pH-dependent gating of ROMK (Kir1.1) channels involves conformational changes in both N and C termini.

ROMK channels (Kir1.1) are members of the superfamily of inward rectifier potassium channels (Kir) and represent the channels underlying K+ secretion in the kidney. As their native counterparts, Kir1.1 channels are gated by intracellular pH, with acidification leading to channel closure. Although a lysine residue (Lys80) close to the first hydrophobic segment M1 has been identified as the pH sensor, little is known about how opening and closing of the channel is accomplished. Here we investigate the gating process of Kir1.1 channels exploiting their state-dependent modification by water-soluble oxidants and sulfhydryl reagents. Mutagenesis of all intracellular cysteines either alone or in combination revealed two residues targeted by these reagents, one in the N terminus (Cys49) and one in the C terminus (Cys308) of the channel protein. Both sites reacted with the thiol reagents only in the closed state and not in the open state. These results indicate that pH-dependent gating of Kir1.1 channels involves movement of protein domains in both N and C termini of the Kir1.1 protein.

ROMK channels (K ir 1.1) are members of the superfamily of inward rectifier potassium channels (K ir ) and represent the channels underlying K ؉ secretion in the kidney. As their native counterparts, K ir 1.1 channels are gated by intracellular pH, with acidification leading to channel closure. Although a lysine residue (Lys 80 ) close to the first hydrophobic segment M1 has been identified as the pH sensor, little is known about how opening and closing of the channel is accomplished. Here we investigate the gating process of K ir 1.1 channels exploiting their state-dependent modification by water-soluble oxidants and sulfhydryl reagents. Mutagenesis of all intracellular cysteines either alone or in combination revealed two residues targeted by these reagents, one in the N terminus (Cys 49 ) and one in the C terminus (Cys 308 ) of the channel protein. Both sites reacted with the thiol reagents only in the closed state and not in the open state. These results indicate that pH-dependent gating of K ir 1.1 channels involves movement of protein domains in both N and C termini of the K ir 1.1 protein.
Potassium homeostasis is controlled by secretion of K ϩ ions across the apical membrane of cortical collecting duct cells in the kidney. As the channels responsible for K ϩ secretion, low conductance (35 pS) inwardly rectifying K ϩ channels with particular high sensitivity to changes in intracellular pH (pH i ) were identified (1,2). Intracellular acidification in the physiological range reversibly reduced open probability of these channels and is thought to account for the subsequent decrease in K ϩ secretion (1). Thus, the sensitivity of the apical K ϩ channel to pH i is assumed to play a key role in K ϩ homeostasis (2,3).
Efforts to identify channel molecules responsible for renal K ϩ secretion resulted in cloning of ROMK1 (4) and its splice variants ROMK2 and ROMK3 (5,6). Splicing results in variable length of the respective N termini with ROMK2 shortened by 19 amino acids and ROMK3 exhibiting an extension of 7 residues with respect to ROMK1. ROMK channels are members of a superfamily of structurally and functionally related K ϩ channel proteins (K ir channels, for review see Refs. 7-9). As deduced from the primary structure, these K ϩ channel subunits are made up of hydrophilic N and C termini that flank a well conserved core region. The latter consists of two hydrophobic segments (M1 and M2) and a putative P region (4,10). Similar to K v -type K ϩ channels, K ir channels are assembled from four subunits (11,12) and form homo-and heteromultimeric channel proteins (11,13). The common functional property of K ir channels is their inwardly rectifying current-voltage relation (I-V), which may be weak or strong and which is due to a voltage-dependent block of the channel pore by intracellular polyamines (14 -17).
ROMK1 (K ir 1.1a) as well as ROMK2 (K ir 1.1b) and ROMK3 (K ir 1.1c) encode weak inward rectifier K ϩ channels and are differentially expressed in renal tubular cells (5,18). As their native counterparts, K ir 1.1 channels are gated by pH i with acidification leading to channel closure (19 -21). The steadystate current-pH i relation shows a pH i value for half-maximal activation (pH 0.5 ) of 6.9 and a Hill coefficient of around 3, indicating cooperativity of the gating process (19). As the determinant responsible for sensing pH i we have identified a lysine residue (Lys 80 ) close to M1 whose protonation triggers pH-dependent gating (19). Thus, the sensor of pH-dependent gating is known, whereas the processes involved in opening and closing of K ir 1.1 channels are largely unknown. Here we investigate the pH-dependent opening and closing of K ir 1.1 channels using state-dependent chemical modification of cysteine residues (22,23) intrinsic to the K ir 1.1 sequence.

MATERIALS AND METHODS
Mutagenesis and cRNA Synthesis-All mutants were prepared with standard techniques (24) and subcloned into pBF1, 1 and the mutation was verified by sequencing. Capped cRNAs specific for K ir 1.1 wild type or mutant channels were synthesized in vitro using SP6 polymerase (Promega, Heidelberg, Germany) and stored in stock solutions at Ϫ70°C.
Preparation and Injection of Oocytes-Xenopus oocytes were surgically removed from adult females and manually dissected. About 50 nl of a solution containing cRNA was injected into Dumont stage VI oocytes. Oocytes were treated with collagenase type II (Sigma, 0.5 mg/ml) and incubated at 19°C for 2-4 days prior to use.
Cysteine Modification-DTT, 2 DTNB and ATP (potassium salt) were purchased from Sigma, MTSES was purchased from Toronto Research Chemicals (North York, ON, Canada), and PIP 2 was obtained from Boehringer (Mannheim, Germany). For DTT and DTNB, stock solutions (100 mM) were made and stored at Ϫ20°C; the final dilutions were used for about 8 h. MTSES was freshly prepared prior to each experiment and used within 20 min. PIP 2 (100 M) was dissolved by 30 min of sonification in the cytoplasmic solution (25,26).
Electrophysiology and Data Evaluation-Electrophysiological recordings were performed 3-7 days after injection using the giant patchclamp recording as described previously (17). Briefly, pipettes were made from thick walled borosilicate glass had resistances of 0.3-0.6 M⍀ (tip diameter of 20 -30 m) and were filled with 120 mM KCl, 10 mM HEPES and 1.8 mM CaCl 2 ; currents were sampled at 10 kHz and corrected for capacitative transients with an EPC9 amplifier (HEKA electronics, Lamprecht, Germany), with analog filter set to 3 kHz (Ϫ3 db). Cytoplasmic solution was applied to excised patches via a multibarrel pipette and had the following composition: 100 mM KCl, 10 mM HEPES, 10 mM K 2 EGTA, (total K ϩ of 120 mM); pH was adjusted to the * 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 U.S.C. Section 1734 solely to indicate this fact.
values indicated by titration with HCl or KOH, respectively. All experiments were performed at room temperature (approximately 23°C).

RESULTS
K ir 1.1 Channels Are Redox-sensitive-The pH-dependent gating of K ir 1.1 channels as recorded in giant inside-out patches from Xenopus oocytes at a membrane potential of Ϫ80 mV (intermittently stepped to 50 mV for 50 ms) is illustrated in Fig. 1. Intracellular acidification (from pH i 8.0 to pH i 6.0) closes channels down, and alkalinization results in reopening ( Fig. 1A and Ref. 19). Complete recovery from pH inactivation, i.e. reopening of all channels, was seen upon short acidifications (Fig.  1A, left panel), whereas only a fraction of channels could be recovered after longer periods in pH i 6.0 (Fig. 1A, right panel). As shown previously (19,27), unitary conductance and open probability of single channels did not change during acidification or alkalinization.
Because the incomplete recovery was reminiscent of "channel run-down," a phenomenon well known for K ir channels, Mg-ATP (28,29) and the anionic phospholipid PIP 2 (25,26,30), both reported to counteract run-down, were tested. As shown in Fig. 1B, neither reagent was able to restore channel activity. Addition of DTT (100 M) or reduced glutathione (5 mM) to the pH i 8.0 solution led to complete recovery even after prolonged acidification (n ϭ 12, Fig. 1C). This suggests that oxidation occurs during acidification, which subsequently prevents channels from recovery from pH inactivation.
Recovery from pH Inactivation Is Different with Oxidizing and Modifying Reagents-Redox sensitivity of K ir 1.1 channels was further investigated with reagents that differentially interact with sulfhydryl-groups. Cu(II)-1,10-phenantroline (Cu-Phen), which induces formation of disulfide bonds, largely reduced the fraction of channels that spontaneously recovered upon realkalinization (Fig. 2A). Addition of DTT to the pH i 8.0 solution still resulted in complete recovery from pH inactiva- pH-dependent Gating in K ir 1.1 Channels tion ( Fig. 2A, n ϭ 2). In contrast, when reagents that chemically modify cysteine residues such as MTSES or DTNB were applied at acidic pH i , DTT failed to recover channels from pH inactivation (Fig. 2, B and C, n Ͼ 10). The increase in current observed upon realkalinization in the experiments in Fig. 2 (B and C) most likely represented unmodified channels because no recovery was seen when application of DTNB was extended to periods longer than Ϸ2 min (not shown). These results indicate that Cu-Phen-induced oxidation is reversible, whereas modification by DTNB or MTSES irreversibly locks channels in a closed state.
Modification of Channels by DTNB Is State-dependent-To more closely characterize modification by DTNB and its relation to pH-dependent gating, experiments were performed where DTNB was applied at pH i 8.0 either before or after pH-induced inactivation, i.e. when channels were either in the open or pH-inactivated state. As shown in Fig. 3 (A and B), DTNB did not affect open channels at pH i 8.0, nor did preapplication of DTNB change subsequent pH gating (n ϭ 3).
In contrast, when DTNB was applied during DTT-induced recovery from pH inactivation, the recovery process was promptly stopped (Fig. 3C, n ϭ 5). As before (Fig. 3A), DTNB did not induce inactivation in this experiment. Thus, under identical conditions, channels were only susceptible to chemical modification when they were pH-inactivated prior to DTNB application.
This coupling of chemical modification to pH inactivation was confirmed by experiments with a mutation of K ir 1.1, K ir 1.1(K80M), in which the pH sensor has been removed and which therefore did not exhibit pH-dependent gating (19). As shown in Fig. 3 (D-F), no effect of DTNB was observed, neither at basic nor at acidic pH i (Fig. 3, D and F, n ϭ 4). The reversible decrease in current amplitude seen at pH i 6.0 reflects a weak pH dependence also known for pH-insensitive K ir 2.1 (IRK1) channels and at least in part due to block of channels by hydrogen ions (19). Taken together, these results indicate that K ir 1.1 channels are targeted by DTNB in a state-dependent manner, i.e. channels are modified in the pH-inactivated state but not in the open state.
State-dependent Modification Occurs at Residues in the N and C Terminus-This state dependence together with the fact that DTNB specifically modifies cysteine residues was exploited to see which domains of the K ir 1.1 protein move during pH-dependent gating.
For this purpose all cysteines in the K ir 1.1 sequence were replaced by alanine or serine (Fig. 4A). Mutations in the N and C termini outside the "core region" (hydrophobic domains M1 and M2 and the P region) resulted in functional channels gated by intracellular pH, whereas no currents were observed upon expression of the two Cys 3 Ala/Ser exchanges in the P region (C121(A/S) and C153(A/S)).
The functional mutations were tested for recovery from pH inactivation in the presence of DTT after they were inactivated by a 50-s application of the pH i 6.0 solution either in the absence (control) or presence of DTNB. As shown in Fig. 4 (B and C), all mutants recovered very similar to WT under control conditions, and DTNB modification was not abolished by either of the single C to A/S exchanges. However, the fractional recovery from inactivation after DTNB modification observed in C49A and C308A was significantly larger than for WT or any of the other mutants (Fig. 4B) and led to the assumption that DTNB modification may occur at more than one cysteine. This was indeed verified, because in the double mutant K ir 1.1(C49A,C308A) recovery from pH inactivation in the presence of DTT was complete, independent of whether DTNB was added to the inactivation solution (Fig. 5).
These results show that the cysteine residues modified by sulfhydryl reagents in a state-dependent manner are Cys 49 in the N terminus and Cys 308 in the C terminus of the K ir 1.1 protein. Moreover, the results indicate that pH-dependent gating in these channels is accompanied by structural rearrangements in both intracellular N and C termini. DISCUSSION The results presented here show that pH-dependent opening and closing of K ir 1.1 channels is accompanied by conformational changes of the channel protein involving movement of domains in the intracellular N and C termini. This movement was visualized by state-dependent modification of Cys 49 and Cys 308 , which were susceptible to reaction with DTNB in the pH-inactivated state, whereas no modification was observed for open channels.
Besides modification by DTNB or MTSES, channels were also oxidized in a state-dependent manner. Cu-Phen applied at pH i 6.0 prevented channels from recovery from inactivation, an effect that could be fully reversed by application of the reducing agent DTT (Fig. 2). But although oxidation and reduction were induced by agents specific for formation and reduction of disul-

FIG. 3. Modification of channels by DTNB is state-dependent.
A-C, application of DTNB on K ir 1.1 channels at pH i 8.0 either before (A and B) or after acidification (C). Note that DTNB irreversibly stopped recovery from pH inactivation induced by DTT, whereas it exerted no effect when applied prior to acidification. D and F, application of DTNB on K ir 1.1(K80M) mutant channels that do not exhibit pH-induced inactivation. E, same experiment as in F but without DTNB during acidification. Application of DTNB and DTT was as indicated; acidification and alkalinization are given as in Fig. 1; calibration of time and current is as indicated.
pH-dependent Gating in K ir 1.1 Channels fide bonds, no significant alteration of the redox sensitivity was observed in either of the Cys 3 Ala/Ser mutations. Moreover, oxidation was also observed in the double-mutant K ir 1.1(C49A,C308A), which completely abolished state-dependent modification by DTNB (Fig. 5) as well as in the triple mutants K ir 1.1(C308A,C355A,C358A) and K ir 1.1(C49A, C175A,C308A). When all cysteines outside the core region were replaced by alanine, no channel activity was detected upon functional expression. Thus, we could not identify the cysteine(s) involved in oxidation nor exclude a role for a protein putatively associated with ROMK.
The reducing agents DTT and reduced glutathione were required for complete recovery from pH inactivation in excised inside-out patches. In living cells, however, the redox sensitivity most likely has no physiological implication, because reduced glutathione should be available in millimolar concentrations. Moreover, experiments with nitric oxide (NO, SIN-1, and S-nitrosocysteine as NO donors), which replaced DTNB in activation of cyclic nucleotide-gated channels (31, 32), did not reveal any effect on pH-dependent gating of K ir 1.1 channels.
The structural rearrangements that occur during pH gating in both N and C termini of K ir 1.1 represent the first direct evidence for conformational changes in the K ir channel family. Work on GIRK channels (K ir 3.0 subfamily) showed that protein domains in both N and C termini are involved in interaction with G-proteins (33)(34)(35). These domains mediate binding of G␤␥ or G␣␤␥ to the K ir subunit. It is not yet known which parts of the molecule move during G-protein-induced channel gating.
Gating induced by changes in extracellular pH (pK a around 2.5) has recently been related to conformational changes in the two-segment type potassium channel from Streptomyces lividans (KscA, (36)). Using EPR spectroscopy, Perozo and coworkers uncovered structural rearrangements at the C-terminal end of the second transmembrane helix (37,38) that were hypothesized to change the width of the inner vestibule and to thus control ion permeation. Whether such a mechanism also underlies gating in K ir 1.1 channels with rearrangements in N and C termini as secondary processes or whether these cytoplasmic domains form the actual gate of the channel remains to be elucidated. Furthermore it remains to be elucidated whether the conformational changes resulting from pH-dependent gating may affect processes reported to affect channel activity such as phosphorylation by protein kinase A (39,40) or interaction with anionic phospholipids (26).

FIG. 4. Cysteine mutagenesis of state-dependent modification.
A, membrane topology of K ir 1.1 channels as derived from hydropathyplots, all cysteine residues are indicated. B, fractional recovery by DTT in K ir 1.1 WT and mutant channels after DTNB modification at pH i 6.0 (see "Materials and Methods"). C, fractional recovery of oxidized channels by DTT. Note the different scaling of the y axis in B and C. Data are the means Ϯ S.D. of three to five experiments.
FIG. 5. State-dependent modification occurs at residues Cys 49 and Cys 308 . A and B, recovery from pH inactivation in K ir 1.1(C49, 308A) mutant channels and K ir 1.1 WT with (B) and without (A) DTNBmediated cysteine modification. Note that in K ir 1.1(C49A,C308A) mutant channels recovery from pH inactivation was no more affected by DTNB. Experimental protocol was as described for Fig. 1; applications, acidification, and alkalinization were as indicated. C, fractional recovery by DTT with and without DTNB modification. Data are the means Ϯ S.D. of five experiments. Data from wild type and single cysteine mutations (all shown in gray) were added for better comparison.
pH-dependent Gating in K ir 1.1 Channels