Computer-aided Discovery of a New Nav1.7 Inhibitor for Treatment of Pain and Itch

To perform similarity search, two compounds with well-documented selective inhibitor activity toward Nav1.7 were selected as query structures: PF-05089771²⁵  and GX-936.²⁶  Both bind to the same site of voltage sensor domain 4 of Nav1.7 and exhibit a remarkably selective inhibition of Nav1.7 sodium currents. PF-05089771 was advanced to a stage 2 clinical trial;²⁵  GX-936 is characterized by available crystal structure of the ligand-protein complex. For virtual screening, a database of 1.5 million compounds was downloaded from the Chemdiv website (http://www.chemdiv.com/, ChemDiv, Inc., USA; accessed April 3, 2018). The downloaded structures were converted in a three-dimensional Structure Datafile format using LigPrep module from the Schrödinger Suite (Schrödinger, LLC, USA). We have generated multiple states for stereoisomers, tautomers, and protonation at pH range 5 to 9.

Similarity search was performed using the FTrees algorithm (FTrees 3.4 version 2.3.2; BioSolveIT GmbH, Germany). The search for similarity using FTrees is made up of two main steps: (1) defining the compound of the input query and the database where the query compound is screened against the database compounds library and (2) calculating similarity by using the generated feature trees and specifying the compounds as hits with a specified threshold of similarity. The algorithm of FTrees uses the shape and pharmacophore properties of the query molecule to represent them as the interconnected fragments. Several pharmacophore properties are considered in the similarity search, including hydrophobic features, the total number of ring closures, volume, hydrogen bond donors and acceptors, and aromatic properties. For assessing the “local” similarity of particular nodes in the form of Tanimoto coefficient, the values of each node are made with the properties and calculated as a numerical fingerprint. The similarity of the “global” feature tree is determined as the normalized sum of the local similarities of the mapped nodes by mapping the nodes of one feature tree onto the nodes of the other. A similarity threshold value of 0.5 against the compounds from the Chemdiv database (ChemDiv, Inc.) was used to screen for potential Nav1.7 inhibitors.²⁷ 

Molecular docking studies were performed using the crystal structure of the Nav1.7 ion channel obtained from the Protein Database (protein code 5EK0).²⁶  The missing side chains of partially resolved residues and the missing loop sequences were reconstructed by using the Modeller software.²⁸  The refined structure of the protein underwent energy minimization in an explicit water environment, being embedded in a lipid bilayer using an OPLS-2015 force field (Desmond module of Schrödinger suite, Schrödinger, LLC; Desmond Molecular Dynamics System, version 3.8, D. E. Shaw Research, USA). The voltage sensor domain 4 binding site in complex with the cocrystalized ligand (GX-936) was loaded into the protein preparation module of the BiosolveIT software (BiosolveIT GmbH, Germany). The atomic coordinates for the binding site were converged, and the resulting model was then used for docking. Binding positions of each ligand retrieved from the similarity search were generated, scored for predicted binding affinity, and ranked using the FlexX algorithm (LeadIT software package version 2.3.2, BioSolveIT GmbH, Germany). Protein residues interacting with the ligand were noted in each docking simulation. Ten poses were generated for each ligand and scored using SeeSAR software, which uses the HYDE scoring function (SeeSAR version 7.0; BioSolveIT GmbH, Germany, www.biosolveit.de/SeeSAR). In HYDE, the values are estimated based on the difference between the bound and unbound states, based on an atomic logP-based mathematical kernel. The system has not been trained to specific targets; instead, H-bond contribution and dehydration (“desolvation”) are intrinsically balanced without weighting parameters as seen in all force fields. By design, HYDE allows the visualization of ΔG on atoms.²⁹ 

Molecular dynamics simulations with an explicit membrane model with OPLS_2005 force field were conducted using Desmond v3.8 (Desmond Molecular Dynamics System, version 3.8, D. E. Shaw Research, USA). The cell membrane was built for the protein complex according to the membrane coordinates obtained from the membrane orientation (OPM) database (http:/opm.phar.umich.edu/).³⁰  The system was then placed in a 10 Å side-length orthorhombic box solved with simple point charge water and neutralized with a suitable amount of counter ions. OPLS_2005 force field was used for energy minimization of the complex system, with setting of the maximum interaction to 2,000 and setting the convergence threshold to 1.0 kcal × mol^-1 × Å^-1. The system performed 10 ns of isothermal–isobaric ensemble simulation at a temperature of 300 K set by Nose–Hoover thermostat and 1.01325 bar pressure set by Martyna–Tobias–Klein barostats before the simulation, to relax the complexes. For the production run, molecular dynamics simulation was running for 100 ns, where the energy and trajectory have been recorded at 1.2 ps and 4.8 ps, respectively, and the statistical analysis was carried out using the data obtained. To describe and compare the stability of the docking complexes, root mean square deviation, potential energy, ligand-protein interaction, and root mean square fluctuation were used.

C57BL/6 mice (stock No. 000664) and NOD.CB-17-Prkdc^scid mice (stock No. 001303, characterized by an absence of functional T cells and B cells, lymphopenia, hypogammaglobulinemia, and a normal hematopoietic microenvironment) were purchased from the Jackson Laboratory (USA). Both male and female mice were used in in vitro and in vivo experiments. Mice were group-housed on a 12-hour light/12-hour dark cycle at 22 ± 1°C with free access to food and water. Animals were randomly assigned to each group. Two to five mice were housed in each cage. If required by the protocol, animals were euthanized by carbon dioxide followed by decapitation. For behavioral tests, mice were habituated to the testing environment daily for at least 2 days before testing. All of the behavioral tests were conducted between 9:00 am and 5:00 pm. Animal behaviors were tested blindly. All the animal experiments were conducted in accordance with the National Institutes of Health (Bethesda, Maryland) Guide for the Care and Use of Laboratory Animals and approved by Institutional Animal Care Use Committee of Duke University (Durham, North Carolina). Data from the conducted experiments were used to refine future protocols to minimize the risk of animal suffering and death.

Human embryonic kidney–hNav1.7 stable cell line was purchased from SB Drug Discovery (United Kingdom). Human embryonic kidney 293 cells were obtained from the American Type Culture Collection (USA) and maintained at the Duke cell culture facility. Cells were cultured in high-glucose (4.5 g/l) Dulbecco’s Modified Eagle’s Medium containing 10% (by volume) fetal bovine serum (Gibco, USA) and streptomycin/penicillin (Thermo Fisher Scientific, USA). For Nav1.7-expressing human embryonic kidney 293 cells, an additional antibiotic, blasticidin, was included. hNav1.5 transfection was performed by Lipofectamine 2000 (Thermo Fisher Scientific) at 70% confluence using 2 μg cDNA. The transfected cells were cultured in the same growth medium for 48 h before electrophysiologic studies.

Mouse dorsal root ganglia were removed aseptically from mice (4 to 6 weeks) and incubated with collagenase (1.25 mg/ml, Roche, Germany)/dispase-II (2.4 U/ml, Roche) in Hanks’ Balanced Salt solution buffer at 37°C for 90 min. Cells were mechanically dissociated with a flame polished Pasteur pipette in the presence of 0.05% DNAse I (Sigma, USA). Dorsal root ganglion cells were plated on 0.5 mg/ml poly-D-lysine-coated glass cover slips and grown in a neurobasal defined medium (with 10% fetal bovine serum, 2% B27 supplement, Invitrogen, USA) with 5 μM AraC (Sigma) and 5% carbon dioxide at 36.5°C. Dorsal root ganglion neurons were grown for 24 h before use. For paclitaxel pretreatment experiments, cultured dorsal root ganglion neurons were incubated with 1 µM paclitaxel (Sigma) in the culture medium for 16 to 24 h before recordings. Patch clamp recordings were conducted in small-diameter (less than 25 μm) dorsal root ganglion neurons.

Nondiseased human dorsal root ganglia were obtained from two donors through National Disease Research Interchange (USA) with permission of exemption from the Duke Institutional Review Board. Postmortem L3 to L5 dorsal root ganglia were dissected and delivered in ice-cold culture medium to the laboratory at Duke University within 24 to 72 hours of the donor’s death. Upon the delivery, dorsal root ganglia were rapidly dissected from nerve roots and minced in a calcium-free Hanks’ Balanced Salt solution (Gibco). Dorsal root ganglia were digested at 37°C in a humidified oxygen incubator for 120 min with collagenase type II (Worthington, USA, 285 U/mg, 12 mg/ml final concentration) and dispase II (Roche, 1 U/mg, 20 mg/ml) in Hanks’ Balanced Salt solution buffer. Dorsal root ganglia were mechanically dissociated using fire-polished pipettes, filtered through a 100 µM nylon mesh, and centrifuged (500 × g for 5 min). The pellet was resuspended and plated on 0.5 mg/ml poly-D-lysine–coated glass coverslips, and cells were grown in Neurobasal medium supplemented with 10% fetal bovine serum, 2% B-27 supplement, and 1% penicillin/streptomycin.³¹  Patch clamp recordings were conducted in small-diameter (less than 55 μm) dorsal root ganglion neurons.³² 

Whole-cell patch-clamp recordings were conducted at room temperature. Patch pipettes pulled from borosilicate capillaries (World Precision Instruments, Inc., USA) were used to record transient Na⁺ currents with an Axopatch-200B amplifier with a Digidata 1440A (Axon Instruments, USA). The resistance was set to 3 to 4 MΩ when filled with pipette solution. The pipette solution contained (in mmol/l): CsCl 130, NaCl 9, MgCl₂ 1, EGTA 10, and HEPES 10, adjusted to pH 7.3 with CsOH. The external solution for recording transient Na⁺ currents contained (in mmol/l): NaCl 131, tetraethylammonium chloride 10, CsCl 10, CaCl₂ 1, MgCl₂ 2, CdCl₂ 0.3, 4-aminopyridine 3, HEPES 10, and glucose 10, adjusted to pH 7.4 with NaOH. The recording chamber (300 µl) was continuously superfused at 3 to 4 ml/min. Series resistance was compensated (greater than 80%), and leak subtraction was performed. Data were low-pass-filtered at 2 kHz and sampled at 10 kHz. pClamp10 (Axon Instruments) software was used during experiments and analysis. In voltage-clamp experiments, transient Na⁺ currents were evoked by a test pulse to 0 mV from the holding potential (−70 mV).³³ 

The formalin test was conducted following our previously published protocol.³⁴  The animals were acclimatized in a transparent acrylic observation chamber for 2 days before the test. Vehicle or active compound was administered intrathecally, intraperitoneally, or intraplantarly, 30 min before intraplantar injection of 20 μl of 5% formalin (Sigma). After the formalin injection, mice were immediately placed back in the observation chamber, and videoed for 40 min. The videos were analyzed after experiments in a double-blinded way. We assessed formalin-evoked spontaneous pain by measuring the time (in seconds) mice spent on licking or flinching the affected paw every 5 min for 40 min. Phase I and phase II were defined as periods of 0 to 10 min and 10 to 40 min, respectively.

For intrathecal injection, spinal cord puncture was made by a Hamilton microsyringe (Hamilton Co., USA) with a 30-G needle between the L5 and L6 level to deliver reagents (5 µl) to the cerebral spinal fluid. For intraperitoneal injection, vehicle or drug was injected into peritoneal with a 30-G needle. For intraplantar injection, 20 µl vehicle or drug was injected into the surface of the hind paw.

Intraperitoneal injection of paclitaxel (2 mg/kg for multiple injections at days 0, 2, 4, and 6) was given to generate chemotherapy-associated neuropathic pain.³²  Neuropathic pain behavior was tested between 2 to 3 weeks after paclitaxel injection using the von Frey test. Briefly, animals were habituated to the testing environment daily for at least 2 days before the baseline testing. Animals were confined in boxes placed on an elevated metal mesh floor, and the hind paws were stimulated with a series of von Frey hairs with logarithmically increasing stiffness (0.02 to 2.56 g, Stoelting, USA), presented perpendicularly to the central plantar surface. The 50% paw withdrawal threshold was determined by an up-down method. In the experiment involving repeated administration of compound 9, mice received three intrathecal injections of 20 nmol compound 9 with a 24-h interval between the injections; mechanical thresholds were tested 1 h after each injection.

Itch behavior was tested blindly. Mice were habituated to the testing environment daily for at least 2 days before testing. Animals were shaved at the back of the neck in an area of approximately 15 × 10 mm on the day before the injection of pruritic agent. Animals were put in small plastic chambers (14 × 18 × 12 cm) on an elevated metal mesh floor and allowed 30 min for habituation before examination. Mice were then briefly removed from the chamber and given an intradermal injection (50 μl) of 500 μg of histamine (Sigma) in the nape of the neck. After the injection, mouse behaviors were video-recorded and the number of scratches in every 5 min for 30 min was counted. A scratch was counted when a mouse lifted its hind paw to scratch the shaved region and returned the paw to the floor or to the mouth.³⁵ 

We used a recently developed murine xenograft model of cutaneous T cell lymphoma using immune-deficient mice (NOD.CB17-Prkdc^scid, 8 to 10 weeks old, male).³⁶  CD4⁺ MyLa cell line was purchased from Sigma (catalog No. 95051032). The cell line was established from a plaque biopsy of an 82-year-old male with mycosis fungoides stage II by inclusion of interleukin 2 and interleukin 4 in the culture medium. Lymphoma was generated via intradermal injection of CD4⁺ MyLa cells (1 × 10⁵ cells/μl, 100 μl) on the nape of the neck under 2% isoflurane anesthesia. Itch behavior was tested in the chronic phase (day 50) after inoculation of lymphoma cells.³⁶ 

The sample size for each experiment was based on our previous studies that used the same experimental design.^37–40  Therefore, no formal statistical power calculation was conducted. Residual plots and Q-Q plots were used to assess the normality assumption required by parametric tests. Outliers were identified using a robust regression followed by outlier identification method.⁴¹  If outliers were present, results were reported for data including and excluding the outliers. There were no missing data; all data endpoints were collected from each animal as planned by the protocol and accounted for in the statistical analyses. Repeated measures two-way ANOVA was used to model data obtained from repeated measurements over time. For electrophysiology experiments, the overall time × treatment interaction was evaluated. If statistically significant (P < 0.05), then pairwise comparisons of all treatments with control at the last timepoint (the primary outcome) were evaluated using a t-statistic with Bonferroni correction. For behavioral experiments, if the overall time × treatment interaction for the time course was statistically significant, then pairwise comparisons with control at each timepoint were performed using a t-statistic, Bonferroni-corrected for the number of active treatment groups. Inflammatory pain model data were analyzed stratified by phase (where the effect of treatment in phase II was considered the primary outcome). Reported P values are adjusted for multiple testing. The criterion for statistical significance was P < 0.05. All statistical hypotheses were two-tailed unless noted otherwise. All data were expressed as means ± SD.

A tiered approach was utilized to identify selective inhibitors of Nav1.7, which included similarity search, followed by docking and molecular dynamic simulation (fig. 1). Similarity search of the ChemDiv database (ChemDiv, Inc.) containing 1.5 million presynthetized compounds yielded a set of 2,000 compounds structurally similar to the query molecules—known inhibitors of Nav1.7. This set of compounds was subjected to docking to predict the predominant binding mode(s) of a ligand with the target protein. Crystal structure of the homotetrameric NavAb/Nav1.7 chimera was utilized to model the extracellular loops of voltage sensor domain 4.²⁶  Each of the 2,000 compounds was docked into a binding site and assessed for ligand affinity. The top 25 compounds with highest affinity scores were selected to perform molecular dynamic simulations. Molecular dynamic simulations were performed for 100 ns, with the starting point of the ligand being docked into the binding site. Root mean square deviation was used to measure the average change in displacement of atoms for a particular frame with respect to a reference frame. A total of 13 complexes showed stable positioning during molecular dynamic simulations (Supplemental Digital Content, https://links.lww.com/ALN/C426). Of those, nine compounds with unique scaffolds and sufficient quantity (1 mg or more) available in stock were selected and purchased from ChemDiv, Inc. for in vitro and in vivo activity assessment (fig. 2).

The effects of the nine selected compounds on Nav1.7 channel activities were tested using patch-clamp recordings on human Nav1.7 expressing human embryonic kidney 293 cells (figs. 3 and 4). In each cell, sodium currents were recorded for 13 min, and candidate inhibitors (10 μM) were incubated for 2 min after 2 min baseline recording (fig. 3). Out of nine tested compounds, compounds 5, 7, 8, and 9 showed significant inhibition of sodium currents in Nav1.7-expressing cell line (29%, P < 0.05, or better, fig. 3).

Specifically, the top performing compound 9 at the concentration of 10 µM reduced the sodium current by 80% (fig. 3). A dose-response analysis revealed that 50% inhibition concentration for compound 9 was 0.74 µM (95% CI, 0.35 to 1.56 μM; fig. 4A–C). As a positive control,⁶  PF-05089771 (1 µM) showed fast and significant inhibition of sodium currents in Nav1.7-expressing cell line (43% inhibition vs. control, fig. 4, A and B), which is similar to the effects of 1 µM compound 9 (47% inhibition vs. control). Compounds 5, 7, and 8 also showed statistically significant inhibitory activity, although to a lower degree than compound 9 (fig. 3).

Based on these results, compound 9 (3-(1-benzyl-1H-indol-3-yl)-3-(3-phenoxyphenyl)-N-(2-(pyrrolidin-1-yl)ethyl) propanamide, named by authors as “DA-0218”) was selected as a lead compound for further investigation of selectivity against Nav1.5 and assessment of its activity in vivo. Notably, DA-0218 had no effect on sodium currents in Nav1.5-transfected human embryonic kidney 293 cells in patch-clamp experiments (fig. 4, D and E).

Because NaV1.7 is highly expressed by small-sized nociceptive dorsal root ganglion neurons,⁹  we examined whether DA-0218 would also modulate transient sodium currents in dissociated small-sized dorsal root ganglion neurons from mice and human donors. DA-0218 (10 μM) produced a modest (17%; 95% CI, 6 to 28%) but statistically significant (two-way repeated measures ANOVA time × treatment interaction F[41, 369] = 4.072; P < 0.0001) inhibition of peak transient sodium currents in mouse dorsal root ganglion neurons (fig. 5, A and B). Notably, the kinetics of inhibition was slow, reaching a peak inhibition 10 min after the drug incubation (fig. 5B). Notably, paclitaxel pretreatment could significantly increase the amplitude of sodium currents (t = 4.023; df = 40; P < 0.001; fig. 5C), and furthermore, DA-0218 (10 µM) produced a more potent inhibition of sodium currents in paclitaxel-pretreated mouse dorsal root ganglion neurons (44% in treated vs. 17% in nontreated group; t = 2.292; df = 11; P = 0.043; fig. 5D). These important findings suggest that DA-0218 produces state-dependent inhibition of sodium currents, as previously shown for other sodium channel inhibitors such as carbamazepine.^42,43  As expected, the amplitude of transient sodium currents was 5- to 10-fold larger in human dorsal root ganglion neurons than that in mouse dorsal root ganglion neurons (fig. 5, A to E). DA-0218 (10 μM) also produced a moderate (relative current difference, 0.22; 95% CI, 0.08 to 0.35) but statistically significant inhibition of peak transient sodium currents in human dorsal root ganglion neurons (fig. 5, E and F).

After confirming that DA-0218 inhibits sodium current in dorsal root ganglion neurons, we next tested whether DA-0218 could attenuate formalin induced inflammatory pain following spinal intrathecal route via lumbar puncture. Formalin induced a biphasic spontaneous pain, as characterized as phase I (0 to 10 min) and phase II (10 to 40 min) (fig. 6, A and B). Intrathecal administration of DA-0218 produced a substantial and dose-dependent inhibition of the second-phase pain by 80% (95% CI, 68 to 92%; fig. 6, A and B). Intrathecal DA-0218 at the highest dose (30 nmol) also reduced phase I response (fig. 6A). Intraperitoneal injection of DA-0218 (30 mg/kg) also inhibited formalin-induced pain in phase II by 76% (95% CI, 48 to 100%; fig. 6, C and D). Intraplantar injection of 60 nmol of DA-0218 resulted in significant inhibition of formalin-induced pain in phase II (fig. 6, E and F). Thus, either systemic (intraperitoneal) or local (intrathecal or intraplantar) treatment of DA-0218 could effectively inhibit inflammatory pain via peripheral (intraplantar) or central (intrathecal) modulation.

We further tested the efficacy of DA-0218 in neuropathic pain induced by paclitaxel. DA-0218 transiently but completely reversed paclitaxel-induced mechanical allodynia for several hours when administered intrathecally (fig. 7A). Specifically, the dose response effect was evident at 1 h and 3 h after the injection, but not at 5 h. DA-0218 showed effective antiallodynia in both male and female mice after intrathecal injection (fig. 7, B and C). Repeated injections of DA-0218 after the second and third injections did not produce any analgesic effect tested at 1 h after each injection (fig. 7D). We also tested the effects of PF-05089771 on paclitaxel-induced neuropathic pain behavior, but intrathecal injection of PF-05089771 (10 nmol and 30 nmol) failed to produce any antiallodynic effect in the paclitaxel model (fig. 7E). Interestingly, no effect on allodynia was observed when DA-0218 was administered intraperitoneally (30 mg/kg, fig. 7F). Notably, this dose was very effective in reducing formalin-induced inflammatory pain (fig. 6, C and D).

After establishing the effectiveness of DA-0218 in models of inflammatory and neuropathic pain, we tested if DA-0218 is also effective against acute and chronic itch. First, intradermal injection of 500 μg histamine induced acute itch starting at 5 min but declining at 30 min after application (fig. 8A). Intrathecal injection of DA-0218 resulted in a dose-dependent reduction of scratch bouts induced by histamine (fig. 8, A and B). Second, lymphoma induced chronic itch in mice as described previously.³⁶  At day 50, lymphoma caused a robust spontaneous itch, showing greater than 100 bouts per hour (fig. 8C), suggesting this is an animal model of chronic itch. DA-0218 was also effective in suppressing chronic itch after intrathecal injection: At the dose of 30 nmol, this compound resulted in a profound reduction of scratch at 1 h and 3 h, and the anti-itch effect remained statistically significant after 5 h (fig. 8C).

In this study, we have identified several new small molecule Nav1.7 inhibitors using a computer-aided drug design approach. Our best performing molecule DA-0218 has a different binding mechanism than previously described sulfonamide-based Nav1.7 inhibitors, even though it binds within the same pocket of voltage sensor domain 4 of the channel. Our results show that DA-0218 inhibits sodium currents in Nav1.7-expressing cell line at 50% inhibition concentration of 0.74 μM and that systemic, intrathecal, and intraplantar administration of DA-0218 substantially reduced formalin-induced inflammatory pain. Moreover, DA-0218 was effective in reducing paclitaxel-induced neuropathic pain and lymphoma-induced chronic itch in mouse models. Thus, DA-0218 may represent a new class of Nav1.7 inhibitors showing both in vitro and in vivo activities.

Notably, compounds 5, 7, 8, and 9 (DA-0218) that show Nav1.7 inhibitory activity in vitro are all derivatives of 3-(1-benzyl-1H-indol-3-yl)-3-(3-methoxyphenyl)-N-methylpropanamide. Members of this class of compounds share a similar structure motif which consist of three moieties: propanamide, methoxyphenyl, and benzyl indolyl. Our docking experiments showed that these three moieties each occupy a specific area of the binding site: the propanamide moiety is docked in the selectivity pocket, the methoxyphenyl moiety is in the lipid-exposed pocket, and the benzyl indolyl (aryl indole) moiety occupies the anion-binding pocket of the binding site. DA-0218 (3-(1-benzyl-1H-indol-3-yl)-3-(3-phenoxyphenyl)-N-(2-(pyrrolidin-1-yl)ethyl) propanamide) also possesses a pyrrolidine group as its major difference from the other explored compounds.

Comparison of the docking modes of the compounds in the binding pocket reveals features that may be critical for compound activity. As illustrated in figure 9 and table 1, the predicted binding mode for DA-0218 is very similar to that of GX-936. The two pi-pi stacking interactions between TYR1537 and two aromatic rings of DA-0218 seem to be critical for the positioning of the ligand. Anchoring of DA-0218 is further stabilized by an H-bonding of ASP1586 with the amine of the pyrrolidine ring. In compound 5, there are three pi-pi stacking interactions between TYR1537 and indole and benzene rings, as well as an H-bond between ASP1586 and propanamide group of compound 5. These interactions might help for positioning and anchoring of compound 5. Thus, compounds 5 and DA-0218 are exhibiting a similar pattern of interactions with voltage sensor domain 4 peptide chains. These two compounds show an H-bond interaction with ASP1586, which is not observed in the rest of explored compounds. This new feature in ligand recognition suggests the role of ASP1586 in ligand recognition in compounds 5 and DA-0218. The pi-pi interactions present here are significant for maintaining stable conformation of the molecule.

The absence of a sulfonamide group in our compounds 5 and DA-0218 explains an absence of the H-bonds with ARG1602 and ARG1608 via oxygen and the nitrogen atom on the sulfonamide group, typically observed in sulfonamide-based Nav1.7 inhibitors. Without the sulfonamide moiety, our compounds are docked deep into the active site. As opposed to an anionic “warhead”²⁶  of sulfonamides, the head group of compounds 5 and DA-0218 is hydrophobic in nature; compound 5 is making a pi-cation interaction with ARG1608 instead of an H-bond formation.

Compound 8 has two H-bonds that seem to be crucial for its binding. One such bond is between the ASP1586 and propanamide group and the second one is with ARG1602. As with compounds 5 and DA-0218, a pi-pi stacking with TYR1537 was observed, as well as a pi-cation interaction with ARG1608. In contrast with compounds 5 and DA-0218, compound 8’s indole group formed no interaction with the peptide.

Compound 7, despite showing an inhibitory profile in human embryonic kidney 293 cells very similar to compound 8, has a very different binding pattern within the protein pocket. Specifically, compound 7 has one H-bond with ARG1602 (but not with ASP1586) and a pi-cation interaction with ARG1608.

Compound 5 had lower activity compared to DA-0218 in patch-clamp experiments, likely due to the shorter pyridine moiety, which is not as favorable for compound activity as the propanamide with pyrrolidine in DA-0218. Similarly, the furan ring in compound 8 results in some activity but less than the pyrrolidine in DA-0218. All three compounds have a phenoxyphenyl group as a common feature, which is also deemed to favor their activity.

A negative impact on the activity of compound 4 and 6 is likely due to the presence of piperazin in their structure. Further study of the structure-activity relationship will help us to find a clue for optimization.

DA-0218 transiently but completely reversed paclitaxel-induced mechanical allodynia for several hours when administered intrathecally. The surprising result in our study was the acute tolerance induced after repeated injections of DA-0218 in the paclitaxel model. This result suggests the difficulty of developing Nav1.7-selective inhibitors for the treatment of chronic pain. This acute antinociceptive tolerance may be related to unique pharmacokinetics and/or pharmacodynamics of the compound and to a particular pain model. It may also result from desensitization of the channel or the compensation of other sodium channels (e.g., Nav1.6, Nav1.8, and Nav1.9) after the inhibitor treatment. It is possible that a pan-sodium channel inhibitor (avoiding Nav1.5) could be more effective. Notably, loss of Nav1.7 resulted in upregulation of opioid receptor signaling.^44,45  It will be of great interest to investigate the additive analgesic effects of DA-0218 and a nonaddictive opioid (e.g., buprenorphine).

Interestingly, DA-0218 appears to be more effective in inhibiting chronic itch (effect duration greater than 5 h) than inhibiting chronic pain (effect duration 3 h). Therefore, this new inhibitor of Nav1.7 may as well lead to a new line of anti-itch candidates.

Compared to Nav1.7-expressing human embryonic kidney 293 cells, DA-0218 produced much less inhibition of sodium currents in dorsal root ganglion neurons. This discrepancy may result from the fact that dorsal root ganglion nociceptor neurons express a different subtype of sodium channels (e.g., Nav1.6, Nav1.8, Nav1.9). Notably, DA-0218 has similar inhibition rate (around 20%) in both mouse and human dorsal root ganglion neurons. It is important to point out that this inhibitor targets the voltage sensor of sodium channel and, therefore, its efficacy may depend on the activation states of sodium channel. Nav1.7 expression and function in human dorsal root ganglion neurons are upregulated in neuropathic pain conditions, such as paclitaxel-induced neuropathic pain.^33,46  Indeed, in our experiments, paclitaxel pretreatment increased the amplitude of sodium currents, and furthermore, DA-0218 produced a more potent inhibition of sodium currents in paclitaxel-pretreated mouse dorsal root ganglion neurons compared to nontreated controls. These important findings corroborate the hypothesis that DA-0218 produces state-dependent inhibition of sodium currents, as previously shown for other sodium channel inhibitors such as carbamazepine.^42,43  These findings suggest that DA-0218 may produce more profound inhibition of sodium currents and pain in chronic pain conditions than in acute pain states.

Interestingly, we found that intrathecal injection but not systemic injection of DA-0218 could reduce paclitaxel-induced mechanical allodynia (fig. 7). This result suggested that either the central action of the Nav1.7 inhibitor is critical for the analgesic efficacy in neuropathic pain, or different modes of administration result in different concentrations of DA-0218 at the central terminal of the nociceptor. In contrast, systemic injection of DA-0218 was sufficient to reduce inflammatory pain, suggesting that different mechanisms might underlie inflammatory pain versus neuropathic pain or acute pain versus chronic pain. It is also possible that spontaneous pain or spontaneous itch is more sensitive to the Nav1.7 inhibition than evoked pain (e.g., mechanical allodynia), given a critical contribution of Nav1.7 to the generation of action potentials and spontaneous discharges. Future study is needed to define penetration of DA-0218 into the central nervous system.

This study has several important limitations. First, although in silico modeling shows that DA-0218 binds in a specific voltage sensor domain 4 site at Nav1.7, other binding sites are possible, on Nav1.7, other sodium channel subtypes, or nonsodium channel targets. Second, the pharmacokinetic properties of DA-0218 are yet to be determined. Although our in vitro data support that metabolites alone are unlikely to explain the inhibitory effects of DA-0218 on Nav1.7, it is still unknown whether DA-0218 itself, or its metabolites, have the predominant impact on pain behavior and itch in mouse models. Furthermore, the effective in vivo concentrations of DA-0218 and its metabolites need to be measured. Third, selectivity of DA-0218 toward Nav1.7 was evaluated only against Nav1.5. Other subtypes need to be investigated, and selectivity ratios are measured.

In summary, we have successfully utilized a virtual screening approach to identify new inhibitors of Nav1.7 using a subtype-specific binding mechanism. DA-0218 identified during the study exhibited promising in vitro and in vivo activities and may represent a good starting point for lead optimization to discover more potent Nav1.7 inhibitors to treat pain and itch.

The authors thank BioSolveIT Gmbh (Sankt Augustin, Germany) for a no-cost license to use their software suite for this research project (https://www.biosolveit.de/).

This study is supported by internal funding from the Duke University Department of Anesthesiology (Durham, North Carolina).

BioSolveIT Gmbh (Sankt Augustin, Germany) provided free access to software to perform the analyses for this study. Dr. Chandra was a winner of an analytical competition and received €1,000 from the company to attend the Rocky Mountain 2019 Bioinformatics conference. Dr. Ji serves as a consultant for Boston Scientific (Valencia, California) and also serves on the Board of Directors of Ascletis Pharm (Hangzhou, China). Dr. Chandra, Dr. Ji, and Dr. Bortsov are listed as inventors on a provisional patent application 62/946,527 filed December 12, 2019.