Site and Mechanism of ML252 Inhibition of Kv7 Voltage-Gated Potassium Channels

Abstract Kv7 (KCNQ) voltage-gated potassium channels are critical regulators of neuronal excitability and are candidate targets for development of antiseizure medications. Drug discovery efforts have identified small molecules that modulate channel function and reveal mechanistic insights into Kv7 channel physiological roles. While Kv7 channel activators have therapeutic benefits, inhibitors are useful for understanding channel function and mechanistic validation of candidate drugs. In this study, we reveal the mechanism of a Kv7.2/Kv7.3 inhibitor, ML252. We used docking and electrophysiology to identify critical residues involved in ML252 sensitivity. Most notably, Kv7.2[W236F] or Kv7.3[W265F] mutations strongly attenuate ML252 sensitivity. This tryptophan residue in the pore is also required for sensitivity to certain activators, including retigabine and ML213. We used automated planar patch clamp electrophysiology to assess competitive interactions between ML252 and different Kv7 activator subtypes. A pore-targeted activator (ML213) weakens the inhibitory effects of ML252, whereas a distinct activator subtype (ICA-069673) that targets the voltage sensor does not prevent ML252 inhibition. Using transgenic zebrafish larvae expressing an optical reporter (CaMPARI) to measure neural activity in-vivo, we demonstrate that Kv7 inhibition by ML252 increases neuronal excitability. Consistent with in-vitro data, ML213 suppresses ML252 induced neuronal activity, while the voltage-sensor targeted activator ICA-069673 does not prevent ML252 actions. In summary, this study establishes a binding site and mechanism of action of ML252, classifying this poorly understood drug as a pore-targeted Kv7 channel inhibitor that binds to the same tryptophan residue as commonly used pore-targeted Kv7 activators. ML213 and ML252 likely have overlapping sites of interaction in the pore Kv7.2 and Kv7.3 channels, resulting in competitive interactions. In contrast, the VSD-targeted activator ICA-069673 does not prevent channel inhibition by ML252.


Introduction
Neuronal Kv7 (KCNQ) voltage-gated potassium channels encode the noninacti v ating "M-curr ent" that modulates neur onal threshold and firing properties. 1 The primary subunits involved in generation of M-current in the CNS are Kv7.2 and Kv7.3, although alternati v e assemb lies inv olving Kv7.4 and Kv7.5 may also contribute. 2 , 3 Mutations in each of these subunits (Kv7. 2-7.5) have been identified in patients with neurological disorders including benign familial neonatal/infantile seizures, epileptic encephalopathy, and progressive hearing loss. [4][5][6][7][8] Due to their powerful modulator y r ole and r elev ance in disease, neuronal Kv7 channels have been identified as an important therapeutic target, particularly for development of antiseizure medications. 9 Many small molecules that modulate Kv7.2/Kv7.3 channels have been discov er ed, with a predominant focus on acti v ator compounds, [10][11][12][13][14] while far fewer Kv7 inhibitors have been identified. 15 , 16 , 17 Kv7 acti v ators typically shift the voltage-dependence of activation to hyperpolarized v olta ges by di v erse mechanisms, and in some cases increase the maximum macroscopic conductance. 18 Kv7 activ ators ar e di vided into at least tw o cate gories [pore-targeted or v olta g e sensing domain (VSD)-targ eted] based on their site and mechanism of action. 10 , 19-24 Both pore-and VSD-targeted subtypes can suppress seizures in animal models, and retigabine (a pore-targeted activator) is the first Kv7 activator to be appr ov ed for seizure treatment in humans. 10 , 12-14 , 17 , 25-27 Poretargeted acti v ators like r etiga bine or ML213 r equir e a T rp (T rp236 in Kv7.2, Trp265 in Kv7.3) in the pore-forming S5 segment for efficacy. 18 , 21 , 25 A combination of experimental and structural findings suggest the importance of a H-bond formed between these drugs and the Trp side chain. [28][29][30][31] In contrast, VSD-targeted drugs (eg, ICA-069673, ztz-240, ICA-27243) do not depend on this interaction but are sensitive to mutations in the VSD, and recent cr yo-EM structur es hav e highlighted a distinct binding site for ztz-240 in the VSD. 22 , 27 , 29 Endogenous suppression of Kv7 channels is typically mediated by Gq-coupled GPCR signaling that culminates in cleavage of PIP2, leading to increased neuronal excitability. 2 , 32 While some small molecule inhibitors of Kv7 channels have been identified and commonly used, these have not been investigated as extensi v el y as Kv7 acti v ators. Recent functional and structural studies have begun to r ev eal details regarding the mechanism and potential site of action of the most commonly used inhibitors linopirdine and XE991. 15 , 30 XE991 and linopirdine have been used to investigate roles of Kv7 channels in epilepsy, longterm potentiation, and neuroplasticity, and are also frequently used to experimentally confirm/validate the mechanism of candidate Kv7 acti v ators. 15 , 33-35 In addition to XE991 and linopirdine, ML252 has been identified from a high-throughput screen as a Kv7.2/Kv7.3 inhibitor, with stronger potency than XE991, 16 but its structure is distinct from linopirdine and XE991 and little is known about its molecular mechanism of action. In this study, we r e port the mechanism of ML252 inhibition of Kv7.2 and Kv7.3, which arises from interaction with a site that likely overlaps the Trp residue required for sensitivity to pore-targeted acti v ators. This feature leads to distinct functional interactions between ML252 and pore-vs. VSD-targeted acti v ator subtypes in-vitr o (electr ophysiological r ecordings), and in-vi v o (optical monitoring of neur onal acti vity in zebrafish larvae). These findings esta b lish a clear mechanism of action of a poorly characterized Kv7 inhibitor, and also demonstrate the importance of understanding potential interactions that may arise between inhibitors and acti v ators when v alidating mechanisms of action of candidate antiseizure medications.

Animal Ethics
Husbandry of animals was carried out as per the Canadian Council on Animal Care. The use of Xenopus laevis and Danio rerio (Zebrafish) wer e appr ov ed by the Uni v ersity of Alberta institutional Animal Care and Use Committees under protocols, AU00001752 and AUP00000077, r especti v el y.

Drug Storage
ML252 was obtained from Sigma-Aldrich. ICA-069673 and ML213 wer e obtained fr om Tocris. Stock solutions for man ual v olta ge clamp experiments were prepared in DMSO at 50 m m for ML252 and ML213, and 100 m m for ICA-069673. All stock solutions for automated patch clamp experiments (executed at Xenon Pharmaceuticals) wer e pr e par ed at 20 m m in DMSO. Working solutions were prepared in extracellular recording solutions at indicated concentrations on each experimental day.

Two-Electr ode Volta ge Clamp Recordings
Two-electr ode v olta ge clamp of Xenopus oocytes w as performed in a modified Ringer's solution (116 m m NaCl, 2 m m KCl, 2 m m MgCl2, 1.5 m m CaCl2, and 5 m m HEPES, pH 7.6) using an OC-725C v olta ge clamp (Warner). Current traces were acquired using a Digidata 1440A (Molecular Devices) and pClamp 10 softw ar e (Molecular Devices). Recordings were filtered at 2 kHz and sampled at 5 kHz.

Manual Patch-Clamp Recordings
Man ual patch-clamp r ecordings with r apid solution exc hange were performed using HEK cells transfected with plasmids encoding human Kv7.2 and Kv7.2 [W236F], as pr eviousl y described. 20 HEK293 cells were maintained in Dulbecco's modified eagle medium supplemented with 10% FBS and 1% penicillin/str e ptom ycin. Cells w er e gr own in Falcon tissue cultur etreated flasks, in a 5% CO 2 and 37 • C incubator. Cells were seeded into 12-well plates and allowed to settle for 24-48 h before transfection. Cells were transfected using jetPRIME DNA transfec-

CaMPARI Imaging of Neural Activity in Zebrafish
A zebr afish tr ansgenic line (Tg[elavl3: CaMPARI (W391F + V398L)]ua3144) expressing the calcium sensor CaMPARI was generated as previously described, by using the Tol2 vector, pDestTol2-elavl3: CaMPARI (W391F + V398L) (kindl y shar ed by Eric Schr eiter, HHMI J anelia). 27 , 42 , 43 CaMPARI undergoes photoconversion (PC) from a green to red emission when high intracellular calcium coincides with illumination by an intense PC light generated by a 405 nm LED array (Loctite). Zebrafish larvae were placed in E3 media [as per Westerfield (2007), 44 but without methylene blue] and exposed to the PC light for 300 s at a distance of 10 cm from the LED array, in various indicated experimental conditions. PC light was applied 30 min after drug application. During PC, the uncov er ed petri dish containing 10 mL of E3 media floated in a heat sink with w ater (r oom temperatur e). After PC, larv ae wer e anesthetized in 0.24 mg/mL tricaine methanesulfonate (MS-222, Sigma) and embedded in 2% low-gelling a gar ose (Sigma #A9045) for imaging. 27 Ima ges wer e collected as Z-stacks (4 μm ste ps) using a 20 ×/0.8 objecti v e, with a laser-point scanning confocal microscope (Zeiss 700) and are presented as maximum intensity projections. We used Imaris 7.6 (Bitman, Zurich), to identify and analyze the 3-dimensional region of interest corresponding to the hindbrain. Relati v e neural acti vity w as interpr eted by calculating the ratio between the red to green mean fluorescence intensities. Data points were presented as a red/green ratio for each individual fish analyzed. One-way ANOVA and Dunnett's post-hoc test were performed in GraphPad Prism Softw ar e (Version 7, GraphPad, San Diego, CA).

Molecular Docking
Molecular docking was performed using AutoDock Vina 4.2.6. 45 Pub lished cr yo-EM structur es of KCNQ2/Kv7.2 structur es in complex with either r etiga bine (PDB 7CR7) or ztz240 (PDB 7CR4) were used as templates, along with a pdb model of ML252 constructed using Maestr o (Schr odinger Inc.). Grid boxes in the Kv7.2 models w ere c hosen to encompass the ion conducting pore along with the pore-forming domain of a single subunit along with its interfaces with neighboring subunits and the v olta ge sensing domain.

ML252 Interacts with the Pore of Kv7.2 and Kv7.3
There is a paucity of knowledge regarding the mechanism of action of Kv7 inhibitors and their interactions with therapeuticall y r elev ant channel acti v ators. Curr ent inhibitors such as XE991 have some dr awbac ks as their subtype specificity is unclear and the drug onset and washout is r e ported to be slow or irr ev ersib le. 15 We inv estigated ML252, a r ecentl y identified Kv7 inhibitor with a pr eviousl y undescribed site and mechanism of action. 16 We first used TEVC recordings from X. laevis oocytes to assess ML252 inhibition of various Kv7 channels ( Figure 1 ). Kv7.2, Kv7.3 [A315T] (r eferr ed to as Kv7.3 * ), WT Kv7.3, Kv7.2/7.3 heteromers, and Kv7.5 channels wer e clearl y inhibited by ML252 with Kv7.2 channels exhibiting slightly stronger potency and more complete block compared to other Kv7 channel types ( Figure 1 A-C and F). Kv7.2/7.3 heteromers also exhibited more complete ML252 block than WT Kv7.3 or Kv7.3 * channels. In contrast, Kv7.1 was only weakly inhibited by ML252 ( Figure 1 F), which may be noteworthy as prior descriptions of ML252 r e ported Kv7.1 inhibition with appr ecia b le potency ( ∼2 μm ) but did not highlight the weak efficac y to war d Kv7.1. 16 To investigate the onset of ML252 inhibition, we pulsed channels to + 20 mV (1 s pulses, 0.1 Hz) during drug w ash-in ( Figur e 1 D and E) and observed that ML252 effects have a rapid onset, with complete block typically observed after the first pulse during drug wash-in. This rapid onset of ML252 contrasts with the pr eviousl y described slow onset of XE991 and linopirdine, which exhibit use-dependence and r equir e pr olonged channel acti v ation for significant channel block to occur. 15 , 46 We confirmed this finding, shown by exemplar traces in Figure 1 G in which oocytes expressing Kv7.2 were exposed to ML252 or XE991. ML252-mediated inhibition was complete after 1 pulse, whereas XE991 was only partially effecti v e ov er these short applications. Although these findings do not explicitl y measur e rates of ML252 binding to closed vs. open channels, they illustrate much faster inhibition by ML252 and suggest that unlike XE991, ML252 can bind to closed channels during the interpulse interval. If channels are exposed to the drugs for a sufficient duration, the extent of the block of heteromeric Kv7.2/Kv7.3 channels is similar for ML252 and XE991 in our experimental system ( Figure 1 H). We also compared XE991 and ML252 inhibition in HEK cells using an automated planar patch clamp approach. Both drugs caused similar maximal inhibition of Kv7.2/Kv7.3 or Kv7.3/Kv7.5, but ML252 exhibited more potent inhibition of Kv7.3/Kv7.5 heteromeric channels ( Figure S5).
To investigate the location of ML252 binding to Kv7 channels, we docked the compound into a r ecentl y pub lished structure of Kv7.2. 29 We were especially interested in drug poses near the region of the r etiga bine binding site (Kv7.    Figure  S1). We also used fast solution switching with patch-clamp recordings of transfected HEK293 cells to assess the kinetics and sensitivity of ML252 inhibition. In these HEK293 recordings, Kv7.2[W236F] was also resistant to ML252 ( Figure 2 D to H). ML252 (30 μm ) generated nearly complete inhibition of WT Kv7.2 with a rapid onset ( Figure 2 D, G, and H), whereas Kv7.2[W236F] channels exhibited slower and incomplete bloc k ( F igure 2 E, G, and H). In these experiments, it is note worth y that the Trp236Phe mutation did not completely abolish ML252 sensitivity, as ther e wer e v aria b le lev els of ML252 inhibition of Kv7.2[W236F] ( Figure 2 E and H). This observation may indicate the presence of lower affinity sites for ML252 inhibition (either in the mutated r etiga bine binding site, or elsewhere in the channel), as the concentration used in these rapid perfusion experiments is high r elati v e to the ML252 IC50 determined in Figure 1 (0.88 μm ), although we have not explored this further. Kv7.2 channels also exhibit v aria b le degr ees of rundown in HEK293 cells and this may contribute to the v aria bility of ML252 inhibition.
Inter estingl y, ML252 washout from WT Kv7.2 channels exhibited a prominent delay leading to sigmoidal washout kinetics, wher eas w ashout fr om Kv7.2[W236F] channels w as monoexponential and faster ( Figure 2 F, inset). This finding may reflect the lower affinity modes of ML252 binding stated a bov e and could imply that multiple ML252 unbinding events are required for relief of inhibition of WT Kv7.2 channels. Regardless, considering the large impact of mutating Trp236, these results suggest that the Trp236 in Kv7.2 (or Trp265 in Kv7.3), is critical for ML252 inhibition in addition to its esta b lished r ole in sensiti vity to por etargeted acti v ators. These findings also pr ovide an explanation for the enhanced ML252 sensitivity of Kv7.2 and Kv7.3 r elati v e to Kv7.1 (which has a Leu at the pore position equi v alent to Kv7.2[Trp236], Figure 1 F). 16 , 18 , 47 Competition Between ML252 and Pore-Targeted Activ a tors Based on the potential overlap of binding sites for ML252 and other pore-targeted activators, we tested the functional relationship between ML252 and a por e-targeted acti v ator, ML213 ( Figure 3 ). Application of ML213 enhanced Kv7.2 channel activation as demonstrated by conductance of currents at negati v e v olta ges ( Figur e 3 A, middle), due to a hyperpolarizing shift of the v olta ge-de pendence of acti v ation pr eviousl y described for this drug. 27 , 28 However, application of ML252 (10 μm ) in combination with ML213, failed to inhibit channels ( Figure 3 A and B, recordings from oocytes). This observation suggested that ML252 interaction with Trp236 could be pr ev ented by ML213, consistent with the idea that these compounds compete for binding to the modulator y por e site (rather than ML252 acting as a por e b locker). Further experiments in oocytes demonstrated that ML252 inhibition of Kv7.3 * (Figure S2) or Kv7.2 ( Figure S3) was attenuated by ML213. Mor eov er, this functional interaction was also observed if drugs were applied sequentially, irrespective of the order of application. This is illustrated using rapid perfusion switches in patch clamp of HEK cells held at + 20 mV ( Figure 3 C to E), in which ML252 application either after ( Figure 3 D) Figure S3A). In these experiments ML252 application to channels "pr e-acti v ated" with ML213 did not suppr ess curr ent (Figur es S2C and S3A). Similarl y, if channels were first inhibited by ML252, subsequent ML213 application led to rapid rescue of currents ( Figures S2D and S3B). These results suggest a functional inter action betw een ML213 and ML252, likely arising from competition for association with Kv7.2 Trp236 (of Kv7.3 Trp265).

VSD-Targeted Activ a tors Do Not Influence ML252 Inhibition
Acti v ators of Kv7.  Figure S3D), in contrast to the effects of ML213 described in Figure 3 . A consistent explanation for the different functional interactions ML213 or ICA-069673 is that these acti v ators occupy different binding sites, and only ML213 is a b le to pr ev ent binding and inhibition by ML252 in the pore site.
We recognized that the drugs tested hav e differ ent affinities and this ma y ha ve impacted our findings in experiments conducted at single concentrations of the individual drugs. In order to test our findings over wider concentration ranges, and better c har acterize the functional interactions of ML252 with poreor VSD-targeted acti v ators, we used automated patch clamp of HEK cells sta b l y expr essing Kv7.  Figure S4D). In contrast, the GV shift mediated mediated by ICA-069673 is not attenuated by ML252 ( Figure  S4C).

In-Vivo Validation of ML252-mediated Kv7.2/Kv7.3 Inhibition
We tested ML252 inhibition and its interaction with acti v ator drugs in-vi v o, using zebrafish larv ae. We used a pr eviousl y described tr ansgenic zebr afish line expressing a calcium sensor, 27 , 42 , 43 CaMPARI, dri v en by a pan-neur onal pr omoter for CNSr estricted expr ession. CaMPARI under goes photo-s witching by 405 nm light applied in the presence of high calcium, converting fr om gr een emission to r ed emission. The red to green ratio reflects neural activity during the PC period, and can be manipulated by application of pro or anticonvulsant drugs. 27 , 48 , 49 We tested the effects of various combinations of ML252, ICA-069673, and ML213 in a CaMPARI photoconversion assay ( Figure 6 ). ML252 (30 μm ) increased neural acti vity measur ed in the hindbrain area, likely reflecting enhanced excitability caused by Kv7 inhibition ( Figure 6 A). Co-application with ML213 attenuated ML252-mediated effects on neuronal activity (r ed:gr een r atio, F igure 6 B). In contr ast, co-application of ICA-069673 did not pr ev ent ML252 effects ( Figure 6 ). Consistent with in-vitro electrophysiological findings, these data indicate that a functional inter action betw een ML213 and ML252 can attenuate ML252-mediated inhibition, whereas ML252 r emains effecti v e in the pr esence of the VSD-targeted acti v ator ICA-069673.

Discussion
Screening and c har acterization of the pharmacology of Kv7 channels has primarily focused on activator compounds like r etiga bine , whic h have ther apeutic benefits for treatment of e pile psy and other neurological disorders. 49 , 50 Kv7 inhibitors have been less studied in terms of basic mechanism or biological outcomes, although the most widely used Kv7 inhibitors  have been investigated in the context of cogniti v e enhancement, Parkinson's disease, and neuroplasticity. 34 , 35 These compounds are also frequently used as experimental tools to validate the molecular target of acti v ators (by anta gonizing the effects of a candidate acti v ator). Recent strides hav e been made in terms of understanding the mechanism of action of XE991 and linopirdine, including the demonstration that these drugs exhibit state-dependent inhibition, requiring prolonged or repetitive acti v ation of Kv7 channels for drug onset to occur. 15 Also, a potential binding site of linopirdine near the intracellular gate of Kv7.4 was identified using cryo-EM, although it is noteworthy that prior functional work indicates an extracellular site of action of linopirdine, suggesting that our understanding of these classically used Kv7 inhibitors remains incomplete. 30 , 51 ML252 is a more recently identified Kv7 inhibitor, and thus it has been less fr equentl y used, and there have been few prior insights into its underlying mechanism. 16 Our study has r ev ealed a likely site and mechanism of action of ML252, along with implications related to its experimental use for validating candidate Kv7 acti v ators. Computational appr oaches, along with functional experiments with mutant channels ( Figures 1  and 2 ), suggest that ML252 inhibition inv olv es interaction with the pore Trp (Kv7.2 Trp236, Kv7.3 Trp265) r equir ed for sensiti vity to pore-targeted activator drugs. There is general consensus regarding the likely importance of an H-bond interaction between the Kv7.2 Trp236 indole N-H and carbonyl/carbamate moieties in drugs including ML213 and r etiga bine. 28 , 29 , 31 Similarly, ML252 comprises a carbonyl that is predicted to interact with the Trp236 side chain, but other features of the drug must cause it to act as an inhibitor rather than an acti v ator. This is consistent with the initial c har acterization of ML252, 16 which r e ported that a small change of an ethyl group to a hydrogen or a  fluorine switched the effects of ML252 from inhibition to potentiation. Similarly, small modifications to activator drugs like r etiga bine hav e been r e ported to switch their effects to inhibition. 50 Overall, ML252 appears to occupy a region within or overlapping the modulatory site that accommodates pore-targeted acti v ator drugs, but with a distinct functional outcome. Thus, our findings suggest that ML252 inhibition is caused by an allosteric effect on gating, rather than direct occlusion of the c hannel pore .
The overlapping site of action of ML252 and pore-targeted acti v ators leads to distinct functional outcomes when inhibitors and acti v ators ar e applied in combination. In-vitro ( Figures 3 -5 ) and in-vivo ( Figure 6 ) studies illustrated that por e-targeted acti v ators can effecti v el y counteract ML252 inhibition, likely due to competiti v e displacement of the inhibitor at the overlapping binding site ( Figure 7 ). In contrast, the VSD-targeted acti v ator ICA-069673 did not compete with the inhibitor ( Figure 7 ). In the context of the growing understanding of the comple x pharmacolo gy of Kv7 acti v ators, these distinct outcomes ar e consistent with functional and structural evidence for multiple sites of action of pore-vs. VSD-targeted activator drugs. 22 , 23 , 29 , 30 , 52 It is now clear that ther e ar e m ultiple potential mechanisms that can be exploited in the design of Kv7 acti v ators, either by trapping the acti v ated VSD conformation, or stabilizing the open pore conformation. In terms of practical applications of ML252 vs. XE991 or linopirdine for mechanistic validation of acti v ator compounds, the r ecognition that ML252 will have distinct interactions with pore-or VSD-targeted acti v ators is an important consideration for experimental design.
In comparison to the more widely used inhibitors linopirdine and XE991, ML252 has certain benefits. Firstly, ML252 inhibition and washout is rapid and likely binds to both open and closed states of the c hannel ( F igures 1 and 2 ). In contrast, XE991 and linopirdine exhibit use-dependence and require prolonged channel activation or repetitive stimulation for the block to set in, and washout of these drugs is also slow and often incomplete. 15 A second benefit of ML252 is that it is effecti v e tow ard most Kv7 channels including Kv7.2/Kv7.3, and Kv7.3/Kv7.5 ( Figures 1 ; Figure S5), and has been previously used to inhibit neuronal channels, 53 but only weakly blocks Kv7.1. In contrast, XE991 and linopirdine inhibit Kv7.1 expressed in cardiom y ocytes at concentr ations < 1 μm (although this is r e ported to be weaker when Kv7.1 is associated with modulatory KCNE subunits). 16 , 54 However, these benefits should be balanced with the aforementioned potential for interactions if ML252 is planned to be used in combination with a Kv7 activator, as is often done in target validation experiments. That is, ML252 can antagonize a pore-targeted activator, but our findings show that this r equir es m uc h higher concentr ations than r equir ed if applied alone, and this could be pr ob lematic in terms of deli v ering drugs in an in-vi v o model. In contrast, ML252 appears to effecti v el y anta gonize the effects of a VSD-targeted acti v ator. Ov er all, understanding the mec hanism of action of a potential acti v ator drug may influence the choice and concentration of inhibitors used in combination for target validation or other applications.
In summary, our work reveals detailed features of the site and mechanism of ML252 inhibition of Kv7.2/Kv7.3 channels. ML252 has a rapid onset and targets a modulatory site that is also affected by pore-targeted Kv7 activators. ML252 sensitivity r equir es a conserv ed Trp r esidue that interacts with por etargeted acti v ators, and this leads to distinct functional outcomes when ML252 is co-applied with pore-vs. VSD-targeted acti v ator drugs. Overall, these findings highlight the effectiveness of ML252 as an inhibitor of Kv7.2 and Kv7.3, and solidify a gro wing bod y of evidence distinguishing Kv7 modulators based on sites of action in the pore or VSD.