Research ArticlePain

Selective activation of TWIK-related acid-sensitive K+ 3 subunit–containing channels is analgesic in rodent models

See allHide authors and affiliations

Science Translational Medicine  20 Nov 2019:
Vol. 11, Issue 519, eaaw8434
DOI: 10.1126/scitranslmed.aaw8434

A pain-relieving task(3)

Potassium channels play a major role in determining neuronal excitability, and recent evidence suggests a major role for potassium currents in the pathophysiology of acute and chronic pain. In a new study, Liao et al. developed and characterized a selective agonist for TWIK-related acid-sensitive K+ 3 (TASK-3) channel called CHET3, with analgesic properties. Systemic administration of CHET3 in rodent models of acute and chronic pain had potent analgesic activity by reducing nociceptive neuron excitability. The results suggest that TASK-3 might be an effective therapeutic target for treating acute and chronic pain.


The paucity of selective agonists for TWIK-related acid-sensitive K+ 3 (TASK-3) channel, a member of two-pore domain K+ (K2P) channels, has contributed to our limited understanding of its biological functions. By targeting a druggable transmembrane cavity using a 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 homomers and TASK-3/TASK-1 heteromers. CHET3 displayed potent analgesic effects in vivo in a variety of acute and chronic pain models in rodents that could be abolished pharmacologically or by genetic ablation of TASK-3. We further found that TASK-3–containing channels anatomically define a unique population of small-sized, transient receptor potential cation channel subfamily M member 8 (TRPM8)–, transient receptor potential cation channel subfamily V member 1 (TRPV1)–, or tyrosine hydroxylase (TH)–positive nociceptive sensory neurons and functionally regulate their membrane excitability, supporting CHET3 analgesic effects in thermal hyperalgesia and mechanical allodynia under chronic pain. Overall, our proof-of-concept study reveals TASK-3–containing K2P channels as a druggable target for treating pain.


Currently available analgesics are not completely effective in treating chronic pain and/or they present severe adverse effects (1). Thus, discovering therapeutic targets to treat a variety of pain manifestations with similar potency but fewer side effects than μ-opioid receptor (μOR) are in urgent need. In this regard, two-pore domain potassium (K2P) channels hold great promise (2) because they give rise to leak K+ currents (3) and their activation in nociceptors can potentially inhibit pain signaling (46). The expression of TWIK-related acid-sensitive K+ 3 (TASK-3, Kcnk9) channel, a K2P channel, has been detected in the peripheral and central nervous systems (7, 8), including human dorsal root ganglia (9). Recent evidence has suggested that TASK-3 is involved in the perception of cold (10), and polymorphisms in the Kcnk9 gene are associated with breast pain in patients with breast cancer (11). However, its functional and anatomical involvement in chronic pain remains largely unknown. In addition, the paucity of selective agonists limits the drug target validation of TASK-3, leaving the notion that selective activation of TASK-3 alleviates pain, a question that remains to be tested experimentally. Here, we sought to discover selective activators for TASK-3 and to use them as tool compounds to reveal the translational potential and the underlying mechanisms of TASK-3 in treating pain.


Discovery of the selective activator CHET3 for TASK-3–containing channels

We set out to discover selective activators for TASK-3 via structure-based virtual screening. Because no crystal structure of TASK-3 has yet been determined, we sought to build a structural model using homology modeling. First, a crystal structure was chosen as the template. To this end, we applied Fpocket 2.0 server (12) to detect pockets in the reported crystal structure of K2P channels. In this computation, a druggability score greater than 0.5 (the threshold) means that the pocket might be druggable (12). We found that the cavity under the intracellular side of the filter and the nearby crevice between transmembrane segments 2 (TM2) and 4 (TM4) in four crystal structures [with Protein Data Bank (PDB) codes 4RUE, 3UKM, 4XDK, and 4XDL] (1315) had druggability scores greater than 0.5 (fig. S1). Thus, this cavity may be a drug-binding pocket. Among the four crystal structures, the structure of TWIK-related K+ channel 2 (TREK-2) (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 (3 × 10−32) in the sequence alignment generated using the BLAST program (blastp algorithm) and the Clustal Omega server (16, 17). Moreover, this TREK-2 structure stood out from the template-searching results (with the best quaternary structure quality estimation value of 0.66) in the SWISS-MODEL server (18, 19). Thus, the structural model of TASK-3 was built on the basis of this crystal structure with Modeller (20). Then, on the basis of this model, we performed virtual screening targeting the pocket (Fig. 1, A and B) with the SPECS and ChemBridge databases. A few hits were selected for the whole-cell patch-clamp electrophysiological tests in human embryonic kidney (HEK)–293 T cells overexpressing recombinant human TASK-3, which led to the discovery of a biguanide compound CHET3, a TASK-3 activator (Fig. 1C). Representative traces recorded using voltage ramp protocol are shown in Fig. 1D, and the time course of continuous traces recorded at 0 mV is shown in Fig. 1E. CHET3 enhanced TASK-3–mediated K+ currents by ~4-fold at 10 μM in comparison to the control [P < 0.001 (Fig. 1D), P < 0.001 (Fig. 1E)]. CHET3 activated TASK-3 with half-maximum effective concentration (EC50) of 1.4 ± 0.2 μM (Fig. 1F). CHET3-mediated currents could be significantly reversed by PK-THPP [P < 0.001 (Fig. 1D), P < 0.001 (Fig. 1E)]. PK-THPP is a selective inhibitor of TASK-3 against other K2P and K+ channels including K+ voltage-gated channel subfamily A member 5 (Kv1.5), human ether-à-go-go–related gene (hERG), and G protein–activated inward rectifier potassium channel 4 (KATP) (21). We further showed that 0.5 μM PK-THPP had no effect on voltage-gated K+ channel subfamily B member 1 (Kv2.1) and large-conductance Ca2+-activated K+ (BK) channels (fig. S2). In single-channel recordings by inside-out patches, CHET3 increased the channel open probability of TASK-3, with no change in single-channel conductance (Fig. 1, G to K), further supporting that CHET3 directly activated TASK-3.

Fig. 1 Structure-based ligand discovery of CHET3.

(A) Pocket in a TASK-3 homology model used for virtual screening. Representative docking poses are shown. (B) Cytoplasmic expanded view of the pocket and the docking poses. (C) Chemical structure of CHET3. (D and E) Example whole-cell path-clamp recordings and quantifications showing the activation of TASK-3 by 10 μM CHET3 and blockade of 0.5 μM PK-THPP [(D) n = 6, (E) n = 5; paired t test]. (F) CHET3 dose-response curves for TASK-3 (n = 7) and TASK-3/TASK-1 (n = 5), and the effects of 10 μM CHET3 on TASK-1 (n = 7, P = 0.569, paired t test for 10 μM CHET3 compared with the control). (G and H) Representative single-channel current traces from inside-out patches showing the activation of TASK-3 by CHET3 at −60 mV (G) and +60 mV (H). (I) Amplitude histograms for the patch recording shown in (G and H), which were fitted by Gaussian distributions. (J and K) Analysis of channel conductance changes and NPo (channel number times open probability) for the single-channel recordings (n = 9; paired t test). (L) Summary for the effects of CHET3 on several other K2P channels as well as hERG, Kv2.1, BK, TRPM8, and TRPV1 channels [n = 5 to 10; P < 0.001 (TASK-3 versus other channels), one-way ANOVA with Dunnett’s post hoc test]. Data in (D) to (F) and (J) to (L) are shown as means ± SEM. n.s., not significant, ***P < 0.001.

In addition to forming homomer channels, the TASK-3 subunit can efficiently form heteromer channels with the TASK-1 subunit (22). Electrophysiological assays showed that CHET3 activated TASK-3/TASK-1 (23, 24) with an EC50 value of 2.5 ± 0.2 μM (Fig. 1F). CHET3 activation could also be significantly blocked by PK-THPP (P < 0.001; fig. S3A). However, CHET3 did not significantly increase the currents of TASK-1 channels up to 10 μM in comparison to the control (P = 0.569; Fig. 1F and fig. S3B). Thus, CHET3 is an activator specific for the TASK-3 homomers and TASK-3/TASK-1 heteromers, 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 homomers and TASK-3/TASK-1 heteromers.

Next, we further examined the selectivity of CHET3. Electrophysiological assays revealed that 10 μM CHET3 has selectivity against several K2P channels, including TREK-1, TREK-2, TWIK-related arachidonic acid-stimulated K+ channel, TWIK-related spinal cord K+ channel (TRESK), and tandem pore domain halothane-inhibited K+ channel 1 (Fig. 1L and fig. S3B). Furthermore, 10 μM CHET3 has selectivity against hERG, Kv2.1, and BK channels, which are three K+ channels sharing similar filter structure and dynamics with K2P (25), as well as transient receptor potential cation channel subfamily M member 8 (TRPM8) and TRP cation channel subfamily V member 1 (TRPV1) (Fig. 1L and fig. S3, C to G).

We also excluded agonist or antagonist functions of CHET3 on pain-related G protein–coupled receptors (GPCRs) by testing the cellular function of μOR, 5-hydroxytryptamine receptor 1B (5-HT1BR), and cannabinoid receptor type 1 (CB1R) upon treatment with 10 μM CHET3 (fig. S4). 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. S5), which revealed the predominant binding mode of CHET3 within the pocket (Fig. 2, A and B). We next examined the ligand-channel interactions in this binding mode using mutagenesis experiments and RosettaLigand (26, 27). Residues T93 and T199 indirectly interacted with CHET3 by water bridges (Fig. 2, A and B). The two residues belong to the filter region, and T93A and T199A mutations led to nonfunctional channels (fig. S6, 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 contributing to 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, whereas the T198A mutant should not (Fig. 2C). A saturating concentration of CHET3 (10 μM) showed no significant activation on F125A (P = 0.502), I235A (P = 0.936), and F238A (P = 0.135) mutants in comparison to the control (Fig. 2D). CHET3 (10 μM) displayed reduced activation on L239A (P = 0.002) and I118A (P = 0.005) compared with the wild type (WT), whereas CHET3 activated T198A similarly to the WT (P = 0.673) (Fig. 2D). Mutant L232A is nonfunctional (fig. S6C). The experimental results on mutants are consistent with the computational predictions. To gain further insight into the action mechanism of CHET3, we carried out MD simulations on the apo (ligand-free) TASK-3 for comparison with the CHET3-bound TASK-3 (Fig. 2, E and F). In two of three independent simulations for the apo system, the channel selectivity filter stayed in a nonconductive-like conformational state (Fig. 2, E and F). In contrast, in all three simulations for the CHET3-bound system, the channel filter adopted a conductive-like state (Fig. 2, E and F). Furthermore, our simulations supported the previous report by González et al. (28) that the H98 residue plays a role in modulating the extracellular ion pathway in TASK-3. In the simulations of the CHET3-bound TASK-3, all the H98 residues adopted a conformation opening the extracellular ion pathway (Fig. 2E and fig. S7A). In contrast, in the simulations of the ligand-free mode, three of the six H98 residues adopted a conformation closing this pathway (Fig. 2E and fig. S7B).

Fig. 2 Activation mechanism of CHET3 on TASK-3.

(A and B) Three-dimensional and two-dimensional diagrams showing interactions between CHET3 and TASK-3. Hydrogen bond (red dash) and π-π interaction (green dash) are shown. (C) Computations showing the contributions of seven residues and their mutations to CHET3 binding (n = 10, top 10 docking models). EVDW represents energy distributions of the average van der Waals (VDW) interactions. Energy unit is Rosetta Energy Unit (R.E.U.). (D) Dose-response curves of six mutations on CHET3 activity (n = 6; paired t test for F238A, F125A, and I235A compared with the control; unpaired t test or Mann-Whitney test for T198A, I118A, and L239A compared with the WT). Data are shown as means ± SEM. (E) Selectivity filter conformations of the apo TASK-3 and the CHET3-bound TASK-3 revealed by MD simulations, including bound potassium ions (blue spheres), carbonyl oxygen (red sphere) rotation of filter residues, and movements of residue H98 (purple sticks). (F) Schematic representation of the action mechanism of CHET3 on TASK-3.

CHET3-induced analgesia in rodents

Next, we systematically evaluated CHET3 in analgesia. The antinociceptive effect of CHET3 was first assessed in the tail immersion test at 52°C in mice. CHET3 displayed dose-dependent analgesia with a fast onset (30 min) after intraperitoneal injection, with a maximal effect at a dose of 10 mg/kg (Fig. 3A). Here, 10 mg/kg intraperitoneal injection was used for most of the following animal studies. CHET3 was only effective in response to a noxious cold stimulus (5°C) or a noxious heat stimuli (46° and 52°C) but not to physiological stimuli (20° and 40°C) (Fig. 3A). The CHET3 analgesia was fully blocked by the coadministration of PK-THPP, and PK-THPP alone also produced a nociceptive effect in the tail immersion test at 46°C (Fig. 3A). Next, both the early and late phases (29) of acute inflammatory pain induced by formalin were attenuated by CHET3 (Fig. 3B), suggesting a peripheral effect of CHET3. The paw pressure test revealed that CHET3 reduced mechanical pain 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 (P = 0.026) reduced the frequency of hindpaw lifting (Fig. 3D), an indicator of spontaneous/ongoing pain behavior in the SNI model (30). In the cold plantar test, CHET3 attenuated the cold hyperalgesia during SNI development (SNI 7 days) and maintenance (SNI 14 and 21 days), which could be reversed by PK-THPP. PK-THPP alone, however, had no effect in the cold plantar test in SNI mice (Fig. 3E). CHET3 was more effective in relieving cold hyperalgesia than pregabalin, a first-line agent for the treatment of neuropathic pain (Fig. 3F). In SNI mice, CHET3 had no effect on alleviating mechanical allodynia in the von Frey test (fig. S8). However, in SNI rats, 10 mg/kg CHET3 attenuated mechanical allodynia throughout the different stages of chronic pain in the von Frey test, which could be reversed by PK-THPP (Fig. 3G). PK-THPP alone had no effect on pain in the von Frey test in SNI rats (Fig. 3G). The analgesic effect of CHET3 at a dose of 20 mg/kg exhibited a faster onset (30 min after injection) with similar efficacy to 10 mg/kg CHET3 (fig. S8) 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).

Fig. 3 Analgesic effects of CHET3 in rodents.

(A) Left: Time course for dose-dependent analgesia by CHET3 in tail immersion test at 52°C (n = 6 to 10; unpaired t test or Mann-Whitney test); middle: CHET3 analgesia in tail immersion tests at different temperatures (n = 9 to 11; unpaired t test or Mann-Whitney test); right: summary for PK-THPP effects (n = 8 to 10; paired t test). (B) Summary for CHET3 analgesia in formalin (2.5%, 20 μl) test (n = 8; unpaired t test). (C) Effects of CHET3 and PK-THPP on the paw withdrawal latency in paw pressure test (measured at 45 min after injection, n = 8 to 10; unpaired t test). (D) Effects of CHET3 on spontaneous pain within a 5-min duration in mice (measured at 35 min after injection, n = 6 to 10; unpaired t test). (E) Left: Time course for the effects of CHET3 on cold hyperalgesia in mice (n = 9 to 11; paired t test or Wilcoxon signed rank test); middle: summary for the effect of CHET3 on cold hyperalgesia at different stages in SNI (n = 9 to 11; unpaired t test or Mann-Whitney test); right: summary for PK-THPP (n = 7 to 10; paired t test or Wilcoxon signed rank test). (F) Comparison of the effects of CHET3 and pregabalin (30 mg/kg, intraperitoneal) on cold plantar test in mice (n = 11 to 12; unpaired t test). AUC, area under the curve. (G) Left: Time course for the effects of CHET3 on mechanical allodynia in rats (n = 13; one-tailed paired t test or Wilcoxon signed rank test); middle: summary for the effects of CHET3 on mechanical allodynia at different stages in SNI rats (n = 8 to 13; one-tailed Wilcoxon signed rank test); and right: summary for the effects of PK-THPP (n = 8; Wilcoxon signed rank test). (H) Left: Time course for the effects of CHET3 on heat hyperalgesia (n = 10 to 11; paired t test or Wilcoxon signed rank test); right: summary for the effects of PK-THPP (n = 11; paired t test). (I and J) Effects of CHET3 on anxiety-like behaviors in elevated plus maze test (I) (n = 8 to 9; one-way ANOVA with Dunnett’s post hoc test) and in light/dark box test (J) (n = 6 to 8; one-way ANOVA with Dunnett’s T3 post hoc test). CHET3 (10 mg/kg) and PK-THPP (15 mg/kg) were administrated via intraperitoneal injections unless specified. Data are shown as means ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.001.

Chronic pain may induce anxiety (31, 32). Compared with sham mice, 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. The administration of CHET3 30 min before the test significantly (P = 0.0497 and P = 0.002) alleviated anxiety-like behaviors in both tests (Fig. 3, I and J). Together, our data suggest that CHET3 potently and efficaciously attenuated acute and chronic pain and pain-associated anxiety in rodents, and the analgesic effects of CHET3 could be pharmacologically blocked by the TASK-3 blocker PK-THPP. CHET3 had no effects on grip strength, rotarod, and open field tests (fig. S9, A to C), suggesting that CHET3 had no effect on the locomotion activities in mice. Because TASK-3 was found to be expressed in mouse carotid body type 1 cells (33), we also evaluated the possible side effects of CHET3 on respiratory and cardiovascular function in mice or rats. CHET3 was undistinguishable from vehicle on respiratory frequency, tidal volume, and minute ventilation in rats within 150 min after administration (fig. S9, D to G), suggesting that CHET3 has no effect on respiration. We monitored blood pressure and heart functions using echocardiography, and we did not observe any significant (P > 0.5) change in blood pressure (fig. S9, H to J) or heart functions including ejection fraction and fractional shortening (table S1) in a postinjection time window between 45 and 90 min, during which CHET3-induced analgesia peaked in most cases. We also monitored the change in body temperature after CHET3 systemic administration, and no hyperthermia or hypothermia was observed (fig. S9K).

On-target validation using chemical and genetic approaches

Was CHET3 truly targeting TASK-3 containing K+ channels as an analgesic? We next 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 and B), the dioxane ring may form a π-π interaction with the F125 residue. CHET3-1, in which the dioxane ring is replaced with an aromatic ring, should maintain the π-π interaction. CHET3-2 should lose the π-π interaction because the dioxane ring is 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 to that of CHET3, whereas that of CHET3-2 decreased (fig. S10). In accordance, CHET3-1 activated TASK-3 with an EC50 value of 0.5 ± 0.1 μM, whereas 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 act as an analgesic. CHET3-1 (10 mg/kg, intraperitoneal injection) attenuated cold hyperalgesia in SNI mice (Fig. 4C), mechanical allodynia in SNI rats (Fig. 4D), and heat hyperalgesia in CFA mice (Fig. 4E), and all of these effects could be reversed by PK-THPP (fig. S11). In contrast, CHET3-2 (10 mg/kg, intraperitoneal injection) was completely inactive in all the experiments above (Fig. 4, C to E).

Fig. 4 Validation of TASK-3 as analgesic target by using CHET3 derivatives.

(A) Chemical structures of CHET3 derivatives. (B) Dose-response relationships for the effects of CHET3-1 (n = 6) and CHET3-2 (n = 6) on TASK-3. (C to E) Effects of CHET3-1 and CHET3-2 on cold hyperalgesia (C) (n = 9 to 11; paired t test or Wilcoxon signed rank test), mechanical allodynia (D) (n = 7 to 8; paired t test or Wilcoxon signed rank test), and heat hyperalgesia (E) (n = 9 to 10; paired t test). Data are shown as means ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.001.

We also generated Kcnk9 gene knockout (TASK-3 KO) mice (fig. S12). Knocking out Kcnk9 should abolish the function of TASK-3 homomers and TASK-3/TASK-1 heteromers in vivo. In TASK-3 KO mice and 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), and von Frey tests in naïve animals (Pre SNI, Fig. 5D) did not reveal any significant (P > 0.05) 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). As expected, CHET3 was completely inactive in TASK-3 KO mice in all the tests described above (Fig. 5, A to G). Thus, using tool compounds and mouse genetics, we provide strong evidence showing that CHET3 targets TASK-3–containing channels, thus acting as an analgesic, and the loss of TASK-3 contributed to the generation/maintenance of both acute and chronic pain.

Fig. 5 Effects of systemic administration of CHET3 in TASK-3 KO mice.

(A and B) Tail immersion tests of CHET3 in TASK-3 KO mice (n = 9 to 12; paired t test). (C) Effects of CHET3 on paw pressure tests in TASK-3 KO mice (n = 8 for KO, n = 7 for WT; two-way ANOVA with Tukey’s post hoc test). Note that there is no change of baseline sensitivity in nociception for TASK-3 KO mice in (A) to (C). (D) Effects of CHET3 on mechanical allodynia in TASK-3 KO mice in SNI model (up-down method, n = 8 for KO, n = 10 for WT; Mann-Whitney test). (E) Effects of CHET3 on spontaneous pain activity of TASK-3 KO mice in SNI model (n = 6 to 7; two-way ANOVA with Tukey’s post hoc test). (F and G) Effects of CHET3 on cold plantar test (n = 6 to 7; paired t test and unpaired t test) in SNI model and Hargreaves test (n = 8 to 12; paired t test and unpaired t test) of CFA model in TASK-3 KO mice. Note that TASK-3 KO mice had shorter paw withdraw latencies in both tests (paired t test or Wilcoxon signed rank test) in basal conditions. Data are shown as means ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.001.

Distribution of TASK-3–containing channels in sensory neurons

Our pharmacokinetic profile of CHET3 (tables S2 and S3) showed a negligible brain concentration of CHET3 and a high concentration of CHET3 in the plasma in both the naïve and SNI 7-d mice, suggesting that CHET3 mainly acted peripherally. The peripheral effect of CHET3 is also supported by the fact that CHET3 attenuated the early phase of formalin-induced pain. These findings, along with the previous finding that TASK-3 in dorsal root ganglion (DRG) neurons mediates cold perception (10), strongly suggest that peripheral TASK-3–containing channels contribute largely, if not entirely, to the analgesic effects of CHET3. Therefore, we evaluated the TASK-3 functions/distributions in the 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 a positive control probe and a negative control probe (see Materials and 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. A much higher expression of Kcnk9 (~14% of total neurons) was found in TG (Fig. 6, A and B). In DRG, about 95% of Kcnk9+ neurons express the TASK-1 subunit, suggesting possible formation of the TASK-3/TASK-1 heteromers in DRG, and about 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). Furthermore, about 50% of Kcnk9+ neurons express tyrosine hydroxylase (TH), a marker for c-low threshold mechanoreceptors (c-LTMRs) predominantly found in nonpeptidergic nociceptors (35), whereas Kcnk9 rarely colocalizes with P2rx3 (P2X3), which labels mainly TH, isolectin B4 (IB4)+ nonpeptidergic nociceptors (36), nor does Kcnk9 colocalize with Ntrk2 (TrkB), a marker for Aδ-LTMRs (35). Thus, TASK-3 identifies a unique subpopulation of both peptidergic and nonpeptidergic nociceptive sensory neurons enriched in thermal sensors (TRPV1 and 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 agreement with a previous study (37), we found that Kcnk9 expression was down-regulated in SNI mice and CFA mice (fig. S13), further suggesting that the down-regulation of TASK-3–containing channels contributes to the generation/maintenance of chronic pain.

Fig. 6 Distribution of TASK-3 in DRG neurons.

(A and B) Images and quantifications showing Kcnk9 expression in sensory neurons using RNAscope (n = 7 sections from three mice) and TG (n = 8 sections from three mice) (Mann-Whitney test). Data in (B) are shown as means ± SEM. **P < 0.01. (C) Quantification of the cell sizes of Kcnk9+ neurons (n = 6 sections from three mice). (D) Representative images showing Kcnk9+ neurons expression in different subtypes of DRG neurons using RNAscope. (E) Bar graph summary for experiments in (D) (n = 4 to 9 DRG sections from three to eight mice for each condition). Data are shown as means ± SEM. (F) Schematic summary for the distribution of Kcnk9+ neurons in DRG.

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 DRG neurons (diameter of ~20 μm, cell capacitance of ~30 pF) based on the ISH data. To isolate K+ currents, we applied voltage ramps from −120 to −30 mV. In total, 89 cells were recorded, and 16 cells responded to CHET3 (21.0 ± 7.2%, 11 rats; Fig. 7A, left). In the CHET3-sensitive cells, CHET3 enhanced the whole-cell current density by about 18%, which could be further inhibited by PK-THPP by about 38% at −30 mV (Fig. 7A, middle and right). We subtracted the CHET3-sensitive current and found that this current was strongly outwardly rectifying and was negligible between −120 and −60 mV, making the reversal potential of the CHET3-sensitive current difficult to resolve (Fig. 7B, left). We further sought to isolate the current carried by TASK-3–containing channels by subtracting PK-THPP–sensitive current. We consistently found a similar profile for PK-THPP–sensitive current (fig. S14A), further suggesting the low basal conductance at hyperpolarized membrane potentials, whereas 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 driving force of K+ currents at hyperpolarized potentials, we increased the extracellular K+ concentration to 143 mM. Under this condition, the CHET3-sensitive current reversed at about 6.7 mV, which was close to the theoretical value of 1.5 mV for K+ conductance (Fig. 7B, right).

Fig. 7 Functional roles of TASK-3 in nociceptive neurons.

(A) Left: Quantification for CHET3-positive DRG neurons responding to CHET3 (n = 89 cells from 11 rats); middle: representative electrophysiological traces showing the effects of CHET3 (10 μM) and PK-THPP (1 μM) on K+ currents in DRG neurons; right: bar graph summary for experiments in “middle” (n = 9 cells in six rats; Friedman test with uncorrected Dunn’s post hoc test). (B) Representative traces showing CHET3-sensitive currents at different extracellular K+ concentrations (VRev was determined from n = 5 cells in three rats). (C) Representative traces and scatter plots showing the effects of vehicle, CHET3, or PK-THPP on resting membrane potential (RMP) changes (n = 12 cells for control and CHET3, n = 11 cells for PK-THPP in five rats for each condition; one-way ANOVA). (D and E) Traces and bar graph showing the effects of CHET3 on rheobase and firing frequency in nociceptive neurons [n = 12 cells in five rats; Wilcoxon signed rank test in (D), paired t test in (E)]. (F and G) Traces and bar graph showing the effects of coapplication of CHET3 and PK-THPP on rheobase and firing frequency in nociceptive neurons [n = 11 cells in three rats; Wilcoxon signed rank test in (F), paired t test in (G)]. (H) Individual Ca2+ imaging traces from small-sized DRG neurons in representative fields of view in response to heat (25° to 43°C) or cool (37° to 15°C). (I) Bar graph summary for experiments in (H) (heat, n = 143 cells in 11 coverslips for control, n = 99 cells in 9 coverslips for CHET3, Mann-Whitney test; cool: n = 87 cells in 9 coverslips for control, n = 46 cells in 9 coverslips for CHET3, Mann-Whitney test; both experiments were from three independent preparations from six rats). Data are shown as means ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.001.

The low-amplitude CHET3- or PK-THPP–mediated currents at about −60 mV suggests that the basal activity of TASK-3–containing channels around the resting membrane potential (RMP) range was low and, thus, that CHET3 or PK-THPP is unlikely to be able to regulate the RMP. To systematically evaluate the regulatory role of CHET3 on the excitability of nociceptive neurons, we first applied a cocktail solution containing menthol and capsaicin, two agonists for TRPM8 and TRPV1 (38, 39), respectively, to better identify the nociceptive neurons that likely express TASK-3–containing channels. Only neurons responding to the cocktail (fig. S14B) were studied in the subsequent experiments. Consistent with the low activity of TASK-3–containing channels at approximately −60 mV, the application of CHET3 or PK-THPP or vehicle (control) did not hyperpolarize the RMP; rather, they all slightly depolarized the membrane by ~2 mV with no significant (P = 0.87) difference among the three groups, suggesting that CHET3 or PK-THPP had no specific roles in altering RMP (Fig. 7C). Next, we explored how CHET3 regulates action potentials (APs). In 12 of 27 neurons, the application of CHET3 markedly increased the rheobase required to elicit the APs by ~70% and decreased the frequency of APs evoked by suprathreshold current injections by ~65% (Fig. 7, D and E). In the other 15 cells, CHET3 had no effect on the rheobase and slightly increased the frequency of APs evoked by suprathreshold current injections by 10% (fig. S14, C and D). In 7 of these 12 CHET3-sensitive cells, we were able to further apply PK-THPP, which reversed the effects of CHET3 (fig. S14, E and F). Furthermore, in another independent set of experiments, we coapplied CHET3 and PK-THPP in naïve cells. In 11 of 27 cells, the coapplication of CHET3 and PK-THPP markedly decreased the rheobase by ~40% and increased the frequency of APs evoked by suprathreshold current injections by ~50% (Fig. 7, F and G); in the other 16 cells, coapplication 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. S14, G and H). Collectively, our electrophysiological data suggest the functional presence of K+ currents mediated by TASK-3–containing channels, 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 with an F340/F380 ratio ≥0.2 were considered responding cells shown in black, and those with an F340/F380 ratio <0.2 were considered nonresponding cells shown in gray). We confirmed that these Ca2+ signals were temperature dependent and were mediated by TRP channels, because the heat-induced responses could be blocked by 5 μM AMG9810 (TRPV1 antagonist) (fig. S15, A and B) (40), and the cool-induced responses could be blocked by 10 μM BCTC (TRPM8 antagonist) (38) and 20 μM HC030031 (TRPA1 blocker) (fig. S15, A and B) (41). Bath application of 10 μM CHET3 significantly (P = 0.008 and P < 0.001) and markedly inhibited the Ca2+ signals evoked by cool or heat stimulation in small-sized DRG neurons (Fig. 7, H and I), suggesting that the activation of TASK-3–containing channels was able to lower the excitability of the nociceptive neurons in response to external thermal stimulations.


The current study revealed three major findings: First, we discovered selective agonists for TASK-3–containing channels by targeting a transmembrane cavity under the selectivity filter using 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 role of peripheral TASK-3–containing channels in controlling the excitability of nociceptive neurons.

Very recently, Schewe et al. (25) reported a class of negatively charged activators (NCAs) that could activate K2P channels, the hERG channel, and the BK channel and revealed that the site below the selectivity filter is the binding site of the NCAs. In the present work, our virtual screening identified CHET3, a noncharged compound that acts on this site, further supporting the finding that the site below the selectivity filter is a ligand binding site. Note that NCAs are nonselective activators for a variety of K+ channels, whereas CHET3 is highly selective for TASK-3–containing channels, suggesting the versatility of this binding site. In addition, NCAs and CHET3 may share some common activation mechanisms on K2P channels, because they both influence the conformation of the selectivity filter. The activation mechanism we describe in this study does not fully explain the selectivity of CHET3.

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 induce modifications to other genes. These off-target genetic side effects discredit target validation work. This is particularly the case in the field of pain medicine. For example, Nav1.7-null mice and humans exhibited insensitivity to pain, whereas potent selective antagonists have weak analgesic activity (42, 43). Another example related to the K2P field is that migraine-associated TRESK mutations lead to the inhibition of TREK-1 and TREK-2 through frameshift mutation-induced alternative translation initiation to increase sensory neuron excitability and are linked to migraine (44). Using chemical probes to validate targets paves an alternative way for later translational research. Regarding in vivo applications of chemical probes in target identification and validation, a major issue is whether the observed phenotypes are relevant to the on-target of the probes. In this study, we provide three independent lines of evidence showing that CHET3 targets TASK-3–containing channels to act as an analgesic. First, the TASK-3 inhibitor PK-THPP could block CHET3-induced analgesia. Second, two structurally similar analogs were discovered and used in the in vivo tests. CHET3-1, a TASK-3 activator structurally similar to CHET3, is bioactive in analgesia and could also be blocked by PK-THPP. CHET3-2, another analog that is highly structurally similar to CHET3, did not activate TASK-3 and was completely ineffective in all the analgesia tests. Last, CHET3 had no analgesic effect in TASK-3 KO mice in all the tests. Collectively, our data suggest that the on-target activity of CHET3 is linked 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 homomers and TASK-3/TASK-1 heteromers likely contribute to CHET3-induced analgesia for the following reasons: (i) Kcnk9 is highly colocalized with Kcnk3 in DRG, and (ii) TASK-3/TASK-1 heteromers have been found assembled efficiently and functionally in cerebellar granule cells (45), motoneurons (46), and carotid body glomus cells (47).

We found that CHET3 decreased the excitability without changing the RMP of nociceptive neurons. The lack of change in RMP could be explained by the fact that CHET3- or PK-THPP–mediated currents are negligible at about −60 mV. One may argue that there may be 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 because PK-THPP, a molecule that is structurally distinct from CHET3, is unlikely to produce 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 (48). Although the use of CHET3 and its derivatives as preclinical candidate compounds requires further assessment with systematic nonclinical safety tests performed according to 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 the respiratory and cardiovascular system, where TASK-3–containing channels are also expressed.

The current study has a few limitations. First, the current structure-function data are insufficient to explain the insensitivity of TASK-1 to CHET3, a channel that has a high structural similarity with TASK-3 in the binding site of CHET3. Second, the role of TASK-3 in the regulation of the RMP requires further investigation using complementary methods such as TASK-3 fluorescent reporter mice. Third, CHET3 alleviated mechanical allodynia using the von Frey test only in SNI rats but not in SNI mice. The involvement of TASK-3 in mechanical allodynia needs to be further assessed with other experimental approaches and on other species. Fourth, further studies are needed to evaluate the translational potential of TASK-3 activation (TASK-3/TASK-1) in TG to treat chronic pain related to trigeminal neuralgia and migraine. Last, although TASK-3 is expressed in human DRG (9) and variation in KCNK9 is involved in breast pain in patients with breast cancer (11), direct evidence for the functional involvement of TASK-3 in pain signaling in humans is still lacking. Future functional studies on human tissue or studies with genetic screening of TASK-3–related mutations in humans would greatly aid in assessing the translational potential of TASK-3 for treating pain in humans.


Study design

Rationale and design of study. 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. The 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, ISH with the 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 K+ currents, action potential firing, and sensitivity to thermal stimulation in nociceptive neurons.

Randomization. Animals were randomly assigned, and the experimenter was not blinded.

Sample size and replicates. For single-cell–based experiments, at least five cells per condition were tested. For in vivo studies in animals, 6 to 10 animals per condition were used. Sample sizes were determined on the basis of prior experience with similar studies and pilot experiments.

Statistical analysis

Statistical analyses were carried out using Origin 9.0 software (Origin Lab Corporation). Data were analyzed as described in the figure legends. Differences in measured variables were assessed by using a one-tailed or two-tailed Student’s t test or nonparametric test for single comparisons, and one-way or two-way analysis of variance (ANOVA) or nonparametric test followed by indicated post hoc corrections for multiple comparison testing. The normality of the data distribution was determined using the Shapiro-Wilk test before appropriate statistical methods were chosen. No statistical methods were used to predetermine sample sizes. Exact P values are given in data file S1.


Materials and Methods

Fig. S1. A potential druggable pocket identified in several structures of K2P channels.

Fig. S2. Selectivity of PK-THPP against hERG, Kv2.1, and BK channels.

Fig. S3. Representative current traces of whole-cell recordings from several K2P channels and other ion channels.

Fig. S4. Effects of 10 μM CHET3 on three pain-related GPCRs.

Fig. S5. Binding modes of CHET3 suggested by docking and MD simulations.

Fig. S6. Characterization of three TASK-3 mutants.

Fig. S7. Conformations of the extracellular ion pathway in MD simulations.

Fig. S8. Analgesic effects of CHET3 on mechanical allodynia in SNI mice or rats.

Fig. S9. Effects of CHET3 (10 mg/kg, intraperitoneal injection) on the locomotion activities, respiration, blood pressure, and body temperature in rodents.

Fig. S10. Comparison of the binding of CHET3, CHET3-1, and CHET3-2.

Fig. S11. Blockade of CHET3-1 (10 mg/kg, intraperitoneal injection) analgesia by PK-THPP (15 mg/kg, intraperitoneal injection).

Fig. S12. Generation and characterization of TASK-3 gene (Kcnk9) KO mice.

Fig. S13. Expression of peripheral TASK-3 under chronic pain.

Fig. S14. Effects of CHET3 and PK-THPP on nociceptive neurons.

Fig. S15. Effects of blockers of TRP channels on thermal stimulation–induced Ca2+ signals of DRG neurons.

Table S1. Echocardiographic evaluation after CHET3 administration (10 mg/kg, intraperitoneal injection) in mice.

Table S2. CHET3 pharmacokinetics in plasma and brain after a single intraperitoneal administration to naïve male C57BL/6 mice.

Table S3. CHET3 pharmacokinetics in plasma and brain after a single intraperitoneal administration to SNI 7-d male C57BL/6 mice.

Data file S1. Raw data.

References (4952)


Acknowledgments: We thank Bioray Laboratories for technical support in preparing Kcnk9 KO mice. We thank T. Li (West China Hospital) for technical assistance with the blood pressure test; L. Gong (SIMM) and C. Zhou (West China Hospital) for technical assistance with the whole-body plethysmography; J. Zhang, Z. Yang, and L. Bai (Histology and Imaging platform, Core Facility of West China Hospital) for assistance with acquiring some ISH images; and R. Cui (SIMM) for assistance on the PK test. We thank the support of ECNU Multifunctional Platform for Innovation (001 and 011). Funding: Supported by the National Science and 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. and 31600832 to R.J.), Thousand Talents Plan in Sichuan Province (to R.J.), 1.3.5 Project for Disciplines of Excellence of West China Hospital of Sichuan University (ZY2016101 to J.L.), the “XingFuZhiHua” funding of ECNU (44300-19311-542500/006 to H.Y.), 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) (no. U1501501 to H.Y.), and the State Key Laboratory of Bioorganic and Natural Products Chemistry (to H.Y.). Author contributions: P.L. performed behavioral tests, ISH, and Ca2+ imaging. Y.Q. performed drug design and computations. Y.M. performed mouse genetics on TASK-3 KO mice, behavioral tests, and whole-body plethysmography. J.F. performed electrophysiology. Z.S. performed electrophysiology, behavioral tests, and ISH. L.H. performed electrophysiology. S.B., Y.W., and B.S. performed Ca2+ imaging. J.-J.Z. and W.-G.L. performed elevated plus maze tests. Z.C. and N.P. assisted with behavioral tests and cell cultures. E.-Y.S. performed dark/light box tests. L.Y. assisted with behavioral 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 behavioral 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 coauthors. Competing interests: H.Y., R.J., and Q.Z. are inventors on patent applications (201811218997.2, 201910109889.X, and PCT/CN2018119640) submitted by West China Hospital of Sichuan University, East China Normal University, and Shaoxing ZeroIn Biomedicines Co. Ltd. that cover the potential usage of CHET3 and its derivatives. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article