Design of potent and specific integrin antagonists. Peptide antagonists with high specificity for glycoprotein IIb-IIIa.

Members of the snake venon-derived, "disintegrin" peptide family containing the Arg-Gly-Asp (RGD) amino acid sequence are among the most potent inhibitors of the binding of adhesive proteins to platelet glycoprotein (GP) IIb-IIIa. However, GPIIb-IIIa antagonists containing the RGD sequence are not integrin specific and inhibit the adhesive functions of many other RGD-dependent integrins. The single disintegrin peptide, barbourin, containing a conservative amino acid substitution of Lys (K) for Arg (R) in the RGD sequence, is however, highly specific for GPIIb-IIIa. Using this information we have tested the hypothesis that both structural and conformational elements of barbourin are important for its high affinity and selectivity for platelet GPIIb-IIIa by synthesizing a series of conformationally constrained, disulfide-bridged peptides containing the KGD amino acid sequence. Incorporation of the KGD sequence into a cyclic peptide template, followed by systematic optimization of the cyclic ring size, optimization of secondary hydrophobic binding site interactions, and the derivatization of the lysyl side chain functionality of the KGD sequence has resulted in peptide analogs which display inhibitory potency and GPIIb-IIIa selectivity comparable to that of barbourin. This study demonstrates that the specificity and potency of the disintegrin family of antagonists, in particular barbourin, can be mimicked by small, conformationally restrained peptides.


Design of Potent and Specific Integrin Antagonists
Members of the snake venon-derived, "disintegrin" peptide family containing the Arg-Gly-Asp (RGD) amino acid sequence are among the most potent inhibitors of the binding of adhesive proteins to platelet glycoprotein (GP) IIb-IIIa. However, GPIIb-IIIa antagonists containing the RGD sequence are not integrin specific and inhibit the adhesive functions of many other RGD-dependent integrins. The single disintegrin peptide, barbourin, containing a conservative amino acid substitution of Lys (K) for Arg (R) in the RGD sequence, is however, highly specific for GPIIb-IIIa.
Using this information we have tested the hypothesis that both structural and conformational elements of barbourin are important for its high affinity and selectivity for platelet GPIIb-IIIa by synthesizing a series of conformationally constrained, disulfide-bridged peptides containing the KGD amino acid sequence. Incorporation of the KGD sequence into a cyclic peptide template, followed by systematic optimization of the cyclic ring size, optimization of secondary hydrophobic binding site interactions, and the derivatization of the lysyl side chain functionality of the KGD sequence has resulted in peptide analogs which display inhibitory potency and GPIIb-IIIa selectivity comparable to that of barbourin. This study demonstrates that the specificity and potency of the disintegrin family of antagonists, in particular barbourin, can be mimicked by small, conformationally restrained peptides.
Disintegrins are a family of homologous peptides derived from viper venoms that inhibit the binding functions of integrins when present in low nanomolar concentrations (1). All but one member of this family contains the Arg-Gly-Asp (RGD) recognition sequence, which appears to mediate the binding of these antagonists to the integrins. Disintegrins inhibit the adhesive interactions of multiple integrins, as has been previously shown for small RGD-containing peptides (1)(2)(3). Although the three-dimensional structure of several disintegrin peptides have been deduced through two-dimensional nuclear magnetic resonance spectroscopy in solution, the RGD-containing loops appear to contain no regular secondary structure and the basis of their potency is still speculative (4, 5 ) . It has been postulated that the conformation of * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom all correspondence should be addressed COR Therapeutics, Inc., 256 E. Grand Ave., Suite 80, South San Francisco, CA 94080. § Present. address: Gladstone Institute of Cardiovascular Disease, P. 0. Box 40608, San Francisco, CA 94140. the RGD amino acid sequence within the disintegrins accounts for the fact that they are up to 1000 times more potent than linear RGD-containing peptides. Consistent with this hypothesis are the observations that these snake venom derived peptides are rich in disulfide bridges and that their inhibitory activity is greatly diminished upon disulfide reduction (6, 7).
We have recently described the isolation and characterization of the novel disintegrin, barbourin, which is a 73-aminoacid, disulfide-rich peptide that is highly homologous to other members of the disintegrin family, but is unique in that it contains the KGD rather than the RGD recognition sequence (8). We have determined that this single Lys for Arg substitution is the sole structural feature responsible for the specificity of barbourin for glycoprotein (GP)' IIb-IIIa (8). It was not known, however, whether small peptides containing KGD sequences would also be specific for GPIIb-IIIa.
In this study we have found that small, conformationally constrained peptides containing the KGD sequence are potent and specific inhibitors of platelet GPIIb-IIIa versus other closely related integrins. Using synthetic peptide templates which incorporate the KGD sequence into conformationally restricted cyclic peptides, we have found that an optimal display of this sequence results in analogs with both high affinity and selectivity for GPIIb-IIIa. In conjunction with our previous observations with a series of disintegrins from American pit viper venoms (9), systematic evaluation of additional residues surrounding the KGD sequence has also revealed secondary binding sites which could be optimized with hydrophobic amino acid residues to increase the apparent affinity for GPIIb-IIIa. Finally, derivatization of the Lys side chain in the peptide antagonists has been found to enhance the affinity of inhibitors for GPIIb-IIIa without compromising their integrin specificity. The more potent analogs in this study display affinity and specificity of small peptide inhibitors for GPIIb-IIIa which are virtually identical to that of barbourin and are excellent candidates for evaluation as acute antithrombotic agents.
Peptide Synthesis-All synthesis were performed on the 0.5 mmol scale starting with 4-methylbenzhydrylamine resin (0.6-0.7 meq/g of resin loading). The syntheses were carried out in a stepwise manner using symmetrical anhydride coupling in dimethylformamide employing the 430.4 synthesizer or with hydroxybenzotriazole active esters in N-methylpyrolidinone employing the 431A peptide synthesizer. Side chain protection was arginine (Tos), cysteine (4-methylbenzyl), lysine (Cl-CBZ), aspartic acid (OcHex), glutamic acid (OcHex), and histidine (im-Boc). After addition of the amino-terminal residue in each synthesis, the Boc group was removed with trifluoroacetic acid and the peptide resin dried.
H F Cleauage and Cyclization-For cyclic peptides, the assembled peptide resins (1.0 g) were suspended in HF (10 ml/g resin) containing 10% by volume anisole and 2% methylethylsulfide employing a type IB HF cleavage apparatus (Protein Research Foundation). The liquid HF was condensed into the cleavage vessel with the aid of liquid nitrogen cooling and was then maintained at -10 "C for 30 min at at 0 "C for an additional 30 min. The HF was removed in uacuo and the resin transferred to a fritted glass funnel. The resin was washed alt.ernately with diethyl ether and chloroform three times followed by extraction of the peptide from the resin with 2.0 M acetic acid. Lyophilization of the peptide-containing acetic acid solutions gave the crude, uncyclized peptides. Crude linear peptides were dissolved in 10 mM NH40Ac buffer, pH 7.9, at a concentration of 0.5 mg/ml of crude peptide, and cyclization was accomplished by the slow addition of potassium ferricyanide (K3Fe(CN)6, 0.01 M ) to this stirred solution until excess ferricyanide ion was evidenced by retention of the bright yellow color. The peptide solution was then treated with DOWEX AG3X4 anion-exchange resin (10 g/liter of cyclization solution) for 20 min with stirring, filtered, and the filtrate diluted with water and lyophilized.
Purification of Synthetic Peptides-Linear peptides from HF cleavage or cyclized peptides from ferricyanide cyclizations were desalted on columns of Sephadex G-25 using 0.5 M acetic acid. Following lyophilization, these materials were either purified on open column ion-exchange resins (CM-Sepharose for basic peptides and DEAE-Sepharose for acidic peptides) or directly purified on semipreparative or preparative CIS reversed-phase columns. For ClS purifications of peptides, the semipreparative column was Vydac Cla 218TP510 (1.0 X 25 cm, 5 pm) and the preparative column was C1, Waters Delta Pak (19 mm X 30 cm). The semipreparative column was run at 3.5 ml/min, and the preparative column was run at 18 ml/min. Gradients of acetonitrile ( 5 6 0 % ) in 0.1% trifluoroacetic acid were run to effect separation and purification of the peptide mixtures. Reanalysis of the purified peptides on an analytical Vydac C,, 218TP54 (0.46 x 25 cm) column was performed to ensure homogeneity of the peptides.
Solid-phase Integrin Binding Assays-The ability of synthetic peptides to inhibit the binding of adhesive proteins to purified GPIIb-IIIa, a,/&, and a& was determined using solid-phase microtiter assays, as described (8-11).
Cell Attachment Assays-The ability of synthetic peptides to block scribed (11).
M21 melanoma cell adhesion to vitronectin was measured as de-

RESULTS
Activity of Linear and Cyclic KGD-containing Peptides-Our strategy for the synthesis of small peptides containing the KGD sequence was based on the amino acid sequence of the GPIIb-IIIa selective antagonist barbourin (8). We initially synthesized a linear Lys-containing tetrapeptide, Lys-Gly-Asp-Trp-NH, (KGDW-NH2) as well as the corresponding Arg-containingpeptide Arg-Gly-Asp-Trp-NH2 (RGDW-NH2) and compared their ability to inhibit fibrinogen binding to GPIIb-IIIa in solid-phase integrin ligand binding assays (8-11). Although KGDW-NH, was inhibitory in this assay, it was at least five times less potent than the RGDW-NH2 peptide (Fig. 1A). The reduced inhibitory potency of linear KGD versus RGD peptides for GPJIb-IIIa is consistent with the observations of Ginsberg et al. (12). Since previous studies have revealed that incorporation of RGD sequences into cyclic peptide structures could yield up to a 10-fold increase in the apparent affinity of this peptide for RGD-dependent integrins (13-15), we synthesized the KGD-containing cyclic peptide 1, cyclo(S,S)-G-C-G-K-G-D-W-P-C-A-NH,,, as well as the corresponding RGD-containing peptide, 2, cyclo(S,S)-G-C-G-R-G-D-W-P-C-A-NH,, with a disulfide bridge connecting the carboxyl-and amino-terminal cysteine residues of these peptides. Evaluation of each of these peptides in the binding assays revealed a significant enhancement of potency for the cyclic Lys-containing analog 1, and a more modest enhancement of potency for the cyclic Arg-containing analog 2, (Fig.  1B). The ability of both the linear and cyclic forms of these peptides to inhibit vitronectin binding to a& was also examined to ascertain their integrin inhibitory selectivity. While cyclization of the KGD peptide increased its apparent affinity However, the cyclic RGD peptide 2 was found to be a potent inhibitor of this cellular attachment ( Fig. 2) with halfmaximal inhibition occurring between 5-10 PM. Exocylclic Residues and Optimization of Cyclic Ring Size-Because of the GPIIb-IIIa selectivity of analog 1, this cyclic peptide was an appropriate starting point for optimizing the conformational display of the KGD sequence within this framework. Additional analogs were prepared to explore the effects of decreasing the ring size encompassing the KGDW sequence, as well as the contributions of exocyclic residues in the interaction of peptide analogs with GPIIb-IIIa. The decapeptide 1, which contains an octapeptide ring system was systematically contracted by removal of the Gly residue amino-terminal to the KGDW sequence and the Pro residue carboxyl-terminal to the KGDW sequence affording analogs 3 and 4, respectively. The desGly-analog 3 was found to be equipotent to 1 in inhibiting the binding of fibrinogen to GPIIb-IIIa in ligand binding assays and in inhibiting human platelet aggregation (Table I). However, removal of the Pro residue of 1 resulted in analog 4, a modification which dramatically reduced its potency greater than 50-fold. The exocyclic amino acid residues of 1 were also removed yielding octapeptide analog 5 . This analog was found to be equipotent to analog 1 ( Table I). Combination of these allowable modifications in analogs 3 and 5 afforded the cyclic heptapeptide, 6, cyclo(S,S)-C-K-G-D-W-P-C-NH2, which was also found to be equipotent with decapeptide 1. Further contraction of the heptapeptide cyclic ring by additional deletion of the Pro residue of the heptapeptide analog 6 did not yield a n analog with appropriate solubility to reliably test in the assays. A linear form of analog 6, in which the Cys residues were protected by the S-acetamidomethyl group Cys(Acm), 7, was found to be 15-20-fold-weaker in its ability to inhibit fibrinogen binding to GPIIb-IIIa and in its inhibition of platelet aggregation confirming the importance of the disulfide bridge in the conformational display of the KGDW sequence. These data are summarized in Table I. ing to immobilized GPIIb-IIIa and vitronectin binding to a& using the solidphase microtiter assay. Various doses of linear and cyclic peptides were allowed t o compete for binding of biotinylated adhesive ligand. Round ligand was quantitated by enzyme-linked immunosorbant assay using an anti-biotin antibody conjugated to alkaline phosphatase. Measurements were performed in quadriplicate, and standard deviations were 10% of the mean or less. Each peptide was examined at least three times. of the specific side chains within the KGD recognition sequence. The following single residue substitution analogs of 6 were prepared (Om, His, Leu, Lys(c-CHO)) for Lys in analogs 8-11, (Ala, D-Ala, Sar) for the Gly residue in analogs 12-14, and (Glu) for Asp in analog 15. None of the Lys substitution analogs retained appreciable activity (Table 11). Within the Gly substitution analogs, only analog 12 (Ala for Gly) and analog 14 (Sar for Gly) displayed significant activity in inhibiting platelet aggregation or in inhibiting fibrinogen binding t o GPIIb-IIIa in solid-phase ligand binding assays. The Glu for Asp substituted analog was found to be completely inactive in these assays. Similar to previous studies with RGD peptides, amino acid substitutions other than those with minor  optimization of cyclic ring size Inhibition of platelet aggregation in platelet-rich plasma was determined for each of the synthetic peptides. The ICs0 value for platelet aggregation inhibition is the concentration necessary to inhibit platelet aggregation to 50% of control aggregation induced by ADP (20 PM). modifications of the KGD sequence dramatically reduced or completely eliminated the ability of these analogs to inhibit ligand binding or platelet aggregation (12) ( Table 11). Contribution of the Trp and Pro Flanking Residues-Previous structure-activity studies of linear RGDX compounds have revealed the importance of the specific amino acid residue carboxyl-terminal to the RGD sequence in determining the affinity of peptides for GPIIb-IIIa. Hydrophobic amino acids in the X position have been found to significantly enhance the inhibitory activity of RGD analogs (9, 16, 17). We found that single substitutions of Phe, Tyr(OMe), Leu, Ile, Val, or 2-Nal for the Trp residue of the cyclic analog 6 (analogs 16-21) afforded analogs with varying degrees of inhibitory activity which appeared to correlate well with the hydrophobicity of the given side chain of the X residue ( Fig.  3 and Table 111). The most potent analogs of this series, 6 (Trp), 17 (TyrOMe), and 21 (2-Nal) contain the most hydro- Inhibition of platelet aggregation in platelet-rich plasma was determined for each of the synthetic peptides. The ICso value for platelet aggregation inhibition is the concentration necessary to inhibit platelet aggregation to 50% of control aggregation induced by ADP (20  Inhibition of platelet aggregation in platelet-rich plasma was determined for each of the synthetic peptides. The IC50 value for platelet aggregation inhibition is the concentration necessary to inhibit platelet aggregation to 50% of control aggregation induced by ADP (20 phobic aromatic side chains in this series. The secondary amino acid, Pro, contained within the cyclic heptapeptide 6, might be expected to strictly play a conformational role in the overall folding of the peptide backbone within the ring of heptapeptide analogs rather than contribute any direct interaction of its side chain with the GPIIb-IIIa binding pocket. Several analogs with single substitutions for the Pro residue of analog 6 were prepared (22 Gly, 23 Sar). Analog 22, containing the conformationally flexible Gly residue, was greater than 30-fold weaker in inhibiting platelet aggregation in platelet-rich plasma (ICs, > 200 pM). However, an analog containing the less flexible, N-methylated derivative of glycine, sarcosine, (Sar), (analog 23) was approximately 5-fold weaker than analog 6 in inhibiting platelet aggregation (Table 111). These results suggest that the Pro residue within this class of cyclic peptides may be playing an important role in restraining the conformational flexibility of the heptapeptide ring.

Analog
Contribution of the Disulfide Bridging Residues-The enhanced activity of the cyclic versus linear KGD analogs clearly established the importance of the conformational display of this sequence in potent analogs. However, the bridging groups which enable an optimized display of the KGD sequence were not apparent from our analysis to this point. We prepared an additional series of disulfide bridged analogs based on analog 6 wherein the stereochemistry of the bridging group was modified (D-enantiomer for L-enantiomer), and additional conformational constraints in the side chain of the bridging groups were introduced. We also examined the role of the Namino group at the amino-terminal Cys in modulating the potency of analogs.
Acetylation of analog 6 (analog 24) increased the inhibitory potency approximately 2-fold, suggesting that there was little if any contribution of the positively charged N-a-amino group to GPIIb-IIIa interactions (Table  IV). Removal of the aamino group by substitution of'the NHP-terminal Cys residue with the mercaptopropionyl residue afforded analog 25, which had an additional 2.5-fold increase in potency (IC& = 1.0 p~) in inhibiting ADP-induced aggregation. These results suggest that the presence of the N-a-amino group of 6 is in fact detrimental to its binding to GPIIb-IIIa and that complete removal of the amino functionality could slightly enhance analog affinity for GPIIb-IIIa (Table IV).
The effects of altering the chirality of the carboxyl-terminal Cys residue were also investigated. Substitution of D-CYS for L-Cys of the Mpr-containing analog 25, gave analog 26, with a corresponding 2-fold reduction in potency for inhibiting platelet aggregation (I& = 2.5 pM) as compared to analog 25. The minor effect of this conformational alteration on activity suggests that conformational orientation of the KGDX which is influenced by the bridging residues may not be extremely important in determining the optimal orienta-

TABLE IV Inhibition of human platelet aggregation by cyclic KGD peptides:
disulfide bridging residues and D-amino acid substitutions Inhibition of platelet aggregation in platelet-rich plasma was determined for each of the synthetic peptides. The ICso value for platelet aggregation inhibition is the concentration necessary to inhibit platelet aggregation to 50% of control aggregation induced by ADP (20 uM).

Analog
IC,,, We also explored the effects of additional conformational constraint in the side chains of each of the bridging disulfide groups. Individually, we substituted the mercaptovaleryl (Mvl) group for the Mpr group of analog 25, affording analog 27 and substituted the carboxyl-terminal Cys of analog 25 with the corresponding penicillamine group (Pen) to yield analog 28. The amino-terminal Mvl group and the Pen residue, both contain gem-dimethyl groups at the &carbon atoms of their side chains. This type of side chain alteration been previously shown to restrict the conformational freedom of other disulfide-bridged peptides leading to highly constrained ring systems (18). The Mvl-containing analog 27 was found t o be approximately 2-fold less potent than the Mpr-containing analog 25 in inhibiting platelet aggregation (ICso = 2.3 pM, Table IV). However, the Pen-containing analog 28 was nearly identical to analog 25 in inhibiting platelet aggregation (Table IV, Fig. 4), suggesting that the gem-dimethyl substitutions in the carboxyl-and amino-terminal Cys residues do not lead to additional conformational constraints of this ring system which increase analog affinity for GPIIb-IIIa. Combination of the Mvl residue at the amino terminus with a Pen substitution afforded analog 29, which did not display greater potency than either analog 25 or 28.
This series of analogs demonstrated that an NH2-terminal amino group was not required for high affinity interaction of analogs with GPIIb-IIIa and that further conformational restriction of the cyclic ring system by introduction of groups which restrict the flexibility of the side chains of the disulfide bridging groups was not productive in preparing analogs with increased potency in inhibiting platelet GPIIb-IIIa.
Substitution of D-Amino Acid Residues within the Ring-D-Amino acid substitutions are frequently employed in structure-activity studies to explore the effects this conformational constraint on the activity of analogs (19, 20). Single D-amino acid substitutions were introduced into analog 28 which afforded analogs (D-LYS, 30; D -A s~, 31; D-Tw, 32; and D-Pro, 33). Only the D-LYS analog 30 retained significant activity and was approximately 5-fold weaker than analog 28 in inhibiting platelet aggregation. The other D-amino acid substituted analogs were at least 100-fold weaker in their ability to inhibit platelet aggregation (Table IV). noted in earlier analogs of this study that Lys-containing analogs were consistently weaker inhibitors of GPIIb-IIIa than the Arg-containing analogs but that the Arg-containing analogs were not integrin specific. The reason for this difference in integrin selectivity was not readily apparent from our structure-activity analysis to this point. In order to explore this phenomena in more detail and to increase the potency of the KGD analogs without altering their specificity for GPIIb-JIIa, we prepared an additional series of analogs in which the e-amino group of the Lys side chain was modified.

Modifications of the <-Amino Group of Lys-We previously
Guanidation of the e-amino group of Lys-containing analogs 25 and 28 in solution with the guanidating reagent, 3,5dimethylpyrazole-1-carboxamidine nitrate, afforded the homoarginine-(Har) containing analogs 34 and 35, respectively (21). Both analogs were found to be between 5-10-fold more potent than the parent Lys-containing analogs in inhibiting platelet aggregation and were approximately 5-10-fold more potent in inhibiting fibrinogen or von Willebrand factor binding to GPIIb-IIIa compared to the corresponding Lys-containing analogs (Figs. 3 and 5, A and C ) .
The Lys side chain of analogs 25 and 28 were also modified employing the imidating reagents methyl acetimidate hydrochloride and methyl benzimidate hydrochloride to afford the acetimidate analogs 36 and 37 and the phenylimidate analogs 38 and 39, respectively (22). The acetimidyl-Lys-modified analogs were approximately 2-fold more potent than the parent Lys-containing analogs in inhibiting platelet aggregation (Table V) as well as fibrinogen and von Willebrand factor binding to GPIIb-IIIa (Fig. 5 , B and C). The corresponding phenylimidyl-Lys analogs 38 and 39 were equipotent to the parent Lys-containing analogs. We also prepared the RGD analog of 28 (analog 40) for comparison with the Lys and Lys-modified analogs.
The effect of modification of the Lys side chain on integrin specificity was also examined in the a& binding assay employing biotinylated vitronectin and von Willebrand factor as ligands (Fig. 6, A and B ) and in the effects on adherence of M21 cells to vitronectin (Fig. 7). Lys-containing analogs that were extremely potent inhibitors of GPIIb-IIIa had virtually no effect on a,@,-mediated binding. However, guandiation of the lysine side chain decreased the specificity of these compounds, although they remained 500-600 times less reactive with CY& than the corresponding Arg-containing analog 40 (Figs. 5 and 6). In contrast, the imidated Lys-containing analogs were more specific for GPIIb-IIIa than the parent Lys-containing analogs. This group of analogs with Lys side chain modifications was found to be the most potent GPIIb-IIIa antagonists prepared in this study and retained nearly absolute specificity for GPIIb-IIIa uersus other RGD-dependent integrins. DISCUSSION We have recently described barbourin, a novel member of the disintegrin family of peptides which, unlike all other disintegrins, is a specific antagonist of GPIIb-IIIa. The structural basis for this specificity has been attributed to the substitution of a Lys for Arg residue in the Arg-Gly-Asprecognition sequence of the peptide (8). In the current study we demonstrate that the GPIIb-IIIa specificity of barbourin is retained in small, KGD-containing cyclic peptides. We have further found that specific modifications of the Lys side chain alter the integrin selectivity of these peptides, while the cyclic ring size and residues adjacent to the KGD sequence have profound effects on the affinity of these peptides for GPIIb-IIIa. Optimization of these modifications has led to the design of a series of small, conformationally constrained peptides 0 .00 0.10 0 . 2 0 0.30 0.40 0.50 . -

Peptide (kM)
Peptide Antagonists of GPIIb-IIIa which are potent and highly specific antagonists of GPIIb-IIIa and which closely mimic the properties of barbourin.
At the time we began these studies, a single report had demonstrated that the incorporation of the RGD recognition sequence into a disulfide-containing, cyclic octapeptide ring system could increase the affinity of these analogs compared to the corresponding linear RGD-containing peptide as measured by the inhibition of attachment of normal rat kidney cells to vitronectin (13). Interestingly, this cyclic RGD modification led to a decrease in potency in inhibiting cellular attachment to fibronectin, and it was proposed that various integrins can recognize different conformational displays of the RGD sequence in synthetic, and perhaps native adhesive ligands. More recently, other groups using different approaches have reported that constrained cyclic ring systems containing RGD or modified RGD sequences can yield potent inhibitors of GPIIb-IIIa. Utilizing a cyclic peptide system similar to Pierschbacher and Rouslahti (13) as a model, we prepared several cyclic peptide analogs containing RGDW and KGDW sequences to test their relative GPIIb-IIIa inhibitory potency and integrin selectivity. We chose the RGDW and KGDW sequences because of our previous observations that RGDW and KGDW containing snake venom disintegrins from Crotalus and Sistrurus species appeared to bind to GPIIb-IIIa with higher affinity than other RGDX-containing disintegrins (9). The Lys for Arg substitution in the RGD sequence has also been previously examined and found to significantly reduce the apparent affinity for these peptides for integrins (12, 23).
Our initial decapeptide analogs which were cyclized through a disulfide bridge were found to be at least 10-fold more potent in inhibiting fibrinogen binding to GPIIb-IIIa and in inhibiting platelet aggregation when compared to the corresponding linear RGDWor KGDW-containing peptides. More important, however, than the significant increase in inhibitory potency caused by cyclization was maintenance of high selectivity for GPIIb-IIIa in the cyclic KGDW-containing analogs. Further optimization of the affinity of these antagonists for GPIIb-IIIa by systematically contracting the ring size of the disulfide bridged ring was investigated. We found that the optimal ring size for antagonists in this series contained a heptapeptide ring system that was disulfide bridged. Other investigators have also reported potent cyclic heptapeptide sequences utilizing a different approach (24, 25). However, similar potency for the inhibition of GPIIb-IIIa can be obtained from cyclic pentapeptide templates (23), suggesting that similar orientations of important functional groups of the inhibitors interacting with GPIIb-IIIa can be obtained within very different conformational templates.
The secondary amino acid residue, Pro, which was carboxylterminal to the KGDX sequence, the residue which we speculated had little direct interaction with the GPIIb-IIIa binding pocket in these analogs, was found to be extremely important for maintaining high inhibitory potency in this series of analogs. The Pro residue presumably functions to position the KGDW residues for optimal interaction with the binding pocket of GPIIb-IIIa. Substitution of other residues for the Pro consistently decreased the inhibitory potency of these analogs. Although the conformation of the heptapeptide ring in these highly active analogs in not known, a Pro residue should limit the conformational freedom in this small ring system to a greater degree than in analogs in which the Pro is substituted by other primary amino acids.
We also reexamined the contribution of the individua1 side chains of the RGD/KGD recognition sequence on the inhibitory activity of analogs. As others have reported, we found that the basic amino acid residues Arg, Lys, or substituted Lys are of critical importance for the biological activity of these peptides. Removal of the charged side chain or alteration of the chirality of this residue had very deleterious effects on analog potency. The Asp residue was also found to be extremely important as previously demonstrated, and substitution by the side chain methylene extended homolog, Glu, completely removes the ability of analogs to function as integrin antagonists (12). The Gly residue situated between the Arg/Lys and Asp residues appeared to tolerate some degree of alteration since Ala or Sar substitution afforded active analogs. However, Gly was the optimal residue for high affinity interaction of analogs with integrins. These observations are in contrast with those recently reported for cyclic pentapeptide GPIIb-IIIa inhibitors where substitution for the Gly of the Arg-Gly-Asp sequence was not tolerated (23). It is not surprising, however, that different conformational tem- plates may yield different backbone requirements within the active site of the inhibitor molecule.
Previous studies have noted that large hydrophobic side chains in the X position of the RGDX sequence are optimal for high affinity interactions with GPIIb-IIIa and possibly other integrins (9, 16, 17,23). In this study, we found that the Trp residue found in many disintegrin peptides (8, 9) is a nearly optimal residue at the X position for our series of cyclic peptides. However, a variety of large hydrophobic amino acid residues at this position are also acceptable substitutions.
Because of this observation, Trp was retained in many of the analogs as other means for optimization of affinity for GPIIb-IIIa were explored. Although attempts to increase the affinity of analogs for GPIIb-IIIa were approached by altering the disulfide bridging residues through alteration of chirality of the bridging groups or by introducing conformational constraints into their side chains, only small additional increases in potency were achieved by this approach. An amino terminal 3-mercaptopropionyl group (Mpr) and a carboxyl-terminal cysteineamide residue appeared to be sufficient to obtain peptides that approached the affinity of the disintegrin peptide, barbourin.
Our last series of modifications of the cyclic Lys-containing analogs involved derivatization of the €-amino group of the Lys side chain, which led to significant increases in the apparent affinity of these peptides for GPIIb-IIIa. Several of these analogs inhibited ADP-induced platelet aggregation a t submicromolar concentrations. Guanidation of the Lys-containing analogs which afforded homoarginine containing inhibitors led not only to significant increases in the affinity of analogs for GPIIb-IIIa, but also increased the affinity of these analogs for a,& as assessed in solid-phase binding assays. However, the specificity of Har-containing analogs for GPIIb-IIIa was still found to be extremely high. Even greater GPIIb-IIIa specificity of analogs was achieved when the Lys side chain was imidated affording either the phenyl or methylimidate derivatives. Analogs with this side chain modification were also found to be extremely potent inhibitors of GPIIb-IIIa but do not appear to interact to any significant degree with the RGD-sensitive vitronectin receptor, a&.
Small peptides containing the RGD amino acid sequence have been shown to inhibit not only the binding of fibrinogen and von Willebrand factor to platelet GPIIb-ITIa but to also described in this study are ideal candidates to be investigated inhibit the adhesive functions of other RGD-dependent inte-in vivo as potential platelet aggregation inhibitors where their grins (12). Previous attempts to design potent platelet aggre-value as antithrombotic agents can be assessed. gation inhibitors have relied on the RGD sequence. Because of these approaches, throughout this structure-activity study