How Phosphorylation and ATPase Activity Regulate Anion Flux though the Cystic Fibrosis Transmembrane Conductance Regulator (CFTR)*

The cystic fibrosis transmembrane conductance regulator (CFTR, ABCC7), mutations of which cause cystic fibrosis, belongs to the ATP-binding cassette (ABC) transporter family and works as a channel for small anions, such as chloride and bicarbonate. Anion channel activity is known to depend on phosphorylation by cAMP-dependent protein kinase A (PKA) and CFTR-ATPase activity. Whereas anion channel activity has been extensively investigated, phosphorylation and CFTR-ATPase activity are still poorly understood. Here, we show that the two processes can be measured in a label-free and non-invasive manner in real time in live cells, stably transfected with CFTR. This study reveals three key findings. (i) The major contribution (≥90%) to the total CFTR-related ATP hydrolysis rate is due to phosphorylation by PKA and the minor contribution (≤10%) to CFTR-ATPase activity. (ii) The mutant CFTR-E1371S that is still conductive, but defective in ATP hydrolysis, is not phosphorylated, suggesting that phosphorylation requires a functional nucleotide binding domain and occurs in the post-hydrolysis transition state. (iii) CFTR-ATPase activity is inversely related to CFTR anion flux. The present data are consistent with a model in which CFTR is in a closed conformation with two ATPs bound. The open conformation is induced by ATP hydrolysis and corresponds to the post-hydrolysis transition state that is stabilized by phosphorylation and binding of chloride channel potentiators.

nucleotide binding domain (NBD). As other members of the ABCC family, CFTR lacks the catalytic glutamine in NBD1 and carries it only in NBD2 (Glu-1371) (2). In NBD1, glutamine is replaced by serine (Ser-573), which binds nucleotides tightly, but significantly reduces the rate of ATP hydrolysis (3). The two halves of the protein are linked by an intrinsically disordered (4) regulatory (R) domain, unique to CFTR. This leads to the overall domain organization TMD1-NBD1-R-TMD2-NBD2 (5). The R-domain is highly charged and contains numerous serine residues that can be phosphorylated by protein kinases, particularly by the cAMP-dependent protein kinase A (PKA) (6). A low resolution structure of CFTR, obtained by electron crystallography in the absence of nucleotides and phosphorylation (7), revealed strong similarity to the nucleotide-bound structure of the homodimeric bacterial ABC transporter Sav1866 (8).
The function of the CFTR anion channel has been investigated in great detail by different electrophysiological methods, including iodide efflux measurements with iodide-sensitive electrodes, Ussing chamber measurements, and most importantly patch clamp techniques (9). Although CFTR gating has long been known to depend on the hydrolysis of nucleoside triphosphate (generally ATP) (10,11) and on phosphorylation of the R-domain (12), information on direct measurements of the two processes is sparse.
CFTR ATPase activity could so far be measured only with radioactively labeled nucleotides in inside-out membrane vesicles, as first demonstrated by Riordan and co-workers (13). The maximum rate of ATP hydrolysis by phosphorylated CFTR was determined as V max ϭ 53.8 nmol mg Ϫ1 min Ϫ1 (14), in good agreement with the catalytic rate constant or turnover number determined as k cat Ϸ0.14 s Ϫ1 (15). However, CFTR shows significant (80 -90%) functional attrition during reconstitution, and therefore the turnover number was later re-estimated as k cat ϭ 1-2 s Ϫ1 (16), which brings it into the time frame for channel gating (17). Further insight into the role of ATP hydrolysis was gained indirectly by investigating channel gating by means of electrophysiological methods in the presence of different nucleotides (3,18).
CFTR phosphorylation by PKA was first measured with radioactively labeled ATP and subsequent immunoprecipitation of phosphorylated CFTR in inside-out vesicles (13,19) and cells (6). Because separation of the two ATP-consuming processes, CFTR-ATPase activity and PKA activity, is difficult, the quantitative information of these approaches is limited. PKAdependent CFTR activation was therefore analyzed mainly indirectly, again using electrophysiological methods (20,21).
Mutation of serine residues (Ser-660, Ser-737, Ser-795, and Ser-813) revealed that none of the sites alone was essential. Simultaneous mutation of all four serine residues substantially reduced the PKA-stimulated current (22); yet mutation of as much as 10 serine residues did not fully eliminate the PKAstimulated current (23). Notably, CFTR constructs lacking the R-domain were still conductive and even exhibited a somewhat higher affinity to nucleotides but shorter open times (24). The unphosphorylated R-domain was therefore assumed to inhibit CFTR channel opening by interfering with nucleotide binding and dimerization, rather than by stabilizing NBD dimer formation, as proposed earlier (25,26).
Presently, the most accepted model for CFTR function assumes channel opening upon ATP binding at the interface between the two NBDs, with concomitant formation of an outward facing form (5,27), and channel closing upon ATP hydrolysis at NBD2, with subsequent formation of an inward facing form (28,29). The model corresponds to the alternating access mechanism (30), which has also been adapted to other ABC transporters (31). This model is supported by the outward facing structures of Sav1866 with two identical nucleotides bound (8,32) and the inward facing (outward closed) structure of the nucleotide-free P-glycoprotein (33). Alternating access to CFTR was confirmed recently by testing the accessibility of cysteines in the transmembrane domain with cysteine-reactive probes. However, surprisingly, the "open state" of CFTR was facing inward, and the "closed state" of CFTR was facing outward (34). This finding raises new questions as to the role of the ATP hydrolysis cycle for CFTR anion channel function.
To circumvent problems with the susceptibility of CFTR to reconstitution, we measured the CFTR-related ATP hydrolysis rate in live Chinese hamster ovary (CHO) and baby hamster kidney (BHK) cells, stably transfected with human CFTR (CHO-CFTR and BHK-CFTR cells). As ATP is re-synthesized directly on demand, ATP hydrolysis can be monitored by measuring the extracellular acidification rate (ECAR), if cells work under glycolytic conditions (35)(36)(37)(38), or the oxygen consumption rate (OCR), if cells work under conditions of oxidative phosphorylation (respiration) (39). In parallel, we measured iodide efflux under comparable experimental conditions. As phosphorylation agents, we used CPT-cAMP and forskolin. CPT-cAMP causes dissociation of the regulatory and catalytic subunits of PKA, which leads to R-domain phosphorylation by the catalytic subunit, whereas forskolin acts indirectly by stimulating cAMP synthesis through adenylate cyclase. In the presence of CPT-cAMP, the energy required to drive CFTR activity (total CFTR-related ATPase activity) can thus be subdivided into the energy consumed to drive PKA and CFTR-ATPase activity, respectively. In the presence of forskolin, adenylate cyclase activity may also contribute.
The following questions were asked. (i) How much ATP is consumed by the individual processes contributing to the total CFTR-related ATP consumption rate, PKA, CFTR-ATPase, and adenylate cyclase activity, respectively? (ii) At which step of the CFTR-ATPase activity cycle does phosphorylation of the R-domain occur? (iii) How are phosphorylation of the R-domain by PKA and CFTR-ATPase activity related to anion efflux? The answers to these questions allowed unraveling the complex interplay between CFTR phosphorylation, CFTR-ATPase activity, and anion flux. A new model for the CFTR-ATPase activity cycle is proposed.
Growth and Flow Media-For cell growth, MEM␣ without ribonucleosides and deoxyribonucleosides, containing 10% heat-inactivated FBS, was used.
For flow medium preparation, commercially available dry powder MEM␣ (for CHO cells) or DMEM (for BHK cells) was used. Both flow media were without FBS and without sodium bicarbonate to maintain a very low buffer capacity. Sodium chloride was used as a substitute to preserve osmotic balance. The pH was adjusted to 7.4 at 37°C.
Cell Culture-Non-transfected CHO cells were grown in MEM␣ at 37°C and 5% CO 2 . Stably transfected cell lines were cultured in the same medium supplemented with 50 M methotrexate. Cells were split 1:15 and passaged every 2-3 days. BHK cell lines expressing CFTR variants were grown in DMEM/F-12, 1:1, supplemented with 10% FBS and 500 M methotrexate under the same conditions as CHO cells.
ECAR and OCR Measured with a Bionas-The Bionas Discovery 2500 cell-analyzing system (39) consists of six parallel measuring chambers. Cells were directly seeded on Bionas SC1000 metabolic chips with oxygen, pH, and impedance sensors to reach a final number of n ϭ 3⅐10 5 cells per chip, which all contributed to the ECAR. The pH sensors measure the voltage change as a function of time that is linearly related to a pH change as a function of time. Oxygen sensors measure the current change as a function of time that is linearly related to oxygen concentration in solution as a function of time (for calibration, see below).
Cells were incubated for 4 h with growth medium. To start a stimulation cycle, cells were switched to flow medium with a flow rate, r ϭ 110 l/min in each measuring chamber, until a stable basal metabolic rate was reached (about 2 h). Cells were then exposed to medium containing phosphorylation agents, until a steady state was reached. After each stimulation cycle, cells were flushed again with pure medium until basal activity was reached, to ensure that stimulations were fully reversible. Arrival of the titrant in the measurement chamber was monitored from changes in impedance values and was moreover calculated taking into account the dimensions of the tubing system. Pump cycles consisted of a 2-min go and a 2-min stop, yielding a time resolution of 4 min. Acidification was measured in the stop phase.
ECAR Measured with the Cytosensor-The Cytosensor microphysiometer (Molecular Devices) (35,40) consists of eight light-addressable potentiometric sensors. It offers a second tubing system that was filled with flow medium containing the titrant (e.g. CPT-cAMP or forskolin) and allows applying agents with little delay, as a valve close to the measurement chamber can be switched between the two channels. Because of the shorter tubing system, compound adsorption to the walls of the tubes is comparatively low.
Cells were grown as a layer on 3.0-m polycarbonate membranes (inner diameter, 12 mm, Corning) on the bottom of a flow chamber in diffusive contact with the sensors. The sensors again measure the voltage change as a function of time. Cells were seeded to reach a final number of N C ϭ 5⅐10 5 cells per membrane, corresponding to 80 -90% confluency; however, because of the flow cell geometry, only ϳ33% of the cells contribute to the ECAR signal detected by the sensors (N C ϭ 1.65⅐10 5 cells). Cells were incubated either for 4 h or overnight with growth medium, which was then switched to flow medium with a flow rate, r ϭ 100 l/min. Extracellular acidification rates were calculated using Molecular Devices Cytosoft, and the slope between 5 s after stopping and 2 s before restarting the pump was evaluated to avoid artifacts caused by pump switching (for pump cycle see Fig. 1 in Ref. 41). The Cytosensor showed a time resolution of 2 min. To quantify the observed ECAR changes, four measurement points at the end of each drug exposure period were averaged and normalized to the average of the last four points of the basal ECAR before substance application, taken as 100%.
Quantification of Acid Export-The measured voltage change per time ECAR V (V s Ϫ1 ) was transformed to the effective ECAR (protons released cell Ϫ1 s Ϫ1 ) based on a calibration (cal) showing that 61.0 Ϯ 1.1 mV corresponded to 1 pH unit in the Cytosensor (37) and 48.0 Ϯ 2.5 mV to 1 pH unit in the Bionas as shown in Equation 1, The buffer capacity of the flow medium was determined as ␤ ϭ 0.596 mM, taking into account phosphate salts (0.91 mM) and the amino acids glutamine (2 mM), histidine (0.2 mM), and cysteine (0.57 mM) (42). The volumina of the flow chambers were V ϭ 2.8 l for the Cytosensor, and V ϭ 5.65 l for the Bionas. N A is the Avogadro number, and N C is the number of cells contributing to the ECAR (N C ϭ 1.65⅐10 5 in Cytosensor and N C ϭ 3⅐10 5 in Bionas).
Lactate Quantification-Experiments in the Cytosensor were carried out as described above, but the flow medium was replaced by phosphate buffer (0.3 mM CaCl 2 , 0.6 mM MgCl 2 , 0.5 mM KH 2 PO 4 , 3 mM KCl, 0.5 mM Na 2 HPO 4 , 130 mM NaCl, 10 mM glucose). The buffer was collected after leaving the measurement chamber. All cells contributed to lactate efflux. The samples were lyophilized and dissolved again in 3 ml of buffer (1 M glycine, 0.6 M hydrazine, 5.6 mM EDTA) at pH 9.5. Lactate dehydrogenase (5000 units/ml) and NAD (2.5 mM) were added to the buffer, and the absorbance increase at 340 nm was monitored. Lactate concentrations were calculated from the stable absorption at 340 nm reached at the end of the reaction (37).
Quantification of Oxygen Consumption-Assuming that the measured current I(k) changes linearly with the oxygen concentration in solution, the oxygen consumption per time ⌬n/⌬t (mol s Ϫ1 ) was determined as shown in Equation 2, where I(k) corresponds to ϳ1 nA and was set to 100% in each go phase to correct for the potential drift; C O2, S is the concentration of oxygen in aqueous solution at 37°C (C O2, S ϭ 0.214 mM); ⌬I(k) is the measured change in current (e.g. ⌬I(k) ϭ 1.5 pA) reflecting the change in oxygen consumption, and V r, chip is the volume of the flow chamber (V r, chip ϭ 5.65 l). ATP Hydrolysis Rates-ATP hydrolysis rates were calculated taking into account that glycolysis yields two molecules of lactic acid per molecule of glucose consumed and produces two molecules of ATP. Respiration yields ϳ36 molecules of ATP per molecule of glucose, whereas six molecules of dioxygen are consumed (35). One molecule of dioxygen consumed thus corresponds to six molecules of ATP produced (42).
Iodide Efflux Measurements-Iodide efflux measurements were carried out according to Chen et al. (43) with small modifications. CHO cells were grown to 90% confluence in 60-mm culture dishes and were washed three times with 2.5 ml of loading buffer, pH 7.4, in which chloride was replaced by iodide (136 mM NaI, 3 mM KNO 3 , 2 mM Ca(NO 3 ) 2 , 11 mM glucose, 20 mM HEPES/NaOH). After incubation for 3 h in the dark at 37°C, cells were washed carefully 30 times with 2.5 ml of efflux buffer, pH 7.4, in which chloride was replaced by nitrate (136 mM NaNO 3 , 3 mM KNO 3 , 2 mM Ca(NO 3 ) 2 , 11 mM glucose, 20 mM HEPES/NaOH) to remove excess iodide. For iodide efflux measurements, efflux buffer was exchanged in 1-min intervals. After 5 min, efflux buffer was replaced by stimulation buffer (i.e. efflux buffer containing CPT-cAMP or forskolin). Samples were collected and stored in the dark at room temperature. An iodide-selective electrode (Mettler Toledo, perfectIon combination I Ϫ ) was used to measure voltages of the samples. Iodide concentrations were calculated using a calibration curve obtained with efflux buffer containing various concentrations of NaI. All experiments were carried out at 37°C. Models-PKA activation by CPT-cAMP was fitted with Equation 3 as proposed previously for PKA in solution (44), The term in brackets reflects the activation of PKA by binding four cAMP molecules to the regulatory subunit and subsequent dissociation of regulatory and catalytic subunits. The CFTR-ATPase activity versus modulator concentration curves were fitted with Equation 4, which is based on a two-site binding model previously proposed for P-glycoprotein (38,45).
Here V is the velocity reflecting the change in ATPase activity or ECAR; K 1 is the concentration of half-maximum activation; K 2 is the concentration of half-maximum inhibition; V 0 is the normalized basal activity; V 1 is the maximum activity; V 2 is the minimum activity, and C the allocrite concentration in aqueous solution. Experiments were performed under steady state conditions.

Results
Glycolysis and Oxidative Phosphorylation-Cultured cells generally work under glycolytic conditions in the presence of glucose as a carbon source (35). However, they possess the ability to shift dynamically between glycolysis and oxidative phosphorylation (or respiration) (46). To test which of the two metabolic pathways predominates under the present conditions, we measured the ECAR and the OCR of CHO-K1 and CHO-CFTR cells in flow medium without and with the phosphorylation agents CPT-cAMP and forskolin. Seven stimulation cycles were measured in sequence with increasing concentrations of CPT-cAMP (Fig. 1, A and B) and forskolin ( Fig. 1, C and D), using a Bionas Discovery 2500. CPT-cAMP barely influenced the ECAR of CHO-K1 control cells (Fig. 1A). Conversely, it strongly stimulated the ECAR of CFTR-transfected cells (Fig.  1B). ECAR maxima were reached at incubation times, t ϳ20 min. The OCR increased only at the highest CPT-cAMP concentrations (C CPT Ͼ 100 M) and was delayed by about 24 min relative to the ECAR.
The effect of forskolin was again small in CHO-K1 cells (Fig.  1C) and significantly more pronounced in CHO-CFTR cells ( Fig. 1D). At higher forskolin concentrations (5th stimulation), the ECAR started to decrease and the OCR to increase in CHO-K1 and CHO-CFTR cells, indicating a decrease in glycolysis and an increase in respiration in both cell types. At even higher concentrations (6th stimulation), respiration became predominant. Washout (i.e. flushing cells with pure medium) after stimulation with high forskolin concentrations led to dilution of the compound and, as a consequence, to an ECAR rebounce. Whereas basal OCR was reached relatively rapidly (within a few minutes), the basal ECAR was reached only slowly (within more than an hour). Analogous washout effects, although with smaller amplitudes, were observed in CHO-K1 cells (see 6th stimulation).
In conclusion, CHO-K1 and CHO-CFTR cells worked under glycolytic conditions (yielding two molecules of ATP per molecule of glucose) in pure medium, as well as in the presence of CPT-cAMP up to the highest concentrations at incubation times, t Ͻ24 min. In the presence of low forskolin concentrations (C fsk ϭ 0.1-1 M), the situation was identical. However, at higher concentrations, forskolin reduced cellular glucose uptake (47)(48)(49) and as a consequence cells shifted from glycolysis to the more efficient respiration (yielding ϳ36 molecules of ATP/molecule of glucose) ( Table 1, end).

ECAR in CHO-K1 and CHO-CFTR Cells Measured as a Function of CPT-cAMP and Forskolin Concentration-For
quantitative data evaluation, we repeated titrations of CHO-K1 and CHO-CFTR cells with CPT-cAMP and forskolin using a Cytosensor (that measures only the ECAR), because it shows less adsorption of hydrophobic compounds to the tubing system and has a higher time resolution. The ECAR of CHO-K1 and CHO-CFTR cells is shown as a function of the CPT-cAMP ( Fig. 2A) and forskolin ( Fig. 3A) concentration, respectively. As in Fig. 1A, CHO-K1 cells showed only a low response with a maximum of 142 Ϯ 9% at high CPT-cAMP concentrations (C CPT Ն200 M), whereas CHO-CFTR cells again showed a significant ECAR increase with a maximum ECAR of 195 Ϯ 20% at C CPT ϭ 400 M.
In the case of forskolin, the ECAR of CHO-K1 cells remained at the basal level at low concentrations, increased only slightly (maximum 125 Ϯ 18%) at higher concentrations up to C fsk ϭ 10 M, and then decreased below basal values, reaching only ϳ50% of initial basal value at C fsk ϭ 100 M. The ECAR of CHO-CFTR cells increased already at low concentrations (C fsk Ͻ1 M) and reached a maximum of 195 Ϯ 20% at the concentration C fsk ϭ 1 M. At higher concentrations, the ECAR started to decrease as for CHO-K1 cells and reached only 30% of the initial basal value at the concentration C fsk ϭ 100 M. As seen in Fig. 1, C and D, the decrease in ECAR is compensated by an increase in OCR and can be attributed to a shift from glycolysis to respiration in CHO-K1 and CHO-CFTR cells.
ECAR Corresponds to Lactate Efflux and Reflects the Rate of ATP Hydrolysis-To test whether the ECAR induced upon stimulation with CPT-cAMP corresponds to lactate export, we superfused CHO-CFTR cells with 50 M CPT-cAMP for 2 h and collected the medium after it had passed the measurement chambers. Lactate was then quantified as described under "Experimental Procedures" (37). Basal acidification rates were determined as v L Ϸ 4⅐10 6 lactic acid molecules extruded cell Ϫ1 s Ϫ1 , which corresponds to v ATP Ϸ 4⅐10 6 ATP molecules cell Ϫ1 s Ϫ1 hydrolyzed via glycolysis. Upon stimulation with CPT-cAMP (C CPT ϭ 50 M), lactate efflux increased to 180 Ϯ 10% relative to basal values (100%) corresponding to v L ϭ 7.2⅐10 6 1). As shown above, the protons measured corresponded to lactic acid molecules extruded.
The basal oxygen consumption rates of CHO-CFTR and CHO-K1 were determined as v bas, ox ϭ 0.63 fmol cell Ϫ1 min Ϫ1 for both cell lines (see Equation 2). This value is within the range of published basal oxygen consumption rates (v oxygen ϭ 0.1-10 fmol cell Ϫ1 min Ϫ1 ) (50). The basal ECAR and OCR values (100%) were used for further quantification.
The ATP hydrolysis rate as a function of the CPT-cAMP concentration as derived from ECAR values ( Fig. 2A) is shown in Fig. 2B. The ATP hydrolysis rate at t ϭ 40 min was calculated by taking into account the OCR measured with the Bionas (last stimulation points in Fig. 1B). After correction for OCR, the ATP hydrolysis rates measured at the two different time points (t ϭ 20 and 40 min) were identical, suggesting that phosphorylation reached a steady state at 20 min.
The ATP hydrolysis rate as a function of forskolin concentration (Fig. 3B) was estimated, combining ECAR (Fig. 3A) and Measurements were performed at 37°C. Rates were measured for 2 min, followed by 2 min of flushing with flow medium; a measurement point was thus generated every 4 min. Horizontal lines indicate the presence of phosphorylation agents in flow medium. Cells were stimulated with phosphorylation agents during 24 min at low concentrations and 32 min at higher concentrations. The concentrations given correspond to the concentrations applied; as the hydrophobic forskolin adsorbed to the tubing system, the concentrations reaching the measuring chambers were somewhat lower. Washout with pure flow medium after application of high forskolin concentrations induced an ECAR re-bounce due to forskolin dilution. JULY 8, 2016 • VOLUME 291 • NUMBER 28

JOURNAL OF BIOLOGICAL CHEMISTRY 14487
OCR measurements (Fig. 1D). Data in the intermediate concentration range could not be determined accurately and were omitted.    0.49 fmol cell Ϫ1 min Ϫ1 ) (Fig. 2, A and B)  Notably, the apparent catalytic rate constant obtained in the presence of CPT-cAMP consists of the phospho group transferase activity of PKA and the ATPase activity of CFTR. PKA Activity Is the Major Contribution and CFTR-ATPase Activity Is the Minor Contribution-To assess the relative contribution of the PKA activity to the total CFTR-related ATP hydrolysis rate, we inhibited the latter with two well known protein kinase inhibitors, the myristoylated PKA inhibitor peptide PKI(14 -22) (K i ϭ 36 nM) and the PKA inhibitor H-89 (K i ϭ 135 nM). Both were applied at a concentration (C i ϭ 20 M) previously proven to completely inhibit PKA activity in live cells (51)(52)(53). Cells were incubated with the inhibitors for 30 -40 min before stimulation with CPT-cAMP or forskolin. Both inhibitors influenced the ECAR and thus the cell metabolism. PKI (14 -22) transiently enhanced the ECAR to about 190%, most likely due to its detergent-like nature; however, the signal returned to basal values during the incubation period. H-89 enhanced the ECAR of all cell lines to about 180% and stabilized at a high level (data not shown). In the case of H-89, observations were therefore normalized to this altered basal activity. Both inhibitors reduced the stimulation with either CPT-cAMP or forskolin to low values of ϳ110% as summarized in Table 2. The major ATP hydrolysis rate (Ն90%) could thus be attributed to phospho group transferase activity of PKA. The strong down-regulation of the ECAR at high forskolin concentrations (C fsk Ͼ10 M) was not influenced by the PKA inhibitors.
Although we could not fully exclude a low prevailing PKA activity, we assumed that the minor (Յ10%) remaining contribution to the total CFTR-related ATP hydrolysis rate was due to ATP hydrolysis of partially phosphorylated CFTR. The catalytic rate constant of CFTR-ATPase was estimated as k cat, CFTR Յ14 s Ϫ1 . This is in the order of the catalytic rate constant of P-glycoprotein (36,54,55) and is also in broad agreement with a previous estimate of the catalytic rate constant of fully phosphorylated CFTR, k cat, CFTR ϭ 1-2 s Ϫ1 (16).
CFTR Phosphorylation by PKA-Because PKA activity was responsible for almost the total ECAR or ATP hydrolysis rate in the presence of CPT-cAMP, the ECAR versus CPT-cAMP concentration curves obtained with CHO-CFTR and BHK-CFTR cells (Figs. 2, A and B, and 4A) could be well fitted with Equation 3 (44) describing cAMP dissociation from PKA (i.e. PKA activation). The resulting apparent dissociation constants, K d, app , of CPT-cAMP from PKA in both cell lines (for data see Table 1) were in good agreement with previous data, obtained with purified PKA in aqueous solution (56). The good agreement between the behavior of purified and cellular PKA further supports the assumption that the major contribution to CFTRrelated ATP hydrolysis was due to PKA activity.
The concentration of half-maximum channel activation by CPT-cAMP was determined as K 0.5 Ϸ40 M at 20 and 40 min. With decreasing incubation times, the parameter, K 0.5 , increased and reached a value of K 0.5 ϳ83 M at 4 min (Table 1), which is in good agreement with a value obtained by patch clamp measurements (9).   CFTR Phosphorylation Occurs in the Post-hydrolysis Transition State-To test whether CFTR-ATPase activity was required for phosphorylation, we used the ATPase-deficient CFTR mutant CFTR-E1371S. It was demonstrated to express well in CHO cells (1), and to exhibit prolonged open times (17). The present experiments were performed with BHK cells under the conditions described above for CHO cells, except that DMEM (without bicarbonate) was used as flow medium instead of MEM␣ to preserve the conditions used for cell culture. The response of BHK-CFTR cells to stimulation with CPT-cAMP was qualitatively similar to that of CHO-CFTR cells; however, the maximum ECAR was lower (123 Ϯ 7% at C CPT ϭ 100 M) (Fig. 4A). The response to forskolin (Fig. 4B) was also similar to that in CHO cells, but the ECAR was again lower and reached a maximum stimulation of 117 Ϯ 4% around C fsk ϭ 1 M. At higher forskolin concentrations, the ECAR started to decrease as shown above in CHO cells (Fig. 3A).

FIGURE 4. A and B. ECAR of BHK cells overexpressing CFTR variants as a function of CPT-cAMP and forskolin concentration. ECAR of BHK-CFTR (f) and BHK-E1371S (E) cells measured in a Molecular
Importantly, BHK-E1371S cells showed no ECAR increase upon addition of phosphorylation agents (Fig. 4). Hence, BHK-E1371S with the hydrolysis-deficient NBD2 was not only unable to hydrolyze ATP, as expected, but moreover could not be phosphorylated. As phosphorylation of BHK-E1371S did not occur, we concluded that ATPase activity is required for phosphorylation and that phosphorylation most likely occurs in the post-hydrolysis transition state. The slight but distinct ECAR decrease (or alkalinization) to 84 Ϯ 4% at C CPT ϭ 50 M in Fig.  4A was most likely due to direct efflux of bicarbonate ions by CFTR-E1371S and possibly also by chloride/bicarbonate exchange proteins regulated indirectly (57,58).
CFTR-ATPase Activity Is Enhanced by Anion Channel Inhibitors-Next, we investigated how the CFTR-ATPase reacts to chloride channel inhibitors, known to specifically interact with CFTR, including glibenclamide (Fig. 5) glipizide (Fig. 6), and CFTRinh-172 (Fig. 7). For this purpose, we phosphorylated CHO-K1 control cells and CHO-CFTR cells with different CPT-cAMP concentrations (C CPT ϭ 5-100 M) as described above, whereby phosphorylation was kept constant throughout the subsequent titration experiment. Cells were incubated for 20 min with the particular drugs at each concentration. The interval length and conditions were chosen to ensure that a stable ECAR was reached and the cell metabolism remained glycolytic. After each stimulation period, cells were flushed with medium containing only CPT-cAMP until the ECAR returned to basal values, showing that ECAR changes were reversible at all conditions.
In CHO-K1 cells, anion channel inhibitors induced only negligibly small effects (Ͻ120%) except CFTRinh-172, which reached ECAR values of ϳ138% at C CPT ϭ 50 M. In CHO-CFTR cells, all there anion channel inhibitors induced bellshaped CFTR-ATPase activity curves with ECAR maxima up to ϳ210% (Figs. 5-7). Titration curves could be well fitted with Equation 4 that assumes ATPase activation at low and inhibition at high allocrite concentrations as shown previously for P-glycoprotein in inside-out lipid vesicles (36,45,59) and in cells (36,38). Kinetic analysis (Equation 4) of the titration curves in the presence of different concentrations of CPT-cAMP (Figs. 5-7) yielded the concentration of half-maximum CFTR-ATPase activation, K 1 , and inhibition, K 2 , as well as maximum, V 1 , and minimum activity, V 2 , summarized in Table 3.
CFTR-ATPase Activity Is Reduced by Anion Channel Potentiators-Analogous titrations of CHO-K1 and CHO-CFTR cells, respectively, were performed with the potentiators genistein (Fig. 8A) and capsaicin (Fig. 8B). In contrast to inhibitors, potentiators showed no ECAR increase. Higher concen- trations even reduced the ECAR of CHO-K1 and CHO-CFTR cells down to a minimum of ϳ55 and 35%, respectively, in the case of genistein (Fig. 8A) and 65 and 50%, respectively, in the case of capsaicin (Fig. 9A). The ECAR decrease at high potentiator concentrations was fully reversible under all conditions in both cell lines.
Anion Channel Potentiators Induce Cellular Respiration-At concentrations relevant for anion channel activation, potentiators such as genistein (Fig. 8) and capsaicin (Fig. 9) significantly reduced ECAR in both CFTR overexpressing (CHO-CFTR) and control (CHO-K1) cells. As for forskolin (47)(48)(49), inhibition of glucose import has been observed for several flavonoids, including genistein, e.g. in U937 cells (60) and CHO cells (61). To our knowledge, the influence of capsaicin on glucose metabolism has not yet been described. Both potentiators thus seemed to reduce glucose uptake, which leads to a reduction in glycolysis and to an increase in respiration as observed for forskolin (Fig. 1D).
CPT-cAMP and Forskolin Not Only Act as Phosphorylation Agents but also as CFTR Modulators-The CFTR-ATPase activity curves were measured at different CPT-cAMP concentrations (Figs. 5-7). With increasing concentrations of CPT-cAMP, the apparent concentration of half-maximum CFTR-ATPase activation, K 1 , decreased. The shift to lower values was particularly evident for the more hydrophilic glibenclamide and glipizide ( Fig. 7 and Table 3), and it was less pro-nounced for the more hydrophobic CFTRinh-172. The maximum ATPase activity, V 1 , of CFTR, induced by the chloride channel inhibitors (i.e. ATPase activators) increased at very low CPT-cAMP concentrations and decreased at higher concentrations (C CPT Ն25 M). Analogous shifts of kinetic constants of CFTR with the CPT-cAMP concentration were observed previously for the potentiator genistein (62).
The phenomenon described above is typical for ABC transporters such as P-glycoprotein, which are able to simultaneously accommodate different molecules in their binding   (89). F, concentration range reported to cause inhibition of CFTR-dependent short-circuit current after stimulation with 100 M CPT-cAMP (89). G, concentration range described to block CFTR in excised patches from transfected mouse embryo fibroblasts (92). Data obtained from biological duplicates are shown with standard deviations.

CFTR Phosphorylation, CFTR-ATPase Activity, and Anion Flux
regions (63)(64)(65). The enhanced CFTR-ATPase activity in the presence of very low concentrations of CPT-cAMP and the reduced CFTR-ATPase activity in the presence of higher concentrations of CPT-cAMP thus likely arise from a direct combined modulatory interaction of CPT-cAMP and e.g. glibenclamide or glipizide with CFTR. Evidence for binding of cAMP to CFTR has been provided previously (66). Moreover, it is known that ABCC4, which is the ABC transporter most closely related to CFTR, is a cAMP and cGMP exporter (67,68). CPT-cAMP as well as forskolin may thus not only act as phosphorylation agents but also as modulators of the CFTR-ATPase.
Iodide Efflux at Different Concentrations of Phosphorylation Agents-To be able to directly compare ATPase activity and anion channel activity of CFTR, we used the same cell lines (CHO-CFTR and CHO-K1 cells) for both types of experiments.
Moreover, we tested whether cells responded equally in Cytosensor experiments if MEM␣ was replaced by iodide efflux buffer (2 mM HEPES versus 20 mM HEPES). As the basal ECARs were similar under the two conditions, we concluded that a direct comparison of ATP hydrolysis and iodide efflux was warranted. Fig. 10, A and B, shows the iodide flux at different concentrations of CPT-cAMP and forskolin, respectively, measured as a function of time. At low concentrations, maximum iodide efflux was reached after a 3-4-min incubation time; at high concentrations (C fsk ϭ 50 -100 M) maximum efflux shifted to shorter incubation times of maximally 1 min (data not included). Iodide efflux as a function of time is given in Fig. 10.

Kinetic constants of the CFTR-ATPase derived from microphysiometry experiments with CHO-CFTR cells in MEM␣ at pH 7.4 and T ‫؍‬ 37°C
The concentration of half-maximum activation (inhibition) K 1 (K 2 ) and the rate of activation (inhibition) V 1 (V 2 ) were obtained by fitting Equation 1 to the data.   efflux (V I ϭ 27.9 fmol cell Ϫ1 min Ϫ1 ). The highest CPT-cAMP concentration measured (C CPT ϭ 1 mM) yielded similar although a slightly lower iodide efflux value, suggesting a second plateau. As the experiment at C CPT ϭ 1 mM was not fully reversible, the value was not included in Fig. 11A. Fig. 11B shows an analogous plot for forskolin. The iodide efflux increased with forskolin concentration up to a first plateau (V I Ϸ18.9 fmol cell Ϫ1 min Ϫ1 ) in the concentration range C fsk ϭ 3-10 M. A concentration of C fsk ϭ 10 M is generally used (9) to obtain full CFTR phosphorylation. At higher concentrations, again a distinct increase in iodide efflux was observed which lead to a second plateau (V I Ϸ25.6 fmol cell Ϫ1 min Ϫ1 ).
It is interesting to note that iodide efflux increased with the concentration of both phosphorylation agents, first in an ATPdependent manner up to similar iodide efflux values of ϳ18 fmol min Ϫ1 cell Ϫ1 (first plateau in Fig. 11, A and B) and second by an ATP-independent manner again up to comparable values (second plateau in Fig. 11). Whereas the first plateau is reached at full phosphorylation of CFTR, the second ATP-independent increase in iodide efflux (see also Fig. 12) can be explained by the modulatory (ATPase inhibitory) effect of phosphorylation agents. The potentially small decrease in iodide efflux at the highest concentrations of phosphorylation agents (C CPT ϭ 1 mM see text and C fsk ϭ 100 M, see Fig. 11B) was rather due to a general metabolic effect than to specific inhibitory phosphorylation sites in CFTR (69).

Iodide Efflux at Different Concentrations of Chloride Channel
Inhibitors and Potentiators-Iodide efflux experiments in the presence of chloride channel inhibitors and activators were again performed under conditions closely resembling those in microphysiometry, i.e. cells were pre-incubated with CPT-cAMP and were then titrated with the compound of interest. Because higher CPT-cAMP concentrations (C CPT Ն50 M) strongly enhanced iodide efflux, and lead to a complete iodide efflux from cells within 3 min or less, efflux experiments with chloride channel inhibitors and potentiators were performed with lower CPT-cAMP concentrations. In the presence of CPT-cAMP (C CPT ϭ 25 M), glibenclamide (C glib ϭ 2-50 M) (Fig. 5D), glipizide (C glipz ϭ 1-20 M) (Fig. 6C), and CFTRinh-172 (C inh172 ϭ 0.1-1 M) (Fig. 7D) did not potentiate CFTR-dependent iodide efflux in CHO-CFTR cells. Conversely, genistein (C gen Ն10 M) distinctly increased iodide efflux under the same conditions (Fig. 8C).
As seen from literature data included in the legends to Figs. 5-7, chloride channel inhibitors reduce anion channel activity already in the lower concentration range where ATPase activity increased and reached a maximum. Inhibition persisted at higher concentrations, where ATPase activity decreased again. In the low concentration range where ATPase activity was high, occlusion of the channel is most likely the result of a dynamic process (see "Discussion"). In the high concentration range, occlusion can be explained in terms of the two-site binding model (Equation 4) and most likely results from "static" channel occlusion by inhibitors as described previously (70). As many  CFTR chloride channel inhibitors are anionic amphiphiles, inhibition may be enhanced by repulsive electrostatic interactions.
Conversely, chloride channel potentiators barely enhanced the ATPase activity of CFTR at low concentrations, in agreement with previous investigations (71), and even reduced it at higher concentrations. At the highest concentrations the potentiators also inhibited chloride flux (62) by channel occlusion, most likely again due to more than one molecule bound to the TMDs.
Contribution of Adenylate Cyclase Activity-Whereas CPT-cAMP directly interacts with PKA, forskolin stimulates cAMP synthesis from ATP by activating adenylate cyclase. Because adenylate cyclase activity (72) and phospho group transferase activity of PKA (73) both work in the time frame of minutes, experimental differentiation between the two processes was not possible. To solve this problem, we compared the ATP consumed to induce a given iodide efflux with either CPT-cAMP or forskolin. As seen in Fig. 12, the ATP hydrolysis rates required for a given iodide flux were identical at low concentrations of phosphorylation agents. However, at higher forskolin concentrations (C fsk Ͼ1 M), the ATP hydrolysis rate significantly increased and was clearly higher than in the presence of CPT-cAMP. The higher energy cost in the presence of forskolin can be attributed to the activity of adenylate cyclase.
Estimating CFTR-mediated Iodide Efflux from Loaded Cells-The present data allow a crude estimate of the iodide molecules passing through CFTR per ATP hydrolyzed. If we assume that the catalytic rate constant of the CFTR-ATPase in the presence of high concentrations of CPT-cAMP is on the order of 1% of the total CFTR-related ATP hydrolysis rate (k cat, CFTR ϳ1 ATP s Ϫ1 ) and that, under these conditions, iodide efflux is Ն30 fmol min Ϫ1 cell Ϫ1 , the number of iodide molecules flowing out of the iodide-loaded cells through CFTR was estimated as n Ն10 4 s Ϫ1 . This number is lower than the average number (n Ϸ10 6 s Ϫ1 ) estimated by patch clamp experiments (74). Nevertheless, the data show that the number of anions passing through the channel during one catalytic cycle is high.

Discussion
Here, we provide a detailed analysis of the CFTR-ATPase activity and CFTR phosphorylation by PKA in live cells, using the ECAR and OCR as a proxy for the ATP consumption/ hydrolysis rate. In the unperturbed cellular environment, the different ATP-consuming processes could be measured with a good signal to noise ratio that allowed unraveling their influence on CFTR channel gating in an unambiguous manner. Taken together, these results lead to a new model for the catalytic cycle of CFTR.
CFTR Phosphorylation, CFTR-ATPase Activity, and Anion Flux-First, we showed that the total CFTR-related ATP hydrolysis rate (or total, apparent catalytic rate constant) in the presence of CPT-cAMP was k cat, app ϭ 140 s Ϫ1 . The major contribution (Ն90%) was due to phosphorylation of the approximately 10 serine residues in the R-domain of CFTR. If phosphorylation was stimulated with forskolin, the rate of ATP hydrolysis was even higher, because adenylate cyclase also contributed (Fig. 12).
The catalytic rate constant of the partially phosphorylated CFTR-ATPase was determined as k cat, CFTR Յ14 s Ϫ1 . Whether ATPase activity increased (16), stayed constant (75), or even decreased with increasing phosphorylation was at first sight difficult to decide, because of the low basal activity of the CFTR-ATPase compared with the activity of PKA. The good fit of Equation 3 to PKA activity in Figs. 2A and 4A supports a constant CFTR-ATPase activity. However, data summarized in Table 3 show that CPT-cAMP and forskolin act as modulators in combination with other compounds. The modulator-induced CFTR-ATPase activity showed bell-shaped dependence on concentration and decreased at high concentrations.
Second, we showed that the ATPase-incompetent mutation, CFTR-E1371S, cannot be phosphorylated, suggesting that phosphorylation occurs only if the post-hydrolysis transition state can been reached. Although CFTR-ATPase activity was required for phosphorylation, it was not required for anion flux through CFTR-E1371S, most likely because the mutation as such influences the stability of the protein (76) and may open the ion channel constitutively as shown for a related mutation CFTR-E1371Q (77).
Third, we demonstrated that the ATPase activity of phosphorylated CFTR could be greatly enhanced with anion channel inhibitors but not with anion channel potentiators. The latter rather reduced the ATPase activity of CFTR and, as a side effect, induced a shift from glycolysis to cellular respiration.
Iodide efflux increased with the concentration of phosphorylation agents and reached a first plateau at complete phosphorylation (Fig. 11). Interestingly, further addition of phosphorylation agents led to a further increase in anion efflux without FIGURE 12. Total CFTR-related ATP hydrolysis rate versus maximum iodide efflux. The total CFTR-related ATP synthesis/hydrolysis rate was derived from ECAR and OCR measurements at superfusion with phosphorylation agents for 6 min to approximate the time frame of iodide efflux measurements more closely. Iodide efflux increased with the total CFTR-related ATP hydrolysis rate up to a certain value in the presence of both CPT-cAMP (f) and forskolin (E). At high concentrations of both phosphorylation agents, the iodide efflux increased in an ATP-independent manner. The amount of ATP required for a given iodide efflux was significantly higher upon phosphorylation with forskolin than upon phosphorylation with CPT-cAMP. Because it can be assumed that PKA and CFTR require the same amount of ATP under both conditions, the surplus in ATP hydrolysis in the presence of forskolin can be attributed to the synthesis of cAMP from ATP catalyzed by adenylate cyclase. Dotted lines were added to guide the eye.
further stimulation of ATP hydrolysis (Figs. 11 and 12). This further increase in iodide efflux can be explained by the modulatory (ATPase-inhibitory) effect of phosphorylation agents at high concentrations.
New Model for the Catalytic Cycle of CFTR-The strong similarity between ATPase activity curves of CFTR (Figs. 5-7) and ATPase activity curves of P-glycoprotein (59) and Sav1866 (78,79), a bacterial homolog of P-glycoprotein, suggests that CFTR binds and flips (or rather flops) amphiphilic compounds from the inner to the outer membrane leaflet as the other two transporters. The fact that transport by CFTR has been difficult to observe may be due to fast diffusion of these compounds relative to transport (80). As for P-glycoprotein, CFTR binds and transports a mixed set of allocrites, including, for example, chloride channel inhibitors and the phosphorylation agents CPT-cAMP and forskolin.
Because of these similarities, we tested whether a model for the catalytic cycle of Sav1866, based on molecular dynamics simulations (81), could also be applied to CFTR. The model (Scheme 1) assumes that the transporter remains more or less associated at the NBDs during the whole catalytic cycle (82) and alternates between an outward closed conformation with two identical nucleotides bound (e.g. two ATPs or two ADPs) and an outward open conformation, induced by ATP hydrolysis at a single site. This model is in contrast to previous models (see Introduction); however, it is in agreement with the recent investigation of the accessibility of cysteines in CFTR's TMDs by cysteine-reactive probes (34), and it also agrees with the finding that addition of ADP at higher (millimolar) concentrations closes the anion channel of CFTR (83), most likely due to binding of two ADPs.
The fact that hydrolysis was required for phosphorylation to occur pointed to the relevance of the post-hydrolysis transition state. We therefore assumed that the post-hydrolysis transition state corresponds to the open state of the anion channel. The assumption is consistent with previous measurements in inside-out membrane patches showing that efficient stabilization of the open channel was achieved by orthovanadate, known to bind to the post-hydrolysis transition state (84 -86). Moreover, it is supported by the observation that non-hydrolyzable ATP analogs do not open (83) CFTR.
Here, we demonstrated that highly ATPase-activating compounds act as chloride channel inhibitors at low and intermediate concentrations (see activating branch of the ATPase activity curve in Figs. 5-7). Inhibition of anion flux occurs most likely, because the dwell time in the outward open conformation (i.e. post-hydrolysis transition state) is short compared with the dwell time in the closed conformation. Mechanistically this dynamically closed "state" differs from the occluded state described previously (70). The occluded state is consistent with conditions found at high concentrations of chloride channel inhibitors (i.e. decreasing branch of the ATPase activity curve) where most likely more than one molecule is bound to CFTR.
Hence, to obtain significant anion flux, consisting of Ͼ Ͼ10 4 s Ϫ1 , the post-hydrolysis transition state has to be stabilized. We suggest that the role of the highly negatively charged phosphorylated R-domain is to delay nucleotide exchange in the post-hydrolysis transition state and to thereby prolong channel open time. An analogous phosphorylation-dependent slowdown in activity was observed for Rap1 (87).
The open state can also be prolonged by binding of potentiators that reduce the ATPase activity and slow down flopping and drug release (rate constant k 2 ). As shown for P-glycoprotein, drug release is rate-determining (88).
Upon exchange of ADP at NBD2 for ATP, which leads to a state where two ATPs are bound to the NBDs, the TMDs close at the extracellular side and thereby squeeze out the compounds that have not yet left the cavity toward the extracellular lipid leaflet. At this point the catalytic cycle can restart.
With the present analysis, we demonstrated that measuring the CFTR-related ATP hydrolysis rate in live cells is complementary to electrophysiological measurements of CFTR. Together, the different types of techniques provided a coherent view of the complex interplay between CFTR phosphorylation, CFTR-ATPase activity, and anion channel gating.   (81). In this conformation a drug, D (e.g. chloride channel potentiator or inhibitor), can enter the cavity from the cytosolic leaflet of the membrane. Upon hydrolysis and release of inorganic phosphate, P i , the post-hydrolysis transition state of CFTR is reached (CFTR(D)(ATP/ADP)). The post-hydrolysis transition state corresponds to the outward open conformation that allows for anion flux through CFTR. In cells the post-hydrolysis transition state is stabilized by the R-domain that adopts a net negative charge upon phosphorylation as indicated by the red circle. The high net negative charge in the vicinity of the NBDs may prevent nucleotide exchange by electrostatic repulsion and may delay anion channel closing. CFTR with an empty NBD2 (CFTR(ATP/Ϫ) is still open but short lived. The rate constant, k 1 , includes drug, D, binding from the lipid membrane to the TMDs of CFTR and release of inorganic phosphate upon ATP hydrolysis. The rate constant, k 2 , includes drug transport or flopping and release, which may be rate-limiting (36,45) as well as ADP release. The rate constant, k 3 , corresponds to ATP binding and concomitant squeeze-out of the drug, D, that has not left the transporter cavity before. The apo-conformation was not considered, because the cellular ATP concentration is typically C ATP ϭ 1-10 mM (96,97), and the concentration of half-maximum activation of CFTR by ATP was determined as K 0.5 Ϸ50 M for CFTR (27).