Structural Basis for the Inhibition of Voltage-gated Sodium Channels by Conotoxin μO§-GVIIJ*

Cone snail toxins are well known blockers of voltage-gated sodium channels, a property that is of broad interest in biology and therapeutically in treating neuropathic pain and neurological disorders. Although most conotoxin channel blockers function by direct binding to a channel and disrupting its normal ion movement, conotoxin μO§-GVIIJ channel blocking is unique, using both favorable binding interactions with the channel and a direct tether via an intermolecular disulfide bond. Disulfide exchange is possible because conotoxin μO§-GVIIJ contains an S-cysteinylated Cys-24 residue that is capable of exchanging with a free cysteine thiol on the channel surface. Here, we present the solution structure of an analog of μO§-GVIIJ (GVIIJ[C24S]) and the results of structure-activity studies with synthetic μO§-GVIIJ variants. GVIIJ[C24S] adopts an inhibitor cystine knot structure, with two antiparallel β-strands stabilized by three disulfide bridges. The loop region linking the β-strands (loop 4) presents residue 24 in a configuration where it could bind to the proposed free cysteine of the channel (Cys-910, rat NaV1.2 numbering; at site 8). The structure-activity study shows that three residues (Lys-12, Arg-14, and Tyr-16) located in loop 2 and spatially close to residue 24 were also important for functional activity. We propose that the interaction of μO§-GVIIJ with the channel depends on not only disulfide tethering via Cys-24 to a free cysteine at site 8 on the channel but also the participation of key residues of μO§-GVIIJ on a distinct surface of the peptide.

Marine snails of the genus Conus employ a complex venom mixture to subdue prey and as an effective means of defending against predation. The active venoms contain an array of peptides that bind to and modulate the properties of ion channels, G-protein-coupled receptors, and neurotransmitter receptors (1). Several peptides modulate the activities of voltage-gated sodium channels (VGSCs), 4 which are implicated in numerous neurological disorders, as well as neuropathic pain. Four classes of conopeptides have been shown to affect VGSC activity. Peptides belonging to theand ␦-families promote activation and inhibit inactivation, respectively, whereas theand O-conotoxins inhibit VGSCs by either blocking the Na ϩ conductance pore or preventing channel activation, respectively (2,3). Recently, the founding member of a fifth class of VGSC inhibitors was identified that blocks the channel through interaction with a previously unidentified neurotoxin binding site, site 8 ( Fig. 1A) (4).
O §-GVIIJ is a 35-residue peptide isolated from the venom of the piscivorous snail Conus geographus (Fig. 1B). In vitro folding of linear O §-GVIIJ with thiol-reactive oxidants (i.e. glutathione, L-cystine, or cystamine) resulted in adducts where Cys-24 was disulfide-bonded with glutathione, cysteine, or cysteamine, respectively (abbreviated as GVIIJ SSG , GVIIJ SSC , and GVIIJ SSEA , respectively; see Fig. 1C for structures). Of these, the GVIIJ SSC variant most closely resembled the native peptide, which also has a Cys disulfide-bonded to its Cys-24 residue. For largely historical reasons, the glutathione adduct (GVIIJ SSG ) was tested by two-electrode voltage clamp electrophysiology against rat Na V 1.1-1.8 expressed in Xenopus laevis oocytes, and it was found to block all tested Na V 1 isoforms with a K d or IC 50 Ͻ 0.4 M except Na V 1.5, which was blocked with an IC 50 of 207 M, and Na V 1.8, which was not blocked at all (IC 50 Ͼ 1 mM). When screened against the neuronal VGSC subtype Na V 1.2, the three analogs possessed nearly identical off rates (k off values) but different on rates (k on values) (4). An explanation consistent with these results is that the glutathione, cysteine, or cysteamine moiety disulfide-bonded to Cys-24 of the peptide acts as a leaving group when, by disulfide exchange, Cys-24 forms a disulfide bridge with a free cysteine on the ␣-subunit (specifically Cys-910, in the case of rNa V 1.2 (4)). This explanation was supported by the results of experiments that tested four additional adducts with different groups disulfidebonded to Cys-24 against both rat Na V 1.2 and mouse Na V 1.6, where it was observed that for a given Na V 1 isoform, the adducts had widely varying k on values but the same k off value (5).
Here we have determined the solution structure of O §-GVIIJ using an analog of this peptide, GVIIJ[C24S], which would not undergo dimerization (5) during NMR studies. The solution structure of GVIIJ[C24S] closely resembled that of the inhibitor cystine knot (ICK) class of peptides in that it possessed two short ␤-strands and was cross-linked by three disulfide bridges in the "core" of the molecule (6 -8). The ICK motif is found in peptides from numerous phyla and is of particular interest for pharmaceutical development because of its inherent stability and amenability to chemical modification for enhanced pharmacological effect (8). Importantly, this structure identified the location and position of residue 24 in a loop linking the two ␤-strands, where it would be available to interact with the channel.
To complement these structural studies, we performed detailed structure-activity relationship (SAR) studies to identify amino acid residues critical for inhibition of rat Na V 1.2 (rNa V 1.2), a VGSC isoform found in the central nervous system. SAR studies were performed on rNa V 1.2 exogenously expressed in X. laevis oocytes using mutants of a potent analog of O §-GVIIJ, GVIIJ SSEA , in which Cys-24 was modified by cysteamine (Fig. 1C). These studies identified three functionally important residues, Lys-12, Arg-14 and Tyr-16, located on a face of the peptide adjacent to the loop with residue 24. Thus, these results suggest a "functionally bipartite" mechanism of interaction by O §-GVIIJ, where a disulfide bridge "tethers" the peptide to a channel cysteine at site 8, whereas residues on a different surface of the peptide are also important for interaction with the channel.
To demonstrate that site 8 was physically distinct from site 1, where tetrodotoxin, saxitoxin, and -conotoxins bind and plug FIGURE 1. Binding site of O §-GVIIJ on the channel and its peptide sequence. A, binding sites of VGSC-inhibiting conotoxins. Pore-blocking -conotoxins bind deep within the Na ϩ conductance pore and compete with tetrodotoxin for neurotoxin binding site 1. Site 1 is located near the bottom of the reentrant loops in all four domains (domains I-IV), but only one site of interaction is shown in this figure. In contrast, O-conotoxins MrVIA/B bind to the voltage sensor of domain II at neurotoxin binding site 4 (3), whereas O §-GVIIJ, whose mechanism of block remains to be determined, binds to the loop in the pore module of domain II at site 8 (4). B, photograph of the shell of the fish-hunting cone snail C. geographus. C, primary sequence of O §-GVIIJ, displaying the previously determined disulfide framework (4). Intercysteine loops 1-4 are underlined. X represents modifications at position 24 used in the structural and pharmacological studies described in this report; their structures are shown at the bottom.
the Na ϩ -conducting pore (Fig. 1A), we previously employed a "leaky" -conotoxin, -KIIIA[K7A] (9,10), and showed that pre-equilibrating rNa V 1.2 with this peptide did not interfere with the block by O §-GVIIJ SSG (4). However, as noted (4), -KIIIA[K7A] has 16 amino acid residues, so these results did not exclude the possibility that a larger -conotoxin, such as -GIIIA (with 22 residues), might intrude into the binding space of O §-GVIIJ. We examined this possibility in the present study with a point mutant of -PIIIA (which, like -GIIIA, has 22 residues and blocks rNa V 1.4 with high affinity), namely -PIIIA[R14Q]. We chose this mutant because others have shown that it binds to rNa V 1.4 reasonably well with a significant residual current (rI Na or "leak") (11). We show here that pre-equilibrating rNa V 1.4 with nearly saturating levels of -PIIIA[R14Q] did not interfere with the block by O §-GVII-J SSG . In contrast, the rate of block by a dimer of O §-GVIIJ, (O §-GVIIJ) 2 , whose monomers were disulfide-bonded to each other via the thiols of their Cys-24 residues (5), was decreased by the presence of -PIIIA[R14Q]. These results lead us to conclude that sites 1 and 8 are distinct although not very far apart.
Peptide Cleavage, Purification, and Folding-Peptides were cleaved from the resin by treatment with reagent K (trifluoroacetic acid (TFA), H 2 O, phenol, thioanisole, 1,2-ethanedithiol; 82.5/5/5/5/2.5 by volume) for 3 h at room temperature. The crude peptide was separated from resin by vacuum filtration. The cleavage product was precipitated in cold methyl-tert-butyl ether, centrifuged, and washed again with the ether. Crude peptide was purified by reversed phase HPLC using a Vydac C18 semipreparative column (218TP510, 250 ϫ 10 mm, 5-m particle size) eluted with a linear gradient ranging from 15 to 45% solvent B (90% acetonitrile in 0.1% TFA) in 30 min. Oxidative folding of the linear peptide was performed in a buffered solution containing 20 M peptide, 0.1 M Tris-HCl, pH 7.5, and 1 mM EDTA. Oxidizing reagents used included a 1:1 mM mixture of reduced:oxidized glutathione (GVIIJ[C24S]), 1 or 2 mM cystamine dihydrochloride (GVIIJ SSEA ), or 2 ml of an L-cystinecontaining solution (6 mg/ml) in 5% (v/v) acetonitrile and 0.1% TFA (GVIIJ SSC ), as described previously (4,5). Oxidative folding was carried out overnight at room temperature. The oxida-  (2-aminoethanethiol) or cysteine, respectively. Retention time (RT) was assessed by RP-HPLC with respect to the start of the HPLC run using an analytical C18 column and an elution gradient ranging from 15 to 45% of solvent B in 30 min at different stages of the project. All "poorly" folding analogs of GVIIJ were characterized either by HPLC peak broadening or by tailing of the peak. tion reaction was quenched by acidification with formic acid (final concentration, 8% (v/v)). Folded analogs were purified by semipreparative reversed phase HPLC. Purities for analogs used in NMR studies were greater than 95%, whereas those used in electrophysiological assays ranged from 80 to 99%. The identities of the oxidized peptides were confirmed by MALDI-TOF mass spectrometry, except GVIIJ[D23␥] SSEA , for which electrospray mass spectrometry was employed.
Other Toxins-Syntheses of -GIIIA, and -PIIIA[R14Q] were performed essentially as previously described (12,13). The synthesis of (O §-GVIIJ) 2 (a dimer in which Cys-24 of one monomer was disulfide-bonded to its counterpart on the other monomer) was described recently (5). and then recording a time course of one-dimensional spectra, followed by acquisition of a 70-ms TOCSY spectrum at 298 K. The spectra were processed using TOPSPIN (version 3.2) and were analyzed in CcpNmr-Analysis (version 2.1.5). The spectra were referenced either directly or indirectly to the DSS methyl signal at 0.0 ppm. Chemical shifts were deposited in the BioMagResBank Data Bank (14,15) with accession numbers 26674 and 26675 for GVIIJ[C24S] and GVIIJ SSEA , respectively.

Structural Constraints and Calculations
Distance constraints for structural calculations were generated from assigned cross-peaks in NOESY spectra (150-ms mixing time) acquired at 298 K, pH 3.2. Dihedral angle constraints ( and ) were generated from TALOS-N predictions (16) and from 3 J HN-H␣ coupling constants according to the following criteria: 3 J HN-H␣ Ͼ 8 Hz, ϭ Ϫ120 Ϯ 40°; 3 J HN-H␣ Ͻ 6 Hz, ϭ Ϫ60 Ϯ 40°. Disulfide bond constraints were added according to the previous experimentally determined connectivity illustrated in Fig. 1C (4). Deuterium exchange rates and temperature coefficients were used to identify H-bonded backbone amides, where H-bond acceptors could be identified from initial rounds of structural calculation (present and consistent in Ն80% of structures) and were included in subsequent structural calculations. Initial structure calculations were optimized for a low target function using the noeassign macro in CYANA 3.0 (17). Structures generated by CYANA were then used to resolve the assignment of any remaining ambiguous inter-residue NOEs. Following optimization, the final set of constraints was entered into CNS version 1.3 (18), and an ensemble of 100 structures was generated. Of these, the 20 lowest energy structures without violations were selected to represent the solution structure of GVIIJ[C24S]. Validation of the final calculated structures was accomplished using PROCHECK-NMR (19). Secondary structure prediction was performed using DSSP (20) and PROMOTIF (21) based on the closest to average structure of GVIIJ[C24S]. Classification of ␤-turns was based on the criteria reported in Ref. 22. Structures were analyzed using MOL-MOL (23), and structural representations were constructed using UCSF Chimera.

Structure-Activity Studies on GVIIJ SSEA and GVIIJ SSC with rNa V 1.2 and Competition Experiments ofversus O §-Conotoxins with rNa V 1.4
Two-electrode Voltage Clamp Electrophysiology-All O §-GVIIJ analogs were screened against rat Na V 1.2 exogenously expressed in X. laevis oocytes. Oocytes were also used to express rat Na V 1.4 in experiments to assess competition between -PIIIA[R14Q] and O §-GVIIJ SSG or O §-GVIIJ dimer and, in positive control tests, between -PIIIA[R14Q] and -GIIIA. The Na V 1.2 (NM_012647) and Na V 1.4 (NM_013178) clones were obtained from Alan Golden (University of California, Irvine). The oocytes were injected with 30 -50 nl of Na V 1.2 or Na V 1.4 cRNA (1.5 or 0.6 ng, respectively) in RNase-free water and incubated overnight at 16°C in ND96 (96 nM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , and 5 mM HEPES, pH 7.5), supplemented with antibiotics (100 units/ml penicillin, 0.1 mg/ml streptomycin). Voltage clamping was performed with a Warner OC-725C amplifier (Warner Instruments) using 3 M KCl-filled microelectrodes (resistance Ͻ 0.5 M⍀) (24). An oocyte was placed in a 4-mm diameter well (30 l volume) filled with ND96 and clamped at a holding potential of Ϫ80 mV. Voltage-gated sodium currents (I Na ) were induced at 20-s intervals with a 50-ms depolarizing pulse to Ϫ10 mV. The peptides to be tested were dissolved in ND96 at 10 times the desired final concentration (0.1-333 M) and applied in 3-l volumes to the static bath containing the voltage clamped oocyte. A static bath was used to conserve the limited quantities of analogs. Immediately following peptide application, the solution in the well was gently mixed through repeated pipetting of the bath solution. All experiments were performed at room temperature. Washout of peptide was performed by perfusing the chamber initially at a rate of 1 ml/min for 1 min followed by 0.2 ml/min for 19 min. Use of X. laevis frogs, which provided oocytes for this study, followed protocols approved by the University of Utah Institutional Animal Care and Use Committee, which conform to the National Institutes of Health Guide for the Care and Use of Laboratory Animals.
Analysis of Electrophysiological Data-The time course of peak I Na was plotted before and during toxin exposure, as well as following toxin washout. The observed rate constants (k obs values) were determined by fitting of the onset of block to a single-exponential function for Ն3 oocytes at each of three different peptide concentrations. Because all O §-GVIIJ analogs tested were very slowly reversible (i.e. Ͻ 50% recovery after 20 min washout or k off Ͻ 0.035 min Ϫ1 ), off rates (k off values) were estimated from the level of recovery after 20 min washout for Ն 9 oocytes, assuming a single-exponential time course (25). To minimize effects of baseline drift, times longer than 20 min were not used. k obs values were plotted against [peptide], and k on values were obtained from the slope of the linear regression line, assuming the equation, k obs ϭ k on [peptide] ϩ k off (26). Electrophysiology data are presented as the means Ϯ S.D., unless specified otherwise, and statistical significance was determined using the two-tailed unpaired t test in GraphPad Prism or Excel. Percentage block of the peak sodium current (I Na ) by conotoxins or their analogs was calculated as defined in Ref. 25 by the following equation: % block (peak I Na ) ϭ [(average Ն 3 traces at steady state)/(average Ն 3 baseline traces)] ϫ 100.

Peptide Synthesis and Analysis-GVIIJ[C24S]
and GVIIJ SSEA , which were used for NMR studies, were synthesized as described under "Experimental Procedures." All peptide analogs used in SAR studies were synthesized following folding and purification protocols described for GVIIJ SSEA . These syntheses generated, on average, between 500 and 900 nmol of purified linear peptide per 100 mg of resin cleaved. Oxidative folding of most of the linear analogs produced a single major product comprising between 40 and 60% of the total folding mixture. The exceptions were the [O6A], [D23N], and [D23␥] analogs, for which the major product comprised only 9, 17, and 28% of the mixture, respectively. Folding of the [F21A], [T26A], and [T28A] analogs in the presence of cystamine dihydrochloride did not lead to a single major product as monitored by reversed phase HPLC. However, oxidation of these analogs in the presence of L-cystine showed modest improvement in producing a major folding product. The [S19A] analog exhibited poor folding properties independent of the oxidant used, with the major peak being just 8% of the total mixture. Broadening and "tailing" of the purified HPLC peaks characterized all "poorly" folding analogs, resulting in final purities of 80 -93%.
NMR Spectroscopy-Good quality 1 H NMR spectra were obtained for GVIIJ [C24S] and GVIIJ SSEA at pH 3.2. Comparison of NOESY spectra of GVIIJ[C24S] at pH 3.2 with spectra at pH 7.4 showed that several key long-range NOEs involving pairs of nonexchangeable protons (e.g. Lys-30 H ␤ -Asp-23 H ␤ ; Cys-29 H ␤ -Hyp-6 H ␣ ; Ala-8 H ␤ -Cys-29 H ␣ ; Thr-9 H ␣ -Thr-28 H ␥2 ; Gly-7 H ␣ -Thr-28 H ␣ ) were present at both pH values, implying that the structure determined at pH 3.2 was representative of that at physiological pH. Sequential assignments were derived from two-dimensional TOCSY and NOESY spectra. The temperature dependence of amide proton chemical shifts is shown in Fig. 2A. Deviations from random coil values (27, 28) for back-   Fig. 3. The near identity of these plots confirms adoption of similar backbone structures for these two analogs. Both structures have chemical shifts close to random coil values at the N and C termini. Small differences between the two peptides were observed for the residues of loop 2 (Gly-11-Tyr-16) and loop 4 (Asp-23-Thr-28). Solution Structures-Structures were generated in CYANA and then refined in CNS (18) using a total of 401 NOE-derived distance constraints, 39 dihedral, 9 1 angle constraints (from 3 J HN-H␣ J-coupling measurements and TALOS-N predictions), and 5 hydrogen bond restraints (from amide temperature coefficients and 2 H exchange experiments). A summary of experimental constraints and structural statistics for GVIIJ[C24S] is provided in Table 2.
Within loop 1 (Gly-4 -Thr-9), adjacent to the ␤2-strand, is a small segment of the peptide that exhibits modest levels of extended structure (Fig. 4B)

Inhibition of Voltage-gated Sodium Channels by O §-GVIIJ
MARCH 25, 2016 • VOLUME 291 • NUMBER 13 JOURNAL OF BIOLOGICAL CHEMISTRY 7211 region was too short to be designated as such by DSSP or PROMOTIF secondary structure prediction programs.
Other secondary structure features include a type IV ␤-turn encompassing Cys-10 -Leu-13 within intercysteine loop 2 (Gly-11-Tyr-16) (Fig. 4C). This loop was less well defined than the remainder of the peptide, with angular order parameters for and Ͻ 0.8 and RMSD of 1.2 Å over the backbone heavy atoms N, CA, and C of residues Gly-11-Tyr-16 (Fig. 5, D-F). Several rapidly exchanging backbone amide protons were identified in this region by hydrogen-deuterium exchange experiments (Figs. 2, B and C, and 5A), indicating that this loop was solvent-exposed. Backbone angular order parameters for residues at the N and C termini were also Յ0.8 (Fig. 5, D-F).
Oxidative Folding of O §-GVIIJ Analogs-Residue 24, which in the native peptide presents the "extra" cysteine responsible for tethering to the channel, is located within the ␤-turn of loop 4 (Fig. 6A). Asp-23, at the start of the turn, was critical for folding of the GVIIJ SSEA analogs, with substitution by either lysine or alanine resulting in inefficient formation of a major folding product (Fig. 6B). A major product was observed in the [D23N] and [D23␥] analogs, albeit in lower overall yields compared with efficiently folding analogs. This restoration in folding was presumably due to interactions between O ␥ in the side chain of Asp-23 and the backbone H N of Tyr-25 (i ϩ 2) and H N of Thr-26 (i ϩ 3) (22). Hydrogen bonds between the side chain of Asp-23 and the H N of Tyr-25 were observed in half the structures and between Asp-23 and Thr-26 in ϳ25% of structures. Replacement of Tyr-25 by Asp made the peptide more prone to dimerization, decreasing the yield for the desired product by ϳ10%. In many type I ␤-turns, Asp is one of three preferred residues for electrostatic interactions with the backbone H N of residue 25 (22); it may be that this substitution further exposed Cys-24, where it could conceivably undergo dimerization. Restoration of the folding yield was observed for the Arg-25 (i ϩ 2) mutation, which may have resulted from stabilization of loop 4 through a weak interaction between N ⑀ of Arg-25 and the main chain oxygen of Asp-23.
Several residues in close proximity to this ␤-turn also appeared to be important for stabilization of loop 4 (Asp-23-Thr-28). In all 20 structures, the side chain of Lys-30 points back into the loop where it could participate in electrostatic interactions with either Asp-23 or Thr-26. In the closest to average structure of GVIIJ[C24S], the H of Lys-30 and the side chain oxygen atoms of Asp-23 or Thr-26 were separated by just 3.4 and 4.4 Å, respectively (Fig. 6A). The importance of such interactions was illustrated by the inability of the [K30D] analog to produce a single, major folding product. Replacement of Lys-30 with an acidic residue such as aspartate presumably had a destabilizing effect on loop 4 (Asp-23-Thr-28), but substitution with a small, uncharged residue in the [K30A] analog restored efficient folding. Other residues found to be important for the folding of O §-GVIIJ included Thr-26 and Thr-28. Neither [T26A] nor [T28A] yielded a single major product following cystamine oxidation. Again, based on the calculated distances between the side chains in the closest to average structure, this was probably caused by removal of stabilizing interactions between the side chains of these residues and H of Lys-30. The use of a different oxidant (i.e. L-cystine) resulted in formation of a major product in each case, albeit in significantly lower yields compared with other analogs used in this study (Fig. 6B). is stabilized by numerous H-bonds between backbone atoms, as well as by side chain to backbone interactions between the residues in this loop (Phe-21-Thr-28). A type I ␤-turn between two antiparallel ␤-strands is stabilized by interactions between residues in this turn (Asp-23-Thr-26). B, HPLC elution profiles of GVIIJ SSEA analogs following folding. Conventional folding of Ala-replacement mutants of Phe-21, Asp-23, Thr-26, and Thr-28 did not yield a single, major folding species (black trace); however, folding in the presence of L-cystine (red trace) resulted in a major product (red asterisk), although with significantly lower yields compared with other mutants (black asterisks). The [D5K] analog was included to show that charge reversal did not affect accumulation of a major product when the substitution was made to residues spatially distinct from loop 4. C, comparison of Na V 1.2 blockade by GVIIJ SSEA mutants in the Cys-24 containing loop. The data are from Table 3. Analogs constructed on the GVIIJ SSC background are denoted by a red asterisk. Nearly all analogs were functionally equipotent with the unmodified peptide, suggesting that residues in this region are important for stability, but not activity. One exception to this was the [T28A] analog, which exhibited a K d value of 8.60 Ϯ 2.30 nM, significantly higher (p value ϭ 0.0212) than that of unmodified peptide. Selection of some amino acid replacements in this loop was based on modest sequence homology between loop 4 of O §-GVIIJ and rat or human Na V ␤2 and ␤4-subunits, whose partial homologous sequences are shown in the inset.

Structure-Activity Relationship Studies of the Block of Na V 1.2 by O §-GVIIJ
Analogs-The ability of peptides to block Na V 1.2 expressed in Xenopus oocytes was assessed electrophysiologically as described under "Experimental Procedures." We used rNa V 1.2 as the target channel to maintain consistency with the previous studies of O §-GVIIJ analogs (4,5). The GVIIJ[C24S] analog, which was used in NMR studies, blocked Na V 1.2 rapidly and very reversibly (Fig. 7A). The rapidity precluded accurate determination of k on and k off values. The IC 50 for GVIIJ[C24S] was 4.8 M (with 95% confidence interval (CI) of 3.7-6.2 M) (Fig. 7B).
All SAR studies were performed with analogs of GVIIJ SSEA , which itself is slightly more potent than GVIIJ SSC , against Na V 1.2 (K d ϭ 2.3 versus 3.4 nM) (5). An initial series of analogs was designed and synthesized to investigate the importance of the residues of loop 4 (Asp-23-Thr-28) in the immediate vicinity of Cys-24. Amino acid substitutions in loop 4 appeared to have significant effects on the structural stability of the peptide, as evidenced by the inability of some of the variants to form a single, major folding product (Fig. 6B). However, most of these substitutions had little effect on the ability of the peptide to block rNa V 1.2 (Fig. 6C). The on rate constants varied considerably across all analogs, whereas recovery from block (reflected in k off ) was very slow and showed little variation (Table 3 and  The solid line is the best fit curve to the equation % peak I Na blocked ϭ Y max /(1 ϩ (IC 50 /[peptide])), where Y max , the extrapolated level of block at saturating peptide concentrations, was 82.8% (with 95% CI of 77.5-88.10), and the IC 50 was 4.8 M (with 95% CI of 3.7-6.2 M). C, GVIIJ SSEA (1 M) rapidly blocked ϳ70% of the sodium current. In contrast to the [C24S] analog, this analog exhibited nearly irreversible block of Na V 1.2 following washout, which is attributed to tethering to site 8 of the channel (4,5). D-F, substitution of critical amino acids Lys-12, Arg-14, and Tyr-16 resulted in analogs with lower k obs for the inhibition of Na V 1.2 (see also Table 3). The red bar represents when the oocyte was exposed to peptide. The inset in each panel shows the current traces before (black) and during (red) exposure to the indicated peptides. . The lower k on values observed with these analogs, in turn, resulted in increased K d values ranging from ϳ50to 280-fold higher than GVIIJ SSEA (Fig. 7, C-F, and Table 3).  (Fig. 8A). At saturating concentrations, -PIIIA[R14Q] blocked incompletely and left a residual current (rI Na ) of 27 Ϯ 3.5% (Fig. 8A). As a positive control for competition between -PIIIA[R14Q] and another site 1 antagonist, we used -GIIIA, the first -conotoxin used to help define site 1 (12,30). -GIIIA (10 M) rapidly blocked the I Na of rNa V 1.4 (Fig. 8B, inset). However, when rNa V 1.4 was pre-equilibrated with 33 M -PIIIA[R14Q], the block of the resulting rI Na by -GIIIA was significantly slower (Fig. 8B and Table 4), as would be expected if the two -conotoxins competed for the same site on the channel. In contrast, the rate of block by 3.  Table 4).

Discussion
The structural and functional studies of O §-GVIIJ described here have revealed several important features of the peptide that contribute to its functional activity. The solution structure of GVIIJ[C24S] is a classical ICK motif (6, 7), exhibiting two antiparallel ␤-strands cross-linked by three disulfide bridges that formed a knot-like structure in which the Cys-17-Cys-29 disulfide crossed the macrocycle formed by the remaining two disulfide bridges and the interconnecting backbone. ICK peptides are abundant in nature, with many being observed in toxin or toxin-like peptides (Fig. 9), and are of significant interest as therapeutic scaffolds because of their relative ease of synthesis, stability and amenability to sequence mutations (8). Important in the structure of GVIIJ[C24S] was the presentation of residue 24, within the ␤-turn of loop 4 (Asp-23-Thr-26), such that it could readily interact with the channel. It was shown recently that the cysteine residue at this position is a key determinant of the off rate of the peptide (4,5). Our results lend support to this because all peptide analogs investigated here exhibited k off values that were close to that of unmodified GVIIJ SSEA (Table 3). Presumably this is because, once bound to the channel, the energetics of dissociation are resilient to changes in individual amino acid side chains. These results suggested that the structural features that had the greatest influence on the biological activity of O §-GVIIJ are those that affect k on .
The solution structure of O §-GVIIJ was solved using an analog in which Cys-24 was replaced with serine to avoid the risk of dimerization during the course of NMR studies. The block by GVIIJ[C24S] was rapidly reversible, in contrast to that of GVIIJ SSC or GVIIJ SSEA , which blocked nearly irreversibly. The differences in kinetics between GVIIJ[C24S] and GVIIJ SSC or GVIIJ SSEA can be attributed to the inability of GVIIJ[C24S] to tether to its binding site on the channel. Given the low nanomolar potency of GVIIJ SSEA for Na V 1.2, this analog was used as the basis for SAR studies to assess the importance of individual non-cysteine residues. Deviations from random coil chemical shift plots were nearly identical for GVIIJ[C24S] and GVIIJ SSEA (Fig. 3), which validated their use for structural and functional studies, respectively.
SAR studies focused initially on mutations of residues in loop 4 (Asp-23-Thr-28) in the vicinity of Cys-24. Alignment of this loop with the free cysteine-containing loops of Na V ␤2 (Phe-23-Thr-28) or ␤4 (Phe-55-Gly-60) revealed modest levels of sequence homology (Fig. 6C) (31,32). This is of interest in light of previous work showing that coexpression of Na V 1 subtypes with Na V ␤2 or ␤4 inhibited the ability of O §-GVIIJ analogs to block (4), suggesting that O §-GVIIJ and the ␤-subunits may interact at the same site on the ectodomain of the ␣-subunit. Numerous residues in this region of the peptide had significant effects on the ability of the analog to fold into a major product, implying that interactions among residues within the ␤-turn of this loop (Asp-23-Thr-26) were important for stabilization of the peptide, even though they had much less bearing on the functional activity of the peptide. Thus, little difference was observed between K d values of loop 4 mutants and that of GVIIJ SSEA (Table 3). Our SAR results nonetheless identified a number of functionally important residues in GVIIJ SSEA , most of which are located in the structurally less well defined loop 2 (Gly-11-Tyr-16). The [K12D], [R14D], and [Y16A] analogs in this loop exhibited on rate constants that were significantly lower than that of the unmodified peptide, giving rise to higher K d values. There is a dearth of medium-and long-range NOEs in this region, and theand -angular order parameters for residues in this region fall below 0.8 (Fig. 5). This loop may adopt a more rigid structure upon interaction with the channel, but further studies will be required to confirm this.
As described in the introduction, the lack of overlap of sites 1 and 8 was observed initially with -KIIIA[K7A] and rNa V 1.2 (4). We had shown previously that -KIIIA and its analogs block rNa V 1.2 and compete for, and can even co-occupy, site 1 with the guanidinium alkaloids tetrodotoxin, saxitoxin, and saxitoxin congeners (9,10,33). In the present experiments, we employed -PIIIA[R14Q] and rNa V 1.4. The site 1 blocker -PIIIA[R14Q] was first characterized at the laboratory of French and co-workers (11), who performed single-channel measurements in planar lipid bilayers to show that this peptide, like -GIIIA[R13Q], blocked rNa V 1.4 with a significant residual current, but unlike -GIIIA[R13Q], which has ϳ100-fold lower affinity than its parent -GIIIA (34), the affinity of  2 blocked rI Na slightly more slowly than it did control I Na (see Table 4 for average k obs values).   Fig. 8, B-D). b Block of rI Na of oocytes pre-exposed to 33 M -PIIIA[R14Q] (see examples in Fig. 8, B-D). c k obs was too large to measure directly, so value shown was obtained using k obs ϭ k on (peptide) ϩ k off with previously published k on and k off values of -GIIIA (25). d Significance of the difference in k obs for the block of control I Na versus that of residual I Na . counterpart (Arg-13 of -GIIIA and Arg-14 of -PIIIA are homologous residues). Thus, we chose to use -PIIIA[R14Q] for our studies here to further explore the distinction between sites 1 and 8. In oocytes expressing rNa V 1.4, -PIIIA[R14Q] blocked with an IC 50 only 7-fold higher than that of native -PIIIA (0.25 versus 0.036 M; Fig. 8A and Ref. 25, respectively). At saturating concentrations, -PIIIA[R14Q] blocked with an rI Na of 27% (Fig. 8A), whereas the native peptide, -PIIIA, blocked with essentially no rI Na (25). -GIIIA blocked the rI Na of -PIIIA[R14Q] with a k obs much smaller than that of control I Na (Fig. 8B and Table 4), consistent with the two peptides compet-ing for the same site (site 1) on the channel. In contrast, the k obs of the block by O §-GVIIJ SSG was not affected by the presence of -PIIIA[R14Q] (Fig. 8C and Table 4), indicating that -PIIIA[R14Q] did not interfere with the block by O §-GVIIJ SSG and therefore that sites 1 and 8 are distinct. These results are consistent with those of earlier competition experiments performed with -KIIIA[K7A] and rNa V 1.2 (4). Finally, the k obs of the dimer of O §-GVIIJ SSG was slightly, but significantly, decreased by the presence of -PIIIA[R14Q] (Fig. 8D and Table 4), suggesting that sites 1 and 8 are close to each other.
Previously we tested seven O §-GVIIJ SSR derivatives (where SR was a different R-group disulfide-bonded to Cys-24, includ- ing Cys-24 of another O §-GVIIJ monomer to form the (O §-GVIIJ) 2 dimer). They all blocked rNa V 1.2 with the same k off and mNa V 1.6 with a k off 17-fold larger than for rNa V 1.2 (5). These results led us to propose that for all seven peptides, the same peptide-channel complex was formed (5). We report here that (O §-GVIIJ) 2 blocked rNa V 1.4 with a k off of 0.0025 Ϯ 0.0009 min Ϫ1 , a value essentially the same as that of GVIIJ SSG , which was 0.0016 Ϯ 0.0008 min Ϫ1 (4) (p ϭ 0.25). Thus, we expand to rNa V 1.4 our proposal that monomeric and dimeric GVIIJ (O §-GVIIJ SSR and (O §-GVIIJ) 2 , respectively) block by forming the same peptide-channel complex.
Combining our structural and functional results, it appears that O §-GVIIJ may interact with two distinct subsites on the channel. Substitution of specific residues, predominantly located in a less well defined region of the peptide (loop 2), had the greatest effects on k on (Fig. 10A). Analogs containing substitutions of important residues in this loop were still very slowly reversible but exhibited much slower on rates and subsequently lower potency compared with GVIIJ SSEA . Comparison of the activity of GVIIJ[C24S] with analogs containing a cysteine at position 24 showed that the identity of the residue at this position was important for k off (Fig. 10B). The C24S replacement led to rapid reversibility of the peptide but did not prevent VGSC inhibition. Presumably, this was because the peptide was no longer capable of undergoing covalent attachment to the channel, but by retaining key residues, the analogs were still able to elicit VGSC inhibition. Previous studies with the free thiol form of O §-GVIIJ (GVIIJ SH ) showed similar results (4,5). The absence of a leaving group attached to Cys-24 prevented efficient disulfide bond formation between the channel and the peptide, which led to rapid reversibility of GVIIJ SH upon washout. Likewise, the important residues in loop 2 (i.e. Lys-12, Arg-14, and Tyr-16) were retained in GVIIJ SH , and therefore the peptide was still capable of block. Thus, our data suggest a "functionally bipartite" mechanism of action where disulfide bond formation through Cys-24 and interactions between Lys-12, Arg-14, and Tyr-16 and the channel both contribute to VGSC blockade (Fig. 10).
O §-GVIIJ and its analogs present a unique opportunity to probe the structure and function of the newly described site 8 on the sodium channel. In addition to the identification of residues that are important for VGSC blockade, SAR studies also identified potential sites for modification (i.e. residues that are noncritical for inhibition), such as residues near the C terminus (e.g. Lys-30 and Asp-31), which could be replaced by residues with reporter groups (e.g. fluorescently labeled amino acids) for the development of peptidic probes to identify the presence or absence of specific VGSC subtypes in different tissue preparations.
Given that O §-GVIIJ is the first conotoxin found to bind by the described "tethering" mechanism, an untapped and significant opportunity now exists to identify additional ligands for site 8. Efforts are ongoing to improve the VGSC selectivity profile of O §-GVIIJ. In addition, mining of venoms of closely FIGURE 10. "Functionally bipartite" mechanism of action for O §-GVIIJ. A, locations of amino acid residues important for on rates (k on ) of O §-GVIIJ against Na V 1.2. Residues with the slowest k on values are colored red. B, location of residue 24 (magenta), which is responsible for the covalent interactions with a free cysteine on the channel (Cys-910). Replacement of Cys-24 with serine prevented disulfide bridge formation between the peptide and the channel and led to rapid reversibility upon washout, indicating that the identity of the residue in position 24 is critical for k off . C, backbone representation of GVIIJ[C24S], which shows the locations of residues deemed particularly important for the activity of O §-GVIIJ. Residues important for the on rate (Lys-12, Arg-14, and Tyr-16) of the peptide are located in loop 2 (red), whereas that affecting the off rate is located within the ␤-turn of loop 4 (magenta).
related Conus species is also underway to identify other members of this peptide family with improved selectivity for painrelevant VGSC subtypes (e.g. Na V 1.3 and 1.7). From a therapeutic perspective, such efforts might prove useful in the identification and development of peptidic drug leads to combat disease states stemming from VGSC dysfunction, such as neuropathic pain, epilepsy, and multiple sclerosis (35,36).