Stapling Mimics Noncovalent Interactions of γ-Carboxyglutamates in Conantokins, Peptidic Antagonists of N-Methyl-d-Aspartic Acid Receptors*

Background: Can dicarba bridges (stapling) replace noncovalent interactions that stabilize helical conformation of neuroactive peptides? Results: A rational design, synthesis, structural, and functional characterization of stapled conG analogs that target NMDA receptors is reported. Conclusion: Stapled conG analogs are potent antagonists of NMDA receptors and anticonvulsant compounds. Significance: Stapling can be successfully applied to convert neuroactive peptides into drug leads. Conantokins are short peptides derived from the venoms of marine cone snails that act as antagonists of the N-methyl-d-aspartate (NMDA) receptor family of excitatory glutamate receptors. These peptides contain γ-carboxyglutamic acid residues typically spaced at i,i+4 and/or i,i+7 intervals, which by chelating divalent cations induce and stabilize helical conformation of the peptide. Introduction of a dicarba bridge (or a staple) can covalently stabilize peptide helicity and improve its pharmacological properties. To test the hypothesis that stapling can effectively replace γ-carboxyglutamic acid residues in stabilizing the helical conformation of conantokins, we designed, synthesized, and characterized several stapled analogs of conantokin G (conG), with varying connectivities in terms of staple length and location along the face of the α-helix. NMR studies confirmed that the ring-closing metathesis reaction yielded a single product with the Z configuration of the olefinic bond. Based on circular dichroism and molecular modeling, the stapled analogs exhibited significantly enhanced helicity compared with the native peptide in a metal-free environment. Stapling i,i+4 was benign with respect to effects on in vitro and in vivo pharmacological properties. One analog, namely conG[11–15,Si,i+4S(8)], blocked NR2B-containing NMDA receptors with IC50 = 0.7 μm and provided significant protection in the 6-Hz psychomotor model of pharmacoresistant epilepsy in mice. Remarkably, unlike native conG, conG[11–15,Si,i+4S(8)] produced no behavioral motor toxicity. Our results extend the applications of peptide stapling to helical peptides with extracellular targets and provide a means for engineering conantokins with improved pharmacological properties.

The conantokins are a large family of neuroactive peptides found in the complex venoms of marine snails, genus Conus, which antagonize N-methyl-D-aspartate (NMDA) receptors (1). Conantokins are small helical peptides, sharing N-terminal sequence identity, but in contrast to most Conus peptides are not stabilized by disulfide cross-links. The biochemical signature of conantokin peptides is their high content of post-translationally modified glutamate in the form of ␥-carboxyglutamate (Gla). 2 The Gla residues found in conantokin peptides play both structural and functional roles. The malonate head groups of Gla permit such side chain residues to effectively chelate metal ions and impose a high degree of structural rigidity (2). Of the conantokins discovered and characterized thus far, up to five Gla residues have been observed in a single conantokin, distributed at i,iϩ4, i,iϩ7, and/or i,iϩ11 intervals along the primary sequence. When Ca 2ϩ is present, Gla residues act as a "zipper" by coordinating Ca 2ϩ , stabilizing the ␣-helical conformation (3,4).
The helical conformation of a peptide can be stabilized by both noncovalent and/or covalent interactions, including metal ion complexation by the side chains of the peptide (5, 6) (described earlier (4)), formation of a salt bridge (7)(8)(9), or hydrophobic (10)/hydrophilic (11) interactions between side chains, lactam bridges, or disulfide bonds or through hydrocarbon staples (12)(13)(14)(15). The hydrocarbon staple is introduced by incorporation of two ␣-methyl, ␣-alkenyl amino acids to create approximately one (i and iϩ3 or i,iϩ4), two (i,iϩ7), or three turns (i,iϩ11) of the helix, following ruthenium catalyzed ringclosing metathesis (RCM) (16). Incorporation of an all-hydrocarbon staple cannot only improve the helical content of a peptide, relative to the native form, but also cellular uptake, binding affinity for the target receptor, serum stability, and in vivo halflife. Examples of previously described stapled peptides include: i,iϩ7 stapling of the ␣-helix p53 (SAH-p53) peptide (17); i,iϩ4 stapling of a pro-apoptotic BH3-only protein, SAHB A (17), MCL-1 (18,19), BAD (20), CAI (21), and stapled peptides that bind at the co-activator protein-binding site of endoplasmic reticulum (22). Recently, a double i,iϩ4 stapling was applied to T649v (13) to produce an analog SAH-gp41 (626 -662) (A,B) with improved activity against neutralization-resistant HIV-1 virus and enhanced pharmacokinetics compared with its unmodified form. The stapled analogs mentioned above exhibited potent bioactivity in addition to well defined ␣-helical structure, protease resistance and cell penetration, indicating stapling as an attractive way of improving structural and pharmacological properties of peptides.
In this work, we applied the stapling strategy to the conantokin peptides to examine the effects of replacing Gla residues with the dicarba bridge on conantokin helical conformation, stability, and bioactivity. As a model conantokin for this study, we selected conantokin G (conG), a potent inhibitor of NR2Bcontaining NMDA receptor ion channels. ConG is a 17-amino acid peptide containing five Gla residues. The structure of conG has been well studied and characterized by NMR (3,23), crystallography (24), and electrophysiology (25) experiments. Based on the structural information, we have designed several analogs of conG with the dicarba bridge incorporated across one (i,iϩ4) and two (i,iϩ7) turns (see Fig. 1) and found that stapled conG analogs can exhibit greater helicity, similar in vitro activity, improved in vivo bioactivity, and reduced behavioral toxicity compared with native conG.
Ring-closing Metathesis, Peptide Cleavage, and Purification-10 mg of resin-bound, fully protected linear peptide (substitution 0.38 mmol g Ϫ1 ) was placed in a glass vial, and a solution of 40 mol% (1.25 mg) Grubb's First Generation Catalyst (Sigma-Aldrich) in 0.3 ml of dichloromethane was added under argon atmosphere. The reaction vial was placed on an orbital shaker and gently mixed for 48 h at room temperature. The progress of the reaction was monitored by reversed phase analytical C 18 RP-HPLC. After the reaction was completed, the resin was washed in a fritted syringe using DMF (5 ϫ 1 ml), methanol (MeOH; 5 ϫ 1 ml), and dichloromethane (5 ϫ 1 ml). Next, the resin was placed in a vial, 0.5 ml of 20% piperidine in DMF was added, and the mixture was gently stirred using an orbital shaker for 30 min. Then resin was washed with DMF (5 ϫ 1 ml), MeOH (5 ϫ 1 ml), dichloromethane (5 ϫ 1 ml), and diethyl ether (5 ϫ 1 ml). Finally, peptide was removed from the resin by a 3-h treatment with reagent K (82.5/5/5/5/2.5 v/v/v/ v/v, trifluoroacetic acid/water/phenol/thioanisole/ethanedithiol) and then precipitated and washed with cold methyl tertbutyl ether. Crude peptides were purified by RP-HPLC using a semi-preparative column (Vydac C 18 , 218TP510, 250 mm ϫ 10 mm, 5-m particle size) and solvents (solvent A contained 0.1% TFA, and solvent B contained 90% acetonitrile, 0.1% TFA). Analogs were purified over a gradient of solvent B: analog conG[10 -14,S i,iϩ4 S(8)]: 20 -50% in 30 min, conG [11-15, S i,iϩ4 S(8)]: 10 -50% in 40 min, conG[7-14,R i,iϩ7 S(11)]: 15-45% in 30 min, with a flow rate of 4 ml/min. The absorbance of the eluate was monitored at 220 nm. Purity of peptide was assessed by analytical RP-HPLC using a Vydac C 18 column (218TP54, 250 mm ϫ 4.6 mm, 5-m particle size) with a flow rate of 1 ml/min using linear gradients described above. The peptides were quantified against a reference peptide using the same gradient. Molecular masses of all analogs were confirmed by electrospray ionization-MS (see Table 1).
Double Bond Saturation-43 mg of the fully protected, resin bound, metathesized conG[10 -14,S i,iϩ4 S(8)] was placed in a 2 ml glass vial and treated with 0.5 ml of the following mixture: 1.4 M piperidine, 0.7 M 2,4,6-triisopropylbenzenesulfonyl hydrazide in anhydrous N-methyl-2-pyrrolidone for 2 h at 47°C. The solution was replaced with a fresh mixture after 2 h and the reaction was heated for another 2 h (the procedure was repeated one more time). Finally, the resin was washed with DMF (5 ϫ 1 ml), MeOH (5 ϫ 1 ml) and diethyl ether (5 ϫ 1 ml) and the peptide was removed from the resin with reagent K as described above. Peptide was purified using RP-HPLC equipped with a Vydac semi-preparative C 18 column over a gradient of solvent B ranging from 20% to 50% in 30 min with 4 ml/min flow rate. Calculated mass was confirmed by electrospray ionization-MS (see Table 1).
Circular Dichrosim Spectroscopy-CD spectra were recorded on an AVIV model 410 spectropolarimeter, using the method and the parameters described in the CD studies of conRl-A (26). Briefly, the peptides were dissolved at 100 M final concentration in 10 mM HEPES buffer, pH 7.4, with or without 2 mM CaCl 2 , and the measurements were taken at room temperature.
The spectra were measured five times and averaged for the sample and reference, respectively. Subtracting the CD signal of the buffer from the peptide CD signal eliminated the contribution of the buffer to the peptide CD signal. The spectral intensities were expressed as mean residue ellipticities using the equation reported elsewhere (27), and a molar ellipticity of Ϫ35086.66 degrees cm 2 dmol Ϫ1 was estimated to be a perfect ␣-helix (100% ␣-helix). The percentage of helical conformation was calculated by assuming a linear relationship in comparison with 100% ␣-helicity.
NMR-One-dimensional NMR spectra were recorded with an Inova 600 NMR spectrometer equipped with a cryogenic probe. Dried and RP-HPLC-purified peptides conG [10 -14,S (28) and ff99SB AMBER force field (29). The structures were built and minimized in vacuo, followed by molecular dynamics (MD) with implicit (generalized Born) solvation (igb ϭ 5) (30). Models were subjected to 2.2-ns MD simulation with constant temperature t ϭ 300 K and also replica exchange MD simulation, which starts several independent simulations at different temperatures (t ϭ 300, 350, 400, 450, 500, 550, and 600 K) in parallel. The replica exchange MD method allowed for exploring of the conformational space of the peptides. A cutoff of 9 Å was applied, and the temperature was controlled through a Langevin thermostat (31) with a factor of 1 ps. A time step of 1.0 fs was applied during the MD simulations. The analysis of the helicity was also performed with the DSSP method, developed by Kabsch and Sander (32).
Heterologous Expression of NMDA Receptors in Xenopus Oocytes-The rat NMDA receptor clones contained within a pSGEM vector for NR1-2b, NR2A, NR2B, NR2C, and NR2D subtypes used were: GenBank TM numbers U08266, AF001423, U11419, U08259, and U08260, respectively. cRNA for each NMDA subtype was prepared using in vitro RNA transcription kits (Ambion, Inc., St. Louis, MO) to a final concentration of 200 g/nl according to the manufacturer's protocol. NMDA receptors were heterologously expressed by nano-injecting 2-5 ng of each NR1/NR2 subunit cRNA per oocyte of Xenopus laevis. The protocol for Xenopus oocyte harvesting was described previously in detail (33). Oocytes were stored in a Petri dish containing ND-96/Pen/Strep/Gent (100 units/ml penicillin G (Sigma), 100 mg/ml streptomycin (Sigma), and 100 mg/ml gentamycin (Invitrogen) at 17°C and left for 1-5 days to express.
Two Electrode Voltage Clamp Electrophysiology-Voltage clamp recording of Xenopus oocytes was conducted as described in detail previously (33). Briefly, all of the oocytes were voltage clamped at Ϫ70 mV at room temperature. The oocytes were gravity-perfused with Mg 2ϩ -free ND96 buffer (96.0 mM NaCl, 2.0 mM KCl, 1.8 mM CaCl 2 , and 5 mM HEPES, pH 7.2-7.5). Mg 2ϩ was omitted from the ND96 buffer to prevent the voltage-dependent blockade of NMDA receptors at Ϫ70 mV. Bovine serum albumin (BSA) (0.1 mg/ml) was added to reduce nonspecific absorption of peptide. One-second pulses of gravity-perfused agonist solution (200 M glutamate and 20 M glycine in Mg 2ϩ free ND-96 with BSA) were used to elicit NMDA receptor-mediated current. Agonist was applied at saturated concentration for all four subtypes and elicited similar response. 3 To measure the effect of stapled conG analogs on currents elicited from oocytes expressing NMDA receptors, the buffer flow was halted, and the peptides were applied in a static bath for duration sufficient to reach equilibrium or a minimum of 5 min. The inhibition of NMDA receptor-mediated current by peptides was measured by normalizing the response of the first agonist pulse following static bath to the base-line response (the average of three agonist-elicited currents in response to agonist prior to peptide application). A virtual instrument made by Dr. Doju Yoshikami at the University of Utah was used for data acquisition, and concentration-response curves were generated using Prism 4 for Windows (GraphPad Software, Inc., La Jolla, CA). The following equation, where n H is the Hill coefficient, and IC 50 is the concentration required to achieve half-maximal block, was used to fit concentration-response curves: % response ϭ 100/[1 ϩ ([peptide]/IC 50 Serum Stability Assay-The stability of peptides in the presence of 25% rat blood serum was evaluated for conG and conG[10 -14,S i,iϩ4 S(8)] by incubation at 37°C for 0 min, 30 min, 1 h, 2 h, 4 h, and 8 h. Samples were prepared by adding 5 g of the peptide, resuspended in nH 2 O, to preheated tubes containing 25% rat blood serum and 0.1 M Tris-HCl, pH 7.5. At the appropriate time points, the reactions were quenched by precipitation of serum proteins through addition of 100 l of isopropanol/water/trichloroacetic acid (45%:40%:15% v/v/v). The samples were then incubated at Ϫ20°C for 20 min, followed by centrifugation at 10,000 rpm for 3 min to remove serum proteins. Supernatants were then removed and were analyzed by analytical HPLC equipped with a Waters YMC ODS-A 5-m 120 Å column using a gradient ranging from 5 to 95% solvent B in 45 min including a 15-min pre-equilibration. Metabolic stability of each peptide was assessed by determining a time course of the disappearance of an intact peptide. Half-lives (t1 ⁄ 2 ) for each peptide were determined from at least three independent time course experiments using the equation below (where m is the slope of the line, and b is the y intercept).
Anticonvulsant Activity of Stapled ConG Analogs-The 6-Hz partial psychomotor seizure test was performed to assess the anticonvulsant potential of stapled conG analogs as described previously (10). Stock solutions of the peptides were prepared in 0.9% saline and were diluted to the required concentration prior to intracerebroventricular (i.c.v.) injections. For intracerebroventricular administrations, the test solution was administered in a volume of 5 l, using a Hamilton syringe (size number 701), directly through the skull to a depth of 3 mm into the lateral ventricle of the brain. Each peptide was tested on groups of eight mice. A current of 32 mA was administered via corneal electrodes for 3 s to elicit psychomotor seizure characterized by forelimb clonus, twitching of the vibrissae, and Straub tail. Animals not displaying seizure activity were considered "protected." Rotorod Testing-The rotorod procedure was used to disclose minimal muscular or neurological impairment. Briefly, each mouse (n ϭ 8) was observed for 1 min on a rod ϳ18 inches above the lab bench that rotated at a speed of 6 rpm. The animal was considered motor impaired if it fell off of the rotating rod three or more times during the 1-min observation period.
Animal Care-Adult male CF No. 1 albino mice (26 -35 g), obtained from Charles River (Portage, MI), were utilized for behavioral testing in both the 6-Hz test and the rotorod procedure. The animals were maintained on an adequate diet (Prolab RMH 3000) and allowed free access to food and water, except during the short time they were removed from their cage for testing. The animals were housed, fed, and handled in a manner consistent with institutional animal care and use committeeapproved protocol.

RESULTS
Design of Stapled ConG-The conformation of conG is a distorted curvilinear 3 10 helix, which in the presence of Ca 2ϩ transitions into a linear ␣-helix, exposing linearly aligned residues Gla 3 , Gla 7 , Gla 10 , and Gla 14 on one face of the helix (3) (Fig. 1). Antagonistic activity of conG against NMDA receptors strongly depends on the N-terminal residues, especially Gla 3 and Gla 4 (25,35,36). From a structure activity relationship study conducted by Blandl et al. (25), replacement of Gla 7 , Gla 10 , and Gla 14 did not affect potency of the peptide. Moreover, when Gla 7 was replaced with Ala, it led to a more active analog. Taking into account the critical role played by Gla 3 in the antagonistic activity of the peptide, only residues Gla 7 , Gla 10 , and Gla 14 were substituted in this work. Furthermore, the last two Gla residues were known to form a tight metal cation-binding site but did not play a role in receptor binding (37). Table 1 summarizes all analogs described within this study. ConG[10 -14,S i,iϩ4 S(8)] and conG [11-15,S i,iϩ4 S(8)] were synthesized to examine stapling of one helical turn. Knowing the importance of Gla residues in metal binding, Gla 10 and Gla 14 of the first analog were replaced with ␣,␣-disubstituted amino acids. Selection of the ␣,␣-disubstituted amino acids participating in the RCM reaction was based on the results published by Schafmeister et al. (38), which indicated the optimal stereochemistry at the ␣-carbon and length of the alkyl tethers of the amino acids participating in the RCM reaction. In the second analog, conG[11-15,S i,iϩ4 S(8)] amino acids creating the staple were shifted by one residue toward the C terminus to localize  the hydrocarbon chain to an adjacent face of the peptide, leaving Gla 10 and Gla 14 present in the sequence. ConG[7-14,R i,iϩ7 S(11)] had replaced Gla 7 and Gla 14 to stabilize two helical turns. Constraining helical conformation of conG by an unsaturated, rigid dicarba bridge was expected to influence not only the structure but also the biological activity of the peptide. Based on the electrophysiology data obtained for the first three analogs, conG[10 -14,S i,iϩ4 S(8)], was selected to explore the effects of unsaturated versus saturated dicarba bridges. This analog exhibited increased potency for multiple NMDA receptor subtypes.
Chemical Synthesis of Stapled ConG-Chemical synthesis of stapled conG analogs was performed on a solid support as described in detail under "Experimental Procedures." Briefly, using Fmoc-SPPS, linear peptides were synthesized and then subjected to the RCM reaction using Grubb's first generation ruthenium catalyst (Fig. 2). RCM was followed by Fmoc deprotection, resulting in a single product with a conversion of more than 95% for both i,iϩ4 analogs conG[10 -14,S i,iϩ4 S(8)] and conG [11-15,S i,iϩ4 S (8)]. To ensure that a single product was obtained, analytical HPLC at elevated (45°C) and low temperature (3°C) was run, resulting in a single chromatographic peak. Reduction of the double bond initially failed when palladium on activated carbon was used under H 2 atmosphere. However, saturation was achieved using a mixture of piperidine and 2,4,6-tri-isopropylbenzene sulfonyl hydrazine, producing the desired analog, conG[10 -14,S i,iϩ4 S(8)] sat , with 81% conversion. For the i,iϩ7 analog conG[7-14,R i,iϩ4 S(11)], the same synthetic approach was applied. For both the linear and the metathesized peptides, a small amount of an additional peak with the same molecular weight as the desired product was observed. Further investigation suggested that the second peak might be a conformational isomer.
Structural Studies-Native conG and the stapled analogs were analyzed by CD, NMR, and molecular modeling. The results of CD experiments are presented in Fig. 3. It is known that incorporation of two ␣,␣-disubstituted amino acids into the peptide sequence increased the helical content with respect to the unmodified peptide (38,39). As expected, all unmetathesized analogs exhibited increased ␣-helical content compared with the native peptide (data not shown showed reduced helicity compared with the unsaturated analog. It was previously observed for different peptides that such manipulation did not influence the peptide helicity (38). As previously reported, the amount of ␣-helicity in conG increased dramatically in the presence of Ca 2ϩ (40). Most of the stapled analogs showed a slight increase in helical content in the presence of Ca 2ϩ . The enhancement was particularly significant for conG [11-15,S    100%. This effect might be explained by Ca 2ϩ chelation by Gla 10 and Gla 14 present in the sequence, which further stabilized the helix.
The heteronuclear single quantum coherence NMR experiment of conG[10 -14,S i,iϩ4 S(8)] (Fig. 4A) confirmed that the RCM reaction produced a single product, and a single set of resonances was observed (data not shown). The double bond of the dicarba bridge was confirmed by 1 H/ 13 C chemical shifts of 5.43/126.4 and 5.34/128.4, respectively. The measured scalar coupling constant between the olefinic protons, 3 J x,y 9 Hz, was consistent with the Z-isomer (for a single olefin model compound with R 1 R 2 ϭ CH 3 , the Z and E conformers give coupling constants of 10.9 and 15.1 Hz, respectively (22,41)). Similarly, for conG[7-14,R i,iϩ7 S(11)], configuration of the double bond was confirmed to be Z (estimated coupling constant 3 J x,y 9 Hz; 1 H chemical shifts 5.21 and 5.37 ppm) (Fig. 4C). Upon saturation, the olefinic signals of conG[10 -14,S i,iϩ4 S(8)] sat shifted up-field (Fig. 4B).
NMR also showed a subset of amides that were protected from a solvent exchange (NH 3 ND) after 30 -60 min in D 2 O (Fig. 4). Fig. 4 (A and B) shows five and four protected amides, respectively, which is consistent with stabilization of a single helical turn by the dicarba bridge. This is further supported in Fig. 4C, where nine amides are protected in two helical turns stabilized in the peptide. Amide exchange was not significantly different after double bond saturation (Fig. 4, compare A  and B).
Molecular modeling simulations also confirmed a stabilizing effect of the all-hydrocarbon staple on the conG structure (Fig. 5). All-atom implicit solvent molecular dynamics simulation was performed on two stapled peptides: conG[10 -14,S i,iϩ4 S(8)] and conG [11-15,S i,iϩ4 S(8)], as well as the native peptide (similar work was done by Guo (43) to study a series of stapled p53 peptides). One of the problems was sampling the conformational space of the peptide during the simulations and avoiding trapping in the local minima. To assure the correctness of the simulations, two approaches were taken: 1) standard MD simulation and 2) replica exchange molecular dynamics, also called replica exchange MD simulation. The second approach was applied in series of simulations, which were run in parallel at different temperatures and allowing exchange of the replicas during the runs. The results confirmed that stapling of conG increased the helicity of the peptide. Within the native conG, we saw two helical fragments between residues 3-5 with lower helicity (30%) and 9 -15 with higher helicity (60%). For conG[10 -14,S i,iϩ4 S(8)], a helicity increase for both fragments was observed; helicity of the first fragment was increased to 45%, whereas the second fragment increased to 80%. Moreover, for conG [11-15,S i,iϩ4 S(8)], an increase in helicity was observed throughout the whole peptide from residues 2 to 16, eliminating the division in helicity observed for the native peptide and conG [10 -14,S i,iϩ4 S(8)]. The C-terminal region was consistently the most helical (80%).
Discrepancies in the helical data from experimental (CD) and theoretical (molecular modeling) methods were noticed. Molecular dynamic simulations predicted that conG [11-15,S i,iϩ4 S(8)] should be the most helical analog. At this time, the cause of such differences is undetermined; however, follow-up studies will be carried out to better explain this phenomenon.
Metabolic Stability-One of the advantages of stapling is the improvement of peptide stability. To determine how the stapling influenced metabolic stability of conantokins, we carried out an in vitro serum stability assay. The analogs were incubated in 25% rat blood serum at 37°C, and at appropriate time points (0, 0.5, 1, 2, 4, and 8 h), aliquots were removed and treated with a mixture of trichloroacetic acid/isopropanol/ nH 2 O (15:45:40, v/v/v). The disappearance of the peptide was measured as the decrease in peak area over time using analytical HPLC methods. Both peptides, conG[10 -14,S i,iϩ4 S(8)] and conG, exhibited t1 ⁄ 2 values of Ͼ10 h (Fig. 7). The data were analyzed with an unpaired t test in GraphPad Prism version 5 for Windows (GraphPad Software, La Jolla, CA), and the difference between those two peptides was not found to be statistically significant (p value ϭ 0.6958).
Anticonvulsant Activity of Stapled ConG Analogs-Conantokins are potent anticonvulsant compounds (reviewed in Ref. 1). Given the high subtype selectivity of stapled conG analogs for NR2B, these peptides were tested for potential anticonvulsant activity in the 6-Hz (32 mA) mouse model of pharmacoresistant epilepsy. At 32 mA, which is 1.5 times the convulsive current required to evoke a seizure in 97% of mice tested (CC97), the 6-Hz model is resistant to phenytoin and lamotrigine, while maintaining its sensitivity to ethosuximide, laco-

TABLE 2 Electrophysiological and anticonvulsant characterization of stapled conantokins
The IC 50 (M) values recorded on NMDA receptors containing four distinct NR2 subunits coexpressed with the NR1-2b splice variant are listed for each stapled peptide analog. In addition, the behavioral data from both the 6-Hz (32 mA) mouse model of epilepsy and the rotorod test are shown for each stapled peptide. Analogs were administered intracerebroventricularly to groups of eight mice at a concentration of 0.1 nmol/5 l peptide. Rotorod assessment was performed immediately before the 6-Hz (32mA) test. Motor impairment refers to the percentage of animals that failed the rotorod test, whereas anticonvulsant activity refers to the percentage of animals not exhibiting seizure activity (protected) in response to the 6-Hz (32 mA) corneal stimulation at 1 h after intracerebroventricular administration of each stapled analog. The motor impairment data were analyzed with Kruskal-Wallis post hoc test using GraphPad Prism version 5.0c for Mac OS X (GraphPad Software, La Jolla, CA). samide, levetiracetam, and valproic acid (44 -46). The data obtained with stapled analogs of conG are summarized in Table  2. For all of the stapled analogs, at a dose of 0.1 nmol following i.c.v. injection, more than 75% mice were protected (n ϭ 8) from seizures at a time to peak effect of 1 h, whereas no control mice (n ϭ 8) were protected (5 l of saline i.c.v.). To determine the degree of behavioral toxicity, the rotorod test was also performed. At the same dose of 0.

DISCUSSION
Recently, stapling was demonstrated to be a valid method for improving the pharmacological properties of peptides, such as binding affinity, serum stability, and cellular uptake. This has been demonstrated across many different signaling pathways, from a trigger of apoptotic pathway in leukemia cells to an HIV-1 fusion inhibitory peptide (17,13). In this work we show that replacing Gla residues with dicarba bridges in conantokins is promising with respect to improving conformation (expected) and pharmacology (unexpected), providing strong motivation for the continued study of stapled conopeptides as pharmacological tools or even perhaps drug lead candidates.
Helical conformation is a defining feature of conantokins. Most adopt helical conformations in the presence of divalent cations. However, there are a few known conantokins that are inherently helical due to ␥-carboxyglutamate residue 7 being replaced with lysine; examples include conRl-A, conPr-C, and conT (26,27,47). Other elements that have been shown to contribute to conantokin structure are disulfide bridges; examples include conR and conP (48,49), although structural analysis of conR showed that the C-terminal disulfide bridge actually disrupted, rather than stabilized, ␣-helical content (25). ConG is unstructured in the absence of divalent cations (i.e. calcium) and only adopts a helical conformation in their presence, representing the unique paradigm of metal-dependent helical transition in peptides. This behavior is attributed to the presence of multiple Gla residues in the sequence, which chelate calcium by tetravalent interaction and restrict the conformation of the peptide by inducing an ␣-helix (24). By applying the stapling technology, we were able to covalently stabilize the helical conformation of conG in the absence of divalent cations in solution. Our CD data show a drastic increase in helical content for the stapled analogs compared with the native peptide, resulting from just the insertion of two ␣-methyl-␣-alkenyl amino acid residues. In our opinion, these residues can be viewed as analogs of Aib (2-methyl alanine) (50), which is known to promote helicity in various other peptides. A similar increase in helical content was observed by Schafmeister et al. (38) and Bird et al. (13). Paraphrasing the statement of Kaul and Balaram (39), we can conclude that the nonstandard amino acids have suitable stereochemical properties, which act as conformational directors of conantokin folding. Metathesized analogs demonstrated an additional increase in helical content exceeding, in most cases, the theoretical value of 100%. Furthermore, the one-dimensional NMR data confirmed a stabilizing effect of the staple on the conantokin structure. We observed a protection of the amide protons by the dicarba bridge for both i,iϩ4 and i,iϩ7 types of stapled conG (Fig. 4), at sites localized within or next to the amino acids creating the bridge. These results are in agreement with previous findings by Bird et al. (13), who studied stapling in the context of increased proteolytic stability of a peptide.
Conantokins are one of two highly disparate biological systems possessing large numbers of post-translationally modified glutamate residues. In both conantokins and mammalian blood clotting factors, the Gla residues are positioned every three to four amino acids along the primary sequence. The novelty of our work lies in taking advantage of this ancient drug design strategy of Gla spacing toward engineering peptide analogs with improved physicochemical and pharmacological properties. Strikingly, i,iϩ4 stapling appeared the most effective in retaining biological activity compared with the i,iϩ7 stapling  motif. Conantokins are known to be highly specific antagonists of NMDA receptors, with conG being a potent antagonist of the NR2B subtype. All of the stapled analogs retained potency and selectivity toward NR2B-containing NMDA receptors. Of particular interest is that three of four analogs showed exclusive selectivity for NR2B, making them useful tools to differentiate between other subtypes of NMDA receptors. Unexpectedly, conG[10 -14,S i,iϩ4 S(8)] showed a 3-fold increase in selectivity toward the NR2D subtype compared with native conG.
Stapled analogs were shown to be active following central administration in the 6-Hz mouse model of pharmacoresistant epilepsy. Because NMDA receptor targets of conantokins have been strongly implicated in epileptogenesis (34,42,51), the pharmacological activity of the stapled conG analogs was speculated to result from antagonism of these receptor targets. Compared with the native peptide, stapled conG analogs possessed similar efficacy in suppressing seizure activities. Importantly, administration of stapled analogs resulted in little or no motor toxic effects. There was a high correlation between increased selectivity toward NMDA subtype and decreased motor toxicity. Although the greatest degree of variability in the results was observed for conG[10 -14,S i,iϩ4 S(8)], these data suggest that stapling improves in vivo pharmacological properties of conantokins. Also, it was known that ␥-carboxyglutamate residues in conantokins make them resistant to degradation by endogenous peptidases (48). Employing an in vitro serum stability assay, we determined the t1 ⁄ 2 of native conG to be Ͼ10 h.
Taken together, our work shows how "nature-guided" engineering of bioactive peptides may improve their properties. In this case, our findings warrant further use of dicarba bridges as means to manipulate receptor selectivity and pharmacological profiles for conantokins.