Direct Sensing of Intracellular pH by the Cystic Fibrosis Transmembrane Conductance Regulator (CFTR) Cl− Channel*♦

In cystic fibrosis (CF), dysfunction of the cystic fibrosis transmembrane conductance regulator (CFTR) Cl− channel disrupts epithelial ion transport and perturbs the regulation of intracellular pH (pHi). CFTR modulates pHi through its role as an ion channel and by regulating transport proteins. However, it is unknown how CFTR senses pHi. Here, we investigate the direct effects of pHi on recombinant CFTR using excised membrane patches. By altering channel gating, acidic pHi increased the open probability (Po) of wild-type CFTR, whereas alkaline pHi decreased Po and inhibited Cl− flow through the channel. Acidic pHi potentiated the MgATP dependence of wild-type CFTR by increasing MgATP affinity and enhancing channel activity, whereas alkaline pHi inhibited the MgATP dependence of wild-type CFTR by decreasing channel activity. Because these data suggest that pHi modulates the interaction of MgATP with the nucleotide-binding domains (NBDs) of CFTR, we examined the pHi dependence of site-directed mutations in the two ATP-binding sites of CFTR that are located at the NBD1:NBD2 dimer interface (site 1: K464A-, D572N-, and G1349D-CFTR; site 2: G551D-, K1250M-, and D1370N-CFTR). Site 2 mutants, but not site 1 mutants, perturbed both potentiation by acidic pHi and inhibition by alkaline pHi, suggesting that site 2 is a critical determinant of the pHi sensitivity of CFTR. The effects of pHi also suggest that site 2 might employ substrate-assisted catalysis to ensure that ATP hydrolysis follows NBD dimerization. We conclude that the CFTR Cl− channel senses directly pHi. The direct regulation of CFTR by pHi has important implications for the regulation of epithelial ion transport.

In cystic fibrosis (CF), dysfunction of the cystic fibrosis transmembrane conductance regulator (CFTR) Cl ؊ channel disrupts epithelial ion transport and perturbs the regulation of intracellular pH (pH i ). CFTR modulates pH i through its role as an ion channel and by regulating transport proteins. However, it is unknown how CFTR senses pH i . Here, we investigate the direct effects of pH i on recombinant CFTR using excised membrane patches. By altering channel gating, acidic pH i increased the open probability (P o ) of wild-type CFTR, whereas alkaline pH i decreased P o and inhibited Cl ؊ flow through the channel. Acidic pH i potentiated the MgATP dependence of wild-type CFTR by increasing MgATP affinity and enhancing channel activity, whereas alkaline pH i inhibited the MgATP dependence of wildtype CFTR by decreasing channel activity. Because these data suggest that pH i modulates the interaction of MgATP with the nucleotide-binding domains (NBDs) of CFTR, we examined the pH i dependence of site-directed mutations in the two ATPbinding sites of CFTR that are located at the NBD1:NBD2 dimer interface (site 1: K464A-, D572N-, and G1349D-CFTR; site 2: G551D-, K1250M-, and D1370N-CFTR). Site 2 mutants, but not site 1 mutants, perturbed both potentiation by acidic pH i and inhibition by alkaline pH i , suggesting that site 2 is a critical determinant of the pH i sensitivity of CFTR. The effects of pH i also suggest that site 2 might employ substrate-assisted catalysis to ensure that ATP hydrolysis follows NBD dimerization. We conclude that the CFTR Cl ؊ channel senses directly pH i . The direct regulation of CFTR by pH i has important implications for the regulation of epithelial ion transport.
The ATP-binding cassette (ABC) 3 transporter cystic fibrosis transmembrane conductance regulator (CFTR) (1) is a multi-functional protein best known as a regulated Cl Ϫ channel (2). CFTR is assembled from five domains: two membrane-spanning domains (MSDs) that form an anion-selective pore, two nucleotide-binding domains (NBDs) that bind and hydrolyze ATP to control channel gating, and a unique regulatory domain (RD), whose phosphorylation by PKA is critical for CFTR activation (2,3). CFTR is principally expressed in the apical membrane of epithelia throughout the body where it plays a fundamental role in fluid and electrolyte movements (4). Malfunction of CFTR causes the common genetic disease cystic fibrosis (CF) (4).
Previous studies demonstrate that the regulation of intracellular pH (pH i ) is defective in CF epithelial cells (e.g. Ref. 5). They also reveal that expression of recombinant CFTR in heterologous cells modulates pH i (e.g. Ref. 6). Analysis of the literature suggests that CFTR modulates pH i in three main ways. First, CFTR itself directly transports HCO 3 Ϫ ions with a modest permeability (P HCO 3 Ϫ/P Cl Ϫ ϳ0.26 (7)). Second, CFTR regulates the Na ϩ /H ϩ exchanger isoform 3 (NHE3), which contributes to Na ϩ -dependent HCO 3 Ϫ reabsorption in pancreatic duct epithelia. CFTR stabilizes NHE3 expression at the cell surface and inhibits NHE3 activity by a cAMP-dependent mechanism when pancreatic HCO 3 Ϫ secretion is stimulated (8). Of note, the regulation of NHE3 by CFTR involves the association of CFTR and NHE3 with the scaffolding protein EBP50 to form a macromolecular complex (8). Third, CFTR regulates the Cl Ϫ /HCO 3 Ϫ (anion) exchanger (AE), which plays a central role in pancreatic HCO 3 Ϫ secretion. CFTR regulation of AE requires the cell surface expression and cAMP-dependent phosphorylation of CFTR, but not its transport of anions (6). Interestingly, Ko et al. (9) demonstrated that CFTR and members of the SLC26 family of AEs coordinate their activities through the interaction of the, phosphorylated RD of CFTR with the STAS (sulfate transporter and antisigma-factor antagonist) domain of SLC26 transporters. Thus, CFTR modulates pH i through its roles as an ion channel and regulator of transport proteins.
A key unresolved question is how CFTR senses changes in pH i . As described above, CFTR might detect pH i changes indirectly through its interactions with NHE3 and SLC26 transporters. Consistent with this idea, Reddy et al. (10) demonstrated that pH i modulates indirectly the CFTR Cl Ϫ conductance of sweat duct epithelia by altering the enzymatic activities of the protein kinases and phosphatases that control the phosphorylation status of CFTR. However, the ATPase activity of some ABC transporters (e.g. the transporter associated with antigen processing (TAP)) is pH i -dependent (11), while the ATPase activity of an NBD1-RD-CFTR fusion protein is optimal at neutral pH (12). These data raise the possibility that pH i might regulate directly CFTR activity. To test this hypothesis, we used excised inside-out membrane patches from cells expressing recombinant wild-type human CFTR. When we observed that pH i has multiple effects on the singlechannel activity of CFTR, we used CFTR constructs containing site-directed mutations to explore the molecular mechanisms of CFTR regulation by pH i .

EXPERIMENTAL PROCEDURES
Cells and CFTR Expression-In this study we used mammalian cells heterologously expressing CFTR constructs. Where available, we used cells stably expressing CFTR constructs. These included (i) mouse mammary epithelial cells (C127 cells) expressing wild-type human CFTR, the CFTR variant ⌬R-S660A (13) or the CF mutant G1349D (14), (ii) Fischer rat thyroid epithelial cells expressing the CF mutant G551D (15), and (iii) NIH-3T3 cells expressing the CFTR construct K1250M (16). To study the CFTR variants K464A, D572N, and D1370N, we employed the vaccinia virus/bacteriophage T7 hybrid expression system to transiently express CFTR variants in HeLa cells as described previously (17,18). As a control, we verified that the single-channel behavior of wild-type human CFTR in different mammalian cells was equivalent (for further information, see supplemental "Results" and Figs. 1 and 2).
HeLa cells were purchased from the American Type Culture Collection (Manassas, VA). C127 cells expressing wildtype CFTR were a generous gift of Dr. C. R. O'Riordan (Genzyme, Framingham, MA), whereas Fischer rat thyroid cells were a generous gift of Drs. L. J. V. Galietta and O. Zegarra-Moran (Istituto Giannina Gaslini, Genova, Italy). Other cells and the vaccinia virus expression plasmid pTM1-CFTR4-expressing CFTR mutants were generous gifts of Professor M. J. Welsh (University of Iowa, Iowa City, IA). Cells were cultured and used as described (14 -16, 19).
Electrophysiology-CFTR Cl Ϫ channels were recorded in excised inside-out membrane patches using an Axopatch 200A patch-clamp amplifier and pCLAMP software (both from MDS Analytical Technologies, Union City, CA) as described (20).
After excision of inside-out membrane patches, we added PKA (75 nM) and ATP (1 mM) to the intracellular solution within 5 min of patch excision to activate CFTR. Once CFTR Cl Ϫ channels were fully activated, we varied the pH i value and/or [ATP] of the intracellular solution depending on the experimental protocol. Because MgATP controls CFTR channel gating (e.g. Refs. 21 and 22), we maintained the [MgATP] constant in different pH i solutions by adding different amounts of Na 2 ATP (for further information, see supplemental "Experimental Procedures" and Table 1). To minimize the rundown of CFTR Cl Ϫ channels in excised membrane patches, we added PKA to all intracellular solutions. However, for studies of some CFTR constructs and wild-type CFTR in the absence of Mg 2ϩ ions, it was also necessary to increase the [ATP] to 1 or 3 mM, respectively, to sustain channel activity.
To investigate how pH i modulates CFTR Cl Ϫ currents, we used membrane patches containing large numbers of active channels. For all other studies, we used membrane patches containing Յ5 active channels. We determined the number of channels in a membrane patch from the maximum number of simultaneous channel openings observed during the course of an experiment. To minimize errors when counting the number of active channels, we employed the strategies described by Cai et al. (14). For further information, see supplemental "Experimental Procedures." In all experiments, intracellular solutions titrated to different pH i values were compared with the average of pre-and postintervention control periods titrated to pH i 7.3, which contained the same concentrations of MgATP (or Mg 2ϩ -free ATP) and PKA.
To investigate the effects of voltage on the CFTR Cl Ϫ channel at different pH i , we bathed membrane patches in symmetrical Cl Ϫ -rich solutions (extracellular pH (pH e ) 7.3; pH i 7. CFTR Cl Ϫ currents were initially recorded on digital audiotape using a digital tape recorder (Biologic Scientific Instruments, model DTR-1204; Intracel Ltd., Royston, UK) at a bandwidth of 10 kHz. On playback, records were filtered with an eight-pole Bessel filter (model 902LPF2; Frequency Devices, Inc., Ottawa, IL) at a corner frequency of 500 Hz and acquired using a Digidata 1200 interface (MDS Analytical Technologies) and pCLAMP at sampling rates of 2.5 kHz (macroscopic currents) or 5 kHz (single-channel currents). For the purpose of illustration, current records were filtered at 500 Hz and digitized at 1 kHz.
To analyze the effects of pH i on CFTR Cl Ϫ currents, we determined the average CFTR Cl Ϫ current (I CFTR ) at a specific pH i by averaging all the data points collected at that pH i and subtracting basal currents recorded in the absence of ATP and PKA. To plot the relationship between pH i or [H i ϩ ], and CFTR Cl Ϫ current, current values at different pH i or [H i ϩ ] were ex-pressed as a percentage of the control CFTR Cl Ϫ current recorded at pH i 7.3. To measure i, Gaussian distributions were fit to current amplitude histograms.
For NP o , P o , and burst analyses, lists of open and closed times were created using a half-amplitude crossing criterion for event detection as described under supplemental "Experimental Procedures." Burst analysis was performed as described by Carson et al. (16) using membrane patches that contained a single active channel and a t c (the time that separates interburst closures from intraburst closures) determined from analyses of closed time histograms (supplemental Table 2). The mean interburst interval (IBI or T IBI ) was calculated using Equation 1, where T b ϭ (mean burst duration) ϫ (open probability within a burst). Mean burst duration (MBD or T MBD ) and open probability within a burst (P o(burst) ) were determined directly from experimental data using pCLAMP software. P o was calculated from open and closed times as described (23). However, for G551D and G1349D, because the number of active channels in a membrane patch was unknown, we measured NP o instead of P o .
To calculated the apparent voltage-dependent dissociation constant (K d ) for OH Ϫ ion inhibition of CFTR and estimate the location of the binding site for OH Ϫ ions within the CFTR pore, we employed the methods described by Sheppard and Robinson (20). For further information, see supplemental "Experimental Procedures." Reagents-With the exception of PKA purified from bovine heart (Promega, Southampton, UK), chemicals were purchased from Sigma-Aldrich (Gillingham, UK). Stock solutions of ATP were prepared immediately before each experiment. When studying CFTR channel gating at ATP ϭ 10 mM, we added MgSO 4 (7 mM) to the intracellular solution. For nominally Mg 2ϩ -free intracellular solutions, we omitted MgCl 2 (3 mM) from the intracellular solution, titrated to pH 7.3 with HCl, and added EDTA (1 mM) to buffer free [Mg 2ϩ ] to Ͻ6 nM (22) before adjusting pH i to 6.3 with H 2 SO 4 and to pH 7.3 and 8.3 with Tris.
Statistics-Results are expressed as means Ϯ S.E. of n observations. To compare sets of data, we used Student's t test. Differences were considered statistically significant when p Ͻ 0.05.

RESULTS
Multiple Effects of pH i on CFTR Activity-To investigate whether pH i regulates directly the CFTR Cl Ϫ channel, we studied CFTR Cl Ϫ currents in membrane patches excised from C127 cells expressing wild-type human CFTR. After CFTR activation by PKA-dependent phosphorylation at pH i 7.3, we varied pH i over the range pH i 5.8 to 8.8 and recorded CFTR Cl Ϫ currents while keeping the MgATP and PKA concentrations constant (for further information, see supplemental "Experimental Procedures" and Table 1. Fig. 1A demonstrates that changing pH i altered the magnitude of CFTR Cl Ϫ current in a readily reversible manner. At acidic pH i , CFTR Cl Ϫ currents were enhanced robustly at pH i 6.8 and 6.3, but decreased at pH i 5.8 (Fig. 1B). By contrast, at alkaline pH i , CFTR Cl Ϫ currents decreased steadily (Fig. 1B). To examine the pH i sensitivity of CFTR, we explored the relationship between CFTR Cl Ϫ current and the proton concentration ([H ϩ ]) of pH i solutions. Over the pH i range pH i 6.3-8.3, the data were best fit by a hyperbolic function with a half-maximal stimulatory concentration (K s ) at pH i 7.06 (Fig. 1C). We interpret these data to suggest that CFTR activity is regulated by pH i and that this regulation involves the protonation of CFTR.
In principle, pH i might alter CFTR Cl Ϫ current in one or more of three ways: (i) by regulating the number of active channels (N), (ii) by modulating current flow through open channels, and (iii) by controlling channel gating and, hence, P o . To understand how pH i regulates CFTR, we investigated the effects of pH i on the single-channel activity of CFTR. Fig. 2 shows representative recordings of a single wild-type CFTR Cl Ϫ channel at different pH i . Visual inspection of these records suggests that varying pH i had striking effects on CFTR channel gating, little effect on single-channel current amplitude (i), and none on N. However, when we measured i, we found that i was unaltered at acidic pH i but diminished progressively at alkaline pH i (Fig. 3A), suggesting that OH Ϫ ions inhibit the CFTR Cl Ϫ channel.
The pattern of gating of wild-type CFTR at pH i 7.3 is characterized by bursts of channel openings interrupted by brief flickery closures and separated by longer closures between bursts (e.g. Ref. 24 and Fig. 2, third trace from the top). To quantify the pH i dependence of channel gating, we measured P o and performed an analysis of bursts. Over the pH i range pH i 6.3-8.3, P o decreased steadily as pH i increased (Fig. 3B). However, between pH i 6.3 and 5.8, P o fell sharply, whereas between pH i 8.3 and 8.8, it rose, returning to a value similar to that of pH i 7.8 ( Fig.  3B). At acidic pH i , MBD increased dramatically, whereas at alkaline pH i , MBD decreased (Fig. 3C). By contrast, Fig. 3D demonstrates that the relationship between pH i and IBI is complex. At acidic pH i , IBI decreased at pH i 6.3, but increased markedly at pH i 5.8, whereas at alkaline pH i , IBI increased at pH i 8.3, but was unchanged at pH i 8.8 (Fig. 3D). Thus, pH i has intricate effects on CFTR channel gating. However, the effects of pH i on the single-channel behavior of CFTR account for the pH i dependence of CFTR Cl Ϫ currents (supplemental Table 3). Based on the control experiments described under supplemental "Results", we suggest that pH i regulates directly the CFTR Cl Ϫ channel.
OH Ϫ Ions Occlude the CFTR Pore- Fig. 3A demonstrates that the i of CFTR diminishes progressively with alkalinization of the intracellular solution, suggesting that OH Ϫ ions might inhibit the CFTR Cl Ϫ channel by acting as an openchannel blocker (25). To test this hypothesis, we investigated the effects of membrane voltage on the i of wild-type CFTR at alkaline pH i . We reasoned that negative voltages would drive OH Ϫ ions from the intracellular solution into the CFTR Cl Ϫ channel where they might bind, occlude the channel pore, and block Cl Ϫ permeation. Conversely, positive voltages would drive Cl Ϫ ions into the CFTR Cl Ϫ channel from the extracellular solution, flushing blocking OH Ϫ ions from the channel pore and relieving inhibition. Fig. 4A shows representative recordings of a single CFTR Cl Ϫ channel at Ϫ80 and ϩ80 mV, Fig. 4B shows the single-channel current-voltage (I-V) relationship of wild-type CFTR at pH i 7.3 and 8.3, and Fig. 4C shows the voltage dependence of the fraction of i inhibited at pH i 8.3. Fig. 4, A-C, demonstrates that at negative voltages there was a small, but significant, decrease in the i of wild-type CFTR at pH i 8.3 compared with that at pH i 7.3. By contrast, at positive voltages the inhibition of i was completely relieved (Fig. 4, A-C) (similar results were observed at pH i 8.8 over the voltage range Ϫ80 to ϩ80 mV (data not shown)). Block of i at negative voltages, but relief at positive voltages suggests that OH Ϫ ions occlude the intracellular vestibule of the CFTR pore.
To determine the potency with which OH Ϫ ions block the CFTR pore, we calculated the apparent voltage-dependent dissociation constant (K d ) for OH Ϫ ion inhibition of wild-type  CFTR. Fig. 4D demonstrates that apparent K d (V) values for CFTR inhibition by OH Ϫ ions are weakly voltage-dependent. The data also indicate that the potency of CFTR inhibition by OH Ϫ ions (apparent K d (0 mV) ϭ 54 Ϯ 6 M; n ϭ 9) approaches that of the best-studied CFTR blocker, the sulfonylurea glibenclamide (K d (0 mV) ϭ 37 M) (20).
To estimate the location of the binding site for OH Ϫ ions within the CFTR pore, we calculated the electrical distance across the membrane sensed by OH Ϫ ions using the Woodhull equation (26), as described under supplemental "Experimental Procedures." Using the data in Fig. 4D and assuming a single binding site for OH Ϫ ions, zЈ ϭ 0.23 Ϯ 0.03 (n ϭ 9), measured from the inside of the membrane over the voltage range Ϫ120 to Ϫ40 mV. These data suggest that the penetration of OH Ϫ ions into the transmembrane electric field from the intracellular side is 23%, similar to that of SCN Ϫ ions (20%) (27). Thus, at alkaline pH i , OH Ϫ ions cause open-channel block of the CFTR Cl Ϫ channel by occupying a superficial site within the intracellular vestibule of the CFTR pore.
In contrast to the effects of pH i on i (Figs. 3A and 4, B and C), the pH i dependence of P o was voltage-independent ( Fig. 4E). At all voltages tested, the P o of wild-type CFTR was diminished at pH i 8.3, but augmented at pH i 6.3. The data also suggest that inhibition of CFTR Cl Ϫ current at alkaline pH i results predominantly from a diminution of P o rather than a decrease in i (supplemental Table 3).
pH i Modulates the Function of the NBDs and RD-CFTR channel gating is controlled by the NBDs and RD (2, 3). Therefore, we reasoned that pH i might modulate CFTR activity by altering the function of these domains. To test this hypothesis, we adopted two strategies; first, we employed the CFTR construct ⌬R-S660A that deletes a large part of the RD and is not regulated by PKA-dependent phosphorylation (13) to explore separately how pH i influences the function of the RD and NBDs. Second, we investigated the effects of pH i on the ATP dependence of CFTR channel gating. For these experiments, we studied CFTR potentiation at pH i 6.3 and CFTR inhibition at pH i 8.3. Fig. 5A shows representative recordings of a single ⌬R-S660A-CFTR Cl Ϫ channel at different pH i , whereas Fig. 5, B-D quantifies the effects of pH i on channel gating. Consistent with previous studies (13), at pH i 7.3 the P o of ⌬R-S660A-CFTR was attenuated markedly compared with that of wild-type CFTR  because its MBD was reduced 0.4-fold, while its IBI was prolonged 9-fold (Fig. 5, B-D). Acidifying to pH i 6.3 potentiated the P o of wild-type CFTR 0.4-fold, but that of ⌬R-S660A-CFTR 3.3-fold (Fig. 5B). Fig. 5, C and D, demonstrates that the marked potentiation of ⌬R-S660A-CFTR P o at pH i 6.3 was primarily caused by a 0.9-fold decrease in IBI, but it was also enhanced by a 1.5-fold increase of MBD. Alkalinizing to pH i 8.3 decreased the P o of wild-type CFTR 0.3-fold, but was without effect on the P o of ⌬R-S660A-CFTR (Fig. 5B). An explanation for this striking difference between wild-type and ⌬R-S660A-CFTR is shown in Fig. 5D. Between pH i 7.3 and 8.3, the IBI of wild-type CFTR increased 0.3-fold, whereas that of ⌬R-S660A-CFTR decreased 0.5-fold (Fig. 5D). Fig. 5D also reveals that the relationship between IBI and pH i is linear for wild-type CFTR, but bell-shaped for ⌬R-S660A-CFTR over the pH i range 6.3-8.3. We interpret the failure of pH i 8.3 to attenuate ⌬R-S660A-CFTR channel gating to suggest that alkaline pH i might inhibit CFTR channel gating, at least in part, by modulating RD function. However, the robust potentiation of ⌬R-S660A-CFTR channel gating at pH i 6.3 argues that acidic pH i likely enhances CFTR channel gating by acting on sites distinct from the RD.
To investigate whether pH i influences NBD function, we studied the ATP dependence of CFTR channel gating. Fig. 6A shows the relationship between [MgATP] and P o for wild-type CFTR at pH i 6.3, 7.3, and 8.3. At each pH i tested, as the [MgATP] increased, P o values enlarged. However, altering pH i influenced both the MgATP sensitivity and channel activity of wild-type CFTR (Fig. 6A). These differences are best illustrated by considering values of K D (the [MgATP] required for halfmaximal P o , which describes the apparent affinity of CFTR for MgATP) and P o max (the maximum P o ) determined from Michaelis-Menten fits to the mean data (pH i 6.  Fig. 6A). Two conclusions are apparent from these data. First, acidifying pH i enhances markedly the apparent MgATP affinity of CFTR, whereas alkalinizing pH i has little effect. Second, acidifying pH i augments channel activity, whereas alkalinizing pH i inhibits CFTR.
The Role of Mg 2ϩ Ions in Determining the pH i Sensitivity of the NBDs-Mg 2ϩ ions are a pre-requisite for the hydrolysis of ATP that drives CFTR channel closure (21,22). To learn whether pH i modulates CFTR channel gating by affecting the interaction of Mg 2ϩ ions with the NBDs, we examined the single-channel activity of CFTR in the absence of Mg 2ϩ ions at different pH i (Figs. 6B and supplemental Fig. 3A). At pH i 7.3 removal of Mg 2ϩ ions attenuated markedly the P o of CFTR (Fig.  6B). Nevertheless, acidification to pH i 6.3 enhanced strongly P o , whereas alkalinization to pH i 8.3 was without effect (Fig. 6B). These data suggest that the presence of Mg 2ϩ ions is critical for the inhibition of CFTR channel gating at alkaline pH i . By contrast, the potentiation of CFTR channel gating at pH i 6.3 is both dependent on, and independent of, Mg 2ϩ ions.
To understand better the role of Mg 2ϩ ions in determining the pH i sensitivity of the NBDs, we studied Asp-572 and Asp-1370, the Walker B aspartates that coordinate Mg 2ϩ ions in ATP binding sites 1 and 2 of the ATP-driven NBD dimerization model of CFTR channel gating (3,28). Previous studies have demonstrated that the mutations D572N-and D1370N-CFTR abolish Mg 2ϩ binding to the NBDs (21,22). Figs. 6C and 8 and supplemental Fig. 3, B and C, demonstrate that the gating behavior of D572N-and D1370N-CFTR Cl Ϫ channels at different pH i diverges from that of wild-type CFTR in several important respects. First, at pH i 7.3, the P o of D572N-CFTR was the same as wild-type CFTR, whereas that of D1370N-CFTR was reduced (Fig. 6C). Second, at pH i 6.3, the P o of D572N-CFTR was potentiated markedly because IBI was decreased 0.7-fold and MBD was increased 1.4-fold (Figs. 6C and 8 and supplemental Fig. 3B). By contrast, for D1370N-CFTR at pH i 6.3, gating behavior and, hence, P o were unchanged ( Fig. 6C and supplemental Fig. 3C). Third, in striking contrast to wild-type CFTR, at pH i 8.3 D572N-CFTR channel gating was enhanced because MBD was increased 0.6-fold and IBI decreased 0.3fold, whereas that of D1370N-CFTR was unaltered (Figs. 2 and 8 and supplemental Fig. 3, B and C). As a result, at pH i 8.3, the P o of wild-type CFTR decreased, that of D1370N-CFTR was unchanged, whereas that of D572N-CFTR increased (Fig. 6C).
Several conclusions can be drawn from the data in Fig. 6C. First, the potentiation of D572N-CFTR channel activity at pH i 8.3 suggests that the binding of Mg 2ϩ ions at site 1 is essential for the inhibition of CFTR channel gating at alkaline pH i . Second the lack of effect of pH i on D1370N argues that the pH i sensitivity of CFTR channel gating is dependent on Mg 2ϩ binding to site 2. This suggests that pH i might exert its effects on CFTR channel gating by altering the ATPase activity of site 2.  (29 -32) and functional (16,21) studies suggest that the phosphate groups of ATP are key determinants of ATP binding to sites 1 and 2. The Walker A lysine residues Lys-464 (site 1) and Lys-1250 (site 2), which bind the ␥-phosphate of ATP, are highly conserved and critical for ATP binding and hydrolysis (16,21,29). Moreover, the LSGGQ motifs contain the highly conserved glycine residues Gly-551 (site 2) and Gly-1349 (site 1) that clamp the phosphate tail of ATP during ATP-driven NBD dimerization (29,30,32). Of note, the CF mutations G551D and G1349D perturb severely CFTR channel gating (14,33). Using CFTR constructs bearing site-directed mutations, we examined the roles of the residues Lys-464, Lys-1250, Gly-551, and Gly-1349 in determining the pH i sensitivity of CFTR channel gating.

ATP Binding Site 2 Is a Crucial Determinant of the pH i Sensitivity of CFTR Channel Gating-Structural
Consistent with previous studies (14,33), G1349D-CFTR and especially G551D-CFTR attenuated strongly CFTR channel gating at pH i 7.3, with brief, poorly resolved channel openings separated by very long-lasting closures (Fig. 7, A and C). Fig. 7, A-D, demonstrates that the gating behavior and, hence, NP o of G551D-and G1349D-CFTR were unaffected by pH i . Because tight dimerization of the NBDs is a prerequisite for channel opening (28) and because G551D and G1349D perturb severely channel gating (14,33), we speculate that the effects of pH i on CFTR channel gating are dependent on the formation of an NBD1:NBD2 dimer. Fig. 7, E and G, show the hallmark effects of mutating the Walker A lysines in NBD1 and NBD2 on CFTR channel gating (16,21). At pH i 7.3, bursts of K464A-CFTR channel openings were separated by prolonged channel closures, whereas dramatically prolonged bursts of K1250M-CFTR channel openings were separated by very long-lived channel closures (Figs. 7,  E and G, and 8). Although P o values of K464A-CFTR were less than those of wild-type CFTR at all pH i values tested, the effects of acidic and alkaline pH i on K464A-and wild-type CFTR were similar (Fig. 7, E and F). This suggests that K464A-CFTR does not change the pH i sensitivity of CFTR. By contrast, the pH i sensitivity of K1250M-CFTR differed strikingly from that of wild-type CFTR. At pH i 6.3, the P o of K1250M-CFTR was reduced because MBD was decreased 0.7-fold, whereas IBI was unchanged (Figs. 7, G and H, and 8). At pH i 8.3, the P o of K1250M-CFTR was increased, albeit slightly, as a result of a small increase in MBD and no change in IBI (Figs. 7H and 8). Fig. 7H also reveals that the P o of K1250M-CFTR was similar to that of wild-type CFTR at pH i 8.3, but greatly diminished at pH i 7.3 and 6.3. Thus, the pH i sensitivity of K1250M-CFTR is the converse of that of wild-type CFTR.
Taken together, our data argue that ATP binding site 2 is a crucial determinant of the pH i sensitivity of the CFTR Cl Ϫ channel. They also suggest that MgATP binding at site 1 might be important for mediating the inhibitory effects of alkaline pH i and cooperating with site 2 to drive channel gating.

DISCUSSION
In this study we investigated whether pH i regulates directly CFTR activity and the responsible mechanisms. Our data demonstrate that acidic pH i stimulates CFTR Cl Ϫ currents by potentiating channel gating, whereas alkaline pH i attenuates CFTR Cl Ϫ currents predominantly by inhibiting channel gating, but also by obstructing Cl Ϫ flow through the CFTR pore. Using CFTR constructs, we demonstrate that ATP-binding site 2 is a primary determinant of the pH i sensitivity of CFTR. Our data suggest that the mechanism by which pH i regulates CFTR is distinct from other ion channels. Instead of protons modifying channel gating by interacting with residues within (e.g. ClC-0 (34)) or close to the intracellular vestibule (e.g. Kir1.1 (35)) of the channel pore, in CFTR protons modulate channel gating by altering the function of ATP-binding site 2 in the NBDs.
Molecular Mechanisms of CFTR Regulation by pH i -Our data demonstrate that pH i has multiple effects on the CFTR Cl Ϫ channel. They also reveal that pH i modulates CFTR function by altering the activity of each of the domains from which CFTR is assembled. In Fig. 9 we summarize the molecular mechanisms by which pH i regulates current flow through OH Ϫ ions inhibit CFTR function in several ways. First, OH Ϫ ions are open-channel blockers of the CFTR pore that disrupt Cl Ϫ flow by interacting with a superficial site within the intracellular vestibule of the CFTR pore. Second, studies of the CFTR construct ⌬R-S660A suggest that OH Ϫ ions might inhibit CFTR channel gating, at least in part, by modulating RD function. Third, OH Ϫ ions inhibit CFTR function by regulating the control of channel gating by ATP-driven NBD dimerization (3,21). The decrease in channel activity at alkaline pH i argues that OH Ϫ ions might destabilize ATP binding at sites 1 and 2, promoting dissociation of the NBD dimer and channel closure. However, our data also suggest that the binding of Mg 2ϩ ions at site 1 is essential for CFTR inhibition by OH Ϫ ions.
In contrast to the effects of OH Ϫ ions, H ϩ ions mediate their effects on CFTR predominantly by modulating the control of channel gating at the NBDs. Our studies argue that ATP-binding site 2 is a crucial determinant of the pH i sensitivity of CFTR. The data suggest that H ϩ ions enhance the affinity with which MgATP binds at site 2 and stabilize ATP binding at this site with the result that the frequency of channel openings is increased and their duration prolonged. Thus, pH i regulates directly CFTR function. Importantly, this regulation occurs within the physiological pH i range (36), consistent with previous work demonstrating a significant role for CFTR in the regulation of pH i (e.g. Ref. 5, 6, 37, and 38).
Open-channel Block of the CFTR Pore by OH Ϫ Ions-At negative voltages OH Ϫ ions caused a small reduction of the i of CFTR. Given the low concentration of OH Ϫ ions in our test solutions (see "Experimental Procedures"), the potency of OH Ϫ ion inhibition of CFTR is surprisingly strong, exceeding that of many open-channel blockers of CFTR (e.g. diphenylamine-2carboxylate and flufenamic acid (39)). Interestingly, the location of the OH Ϫ ion-binding site within the transmembrane electric field (23%) is similar to that of SCN Ϫ ions (20%) (27). These data argue that OH Ϫ and SCN Ϫ ions interact with a superficial site(s) within the intracellular vestibule of the CFTR pore (Fig. 9). The shallow penetrance of OH Ϫ and SCN Ϫ ions into the CFTR pore contrasts with that of large organic anions (e.g. glibenclamide (20)), which penetrate deep into the intracellular vestibule to interact with a site located halfway across the electric field of the membrane from the inside. Site-directed mutagenesis of Lys-95, located toward the extracellular end of the first transmembrane segment, reveals that the positive charge at Lys-95 is essential for channel block by glibenclamide and four other open-channel blockers of CFTR with unrelated chemical structures (40). However, studies of the kinetics of channel block (41)(42)(43) demonstrate that the interaction of open-channel blockers with CFTR is complex and likely to involve multiple binding sites within the intracellular vestibule of the CFTR pore. One of these binding sites might overlap with that of OH Ϫ ions.
Intricate Effects of pH i on CFTR Channel Gating-Our data reveal that pH i has complex effects on CFTR channel gating. Over the pH i range 6.3 to 8.3, the P o of wild-type CFTR decreases steadily because of reciprocal changes in IBI and MBD. This suggests that between pH i 6.3 and 8.3, the effects of pH i on MBD and IBI are coordinated. By contrast, our data suggest that in strongly acidic and alkaline intracellular solutions, MBD values are stable, whereas those of IBI appear unstable.
At pH i 5.8 the IBI of wild-type CFTR was prolonged strikingly, indicating that channel opening is very difficult at strongly acidic pH i . At pH i 5.8, MgATP 2Ϫ is partially protonated at the ␥-phosphate to become MgHATP Ϫ (44). This suggests that protonated ATP at pH i 5.8 might interfere with the formation of H-bonds between CFTR and the ␥-phosphate of ATP, attenuating ATP binding and impeding channel opening. However, we speculate that protonation of phosphate groups in other parts of the CFTR gating pathway or on the RD at strongly acidic pH i might also contribute to this result.
Interestingly, at pH i 8.8 the IBI of wild-type CFTR was attenuated with the result that the P o of wild-type CFTR at pH i 8.8 enlarged to become equivalent to that at pH i 7.8. Because negative surface charge increases with alkalinization of the intracellular solution, the diminution of IBI at pH i 8.8 might represent a change in RD conformation or the interaction of the RD with other domains of CFTR (Fig. 9). For example, manipulation of the surface charge on the RD might mimic the effects of RD phosphorylation and disrupt interactions between the RD and NBDs leading to channel activation (13,45,46). Alternatively, increased negative surface charge at alkaline pH i might alter the balance between stimulatory and inhibitory RD phosphorylation sites to favor channel activation (47).
In contrast to the effects of strong alkaline pH i on wild-type CFTR, our studies of the ⌬R-S660A-CFTR Cl Ϫ channel at pH i 8.3 argue that alkaline pH i inhibits CFTR channel gating, at least in part, by modulating RD function. The reason why ⌬R-S660A-CFTR is not inhibited at pH i 8.3 is that its prolonged IBI is attenuated, not extended, at this pH i value. Further studies are required to understand the underlying mechanisms. However, it is interesting to note that for both ⌬R-S660A-CFTR and the ATP-binding site 1 mutant D572N-CFTR, the relationship between pH i and IBI is bell-shaped, not linear, between pH i 6.3 and 8.3 (Figs. 5D and 8B). We interpret these data to suggest that the pH i dependence of IBI reflects complex interactions between the RD and NBDs.
Use of the ATP-driven NBD Dimerization Model of CFTR Channel Gating to Explain the Effects of pH i -Like other ABC transporters, the two NBDs of CFTR likely form a head-to-tail dimer that sandwiches ATP molecules within two binding sites (termed site 1 and site 2) located at the NBD1:NBD2 interface (3,21). In CFTR, site 1 has no ATPase activity, but site 2 cyclically hydrolyzes ATP to drive channel gating (3,21). In this asymmetric gating model, ATP-induced dimerization at site 1 likely occurs before that at site 2, which in turn precedes CFTR opening (48).
Structural studies of ABC transporters (29 -32) demonstrate that within the ATP-binding sites the cis NBD anchors the phosphate tail of an ATP molecule by a network of hydrogen bonds (H-bonds) and hydrogen-mediated salt bridges from the Walker A motif. By contrast, in the trans NBD, only H-bonds from the LSGGQ motif clamp the ␥-phosphate of ATP. These data suggest that perturbation of the LSGGQ motif might have a greater impact on NBD dimerization than disruption of the Walker A motif. Accordingly, the Walker A lysine mutant K1250A-CFTR is an ATPdependent channel with moderate P o (16), whereas the LSGGQ motif mutant G551D-CFTR is an ATP-independent channel with extremely low P o (14,33). Thus, the loss of pH i sensitivity by G551D-and G1349D-CFTR suggests that correct alignment of the NBD1:NBD2 dimer is required for the potentiation of CFTR channel gating by acidic pH i and inhibition by alkaline pH i .
Based on several lines of evidence, we suggest that ATP-binding site 2 is a crucial determinant of the pH i sensitivity of CFTR. First, acidifying pH i enhances markedly the apparent MgATP affinity of CFTR, whereas alkalinizing pH i has little effect. Vergani et al. (49) (24) demonstrated that the opening rate of CFTR, not its closing rate, is ATP-dependent. These considerations suggest that the pH i sensitivity of the apparent MgATP affinity of CFTR likely reflect the effects of H ϩ ions on the interaction of MgATP with site 2.
Second, our studies of CFTR channel gating in the absence of Mg 2ϩ ions demonstrate that Mg 2ϩ ions are critical for channel inhibition at alkaline pH i and modulate channel potentiation at acidic pH i . Because MgATP binding at both ATP binding sites is a prerequisite for channel opening (50), these data suggest that pH i has effects on both ATP binding sites. However, Mg 2ϩ ions are also a prerequisite for the hydrolysis of ATP that closes the channel (50). This argues that the potentiation of channel gating at pH i 6.3 is evidence for H ϩ ions mediating their effects, at least in part, via site 2.
Third, H ϩ ions potentiate the gating behavior of CFTR constructs bearing site-directed mutations in ATP-binding site 1 (K464A-and D572N-CFTR). By contrast, H ϩ ions are either without effect (D1370N-CFTR) or inhibit (K1250M-CFTR) the gating behavior of site-directed mutations in ATP-binding site 2. As discussed above, the failure of H ϩ ions to potentiate G551D-(site 2) and G1349D-CFTR (site 1) is likely a consequence of the profound disruption of NBD dimerization and, hence, channel gating by these constructs. Taken together, the simplest interpretation of our data is that ATP-binding site 2 is  DECEMBER 18, 2009 • VOLUME 284 • NUMBER 51 a crucial determinant of the pH i sensitivity of the CFTR Cl Ϫ channel (Fig. 9). This suggests that H-bond formation at site 2 enhances channel activity by accelerating NBD dimerization and stabilizing the ATP bound conformation of the NBD1: NBD2 dimer. Consistent with this idea, acidic pH enhanced greatly the formation of an engineered TAP1 homodimer (11).

CFTR Senses Directly pH i
Mechanism of ATP Hydrolysis at Site 2-A common feature of the ATP binding sites of the ABC transporters CFTR and TAP is their asymmetric ATPase activity (11,29). In an engineered human TAP1 homodimer, the ATP-binding site with ATPase activity contains a catalytic dyad composed of a histidine residue in the H-loop and a glutamic acid residue distal to the Walker B motif (11,31). However, instead of hydrolyzing ATP by general base catalysis, this dyad employs substrate-assisted catalysis (SAC) for this reaction (11,31). In SAC, the interaction of the substrate, ATP, with catalytic water via its ␥-phosphate is stabilized by an H-bond between the dyad histidine and the ␥-phosphate of ATP (31). A key advantage of SAC over general base catalysis is that it prevents ATP hydrolysis occurring before NBD dimerization (31). In this way socalled "futile ATP hydrolysis" by ABC transporters is avoided.
It is intriguing to consider whether CFTR hydrolyzes ATP at site 2 by SAC. In support of this idea, site 2 but not site 1 contains a catalytic dyad like that found in human TAP1 (11,29). Although the duration of bursts of channel openings are not determined solely by the ATPase activity of the NBDs (51), the effects of pH i on the MBD of CFTR (Fig. 3C) suggest that alkaline pH i might enhance ATP hydrolysis at site 2, whereas acidic pH i has the converse effect. This raises the interesting possibility that the pH i sensitivity of the ATPase activity at site 2 of CFTR might be similar to the ATP-binding site of an engineered TAP1 homodimer that uses SAC to hydrolyze ATP (11). Finally, in SAC, OH Ϫ ion attack of the ␥-phosphate is the ratelimiting step for ATP hydrolysis (11,31). Taken together, the data suggest that SAC might provide an efficient way to couple NBD dimerization and ATP hydrolysis in the CFTR Cl Ϫ channel. Future studies should explore this possibility.
The pH i Sensitivity of CFTR Reveals Cross-talk between ATP Binding Sites 1 and 2-An interesting aspect of our data is the pH i sensitivity of K1250M-CFTR, which is the reverse of that of wild-type CFTR. This observation suggests that there might be hidden minor pH i effects, masked by the major effects of pH i on wild-type CFTR. If K1250M-CFTR severely disrupts the function of site 2 (16), these hidden minor pH i effects might originate from site 1. Our data also reveal that D572N-CFTR has a very unusual response to pH i : exaggerated potentiation of channel gating at acidic pH i and potentiation, not inhibition, at alkaline pH i . In ABC transporters, the Walker B motif is located adjacent to the highly conserved D-loop motif (30,31). Structural studies of ABC transporters (30,31) suggest that in CFTR the D-loop of NBD1 (site 1) might (i) interact with the Walker A motif of NBD2 (site 2) and (ii) sense the functional state of site 1 via the H-loop motif of NBD1. In support of this idea, the effects of acidic and alkaline pH i on D572N-CFTR are reminiscent of the enhanced activity of P574H-CFTR, a CF mutant affecting a residue in the Walker B-D-loop region of NBD1 (52). Thus, the D-loop might mediate cross-talk between the ATP binding sites of CFTR and influence the pH i sensitivity of CFTR.
Identification of pH i -sensitive Amino Acids in CFTR-The changes in CFTR channel gating at pH i 6.3 and 8.3 suggest that histidine (pK a ϳ 6.1) and cysteine (pK a ϳ 8.3) might serve as pH i -titratable amino acid residues. Examination of the NBD1: NBD2 dimer identifies a number of histidine and cysteine residues located close to the ATP binding sites. Site 1 contains two histidine residues in its LSGGQ motif (His-1348 and His-1350) and a cysteine residue Cys-491 (1,29). Because Cys-491 is located spatially close to the Walker B aspartate, Asp-572, of site 1 (29), it might modulate the role of Asp-572 in CFTR channel gating by sensing alkaline pH i . Interestingly, the Hloop of site 2 contains histidine (His-1402) and cysteine (Cys-1400) residues that are lacking in the H-loop of site 1 (29). This raises the interesting possibility that Cys-1400 and His-1402 might act as a pH i sensor for site 2. However, the potential roles of other histidine and cysteine residues in the control of CFTR channel gating should not be overlooked. For example, the sulfhydryl-modifying reagent N-ethylmaleimide rapidly and irreversibly potentiates CFTR channel gating by covalently modifying Cys-832 in the RD (53). Moreover, analyzes of the pH i dependence of other ion channels (e.g. Ref. 35) suggest that other amino acid residues might act as pH i sensors for CFTR channel gating.
Physiological Significance-Reddy et al. (10) demonstrated that pH i regulates indirectly the CFTR Cl Ϫ conductance of sweat duct epithelia by altering the activity of protein kinases and phosphatases that control RD phosphorylation. By contrast, our data reveal that pH i regulates directly CFTR channel gating. Taken together, these data suggest that CFTR activity is regulated by pH i via two different control mechanisms (Fig. 10). First, at the cell membrane localized pH i fluctuations are detected directly by CFTR, leading to prompt changes in its behavior, coordinated with alterations in the activity of NHE3 and SLC26 transporters. Second, when there are global changes in pH i , the effects of pH i on the enzymes controlling the phosphorylation status of CFTR regulate CFTR indirectly. Fig. 10 also suggests how the direct and indirect mechanisms of CFTR regulation by pH i might be integrated in the control of HCO 3 Ϫ secretion by epithelial cells. HCO 3 Ϫ secretion is driven by the coordinated activity of CFTR and electrogenic SLC26 transporters at the apical membrane (9). Cl Ϫ that exits the cell through CFTR is harnessed by SLC26 transporters to drive HCO 3 Ϫ secretion. This efflux of HCO 3 Ϫ acidifies pH i locally in the vicinity of the cell membrane (7), enhancing CFTR channel gating and, hence, Cl Ϫ secretion (Fig. 10). Because CFTR activity is reciprocally coupled to that of SLC26 transporters (9), our data suggest that the two transporters form a positive feedback loop to promote HCO 3 Ϫ secretion (Fig. 10). To break this positive feedback loop, a further CFTR control mechanism is required. This is provided by pH i regulating indirectly RD phosphorylation by controlling enzymatic activity through a negative feedback loop (Fig. 10).
The two-loop model of CFTR regulation by pH i (Fig. 10) provides a mechanism to control the magnitude and duration of epithelial HCO 3 Ϫ secretion. This model might also be exploited to understand better how CF causes aberrant HCO 3 Ϫ secretion (37,38), abnormal pH i regulation (5,6), and defects in acidic pH i -induced apoptosis (54) and bacterial killing (55). Thus, by investigating the effects of pH i on CFTR, we understand better its mechanism of action, role in fluid and electrolyte transport, and malfunction in CF.