Comprehensive engineering of the tarantula venom peptide huwentoxin-IV to inhibit the human voltage-gated sodium channel hNa_v1.7

Abstract

Pain is a significant public health burden in the United States, and current treatment approaches rely heavily on opioids, which often have limited efficacy and can lead to addiction. In humans, functional loss of the voltage-gated sodium channel Na_v1.7 leads to pain insensitivity without deficits in the central nervous system. Accordingly, discovery of a selective Na_v1.7 antagonist should provide an analgesic without abuse liability and an improved side-effect profile. Huwentoxin-IV, a component of tarantula venom, potently blocks sodium channels and is an attractive scaffold for engineering a Na_v1.7-selective molecule. To define the functional impact of alterations in huwentoxin-IV sequence, we produced a library of 373 point mutants and tested them for Na_v1.7 and Na_v1.2 activity. We then combined favorable individual changes to produce combinatorial mutants that showed further improvements in Na_v1.7 potency (E1N, E4D, Y33W, Q34S–Na_v1.7 pIC₅₀ = 8.1 ± 0.08) and increased selectivity over other Na_v isoforms (E1N, R26K, Q34S, G36I, Na_v1.7 pIC₅₀ = 7.2 ± 0.1, Na_v1.2 pIC₅₀ = 6.1 ± 0.18, Na_v1.3 pIC₅₀ = 6.4 ± 1.0), Na_v1.4 is inactive at 3 μm, and Na_v1.5 is inactive at 10 μm. We also substituted noncoded amino acids at select positions in huwentoxin-IV. Based on these results, we identify key determinants of huwentoxin's Na_v1.7 inhibition and propose a model for huwentoxin-IV's interaction with Na_v1.7. These findings uncover fundamental features of huwentoxin involved in Na_v1.7 blockade, provide a foundation for additional optimization of this molecule, and offer a basis for the development of a safe and effective analgesic.

Introduction

According to a Centers for Disease Control and Prevention analysis of the 2016 National Health Interview Survey an estimated 50 million people in the United States experience chronic pain, 19.6 million of which experience pain severe enough to limit their daily activities (1). Individuals experiencing such severe pain are more likely to have a poor health state and report exhaustion, depression, and anxiety. In addition, these persons have a greater likelihood of needing bed-disability days, accessing medical care, and visiting an emergency room compared with persons experiencing lower pain states (2). This not only diminishes an individual's quality of life and their ability to conduct daily activities, it represents a significant public health burden in the United States with a financial cost of more than 500 billion dollars (3, 4).

The World Health Organization endorses a three-step approach to the pharmacological management of chronic pain, recommending nonsteroidal anti-inflammatory drugs and other nonopioid analgesics for mild pain, the addition of weak opioids for mild to moderate pain, and reserving strong opioids for patients with moderate to severe pain (5). Use of opioids, however, is commonly accompanied by adverse effects such as nausea, constipation, and respiratory depression (6). With sustained use, efficacy can be limited by the development of tolerance and opioid-induced hyperalgesia (6, 7). In addition, prolonged opioid use increases the probability of misuse and addiction, behaviors that have contributed to the current epidemic of opioid abuse (8, 9). Thus, the development of efficacious, nonopioid analgesics could help mitigate this public health crisis and address a significant unmet medical need.

The voltage-gated sodium channel Na_v1.7 (VGSC,⁷ Na_v1.7) is preferentially expressed in human pain fibers and is critical for amplifying the sensory input they receive (10, 11). Mutations in SCN9A, the gene that codes for Na_v1.7, reveal its fundamental role in determining pain perception in humans. Na_v1.7 polymorphisms can cause recurrent episodes of severe pain (12–15), and nonsense mutations convey a complete insensitivity to pain without cognitive or autonomic defects (16). As such, selective blockers of Na_v1.7 could provide powerful analgesia with minimal side effects and provide an alternative to opioid analgesics.

Despite significant drug-discovery efforts to target Na_v1.7 with small molecules, no selective Na_v1.7 antagonist to date has moved past phase II clinical trials (17). Strategies focused on engineering large molecules have emerged as an alternative path for achieving selective Na_v1.7 block (18). A rich source of good starting points for drug discovery is the venom of animals (19). Of particular interest for Na_v1.7-focused drug discovery are peptides derived from spider venoms that block voltage-gated sodium channels (20, 21).

Huwentoxin-IV (HwTX-IV) is an inhibitory cysteine knot peptide derived from the venom of the tarantula Haplopelma schmidti and a broad inhibitor of tetrodotoxin-sensitive sodium channels (22). We previously identified key structural and functional features of HwTX-IV critical for its activity for both Na_v1.7 and Na_v1.2 (23, 24). Since then, variants of HwTX-IV that further elaborate its structure–activity relationship with Na_v1.7 have been published (25) as well as peptides that have been optimized for potency on Na_v1.7 and/or selectivity for Na_v1.7 over other sodium channel family members (26–28). Here, we present our comprehensive engineering efforts focused on the design of potent and Na_v1.7-selective variants of HwTX-IV, as well as functional studies describing the activities of these libraries on Na_v1.7. In addition, we present the activities of these peptides on Na_v1.2 to highlight their selectivity for Na_v1.7 over one of the most abundant sodium channel isoforms in the central nervous system (29). We also describe the broader selectivity of some of the peptides with the most favorable modifications by measuring peptide activity against other sodium channel members. These data are a road map for further optimization of HwTX-IV and ultimately may provide a basis for the discovery of a novel, peptide-based nonopiate analgesic.

Results

Recombinant production of HwTx-IV

As described previously, we produced recombinant huwentoxin (rHwTx-IV) in a mammalian expression system (23). This peptide differs from native HwTx-IV in three capacities. 1) Because this peptide was produced as a fusion protein to optimize expression levels, it retains a glycine and proline on its N terminus (remains of an HRV3C cleavage site). 2) This peptide contains two additional residues—a Gly at position 36 and a Lys at position 37 on its C terminus. 3) The C terminus is not amidated (Fig. 1A).

Figure 1.

Activity profiles of synthetic and recombinant HwTX-IV in FLIPR. A, basic structure and sequence comparison of synthetic and recombinantly produced HwTX-IV. B, concentration response of synthetic (red squares) and recombinant (blue circles) HwTX-IV versus Na_v1.7. C, concentration response of synthetic (red squares) and recombinant (blue circles) HwTX-IV versus Na_v1.2. Data mean %I ± S.E. from representative experiments.

This peptide was screened for activity against Na_v1.7 and Na_v1.2 in a FRET-based FLIPR assay. In this assay, the mean pIC₅₀ values for rHwTx-IV on Na_v1.7 and Na_v1.2 were 7.1 ± 0.07 and 6.6 ± 0.12, respectively. As reported previously, these values are right-shifted relative to potencies observed when testing synthetically produced, amidated huwentoxin (Peptides International) in this assay (Na_v1.7, 7.4 ± 0.04, and Na_v1.2, 7.4 ± 0.03, Fig. 1, B and C) (23).

Amino acid–scanning mutagenesis of HwTx-IV

To define the impact of alterations in HwTx-IV sequence on its ability to impair sodium channel gating, we systematically substituted amino acids (except cysteine and methionine) at noncysteine positions and tested them for activity in FLIPR on hNa_v1.7 and hNa_v1.2. Mean Na_v1.7 and Na_v1.2 pIC₅₀ values for 373 mutant peptides are presented in Figs. S1 and S2, respectively. To better understand general trends in HwTX-IV's SAR, we first interrogated the dataset to determine the mean impact of all mutations in each of HwTX-IVs loops (as well as the N and C termini) on Na_v1.7 and Na_v1.2 activity. Except for the N-terminal mutations in Na_v1.2, changes in HwTX-IV decreased activity on both Na_v1.2 and Na_v.17 with the most robust effects resulting from mutations in HwTX-IV's loop 4 and the C terminus (Fig. 2A).

Figure 2.

HwTX loop SAR, mean impact of amino acid property substitutions on HwTX-IV activity. A, mean change in pIC₅₀ (±S.D.) caused by all mutations generated in a given loop versus Na_v1.7 (left) and Na_v1.2 (right) compared with recombinant HwTX-IV (rH-IV). B–F, mean change in pIC₅₀ (±S.D.) caused by substitution of amino acids with common general properties into specific segments of HwTX-IV versus Na_v1.7 (left) and Na_v1.2 (right). *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001 versus recombinant HwTX-IV using ANOVA with Dunnett's post hoc test.

Next, we looked at the mean impact of the substitution of amino acids with similar general properties in each HwTX-IV segment (Fig. 2, B–F). This property-focused loop analysis showed differential effects on HwTX-IV's Na_v1.7 and Na_v1.2 activity that were not detected by the original loop analysis. Of note, introduction of basic and nonpolar amino acids on HwTX-IV's N terminus (position Glu-1) significantly increased potency on Na_v1.2 (Fig. 2B). In loop 2, substitution of acidic and basic functionalities significantly diminished Na_v1.7 activity (Fig. 2C). Introduction of acidic amino acids in loop 3 significantly decreased both Na_v1.7 and Na_v1.2 potency (Fig. 2D). Loop 4 was the least-tolerant segment, with acidic and uncharged polar substitutions decreasing both Na_v1.7 and Na_v1.2 potency and nonpolar insertions reducing Na_v1.7 potency (Fig. 2F). In the C terminus, acidic functionalities selectively decreased HwTX-IV Na_v1.7 potency, and amino acids with uncharged polar properties decreased potency on both isoforms (Fig. 2F). No significant changes in activity were seen with the property-based analysis in loop 1 (data not shown).

To further explore HwTX-IV's SAR, we also examined the mean change in pIC₅₀ caused by all mutations at each HwTX-IV amino acid position for both Na_v1.7 and Na_v1.2. In this analysis, no significant increases in mean potency for either Na_v isoform were observed at any position. Mutations at positions 5, 6, 11, 14, 25, 27, 28, 30, 32, 33, and 35 significantly decreased Na_v1.7 potency. Likewise, mutation of positions 6, 14, 26, 27, 30, and 32, 33, and 35 significantly decreased Na_v1.2 potency (Fig. 3, A and B).

Figure 3.

HwTX SAR, mean impact of amino acid substitutions on Na_v activity at individual positions. Mean change in pIC₅₀ (± S.D.) is caused by all mutations generated at a given position versus Na_v1.7 (A) and Na_v1.2 (B) compared with recombinant HwTX-IV. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001 versus recombinant HwTX-IV using ANOVA with Dunnett's post hoc test.

Both loop and position-focused analyses of the differential impact of mutations on Na_v1.7 and Na_v1.2 activity were also performed using a linear regression where the location of amino acid replacement was used as a predictor variable for estimating the change in Na_v1.2 pIC₅₀ values. The corresponding change in Na_v1.7 pIC₅₀ values that result from the replacement was also included as a predictor variable in the model. This allowed us to identify when the changes in Na_v1.2 pIC₅₀ values were greater than expected given the observed change in Na_v1.7 pIC₅₀ values that resulted from the amino acid replacement.

When applied to HwTX-IV's loops, the above analysis showed that alterations in the N terminus and loop 1 shifted huwentoxin's activity preference toward Na_v1.2 (Fig. 4A). These trends are also reflected in the position-based analysis as seen in of all noncysteine positions through Ser-12, which show significant shifts of huwentoxin's activity toward Na_v1.2. The remaining positions show either nominal impact or increased selectivity for Na_v1.7. The most robust of these was Arg-26, which improved selectivity for Na_v1.7 by more than 3-fold (Fig. 4B).

Figure 4.

HwTX SAR, impact of amino acid substitutions on Na_v selectivity at individual positions. Mean difference in change from recombinant HwTX-IV IC₅₀ between Na_v1.7 and Na_v1.2 (± S.D.) caused by all mutations in a given segment (A) or HwTX-IV position (B). ■, p < 0.1; *, p < 0.05; **, p < 0.01; ***, p < 0.001 versus recombinant HwTX-IV using a two sample t test on the regression coefficients obtained from a simple linear regression of the Na_v1.7 and Na_v1.2 ΔpIC₅₀ values.

Examination of the effect of individual mutations, however, revealed unique behaviors at each position that were not predicted by the more general analyses. These data allowed us to narrow our focus to a smaller subset of mutants that contain an attractive combination of two specific properties: potency toward Na_v1.7 and selectivity against other Na_v1.2. Specifically, we identified 17 point mutations that produced 3-fold or better improvements on HwTX's Na_v1.7 potency compared with recombinant HwTx-IV (Table 1). In addition, we found multiple point mutations that caused differential effects on Na_v1.7 and Na_v1.2 activity with 11 mutant peptides exhibiting 10-fold or greater selectivity for Na_v1.7 over Na_v1.2 (Table 2).

Table 1.

Individual mutations that improve HwTX-IV potency versus Na_v1.7 and selectivity versus Na_v1.2

pIC₅₀ values of mutations that produce improvements in the HwTX-IV's potency versus Na_v1.7 3-fold or greater versus parent recombinant peptide are shown. n = number of independent concentration-response experiments performed.

Peptide	1.7 pIC50	n
rHwTX-IV	7.0 ± 0.14	6
E1N	7.6 ± 0.1	4
E4R	7.9 ± 0.06	3
E4N	7.6 ± 0.03	1
E4Q	7.5 ± 0.10	1
A8R	7.7 ± 0.02	1
S19N	7.7 ± 0.04	2
S19P	7.5 ± 0.09	2
S19Q	7.4 ± 0.22	1
S20N	7.5 ± 0.05	2
K21R	7.9 ± 0.13	2
S25I	7.7 ± 0.07	2
Y33W	7.7 ± 0.03	3
Q34S	7.5 ± 0.07	2
Q34F	7.7 ± 0.18	2
Q34L	7.5 ± 0.04	2
K37R	7.7 ± 0.03	1

Table 2.

Individual mutations that improve HwTX-IV potency versus Na_v1.7 and selectivity versus Na_v1.2

Mean pIC₅₀ values ± S.E. of mutations that conferred a 10-fold or greater improvement in selectivity for Na_v1.7 versus Na_v1.2 are shown. Mutations in bold were selected for follow-up.

Peptide	1.7 pIC50	n	1.2 pIC50	n	Selectivity
N13G	6.8 ± 0.07	1	5.8 ± 0.43	1	1.0
D14P	6.6 ± 0.03	1	5.5 ± 0.51	2	1.1
K18F	6.5 ± 0.03	2	5.0 ± 2.6	2	1.5
S19Q	7.6 ± 0.21	1	6.4 ± 1.1	1	1.2
R26K	6.8 ± 0.22	1	5.8 ± 0.13	1	1.1
R26G	6.6 ± 0.06	2	>5.4	2	>1.2
K27W	7.2 ± 0.06	1	5.8 ± 0.31	2	1.4
W30K	6.9 ± 0.12	1	>5.5	2	1.4
W30Y	5.7 ± 0.08	2	>4.5	1	>1.2
Y33T	7.2 ± 0.07	1	6.2 ± 0.05	1	1.0
G36I	7.0 ± 0.07	2	5.4 ± 0.5	2	1.4

Patch-clamp confirmation of mutant peptide activity on channel gating

Although the FRET-based FLIPR assay is a powerful screening tool, it measures changes in membrane potential, an event initiated by sodium channel gating but easily confounded by the subsequent activation of other voltage-sensitive conductances. In addition, a lack of voltage control in this assay limits its ability to accurately determine the potency of voltage-dependent compounds. Therefore, we verified the activity of recombinant HwTX-IV and six peptides with key individual mutations on sodium channel gating using the QPatch HT automated patch-clamp system. For both Na_v1.7 and Na_v1.2, we saw good agreement between FLIPR and QPatch data for most peptides with a linear fit yielding a R² = 0.80, m = 1.25, and b = −1.97 for Na_v1.7 and R² = 0.84, m = 0.52, and b = 3.79 for Na_v1.2 (Fig. S3) with the discrepancies largely accounted for by assay variability. FLIPR did, however, underestimate the potency of peptides with a lower Na_v1.2 potency, underscoring the need for verification of potency in QPatch for compound selectivity.

Combinatorial mutation of HwTx-IV

The full amino acid scan uncovered a wide range of position-dependent activity relationships throughout the parent peptide. Thus, we introduced multiple individual mutations with favorable effects to look for additive or synergistic effects on Na_v1.7 potency and/or selectivity over Na_v1.2. To this end, we produced a small library of HwTX-IV–based peptides that contained varied combinations of E1N, E4D, Q34S, R26K, Y33W, and G36I. These peptides were tested in FLIPR against Na_v1.7 and Na_v1.2 (see Table S1 for full table of FLIPR activities), and we identified a peptide with improved Na_v1.7 potency (E1N, E4D, Y33W, Q34S-Na_v1.7 pIC₅₀ = 8.1 ± 0.08, and Na_v1.2 pIC₅₀ = 7.8 ± 0.09) and a peptide that showed selectivity for Na_v1.7 over Na_v1.2 (E1N, R26K, Q34S, G36I, Na_v1.7 pIC₅₀ = 7.2 ± 0.1, Na_v1.2 pIC₅₀ = 6.1 ± 0.18). The activity of these peptides was verified in QPatch with good agreement (E1N, E4D, Y33W, Q34S, Na_v1.7 pIC₅₀ = 8.3 ± 0.26, and Na_v1.2 pIC₅₀ = 7.8 ± 0.06, E1N, R26K, Q34S, G36I, and Na_v1.7 pIC₅₀ = 7.4 ± 0.10, Na_v1.2 pIC₅₀ = 5.9 ± 0.14, Fig. 5, A and B, respectively).

Figure 5.

Combinatorial mutation of HwTX-IV improves potency versus Na_v1.7, selectivity versus Na_v1.2. A, left: concentration-response curves from QPatch experiments of recombinant HwTX-IV (blue circles) and E1N, E4D, Y33W, Q34S (green squares) versus Na_v1.7 (left) and Na_v1.2 (right). Below both graphs are representative traces showing current amplitudes in the absence (black) and presence of 30 nm recombinant HwTX-IV (blue) or 300 nm, E1N, E4D, Y33W, Q34S. B, left: concentration-response curves of recombinant HwTX-IV (red circles) and E1N, R26K, Q34S, G36I (orange squares) versus Na_v1.7 (left) and Na_v1.2 (right). Data are mean ± S.E. % inhibition. Below both graphs are representative traces showing current amplitudes in the absence (black) and presence of 300 nm recombinant HwTX-IV (red) or 300 nm E1N, R26K, Q34S, and G36I (orange).

Broader selectivity profiles of best-combinatorial mutants

To further characterize the selectivity profiles of these peptides, we tested them in QPatch against Na_v1.1, 1.3, 1.4, 1.5, and 1.6. Compared with its Na_v1.7 activity, E1N, E4D, Y33W, and Q34S remained active on Na_v1.1 (pIC₅₀ = 8.1 ± 0.10), 1.3 (pIC₅₀ = 8.2 ± 0.14), and 1.6 (pIC₅₀ = 7.5 ± 0.08) but showed diminished activity on Na_v1.4 (pIC₅₀ = 6.0 ± 0.12) and Na_v1.5 (62% inhibition at 10 μm). In contrast, relative to its Na_v1.7 activity, E1N, R26K, Q34S, G36I only showed strong Na_v1.1 (pIC₅₀ = 7.3 ± 0.15) and 1.6 activity (pIC₅₀ = 8.1 ± 0.65) and reduced or no Na_v1.3 (pIC₅₀ = 6.4 ± 1.0), 1.4 (inactive at 3 μm), or 1.5 (inactive at 10 μm) activities (Fig. 6A).

Figure 6.

Broader selectivity profile of combinatorial mutants. A, mean QPatch pIC₅₀ values of HwTX-IV and two combinatorial mutants versus Na_v1.1–Na_v1.7. B, fold selectivity over Na_v1.7 of HwTX-IV and two combinatorial peptides. Data are derived from logistic Hill fits of concentration-response experiments, n = x-y cells/concentration. *** denotes estimated pIC₅₀/selectivity values based upon single concentration inhibition data with an assumption of a Hill slope of 1.

These mutations also translated to improvements in selectivity for Na_v1.7 over some Na_v isoforms compared with WT HwTX-IV. For E1N, E4D, Y33W, and Q34S, the most substantial gains were over Na_v1.2 (increased from 0.4- to ∼2-fold) and Na_v1.6 (equipotent to 4-fold) with selectivity over other isoforms largely unchanged. E1N, R26K, Q34S, and G36I also showed the most robust improvement in selectivity for Na_v1.7 over Na_v1.2 (increased from 0.4- to 30-fold) but also showed meaningful improvements in selectivity over Na_v1.3 (increased from equipotent 10-fold) and Na_v1.6 (increased from equipotent to 3-fold) (Fig. 6B).

Grafting of additional mutations into the best HwTx-IV combinatorial mutants

To further improve potency and selectivity of the two best-combinatorial peptides, we grafted additional point mutations into the two scaffolds. These peptides were tested in QPatch against Na_v1.7. Most of the substitutions we introduced, however, significantly decreased Na_v17 potency (Table S2). Significant improvements in the E1N, E4D, Y33W, and Q34S's Na_v1.7 potency were observed with the following mutations: K37Q, R26T, K37S, R26W, R29N, and R29H. In contrast, no changes to E1N, R26K, Q34S, and G36I resulted in significant improvements in potency. Select peptides were subsequently tested against Na_v1.2. This identified R26T as a favorable addition to E1N, E4D, Y33W, and Q34S (QPatch: Na_v1.7 pIC₅₀ = 8.9 ± 0.11, and Na_v1.2 pIC₅₀ = 7.7 ± 0.11, see Fig. 7).

Figure 7.

Improvement of combinatorial peptides. Concentration-response curves from QPatch experiments testing E1N, E4D, Y33W, and Q34S (green circles) and E1N, E4D, R26T, Y33W, and Q34S (purple squares) versus Na_v1.7 (left) and Na_v1.2 (right) n = 3–5 cells per concentration. Below both graphs are representative QPatch traces showing current amplitudes in the absence (black) and presence of E1N, E4D, Y33W, and Q34S (green) or E1N, E4D, R26T, and Y33W (purple). Representative Na_v1.7 traces were recorded in the presence of 3 nm peptide and Na_v1.2 in the presence of 30 nm peptide.

Docking model of HwTX-IV on Na_v1.7

HwTx-IV was docked onto a Na_v1.7 homology model generated using a composite method with a cryo-EM structure of Na_v1.7 (PDB 6J8G) serving as the template for most of the structure except for the domain II voltage sensor (VSD2), which used as a template the deactivated voltage sensor (PDB 6N4R). Once the WT Nav1.7 model was generated, hwentoxin-IV (PDB 1MB6) was manually docked to the Na_v1.7 homology model using the protoxin-II–binding site from the Na_v1.7 VSD2–NavAb channel chimera protein structure in complex with protoxin-II (PDB 6N4I) as a guide. HwTx-IV was initially structurally aligned to the protoxin-II and then manually adjusted to generate a docking pose. The entire system was further minimized and refined using molecular dynamics.

The final docked and refined model of HwTX-IV to Na_v1.7 shows key interactions involving residues that impact HwTX-IV potency and/or selectivity over Na_v1.2. Lys-27 forms a salt bridge with Glu-818 (with proximity to Asp-816), and Lys-32 forms a salt bridge with Glu-811 (Fig. 8A); this area is key to VSD2-gating movement. Trp-30 sits in a hydrophobic pocket formed by Ala-766, Ile-767, Leu-770, and Leu-812 (Fig. 8B). The indole nitrogen from the Trp-30 also forms a hydrogen-bonding interaction with the backbone carbonyl of Leu-812. Finally, Arg-26 makes a salt-bridge interaction with Glu-760, whereas Arg-29 primarily makes an interaction with the polar headgroups of the lipid membrane, although within the molecular dynamic's simulations, 30% of the time Arg-29 rotated and also made a salt-bridge interaction with Glu-760 (Fig. 8C).

Figure 8.

Model of key interactions of huwentoxin-IV with the Na_v1.7 VSDII. A, Lys-27 and Lys-32 interact with Glu-811, Glu-818, and the previously identified LFLAD motif in Na_v1.7 to electrostatically impair channel gating. B, Trp-30 docks into a hydrophobic pocket composed of Ala-766, Leu-770, and Leu-812 and also forms a hydrogen bonding interaction between the nitrogen of the indole and the backbone carbonyl of Leu-812. C, Arg-26 forms a salt bridge with Glu-760 and Arg-29 engages in a hydrogen bond interaction with Asn-763.

Noncoded amino acids

We also selected positions 6, 19, 20, 21, 27, 30, and 32 in which to substitute noncoded amino acid analogs into E1N, K21F, R26K, Q34S, and G36I. (Na_v1.7 pIC₅₀ = 7.2 ± 0.2 and Na_v1.2 pIC₅₀ = 6.1 ± 0.1). Selection was based upon apparent participation in a Na_v channel epitope, phospholipid interaction, and/or their location in regions of potential metabolic liability (i.e. unstructured secondary structure). It should be noted that the selections of amino acid types (see Fig. S4) were strongly influenced by resource limitations and are therefore far from comprehensive. In addition to highlighting various sequence–activity relationships, the full scan identified position-specific tolerance to substitution. Four of the positions selected for noncoded amino acid substitution are in positions of low tolerance (6, 27, 30, and 32). However, as Table 3 illustrates, specific noncoded amino acid substitutions can be beneficial in places where coded amino acid substitutions are not.

Table 3.

Substitution of noncoded amino acid analogs at select positions in a combinatorial HwTX-IV peptide

Impact of substitutions on peptide potency versus Na_v1.7 and Na_v1.2 was measured in QPatch. n = 3–8 cells per peptide concentration (except for Trp-30-(5-methoxy-Trp) versus Na_v1.2; n = 1 cell).

Substitution	QPatch (%I at 1 μm) or pIC50± S.E.
	Nav1.7	Nav1.2
E1N, K21F, R26K, Q34S, G36I (Parent)	7.2 ± 0.2	6.1 ± 0.1
Phe-6-(pyridylalanine)	6.3 ± 0.1	(7.2 ± 9.3%)
Phe-6-(trifluorophenylalanine)	8.0 ± 0.1	6.7 ± 0.1
Ser-19-(3-hydroxyproline)	7.7 ± 0.2	6.4 ± 0.2
Ser-19-(4-aminoproline)	5.5 ± 0.1	(34.8 ± 9.8%)
Ser-19-(azetidine 2-carboxylic acid)	8.0 ± 0.1	6.2 ± 0.2
Ser-20-(3-hydroxyproline)	8.0 ± 0.2	6.4 ± 0.2
Ser-20-(4-aminoproline)	7.3 ± 0.1	6.0 ± 0.1
Ser-20-(azetidine 2-carboxylic acid)	7.7 ± 0.1	6.2 ± 0.1
Lys-21-(4-pyridylalanine)	7.7 ± 0.1	6.0 ± 0.4
Lys-27-(Lys(Me)2)	7.8 ± 0.4	6.0 ± 0.1
Lys-32-(Lys(Me)2)	(9.0 ± 6.5%)	(4.9 ± 17.5%)
Lys-32-(ornithine)	(43.6 ± 7.8%)	(−6.6 ± 7.5%)
Trp-30-(5-methoxy-Trp)	8.3 ± 0.4	(65.2%)
All D versions	(6.2 ± 9.7%)	(6.2 ± 9.7%)
Trp-30-(fluoro-Trp)	8.0 ± 0.1	6.8 ± 0.7

Although the full scan identified position 6 to be nontolerant of change from phenylalanine, modeling indicates position 6 to be at the VSD/membrane interface. Assuming ring electronics and quadrupoles drive cation–π and π–π interactions, which may exist between the peptide and channel and/or membrane, substitutions were made that maintain aromaticity and shape but modify the quadrupole of the parent phenyl ring. Despite its structural similarity 4-pyridylalanine substituted at position 6 decreased both Na_v1.7 potency and selectivity toward Na_v1.2. In contrast, 2,4,6-trifluorophenylalanine at position 6 slightly enhanced both potency toward Na_v1.7 and selectivity against Na_v1.2.

Residues 19 and 20 lie within loop 3 and may initiate a β-turn that precedes anti-parallel β-strand 1. We substituted 3-hydroxyproline, 4-aminoproline, and azetidine 2-carboxylic acid at these positions to enhance the β-turn while maintaining solubility and hydrogen-bond donor properties, as well as to minimize the amount of unstructured sequences visible to proteases. Substitution of azetidine 2-carboxylic acid and 3-hydroxyproline at positions 19 and 20 had little effect. In contrast, 4-aminoproline greatly reduced activity at both Na_v1.7 and Na_v1.2 when substituted at position 19. When substituted at position 20, potency at Na_v1.7 was maintained.

Although 4-pyridylalanine was not a good phenylalanine substitution at position 6, at position 21 its physicochemical properties proved to be a favorable analog of the parent lysine. An increase in both Na_v1.7 potency and Na_v1.2 selectivity was observed. This mirrors the results of the position 21 phenylalanine mutant from the full scan.

Modulating basicity of the lysine side-chain amine was of interest not only at position 21 but at positions 27 and 32 as well. Amine basicity may play a role in peptide channel–binding in a variety of ways: strength of cation–π interactions between peptide and channel and de-solvation of peptide to name a few. N,N-ϵ-Dimethyl-lysine replaced lysine at position 27 with good results; potency was maintained, and selectivity over Na_v1.2 was dramatically increased. However, the same substitution at position 32 effectively killed activity at both Na_v1.7 and Na_v1.2. The shorter side-chain lysine analog, ornithine, showed similar negative results. These results, in combination with the full scan data, strongly indicate position 32 to be intolerant to any residue other than lysine.

Discussion

Peptides contained in spider venoms have been refined by many years of evolution to modulate biological targets (19). These evolutionary pressures, however, drive toxins toward functional profiles that confer a selective advantage in predation. Huwentoxin exemplifies this trend as a potent, nonselective blocker of sodium channels which can disrupt muscle and neuronal functions. Given this sustained optimization, it is unsurprising that analyses that looked at the mean effect of all changes in HwTX-IV's loops or at specific positions only revealed decreases in potency and that only introduction of specific properties or substitutions produced improvements.

We previously reported a critical role for Phe-6, Pro-11, Asp-14, Leu-22, Ser-25, Trp-30, Lys-32, Tyr-33, and Ile-35 in maintaining huwentoxin activity on Na_v1.7 and the same residues, except Ile-35, critical for Na_v1.2 block (23). The excellent agreement between these findings and this current study is notable, where all positions previously flagged as being critical for Na_v1.7 activity, except Leu-22, caused significant decreases in mean pIC₅₀ Na_v1.7. This study, however, also identifies Ile-5, Lys-27, and Thr-28 as being important for Na_v1.7 block. We also confirmed the importance of all residues except Pro-11, Leu-22, and Ser-25 for Na_v1.2 block as well as identifying and Arg-26, Lys-27, and Ile-35 as fundamental to Na_v1.2 activity.

From our engineering efforts we selected a compact list of functionally important mutations (E1N, E4R, Y33W, Q34S, and G36I) and methodically combined them. This effort led to the discovery of two lead peptides: E1N, E4R, Y33W, Q34S, which had an improved Na_v1.7 potency but lacked selectivity, and E1N, R26K, Q34S, G36I, which displayed greater selectivity for Na_v1.7 over Na_v1.2. Our subsequent grafting of favorable mutations onto the combinatorial peptides was successful in significantly improving potency and selectivity in several peptides, the best of which was the addition of R26T to E1N, E4R, Y33W, and Q34S.

It is notable that the slope of the concentration-response curves was increased for several of the peptides we tested. It is possible that this change reflects an affinity for additional Na_v VSDs for these peptides. Significant additional experimentation would be required to confirm this and could be an interesting focus for future studies.

Many of the changes in activity induced by mutation of HwTX-IV can be explained by the docking model we present in this paper. Previous studies have shown that huwentoxin inhibits sodium channel gating by trapping the channel's domain II voltage sensor in a resting state (36) through an interaction with the S3–S4 linker (32). More recently, studies have illuminated the location where and the general orientation in which HwTX-IV docks on the Na_v1.7 VSD2 and key interactions that mediate the highly-homologous peptide ProTX-II's interaction with VSD2 (37, 38). Based upon these new data, we developed a new model of HwTX-IV binding to the VSD that differs substantially from our previously published docking pose. The finding that loop 4 and the C terminus of HwTX are generally intolerant to mutation aligns well with our current model because these are the segments of HwTX-IV that are most proximal to the VSD and the cell membrane and contain the bulk of the residues that mediate key interactions with the Na_v1.7 VSD. In addition, our new model provides a framework for understanding the functional behavior of point mutants as well.

Xu et al. (37) identified five residues in protoxin-II that they describe as a hydrophobic “anchor,” which facilitates membrane partitioning and ultimately proper docking of the toxin into the VSD. Our current hypothesis is that Ile-5, Phe-6, Trp-30, and Ile-35 are analogs to Trp-5, Trp-7, Trp-24, and Leu-29 in protoxin-II and thus generate affinity for the lipid membrane in a similar manner. Ile-5, Phe-6, Trp-30, and Ile-35 are all extraordinarily intolerant to mutation in our hands, suggesting that each residue plays a critical role in VGSC interaction. Of these positions the only mutation that confers improvement in Na_v1.7 and Na_v1.2 activity is insertion of a tyrosine at position 5. This would not substantially affect lipophilicity at this position, suggesting that Ile-5 can mediate a more specific interaction in addition to functioning as a membrane anchor.

Other groups have shown that enhancing the affinity of HwTX-IV for lipid membranes correlates with increased potency for Na_v1.7 (26). This is confirmed by the mutations E4Y, A8Y, Y33W, Q34F, and G36I, all of which dramatically improve Na_v1.7 potency. Given that these residues flank Ile-5 and Ile-35, it seems reasonable that these changes would improve HwTX-IV's ability to embed in the membrane, further strengthening the correlation between hydrophobicity and potency. Notably, however, the introduction of lipophilic amino acids did not appreciably alter activity on Na_v1.7 or Na_v1.2 in the loop analysis other than a small increase in Na_v1.2 potency with nonpolar substitutions in Glu-1. Thus, it is clear that the correlation between lipophilicity and potency is not universal to the entire HwTX-IV molecule. Rather, it is likely largely driven by hydrophobic substitutions in membrane-adjacent regions of the toxin.

Xu et al. (37) also revealed a polar interface for protoxin-II which is responsible for the electrostatic repulsion that antagonizes Na_v1.7's S4 helix gating charge movement. The residues that mediate this gating modulation are closely mirrored by Lys-27 and Lys-32 in our model. Lys-27 is well-positioned to interact with Glu-818 and Lys-32 with Glu-811 in the VSD (Fig. 8A). These data align nicely with a previous mutagenesis study that identified Glu-811 and Glu-818 as points of interaction critical for maintaining HwTX-IV's Na_v1.7 activity (39). It is notable that HwTX-IV's activity is decreased if lysine is replaced by other basic amino acids at Lys-27. In fact, only aromatic substitutions maintain activity in this location, likely because they add lipophilicity to the peptide without disrupting Lys-32's ability to impair gating. This hypothesis is strengthened by the noncoded amino acid data where increasing lipophilicity of Lys-27 via dimethyl addition maintained activity.

Lys-32 is the only position in HwTX-IV in which every modification we attempted (even the close analog ornithine) completely inhibited the toxin's ability to block both Na_v1.7 and Na_v1.2. We previously established that substitution of an alanine at this position does not cause a gross disruption of HwTX-IV structure (23). Given that in this mutant Lys-27 is intact and that aromatic substitutions do not maintain toxin activity, this strongly suggests that Lys-32 is the lynchpin of HwTX-IV's electrostatic modulation of Na_v1.7 gating.

Trp-30 is unique in that in addition to orienting the hydrophobic face of HwTX to the membrane, it is well-positioned to interact with the previously described hydrophobic “shelf” between the S2 and S3 VSD helices composed of Ala-766, Leu-770, and Leu-812 as shown in Fig. 8B (37). It is interesting to note, however, that the sole change (other than 5-methoxy-Trp and 5-fluoro-Trp) that can be made at this position that does not fully disrupt HwTX-IV's Na_v1.7 activity is W30K. This finding is particularly striking because this mutant's Na_v1.7 potency is only marginally decreased but is decreased >30-fold on Na_v1.2.

Earlier studies have highlighted positions Glu-1 and Glu-4 as locations where the addition of nonpolar amino acids (Ala and Glu) produce an increase in HwTX-IV potency (23, 27, 28). This dataset confirms these findings for position Glu-1 on Na_v1.2. We also show that introduction of basicity is generally more favorable for Na_v1.2 activity than nonpolar substitutions at both Glu-1 and Glu-4. We did not, however, test E1G and E4G, which have been reported as modifications that robustly improve HwTX-IV potency (27, 28).

This study also identified positions where mutations created a preference for HwTX-IV to block Na_v1.7 over Na_v1.2. The most prominent residue where this occurred was at Arg-26, where the mean effect of mutation was a significant reduction in Na_v1.2 potency with a nominal effect on Na_v1.7 potency. This is most evident when Arg-26 was mutated to a Lys and showed a small reduction in Na_v1.7 potency but a 10-fold reduction in Na_v1.2 activity. Intriguingly, in our model, Arg-26 is situated such that it could potentially exploit a nearby site that differs between the Na_v1.7 and Na_v1.2 VSDs: Glu-760 and Gln-786, respectively. The lack of the charge–charge interaction in Na_v1.2 may contribute to this selectivity (Fig. 8C).

HwTX-IV's potency for VGSCs renders it an attractive starting scaffold for engineering a peptide that is selective for Na_v1.7. Here, we have identified general trends as well as specific mutations that improve HwTx's Na_v1.7 potency and selectivity for Na_v1.7 over other VGSC isoforms. In addition, through molecular modeling we have hypothesized key structural determinants that dictate HwTX-IV's sodium channel activity. Taken together, these data provide valuable information for additional study and improvement of HwTX-IV and can serve as guide for future work focused on the discovery of a Na_v1.7-selective analgesic.

Experimental procedures

Synthesis/production of HwTx-Iv peptide variants

Recombinant peptides were produced and purified as described previously (23). Synthetic peptides were produced by using a Fmoc–t-butyl strategy using Fmoc–Rink–mBHA resin (Peptides International, Louisville, KY). Fmoc deprotection was facilitated with 20% piperidine in N,N-dimethylformamide. The linear peptide was assembled using diisopropyl carbodiimide/6-Cl-1-hydroxybenzotriazole–based activation for each amino acid. All Cys residues were trityl-protected on their sulfhydryl groups. Following assembly of each of the linear peptide chains, the peptides were cleaved from the solid-phase resin and simultaneously side-chain deprotected using a reagent K acidolytic cleavage step for 3 h at room temperature (30). The crude peptide was isolated by filtration through a Buechner fritted glass filter to remove the resin beads and subsequently precipitated into ice-cold diethyl ether. The precipitated peptide was isolated by centrifugation and was resuspended and washed three times with ice-cold diethyl ether. The crude linear peptide was subsequently dissolved in 50% aqueous AcOH and purified by preparative RP-HPLC to >75% purity as determined by analytical RP-HPLC. The purified linear peptide was diluted into a buffer composed of 0.1 m Tris-HCl, 0.5 m guanidinium HCl, containing 0.5 mm GSH (oxidized) and 2 mm GSH (reduced). After dilution of the purified linear peptide into the folding buffer, the pH was adjusted with NH₄OH to 7.8 to facilitate oxidative folding for 18 hr at room temperature. The cyclized product was subsequently purified by preparative RP-HPLC (Phenomenex Luna C18) using a gradient form 10–35% MeCN versus 0.05% TFA in H₂O at 50 °C. Fractions with the targeted mass spectrum for each peptide and a purity of >95% as determined by analytical RP-HPLC were pooled and lyophilized.

Cell culture

HEK293 cells stably expressing human Na_v1.7 (Millipore), human Na_v1.2 (supplied by Dr. H. A. Hartmann, University of Maryland Biotechnology Institute), and human Na_v1.4 and human Na_v1.5 (Dr. A. George, University of Pennsylvania) were cultured in Dulbecco's modified Eagle's medium/F-12 medium with l-glutamine, supplemented with 10% fetal bovine serum (FBS), 100 μg/ml penicillin/streptomycin, 400 μg/ml geneticin, and 100 μm nonessential amino acids (NEAAs) at 37 °C and in 5% CO₂. Tetracycline-inducible CHO cell lines (human Na_v1.1 and human Na_v1.6, Chantest, Cleveland, OH) were cultured in Ham's F-12 media (Mediatech) supplemented with 10% FBS, 100 μg/ml penicillin/streptomycin, 0.01 mg/ml blasticidin, 0.4 mg/ml Zeocin. CHO cells stably expressing human Na_v1.3 were maintained in Iscove's modified Dulbecco's medium supplemented with 10% FBS, 1% sod supplement, 1% NEAAs, and 400 μg/ml geneticin. All culture reagents were from Invitrogen unless otherwise specified.

FLIPR

Cells were seeded into a 384-well poly-d-lysine–coated optical plates (Corning) at a density of 25,000 cells/well and allowed to adhere for 24 h. On the day of experimentation, cells were washed with assay buffer (135 mm NaCl, 4 mm KCl, 3 mm CaCl₂, 1 mm MgCl₂, 5 mm glucose, and 10 mm HEPES, pH 7.4, 300 mosm) using a Biotek EL405 plate washer (Biotek, Winooski, VT). Cells were then loaded with voltage-sensitive FRET dyes as described previously (31). Briefly, 8-octadecyloxypyrene-1,3,6-trisulfonic acid DMSO stock (PTS₁₈) was diluted to 6 μm (2× final concentration) by adding the appropriate volume of stock solution into in equal volume of 10% Pluronic F127 and diluting with assay buffer. Dye solution was then diluted into an equal volume of assay buffer in the cell plates. Plates were incubated with the dye for 30 min at room temperature in the dark and washed again with assay buffer. Bis-(1,3-diethylthiobarbituric acid)trimethine oxonol (Molecular Probes) was solubilized at 20 μm, and the blocking dye VABSC-1 was added to the oxonol solution at 800 μm (2× final concentrations). The oxonol/VABSC-1 solution was transferred to intermediate plates where test compounds were added at 2× their final concentration. This mixture was then diluted in an equal volume of assay buffer in the cell plates and incubated for a 1 h in the dark.

To test compounds, cell plates were placed into a FLIPR_TETRA, and baseline fluorescence (excitation filter = 400 nm and emission filter = 580 nm) was recorded for 10 s. Cells were subsequently challenged with veratridine to activate voltage-gated sodium channels, and the changes in fluorescence were recorded at 0.5 Hz for 110 s. HwTX-IV inhibition was determined by normalizing the fluorescent response (R) in each well of a serial dilution to the response of negative (NC, 0.5% DMSO) and positive controls (PC, 10 μm tetracaine).

In our cell lines, the Na_v channel expression can vary from passage to passage. In addition, large depolarizations can impair HwTX-IV's ability to block VGSCs (32). Therefore, prior to testing experimental peptides, an initial plate was used to establish the appropriate concentration of veratridine necessary to activate channels in a consistent manner across experiments. To achieve this end, eight identical concentration-response experiments of synthetic HwTX-IV (Peptides International, 1 μm to 0.02 nm, 3-fold dilution steps) were prepared. Each HwTX concentration-response experiment was then challenged with a different concentration of veratridine (100–0.4 μm, 2-fold dilution steps). The concentration of veratridine that produced an IC₅₀ value for HwTX-IV within the 10–30 nm range was selected for use with experimental peptides in subsequent plates.

The % inhibition (%I) values were calculated as shown in Equation 1.

$$\%\text{I}=100\times \left(\frac{\text{R}-\text{NC}}{\text{PC}-\text{NC}}\right)$$ (Eq. 1)

The % inhibition data from each FLIPR experiment were aggregated and analyzed by fitting with a nonlinear regression in Origin (Microcal) to determine the pIC₅₀ values using Equation 2,

$$y=A1+\frac{A2-A1}{1+{10}^{\left(\log \times 0-x\right)p}}$$ (Eq. 2)

where A1 is bottom asymptote; A2 is = top asymptote; log × 0 is center; and p = Hill slope. Upper and lower asymptotes, log × 0, and p were all allowed to vary. pIC₅₀ values for each peptide were then averaged to determine the mean pIC₅₀ and the associated standard error of the mean.

QPatch

Cell preparation for QPatch assays was performed as described previously (23). Briefly, cells were dissociated and resuspended in extracellular solution containing 137 mm NaCl, 4 mm KCl, 1 mm MgCl₂, 2 mm CaCl₂, 5 mm glucose, and 10 mm HEPES, pH 7.3, 300 mosm/liter. The intracellular solution contained 135 mm CsF, 10 mm CsCl, 5 mm EGTA, 5 mm NaCl, and 10 mm HEPES, pH 7.3, 290 mosm/liter.

To elicit sodium currents, cells were first hyperpolarized from the holding potential to −120 mV for 2 s and then depolarized to 0 mV for 5 ms before returning to the holding potential (−75 mV for Na_v1.7 and −65 mV for Na_v1.2, which are the V_½ values for steady-state inactivation for the respective channels; data not shown). This protocol was repeated once every 60 s during liquid applications. Cells were otherwise held at −75 mV (Na_v1.7) or −65 mV (Na_v1.2) when the voltage protocol was not in process.

Upon establishment of the whole-cell recording configuration, a total of five applications of the extracellular solution were administered to the cell; 1 × buffer + 0.1% BSA (NC), 3 × peptide (or NC), 1 × 1 μm tetrodotoxin (PC). All solutions contained 0.1% BSA except for the tetrodotoxin solution. The voltage protocol was executed 10 times after each application. Currents were sampled at 25 kHz and filtered at 5 kHz with an 8-pole Bessel filter. The series resistance compensation level was set at 80%. All experiments were performed at room temperature (∼22 °C). The % inhibition values for each cell was calculated based upon peak current amplitude in the presence of compound relative to the tetrodotoxin-sensitive component of the current as described for FLIPR above. Raw % inhibition (%I_raw) values were adjusted for current rundown based upon the mean % inhibition calculated from associated control experiments from each batch of cells where only 0.1% BSA was applied (%I_BSA) as shown in Equation 3.

$$\%\text{I}=100\times \left(\frac{\%{\text{I}}_{\text{raw}}-\%{\text{I}}_{\text{BSA}}}{100-\%{\text{I}}_{\text{BSA}}}\right)$$ (Eq. 3)

pIC₅₀ values from QPatch were derived from a nonlinear mixed effects model fit by maximum likelihood as implemented in the function nlme in R (33). In this model, the relationship between the logarithm of concentration and percent inhibition was modeled as a logistic curve defined by lower and upper asymptotes, logarithm of the ED₅₀, and slope. The mixed effects model featured overall slope and EC₅₀ as fixed effects and slope and ED₅₀ across experiments as random effects. This design enabled us to account for inter-experimental variation. Upper asymptote (i.e. maximal inhibition) was always fixed at 100%. In cases when this model failed to converge, we fitted an alternative model with fixed upper and lower asymptotes at (100 and 0% response). For all experiments chemicals were obtained from Sigma unless otherwise specified.

Homology model of Na_v1.7 with huwentoxin-IV

Using Schrodinger Maestro's (34) prime module, a WT Na_v1.7 (UniProt Q15858) homology model with huwentoxin-IV bound was generated using a composite homology model with a cryo-EM structure of Na_v1.7 (PDB 6J8G) serving as the template for the majority of the structure except for domain II voltage sensor (residues 720–831), which used as a template the deactivated voltage sensor (PDB 6N4R). Once the WT Na_v1.7 model was generated, huwentoxin-IV (PDB 1MB6) was manually docked to the Na_v1.7 homology model using the protoxin-II–binding site from the Na_v1.7 VSD2–NavAb channel chimera protein structure in complex with protoxin-II (PDB 6N4I) as a guide.

The complete Na_v1.7–huwentoxin-IV homology model was then prepared for molecular dynamics simulation by using Schrodinger's protein preparation wizard with default settings of assigning bond orders, adding hydrogen bonds, creating disulfide bonds, and capping termini for Na_v1.7 (huwentoxin-IV's termini were not capped), and generating het states using Epik. Hydrogen bonds were assigned using PROPKA at pH 7.0. The system was then minimized using restrained minimization with convergence of heavy atoms to 0.3 Å RMSD using OPLS3e. A molecular dynamics system was then built using this prepared protein structure with Schrodinger's System Builder wizard with a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine membrane (placed based on Na_v1.7 helices), an orthorhombic box with 15 Å buffer in all dimensions; ions were added to neutralize the system; and NaCl added to a salt concentration of 0.15 m.

Molecular dynamics simulations were done using Schrodinger's Desmond (35). The Na_v1.7-HwTxIV system was relaxed before simulation and equilibrated for 1 ns using NPT with pressure set to 1.01325 bar and temperature set to 300 K with backbone Cα atoms for all Na_v1.7 residues restrained with a force constant of 1 kcal/mol except for the residues corresponding to VSD2 (residues 730–850). Following this initial equilibration, a 100 ns production simulation was run using NVT with a temperature of 300K, again with backbone CA atoms for all Na_v1.7 residues restrained with a force constant of 1 kcal/mol, except for the VSD2 residues. Following the production simulation, the trajectory was clustered based on backbone RMSD of Na_v1.7. The top clustered model was then further minimized using Desmond for 250 ps to generate the final model of HwTxIV bound to Na_v1.7 (Fig. S5).

Author contributions

R. A. N., M. F., A. G., M. J. H., and A. D. W. conceptualization; R. A. N., N. A. M., and Y. L. data curation; R. A. N., N. A. M., Y. L., O. L., and S. Y. formal analysis; R. A. N., M. F., A. Y. S., N. A. M., and Y. L. investigation; R. A. N. and N. A. M. visualization; R. A. N., M. F., A. Y. S., Y. L., R. F., and M. W. P. methodology; R. A. N., M. F., and A. G. writing-original draft; R. A. N., M. F., and A. D. W. project administration; R. A. N., M. F., A. G., A. Y. S., N. A. M., Y. L., R. F., O. L., S. Y., M. W. P., M. J. H., and A. D. W. writing-review and editing; M. J. H. and A. D. W. supervision; A. D. W. funding acquisition; A. Y. S. in silico modeling.