PIP₂ controls voltage-sensor movement and pore opening of Kv channels through the S4–S5 linker

Abstract

Voltage-gated K⁺ (Kv) channels couple the movement of a voltage sensor to the channel gate(s) via a helical intracellular region, the S4–S5 linker. A number of studies link voltage sensitivity to interactions of S4 charges with membrane phospholipids in the outer leaflet of the bilayer. Although the phospholipid phosphatidylinositol-4,5-bisphosphate (PIP₂) in the inner membrane leaflet has emerged as a universal activator of ion channels, no such role has been established for mammalian Kv channels. Here we show that PIP₂ depletion induced two kinetically distinct effects on Kv channels: an increase in voltage sensitivity and a concomitant decrease in current amplitude. These effects are reversible, exhibiting distinct molecular determinants and sensitivities to PIP₂. Gating current measurements revealed that PIP₂ constrains the movement of the sensor through interactions with the S4–S5 linker. Thus, PIP₂ controls both the movement of the voltage sensor and the stability of the open pore through interactions with the linker that connects them.

Voltage-gated K⁺ (Kv) channels are tetrameric integral membrane proteins critical to membrane excitability that respond rapidly to changes in membrane potential to control membrane permeability to potassium ions. Upon membrane depolarization, a voltage sensor in each subunit undergoes a transition from a resting to an activated state followed by a concerted transition leading to the opening of the pore (1–4). The voltage-sensing domain [i.e., the S1–S4 transmembrane (TM) helices] of Kv channel subunits harbors within its S4 helix several positively charged residues that respond directly to changes in membrane voltage (5–7). The movement of these charges can be monitored by the gating current they produce, and the opening of the pore is monitored by the ionic current that follows. The S4–S5 linker couples the movement of the voltage sensor to the opening of the pore.

X-ray structures of Kv channels have shown that the S1–S4 voltage-sensing domains are exposed to lipids when embedded in a membrane (8, 9). A number of studies have suggested that, after depolarization, interactions of the S4 charges with lipids in the outer leaflet of the membrane are important in the stabilization of the sensor in the activated state (10–12).

Phosphatidylinositol-4,5-bisphosphate (PIP₂), a phospholipid that affects the activity of many types of ion channels (13, 14), acts as a docking platform for the N-terminal domain of fast-inactivating Kv channels (15). Activation of Ciona intestinalis voltage-sensitive phosphatase (Ci-VSP), which contains a voltage-sensing domain (S1–S4) coupled to a cytoplasmic phosphatase domain rather than a TM pore, shows a dependence on membrane depolarization similar to that of voltage-gated channels (16). PIP₂ modulates the motions of the Ci-VSP voltage-sensor domain and its coupling to the phosphatase domain by interacting with the linker that connects the voltage sensor and phosphatase domains (17).

In the present study we set out to determine whether PIP₂ modulates the gating mechanism of Kv1.2 channels and to identify specific regions where the lipid might interact to exert its effects. We used a number of different approaches to study PIP₂ depletion effects on the activity of Kv1.2 channel expressed in Xenopus oocytes, using the excised patch mode of the patch-clamp technique (18) or the two-electrode voltage-clamp technique on intact oocytes (7).

Results

Dual Effect of PIP₂ on Kv1 Channels.

The effects of PIP₂ depletion on Kv1.2 channels expressed in Xenopus oocytes were investigated first in excised inside-out macropatches.

Rundown.

Following patch excision, into a symmetrical high-K⁺ solution (ND96K), current mediated by Kv1.2 channels and activated by depolarizing steps to +60 mV successively decreased in amplitude (to 75% of the cell-attached value) (Fig. 1 A and B Upper), and the tail current deactivation was slowed successively [∼3.5-fold increase in the deactivation time constant (τ_deact)] (Fig. 1 C and D and Fig. S1). Similarly, the steady-state activation curve of Kv1.2 was shifted largely to the left following formation of inside–out patches (Fig. 1G). The midpoint voltage of activation (V_1/2) was approximately −67 mV when recorded immediately after patch excision; however, within 10 min of formation of the inside–out patch, the V_1/2 had shifted by −19 mV, to approximately −86 mV. The time course of rundown for this experiment (τ = 10.7 min) (Fig. S1A) was different from the time course of the tail current deactivation and the V_1/2 shift (τ = 2.3 min and 2.9 min, respectively) (Fig. S1B), indicating the existence of two kinetically distinct effects, one on the current amplitude and the other on the voltage sensitivity (V_1/2 shift and tail current deactivation), following inside–out patch rundown.

Fig. 1.

Effects of ND96K or FVPP perfusion of excised inside–out patches expressing Kv1.2 channels. (A) Temporal course of the current amplitude at +60 mV in ND96K (black squares) or FVPP (red circles) solutions. Current is normalized to that obtained immediately after patch excision (labeled as “control” in B). (B) Average current amplitude after a 10-min perfusion relative to control in ND96K (Upper) and FVPP (Lower) solutions (± SEM, n = 5). (C) Representative traces recorded at the time points indicated by the arrows in A for ND96K solution. (D) Summary of the time constants of tail current deactivation (_deact) from currents similar to those shown in C in control (black) and ND96K-perfused(red) patches for 10 min (± SEM, n = 5). (E) Representative traces recorded at the time points indicated by the arrows in A for FVPP solution. (F) Summary of the time constants of tail current deactivation from currents similar to those shown in E in control (black) and FVPP-perfused (red) patches for 10 min (± SEM, n = 5). (G) Steady-state activation curves of Kv1.2 channels determined in control patches (black) and after patches were perfused with ND96K solution for 10 min (red). (H) Steady-state activation curves of Kv1.2 channels determined in control patches (black) and after patches were perfused with FVPP solution for 10 min (red). Data points in G and H are mean ± SEM (n = 5). Lines are fits of a Boltzmann function to the data. *P < 0.05 versus control.

To limit the known enzymatic degradation of phospholipids in perfused inside–out patches, a phosphatase inhibitor-containing solution (fluoride, vanadate, pyro-phosphate; FVPP) was used (19, 20). Excision in FVPP solution largely slowed the decay of the Kv1.2 current observed under control conditions (Fig. 1 A and B Lower). However, the slowdown of the tail current deactivation (Fig. 1 E and F) and the left shift of the steady-state activation curve (Fig. 1H) were still present under these conditions. Thus, phosphatase inhibitors largely delayed the current rundown but did not prevent the changes in Kv1.2 channel tail current deactivation and V_1/2, suggesting differential PIP₂ sensitivities of these two PIP₂-dependent effects.

PIP₂ antibody.

We then examined the effect of reducing endogenous PIP₂ levels by applying an antibody specifically recognizing PIP₂ (PIP₂-Ab) directly to inside–out patches (19). Application of PIP₂-Ab caused a 25% reduction of the current amplitude at +60 mV (Fig. 2 A–C), an effect similar to the rundown results shown above (i.e., Fig. 1 A and B Upper). In addition to the blocking of current amplitude, effects on the tail current deactivation kinetics and V_1/2 (similar to the rundown results shown in Fig. 1) were observed. PIP₂-Ab caused a slowdown of the tail current deactivation kinetics (Fig. 2 D and F) and an ∼14-mV leftward shift in the voltage dependence of activation (Fig. 2E), indicating that the reduction in PIP₂ levels caused a pronounced stabilization of the voltage-dependent activated state of the channel. All three effects were reversed to the control conditions by applying exogenous PIP₂ to the excised patches (Fig. 2). Similar results were obtained with the Shaker K⁺ channel (a Shaker variant in which the inactivation domain was removed; Shk-IR) (Fig. S2).

Fig. 2.

PIP₂-Ab–mediated effects on voltage-dependent gating and current amplitude of Kv1.2 channels are reversed by PIP₂. (A) Temporal course of the Kv1.2 current amplitude at +60 mV before (black) and during (red) PIP₂-Ab application and during PIP₂ (blue) application. (B) Summary of the effects of PIP₂-Ab and PIP₂ on current amplitude (± SEM, n = 5). (C) Representative traces recorded at the time points indicated by the arrows in A. (D) Deactivation of Kv1.2 tail currents in control channels and with PIP₂-Ab and PIP₂ application. (E) Steady-state activation curves of Kv1.2 channels determined before (black) and after application of 10 μg/mL PIP₂-Ab (red) and during subsequent application of 10 μM PIP₂ (blue). Data points are mean ± SEM (n = 6). Lines are fits of a Boltzmann function to the data. (F) Summary of the time constants of deactivation (τ_deact) from tail currents similar to those shown in D in control channels and with PIP₂-Ab and PIP₂ application (± SEM, n = 5). *P < 0.05 versus control.

Wortmannin.

In addition to manipulating PIP₂ levels in inside–out patches, we examined the effects of reducing PIP₂ levels in the plasma membrane of intact oocytes. The activity of Kv1.2 channels expressed in Xenopus oocytes was recorded with the two-electrode voltage-clamp technique under control conditions and after the oocytes were preincubated for 2 h with 20 μM wortmannin, an inhibitor of the type-III PI 4-kinase at micromolar concentrations (21). As illustrated in Fig. 3 A and B, Kv1.2 channels were activated upon depolarizing voltage steps. Wortmannin caused a decrease in the current amplitude, measured in a voltage step where all of the channels were fully activated (Fig. 3C). Independently of this effect on the current level, Kv1.2 channels in oocytes pretreated with wortmannin required lower depolarizing voltage steps for equivalent activation than channels recorded under control conditions. Wortmannin treatment led to a left shift of channel activation by >10 mV, with V_1/2 = −21.64 mV and −31.90 mV as determined with Boltzmann fits to the mean steady-state activation of control and wortmannin-treated Kv1.2 channels, respectively (Fig. 3D). In addition, wortmannin caused a decrease in the rate of the tail current deactivation (Fig. 3E).

Fig. 3.

Modulation of Kv1.2 channel currents by wortmannin. (A and B) Representative whole-cell current traces from oocytes expressing Kv1.2 channels evoked by depolarizing voltages from −80 to +60 mV followed by a step to −80 mV for tail current analysis preincubated for 120 min in control (A) or wortmannin (B) solutions. (C) Average amplitude of Kv1.2 channels at +60 mV (± SEM, n = 5). (D) Steady-state activation curves of Kv1.2 channels determined in control (black) and wortmannin-treated (red) oocytes. Data points are mean ± SEM (n = 5). Lines are fits of a Boltzmann function to the data. (E) Summary of time constants of tail current deactivation (τ_deact) in control and wortmannin-treated oocytes (± SEM, n = 5). *P < 0.05 versus control.

Ci-VSP.

We also depleted PIP₂ in intact oocytes with Ci-VSP (16). Reversible PIP₂ depletion can be achieved after membrane depolarization to activate Ci-VSP. Ci-VSP was coexpressed with the channel, and its activity was controlled with the voltage protocol shown in Fig. 4A, Upper in the whole-cell configuration, using the two-electrode voltage-clamp technique. An up-ramp protocol from −80 to +60 mV was applied first and was used as a control (resting PIP₂ levels before the voltage protocol reached 0 mV), and a second, down-ramp protocol from +60 to −80 mV was applied after the oocytes were held at +60 mV for 3 s for partial depletion of PIP₂ from the plasma membrane. Application of the up- and down-ramps in oocytes expressing Kv1.2 channels alone produced nearly indistinguishable linear currents (Fig. 4A, Lower); the difference in the voltage at 10% of activation (ΔV_1/10) between the up- and down-ramps was 2.28 ± 0.43 mV (n = 4). Activation of Ci-VSP during the up-ramp (>0 mV) caused a decrease in current amplitude, as can be seen by the deviation of current from linearity, whereas application of the down-ramp in addition to a further decrease in current amplitude also showed a shift in the voltage dependence of activation (Fig. 4B), with a ΔV_1/10 of 13.44 ± 1 mV (n = 6). Finally, coexpression of Kv1.2 with the catalytically inactive Ci-VSP mutant (C363S) showed currents similar to those in oocytes expressing Kv1.2 alone (Fig. 4C), with a ΔV_1/10 of 2.34 ± 0.3 mV (n = 6). These results in whole-cell recordings recapitulated and confirmed the results in excised patches with regard to the PIP₂-mediated effects on the current amplitude and voltage sensitivity of Kv1.2 channels. Just as with excised patches, the two effects on current amplitude and voltage dependence of activation mediated by Ci-VSP were kinetically distinct: Although the recovery of the shift in the voltage dependence of activation took only several milliseconds (Fig. S3 A and B), the recovery of current amplitude required >25 s of repolarization before reaching control levels (Fig. S3 C and D). Taken together, these results in both excised patches and intact oocytes suggest that PIP₂ modulates the current amplitude and voltage dependence of activation of Kv channels by distinct mechanisms (Figs. 1–4 and Figs. S1–S3).

Fig. 4.

Modulation of Kv1.2 channel voltage-dependent gating and current amplitude by Ci-VSP. (A) Representative current traces evoked by the ramp pulses (Upper) for oocytes expressing Kv1.2 channels alone (Lower), coexpressed with Ci- VSP (B), and with the catalytically inactive mutant C363S of Ci-VSP (C). Rightmost panels in A, B, and C show the current–voltage relationships (up-ramp is black and down-ramp is red) from the representative traces.

Voltage Dependence and Channel Gating Possess Distinct Sensitivities to PIP₂.

The rundown experiments in ND96K and FVPP solutions (Fig. 1) suggested differential sensitivities of the two PIP₂-dependent effects on the phosphoinositide. The soluble dioctanoylglycerol-PIP₂ (diC8-PIP₂) is commonly used to quantify the sensitivity of a channel to PIP₂; unlike long-chain PIP₂ analogs, the effects of diC8-PIP₂ are readily reversible, making it possible to construct dose–response curves. We examined the diC8-PIP₂ sensitivity of the two related effects that we observed in Kv1.2 channels. The application of increasing concentrations of diC8-PIP₂ to inside–out patches speeds up the tail current deactivation kinetics (Fig. 5) and shifts back the voltage-dependent activation to depolarized potentials with an estimated EC₅₀ of 71.4 ± 15.2 μM (Fig. 5B; the maximum ΔV_1/2 was set to 14 mV, the value obtained in our PIP₂-Ab experiments). Interestingly, application of diC8-PIP₂ did not recover the current amplitude of the channel even at concentrations as high as 300 μM, suggesting that the current amplitude and voltage dependence of activation possess distinct sensitivities to PIP₂. Fig. 5 C and D shows temporal courses of the current amplitude measured at +60 mV (Fig. 5C) and of the rate of tail current deactivation (Fig. 5D) after application of 10 μg/mL PIP₂-Ab, 30 μM diC8-PIP₂, and 10 μM long-chain PIP₂. The application of PIP₂-Ab decreased the Kv1.2 current amplitude and the rate of tail current deactivation, as was seen in our previous experiments; however, application of 30 μM diC8-PIP₂ did not have any effect on current amplitude but partially recovered the rate of tail current deactivation. Subsequent application of long-chain PIP₂ recovered both the current amplitude and rate of tail current deactivation to levels comparable with those before the application of PIP₂-Ab (Fig. 5 C and D).

Fig. 5.

diC8-PIP₂ restores the voltage-dependent gating of Kv1.2 channels but fails to recover the current amplitude. (A) Representative current traces evoked by a pulse to +60 mV from a holding potential of −120 mV before (black) and after (red) application of 10 μM PIP₂-Ab and during subsequent application of 10 μM diC8-PIP₂ (green). (B) Dose–response curve for the ΔV_1/2 as a function of the concentration of diC8-PIP₂. Data points are mean ± SEM (n = 3–4). Data were fit with the Hill equation. The maximal ΔV_1/2 was fixed at 14 mV (the maximum shift recorded) during the adjustment). (C) Temporal course of the current amplitude at +60 mV in control (black) channels and with PIP₂-Ab (red), diC8-PIP₂ (green), or long-chain PIP₂ (blue) application. Current is normalized to that obtained immediately after patch excision. (D) Temporal course of the time constant of tail current deactivation (τ_deact) color-coded as in C. Experiments in C and D were repeated four times.

PIP₂ Increases Activity by Affecting the Open Probability of Kv1.2 Channels.

PIP₂ rundown or depletion by chelating agents (e.g., PIP₂-Ab) decreased the macroscopic conductance, whereas PIP₂ perfusion had the opposite effect. We turned to single-channel recording to examine whether the effects of PIP₂ involved the channel open probability, unitary conductance, or the number of active channels in the patch (I = NP_oi, where I is the macroscopic current, N the number of active channels, Po the open probability, and i the unitary current). Fig. 6A shows representative activity from a single channel in an inside–out patch held at either −60 mV or +60 mV under rundown conditions or following application of PIP₂. At both voltages PIP₂ increased the open probability without changing the unitary current size. A comparison of the unitary conductance from patches under rundown conditions vs. conditions of PIP₂ application showed similar values at ∼15 pS (Fig. 6B). In contrast, PIP₂ yielded a fourfold increase in open probability from patches containing a single channel (Fig. 6C). Patches that contained multiple channels yielded an ∼fourfold stimulation in NPo similar to that seen in single-channel patches, suggesting that the open probability, rather than the number of active Kv1.2 channels, was affected by PIP₂.

Fig. 6.

PIP₂ increases open probability but not the unitary conductance or number of active channels. (A) Comparison of Kv1.2 single-channel current traces before (at the end of the rundown period) (Left) and after application of PIP₂ (Right), held at a membrane potential of either −60 (Lower) or +60 mV (Upper). Single-channel current traces of Kv1.2 were recorded from excised inside–out patches. Arrows indicate the zero current level (closed state of channels). (B) Comparison of Kv1.2 i–V relationships between unitary conductances in the absence (γ = 15.5 pS, n = 5) (red) and presence (γ = 14.2 pS, n = 6) of PIP₂ (blue). i–V relationships were measured with symmetrical 96 mM K-MeS. (C) Summary of the single-channel open probability (Po) for Control (following patch excision; NPo = 45.2 ± 4%), rundown (just before PIP₂ application; NPo = 8.4 ± 4%), and after application of 10 μM PIP₂ in the bath (NPo = 32.6 ± 7%). (D) Activity (NPo) versus time plot of a representative inside–out patch containing multiple Kv1.2 channels. The filled horizontal bar indicates the time period of PIP₂ application in the bath. The open horizontal bars indicate the same length of time used to calculate NPo from the first appearance of activity in the rundown or the PIP₂ conditions. **P < 0.01, *P < 0.05.

Molecular Determinants for Dual Effects of PIP₂ on Kv1 Channels.

We next asked whether the two distinct PIP₂-dependent effects on Kv channels we have described are controlled by different molecular determinants. We hypothesized that, as with other proteins, positively charged amino acids could be involved in electrostatic interactions with PIP₂ in the plasma membrane and that these interactions might be responsible for the PIP₂ effects observed in Kv1.2 channels (22). To identify molecular determinants involved in PIP₂ regulation, we mutated a number of positively charged residues in the N and C termini and in the S4–S5 linker of the Kv1.2 channel. Fig. 7 A and B shows that K322Q (in the S4–S5 linker) and R147Q (in the N terminus) caused a larger shift in the voltage dependence of activation, as compared with wild-type channels, when PIP₂-Ab was applied to inside–out patches. Similar results were obtained in whole-cell recordings coexpressing the Kv1.2 channels with Ci-VSP (Fig. S4). Fig. 7 C and D shows that K322Q, but not R147Q, also exhibited a significant reduction in the current amplitude seen by PIP₂-Ab application compared with the wild type. Moreover, the S4–S5 linker mutant R326Q caused a significant change in the current amplitude but not in the voltage sensitivity. These data indicate that R147Q and R326Q affect only voltage sensitivity and current amplitude, respectively, but K322Q affects both.

Fig. 7.

Distinct determinants of the dual effects of PIP₂ and structural consequences of channel–PIP₂ interactions via the S4–S5 linker. (A) Steady-state activation curves of Kv1.2 wild-type and mutant (K322Q and R147Q) channels determined before (black) and after (red) application of 10 μg/mL PIP₂-Ab. Data shown are ± SEM; n = 4–5. (B) Average shift in V_1/2 shown in response to PIP₂-Ab in Kv1.2 wild-type and mutant channels. Data shown are ± SEM; n = 3–5. (C) Representative current traces at +60 mV before and after PIP₂-Ab application in Kv1.2 wild-type and mutant (K322Q and R326Q) channels. (D) Fraction of current not blocked by PIP₂-Ab (antibody/control) in Kv1.2 wild-type and mutant channels. Data shown are ± SEM; n = 3–5. *P < 0.05 versus control.

Interactions of the S4–S5 Linker with PIP₂ Are Coupled to the Voltage Sensitivity and Gating Machineries of Kv1 Channels.

Next we investigated whether the presence of PIP₂ could cause changes in the Kv channel structure. Crystal structures of part of the Kv1.2 channel in the open state have been reported (8, 23). We proceeded to dock PIP₂ to a Kv1.2 homology model of the open conformation based on these structures (Methods) as well as to a model of the closed state (24, 25). Localization of the three residues whose mutation altered PIP₂-dependent effects on Kv1 channels suggested that PIP₂ could engage all three residues in the closed (Fig. 8 A and C) but not in the open (Fig. 8B) conformation. In the open conformation K322 still could interact with PIP₂, but R147 has moved quite far from the putative PIP₂ binding site, and the rotation of R326 in this conformation would prevent it from interacting with PIP₂ (Fig. 8B). The crystal structure of the Kv1.2/Kv2.1 chimera depicts a lipid bound to the S4–S5 region, but this lipid does not interact with any of the three residues we identified (8). Interestingly, in the model of the closed conformation of Kv1.2 (24, 25), the three key residues (R147, K322, and R326) are oriented so that they all can interact with PIP₂ simultaneously (Fig. 8 A and C), similar to the manner in which PIP₂ interacts with Kir2.2 in the crystallographic structure (Discussion) (26). We next performed molecular dynamics (MD) simulations of our open-state homology model in the absence (Apo) and presence (Holo) of PIP₂ and a combined principal components analysis (PCA), using a trajectory from the MD simulations that concatenated the equilibrated trajectories (50–100 ns) in the Apo and Holo conditions. Fig. 8D displays the first eigenvector of the combined PCA for one subunit in the absence (red) and presence (blue) of PIP₂ to depict the collective motions of Kv1.2 caused by PIP₂. Significant PIP₂-induced movements were seen in several segments of the channel implicated in the gating mechanism of Kv1 channels.

Fig. 8.

PIP₂ interactions with critical Kv1.2 residues in the closed conformation affect the motion of critical components of the gating machinery. (A) A model of the closed conformation of Kv1.2 (24, 25) showing the distance between the side-chains of R147 and K322. (B) A model of the open conformation of the Kv1.2 (Methods) showing the distance between the side-chains of R147 and K322. (C) A model of the closed structure of Kv1.2 illustrating how PIP₂ could bind to the channel, engaging all three critical residues identified by the experiments shown in Fig. 7. (D) PCA analysis showing the collective movements of a Kv1.2 subunit from the absence of PIP₂ (red) to the presence of PIP₂ (blue). Flexible loops were removed from the structures shown to illustrate better the PIP₂-induced collective movements of less flexible regions.

The presence of PIP₂ caused significant movements; furthermore, these movements were correlated. The largest changes observed in the correlated movement of K322 in the S4–S5 linker were with residues in the selectivity filter (SF) and the S2 and S3 regions following PIP₂ binding (Fig. S5 A and C). Quantification of these changes revealed increased correlation in the movements of the S4 helix (at K322) with the S2 and S3 residues but a decrease in the correlation with the SF. Fig. S5D highlights these residues in the open conformation of the channel.

Disruption of Critical S4–S5 Linker Interactions with PIP₂ Removes a Restraint in Voltage-Sensor Movement.

Could PIP₂ directly affect the voltage-sensor movement, thereby controlling the voltage sensitivity of the channel? To address this question, we examined the effect of reducing endogenous PIP₂ levels or channel–PIP₂ interactions on gating currents. For these experiments, we used the well-studied ShK-IR channel for ionic current meaurements and its nonconducting W434F mutant (ShK-W434F) for gating current measurements (Fig. 9 A and B, respectively). Wortmannin treatment induced a leftward shift in the voltage dependence of activation of Shaker ionic and gating currents (Fig. 9 C and D, respectively). Following PIP₂ depletion, just as with Shaker ionic current deactivation under PIP₂-Ab conditions (Fig. S2D), the rate of the off-gating currents was slowed under wortmannin conditions (Fig. 9I). Prompted by effects of S4–S5 linker mutants on Kv1.2 channels, we mutated two positively charged residues in the Shaker S4–S5 linker, creating a chimeric channel with a linker identical to that of Kv1.2 (K322Q). The Shaker S4–S5 R387Q-K390Q double mutant caused a leftward shift in the voltage dependence of activation of Shaker ionic and gating current (Q_on) (Fig. 9 E and F, respectively) as well as a slow down in the off-gating current (Fig. 9I). These PIP₂-mediated effects on ionic and gating currents have been summarized in the ionic current-to-voltage (I–V) and gating current-to-voltage (Q–V) curves shown in Fig. 9 G and H, respectively. Thus, these experiments suggest that PIP₂ constrains the voltage-sensor movement of Kv channels through interactions with positively charged residues in the S4–S5 linker (i.e., K322) and possibly other participating sites (e.g., R147) of Kv channels that may come together in the 3D closed conformation to contribute to the PIP₂ binding site.

Fig. 9.

Effect of PIP₂ depletion on ionic and gating currents of Shk-IR. (A and B) Representative traces of Shk-IR ionic currents (A) evoked by depolarizing voltages from −120 to +50 mV followed by a step to −90 mV for tail current analysis and of ionic current-removed Shaker-W434F (Shk-W434F) gating currents (B) evoked by depolarizing voltages from −120 to +50 mV followed by a step to −90 mV. (C and D) Representative traces of Shk-IR ionic currents (C) and Shk-W434F gating currents (D) after wortmannin treatment (2 h in 50 μM wortmannin). (E and F) Representative traces for Shk-IR (R387Q-K390Q) ionic currents (E) and for Shk-W434F(R387Q-K390Q) gating currents (F). (G) Current–voltage relationships for Shk-IR control, wortmannin-treated, and Shk-IR(R387Q-K390Q) channels. Data shown are ± SEM; n = 4–7. (H) Charge–voltage relationships for Shk-W434F controls, wortmannin-treated, and Shk-W434F(R387Q-K390Q) channels. Data shown are ± SEM; n = 4–5. (I) Summary of the time constants (τ) of the off-gating currents for Shk-W434F (control), wortmannin-treated, and the R387Q-K390Q mutant. Data shown are ± SEM; n = 4–5. *P < 0.05 versus control.

Discussion

In the present work we found that PIP₂ exerts two distinct effects on Kv channels. First, PIP₂ stabilizes the voltage sensor of Kv channels in a state of decreased sensitivity (Q–V and I–V are right shifted), favoring the closed state of the channel. Second, PIP₂ stabilized the Kv channel pore in the conducting state. These two apparently contradictory effects could be separated by differences in kinetics (Figs. S1 and S3), sensitivities to PIP₂ (Figs. 1 and 5), and molecular determinants (e.g., R147 influenced only the PIP₂ effect on voltage sensitivity but not on the macroscopic current; Fig. 7 and Fig. S4).

Most actions of PIP₂ that occur through binding to specific protein sites involve electrostatic interactions between the negatively charged head-group of PIP₂ and positively charged amino acid residues at the target site. This interaction has been demonstrated for direct PIP₂ regulation of several ion channels (27). PIP₂ interacts with Kir channels engaging several intracellular sites, most of which are positively charged residues (20, 28). Similarly, positive residues just after the end of S6 are thought to contribute to the PIP₂ sensitivity of KCNQ1/KCNE1 channels (29). Here our data suggest that the PIP₂-induced effects involve interactions with positively charged residues in the S4–S5 linker (Figs. 7–9), a region responsible for the coupling between the voltage sensor and the pore domain in Kv channels. This interaction directly modulates the movement of the voltage sensor (Fig. 9) and maintains high activity of the channel, possibly by stabilizing the coupling between the sensor and the pore. It has been shown that prolonged depolarizations shift the Q–V relationship of voltage sensor-containing proteins to the left, a process that has been referred to as “relaxation” (30). The effects of PIP₂ depletion or of weakening the interactions of the S4–S5 linker on the Q–V relationship (Fig. 9H) were reminiscent of relaxation. Whether prolonged depolarization causes relaxation by decreasing the affinity of the S4–S5 linker to PIP₂ through electrostatic repulsion remains to be examined.

The voltage sensitivity of Kv channels has been shown to be regulated by negatively charged lipids other than PIP₂, including ceramide-1-phosphate (11), and polyunsaturated fatty acids (31). However, this modulation involves channel interactions with the external leaflet of the bilayer, where these lipids are localized. In contrast, the modulation of the channel by PIP₂ is confined to the inner leaflet of the plasma membrane, and its magnitude could be under dynamic control.

Collectively, our data support a model in which positive residues in the N terminus and the S4–S5 linker (e.g., R147, K322, and R326) interact with PIP₂. In the Kir2.2 structure in complex with PIP₂, R78 and R80 are located at the interface between the TM and cytosolic domains of Kir2.2, and their side-chains point in the same direction (26). Similarly, in the proposed model of the PIP₂ binding site in the closed conformation of Kv1.2 (Fig. 8C), the 1′ phosphate of PIP2 interacts with K322 and R326, which are located in the S4–S5 linker in proximity to the loop that connects the linker to the S5 TM helix and at the interface between the cytosolic and TM domains. Like the orientation of R78 and R80 in the structure of Kir2.2 in complex with PIP₂, the side-chains of K322 and R326 point in the same direction in the closed conformation of Kv1.2. Furthermore, like the binding site of PIP₂ in Kir2.2, where the acyl chains interact with hydrophobic residues in the TM helices, in the proposed binding site of PIP₂ in Kv1.2 the acyl chains insert into the membrane where they can interact with the hydrophobic amino acids in S3, S5, and possibly S4 (Fig. S5). With the acyl chains and 1′ phosphate of PIP₂ interacting with the TM domain, the inositol ring is oriented toward the cytosolic domain (as in Kir2.2) and interacts with R147.

As the voltage sensor transitions from the resting to the activated state, the S4–S5 residue interactions exert a dual effect. On one hand, they constrain the movement of the voltage-sensing domain; on the other hand, they stabilize the opening of the pore, presumably by constraining the linker with PIP₂ and achieving stronger coupling between the voltage-sensing domain and the pore. K322 was the only residue identified that is predicted to interact with PIP₂ in both the open and closed states. K322 neutralization also was the only residue mutation that affected the stability of both the voltage-sensing domain and the pore. In contrast, R326, which was predicted to interact with PIP₂ only in the closed-state model, did not affect voltage sensitivity, possibly because the upstream interaction between K322 and PIP₂ was still intact. It is possible that mutation of R326 affects the PIP₂-dependent pore stability indirectly by altering the K322–PIP₂ interactions. This suggestion will require validation by further experimental tests. Furthermore, our mutagenesis efforts did not yield insights on molecular determinants of the slow PIP₂ effects on the macroscopic current. Could the stability of the pore also involve cytosolic determinants other than the S4–S5 linker, as is the case in Kir channels? The lack of experimental and structural insights implicating the cytosolic domains of Kv1 channels in interactions with PIP₂ makes it difficult to answer this question at the present time. Clearly, a more complete molecular understanding of how PIP₂ exerts its slow control of the conducting open conformation of the channel will be needed.

The dual effect of PIP₂ on voltage sensitivity and on the stability of the open state also has been shown for voltage-gated N- and P/Q-type Ca²⁺ channels (32). Although currently untested, it is possible that the S4–S5 linker of Ca_v channels, as in Kv channels, is critically involved in mediating the effects of PIP₂. On the other hand, as is the case for Ca_v channels, it is likely that activation of membrane receptors that dynamically deplete PIP₂ levels in excitable cells also has important consequences for cell excitability through modulation of Kv channel activity.

It is likely that recognition of the key role PIP₂ plays in the activation of highly voltage-dependent channels, such as Ca_v and Kv channels, will stimulate the search for a deeper structural understanding of how PIP₂ interacts with full-length Kv channels and how physiological modulation of phosphoinositide levels affects cellular excitability.

Methods

Molecular Biology and Channel Expression in Oocytes.

Plasmids rKv1.2-pXoom, Ci-VSP-sp64t, Shaker H4-pBSTA with N-type inactivation removed (Δ6–26) (Shk-IR), and Shaker W434F-pBSTA were used. The cDNA was linearized, and cRNA was prepared using the MessageMachine kit (Ambion). Point mutants were introduced by Pfu-based mutagenesis using the QuikChange TM kit (Stratagene). The mutations were verified by sequencing. Xenopus oocytes were injected with different concentrations of the RNAs used, depending on the expression level of the given protein.

Electrophysiology.

Recordings in Xenopus oocytes were performed 24–72 h after RNA injection. Currents in whole oocytes were measured by conventional two-electrode voltage-clamp with a GeneClamp 500 amplifier (Molecular Devices). Electrodes were filled with 1.5% (wt/vol) agarose in 3 M KCl. The electrodes had resistances of 0.5–1 MΩ. Oocytes were perfused with a high-potassium solution containing (in mM) 96 KCl, 1 NaCl, 1 MgCl₂, and 5 Hepes, pH 7.4. A low-potassium solution was used in experiments coexpressing Kv1.2 channels and Ci-VSP, containing (in mM) 96 NaCl, 2 KCl, 1 MgCl₂, and 5 Hepes, pH 7.4. Data acquisition and analysis were carried out using pClamp8 (Molecular Devices) and Origin 7 (Microcal) software.

Macropatch channel activity was recorded under the cell-attached and inside–out configurations of the patch-clamp technique using an Axopatch 200A patch-clamp amplifier and pClamp8 data acquisition software (Molecular Devices). Electrodes were made from borosilicate glass (World Precision Instruments) using a Sutter P-97 microelectrode puller (Sutter Instrument) and had a tip diameter of 5–25 μm. Three sets of solutions were used on both sides of the membrane: (i) ND96K solution: (in mM) 96 KCl, 1 MgCl₂, 5 EGTA, and 10 Hepes (pH 7.4); (ii) FVPP solution: (in mM) 96 KCl, 5 EDTA, 10 Hepes, 5 NaF, 0.2 Na₃VO₄, and 10 Na₂PO₇ (pH 7.4) to delay PIP₂ dephosphorylation (19); and (iii) gating current solution: (in mM) 120 N-methyl-D-glucamine (NMG⁺), 2 EGTA, 2 Ca²⁺MeSO⁻₃, and 10 Hepes (pH 7.4). Data were analyzed using the Clampfit 8 and Origin software.

Kv1.2 single-channel currents were recorded from oocytes under the standard inside–out patch configuration. The pipette solution contained (in mM) 96 K-MeS, 10 Hepes, and 2 mM MgCl₂, pH 7.4, adjusted with KOH. The composition of the bath solution was (in mM) 96 potassium 4-morpholineethanesulfonate (K-MeS), 10 Hepes, and 5 EGTA, pH 7.4, adjusted with KOH.

Single-channel conductance values were determined by the slope of current–voltage (i–V) curves where i–V data could be well fitted to a linear line. Open probability was determined simply by dividing the total open time by the sum of the open and closed times in patches containing only one channel. When multiple channels were present in the patch, open probability was determined from the amplitude histogram as P_o (%) = (1 − P_C^1/N) × 100, where P_C is the fraction of area under the closed state, and N is the maximal number of simultaneous channel openings observed when the open probability was high (e.g., at positive voltages).

Chemicals.

PIP₂ and diC8-PIP₂ and PIP₂ antibody were purchased from Avanti Polar Lipids, PIP₂-Ab was purchased from Enzo Life Sciences (formerly Assay Designs), and wortmannin was purchased from Sigma-Aldrich. PIP₂, diC8-PIP₂, and PIP₂-Ab were prepared as described previously (19, 28).

Data Analysis.

Data in all figures are expressed as mean ± SEM. Statistical significance was evaluated by Student´s t test.

Homology Modeling for Kv1.2 Channel.

Homology model construction.

The crystal structure of the Kv1.2 channel [Protein Data Bank (PDB) ID 2A79; resolution: 2.9 Å] (23) is missing electron density in segments S1 and S3 and in the loops connecting segments S1, S2, S3, and S4. The most complete Kv channel crystal structure is the Kv1.2 and Kv2.1 chimera (PDB ID 2R9R; resolution: 2.4 Å) (9). The Kv1.2/2.1 channel chimera structure was used as a template to develop homology models for the Kv1.2 channel. We used the Kv1.2/2.1 chimera and the Kv1.2 channel (Genlnfo Identifier: 52000923) for sequence alignment [ClustalW server (http://www.ebi.ac.uk/Tools/clustalw2/index.html)]. The Kv1.2/2.1 chimera was constructed mostly by the Kv1.2 channel. The sequence identity between the Kv1.2/2.1 chimera and the Kv1.2 channel is 93%. Only a small region (27 residues) between the S3 and S4 helices of the chimera structure came from the Kv2.1 channel. We replaced the Kv2.1 residues on the S3 and S4 helices of the Kv1.2/2.1 chimera structure with the corresponding Kv1.2 residues and regenerated an S3–S4 linker by using the loop structural template database search in the SYBYL graphic software package (Tripos Inc.). This regeneration was followed by energy minimizations using the SYBYL software package. The Kv1.2 model structure was evaluated by the PROCHECK program (33, 34) for the quality of the stereochemistry. The phi and psi distributions showed that in the final model 91% of residues were in the most favored regions of the Ramachandran plot, 9.0% of residues were in allowed regions, and none was in disallowed regions. This result compared favorably with the structure of the Kv1.2/2.1 crystal template, where 89.9% of residues were in the most favored regions, 11.1% of residues were in the allowed regions, and none was in the disallowed regions. The overall PROCHECK G-factor of the model was 0.0. All these assessments indicated that the model structure was reasonable.

PIP₂ and Kv1.2 Interactions.

We used the AUTODOCK program (35) to dock the PIP₂ head group into the Kv1.2 model structure. AUTODOCK is an automatically flexible molecular docking program that can account for the flexibility of the PIP₂ head group during the docking simulations. The grid-based potential maps that were generated for the Kv1.2 channel using CHNOP (i.e., carbon, hydrogen, nitrogen, oxygen, and phosphorus) elements, sampled on a uniform grid containing 120 × 120 × 120 points, were 0.375 Å apart for the free-energy calculations. The grid box was centered at the geometrical center of residues K147, K322, and R326 of Kv1.2, which were found by our functional study to be important for PIP₂ sensitivity. The Lamarckian Genetic Algorithm was used to identify the binding conformations of the PIP₂ head group. One hundred docking simulations were performed. The final docked PIP₂ head group configurations were selected based on docked binding energies and cluster analysis. A potential binding site of the Kv1.2 channel for PIP₂ was identified by docking simulations, formed by positively charged residues K134, R138, K312, K322, and R419.

MD simulations.

MD simulations (100 ns in duration) were conducted on Kv1.2 in the presence and absence of PIP₂. The channel was immersed in an explicit palmitoyl-oleoyl phosphatidylcholine bilayer of ∼35-Å thicknesses. After being solvated with single point charge water molecules, neutralized by Na⁺ as the counter ions, and including K⁺ located in the selectivity filter obtained from the Kv1.2/2.1 crystal, each system used in the MD simulations had ∼175,000 atoms. GROMACS v4.0.5 (36) was used to conduct the simulation with GROMOS96 53a6 force field (37).