A conserved inhibitory and differential stimulatory action of nucleotides on K(IR)6.0/SUR complexes is essential for excitation-metabolism coupling by K(ATP) channels.

The mechanism by which ubiquitous adenine nucleotide-gated K(IR)6.0(4)/SUR(4) channels link membrane excitability with cellular metabolism is controversial. Is a decreased sensitivity to inhibitory ATP required, or is the Mg-ADP/ATP-dependent stimulatory action of the ATPase, sulfonylurea receptor (SUR), on K(IR) sufficient to elicit a physiologically significant open channel probability? To evaluate the roles of nucleotide inhibition versus stimulation, we compared K(IR)6.1-based K(NDP) channels with K(IR)6.2-based K(ATP) channels and all possible K(IR)6.1/6.2 hybrids. Although K(NDP) channels are thought to be poorly sensitive to inhibitory ATP and to require Mg-nucleotide diphosphates for activity, we demonstrate that, like K(ATP), and hybrid channels, they are inhibited with an IC(50(ATP)) 100-fold lower than [ATP](i). K(IR)6.1 is, however, more efficiently stimulated by SUR than K(IR)6.2, thus providing a mechanism for differential nucleotide regulation, in addition to the known differential interactions of Mg-nucleotides with SUR isoforms. The on-cell and spontaneous activities of K(NDP), K(ATP), and hybrid channels identified in native cells, are different; thus, their similar IC(50(ATP)) values argue the regulatory "beta" SUR subunits play a preeminent role in coupling excitation to metabolism and pose questions about the physiologic significance of models, which assume the ATP insensitivity of open K(IR)s.

shortening the action potentials of cardiac and skeletal myocytes with lower membrane resistance.
The primary signaling mechanism that attunes the P O to the cellular metabolic status remains controversial. Observations that K ATP channels burst spontaneously in inside-out patches and show reversible inhibition by submillimolar ATP (4 -6) suggested that opening these K IR s in vivo, in the face of millimolar nucleotide, would necessitate a significant decrease in their sensitivity to ATP. The reduction in the ATP sensitivity of isolated K ATP channels upon treatment with numerous substances has been interpreted as evidence for the plasticity of ATP-inhibitory gating in vivo. Following the report (7) that ATP weakly inhibits K IR 6.2-based channels in the absence of SUR (IC 50(ATP) ϭ ϳ0.2 mM versus ϳ5-20 M with SURs; Refs. 8 and 9), K IR -based plasticity of ATP inhibition has been considered a possible cause of the variable activity observed for K ATP channel isoforms in situ.
An alternative idea is that SURs, obligatory regulatory subunits of K ATP channels and members of the ABC superfamily (10,11), act as nucleotide sensors and, in vivo, SURs would exert a Mg-ADP/ATP-dependent stimulatory action sufficient to overcome the saturated inhibitory effect of nucleotides at noncanonical nucleotide binding site(s) on the K IR (1). This hypothesis is consistent with only a small fraction, ϳ1%, of the maximal NP O for K ATP channel population participating in the regulation of insulin secretion from pancreatic ␤-cells (12) or shortening of action potentials of metabolically inhibited ventricular cardiomyocytes (13,14). The hypothesis is consistent with the marginal effect of a decreased submembrane concentration of ATP on K ATP channel activity in situ, e.g. the enhancement of K ATP currents observed when adenylyl cyclases are maximally stimulated, thus reducing ATP without producing stimulatory MgADP, in cardiomyocytes dialyzed with submillimolar ATP, but not in the perforated patch or conventional whole-cell configuration with millimolar [ATP] i (15). The physiologic significance of the Mg-nucleotide-dependent stimulation of K IR 6.2 has been established by the identification of SUR1 mutations that compromise the stimulatory, but not the inhibitory action of Mg-nucleotides on pancreatic ␤-cell K ATP channels (16), thus eliminating their short burst activity in situ and causing persistent hyperinsulinemic hypoglycemia of infancy (17).
The lower unitary conductance (g) K IR 6.1-based "K NDP " channels are reported to require Mg-nucleotide diphosphates for activity and show poor sensitivity to ATP (18), suggesting their nucleotide-sensing mechanism might be distinct from that of K IR 6.2-based channels (19,20). This fostered the idea that the basal activity of K NDP channels in intact cells was a result of their poor interaction with inhibitory ATP and that heteromultimerization of K IR 6.0 tetramers might generate novel ATP-inhibitory machinery. However, mutations that pro-duce the largest increase in IC 50(ATP) without affecting the spontaneous P O of K IR 6.2-based channels (21)(22)(23) are at conserved residues, thus arguing for a similar inhibitory mechanism (24). In addition, spontaneously active hybrid channels containing K IR 6.1 and at least one K IR 6.2 show K ATP channellike inhibition by ATP (24), indicating both K IR s sense inhibitory nucleotides via a common mechanism, but respond differentially to the Mg-nucleotide-dependent stimulatory action of SUR. These results are consistent with dual, separable, stimulatory, and inhibitory effects of Mg-nucleotides on both channels (see also Ref. 25). An unambiguous analysis of ATP inhibition of K IR 6.1-based channels requires measurable activity in the absence of stimulatory Mg-nucleotides and openers, ideally on a pore that has not been mutated.
Here we describe first, the use of complimentary approaches to obtain spontaneously bursting K IR 6.1-based channels, without pore mutations, and determine their steady-state, Mg-independent sensitivity to inhibitory ATP. Analyses of K NDP transients induced by rapid wash-out of Mg-ATP were used to verify that nucleotide-inhibitory and stimulatory gating are separable in K IR 6.1/SUR1 channels. Second, SUR-containing channels, the same channels in cell-attached and inside-out configurations, were used to compare the efficiency of the Mg-ATP(ADP)-dependent stimulation of K IR 6.1 versus K IR 6.2based channels in situ. Third, K IR 6.1/6.2 concatemers were used to establish the basic properties of all possible hybrid K ATP channels. Finally, we examined single-channel patches of native cells to determine whether hybrid K IR 6.0-based channels are physiologically relevant.

EXPERIMENTAL PROCEDURES
Molecular Biology-The generation of cDNA constructs and their expression in COSm6 cells have been described (23,24).
Cardiomyocyte Preparation-Human cardiac myocytes were isolated from specimens of myocardium taken from six patients, aged 6 -44 years, during cardiac surgery involving atrial appendage, as described previously for isolation of human ventricular cells (26), except a 2-fold lower concentration of Pronase E was used. Neonatal rat cardiomyocytes were isolated and cultured as described previously (27).
Electrophysiology and Single-channel Kinetic Analysis-Patchclamp recordings were done as described previously (23,24,28). The pipette solution contained (in mM): KCl, 145; MgCl 2 , 1; CaCl 2 , 1; HEPES, 10; pH 7.4 (KOH). The bath, "intracellular" solution contained (in mM): KCl, 140; MgCl 2 , 1; EGTA, 5; HEPES, 5; KOH, 10; pH 7.2 (KOH). The [Mg 2ϩ ] i was kept at a quasi-cytosolic level of ϳ0.7 mM by adding MgCl 2 . The Mg 2ϩ -free internal solution contained (in mM): KCl, 140; EDTA, 5; HEPES, 5; KOH, 10; pH 7.2 (KOH). The holding potential was Ϫ40 mV unless otherwise noted. In all figures, the vertical arrow indicates isolation of an inside-out patch, and the horizontal dotted line gives the level where all K IR 6.0-containing channels are closed. 2Ј-(or 3Ј)-O-(Trinitrophenyl)-ATP (TNP-ATP), ATP-[␥]4-azidoanilide (ATP-␥AA) were from Molecular Probes, Inc. (Eugene, OR) and ALT, Inc. (Lexington, KY), respectively. Other nucleotides, diazoxide, and glibenclamide were from Sigma-Aldrich, Inc. Solutions were exchanged within 2 ms, and the patch currents were recorded at 2-10 kHz (digitized at 20 -100 kHz) to determine the unitary conductance, g, single-channel kinetics, and nucleotide dose responses. The mean P O was estimated from all-points single-channel current amplitude histograms and/or calculated from macrocurrents by fitting the equation: P O ϭ 1 Ϫ 2 *I Ϫ1 *i Ϫ1 , where 2 , I, and i are the K ATP current variance, their mean, and the single-channel current amplitude, respectively. Estimation of the intraburst, but not interburst, kinetics from multiple-channel current traces showing a single level openings is valid and allowed comparative analysis of the O and Cf for K ATP versus K NDP in the presence of ϳ0.1 versus 1 mM Mg-ATP. The differences in averaged values (expressed/plotted as mean Ϯ S.D.) with p Ͻ 0.05 were considered significant.
Metabolic Inhibition of Patch-clamped Cells-COSm6 cells expressing K IR 6.1/SUR1 or K IR 6.2/SUR1 channels were perifused with the oxidative phosphorylation uncoupler, carbonyl cyanide p-(trifluoro-methoxy)phenylhydrazone (FCCP, 0.5 M) in substrate-free, K ϩ -rich bath solution to zero the resting potential, as described previously (28). This method induces a less steep rise in the activity of K ATP channels com-

RESULTS
K NDP Versus K ATP : Analysis of Gating Components-K IR 6.1/ SUR1, unlike K IR 6.2/SUR1 channels, displayed substantial steady-state activity in cell-attached, but not excised patches, in nucleotide-free solution (Fig. 1). K IR 6.1-containing channels could be maximally activated by 10 mM MgUDP and displayed substantial activity in the presence of a quasi-cytosolic concentration of MgATP (1 mM). This activity was strongly increased by addition of 0.2 mM Mg-ADP, conditions that simulate severe metabolic inhibition. Similar observations have been interpreted as poor sensitivity to inhibitory ATP (19,20). However, physiologic concentrations of Mg-ATP reduced the K NDP currents activated by Mg-UDP (Fig. 1, bottom trace), which exerted only a weak inhibitory action on K IR 6.2/SUR1 channels (29) (Fig. 1, upper trace). Rapid wash-out of millimolar MgATP from inside-out patches with "fast" diffusion access induced K IR 6.1/SUR1 current transients (24) not resolved in patches with "slow" access ( Fig. 1, middle versus bottom trace). The fast deactivation of K NDP channels was not the result of a K IR 6.1specific "ultrarapid run-down" (19), as completely deactivated K IR 6.1/SUR1 channels are reopened by MgUDP (24) which, unlike MgATP, is not a substrate for kinases and does not "refresh" spontaneously inactivated K ATP channels (30). The rapid decay of the transient reflects a requirement for stimulatory Mg-nucleotide(s) to overcome the greater stability of the long-lived closed state(s) of K NDP versus K ATP channels, both of which spontaneously inactivate in inside-out patches with a time constant of ϳ10 2 s. The transients could be accounted for by a difference in the off-rate constants for inhibitory versus stimulatory Mg-nucleotides. The unbinding of inhibitory nucleotide from the K IR was rapid and the rise of the transient, like the instantaneous K IR 6.2 currents induced by wash-out of ATP, appeared to be "diffusion-limited," whereas release of stimulatory nucleotide(s) from SUR1 was slower. In support of this explanation, the fastest rates of rise were seen in transients whose peak currents, upon removal of saturating Mg-ATP (10 mM), approached the maximal currents seen in 10 mM MgUDP (in eight different patches). The additional observations that the steady-state NP O values before removal of 1 mM MgATP were Ͻ10% of the peak transient NP O values (see, for example, Fig. 1, middle trace) and that removal of submillimolar (0.1 or 0.3 mM) Mg-ATP from two fast patches induced approximately a 10-fold transient increase in the NP O , further argue that, rather than K NDP channels being poorly sensitive to ATP, their IC 50 for inhibition by Mg-ATP is below 10 Ϫ4 M.
Analyses of the single-channel transients ( Fig. 2) confirmed that K NDP channels were sensitive to submillimolar ATP and demonstrated that their nucleotide-and (V m -E K )-sensitive gating kinetics were separable, as shown previously for K ATP channels (23). In 0.3 mM MgATP, the P O of a single K IR 6.1/ SUR1 channel was very low and rapid wash-out of nucleotide induced a burst of openings ( Fig. 2A). The integral current constructed from a number of sweeps resembled a macrocurrent transient. The results imply that the off-rate constants for release of the inhibitory nucleotides were greater than those for stimulatory Mg-nucleotides; thus, the unbinding of ATP from the inhibitory sites was faster. The stimulatory nucleotide(s) and Mg 2ϩ that remain bound to SUR in the absence of inhibitory ATP promoted the gating transitions that initiate bursts. The eventual nucleotide release from SURs prevented re-entry into a burst, leaving the channel in a nonconducting conformation. We determined, for the first time, the "intrinsic" mean open ( O ) and intraburst closed ( Cf ) times of K IR 6.1-based channels in nucleotide-free solution from these single-channel recordings during a transient (Fig. 2C). The V m -dependent intraburst dwell times were negligibly affected by varying mixtures and concentrations of nucleotides, which specified markedly different P O values by modulating the V m -independent interburst kinetics (Fig. 2B). These results show that inhibition of native K IR 6.1-based channels, in the presence of a physiologic concentration of Mg 2ϩ , occurred via nucleotide binding to a site with a submillimolar K D , which controls gating transitions that are independent of the K ϩ driving force. Similar conclusions have been presented for K IR 6.2-based channels (23), thus arguing the ligand-gating mechanism of these K IR s is the same even though the intrinsic stability of their closed tetramers is different. To demonstrate unambiguously the K ATP -like ATP inhibition of K NDP channels in the absence of Mg 2ϩ , we developed two complementary approaches to induce spontaneous bursting of K IR 6.1-based channels.
⌬NK NDP Versus ⌬NK ATP Channels: Different Spontaneous P O , Similar Inhibition by ATP-The first approach utilized our finding that short deletions from the N termini of K IR 6.2 tetramers does not affect their functional properties and will decrease the occupancy of the interburst closed state of K IR 6.2/ SUR1 channels in the absence of nucleotides (23). Deletion of the 2nd through the 5th residues of K IR 6.1 produced low activity, P O ϳ 10 Ϫ2 , ⌬NK IR 6.1/SUR1 channels after wash-out of nucleotides. ⌬N8K IR 6.1/SUR1 channels exhibited greater spontaneous activity (Fig. 3A) and displayed current transients upon rapid isolation of an inside-out patch or upon MgATP wash-out. The NP O values, of the same channels, determined in the "on-cell" configuration and after excision in 1 mM MgATP, or 0.1 mM ATP without Mg 2ϩ , were comparable. Together the results, demonstrate that truncation of K IR 6.1, like K IR 6.2 (29), does not compromise the response to Mg-nucleotidedependent stimulation by SUR1. Stimulation was more efficient for ⌬NK IR 6.1-versus ⌬NK IR 6.2-containing channels, saturating at Ն10 mM (as seen for intact K NDP in Fig. 1, bottom trace). The observation that 1 mM MgATP decreased steadystate currents following the increase induced by 10 mM MgATP indicates that the inhibitory action of MgATP is saturated at physiological concentrations, whereas its stimulatory action is not. After wash-out of Mg 2ϩ , the spontaneous activity of ⌬N8K IR 6.1/SUR1 channels was markedly reduced by 0.1 mM ATP, consistent with an apparent K D for inhibitory ATP of Ͻ10 Ϫ4 M. Progressively larger deletions, ⌬N13, ⌬N21, and ⌬N33, increased the P O(max) with an apparent saturation of the effect by ⌬N33. Although spontaneously active, the ⌬NK IR 6.1/ SUR1 channel P O(max) values are low in comparison with FIG. 2. K ATP -like, separable nucleotide-sensitive, and V m -dependent single-channel kinetics of K NDP . A, response of a K IR 6.1 4 / SUR1 4 channel to the rapid wash-out of MgATP, superimposed on the sum of similarly induced sweeps collected over ϳ2 min. The transient can be accounted for by unbinding of ATP from inhibitory sites on K IR 6.1, occurring simultaneously with and independently of nucleotide release from the NBFs of SUR1. The latter process is slower, and, thus, the decay of the transients is not limited by diffusional access to the inner side of a patch. B, the activity of the same channel in different Mg-nucleotides. The equivalent pattern-coded bars are labeled once. C, the O and Cf determined from records as in A and B (3 different patches). The (V m -E K )-dependent dwell times, determined from either the first or second halves of transients in the nucleotide-free solution, were similar; the latter are labeled "Mg 2ϩ ." ⌬NK IR 6.2/SUR1 channels. Determination of the inhibitory ATP dose-response curves for a series of ⌬NK IR 6.1 versus ⌬NK IR 6.2-containing channels in the absence of Mg 2ϩ demonstrates that the "longer" channels exhibit lower IC 50(ATP) val-ues and that comparable length ⌬NK IR 6.1-and ⌬NK IR 6.2containing channels have indistinguishable IC 50(ATP) values despite as much as a 50-fold difference in occupation of their interburst closed states. The results are presented as logarithmic plots of the P O values from individual patches versus [ATP] fit with pseudo-Hill curves (Fig. 3B). The IC 50(ATP) for ⌬N5K IR 6.1/SUR1 is ϳ10 Ϫ5 M; using this approach, an extrapolated value for full-length K IR 6.1/SUR1 channels is ϳ5-10 M.
K IR 6.0 Tetramers: Conserved Noncanonical, Adenine-selective Nucleotide Binding Sites-A second approach to obtain measurable spontaneous activity from K IR 6.1 4 /SUR1 4 channels comes from our recent observation that concatenated K IR 6.2-2-2-2/SUR1 4 channels exhibit an increased P O(max) (24). From a survey of a large number of patches, we obtained inside-out membrane fragments that contained Ͼ10 2 concatenated K IR 6.1-1-1-1/SUR1 4 channels and had negligible background currents. These patches were used to characterize the low P O(max) activity of concatemeric channels in Mg-free internal solution (Fig. 4A) and demonstrate they are reversibly inhibited by ATP with an IC 50(ATP) value of ϳ10 Ϫ5 M, consistent with FIG. 3. ATP sensitivity of ⌬NK IR 6.1-versus ⌬NK IR 6.2-containing channels. A, ⌬NK IR 6.1 4 /SUR1 4 channels exhibit a substantial steady-state activity in Mg-nucleotide-free solution with O and Cf similar to those in the presence and absence of ATP in Fig. 2. B, K IR 6.2 4 /SUR1 4 -containing patches (gray crosses fit with pseudo-Hill functions) were tested before and after recording from three ⌬NK IR 6.1 4 / SUR1 4 -(in black) versus three comparable ⌬NK IR 6.2 4 /SUR1 4 -containing patches (in gray) derived from the same transfection. The K IR type and time after transfection are given by the symbol/line type and size code, respectively. No time-dependent shifts were observed, validating the IC 50(ATP) estimates from averaged ATP dose responses (23,29). Consistent with these reports, the pseudo-Hill coefficients (h) were somewhat higher for the ⌬N versus wild-type channels. This may reflect modified interactions between the cytoplasmic domains of ⌬NK IR 6.0, but differences in h are an ambiguous reporter of changes in the cooperativity or stoichiometry of ligand binding or K IR inhibition and we do not interpret them further. The averaged IC 50(ATP) for ⌬N5, ⌬N8, ⌬N13, ⌬N21, and ⌬N33K IR 6.1 were 10.3 Ϯ 0.7, 20.5 Ϯ 2.9, 44.8 Ϯ 3.3, 62.9 Ϯ 10.5, and 83.6 Ϯ 9.1 M, respectively, and for ⌬N5, ⌬N10, ⌬N20, ⌬N32, and ⌬N44K IR 6.2 were 11.1 Ϯ 1, 21.1 Ϯ 3.7, 57.7 Ϯ 1.7, 90.5 Ϯ 2.9, and 92.7 Ϯ 4.8 M, respectively. the IC 50(ATP) values estimated above and indistinguishable from those of similarly concatenated, although more active, K IR 6.2 4 /SUR1 4 channels (24). The availability of spontaneously active K IR s allowed a direct, unambiguous comparison of the inhibition of K NDP versus K ATP channels by different nucleotide triphosphates. Inhibition of native K ATP channels gave a rank order of effectiveness of ATP ϳ ATP␥S Ͼ GTP (30,31), suggesting the importance of the adenine ring. We observed that 8-azido-ATP, with its bulky substitution on the adenine ring, was ϳ10-fold less effective than ATP as an inhibitor of K IR 6.2/ SUR1 channels, although it has a similar affinity for SUR (32)(33)(34). ATP-␥AA is poorly recognized by SUR, but inhibits homomeric K IR 6.2 channels as effectively as ATP (35). We compared the dose-response curves of these ATP analogs, and TNP-ATP, from spontaneously active K IR 6.1-1-1-1/SUR1 4 and K IR 6.2-2-2-2/SUR1 4 channels (Fig. 4B). The 8-azido-group on the adenine ring reduced the inhibitory effectiveness of ATP on both channel types, whereas neither the bulkier azidoanilide group on the ␥-phosphate nor the TNP group on the ribose ring significantly affected the IC 50 values. In addition, single concentration tests on the same patch suggest a similar rank order of effectiveness, ATP Ͼ GTP Ͼ UTP, for K IR 6.1-1-1-1/SUR1 4 and K IR 6.2-2-2-2/SUR1 4 channels (three independent experiments each; Fig. 4C). The results show that the gating machinery of K IR 6.1 and K IR 6.2 is controlled by similar noncanonical, adenine-selective, inhibitory nucleotide-binding sites.
Enhanced Adenine Nucleotide-dependent Stimulation of K IR 6.1 by SUR-The K IR 6.2-like ATP-sensitivity of K IR 6.1 channels can be reconciled with their higher activity in situ (compare currents before patch excision in Fig. 1, top versus middle traces) if K IR 6.1 is more responsive to the Mg-nucleotidedependent stimulatory action of SUR. The activity of K NDP channels in intact cells was comparable with that observed in inside-out patches in a quasi-cytosolic concentration of Mg-ATP, or in a 10-fold lower concentration of ATP without Mg 2ϩ (Fig. 1, middle trace, and Fig. 3A), revealing a substantial stimulation of K IR 6.1/SUR1 channels by Mg-ATP. The nucleotide stimulatory action saturates at a much higher concentration of ATP than inhibition, and was more efficient for ⌬NK IR 6.1-versus ⌬NK IR 6.2 channels as shown in Fig. 3 (this comparison is valid, as N-terminal truncation did not alter the stimulatory action of nucleotides). Moreover, 0.2 mM MgADP added to 1 mM MgATP stimulated K IR 6.1/SUR1 channels Ͼ7fold, whereas the same concentration of MgADP, applied to K ATP channels whose P O was preset to a value close to that of K NDP channels by varying the concentration of MgATP, stimulated K IR 6.2/SUR1 channels Ͻ3 times (5 versus 5 patches; see an example in Fig. 1). The O and Cf values of K NDP versus K ATP channels were comparable irrespective of a 10-fold difference in the concentration of Mg-ATP (1.69 Ϯ 0.27 versus 2.16 Ϯ 0.38 ms, and Cf 0.21 Ϯ 0.03 versus 0.25 Ϯ 0.03 ms, filtered at 5 kHz, respectively; 3 versus 3 patches).
Further evidence for more efficient stimulation of K IR 6.1based channels by adenine nucleotides in nondisrupted cells, where the concentrations of ATP and ADP (and thus their ratio) are expected to be comparable, was obtained by measuring the activity of K NDP versus K ATP channels in on-cell patches upon poisoning with the metabolic inhibitor, FCCP (see "Experimental Procedures"). The normalized increases in NP O , for similar duration of FCCP perifusion, were larger for K IR 6.1-based channels (Fig. 5A).
The K ATP opener, diazoxide, strongly stimulated K NDP currents even in the presence of 10 mM Mg-ATP (the highest accurately testable concentration) where K ATP currents were practically immeasurable. In patches with a high density of K ATP channels, which is required to obtain accurately measur-able currents at a physiologic concentration of MgATP, diazoxide (300 M) induced a greater than 3-fold increase in the activity of K IR 6.1/SUR1 versus K IR 6.2/SUR1 channels in 3 mM MgATP (Fig. 5B). This test, using subsaturating concentrations of either stimulatory ligand, provides an additional line of evidence that SUR more efficiently stimulates K IR 6.1 consistent with the hypothesis that a differential response to the Mg-nucleotide-dependent stimulatory action of the regulatory subunit can explain the observed differences in K IR 6.1/SUR1 versus K IR 6.2/SUR1 activity in situ.
Hybrid K ATP Channels: Differential Nucleotide Stimulation Sets the P O in Situ Irrespective of the K IR 6.1/K IR 6.2 Ratio-dependent Stability of Closed K IR s-We examined the nucleotide gating of all possible hybrid K IR 6.0-0-0-0 concatemers, expressed with SUR1, in which the number and positions of K IR 6.1 and K IR 6.2 were systematically varied (Fig. 6A, left  column). For each hybrid we determined that there were no subconductance states resolvable from single-channel current fluctuations within a 10-kHz bandwidth (Fig. 6B illustrates one test for K IR 6.2-2-1-1). These results complement our data for K IR 6.2-2-2-2 and K IR 6.1-0-0-2 (24), where no subconducting states were found in the presence ATP, indicating that reorientations of the pore-lining domains of these K IR s during gating must be highly coordinated, as has been argued for pH i -gated KcsA channels (36,37). The single peak i-distributions allow calculation of the mean P O from currents through representative macropopulations of hybrid channels (see "Experimental Procedures"). Fig. 6C gives a side-by-side comparison of the macrocurrents through a subset of hybrid channels (K IR 6.2-0-0-1/SUR1 4 ) with identical K IR 6.2 heads, K IR 6.1 tails, and -GGGSGGGA-linkers (24). An increase in the number of K IR 6.1 subunits, irrespective of their position, is positively correlated with channel activity in the cell-attached configuration and negatively correlated with steady-state activity in excised patches in nucleotide-free solution. The spontaneous activity of all hybrid channels was similarly reduced by 0.1 mM ATP. The mean P O values calculated from segments of macro- FIG. 6. Concatemers of K IR 6.0 with all possible K IR 6.1 (red):K IR 6.2 (blue) ratio and subunit order co-assemble with SUR (gray) spontaneously bursting channels. A, the g (normalized as in Ref. 24) and IC 50(ATP) values for different K IR 6.0-0-0-0/SUR1 4 . Individual patches were tested without knowing the type of expressed concatemer. B, a single K IR 6.2-2-1-1/SUR1 4 channel current at Ϫ90 mV, and its i-histogram; some longlived interburst intervals were deleted from the trace before construction of the histogram to obtain comparable areas under the peaks and increase the significance of the multi-Gaussian fit. C, macrocurrents through all K IR 6.2-0-0-1/SUR1 4 . The density of K IR 6.1-containing channels was usually lower than that of K IR 6.2-based channels (24); thus, hybrids were tested in macropatches without resolving current transients. Segments of records in which the amplification was changed to obtain maximal signal resolution upon transition from cell-attached (c-a) to inside-out (i-o) configuration have been deleted. The equivalent time and pattern-coded bars are labeled once. S.D. values of the idealized P O responses, shown on the right, are given by the dotted lines. Nucleotide Regulation of the K ATP Channel Family 49088 currents (see "Experimental Procedures") were used to construct averaged P O responses (3 patches for each channel type; plots are to the right of each macrocurrent trace in Fig. 6C). These plots show that the P O , in situ, from similar cells is significantly higher for channels containing more K IR 6.1 subunits, although the spontaneous P O(max) of the hybrid channels, determined in inside-out patches, is progressively reduced with increasing K IR 6.1/K IR 6.2 ratio. The maximal P O at 10 mM MgUDP, which saturates stimulation, was Ͼ0.5 in all experiments. Figs. 7A and 6A summarize the values of g and IC 50(ATP) for all of the possible hetero-versus homoconcatemers. Six populations of channels can be distinguished, two with wildtype and four with intermediate conductivity, verifying that the channel pore structure is determined by concatemer composition (24). All of the hybrid channels are inhibited with an IC 50(ATP) of ϳ10 Ϫ5 M, equivalent to that of K IR 6.2-2-2-2/SUR1 4 channels.
Identification of Native Hybrid K ATP Channel Activity in Intact Cells-Overexpressed K IR 6.1 and K IR 6.2 co-immunoprecipitate and co-assemble with SUR2B channels (38), which conduct like our hybrid K ATP channels (24). The present results with heteroconcatemeric K IR s demonstrate that all of the possible heteromultimeric K ATP channels are functional and define their basic properties. Confirmation of the existence of native hybrid K ATP currents requires a demonstration that the same channel that bursts in an intact cell displays intermediate g openings (g values of 50 -65 picosiemens between Ϫ80 and Ϫ20 mV with [K ϩ ] o ϳ [K ϩ ] i ϳ 150 mM) in the inside-out configuration, which are inhibited Ͼ50% by 30 M ATP in the absence of Mg 2ϩ . The presence of both K IR 6.0 messages in mammalian cardiomyocytes, and the existence of "novel ATPsensitive K ϩ channels" has been documented in rabbit nonventricular and neonatal rat cardiomyocytes (39 -41). To identify native hybrid K ATP channels, we searched for glibenclamidesensitive weak inward rectifiers in intact human atrial and cultured neonatal rat cardiomyocytes. The latter cells were advantageous for single K ATP recording, as their K IR density is lower than in adult ventricular cells. Records with a single level of channel openings over Ն2 min of continuous recording were selected for off-line analysis from a large number of patchclamp records, obtained in different laboratories (see "Acknowledgments") under the same experimental conditions used to analyze the recombinant hybrids. A dozen channels were obtained, which satisfied all the conservative criteria listed above for identification of a native hybrid K ATP channels (Fig. 7B). The distinctive g levels of these channels match those of recombinant hybrids (compare with Fig. 6A). The P O(max) values of the native hybrids were lower than those of K IR 6.2-based channels despite their comparable inhibition by ATP (8,9,28), ruling out the interpretation that the on-cell activity of the native channels was the result of stabilization of an ATPinsensitive open state by locally altered membrane phospholipids (42)(43)(44). DISCUSSION The results demonstrate the following points. 1) K NDP channels have K ATP -like sensitivity to inhibitory ATP. 2) Mg-nucleotide-dependent stimulation, by the same SUR, is more efficient for K IR 6.1 than for K IR 6.2. 3) Hybrid channels are uniformly sensitive to inhibitory ATP, irrespective of their K IR 6.1 content, and irrespective of the dissimilar intrinsic stabilities of their closed state(s), which are specified by the K IR 6.1/K IR 6.2 ratio. On the other hand, the efficiency of Mg-ADP/ATP-dependent stimulation of these same hybrid channels, determined in on-cell patches, correlates positively with K IR 6.1 content. 4) We have identified native channels in intact cardiac cells that are glibenclamide-sensitive, have the intermediate conductivity and basal activity expected for hybrid K ATP channels, and are inhibited Ͼ50% by 30 M ATP after removal of Mg 2ϩ i . Together, these findings strongly favor a regulatory mechanism in which the inhibitory action of adenine nucleotides on K IR 6.0 pores is always saturated and channel openings are stimulated by SUR in response to changes in Mg-ADP/ATP.
The demonstration that all of the possible K IR 6.1-and K IR 6.2-containing channels are similarly sensitive to inhibitory ATP is consistent with earlier arguments that the two SUR-K IR gene complexes evolved from a common pair of ancestral genes and strongly implies their ligand-gating machinery has been conserved. The results rule out suggestions that K NDP channels are poorly sensitive to inhibitory ATP and that heteromultimerization might have increased hybrid K ATP conductance either by elevating the apparent K D for ATP or by reducing the cooperativity of Mg-independent ATP inhibition. The demonstration that the ATP-inhibitory curves, over a physiologic range of concentrations (up to 3 mM), are essentially identical for K IR s that have a nearly 50-fold difference in their P O(max) values reinforces arguments (8) against the validity of linear kinetic schemes, which require interburst closed-state delimited ATP binding. The data imply an ATP-binding site, which is accessible before the K IR transitions to a nonconducting "interburst" conformation. Defining this noncanonical inhibitory nucleotide-binding pocket will require structural analysis of the C terminus and the proximal N-terminal domain (from the 45th residue; Ref. 23) of K IR 6.0.
The finding that the Mg-nucleotide-dependent stimulation of K IR 6.1 by SUR is more efficient than for K IR 6.2 complements the evidence that "classic" K ATP channel subtypes are differentially stimulated by SUR isoforms with SUR1/K IR 6.2 ϳ SUR2B/K IR 6.2 Ͼ SUR2A/K IR 6.2 (compare the MgATP versus ATP responses in Refs. 28, 29, and 45). Differential stimulation can account for the higher apparent IC 50 for ATP inhibition of native and recombinant ␤-cell K IR 6.2 4 /SUR1 4 versus K IR 6. The origin of this differential stimulation could be differences in the cooperative interactions of Mg-nucleotides with the NBFs of SUR isoforms in which at least NBF2 may hydrolyze ATP (47). Comparison of K ATP channel subtypes (above), would predict that the hydrolytic activity of SUR1 and SUR2B is higher than the basal ATPase activity of SUR2A. We propose that two additive mechanisms specify the differential responses of SUR/K IR 6.0 channel variants to [ATP]/[ADP] in physiologic Mg 2ϩ , variations in the binding and/or rate of hydrolysis of ATP by SUR isoforms, and differences in the efficiency of response of K IR 6.1 versus K IR 6.2 to stimulation by SURs liganded with ATP/ADP. This could account for the observations that K IR 6.1/SUR2B channels in smooth muscle cells and K IR 6.2/SUR1 channels in ␤-cells and neurons display a basal activity, whereas K IR 6.2/SUR2A channels in intact striated myocytes are virtually silent (reviewed in Ref. 1).
The analysis of hybrid K ATP channels shows there is no barrier to co-assembly of K IR 6.1 and K IR 6.2 and indicates that hybridization can provide a mechanism for diversification of K IR 6.0/SUR channel currents in situ. It is unclear whether the biogenesis and trafficking of K IR 6.1-based channels is equivalent to their K IR 6.2-based counterparts (see for example Ref. 48 versus Ref. 38). Heteromultimerization did not increase channel density (Ref. 24 and Fig. 6C); thus, the principal mechanism by which hybridization may up-regulate K ATP conductance is through enhancement of the stimulatory action of Mgnucleotides. Interestingly, sporadic observations of native K ATP channel openings in millimolar ATP solutions containing low Mg 2ϩ have been reported, but were interpreted as evidence for "hyposensitization" of K IR 6.2 to ATP by a local increase in the concentration of membrane PIP 2 (Ref. 49 versus Ref. 44). These rare currents, mostly from cardiac cells, have not been well characterized at the single-channel level, and it is possible they are caused by rare hybrid K IR 6.1-containing K ATP channels. We identified native currents with the characteristics expected for hybrid K ATP channels, based on the properties of the K IR 6.0 FIG. 7. Properties of heteroconcatemeric and native hybrid K ATP channels. A, summary plot of data from an analysis of records done without knowing the concatemer type. Subunits are colorcoded as in Fig. 6. Spheres identify individual data points. The g values were determined as in Ref. 24. The normalized g values are pseudocolored from g ϭ 0.515 (red) through g ϭ 1 (blue). White lines connect data points from each K IR 6.0-0-0-0 subtype; the gray lines are their projections. The white squares and bars on the back wall give the means Ϯ S.D. for the distinguishable g values. The IC 50(ATP) values, yellow drop lines, are nearly normally distributed; the yellow squares and bars on the floor give their means Ϯ S.D. All of these IC 50(ATP) values are slightly higher than those determined for nonconcatenated SUR1-containing K ATP channels (24). We presume this minor effect reflects the restricted position/ mobility of the K IR cytoplasmic domains, as we have verified the wild-type properties of channels co-assembled from SUR1 and K IR 6.0 with the attached "half"-linker (24), and a similar shift was observed for channels assembled from SUR1-K IR 6.2 fusion proteins (56). B, characteristics of individual native K ATP -like channels from adult atrial (a) and neonatal (n) cardiomyocytes sorted by their g values after normalization to that of classic K ATP channels (i.e. Ͼ70 picosiemens, silent in the cell-attached configuration, but active in the inside-out configuration in the absence of nucleotide) from adult human and rat ventricular cells, respectively, tested in parallel.
concatemers. In intact atrial and developing cardiomyocytes, these channels display a basal activity, providing additional evidence for the physiologic relevance of differential stimulation of SUR/K IR 6.0 variants by Mg-nucleotides. Further studies are required to explain the negligible surface expression of K IR 6.1-containing channels in mature ventricular cardiomyocytes documented in our previous analysis (28) and the lack of significant K ATP macrocurrents in ventricular myocytes from K IR 6.2 Ϫ/Ϫ mice (50).
The present results strongly support the idea that nucleotides exert a dual action on K ATP channel gating (reviewed in Ref. 1) and extend this concept to all possible K IR 6.0/SUR channels. The essential idea, which resolves the longstanding paradox of how K ATP channels with an IC 50(ATP) ϳ100-fold lower than the estimated cytosolic concentration of ATP can couple membrane excitation with metabolism, is that nucleotides, acting through separate binding sites on K IR s and SURs, exert both Mg-independent inhibitory and Mg-dependent stimulatory actions, respectively. The demonstration that C-terminally truncated K IR 6.2 was inhibited by ATP in the absence of SUR provided the first evidence for an inhibitory nucleotide binding site on K IR 6.2 (7), a result consistent with the low affinity photolabeling of K IR 6.2 with azido-analogs of ATP (35,51). Our data with ATP and other nucleotides argue this noncanonical (see also Ref. 32), purine ring-selective inhibitory site is conserved in both K IR 6.1 and K IR 6.2. SUR plays two roles in nucleotide-dependent gating: 1) assembly of K IR 6.2 with SUR1 potentiates the inhibitory effect of ATP, as shown by steadystate IC 50(ATP) values of ϳ200 M for K IR 6.2 alone versus ϳ5 M for K IR 6.2/SUR1 channels (8,9); and 2) SUR is required for Mg-nucleotide-dependent stimulation of K ATP channel activity. Numerous mutations in the NBFs that abolish the stimulatory action of SUR do not alter the inhibitory action (for example the G1479R mutation in NBF2 of SUR1; Ref. 16). In vivo the inhibitory effect of nucleotides is nearly saturated with K ATP channels operating at the extreme of their inhibitory curve. This includes normal, hypoxic, ischemic, or low blood sugar conditions, at submembrane concentrations of Mg-ADP up to ϳ0.1-0.2 mM at Mg-ATP concentrations of ϳ1-2 mM. These concentrations of MgADP/ATP are near the middle of the stimulatory dose-response curves, thus allowing optimal sensing of the metabolic ratio by the regulatory subunit. This is consistent with evidence that physiologic concentrations of adenine nucleotides do not saturate the stimulatory action of SUR. A variable Mg-nucleotide stimulation, when the inhibitory action is saturated, means that K IR 6.0/SUR channels monitor the metabolic index, i.e. the ADP/ATP ratio, rather than [ATP] i itself. This is a critical point as variation in inhibitory [ATP] i continues to be cited as the basis for physiologic regulation of K ATP channels.
[ATP] i "depletion" in vivo, even under severe ischemic conditions, does not fall to the submillimolar levels achieved during conventional whole-cell or inside-out patch recording, and even severe metabolic poisoning fails to activate K IR 6.2/SUR1 channels with disabled NBFs (16).
From a signaling viewpoint, K ATP channels would be poor sensors and noisy effectors unsuited for precise coordination of changes in V m with metabolic index, if they were regulated by ATP-binding with an IC 50(ATP) closer to cytoplasmic [ATP], i.e. ϳ0.2 mM, as determined for K IR 6.2 tetramers. The increased sensitivity of K IR 6.0/SUR complexes to inhibitory ATP may have evolved as a safeguard against expression of "leaky" K IR s at the cell surface where their overactivity is deleterious (9,23,52). K IR 6.2 tetramers are retained in the endoplasmic reticulum when expressed alone in Xenopus oocytes (53) and fail to reach the cell surface in SUR1 knock-out mice (54). Whether antagonism of the action of inhibitory ATP, e.g. by high [PIP 2 ], represents a physiologically significant mechanism with which to facilitate excitation-metabolism coupling remains to be established. It is unclear, for example, how impaired metabolism could increase endogenous [PIP 2 ]. Moreover, experiments designed to strongly increase phosphatidylinositol-4-phosphate 5-kinase activity resulted in approximately a 5-fold decrease in ATP sensitivity of K IR 6.2/SUR1 channels, which is a marginal effect in comparison with the ϳ500-fold increase in IC 50(ATP) for these channels in similar patches treated by exogenous PIP 2 micelles (Ref. 55 versus Ref. 44). We argue that the dual action of nucleotides on the entire K IR /SUR channel family, rather than hyposensitization of K IR s to inhibitory ATP, is essential for the high fidelity coupling of metabolism to excitation.