Pharmacological characterization of potent and selective NaV1.7 inhibitors engineered from Chilobrachys jingzhao tarantula venom peptide JzTx-V

Identification of voltage-gated sodium channel NaV1.7 inhibitors for chronic pain therapeutic development is an area of vigorous pursuit. In an effort to identify more potent leads compared to our previously reported GpTx-1 peptide series, electrophysiology screening of fractionated tarantula venom discovered the NaV1.7 inhibitory peptide JzTx-V from the Chinese earth tiger tarantula Chilobrachys jingzhao. The parent peptide displayed nominal selectivity over the skeletal muscle NaV1.4 channel. Attribute-based positional scan analoging identified a key Ile28Glu mutation that improved NaV1.4 selectivity over 100-fold, and further optimization yielded the potent and selective peptide leads AM-8145 and AM-0422. NMR analyses revealed that the Ile28Glu substitution changed peptide conformation, pointing to a structural rationale for the selectivity gains. AM-8145 and AM-0422 as well as GpTx-1 and HwTx-IV competed for ProTx-II binding in HEK293 cells expressing human NaV1.7, suggesting that these NaV1.7 inhibitory peptides interact with a similar binding site. AM-8145 potently blocked native tetrodotoxin-sensitive (TTX-S) channels in mouse dorsal root ganglia (DRG) neurons, exhibited 30- to 120-fold selectivity over other human TTX-S channels and exhibited over 1,000-fold selectivity over other human tetrodotoxin-resistant (TTX-R) channels. Leveraging NaV1.7-NaV1.5 chimeras containing various voltage-sensor and pore regions, AM-8145 mapped to the second voltage-sensor domain of NaV1.7. AM-0422, but not the inactive peptide analog AM-8374, dose-dependently blocked capsaicin-induced DRG neuron action potential firing using a multi-electrode array readout and mechanically-induced C-fiber spiking in a saphenous skin-nerve preparation. Collectively, AM-8145 and AM-0422 represent potent, new engineered NaV1.7 inhibitory peptides derived from the JzTx-V scaffold with improved NaV selectivity and biological activity in blocking action potential firing in both DRG neurons and C-fibers.

Introduction Melusine ref. MLU_020110). Venom samples were collected, lyophilized, and dissolved in 0.1% trifluoroacetic acid (TFA) in water to approximately 1 mg venom/mL. The crude venom solutions were desalted by solid-phase extraction (SPE) with Sep-Pak C18 cartridges (Waters, Milford, MA, USA) equilibrated in 0.1% TFA and eluted with 60% aqueous acetonitrile and then evaporated. The dried material was dissolved in 0.1% TFA to approximately 1 mg venom/mL concentration, and higher molecular weight components were removed with a 10 kDa molecular weight cut-off Ultrafree-CL centrifugal filter (Millipore). The <10 kDa venom extract was then dried under vacuum and stored at -80˚C.
The venom fractions were screened for activity in a Na V 1.7 IonWorks Quattro (IWQ) assay. Several fractions with significant (>50% of control) Nav1.7 inhibitory activity were identified. A second aliquot of the fractions was tested against Na V 1.7, Na V 1.4, and Na V 1.5 IWQ assays to confirm the activity of the initial hit and evaluate selectivity. The most potent and selective fraction, fraction 36, was analyzed by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) using an α-cyano-4-hydroxycinnamic acid matrix. The concentration of peptide in this fraction was not determined. The major signal indicated a molecular weight (3602.34 Da) consistent with the known Na V inhibitory peptide toxin JzTx-V (3602.65 Da, UniProt Accession Number Q2PAY4) [25]. To confirm the identity and biological activity of JzTx-V, the peptide was chemically synthesized and tested.

Peptide synthesis, oxidative folding, and purification
Rink Amide Chem Matrix resin (0.2 mmol, 0.45 mmol/g loading, 0.444 g, Matrix Innovation) was weighed into a CS BIO reaction vessel. The reaction vessel was connected to a channel of the CS BIO 336X automated peptide synthesizer, and the resin was washed twice with DMF and allowed to swell in DMF for 15 min. Fmoc-amino acid (1.0 mmol, Midwest Biotech or Novabiochem) was dissolved in 2.5 mL of 0.4 M 6-chloro-1-hydroxybenzotriazole (6-Cl-HOBt, Matrix Innovation) in DMF. To the solution was added 1.0 mL of 1.0 M 1,3-diisopropylcarbodiimide (DIC, Sigma-Aldrich) in DMF. The solution was agitated with nitrogen bubbling for 15 min to accomplish pre-activation and then added to the resin. After the mixture was shaken for 2 h, the resin was filtered and washed 3 x DMF, 2 x DCM, and 3 x DMF. Fmocremoval was accomplished by treatment with 20% piperdine in DMF (5 mL, 2 x 15 min, Fluka). The resin was filtered and washed 3 x DMF. All residues were single coupled through repetition of the Fmoc-amino acid coupling and Fmoc removal steps described above.
After final Fmoc-removal from the N-terminal residue, resin-bound linear peptide (0.2 mmol scale) was transferred to a 25 mL SPE filter tube, washed 3 x DMF and 3 x DCM, and dried under vacuum. To the resin were added triisopropylsilane (1.0 mL), 3,6-dioxa-1,8octane-dithiol (DODT, 1.0 mL), water (1.0 mL), TFA (15 mL), and a stir bar, and the mixture was stirred for 3 h. The mixture was filtered into a 50 mL centrifuge tube. The resin was washed with TFA (~5 mL), and the combined filtrate was concentrated by rotary evaporation in a Genevac HT-12 (30˚C chamber temperature, pressure ramp from 500 to 50 mbar over 40 min and a final pressure of 8 mbar for 2 h). To the residue (~5 mL) was added 40 mL cold diethyl ether, leading to the formation of a white precipitate. The white ether suspension was stirred and centrifuged (4 min, 4,400 rpm), and the ether was decanted. To the tube was added another 40 mL of cold ether, and the precipitate was stirred. The mixture was centrifuged, and the ether was decanted. The solid was dried overnight under vacuum. The crude linear peptide was purified by preparative LC-MS. The filtered sample (300 mg in 5 mL DMSO with 0.1 M tris(2-carboxyethyl)phosphine hydrochloride (TCEP, Sigma-Aldrich)) was injected onto a preparative HPLC column (Phenomenex Synergi 4um MAX-RP 80A AXIA, 250 x 30 mm). The peptide was eluted with a 10-40%B over 60 min gradient at 30 mL/min, followed by a 10 min flush and a 10 min equilibration. The fractions were analyzed by LC-MS, pooled, and lyophilized to afford the pure linear peptide precursor.
In a 1L polypropylene bottle was prepared a folding buffer with water (800 mL), acetonitrile (100 mL), cysteine (1 mL of a 1 M stock solution in water), and cystine dihydrochloride (6.667 mL of a 150 mM stock solution in water). To the pure linear peptide (100 mg) was added 5 mL acetonitrile and 5 mL water. The mixture was vortexed and sonicated to complete dissolution of the peptide. The peptide solution was added to the buffer followed by 1M Tris-HCl pH 8.0 (100 mL), (0.1 mg/mL peptide concentration, 1 mM cysteine, 1 mM cystine, 10% v/v acetonitrile, 0.1 M Tris pH 8.0). The pH value was measured to be 8.0. The folding mixture was allowed to stand at 4˚C for 18 to 72 h. A small aliquot was removed, and the sample was analyzed by LC-MS to ensure that the folding was complete. The solution was quenched by the addition of 4 mL acetic acid (HOAc) and 4 mL TFA (pH = 2.5). The aqueous solution was filtered (0.45 μM cellulose membrane).
The filtered solution (1000 mL, 100 mg peptide) was loaded onto a preparative HPLC column (Phenomenex Synergi 4um MAX-RP 80A AXIA, 250 x 30 mm) at 30 mL/min using an Agilent preparative loading pump. The column was flushed for 10 min with 10% B at 30 mL/ min to elute the HOAc/TFA. The column was attached to a preparative HPLC, Agilent/LEAP prep LC-MS, and the peptide was eluted with a 10-40% B gradient over 60 min, followed by a 10 min flush and a 10 min equilibration. The fractions were analyzed by LC-MS, pooled, and lyophilized to afford pure folded peptide.
Final LC-MS analysis (Phenomenex Jupiter 20 × 2 mm, 100 Å, 5 micron column eluted with a 10 to 60% B over 10 min gradient at a 0.750 mL/min flow rate monitoring absorbance at 220 nm) was performed. Peptides with > 95% purity and correct (m/z) ratio were screened.

NMR structural analysis
The structures of peptides 3 and AM-8145 were obtained by high resolution NMR spectroscopy using Bruker Avance III 600 MHz and 500 MHz NMR systems both utilizing a TCI cryoprobe (Billerica, MA). Samples were prepared in 20 mM sodium phosphate, pH 7.0 containing 50 mM NaCl and 10% D 2 O. The temperature of all NMR spectra were optimized (320K for peptide 3 and 325K for AM-8145) to minimize effects of exchange broadening and facilitate signal resolution.
The assignment of chemical shifts correlating to the backbone and side-chain (all protons and Cα/Cβ only) resonances of both peptides were determined using standard methods from a combination of the two-dimensional NOESY, TOCSY, HSQC, and HMBC spectra. NOEbased distance restraints were determined from NOESY spectra utilizing mixing times of 500 ms. Additional NOEs were obtained from the ω 1 -decoupled TOCSY-NOESY experiment that provided dipolar correlations between cysteine residues [28]. These restraints were incorporated manually. Stereospecific proton assignments were obtained based on the coupling patterns between C 0 -Hβ and Hα-Hβ evaluated from 13 C-HSQC and HMBC experiments [29]. Dihedral angle constraints for the χ 1 angles (N-Cα-Cβ-S) were determined, when possible, using the NOE patterns between Hα-Hβ and Hα-Hβ 0 , along with the stereospecific assignments and 3 J HαHβ and 3 J HαHβ 0 coupling constants measured from the high resolution NOESY or TOCSY spectra.
Structure determination for both peptides proceeded by manually assigning NOEs using standard methods. The NOE assignments and structure calculations were carried out from extended strand starting structures using Cyana 3.0 [30] and performed iteratively along with interactive assignment by the user. In each of seven rounds, 100 structures were calculated using 50,000 steps of torsion angle dynamics. Once structures had begun to converge, hydrogen bond restraints were incorporated manually. At the final stage, disulfide bond constraints were added according to the determined connectivity patterns. χ 1 angle constraints were added based on ranges determined using the Karplus equation [31] and were incorporated with relatively loose restraints (60 ± 30˚). Structures were re-calculated as above, and the 10 structures with the lowest target functions were selected for analysis.

Radioligand binding assay
125 I-ProTx-II was purchased from PerkinElmer Life Sciences with a specific activity of 2,200 Ci/mmol. HEK293 cells stably expressing human Na V 1.7 were seeded at 75,000 cells/well in 96-well poly-D-lysine-coated plates (Corning) and allowed to attach for 16-20 h at 37˚C. Prior to assay, cell culture medium was replaced with assay buffer containing DMEM/F12 (Gibco) with 0.1% fetal bovine serum (Gibco). For saturation binding experiments, cells were incubated with increasing concentrations of 125 I-ProTx-II for 3 hours and then washed twice with 200 μl of assay buffer. 100 μl of liquid scintillation fluid (Microscint-40, PerkinElmer Life Sciences) was added, and radioactivity was counted using a TopCount TM microplate scintillation counter (PerkinElmer). Triplicate samples were averaged for each experimental data point. Total binding was defined as counts in the presence of 125 I-ProTx-II alone, nonspecific binding was defined as counts remaining when 125 I-ProTx-II was co-incubated with cold 1 μM AM-8145, and specific binding was obtained by subtracting nonspecific binding from total binding counts. K d and B max values were calculated using Graphpad Prism 7 (GraphPad Software Inc.). For competition binding experiments, 0.5 nM 125 I-ProTx-II was incubated with cells for 3 h in the absence or presence of increasing concentrations of ProTx-II, HwTx-IV, GpTx-1 and JzTx-V peptides. Inhibition constant (K i ) values were calculated using GraphPad Prism 7.

Electrophysiology
Human Na V (hNa V ) stably transfected cell lines in HEK293 or CHO backgrounds and cell culture methods were as previously described [32]. hNa V 1.1-hNa V 1.7 cells were obtained from Eurofins Pharma Discovery Services (St. Charles, MO), and hNa V 1.8 cells were generated in house. Whole cell patch clamp, PatchXpress, and IWQ electrophysiology voltage protocols as well as DRG neuron isolation were carried out as previously described [24]. Na V 1.7-Na V 1.5 chimeras were generated and evaluated in BacMam transduced U2-OS cells on the IWQ platform as previously described [23].

Multi-electrode array (MEA)
Action potentials of cultured postnatal day 2 rat DRG neurons (Lonza, Walkersville, MD) were detected using the Maestro multi-well plate-based MEA system (Axion Biosystems, Atlanta, GA). A 48-well MEA plate containing 16 electrodes/well was pre-coated with 10μL 0.1% PEI (Sigma-Aldrich, St. Louis, MO) in borate buffer (50mM boric acid (Fisher Scientific, Waltham, MA), 12mM sodium tetraborate (Sigma-Aldrich), pH 8.4). The plate was incubated at 37˚C for 2h, washed 3-times with 200μL sterile ddH 2 O and then air dried. Next, 10μL of 20μg/mL laminin (Sigma) in Primary Neural Basal Medium (PNBM) (Lonza) was spotted over the PEI-treated electrode surface and incubated for 2h at 37˚C. Laminin was removed and immediately replaced with 10μL/well of 7x10 6 cells/mL rat DRGs in PNBM. After 1h at 37˚C, 300μL of PNBM (supplemented with 2 mM L-glutamine (Lonza), 1% gentamycin/ amphotericin (Lonza), 2% NSF-1 (Lonza), 100 ng/mL NGF-β (Sigma-Aldrich), and an additional 50mM NaCl to yield a final NaCl concentration of~127mM) was added and cells were maintained at 37˚C, 5% CO 2 for a total of 8-9 days, with a fresh media exchange after 4 days. Compound inhibition of capsaicin-induced action potential firing was measured at 37˚C. Baseline spontaneous firing was recorded for approximately 1 min. Twenty-five μL of PNBM, 325nM capsaicin (MP Biomedicals, 25nM final, EC 80 ) in PNBM, or 325nM capsaicin plus test compound was then added, and action potentials were continuously recorded for another 2 min with six replicate wells used per condition. MEA data were analyzed using Axion Integrated Studio AxIS2.1 (Axion Biosystems) and GraphPad Prism version 7.02 for Windows (GraphPad Software, La Jolla, CA).

Ethics statement
Experimental procedures were approved by Amgen Inc.'s Institutional Animal Care and Use Committee in accordance with the National Institutes of Health's Guide for the Care and Use of Laboratory Animals and conducted at an AALAC-accredited facility.

Skin-saphenous nerve preparation
Male C57BL/6 mice (Charles River Laboratories, San Diego, CA) between 2-3 months old (136 total mice used in study) were housed under standard temperature conditions (22-24˚C) and illumination (12-hour light/dark cycle with lights on at 6:30 AM) with ad libitum access to fresh water and food. Mice were euthanized by CO 2 followed by cervical dislocation. Experimental procedures for mouse skin-saphenous nerve studies were conducted as previously described with the use of 0.1% BSA (Acros) in the bath solution to decrease non-specific loss of peptide to chamber components [19] [32]. Briefly, the saphenous nerve and skin from the left medial dorsum of the hind paw were dissected and mounted in a custom-made organ bath chamber. The proximal saphenous nerve end was passed through a hole into a separate recording chamber, desheathed and teased into fine fibers for extracellular recording. Mechanical responses were evoked by square force waves using a mechanical stimulator before and following peptide application to a small area of the corium (dermis) side of the skin in a custommade stainless steel ring. Recording and analysis of C-fiber-mediated action potentials were performed in a blinded manner with respect to peptide treatment.

Screening venom fractions to identify JzTx-V
Our previous experience with GpTx-1 had shown spider venoms from the Theraphosidae family were enriched in Na V inhibitory peptides [24]. We acquired fractionated venoms from a number of tarantula species to add to our natural peptide collection and screened them for Na V 1.7 inhibitory activity with the intention of identifying a more potent starting lead molecule. Over eighty percent inhibition of Na V 1.7 current was observed for fraction 36 from the Chinese earth tiger tarantula C. jingzhao (S1 Fig JzTx-V is a 29 amino acid polypeptide with a C-terminal amide and 6 cysteine residues engaged in 3 disulfide bonds to form an ICK motif and is a member of NaSpTx family 3 ( Fig 1A) [6].

Potency and selectivity of JzTx-V
JzTx-V was originally described as a non-selective inhibitor of Na V channels in rat DRG neurons [25]. Synthetic JzTx-V (HPLC profile shown in S3 Fig) was evaluated against human Na V 1.7 heterologously expressed in HEK293 cells on a PatchXpress automated electrophysiology platform, using a voltage protocol in which 20% of channels were in the inactivated state, and yielded an IC 50 of 0.63 ± 0.17 nM (n = 4). The potency of JzTx-V against Na V 1.4 revealed 3-to 4-fold selectivity over Na V 1.7 (Na V 1.4 IC 50 = 2.2 ± 0.4 nM, n = 3), and the potency of JzTx-V against Na V 1.5 revealed nearly 4,000-fold selectivity over hNa V 1.7 (Na V 1.5 IC 50 = 2,350 ± 480 nM, n = 3). Manual patch clamp electrophysiology studies were conducted to evaluate the mechanism of action for JzTx-V channel blockade of hNa V 1.7. The potency of JzTx-V inhibition of Na V 1.7 was 0.15 ± 0.05 nM (n = 2) by manual patch, using the same voltage protocol as above; this value is slightly lower than obtained on the PatchXpress platform and likely due to improved cell perfusion. JzTx-V inhibition of hNa V 1.7 in the resting/closed state (0.3 nM JzTx-V blocked 83 ± 2% current at a holding potential of -140 mV) or a partially-inactivated state (0.3 nM JzTx-V blocked 83 ± 6% current at a holding potential of -80 mV) was comparable, indicating peptide block was not state-dependent across these voltages and proceeded via interaction with a closed state (Fig 1B). High frequency strong depolarizations to +100 mV partially reversed JzTx-V block of Na V 1.7, indicating lower peptide affinity for the channel open state(s) and displacement of the peptide from its binding pocket upon the closed to open gating state transition (Fig 1C).

Na V isoform selectivity engineering to discover AM-8145 and AM-0422
Since the selectivity of native JzTx-V for Na V 1.7 over Na V 1.4 was only 3-4 fold, we set out to improve Na V 1.4 isoform selectivity by the single residue mutation attribute-based positional scanning paradigm we previously described [22]. Alanine scanning mutagenesis of all noncysteine residues via chemical synthesis and refolding was performed and the resulting peptides were tested against Na V 1.7, Na V 1.4 and Na V 1.5 using the IWQ platform. The resulting IC 50 data identified key residues for Na V 1.7 block, exemplified by Trp5, Leu19, Trp24 and Arg26 (Fig 2A, S1 Table). Similar to the parental JzTx-V peptide, Ala-mutants did not block Na V 1.5 function. However, none of the Ala-mutants conferred significant selectivity over Na V 1.4. Attribute-based positional scanning of tarantula toxin GpTx-1 showed maximum disruption of Na V activity with the negatively charged glutamic acid residue [22]. Therefore, we prepared and tested Glu-mutants of JzTx-V as above. The Na V 1.7 IC 50 data showed Met6, Thr8, Asp10, Arg13 and Leu23 were additionally involved in the interaction with Na V 1.7 (Fig  2A, S1 Table). Interestingly, Glu-scanning mutagenesis revealed a significant advance in generating selective Na V 1.7 inhibitors from the JzTx-V scaffold in the form of the Ile28Glu mutation that showed good selectivity over Na V 1.4. Peptide 1, [Glu28]JzTx-V(1-29), potently blocked Na V 1.7 (IC 50 = 0.6 nM), was 500-fold selective against Na V 1.4 (IC 50 = 301 nM) and was a weak blocker of Na V 1.5 (IC 50 = 8,800 nM) on the PX platform.
We next sought to identify suitable locations on the JzTx-V sequence where we could attach handles for potential derivatization by incorporating an alkyne side-chain containing residue in the sequence, e.g. propargylglycine (Pra) [23]. The Glu-scan data suggested specific positions where such mutations would have minimum impact on Na V 1.7 potency (exemplified by Glu-scan analog Na V 1.7 IC 50 's < 50 nM, represented in green in Fig 2A). The N-terminus, positions 1, 11, 14 and 17 were selected for the incorporation of Pra. For this study, we replaced Met6 with the isosteric residue norleucine, to avoid complications from side-chain sulfide oxidation during synthesis and folding. The Pra analogs of [Glu28]JzTx-V(1-29) were tested against Na V 1.7, Na V 1.4 and Na V 1.5 (Table 1). Adding a Pra-residue to the N-terminus gave peptide 2 (AM-8145), which was equipotent to WT JzTx-V and demonstrated excellent selectivity over Na V 1.4 (300-fold) and Na V 1.5 (6,000-fold). For an exact comparison, the peptide was remade without the key selectivity-determining Glu28 residue. Peptide 3, which has the native Ile residue at position 28 and thus differs from AM-8145 by only one amino acid,

JzTx-V analog NMR structure
The NMR structures of JzTx-V peptide analogs were determined with and without the Na V 1.4 selectivity-conferring Glu28 residue to evaluate the impact of this modification on peptide conformation. The structure of peptide 3, Pra-[Nle6]JzTx-V(1-29) (PDB 6CHC), containing the wild-type Ile28 residue was determined at 320K (RMSD for residues 5-30 was 0.45 ± 0.07Å) and confirmed the ICK folded motif with disulfide connectivity identified between Cys2-Cys16, Cys9-Cys21 and Cys15-Cys-25 ( Fig 2B). Super-imposing averaged Ala-and Gluscan analog heat map data onto the NMR structure of peptide 3 revealed residues with the greatest loss in Na V 1.7 potency were clustered on one face of the folded peptide, the purported ion channel binding face (Fig 2C). The back or reverse face of the peptide is comprised of residues shown to be replaceable without compromising Na V 1.7 inhibition, and appear to be reasonable locations for Pra insertions for further derivatization.
The NMR structure of the Glu28 analog of peptide 3, AM-8145 (PDB 6CGW), was solved at 325K (RMSD of 0.71 ± 0.14 Å, Fig 2D). The signals from residues Lys12-Arg13-Ala14 were broadened suggesting local disorder and thus complete assignment of that region was not possible. The Ile28Glu modification caused a change in the structure of the C-terminal region of the peptide (Fig 2E). The acidic Glu28 side chain is reoriented towards the basic Arg20 functionality. In the AM-8145 structure, Leu19 packs into the core of the peptide adjacent to Trp5 and towards the purported ion channel binding face. We also observed a modest reorientation of the sidechains of Trp24 and Arg26 within the mapped binding surface. Residues critical for activity continue to cluster on one face of the peptide; though the overall shape of the presumed binding face is changed. Significant changes in the N-terminal conformation and Asp10-Ala14 loop were also observed; however, these changes were in a region of the AM-8145 peptide that showed local disorder and was tolerant to substitution, as mutation to Ala or Pra had minimal effect on activity (peptides 5-7). The conformational change induced by the Ile28Glu substitution and subtle alteration of the presumed binding face may help rationalize the Nav1.4 selectivity gains achieved with this amino acid change.
JzTx-V binding to Na V 1.7 cells Na V 1.7 inhibitory peptide binding to live HEK293 cells stably expressing Na V 1.7 was evaluated. Using the radiolabeled peptide 125 I-ProTx-II and cold peptide AM-8145 (1 μM), specific interaction with HEK293-Na V 1.7 but not parental HEK293 cells was observed (Fig 3A). 125 I-ProTx-II selectivity to Na V 1.4. Cys residues were not mutated. B. NMR structural ensemble of peptide 3 containing the Ile28 residue. Disulfides are in yellow, the N-terminal region in blue and the C-terminal region in red. C. Average of Ala-and Glu-mutant Na V 1.7 IC 50 's mapped onto a space-filling model of peptide 3 based on the NMR structure, showing key residues that lose Na V 1.7 activity upon modification cluster on one face of the peptide. Colors correspond to IC 50 ranges in panel A. Residues tolerant to modification are in green and located on the reverse side of the putative binding face. The Ile28 residue is colored magenta. D. NMR structural ensemble of peptide AM-8145 containing the Glu28 residue that imparts selectivity over Na V 1.4. Colors are as in panel B. E. Comparison of key amino acid side chains from peptide 3 and AM-8145 NMR structures. The Glu28 modification induced a change in overall conformation, with the acidic Glu28 side chain pointing down towards the basic Arg20 in AM-8145. https://doi.org/10.1371/journal.pone.0196791.g002 Engineered JzTx-V peptide inhibitors of Na V 1.7 binding had a K d of 0.61 ± 0.2 nM (n = 3) and B max of 540 ± 62.8 (2,700 sites per cell) in saturation binding experiments (Fig 3B). Using this assay, a panel of cold Na V 1.7 inhibitory peptides was evaluated, including the tarantula venom peptides ProTx-II, HwTx-IV and GpTx-1. These diverse peptides all competed for 125 I-ProTx-II binding, indicating interaction with the same Na V 1.7 binding site (Fig 3C). Extension of this panel to encompass a collection of JzTx-V peptide analogs, including AM-0422 and AM-8145, revealed a similar rank order between binding (K i ) and functional block by electrophysiology on the PatchXpress platform (IC 50 ) (Fig 3D). Collectively, these data demonstrate that JzTx-V peptides specifically bind to Na V 1.7 channels and that peptide binding reflects channel block as assessed by electrophysiology.

AM-8145 chimera mapping to the Na V 1.7 second voltage-sensor domain
Experiments were conducted to map the binding site of JzTx-V peptide AM-8145 on Na V 1.7. Given the lack of robust bioactivity of AM-8145 on Na V 1.5, we evaluated AM-8145 functional block on a battery of human Na V 1.7-Na V 1.5 chimeric channels, in which individual voltage sensor domains (transmembrane domains S1-S4 of domains I-IV) or the entire pore domain (transmembrane domains S5-S6 of domains I-IV) of Na V 1.7 were inserted into the Na V 1.5 backbone. As shown in Fig 4, AM-8145 potently blocked wild-type Na V 1.7 (IC 50 = 0.05 ± 0.01 μM) but not wild-type Na V 1.5 (IC 50 > 3 μM) when evaluated on the IWQ platform. Chimeras with all four Na V 1.7 voltage-sensor domains (IC 50 = 0.18 ± 0.03 μM) or only the second Na V 1.7 voltage-sensor domain (IC 50 = 0.10 ± 0.01 μM) showed potent channel block by AM-8145 similar to wild-type Na V 1.7, whereas chimeras with only the first, third, or fourth Na V 1.7 voltage-sensor domain did not exhibit potent block (IC 50 > 3 μM). A chimera with the Na V 1.7 pore domain exhibited weak block by AM-8145 with an IC 50 value that was right-shifted over 20-fold (IC 50 = 1.2 ± 1.7 μM) compared to wild-type Na V 1.7. Taken together, these data indicate that AM-8145 interacts with the second voltage-sensor domain of Na V 1.7.

AM-8145 Na V isoform selectivity
The potency and selectivity of AM-8145 on human Na V isoforms (Na V 1.1-Na V 1.8) as well as native Na V channels sensitive to tetrodotoxin (TTX-S) or resistant to tetrodotoxin (TTX-R) in mouse DRG neurons were evaluated using a manual patch clamp electrophysiology voltage protocol in which 20% of Na V channels were inactivated. AM-8145 exhibited an IC 50 of 0.58 nM on Na V 1.7 and was 30-to 120-fold selective over other human TTX-S Nav channels IC 50 values of JzTx-V analogs on human Na V 1.7, Na V 1.4 and Na V 1.5 as determined on the PX platform. IC 50 values were determined from at least ten different cells with two to three data points per peptide concentration as previously described [24]. n.d. = not determined. All peptides contain C-terminal amides. https://doi.org/10.1371/journal.pone.0196791.t001 Engineered JzTx-V peptide inhibitors of Na V 1.7 including Na V 1.1, Na V 1.2, Na V 1.3, Na V 1.4, and Na V 1.6 ( Fig 5A and 5C). AM-8145 potently inhibited native TTX-S channels, largely comprised of native Na V 1.7 [33], in mouse DRG neurons with an IC 50 of 10.3 nM. AM-8145 was over 1,000-fold selective over human TTX-R channels including Na V 1.5 and Na V 1.8 as well as native TTX-R Na V 1.8 channels in mouse DRG neurons with IC 50 values > 1 μM (Fig 5B and 5D). In summary, AM-8145 potently blocks human Na V 1.7 and native TTX-S channels in mouse DRG neurons, exhibits 30-to 120-fold selectivity over other human TTX-S channels, and exhibits over 1,000-fold selectivity over other human TTX-R channels.

AM-0422 and less active analog AM-8394 potency on rodent DRG neurons
To further evaluate the pharmacology of JzTx-V peptides in biological assays measuring action potential firing in rodent DRG neurons or peripheral nerve fibers, we generated a pair of  Potency of AM-8145 inhibition of wild-type Na V 1.7, wild-type Na V 1.5, or chimeric Na V 1.7/Na V 1.5 channels (red illustrates Na V 1.7 sequence and gray illustrates Na V 1.5 sequence) on the IWQ platform (mean ± SEM). D = "Domain"; S = "Sensor" corresponding to transmembrane segments 1-4; P = "Pore" corresponding to transmembrane segments 5-6. AM-8145 block of wild-type Na V 1.7 is similar to chimera DIIS containing the second voltage sensor domain of Na V 1.7 embedded in the Na V 1.5 backbone. https://doi.org/10.1371/journal.pone.0196791.g004 Engineered JzTx-V peptide inhibitors of Na V 1.7

Fig 5. Selectivity of AM-8145 for human Na V channels in heterologous cells and native Na V currents in mDRG neurons.
A. Concentration response curves for TTX-S channels including Na V 1.1-Na V 1.4, Na V 1.6-Na V 1.7 and mDRG TTX-S. IC 50 values with fold-selectivity compared to Na V 1.7 are listed in parenthesis. AM-8145 was 30-to 120-fold structurally similar peptides with disparate bioactivities. Since the Na V 1.7 potency and selectivity over Na V 1.4 and Na V 1.5 were comparable for AM-8145 and AM-0422, which mainly differ in positioning of the Pra residue for future derivatization, we selected AM-0422 as the active peptide for these experiments. AM-0422 potently blocked native TTX-S Na V channels in mouse (IC 50 27.6 ± 7.7 nM, n = 4) and rat (IC 50 9.3 ± 1.3 nM, n = 2) DRG neurons whereas AM-8145 peptide analog AM-8394 (Table 1), containing a Leu19Glu modification, Pra-[Nle6; Glu19,28]JzTx-V , was inactive at the highest concentration tested (1 μM AM-8394 caused 3.5 ± 1.1% inhibition), when assessed by manual patch clamp whole cell electrophysiology using a voltage protocol where 20% of Na V channels were inactivated (Fig 6). AM-8394 was designed and synthesized for this study based on the loss in Na V 1.7 potency when the wild-type Leu19 residue was replaced by either Ala or Glu during scanning experiments (Fig  2A). When evaluated against human Na V 1.7 and mouse Na V 1.7 channels in heterologous HEK293 cells, AM-8394 did not potently block sodium currents at the highest concentration tested (1 μM). Thus, AM-8394 represents a Ile28Glu containing JzTx-V analog that can be used as a negative control peptide in pharmacology studies.

AM-0422 and less active analog AM-8394 inhibition of action potential firing in rat DRG neurons by MEA
Experiments were performed to evaluate the ability of AM-0422 to block action potential firing in response to the nociceptive agent capsaicin in neurons that natively express Na V 1.7 channels. Using cultured rat DRG neurons grown on a multi-electrode array, 25 nM capsaicin (corresponding to an EC 80 concentration), but not buffer control, increased the frequency of action potential firing for a period of two minutes following application (Fig 7A, 7C, 7D and 7G). When co-applied with capsaicin, the active peptide AM-0422, but not the less active peptide AM-8394, dose-dependently decreased capsaicin-induced action potential firing (Fig 7A,  7B and 7E-7G). The IC 50 for AM-0422 block was 21.1 ± 5.7 nM whereas the IC 50 for AM-8394 was >1 μM. Taken together, these data indicate that AM-0422 specifically blocks capsaicininduced rat DRG neuron action potential firing.

AM-0422 and less active analog AM-8394 inhibition of action potential firing in skin-saphenous nerve assay
The mouse skin-saphenous nerve tissue preparation represents an ex vivo experimental system to evaluate action potential firing in nerve fibers that originate from peripheral receptive fields in the skin and are transmitted along the saphenous nerve. We previously demonstrated that both TTX and a Na V 1.7-specific small molecule sulfonamide gating modifier were able to inhibit mechanically-induced C-fiber action potential firing in this preparation [32]. Therefore, experiments were performed to evaluate the impact of AM-0422 in this system. When locally applied to a receptive field on the skin, AM-0422 dose-dependently blocked spiking in response to a 150 mN mechanical stimulus compared to 0.1% BSA vehicle control (Fig 8; selective for Na V 1.7 over Na V 1.1-Na V 1.4 and Na V 1.6. B. Concentration response curves for TTX-R channels including Na V 1.5, Na V 1.8 and mDRG TTX-R. AM-8145 was over 1,700-fold selective for Na V 1.7 over Na V 1.5, Na V 1.8 and mDRG TTX-R. Curves in A and B were derived from whole cell patch clamp experiments. Data are presented as mean ± SEM with n = 2-4 cells per data point. C. Representative current traces for TTX-S channels before (black) and after (red) 10 nM AM-8145 addition. Cells were held at a voltage yielding 20% channel inactivation and stepped to -10 mV for 14 msec. D. Representative current traces for TTX-R channels before (black) and after (red) 100 nM AM-8145 addition. Cells were held at a voltage yielding 20% channel inactivation and stepped to -10 mV for 14 msec (Na V 1.5), or 0 mV for 15msec (Na V 1.8 and mDRG TTX-R). Studies with mDRG TTX-R included 0.5 uM TTX to block endogenous TTX-S currents. 10 nM AM-8145 preferentially inhibited Na V 1.7.
repeated measures two-way ANOVA, F(5,130) = 34, P = <0.0001) with an IC 50 of 880 nM. When evaluated at 16 μM, the highest concentration of AM-0422 tested, the less active peptide AM-8394 had no effect on spiking. Taken together, these data indicate that AM-0422 specifically blocks mechanically-induced C-fiber action potential firing from afferent nerve terminals.

Discussion
AM-8145 and AM-0422 represent potent and selective engineered JzTx-V-derived peptide inhibitors of recombinant human Na V 1.7 channels in heterologous cells and native TTX-S Na V channels in mouse and rat DRG neurons. AM-0422 inhibited capsaicin-induced action potential firing in rat DRG neurons and mechanically-induced C-fiber action potential firing in a mouse skin-nerve preparation. The inactive control peptide AM-8394 did not block action potential firing, pointing to a specific requirement for Na V target engagement in these endpoints. This represents the first report characterizing the pharmacology of an active and inactive Na V 1.7 peptide pair in a collection of in vitro and ex vivo experiments. JzTx-V was originally described as an inhibitor of both TTX-S (IC 50 = 28 nM) and TTX-R (IC 50 = 30 nM) sodium channels in rat DRG neurons with additional inhibitory activity against the voltage-gated potassium channels K V 4.1, K V 4.2 and K V 2.1 [25] [34]. Consistent with our findings, JzTx-V exhibited weak inhibition of Na V 1.5 and had nominal selectivity over Na V 1.4 when evaluated in heterologous HEK293 cells [26]. In order to identify a better compound for further interrogation of Na V 1.7 biology and encouraged by the inherent potency of the parent JzTx-V peptide, we sought to engineer improved selectivity against Na V 1.4 using the JzTx-V scaffold. Given the lack of structural information of human Na V channels at the time of this work from X-ray crystallography or cryo-electron microscopy needed to elucidate atomic-level structural differences between the homologous TTX-S sodium channels Na V 1.7 and Na V 1.4, we focused on an empirical attribute-based positional scan analoging strategy previously successfully employed for K V 1.3 and Na V 1.7 [22,27]. The effect of a JzTx-V Ala-scan on rat Na V 1.4 block has been reported [26], and our Ala-mutant data validated the importance of residues Trp5, Trp24 and Arg26 in peptide block of human Engineered JzTx-V peptide inhibitors of Na V 1.7 TTX-S channels. However, we did not observe any indication of differential peptide activity against Na V 1.7 and Na V 1.4 with Ala scan mutants. In our previous GpTx-1 engineering study, scanning with the negatively charged glutamic acid residue had the most disruptive effect on bioactivity [22]. Like GpTx-1 and other tarantula venom-derived Na V inhibitory peptides, JzTx-V is basic in nature and we reasoned that a single residue scan with a carboxyl groupcontaining side chain could lead to novel peptides with modified Na V activity. Therefore, we synthesized and tested the full panel of Glu-scan mutants. The set of Glu JzTx-V single mutants increased the topographic map of residues that may interact with Na V 1.7 (e.g. residues Met6, Thr8, Leu23), and revealed a key Ile28Glu mutation with >100-fold selectivity for Na V 1.7 over Na V 1.4. This finding is a testament to the power of attribute-based positional analoging in peptide lead discovery and optimization when high resolution structural information is unavailable.
Combining scan analoging with NMR structural data showed that key residues for Na V 1.7 block were clustered on one face of the folded peptide while modifications on the opposite or back face were more tolerated. We identified locations on the back surface for attaching  synthetic handles for future peptide derivatization that retained Na V 1.7 potency and selectivity. This effort yielded the lead JzTx-V peptides AM-8145 and AM-0422 for further investigation of Na V 1.7 biology. Interestingly, the Leu19Glu analog of AM-8145 (AM-8394) lost bioactivity, despite the modified residue mapping to the peptide core based on the AM-8145 NMR structure, indicating that the contribution of certain residues in blocking ion channel function may be due to indirect conformational effects. In any case, the outcome afforded access to the inactive JzTx-V mutant AM-8394 as a negative control in biological experiments.
To further understand Nav1.4 selectivity gains of the Ile28Glu substitution, we solved the NMR structures of peptide 3 and its Glu28 analog, AM-8145. The Ile28Glu substitution reorients the C-terminus of the peptide with Glu28 in close proximity to Arg20. This results in the reorientation of Leu19 and Arg26, a subtle shift in the binding face of AM-8145 as well as larger conformational changes on the peptide back face. The change in the geometry and composition of the binding face may contribute to the Nav1.4 selectivity observed with this single residue change. Presently, no high-resolution structure of a mammalian Na V containing the second voltage-sensor domain is available. Given the dynamic nature of Na V gating and the rigid protein-protein docking methods available to us, we did not have sufficient confidence that docking studies would afford the resolution or accuracy necessary to explain the observed Na V 1.4 selectivity.
JzTx-V functions as a gating modifier peptide toxin that stabilizes Na V 1.7 in a closed conformation with state-independent block between -140 and -80 mV. JzTx-V has lower affinity for the channel open state(s) and high frequency strong depolarizations that cycle Na V 1.7 through closed, open, and inactivated conformations partially reversed JzTx-V channel block. Using a ProTx-II radiolabeled cell-based binding assay, a diverse panel of cold Na V 1.7 inhibitory peptides including JzTx-V derivatives (AM-8145 and AM-0422), GpTx-1, HwTx-IV, and ProTx-II competed for the same binding site with a good correlation between Na V 1.7 IC 50 values by electrophysiology and K i values by binding. A panel of Na V 1.7-Na V 1.5 chimeras was employed to map AM-8145 to the second voltage-sensor domain of Na V 1.7. Further Na V 1.7--Na V 1.5 chimeras with specific extracellular loop swaps did not narrow Na V 1.7 regions necessary for peptide block in preliminary experiments. Collectively, the landscape of different tarantula-derived Na V 1.7 inhibitory peptides including the JzTx-V derivatives described here, ProTx-II, HwTx-IV, GpTx-1 and additional newly described ICK peptides all share similar biophysical mechanisms of action including channel block via gating modulation, lower affinity for channel open states, and mapping to specific channel voltage-sensor domains in particular the second voltage sensor region [23,[35][36][37][38][39][40][41][42][43][44][45][46].
AM-0422 dose-dependently blocked action potential firing induced by the algogen capsaicin in cultured rat DRG neurons in a MEA assay. Nominal inhibition was observed with the inactive control peptide AM-8394. The half-maximal inhibitory concentration of AM-0422 in the spiking MEA assay was 21 nM, which was two to three-fold higher than the native TTX-S Na V IC 50 of 9 nM in rat DRG neurons. Similarly, AM-0422 dose-dependently blocked mechanically-induced action potential firing in mouse C-fibers in a skin-nerve preparation assay, and no inhibition was observed with the inactive peptide AM-8394. The half-maximal inhibitory concentration of AM-0422 in the skin nerve assay was 880 nM, which was thirtyfold higher than the native TTX-S Na V IC 50 in mouse DRG neurons. We were unable to evaluate AM-0422 block of capsaicin-induced action potential firing in the skin nerve preparation due to desensitization of capsaicin responses that precluded repeated testing before and after peptide pharmacological challenge. Higher concentrations of AM-0422 needed to block spiking in nerve fibers in the skin-nerve preparation compared to dissociated DRG neurons in culture could be attributable to dermal barriers that limit peptide access to Nav channels in ex vivo tissue systems, as previously described for the Na V 1.7 small molecule sulfonamide antagonist AMG8379 [32] and the Na V pore blocker TTX [47]. Consistent with this proposal of limited biodistribution to Na V channels, AM-0422 effects required 20-25 minutes of constant peptide perfusion to achieve full block, and the Na V 1.7 inhibitory peptide ProTx-II only impacted electrically-induced C-fiber action-potential firing following desheathing of the saphenous nerve to enable peptide access [48]. The highest concentration of AM-0422 tested (16 uM) nearly fully inhibited C-fiber action potential firing, whereas both AMG8379 and Nav1.7 KO mice yielded 60 percent C-fiber inhibition [19,32]. The additional block by AM-0422 may be due to engagement of other TTX-S Nav channels in C-fibers, such as Na V 1.1 and Na V 1.6 [49][50][51][52].
In summary, using a combination of attribute-based positional scan analoging and NMR structural analyses, we have engineered potent and selective Na V 1.7 inhibitory peptides derived from the JzTx-V scaffold. AM-8145 and AM-0422 represent research tools with sub-nM potency on human Na V 1.7 and 100-1000-fold selectivity over Na V 1.4 and Na V 1.5 that can be utilized to interrogate Na V 1.7 biology in cell and tissue-based preparations. Continued peptide optimization efforts to develop analogs that can engage Na V 1.7 in vivo and block pharmacodynamic endpoints are ongoing.
Supporting information S1 Table. Potency of JzTx-V Ala and Glu scan mutations. IC 50 values of JzTx-V peptides against human Na V 1.7, Na V 1.4, and Na V 1.5 as determined on the IWQ platform. Wild-type JzTx-V peptide sequence is listed in Column 1. Cysteine residues were not mutated (shaded in green).