Inhibition of the mitochondrial KATP channel by long-chain acyl-CoA esters and activation by guanine nucleotides.

The mitochondrial KATP channel (mitoKATP) is highly sensitive to ATP, which inhibits K+ flux with K1/2 values of 20-40 μM. This raises the question, how can mitoKATP be opened in the presence of physiological concentrations of ATP? We measured K+ flux in liposomes reconstituted with purified mitoKATP and found that guanine nucleotides are potent activators of this channel. ATP-inhibited K+ flux was completely reactivated by both GTP (K1/2 = 7 μM) and GDP (K1/2 = 140 μM). These ligands had no effect in the absence of ATP. The K1/2 for ATP inhibition exhibited quadratic dependence on [GTP] and [GDP], consistent with two binding sites for guanine nucleotides. We also found that palmitoyl-CoA and oleoyl-CoA inhibited K+ flux through reconstituted mitoKATP with K1/2 values of 260 nM and 80 nM, respectively. This inhibition was reversed by GTP (K1/2 = 232 μM) as well as by the K+ channel openers cromakalim (20 μM) and diazoxide (10 μM). Inhibition of mitoKATP by long-chain acyl-CoA esters, like that of ATP, exhibited an absolute requirement for Mg2+ ions. We propose that the open-closed state of the mitochondrial KATP channel is determined by the relative cytosolic concentrations of GTP and long-chain acyl-CoA esters.

The mitochondrial K ATP channel (mitoK ATP ) is highly sensitive to ATP, which inhibits K ؉ flux with K1 ⁄2 values of 20 -40 M. This raises the question, how can mitoK ATP be opened in the presence of physiological concentrations of ATP? We measured K ؉ flux in liposomes reconstituted with purified mitoK ATP and found that guanine nucleotides are potent activators of this channel. ATPinhibited K ؉ flux was completely reactivated by both GTP (K1 ⁄2 ‫؍‬ 7 M) and GDP (K1 ⁄2 ‫؍‬ 140 M). These ligands had no effect in the absence of ATP. The K1 ⁄2 for ATP inhibition exhibited quadratic dependence on [GTP] and [GDP], consistent with two binding sites for guanine nucleotides. We also found that palmitoyl-CoA and oleoyl-CoA inhibited K ؉ flux through reconstituted mi-toK ATP with K1 ⁄2 values of 260 nM and 80 nM, respectively. This inhibition was reversed by GTP (K1 ⁄2 ‫؍‬ 232 M) as well as by the K ؉ channel openers cromakalim (20 M) and diazoxide (10 M). Inhibition of mitoK ATP by longchain acyl-CoA esters, like that of ATP, exhibited an absolute requirement for Mg 2؉ ions. We propose that the open-closed state of the mitochondrial K ATP channel is determined by the relative cytosolic concentrations of GTP and long-chain acyl-CoA esters.
Mitochondrial K ATP channels (mitoK ATP ) 1 were first discovered in 1991 (1-3). Inoue et al. (1) reported electrophysiological evidence from patch clamp studies of fused mitoplasts, and we described reconstitution of a highly purified mitoK ATP (2,3). In our protocols, which have the advantage of being free of the complexities of intact mitochondria, K ϩ flux is measured using steady-state spectroscopy of the fluorescent probe PBFI. These measurements have permitted initial characterization of the kinetics and regulation of mitoK ATP (4). The K m for K ϩ is 32 mM, and the channel is highly selective for K ϩ (Na ϩ and TEA ϩ are neither transported nor do they affect K ϩ flux through mitoK ATP ). MitoK ATP is inhibited with high affinity by ATP and ADP, and this inhibition exhibits an absolute requirement for divalent cations. We have recently shown that ATP inhibition of K ϩ flux through mitoK ATP is reversed by submicromolar levels of K ϩ channel openers (5).
We also demonstrated mitoK ATP activity in respiring rat liver mitochondria (6). The confounding and unavoidable coexistence of K ϩ diffusion (leak) was controlled by comparing K ϩ flux to TEA ϩ flux, for which there are no endogenous pathways other than diffusive leak. We made the simple demonstration that ATP inhibited K ϩ uptake to rates similar to those of TEA ϩ uptake and had no effect on TEA ϩ uptake itself.
These studies left us with a conundrum: given the high affinity for ATP, how can mitoK ATP ever be opened under normal physiological conditions? We hypothesized (4) that endogenous activators of mitoK ATP must exist to overcome the high affinity for ATP, and we now present support for this hypothesis. K ϩ flux through the MgATP-inhibited channel is restored to full activity by GTP and GDP, neither of which has any effect in the absence of MgATP. GTP and GDP are competitive with ATP, and their reversal of ATP inhibition exhibits hyperlinear kinetics consistent with two guanine nucleotide binding sites. We also report that palmitoyl-CoA and oleoyl-CoA inhibit mitoK ATP with high potency, and this inhibition is also reversed by GTP and by the potassium channel openers, cromakalim and diazoxide. Inhibition by long-chain acyl-CoA esters, like inhibition by ATP, exhibits an absolute requirement for Mg 2ϩ ions and is immediately reversed upon chelation of Mg 2ϩ . From these findings, we infer that GTP and longchain acyl-CoA esters may be the physiological regulators of mitoK ATP and that this channel may play a role in vivo in regulating fatty acid oxidation.

EXPERIMENTAL PROCEDURES
Extraction, Purification, and Reconstitution of MitoK ATP from Rat Liver Mitochondria-Purification and reconstitution generally followed protocols previously described (4,7). Rat liver mitochondria were purified on a linear sucrose density gradient and used to prepare inner membrane vesicles according to the procedure of McEnery et al. (8). Mitochondria were suspended in 220 mM D-mannitol, 70 mM sucrose, 0.5 mg/ml bovine serum albumin, 20 mM HEPES (K ϩ ), pH 7.4, sonicated on ice, and centrifuged for 10 min at 10,000 ϫ g. The supernatant was recentrifuged at 210,000 ϫ g for 30 min, and the resulting membrane pellet was washed three times in PA buffer (0.15 M potassium phosphate, 1 mM ATP, 25 mM EDTA, 0.5 mM dithiothreitol, 5% ethylene glycol, pH 7.9). The final membrane pellet was suspended to 15 mg/ml in PA buffer and stored at Ϫ70°C until needed. Prior to use, the vesicles were incubated in PA buffer containing 3 M guanidine HCl to remove F1-ATPase and bound chaperonins. Treated vesicles were then washed three times in PA buffer (30 min at 138,000 ϫ g) and finally solubilized (2 mg of protein/ml) in 3% Triton X-100, 20% glycerol, 0.1% ␤-mercaptoethanol, 0.2 mM EGTA, 1 mM MgCl 2 , and 50 mM Tris-HCl, pH 7.2.
After incubation on ice for 20 min, the mixture was centrifuged at 120,000 ϫ g for 35 min. Ten ml of the supernatant, typically containing 50 -80 mg of extracted proteins, were loaded onto a DEAE-cellulose column (10-ml bed volume) that had been equilibrated with a buffer containing 1% Triton X-100, 0.1% ␤-mercaptoethanol, 1 mM EDTA, and 50 mM Tris-HCl, pH 7.2. The active fraction, eluting at 250 mM KCl, was * This research was supported in part by Grants GM 31086 and HL 36573 (to K. D. G.) from the National Institutes of Health and a Postdoctoral Fellowship (to P. P.) from the Oregon Affiliate of the American Heart Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The purified mitoK ATP fraction was added to a 10:1 mixture of L-␣lecithin (Avanti) and cardiolipin in 10% octylpentaoxyethylene. The buffer composition at this stage defines the internal medium, which contained 300 M PBFI, 100 mM TEA-SO 4 , 0.14 mM KCl, 1 mM TEA-EDTA, and 25 mM TEA-HEPES, pH 6.8. This mixture was loaded onto a 2-ml Bio-Beads SM-2 column (Bio-Rad) to remove detergent and form proteoliposomes. After incubation for 90 min at 0 -4°C, the column was centrifuged at 400 ϫ g for 2 min to collect the proteoliposomes. To remove extravesicular PBFI, 200-l aliquots of the proteoliposome suspension were passed twice through 4-ml Sephadex G-25-300 columns. The final stock vesicle suspension (nominally 50 mg of lipid/ml) was stored on ice during the experiment. Protein content, measured by the amido black method (9), was normally 10 ng of protein per mg of lipid. Intraliposomal volume of each preparation was estimated from the volume of distribution of PBFI and was normally found to be 1 l per mg of starting lipid.
Assay of K ϩ Flux through Reconstituted MitoK ATP -15 l of stock vesicles were added to 2 ml of external medium containing 150 mM KCl, 1 mM TEA-EDTA, and 25 mM TEA-HEPES, pH 7.4. Electrophoretic K ϩ flux was initiated by 1 M FCCP to provide charge compensation via H ϩ flux. K ϩ flux was quantitated from the fluorescence of intraliposomal PBFI, which increases with increasing [K ϩ ] in . Fluorescence was followed with an SLM/Aminco 8000C spectrofluorometer. The K ϩ response of intravesicular PBFI was calibrated by stepwise additions of KCl to proteoliposomes in internal medium in the presence of nigericin and tributyltin (7).
Chemicals and Reagents-Tris salts of adenine and guanine nucleotides were titrated to pH 7.2 with Tris base. PBFI was from Molecular Probes Inc. (Eugene, OR). All other chemicals were obtained from Sigma unless otherwise indicated. Fig. 1 demonstrate activation of ATP-and ADP-sensitive K ϩ flux by GTP and GDP. The following observations can be made from these data: (i) K ϩ flux was completely restored by both GTP and GDP, and (ii) GTP was 20 -30 times more potent than GDP, irrespective of whether ATP or ADP was used to inhibit K ϩ flux. In 0.5 mM ATP, the K1 ⁄2 values for GTP and GDP activation were 6.9 M and 143 M, respectively. In 0.5 mM ADP, the K1 ⁄2 values for GTP and GDP activation were 0.12 M and 3.4 M, respectively.

Activation of the ATP-inhibited K ATP Channel by GTP and GDP-The results in
Additional experiments (not shown) further characterize guanine nucleotide reversal of ATP inhibition of K ϩ flux through mitoK ATP . (i) Guanine nucleotides had no effect on K ϩ flux through the open channel, measured in the absence of MgATP. (ii) Activation required that guanine nucleotides be added to the same side as MgATP. Thus, external GTP had no effect on K ϩ flux when it was inhibited by internal MgATP. (iii) GTP or GDP activated K ϩ flux when added 30 s after inhibition by MgATP had already been established.
Kinetics of Guanine Nucleotide Activation of the K ATP Channel-To examine the kinetics of activation, we measured ATP inhibition of K ϩ flux in the presence of 3 mM Mg 2ϩ and different concentrations of GTP or GDP. Fig. 2 contains representative dose-response curves.
In the absence of GTP, ATP inhibited K ϩ flux through reconstituted mitoK ATP with a K1 ⁄2 of 21 M in this experiment (solid circles, Fig. 2). We observed K1 ⁄2 values for ATP ranging between 20 and 30 M in 4 independent experiments, and the Hill coefficient was always 1.0 Ϯ 0.1. These K1 ⁄2 values are lower than our previously reported value of 39 M (4) because they are calculated from ATP-sensitive, rather than total, K ϩ flux. We have recently established that 10 -15% of channels are reconstituted with their regulatory sites facing inward and are, therefore, inaccessible to external ATP (10,11).
In the presence of increasing doses of GTP, the K1 ⁄2 value for ATP inhibition was shifted sharply higher (see Fig. 2). It is striking that 20 M GTP increased the K1 ⁄2 for ATP inhibition from 21 M to 6 mM. ATP was ineffective in the presence of 3 mM GTP (not shown). Fig. 3 contains a summary of the results of five experiments in which the K1 ⁄2 for ATP inhibition of mitoK ATP was measured at various concentrations of GTP (q) or GDP (E). These data show that the apparent affinity of mitoK ATP for ATP decreases (K1 ⁄2 increases) in a quadratic manner with guanine nucleotides. In order to extract parameters from the data in Fig. 3, we constructed a simple model for nucleotide interaction with the mitoK ATP receptor, R: This model is consistent with available data. For example, if ATP binds to a second binding site, its affinity is too low to be detected. Solving the kinetic equations for K1 ⁄2 (ATP), FIG. 1. Activation of ATP-and ADP-inhibited mitoK ATP by GTP and GDP. The relative ATP-sensitive K ϩ uptake into liposomes reconstituted with mitoK ATP , ⌬J/⌬J max , is plotted versus concentration of GTP or GDP. All assay media contained 3 mM Mg 2ϩ , and K ϩ influx was initiated by adding 1 M FCCP to assay medium at 10 s. Nucleotides were added to assay medium. A, activation of ADP-inhibited K ϩ flux. GTP or GDP was added to assay medium containing 0.5 mM ADP. ⌬J max is the difference between control fluxes in the absence or presence of 0.5 mM ADP, which inhibited total K ϩ flux by 65% (4). ⌬J is the difference between fluxes in the presence or absence of guanine nucleotide measured in the presence of 0.5 mM ADP. The K1 ⁄2 values and Hill slopes (in parentheses) for activation were 0.12 M (1.0) for GTP and 3.4 M (1.6) for GDP. B, activation of ATP-inhibited K ϩ flux. GTP or GDP was added to assay medium containing 0.5 mM ATP. ⌬J max is the difference between control fluxes in the absence or presence of 0.5 mM ATP, which inhibited total K ϩ flux by 85% to 90% (4). ⌬J is the difference between fluxes in the presence or absence of guanine nucleotide with both fluxes measured in the presence of 0.5 mM ATP. The K1 ⁄2 values and Hill slopes (in parentheses) for activation were 6.9 M (1.2) for GTP and 140 M (1.2) for GDP.
where [G] refers to GTP or GDP concentrations. The data were fit to this equation (solid lines in Fig. 3), using K i (ATP) Ϸ 21 M. The derived dissociation constants for GTP were k 1 Ϸ 0.18 M and k 2 Ϸ 14 M. For GDP, the values were k 1 Ϸ 21 M and k 2 Ϸ 25 M. A simple qualitative interpretation of these results is that GTP reacts at a high-affinity and a low-affinity site, whereas GDP reacts with two low-affinity sites. The low-affinity sites appear to have similar affinities for ATP, GTP, and GDP. Fig. 4 contains the results of experiments on the effects of oleoyl-CoA and palmitoyl-CoA on K ϩ flux through reconstituted mitoK ATP . Long-chain acyl-CoA esters are known to inhibit other ATP-binding transport proteins in mitochondria (12), and they are potent inhibitors of K ϩ flux through mi-toK ATP . Oleoyl-CoA inhibited with K1 ⁄2 Ϸ 80 nM and Hill coefficient of 1.7. Palmitoyl-CoA inhibited with K1 ⁄2 Ϸ 260 nM and Hill coefficient of 2. These experiments were carried out in the presence of 3 mM Mg 2ϩ . Strikingly, acyl-CoA esters had no effect on K ϩ flux in the absence of Mg 2ϩ (0.5 mM EDTA).  Fig. 5B contains fluorescence traces from experiments designed to determine whether palmitoyl-CoA inhibition of K ϩ flux is reversed by GTP. Control flux (trace a) was inhibited by 1 M palmitoyl-CoA (trace b), and this inhibition was prevented by the inclusion of 1 mM GTP in the assay medium (trace c). GTP and K ϩ channel openers also activated K ϩ flux when added after flux was inhibited by palmitoyl-CoA (not shown).

Activation of the Palmitoyl-CoA-inhibited K ATP Channel by GTP and K
The dose-response curves in Fig. 6 demonstrate GTP activation of K ϩ flux inhibited by ATP (q), palmitoyl-CoA (E), or a combination of ATP and palmitoyl-CoA (å). In these experiments, the K1 ⁄2 values were 4 M (in 0.5 mM ATP), 232 M (in 1 M palmitoyl-CoA), and 283 M (ATP plus palmitoyl-CoA). The two important features of these results are that palmitoyl-CoA moved the K1 ⁄2 for GTP activation toward the physiological range of GTP concentration and that ATP had no effect on the K1 ⁄2 for GTP in the presence of palmitoyl-CoA.

DISCUSSION
Regulation of the Mitochondrial K ATP Channel-The purpose of these experiments was to explore regulation of mitoK ATP by physiological ligands. MitoK ATP is inhibited by ATP, ADP (4), and long-chain acyl-CoA esters. Inhibition of mitoK ATP by longchain acyl-CoA esters with high affinity is consistent with a proposed signaling role of this channel in regulating ␤-oxidation of fatty acids (13). Inhibition by these ligands exhibits an absolute requirement for Mg 2ϩ ions, and Mg 2ϩ reduces the apparent affinity for glibenclamide in inhibiting K ϩ flux through mitoK ATP (4). These findings suggest that Mg 2ϩ interacts separately with the mitoK ATP complex, because acyl-CoA esters and glibenclamide are not Mg 2ϩ chelators. It is noteworthy that ADP and acyl-CoA esters, which are chemical analogues, exert opposite effects on K ATP channels from mitochondria and plasma membranes. They inhibit mitoK ATP (4, this paper), but they activate the plasma membrane K ATP channels of pancreatic ␤ cells (14).
Inhibition by adenine nucleotides or acyl-CoA esters can be fully overcome by GTP and GDP, and by the pharmacological agents known as K ϩ channel openers (5, this paper). Guanine nucleotide activation is competitive with ATP, with kinetics indicating two nucleotide binding sites. The effects on the K1 ⁄2 for ATP inhibition (Fig. 3) suggest both high-affinity and lowaffinity GTP sites.
It is characteristic of all K ATP channels that the K1 ⁄2 values for ATP inhibition are roughly 2 orders of magnitude lower than normal cytosolic [ATP]. We now show that the K1 ⁄2 values for GTP reversal of ATP inhibition of mitoK ATP are 2 orders of magnitude less than normal cytosolic [GTP]. These results can, however, be rationalized by the simple consideration that the nucleotide binding sites will be occupied in situ by high-affinity ligands other than ATP. The data suggest that ATP cannot inhibit mitoK ATP in the presence of physiological [GTP], raising the possibility that ATP is not a physiological regulator of mitoK ATP . On the other hand, when long-chain acyl-CoA esters and GTP are present together, as in the experiments of Fig. 6, their K1 ⁄2 values fall within their respective physiological ranges. We infer from our results that the nucleotide binding sites on mitoK ATP are fully occupied by GTP or long-chain acyl-CoA esters under physiological conditions, and that the fraction of open channels is determined by the balance between these regulators.
Mitochondrial Volume Is Controlled by the Potassium Cycle-The mitochondrial K ϩ cycle consists of electrophoretic K ϩ uptake and electroneutral K ϩ efflux across the inner membrane. Any net K ϩ flux will be accompanied by electroneutral flux of anions and osmotically obligated water (15). Because matrix [K ϩ ] is about 180 mM, net K ϩ transport will have little effect on the matrix concentration of K ϩ , but it will have a profound effect on matrix volume. Thus, the redox energy consumed by the K ϩ cycle is the cost of regulating matrix volume (15). The K ϩ cycle is mediated by two highly regulated processes. Efflux is mediated by the K ϩ /H ϩ antiporter, whose existence was predicted by Mitchell (16) and first demonstrated by Garlid (17) nearly 20 years later. Influx is mediated by the mitochondrial K ATP channel (mitoK ATP ), which was described by Inoue et al. (1) and Paucek et al. (4).
A primary role of regulated K ϩ /H ϩ antiport is to compensate for unregulated K ϩ leak into the matrix, driven by the high voltages required for oxidative phosphorylation. Uncompensated K ϩ uptake amounting to as little as 10% of proton pumping would double matrix volume within 1-2 min (18). The K ϩ /H ϩ antiporter is inhibited by matrix Mg 2ϩ (K i Ϸ 300 M) as well as by matrix protons, and the concentrations of these inhibitors decrease with uptake of K ϩ salts, causing compensatory activation of K ϩ efflux (15). Thus, the K ϩ /H ϩ antiporter is responsible for volume homeostasis and is essential for maintaining vesicular integrity in the face of high ionic traffic across the inner membrane.
The discovery of mitoK ATP has profound new implications for mitochondrial physiology, because the existence of a regulated FIG. 4. Oleoyl-CoA and palmitoyl-CoA inhibit K ؉ flux through mitoK ATP . The relative ATP-sensitive K ϩ uptake, ⌬J/⌬J max , into liposomes reconstituted with mitoK ATP is plotted versus concentrations of oleoyl-CoA and palmitoyl-CoA. ⌬J max is the difference between control fluxes in the absence or presence of 0.5 mM ATP, which inhibited total K ϩ flux by 90%. ⌬J is the difference between fluxes in the presence or absence of acyl-CoA ester measured in the absence of ATP. K1 ⁄2 values for oleoyl-CoA and palmitoyl-CoA inhibition were 260 nM and 80 nM, respectively (averages of three independent experiments). Assay medium contained 3 mM Mg 2ϩ . Oleoyl-CoA and palmitoyl-CoA had no effect on K ϩ flux in the absence of Mg 2ϩ . K ϩ influx pathway permits volume regulation. For example, opening mitoK ATP will transiently shift the balance between K ϩ uniport and K ϩ /H ϩ antiport until the antiport catches up with the higher rate of K ϩ influx. This will cause transient swelling to a higher steady-state volume that will persist for as long as mitoK ATP remains open. Such a "regulated interplay" between K ϩ uniport and K ϩ /H ϩ antiport was correctly postulated many years ago by Brierley (19).
Matrix Volume Regulates Electron Transport-Fatty acids are the fuel for thermogenesis by brown adipose tissue mitochondria, and their rate of oxidation is strictly controlled by matrix volume (20). A thorough characterization of this phenomenon by Halestrap (21) has demonstrated that increasing matrix volume, over the narrow range thought to obtain in vivo, greatly stimulates activity of the respiratory chain in both heart and liver mitochondria. ␤-oxidation of fatty acids is particularly sensitive to matrix volume. The site of activation has been localized to membrane enzymes that feed electrons to ubiquinone. The molecular mechanism is not known, but may involve a stretch receptor. Matrix volume changes have been observed in vivo during respiratory stimulation secondary to hormonal activation of liver (21) and brown adipose tissue (22).
A role for mitoK ATP in regulating cellular bioenergetics has been suggested by Halestrap (23), Szewczyk et al. (24), and Garlid (13), and the exquisite sensitivity of mitoK ATP to longchain acyl-CoA esters dovetails nicely with this hypothesis. A plausible scenario is that mitoK ATP will open in the glucosedepleted state, where long-chain acyl-CoA esters are low. The resulting matrix expansion will activate ␤-oxidation and direct energy to support gluconeogenesis in liver, increased mechanical work in heart and skeletal muscle, and thermogenesis in brown adipose tissue. Conversely, elevated long-chain acyl-CoA esters in the fed state may inhibit mitoK ATP , and, together with inhibition of carnitine palmitoyltransferase I (25), promote diversion of energy to fatty acid esterification in hepatocytes, adipocytes, and pancreatic ␤-cells, and to glycolysis in heart and skeletal muscle.