The Arabidopsis AtPP2CA Protein Phosphatase Inhibits the GORK K+ Efflux Channel and Exerts a Dominant Suppressive Effect on Phosphomimetic-activating Mutations*

The regulation of the GORK (Guard Cell Outward Rectifying) Shaker channel mediating a massive K+ efflux in Arabidopsis guard cells by the phosphatase AtPP2CA was investigated. Unlike the gork mutant, the atpp2ca mutants displayed a phenotype of reduced transpiration. We found that AtPP2CA interacts physically with GORK and inhibits GORK activity in Xenopus oocytes. Several amino acid substitutions in the AtPP2CA active site, including the dominant interfering G145D mutation, disrupted the GORK-AtPP2CA interaction, meaning that the native conformation of the AtPP2CA active site is required for the GORK-AtPP2CA interaction. Furthermore, two serines in the GORK ankyrin domain that mimic phosphorylation (Ser to Glu) or dephosphorylation (Ser to Ala) were mutated. Mutations mimicking phosphorylation led to a significant increase in GORK activity, whereas mutations mimicking dephosphorylation had no effect on GORK. In Xenopus oocytes, the interaction of AtPP2CA with “phosphorylated” or “dephosphorylated” GORK systematically led to inhibition of the channel to the same baseline level. Single-channel recordings indicated that the GORK S722E mutation increases the open probability of the channel in the absence, but not in the presence, of AtPP2CA. The dephosphorylation-independent inactivation mechanism of GORK by AtPP2CA is discussed in relation with well known conformational changes in animal Shaker-like channels that lead to channel opening and closing. In plants, PP2C activity would control the stomatal aperture by regulating both GORK and SLAC1, the two main channels required for stomatal closure.

assembly of four subunits encoded by a family of nine genes in Arabidopsis thaliana. A Shaker subunit comprises a short N-terminal sequence, six transmembrane segments (S1-S6), a pore domain between S5 and S6, and a large intracytoplasmic C-terminal region that comprises about two-thirds of the protein, including a C-linker involved in channel targeting and activity (23), a cyclic nucleotide-binding domain (CNBD), an ankyrin protein-protein interaction domain (absent in KAT1, KAT2, and AtKC1), and a K HA (HA for hydrophobic/acidic) domain (24). Depending on the subunits used, inwardly or outwardly rectifying channels can be formed (25,26). Most genes encoding subunits involved in inwardly rectifying channel assembly (KAT1, KAT2, AKT1, and AKT2) are expressed in guard cells (5,27,28), whereas GORK is the only outwardly rectifying channel in this cell type (29). Because GORK is activated by membrane depolarization, it is commonly thought that anion efflux initiates K ϩ efflux through this channel (30). In root peripheral cells, AKT1, together with AtKC1, is involved in K ϩ uptake from the soil. GORK is also located in these cells, although its role remains elusive. GORK may initiate plasma membrane repolarization after depolarization evoked by environmental changes (26,31).
Two Shaker channels, AKT1 and AKT2, have been reported to be regulated by PP2C phosphatases, and a direct interaction has been observed between the phosphatase AtPP2CA and AKT2 (32,33). Another closely related phosphatase, AIP1/ HAI2, binds and regulates AKT1 (34,35). The AKT1 model predicts that different calcineurin B-like interacting protein kinases (CIPKs) interact with the ankyrin domain to phosphorylate the channel (34,35), whereas PP2C phosphatases (including AtPP2CA) inactivate AKT1 by interacting specifically with some of these CIPKs. The direct effect of AIP1 on AKT1 activity was not addressed.
Inward channel regulation by PP2Cs is now well studied, but nothing is known about outward channel/PP2Cs relationships. Previous studies have shown that GORK and its SKOR (stelar K ϩ outward rectifier) homolog are sensitive to different signaling factors, such as reactive oxygen species (31,36,37), external K ϩ (26,38,39), and internal and external pH (38,40), and that clustering of GORK channels is correlated with changes in its gating properties (41). Regarding PP2Cs, early physiological studies in guard cells of Nicotiana benthamiana provided evidence that outward K ϩ fluxes were not driven only by depolarization and that PP2C phosphatases could take part in their control. Overexpression in N. benthamiana of the abi1-1 dominant interfering mutant phosphatase conferring ABA-independent constitutive activity led to a strong and specific decrease of the amplitude of guard cell outward K ϩ currents independent of the cell membrane potential and not observed with K ϩ inward and anionic currents (42). This indicated that the K ϩ efflux channel, later identified as GORK, was inhibited by ABI1-1 independent of the membrane potential and ABA signaling pathway.
In this study, we provide evidence for direct GORK inhibition by AtPP2CA. We also show a strong activation of GORK using phosphomimetic mutations and uncover an unexpected suppression of the activating effect of these mutations by co-expression with AtPP2CA.

Experimental Procedures
Plant Material-The atpp2ca-1 (SALK_028132) and atpp2ca-2 (SAIL_609_G12) mutants were obtained from the Salk Institute for Biological Studies (La Jolla, CA) (43). Homozygous plants were selected by PCR using primers surrounding the insertion site. Both insertions were located in the second exon.
Plant Transpiration Measurements-Plants were placed under short-day conditions in a growth chamber (21°C, 70% relative humidity, 8-/16-h light/dark, 300 E.m Ϫ2 .s Ϫ1 ) in lighttight pots filled with soil and sealed by a screw lid to prevent evaporation. One hole was drilled in the lid through which a single plant could grow (44). The plants were grown for 5 weeks (late rosette stage), during which they were watered periodically to keep the whole pot weight constant. The rosette surfaces were calculated using ImageJ software from pictures taken before starting the experiments. Transpiration was measured by weighing the plants at different time points, as indicated under "Results." Plants for excised rosette experiments were grown under the same conditions. The whole rosette was cut from the pot and weighed at different time points to define the water loss.
Two-hybrid System Experiments-The cDNA encoding the GORK C-terminal cytosolic domain (from Gly-313) and the AtPP2CA full-length ORF were inserted into pGBT9 (45) and pGAD10 (Clontech). Yeast transformation was performed according to Ref. 46. After transformation, yeast suspensions were plated on selective agar medium without leucine and tryptophan but containing histidine (47). The yeast strains used were Y190 for quantitative two-hybrid tests and AH109 for drop tests. Quantitative two-hybrid tests were performed using O-nitrophenol-␤-Dgalactopyranoside as a substrate (45). For drop tests, overnight liquid precultures (without tryptophan and leucine) were diluted 10 times in the same medium, grown for 6 h, and then washed with water. The optical density (600 nm) was adjusted to 0.2, and then serial dilutions (1/3, 1/9, 1/27, 1/81) were prepared before pouring drops on the selective agar medium.
Co-purification of AtPP2CA and GORK-AtPP2CA was expressed in Escherichia coli as a fusion protein with an N-terminal His 6 tag. The bacterial extract was loaded onto a minicolumn made of nickel-coated beads as described in Ref. 33. GORK was expressed in yeast (L40 strain) as a fusion with LexA (in the pBTM116 vector). Yeast cells were grown in 500 ml of liquid medium supplemented with amino acids but without tryptophan (47). The centrifugation pellet was resuspended in 5 ml of CelLytic Y cell lysis solution (Sigma) supplemented with dithiothreitol (5 mM), a protease inhibitor mixture (Sigma, catalog no. P8215, 1/100°v/v), leupeptin (10 M), and EDTA (1 mM). Glass beads (3.5 ml, 0.5 mm in diameter) were added to the yeast suspension. Yeast cells were broken by six cycles of 30 s of vortexing/1 min of chilling on ice. After centrifugation, an aliquot of the pellet was loaded onto the mini-column. Proteins were washed and eluted as described previously (33). Aliquots of the different purification steps (15 l) were loaded onto SDS-PAGE gels (10%). Proteins were then transferred onto a nitrocellulose filter and revealed by Western blotting with an anti-LexA antibody (rabbit polyclonal antibody (Invit-rogen; catalog no. 46-0710; lot no. 1200262; dilution, 1/5000) and goat anti-rabbit IgG/alkaline phosphatase conjugate. The second antibody and chemiluminescent substrate were provided in the Aurora Western blotting kit (ICN Biochemicals, catalog no. 821539, antibody lot no. 6230).
Electrophysiological Recordings-GORK, AtPP2CA, ABI2, and OST1 cDNAs were cloned in a vector derived from pGEMHE (48) between the 5Ј and 3Ј untranslated regions of the Xenopus ␤-globin gene. The GORK S722A, GORK S722E, and AtPP2CA point mutations were generated by PCR mutagenesis using sense and antisense mutated primers (Stratagene) and Pfu Ultra polymerase. In vitro transcription was performed on the linearized plasmid using the mMESSAGE mMACHINE T7 Ultra kit (Ambion). Stage V-VI oocytes were extracted from mature Xenopus laevis, and then the follicular cell layer was digested with 1 mg.ml Ϫ1 collagenase (type 1A, Sigma) for 1 h. Oocytes were injected with a final volume of 50 nl of cRNA preparation using a micropipette connected to an oocyte injector (Nanoliter 2000, WPI). Injected oocytes were maintained at 19°C for 3 or 4 days in ND-96 solution (2 mM KCl, 96 mM NaCl, 1 mM MgCl 2 , 1.8 mM CaCl 2 , 5 mM HEPES, and 2.5 mM sodium pyruvate (pH 7.4) supplemented with 5 g/ml gentamycin). Currents through the oocyte membrane were then recorded using the two-electrode voltage clamp technique as described previously (49). During measurements, oocytes were placed under continuous perfusion in a bath solution containing, unless stated otherwise, 10 mM KCl, 90 mM NaCl, 1 mM CaCl 2 , 1.5 mM MgCl 2 , and 10 mM HEPES (pH 7.4). The 100 mM KCl solution used for some experiments did not contain NaCl. For oocyte single-channel recordings, oocytes were peeled with tweezers and analyzed by patch clamp according to Ref. 40. The pipette solution contained 100 mM KCl, 3 mM MgCl 2 , 5 mM EGTA, and 10 mM Hepes-NaOH (pH 7.2). The bath solution contained 100 mM KCl, 3 mM MgCl 2 , 1 mM CaCl 2 , and 10 mM Hepes-NaOH (pH 7.4). All voltage pulse protocol designs, data acquisitions, and analyses were done using pCLAMP 9.0 software (Axon Instruments). Patch clamp experiments in COS cells (whole-cell mode) were performed as described previously (50), with 10 mM KCl in the bath solution and 150 mM KCl in the pipette solution.
Isolation of Oocyte Plasma Membranes and Western Blotting-Oocytes (150 -200 per assay) were injected with cRNAs and left for 3 days in ND-96 solution for protein expression. They were then crushed and homogenized at 4°C in 5 mM sodium phosphate (pH 7.4) buffer containing 5 mM MgCl 2 , 250 mM sucrose, 0.5 mM Pefablock, 10 g of leupeptin, 10 g of aprotinin, 10 g of pepstatin, and 60 g of chymostatin (homogenization buffer). Homogenates were centrifuged at 500 ϫ g for 5 min at 4°C. The supernatants were then overlaid on top of a discontinuous sucrose gradient (50% and 20% sucrose in homogenization buffer) and centrifuged at 4°C at 22,100 rpm (50,000 ϫ g) for 4 h in a swinging SW60Ti Beckman rotor. The 20% interface enriched in plasma membrane and the 50% interface enriched in endoplasmic reticulum (51) were collected, resuspended in 5 volumes of homogenization buffer without sucrose, and centrifuged at 4°C, 150,000 ϫ g for 2 h. The resulting membrane pellets were resuspended in homogenization buffer before protein determination by Bradford assay.
For Western blotting analyses, about 50 g of membrane proteins were solubilized with 2ϫ Laemmli solubilization buffer, and proteins were separated by SDS-PAGE and electrotransferred on a Millipore Immobilon-P PVDF 0.45-m membrane. GORK channels were then detected using a rabbit anti-GORK antibody (1/100) and goat antibodies (1/10,000) coupled to horseradish peroxidase. Polyclonal anti-GORK antiserum was raised against a GORK-specific synthetic peptide (amino acids 50 -61) and tested as described in Ref. 41. The conjugated antibodies were from Agrisera (catalog no. AS09 602, lot no. 1309).

Atpp2ca Mutants Display a Decreased Transpiration Phenotype Opposite to That of the Gork Mutant-Previous studies have
indicated that gork mutant plants display larger stomatal apertures than wild-type plants (29). The difference was slightly more pronounced after the light-to-dark transition (29). Whole plants were grown in soil to measure water loss in atpp2ca mutants using individual sealed pots (44). Under constant light, the transpiration rate was reduced slightly (by about 6%) in the atpp2ca mutants (Fig. 1A, left set of columns) compared with the wild type. The relative difference in transpiration rate between the wild type and the atpp2ca mutant plants increased after the light was switched off (Fig. 1A, center and right sets of columns; 17.3% and 12.6% reduction compared with the wild type during the first 15 min after light extinction and 20.9% and 11.9% between 15 and 45 min). This shows an increased efficiency of stomatal closing in the mutants compared with the wild type. Similarly, in measurements on excised rosettes, atpp2ca mutant plants lost less water than the wild type (Fig. 1B).
Physical Interaction between GORK and AtPP2CA-We next tested for a direct interaction between GORK and AtPP2CA with the yeast two-hybrid system. The C-terminal part of GORK downstream of the transmembrane domain (from Gly-313) and the full-length sequence of AtPP2CA were fused to the DNA-binding domain or activator domain of GAL4. A positive signal was detected using a ␤-galactosidase assay ( Fig. 2A), showing an interaction between GORK and AtPP2CA. These results were confirmed by growth assay using the HIS3 reporter gene (Fig. 2B), which proved to be the best assay for detecting interactions with GORK. His trap purification was then performed to confirm the interaction. A product of the expected size of about 80 kDa (22 kDa for the lexA DNA-binding domain and 57 kDa for the GORK C-terminal part) was retained specifically on a column bearing AtPP2CA fused to the His tag ( Fig.  2C). Therefore, the two-hybrid tests provided evidence for physical interactions between AtPP2CA and the GORK C-terminal domain, and using His trap purification gave further evidence to support these results.
The Effect of AtPP2CA on GORK Activity-In Xenopus oocytes, GORK expression resulted in outwardly rectifying currents displaying a typical sigmoidal activation profile, in agreement with previous functional analyses (29,38) (Fig. 3A). Coexpression of AtPP2CA with GORK resulted in inhibition of GORK-mediated currents (Fig. 3B) with a mean efficiency of ϳ45% (Fig. 3C). The current/voltage curves did not reveal a shift in activation potential upon inhibition. Inhibition of GORK by AtPP2CA was confirmed using COS cells (Fig. 3D). Therefore, using two independent heterologous systems, we showed evidence for GORK inhibition by AtPP2CA.
To evaluate the specificity of AtPP2CA, we tested another clade A PP2C, ABI2, which is located on the other branch of the phylogenetic tree (3). Like AtPP2CA, ABI2 inhibited GORK currents (Fig. 3E), suggesting that there is a redundancy in the action among the clade A PP2Cs.
AtPP2CA interacts with the OST1 kinase in the ABA signaling pathway (18). We therefore wanted to test this kinase against GORK. No activation of GORK currents was observed ( Fig. 3F). This is in agreement with the absence of an interaction between GORK and OST1 in previous bimolecular fluorescence complementation tests (19).
The Effect of AtPP2CA Single Mutations on the GORK/ AtPP2CA Interaction-We developed amino acid substitution mutations in the AtPP2CA sequence on the basis of mutations identified previously for their negative effect on phosphatase activity in other PP2C proteins. The G139D (G174D in ABI1) mutation of AtPP2CA, close to the DG (aspartate-glycine) metal-coordinating residues (DGHG motif), is commonly used (6,18,35). The G139D mutation inactivates AtPP2CA, but it does not interfere with its expression in yeast (18,35) or its interaction abilities with kinase partners (6,18,35). We found that this substitution mutation prevents GORK inhibition in oocytes (Fig. 4A). Interestingly, the interaction between AtPP2CA and FIGURE 2. Physical interaction between AtPP2CA and the C-terminal cytosolic region of GORK. A, quantitative two-hybrid ␤-galactosidase activity assays with AtPP2CA as the bait. The reciprocal interactions with GORK as the bait could not be assayed because of the low sensitivity of the test. ␤-Galactosidase activities (n ϭ 7, mean Ϯ S.D.) are expressed as A 420 per min ϫ 1000/A 600 of the culture medium. B, two-hybrid tests using HIS3 as the reporter gene in strain AH109. Drop tests (dilution series, from A 0,066 at 600 nm (1/3)) showing growth of cotransformed yeast cells on medium without tryptophan and leucine (selection of plasmids, control medium) or also without histidine (interaction tests). GORK was used as a positive control (strong GORK-GORK interaction because of subunit assembly via the C-terminal intracytoplasmic domains (25)). In this experiment and all subsequent drop test experiments, each series of dilutions was made in duplicate with two different yeast colonies, and tests were repeated with colonies from a different yeast transformation. DBD, DNA-binding domain of GAL4; ACT, activator domain. ACT alone indicates transformations with empty pGAD10 vector. C, co-purification on a His trap column. An E. coli extract containing either the His 6 tag alone, or the AtPP2CA-His 6 tag fusion protein was loaded onto nickel-coated beads (control column and AtPP2CA column, respectively). After washing, Saccharomyces cerevisiae extract containing lexA::CT-GORK protein (CT) was loaded on both columns. Unbound proteins were washed again, and aliquots (15 l) of the last wash fraction (W), the first elution (E1), and the second elution (E2) fractions were loaded on SDS-polyacrylamide gel. Right panel, the Western blot obtained from the gel, revealed with anti-LexA antibody. Left panel, Western blot analysis performed with an extract obtained from yeast expressing the LexA tag alone (yeast transformed with empty vector, EV).
GORK in the two-hybrid assay was disrupted by this mutation (Fig. 4B, third lane). To test whether the phosphatase activity of AtPP2CA is linked to its ability to interact with GORK, we tried other mutations leading to complete or substantial loss of PP2C activity ( Fig. 4B and Fig. 5). The dominant interfering G145D mutation in AtPP2CA (G180D or abi1-1 in ABI1, G168D in ABI2, and G246D in HAB1) is known to prevent interaction of PP2Cs with PYR/PYL receptors (13,14). In a two-hybrid assay, the G145D mutation also disrupted the interaction between AtPP2CA and GORK (Fig. 4B, fifth lane). We also tested the D142N, D327N, and D380N mutations in PP2CA because we knew that they target conserved residues in PP2Cs beyond the plant kingdom and correspond to human PP2C␣ amino acids (Asp-60, Asp-239, and Asp-282) that are essential for metal coordination and catalysis (52). The mutations A294F, P308L, R284C, and G287D in AtPP2CA, corresponding to mutations in ABI1 (Fig. 5) that were shown to reverse the abi1-1 mutation phenotype, were also tested in a two-hybrid assay. These four mutations have a strong negative effect on phosphatase-specific activity, but the R304C and G307D mutant forms in ABI1 retain significant residual activity (53). Finally, in two-hybrid assays against the C-terminal part of GORK, all mutations tested completely suppressed the interaction (Figs. 4B and Fig.  5).

Potential Phosphorylation Sites in the GORK C-terminal Part
Act as Molecular Switches Inactivated by AtPP2CA-To investigate the mechanism of AtPP2CA action in more detail, we targeted phosphorylation sites in the GORK protein sequence. In GORK, one phosphopeptide has been identified as a target for a plant kinase. The sequence SDFLKRLLSSGMNPN contains the LXRXXS motif common to many kinase sites (54) and has been shown to be phosphorylated by protein kinases in the CPK family (55). According to a model of ankyrin repeats published previously (56), the phosphorylatable serine (Ser-649) is at the end of the second helix of the fourth repeat in the GORK ankyrin domain. Another serine located just downstream of the ankyrin domain was also pinpointed as a possible phosphorylation site. This serine (Ser-722) is the only phosphorylatable residue in GORK predicted by the Netphos server that agrees with dephosphorylation sites for HAB1, a close relative of AtPP2CA (57). In the Arabidopsis Shaker channel family, the first site is specific to GORK (not predicted as a phosphorylation site in SKOR), whereas the second site is common to the two subunits forming the outwardly rectifying channels SKOR and GORK but is not present in other subunits. For each phosphorylation site, two mutations were designed: serine to alanine (mimicking the dephosphorylated state) and serine to glutamate (mimicking constitutive phosphorylation). We first aimed to determine whether the mutations have an effect on GORK activity and whether dephosphorylation of one of the two sites by AtPP2CA would explain the inhibition of GORK by the phosphatase.
A yeast two-hybrid assay was performed first with AtPP2CA and the S649E and S649A mutant forms of GORK. None of these mutations changed AtPP2CA binding to the GORK C-terminal region (Fig. 6A). We then evaluated the expression of the two mutant forms of GORK in Xenopus oocytes. The S649A GORK mimicked a constitutive dephosphorylated state and displayed the same current/voltage curve as the GORK wild-type form; it did not modify GORK activity in this system (Fig. 6B). Strikingly, the constitutively phosphorylated mimic mutant had significantly higher activity (Fig. 6C). This suggests that the constitutive phosphorylation of this residue increases GORK activity. These data support the hypothesis that the GORK channel is not phosphorylated in Xenopus oocytes at this site but can be activated by Ser-649 phosphorylation.
The AtPP2CA effect on the GORK mutants was also examined in Xenopus oocytes (Fig. 6, B and C). The S649A and S649E GORK mutants were not insensitive to inhibition by AtPP2CA. This confirms that GORK inhibition by AtPP2CA is not due to the dephosphorylation of the Ser-649 residue.
In this context, for the mutant S649E, we had expected a level of inhibition (Fig. 6D, a), as indicated by the white arrow. Interestingly, we observed a much higher level of inhibition of GORK activity (Fig. 6, C and D). The level of GORK S649E ϩ AtPP2CA activity was significantly lower than the theoretical estimate on the basis of the assumption that the inhibition would be proportional to GORK activity (Fig. 6D, gray arrow). In fact, the effect of AtPP2CA was dominant over that of the S649E mutation, which was suppressed completely (Fig. 6, C  and D). In summary, our data show that the levels of channel activity are similar for GORK ϩ AtPP2CA and GORK S649E ϩ AtPP2CA.
The effect of similar mimetic mutations in the Ser-722 residue of GORK was also investigated. Our two-hybrid results show that neither the GORK S722A nor the S772E mutations impaired the GORK/AtPP2CA interaction. Even changing Ser-772 to a stop codon did not prevent the interaction (Fig. 7A). The effects of the S722A and S722E mutations on GORK activity and AtPP2CA interaction were investigated in Xenopus oocytes in the same way as for the GORK S649 mutations. Again, we observed no difference between the S722A-mutated GORK channel and the native channel concerning both their activity and inhibition by AtPP2CA (Fig. 7B). The S722E mutation also increased channel activity even more than the corresponding substitution in Ser-649 (Fig. 7C). AtPP2CA also strongly reduced the activity of the GORK mutant (mean inhibition of 77% in seven experiments). Most importantly, the GORK S722E mutation systematically lost its effect in the presence of AtPP2CA (Fig. 7C). Because GORK activity and clustering depend on K ϩ external concentration, two different concentrations of 10 and 100 mM were tested in this system. Neither the increase of GORK activity because of the "E" mutation nor the effect of AtPP2CA on native or mutated GORK forms were found to depend on the K ϩ concentration of the medium (Fig. 7, B and C).
We next investigated the increased GORK activity seen in the experiments with the GORK S722E mutant. We tested for an increase of GORK expression or export to the membrane, which could lead to a change of transport efficiency at the oocyte membrane, using Western blotting of oocyte fractioned membranes to verify GORK S722E and S722A expression compared with the native GORK (Fig. 7D). All protein levels were the same under our experimental conditions. We next asked whether these mutations change GORK gating properties. This question was addressed by analyzing GORK activity at the single-channel level (Fig. 8). Our recordings indicated that the mutation did not change GORK unitary conductance, which was close to 20 pS (Fig. 8B). AtPP2CA, which reverses the effect of the mutation, also had no effect on the conductance of the mutant (Fig. 8B). Furthermore, the frequency of opening events was increased by the mutation, and this effect was inverted by co-expression with AtPP2CA (Fig. 8,  A and C). In conclusion, the S722E mutation increases the GORK opening frequency without changing its unitary conductance, whereas AtPP2CA neutralizes this effect.

Discussion
AtPP2CA, GORK, and Stomatal Closing-Because of the presence of multiple PP2Cs in guard cells, the transpiration phenotype of atpp2ca mutants is not predicted to be strong. Conversely, in the triple abi1 hab1 pp2ca loss-of-function mutant, only a low residual conductance has been measured (58). In this study, small but significant decreases of transpiration rates in atpp2ca plants compared with wild-type plants were detected. These differences increased after light/dark transition, which might be explained by a reduced pool of redundant, free active PP2Cs during that period (Fig. 9). The gork mutant displays an increase in transpiration rate compared with the wild type. This opposite phenotype, compared with atpp2ca mutants, is consistent with GORK inhibition by AtPP2CA. Because GORK activity can be inhibited by other close clade A PP2C members (Fig. 3E), our data are also consistent with published results of the inhibition of K ϩ outward currents in tobacco plants overexpressing the ABI1-1 protein, a dominant interfering mutant of ABI1 (42). SLAC1 and GORK can work together to control stomatal closure. SLAC1 initiates a flux of anions that depolarizes the membrane, and this triggers potassium efflux out of the guard cells through GORK. Interestingly, both SLAC1 and GORK are inhibited by AtPP2CA (18) and by other PP2Cs, such as ABI1 and/or ABI2 (19). The transpiration phenotype of the atpp2ca mutants is consistent with a co-regulation of GORK and SLAC1 by AtPP2CA. Therefore, PP2Cs plays a central role in the con-trol of stomatal closure by regulating the activities of the three major actors of this process: SnRK2.6/OST1 kinase (16,17), anion channels (18), and K ϩ channels. Because of its regulation by PP2Cs (Figs. 3 and 6 -8), GORK gating is not only the consequence of anion efflux and membrane depolarization. GORK inhibition by PP2Cs also helps to slow down the stomatal closure rate and, perhaps, maintain depolarization during that process as well.
The Relationship between the Activity of AtPP2CA and Its Effect on GORK-PP2Cs do not always need to dephosphorylate their substrates to inhibit their function. A mechanism of OST1 inhibition by physical interaction with phosphatase has been described for AtPP2CA (34), and the inhibition of CIPKs by PP2Cs is also thought to involve both physical interaction and dephosphorylation of the kinases (35). The most studied example of a PP2C/substrate interaction is between HAB1 and OST1. On the basis of the use of an inactive mutant of the HAB1 phosphatase, a two-step mechanism has been proposed for inhibition of OST1 by HAB1. At low concentrations of HAB1, OST1 is inhibited by dephosphorylation of its Ser-175 residue located within its activation loop anchored into the HAB1 catalytic site. If the phosphatase is in excess, then a second inhibition mechanism by physical interaction and mutual packing of the HAB1 and OST1 catalytic sites takes place (16). This mechanism involves a conserved tryptophan in the PYL interaction loop of HAB1 that blocks access of the substrate to the kinase (16). The active site of PP2Cs and the PYL interaction loop are also involved in the interaction with PYR/PYL receptors (16) (Fig. 5). Mutations that inactivate PP2Cs have been used to address their mechanism of action (16,18,35,52,59). The mutants we designed reproduce amino acid substitutions in critical residues characterized in AtPP2CA or homologous PP2Cs. Mutant proteins could be expressed in protoplasts or heterologous systems for kinetic analyses (6,52,53), and they often retain residual activity (6,52,53,57). The dominant interfering mutant forms in ABI1 and HAB1 (equivalent to G145D in AtPP2CA) retain 2.5% and 4% of the activity against casein as a substrate (53,60), but HAB1 G246D only displays a 2-to 3 times reduction of OST1 inactivation (57). The AtPP2CA G139D G145D double mutant, which has completely lost its PP2C activity (18), retains the ability to bind OST1 (18) and CIPK6 (35) in yeast. It also efficiently inhibits the autophosphorylation of OST1 in vitro (18) and the activation of AKT1 by CIPK6 in Xenopus oocytes (35). Surprisingly, all nine mutations tested in AtPP2CA disrupted the interaction with the GORK C-terminal domain (Fig. 4B). However, it is unlikely that these results are due to all AtPP2CA mutants having lost the ability to interact with GORK simply because of a drastic reduction in expression or conformational change. These mutations target sites in conserved domains that are either involved directly in the catalytic activity of the phosphatase (G139D, D142N, G145D, D327N,  and D380N) or close to residues forming the catalytic site FIGURE 6. Effect of mutations on Ser-649 on GORK activity and inhibition by AtPP2CA. A, two-hybrid interaction assay between GORK (or GORK mutated on serine 649) and AtPP2CA. Yeast cells were grown on medium without tryptophan, leucine, and histidine. Details of the experimental procedures are in the legend for Fig. 2B. DBD, DNA-binding domain; ACT, activator domain. B, current/voltage curves with oocytes expressing GORK alone, GORK with AtPP2CA, GORK S649A alone, or GORK S649A with AtPP2CA (n Ն 10). The holding potential is Ϫ60 mV. Applied voltages increase from Ϫ90 to ϩ50 mV with an increment of ϩ10 mV. C, the same experiment as in B with GORK S649E instead of 649A (n Ն 12). B and C, oocytes were injected with 10 ng of GORK cRNA and 20 ng of AtPP2CA cRNA. D, quantitative aspects of the experiments in B and C (1 and 2, respectively). Data were normalized with the level of GORK current in each experiment. a, the basic level of inhibition of GORK by AtPP2CA; b, the increase of activity because of the mutation S649E. The white and gray arrows point, respectively, to the expected levels of GORK S649E ϩ AtPP2CA if the effect of AtPP2CA on GORK were completely independent on the activation of the Ser-649 site or if it were proportional to the activity of the channel. The difference between the level of the gray arrow (0.87 Ϯ 0.05) and that of GORK S649E ϩ AtPP2CA (0.62 Ϯ 0.05) is statistically significant (Student's t test, p ϭ 0.002). In reality, the increase of inhibition (b') is roughly equivalent to b, highlighting a complete suppression of the positive effect of the mutation. Data are mean Ϯ S.E.
(R284C, G287D, A294F, and P308L in this study) (53) (Fig. 5). Arg-284 and Gly-287 are located in the PYL interaction loop near the conserved tryptophan (Fig. 5). The total absence of interaction between GORK and all AtPP2CA mutants we tested strongly suggests that the proper conformations of the active site and the PYL interaction loop are required for GORK/ AtPP2CA interaction and GORK inhibition. The difference between OST1 and GORK in their ability to bind AtPP2CA mutants is likely due to a second binding site for PP2Cs in the ABA box of OST1 (16). Mutations in several OST1-interacting residues in HAB1 are inefficient at disrupting OST1-HAB1 binding unless they are associated with a mutation in the C-terminal end of HAB1 that is required for interaction with the ABA box of OST1 (16).
Phosphomimetic Mutations Support a Regulation of GORK by Phosphorylation-The S649E and S722E mutations in GORK increase its activity (Figs. 6 and 7). This strongly suggests that these serine residues can be phosphorylated in plant cells to increase the activity of the channel. The phosphorylation by plant CPKs of serine 649 in a peptide derived from GORK sequence, reported previously (55), further supports this idea.
Kinases that phosphorylate GORK in plants remain to be identified. Although AtPP2CA interacts physically with OST1 (18), our results indicate that this kinase is not involved directly in GORK activation (Fig. 3F) like it is for SLAC1 (18). Other kinases, such as OST1-related kinases, CIPKs, and CPKs, which regulate PP2C phosphatases and/or the anion channel SLAC1 and its SLAH3 homolog (21,22,34,35,61), would be good candidates for other regulators of GORK phosphorylation.
Inhibition of GORK by AtPP2CA Uncovers a Dominant Effect-Our results indicate that Ser-649 and Ser-722 are important sites for GORK activity. We found that GORK inhibition can occur in the absence of dephosphorylation of the two serines, which means that Ser-649 and Ser-722 are not good candidates for dephosphorylation by AtPP2CA in Xenopus oocytes. Instead, these results reveal an additional mechanism involving the interplay between AtPP2CA and the two serines, as described below.
Indeed, two routes of inhibition of GORK by AtPP2CA can be distinguished. The first corresponds to the inhibition observed with native GORK in oocytes (Fig. 6D, a). Mutations leading to partial or complete loss of AtPP2CA activity did not allow discrimination between inhibition by dephosphorylation or physical interaction involving the active site without dephosphorylation. A phosphorylation-dependent mechanism would suppose that endogenous oocyte kinases are able to phosphorylate GORK. These kinases can phosphorylate a limited number of commercial exogenous animal protein substrates (62) and potassium leak channels (63) and modulate the activity of plant transport systems such as aquaporins (64,65), sucrose transporters (66), and the AKT2 K ϩ channel (67).
The second level of inhibition (Fig. 6D, b), which was studied in more detail, is observed with E mutants mimicking phosphorylation states of serines in the C-terminal domain of GORK. This mechanism does not involve dephosphorylation of Ser-649 and Ser-722. It is interesting to note that the two mutated serines behaved similarly, suggesting that other phosphorylation sites might also be concerned. Our results suggest that GORK Ser-649 and Ser-722 are not phosphorylated in oocytes. The extra activity of GORK because of the E mutations was suppressed systematically by AtPP2CA, whatever the percent- age of native GORK inhibition in the same experiment (Figs. 6, C and D, and 7C). This suppression of GORK activation is therefore dominant over the effects of the mutations. This might not be restricted to the interaction with GORK. It is worth noting that SLAC1 and its SLAH3 homolog systematically exhibit a dominance of the effect of the phosphatases versus kinases, resulting in a complete or almost complete suppression of SLAC1 and SLAH3 kinase activation in the presence of PP2Cs, even with equal amounts of injected phosphatase and kinase cRNAs (19 -22). When significant SLAC1 activity was detected in Xenopus oocytes in the absence of kinase, it was possible to measure direct SLAC1 inhibition by AtPP2CA (18). Interestingly, the level of SLAC1 activity after inhibition by the phosphatase was the same in the absence and in the presence of the OST1 kinase that is known to activate SLAC1 (SLAC1 ϩ OST1 ϩ AtPP2CA ϭ SLAC1 ϩ AtPP2CA) (18).
The activation of GORK by the S722E mutation and its inactivation by AtPP2CA do not change the channel conductance but modulate the number and/or duration of opening events (Fig. 8). Because the mutation does not change the amount of GORK protein at the plasma membrane, this increase in open states cannot be due to an increase in the number of GORK channels. The conclusion is that the GORK opening frequency is increased by the mutation and decreased reversibly by AtPP2CA. In Shaker channels, voltage gating is driven by the interplay between the voltagesensing module (S1-S4 segments) and the pore-forming module (S5-pore-S6) (24). Some regions of the C-terminal region of KAT1 and KAT2 Shaker channels (downstream of S6) are also important for channel activity and gating, such as the C-linker (23,68) and the CNBD (68). The increase in channel activity by mutations in or in the vicinity of the ankyrin domain (absent in KAT1, KAT2, and AtKC1) underscores the role of this protein-protein interaction domain, which is located far in the peptide sequence from the transmembrane domains (about 200 amino acid down-stream of S6). With respect to GORK gating and regulation, these results suggest a connection between the ankyrin domain and the transmembrane segments. This is in agreement with studies performed with the closest animal homologs of plant Shaker channels (HCN (hyperpolarization-activated cyclic nucleotide-gated), CNG (cyclic nucleotide-gated), and KCNH (K ϩ voltage-gated channel, subfamily H) families), demonstrating that the C-terminal cytosolic region is mediating channel gating. The first two domains after the S6 segment, the C-linker and the CNBD, play a crucial role. Changes in the conformation of the CNBD, either by a cyclic nucleotide ligand (69), a short intrinsic ␤ strand of this domain (70), or mutations in residues of the C-linker and CNBD (71), result in a modulation of channel gating properties (69,70,71), and these changes in conformation of the C-linker are coupled to pore opening (69,72). Similarly, in plant Shaker channels, the first helix in the C-linker has been found to be essential for channel activity (23). The ankyrin domain, just downstream from the CNBD, is specific to plant Shaker channels, and its function in channel gating is currently unknown. The changes we observed in GORK gating properties with mutations around the ankyrin domain and their reversibility by AtPP2CA binding suggest the possibility that conformational constraints of this domain can be transmitted to the other C-terminal domains in the channel to modulate its gating.
In conclusion, this study reveals that GORK is regulated by PP2C protein phosphatases, that AtPP2CA active site integrity is likely to be required for the interaction, and that phosphomimetic mutations in or close to the ankyrin domain lead to an activation of the channel that is suppressed by AtPP2CA, suggesting an involvement of the ankyrin domain in channel gating. One of the most surprising observations was the dominant effect of the phosphatase over the activation conferred by phosphomimetic mutations. Such a mechanism of phosphorylationdependent inactivation without dephosphorylation of these sites would help GORK respond in real time to environmental constraints because it allows rapid and cost-effective regulation at multiple sites within the channel.