An RGD to LDV Motif Conversion within the Disintegrin Kistrin Generates an Integrin Antagonist That Retains Potency but Exhibits Altered Receptor Specificity

Integrin ligands almost invariably employ a variant of either the RGD or LDV motif as a key element of their receptor recognition site. These short acidic peptide sequences collaborate with specific nonhomologous flanking residues and spatially separate “synergy” sequences to determine receptor binding specificity. Although the consensus sequences for RGD and LDV motifs are quite different, their common use suggests that they might share a critical role in receptor-ligand engagement. To date, the effects of interconversion of the two motifs within a natural protein framework have not been tested; however, in this study, we have converted the natural RGD site found in the snake venom disintegrin kistrin into an LDV motif and examined the effects of the change on the specificity of integrin recognition and on disintegrin potency. While an assessment of receptor binding using cell adhesion and purified integrin solid-phase assays demonstrated recognition of recombinant RGD kistrin by αVβ3 and α5β1, a series of LDV kistrin chimeras did not bind to these integrins, but instead were recognized specifically by α4β1. The minimal change to elicit this distinct switch in receptor specificity was found to involve alteration of only three residues within kistrin. Alanine scanning mutagenesis was used to provide further information on the functional contribution of the three residues. More important, the LDV kistrin chimeras also retained much of the characteristic potency of RGD kistrin, indicating that the kistrin scaffold is optimized for presentation of both RGD and LDV sequences. These findings provide evidence for similarities in motif pharmacophore and reinforce the hypothesis that RGD and LDV sites have an equivalent functional role in receptor binding. They also demonstrate the potential for other disintegrin-containing proteins, perhaps from the ADAM family, to employ LDV sequences for integrin binding.

Integrin ligands almost invariably employ a variant of either the RGD or LDV motif as a key element of their receptor recognition site. These short acidic peptide sequences collaborate with specific nonhomologous flanking residues and spatially separate "synergy" sequences to determine receptor binding specificity. Although the consensus sequences for RGD and LDV motifs are quite different, their common use suggests that they might share a critical role in receptor-ligand engagement. To date, the effects of interconversion of the two motifs within a natural protein framework have not been tested; however, in this study, we have converted the natural RGD site found in the snake venom disintegrin kistrin into an LDV motif and examined the effects of the change on the specificity of integrin recognition and on disintegrin potency. While an assessment of receptor binding using cell adhesion and purified integrin solidphase assays demonstrated recognition of recombinant RGD kistrin by ␣V␤3 and ␣5␤1, a series of LDV kistrin chimeras did not bind to these integrins, but instead were recognized specifically by ␣4␤1. The minimal change to elicit this distinct switch in receptor specificity was found to involve alteration of only three residues within kistrin. Alanine scanning mutagenesis was used to provide further information on the functional contribution of the three residues. More important, the LDV kistrin chimeras also retained much of the characteristic potency of RGD kistrin, indicating that the kistrin scaffold is optimized for presentation of both RGD and LDV sequences. These findings provide evidence for similarities in motif pharmacophore and reinforce the hypothesis that RGD and LDV sites have an equivalent functional role in receptor binding. They also demonstrate the potential for other disintegrin-containing proteins, perhaps from the ADAM family, to employ LDV sequences for integrin binding.
Cell-surface adhesion receptors mediate the cell-cell and cellmatrix interactions that are fundamental to normal cell behavior and tissue organization. The integrins, a superfamily of structurally related ␣␤ heterodimers, represent a major class of adhesion receptors (for reviews, see Refs. [1][2][3][4]. Many studies of the molecular basis of integrin-ligand binding have now been performed, and frequently, integrins have been found to recognize short peptide motifs within their ligands. The RGD motif, which contains a critical aspartate residue, was first identified in the central cell-binding domain of fibronectin (5-7), but has subsequently been shown to be functional in other molecules, including vitronectin and von Willebrand factor. The molecular mechanisms involved in the recognition of the RGD sequence are complex since it binds to a number of different integrins, and receptor specificity varies between different matrix molecules. A second peptide motif, LDV, present within the alternatively spliced CS1 sequence of the IIICS region of fibronectin, also contains a crucial aspartate and is primarily recognized by integrin ␣4␤1 (8 -10). The integrin-binding members of the immunoglobulin superfamily possess motifs that are homologous to LDV, including the tetrapeptide sequence IDSP in domains 1 and 4 of VCAM-1 (11)(12)(13)(14) and glutamate-containing sequences in the membranedistal domains of intercellular adhesion molecules (15)(16)(17). A third aspartate-dependent motif, which contains QAGDV as the minimal active sequence, is found near the C terminus of the ␥ chain of human fibrinogen and is recognized by platelet integrin ␣IIb␤3 (18).
The active-site motifs of integrin ligands can be reproduced synthetically in the form of peptides, and these reagents have been instrumental in providing information on the specificities of receptor-ligand binding and for the development of therapeutic agents for diseases involving aberrant adhesion (19). In addition, the insertion of peptide motifs into nonadhesive protein scaffolds has previously been reported to generate adhesive activity. The RGD sequence has been incorporated into a long exposed loop within human lysozyme by site-directed mutagenesis, and the resultant mutant protein was able to promote cell adhesion (20). Insertion of RGD into the finger-like structure of hirudin produced a molecule capable of inhibiting platelet aggregation (21). Both the RGD-and LDV-containing CS1 sequences have been successfully grafted into staphylococcal protein A, with the creation of the cell adhesive activities of both motifs (22,23).
Although short sequence motifs such as RGD and LDV are critical for integrin recognition, they do not account for the entire binding event, as synergistic sequences have been identified that collaborate with them (24,25), and nonfunctional RGD and LDV sequences are also present in many molecules. The affinity and selectivity of linear peptides can be improved by the introduction of a conformational constraint, for example, by cyclization (7,26). Thus, differences in conformation, flank-ing residues, and the use of synergy sequences may all contribute to the binding specificities of different integrin-ligand combinations.
A conformationally constrained, functional RGD motif is also found within many disintegrins, a family of naturally occurring polypeptide integrin antagonists present in snake venoms (for reviews, see Refs. [27][28][29]. Disintegrins inhibit platelet aggregation by blocking the adhesive function of integrin ␣IIb␤3 (30), but are also able to disrupt the adhesive functions of other RGD-dependent integrins and are therefore relatively nonspecific (31)(32)(33). It is notable that disintegrins are up to 1000-fold more potent than linear RGD-containing peptides (34), probably because the RGD sequence exists in a favorable conformation at the apex of a long loop across the surface of the molecule (35).
The disintegrin molecule may be considered as a model scaffold for the presentation of an integrin recognition motif, and in this study, we have examined the effects of engineering the replacement of the RGD sequence in the disintegrin kistrin with LDV to assess whether it is possible to retain potency in an integrin antagonist while altering specificity. We have reported in preliminary form the design of synthetic kistrin cDNA and the expression of the molecule as a recombinant protein in which specific residues or segments can be mutated to create chimeric kistrins (36).
Cloning and Expression of Recombinant Kistrin-Kistrin cDNA was generated from complementary oligonucleotides designed from the protein sequence using Escherichia coli codon usage data. BamHI and EcoRI restriction enzyme sites were incorporated at the ends of the cDNA sequence to allow cloning into the phagemid vector pUC118. Internal BspMII, NarI, BglII, and KspI restriction sites were also incorporated to allow exchange of different segments of the molecule (Fig. 1). The complementary phosphorylated oligonucleotides were annealed and ligated into the dephosphorylated vector pUC118. Competent E. coli DH5␣FЈ cells were transformed with the ligation product, and the kistrin cDNA sequence was verified before subcloning into the pGEX2T expression vector (Pharmacia Biotech, Milton Keynes, UK) using the same restriction sites. DH5␣FЈ cells were transformed, and glutathione S-transferase-kistrin fusion proteins were induced and isolated as described (41). Briefly, a 40-ml overnight culture of transethyl]glycine; PBS, phosphate-buffered saline; BSA, bovine serum albumin; r-kistrin, recombinant kistrin. formed DH5␣FЈ cells was diluted 1:10 with fresh LB medium containing 50 g/ml ampicillin and cultured for 1 h at 37°C. Isopropyl-␤-D-thiogalactopyranoside was added to 0.1 M and cultured for a further 4 h. Cells were then centrifuged, resuspended in divalent cation-free Dulbecco's phosphate-buffered saline, and lysed by sonication. The extract was cleared by centrifugation and applied to a glutathione-agarose affinity column (Sigma, Poole, UK) pre-equilibrated with 150 mM NaCl and 10 mM Tris-HCl, pH 7.5. The column was washed with 150 mM NaCl and 10 mM Tris-HCl, pH 7.5, and the fusion protein was eluted with 5 mM reduced glutathione and 50 mM Tris-HCl, pH 7.5. The glutathione S-transferase carrier was removed by cleavage with human thrombin (Sigma) for 3 h at room temperature using an enzyme/substrate ratio of 1:500 (w/w). A second glutathione-agarose column was used to separate kistrin from GST, with the kistrin appearing in the column flowthrough fraction. Protein concentrations were measured using the BCA assay (Pierce, Chester, UK), and peak fractions were stored at Ϫ70°C.
Production of Mutant Kistrins-The segment of cDNA encoding the kistrin RGD loop was removed by digestion with BglII and KspI restriction endonucleases and replaced with a double-stranded insert encoding a mutant sequence. The insert was generated from two complementary oligonucleotides that were annealed, phosphorylated, and ligated into the dephosphorylated vector. Competent DH5␣FЈ cells were transformed with the ligation product, and the correct sequence was verified before expression of the mutant kistrins as described above.
Characterization of Recombinant Kistrins-The purity of the recombinant kistrins was assessed by discontinuous SDS-polyacrylamide gel electrophoresis using a Tris/Tricine buffer system (42). The fidelity of bacterial translation and thrombin cleavage were validated by N-terminal sequencing and mass spectroscopy.
N-terminal Sequencing-Prior to analysis, recombinant kistrin was dialyzed extensively at 4°C against 0.1% (v/v) trifluoroacetic acid. Approximately 1-nmol samples were sequenced by automated Edman degradation and high pressure liquid chromatography using a Model 476A Protein Sequencer (Applied Biosystems, Warrington, Cheshire, UK). The sequence obtained for recombinant kistrin was GSGKEX-DXS(G/S)PENPXXD, which is identical to that of authentic kistrin with the exception of the ambiguous position, which is actually Ser, and the initial Gly and Ser, which are derived from the GST fusion partner following thrombin cleavage. The four unidentified residues (X) are cysteines that are destroyed during sequencing.
Mass Spectrometry-Recombinant kistrins were analyzed by matrixassisted laser desorption time-of-flight mass spectrometry using a VG-TofSpec E spectrometer (Fisons Instruments, Manchester, UK). A mass of 7469 obtained for recombinant RGD kistrin was within 0.1% of the calculated mass of 7474.5 based on the amino acid sequence. For ILDV kistrin, the value of 7451 obtained was also within 0.1% of the calculated mass of 7474.5.
Solid-phase Receptor-Ligand Binding Assay-Assays were performed using a method based on that of Charo et al. (43). The 80-kDa fragment of fibronectin (500 g/ml in PBS) or the H/120 recombinant fragment of fibronectin (450 g/ml) was biotinylated by mixing with an equal mass of sulfo-N-hydroxysuccinimidobiotin (Pierce). After rotary mixing for 40 min at room temperature, excess biotin was removed from the mixture by dialyzing against several changes of 150 mM NaCl and 25 mM Tris-HCl, pH 7.4. A 96-well enzyme-linked immunosorbent assay plate (Immulon-3, Dynatech, Billingshurst, UK) was incubated overnight at room temperature with 100-l aliquots of purified integrins diluted with PBS to a concentration of ϳ5 g/ml. The wells were then blocked with 200 l of 5% (w/v) BSA, 150 mM NaCl, 0.05% (w/v) NaN 3 , and 25 mM Tris-HCl, pH 7.4, for 2 h at room temperature and washed three times with 200 l of 150 mM NaCl, 5 mM MnCl 2 , and 25 mM Tris-HCl, pH 7.4, containing 1 mg/ml BSA (buffer A). 100-l aliquots of biotinylated H/120 (ϳ0.2 g/ml), diluted with buffer A, with or without recombinant kistrins, were then added to the wells and incubated at 30°C for 3 h. Unbound biotinylated ligand was removed, and the wells were washed three times with buffer A. Bound ligand was quantitated by addition of 100 l of ExtrAvidin-peroxidase conjugate (Sigma) diluted 1:200 in buffer A for 10 min at room temperature, following which the wells were washed four times with buffer A. The color was developed by addition of 2,2Ј-azinobis(3-ethylbenzthiazoline-6-sulfonic acid) (Sigma), and the absorbance was measured at 405 nm. Nonspecific binding was determined from the level of ligand binding to BSA alone, and these values were subtracted from those obtained with integrincoated wells. Each point represented the mean Ϯ S.D. of four replicate wells.
Cell Spreading Assay-The A375-SM human melanoma cell line (provided by I. J. Fidler, M. D. Anderson Hospital and University of Texas, Houston, TX) was maintained as described (44) in Eagle's min-imal essential medium supplemented with 10% (v/v) fetal bovine serum, minimal essential medium vitamins, nonessential amino acids, 2 mM L-glutamine, and 1 mM sodium pyruvate. HT-1080 human fibrosarcoma cells were obtained from the European Collection of Animal Cell Cultures (Porton Down, UK) and cultured in Dulbecco's modified Eagle's medium containing 0.11 g/liter sodium pyruvate, 10% (v/v) fetal bovine serum, and 2 mM L-glutamine. Assays were performed as described by Humphries et al. (45) using 96-well tissue culture plates (Costar, High Wycombe, Bucks, UK). Ligands were diluted in Dulbecco's phosphatebuffered saline (PBS), and 100-l aliquots were used to coat the wells for 1 h at room temperature. The ligand solution was then removed, and sites that could support nonspecific cell adhesion were blocked with 100 l of 10 mg/ml heat-denatured BSA (45). After incubation for 30 min at room temperature, the BSA was removed, and 50-l aliquots of inhibitors in PBS were added. Cells were detached using 0.05% (w/v) trypsin and 0.02% (w/v) EDTA and then washed and resuspended to 2 ϫ 10 5 cells/ml in Dulbecco's minimal essential medium. Cell suspensions were allowed to recover at 37°C for 10 min before addition of 50-l aliquots to the wells. Following incubation at 37°C for 90 min in a humidified atmosphere of 6% CO 2 , cells were fixed with 5% (w/v) glutaraldehyde in PBS. Using phase-contrast microscopy, the percentage of cells with a flattened, phase-dark morphology was estimated. Each point was obtained by counting 3 ϫ 100 cells in random fields. No cell spreading was observed on wells coated only with heat-denatured BSA.

RESULTS
The disintegrin kistrin was selected for use as a scaffold for the insertion of the LDV motif because the sequence of the RGD loop shows some homology to the LDV-containing CS1 sequence in fibronectin. In both sequences, the critical aspartate is immediately followed by a residue with a hydrophobic side chain, and a proline residue is present on either side of the motif: kistrin RGD loop sequence, RIPRGDMPDDR; and CS1 sequence, PEILDVPSTV. Synthetic kistrin cDNA was generated from overlapping oligonucleotides and expressed as a recombinant protein in E. coli. Samples of recombinant kistrin (r-kistrin) from different stages of purification were analyzed by SDS-polyacrylamide gel electrophoresis (Fig. 2). The r-kistrin was resolved as a slightly diffuse band of ϳ7.5 kDa under both reducing and nonreducing conditions, which corresponds with its expected molecular mass. The band at 26 kDa was the major contaminant, but was removed by repeating the glutathione-agarose affinity chromatography and therefore appeared to be GST (data not shown).
Following the expression of kistrin as a recombinant protein, the activity of r-kistrin was compared with that of native kistrin isolated from venom. The ability of both molecules to inhibit the spreading of A375-SM cells on the RGD-containing 80-kDa fragment of fibronectin, which is recognized by inte- grins ␣5␤1 and ␣V␤3, was examined (Fig. 3). Half-maximal inhibition of cell spreading was obtained at 2 M for native kistrin and at 7 M for the recombinant molecule. When compared with the published value of 100 M for the linear peptide GRGDS (34), these data demonstrate that both native and recombinant kistrins are potent integrin antagonists; they also suggest that the recombinant molecules are largely correctly folded. The biological activity of disintegrins appears to be dependent on correct disulfide bond-mediated folding, as suggested by the dramatic loss of activity following reduction and alkylation of the disulfide bonds, but nevertheless, E. coli appears to be a suitable host for their recombinant expression (31,46).
Having established the anti-adhesive activity of recombinant kistrin, the synthetic cDNA was used as a template for the creation of chimeras with different integrin recognition motifs. In the initial chimera targeting the RGD loop, the entire 11residue loop sequence was removed by excision of the segment of cDNA encoding this region and replaced with a segment encoding an 11-amino acid sequence from the C-terminal end of the CS1 peptide of fibronectin. A second chimera in which the RGDM tetrapeptide sequence from the loop was replaced by ILDV was also generated, the minimal change needed to convert RGD to LDV. In further mutagenesis, the contribution of individual residues to the activity of the molecule was investigated by replacement with alanine. The active-site aspartate was retained in the same position as in the native kistrin sequence in all mutants. The amino acid sequences of the RGD loop were as follows: r-kistrin, CRIPRGDMPDDRC; CS1 kistrin, CGPEILDVPSTVC; ILDV kistrin, CRIPILDVPDDRC; ALDV kistrin, CRIPALDVPDDRC; IADV kistrin, CRIPIADVP-DDRC; ILAV kistrin, CRIPILAVPDDRC; and AAAV kistrin, CRIPAAAVPDDRC.

Promotion of Cell Spreading by Recombinant Kistrin Fusion
Proteins-To compare the abilities of mutant kistrins to promote cell spreading, they were purified as fusion proteins retaining the GST carrier and immobilized on the surface of 96-well plates. A high maximal level (Ͼ95%) of HT-1080 cell spreading was observed on GST-r-kistrin containing the native sequence, with half-maximal spreading at a coating concentra-tion of 1.1 g/ml. HT-1080 cell spreading was not supported by any of the mutants (Fig. 4 and data not shown).
Since A375-SM cells express integrin ␣4␤1 and recognize the ILDV motif within the CS1 sequence of fibronectin, spreading assays were performed to determine whether these cells were able to recognize the ILDV motif in kistrin chimeras (Fig. 5). Dose-dependent promotion of spreading was observed, but activity varied between the different mutant kistrins. GST-rkistrin supported up to 98% spreading, with half-maximal spreading observed at a coating concentration of 0.7 g/ml. Half-maximal spreading was supported at 7.4 g/ml by the ILDV kistrin chimera and at 19 g/ml by the ALDV kistrin mutant. Cell spreading on either the IADV kistrin or CS1 kistrin fusion proteins reached a maximum of 20% at ϳ200 g/ml (data not shown). No cell spreading was observed on GST alone or on the two mutants AAAV kistrin and ILAV kistrin. The results of the cell spreading assay were mirrored by A375-SM cell attachment assays (data not shown).
Effect of Anti-integrin Antibodies on Cell Spreading on Fusion Proteins-To investigate which receptors were used by the HT-1080 and A375-SM cells to spread on the kistrin chimeras, the effects of function-blocking anti-integrin antibodies were tested. The effects of anti-integrin antibodies on the spreading of HT-1080 cells on RGD kistrin are shown in Fig. 6. Spreading was partially inhibited by function-blocking anti-␣5 (33%), anti-␣V (43%), anti-␤1 (46%), and anti-␣V␤3 (40%) antibodies, but not by anti-␣4 or anti-␣2 antibodies. Substantial inhibition was observed when anti-␤1 and anti-␣V␤3 (97%) antibodies or when anti-␣5 and anti-␣V (95%) antibodies were used together. A combination of anti-␣5 and anti-␣V␤3 antibodies gave ϳ77% inhibition of spreading. These data suggest that the receptors used by HT-1080 cells to recognize r-kistrin are integrins ␣V␤3, ␣5␤1, and possibly ␣V␤1.
Inhibition of Integrin-Ligand Binding by Recombinant Kistrins in Solid-phase Assay-To measure the potency of recombinant kistrins and to exclude the possibility that the mutant kistrins were exerting their effects indirectly, for example, via another cell-surface receptor, the interactions of the mutant kistrins with purified integrins in solid-phase assays were examined. RGD kistrin was found to be a potent inhibitor of the binding of the RGD-containing 80-kDa fibronectin fragment to immobilized ␣5␤1 (Fig. 8), with half-maximal inhibition being observed at 0.01 M. None of the mutant chimeras showed any inhibitory activity when tested at concentrations of up to 1.2 M (Fig. 8 and data not shown).
ILDV kistrin was the most potent inhibitor of the binding of the LDV-containing H/120 fibronectin fragment to immobilized ␣4␤1 (with half-maximal inhibition at ϳ0.1 M). The relative activities of the alanine replacement mutants followed the same pattern as for the promotion of melanoma cell spreading and attachment, with half-maximal inhibition observed at ϳ0.12 M for ALDV kistrin and at 0.79 M for IADV kistrin (Fig. 9A). Half-maximal inhibition was not achieved with either AAAV kistrin or ILAV kistrin at concentrations up to 3.6 M. The ability of the chimeras to block the function of purified ␣4␤1 therefore correlated with the ability to promote ␣4␤1-dependent cell spreading.
RGD kistrin and CS1 kistrin were less potent antagonists of ␣4␤1 than ILDV kistrin (Fig. 9B). Half-maximal inhibition was not achieved at concentrations up to 3.5 M for CS1 kistrin, but was observed at 2.7 M for RGD kistrin. The different potencies of the recombinant kistrins in Fig. 9 are likely to be a reflection of their relative affinities for the integrin receptors, but it is also conceivable that this might be attributed to selective inhibition of different integrin activation states. Although Mn 2ϩ was included in all of the solid-phase assays to maximize integrin activation, a proportion of the integrin population may nevertheless have remained in an inactive conformation. DISCUSSION The aims of this study were to express kistrin (and kistrin chimeras) in recombinant form and to examine for the first time the effects of interconverting integrin recognition motifs within a natural framework on receptor binding potency and specificity. Our major findings are as follows. (i) Recombinant and native kistrins exhibit similar potency and specificity, confirming previous studies that have reported successful bacterial expression of disintegrins. (ii) A chimeric kistrin containing a minimal RGD to LDV motif change exhibits a distinct switch in receptor specificity: RGD kistrin recognizes ␣5␤1 and ␣V␤3, whereas ILDV kistrin binds specifically to ␣4␤1. (iii) The ILDV kistrin chimera retains much of the potency of the parent RGD kistrin, suggesting that the structural features required for a potent RGD activity also apply to the LDV motif. This provides further evidence that the RGD and LDV motifs may be functionally equivalent. (iv) Alanine replacement mutants within the LDV motif have provided information on the relative contributions of individual flanking residues to integrin binding activity.
The receptor binding specificity of recombinant kistrins was assessed both by solid-phase assays and by inhibition of cell spreading by anti-integrin antibodies. Spreading of both HT-1080 and A375-SM cells on r-kistrin was inhibited by anti-␣V␤3, anti-␣V, anti-␤1, and anti-␣5 antibodies, and r-kistrin was able to bind to purified ␣5␤1 in solid-phase assays. These data suggest that r-kistrin is recognized by integrin ␣V␤3 and with a lower affinity by ␣5␤1 and also possibly by ␣V␤1. Previous studies have suggested that the hydrophobic methionine residue immediately C-terminal to RGD in kistrin is responsible for a reduced affinity for ␣5␤1 and ␣V␤3 compared with ␣IIb␤3 (47, 48); however, ␣V␤3 seems to be tolerant of a more diverse flanking region than ␣5␤1 (49), which may explain the significant contribution of ␣V␤3 to the adhesion of both A375-SM and HT-1080 cells to r-kistrin.
The ILDV kistrin chimera was unable to bind to purified ␣5␤1, but bound to ␣4␤1, and cell spreading on ILDV kistrin was inhibited only by anti-␣4 and anti-␤1 antibodies. These results suggest that ILDV kistrin is recognized by ␣4␤1, but not by ␣V␤3 or ␣5␤1, indicating that alteration of only three The coating concentration of GST-kistrin was 10 g/ml. C represents spreading in the absence of antibody. Antibodies were used at 10 g/ml, with the exception of HP2/1, which was used at 20 g/ml, and 17E6, which was used at 5 g/ml. Error bars indicate S.D.

FIG. 7
residues is sufficient to achieve a major change in specificity. The results of the alanine replacement mutagenesis reflect the dependence on the aspartate residue for the activity of ILDV kistrin and the relative importance of the isoleucine and leucine residues, the requirement for the leucine being greater than for isoleucine.
Although the ILDV kistrin chimera is potent, the CS1 kistrin chimera is considerably less active, although both share the ILDVP pentapeptide sequence at the same position in the loop. The differences in sequence between the two lie in the flanking sequences RIP, immediately N-terminal, and DDR, immediately C-terminal to ILDVP, and it is likely that they determine the potency of the kistrin molecule. Further mutagenesis will reveal the contribution of these sequences to the activity of ILDV kistrin. It has been suggested that the RGD tripeptide is chiefly responsible for the activity of kistrin (50), but there is also evidence that the surrounding amino acids contribute to the affinity and selectivity of disintegrins (51)(52)(53). It may also be possible that the residues in the loop contribute differently to the binding of ␣4␤1 to ILDV kistrin than to the binding interactions between ␣V␤3 and r-kistrin. For example, the proline residues lying N-and C-terminal to the RGD motif in kistrin are not present in the known RGD loop sequences of most other disintegrins, and alanine replacement mutagenesis suggests that they are not critical to the potency of kistrin (50), but may contribute to the activity of ILDV kistrin.
The kistrin RGD loop is conformationally flexible but more constrained than a linear peptide, possibly contributing to the potency of the molecule. The concentration of the linear CS1 peptide required to give half-maximal inhibition of H/120 binding to ␣4␤1 in a solid-phase assay has been quoted as 4 g/ml (1.5 M) (39) and is therefore significantly more active than the CS1 kistrin chimera. The 11-residue sequence inserted into the kistrin RGD loop may be unfavorably constrained, resulting in low affinity interactions with ␣4␤1. Clearly, specific residues or sequences must be present within the loop to produce a potent integrin inhibitor.
Our finding that it is possible to exchange RGD and LDV motifs (by exchanging only three residues) in an integrin ligand and alter specificity but retain activity suggests a functional equivalence between the two motifs. Earlier evidence for this was obtained from cross-inhibition studies in which short linear RGD peptides were shown to inhibit ␣4␤1-LDV interactions competitively (34), and certain cyclic RGD peptides have been produced that were also able to block ␣4␤1 (54,55). This may explain our finding that r-kistrin was able to inhibit the adhesive activity of purified ␣4␤1 in the solid-phase assay. The QAGDV motif is also thought to be functionally equivalent to RGD since QAGDV-and RGD-containing peptides share common or mutually exclusive binding sites on ␣IIb␤3 (56).
The aspartate residue critical to the activity of all three motifs may be central to a common integrin-binding mechanism used by these sequences. Integrin-ligand binding and receptor activation are dependent on divalent cations, and ligand-binding sites within integrins coincide with putative cation-binding sites (57,58). It has been proposed that a divalent cation could be coordinated simultaneously by the ligand motif and the active site within the integrin, with the acidic aspartate side chain providing a cation-coordinating group (59,60). Subtle differences in motif conformation between ligands and the use of synergy sequences may provide the basis for receptor specificity. In the future, it may be possible to exploit kistrin (or other disintegrin) scaffolds to display different integrin recognition sequences and to generate potent inhibitors that can be used in vivo as probes of adhesive function or for structural analysis.
Finally, it is conceivable that other disintegrins might naturally express an LDV motif. A gene family known as ADAM (for proteins containing a disintegrin and metalloproteinase domain) encodes a family of membrane-anchored proteins in mammals that may play roles in cell-cell interactions (61,62). Whereas almost all snake venom disintegrins contain an RGD active site, the only ADAM protein that contains RGD in its disintegrin domain that has been identified to date is the molecule metargidin (63). Other ADAM proteins lack this sequence and instead have a motif that includes highly conserved cysteine and aspartate residues and that bears some homology to an LDV motif. As yet, the integrin binding specificity of ADAM proteins and the role of this active site sequence are poorly understood, but it is conceivable that they may be natural LDV forms of disintegrins.