Activity of KCNQ (Kv7) channels requires binding of phosphatidylinositol 4,5-bisphosphate (PIP2) from the plasma membrane. We give evidence that Mg2+ and polyamines weaken the KCNQ channel–phospholipid interaction. Lowering internal Mg2+ augmented inward and outward KCNQ currents symmetrically, and raising Mg2+ reduced currents symmetrically. Polyvalent organic cations added to the pipette solution had similar effects. Their potency sequence followed the number of positive charges: putrescine (+2) < spermidine (+3) < spermine (+4) < neomycin (+6) < polylysine (≫+6). The inhibitory effects of Mg2+ were reversible with sequential whole-cell patching. Internal tetraethylammonium ion (TEA) gave classical voltage-dependent block of the pore with changes of the time course of K+ currents. The effect of polyvalent cations was simpler, symmetric, and without changes of current time course. Overexpression of phosphatidylinositol 4-phosphate 5-kinase Iγ to accelerate synthesis of PIP2 attenuated the sensitivity to polyvalent cations. We suggest that Mg2+ and other polycations reduce the currents by electrostatic binding to the negative charges of PIP2, competitively reducing the amount of free PIP2 available for interaction with channels. The dose–response curves could be modeled by a competition model that reduces the pool of free PIP2. This mechanism is likely to modulate many other PIP2-dependent ion channels and cellular processes.
Phosphatidylinositol 4,5-bisphosphate (PIP2) is a minor but essential signaling lipid that is needed for many plasma membrane processes, including the function of ion channels and transporters, control of endocytosis, and the nucleation of cytoskeletal networks (Hilgemann et al., 2001; Suh and Hille, 2005). This paper describes the effect of intracellular Mg2+ on the availability of PIP2 for ion channels, as revealed by using KCNQ channels as an indicator. M-type K+ channels (KCNQ, Kv7) are low-threshold voltage-activated channels important in the physiology and pathophysiology of neurons, heart, and epithelia (Brown, 1988; Delmas and Brown, 2005; Lerche et al., 2005). They control the membrane potential and cell excitability and can be turned off by activating receptors that couple to PLC (Marrion et al., 1989; Bernheim et al., 1992; Jones et al., 1995; Delmas and Brown, 2005). The principal mechanism of this regulation is that the channels require PIP2 to function and PLC depletes the essential PIP2 from the plasma membrane (Suh and Hille, 2002; Zhang et al., 2003; Ford et al., 2004; Winks et al., 2005; Horowitz et al., 2005; Suh et al., 2006). Hence the magnitude of the current in KCNQ channels tracks the availability of PIP2 in the plasma membrane.
Magnesium ion, the most abundant intracellular alkaline earth cation, interacts with polyvalent anions, notably with organic polyphosphates such as ATP and PIP2. It is a necessary cofactor for protein and lipid kinases, phosphatases, and ATPases; magnesium ion helps stabilize the organization of negatively charged DNA, RNA, and membranes (Alvarez-Leefmans et al., 1987). Although the total cellular Mg2+ content is typically >10 mM, most of this is bound to nucleotides and nucleic acids. Mammalian cells are believed to have resting free Mg2+ concentrations of 0.5–1.0 mM (Gupta et al., 1984; White and Hartzell, 1989; London, 1991; Romani and Scarpa, 2000). Changes in intracellular Mg2+ observed in response to specific stimuli are typically considered small and slow, but the lack of good Mg2+ indicators might hide more interesting behavior. Some Mg2+-permeable ion channels TRPM6 and TRPM7 contribute to homeostasis of intracellular Mg2+ (Schlingmann et al., 2002; Walder et al., 2002; Schmitz et al., 2003; Voets et al., 2004). There are reports that free [Mg2+]i can rise several millimolar during the activation of some receptors (Brocard et al., 1993; Takaya et al., 1998) or as a consequence of intracellular Ca2+ concentration increases (Murphy et al., 1989a; Gotoh et al., 1999); and Mg2+ can be released from mitochondria when their membrane potential falls (Kubota et al., 2005). Cytoplasmic Mg2+ influences the activity of ion channels in several ways. Notably, intracellular Mg2+ and organic polycations block outward current in some K+ channels by voltage-dependent binding within the inner vestibule of the pore (Vandenberg, 1987; Lu and MacKinnon, 1994; Voets et al., 2003; Obukhov and Nowycky, 2005; Zhang et al., 2006). This mechanism underlies fast inward rectification. In addition, intracellular Mg2+ exerts “slow” inhibitory effects in KCNQ1/KCNE1 channels, TRPV5 and TRPM7 channels, and Kir2.3 (IRK3) channels (Chuang et al., 1997; Shen and Marcus, 1998; Nadler et al., 2001; Loussouarn et al., 2003; Du et al., 2004; Lee et al., 2005). Each of these is a PIP2-requiring ion channel. Thus physiological effects of Mg2+ on membrane excitability merit deeper study.
Intracellular Mg2+ has several known roles in the receptor-mediated modulation of KCNQ channels. First, submillimolar Mg2+ is needed for onset and termination of G-protein signaling, where it is involved in the conformational changes leading to G-protein subunit dissociation and in the GTPase step leading to deactivation of Gα subunits (Gilman, 1987; Suh et al., 2004). Millimolar Mg2+ is required for the phosphoinositide kinases that restore PIP2 pools and hence mediate M-current recovery (Yamakawa and Takenawa, 1988; Suzuki et al., 1991; Downing et al., 1996). Therefore cytoplasmic Mg2+ can be a limiting factor for both inhibition and recovery of KCNQ channels when the receptor is stimulated (Suh et al., 2004). In addition, one must consider the possibility of the fast block and the slow inhibition already mentioned for other channels. Here we show that internal Mg2+ depresses KCNQ currents. We argue that internal Mg2+ and other polyvalent cations regulate KCNQ channel activity by reducing the availability of PIP2 for binding to the channel. This may correspond to the slow inhibitory mechanism reported in other channels.
Materials And Methods
Cell Culture and Transfection
Transformed human embryonic kidney tsA-201 (tsA) cells were cultured and transiently transfected using Lipofectamine 2000 (Invitrogen) with various cDNAs (Suh et al., 2004) including mouse M1-muscarinic receptor (1 μg, from N. Nathanson, University of Washington, Seattle, WA), the channel subunits human KCNQ2 and rat KCNQ3 (Kv7.2 and Kv7.3; 1 μg, from D. McKinnon, State University of New York, Stony Brook, NY), and when necessary, GFP (0.1 μg) as a marker for transfection. In some experiments with confocal microscopy, we monitored PIP2 and its cleavage products by transfecting with fluorescent translocation probes, either PH-PLCδ1-EGFP (EGFP-PHPLC, 0.25 μg, from P. De Camilli, HHMI, Yale University, New Haven, CT), which binds to PIP2 and IP3, or PKC-C1a-EGFP (GFP-C1PKC, 0.25 μg, from T. Meyer, Stanford University, Stanford, CA), which binds to diacylglycerol. tsA cells were maintained in DMEM (Invitrogen) supplemented with 10% FCS and 0.2% penicillin/streptomycin.
Reagents and Bathing Solutions
The muscarinic receptor agonist oxotremorine-M was applied at 10 μM. Chemicals were purchased from Sigma-Aldrich. We used 30,000–70,000 MW poly-l-lysine (Sigma-Aldrich). The external Ringer's solution used for confocal microscopy and for current recording contained (in mM) 160 NaCl, 2.5 KCl, 2 CaCl2, 1 MgCl2, 10 HEPES, and 8 glucose, adjusted to pH 7.4 with NaOH. The high K+ Ringer's solution contained (in mM) 132.5 NaCl, 30 KCl, 2 CaCl2, 1 MgCl2, 10 HEPES, and 8 glucose, adjusted to pH 7.4 with NaOH.
Confocal Imaging and Analysis
tsA cells were imaged 24–36 h after transfection on polylysine-coated coverslips (Suh et al., 2004). Images were taken every 5 s on a Leica TCS/MP confocal microscope at room temperature and processed with Metamorph (UIC) and Igor Pro 4.0 (Wavemetrics, Inc.). To obtain averaged time courses, the fluorescence intensity F over a given region of the cytoplasm or nucleus was usually normalized to the average baseline intensity for the 30 s before agonist application F0 (F/F0). Images are shown here in negative contrast so that the brightest fluorescence appears black.
tsA cells were whole-cell clamped at room temperature 24–48 h after transfection (Suh et al., 2004). The bath solutions exchanged in ∼6 s. The standard pipette solution contained (in mM) 175 KCl, 5 MgCl2, 5 HEPES, 0.1 BAPTA, 3 Na2ATP, 0.1 Na3GTP, pH 7.4, adjusted with KOH. This solution had a free Mg2+ concentration of 2.08 mM (estimated with the WinMaxC program). To vary the free Mg2+, other pipette solutions contained 1, 3, and 10 mM added MgCl2 or 1 mM EDTA and no added MgCl2 but 0.65 mM CaCl2. Pipettes were pulled to a resistance of 1.8–2.3 MΩ, and experiments were not continued if the series resistance was >3 MΩ. Series resistance was 60% compensated. Recordings were performed using a HEKA EPC-9 amplifier with Pulse software (HEKA Elektronik). Data analysis was done using Igor Pro 4.0. Cells were held at a holding potential of −20 mV, and 500-ms test pulses to −60 mV were given every 4 s. Reported values (I−20 mV) are mean outward current taken at −20 mV just before each step to −60 mV. Recordings started 3–5 min after breakthrough.
Student's unpaired t test (two tailed), or, when indicated, a one-way ANOVA was used to test for significance. Where error bars are shown, they represent SEM.
KCNQ Current Is Sensitive to Intracellular Mg2+ and Polyamines
We studied KCNQ K+ currents using whole-cell recording from tsA cells transfected with KCNQ2/KCNQ3 channel subunits. Fig. 1 A shows families of outward currents obtained with depolarizing voltage steps from −70 mV at 10-mV intervals in a standard Ringer's bathing solution containing 2.6 mM K+. The insets show currents elicited by the classic M-current deactivation voltage protocol with the membrane held at −20 mV, where KCNQ channels are open and pass a steady noninactivating current, and then stepped to −60 mV where channels deactivate, and back to −20 mV. The phenomenon we wanted to study is apparent when comparing experiments where the pipette solution contained the 5 mM control value of total Mg2+ (left) with those where the pipette contained EDTA and no added Mg2+ (right). Initially the KCNQ currents are the same in these exemplar cells, but 3 min after breakthrough, the currents have decreased with the 5 mM Mg2+ pipette solution, whereas they increased with the Mg2+-free (EDTA) pipette solution. These changes in amplitude were not accompanied by shifts in the voltage dependence of channel activation (midpoint −28.3 ± 0.7 mV with 5 mM Mg2+ vs. −27.6 ± 0.5 mV with EDTA, n = 5) or in the rates of channel deactivation or activation (Fig. 1 A, insets). Similar Mg2+ dependence was seen in cells treated with the PLC inhibitor U73122 or with the phosphoinositide (PI) 4-kinase inhibitor wortmannin (unpublished data), implying that the effect was not due to alterations of PIP2 metabolism by PLC or lipid kinases. The effect developed within a minute or two after breakthrough (Fig. 1 B), approximately the time expected for dialysis of ions from the pipette into the cytosol.
In this paper we usually give the total Mg2+ in the pipette solutions. Much of that becomes bound to the 3 mM ATP present. Thus, the 5 mM Mg2+ pipette solution has a calculated free Mg2+ concentration of only 2.08 mM. Plotting the relative change of current amplitude against the calculated free Mg2+ concentration (Fig. 1 C) reveals a graded dependence on free Mg2+ that crosses the null point at 0.47 mM free Mg2+. This is near the range of internal free Mg2+ estimated for resting cells (Gupta et al., 1984; White and Hartzell, 1989; London, 1991; Romani and Scarpa, 2000). The effect of high Mg2+ could be mimicked by other polyvalent cations. Fig. 1 D shows that the addition of 1 mM of the aminoglycoside antibiotic neomycin to the pipette solution almost completely inhibited the current within 5 min. Neomycin has a maximum charge of +6 at low pH and a charge of +4.5 at neutral pH (McLaughlin and Whitaker, 1988). The polyamines spermine (+4) and putrescine (+2) also inhibited but with lower potency. Dose–response experiments give 50% inhibitory concentrations (IC50 values) of 0.25, 0.86, and 2.4 mM for inhibition by neomycin, spermine, and putrescine, respectively, in the presence of the 2.08 mM free Mg2+ already in standard pipette solution (Fig. 1 E). The potency increases as the positive ionic valence increases.
Amplitude Changes of KCNQ Current Are Not Voltage Dependent
Previous studies have shown that Mg2+ and other polyvalent cations block some ion channels in a fast voltage-dependent manner by entering into the inner or outer mouth of the pore. For example, this is the principal mechanism of physiological inward rectification in inwardly rectifying K+ channels (Vandenberg, 1987; Lopatin et al., 1994). There, internal polyvalent cations are drawn into the inner mouth of the pore at positive potentials, blocking outward flow of K+, and are expelled from the inner mouth back into the cytoplasm at negative potentials, allowing inward flow of K+. Characteristic of such block is its fast rectifying nature. Therefore we asked whether Mg2+ blocks current in KCNQ channels in this way by comparing the block of inward and outward K+ currents. Increasing extracellular K+ concentration to 30 mM raised the K+ equilibrium potential EK from −108 to −45 mV, so we could observe inward currents negative to −45 mV. The tail currents at −70 mV were inward (Fig. 2 A). Elevating the K+ concentration did not change the voltage dependence of activation (Fig. 2 B; midpoints, −26.7 ± 0.5 mV in 2.6 mM K+, −27.7 ± 1.6 mV in 30 mM K+) or the time course of activation and deactivation (Fig. 2 C). Inward and outward currents could still be suppressed by addition of the muscarinic agonist oxotremorine-M (Oxo-M), and they recovered upon removal of Oxo-M (Fig. 2 D). Finally, the time course of muscarinic suppression of the inward and outward currents in high-K+ solution (Fig. 2 E) was very much like the time course of suppression in standard, low-K+ Ringer's solution (compare Suh et al., 2004), both with standard pipette solution and with the EDTA pipette solution. As we have reported before (Suh et al., 2004), muscarinic suppression of KCNQ current was extremely slow and persistent when Mg2+ was removed with internal EDTA, an effect we have attributed to the Mg2+ requirement for the G-protein cycle. Thus elevated K+ and changes of direction of K+ flow leave the gating and the muscarinic modulation of KCNQ channels unaltered.
Now we could test for any current rectification due to Mg2+. Fig. 3 (A and B) shows that including 10 mM Mg2+ in the pipette decreased inward and outward KCNQ currents symmetrically, and removing Mg2+ (Mg2+-free EDTA) increased them symmetrically. There is no sign of voltage-dependent block or block that is dependent on the direction of current. Polylysine, a high molecular weight polybasic amine often used to sequester PIP2, also reduced KCNQ currents symmetrically. Polylysine produces a local positive electrostatic potential when it binds to the membrane and strongly attracts and effectively sequesters multivalent PIP2 (Wang et al., 2004). Similarly the polyamines neomycin, spermine, spermidine, and putrescine inhibited currents symmetrically (Fig. 3, C and D). These results suggest that polyvalent cations, including Mg2+, do not decrease the KCNQ current by a block of the channel pore from inside the cell but instead possibly act as polylysine is assumed to act, i.e., by binding to PIP2 electrostatically, thus reducing the availability of PIP2 for direct interaction with the channels.
Mg2+ Effects Are Reversible
The actions of Mg2+ were readily reversible. Fig. 4 shows experiments where the same cell was patched twice by pipettes containing different Mg2+ concentrations. In Fig. 4 A, the first pipette with a high concentration of Mg2+ decreased the current and the second pipette with Mg2+-free EDTA led to fast enhancement of current. In Fig. 4 B, the order is reversed with the same observation of full reversibility. These reversible changes of KCNQ current also occurred in cells treated with wortmannin (inside and outside) (Fig. 4 C), a treatment that would block the recovery of KCNQ current from muscarinic suppression by inhibiting the enzyme type III PI 4-kinase (Fig. 4 D; see also Suh and Hille, 2002; Zhang et al., 2003; Ford et al., 2004; Winks et al., 2005). That result suggests that the changes of current do not reflect degradation or synthesis of PIP2 in the plasma membrane.
Block with Internal Mg2+ Differs from Block with Internal TEA
Many K+ channels are blocked by TEA, which binds reversibly to independent sites in the outer and inner mouths of the pore (Armstrong, 1966; Hille, 1967; MacKinnon and Yellen, 1990; Holmgren et al., 1997; Hadley et al., 2000). When 30 mM TEA was included in the extracellular, high-K+ bath solution, inward and outward KCNQ currents were symmetrically reduced ∼50% without changes in the rates of activation or deactivation of the current (Fig. 5 A, also see inset). On the other hand, when the cells were dialyzed internally with TEA, there were changes both in current waveform and in rectification. With a low pipette concentration of TEA (1 mM), the asymmetry was not so obvious (Fig. 5 B) but the changes in kinetics were clear (Fig. 5 C). Deactivation of the current appeared greatly slowed whereas activation appeared speeded. For comparison, we tried the KCNQ channel blocker linopirdine. It blocked outward and inward KCNQ current symmetrically when applied in the bath, and it had no action on amplitude or time course of current when applied inside (Fig. 5, D–F). Experiments with XE991 also showed that the KCNQ current was blocked only when the drug was applied outside (unpublished data). Of all the blockers we tried, only internal TEA slowed deactivation and speeded activation (Fig. 5, C and G), whereas polylysine, EDTA, high Mg2+, putrescine, spermidine, and linopirdine did not (Fig. 5 G).
The action of 10 mM intracellular TEA was like the classical, open-channel block described in squid giant axons by Armstrong (1966). There the current-dependent kinetics of blocker entering and leaving the activated channel adds another transition to the gating scheme and distorts the current waveform. Fig. 6 shows that after 3 min of TEA dialysis, the inward tail current had developed a pronounced “hook” at −70 mV. The small initial size of tail current reflects the amount of channel block developed during the preceding depolarization, and the rising phase of the hook is understood as a time-dependent unblock of channels at −70 mV followed by normal deactivation gating of the unoccupied channel. The hook is larger after a depolarization to +40 mV than it is after a depolarization to −20 mV, showing that a larger fraction of channels is plugged by a TEA ion during the more positive step. Such hooks were not induced by 10 mM Mg2+ or polylysine in the pipette, which appear to reduce KCNQ currents by a different mechanism.
Raising Membrane PIP2 Prevents the Modulation of KCNQ Current by Internal Mg2+
Since intracellular Mg2+ ion can bind to the negative phosphates of PIP2 (Hendrickson and Fullington, 1965; Toner et al., 1988), it might be reducing KCNQ current by making free PIP2 less available. We asked if we could overcome the sensitivity to Mg2 by augmenting PIP2 production. Half the cells were transiently transfected with the enzyme phosphatidylinositide 4-phosphate (PIP) 5-kinase Iγ (PIPKIγ; Aikawa and Martin, 2003; see Winks et al., 2005; Suh et al., 2006), which converts PIP to PIP2. Two groups of cells also were cotransfected with a fluorescent translocation probe, either GFP-PHPLC or GFP-C1PKC for study with confocal microscopy. In control cells (no PIPKIγ), the GFP-PHPLC probe, which binds PIP2 and IP3, sits at the plasma membrane (Fig. 7 A, top), whereas the GFP-C1PKC probe, which binds to diacylglycerol, remains in the cytoplasm (no diacylglycerol) (Fig. 7 B, top). Upon activation of PLC in control cells, the GFP-PHPLC probe translocates from plasma membrane to cytoplasm, indicating significant depletion of membrane PIP2 and production of cytoplasmic IP3. The GFP-C1PKC probe translocates from cytoplasm to membrane, indicating production of diacylglycerol (GFP-C1PKC) in the plasma membrane. Cells overexpressing both PIPKIγ and GFP-PHPLC showed strong fluorescence at the plasma membrane at rest, as expected if PIP2 is high there. However, they also showed a few very intense regions of fluorescence in the cytoplasm, suggesting formation of abnormal intracellular pools of PIP2 by overexpressed PIPKIγ (Fig. 7 A, bottom). When cells transfected with PIPKIγ were treated with Oxo-M, the translocation of CFP-C1PKC was normal, whereas the translocation of the GFP-PHPLC probe was only 30% of the control amount (Fig. 7, A and C). This suggested that PLC was activated and hydrolyzing PIP2 to produce diacylglycerol normally, but because the supply of PIP2 was so greatly augmented, the PLC reaction was not fast enough to deplete all of it (Fig. 7, B and C). From our modeling, we believe that the PIP2 pool must have been increased many-fold by PIPKIγ (see later). Similarly modulation of KCNQ current by Oxo-M was retarded and reduced to 40% in PIPKIγ-overexpressing cells (Fig. 7 D), again implying that the PIP2 pool was too large to be fully depleted by PLC. In other observations, PIPKIγ increased the open probability of KCNQ channels in the following ways: it shifted the voltage dependence of activation by 10 mV to more negative potentials (Fig. 7, E and F), speeded the time course of activation, and slowed deactivation of channels (Fig. 7 G).
Overexpression of PIPKIγ profoundly reduced the sensitivity of KCNQ current to changes of internal Mg2+. Neither the 10 mM Mg2+ pipette solution nor the Mg2+-free EDTA pipette solution had much effect on current (Fig. 8, A–C). In addition, PIPKIγ overexpression diminished the current inhibition by neomycin (Fig. 8, D and E). However, it did not diminish the rectifying nature of block by intracellular TEA (Fig. 8 F). Apparently current regulation by Mg2+, polyamines, and PLC are attenuated by overproduction of PIP2, whereas pore block by TEA was not changed.
We found that KCNQ2/KCNQ3 current is sensitive to the cytoplasmic free Mg2+ concentration. Current rises when free Mg2+ is lowered and falls when free Mg2+ is raised. The null point is near 0.47 mM free Mg2+. On average the KCNQ2/3 currents in a resting cell are only 56% of what can be achieved by removing Mg2+. A lack of voltage dependence of the current depression suggests that the mechanism is not a plugging of the pore from the inside. Mimicry by many polycationic amines and independence from inhibition of PI 4-kinase or PLC suggest that the mechanism does not involve changes of metabolism, especially not through Mg2+-dependent enzymes nor by changing the concentration of essential Mg2+–ATP complexes. Virtual elimination of the Mg2+ and polyvalent cation sensitivity by raising the synthesis of PIP2 and other arguments given later favor a mechanism with Mg2+ and other polyamines interacting electrostatically with PIP2 to reduce its availability.
As an aside to the principal theme of this paper, we also observed that intracellular TEA blocks KCNQ current by a mechanism that is distinct from the inhibition by polyvalent cations. The block by TEA follows all the rules of open-channel block as first described by Armstrong (1966) for squid delayed rectifier K+ channels. Presumably this means that KNCQ channels have a hydrophobic inner vestibule that shares many features with other K+ channels of the KV family. On the other hand, linopirdine and XE991 blocked only from the outside.
Comparison with Previous Work
Intracellular Mg2+ is reported to reduce currents in many channels. We focus first on several well-studied cases where the channels are PIP2 requiring and the mechanism is clearly distinct from the fast, voltage-dependent, strongly rectifying pore block found in many channels (Nowak et al., 1984; Vandenberg, 1987; Lu and MacKinnon, 1994; Voets et al., 2003; Obukhov and Nowycky, 2005; Zhang et al., 2006).
(a) An early example was the study by Chuang et al. (1997) on Kir2.3 (IRK3) channels expressed in Xenopus oocytes. The parallel with our KCNQ results is striking. Like us, they reported that Mg2+ slowly reduced current and EDTA increased current. From single-channel analysis they called the Mg2+ state an inactivated state. They determined a null point of 0.5 mM free Mg2+, which they suggested would be the normal resting concentration in the oocyte. They also showed that activating M1 muscarinic receptors inhibited Kir2.3 currents in a phenomenologically similar way to Mg2+. Finally they proposed that activating M1 receptors raises free Mg2+ as a novel second messenger acting on the channel, although no receptor mechanism that raises free Mg2+ was known. Subsequent work showed that activating M1 receptors would close Kir2.3 channels by depleting PIP2 from these PIP2-requiring channels (Du et al., 2004), providing a different explanation for the effects of receptor activation. That study also showed that the sensitivity to Mg2+ was highest in versions of the channel that had the lowest PIP2 affinity and proposed that elevated Mg2+ might inhibit these channels by stimulating lipid phosphatases to deplete PIP2. However, the finding that multivalent organic cations mimic the Mg2+ effects in several examples means that phosphatases cannot be the general explanation for “slow” channel inhibition by polycations.
(b) Lee et al. (2005) studied slow reversible inhibition of PIP2-dependent TRPV5 channels expressed in CHO cells. Pipette Mg2+ inhibited current with an IC50 of 0.29 mM free Mg2+ in whole-cell recording. With excised patches, addition of PIP2 enhanced the current and greatly diminished the sensitivity to Mg2+, whereas allowing depletion of PIP2 reduced the current and increased the sensitivity to Mg2+. In addition they found a fast, voltage-dependent block of the pore by Mg2+. They suggested that the fast block involves Mg2+ binding to an aspartic acid in the channel, and that removal of PIP2 could favor a slow conformational change of this Mg2+-bound channel to a more persistent inhibited state.
(c) Endogenous TRPM7 channels in RBL cells are known to be PIP2 dependent (Runnels et al., 2002) and Mg2+ sensitive (Nadler et al., 2001; Kozak and Cahalan, 2003). Kozak et al. (2005) found that the slow inhibition by Mg2+ could be mimicked by other divalent and trivalent metal cations and by all the polyvalent amine cations that we tested. These cations did not induce fast voltage-dependent pore block, whereas internal TEA did. They hypothesized that Mg2+ might act by electrostatic screening of PIP2. This hypothesis is very close to the one we adopt below.
(d) Finally, we mention two studies on KCNQ1/KCNE1 (IsK/KvLQT1) channels, whose suppression by activation of M1 muscarinic receptors (Selyanko et al., 2000) suggests they have a PIP2 requirement. Adding Mg2+ to the cytoplasmic side of an excised membrane patch accelerates rundown of KCNQ1/KCNE1 currents from native inner ear cells (Shen and Marcus, 1998) and expression systems (Loussouarn et al., 2003). This Mg2+ effect was considered not due to endogenous Mg2+-dependent protein phosphatases or kinases because it was readily reversible and repeatable even while the membrane patch was bathed in a simple salt solution lacking Mg-ATP and enzymes.
Is Mg2+ Interacting with PIP2?
We now consider a sequence of arguments for the hypothesis that Mg2+ depresses KCNQ current by reducing the availability of polyanionic PIP2. The PIP2 requirement for KCNQ channel function seems absolute (Suh and Hille, 2002; Zhang et al., 2003; Ford et al., 2004; Horowitz et al., 2005; Winks et al., 2005; Suh et al., 2006). When PIP2 is removed there is no current. The Mg2+ ion readily forms complexes with all phosphates, especially polyphosphates. Thus the dissociation constant for the complex with inorganic
Does this interaction make PIP2 less available for interaction with PIP2-dependent proteins? We have described studies in four other channels, in addition to our work on KCNQ2/3, showing similar depression of currents by divalent metals and by a specific group of polyvalent organic cations. This similar pattern suggests that their Mg2+ sensitivity results from some common feature, which we suggest is their PIP2 dependence. In each of the channels where it has been studied, clusters of basic residues in the C terminus of the channel are presumed to interact electrostatically with the three phosphate groups of PIP2. Removing positive charges from these clusters on the channel decreases the PIP2 affinity (Shyng et al. 2000; Du et al., 2004; Rohacs et al., 2005; Nilius et al., 2006). Adding competing polyvalent cations will diminish such electrostatic interaction between channel and PIP2 by lowering the negative charge of the lipid. Since the true geometry of PIP2 binding is not known for any channel, this last conclusion remains model dependent. In any case we have now shown by overexpression of the lipid kinase PIP5KIγ that the sensitivity to multivalent cations can be overcome by raising the cellular PIP2 levels, as would be expected if the cations acted by reducing the fraction of PIP2 available.
A confounding factor is that even if Mg2+ does bind to PIP2, it necessarily interacts with many other acidic metabolites and acidic residues of proteins as well. This shows up in our assays of KCNQ current as additional components of change that we have not emphasized in this paper. For example, in Suh et al. (2004) we studied the effect of intracellular Mg2+ on the kinetics of muscarinic inhibition of KCNQ currents. We found that very low free Mg2+ severely slowed the onset of muscarinic inhibition and the recovery after inhibition (as in e.g., Fig. 2 E and Fig. 8 B of this paper). Elevated Mg2+ did not have large effects on inhibition. These effects were successfully described in a kinetic model for receptor-mediated inhibition in terms of known high-affinity (20 μM) Mg2+ requirements for the activation of G proteins, hydrolysis of GTP by G proteins, and phosphorylation and dephosphorylation of phosphoinositides by lipid kinases and phosphatases (Suh et al., 2004). While doing those experiments, we also discovered the magnesium effects elaborated in the present paper. Therefore the 2004 measurements with Oxo-M were made only after 5 min of whole-cell dialysis when the amplitude changes of KCNQ current (Fig. 1 B) were nearly complete. In a higher-order analysis, the phenomena of the 2004 paper and of this paper probably do interact. The Mg2+ complexes with PIP2 (and PIP) probably change the ease of hydrolysis by PLC and of phosphorylation and dephosphorylation by kinases and phosphatases and thus could affect the rates of muscarinic inhibition and recovery. Further, the direct requirement for Mg2+ of lipid kinases and phosphatases probably establishes new set points for the sizes of phosphoinositide pools during the experiments of this paper, giving additional slow components of relaxation of current amplitudes. For that reason, in this paper we emphasized changes that occur within 1–3 min and have not pointed to additional slower changes that might occur with long recording.
Most papers we have cited concerning an inhibition by Mg2+ that is not pore block, have called this “slow block.” Indeed it develops over a couple of minutes during whole-cell dialysis. This time course is distinct from the fast pore block that can be induced sometimes in the same channels by for example quick steps of membrane potential. Nevertheless, the slowness may primarily reflect the speed of dialysis by diffusion from the pipette. The model we describe below attributes the block to formation of Mg2+ complexes with PIP2, a process that could be intrinsically fast unless it also has to wait for dissociation of bound PIP2 from a large reservoir of macromolecular complexes.
An Electrostatic Model
If the depression of current is as simple as forming a Mg2+–PIP2 complex, we should be able to write a simple equilibrium model. Such a model is described mathematically in the Appendix. We describe the logic here. The most elementary postulates would be that (a) Mg2+ and KCNQ are competing ligands for free PIP2 in the membrane, (b) the KCNQ–PIP2 complex can conduct current, and (c) the uncomplexed KCNQ does not conduct current. That model gives the required reversible loss of current at high Mg2+, but the current would turn off much more sharply as Mg2+ rises than is observed. Fig. 1 C shows that the effect of Mg2+ develops over four decades of concentration. To rectify this discrepancy we assumed that PIP2 binds one Mg2+ at low Mg2+ concentration, and a second Mg2+ at higher concentration (Fig. 9). That generates two complexes with Mg2+ that develop over a wider concentration range. Further, this model needs to allow the PIP2 complex with one Mg2+ to bind to KCNQ channels also, but with a lower affinity than the free PIP2 does. The line drawn in Fig. 1 C is the output of this model using constants given in Table I in the Appendix.
The next step is to consider the polyvalent amines. The dose–response curves for neomycin, spermine, and putrescine are much steeper than that for Mg2+, which suggests that binding of one polyatomic, polyvalent amine suffices to “hide” a molecule of PIP2. Therefore the model was extended to allow the polyvalent organic cations to compete with the first Mg2+ for free PIP2 (Fig. 9 A). The Mg2+ and channel-related assumptions remain as determined before in fitting Fig. 1 C, and Mg2+ was included in the calculation. This model generates the three curves in Fig. 1 E. The success of the simple model shows that the concept of reduced PIP2 availability in elevated Mg2+ and polyamine solutions can provide a quantitative explanation of the observations. We should point out that the null point determined here and in several published studies by patching with different free Mg2+ concentrations in the pipette (Fig. 1 C) may slightly overestimate the resting free Mg2+ of cytoplasm. This is because the starting condition (the resting cell before dialysis) includes not only the endogenous Mg2+ but also a significant concentration of endogenous organic polyamines (spermine and spermidine) that are normally in cytoplasm.
Finally we consider the effect of overexpressing the lipid 5-kinase PIPKIγ, which greatly reduced the sensitivity to polyvalent cations. Presumably the total PIP2 of those cells rose considerably as suggested in previous work (Suh et al., 2006; Winks et al., 2005) and by the reduced muscarinic suppression of M-current and of PH-domain translocation (Fig. 7, C and D). In our model, raising the total PIP2 for example 15-fold reduces the effect of 10 mM Mg2+ pipettes from a 67% inhibition to a 21% inhibition, and of a 1 mM neomycin pipette, from an 80% inhibition to a 44% inhibition (see Fig. 10). Similarly, the effect of an EDTA pipette is reduced from a 78% enhancement to a 7% enhancement. These calculated effects are in qualitative agreement with the measurements in Fig. 8.
Overexpressing the lipid 5-kinase also retarded and greatly reduced the ability of Oxo-M to suppress KCNQ current and to cause translocation of the PH-domain probe (Fig. 2 E and Fig. 8 B; Suh et al., 2006; Winks et al., 2005). Qualitatively, one might anticipate difficulty in depleting PIP2 below the thresholds for channel activation and PH-domain binding when the rate of PIP2 production is speeded up by an order of magnitude; however, a simulation with the complete kinetic model we have used previously for simulating PIP2 depletion by PLC (Suh et al., 2004; Horowitz et al., 2005) predicts much less slowing and reduction than is observed experimentally. This is partly because the modeled PLC reaction is fast and first order. When PIP2 is elevated, PLC simply becomes proportionately faster. Significant improvement in the simulation is gained by assuming that the PLC reaction becomes half saturated at the normal resting PIP2 concentration so the speeding up at higher concentration is quite limited. Further improvements can be realized by assuming “buffering” of PIP2 concentrations from the significant abnormal intracellular pools of PIP2 reported by the PH-domain probe, up-regulation of PI 4-kinase, or down-regulation of lipid phosphatases and PLC. Tests of these ideas are beyond the scope of this work.
Intracellular free Mg2+ Concentration Changes in Normal and Pathological Conditions
Changes in the concentration of intracellular ionized Mg2+ on cell physiology are not extensively studied and are difficult to document for the lack of a good Mg2+ indicator. Much intracellular Mg2+ is bound, so free Mg2+ levels represent only 2–5% of the total cellular Mg2+ in resting cells. Therefore, given the high concentration of metabolic intermediates or nucleotides, including Mg-ATP, in cells, it can be expected that conditions that substantially lower those levels, such as strong exercise, would also release free Mg2+. For example, during ischemia in rat cardiac myocytes, free Mg2+ increases from 0.6–0.7 mM to 2.1–2.3 mM in parallel with the decline in ATP levels (Murphy et al., 1989b; Headrick and Willis, 1991). Possibly the cytoplasmic concentration of endogenous amines (spermine and spermidine) is subject to regulation as well. Raising their concentration would free some Mg2+ by competition at polyphosphates. Intracellular free Mg2+ also increases during transient cytoplasmic acidification (Freudenrich et al., 1992) and in hypertension (Ebel and Gunther, 2005). Whether intracellular Mg2+ concentration changes are used in physiological signaling is not clear, but activation of several hormonal receptors or of protein kinases with phorbol esters can alter Mg2+ influx and hence the level of free Mg2+ (Elliott and Rizack, 1974; Erdos and Maguire, 1983; Grubbs and Maguire, 1986; Maguire, 1987). Additionally the buffering capacity for Mg2+ may be altered by changes in metabolic state or signaling, such as local release of Ca2+ from the internal stores or transient change of intracellular pH, thereby altering the free Mg2+ concentration either globally or locally (Flatman, 1991; Murphy et al., 1991). Thus intracellular free Mg2+, integrating the signals from hormone activity, cellular metabolism, and systemic ion homeostasis, could potentially regulate many Mg2+-requiring enzymes and substrates as well as the free PIP2 in cells.
We find that intracellular Mg2+ and polyamines inhibit KCNQ2/3, a PIP2-requiring ion channel. Similar “slow” inhibition by polyvalent cations is seen in several other PIP2-requiring ion channels. We find that elevated membrane PIP2 decreases the sensitivity of KCNQ to inhibition by these cations. Conversely it is reported that partial depletion of PIP2 increases sensitivity to inhibition by Mg2+ (Lee et al., 2005). Other channels, including most inward rectifier potassium (Kir) channels, TRP channels, ENaC channels, HCN channels, and probably Ca2+ channels, share a PIP2 requirement (Fan and Makielski, 1997; Kobrinsky et al., 2000; Lei et al., 2001; Gamper et al., 2004; Delmas et al., 2005; Suh and Hille, 2005; Pian et al., 2006). We suggest that all PIP2-requiring cellular functions, including ion channels, transporters, cytoskeletal regulators, and membrane traffic components (Hilgemann et al., 2001; Suh and Hille, 2005), should exhibit sensitivity to Mg2+ and other polyvalent cations. This sensitivity will be less obvious for proteins that have a high affinity for PIP2, which might remain fully PIP2 bound even when available PIP2 declines. It will be most obvious for proteins like KCNQ2/3 that have a low PIP2 affinity and are only partly saturated. Such proteins may be subject to physiological regulation in conditions that change cellular polyvalent cation concentrations, and when studied in vitro, their properties including apparent PIP2 affinity will depend on the polyvalent cation composition of the test solutions. Care should be taken to use solutions that match cellular Mg2+ and polycation levels in quantitative studies of any PIP2-dependent phenomena.
This Appendix describes the mathematical models for (a) the equilibrium binding of Mg2+ and other polycations to PIP2 and (b) the equilibrium binding of KCNQ channel subunits to PIP2. The rationale for the models is given in the main text.
Cation Binding to PIP2
The scheme for binding of cations to PIP2 is drawn in Fig. 9 A. It shows four states of PIP2: free, complexed with one Mg2+, complexed with two Mg2+, and complexed with a polyamine of valence z+. The equilibrium constants K indicated by each arrow are dissociation constants in units of molar. We use conventional equations for multiple simultaneous equilibrium binding. If a is the total PIP2 concentration relative to that in a standard cell, and Mg and Amz+ are the local concentrations of the Mg2+ and polyamine ligands in the vicinity of the PIP2, the solution of this equilibrium for a standard cell is given by: a = 1; b = a × Mg/KPIP2.Mg; c = b × Mg/KPIP2.2Mg; d = a × Amz+/KPIP2.Amz+; D = a + b + c + d; PIP2 = a/D; PIP2.Mg = b/D; PIP2.2Mg = c/D; PIP2.Amz+ = d/D. When a = 1, i.e., with a standard resting amount of total PIP2, the calculated values for PIP2, PIP2.Mg, PIP2.2Mg, and PIP2.Amz+ are numerically equal to the fraction of PIP2 in each of the four forms. When neither Mg2+ nor organic polyamines are present, the free PIP2 would then be PIP2 = 1.
The model also assumes that there is a local negative potential ψ at the cation binding site. This local potential raises the local cationic ligand concentration by an equilibrium Boltzmann factor that depends on the valence of the cation: Mg = [Mg2+]bulkexp(−2qeψ/kT) and Amz+ = [Amz+]bulkexp(−zqeψ/kT), where qe, k, and T are the charge of an electron, Boltzmann's constant, and the absolute temperature, respectively. This assumption of a local negativity is not required to make a workable model, but it was invoked because the region around a PIP2, with its three phosphate groups, will certainly be negative, and it gives a natural explanation for why the apparent affinity of polyvalent cations increases with the charge of the cation.
Table I shows values of constants that predict the concentration dependence for Mg2+ binding illustrated in Fig. 9 C. There is a 150-fold separation of the two Mg2+ binding steps on the concentration axis. The three sets of dissociation constants appropriate for different local potentials give the same binding curves.
Binding of PIP2 Species to KCNQ Subunits
The purpose of the model was to simulate the depression of KCNQ current by the addition of cytoplasmic polyvalent cations. Allowing both free PIP2 and singly occupied PIP2.Mg to bind to channels as diagramed in Fig. 9 B allowed us to achieve the broad dose–response curve seen empirically (Fig. 1 C). The equations take the same format as those for PIP2–polycation interaction: f = 1; g = f × PIP2/KKCNQ.PIP2; h = f × PIP2.Mg/KKCNQ.PIP2.Mg; D = f + g + h; KCNQ = f/D; KCNQ.PIP2.Mg = g/D; KCNQ.PIP2.2Mg = h/D. The dissociation constants were chosen as KKCNQ.PIP2 = 0.2 and KKCNQ.PIP2.Mg = 1.0. These numbers are dimensionless, where 0.2 means that the channel would be half saturated when 20% of the standard PIP2 is free. The broad, predicted dose–response curve for Mg2+ experiments is drawn as a solid line in Fig. 1 C and as a dashed line in Fig. 9 C. The predicted dose–response curves for experiments with neomycin, spermine, and putrescine are drawn as solid lines in Fig. 1 E. In this latter calculation, the concentration of individual polyamines is varied in the constant presence of 2 mM free Mg2+ to mimic the conditions of the actual experiments. The cations were assumed to be fully ionized at the local pH near the binding site, which would be a little more acidic than that of the cytoplasmic solution when there is a negative surface potential.
Effect of Elevated PIP2
The simulation of the effect of overexpressing PIPKIγ described in text was done by elevating the PIP2 concentration from the standard value of 1 to 15 in the binding model. Fig. 10 shows that the dose–response curve for Mg2+ acting on current becomes nearly flat.
Discussion of the Binding Model
This model is simplified in several respects. One simplification is that the polyvalent ions are assumed only to form chemical complexes with PIP2 rather than to both screen surface negative charge of the membrane and bind to PIP2 in the sense of the Gouy-Chapman-Stern double-layer model used by Toner et al. (1988). If the local potential is small, the double-layer screening effects would not be strong; however to model large potentials such as −40 mV, a full Gouy-Chapman-Stern treatment would be proper. The assumption that up to 2 Mg2+ ions can bind per PIP2 is not a new one. It was used by Hendrickson and Fullington (1965) and Toner et al. (1988). Our model gives quite different dissociation constants depending on the assumed local potential. The two intrinsic Mg2+ dissociation constants in the model of Toner et al. are on the order of 2 and 10 mM at their zeta potential of −38 mV. These considerations point to how model dependent the values of such parameters are. The success of our model shows that direct binding of polyvalent cations to PIP2 can explain the depression of K+ currents, but a range of formulations and specific parameter values probably would work.
A second simplification is that the current is directly proportional to the fraction of channels with a bound PIP2 or PIP2.Mg. Presumably in fact the four subunits of KCNQ channels each can bind one or several molecules of phosphoinositide, and the activity of channel tetramers might be more complicated than just a linear function of occupancy.
A third simplification is that the cation-binding scheme of Fig. 9 A is solved first and then the channel-binding scheme of Fig. 9 B is solved assuming that it does not perturb the population of cation-binding states. This is equivalent to saying that the number of channels is much less than the number of PIP2 molecules. The kinetic model we have used in the past for PIP2 metabolism assumes that there are 5,000 PIP2 molecules per square micrometer in a standard cell at rest. As an order-of-magnitude calculation, we note that the highest current density observed in our experiments (100 pA/pF) is equivalent to 1 pA/μm2. Only one or two open channels per square micrometer is needed for that current.
We thank Lea Miller for technical assistance, Greg Martin in the Keck Center for Neural Imaging for help with microscopy, and Drs. Y. Aikawa and T.F. Martin for the PIPKIγ construct.
This work was supported by National Institutes of Health grant NS08174 from the National Institute of Neurological Disorders and Stroke.
Edward N. Pugh Jr. served as editor.
Abbreviations used in this paper: IP3, inositol 1,4,5-trisphosphate; PIP, phosphatidylinositol 4-phosphate; PIP2, phosphatidylinositol 4,5-bisphosphate; Oxo-M, oxotremorine-M; TEA, tetraethylammonium ion; PIPKIγ, phosphatidylinositol 4-phosphate 5-kinase Iγ.