Three-dimensional Solution Structure of the Sodium Channel Agonist/Antagonist (cid:1) -Conotoxin TxVIA* □ S

The three-dimensional solution structure of (cid:1) -cono-toxin TxVIA, a 27-mer peptide agonist/antagonist of sodium channels, was determined by two-dimensional 1 H NMR spectroscopy with simulated annealing calculations. A total of 20 converged structures of (cid:1) -conotoxin TxVIA were obtained on the basis of 360 distance constraints obtained from nuclear Overhauser effect connectivities, 28 torsion angle constraints, and 27 constraints associated with hydrogen bonds and disulfide bonds. The atomic root mean square difference about the averaged coordinate positions is 0.35 (cid:2) 0.07 Å for the backbone atoms (N, C (cid:3) , C) and 0.98 (cid:2) 0.14 Å for all heavy atoms of the entire peptide. The molecular structure of (cid:1) -conotoxin TxVIA is composed of a short triple-stranded antiparallel (cid:4) -sheet. The overall (cid:4) -sheet topology is (cid:5) 2x, (cid:6) 1, which is the same as those for other conotoxins. However, the three-dimensional structure of (cid:1) -conotoxin TxVIA has an unusual hydrophobic patch on one side of the molecule, which may play an important role in the sodium channel binding. These results provide a molecular basis for understanding the mechanism stand- ard pulse sequences and phase cycling on either a Bruker AMX-500 or Avance-500 spectrometer operating at 500 MHz for the proton fre- quency. All two-dimensional NMR spectra were acquired in a phase-sensitive mode using the time-proportional phase incrementation (21) for quadrature detection in the t 1 dimension. NOESY spectra (22, 23) were recorded at temperatures of 15 and 25 °C with mixing times of 50, 100, 200, and 300 ms, respectively. TOCSY spectra were recorded using a MLEV-17 pulse scheme (24) with isotropic mixing times of 50 and 80 ms. The suppression of the solvent resonance was achieved by using the WATERGATE scheme in both the NOESY and TOCSY spectra (25). A DQF-COSY spectrum (26) was recorded to obtain the constraints for torsion angles. The suppression of the solvent resonance was achieved by using coherence selection with a gradient enhanced method (27, 28). No selective irradiation during the relaxation delay period was used to suppress the solvent resonance. A PE-COSY (29) spectrum was recorded for stereospecific assignment in 99.96% 2 H 2 O. The data process- ing was performed on either a Bruker X-32 UNIX work station with UXNMR software or a Silicon Graphics O 2 work station with XWIN- NMR software. For the slowly exchanging backbone amide protons, the sample lyophilized from H 2 O was redissolved in 2 H 2 O and was identi- fied by analyses of NOESY spectra recorded at time scales of 0.5, 3.0, 6.0, and 12 h. Chemical shifts were referenced to the methyl resonance of 4,4-dimethyl-4-silapentane-1-sulfonic acid, used as an internal standard. Distance Constraints and Structure Calculations— Interproton distance restraints were obtained from the NOESY spectra with mixing times of either 100 or 300 ms. NOE data were divided into three classes, strong, medium, and weak, corresponding to upper limits of 2.5, 3.5, and 5.0 Å in the interproton distance restraints. Pseudo-atoms were used for non-stereospecifically

The three-dimensional solution structure of ␦-conotoxin TxVIA, a 27-mer peptide agonist/antagonist of sodium channels, was determined by two-dimensional 1 H NMR spectroscopy with simulated annealing calculations. A total of 20 converged structures of ␦-conotoxin TxVIA were obtained on the basis of 360 distance constraints obtained from nuclear Overhauser effect connectivities, 28 torsion angle constraints, and 27 constraints associated with hydrogen bonds and disulfide bonds. The atomic root mean square difference about the averaged coordinate positions is 0.35 ؎ 0.07 Å for the backbone atoms (N, C ␣ , C) and 0.98 ؎ 0.14 Å for all heavy atoms of the entire peptide. The molecular structure of ␦-conotoxin TxVIA is composed of a short triplestranded antiparallel ␤-sheet. The overall ␤-sheet topology is ؉2x, ؊1, which is the same as those for other conotoxins. However, the three-dimensional structure of ␦-conotoxin TxVIA has an unusual hydrophobic patch on one side of the molecule, which may play an important role in the sodium channel binding. These results provide a molecular basis for understanding the mechanism of sodium channel modulation through the toxinchannel interaction and insight into the discrimination of different ion channels.
Voltage-dependent sodium channels are integral plasma membrane proteins responsible for the rapidly rising phase of action potentials in most excitable tissues and are specifically targeted by many neurotoxins. These toxins bind to different receptor sites on the sodium channel and have facilitated its functional mapping and revealed its subtype diversity (reviewed by Catterall (1) and Gray et al. (2)). There are at least six reported neurotoxin receptor sites on the sodium channel. Five neurotoxin receptor sites were defined by Catterall (3), and another new receptor site for the ␦-conotoxins has been proposed by Fainzilber et al. (4,5). Several kinds of conotoxins have been found to bind to this new receptor site: ␦-conotoxin TxVIA (␦-CTX TxVIA) 1 from Conus textile (6), ␦-CTX GmVIA from Conus gloriamaris (7), ␦-CTX PVIA from Conus purpurascens (8), and NgVIA (5) (Fig. 1). Among these conotoxins, ␦-CTX TxVIA was originally found as a mollusk-specific conotoxin that slows sodium channel inactivation exclusively in mollusk neuronal membranes (9 -11). However, this peptide also exhibits high affinity binding to rat brain (silent binding) neuronal membranes, despite its lack of any toxic effect in vertebrates in vivo and in vitro (4). Binding studies and electrophysiological tests with both vertebrate muscle and insect neuronal preparations have indicated that the silent binding sites of ␦-CTX TxVIA are highly conserved among a wide range of distinct vertebrate and insect sodium channels. The latter finding indicates the existence of a pharmacological distinction between the silent and effective binding sites of ␦-CTX TxVIA and highlights some structural differences that may be functionally important between the molluscan and rat brain sodium channels.
␦-CTX TxVIA consists of 27 amino acid residues with three disulfide bonds. The cysteine topology of ␦-CTX TxVIA is similar to those of -conotoxin MVIIA (-CTX MVIIA) obtained from Conus magus (12) and -conotoxin GVIA (-CTX GVIA) from Conus geographus (13), which are specific antagonists for N-type calcium channels (see Fig. 1). However, the amino acid residues of ␦-CTX TxVIA are quite different from those of the other two peptides. For example, ␦-CTX TxVIA has three negatively charged residues and a total charge of minus two, whereas -CTX MVIIA and -CTX GVIA have overall plus charges. Moreover, ␦-CTX TxVIA has many hydrophobic residues, unlike in -CTX MVIIA and -CTX GVIA. Such differences in the amino acid sequences may play an important role in discriminating the specific target channel from the many other kinds of ion channels.
In the present study, we determined for the first time the three-dimensional structure of one of the ␦-conotoxins, ␦-CTX TxVIA, in aqueous solutions by using two-dimensional NMR with simulated annealing. Based on analyses of the determined secondary and tertiary structure elements, the backbone conformation of ␦-CTX TxVIA is similar to those of -CTX GVIA (14 -17) and -CTX MVIIA (18,19) reported previously. However, novel charge distribution and hydrophobic characteristics are present in ␦-CTX TxVIA. A comparison of the ␦-CTX TxVIA and other toxin structures and the structure-function relationships of calcium channel blockers are also discussed.

EXPERIMENTAL PROCEDURES
Peptide Synthesis-␦-CTX TxVIA was chemically synthesized as described previously for TxVII (20), except for a modification of the folding solution. Briefly, a linear precursor of ␦-CTX TxVIA was assembled by Fmoc (N-(9-fluorenyl)methoxycarbonyl) solid phase synthesis and was dissolved to a final concentration of 0.05 mM in 0.4 M Tris-HCl buffer (pH 8.2) containing 20% methanol, 25% acetonitrile, EDTA (final 1 mM), and reduced and oxidized glutathione (each final 1.2 mM). The folding solution was stirred at 4°C for 3 days and then acidified and filtrated. The peptides were concentrated on an ODS column by a medium pressure pump, and ␦-CTX TxVIA was subsequently purified by chromatography on Sephadex G-50F and preparative reversed phase HPLC on an ODS column. The synthetic ␦-CTX TxVIA was identified by coinjection with native ␦-CTX TxVIA on reversed phase HPLC and by matrixassisted laser desorption ionization time-of-flight mass spectroscopy analysis.
The disulfide bond pairings of ␦-CTX TxVIA were determined by the same method as used for the determination of the disulfide bond pairings of -CTX TxVII (20). The authentic peptide, WCKQ/LDQNCCDG/ VCT, which was connected by two definite disulfide bonds, was synthesized for the comparison with the fragment cut out from ␦-CTX TxVIA by enzymatic digestions.
NMR Spectroscopy-The samples for NMR experiments were prepared at a concentration of ϳ5 mM in either 99.96% 2 H 2 O or 90% H 2 O, 10% 2 H 2 O at pH 7.0. NMR measurements were performed using standard pulse sequences and phase cycling on either a Bruker AMX-500 or Avance-500 spectrometer operating at 500 MHz for the proton frequency. All two-dimensional NMR spectra were acquired in a phasesensitive mode using the time-proportional phase incrementation (21) for quadrature detection in the t 1 dimension. NOESY spectra (22,23) were recorded at temperatures of 15 and 25°C with mixing times of 50, 100, 200, and 300 ms, respectively. TOCSY spectra were recorded using a MLEV-17 pulse scheme (24) with isotropic mixing times of 50 and 80 ms. The suppression of the solvent resonance was achieved by using the WATERGATE scheme in both the NOESY and TOCSY spectra (25). A DQF-COSY spectrum (26) was recorded to obtain the constraints for torsion angles. The suppression of the solvent resonance was achieved by using coherence selection with a gradient enhanced method (27,28). No selective irradiation during the relaxation delay period was used to suppress the solvent resonance. A PE-COSY (29) spectrum was recorded for stereospecific assignment in 99.96% 2 H 2 O. The data processing was performed on either a Bruker X-32 UNIX work station with UXNMR software or a Silicon Graphics O 2 work station with XWIN-NMR software. For the slowly exchanging backbone amide protons, the sample lyophilized from H 2 O was redissolved in 2 H 2 O and was identified by analyses of NOESY spectra recorded at time scales of 0.5, 3.0, 6.0, and 12 h. Chemical shifts were referenced to the methyl resonance of 4,4-dimethyl-4-silapentane-1-sulfonic acid, used as an internal standard.
Distance Constraints and Structure Calculations-Interproton distance restraints were obtained from the NOESY spectra with mixing times of either 100 or 300 ms. NOE data were divided into three classes, strong, medium, and weak, corresponding to upper limits of 2.5, 3.5, and 5.0 Å in the interproton distance restraints. Pseudo-atoms were used for non-stereospecifically assigned protons, and intra-residue and long range correcting factors were added to the distance restraints, respectively (30). In addition, 0.5 Å was added to the upper limits for distance restraints involving methyl protons (31). Nine additional constraints were added to define the three disulfide bonds involved in ␦-CTX TxVIA. For each disulfide bond, there are three distance constraints, S(i)-S(j), S(i)-C ␤ (j), and S(j)-C ␤ (i), whose target values were set to 2.02 Ϯ 0.02 Å, 2.99 Ϯ 0.5 Å, and 2.99 Ϯ 0.5 Å, respectively (32). After the initial calculation, 10 distance constraints for hydrogen bonds in the ␤-sheet, which were unambiguously defined, were added as target values of 1.8 -2.3 Å for NH(i)-O(j) and 2.8 -3.3 Å for N(i)-O(j), respectively.
All calculations were carried out with the X-PLOR 3.1 program (33). The three-dimensional structures were calculated on the basis of the experimentally derived distance and torsion angle constraints, using a dynamical simulated annealing protocol starting from a template structure with randomized backbone and torsion angles. For a structural comparison, the coordinates of -CTX TxVII were obtained from the Protein Data Bank Entry 1F3K (34).

RESULTS
Structure Determination-As an initial approach to the structural analysis, we determined the disulfide bonding pattern of ␦-CTX TxVIA by an enzymatic fragmentation method (20). We confirmed that ␦-CTX TxVIA has the same disulfide bonding pattern that is generally observed in other conotoxins, namely, Cys 2 -Cys 17 , Cys 9 -Cys 21 , and Cys 16 -Cys 26 . Sequencespecific resonance assignments were achieved according to the standard method established by Wü thrich and co-workers (35). The complete sequence-specific resonance assignments of ␦-CTX TxVIA are summarized in the Supplementary Material. To determine the three-dimensional structure, we carried out the structural calculation using the X-PLOR-simulated annealing protocol (36). The input data of the NMR experimental constraints consisted of 387 distance and 28 dihedral constraints. Simulated annealing calculations were started from 100 initial random structures. We selected 20 final structures with lower energies. Structural statistics for the mean and 20 converged structures were evaluated in terms of structural parameters, as shown in Table I. The deviations from idealized covalent geometry were very small, and the Lennard-Jones van der Waals energy was large and negative, indicating that no distortions and no non-bonded bad contacts existed in the converged structures. The r.m.s. differences from the averaged coordinate positions were 0.35 Ϯ 0.07 Å for the backbone atoms (N, C ␣ , and C) and 0.98 Ϯ 0.14 Å for all heavy atoms (Table II).
Description of the Three-dimensional Structure- Fig. 2, a and b, show stereopair representations of the best-fit superposition of the backbone atoms (N, C ␣ , C) for the 20 converged structures. An analysis of the 20 converged structures indicates that the molecular structure of ␦-CTX TxVIA contains a ␤-sheet region composed of three short ␤-strands, i.e. ␤-strand 1 (Glu 7 to Cys 9 ), ␤-strand 2 (Tyr 20 to Ile 22 ), and ␤-strand 3 (Val 25 to Thr 27 ). It can be classified as a triple-stranded antiparallel ␤-sheet of the topology ϩ2x, Ϫ1 (37), as shown in Fig.  2, a and b. The ␤-sheet is the best defined region through the entire molecule, as indicated by the small values of the corresponding backbone r.m.s.d. values (Table II), and the extent of each ␤-strand is limited by the presence of the characteristic turn structures.
We identified the turns by using the standard definition, that the distance between C ␣ (i) and C ␣ (i ϩ 3) is less than 7 Å (38), and by evaluating the characteristic distance connectivities of the backbone protons (35) for the corresponding turn segments in the 20 final converged structures and their mean structure. These analyses led to the identification of four ␤-turns, which were classified according to Wilmot and Thornton (39). The four ␤-turns involve residues Gln 4 to Glu 7 (type II), Asn 10 to Asp 13 (type I), Cys 16 to Gly 19 (type I), and Ile 22 to Val 25 (type IIЈ). These turns are stabilized by the hydrogen bonds between the carbonyl group (i) and the amide proton (i ϩ 3) of the peptide backbone. The hydrogen-deuterium exchange experiment led to the identification of the slowly exchanging amide protons originating from these four turns. The characteristic and angle ranges of the ␤-turns are conserved in all 20 converged structures and show high convergence in a Ramachandran plot (not shown).
As described above, the molecular architecture of ␦-CTX TxVIA is composed of a number of intramolecular hydrogen bonds and three disulfide bonds, resulting in a compact, well defined structure over the entire molecule. The disulfide bonds between Cys 2 and Cys 17 and between Cys 16 and Cys 26 constrain the peptide backbone into close spatial contact between the third ␤-turn (Cys 16 to Gly 19 ) and the N and C termini, respectively, and the disulfide bond between Cys 9 and Cys 21 interconnects ␤-strands 1 and 2. Consequently, the disulfide bonds between Cys 2 and Cys 17 and between Cys 9 and Cys 21 are located at the surface, and the disulfide bond between Cys 16 and Cys 26 is buried within the molecule.

Structural Comparison of ␦-CTX TxVIA with Other Ion
Channel Blockers-In the present study, we have determined the three-dimensional structure of ␦-CTX TxVIA in aqueous solution by using 1 H NMR spectroscopy and simulated annealing calculations. ␦-CTX TxVIA is composed of a short triplestranded antiparallel ␤-sheet and several turns. The overall ␤-sheet topology is ϩ2x, Ϫ1, which is the same as those reported for -CTX GVIA and -CTX MVIIA, N-type calcium channel blockers. As shown in Fig. 3, the global conformation of ␦-CTX TxVIA is similar to those of the structurally well characterized -CTX GVIA (14 -17) and -CTX MVIIA (18,19). Despite the relatively low amino acid sequence homology among the many conotoxins, the toxin structures exhibit similar locations and orientations of the secondary structure elements in their three-dimensional coordinates.
In terms of the numbers of amino acids between Cys residues, ␦-CTX TxVIA is the same as -CTX MVIIA. The numbers of amino acids between each Cys residue are identical, i.e. six, six, zero, three, and four for both peptides (see Fig. 1). Therefore, the backbone structures of both peptides can be estimated to be similar to each other. In fact, the backbone r.m.s.d. between residues 2 and 26 of ␦-CTX TxVIA and residues 1 and 25 of -CTX MVIIA is 0.9 Å, indicating that the backbone architecture is almost the same between these two toxins. However, the types of side chains of both peptides are completely different between these two peptides, except for the six Cys residues, Lys 3 , and Gly 18 (in ␦-CTX TxVIA). These facts indicate that the cysteine knot motif, which is often seen among the conotoxins, is important for the backbone architecture formation, but not for the selectivity of channel types. This is supported by former studies. (i) It was reported that changes in the number and the pairings of disulfide bonds resulted in partial disruption of the biologically active structure, leading to a reduction of the channel-blocking activity in the case of -CTX GVIA (40,41). (ii) Pallaghy et al. (15) found that the structures of disulfide isomers of -CTX GVIA displayed some conformational heterogeneity in solution, with a significant loss of the structural features of the native molecule. (iii) Linear analogs of -CTX MVIIA and -CTX GVIA, with full deletions of the three disulfide bonds, showed a complete loss in their binding affinity for N-type calcium channels.
Comparison of ␦-CTX TxVIA with -CTX TxVII-We have recently determined the solution structure of -CTX TxVII, another toxin from Conus textile (34). This peptide is known to bind to mollusk L-type calcium channels. However, -CTX TxVII is similar to ␦-CTX TxVIA, in that both peptides have many hydrophobic residues. Therefore, we compared the structure of ␦-CTX TxVIA with that of -CTX TxVII. Fig. 4, a-d, show the positive, negative, and hydrophobic residue represen-   tations of ␦-CTX TxVIA and -CTX TxVII. In these figures, both peptides have a hydrophobic patch on the bottom side of the molecules. However, the distributions of other hydrophobic residues are rather different between these two peptides. The residues corresponding to Ala 4 and Met 25 of -CTX TxVII are the hydrophilic (white color) Ser 5 and Thr 27 of ␦-CTX TxVIA, respectively. Furthermore, -CTX TxVII has an extra hydrophobic residue, Trp 26 . These three hydrophobic residues of -CTX TxVII form another hydrophobic patch on the left side of the molecule in Fig. 4c. However, there are hydrophilic residues on the left side of the ␦-CTX TxVIA molecule in Fig. 4a. Furthermore, the corresponding residues Trp 1 of ␦-CTX TxVIA and Leu 13 of -CTX TxVII are not hydrophobic residues, respectively. As for charged residues, there is a prominent difference between these two peptides. -CTX TxVII has no counterpart of the negatively charged Asp 18 in ␦-CTX TxVIA (Fig. 4, a   and c). These residues may cause the different specificity for ion channels between ␦-CTX TxVIA and -CTX TxVII.
Structure-Function Relationships of ␦-CTX TxVIA-␦-CTX TxVIA has many hydrophobic residues, i.e. Trp 1 , Met 8 , Leu 11 , Leu 12 , Tyr 20 , Ile 22 , Val 23 , Leu 24 , and Val 25 (Fig. 1). This is characteristic of the ␦-conotoxins, ␦-CTX TxVIA, ␦-CTX GmVIA ␦-CTX PVIA, ␦-CTX GmVIA, and NgVIA, which affect sodium current inactivation. Fig. 4b shows the distribution of the hydrophobic residues on the three-dimensional structure. Although the hydrophobic residues are dispersed on the primary structure, they are clustered and form a hydrophobic patch (Fig. 4b) on the three-dimensional structure. Some reasons why the hydrophobic residues are clustered on the three-dimensional structure may be as follows. (i) ␦-CTX TxVIA may dimerize via the hydrophobic patch. (ii) ␦-CTX TxVIA may bind to sodium channels via the hydrophobic patch. However, we did not observe the dimerization of ␦-CTX TxVIA, as confirmed by gel filtration at a neutral pH and at an even higher concentration of the peptide (data not shown). Therefore, the hydrophobic patch of ␦-CTX TxVIA may play an important role in binding to the sodium channels.
The sodium channel from rat brain is a heterotrimeric protein with an ␣␤ 1 ␤ 2 composition, but it is known that the ␣-subunit is sufficient to provide voltage-sensitive sodium gating in vitro (3,42). The ␣-subunit contains four homologous transmembrane domains connected by cytoplasmic linkers. The intracellular linker between domains III and IV has been identified as the inactivation domain, as revealed by the following facts. (i) Antibodies directed against the III-IV linker completely block fast inactivation (43,44). (ii) Inactivation does not occur in channels in which this linker is clipped and domain IV is expressed independently of domains I-III (45). (iii) Deletions or mutations in the III-IV linker show slow or no inactivation (46,47). A hydrophobic triad of amino acids, IFM (residues 1488 -1490), has been identified as a critical motif for inactivation (48 -50). Thiol modification experiments indicated that the III-IV linker undergoes a conformational change associated with inactivation that renders the phenylalanine in the IFM motif inaccessible, leading to the proposal that the III-IV linker functions as a hinged lid with the IFM motif serving as a hydrophobic latch that occludes the ion pore (51). Recently, the solution structure of the inactivation gate of the rat sodium channel was reported (52). The structured region of the gate peptide includes the hydrophobic IFM motif and extends C-terminally to Ser 1506 , which is required for modulation of inactivation by protein kinase C (53). The backbone of the inactivation gate forms an ␣-helix capped at its N terminus by a turn structure, with the hydrophobic IFM motif immediately N-terminal to this turn. These residues are important for the inactivation of the sodium channel, as revealed by the fact that mutants in which these residues are modified by a bulky reagent did not show any inactivation (52).
Taken together, the mechanism of the slow inactivation of the sodium channels by ␦-CTX TxVIA may be considered as follows (1). ␦-CTX TxVIA binds directly to the hydrophobic residues, such as the IFM motif, of the inactivation gate through its hydrophobic patch (2). ␦-CTX TxVIA binds to the inactivation gate binding interface of the sodium channel (3). ␦-CTX TxVIA binds to another hydrophobic region of the sodium channel and allosterically inhibits the inactivation. In each of these scenarios, the hydrophobic patch of ␦-CTX TxVIA, found in this study, may play an important role in the binding with the sodium channels.
There are several kinds of conotoxins that affect sodium current inactivation. These conotoxins may be classified in two categories: 1) conotoxins that affect the sodium channel inac-  (c and d). Blue and red regions represent positive and negative residues, respectively. Green regions represent hydrophobic residues. a is in the same orientation as the stereopairs in Fig. 2a. b was obtained by a 180°rotation about the vertical axis of a. c is in the orientation fit to a. d was obtained by a 180°rotation about the vertical axis of c. These figures were prepared using the program GRASP (55). tivation in rat brain. (2). Conotoxins that affect sodium channel inactivation in mollusks, but act as antagonists in the rat brain. ␦-CTX PVIA and NgVIA are classified in the first category, and ␦-CTX TxVIA and ␦-CTX GmVIA are classified in the second category. What are the differences between these two categories? The amino acid sequences of ␦-CTX PVIA and NgVIA are similar. Especially, both peptides have the same sequence between residues Hyp 6 and Ser 19 . In these sequences a hydrophobic residue, Leu 12 , in ␦-CTX TxVIA is replaced by a positively charged residue, Lys, in both ␦-CTX PVIA and NgVIA. Moreover, the conserved residues, Gln 14 and Asn 15 , in ␦-CTX TxVIA and ␦-CTX GmVIA are replaced by Gly and Leu in ␦-CTX PVIA and NgVIA, respectively. All of these residues are in the proximity of the hydrophobic patch, the putative sodium channel binding site. Therefore, these residues may play an important role in discriminating the mollusk sodium channel from the rat brain sodium channel. To discuss this in more detail, the three-dimensional structure of a toxin in the first category, ␦-CTX PVIA or NgVIA, should be determined.
The solution structure of ␦-CTX TxVIA reported here should contribute to our understanding of the structure and action of sodium channel agonists and antagonists and will be useful in designing synthetic organic compounds to control sodium current inactivation.