Novel Kv7.1-Phosphatidylinositol 4,5-Bisphosphate Interaction Sites Uncovered by Charge Neutralization Scanning*

Background: Phosphatidylinositol 4,5-bisphosphate (PIP2) regulates a variety of ion channels including Kv7 channels. Results: Several Kv7.1 mutations impair functional interaction with PIP2. Conclusion: PIP2 interacts with several Kv7.1 amino acids, indicating the existence of at least two binding pockets. Significance: Impaired molecular interaction between mutant Kv7.1 and PIP2 may underlie heart disease. Kv7.1 to Kv7.5 α-subunits belong to the family of voltage-gated potassium channels (Kv). Assembled with the β-subunit KCNE1, Kv7.1 conducts the slowly activating potassium current IKs, which is one of the major currents underlying repolarization of the cardiac action potential. A known regulator of Kv7 channels is the lipid phosphatidylinositol 4,5-bisphosphate (PIP2). PIP2 increases the macroscopic current amplitude by stabilizing the open conformation of 7.1/KCNE1 channels. However, knowledge about the exact nature of the interaction is incomplete. The aim of this study was the identification of the amino acids responsible for the interaction between Kv7.1 and PIP2. We generated 13 charge neutralizing point mutations at the intracellular membrane border and characterized them electrophysiologically in complex with KCNE1 under the influence of diC8-PIP2. Electrophysiological analysis of corresponding long QT syndrome mutants suggested impaired PIP2 regulation as the cause for channel dysfunction. To clarify the underlying structural mechanism of PIP2 binding, molecular dynamics simulations of Kv7.1/KCNE1 complexes containing two PIP2 molecules in each subunit at specific sites were performed. Here, we identified a subset of nine residues participating in the interaction of PIP2 and Kv7.1/KCNE1. These residues may form at least two binding pockets per subunit, leading to the stabilization of channel conformations upon PIP2 binding.

The voltage-gated potassium channel K v 7.1 (also called KCNQ1 and K v LQT1) is one of five members of the K v 7 subfamily. It is foremost expressed in the heart and epithelial tissues of many organs (1,2). Although K v 7.1 can assemble with all five members of the KCNE ␤-subunit family, its interaction with KCNE1 (also called minK or IsK) is best described (1,(3)(4)(5)(6)(7). The noninactivating K v 7.1/KCNE1 channel complex shows slow activation and deactivation kinetics (1,4,8,9). These channels mediate the slowly activating cardiac potassium current I Ks (1,3), which contributes to the repolarization of myocyte membranes during phase 3 of the cardiac action potential (10,11). Many mutations in K v 7.1-and KCNE1-encoding genes have been reported to result in loss of function and to cause long QT syndrome (LQTS), 2 which is associated with cardiac arrhythmia, syncope, and sudden cardiac death (12)(13)(14)(15)(16)(17)(18).
All K v 7 channels are inhibited after stimulation of G q -or G 11 -protein-coupled receptors (19 -22). This is assumed to be primarily due to depletion of the plasma membrane lipid phosphatidylinositol 4,5-bisphosphate (PIP 2 ) upon activation of phospholipase C (23)(24)(25). PIP 2 is important for the function of many proteins, including trafficking molecules such as small GTPases (26,27) and cytoskeletal proteins (28,29), as well as transporters and ion channels (30).
Among the ion channels regulated by PIP 2 are transient receptor potential channels, (32), inward-rectifier potassium (K ir ) channels (33), and voltage-gated potassium channels such as K v 7.1. The lipid presumably exerts its effect on the latter by stabilizing the open channel conformation, thereby slowing the deactivation kinetics and shifting the activation curve to more negative potentials (34).
To date, the interaction has been localized to the C terminus of K v 7.1, in which four amino acids have been identified to interact with PIP 2 (35). However, it is not known whether the interaction site is confined to these amino acids. The aim of the present study, therefore, was the identification of residues within K v 7.1 responsible for the interaction with PIP 2 . Our experiments were based on the assumption that direct interaction is caused by electrostatic forces between negatively charged PIP 2 head groups and positively charged amino acid residues. Hence, we generated neutralizing mutants of 13 positively charged amino acids. These were coexpressed with KCNE1 in Xenopus laevis oocytes to form slowly activating (I Ks ) channels. Effects of injected diC 8 -PIP 2 on wild type and mutant channels was investigated by means of the two-electrode voltage-clamp. Included in the charge neutralization scan were the residues identified by Thomas et al. (35). Nine of the thirteen amino acids proved positive for a functional interaction with the lipid. Five of these had been identified previously as LQTS mutants. Electrophysiological analysis revealed an impaired PIP 2 interaction with two of these. Two regions containing examined amino acids bound PIP 2 and other lipids in lipid binding assays. To gain insight into the underlying structural mechanisms, we conducted molecular dynamics simulations of K v 7.1 tetramers in complex with four KCNE1 ␤-subunits, both with and without PIP 2 molecules. The results implied that binding of PIP 2 to at least eight specific binding sites causes the stabilization of open and closed channel conformations.
PIP Strips membranes were blocked with TBS-T containing 10 mmol/liter Tris-HCl, pH 8.0, 150 mmol/liter NaCl (containing 0.1% (v/v) Tween 20 detergent) ϩ 3% fatty acid-free BSA and gently agitated for 1 h at room temperature protected from light. The peptides (50 g/ml) were each solved in TBS-T ϩ 3% fatty acid-free BSA and incubated on the membranes for 1 h at room temperature protected from light, using gentle agitation. The membranes were then washed three times for 10 min each with TBS-T ϩ 3% fatty acid-free BSA, protected from light, and gently agitated. Peptide binding was detected by analysis of fluorescence signal.
Molecular Dynamics Simulations-The K v 7.1/KCNE1 structural models from Kang et al. (38) were used. Eight PIP 2 molecules were introduced in proximity to the eight residue clusters identified by electrophysiology. Structures with and without PIP 2 were inserted into a membrane (composition PEA/PSE/ PCH ϭ 1/1/1) and incorporated into a simulation box filled with NaCl-H 2 O (0.9%). All-atoms-mobile simulations were performed to reach stable conformations using YASARA Structure version 10.1 (YASARA Biosciences GmbH, Vienna, Austria) as previously described (39).
Data Analysis-Mean current-voltage curves were analyzed using the Boltzmann equation multiplied with the driving force, considering increased current amplitudes caused by increased applied voltages, where I is the current amplitude, V is the applied voltage, E K is the potassium equilibrium potential, V1 ⁄ 2 is the half-maximal activation voltage, is the time constant, and A and B are the maximal and minimal current amplitudes, respectively.
The variation of structures during the molecular dynamics simulations is presented as the root mean square deviation, which is a measure of the deviation of the atom positions compared with the first structure, where t 1 ϭ 0, N is the number of atoms, r ជ i is the position of atom i at time t, m i is the atom mass, and M is the sum of the atom masses.
Because oocyte batches may vary in their endogenous PIP 2 levels and hence binding sites may be occupied to different extents, we included only those batches in which increased WT currents were observed after PIP 2 application. In these batches, currents of the mutants were increased, decreased, or unchanged ( Fig. 2A). To discover different properties upon diC 8 -PIP 2 application, we analyzed current amplitudes. The amplitudes at 40 mV were normalized to the respective current without the PIP 2 analog diC 8 -PIP 2 of each individual batch (Fig.  2B). diC 8 -PIP 2 significantly increased the WT (Ϫcys) current amplitude to ϳ160% on average, as well as the amplitudes of four mutants shown in gray (H126C, R195C, K196C, and K358C). Supposedly, the stimulation indicates that diC 8 -PIP 2 was able to stabilize the open channel conformation of these mutants. Four mutants showed reduced stimulation (marked in blue; K183C, R190C, R249C, and K354C) to values between 120 and 150%. In contrast, currents of mutants shown in red (R181C, R192C, R259C, R360C, and K362C) were either not affected or even significantly decreased by up to 25% upon diC 8 -PIP 2 injection, indicating an impaired functional interaction with PIP 2 . Further analysis of the diC 8 -PIP 2 effects on FIGURE 1. Intracellularly located mutations modulate K v 7.1/KCNE1 channel currents. Currents were recorded 3 days after injection. A, topology model of K v 7.1, showing all six transmembrane segments S1-S6 with pore loop and selectivity filter sequence GYG. Shown are all amino acids included in this study. Red, necessary for interaction with PIP 2 ; black, not relevant for interaction with PIP 2 (according to our electrophysiological data; see Fig. 2). The cysteine-free clone used in this study is based on isoform 0, whose N terminus is 95 amino acids shorter than isoform 1 and in which the first 11 amino acids are altered (green). B, representative current traces and voltage protocol with steps from Ϫ100 to 40 mV to Ϫ120 mV. Current of mutant H126C (red) was slightly decreased compared with wild type (black), whereas the current of mutant R181C was increased. C, mean data Ϯ S.E. of current-voltage relationships between wild type (emphasized) and mutants are similar. The arrow in voltage-step protocol marks the time point of analysis. D, currents at 40 mV were normalized to the corresponding wild type. The mean data Ϯ S.E. (all wild type recordings averaged) show significantly increased (R181C, K183C, R190C, R192C, K196C, R249C, R259C, and R360C) or unchanged (H126C, R195C, K354C, K358C, and K362C) current amplitudes. n, number of oocytes. Asterisks indicate significance: ***, p Ͻ 0.001; **, p Ͻ 0.01; *, p Ͻ 0.05; ns, not significant. mutant channel currents compared with wild type emphasized these differences (Fig. 2C). Here, current amplitudes were normalized to the WT (Ϫcys) current with diC 8 -PIP 2 of each individual batch. This analysis indicated no significant differences between current amplitudes of wild type and mutants shown in gray, suggesting that these residues do not play a significant role in the protein-lipid interaction. In contrast, current amplitudes of mutants depicted in blue were stimulated to a lesser extent compared with WT channels, implying an involvement in the interaction. Intriguingly, currents carried by several mutants shown in red were unchanged or reduced by diC 8 -PIP 2 injection, suggesting an important role for these residues in PIP 2 coordination.
Next we analyzed current-voltage relationships of the neutralizing mutants after injection of diC 8 -PIP 2 . To examine differences in voltage dependence of activation caused by diC 8 -PIP 2 , we fit our data according to the Boltzmann equation multiplied by the driving force (representative fits in Fig. 3A). Analysis of the half-maximal activation voltage V1 ⁄ 2 showed that the values of the various mutants were distributed between Ϫ20 and 10 mV compared with Ϫ4 mV of the WT (Ϫcys) (Fig. 3B). The voltage dependences of WT (Ϫcys) and mutants R181C, R190C, R249C, R259C, and K358C were unchanged by diC 8 -PIP 2 . The voltage dependences of mutants R192C, K354C, and R360C were shifted by diC 8 -PIP 2 to more negative potentials. Half-maximal activation voltages of mutants H126C, K183C, R195C, K196, and K362C were shifted to more positive potentials. However, significances could not be determined. Comparison between half-maximal activation voltage upon diC 8 -PIP 2 injection and current-amplitude changes revealed no correlation (Fig. 3C).
As a control, we compared the effect of diC 8 -PIP 2 on K v 7.1 WT and Ϫcys (Fig. 4A). Both current amplitudes were augmented, however, with Ϫcys augmented to a slightly but significantly lesser extent.
Notably, there are known LQTS mutants at amino acid positions 190,192,195,354,360, and 362. Because we could observe an impaired diC 8 -PIP 2 regulation when neutralizing these residues, we hypothesized that the impaired regulation is the underlying cause for channel dysfunction. Therefore, we also investigated these mutations electrophysiologically under the influence of diC 8 -PIP 2 (Fig. 4B). Indeed, mutations R190L, R192H, and R195W showed less increased current amplitudes. Mutations K354R and K362R showed even enhanced current amplitude increases (not significantly and significantly, respectively). Mutations R360T and R360M did not yield functional currents (data not shown). Lipid Binding Assay-The S2-S3 linker, the S4-S5 linker, and the C-terminal part of S6, all containing the examined amino acids, were subjected to lipid binding assays (Fig. 4C). Both the S2-S3 linker and S6 were shown to bind to all phosphatidylinositol (PtdIns) phospho derivatives, including PIP 2 . Additionally, the S6 fragment was positive for phosphatidyl serine binding. The S4-S5 linker did not bind to any of the tested lipids.
Molecular Dynamics Simulation-We based molecular dynamics simulations on K v 7.1/KCNE1 models (4:4 relation because of the ongoing discussion on stoichiometry (40)) in a closed and an open conformation as described by Kang et al. (38) and introduced eight PIP 2 molecules into the models. Conventional molecular dynamics simulations on channels in a membranous environment were performed until stable structures were reached. The last nanosecond of each simulation was averaged (Fig. 5A, cyan without PIP 2 and yellow with PIP 2 ) and compared with the starting structure (Fig. 5A, green). In the open conformation, differences between the starting structure and the average without the lipid were slightly larger than the differences between the starting structure and the average with PIP 2 . In the closed conformation, the structural differences without PIP 2 were much larger than in presence of PIP 2 .
Root mean square deviation (RMSD) analysis, which indicates the structural variation over time, showed that all structures were in equilibrium for at least the last nanosecond of simulation (Fig. 5B). In equilibrium, the RMSD of all simulations were ϳ4 Å, except for the closed conformation with PIP 2 with an RMSD of ϳ3 Å, agreeing with the direct visual comparison of the structures. The RMSD values of the simulations with and without PIP 2 in the open conformation (Fig. 5C, green and blue) were almost identical, whereas there was a large difference of ϳ1 Å in the closed conformation (Fig. 5C, black and  violet). Residue-resolved RMSD values were averaged over time and all four subunits (Fig. 5C). These values confined these results to the pore region, beginning with the S4-S5 linker, because the voltage sensor itself is much more flexible. Both regions seemed more flexible when all four subunits were averaged, although this was not the case for each individual subunit (data not shown). The results were confirmed by the averaged RMSD of the S4-S5 linker (residues 249 -261) and pore helix to the upper part of helix S6 region (residues 305-343; Fig. 4D). RMSD values without PIP 2 were enlarged by factors of ϳ1.8 and 1.3 (linker) and 1.5 and 1.2 (S6) in the closed and open conformation, respectively.
In addition, both PIP 2 -free structures were distorted. That is, ␣-helices were bent or in part degraded, hinting at a destabilization of the channel in the absence of PIP 2 . Recently, Hansen et al. (41) and Whorton and MacKinnon (42) demonstrated electrostatic interactions as the basis for PIP 2 binding in K ir channels, underscoring our assumption. Our results agreed with these data, because some identified residues surrounded the PIP 2 head groups (Fig. 5E).

DISCUSSION
PIP 2 is a known regulator of diverse ion channels and transporters (30). Recently, two crystal structures of ion channels in complex with PIP 2 have been resolved. Key to the interaction of channels with PIP 2 are positively charged residues that enable electrostatic interactions with the negatively charged phosphates of PIP 2 (41,42). There are several charged residues at the predicted membrane border of the inner membrane leaflet, which potentially represent PIP 2 interaction sites. In this study, we scanned for amino acids in K v 7.1 that functionally interact and potentially bind to PIP 2 . Wild type current amplitudes were up-regulated by PIP 2 to different extents, probably depending on the amount of endogenous PIP 2 and whether its concentration was saturating.

JOURNAL OF BIOLOGICAL CHEMISTRY 22753
dues Lys-183, Arg-190, Arg-249, and Lys-354 to be necessary for PIP 2 regulation, because their currents were increased by diC 8 -PIP 2 to a lesser extent than WT currents. Finally, amino acids Arg-181, Arg-192, Arg-259, Arg-360, and Lys-362 seemed even more important because diC 8 -PIP 2 failed to stimulate channel currents or even inhibited function. In the latter case, binding of PIP 2 to the mutants may be so unfavorable that the equilibrium shifts toward the closed instead of the open state. Interestingly, these mutants were the same as those that showed increased current amplitudes compared with WT currents with only three exceptions (K196C, K354C, and K362C). This suggests that endogenous PIP 2 significantly influences current amplitudes.
To determine whether changed effects on current amplitudes of the neutralizing mutants can be explained by differences in channel kinetics, we analyzed the half-maximal activation voltage of WT and mutant channels. First of all, halfmaximal activation voltages of many mutants differ largely from the wild type, which might be influenced by changed interaction with endogenous PIP 2 . Half-maximal activation voltage of WT was unchanged by application of diC 8 -PIP 2 . However, half-maximal activation voltages of several mutants were either shifted to more positive or negative potentials. Nevertheless, these V1 ⁄ 2 shifts do not correlate with the currentamplitude change upon diC 8 -PIP 2 application. One factor may be weak analysis of the V1 ⁄ 2 values, because Boltzmann fits were difficult to obtain, since conductivity-voltage relationships did not reach saturation (43), compromising detailed interpretation. The lack of V1 ⁄ 2 -I PIP 2 correlation suggests either complete or partial independence and too complex interaction of PIP 2 molecules to be analyzed by the techniques used here.
Mutations of several amino acids that we identified as PIP 2 interaction sites have been reported to be associated with LQTS (R190L, R190W, and R190Q; R192H and R192P; R195W; R259H, R259L, and R259C; K354R; R360G, R360T, and R360M; and K362R) (51)(52)(53)(54)(55)(56)(57)(58). It is plausible that disturbed PIP 2 interaction may contribute to an impaired channel function responsible for LQTS. For all of these residues, except for Arg-195, we demonstrated that the current amplitude was not changed by diC 8 -PIP 2 or significantly less increased than WT current amplitude, further supporting the notion that impaired PIP 2 binding, and hence regulation is involved in the pathogenesis of LQTS. To test this hypothesis, we investigated some of the mentioned mutations. Indeed, in mutants R190L, R192H, R195W, and R259C, regulation by PIP 2 was disturbed, corroborating our assumption. It is noteworthy that also R195W was up-regulated to a lesser extent, although we could not see a difference between R195C and the wild type. Therefore, it is presumable that not the neutralization of the arginine is responsible for the disturbed regulation but rather conformational changes introduced by tryptophan. Analysis of both conserved mutations K354R and K362R, on the other hand, revealed preserved PIP 2 regulation, which can be easily explained by the conservation of positive charges. Therefore, a disturbed PIP 2 regulation cannot explain channel dysfunction and LQTS development in these cases.
To date, PIP 2 binding sites have been identified mainly in the C terminus of ion channels. That is, residues in the C terminus of various inward rectifier potassium (K ir ) channels have been shown to participate in PIP 2 binding (33, 29, 44 -48). Recently, the binding site in K ir channels has been located to the interface between transmembrane domain and C terminus (41,42). Also, a cluster of basic amino acids in the C terminus of K v 7.2 and K v 7.3 has been demonstrated as a PIP 2 binding site (49). Moreover, the C-terminal residues Arg-539 and Arg-555 within K v 7.1 have been shown to participate in the interaction with PIP 2 (50). The amino acids identified here are located in the C-terminal portion of helix S6 beneath the membrane (Lys-354, Arg-360, and Lys-362) but also in the intracellular linkers between S2 and S3 (Arg-181, Lys-183, , as well as between S4 and S5 (Arg-249 and Arg-259; Fig. 2C). Unspecific binding to PtdIns was shown for both the S2-S3 linker and the C-terminal fragment of S6. This is in contrast to an earlier study by Thomas et al. (35), who reported that PIP 2 binds exclusively in the C terminus, because they did not observe lipid binding of the C-terminally truncated protein in lipid binding assays. A possible explanation for this conflict might be an incorrect folding of the truncated protein. However, Zaydman et al. (59) analyzed the same regions as in this study and also concluded that these regions form PIP 2 binding sites. In contrast to the electrophysiological data, the S4-S5 linker did not bind to any of the tested lipids. A possible explanation is that the investigated region is not sufficient for PIP 2 binding but requires further protein parts, for example Arg-243, which was already demonstrated to reduce PIP 2 affinity (50). However, amino acids Arg-249 and Arg-259 may affect PIP 2 regulation allosterically.
Residues outside of the C terminus have been detected before, for instance in the N terminus of K ir channels (42,46,48) and at the end of S4 in K v 7.1 (50). Thus, it is possible that all basic residues in a certain distance to the membrane mediate interaction with PIP 2 without forming a specific binding site. However, in our model obtained from the molecular dynamics simulations, the identified residues were clustered and might form distinct binding pockets because of their tertiary structure. The PIP 2 molecules were in direct contact with most identified residues. These residues might form two binding pockets per subunit together with noncharged residues or the protein backbone. The positions of these proposed binding sites are indicated in Fig. 6. Binding site I in Fig. 6 seemed to be close to the pore domain, which is in line with recent work by Hansen et al. (41), as well as Whorton and MacKinnon (42) showing PIP 2 molecules in this region in K ir channel crystal structures. Our results suggest a second binding site in each subunit located further outside at the interface between voltage sensor and plasma membrane.  AUGUST 15, 2014 • VOLUME 289 • NUMBER 33

JOURNAL OF BIOLOGICAL CHEMISTRY 22755
We found that the absence of PIP 2 led to a higher structural flexibility of the pore domain and a distortion of the channel in both the open and closed conformation, suggested by conventional molecular dynamics simulations (Fig. 5). Thus, PIP 2 presence might result in a stabilization of the overall channel structure in both conformations. However, in the absence of the lipid, the structural stability of the closed conformation was decreased as suggested by the ϳ1 Å larger RMSD values compared with the structure in the presence of PIP 2 . Consequently, the closed conformation might be structurally more stable in the absence of PIP 2 . Hence, the thermodynamic equilibrium may be shifted to the closed state, increasing V1 ⁄2 and thereby decreasing the open probability of the channel. Thomas et al. (35) have previously shown that four amino acids within K v 7.1 helix S6, Lys-354, Lys-358, Arg-360, and Lys-362, directly interact with PIP 2 . Based on double mutant analysis, they identified residues Lys-358 and Arg-360 as the most important ones. This is largely in agreement with our results, confirming the putative participation of Lys-354, Arg-360, and Lys-362 in PIP 2 interactions. Moreover, the importance of Arg-360 is underlined by the observation of completely abolished PIP 2 regulation when mutated. However, we could not find any significant participation of the residue Lys-358, because we observed no difference in PIP 2 sensitivity compared with WT. A possible explanation for this contradictory finding might be a weak interaction between PIP 2 and Lys-358 that can be functionally compensated by other amino acids interacting with PIP 2 . However, if both residues are lost, compensation might not be possible anymore, increasing the impairment in PIP 2 regulation. However, these interpretations require further investigation. Furthermore, Zaydman et al. (59) included the same amino acids as in our study. However, comparison of both studies reveals discrepancies as to which amino acids participate in the interaction and which do not. However, it has to be considered that Zaydman et al. (59) did not analyze the influence of PIP 2 directly but rather hypothesized endogenous PIP 2 as the cause for loss of function of the mutant channels. Therefore, other influences cannot be excluded.
In summary, we showed that several K v 7.1 amino acids participate in the PIP 2 regulation of the cardiac I Ks channel K v 7.1/ KCNE1. These are located in both the S2-S3 and S4-S5 linker, as well as in the intracellular region of helix S6. Together with molecular dynamics simulations, our results suggested the existence of at least two specific PIP 2 binding sites per subunit. Disturbed binding caused by mutations of these amino acids may contribute to the manifestation of LQTS.