INTRODUCTION

The cystic fibrosis transmembrane conductance regulator (CFTR)¹ is an integral membrane protein that resides on the apical membrane of several types of epithelial cells, and its absence or dysfunction in cystic fibrosis is thought to impair severely the salt and water transport capacity of these cells (1). CFTR shares structural similarities with other members of the ATP-binding cassette (ABC) superfamily of traffic ATPases; namely, it possesses two repeats, each consisting of a membrane-spanning domain followed by a nucleotide-binding domain (2, 3). Furthermore, CFTR possesses a unique cytoplasmic domain, the R domain, which possesses several putative sites for PKA and PKC phosphorylation and links the two halves of the molecule (2).

Unlike other members of the ABC superfamily, CFTR exhibits chloride channel activity. The chloride channel function of CFTR was implicated in mutagenesis studies in which substitution of certain charged residues within the putative transmembrane-spanning domains caused changes in single channel conductance or anion selectivity (4, 5). Further convincing evidence of the chloride channel activity of CFTR came from our electrophysiological studies of purified, reconstituted CFTR (6). Fusion of liposomes containing purified CFTR with planar lipid bilayers resulted in the appearance of chloride-selective channels that exhibited biophysical features identical to those associated with CFTR expression in epithelial cell membranes (7, 8).

Phosphorylation by PKA is required but not sufficient to activate CFTR chloride channel function. Electrophysiological evidence supports the hypothesis that hydrolysis of ATP is essential for normal opening and closing or gating of the channel (9, 10, 11, 12). First, the addition of Mg-ATP, but not nonhydrolyzable ATP analogs, to PKA-phosphorylated CFTR causes channel gating (10). Second, the subsequent addition of nonhydrolyzable ATP analogs to ATP-activated CFTR channels interferes with normal channel gating kinetics (13). Third, chelation of magnesium ion, an essential cofactor in most reactions involving nucleotide binding and/or hydrolysis, causes alterations in the rates of channel opening and closing (14, 15). However, other research groups have reported electrophysiological studies with conflicting results, namely that nonhydrolyzable analogs of ATP fail to compete with the opening or closing of CFTR channels (16) and that gating persists, albeit at a slower rate, after removal of magnesium (17, 18). As it has been widely hypothesized that CFTR is functionally coupled to other membrane proteins through direct protein interactions (19), it is possible that the sensitivity of CFTR channel gating to ATP is conferred or modulated by associated membrane proteins. Hence, contradictory reports such as those described above may reflect variabilities in the membrane environment for CFTR.

To eliminate possible modulatory effects of associated membrane proteins, we examined the role of ATP hydrolysis in CFTR channel gating using purified, reconstituted CFTR protein. We present novel biochemical evidence that the CFTR protein possesses intrinsic ATPase activity and that this catalytic activity is coupled directly to the channel function of this protein.

EXPERIMENTAL PROCEDURES

Two liters of Sf9 cells were infected with recombinant baculovirus containing the complete coding sequence for CFTR as described previously (6). CFTR purification was performed according to our published procedure (6) with the following modifications. Infected cells were harvested after 48 h, and the pellet was treated with a phosphate-buffered saline containing 2% Triton X-100, 2 units/ml DNase, 5 mM MgCl₂, 2 mM dithiothreitol, 10 µg/ml leupeptin, 10 µg/ml aprotinin, 1 mM benzamidine, and 10 µM E64. The mixture was stirred for 2 h at 15 °C after which the insoluble material was centrifuged at 100,000 × g for 2 h. The resulting pellets were treated with 180 ml of 2% SDS, 3% mercaptoethanol in 10 mM sodium phosphate, pH 7.2, and the mixture was stirred overnight at 4 °C. Insoluble material was centrifuged at 60,000 × g, and the supernatant was filtered through a 0.22-µm filter before being applied at 1 ml/min to a ceramic hydroxyapatite column. Washing, elution, and identification of the CFTR-containing peaks were performed as described previously (6). The CFTR-containing fractions from the ceramic hydroxyapatite column were concentrated and applied to a Superose 6 column as the final purification step. The purified protein was quantitated by amino acid analysis. NH₂-terminal sequencing and amino acid analysis were performed by the HSC-U of T-Pharmacia Biotechnology Centre. An identical procedure was used for the purification of the CFTR variant CFTRG551D.

SDS-polyacrylamide gel electrophoresis of purified CFTR and CFTRG551D protein was performed using 6% acrylamide gels as described previously (20). Total protein was detected with silver stain as described previously. For immunoblotting, protein was transferred to nitrocellulose, and CFTR was probed with an anti-CFTR monoclonal antibody as described previously (20). Immunoreactive bands were visualized by chemiluminescence using the ECL system (Amersham Corp.)

The CFTR sample (5 µg) in 1 ml of solution containing 10 mM Tris buffer, pH 7.8, containing 0.25% lithium dodecyl sulfate and 100 mM NaCl was concentrated in a Centricon concentrator (molecular mass cutoff, 100 kDa) to a volume of 100 µl. The protein sample was then diluted to 1,000 µl with 8 mM Hepes buffer containing 0.5 mM EGTA, pH 7.2 (buffer A), and reconcentrated to a 100-µl volume. The final lithium disulfate concentration was 0.025%. Liposomes, 10 mg/ml, were prepared in buffer A as described previously using a lipid mixture of PE:PS:PC:ergosterol 5:2:1:1 by weight. 100 µl of liposomes containing 1 mg of lipid was added to 100 µl of the CFTR sample yielding a final lithium dodecyl sulfate concentration of 0.0125%. The mixture was dialyzed (Spectrapor; molecular mass cutoff, 50 kDa) for 18 h against 2 liters of buffer A containing 1.5% sodium cholate. Dialysis was then continued against 4 liters of buffer A for 2 days with daily changes of buffer. For lipid bilayer studies, nystatin was introduced into the reconstituted proteoliposomes as described previously (6). An identical protocol was employed for the reconstitution of CFTRG551D.

ATPase activity was measured as the production of [-³²P]ADP from [-³²P]ATP using polyethyleneimine-cellulose chromatography for separation of the nucleotides. Unless otherwise stated, the assay was carried out in a 15-µl reaction mixture containing 50 mM Tris, 50 mM NaCl, pH 7.5, 2 mM MgCl₂, 10% glycerol, 0.5 mM CHAPS, and 8 µCi of [-³²P]ATP. Reaction mixtures were incubated at 30 °C and were stopped by the addition of 5 µl of 10% SDS. One-µl samples were spotted on a polyethyleneimine-cellulose plate and developed in 1 M formic acid, 0.5 M LiCl. The location and quantitation of the radiolabeled ATP and ADP were determined with a Molecular Dynamics PhosphorImager. Data were analyzed using the ImageQuant software package (Molecular Dynamics).

CFTR proteoliposomes and liposomes without CFTR, both in 50 mM Tris, 50 mM NaCl at pH 7.4, were mixed with catalytic subunit of PKA (final concentration, 200 nM), 20 µM ATP, and 5 mM MgCl₂, sonicated briefly, and incubated for 1 h at room temperature. These conditions mimic those used for in vitro phosphorylation of CFTR immunoprecipitated from Chinese hamster ovary cells (21). To confirm that purified, reconstituted CFTR was phosphorylated under these conditions, [-³²P]ATP was used in one experiment to phosphorylate CFTR. An autoradiograph of the phosphorylated protein run on 6% SDS-polyacrylamide gel (22) showed a single intense band corresponding to CFTR. To remove PKA after the phosphorylation reaction, CFTR proteoliposomes and control protein-free liposomes were airfuged at 100,000 × g for 30 min and then washed twice by sonication with buffer and pelleted in the Airfuge. Pellets were resuspended in the appropriate buffer for ATPase measurements. For bilayer studies of channel function, CFTR was phosphorylated as described in the preceding paragraph. Subsequently, PKA was separated from the proteoliposomes by microspin chromatography using Sephadex G-50. To validate complete removal of PKA, we assessed the presence of ¹²⁵I-labeled catalytic subunit of PKA in the proteoliposome sample and determined that three cycles of sonication followed by separation on a microspin column were needed for complete removal of the labeled subunit of PKA. To dephosphorylate CFTR, proteoliposomes in 50 mM Tris, 50 mM NaCl (1 µg of purified protein/50 µl, pH 7.4) were mixed and incubated with protein phosphatase 2A (3.75 units/ml; Calbiochem) for 2 h at room temperature. This concentration of protein phosphatase 2A has been shown to dephosphorylate CFTR in biological membranes (23). Prior to treatment, the phosphatase was dialyzed against the above Tris buffer to partially remove glycerol in the storage medium.

Purified CFTR was reconstituted into proteoliposomes containing a phospholipid mixture PE:PS:PC:ergosterol (5:2:1:1) by weight for bilayer studies. Nystatin (120 µg/ml) was introduced into the proteoliposomes for bilayer studies to promote proteoliposome fusion and to facilitate detection of these fusion events (24). Fusion of nystatin-containing liposomes was indicated by the appearance of transient ``nystatin spikes'' in bilayer conductance. As in our previous experiments, a 10 mg/ml solution of phospholipid (PE:PS at a ratio of 1:1, Avanti Polar Lipids) in n-decane was painted over a 200-µm aperture in a bilayer chamber to form bilayers. Bilayer formation was monitored electrically by observing the increase in membrane capacitance. In all experiments, bilayer capacitance was greater than 200 picofarads. Fusion of liposomes was potentiated with the establishment of an osmotic gradient across the lipid bilayer. The cis compartment of the bilayer chamber, defined as that compartment to which liposomes were added, contained 300 mM KCl; the trans compartment, connected to ground, contained 50 mM KCl. Single channel currents were monitored at a holding potential of 40 mV applied to the cis compartment and detected with a bilayer amplifier (custom made by M. Shen, Physics Laboratory, University of Alabama). Data were recorded and analyzed using pClamp 6.0.2 software (Axon Instruments, Inc). Prior to analysis of open probability and dwell times, single channel data were filtered digitally at 100 Hz. Ideal records were created by use of a half-height transition protocol (pClamp 6.0.2 software).

Open and closed time histograms were created with a logarithmic x axis with 10 bins/decade and a lower limit of 10 ms. The maximum likelihood method was used to fit the data with one or two exponentials (pClamp 6.0.2 software). The goodness of fit was assessed using the log likelihood ratio test. The mean open and closed time constants derived from histograms generated from single experiments (approximately 5 min in duration) were comparable to values derived from histograms generated from four combined experiments (i.e. four separate protein preparations). Hence, we have shown open and closed time histograms of combined experiments. Further, combination of data from multiple experiments permits the generation of burst duration histograms. Burst analysis was performed using tc = 60 ms. tc, the time that separates short, intraburst closures from long, interburst closures, was estimated using several approaches (25, 26). The latter approach described by Winter et al. for CFTR in biological membranes yielded a value for tc which led to the optimal separation of intraburst and interburst closed times. According to Winter and co-workers, tc can be determined by examination of the biexponential fit of the closed time histogram for single channel activity. tc corresponds to the nadir between the two peaks of the relationship defining the first and second closed time constants. A value for tc of 60 ms was determined as the mean from four different closed time histograms corresponding to the four different protein preparations (60.5 ± 2.1 ms) as well as from the closed time histogram generated from the combination of the four experiments (60.1 ± 0.5 ms).

Results are expressed as means ± S.D. Statistical significance was assessed using the log likelihood ratio test or Student's t test.

RESULTS

The scheme for purification of recombinant CFTR from Sf9 cells and reconstitution in phospholipid liposomes was published previously (6). We used a similar purification and reconstitution strategy in our present studies of CFTR ATPase activities. We confirmed the homogeneity of the purification products used in these experiments by evaluation of silver-stained gels overloaded with CFTR protein and by Western blot analysis of the reconstituted protein (Fig. 1A). In Fig. 1B we show that reconstituted, purified CFTR protein does possess intrinsic ATPase activity. The production of radiolabeled dinucleotide [-³²P]ADP from [-³²P]ATP was determined following separation of the nucleotides on thin layer chromatography plates, a technique commonly used for the measurement of GTPase activity of purified G proteins (27, 29). The rate of [-³²P]ADP production by liposomes containing purified, PKA-phosphorylated CFTR was linear over a 5-h time period (Fig. 1B). ATPase activity was conferred by purified CFTR protein as there was no detectable ATPase activity in suspensions of protein-free liposomes. In Fig. 1C we show that our measurements of CFTR ATPase activity are reproducible. The mean ATPase activity exhibited by three distinct preparations of purified CFTR is significantly greater than that detected in preparations of protein free liposomes (p < 0.02).

As in the case of CFTR chloride channel activity, CFTR ATPase activity is dependent upon phosphorylation (Fig. 2). Pretreatment of purified CFTR with the serine/threonine protein phosphatase 2A caused marked inhibition of ATP hydrolysis (Fig. 2A). Hence, purified CFTR likely remains partially phosphorylated throughout the purification procedure. Modulation of CFTR ATPase activity by phosphorylation is shown in further studies where CFTR is fully phosphorylated with PKA catalytic subunit. PKA pretreatment of CFTR caused a 2-3-fold activation of the rate of ATP hydrolysis. The substrate dependence for ATPase activity of both partially phosphorylated (not pretreated with PKA) and fully phosphorylated (PKA-treated) protein was determined and the data fitted by nonlinear regression analysis using the Hill equation, v = S^mV_max/S^m + K, where v and V_max are the initial and maximum initial rates, respectively; S is the ATP concentration, and m is the Hill coefficient. The constant K is related to K_m by the equation K = (K_m)^m. As apparent in Fig. 2B and Table I, PKA treatment modifies ATPase activity of CFTR by causing a reduction in apparent K_m from 1 mM to 0.3 mM, presumably by increasing the affinity of CFTR for ATP. The shape of the function describing ATP dependence of ATPase activity changed from hyperbolic, for the partially phosphorylated protein, to sigmoidal for the PKA-treated, fully phosphorylated protein. This change in shape was reflected by a change in the Hill coefficient from 1 to 1.7. Therefore, full phosphorylation of CFTR appears to induce positive cooperativity between two sites of ATPase activity. The V_max for CFTR ATPase activity, 50 (nmol/mg of purified protein/min), was not altered by PKA treatment. It is probable that we have underestimated the ATPase activity of CFTR as it is likely that only a fraction of purified protein has been reconstituted completely to its active configuration. Previously, we estimated on the basis of reconstitution of single channel activity, that 20-40% of total purified protein was functional following fusion to planar lipid bilayers (6). Hence, the ATPase activity by CFTR may be as high as 125-250 nmol/mg/min, which corresponds to a turnover number of approximately 0.5-1 molecules of ATP hydrolyzed/s.

Table I. PKA modification of CFTR ATPase activity Kinetic parameters Vmax Kma Hill coefficientb nmol/mg protein/min µM CFTR 51.3 989.7 1 CFTR-Pc 53.8 303 1.73 a  Calculated from K = (Km)m. b  Obtained by nonlinear regression analysis to the Hill equation. c  Phosphorylated CFTR.

We have shown in previous publications that purified, reconstituted CFTR exhibits channel activity with biophysical properties identical to those observed for CFTR in cell membranes; namely, it is anion-selective, exhibits a low unitary conductance of approximately 10 picosiemens, and requires phosphorylation by the addition of exogenous PKA plus Mg-ATP for activity (6, 28, 29, 30). Addition of Mg-ATP but not its nonhydrolyzable analog, Mg-AMPPNP, caused channel activity of phosphorylated CFTR reconstituted in planar lipid bilayers (data not shown), supporting the hypothesis that ATP hydrolysis rather than binding is required to open the CFTR channel.

It has been postulated that ATP is utilized by CFTR to cause gating of its chloride channel activity. Hence, the rate of ATP hydrolysis by CFTR should be comparable to the rate of channel gating. As we have quantitated ATP hydrolysis by purified CFTR we can examine this hypothesis directly in the same protein specimen. Our planar bilayer studies of purified protein verified patch-clamp studies that showed that CFTR exhibits a bursting pattern of gating in the presence of 1 mM Mg-ATP (Fig. 3A). As shown in Fig. 3B, there is a single population of channel-open times and two populations of channel-closed times, a short and a long closed time. Only the longer of the two closed times exhibits ATP dependence, suggesting that only the transition from the long duration closed time to the channel-open state is substrate-dependent. For subsequent burst analysis, we used a value of 60 ms as the burst delimiter (tc). tc was determined according to the method described by Winter et al. (26). Briefly, tc was derived from the function describing the closed dwell time histogram as the nadir between the peak values that define the mean short closed (intraburst) times and the long closed (interburst) times. Following analysis of the ATP dependence of burst and interburst duration histograms (Fig. 3C), we found that both the transition rate to the open burst (1/between burst duration) and the transition rate to the interburst, closed state (1/burst duration) were altered by increasing ATP concentration. The effective bursting rate increased with increasing ATP concentration, and this relationship was fit by nonlinear regression analysis using the Hill equation to yield a K_m of 584 µM, maximal opening rate of 10/s, and a Hill coefficient of 2. The Hill coefficient of 2 suggests that there is positive cooperativity between two sites through which ATP can act to promote channel bursting. The effective closing rate from an open burst decreased slightly with increasing ATP concentrations, and these data were fit using an exponential decay function to derive an estimate of the ATP concentration required for half-maximal inhibition of closing, 434 µM. Assuming that the overall transition rate between the two conductance states of CFTR channel is limited by the slowest rate, i.e. bursting or closing rate, we predict that at 1 mM ATP, the slow closing rate will limit gating of CFTR to 1-2 transitions/s. Our quantitation of the rate of CFTR gating is comparable to the rate of ATP hydrolysis by CFTR, namely, 0.5-1 molecules of ATP hydrolyzed/s. Hence, we argue on the basis of our kinetic analyses that CFTR channel activity is coupled to ATP turnover.

Further evidence for coupling between CFTR channel and ATPase functions is found in our studies of the sensitivity of these two activities for magnesium ion and sodium azide. In Fig. 4A we show that chelation of magnesium ion with EDTA inhibits channel opening as well as ATPase activity. Phosphorylated CFTR protein reconstituted in planar lipid bilayers showed channel activity when exposed to MgCl₂ (10 µM) plus Na-ATP (100 µM), where the free Mg²⁺ concentration was estimated at 4 µM (31). The subsequent addition of EDTA (1 mM) reduced free Mg²⁺ to the concentration of 4 nM and abolished channel gating. Similarly, ATPase activity of phosphorylated CFTR was reduced upon chelation of Mg²⁺. As further evidence for coupling between CFTR channel and ATPase activities, we found that both functions exhibited sensitivity to sodium azide (Fig. 4B). Sodium azide (NaN₃) (1 mM), a well known inhibitor of the F₁F₀-ATPase (32), inhibited opening of the purified CFTR chloride channel as well as CFTR ATPase activity.

Mg

Finally, we assessed coupling between the ATPase and channel functions of CFTR by determining the effect of a site-directed mutation on both activities. The relatively common cystic fibrosis mutation, CFTRG551D, leads to severe disease (33). Glycine 551 lies within a sequence in the first nucleotide-binding fold, D-X-[G/A]-G-Q, which shows remarkable conservation in ABC transporters and in G proteins and has been implicated in NTP-binding and hydrolysis (34, 35, 36). In the present studies, we found that purified CFTRG551D exhibits altered chloride channel activity when reconstituted in lipid bilayers (Fig. 5A). Although the unitary conductance of CFTRG551D is comparable to that of the wild type protein, approximately 10 picosiemens, arguing against global changes in protein structure, the open probability of the mutant protein was much lower than that determined in the wild type protein, 0.011 versus 0.48, respectively. Similarly, reconstituted CFTRG551D protein exhibits altered ATPase activity (Fig. 5B). At 1 mM Mg-ATP the rate of ATP hydrolysis by the mutant protein occurs at approximately 10% that of the wild type (2.5 nmol/mg of protein/min versus 20 nmol/mg of protein/min). Hence, we have shown that a mutation within a sequence conserved among NTPases causes a reduction both in channel gating and ATPase activity.

DISCUSSION

In these studies of purified, reconstituted CFTR protein we have provided the first direct biochemical evidence that this protein functions as an ATPase. We measured the ATPase activity of purified CFTR as 50 nmol/mg/min and estimated that this activity may be as high as 125-250 nmol/mg/min. Significantly, Ko and Pedersen (37) recently reported that a fusion protein containing the first nucleotide-binding fold of CFTR was also capable of hydrolyzing ATP, albeit at a somewhat slower rate of 30 nmol/mg/min. Our estimate of CFTR ATPase activity is less than that measured for P-glycoprotein, a closely related member of the ABC superfamily of transporters, with hydrolytic rates measured as 300 (nmol/mg/min) (38) and 1,650 (nmol/mg/min) (39) and less than rates determined for purified P-type ATPases; Na⁺,K⁺-ATPase (800 nmol/mg/min) and Ca²⁺-ATPase (600 nmol/mg/min) (40). On the other hand, CFTR ATPase activity is similar to that reported for the intrinsic GTPase rates of dynamin (41) and much higher than intrinsic rates reported for the low molecular weight GTPases such as ras (42). The low rate of ATP hydrolysis by CFTR may explain why it cannot be measured in native cell membranes² nor in some preparations of fusion proteins containing the first nucleotide-binding fold of CFTR (43).

The results presented in this paper support the hypothesis that CFTR ATPase and channel activities are coupled. Both CFTR ATPase activity and channel gating are regulated by PKA phosphorylation. Our kinetic analyses suggest that PKA phosphorylation enhances CFTR ATPase activity by increasing affinity for ATP, possibly through exposure of a second catalytic site. Although this hypothesis must be examined directly in future studies, we speculate that the phosphorylated R domain may function to coordinate the activities of the two nucleotide-binding folds of CFTR. It is well known that CFTR channel gating requires phosphorylation, and a role of the R domain in the coordination of ATP-dependent channel gating has also been implicated (12). Further, we determined that the rates of ATP turnover and channel gating by phosphorylated, purified CFTR are similar, approximately 1/s, attesting to the proposal that these activities are coupled.

The requirement for CFTR ATPase activity in channel gating was assessed by inhibition of ATPase activity by site-directed mutagenesis. Glycine 551 of CFTR lies within a motif that is conserved in nucleotidases, and mutation of this residue to aspartic acid caused a marked reduction in both ATPase activity and channel gating. Hence, the mutation CFTRG551D likely causes altered channel activity (44) and human disease (33) because it possesses defective ATPase activity. Further, chemical modulation of CFTR ATPase activity caused comparable inhibition of the rate of channel gating. In contrast to some electrophysiological studies of CFTR in biological membranes (17, 18), we could observe a clear dependence of channel activation on magnesium ion. This difference may relate to an improved ability to regulate free magnesium ion concentrations in bilayer studies of purified protein.

To understand how CFTR functions in cells it is important to determine which of its putative activities are intrinsic or dependent upon the context of other membrane proteins. This issue is of particular relevance to CFTR and other members of the ABC superfamily of membrane proteins as it has been hypothesized that they may mediate cellular transport through direct protein-protein interactions within the membrane (19). Studies of purified, reconstituted CFTR can be used to verify or to argue against putative activities of CFTR (6, 30). For example, the chloride channel activity of CFTR was originally verified in planar lipid bilayer studies of purified CFTR (6). Conversely, the putative ATP channel function of CFTR could not be confirmed in our recent studies using purified protein (30), leading us to speculate that CFTR-mediated cellular efflux of ATP reported in some studies (45) may be due to an effect of CFTR on associated membrane proteins.

Finally, the results in the present paper clearly show that CFTR possesses intrinsic ATPase activity and support the hypothesis that CFTR utilizes ATP to gate its channel function. Hence, the ATP dependence of channel activity is unlikely to be mediated by neighboring membrane proteins. These results pave the way for more detailed investigations of the mechanism through which ATP is utilized to drive CFTR channel gating. Furthermore, other putative activities for this protein, such as its possible function as a pump for organic substrates, can now be tested directly (3).