Development of a O-Conotoxin Analogue with Improved Lipid Membrane Interactions and Potency for the Analgesic Sodium Channel NaV 1

Jennifer R. Deuis, Zoltan Dekan, Marco C. Inserra, Tzong-Hsien Lee, Marie-Isabel Aguilar, David J. Craik, Richard J. Lewis, Paul F. Alewood, Mehdi Mobli , Christina I. Schroeder, Sónia Troeira Henriques, and Irina Vetter From the Institute for Molecular Bioscience and Centre for Advanced Imaging, The University of Queensland, St. Lucia, Queensland 4072, Australia, School of Pharmacy, The University of Queensland, Woolloongabba, Queensland 4102, Australia, and Department of Biochemistry and Molecular Biology, Monash University, Wellington Road, Clayton, Victoria 3800, Australia

distinct expression profile and subsequent functional role (1). For example, Na V 1.4 is predominantly expressed in skeletal muscle where it is important in the generation and propagation of action potentials that initiate muscle contraction (2), whereas Na V 1.5 is predominately expressed in cardiac tissue where it is crucial in setting the electrical conduction properties of the heart (3). Several subtypes (Na V 1.1 and Na V 1.6 -1.9) are expressed in peripheral sensory neurons, and it is due to this expression profile that some of these subtypes have gained attention as potential therapeutic targets for pain (4). Na V 1.8 is of particular interest as an analgesic target because it is expressed almost exclusively in peripheral sensory neurons and is present in the majority of nociceptive neurons (5,6) where it is a major contributor to the sodium current underlying action potentials (7)(8)(9). Accordingly, Na V 1.8-null mice have reduced responses to nociceptive mechanical and thermal stimuli as well as deficits in the development of inflammatory pain, visceral pain, and to a lesser extent neuropathic pain (10 -14). Therefore, pharmacological inhibition of Na V 1.8 is considered a promising therapeutic strategy for the treatment of pain.
The recently discovered O-conotoxin MfVIA (15) and the related MrVIA and MrVIB (16) (see Fig. 1A) are some of a few venom-derived peptides that inhibit Na V 1.8 and the only peptides known to selectively inhibit Na V 1.8 over the other neuronal Na V subtypes expressed in peripheral sensory neurons (15,17,18). However, despite this unique selectivity, little is known about structure-activity relationships of the O-conotoxins. This gap in knowledge arises from difficulties in the synthesis of sufficient quantities to permit detailed studies (19), which is in part due to the hydrophobic nature of these peptides. Characterization of the structural elements contributing to the high affinity inhibition of Na V 1.8 by the O-conotoxins is of particular interest given that the O-conotoxins also potently inhibit the skeletal muscle isoform Na V 1.4, and reduced activity at Na V 1.4 while maintaining activity at Na V 1.8 would be required for therapeutic use (20).
The O-conotoxins are gating modifier toxins that bind to site 4 of the sodium channel located on the extracellular loops in Domain II and block Na ϩ conductance by trapping the Domain II voltage sensor in the resting state that leads to inhibition of sodium channel opening (21,22). Although the exact binding site remains to be elucidated, the high hydrophobicity of the O-conotoxins suggests that they bind to the voltage sensor at the interface of, or within, the lipid membrane. Indeed, the ability to partition into the lipid membrane has been shown to be important for the activity of several spider toxins that also target voltage sensor domains, including hanatoxin and SGTx1 (23,24), but the importance of this property for the activity of the O-conotoxins has not been assessed previously. Thus, we set out to determine the contributions of peptide-lipid interactions to the activity of the O-conotoxin MfVIA.
Having recently developed a novel approach to obtain synthetic MfVIA in sufficient amounts for pharmacological characterization (15), we used an alanine-scanning mutagenesis strategy to determine which surface-exposed residues contribute to specific interactions with Na V 1.8 and to define the interaction of MfVIA with lipid membranes. From our mutagenesis studies, we produced an MfVIA analogue with increased potency at Na V 1.8 and demonstrated that this peptide retains analgesic activity in vivo. Our results suggest that the pharmacological effect of MfVIA is driven not only by peptide-ion channel interactions but is also modulated by peptide-lipid interactions.

Experimental Procedures
Peptide Synthesis-MfVIA analogues were synthesized based on the method described previously (15). Chain assembly was performed using standard Fmoc (9-fluorenylmethoxycarbonyl) protocols on 2-chlorotrityl resin. The cysteine side chain-protecting groups used were 4-methylbenzyl, acetamidomethyl, and trityl for the Cys I,IV , Cys II,V , and Cys III,VI pairs, respectively.
Cleavage from the resin was accomplished by treatment with 10% AcOH, 10% trifluoroethanol, dichloromethane for 1 h at room temperature followed by precipitation with n-hexane. The product was redissolved in 1,1,1,3,3,3-hexafluoropropan-2-ol (5 ml) and added dropwise to a stirred solution of I 2 (5 eq) in 10% 1,1,1,3,3,3-hexafluoropropan-2-ol, dichloromethane (20 ml) over 5 min. Stirring was continued for a further 5 min before it was poured into a solution of 0.2 M ascorbic acid, 0.5 M NaOAc in H 2 O. The aqueous phase was extracted with dichloromethane (two times), and the combined organic layers were washed with water (two times). Following removal of solvent under reduced pressure, the product was lyophilized from 1,4dioxane/MeCN/H 2 O. This product was redissolved in N,N-dimethylformamide (20 ml) and added dropwise to a stirred solution of I 2 (8 eq) in N,N-dimethylformamide (20 ml) over 25 min, and stirring was continued for a further 25 min. The solution was then poured into an ice-cold solution of ascorbic acid in water. The resulting precipitate was collected by filtration, washed with water, and lyophilized from 1,4-dioxane/MeCN/ H 2 O. To effect removal of the remaining non-cysteine side chain-protecting groups, the above product was treated with 95% TFA, 2.5% triisopropylsilane, 2.5% H 2 O for 2 h at room temperature. After most of the cleavage solution was evaporated under a stream of N 2 , the product was precipitated and washed with Et 2 O and lyophilized. Cys(4-methylbenzyl) groups were removed by treatment with 90% HF, 10% p-cresol for 1 h at 0°C, and following removal of HF the product was precipitated and washed with Et 2 O and lyophilized. The crude product was dissolved in 0.1% TFA, 50% MeCN, H 2 O, and a solution of I 2 in MeCN (10 mg/ml) was added until a yellow color persisted. After stirring for 5 min, ascorbic acid was added until the solution became colorless, and product was isolated by preparative reversed phase HPLC.
A NOESY spectrum (200 ms) of MfVIA run at 298 K was used for assignment and peak integration. Following assignment and integration of sequential and intraresidue NOEs for MfVIA, the remaining peaks were integrated, and a list of interproton distances was generated from chemical shifts and NOE intensities using the AUTO function in CYANA 3.0 (31). When more than 92% of all the picked peaks (674 of 727) were appropriately assigned, the chemical shift and distance restraint lists generated by the AUTO function in CYANA were used for further structure calculations, and additional restraints, including disulfide bonds and dihedral angle restraints, were included.
Constraints for the and the backbone dihedral angles were generated using TALOS-N (32) from H␣, C␣, C␤, HN, and nitrogen chemical shifts derived from 1 H-13 C HSQC and 1 H-15 N HSQC spectra. In total, 25 and 22 backbone dihedral angles were included in the structure calculation. The ANNEAL function in CYANA was used to perform 15,000 steps of torsion angle dynamics to generate an ensemble of an initial 200 structures from which 20 of the structures with the lowest energy were chosen for analysis. Several rounds of structure calculations were performed to resolve distance and angle constraint violations. Using protocols from the RECOORD database (33), an ensemble of 50 structures was subsequently calculated within CNS (34) using the force field distributed with Haddock 2.0 (35), and the 50 structures generated were further refined in a water shell (36). A final set of 20 structures based on the lowest energy and one NOE violation greater than 0.2 Å and two dihedral violations greater than 2°was selected for submission to the Protein Data Bank (Protein Data Bank code 2N7F; BioMagResBank accession number 25804). The structures were visualized using MOLMOL (37), and figures were generated using MOLMOL and MacPyMOL (PyMOL Molecular Graphics System, version 1.6, Schrödinger, LLC).
Cell Culture-HEK293 cells stably expressing human Na V 1.8 (SB Drug Discovery, Glasgow, UK) were cultured in minimum essential medium Eagle containing 10% (v/v) fetal bovine serum supplemented with 2 mM L-glutamine and selection antibiotics as recommended by the manufacturer. CHO cells stably expressing human Na V 1.8 in a tetracycline-inducible system (ChanTest, Cleveland, OH) were cultured in Ham's F-12 containing 10% (v/v) fetal bovine serum and selection antibiotics as recommended by the manufacturer. To induce hNa V 1.8 expression, cells were cultured in the presence of tetracycline (1 g/ml) for 24 h at 27°C. Cells were grown in a humidified 5% CO 2 incubator at 37°C, grown to 70 -80% confluence, and passaged every 3-4 days using TrypLE Express (Invitrogen). Functional characterization of the cell lines confirmed stable expression of Na V channels consistent with the expected pharmacology of Na V 1.8 (data not shown).
FLIPR Membrane Potential Assay-HEK293 cells stably expressing hNa V 1.8 were plated 48 h before the assay on 384well black walled imaging plates at a density of 10,000 -15,000 cells/well and loaded with red membrane potential dye (Molecular Devices, Sunnyvale, CA) diluted in physiological salt solution (140 mM NaCl, 11.5 mM glucose, 5.9 mM KCl, 1.4 mM MgCl 2 , 1.2 mM Na 2 H 2 PO 4 , 5 mM NaHCO 3 , 1.8 mM CaCl 2 , 10 mM HEPES ) according to the manufacturer's instructions for 30 min at 37°C. MfVIA and MfVIA analogues were diluted in 0.1% bovine serum albumin (BSA) to avoid adherence to plastic surfaces. After addition of MfVIA analogues using the FLIPR-TETRA (Molecular Devices), cells were incubated for a further 30 min before stimulating with deltamethrin (150 M) to ensure steady-state inhibition of Na V responses. Changes in membrane potential were assessed using the FLIPR TETRA (excitation, 515-545 nm; emission, 565-625 nm) every 2 s for 30 min after adding the agonist.
Electrophysiology-CHO cells stably expressing hNa V 1.8 were passaged 48 h prior to the assay in a T-175 flask and cultured in selection antibiotic-free medium at 37°C. CHO cells were used for electrophysiology assays as Na V 1.8-HEK cells are unsuitable for patch clamp experiments using the QPatch (Sophion Bioscience, Ballerup, Denmark). Cells were harvested at 70 -80% confluence by washing with Ca 2ϩ -and Mg 2ϩ -free Dulbecco's phosphate-buffered saline and incubated with Detachin (Bio-Scientific, New South Wales, Australia) at 37°C for 2-5 min. Dissociated cells were then resuspended in Ex-Cell animal component-free CHO medium with 25 mM HEPES (Sigma-Aldrich), transferred to the QStirrer (Sophion Bioscience), and allowed to recover for 30 min.
The extracellular solution contained 145 mM NaCl, 4 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES, and 10 mM glucose. The pH was adjusted to 7.4 with NaOH, and osmolarity was adjusted with sucrose to 305 mosM. The intracellular solution contained in 140 mM CsF, 1 mM EGTA, 5 mM CsOH, 10 mM HEPES, and 10 mM NaCl. The pH was adjusted to 7.3 with CsOH, and osmolarity adjusted with sucrose to 320 mosM. MfVIA and (E5K,E8K)MfVIA were diluted in extracellular solution with 0.1% BSA at the concentrations stated.
Whole-cell patch clamp experiments were performed on a QPatch-16 automated electrophysiology platform using 16channel planar patch chip plates (QPlates, Sophion Bioscience) with a patch hole diameter of 1 m and resistance of 2 Ϯ 0.02 megaohms. Cell positioning and sealing parameters were set as follows:positioningpressure,Ϫ60millibars;minimumsealresistance, 0.1 gigaohm (average, 1.2 Ϯ 0.1 gigaohms); holding potential, Ϫ100 mV; holding pressure, Ϫ20 millibars. Wholecell currents were filtered at 5 kHz and acquired at 25 kHz. Current (I)-voltage (V) relationships were obtained with a holding potential of Ϫ80 mV followed by a prepulse of Ϫ100 mV for 50 ms and a series of 50-ms step pulses that ranged from Ϫ80 to 50 mV in 5-mV increments before returning to a holding potential of Ϫ80 mV (repetition interval, 5 s). To investigate the effects of MfVIA on voltage dependence of fast inactivation, cells were clamped at a holding membrane potential of Ϫ90 mV before a series of single prepulses of 500 ms ranging from Ϫ120 to Ϫ10 mV in 10-mV increments followed by a test pulse of ϩ10 mV (repetition interval, 30 s). For concentration-response curves at Na V 1.8, a holding potential of Ϫ80 mV was used followed by a prepulse of Ϫ120 mV for 100 ms and then a test pulse of ϩ10 mV for 20 ms (repetition interval, 20 s). Toxin effects were compared with pretoxin control parameters within the same cell once steady-state inhibition was achieved.
Peptide-Lipid Interactions Followed by SPR-SPR measurements were conducted at 25°C with an L1 biosensor chip in a Biacore 3000 instrument (GE Healthcare). Peptides were injected over lipid bilayers deposited onto the chip surface using 10 mM HEPES buffer, pH 7.4, containing 150 mM NaCl as running buffer and following protocols described previously (40,41). Response units were normalized to peptide-to-lipid ratio to normalize the response for the different peptides and lipid systems studied (40,41). Because of SPR and methodology detection limitations, the peptide concentrations required to study and detect peptide-membrane interactions are much higher (low to high M range) than their inhibitory potency (sub-to low M range). In addition, peptide-lipid interactions are often complex and do not typically occur in 1:1 stoichiometry as each peptide molecule is likely to interact with several lipids in the bilayer. Thus, a direct comparison of fitted K d values obtained from SPR with IC 50 values for Na V inhibition is potentially confusing. We therefore present binding data as the normalized peptide-to-lipid ratio to enable comparison between the various peptide-lipid systems (40,41).
Birefringence (⌬n f ) Analysis-The impact of each peptide on the molecular packing order (birefringence) of lipid bilayers adsorbed onto the silicon oxynitride chip surface were studied by dual polarization interferometry. Solutions of 5, 10, 20, and 40 M peptide in 10 mM HEPES, pH 7, 150 mM NaCl were injected consecutively onto the lipid bilayers at 20°C. The mass of an adsorbed molecular layer was calculated with the use of the De Feijter formula (42) where the dn/dc values of 0.135 and 0.182 ml/g were used for calculating the adsorbed lipids and peptides, respectively (43,44). Birefringence (⌬n f ) was obtained by calculating the difference between two effective refractive indices, namely refractive index of transverse magnetic (TM) waveguide mode (n TM ) and refractive index of transverse electric (TE) waveguide mode (n TE ) (43)(44)(45). The respective effective adlayers n TM and n TE corresponding to the measured TM and TE phase changes are calculated by fitting data to a waveguide equation (44 -46). For an anisotropic layer, the difference between the n TM and n TE will be the true effective birefringence of the adlayer (44 -46). A fixed refractive index value of 1.47 was used to obtain the birefringence of adsorbed lipid bilayer.
Exposition of the Trp residue to the aqueous environment was examined by Trp fluorescence emission quenching induced by acrylamide, and the membrane depth location of the peptides was examined by a differential fluorescence quenching approach using 5-and 16-doxyl stearic acid (5-and 16-NS) as before (47). The fluorescence intensity (excitation at 290 nm; emission at 345 nm) of 25 M MfVIA or (E5K, E8K)MfVIA in the presence of 2 mM POPC/POPS (4:1) vesicles was monitored upon titration with acrylamide, 5-NS, or 16-NS stock solutions. Data were represented with a Stern-Volmer plot and fitted with the linear Stern-Volmer equation or with the modified Lehrer equation when plots showed a negative deviation (47). The quenching efficiency was quantified by the Stern-Volmer constant (K SV ) (47).
Formalin Test in Mice-For behavioral assessment, we used adult male C57BL/6J mice aged 6 -8 weeks. Animals were housed in groups of three or four per cage under 12-h light-dark cycles and had standard rodent chow and water ad libitum.
Ethical approval for in vivo experiments in animals was obtained from the University of Queensland animal ethics committee. Experiments involving animals were conducted in accordance with the Animal Care and Protection Regulation Queensland (2012), the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes, 8th edition (2013), and the International Association for the Study of Pain Guidelines for the Use of Animals in Research.
Formalin (16% (w/v) formaldehyde; Thermo Scientific, Australia) was diluted in saline, 0.1% (w/v) BSA to a 1% (w/v) formaldehyde solution Ϯ (E5K,E8K)MfVIA (10 M) and administered by intraplantar injection to the left hind paw of mice in a volume of 20 l under light isoflurane (3%) anesthesia. Immediately after injection of formalin, mice were placed individually into acrylic boxes (10 ϫ 10 ϫ 10 cm), and spontaneous pain was quantified by counting the number of pain behaviors (paw lifts, licks, shakes, and flinches) by a blinded observer (unaware of each individual animal's treatment) in 5-min intervals from video recordings. Phase I and Phase II were defined as the cumulative pain behaviors that occurred from 0 to 10 min and from 10 to 45 min postinjection, respectively.
Parallel Rod Floor Test-Motor performance was assessed using the parallel rod floor test and analyzed using ANY-Maze software (Stoelting Co., Wood Dale, IL). (E5K,E8K)MfVIA (30 M) or saline, 0.1% BSA (vehicle control) was administered by intraplantar injection in a volume of 20 l under light isoflurane (3%) anesthesia 10 min prior to the parallel rod floor test. Mice were then placed in the parallel rod floor test apparatus, and the distance traveled (m) and number of foot slips were recorded over 1 min using the ANY-Maze software. The ataxia index, indicative of motor performance, was defined as the number of foot slips per meter traveled.
Data Analysis and Statistics-The data were plotted and analyzed by GraphPad Prism, version 6.0. For concentration-response curves, a four-parameter Hill equation with variable Hill coefficient was fitted to the data using the following equation: IC 50 values were presented as the mean Ϯ S.E. of the negative logIC 50 (pIC 50 ) from at least three independent experiments with n ϭ 3 points per concentration. Current (I norm ) plotted for the I-V curve and fast inactivation analysis was normalized using the following equation: I norm ϭ I/I max where I is the current and I max is peak current. Conductance (G) plotted for the G-V curve was normalized using the following equation: G ϭ where V is the test voltage and V rev is the reversal potential. Normalized G-V and voltage dependence of inactivation data were fit to the Boltzmann equation. Statistical significance was defined as p Ͻ 0.05 and was determined by an unpaired t test assuming equal variance or one-way analysis of variance analysis with Dunnett's post-test as indicated. Data are expressed as the means Ϯ S.E.

MfVIA Is a Gating Modifier Toxin That Interacts with Lipid
Membranes-The O-conotoxins are typically considered gating modifier toxins; however, the mechanism of action of MfVIA has not been previously determined. We thus examined the effect of MfVIA on the I-V relationship as well as the voltage dependence of activation and fast inactivation at Na V 1.8, the isoform most potently inhibited by the O-conotoxins (15,18). MfVIA inhibited peak current ( Fig. 2A) and caused a small leftward shift in the I-V curve (Fig. 2B) with a corresponding small but significant shift in the voltage dependence of activation (V 1/2 Na V 1.8: control, Ϫ14 Ϯ 0.5 mV; MfVIA (3 M); Ϫ22 Ϯ 0.6 mV; p Ͻ 0.01 determined by t test ; Fig. 2C) and a small but significant shift in the voltage dependence of fast inactivation (V 1/2 Na V 1.8: control, Ϫ49 Ϯ 0.7 mV; MfVIA (3 M), Ϫ53 Ϯ 0.8 mV; p Ͻ 0.01 determined by t test ; Fig. 2C), confirming gating modifier activity. In addition, MfVIA inhibited Na V 1.8 using a FLIPR membrane potential assay (pIC 50 : HEK-Na V 1.8, 5.74 Ϯ 0.09; CHO-Na V 1.8, 5.68 Ϯ 0.02) with similar potency to the patch clamp assay, validating that this high throughput approach can be used to assess the activity of MfVIA and compare it with that of analogues.
The voltage sensor domains targeted by gating modifier toxins are embedded within the cell membrane, and it has been proposed that gating modifier peptides act by inserting into the lipid bilayer to access the voltage sensor domain. Thus, to determine whether MfVIA partitions into lipid membranes, we examined the membrane binding affinity of MfVIA using SPR and model membranes (Fig. 2D). Membranes of eukaryotic cells are fluid with the outer leaflet rich in lipids with phosphatidylcholine headgroups, raft domains rich in Chol and SM, and an inner leaflet that is negatively charged due to the presence of phosphatidylserine phospholipids. The three lipid systems tested were POPC, which mimics the overall membrane fluidity and the neutral charge of the outer layer in eukaryotic cells; POPC/Chol/SM (27:33:40 molar ratio), which mimics raft domains and their liquid-ordered phase; and POPC/POPS (80:20 molar ratio), which mimics the negative charge in the inner leaflet. MfVIA had higher affinity for POPC and POPC/ POPS than for POPC/Chol/SM, demonstrating a preference to bind to membranes in the fluid phase rather than membranes in the liquid-ordered phase (Fig. 2D). A preference for the fluid phase over more ordered phase is common in peptides and suggests that MfVIA has preference for fluid regions over raft domains. The linear reduced form of MfVIA had a significantly lower ability to bind to membranes compared with the native form (Fig. 2E), indicating that the overall three-dimensional structure of MfVIA is important for efficient binding to membranes. These results suggest that interaction of MfVIA with lipid membranes might play a role in its pharmacology.
NMR Characterization of MfVIA-We carried out chemical shift analysis and full structure determination on native MfVIA. NMR experiments were run in 70% H 2 O, 30% CD 3 CN at 298 K for optimal solubility of the peptides. The H␣ secondary chemical shifts of MfVIA were similar to published shifts of MrVIB (Protein Data Bank code 1RMK) (17) (Fig. 1B), suggesting a similar structural fold and disulfide bond connectivity (Cys I -Cys IV , Cys II -Cys V , and Cys III -Cys VI ). A set of 50 structures was calculated using 438 restraints, including 117 intraresidue, 159 sequential, 34 medium range (i Ϫ j Ͻ 5) and 128 long range distance restraints (i Ϫ j Ն 5), and 47 dihedral angle restraints with 25 ⌽ and 22 angles derived from TALOS-N. The 20 lowest energy structures with the fewest violations (Fig. 1C) (49) surrounded by a number of loops and turns. A short antiparallel ␤-sheet is present in loop 4, including residues 25-26 and 31-32, which is characteristic of peptides containing the inhibitory cystine knot motif (49).

Alanine Scanning of MfVIA Reveals Important Functional
Residues-To determine which MfVIA residues are important for binding to Na V 1.8, we performed an alanine scan on MfVIA surface-exposed residues identified based on the NMR structure, including residues Glu-5, Lys-6, Trp-7, Glu-8, Ile-14, Leu-15, Phe-17, Val-18, Tyr-19, and Phe-29, as these residues are most likely to interact with the channel binding site and/or lipid membrane. All the MfVIA analogues possess identical H␣ secondary chemical shifts to MfVIA, suggesting a similar structural fold and disulfide bond connectivity (data not shown). Analogues with decreased lipid membrane interactions (W7A and F29A) showed a decrease in potency at Na V 1.8, consistent with membrane interaction being important for pharmacological activity (Fig. 3, A and B, and Table 2). Despite a loss of potency at Na V 1.8, interactions of (I14A)MfVIA and (Y19A)-MfVIA with lipid membranes were not significantly affected (Fig. 3C), suggesting that these residues may be involved directly with binding to the Na V channel or that these residue replacements introduce minor structural perturbations that affect binding to the channel but that are not sufficient to alter membrane binding. It should be noted that replacement of a single residue does not abolish the interaction with membranes, suggesting that overall hydrophobicity, rather than the individual residues, is important for the ability of MfVIA to bind to lipid bilayers.
Replacement of the negatively charged glutamic acid residues Glu-5 and Glu-8 with an alanine residue led to a small increase in potency at Na V 1.8 with a small reduction in lipid membrane binding (Fig. 3C), suggesting that a negative charge at these positions is not required for activity at Na V 1.8. We thus chose to further explore the contribution of Glu-5 and Glu-8 as well as Tyr-19 because the (Y19A)MfVIA mutant lost activity at Na V 1.8 by generating additional analogues.

Positively Charged Residues and Increased Affinity for Lipid
Membranes Improve MfVIA Potency at Na V 1.8 -To explore the contribution of Tyr-19 to Na V activity and membrane interactions, we replaced the hydrophobic tyrosine residue at position 19 with phenylalanine or with tryptophan by producing the analogues (Y19F)MfVIA and (Y19W)MfVIA (Fig. 4, A and B, and Table 2). Replacing tyrosine with a phenylalanine at position 19 did not significantly affect potency at Na V 1.8, suggesting that the polar hydroxyl group on Tyr-19 is not essential for binding. Increasing the hydrophobicity of Tyr-19 by replacing the tyrosine with a tryptophan did not improve activity further and in fact led to a 2-fold loss of potency at Na V 1.8. In addition, MfVIA and (Y19W)MfVIA have identical affinity for lipid bilayers (data not shown), providing evidence that Tyr-19 is more important for interacting with the binding site rather than the lipid membrane.
As the replacement of the negatively charged residues Glu-5 and Glu-8 with the neutral amino acid alanine led to an increase in potency at Na V 1.8, we assessed the effect on potency and selectivity when Glu-5 and Glu-8 are replaced with positively charged residues. We chose to make a double (E5K,E8K)MfVIA mutant as the O-conotoxins MrVIA and MrVIB already have a lysine at the equivalent Glu-5 position, suggesting that mutating this residue alone would not greatly improve activity. Replacement of both negatively charged glutamic acids with positively charged lysines at positions 5 and 8 led to a 4-fold increase in potency at Na V 1.8 (Fig. 4, A and B). The affinity of (E5K,E8K)MfVIA for POPC and for the anionic POPC/POPS membrane increased significantly (Fig. 4C) compared with MfVIA, suggesting that Na V 1.8 potency can be driven by enhanced peptide-lipid interactions.
(E5K,E8K)MfVIA Inserts More Deeply in the Membrane than MfVIA-To gain additional insights into the membrane binding properties and the membrane depth location of MfVIA and (E5K,E8K)MfVIA, we examined their intrinsic fluorescence properties, specifically the fluorescence emission of residue Trp-7 in the absence/presence of vesicles. For comparison, we also examined (F29A)MfVIA, an analogue with lower affinity for membranes than the wild-type peptide. The fluorescence emission maxima of MfVIA and its analogues in buffer (10 mM HEPES, 150 mM NaCl, pH 7.4) were the same (357 nm) and identical to L-Trp alone (Fig. 6A), suggesting that Trp-7 in . Functional Na V activity and lipid membrane binding affinity of MfVIA and its Ala mutants. A, functional activity of MfVIA and its Ala mutants at Na V 1.8 assessed using membrane potential assays. Data are presented as mean Ϯ S.E. from three independent experiments. Statistical significance was determined using one-way analysis of variance analysis with Dunnett's post-test. *, p Ͻ 0.05 compared with native MfVIA. Error bars represent S.E. B, representative concentration-response curves for MfVIA and its Ala mutants at Na V 1.8 from the membrane potential assay. C, binding of MfVIA and its Ala mutants to POPC membranes measured by SPR. Peptide-to-lipid ratio (mol/mol) obtained at the end of the peptide injection was plotted as a function of the peptide concentration injected over the lipid surface. MfVIA, and in its analogues, is solvent-exposed and therefore can be used to examine insertion into the membrane upon titration with lipid. The fluorescence emission spectra of (E5K, E8K)MfVIA upon titration with POPC/POPS (4:1) vesicles displayed a blue shift (357 nm in buffer to 344 nm with 4 mM lipid) and an increase in the quantum yield (Fig. 6, B and C), typical of Trp residues moving toward a more hydrophobic environment (38). MfVIA showed a shift in the fluorescence emission spectra (358 nm in buffer compared with 351 nm with 4 mM lipid) but not in the quantum yield, whereas (F29A)MfVIA had neither a shift nor an increase in the quantum yield (see Fig. 6, B and C). Overall, these results suggest that the Trp residue in MfVIA is located close to, but not deeply inserted into, the membrane, whereas in (E5K,E8K)MfVIA the Trp is involved in insertion into the membrane, and in (F29A)MfVIA the Trp residue has a weak interaction with the membrane. To gain information on the depth of MfVIA and (E5K, E8K)MfVIA within the membrane, we used a differential quenching strategy in which we compared the fluorescence quenching efficiency induced by three fluorescence quenchers, namely acrylamide, 5-NS, and 16-NS, each having distinct locations in the lipid membrane. Acrylamide, an impermeable quencher that does not insert into membranes, was used to quench the fluorescence emission of Trp exposed in solution, whereas 5-NS quenches the fluorescence of Trp that is located closer to the membrane interface, and 16-NS quenches the fluorescence emission of Trp that is more deeply inserted (47). The quenching efficiency was quantified by K SV . Quenching efficiency by acrylamide (Fig. 6D) followed the order L-Trp Ͼ MfVIA Ͼ (E5K,E8K)MfVIA, confirming that in the presence of lipid vesicles the Trp residue in MfVIA is not fully exposed to aqueous solution when lipid vesicles are present, and it is even less exposed in (E5K,E8K)MfVIA. The Trp fluorescence of (E5K,E8K)MfVIA was more efficiently quenched by both 5and 16-NS (Fig. 6E) than the Trp fluorescence of MfVIA, confirming that (E5K,E8K)MfVIA inserts more deeply in the membrane compared with MfVIA. The deeper insertion of (E5K,E8K)MfVIA into the membranes was further examined by changes in the molecular packing order of lipid bilayers. (E5K,E8K)MfVIA caused a greater bilayer destabilization to both POPC and POPC/POPS membranes at concentrations greater than 10 M, whereas MfVIA did not, confirming that the double mutant inserts into the membrane bilayer (Fig. 6, F-I).

Discussion
The O-conotoxins are among the few venom-derived peptides with activity at the pain target Na V 1.8; however, little is known about the residues or structural elements that define their Na V selectivity. In this study, we report the first membrane binding characterization and structure-activity study of a O-conotoxin in which we utilized an alanine-scanning strategy of the main solvent-exposed residues of MfVIA to identify residues that are most important for Na V 1.8 activity and for affinity for lipid bilayers. This led to the mutant (E5K, E8K)MfVIA, which was 3-fold more potent at Na V 1.8 and had greater interactions with lipid membranes.
We used the three-dimensional solution structure of MfVIA to evaluate whether certain surface characteristics could explain the enhanced binding of (E5K,E8K)MfVIA to lipid membranes and the increase in Na V 1.8 selectivity compared with its native counterpart. MfVIA differs from MrVIB only by four residues in the N-terminal region (see Fig. 1A). MfVIA has two residues, Arg-1 and Asp-2, preceding the first Cys residue compared with MrVIB, which only has one, Ala-1. The two residues following Cys I are Gln-4 and Glu-5 in MfVIA compared with Ser-3 and Lys-4 in MrVIB. Although these differences appear minor, they greatly affect the surface characteris-   tics of the two peptides. Both peptides are extremely hydrophobic, resulting in the challenging synthesis mentioned previously. Although MrVIB has two Lys residues in positions 4 and 5 (MrVIB numbering), they do not appear to be as surfaceexposed as expected; hence, the overall surface of profile of MrVIB is mainly hydrophobic (Fig. 4F). The additional Arg-1, Asp-2, and Gln-4 residues in MfVIA therefore play a significant role in the overall surface charge of the native peptide MfVIA, and by replacing the negatively charged Glu-5 and Glu-8 with positively charged Lys, the surface of the peptide appears much more similar to the amphipathic spider toxins previously described to interact with model membranes (Fig. 4, D and E) (24). Loop 2 of MfVIA was found to be disordered similar to MrVIB (17). This is not surprising as this loop is the longest in the peptide with very few stabilizing restraints. This phenomenon has been observed previously for other inhibitory cystine knot peptides, including -conotoxin MVIIA (52) and -conotoxin PVIIA (53). In the case of -conotoxin MVIIA, this flexible loop includes Tyr-13, the most critical residue for binding, whereas our structure-activity data suggest that loop 2 of MfVIA is less involved in binding to the receptor or lipid membranes.
The O-conotoxins MrVIA and MrVIB inhibit Na ϩ current without significantly shifting the voltage dependence of activation or inactivation, which is characteristic of toxins that inhibit Na ϩ current by physically occluding the pore of the channel (20,54). However, domain swapping, site-directed mutagenesis, and competitive binding experiments have localized the O-conotoxin binding site to the voltage-sensing regions located on Domain II, providing evidence that O-conotoxins act as gating modifiers that inhibit Na ϩ current by trapping the Domain II voltage sensor in the closed position to prevent opening of the channel (21,22). This mechanism of action is also reported for several spider peptides that inhibit Na V channels by binding to the extracellular loops of the same domain (55)(56)(57). Accordingly, MfVIA inhibited peak Na ϩ current, causing small hyperpolarizing shifts in the voltage dependence of activation and inactivation at Na V 1.8. Hyperpolarizing shifts in the voltage dependence of activation are unexpected for an inhibitor of Na V channels, although the physiological relevance of this small shift is likely negligible given that Na V 1.8 is only activated at very depolarized membrane potentials in comparison with the other Na V subtypes (9). Indeed, no spontaneous pain behaviors, like those seen after administration of the nonselective Na V activators veratridine and grayanotoxin III (58), were observed after intraplantar injection of MfVIA alone in naïve mice (data not shown), consistent with the lack of any significant excitatory activity of MfVIA.
Domain II comprises four segments (S1-S4) embedded in the membrane that form the voltage-sensing segments; these are connected by two extracellular loops known as the S1-S2 linker and the S3-S4 linker. The precise binding site of the O-conotoxins is unknown but probably involves interactions with the extracellular loop(s) and part of the transmembrane regions as MfVIA binds to model lipid membranes. In addition, single mutation of Trp-7 or Phe-29, which are located on the same side on the molecule, with alanines reduces both the functional activity and membrane binding affinity of MfVIA, sug-gesting that membrane interactions are important for access of the peptide to parts of the voltage sensor located in the membrane. Indeed, peptide-membrane interactions are important for the functional activity of the spider peptides hanatoxin and SGTx1, which bind to the equivalent S3-S4 linker site on voltage-gated potassium (K V ) channels (24).
As the binding site of MfVIA appears to involve voltagesensing segments buried in lipid, it would be expected that simply altering the hydrophobicity, and thus the binding affinity for lipid membranes, would affect the potency of MfVIA. However, not all mutations affecting hydrophobicity led to differential effects on potency at Na V 1.8 as replacement of the hydrophobic residues Ile-14 and Tyr-19 with an alanine led to reduction in potency but had no significant effect on membrane binding, suggesting that these residues form interactions with the channel binding site. In addition, (Y19F)MfVIA and MfVIA had similar potency at Na V 1.8, indicating that a hydrogen bond is not formed with the binding site at this position. Interestingly, by increasing both size and hydrophobicity, (Y19W)MfVIA lost potency at Na V 1.8, indicating that the aromatic phenyl ring at this position may form a specific interaction with the binding site.
The double mutant, (E5K,E8K)MfVIA, in which the negatively charged Glu-5 and Glu-8 residues were replaced with lysine residues, had increased lipid binding affinity, inserted more deeply into model membranes, and induced a greater membrane disorder of bilayer packing compared with MfVIA. Interestingly, (E5K,E8K)MfVIA also showed increased potency at Na V 1.8, although full selectivity of (E5K,E8K)MfVIA at human Na V 1.1-Na V 1.7 remains to be determined. Although (E5K,E8K)MfVIA has increased membrane binding affinity and enhanced inhibitory potency at Na V 1.8, its Na V 1.8 inhibition mechanism is identical to that of MfVIA (see Figs. 2C and 5F), suggesting that both peptides have the same binding site.
As well as having higher affinity for membranes, (E5K, E8K)MfVIA has an apparent decrease in depolarization-induced dissociation compared with MfVIA. Relief of block induced by strong depolarizations has been reported previously for MrVIA and was attributed to a membrane potential-dependent effect on K off (59). Similarly, we observed reversal of block during repetitive depolarizations with MfVIA but not with (E5K,E8K)MfVIA (Fig. 5, C and D). This suggests that for (E5K,E8K)MfVIA deeper insertion into the membrane may play a role in increasing the concentration of peptide in the vicinity of its channel and improve the interaction with its binding site, which in turn leads to enhanced potency at Na V 1.8. Given that a similar depolarization-induced dissociation for MrVIB has been reported for Na V 1.4 (22), it is plausible that improved membrane affinity and potency do not intrinsically lead to improved selectivity for Na V 1.8. Further combinations of mutations that improve membrane binding and mutations that are associated with apparent specific receptor interactions, such as Ile-14 and Tyr-19, may lead to further improvements in potency as well as selectivity for Na V 1.8. An important point to note is that, because Na V 1.8 is expressed in sensory nerve terminals whereas Na V 1.4 is the predominant skeletal muscle isoform, the membrane composition of native tissues where these isoforms are expressed may differ substantially. It is possible that such differences may contribute to the relative lack of skeletal muscle side effects we observed after local injection, although this remains to be validated.
(E5K,E8K)MfVIA significantly reduced formalin-induced pain behaviors in mice; however, the in vivo target mediating the analgesic effects of (E5K,E8K)MfVIA remains to be confirmed. It is plausible that the reduction of formalin-induced nocifensive responses is mediated by Na V 1.8 given that nonselective Na V inhibitors like local anesthetics completely abolish formalin-induced flinching (60,61). Although (E5K,E8K)-MfVIA significantly reversed formalin-induced pain behavior, only partial analgesia could be achieved even at a high intraplantar concentration. It remains unclear whether unfavorable pharmacokinetics contribute to this effect. Intraplantar injection of MrVIB caused a similar reduction in formalin-induced pain behavior, suggesting that inhibition of Na V 1.8 only achieves partial analgesia in the formalin model (18).
We chose the formalin test to assess the analgesic effect of (E5K,E8K)MfVIA based on previous reports that the O-conotoxin MrVIB has similar efficacy in the same model. In addition to being a well validated pain target in inflammatory pain and visceral pain and to a lesser extent neuropathic pain (10 -14), Na V 1.8 is resistant to cooling-induced inactivation, making it the only Na V isoform available for conduction of action potentials in cold-associated pain (62). Thus, Na V 1.8 inhibitors such as (E5K,E8K)MfVIA could be particularly effective in conditions associated with thermal pain, including cold pain and cold allodynia, although broader analgesic efficacy in other pain modalities remains to be validated in future studies. Na V 1.8 is a potential therapeutic target for the treatment of pain, but relatively few selective inhibitors are available due to the high sequence similarity between isoforms (Na V 1.1-1.9). Here we show that inhibition of Na V 1.8 by O-conotoxin MfVIA is driven not only by peptide-ion channel interactions but also by peptide-lipid interactions, suggesting that the putative binding site is located in the membrane-embedded voltagesensing domain. Building on these interactions, we have established as the more potent mutant (E5K,E8K)MfVIA, which may guide the design of other novel selective inhibitors of Na V 1.8 with therapeutic potential.