Molecular Basis and Structural Insight of Vascular KATP Channel Gating by S-Glutathionylation*

The vascular ATP-sensitive K+ (KATP) channel is targeted by a variety of vasoactive substances, playing an important role in vascular tone regulation. Our recent studies indicate that the vascular KATP channel is inhibited in oxidative stress via S-glutathionylation. Here we show evidence for the molecular basis of the S-glutathionylation and its structural impact on channel gating. By comparing the oxidant responses of the Kir6.1/SUR2B channel with the Kir6.2/SUR2B channel, we found that the Kir6.1 subunit was responsible for oxidant sensitivity. Oxidant screening of Kir6.1-Kir6.2 chimeras demonstrated that the N terminus and transmembrane domains of Kir6.1 were crucial. Systematic mutational analysis revealed three cysteine residues in these domains: Cys43, Cys120, and Cys176. Among them, Cys176 was prominent, contributing to >80% of the oxidant sensitivity. The Kir6.1-C176A/SUR2B mutant channel, however, remained sensitive to both channel opener and inhibitor, which indicated that Cys176 is not a general gating site in Kir6.1, in contrast to its counterpart (Cys166) in Kir6.2. A protein pull-down assay with biotinylated glutathione ethyl ester showed that mutation of Cys176 impaired oxidant-induced incorporation of glutathione (GSH) into the Kir6.1 subunit. In contrast to Cys176, Cys43 had only a modest contribution to S-glutathionylation, and Cys120 was modulated by extracellular oxidants but not intracellular GSSG. Simulation modeling of Kir6.1 S-glutathionylation suggested that after incorporation to residue 176, the GSH moiety occupied a space between the slide helix and two transmembrane helices. This prevented the inner transmembrane helix from undergoing conformational changes necessary for channel gating, retaining the channel in its closed state.

ATP-sensitive K ϩ (K ATP ) 7 channels are expressed in a variety of tissues, including smooth muscles, pancreatic ␤-cells, myocardium, and neurons, where they play an important role in cellular function (1,2). Activity of the K ATP channels is tuned by physiological or pathophysiological stimuli, including hypoxia, hyperglycemia, ischemia, and oxidative stress, allowing a regulation of cellular excitability according to the metabolic state (3). The vascular smooth muscle (VSM) isoform of K ATP channels regulates vascular tones (4,5). Activation of the channel by vasodilators produces hyperpolarization of VSM cells, reduces activity of the voltage-dependent Ca 2ϩ channels, and relaxes VSMs. Inhibition of the channel leads to constriction of VSMs. Disruption of the vascular K ATP channel in mice results in vasospasm in coronary arteries and sudden cardiac death (6,7).
Other studies have further shown that disruption of the vascular K ATP channel has drastic effects on the systemic response to septic stress. With a forward genetic approach by genomewide random chemical mutagenesis, Croker et al. (8) screened a large population of mice and found four strains that are highly susceptible to multiple septic pathogens, including lipopolysaccharides (LPSs). The LPS hypersensitivity phenotype of these mice is due to a null allele of Kcnj8, encoding the Kir6.1 subunit of the vascular K ATP channel (8). Similar septic susceptibility has been observed in Kcnj8-knock-out mice that also show coronary hypoperfusion and myocardial ischemia during LPS exposure (9). These studies thus indicate that the vascular K ATP channel not only contributes to the vascular tone regulation at physiological conditions but also affects critically systemic stress responses.
Our recent studies have shown that the vascular K ATP channel is strongly inhibited in oxidative stress by S-glutathionylation (10). S-Glutathionylation is a post-translational modification mechanism occurring in a variety of physiological or pathophysiological conditions (11). This protein modulation mechanism is remarkable especially in vasculatures because oxidative stress is a major contributing factor to several cardiovascular diseases, in which S-glutathionylation plays an important role (12). Although S-glutathionylation is often associated with the adverse effects of oxidative stress, such a protein modulation is reversible under certain circumstances and can act as a functional modulation mechanism like protein phosphorylation (11). Thus, demonstration of how S-glutathionylation affects protein function should have broad significance. As for the vascular K ATP channel S-glutathionylation, the molecular basis remains unclear. Therefore, we performed studies to elucidate the molecular basis of and to provide structural insight into the K ATP channel S-glutathionylation.

MATERIALS AND METHODS
Chemicals and Reagents-Chemicals or reagents used in this study were purchased from Sigma-Aldrich unless stated otherwise. High concentration stocks were prepared in double-distilled water or DMSO. The final concentration of DMSO in the experimental solution was less than 0.1%, which did not cause any detectable effect on the channel activity.
Electrophysiology-Patch clamp experiments were performed according to our previous reports (13)(14)(15)(16)(17)(18). Whole-cell currents were recorded in voltage clamp with a holding potential of 0 mV and a hyperpolarizing step to Ϫ80 mV. Symmetric concentrations of K ϩ (145 mM in total) were applied to both bath and pipette solution. The bath solution contained 10 mM KCl, 135 mM potassium gluconate, 5 mM EGTA, 5 mM glucose, and 10 mM HEPES (pH 7.4), whereas the pipette solution had 10 mM KCl, 133 mM potassium gluconate, 5 mM EGTA, 5 mM glucose, 1 mM K 2 ATP, 0.5 mM NaADP, 1 mM MgCl 2 , and 10 mM HEPES (pH 7.4). All recordings were made with an Axopatch 200B amplifier (Axon Instruments Inc., Foster City, CA). The data were low pass-filtered (2 kHz; Bessel 4-pole filter; Ϫ3 dB) and digitized (10 kHz, 16-bit resolution) with Clampex 9 (Axon Instruments Inc.). Macroscopic currents were recorded from giant inside-out patches, and single-channel currents were recorded from regular inside-out patches with a constant single voltage of Ϫ80 or Ϫ60 mV. In the inside-out patch configurations, 1 mM K 2 ATP and 0.5 mM NaADP were included in the bath solution to maintain the channel activity. A higher sampling rate (20 kHz) was used for the currents recorded from the inside-out patches. Data were analyzed using Clampfit 9 (Axon Instruments Inc.).
Mutagenesis and Construction of Chimeras-Construction of chimeras was described previously (16). Briefly, Kir6.x channels were divided into three regions (N terminus, core domain containing two transmembrane domains and the pore loop, Site-directed mutagenesis kits (Stratagene) were used to introduce mutations. In these experiments, two oligonucleotide primers (30 -35 bp) were synthesized, each of which contained the same mutation and annealed to the same sequence on opposite strands of the cDNA. Pfu DNA polymerase was used to extend and incorporate the mutagenic primers, leading to two nicked circular strands. DpnI was used to digest the parental cDNA. The circular, nicked double-strained DNA was transformed to XL1-Blue competent cells for amplification. The correct constructs and mutations were confirmed by DNA sequencing.
Streptavidin Pull-down Assay and Western Blot-HEK cells expressing WT or mutant Kir6.1/SUR2B channels were used for these experiments. Cells were allowed to grow to ϳ80 -90% confluence (about 10 6 cells) in the 35-mm dish. Before experiments, cells were incubated with fresh medium for 2 h. Biotinylated glutathione ethyl ester (BioGEE; 250 M) (Invitrogen) then was added to the medium for a 1-h incubation followed by a 15-min H 2 O 2 (750 M) challenge. Excess free BioGEE was removed by three washes (10 min each) with PBS (containing 0.3% Triton X-100). Radioimmune precipitation assay buffer (100 l) (Sigma) was used for cell lysis. Protein concentration was measured using a BCA protein assay system (Thermo Scientific, Waltham, MA) and ranged from 2 to 8 mg/ml in our preparation. All of the WT and mutant protein samples were diluted to 1 mg/ml so that at this time point all of the protein concentrations were comparable. For the Western blot before pull-down, 15-l samples of both WT and mutant channels were used for immunoblotting. Samples were run on a 10% SDS-polyacrylamide non-reducing gel and then transferred to a nitrocellulose membrane (Bio-Rad). Rabbit primary antibodies against Kir6.1 (1:500) (Sigma, P0874), monoclonal antibodies against ␤-actin (1:5000) (Sigma, A5441), and secondary antibodies conjugated with alkaline phosphatase (1:10,000) (Jackson ImmunoResearch) were used. SuperSignal West Pico chemiluminescent substrate (Thermo Scientific) was used for signal visualization. Experiments were repeated at least four times.
Streptavidin-Dynabeads (Invitrogen, M-280) were used to pull down the biotin-GSH-conjugated proteins. Briefly, before the immobilization, the Dynabeads were washed three times (10 min each) with washing buffer to remove preservatives. Protein samples (100 l, in a concentration of 1 mg/ml) from HEK cells expressing either WT Kir6.1 channels or C176A mutant channels were then mixed with 100 l of Dynabeads (10 mg/ml), respectively. The mixture was then incubated at room temperature with gentle rotation for 30 min. The Dynabeadprotein complex was separated as pellet from other unlabeled proteins by a magnet. The pellet containing labeled proteins was resuspended, followed by three washes, whereas the supernatant was discarded. At last, the pellets were mixed with 20 l of loading buffer containing 0.1% SDS for boiling to release the glutathionylated proteins into solvent. The solvent containing the pull-down samples (15 l) was loaded to the gel for Western blot as described above. In this procedure, the same amount of protein (100 l, 1 mg/ml) was treated with the same amount of streptavidin beads (100 l, 10 mg/ml) for both WT and mutant channels.
The photodensity of the WT and mutant channel bands before and after pull-down was quantified by ImageJ (National Institutes of Health). The photodensity of the WT channel after pull-down was divided by the photodensity of the WT channel before pull-down (designated as R WT ). Similarly, the photodensity of the C176A mutant channel after pull-down was divided by the photodensity of the C176A mutant channel before pulldown (R MT ). R MT was then divided by R WT to obtain the relative (percentage) density.
Structural Modeling-The native Kir6.1 structure in its closed state was generated using the I-TASSER protein structure prediction server (19 -21) (25,26). The native Kir6.1 structural model was further modified by incorporating the GSH molecule onto the thiol group of Cys 176 . Several candidate locations and orientations of GSH in the Kir6.1 structural model were considered based on the interaction of charged groups of GSH with that of Kir6.1 proteins and refined using energy minimization (27). The one with the lowest binding energy between GSH and the Kir6.1 protein was chosen for further analysis and presentation. The sequence alignment was created by ClustalW2. Structural images were generated with the PyMOL molecular graphics system (DeLano Scientific, LLC, San Carlos, CA). Distances between molecules were measured with PyMOL.
Data Analysis-Data are presented as means Ϯ S.E. Differences were evaluated with Student's t tests or analysis of variance, and the statistical significance was accepted when p was Ͻ0.05.

RESULTS
Subunit Identification-Functional K ATP channels consist of four pore-forming Kir6.x subunits (Kir6.1 or Kir6.2) and four regulatory sulfonylurea receptors (SUR1, SUR2A, or SUR2B) (2, 4). The Kir6.1/SUR2B channel is considered to be the major vascular isoform of the K ATP channels, although Kir6.2 is also found in VSM cells (28,29). Our recent studies demonstrate that the Kir6.1/SUR2B isoform of vascular K ATP channel is targeted by reactive oxygen species through S-glutathionylation (10). To further identify the specific subunit(s) or isoform(s) targeted by oxidants, we expressed Kir6.1/SUR2B and Kir6.2/ SUR2B channels in HEK cells separately. Both Kir6.1/SUR2B and Kir6.2/SUR2B currents were activated by the specific K ATP channel activator pinacidil (10 M) and inhibited by the channel inhibitor glibenclamide (10 M) (Fig. 1, A and B, Glib). The currents of interest were normalized to the window of pinacidil and glibenclamide effects.
Intracellular Versus Extracellular Location-To determine the intracellular versus extracellular location of the S-glutathionylation sites, the effects of membrane-permeable and -impermeable PDSs were compared. Under the same recording conditions as for 2-DTP, a high concentration of 5,5Ј-dithiobis-2-nitrobenzoic acid (DTNB) (200 M), a membrane-impermeable PDS, marginally inhibited the channel activity by 28.7 Ϯ 5.2% (n ϭ 5) (Fig. 1, D and H). These different effects of DTNB and 2-DTP were not due to different potencies of these PDSs because 2-DTP and DTNB inhibited Kir6.1/SUR2B channel activity to almost the same extent when they were applied to the internal membranes of inside-out patches (Fig. 1, E, F, and I).
The different effects of 2-DTP and DTNB in whole-cell recording versus inside-out patches therefore indicate that the modulation of thiol groups takes place mainly on the intracellular side of the plasma membranes.
Determination of Critical Protein Domains-To determine the critical domains for channel inhibition, we took the advantage of the different sensitivity of the Kir6.1 and Kir6.2 to thiol oxidation and performed experiments using six Kir6.1/Kir6.2 chimerical constructs created in our laboratory (16). In these chimeras, the N terminus, transmembrane core region, and C terminus were swapped between Kir6.1 and Kir6.2. All constructs expressed functional channels with SUR2B. The responses of these chimerical channels to 2-DTP were then examined. In comparison with 222 (or Kir6.2; 12.9 Ϯ 0.5% inhibition; n ϭ 4) (Fig. 2, A and B), chimeras 122 (the N terminus of Kir6.2 was substituted with that of Kir6.1; other chimeras are named similarly) and 212 gained substantial sensitivity to 2-DTP with the N terminus or core domain from Kir6.1 (68.5 Ϯ 6.2 and 57.6 Ϯ 5.9% inhibition, respectively; n ϭ 4; p Ͻ 0.05) (Fig. 2, A, C, and D). However, chimera 221 did not show any significant difference in its response to 2-DTP compared with 222 (Fig. 2 (Fig. 2A). When both the N terminus and the core domain existed, the 112 had 2-DTP sensitivity (91.9 Ϯ 8.0% inhibition; n ϭ 4) as great as the 111 ( Fig. 2A). This chimerical analysis thus suggests that both the N terminus and the core domain are crucial for the channel inhibition by 2-DTP.
Cys 176 in the Kir6.1 Subunit-Systematic mutagenesis was carried out on all cysteine residues of the Kir6.1 (Fig. 3, A and  B). All of the mutants exhibited functional channel activity when expressed with SUR2B in HEK cells. By 2-DTP screening, we found that one cysteine residue in the core domain (Cys 176 ) and another in the N terminus (Cys 43 ) were important, and mutations of these residues disrupted the channel sensitivity to 2-DTP (50 M) in whole-cell recordings (Fig. 3, B-D). Mutation of any other cysteine residues did not result in a significant change in 2-DTP sensitivity using analysis of variance (Fig. 3B). Of these two cysteine residues, the Cys 176 was more prominent in the oxidant-mediated channel inhibition because the Kir6.1-C176A/SUR2B channel lost most of its 2-DTP sensitivity (26.0 Ϯ 6.5% inhibition, p Ͻ 0.001) (Fig. 3, B and C). Consistently, this mutant also lost its sensitivity to H 2 O 2 (3.5 Ϯ 4.7% inhibition; p Ͻ 0.001).
Cys 43 and Cys 120 in Oxidant Sensitivity-The Cys 43 mutation (C43A) reduced the channel sensitivity to 50 M 2-DTP (69.0 Ϯ 3.8% inhibition; n ϭ 5) (Fig. 3, B and D), although the effect was significantly smaller than that of Cys 176 (p Ͻ 0.001). We further tested the effect of GSSG on the C43A mutation and found that the response of the Kir6.1-C43A mutant to GSSG was comparable with its response to 2-DTP, showing rather modest inhibition of the channel activity (60.9 Ϯ 3.2% inhibition; n ϭ 4) (Fig. 4, C and E). When the Kir6.1-C176A/C43A double mutation channel was tested, we found that the GSSG effect was completely eliminated (Ϫ3.9 Ϯ 3.5% inhibition; n ϭ 4) (Fig. 4, D  and E).
In addition to Cys 176 and Cys 43 , Cys 120 appeared to be important for the channel sensitivity to extracellular oxidants. The Kir6.1-C120S/SUR2B mutant channel lost substantial 2-DTP sensitivity (77.6 Ϯ 5.1% inhibition; n ϭ 4) (Fig. 3E), although statistical significance was not found using analysis of variance. When the data were tested with Student's t test, however, its response to 2-DTP was significantly smaller than that of the Kir6.1/SUR2B WT channel (p Ͻ 0.01). Therefore, further studies were conducted on this site. In inside-out patches, the Kir6.1-Cys 120 mutant did not show any reduction in its sensitivity to GSSG in comparison with Kir6.1/SUR2B WT channel (87.0 Ϯ 5.8% inhibition (n ϭ 4) versus 84.3 Ϯ 8.7% inhibition (n ϭ 5), respectively; p Ͼ 0.05) (Fig. 4E). When this mutant was examined in whole-cell recordings using membrane-impermeable DTNB (200 M) applied extracellularly, the effect of DTNB was eliminated almost completely (2.7 Ϯ 6.0% inhibition; n ϭ 6) (Fig. 3F) in comparison with the Kir6.1/SUR2B WT channel (28.7 Ϯ 5.2% inhibition by DTNB; n ϭ 5) (Fig. 1D). These data thus suggest that Cys 120 may serve for extracellular redox sensing but does not seem to be an S-glutathionylation site.
Biochemical Evidence for Cys 176 Modification-Because Cys 176 was the dominating residue mediating the S-glutathi-  According to our previous study, a strong reactive band of ϳ32 kDa is Kir6.1-specific (10). The specificity of the antibodies is also confirmed by internal experiments performed by Sigma-Aldrich. With ␤-actin as loading control, we found that the density of the Kir6.1-C176A-reactive band was comparable with that of the Kir6.1 WT band (106 Ϯ 8.5%; n ϭ 5), indicating that the C176A mutation did not change the protein expression pattern of the Kir6.1 subunit (Fig. 5A). These constructs thus were subjected to a streptavidin pull-down assay. The HEK cells transfected with Kir6.1/SUR2B or Kir6.1-C176A/SUR2B were incubated with BioGEE (250 M) for 1 h, followed by 15 min of H 2 O 2 (750 M) challenge as described previously (10,32). If BioGEE was incorporated into the channel proteins, streptavidin-beads then should pull down the channel-BioGEE complex, which would be further detected by Kir6.1 antibodies in Western blot. On the other hand, if the mutation impaired the protein S-glutathionylation, then the mutation should decrease the binding of Kir6.1 protein to BioGEE, resulting in a weaker band or even no band in Western blot. Indeed, we observed different band densities between these two constructs. After streptavidin pull-down, the density of the Kir6.1-C176A-reactive band was much weaker (38.6 Ϯ 3.8%; n ϭ 5) (Fig. 5, A and B) compared with the band of the Kir6.1 WT channel after normalizing to protein inputs (Fig. 5, A and B). The data further demonstrate that the C176A mutation impaired the Kir6.1 protein S-glutathionylation.
As a control, the washout of the streptavidin bead-protein mixture was also used for Western blot. No clear bands were   Simulation Modeling-Several Kir channel protein crystal structures, including the eukaryotic Kir2.2 channel structure, have been resolved recently. Although N and C termini show notable variation among different Kir channels, the core domains are rather conservative. To gain a structural insight into how the S-glutathionylation of Cys 176 affects the channel activity, we modeled the Kir6.1 structure with a GSH moiety bound at residue 176 in its closed state (Fig. 6, A and C-G). Our model showed that after incorporation, the GSH moiety occupied the space between the M0, M1, and M2 helices (Fig. 6,  C-G). The 360°view of the GSH-bound local structure is also presented (supplemental Fig. 1). Because the channel opening I, schematic view of K ATP channel in its closed state. The "activation gate" Phe 178 blocks the pore and prevents the ion from passing through. J, the administration of K ATP channel opener causes a conformational change of K ATP channel; the inner helix bends at Gly 175 and moves the gate (Phe 178 ) away from the ion conduction pathway to facilitate the passage of potassium ions. K, S-glutathionylation at residue Cys 176 results in the incorporation of GSH that occupies the space between the inner and outer helices. The addition of GSH prevents the channel from entering its open state. requires a movement of the M2 inner helix (23,25), the addition of GSH at residue Cys 176 prevented the M2 inner helix from undergoing such a necessary movement for channel opening. To further validate this idea, we also modeled the Kir6.1 structure in its open state (Fig. 6, B and H) and compared it with the Kir6.1 model in its closed state. In the Kir6.1 open state model, the M2 inner helix was bent and twisted, making the Cys 176 residue face the M1 outer helix directly (Fig. 6H). After energy minimization, the distance between Cys 176 and the closest residue (Leu 73 ) on the M1 outer helix was 2.8 Å (Fig. 6H). When there was a GSH moiety on Cys 176 , the channel open state with such a short distance between Cys 176 and Leu 73 could not be achieved. Consequently, the channel was retained in its closed state.

DISCUSSION
S-Glutathionylation is a post-translational modulation mechanism that is involved in a variety of physiological or pathophysiological events (33). We have recently shown that the VSM K ATP channel, a vascular tone regulator, is modulated by S-glutathionylation. Our data in the present study reveal that the Kir6.1/SUR2B channel modulation by S-glutathionylation is likely to take place primarily at the Cys 176 residue of the Kir6.1 subunit. Because of its critical location, the incorporation of the GSH moiety into Cys 176 prevents the pore-forming inner helices from undergoing necessary conformational change for opening and retains the channel in its closed state.
Excessive reactive oxygen species produced during oxidative stress can result in structural modification of proteins affecting protein function (12). Although some studies have shown that the K ATP channels are targeted by redox regulation, data are rather inconsistent regarding the effect of oxidants or thiol oxidation on K ATP channels from different tissues. Reactive oxygen species lower the K ATP channel activity in cerebral arterioles (34,35) and coronary arteries (36) but facilitate its opening in cardiac myocytes (37) and pancreatic ␤-cells (38). Our results suggest that the differential responses of K ATP channels to reactive oxygen species in these tissues are likely to be due to different isoforms of K ATP channels expressed.
K ATP channels consist of Kir6.1-containing channels and Kir6.2-containing channels. Although the Kir6.2 is the closest family member of Kir6.1, the intrinsic properties of these two groups of channels are quite different (39). For example, they have different biophysical properties and are modulated differently by PKA and PKC (40 -42); the sensitivities of these channels to ATP and ADP are distinct (43); and without ATP, the Kir6.2 channel often opens automatically, whereas the Kir6.1 channel generally has low basal activity (16,43,44). Therefore, the studies of the Kir6.2 channel often take the advantage of its automatic opening, whereas the studies of the Kir6.1 channel require the K ATP channel openers to activate the channel. In this current study, we found another major difference between these channels. The Kir6.1/SUR2B channel is inhibited by oxidants and reactive disulfides strongly, whereas the Kir6.2/ SUR2B channel is barely inhibited by these reagents. These different responses may be attributable to the difference between Kir6.1 and Kir6.2 subunits. However, we cannot rule out the possibility of the involvement of the SUR subunit. The close coupling between Kir6.x and SURx is well recognized. The SUR is known to interact with the N terminus but not the C terminus of Kir6.x to affect channel gating (45,46). Such an interaction may rely on certain N-terminal residues and produce different conformational supports for distinct Kir6.x subunits. However, our data do suggest that the cysteine residues in the SUR2B subunit are not functionally S-glutathionylated, which is supported by our systematic mutagenesis for all the intracellular cysteine residues in the SUR2B subunit (supplemental Fig. 2).
Previous studies of the oxidant sensitivity of K ATP channels were mainly focused on Kir6.2-containing channels (2). Studies of Kir6.2-containing channels in native tissues suggest that the application of H 2 O 2 facilitates the opening of these channels, preferentially through changing the ATP/ADP sensitivity of the Kir6.2-containing channels (38,47 Biochemical experiments, including Western blot and pulldown assays, require the use of Kir6.1 antibodies. Kir6.1 antibodies have been generated from different laboratories and are also commercially available. Interestingly, although specific, antibodies from different sources detect bands of distinct sizes in different tissue samples. For example, a single band of 35, 44, or 49 kDa has been detected in rat heart tissue from three independent studies (48 -50), respectively. In rat brain samples, a single band of 51 kDa has been detected (48), whereas another study has detected the Kir6.1 band between 37 and 50 kDa in adult mouse hippocampus (51). In the present study, we used the Kir6.1 antibodies purchased from Sigma-Aldrich. The size of the Kir6.1-reactive band detected in our study is close to the band detected from heart tissue by Sun et al. (48). It is suggested that the post-translational modification, the natural truncation of Kir6.1 protein during the preparation, and the methods of sample treatment (e.g. lysis reagents, reducing or non-reducing preparation, etc.) may affect the size of Kir6.1 protein from different tissues. K ATP channels have been modeled previously using crystallized bacterial channels as templates (52)(53)(54)(55). In this study, the closed state of the Kir6.1 is modeled using the most recently crystallized eukaryotic Kir2.2 together with bacterial Kir channels as templates (see supplemental Fig. 3  for sequence alignments). The open state model is generated based on the open state KirBac1.1 structural model provided by Dr. Venien-Bryan (25,26). With the information from these structures, a model of the S-glutathionylation-mediated Kir6.1/SUR2B channel gating is proposed (Fig. 6, I-K, and supplemental Movie 1). A conservative phenylalanine residue (Phe 146 of KirBac1.1; Phe 178 of Kir6.1) located in the narrowest region of the ion conduction pathway is likely to serve as "blocking residue/activation gate" that prevents K ϩ from passing through when the channel is closed (23,56). When the channel is open, the slide helix moves laterally and exerts strain on the bottom of the inner helix, resulting in the bending of the inner helix at a weak point (e.g. a glycine residue (Gly 175 of Kir6.1)). This bending moves the side chain of the blocking residue (Phe 178 of Kir6.1) away from the center of the ion conduction pathway and allows K ϩ to pass through (25). In the channel open state, Cys 176 of the M2 helix makes a close contact with the M1 helix. The closest distance between Cys 176 and Leu 73 in the M1 helix is measured to be 2.8 Å. Such a short distance/small space cannot accommodate a GSH moiety (or even a smaller thiol modulation regent, 2-DTP). Therefore, when the channel is S-glutathionylated, the open conformation of the channel cannot be achieved.
In the Kir6.2 channel, Cys 166 (corresponding to Cys 176 in Kir6.1) is suggested to be involved in the intrinsic channel gating (57). The channel with the C166S mutation lost most of its sensitivity to both K ATP channel opener and inhibitor with a drastic augmentation of the channel open probability (57). However, Cys 176 in the Kir6.1/SUR2B channel does not seem to be a general gating site. We have found that Kir6.1/SUR2B channels with the C176A mutation are still sensitive to both K ATP channel opener and inhibitor, indicating that the general channel gating machinery of Kir6.1/SUR2B channel is still intact with the Cys 176 mutation.
In contrast to the conserved core domain, the N terminus of the Kir channel shows considerable variations (supplemental Fig. 3). The crystal structure of the N terminus of Kir channels cannot be modeled with a decent resolution. Therefore, we did not attempt to study how the S-glutathionylation of Cys 43 affects the protein conformation.
Unlike Cys 176 and Cys 43 , Cys 120 does not seem to be involved in S-glutathionylation. In the Kir2.1 channel, the mutation of Cys 122 , the counterpart of Cys 120 in Kir6.1, results in an absence of ionic currents although the channels are still expressed (58,59). In our studies, we were able to record the currents from the C120S mutant, although the currents were rather small compared with most of the other mutants. Based on the Kir protein structure, Cys 120 is located on the extracellular interface of the cellular membranes, accessible to extracellular environments. The accessibility of this residue to extracellular oxidants as well as membrane-impermeable PDS, but not intracellular GSSG, suggests that Cys 120 could be a site for extracellular redox modulation rather than intracellular S-glutathionylation.
In conclusion, our studies indicate that S-glutathionylation inhibits the Kir6.1/SUR2B channel by targeting mainly Cys 176 of the Kir6.1 subunit. S-Glutathionylation at Cys 176 is likely to structurally prevent the pore-forming inner helix from undergoing necessary conformational change for channel gating, thus retaining the channel in its closed state. The demonstration of the molecular mechanism underlying S-glutathionylation of the vascular K ATP channel with structural insight into the channel gating should have a profound impact on the understanding of the post-translational modifications of ion channels.