NMR Solution Structures of (cid:1) -Conotoxin EVIA from Conus ermineus That Selectively Acts on Vertebrate Neuronal Na (cid:2) Channels* □ S

(cid:1) -Conotoxin EVIA, from Conus ermineus , is a 32-resi-due polypeptide cross-linked by three disulfide bonds forming a four-loop framework. (cid:1) -Conotoxin EVIA is the first conotoxin known to inhibit sodium channel inactivation in neuronal membranes from amphibians and mammals (subtypes rNa v 1.2a, rNa v 1.3, and rNa v 1.6), without affecting rat skeletal muscle (subtype rNa v 1.4) and human cardiac muscle (subtype hNa v 1.5) sodium channel (Barbier, J., Lamthanh, H., Le Gall, F., Favreau, P., E., Chen, H., N.,

The new ␦-conotoxin EVIA (␦-EVIA), 1 a 32-amino acid conopeptide isolated from the venom of Conus ermineus, is the first conotoxin demonstrated to inhibit sodium channel inactivation in neuronal membranes from amphibians and mammals (subtypes rNa v 1.2a, rNa v 1.3, and rNa v 1.6) without affecting rat skeletal muscle (subtype rNa v 1.4) and human cardiac muscle (subtype hNa v 1.5) sodium channel subtypes (1). This important recent discovery makes ␦-EVIA a unique tool to study the modulation mechanisms of neuronal Na ϩ channels. As a consequence, ␦-EVIA may also serve as a new lead molecule for the design of new drugs to treat neurological diseases characterized by defective nerve conduction, especially those causing an axonal demyelinization (2,3). Nerve conduction could be facilitated by specific inhibition of Na ϩ channel inactivation. The knowledge of the detailed three-dimensional structure is therefore the first step necessary to understand the structureactivity relationships of this new lead conotoxin.

and this work).
Despite the differences in the loop length between cysteine residues (especially for the loops 2-4), their three-dimensional solution structures reveal a common scaffold consisting of a small ␤-hairpin structure and several types of tight turns. All conotoxins of this group exhibit a rather well defined backbone conformation, stabilized by a number of hydrogen bonds located in the different secondary structures, and by the three disulfide bridges. The scaffold forms the classical "cystineknot" motif known for toxic and inhibitory polypeptides (29). A remarkable feature of ␦-EVIA is the length of the loop 2 is made of nine residues instead of six for the other conotoxins. A variability in the cystine-knot scaffold, both in sequence and length, was already described for the loop 4 of conotoxin -GS (27) or the ␦-atracotoxin-Hv1 (30, Fig. 1) but not for the loop 2.
We describe here the three-dimensional NMR structure in aqueous solution of ␦-EVIA. The specific difference with the solution conformation of the other members of the four-loop family of conotoxins is located in loop 2, which is characterized by an apparent low conformational order corroborated by a 1:1 cis/trans isomerism of the Leu 12 -Pro 13 peptide bond in slow exchange on the NMR chemical shift time scale. Both cis Leu 12 -Pro 13 and trans Leu 12 -Pro 13 bonds were separately solved based on two separate sets of experimental restraints. For comparison purposes, a Pro 13 3 Ala single mutant was also prepared. The NMR data for P13A mutation are consistent with a 12-13 peptide bond purely trans, whereas the global structure of the toxin is maintained. The binding of the P13A ␦-EVIA to rat brain synaptosomes as well as its activity, measured by intracerebroventricular injection to mice, is reduced about 2-fold.

EXPERIMENTAL PROCEDURES
Toxin synthesis-solid phase synthesis of the linear ␦-EVIA and P13A ␦-EVIA peptides were done with an ABI 430A synthesizer (PerkinElmer Life Sciences), using Fmoc chemistry with 1,3-dicyclohexylcarbodiimide/HOAt as the condensation reagent (31) and 0.1 mmol of Rink amide resin. After the last Fmoc-amino acid coupling, the Fmoc group was kept bound to the N-terminal amino acid of the linear peptide in order to simplify the read out of the HPLC profile during its purification. The N-␣-Fmoc linear peptide was cleaved from the resin and purified with a reverse phase preparative column (Vydac, 218TP, 250 ϫ 25 mm). The peptide was eluted by a linear gradient of A and B solvents, 30 -100% B in 70 min; A is an aqueous solution of 0.1% (v/v) trifluoroacetic acid, and B is an aqueous solution containing 60% (v/v) acetonitrile and 0.1% trifluoroacetic acid. After elution, the major component (151 mg), corresponding to the target Fmoc peptide (mass spectrum, m/z 3515.32), was deprotected by incubation (5 min, 22°C) with a mixture composed of N,N-dimethyl formamide (5 ml) containing 20% piperidine (v/v) and 1% 1,8-diazabicyclo [5.4.0]undec-7-ene (v/v) and 5 ml of mercaptoethanol (72 mmol). After deprotection, the medium was diluted with deionized water (final volume ϳ500 ml) and acidified to pH 4 with trifluoroacetic acid, and the linear peptide was desalted by reverse phase preparative HPLC, as described previously. A pool of the ␦-EVIA-containing fractions (ϳ100 ml) was diluted with deionized water (final volume, ϳ500 ml) containing the redox couple cysteine/cystine, 2/0.5 mmol. The volume was adjusted to 1 liter with degassed buffer (0.1 M NH 4 SO 4 , 0.1 M ammonium acetate, 1 mM EDTA, pH 8.5), and the solution was adjusted to pH 8.5 with NH 4 OH and renatured by incubation (48 h at 4°C, followed by 12 h at 22°C). The mixture of oxidized peptides was adjusted to pH 4 with trifluoroacetic acid, loaded on reverse phase preparative column (Vydac, 218TP, 250 ϫ 25 mm), and eluted with the acetonitrile gradient described previously. Based on HPLC co-elution experiments and electrospray ionization mass spectrometry, ␦-EVIA was identified as a minor product under the oxidation conditions used. The peptide was further purified with a semi-preparative column (Zorbax SB C18, 250 ϫ 9.4 mm).
Disulfide Pairing Assignment-␦-EVIA disulfide pairing pattern was determined after partial reduction with tris(2-chloroethyl)phosphate (40°C, 5 min), as reported previously (32). The reverse phase chromatographic profile of the mixture is shown in Fig. 2A. The fingerprint pattern displayed a huge peak of nonreduced ␦-EVIA and three intermediates with one or two open disulfide bonds (peaks A-C). The fully reduced peptide (SH) 6 eluted earlier than the nonreduced ␦-EVIA and the three intermediates. After the intermediates in peaks A-C were alkylated with a large excess of iodoacetamide and purified by reverse phase HPLC, the amino acid sequences of their Cys (S-carboxamidomethyl) derivatives were determined (Fig. 2B). Sequencing the Cys (Scam) derivative in peak A indicated SS bridging of Cys 3 and Cys 21 . Sequencing the Cys (S-cam) derivative in peak B revealed an increased signal for Cys (S-cam) at cycles 3, 10, 21, and 25. Thus, peak B has the disulfide bridge Cys 10 , Cys 25 in addition to the disulfide bridge Cys 3 , Cys 21 . Analysis of peak C yielded a significant increase in the Cys (S-cam) signal at cycles 10 and 25, due to the open disulfide bridge Cys 10 , Cys 25 . Therefore, we deduce SS bridging between Cys 20 , Cys 29 , in addition to the detected Cys 3 , Cys 21 and Cys 10 , Cys 25 disulfide bonds. The cystine framework of ␦-EVIA was summarized as Cys 3 , Cys 21 , Cys 10 , Cys 25 , and Cys 20 , Cys 29 .
Radioiodination and Binding Assays-␦-Conotoxin TxVIA was radioiodinated by using 1 nmol of toxin, 0.5 mCi of carrier-free Na 125 I in a potassium phosphate buffer, pH 7.25, containing H 2 O 2 (10 l of 1:50,000 solution) and lactoperoxidase (0.7 unit, EC 1.11.1.7 from bovine milk) for a 2-min incubation time. The monoiodotoxin was purified on a Vydac C18 column. Rat brain synaptosomes were prepared from adult Sprague-Dawley rats (300 g), according to the method described by Kanner (33). Equilibrium competition assays were performed using increasing concentrations of unlabeled toxins in the presence of a constant low concentration of the radioactive toxin. Competition binding experiments were analyzed by the program Kaleidagraph (Synergy Software) by using a non-linear Hill equation (for IC 50 determination). The K i values of ␦-EVIA were calculated by the equation K i ϭ IC 50 /(1 ϩ (L*/K d )) where L* is the concentration of the hot ␦-TxVIA, and K d is its dissociation constant (34). Standard binding medium composition was (in mM) as follows: choline Cl 130, CaCl 2 1.8, KCl 5, MgSO 4 0.8, HEPES 50, glucose 10, and 2 mg/ml bovine serum albumin. Following incubation for the designated times, the reaction was terminated by dilution with 2 ml of ice-cold wash buffer of the following composition (in mM): choline Cl 140, CaCl 2 1.8, KCl 5.4, MgSO 4 0.8, HEPES 50, pH 7.2, 5 mg/ml bovine serum albumin. Separation of free from bound toxin was achieved by rapid filtration under vacuum using Whatman GF/C filters preincubated with 0.3% polyethyleneimine. The filter discs were then rapidly washed twice with 2 ml of buffer. Nonspecific toxin binding was determined in the presence of a high concentration of the unlabeled toxin.
Biological Activity-To quantify the biological activity of native and synthetic conotoxins, Swiss-Webster mice (Ϸ15 g) were injected intracerebroventricularly with a stereotaxic system (Harvard/ASI Apparatus, UK). The ED 50 value was defined as the dose that produces hyperactivity in 50% of the tested animals within 12 h postinjection.
NMR Spectroscopy-␦-EVIA was dissolved at 2.2 mM (4 mg/550 l) in either 90% H 2 O, 10% D 2 O, or 100% D 2 O. pH was adjusted to 3.0 (direct uncorrected pH-meter reading) using microliter increments of 0.1 N HCl. The P13A analog was prepared at 0.6 mM in 90% H 2 O, 10% D 2 O, pH 3.0. NMR spectra were recorded on a Bruker Avance DRX 500 spectrometer using a 5-mm ( 1 H, 13 C, 15 N) triple-resonance probe head, equipped with a supplementary self-shielded z gradient coil. Spectra were processed using Bruker XWINNMR and GIFA V.4 (35) software. The solvent signal was suppressed with the WATERGATE sequence using a 3-9-19 pulse sequence with z gradient (36,37). The DQF-COSY experiment (38) was recorded at 283 K in H 2 O with a very low power presaturation of the water resonance during the recycle delay to minimize the radiation damping effect. TOCSY/HOHAHA experiments (39,40) were collected in both H 2 O and D 2 O at 283 K with a spin-lock time of 40 and 80 ms and at 290 and 297 K with a spin-lock time of 80 ms. An MLEV pulse sequence was used for the isotropic mixing arranged as the clean-TOCSY scheme (41). NOESY spectra (42,43) were collected at 283, 290, and 297 K with a 1.5-s recycle time and a 150-ms mixing time. Supplementary NOESY spectra were collected at 283 K in H 2 O and D 2 O with mixing time of 75 and 250 ms, respectively. In these NOESY experiments, a selective flip-back pulse was applied on water resonance, and gradients were added during t 1 to minimize the radiation damping effect (44). Except for the NOESY experiment at 283 K in H 2 O where 96 scans per t 1 increments were accumulated, spectra were registered with 512 (t 1 ) ϫ 1024 (t 2 ) complex data points and 32 scans per t 1 . The quadrature detection in the t 1 dimension was achieved using the States-TPPI method (45). Chemical shifts were quoted relative to the solvent (H 2 O) chemical shift at the respective temperature (4.92 ppm at 283 K). The spectral width of all experiments was 10.96 ppm (5482.5 Hz) with a carrier frequency on-resonance with the water resonance. For the P13A conotoxin ␦-EVIA, the TOCSY spectrum was collected with a spin-lock time of 80 ms, and NOESY spectrum was recorded with a 150-ms mixing time.
Deuterium exchange of labile protons was monitored just after dissolution of the lyophilized sample from H 2 O at pH 3.0 in neat D 2 O. The residual NH signal was followed with time at 283 K by analysis of TOCSY spectra recorded at 0.5, 1, 3, 19, 36, and 54 h.
Experimental NMR Restraints-Interproton-distance constraints were classified into four categories according to the cross-peak intensity of the NOESY spectrum at 283 K and 150-ms mixing time in H 2 O and D 2 O. Upper bounds were fixed at 2.7, 3.3, 5.0, and 6.0 Å for strong, medium, weak, and very weak correlations, respectively. The lower bound for all restraints was fixed at 1.8 Å, which corresponds to the sum of the hydrogen van der Waals' radii. The calibration for the NOE intensities was achieved using the cross-peak intensity H␦-H⑀ of Tyr 7 . Pseudo-atom corrections (46) of the upper bounds were applied for magnetically equivalent aromatic protons (ϩ2 Å) and unresolved diastereotopic methyl or methylene protons (ϩ 1 Å). 0.3 Å was added to NOEs involving amide protons. Dihedral angle restraint 1 was deduced from the stereospecific assignments of diastereotopic ␤-protons (Ϯ40°from the ideal staggered conformation) (47,48). The dihedral angle restraints were set to Ϫ65 Ϯ 25°for 3 J HN-Ha Ͼ8 Hz and to Ϫ120 Ϯ 40°for 3 J HN-Ha Ͻ5 Hz. After the initial calculation, 22 distance restraints were applied to restrain 11 hydrogen bonds that were unambiguously defined by both the exchange rate and the NOEs, and by the fact that they appear in ϳ75% of the models. Target values of 1.

5-2.3 Å for NH(i)-O(j) and 2.5-3.3 Å for N(i)-O(j) were used.
Structure Calculations-Models were calculated following a protocol described previously (49) using the X-PLOR software version 3.851 (50). Structures were analyzed using aqua/procheck-nmr (51) and promotif (52) and displayed using MolMol version 2.4 (53) and Molscript (54). The structure was generated by the hybrid distance geometry dynamically simulated annealing method (55,56). In a first stage, the substructures generated using metric matrix distance geometry algorithms were regularized and refined by a high temperature simulated annealing protocol, using the parallhdg.pro force field of X-PLOR. The nonbonded van der Waals' interactions were represented by a simple repulsive quadratic term (55,57). The experimental distance restraints were represented as a soft asymptotic potential, and electrostatic interactions were ignored. The force constant associated with the distance restraints was kept to 50 kcal⅐mol Ϫ1 ⅐Å Ϫ2 throughout the protocol. One cycle of simulated annealing refinement consisted of 1,500 steps of 3 fs at 1,000 K followed by 3,000 cooling steps of 1 fs from 1,000 to 300 K. At the end, each structure was subjected to 1,500 steps of conjugate gradient energy minimization.
In the second stage of the calculation, structures with good experimental and geometric energies were further refined using the full CHARMM22 force field of X-PLOR. In this stage of the calculation, the non-bonded interactions such as electrostatic interactions and van der Waals' interactions (described by the Lennard-Jones empirical energy function) were taken into account. An approximate solvent electrostatic screening effect was introduced by using a distance-dependent dielectric constant and by reducing the electric charges of the formally charged amino acid side chain (Asp and Lys) to 20% of its nominal charges defined in the CHARMM22 force field. The force constant used for the NOE potential was reduced to 25 kcal⅐mol Ϫ1 ⅐Å Ϫ2 . After 1,500 steps of conjugate gradient energy minimization, the dynamic was initiated at 750 K. The system was equilibrated for 0.5 ps with an integration step of 1 fs and then coupled to a heat bath at 750 K, and the molecule was allowed to evolve for 10 ps before a slow cooling to 300 K for a period of 5.4 ps and allowed to evolve again for 15 ps. Finally, the structures were energy-minimized by 1,500 steps of the conjugate gradient algorithm.

RESULTS
␦-EVIA was synthesized using Fmoc chemistry, based on its amino acid sequence. The N ␣ -Fmoc linear/denatured and reduced peptide was obtained in high yield (151/350 mg of calculated Fmoc peptide). Its folding/renaturation was the limiting step in the synthesis of bioactive ␦-EVIA. Typically, the folding procedure yielded only 5-10 mg of synthetic and bioactive ␦-EVIA (3.3-6.6% yield, based on 151 mg of Fmoc linear peptide). The HPLC pattern of the preparation obtained after the folding procedure indicated the presence of a large quantity of monomer isomass peptides (data not shown), probably because of misfolding of the linear, reduced peptide. Synthetic ␦-EVIA co-eluted with native ␦-EVIA (Fig. 2C), and the predicted MW of synthetic ␦-EVIA was confirmed by ESI-MS (observed MH ϩ 3288.1; calculated MH ϩ 3288.4).
Sequence Resonance Assignments of the Spin Systems-From the DQF-COSY and TOCSY spectra, we first noticed the presence of 53 amide protons (H N ), instead of the 30 expected. The sequence-specific resonance assignment (46) was undertaken and revealed two distinct NOESY walking with only Asp 1 , Asp 2 , Lys 5 , Hyp 6 , Gly 8 , Cys 21 , and Leu 32 common to the two systems. The d ␣N connectivities proceeded unambiguously for the two sets of resonances except for Cys 21 /Ser 22 . Tyr 7 , Phe 9 , Pro 13 , Ala 24 , and Ala 30 spin systems were used as starting points for the sequential assignment due to their specific pattern. In one of the two systems, the Leu 12 -Pro 13 peptide bond was unambiguously assigned to the cis peptide conformation due to the observation of a strong d ␣ ␣ and a weak d ␣ ␦ in NOESY spectra. For the second system, the peptide bond conformation was unambiguously assigned to the trans conformation due to the observation of a strong d ␣ ␦ with no d ␣ ␣ . Then the two systems clearly resulted from a Leu 12 -Pro 13 cis/trans mixture of ␦-EVIA in slow exchange on the time scale of the NMR chemical shift. No exchange peaks between the two conformers were detected either in the TOCSY or the NOESY experiments (Fig. 3). The chemical shift differences (where ⌬␦ c-t indicates difference of the chemical shifts between cis and trans conformers) observed in the NMR spectra between ␦-EVIA 12-13cis and 12-13trans are the most important for the H N compared with the C ␣ H protons and very small for all the side chain protons but four of the nine residues in loop 2 (Ser 11 , Leu 12 , Pro 13 , and Asn 17 ) (see "Discussion"). By taking into account the experimental uncertainties and using the relative intensities of the TOCSY crosspeaks of the same proton pair in the two different conformers (the spin system of Cys 10 , Ser 22 , Gly 27 , and Cys 29 were the best resolved), a 1:1 ratio of each population was estimated. This ratio is not temperature-dependent between 283 and 313 K. The assignment of the Hyp 6 was straightforward with strong d ␤ ␦ and no d ␣ ␣ between Lys 5 and Hyp 6 , indicating that the peptide bond is in a trans conformation. However, the observation of d ␣ ␦ between Lys 5 and Hyp 6 could not be observed due to the superposition of the strong intra-residual NOE C ␣ H-C ␣ HЈ of Gly 23 . A number of cross-peaks were not found in the H Naliphatic region of the DQF-COSY spectrum. It was the case for Lys 5 , Lys 16 , Cys 20 , Cys 21 , and Ala 30 . The two sequence-specific resonance assignments of ␦-EVIA are given as Supplemental Material.
Secondary Structure and Molecular Topology-On the basis of the NMR data ( Fig. 4 and Fig. 5), we identified the presence of a small ␤-sheet composed of three short antiparallel strands, involving Gly 8 -Cys 10 (strand 1), Gly 23 -Val 26 (strand 2), and Val 28 -Asp 31 (strand 3). The double-stranded anti-parallel ␤-sheet comprising strands 2 and 3 are connected by a typical ␥-turn Val 26 -Val 28 and forms a ␤-hairpin structure. Finally, two ␤-turns including amino acids Lys 5 -Gly 8 and Cys 20 -Gly 23 were identified according to the standard criteria (58).
Structure Calculations-The input data for the structural calculation consisted of 206 distance and 24 dihedral restraints. The 206 distance restraints include 56 intraresidual, 68 sequential (i, i ϩ 1), 27 medium range (i, i ϩ 2 to i, i ϩ 4), and 55 long range (Ͼi, i ϩ 4) restraints (Fig. 5A) for the trans conformer. The number of distance restraints is a little different for the cis conformer with 134 and 56 distance restraints for the sequential and medium range restraints, respectively. Sixteen slowly exchanging amide protons were assigned from the D 2 O exchange experiment, and 11 of them were used to define 22 hydrogen-bond restraints (see under "Experimental Procedures"). These 11 with the 5 remaining slowly exchanging amide protons will be discussed below. Dihedral angle restraints included 15 1 angles and 9 angles.
Both cis and trans Pro 13 were modeled separately with the appropriate angle restraints in the simulated annealing calculations using parallhdg.pro force field. Thirty initial structures were calculated, leading to 16 and 18 final structures for the trans and the cis conformer, respectively. None of these models have an NOE violation exceeding 0.15 Å or a dihedral angle violation exceeding 5°. These structures were refined with the CHARMM22 force field and show a few distance violations lower than 0.25 Å and no dihedral angle violations exceeding 5°. The geometric and energetic statistics of these final structures, given in Table I, satisfy both the force fields and the experimental restraints. Side-by-side views of the final structures (Fig. 6A) show that residues 3-10 and 20 -30 (corresponding to loops 1, 3, and 4) have well defined backbone dihedral angles (Fig. 7, B, C, and E) and low r.m.s.d. for the backbone atoms (Fig. 7D), whereas loop 2 (residues 11-19) appears completely disordered. The two residues of the N-and C-terminal parts also exhibit higher disorder. This correlates with the small number of medium and long range NOEs for these parts of the molecule (Fig. 7A).
Analysis of the Hydrogen Bond Network-The hydrogen-exchange rates in D 2 O identified 16 slowly exchanging amide protons in the structure. It concerned the Ile 4 , Lys 5 , Tyr 7 , Gly 8 , Phe 9 , Cys 10 , Ile 14 , Leu 15 , Cys 21 , Ser 22 , Gly 23 , Val 26 , Val 28 , Cys 29 , and Asp 30 amide protons. Eleven of them were used as restraints in the structure calculation from which 7 were involved in the triple-stranded antiparallel ␤-sheet (Fig. 4). Two amide protons, Gly 8 and Gly 23  The four remaining slow exchanging amide protons (Lys 5 , Phe 9 , Ile 14 , and Leu 15 ) are more rapidly exchanged than the preceding ones, except for Lys 5 as shown in Fig. 5. The two first amide protons corresponding to the second and the last residues of the first loop are also slowly exchanged in -MVIIA and -SVIB conotoxins (25), but a posteriori no justification was given. The slow exchange of Ile 14 and Leu 15 amide protons, located in the disordered loop, could be due to the high hydrophobicity in this region (see "Discussion" below).
In addition, one ␥and two ␤-turns were characterized in the models. Based on the analysis of the and angles (60), the models describe a type II ␤ 5-8 and a type I ␤ 20 -23 turn (Fig.  7, B and C). Also, the hydrogen bond between Val 28 (H N ) and Val 26 (O) defined a classic ␥ 26 -28 turn (␥ c -turn).
The buried side chains correspond to residues Ile 4 , Hyp 6 , Ser 11 , Ser 22 , Ala 24 , Val 26 , Val 28 , and Ala 30 and the six cystines. When calculated on the heavy atoms of the Cys 3 -Cys 10 plus Cys 20 -Asp 30 segments, the r.m.s.d. with respect to the mean coordinate positions is only equal to 0.59 Å because of the restriction of 14 1 angles (Fig. 5). The 1 angles of all the Cys residues were restricted and thus contribute to stabilize the core of the molecule. As observed for the majority of the conotoxins, the Cys 10 -Cys 25 disulfide bridge is a right-handed spiral motif, whereas the Cys 20 -Cys 29 disulfide bridge is a lefthanded spiral motif (52). The third disulfide bridge Cys 3 -Cys 21 is less well defined due to the proximity of the undefined N-terminal part. It appears to be represented by two conformations, with 7 structures displaying a right-handed hook conformation and 9 structures having a non-standard conformation. The ordering of the three disulfide bridges with the three-stranded antiparallel ␤-sheet form the so-called "inhibitor cystine-knot motif " described by Pallaghy et al. (29).
NMR Analysis of the Pro 13 3 Ala ␦-EVIA-The P13A mutation restores a unique NMR spin system (given as Supplemental Material), and the sequential NOEs are characteristic of a trans peptide bond between Leu 12 and Ala 13 . The chemical shifts are very close in frequency to the trans wild-type toxin; this observation indicates a conserved global solution structure. The chemical shift similitude between the Leu 12 -Ala 13 peptide bond in P13A mutant and the Leu 12 -Pro 13 trans conformer in wild-type conotoxin is particularly noteworthy for the H N proton of Ile 14 , which has the highest difference with the cis isomer ( (⌬␦ c-t ϭ 0.73 ppm). Most of the observed chemical shift a F bond is the bond length deviation energy; F angle is the valence angle deviation energy; F impr deviation energy for the improper angles used to maintain the planarity of certain groups of atoms; F coulombic is the coulombic energy contribution to F total ; F L-J is the Lennard-Jones van der Waals energy function, and F NOE is the experimental NOE function calculated using a force constant of 50 and 25 kcal⅐mol Ϫ1 . Å Ϫ2 in the case of the allhdg.pro and the CHARMM22 force field, respectively. b In the case of the allhdg.pro force field, only the repulsion term is given. differences with the wild-type conotoxin are essentially located in the region of the Leu 12 -Ala 13 dipeptide (given as Supplemental Material).
Biological Activity Pro 13 3 Ala ␦-EVIA-Competition curves for 125 I-labeled ␦-TxVIA binding inhibition by ␦-EVIA and its P13A variant on rat brain synaptosomes revealed that both conopeptides, in a dose-dependent manner, fully displaced the radiolabeled conotoxin from its receptor site (Fig. 8). The calculated K i values were about two times higher for the P13A variant (1,100 Ϯ 140 nM, n ϭ 3) than for ␦-EVIA (475 Ϯ 75 nM, n ϭ 3). These results indicate that the P13A variant binds to receptor site 6 of voltage-dependent Na ϩ channels, like ␦-EVIA, but with lower affinity. Consistent with this finding, a 2-fold reduction in the ED 50 for the excitotoxic activity of the P13A variant was observed (following intracerebroventricular injection to mice), when compared with natural and synthetic ␦-EVIA (Table II). Furthermore, experiments performed on isolated frog neuromuscular preparations revealed that 5 M P13A analog selectively increased nerve terminal excitability, without affecting directly elicited muscle action potentials and excitability in muscle fibers (data not shown). These results indicate that the mutant toxin keeps it selectivity for neuronal Na ϩ channels but is less active than reported for ␦-EVIA (1). Fig. 2A, the retention time (R t ) for natively folded and biologically active ␦-EVIA was significantly longer than that of the linear, fully reduced peptide, in contrast to the observed R t for the natively folded hydrophilicor -conotoxin (20,61). This observation suggests (i) that the folding process increases the ␦-EVIAaccessible hydrophobic area and (ii) the hydrophobic character of ␦-EVIA likely contributes to the low yield folding process. The detergent-assisted oxidative folding process recently described by De la Cruz et al. (63) could therefore improve the ␦-EVIA folding yield. In this respect, we have observed a higher amount of folded P13A ␦-EVIA variant compared with that of the wild-type ␦-EVIA, suggesting a folding effect of Pro 13 .

Chemical Synthesis and Folding-As shown in
Comparison to Other Toxins-The three-dimensional structure of ␦-EVIA consists of a ␤-hairpin, involving residues Gly 23 to Asp 31 (Fig. 6B), and several turns. Such a conformation, stabilized by a number of disulfide bridges and hydrogen bonds, is adopted by several other small proteins (Fig. 1) coming from phylogenetically divergent species including spiders and cone shells but also fungi or plants (29). The r.m.s.d. of the three disulfide bridges with or without loops 1 and 3 was calculated for these different peptides relative to the ␦-EVIA (Table III). Despite the relatively low amino acid sequence homology (Fig. 1), it reveals a strong correlation between this four-loop family and the scaffold of the peptidic backbone. In particular, the loop 1, stabilized by a ␤-turn, and the disulfide bridges appear very close. The loop 3 is constituted of a type I (-conotoxin PVIIA, huwentoxin-I, and ␦-EVIA), type VIa (␦atracotoxin-Hv1), or type VIII (-conotoxin GVIA, -conotoxin MVIIA, and -agatoxin) ␤-turn.
The major structural differences between these molecules result from the different sizes of the loops 2 and 4 (Fig. 1). The resulting secondary structure in loop 2, except for peptides with nine amino acids, is a type I ␤-turn in the majority of the cases. ␦-EVIA, which is the first structurally resolved conotoxin with a large second loop, displays a disordered loop from Ser 11 to Leu 19 (Figs. 6A and 9). This loop is devoid of medium range and long range interactions with the rest of the molecule and therefore protrudes from the core of the molecule without a well defined orientation. Since the other conotoxins with such a large second loop (conotoxins MrVIA, MrVIB, and ␦-PuIA) also contain a Pro residue in the third position (Fig. 1), it can be supposed that the same cis/trans isomerism could occur for this residue. The earlier structural studies of the four-loop peptides all concern peptides with at least four amino acids in the loop 4. This loop is characterized by a type I (-conotoxin GVIA and -conotoxin PVIIA), type IЈ (-conotoxin MVIIA and -conotoxin MVIIC), or type VI (-conotoxin GS and -agatoxin) ␤-turn and is involved in a ␤-hairpin structure. Despite the presence of only three amino acids in the loop 4 of ␦-EVIA, the peptide backbone adopts the same topology between the strands 2 and 3, as shown in Fig. 9, with a similar hydrogen bond network, constraining in this case the hairpin turn to form a ␥-turn.
The cis/trans Isomerism of Pro 13 in ␦-EVIA-Considering the NMR ensemble of ␦-EVIA, the great disorder of loop 2 could correlate with the Leu 12 -Pro 13 peptide bond cis/trans isomerism. The absence of exchange peaks between the two conformers indicates, however, a very slow exchange rate relative to the NMR time scales. For instance, such a feature was already observed for the Lqh-8/6, a toxin from a scorpion venom (64), or for a cyclic peptide (65). Several statistical studies (66,67) tentatively suggested possible correlations between the steric and electronic properties of the residues surrounding the proline and the stability of the cis isomer, by examination of peptide bonds from the Protein Data Bank (68). According to these studies, a cis peptide bond in a sequence Leu 12 -Pro 13 -  Ile 14 of ␦-EVIA would not be frequent due to the presence of branched aliphatic residues preceding and following the proline (67).
In ␦-EVIA, amide protons of Ile 14 and Leu 15 slowly exchanged with D 2 O. Unfortunately, the H N of Ile 14 is degenerate in the trans conformer with the intra-residual C ␣ H/H N crosspeak of Gly 23 , so we cannot conclude about a similar exchange for this conformer (Fig. 4). Two additional weak NOEs, Leu 12 (C ␣ H)-Ile 14 (H N ), and Leu 12 (C ␣ H)-Ile 14 (C ␣ H), are found for the cis isomer relative to the trans isomer. Calculation of structures for the cis conformer induced the formation of a nonhydrogen-bonded type VIb-3 ␤-turn involving Ser 11 to Ile 14 (68,67), and a significant decrease of the disorder for the and angles of Leu 12 (Ϫ122 Ϯ 9 and 168 Ϯ 8°) and Pro 13 (Ϫ65 Ϯ 8 and 157 Ϯ 11°) (see Fig. 7, B and C, for comparison with the trans conformer). However, the presence of this ␤-turn is not sufficient to render ordered the entire loop 2. The slow exchange observed for the H N of Ile 14 and Leu 15 would then originate in the high local hydrophobicity (Leu 12 -Pro 13 -Ile 14 -Leu 15 ) rather than the formation of hydrogen bonds.
As a consequence of the cis/trans isomerism of Pro 13 , the ⌬␦ c-t is not uniformly distributed along the sequence (data given as Supplemental Material). Ile 14 HN is the most affected as well as HN of Asn 17 , Gly 18 , and Leu 19 located in front of Pro 13 in the loop 2. In addition to the backbone protons, the ⌬␦ c-t of the side chain protons of Cys 10 , Ser 11 , Leu 12 , Pro 13 , and especially Asn 17 residues are relatively high (0.10 to 0.30 ppm). The comparison of the structure of the two conformers in this region show more  Table III. B, surface-associated electrostatic potentials of conotoxin ␦-EVIA represented as a MolMol contact surface (53) from electronegative to electropositive by a red to blue continuous color range, respectively. The molecule is shown in the same orientation as in Figs. 7 and 8. Specific amino acids associated with loop 2 and/or the hydrophobic patch are indicated as well as charged amino acids.  hydrogen bonds for the cis conformer involving the side chain of Asn 17 . The ␦NH 2 protons are hydrogen-bonded in the models with Ser 11 (O␥), Ser 11 (O), and Pro 13 (O) in 6, 6, and 10 of the 18 final structures, and they are practically absent for the trans conformer. Moreover, half of the final structures for the cis conformer have a Leu 15 (H N )-Asn 17 (O␥) hydrogen bond, although it does not appear for the trans conformer. The expected ⌬G of the cis-to-trans interconversion for an isolated Xaa-Pro bond is in the order of 13 kcal/mol, and the cis/trans ratio was calculated to be about 1:3 (71)(72)(73). The additional stabilization in the cis conformer could explain the 1:1 cis/trans ratio observed. It is also likely that the additional hydrogen bonds in the cis conformer could increase the interconversion energy barrier leading to a very slow exchange rate.
The fact that the Pro 13 3 Ala mutant ␦-EVIA gives a better yield of the native structure upon the air-oxidation step of the free cysteines could be interpreted as a propensity of Pro 13 3 Ala ␦-EVIA to favor low potential energy intermediates, productive for the native structure. A prolyl residue at the position 13 would slow a folding step as observed recently in the case of the systematic study of the Pro 3 Ala mutations of the three prolines on the folding kinetics of the cellular retinoic acidbinding protein I (74).
Interaction with the Na ϩ Channel-As for Oand ␦-conotoxins targeting sodium channels, ␦-EVIA displays a high degree of hydrophobicity and a small number of charges (Figs. 1 and 10). This fact was recently discussed for the NMR structure of ␦-TxVIA (12), a 27-mer peptide that competes with ␦-EVIA for the same receptor site 6 on the Na ϩ channels (1) (Fig. 8). In particular, a characteristic of ␦-TxVIA is to bind strongly to the receptor site 6 of Na ϩ channels of mammals without modifying the spontaneous inactivation period, while it delays significantly this period in mollusks (75)(76)(77). A hydrophobic cluster was revealed on the surface ␦-TxVIA including, in particular, Met 8 , Leu 11 , Leu 12 , Tyr 20 , Val 23 , Leu 24 , and Val 25 . Among these seven amino acids, five are homologously conserved in ␦-EVIA (Phe 9 , Pro 13 , Ile 14 , Val 26 , and Val 28 , Fig. 1). The hydrophobic patch revealed on the surface ␦-TxVIA is even extended by Ile 4 , Tyr 7 , and Leu 12 as shown in Fig. 10. The positively charged amino acids are located at the beginning of the loop 1 (Lys 5 ) and in the loop 2 (Lys 16 ), whereas the negatively charged amino acids are located at the opposite N and C termini (Asp 1 , Asp 2 , and Asp 31 ), making ␦-EVIA quite amphipathic.
The apparent disorder observed in the loop 2 of ␦-EVIA, which includes Leu 12 , Pro 13 , and Ile 14 participating in the hydrophobic patch, could be important for receptor binding as it was suggested for the ␣A-conotoxin PIVA (78), which has a flexible 3-11 loop and a rigid 14 -23 loop. Such a flexible region has already been proposed to be stabilized upon interaction with the receptor (79 -81). In particular, a cis/trans proline isomerization was supposed to occur in the free form of the human acidic fibroblast growth factor (80), but when complexed to the sucrose octasulfate, the slow conformational motion corresponding to this isomerization is thwarted by this binding leading to a more rigid conformation. Also, conformational changes of small peptidic ligands upon binding have been observed in other systems (82)(83)(84)(85). It can be presumed that the loop 2, which is quite flexible in solution, might allow structural rearrangements to accommodate its binding, whereas the rigid core of ␦-EVIA serves as a delivery scaffold, with possible direct additional interactions with the receptor.
Due to the peculiar apparent low structural order and the hydrophobic character of the loop 2 from Leu 12 to Leu 15 , and to the specific cis/trans isomerism of the Leu 12 -Pro 13 peptide bond, another question arose about a possible biological and functional interest of such a structural singularity. The fact is that a trans, or a cis, amino-acyl proline peptide bond does not offer at all the same local binding abilities. In the trans conformation, the peptide carbonyl (a potent hydrogen bond acceptor) of the preceding residue, is trans to the hydrophobic cyclic side chain of the proline amino acid (i.e. on the two opposite sides of the main peptide chain). In the cis conformer, the carbonyl of the peptide bond is cis to the hydrophobic proline side chain (i.e. on the same side of the main peptide chain). In terms of molecular interactions between the neurotoxin and the receptor, it could make strong and significant differences. Such differences have been demonstrated, for instance, in a study related to an opioid peptide (86) in which increasing the proportion of the cis conformation of a prolyl peptide bond affects its affinity for the receptor. If we hypothesize the requirement of a cis amino-acyl proline peptide motif for a tight molecular contact within the neurotoxin-receptor complex, the small but significant change in activity of the P13A mutant ␦-EVIA could be interpreted as a deficiency in presenting on the same side of the main peptide chain both a hydrophobic cyclic proline side chain and a peptide bond carbonyl, a potent acceptor of a hydrogen bond. Further explorations and probing of this hypothesis are currently in progress using peptido-mimics simulating either cis or trans amino-acyl proline as well as using other single-point mutations in the regions expected to be important for the toxin-receptor interactions.