The Carboxyl Termini of KATP Channels Bind Nucleotides*

ATP-sensitive potassium (KATP) channels are expressed in many excitable, as well as epithelial, cells and couple metabolic changes to modulation of cell activity. ATP regulation of KATP channel activity may involve direct binding of this nucleotide to the pore-forming inward rectifier (Kir) subunit despite the lack of known nucleotide-binding motifs. To examine this possibility, we assessed the binding of the fluorescent ATP analogue, 2′,3′-O-(2,4,6-trinitrophenylcyclo-hexadienylidene)adenosine 5′-triphosphate (TNP-ATP) to maltose-binding fusion proteins of the NH2- and COOH-terminal cytosolic regions of the three known KATP channels (Kir1.1, Kir6.1, and Kir6.2) as well as to the COOH-terminal region of an ATP-insensitive inward rectifier K+ channel (Kir2.1). We show direct binding of TNP-ATP to the COOH termini of all three known KATP channels but not to the COOH terminus of the ATP-insensitive channel, Kir2.1. TNP-ATP binding was specific for the COOH termini of KATP channels because this nucleotide did not bind to the NH2 termini of Kir1.1 or Kir6.1. The affinities for TNP-ATP binding to KATP COOH termini of Kir1.1, Kir6.1, and Kir6.2 were similar. Binding was abolished by denaturing with 4 m urea or SDS and enhanced by reduction in pH. TNP-ATP to protein stoichiometries were similar for all KATP COOH-terminal proteins with 1 mol of TNP-ATP binding/mole of protein. Competition of TNP-ATP binding to the Kir1.1 COOH terminus by MgATP was complex with both Mg2+ and MgATP effects. Glutaraldehyde cross-linking demonstrated the multimerization potential of these COOH termini, suggesting that these cytosolic segments may directly interact in intact tetrameric channels. Thus, the COOH termini of KATPtetrameric channels contain the nucleotide-binding pockets of these metabolically regulated channels with four potential nucleotide-binding sites/channel tetramer.

Although the SUR/fibrosis transmembrane conductance regulator subunits contain nucleotide-binding folds (11,12), this subunit is not required for ATP-mediated inhibition of K ϩ channel activity. For example, deletion of the last 36 amino acids from the COOH terminus of Kir6.2 (Kir6.2⌬C36) produces functional K ϩ channels in the absence of coexpressed SURs that are sensitive to ATP (13). Nevertheless, SUR subunits are required for ADP-mediated activation of K ATP channels (14 -16). Thus, ATP inhibition of K ATP channel activity is thought to involve direct interaction with Kir subunits despite the lack of identifiable nucleotide-binding motifs. The recent demonstration of the photoaffinity labeling of Kir6.2 channel by 8-azido-[␥-32 P]ATP (17,18) also supports the direct binding of ATP with the pore-forming subunit of K ATP channels. In addition, mutations in both the NH 2 -and COOH-terminal regions of the Kir6.2 (13, 19 -23) and Kir1.1 (24) subunits alter the EC 50 for ATP-mediated channel gating. Because ATP-mediated inhibition of channel activity must be a complex process involving residues that form an ATP-binding pocket and others that may be required for linking ATP binding to channel closure, those mutational studies of channel gating by nucleotides do not provide unequivocal evidence for direct involvement of those residues in ATP binding.
In the present study, we assessed the direct binding of fluorescent 2Ј,3Ј-O-(2,4,6-trinitrophenylcyclo-hexadienylidene) adenosine triphosphate (TNP-ATP) to purified maltose-binding fusion proteins of the cytosolic NH 2 and COOH termini of the three known K ATP channels and the COOH terminus of a ATP-insensitive inward rectifier K ϩ channel, Kir2.1 (25). We provide herein what we believe to be the first evidence of direct binding of ATP to cytosolic domains of the pore-forming subunits of K ATP channels and show that the COOH termini, but not the NH 2 termini, of Kir subunits of K ATP channels bind TNP-ATP. The kinetic analyses of TNP-ATP binding suggest that the COOH termini have a single nucleotide-binding site. Based on glutaraldehyde cross-linking studies, the COOH termini of these three ATP-sensitive channels also exhibit multimerization potential so that they may interact in these intact tetrameric channels.
Production and Purification of Maltose-binding Fusion Proteins-We constructed MBP fusion proteins containing the NH 2 (MBP_1.1N and MBP_6.1N) or the COOH (MBP_1.1C and MBP_6.1C) terminus of rat Kir1.1 and Kir6.1, respectively, and the COOH termini of mouse Kir2.1 (MBP_2.1C) and rat Kir6.2C⌬36 (MBP_6.2C⌬36) channels. We used the MBP_6.2C⌬36 construct for these studies because deletion of the last 36 amino acids from the end of the COOH terminus of Kir6.2 gives rise to functional and ATP-sensitive channel activity in cells in the absence of SUR1 (13). Recombinant proteins were expressed using the pMBPT vector as per the manufacturer's instructions (New England Biolabs). Briefly, 1 liter of Luria-Bertani medium with 0.1 mg/ml ampicillin and 0.5% glucose was inoculated with 10 ml of an overnight culture of Epicurian coli® BL21-CodonPlus TM -RIL-competent cells (Stratagene) expressing the fusion vector and grown to an A 600 of ϳ0.5 at 37°C. Induction was performed with 0.3 mM isopropyl ␤-D-thiogalactoside at 37°C for 2.5 h. The cells were harvested and centrifuged at 4,000 ϫ g for 20 m at 4°C. The cell pellet was resuspended in 50 ml of column buffer (20 mM Tris-Cl, 200 mM NaCl, 1 mM EDTA, pH 7.4) and frozen overnight at Ϫ20°C. The sample was thawed in ice water and lysed with a probe sonicator (four times for 30 s, with 30-s intervals in an ice water bath. The sample was then centrifuged at 9,000 ϫ g for 30 m at 4°C. The supernatant was kept and diluted 1:5 with column buffer. The diluted extract was loaded into a 25-ml column containing 15 ml of amylose resin and washed with 12 column volumes of column buffer. The fusion protein was eluted with column buffer with 10 mM maltose, and 1.5-ml fractions were collected. The protein was detected by UV absorbance at 280 nm, dialyzed against 50 mM Tris-HCl, pH 7.5, and kept at Ϫ80°C until the experiments were performed. The yields of purified recombinant fusion proteins were 15-25 mg/liter. TNP-ATP Binding-To assess the binding of ATP to these recombinant fusion proteins, we used fluorescent TNP-ATP (Molecular Probes, Inc.) (29,30), which has been widely employed to study nucleotide binding to enzymes and other proteins (31)(32)(33)(34). The binding of TNP-ATP to recombinant proteins was performed generally as described by Faller (32). Briefly, 5 M recombinant protein was dissolved in 50 mM Tris-Cl at pH 7.5 or 5 mM MES monohydrate (Sigma) at pH 6.5, and TNP-ATP binding was detected by the increase in fluorescence upon binding to recombinant protein using a SPEX Fluromax-3 spectrofluorometer (Jobin Yvon Inc., Edison, NJ). The fluorescence units reported here were scaled by 1000. Excitation wavelength (403 nm) and emission wavelength (546 nm) were determined for the Kir1.1 COOH terminus fusion protein and utilized for all recombinant proteins (slit widths, 5 nm) because they did not vary significantly among proteins examined. A typical 10-nm blue shift in emission wavelength was detected upon binding of TNP-ATP to proteins (32). The temperature was maintained at 22 Ϯ 0.1°C by a circulating water bath (Neslab, Newington, NH). Incremental additions of TNP-ATP were delivered to polystyrene cuvettes (Elkay Products Inc., Shrewsbury, MA) from stock solutions (0.2-1.0 mM). Total fluorescence was measured 30 s after the additions to allow for equilibration. All of the titrations were corrected for dilution. TNP-ATP fluorescence was also measured in the presence of 5 mM MgATP or by denaturing the protein with 4 M urea. MgATP was added from a stock solution of 0.2 M adjusted to pH 7.5 or 6.5, as indicated.
Free TNP-ATP is weakly fluorescent in buffer, but upon binding to proteins fluorescence is enhanced severalfold with the absolute magnitude dependent on the specific protein environment within the nucleotide-binding pocket (31,32). The fluorescence enhancement factor (␥), TNP-ATP to protein subunit stoichiometry (N o ), and dissociation constant (K d (M)) were determined by least squares fitting to a modified version of the binding equation derived by Faller (32) using GraphPad PRISM TM 3.0 software. The observed fluorescence intensity (F obs ) in arbitrary units is given by the following equation.
where P is the protein concentration (M). Q and Q 2 are constants (fluorescence intensity/M or M 2 of free TNP-ATP, respectively) derived independently from the concentration dependence of TNP-ATP fluorescence intensity in buffer alone (F Buffer ) and account for the "inner filter" effect (32) We independently determined the enhanced factor (␥) by measuring the increase in F obs with increasing protein concentration at a fixed concentration of TNP-ATP (5 M). The F obs data were corrected for light scatter and were fit well by a single exponential. F obs max was determined as F obs at infinite protein concentration when all TNP-ATP would be bound. The enhancement factor was then calculated as follows.
Using this enhancement factor we calculated the concentrations of free ([F]) and bound ([B]) TNP-ATP as described by Moczydlowski and Fortes (31) taking into account the inner filter effect.
Free TNP-ATP is then the difference between total [TNP-ATP] and [B].
Bound versus free TNP-ATP plots were analyzed using a standard binding model that follows mass action.
where B max is the maximal TNP-ATP binding. The data were also plotted for Scatchard or Hill analyses (36) as described (31,37,38). For noncompetitive binding the Scatchard analysis is linear as described by Moczydlowski and Fortes (31).
where N is the number of TNP-ATP binding sites in mol/mg. For MgATP, NaATP, or MgCl 2 competition of TNP-ATP binding, we used a two-site model as described by Faller (39).
where ⌬F obs /⌬F obs max is the fractional change in fluorescence intensity, S frac is the fraction of binding sites in the first site, and K 1 and K 2 are the apparent substrate affinities for the first and second sites, respectively.
8-Azido-[␥-32 P]ATP Labeling-Photoaffinity labeling of recombinant proteins with 8-azido-[␥-32 P]ATP was performed as described previously (40,41). 5 g of the purified protein was added to solution A (50 mM HEPES, 10 mM Tris, pH 7.4, 10 mM CaCl 2 , 0.5 mM MgCl 2 , and 2 Ci of [␥-32 P]azido-ATP; ICN Biochemicals, Inc.) and incubated for 15 min in the dark at 4°C. The reaction mixture was irradiated with UV light at 350 nm for 1 min at room temperature to covalently link the azido-ATP to neighboring amino acid residues. The labeled protein was resolved by SDS-PAGE and visualized by autoradiography.
Cross-linking-Cross-linking of fusion proteins with glutaraldehyde was performed as described previously (42). Briefly, 0.15 g of purified MBP fusion proteins (total volume, 40 l) were incubated with different concentrations (final concentrations, 0, 0.005, 0.01, 0.025, 0.05, 0.075, and 0.1%) of glutaraldehyde in phosphate-buffered saline on ice for 30 min. The cross-linking was quenched with the addition of 100 mM glycine, pH 8.0. The proteins were solubilized in Laemmli buffer with 5% ␤-ME and resolved by SDS-7.5% PAGE. The proteins were transferred to polyvinylidene difluoride membrane (Bio-Rad), blocked with 5% milk in a shaker at room temperature for 1 h, incubated with rabbit anti-MBP antibody (1:10,000; New England Biolabs) overnight at 4°C on a rocker, and then incubated with horseradish peroxidase-conjugated donkey anti-rabbit Ig (1:10,000; Amersham Biosciences) for 1 h at room temperature on a rocker. The proteins were visualized by ECL (Amersham Biosciences).
Electrophysiology-Inside out patch-clamp experiments were performed at room temperature (22-24°C) as described (ϪV p ϭ Ϫ40 mV) (43) to assess the effects of TNP-ATP on apical K ATP channel activity in rat cortical collecting ducts principal cells. Briefly, Sprague-Dawley rats (80 -100g) were obtained from Taconic Farms Inc. and kept on normal chow diet (PMI Nutrition International, Inc.) for 7-10 days before experiments. The animals were euthanized, their kidneys were removed, and coronary slices were cut and placed in ice-cold dissection solution. Individual cortical collecting ducts were dissected at room temperature, and the tubules were immobilized on a 5 ϫ 5-mm cover glass coated with Cell Tac (Becton Dickinson) and then transferred to a perfusion chamber mounted on the stage of an inverted microscope (IMT-2; Olympus). The tubules were opened with a sharpened pipette to gain access to the apical membrane. The principal cells were identified by their hexagonal shape and large flat surface. The bath solution contained 140 mM NaCl, 5 mM KCl, 1 mM EGTA, 10 mM HEPES, 0.2 mM MgATP, pH 7.4. The pipette solution contained 140 mM KCl, 1.8 mM MgCl 2 , 10 mM HEPES, pH 7.4. TNP-ATP (0 -1000 M) was added to the bath solution where indicated. MgATP is required in the bath solution to keep the K ATP channels in principal cells from running down (43).
Chemicals-All of the chemicals were research grade or better and were from Sigma unless otherwise stated.

RESULTS
ATP Binds to the COOH Terminus of Kir1.1-All MBP fusion proteins were efficiently expressed in bacteria and could be highly purified at milligram quantities (5-25 mg/liter of bacterial culture) without exposure to detergents or denaturing agents (28). The recombinant MBP and the NH 2 -terminal (MBP_1.1N and MBP_6.1N) and COOH-terminal (MBP_1.1C, MBP_6.1C, MBP_6.2C⌬36, and MBP_2.1C) MBP fusion proteins ran at their expected molecular masses as shown in Fig.  1. MBP_6.2C⌬36 consistently produced the lowest yield of 5-10 mg/liter, whereas the yields of MBP_1.1C and MBP_6.1C were 15-25 mg/liter. Cleaving the MBP from the channel protein at the thrombin site resulted in insoluble protein under our current buffer conditions, probably because of the hydrophobicity of these cytosolic NH 2 and COOH termini. Thus, all of the experiments were performed using the MBP fusion proteins.
We used fluorescent TNP-ATP to assess the binding of ATP to the cytosolic domains of Kir channels (31)(32)(33)(34). The concentration dependence relationships of TNP-ATP fluorescence with MBP_1.1C, MBP_1.1N, and MBP alone at pH 7.5 are shown in Fig. 2. F obs for unbound TNP-ATP in buffer without protein was low and increased in a nonlinear, concentrationdependent manner (Fig. 2, A and B), consistent with the intrinsic fluorescence of this ATP analogue and the inner filter effect (29,30,31). All of the buffer data were well fit using a second order polynomial that accounts for this inner filter effect (see "Materials and Methods"; r 2 Ն 0.99). In contrast, F obs was significantly enhanced over the buffer control in the presence of MBP_1.1C (Fig. 2, A and B, F P ), consistent with binding of TNP-ATP to this fusion protein. (1:1) did not significantly affect the affinity for TNP-ATP binding (control K d ϭ 1.84 Ϯ 0.14, (n ϭ 6); mixing K d ϭ 1.63 Ϯ 0.22; (n ϭ 5); data not shown).
Further support for nucleotide binding to MBP_1.1C was obtained by photoaffinity labeling by 8-azido-[␥-32 P]ATP as shown in Fig. 3A. The 8-azido-[␥-32 P]ATP labeling was competed with unlabeled MgATP consistent with specific labeling of MBP_1.1C with this nucleotide analogue. We also examined the ability of MgATP to compete the TNP-ATP binding to MBP_1.1C. The TNP-ATP concentration-dependent increase in F obs with MBP_1.1C was reduced by 5 mM MgATP (Fig. 3B, triangles), and the K d for TNP-ATP binding affinity was significantly increased; K d increased from 3.0 Ϯ 0.2 (F P ) to 6.9 Ϯ 1.9 (F P 5 mM MgATP ; n ϭ 13). Increasing MgATP concentration to 50 mM virtually abolished TNP-ATP fluorescence enhancement with MBP_1.1C (K d ϭ 50.9 Ϯ 14.7 M; Fig. 3B; n ϭ 5). We also assessed the competition of TNP-ATP binding to MBP_1.1C by MgATP (Fig. 3C). Increasing concentrations of MgATP reduced ⌬F obs /⌬F obs max in a concentration-dependent manner. The shape of the MgATP competition curve was complex, suggesting multiple binding interactions; the data were well fit, however, using the two-site model described by Equation 7 (r 2 ϭ 0.99). K 1 and K 2 were 71 Ϯ 5 and 3.8 Ϯ 0.8 mM, respectively, and S frac was 0.77 Ϯ 0.02.
A fraction of MgATP will dissociate in our buffer solution to free Mg 2ϩ and ATP anion (43), and Mg 2ϩ has been shown to modulate TNP-ATP binding or fluorescence enhancement in several nucleotide-binding proteins ( 3C). MgCl 2 reduced ⌬F obs /⌬F obs max in a concentration-dependent manner to 48% of the control with an EC 50 of 61 Ϯ 2 M, a value virtually identical to K 1 observed with MgATP competition. This result suggests that the MgATP competition curve is composed of both free Mg 2ϩ (K 1 ) and MgATP/ATP anion (K 2 ) components. Accordingly, the EC 50 for MgATP competition of TNP-ATP binding to MBP_1.1C is 3.8 mM (K 2 ). This EC 50 value (K 2 ) for MgATP competition is consistent with the ϳ50% reduction in TNP-ATP binding by 5 mM MgATP shown in Fig. 3B and with our previous observations of MgATP inhibition of Kir1.1 channel activity expressed in Xenopus laevis oocytes (EC 50 of ϳ3.5 mM) (24).
We also assessed the ability of NaATP to compete TNP-ATP binding to MBP_1.1C. In contrast to MgATP, NaATP has little effect on the activity of either the Kir1.1 channel expressed in oocytes (24) or the native kidney K ATP channel (43) at concentrations less than 10 mM. As shown in Fig. 3C, NaATP reduced TNP-ATP fluorescence in a concentration-dependent manner; however, 20 mM NaATP reduced ⌬F obs /⌬F obs max by only 66 Ϯ 2%. The competition data were well fit by either a single-site or a two-site model (Equation 7) yielding an estimated EC 50 of Ն17.5 Ϯ 2.6 mM ( Fig. 3C; n ϭ 7). Based on the low affinity of NaATP, the K 2 value for MgATP competition (3.8 mM) was likely due to MgATP complex rather than ATP anion. Thus, the affinity profile for ATP binding to MBP_1.1C is: TNP-ATP Ͼ Ͼ (Mg 2ϩ ) MgATP Ͼ Ͼ NaATP. The TNP-ATP affinity for some other nucleotide-binding proteins is also higher than for unmodified ATP (34,39,46).

TNP-ATP Inhibits the Secretory K ATP Channel in Principal Cells of Rat Cortical Collecting Duct with Higher Affinity than
MgATP-Given our biochemical evidence for direct binding of TNP-ATP to MBP_1.1C with a higher affinity than MgATP, we assessed TNP-ATP inhibition of native K ATP channel activity believed to be formed by Kir1.1 (43). Inside-out patches from apical membranes of rat principal cells containing the typical low conductance K ϩ channels (SK) were exposed to varying TNP-ATP concentrations. Fig. 4A shows a representative trace from an inside-out excised apical patch demonstrating that 1 mM TNP-ATP added to the bath (cytosolic side) reversibly inhibited SK channel activity. The TNP-ATP concentration-dependent inhibition of the SK channel is shown in Fig. 4B (n ϭ  6). The EC 50 for channel inhibition was 170 M, a value three to four times lower than for unmodified MgATP (43). This EC 50 is consistent with the observed affinity for TNP-ATP binding to MBP_1.1C being greater than for MgATP (Fig. 3C). It is likely, however, that the affinity for TNP-ATP inhibition of the SK channel was underestimated in these experiments because 0.2 mM MgATP (and free Mg 2ϩ ; TNP-ATP competitors) was present in the bath solution to keep these K ATP channels from running down (43).
The Kinetics and Stoichiometry of TNP-ATP Binding to MBP_1.1 at pH 7.5 and 6.5-The TNP-ATP to MBP_1.1C protein stoichiometry can be estimated from the TNP-ATP concentration-dependent increases in F obs shown in Fig. 2B. Using Equation 1 (32), the stoichiometry (N o ) for TNP-ATP binding to MBP_1.1C was 0.89 Ϯ 0.02 mol of TNP-ATP/mol of protein (n ϭ 11). An additional estimate of N o can be made from the intersection of linear fits to the initial and final F obs values as suggested by Faller (32). This is possible because F obs initially increased linearly with 0 -1 M TNP-ATP concentrations, indicating that nearly all of the TNP-ATP was bound to the fusion protein over this range and was flat at TNP-ATP concentration above 15 M (Fig. 2B, dashed lines; r 2 ϭ 0.99; n ϭ 11; p Ͻ 0.001). The intersection gave a maximal TNP-ATP binding of 4.1 M at a MBP_1.1C protein concentration of 5 M (Fig. 2B) 6) was 11.6 Ϯ 0.2 nmol of TNP-ATP bound per mg of protein with a 95% CI of 11.1-12.0. Based on the calculated molecular weight of MBP_1.1C (15.06 nmol/mg), the stoichiometry (mol of TNP-ATP/mol of protein) for TNP-ATP binding to MBP_1.1C was 0.77, ranged from 0.74 to 0.80 (95% CI), and was similar to that derived using Equation 1 from the F obs data in Fig. 2B.
TNP-ATP binding to MBP_1.1C was significantly enhanced by lowering pH from 7.5 to 6.5 and reducing the salt concentration from 50 mM Tris-Cl to 5 mM MES (Fig. 5). The enhancement factor, calculated from the MBP_1.1C protein titration of 5 M TNP-ATP, increased from 7.7 Ϯ 0.3 at pH 7.5 (Fig. 5A) to 34.5 Ϯ 1.6 at pH 6.5 ( Fig. 6A; n ϭ 5). The K d calculated from the TNP-ATP concentration dependence of F obs at pH 6.5 ( Fig. 6B; n ϭ 5) using Equation 1 and a ␥ value of 34.5 was 1.0 Ϯ 0.1 M or less than half of the K d at pH 7.5. Fit of the Scatchard data by Equation 6 gave a similar TNP-ATP binding affinity of 1.0 Ϯ 0.1 M (Fig. 6C).
TNP-ATP Binding to the COOH Terminus of Kir6.1-The binding of TNP-ATP to the COOH terminus, but not the NH 2 terminus, of Kir1.1 (Fig. 2) and the photoaffinity labeling of MBP_1.1C by 8-azido-[␥-32 P]ATP (Fig. 3A) Fig. 7 shows that F obs increased in a TNP-ATP concentrationdependent and saturable fashion with MBP_6.1C (Fig. 7A) but not MBP_6.1N (Fig. 7B) at a pH of 7.5. Both 5 mM MgATP and 4 M urea significantly reduced the increase in F obs with MBP_6.1C, but urea had little effect on the low F obs with MBP_6.1N (Fig. 7B). Given that the COOH termini of both Kir1.1 and Kir6.1 bind TNP-ATP, we assessed the specificity for TNP-ATP interactions with K ATP COOH termini by determining TNP-ATP-dependent increases in F obs with the COOH terminus of an ATP-insensitive inward rectifier K ϩ channel (Kir2.1; MBP_2.1C) (25). The TNP-ATP concentration-dependent increases in F obs with MBP_2.1C (Fig 7C) were small and unaffected by 5 mM MgATP or 4 M urea. Thus, unique amino acid sequence(s) specific to the COOH termini of these K ATP channels determines their ability to bind nucleotides.
The kinetics of TNP-ATP binding to MBP_6.1C at pH of 7.5 is shown in Fig. 8. MBP_6.1C titration of 5 M TNP-ATP (Fig.  8A) yielded a ␥ of 17.3 Ϯ 0.7 (n ϭ 3), significantly higher than for MBP_1.1C (7.7 Ϯ 0.3; Fig. 5A). The K d and stoichiometry for TNP-ATP binding to MBP_6.1C were calculated by fitting the F obs data (Fig. 8B) to Equation 1 using a ␥ of 17.3 (Fig. 8A) Fig. 8C; the K d calculated from Equation 6 was 3.4 Ϯ 0.3, indistinguishable from that derived using Equation 1 from the F obs data in Fig. 8B. The TNP-ATP binding stoichiometry derived from Equation 6 was 12.16 Ϯ 0.36 with a 95% CI of 11.3-13.0. Based on the calculated molecular weight of MBP_6.1C (14.33 nmol/mg), the stoichiometry (mol of TNP-ATP/mol of protein) for TNP-ATP binding to MBP_6.1C was 0.87, and ranged from 0.79 to 0.91 (95% CI), and was similar to that derived using Equation 1 from the F obs data in Fig. 8B.
The kinetics of TNP-ATP binding to MBP_6.1C at pH 6.5 is shown in Fig. 9. Similarly to MBP_1.1C (Fig. 6), the fluorescence enhancement factor for TNP-ATP binding to MBP_6.1C at pH of 6.5 was significantly increased over that at pH 7.5: ␥ ϭ 32.8 Ϯ 0.8 ( Fig. 9A; n ϭ 5; pH 6.5) versus 17.3 Ϯ 0.7 ( Fig. 8A; pH 7.5; p Ͻ 0.01). The TNP-ATP concentration-dependent increase in F obs upon binding to MBP_6.1C at pH 6.5 is shown in Fig. 8B. The F obs data were well fit by Equation 1 (r 2 ϭ 0.998) using the ␥ of 32.8 and gave a K d ϭ 1.0 Ϯ 0.1 M, a significantly higher affinity than for TNP-ATP binding to MBP_6.1C at a pH of 7.5 (Fig. 8B). The increase in F obs at pH 6.5 was abolished by 4 M urea (data not shown). The ␥ of 32.8 was used to calculate TNP-ATP Binding to the COOH Terminus of Kir6.C⌬36 -Previous reports (17,18) have suggested that ATP can directly interact with Kir6.2, based on the photoaffinity labeling of the entire Kir6.2 channel subunit by 8-azido-[␥-32 P]ATP. In addition, mutations in the COOH terminus alter the EC 50 for ATP inhibition of channel activity (13, 19 -23). We used the COOH terminus of the functional, ATP-sensitive deletion mutant Kir6.2C⌬36 [MBP_6.2C⌬36] (13) to assess TNP-ATP binding to Kir6.2 (Fig. 10). F obs values increased in a concentration-dependent manner with MBP_6.2C⌬36 and were significantly enhanced over the buffer (Fig. 10A; F B ) at either pH of 7.5 (Fig.  10A, white squares and dashed line) or 6.5 (Fig. 10A, solid  squares and solid line). The TNP-ATP concentration-dependent increases in F obs were significantly reduced by 5 mM MgATP or 4 M urea ( Fig. 10B; pH 6.5 shown; similar results were obtained at pH 7.5 but are not shown). The ␥ value was significantly increased at pH 6.5 to 46.1 Ϯ 0.3 ( Fig. 10C; n ϭ 3) compared with 11.4 Ϯ 0.3 at pH 7.5 ( Fig. 10C; n ϭ 5). K d for TNP-ATP binding to MBP_6.2C⌬36 at pH 7.5 and 6.5 calculated using Equation 1 for the F obs data in Fig. 10A were, at pH 7.5, K d ϭ 6.8 Ϯ 0.6 M, N o ϭ 0.51 Ϯ 0.02 mol of TNP-ATP/mol of protein and, at pH 6.5, K d ϭ 1.4 Ϯ 0.1 M. Scatchard plots of the bound and free TNP-ATP concentrations at both pH values are shown in Fig. 10D. K d and N values calculated using Equation 6 were, at pH 7.5, K d ϭ 4.9 Ϯ 0.2 M, n ϭ 6.86 Ϯ 0.14 nmol/mg and, at pH 6.5, K d ϭ 1.6 Ϯ 0.1 M. Using the calculated molecular weight of MBP_6.2C⌬36 (15.90 nmol/mg) yielded a stoichiometry (N o ) of 0.43 mol of TNP-ATP/mol of protein (pH 7.5), consistent with one TNP-ATP-binding site/Kir6.2 COOH terminus with ϳ50% of the protein being active.
Multimerization Potential of MBP_1.1C, MBP_6.1C, and MBP_6.2C⌬36 -K ATP channel pores are formed of four identical Kir subunits (47,48). To assess whether the COOH termini of MBP_1.1C, MBP_6.1C, and MBP_6.2C⌬36 proteins have the capacity to self-assemble into oligomers in the absence of the NH 2 termini and transmembrane spanning segments and the pore, we analyzed dilute solutions of these fusion proteins by SDS- PAGE in the presence of dithiothreitol (DTT) followed by Western blotting using anti-MBP as described (42). In the absence of cross-linking agents and disulfide bond formation, the three fusion proteins exhibited oligomeric structures (Fig. 11, A, C, and D, first lanes). Oligomerization was enhanced with crosslinking using glutaraldehyde (Fig. 11). At concentrations of glutaraldehyde of 0.005-0.025%, the trimer and tetrameric forms became dominant. At high concentrations of glutaraldehyde (Ն0.05%), higher order multimers were produced that either did not enter the gel or migrated near the top of the gel. The oligomerization of these proteins was specific for the COOH termini because MBP has been shown not to oligomerize with glutaraldehyde concentrations up to 1% under our conditions (42).
Although oligomerization of the COOH termini of these fusion proteins does not depend on disulfide bridge formation, Kir1.1 channels are redox-sensitive with pH-mediated channel closure resulting in exposure of a COOH-terminal cysteine (Cys 308 ) that forms a disulfide bond and locks the channel in the closed state (49). Therefore, we examined whether reducing agents alter TNP-ATP concentration-dependent increase in F obs with MBP_1.1C. One mM DTT with 10 mM ␤-ME did not significantly change the K d for TNP-ATP binding to MBP_1.1C (Fig. 11B): ϪDTT/␤-ME, 2.7 Ϯ 0.3 M, n ϭ 23; ϩDTT/␤-ME, 1.8 Ϯ 0.2 M, n ϭ 3.

DISCUSSION
Our results provide direct evidence for high affinity TNP-ATP binding to the cytosolic COOH-terminal domains of the pore-forming subunits of K ATP channels: Kir1.1, Kir6.1, and Kir6.2⌬C36. NH 2 termini of the Kir1.1 and Kir6.1 K ATP channels did not bind TNP-ATP, demonstrating that the nucleotidebinding domain is restricted to COOH termini. A summary of TNP-ATP binding to these COOH termini is shown in Fig. 12. Fig. 12A shows the relative increases in ⌬F obs /⌬F obs max for all three COOH termini at both pH 7.5 and 6.5. The higher affinities for TNP-ATP binding at pH 6.5 are apparent. Fig. 12B shows the Scatchard plots and summaries of the stoichiometry (N o ) and K d values. The TNP-ATP affinity profile at pH 7.5 was: MBP_1.1C Ͼ MBP6.1C Ͼ MBP6.2C⌬36. At a pH of 6.5, however, the K d values for all three proteins were similar at ϳ1 M. Reducing pH to 6.5 also increased the enhancement factor (␥) for TNP-ATP binding.
Several lines of evidence indicate that the COOH termini of K ATP channels are necessary and sufficient to bind TNP-ATP. First, neither MBP alone (Fig. 2C)   See the legend to Fig. 9 for a general explanation. A, the TNP-ATP concentration-dependent increases in F obs with 5 M MBP_6.2C⌬36 (F P ) at pH 7.5 (white squares) and pH 6.5 (black squares). The solid lines were calculated according to Equation 1. F obs was significantly more enhanced at pH 6.5 than at 7.5. Intersection of linear fits of initial and final TNP-ATP concentrations (dotted lines) is shown (see text for discussion). For comparison, F obs is shown for TNP-ATP in buffer (diamonds and dashed line, second order polynomial). B, denaturing the MBP_6.2C⌬36 fusion protein at pH 6.5 with 4 M urea (black inverted triangles and dashed line; F P Urea ) or addition of 5 mM MgATP (black triangles and dashed line; F P MgATP ) significantly reduced the increases in F obs . C, MBP_6.2C⌬36 protein titration of 5 M TNP-ATP at pH values of 7.5 (diamonds) and 6.5 (squares). F obs was corrected for protein light scatter. The lines were calculated using an exponential fits, and fluorescence at infinite protein concentration (Pϱ) was determined. ␥ was calculated as F obs Buffer /F Pϱ Buffer . The gray bar represents the intrinsic fluorescence of 5 M TNP-ATP in buffers at pH of 7.5 and 6.5. D, Scatchard plots for TNP-ATP binding to MBP_6.2C⌬36 at pH of 7.5 (squares) and 6.5 (diamonds). The lines were calculated according to Equation 6. mutations could influence nucleotide binding in the functional tetrameric channel as a result of interactions between the NH 2 and COOH termini that was not observed in our mixing experiments using MBP fusion proteins.
The affinities for TNP-ATP binding to the K ATP channel COOH termini (Fig. 12B) were higher than the IC 50 values for ATP-mediated inhibition of Kir6.x or Kir1.1 channels exogenously expressed in oocytes or mammalian cells (1,3,4,24). TNP-ATP binds to several other proteins with higher affinity than unmodified ATP (31,32). In this regard, the EC 50 for MgATP competition (Fig. 3C) was Ͼ10 3 higher than the K d for TNP-ATP binding to MBP_1.1C, but similar to the EC 50 (ϳ3.5 mM) for MgATP inhibition of Kir1.1 expressed in oocytes (24). Coexpression of Kir1.1 with fibrosis transmembrane conductance regulator in Xenopus oocytes significantly enhances the sensitivity of the channel to nucleotides (EC 50 Ϸ 0.5 mM) (9), similar to the native renal K ATP channel (43). Similarly, the IC 50 for ATP inhibition of Kir6.2 is reduced from ϳ100 to 10 M by coexpression with SUR1 (50), a value that is similar to the K d for TNP-ATP binding to MBP_6.2C⌬36 (Fig. 12B). Thus, interactions with fibrosis transmembrane conductance regulator or SUR either modify the binding of nucleotides to COOH termini or enhance channel gating following nucleotide binding. In addition, phosphatidylinositol phospholipids, like phosphatidylinositol 4,5-bisphosphate, can dramatically reduce ATP sensitivity of native or exogenously expressed K ATP channels (51)(52)(53). We have recently shown that phosphatidylinositol phosphates, like phosphatidylinositol 4,5-bisphosphate, compete off TNP-ATP binding to MBP_1.1C and MBP_6.1C and could account for the effect of phosphatidylinositol 1,4,5bisphosphate to enhance channel activity (54).
MgATP competition of TNP-ATP binding to MBP_1.1C was complex (Fig. 3) and indicated that at least two distinct binding events were occurring: one for Mg 2ϩ with M affinity and one for MgATP with mM affinity similar to that found for MgATP inhibition of Kir1.1 expressed in Xenopus oocytes (24). The displacement of TNP-ATP binding to MBP_1.1C by free Mg 2ϩ (Fig. 3C) and the ability of a saturating concentration of free Mg 2ϩ (1 mM) to decrease TNP-ATP binding affinity nearly 4-fold without a decrease in ␥ (Fig. 3D) is consistent with a direct effect of Mg 2ϩ to compete TNP-ATP binding to MBP_1.1C. The influence of free Mg 2ϩ on TNP-ATP interactions has been well documented with other proteins that bind MgATP (32,34,39,44,45). Although the specific mechanism by which Mg 2ϩ alters TNP-ATP binding cannot be deduced from our present studies, both polyanionic and polycationic (including Ca 2ϩ and Mg 2ϩ ) charges affect nucleotide gating of K ATP channels (55,56). Our present results are consistent with the suggestion by Deutsch et al. (55) that Mg 2ϩ modulation of surface charge on K ATP channels influences nucleotide interactions and gating.
Lowering pH from 7.5 to 6.5 enhanced TNP-ATP binding affinity (Fig. 12B) for each of the K ATP channel COOH termini. This was associated with significant increases in ␥ values (Fig.  12A), consistent with pH-dependent alterations in the specific protein environments surrounding the nucleotide binding pockets in these COOH termini. Interestingly, although ␥ values for TNP-ATP binding were significantly different at pH 7.5 (MBP_6.1C Ͼ MBP_6.2C⌬36 Ͼ MBP_1.1C; Fig. 12A), the ␥ at pH 6.5 were similar. The latter would suggest that the protein environments forming the nucleotide-binding pockets may be fundamentally similar in these COOH termini. The effect of lowering pH on ␥ suggests that residues titratable over the range 7.5-6.5 are involved in TNP-ATP binding. Further analysis will require defining the specific amino acid residues involved in nucleotide binding, which is now feasible using our current approach.
The stoichiometries of 0.4 -0.9 mol of ATP/mol of protein for TNP-ATP binding to the COOH termini of K ATP channels indicate that each COOH terminus is capable of binding at least one TNP-ATP molecule. Given the homotetramerization of Kir subunits to form K ATP channels, the simplest model based on our results would be a channel in which each of the four COOH termini contributes an identical single ATP-binding site. This model is consistent with the proposed ATP stoichiometry of 1:4 for Kir6.2 channel gating in a recent study (57) based on mixing experiments with wild-type and G334D, ATP-insensitive, mutant constructs. The authors suggested that ATP interaction with one of four identical sites was sufficient to gate the channel closed. Nichols and co-workers (23) have suggested that only two sequential nucleotide-binding steps are required to give the steep dependence of Kir6.2 channel activity on ATP. Moreover, in a subsequent communication (58), Nichols' group demonstrated that a linear four-site model predicted the data for ATP sensitivity of Kir6.2 ϩ SUR1 channels better than a one-site linear model. The oligomerization potential for K ATP COOH termini demonstrated in Fig. 11 suggests the possibility that COOH termini could interact directly in intact channels and that this association could modulate nucleotide binding or gating. The notion that the COOH terminus of K ATP channels may form a tetrameric nucleotide-binding domain is also supported by the proposed tetrameric structural models of the COOH-terminal domains of the bacterial KcsA (59) and the plant ATK1 (35) potassium channels.