Gating modifier toxins isolated from spider venom: Modulation of voltage-gated sodium channels and the role of lipid membranes

Gating modifier toxins (GMTs) are venom-derived peptides isolated from spiders and other venomous creatures and modulate activity of disease-relevant voltage-gated ion channels and are therefore being pursued as therapeutic leads. The amphipathic surface profile of GMTs has prompted the proposal that some GMTs simultaneously bind to the cell membrane and voltage-gated ion channels in a trimolecular complex. Here, we examined whether there is a relationship among spider GMT amphipathicity, membrane binding, and potency or selectivity for voltage-gated sodium (NaV) channels. We used NMR spectroscopy and in silico calculations to examine the structures and physicochemical properties of a panel of nine GMTs and deployed surface plasmon resonance to measure GMT affinity for lipids putatively found in proximity to NaV channels. Electrophysiology was used to quantify GMT activity on NaV1.7, an ion channel linked to chronic pain. Selectivity of the peptides was further examined against a panel of NaV channel subtypes. We show that GMTs adsorb to the outer leaflet of anionic lipid bilayers through electrostatic interactions. We did not observe a direct correlation between GMT amphipathicity and affinity for lipid bilayers. Furthermore, GMT–lipid bilayer interactions did not correlate with potency or selectivity for NaVs. We therefore propose that increased membrane binding is unlikely to improve subtype selectivity and that the conserved amphipathic GMT surface profile is an adaptation that facilitates simultaneous modulation of multiple NaVs.

Gating modifier toxins (GMTs) 2 extracted from spider venom are a class of peptides that are valuable probes for studying the physiology and pharmacology of voltage-gated ion channels (1)(2)(3)(4). GMTs alter the gating kinetics of voltage-gated ion channels (4), which are transmembrane proteins integral to a range of physiological processes in humans (1,5,6). These GMTs contain six Cys residues arranged to form an inhibitory cystine knot motif (7). In addition, these peptides share a conserved amphipathic surface profile characterized by a high proportion of hydrophobic amino acid residues, such as Trp, Tyr, and Phe, surrounded by a ring of cationic residues, including Arg and Lys, that typically promote peptide-membrane interactions (Fig. 1, A and B) (8).
Several GMTs, including GsMTx-IV, HaTx-I, VsTx-I, ProTx-I, ProTx-II, and SgTx-I, have been shown to bind to model membranes (8 -12). The concept of a trimolecular lipid-peptidechannel complex has subsequently been proposed to exist in interactions between GMTs, voltage-gated ion channels, and the lipid membrane (10,13,14). A trimolecular complex presents the possibility for a novel approach to rational drug design whereby the cell membrane is considered as a third component in addition to the traditional approach, which only takes into account the transmembrane protein and the modulatory ligand (10,14,15).
The objective of the current study was to determine whether there is an overall relationship between the amphipathic surface profile of GMTs and their ability to bind lipid membranes and modulate voltage-gated ion channels. Electrophysiology was used to examine potency of the peptides at voltage-gated sodium channel 1.7 (Na V 1.7), a transmembrane protein that is currently being pursued as a target for development of therapeutics for chronic pain (5,16). GMTs are likely to show toxicity in mammals if they show activity at off-target Na V s; therefore, Na V 1.2, Na V 1.4, Na V 1.5, and Na V 1.6 were included to study whether there is a relationship between GMT amphipathicity, affinity for lipid bilayers, and off-target selec-tivity. Previous reports have examined lipid-binding properties of small GMT cohorts (10,11,17,18), but the current study represents the first concerted effort to study the membrane binding of a cohort of native GMTs in model membranes chosen to mimic physiological properties (15).
CcoTx-I, CcoTx-II, PaurTx-III, and ProTx-I were chosen because these GMTs are known promiscuous modulators of Na V s (19,20) and would therefore provide information on the relationship between GMT amphipathicity, promiscuity for Na V s, and affinity for lipid bilayers. HnTx-IV, HwTx-IV, and GpTx-I were chosen for their known potency at Na V 1.7 (21)(22)(23), and although we have recently examined the lipid affinity of HwTx-IV (24), we have not done so in the context of a comparative analysis with native GMTs. GsMTx-IV is a known modulator of stretch-activated mechano-sensitive channels with a mechanism of action that is primarily related to interactions with the lipid bilayer (12,25) and was included in the present study as a positive control for the lipid affinity studies and as a negative control for the studies on Na V s. SgTx-I is a known modulator of K V channels and is an inhibitor of the voltage of inactivation of Na V 1.2 (26 -28) and, like GsMTx-IV, was included as a negative control for the studies on Na V s. Our results suggest that the conserved amphipathic surface profile of spider-derived GMTs is most probably an adaptation that allows the concomitant modulation of several voltage-gated ion channels and that GMTs have preferential affinity for anionic model membranes.

Oxidative folding of GMTs
To optimize the yield of GMTs with correct disulfide connectivity, each peptide was subjected to 3-5 oxidative folding trials in which the effects of temperature and folding buffer were examined (pH was always 7.7-8.0) (Table S1). Peptide folding was monitored using analytical reverse-phase HPLC (RP-HPLC) and LC/MS ( Fig. S1 and Table S1) Most GMTs were successfully folded at 25°C in buffer containing 7.5% (v/v) acetonitrile (ACN) (method A). GsMTx-IV was folded using both method A and B but formed aggregates using method B, probably because of the lack of organic solvent (Table S1). Furthermore, oxidation using method B took 24 h for GsMTx-IV compared with 16 h for method A. Formation of disulfide bonds of both HnTx-IV and HwTx-IV proceeded in the absence of organic solvent at room temperature in 16 h (methods C and D), whereas SgTx-I and ProTx-I required a slow reaction in 0.1 M ammonium acetate buffer, using 2 M urea at 4°C for 72 h (method E).

NMR analysis of the GMTs
One-dimensional 1 H spectra (Fig. S2) of all GMTs revealed good dispersion of amide-proton resonances (7-10 ppm), suggesting that the peptides were folded and structured. H␣ chemical shifts derived from two-dimensional TOCSY and NOESY  (30), and SgTx-I (PDB code 1LA4) (32). A, surface representations of the GMTs colored by residue type: hydrophobic (green), positively charged (blue), negatively charged (red), and uncharged (white). All of the peptides are in the same orientation. B, alignment of GMT sequences relative to HnTx-IV and HwTx-IV. Cysteine residues are highlighted in gray. Residues that contribute to the hydrophobic patch (green) and positively charged ring (blue) as well as negatively charged residues (red) are labeled. NT and CT, N and C termini of the peptides, respectively. *, C-terminal amidation (residue numbers are shown below the table).

GMT amphipathicity and Na V inhibition
spectra were used for sequential assignment of the peptides (Fig. 2), and H␣ shifts for GpTx-I (23), HnTx-IV (PDB code 1NIY) (21), HwTx-IV (PDB code 2M4X), (29), and ProTx-I (PDB code 2M9L) (30) were in good agreement with literature values, with small differences attributable to differences in chemical shift referencing or pH (Fig. 2). Structures of GsMTx-IV and SgTx-I are available (31,32), but H␣ shifts have not been reported (Fig. 2). The solution structures and chemical shift assignments for CcoTx-I (PDB code 6BR0; BMRB 30376) and CcoTx-II (PDB code 6BTV; BMRB 30377) are reported here for the first time and have been submitted to the Protein Data Bank and to the BioMagnetic Resonance Bank, respectively ( Fig. 3A and Table S2). The disulfide bridges in these GMTs form a classical inhibitory cystine knot motif in which the cystine knot stabilizes a structure composed of loops, turns, and two antiparallel ␤ strands (Tyr 20 -Cys 22 and Cys 29 -Tyr 31 ). The hydrophobic residues form a cluster with the side chains showingstacking and attractive hydrophobic interactions (Fig. 3B). The structure for PaurTx-III, which we recently solved and submitted to the PDB (PDB code 5WE3; BMRB 30317) is also shown (Fig. 3) (33).

GMT-lipid bilayer interactions
Surface plasmon resonance (SPR) was used to compare GMT affinity for different model membranes. Comparisons are made relative to GsMTx-IV, which had the highest affinity for all of the model membranes ( Fig. 4 and Table S3). Within the context of the current study, affinity refers to amount of peptide bound to lipid (peptide/lipid (P/L) (mol/mol)) and the rate of dissociation of the peptide from the lipid bilayer ( Fig. 4 and Table S3). Here, "weak affinity" refers to GMTs that either dissociate from the lipid bilayers rapidly or do not reach a P/L (mol/mol) that is as high as GsMTx-IV. In some instances, we compare the affinity of specific GMTs with different model membranes, in which case "weak affinity," "high affinity," or "preferential affinity" compares a specific peptide's affinity for one model membrane with another model membrane.
Our examination of GMT interactions with model membranes began with lipid bilayers composed of zwitterionic 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), as phospholipids containing phosphatidylcholine headgroups are the most abundant phospholipids in the outer leaflet of healthy mammalian cells (34). Sensorgrams and concentrationresponse curves revealed that GsMTx-IV had the highest affinity (highest amount of maximum peptide bound to lipid (i.e. 0.071 P/L (mol/mol)) with the slowest dissociation from the POPC model membranes) (Fig. 4 and Table S3). The remaining peptides had weak affinity (0.002-0.031 P/L (mol/mol)) for POPC lipid bilayers ( Fig. 4 and Table S3).
There is evidence that voltage-gated ion channels are embedded within lipid rafts that form domains around these transmembrane proteins for functional and structural integrity (35,36). A mixture of POPC/sphingomyelin (SM)/cholesterol (CHOL) (2.7:4:3.3 molar ratio) was therefore used to mimic the environment formed by lipid rafts (37,38). GsMTx-IV and CcoTx-II showed the highest amount of peptide bound to lipid (0.045 and 0.041 P/L (mol/mol), respectively), although  Overlay of the ensemble of 20 conformers was selected on the basis of lowest energy of minimization and best MolProbity scores. A, schematic of CcoTx-I, CcoTx-II, and PaurTx-III with disulfide bridges shown in gold and cysteines labeled. NT and CT, N and C termini, respectively. B, the three GMTs are oriented the same way as in A with side chains of aromatic residues forming the hydrophobic patch shown in white.
We also examined GMT affinity for anionic POPC/ceramide 1-phosphate (C1P)/CHOL (2.7:4:3.3) model membranes, in which C1P possesses a negatively charged headgroup. It was previously shown that GMTs have stronger affinity for voltagegated ion channels when sphingomyelinase D (SMase D), an enzyme found in the venom of sicariid spiders, hydrolyzes SM to C1P (35,36,38). The GMTs had the highest affinity for this model membrane compared with other model membranes used in this study except for PaurTx-3, which showed weak affinity for this lipid type (0.001 P/L (mol/mol)) ( Fig. 4 and Table S3).

GMT interactions with model membranes upon variation of ionic strength
To examine the importance of electrostatic interactions in driving GMT binding to model membranes, we compared GMT binding to POPC/POPS (4:1) lipid bilayers in buffers of varying ionic strength. All GMTs showed increased binding to lipid vesicles formed in lower-ionic strength buffer (50 mM NaCl) compared with lipid vesicles formed in high-ionic strength buffer (300 mM NaCl) (Fig. 5). PaurTx-III, HnTx-IV, and GpTx-I had fast dissociation rates from lipid bilayers prepared in low ionic strength buffer, suggesting that these three GMTs form weaker electrostatic interactions with POPC/ POPS (4:1) lipid bilayers than the remaining six GMTs (Fig. 5).

Analysis of environment surrounding GMT Trp residues upon lipid titration
The fluorescence emission spectrum of Trp residues depends on their local environment. As a Trp residue on a peptide moves from a polar to an apolar environment, an increase in quantum yield and a leftward (blue) shift of the fluorescence emission spectrum is expected (42,43). These properties were used to examine whether Trp residues found on GMTs insert into the hydrophobic acyl chains of model membranes. Large unilamellar vesicles (LUVs) were used to mimic a planar bilayer surface (42,44). All nine GMTs have a fluorescence emission spectrum with a maximum near 350 nm, which is close to the fluorescence emission of L-Trp in buffer (354 nm), suggesting the presence of solvent-exposed Trp residues and that the fluorescence spectra can be used to gauge whether the peptides move from a polar to an apolar environment. None of the GMTs showed a significant change in the fluorescence emission properties of their Trp residues upon titration with POPC/ POPS LUVs (Fig. 6), suggesting that the Trp residues do not deeply insert into the acyl core of the lipid bilayers but rather adopt a superficial position on the lipid bilayer. The conclusion that the peptides adopt a shallow position is supported by the SPR data for the peptide affinity at POPC/POPS model membranes (Fig. 4). In this assay, none of the peptides bound to the lipid at a ratio of Ͼ0.1 P/L (mol/mol), which suggests that the peptides did not insert deeply into the lipid bilayer. For comparison, [E5K,E8K]MfVIA, a peptide that we have recently reported on, showed both an increase in quantum yield and a blue shift in the emission spectrum (Fig. 6), accompanied by a maximum P/L mol/mol ϳ0.5, as calculated using SPR in the presence of POPC/POPS model membranes (45).

Na V subtype selectivity of GMTs
Two-electrode voltage-clamp electrophysiology was used to examine the effect of GMTs on Na V 1.2, Na V 1.4, Na V 1.5, and Na V 1.6 expressed in Xenopus oocytes (difficulties with expression of Na V 1.7 in oocytes precluded two-electrode voltage clamp studies). The GMTs showed an absence of selectivity when studied against the Na V channels ( Fig. 8 and Table 1). ProTx-I, CcoTx-I, PaurTx-III, and CcoTx-II inhibited 80-100% of current from all of the Na V channels examined (Fig. 8 (A-D) and Table 1). HwTx-IV, HnTx-IV, and GpTx-I had the best overall selectivity, as they only showed complete current inhibition of Na V 1.2 and Na V 1.6 but did not completely inhibit current from Na V 1.4 and Na V 1.5, as seen for CcoTx-I, CcoTx-II, PaurTx-III, and ProTx-I ( Fig. 8 and Table 1). HwTx-IV was equipotent on Na V 1.2 and Na V 1.6 ( Fig. 8 and Table 1), whereas HnTx-IV was slightly more potent at Na V 1.2 and had 2-fold lower potency for Na V 1.6 ( Fig. 8 and Table 1). GpTx-I had the best overall selectivity, with a 6-fold difference in potency between Na V 1.6 and Na V 1.2 ( Fig. 8 and Table 1). GsMTx-IV failed to inhibit 100% of current from any Na V channel subtype, and SgTx-I caused a delay in channel inactivation accompanied by an increase in sodium peak current. Despite these differences between GsMTx-IV and SgTx-I compared with the remaining seven GMTs, both GsMTx-IV and SgTx-I were not selective ( Fig. 8 and Table 1).

Examination of the physicochemical properties of the GMTs
All nine GMTs have similar net positive charge with no observable relationship between net charge and affinity for lipid membranes ( Table 2 and Fig. 9A). GMT hydrophobicity was compared using their RP-HPLC retention times, and the proportion of apolar solvent-accessible surface area (SASA) was used to measure the solvent-accessible hydrophobic area of each peptide. There was no correlation between GMT hydrophobicity or apolar SASA and membrane binding ( Table 2 and Fig. 9 (B and C)). For instance, CcoTx-I and PaurTx-III have retention times of ϳ27 min and apolar SASAs of 57 and 65%, respectively, but these peptides show different affinities for the lipid bilayers examined (Table 2 and Figs. 4 and 9). Likewise, PaurTx-III and GsMTx-IV have the highest percentage of apolar SASAs (65 and 66%, respectively) but consistently had the

GMT amphipathicity and Na V inhibition
weakest and strongest affinities, respectively, for the model membranes studied (Table 2 and Figs. 4 and 9).

Discussion
GMTs isolated from spider venom share a conserved hydrophobic patch surrounded by a ring of positively charged residues that together have been proposed to promote peptide affinity for lipid membranes (Fig. 1) (8, 47). Here, we examined the extent to which nine GMTs bound to a series of model membranes, whether there is a relationship between membrane binding and inhibition of Na V 1.7, and whether GMT amphipathicity is important for membrane binding and/or inhibition of Na V channels.
GMTs showed preferential affinity for negatively charged POPC/C1P/CHOL model membranes compared with those made from POPC/SM/CHOL ( Fig. 4 and Table S3), as observed previously for ProTx-II (38). Interestingly, the pharmacology of some voltage-gated ion channels has previously shown sensitivity to the enzymatic conversion of SM to C1P by SMase D (35,36); it is possible that sicariid spiders take advantage of the presence of SMase D, and perhaps additional venom components, to increase GMT binding to both cell membranes and the voltage-gated ion channels to optimize activity (38). This hypothesis is supported by a study showing that ProTx-I has increased affinity for voltage-gated ion channels upon conversion of SM to C1P by SMase D (36).   . Na V subtype selectivity of GMTs. Two-electrode voltage clamp electrophysiology on oocytes was used to examine the effects of GMTs Na V 1.2, Na V 1.4, Na V 1.5, and Na V 1.6. All GMTs promiscuously modulated the activity of all Na V channels tested, with GpTx-I, HwTx-IV, and HnTx-IV showing the best overall selectivity. SgTx-I showed an increase in current influx for Na V channel subtypes; therefore, I 30 ms /I peak is shown for this peptide. Data for the remaining eight peptides were normalized to maximum current inhibition. Error bars, S.E., where n ϭ 3.

GMT amphipathicity and Na V inhibition
The preferential affinity of GMTs for anionic compared with zwitterionic model membranes was further observed in a higher GMT affinity for POPC/POPS and POPC/PI compared with POPC lipid bilayers (Fig. 4). This affinity is presumably driven by electrostatic interactions between the positively charged peptides and anionic lipid headgroups and may be indicative of GMT affinity for negative moieties on the outer leaflet of cell membranes ( Table 2). The importance of electrostatic interactions is further supported by the observed increase in GMT affinity for POPC/POPS with low ionic strength (see results with 50 mM NaCl; Fig. 5). These results agree with previous studies that also emphasized the importance of electrostatic interactions in driving GMT binding to lipid bilayers (10,18,38,39).

Analysis of the environment surrounding GMT Trp residues in anionic lipids
The absence of a blue shift or change in quantum yield in the fluorescence spectra of the GMTs in the presence of anionic lipids (Fig. 6) suggests that they adopt a shallow position on anionic lipid bilayers as reported for ProTx-II and ProTx-I (38,39). Although SgTx-I and HaTx-I were previously shown to insert into lipid membranes up to a distance of ϳ9 Å from the center of the bilayer, these studies employed model membranes containing 50% anionic PG-phospholipids, and model membranes were prepared in solutions without NaCl (9,48). Some GMTs like VsTx-I modulate the function of bacterial voltagegated potassium channels (3,13,49); therefore, the use of PGphospholipids can be justified in the studies of the trimolecular complex formed by these peptides. However, PG-phospholipids are unlikely to be found on the outer leaflet of mammalian cell membranes (34) and therefore were not included in the current study. Taken together, the present work indicates that GMTs adsorb to the outer leaflet of anionic lipid bilayers.

Relationship between GMT-membrane binding and Na V channel inhibition
Potency of the panel of GMTs for Na V 1.7 appears not to be correlated to affinity for any model membrane. For example, GsMTx-IV and SgTx-I are both known poor inhibitors of Na V 1.7 but had different affinities for the model membranes. The other GMTs studied were mid-nanomolar inhibitors of Na V 1.7 ( Fig. 7) but had varied affinities for the model membranes. We previously showed that reducing the membrane affinity of ProTx-II results in analogues with weaker potency at Na V 1.7 (38). We also demonstrated that the membrane binding and potency of HwTx-IV can be increased without changing the pharmacophore of this particular GMT (24). Thus, although membrane binding is not essential for inhibition of Na V 1.7, modulating affinity of individual GMTs for lipid bilayers can be exploited to increase or decrease potency.
We next examined whether differences in membrane binding properties might be important in dictating Na V subtype selectivity. None of the GMT studied displayed exceptional selectivity for any Na V channel subtype ( Fig. 8 and Table 1), affirming previous reports on the promiscuity of most of these peptides (14,19,20). There was also no direct correlation between selectivity and membrane binding. However, GpTx-I, HwTx-IV, and HnTx-IV, three peptides with an overall weak affinity for model membranes in the current study (Fig. 4), had the best overall selectivity, as they achieved 100% current inhibition and strong potency only for Na V 1.2 and Na V 1.6 (Fig. 8). Conversely, ProTx-I, CcoTx-I, and CcoTx-II, three peptides that achieved 100% current inhibition for at least three Na V chan-

GMT amphipathicity and Na V inhibition
nel subtypes ( Fig. 8 and Table 1), all had overall higher affinity for model membranes compared with HnTx-IV, HwTx-IV, and GpTx-I (Fig. 5). Therefore, it appears that increasing membrane binding is unlikely to improve the Na V subtype selectivity of spider venom GMTs.

Deductions on the role of GMT amphipathicity in Na V inhibition
There was also no apparent correlation between GMT physicochemical properties and affinity for the lipid membranes studied (Fig. 9), and these results are in agreement with our previous work involving a smaller sample of GMTs (39). Because the global physicochemical properties of native GMTs are not correlated with their propensity to interact with cell membranes, and there is no correlation between membrane binding and the modulation of Na V channels, the combination of a hydrophobic patch and positively charged ring is a conserved feature more likely to promote simultaneous potency for several voltage-gated channels. The current consensus is that GMTs employ certain residues incorporated in the hydrophobic patch and positively charged ring to bind to conserved sequences consisting of hydrophobic and anionic amino acid residues on voltage-gated ion channels (Fig. 10) (23, 29, 46, 50 -52). This hypothesis is supported by two observations: (i) the overall potency of most GMTs is comparatively lower for Na V 1.4, Na V 1.5, or both (Fig. 8), and these channels have fewer anionic residues in the GMT binding region of their voltagesensor domains (Fig. 10); (ii) SgTx-I and GsMTx-IV, which either have a different mechanism of action or do not potently inhibit Na V s, respectively (Fig. 8), have fewer positively charged residues in their C-terminal regions (Fig. 1B). The C-terminal region contains residues that contribute to the putative phar-macophores of some GMTs that inhibit Na V channels, including HnTx-IV, HwTx-IV, and GpTx-I for Na V 1.7 (23,29,53).

Conclusions and summary
This study revealed that some GMTs adsorb via electrostatic interactions to the outer leaflet of anionic lipid membranes and that the amphipathic surface profile of the peptides is most probably an adaptive strategy that enables simultaneous inhibition of several voltage-gated ion channels (Fig. 10), in a multipronged approach to immobilize insect prey (54). From a pharmacological perspective, application of the trimolecular complex to rational drug design cannot be based on broad generalizations about GMTs, but each GMT will require focused studies to explore the possibilities that membrane interactions present.
Most studies on the trimolecular complex have thus far been limited by distinct studies of GMT-model membrane and GMT-ion channel interactions. However, as novel systems continue to be developed, including the use of nanodiscs containing the transmembrane voltage-gated ion channels in combination with rationally chosen model membranes, techniques such as cryo-EM and NMR might be able to provide better insights into the tripartite interactions occurring between GMTs, lipids, and the voltage-gated ion channels (13,15,55).
Optimal conditions for oxidative folding of GMTs were chosen from one of several buffer formulations (Table 1). In each case, peptides were added into the oxidation buffer dropwise to achieve a final peptide concentration of 0.1 mg/ml and oxidized at 25°C unless otherwise stated. CcoTx-I, CcoTx-II, GpTx-I, GsMTx-IV, and PaurTx-III were dissolved in 50% (v/v) ACN and added to a buffer containing 7.5% (v/v) ACN and 0.1 M Tris, 0.81 mM GSH, and 0.81 mM GSSG at pH 7.7. Reactions were quenched by lowering the solution to pH 4 after 16 h using 2 M acetic acid (method A) (23). GsMTx-IV was also successfully oxidized by dissolving the peptide in 50% (v/v) ACN and adding it to 0.1 M Tris, 10 mM GSH, and 1 mM GSSG at pH 7.8. The reaction was quenched by lowering the solution to pH 2 after Figure 10. The hydrophobic patch and surrounding cationic ring in spider venom GMTs promotes Na V channel promiscuity. HnTx-IV (PDB code 1NIY) is shown as a representative GMT with hydrophobic patch (green), positively charged residues (blue), and anionic amino acid residues (red). A cartoon representation of a Na V channel is shown (adapted from Ref. 24), with the putative GMT target sequences of Na V 1.2, Na V 1.4, Na V 1.5, Na V 1.6, and Na V 1.7 also shown to highlight the conserved hydrophobic amino acid residues (boldface yellow type with cross below) and anionic amino acid residues (boldface red type with asterisk below).
Folded peptides were purified via RP-HPLC using a preparatory C18 column (Agilent, 250 ϫ 30 mm, 100 Å; Agilent Technologies) with a 1%/min gradient of 10 -70% solvent B at a flow rate of 8 ml/min and further using a semipreparatory C18 column (Phenomenex Gemini, 250 ϫ 10 mm, 5 m) at 3 ml/min on a 0.5%/min gradient of 15-45% solvent B. Relevant fractions were pooled, lyophilized, and stored at Ϫ20°C. Folding of the peptides was confirmed using NMR spectroscopy.
Additional NMR spectra acquired to determine the solution structures of CcoTx-I and CcoTx-II included natural abundance 2D 1 H-15 N HSQC and 1 H-13 C HSQC (samples in 90% (v/v) H 2 O and 10% (v/v) D 2 O) and 1 H-1 H, TOCSY, NOESY, and E.COSY (samples in 100% (v/v) D 2 O). A series of TOCSY spectra was obtained in 5°C temperature increments (from 10 to 35°C) to obtain amide proton temperature coefficients. The structures of the two peptides were calculated from the NMR data using methods described previously (24,33). The structure of PaurTx-III (PDB code 5WE3) has recently been published (33).

Calculations of physicochemical properties
The SASA of each peptide was calculated with GETAREA (58), using a 1.4-Å water probe to compare polar and apolar surface areas of the peptides. Propka (Jensengroup) (59) was used to assign protonation states to the side chains at pH 7.4 for net charge calculations. Hydrophobicity was assumed to be directly proportional to retention time of the peptides when eluted using a 1%/min gradient of 0 -60% solvent B at a flow rate of 0.3 ml/min on analytical RP-HPLC using a C18 column (Grace Vydac, 150 ϫ 2.1 mm, 5 m).

Preparation of lipid vesicles
Lipids used included CHOL (Sigma-Aldrich), PI (soy extract), POPC, POPS, C1P, and SM (Avanti Polar Lipids, Alabaster, AL). Molar ratio mixtures of the lipids were prepared in chloroform, the solvent was evaporated under a stream of N 2 , and then lipids were dried overnight in vacuo. Lipid films were hydrated with buffer and lipid suspensions subjected to eight freeze-thaw cycles. Liposomes were then sized by extrusion through a polycarbonate filter (50 nm to produce small unilamellar vesicles (SUVs) or 100 nm for LUVs).

GMT-membrane interactions
SPR (Biacore 3000, GE Healthcare Life Sciences) was used to examine the binding of GMTs to model membranes made from molar ratio mixtures of POPC, POPC/POPS (4:1), POPC/PI (4:1), POPC/SM/CHOL (2.7:4:3.3), and POPC/C1P/CHOL (2.7:4:3.3). Unless otherwise stated, SUVs were prepared in HEPES buffer solution (10 mM HEPES, 150 mM NaCl, pH 7.4) and deposited on L1 sensor chips (2 l/min, 2600 s). Peptides were injected at various concentrations (0 -64 M) over the lipid surface at a rate of 5 l/min for 180 s; the dissociation phase was followed for 600 s. The effect of ionic strength on GMT affinity for lipid membranes was examined by comparing peptide binding to POPC/POPS (4:1) bilayers in 10 mM HEPES, pH 7.4, at varying concentrations of NaCl (50, 150, or 300 mM). The same buffer was used to prepare SUVs and peptide samples and as running buffer in SPR. All experiments were conducted at 25°C. Binding affinity was compared by calculating the peptide/lipid ratio near the end of peptide injection and close to equilibrium (t ϭ 170 s). Data were corrected for buffer contribution and normalized using the assumption that 1 response unit ϭ 1 pg/mm 2 of lipid deposited (or peptide bound) (61). Concentration-response curves were fitted with a standard Hill equation using Prism version 7 (GraphPad Software Inc., La Jolla, CA). This is the first time we have reported on the current data for peptide-lipid affinity for CcoTx-I, CcoTx-II, GpTx-I, GsMTx-IV, HnTx-IV, PaurTx-3, and SgTx-I (Fig. 4); however, the data for peptide-lipid affinity for HwTx-IV as shown for POPC, POPC/POPS, POPC/SM/CHOL, and POPC/C1P/CHOL are in agreement with our recent study on HwTx-IV ( Fig. 4 and Table S3) (24).

GMT amphipathicity and Na V inhibition Fluorescence spectroscopy
Trp fluorescence emission spectra ( excitation ϭ 280 nm) of each GMT (25 M, or 12.5 M for PaurTx-III because the fluorescence emission spectrum for this GMT was saturated at 12.5 M) were acquired upon titration with LUV suspensions (0 -4 mM) to examine whether the environment surrounding the peptide Trp residue changes upon lipid titration (62). Data were corrected for dilution and light dispersion upon titration with vesicle suspension. The overall area of the emission spectra was used to compare quantum yield, and spectra were normalized to maximum fluorescence emission wavelength to check for presence of a blue shift. The fluorescence spectroscopy experiments for the eight peptides are represented here for the first time (Fig. 6), and the data for HwTx-IV are in agreement with our recently published work (24). Previously published spectra for [E5K,E8K]MfVIA are included as a positive control (45).

Whole-cell patch-clamp electrophysiology on HEK cells
The inhibitory potency of GMTs on hNa V 1.7 stably expressed in HEK293 cells was examined using an automated QPatch-16 electrophysiology platform (Sophion Bioscience, Ballerup, Denmark) at room temperature (ϳ25°C). The intracellular solution comprised 140 mM CsF, 1 mM EGTA, 5 mM CsOH, 10 mM HEPES, and 10 mM NaCl, pH 7.4, with CsOH (320 mosM). The extracellular solution comprised 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES, 4 mM KCl, 145 mM NaCl, pH 7.4, with NaOH (305 mosM). Solutions were filtered using a 0.22-m membrane filter (Merck Millipore, Cork, Ireland). Before recordings, cells were detached from culture flasks with Detachin (Genlantis, San Diego, CA). Na ϩ currents were acquired at 25 kHz and filtered at 4 kHz. Cells with Ͻ1 nA peak current were excluded. To determine IC 50 values, Na V 1.7 cells were held at Ϫ90 mV in the presence of varying concentrations of GMT and then stepped to Ϫ120 mV for 200 ms, followed by a 20-ms test depolarization to 0 mV. Na ϩ currents were normalized after leak subtraction. Offline analysis was performed using Microsoft Excel and GraphPad Prism version 7. Data are presented as mean Ϯ S.E. of at least three independent experiments.

Two-electrode voltage-clamp electrophysiology
Harvesting of stage V-VI oocytes from ovarian lobes of Xenopus laevis and the subsequent expression of Na V 1.2, Na V 1.4, Na V 1.5, and Na V 1.6 channels were performed as described (63). After injection with cRNA (50 nl of 1 ng/nl) using a microinjector (Drummond Scientific, Broomall, PA), oocytes were incubated in 96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 2 mM MgCl 2 , and 5 mM HEPES, pH 7.4, supplemented with 50 mg/liter gentamycin sulfate. Whole-cell currents were recorded after 1-4-day incubation periods at room temperature using twoelectrode voltage-clamp electrophysiology recordings using a Geneclamp 500 amplifier (Molecular Devices, Downingtown, PA) together with pClamp data acquisition software (Axon Instruments, Union City, CA). The bath solution contained 96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 2 mM MgCl 2 , and 5 mM HEPES, pH 7.4. Voltage and current electrodes were filled with 3 M KCl (resistance 0.8 -1.5 megaohms) (63). A four-pole lowpass Bessel filter was used to filter currents at 1 kHz with sampling at 20 kHz. A ϪP/4 protocol was used for leak subtraction. Currents for Na V channels were elicited from the oocytes using 30-ms step depolarizations from Ϫ90 mV to 70 mV in 5-mV increments. Current traces were obtained by 50-ms depolarizations to V max , the voltage at maximal current in control settings. Modulation of current by GMTs was examined by normalizing the data to peak current amplitude (I 30 ms /I max ), and concentration-response curves were used to obtain EC 50 values as before (63). Data are presented as mean Ϯ S.E. of at least three independent experiments. Mature female X. laevis were purchased from Nasco (Fort Atkinson, Wisconsin) and were housed in the Aquatic Facility (KU Leuven) in compliance with the regulations of the European Union concerning the welfare of laboratory animals as declared in Directive 2010/63/EU. The use of X. laevis was approved by the Animal Ethics Committee of KU Leuven (license number LA1210239, project number P038/2017).