The structure, dynamics and selectivity profile of a NaV1.7 potency-optimised huwentoxin-IV variant

Venom-derived peptides have attracted much attention as potential lead molecules for pharmaceutical development. A well-known example is Huwentoxin-IV (HwTx-IV), a peptide toxin isolated from the venom of the Chinese bird-eating spider Haplopelma schmitdi. HwTx-IV was identified as a potent blocker of a human voltage-gated sodium channel (hNaV1.7), which is a genetically validated analgesic target. The peptide was promising as it showed high potency at NaV1.7 (IC50 ~26 nM) and selectivity over the cardiac NaV subtype (NaV1.5). Mutagenesis studies aimed at optimising the potency of the peptide resulted in the development of a triple-mutant of HwTx-IV (E1G, E4G, Y33W, m3-HwTx-IV) with significantly increased potency against hNaV1.7 (IC50 = 0.4 ± 0.1 nM) without increased potency against hNaV1.5. The activity of m3-HwTx-IV against other NaV subtypes was, however, not investigated. Similarly, the structure of the mutant peptide was not characterised, limiting the interpretation of the observed increase in potency. In this study we produced isotope-labelled recombinant m3-HwTx-IV in E. coli, which enabled us to characterise the atomic-resolution structure and dynamics of the peptide by NMR spectroscopy. The results show that the structure of the peptide is not perturbed by the mutations, whilst the relaxation studies reveal that residues in the active site of the peptide undergo conformational exchange. Additionally, the NaV subtype selectivity of the recombinant peptide was characterised, revealing potent inhibition of neuronal NaV subtypes 1.1, 1.2, 1.3, 1.6 and 1.7. In parallel to the in vitro studies, we investigated NaV1.7 target engagement of the peptide in vivo using a rodent pain model, where m3-HwTx-IV dose-dependently suppressed spontaneous pain induced by the NaV1.7 activator OD1. Thus, our results provide further insight into the structure and dynamics of this class of peptides that may prove useful in guiding the development of inhibitors with improved selectivity for analgesic NaV subtypes.

Introduction complete subtype selectivity is not known. Similarly, although the structure of the wild type (WT) peptide and several variants have been determined [19,20], the effect of the triple mutation on the pharmacophore is not known and a high quality structure would provide valuable information for rationalising the observed gain in potency.
To this end, we produced a recombinant, isotopically labelled version of m 3 -HwTx-IV, which allowed us to determine (1) a high-resolution 3D structure of the peptide using heteronuclear NMR and (2) the dynamics of the peptide using 15 N relaxation experiments. Although the overall fold of m 3 -HwTx-IV remained the same as the parent toxin [14], our dynamics data show large amplitude motion in the pharmacophore region indicating that key residues undergo conformational exchange in the μs-ms time scale. In vitro and in vivo functional data indicate that m 3 -HwTx-IV has the same Na V selectivity profile as the WT toxin, and we show for the first time that the peptide is a potent inhibitor of Na V 1.6 and Na V 1.1 channels. The full selectivity profile of the channel reveals that Na V isoforms 1.1, 1.2, 1.3, 1.6 and 1.7 are sensitive to the toxin whilst isoforms 1.4, 1.5 and 1.8 are not. Finally, we show that the peptide has Na V 1.7 on-target activity in vivo using a recently developed Na V 1.7 target-engagement mouse model [21]. Our findings provide further insight into the structure, function and activity of this fascinating class of peptides in the pursuit of turning venoms into analgesic drugs.

Recombinant peptide production
A codon-optimised synthetic gene encoding m 3 -HwTx-IV was synthesised by GeneArt (Thermo Fisher Scientific) and subcloned into an in-house, modified version of pLIC expression vector, between the KpnI and SacI restriction sites, in frame with a MalE signal sequence to direct secretion of the fusion-protein to the periplasmic space of E. coli, where disulfidebond formation and folding of the molecule takes place [22]. Competent E. coli BL21 (DE3) cells were transformed with the plasmid encoding His 6 -MBP-m 3 -HwTx-IV fusion protein. Cells were grown on a LB agar plate supplemented with 100 μg/mL ampicillin (Amp). LB/ Amp (100 μg/mL) medium was inoculated from a single colony or glycerol stock and the cells were grown for~18 h (overnight) at 30˚C whilst shaking at 120 rpm until reaching an OD 600 of~2. Fresh LB/Amp solution was inoculated with the pre-culture to an OD 600 of~0.05 under agitation at 120 rpm in 5 litre baffled flasks at 37˚C until reaching an OD 600 of~0.4, at which time the growth temperature was reduced to 16˚C. Protein production was induced at OD 600 of~0.8 by addition of isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.2 mM and growth was continued overnight at 16˚C. The cells were harvested by centrifugation (6,000 g for 15 min at 4˚C) and the cell pellets stored at -80˚C for subsequent purification. In order to produce uniformly 13 C/ 15 N-labelled m 3 -HwTx-IV peptide, LB broth was replaced with M9 minimal medium supplemented with 15 NH 4 Cl and 13 C 6 -glucose as the sole nitrogen and carbon sources, respectively [23].

Peptide purification
Cells overexpressing His 6 -MBP-m 3 -HwTx-IV fusion protein were resuspended by stirring in lysis buffer (40 mM Tris-HCl, pH 8.0, and 400 mM NaCl) at 4˚C for 1 h, or until reaching complete homogeneity. Cells were lysed by continuous flow cell disruption (Constant Systems Ltd., TS Series Benchtop; triple passage at constant pressure of 32 kpsi). The cell lysate was centrifuged (40,000 g, 30 min, 4˚C) to remove cell debris. The supernatant was loaded onto a Ni-NTA column equilibrated with lysis buffer. The column was washed with 3-5 column volumes (CV) of lysis buffer, then non-specifically bound proteins were eluted with the same buffer containing 40 mM imidazole. The fusion protein was then eluted from the column with lysis buffer supplemented with 250 mM imidazole. The peak fractions were pooled and then desalted and concentrated using an Amicon Ultra-15 centrifugal filter unit (Merck Millipore) with a 30 kDa cut-off membrane. The protein was then digested to yield His 6 -MBP and m 3 -HwTx-IV peptide using 50 μg/ml tobacco etch virus (TEV) protease in a freshly prepared redox buffer containing 3 mM reduced glutathione (GSH) and 0.3 mM oxidised glutathione (GSSG), gently shaking overnight, at room temperature. Fusion protein cleavage was confirmed by SDS-PAGE, then the cleavage reaction mixture was acidified by adding 1% trifluoroacetic acid (TFA). Precipitated protein was removed by centrifugation followed by membrane filtration (0.45 μm). Reverse-phase (RP) HPLC purification of m 3 -HwTx-IV was performed using a C 4 semi-preparative column (Phenomenex). The bound peptide was eluted at 3 ml/min with a linear gradient of 5-80% solvent B (90% Acetonitrile (ACN), 0.043% TFA) in solvent A (0.05% TFA in water). UV absorbance was monitored at 214 and 280 nm and fractions corresponding to the peptide of interest were pooled, lyophilised and resuspended in pure water. UV absorbance was measured (Thermo Scientific™ NanoDrop 2000) at 280 nm, and the concentration was calculated using the predicted extinction coefficient of the peptide. Identity and purity of the peptide was further characterised by liquid chromatography mass spectrometry (LC-MS), matrix-assisted laser desorption/ionisation time-of-flight (MALDI-TOF) mass spectrometry and analytical RP-HPLC with a C 18 analytical column (Phenomenex) using a gradient of 1% solvent B/min at a flow rate of 1 ml/min.

Mass spectrometry
MALDI-TOF MS performed on a Bruker autoflex speed High-Performance MALDI-TOF System (Bruker, MA, USA) was employed to determine peptide mass. RP-HPLC fractions were mixed (2:1 v/v) with α-cyano-4-hydroxycinnamic acid (CHCA) matrix (5 mg/ml in 50/50 ACN/H 2 O) on a MALDI plate spot and dried at room temperature. The MALDI-TOF mass spectrum was acquired in positive reflector mode. Applied Biosystems Data Explorer™ was used for analysis of monoisotopic [M + H] + ions MS data.

NMR structure determination
All NMR data were recorded using a 300 μl samples, added into a susceptibility-matched 5 mm micro tube (Shigemi Inc. Japan), prepared by dissolving 15 N or 15 N/ 13 C-labeled m 3 -HwTx-IV to a final concentration of 400 μM in 20 mM sodium acetate solution, pH 5 with 5% (v/v) D 2 O. All spectra were acquired at 25 o C using a 900 MHz Bruker Avance NMR spectrometer equipped with a cryogenically cooled probe (cryoprobe). Resonance assignments were obtained using two-dimentional (2D) 1 H-15 N heteronuclear single quantum coherence (HSQC), 2D 1 H-13 C-HSQC, 3D CBCA(CO)NH, 3D HNCACB, 3D HNCO, 3D HBHA(CO)NH and fourdimensional (4D) HCC(CO)NH-TOCSY spectra. 3D and 4D spectra were acquired using nonuniform sampling [24,25] and processed using the Rowland NMR Toolkit (http://rnmrtk.uchc. edu/rnmrtk/RNMRTK.html). Maximum entropy parameters were selected automatically as described previously [26]. 3D 15 N and 13 C (separate spectra for aliphatic and aromatic regions) filtered NOESY-HSQC spectra (200 ms mixing time) were acquired using linear sampling and processed using the Rowland NMR Toolkit. Data were analysed in CcpNmr Analysis v2.4 (University of Cambridge, UK) [27]. NOESY cross peaks were manually picked and supplemented with dihedral angles derived from chemical shifts using the TALOS software [28], and used for structure calculation using CYANA v3.0. [29] The NOEs were automatically assigned by CYANA. 200 structures were calculated and an ensemble of 20 structures with the lowest CYANA target function were selected.
Dynamics of the peptide NMR spin relaxation. NMR experiments for measuring longitudinal (R1) and transverse relaxation rates (R2) and 1 H-15 N steady-state NOEs were acquired using a single-scan interleaved pulse sequence on a 700 MHz Bruker spectrometer equipped with a z-gradient cryoprobe. Data were acquired at 25˚C using a 200 μM sample of 15 N labeled m 3 -HwTx-IV. R1 spectra were recorded with 7 delay times (10 Ã , 50, 300 Ã , 500, 700, 900 Ã , 1100 ms) as were R2 spectra (16 Ã , 32, 64, 96 Ã , 128, 160, 192 ms), where delays marked with an asterisk were recorded twice to confirm data reproducibility. Steady-state 1 H-15 N heteronuclear NOEs were acquired with a relaxation delay of 9 s. Spectra were processed using the Rowland NMR toolkit and analysed using CcpNMR [27]. hNa V 1 activity using patch-clamp electrophysiology Patch-clamp experiments used HEK293 cells expressing either of hNa V 1.1 -hNa V 1.5 and the β1 auxiliary subunit (SB Drug Discovery, Glasgow, UK) or Chinese Hamster Ovary (CHO) cells expressing hNa V 1.6, hNa V 1.7 and hNa V 1.8/β3 (ChanTest Corp). Na + currents were measured by automated whole-cell patch clamp (QPatch 16X; Biolin Scientific A/S, Ballerup, Denmark). The extracellular solution comprised (in mM) 1 CaCl 2 , 1 MgCl 2 , 5 HEPES, 3 KCl, 140 NaCl, 0.1 CdCl 2 and 20 TEA-Cl at pH 7.3 and 320 mOsm, and the intracellular solution comprised (in mM) 140 CsF, 1/5 EGTA/CsOH, 10 HEPES and 10 NaCl at pH 7.3 and 320 mOsm. The elicited currents were sampled at 25 kHz and filtered at 4 kHz. Cells were maintained at a holding potential -80 mV and Na + currents elicited by 20 ms voltage steps to 0 mV (+10 mV for Na V 1.8) from a -120 mV conditioning pulse applied for 200 ms. To obtain concentrationresponse curves, cells were incubated for 5 min with increasing concentrations of toxin. All experimental data was analysed using QPatch Assay Software v5.0 (Biolin Scientific A/S, Ballerup, Denmark). hNa V 1 activity using FLIPR membrane potential assay Human embryonic kidney (HEK) 293 cells heterologously expressing human Na V 1.1-1.8 (SB Drug Discovery) were cultured in minimal essential medium (MEM) containing 10% v/v fetal bovine serum (FBS) and selection antibiotics as recommended by the manufacturer. Cells were grown in a humidified 5% CO 2 incubator at 37˚C, grown to 70-80% confluence, and passaged every 3-4 days using TrypLE Express (Invitrogen). Cells were plated in 384-well clear-bottom black imaging plates (Corning, NY, USA) at a density of 10,000 to 15,000 cells per well. After 48 h, cells were loaded with (20 μl per well) red membrane potential dye (Molecular Devices, Sunnyvale CA, USA) diluted in physiological salt solution [PSS; composition (in mM) 140 NaCl, 11.5 glucose, 5.9 KCl, 1.4 MgCl 2 , 1.2 NaH 2 PO 4 , 5 NaHCO 3 , 1.8 CaCl 2 and 10 HEPES pH 7.4] and incubated for 30 min at 37˚C in a humidified 5% CO 2 incubator. m 3 -HwTx-IV (10 nM-30 μM) was diluted in PSS/0.1% bovine serum albumin (BSA) and preincubated for 5 min before stimulating Na V 1.1-1.7 using veratridine (60 μM) or Na V 1.8 using deltamethrin (150 μM). Changes in membrane potential were assessed using a fluorescence imaging plate reader (FLIPR TETRA , Molecular Devices) (excitation 515 to 545 nm, emission 565 to 625 nm) every second for 300 s after adding agonists. OD1 model. The Na V 1.7 activator OD1 (300 nM) was co-administered alone (control; n = 5) or with m 3 -HwTx-IV at a concentration of 100 nM (n = 4) or 300 nM (n = 3) diluted in saline/0.1% BSA by intraplantar (i.pl.) injection in a volume of 40 μL under light isoflurane (3%) anesthesia in mice as previously described [21]. Mice were then individually placed in polyvinyl boxes (10 x 10 x 10 cm) and allowed to recover from anesthesia. Immediately upon recovery, spontaneous pain behaviours (number of paw lifts, licks, shakes, and flinches) were recorded on video, and later counted by an investigator, unaware of the treatment each individual animal received, for 40 min post-injection at 5-min intervals. Data were plotted and analysed using GraphPad Prism. Statistical significance was defined Ã P < 0.05 and was determined using one-way ANOVA with Dunnett's post-test compared to control.

Production of recombinant m 3 -HwTx-IV
An efficient method was required to provide sufficient amounts of m 3 -HwTx-IV to perform the structural and functional studies. Therefore, we created a construct that enables expression of a His 6 -MBP-m 3 -HwTx-IV fusion protein in the periplasm of E. coli, where the disulfidebond machinery of the cell is located [22]. The fusion protein was purified from the soluble cell fraction using Ni-NTA affinity chromatography ( Fig 1C). m 3 -HwTx-IV was cleaved from the His 6 -MBP fusion tag using TEV protease, then semi-preparative RP-HPLC ( Fig 1D) was employed to yield a single dominant disulfide-bond isoform with a purity of >98% as assessed by analytical RP-HPLC and MALDI-TOF MS ( Fig 1E). The activity of the peptide against hNa V 1.7 was initially confirmed using a FLIPR assay, then by automated patch-clamp electrophysiology as described below. The final yield of recombinant m 3 -HwTx-IV was~0.6 mg per litre of bacterial culture.

Structure of m 3 -HwTx-IV
The structure of m 3 -HwTx-IV was determined using multidimensional heteronuclear NMR spectroscopy. NMR has been the dominant approach for determining the structure of peptides smaller than 5 kDa, including disulfide-rich venom peptides [22,30]. Employing bacterial expression enabled us to produce uniformly 15 N/ 13 C labelled m 3 -HwTx-IV. Two-dimensional (2D) 1 H-15 N HSQC (Fig 2A) and 3D HNCACB, CBCA(CO)NH, and HNCO experiments were used to assign the backbone resonances of m 3 -HwTx-IV. A 4D HCC(CO)NH-TOCSY experiment, which has the advantage of providing sidechain 1 H-13 C correlations [25], was used to provide side-chain assignments. Ultimately, structure calculations were based on interproton distances derived from analysis of 3D 15 N-edited and 13 C-edited HSQC NOESY spectra and backbone dihedral-angle restraints derived from analysis of 1 H, 13 C, and 15 N chemical shifts.
All non-exchanging and slow exchanging hydrogen atoms were assigned, as well as all protonated carbon and nitrogen atoms (BMRB-ID: 30160). These assignments were supplemented with assignments of all C' atoms from the HNCO experiment, which excludes assignments from residues preceding the prolines and the C-terminal residue (N10, I35). The backbone chemical shifts were used to derive 46 dihedral angle predictions (23 ϕ, 23 ψ) by TALOS-N [31]. A total of 1575 NOEs were manually picked from the three NOESY spectra  13 C aromatic, 13 C aliphatic) and 1537 (98%) automatically assigned by CYANA [29]. These were translated to 790 unique distance limits. The disulfide bond connectivities were derived from unambiguous NOE assignments across the disulfide bond [32] and introduced into the structure calculations as three upper and lower distance limits across the disulfide bond The structure calculations generated an ensemble of 20 structures (see Table 1 for statistics, PDB-ID: 5T3M). The ensemble was well defined with low backbone and sidechain RMSDs (0.02 Å and 0.48 Å respectively for residues . This compares favourably with the previously determined structure of the WT peptide (PDB ID 1MB6; RMSDs of 0.34 Å and 1.01 Å for the backbone and sidechain atoms of residues 2-34 [19]). The m 3 -HwTx-IV structure shows that the aromatic ring of W30 is directly above the H β atom of T28, explaining the unusual upfield chemical shift of this atom at 2.48 ppm (compared with the random coil value of 4.16 ppm).

Dynamics of m 3 -HwTx-IV
Heteronuclear spin relaxation is a powerful technique to probe the overall and internal dynamics of macromolecules [33]. In order to study the dynamics of m 3 -HwTx-IV, we determined the 15

Na V 1 subtype selectivity of m 3 -HwTx-IV
The full selectivity profile of m 3 -HwTx-IV at all human Na V channel subtypes (hNa V 1. not. The selectivity profile matched that of the parent (HwTx-IV) peptide, with the exception that activity against the Na V 1.1, Na V 1.6 and Na V 1.8 had previously not been measured. The effects of m 3 -HwTx-IV on Na V channel fast inactivation at 0 mV are shown in S1 Fig for the toxin sensitive channels (tested at~IC 50 concentrations, expect for Na V 1.4 which was assessed at 1 μM). m 3 -HwTx-IV had no substantial effect on the time constant of fast inactivation (τ) of any of the channels tested (S1 Fig).

Discussion
Structure of HwTx-IV and mutants m 3 -HwTx-IV was developed by Revell et al. [14] to improve potency of the parent compound (HwTx-IV) at Na V 1.7. Here we determined whether the improved activity affected the structure or selectivity of HwTx-IV. The structure of WT HwTx-IV was previously solved by two independent groups (PDB IDs 1mb6 and 2m4x for the original and more recent HwTx-IV structures, respectively). Notably, when these two structures are aligned over the backbone atoms in the structured region (residues 5-33), the RMSD was found to be 1.84 Å, which is remarkably high (i.e., indicative of significant differences), considering that both structures were solved using primarily 2D homonuclear NOESY data. To structurally characterise m 3 -HwTx-IV we used a recombinant expression method [22] to isotope label the peptide with 13  heteronuclear 3D/4D NMR methods as well as 15 N relaxation measurments. The structure obtained here using heteronuclear NMR aligns well with the original native peptide structure, 1mb6, with an RMSD of 0.7 Å over the backbone atoms (5-33), but it aligns poorly with the more recent 2m4x structure, with an RMSD of 1.75 Å over the same range of backbone atoms. The large difference between our structure and the more recent of the two structures of the WT peptide is apparent in 3 of the 4 loops of the peptide as well as the termini. It is unclear why the 2m4x structure deviates to such an extent from the other two structures. We note, however, that for the 1mb6 structure the authors recorded separate 2D NOESY spectra in H 2 O and D 2 O and it may be that this approach provided additional 1 H NOEs from Hα atoms in the β-sheet regions which are often obscured in 2D NOESY spectra by the strong H 2 O signal. In our 3D filtered NOESY data these NOEs are obscured along the direct dimension but can often be assigned along the indirect dimension. Given the discrepancy, and the additional experimental data used in the 1mb6 structure, we have below only made comparisons to 1mb6.

Structural consequences of mutations
In the m 3 -HwTx-IV structure the position of W33 with respect to K21 is well defined, whereas in the WT structure the aromatic ring of Y33 is poorly defined and predominantly positioned facing the sidechains of K32 and F6. In our structure there are clear NOEs between the sidechain H β protons of K21 and the H δ 1 and H ε 1 protons of W33 in the 3D 13 C aromatic NOESY, 3D 13 C aliphatic NOESY and the 3D 15 N NOESY spectra. The chemical shifts of these atoms are unique in the 3D NOESY spectra and the NOE assignments are therefore unambiguous, strongly supporting the alternative sidechain conformation in our model (Fig 6).
The NMR relaxation studies revealed no evidence of structural or conformational flexibility in the C-terminal tail, except for the last residue, which appears less ordered. It is therefore unclear wether the W33/K21 interactions are not present in the WT peptide or whether these interactions were simply not resolved in the previous homonuclear NMR studies. We note that in the 2D NOESY data the H β atoms of K21 are in a crowded region, which may have obscured NOEs in the 1mb6 data. It had previously been speculated that cation-π interactions may serve to reduce the loss in energy upon binding due to the desolvation of lysine residues. Here, the conformation of these two residues would not provide such stabilisation, as the charged NH 3 + group of the lysine is instead proximal to the carboxyl-terminus of the peptide.
The other residues that were mutated in m 3 -HwTx-IV are two glutamic acid residues in the N-terminal section of the peptide. Curiously the E4 residue appears to form a salt bridge with K7 in the WT toxin, and it might be expected that this charge pairing stabilises the fold of the peptide. However, in our structure K7 assumes a nearly identical position as in the WT peptide whilst there are no major changes to the conformation of loop 1. Thus, the improved Na V 1.7 activity from mutating these acidic residues to Gly likely results from removal of unfavourable interactions with either VSD II of Na V 1.7 or the surrounding membrane. Indeed, previous work has shown that loop 1 residues are important in lipid binding and it is possible that the acidic residues either interact unfavourably with anionic components of the membrane leaflet or reduce the permeability of the peptide in the lipid bilayer [35].
The C-terminus of HwTx-IV is amidated and previous structure/activity studies have shown that removal of this amidation causes a 20-50 fold reduction in activity at Na V 1.7 [14,36]. Since m 3 -HwTx-IV was produced recombinantly, the C-terminus is not amidated and we would expect the activity of the peptide to be less than that of the amidated peptide. Indeed we found that the recombinant peptide inhibited Na V 1.7 with an IC 50 of 3.3 nM, which is 8-fold higher than that previously reported for the synthetic amidated version but still more potent than WT HwTx-IV (IC 50 26 nM). Thus, m 3 -HwTx-IV is one of the most potent recombinantly produced Na V 1.7 blockers reported, and in contrast to ProTx-II, it does not require refolding [37].

Dynamics of the pharmacophore region
The Na V pharmacophore in this class of peptides has been extensively characterised and it is clear that the W30/K32 pair are the most functionally critical residues [14,20]. In our structure these residues assume a conformation similar to that in WT peptide, and we conclude that the three mutations do not affect the overall configuration of these critical residues. Instead, our relaxation data show clear evidence of conformational mobility in loop 4 residues including W30. Remarkably, we find that the largest discrepancy between our mutant structure and the WT structure is in this loop, and when this loop is omitted in the structural alignment the RMSD falls from 0.7 Å to 0.4 Å. This suggests that this loop is flexible and that the difference between the structures may simply reflect different conformational states of loop 4 (see also Fig 2B). Given that this dynamic region coincides with the active site of the peptide we speculate that this mobility provides additional conformational states that may stabilise binding of the peptide to VSD II . These data also suggest that the absolute configuration of the pharmacophore residues in the solution structure of HwTx-IV may differ from that of the channelbound peptide. Should this be the case, the solution structures may not be suitable design templates for development of peptide mimetics or small molecules inhibitors, unless sufficient flexibility is introduced with respect to the position of these key functional residues.

Subtype selectivity
The binding site of HwTx-IV was characterised in detail by Xiao et al. [16]. The authors found that HwTx-IV inhibits rat (r) Na V 1.2 and rNa V 1.3 as well as human Na V 1.7, but is far less potent on rNa V 1.4 and rNa V 1.5. Based on these results and careful mutagenesis work, Xiao et al. were able to identify residues on Na V s that conferred selectivity [17]. They showed that the peptide binds to VSD II and that one residue on the S1-S2 loop (E753), together with four residues on the S3-S4 loop (E811, L814, D816, E819), are particularly important for peptide binding. This recognition motif "EELDE" is conserved in Na V 1.6, whilst only differing in a Dto-N change (EELNE) in Na V 1.1, Na V 1.2, Na V 1.3. In contrast, Na V 1.4 and Na V 1.5 contain two changes (EELNQ and EELRS respectively), and Na V 1.8 contains three changes (EEVKG) with Na V 1.9 having the biggest difference with four changes (EDVQR).
From the above analysis it should therefore be expected that Na V 1.6 should be equally sensitive to m 3 -HwTx-IV, and since Na V 1.2 is sensitive to the toxin, we would expect Na V 1.1, which has the same recognition sequence, to also be m 3 -HwTx-IV sensitive. Finally, based on the sequence analysis, Na V 1.8 and Na V 1.9 are least likely to be sensitive to the toxin. Using a FLIPR membrane potential assay we compared subtype selectivity profiles across Na V isoforms and as predicted by the work in the Cummins group [17,38] the previously uncharacterised hNa V 1.6 and hNa V 1.1 are indeed sensitive to the toxin whilst hNa V 1.8 is, again as predicted, insensitive to the toxin. To further characterise the selectivity of m 3 -HwTx-IV we measured its activity against Na V 1.1-1.8 using whole cell patch-clamp electrophysiology. We found that the peptide is very potent on Na V 1.1-1.3, Na V 1.6 and Na V 1.7. Thus, it appears that the D-to-N mutation at position 816 is well tolerated. Comparing the activity on Na V 1.4/5 and Na V 1.1 we can conclude that final E-to-Q/S mutation is not tolerated and the presence of a negative charge is essential in this position. The full range of Na V subtypes sensitive to HwTx-IV only differs from tetrodotoxin (TTX) in its activity on Na V 1.4. This may prove useful as differential inhibition of sodium currents by HwTx-IV and TTX can be attributed to Na V 1.4.
Our results show that, despite claims of improved selectivity and potency, the subtype selectivity of m 3 -HwTx-IV is not significantly altered compared with native HwTx-IV. From the sequence analysis our data would support a model where the peptide binds to residues within the cavity formed between the four helices of the VSD II [39]. This model would suggest that there is a large interface between the peptide and the channel, indicating that it should in theory be possible to develop subtype selective peptides. Current peptide engineering efforts, however, rely on exhaustive mutagenesis of the peptide and have unfortunately shown limited success, with the common observation that improvements in affinity against one of the sensitive subtypes often leads to a correlated increase in affinity against other subtypes [40]. To overcome these challenges further structural details of the VSD II in its peptide bound form are required to enable rational peptide design.

In vivo activity
The in vivo activity of HwTx-IV was previously reported in rat models of formalin-induced pain and spared nerve injury, where it blocked pain behaviours at doses between 25 and 200 μg/kg [41]. This exquisite in vivo potency may at least in part be due to the strain of animals used, as morphine also was effective at doses at least 20 times lower than those usually required to achieve analgesia in the same models [42][43][44][45]. Given the poor selectivity of HwTx-IV, in particular over Na V 1.6, we used a recently developed Na V 1.7 target engagement assay to assess the biological activity of recombinant m3-HwTx-IV following local intraplantar injection, as systemic administration of peptides with activity at Na V 1.6 typically leads to adverse side-effects. Intraplantar injection of OD1 at the concentration used here evokes spontaneous pain behaviours that are reduced by~70% in Na V 1.7 knockout mice or after treatment with the selective Na V 1.7 inhibitor PF-04856264, with residual pain responses likely attributable to Na V 1.6 [21]. Accordingly, the observed in vivo effects of m3-HwTx-IV are entirely consistent with our pharmacological characterization of m3-HwTx-IV showing potent effects at Na V 1.6 and Na V 1.7, and indicate target engagement of Na V 1.7 and to lesser extent Na V 1.6 on peripheral nerve terminals in vivo. The in vivo effects of m3-HwTx-IV were only assessed in male mice, as sex-related differences in Na V 1.7-mediated pain are not expected [46].

Conclusions
We report here an efficient method of recombinant production of a highly potent Na V 1.7 inhibitor, m 3 -HwTx-IV. Our high-resolution solution-state NMR structure revealed that the mutations introduced into HwTx-IV do not alter the structure of the peptide. Thus, with perhaps the exception of the Y33W mutation, the gain in potency for the mutant peptide appears to be due to removal of unfavourable contacts from the acidic moieties (E1G, E4G). Furthermore, we provide the first NMR relaxation studies of this family of sodium channel inhibitors, which revealed significant conformational flexibility in the critical pharmacophore region. Detailed functional studies showed that the subtype selectivity of m 3 -HwTx-IV differs from TTX only at Na V 1.4, which was not sensitive to m 3 -HwTx-IV. Finally, we found that m 3 -HwTx-IV provides analgesia in a mouse model of Na V 1.7-mediated pain. While recombinant m 3 -HwTx-IV is likely to be a useful pharmacological tool for structural and pharmacological studies, its off-target activity makes it a poor clinical candidate. Further structural details of the peptide-channel complex are likely required to enable design of HwTx-IV analogues with improved selectivity profiles.
Supporting information S1 Fig. m 3 -HwTx-IV has no effect on Na V channel fast inactivation when tested near the IC 50 value. Na + currents were elicited by a 20 ms depolarising pulse to 0 mV from a 200 ms pre-pulse at -120 mV. Holding potential was -80 mV. The time constant of fast inactivation (τ) was calculated from a fit of the inactivation phase to a mono exponential equation and is presented as mean +/-SEM (n = 3 repetitions). (DOCX)