Combining endocannabinoids with retigabine for enhanced M-channel effect and improved K_V7 subtype selectivity

Epilepsy is a disorder of the nervous system that affects an estimated 50 million people worldwide. Although the majority of patients with epilepsy are adequately helped by available treatment (Brodie et al., 2012), over 30% continue to have seizures (Brodie et al., 2012), which motivates major efforts to develop new anti-epileptic treatment strategies. Retigabine was approved by the US Food and Drug Administration in 2011 as the first anti-epileptic drug targeting the neuronal M-channel (Main et al., 2000; Gunthorpe et al., 2012), a potassium channel important for dampening neuronal excitability by contributing to the negative resting membrane potential. Retigabine showed anti-excitable and anti-epileptic effects in cellular and animal models as well as in clinical trials (Rostock et al., 1996; Rundfeldt, 1997; Brodie et al., 2010; French et al., 2011). Retigabine was used as an adjunctive treatment in patients with partial-onset seizures and had advantages, such as less drug interaction with other anticonvulsants, because it is not metabolized through the P-450 system (Hempel et al., 1999; Łuszczki, 2009). Unfortunately, retigabine was associated with adverse effects including bladder dysfunction and changes in retinal pigmentation (Splinter, 2012; Garin Shkolnik et al., 2014), which limited its clinical use. Retigabine was withdrawn from the market in 2017. Extensive studies using retigabine have established the M-channel as a target for new anti-epileptic drugs but have also highlighted the need to develop strategies to reduce adverse effects of anticonvulsants.

Metabolic instability of retigabine (Groseclose and Castellino, 2019), retigabine effects on M-channels in peripheral nervous tissue (Tykocki et al., 2019), and retigabine off-target effects on other K_V7 channel subtypes (Rode et al., 2010) all likely contribute to adverse effects produced by the drug. The human K_V7 (hK_V7; also called KCNQ) channel family contains a total of five voltage-sensitive potassium channels, hK_V7.1 to hK_V7.5, which are expressed in various tissues in the body (Robbins, 2001). The M-channel is primarily composed of hK_V7.2 and hK_V7.3 subunits (Wang et al., 1998), which coassemble as heterotetramers. In some neuronal compartments, hK_V7.5 contributes to the M-current (Shah et al., 2002). Retigabine’s anti-excitable effect is caused by retigabine binding to a tryptophan in the ion-conducting pore domain of the M-channel (W236 in hK_V7.2 and W265 in hK_V7.3; Schenzer et al., 2005; Wuttke et al., 2005; Kim et al., 2015), stabilizing the open conformation of the channel. Although hK_V7.1 lacks this tryptophan (rendering it insensitive to retigabine), the tryptophan is conserved in hK_V7.2–5, which are all activated by retigabine (Schenzer et al., 2005). This poses a potential risk of off-target effects on smooth muscle cells, outer hair cells, and retinal pigment epithelial cells, which have been shown to express hK_V7.4 and/or hK_V7.5 channels (Holt et al., 2007; Xu et al., 2007; Zhang et al., 2011; Brueggemann et al., 2014; Chadha et al., 2014; Provence et al., 2018).

A means to reduce the adverse effects of retigabine is to improve the metabolic stability and subtype selectivity of retigabine. However, developing retigabine analogues that better differentiate between hK_V7 subtypes is challenging because of conserved residues in the retigabine binding site. Moreover, promising retigabine analogues with improved hK_V7 subtype selectivity characterized in heterologous cell systems (Kumar et al., 2016) require extensive additional studies before their possible clinical potential is shown. Because retigabine and other M-channel activators could potentially be tolerated in low doses as anti-epileptic drugs, a parallel strategy would be to develop combined treatments to preserve the anticonvulsive M-channel effect while limiting adverse effects. In support of this, Manville and Abbott (2018) elegantly demonstrated that significant activating effects on the M-channel are achieved by a low concentration of retigabine if coapplied with components of herbal extracts binding to a site overlapping with the retigabine site. We hypothesize that the concept of combined targeting of the M-channel with dual drugs could be further developed if retigabine is strategically combined with M-channel activators with K_V7 subtype selectivity different from that of retigabine. Specifically, we propose that the subtype selectivity of such a combination treatment could be improved if retigabine is used in conjunction with M-channel activators that act on a site distinct from the retigabine site. In line with this notion, improved drug selectivity with potentially fewer adverse effects is one of the highlighted advantages of positive allosteric modulators targeting ligand-gated ion channels and G protein–coupled receptors (Abdel-Magid, 2015; Changeux and Christopoulos, 2016; Foster and Conn, 2017). These positive allosteric modulators are developed to bind to a site that is distinct from and less conserved than that of the natural ligand, with the aim of enhancing the response of a specific receptor to its ligand.

In this work, we test the potential of endocannabinoids as a novel class of M-channel activators that could be used in conjunction with retigabine. Common endocannabinoids, such as 2-arachidonoylglycerol (2-AG) and N-arachidonoylethanolamine (anandamide; AEA), are composed of an arachidonic acid tail connected to a head group and signal via cannabinoid receptors (Devane et al., 1992; Mechoulam et al., 1995). In addition, several endogenous arachidonic acid–based compounds related to 2-AG and AEA have been identified (Bisogno et al., 2000; Huang et al., 2001; Milman et al., 2006), with weak to strong potential to activate cannabinoid receptors. In this study, we refer to all these endogenous ligands as “endocannabinoids.” Besides the neurobehavioral effects of endocannabinoids mediated through the cannabinoid receptors, there is increasing evidence that endocannabinoids have a broader regulating function in the nervous system through interaction with so-called noncanonical targets. For instance, 2-AG has been shown to increase the firing of dopaminergic neurons by inhibiting A-type potassium channels (Gantz and Bean, 2017). Also, endothelium-dependent relaxation by AEA has been suggested to be caused by direct activation of BK channels (Bondarenko et al., 2017). If also the M-channel is a noncanonical target of endocannabinoids, the different chemical properties of retigabine and endocannabinoids make it likely that endocannabinoids act on a site different from that of retigabine.

We find that specific members of the endocannabinoid family activate the human M-channel (formed by coexpressed hK_V7.2 and hK_V7.3) expressed in Xenopuslaevis oocytes. Especially N-arachidonoyl-L-serine (ARA-S), a weak endogenous activator of cannabinoid receptor type 1 (Milman et al., 2006), induced large activating effects at low concentrations. We uncovered that ARA-S acts through a different mechanism and displays a different pattern of hK_V7 subtype selectivity compared with that of retigabine. Finally, we demonstrate how to use this by combining low concentrations of ARA-S and retigabine for pronounced activation of the M-channel while limiting activation of the other hK_V7 subtypes. Altogether, our study suggests that effective activation of the M-channel with improved hK_V7 subtype selectivity can be achieved by strategically combining M-channel activators with additive effects on the M-channel and nonadditive effects on the other hK_V7 subtypes.

All animal experiments were approved by the Linköping Animal Care and Use Committee (Permit #1941) and conform to national and international guidelines.

All chemicals were purchased from Sigma-Aldrich if not stated otherwise. 2-AG, AEA, ARA-S, N-arachidonoyl dopamine (NADA), N-arachidonoyl-γ-aminobutyric acid (NAGABA), arachidonoyl serinol (ARA-Serinol), and N-oleoyl-L-serine (OLE-S) were bought from Cayman Chemicals. Retigabine dihydrocloride was bought from Alomone Labs. N-docosahexaenoyl-L-serine (DOC-S), N-linoleoyl-L-serine (LIN-S), N-arachidonoyl-D-serine, and N-arachidoyl-L-serine (arachidoyl-S) were synthesized in house. The synthesis methods have been partly described previously (Silverå Ejneby et al., 2018). Most of the reagents for synthesis of new compounds were from Sigma-Aldrich except O-(benzotriazol-1-yl)-N,N,N′,N′-tetramethyluronium tetrafluoroborate, which was from Fluka (Fisher Scientific GTH AB). Preparative liquid chromatography (LC) was run on a Waters system with an XSELECT Phenyl-Hexyl column (250 × 19 mm, 5 μm), under neutral condition using gradient CH₃CN/water as eluent (A, water phase: 95: 5 water/CH₃CN, 10 mM NH₄OAc; B, organic phase: 90: 10 CH₃CN/water, 10 mM NH₄OAc). NMR spectra were recorded on a Varian Avance 300 MHz with solvent indicated. Chemical shift was reported in ppm on the δ scale and referenced to solvents peak (CDCl₃: δ_H = 7.26 ppm, δ_C = 77.16 ppm; methanol-d4: δ_H = 3.30 ppm, δ_C = 49.50 ppm).

To the saturated or unsaturated aliphatic acid and 2.20 equiv triethylamine in acetonitrile (20 ml) and/or dimethylformamide (DMF; 2–4 ml) was added 1.05 equiv O-(benzotriazol-1-yl)-N,N,N′,N′-tetramethyluronium tetrafluoroborate, and the reaction mixture was stirred at room temperature (rt) for ∼1 h. D-serine or L-serine (1.20 equiv) was added at rt and stirred over 1–7 nights. The solution was concentrated, and 5 ml water was added. The mixture was extracted with ethyl acetate (EA; 5 ml × 3). The organic layers were combined, concentrated, and purified using preparative LC (40–100%). The desired fractions were combined and concentrated to remove most of the acetonitrile. Another 5 ml water was added to the residue. The resulting solution was adjusted to pH at ∼5 using 1 N HCl aqueous solution and extracted with EA (15 ml × 2). The organic layers were concentrated to give the desired product.

Following the general procedure with docosahexaenoic acid (100 mg, 0.304 mmol), L-serine as starting materials, and DMF (2 ml) as solvent, the reaction mixture was stirred overnight at rt, filtered, and purified using preparative LC without extraction to give the product as a syrup (16.8 mg, 13% yield). ¹HNMR (CDCl₃/CD₃OD 3:2, 300 MHz): δ 5.43–5.14 (m, 12H), 4.48 (t, J = 4.2 Hz, 1H), 3.89 (dd, J = 11.4, 4.2 Hz, 1H), 3.76 (dd, J = 11.4, 4.2 Hz, 1H), 2.85–2.57 (m, 12H), 2.39–2.30 (m, 2H), 2.28–2.20 (m, 2H), 2.06–1.92 (m, 2H), 0.90 (t, J = 7.5 Hz, 3H). ¹³CNMR (CDCl₃/CD₃OD 3:2, 75 MHz): δ 173.4, 172.2, 131.9, 129.2, 128.4, 128.2, 128.15, 128.1, 128.0, 127.96, 127.9, 127.8, 126.9, 62.4, 54.4, 35.9, 25.5, 25.43, 25.41, 23.1, 20.4, 14.1. MS (ESI⁻): m/z calcd for C₂₅H₃₆NO₄ (M-H⁻) 414.26, found 414.51.

Following the general procedure with linoleic acid (280 mg, 0.998 mmol), L-serine (126.1 mg, 1.200 mmol) as starting materials, and acetonitrile (20 ml) as solvent, the reaction mixture was stirred over seven nights at rt, purified using preparative LC to give the product as a syrup (11.1 mg, 3%). ¹HNMR (CDCl₃, 300 MHz): δ 6.65 (br d, J = 6.6 Hz, 1H, NH), 5.45–5.25 (m, 4H), 4.50 (m, 1H), 4.14 (dd, J = 11.4, 3.6 Hz, 1H), 3.85 (dd, J = 11.4, 3.6 Hz, 1H), 2.82–2.70 (m, 2H), 2.30 (d, J = 7.5 Hz, 2H), 2.15–2.00 (m, 4H), 1.70–1.60 (m, 2H), 1.41–1.20 (m, 14H), 0.87 (t, J = 6.9 Hz, 3H). ¹³CNMR (CD₃OD/CDCl₃ 2:1, 75 MHz): δ 175.8, 173.1, 130.7, 130.6, 128.75, 128.69, 62.8, 55.5, 36.8, 32.3, 30.4, 30.11, 30.05, 30.0, 29.9, 27.9, 26.5, 26.3, 23.3, 14.3. MS (ESI⁻): m/z calcd for C₂₁H₃₆NO₄ (M-H⁻) 366.26, found 366.41.

Following the general procedure with arachidic acid (312.5 mg, 1.000 mmol), L-serine (126.1 mg, 1.200 mmol) as starting materials, acetonitrile (20 ml) and DMF (4 ml) as solvents, the reaction mixture was stirred over seven nights at rt. The solubility of this compound is poor. Part of the reaction mixture was purified using preparative LC to give the product as a white solid (5.0 mg). The yield was not calculated. ¹HNMR (CDCl₃, 300 MHz): δ 4.41 (dd, J = 4.8, 4.2 Hz, 1H), 3.89–3.75 (m, 2H), 2.29–2.22 (m, 2H), 1.35–1.25 (m, 34H), 0.89 (t, J = 7.2 Hz, 3H). MS (ESI⁻): m/z calcd for C₂₃H₄₄NO₄ (M-H⁻) 398.33, found 398.67.

Following the general procedure with docosahexaenoic acid (50 mg, 0.164 mmol), D-serine (20.7 mg, 0.197 mmol) as starting materials, and DMF (4 ml) as solvent, the reaction mixture was stirred overnight at rt, filtered, and purified using preparative LC without extraction to give the product as a syrup (21.5 mg, 33% yield). ¹HNMR (CDCl₃:CD₃OD 85:15, 300 MHz): δ 5.40–5.20 (m, 8H), 4.49 (t, J = 3.6 Hz, 1H), 3.89 (dd, J = 11.1, 3.6 Hz, 1H), 3.77 (dd, J = 11.1, 3.6 Hz, 1H), 2.82–2.68 (m, 6H), 2.21 (t, J = 7.8 Hz, 2H), 2.10–1.90 (m, 4H), 1.72–1.58 (m, 2H), 1.36–1.16 (m, 6H), 0.82 (t, J = 6.6 Hz, 3H). ¹³CNMR (CDCl₃:CD₃OD 85:15, 75 MHz): δ 174.1, 172.4, 130.5, 129.0, 128.8, 128.6, 128.3, 128.2, 127.9, 127.5, 62.6, 54.5, 35.7, 31.5, 29.3, 27.2, 26.7, 25.6, 25.0, 22.5, 14.0. MS (ESI⁻): m/z calcd for C₂₃H₃₆NO₄ (M-H⁻) 390.26, found 390.60.

All compounds were either delivered as or diluted as ethanol stock solutions except for 2-AG, which was delivered in acetonitrile. Stock solutions were stored at −20°C except for 2-AG, which was stored at −80°C.

Xenopus oocytes were surgically isolated and prepared at Linköping University alternatively purchased from Ecocyte Bioscience. Isolated Xenopus oocytes were injected with 50 nl RNA. Each oocyte was injected with 2.5 ng of hK_V7.2 (GenBank accession no. NM_004518) and 2.5 ng of hK_V7.3 (GenBank accession no. NM_004519) RNA for hK_V7.2/3. For other channels, 2.5–50 ng of RNA was injected per oocyte (GenBank accession nos. NM_000218 for hK_V7.1; NM_004700 for hK_V7.4; and NM_001160133 for hK_V7.5). Sequence alterations were introduced using site-directed mutagenesis (QuikChange II XL with 10 XL Gold cells; Agilent), and constructs were sequenced at the Core Facility at Linköping University to ensure correct sequence. cRNA was prepared using a T7 mMessage mMachine transcription kit (Invitrogen). RNA concentration was quantified using spectrophotometry (NanoDrop 2000c, Thermo Scientific). Injected oocytes were incubated at 8°C or 16°C for 2–4 d before being used for experiments. The two-electrode voltage clamp recordings were performed using a Dagan CA-1B Amplifier (Dagan). Currents were filtered at 500 Hz and sampled at 5 kHz. The holding voltage was generally set to −80 mV. Activation curves were generally generated in steps between −100 and +50 mV in increments of 10 mV (2-s duration). The tail voltage was generally set to −30 mV. In experiments where a large shift in half-maximal activation (V₅₀) occurred, a prepulse to −140 mV, −120 mV, or −100 mV (2-s duration) was included before the test voltage steps to close the channels. The control solution contained 88 mM NaCl, 1 mM KCl, 15 mM HEPES, 0.4 mM CaCl₂, and 0.8 mM MgCl₂. pH was set to 7.4 using NaOH. When experiments were performed at lower or higher pH, pH was set the same day as the experiment using HCl or NaOH, respectively. The two-electrode voltage clamp recordings were performed at rt. Control and test solutions (with endocannabinoids or/and retigabine or ICA73) were applied continuously using a pump (model ISM597D; Labinett Lab AB) with a perfusion rate of 0.5 ml/min during all recordings. Test solution was applied during an application protocol (depolarizing steps to −40 mV every 10 s) to ensure steady-state effects. If no effect was observed during application, the compound was applied for a minimum of 5 min.

Average values are expressed as mean ± SEM. Statistical analyses were done using one-way ANOVA followed by a multiple comparison test when comparing multiple data points. Dunnett’s multiple comparisons test was used when compared to defined reference data. Statistical analyses were done using Student’s t test when comparing two data points against each other or against a hypothetical value of zero. P < 0.05 was considered statistically significant. In case of a large number of Student’s t tests, the significance level was adjusted with a Bonferroni correction to reduce the risk of a type 1 error.

Fig. S1 shows that specific endocannabinoids do not activate the hK_V7.2/3 channel. Fig. S2 shows the effect of ARA-S and NAGABA on hK_V7.2/3. Fig. S3 shows the effect of N-arachidonoyl-D-serine on hK_V7.2/3. Fig. S4 shows that ARA-S activates homomeric hK_V7.2 and hK_V7.3 tryptophan mutants. Fig. S5 shows that the specific charge-neutralizing mutation in hK_V7.2 reduces ARA-S effect. Fig. S6 shows that the specific charge-neutralizing mutation in hK_V7.3 reduces ARA-S effect. Fig. S7 shows that coapplication of low concentrations of ARA-S and retigabine limits the off-target effect on other hK_V7 subtypes. Fig. S8 shows that coapplication of low concentrations of ARA-S and ICA73 limits the off-target effect on other hK_V7 subtypes. Table S1 summarizes the biophysical properties of used constructs.

To study whether endocannabinoids target the M-channel, we initially tested the effect of five arachidonic acid–based endocannabinoids on hK_V7.2 and hK_V7.3 coexpressed in Xenopus oocytes (hereafter referred to as hK_V7.2/3): 2-AG, AEA, ARA-S, NADA, and NAGABA (Fig. 1 a). All five compounds share an arachidonic acid tail (which makes them likely to be among the more prevalent endocannabinoids because of the high abundance of arachidonic acid in membranes; Di Marzo et al., 2004) yet contain diverse head groups (which may indicate different functional effects; Bohannon et al., 2020). Each compound was applied at two different concentrations, 10 µM and 100 µM. From this initial screen, we identified two compounds that activated the hK_V7.2/3 channel. 10 µM and 100 µM ARA-S and NAGABA significantly shifted the voltage for V₅₀ toward negative voltages (Fig. 1 b), thus allowing the channel to open at more negative voltages. ARA-S also significantly increased the maximal conductance (G_MAX; Fig. 1 c). In contrast, 2-AG, AEA, and NADA had no effect on V₅₀ or G_MAX of the hK_V7.2/3 channel (Fig. 1, b and c; and Fig. S1, a–c).

We investigated the activating effects of ARA-S and NAGABA further. 10 µM ARA-S induced a clear increase in the overall current amplitude of hK_V7.2/3 (Fig. 1 d). The onset of the ARA-S effect was relatively fast, with stable effects achieved within 2 min (Fig. S2 a). The washout of the ARA-S effect using control solution was relatively slow, with roughly 10 min of washout required for full reversibility (Fig. S2 b). The washout was faster when albumin, which binds endocannabinoids (Bojesen and Hansen, 2003), was added to the control solution (Fig. S2 b). ARA-S sped up channel-opening kinetics and slowed down the closing kinetics (Fig. S2 c; for opening at +20 mV, τ_ARA-S/τ_ctrl = 0.76 ± 0.02; for closing at −100 mV, τ_ARA-S/τ_ctrl = 1.45 ± 0.11; n = 3). The conductance versus voltage (G(V)) curve clearly demonstrates the ability of ARA-S to shift V₅₀ and increase G_MAX (Fig. 1 e). On average, 10 µM ARA-S shifted V₅₀ by −19.5 ± 1.5 mV and increased G_MAX by 53% ± 7%. The shift in V₅₀ induced by ARA-S was significant at concentrations as low as 300 nM (−1.4 ± 0.4 mV, P = 0.03 using Student’s t test; Fig. 1 f). The concentration-response curve for the V₅₀ effect of ARA-S suggests an estimated maximal shift of V₅₀ by −31.9 ± 4.3 mV with 50% of the maximal shift (EC₅₀) caused by 7 µM (Fig. 1 f). The concentration-response curve for the G_MAX effect of ARA-S suggests an estimated maximal increase of G_MAX by 66% ± 8%, with an EC₅₀ of 3 µM (Fig. 1 g). Both the shift in V₅₀ and the increase in G_MAX occur at similar concentrations and together contribute to the overall increase of the current. On average, 10 µM ARA-S induced a sevenfold increase in the current at −60 mV (Fig. 1 h), suggesting that ARA-S effectively activates hK_V7.2/3 at voltages around the resting membrane potential of neurons.

NAGABA produced overall similar but smaller effects compared with ARA-S (Fig. S2, d–g). On average, 10 µM NAGABA shifted V₅₀ by −6.9 ± 0.5 mV but did not significantly increase G_MAX. The concentration-response curve for the V₅₀ effect of NAGABA suggests an estimated maximal shift of V₅₀ by −20.8 ± 5.2 mV with an EC₅₀ of 24 µM (Fig. S2 e). A concentration-response curve for the G_MAX effect of NAGABA could not be obtained as there was no consistent effect of NAGABA on G_MAX (Fig. S2 f). On average, 10 µM NAGABA induced a threefold increase in the current at −60 mV (Fig. S2 g).

We tested which molecular determinants of the studied endocannabinoids are necessary to induce activating effects on the hK_V7.2/3 channel. Modifying either the head or the lipid tail of the endocannabinoid could potentially change the effect of the compound. For instance, the head group of both ARA-S and NAGABA contains a carboxyl group, which can lose a proton to become negatively charged depending on the acid dissociation constant (pKa) of the head group and the pH of the surrounding solution. To test whether the negative charge of the head group is important for the ARA-S effect, we tested ARA-Serinol, which has a head group that lacks the carboxyl group of ARA-S and therefore is uncharged (Fig. 2 a). 10 µM ARA-Serinol had no significant effect on V₅₀ or G_MAX of hK_V7.2/3 (Fig. 2, a–c; P > 0.05 using Student’s t test), suggesting that the carboxyl group is required for the activating effect.

We tested the effect of ARA-S and NAGABA on V₅₀ at different extracellular pH to alter the degree of carboxyl group protonation. For ARA-S, the effect on V₅₀ was not increased when pH levels were increased from pH 7.4 to pH 9.0 or 10 (Fig. 2 d). However, the effect was reduced when pH was decreased from pH 7.4 to pH 5.5 (Fig. 2 d). The estimated apparent pKa for ARA-S at the channel is 5.4 (Fig. 2 d), which suggests that ARA-S is fully deprotonated at physiological pH. In contrast, the effect of NAGABA was increased when pH was increased from pH 7.4 to pH 9.0 or 10, and the effect of NAGABA was eliminated when pH was decreased from pH 7.4 to pH 6.5 (Fig. 2 e). The estimated apparent pKa for NAGABA at the channel is 7.3 (Fig. 2 e), suggesting that only around 50% of the NAGABA molecules are deprotonated at physiological pH. The difference in estimated pKa could partly explain the difference in effect of ARA-S and NAGABA on V₅₀ at pH 7.4.

Since the head group of ARA-S contains a serine moiety, ARA-S could be either L or D enantiomer. The naturally occurring enantiomer (Milman et al., 2006) tested in the initial experiments was the L enantiomer of ARA-S. To test if the chirality affects the effect of ARA-S on hK_V7.2/3, we tested the D enantiomer of ARA-S. The D enantiomer of ARA-S induced effects comparable to those of the L enantiomer on V₅₀ and G_MAX of hK_V7.2/3 (Fig. S3, a–c). Altogether, these findings suggest that the negative charge of the endocannabinoid head group is important for the activating effect on the hK_V7.2/3 channel, but not the chirality of the head. The importance of the negative charge could explain why 2-AG, AEA, and NADA do not activate the hK_V7.2/3 channel, as none of these compounds can be negatively charged.

We next tested the impact of the fatty acid tail on the effect of ARA-S on hK_V7.2/3. We kept the serine head but exchanged the arachidonic acid tail of ARA-S to different lipid tails. We tested the polyunsaturated docosahexaenoic and linoleic tails (DOC-S and LIN-S), the monounsaturated oleic tail (OLE-S), and the saturated arachidic tail (arachidoyl-S; Fig. 2 f). At a concentration of 10 µM, exchanging the arachidonic tail for the docosahexaenoic tail preserved the ability to shift V₅₀ and increase G_MAX (Fig. 2, g and h). The linoleic tail reduced the ability to shift V₅₀ but preserved the ability to increase G_MAX (Fig. 2, g and h). The oleic tail reduced the ability to shift both V₅₀ and increase G_MAX (Fig. 2, g and h). The arachidic tail eliminated the activating effect on hK_V7.2/3 (Fig. 2, g and h). Altogether, these data suggest that the composition of the lipid tail alters the effect on hK_V7.2/3. Decreasing the number of double bonds in the fatty acid tail tended to decrease the effect of the serine compounds on hK_V7.2/3. It should be noted that arachidoyl-S was difficult to dissolve, suggesting that the absence of effect of arachidoyl-S could be caused in part by poor solubility.

As previously described, retigabine (structure in Fig. 3 a) shifts V₅₀ of hK_V7.2/3 to negative voltages and increases the overall current (Main et al., 2000). In our experiments, retigabine shifted V₅₀ of hK_V7.2/3 with an estimated EC₅₀ of 2 µM and a ΔV₅₀ of −21.7 ± 0.5 mV induced by 3 µM retigabine (Fig. 3, a and b). As has been previously demonstrated (Schenzer et al., 2005), removing the tryptophan in S5 that is important for retigabine effects (by constructing the hK_V7.2_W236L and hK_V7.3_W265L mutants and coexpressing them as hK_V7.2_W236L/hK_V7.3_W265L) rendered the channel insensitive to 3 µM retigabine (Fig. 3, c and d). ARA-S, on the other hand, retained its effect on the hK_V7.2_W236L/hK_V7.3_W265L channel (Fig. 3, e and f) as well as on homomeric hK_V7.2_W236L and hK_V7.3_W265L channels (Fig. S4, a–d). These experiments demonstrate that the tryptophan in S5 is not important for the activating effect of ARA-S and suggest that ARA-S and retigabine activate the channel through distinct mechanisms. Moreover, ARA-S retained its effect on the hK_V7.2_F168L channel (Fig. S4, e and f), which was previously shown to render the channel insensitive to the M-channel activator ICA73 (ICA-069673) by removing the phenylalanine in S3 (Wang et al., 2017; Wang et al., 2018). This suggests that ARA-S activates the channel through a mechanism different from that of ICA73.

In previous work, we showed that polyunsaturated fatty acids and their analogues activate K_V channels (Liin et al., 2015; Liin et al., 2016; Elinder and Liin, 2017). In hK_V7.1, we have suggested that the activation is caused by two distinct interactions between polyunsaturated fatty acids and the channel (Liin et al., 2018): (1) interaction with the outermost arginines in S4 of the voltage-sensing domain facilitates outward S4 movement to shift V₅₀ toward more negative voltages (illustrated in Fig. 4 a, top row), and (2) interaction with a lysine in S6 of the pore domain increases G_MAX, possibly by stabilizing the conducting selectivity filter (illustrated in Fig. 4 a, bottom row). Both the negative charge of the fatty acid and the polyunsaturation of the lipid tail are important for these effects (Börjesson et al., 2008; Börjesson and Elinder, 2011; Liin et al., 2015; Bohannon et al., 2019). Because of the similarity in chemical properties of polyunsaturated fatty acids and endocannabinoids and their overall similar pattern of chemical properties required to activate K_V7 channels, we tested whether the corresponding residues in hK_V7.2 and hK_V7.3 are important for the ARA-S effect. In hK_V7.2, neutralization of either the first or second S4 arginine (R198 and R201, respectively) significantly reduced the ability of 10 µM ARA-S to shift V₅₀ (Fig. 4 b and Fig. S5). Note that the Q204R mutation was introduced together with the R201Q mutation, as the R201Q single mutant is voltage insensitive (Miceli et al., 2008). Guided by studies in hK_V7.1 (Panaghie and Abbott, 2007; Wu et al., 2010), we found that voltage sensitivity of the R201Q mutant was retained by introducing a compensatory positive charge at the Q204 position, which allowed probing of ARA-S effects on V₅₀ of this mutant. The Q204R mutation per se did not appear to alter the ARA-S effect on V₅₀ (compare effects on R198Q and R198Q/Q204R in Fig. 4 b). In contrast, neutralization of R291 in S6 (corresponding to the lysine in S6 of hK_V7.1) significantly reduced the ability of 10 µM ARA-S to increase G_MAX (Fig. 4 c and Fig. S5). In hK_V7.3, neutralization of either the first or second S4 arginine (R227 and R230, respectively) alone did not alter the ARA-S effect on V₅₀ (Fig. 4 d and Fig. S6). However, simultaneous neutralization of R227 and R230 significantly reduced the ability of 10 µM ARA-S to shift V₅₀ (Fig. 4 d and Fig. S6). As with hK_V7.2, a compensatory positive charge at the Q233 position was introduced to retain voltage sensitivity.

For hK_V7.3, the effect of mutation on G_MAX was less clear, with no significant reduction in the ability of 10 µM ARA-S to increase G_MAX by the R330Q mutation (Fig. 4 e and Fig. S6). However, interpretation of alteration in the G_MAX effect in hK_V7.3 is possibly complicated by the A315T mutation introduced to enable studies of homotetrameric hK_V7.3 channels. The A315T mutations have been proposed to increase the K⁺ current generated by hK_V7.3 by stabilizing the conducting pore (Choveau et al., 2012a, 2012b), possibly together with increasing membrane expression (Gómez-Posada et al., 2010). Because the ARA-S effect on G_MAX is expected to be dependent on the intrinsic open probability of the channel (Liin et al., 2018), possible varying intrinsic open probability of the hK_V7.3 mutants will impact the ability of ARA-S to increase G_MAX. Altogether, experiments on the set of mutants with neutralized positive residues at specific positions suggest that positively charged amino acids in S4 and S6 are important for the ability of ARA-S to shift V₅₀ (in both hK_V7.2 and hK_V7.3) and increase G_MAX (in hK_V7.2), respectively.

Retigabine was previously shown to activate all members within the hK_V7 family except hK_V7.1 (Tatulian et al., 2001; Schenzer et al., 2005). Because our results indicate that the activation of hK_V7.2/3 induced by ARA-S and retigabine occurs via different mechanisms, ARA-S might show a different hK_V7 subtype selectivity than retigabine. We tested the effect of ARA-S on each member of the hK_V7 channel family expressed as homomeric channels in Xenopus oocytes. 10 µM ARA-S significantly activated hK_V7.1, hK_V7.2, hK_V7.3_A315T, and hK_V7.5 by shifting V₅₀ toward more negative voltages and increasing G_MAX (Fig. 5, a–c and e–g). In contrast, ARA-S had a slight inhibiting effect on hK_V7.4. 10 µM ARA-S shifted V₅₀ of hK_V7.4 by +6.7 ± 2.0 mV (Fig. 5 d,Fig. 5, d and f; P < 0.009 with Student’s t test) but had no effect on G_MAX (Fig. 5 g). Fig. 5, h and i shows the corresponding effects of 3 µM retigabine on V₅₀ and G_MAX, which are in overall concordance with previously reported effects for this retigabine concentration (Schenzer et al., 2005). Altogether, these experiments reveal that ARA-S has a different subtype selectivity in its effect on the hK_V7.1 and hK_V7.4 channels compared with retigabine.

We hypothesized that one could leverage upon the similar effect of ARA-S and retigabine on the M-channel and the different effect on the other hK_V7 subtypes by coapplying low concentrations of ARA-S and retigabine for retained M-channel effect with limited off-target effects. As a first step, we compared the effect of low concentrations of ARA-S or retigabine alone with both compounds in combination on hK_V7.2/3. 1 µM ARA-S alone shifted V₅₀ of hK_V7.2/3 by −6.7 ± 0.7 mV and increased G_MAX by 20% ± 4% (Fig. 6). 1 µM retigabine alone shifted V₅₀ of hK_V7.2/3 by −8.1 ± 0.7 mV and did not affect G_MAX (Fig. 6). In contrast, coapplication of 1 µM ARA-S and 1 µM retigabine shifted V₅₀ by −19.5 ± 1.3 mV and increased G_MAX by 31% ± 9% (Fig. 6). The improved effect of coapplication of low concentrations of ARA-S and retigabine was clear when comparing the magnitude of hK_V7.2/3 current fold increase at negative voltages. On average, coapplication of 1 µM ARA-S and 1 µM retigabine induced a 20-fold increase of the current at −60 mV, compared with a three- to fourfold increase for 1 µM of either of the compounds individually (Fig. 6 d). Notably, the 20-fold increase of the current at −60 mV induced by coapplication is almost three times the increase of the current induced by 10 µM ARA-S alone (compare with Fig. 1 h). Thus, the combined activating effect of low concentrations of ARA-S and retigabine substantially increased the current generated by hK_V7.2/3 at voltages corresponding to the resting membrane potential of neurons.

As a second step, we compared the off-target effect of coapplying ARA-S and retigabine on hK_V7.1, hK_V7.4, hK_V7.5, and hK_V7.4/5 channels with the effects of the compounds individually. hK_V7.4/5 was included in this analysis because heterotetramers of hK_V7.4 and hK_V7.5 have been implicated in regulation of smooth muscle contraction (Brueggemann et al., 2014; Chadha et al., 2014; Provence et al., 2018). We compared the effect of either 10 µM ARA-S or 3 µM retigabine individually with that of coapplication of 1 µM ARA-S and 1 µM retigabine, as these three treatments all shifted V₅₀ of the hK_V7.2/3 channel by about −20 mV (Fig. 7 a). We will refer to these three treatments as ARA-S_high, retigabine_high, and coapplication. As was shown in Fig. 5, ARA-S_high primarily induced off-target effects on hK_V7.1 and hK_V7.5 by shifting V₅₀ by about −30 mV (Fig. 7 a). ARA-S_high induced similar off-target effects on hK_V7.4/5 by shifting V₅₀ by about −25 mV (Fig. 7 a). Retigabine_high, on the other hand, primarily induced off-target effects on hK_V7.4 by increasing G_MAX by 350% (Fig. 7 b and Fig. S7 a). Retigabine_high also induced off-target effects on hK_V7.4/5 by increasing G_MAX by 180% (Fig. 7 b). In contrast, all off-target effects were reduced during coapplication: coapplication shifted V₅₀ of hK_V7.1, hK_V7.5, and hK_V7.4/5 by less than 4 mV (Fig. 7 a and Fig. S7 c) and reduced the increase in G_MAX of hK_V7.4 and hK_V7.4/5 to 75% and 35%, respectively (Fig. 7 b and Fig. S7 b).

We also tested the concept of coapplication using ICA73 (structure in Fig. S8 a). Although 20 µM of ICA73 alone (referred to as ICA73_high) activated hK_V7.2/3 by shifting V₅₀ by −15 mV (Fig. S8 a), ICA73_high also induced off-target effects on hK_V7.4 by shifting V₅₀ by −15 mV and increasing G_MAX by 175% (Fig. S8, a and c). Like ICA73_high, coapplication of 2 µM ARA-S and 6 µM ICA73 shifted V₅₀ of hK_V7.2/3 by about −15 mV (Fig. S8 a). However, coapplication did not shift V₅₀ of hK_V7.4 (P > 0.05 with Student’s t test) and reduced the increase in G_MAX of hK_V7.4 to 66% (Fig. S8, a and c). Altogether, these experiments demonstrate that ARA-S can be combined with either retigabine or ICA73 to preserve the activating effect on the hK_V7.2/3 channel and limit off-target effects on other hK_V7 subtypes.

In this study, we make three important observations. First, we uncover that specific endogenous compounds belonging to the endocannabinoid family activate the hK_V7.2/3 channel expressed in Xenopus oocytes. ARA-S in particular was identified as a potent hK_V7.2/3 activator, which activated the channel by both shifting the voltage dependence of channel opening toward negative voltages and increasing the maximal conductance. Second, we found that ARA-S activates hK_V7.2/3 through a different mechanism than retigabine and that these two compounds differ in their hK_V7 subtype selectivity. The activating effect of retigabine on hK_V7.2/3 has been extensively studied. Retigabine activates hK_V7.2/3 by binding to an internal site in the pore domain, in which a tryptophan is critical for binding, to stabilize the open channel (Schenzer et al., 2005; Wuttke et al., 2005; Kim et al., 2015). For endocannabinoids, we here describe that conserved positively charged residues in S4 and S6, distinct from the retigabine site but corresponding to those previously shown to be important for polyunsaturated fatty acid effects in hK_V7.1 (Liin et al., 2018), are important for ARA-S–induced activation. We also describe that the negative charge of the endocannabinoid head is critical for the effect on hK_V7.2/3, which could explain why 2-AG, AEA, and NADA are ineffective, and that a saturated tail appears to eliminate the effect. Given the corresponding important channel residues identified and the shared chemical properties of polyunsaturated fatty acids and endocannabinoids, we find it likely that endocannabinoids act through a similar mechanism to that previously described for polyunsaturated fatty acids (Liin et al., 2018). Thus, we hypothesize that the negative charge of the endocannabinoid head group is required to enable electrostatic interaction with S4 arginines (to facilitate outward S4 movement) and with the S6 arginine (to induce structural rearrangements in the pore that enhance potassium conductance; Liin et al., 2018). The flexibility of a polyunsaturated tail to adopt bent conformations has been reported to allow intimate and high-affinity interaction of polyunsaturated fatty acids with other voltage-gated ion channels (Tian et al., 2016; Yazdi et al., 2016). Possibly, the more rigid nature of monounsaturated and saturated endocannabinoid tails limits bent conformations needed for interaction with hK_V7.2/3. Future studies are needed to further resolve molecular details for endocannabinoid interactions with hK_V7.2/3. We note that the electrostatic nature of the proposed interaction between the negative endocannabinoid head group and positive residues in the channel is reminiscent of the interaction between PIP₂ (phosphatidylinositol 4,5-bisphosphate) and K_V7 channels (Zaydman and Cui, 2014; Taylor and Sanders, 2017). However, whereas PIP₂ interacts with positively charged residues in the inner half of the channel (because PIP₂ is localized to the inner leaflet of the plasma membrane), we propose that endocannabinoids primarily act on residues in the outer half of the channel. Third, we demonstrate how to leverage upon the different hK_V7 subtype selectivities through coapplication of ARA-S and retigabine for efficient activation of hK_V7.2/3 while limiting the off-target effect on hK_V7.1, hK_V7.4, hK_V7.5, and hK_V7.4/5. Beneficial effects of coapplication were also achieved by combining ARA-S and ICA73.

Can coapplication of low doses of two compounds targeting the M-channel be a viable approach to reach adequate anti-convulsant effect while reducing adverse effects usually caused by higher doses of these compounds? Manville and Abbott (2018) lend support to efficient anti-convulsant effect in an animal epilepsy model of coadministration of multiple compounds targeting the M-channel. Our experiments on hK_V7 channels expressed in Xenopus oocytes show that combined application of low concentrations of ARA-S and retigabine effectively activates the M-channel and that strategic combination of M-channel activators with different subtype selectivity limits off-target effects on other hK_V7 subtypes. Developing new anticonvulsants is a costly and exhaustive process. Withdrawal of a drug that has passed critical safety testing is a major setback. Our strategy of targeting the M-channel with dual compounds acting through a distinct mechanism, here tested in Xenopus oocytes, might offer ways for promising drugs with some degree of off-target effect to be tolerated clinically. Therefore, we encourage further exploration of this concept in more complex experimental systems. Such studies need to address important questions, such as which M-channel activators should ideally be combined (considering critical aspects such as ion channel selectivity, apparent affinity, and efficacy) and the balance between anti-excitable effects in neurons and possible off-target effects in other cell types. Clinical challenges with combined treatment using two or more drugs include possible varying pharmacodynamic profiles of the drugs used and possible drug interaction. However, as ∼20% of patients suffering from epilepsy are treated using two or more drugs (Brodie et al., 2012), these are challenges already faced by current treatment strategies.

We note that coapplication of 1 µM ARA-S and 1 µM retigabine induced hK_V7.2/3 effects that were slightly larger than expected from additive effects (Fig. 6, b–d). In particular, the increase of current amplitude at −60 mV was unexpectedly large upon coapplication (Fig. 6 d). Manville and Abbott (2018) described synergistic effects on hK_V7.2/3 upon coapplication of retigabine and herbal extracts. They proposed that the ability of such compounds to simultaneously occupy a shared binding pocket in combination with prominent effects on different subunits in the hK_V7.2/3 channel (hK_V7.3 for retigabine and hK_V7.2 for herbal components) underlies synergistic effects (Manville and Abbott, 2018). We observed larger ARA-S effects on hK_V7.2 than on hK_V7.3. However, because ARA-S and retigabine use different binding sites, it is not clear if different preference for different hK_V7.2/3 subunits contributed to the apparent synergy in ARA-S and retigabine effects. Alternatively, presence of one of the compounds may facilitate action of the other compound (e.g., by improving affinity or efficacy). However, at this point, we are careful with interpretation of additive versus synergistic effects in our data, as the most prominent indication of synergy was observed for current amplitudes at −60 mV, which is at the foot of the G(V) curve, meaning that small variations in V₅₀ between cells may impact the magnitude of current increase. Further characterization of additive/synergistic effects and possible underlying mechanisms is worth exploring in future studies.

Because ARA-S is an endogenous compound, this raises several questions. For example, does ARA-S have physiological or pathophysiological effects in the body at the concentrations used in this study, and could these effects be tuned by regulating the concentration of ARA-S? Although all five arachidonic acid–based endocannabinoids studied in this work have been detected in neuronal tissue from mammals (Devane et al., 1992; Kondo et al., 1998; Bisogno et al., 2000; Huang et al., 2001; Milman et al., 2006), the understanding of the endogenous functions of some of them is less clear. The endocannabinoids 2-AG and AEA have been shown to signal noncanonically through other pathways than cannabinoid receptors (Bondarenko et al., 2017; Gantz and Bean, 2017). ARA-S has been isolated from bovine brain (Milman et al., 2006), but it is not known if ARA-S is present in human tissue and, if so, at what concentrations. This makes it difficult to evaluate whether ARA-S might have physiological or pathophysiological functions in the human body. Future studies will hopefully shed light on whether ARA-S endogenously reaches concentrations triggering noncanonical signaling through interaction with ion channels such as hK_V7.2/3, the BK channel (Godlewski et al., 2009), and the N-type Ca_V channel (Guo et al., 2008).

To conclude, this study shows that specific members within the endocannabinoid family, most notably ARA-S, target and activate the neuronal M-channel. The pronounced activation of the M-channel by low concentrations of ARA-S highlights ARA-S as an interesting model compound for future development of new treatment strategies for epilepsy targeting the M-channel. The successful combination of ARA-S and retigabine (as well as ARA-S and ICA73) to achieve M-channel activation with improved hK_V7 subtype selectivity allows for possibilities to use such ARA-S mimetic compounds either alone or in conjunction with other M-channel activators. Future studies are required to study the concept of combined treatment in more complex systems.

Joseph A. Mindell served as editor.

We thank Dr. H. Peter Larsson at the University of Miami and Dr. Fredrik Elinder at Linköping University for valuable comments on the manuscript. We thank Dr. Peter Konradsson at Linköping University for valuable discussion regarding synthesis strategies. The clones for human K_V7.1–7.5 were kind gifts from Dr. Nicole Schmitt at the University of Copenhagen, Copenhagen, Denmark.

This work was supported by the Swedish Society for Medical Research and the Swedish Research Council (2017-02040).

A patent application (#62/032,739) including a description of the interaction of charged lipophilic compounds with the K_V7.1 channel has been submitted by the University of Miami, with S.I. Liin identified as one of the inventors.

Author contributions: J.E. Larsson: concept and design, acquisition of data, analysis and interpretation of data, and drafting and revising of the article. U. Karlsson: concept and design, interpretation of data, and drafting and revising the article. X. Wu: contributing new compounds and drafting and revising the article. S.I. Liin: concept and design, acquisition of data, analysis and interpretation of data, and drafting and revising the article.