The “Linker” Region (Amino Acids 38-47) of the Disintegrin Elegantin Is a Novel Inhibitory Domain of Integrin α5β1-Dependent Cell Adhesion on Fibronectin

Disintegrins are a family of potent inhibitors of cell-cell and cell-matrix adhesion. In this study we have identified a region of the disintegrin elegantin, termed the “linker domain” (amino acids 38-47), with inhibitory activity toward α5β1-mediated cell adhesion on fibronectin (Fn). Using a chimeric structure-function approach in which sequences of the functionally distinct disintegrin kistrin were introduced into the elegantin template at targeted sites, a loss of inhibitory function toward α5β1-mediated adhesion on Fn was observed when the elegantin linker domain was substituted. Subsequent analysis comparing the inhibitory efficacies of the panel of elegantin-kistrin chimeras toward CHO α5 cell adhesion on recombinant Fn III6-10 fragments showed that the loss of inhibitory activity associated with the disruption of the elegantin linker domain was dependent upon the presence of a functional Fn III9 synergy site within the Fn III6-10 substrate. This suggested that the elegantin linker domain inhibits primarily the activity of the Fn synergy domain in promoting α5β1 integrin-mediated cell adhesion. Construction of a cyclic peptide corresponding to the entire region of the elegantin linker domain showed that this domain has intrinsic α5β1 inhibitory activity comparable with the activity of the RGDS peptide. These data demonstrate a novel biological function for a disintegrin domain that antagonizes integrin-mediated cell adhesion.

Disintegrins are a family of potent inhibitors of cell-cell and cell-matrix adhesion. In this study we have identified a region of the disintegrin elegantin, termed the "linker domain" (amino acids 38 -47), with inhibitory activity toward ␣ 5 ␤ 1 -mediated cell adhesion on fibronectin (Fn). Using a chimeric structurefunction approach in which sequences of the functionally distinct disintegrin kistrin were introduced into the elegantin template at targeted sites, a loss of inhibitory function toward ␣ 5 ␤ 1 -mediated adhesion on Fn was observed when the elegantin linker domain was substituted. Subsequent analysis comparing the inhibitory efficacies of the panel of elegantin-kistrin chimeras toward CHO ␣ 5 cell adhesion on recombinant Fn III 6 -10 fragments showed that the loss of inhibitory activity associated with the disruption of the elegantin linker domain was dependent upon the presence of a functional Fn III 9 synergy site within the Fn III 6 -10 substrate. This suggested that the elegantin linker domain inhibits primarily the activity of the Fn synergy domain in promoting ␣ 5 ␤ 1 integrin-mediated cell adhesion. Construction of a cyclic peptide corresponding to the entire region of the elegantin linker domain showed that this domain has intrinsic ␣ 5 ␤ 1 inhibitory activity comparable with the activity of the RGDS peptide. These data demonstrate a novel biological function for a disintegrin domain that antagonizes integrin-mediated cell adhesion.
Cell adhesion is a complex process involving several classes of molecular complexes, including the integrin family of adhesion receptors. The regulation of integrin-ligand binding is a targeted process for both physiological and pathophysiological mediators. Disintegrins were originally described as inhibitors of platelet aggregation, an integrin-dependent process, isolated as monomeric proteins of 5-8 kDa from viper venoms (1). Subsequently, it has been demonstrated that these molecules are potent integrin ligands and are found in snake venoms as both homo-and heterodimers (2,3), as fusion proteins with metalloproteases, and in mammals as modules within the disintegrin-metalloprotease family. Currently, the monomeric snake venom disintegrin family can be conveniently divided into three different groupings according to their length and prevalence of disulfide bonds, including short (41-51 residues), medium (ϳ70 residues), and long (84 residues) (4).
An archetypal structure within all disintegrins is the presence of a solvent-exposed ␤-loop containing an "integrinbinding" tri-peptide motif (IBM) 5 that almost invariably contains an acidic residue. The exceptions to this are the R/KTS disintegrins, which display specificity for the ␣ 1 -integrins (5,6). The most ubiquitous motif observed in disintegrins is the RGD sequence, which is also found in fibronectin (Fn) and other extracellular matrix and plasma proteins. It is, therefore, assumed that disintegrins bind to integrins through an analogous mechanism to physiological ligands whereby the RGD loop interacts with amino acids of both the ␣and ␤-subunits at the subunit interface of the integrin heterodimer (7,8).
Structure-function studies with disintegrins have shown that the positioning of the IBM at the apex of the ␤-loop and the correct pairing of disulfide bonds are essential features for biological activity (reviewed in Ref. 4). Synthetic peptides corresponding to the entire RGD loop of disintegrins possess only between 6 and 20% of the potency of their parent protein even after cyclization (9). The molecular details of how the structure of disintegrins contribute to their full biological activity still remain obscure, because a high resolution structure of a disintegrin in complex with an integrin receptor is unavailable at present.
The integrin selectivity of disintegrins has been shown to be dependent upon the composition of the amino acid environment surrounding the acidic residue, which comprises the IBM within the ␤-hairpin loop. The disintegrin barbourin, which contains a KGD IBM, was shown to possess a high degree of selectivity toward the integrin ␣ IIb ␤ 3 as apposed to the related integrin ␣ v ␤ 3 (10), now known to be due to the additional length of the aliphatic lysine side chain (8). Tselepis et al. (11) showed that the substitution of the IBM derived from the CS-1 domain of Fn (LDV) in place of the RGD sequence in kistrin altered the integrin specificity of this disintegrin from the integrin ␣ v ␤ 3 toward the integrin ␣ 4 ␤ 1 . Furthermore, several studies have demonstrated that the amino acid residues positioned N-and C-terminal to the RGD sequence in disintegrins modulates the specificity of their binding to integrin complexes (12,13). This specificity has a number of facets, including specificity toward distinct integrin affinity states (12) and the inhibitory preferences toward specific physiological integrin-ligand partnerships (14). Furthermore, recombinant disintegrins with distinct RGD motifs show several mechanisms of competitive behavior for binding to the integrin ␣ IIb ␤ 3 indicative of allosteric modulation (15,16).
In addition to the ␤-hairpin loop containing the IBM, other regions of the disintegrin molecule have been postulated to play a role in modulating their binding to integrins. For example, several studies have implicated a role for the residues at the extreme of the C terminus of short disintegrins such as echistatin and eristostatin as regulators of disintegrin binding to integrins (13,17). These studies illustrate that the C-terminal residues may be involved in the promotion of high affinity conformational states within the integrin complex as detected by the expression of ligand-induced binding site epitopes. However, as studies with Fn have shown, physiological ligands are postulated to interact with integrins in a complex manner involving synergy contacts between receptor and ligand (18 -21). Disintegrin inhibition of physiological integrin-ligand partnerships may necessitate down-regulation of these synergy interactions in addition to direct competition between the IBM containing loops of the disintegrin and the physiological ligand. However, little is known about which regions of the disintegrin structure antagonize the activity of these synergy domains.
The purpose of the present study was to test the hypothesis that disintegrins inhibit and thereby mimic receptor-ligand interactions at synergy sites. We also wanted to identify the structure-function relationship of this biological activity. We, therefore, constructed a panel of chimeric disintegrins based upon the structures of two well characterized "medium-sized" family members elegantin and kistrin with differing inhibitory properties toward ␣ 5 ␤ 1 -mediated cell adhesion. Kistrin sequences were introduced at sites of sequence variation within the elegantin IBM loop: the C-terminal peptide and the "linker domain" (amino acid 38 -47), which connects the N-and C-terminal portions of the molecule. The aim was to identify a potential region of elegantin that functioned as an additional inhibitory domain for the integrin ␣ 5 ␤ 1 , supplementing activity of the IBM-containing loop. The data presented here provide compelling evidence that the linker domain of elegantin (amino acids 38 -47) harbors inhibitory activity toward the integrin ␣ 5 ␤ 1 . This activity is of biological significance through the inhibition of Fn synergy site activity in promoting ␣ 5 ␤ 1 -mediated cell adhesion.
Given that the selectivity of the IBM of barbourin underscored its success as a template for the development of a highly successful pharmacological reagent to treat thrombotic episodes, the potential for the linker domain of elegantin to inhibit ␣ 5 ␤ 1 -Fn synergy interaction identifies a potential for disintegrin linker domains to serve as potential templates for the construction of a novel integrin antagonists.

EXPERIMENTAL PROCEDURES
Materials-Fibrinogen was obtained from Kabi (Stockholm, Sweden, Grade L) and fibronectin was from Sigma. The purity and integrity of these glycoproteins were determined by SDS-PAGE. The ␣ IIb ␤ 3 -directed monoclonal antibody AP2 was a generous gift from Dr. Tom Kunicki (Scripps Research Institute, La Jolla, CA). All other antibodies were obtained from Chemicon (Harrow, UK).

Construction and Expression of Elegantin-Kistrin Chimeras-
The construction, expression, and characterization of recombinant elegantin and variants constructs has been described previously (12,16). The elegantin-kistrin chimeric constructs described in the present study were prepared using these elegantin constructs (12) as templates with additional modifications prepared by oligonucleotide-based site-directed mutagenesis using the Transform mutagenesis procedure (Clontech, Basingstoke, Herts, UK) according to the manufacturer's instructions. In targeting the various regions of the elegantin sequence for amino acid substitution, only non-conserved residues were altered. In some instances, successive rounds of mutagenesis were necessary to obtain full sequence substitution. Recombinant elegantin-kistrin chimeras were subcloned into pGEX-3X (Amersham Biosciences) vectors or pUBHis10 (gift from Dr. T. Butt, LifeSensors Inc., Philadelphia, PA) and expressed. Recombinant fusion proteins were prepared from the lysates of Epicurean coli (DE3)pLysS (Stratagene, La Jolla, CA) by affinity chromatography on glutathione-Sepharose (Amersham Biosciences) or nickel chelating resins (Sigma) using a batch procedure. Purified recombinant disintegrins were analyzed by SDS-PAGE indicating that the preparations were Ͼ95% homogeneous with minimal proteolytic degradation of the sample observed.
Protein Modification and Peptide Synthesis-NHS-fluorescein (5-and 6-carboxyfluorescein, succinimidyl ester, Pierce & Warriner Ltd., Chester, UK) was used to conjugate glutathione S-transferase disintegrins (1 mg/ml) in 50 mM sodium bicarbonate buffer, pH 8.5. The conjugation reaction was carried out at 4°C for 2 h. The fluorescein-conjugated protein was separated from unconjugated fluorescein by gel filtration on PD 10 columns (Amersham Biosciences). Fluorescence:protein ratios were determined for each protein which was stored at Ϫ40°C. Cyclic synthetic peptides corresponding to the elegantin linker domain were synthesized by Alta Biosciences Ltd. (Birmingham, UK) using conventional Fmoc (N-(9-fluorenyl)methoxycarbonyl) solid-phase chemistry. Wild-type elegantin linker sequence peptide (amino acids 38 -47) P L (-S-CRFKKKRTIC-S-) and a scrambled version P C (-S-CRRTKIFKKC-S-) were reconstituted in water and stored at Ϫ20°C.
Cell Adhesion and Binding Assays-K562 cells were obtained from the European Collection of Animal Cell Cultures (Porton Down, UK) and maintained in RPMI 1640 plus Glutamax, supplemented with 10% (v/v) fetal bovine serum. CHO ␣ 5 and CHO ␣ v /␣ 5 cells have been described previously (26) and were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and G418 (200 g/ml). Harvested cells were washed in phosphate-buffered saline containing and resuspended in Tyrode's buffer (10 mM Hepes, pH 7.35, 150 mM NaCl, 5 mM KCl) containing either 1 mM MgSO 4 and 2 mM CaCl 2 (standard physiological cation conditions) or an appropriate concentration of MnCl 2 (as indicated under "Results"). For adhesion assays, cells were treated with or without inhibitor (disintegrins, peptide, or monoclonal reagents) for 30 min prior to application to microtiter wells (100 l of 1 ϫ 10 6 cells/ml), which were pre-coated with either fibrinogen or fibronectin (10 -20 g/ml) or recombinant Fn III 6 -10 fragments (5 g/ml). After washing, adherent cells were quantified by measurement of endogenous acid phosphatase. Fluorescent ligand binding were performed as previously described (16). For studies of ligand association, cells (1 ml of 1 ϫ 10 6 /ml) were suspended in appropriate buffers and incubated with 100 nM of recombinant fluorescein isothiocyanate-conjugated disintegrin for various time periods, and the reaction was stopped by the addition of 2% (v/v) final concentration of ice-cold paraformaldehyde to the incubation mixture. Cells were incubated for 30 min on ice and then washed (four times) in phosphate-buffered saline and analyzed by flow cytometry.

Construction of Elegantin-Kistrin
Chimeras-Elegantin is a disintegrin originally isolated as an inhibitor of platelet aggregation (22) and was subsequently shown to be a highly specific antagonist of the integrin ␣ 5 ␤ 1 (12,23). To identify potential ␣ 5 ␤ 1 integrin synergy binding domains on elegantin, a panel of chimeric disintegrins was constructed employing the elegantin genetic template upon which sequences derived from the close family member kistrin were introduced at structurally homologous regions. This exercise was facilitated by the construction of a molecular model of elegantin from the coordinates of the crystal structure of a closely homologous (93% sequence identity) disintegrin flavoviridin and structural comparison to the solution structure of kistrin (Fig. 1A). Based upon this structural analysis, two additional regions of the elegantin structure were targeted for the introduction of amino acid residues from analogous structures in kistrin due to their proximity to the RGD loop. These were the C-terminal peptide and the peptide that connects the N-terminal portion of the disintegrin structure to the C-terminal portion, which we designated the linker domain (Fig. 1A). Comparison of the amino acid sequences in these regions of the disintegrin structure also showed significant variation. Fig. 1B shows the entire amino acid sequences of the chimeric constructs that were expressed as bacterial fusion proteins. A total of seven novel chimeras were constructed to supplement the existing recombinant panel. The RGD loop of elegantin had been targeted in a previous study where we reported that substitution of Ala 50 to Pro drastically reduced the binding affinity of elegantin for the low affinity conformation of the integrin ␣ 5 ␤ 1 expressed on K562 cells (12). The substituted domains were expressed as elegantin constructs both in isolation and in various combinations with the construct termed C-P107 comprising the linker, IBM, and C terminus of kistrin thereby differing with the wild-type kistrin sequence by only two non-conserved residues and two additional residues at the N terminus (Fig. 1B).
The bacterial fusion elegantin-kistrin chimeras were purified to homogeneity, and the genetic substitutions were confirmed at the protein level by mass spectrometry (data not shown). An initial functional screen for activity was performed by assessing the efficacy of the panel of elegantin chimeras as inhibitors of platelet aggregation. Table 1 shows the IC 50 values for each recombinant disintegrin fell within the range observed for most naturally isolated disintegrins indicating faithful recombinant expression as previously shown (12,16).
Elegantin-Kistrin Chimeras Bind to ␣ 5 ␤ 1 -The efficacy of the panel of elegantin-kistrin chimeras as inhibitors of ␣ 5 ␤ 1 -mediated K562 cell adhesion was also assessed. Under physiological cation buffer conditions under which the integrin ␣ 5 ␤ 1 assumes a low affinity conformation in this cell line (25), the recombinant elegantin-kistrin chimeras expressing the wild-type elegantin RGD motif ( 50 ARGDN 54 ) were strong inhibitors (IC 50 range, 5-25 nM) of K562 cell adhesion on Fn irrespective of the sequences present in the linker domain or at the C terminus ( Fig. 2A). In agreement with our previous observations (12), replacement of Ala 50 with the corresponding kistrin Pro residue within the RGD loop drastically reduced inhibitory potency (Fig. 2B). The one exception being the chimera designated Eg P107, which contains the elegantin C terminus and showed a more potent inhibitory activity compared with other chimeras containing Pro 50 . However, this was still considerably weaker (ϳ5-to 10-fold, IC 50 ϭ 175 nM) than the elegantin chimeras containing the wild-type (Ala 50 ) elegantin RGD motif. Using an ␣ 5 ␤ 1 monoclonal reagent JBS-5 to block disintegrin-␣ 5 ␤ 1 engagement, direct binding of fluorescently labeled elegantin-kistrin chimeras to K562 cells confirmed that the chimeras harboring the kistrin RGD motif failed to interact significantly with K562 cells via the ␣ 5 ␤ 1 integrin with the exception of Eg P107 (Fig. 3).
Under buffering conditions containing Mn 2ϩ ions, K562 cells adhere on Fg in an ␣ 5 ␤ 1 -dependent manner (12). The effect of Mn 2ϩ is to moderately activate the integrin through promoting conformational changes in the ␤-subunit A-domain through displacement of the Ca 2ϩ ion coordinating at the ADMIDAS site (7). Although under these conditions, K562 cell adhesion on Fn is enhanced, we observed previously (12) that the integrin dependence of the cell adhesion was no longer solely due to the activity of the ␣ 5 ␤ 1 integrin but contained a significant ␣ 4 ␤ 1 component, complicating interpretation of the disintegrin inhibitory activity. The ␣ 5 ␤ 1 -dependent adhesion on an Fg substrate has been shown to be mediated by the RGD sequence in Fg located at amino acids 572-574 in the A␣ chain (24).
Under these experimental conditions, all elegantin-kistrin chimeras containing either the elegantin wild-type or kistrin RGD motif, linker, and C-terminal domains were strong and equipotent inhibitors (IC 50 range, 5-10 nM, Fig. 2, C and D). Therefore, these experiments demonstrate that all the recombinant chimeras retain the capacity to ligate the integrin ␣ 5 ␤ 1 and inhibit cell adhesion when a moderately high affinity conformation is induced with Mn 2ϩ . They also demonstrate that the kistrin RGD motif can discriminate between different activation states of the integrin. These observations are in agreement with our previous observations that partial activation of the ␤ 1 -subunit A-domain by Mn 2ϩ is necessary to accommodate the Pro residue of the kistrin RGD motif at residue position 50 in the elegantin primary structure (12).
Inhibition of Integrin ␣ 5 ␤ 1 -Fn Binding Involves the Elegantin Linker Domain-Although the use of Mn 2ϩ ion to promote moderate activation of the ␣ 5 ␤ 1 complex expressed on K562 cells allows for recognition of elegantin-kistrin chimeras with the kistrin RGD motif, previous studies have suggested that the interaction of the integrin ␣ 5 ␤ 1 with Fn does not involve the participation of the synergy PHSRN site located in the ninth type III repeat in these cells when the integrin is not fully activated (25). Furthermore, the interaction of the integrin ␣ 5 ␤ 1 with Fg via the RGD site in the A␣ chain does not involve an additional synergy interaction due to the absence of the PHSRN sequence in Fg. Therefore, to assess the impact of the Fn synergy interaction upon the inhibitory efficacy of the panel of elegantin-kistrin chimeras, we exploited a system expressing ␣ 5 ␤ 1 in a different cellular context, which was known to employ the activity of the Fn synergy site (26). Adhesion experiments were performed using CHO cells expressing the human ␣ 5 -subunit in partner with endogenous hamster ␤ 1 (CHO ␣ 5 ). For comparison, the inhibitory efficacy of the panel elegantin chi-meras was also assessed toward the adhesion of CHO cells expressing a human-hamster hybrid ␣ v ␤ 1 heterodimer (CHO ␣ v ) on Fn, which does not involve interaction with the Fn synergy site (26).
In these experiments, CHO ␣ 5 cell adhesion on Fn was inhibited strongly by Eg WT and the chimera Eg C-WT, which differs from native elegantin (Eg WT) only by the inclusion of the kistrin C terminus (Fig. 4A). Therefore, the presence of the kistrin C terminus did not alter the inhibitory efficacy of the chimeric disintegrin toward the integrin-ligand interaction. Conversely, chimeras Eg P106 and Eg C-P106, which both contain the native elegantin RGD motif but in combination with the kistrin linker domain, were weak inhibitors of CHO ␣ 5 adhesion on Fn, achieving a maximal inhibition of cell adhesion of only 40% (Fig. 4A). Indeed, these chimeras showed a loss of inhibitory activity of a similar magnitude to the elegantin-kistrin chimeras expressing the kistrin RGD motif (Fig. 4B). This observation, therefore, suggested that the elegantin linker domain was indispensable to the inhibitory efficacy of elegantin toward ␣ 5 ␤ 1 integrindependent adhesion on Fn. Significantly, for the CHO ␣ v cell adhesion on Fn, all elegantin-kistrin chimeras were efficient inhibitors achieving maximal levels of inhibition at ϳ100 nM albeit with differing IC 50 values (Fig. 4, C and D). These data show that, for CHO ␣ v -Fn adhesion, both elegantin and kistrin are efficient antagonists and that the presence of the kistrin linker sequence within the elegantin template does not lead to a loss of inhibitory activity as with CHO ␣ 5 -mediated adhesion on Fn. These data implied that the inhibitory activity of the linker domain may relate to the antagonism of the function of the synergy domain of Fn in promoting ligation to the ␣ 5 ␤ 1 integrin.
To test this hypothesis, the panel of elegantin-kistrin chimeras was assessed for their capacity to inhibit CHO ␣ 5 cell adhesion on the recombinant fragment of Fn comprising type III repeats 6 -10 (Fn III 6 -10 ) in which the PRSHN site in the ninth type III repeat was mutated to SPSDN, an inactive sequence derived from domain III 8 . However, it is known that the general level of adhesion of cells expressing ␣ 5 ␤ 1 to this recombinant fragment is low compared with the wild-type fragment. This is thought to reflect a reduction in the affinity of the binding interaction between the integrin and its ligand (27). Previous reports (25) have suggested that treatment of K562 cells with the activating monoclonal TS2/16 or phorbol 12-myristate 13-acetate  can restore equivalent levels of adhesion to both wild-type and mutant fragment. However, in our hands we were unable to achieve a significant improvement in the level of adhesion of K562 cells to Fn III 6 -10 SPSDN by treatment with either TS2/16 or phorbol 12-myristate 13-acetate or both reagents in combination (data not shown). Furthermore, CHO ␣ 5 cells also did not adhere to Fn III 6 -10 SPSDN in a comparable manner to Fn III 6 -10 PRSHN in a Ca 2ϩ /Mg 2ϩ buffer (Fig. 5A). However, when CHO ␣ 5 cells were suspended in a buffer containing 100 M Mn 2ϩ , similar levels of adhesion on both substrates were observed with the retention of ␣ 5 ␤ 1 specificity (Fig. 5B).
Under the experimental conditions shown in Fig. 5, the inhibitory efficacy of the panel of elegantin chimeras toward CHO ␣ 5 cell adhesion on both substrates was assessed. These studies showed that adhesion on Fn III 6 -10 PHSRN was inhibited strongly by Eg WT and Eg C-WT. However, chimeras Eg P106 and Eg C-P106 were comparatively poor inhibitors achieving a maximal inhibition of cell adhesion of only 65% and with ϳ3-fold elevated IC 50 values (Fig. 6A). In contrast, all four elegantin-kistrin chimeras were potent inhibitors of CHO ␣ 5 cell adhesion on Fn III 6 -10 SPSDN achieving a maximal level of inhibition of cell adhesion of 100% at 100 nM with similar IC 50 values (Fig. 6B). These data, therefore, in agreement with the observations for CHO ␣ 5 cell adhesion on Fn under Ca 2ϩ / Mg 2ϩ conditions (Fig. 4), confirm that the loss of inhibitory function toward CHO ␣ 5 cell adhesion associated with the substitution of the elegantin linker domain is only apparent when a Fn substrate contains a functioning synergy site in the III 9 module. We postulated that, if the elegantin linker domain antagonized cell adhesion dependent upon the activity of the Fn III 9 synergy site, then the linker domain and the Fn synergy site may make contact with the same region of the ␣ 5 ␤ 1 complex. Previously, Mould et al. (26) showed, using a chimeric integrin ␣ v subunit comprising the N-terminal residues 1-232 of human ␣ 5 stably expressed in CHO cells with endogenous ␤ 1 (CHO ␣ v /␣ 5 cells), that a region comprising the Fn synergy interaction site on the integrin ␣ 5 subunit was contained within this N-terminal region. In that study, the CHO ␣ v /␣ 5 cell adhesion profile on the Fn III 6 -10 fragments mimicked that of CHO ␣ 5 cells. We, therefore, tested the inhibitory efficacy of our elegantin-kistrin chimeras upon CHO ␣ v /␣ 5 cell adhesion on Fn III 6 -10 PHSRN and Fn III 6 -10 SPSDN. As shown in Fig. 6C, the cell adhesion on the Fn III 6 -10 PHSRN fragment was strongly inhibited by  Eg WT and Eg C-WT with 100% maximal inhibition attained at 150 nM. In contrast, chimeras Eg P106 and Eg C-P106 were comparatively poor inhibitors with only a maximal 35% inhibition of cell adhesion at 200 nM. In contrast, CHO ␣v/␣ 5 cell adhesion on FN III 6 -10 SPSDN was inhibited strongly by all four elegantin-kistrin chimeras with or without the wild-type linker sequence, displaying a maximal inhibition of 100% at 150 nM (Fig. 6D). However, in contrast with CHO ␣ 5 cell adhesion upon FN III 6 -10 SPSDN, Eg WT and Eg C-WT did appear to be slightly more efficacious inhibitors than Eg P106 and Eg C-P106, which displayed a 2-to 3-fold increase in IC 50 values. In spite of this difference between the two types of CHO cell lines, the results of the CHO ␣ v /␣ 5 studies largely parallel those obtained with CHO ␣ 5 cells highlighting a loss of inhibitory efficacy toward ␣ 5 ␤ 1 -dependent cell adhesion when the Fn substrate contains an active synergy sequence within module III 9 . The results also suggest that the elegantin linker domain binding site on the integrin ␣ 5 ␤ 1 is also located within the N-terminal 232 residues of the ␣ 5 -subunit, because the linker domain inhibitory activity was not significantly altered in CHO ␣ v /␣ 5 cells.
The significance of the elegantin linker domain in promoting efficacious inhibition of CHO ␣ 5 cell adhesion on Fn was further supported by the employment in this experimental system of a structurally unrelated snake venom-derived integrin antagonist dendroaspin, which has no structural domain comparable to the linker domain of elegantin (14). This antagonist is a naturally occurring inhibitor of the integrin ␣ IIb ␤ 3 and has an identical RGD motif to kistrin rendering it unable to ligate low affinity integrin ␣ 5 ␤ 1 . In a previous study (28), we substituted the wild-type dendroaspin RGD motif with the elegantin RGD motif by site-directed mutagenesis to generate a variant dendroaspin (Ala 42 , Asn 46 ) with a capacity to engage low affinity ␣ 5 ␤ 1 . 6 In adhesion assays, dendroaspin (Ala 42 , Asn 46 ) showed very weak inhibitory efficacy toward CHO ␣ 5 adhesion on Fn III 6 -10 PHSRN (Fig. 7). In contrast, dendroaspin (Ala 42 , Asn 46 ) was a comparable inhibitor to the elegantin-kistrin chimeras at blocking CHO ␣ 5 adhesion upon Fn III 6 -10 SPSDN. This observation supports the contention that the potent inhibitory efficacy of elegantin toward ␣ 5 ␤ 1 -Fn-mediated adhesion is associated with its dual abrogation of the RGD and synergy site activities.
Elegantin Linker Domain Has Intrinsic Integrin ␣ 5 ␤ 1 Inhibitory Activity-Because the inhibitory activity of the linker domain depends upon the biological action of the Fn synergy site within the fragment Fn III 6 -10 (Figs. 4 and 6), the mechanism of inhibition of Fn binding to ␣ 5 ␤ 1 by elegantin could be based upon direct competition with the III 9 synergy domain. However, Takagi et al. (34) have proposed that the Fn III 9 synergy domain does not comprise an extended contact surface with the ␣ 5 -subunit but rather promotes the association of the III 10 RGD site with the integrin. If the linker domain of elegantin assumes a similar role, then the weak inhibitory activity of Eg P106 and Eg C-P106 in our adhesion experiments using Fn III 6 -10 PHSRN as a substrate could be accounted for by a reduced on-rate of the chimeras lacking the elegantin linker sequence thereby reducing the disintegrin competitiveness. We decided, therefore, to measure the comparative rates of association of the elegantin chimeras with CHO ␣ 5 cells. As shown in Fig. 8A, rather than having a reduced association, Eg P106 and Eg C-P106 binding to CHO ␣ 5 cells was considerably faster than both Eg WT and Eg C-WT. The data demonstrate that the presence of the linker domain in elegantin slows the interaction of the disintegrin with ␣ 5 ␤ 1 suggesting that it may participate in a multicontact binding mechanism. These data, therefore, rule out the possibility that the loss of function observed in elegantin-kistrin chimeras lacking the elegantin linker domain sequence is due to reduced on-rates leading to poorer competitiveness. These data are consistent with the contention that the elegantin linker domain is inhibitory toward the Fn synergy domain biological activity of enhancing Fn interaction with the integrin ␣ 5 ␤ 1 .
To test this hypothesis, we synthesized a cyclic peptide encompassing the entire linker domain sequence (amino acid 38 -47), including flanking cysteines that were cyclized through oxidation of their free thiol groups. We assessed the capacity of the linker domain peptide, designated P L , to block CHO ␣ 5 adhesion on Fn III 6 -10 PHSRN and Fn III 6 -10 SPSDN (Fig. 8, B and C). In these experiments, we observed that P L was unable to inhibit CHO ␣ 5 cell adhesion on Fn III 6 -10 PHSRN to any measurable extent at concentrations up to 1.0 mM peptide when the substrate was coated at 5.0 g/ml. Under the same experimental conditions, RGDS at similar peptide concentrations was also unable to block cell adhesion to any measurable extent (data not shown). This suggested that the coating density promoted an adhesion strength that could not be competed by both peptides under the assay method. We, therefore, reduced the coating density of the Fn III 6 -10 PHSRN substrate to 0.75 g/ml, which supported cell adhesion to ϳ30 -40% of that observed at 5.0 g/ml (data not shown). At this adhesion strength, both P L and RGDS blocked CHO ␣ 5 cell adhesion on Fn III 6 -10 PHSRN with similar efficacy with an approx maximal inhibition of 80% and IC 50 values of ϳ50 M (Fig. 8B). By comparison, P L strongly inhibited CHO ␣ 5 cell adhesion on Fn III 6 -10 SPSDN to a maximal level of 100% with an IC 50 of ϳ10 M even when the substrate coating was maintained at 5.0 g/ml (Fig. 8C). A scrambled cyclic peptide P C , comprising the same residues between the cysteines showed no measurable inhibitory activity to either Fn III 6 -10 PHSRN or Fn III 6 -10 SPSDN substrates. Under these experimental conditions RGDS showed approximately similar inhibitory potency with P L . These data, therefore, demonstrate that the elegantin linker domain can block integrin ␣ 5 ␤ 1 -Fn binding in the presence and absence of the synergy domain indicating that the linker sequence has intrinsic integrin ␣ 5 ␤ 1 -inhibitory activity.

DISCUSSION
This study has demonstrated that the medium-sized disintegrin elegantin, a potent RGD-based antagonist of the integrin ␣ 5 ␤ 1 , contains a region encompassing amino acids 38 -47 that contributes to its efficacy as an inhibitor of cell adhesion on Fn. Our data suggest that this region of elegantin binds to the ␣ 5 ␤ 1 complex and inhibits the interaction with Fn through a mechanism that antagonizes primarily the biological activity of the Fn synergy site located in module III 9 . This conclusion is based upon the following observations. Elegantin-kistrin chimeras lacking the elegantin linker domain but retaining the elegantin RGD loop with Ala 50 showed reduced inhibitory efficacy  toward CHO ␣ 5 cell adhesion on Fn comparable to the inhibitory efficacy of Pro 50 chimeras (kistrin RGD motif). This phenomenon, therefore, is not related to the recognition of RGD motif by the ␣ 5 ␤ 1 complex as with the case of the presence of Pro 50 in the RGD motif of the elegantin-kistrin chimeras (Fig.  4). Rather, replacement of the linker domain of elegantin with the corresponding sequence of kistrin reduced the inhibitory efficacy of the disintegrin toward CHO ␣ 5 cell adhesion on Fn III 6 -10 PHSRN as apposed to Fn III 6 -10 SPSDN indicating that the loss of inhibitory function is dependent upon an active Fn synergy site. It is also indirectly supported by the observation that a potent snake venom antagonist, dendroaspin (Ala 42 , Asn 46 ), containing the elegantin RGD motif but lacking a linker domain sequence, showed weak inhibitory activity under the same experimental conditions using Fn III 6 -10 PHSRN as a substrate but was a strong inhibitor of CHO ␣ 5 adhesion on Fn III 6 -10 SPSDN. Finally, a cyclic peptide corresponding to the linker region of elegantin inhibited CHO ␣ 5 cell adhesion upon Fn III 6 -10 PHSRN and Fn III 6 -10 SPSDN with comparable potency to the linear RGDS peptide, thus demonstrating intrinsic integrin inhibitory activity.
The binding mechanism of RGD ligands to integrin complexes has been the subject of intense study over the last decade. The biological significance of the role of the RGD IBM was established predominantly through biochemical and functional approaches, which have been consolidated and refined through the application of structural methods to study integrin-ligand binding. However, biochemical studies have also demonstrated the importance of additional ligand structures that support or "synergize" with the RGD ␤-loop. The most intensely studied synergy domain is the region present in the III 9 module of Fn with its core residues mapped as corresponding to the sequence PHSRN (18 -21). These studies demonstrated the pivotal role for the Fn synergy region in promoting ␣ 5 ␤ 1 -dependent cell adhesion. The significance of the Fn synergy sequence to the biological activity of Fn has been confirmed by several additional studies. For example, Sechler et al. (29) showed that the synergy site in Fn was essential for the formation of an ␣ 5 ␤ 1driven accumulation of an Fn fibrillar matrix. Mesendoderm extension was also shown to be dependent on the activity of the Fn III 9 synergy site during Xenopus gastrulation (30). The synergy interaction of Fn with ␣ 5 ␤ 1 is also thought to promote nodule initiation during bone morphogenesis (31). The latter finding was supported by the recent observation that co-localization of the RGD and PHSRN peptide sequences within a synthetic matrix enhanced osteoblast adhesion, spreading, and focal adhesion formation compared with RGD alone (32).
Whereas the important role of the Fn synergy region is well established through biochemical and functional studies, the mechanism of its action in the ligand binding process remains controversial. Mould et al. (26) showed that, in a structural model of a molecular complex of the truncated head region of  Fig. 6 (n ϭ 3). Inset, an expanded abscissa displaying ligand association during over 2 min. B and C, CHO ␣ 5 cells (1 ϫ 10 6 /ml) were preincubated with varying concentrations of the synthetic peptides as shown for 30 min prior to addition to microtiter wells coated with either Fn III 6 -10 PHSRN (B) or Fn III 6 -10 SPSDN (C). After 1 h at room temperature, adherent cells were quantified by colorimetric measurement of endogenous acid phosphatase. Background cell adhesion was determined by incubation of a sample of the cell suspension in the presence of 10 mM EDTA. Results are expressed as percent inhibition. All data points were performed in quadruplicate Ϯ S.E. Data shown are for a representative experiment (n ϭ 3). ␣ 5 ␤ 1 with Fn III 6 -10 obtained by x-ray scattering, the III 9 domain interacts with the ␤-propeller domain of the ␣ 5 -subunit, consistent with previous work, which mapped the synergy interaction site to propeller blades 2 and 3. Furthermore, substitution of residues Tyr 204 3 Ala and Ile 210 3 Ala located in a loop within blade 3 reduced the affinity of the interaction between ␣ 5 ␤ 1 and Fn III 6 -10 by 5-and 200-fold, respectively, while having little affect upon the interaction of the integrin with Fn III 6 -10 SPSDN. These observations support a model in which the Fn synergy site forms part an extended contact surface between the integrin ␣ 5 ␤ 1 and Fn III 9 -10 . Support for this notion was extended using electron microscopy to study a complex comprising the entire ectodomain of the integrin ␣ v ␤ 3 in complex with Fn III [7][8][9][10] . In this molecular complex, the Fn synergy region was also positioned in close proximity to the surface of the ␣ v -subunit ␤-propeller domain (33). However, Takagi et al. (34) observed that in complexes of truncated ␣ 5 ␤ 1 with Fn 7-10, no extensive contact between Fn III 9 and the ␣ 5 -subunit was present. These workers suggested that the synergy site in Fn promotes cell adhesion by promoting the association phase of the ligand binding process either by optimization of the RGD loop conformation or by electrostatic steering.
The subtle relationship between the RGD and synergy sites of Fn for acquiring full ␣ 5 ␤ 1 -dependent adhesive activity have been demonstrated. Altroff and co-workers (35) showed that targeted mutagenesis of residues within Fn III 9 , which effect domain stability, and at the interface between III 9 and III 10 that alter the angle of interdomain tilt, abrogate ␣ 5 ␤ 1dependent adhesion to levels to those observed for Fn synergy domain mutants (36).
The precise interaction site for the elegantin linker domain on the integrin ␣ 5 ␤ 1 was not identified in the present work. However, our data suggest that the site of interaction is located within amino acids 1-232 of the ␣ 5 -subunit, a region that has been previously identified as harboring structural elements conferring ␣ 5 ␤ 1 receptor specificity and the putative contact region for the Fn synergy site (26). Furthermore, our data suggest that the linker domain of elegantin does not promote the association phase of the binding mechanism but rather slows the interaction considerably. This observation suggests that binding of elegantin to the ␣ 5 ␤ 1 integrin involves a multicontact-mutlistep interaction mechanism giving rise to greater inhibitory specificity toward the physiological ligand Fn.
The precise mechanism by which the linker domain of elegantin down-regulates ␣ 5 ␤ 1 -dependent adhesion remains unknown and requires further investigation. However, our data suggest that the linker region in elegantin, a highly basic domain, negatively regulates the synergy site interaction on the integrin ␣ 5 ␤ 1 either directly or allosterically. Under the experimental conditions employed in this study, both the linker domain peptide P L and the RGDS peptide were comparable inhibitors of CHO ␣ 5 cell adhesion on either Fn III 6 -10 PSHRN or Fn III 6 -10 SPSDN substrates, although significantly greater concentrations of peptide were required to block adhesion on Fn III 6 -10 PSHRN. Because the linker domain cyclic peptide blocked CHO ␣ 5 adhesion on Fn III 6 -10 SPSDN, it appears to show a capacity to negatively modulate the interaction of the ␣ 5 ␤ 1 interaction with the RGD loop of Fn III 10 . The potency of P L as an integrin inhibitory molecule is striking considering that it plays an ancillary or synergistic role to the RGD ␤-loop. Furthermore, it is highly unlikely that P L assumes the native, extended conformation of the linker domain observed for the analogous region of the related disintegrin flavoviridin and as predicted in our model of elegantin as a consequence of the interdomain disulfide and hydrogen bonding that is present in the parent disintegrin (Fig. 1).
Structural studies have shown that the RGD tri-peptide binds to a pocket at the interface of the ␣and ␤-subunits of the integrin heterodimer within the globular head region (7,8). Binding of the RGD sequence and chemical mimetics invariably involves coordination of the metal ion-dependent adhesion site Mg 2ϩ ion by a carboxyl group adjacent to a basic moiety or amino acid. It is highly unlikely, therefore, that the linker domain peptide (CRFKKKRTIC) behaves as a RGD mimetic due to the absence of a free carboxyl group in the peptide. However, because the linker peptide P L was able to inhibit CHO ␣ 5 cell adhesion on Fn III 6 -10 SPSDN, a solely RGD-dependent event, competition between the two adhesive peptides appears to occur. Our favored model, therefore, is that the linker domain most likely interacts with the ␣ 5 -subunit and acts as an inhibitor of Fn binding by interfering with the synergy site interaction directly and RGD interaction site allosterically. However, it is possible that the linker peptide binds to a site on the ␣ 5 -subunit that blocks both Fn synergy and RGD interactions simultaneously by interfering, in the latter case, with the predicted hydrogen bonding between the guanidinium moiety of arginine and an acidic residue in the ␣ 5 -subunit. In support of the view that peptide P L binds to the ␣ 5 -subunit, we observed that the pattern of inhibition of CHO ␣ v /␣ 5 cells was essentially similar to CHO ␣ 5 cells. Furthermore, Mould et al. (37) showed that mutations in the predicted region of the ␣ 5 -subunit reduced the affinity of ␣ 5 ␤ 1 for Fn III 6 -10 SPSDN, suggesting that an allosteric relationship may exist between the synergy and RGD interaction sites on the ␣ 5 ␤ 1 complex. Through construction of a molecular model of the ␣ 5 -subunit built upon the crystal structure of the ␣ v -subunit, we have identified a highly acidic face in the ␣ 5 -subunit ␤-propeller domain located at the interface with the ␤ 1 -subunit that could potentially form electrostatic bonds with the highly basic linker domain of elegantin (data not shown). Further studies are needed to identify the precise location of the elegantin linker domain interaction site on the ␣ 5 ␤ 1 complex, which would facilitate the design of experiments to delineate the distinctive mechanisms of action of the linker region and RGD sequence.
The present work has identified a domain within elegantin with intrinsic integrin ␣ 5 ␤ 1 inhibitory activity. This widens the number of disintegrin-derived sequences known to interact and regulate integrin function. Previous work using the barbourin IBM as a template for the construction of therapeutic antagonists of the integrin ␣ IIb ␤ 3 highlighted the importance of disintegrins as model integrin ligands (10). The discovery of the biological activity of the elegantin linker domain could, therefore, offer further insight for the development of novel integrin antagonists.