ATP and AMP Mutually Influence Their Interaction with the ATP-binding Cassette (ABC) Adenylate Kinase Cystic Fibrosis Transmembrane Conductance Regulator (CFTR) at Separate Binding Sites*

Background: Cystic fibrosis transmembrane conductance regulator (CFTR) has adenylate kinase activity (ATP + AMP ⇆ 2 ADP). Results: ATP enables CFTR photolabeling by 8-N3-AMP, and AMP increases 8-N3-ATP photolabeling at ATP-binding site 2. Conclusion: AMP interacts with CFTR in an ATP-dependent manner and alters ATP interaction with the adenylate kinase active center ATP-binding site. Significance: These findings exemplify nucleotide interactions with an ABC adenylate kinase.

the presence of physiologically relevant concentrations of AMP, adenylate kinase activity regulates CFTR channel function (19 -22). The data further suggested that AMP interacted with a binding site distinct from the two ATP-binding sites and that ATP at ATP-binding site 2 was involved in the adenylate kinase activity (19,20,22). Both ATPase and adenylate kinase activity have been reported for two other ABC proteins, Rad50 (23) and SMC (structural maintenance of chromosome protein) (24). The active center of an adenylate kinase comprises separate ATP-and AMP-binding sites (25)(26)(27). The first structural view of an ABC adenylate kinase became available when Lammens and Hopfner (24) solved the crystal structure of the NBD of the Pyrococcus furiosus SMC protein in complex with the adenylate kinase inhibitor P 1 ,P 5 -di(adenosine-5Ј) pentaphosphate (Ap 5 A). Ap 5 A contains two adenosine groups connected by five phosphate groups, allowing it to bind simultaneously to the ATP-and the AMP-binding site (28,29). These features make it a valuable experimental probe for an adenylate kinase active center. The SMC-NBD structure showed the two adenosine moieties of Ap 5 A attached to two binding sites separated by ϳ15 Å. A Mg 2ϩ ion, one adenosine, plus ␣-, ␤-, and ␥-phosphates of Ap 5 A bound the canonical Mg 2ϩ -ATP-binding site on lobe I of the NBD. The other adenosine stacked onto the side chain of a conserved glutamine of the Q-loop at the interface of lobe I and lobe II. The authors also examined nucleotide-induced conformational changes. ATP did not alter the NBD structure. In contrast, binding of Ap 5 A rotated the two lobes of the NBD by ϳ15°. These results suggest that adenylate kinase activity induces distinct conformational changes in SMC.
Although ATP and AMP bind separate sites, binding of one nucleotide influences the interaction with the other in non-ABC adenylate kinases. A number of different experimental approaches showed that ATP greatly increased binding of AMP to the AMP-binding site (30 -33). Other studies demonstrated that binding of AMP to the AMP-binding site induced conformational changes in the ATP-binding domain (34 -37).
Based on data from non-ABC adenylate kinases, we hypothesized that ATP and AMP mutually influence their interaction with CFTR at separate binding sites (Fig. 1). Based on the data from CFTR, we further hypothesized that only one of the two ATP-binding sites of CFTR is involved in adenylate kinase activity. To test these hypotheses, we employed the patch clamp technique with excised membrane patches containing CFTR and also photolabeling of CFTR using radioactive nucleotides with a photoactivatable azido (N 3 )-group attached to the adenine ring. The N 3 -group absorbs UV light, resulting in photolysis and formation of a reactive intermediate, which reacts covalently with nearby amino acid residues (38,39). An advantage of these approaches is that they allow the interactions of nucleotides with CFTR to be studied while CFTR is embedded in the membrane.

EXPERIMENTAL PROCEDURES
Materials-The radioactive, photoactivatable azido-nucleotides, dissolved as triethylammonium salt in absolute methanol, were from Affinity Photoprobes, LLC (Lexington, KY). Immediately before use, the methanol was evaporated under a stream of argon, and the azido-nucleotide was dissolved in a buffer of 20 mM Hepes (pH 7.5), 50 mM NaCl, 3 mM MgCl 2 . Non-radioactive ATP, AMP, P 1 ,P 4 -di(adenosine-5Ј) tetraphosphate (Ap 4 A), and Ap 5 A were from Sigma-Aldrich. ATP was used as magnesium salt. AMP and Ap 5 A were sodium salts, and Ap 4 A was an ammonium salt. The protease inhibitors used in this study were purchased from Sigma-Aldrich. Endoproteinase Arg-C from Clostridium histolyticum (sequencing grade) was from Roche Applied Science. Protein kinase A, catalytic subunit (PKA), purified from bovine heart, was from EMD Millipore Corp. (Billerica, MA). The monoclonal CFTR antibodies used were from R&D Systems, Inc. (Minneapolis, MN) (13-1 (40)), EMD Millipore (M3A7, L12B4 (41) and 13-4 (42)), and the Cystic Fibrosis Foundation in conjunction with the University of North Carolina (Chapel Hill, NC) (596 (42)).
CFTR Azido-nucleotide Photolabeling-Membranes (containing 25-35 g of protein to detect photolabeling of fulllength CFTR or 150 -270 g of protein to detect labeling of proteolytic CFTR fragments) were incubated on ice in 20 mM Hepes (pH 7.5), 50 mM NaCl, 3 mM MgCl 2 with the radioactive azido-nucleotide and non-radioactive nucleotides as described in the figure legends in a total volume of 30 l. Individual reactions were then irradiated with UV light (302 nm, 8-watt lamp) at a distance of 5 cm unless otherwise indicated. The duration of UV irradiation varied between the different azido-nucleotides and is indicated in the figure legends. After exposure to UV light, 20 l of stop buffer (25 mM dithiothreitol, 4% SDS, 20 mM Hepes (pH 7.5), 50 mM NaCl, 125 g/ml benzamidine, 4 g/ml aprotinin, 2 g/ml leupeptin, 100 g/ml Pefabloc, 7 g/ml E-64) followed by 875 l of solubilization buffer (1% Triton X-100 in 20 mM Hepes (pH 7.5), 50 mM NaCl, 125 g/ml benzamidine, 4 g/ml aprotinin, 2 g/ml leupeptin, 100 g/ml Pefabloc, 7 g/ml E-64) were added. Samples were stored at Ϫ80°C overnight and thawed on ice before adding CFTR antibodies for immunoprecipitation. CFTR was immunoprecipitated as described (22) using monoclonal CFTR antibodies to its regulatory domain (13-1; 0.2 g/sample) and NBD2 (M3A7; 1 g/sample).
Gel Electrophoresis and Autoradiography-Immunoprecipitated full-length CFTR was fractionated on 6 or 8% SDS-polyacrylamide gels. Proteolytic digestion products of CFTR were fractionated on Novex 16% Tricine gels (Invitrogen) (44). After electrophoresis, gels were either dried and then subjected to digital autoradiography using a FLA-7000 imaging system (Fuji Photo Film Co., Ltd., Tokyo, Japan) or transferred onto a PVDF membrane (Immobilon-FL transfer membrane, EMD Millipore, Billerica, MA) for Western blotting. For quantitative image analysis, Multi Gauge analysis software (version 3.0; Fuji Photo Film) was used. Region intensities corresponding to protein or peptide bands were quantified in linear arbitrary units. Background intensities were subtracted.
Patch Clamp Experiments-CFTR Cl Ϫ currents were studied using excised, inside-out membrane patches from HeLa cells transiently expressing either wild-type or mutant CFTR using a vaccinia virus/T7 RNA polymerase expression system (43) as described previously (20,45). The pipette (extracellular) solution contained 140 mM N-methyl-D-glucamine, 2 mM MgCl 2 , 5 mM CaCl 2 , 100 mM L-aspartic acid, and 10 mM Tricine, pH 7.3, with HCl. The bath (intracellular) solution contained 140 mM N-methyl-D-glucamine, 3 mM MgCl 2 , 1 mM CsEGTA, and 10 mM Tricine, pH 7.3, with HCl. Following patch excision, CFTR channels were activated with 25 nM protein kinase A catalytic subunit (PKA) and ATP. PKA was present in all cytosolic solutions that contained ATP. Experiments were performed at room temperature (23-26°C). Macropatch recordings were low pass-filtered at 100 Hz for analysis using an 8-pole Bessel filter (model 900, Frequency Devices, Inc. (Haverhill, MA)). Recordings from patches containing very few channels (A462F CFTR) with up to five simultaneous channel openings were low pass-filtered at 500 Hz for analysis. For these recordings, NP o was determined using the pCLAMP software package (version 9.2, Axon Instruments, Inc., Union City, CA).
Construction of a Three-dimensional Model of the NBD1-NBD2 Heterodimer-We constructed a putative molecular model of the human CFTR NBD1-NBD2 heterodimer in the following way. The crystal structure of NBD1 in complex with ATP (PDB code 1R0X) (46) was used as a template for constructing a homology model of NBD2 using the modeling program SWISSMODEL (47). Missing loop regions for NBD2, which were identified using the SEQATOMS Web server (48), were added using the loop-modeling program Loopy (49). Following the work of Callebaut et al. (50), a model of the CFTR NBD1-NBD2 heterodimer was assembled using the homodimeric structure of the MJ0796 ABC transporter from Methanocaldococcus jannaschii as a template (PDB code 1L2T) (9). Specifically, the NBD1-ATP crystal structure was superimposed onto chain A of the 1L2T structure using the "Iterative Magic Fit" tool in the modeling program Swiss-PdbViewer (51), with the C␣ coordinates being used for the superposition. Similarly, the homology-modeled NBD2-ATP structure was superimposed onto chain B of the 1L2T structure. The resulting model of the NBD1-NBD2 heterodimer, together with the bound ATP molecules, is shown in Fig. 10.
Data Presentation and Statistics-Data are presented as means Ϯ S.E. p values of Ͻ0.05 were considered statistically significant. For statistical analysis, SigmaStat software (version 3.0, SPSS Inc., Chicago, IL) was used.

ATP Enhances 8-N 3 -AMP Photolabeling and AMP Enhances 8-N 3 -ATP Photolabeling-
To test the hypothesis that ATP and AMP mutually influence their interaction with CFTR at separate binding sites, we expressed CFTR in HeLa cells using a double vaccinia virus/T7 RNA polymerase system (43) and collected cell membranes (22). Western blot analysis confirmed the presence of CFTR ( Fig. 2A). The majority of the CFTR protein migrated as the highly glycosylated band C (40,52). To investigate the interactions of ATP and AMP with CFTR, we used photolabeling with their azido (N 3 ) nucleotide analogues. For ease of reading, we will refer to 8-N 3 -[ 32 P]AMP as *AMP, and to 8-N 3 -[␣-32 P]ATP and 8-N 3 -[␥-32 P]ATP as *ATP.
We tested the interaction of AMP with CFTR using *AMP. We detected photolabeling that increased in the presence of ATP (Fig. 2, compare lanes 4 and 5 in B, lanes 4 and 6 in C, and lanes 8 and 11 in C). Photolabeling in the absence of ATP was not reduced by non-radioactive AMP (Fig. 2C, compare lanes 4 and 5). On the contrary, photolabeling increased in the presence of AMP, a phenomenon we will discuss below. Ap 4 A and Ap 5 A also failed to reduce labeling (Fig. 2C, compare lane 1 with lanes 2 and 3). These data indicate that when *AMP is present alone, there is no specific photolabeling of an AMPbinding site. In the presence of ATP, non-radioactive AMP reduced photolabeling (Fig. 2C, compare lanes 6 and 7). Fig. 2D shows quantitative data. These results suggest specific labeling of an AMP-binding site in the presence of ATP. They further indicate that ATP-and AMP-binding sites are not identical.
These data suggest that AMP interacts with CFTR in an ATP-dependent manner. They further suggest that AMP alters the interaction of ATP with CFTR.

Ap 5 A and Ap 4 A Inhibit 8-N 3 -AMP and 8-N 3 -ATP Photolabeling-As an additional test for interactions of AMP and ATP
with sites that may be involved in adenylate kinase activity, we asked whether Ap 5 A and Ap 4 A would inhibit photolabeling by *AMP (in the presence of ATP) and *ATP. Ap 4 A, like Ap 5 A, is a double substrate adenylate kinase inhibitor that binds simultaneously to an ATP-and an AMP-binding site (PDB code 2C95). 4 We previously showed that both Ap 4 A and Ap 5 A inhibit CFTR cur-  The sample of lane 5 was not UV-irradiated. CFTR was solubilized, immunoprecipitated, and fractionated on 8% (lanes 1-4) and 6% (lanes 5-8) SDSpolyacrylamide gels as described under "Experimental Procedures." B, summary data. Experiments were performed as in A. Radioactivity incorporated into CFTR was normalized to radioactivity for the conditions indicated below bars 2 and 6. *, p Յ 0.001 compared with control (bars 2 and 6) (one-way repeated measures ANOVA followed by Holm-Sidak's method of multiple comparisons versus control group, n ϭ 6 -7). Error bars, S.E. rent (19). We found that both compounds inhibited *AMP labeling (Fig. 2C, compare lanes 9 and 10 versus lane 8). They also reduced photolabeling by *ATP (Fig. 3A, compare lanes 7 and 8  with lane 6). These results are consistent with an interaction of Ap 5 A and Ap 4 A with both an AMP-and an ATP-binding site. However, Ap 5 A and Ap 4 A reduced labeling by *ATP to a lesser extent than did ATP, suggesting that they might not compete with *ATP at both ATP-binding sites.
To further assess this possibility, we labeled CFTR with *ATP, partially proteolyzed CFTR with Arg-C, and separated the digestion products on a 16% Tricine gel (Fig. 4A). We detected several labeled proteolytic CFTR fragments and focused on five small, labeled fragments (Fig. 4A, bands 1-5).
We tested bands 1-5 with antibodies to NBD1 and NBD2 (Fig.  5). Bands 1 and 5 contained NBD2 sequence, whereas bands 3 and 4 contained NBD1 sequence. Band 2 was recognized by antibodies against NBD1 (L12B4) and NBD2 (596), suggesting that band 2 may represent two fragments of similar size originating from NBD1 and NBD2. Non-radioactive ATP reduced *ATP photolabeling of bands 1-5 (Fig. 4, A (compare lanes 1   and 2) and B (quantitative data)), indicating that the CFTR fragments of bands 1-5 interacted with or were in close proximity to ATP bound to ATP-binding site 1 or to ATP-binding site 2 in the intact CFTR protein. However, non-radioactive Ap 4 A and Ap 5 A only reduced labeling of bands 2, 3, and 4 ( Fig. 4, A (compare lanes 3 and 4 with lane 2) and B (quantitative data)). Thus, the binding sites for ATP and Ap 4 A/Ap 5 A overlapped, but there were more binding sites for ATP than for Ap 4 A and Ap 5 A. Because CFTR contains two ATP-binding sites, these results suggest that Ap 4 A and Ap 5 A interacted with one of the two ATP sites.
Ap 4 A and Ap 5 A Interact with an ATP-and an AMP-binding Site-To corroborate these results and to further characterize the interactions of Ap 4 A and Ap 5 A with CFTR, we used photolabeling with an N 3 -analog. Because a radioactive N 3 -Ap 5 A was not commercially available, we used 8-N 3 -[␤ 3 -32 P]Ap 4 A. For ease of reading, we will refer to 8-N 3 -[␤ 3 32 P]Ap 4 A as *Ap 4 A. We found that *Ap 4 A readily labeled CFTR (Fig. 6A, lane 3). Labeling was UV light-dependent, and non-radioactive Ap 4 A and Ap 5 A reduced photolabeling. These results indicate that both Ap 4 A and Ap 5 A can interact with CFTR in the absence of other nucleotides. We also found that ATP and AMP reduced *Ap 4 A photolabeling. However, ATP interfered with *Ap 4 A photolabeling more effectively than AMP (Fig. 6A, lane 3 versus  lanes 6 and 7 compared with lane 3 versus lanes 8 and 9). Previous studies showed that ATP was also more effective than AMP at blocking *Ap 4 A labeling in chicken muscle adenylate kinase (56). These results suggest that when AMP is added in the absence of ATP, it interacts with CFTR with a lower affinity than ATP, Ap 4 A, and Ap 5 A.
To test the hypothesis that *Ap 4 A labeled both an ATP-binding site and a separate AMP-binding site, we compared the proteolytic fragments labeled by *ATP and *AMP with those labeled by *Ap 4 A. We found the following. (a) Again, *ATP

. Ap 5 A and Ap 4 A do not interact with all CFTR ATP-binding sites.
A, left, autoradiograph. Photolabeling of CFTR with 50 M 8-N 3 -[␣-32 P]ATP was performed as described in the legend to Fig. 3. 5 mM non-radioactive ATP, Ap 4 A, or Ap 5 A were present as indicated below the lanes of the autoradiograph. After photolabeling, CFTR was solubilized, immunoprecipitated, and subjected to partial proteolysis with the proteinase Arg-C. The digestion products were fractionated on a 16% Tricine gel. To compare with undigested photolabeled and immunoprecipitated CFTR, see Fig. 3. Middle, summary of Western blot analysis results (see Fig. 5). Right, quantitative data in linear arbitrary units (LAU) for radioactivity incorporated into the different CFTR fragments (bands 1-5) after labeling with 8-N 3 -[␣ 32 P]ATP. *, p Յ 0.024 compared with background control of no CFTR present (Mann-Whitney rank sum test, n ϭ 3-8). B, quantitative data for photolabeling bands 1-5 with 50 M 8-N 3 -[␣-32 P]ATP in the presence or absence of 5 mM non-radioactive ATP, Ap 4 A, or Ap 5 A. Amount of radioactivity incorporated into each band was normalized to radioactivity for control conditions (i.e. absence of non-radioactive ATP, Ap 4 A, or Ap 5 A). *, p Յ 0.001 (one-way repeated measures ANOVA followed by Holm-Sidak's method of multiple comparisons versus control group, n ϭ 2-5). Error bars, S.E. labeled bands 1-5 (Fig. 7, A (lane 4) and B (white bars)). (b) The labeling pattern with *AMP was different from that with *ATP (Fig. 7, A (compare lanes 4 and 5) and B (compare white and light gray bars)). *AMP photolabeled a proteolytic fragment of ϳ10 kDa (Fig. 7, A and B, band 6) that was not labeled by *ATP. Note that as described above, non-radioactive ATP was present when band 6 was labeled by *AMP (Fig. 7A, lane 5). This fragment was not detected in the absence of CFTR. It was not recognized by the NBD antibodies L12B4, M3A7, and 596 that reacted with fragments 1-5 (Fig. 5), presumably because it did not contain the epitope for those antibodies. These data suggest that CFTR contains a binding site for AMP that is distinct from the ATP-binding sites. (c) *Ap 4 A photolabeled a subset of fragments marked by *ATP (bands 2-4; Fig. 7, A (compare lanes 3  and 4) and B (compare white and dark gray bars)), consistent with *Ap 4 A labeling only one of the two ATP-binding sites. In addition, like *AMP, *Ap 4 A labeled band 6 ( Fig. 7, A (compare  lanes 3 and 5) and B (compare light and dark gray bars)). These data suggest that *Ap 4 A labels both an ATP-and an AMPbinding site. (d) ATP, as well as Ap 4 A and Ap 5 A, prevented labeling by *Ap 4 A (Fig. 7, A (compare lanes 8 -10 with lane 7) and B (dark gray bars)). Interestingly, non-radioactive ATP prevented photolabeling of band 6 by *Ap 4 A, although *ATP did not label band 6. These data suggest that photolabeling of band 6 by *Ap 4 A required that *Ap 4 A also interacted with an ATP-binding site at the time of labeling. This result is consistent with the finding that ATP facilitated *AMP labeling (Fig. 2).

AMP Enhances 8-N 3 -ATP Photolabeling at the Site Labeled by 8-N 3 -Ap 4 A-Ap 5 A inhibits adenylate kinase activity by
binding to the enzyme's active center (i.e. the phosphate donor ATP-binding site and the phosphate acceptor AMP-binding site) (26,29). We hypothesized that the increase in CFTR photolabeling with *ATP in the presence of AMP (Fig. 3) involved the same ATP-binding site that interacted with Ap 4 A/Ap 5 A. As shown above, *Ap 4 A labeled a subset of the proteolytic fragments that were labeled by *ATP (i.e. bands 2-4) (Fig. 7), and Ap 5 A reduced *ATP photolabeling of bands 2-4 but not of bands 1 and 5 (Fig. 4). We therefore predicted that AMP would enhance *ATP labeling of band 2, 3, or 4. Consistent with that prediction, AMP mainly increased *ATP photolabeling of band 4 as well as, to a small extent, of band 3, whereas labeling of bands 1, 2, and 5 did not significantly change (Fig. 8A, compare  lanes 2 and 3). Fig. 8B shows quantitative data. Although this assay has the limitation that it may not detect differences smaller than 25%, the results indicate that AMP enhances *ATP photolabeling at the same ATP-binding site that interacts with Ap 4 A/Ap 5 A (i.e. that interacts with the phosphate donor ATP).
Ap 5 A Interacts with ATP-binding Site 2-As previously shown, Ap 5 A inhibited wild-type CFTR Cl Ϫ current (19, 20, 57) (Fig. 9A, middle and right). To learn whether inhibition was due   (12.4 Ci). After photolabeling, CFTR was solubilized, immunoprecipitated, and subjected to partial proteolysis with the proteinase Arg-C. The digestion products were fractionated on a 16% Tricine gel. B, quantitative data in linear arbitrary units (LAU) for radioactivity incorporated into the different CFTR fragments after labeling with *Ap 4 A (dark gray bars), *ATP (white bars), and *AMP (light gray bars). *, p Յ 0.005 compared with 8-N 3 -[␤ 3 -32 P]Ap 4 A labeling under control conditions (i.e. absence of non-radioactive ATP, Ap 4 A, or Ap 5 A (dark gray bars)) (one-way repeated measures ANOVA followed by Holm-Sidak's method of multiple comparisons versus control group, n ϭ 3-5). Error bars, S.E.
to an interaction with ATP-binding site 1 or site 2, we studied the effect of phenylalanine substitutions in the phosphatebinding loops (Walker A motifs). ATP-binding site 1 contains the phosphate-binding loop of NBD1, whereas ATP-binding site 2 contains the phosphate-binding loop of NBD2 (Fig. 9A,  left). Berger et al. (15) found that substituting alanine at position 462 in NBD1 with phenylalanine (A462F mutation; Fig. 9B, left) abolished nucleotide interaction with ATP-binding site 1. Substituting serine at position 1248 in NBD2 with phenylalanine (S1248F mutation; Fig. 9C, left) abolished nucleotide interaction with ATP-binding site 2. Both mutants exhibited a low, ATP-dependent open probability due to a reduced opening rate with a normal burst duration (15). The A462F, but not the S1248F mutation interfered with processing and trafficking to the cell membrane (supplemental Fig. S1), and hence, the number of channels in excised membrane patches was small; there-fore, we quantified channel activity as NP o . We found that Ap 5 A reduced the NP o of A462F CFTR (Fig. 9B, middle and right). On the other hand, Ap 5 A had no effect on ATP-dependent current of S1248F CFTR (Fig. 9C, middle and right). Thus, preventing the interaction of nucleotides with ATP-binding site 2 eliminated Ap 5 A inhibition. Our results are consistent with the previous observations that mutations of conserved residues in the Walker A and B motifs of ATP-binding site 2, K1250A and  . C, left, model of S1248F CFTR. This mutant contained an N-terminal 6-histidine tag between CFTR amino acids 2 and 3. Middle, current recording (100 ms averages) from an excised inside-out membrane patch containing multiple S1248F CFTR channels. ATP and Ap 5 A were present during the times and at the concentrations indicated by bars. ATP was added together with PKA catalytic subunit. Holding voltage was Ϫ80 mV. Right, S1248F CFTR Cl Ϫ current before and after adding 1 mM Ap 5 A. Experiments were performed as shown in the middle panel with 0.3 mM ATP and PKA present. No significant differences were detected (p ϭ 0.463, Wilcoxon signed rank test, n ϭ 14 paired experiments of current measurements before and after adding Ap 5 A obtained from four membrane patches). Error bars, S.E. D1370N, abolished Ap 5 A inhibition of current, whereas the homologous mutations in ATP-binding site 1, K464A and D572N, did not. These results indicate that Ap 5 A interacts with ATP-binding site 2 and not site 1 to inhibit current.

DISCUSSION
Our data indicate that ATP and AMP interact with distinct sites in CFTR, and they mutually influence their interactions. Furthermore, our patch clamp data suggest that the adenylate kinase active center contains ATP-binding site 2.
We found that *AMP photolabeled CFTR in an ATP-dependent manner. ATP may have assisted AMP binding or azidogroup cross-linking. This effect may require ATP to interact with both ATP-binding sites or with only one of them. Of note, a number of approaches have demonstrated an ATP-dependent interaction with AMP in soluble, non-ABC adenylate kinases. (a) Several studies indicate that ATP increases binding of AMP to the AMP-binding site. Studies with beef heart mitochondrial adenylate kinase and a closely related GTP:AMP phosphotransferase showed that AMP bound to the enzyme only in the presence of nucleotide triphosphate (30,31). (b) The Michaelis constant (K m ) for AMP of bakers' yeast adenylate kinase is significantly higher if ATP is substituted by GTP. GTP binds with lower affinity than ATP and generates a lower catalytic rate (33), indicating that substrate binding at the nucleotide triphosphate binding site influenced the interaction of AMP at the AMP-binding site. Furthermore, mutations within the ATP-binding site that increase K m for ATP also increase K m for AMP (29).
In addition, we observed some augmentation of *AMP photolabeling when non-radioactive AMP was added in the absence of ATP (Fig. 2C, compare lanes 4 and 5). We speculate that AMP might bind the ATP-binding site under these conditions, which may increase *AMP labeling at the AMP-binding site(s). This speculation is founded on results with bakers' yeast and porcine muscle adenylate kinase showing that (a) ATP facilitated AMP binding, and (b) in the absence of ATP, AMP bound preferentially to the ATP-binding site although with much lower affinity than ATP (32,33,58).
Our experiments also showed that AMP increased CFTR photolabeling by *ATP. The data suggest that *ATP photolabeling increased at the same ATP-binding site that binds Ap 5 A. We speculate that an AMP-induced conformational change might be responsible (i.e. one similar to that described for SMC) (24). These biochemical data complement previous electrophysiological data: patch clamp experiments had shown that AMP induced positive cooperativity into the relationship between ATP concentration and current, indicating that AMP affected the interaction of ATP with CFTR. AMP-induced conformational changes affecting the interaction of ATP with the enzyme have been proposed from structural data or experimentally demonstrated for other adenylate kinases (34 -37). For example, in a study using an ATP analog with an environment-sensitive fluorescent group to probe the ATP-binding site of rabbit muscle adenylate kinase, Chuan et al. (35) showed that fluorescence increased in the presence of AMP. This result indicated that the environment of the fluorescent probe at the ATP-binding site changed when AMP was added, suggesting a conformational change (35).
Our results specify that the adenylate kinase inhibitor Ap 5 A interacts with only one of the two ATP-binding sites of CFTR, ATP-binding site 2. Because Ap 5 A binds to the active center of adenylate kinases, these data suggest that ATP-binding site 2 interacts with the phosphoryl group donor ATP. Previous studies are consistent with this conclusion. (a) The S1248F mutation, which interfered with the binding of ATP to site 2 (15) and abolished the inhibition of CFTR current in the presence of Ap 5 A in our study (Fig. 9C), also disrupted CFTR adenylate kinase activity (22). (b) Patch clamp studies showed that CFTR mutations K1250A and D1370N, located within the conserved Walker A and B motifs of ATP-binding site 2, abolished the effects of Ap 5 A and AMP on CFTR current. However, the homologous mutations in ATP-binding site 1 (K464A and D572N) did not (19). Our data plus those observations indicate that adenylate kinase activity involves the same ATP-binding site at which ATP is hydrolyzed in the absence of AMP (15)(16)(17)(18).
In current models of CFTR, both ATP molecules are bound between the phosphate-binding loop (Walker A motif) of one NBD and the ABC signature motif of the other NBD (Fig. 9A, left) (50,59,60). Thus, both NBDs contribute to both binding sites. These models are based on the crystal structures of other ABC proteins and corroborated by patch clamp data suggesting ATP-driven NBD1-NBD2 dimerization opening the CFTR channel pore (61). Our patch clamp results suggest that Ap 5 A interacts with the phosphate-binding loop (Walker A motif) of NBD2 but not with the phosphate-binding loop of NBD1. In addition, we showed that Ap 5 A prevented *ATP photolabeling of CFTR proteolytic fragments containing NBD1 sequences (bands 2-4 in Figs. 4 and 5). *ATP photolabeling occurs at protein sequences in close proximity to the N 3 -group at the adenine ring. These results are consistent with Ap 5 A and ATP binding at the interface of an NBD1-NBD2 dimer.
To speculate about the structural basis of our findings, we constructed a three-dimensional homology model of the CFTR NBD1-NBD2 heterodimer ( Fig. 10 and supplemental Movie S1). The model revealed a central cavity formed by residues from both NBD1 and NBD2, particularly Gly-550, Ser-573, Gly-576, Tyr-577, Glu-655, Gln-1291, Gln-1352, Glu-1371, Ala-1374, and His-1402. If one adenosine moiety of Ap 5 A were bound to ATP-binding site 2, the other adenosine, the "AMP" adenosine, could extend into the central cavity. Although the conformation of the bound Ap 5 A cannot be predicted, it would probably interact with a subset of the residues lining the cavity. The same residues would be expected to interact with AMP. Of note, Gln-1291 is the conserved Q-loop glutamine of NBD2 (8,62). In the crystal structure of an SMC-NBD, the "AMP" adenosine of Ap 5 A is bound to the SMC Q-loop glutamine (24). Interestingly, mutations of several residues lining the central cavity have been observed in human disease (see the Cystic Fibrosis Mutation Database, Cystic Fibrosis Centre at the Hospital for Sick Children, Toronto, Canada). Mutation Q1291R has been found in patients with cystic fibrosis. Mutations G576A, G550R, and Q1352H have been described in patients with congenital bilateral absence of the vas deferens, a condition that affects men with cystic fibrosis but can also occur in the absence Nucleotide Interactions with the ABC Adenylate Kinase CFTR SEPTEMBER 20, 2013 • VOLUME 288 • NUMBER 38 of other disease manifestations. Because in our model, both NBDs contribute to the central cavity, the interaction of AMP with CFTR is predicted to be influenced by NBD1-NBD2 dimerization, which is linked to ATP binding. Our finding that *AMP photolabeled CFTR in an ATP-dependent manner is consistent with this model. Our model shows His-1348 in a position where it might prevent Ap 5 A from interacting simultaneously with ATP-binding site 1 and central cavity residues. We speculate that this structure may explain our results indicating that Ap 5 A interacts only with ATP-binding site 2.
Previous work showed that Ap 5 A partially inhibits wild-type CFTR channel activity (19,57). In the case of CFTR mutants, the interaction with Ap 5 A may increase current. An example is L1254A CFTR, which showed an increase in current after adding Ap 5 A due to a reduced channel closing rate. The mutation is predicted to affect the interaction of ATP and Ap 5 A at ATPbinding site 2 (45). Although Ap 5 A itself is not cell-permeable and is not therefore a potential pharmaceutical, our studies may lead to the development of therapeutic strategies to positively or negatively modulate the channel activity of CFTR. Those may be of potential value for the treatment of diseases associated with decreased (cystic fibrosis) or increased (cholera toxininduced secretory diarrhea (55)) CFTR chloride currents.