ATP Modulates Interaction of Syntaxin-1A with Sulfonylurea Receptor 1 to Regulate Pancreatic β-Cell KATP Channels*

ATP-sensitive potassium (KATP) channels are regulated by a variety of cytosolic factors (adenine nucleotides, Mg2+, phospholipids, and pH). We previously reported that KATP channels are also regulated by endogenous membrane-bound SNARE protein syntaxin-1A (Syn-1A), which binds both nucleotide-binding folds of sulfonylurea receptor (SUR)1 and 2A, causing inhibition of KATP channel activity in pancreatic islet β-cells and cardiac myocytes, respectively. In this study, we show that ATP dose-dependently inhibits Syn-1A binding to SUR1 at physiological concentrations, with the addition of Mg2+ causing a decrease in the ATP-induced inhibitory effect. This ATP disruption of Syn-1A binding to SUR1 was confirmed by FRET analysis in living HEK293 cells. Electrophysiological studies in pancreatic β-cells demonstrated that reduced ATP concentrations increased KATP channel sensitivity to Syn-1A inhibition. Depletion of endogenous Syn-1A in insulinoma cells by botulinum neurotoxin C1 proteolysis followed by rescue with exogenous Syn-1A showed that Syn-1A modulates KATP channel sensitivity to ATP. Thus, our data indicate that although both ATP and Syn-1A independently inhibit β-cell KATP channel gating, they could also influence the sensitivity of KATP channels to each other. These findings provide new insight into an alternate mechanism by which ATP regulates pancreatic β-cell KATP channel activity, not only by its direct actions on Kir6.2 pore subunit, but also via ATP modulation of Syn-1A binding to SUR1.

The ATP-sensitive potassium (K ATP ) 3 channel couples intracellular metabolic changes to plasma membrane electrical activity in many cell types (1)(2)(3). In pancreatic ␤-cells, increase in ATP/ADP ratio from glucose metabolism closes plasma membrane K ATP channels, causing membrane depolarization that leads to opening of L-type voltage-dependent calcium channels. Ensuing calcium influx triggers exocytosis of docked and primed insulin granules in part by acting on the exocytic SNARE (soluble NSF attachment protein receptor) complex (4). The pancreatic ␤-cell K ATP channel is a heterooctamer of pore-forming Kir 6.2 and regulatory sulfonylurea receptor 1 (SUR1) subunits, with a 4:4 stoichiometry (5-7). SURs are members of the ATP-binding cassette protein superfamily (8), including cystic fibrosis transmembrane conductance regulator and P-glycoprotein/MDR1 (multidrug resistance 1). These ATP-binding cassette proteins uniformly contain two nucleotide-binding folds (NBF1 and NBF2), with each NBF containing Walker A and B motifs, which are sites of the action by adenine nucleotides. Regulation of K ATP channels by adenine nucleotides is complex, with ATP and ADP having both stimulatory and inhibitory effects (9 -11).
Structure-function studies have demonstrated that ATP inhibits channel activity by interactions with Kir6.2 (12), whereas ADP acts on the SUR subunit to stimulate channel activity in an Mg 2ϩ -dependent manner (13).
The importance of SUR1 as a regulator of K ATP channel activity is demonstrated by the fact that loss-of-function mutations decrease K ATP channel activity, leading to continuous insulin secretion, thus causing hyperinsulinism of infancy (14). Conversely, gain-of-function mutations enhance channel activity, leading to reduction of insulin secretion, thereby causing neonatal diabetes (15). SUR1 regulates K ATP channel activity through several mechanisms: modulating channel pore opening kinetics by enhancing channel sensitivity to ATP and open probability (16), enabling functional expression of the channel at the cell surface (17) and conferring channel sensitivity to MgATP, MgADP and therapeutic drugs (sulfonylureas) (12).
In addition to adenine nucleotides, K ATP channels can also be regulated by a variety of cytosolic factors, including phospholipids (18), long-chain acyl-coenyzme A esters (19), and pH (20). These cytosolic factors undergo dynamic changes in pancreatic ␤-cells in health and diabetes. Remarkably, we have reported that pancreatic ␤-cell K ATP channels can also be regulated by endogenous plasma membrane-bound syntaxin-1A (Syn-1A), a SNARE protein originally described to mediate exocytic membrane fusion. In contrast to the cytosolic factors that act on the Kir6.2 subunit (18 -20), Syn-1A binds the SUR subunits at both NBFs to regulate K ATP channel gating (21,22).
In the present study, we examined the interplay between cytosolic ATP and membrane-bound Syn-1A on regulating K ATP channel gating. We show that ATP and Syn-1A influence ␤-cell K ATP channel sensitivity to each other, with physiological ATP concentrations disrupting the Syn-1A binding to SUR1.
In Vitro Binding Assay and Western Blotting-In vitro binding assays were performed according to the methods we described previously (22). Briefly, GST (as a negative control) and GST-Syn-1A (350 pmol protein each) were bound to glutathione-agarose beads and incubated with 0.5 ml SUR1 overexpressing HEK293 cell lysate extract (400 g protein in lysis buffer containing 20 mM HEPES (pH 7.4), 100 mM KCl, 1.5% Triton X-100, 1 g/ml pepstatin A, 1 g/ml leupeptin, and 10 g/ml aprotinin) in the presence or absence of different concentrations of ATP (Sigma) along with or without 2 mM MgCl 2 at 4°C for 2 h with constant agitation, followed by washing the beads three times with washing buffer containing 20 mM HEPES (pH 7.4), 150 mM KOAC, 1 mM EDTA, 5% glycerol, and 0.1% Triton X-100. The samples were then separated on 10% SDS-PAGE, transferred to nitrocellulose membrane, and identified with specific primary antibodies against SUR1 (1:1000; a gift from J. Ferrer, Barcelona, Spain). The blots were quantified by densitometry scanning followed by analysis with Scion Image (beta release version 4.0.2; Scion Corp.). Data analysis and curve fitting were performed using Origin (version 6.0; Microcal Software Inc.).
FRET Imaging-Transfected cells were imaged in intracellular buffer containing 20 mM HEPES, 5 mM NaCl, 140 mM K-gluconate, and 2 mM MgCl 2 and pre-equilibrated with O 2 : CO 2 ϭ 95:5, pH 7.4 at 37°C using a Nikon TE-2001U inverted microscope fitted with an argon laser unit (480 Ϯ 10 nm, Spectra-Physics), an He-Ne laser unit (545 Ϯ 10 nm, Melles Griot), and a Quantum 512SC charge-coupled device camera (Photometrics). For FRET measurements, a Plan Apo 60ϫ oil immersion objective (1.65 numerical aperture), a Dual-View imaging system (Optical Insights) containing a dichroic splitter (565dcxr) and two emission filters (HQ530/30 for EGFP and HQ630/50 for mCherry) were combined together to allow simultaneous two-channel monitoring of emission fluorescence. The cells were permeabilized by incubation in intracellular buffer containing 10 M digitonin for 4 min at 37°C and then were applied for FRET imaging in the absence or presence of 0.5 or 2 mM ATP. For FRET analysis, EGFP fused with SUR1 was used as FRET donor and mCherry fused with Syn-1A as an acceptor. Four images, including donor excitation/donor emission (Dd), donor excitation/acceptor emission, acceptor excitation/acceptor emission (Aa), and acceptor excitation/donor emission were acquired in the same condition and a fixed 500-ms exposure time in each FRET experiment. In addition, the donor-only sample and acceptor-only sample were imaged before each experiment for calculation of bleed-through. FRET efficiency was used to indicate the interaction of the two proteins, calculated as follows: FRET corr ϭ FRET raw Ϫ (CoB ϫ Dd FRET ) Ϫ (CoA ϫ Aa FRET ). FRET efficiency % ϭ (FRET corr /Dd FRET ) ϫ 100%. Here, CoA is the amount of acceptor bleed-through in the absence of a donor, and CoB is the amount of donor bleed-through in the absence of an acceptor.
Electrophysiology-Rat islet ␤-cells isolated from male Sprague-Dawley as described previously were voltage-clamped in a whole-cell configuration (23). The typical pipette resistance was 3-5 M⍀. The external solution contained the following: 140 mM NaCl, 4 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 2 mM glucose, 10 mM HEPES, pH 7.4 with NaOH. The pipette solution contained the following: 140 mM KCl, 1 mM MgCl 2 , 1 mM EGTA, 0.3 or 1.0 mM MgATP, and 10 mM HEPES, pH 7.3 with KOH. GST or GST-Syn-1A fusion proteins (all at 1 M) were added to the pipette solution for dialysis into the cells. Membrane potential was held at Ϫ80 mV and a pulse of Ϫ140 mV (500 ms) was given every 10 s to monitor K ATP current magnitude. As the pipette contained a low concentration of ATP (when adding 0.3 or 1.0 mM MgATP), which gradually equilibrated with cell cytosol, K ATP currents gradually developed to the maximum amplitude.
INS-1E cells (a gift from C. Wollheim, Geneva, Switzerland) were used for standard inside-out patch-clamp recordings. The cells were transfected with EGFP and BoNT/C1 (to cleave endogenous Syn-1A) or EGFP alone as indicated under "Results." To increase macroscopic K ATP currents, all cells were subsequently co-infected with three recombinant adenoviruses containing Kir6.2, tetracycline-inhibited transactivator, and tetracycline-inhibited transactivator-regulated gene expressing fSUR1 (FLAG epitope-tagged SUR1) (24). 24 h after infection, the cells were patch clamped in symmetrical K ϩ conditions (140 mM KCl, 1 mM EGTA, 1 mM EDTA, and 10 mM HEPES, pH 7.4) with or without MgATP, as indicated. The typical pipette resistance when filled with solution was 1.0 -1.5 M⍀. Syn-1A or GST recombinant proteins were added to bath solutions as indicated. Macroscopic currents were recorded with membrane patches held at Ϫ50 mV to evoke inward currents. All electrophysiological experiments were performed at room temperature (22-24°C) using an EPC9 amplifier with Pulse acquisition software (version 8.6; HEKA Electroniks, Mahone Bay, Nova Scotia, Canada).
Data Analysis-Data are presented as means Ϯ S.E. Differences in means were compared using unpaired (electrophysiological studies) or paired (FRET assay) Student's t test, with p Ͻ 0.05 considered as statistically significant.

ATP Dose-dependently Inhibits Syn-1A Binding to SUR1-
Our previous studies (22,23) showed that Syn-1A binds SUR1 at its NBF1 and NBF2 domains and would not bind Kir6.2. In addition, the physical interaction between SNAP-25 and SUR1 was not observed. These results indicate that the interaction of Syn-1A with NBFs of SUR1 does not require a ternary protein. Because ATP (and ADP), in addition to its direct actions on Kir6.2, can also bind NBFs to modulate K ATP channel activity, we speculated that ATP might induce conformational changes in the NBFs, which could influence SUR1 binding to Syn-1A with consequent effects on K ATP channel gating.
First, we used GST-Syn-1A bound to agarose glutathione beads to pull down SUR1 from SUR1/Kir6.2 stably expressing BA8 cell lysate extract in the presence of increasing concentrations of ATP. As shown in Fig. 1A, ATP (without Mg 2ϩ ) dose-dependently inhibited the binding of Syn-1A to SUR1 (n ϭ 4). As Mg 2ϩ is present physiologically to influence ATP actions, we examined ATP inhibition of Syn-1A binding to SUR1 in the presence of a cellular level of Mg 2ϩ . Addition of 2 mM MgCl 2 shifted the dose-response curve to higher ATP concentrations (Fig. 1B), suggesting a reduction in the ATPinduced inhibitory effect on Syn-1A binding to SUR1. The actions of Mg 2ϩ in regulating K ATP channels are complex, and in particular, MgATP would activate K ATP channels, thus counteracting free ATP inhibition of K ATP channels (9,25 . 1B), suggesting that free Mg 2ϩ in this condition may not be relevant; and this caused a 48% inhibition (Fig. 1B), whereas in absence of Mg 2ϩ (Fig. 1A), the estimated inhibition was stronger at ϳ74%. At lower [ATP] of 0.5-1 mM, where we observed significant effects on SUR1-Syn-1A binding in the absence of Mg 2ϩ (Fig. 1A), adding 2 mM MgCl 2 , greatly reduced these effects, causing a rightward shift of the dose response (Fig. 1B), and here, the calculated free [Mg 2ϩ ] in these conditions, which ranged between 1.1-1.5 mM are quite similar and close to the physiologic values of free [Mg 2ϩ ] reported in the literature (26). We have indicated two arrows in Fig. 1B corresponding to the [ATP] (0.5 and 2 mM, respectively) in the presence of 2 mM MgCl 2 used in the FRET studies in Fig. 2. As a negative control, GST did not bind to Syn-1A in all conditions tested (data not shown).  ATP Disrupts Syn-1A Binding to SUR1 in Living Cells-Because Syn-1A is known to be a "sticky" protein that could bind nonspecifically to other proteins in protein-binding assays, we proceeded to examine the specific interaction in living cells expressing these interacting proteins in their physiological configuration targeted to their native compartment, the plasma membrane. These proteins were fluorophoretagged (EGFP or mCherry) and their bona fide cellular interactions were examined by FRET analysis in the presence and absence of ATP. HEK293 cells were co-transfected with SUR1 tagged with EGFP as a donor and Syn-1A tagged with mCherry as an acceptor, plus Kir6.2 for complete and functional K ATP channel expression at the plasma membrane. A previous study has shown that spectral properties of EGFP and mCherry are well suited for measuring molecular rearrangements by the FRET technique (27). Due to some overlap in EGFP and mCherry spectra, measured mCherry emission caused by FRET was contaminated by both direct excitation of mCherry and by EGFP cross-talk emission in the mCherry range. To overcome this problem, a sensitive spectral method for FRET efficiency quantification was used as described under "Materials and Methods." To examine the molecular interactions of Syn-1A and SUR1 on the surface of the plasma membrane, we used total internal reflection fluorescence microscopy to optically isolate the plasma membrane, enabling high spatial image resolution of the protein interactions on the entire surfaces of the plasma membrane. Total internal reflection fluorescence microscopy, with an exponentially decaying evanescent wave, illuminates the FRET signal within a thin layer (Ͻ 200 nm) at and beneath the plasma membrane, thus significantly reducing out-of-focus background fluorescence.
First, we examined the FRET in HEK293 cells co-transfected with SUR1-EGFP, Syn-1A-mCherry and Kir6.2, and permeabilized with digitonin in the buffer without ATP. We applied a 488-nm laser to excite EGFP and then recorded the mCherry signal (FRET) with a 605-655 nm band-pass filter, which elicited strong FRET signals on the plasma membrane of the cell (Fig. 2, A, ia), indicating very close proximity between SUR1 and Syn-1A, and thus their presumed interactions. The FRET efficiency, expressed as the mean value of FRET corr /Dd FRET , was 44.20 Ϯ 3.91% (n ϭ 12) (Fig. 2, A, ib,  and B). Note that the discrete punctuate spots indicated by arrows showing the highest FRET signals. When 0.5 mM ATP (in presence of 2 mM MgCl 2 , giving free [ATP] ϭ 0.04 mM, free [Mg 2ϩ ] ϭ 1.54 mM, and [MgATP] ϭ 0.46 mM) was applied to these permeabilized HEK cells for 3 min, we observed a dramatic decrease in the FRET signal (purple indicating background FRET) precisely at the same high FRET spots (indicated by arrows, compare Fig. 2, A, ia with 2, A, iia, and Fig. 2, A, ib, with 2, A, iib). Specifically, FRET efficiency decreased by 55.2% from 44.2 Ϯ 3.91% to 19.8 Ϯ 2.32% (Fig. 2B) (Fig. 2, A, iiia, 2, A, iiiB,  and Fig. 2B). As a negative control, no FRET signal was ob-   FEBRUARY 18, 2011 • VOLUME 286 • NUMBER 7 served in HEK293 cells co-transfected with EGFP and mCherry (Fig. 2, B, iva), showing 0% FRET efficiency (Fig. 2 Fig. 1B) in the patch pipette, K ATP currents gradually developed in rat ␤-cells, reaching a maximum current density of 183 Ϯ 17 pA/pF (picoampere/ picofarad) (Fig. 3). Dialyzing 1 M Syn-1A into the cell reduced the currents by 33% (Fig. 3, A and C), similar to our previous report (23). Addition of a higher concentration of 1 mM MgATP (in presence of 1 mM MgCl 2 , giving free [ATP] ϭ 0.10 mM, free [Mg 2ϩ ] ϭ 1.00 mM and [MgATP] ϭ 0.90 mM) in the pipette, maximum current density developed was 53 Ϯ 13 pA/pF. However, dialyzing 1 M Syn-1A into the cells did not significantly reduce the currents (Fig. 3, B and C), though the kinetics of current development (Fig. 3B) appeared slower than control. Thus, addition of 1 mM MgATP shown to cause only moderate inhibition of Syn-1A binding to SUR1 (Fig. 1B) was sufficient to abrogate Syn-1A inhibition of the ␤-cell K ATP channel. Taken together, these results indicate that reduced cellular ATP concentration enhances ␤-cell K ATP channel sensitivity to Syn-1A inhibition.

Interactions between Syn-1A and ATP in K ATP Channel Regulation
Dynamic Interaction of Endogenous Syn-1A and ATP in Regulation of ␤-Cell K ATP Channels-We then explored whether endogenous Syn-1A levels could influence ␤-cell K ATP channel sensitivity to ATP. This is of clinical relevance as islet Syn-1A in diabetics has been shown severely reduced by 79% (28). Here, we depleted endogenous Syn-1A in INS-1E cells by transfection of the cells with BoNT/C1 and EGFP, the latter to identify the BoNT/C1-transfected cells. EGFP alonetransfected cells were used as control. We also performed parallel transfection of HEK293 cells with Syn-1A. 48 h after transfection, Syn-1A was observed to be cleaved completely in INS-1E cells (Fig. 4, Aii) and with only a small residual Syn-1A (85% reduction) in Syn-1A-overexpressing HEK293 cells (Fig. 4, Ai).
Next, we performed inside-out patch recording on the INS-1E cells to test the effects of different ATP concentrations. Note that K ATP channels from inside-out patches are known to be much more sensitive (i.e. 7-26 times) to ATP than those in whole cell configuration in part due to reduced levels of inositol phospholipids (3,29); hence, we used lower ATP concentrations. As shown in Fig. 4, B-D, membrane patches were first exposed to 500 M MgATP In control cells (Fig. 4, Bi), addition of exogenous Syn-1A did not affect K ATP currents compared with GST at either 1 M MgATP (⌬ 1 ) or 20 M MgATP (⌬ 2 ). In contrast, in BoNT/ C1-transfected cells (Fig. 4, Ci), application of 0.3 M exogenous Syn-1A caused much further K ATP channel inhibition at 1 M MgATP as compared with GST (⌬ 1 ). However, at a higher concentration of 20 M MgATP, the exogenous Syn-1A could not further inhibit the K ATP channel current compared with control cells (⌬ 2 ). We summarized these data in Fig. 4, Bii and Cii. Here, ⌬ 1 and ⌬ 2 , which indicate reduc-  tion in the K ATP channel current caused by exogenous Syn-1A compared with GST, were calculated as a percentage of the maximal current (I max ) of each individual patch. Note that ⌬ 1 (at 1 M MgATP) Syn-1A inhibition was much greater in BoNT/C1 transfected cells than the control cells. Our findings therefore show that the effects of ATP on Syn-1A modulation of native K ATP channels is apparent only when endogenous Syn1A is depleted. This suggests that endogenous Syn-1A can rapidly interact with SUR1 when ATP levels are reduced. These results indicate an important role for endoge-nous Syn-1A and its dynamic interaction with ATP in regulating K ATP channels.

DISCUSSION
Syn-1A is a versatile protein that serves as a platform to which exocytic (30) and ion channel (31,32) proteins bind and become activated or form functional complexes, consequently modulating the different steps in exocytosis and ion channel gating. Our previous studies have shown that Syn-1A binds to NBF1 and NBF2 domains of SUR subunits to regu- late K ATP channel activity in both rat pancreatic ␤-cells (22,23) and cardiac ventricular myocytes (21,33). In the present study, we demonstrated that ATP can regulate the physical association of Syn-1A with SUR1 and modulate the inhibitory actions of Syn-1A on K ATP channels. There are two major features. First, ATP is able to dose-dependently inhibit Syn-1A binding to SUR1. Addition of Mg 2ϩ , generating MgATP and free Mg 2ϩ reduced this ATP-induced inhibition. Second, our electrophysiological studies revealed that ATP modulated K ATP channel sensitivity to endogenous Syn-1A inhibition. Below, we discuss each of these features.
ATP Effects on Syn-1A-SUR1 Physical Interactions-Our results examining the effect of ATP on the physical association between Syn-1A and SUR1 clearly demonstrate an ATP concentration-dependent reduction in the interaction. This ATP-induced inhibitory effect is not dependent on Mg 2ϩ , and in fact, addition of Mg 2ϩ reduced Syn-1A-SUR1 binding. A number of studies have shown that Mg 2ϩ play an important role in regulation of many ion channels, such as the voltagedependent calcium channel (34) and K ATP channel (9). ATP and ADP regulate K ATP channel activity by interaction with two physically distinct sites on the channel. Interaction with one site causes channel closure, whereas interaction with the other site leads to channel opening (10). The properties of these two sites differ in their requirement for Mg 2ϩ (9). Our results show that Mg 2ϩ can decrease ATP-induced inhibitory effect on Syn-1A binding to SUR1. Although our finding cannot yet provide the precise Mg 2ϩ acting sites on SUR1, we speculate there are three possible explanations for the reduction of ATP-induced Syn-1A binding to SUR1 by Mg 2ϩ . The first and most likely explanation is that Mg 2ϩ binding to ATP results in reduction of free ATP concentration (see Fig. 1B) and the generation of MgATP, the former known to be the potent inhibitor of K ATP channels and the latter to activate the channel (9,25), thus counteracting the ATP inhibitory effect. Second, Mg 2ϩ may alter SUR1 conformation to form a complex with Syn-1A that becomes resistant to ATP inhibition. Third, Mg 2ϩ increases the affinity of SUR1 to Syn-1A, but this Syn-1A-SUR1 complex remains receptive to ATP disruption. Further detailed work is needed to investigate how these complex interactions of ATP and Mg 2ϩ on SUR1 translate to actual regulation of K ATP channel activity.
When comparing the FRET and binding assays, the ATP sensitivity, in the presence of 2 mM MgCl 2 for abolishing the Syn-1A⅐SUR1 complex, is higher in the FRET study in live cells, particularly at the lower ATP concentration (0.5 mM ATP plus 2 mM MgCl 2 ), wherein disruption of the Syn-1⅐SUR1 complex was 55.2 versus 3.1% for FRET versus binding assays, respectively. The reason for this apparent discrepancy may be that Mg 2ϩ binds to other endogenous proteins in intact cells employed in the FRET assay; thus, the higher concentration of free ATP would cause a stronger inhibitory effect on Syn-1A binding to SUR1. To address this possibility, we tested whether Mg 2ϩ per se can affect the Syn-1-SUR1 interaction. As shown in supplemental Fig. 1, 2 mM MgCl 2 in itself had no effect on the Syn-1A-SUR1 FRET signal. However, addition of 2 mM MgCl 2 to 0.5 mM ATP indeed reduced the ability of 0.5 mM ATP to disrupt the Syn-1A-SUR1 FRET signal. This supports our thinking that Mg 2ϩ could affect Syn-1A-SUR1 interaction only in the presence of ATP and that the cytosolic proteins very likely might be chelating endogenous Mg 2ϩ as to have caused the increased inhibitory effects of low ATP (0.5 mM ATP) in the FRET experiments. Of course, there may be other undefined reasons.
Physiological Implications-Based on the binding data discussed above, one would predict that at physiological concentrations of ATP, there may be only little Syn-1A binding to SUR1. In inside-out patch recording experiments, Syn-1A modulation of K ATP channel activity was only apparent after depletion of endogenous Syn-1A with BoNT/C1. Taken together, these results suggest that endogenous Syn-1A can rapidly complex with SUR1 in ␤-cells when ATP levels are lowered. Type 2 diabetes is associated with lower ATP levels in ␤-cells as a result of diminished efficiency of mitochondrial ATP production in part due to perturbation of uncoupling proteins (35). Here, lower ␤-cell ATP concentration in diabetes patients would increase K ATP channel sensitivity to Syn-1A-induced inhibition, consequently increasing insulin secretion from pancreatic ␤-cells during states of low ATP generation, as a compensatory mechanism.
Cytosolic ATP levels undergo considerable changes or lack of changes in health and disease in many cell types (pancreatic islet ␤and ␣-cells, heart, and brain) to regulate their K ATP channels and consequent membrane excitability. The present study demonstrates that Syn-1A and ATP could modulate the sensitivity of K ATP channels to either mediator in an autoregulatory fashion, thus providing a more versatile mechanism to regulate the K ATP channel in different cell types.