Manipulation of a spider peptide toxin alters its affinity for lipid bilayers and potency and selectivity for voltage-gated sodium channel subtype 1.7

Huwentoxin-IV (HwTx-IV) is a gating modifier peptide toxin from spiders that has weak affinity for the lipid bilayer. As some gating modifier toxins have affinity for model lipid bilayers, a tripartite relationship among gating modifier toxins, voltage-gated ion channels, and the lipid membrane surrounding the channels has been proposed. We previously designed an HwTx-IV analogue (gHwTx-IV) with reduced negative charge and increased hydrophobic surface profile, which displays increased lipid bilayer affinity and in vitro activity at the voltage-gated sodium channel subtype 1.7 (NaV1.7), a channel targeted in pain management. Here, we show that replacements of the positively-charged residues that contribute to the activity of the peptide can improve gHwTx-IV's potency and selectivity for NaV1.7. Using HwTx-IV, gHwTx-IV, [R26A]gHwTx-IV, [K27A]gHwTx-IV, and [R29A]gHwTx-IV variants, we examined their potency and selectivity at human NaV1.7 and their affinity for the lipid bilayer. [R26A]gHwTx-IV consistently displayed the most improved potency and selectivity for NaV1.7, examined alongside off-target NaVs, compared with HwTx-IV and gHwTx-IV. The lipid affinity of each of the three novel analogues was weaker than that of gHwTx-IV, but stronger than that of HwTx-IV, suggesting a possible relationship between in vitro potency at NaV1.7 and affinity for lipid bilayers. In a murine NaV1.7 engagement model, [R26A]gHwTx-IV exhibited an efficacy comparable with that of native HwTx-IV. In summary, this study reports the development of an HwTx-IV analogue with improved in vitro selectivity for the pain target NaV1.7 and with an in vivo efficacy similar to that of native HwTx-IV.

Voltage-gated sodium channels (Na V s) 3 are involved in almost all aspects of human physiology, in particular in initiation and propagation of neurotransmission. To date, nine Na V subtypes (Na V 1.1-1.9) have been described based on their distribution and physiological function (1), with loss-and gainof-function mutations in individual subtypes resulting in a number of pathophysiological conditions, including pain (2, 3), epilepsy (4,5), and neuromuscular (6,7) and cardiac disorders (8). More specifically, Na V 1.7-1.9 are being pursued as targets for pain therapeutics because of their involvement in pain-related pathologies, including neuropathic pain (3,(9)(10)(11), mechanical pain associated with inflammatory bowel disease (12), diabetic neuropathy (13)(14)(15), and inflammation (16,17). The Na V architecture is typified by four domains (domains I-IV), connected via intracellular loops of various lengths, with each domain containing six segments (S) such that S1-S4 form the voltage sensor domains, and S5 and S6 form the pore domain ( Fig. 1A) (18). Because of high-sequence homology between subtypes Na V 1.1-1.7 (Na V 1.8 and Na V 1.9 display lowsequence homology across known binding sites) (Fig. 1C) (19), only highly-specific inhibitors would be of interest in order to avoid unwanted off-target effects.
In the search for novel modulators of these channels, gating modifier toxins (GMTs) discovered in spider venom have proven to be useful pharmacological probes and drug leads for conditions mediated by Na V function due to their high specificity and selectivity for the therapeutically-relevant sodium channel subtype Na V 1.7 (20 -25). Most spider GMTs are disulfide-rich and fold to adopt a knottin motif consisting of disulfide bridges formed between Cys I-IV, Cys II-V, and Cys III-VI (26,27). GMTs also share a hydrophobic patch and charged ring on their surface profiles (Fig. 1B), reminiscent of membrane-interacting peptides (28 -31).
As some GMTs bind to lipid bilayers (28 -30), the concept of a three-way interaction between GMTs, voltage-gated ion channels, and the lipid membrane has arisen (28,30,(32)(33)(34)(35). Using protoxin-II (ProTx-II), a GMT with affinity for lipid membranes (36,37), we recently showed that reducing the lipid binding of this particular GMT reduces its potency at Na V 1.7 (37). However, huwentoxin-IV (HwTx-IV), a GMT found in the venom of the Chinese bird spider Haplopelma schmidti that acts selectively on tetrodotoxin-sensitive sodium channels, is known to have weak affinity for lipid membranes (28,34,38). We recently designed a HwTx-IV analogue, gHwTx-IV ([E1G,E4G,F6W,Y33W]HwTx-IV), displaying increased lipid affinity and an increase in potency at Na V 1.7 (34). These studies contributed to the perspective that the lipids surrounding Na V s and other voltage-gated ion channels can attract, concentrate, and manipulate the binding of GMTs near their active sites at the channels for improved potency (33, 34, 37, 39 -42).
A largely-unanswered question is how increased GMT-lipid affinity affects selectivity for Na V 1.7. GMT selectivity for Na V 1.7 is a desirable feature because of the reported toxicity of some of these peptides in vivo, probably associated with offtarget effects (43)(44)(45). A difficulty with improving the selectivity of HwTx-IV, and similar GMTs, is that these peptides are attracted to a sequence of highly-conserved anionic amino acid residues (Glu-811, Asp-816, and Glu-818 (human Na V 1.7 primary accession number Q15858)) located on the extracellular loop between S3 and S4 of domain II (Fig. 1C) (46,47).
Previous mutagenesis studies have deduced that the positively-charged amino acids on loop 4 of HwTx-IV are of particular importance for activity ( Fig. 1B) (24,48,49). Deng et al. (49) noted that among other residues on loop 4, the R29A mutant lost ϳ260-fold of activity at tetrodotoxin-sensitive Na V s, including Na V 1.7 (examined on dorsal root ganglion cells) (49). The same study also hypothesized that Lys-27 played an important role in the bioactivity of HwTx-IV (49). Additional studies by Minassian et al. (48) showed that R26A and K27A mutations of HwTx-IV resulted in some loss of potency at Na V 1.7, accompanied by larger reductions in potency at Na V 1.2. The R29A mutation resulted in no change in potency at Na V 1.7, but a considerable reduction in potency at Na V 1.2 (these studies were conducted on Na V s expressed in HEK293 cells) (48). The main difference between Na V 1.7 and Na V 1.2 at the putative HwTx-IV-binding site on S3 and S4 of DII is that Na V 1.2 only has two anionic residues compared with three on Na V 1.7 in the putative GMT-binding site (Na V 1.2 contains Asn-816 in place of Asp-816) (Fig. 1C). Therefore, the reduction in potency for the Arg-29 analogue could be due to weakened electrostatic interactions between HwTx-IV and Na V 1.2 (46). Recent studies by Tzakoniati et al. (50) and Neff et al. (51) propose that interactions between anionic residues on both the S1 and S2 loop and S3 and S4 loop and cationic residues Arg-26, Arg-29, and Figure 1. Voltage-gated sodium channels and spider gating modifier toxins. A, Na V domains I-IV and transmembrane segments S1-S6 are shown. Extracellular loops between S1 and S2 (green), S3 and S4 (red), and S5 and S6 (black) are also shown. Figure was adapted from Agwa et al. (33). B, gHwTx-IV (PDB code 5TLR) is shown as a representative GMT with a surface profile highlighting residues forming the hydrophobic patch (green) and the positively-charged ring (blue). A ribbon representation of gHwTx-IV is also shown with cystines forming the inhibitor cystine knot motif labeled in roman numerals. The loops formed by the disulfide bridges (yellow) are also highlighted as follows: loop 1, black; loop 2, purple; loop 3, gray; and loop 4, blue. The positively-charged residues (blue) on loop 4, which are involved in activity of gHwTx-IV and HwTx-IV, are shown. C, alignment of Na V 1.1-Na V 1.9 at the extracellular S3 and S4 loop of domain II is also shown. Red amino acids are the residues where the GMTs putatively bind (46,48,49). Black shows equivalent positions on Na V s lacking the conserved anionic residues.

Manipulating binding of a spider toxin to sodium channels
Lys-27 are important in HwTx-IV binding (50,51), although there is some disagreement on the specific interactions across existing models (48,50,51). Based on these observations, we hypothesized that manipulation of the basic residues on loop 4 of HwTx-IV may aid in improving selectivity by modulating salt-bridge interactions between basic amino acids on loop 4 of the GMT and anionic amino acids in the conserved S3 and S4 extracellular loop of DII on off-target Na V s.
We used gHwTx-IV (24, 34) as a starting molecule to capitalize on the increased potency of this molecule at Na V 1.7, which is accompanied by stronger affinity for the lipid bilayer compared with HwTx-IV. We previously postulated that the lipid bilayer can attract and concentrate gHwTx-IV near its binding site at Na V 1.7 to increase its apparent potency (34). Even though Arg-26, Lys-27, and Arg-29 mutations on HwTx-IV did not always display favorable potency at Na V 1.7 (48,49), we were interested in examining whether point mutations to Arg-26, Lys-27, and Arg-29 of gHwTx-IV would improve selectivity while maintaining potency at Na V 1.7 through interactions with the lipid bilayer. To this end, we synthetically produced HwTx-IV, gHwTx-IV, [R26A]gHwTx-IV, [K27A]gHwTx-IV, and [R29A]gHwTx-IV (Fig. 1D). We ensured correct folding of the analogues using NMR, examined lipid binding of the peptides using surface plasmon resonance (SPR), and the activity of all peptides on human Na V 1.7 was assessed using automated patch-clamp electrophysiology. In addition, we were specifically interested in examining selectivity for Na V 1.7 over off-target Na V s that possess high-sequence homology to Na V 1.7 at the putative binding site for HwTx-IV; therefore, peptides were tested on human Na V 1.1, Na V 1.2, Na V 1.3, Na V 1.4, Na V 1.5, and Na V 1.6 for off-target activity (Fig.  1C). To examine the ability of HwTx-IV and the analogues to inhibit activity of the channels, the most potent analogue at Na V 1.7 was determined using automated patch-clamp electrophysiology, and its potency at Na V 1.1-1.7 was subsequently determined. The Na V 1.1-1.7 selectivity profiles for HwTx-IV and all analogues were determined using a membrane potential assay on FLIPR TETRA (52). Taken together, this body of work demonstrates that for gHwTx-IV, mutations to amino acids involved in activity can improve selectivity for Na V 1.7 and that the same modifications decrease affinity for lipid bilayers. The most potent analogue, [R26A]gHwTx-IV, was further tested for on-target activity in vivo using a mouse model of Na V 1.7-mediated nociception, and it showed improved efficacy compared with gHwTx-IV and similar efficacy to HwTx-IV.
[R29A]gHwTx-IV was synthesized with S-trityl (Trt) protecting groups on Cys III and Cys VI, acetamidomethyl (Acm) protecting groups on Cys II and Cys V, and 4-methylbenzyl (4-MeBzl) protecting groups on Cys I and Cys IV. A slow-fastslow kinetic approach was used to oxidize one disulfide bridge at a time (Fig. 2). DMSO was chosen for the slow oxidation steps (step 1, Cys III-Cys VI (85% yield), and step 3, Cys I-Cys IV (after cleavage of 4-MeBzl using hydrogen fluoride (20% yield)) because it offered a low-risk approach with no observed side reactions. I 2 was used for the removal of Acm protecting groups from Cys II and Cys V (step 2) and the subsequent oxidation of the pair. We found that a 10-min reaction using 20 molar eq of I 2 in 15% (v/v) methanol (MeOH), 42.5% (v/v) acetic acid (AcOH), and 42.5% (v/v) H 2 O was optimal to maximize the oxidized product (30% yield) and to minimize side reactions between I 2 and the indole groups on the Trp residues in the peptide.
One-dimensional (1D) 1 H NMR analysis of the peptides confirmed that they were folded (Fig. S1). Secondary chemical shift analysis obtained from the sequential assignment of two-dimensional (2D) TOCSY and NOESY spectra showed that the overall structure of the peptides was maintained (Fig. 3). Free cysteines Cys III and Cys VI were oxidized slowly in DMSO, followed by a fast removal of Acm protecting groups with subsequent oxidation of Cys II and Cys V using I 2 . After HF cleavage of 4-MeBzl, Cys I and Cys IV were oxidized slowly in DMSO to give the final product with correct disulfide connectivity. MALDI-TOF was used to ascertain that the correct product was formed at each step, and RP-HPLC analytical traces show a leftward shift in retention time after each oxidation step.

Manipulating binding of a spider toxin to sodium channels
Expected local differences in the backbones of the peptides are observed between gHwTx-IV and HwTx-IV at the E1G, E4G, F6W, and Y33W mutations as before (34), and between gHwTx-IV and the three analogues at R26A, K27A, and R29A (Fig. 3).

Affinity of GMTs for lipid bilayers
SPR was used to examine the affinity of the GMTs for lipid bilayers by measuring the amount of peptide bound to lipid (peptide/lipid (P/L mol/mol)) using zwitterionic 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) and anionic POPC/POPS (4:1 molar ratio) model membranes. The rationale behind the choice of lipids for these studies has previously been discussed (28,33,34,37,53). Briefly, phospholipids containing phosphatidylcholine headgroups are abundant on the outer leaflet of mammalian cells, and phospholipids containing phosphatidylserine headgroups mimic the anionic environment surrounding membrane-embedded voltage-gated sodium channels (54).
As before ( HwTx-IV. We previously described the membrane-binding face of gHwTx-IV to include Gly-1, Gly-4, Trp-6, and Trp-33 among other residues on the hydrophobic patch (34). The results herein suggest that the positively-charged amino acid residues on loop 4 of the GMT also contribute to affinity for lipid bilayers and that substitution of these residues for alanine residues on gHwTx-IV reduces their affinity for lipid bilayers.

Potency and selectivity of the GMTs
Potency at human Na V 1.7 was improved by each of the mutations compared with HwTx-IV, as determined by automated patch-clamp electrophysiology ( Fig. 5A and Fig. S3A). As before (34), gHwTx-IV was ϳ4-fold more potent than HwTx-IV (HwTx-IV IC 50
Selectivity of [R26A]gHwTx-IV for human Na V 1.7 over human Na V 1.1-1.6 was then determined by automated patchclamp electrophysiology ( Fig. 5B and Fig. S3B). Although Na V 1.8 and Na V 1.9 are both validated pain targets (1,11,16,17), these subtypes were not included in this study as they have low sequence homology across the proposed binding site compared with Na V 1.7 (19). Furthermore, Na V 1.9 is notoriously challenging to express and is only available in a handful of laboratories in the world (55,56).
[R26A]gHwTx-IV showed 11and 13-fold selectivity for Na V 1.7 compared with Na V 1.6 and Na V 1.1, respectively. Selectivity was 27-fold greater at Na V 1.2 and 64-fold greater at Na V 1.3. The peptide showed no activity at Na V 1.4 or Na V 1.5 up to 300 nM; therefore, selectivity at both subtypes is at least 176-fold ( Fig. 5B and Table 2). This results in Time constants of block () were determined using automated patch-clamp electrophysiology for HwTx-IV, gHwTx-IV, and the most potent analogue at Na V 1.7, [R26A]gHwTx-IV (Fig. 5C). Peptides at their respective IC 90 concentrations (HwTx-IV, 100 nM; gHwTx-IV, 20 nM; [R26A]gHwTx-IV, 10 nM) were added to cells to achieve near-complete block, and the decrease in current was measured over a period of 20 min, during which all peptides reached a steady state. It is necessary to test each peptide at its IC 90 because values are concentration- values were calculated for each peptide by fitting to a one-phase decay. A subsequent stepwise wash-off was conducted for 30 min, and the percentage of inhibition remaining at the end of the wash-off protocol is shown in a bar graph. D, time course of the binding of the peptides was conducted as in C, but all peptides were measured at 100 nM. All values expressed as mean Ϯ S.E. (n Ն 3). E, representative Na V 1.7 current trace of a single cell before and after addition of 10 and 100 nM HwTx-IV.

Table 1
Potency of HwTx-IV, gHwTx-IV, and analogues on Na V 1.7 Data were measured using automated patch-clamp electrophysiology on HEK293 cells, n Ն 3.
Manipulating binding of a spider toxin to sodium channels  5C). After continuous wash-off for 30 min, gHwTx-IV showed the highest percentage of sustained inhibition (68.1% Ϯ 7.6%). This is similar to the remaining inhibition of HwTx-IV (52.1% Ϯ 13.3%), but [R26A]gHwTx-IV had a significantly lower percentage of remaining inhibition (28.7% Ϯ 8.8%). This suggests that the mutations that make up gHwTx-IV do not affect the peptide's dissociation from the channel compared with HwTx-IV, whereas the R26A mutation has a significant effect on peptide dissociation.
IC 50 values of the same three peptides were measured with holding potentials of both Ϫ90 mV and Ϫ55 mV to determine whether binding is state-or voltage-dependent, showing no significant difference (Fig. S4). This suggests that although there is strong evidence that HwTx-IV binds preferentially to the closed conformation of voltage-sensor DII, the modified toxins could experience some alterations in binding to the activated voltage-sensor, and unlike for some other GMTs (57,58), the activity of the peptides tested in this study appear not to be influenced by channel state or membrane potential.
A complete selectivity screen was performed for each peptide using the FLIPR TETRA membrane potential assay (Table 3, Fig.  6). The addition of veratridine is known to underestimate the potency of GMTs compared with electrophysiology measurements, as reported previously (59). Although this assay gives a median 15-fold underestimation of the potencies of GMTs compared with patch-clamp electrophysiology assays (as observed in this study) (Figs. S5 and S6 and Table S1), it is a lower-cost, high-throughput assay that can reliably be used to measure relative potencies and therefore selectivity. Overall, the observed trend of selectivity for Na V 1.7 using both assays is in good agreement.
[R26A]gHwTx-IV and [R29A]gHwTx-IV had a 92-fold (overestimated compared with 64-fold from patch-clamp electrophysiology) and 273fold selectivity (respectively) for Na V 1.7 compared with Na V 1.3 (Tables 2 and 3). In this case, the selectivity of [K27A]gHwTx-IV was 11-fold, which was less than the selectivity of gHwTx-IV (83-fold). HwTx-IV did not show inhibitory activity at Na V 1.3 up to 10 M in our hands, making it at least 41-fold more selective for Na V 1.7 over Na V 1.3 (Table  3), although an IC 50 value of 338 nM was previously reported at rat Na V 1.3 (46). In comparison with Na V 1.7 to Na V 1.4, [R26A]gHwTx-IV showed the greatest improvement in selectivity for Na V 1.7, showing no inhibitory activity on Na V 1.4 up to 10 M, thus being at least 392-fold more selective (or at least 176-fold, from patch-clamp electrophysiology) (Tables 2 and 3).
These results show that a combination of mutations to HwTx-IV, including gHwTx-IV (E1G, E4G, F6W, and Y33W), and point mutations at the positively-charged residues in loop 4 lead to modified potency and selectivity of this GMT at Na V 1.7.
Taking these results together with the GMT-lipid affinity studies, removal of positively-charged residues that contribute to the activity of gHwTx-IV reduces, but does not eliminate, affinity for the lipid bilayer. These mutations also improve the overall selectivity for Na V 1.7 with respect to off-target Na V s containing a string of anionic residues at HwTx-IV's putative binding site (Fig. 1, C and D).

Manipulating binding of a spider toxin to sodium channels Activity of HwTx-IV, gHwTx-IV, and [R26A]gHwTx-IV in murine model of Na V 1.7-mediated nociception
We were interested in examining whether the most selective peptide was efficacious in vivo; therefore, the effects of HwTx-IV, gHwTx-IV, and [R26A]gHwTx-IV were examined in a mouse model of Na V 1.7-mediated nociception after intraplantar administration of OD1 (300 nM), an ␣-scorpion toxin that selectively impairs inactivation and enhances current from Na V 1.7 (Fig. 7) (52,60). There is 100% homology between human and mouse Na V 1.7 across the S3 and S4 helix sequence in DII, which includes the S3 and S4 extracellular loop previously described as the GMT-binding site, suggesting that a mouse model provides a suitable comparison for human Na V 1.7 (UniProt ID Q15858 hNa V 1.7 and Q62205 mNa V 1.7) (Fig. S7). HwTx-IV significantly reduced nocifensive behaviors (cumulative nocifensive behavior count 132.0 Ϯ 44.3, p Ͻ 0.01) compared with the control (OD1; 518.3 Ϯ 59.1). There was a significant reduction in nocifensive behaviors for [R26A]gHwTx-IV (182.7 Ϯ 10.0, p Ͻ 0.01), comparable with that of HwTx-IV, at the same dose (Fig. 7). There was a small but significant difference in nocifensive behavior for gHwTx-IV (308.0 Ϯ 28.2, p Ͻ 0.05), although this peptide showed a delayed inhibitory effect during the first 10 min post-injection, with a nocifensive behavior count comparable with the control (Fig.  7). After this delay period, nocifensive behaviors were reduced to a similar level to HwTx-IV and [R26A]gHwTx-IV for the remaining 20 min of the experiment.

Discussion
Na V 1.7 is a promising target for the development of novel therapeutics for the treatment of pain (19,61,62). Spider GMTs, which are potent Na V modulators, are being pursued as peptide drug leads for Na V 1.7 (63). Because of the proposed affinity of some GMTs for lipid bilayers surrounding Na V s (28,33,34,37,53,64), here we have used gHwTx-IV to examine how mutations to specific GMT residues known to interact with the channels (38,48) can also influence affinity for lipid bilayers and selectivity for Na V 1.7.
Point mutations to the positively-charged residues on loop 4 of gHwTx-IV reduce but do not eliminate affinity for model lipid bilayers (Fig. 4). This places [R26A]gHwTx-IV, [K27A]gHwTx-IV, and [R29A]gHwTx-IV in between gHwTx-IV and HwTx-IV with respect to affinity for neutral POPC and anionic POPC/POPS lipid bilayers. When affinity of HwTx-IV for model lipid membranes was first examined, this GMT showed no binding to lipid bilayers (34,38). We then engineered gHwTx-IV to increase affinity for model lipid membranes by reducing the anionic contribution to the overall surface charge of the peptide and by increasing the size of the hydrophobic patch (34). At the time, we proposed that the peptide may bind to the lipid bilayer using a membrane interaction face (Fig. 8A), consisting primarily of hydrophobic residues in a manner similar to ProTx-I and ProTx-II (34,37,53). The observed change in affinity for lipid bilayers of the three new analogues compared with gHwTx-IV suggests that Arg-26, Lys-27, and Arg-29 also contribute to the membrane-binding face residues important for affinity for the lipid bilayers, most probably through electrostatic interactions with anionic moieties on the lipid bilayer (Fig. 8A). In a study conducted by Henriques et al. (37), [E17K]ProTx-II showed increased affinity for POPC/ POPS bilayers compared with the WT toxin. They propose that the introduction of a cationic residue increases the toxin's affinity for the anionic phospholipid headgroups (37), which is in agreement with the results of this study. The opposite effect was observed by Nishizawa et al. (65), who found that a mutant of GsMTx4 with increased negative charge, [K15E]GsMTx4, showed an increase in POPC/1palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1Ј-rac-glycerol) affinity.
[R26A]gHwTx-IV, [K27A]gHwTx-IV, and [R29A]gHwTx-IV displayed improved activity at Na V 1.7 compared with activity of HwTx-IV at the same channel (Fig. 5). As before, gHwTx-IV was 4-fold more potent than HwTx-IV (34). [R29A]gHwTx-IV and [K27A]gHwTx-IV displayed similar potency compared with gHwTx-IV, and [R26A]gHwTx-IV was 21-fold more potent than HwTx-IV at Na V 1.7 ( Fig. 5 and Table 1). As mentioned, the basic residues on loop 4 of HwTx-IV have been proposed to form important interactions with anionic amino acids on the S3 and S4 loop of domain II on Na V s (46,48), as well as the anionic amino acids on the S1 and S2 loop (50,51 (24) further optimized HwTx-IV potency at Na V 1.7 by generating a [E1G,E4G,Y33W]HwTx-IV triple mutant, which was also investigated by Rahnama and co-workers (59) and shown to be 2-3-fold selective for Na V 1.7 over Na V 1.1, Na V 1.2, Na V 1.3, and Na V 1.6. Here, we have shown that Table 3 Fold selectivity of HwTx-IV, gHwTx-IV, and analogues determined by membrane potential assay Fold Na V 1.7 selectivity is IC 50 Na V 1.x/ IC 50 Na V 1.7. Potency at voltage-gated sodium channels was measured using FLIPR TETRA membrane potential assay on HEK293 cells expressing each respective channel where n Ն 3 experiments and three replicates per experiment.

Peptide
Na V 1. This could be because potential losses to affinity for the channel may be compensated for by increased lipid membrane affinity. The lipid bilayer may anchor and concentrate these gHwTx-IV analogues in the lipid membrane near the channel active site. It is possible that previous studies did not see the same effect because those mutations (R26A, K27A, and R29A) were on HwTx-IV, a peptide with less affinity for the lipid bilayer compared with gHwTx-IV. The specific interactions between the basic residues on loop 4 of HwTx-IV and the Na V 1.7 voltage-sensor domain remain unclear. Multiple binding models have been proposed using different methods, but no consensus has been reached (48,50,51). Additionally, because the mutations that make up gHwTx-IV and its analogues may alter the position of the peptide when binding, some interactions may differ. Although gHwTx-IV has an increased affinity for the lipid bilayer compared with HwTx-IV, gHwTx-IV did not show an improved selectivity profile compared with HwTx-IV, with a comparable lack of selectivity against Na V 1.2 and Na V 1.1 and a reduction of selectivity against Na V 1.3 and Na V 1.6 when compared with Na V 1.7 (Table 3). This lack of selectivity is most probably because gHwTx-IV is anchored in the lipid bilayer in a similar fashion for all the Na V channels, which is likely to be the same across the channel sub-types particularly in this study where all channels are studied in HEK293 cells. We propose that the improved selectivity profile for the new gHwTx-IV analogues comes from weakening some electrostatic interactions with the voltage-gated ion channels, combined with having the right amount of lipid affinity to anchor the peptides, contributing to an increase in potency at Na V 1.7.
We examined the in vivo efficacy of HwTx-IV, gHwTx-IV, and the most potent analogue, [R26A]gHwTx-IV, using a Na V 1.7 engagement model based on intraplantar administration of the Na V 1.7 agonist OD1 (52). HwTx-IV and [R26A]gHwTx-IV significantly decreased nocifensive behaviors compared with the control to a similar extent. Thus, the anti-nocifensive effects observed for [R26A]gHwTx-IV are comparable with those observed for other GMTs using the same model (21,52,59). Interestingly, gHwTx-IV had a delayed  Optimizing GMTs to access their targets in vivo remains an area for future work if these peptides are to be converted into efficacious drugs. Both HwTx-IV and the most selective analogues of ProTx-II remain toxic in animal models, resulting in death and possible histaminergic effects, respectively (43,45). Gonçalves and co-workers (44) have recently proposed that HwTx-IV's off-target activity at Na V 1.6 is responsible for neuromuscular toxicity (44); therefore, an area for future investigations could be to examine whether the selectivity and efficacy of [R26A]gHwTx-IV is accompanied by a loss of neuromuscular toxicity in vivo or whether systemic dosing leads to motor side effects.
Taking the in vivo and in vitro data together, we have shown here that it is possible to engineer selective modulators of Na V 1.7 while considering interactions with the lipid bilayers. The in vitro data give only a partial understanding of GMT behavior and further in vivo target engagement studies are therefore crucial to fully appreciate the biological activity of Na V 1.7 modulators.

Conclusion
Here, we have considered all three components of the trimolecular complex by investigating peptide-lipid bilayer as well as peptide-channel interactions in the quest to design potent and selective modulators of Na V 1.7. As seen for [R26A]gHwTx-IV, [K27A]gHwTx-IV, and [R29A]gHwTx-IV, peptides do not require maximal affinity for the lipid bilayer to display an increase in potency at Na V 1.7. Instead, for this suite of peptides, an intermediate affinity for lipid bilayers somewhere between gHwTx-IV and HwTx-IV is optimal for potency and selectivity. GMTs are probably adapted more for contact with the channel than for contact with the lipid bilayer (28); therefore, we propose that the tripartite relationship formed by gHwTx-IV and its analogues occurs such that the peptides are nestled between the extracellular loops formed by S3 and S4 and S1 and S2 and probably interact with the phospholipid headgroups in this region (Fig. 8B). We have previously discussed the possibility of this binding mode (33), and others have used molecular dynamics combined with channel mutagenesis studies on similar GMTs to demonstrate the same binding mode (48,66,67). X-ray crystallography has recently been used to support this GMT-binding mode, showing ProTx-II interacting with the membrane (42). Alternatively, the GMTs are first attracted to the anionic environment surrounding voltagegated ion channels via electrostatic interactions, then concentrated in the lipid bilayer, and oriented externally near the S3and S4-binding site on domain II of Na V 1.7 (33, 34, 37, 39 -42). In our hands, every spider GMT has shown a distinct set of behavioral patterns with respect to the lipid bilayer and voltagegated sodium channels (28); therefore, we believe it is not possible to make sweeping statements about GMT behavior. Nevertheless, as we have shown here with HwTx-IV and previously with ProTx-II (37), detailed studies on individual GMTs can produce novel insight and approaches to optimizing these peptides for use as potential drugs for pain and other Na V -dependent pathologies. We anticipate that new technologies, including cryogenic EM, will help us to overcome current limitations and allow future studies to focus on investigating the three components (GMTs, model lipid bilayers, and voltage-gated ion channels) simultaneously (42,68), providing a complete picture of the different interactions.

Peptide synthesis
Peptides were assembled using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry on a Rink amide resin (for C-terminal amidation) on a scale of 0.25 mmol/g using an auto-  (34) and an updated membrane-binding face (right) are proposed with residues that probably contribute to interactions with lipid bilayers shown. B, we propose that with a membrane-binding face that includes some of the residues involved in activity, gHwTx-IV probably interacts with the phospholipid heads found between extracellular loops formed by S1 and S2 and by S3 and S4. Figure is adapted from Agwa et al. (34).
For [R29A]gHwTx-IV (Fig. 2), the Trt-protecting groups on Cys III and Cys VI were removed during cleavage, and the disulfide bridge between these free cysteines was formed by incubating the peptide (0.1 mg/ml) in 30% (v/v) DMSO/H 2 O for 24 h. The peptide was purified using previously described semi-preparatory methods (28) and lyophilized prior to removal of Acm protecting groups from Cys II and Cys V with simultaneous oxidation of the free cysteines (1 mg/ml) using 20 molar eq of I 2 in 15% (v/v) MeOH, 42.5% (v/v) AcOH, and 42.5% (v/v) H 2 O. 0.1 M ascorbic acid was used to quench the reaction after 10 min, and the peptide was purified (28) and lyophilized. Finally, the 4-MeBzl groups were removed at 0°C for 1 h using hydrogen fluoride for cleavage and p-cresol as a scavenger. The peptide was extracted from the cleavage solution using diethyl ether and extracted from the diethyl ether using phase separation in 45% (v/v) ACN, 0.05% (v/v) TFA, in H 2 O, and the Cys I-Cys IV pair was oxidized over 24 h by diluting to 0.05 mg/ml peptide, 22.5% (v/v) ACN, and 33% (v/v) DMSO in H 2 O. Each step of peptide oxidation was monitored using MALDI-TOF MS and analytical RP-HPLC as described previously (28). Once oxidation was complete, the peptide was purified as before (28), lyophilized, and stored at Ϫ20°C.

Affinity of HwTx-IV and analogues for model lipid bilayers
As described previously (28,34,37,53,69), model lipid bilayers were prepared using POPC and POPS (Avanti Polar Lipids, Alabaster, AL). SPR (Biacore 3000, GE Healthcare) was used to examine the amount of peptide bound to zwitterionic POPC and anionic POPC/POPS (4:1 molar ratio) deposited on an L1 sensor chip. Peptides were then injected onto the lipid surface and followed for the duration of association and dissociation. Data at 170 s (near the end of association) were collected, corrected for buffer contribution, and normalized using the assumption that 1 response unit ϭ 1 pg/mm 2 of lipid deposited (or peptide injected). As described previously (34,73), data were fitted to a one-site specific binding curve, and maximum P/L (mol/mol) was extrapolated from the line-of-best-fit using GraphPad Prism, version 8.0.2 (GraphPad Software Inc, San Diego, CA).
Peptides were diluted in extracellular solution with 0.1% BSA. Assays were run with a holding potential of Ϫ90 mV pulsed to Ϫ20 mV for 50 ms at 0.05 Hz. Activity assays were Manipulating binding of a spider toxin to sodium channels performed with 5-min incubation periods after each peptide addition. To measure values, 100 nM HwTx-IV, 20 nM gHwTx-IV, and 10 nM [R26A]gHwTx-IV were added to cells and incubated for 20 min. Inhibition was quantified by measuring peak current, and all values were normalized to buffer control. Concentration-response curves were plotted with GraphPad Prism, version 8.0.2 (GraphPad Software Inc.), fitted with a four-parameter Hill equation with variable Hill coefficient. Time constants of block data were also plotted with GraphPad Prism, and values were calculated by fitting data to a one-phase decay. Data are presented as mean Ϯ S.E.

Examination of subtype selectivity using FLIPR
The FLIPR TETRA plate reader (Molecular Devices, Sunnyvale, CA) was used to examine activity of the peptides at Na V channels as described previously (52). In brief, freshly-dissociated cells were plated into 384-well clear-bottom black-walled imaging plates (Corning, NY) at a density of 10,000 -15,000 cells per well. After 48 h, the growth media were removed from the wells, and the cells were loaded with 20 l per well red membrane potential dye (Molecular Devices) diluted in physiological salt solution (PSS; composition (mM): NaCl 140, glucose 11.5, KCl 5.9, MgCl 2 1.4, Na 2 H 2 PO 4 1.2, NaHCO 3 5, CaCl 2 1.8, HEPES 10 (pH 7.4)) according to the manufacturer's instructions for 30 min at 37°C, 5% CO 2 . Peptides were diluted in PSS, 0.1% (w/v) BSA and added to the cells to achieve concentrations ranging from 10 M to 0.3 nM and incubated for 5 min. Na V channels were stimulated using 60 M veratridine, and changes to membrane potential were measured with a cooled CCD camera (excitation 515-545 nm, emission 565-625 nm) with reads taken every 1 s for 10 s before (baseline values) and for 300 s after addition of veratridine. PSS and 0.1% (w/v) BSA was used as a negative control. ScreenWorks 3.2.0.14 (Molecular Devices) was used to compute the area under the curve over 300 s, and the data were plotted and analyzed using GraphPad Prism, version 8.0.2 (GraphPad Software Inc.), to quantify inhibitory effects of the peptides. To calculate concentration-response curves, a four-parameter Hill equation with variable Hill slope was fitted to the data. All data are expressed as the mean Ϯ S.E. and are representative of at least three independent experiments (three wells per experiment). Potency of each peptide at off-targets is compared with potency of the specific peptide at Na V 1.7 (Table 3).

In vivo Na V 1.7 target engagement
Animal ethics approval was obtained from the Animal Ethics Committee of University of Queensland. All experiments were conducted in accordance with local and national regulations and the International Associations for the Study of Pain Guidelines for the Use of Animals in Research. Male C57BL/6J mice aged 6 -8 weeks (20 -25 g) were housed in 12-h light/dark cycles with access to food and water ad libitum. To assess the in vivo effect of peptide analogues, an OD1-induced model of Na V 1.7 target engagement was used as described previously (52).