Chlorpromazine binding to the PAS domains uncovers the effect of ligand modulation on EAG channel activity

Ether-a-go-go (EAG) potassium selective channels are major regulators of neuronal excitability and cancer progression. EAG channels contain a Per–Arnt–Sim (PAS) domain in their intracellular N-terminal region. The PAS domain is structurally similar to the PAS domains in non-ion channel proteins, where these domains frequently function as ligand-binding domains. Despite the structural similarity, it is not known whether the PAS domain can regulate EAG channel function via ligand binding. Here, using surface plasmon resonance, tryptophan fluorescence, and analysis of EAG currents recorded in Xenopus laevis oocytes, we show that a small molecule chlorpromazine (CH), widely used as an antipsychotic medication, binds to the isolated PAS domain of EAG channels and inhibits currents from these channels. Mutant EAG channels that lack the PAS domain show significantly lower inhibition by CH, suggesting that CH affects currents from EAG channels directly through the binding to the PAS domain. Our study lends support to the hypothesis that there are previously unaccounted steps in EAG channel gating that could be activated by ligand binding to the PAS domain. This has broad implications for understanding gating mechanisms of EAG and related ERG and ELK K+ channels and places the PAS domain as a new target for drug discovery in EAG and related channels. Up-regulation of EAG channel activity is linked to cancer and neurological disorders. Our study raises the possibility of repurposing the antipsychotic drug chlorpromazine for treatment of neurological disorders and cancer.

EAG channels are part of the KCNH voltage-gated potassium channel family that also includes EAG-related gene (ERG), also known as Kv11, and EAG-like K ϩ , also known as Kv12, subfamilies (14). KCNH channels are tetramers, with each of the four subunits containing six membrane spanning segments (S1-S6) and an intervening pore-forming loop ( Fig. 1a) (14 -16). The S1-S4 segments form a voltage sensor (VS) domain, whereas the S5 and S6 of all four subunits together with the pore-forming loops form a centrally located pore domain. The signature feature of KCNH channels is the presence of the PAS domain in their N-terminal region and a cyclic nucleotide-binding homology (CNBH) domain in the C-terminal region. These intracellular PAS and CNBH domains harbor numerous genetic mutations linked to cancer in EAG channels and cardiac arrhythmias in ERG channels (17)(18)(19). The CNBH domains of KCNH channels are structurally similar to the cyclic nucleotide-binding domains of cyclic nucleotide-gated (CNG) and hyperpolarization-activated cyclic nucleotide-gated (HCN) channels (14 -16, 20). However, unlike CNG and HCN channels, KCNH channels are not modulated by direct binding of cyclic nucleotides to the CNBH domain (14,21). Instead, a short ␤-strand, known as an intrinsic ligand, occupies the cavity where cyclic nucleotides would bind in HCN and CNG channels (22,23).
The PAS domains of KCNH channels share structural similarity with PAS domains in other proteins (14,24). Although in other proteins PAS domains frequently function as ligandbinding domains (25,26), the PAS domains' ability to regulate KCNH channel function via ligand binding has not been reported before. Determining whether the PAS domain can regulate KCNH channels via ligand binding is crucial for understanding the molecular pathways involved in KCNH channel gating (opening and closing) and also for directing drug discovery efforts targeting EAG channels. This work was supported by the NIGMS, National Institutes of Health Grant R01GM124020 (to T. I. B.). The Biacore Molecular Interaction Shared Resource is supported by National Institutes of Health Grants P30CA51008 and 1S10OD019982-01. The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains Table S1. 1

cro ARTICLE
In this study, we used surface plasmon resonance (SPR) and tryptophan fluorescence to show that a small molecule chlorpromazine hydrochloride (CH) binds to the PAS domain of EAG channels. CH inhibited currents recorded from EAG channels in a concentration-dependent manner. Deletion of the PAS domain significantly decreased the apparent affinity and potency of CH inhibition. These results suggest that the PAS domain can regulate EAG channel function by binding small molecule ligands. Importantly, our findings indicate that CH, which is a widely used antipsychotic drug, can be repurposed for treatment of cancer and neurological disorders associated with increased EAG channel activity.

Chlorpromazine binds to the PAS domain of EAG channels in a concentration-dependent manner
In a recent study, we screened small molecules in the Spectrum Library of chemical compounds for binding to the intracellular PAS and CNBH domains of KCNH channels using the SPR method (27). From this screen, CH (structure shown in Fig. 1b) emerged as a potential EAG PAS domain small molecule binder. To further investigate the concentration dependence and specificity of CH binding, we immobilized the isolated PAS and CNBH domains of mouse EAG (mEAG) and human ERG (hERG) channels on a NTA sensor chip (Fig. 1c) Figure 1. CH binding to the PAS domain of EAG channels detected with SPR. a, ribbon representation of the full-length cryo-EM structure of rat EAG channels (Protein Data Bank accession no. 5K7L) viewed from the side and down the pore of the channel. The PAS domain is green, the CNBH is blue, and the transmembrane segments are gray. Ribbon representations were created using PyMOL. b, chemical structure of CH. c, schematic of the PAS and CNBH domains immobilized on the NTA sensor chip using Ni 2ϩ -NTA coupling. d and e, SPR sensorgrams recorded for the immobilized PAS domain of mEAG channel (d) and the corresponding dose dependence of the SPR response (e). f-h, SPR sensorgrams for the immobilized PAS domain of ERG (f) and the CNBH domains of EAG (g) and ERG (h) channels recorded with the indicated concentrations of CH.

Mechanism of EAG channel inhibition by chlorpromazine
response over a range of CH concentrations for the four target proteins. Application of CH to the PAS domain of mEAG channels increased the SPR response in a concentration-dependent manner (Fig. 1, d and e) via direct binding to the immobilized proteins. It is noteworthy that the PAS domains of mEAG and human EAG1 channels are 99% identical. Therefore, we expect our findings to hold for human EAG channels as well. No substantial direct binding was observed for the PAS domain of hERG channels or for the CNBH domains of mEAG and hERG channels ( Fig. 1, f-h). The PAS and CNBH domains have different structural folds ( Fig. 1a) (14 -16). Therefore, CH binding is PAS domain structurespecific. At concentrations of Ͼ30 M, CH displayed a nonspecific binding to the NTA chip, resulting in a higher SPR response for the control surface than for the surfaces with immobilized proteins. Nonspecific bindings of analytes to the control surface of the Ni 2ϩ -NTA sensor chip can limit the applicability of the SPR-based experiments (28). Because of this limitation at higher CH concentrations, we were unable to obtain a full dose-response and determine the CH binding affinity for the isolated mEAG PAS domain with SPR.
To further examine CH binding with a method that does not require PAS domain immobilization, we used tryptophan fluorescence as a reporter of CH binding. Binding of a ligand in the vicinity of a Trp residue could change tryptophan fluorescence because of its environmental sensitivity (21,29,30). The PAS domain of mEAG channels contains two endogenous Trp residues (Fig. 2a). Therefore, we tested whether CH binding to the PAS domain can be detected by changes in the fluorescence of the Trp residues.
Upon excitation at 290 nm, a commonly used wavelength for Trp excitation, the emission spectra of the mEAG PAS domain displayed a strong fluorescence signal with a peak at 343 nm. Application of CH decreased the peak fluorescence intensity in a concentration-dependent manner (Fig. 2b). To test whether the changes in the fluorescence signal are specific to the PAS domain protein rather than a nonspecific effect such as an inner filter effect, we acquired the emission spectra of free tryptophan in solution. Unlike for the PAS domain, for free tryptophan application of CH had no concentration-dependent effect on the fluorescence signal (Fig. 2c). Therefore, the observed decrease in the fluorescence intensity of the mEAG PAS domain is due to the changes in the environment of Trp residues caused by a specific binding of CH to the PAS domain. To determine CH-binding affinity, plots of the changes in the peak fluorescence intensity versus the total CH concentration were fitted with Equation 2. This analysis revealed the binding affinity of 1 Ϯ 0.7 M for CH (Fig. 2d). Taken together, the results of the SPR and fluorescence-based experiments indicate that CH is a small molecule binder of the PAS domain of mEAG channels that binds in a concentration-dependent and structure-specific manner.

CH inhibits currents from EAG channels in a concentrationdependent manner without affecting kinetics of deactivation
To determine the functional effect of CH on EAG channels, full-length mEAG channels were expressed in Xenopus laevis oocytes, and currents from mEAG channels were recorded in the absence and presence of 50 M CH (Fig. 3a). The inside-out con- The peak emission intensity was at 343 nm for mEAG PAS and at 353 nm for free tryptophan. The binding affinity for CH was 1 Ϯ 0.7 M. For clarity, the data point for Trp fluorescence in the presence of 1 M CH was moved slightly to the right to prevent overlap with the data for mEAG PAS.

Mechanism of EAG channel inhibition by chlorpromazine
figuration of the patch-clamp technique was used for the current recordings so that the ligand can be directly applied to the intracellular side of the channels containing the PAS domains. CH dramatically inhibited steady-state and tail currents from mEAG channels (Fig. 3a). The inhibition was concentration-dependent with an IC 50 of 3.7 Ϯ 0.7 M (Fig. 3b, n ϭ 4). The effect of CH on the tail-current deactivation kinetics was not statistically significant (Fig. 3, c and d). The averaged deactivation time constant was 3.6 Ϯ 0.3 ms in the absence and 4.2 Ϯ 0.1 ms in the presence of 10 M CH for tail currents recorded following a ϩ30-mV test pulse and 3.5 Ϯ 0.3 ms in the absence and 4.7 Ϯ 0.2 ms in the presence of 10 M CH for tail currents recorded following a ϩ70-mV test pulse (p ϭ 0.5 for both ϩ30 mV and ϩ70 mV by Student's t test). The tail currents recorded in the presence of 50 M CH were too small to be included in the analysis of the deactivation kinetics. The lack of CH effect on the kinetics of tail currents suggests that CH does not affect the return of the VS domain to the resting state.

Voltage dependence of CH inhibition of currents from EAG channels
In the absence of CH, EAG channels activated with voltage with the V1 ⁄ 2 of Ϫ35.8 mV. Application of CH decreased the channel conductance and shifted the conductance versus voltage plots to more depolarized potentials (Fig. 4a). Importantly, current inhibition by CH was mostly voltage-independent (Fig.  4b). For instance, for tail currents recorded in the presence of 10 M CH, the inhibition was 45.3 Ϯ 7.2% following a 0-mV test

Mechanism of EAG channel inhibition by chlorpromazine
pulse and 59.6 Ϯ 3.8% following a ϩ70-mV test pulse, which are not statistically significantly different values (p ϭ 0.13 by Student's t test). The voltage independence of the inhibition suggests that the EAG current inhibition by CH is caused by changes in the gating mechanism rather than a pore block of EAG channels.

Deletion of the PAS domain significantly decreases CH inhibition of EAG currents
To determine whether binding of CH to the PAS domain is directly involved in the inhibition of CH currents, we examined the effect of CH on a mutant mEAG channel with a deletion of the PAS domain (⌬PAS). In our hands, deletion of the PAS domain substantially decreased the surface expression of mEAG channels, making it difficult to study the inhibition by recording currents from excised inside-out patches. To circumvent this issue, we recorded currents from the WT and ⌬PAS mutant mEAG channels using two-electrode voltageclamp (TEVC) technique. Taking advantage of the membrane permeability of CH, we then studied the effect of CH on mEAG channel currents over a range of CH concentrations applied to the bath solution. mEAG tail currents recorded with TEVC deactivated too fast to be used for a tail current-based channel gating analysis (Fig. 5a), consistent with the previous report (31). Therefore, only steady-state currents were analyzed for experiments carried out using TEVC.
Consistent with the results for tail currents obtained using excised inside-out configuration of the patch-clamp technique, steady-state currents from WT mEAG channels recorded with TEVC were also inhibited by CH (Fig. 5, a and c). Notably, the IC 50 of the CH inhibition for WT channels was ϳ10-fold higher than the IC 50 for the excised inside-out patches. Most likely this difference reflects the decrease in the effective concentration of CH caused by the oocyte volume and various intracellular molecules that could bind CH, decreasing the effective concentration available to affect EAG channels.
Similar to the results for excised patches, CH shifted the conductance versus voltage plots to more depolarized potentials and decreased the channel conductance, which was largely voltage-independent from 0 to ϩ50 mV. At voltages higher than ϩ50 mV, both in the absence and in the presence of CH, EAG channel conductance started to decrease (Fig. 5d). Most likely this reflects a voltage-dependent block by various potential channel pore-blocking molecules present in the oocytes. This voltage-dependent decrease was absent for currents recorded in the excised inside-out patch-clamp configuration, where the intracellular and extracellular solutions were devoid of impurities (Fig. 4a).
Deletion of the PAS domain uncovered an inactivation-like behavior at high voltages not seen in WT channels and shifted the current-voltage relationship to more depolarized potentials (Fig. 5, b, d, and e), as reported previously (32,33). Importantly, deletion of the PAS domain significantly decreased CH inhibition of mEAG currents (Fig. 5, b and c) and eliminated the shift of the conductance versus voltage plots to more depolarized potentials (Fig. 5e). The IC 50 of CH inhibition was 29.7 Ϯ 0.7 M for WT and 53.6 Ϯ 8.2 M for ⌬PAS mEAG channels. In addition, the magnitude (potency) of the current inhibition significantly decreased in the absence of the PAS domain. Although the averaged current inhibition observed at 300 M CH was 63.3 Ϯ 4.9% for WT channels, it decreased to 25.1 Ϯ 2.9% for ⌬PAS channels. Moreover, the residual inhibition of ⌬PAS channels that is apparent only at voltages higher than ϩ20 mV (Fig. 5, e and f) could be due to the voltage-dependent pore block caused by various intracellular molecules present in oocytes, as suggested by the results in Fig. 5d. Overall, the current inhibition by CH was very weakly voltage-dependent (Fig.  5f), consistent with the results obtained for currents recorded from excised inside-out membrane patches (Fig. 4b). The voltage dependence of the current inhibition was the same for the WT and ⌬PAS channels, suggesting that the mutation did not cause drastic changes in the pore energetics and CH pore accessibility. Taken together, the results of the TEVC experiments demonstrate that the PAS domain deletion significantly decreased the effect of CH on EAG currents, strongly suggesting that the PAS domain is directly involved in the CH regulation of EAG currents.

Discussion
Using a combination of SPR and tryptophan fluorescence, we show that a small molecule CH directly binds to the PAS domain of EAG channels. When applied to the excised inside-b a

Mechanism of EAG channel inhibition by chlorpromazine
out patches of membranes expressing EAG channels, CH inhibits currents from EAG channels. Deletion of the PAS domain substantially decreases the potency and apparent affinity of CH inhibition. These results strongly suggest that CH binding to the PAS domain modulates EAG channel currents. To our knowledge, this is the first indication of the PAS domain function as a ligand-binding domain in EAG channels and its capacity to modulate EAG channel currents via the ligand binding. CH is clinically used as an antipsychotic medication. Therefore, our study demonstrates a potential of repurposing CH for treat-ment of cancer and neurological disorders associated with increased EAG channel activity.
PAS domains of EAG and other KCNH channels contain a cavity formed by the ␤-strands and flanking ␣-helices (Fig. 2a,  gray mesh). It is tempting to speculate that the cavity might provide the CH-binding site. However, a structural analysis of the PAS domains suggested that the cavity lacks polar residues necessary for supporting small molecule binding (24,34). Structural studies are necessary to determine whether CH binds inside or outside of the PAS domain cavity.

Mechanism of EAG channel inhibition by chlorpromazine
CH directly inhibited currents from EAG channels in a concentration-dependent and weakly voltage-dependent manner. CH inhibition of EAG channels could result from the effect of CH on EAG channel gating via the PAS domain or a voltagedependent channel pore block. We feel that our data favor the former mechanism. A voltage-dependent pore block would decrease channel conductance as membrane voltage increases. However, our results indicate that for currents recorded from excised inside-out patches, in the presence of CH, EAG channel conductance does not decrease with increases in membrane voltage (Fig. 4a). In fact, for voltages of Ϫ30 mV and higher, EAG channel conductance is essentially constant and voltageindependent in the presence of CH. Additional evidence supporting the effect of CH through the binding to the PAS domain comes from experiments on the mutant EAG channels with the deletion of the PAS domain (⌬PAS). The deletion of the PAS domain essentially removed CH inhibition of EAG channels at voltages of Ͻ30 mV (Fig. 5e) and significantly decreased the effect of CH on EAG currents at voltages of Ն30 mV (Fig. 5c). We think that the residual CH inhibition of EAG currents at voltages of Ͼ30 mV is due to a voltage-dependent pore block by various small molecules, other than CH, present intracellularly in oocytes. This explanation is further strengthened by the voltage-dependent decline of currents from WT EAG channels detected at high voltages in the absence of CH (Fig.  5d). Although it is possible that the ⌬PAS mutation could directly or allosterically affect the energetics of the pore, similar voltage dependence of the CH inhibition for the WT and ⌬PAS channels suggests that drastic changes to the pore energetics are unlikely. It is noteworthy that CH also inhibits hERG channels in a voltage-dependent manner with the current inhibition increasing with membrane depolarization (35). hERG channels are notorious for being susceptible to inhibition by various compounds (36). Unlike for hERG channels, very few inhibitors (especially clinically relevant ones) have been identified for EAG channels. Our SPR-based results indicate that CH does not bind to the PAS domain of hERG channels (Fig. 1f), suggesting that the proposed PAS domain-dependent CH inhibition is unique to EAG channels.
How can CH binding to the PAS domain inhibit EAG currents? In KCNH channels, PAS domains form intersubunit interactions with the CNBH domains, where a PAS domain from one subunit interacts with the CNBH domain from the adjacent subunit (Fig. 1a) (15,(37)(38)(39). In EAG and ERG channels, the PAS/CNBH domain interaction favors the open state of the channel (31, 40) (Fig. 6, left half). CH binding to the PAS domain, either directly or allosterically, could weaken the PAS/CNBH domain interaction, causing decreases in EAG channel currents (Fig. 6, right half). In KCNH channels the PAS domain also forms interactions with the VS domain (Figs. 1a and 6, left half) (15). The PAS/VS domain interactions are functionally important because deletion of the PAS domain, either directly or allosterically, shifts the current-voltage relationship of EAG channels to more depolarized potentials (Ref. 33 and Fig. 5, d  and e). Therefore, CH binding, either directly or allosterically, could change the PAS/VS domain interactions, causing changes in the pore energetics (Fig. 6, right half). Future studies are necessary to test the possible pathways of CH action in EAG channels.
Inhibition of the EAG channel activity using RNAi and EAG channel-specific antibody blockers decreases tumor growth (12,13). It is noteworthy that all known and functionally examined mutations in EAG channels associated with epilepsy and Zimmerman-Laband and Temple-Baraitser syndromes are gain-of-function mutations that increase EAG channel activity (3,4,41). Therefore, EAG channel inhibitors have high therapeutic potential for treatment of both cancer and neurological disorders. CH is an Food and Drug Administration-approved drug that has been extensively used as an antipsychotic medication (42). Our study presents a strong rational for repurposing CH for treatment of cancer and neurological disorders.
In summary, we show that CH is a novel ligand of EAG channels. To our knowledge, our study presents the first evidence that a ligand can bind to the PAS domain and regulates EAG channel activity via this binding. Therefore, our study sets a precedent of KCNH channel regulation via ligand binding to the PAS domain and indicates that this mechanism should be taken into consideration for future studies of KCNH channel gating and drug discovery efforts targeting KCNH channels.

Protein purification
PAS and CNBH domains of mEAG1 and hERG1 channels were purified with Ni 2ϩ -NTA and size-exclusion chromatography, as previously described (21,22,27). The start and end of the amino acid sequence of the domains used in this study and their gene identifier numbers are indicated in the Table S1. Briefly, the genes encoding the domains were subcloned into pETM11 bacterial expression vector containing an N-terminal His 6 affinity tag and expressed in BL21 (DE3) cells. Protein expression in BL21 (DE3) cells was induced with 1 mM isopropyl ␤-D-1-thiogalactopyranoside at 18°C. The cells were harvested by centrifugation and resuspended in buffer A (150 mM KCl, 10% glycerol, 1 mM tris(2-carboxyethyl)phosphine, 30 mM HEPES, pH 7.5) supplemented with 1 mM 4-(2-aminoethyl) benzenesulphonyl fluoride hydrochloride and 2.5 mg/ml DNase I. The cells were lysed with an EmulsiFlex C-5 homogenizer (Avestin), and insoluble protein was separated by centrifugation at 30,000 rpm for 1 h at 4°C in a Beckman 45 Ti rotor. The supernatant was loaded onto His-Trap HP column (GE Healthcare). The column was washed with buffer A, and the proteins were eluted with buffer A with added 500 mM imidazole. The proteins were further purified with sizeexclusion chromatography on a Superdex 200 Increase column (GE Healthcare) equilibrated with buffer A. The molecular weight of the purified proteins was verified with Coomassie Blue-stained gels and with MS (electrospray) at Georgetown Proteomics and Metabolomics Core Facility. The purified proteins were stored at Ϫ80°C in small aliquots before use.

Protein immobilization for SPR
A Biacore 4000 SPR instrument was used in all SPR experiments. The purified PAS and CNBH domains of mEAG and hERG channels were immobilized on a NTA chip (GE Healthcare), as previously described (27,43). Immobilizations of the proteins were performed in HBS-P buffer (150 mM NaCl, 10 mM HEPES, 0.05% (v/v) surfactant P20, pH 7.4). The NTA sensor surface was first activated with a 1-min injection of 0.5 mM NiCl 2 . The coupling of the Ni 2ϩ -NTA chip surface groups with the His 6 -tagged proteins was then achieved by 2.5-min injections of the proteins at 10 -200 nM concentrations. After the initial capturing, the proteins were covalently cross-linked via 20-s injections of NHS-EDC carboxyl-reactive cross-linkers to prevent protein loss from the chip surface with successive analyte and buffer injections. This was followed by 20-s injection of 1 M ethanolamine to block the remaining reactive sites. The proteins were captured at ϳ1000 -3000 response units (RU) (1 RU ϭ 1 pg of protein/mm 2 ). The SPR data were doubly corrected for the nonspecific binding by subtracting the SPR response to the blank (buffer only) injections and also response to the control surface with no immobilized protein.

Fluorescence measurements
Fluorescence intensity was recorded in a quartz cuvette with a 100-l chamber using a PTI QuantaMaster spectrofluorom-eter (Horiba Jobin Yvon) and Felix GX 4.9 software. The sample was excited at 290 nm, and the emission spectra were recorded from 300 to 450 nm for 5 M mEAG PAS and 5 M free tryptophan in the absence and presence of various concentrations of CH, and also for CH alone over the range of the examined CH concentrations. In subsequent analysis, the fluorescence intensities for mEAG PAS and free tryptophan samples were background subtracted. For the background subtraction, the fluorescence intensity of CH at a given concentration was subtracted from the fluorescence intensity of the sample with the same CH concentration. Observed emission intensity could be smaller than expected because of the optical density of the sample-the so-called inner filter effect. However, as indicated by the absence of CH concentration-dependent changes in the free tryptophan fluorescence, the inner filter effect was negligible in our experiments. Therefore, no inner-filter correction was performed.
To estimate the binding affinity (K d ), plots of the change in the peak fluorescence intensities versus total CH concentration for the mEAG PAS domain were analyzed using the equations below, as previously described (21,29), where RL is the concentration of the free receptor-ligand complex, R t and L t are total receptor and ligand concentration, ⌬F is the peak fluorescence change, and x is a scaling factor. The data analysis and fitting of the plots was performed in Origin (Microcal Software, Inc.). Each of the experiments was repeated at least three times. The error bars on the figures correspond to the S.E.

Electrophysiology
The cDNA encoding mEAG1 channels in pGH19 vector was kindly provided by G. Robertson (University of Wisconsin-Madison, Madison, WI). The mutant mEAG1 channel with the PAS domain deletion (⌬2-173) in pGH19 was generated by Bio Basic Inc. (Canada) and verified by sequencing (Genewiz). The cRNA was transcribed using the T7 mMessage mMachine kit (Thermo Fisher Scientific). Defolliculated X. laevis oocytes were purchased from Ecocyte Bioscience (Austin, TX) and injected with the cRNA using a Nanoinject II oocyte injector (Drummond).
For current recordings in the inside-out patch configuration (44), following a manual removal of the vitelline membrane oocytes were transferred to a handmade chamber containing bath solution for current recording. Currents were recorded with Axopatch 200A patch-clamp amplifier (Molecular Devices) and pClamp10 software (Molecular Devices). The signals were digitized using Digidata 1550 (Molecular Devices). Patch pipettes were pulled from borosilicate glass and had resistances of 0.5-1.2 M⍀ after fire polishing. The intracellular (bath) and extracellular (pipette) solutions contained 130 mM KCl, 10 mM HEPES, 0.2 mM EDTA, pH 7.2.

Mechanism of EAG channel inhibition by chlorpromazine
The EAG currents were elicited by applying a series of 0.1-s voltage pulses (ranging from Ϫ100 to ϩ70 mV in 10-mV increments) from a holding potential of Ϫ80 mV, followed by a 0.15-s voltage pulse to Ϫ100 mV. The currents were not leak-subtracted.
To analyze voltage dependence of the tail currents, peak tail-current amplitudes recorded in the absence or presence of CH were normalized to the largest peak tail-current amplitude recorded in the absence of CH for the given membrane patch (G max ). These normalized data were then plotted against the test voltage and were fit with a Boltzmann equation, where V represents the test voltage (mV), V1 ⁄ 2 is the midpoint activation voltage (mV), and s is the slope of the relation (mV).
To determine the deactivation time constant, tail currents were fit with a single-exponential function. For currents recorded with TEVC technique oocytes were placed into a RC-3Z chamber (Warner Instruments). The currents were recorded with OC-725C amplifier (Warner Instruments) and pClamp11 software (Molecular Devices). The signals were digitized using Digidata 1550 (Molecular Devices). Patch pipettes were pulled from borosilicate glass and had resistances of 0.7-1.5 M⍀ after fire polishing. The recording (bath) solution contained 96 mM NaCl, 4 mM KCl, 0.1 mM CaCl 2 , 1.8 mM MgCl 2 , and 5 mM HEPES, pH 7.5. Pipette solution contained 3 M KCl. The EAG currents were elicited using the same protocol as for the inside-out patches.
To analyze voltage dependence of the steady-state currents recorded with TEVC, the conductance (G) was calculated as G ϭ I ss /(V Ϫ V rev ), where I ss is the steady-state current recorded at the end of the 0.1-s voltage pulses, V is the test voltage, and V rev is the membrane reversal potential for K ϩ selective channels. For our experiments V rev was Ϫ83.9 mV, calculated based on the bath concentration of K ϩ of 4 mM and the intracellular K ϩ concentration of 109.5 mM (45). The conductance in the absence or presence of CH was normalized to the largest conductance in the absence of CH for the given oocyte (G max ). The normalized conductance was then plotted against the test voltage, and the plots were fit to the Boltzmann equation (Equation 3).
CH was purchased from Alfa Aesar. CH stock was prepared in high purity water and then diluted with the bath solution to obtain the range of concentrations used for the dose-response experiments. The bath solution was changed using a gravity-fed solution changer for both patch-clamp and TEVC experiments. To determine the IC 50 , the concentration of the compound at half-maximal current inhibition, for both excised patch-clamp and TEVC experiments, the plots of the percentage of current inhibition versus the concentration of CH were fitted with a Hill equation, where Y 0 represents the minimum percentage of inhibition, Y ∞ is the maximum percentage of inhibition, and n is the Hill coefficient.
The data analysis and fitting of the plots were performed in Clampfit (Molecular Devices) and Origin (Microcal Software, Inc.). The error bars on the figures correspond to the S.E. Statistical analysis was performed using Student's t tests. p values Ͻ 0.05 were considered significant. n represents the number of recordings from different oocytes.