Discovery of a selective, state-independent inhibitor of NaV1.7 by modification of guanidinium toxins

The voltage-gated sodium channel isoform NaV1.7 is highly expressed in dorsal root ganglion neurons and is obligatory for nociceptive signal transmission. Genetic gain-of-function and loss-of-function NaV1.7 mutations have been identified in select individuals, and are associated with episodic extreme pain disorders and insensitivity to pain, respectively. These findings implicate NaV1.7 as a key pharmacotherapeutic target for the treatment of pain. While several small molecules targeting NaV1.7 have been advanced to clinical development, no NaV1.7-selective compound has shown convincing efficacy in clinical pain applications. Here we describe the discovery and characterization of ST-2262, a NaV1.7 inhibitor that blocks the extracellular vestibule of the channel with an IC50 of 72 nM and greater than 200-fold selectivity over off-target sodium channel isoforms, NaV1.1–1.6 and NaV1.8. In contrast to other NaV1.7 inhibitors that preferentially inhibit the inactivated state of the channel, ST-2262 is equipotent in a protocol that favors the resting state of the channel, a protocol that favors the inactivated state, and a high frequency protocol. In a non-human primate study, animals treated with ST-2262 exhibited reduced sensitivity to noxious heat. These findings establish the extracellular vestibule of the sodium channel as a viable receptor site for the design of selective ligands targeting NaV1.7.

www.nature.com/scientificreports/ recovery from inactivation ( Fig. 1B) [13][14][15][16] . By contrast, cationic guanidinium toxins and peptide cone snail toxins inhibit ion conduction by sterically occluding the extracellular vestibule of the channel pore (Site 1). The former are a unique collection of small molecule natural products exemplified by saxitoxin and tetrodotoxin-high affinity, state-independent blockers against six of nine Na V subtypes ( Fig. 1C) 17 .
In the pursuit of isoform-selective inhibitors of Na V 1.7, two binding sites, the cystine knot toxin site at VSD II and the sulfonamide site at VSD IV, have been heavily interrogated. Certain cystine knot toxins that engage VSD II such as HwTx-IV, Pn3a, and ProTx-II exhibit 6-1,000× selectivity for Na V 1.7 over other channel isoforms. Potency and selectivity for this target have been improved with synthetic peptide toxin derivatives [22][23][24][25][26] . Among small, Lipinski-type molecules, only the aryl and acyl sulfonamides pioneered by Icagen/Pfizer and subsequently investigated by Amgen, Chromocell, Genentech/Xenon, Merck, and others have shown evidence of significant Na V 1.7 isoform selectivity 7,16 . Within the sulfonamide series, selectivity levels are > 1,000× over certain off-target isoforms including the cardiac isoform, Na V 1.5, but generally 10-50× against Na V 1.2 and Na V 1.6. Many but not all sulfonamide Na V 1.7 inhibitors suffer from undesirable pharmaceutical properties, including high plasma protein binding (e.g. > 99.8%), cytochrome p450 inhibition, in vitro hepatotoxicity and high unbound clearance 27,28 , which have hindered clinical development. Although a number of candidates have been advanced to human testing, one compound has been discontinued after a Phase 2 study likely due to limited efficacy (PF-05089771); others have been discontinued after Phase 1 trials for reasons that may be related to safety liabilities such as elevated expression of liver transaminases and hypotension (GDC-0276) 29,30 .
Electrophysiology studies with naturally occurring Site 1 ligands against different wild-type and mutant Na V isoforms have identified the extracellular vestibule of Na V 1.7 as a promising locus for selective inhibitor design [31][32][33] . The outer mouth of the channel is formed from residues that link the S5-S6 helices (referred to as pore loops) from each of the four domains. The domain III pore loop of human Na V 1.7 contains a T1398/I1399 sequence motif that is not present in other human Na V subtypes (which contain MD at equivalent positions, Suppl Table 1) 31 . Comparison of the amino acid sequence of the domain III pore loop across species indicates that the sequence motif in hNa V 1.7 is unique to primates. The half-maximal inhibitory concentration (IC 50 ) value for saxitoxin (STX) is markedly altered (250-fold change) depending on the presence or absence of the T1398 and I1399 residues. Against rNa V 1.4, the IC 50 of STX is 2.8 ± 0.1 nM compared to 702 ± 53 nM for hNa V 1.7 31 . Introduction of the alternative domain III pore loop sequence by mutagenesis restores potency (hNa V 1.7 T1398M/I1399D IC 50 = 2.3 ± 0.2 nM). These findings suggest that it may be possible to capitalize on structural differences in the extracellular vestibule between hNa V isoforms to design Na V 1.7-selective inhibitors.
Recent advances in the de novo synthesis of guanidinium toxin analogues have enabled systemic examination of the structure-activity relationship (SAR) properties that govern hNa V 1.7 potency and isoform selectivity [34][35][36][37] . Prior to 2016, the binding orientation of STX proposed in the literature indicated that the C11 methylene carbon was positioned proximally to domain III pore loop residues [38][39][40] . SAR and mutant cycle analysis studies posited a revised pose in which the C13 carbamate moiety abuts DIII 32 . This revised binding pose was recently confirmed by cryoelectron microscopy (cryo-EM) structures of STX bound to Na V PaS and hNa V 1.7 18,41 . In the present study, analogues of STX substituted at both the C11 and C13 positions were investigated to understand the requirements for selective inhibition of hNa V 1.7. These efforts led to the discovery of ST-2262, a potent and selective inhibitor of Na V 1.7 that reduces sensitivity to noxious heat in a preliminary study in non-human primates (NHPs).

Discovery of ST-2262.
ST-2262 was discovered through a rational design strategy aimed at identifying derivatives of natural bis-guanidinium toxins that preferentially inhibit hNa V 1.7 over other off-target hNa V isoforms 31 . Mutagenesis, homology modeling, and docking studies conducted prior to 2016 suggested that bis-  [38][39][40] . Exploration of substitution at C11 of decarbamoyl saxitoxin (dcSTX) led to the identification of a series of analogues bearing aryl amide groups at this site. Certain compounds, as exemplified by ST-282, show excellent potency against hNa V 1.7 but minimal selectivity (~ 1:1) over off-target isoforms such as hNa V 1.4 (Fig. 2B). The finding that hNa V 1.7 isoform selectivity could not be achieved by modification of the C11 substituent led us to investigate SAR at alternative positions. These studies followed evidence that the proper binding orientation of STX is rotated ~ 180° from earlier models, thus placing the C13 substituent in close proximity to domain III ( Fig. 2A, Revised pose) 32 .
Derivatives of STX bearing amide, carbamate, ester, ether, and urethane substituents at the C13 position were prepared in an effort to identify compounds with improved selectivity for hNa V 1.7. Insight from studies of a naturally-occurring STX C13 acetate congener, STX-OAc, helped guide selection of compounds for synthesis (Suppl Figure 1) 32 . The difference in potencies between STX and STX-OAc is striking considering that these two structures vary at a single position (NH 2 → CH 3 ). Following this lead, we explored substituents at C13 that could replace the hydrolytically unstable acetate group. Ultimately, the C13 succinimide was discovered as a suitable acetate isostere, which was paired with a C11 tetrahydronaphthyl amide to generate ST-2262, the focus of the present study.  Table 2). Potencies against off-target sodium channel isoforms (hNa V 1.1-1.6, hNa V 1.8) were determined following a similar protocol. Activity against hNa V 1.9 was not evaluated due to the difficulty of expressing this subtype heterologously 42 . ST-2262 was determined to be > 200-fold selective over hNa V 1.6 (IC 50 = 17.9 µM, 95% CI 14.8-22.1), > 900-fold selective over hNa V 1.3 (IC 50 = 65.3 µM, CI 62.7-68.1), and > 1,000-fold selective over all other Na V isoforms tested. Similar IC 50 values against the eight hNa V subtypes were obtained in an independent study using the PatchXpress automated electrophysiology platform (Suppl Table 3).

ST-2262 is a potent and selective inhibitor of hNa
Exposure of hNa V 1.1 and hNa V 1.2 to high concentrations of ST-2262 (10-100 µM) resulted in a reduction of the rate of fast inactivation; a similar effect was noted, albeit to a lesser degree, with hNa V 1.3 and hNa V 1.4 (Suppl Figure 2). Lower concentrations of ST-2262 (1-3 µM), which remain sufficiently high to achieve > 90% inhibition of hNa V 1.7, had no measurable effect on fast inactivation of hNa V 1.1 and hNa V 1.2. It is possible that elevated concentrations of ST-2262 result in a secondary mode of binding against these Na V subtypes, however, efforts have not been made to examine such a mechanism at this time. To our knowledge, changes in the rate of fast inactivation have not been observed with STX.
To investigate whether the potency of ST-2262 was dependent on the membrane holding potential or frequency of stimulus, an IC 50 value was measured against hNa V 1.7 using a two-pulse protocol with a pre-pulse to the voltage at half-inactivation (8 s step) and with a protocol that depolarizes the cell at high frequency (30 Hz stimulus). The potency of ST-2262 was not appreciably altered using either stimulation protocol (IC 50 Table 4). These results indicate that ST-2262 is a selective, use-independent inhibitor of hNa V 1.7.
Species variation in potency and mutagenesis. The potency of ST-2262 was assessed against a panel of species variants of Na V 1.7, including mouse, rat, and cynomolgus monkey (Suppl Table 5). Consistent with the hypothesis that Na V 1.7 potency is affected by the presence of the T1398/I1399 sequence motif in the DIII pore loop, the IC 50 of ST-2262 against cynoNa V 1.7 (0.101 µM, 0.073-0.140) was similar to human. In contrast, ST-2262 was > 50 × less potent against mouse (IC 50 = 3.78 µM, 3.23-4.43) and rat Na V 1.7 (IC 50 = 4.95 µM, 4.17-5.87) than the human ortholog. Affinity was restored within twofold of the hNa V 1.7 potency by introduction of domain III MD-TI mutations to mouse Na V 1.7 (IC 50 = 0.130 µM, 0.055-0.307; Fig. 3C, Suppl Table 6).
Multiple lines of evidence suggest that ST-2262 binds to the extracellular vestibule of the sodium channel (i.e., Site 1) including: (i) the structural similarity of ST-2262 to natural bis-guanidinium toxin ligands, (ii) the   Fig. 3D, Suppl Table 6). Collectively, these results indicate that ST-2262 binds to the extracellular vestibule of Na V 1.7, displaying significant species variation in potency and isoform selectivity in large part due to molecular interactions with residues T1398 and I1399, which are unique to human and nonhuman primate Na V 1.7 orthologs 31,32 .   www.nature.com/scientificreports/ Na V 1.7 in species that lack the T1398/I1399 sequence motif (Suppl Table 5). A NHP model of acute thermal pain was identified that uses a heat lamp to deliver a stimulus to the dorsal surface of the hand of lightly anesthetized cynomolgus macaques and measures the time to withdrawal 47 . Prior to advancing ST-2262 into the NHP acute thermal pain model, a standard battery of preclinical assays was completed to evaluate ADME and pharmacokinetic properties of this compound in cynomolgus macaques (Suppl Table 7). Off-target activity of ST-2262 using a commercially available radioligand binding assay panel against 68 different targets was also measured (LeadProfilingScreen, Eurofins, Taipei, Taiwan). No hits were identified on the off-target panel, defined as > 50% inhibition with 10 µM ST-2262 (Suppl Table 8).

ST-2262 increases withdrawal latency in a nonhuman
Male cynomolgus monkeys were anesthetized with propofol to a level in which the withdrawal reflex of the hand occurred at a consistent latency of approximately 3 s, a response time that was comparable to the detection of sharp pain from Aδ fibers when tested in prior studies on human volunteers 48,49 . The dorsal surface of the hand was exposed to a thermal stimulus that selectively activates Aδ-fiber nociceptors (Fig. 4A-C) 47,50 . The thermal stimulus was turned off at 5 s to prevent tissue damage. Heart rate was monitored throughout the study, and presentation of the noxious thermal stimuli consistently led to a transient increase in heart rate that peaked seconds after the stimulus and then returned to baseline (ΔHR). Acute noxious thermal stimuli transiently increase heart rate in human subjects; the percent change in heart rate correlates with subjective pain score 51 .
ST-2262 hydrochloride administered IV increased the withdrawal latency to noxious thermal stimuli (Fig. 4A). Efficacy was assessed in one subject at four dose levels (0.01, 0.05, 0.25, 1.25 mg/kg), in two subjects at the three higher dose levels (0.05, 0.25, 1.25 mg/kg), and in one additional subject at the highest dose level only (1.25 mg/kg). At the highest dose of 1.25 mg/kg, all four animals showed no hand withdrawal prior to the 5 s cut-off latency (Fig. 4A) Plasma samples were obtained from animals to assess the PK/PD relationship between drug exposure and thermal withdrawal latency. We found that 0.25 mg/kg ST-2262 resulted in ~ 1,400 ng/ml in plasma at the 5 min time point (n = 2), which corresponds to 7× the IC 50 value of ST-2262 against cynoNa V 1.7, corrected for plasma protein binding (cyno PPB = 73.5%). The unbound exposure of drug was reduced to 3.4 × cynoNa V 1.7 IC 50 at the 30 min time point. At a dose of 1.25 mg/kg, the total plasma concentration was ~ 7,000 ng/ml at 5 min (n = 2), which corresponds to an unbound exposure of 32 × cynoNa V 1.7 IC 50 , and was maintained above 15× cynoNa V 1.7 IC 50 for over 100 min (Fig. 4C). Lumbar CSF samples collected from two animals receiving the 1.25 mg/kg dose indicated that ST-2262 was peripherally restricted, with CSF:plasma ratios < 10 -3 (n = 2; [ST-2262] 0.8, < 0.5 ng/ ml in CSF).
By adjusting radiant heat parameters, the noxious heat model can be used to selectively assess responses to cutaneous C-fiber nociceptor activation, which produces a burning pain in volunteers 48,49 . The effect of ST-2262 on C-fiber induced hand withdrawal and heart rate change was investigated on two cynomolgus subjects 47 . As with the Aδ nociceptive response, 1.25 mg/kg ST-2262 completely abolished the C-fiber-mediated hand withdrawal and ΔHR (Fig. 4D,E). Collectively, these results are consistent with the hypothesis that pharmacological block of Na V 1.7 reduces sensitivity to noxious heat, phenotypically analogous to studies of Na V 1.7 lossof-function in CIP patients 2 . In addition, analysis of the PK/PD relationship of ST-2262 in this model provides insight into the level of Na V 1.7 target occupancy that may be necessary to achieve a pharmacodynamic effect. Recognizing the limited number of animals tested due to the challenge of working with non-human primates, additional work is warranted to further define the relationship between pharmacological inhibition of Na V 1.7 and sensitivity to noxious thermal stimuli.

Discussion
The finding that humans lacking functional Na V 1.7 exhibit an inability to experience pain raises the intriguing possibility that selective inhibitors of Na V 1.7 may be potent analgesics [1][2][3] . In the present study, we describe the discovery and characterization of ST-2262, a selective pore blocker of hNa V 1.7 advanced through rational modification of a natural small molecule toxin lead, STX. In whole cell voltage clamp recordings, ST-2262 exhibited > 200-fold selectivity for hNa V 1.7 over hNa V 1.1-1.6 and hNa V 1.8. The selectivity of ST-2262 was not examined against hNa V 1.9, a channel subtype that is difficult to express in heterologously. hNa V 1.9 contains a residue in the domain I p-loop, S360, that confers resistance to STX and lacks the domain III threonine/isoleucine sequence motif that is essential for high potency of ST-2262 against hNa V 1.7. Thus, inhibition of Na V 1.9 by ST-2262 is unlikely 42 .
The properties of ST-2262 are in contrast to other preclinical and clinical inhibitors of Na V 1.7, which preferentially bind to an inactivated conformation(s) of the channel 52 . Mutagenesis experiments indicate that specific residues in the extracellular pore of Na V 1.7, including a two amino acid sequence variation in the domain III pore loop that is unique to primates, are required for ST-2262 binding to cyno-and human Na V 1.7 31,39 . These findings establish the extracellular vestibule of Na V 1.7 as a viable receptor site for the design of potent and selective channel inhibitors.
Whereas congenital insensitivity to pain in humans is the result of complete and permanent Na V 1.7 loss-offunction, inhibition by small molecule agents is incomplete and transient. This difference raises several important questions regarding the pharmacology of Na V 1.7: (1) is transient inhibition sufficient for analgesia, (2) what level of target engagement is required for efficacy, and (3)  www.nature.com/scientificreports/ is sufficient to produce analgesia 45 . Furthermore, recognizing that ST-2262 is a polar small molecule with low membrane permeability and therefore unlikely to reach efficacious concentrations in the CNS (analysis of CSF samples obtained during NHP experiments gave a CSF:plasma ratio of < 10 -3 ), the observed effects on thermal withdrawal latency and ΔHR are likely the result of peripheral inhibition. Our findings, however, do not rule out an additional role for Na V 1.7 at the central terminals of primary afferents or in dorsal horn neurons, as has been suggested 53 .
In the present study, the effect of ST-2262 on withdrawal latency to noxious heat was measured in NHPs at doses of 0.01, 0.05, 0.25 and 1.25 mg/kg IV. Doses of 0.05, 0.25 and 1.25 mg/kg resulted in unbound plasma concentrations of ST-2262 of 0.7×, 3.4× and 16× the IC 50 value against cynoNa V 1.7 at a time point 30 min following drug administration. Assuming a unitary Hill coefficient, which is consistent with the dose-response relationship for ST-2262 in whole cell recordings against cyno-and human Na V 1.7, these unbound exposures correspond to 41%, 78% and 94% inhibition of Na V 1.7, respectively. Further work to understand whether a similar relationship exists between Na V 1.7 target occupancy and analgesic pharmacodynamic effects in other preclinical pain models is ongoing.  47 . In two subjects, the C-fiber-induced hand withdrawal response was replicable for testing. The efficacy endpoints measured were withdrawal latency (A) and heart rate change (B).