Anionic phospholipids activate ATP-sensitive potassium channels.

The ATP-sensitive potassium channel (KATP) controls insulin release in pancreatic β-cells and also modulates important functions in other cell types. In this study we report that anionic phospholipids activated KATP in pancreatic β-cells, cardiac myocytes, skeletal muscle cells, and a cloned KATP composed of two subunits (SUR/Kir6.2) stably expressed in a mammalian cell line. The effectiveness was proportional to the number of negative charges on the head group of the anionic phospholipid. Screening negative charges with polyvalent cations antagonized the effect. Enzymatic treatment with phospholipases that reduced charge on the lipids also reduced or eliminated the effect. These results suggest that intact phospholipids with negative charges are the critical requirement for activation of KATP, in distinction from the usual cell signaling pathway through phospholipids that requires cleavage. Mutations of two positively charged amino acid residues at the C terminus of Kir6.2 accelerated loss of channel activity and reduced the activating effects of phospholipids, suggesting involvement of this region in the activation. Metabolism of anionic phospholipids in plasmalemmal membrane may be a novel and general mechanism for regulation of KATP and perhaps other ion channels in the family of inward rectifiers.

KATP, however, also depends on another less well characterized regulatory process, because KATP gradually inactivates when the integrity of the cell is disrupted by, for example, the excised patch method of voltage clamp (11)(12)(13). This property is shared not only by most KATP channels from different tissues, but also by other channels of the inward rectifier potassium (Kir) superfamily (for reviews, see Refs. 14 and 15). Despite intense investigation, the mechanisms for maintaining activation of KATP have not been fully identified and characterized, although previous work has focused on protein phosphorylation processes on the channel or related subunit (for a review, see Ref. 16). We found that anionic phospholipids, especially phosphoinositides, are necessary and sufficient to activate and maintain KATP activity in native and cloned KATP and in two other Kir channels. We also propose and test a novel hypothesis involving the cytoplasmic tail of the pore-forming subunit to account for this regulation. Modulation of channel activity through the composition and phosphorylation of membrane phospholipids may represent an important regulation for KATP in particular, and for channels in the Kir superfamily in general.

EXPERIMENTAL PROCEDURES
Cell Isolations-Single pancreatic ␤-cells were enzymatically (0.05% trypsin) dispersed from isolated islets of Wistar rats collected by centrifugation through a discontinuous gradient of Ficoll (17). Single ventricular myocytes were isolated from rabbit hearts by an enzymatic (0.1% collagenase, Williton) dissociation method (17,18). Single skeletal muscle cells were dissociated enzymatically (0.3% collagenase, Williton) from the flexor digitorum brevis muscles of the hind legs of Wistar rats (13). Isolated single skeletal muscle cells were suspended in Dulbecco's modified Eagle's medium until they were used. Membrane vesicles were formed by washing the isolated cells with high K ϩ solution.
Solutions and Chemicals-The pipette solution (extracellular side, 140-K ϩ -pipette) contained (in mM) KCl (140.0), CaCl 2  , and the pH was adjusted to 7.4. K 2 -ATP (Sigma) was dissolved into the bath solution just before use, and the pH of the solution was readjusted. DNase I and cytochalasin B were products of Sigma. Polyvalent cations were prepared in stock solutions and added to the solution prior to use. The concentration of free cations was calculated (19). A 2-ml bath was perfused with a solution of dispersed lipids at concentrations of 0.005-1 mg/ml. PPIs (a mixture of PI(4,5)P 2 , PI(4)P (15-20%), PI, PS, and PC, from bovine brain; Sigma) were used. Other lipids were as follows: PI(4,5)P 2 (Boehringer Mannheim), PI(4)P (sodium salt, from bovine brain; Sigma), PI (from bovine liver; Avanti Polar Lipids), PS (from brain; Avanti Polar Lipids), PC, PE (from bovine heart; Avanti Polar Lipids), IP 3 (potassium salt, from bovine brain; Sigma), DOG (8:0 DG), and DPG (16:0 DG; Sigma). Anionic phospholipids were dispersed in the solution with 30-min sonication (half set-ting) on ice. The following uncharged lipids were dissolved first in dimethylsulfoxide and then sonicated for 30 min: PLC-PI (from Bacillus cereus; Sigma), PLC (Type IV, from Bacillus cereus; Sigma), PLA 2 (Sigma), and PLD (Sigma). All enzymes were dissolved directly into reaction solution. Experiments were done at room temperature (21-24°C).
cDNA Expression and Mutagenesis-The wild type recombinant complementary DNAs encoding the rat SUR (rSUR), and rat Kir6.2 (rKir6.2) were cloned from a cDNA library of insulinoma beta cell tumor line, and a rat islet cDNA library (7), respectively. A polymerase chain reaction-based site-directed mutagenesis kit (ExSite, Stratagene) was used to generate the site-directed point mutations, which were confirmed by sequencing. The cDNA of SUR (in a pcDNA3 vector, Invitrogen) and wild type Kir6.2 or its mutants (in a pCR3 vector, Invitrogen) were co-transfected and heterologously expressed in mammalian cells either transiently (tsA201 cell line (20)) or stably (a human embryonic kidney cell line, HEK) using a calcium-phosphatide transfection kit (Life Technologies, Inc.). The transfected cells were cultured in Dulbecco's modified Eagle's medium for 12 h with the precipitated cDNA. The transiently transfected cells were kept in culture for another 24 -48 h before being used in electrophysiological experiments. Stably transfected cells were selected by growth in media containing 400 -600 g/ml active G418 (Life Technologies, Inc.) for at least 1 week, and single colonies surviving were isolated and grown to harvest in the continuous pressure of selection with 200 -300 g/ml active G418.
Electrophysiological Recordings and Data Analysis-Currents were recorded using inside-out patch clamp (cytoplasmic membrane exposed to the bath) or whole cell clamp (external membrane exposed to bath) through a patch clamp amplifier (Axopatch 1-D or Axopatch 200A, Axon Instruments) and filtered through a built-in low-pass filter at 1 kHz, unless indicated otherwise. Leak current was compensated electronically on-line or subtracted off-line. Data were acquired by digitizing at 2 kHz and analyzed by a pClamp 6.0 (Axon Instruments) and graphic plotting software running on a PC-compatible computer (Gateway 2000).

RESULTS
Currents of native KATPs were recorded from inside-out patches of pancreatic ␤-cells (Fig. 1A), cardiac ventricular myocytes (Fig. 1B), and sarcomeric membrane vesicles from skeletal muscle cells (Fig. 1C). KATP current expressed in HEK cells stably transfected by cloned cDNA of rSUR and rKir6.2 was also studied (Fig. 1D). Open channel activity was increased within 1-2 s to a maximal value after excising the patch membrane into an ATP-free solution; this was followed by decreased channel activity for pancreatic cells, cardiac cells, and transfected HEK cells. In the vesicles from skeletal muscle cells, the patch was initially devoid of KATP activity in most cases (Fig.  1C), presumably because these channels had inactivated during the vesicle formation process. An anionic phospholipid-rich tissue extraction containing PPIs (0.05-1 mg/ml) consistently activated KATP in cells from all three tissues and also the cloned KATP. Maximal effects were reached within 30 s to 2 min and did not change with time if the perfusion of the lipid-containing solution was maintained. After wash-out of PPIs from the bath solution, the activation effect remained for a time but diminished slowly with a time course that depended upon the previous concentration of PPIs. PPIs had no effects on single channel current amplitudes or ATP sensitivity. A non-KATP inwardly rectifying K ϩ channel, presumably the cardiac inward rectifier, or IRK1, noted coincidentally in Fig. 1B (small conductance channel observed before excision) also lost activity and was then activated by PPIs (small conductance apparent after ATP suppressed KATP). In separate experiments, PPIs activated currents expressed from clones of mouse inwardly rectifying K ϩ channels mIRK1 and the ATP-regulated inward rectifier hROMK1 (Fig. 2). The preparation of PPIs was a mixture of PI(4,5)P 2 , PI(4)P, PI, PS, and PC; to evaluate the contribution of the individual components to the activation effect, we individually tested purified PI(4,5)P 2 (5 negative charges), PI(4)P (3 negative charges), PI (1 negative charge), PS (1 negative charge), PC (neutral), and another neutral phospholipid, PE, using the same protocol. All negatively charged phospholipids tested activated KATP; Fig. 3A shows an example for PI(4,5)P 2 , and results for all lipids tested are summarized in Fig. 3D. Similar to PPIs, the activation effect also remained after washing the lipids from the bath solution. The most effective anionic phospholipid was PI(4,5)P 2 , and others were less effective. The effectiveness sequence correlated with the number of negative charges in the head groups, suggesting a role for electrostatic force. Uncharged PC (1 mg/ml) had no effect on KATP activity, while uncharged PE (1 mg/ml) induced a slight depression of KATP activity. In intact cells, phospholipase C hydrolyzes phospho-lipids such as PI(4,5)P 2 into the cascade products, soluble IP 3 and membrane-delimited diacylglycerol, which as important second messengers, may affect channel function. Although IP 3 has 6 negative charges/molecule, it had no effect on KATP activity (0.1 mg/ml) (Fig. 3B). The diacylglycerols (DPG and DOG) actually reduced activity of KATP (results summarized in Fig. 3D). This indicates that an intact unit of a hydrophobic tail and a charged hydrophilic head are a basic structural requirement for the anionic phospholipids to activate KATPs. Other cascade products of phospholipids, arachidonates, had inconsistent effects on KATPs in cardiac cells (21) and insulinoma cells (22). The effect of intact phospholipids to activate KATP was also tested by using phospholipase-treated lipids (Fig. 3C). PLC-PI, PLC, or PLA 2 was incubated with PI(4,5)P 2 , PI(4)P, and PS. The enzyme-treated lipids were much less potent than the untreated lipids (Fig. 3E). Interestingly, PLD treatment of PC increased KATP activity (Fig. 3E). PLD adds NP open measured after the inactivation and just before the application of lipids was taken as a control. Control mean value is indicated by a dotted line in both D and E. Most data are from cardiac cells, with some data from pancreatic ␤-cells included in the effects of PPIs, PI(4,5)P 2 , PI, PS, and PE. Data are expressed as mean Ϯ S.D., with numbers from 3 to 5 for each point except for the control (n ϭ 67), and for DOG, PI(4)P plus PLC-PI, and PC plus PLD (n ϭ 1). Experimental conditions were the same as described in the Fig. 1 legend, except in some cardiac cells EGTA was included. The lipids were sonicated immediately before the experiment. The lipids were sonicated again after incubation with phospholipase and before the experiments. an anion to the head group of neutral PC, making this observation consistent with the effectiveness sequence depending upon net negative charge. Finally, exogenous PLA 2 decreased KATP activity in inside-out patches, consistent with a recent report (23). This effect of PLA 2 was reversed by subsequent application of PPIs.
We further tested the hypothesis that the negative charges in the anionic heads of the anionic phospholipids are the functional group critical to the activation of KATP by screening those charges with polyvalent cations. As an example (Fig. 4A), the polyvalent cation antibiotic neomycin decreased KATP activity, and PPIs could reverse and antagonize this effect. Divalent cations such as Ca 2ϩ and Mg 2ϩ are already known to accelerate the decrease of KATP activity (24,25) by an unknown mechanism. We found that trivalent cations such as Gd 3ϩ and La 3ϩ (ED 50 ϭ 70 nM and 0.3 M, respectively; ED 50 defined as 50% inhibition at 1 min) were even more potent inhibitors of KATPs than divalent cations (Fig. 4B). Other polyvalent cations including aminoglycoside antibiotics and polyamines used as negative charge chelators for phospholipids also have the similar effect. Neomycin (ED 50 ϭ 20 M), gen-tamicin (ED 50 ϭ 60 M), and spermine (ED 50 ϭ 0.94 mM) inhibited KATP activity irreversibly (Fig. 4B). A decreased potency for the bigger molecules may reflect the influence of the molecular sizes in chelating negative charges of the phospholipids for these polyvalent cations. Application of exogenous PPIs activated KATPs previously inhibited by polyvalent cations, and application of polyvalent cations with PPIs reduced the effect of PPIs.
The effects of phospholipids on KATP were rapid, sustained, and general to all anionic phospholipids and all types of KATPs tested. The mechanism for the effect of these lipids appeared less likely to involve a very specific or complex cellular process requiring many steps. We hypothesized that PPIs directly interact with KATPs or one of its subunits, mediated by electrostatic force, to maintain the channel in a functional state. We noted a segment of Kir6.2 composed of a high concentration of positively charged residues at the beginning of the cytoplasmic C terminus (Fig. 5A). This positioning of a cluster of positive charges has been shown to be an important determinant of protein topology in other membrane proteins (26,27). To test our hypothesis, two mutants were made in this region: 1) adjacent positively arginine residues at positions 176 and 177 were mutated to neutral alanines (R176A,R177A), and 2) the arginine at 176 was mutated to the negatively charged glutamic acid (R176E). Wild type Kir6.2 co-transfected with SUR expressed a peak KATP current of 1.59 Ϯ 0.3 nA (n ϭ 8) recorded by whole cell clamp at 0 mV with no ATP in the pipette. The current was blocked by glibenclamide and also exhibited the usual loss of activity with time. The R176A,R177A mutant co-transfected with rSUR also expressed a glibenclamide-sensitive K ϩ current, but the current was much smaller, with a mean value of 0.32 Ϯ 0.07 nA (n ϭ 4, p ϭ 0.008, compared with wild type) at 0 mV at maximum. Loss of channel activity was much faster than wild type, and the current showed more fluctuation. Examples of whole cell wild type and R176A,R177A currents are shown in Fig. 6, panels A and B, respectively. In inside-out patches, exogenous application of PPIs also activated the channels expressed from the R176A, R177A mutant, but activation was less effective than for wild type, as demonstrated by a lower open probability and by the transient nature of the activation. Single channel recordings (Fig. 7A) from excised patches showed that the lower open probability in the mutant was caused by a decrease in open times (Fig. 7B) with a possible increase in the closed times. No glibenclamide-sensitive current could be recorded from the cells co-transfected with the R176E mutant and rSUR, suggesting that either this mutant did not express or that the mutation rendered it incapable of opening.

DISCUSSION
Activation/Inactivation of KATP-Gradual inactivation of KATP with time was recognized with the first description of KATP (1) using the inside-out patch clamp technique, a method that severely disrupts the cytoplasmic environment. This inactivation, sometimes called run-down, has been reported or addressed in nearly all research on KATP. Loss of activity is a consistent feature of all native and cloned KATP (Fig. 1, and Refs. 11-13 and 16). Despite much study, the mechanism required to maintain KATP activity is still unclear. Mg-ATP (28,29) was found to transiently and partially activate and maintain KATP in cardiac, pancreatic, and skeletal muscle cells. Some anions (F Ϫ , VO Ϫ ) (30,31) were also shown to have similar effects. Trypsin treatment was found to dramatically maintain KATP activation (18,32). Mg 2ϩ was found to be necessary for ATP activation of KATP; therefore, many researchers speculated that phosphorylation/dephosphorylation was the key to maintenance of KATP channel activity (16). Mg-ATP, however, was often ineffective in activating KATP in rat skeletal muscles (13). In addition, intensive searches for protein kinases and the protein phosphorylation targets have not been successful (32). These results suggest that although phosphorylation may be a part of the regulatory process, additional key elements are involved. In rat skeletal muscle cells, KATP was found to be activated by high concentrations (Ͼ50 mM) of gluconate, with the effect proportional to the length of the carbon chain (13). This finding suggested to us the hypothesis that native anionic FIG. 5. Schematic diagrams for structure and amino acid sequence of Kir channels to account for electrostatic channel-membrane lipid interactions. A, block diagram of the linear amino acid sequence of Kir channels indicating the two transmembrane (M1 and M2) and pore-forming (H5) domains. The amino acid sequence for the initial C terminus (shaded area) for six representatives of the Kir family are shown. B, a diagram of two Kir subunits indicating how positively charged residues might interact with anionic membrane phospholipids to "tether" a "gate" away from the pore, allowing the channel to be active. phospholipids might be an endogenous mediator of this effect, because like gluconate they have an anionic head with a hydrophobic tail.
Intact Anionic Phospholipids Maintain KATP Activity-Activation of KATP in cardiac cells by PI(4,5)P 2 and Mg-ATP/PI has been recently reported (33,34). All negatively charged intact phospholipids we tested possess the ability to activate KATP, including PS, PI, PI(4)P, and PI(4,5)P 2 , directly, and in the absence of magnesium and ATP. Mg-ATP and PI, therefore, are not critical requirements to activate the channel. Although Mg-ATP is not required to mediate activation by anionic phospholipids, the anionic lipid hypothesis can account for the well known Mg-ATP activation effect on KATPs. It has previously been reported that nonphosphorylating analogs of ATP, AMP-PNP and AMP-PCP, are incapable of increasing KATP activity, suggesting that Mg-ATP acts as a phosphoryl group donor in kinase-mediated phosphorylation reaction, although such kinases have not yet been identified (29). Our effectiveness se- The open times were fitted to two components of the exponential. The time constants were 4.3 and 121 ms for Kir6.2/ SUR, and were 3.8 and 32 ms for R176A,R177A/rSUR, respectively. quence of PI(4,5)P 2 Ͼ PI(4)P Ͼ PI suggests the possibility that in vivo, Mg-ATP potentiates the activity of KATPs by serving as a substrate for the phosphoinositide kinases (35) and aminophospholipid translocases that maintain sufficient membrane concentrations of PPIs to maintain KATP by increasing net negative membrane charge. PI(4,5) 2 and phosphorylation of PI may perhaps be the important physiological regulators of KATP activity in this mechanism, a subject that requires further study.
Possible Molecular Mechanisms for the Anionic Lipid Effect-Do anionic phospholipids activate KATPs through an intermediate structure or process or through a direct structural interaction with the channel? Some components of PPIs, especially PI(4,5)P 2 , are known to directly regulate various cellular proteins. One of the most important proteins interacting with the anionic phospholipids is protein kinase C, which uses Ca 2ϩ as part of its signaling pathway. In our cell-free patch clamp experiments, however, ATP was not included in the perfusion solution to the cytoplasmic side of the membrane, and Ca 2ϩ was chelated by EGTA to less than 5 ϫ 10 Ϫ10 M. Indeed, regardless of the presence of ATP, raising the cytoplasmic Ca 2ϩ concentration decreased, rather than increased, the effect of PPIs. Thus, ATP and Ca 2ϩ are not required for the phospholipid effect on KATP, making it unlikely that the protein kinase C signaling pathway is involved in these effects of anionic lipids. Our observation that the protein kinase C activator diacylglycerol was ineffective in activating KATPs is consistent with noninvolvement of protein kinase C. Likewise, the immediate requirement of other phosphorylation processes in maintaining KATP activity is unlikely because of the ATP-free conditions of our study. Another important protein group known to interact with PI(4,5)P 2 is the cytoskeletal regulating proteins such as gelsolin. The possibility of involvement of cytoskeletal structure in the KATP activity has been suggested (33). The effects of PPIs remained even after application of high concentrations of the cytoskeleton disrupters DNase I and cytochalasin B (data not shown). Also, PS, which does not interact with most cytoskeletal proteins (36), was effective in activating KATPs in our experiments. Participation of the cytoskeleton network is evidently not a requirement for the effect, although a modulatory role has not been excluded.
The apparent importance of electrostatic interactions suggested by our data lead us to propose a molecular/physical model involving the channel protein (Fig. 5B). Kir channels that are not thought to associate with SUR (mIRK1, but hROMK1) were also activated by anionic phospholipids. This caused us to focus on the channel structure (Kir6.2) rather than SUR for the mechanisms of the effect. In our model, the positively charged residuals at the beginning of the C terminus of Kir6.2 are anchored or tethered by the electrostatic force of the anionic phospholipid head groups at the cytoplasmic face of the membrane. A decrease in the concentration of anionic phospholipids causes a release or conformation change of the tethered portion of the C terminus, triggering the formation of a gate that closes off the inner vestibule of the channel pore. This model is highly speculative, but protein-membrane electrostatic interactions at this portion of membrane proteins have been proposed previously (26,27). The model also has value because it can account for the known experimental data such as the inhibitory effects of pH and di-and trivalent ions (by charge screening or neutralization), the anionic charge effectiveness sequence, and suggests possible explanations for such effects as trypsin (cleaving the gate) and Mg-ATP (increasing net negative membrane charge by phosphorylation of membrane phospholipids). This model was tested by the Kir6.2 mutant R176A,R177A, which reduced positive charge on the cytoplas-mic tail. As the model predicted, the mutant channel activity was reduced and less sensitive to PPIs.
How widely might this protein-lipid interaction mechanism apply to ion channel activity in general? To date, KATP has been cloned from pancreatic cells (6,7) and cardiac cells (9,37); both require a sulfonylurea receptor and Kir6.2 to function. Kir6.2 is one member of the Kir family with functional and structural similarities, including a concentration of positive charges at the beginning section of the C terminus (Fig. 5A). Loss of channel activity has been generally found in most members of the Kir family, in both native channels (Refs. 38 -40, for example) and expressed channels (41,42). Our demonstration that PPIs activated native non-KATP inward rectifier (Fig. 1B) and also two cloned non-KATP inward rectifiers (Fig. 2) indicates that this mechanism may be general to all channels of this family. The higher effectiveness for more highly charged phospholipids suggests a role for lipid phosphorylation in this regulation. Thus, the PPIs, especially PI(4,5)P 2 and phosphatidylinositol 3,4,5-triphosphate, along with enzymes maintaining them such as phosphoinositide kinases (35), may be important cellular regulators by maintaining a channel protein-anionic lipid interaction required to activate channels.