Redox Reagents and Divalent Cations Alter the Kinetics of Cystic Fibrosis Transmembrane Conductance Regulator Channel Gating*

Gating of the cystic fibrosis Cl− channel requires hydrolysis of ATP by its nucleotide binding folds, but how this process controls the kinetics of channel gating is poorly understood. In the present work we show that the kinetics of channel gating and presumably the rate of ATP hydrolysis depends on the species of divalent cation present and the oxidation state of the protein. With Ca2+ as the dominant divalent cation instead of Mg2+, the open burst duration of the channel is increased approximately 20-fold, and this change is reversible upon washout of Ca2+. In contrast, “soft” divalent cations such as Cd2+ interact covalently with cystic fibrosis transmembrane conductance regulator (CFTR). These metals decrease both opening and closing rates of the channel, and the effects are not reversed by washout. Oxidation of CFTR channels with a variety of oxidants resulted in a similar slowing of channel gating. In contrast, reducing agents had the opposite effect, increasing both opening and closing rates of the channel. In cell-attached patches, CFTR channels exhibit both oxidized and reduced types of gating, raising the possibility that regulation of the redox state of the channel may be a physiological mode of control of CFTR channel activity.

The cystic fibrosis transmembrane conductance regulator (CFTR) 1 belongs to the superfamily of ATP-binding cassette transporters, members of which are characterized structurally by two ATP-binding motifs and functionally by their ability to couple ATP hydrolysis to transport of a substrate molecule across a membrane (1,2). Although a role for CFTR in solute transport has been proposed (3)(4)(5)(6), no substrate has yet been identified. CFTR remains unique among ATP-binding cassette transporters in that it couples ATP hydrolysis to the opening of a Cl Ϫ channel (6,7), reviewed in (8). The role of ATP hydrolysis in the CFTR gating process was inferred initially from studies of CFTR gating in electrophysiological experiments using poorly hydrolyzable ATP analogs and transition state phosphate analogs (9 -12). Direct measurement of ATPase activity in purified reconstituted CFTR has confirmed that it functions as an ATPase, and the measured turnover rate of the phosphorylated form (0.5-1 s Ϫ1 ) is consistent with a model in which opening and closing of the CFTR channel gate is coupled directly to ATP hydrolysis (13).
The role of ATP hydrolysis by the two nucleotide-binding folds (NBF1 and NBF2) in mediating channel gating has been elucidated by studies examining CFTR channels individually mutated in each of the NBFs (14 -16). These papers presented models in which two ATP hydrolytic events, one occurring at NBF1 and one at NBF2, are involved first in opening and then closing the CFTR channel. In one model a single cycle of ATP hydrolysis occurs at NBF1 to directly open the channel (8,14,16,17). Alternatively, a second model proposes that ATP hydrolysis at NBF1 converts the channel into a gatable state from which nucleotide binding at NBF2 can then open the channel (15). In both models, once the channel is open, the duration of the open burst is regulated by the hydrolysis rate of ATP at NBF2, with hydrolysis at NBF2 terminating the burst. Channel closure may occur either directly through the release of hydrolysis end products (ADP and inorganic phosphate) from NBF2 or indirectly by altering the hydrolysis rate or product off-rate at NBF1.
In addition to the nucleotide-dependent gating of CFTR, phosphorylation of the R domain plays an essential role in regulating channel gating. In its highly phosphorylated state, CFTR channels display a robust activation by ATP, whereas no channel gating occurs in an unphosphorylated state even in the presence of large concentrations of ATP (9 -12). Although phosphorylation is critical for channel gating, the mechanism by which the phosphorylated R domain interacts with ATP hydrolysis by the NBFs to control channel gating is poorly understood. In addition, it has been observed in patch clamp experiments that CFTR channel openings have a heterogeneity that is not completely explained by ATP concentrations or phosphorylation state (18 -20), so that there are probably other factors that regulate CFTR channel activity. One relatively unexplored source of control over CFTR channel gating is the redox potential within the cell. Perfusion of CFTR expressing cells with oxidized or reduced forms of pyridine nucleotides has been shown to modulate both basal and forskolin-stimulated Cl Ϫ conductance (21). Perfusion of oxidized purines through the patch pipette increased forskolin-stimulated whole cell Cl Ϫ current, whereas reduced purines inhibited the current. Moreover, cell-permeant oxidizing agents elevated the open probability (Po) of CFTR channels in cell-attached patches (21). However, in an apparent contrast, Koettgen et al. (22) have * This work was supported by a grant from the Cystic Fibrosis Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  ʈ Established Investigator of the American Heart Association. 1 The abbreviations used are: CFTR, cystic fibrosis transmembrane conductance regulator; NBF, nucleotide-binding fold; PKA, protein kinase A; ␤-ME, ␤-mercaptoethanol; NEM, N-ethylmaleimide; DTT, dithiothreitol; SNAP, S-nitroso-N-acetyl-penicillamine; ATP␥S, adenosine-␥-thiotriphosphate; DTNB, 5,5Ј-dithiobis-2-nitrobenzoic acid; Po, open probability; MOPS, 4-morpholinepropanesulfonic acid; POPE, reported that derivatives of cysteine with anti-oxidant properties increase the Cl Ϫ conductance of airway epithelial cells.
Although agents that alter the redox environment of the cell seem to affect CFTR channel activity, exactly what causes that effect remains unclear. The obvious targets for redox modulation with the potential to alter channel activity are cysteine residues in the channel protein. The CFTR sequence contains 18 cysteines, 14 of which are predicted to be intracellular. A role for cysteine residues in modulating CFTR channel activity has been demonstrated by Cotten and Welsh (23) who reported that the cysteine-modifying reagent N-ethylmaleimide (NEM) activates CFTR channels in excised patches, an effect that was eliminated by mutation of one cysteine, Cys 832 in the R domain.
In the present paper we show that the kinetics of channel gating depend on the species of divalent cation present and the oxidation state of the protein. When Mg 2ϩ is replaced by Ca 2ϩ or in the presence of sulfhydryl-reactive cations or oxidizing agents, the channel opens less frequently and closes more slowly, whereas reducing agents increase both the opening and closing rates of CFTR Cl Ϫ channels.
Microsome Preparation-Microsomes were prepared as described in Gunderson et al. (12) except that the final resuspension step was in a solution of 300 mM sucrose, 1 mM EDTA, 10 mM MOPS, 0.1% ␤-ME, pH 7.2. Prephosphorylation of the microsomes was performed upon thawing and addition of 4 mM ATP, 4 mM MgCl 2 , and 0.1 unit PKA/l. Microsomes stored in liquid N 2 were good for up to a year.
Planar Lipid Bilayers-Planar bilayers were formed by painting synthetic lipids (POPE:POPS, 7:3) dissolved in n-decane (20 mg/ml) across an aperture in a polycarbonate cup separating the cis and trans chambers. The average capacitance of the bilayers formed was ϳ100 picofarad. Addition of prephosphorylated microsomes to the cis chamber under stirring and a Ϫ60 mV holding potential led to fusion of CFTR containing vesicles within 10 -20 min. CFTR channels usually fused in with their cytoplasmic side facing the cis chamber as assayed by the ATP sensitivity of the cis versus trans chambers.
Instrumentation and Data Analysis-Bilayer current recordings were detected with a Warner bilayer clamp (BC 525A), prefiltered at 5000 Hz (Warner LPF2A), and stored on a modified DAT recorder (Sony DTC700). The desired records were played back from tape, filtered at 50 Hz, and digitized at 200 Hz with the Digidata 1200 A/D converter. Digitized data were analyzed with PCLAMP6 software. Open burst analysis used a 40-ms delimiter to separate closings within a burst from interburst closings, except with long calcium-mediated open bursts, which were performed manually. Mean dwell times were computed from exponential fits to the dwell time histogram. Records analyzed were from 2 to 20 min long.
Patch Clamp-Cell-attached and inside-out patch recording was done with HEK-293 cells stably transfected with the human CFTR protein. Current traces were collected with and digitized at 500 Hz with filtering at 100 Hz. Digitized data were analyzed with PCLAMP6 software (Axon Instruments, Foster City, CA) with filtering at 50 Hz. For burst duration analysis, the burst delimiter of 100 ms was determined from a plot of burst delimiter versus closings per burst as described in Sigurdson et al. (24). Analysis of the data using burst delimiters of 90 and 110 ms produced similar results. Open time and burst duration time constants were derived from fits of one or two exponentials using the maximum likelihood method. Burst and open time analysis was performed on patches with single channels or on unsuperimposed openings from multi-channel patches. The patch clamp buffer consisted of 135 mM N-methyl-D-glucamine, ϳ135 mM HCl, 10 mM HEPES, 3 mM MgCl 2 , pH 7.5. For cell-attached patches cells were bathed in 10 M forskolin ϩ 5 M phorbol ester to activate protein kinase C, and channels were recorded with a pipette potential of Ϫ60 to Ϫ70 mV. For inside-out patches, the solution bathing cytoplasmic face of the channel contained 1 mM ATP plus 5 mM MgCl 2 , and channels were recorded with a pipette potential of 60 -70 mV. Patches with low or no basal activity on excision were treated with 250 units/ml PKA prior to recording. SNAP, DTNB, and NEM were dissolved in dimethyl sulfoxide within 1 min of use and discarded afterward. All other drugs were diluted in patch clamp buffer. For some patches treated with drugs dissolved in dimethyl sulfoxide such as SNAP and NEM, in some cases the drugs were washed off after 5-7 min exposure and before channel activity was recorded to reduce noise and preserve the life of the patch.

Different Divalent Cations Have Different Effects on CFTR
Channel Gating-Enzymes that hydrolyze nucleotide triphosphates require divalent cations such as Mg 2ϩ to coordinate the ␥ phosphate of the nucleotide during enzymatic cleavage (25)(26)(27). Because of this requirement and the potential for specificity of ATP hydrolysis reactions for one cation versus another, different divalent cations may serve as useful probes for investigating the role of ATP hydrolysis in gating the CFTR channel. Therefore we examined the effect of replacing Mg 2ϩ with other divalent cations (Fig. 1). In experiments with CFTR channels fused into planar lipid bilayers, channel gating in the presence of 2 mM Mn 2ϩ or Co 2ϩ was indistinguishable from that observed with 2 mM Mg 2ϩ . No significant difference was observed in the open probability, or mean burst duration for any of these three divalent cations (Fig. 1, B, C, and E). In contrast, Ca 2ϩ increased open burst duration by more than an order of magnitude, from ϳ500 ms with Mg 2ϩ to almost 13 s in Ca 2ϩ (Fig. 1,  D and E). This dramatic increase in burst length results from Ca 2ϩ slowing the closing rate of the channel. Somewhat surprisingly, Ca 2ϩ had only a modest effect on opening rate of the channel because the mean closed dwell time was hardly affected (Fig. 1F).
To determine whether it was the absence of Mg 2ϩ or the presence of Ca 2ϩ that led to the prolonged open bursts, we examined the effect of mixed concentrations of Mg 2ϩ and Ca 2ϩ on CFTR gating. As shown in Fig. 2B, the presence of Ca 2ϩ at 0.8 mM induced long open burst durations even in the presence of a substantial Mg 2ϩ concentration (0.2 mM, enough for normal gating). A histogram of burst durations for the mixed Ca 2ϩ and Mg 2ϩ condition shows two distributions as compared with the single distribution of Mg 2ϩ alone (Fig. 2B). These data indicate that once Ca 2ϩ or Mg 2ϩ has induced an open burst, each cation remains tightly bound to the channel and is not readily exchanged for the other, resulting in two separate and characteristic burst lengths.
Ca 2ϩ , Co 2ϩ , and Mn 2ϩ represent one type of divalent cation, the so-called "hard" divalent cations, which do not readily share their electron density and are usually stabilized by simple electrostatics. Another type of divalent cation characterized as "soft" divalents are those such as Zn 2ϩ , Cu 2ϩ , Ni 2ϩ , and Cd 2ϩ , which have an easily polarizable electron cloud. These soft metals more easily participate in electron sharing and exhibit a greater degree of covalency in binding to ligands such as the sulfhydryl groups of proteins. An indicator of covalent modification is that the effect persists even after the cation has been washed out. In channels measured from both cell-free patches and planar lipid bilayers, the presence of soft cations had large effects on channel gating even in the presence of a 50-fold excess of Mg 2ϩ . As shown in Fig. 3, addition of Cd 2ϩ to CFTR channels in excised, inside-out patches dramatically altered channel gating, increasing the duration of the CFTR channel open bursts while prolonging interburst closed times (Fig. 3). In inside-out patches the durations of the open bursts of the CFTR channel are highly heterogeneous, and histograms of burst durations are fit by two exponentials resulting in two burst time constants, 1 and 2 . At 0.1 mM, Cd 2ϩ increased the length of the second component ( 2 ) from 4 s to nearly 8 s and substantially increased the percentage of the histogram fit by 2 from about a quarter of the opening events to more than twothirds of them. These changes correspond to an increase in both the average burst length (as shown by the change in percentage of events fit by 2 ) as well as the maximum burst length (as shown by the increase in the length of 2 ). Washout of Cd 2ϩ does not eliminate these effects or restore normal channel gating, indicating that these changes are most likely due to the covalent modification of the channel rather than an effect of Cd 2ϩ substituting for Mg 2ϩ during ATP hydrolysis. Consistent with the distinct characteristics of these types of metals, other soft divalent cations such as Zn 2ϩ and Ni 2ϩ had similar effects on the gating of CFTR channels fused into lipid bilayers. Zn 2ϩ and Ni 2ϩ as well as Cd 2ϩ prolonged open bursts of CFTR channels in bilayers, and the effects persisted even after the cations were washed out (data not shown).
Oxidizing and Reducing Agents Alter Gating Kinetics of CFTR Channels-The effects of Cd 2ϩ and other soft divalent cations on CFTR channel gating as well as the fact that these metals can covalently modify exposed sulfhydryl groups prompted us to examine the effect of other sulfhydryl modifiers such as reducing and oxidizing agents. In excised, inside-out patches, treatment with 10 mM ␤-ME had two effects on channel activity. First, it dramatically increased the activity of channels, even to the point of activating previously silent (low Po) channels. In many cases patches that appeared to have only one or two active channels would, after addition of ␤-ME, suddenly have openings of four or more channels (Fig. 4, A and B). The channels activated by ␤-ME were almost certainly CFTR channels, because the conductance of channel openings did not change and channel activity required both the presence of ATP and phosphorylation by PKA, two hallmarks of CFTR channel gating (Fig. 5). In patches with a single active CFTR channel, the presence of ␤-ME increased the opening rate of the channel resulting in a higher open probability (Fig. 6A). The stimulation of channel opening by ␤-ME was not mediated by activation of PKA to increase phosphorylation, because it was observed in 9 of 14 patches with no exogenous kinase present, (including the sample patches shown in Figs. 4 and 5) and in two of three patches in the presence of the specific PKA inhibitor, PKI (Fig. 4, B and C).
In addition to stimulating channel gating, a second effect of ␤-ME was a dramatic change in channel gating kinetics. In the presence of ␤-ME, the open burst duration was dramatically shortened, which, combined with the increase in opening rate, resulted in frequent, short channel openings Figs. 4 and 6). In contrast, oxidizing agents had effects essentially the opposite of reducing compounds, yielding channels with very long open bursts, separated by long silent periods in which the channel remained closed for minutes at a time (Fig. 6). Oxidizing conditions included treatment the strong oxidizer KMnO 4 , as well as the thiol-specific oxidizing agent DTNB and oxidized glutathione, a cellular oxidizer. All of the oxidizing agents had similar effects on CFTR channel gating, and channel gating in oxidized conditions was very like that seen with Cd 2ϩ and other tograms from channels in freshly excised patches (Fig. 7A). An exponential fit to burst duration histograms in reducing conditions results in a single time constant ( o ) that approximates the short component ( 1 ) of the control conditions. In contrast, the effect of oxidizing agents (DTNB, KMnO 4 , or oxidized glutathionel; data pooled together) on channel burst kinetics resembles that of Cd 2ϩ . In oxidizing conditions, both the length of the second burst duration time constant ( 2 ) and the percentage of burst events that are fit by the long component of the distribution were increased (Fig. 7B). Qualitatively similar results were observed with CFTR channels fused into lipid bilayers. Oxidizing agents dramatically lengthened CFTR channel open bursts, an effect that was rapidly reversed when bilayers were treated with ␤-ME or DTT (data not shown).
Reversibility of the Effects of Reducing and Oxidizing Agents-CFTR channels treated with ␤-ME or DTT exhibit changes in channel gating kinetics that are reversed when the reducing agents are washed out, suggesting that the redox potential of the inside-out patch preparation is relatively oxidizing. In general, CFTR channels in freshly excised inside-out patches tend to exhibit the long open bursts and long closed periods characteristic of oxidized channels. Moreover, the shorter burst durations caused by reducing agents were maintained only as long as the channel remained in highly reducing conditions. Once the reducing agent was washed out, channel gating resumed the longer bursts and longer interburst intervals characteristic of basal conditions (Fig. 8, A and B). In contrast, when treated with oxidizing agents, the channels maintained the slower kinetics even after washout; simply removing the oxidizing agent did not alter channel gating or return it to its basal state (data not shown).
In addition to being oxidized in their basal state, channels in inside-out patches were also resistant to chemical reduction and required millimolar concentrations of ␤-ME or DTT to alter their gating. Although 10 mM ␤-ME completely eliminated long duration open bursts (the burst duration histogram had only a single distribution), 1 mM ␤-ME had a much smaller effect (Fig.  8B). Like the basal conditions, the burst duration histogram from channels treated with 1 mM ␤-ME was fit by two exponentials, indicating that both long and short components of the burst duration histogram are present. However, the longer component of the histogram is much shorter in the presence of 1 mM ␤-ME than in the control, so that the effect on the burst duration is intermediate between the basal condition and in the The long open bursts exhibited by CFTR channels under oxidizing conditions resemble the "locked-open" state of the channel, which occurs when ATP hydrolysis is inhibited by poorly hydrolyzable ATP analogs (10 -12), raising the possibility that they work through a common mechanism. Because treatment with reducing agents eliminates the long open bursts characteristic of oxidizing conditions, we examined whether the locked-open channels induced by nonhydrolyzable analogs were similarly sensitive to reducing agents. As shown in the sample trace in Fig. 8C, 0.1 mM ATP␥S was able to lock channels open under reducing conditions when channel openings normally are shortened. When ATP␥S was added to patches already treated with 10 mM ␤-ME, the channels became locked open in three of three patches tested, indicating that the increase in channel closing rate caused by ␤-ME can be reversed by blocking ATP hydrolysis. Apparently when ATP hydrolysis is "locked" at a low rate, ␤-ME is unable to alter the closing rate of the channel. These data suggest that the effects of oxidizing and reducing agents are due to alterations in the rate of ATP hydrolysis, rather than a change in the pore properties of the channel.
The Effect of Reducing Agents Is Not Blocked by NEM-Data showing that the opening and closing rates of the CFTR channel can be altered by changing the oxidation state of the protein strongly suggest that cysteine residues in the channel modu-late gating. Recently Cotten and Welsh (23) reported that NEM activates CFTR channels by covalently modifying Cys 832 , a cysteine residue near the C-terminal end of the R domain. To test whether the effects of reducing agents are due to modification of Cys 832 , we examined the effect of NEM pretreatment on patches treated with ␤-ME. Because NEM irreversibly alkylates cysteines, residues that are already modified by NEM cannot subsequently be reduced. Pretreatment with NEM for 5-10 min did not block the effect of 10 mM ␤-ME on CFTR channels, because the burst duration histogram for channel openings in NEM plus ␤-ME was virtually identical to that of ␤-ME alone (Fig. 9, B and C). Because our data suggest that cysteine residues can be oxidized in inside-out patches and alkylation by NEM requires a reduced sulfhydryl, it is possible that in basal conditions critical cysteines are unavailable for NEM modification. To test this possibility we reversed the order of treatment: treating patches first with ␤-ME at 10 mM to ensure that all sulfhydryl moieties were reduced and then adding 100 M NEM. However, even when the channels were reduced first, the channel gating kinetics in the presence of ␤-ME and NEM together were no different than those of ␤-ME alone (data not shown). Although the dramatic effects of specific sulfhydryl reducing agents on CFTR channels indicates that cysteine residues are important modulators of channel gating kinetics, our results suggest that the modulation of channel gating by ␤-ME involves one or more residues that are FIG. 8. The effect of reducing agents requires a high concentration and is reversed upon washout. A, 60-s sample traces from an inside-out patch showing activity of CFTR channels before and after treatment with ␤-ME and then after washout. B, a graph of time constants is shown for exponential fits to burst duration histograms of 5000 events each from channels in control conditions and after washout of ␤-ME (10 mM), as well as patches treated with ␤-ME at 1 and 10 mM.  9. Effect of ␤-ME is not blocked by pretreatment with NEM. A, 60-s sample traces from a single inside-out patch treated with 0.1 mM NEM, and NEM plus 10 mM ␤-ME. B, burst duration histograms of 5000 channel events in the presence of NEM or NEM plus ␤-ME. C, graph of time constants from fits of burst duration histograms for 5000 events each from patches treated with NEM, ␤-ME and NEM ϩ ␤-ME. The shaded region represents the short component ( 1 ), and the white bar is the long component ( 2 ). For data taken in the presence of ␤-ME, the histogram was best fit by only one exponential. The pie graphs illustrate the percentage of burst events fit by each component.

resistant to modification by NEM.
Cell-attached Patches Show Both Oxidized and Reduced Types of Gating Behavior-CFTR channels in excised patches are activated by reducing agents and have dramatically different gating characteristics under reducing and oxidizing conditions. We were curious as to whether the gating of channels in cell-attached patches could be altered in a similar manner. To test this we examined the effects of reducing conditions cellattached patches. In cells treated with forskolin and phorbol ester to maximize protein kinase activation, the presence of 5 mM ␤-ME caused a rapid activation of CFTR channels in 14 of 19 cell-attached patches, and a sample experiment is shown in Fig. 10A. The activation in cell-attached patches was very similar to the effect of ␤-ME on channels in inside-out patches (Fig. 4). The apparent increase in the number and activity of channels was accompanied by a shortening of the open burst duration, which was similar to the effect of ␤-ME on channels in inside-out patches (Fig. 10B). Cell-attached patches from cells treated with forskolin and phorbol ester exhibit both the long open bursts characteristic of the oxidized channel and the short open bursts seen under reducing conditions (Fig. 10). Treatment of cells with ␤-ME resulted in shorter channel open bursts in cell-attached patches, just like excised patches. In a histogram of burst durations, channels in cell-attached patches show the same range of burst durations as in excised patches. Moreover, treatment with 5 mM ␤-ME reduced the frequency of long open bursts and shortened the time constant of the long component in a manner similar to that seen with inside-out patches (Fig. 10C).
The cytoplasm of cultured cells contains large concentrations (up to 10 mM) of redox buffers such as glutathione, most of it in the reduced state (28). Although the cytosol is a reducing environment, these data with cell-attached patches suggest that CFTR channels in the plasma membrane of intact cells exhibit striking heterogeneity in their oxidation state.
NO ϩ and S-Nitrosothiol Donors Oxidize CFTR Channels-The production of NO and its metabolic byproducts such as S-nitrosothiols has been hypothesized to support formation oxidized sulfhydryls even in the strongly reducing environment inside the cell (29). Therefore we used the NO ϩ and S-nitrosyl donor S-nitroso-N-acetyl-penicillamine (SNAP) to test whether these possible endogenous oxidizing agents could modify CFTR channel gating. As shown in Fig. 10, SNAP had the same effect on CFTR channel gating as Cd 2ϩ and oxidizing reagents. Treatment with 0.1 mM SNAP increased the length of the open bursts while decreasing the frequency of opening, resulting in long channel openings separated by long silent periods. The increase in burst length was reflected in an increase in the time constant for the long component of the burst duration histogram. In addition, like other oxidizing agents, SNAP also increased the portion of the burst events in the histogram that were fit by the long component (Fig. 11). These results indicate that in the presence of SNAP, longer bursts were more frequent, and there was an increase in the length of the longest bursts. The effect of SNAP persisted on washout but was rapidly reversed by treatment with ␤-ME, indicating that the effect on the channels was due to oxidation of cysteine residues. These data add support to the hypothesis that NO ϩ or its metabolites could modulate CFTR channel gating by affecting the redox state of cysteine residues in the protein. DISCUSSION In this paper we investigated the effects of divalent cations and sulfhydryl modifying reagents on the gating of CFTR channels. The principal findings of this study are that CFTR channel gating kinetics can be influenced by covalent modification of sulfhydryl residues and that channel activity in vivo is subject to modulation by intracellular redox status. Our data also show that Ca 2ϩ ions when substituted for Mg 2ϩ dramatically alter CFTR gating via a noncovalent interaction with the protein.
Divalent cations serve as useful probes for investigating the role of ATP hydrolysis in gating the CFTR channel because of their requirement in ATP hydrolysis reactions. Many well studied ATPases including P-type ATPases (30), V-type ATPases (31), and ATP-binding cassette transporters such as P-glycoprotein (32,33) and the maltose transporter (34) exhibit similar rates of hydrolysis with Mg 2ϩ , Mn 2ϩ , and Co 2ϩ but are orders of magnitude slower at hydrolyzing Ca 2ϩ ATP. Our finding that replacement of Mg 2ϩ with Ca 2ϩ leads to a large (ϳ20-fold) increase in the open burst duration of CFTR channels most likely reflects the dramatically slower rate of hydrolysis of Ca 2ϩ ATP compared with Mg 2ϩ ATP and supports previously proposed models of CFTR gating in which an open burst may be terminated by hydrolysis of ATP at NBF2 (14 -16).
Although an essential role for ATP hydrolysis in initiating an open burst is strongly supported by studies in which ATP is completely replaced with hydrolysis-resistant ATP analogs (10 -12), the stoichiometry of the coupling of ATP hydrolysis to channel opening is controversial. According to one model (8,14,16,17), initiation of each open burst is directly coupled to hydrolysis of ATP by NBF1. The longer closed times observed in CFTR channels mutated at the invariant lysine (Lys 464 ) in the NBF1 Walker motif are consistent with a role for ATP hydrolysis at NBF1 in the initiation of an open burst. However, the effect of these Lys 464 mutations on opening rate is relatively small (2-4-fold), compared with the large (1-2 orders of magnitude) effect of the analogous mutation in NBF2 (Lys 1250 ) on open burst duration. Furthermore, the large effect of mutations in the corresponding invariant "P-loop" lysines on the hydrolysis rate of other GTPases and ATPases (35)(36)(37) indicates that the hydrolysis rate of Lys 464 mutants is likely to be very low. This discrepancy between the predicted hydrolysis rate and the effect on channel gating led us to propose that ATP hydrolysis at NBF1, while essential for CFTR activity, is not directly coupled to the initiation of each open burst (15). This model is strengthened by the observation in this paper that the kinetics of channel opening are unaffected by replacement of Mg 2ϩ ATP with Ca 2ϩ ATP, even while the closing rate is greatly reduced.
Our data also suggest that CFTR gating can be modulated by changes in the oxidation state of sulfhydryl residues. Reducing conditions accelerate both the opening rate of the channel (causing more frequent openings) and the closing rate of the channel (resulting in very short openings), whereas oxidizing conditions or treatment with "soft" divalent cations reverses these effects. The more frequent channel opening events in reducing conditions increased average channel Po in both cell-attached and excised patches despite the shorter length of the openings. In contrast, in oxidizing conditions, channels open into bursts that can last for minutes at a time, indicating that the closing rate of CFTR channels is significantly slower than in reduced conditions. In addition, oxidized channels also have a slower opening rate compared with reduced conditions, so that the closed times are dramatically lengthened. The fact that in freshly excised patches, channel gating behavior falls largely in the middle of the oxidized and reduced modes of gating probably reflects the heterogeneity of the channel in patches when the redox state is not controlled.
It is possible that the treatment with sulfhydryl-specific drugs alters CFTR gating not through a direct effect on the channel but through the action of some other modulating protein such as a kinase or phosphatase. However, the activating effects of reducing agents are probably not due to increased phosphorylation because they occurred even in the absence of added kinase and in the presence of the PKA inhibitory peptide PKI. Although it is conceivable that an endogenous kinase other than PKA may be involved, increased phosphorylation as would be expected under reducing conditions has been shown to lengthen the duration of channel openings rather than shortening them (20,38). Similarly, if oxidizing conditions simply reduced the level of phosphorylation of channels, then one would expect to see only a reduction in channel opening events not the changes in burst kinetics that we observed. Moreover, CFTR channels fused into lipid bilayers showed the same alterations in gating kinetics in the presence of oxidizing and reducing agents as channels in inside-out patches, indicating that the effects of redox are not dependent on the particular proteins present in excised patches.
Our data support a model in which modification of sulfhydryl residues alters the hydrolysis rate of ATP at one or both NBFs with reducing conditions favoring faster hydrolysis and oxidizing agents decreasing the rate of ATP hydrolysis. This hypothesis is supported by our finding that reducing agents are unable to shorten the long open bursts caused by nonhydrolyzable analogs. Apparently, when the ATP hydrolysis rate is "locked" at a low rate, reducing agents are unable to alter channel gating.
Could changes in open burst duration through regulation of the redox state of the channel be a way for cells to control the open probability of CFTR? In cell-attached patches we observe the same range of long and short burst durations as in excised patches, and others have observed similarly long open bursts in cell-attached patches (18,39). Moreover, we observed a decrease in long open bursts in cell-attached patches when cells were treated with ␤-ME, raising the possibility that channels in the plasma membrane may not be in a uniformly reduced state. There have been reports of other membrane proteins in which intracellular cysteines appear to be oxidized at the plasma membrane. Formation and breakage of an intracellular disulfide bond has been proposed as a mechanism for regulation of the coated vesicle vacuolar H ϩ -ATPase as it cycles from the plasma membrane to clathrin-coated vesicles and back to the plasma membrane (40). Despite the typically reducing environment of the cell, it is possible for intracellular cysteines to be oxidized, either through the formation of an intramolecular disulfide bond or formation of a mixed disulfide with a redoxbuffering peptide such as glutathione. Redox potential in cells is maintained largely by glutathione, which exists in the cytoplasm in both oxidized and reduced forms. Although the ratio of reduced to oxidized glutathione in the cytoplasm is on the order of 20:1 (28), conditions at the plasma membrane may be significantly more oxidizing than the bulk cytoplasm (41). In addition, the ability of a pair of cysteine residues to form a disulfide bond is determined by their oxidation potential, so that residues with a high oxidation potential can form disulfide bonds even in a reducing environment (40). The fact that large concentrations of reducing agents are required to affect CFTR channel gating indicates that the sulfhydryls being reduced have a very high redox potential and could potentially be oxidized at physiological concentrations of reduced glutathione. In addition, the ratio of oxidized to reduced forms of biological thiols is very sensitive to intracellular redox potential, and small changes can lead to significant alterations in the ratio of oxidized to reduced forms (42).
Oxidizing pressure in the cell can come from oxygen free radicals (43) or NO ϩ , a byproduct of intracellular nitric oxide production (29). NO ϩ can interact with thiol groups of cysteine residues to form S-nitrosothiols. These nitrosothiol groups can then be transferred between and among proteins or glutathione molecules tremendously extending the lifetime of NO-mediated effects on intracellular proteins (44). NO ϩ -mediated oxidation of sulfhydryl groups on intracellular cysteine residues has been implicated in the regulation of a number of proteins including olfactory cyclic nucleotide-gated channels (45), L-type calcium channels (46), calcium-dependent potassium channels (47), type I adenylyl cyclase (48), and p21 ras (49). Epithelial cells of the type that express CFTR have the enzyme nitric oxide synthetase and can produce quantities of nitric oxide, particularly in the lung (50 -52). Therefore, it is possible that redox modulation through the production of intracellular nitric oxide may play a role in regulation of CFTR in vivo. Our experiments with the NO donor SNAP support this possibility by demonstrating that cysteine residues in CFTR can be oxidized by exposure to NO ϩ and nitrosothiols.
The effects of redox state that we have described may shed some light on the variable behavior of CFTR channels in different laboratories and in different preparations. Redox potential is rarely controlled in patch clamp experiments, and so conditions are likely to be mostly oxidizing. In contrast, preparation of microsomes for fusion of channels into bilayers is frequently done under reducing conditions that would reverse any cellular oxidative changes and might result in a channel with quite different gating characteristics. As a further complication, protein kinase subunits added to increase channel activity are frequently kept in an active state by storage in as much as 50 mM of a reducing agent such as DTT, so that the addition of the kinase to patches or bilayers may affect more than the phosphorylation state of the channel. Taken together, our data demonstrate that changes in the oxidation state of CFTR affect both the opening and the closing rate of the channel and indicate that these changes may have some physiological relevance to the cellular regulation of channel activity.