Phosphorylation of the ATP-sensitive, inwardly rectifying K+ channel, ROMK, by cyclic AMP-dependent protein kinase.

Activity of the recently cloned ATP-sensitive epithelial K+ channel, ROMK (Ho, K., Nichols, C. G., Lederer, W. J., Lytton, J., Vassilev, P. M., Kanazirska, M. V., and Hebert, S. C. (1993) Nature 362, 31-38), is regulated by phosphorylation-dephosphorylation processes with cAMP-dependent protein kinase (PKA)-dependent phosphorylation events being required for maintenance of channel activity in excised membrane patches (McNicholas, C. M., Wang, W., Ho, K., Hebert, S. C., and Giebisch, G. (1994) Proc. Natl. Acad. Sci. U.S.A. 91, 8077-8081; Kubokawa, M., McNicholas, C. M., Higgins, M. A., Wang, W., and Giebisch, G. (1995) Am. J. Physiol. 269, F355-F362). To determine whether this channel is a substrate for PKA, ROMK tagged with the hemagglutinin epitope was transiently transfected into HEK293 cells. In vitro labeling of immunoprecipitated proteins from transfected cells showed that ROMK could be phosphorylated by PKA. Metabolic labeling of ROMK resulted in a significantly increased phosphorylation upon pretreatment of the cells with forskolin, consistent with an action of cAMP-dependent protein kinase. Phosphoamino acid analyses of the ROMK phosphoproteins revealed that phosphate was attached exclusively to serine residues. Three putative PKA phosphorylation sites containing serine residues in the predicted ROMK proteins are shown directly to be substrates for PKA. Site-directed mutagenesis of each of these sites or double mutation of any two sites showed that ROMK proteins retained the ability to be phosphorylated by PKA both in vivo and in vitro to a variable extent, while triple mutation of all three PKA sites abolished the phosphorylation induced by cAMP agonists in transfected cells. Two-electrode voltage clamp experiments showed that PKA-dependent phosphorylation was required for ROMK channel activity and that at least two of the three sites were required for channel function when expressed in X. laevis oocytes. Taken together, these results provide strong evidence that direct phosphorylation of the channel polypeptide by PKA is involved in channel regulation and PKA-dependent phosphorylation is essential for ROMK channel activity.

To determine whether this channel is a substrate for PKA, ROMK tagged with the hemagglutinin epitope was transiently transfected into HEK293 cells. In vitro labeling of immunoprecipitated proteins from transfected cells showed that ROMK could be phosphorylated by PKA. Metabolic labeling of ROMK resulted in a significantly increased phosphorylation upon pretreatment of the cells with forskolin, consistent with an action of cAMP-dependent protein kinase. Phosphoamino acid analyses of the ROMK phosphoproteins revealed that phosphate was attached exclusively to serine residues. Three putative PKA phosphorylation sites containing serine residues in the predicted ROMK proteins are shown directly to be substrates for PKA. Site-directed mutagenesis of each of these sites or double mutation of any two sites showed that ROMK proteins retained the ability to be phosphorylated by PKA both in vivo and in vitro to a variable extent, while triple mutation of all three PKA sites abolished the phosphorylation induced by cAMP agonists in transfected cells. Twoelectrode voltage clamp experiments showed that PKAdependent phosphorylation was required for ROMK channel activity and that at least two of the three sites were required for channel function when expressed in X. laevis oocytes. Taken together, these results provide strong evidence that direct phosphorylation of the channel polypeptide by PKA is involved in channel regulation and PKA-dependent phosphorylation is essential for ROMK channel activity.
ATP-sensitive (K ATP ) potassium channels have been identified in apical membranes of several renal epithelial cells, where they are in a position to play critical roles in mediating and regulating K ϩ secretion (Misler and Giebisch, 1992). These epithelial secretory K ATP channels are characterized by a high open probability, inward rectification, exquisite pH sensitivity, and inhibition by cytosolic ATP (Misler and Giebisch, 1992;Wang et al., 1992). The low conductance (25-35 pS) 1 K ATP channel in apical membranes of thick ascending limbs of Henle is critical to NaCl absorption, as it ensures that adequate luminal potassium is provided for efficient function of the Na ϩ : K ϩ :Cl Ϫ cotransporter (Hebert and Andreoli, 1984;Wang, 1994b). A similar apical K ATP channel has been identified in principal cells in the cortical collecting duct, where it facilitates potassium secretion (Frindt and Palmer, 1987;Wang et al., 1990;Misler and Giebisch, 1992).
An inwardly rectifying, ATP-regulated K ϩ channel, ROMK1 was recently cloned from the outer medulla of rat kidney (Ho et al., 1993). ROMK, along with other subsequently identified K ϩ channel genes (Dascal et al., 1993;Kubo et al., 1993aKubo et al., , 1993bAshford et al., 1994;Suzuki et al., 1994;Zhou et al., 1994. Takumi et al., 1995 define a new family of inward rectifying K ϩ channels. The inward rectifying K ϩ channel protein contains an H5-like "pore-forming" region related to the voltagegated K ϩ channels and exhibits a characteristic topology featuring only two potential membrane-spanning segments. ROMK1 channels expressed in Xenopus laevis oocytes display properties similar to those of the low conductance K ATP channels identified in renal epithelia (Ho et al., 1993;Nichols et al., 1994;McNicholas et al., 1994;Kubokawa et al., 1995). Recently, splice variants of ROMK1 denoted ROMK1-3 have been identified (Zhou et al., 1994;Boim et al., 1995), which display alternative splicing at the 5Ј end and give rise to channel proteins differing in their amino-terminal amino acid sequences. These isoforms are differentially expressed along the loop of Henle and distal nephron in the kidney; functional expression in X. laevis oocytes showed that they all form functional Ba 2ϩ -sensitive K ϩ channel (Boim et al., 1995).
Both the secretory K ATP channel in renal epithelia Giebisch, 1991a, 1991b) and ROMK channels expressed in X. laevis oocytes are regulated by phosphorylation and dephosphorylation processes, with activation of channel activity by cAMP-dependent protein kinase (PKA) (McNicholas et al., 1994). The predicted ROMK channel protein contains three PKA consensus phosphorylation sites, suggesting that ROMK may be a substrate for PKA and that direct phosphorylation of the channel polypeptide may play a role in channel regulation by this serine-threonine kinase. Phosphorylation of specific amino acid residues on other ion channels is one mechanism of regulating channel properties (for review, see Levitan (1988Levitan ( , 1994. Thus, in the present report we investigated whether the modulation of ROMK channel activity by PKA is associated with direct phosphorylation of the ROMK channel polypeptide. A functional HA-tagged ROMK1 channel cDNA construct was transiently expressed in HEK293 cells and expression confirmed biochemically. We observed phosphorylation of ROMK1 protein both in vivo and in vitro, as indicated by PKA-induced [ 32 P]phosphate ( 32 P i ) incorporation. We also found that phosphate is attached exclusively to serine residues and the extent of 32 P i incorporation is enhanced by preincubation of cells with forskolin. Site-directed mutagenesis, coupled with phosphopeptide mapping, identified three serine phosphorylation sites in ROMK2. Expression of these serine mutants of ROMK2 channels in X. laevis oocytes showed that at least two sites were required for channel function. Mutation of all three PKA phosphorylation sites rendered the channel inactive and abolished the phosphorylation of channel protein induced by cAMP agonists in transfected HEK293 cells.

MATERIALS AND METHODS
Construction of HA-tagged ROMK-The HA epitope of influenza virus hemagglutinin (Wilson et al., 1984;Meloche et al., 1992) was introduced into the 3Ј end of the ROMK coding region by polymerase chain reaction (PCR) using ROMK1/pSPORT and ROMK2/pSPORT as templates with primers P-1 (TGACCGTGTTCATCACAGC) and P-HA (TCAGCTAGCTAAGCATAATCAGGAACATCATAAGGATACATCTG-GGTGTCGTCCG). Amplification was performed on a PTC-100™ programmable controller (MJ Research, Inc.) using the cycling parameters as described previously (Boim et al., 1995). The PCR products were digested with MscI and NheI and subcloned into MscI and NheI digested wild-type ROMK to generate the HA-tagged ROMK (ROMK-HA/ pSPORT). The MscI-NheI fragment containing HA was also subcloned into both ROMK1/pSVL and ROMK2/pSVL for transient transfection. Fig. 1A shows a schematic representation of the 3Ј HA epitope-tagged ROMK1 cDNA construct. Sequencing of the ROMK-HA/pSPORT constructs (Sequenase™ kit, version 2.0, U. S. Biochemical Corp.) demonstrated that the insert was free of mutations and in-frame.
Site-directed Mutagenesis of ROMK2 PKA Sites-Site-directed mutagenesis of ROMK2 PKA sites was performed in ROMK2/pSPORT according to method of Kunkel (1985). For single mutations, which modified ROMK2 by mutating individually the three potential PKA phosphorylation sites, three primers were used to change the indicated serine to alanine: S25A, TCTTCCTTCTTTGGCCACCAGCCTTG; S200A, CTGCCAATCAGTAAGGCCTTCCTAAGAT; and S294A, CCTCTGGGACATATGCCGTGCGGACC. Three silent restriction sites (S25A, MscI; S200A, StuI; S294A, NdeI) generated by these primers were used to confirm subcloning. Mutants containing double or triple changes were accomplished by PCR using the relevant single mutant ROMK2 as template and subcloned into the single or double mutant ROMK2. For expression in HEK293 cells, the ROMK cDNA mutants were transferred into the eukaryotic expression vector ROMK2/pSVL and tagged with the HA epitope. All mutants were verified by DNA sequencing.
K ϩ Currents Recorded by Two-electrode Voltage Clamp-Ba 2ϩ sensitive K ϩ currents in X. laevis oocytes injected with wild-type, HA-tagged, and mutant forms of ROMK cRNA were functionally examined by two-electrode voltage clamping as described (Ho et al., 1993;Boim et al., 1995). Oocytes were injected with 50 nl of cRNA (0.5 g/l) and treated with collagenase (2 mg/ml) for 2 h at room temperature. Electrophysiological recordings were performed 36 -72 h later at 22 Ϯ 2°C by two-electrode voltage clamp at a holding potential of Ϫ80 mV (Axoclamp-2A, Axon Instruments); the bath solution was 48 mM NaCl, 50 mM KCl, 0.3 mM CaCl 2 , 1 mM MgCl 2 , 5 mM HEPES, with or without 5 or 10 mM Ba 2ϩ (pH 8.0).
In Vitro Translation of ROMK cDNA-Wild-type or mutant ROMK cDNA constructs were translated in vitro using TNT™-coupled Reticulocyte Lysate Systems and [ 35 S]methionine as per manufacturer's instruction (Promega). ROMK cDNA was linearized for this purpose with NotI. Reaction mixtures were incubated at 30°C for 120 min after addition of [ 35 S]methionine. Protein products were resolved by 8% Laemmli SDS-polyacrylamide gel electrophoresis (Laemmli, 1970) and the [ 35 S]methionine-labeled ROMK proteins were visualized by autoradiography.
Transient Expression of Wild-type and Mutant Forms of ROMK in HEK293 Cells-HEK293 cells were transfected with wild-type or mutant ROMK cDNA constructs using LipofectAMINE™ according to the manufacturer's instruction (Life Technologies, Inc.). The transfected cells were analyzed 50 h after initial transfection. DNA-free transfection (untransfected cells) underwent each of the transfection steps without addition of cDNA. Metabolic labeling was performed using 200 Ci/ml 32 P i for 4 h in a phosphate-free Dulbecco's modified Eagle's medium in the presence of 50 M sodium orthovanadate. To activate endogenous adenylate cyclases and cyclic AMP production, transfected cells were treated with 20 M forskolin and 100 M 3-isobutyl-1-methylxanthine (IBMX; Sigma) for 15 min prior to cell lysis. For cell-surface biotinylation, intact cells were incubated with 0.1 M NHS-biotin (Sigma) in phosphate-buffered saline at room temperature for 30 min (Gottardi et al., 1995).
Plasma Membrane Isolation and Quantitative Immunoblotting-Crude membranes (plasma and cytosolic membranes) from transiently transfected HEK293 cells were isolated according to Haas et al. (1991). The resulting membranes were suspended in 30 l of sucrose-histidine buffer and stored at Ϫ70°C. Protein concentration of the suspensions were determined by the method of Lowry et al. (1951). Aliquots of membrane extracts were separated in adjacent lanes on 8% SDS-polyacrylamide gel and electrophoretically transferred to polyvinylidene fluoride membranes (Bio-Rad). The blots were incubated with monoclonal anti-HA antibody (clone 12CA5, Boehringer Mannheim) followed by peroxidase-conjugated anti-mouse antibody (Amersham Corp.) and visualized with enhanced chemiluminescence (ECL, Amersham Corp.).
Immunoprecipitation and Biochemical Analysis of ROMK-Transiently transfected cells labeled with biotin were washed twice with ice-cold phosphate-buffered saline buffer and lysed in a lysis buffer containing 20 mM Tris/HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40 (v/v). For metabolic labeling, the lysis buffer was supplemented with 10 mM sodium pyrophosphate and 1 mM sodium orthovanadate. Cells were lysed at 4°C for 20 min, followed by centrifugation at 10,000 ϫ g for 15 min to remove cell debris. The ROMK proteins were immunoprecipitated by incubation of the supernatants with anti-HA monoclonal antibody for 4 h and with protein A-Sepharose for overnight at 4°C. For in vitro phosphorylation of ROMK, immunoprecipitates were incubated with 20 ng of PKA catalytic subunit (Sigma) in 50 l of kinase buffer (50 mM Tris/HCl, pH 7.5, 10 mM MgCl 2 , 100 mM bovine serum albumin) at 30°C for 60 min in the presence of 10 Ci of [␥-32 P]ATP (Cheng et al., 1991). The reactions of biochemical labeling were stopped by washing the protein A beads once with RIPA buffer (50 mM Tris/HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS) and twice with RIPA buffer containing 1 M NaCl. The ROMK protein was recovered by incubation of protein A with 2 ϫ Laemmli SDS sample buffer at 55°C for 15 min and separated by 10% SDS-polyacrylamide gel. Phosphoproteins resulting from metabolic and in vitro labeling were analyzed by PhosphorImager with the ImageQuant™ program (Molecular Dynamics).
Phosphoamino Acid Analysis-Phosphoamino acid analysis of in vivo or in vitro phosphorylated ROMK was performed as described by Hunter and Sefton (1980) and Huganir et al. (1984). The ROMK proteins were recovered from SDS-polyacrylamide gels and acid hydrolyzed in 4 N HCl under vacuum at 110°C for 3 h. The acid hydrolysates were suspended in a marker mixture containing 15 g each phosphoserine, phosphothreonine, and phosphotyrosine. The samples were spotted 4 cm from the right edge on 20 ϫ 20-cm thin-layer chromatography plates (Kodak) and electrophoresed at 375 V for 90 min in 1% pyridine, 10% acetic acid (v/v) buffer, pH 3.5. The plates were air-dried, and the markers were visualized with ninhydrin (Sigma).
Phosphopeptide Mapping-Two-dimensional tryptic phosphopeptide map analysis was performed as described by Hunter and Sefton (1980) and Cheng et al. (1991). Recovered proteins were treated with L-1tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin at a final concentration of 0.15 mg/ml for 12 h at 37°C. Tryptic digests were resolved in two dimensions with electrophoresis in 1% ammonium carbonate, pH 8.9, and chromatography at pH 3.5 in the chromatography solvent containing n-butanol/glacial acetic acid/pyridine/H 2 O (v/v: 75:15:50:60).

RESULTS
Epitope Tagging of ROMK Protein-Since our antibodies to the wild-type ROMK channel protein are not immunoprecipitating, we tagged ROMK1 and ROMK2 channels with the influenza virus hemagglutinin epitope, HA, at the carboxyl termini (Fig. 1A). This allowed for immunoprecipitation of ROMK protein with anti-HA monoclonal antibody (Meloche et al., 1992). Fig. 1B shows that in vitro translation of equal amounts of cDNA from the untagged (lane 1) and HA-tagged (lane 2) ROMK1 constructs yielded bands of similar intensity and at the expected apparent molecular mass (45 kDa) of the wild-type ROMK1 translation product (Ho et al., 1993). The ϳ1.1-kDa increase in molecular mass for the HA-tagged ROMK1 protein was not resolved on this gel. Immunoprecipitation of the in vitro translated proteins with anti-HA antibody identified a 45-kDa band only for the HA-tagged construct (Fig.  1B, lanes 3 and 4). Thus, the HA tag has no effect on translation of ROMK cDNA in vitro, and the anti-HA antibody immunoprecipitates only the tagged ROMK protein.
Functional Expression of HA-tagged ROMK1 in X. laevis Oocytes-To determine whether the added epitope would disrupt ROMK1 protein targeting and/or function, we evaluated Ba 2ϩ -sensitive K ϩ currents (I K(Ba) ) by two-electrode voltage clamping in X. laevis oocytes injected with cRNA transcribed from the ROMK1-HA cDNA. The HA-tagged ROMK1 cRNAinjected oocytes exhibited large Ba 2ϩ -sensitive, tetraethylammonium-insensitive currents ( Fig. 2A), consistent with the properties of ROMK1 (Ho et al., 1993). Current-voltage relationships were inwardly rectifying at low external K ϩ concentrations ([K ϩ ] o ) and linear at high [K ϩ ] o (Fig. 2B), typical of this epithelial K ATP channel (Nichols et al., 1994). Moreover, the reversal potential (E rev ) was Nernstian with [K ϩ ] o (Fig. 2C), consistent with a high potassium selectivity. These electrophysiological data demonstrate that the ROMK1-HA cDNA encodes a product whose functional properties are indistinguishable from those of the wild-type ROMK1 channel (Ho et al., 1993). Thus, tagging with the HA epitope at the carboxyl terminus of ROMK1 does not significantly alter the basic properties of this K ϩ channel in oocytes.
Heterologous Expression of the ROMK1 Protein in HEK293 Cells-Although the physiological studies shown in Fig. 2 demonstrate that the HA-tagged ROMK functions as a K ϩ channel in X. laevis oocytes, we wished to study the phosphorylation of the HA-tagged ROMK protein in a mammalian cell line. Thus, we also assessed whether the HA-tagged channel protein could be expressed at the plasma membrane of HEK293 cells. First, we were able to detect the 45-kDa ROMK1-HA protein in crude membrane preparations by Western blot (Fig. 3A). Furthermore, we used biotin, which specifically labels plasma mem-brane proteins, to assess the cell-surface expression of ROMK1-HA. Fig. 3B shows that the 45-kDa ROMK1-HA protein was detected by cell-surface labeling with biotin in transiently transfected, but not untransfected, HEK293 cells. These experiments demonstrate that the ROMK1-HA channel protein was both synthesized and transported to the plasma membrane of these mammalian cells. Phosphorylation of the ROMK1 Protein by PKA-We next determined whether the expressed ROMK1 protein could be phosphorylated either in vitro or in vivo. For in vitro phosphorylation, HA-tagged ROMK proteins expressed in transfected HEK293 cells were immunoprecipitated with anti-HA antibody and the resultant immunoprecipitates were labeled by the catalytic subunit of PKA in the presence of [␥-32 P]ATP. As shown in Fig. 4A, a 45-kDa ROMK1 phosphoprotein was detected in transfected cells and not in untransfected control cells. In addition, no 45-kDa phosphorylation was observed in the absence of PKA under these conditions used for this assay (data not shown). These data suggest that at least in the in vitro conditions, the ROMK protein is a substrate for PKA. However, phosphorylation of the detergent-denatured ROMK1 by exogenous protein kinases may not reflect phosphorylation of ROMK in intact cells. Thus, we determined whether the ROMK protein could be phosphorylated in vivo by activation of endogenous PKA in HEK293 cells. Fig. 4C shows that when transiently transfected HEK293 cells were incubated with 32 P i prior to cell lysis and immunoprecipitation, a basal level of phosphorylation of the ROMK proteins was observed in transfected cells and the amount of labeling increased markedly upon treatment of the cells with forskolin and IBMX, consistent with an action of endogenous, cAMP-dependent protein kinase. The weak phosphorylation detected in non-forskolin treated cells transfected with the HA-tagged construct may represent a low level of endogenous adenylate cyclase activity present in cells exposed to fetal bovine serum in the culture medium. This is supported by the finding that culturing HEK293 cells in serum-free medium 12 h prior to in vivo labeling significantly decreased the basal level of phosphorylation (data not shown). The basal level of labeling may also be due to PKC-dependent phosphorylation, since we observed phosphorylation of the transfected ROMK proteins by PKC (data not shown). Several lighter bands were observed in the forskolin-treated sample, which may represent either the glycosylated ROMK or nonspecific labeling of other membrane proteins. These latter bands were also present in the forskolintreated ROMK2 shown in Fig. 7. Equal amounts of protein were resolved by SDS-polyacrylamide gel electrophoresis, and the 32 P i -incorporation in the 45-kDa band was measured by a PhosphorImager. This showed an approximately 3-fold increase in the phosphate labeling after forskolin treatment.
Previous investigations have shown that phosphorylation by PKA may promote assembly, and insertion, of mature proteins to the plasma membrane Ross et al., 1991;Levin et al., 1995). Thus, we examined whether the magnitude of surface biotinylation of ROMK1 proteins expressed in HEK293 cells was altered by forskolin and IBMX treatment. Treatment of transfected cells with forskolin and IBMX under the same conditions as in the in vivo phosphorylation experiments did not significantly increase the amount of biotinylated ROMK1 protein expression at the plasma membrane of HEK293 cells (data not shown). Thus, the observed increase in 32 P i incorporation in response to forskolin is due primarily to the enhancement of PKA-dependent phosphorylation of the ROMK1 proteins at the plasma membrane. Phosphoamino acid analyses were performed on the ROMK1 proteins, which were labeled either in vitro or in vivo. In both cases, we observed that the 32 P i was exclusively incorporated into serine residues (Fig.  4, B and D). No 32 P i -phosphothreonine was detected, even when the TLC plate was exposed to film for up to 30 days. In addition, there were no detectable counts associated with phosphotyrosine, suggesting that the basal level of phosphorylation in the absence of forskolin and IBMX (Fig. 4C) is not due to an endogenous tyrosine kinase-catalyzed phosphorylation. These data provide strong evidence that only those putative PKA phosphorylation sites in ROMK containing serine residues are the potential phosphate receptors.
Site-directed Mutagenesis of Putative PKA Sites-Three potential PKA phosphorylation sites with serine residues can be identified in ROMK1 or ROMK2 respectively). We prepared seven ROMK2 mutant constructs, including three single mutants, three double mutants, and one triple mutant form, in which all three serine residues were mutated to alanine residues (see "Materials and Methods"). Mutant forms of ROMK2 in pSVL vector were HA-tagged in order to immunoprecipitate the mutant ROMK proteins expressed in transfected HEK293 cells. First, we assessed whether the PKA phosphorylation site ROMK2 mutants affect the translation of these cDNAs both in vitro and in HEK293 cells. Fig. 5 shows that in vitro translation of equal amounts of cDNAs (1 g) from the wild-type and mutant ROMK2 constructs yielded the ϳ42-kDa bands (the apparent molecular mass of the ROMK2) of similar intensity. We also transiently transfected the wild-type or mutant HAtagged ROMK2 cDNA constructs into HEK293 cells and measured protein expression by cell-surface labeling with biotin. The results shown in Fig. 6 indicate that all four mutants were expressed at the plasma membrane of HEK293 cells at a level indistinguishable from that of wild-type HA-tagged ROMK2 (similar results were observed in the other two single mutants and one double mutant form; data not shown). These experiments demonstrate that the mutation of PKA phosphorylation sites does not produce structural alterations that prevent the synthesis and transport of the HA-tagged mutant ROMK proteins to the plasma membrane of HEK293 cells.
We next examined the ability of each mutant to be phosphorylated following exposure to forskolin and IBMX. Fig. 7A shows that the single mutant ROMK2 channels, R2-S25A and R2-S200A, exhibited ϳ3-fold increases in 32 P i incorporation in FIG. 4. In vivo or in vitro phosphorylation of transfected ROMK1 by PKA. A, ROMK1 cDNA was transfected into HEK293 cells, immunoprecipitated by anti-HA antibody, and phosphorylated in vitro by PKA in the presence of [␥-32 P]ATP. B, phosphoamino acid analysis of the in vitro phosphorylated ROMK1 phosphoprotein by electrophoresis at pH 3.5. C, ROMK1 cDNA was transfected into HEK293 cells, metabolically labeled with 32 P i , and treated with or without forskolin and IBMX. D, phosphoamino acid analysis of the in vivo phosphorylated ROMK1 phosphoprotein by electrophoresis at pH 3.5. Molecular mass is given in kDa. The position of the ROMK1 bands is indicated by arrows. P-Ser, phosphoserine; P-Thr, phosphothreonine; P-Tyr, phosphotyrosine; FSK, forskolin. response to cAMP agonists in vivo (a similar result was observed with the R2-S294A mutant; data not shown). Double mutation of the potential PKA phosphorylation sites (R2-S25A/ S200A, R2-S25A/S294A, and R2-S200A/S294A; Fig. 7B) re-sulted in approximately a 50% decrease in the intensity of phosphorylation with forskolin and IBMX compared with the wild-type ROMK2 protein. The triple mutant (R2-S25A/S200A/ S294A), which lacks all putative PKA phosphorylation sites, had only a basal level of phosphorylation and failed to show enhanced phosphorylation intensity after forskolin and IBMX treatment (Fig. 7B, lanes 9 and 10). These results indicate that in the absence of all three putative phosphorylation sites, the HA-tagged ROMK2 protein is not a substrate for PKA. Phosphorylation of the wild-type and mutant HA-tagged ROMK2 proteins are summarized in Table I.
Tryptic Phosphopeptide Mapping of ROMK Labeled in Vitro-In order to verify that all three serine residues are substrates for PKA-mediated phosphorylation, we compared two-dimensional maps of tryptic phosphopeptides from wildtype ROMK2 and from individual mutant and triple mutant forms of ROMK2 (Fig. 8). Serine mutant forms of the HAtagged ROMK2 cDNA were transfected into HEK293 cells and in vitro labeled using PKA and [␥-32 P]ATP. The resultant ROMK2 phosphopeptides were excised from SDS-polyacrylamide gel, digested with L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin, and fingerprinted by horizontal electrophoresis and ascending chromatography. By comparison of the wild-type and mutant HA-tagged ROMK2 fingerprints, specific spots on autoradiograms were assigned to each serine residue. As shown in Fig. 8, we were able to identify three phosphopeptides that were clearly associated with the three PKA consensus phosphorylation sites in the HA-tagged ROMK2 protein. These findings suggest that, at least in vitro, all three serine residues can be phosphorylated by PKA.
Functional Expression of the Wild-type and Mutant ROMK2 cRNA in X. laevis Oocytes-We determined the effects of the PKA phosphorylation site mutations on ROMK channel activity in X. laevis oocytes by two-electrode voltage clamp analysis. Oocytes were injected with equal amounts of channel cRNA and examined on the same day after injection. Endogenous cAMP levels in our oocytes are sufficient to give maximal or near-maximal levels of ROMK activity (e.g. forskolin does not enhance ROMK current). 2 Fig. 9 shows that the Ba 2ϩ -sensitive K ϩ current from single-site mutant HA-tagged ROMK2 was ϳ40% lower than currents observed in oocytes expressing wildtype ROMK2. On the other hand, oocytes injected with cRNA from double-or triple-site mutant ROMK2 showed no detectable Ba 2ϩ -sensitive K ϩ currents. These results indicate that phosphorylation of ROMK2 is absolutely required for channel activity and that at least two of the three serine residues must 2 Z.-C. Xu, Y. Yang, and S. C. Hebert, unpublished studies.
Ϫ Ϫ Ϫ a Metabolic and in vitro labeling were quantitatively analyzed by PhosphorImager with the ImageQuant™ program (Molec-ularDynamics).
b The number of plus (ϩ) signs indicates relative channel activity or level of phosphorylation. The minus sign (Ϫ) indicates no detectable activity or phosphorylation.

Phosphorylation of ROMK by PKA
be phosphorylated in order for the channel to be active. This is consistent with the previous patch clamp study by McNicholas et al. (1994) showing that PKA-dependent phosphorylation is essential for ROMK channel function. Moreover, the significant difference in magnitude of current between the wild-type and single-site mutant channels suggests that maximal channel activity requires phosphorylation of all three serine residues. Finally and interestingly, these results suggest that none of the sites is more or less important for channel function, i.e. any combination of two sites gives rise to the same level of channel activity. These functional effects of PKA phosphorylation site mutants are summarized in Table I. DISCUSSION It is well established that a variety of voltage-gated and ligand-gated ion channels are substrates for protein kinases, and phosphorylation of ion channel proteins on serine, threonine, or tyrosine residues is considered a ubiquitous mechanism of modulating ion channel activity (for review, see Catterall (1988) and Levitan (1988Levitan ( , 1994). Studies on renal tubules have shown that the low conductance, ATP-sensitive K ϩ channel (K ATP ) present in apical membranes of rat cortical collecting duct and MTAL cells is activated by PKA (Wang and Giebisch, 1991a;Kubokawa et al., 1995a;Wang, 1994a). A recent patch clamp study of ROMK2 expressed in X. laevis oocytes demonstrated that addition of the catalytic subunit of PKA and MgATP was required to restore channel activity following phosphatase-induced channel run-down (McNicholas et al., 1994). Thus, these physiological studies suggest that, similar to the native low conductance K ATP channel, ROMK channels are regulated by PKA-mediated phosphorylation and dephosphorylation processes, and that PKA-dependent phosphorylation is required for maintaining channel activity. Phosphorylation of K ATP channels in non-renal cells is also thought to be an important mechanism for modulation of these metabolically regulated channels (Ashcroft, 1988;Misler and Giebisch, 1992). However, it is not known whether ROMK, or indeed any of the cloned inwardly rectifying K ϩ channels, are substrates for protein kinases. In the present study, we describe the direct phosphorylation of ROMK K ATP channel protein in transiently expressed HEK293 cells by PKA.
We epitope-tagged ROMK1 and ROMK2 at the carboxyl termini in order to immunoprecipitate ROMK with anti-HA antibody. It should be noted that although ROMK1 and another isoform, ROMK2, differ at the amino terminus due to alternative splicing, they both form similar functional K ϩ channels when expressed in X. laevis oocytes (Boim et al., 1995) and both isoforms require PKA-mediated phosphorylation processes for maintenance of channel activity. 3 Although the HA epitope tag has been used frequently for verifying the expression of membrane proteins (Attisano et al., 1993;Wrana et al., 1994), the caveat with epitope tagging of proteins is that the added epitope may disrupt protein sorting and function. We have shown, however, that the HA-tagged ROMK1 and ROMK2 channels can be functionally expressed in X. laevis oocytes (Fig.  2 (ROMK1) and Fig. 9 (ROMK2)) and that these epitope-tagged constructs are detected by cell-surface biotinylation in HEK293 cells (Fig. 3B (ROMK1) and Fig. 6 (ROMK2)), indicating that the ROMK-HA proteins are functional and are expressed at plasma membranes.
The observations that the immunoprecipitated ROMK1 protein from transiently transfected HEK293 cells could be phosphorylated in vitro by PKA (Fig. 4A) and that forskolin with IBMX increased 32 P i incorporation into the ROMK1 (Fig. 4C) and ROMK2 (Fig. 7) proteins isolated from transfected FIG. 8. Two-dimensional tryptic phosphopeptide mapping of ROMK2 labeled in vitro. The wild-type or mutant ROMK2 cDNA constructs were transiently transfected into HEK293 cells, immunoprecipitated by anti-HA antibody, and phosphorylated in vitro by the catalytic subunit of PKA in the presence of [␥-32 P]ATP. The ROMK phosphoproteins were gel-purified and digested with trypsin. The tryptic peptides were then resolved by electrophoresis and chromatography in two dimensions. The origins are marked with a circle.
HEK293 cells in vivo demonstrate that ROMK channels are substrates for PKA-mediated phosphorylation and are consistent with the electrophysiological studies indicating that ROMK channels are regulated by PKA-dependent processes (McNicholas et al., 1994). Phosphoamino acid analyses of both in vitro and in vivo phosphorylated ROMK1 proteins demonstrated that phosphate is incorporated only at serine residues Ser-44, Ser-219, and Ser-313 by PKA (Fig. 4, B and D).
It is quite common that mutant forms of proteins fail to be transported to the correct cellular location (Cheng et al., 1990;Welsh and Smith, 1993). Thus, we tested for plasma membrane expression of ROMK2 in transiently transfected HEK293 cells by cell-surface biotinylation. The results in Fig. 6 clearly show that the serine-to-alanine mutations did not prevent the ROMK2 mutant proteins from sorting to the plasma membrane of HEK293 cells. Several lines of evidence from in vitro or in vivo phosphorylation strongly suggest that all three PKA sites are directly involved in phosphorylation of ROMK. First, the in vitro phosphorylated wild-type ROMK proteins examined by two-dimensional TLC analysis show three phosphopeptides, which represent three PKA sites that can be abolished by site-specific mutagenesis (Fig. 8A). Second, in vivo phosphorylation of the transfected ROMK2 mutant constructs indicate that the serine-to-alanine substitutions at these three PKA sites resulted in a decrease in 32 P i incorporation with stimulation with forskolin and IBMX. There was a 50% decrease in 32 P i incorporation in the double-site mutant forms and complete loss of 32 P i incorporation in the triple mutant form compared with the wild-type ROMK2 (Fig. 7, A and B). Failure to detect a significant decline in the 32 P i incorporation in the single mutant ROMK2 is due possibly to the high level of phosphorylation. In addition, we were not able to detect PKA-dependent phosphorylation in vitro when the triple mutated ROMK2 was examined by phosphopeptide analysis (Fig. 8E). Similarly, the magnitude of Ba 2ϩ -sensitive K ϩ currents observed in X. laevis oocytes was dependent on the number of serine-phosphorylation residues mutated.
From analysis of these results, which are summarized in Table I, two major conclusions can be reached. First, like the CFTR Cl Ϫ channel and the insulin receptor (Cheng et al., 1991;Zhang et al., 1991), phosphorylation of ROMK by PKA is degenerate, meaning that no one individual site is essential, and yet more than one site is required for maintaining channel activity. Second, studies on the triple mutant ROMK2 clearly indicate that there is a direct correlation between the ROMK phosphorylation and channel activity, and no other sites are detectable upon phosphorylation of ROMK by PKA. At present, we have not yet examined the ROMK single-channel properties using patch clamp techniques. These latter studies may reveal other aspects of channel function, which are modulated by specific phosphorylated residues (e.g. the characteristics of MgATP-or pH-mediated channel inhibition) .
In summary, the present study demonstrates that ROMK is a phosphoprotein, that the channel can be phosphorylated by PKA, and that three PKA sites containing serine residues are essential for ROMK channel activity. Given the critical importance of this channel for renal K ϩ secretion and recycling, these findings provide important insights into the functional regulation of ROMK and possibly other K ATP channels.