Ancient and modern anticonvulsants act synergistically in a KCNQ potassium channel binding pocket

Abstract

Epilepsy has been treated for centuries with herbal remedies, including leaves of the African shrub Mallotus oppositifolius, yet the underlying molecular mechanisms have remained unclear. Voltage-gated potassium channel isoforms KCNQ2-5, predominantly KCNQ2/3 heteromers, underlie the neuronal M-current, which suppresses neuronal excitability, protecting against seizures. Here, in silico docking, mutagenesis and cellular electrophysiology reveal that two components of M. oppositifolius leaf extract, mallotoxin (MTX) and isovaleric acid (IVA), act synergistically to open neuronal KCNQs, including KCNQ2/3 channels. Correspondingly, MTX and IVA combine to suppress pentylene tetrazole-induced tonic seizures in mice, whereas individually they are ineffective. Co-administering MTX and IVA with the modern, synthetic anticonvulsant retigabine creates a further synergy that voltage independently locks KCNQ2/3 open. Leveraging this synergy, which harnesses ancient and modern medicines to exploit differential KCNQ isoform preferences, presents an approach to developing safe yet effective anticonvulsants.

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1

Multiple M. oppositifolius leaf extract compounds activate KCNQ2/3. a KCNQ1–KCNQ3 chimeric structure model. b KCNQ topology (two of four subunits shown). VSD voltage-sensing domain. c Structure and electrostatic surface potential (blue, positive; green neutral; red, negative) of M. oppositifolius leaf extract components. Open red circles highlight strongly negative electrostatic surface potential. d Averaged KCNQ2/3 current traces in response to voltage protocol (upper inset) when bathed in the absence (Control) or presence of M. oppositifolius leaf extract components (n = 4–16). Dashed line indicates zero current level in this and all following current traces. e, f Mean effects of leaf extract components (as in d; n = 4–16) on: e KCNQ2/3 raw tail currents at −30 mV after prepulses as indicated; fG/Gmax. Error bars indicate SEM. Red boxes indicate KCNQ2/3 activation; black box indicates KCNQ2/3 inhibition

Fig. 2

MTX preferentially activates KCNQ2. a Voltage dependence of KCNQ2/3 current fold-increase by MTX (30 µM), plotted from traces as in Fig. 1 (n = 9). b Dose response of KCNQ2/3 channels at −60 mV for MTX (calculated EC50 = 11.5 µM; n = 4–9). c Exemplar −60 mV KCNQ2/3 current (left) during wash-in/washout of MTX; right, during wash-in of MTX followed by XE991. d Mean activation (left) and deactivation (right) rates for KCNQ2/3 before (Ctrl) and after wash-in of MTX (n = 9); ***p < 0.001. Activation rate was quantified using voltage protocol as in Fig. 1d. Deactivation rate was quantified using voltage protocol shown (lower right inset). e MTX dose-dependently hyperpolarizes resting membrane potential (EM) of unclamped oocytes expressing KCNQ2/3; n = 9. f MTX has no effect on (left) endogenous mean current or (right)EM of water-injected control oocytes (n = 5). Voltage protocol as in Fig. 1d. g MTX has no effect on (left) averaged current traces or (right) G/Gmax of oocytes expressing KCNA1 (n = 5). Voltage protocol as in Fig. 1d. h Averaged current traces for homomeric KCNQ2–5 channels in the absence (Control) or presence of MTX (30 µM) (n = 5–10). Voltage protocol as in Fig. 1d. i Mean effects of MTX (30 µM) on −30 mV tail currents for channels and voltage protocol as in h (n = 5–10).j. Mean voltage dependence of 30 µM MTX (structure and surface potential, right) activation of homomeric KCNQ2–5 at −60 mV, recorded from tail currents as in i (n = 5–10). k MTX dose response at −60 mV for homomeric KCNQ2–5, quantified from data as in i (n = 5–10). All error bars indicate SEM

Fig. 3

MTX activates KCNQ2/3 by binding close to the pore. a KCNQ topology (two of four subunits shown) indicating approximate position of KCNQ3-W265. VSD voltage-sensing domain. b Binding position of (upper) retigabine and (lower) MTX in KCNQ3 predicted by SwissDock using a chimeric KCNQ1–KCNQ3 structure model. c Effects of MTX (30 µM) on tail current and G/Gmax relationships for single- and double-W/L mutant KCNQ2/3 channels as indicated (n = 3–5). Voltage protocol as in Fig. 1d. d Dose response for mean ΔV0.5 of activation induced by MTX for wild-type KCNQ2/3 and mutant channels as in c (n = 3–9). eLeft, exemplar traces; right, mean I/V relationships for KCNQ2/3 channels bathed in 100 mM K+, Rb+, Cs+, or Na+ in the presence or absence (Control) of MTX (30 µM); n = 4–7. f Relative ion permeabilities of KCNQ2/3 channels in the presence or absence (Ctrl) of MTX (30 µM); n = 4–7. Quantified from traces and plots as in panel e. g Relative Rb+ to K+ permeabilities of KCNQ2/3 channels in the presence or absence (Ctrl) of MTX (30 µM); n = 4–8. All error bars indicate SEM

Fig. 4

MTX outcompetes 2-mercaptophenol to activate KCNQ2/3. a Binding position of GABOB predicted by SwissDock using a chimeric KCNQ1–KCNQ3 structure model. b Exemplar traces showing effects of MTX (30 µM) with GABOB (100 µM) on KCNQ2/3 channels. Voltage protocol as in Fig. 1d. c Effects of MTX (30 µM) with GABOB (100 µM) on mean tail current (left) and G/Gmax (right) relationships for KCNQ2/3 (n = 6) calculated from traces as in panel b. d Current fold-change at −60 mV exerted by MTX (30 µM) alone (from Fig. 2a) or with 100 µM GABOB, from data as in panel c (n = 6). e Binding position of 2-mercaptophenol predicted by SwissDock using a chimeric KCNQ1–KCNQ3 structure model. f Exemplar traces showing effects of 2-mercaptophenol (100 µM) on KCNQ2-W236L/KCNQ3-W265 (WL-WL) channels. Voltage protocol as in Fig. 1d. g Effects of 2-mercaptophenol (100 µM) on mean tail current (left) and G/Gmax (right) relationships for KCNQ2-W236L/KCNQ3-W265 (WL-WL) channels (n = 9) calculated from traces as in panel f. h Exemplar traces showing effects of MTX (30 µM) with 2-mercaptophenol (100 µM) on KCNQ2/3 channels. Voltage protocol as in Fig. 1d. i Effects of MTX (30 µM) with 2-mercaptophenol (100 µM) on mean tail current (left) and G/Gmax (right) relationships for KCNQ2/3 (n = 9) calculated from traces as in panel h. j Current fold-change at −60 mV exerted by MTX (30 µM) alone (from Fig. 2a) or with 100 µM 2-mercaptophenol, from data as in panel i (n = 9). All error bars indicate SEM

Fig. 5

IVA activates neuronal KCNQs with preference for KCNQ2. aLeft, Valeriana officinalis. Right, structure (upper) and electrostatic surface potential (red, negative; blue, positive) (lower) of isovaleric acid (IVA). b Binding position of IVA in KCNQ3 predicted by SwissDock using a chimeric KCNQ1–KCNQ3 structure model. c Mean tail current versus prepulse voltage relationships recorded by TEVC in Xenopus laevis oocytes expressing homomeric KCNQ1–5 channels in the absence (black) and presence (blue) of IVA (n = 4–7). Voltage protocol as in Fig. 1d. d IVA dose response at −60 mV for KCNQ2–5, quantified from data as in c (n = 4–7). e Mean tail current versus prepulse voltage relationships for wild-type KCNQ2/3 (left) or KCNQ2-W236L/KCNQ3-W265L (right) channels in the absence or presence of IVA as indicated (n = 4–6). Voltage protocol as in Fig. 1d. f Dose response for current increase at −60 mV in response to IVA for channels as in e. g Dose response for the V0.5 of activation shift induced by IVA versus MTX in wild-type KCNQ2/3 versus KCNQ2-W236L/KCNQ3-W265L (WL/WL) channels. IVA data (n = 4–6) quantified from e; MTX data from Fig. 3d. h Averaged traces for KCNQ1 in the absence or presence of IVA (500 µM); n = 6. i Mean data from traces as in h. j Exemplar traces showing effects of IVA (500 µM) with GABOB (100 µM) on KCNQ2/3 channels. Voltage protocol as in Fig. 1d. k Effects of IVA (500 µM) with GABOB (100 µM) on mean tail current (left) and G/Gmax (right) relationships for KCNQ2/3 (n = 5) calculated from traces as in panel j. l Current fold-change at −60 mV exerted by IVA (500 µM) alone (from panel f) or with 100 µM GABOB, from data as in panel k (n = 5). mRight, dose responses for the shift in V0.5 of KCNQ2/3 activation induced by the leaf extract compounds shown on left, calculated from traces as shown in Figs 1, 2 and 5 (n = 4–16). All error bars indicate SEM

Fig. 6

MTX and IVA synergize to activate KCNQ2/3 and protect against seizures. a Averaged traces showing effects of IVA and MTX on KCNQ2/3 (n = 5). Voltage protocol as in Fig. 1d. b Effects at −60 mV highlighted, from traces as in a. c Mean tail current and G/Gmax from traces as in a (n = 5). d Mean effect of IVA (500 µM) + MTX (30 µM) on KCNQ2/3 activation at +40 mV and deactivation at −80 mV (n = 5). ***p = 0.0009; **p = 0.001. e Mean KCNQ2/3 current fold-increase versus voltage induced by IVA and MTX alone (from Figs. 2 and 4) or in combination (from traces as in a); n = 4–9. f Averaged traces showing effects of leaf extract cocktail (compounds shown in g) on KCNQ2/3 (n = 7). Voltage protocol as in Fig. 1d. g Effects at −60 mV highlighted, from traces as in f. h Mean tail current and G/Gmax from traces as in f (n = 7). i Mean effect of leaf extract cocktail on rates of KCNQ2/3 activation (left) and deactivation (center; voltage protocol on right) (n = 7). *p < 0.05; **p < 0.01. j Binding position of IVA and MTX in KCNQ3 predicted by SwissDock using a chimeric KCNQ1–KCNQ3 structure model. k–m Effects of vehicle (n = 35) compared to IVA and MTX alone or in combination (n = 11–12) on k clonic seizure incidence, l tonic seizure incidence, and m seizure assay survival in a mouse PTZ chemoconvulsant assay. *p < 0.05; **p < 0.01; ***p < 0.001. Survival statistical analysis by chi-squared, all others by one-way ANOVA. All error bars indicate SEM. All box and whisker plots: box range, 25–75%, coefficient 1; whisker range, 5–95%, coefficient 1.5

Fig. 7

IVA and MTX synergize with RTG to lock open KCNQ2/3. a Averaged traces showing effects of high-dose RTG, IVA, and MTX alone or in combination on KCNQ2/3 (n = 5–31). Voltage protocol upper inset. b–e Analysis of traces as in a: b peak current; c tail current; dG/Gmax; e current fold-change versus voltage; compounds and combinations color-coded as in a. n = 5–31. f Mean effects of low-dose 1 µM RTG versus 1 µM of each of RTG, IVA, and MTX (1 + 1 + 1) versus 1 µM RTG + 10 µM IVA + 10 µM MTX (1 + 10 + 10) on KCNQ2/3 tail currents and G/Gmax versus prepulse voltages; n = 8–13. g KCNQ2/3 current fold-increase versus voltage induced by compounds as indicated alone or in combination; n = 8–13. **p < 0.01. h KCNQ2/3 current fold-increase versus voltage induced by compounds as indicated alone or in combination; n = 8–13. **p < 0.01. All error bars indicate SEM. All comparisons by one-way ANOVA

Fig. 8

RTG + MTX + IVA alter the pore conformation of KCNQ2 and KCNQ3. a, bLeft, exemplar traces; right, mean I/V relationships for KCNQ2/3 (Q2/Q3) or homomeric KCNQ2 or KCNQ3 channels as indicated, bathed in 100 mM K+, Rb+, Cs+, or Na+ in the (a) absence or (b) presence of RTG (10 µM) + MTX (30 µM) + IVA (500 µM); n = 6–7. c Relative ion permeabilities of KCNQ2/3 channels in the presence (green) or absence (black) of RTG (10 µM) + MTX (30 µM) + IVA (500 µM); n = 6–7. Quantified from traces and plots as in panels a, b. All error bars indicate SEM

Fig. 9

Leveraging heteromeric channel composition to lock open KCNQ2/3. a Averaged traces showing effects of RTG, IVA, and MTX alone or in combination, doses as indicated, on homomeric KCNQ2–5 channels (n = 4–11). Voltage protocol as in Fig. 1d. b–d Analysis of traces as in a: b tail current; cG/Gmax; d current fold-change versus voltage; compounds and combinations color-coded as in a. n = 4–11. e, f Effects of high-dose RTG + MTX + IVA on KCNQ2/3 versus KCNQ3* held at −120 mV for 25 s. e Representative traces; f mean peak (0.5 s) versus steady-state (25 s) current. **p < 0.01; n = 4. Box and whisker plots: box range, SEM, coefficient of 1; whisker range 5–95%, coefficient of 1.5. g Model summarizing findings. Squares represent subunits within tetrameric KCNQ2/3 channels (yellow, KCNQ2; pale blue, KCNQ3). h Possible distinct binding positions of RTG, IVA, and MTX in one binding site in KCNQ3 (left) versus a lower-positioned RTG binding site (center) that would overlap with binding sites for IVA and RTG (right), predicted by SwissDock using a chimeric KCNQ1–KCNQ3 structure model. Red, KCNQ3-W265; magenta, IVA; blue, MTX; yellow, RTG. Space-filling omitted from molecules in right panel for clarity. i Further possible poses RTG (yellow, no spacefill) that would overlap with IVA (magenta) and MTX (blue) in KCNQ3 chimera model as predicted by SwissDock. All error bars indicate SEM. All comparisons by one-way ANOVA