The Sensorless Pore Module of Voltage-gated K+ Channel Family 7 Embodies the Target Site for the Anticonvulsant Retigabine*

KCNQ (voltage-gated K+ channel family 7 (Kv7)) channels control cellular excitability and underlie the K+ current sensitive to muscarinic receptor signaling (the M current) in sympathetic neurons. Here we show that the novel anti-epileptic drug retigabine (RTG) modulates channel function of pore-only modules (PMs) of the human Kv7.2 and Kv7.3 homomeric channels and of Kv7.2/3 heteromeric channels by prolonging the residence time in the open state. In addition, the Kv7 channel PMs are shown to recapitulate the single-channel permeation and pharmacological specificity characteristics of the corresponding full-length proteins in their native cellular context. A mutation (W265L) in the reconstituted Kv7.3 PM renders the channel insensitive to RTG and favors the conductive conformation of the PM, in agreement to what is observed when the Kv7.3 mutant is heterologously expressed. On the basis of the new findings and homology models of the closed and open conformations of the Kv7.3 PM, we propose a structural mechanism for the gating of the Kv7.3 PM and for the site of action of RTG as a Kv7.2/Kv7.3 K+ current activator. The results validate the modular design of human Kv channels and highlight the PM as a high-fidelity target for drug screening of Kv channels.

We selected the PM of the Kv7 (KCNQ) channel family because the amino acid sequence of its pore and voltage sensor modules is similar to that of KvLm, a bacterial Kv for which we solved the structure of the pore-only module in a lipid environment at atomic resolution (22). We exploit the structural independence of the pore and sensor modules of Kvs (22)(23)(24) to produce homomers (6, 7, 13-15, 20, 21, 25-27) and heteromers (6, 7, 13-15, 20, 25-27) of Kv7.2 and Kv7.3 PMs by in vitro transcription and translation. Single-channel activity of the purified pore-only module (PM) was characterized after reconstitution in lipid bilayers. Single-channel currents through purified Kv7.2, Kv7.3, and Kv7.2/Kv7.3 PMs are K ϩ -selective and blocked by specific Kv7 channel blockers (28). Interestingly, RTG potentiates channel function by increasing single channel open probability in all functional assemblies of Kv7 PMs, validating that the site of action of RTG is embodied in the sensorless PM. Furthermore, a key mutation in Kv7.3 PM W265L renders the channel insensitive to RTG, further confirming the PM of Kv7 channels as the interaction site for the drug.

Experimental Procedures
Gene Synthesis of Kv7.2, Kv7.3 PM WT, and Mutants-The DNA sequences of the human Kv7.2 (gi 26051260) and Kv7.3 (gi 5921785) PMs were synthesized codon-optimized for Escherichia coli expression with a C terminus His 6 tag by GENEWIZ and cloned into the pT7-SC1 vector (29). The amino acid sequences of the Kv7.2 and Kv7.3 PMs extended from Met-211 to Arg-333 and from Asp-241 to Lys-373. Mutants were synthesized using the QuikChange site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA) according to the instructions of the manufacturer.
Protein Expression and Purification-Cell-free (in vitro transcription/translation) expression of the PMs was accomplished using the E. coli T7 S30 extract system (Promega, Madison, WI) according to the instructions of the manufacturer but supplemented at the reaction surface with liposomes (10% molar 1,2diphytanoyl-sn-glycero-3-phospho-(1Ј-rac-glycerol) plus 90% molar 1,2-diphytanoyl-sn-glycero-3-phosphocholine) in 0.2 M KCl, 10 mM HEPES (pH 7.4) to a final concentration of 2 mg/ml. All lipids used were purchased from Avanti Polar Lipids (Alabaster, AL). At the end, the reaction was incubated with DNase (1 g/ml final) for 10 min at room temperature, followed by incubation in 1 mM dodecyl maltoside (Anatrace, Maumee, OH) for 30 min at room temperature. The reaction was then subjected to centrifugation at 1000 ϫ g, and the supernatant was incubated with His-60 nickel magnetic beads (Clontech, Mountain View, CA) for 30 min. Following a wash with 0.2 M KCl, 10 mM imidazole, 10 mM HEPES (pH 7.4), and 1 mM dodecyl maltoside, proteoliposomes were eluted with 0.2 M KCl, 0.3 M imidazole, 10 mM HEPES (pH 7.4), and 1 mM dodecyl maltoside. Purification of the His-tagged PMs was verified by Western blotting utilizing a primary rabbit anti-penta-His BSA-free antibody (Qiagen, Valencia, CA, catalog no. 34660) and a secondary goat anti-mouse, Alexa Fluor 680-conjugated fluorescent antibody (Life Technologies, catalog no. A21058) (Fig. 1A).
Single-channel Recordings Using Droplet Lipid Bilayers-Liposomes were prepared as described previously (22,23,30,31). For reconstitution, the protein was diluted ϳ100to 200fold into preformed liposomes in 0.5 M KCl, 10 mM HEPES (pH 7.4). The experiments were performed in 90% 1,2-diphytanoylsn-glycero-3-phosphocholine supplemented with 10% 1,2-dioleoyl-sn-glycero-3-phosphate symmetric bilayers in 0.5 M KCl to increase the signal-to-noise ratio. Single-channel currents were recorded from droplet lipid bilayers as described previously (22, 23, 30 -32), with proteoliposomes always added to the cis aqueous compartment. The electrode in the cis compartment was connected to the grounded end of the amplifier head stage. The electrode in the trans compartment was connected to the working end of the amplifier. In the indicated experiments, NaCl was injected to a final concentration of 0.2 M, whereas amitriptyline (Sigma-Aldrich, Carlsbad, CA) and linopirdine (1,3-dihydro-1-phenyl-3,3-bis(4-pyridinylmethyl)-2Hindol-2-one) (Sigma-Aldrich) were injected to a final concentration of 10 and 50 M respectively, using a Nano injector (WPI, Sarasota, FL). Linopirdine and amitriptyline stocks of 1 mM were made in 200 mM KCl, 10 mM HEPES (pH 7.4). The known sided-block by linopirdine was used to establish the channel orientation in lipid bilayers; therefore, the extracellular side of the channels was assigned to the cis droplet. Only records in which this channel orientation was observed were analyzed. A retigabine (LGM Pharma, Beit Shemesh, Israel) stock of 3 mM was made in dimethyl sulfoxide and diluted further to 10 M in 10% 1,2-dioleoyl-sn-glycero-3-phosphate containing liposomes. The liposomes containing RTG were extruded through 100-m filter (Avanti Polar Lipids). The protein was added to the RTG-containing liposomes and incubated for 2-5 h prior to the single-channel recordings.
Single-channel Acquisition and Analysis-Single-channel acquisition and analysis were performed as described previously (22, 23, 30 -32) with slight modifications. Segments of continuous recordings in the range of 50 s Յ t Յ 500 s were used for analysis. The currents were sampled at 20 kHz and filtered to 2 kHz. Additional offline filtering of 1 kHz was applied to all recordings. To avoid the detection of erroneous events, the receiver's dead time (t d ) was set at 300 s. All current recordings were measured at V ϭ ϩ100 mV unless indicated otherwise. The calculated values are reported as mean Ϯ S.D. Statistical

Results
The Sensorless PM of Kv7.3 Forms K ϩ -selective Channels after Reconstitution in Lipid Bilayers-Kv7.3 PM was expressed by coupled in vitro transcription and translation and purified by a nickel affinity column (see "Experimental Procedures"). To verify the molecular weight of the synthesized and purified sample, a small volume was loaded on a 4 -20% Mini-PROTEAN TGX SDS-PAGE gel (Bio-Rad) and visualized by Western blotting (Fig. 1A). A distinct band with an apparent M r of ϳ15 was observed in the eluate, in agreement with the calculated mass of the Kv7.3 PM of 16.3 kDa. We adopted the purified eluted sample for subsequent protein reconstitution in droplet lipid bilayers. A family of single-channel currents of the Kv7.3 PM recorded at the indicated depolarizing voltages in symmetrical 0.5 M KCl is illustrated (Fig. 1, B-E). The PM opens frequently, and the pattern of activity is characterized by bursts of openings interrupted by quiescent periods. The single-channel current increases with voltage in an ohmic manner with a slope conductance (␥) of 27 Ϯ 3 pS. The corresponding normalized allpoint histograms (Fig. 1, B-E, right panels) show that the open probability of the channel (P o ) increases from 0.30 Ϯ 0.01 at 30 mV to 0.50 Ϯ 0.01 at 150 mV. Therefore, in the absence of the sensor, the Kv7.3 PM retains only vestigial voltage-gating, as observed previously in the PM of KvLm (24). The Kv7.3 PM also retains the classical K ϩ selectivity over Na ϩ . In symmetric 0.5 M KCl solutions, ␥ ϭ 28 Ϯ 3 pS at ϩ50 mV ( Fig. 2A). Injection of 0.2 M NaCl to the trans aqueous compartment reduces ␥ to 10 Ϯ 2 pS, with the openings featuring a brief mean open time ( open ) (Fig. 2B).
In a separate set of experiments (Fig. 2C), under bi-ionic conditions (0.5 M KCl in the cis compartment and 0.5 M NaCl in the trans compartment), no current was detected at ϩ50 mV when Na ϩ was the expected current carrying species, whereas channel activity was evoked immediately at Ϫ50 mV, conditions under which K ϩ is the current carrying ion. At the end of the recording, the voltage was switched to ϩ50 mV, and no current was detected, as anticipated for a K ϩ -selective channel (Fig.  2C). To determine whether the Kv7.3 PM retains the permeation properties under physiological ionic conditions, channel activity was recorded in 0.15 M KCl (Fig. 2D). The ␥ of 9 Ϯ 2 pS at ϩ100 mV recapitulates the previously measured ␥ for heterologously expressed Kv7.3 recorded in 0.15 M KCl (34), suggesting that deletion of voltage sensors had little effect on the permeation properties of the Kv7.3 PM channel and that high salt conditions (0.5 M KCl) are an appropriate means to enhance the signal-to-noise ratio for the purpose of single-channel analysis.
To ascertain that the Kv7.3 PM exhibits the pharmacological specificity of the full-length channel, we explored the activity of two known blockers/modulators. To determine the orientation of the channel in the lipid bilayer, we utilized the Kv7 channel blocker linopirdine, known to block Kv7 channels expressed in cells when applied from the extracellular side but not from the cytosol (28). Injection of 50 M linopirdine into the trans compartment (Fig. 2E, center panel) has no discernible effect on ␥ and P o of the reconstituted PM. In sharp contrast, injection of 50 M linopirdine into the cis compartment during the same experiment blocks the single-channel currents (Fig. 2E, right  panel). This indicates that the Kv7.3 PM is oriented in the lipid bilayer predominantly with the intracellular side facing the trans compartment, whereas the extracellular side faces the cis compartment. Therefore, at positive potentials, the physiologically relevant outward K ϩ flux is assayed. The result that linopirdine blocks only when applied to the compartment corre-  sponding to the extracellular side of the channels establishes that the Kv7.3 PM retains the linopirdine sensitivity of the authentic full-length channel when expressed in a cellular context.
Next we assayed the effect of the widely prescribed tricyclic antidepressant amitriptyline, a Kv7.3 channel blocker shown to act from the extracellular side (35), on the single channel currents of the Kv7.3 PM. Injection of 10 M amitriptyline in the cis compartment (Fig. 2, F and G), equivalent to the extracellular aqueous compartment in cells, decreased single-channel currents to an undetectable amplitude, confirming amitriptyline as a blocker of the Kv7.3 PM and validating the pharmacological integrity of the Kv7.3 PM.
Retigabine Stabilizes the Conductive Conformation of the Reconstituted Kv7 Pore Module-To determine the functional intactness of the Kv7.3 PM, we used RTG as a selective probe, given its well characterized activity as Kv7.3 channel opener in cellular assays (15,20). Specifically, we asked whether the interaction between RTG and the Kv7.3 PM suffices to recapitulate the RTG action on the full-length Kv7.3 channel expressed in cells. Accordingly, the Kv7.3 PM was incubated with 10 M RTG prior to the single-channel recordings. As shown in Figs. 3 and 4, RTG promotes prolonged bursts of discrete channel activity with a 3-fold increased longer burst length, 1.4-fold increased P o , 7-fold increased open , 2-fold increase in 1closed , and 3-fold increase in 2closed (Figs. 3B and 4). These results define the activity of RTG in the Kv7.3 PM as a stabilizer of the conductive conformation of Kv7.3 PM that emulates the RTG sensitivity of the full-length Kv7.3 channel (15,20). The implication is that the sensorless PM is sufficient to confer RTG sensitivity to the authentic Kv7.3 channel.
Given that Kv7.2/3 heteromers are thought to predominantly determine the neuronal M current, we next characterized the single-channel properties of Kv7.2 PM homomers and Kv7.2/3 PM heteromers in the absence and presence of RTG (Figs. 3, C-F, and 4). Kv7.2 PM homomers exhibit a ␥ of 9 Ϯ 1 pS, whereas Kv7.2/3 PM heteromers exhibit a ␥ of 15 Ϯ 3 pS in symmetric 0.5 M KCl at ϩ50 mV (Fig. 3, C and E). Both assemblies also have distinct gating patterns. P o and burst length for the Kv7.2/3 PM heteromer increased ϳ2.3-fold and ϳ4.3-fold in comparison with Kv7.2 PM homomers (P o ϭ 0.24 Ϯ 0.03 and burst length ϭ 73 Ϯ 8 ms, Fig. 4, A and B). Notably, a single population of conductance and gating parameters was observed in records of Kv7.2/3 PM heteromer assemblies, suggesting that Kv7.2 and Kv7.3 subunits oligomerize with a predominantly unique stoichiometry when coexpressed in a cellfree in vitro transcription and translation system to generate functional assemblies with gating and permeation features different from either Kv7.2 or Kv7.3 homomers. In additional semblance to the reported features of the heteromer expressed in cells (6, 7, 13-15, 20, 25-27), the Kv7.2/3 heteromeric PM (Fig. 3E) displays a statistically significant longer burst length than either homomeric assembly (Figs. 3, A and C, and 4, A and  B), implying that co-assembly of the Kv7.2 and Kv7.3 subunits generates a more stable conductive conformation of the PM (Fig. 4).  (Figs. 3, C and D, and 4, A-C). In comparison, when the Kv7.2/3 PM heteromers were incubated with RTG, the P o increased 1.5-fold, open increased 4.7-fold, and burst length increased 3-fold (Figs. 3, E and F, and 4). The implication is that RTG stabilizes the open conformation of both Kv7.2 PM and Kv7.2/3 PM channels, as observed for the native channels in the cellular context (13,14,20,21,36).
How Does RTG Modulate the PM Conformation?-A tryptophan at position 265 of Kv7.3, conserved in Kv7.2 (Trp-236) (21) to Kv7.5 but absent in Kv7.1, was identified as a necessary constituent of the RTG binding site (20). Accordingly, a mutant Kv7.3, W265L, was studied, and, as shown in Fig. 3, G and H, it is practically insensitive to RTG given that, in its absence, the characteristic ␥, P o , burst length, open , 1closed , and 2closed are all equal, within error, to those measured after incubation with 10 M RTG (Figs. 3, G and H, and 4). Importantly, the P o of the mutant is higher than that of the Kv7.3 WT and equal, within error, to that observed for the Kv7.3 WT in the presence of RTG (Fig. 4A). To further examine this, we turned to homology modeling, using as templates the open pore conformation of the Kv1.2-2.1 channel (PDB code 2R9R) (37), the closed structure of KvLm PM (PDB code 4H33) (22), and in silico docking in YASARA (YASARA Biosciences). The solvent-exposed surface depiction of the homology model on the basis of the KvLm closed channel structure shows that the Kv7.3 channel is closed (Fig. 5A). In this closed conformation model, Trp-265 is shown to form two distinct intrasubunit -( stacking) interactions: one with Phe-269 within the same transmembrane helix (S5 following the convention of full-length channels) and the other with Phe-343 in the sixth transmembrane segment (S6 in fulllength channels) (Fig. 5B). In contrast, a solvent-accessible surface representation of the Kv7.3 homology model utilizing the PDB code 2R9R open channel template shows that the channel is open (Fig. 5C). In this open model conformation, the indole ring of Trp-265 in a Kv7.3 subunit forms twointrasubunit interactions with Tyr-266 and Phe-269 within the same transmembrane segment, S5 (Fig. 5D). Therefore, the models suggest that, in transitioning from closed to open, ainteraction that holds S5 to S6 within the same subunit is lost in the Kv7.3 PM and replaced by ainteraction within S5.
The described Kv7.3 PM gating model proposes the testable hypothesis that theinteraction between Trp-265 in S5 and Phe-343 in S6 is a major determinant of the stability of the closed conformation of the PM. In experimental agreement, the P o of the reconstituted Kv7.3 W265L mutant PM is observed to be larger than the WT. When expressed heterologously in cells, the Kv7.3 mutant W265L has been shown to have a left-shifted voltage of activation relative to the wild type (20), demonstrating a higher probability of finding the channel open at lower potentials and justifying the proposed role for Trp-265 in determining closed structure stability. To understand how a presumed RTG interaction with Trp-265 in the reconstituted PM and also in cells destabilizes the closed conformation of the Kv7.3 PM (13-15), we docked RTG on the open model structure. The two lowest-energy docking models in which RTG interacts with Trp-265 suggest that the aromatic ring in RTG forms ainteraction with Trp-265 (Fig. 5, E and F). The model presented here differs from an earlier model proposed by Lange et al. (38), particularly in the orientation of RTG within the same binding pocket. In their model, RTG is not positioned to form ainteraction with Trp-265. Because, in all our calculated models, RTG fails to form a strong interaction with Phe-343 in S6, we propose that RTG stabilizes Kv7.3 PM by preventing Trp-265 from interacting with Phe-343 to stabilize the closed pore structure mediated through ainteraction. This hypothesis awaits further experimental validation.

Discussion
RTG, an anticonvulsant drug currently prescribed for the treatment of epilepsy, has been shown to prolong the open conformation of neuronal Kv7.2/3 channels (13)(14)(15) and its effect demonstrated to depend on the presence of Trp-265 in Kv7.3 (20) and/or Trp-236 in Kv7.2 (21). However, the mechanism by which RTG produces its effect has not yet been outlined at the molecular level. Here we show that RTG increases the open probability of the Kv7.2 PM, Kv.7.3 PM, and Kv7.2/7.3 heteromeric PMs in the absence of the voltage sensor module, suggesting that its sufficient target site resides exclusively on the PM. But how does an interaction between RTG and the pore module subunits of Kv7.3 or Kv7.2 produce this effect? Homology modeling of the closed structure of Kv7.3, using as a template the only available closed structure of a Kv channel with significant homology to Kv7.3 (22), allows us to propose a mechanism. In the modeled closed conformation of Kv7.3, we identify ainteraction between Trp-265 and Phe-343 within the same subunit. We propose that destabilizing suchinteraction either by mutations or by RTG favors the open channel conformation. These findings outline a new path forward to design and test drugs that modulate the function of Kv7 channels and the ensued signaling networks. E and F, two low-energy docking models of RTG onto the open channel module (side views), both involving ainteraction between Trp-265 and RTG, with amino acids forming the pocket (distance between protein and RTG, Ͻ5 Å) labeled explicitly.