Phosphatidylinositol-4,5-bisphosphate, PIP₂, controls KCNQ1/KCNE1 voltage-gated potassium channels: a functional homology between voltage-gated and inward rectifier K⁺ channels

Abstract

Phosphatidylinositol-4,5-bisphosphate (PIP₂) is a major signaling molecule implicated in the regulation of various ion transporters and channels. Here we show that PIP₂ and intracellular MgATP control the activity of the KCNQ1/KCNE1 potassium channel complex. In excised patch–clamp recordings, the KCNQ1/KCNE1 current decreased spontaneously with time. This rundown was markedly slowed by cytosolic application of PIP₂ and fully prevented by application of PIP₂ plus MgATP. PIP₂-dependent rundown was accompanied by acceleration in the current deactivation kinetics, whereas the MgATP-dependent rundown was not. Cytosolic application of PIP₂ slowed deactivation kinetics and also shifted the voltage dependency of the channel activation toward negative potentials. Complex changes in the current characteristics induced by membrane PIP₂ was fully restituted by a model originally elaborated for ATP-regulated two transmembrane-domain potassium channels. The model is consistent with stabilization by PIP₂ of KCNQ1/KCNE1 channels in the open state. Our data suggest a striking functional homology between a six transmembrane-domain voltage-gated channel and a two transmembrane-domain ATP-gated channel.

Introduction

The KCNQ1/KCNE1 channel complex underlying the delayed rectifier potassium current, I_Ks, is key in controlling the duration of the action potential of the human heart (Barhanin et al., 1996; Sanguinetti et al., 1996) and also potassium secretion in the endolymphatic space of the inner ear (Vetter et al., 1996). Genetic mutations affecting either KCNQ1 (the α-subunit) or KCNE1 (the β regulatory subunit) are responsible for long QT syndrome, a rare, albeit severe, human genetic disorder, and also for congenital deafness (Towbin and Vatta, 2001).

The phospholipid phosphatidylinositol-4,5-bisphos phate (PIP₂) is an important regulator of several ion channels and transporters (Shyng and Nichols, 1998; Hilgemann et al., 2001). In particular, PIP₂ has been reported to modulate the voltage dependency of recombinant HERG channels and to stabilize the activity of the voltage-dependent potassium channels HERG and Kv2.1 in excised patches (Bian et al., 2001; Hilgemann et al., 2001). The KCNQ family of voltage-dependent potassium channels is also regulated by PIP₂ (Suh and Hille, 2002; Zhang et al., 2003). The regulation by PIP₂ of the pancreatic ATP-sensitive K⁺ channel (constituted by the assembly of Kir6.2 α-subunits and SUR1 proteins), a two transmembrane-domain inward rectifier channel, has been investigated in greater detail (Baukrowitz and Fakler, 2000). In the case of the Kir6.2/SUR1 channels, a model has been established which proposes that PIP₂ stabilizes the open state of the channel, whereas ATP stabilizes the closed state. This model recapitulates many characteristics of the channel biophysics, including its maximum open probability and ATP sensitivity for varying membrane PIP₂ levels and various pore mutations (Enkvetchakul et al., 2000).

In the present study, our objective was to investigate whether a common mechanism would underlay regulation by PIP₂ of voltage-gated and inward rectifier potassium channels, despite their different structural characteristics. We demonstrate that PIP₂, but also MgATP, is necessary for maintaining KCNQ1/KCNE1 complex activity. We show that the effects of PIP₂ on the characteristics of the KCNQ1/KCNE1 currents are fully recapitulated by a model derived from the Kir6.2/SUR1 model, in which regulation by voltage (in the case of KCNQ1 α-subunits) replaces regulation by intracellular ATP (in the case of Kir6.2 α-subunits). The model is also consistent with the interpretation that PIP₂ stabilizes the open state of the KCNQ1/KCNE1 channel. Our results suggest a strong functional homology, conserved between a six transmembrane-domain voltage-gated channel (KCNQ1) and a two transmembrane-domain inward rectifier (Kir6.2).

Results

PIP₂-dependent and -independent rundown of KCNQ1/KCNE1 channels

KCNQ1 and KCNE1 channel proteins were co-expressed in COS-7 cells. We first examined the effect of patch excision on channel activity. Because of the small conductance and the possible clustering of the channels (Yang and Sigworth, 1998), sizeable KCNQ1/KCNE1 currents were recorded in only 15% of the cell-attached patches (from 663 successful patches, 106 demonstrated sizeable currents). Figure 1A illustrates a representative inside-out recording after excision in a divalent ion-free high potassium medium. Complete rundown of the channel activity occurred within 5 min (Figure 1B). This time-frame for rundown was very comparable to that observed for endogenous I_Ks channels in inner ear epithelial cells (Shen and Marcus, 1998). We did not observe any shift in the leak-subtracted reversal potential, suggesting that the K⁺ selectivity was not modified by rundown. Application at the inner face of the membrane of Mg²⁺ alone or Mg²⁺ plus Ca²⁺ accelerated rundown (Figure 1B and C), similar to the observation previously made by Shen and Marcus (1998) with the native current.

Fig. 1. KCNQ1/KCNE1 channel rundown is accelerated by divalent cations and slowed by PIP₂. (A) Representative inside-out recording of KCNQ1/KCNE1 currents at various time after excision in a control solution (145 mM K-gluconate, 10 mM K-HEPES, 1 mM K-EGTA, pH 7.2). The membrane potential was stepped from a holding potential of –80 to +40 mV (1000 ms) and then back to –40 mV (500 ms), every 5 s. Zero current is indicated by a solid line. (B) Average time-dependent currents measured at the end of the depolarizing step relative to their maximum value and plotted against various times after patch excision. Patches were excised in control solution (open squares, n = 9), control solution plus 0.6 mM free Mg²⁺ (open circles, n = 6), control solution plus 1 mM free Mg²⁺ and Ca²⁺ (open triangles, n = 5) or control solution plus 5 µg/ml PIP₂ (filled squares, n = 5). (C) Mean ± SEM values of τ_rundown determined from a monoexponential fit of rundowns in the different conditions presented in (B). *P < 0.001 compared with control. (D) Individual time-dependent currents measured at the end of the depolarizing step relative to their maximum value measured after patch excision. Three individual traces are presented, one with PIP₂ application just before excision, and two with PIP₂ application at 3 or 12 min after excision (arrows).

In order to investigate the effects of PIP₂ on channel rundown, PIP₂ (5 µg/ml) was added to the cytosolic solution just prior patch excision. Addition of PIP₂ markedly slowed but did not entirely suppress the rundown (Figure 1B and C). Strikingly, we observed that when we applied PIP₂ after the patch was excised as illustrated in Figure 1D, the current raised up to a level that coincided with a slower rundown curve uncovering a PIP₂-independent rundown phenomenon. A very slow PIP₂-independent rundown exists in the case of Kir6.2/SUR1 channels in the presence of the same concentration of PIP₂ (Shyng and Nichols, 1998).

Because PIP₂ is known to activate actin polymarization (Yin and Janmey, 2003), we hypothesized that KCNQ1/KCNE1 activity could depend on the assembly/disassembly of the actin cytoskeletal network. To test this hypothesis, we investigated the effects of cytochalasin D on the channel activity in the perforated patch condition. Monitoring KCNQ1/KCNE1 activity during application of 10 µM cytochalasin for 10 min did not show any effect of actin depolymerization when compared with control conditions (n = 4 and 6, respectively; data not shown). This is consistent with the absence of effects of cytochalasin on the amplitude of KCNQ1 currents expressed in Xenopus oocytes (Grunnet et al., 2003). These data suggest that PIP₂ regulation of the channel does not occur through the cytoskeletal network.

PIP₂-dependent rundown, but not PIP₂-independent rundown, is associated with an acceleration in deactivation

In Kir6.2/SUR1 channels, it is well known that variations in membrane PIP₂ levels affect not only the current amplitude, but also the ATP sensitivity of the channel and the single channel kinetics (Enkvetchakul et al., 2000). These modifications can be interpreted as resulting from a unique event, i.e. stabilization by PIP₂ of the open conformation of the ATP-controlled gate. If PIP₂ acts on the KCNQ1/KCNE1 complex through stabilization of the open conformation of the voltage controlled gate, a decrease in PIP₂ levels during rundown should be accompanied by changes in the activation and/or deactivation rates, and by a shift of the voltage dependency. It is known that channels made of KCNQ1 and KCNE1 do not exhibit substantial inactivation (Tristani-Firouzi and Sanguinetti, 1998). PIP₂-related rundown was too fast to permit accurate measurement of the half-maximum activation potential, V_0.5. In contrast, the activation/deactivation kinetics could be convincingly quantified. Current traces were normalized during activation (from –80 to +40 mV) and deactivation (from +40 to –40 mV) in the presence and absence of PIP₂. As shown in Figure 2, there was no alteration in the activation kinetics either during PIP₂-dependent or PIP₂-independent rundown (Figure 2C). In contrast, close examination of the tail current revealed that deactivation was markedly accelerated during PIP₂-dependent rundown but not in presence of PIP₂ (Figure 2D). When PIP₂ was added at different times after excision (as in Figure 1D), deactivation initially accelerated (τ_deact decreased from 564 ± 230 to 131 ± 9 ms) until PIP₂ was applied. At this point, deactivation slowed, only partially reaching pre-rundown values (τ_deact 278 ± 68 ms; n = 3). It is likely that PIP₂-independent rundown prevented recording of PIP₂-induced complete recovery of deactivation.

Fig. 2. PIP₂-dependent, but not PIP₂-independent, rundown is associated with an acceleration of the deactivation. (A and B) Representative normalized inside-out recordings of KCNQ1/KCNE1 currents at various time after excision in the absence (A) or presence (B) of PIP₂ (5 µg/ml). Activating pulse currents measured at different time after excision were normalized to 1 at the end of the depolarizing step, in order to compare activation kinetics (left part of the curve). Deactivating tail currents at different time after excision were normalized to 1 at the beginning of the repolarizing step, so as to compare deactivation kinetics (right part of the curve). (C) Mean ± SEM values of τ_act determined from a fit of the activating current based on the Hodgkin and Huxley model for a voltage-dependent potassium channel activation (cf. Materials and methods). Black bars (0 µg/ml PIP₂) and white bars (5 µg/ml PIP₂) represent the average τ_act as a function of the remaining current during rundown (100, 75, 50 and 25% stand for the ranges 75–100%, 50–75%, 25–50% and 0–25%, respectively (n = 5–16). (D) Mean ± SEM values of τ_deact measured from a monoexponential fit of the deactivating tail current. Black bars (0 µg/ml PIP₂) and white bars (5 µg/ml PIP₂) represent the average τ_deact as a function of the remaining current during rundown (n = 4–16). *P < 0.01.

Altogether, these observations confirm the presence of two types of rundown: a PIP₂-dependent rundown associated with accelerated deactivation kinetics and a PIP₂-independent rundown not associated with alteration in activation or deactivation kinetics.

PIP₂-independent rundown is neither okadaic acid nor calmodulin sensitive

We investigated further PIP₂-independent rundown and tentatively prevented this slow rundown process. It is known that KCNQ1/KCNE1 activity is modulated by phosphorylation (Potet et al., 2001; Marx et al., 2002). Although in our experiments, channels were not expressed with an AKAP (a kinase anchoring protein) to target protein kinase A (PKA) at the vicinity of the channel, a less specific phosphorylation may have occurred. If PIP₂-independent rundown was caused by the activity of a phosphatase, inhibition of this phosphatase should reduce the rundown. Okadaic acid (300 nM), an inhibitor of phosphatase 1 (PP1) and PP2A did not modify either PIP₂-dependent or PIP₂-independent rundown, thereby excluding the involvement of an okadaic acid-sensitive phosphatase (Figure 3A and C). In the yeast two-hybrid assay, different groups, including ours (F.Le Bouffant, I.Baró and D.Escande, unpublished results), found that calmodulin (CaM) interacts with the intracellular C-terminal region of several members of the KCNQ family (KCNQ2, KCNQ3, KCNQ5 and, less clearly, KCNQ1 and KCNQ4; Yus-Najera et al., 2002). In KCNQ2/3 channels, Wen and Levitan (2002) demonstrated a correlation between CaM binding and channel function. Although KCNQ1–CaM interactions were possibly weak, we supposed that PIP₂-independent rundown could be due to dilution of CaM upon excision. Such a dilution of calmodulin has been shown to cause loss of channel inhibition by Ca²⁺ in HERG channels (Schonherr et al., 2000). We observed that application of CaM 500 nM did not reverse PIP₂-independent rundown (Figure 3B and C).

Fig. 3. PIP₂-independent rundown is okadaic acid and calmodulin insensitive. (A) Representative PIP₂-dependent and -independent rundowns of the KCNQ1/KCNE1 current in the presence of 300 nM okadaic acid. The KCNQ1/KCNE1 current is calculated as in Figure 1B and D. The arrow indicates PIP₂ application to the cytosolic side of the patch. The solid line represents the average rundown in presence of PIP₂ in control cells. (B) Representative PIP₂-dependent and -independent rundown of the KCNQ1/KCNE1 current. The first arrow indicates PIP₂ application to the cytosolic side of the patch, the second arrow indicates 500 nM CaM cytosolic application. The solid line represents the average rundown in presence of PIP₂ in control cells. (C) Mean ± SEM values of τ_rundown from a monoexponential fit of rundowns in the different conditions presented in (A) and (B) (n = 5–9). *P < 0.001 compared to control.

Cytosolic application of Mg-ATP prevents PIP₂-independent rundown

MgATP is implicated in the regulation of many ion channels and transporters through a variety of mechanisms (Hilgemann, 1997). We tested the effects of a physiological MgATP concentration on PIP₂-independent rundown. In COS-1 cells, intracellular MgATP concentration has been estimated close to 1 mM (Sippel et al., 1997). Co-application of MgATP (1.4 mM) together with PIP₂ entirely abolished the rundown of KCNQ1/KCNE1 channels (Figure 4A). Application of MgATP alone only slowed the rundown process similar to PIP₂, suggesting that the channels require both PIP₂ and MgATP to remain functional. These regulations involve different mechanisms, since only PIP₂-dependent rundown is associated with an accelerated deactivation.

Fig. 4. The effects of PIP₂ on the voltage-dependency of KCNQ1/KCNE1 currents. (A) Average time-dependent currents measured at the end of the depolarizing step relative to their maximum value measured after patch excision. Patches were excised in control solution (open circles, n = 9), control solution plus 1.4 mM MgATP (0.6 mM free Mg²⁺; filled circles, n = 3), control solution plus 1.4 mM MgATP plus 5 µg/ml PIP₂ (filled squares, n = 9) and control solution plus 5 µg/ml PIP₂ (dashed line, n = 5). Gaps correspond to acquisition interruptions for other protocols recordings. (B) Representative effects of PIP₂ on Kir6.2/SUR1 channels expressed in a COS-7 cell excised patch. PIP₂ was applied as indicated by the arrow. The solid line indicates zero current level. Two hundred micromolar cytosolic ATP repetitive applications are indicated by open boxes. (C and D) Superimposed KCNQ1/KCNE1 currents recorded when membrane potential was stepped, in 10-mV increments, from a holding potential of –80 mV to various potentials between –100 and +60 mV, and then stepped back to –40 mV. Voltage protocols were performed (C) before and (D) 10 min after patch excision in a MgATP + PIP₂-containing solution. Insets, detail of the tail current at –40 mV. Horizontal bar 200 ms, vertical bar 100 pA. (E) Activation curves calculated from the tail current amplitude presented in (C) and (D) before (open circles) and after (filled circles) patch excision in a MgATP + PIP₂-containing solution. (F) Potential for half-maximal activation (V_0.5) calculated from activation curves such as presented in (E) before and at 3 and >5 min after patch excision. *P < 0.05 compared with value at t = 0.

PIP₂ shifts the activation curve of the channel

Prevention by MgATP of PIP₂-independent rundown allowed us to explore the effects of PIP₂ over a longer time of application (tens of minutes are required to study PIP₂ effect on Kir6.2/SUR1 channels; see Figure 4B). In particular, we were able to perform step protocols at different times after patch excision. In COSm6 cells expressing Kir6.2/SUR1 channels, application of 5 µg/ml PIP₂ induced a run-up of the current and a decrease in its ATP sensitivity (Baukrowitz et al., 1998; Shyng and Nichols, 1998), consistent with an increase in membrane PIP₂ above the physiological levels. As a control, we successfully reproduced these experiments (Figure 4B) in which the preparation of PIP₂ has been considered as critical (Larsson et al., 2000). Then, we performed step protocols in the cell-attached configuration (Figure 4C) and at different time after patch excision in a solution containing PIP₂ in excess (5 µg/ml) and MgATP (1.4 mM; Figure 4D). Examination of the tail current demonstrated a progressive shift in the activation curve towards negative potentials (Figure 4E and F) and a concomitant increase in τ_deact (408 ± 55 ms in control versus 552 ± 29 ms with PIP₂; P < 0.05; n = 8 and 5). We thus concluded that prolonged application of PIP₂ above physiological concentration altered the KCNQ1/KCNE1 voltage dependency and slowed deactivation.

A model of PIP₂ regulation of Kir6.2/SUR1 channels applies for KCNQ1/KCNE1

Nichols and colleagues have previously proposed a kinetic model that accounts for many observations made on Kir6.2/SUR1 K_ATP channels. This model was established on the following observations: (i) the ATP dose–response curve of channel inhibition suggested an ATP binding site for each channel Kir6.2 α-subunit; (ii) the mean open time measured on a single channel did not depend on [ATP]; and (iii) some Kir6.2 pore mutants and PIP₂ application decreased the channel ATP sensitivity (Enkvetchakul et al., 2000). We established a similar kinetic model by replacing the regulation of Kir6.2/SUR1 channels by ATP with the regulation of KCNQ1/KCNE1 channels by membrane potential (Figure 5A). The KCNQ1/KCNE1 channel model nicely accounted for our experimental observations, including the effects of PIP₂ on the current amplitude and on the deactivation kinetics with no associated change in the activation kinetics. Like the Kir6.2/SUR1 model, the KCNQ1/KCNE1 model is based on the assumption that PIP₂ acts only by stabilizing the open state of the channels (Enkvetchakul et al., 2000). We adjusted k_PIP2 (corresponding to a decrease in PIP₂ levels; Figure 5B) in order to fit the current amplitude at +40 mV at 1, 2 and 3 min after excision. Adjustment of k_PIP2 constant to rundown data changed the deactivation rate in a way that suitably fitted the experimental deactivation curves (Figure 5C). The KCNQ1/KCNE1 model also showed no change in the activation rate as a function of PIP₂ level, which again was consistent with experimental data. These findings suggest that PIP₂ stabilizes the open state of both a two transmembrane-domain Kir6.2 ATP-gated K⁺ channel and a six transmembrane-domain KCNQ1 voltage-gated channel, further strengthening the idea of similar gating processes among potassium channels (Loussouarn et al., 2002).

Fig. 5. A model based on the stabilization of the open state by PIP₂ recapitulates the characteristics of the KCNQ1/KCNE1 currents. (A) The kinetic scheme presented here is based on previous models of Kir6.2/SUR1 channel regulation by PIP₂. This simple model is based on the assumption that PIP₂ does not affect the voltage sensor (k_S4, k′_S4), but only a closed state to an open state transition when the four voltage sensors are in the permissive state. Hence k_S4, k′_S4 and k′ are PIP₂-independent and only k_PIP2 varies during the simulated rundown. k_S4= 3.56/s during activation and is negligible during deactivation; k′_S4 = 7.47/s during deactivation and is negligible during activation; and k′ = 87.3/s. (B) Current traces from Figure 1A, to which the leak current was subtracted, were superimposed with the simulated current (solid lines). k_PIP2 was fixed to 592.74, 176.43, 25.84 and 4.53/s to best fit the decrease in current amplitude during rundown, as shown in the inset. Inset: simulated (circles) and observed current (solid line) amplitudes as a function of time after patch excision. (C) Traces in (B) were normalized to compare the observed and simulated kinetics of activation and deactivation.

Decreased I_Ks current during metabolic poisoning is related to MgATP regulation

In guinea pig cardiomyocytes recorded in the whole-cell configuration, uncoupling of oxidative phosphorylation with dinitrophenol leads not only to activation of the K_ATP channel, but also to a decreased activity of a delayed rectifier potassium current (Escande, 1989). I_Ks is a major component of the delayed rectifier in guinea pig myocytes (Zicha et al., 2003) and may correspond to the component sensitive to metabolic inhibition. We recorded I_Ks current in guinea pig cardiac myocytes submitted to dinitrophenol challenge. We initially checked that under our experimental conditions, brief applications of 10 µM dinitrophenol (∼30 s) reversibly activated ATP-sensitive K⁺ current (not illustrated). Further experiments were then performed in the presence of glibenclamide to prevent K_ATP channel activation and also in the presence of E4031 to block the fast component of the delayed rectifier I_Kr. Within the same time-frame as required for K_ATP channel activation, dinitrophenol led to a reversible inhibition of I_Ks by 40.2 ± 9.5% (n = 6). As illustrated in Figure 6, tail current deactivation kinetics were not modified by metabolic poisoning (τ_deact = 179.4 ± 24.9 ms in dinitrophenol versus 202 ± 24.9 ms in control; n = 6 and 5, respectively). During metabolic inhibition, MgATP but also PIP₂ decreases (Loussouarn et al., 2001). The absence of change in deactivation kinetics in cardiac myocytes suggests that metabolic poisoning affects I_Ks through decreased MgATP concentration independently of PIP₂ levels.

Fig. 6. Metabolic poisoning of cardiomyocytes leads to rundown of I_Ks. (A) The membrane potential was stepped from a holding potential of –40 to +40 mV (3 s) and then back to –40 mV (1 s), every 4.5 s. Zero current is indicated by a solid line. Cells were perfused permanently with glibenclamide (3 µM), E4031 (0.3 µM) and CoCl₂ (1.8 mM) to inhibit I_KATP, I_Kr and I_ca,L, respectively. Addition of dinitrophenol (DNP; 10 µM) is followed by a decrease in I_Ks. Note the variation of the steady state current at –40 mV, suggesting that another channel may be affected by DNP as well, such as the inward rectifier I_K1. (B) Normalization of the tail current of I_Ks shows no changes in deactivation kinetics.

Discussion

Dual rundown suggests that two factors are required for functional KCNQ1/KCNE1 channels

The investigation of the rundown process upon patch excision has previously contributed to the identification of many physiological mechanisms involved in ion channels and transporters regulation, including regulation of ion channels by phosphorylation (Becq, 1996). In the present work, we show that KCNQ1/KCNE1 channel activity depends on two distinct intracellular factors. We first unveiled the requirement of PIP₂ to maintain channel activity. In addition, intracellular PIP₂ application uncovered a slower rundown. PIP₂-sensitive rundown and PIP₂-insensitive (but MgATP-sensitive) rundown can easily be distinguished, since only PIP₂-dependent rundown is associated with changes in the channel deactivation kinetics. PIP₂-independent channel rundown could reflect many processes, including the dilution of a cytosolic activator of the channel such as CaM (Schonherr et al., 2000), the dephosphorylation of the channel (Sansom et al., 1997) or an alteration of the cytoskeleton (Furukawa et al., 1996). Intracellular MgATP abolishes PIP₂-independent rundown, which was not affected either by inhibition of the phosphatases usually implicated in channel rundown or by calmodulin application. Although these observations narrow the array of potential mechanisms implicated in KCNQ1/KCNE1 activity, it remains possible that cytoplasmic MgATP participates in many processes, as reviewed by Hilgemann (1997). Additional experiments are required to identify the role of MgATP in maintaining the channel activity.

The bell-shaped effect of intracellular Mg²⁺ on cardiac I_Ks current

In guinea pig cardiac cells, whole-cell patch–clamp recordings of the I_Ks current carried by KCNQ1/KCNE1 showed a bell-shaped dependence on intracellular Mg²⁺ concentration (Hirahara et al., 1998). With 0.03–1 mM [Mg²⁺]_i, I_Ks is relatively stable upon patch excision, but decreased rapidly when [Mg²⁺]_i is below or above this range. These observations can be re-interpreted in the light of the dual regulation of the channel by MgATP and PIP₂. In our experiments, KCNQ1/KCNE1 currents were stable when the intracellular solution contained 5 µg/ml PIP₂ and 1.4 mM MgATP (∼0.6 mM free Mg²⁺, which is in the 0.03–1 mM range). Removing MgATP (Figure 1B, filled squares) or PIP₂ (Figure 4A, filled circles), but also screening PIP₂ by divalent ion application, decreased the KCNQ1/KCNE1 current. Mg²⁺ ion depletion may depress the cardiac I_Ks current through a depletion in MgATP. Conversely, high intracellular Mg²⁺ may depress the cardiac I_Ks current through the screening of PIP₂ negative charges by divalent cations.

Potential physiological significance of KCNQ1/KCNE1 channel regulation by PIP₂

Recent studies have revealed the implication of PIP₂ in the physiological regulation of ion channels and transporters activity. For example, inhibition of K_ATP channels by α1-adrenoceptor stimulation in rat myocytes (Haruna et al., 2002) and inhibition of voltage-gated calcium channel by LHRH in bullfrog neurons (Wu et al., 2002) are mediated by PIP₂. The requirement of PIP₂ for KCNQ/KCNE1 activity raises the possibility of physiological regulation of the channel complex by receptor-coupled phospholipase C (PLC). In neurons, inhibition of the KCNQ2/KCNQ3 complex by a muscarinic receptor implies PLC and is reversed by PIP₂ (Suh and Hille, 2002; Zhang et al., 2003). PLC-coupled receptors are diverse including α1-adrenergic, endothelin or angiotensin II receptors. An orphan receptor APJ is also likely coupled to a PLC (Szokodi et al., 2002). Further studies are required to establish a correlation between PLC and KCNQ1/KCNE1 activity. In vestibular dark cells for example, evidence was provided that was consistent with a direct effect of the PKC branch of the PLC pathway on the I_Ks current in response to activation of the apical P2U receptor (Marcus et al., 1997).

Homologous effect of PIP₂ on a two and a six transmembrane-domain potassium channel

All the data gathered on the effects of PIP₂ on Kir6.2/SUR1 channels can be interpreted by a stabilization of the channel in the open conformation (Enkvetchakul et al., 2000). Similarly, our data can also be interpreted by a model based on the stabilization by PIP₂ of the open conformation of KCNQ1/KCNE1 channels. This model is also comparable to Shaker channels kinetic models, i.e. one or more transitions of the four subunits followed by one or more additional concerted transitions (Schoppa and Sigworth, 1998). In the model that we used, one transition of the four subunits and one concerted transition were sufficient to convincingly fit the data. The model illustrates how variations in PIP₂ levels are accompanied by changes in deactivation with no changes in activation kinetics. The requirement of the four domains to be in the ‘on’ state to allow the channel to finally open makes this transition slow and rate limiting. Since this step is PIP₂-independent (k_S4 is constant and k′_S4 negligible), there are no effects of PIP₂ on activation kinetics. On the contrary, deactivation strongly depends on the value of k_PIP2: when PIP₂ levels are elevated, k_PIP2 is much higher than k′_S4 (592.74 and 7.47/s, respectively), the frequent backwards transitions to the open state slow down deactivation; when PIP₂ levels are low, k_PIP2 is much lower, and the channels directly close without backwards transition to the open state.

Similarities between the Kir6.2/SUR1 and KCNQ1/KCNE1 channel models suggest homologous effects of PIP₂ on a six transmembrane-domain, voltage-dependent potassium channel and a two transmembrane-domain, ATP-dependent channel. This homology strengthens the idea that, despite their rather different molecular structure, potassium (and more generally cation) channels share a common fundamental pore architecture (Lu et al., 2001; Loussouarn et al., 2002). This functional homology may give some insight on the nature of PIP₂ regulation of KCNQ1/KCNE1 channels. Direct interaction of PIP₂ with the C-terminus of several inwardly rectifying K (ROMK, GIRK and IRK) channels has already been shown (Huang et al., 1998). An alanine scan (Shyng et al., 2000; Cukras et al., 2002) of both the C- and N-terminus of Kir6.2 revealed that neutralization of 16 positively charged amino acids (of a total of 38) generally in the near C- and N-termini of Kir6.2 changes the PIP₂ sensitivity of the channel, suggesting that they belong to a potential PIP₂-binding domain and may stabilize the channel open state through electrostatic interaction with membrane PIP₂ (Enkvetchakul et al., 2000). This potential PIP₂-binding domain may share the structure of the recently crystallized cytoplasmic pore of GIRK1 (Nishida and MacKinnon, 2002). KCNQ1 presents 17 positively charged amino acids in the N-terminus and 49 in the C-terminus. Identification of the residues implicated in PIP₂ interaction would require a large alanine scan of the channel protein. In favor of the existence of a PIP₂ binding domain in KCNQ1 is the clustering of positive charges at the bottom of M2 and S6 domains of Kir6.2 and KCNQ1, respectively (Figure 7). In Kir6.2, neutralization of these positively charged residues alters the channel PIP₂ sensitivity (Shyng et al., 2000), suggesting that they interact electrostatically with PIP₂. The alignment presented in Figure 7 strongly suggests a similar role for the positive amino acids located at the bottom of the S6 domain of KCNQ1, which resembles the M2 domain in Kir6.2 (Lu et al., 2001; Loussouarn et al., 2002). Interestingly, in KCNQ2/3, neutralization of one histidine in the same region (aligned with the KCNQ1 histidine boxed in Figure 7) leads also to decreased PIP₂ sensitivity (Zhang et al., 2003), suggesting that the cluster of positive charges in KCNQ1 may also interact with PIP₂ to stabilize the open state. Much more experiments, including an alanine scan of the 66 positive residues in the C- and N-terminus, are required to clarify the structural homology behind the functional homology.

Fig. 7. Alignment of the pore and M2/S6 domains of KCNQ1 with Kir6.2 showing the similar clustering of positive charges at the bottom of the putative pore-lining domains S6 (for KCNQ1) and M2 (for Kir6.2). The boxed histidine is aligned with the KCNQ2 histidine that decreases PIP₂ affinity when neutralized (see text).

Materials and methods

Cell isolation, cell culture and cell transfection with KCNQ1/KCNE1 and Kir6.2/SUR1

Single cells were obtained from the heart of guinea pigs by the use of an enzymatic dissociation method as described previously (Baró and Escande, 1989). The concentration of the enzyme used was adjusted to give an activity of 350 U/ml for collagenase (Type I; Sigma) and 0.61 U/ml for protease (Type XIV; Sigma).

The African green monkey kidney-derived cell line COS-7 was obtained from the American Type Culture Collection (Rockville, MD) and cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal calf serum and antibiotics (100 IU/ml penicillin and 100 µg/ml streptomycin; all from Gibco, Paisley, UK) at 37°C in a humidified incubator. They were subcultured regularly by enzymatic treatment. Cells were transfected when the culture reaches 60–80% confluence, with the plasmids (2 µg per ml of culture medium) complexed with JetPEI (Polyplus-transfection, Strasbourg, France), according to the standard protocol recommended by the manufacturer. The human KCNQ1 and KCNE1 cDNAs were subcloned into the mammalian expression vector, pCI and pCR, respectively (Promega, Madison, WI) under the control of a cytomegalovirus enhancer/promoter. The mouse pCMV-Kir6.2 (Inagaki et al., 1995) and the hamster pECE-SUR1 clones were kindly provided by S.Seino and C.G.Nichols, respectively. The plasmid coding for the green fluorescent protein (GFP), used to identify transfected cells was purchased from Clontech. For KCNQ1/KCNE1 current measurement, relative DNA composition was 45% KCNQ1/45% KCNE1/10% GFP. For Kir6.2/SUR1 current measurement, relative DNA composition was 30% Kir6.2/30% SUR1/40% GFP.

Electrophysiology

Twenty-four to 72 h post-transfection, COS-7 cells were mounted on the stage of an inverted microscope and constantly perfused with the Tyrode solution at a rate of 2 ml/min. A microperfusion system allowed local application and rapid change of the different experimental solutions. The bath temperature was maintained at 22.0 ± 1.0°C. Micropipettes (tip resistance 0.8–1.5 MΩ) were pulled from thin-walled glass (Kimble; Vineland, NJ) on a vertical puller (P30; Sutter Instruments, Co., Novato, CA), fire polished, and electrically connected to a patch–clamp amplifier (RK-300; Biologic Science Intruments, Claix, France). Stimulation, data recording and analyses were performed by Acquis1, a software made by Gérard Sadoc (distributed by Biologic Science Intruments) through an analog-to-digital converter (Tecmar TM100 Labmaster; Scientific Solution, Solon, OH, USA). Rundown of KCNQ1/KCNE1 currents as well are activation and deactivation kinetics were studied with a protocol consisting of depolarized voltage steps from a holding potential (–80 mV) to +40 mV (1 s) and then back to –40 mV (500 ms), every 5 s. To better fit the S-shaped activation current, we used the Hodgkin and Huxley model for a voltage-dependent potassium channel activation: I_k = [1 – exp(–t/τ_act)]⁴ × I_max, with τ_act the activation constant, t the time and I_max the potassium current amplitude when the channels are fully open (Hille, 2001). KCNQ1/KCNE1 deactivation kinetics were obtained by a monoexponential fit. For the half activation potential (V_0.5) calculation, the membrane potential was stepped, in 10-mV increments, from a holding potential (–80 mV) to various voltages (pulse potential) between –100 and +60 mV, and then stepped back to –40 mV, where tail currents are visible. The activation curve was obtained from extrapolation of the tail current to the start of the voltage step to avoid contamination with the capacitive current and fitted to a Boltzmann distribution. Kir6.2/SUR1 currents were measured at a membrane potential of –50 mV (pipette voltage = +50 mV). Inward currents at this voltage are shown as positive-going signals.

Whole-cell patch–clamping of cardiomyocytes was described previously (Baró and Escande, 1989). Isolated myocytes were constantly perfused with Tyrode solution maintained at 35 ± 1.0°C. Micropipettes were pulled narrower than in inside-out experiments to reduce intracellular MgATP dilution in the pipette. They had a higher tip resistance (1.5–4 MΩ) when filled with the intracellular solution. Patch–clamp experiments are presented as the mean ± SEM. Statistical significance of the observed effects was assessed by the Student’s t-test. Off-line analysis was performed using Acquis1 and Microsoft Excel programs. Microsoft Solver was used to fit data by a least-squares algorithm.

Kinetic model of PIP₂ regulation of KCNE1/KCNQ1

Simulations of KCNQ1/KCNE1 currents during rundown were generated from the model presented in Figure 5A. KCNQ1/KCNE1 currents during a depolarizing step are based on the assumption that all the S4 domains are in the ‘off’ conformation (C_{s4 off} = 1, C_{s4 on} = 0 and O = 0) at the beginning of the depolarizing step. KCNQ1/KCNE1 tail currents during repolarization are calculated from the Po value obtained at the end of the depolarizing step. Only k_PIP2 is PIP₂ dependent and its values during the simulated rundown were chosen to best fit the decrease in the maximal current amplitude at 40 mV during rundown (cf. legend of Figure 5). Microsoft Excel was used to perform the currents simulation.

Solutions and drugs

The pipette (extracellular) solution used in experiments on transfected COS-7 cells had the following composition: 145 mM Na-gluconate, 4 mM K-gluconate, 4 mM Mg_1/2-gluconate (1 mM free Mg²⁺), 7 mM Ca_1/2-gluconate (1 mM free Ca²⁺), 5 mM K-HEPES, 1 mM K-EGTA, pH 7.4 with NaOH. The standard microperfusion (intracellular) solution contained 145 mM K-gluconate, 10 mM K-HEPES, 1 mM K-EGTA, pH 7.2 with NaOH, with additions as described. The pipette (intracellular) solution used in experiments on cardiomyocytes cells had the following composition (in mM): K-gluconate 120, KCl 15, HEPES 5, EGTA 5. The local perfusion contained (in mM): NaCl 132, KCl 4, HEPES 10, mannitol 20 and glucose 5, MgCl₂ 1.2, CoCl₂ 1.8 and E4031 0.3 µM, glibenclamide 3 µM. Metabolic inhibition was performed by replacing glucose by 10 µM dinitrophenol. Free activities were calculated using a software designed by G.L.Smith (University of Glasgow, Glasgow, UK). PIP₂ (Roche Molecular Biochemical) was sonicated in ice for 30 min before application to inside-out patches. Okadaic acid and CaM (Calbiochem) containing solution were prepared from stock solution in water stored at –20°C (300 and 50 µM, respectively). All other products were purchased from Sigma.