Substitutions of proline 42 to alanine and methionine 46 to asparagine around the RGD domain of the neurotoxin dendroaspin alter its preferential antagonism to that resembling the disintegrin elegantin.

Previous studies have shown that the neurotoxin dendroaspin and the disintegrin kistrin, which show little overall sequence homology but similar residues around RGD (PRGDMP), preferentially inhibited platelet adhesion to fibrinogen. In contrast, the elegantin which has different amino acids around RGD (ARGDNP) preferentially inhibited platelet adhesion to fibronectin. To investigate further the role of amino acids around RGD in disintegrins, we have constructed the genes of a wild-type and of two mutant dendroaspins with substitutions around the RGD, namely [Asn46]- and [Ala42,Asn46]-dendroaspins. Proteins were expressed in Escherichia coli as glutathione S-transferase fusion recombinants and purified to homogeneity by affinity chromatography and reversed phase high performance liquid chromatography. Platelet aggregation studies revealed that wild-type dendroaspin showed an IC50 value similar to that of native dendroaspin, with [Ala42,Asn46]-dendroaspin showing an IC50 value similar to that of elegantin. Interestingly, in platelet adhesion assays, the mutants showed a progressive shift in inhibitory preference, in particular, [Ala42,Asn46]dendroaspin showed nearly identical behavior as elegantin when fibronectin was the immobilized ligand (IC50 = 0.33 microM and 0.6 microM, respectively, compared with 20 microM for native dendroaspin). Native and recombinant wild-type dendroaspin bound to a single class of binding site exhibiting a Kd = 67 nM; [Asn46]- and [Ala42,Asn46]dendroaspins, however, both produced biphasic isotherms with Kd values = 87 nM and 361 nM for [Asn46]dendroaspin and 33 nM and 371 nM for [Ala42,Asn46]dendroaspin, which are close to those of elegantin (Kd values = 18 nM and 179 nM). These studies prove that the amino acids flanking RGD provide an extended locus that regulate the affinity and selectivity of RGD protein dendroaspin.

Integrins are a family of cell surface receptors that mediate adhesion of cells to each other or to extracellular matrix substrate (1)(2)(3)(4)(5). They are composed of noncovalently associated ␣ and ␤ transmembrane subunits selected from among 16 ␣ and 8 ␤ subunits that heterodimerize to produce 20 receptors (6).
Recently, many proteins from a variety of snake venoms have been identified as potent inhibitors of platelet aggregation and integrin-dependent cell adhesion. The majority of these proteins which belong to the disintegrin family share a high level of sequence homology, are small (4 -8 kDa), cysteine-rich, and contain the sequence RGD (16) or KGD (17). In addition to the disintegrin family, a number of non-disintegrin RGD proteins of similar inhibitory potency, high degree of disulfide bonding, and small size have been isolated from both the venoms of the Elapidae family of snakes (18,19) and leech homogenates (20). All of these proteins are approximately 1000 times more potent inhibitors of the interactions of glycoprotein ligands with the integrin receptors than simple linear RGD peptides, a feature that is attributed to an optimally favorable conformation of the RGD motif held within the protein scaffold. The NMR structures of several inhibitors including kistrin (21)(22)(23), flavoridin (24), echistatin (25)(26)(27)(28), albolabrin (29), decorsin (30), and dendroaspin 1 (31,32) have been reported, and the only common structural feature elucidated so far is the positioning of the RGD motif at the end of a solvent exposed loop, a characteristic of prime importance to their inhibitory action.
Recent studies have implied a role for the amino acids around the tripeptide RGD in regulating the ligand binding specificity shown by snake venom proteins. Scarborough et al. (33) examined a range of disintegrins and observed that those containing RGDW were very effective at inhibiting the interactions of fibrinogen to purified ␣ IIb ␤ 3 but not of vitronectin and fibronectin to purified ␣ v ␤ 3 and ␣ 5 ␤ 1 , respectively, whereas the converse was true for disintegrins containing the sequence RGDNP. Other regions of amino acid sequence divergence may also be contributory (33). We have reported that dendroaspin, a short chain neurotoxin analogue containing the RGD sequence, and the disintegrin kistrin, which show little overall sequence homology but have similar amino acids flanking the RGD sequence (PRGDMP), are both potent inhibitors of platelet adhesion to fibrinogen but poor antagonists of the binding of plate-lets to immobilized fibronectin (34). In contrast, elegantin, which has 65% sequence homology to kistrin but markedly different amino acids around RGD (ARGDNP), preferentially inhibited platelet adhesion to fibronectin as opposed to fibrinogen and binds to an allosterically distinct site on ␣ IIb ␤ 3 complex. These studies suggested that the amino acids around the RGD determine the affinity and selectivity of these RGD proteins. In addition to RGD domains, a number of recent studies have suggested that amino acids at the carboxyl terminus of these proteins may affect their interactions with integrins. Deletion of the PRNP sequence from echistatin has been reported to reduce its ability to inhibit platelet aggregation, implying a reduction in the binding affinity (16). Furthermore, the complete carboxyl-terminal peptide (PRNPHKGPAT) of echistatin not only competed with the binding of echistatin to the ␣ IIb ␤ 3 complex but also enhanced the binding of fibronectin and vitronectin to the purified ␣ IIb ␤ 3 , indicating that this non-RGD component of the protein was able to alter the integrin affinity for glycoprotein ligands (35). However, the mechanism by which amino acids at the carboxyl terminus of these proteins interact with their receptors and their binding characteristics are not yet understood.
In this study we examined the role of amino acids flanking the RGD sequence by expressing the neurotoxin variant dendroaspin in Escherichia coli and using site-directed mutagenesis. Dendroaspin, unlike echistatin, does not have any appreciable sequence at its carboxyl-terminal after Cys 57 making it an excellent model to study the functional role of the nature of the amino acids flanking the sequence RGD. We show that recombinant dendroaspin has inhibitory properties identical with native dendroaspin indicating that the expressed protein has the correct folding and disulfide bonding and that substituting Met 46 3 Asn (PRGDMP 3 PRGDNP) or Met 46 3 Asn and Pro 42 3 Ala (PRGDMP 3 ARGDNP) dramatically altered the preferential inhibitory properties and binding characteristics of the protein to the ␣ IIb ␤ 3 complex to that of the disintegrin elegantin containing the sequence ARGDNP. These studies prove that the amino acid flanking sequence RGD provide an extended locus that determines the preferential selectivity of dendroaspin.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes, T4 polynucleotide kinase, T4 DNA ligase, isopropyl-1-thio-␤-D-galactopyranoside, and DH5␣ competent cells were purchased from Life Technologies Ltd. (Paisley, UK) or Promega Ltd. (Southampton, UK). Vent (exo-) DNA polymerase was supplied by New England Biolabs Ltd. (Hitch, UK). Proteinase Factor Xa was purchased from Boehringer Mannheim (Sussex, UK). Human fibrinogen (grade L) was purchased from Kabi (Stockholm, Sweden). Human fibronectin was supplied by Bioproducts Laboratories (Herts, UK). Lyophilized snake venoms were obtained from either Latoxan (Rosans, France) or Sigma Ltd. (Dorset, UK). The monoclonal reagents PM6/248 and PM6/13, which have specificities for the native ␣ IIb ␤ 3 complex and ␤ 3 subunit, respectively, have been described previously (36). Oligonucleotides were made either in King's College School of Medicine and Dentistry (London, UK) or in Cruachem Ltd. (Glasgow, UK) and further purified by denaturing polyacrylamide gel electrophoresis on a 15% acrylamide, 8 M urea gel. Deoxynucleotide triphosphates (dNTPs), dideoxynucleotide triphosphates (ddNTPs), and plasmid pGEX-3X, a vector that expresses a cloned gene as a fusion protein linked to glutathione S-transferase (GST), and glutathione-Sepharose CL-4B were purchased from Pharmacia Biotech Ltd. (Herts, UK). The Construction of Synthetic Dendroaspin and Mutant Dendroaspin Genes and the Expression Vectors-On the basis of the known amino acid sequence of dendroaspin (18,19), we designed a synthetic den-droaspin gene and mutant dendroaspin genes with high codon usage in E. coli (37) utilizing 5 oligonucleotides each for the coding and the complementary strands. The two strands of the dendroaspin gene were expanded by 7 codons at the 5Ј end to generate a BamHI restriction site and by 6 codons at the 3Ј to produce an EcoRI restriction site, thereby allowing directional cloning. In addition, two oligonucleotides were designed to contain the nucleotide sequence coding for the tripeptide RGD and flanking amino acid residues, allowing introduction of specific replacement codons for the production of mutant dendroaspin variants. Each purified oligonucleotide was phosphorylated at 37°C for 60 min in the presence of 1 mM ATP and T4 polynucleotide kinase. Each pair of overlapping phosphorylated oligonucleotides were annealed separately on a Perkin Elmer thermal cycler. The following program was used: 95°C, 5 min, 70°C, 30 s, then slowly cooling to room temperature. Ligation was performed at 16°C for 15 h in a total volume of 50 l containing approximately 1 nM concentration of each annealed fragment, 50 mM Tris-HCl (pH 7.6), 10 mM MgCl 2 , 1 mM ATP, and 5% PEG 8000 and 5 units of T4 ligase. After ligation, the gene of dendroaspin (or mutant dendroaspin) was amplified by polymerase chain reaction, using 1 l of ligation mixture as a template, 1 l of two 5Ј-overhanging oligonucleotides as primers, and 2 units of vent polymerase. The following program was applied: one cycle of 3 min at 94°C and 1 min at 72°C, followed by 39 cycles of 30 s at 94°C, and 2 min at 72°C. The amplification product was checked and found to be of the expected size (216 bp) as ascertained on a 2% agarose gel and further purified on a 2% low-melting-point agarose gel. The gene of dendroaspin (or of mutant dendroaspins) was digested with EcoRI and BamHI and then cloned into the restricted vector pGEX-3X at the carboxyl terminus of the glutathione S-transferase (GST) gene. The Factor Xa cleavage sequence was positioned 5Ј to the gene coding for the recombinant proteins to produce recombinant plasmid pGEX-dendroaspin gene and pGEX-mutant-dendroaspin gene. The correct orientation and sequence of the genes of dendroaspin and mutant dendroaspins were confirmed by DNA sequencing using the method of Sanger et al. (38).
Transformation and Protein Expression-The cloned vector was used to transform 50 l of E. coli DH5␣ competent cells by standard methods (39). Bacterial culture was carried out as follows; the culture was inoculated with an overnight seed culture (1%, v/v) and grown in LB/ ampicillin medium (100 g/ml) and shaken at 37°C until it reached an A 600 of 0.7, then isopropyl-1-thio-␤-D-galactopyranoside was added to a final concentration of 0.1 mM for induction. The cells were grown for an additional 4 h at 30°C and harvested by centrifugation. In contrast to noninduced transformants, analysis of isopropyl-1-thio-␤-D-galactopyranoside-treated cell lysates by SDS-polyacrylamide gel electrophoresis showed an emergence of a 32-kDa protein corresponding to the GSTfusion protein.
Purification of Elegantin, Dendroaspin, Recombinant Wild-type Dendroaspin, and Mutant Dendroaspins-Elegantin and dendroaspin were purified using reverse-phase HPLC as described previously (40). Recombinant dendroaspins were purified as follows: the cell pellets were suspended in PBS buffer (pH 7.4) containing 1% Triton X-100 and the protease inhibitors phenylmethylsulfonyl fluoride (1 M), pepstatin (5 g/ml), aprotinin (5 g/ml), trypsin inhibitor (1 g/ml), 1 mM EDTA, and sonicated on ice. The sonicated mixture was centrifuged at 7,800 ϫ g at 4°C for 10 min to pellet the cell debris and insoluble material. Recombinant GST-dendroaspin and GST-mutant-dendroaspins from supernatants were purified by affinity chromatography on glutathione-Sepharose CL-4B columns by adsorption in PBS containing 150 mM NaCl and elution with 50 mM Tris-HCl containing 10 mM reduced glutathione (pH 8.0). Elution of the absorbed material with glutathione resulted in the appearance of a major band migrating at 32 kDa (GSTdendroaspin fusion protein) in 12.5% polyacrylamide gels. The appropriate fractions comprising the 32-kDa fusion proteins were then digested in the presence of 150 mM NaCl, 1 mM CaCl 2 , and Factor Xa (1:100, w/w, Factor Xa:fusion protein) at 4°C for 24 h. Treatment of the purified GST-proteins with Factor Xa released recombinant proteins migrating as 7-kDa bands, approximating the size of dendroaspin, and free GST appearing as an intensification of a 28-kDa band on SDSpolyacrylamide gel electrophoresis. The digested mixture was loaded onto a Vydac C 18 reverse-phase HPLC analytical column (TP104) and eluted with a linear gradient of 0 -26% acetonitrile (1.78% per min) containing 0.1% trifluoroacetic acid, followed by 26 -36% acetonitrile in 0.1% trifluoroacetic acid (0.25% per min). When necessary, further analytical columns were run under the same conditions. The fractions from HPLC were freeze-dried, dissolved in water, and assayed for inhibition of ADP-induced platelet aggregation.
Electrospray Ionization Mass Spectrometry of the Dendroaspin Proteins-Electrospray ionization mass spectroscopy was used to deter-mine the molecular sizes of native and mutant dendroaspins. Samples (approximately 50 pmol) were lyophilized and dissolved in 20 l of acetonitrile/water (1:1). Ion-mass spectral analysis was performed with a SSQ 710 mass analyzer (Finnigan Mat, UK) using an injection rate of 5 l/min. Measurement of Platelet Aggregation-Platelet aggregation was measured by the increase in light transmission as described previously (34,40). Briefly, platelet-rich plasma was prepared from citrated human blood obtained from healthy individuals by centrifugation at 200 ϫ g for 15 min. Washed platelets were prepared from platelet-rich plasma and resuspended in adhesion/aggregation buffer (145 mM NaCl, 5 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 10 mM glucose, 3.5 mg/ml bovine serum albumin, and 10 mM HEPES, pH 7.35) and adjusted to a count of 3 ϫ 10 8 /ml. Platelet aggregation (320-l incubations) was induced with 10 M ADP in the presence of 1.67 mg/ml fibrinogen and measured using a Payton dual-channel aggregometer linked to a chart recorder.
Measurement of Platelet Adhesion-Platelet adhesion was measured as described previously (34). Briefly, 96-well plates were coated overnight at 4°C with either human fibrinogen or fibronectin reconstituted in PBS (pH 7.4) at appropriate concentrations (2-10 g/ml, 100 l). Platelets were treated with antagonists at appropriate concentrations for 3 min before the addition (90 l) to the microtiter plates which were preloaded with 10 l of 500 M ADP. After 60 min, the number of adherent platelets was determined by measurement of platelet acid phosphatase activity using 130 l of the developing buffer (sodium acetate, pH 5.5, 10 mM p-nitrophenyl phosphate, 0.1% Triton X-100)/ well, and the absorbance values were read at 410/630 nm on an automated plate reader (34).
Iodination of Ligands and Ligand Binding Studies-Iodination of all proteins used in this study was performed using Enzymobead Radioiodination Reagent (Bio-Rad Laboratories, Ltd., Hertfordshire, UK) according to the manufacturer's specifications. The binding of 125 I-labeled disintegrins, dendroaspin, and mutant dendroaspins to washed platelets was performed under equilibrium conditions essentially as described previously (34). Briefly, the incubation mixture was composed of 300 l of washed platelets (3 ϫ 10 8 /ml), 10 l of agonist (1.75 mM ADP giving a final concentration of 50 M), 10 l of 125 I-labeled protein samples, 5-20 l of resuspension buffer and made to a final volume of 350 l. In antibody inhibition studies, platelet suspensions were treated with antibody for 30 min prior to exposure to ADP and then added to 125 I-protein samples, and the mixture was incubated at room temperature for an additional 60 min. Incubations were terminated by loading onto a 25% (w/v) sucrose, 1% bovine serum albumin cushion and centrifugation at 12,000 ϫ g for 10 min. Both platelet pellets and supernatants were counted to determine the levels of bound and free ligand. Background binding levels were determined in the presence of a 50-fold excess of unlabeled disintegrin or 10 mM EDTA.

Design and Functional Characterization of Recombinant
Wild-type and Mutant Dendroaspins-We had previously noted that dendroaspin and kistrin share similar amino acid residues at positions flanking the tripeptide RGD and speculated that this sequence similarity underpinned the similar functional and binding characteristics of these two structurally unrelated snake venom inhibitors (Table I, Ref. 34). To test this hypothesis, in the present study, we generated dendroaspin variants with specific substitution of the residues at positions flanking RGD to those residues present in elegantin (Table I), a welldefined disintegrin with functional and binding characteristics to the ␣ IIb ␤ 3 complex distinct from those of kistrin and dendroaspin.
The functional characterization of the recombinant wild-type and of the mutant dendroaspins purified from cell lysates of E. coli was determined by platelet aggregation and adhesion assays. In order to verify that the expression system generated was satisfactory, we first compared the functional properties of

RICFNHQSSQPQ-TTKTCSPGESSCYHKQWSDFRGTIIERGCGCP-TVKPGIKLS--CCES-EVCNN
Toxin S 5 C 10  (Table II). We previously observed that measurement of ADP-activated platelet adhesion to immobilized glycoproteins highlights selective inhibitory preferences for RGD snake venom proteins that are less easily discernible using the platelet aggregation assay (34,39). In such experiments, we have shown that dendroaspin and kistrin are potent inhibitors of platelet adhesion to fibrinogen, whereas elegantin preferentially inhibits platelet adhesion to fibronectin (34). Fig. 1 and Table III (showing IC 50  In particular, the latter showed a maximal extent of inhibition and IC 50 values that were markedly similar with those displayed by elegantin. Thus, substituting Pro 42 3 Ala and Met 46 3 Asn in the residues immediately flanking the RGD in dendroaspin altered the inhibitory preferences of dendroaspin to that of elegantin. The presence of asparagine adjacent to aspartic acid would be particularly important in inhibiting the interactions of fibronectin with its receptor. This study strongly supports our previous studies (34). Binding of 125 I-labeled recombinant and mutated dendroaspins and of 125 I-elegantin to activated platelets was studied to determine whether the alterations in functional properties of the mutated dendroaspins were reflected in their binding characteristics. All four 125 I-labeled proteins bound to ADP-activated platelets in a saturable and cation-dependent manner (Fig. 2, insets). Scatchard analysis of the data using the Kinetic, EBDA, Ligand, and Lowry version 4 software programs (BIOSOFT, Cambridge, UK) indicated that recombinant dendroaspin bound to a single class of binding site exhibiting a K d ϭ 67 nM ( Fig. 2A  ]dendroaspin (q), or elegantin (Ⅺ) for 3 min prior to application to microtiter wells coated with either 10 g/ml fibrinogen (A) or fibronectin (B). The number of adherent platelets were determined by measurement of endogenous acid phosphatase as described previously (34). Results are expressed as percent inhibition relative to the number of adherent platelets observed in the absence of inhibitors. All points were performed in quadruplicate, and the mean Ϯ S.E. are shown in Table III (n ϭ 2-4).  (Tables I and II) and have structures unrelated, except for the (A)RGD(N) domain, the close similarity of the dissociation constants is striking.

RICYNHQSNTPA-TTKSCV--ENSCYKSIWADHRGTIIKRGCGCP-RVKSKIK----CCKS-DNCNL
To confirm that both binding sites occupied by the two dendroaspin mutants on ADP-treated platelets were present on two-site fit, R 2 ϭ 0.996 and 0.983, respectively), and 125 I-labeled elegantin (D; twosite fit, R 2 ϭ 0.990 and 0.997, respectively) were incubated with ADP-treated, washed platelets (3 ϫ 10 8 /ml) for 30 min at room temperature in a volume of 350 l. Bound and free levels of 125 I-labeled RGD-containing proteins were determined by loading 300 l of the platelet suspension onto a cushion of 25% (w/v) sucrose, 1% bovine serum albumin and centrifuged for 10 min at 12000 ϫ g. Both the radioactivity in the platelet pellets and supernatants were determined. Insets, saturation isotherms of 125 I-labeled RGD-containing proteins binding to ADPtreated washed platelets. The curves show nonspecific (Ç), specific (Ⅺ), and total (bound ϩ free) binding (É). The values are representive of three similar experiments with all points performed in duplicate. (S.E. were less than 10%.) the ␣ IIb ␤ 3 integrin complex, the effects of two inhibitory antibodies on radioligand binding were monitored. PM6/248, a monoclonal antibody with specificity for the native ␣ IIb ␤ 3 complex (36), effectively inhibited in a dose-related manner the binding of all these 125 I-labeled recombinant dendroaspins by 80 -100% (Fig. 3). In contrast, an anti-␣ 5 ␤ 1 antibody was comparatively ineffective, confirming that the binding parameters observed were specifically associated with the ␣ IIb ␤ 3 complex.
Further evidence for the close similarity in the binding of [Ala 42 ,Asn 46 ]-and [Asn 46 ]dendroaspin and of elegantin to the ␣ IIb ␤ 3 complex was obtained by examining the association kinetics of three ligands (Fig. 4). Native and recombinant dendroaspin show simple and rapid binding, reaching equilibrium by 5 min. However, elegantin, [Asn 46 ]-and [Ala 42 ,Asn 46 ]dendroaspin showed complex association kinetics with approximately 3-to 4-fold higher binding before equilibrium than at equilibrium. The reasons for this complex association pattern are not known at present, but are not due to internalization of the ligand as the binding was fully reversible (data not shown). That [Asn 46 ]-and [Ala 42 ,Asn 46 ]dendroaspin, but not native dendroaspin, behaved in this manner points to this property being solely due to the presence of the ARGDN sequence, and whether other ARGDN-containing disintegrins, e.g. viridin, jararacin, cotiarin (Table I), behave in a similar manner remains to be examined.
These studies report that the amino acids around the RGD motif regulate the affinity and selectivity of the RGD protein dendroaspin and support our earlier studies (34) and those of Scarborough et al. (33). Further details of the mechanisms of integrin/ligand interactions will benefit greatly from the analysis of both wild-type and mutant dendroaspins by x-ray crystallography or NMR spectroscopy. Until the receptor-ligand complexes are available for such structural studies, the further structure/function evaluation of snake venom adhesive ligands may allow us to engineer potent antagonists that show not only ligand specificity but also receptor specificity.