The Inhibition Mechanism of Non-phosphorylated Ser768 in the Regulatory Domain of Cystic Fibrosis Transmembrane Conductance Regulator*

The cystic fibrosis transmembrane conductance regulator (CFTR) is a member of the ATP-binding cassette transporters but serves as a chloride channel dysfunctional in cystic fibrosis. The activity of CFTR is tightly controlled not only by ATP-driven dimerization of its nucleotide-binding domains but also by phosphorylation of a unique regulatory (R) domain by protein kinase A (PKA). The R domain has multiple excitatory phosphorylation sites, but Ser737 and Ser768 are inhibitory. The underlying mechanism is unclear. Here, sulfhydryl-specific cross-linking strategy was employed to demonstrate that Ser768 or Ser737 could interact with outwardly facing hydrophilic residues of cytoplasmic loop 3 regulating channel gating. Furthermore, mutation of these residues to alanines promoted channel opening by curcumin in an ATP-dependent manner even in the absence of PKA. However, mutation of Ser768 and His950 with different hydrogen bond donors or acceptors clearly changed ATP- and PKA-dependent channel activity no matter whether curcumin was present or not. More importantly, significant activation of a double mutant H950R/S768R needed only ATP. Finally, in vitro and in vivo single channel recordings suggest that Ser768 may form a putative hydrogen bond with His950 of cytoplasmic loop 3 to prevent channel opening by ATP in the non-phosphorylated state and by subsequent cAMP-dependent phosphorylation. These observations support an electron cryomicroscopy-based structural model on which the R domain is closed to cytoplasmic loops regulating channel gating.

The cystic fibrosis transmembrane conductance regulator (CFTR) 2 chloride channel is widely distributed in the human organs, including the heart, and mediates the electric response to ATP and protein kinase A or C. As shown in Fig. 1, this protein has two membrane-spanning domains (MSD1 and MSD2), two intracellular nucleotide-binding domains (NBD1 and NBD2), and a unique regulatory (R) domain, although it belongs to the human C subfamily of ATP-binding cassette transporters (1,2). Each MSD consists of six transmembrane helical segments probably extended to four cytoplasmic loops (3). Although recent studies have strongly suggested structural similarities between CFTR and bacterial transporters Sav1866 and MsbA (3)(4)(5), three-dimensional structural information about the whole protein is still unavailable except for the crystal structure of the isolated NBD1 (6). Furthermore, the exact location and relative orientation of the R domain in the whole protein are also unclear because this domain lacks a stably folded globular structure and thus is disordered (7,8).
Ion transport of CFTR is triggered by not only ATP binding and hydrolysis at the interface of a NBD1-NBD2 dimer but also phosphorylation by protein kinase A (PKA) (9). Structures of bacterial NBD homodimers indicate two ATP-binding sites at the NBD1-NBD2 interface, and each site is composed of residues from both NBDs (10). However, most PKA phosphorylation sites are mainly found in the R domain (6,11).
CFTR activity is tightly controlled by interdomain interactions. Several thiol-specific cross-linking studies, based on the crystal structures of Sav1866 and MsbA, have shown that the NBD1-NBD2 dimerization drives channel opening (12). However, chemical cross-linking of NBDs to cytoplasmic loops (CLs) inhibits channel activity ( Fig. 1) (5,13,14). Recent structural studies of CFTR and other ATP-binding cassette transporters suggested rearrangements of CLs that couple dimerization of the NBDs to a change in the MSDs from an inward to an outward facing conformation (4,15,16). Our recent study also demonstrated that a K190C/S mutation from CL1 enhances ATP-independent channel opening induced by a K978C/P/S mutation from CL3 (17). Thus, CLs may function as a key regulatory switch to modulate normal CFTR activity.
The R domain (amino acids 686 -850) has 14 PKA phosphorylation sites exerting multiple effects on channel activity (18,19). Phosphorylation introduces negative charges to the R domain and thus reduces the ␣-helical content (8). It has been reported that PKA does regulate an NBD1-NBD2 interaction (12,13) and that PKA can regulate ATP-independent gating in CFTR constructs with G551D (20) and constructs lacking NBD2 (⌬1198) (17,21,22). Finally, PKA can also regulate the N-ethylmaleimide effect, which potentiates channel activity by modifying Cys 832 (23). Therefore, phosphorylation of the R domain by PKA is a key physiological regulator. Although most phosphorylation sites, including Ser 700 , Ser 795 , Ser 813 , and Ser 660 , stimulate channel activity, Ser 737 and Ser 768 are inhibitory sites (18). Substitutions of these two residues with alanines increase channel activity (18,19,24,25). In addition, removal of residues 760 -783 or 817-838 (NEG2) or much of the R domain (⌬708 -835/S660A) from CFTR eliminates PKA dependence of channel activity (26 -28). Thus, some residues of the R domain may interact with cytosolic domains of CFTR to control channel gating. Naren et al. (29) has indicated that a direct interaction between the N-terminal cytoplasmic tail and the R domain regulates PKA-dependent channel gating. NMR experiments suggested an interaction between the isolated R domain and the NBD1 (11). However, direct biochemical and electrophysiological evidence in a native whole protein was still missing. Moreover, the use of isolated CFTR fragments may lead to an interdomain interaction that might not be present in the native channel. In fact, recent electron microscopy studies demonstrated that the R domain is mainly closed to the cytoplasmic loops (16). However, a low resolution is not enough to illuminate the exact regulation mecha-nisms of the R domain. Indication of the exact interactions of the R domain with other domains of CFTR is still necessary.
My previous study suggested that phosphorylated Ser 768 may enhance the endogenous inhibitory Fe 3ϩ binding at the R-CL3 interface formed by His 950 , His 954 , Cys 832 , Asp 836 , and His 775 (30). Thus, it is very likely that Ser 768 may govern channel gating by interacting with CL3. In order to test this hypothesis, outwardly facing residues Lys 946 , His 950 , Lys 951 , His 954 , Ser 955 , and Gln 958 from CL3 and two well known inhibitory sites, Ser 768 and Ser 737 , from the R domain ( Fig. 2A) were selected to carry out several independent experiments. First, systematic cysteine scanning mutagenesis was used to identify which residue of CL3 forms an inhibitory pair with Ser 768 or Ser 737 of the R domain. Second, a missense alanine mutation of the inhibitory residue was employed to determine if ATP and PKA dependence of channel activity is altered. Finally, the nature of the interaction between Ser 768 and inhibitory residues of CL3 was further investigated. The resulting data first indicated that Ser 768 could inhibit channel opening by ATP in the nonphosphorylated CFTR. A putative H-bond between the imidazole group of His 950 and the hydroxyl group of Ser 768 may be involved. Regulation of channel gating by the putative H-bond was also determined under in vitro and in vivo conditions.

EXPERIMENTAL PROCEDURES
Molecular Biology-The human wild type (WT) CFTR was subcloned into the pCDNA3 mammalian expression vector (Invitrogen). A Cys-free construct without all 18 cysteines was provided by Dr. Gadsby (Rockefeller University) (12). ⌬R-S660A-CFTR was provided by Michael Welsh (University of Iowa) (31). All of the mutants were produced by using the QuikChange TM site-directed mutagenesis kit from Stratagene and confirmed by automated sequencing and Western blotting.
Chemicals and Reagents-MTS reagents were purchased from Toronto Research Chemicals. N-ethylmaleimide, diamide, DTT, cAMP, and forskolin were purchased from Sigma.
Cell Culture and Transfection-Human embryonic kidney (HEK)-293T cells were transiently transfected with wild type or mutant CFTR cDNAs using the Lipofectamine transfection kit (Invitrogen). Cells were cultured in Dulbecco's modified Eagle's medium (Mediatech) supplemented with 10% fetal bovine serum and 1 mM penicillin/streptomycin. For patch clamp recordings, all cells were grown on plastic coverslips and used 1-4 days postseeding. To improve expression, cells expressing Cys-free CFTR-based constructs inserted with V510A (32) were grown for 1-2 days at 24°C and then for 2-5 h at 37°C before measurements.
Patch Clamp Analysis-All of the channels expressed in HEK-293T cells were recorded in inside-out configurations (Axon Instruments, Foster City, CA) for intracellular application of reagents to the cytoplasmic face. CFTR currents were recorded in symmetrical solutions, each containing 140 mM N-methyl-D-glucamine chloride, 3 mM MgCl 2 , 1 mM EGTA, and 10 mM TES (pH 7.3). The resulting resistance of a borosilicate patch pipette was 1-2 megohms in the bath solution. Macroscopic currents were evoked using a ramp protocol

Inhibition of CFTR by Ser 768
from ϩ80 to Ϫ80 mV and filtered at 200 Hz. In vitro single channel recordings were determined at Ϫ60 mV by using inside-out patches and were filtered at 20 Hz. In vivo single channel recordings were done at ϩ60 mV by using cell-attached membrane patch clamp and were also filtered at 20 Hz. All of the experiments were carried out at room temperature (20 Ϯ 1°C). Data were acquired and analyzed using pCLAMP10.2 software (Axon Instruments). Curve fitting was made using Microcal Origin software. Student's t test was used for statistical analysis. A p value of 0.05 was considered as significant. Data are shown as mean Ϯ S.E.
Cysteine Cross-linking-Transfected HEK-293T cells expressing CFTR constructs with a specific single Cys or Cys pairs were washed in divalent-free PBS (Mediatech, Herndon, VA). CFTR proteins were oxidized in intact cells by incubation in 10 mM diamide for 10 min followed by a 5-min incubation in 5 mM N-ethylmaleimide. Cells were then solubilized in 100 l of SDS sample buffer for SDS-PAGE analysis. Half of each sample was treated with 100 mM DTT prior to running on a 4 -15% SDS-polyacrylamide gel. Separated proteins were transferred to PVDF membranes, blocked for 1 h with 5% milk in Tris (25 mM)-buffered saline, and then Western blotted with C-terminal anti-CFTR (monoclonal antibody clone 24-1, R&D Systems) at 1:4000 and detected with goat antimouse HRP-conjugated antibody (Amersham Biosciences) at 1:5000. Blots were extensively washed in TBS, exposed in SuperSignal West Pico Chemiluminescent Substrate (Pierce) for 5 min, and then developed on HXR film (Hawkins X-ray Supply).

RESULTS
Disulfide Cross-linking of the R Domain to CL3-If Ser 768 inhibits channel activity by interacting with outwardly facing residues from CL3, disulfide cross-linking of S768C to a corresponding cysteine inserted in CL3 will be expected to suppress channel activity. As a negative control, S768C cannot form an inhibitory disulfide bond with an inwardly facing cysteine mutated in CL3. To test this hypothesis, a Cys-free CFTR construct was used as a background clone to generate all Cys-containing mutants. S768C or S737C were fixed as an anchor point to search other inhibitory targets from CL3 ( Fig.  2A). S768C/H950C was a representative example (Fig. 2B). Once the mutant channel was activated by ATP (1.5 mM) and PKA (24 units/ml), further phosphorylation was blocked by protein kinase inhibitory peptide (PKI) to make sure that the subsequent reagents targeted not PKA but the CFTR channel. In response to the addition of 20 M diamide (a mild oxidizing reagent) to the cytoplasmic side of an inside-out patch, about 70% of both outward and inward Cl Ϫ currents were inhibited, and inhibition was reversed by 6 mM DTT (Fig. 2B). Finally, both outward and inward currents were blocked by a CFTR-specific inhibitor, CFTR inh172 (10 M). In contrast, both diamide and DTT had no effect on H950C and S768C CFTR constructs (Fig. 2, C and D). These observations clearly suggest that a disulfide bond may be formed between S768C and H950C. In other words, Ser 768 may be close to His 950 . Supporting this argument, internal diamide also inhibited activity of a mutant S768C/K951C, S768C/H954C, or S768C/ S955C, and inhibition was reversed by 4 -6 mM DTT (Fig.  2E). In sharp contrast, both diamide and DTT had no effect on such CFTR constructs as K951C, H954C, and S955C. Furthermore, channel activity of V769C/H950C was also inhibited by diamide (Fig. 2E), suggesting that phosphorylation of Ser 768 may not affect the CL3-R interface. However, S768C could not form an inhibitory disulfide bond with V956C (inwardly facing) or K946C, possibly as a result of a long distance or a poor relative orientation (Fig. 2E). Similarly, diamide also suppressed channel activity of mutants S737C/H954C, S737C/S955C, and S737C/Q958C, and suppression was reversed by DTT (Fig. 2E). These results suggest that Ser 737 may also be close to CL3.
In order to further confirm that Ser 768 and Ser 737 are close to hydrophilic residues from CL3 in the resting cell, intact HEK-293T cells expressing CFTR constructs were exposed to 10 mM diamide and were subsequently examined by SDS-PAGE electrophoresis and Western blotting analysis. Fig. 3 demonstrates that a CFTR construct with a single cysteine S768C, S737C, H950C, or H954C exhibited a clear single band no matter whether diamide or DTT was added. In sharp contrast, CFTR constructs with a cysteine pair (Cys-free background), S737C/H950C, S737C/H954C, S768C/H950C, and S768C/H954C, exhibited an additional cross-linked (Xlinked) band because it was induced by diamide but was weakened by DTT. In contrast, the H950C/V956C mutant exhibited no X-linked band possibly because of a poor relative orientation between H954C and V956C. Therefore, a disulfide bond can be formed between H950C (or H954C) and S768C or between H954C (or H950C) and S737C.
In order to address if the disulfide bond changes the gating kinetics, a two-channel recording of the H950C/S768C construct was done. Fig. 2F demonstrates that channel opening was clearly inhibited by diamide, but inhibition was partially reversed by 4 mM DTT, suggesting that disulfide cross-linking across the R domain and CL3 may prevent channel opening. Taken together, it is concluded that His 950 , Lys 951 , His 954 , Ser 955 , and Gln 958 from CL3 are close to Ser 768 and Ser 737 and may theoretically form inhibitory pairs with Ser 768 or Ser 737 regulating channel gating.
ATP-dependent Channel Opening by Curcumin-To further evaluate the contribution of the above inhibitory candidates to native channel gating, a missense alanine mutation was made, and ATP and PKA dependence was measured. A previous study indicated that curcumin activates a CFTR construct with the R domain deleted (21). However, my preliminary data demonstrated that before PKA was added, curcumin failed to activate WT CFTR even in the presence of ATP (Fig. 4A) until the R domain was removed (Fig. 4B). Therefore, it is very likely that an interaction between the R domain and CL3 may suppress channel activation by curcumin. In other words, curcumin may be a powerful tool to evaluate the relative contribution of inhibitory candidates at the R-CL3 interface to channel gating. The ⌬R construct was a good positive control, whereas WT CFTR was a negative one. Fig. 4C shows that S768A was dramatically activated by curcumin in the presence of ATP, but PKA failed to continue to potentiate channel activity. Thus, Ser 768 is a very strong inhibitory residue. It is very exciting that H950A was also greatly activated by curcumin after pretreatment of ATP, but subsequent PKA further increased channel activity (Fig. 4D). Similarly, curcumin also activated mutants K946A, K951A, S955A, and Q958A to a different extent in the presence of ATP (Fig. 4E). It is worth mentioning that Ser 955 was also a strong inhibitory residue. In contrast, curcumin had no such effect on S737A and H954A mutants, suggesting that they may be weak inhibitory residues, although disulfide cross-linking of S737C to H954C strongly inhibited channel activity (Figs. 2E and 4E). Because curcumin increased initial channel activity of most mutants at the R-CL3 interface after ATP was present, it is reasonable that subsequent PKA dependence was greatly reduced except for S737A and H954A. Especially, S768A and S955A, together with CFTR-⌬R, completely removed PKA dependence (Fig. 4F).
In order to further investigate if ATP is required for the effects of curcumin on S768A and H950A, curcumin was first applied to their intracellular sides before ATP was introduced. Fig. 5, A and B, demonstrates that curcumin alone failed to greatly activate these two mutants until ATP was subsequently added. Thus, the curcumin effects were ATP-dependent. ATP binding to the NBD1-NBD2 interface must be done for normal channel activation by curcumin (Fig. 5C). It is interesting that both H950A and S768A still needed more PKA to be fully activated in this case (Fig. 5D). . SDS-PAGE mobility analysis of CFTR mutants at the CL3-R interface. HEK-293T cells transfected by CFTR mutants were incubated with diamide (10 mM) for 10 min immediately before solubilization in the SDS sample buffer, SDS-PAGE, and Western blotting with a CFTR COOH tail mAb (24-1). The solubilized sample was divided into two equal portions, and half of each sample was reduced with DTT (100 mM) prior to loading the gel.

Ser 768 Involves an Inhibitory Hydrogen Bonding-Because
Ser 768 is a critical inhibitory site, the next question is what kind of interaction between Ser 768 and CL3 is responsible for inhibition. A previous study (24) suggested that preferential phosphorylation of Ser 768 suppresses channel activation at some excitatory PKA phosphorylation sites. Thus, if primary phosphorylation of Ser 768 inhibits channel activation by curcumin in the presence of ATP, S768D, which is equivalent to phosphorylated Ser 768 , should also dampen channel activation by curcumin. However, Fig. 6A demonstrates that S768D was also activated by curcumin with ATP. Therefore, most of the Ser 768 in WT CFTR may be less phosphorylated before application of PKA. In addition, because S768D is negatively charged, an electrostatic attraction between S768D and Lys 946 or Lys 951 may be impossible because modification of S768C with MTSCE (negatively charged) or MTSET (positively charged) failed to change channel activity (supplemental Fig.   S1, A and B). On the other hand, S768D is a strong H-bond acceptor (Table 1). Therefore, it is hypothesized that Ser 768 forms an H-bond with His 950 or Ser 955 , an inhibitory residue from CL3 (Figs. 2E and 4E). Because disulfide cross-linking of S768C to H950C inhibited more channel activity than that of S768C to S955C (Fig. 2E), it is more possible for His 950 to form an inhibitory H-bond with Ser 768 . This notion was supported by the observation that S768R, a strong H-bond donor, failed to be activated by curcumin even in the presence of ATP (Fig. 6B). Thus, Ser 768 may function as a proton donor to form a putative H-bond. Supporting this proposal, modification of S768C with MTSEA (a very strong H-bond donor (33)) also inhibited 25% of channel activity after the channel was activated by ATP and PKA followed by PKI to block further phosphorylation (supplemental Fig. S1, C and D). Because S768T strongly enhanced PKA dependence in the presence of ATP and curcumin, S768T may be a stronger H-bond donor ( Fig. 6B). In agreement with a putative H-bond between Ser 768 and His 950 , H950R was also activated by curcumin upon ATP treatment, but H950D was not (Fig. 6, C and D). Thus, His 950 may serve as an H-bond acceptor. Because H950Q reduced sensitivity to PKA even in the presence of ATP and curcumin, H950Q may be a stronger H-bond acceptor (Fig. 6D). What is more, curcumin also activated H950R/ S768R and H950D/S768D constructs in the presence of ATP (Fig. 6E) because two strong proton donors or acceptors cannot form an H-bond (Table 1). More importantly, even if both H950R and S768D could be activated by curcumin with ATP involvement (Fig. 6, A and C), H950R/S768D was silent in response to curcumin even in the presence of ATP (Fig. 6E), suggesting that a strong electrostatic attraction between H950R and S768D prohibit the channel from activation. Consistent with this notion, H950D/S768R was also not activated by curcumin with ATP (Fig. 6E). Taken together (Fig. 6F), it is proposed that a putative H-bond should be formed between His 950 and Ser 768 . The imidazole group of H950 may serve as an H-bond acceptor, whereas the hydroxyl group of Ser 768 may function as an H-bond donor. It is reasonable that Ser 737 may not form an inhibitory H-bond because S737D was not activated by curcumin even if ATP was added (Fig. 6F), further suggesting that Ser 737 may be a weak inhibitory site.
Effects of the Putative H-bond on ATP and PKA Dependence of Channel Activity-Because the curcumin effect was ATPdependent and resulted from disruption of the putative Hbond between His 950 and Ser 768 , it is fitting to ask if the putative H-bond prohibits channel opening by ATP and PKA. To address this question, sensitivity of His 950 or Ser 768 or combined mutants to ATP and PKA was determined in the ab-sence of curcumin. Fig. 7A shows that H950R/S768R was activated by ATP only, even without curcumin, but H950R or S768R was not (Fig. 7C). More importantly, H950R/S768R completely removed PKA dependence of channel activity (Fig.  7, A and D) no matter whether K978C, which promotes the channel opening without ATP, was inserted and accelerated channel activation by ATP (Fig. 7B) or not. In contrast, ATP failed to activate construct H950D/S768D, although an Hbond cannot be formed between two strong proton acceptors (Fig. 7C). This result may be due to an endogenous Fe 3ϩ binding between S768D and H950D, which also prevented the channel from opening (30). Supporting this hypothesis, Fig.  7C and supplemental Fig. S2 clearly demonstrate that the mutant S768D/H950D can be much activated by ATP once 5 mM EDTA was added to remove the endogenous Fe 3ϩ in the channel. In contrast, H950D and S768D could not be dramatically activated by ATP only even in the presence of 5 mM EDTA (Fig. 7C). Fig. 7D and supplemental Fig. S2 further show that the presence of EDTA clearly weakened the PKA dependence of H950D/S768D channel activity. Thus, a strong electrostatic expulsion between H950R/D and S768R/D promoted channel opening by ATP alone. It is reasonable that both Q958R/S737R and Q958D/S737D were not activated by ATP because Ser 737 was a weak inhibitory site (Fig. 7C). Unlike H950R/S768R and H950D/S768D, which exerted an electrostatic interaction between the R domain and CL3, H950A, S768A, S768D, and H950R were not apparently activated by ATP only (Fig. 7C) but more sensitive to PKA than WT CFTR (Fig. 7D). They only needed 5 units/ml PKA to achieve halfactivation, whereas at least 10 units/ml PKA was necessary for half-activation of WT CFTR. Thus, the putative H-bond suppressed channel activation by PKA.

Role of the Putative H-bond in Channel
Gating-In order to evaluate the contribution of the putative hydrogen bonding at the R-CL3 interface to channel gating, the unitary currents of both WT and CFTR mutants were recorded at Ϫ60 mV from inside-out membrane patches. Fig. 8A indicates that an apparent open probability of WT CFTR was as small as P o ϭ 0.0001 and not increased by ATP (P o ϭ 0.0002). Upon application of PKA, channel activity was dramatically increased but inhib-ited by CFTR inh172 (Fig. 8A). In contrast, an apparent open probability of S768A or H956A was as low as 0.0001 to 0.0005, comparable with that of WT CFTR. However, ATP clearly promoted channel opening of these two constructs up to 0.0040 -0.0132 (Fig. 8, B-D). These results suggest that disruption of the putative H-bond between Ser 768 and His 950 may promote channel opening by ATP binding at the NBD1-NBD2 interface. It is interesting that S768D also promoted channel opening by ATP (Fig. 8, D and E). Because S768D is equivalent to phosphorylated Ser 768 , this finding suggests that phosphorylated Ser 768 should not form an inhibitory H-bond. More importantly, this result suggests that WT CFTR should be less phosphorylated when the inside-out patch was excised from the HEK-293T cell.
Unlike H950A or S768A/D, an apparent open probability of H950R/S768R was higher (P o(app) ϭ 0.0042) than that of WT CFTR even in the absence of ATP, and ATP binding further increased channel opening (P o(app) ϭ 0.198) (Fig. 8, D and E). Similarly, H950D/S768D also exhibited an increased apparent open probability (P o(app) ϭ 0.0132) in the presence of 5 mm EDTA, and ATP continued to promote channel opening (P o(app) ϭ 0.0452) (Fig. 8, D and E). Therefore, although the exact number of activated channels in these constructs was unclear, an energy barrier may be dramatically lowered for spontaneous channel opening without ATP and PKA when an electrostatic expulsion at the R-CL3 interface was introduced. Taken together, these findings indicate that the putative Hbond between His 950 and Ser 768 may inhibit the channel from opening by ATP in the non-phosphorylated state.
Effects of the Putative H-bond on in Vivo Channel Activation-Because CFTR is a cAMP/PKA-regulated Cl Ϫ channel in the apical membrane of epithelial cells, it is interesting to know how the CFTR function will be altered by disrupting the putative H-bond when the protein is in vivo phosphorylated by cAMP-dependent PKA. To address this question, a cell-attached patch clamp was employed to record single channel activities from the HEK-293T cells expressing CFTR constructs. cAMP (300 M) and forskolin (50 M) were subsequently added to the extracellular perfusate to activate PKA in the cells. As shown in Fig. 9, extracellular cAMP failed to influence channel activity apparently because it is not membrane-permeable. However, forskolin clearly activated WT CFTR and mutant constructs because activation was blocked by CFTR inh172 (10 M) or glibenclamide (200 M). Fig. 9A indicates that it took about 3 min for the WT CFTR channel to be activated by forskolin. However, the activation time became significantly shorter for H950A, S768A, and S768D (Fig. 9, B-E). Thus, disruption of the putative H-bond may accelerate channel activation by cAMP-dependent PKA. It is very interesting that apparent basal activity of S768A was not so high and was comparable with that seen with S768D (Fig. 9, C and D). This observation is different from previous reports based on macroscopic recordings from Xenopus oocytes (24,25) and further suggests that fractional phosphoserine Ser 768 in WT CFTR should be very low when expressed in HEK-293T cells. In vivo single channel recordings further supported this notion. Fig. 9G shows that an open probability of WT CFTR was very low (0.00004) in the resting cells, no

TABLE 1 Potential roles in hydrogen bonding at the CL3-R domain interface
Note that mutants whose channel activity was increased by curcumin in the presence of ATP are highlighted in boldface type. This low basal open probability may be due to an endogenous inhibitory Fe 3ϩ binding at the CL3-R interface (30). Although S768A/D disrupted hydrogen bonding with His 950 , the Fe 3ϩ binding was still strong, and S768D may enhance the metal binding affinity (30). Supporting this possibility, H950A increased basal channel opening to a higher level, possibly by weakening Fe 3ϩ binding.

DISCUSSION
Ser 768 is a well known inhibitory PKA phosphorylation site in the CFTR channel. However, the functional role of nonphosphorylated Ser 768 is unclear. This study employed several independent approaches to first indicate that Ser 768 inhibited channel opening by primarily forming a putative H-bond with His 950 of CL3 in the nonphosphorylated state. Although thiolspecific disulfide cross-linking of S768C to H950C or nearby cysteines inserted in CL3 inhibited channel activity primarily by stopping the channel from opening, an electrostatic expulsion between S768R/D and H950R/D clearly promoted chan-nel opening even in the absence of ATP. Thus, Ser 768 is closed to His 950 of CL3. Furthermore, both S768A and S768D increased sensitivity of CFTR activity to ATP, curcumin, and PKA phosphorylation. Therefore, it is not phosphorylated Ser 768 but nonphosphorylated Ser 768 that stops the channel from opening by ATP and subsequent PKA phosphorylation. Finally, both H950R and S768D promoted channel opening by ATP followed by curcumin or PKA phosphorylation, but H950D or S768R could not. Therefore, a putative H-bond between the imidazole group of His 950 and the hydroxyl group of Ser 768 was proposed to inhibit channel opening by ATP and PKA phosphorylation.
Phosphorylation of Ser 768 -Previous studies indicated that Ser 768 can be phosphorylated possibly by AMPK to suppress activation of the WT CFTR channel in the resting cell (18,24,25). However, no evidence has demonstrated that all Ser 768 in the protein has been phosphorylated because previous SDS-PAGE mobility cannot tell the Ser 768 -phosphorylated band from the unphosphorylated R domain (24). In addition, phosphorylation of Ser 768 is regulated by endogenous basal AMPK activity (25). Finally, even if S768D activity is as low as WT CFTR activity under basal conditions (25), functional studies based on the whole-cell recordings cannot distinguish phosphorylated Ser 768 from the unphosphorylated residue if both inhibit channel activity. In this study, excised patches from the HEK-293T cell were used, and thus the CFTR channels have been dephosphorylated by membrane-associated phos- phatases. Fig. 6 clearly demonstrates that S768D, which is equivalent to phosphorylated Ser 768 , is different from WT CFTR in the inside-out patch. First, both S768A and S768D mutants could be activated by ATP followed by curcumin, but WT CFTR could not even be activated in the presence of ATP (Figs. 4 and 6). Second, both S768A and S768D were more sensitive to PKA phosphorylation than WT CFTR (Fig. 7D).
Third, the open probabilities of both S768A and S768D were increased by ATP, whereas that of WT CFTR was not (Fig. 8). Therefore, most of the Ser 768 in the WT CFTR may not be phosphorylated in the excised patch, and thus the H-bond can be formed between Ser 768 and His 950 . In fact, not all Ser 768 may be phosphorylated in the resting cell because the basal in vivo open probability of S768D was a little higher (P o ϭ 0.0004) than that of WT CFTR (P o ϭ 0.00004) (Fig. 9).
Role of Curcumin in Normal CFTR Gating-A previous study (21) showed that curcumin activates mutant CFTR channels, such as G551D, W1282X, ⌬1198, and A462F. These mutants have been shown to strongly inhibit ATP binding to the NBDs or dimerization of NBD1-NBD2 and thus to disrupt the normal ATP-dependent mode of gating. In contrast, slight activation of WT CFTR by curcumin still requires channel phosphorylation by PKA, although ATP seems to be unnecessary (21). Accordingly, it is not at all clear if the curcumin ef- fect has anything to do with normal ATP-dependent gating. However, Figs. 4 and 5 clearly demonstrate that regulation of normal channel gating by curcumin required ATP because ATP binding to the NBDs promoted channel opening of H950A and S768A mutants (Fig. 8). Although the action mechanism is still unclear, curcumin seems to function as a chemical "amplifier." Figs. 4 -8 demonstrate that any small differences in ATP and PKA dependence of channel activity could become dramatic upon pretreatment of curcumin. Therefore, curcumin was a powerful tool in this study to primarily evaluate the putative inhibitory hydrogen bonding between His 950 and Ser 768 by assessing the potency of curcumin to activate those inhibitory pairs at the R-CL3 interface.
H-bond at the CL3-R Interface-The proposal of the putative inhibitory H-bond between His 950 and Ser 768 in the un-phosphorylated CFTR is based on several lines of evidence. First, the imidazole group of His 950 and the hydroxyl group of Ser 768 can theoretically form an H-bond at the physiological pH because both groups can serve as proton donors or acceptors (Table 1). Second, diamide-induced disulfide bond crosslinking of S768C to H950C or its neighboring cysteines, which was confirmed by the SDS-PAGE mobility (Fig. 3), inhibited channel activity (Fig. 2). Thus, Ser 768 is close to His 950 , allowing the formation of the H-bond. Third, missense alanine mutations of Ser 768 and His 950 drove channel opening by ATP, which was enhanced by curcumin (Figs. 4, 5, and 8). In agreement with a proposal of His 950 as an H-bond acceptor and Ser 768 as an H-bond donor, S768D and H950R mutants were more sensitive to ATP or curcumin or PKA phosphorylation than S768R and H950D (Figs. 6 -8). Moreover, modification  (33)), but not with MTSCE (a strong H-bond acceptor), inhibited channel activity, whereas modification with MTSET failed to (supplemental Fig. S1). Finally, both proton donors cannot form an inhibitory H-bond between H950R and S768R. Instead, an electrostatic expulsion between H950R and S768R dramatically increased the ATP-independent open probabilities and sensitivity to ATP and PKA (Figs. 7 and 8). In contrast, channel activity was not potentiated by an electrostatic expulsion between H950D and S768D until EDTA was added to remove potential endogenous Fe 3ϩ binding to S768D and H950D, although both proton acceptors cannot form an inhibitory H-bond ( Fig. 7 and 8). Taken together, it is reasonable to propose that Ser 768 form a putative H-bond with His 950 to inhibit channel opening by ATP in the nonphosphorylated state. As shown in Fig. 10, the imidazole group of His 950 may serve as an H-bond acceptor, whereas the hydroxyl group of Ser 768 may serve as an H-bond donor. Once Ser 768 is phosphorylated by PKA or cAMP-dependent PKA, the H-bond is broken up.
In contrast, the S737A mutation failed to promote channel opening by ATP and curcumin, although it was also closed to CL3 (Figs. 2-4). Therefore, Ser 737 may not form an inhibitory H-bond with outwardly facing hydrophilic residues from CL3. Similarly, His 954 could not form an inhibitory H-bond with Ser 768 because H954A was not dramatically activated by a combination of ATP and curcumin (Fig. 4E). On the other hand, K946A, K951A, S955A, and Q958A still promoted channel opening by ATP followed by curcumin (Fig. 4, E and  F). This result suggests that an electrostatic interaction or other H-bond formation may be involved between these inhibitory residues in CL3 and other inhibitory residues in the R domain. For example, NEG2 (26) has multiple negative charges that could form salt bridges with Lys 946 and Lys 951 or H-bonds with Ser 955 and Gln 958 .
Role of R-CL3 Interfacial Hydrogen Bonding in Gating-Generally, activation of CFTR needs ATP binding to drive dimerization of NBD1-NBD2, and an H-bond between NBD1 and NBD2 plays a pivotal role (34). However, the putative hydrogen bonding between Ser 768 and His 950 stopped channel opening by ATP until it is disrupted by removal of an H-bond donor or an acceptor or by an electrostatic expulsion (Figs. 8 and 10). Our previous investigation suggested that CL3 may trigger channel opening possibly by stabilizing an outwardly facing conformation of MSDs with CL1 (17). Thus, the putative H-bond between the R domain and CL3 may impair the CL1-CL3 interaction, generate an activation energy barrier, and thus lower the channel open probability. Supporting this proposal, either a Fe 3ϩ bridge (30) or a disulfide bond (Figs. 2 and 3) between the R domain and CL3 prohibited channel opening, whereas an electrostatic expulsion between them significantly promoted channel opening even in the absence of ATP (Fig. 8). Therefore, the putative hydrogen bonding between the R domain and CL3 may play a critical role in regulating channel gating.
It is exciting that both S768A and H950A increased PKA sensitivity no matter whether curcumin was present or not (Figs. 4 -7). Because S768D also reduced PKA dependence of channel activity even without curcumin involvement (Figs. 6 and 7), most of the Ser 768 in WT CFTR may not be phosphorylated in the excised patch, and early phosphorylated Ser 768 may not attenuate channel activation by prohibiting PKA phosphorylation at some stimulatory sites, as suggested by Csanády and co-workers (24). In fact, Kongsuphole et al. (25) also demonstrated that phosphorylation of Ser 768 by AMPK cannot affect subsequent phosphorylation by PKA. Therefore, it is disruption of the putative H-bond between His 950 and Ser 768 that may promote PKA phosphorylation at some stimulatory sites. Although phosphoserine Ser 768 cannot form an H-bond with His 950 , the maximal open probability of WT CFTR was found to be lower than that of S768A (24). In this case, phosphorylated Ser 768 may suppress channel opening possibly by enhancing the endogenous Fe 3ϩ binding affinity (30). Thus, a Fe 3ϩ bridge at the interface of R-CL3 may also prevent stimulatory PKA sites from phosphorylation.
Effects of cAMP on Channel Gating-Previous studies demonstrated that the S768A mutant expressed in Xenopus oocytes exhibited weak phosphorylation of the R domain, high base-line activity, substantial activation by isobutylmethylxanthine/forskolin, and slight inhibition by local AMPK activation (18,24,25). However, single channel recordings of CFTR constructs expressed in HEK-293T cells indicated a complex situation. An open probability of WT CFTR was lower in vivo (P o ϭ 0.00004) than in vitro (P o ϭ 0.0001) (Figs. 8 and 9). This difference may not result from phosphorylation of Ser 768 because a basal open probability of S768D was higher (P o ϭ 0.0004) than that of WT CFTR (Fig. 9). Thus, it is proposed that endogenous Fe 3ϩ binding to CFTR stops channel opening in the resting cells. Supporting this proposal, H950A greatly facilitated basal channel opening in the resting cell, possibly because this mutation increased sensitivity to endogenous ATP and promoted phosphorylation at some stimulatory sites primarily by weakening endogenous Fe 3ϩ binding (Fig. 9). It is expected that H950R and H950R/S768R may also exert similar effects on the basal channel opening. Unlike H950A, a basal channel open probability of S768A and S768D was still low (P o ϭ 0.0004) (Fig. 9). Although endogenous ATP in the cell could promote channel opening (Fig. 8), endogenous Fe 3ϩ binding to CFTR may suppress channel opening. A similar observation would be seen with H950D/S768D. Despite this complex involvement, H950A, S768A, and S768D were more sensitive to forskolin than WT CFTR because they were dramatically activated soon after forskolin was introduced ( Fig. 9). This finding further supports the notion that the putative hydrogen bonding at the R domain and CL3 should dampen channel activation. Even if Ser 768 is preferentially phosphorylated, the putative Fe 3ϩ bridge at the R-CL3 interface could continue inhibiting channel opening.