhesion receptors that bind to a variety of ligands, including extracellular matrix proteins and other cell surface molecules (6 -10). MDCs are potential ligands for integrins, since most snake venom disintegrins interact with integrins ␣IIb␤3 and ␣v␤3 (reviewed in Ref. 11 and references therein). However, little is known about the receptor specificity of MDCs, except that mouse egg integrin ␣6␤1 has been proposed as a receptor for fertilin (2). Evans et al. (12) recently expressed recombinant fertilin fragments in bacteria as fusion proteins with maltose-binding protein (12). The recombinant fertilin-␤ fragment has been shown to bind to the egg membrane to which sperm bind and to block sperm from binding to the egg. These results suggest that the disintegrinlike domains of MDCs may be properly folded in bacteria, that glycosylation of the disintegrin-like domain may not be required for interaction with receptors, and that a strategy using recombinant MDC proteins is a viable alternative to those using purified materials that are not easily available.
MDC-15 (metargidin) (13,14) is the only known MDC that has an RGD sequence in its disintegrin-like domain. MDC-15 is widely expressed in various tissues and cells, including human umbilical vein endothelial cells and smooth muscle cells (14). In this study, we determined the receptor specificities of MDC-15 using the recombinant disintegrin-like domain of MDC-15, which was expressed in bacteria. We determined the receptor specificity of the recombinant protein using mammalian cells expressing different recombinant human integrins. We discovered that the recombinant disintegrin-like domain specifically interacts with integrin ␣v␤3 but not with other RGD-dependent or independent integrins tested, including ␣IIb␤3. This is in contrast to snake venom disintegrins, most of which bind to both ␣IIb␤3 and ␣v␤3 (Ref. 11 and references therein). We also found that the integrin specificity of the recombinant disintegrin-like domain is mediated by the RGD tripeptide and its flanking sequence, as in snake venom disintegrins. These results suggest that MDC-15 may be involved in ␣v␤3-mediated cell-cell adhesion.  (13) was amplified by polymerase chain reaction with a human erythroleukemia (K562) cell cDNA library as a template using 5Ј-CGGGATCCATGGCTGCTTTC-TGCGG and 5Ј-CGGGATCCTTACTCGCCATCCCCTAGGCTG as primers. The cDNA fragment was subcloned into the BamHI site of a pGEX-2T vector (Amersham Pharmacia Biotech). Synthesis of the GST fusion protein of the disintegrin-like domain of MDC-15 was induced in Escherichia coli DH5␣ by adding 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside in culture medium as described previously (20). Protein was extracted from the bacterial suspension by sonication and purified using glutathioneagarose (Sigma) affinity chromatography.

Monoclonal Antibodies and
A cDNA fragment of about 1100 nucleotides that encodes the 8 -11th type III repeats of rat fibronectin (Ala-1356 to Thr-1720) was amplified by polymerase chain reaction with rat fibronectin cDNA (provided by J. Schwarzbauer, Princeton University, NJ) as a template using 5Ј-CGG-GATCCGCCGTCCCTCCTCCCACG-3Ј and 5Ј-CGGGATCCTTAGGT-CACTGCAGTCTGAAC-3Ј as primers. The cDNA fragment was subcloned into the BamHI site of a pGEX-2T vector. We expressed the GST fusion protein of rat fibronectin (designated GST-FN) in bacteria and purified it as described above. The GST-FN preparation has a major band with a M r of about 65,000 (approximately 80% of the total) and some minor protein bands (degradation products) (data not shown), which is consistent with the M r of 65,773 calculated from the primary structure of GST-FN.
Absorbance at 280 nm was measured to determine the concentration of purified proteins using A 280 ϭ 1.356 for D-15, A 280 ϭ 1.281 for GST-FN, and A 280 ϭ 1.567 for wild-type (wt) GST. The extinction coefficient for each protein was calculated from the amino acid sequence by counting the number of Tyr, Trp, and Cys residues and using the following values for molar extinction. For Tyr, ⑀ 280 ϭ 1400; for Trp, ⑀ 280 ϭ 5600; and for Cys, ⑀ 280 ϭ 127 for each disulfide bond (2 Cys residues) (21,22).
Adhesion Assays-Wells of 96-well Immulon-2 microtiter plates (Dynatech Laboratories, Chantilly, VA) were coated with 100 l of PBS (10 mM phosphate buffer, 0.15 M NaCl, pH 7.4) containing substrates at a concentration of 20 g/ml and were incubated overnight at 4°C. The remaining protein binding sites were blocked by incubating with 1% bovine serum albumin (Calbiochem) for 1 h at room temperature. Cells (10 5 cells/well) in 100 l of Dulbecco's modified Eagle's medium were added to the wells and incubated at 37°C for 1 h. After gently rinsing the wells three times with PBS to remove unbound cells, bound cells were quantified by measuring endogenous phosphatase activity (31).
Affinity  (33). Cells were incubated with mouse IgG or mAb PT25-2 at 10 g/ml for 30 min at 4°C in Dulbecco's modified Eagle's medium. Then, FITC-labeled fibrinogen was added at a final concentration of 100 g/ml, and the mixture was further incubated for 30 min at room temperature. After washing the cells once with PBS to remove unbound labeled protein, bound protein was quantified by flow cytometry in FACSCan (Beckton-Dickinson).
Other Methods-Site-directed mutagenesis was carried out using the unique site elimination method (34). The presence of mutations was verified by DNA sequencing.

The Recombinant Disintegrin-like Domain of MDC-15-To
prove that MDC-15 mediates cell-cell adhesion through interaction with integrins, we expressed the disintegrin-like domain of MDC-15 (Met-420 to Glu-510), a putative integrin binding site (Fig. 1A), as a fusion protein with GST in bacteria (designated D-15). We obtained soluble D-15 and purified it using affinity chromatography on glutathione-agarose. Fig. 1B shows that the purified D-15 migrates as a monomer with a M r of 36,000 under nonreducing conditions (lane 1), and the control, wt GST protein, migrates as a monomer with a M r of approximately 26,000. The sizes of these proteins match the values calculated from the primary structures of these proteins (35,930 and 26,968, respectively). Although the D-15 preparation contained some degradation products, we used it for adhesion and binding assays without further purification.
Adhesion to D-15 of CHO Cells Expressing Different Recom- binant Integrins-We determined whether D-15 supports integrin-mediated cell adhesion using CHO cells expressing different recombinant integrins. Parent CHO cells express ␣5␤1 as a major integrin but do not express ␤2 or ␤3 integrins (29,35). As shown in Fig. 2A, wt GST (the negative control) does not support adhesion to any of the cells used, but GST-FN (the positive control), which contains the central cell binding domain of rat fibronectin (the 8 -11th type III repeats), supported all of the cell lines used. D-15 supported adhesion of ␤3-CHO cells (that express ␣v␤3) and ␣IIb␤3-CHO cells (that express both ␣IIb␤3 and ␣v␤3) but not parent CHO cells and cells expressing other exogenous integrins (including ␣2␤1, ␣3␤1, ␣4␤1, ␣5␤1, ␣6␤1, ␣6␤4, ␣v␤1, and ␣L␤2). These results indicate that D-15 interacts either with ␣v␤3 or with both ␣IIb␤3 and ␣v␤3. Fig. 2B shows that both D-15 and GST-FN support maximum adhesion of ␤3-CHO cells at the coating concentration of 10 g of protein/ml, suggesting that D-15 and GST-FN support adhesion of ␤3-CHO cells at comparable levels. Fig. 2C shows that adhesion to D-15 of ␤3-CHO cells is completely blocked by function-blocking anti-␣v␤3 mAb LM609, but adhesion to GST-FN is not. These results support the idea that adhesion to D-15 is mediated exclusively by ␣v␤3, whereas adhesion to GST-FN is mediated by multiple receptors that include ␣v␤3, ␣v␤1, and ␣5␤1.
Affinity Chromatography of Solubilized ␣IIb␤3 and ␣v␤3 Integrins on D-15 Immobilized to Agarose-To determine whether D-15 interacts with ␣v␤3 or both ␣IIb␤3 and ␣v␤3, we carried out affinity chromatography on immobilized D-15. Purified D-15 was coupled to glutathione-agarose and incubated with lysates of 125 I-labeled ␣v␤3, ␣IIb␤3, or parent CHO cells. Bound materials were eluted with 20 mM EDTA. Fig. 3 shows that two protein bands corresponding in size to ␣v/␣IIb and ␤3 were eluted with lysates from ␣v␤3and ␣IIb␤3-CHO cells, but very little radioactivity was eluted with the lysate of parent CHO cells. We analyzed the eluted materials by immunoprecipitation using antibodies specific to ␣v, ␣IIb, or ␤3. ␣IIb␤3 was detected only in the unbound fraction with ␣IIb␤3-CHO cells. ␣v␤3 was detected in both the bound and unbound fractions with ␣v␤3-CHO and ␣IIb␤3-CHO cells. These re-

FIG. 2. Adhesion of cells expressing different integrins to the recombinant disintegrin-like domain of MDC-15.
A, wells of 96well microtiter plates were coated with 20 g/ml (in PBS) D-15 (black column), GST-FN (white column), or wt GST (shaded column). Cells homogeneously expressing different human integrins were incubated in wells at 37°C for 1 h. After rinsing the wells to remove unbound cells, bound cells were quantified using endogenous phosphatase activity. GST-FN and purified FN (human or bovine) gave almost identical results. B, adhesion of ␤3-CHO cells to D-15 and GST-FN was determined as a function of the substrate. The data suggest that the D-15 protein used in the above experiment (20 g/ml in PBS) is a saturating concentration and that D-15 is comparable to GST-FN in supporting cell adhesion. C, the effect of anti-␣v␤3 mAb LM609 on adhesion of ␤3-CHO cells to D-15 and GST-FN was determined. Adhesion assays were performed as described above (A). A control mAb KH72 (to integrin ␣5) did not block adhesion of ␤3-CHO cells to D-15 (data not shown). LM609 and KH72 were used at X250 dilution of ascites.

␣v␤3-D-15 Interaction Is Dependent on the Tripeptide RGD and the Flanking Sequence in the Putative Integrin Binding
Site-To determine whether ␣v␤3-D-15 interaction is RGD-dependent, we mutated the D-15 RGD sequence to SGA. The mutant protein (designated D-15/SGA) was expressed as a soluble monomer (Fig. 1B), purified using affinity chromatography, and used for adhesion assays. The mutation completely blocked adhesion of ␤3-CHO cells to the fusion protein (Fig. 4). Increasing the coating concentration of the D-15/SGA mutant did not reverse the effects of the mutation. These results indicate that D-15-␣v␤3 interaction is dependent on the RGD tripeptide in the putative integrin binding site.
It has been reported that receptor specificity of snake venom disintegrins is defined by the sequence flanking the RGD tripeptide (36 -38). To determine whether receptor specificity is determined by the sequence flanking the RGD tripeptide in D-15, we replaced the sequence RPTRGD with NWKRGD, which has been reported to support high affinity binding to ␣IIb␤3 in a phage display system (39). The mutant, designated D-15/NWK, was also expressed as a soluble monomer in bacteria (Fig. 1B), purified by affinity chromatography, and used for adhesion assays. Fig. 5A shows that D-15/NWK supports adhesion of both ␤3-CHO and ␣IIb␤3-CHO cells, suggesting that the mutant interacts with ␣v␤3; however, it is not clear whether the mutant binds to ␣IIb␤3 (␤3-CHO cells express ␣v␤3, and ␣IIb␤3-CHO cells express both ␣v␤3 and ␣IIb␤3). To clarify this point, we examined binding of FITC-labeled wt D-15 and D-15/NWK mutant to cells expressing recombinant ␣IIb␤3. Since CHO cells express an inactive form of ␣IIb␤3, we activated the integrin using the anti-␣IIb␤3 mAb PT25-2 (30). As shown in Fig. 5B, D-15/NWK and fibrinogen (a positive control) bound to ␣IIb␤3-CHO, but wt D-15 did not. These results indicate that the specificity of the recombinant disintegrin-like domain of MDC-15 is dependent on the sequence flanking the RGD tripeptide in the putative integrin binding site and confirm that wt D-15 binds to ␣v␤3 but not to ␣IIb␤3.
Interaction between a Natural Human ␣v␤3 Heterodimer and D-15-Since we used a recombinant hamster ␣v/human ␤3 hybrid integrin on CHO cells to study the receptor specificity of D-15, we determined whether the natural human ␣v␤3 heterodimer recognize wt and mutant D-15. As shown in Fig. 6

DISCUSSION
Although MDC-15 and other members of its family have been proposed to represent a new class of cell adhesion molecules based on deduced amino acid sequences, their function is unclear in most cases (see the Introduction). In this paper we have evaluated the receptor specificity of MDC-15 by studying its recombinant disintegrin-like domain. Although the recombinant fertilin-␤ (MDC-2) fragment shows biological functions (12), it is possible that the native disintegrin-like domains of MDCs purified from mammalian cells or of recombinant MDCs expressed in eukaryotic systems may fold or be modified differently and show additional or different properties. In this study, we have shown that the recombinant disintegrin-like domain of MDC-15 specifically interacts with integrin ␣v␤3 using adhesion assays and affinity chromatography. We also present evidence that the receptor recognition and specificity of the recombinant disintegrin-like domain of MDC-15 is determined by the RGD tripeptide and flanking sequence. Substitution of the RGD tripeptide to SGA completely blocks binding of D-15 to ␣v␤3. Substitution of the sequence flanking the tripeptide RGD changes the receptor specificity (from ␣v␤3 to both ␣IIb␤3 and ␣v␤3). Thus, the binding specificity of the recombinant disintegrin-like domain of MDC-15, like that of snake venom disintegrins and dendroaspin, an RGD-containing neurotoxin variant (38), is mediated by the RGD tripeptide and flanking sequence. NMR structural studies of the snake venom disintegrins kistrin and echistatin show that, in both cases, the region containing the RGD sequence is a loop structure exposed to the outside of the disintegrin molecule (40,41). This peculiar structure probably supports the high affinity interaction of snake venom disintegrins with integrins (36,37). It would be interesting to determine whether the structure of the recombinant disintegrin-like domain of MDC-15 is similar to those of snake venom disintegrins.
It should be noted that the disintegrin-like domain of MDC-15 is unique in its receptor specificity. Most snake venom disintegrins recognize both ␣v␤3 and ␣IIb␤3 (11). Barbourin is the only disintegrin that is specific to ␣IIb␤3. There is no known snake venom disintegrin that is specific to ␣v␤3. Thus, the RGD tripeptide and flanking sequence in the disintegrinlike domain of MDC-15 represent a unique ␣v␤3-specific recognition sequence. The disintegrin-like domains of MDCs have additional Cys residues in the middle of the loops in their putative integrin binding sites (RGDC in MDC-15, for example). It has not been determined whether this Cys residue makes a disulfide link with another Cys residue. The topology of this region of the disintegrin-like domain is probably very different from that of the snake venom class P II disintegrins in that its loop structure is probably less flexible and its conformation more restricted, with an increase in restriction if the Cys residues flanking the RGD tripeptide are involved in a disulfide linkage (in the snake venom class P II disintegrins, the RGD sequence is positioned within an extended, flexible loop structure where there is only limited conformational restriction of the RGD sequence; see Ref. 42 for review). It is possible, therefore, that using short synthetic peptides (cyclic or linear) derived from the RGD and flanking sequences of the disintegrin-like domain of MDC-15 might provide different receptor specificities than those obtained in this study using recombinant or purified disintegrin-like domains.
The interaction of the disintegrin-like domain of MDC-15 with integrin ␣v␤3 may be related to its biological functions. MDC-15 is not expressed in vivo in normal vessels but is up-regulated in lesions of atherosclerosis, where many macrophages are present (14). MDC-15 on cultured endothelial cells undergoes proteolytic processing (14), which appears to be associated with MDC activation (4,43). It is possible that activated MDC-15 on endothelial cells interacts with ␣v␤3 on leukocytes during atherogenesis through its exposed disintegrinlike domain. ␣v␤3 has been shown to be involved in the progression of melanoma and the induction of neo-vascularization by tumor cells. ␣v␤3 antagonists promote tumor regression by inducing apoptosis of angiogenic blood vessels (44,45). It is possible that activated MDC-15 and ␣v␤3 on endothelial cells interact with each other, leading to homotypic aggregation of endothelial cells during angiogenesis. Based on the wide distribution of MDC-15, MDC-15-␣v␤3 interaction may mediate cell-cell interactions in many other instances (e.g. metastasis). The snake venom metalloprotease/disintegrin jararhagin is known to block collagen-induced aggregation of platelets (46). It has been proposed that the inhibition of platelet response to collagen by jararhagin is mediated through the binding of jararhagin to the platelet ␣2-subunit via the disintegrin domain followed by proteolysis of the ␤1 subunit with loss of the integrin structure (conformation) necessary for the binding of macromolecular ligands (47). It has been hypothesized that a fragment of MDC-15 containing the metalloprotease and disintegrin-like domains is released from cultured endothelial cells (14). It is possible that the proteolytic fragment of MDC-15 containing the metalloprotease and disintegrin-like domains interacts with ␣v␤3, as in the case of jararhagin and ␣2␤1, and either modifies the function of the integrin or promotes degradation of the matrix proteins surrounding the cells that express ␣v␤3. Therefore, the proposed ␣v␤3-MDC-15 interaction may be of wide biological importance.