Voltage-dependent activation in EAG channels follows a ligand-receptor rather than a mechanical-lever mechanism

Ether-a-go-go family (EAG) channels play a major role in many physiological processes in humans, including cardiac repolarization and cell proliferation. Cryo-EM structures of two of them, KV10.1 and human ether-a-go-go-related gene (hERG or KV11.1), have revealed an original nondomain-swapped structure, suggesting that the mechanism of voltage-dependent gating of these two channels is quite different from the classical mechanical-lever model. Molecular aspects of hERG voltage-gating have been extensively studied, indicating that the S4-S5 linker (S4-S5L) acts as a ligand binding to the S6 gate (S6 C-terminal part, S6T) and stabilizes it in a closed state. Moreover, the N-terminal extremity of the channel, called N-Cap, has been suggested to interact with S4-S5L to modulate channel voltage-dependent gating, as N-Cap deletion drastically accelerates hERG channel deactivation. In this study, using COS-7 cells, site-directed mutagenesis, electrophysiological measurements, and immunofluorescence confocal microscopy, we addressed whether these two major mechanisms of voltage-dependent gating are conserved in KV10.2 channels. Using cysteine bridges and S4-S5L–mimicking peptides, we show that the ligand/receptor model is conserved in KV10.2, suggesting that this model is a hallmark of EAG channels. Truncation of the N-Cap domain, Per-Arnt-Sim (PAS) domain, or both in KV10.2 abolished the current and altered channel trafficking to the membrane, unlike for the hERG channel in which N-Cap and PAS domain truncations mainly affected channel deactivation. Our results suggest that EAG channels function via a conserved ligand/receptor model of voltage gating, but that the N-Cap and PAS domains have different roles in these channels.

opens (Fig. 1B, middle). Upon repolarization S4-S5 L binds to the S6 T gate and channel deactivates (back to Fig. 1B, left). Here, using two distinct approaches, we observe that this ligand/receptor mechanism, which we originally proposed for hERG, is conserved in K V 10.2 (15).
Second, hERG deactivation is modulated by the channel N terminus. In the N-terminal eag domain, deletion in the N-Cap and/or the PAS domains, profoundly accelerated deactivation with no major effect on maximal current amplitude (16 -19). Moreover, covalent binding of the N-Cap and the S4-S5 L closed the channel (20). These observations suggested that the N-Cap and/or the PAS domains regulate channel deactivation by modulating S4-S5 L and S6 T interaction. In K V 10.2, we show that the eag domain presents quite different functions. Deletions of the N-Cap and PAS domains, separately or altogether, completely abolish channel activity, at least partially due to a defect in membrane trafficking.
In conclusion, we show that slowing activating channels follow an allosteric model, in which the voltage sensor and pore domains are weakly coupled, via a ligand, S4-S5 L and a receptor, S6 T . In K V 10.2, such coupling is not modulated by the eag domain, as proposed for hERG (20).

As in hERG, covalent binding of S4-S5 L to S6 T locks the K V 10.2 channel closed
The similar structures of hERG and K V 10.1 (12,13) showing a nonswapped arrangement of the pore and voltage sensor domains suggest that voltage-gating of these channels does not follow the classical mechanical-lever model in which S4-S5 L constricts the S6 T gate (21). Also, functional evidence obtained by several groups including ours, strongly suggests that hERG channels do not follow the mechanical-lever model (14,15). First, Ferrer and collaborators, observed that introduction of a cysteine in S4-S5 L (D540C) and another cysteine in S6 T (L666C) locks the channel in a closed state in oxidative conditions. This was associated to a restricted movement of the voltage sensor in a oxidative condition, as measured by gating currents, suggesting the formation of a disulfide bridge (14). Also, mutagenesis experiments on hERG suggested electrostatic interaction between Asp-540 and Leu-666 playing a major role in hERG voltage-dependent gating (22). Such observations suggest that in the WT hERG channel, specific interactions between S4-S5 L and the activation gate (S6 T ) stabilize the closed channel. In transfected COS-7 cells, we could reproduce the experiments originally done by Ferrer and collaborators (14) in the Xenopus oocyte model. Moreover we could observe that a S4-S5 L mimicking peptide inhibits the hERG channel, suggesting a ligand/receptor model, in which S4-S5 L , directly under the control of the voltage sensor S4, binds to S6 T to lock the channel in a closed state (see Fig. 1B, right).
To check whether K V 10.2 channels follow the same voltagegating mechanism as hERG channels, we aligned the amino acid sequences of the two channels, and mutated to a cysteine Asp-339 and Met-474 in K V 10.2, corresponding to hERG positions in which cysteines were previously introduced: Asp-540 in Figure 1. Alignment used to design the cysteine mutants and the S4-S5 L peptide, hypothetical ligand/receptor model. A, alignment between hERG and K V 10.2. This alignment was used to (i) introduce the cysteines in S4-S5 L and S6 T and (ii) design the S4-S5 L peptide from the previously identified hERG S4-S5 L inhibiting peptide. S4-S5 L refers to the S4-S5 region, including the S4-S5 linker and a part of S5. S6 T refers to the C-terminal part of the S6 segment, which is the activation gate. The alignment was obtained using Cobalt (43). In red are represented the basic residues, in yellow acidic residues, and in purple the position of the narrowest part of the bundle crossing, also the gating residue (44). The color boxes represent the transmembrane segments. Introduced cysteines in hERG (15) and K V 10.2 in the present study are indicated. Gray line represents the inhibiting S4-S5 L peptide in hERG (15). Red line represents the peptide engineered in K V 10.2 for the present study. B, left: scheme of the hypothetical ligand/receptor model in which S4-S5 L (deep blue) binds to S6 T (light blue) to stabilize the channel in a closed state. Middle, upon membrane depolarization, S4 pulls S4-S5 L out of the S6 T receptor, allowing channel opening. Right, the K V 10.2 S4-S5 L peptide (red) mimics endogenous S4-S5 L , locking the channel in its closed conformation.

Voltage-dependent gating mechanism of EAG channels
S4-S5 L and Leu-666 in S6 T (14,15) (black arrows in Fig. 1A). As for the hERG double cysteine mutant, a 2-h application of 0.2 mM tert-butylhydroperoxide (tbHO 2 ) led to an almost complete inhibition of the D339C/M474C K V 10.2 channel current (Fig. 2, A and B). Knowing the high homology between hERG and K V 10.2 (65%), we supposed that D339C and M474C in K V 10.2 also form a disulfide bridge in oxidative conditions, as a cause of the drastic current reduction. As control experiments, 0.2 mM tbHO 2 application had no effect on tail current amplitude when only one (D339C or M474C K V 10.2) or none (WT K V 10.2) of the two cysteines was introduced (Fig. 2, C-H) and activation curves showed a similar shift in all conditions (Fig.  S1). Because a 2-h tbHO 2 application represents a slow time course for a putative disulfide bridge formation, we also applied a higher concentration of tbHO 2 (2 mM) for a shorter time. As in the Ferrer study (14) on hERG, this concentration led, in around 12 min, to an almost complete inhibition of the D339C/ M474C K V 10.2 channel current but not the WT channel current (Fig. 3). This inhibition was reversed by 10 mM DTT, as in the Ferrer study on hERG (14).
To strengthen the hypothesis of the interaction between S4-S5 L and S6 T in K V 10.2, we also tested the tbHO 2 effect on another double mutant of the K V 10.2 channel. We chose the E343C/M474C K V 10.2 channel, because Glu-343, in S4-S5 L , is aligned with Glu-544 in hERG, which when mutated to a cysteine together with L666C in hERG S6 T (Fig. 1A), also lead to channel inhibition, suggesting the formation of a disulfide bridge (15). We observed an inhibition of the current after 15 min incubation of 2 mM tbHO 2 , which is reversed by 10 mM DTT (Fig. 4, A and B). Control experiments showed no effect either on tail current amplitude or on the activation curve when only one (E343C or M474C K V 10.2) or none (WT K V 10.2) of the two cysteines was introduced (Fig. 4, C-H, Fig. S2). Altogether, these observations confirmed that two different disulfide bridges created between the two introduced cysteines in S4-S5 L and S6 T locks K V 10.2 in a closed state, as in hERG (14,15).

As in hERG, a K V 10.2 S4-S5 L mimicking peptide inhibits K V 10.2 channels
Recently, we observed that (i) co-expressing a hERG S4-S5 L mimicking peptide with hERG channel partially inhibited the generated current and (ii) covalently binding this peptide to the hERG channel completely inhibited it (15). These observations suggest that in hERG channels, S4-S5 L rather acts as a voltagecontrolled ligand that binds to the S6 T gate and stabilizes it in the closed state (Fig. 1B, left, for the endogenous interaction and B, right, for the mimicking peptide). This mechanism of gating is consistent with Ferrer et al. (14) observation that covalent binding of S4-S5 L and S6 T channel regions locks the channel in a closed state. Using the alignment between hERG and K V 10.2 (Fig. 1A), we designed the K V 10.2 S4-S5 L peptide from the position of the S4-S5 L peptide sequence that was inhibiting the hERG channel (15). Co-expressing the K V 10.2 channel with its specific S4-S5 L peptide led to a profound decrease in current density (more than 70%), with no shift in the activation curve, as observed for hERG (Fig. 5, A-E). As an additional control, coexpressing K V 10.2 with a scramble S4-S5 L peptide led to cur-rent density similar to the one observed in the absence of peptide (Fig. 5, A and B).
Because the S4-S5 L peptide is supposed to interfere with the gating, it may be intriguing that activating and tail currents are inhibited to the same extent, and that neither the activation curve, nor the activation kinetics are modified (Fig. 5, C-F). To address this issue, we used a kinetic model mimicking channel activity in the presence/absence of the peptide (Fig. 6), based on a previous model of KCNE1-KCNQ1 and S4-S5 L peptide interaction (23). In the KCNE1-KCNQ1 model, the presence of the peptides was affecting the channel activation curve and activation kinetics. However, when constrained to K V 10.2 kinetics, the model did not show any alteration of activation kinetics or steady-state activation (Fig. 6, B and C). This is because peptide binding and unbinding rates (0.02 and 0.04 s Ϫ1 , respectively) are lower than channel opening/closing rates, because increasing these rates leads to alterations in channel activation kinetics and steady-state activation (Fig. 6, B and C). A peptide with similar binding/unbinding rates impact KCNE1-KCNQ1 channel activation curve and activation kinetics (23) because gating kinetics of this channel are slower than K V 10.2 kinetics. Altogether, these observations support the ligand/receptor model of voltage dependence in K V 10 channels.

As opposed to hERG, N-Cap and PAS domain-deleted K V 10.2 channels are not functional
The results described above suggest that in both K V 10.2 and hERG channels, deactivation is due to S4-S5 L binding to S6 T and consequent stabilization of the closed state. It has been shown by several works that intracellular N-cap and PAS domains of hERG (shown in Fig. 7A for K V 10.2, cf. alignment in Fig. S3) modulate channel deactivation kinetics. Most importantly truncated hERG channels missing N-Cap or the whole eag domain (N-Cap ϩ PAS), when expressed in Xenopus oocytes, showed robust currents but a more than 5-fold acceleration in deactivation (16 -19). Also in mammalian cells, it has been observed that the eag domain is not necessary for hERG channel trafficking, consistent with the observation of robust currents in this model (24,25). Another work showed that N-Cap is close to S4-S5 L (20). Thus, N-Cap may modulate channel deactivation through a direct interaction with this linker. Based on all these observations, we supposed that deletion of both N-Cap and PAS domains in K V 10.2 should give rise to functional channels with accelerated deactivation, as in the study on hERG in mammalian cells (25). Transfection of WT K V 10.2 tagged with 1D4 at the C terminus gave rise to a current similar to a previous description (6) and immunofluorescence experiments using this tag showed plasma membrane enrichment of the channel, compared with the intracellular compartment (Fig. 7, B-F). Surprisingly, deletion of N-Cap domain, PAS domain, but also of both domains all resulted in nonfunctional K V 10.2 channels (Figs. 8B, 9B, and 10B). Immunofluorescence experiments on N-Cap and/or PAS domains truncated channels showed no membrane enrichment of the channels (Figs. 8, C-F, 9, C-F, and 10, C-F). These findings Voltage-dependent gating mechanism of EAG channels suggest a trafficking defect of the N-Cap and/or PAS domains truncated Kv10.2 channels.

As opposed to hERG, coexpressing a K V 10.2 N-Cap mimicking peptide with truncated K V 10.2 does not counteract the effect of channel truncation
For the hERG channel, it has been shown that a peptide corresponding to the first 16 amino acids of the channel is sufficient to reconstitute slow deactivation to hERG lacking this region (19). Similarly, another study has shown that injection of the purified eag domain, corresponding to the first 135 amino acids of hERG, into oocytes expressing eag-truncated hERG, restores the deactivation kinetics to WT-like in more than 24 h (17). Based on these previous observations on hERG, we proposed that co-expression of specific K V 10.2 N-Cap mimicking peptides with the N-Captruncated K V 10.2 channel should recover its expression and activity at the plasma membrane. Surprisingly again, K V 10.2 channel activity was not recovered in the presence of N-Cap mimicking peptide (n ϭ 8). This observation further suggests that N-Cap and PAS domains play distinct roles in hERG and K V 10.2 function.

Co-expression of WT and truncated K V 10.2 channels gives rise to a right shift in the activation curve as compared with homomeric WT channels, but no change in deactivation kinetics
To evaluate the potential effects of N-Cap truncation on channel activity, we co-expressed K V 10.2 missing the N-Cap with the WT channel, in an attempt to generate heteromers. We observed robust voltage-dependent currents, showing a ϳ30-mV shift in the activation curve toward depolarized potential, demonstrating the generation of such heteromers (Fig. 11, C-F). This shift in the activation curve suggests that N-Cap deletion leads not only to a K V 10.2 trafficking defect but also to a gating defect. In the hERG channel, deletion of N-Cap did not lead to a shift of the activation curve, but an acceleration of deactivation (26). In the present experiments on K V 10.2, no change in deactivation kinetics was observed when the N-Captruncated channel was co-expressed with the WT channel (Fig.  12). We also co-expressed the K V 10.2 channel construct lacking both the N-Cap and PAS domains, with the WT channel. Again, we observed a ϳ30-mV shift in the activation curve toward depolarized potentials, but no modification in deactivation (Figs. 11 and 12). Thus, although we are likely recording the  From a holding potential of Ϫ100 mV, followed by a 3-s prepulse at Ϫ40 mV, tail currents were recorded at ϩ80 mV, every 8 s. Following stabilization of the tail current, 2 mM tbHO 2 was perfused (gray arrow), and the step protocol was repeated for 6 min. Following the tbHO 2 application, a fraction of the cells was then perfused with 10 mM DTT, and the step protocol was continued for an additional 6 min. Each data point represents the mean Ϯ S.E. current magnitude normalized to values obtained before tbHO 2 , n ϭ 16 (WT), 19 (D339C/M474C in tbHO 2 ), and 7 (D339C/M474C in DTT). Insets (a-d) correspond to representative recordings at the arrows.
Voltage-dependent gating mechanism of EAG channels combined activities of tetrameric channels containing different ratios of the WT and truncated subunits, it appears that N-terminal deletion of K V 10.2 impacts the steady-state activation curve rather than deactivation kinetics.

Discussion
From the present and previous works, we suggest that among voltage-gated channels, coupling between voltage sensor movement and pore gating falls into two categories: (i) the mechanical-lever model: an obligatory coupling in which the S4 resting state directly translates into S6 gate-closed state. This mechanical-lever model, inferred from structural data in Shaker-like channels (27), also applies to eukaryotic sodium channels, as suggested by recent structural studies (28,29); (ii) the ligand/receptor model: the obligatory coupling cannot hold if the S6 T gate is able to open, even if S4 segments are in the resting state, as shown for hERG and KCNQ1 channels (30 -32), and, vice versa, if the S6 T gate is able to close, even if S4 Voltage-dependent gating mechanism of EAG channels segments are in the activated state (33). We recently demonstrated this ligand/receptor model in hERG channels by using several approaches. Here, we obtained similar results on K V 10.2 using similar approaches.
First, introduction of cysteines in S4-S5 L and S6 T lock the channel in a closed state in oxidative conditions, suggesting the formation of a disulfide bridge, as in hERG. This suggests that the same gating mechanism applies to K V 10.2. Introduction of cysteines in both S4-S5 L and S6 T may lead to a nonnative conformation that favor an S4-S5 L interaction with S6 T , which would not be met in the WT channel. But in the second set of experiments, a S4-S5 L mimicking peptide, without any introduced cysteine, inhibits K V 10.2 channel, also without any introduced cysteine, further suggesting the capability of S4-S5 L to stabilize the channel closed state. Noteworthy, similar channelspecific peptides have the same effect in KCNQ1 and hERG channels (15,23). Complementary experiments in hERG revealed that S4-S5 L peptide effects was on channel gating, and not channel trafficking (15).
Altogether these experiments suggest that K V 10.2 follows the ligand/receptor mechanism observed in hERG. In both channels, S4-S5 linkers are short (12,13), thus it is likely that the part of S5 that is present in the peptide also plays a role in closed channel stabilization. Further structural data of hERG and K V 10.2 channels in the closed state should clarify the residues involved in the S4-S5 L and S6 T interaction. Noteworthy, this ligand/receptor model is consistent with the observation that the voltage-dependent closure of the related K V 10.1 channel, but also of the hERG channel, requires at least a part of the inhibiting S4-S5 L to be covalently linked to the voltage sensor S4 (34,35).
As opposed to similar ligand/receptor gating mechanisms in the two channels, the role of the eag domain (N-Cap ϩ PAS) is quite different between hERG and K V 10.2. In hERG channels, the eag domain (N-Cap ϩ PAS) is not necessary for channel activity and mainly modulates the current deactivation rate (17)(18)(19). Here, we observed that truncation of the K V 10.2 eag domain renders the channel nonfunctional, at least partly due to trafficking defects. In rescue experiments with the WT channel, we observed that eag domain deletion or even only N-Cap deletion are associated with a shift in the activation curve, but no change in channel deactivation. The contrary is observed in hERG channel: an altered deactivation, no shift in the activation curve (26, 36). Our results suggest major differences in functional roles for the N-Cap and PAS domains between K V 10.2 and hERG. Altogether, this work suggests a conserved ligand/receptor (allosteric) model of voltage gating, but divergent roles in eag domains among channels of the EAG family.
In combination with previous work on hERG, this study highlights the voltage-gated channels superfamily, divergent gating mechanisms (obligatory versus allosteric) that matches divergent structures (swapped versus nonswapped domain, respectively) and divergent kinetics (fast versus slow activating channels, respectively). Nonswapped domains (hERG, K V 10) may provide less contact between S4-S5 L and S6 T (12, 13). We

Voltage-dependent gating mechanism of EAG channels
propose that this weak coupling between S4-S5 L and S6 T provides a framework for a two-step channel activation: first, the fast S4 movement drags the ligand S4-S5 L out of its receptor on the S6 T gate, followed by slow gate opening. This allosteric regulation of the S6 T gate by S4-S5 L may explain how in slowly activating channels, movement of S4 is not concomitant to pore opening (37).

Experimental procedures
Plasmid constructs pCDNA6 hK V 10.2 was subcloned into pMT3 vector using the standard PCR overlap extension method (38). A 1D4 immunoaffinity tag (derived from the C terminus of bovine rhodopsin) was added to the C terminus of all constructs (39). The D339C, E343C, M474C, E343C/M474C, and D339C/M474C  (23). a, kinetic model in the absence of peptide, on which optimization has been performed (see "Experimental procedures"). Optimized transition rates are presented in Table 1. b, binding of exogenous S4-S5 L locks the channel and prevents its opening. Peptides are supposed to interact with each monomer in the unlocked states. B, simulated currents during step protocols (same as in Fig. 5), in the absence (Ctrl) or presence of S4-S5 L peptide , at the indicated S4-S5 L on/off rates. C, gray filled circles: experimental activation curves and half-activation times in control condition. Other symbols: simulated values in the absence of peptide (control, open circles), or in the presence of peptides, at the indicated rates of peptide binding/unbinding.

Voltage-dependent gating mechanism of EAG channels
mutations were inserted into the pMT3-K V 10.2 construct using the QuikChange TM site-directed mutagenesis-based technique using Accuprime Pfx polymerase (ThermoFisher Scientific) according to the standard protocol recommended by the manufacturer. Truncation mutants were constructed by deleting residues 2 to 24 (⌬N-CAP), 2 to 134 (⌬eag), and 25 to 134 (⌬PAS) using the standard PCR overlap extension method.
PCR products were digested with HindIII and XbaI and ligated into pCDNA6 and PMT3 vectors. All constructs were confirmed by sequencing. Oligonucleotides encoding K V 10.2 peptides were synthesized by TOP Gene Technologies and contained a XhoI restriction enzyme, followed by a methionine (ATG) for translation initiation, a glycine (GGA) to protect the ribosome-binding site during translation, and the nascent pep- Voltage-dependent gating mechanism of EAG channels tide against proteolytic degradation (40). A BamHI restriction enzyme site was synthesized at the 3Ј end immediately following the translational stop codon (TGA). These oligonucleotides were then ligated into pIRES2-EGFP (Clontech) and sequenced.

Cell culture and transfection
The African green monkey kidney-derived cell line COS-7 was obtained from the American Type Culture Collection (CRL-1651) and cultured in Dulbecco's modified Eagle's medium (GIBCO) supplemented with 10% fetal calf serum and antibiotics (100 IU/ml penicillin and 100 g/ml of streptomycin) at 5% CO 2 and 95% air, maintained at 37°C in a humidified incubator. Cells were transfected in 35-mm Petri dishes when the culture reached 50 -60% confluence, with 4 g of total DNA complexed with 12 l of FuGENE-6 (Roche Molecular Biochemical) according to the standard protocol recommended by the manufacturer. In different experiments, plas- Voltage-dependent gating mechanism of EAG channels mid quantities were optimized to keep current amplitudes in a range that undetectable currents were rare, and large currents inducing incorrect voltage-clamp were also rare. Immunofluorescence and confocal microscopy experiments were done with pCDNA6-K V 10.2 for which channel expression was lower than with pMT3-K V 10. For S4-S5 L peptide experiments, COS-7 cells were co-transfected with 2 g of pMT3-WT K V 10.2 and 2 g of pIRES2-EGFP plasmids encoding or not the S4-S5 L peptide. As an additional control, a pIRES2-EGFP plasmid encoding a scramble S4-S5 L peptide was used. In pIRES2-EGFP plasmids, the second cassette (EGFP) is less expressed than the first cassette, guaranteeing high levels of peptide expression in fluorescent cells Voltage-dependent gating mechanism of EAG channels (23).

Electrophysiology
One day after splitting, COS-7 cells were mounted on the stage of an inverted microscope and constantly perfused by a Tyrode solution (cf. below) at a rate of 1-3 ml/min. The bath Voltage-dependent gating mechanism of EAG channels temperature was maintained at 22.0 Ϯ 2.0°C. Stimulation and data recording were performed with Axon pClamp 10, an A/D converter (Digidata 1440A), and an Axopatch 200B amplifier (all Molecular Devices). Patch pipettes (tip resistance: 2-3 megohms) were pulled from soda lime glass capillaries (Kimble-Chase) and coated with wax. Currents were recorded in the whole-cell configuration, pipette capacitance and series resistance were electronically compensated (by around 75%). Activation protocols were adjusted to the voltage-dependence of the construct as in the previous study on hERG (15). Activation curves were obtained from the tail currents and fitted by Boltzmann equations.

Confocal microscopy
Immunohistological analyses were performed to study cell localization of transfected WT/truncated K V 10.2-1D4 in COS-7 cells. Twenty-four hours after transfection, cells were plated on IBIDI plates for 24 h. Cells were then fixed with 4% formaldehyde, stained for 10 min at room temperature with Alexa Fluor TM 647 conjugated wheat germ agglutinin (WGA; ThermoFisher), a plasma membrane marker, permeabilized with 0.5% saponin and blocked with 1% PBS/BSA. Cells were then incubated with a mouse mAb directed against the 1D4 tag diluted in PBS (Abcam). Secondary antibody staining was per- Voltage-dependent gating mechanism of EAG channels formed using Alexa 488-conjugated anti-mouse antibody. DAPI was used for nuclear staining. Conventional imaging was performed using a LSM710-Confocor3 (Zeiss) and a Nikon Confocal A1RSi microscope system equipped with a SR Apo 100 ϫ 1.49 N.A objective. Images were analyzed with ImageJ software. In figures, but not for analyses, Enhance Local Contrast adjustment was performed on WGA staining to highlight plasma membrane staining. To quantify fluorescence, the line plot was arbitrarily segmented in 3 different regions: the first and last 5-15% of the line plot, corresponding to plasma membrane (M1 and M2), and the remaining intermediate 70% of the line plot, corresponding to the intracellular compartment. For each of these regions in each cell, K V 10.2 fluorescence intensity values were normalized by the average cell K V 10.2 fluorescence intensity signal (41).

Kinetic model
The K V 10.2 kinetic model (Fig. 6A, top) contains two voltage sensor transitions, and two open states, as in the model of I Ks (42). This model was optimized using IChMASCOT (J.A. De Santiago-Castillo and M. Covarrubias) to fit traces of the representative control of Fig. 5. Optimized transition rates are presented in Table 1. Next, another model was designed (Fig. 6A,  bottom), with an additional state in which S4-S5 L mimicking peptide binds to the pre-open state and stabilizes it, as in Ref. 23. Various S4-S5 L binding/unbinding transition rates were applied, and the effects on the biophysical parameters were studied.  Table 1 Optimized transition rates used in the model presented in Fig. 6 The abbreviations use are: F ϭ 96485 C. mol Ϫ1 (Faraday constant); R ϭ 8.314 J. mol Ϫ1 .K Ϫ1 (gas constant); T ϭ 297 K; V (membrane potential) in V.

Voltage-dependent gating mechanism of EAG channels Statistics
All data are expressed as mean Ϯ S.E. Statistical differences between current densities (data points are not normally distributed) were determined using nonparametric Mann-Whitney test. Statistical differences between activation parameters, V 0.5 , K (data points are normally distributed) were determined using unpaired Student's t tests. A value of p Ͻ 0.05 was considered significant.