A Maurotoxin with Constrained Standard Disulfide Bridging INNOVATIVE STRATEGY OF CHEMICAL SYNTHESIS, PHARMACOLOGY, AND DOCKING ON K CHANNELS*

Sarrah M’Barek, Ignacio Lopez-Gonzalez, Nicolas Andreotti, Eric di Luccio, Violeta Visan, Stephan Grissmer, Susan Judge, Mohamed El Ayeb, Hervé Darbon, Hervé Rochat, François Sampieri, Evelyne Béraud, Ziad Fajloun, Michel De Waard, and Jean-Marc Sabatier From the Laboratoire International Associé d’Ingénierie Biomoléculaire, CNRS Unité Mixte de Recherche 6560 Bd Pierre Dramard, 13916 Marseille Cedex 20, France, Inserm EMI 9931, Commissariat à l’Energie Atomique, Institut Fédératif de Recherche 27, Département de Recherche Dynamique Cellulaire, Canaux Ioniques et Signalization, 17 rue des Martyrs, 38054 Grenoble Cedex 09, France, Universität Ulm, Albert Einstein-Allee 11, D-89081 Ulm, Germany, the University of Maryland School of Medicine, Baltimore, Maryland 21201-1509, the Laboratoire des Venins et Toxines, Institut Pasteur de Tunis, P. O. Box 74, 1002 Belvédère, Tunis, Tunisia, Architecture et Fonction des Macromolécules Biologiques, CNRS Unité Propre de Recherche 9039, 31 Chemin Joseph Aiguier, 13402 Marseille, France, and the Laboratoire d’Immunologie, Faculté de Médecine Timone, 27 Bd Jean Moulin, 13385 Marseille Cedex 5, France

sumptive non-implication regarding ion channel recognition, one cannot rule out that they not only affect half-cystine pairings but also the spatial distribution of key toxin amino acid residues.
In the present work, we examined for the first time the sole contribution of the disulfide bridge arrangement on toxin pharmacology by maintaining unchanged the MTX primary structure. The aim was to produce and characterize an MTX variant adopting conventional Pi1-like disulfide bridges, referred to as MTX Pi1 (Fig. 1A). For this purpose, we developed an innovative strategy of solid-phase peptide synthesis based on a temporary chemical modification of the side chain of a trifunctional MTX amino acid residue (Tyr 32 ) expected to guide the type of toxin half-cystine pairings (Fig. 1B). We focused on the Tyr 32 residue because formation of the short Cys 31 -Cys 34 (C7-C8) disulfide bridge (referred to as a 14-member disulfide ring) is very sensitive to local steric hindrance. Indeed, we have previously shown that replacement of the side-chain hydrogen atom of Gly 33 by a larger methyl group of Ala prevents the C7-C8 connection and forces the corresponding synthetic MTX analogue ([Ala 33 ]-MTX) to adopt the conventional Pi1-like disulfide bridge arrangement of other ␣-KTx6 members (18). This novel pattern of half-cystine pairings (C1-C5, C2-C6, C3-C7, and C4-C8) comprises the reorganization of two of the four disulfide bridges (C3-C7 and C4-C8, versus C3-C4 and C7-C8). The following approach was experimentally developed to produce MTX Pi1 (Fig. 1B). (i) A classic stepwise solid-phase assembly of MTX peptide chain using a combination of Fmoc chemistry and t-butyl-type side-chain protecting groups for trifunctional amino acid residues (19). In the case of Tyr 32 , a more acidresistant 2,6 dichloro-benzyl group was used to protect its phenol ring. (ii) A trifluoroacetic acid treatment to remove all t-butyl-type protecting groups and to cleave the peptide from the resin. (iii) An oxidative folding of the Tyr 32 -protected MTX (MTX Tyr ), and (iv) a final trifluoromethanesulfonic acid (TFMSA) treatment of the folded/oxidized MTX Tyr to remove the Tyr 32 side-chain protecting group, thereby generating MTX Pi1 (20). Using this procedure, we succeeded in the chemical production of MTX Pi1 and demonstrate that its novel, but conventional, disulfide bridging is accompanied by marked differences in toxin properties.

Materials
N-␣-Fmoc-L-amino acids, Fmoc-amide resin, and reagents used for peptide synthesis were obtained from PerkinElmer, except N-␣-Fmoc-L-Tyr(2,6 dichloro-benzyl)-OH, which was from Fluka. Solvents were analytical-grade products and purchased from SDS. Enzymes (trypsin and chymotrypsin) were obtained from Roche Applied Science.

Chemical Synthesis and Characterization of MTX Pi1
The MTX Pi1 variant was assembled by the solid-phase technique (19) using a peptide synthesizer (Model 433A; Applied Biosystems Inc.). Peptide chains were assembled stepwise on 0.35 milliequivalent of Fmoc-amide resin (0.66 milliequivalent of amino group/g) using 1 mM Fmoc amino acid derivatives. The side-chain protecting groups used for trifunctional residues were: trityl for Cys, Asn, and Gln; t-butyl for Ser, Tyr, Thr, and Asp; pentamethylchroman for Arg; t-butyloxycarbonyl for Lys; and 2,6 dichloro-benzyl for Tyr in position 32. The Fmoc-amino acid derivatives were coupled (20 min) as their hydroxybenzotriazole active esters in N-methylpyrrolidone (2.8-fold excess). The fully protected peptide resin (2.5 g) was treated for 2.5 h at 25°C with a mixture of trifluoroacetic acid/H 2 O/thioanisole/ethanedithiol (73:11:11:5, v/v) in the presence of crystalline phenol (2.5 g). Under this condition, the peptide is cleaved from the resin and all its side-chain protecting groups are removed, except the 2,6 dichloro-benzyl protecting group of Tyr 32 . After filtration of the mixture, the crude Tyr 32 -protected peptide (MTX Tyr ) was precipitated and washed by adding cold diethyloxide. The crude MTX Tyr was pelleted by centrifugation (3,000 ϫ g; 10 min), and the supernatant was discarded. The reduced MTX Tyr was then dissolved at 2 mM concentration in 0.2 M Tris-HCl buffer, pH 8.3, and stirred under air to allow oxidative folding (72 h, 25°C). The folded/oxidized MTX Tyr peptide was purified by reversed-phase high-pressure liquid chromatography (HPLC) (PerkinElmer, C 18 Aquapore ODS 20 M, 250 ϫ 10 mm) by means of a 60-min linear gradient of 0.08% (v/v) trifluoroacetic acid/0% to 35% acetonitrile in 0.1% (v/v) trifluoroacetic acid/H 2 O at a flow rate of 5 ml/min ( ϭ 230 nm). The purified oxidized MTX Tyr (10 mg) was treated for 10 min at 25°C with a 10.25-ml mixture of trifluoroacetic acid/H 2 O (97.5:2.5, v/v) in the presence of crystalline phenol (1.5 g) and p-cresol (1 g). Next, the mixture was chilled on ice before addition of 1 ml of neat trifluoromethanesulfonic acid (20). The new mixture was then incubated for an additional 20 min to remove the 2,6 dichloro-benzyl protecting groups from oxidized MTX Tyr , thereby yielding MTX Pi1 . The peptide was then filtrated, precipitated, and washed as described for MTX Tyr after the first trifluoroacetic acid cleavage. The homogeneity and identity of MTX Pi1 was assessed by: (i) analytical C 18 reversed-phase HPLC, (ii) amino acid analysis after acidolysis, (iii) Edman sequencing, (iv) mass determination by matrixassisted laser desorption ionization-time of flight mass spectrometry, and (v) enzyme-based cleavage for half-cystine pairing determination. B, strategy used for the chemical synthesis of MTX Pi1 . The phenol ring of Tyr 32 from reduced MTX remains protected with the 2,6 dichlorobenzyl group after trifluoroacetic acid treatment (reduced MTX Tyr ) of the MTX peptide resin, whereas the side-chain protecting groups (tbutyl-type denoted X) of other trifunctional amino acid residues are removed. The reduced MTX Tyr folds/oxidizes to yield the oxidized MTX-Tyr with Pi1-like half-cystine pairings. Removal of the 2,6 dichlorobenzyl group by TFMSA treatment of the folded/oxidized MTX Tyr generates MTX Pi1 .

Assignment of Half-cystine Pairings of MTX Pi1 by Enzyme-based
Cleavage and Edman Sequencing Analysis MTX Pi1 (800 g) was incubated with a mixture of trypsin and chymotrypsin at 10% (w/w) in 0.2 M Tris-HCl, pH 7.4, for 12 h at 37°C. The resulting peptide fragments were then purified by reversed-phase HPLC (Chromolith RP18, 5 M, 4.6 ϫ 100 mm) with a 60-min linear gradient of 0.08% (v/v) trifluoroacetic acid /0 -60% acetonitrile in 0.1% (v/v) trifluoroacetic acid/H 2 O at a flow rate of 1 ml/min ( ϭ 230 nm) and freeze-dried prior to their analyses. These peptide fragments were hydrolyzed by acidolysis (6 N HCl/phenol), b and their amino acid contents were determined (System 6300 amino acid analyzer; Beckman). The fragments were further characterized by mass spectrometry analysis (RP-DE Voyager; Perseptive Biosystems) and Edman sequencing using a gas-phase microsequencer (Applied Biosystems 470A). In standard HPLC conditions for analyzing phenylthiohydantoin (PTH) amino acid derivatives, diPTH-cystine elutes at a retention time of 9.8 min.

Circular Dichroism Analyses of MTX Pi1 , MTX, and Pi1
Circular dichroism (CD) spectra were obtained on a Jasco J-810 spectropolarimeter equipped with a PTC-423S thermostat. A ratio of 2:20 was found between the positive CD band at 290.5 nm and the negative band at 192.5 nm. CD spectra were reported as the absorption coefficient (⌬⑀) per amide. The far UV CD spectra were acquired at 20°C in H 2 O between 185 and 260 nm using a 0.1-cm path length cell. Data were collected twice at 0.6-nm intervals with a scan rate of 50 nm/min. As assessed by amino acid analysis, the concentration of MTX Pi1 , MTX, or Pi1 was 40 nM.
Toxin Docking on Voltage-gated K ϩ Channels Atomic Coordinates-Atomic coordinates of MTX was obtained from the Swiss Protein Data base (Swiss-Prot www.expasy.ch) (number 1TXM).
Molecular Modeling-Molecular modeling of the S5-H5-S6 portions of rat K v 1.1, K v 1.2, K v 1.3, and Drosophila Shaker B channels was achieved on the basis of the crystal structure of the KcsA channel solved at a resolution of 3.2 Å (Swiss-Prot number 1BL8). The three-dimensional structural models of these channels were generated by using KcsA as a template and with the biopolymer homology modeling software of Swiss-model/Deep view 3.7 (Swiss-Prot, Switzerland). Amino acid sequence alignments between KcsA and K v 1.1, K v 1.2, K v 1.3, or Shaker B channels, which were generated by using CLUSTALW (V.1.82, www.ebi.ac.uk/clustalw/), showed that homologies are 69.8, 70.1, 69.1, and 65.6%, respectively. To avoid steric overlaps and clashes, modeled side chains and C␣ backbones of K ϩ channels were subjected to energy refinement (until ⌬⑀⌭ Ͻ 0.05 kJ⅐mol Ϫ1 ⅐Å Ϫ1 ) using, successively, steepest-descent, conjugate gradient, and Newton Raphson algorithms, with the consistent valence force-field as implemented in the INSIGHT II Discover3 module (1998 release, Molecular Simulations Inc., ACCEL-RYS, San Diego, CA). Root mean square deviation values between the KcsA template C␣ backbone and the modeled K v 1.1, K v 1.2, K v 1.3, and Shaker B C␣ backbones were 0.48, 1.93, 0.32, and 1.68 Å, respectively.
A molecular model of Pi1 was obtained on the basis of the threedimensional structure of MTX in solution (Swiss-Prot number 1TXM) by using the homology method of Swiss-Model/Deep view 3.7. Disulfide bridges were assigned using the Biopolymer module of InsightII. Similarly, this module was used to generate the molecular model of MTX Pi1 . Molecular models were relaxed by 5,000 steps of 1 fs of dynamics simulation at 15 K, then minimized by energy refinement (until ⌬⑀⌭ Ͻ 0.05 kJ⅐mol Ϫ1 ⅐Å Ϫ1 ) using the algorithms and force field previously described for K v 1.1, K v 1.2, K v 1.3, and Shaker B channels. Amino acid sequence alignment between Pi1 and MTX (CLUSTALW) points to 88.2% sequence homology. Root mean square deviation values between template MTX C␣ backbone and the modeled Pi1 and MTX Pi1 C␣ backbones were 1.33 and 1.05 Å, respectively. Geometric quality of all models was evaluated using PROCHECK V3.5.4 (21,22).
Protein Docking-Molecular interaction simulations were performed using the BiGGER program (bimolecular complex generation with global evaluation and ranking) (23). In the first step, a 1-Å three-dimensional matrix composed of small cubic cells, which represents the complex shape of each molecule, was generated. The translational interaction space was searched for each relative orientation of the two molecules by systematically shifting the probe matrix (toxin) to the target matrix (ion channel). 5,000 docking solutions were selected after probe rotation of 15°relative to the target, and this surface matching was repeated until a complete non-redundant search was achieved. The algorithm used by BiGGER performs a complete and systematic search for surface complementarities (both geometry complementarities and amino acid residue pairwise affinities are considered) between two potentially interacting molecules and enables an implicit treatment of molecular flexibility. In the second step, the 5,000 putative solutions were ranked according to four different interaction terms: surface matching, side-chain contacts, electrostatic, and solvation energies combined into a global scoring function.
Docking Solution Screening-The 15 best solutions were selected according to (i) the global score from BiGGER, (ii) toxin Lys residue (Lys 23 for MTX and MTX Pi1 , or Lys 24 for Pi1) and ␤-sheet strand orientations toward the ion channel pore, and (iii) the best orientation, considering the electrostatic properties of both the toxin and the K ϩ channel. The GRASP software (24) was used to determine these electrostatic properties (GRASP; Howard Hughes Medical Institute, Columbia University, New York).
Structural Refinement of the Final Complexes-The screened docking solutions were minimized with a rigid-body method (C␣-locked) with steepest-descent algorithms using Deep-view V3.7 (until ⌬⑀⌭ Ͻ 0.05 kJ⅐mol Ϫ1 ⅐Å Ϫ1 ) with a GROMOS96 force field (25) to relieve possible steric clashes and overlaps. During structural refinement, a distancedependent dielectric constant of 4 was used.
Docking Energy Calculations-Final docking energy of each best solution (⑀ toxinϪchannel Ϫ⑀ toxin ϩ ⑀ channel )) was obtained by subtracting the sum of toxin energy alone (⑀ toxin ) and ion channel energy alone (⑀ channel ), after rigid body minimization (C␣ϪC␣ distances locked) until ⌬⑀⌭ Ͻ 0.05 kJ⅐mol Ϫ1 ⅐Å Ϫ1 (GROMOS96 force field) (25), from the final complex energy (⑀ toxinϪchannel ) minimized under identical conditions. Close Interaction Analyses-Details of interactions were analyzed using the LIGPLOT program (26) on each best docking solution given by the screening method.
Linear Regression-Linear regression was computed using the Prism software (GraphPad Prism version 3.0cx for MacOS X; GraphPad Software, San Diego, CA; www.graphpad.com).

Neurotoxicity of MTX Pi1 and MTX Tyr in Mice
The peptides were tested in vivo for toxicity by determining the LD 50 after intracerebroventricular injections into 20 g of C57/BL6 mice (animal testing agreement number 006573, delivered by the Ministère de l'Agriculture et de la Pêche). Groups of six mice per dose were injected with 5 l of MTX Pi1 solution containing 0.1% (w/v) bovine serum albumin and 0.9% (w/v) sodium chloride.

Competitive Inhibition of 125 I-Apamin Binding onto Rat Brain
Synaptosomes by MTX Pi1 , MTX Tyr , and MTX Rat brain synaptosomes were prepared as described by Gray  volume. The incubation buffer was 25 mM Tris-HCl, 10 mM KCl, pH 7.2. The samples were centrifuged, and the resulting pellets were washed three times in 1 ml of the same buffer. Bound radioactivity was determined by ␥ counting (Packard Crystal II). The values expressed are the means of triplicate experiments Ϯ S.D. Nonspecific binding, less than 8% of the total binding, was determined in the presence of an excess (10 nM) of unlabeled apamin.

Preparation and Electrophysiological Recordings of Xenopus Oocytes
Xenopus laevis oocytes at stages V and VI were prepared for cRNA injection and electrophysiological recordings. The follicular cell layer was removed by enzymatic treatment with 2 mg/ml collagenase IA (Sigma) in classic Barth's medium lacking external Ca 2ϩ . The cDNA plasmids were linearized with SmaI (Shaker B), NotI (rat K v 1.1), XbaI (rat K v 1.2), and EcoR1 (rat K v 1.3) and transcribed with either T7 or SP6 RNA polymerase (mMessage mMachine kit; Ambion). The cells were microinjected 1-2 days later with 50 nl of cRNA (0.1 g/l Shaker B, rat K v 1.1, rat K v 1.2, or rat K v 1.3 channels). To favor K ϩ channel expression, cells were incubated at 16°C into a defined nutrient oocyte medium (28) 2-6 days before current recordings. Oocyte currents were then recorded at 20°C by standard two-microelectrode techniques using a voltage-clamp amplifier (GeneClamp 500; Axon Instruments) interfaced with a 16-bit AD/DA converter (Digidata 1200A; Axon Instruments). Electrodes filled with 140 mM KCl had an electric resistance of 0.5-1 M⍀. Voltage pulses were delivered every 15 s from a holding potential of Ϫ80 mV. Current records were sampled at 10 kHz and low pass-filtered at 2 kHz using an eight-pole Bessel filter and stored on computer for subsequent analysis. The extracellular recording solution contained (in mM): 88 NaCl, 10 KCl, 2 MgCl 2 , 0.5 CaCl 2 , 0.5 niflumic acid, 5 HEPES, 0.1% bovine serum albumin, pH 7.4 (NaOH). Leak and capacitive currents were subtracted on-line by a P/4 protocol. Residual capacitive artifacts were blanked for display purposes. Toxin solutions were superfused in the recording chamber at a flow rate of 2 ml/min using a ValveBank4 apparatus (Automate Scientific Inc.). The results are presented as mean Ϯ S.D.

RESULTS AND DISCUSSION
Solid-phase Synthesis and Physicochemical Characterization of MTX Tyr and MTX Pi1 -Stepwise assembly of MTX Tyr was achieved by means of Fmoc/t-butyl chemistry (19). For Tyr 32 , we used the more acid-resistant, but TFMSA-sensitive, 2,6 dichloro-benzyl side-chain protecting group that is not cleaved by trifluoroacetic acid treatment. A double coupling strategy was applied with Fmoc-amino acid hydroxybenzotriazole active esters. The yield of assembly ranged between 80 and 90%. Fig.   2 illustrates the elution profiles by C 18 reversed-phase HPLC of MTX Tyr and MTX Pi1 at different steps of the synthesis: crude reduced MTX Tyr after trifluoroacetic treatment (A), crude oxidized MTX Tyr after oxidative folding (B), MTX Pi1 resulting from TFMSA treatment of MTX Tyr (C), and purified MTX Pi1 (D). These data suggest that the chemical strategy elaborated to synthesize MTX Pi1 appears to be successful. However, a careful physicochemical characterization was required, especially to formally establish that MTX Pi1 exhibits the expected Pi1-like disulfide bridging.
First the relative molecular mass of purified MTX Pi1 was verified by matrix-assisted laser desorption ionization-time of flight mass spectrometry analysis (Fig. 3A). An experimental M r (MϩH) ϩ value of 3613.1 was obtained for MTX Pi1 , in close agreement with its deduced M r (MϩH) ϩ of 3613.3. As expected, this experimental value also agrees with the experimental M r (MϩH) ϩ of 3613.3 obtained for MTX (29). According to amino acid analysis after acidolysis of MTX Pi1 , the amino acid ratios were similar to the deduced values (Fig. 3B). The primary structure of MTX Pi1 was further verified by Edman sequencing (data not shown). To establish the half-cystine pairings of MTX Pi1 , the folded/oxidized peptide was treated with a mixture of trypsin and chymotrypsin. As shown in Fig. 3C, the data demonstrate that, contrary to MTX, MTX Pi1 exhibits half-cystine pairings between Cys 3 -Cys 24 , Cys 9 -Cys 29 , Cys 13 -Cys 31 , and Cys 19 -Cys 34 (which corresponds to the standard C1-C5, C2-C6, C3-C7, and C4-C8 Pi1-like pairings). Thus, as expected, MTX Pi1 differs from MTX by the two last disulfide bridges (C3-C7 and C4-C8, instead of C3-C4 and C7-C8) and adopts a conventional pattern of disulfide bridging that is identical to those of other characterized ␣-KTx6 toxins (Fig. 3D).
Structural Properties of MTX Pi1 -The CD spectrum of MTX Pi1 was recorded to assess its secondary structures and was compared with the CD spectra of MTX and Pi1 (Fig. 4A). Measurements were performed at a wavelength ranging from 185-260 nm. The data obtained correspond essentially to -* and n-* transitions of the amide chromophores of the peptide backbones (30). The CD spectra show large negative contributions between 207 and 230 nm and large positive contributions around 190 nm, indicating the presence of both ␣-helical and ␤-sheet structures. These data are consistent with peptide backbone folding according to ␣/␤ scaffolds (12) for MTX, MTX Pi1 , and Pi1. However, the CD spectra analyses do not point to obvious structural changes between MTX Pi1 and MTX. For the sake of comparison with the three-dimensional structure of MTX (11), we therefore generated a computed molecular model of MTX Pi1 . This model was obtained using MTX as a template; it was relaxed, minimized, and validated as described under ''Experimental Procedures.'' As shown in Fig. 4B, the C␣ backbone of MTX Pi1 does not differ markedly from that of MTX despite the important differences in half-cystine pairings. In contrast, a detailed examination of the side chains of a number of trifunctional amino acid residues suggests some marked differences in their orientations (Fig. 4C). These structural changes may nevertheless be sufficient to significantly impact peptide pharmacology.
Pharmacology of MTX Pi1 -MTX Pi1 was tested in vivo for neurotoxicity by intracerebroventricular injections in C57/BL6 mice. It is lethal in mice, with an LD 50 value of 90 ng/mouse. In comparison, the LD 50 values of MTX (2) and Pi1 (31) are 80 and 200 ng per mouse, respectively. MTX Pi1 remains as fully active as MTX in vivo, indicating that both disulfide bridge patterns provide peptides of equipotent lethality. Interestingly, oxidized MTX Tyr , the intermediate reaction product that generates MTX Pi1 upon TFMSA treatment, is ϳ9-fold less potent than MTX Pi1 for lethal activity in mice, with an LD 50 value of 800 ng/mouse. This result suggests that the integrity of Tyr 32 is key to expression of MTX Pi1 lethality.
To investigate the pharmacology of MTX Pi1 , we first tested its ability to compete with 125 I-apamin for binding onto SK channels of rat brain synaptosomes (Fig. 5). MTX Pi1 inhibits 125 I-apamin binding with an IC 50 value of 17.4 Ϯ 5.6 nM. It is about 4-fold less potent than MTX, which exhibits an IC 50 value of 4.4 Ϯ 3.1 nM, in agreement with previous data (2). Therefore, the disulfide bridge pattern of the peptide (MTXtype versus Pi1-type) appears to mildly affect its binding onto rat brain apamin-sensitive SK channels. Additionally, the presence of the 2,6 dichloro-benzyl protecting group on Tyr 32 sig- nificantly decreased the ability of the peptide to compete with 125 I-apamin for binding to SK channel (IC 50 value of 2.6 Ϯ 0.3 M; 150-fold less potent).
Next, we tested the effects of MTX Pi1 and MTX Tyr onto Shaker B, rat K v 1.1, K v 1.2, and K v 1.3 expressed in Xenopus oocytes, because they are the regular targets of MTX (1, 2). As shown in Fig. 6A, MTX Pi1 blocks Shaker B K ϩ outward currents with high affinity. The peptide starts to be active at 10 pM concentration and achieves the highest current block (97.7%) at about 10 nM. The effect of 10 nM MTX Pi1 is readily reversible upon washout of the peptide (Fig. 6B). The effect of MTX Pi1 is concentration-dependent with an IC 50 value of current inhibition of 0.24 Ϯ 0.12 nM (n ϭ 63; Fig. 6C). This should be compared with the effect of MTX, which acts on Shaker B channels with an IC 50 of 3.4 nM (9) in identical experimental conditions. These data indicate that the MTX peptide is ϳ14fold more potent in binding onto Shaker B channels when reticulated with Pi1-like half-cystine pairings rather than with its wild-type pairings. For rat K v 1.2 channels, the extent of K ϩ current blockage by MTX Pi1 is maximal with an IC 50 value of 2.8 Ϯ 2.1 nM. Compared with MTX, these values correspond to a 46-fold reduction in affinity but to an increase of about 30% in the extent of blockage (16). Altogether, the data obtained for Shaker B and K v 1.2 channels suggest that the change in disulfide bridging of the MTX peptide is accompanied, not only by modifications in affinity, but also by changes in the combined efficacy of ionic pore occlusion and K ϩ efflux by the peptide. This analysis is reinforced by examining the effect of MTX Pi1 on rat K v 1.3 K ϩ currents (Fig. 6E). MTX Pi1 interacts with K v 1.3 channels with an IC 50 value of 102 Ϯ 37 nM (n ϭ 63), which represents a 3-fold increase in affinity as compared with MTX (16). Interestingly, MTX Pi1 also blocks the K ϩ efflux to a greater extent (83 Ϯ 4%) than MTX (ϳ20%). A similar change in blocking efficacy toward K v 1.3 channel had already been observed with a three-disulfide-bridged MTX analog (16), suggesting that the peptide half-cystine pairing pattern may significantly affect ion channel pore occlusion. Finally, we also investigated the effect of MTX Pi1 on rat K v 1.1 K ϩ currents (Fig.  6F) and found it to be mostly inactive, as reported for MTX (16).
To get some insight on the contribution of Tyr 32 residue to MTX Pi1 pharmacology, we also investigated the effects of folded/oxidized MTX Tyr on the various voltage-gated K ϩ channels (Fig. 7). Interesting marked differences in the pharmacological properties of this peptide were observed, as compared with those of MTX Pi1 . Tyr 32 appears to be key with regard to MTX Pi1 affinity for Shaker B channel but not for the extent of K ϩ current blockage (Fig. 7, A and B). Indeed, with an IC 50 value of 1,229 Ϯ 41 nM (n ϭ 70), the folded/oxidized MTX Tyr is about 5,000-fold less potent than MTX Pi1 for K ϩ channel interaction. For rat K v 1.2 channels, an inverted situation is observed (Fig.  7C). The IC 50 value obtained for MTX Tyr is grossly similar to that of MTX Pi1 (6.5 Ϯ 3.6 nM and 2.8 Ϯ 2.1 nM, respectively), contrary to the extent of current blockage, which is markedly decreased from 100% to 34 Ϯ 3% in the case of MTX Tyr . These findings further support a key role of MTX Tyr 32 residue for toxin effect on K v 1.2 channel, as reported previously (8). In contrast, the presence of a 2,6 dichloro-benzyl moiety on the Tyr 32 phenol ring has no significant impact on rat K v 1.3 (Fig.  7D) or K v 1.1 (Fig. 7E) K ϩ channel pharmacology.
Docking of MTX Pi1 onto Voltage-gated K ϩ Channels-We first performed a Blastp (V.2.2.5, us.expasy.org/tools/blast/) search against the whole Protein Data Bank to select the correct template to generate models of the S5-H5-S6 portions of rat K v 1.1, K v 1.2, K v 1.3, and Shaker B channels. The KcsA primary structure (Swiss-Prot number 1BL8) showed the best E-value score for all the voltage-gated K ϩ channels under consideration. In addition, CLUSTALW (V.1.82) amino acid sequence alignments indicate that KcsA channel is a premium template that presents sequence homologies of 69.8% (K v 1. by the PROCHECK software (V. 3.5.4). No amino acid residue was found to be in disallowed regions, thereby validating the structural properties of the models (data not shown). The three-dimensional structure of MTX (11) and the molecular models of both MTX Pi1 and Pi1 were used in docking experiments with the different models of voltage-gated K ϩ channels. Docking energies were calculated according to the procedures described under ''Experimental Procedures.' ' We first detailed the docking of MTX Pi1 on Shaker B channel as it exerts its highest affinity toward this K ϩ channel subtype (IC 50 value of 0.24 nM). Fig. 8 illustrates the amino acid residues of MTX Pi1 (Fig. 8A) that may interact with Shaker B channel residues (Fig. 8B), as identified according to docking simulation. It is worth noting that the Lys 23 and Tyr 32 residues of MTX Pi1 belong to the functional dyad that is reported to be crucial for toxin bioactivity (8,9,32).
Docking simulations suggest that MTX Pi1 and MTX possess similar overall interaction topologies. For example, the Lys 7 and Lys 23 residues share the same interacting residues on Shaker B channel (the pair Thr 406 and Val 408 for Lys 7 , and Thr 407 for Lys 23 ; data not shown for MTX). Interestingly, additional analyses show that MTX Pi1 possess specific molecular contacts (Asn 26 with Gly 404 and Asp 405 ) that are not observed in the MTX docking simulations. Moreover, MTX Pi1 seems to be more stabilized than MTX on Shaker B channel because of a greater number of molecular contacts with the outer loop domain (Glu 380 , Asn 381 , and Ser 382 ). One should note that Lys 27 of MTX Pi1 also interacts with Ser 379 of Shaker B channel, whereas Lys 27 of MTX does not interact with any ion channel amino acid residue. This may reasonably explain the 14-fold difference in IC 50 values observed experimentally for MTX Pi1 and MTX. Next, we correlated the docking energies of MTX Pi1 , MTX, and Pi1 on Shaker B channel with their experimentally observed IC 50 values (Fig. 9A). A high degree of correlation (r 2 ϭ 0.97) was observed between docking energies and IC 50 values, which validates our overall molecular modeling approach. It also indicates that more detailed investigations of the interaction between MTX Pi1 and Shaker B channel will be permitted.
On K v 1.1 channel, MTX Pi1 , MTX, and Pi1 are not significantly active. In agreement with these data, no satisfying docking solutions were obtained for these peptides. Thus, we next investigated the docking properties of the peptides on K v 1.2 channel (Fig. 9B). MTX Pi1 (IC 50 ϭ 2.8 nM) is, respectively, 46-and 6-fold less active than MTX (IC 50 ϭ 0.06 nM) and Pi1 (IC 50 ϭ 0.44 nM) on K v 1.2. Docking simulations indicate that MTX Pi1 , MTX, and Pi1 share basically a common interaction map with K v 1.2, although some subtle differences could be observed that may explain their distinct affinities. The Thr 4 , Lys 7 , and Asp 8 residues of both MTX and MTX Pi1 are in contact with identical amino acid residues, Gly 378 and Asp 379 , of the K v 1.2 ion channel pore (data not shown  (Fig. 9B).
Docking simulations performed with the three peptides on rat K v 1.3 channel correlate well with the actual peptide pharmacologies. Relatively low-scoring interactions between MTX Pi1 and K v 1.3 channel or MTX and K v 1.3 channel were found (data not shown), consistent with their experimental IC 50 values. In addition, computed data on Pi1 docking show the existence of very few contacts between Pi1 and the K v 1.3 channel. As for other docking simulations, a high degree of linear correlation (r 2 ϭ 0.99) was also obtained between the experimental IC 50 values and the docking energies of these peptides (Fig. 9C). In this study, we generated molecular models of the various voltage-gated K ϩ channels using the KcsA structure as template. The structure of a novel K ϩ channel (KvAP) from Aeropyrum pernix has recently been described, after our own structural analyses were completed (34). Of note, this K ϩ channel is voltage-dependent, contrary to the KcsA channel. However, a careful comparison of the pore regions (S 5 -H 5 -S 6 segments) of KvAP and KcsA reveals an almost perfect superimposition of the ␣-carbon traces of both channels. In addition, the selectivity filter is essentially identical to that of KcsA. Despite these marked structural similarities, we also generated molecular models of the Shaker B, K v 1.2, and K v 1.3 channels using the KvAP channel as template instead of KcsA. In each case, models generated using either KcsA or KvAP as template were identical (data not shown), thereby validating the use of KcsA as template in our study.
Concluding Remarks-In the present work, we show that, by using a particular strategy of solid-phase peptide synthesis, one can act on the final half-cystine pairing pattern of a reticulated peptide without altering its chemical structure by either mutations or chemical modifications of specific amino acid residues, or both. Docking experiments ease the understanding of the molecular basis of the toxin to ion channel recognition. In the case of voltage-gated K ϩ channels, it was generally well admitted that this recognition was solely based on the participation of amino acid residues from the toxin ␤-sheet structure. This does not appear to be the case because docking data suggest the contribution of amino acid residues belonging to distinct toxin structural domains (e.g. Ser 6 , Lys 7 , and Tyr 10 of MTX). From this study, the disulfide bridge organization of MTX contributes to its pharmacological action. Though the ␣/␤ scaffold is neither disrupted nor markedly altered, we highlighted some interesting differences in the relative orientation of the side chains of certain ''key'' amino acid residues. Therefore, it is likely that the most significant changes in pharmacological properties observed between MTX Pi1 and MTX may be in part attributed to the side chains of Arg 14 , Lys 27 , and/or Tyr 32 residues. In line with such a view, a recent study (33) based on Brownian dynamics simulations argues in favor of central roles played by Lys 27 and Tyr 32 residues of MTX in its recognition of the Kv1.2 channel.