ABSTRACT
The paucity of selective agonists for TASK-3, a member of two-pore domain K+ (K2P) channels, limited our understanding of its biological functions. Targeting a novel druggable transmembrane cavity using structure-based drug design approach, we discovered a biguanide compound CHET3 as a highly selective allosteric activator for TASK-3-containing K2P channels including TASK-3 homomer and TASK-3/TASK-1 heteromer. CHET3 displayed unexpectedly potent analgesia in vivo in a variety of acute and chronic pain models in rodents that could be abolished by pharmacology or genetic ablation of TASK-3. We further found that TASK-3-containing channels anatomically define a unique subset population of small-sized, TRPM8, TRPV1 or Tyrosine Hydroxylase positive nociceptive sensory neurons and functionally regulate their membrane excitability, supporting the CHET3 analgesia in thermal hyperalgesia and mechanical allodynia under chronic pain. Overall, our proof-of-concept study reveals TASK-3-containing K2P channels as a novel druggable target for treating pain.
One Sentence Summary Identification of a novel drug target and its new hit compounds for developing new-generation non-opioid analgesics.
INTRODUCTION
Currently available analgesics do not treat pain completely and some of them, particularly opioids, provoke social problems (1). It is an urgent need to discover new therapeutic targets for developing new-generation analgesics. In particular, targets that treat a variety of pain with similar potency but less side effects compared to μ-opioid receptor are keenly awaited. In this regard, two-pore domain K+ (K2P) channels hold great promise (2), since they produce background leak K+ currents (3) and the activation of which in nociceptors theoretically inhibits pain signaling (4–6). The expression of TASK-3 (Kcnk9), a K2P channel, has been detected in peripheral and central nervous system (7, 8) including in human dorsal root ganglia (9). Recent evidence suggested that TASK-3 is involved in perception of cold (10), and the variations in Kcnk9 gene are associated with breast pain in breast cancer patients (11). However, its functional and anatomical involvement in chronic pain remain largely unknown. More importantly, the paucity of selective agonists limits the drug target validation of TASK-3, leaving the notion that selective activation of TASK-3 alleviates pain hypothetical. Here we sought to discover selective activators for TASK-3 and to use the activators as tool compounds to reveal and the translational potentials and the underlying mechanisms of TASK-3 in treating pain.
RESULTS
Discovery of selective activator CHET3 for TASK-3-containing channels
We set out to discover selective activators for TASK-3 via structure-based virtual screening. Since no crystal structure of TASK-3 has been determined yet, we sought to build a structural model using homology modeling. Firstly, a crystal structure was chosen as the template. To this end, Fpocket 2.0 server (12) was applied to detect pockets in the reported crystal structures of K2P channels. In this computation, druggability score more than 0.5 (the threshold) means the pocket might be druggable. We found that the cavity under the intracellular side of the filter and the nearby crevice between TM2 and TM4 in four crystal structures (with PDB codes 4RUE, 3UKM, 4XDK and 4XDL) (13–15) have druggability scores more than 0.5 (fig. S1).Thus, this cavity may be a drug binding pocket. Among the four crystal structures, the structure of TREK-2 channel (PDB code 4XDL) is a suitable template for building the structural model of TASK-3, because this structure has good sequence identity (31%) and expectation value (3E-32) in the sequence alignment generated using BLAST program (blastp algorithm) and Clustal Omega server (16, 17). Moreover, this TREK-2 structure stood out from the template searching results (with the best QSQE value: 0.66) in SWISS-MODEL server (18, 19). Thus, the structural model of TASK-3 was built based on this crystal structure with Modeller (20). Then, based on this model, we performed virtual screening targeting the pocket (Fig. 1, A and B) with SPECS and ChemBridge databases. A few hits were selected for the whole-cell patch-clamp electrophysiological tests in HEK-293 cells overexpressing recombinant human TASK-3, which led to the discovery of a biguanide compound CHET3, a novel TASK-3 activator (half-maximum effective concentration (EC50) 1.4 ± 0.2 μM, Fig. 1, C to F). CHET3 enhanced TASK-3-mediated K+ currents maximally by ∼4-fold which can be reversed by washout (Fig. 1D) or with pharmacological blockade by PK-THPP (86 ± 3% inhibition at 0.5 μM, n = 6 cells, representative traces shown in Fig. 1E), a TASK-3 specific inhibitor (21). In single-channel recordings by inside-out patches, CHET3 enhanced the channel openings mediated by TASK-3 (Fig. 1, G to K), further supporting that CHET3 directly activated TASK-3. Increases in open probability and conductance were observed in both the inward and outward directions in response to 3 μM CHET3 (Fig. 1, I to K).
In addition to forming homomer channels, TASK-3 subunit can efficiently form heteromer channels with TASK-1 subunit (22). Electrophysiological assays showed that CHET3 could activate TASK-3/TASK-1 (23, 24) with an EC50 value of 2.5 ± 0.2 μM with a reduced maximal efficacy of ∼2.4 fold (Fig. 1F), which were also blocked by PK-THPP (90 ± 2 % inhibition at 0.5 μM, n = 5 cells, representative traces shown in fig. S2A). However, CHET3 did not activate TASK-1 channels up to 10 μM (Fig. 1F and fig. S2B). Thus CHET3 is an activator specific for TASK-3 homomer and TASK-3/TASK-1 heteromer, two TASK-3-containing channels, with high selectivity against the structurally most related K+ channel TASK-1. In the subsequent sections, we use TASK-3-containing channels to represent TASK-3 homomer and TASK-3/TASK-1 heteromer.
Next, we further examined the selectivity of CHET3. Electrophysiological assays on several other K2P channels, including TREK-1, TREK-2, TRAAK, TRESK and THIK-1, further supported that CHET3 has high subtype selectivity among the K2P family (Fig. 1L and fig. S2B). Further, we found that 10 μM CHET3 has high selectivity against human ether-à-go-go–related gene (hERG) channel, voltage-gated K+ channel subfamily B member 1 (Kv2.1), large conductance Ca2+-activated K+ channel (BK), three K+ channels sharing similarity with K2P in filter structure and dynamics (25), and transient receptor potential cation channel subfamily M member 8 (TRPM8) and transient receptor potential cation channel subfamily V member 1 (TRPV1) (Fig. 1M, fig. S2, C to G).
We also excluded agonizing and antagonizing functions of CHET3 on pain-related G protein-coupled receptors (GPCRs) by testing the cellular function of μ-opioid receptor (μOR), 5-hydroxytryptamine receptor 1B (5-HT1BR) and cannabinoid receptor type 1 (CB1R) upon 10 μM CHET3 (fig. S3). Collectively, these results indicate that CHET3 is a selective activator of TASK-3-containing channels.
Activation mechanism of CHET3
Binding models derived from docking simulation were optimized by molecular dynamics (MD) simulations (fig. S4), which revealed the predominant binding mode of CHET3 within the pocket (Fig. 2, A and B). We next examine the ligand-channel interactions in this binding mode using mutagenesis experiments and RosettaLigand (26, 27). Residues T93 and T199 indirectly interacts with CHET3 by water bridges (Fig. 2, A and B). The two residues belong to the filter region, and the mutations T93A and T199A led to non-functional channels (fig. S5, A and B). F125 may form a π-π interaction with CHET3, and other surrounding residues, including I118, F125, T198, L232, I235, F238 and L239, likely contribute hydrophobic interactions with the ligand. RosettaLigand computations based on this MD mode predicted that the I118A, F125A, L232A, I235A, F238A and L239A mutants decrease CHET3 binding, while the T198A mutant should not (Fig. 2C). Indeed, a saturating concentration of CHET3 (10 μM) showed no activation on the mutants of F125A, I235A, F238A and reduced activation on mutant L239A, whereas CHET3 did activate T198A similarly to the WT (Fig. 2D). Mutant L232A is non-functional (fig. S5C). Although CHET3 did not show reduced activation on I118A (Fig. 2D), the experimental results are generally consistent with the computational predictions. To gain further insights into action mechanism of CHET3, MD simulations were carried out on the apo TASK-3 to compare with the CHET3-bound TASK-3 (Fig. 2, E and F). In two out of three independent simulations for the apo system, the channel selectivity filter tended to stay in a nonconductive-like conformational state (Fig. 2, E and F). By contrast, in all the three simulations for the CHET3-bound system, the channel filter adopted a conductive-like state (Fig. 2, E and F). Further, our simulations supported the previous report by González et al. that the residue H98 plays roles in modulating the extracellular ion pathway in TASK-3 (28). In the simulations of the CHET3-bound TASK-3, H98 adopted a conformation to open the extracellular ion pathway (Fig. 2E and fig. S6A). By contrast, in ligand-free mode, H98 has a high probability to adopt a conformation to close this pathway (Fig. 2E and fig. S6B).
CHET3-induced analgesia in rodents
Next, we evaluated CHET3 in analgesia systematically. The anti-nociceptive effect of CHET3 was firstly assessed by the tail immersion test at 52 ℃. CHET3 displayed dose-dependent analgesia with a fast onset (30 min) after intraperitoneal (i.p.) injection with a maximal effect at a dose of 10 mg/kg (Fig. 3A). Hereafter, i.p. injection with 10 mg/kg was used for most of the following animal studies. Interestingly, CHET3 was only effective in response to a noxious cold stimulus (5 °C) or noxious heat stimuli (46 °C and 52 °C) but not to physiological stimuli (20 °C and 40 °C) (Fig. 3A). Importantly, the CHET3 analgesia was fully blocked by the co-administration of PK-THPP, and PK-THPP alone also produced nociceptive effect in tail immersion test at 46 °C (Fig. 3A). Next, both the early and the late phases (29) of acute inflammatory pain induced by formalin were attenuated by CHET3 (Fig. 3B), suggesting at least a peripheral effect of CHET3. CHET3 reduced mechanical pain revealed by paw pressure test in mice, and the effect was fully blocked by PK-THPP (Fig. 3C). Next, we evaluated the analgesic effects of CHET3 on chronic pathological pain. In the spared nerve injury (SNI)-induced neuropathic pain mouse model, CHET3 significantly reduced the frequency of hind paw lifting (Fig. 3D), an indicator of spontaneous/ongoing pain behavior in SNI model (30). In the cold plantar test, CHET3 attenuated the cold hyperalgesia in SNI at the development (SNI 7 d) and maintenance (SNI 14 d and 21 d) stages of the chronic pain, which could be reversed by PK-THPP. PK-THPP alone, however, had no effect in cold plantar test in SNI mice (Fig. 3E). Compared to Pregabalin, a first line agent for the treatment of neuropathic pain, CHET3 was more effective in relieving cold hyperalgesia (Fig. 3F). In SNI mice, CHET3 has little effect on alleviating mechanical allodynia in the von Frey test. However, in SNI rats, CHET3 at the dose of 10 mg/kg attenuated the mechanical allodynia throughout the different stages of the chronic pain in the von Frey test, which could be reversed by PK-THPP (Fig. 3G). PK-THPP alone had no effect in the von Frey test in SNI rats (Fig. 3G). The analgesic effect of CHET3 at the dose of 20 mg/kg exhibited a faster onset (30 min post injection) with similar efficacy compared to these at 10 mg/kg (fig. S7) in the von Frey test in SNI rats. In chronic inflammatory pain induced by the Complete Freund’s Adjuvant (CFA), CHET3 reduced heat hyperalgesia in the Hargreaves test, which was blocked by PK-THPP (Fig. 3H). In addition, PK-THPP injection alone aggravated heat hyperalgesia (Fig. 3H).
Chronic pain may induce anxiety (31, 32). Compared with the Sham group, the SNI mice spent less time in open arms in the elevated plus maze test, and spent less time in the light box in the light/dark box test, suggesting anxiety-like behaviors in the SNI mice. Administration of CHET3 30 min before the test significantly alleviated anxiety-like behaviors in both tests (Fig. 3, I and J). Together, our data suggest CHET3 potently and efficaciously attenuated acute and chronic pain and pain-associated anxiety in rodents, and the analgesic effects of CHET3 can be pharmacologically blocked by the TASK-3 blocker PK-THPP. Importantly, CHET3 was inactive in grip strength, rotarod and open field tests (fig. S8, A to C), suggesting that CHET3 had no effect on the locomotion activities in mice. Since TASK-3 was found to be expressed in mouse carotid boy type-1 cells (33), we also evaluated the possible side effects of CHET3 on the cardiovascular functions in mice or rats. We monitored the blood pressure as well as heart functions using echocardiography, and we did not observe any significant change in blood pressure (fig. S8, D to F) or heart functions including Ejection Fraction (EF) and Fractional Shortening (FS) (table S1) in a post-injection time window between 45 min–90 min where CHET3-induced analgesia peaked in most cases. We also monitor the body temperature change following CHET3 systemic administration, and no significant hyperthermia or hypothermia was observed (fig. S8G).
Further on-target validation using chemical and genetic approaches
Was CHET3 truly targeting TASK-3 containing K+ channels to be analgesic? We then performed additional target validation experiments using chemical and genetic approaches. Medicinal chemistry yielded CHET3-1 and CHET3-2 (Fig. 4A), two derivatives of CHET3. In the CHET3-TASK-3 binding model (Fig. 2, A to B), the dioxane ring may form a π–π interaction with the residue F125. CHET3-1, where the dioxane ring was replaced with an aromatic ring, should keep the π–π interaction. CHET3-2 should lose it since the dioxane was replaced by a steric bulk tert-butyl group. Binding energy computations based on the binding model suggested that the binding affinity of CHET3-1 to TASK-3 was similar with CHET3, while that of CHET3-2 should decrease (fig. S9). In accord, CHET3-1 activated TASK-3 with an EC50 value of 0.5 ± 0.1 μM, while CHET3-2 was inactive (Fig. 4B), further supporting the putative binding model. We reasoned that CHET3-1 should be bioactive in analgesia whereas CHET3-2 should not, if CHET3 truly targets TASK-3-containing channels to be analgesic. Indeed, CHET3-1 attenuated cold hyperalgesia in SNI mice (Fig. 4C) and mechanical allodynia in SNI rats (Fig. 4D), and heat hyperalgesia in CFA mice (Fig. 4E), all these effects could be reversed by PK-THPP (fig. S10). In contrast, CHET3-2 was completely inactive in all the experiments above (Fig. 4, C to E).
We generated the Kcnk9 gene knockout (TASK-3 KO) mice (fig. S11). Knocking out Kcnk9 should abolish the function of TASK-3 homomer and TASK-3/TASK-1 heteromer in vivo. In the TASK-3 KO and their WT control mice, we measured the basal sensitivity to nociception, thermal hyperalgesia and mechanical allodynia, and we also evaluated the analgesic effect of CHET3 in these mice. Tail immersion (Fig. 5, A and B), paw pressure tests (Fig. 5C) or von Frey tests in naive animals (Pre SNI, Fig. 5D) did not reveal any significant difference in baseline nociceptive sensitivity between TASK-3 KO and WT; however, cold plantar (Pre SNI, Fig. 5F) and Hargreaves tests (Pre CFA, Fig. 5G) revealed increased nociceptive cold and heat sensitivity in TASK-3 KO mice. Furthermore, in the chronic pain models, von Frey, cold plantar and Hargreaves tests revealed that TASK-3 KO mice exhibited aggravated mechanical allodynia (Fig 5D), spontaneous neuropathic pain behavior (Fig. 5E), and thermal hyperalgesia (Fig. 5, F and G). CHET3, as expected, in TASK-3 KO mice, was completely inactive in all the tests described above (Fig. 5, A to G). Thus, using tool compounds (Fig. 4) and mouse genetics (Fig. 5), we provided strong evidence to show that CHET3 targets TASK-3-containing channels to be analgesic, and the loss of TASK-3 contributed to the generation/maintenance of both acute and chronic pain.
Distribution of TASK-3-containing channels in sensory neurons
Our pharmacokinetic profile of CHET3 (table S2) showed a negligible brain concentration (Cmax = 79.1 ng/mL) of CHET3 and a high concentration in the plasma (Cmax = 1112.0 ng/mL), suggesting CHET3 mainly acted peripherally. The peripheral effect of CHET3 was also supported by the fact that CHET3 attenuated the early phase of formalin-induced pain (Fig. 3B). These along with the previous finding that TASK-3 in dorsal root ganglion (DRG) neurons mediates cold perception (10) together strongly suggest that peripheral TASK-3-containing channels contribute largely, if not all, to CHET3 analgesia. Therefore, we evaluated the TASK-3 functions/distributions in peripheral nervous system, particularly in DRG.
We used fluorescence in situ hybridization (ISH) (RNAscope technique) to map the mRNA expression of TASK-3 in DRG and trigeminal ganglia (TG). The specificity of the fluorescent signals was validated by positive control probe and negative control probe (see methods). Kcnk9 was identified in a subset of neurons (∼7% of total neurons) in DRG, predominantly in small-sized neurons (diameter ≤ 20 μm) (Fig. 6, A and C), indicative of its specific expression in nociceptors. Interestingly, in TG, a much higher expression level of Kcnk9 (∼14% of total neurons) was found (Fig. 6, A and B). In DRG, approximately 95% of Kcnk9+ neurons express TASK-1 subunit, suggesting a possible formation of TASK-3/TASK-1 heteromer in DRG, and approximately 50% of Kcnk9+ neurons express TRPV1, a well-known noxious heat sensor predominantly expressed in peptidergic nociceptive sensory neurons (34). More than 95% of Kcnk9+ neurons express TRPM8, and very little Kcnk9+ neurons express TRPA1, two well-known noxious cold sensors (34). Further, approximately 50% of Kcnk9+ neurons express Tyrosine Hydroxylase (TH), a marker for c-low threshold mechanoreceptors (c-LTMRs) predominantly found in non-peptidergic nociceptors (35), whereas Kcnk9 rarely colocalized with P2rx3 (P2X3), which labels mainly TH- negative, IB4+ non-peptidergic nociceptors (36), nor did they colocalize with Ntrk2 (TrkB) a marker for Aδ-LTMRs (35). Thus, TASK-3 marks a unique subpopulation of both peptidergic and non-peptidergic nociceptive sensory neurons enriched in thermal sensors (TRPV1, TRPM8) or mechanoreceptors (TH+ c-LTMRs) (Fig. 6, D to F), in line with its functional involvement in thermal and mechanical sensation in vivo. In line with the previous study (37), we found that Kcnk9 expression was down regulated in SNI mice and CFA mice (fig. S12), further suggesting the down-regulation of TASK-3-containing channels contributes to the generation/maintenance of chronic pain.
Functional roles of TASK-3-containing channels in nociceptive neurons
The functional roles of TASK-3-containing channels were examined by whole-cell patch-clamp recordings in dissociated DRG neurons. Recordings were focused on small sized DRG neurons (diameter of ∼20 μm, cell capacitance of ∼30 pF) based on the ISH data. To isolate K+ currents, the voltage ramps from −120 mV to −30 mV were applied. In total 89 cells recorded, 16 cells responded to CHET3 (20.3 ± 6.3%, 11 mice). In the CHET3-sensitive cells, CHET3 enhanced the whole cell current density by approximately 18%, which could be further inhibited by PK-THPP by approximately 38% at −30 mV (Fig. 7A). We subtracted the CHET3-sensitive current, and we found this current was strongly outwardly rectifying, and was tiny between −120 mV and −60 mV, leaving the reversal potential of the CHET3-sensitive current difficult to be resolved (Fig. 7B). We further sought to isolate current carried by TASK-3-containing channels by subtracting PK-THPP-sensitive current, and consistently, we found a similar profile for PK-THPP-sensitive current (fig. S13A), further suggesting the low basal conductance at the hyperpolarized membrane potentials and strong outwardly rectifying represents an intrinsic property of K+ currents mediated by TASK-3-containing channels in DRG under our experimental conditions. To increase the drive force of the K+ currents in the hyperpolarized potentials, we increased the extracellular K+ concentration to 143 mM. Under this condition, the CHET3-sensitive current was reversed at around 6.7 mV, which was close to the theoretical value of 1.5 mV for K+ conductance (Fig. 7B).
The tiny CHET-3- or PK-THPP-mediated currents around −60 mV suggests that the basal activity of TASK-3-containing channels around the resting membrane potentials (RMP) range was low, and thus that CHET3 or PK-THPP was unlikely to be able to regulate the RMP. To systematically evaluate the regulatory role of CHET3 on the excitability of nociceptive neurons, first, we applied a cocktail solution containing Menthol and Capsaicin, two agonists for TRPM8 and TRPV1, respectively, to better identify the nociceptive neurons that likely express TASK-3-containing channels. Only neurons responding to the cocktail (fig. S13B) were furthered studied in the following experiments. Consistent with low activity of TASK-3-containing channels around −60 mV, application of CHET3 or PK-THPP or Vehicle (Control) did not hyperpolarize the RMP, rather, they all slightly depolarized it by ∼2 mV with no significant difference among the three groups, suggesting CHET3 or PK-THPP had no specific roles in altering RMP (Fig. 7C). Next, we explored how CHET3 regulates action potentials. In 12 out of 27 neurons, application of CHET3 markedly increased by ∼70% the rheobase required for eliciting the action potentials (APs) and decreased by ∼65% the frequency of APs evoked by suprathreshold current injections (Fig. 7, D and E), whereas in the other 15 cells, CHET3 had no effect on the rheobase and slightly increased by 10% the frequency of APs evoked by suprathreshold current injections (fig. S13, C and D). In 7 out of these 12 CHET3-sensitive cells, we were able to further apply PK-THPP, which reversed the effects of CHET3 (fig. S13, E and F). Furthermore, in another independent set of experiment, we co-applied CHET3 and PK-THPP in naive cells. In 11 out of 27 cells, the co-application of CHET3 and PK-THPP markedly decreased by ∼40% the rheobase and increased by ∼50% the frequency of APs evoked by suprathreshold current injections (Fig. 7, F and G), whereas in the other 16 cells, co-application of CHET3 and PK-THPP slightly increased the rheobase by ∼20% but had no effect on the APs frequency evoked by suprathreshold current injections (fig. S13, G and H). Collectively, our electrophysiological data suggest the functional presence of K+ currents mediated by TASK-3-containing channels, and the enhancement of which reduces the excitability of nociceptive neurons without affecting the RMP.
Last, Ca2+ imaging experiments were performed in acutely dissociated DRG neurons to measure how the activation of TASK-3-containing channels contributes to the thermal sensitivity of DRG neurons. Thermal stimulations elicited Ca2+ signals in a portion of small-sized DRG neurons (Fig. 7H, cells having ratio F340/F380 ≥ 0.2 were considered as responding cells shown in black, and those having ratio F340/F380 < 0.2 were considered as non-responding cells shown in grey). We confirmed that these Ca2+ signals were temperature-dependent and were mediated by TRP channels as the heat-induced responses can be blocked by 5 μM AMG9810 (TRPV1 antagonist) (fig. S14, A and B), and the cool-induced responses can be blocked by 10 μM BCTC (TRPM8 antagonist) and 20 μM HC030031 (TRPA1 blocker) (fig. S14, A and B). Bath application of 10 μM CHET3 significantly and markedly inhibited the Ca2+ signals evoked by cool- or heat-stimulations in small-sized DRG neurons (Fig. 7, H and I), suggesting the activation of TASK-3-containing channels was able to lower the excitability of the nociceptive neurons in response to external thermal stimulations.
DISCUSSION
The current study has three major findings: first, we discovered selective agonists for TASK-3-containing channels by targeting a transmembrane cavity under the selectivity filter using the structure-based approaches. Second, in vivo activation of peripheral TASK-3-containing channels displayed potent analgesia, suggesting a TASK-3-based therapeutic strategy for treating chronic inflammatory and neuropathic pain. Third, our anatomical and functional data highlight the roles of peripheral TASK-3-containing channels in controlling the excitability of nociceptive neurons.
Very recently, Schewe et al. reported a class of negatively charged activators (NCAs) that could activate K2P channel, hERG channel and BK channel and revealed that the site below the selectivity filter is the binding site of the NCAs (25). In the present work we obtained CHET3 acting on this site, a non-charged compound, by virtual screening, further supporting the finding that the site below the selectivity filter is a ligand binding site. It is noteworthy that NCAs are non-selective activators for a variety K+ channels while CHET3 is highly selective for TASK-3-containing channels, suggesting the versatility of this binding site. Plus, NCAs and CHET3 may share some common activation mechanisms on K2P channels by both influencing the conformation of selectivity filter. Notably, the activation mechanism we provided in this study could not fully explain the selectivity of CHET3 at present. In particular, TASK-1 and TASK-3 are closest relative to each other, and the residues below selectivity filter are conserved as well as H98. Further studies to elucidate the differential responses of TASK-1 and TASK-3 to CHET3 may be helpful for understanding the selective modulation principle in K2P.
In most cases, the initial proof-of-concept identification of a protein as a potential target is dependent on genetic methods. However, genetic deletion may produce modifications on other genes. This off-target genetic side effects discredit target validation work. This is particularly the case in the field of pain medicine: genetically mutated mice, e.g., Nav1.7-null mice, or human exhibited remarkable insensitivity to pain, whereas potent selective antagonists have weak analgesic activity (38, 39). Another example more related to K2P field is that migraine-associated TRESK mutations lead to the inhibition of TREK-1 and TREK-2 through frame shift mutation-induced alternative translation initiation (fsATI) to increase sensory neuron excitability and is linked to migraine (40). Using chemical probes to validate targets pave another way for the later translational research. For in vivo applications of chemical probes in target identification and validation, a major issue is whether the observed phenotypes are indeed relevant to the on-target of the probes. In this study, we provided three independent lines of evidence to show CHET3 targets TASK-3-containing channels to be analgesic. First, TASK-3 inhibitor PK-THPP could block the CHET-3 induced analgesia. Second, two structurally highly close analogs were discovered and used in in vivo tests. CHET3-1, a TASK-3 activator structurally highly close to CHET3, is bioactive in analgesia, which could be also blocked by PK-THPP. CHET3-2, another analog structurally highly close to CHET3 inactive on TASK-3, was completely inactive in all the analgesia tests. Last, CHET3 was completely inactive in analgesia in TASK-3 KO mice in all the tests. Collectively, our data suggest that the on-target activity of CHET3 links to the analgesic phenotypes.
Although CHET3 has a higher activation efficacy on TASK-3 over TASK-3/TASK-1, we suggest that both TASK-3 homomer and TASK-3/TASK-1 heteromer channels likely contribute to CHET3-induced analgesia based on the following reasons: 1) Kcnk9 is highly colocalized with Kcnk3 in DRG; 2) TASK-3/TASK-1 heteromer has been found assembled efficiently and functionally in cerebellar granule cells (41), motoneurons (42) and carotid body glomus cells (43).
We found that CHET3 decreased the excitability with no change in RMP of the nociceptive neurons. No change of RMP could be well explained by the fact that CHET3-or PK-THPP-mediated currents are negligible around –60 mV. One may argue that there may be a strong depolarizing “off-target” activity of CHET3 through another unknown channel/receptor, thereby masking the hyperpolarizing effect mediated by CHET3 on TASK-3 containing channels. However, if this were the case, one would at least expect PK-THPP to depolarize the RMP, since PK-THPP, a molecule structurally distinct to CHET3, is unlikely produce the hyperpolarizing “off-target” activity through the same unknown channel/receptor.
CHET3 acted mainly on peripheral TASK-3-containing channels. Peripheral targets are much less likely to produce central side effects including dependence/addiction. Although the utility of CHET3 and its derivatives as pre-clinical candidate compounds require to be further assessed with systematically non-clinical safety tests performed in GLP (Good Laboratory Practice) in rodents and other animals, it seems that the activation of peripheral TASK-3-containing channels does not produce obvious severe acute side effects on cardiovascular system where TASK-3-containing channels were also expressed. Interestingly, we found that TASK-3 was expressed in TG with a higher expression rate than DRG. Further studies are needed to evaluate the translational potentials of activation TASK-3 (TASK-3/TASK-1) in TG to treat chronic pain related to trigeminal neuralgia and migraine. Last, although TASK-3 was found expressed in human DRG (9) and variation of KCNK9 was involved in breast pain in breast cancer patients (11), direct evidence for functional involvement of TASK-3 in pain signaling in human remains lacking. Future functional studies on human tissues or studies with genetic screening of TASK-3-related mutations in human would greatly aid the assessment of the translational potential of TASK-3 in treating pain in human.
MATERIALS AND METHODS
Study design
Structure-based drug design methods were used to perform initial virtual screening and patch-clamp electrophysiology was mainly used to study the activity/mechanism of candidate compounds on TASK-3-containing channels. Analgesic effects of TASK-3 activators were then studied in acute and chronic pain models in mice and rats. Pharmacokinetic analysis was performed to assess how CHET3 was distributed. KO mice were used to confirm the on-target activity of CHET3. Last, in situ hybridization with RNAscope technique was used to map the distribution of TASK-3 in DRG and TG. The functional roles of TASK-3 were assessed by measuring how CHET3 and PK-THPP modulate the K+ currents, action potential firings and sensitivities to thermal stimulations in nociceptive neurons.
Sample size and replicates: For mutations of TASK-3 led to no-functional currents, 3 cells per mutation were tested. For other studies of ion channels in cells including these in cell line system and acutely prepared DRG cells, at least 5 cells per condition were tested. For experiments in DRG neurons, at least 3 independent preparations of DRG culture were performed. For in vivo studies with animals, 6-10 animals per condition were used. No power analysis was performed to determine the sample size.
Homology modeling for TASK-3 structure
Sequence alignment was generated by using Clustal Omega server (16). Notably, the two pore domain and selectivity filter sequence motif were highly conserved among the K2P channels, which were largely used to guide the alignment. Conserved residues E30 and W78 in TASK-3 helped to locate the position of non-conserved cap domain.
Virtual screening
Docking was performed by using Schrödinger Glide software (New York, NY, USA). Compounds were screened using the high-throughput virtual screening (HVS) module followed by a standard docking module SP in Glide. Glide G-score was used to rank the result list. Allowed for diversities of molecule structure, binding mode and drug-like properties, twelve hits were selected for bioassay.
Chemicals
PK-THPP was purchased from Axon Medchem. CHET3 purchased from commercial sources was used in the initial electrophysiological screening. Then CHET3 was synthesized in the lab for the following studies in this paper. Synthesis routes and characterization of CHET3 and its derivatives CHET3-1 and CHET3-2 are outlined in the Supplementary Materials.
For electrophysiology, stock solutions of CHET3 and derivatives (50 mM) were prepared in dimethyl sulfoxide (DMSO) and diluted in the extracellular solution before use.
For animal studies, CHET3 and PK-THPP were both dissolved in 10% DMSO, 5% tween80 and 85% saline, CHET3-1 was dissolved in 10% DMSO, 5% Castor Oil Ethoxylated, 35% Poly (Ethylene Glycol) and 50% corn oil, and CHET3-2 was dissolved in 14% DMSO, 5% tween80 and 81% saline. The solvents were used as vehicle controls.
Detailed modeling of the CHET3-TASK-3 binding poses
Initially, the configuration of CHET3 was determined by Ligprep module in the Schrödinger Maestro and Gaussain09 (Gaussian, Inc). Detailed descriptions are displayed in the Supplementary Materials. The configuration of tautomer with lowest energy was adopted to generate multiple ring conformations. CHET3 conformations were docked to the TASK-3 channel model by standard Glide as described for the virtual screening. Two binding modes (G-score values at −8.3 and −7.9, separately) were obtained from docking. In the best pose (1st model in fig. S4A), the guanidyl group in CHET3 establishes hydrogen bond with backbone NH of L232 residues in TM2, while the guanidyl group in the additional mode of binding (2nd model in fig. S4B) faces towards the selectivity filter and interacts with hydroxyl group of T199. To identify the accurate binding mode of CHET3, two models from docking were further studied using molecular dynamics (MD) simulations (see below).
MD simulations
The TASK-3 model obtained from the homology modeling and two binding models of CHET3-bound TASK-3 were used to build the models of the apo TASK-3 and the CHET3-bound TASK-3, respectively. Models were inserted in a POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) lipid bilayer to establish the CHET3-bound system and the apo system, respectively. MD simulations were carried out by using GROMACS 5.1.4 (44) with CHARMM36 parameters (45).
Comparison of the binding of CHET3, CHET3-1 and CHET3-2
RosettaLigand application (26, 27) was applied to dock CHET3, CHET3-1 and CHET3-2. The best binding mode obtained from MD simulations were adopted as the initial docking model. For each docking trial, top 1000 models were sorted by total score and the binding energy between three compounds and the channel were calculated. Also, in silico alanine scan were conducted by individually changing residue to alanine without otherwise changing the conformation of protein or ligands in Rosetta. To explore the distribution of binding interactions between compounds and proteins, the average energy of top 10 models with the lowest binding energies (interface score) were calculated. To compare the binding of CHET3, CHET3-1 and CHET3-2, top 50 models of each compound with the lowest binding energies were used to calculate the total score and interface score.
Electrophysiology
Electrophysiology tests of hTASK-3, hTASK-1, hTREK-1, mTREK-2, hTRAAK, hTHIK-1, hTRESK, hTASK-3/hTASK-1, hTRPM8 and hTRPV1 were done with transiently transfected HEK-293T cells. The cDNA of hTASK-3, hTASK-1, hTHIK-1, hTRESK and hTASK-3/hTASK-1 were subcloned into the pCDNA3 vector (Invitrogen). The cDNA of hTREK-1, mTREK-2, hTRAAK, hTRPM8 and hTRPV1 were subcloned into the pEGFPN1 expression vector (Invitrogen). For the hTASK-3/hTASK-1 concatemer products were designed for the 3’ and 5’ ends of TASK-3 and TASK-1 ensuring that the stop codon of TASK-3 was removed.
Electrophysiology tests of hERG, Kv2.1 and BK were done with stable cell lines. The CHO-hERG stable cell line was generated in-house and was based on a standard CHO-K1 cell line. The HEK293-human Kv2.1 stable cell line and the CHO-human BK stable cell line were generated by Ion Channel Explore (Beijing, China).
Whole-cell recordings of ion channels were performed with patch-clamp amplifiers (EPC10, HEKA or Axon 700B, Molecular Devices) at 23-25 °C. The current signals were filtered at 2 kHz and digitized at 10 kHz. The pipettes for whole cell recordings were pulled from borosilicate glass capillaries (World Precision Instruments) and had a resistance of 3-7 MΩ. For recordings of K+ channels, the standard pipette solution contained (in mM): 140 KCl, 2 MgCl2, 10 EGTA, 1 CaCl2, 10 HEPES (pH 7.3, adjusted with KOH), and the external solution contained (in mM): 150 NaCl, 5 KCl, 0.5 CaCl2, 1.2 MgCl2, 10 HEPES (pH 7.3, adjusted with NaOH). For recordings of the TRPV1 and TRPM8 currents, the internal solution contained (in mM): 140 CsCl, 0.1 CaCl2, 1 MgCl2, 10 HEPES, 5 EGTA (pH 7.2, adjusted with CsOH), and the external solution contained (in mM): 140 NaCl, 5 KCl, 1 MgCl2, 0.5 EGTA, 10 HEPES (pH 7.4, adjusted with NaOH). For recordings of hERG, the outward current of hERG channels was elicited by a 2.5-second depolarization step to +30 mV from a holding potential of −80 mV followed by a 4-second repolarization step to −50 mV to measure the tail current. For recordings of Kv2.1, currents were evoked by a 200-millisecond depolarization step to +60 mV from a holding potential of −80 mV. For recordings of BK, currents were evoked by a 1-second depolarization step to +70 mV from a holding potential of −80 mV. For recordings of TRPV1 and TRPM8, currents were recorded using a ramp protocol from −100 mV to +100 mV over a period of 400-millisecond at a holding potential of 0 mV.
Inside-out patches were done by using EPC10 (HEKA) at 23-25 °C. The pipettes for single-channel recordings had resistances of 7-15 MΩ. The standard pipette and bath solutions contained (in mM): 140 KCl, 1 CaCl2, 2 MgCl2, 10 HEPES, 10 EGTA (pH 7.4, adjusted with KOH). Currents were recorded at a pipette potential of −60 mV and +60 mV respectively.
Ethics statement
All experiments with animals were approved by the Animal Research Committee of East China Normal University (PROTOCOL No.: m20171020 and m20180112) and the Animal Research Committee of West China Hospital of Sichuan University (PROTOCOL No.: 2018175A). For tissue collection, mice were given a lethal dose of pentobarbital intraperitoneally.
Animals
BALB/c mice and Sprague-Dawley rats were used in most animal studies, and TASK-3 KO mice and WT control littermates were on a C57BL/6 background. Male mice or rats aged 8-10 weeks were used for behavior tests, unless stated otherwise. Animals were housed in a conventional facility at 21 °C on a 12 h light-dark cycle with unrestricted access to food and water.
TASK-3 KO mice generation
To generate a Kcnk9 knockout C57BL/6 mouse line by CRISPR-Cas9 genome editing system, two single-guide RNAs (sgRNA-1, 5′-CCGCTTCATGGCCGCGAAGAAGG-3′, and sgRNA-2, 5′-AGGAACCGGCGAATTTCCACTGG-3′) flanking exon1 were designed (Bioray Laboratories). A 241-bp deletion was bound to the exon 1 of the Kcnk9 gene locus, resulting Kcnk9Δ/Δ mice with a frameshift mutation. Additional information is provided upon request.
Spared nerve injury model
Unilateral spared nerve injury (SNI) surgery was performed. The experiment animal was laid on prone position. After disinfection with povidone iodide and 75% ethanol, a minimal skin incision is made at the mid-thigh level in order to exposing the sciatic nerve and its three branches via separating the muscle layers. The tibial and common peroneal nerves were tightly ligated with 5.0 silk threads and a 1-2 mm length section was removed between the proximal and distal parts of nerves. The sural nerve was restrictively preserved to avoiding any harmful injury. The muscle layer and skin were closed after surgery and animals were transferred to a warm pad to recover from anesthesia.
Chronic inflammatory pain model
A volume of 20 μL Complete Freund’s Adjuvant (CFA) (Sigma-Aldrich) was subcutaneously injected into the left hindpaw of mouse to induce chronic inflammatory pain in mouse. After injection the syringe was maintained for at least 30 s to avoid overflow.
Tail immersion
Mice were restrained in the test tube with the tails stretching out and moving freely 15 min twice daily for 3 days. The distal third tails were immersed into a water bath at 5 °C, 20 °C, 40 °C, 46 °C, or 52 °C. The three measurements of tail flick latency (in second) to stimulation as indicated by rapid tail flexion were averaged. A cutoff value of 15 seconds was adopted to prevent unexpected damage.
Formalin test
Mice were housed individually in Plexiglas chambers, after habituation to the testing environment for at least 30 min, the left hindpaw of the mice were injected subcutaneously with formalin (20 μL of 2.5% formalin, diluted in saline) and the mice were put into the chamber of the automated formalin apparatus where movement of the formalin-injected paw was recorded by an action camera (SONY, HDR-AS50). The number of paw flinches was counted at 5 min intervals for 60 min by a blind experimenter. Time spent exhibiting these pain behaviors was recorded for the first (0-10 min) and second phases (10-60 min).
Paw pressure
The effects of the mechanical nociception were evaluated with an Analgesimeter (model 37215; Ugo-Basile, Varese, Italy). Mice were placed in the testing room for 3 continuous days to acclimate environment. The hindpaw of mice was pressed with a constant pressure of 450 g using a cone-shaped paw-presser with a rounded tip and immediately stopped as soon as the animal showed a struggle response, and the reaction latency was recorded in second. The analgesic effects for TASK-3 agonists were evaluated 30 min after i.p. injection.
Spontaneous pain test
After 3 days acclimation, the SNI mice were placed in an elevated transparent cage (20 × 20 × 14 cm) with a wire mesh floor (0.5 × 0.5 cm). A 5 min duration was videoed by an action camera (SONY, HDR-AS50) for each mouse, and the number of left hindpaw flinching was calculated by a blind experimenter.
Cold plantar test
Mice were allowed to acclimate to the testing environment 2-3 h daily for 3 continuous days. The cold probe produced freshly with fine dry ice powder into a 5 mL syringe was held against a 6 mm depth of flat glass. The center of hindpaw was targeted and the withdrawal latency manifesting as quick flick or paw licking was recorded. A cutoff time of 30 s was used to prevent potential tissue damage.
von Frey test
The SNI rats and mice were individually placed in the chamber as described in the spontaneous pain test. Mechanical sensitivity was assessed by two methods:
Method 1 (for all the von Frey tests described except Fig. 5D): The mechanical paw withdrawal threshold was assessed using von Frey filaments with an ascending order. The tip of filament was perpendicularly targeted to the region comprising of the sural nerve territory and sufficient stimulation was held for 1 s. Rapid paw withdrawal or flinching was considered as a positive response and the bending force for which at least 60% of the application elicited positive response was recorded as the mechanical paw withdrawal threshold.
Method 2 (up-down method): Mechanical responses were tested by stimulating the region comprising of the sural nerve territory with von Frey monofilaments by using the up-and-down method, starting with 0.04 g. Biting, licking, and withdrawal during or immediately following the 3 s stimulus were considered as a positive response.
Hargreaves test
The hindpaw sensitivity to thermal noxious stimulus was assessed using a radiant heat source (model 37370; Ugo-Basile, Varese, Italy), stimulus intensity was set to produce an approximate latency of 10 s at baseline, and a cut-off value was set at 20 s to avoid unexpected damage. Mice were allowed to acclimate in Plexiglas chambers with a glass floor for 3 days, and the time to paw withdrawal was measured per mouse with a 5 min inter-stimulation period. Three trials were averaged to yield the withdrawal latency.
RNAscope in situ hybridization
Sequences of target probes, preamplifier, amplifier and label probes are proprietary and commercially available (Advanced Cell Diagnostics). In situ hybridization was performed on frozen DRG sections (10 μm) using RNAscope Multiplex Fluorescent Reagent Kit v2 (ACDbio, Cat#323100) and RNAscope 4-Plex Ancillary Kit for Multiplex Fluorescent Kit v2 (ACDbio, Cat#323120). The hybridization assay as described by vendor’s protocol. In situ probes include: Kcnk3 (Cat#534881), Kcnk9 (Cat#475681), Trpa1 (Cat#400211), Trpv1 (Cat#313331), Trpm8 (Cat#420451), Rbfox3 (Cat#313311), Th (Cat#317621), Ntrk2 (Cat#423611), P2rx3 (Cat#521611). The specificity of the fluorescent signals was validated by RNAscope 3-plex Positive Control Probe (Cat#320881) and RNAscope 3-plex Negative Control Probe (Cat#320871). Fluorescent images were taken using a NIKON A1R+MP Two-photon confocal scanning microscope and were analyzed using ImageJ software.
Acutely dissociated DRG neuron preparation and electrophysiology
3-6 week-old male Sprague-Dawley rats were sacrificed. The DRGs were collected in a 35-mm tissue culture dish and digested in 3% collagenase for 20 min, followed by 1% trypsin for another 30 min. After titrating by sucking up and down, the DRG neurons were cultured in Neurobasal containing 2% B27 medium for 2-4 h. The bath solution contained (in mM): 140 NaCl, 3 KCl, 1.3 MgCl2, 10 HEPES, 2.4 CaCl2, 10 glucose, pH 7.3. The pipette solution contained (in mM): 40 KCl, 10 HEPES, 5 EGTA, 10 NaCl, 95 K-gluconate, 4 Mg-ATP, pH 7.4. To minimize voltage-gated currents, the voltage ramps from −120 mV to −30 mV were applied, and 1mM CsCl was added extracelluarly to block hyperpolarization-activated currents. To determine the reversal potential of the CHET3-sensitive currents, NaCl was replaced with equimolar KCl. The whole-cell recordings were performed with similar hardware settings as described in electrophysiology in HEK-293 cells.
Acutely dissociated DRG neuron preparation and intracellular Ca2+ imaging
2 week-old male Sprague-Dawley rats were sacrificed. The DRGs were collected in a 35-mm tissue culture dish and digested in 2.5 mg/mL papain (Sigma-Aldrich) for 30 min at 37 °C, followed by 2.5 mg/mL collagenase (Sigma-Aldrich) for another 30 min. Digested ganglia were gently washed with neurobasal medium and mechanically dissociated by passage through pipet. Neurons were seeded on laminin-coated wells (Corning) and cultured overnight at 37 °C in 5% CO2 in 2% B27 (Sigma-Aldrich), supplemented neuronbasal medium, containing 50 ng/mL GDNF (PeproTech), 50 ng/mL BDNF (PeproTech).
Changes in intracellular Ca2+ concentration were monitored using ratiometric Fura-2-based fluorimetry. Neurons were loaded with 2 μM Fura-2-acetoxymethyl ester (Yeasen) dissolved in bath solution and 0.02% Pluronic F127 (Sigma-Aldrich) for 30 min at 37°C. Fluorescence was measured during alternating illumination at 340 and 380 nm using Olympus IX73 inverted fluorescence microscopy system. The bath solution contained (in mM): 138 NaCl, 5.4 KCl, 2 CaCl2, 2 MgCl2, 10 glucose, and 10 HEPES, pH 7.4, adjusted with NaOH. At the end of each experiment, cells were subjected to a depolarizing solution containing 50 mM KCl, and non-responsive cells to 50 mM KCl were excluded from analysis. Bar graphs in Fig. 7I and fig. S14B were pooled data from both responding cells and non-responding cells in different conditions.
Thermal stimulation
Coverslip pieces with culture cells were placed in a recording chamber and continuously perfused (about 1 mL/min).
Cool stimulation: the temperature was adjusted with ice bag cool perfuse solution, and controlled by a feedback device. Cold sensitivity was investigated with a ∼2 min duration ramp-like temperature drop from 37 °C to ∼15 °C.
Heat stimulation: the temperature was adjusted with a water-heated device (model TC-324B/344B, America), with the temperature of the perfuse solution raise, and controlled by a feedback device. Heat sensitivity was investigated with a ∼5 min duration ramp-like temperature rise from 25 °C to ∼43 °C.
Statistical analysis
Statistical analyses were carried out using the Origin 9.0 software (Origin Lab Corporation, Northampton, USA). Data were analyzed as described in the figure legends. Normality of the data distribution was determined before appropriate statistical methods were chosen. The drug was assessed as significantly active by using statistical tests performed between values of the baseline and those of given time points unless specified. No statistical methods were used to predetermine sample sizes.
Funding
This work was funded by National Science & Technology Major Project “Key New Drug Creation and Manufacturing Program” of China (2018ZX09711002 to H.J., Q.Z., and H.Y.), the National Natural Science Foundation of China (21422208 to H.Y; 31600832 to R.J.), Thousand Talents Plan in Sichuan Province (to R.J.), 1.3.5 Project for Disciplines of Excellence (ZY2016101), West China Hospital, Sichuan University (to J.L.), the “XingFuZhiHua” funding of ECNU (44300-19311-542500/006), the Fundamental Research Funds for the Central Universities (to H.Y., and 2018SCUH0086 to R.J.) and the Special Program for Applied Research on Super Computation of the NSFC-Guangdong Joint Fund (the second phase) under Grant No.U1501501, and the State Key Laboratory of Bioorganic and Natural Products Chemistry.
Author contributions
P.L. performed behavior tests, ISH, Ca2+ imaging; Y.Q. performed drug design and computations; Y.M. performed mouse genetics on TASK-3 KO mice and behavior tests; J.F. performed electrophysiology; Z.S. performed electrophysiology, behavior tests, ISH; L.H. performed electrophysiology; S.B., Y.W. and B.S. performed Ca2+ imaging; J.Z. and W.G.L. performed elevated plus maze tests; Z.C. and N.P. assisted with behavior tests and cell culture; E.S. performed dark/light box tests; L.Y. assisted with behavior tests; F.T., X.L. and Z.G. performed electrophysiology for some of the initial compound screenings; P.S., Y.C. and Y.M. performed pharmacokinetics study; D.H. performed the qPCR experiments for TASK-3 KO mice; L.Z. performed experiments of μOR; D.Y. performed experiments of 5-HT1BR; W.L. performed experiments of CB1R; T.Y., J.X. and Y.M. performed experiments of echocardiography. Q.Z. prepared the derivatives of CHET3. J.L. oversaw the animal behavior tests. H.J. oversaw the computations. R.J. and H.Y. initiated, supervised the project, analyzed the experiments, and wrote the manuscript with input from all co-authors.
Competing interests
The authors declare no competing interests.
Data and materials availability
All data is available in the main text or the supplementary materials.
Acknowledgements
We thank Bioray Laboratories for technical support in preparing Kcnk9 knockout mice. We thank Dr. Tao Li (West China Hospital) for technical assistance with blood pressure test and Miss Rongrong Cui (SIMM) for assistance on PK test. We thank the supports of ECNU Multifunctional Platform for Innovation (001 and 011).