Inhibitory Effects of MLDG-containing Heterodimeric Disintegrins Reveal Distinct Structural Requirements for Interaction of the Integrin α9β1 with VCAM-1, Tenascin-C, and Osteopontin*

The integrin α9β1 is expressed on epithelial cells, smooth muscle cells, skeletal muscle, and neutrophils and recognizes at least three distinct ligands: vascular cell adhesion molecule 1 (VCAM-1), tenascin-C, and osteopontin. The α9 subunit is structurally similar to the integrin α4 subunit, and α9β1 and α4β1 both recognize VCAM-1 as a ligand. We therefore examined whether the disintegrin EC3, which we have recently shown specifically inhibits the binding of α4 integrins to ligands, would also be a functional inhibitor of α9β1. EC3 and a novel heterodimeric disintegrin that we identified, EC6, both were potent inhibitors of α9β1-mediated adhesion to VCAM-1 and of neutrophil migration across tumor necrosis factor-activated endothelial cells. A peptide containing a novel MLDG motif shared by both of these disintegrins also inhibited α9β1- and α4β1-mediated adhesion to VCAM-1. Surprisingly though, concentrations of EC3 that completely inhibited adhesion of α9-transfected cells to VCAM-1 had little or no effect on adhesion to either of the other α9β1 ligands, osteopontin and tenascin-C. Furthermore, peptides AEIDGIEL and SVVYGLR, which we have previously shown inhibit binding of α9β1-expressing cells to tenascin-C and osteopontin, respectively, had no effect on adhesion to VCAM-1. These data suggest that there are structurally distinct requirements for interactions of the α9β1 integrin with VCAM-1 and the extracellular matrix ligands osteopontin and tenascin-C.

The integrin ␣9 subunit forms a single known heterodimer, ␣9␤1, that is widely expressed in epithelia and smooth and skeletal muscle and on neutrophils (1,2). We and others have identified three distinct ligands for ␣9␤1: the extracellular matrix proteins tenascin-C (3) and osteopontin (4 -6) and the inducible endothelial immunoglobulin family member, VCAM-1 1 (1). We have mapped the ligand-binding site in tenascin-C to an exposed peptide loop in the third fibronectin type III repeat containing the sequence AEIDGIEL (7). We have also mapped the ␣9␤1 ligand-binding site in osteopontin (6). Although an initial report suggested that ␣9␤1 might bind to an RGD-containing sequence in osteopontin (5), we were able to demonstrate by extensive mutagenesis that the binding site is within the linear peptide sequence SVVYGLR immediately adjacent to the RGD site (6).
Structurally, the ␣9 subunit is closely related to the ␣4 subunit, and on the basis of sequence homology ␣4 and ␣9 appear to be the only known members of a subfamily of integrin ␣ subunits that lack both an insertional domain and an extracellular disulfide-linked cleavage site (2). Furthermore, both subunits are expressed on leukocytes and mediate leukocyte migration (1,8). Finally, both integrins recognize VCAM-1 as a ligand (1,9). We have shown that both ␣4␤1 and ␣9␤1 contribute to the chemotactic migration of human neutrophils across endothelial cell monolayers activated by tumor necrosis factor-␣ (TNF␣), an effect that is due at least in part to interaction of these integrins with VCAM-1 induced in response to TNF␣ (1).
Disintegrins are a family of low molecular weight, cysteinerich, anti-adhesive proteins that are present in the venoms of various vipers and selectively block the function of integrins (10,11). Early studies focused on monomeric disintegrins that express an RGD motif within a 13-amino acid putative hairpin loop that is maintained in an appropriate conformation by a disulfide bridge. Not surprisingly, these disintegrins selectively inhibit integrins that bind to ligands through RGD sites, such as ␤3 integrins (for eristostatin, kistrin, and bitistatin) or ␤3 integrins and the fibronectin receptor ␣5␤1 (for echistatin and flavoridin). Recently, however, we isolated a disintegrin from the venom of Echis carinatus called EC3 that potently and preferentially inhibited the interactions of ␣4 integrins with the immunoglobulin family members VCAM-1 and mucosal addressin cell adhesion molecule (12). EC3 is composed of two subunits, A and B, in which the RGD motif is substituted by VGD and MLD sequences, respectively (13). Because of the sequence similarity between ␣9 and ␣4, and because both ␣4 integrins and ␣9␤1 recognize VCAM-1 as a ligand, in this manuscript we examined the effects of EC3 on ␣9␤1-mediated adhesion and neutrophil migration. We also purified a novel related heterodimeric disintegrin, EC6, from E. carinatus venom, demonstrated that, like EC3, it also has a substitution of MLD for the RGD sequence in one subunit, and examined the effects of this disintegrin on integrin-mediated adhesion and migration.

MATERIALS AND METHODS
Reagents-Bovine serum albumin (BSA), formylmethionylleucylphenylalanine (fMLP), and dextran were purchased from Sigma. Recombinant human TNF␣ was obtained from R & D Systems (Minneapolis, MN). A recombinant form of the third fibronectin type III repeat of chicken tenascin-C (14) containing alanine substitution mutations within the RGD site (TNfn3RAA) (14), obtained from Anita Prieto and Kathryn Crossin (Scripps Research Institute, La Jolla, CA), and a recombinant form of the N-terminal fragment of the B splice variant of human osteopontin containing alanine substitution mutations within the RGD site (nOPNb RAA) (6) were prepared in Escherichia coli. A recombinant VCAM-1/IgG chimera was obtained from Ted Yednock (Elan Pharmaceuticals, South San Francisco, CA) and from Biogen, Inc. (15). Ficoll-Hypaque PLUS for isolation of neutrophils from venous blood was purchased from Amersham Pharmacia Biotech and used according to the manufacturer's specifications. Fibronectin was purchased from Calbiochem. Highly purified fibrinogen was a gift of Dr. A. Z. Budzyinski (Department of Biochemistry, Temple University). Peptides AEIDGIEL and SVVYGLR and their controls were synthesized using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry on a peptide synthesizer (model 432A, PerkinElmer Life Sciences) at the Center Laboratory for Research and Education, Osaka University, followed by purification with C18 reversed-phase column chromatography. Peptide CKKAMMLDGLNDYC from EC6A and control peptide CKKAMAA-GLNDYC were synthesized and purified by Sigma.
Flow Cytometry-Cultured cells were harvested by trypsinization and rinsed with Tris-buffered saline containing 1 nM CaCl 2 and 1 mM MgCl 2 (TBS). Nonspecific binding was blocked with normal goat serum at 4°C for 10 min. To evaluate the effects of disintegrins on expression of a ligand binding-dependent epitope on the integrin ␤1 subunit, mockor ␣9-transfected SW480 cells were incubated with or without various concentrations of disintegrins in DMEM for 15 min at 4°C and then incubated with primary antibodies (15/7) at 15 g/ml for 20 min at 4°C, followed by secondary antibodies conjugated with phycoerythrin (Chemicon, Temecula, CA). Between incubations, cells were washed twice with TBS. The stained cells were re-suspended in 100 l of TBS, and fluorescence was quantified on 5,000 cells with a FACScan (Becton Dickinson, Rutherford, N.J.).
Purification of Disintegrins-Monomeric and heterodimeric disintegrins were purified from lyophilized viper venoms provided by Latoxan (Valence, France) by one or two steps of HPLC. Echistatin, EC3, and EC6 were purified from E. carinatus suchoreki venom. Lyophilized venom was dissolved in 0.1% trifluoroacetic acid (30 mg/ml). The solution was centrifuged for 5 min at 37°C to remove insoluble proteins, the pellet was discarded, and the supernatant was applied to a C-18 HPLC column (Vydac, Hesperia, CA). The column was eluted with an acetonitrile gradient (0 -85%) over 45 min. Fig. 1 shows that several fractions were eluted from the column under these conditions. Fraction 1 was identified as echistatin, fraction 3 was called EC3, and fraction 6 was called EC6. To obtain higher purity, fractions were re-chromatographed over the same column with a more gradual gradient. A similar method was used to purify eristostatin and EMF-10 from Eristocophis macmahoni venom, kistrin from Agkistrodon rhodostoma venom, and flavoridin from Trimeresuris flavoriridis venom. The purity and molecular mass of each disintegrin was evaluated by SDS-polyacrylamide gel electrophoresis and mass spectrometry. Amino acid sequencing was carried out by Edman degradation following reduction and ethylpyridylation (13,18). Monomeric disintegrins, tested according to McLane et al. (19), were potent inhibitors of platelet aggregation. IC 50 values for echistatin, eristostatin, flavoridin, and kistrin were 136, 59, 51, and 40 nM, respectively. All of the heterodimeric disintegrins had IC 50 values above 1000 nM.
Cell Adhesion Assays-Initial screening cell adhesion assays were performed at Temple University using cells labeled with 5-chloromethyfluorescein diacetate, as described previously (20). Briefly, various ligands, disintegrins, or antibodies were immobilized on 96-well microtiter plates (Falcon, Pittsburgh, PA) by overnight incubation in phosphate-buffered saline at 4°C. The coating concentrations used were 2 g/ml for VCAM-1, 10 g/ml for fibronectin, and 10 g/ml for fibrinogen. Antibodies were coated at 10 g/ml, and disintegrins were coated at 20 g/ml. Wells were blocked with 1% BSA at 37°C. Cells were labeled by incubation with 12.5 M 5-chloromethylfluorescein diacetate in Hanks' balanced salt solution containing 1% BSA for 15 min. Unbound label was removed by washing in the same buffer. Labeled cells were added to the well in the presence and absence of inhibitors and incubated at 37°C for 30 min. Unbound cells were removed by washing, and bound cells were lysed by the addition of 0.5% Triton X-100. A standard curve was prepared using known concentrations of labeled cells. Plates were read using a Cytofluor 2350 fluorescence plate reader (Millipore, Bedford, MA) with a 485-nm excitation filter and a 530-nm emission filter.
Subsequent adhesion assays were performed in San Francisco and Hiroshima using unlabeled cells. For these assays, wells of non-tissue culture-treated polystyrene 96-well flat-bottomed microtiter plates (Nunc Inc., Naperville, IL) were coated by incubation with 100 l of VCAM-1/Ig (10 g/ml), TNfn3RAA (1 g/ml), or nOPNb RAA (1 g/ml) for 1 h at 37°C. After incubation, wells were washed with phosphatebuffered saline and then blocked with 1% BSA in DMEM at 37°C for 30 min. Control wells were filled with 1% BSA in DMEM. SW480 or CHO cells were detached using trypsin/EDTA and resuspended in serum-free DMEM. For blocking experiments, cells were incubated with or without 10 g/ml of Y9A2 or various disintegrins or with or without peptides for 15 min at 4°C before plating. The plates were centrifuged (top-side up) at 10 ϫ g for 5 min before incubation for 1 h at 37°C in humidified 5% CO 2 . Non-adherent cells were removed by centrifugation top-side down at 48 ϫ g for 5 min. Attached cells were fixed with 1% formaldehyde and stained with 0.5% crystal violet, and the wells were washed with phosphate-buffered saline. The relative number of cells in each well was evaluated after solubilization in 40 l of 2% Triton X-100 by measuring the absorbance at 595 nm in a microplate reader (Bio-Rad). All determinations were carried out in triplicate.
Neutrophil Transmigration Assays-Neutrophils were purified from human peripheral venous blood containing 20 units/ml heparin. Neutrophils were isolated by Ficoll-Hypaque density gradient centrifugation, followed by 3% dextran sedimentation (21). Erythrocytes were subjected to hypotonic lysis, and remaining neutrophils were washed and resuspended in phosphate-buffered saline. The isolated neutrophils were more than 95% pure and more than 95% viable as assessed by Wright-Giemsa staining and trypan blue exclusion, respectively. Transendothelial neutrophil migration was assessed as described by Cooper et al. (22). HUVE cells were plated onto polycarbonate inserts (Transwell, Costar, Cambridge, MA; diameter, 6.5 mm; pore size, 8 m for a 24-well plate) in serum-containing endothelial cell growth medium and allowed to grow to confluence over 72 h. 24 h before assays, upper chambers were washed twice with serum-free medium, and new medium with or without 3 ng/ml TNF␣ was added. Immediately prior to the addition of neutrophils, the upper chambers were washed twice with serum-free DMEM, and medium in the lower chamber was replaced with 500 l of serum-free DMEM with or without 10 nM fMLP. Purified neutrophils were incubated with no antibody, Y9A2 (10 g/ml), IB4 (20 g/ml), or various disintegrins for 15 min at 4°C, and 2 ϫ 10 5 cells in 200 l of medium were added to each upper chamber. After 3 h at 37°C in 5% CO 2 , non-adherent cells in the upper chamber were removed. Medium including migrated neutrophils from the lower chamber was collected, the lower chamber was rinsed several times to collect all of the neutrophils that had transmigrated, and the absence of additional adherent neutrophils was confirmed microscopically. The medium and all washes were pooled and resuspended, and cells were counted with a hemocytometer. All determinations were carried out in duplicate and repeated at least twice. Results were analyzed for statistical significance with an analysis of variance followed by the Fisher exact test.

Isolation and Peptide Sequencing of the Novel Heterodimeric
Disintegrin EC6 -We have previously described a heterodimeric disintegrin, EC3, that specifically inhibited ␣4 integrins, with much less effect on ␤3 integrins or the integrin ␣5␤1, which recognize RGD-containing ligands. In contrast to most of the previously described disintegrins, each chain of EC3 contained a novel sequence in the hairpin loop known to interact with integrins. In one loop the usual RGD (or KGD) sequence was replaced with VGD, and in the other loop RGD (or KGD) was replaced with the even more divergent sequence MLD. As part of an ongoing effort to improve the design of integrin inhibitors based on sequences in naturally occurring snake venom disintegrins, we identified and purified a novel heterodimeric disintegrin, EC6, from the venom of E. carinatus. Purification was performed from crude venom as described previously (19) by reversed-phase high pressure liquid chromatography using a Vydac C-18 column and an acetonitrile gradient. EC6 eluted as a single peak at an acetonitrile concentration of 60% (Fig. 1). After reduction and ethylpyridylation, EC6 yielded two fractions that we called EC6A and EC6B (based on the elution pattern of reduced and alkylated subunits). By mass spectrometry we determined that the molecular mass of EC6 was 14,807 Ϯ 2 Da, composed of subunits EC6A (8524 Da) and EC6B (8411 Da), each containing 10 ethylpyridylated cysteine residues. By SDS-polyacrylamide gel electrophoresis, purified EC6 ran as a single band, and the apparent molecular mass was reduced by ϳ50% upon reduc-tion, consistent with a disulfide-linked dimer ( Fig. 2A). Peptide sequencing after CNBr degradation and proteolytic digestions allowed us to determine the complete amino acid sequence of each subunit, as shown in Fig. 2B. Alignment of the sequences of EC6A and EC6B with the previously described heterodimeric disintegrins EC3 and EMF10 and with the monomeric disintegrins echistatin, eristostatin, kistrin, and flavoridin is shown in Fig. 3. It is apparent that there is considerable homology among all of these disintegrins, including alignment of all 10 conserved cysteine residues in each subunit of each heterodimeric disintegrin (Fig. 3). However, like EC3B, EC6A contains an MLDG (instead of RGDX) motif in the putative hairpin loop region. In contrast to EC3, the other subunit of EC6 contains a typical RGD sequence.
Effects of EC6 on Adhesion Mediated by ␣5␤1 or ␣4␤1 Integrins-To determine whether the MLDG motif was the critical disintegrin motif involved in inhibition of ␣4␤1-mediated adhesion to VCAM-1, we performed adhesion assays with Jurkat cells, which express ␣4␤1 but not ␣4␤7. We compared the ability of EC3 and EC6, each of which contain a subunit expressing the MLDG motif, to five other disintegrins that lack this motif. In parallel, we evaluated the effects of each of these disintegrins on the adhesion of K562 cells to fibronectin (mediated through an RGD site by the integrin ␣5␤1) and of ␣IIb␤3-transfected CHO cells to fibrinogen (an effect that has previously been shown to be inhibited specifically by the disintegrin eristostatin). As we have previously reported (12), EC3 potently inhibited the ␣4␤1-mediated adhesion but was considerably less potent in inhibiting adhesion mediated by ␣IIb␤3 or ␣5␤1 (Fig. 4). With the exception of eristostatin, the RGDcontaining disintegrins preferentially inhibited ␣5␤1-mediated adhesion, with no effect on adhesion mediated by ␣4␤1. However, EC6 inhibited adhesion mediated by ␣4␤1 as well as ␣5␤1. These results suggest that the MLDG motif, uniquely shared by EC3 and EC6, is critical for inhibition of ␣4␤1mediated adhesion to VCAM-1. The potent inhibition of ␣5␤1mediated adhesion by EC6 is probably due to the presence of a typical RGD sequence in the EC6B subunit.
Interaction of EC3 and EC6 with the Integrin ␣9␤1-Because ␣9␤1 is the only integrin other than ␣4 integrins that has been reported to recognize VCAM-1 as a ligand, we next sought to determine whether ␣9␤1 would bind to EC3 and/or EC6. As a first step, adhesion assays were performed to examine the attachment of ␣9or mock-transfected SW480 colon carcinoma cells to immobilized EC3 or EC6 or to immobilized antibodies to ␣9␤1 or ␣5␤1 integrins (Fig. 5). We also examined attachment to immobilized antibody against the integrin ␤1 subunit (as a positive control) and the integrin ␤7 subunit (as a negative control). As expected, both cell lines attached to antibodies to ␣5␤1, because this integrin is highly expressed on SW480 cells, but only the ␣9 transfectants attached to antibody to ␣9␤1. ␣9-transfected cells, but not mock-transfected cells, attached to immobilized EC3, an effect that was completely blocked by antibody to ␣9␤1 or antibody to ␤1 but unaffected by antibody to ␣5 (data not shown). Both cell types attached to immobilized EC6, consistent with the potent effect of EC6 in inhibiting ␣5␤1-mediated cell adhesion, and both cell types attached to EMF10, another potent antagonist of ␣5␤1.
Similar results were obtained when disintegrin binding was assessed by flow cytometry using the ligand binding-dependent epitope on the integrin ␤1 subunit recognized by monoclonal antibody 15/7. Incubation with EC3 only induced 15/7 binding in ␣9-transfected cells, whereas EC6 induced expression in both mock and ␣9 transfectants (Fig. 6). These data are consistent with binding of EC3 preferentially to ␣9␤1. Because SW480 cells also expresss ␣5␤1, which is potently inhibited by EC6, we could not evaluate interaction of EC6 with ␣9␤1 by these assays.
Effects of Disintegrins on the Adhesion of ␣9-Transfected SW480 Cells to VCAM-1-We initially screened a range of concentrations of monomeric and dimeric disintegrins for their ability to inhibit the adhesion of ␣9-transfected SW480 cells to immobilized VCAM-1. Only EC6 and EC3 potently inhibited adhesion, with IC 50 values of approximately 30 nM (Fig. 7). To confirm that these effects were not cell type-specific and to confirm that the inhibitory effect observed was due to the inhibition of ␣9␤1-mediated adhesion, we also examined the effects of EC3, EC6, and the control disintegrin echistatin on adhesion of mock-or ␣9-transfected CHO cells to VCAM-1. Both EC3 and EC6, but not echistatin, potently inhibited adhesion of ␣9-transfected CHO cells, whereas the minimal adhesion of mock-transfected cells was unaffected by any disintegrin (Fig. 8).
An MLDG-containing Peptide Inhibits Adhesion of ␣9and ␣4-Transfected CHO Cells to VCAM-1-EC3 and EC6, the two disintegrins we have found to preferentially inhibit ␣9␤1 and ␣4 integrins, share the novel sequence MLDG within the putative disintegrin loop. To determine whether this sequence is indeed critical, we examined the effects of a range of concentrations of the synthetic peptide CKKAMLDGLNDYC, corresponding to this loop region, on adhesion of mock-transfected CHO cells, ␣9 transfectants, and ␣4 transfectants to either VCAM-1 or the irrelevant integrin ligand fibronectin. We also examined the effects of a mutant form of this peptide in which the MLDG sequence was changed to MAAG (Fig. 9). The MLDG-containing peptide caused concentration-dependent inhibition of adhesion of ␣9 and ␣4 transfectants to VCAM-1 but had no effect on adhesion to fibronectin. In contrast, the MAAG mutant had no effect on adhesion, even at the highest concentration examined (2 mM). These data support the hypothesis that the sequence MLDG is critical for specific inhibition of ␣9␤1 and ␣4 integrin-mediated interactions with VCAM-1.

EC3 and EC6 Inhibit Chemotactic Neutrophil Migration across Activated Endothelial Cells-
We have previously shown that both ␣9␤1 and ␣4␤1 are required for optimal migration of human neutrophils across TNF␣-activated human umbilical vein endothelial cells. We therefore sought to determine whether EC3 and EC6, which potently inhibit cell adhesion mediated by ␣9␤1 and ␣4␤1, would also inhibit neutrophil transendothelial migration. In the absence of a chemotactic gradient, there was little neutrophil migration across TNF␣activated HUVE cells, and the low level of baseline migration was unaffected by disintegrins or by antibodies against ␣9␤1 or ␤2 integrins, as we have reported previously (1) (Fig. 10). However, in the presence of a chemotactic gradient of the neutrophil chemoattractant fMLP, migration was markedly increased, and this fMLP-induced transendothelial migration was potently inhibited by EC3, EC6, or anti-␣9␤1 but largely unaffected by the control disintegrin echistatin. It is noteworthy that neither EC3 nor EC6 has any inhibitory effect on ␤2 integrins at concentrations up to 1 mM. 2 Numerous previous reports have identified an important role for ␤2 integrins in neutrophil attachment and migration. However, as we have reported previously (1), under the conditions utilized in these experiments, migration is only partially inhibited by a blocking antibody against ␤2 integrins, allowing us to examine the important roles of ␣9␤1 and ␣4 integrins in this process.

Specificity of EC3 and ␣9␤1 Recognition Peptides from Tenascin-C and Osteopontin in ␣9␤1-Mediated Adhesion to Each
Ligand-Based on the adhesion data described above, EC3 appeared to be a highly potent and relatively specific inhibitor of ␣9␤1 and ␣4 integrins. We therefore expected EC3 to also potently inhibit ␣9␤1-mediated adhesion to the extracellular matrix ligands tenascin-C and osteopontin. To examine this possibility, we performed adhesion assays with ␣9-transfected 2 C. Marcinkiewicz, unpublished observations. FIG. 2. A, Coomassie Blue-stained nonreduced (lane 1) and reduced (lane 2) SDS-polyacrylamide gel electrophoresis (20%) demonstrating purity of EC 6. The apparent decrease in molecular mass upon reduction is consistent with the prediction of a disulfide-linked dimer. B, peptide map of EC6. EC6 was subjected to reduction and ethylpyridylethylation, and subunits A and B were separated by HPLC. N-terminal Edman degradation estimated 41 N-terminal peptides. Peptides CNBr-1 and -2 were isolated following EC6A cleavage at methionine 42. Peptides K7, K8, K10, and K14 were isolated following Lys-C digestion. Peptides D9, D16a, and D16b were isolated following endoproteinase N digestion. Conserved cysteine residues and the RGD (MLD) site are sown in bold. EP, ethylpyridylated. CHO cells on recombinant fragments of tenascin-C and osteopontin that we have engineered to be specific ligands for ␣9␤1 (by mutating the RGD site in each fragment to RAA). However, cell adhesion assays with ␣9-transfected CHO cells demonstrated that concentrations of EC3 up to 100 nM had little effect on adhesion to either of these ligands (Fig. 11, B and C). We have previously mapped the binding sites of ␣9␤1 in tenascin-C and osteopontin to the linear peptide sequences AEIDGIEL and SVVYGLR, respectively (6,7). Each of these peptides inhibits, albeit with low potency, ␣9␤1-mediated adhesion to the ligand from which it was derived. The differential effects of EC3 on adhesion to VCAM-1 and tenascin-C and osteopontin suggested that each ligand might bind to different sites on ␣9␤1. If true, this would predict that peptides derived from each ligand would preferentially inhibit adhesion to that ligand. On the other hand, if each of the ligands competes for an identical binding site on ␣9␤1, the order of potency of inhibitors should be similar for inhibition of adhesion to all three ligands. The results clearly demonstrate that EC3 preferentially inhibits adhesion to VCAM-1 and that the peptides AEI-DGIEL and SVVYGLR preferentially inhibit adhesion to tenascin-C and to osteopontin, respectively, suggesting that each of these three ligands interacts with distinct binding sites on ␣9␤1 (Fig. 11). DISCUSSION We have purified and sequenced a novel heterodimeric disintegrin from the venom of the viper E. carinatus that we call EC6. This disintegrin, like the related disintegrin EC3, contains a substitution of the sequence MLDG for the usual RGDX present in a putative hairpin loop of one of its two subunits. Like EC3, EC6 potently inhibits adhesion of ␣4␤1-expressing cells to VCAM-1. Furthermore, we now demonstrate that EC3 and EC6 are the only disintegrins examined that inhibit adhesion of the integrin ␣9␤1 to VCAM-1. A peptide composed of the MLDG-containing disintegrin loop of EC6A, but not a mutant peptide containing the sequence MAAG, also inhibits ␣4␤1and ␣9␤1-mediated adhesion to VCAM-1. EC3 and EC6 also potently inhibit chemotactic transendothelial migration of human neutrophils across TNF␣-activated endothelial cells, an effect that we have previously shown to be mediated by ␣4␤1 and ␣9␤1 (1). These results establish the sequence MLDG as a defining feature of specific inhibitors of ␣9␤1and ␣4-containing integrins and identify EC3 and EC6 as by far the most potent inhibitors of either of these integrins described to date. In contrast to EC3, EC6 also potently inhibited adhesion mediated by the integrin ␣5␤1. This effect is probably due to the RGD sequence present in the EC6B subunit and the high degree of homology of the disintegrin hairpin loop to other disintegrin inhibitors of ␣5␤1. Because EC3 specifically inhibited adhesion to each of the ␣4 integrin ligands examined, we expected that EC3 would also be a potent inhibitor of ␣9␤1-mediated adhesion to each of its ligands. Indeed, our initial identification of VCAM-1 as a ligand for ␣9␤1 was based on the sequence similarity between a crit-ical tripeptide in the ␣4␤1-binding site in VCAM-1 (IDS) and a critical tripeptide in the ␣9␤1-binding site in tenascin-C (IDG). It is worth noting, however, that the precise ␣9␤1-binding site in VCAM-1 has not been definitively mapped. Surprisingly, EC3 was not found to be a potent inhibitor of ␣9␤1-mediated adhesion to recombinant fragments from the ligand-binding sites of either tenascin-C or osteopontin. Examination of the inhibitory effects of peptides derived from the binding sites in FIG. 7. Effects of various concentrations of monomeric and dimeric disintegrins on the adhesion of ␣9-transfected SW480 cells to immobilized VCAM-1. Fluorescently labeled cells (10 5 cells/well) were mixed with disintegrins, added to 96-well plates coated overnight with VCAM-1 (2 g/ml), incubated at 37°C for 30 min, and washed. Bound cells were lysed in 0.5% Triton X-100, fluorescence was measured, and percent inhibition of adhesion was calculated in comparison to fluorescence of adherent cells in the absence of disintegrins.  fibronectin (B, D, and F). CHO cells incubated with or without (-) a range of concentrations of each peptide were added to 96-well plates coated with VCAM-1/Ig (10 g/ml) or fibronectin (10 g/ml). Cells were allowed to attach for 60 min, non-adherent cells were removed by centrifugation, and adherent cells were stained with crystal violet and quantified by measurement of absorbance at 595 nm. Data are expressed as the mean ϩ S.D. of triplicate measurements from a single experiment.
each of these ligands on ␣9␤1-mediated adhesion to VCAM-1 made it clear that the order of potency of each of these three inhibitors is distinct, with EC3 most potently inhibiting adhe-sion to VCAM-1, the tenascin-C-derived peptide most potently inhibiting adhesion to tenascin-C, and the osteopontin-derived peptide most potently inhibiting adhesion to osteopontin. These results are not compatible with a model by which all three ligands compete for an identical binding site on ␣9␤1 but rather suggest that each ligand is interacting with somewhat different sites on the integrin. This conclusion contradicts the simplest model proposed for integrin-ligand interactions, in which a negatively charged amino acid residue in an integrin ligand contributes the final coordination site for a divalent cation within a single binding site created by a "Midas-motif" (23). However, in the absence of definitive three-dimensional structural data we cannot determine whether the interaction sites of ␣9␤1 with each ligand are actually distinct or overlapping. The similarities in sequence between the MLDG peptide in EC3 and EC6 and the AEIDGIEL sequence in tenascin-C, as well as the partial inhibition by the AEIDGIEL peptide of ␣9␤1-mediated adhesion to osteopontin (Fig. 10), suggest the possibility that there is some overlap among these binding sites.
Although ␣4 integrins and ␣9␤1 both bind to VCAM-1 (and in both cases binding is potently inhibited by EC3 and EC6), there are also considerable differences in ligand binding specificity among these integrins. For example, ␣4 integrins do not bind to any site on TNfn3, and ␣9␤1 does not bind to the CS-1 site in fibronectin (an ␣4␤1 ligand) or to the ␣4␤7 ligand mucosal addressin cell adhesion molecule. 3 These latter findings are of interest because EC3 and EC6 potently inhibit ␣4 integrin-mediated adhesion to both CS-1 and mucosal addressin cell adhesion molecule. These findings are consistent with numerous reports of dramatic differences among RGD-binding integrins to interact with RGD sequences in different ligands.
This report is not the first to suggest that a single integrin can use different sites to interact with different ligands. For example, Tripani-Lombardo et al. (24) identified a monoclonal antibody, LJ-P5, that effectively blocked the binding of von Willebrand factor, but not fibrinogen, to the platelet integrin ␣IIb␤3 and suggested that these two ligands interacted with FIG. 10. Effects of disintegrins on transmigration of neutrophils across activated human endothelial cell monolayers. Purified human neutrophils that had been incubated with or without echistatin (30 or 300 nM), EC3 (30 or 300 nM), EC6 (30 or 300 nM), anti-␣9␤1 antibody (Y9A2, 10 g/ml), or anti-␤2 antibody (IB4, 20 g/ml) were added to upper chambers above confluent monolayers of HUVE cells that had been incubated with TNF␣ (3 ng/ml) for 24 h. Serum-free DMEM containing fMLP (10 nM) or serum-free DMEM alone was added to the lower chamber. After 3 h of incubation at 37°C in 5% CO 2 , neutrophils that had migrated across the monolayer were collected from the lower chamber and counted. Data are expressed as the mean ϩ S.D. of quadruplicate measurements from two separate experiments. *, p Ͻ 0.05; **, p Ͻ 0.01.
FIG. 11. Effects of disintegrins and synthetic peptides on the adhesion of ␣9-transfected CHO cells to VCAM-1, TNfn3RAA, and nOPNb RAA. ␣9-transfected CHO cells incubated with or without disintegrins (30 or 100 nM) or synthetic peptides (1 mM) were added to 96-well plates coated with VCAM-1/Ig (10 g/ml) (A), TNfn3RAA (1 g/ml) (B), or nOPNb RAA (1 g/ml) (C). Cells were allowed to attach for 60 min, non-adherent cells were removed by centrifugation, and adherent cells were stained with crystal violet and quantified by measurement of absorbance at 595 nm. Data for one of three typical experiments are shown and are expressed as the mean ϩ S.D. of triplicate measurements. distinct sites on the integrin. Furthermore, Rahman et al. (25) showed that the disintegrin elagantin was considerably more effective in inhibiting the adhesion of ␣IIb␤3 to fibronectin than to fibrinogen, whereas a mutant version of the neurotoxin dendrospin was a more potent inhibitor of fibrinogen binding. Again the authors suggested that these findings implied that each ligand interacted with somewhat different sites on the same integrin. Perhaps the most convincing evidence in support of this idea comes from recent reports that echo virus and collagen bind to two distinct sites on the isolated insertional domain from the integrin ␣2 subunit (26), a finding that has been confirmed by differential effects of substitutions within this domain on adhesion to each ligand (27).
Venomous vipers have evolved remarkable overlapping mechanisms to produce tissue injury and prevent appropriate repair in their victims. Viper venoms have now been shown to contain at least four classes of proteins involved in this process. These include C-lectin-like proteins, venom metalloproteinases (reprolysins) that may or may not include a disintegrin domain (28), and monomeric and dimeric disintegrins. Monomeric disintegrins potently inhibit ␤3-containing integrins, including the platelet integrin ␣IIb␤3 (10,11), and by interfering with platelet aggregation, result in extensive tissue hemorrhage at the site of viper bites. C-lectin-like proteins can also potentiate local hemorrhage by interfering with von Willebrand factor binding to platelet glycoprotein GPIb on platelets and endothelial cells (29,30). In addition, these venom proteins can inhibit the collagen-binding integrin ␣2␤1, 4 which could further impair healing of the viper wound. Reprolysins further potentiate local injury by degrading components of the extracellular matrix and in some cases degrading the ␣2␤1 integrin (31). ␣9␤1 and ␣4 integrins play critical roles in the transendothelial and extravascular migration of neutrophils, lymphocytes, eosinophils, and monocytes. Our identification of heterodimeric disintegrins in viper venom that potently inhibit these integrins suggests that vipers have also evolved mechanisms to impair leukocyte-mediated repair of wound sites, further increasing the likelihood that victims will be incapacitated or killed by viper bites. These highly potent, naturally occurring integrin inhibitors provide a basis for the development of more potent drugs that could be used to inhibit the excessive tissue emigration of leukocytes that characterize a number of common diseases.