Mapping the Kidney Potassium Channel ROMK1

ROMK1, also known as Kir 1.1, is an inwardly rectifying K+ channel and is the prototypical member of the large Kir gene family. The accepted model of Kir topology predicts intracellular NH2 and COOH termini, and two membrane-spanning segments, M1 and M2, connected by an intramembranous pore-forming segment, H5. The sequence of H5 is similar in voltage-dependent K+ channels and features a strictly conserved GY/FG in its mid-region, which has been proposed as the selectivity filter of the pore. We have been usingN-glycosylation substitution mutants to map the extracellular topology of ROMK1 biochemically and have described several loci in H5 that were glycosylated. We now report glycosylation at loci Tyr144 and Phe146, which indicates that the signature GYG sequence (143–145) rather than being intramembranous is extracellular. The COOH terminus was predicted to begin at position 178, but contrary to the model, we observed that position 257 was glycosylated and surrounding positions at 199, 222, and 298 were unglycosylated. N-Glycosylation sequon substitution at the latter three positions abolished K+/Na+selectivity. Our results suggest a major revision of the topology of ROMK1 with H5 and the pore signature sequence now completely extracellular. The COOH terminus appears to form two additional membrane-spanning segments and to contribute to the ion conduction pathway.

ROMK1 is a weak, inwardly rectifying ATP-regulated K ϩ channel that was cloned from rat kidney outer medulla (1). There are three isoforms, which differ in their NH 2 terminus as a result of alternative splicing (2,3). The isoforms are differentially expressed along the loop of Henle and distal nephron (4). In the thick ascending loop of Henle, ROMK provides K ϩ to the Na-K-2Cl cotransporter and thereby enhances Na ϩ uptake (5,6). This physiological role is supported by mutations in ROMK that have been linked to the antenatal form of Bartter's syndrome, characterized by hypokalemic metabolic alkalosis, hypotension, renal salt wasting, and hypercalciuria (7,8). Furthermore, characterization of some of these Bartter's ROMK1 mutations revealed disruption of function by phosphorylation, proteolytic processing, and protein transport. 1 The topological model of ROMK1 and related members of the Kir gene family was predicted from hydropathy plots and, in the case of H5, from sequence similarity to Kv channels 2 (Fig.  1A). The NH 2 and COOH termini were cytoplasmic, based on the absence of a signal sequence in the open reading frame, and the ␣-helical transmembrane segments M1 and M2 were separated by a linker containing a 17-residue stretch called H5 (9 -13). In Kv channels, H5 is thought to form an intramembranous hairpin structure with the selectivity filter at the highly conserved GYG, which is at loci 143-145 of ROMK1.
Recently, we showed that several H5 loci of ROMK1 could be glycosylated and that one of them, Q139N, in both its glycosylated and unglycosylated forms was nonselective between K ϩ and Na ϩ (14). That H5 is a hotspot for residues that affect the pore is supported by the report that Leu 117 and Val 121 in ROMK2, corresponding to 136 and 140 in ROMK1, retained K ϩ selectivity but increased single channel conductance and Ba 2ϩ sensitivity (15). However H5 is not the sole pore-determining segment, since most of the important determinants of pore blockade by Mg 2ϩ and polyamines are negatively charged residues located in the carboxyl terminus (16 -19) and M2 (20 -24).
Our previous results (14) suggested that H5 and its flanking regions E1 and E2 formed the first extracellular loop (EL1) of Kir's and that the second transmembrane segment (M2) was shifted toward the COOH terminus (Fig. 1B). The possibility remained that the GYG signature pore sequence was intramembranous, and to test this, we engineered N-glycosylation substitution mutants (GSMs), which are markers of extracellularity in and around this region. We also targeted the COOH terminus thought to become cytoplasmic at position 178 for this type of evaluation. Glycosylation was demonstrated by gel shift assays and reductions in ROMK1 currents before and after tunicamycin (TM) treatment. Our results show that the present topological model requires major revision. We found that position 144, supposedly forming the intramembranous selectivity filter of the pore, was glycosylated as was its immediate neighbor at locus 146. Thus the signature sequence rather than being intramembranous is extracellular. We found that position 257 was glycosylated, whereas surrounding loci at 199, 222, and 298 were not. However, N-glycosylation site substitution at 199, 222, and 298 rendered ROMK1 nonselective for K ϩ over Na ϩ . It appears that the COOH terminus contributes at least two additional transmembrane segments to Kir topology and may form part of the ion conduction pathway.

EXPERIMENTAL PROCEDURES
Materials-TNM-FH insect medium was purchased from JRH Biosciences; Sf9 cells were obtained from American Type Culture Collec-tion; the TA cloning kit was from Invitrogen; the BaculoGold™ transfection kit was from PharMingen; fetal bovine serum and penicillinstreptomycin and pluronic F-68 10% solution were from Life Technologies, Inc.; Zwittergent 3-10 was from Calbiochem; Tween 20, tunicamycin, phenylmethylsulfonyl fluoride, aminobenzamidine, bovine serum albumin, and trypan blue solution were from Sigma; and Qiagen columns were from Qiagen.
Mutagenesis-All ROMK1 mutants were engineered by polymerase chain reaction overlap extension (25) using FLAG/N117Q cDNA as template (14,26). Polymerase chain reaction products were subcloned into pCRII for amplification and sequencing and subsequently subcloned into NotI/BamHI double-digested baculovirus transfer vector (pVL 1392). The M2 FLAG epitope was fused to the NH 2 terminus of ROMK1 for immunoaffinity purification, Western blots, and immunocytochemistry.
Cell Culture and Recombinant Baculoviruses-Sf9 cells were grown in Hink's TNM-FH insect medium containing 10% (v/v) fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, and 0.1% pluronic F-68 under natural atmosphere at 27°C. Cells were maintained in monolayer cultures and passaged every 4 -10 days. Production of recombinant protein was conducted in suspension cultures at a cell density of 1-1.2 ϫ 10 6 cells/ml using viral supernatant from suspension culture amplifications. The recombinant baculoviruses were generated by cotransfection of Sf9 cells with baculovirus vectors containing the ROMK1 construct and BaculoGold™ viral DNA (modified AcNPV) according to the accompanying instructions (PharMingen). Viral seed stocks were prepared by amplification using monolayer cultures. High viral titers were generated by infecting suspension cultures (0.5-0.8 ϫ 10 6 cells/ml) with viral seed stocks. The viral supernatant and/or the infected cells were collected 1-5 days after the infection. Tunicamycin (25 g/ml) was added to cells 5-15 min after inoculation with viral supernatant. Standard procedures for routine subculturing, cotransfections, and infections were followed (27).
Immunoaffinity Purification-Infected Sf9 cells (4 ϫ 10 7 to 5 ϫ 10 7 ) were harvested 19 -28 h postinfection by centrifugation at 2000 ϫ g for 10 min. The pellet was resuspended in cell lysis buffer (50 mM phosphate, pH 7.4, 0.3 M KCl, 2% Zwittergent 3-10), plus 125 l of the M2 antibody affinity gel (Eastman Kodak Co.) resuspended in PBS, which recognizes the FLAG epitope attached to the NH 2 terminus. The requirement of protease inhibitors (100 M phenylmethylsulfonyl fluoride and 200 M aminobenzamidine) in the cell lysis buffer was optional, because degradation was not a problem. The sample was gently rocked overnight. Resin was washed with the solubilization buffer without the detergent (8 ml) and then with PBS. Beads were resuspended in 100 -150 l of reducing SDS sample buffer and heated at 90°C for 4 min.

SDS-Polyacrylamide Gel Electrophoresis and Western
Blotting-Immunoaffinity samples were subjected to electrophoresis on 10% SDSpolyacrylamide gels. The electrophoresed proteins were transferred to Immobilon-P membranes and probed with the M2 antibody (mouse monoclonal IgG, that recognizes the FLAG epitope attached to ROMK1 proteins; Eastman Kodak Co.) as described previously by Schwalbe et al. (14,26). The gels were run until carbonic anhydrase migrated off the gel, ensuring separation of the glycosylated proteins. Only the sections of the gels that contain ROMK1 proteins are shown.
Laser Confocal Microscopy-Suspension cultures infected for 20 -28 h were plated in cell wells containing glass coverslips, allowed to attach for at least 15 min, and then washed with PBS. Cells were then fixed with paraformaldehyde (3%) for 10 min and washed three times with PBS. Cells were permeabilized, and nonspecific binding was blocked using PBS, containing 3% bovine serum albumin, 0.1% Tween 20 and 0.03% Triton X-100. Fixed cells were incubated with the primary antibody M2 (1:100) overnight at 4°C, washed three times with blocking buffer, minus Triton X-100, and incubated with the secondary antibody, rhodamine-conjugated sheep anti-mouse (1:50) for 1 h at room temperature. Cells were washed three times with PBS. The glass coverslips were mounted with Vectorex medium. Cells were examined with a Noran Instruments Odyssey XL confocal laser scanning microscope equipped with a Nikon inverted microscope using a 60 ϫ objective lens. The antibodies were suspended in PBS, plus 3% bovine serum albumin and 0.1% Tween 20.
Patch-Clamp Recordings-Cells were added 20 -24 h postinfection to Petri dishes containing glass coverslips, and currents were measured 23-42 h postinfection. Currents were measured at 23-25°C using an Axoclamp 1B amplifier (Axon Instruments, Foster City, CA) with membrane capacitance and resistance compensations. Fire-polished pipette electrodes had tip resistances of 2-5 and 5-10 M⍀ for whole-cell and single channel measurements, respectively. Voltage pulse protocols and data acquisition and analysis were carried out using the pClamp suite of programs (Axon Instruments). Data were filtered at 1 kHz and subsequently digitized at 5 kHz.

Experimental Approach
In our topological model of ROMK1, the H5 segment thought to be intramembranous was proposed as extracellular (Fig. 1B). To address experimentally the extracellularity of both the GYG signature sequence within H5 and the putative intracellular COOH terminus, we used N-glycosylation tagging. All of the GSMs have the native N-glycosylation site removed (NRT, amino acid residues at positions 117-119) and only one introduced N-glycosylation sequon (26). Sf9 cells were infected with recombinant baculovirus encoding the ROMK1 mutants for expression and characterization (14,26).

Detection of Glycosylation by Structural and Functional Assays
Gel Shift Assays-Proteins were immunoaffinity-purified and detected by Western blotting using the M2 FLAG epitope fused to the NH 2 terminus (WT ROMK1, Y144N/F146S, V199N, I222N, N259T, and D298N) or both termini (F146N/ F148S). The octapeptide attached to the NH 2 terminus (26) or both termini did not interfere with WT ROMK1 function. Pre-FIG. 1. Topological structures of ROMK1. A, the predicted two-dimensional orientation of ROMK1 based on hydropathy plots and sequence homology to Kv channels. M1 runs from residue 83 to residue 104, M2 from residue 156 to 177, and H5 from residue 131 and 147. Two stretches flanking H5 are E1 from residue 105 to 130 and E2 from residue 153 to 156. The NH 2 terminus runs from residue 1 to 82 and the COOH terminus from residue 178 to 391. B, our initial modified topology based on N-glycosylation tagging (14). The orientation of the NH 2 and COOH termini and M1 are similar to the predicted model. The differences are the presence of a single extracellular loop (EL1), containing E1, H5, and E2 and repositioning of M2, which now runs from residue 166 to 185. The numbers represent the first residues in the native and introduced N-glycosylation sites from this study and our previous study (14). Underlined numbers are those examined in this study. The branched structure indicates glycosylation.
viously we have shown that ROMK1 of total Sf9 membranes migrates as a doublet: the upper immunoband (45 kDa) was glycosylated protein and the lower immunoband (43 kDa) was unglycosylated (14,26). These two bands were observed for immunoaffinity samples of WT ROMK1, Y144N/F146S, F146N/F148S, and N259T ( Fig. 2A). The upper band of WT ROMK1, Y144N/F146S, F146N/F148S, and N259T was not detected when N-glycosylation was inhibited with TM. For V199N, I222N, and D298N, only the lower TM-insensitive band was present (Fig. 2B). The degree of glycosylation was different for each mutant, as well as WT ROMK1 and other GSMs as reported in our previous studies (14,26), indicating differences in the accessibility of the introduced sites to the oligosaccharyltransferase. Some of the samples showed a slightly faster migrating band that was degradation product, which was also observed in some total membrane samples (14). Immunobands of F146N/F148S migrated slightly slower due to the additional eight residues attached to the COOH terminus. The only prominent bands in uninfected cells or cells infected with a recombinant virus encoding the human ether-a-go-go gene (HERG), a Kv channel construct without the FLAG epitope, were the heavy and light chains of the mouse monoclonal antibody. Bands at positions of glycosylated and unglycosylated ROMK1 protein were not detected.
Membrane Currents-Inward currents for F146N/F148S, V199N, I222N, N259T, and D298N were time-independent, and outward currents showed weak inward rectification (Fig.  3). These currents are characteristic of ROMK1 channels expressed in Sf9 cells (26) and Xenopus oocytes (1,16). Recently we showed that inhibition of utilized N-glycosylation sequons of ROMK1 at native or introduced sites using TM significantly reduced whole-cell currents (14,26). Consistent with these results we found that TM produced a large reduction in the whole-cell currents for F146N/F148S and N259T, while the currents for V199N, I222N, and D298N were not affected. Whole-cell currents for Y144N/F146S were time-independent, like WT ROMK1, but were very small (3% of WT). After TM treatment the currents could not be distinguished from leak current. The effects of TM were invariant with respect to the presence and absence of glycosylation. However the magnitude of the currents expressed by the different GSMs was variable due to differences in viral titers and postinfection times. Therefore whole-cell currents were measured in the presence and absence of TM using the same viral supernatant at similar post-infection times.
The reduction in whole-cell currents for WT ROMK1 produced by TM, as well as other glycosylated ROMK1 mutants, was due to a decrease in the opening probability, P o , of the unglycosylated counterpart (14,26). F146N/F148S and N259T channels have long open times and brief closures, and the closures were very few for the former, while the latter was similar to WT ROMK1 ( Fig. 4; Ref. 26). Inhibition of glycosylation of WT ROMK1 and GSMs with TM greatly increased the closing times (14,26). The P o showed a large reduction in the presence of TM. The single channel current amplitudes and the slope conductances for glycosylated and unglycosylated F146N/ F148S and N259T were similar to WT ROMK1 (26). For  V199N, I222N, and D298N, the P o was low ( Table I). The single channel current amplitudes were similar to WT ROMK1; however, openings of larger amplitude with considerable flickering were also observed.

K ϩ /Na ϩ Selectivity and Ba 2ϩ Block
Disruptions in the primary sequence of the putative poreforming segment containing the selectivity filter sequence (GYG) may produce nonselective channels and alter the potency of channel blockers. WT ROMK1 currents are highly selective for K ϩ over Na ϩ (1,16). When K ϩ was replaced with Na ϩ , the currents were significantly reduced, and the current reversal potentials were much more negative for F146N/F148S ( Fig. 5A; Table I), N259T, and Y144N/F146S. For I222N (Fig.  5), V199N and D298N ⌬V values were Ϫ0.5 Ϯ 0.5 mV, n ϭ 4; 0.5 Ϯ 1.1 mV, n ϭ 6; and 1.5 Ϯ 0.81 mV, n ϭ 13, respectively. These negligible shifts in the current reversal potentials indicated a lack of selectivity between K ϩ and Na ϩ (Table I).

Cellular Location of ROMK1 H5 Substitution N-Glycosylation Mutants
The small currents for Y144N/F146S suggested that either lower amounts were expressed at the cell surface or fewer channels were in the open state at the cell surface. To distinguish between these possibilities, ROMK1 mutant proteins were localized in Sf9 cells using the M2 monoclonal antibody and fluorescence immunocytochemistry in combination with laser confocal microscopy. The immunofluorescence staining patterns for the mutants were similar, and the protein appeared to be at the cell surface in the majority of the cells. The thin outer immunofluorescence ring for both Y144N/F146S and F146N/F148S indicated that the channel proteins were at the cell surface as we described for WT ROMK1 (Fig. 6). 2 Immunofluorescence was also observed as a thicker inner ring around the nucleus and patches in between the rings. The immunofluorescence around the nucleus was attributed to pro-

N-Glycosylation Substitution Sequons in ROMK1-N-
Linked oligosaccharides were detected for GSMs at positions 144, 146, and 257 (Fig. 7). Further evidence of glycosylation was provided by the gel shifts produced by TM. The different amounts of glycosylated monomer for WT ROMK1, Y144N/ F146S, F146N/F148S, and N259T, and also GSMs from our previous study (14), reveal differences in the accessibility of the native and novel sites to the oligosaccharyltransferase. Since utilization of an N-glycosylation site depends on the placement of a site in the exoplasmic loop, the distance from the NH 2 terminus, and the conformation of a tripeptide sequon (28 -30), the degree of glycosylation may differ. Glycosylation was also consistent with the large reductions in whole-cell currents caused by TM, which correlated with large reductions in P o rather than decreases in channel number as noted previously (14,26). In addition, no apparent difference in channel density at the cell surface of F146N/F148S in the presence and absence of TM was observed by immunocytochemistry measurements. The present results with TM along with our previous studies (14,26) show that at least one glycosylated subunit was incorporated into the functional channels at the cell surface. In addition, immunocytochemistry of Y144N/F146S and F146N/ F148S showed that channels were transported to the plasma membrane. Taken together, we conclude that positions 144, 146, and 257 were glycosylated and assembled as functional channels. We also note that N-linked oligosaccharides positioned anywhere along the ROMK1 sequence stabilized the open state of the channel, consistent with our previous studies FIG. 5. K ؉ /Na ؉ selectivity and Ba 2؉ block of ROMK1 N-glycosylation substitution mutants. A, whole-cell currents recorded in a high K ϩ solution (left side; in mM): potassium aspartate, 140; MgCl 2 , 1; MES, 10; mannitol, 60 (pH 6.3, adjusted with Tris-OH)) and low K ϩ solution (center; in mM) sodium aspartate, 135; potassium aspartate, 5; MgCl 2 , 1; MES, 10, mannitol, 60 (pH 6.3, adjusted with Tris-OH). Current-voltage relationships of the mutants are shown on the right side. Open and filled circles are high and low K ϩ solutions, respectively. B, whole-cell currents in the high K ϩ solution before and after addition of 7 mM BaCl 2 . The first 50 ms of the recordings are shown. Currents were evoked by voltage steps from ϩ80 to Ϫ100 mV, from a holding potential of 0 mV.

TABLE I ROMK1 N-glycosylation mutants and their function
The percent reduction in the whole cell currents was the difference in the currents of GSMs with and without TM, divided by currents of GSMs without TM treatment, at Ϫ120 mV. The single channel opening probability (P o ) at Ϫ100 mV in cells with and without TM. The single channel conductance was obtained at voltage steps from Ϫ160 to Ϫ40 mV at 20-mV increments, from a holding potential of 0 mV. The ⌬V represents the change in the potential of the zero membrane current when the solution perfusing the cell was changed from high to low K ϩ . The time dependency of the barium block is expressed as the ratio between the level of the current at the beginning (peak) and end (steady-state) of a 200-ms pulse at Ϫ120 V.   (14,26). Positions 199, 222, and 298 were not N-glycosylated, as only the lower TM-insensitive band was observed. The whole-cell currents were not affected by TM, and the P o was low as previously reported for unglycosylated ROMK1 channels (14,26). Likewise unglycosylated channels for F146N/F148S and N259T produced by TM had small P o values. Thus the functional phenotype of the unglycosylated channels was distinct from the glycosylated channels. A possible mechanism of the two distinct phenotypes may be that N-linked carbohydrate is needed to direct correct disulfide bond/s formation and consequently stabilize the open state of the glycosylated channels. The significance of N-glycosylation in directing disulfide bond formation has been demonstrated for other proteins (31)(32)(33). In any case, our results do indicate that N-linked carbohydrate at residues in E1, H5, E2, and the putative intracellular COOH terminus can stabilize the open state of the channel and therefore suggest that they are in a similar environment.
K ϩ selectivity was retained when residues 144, 146, 148, and 259 were mutated to nonconserved residues and when oligosaccharides were attached to Asn at positions 144, 146, and 257, whereas substitutions at positions 199, 222, and 298 with Asn caused virtually a complete loss of K ϩ selectivity. These results indicate that neither the side chains at positions 144, 146, 148, and 259 nor the bulk of the carbohydrate at positions 144, 146, and 257 drastically altered the formation of the selectivity filter (GYG; 143-145), while residues 199, 222, and 298 in the putative intracellular COOH terminus influenced selectivity either directly or indirectly by destabilizing the pore structure.
WT ROMK currents were blocked by Ba 2ϩ , as were currents from all of the GSMs. The time dependence of the block at hyperpolarized pulses was retained when residues in H5, positions 144, 146, and 148, or in the cytoplasmic tail at position 259 were mutated. However, three mutations in the intracellular COOH terminus at positions 199, 222, and 298 produced a time-independent block. This result is consistent with a more accessible binding site or a higher affinity of Ba 2ϩ for the V199N, I222N, and D298N channels. Thus, mutations in the COOH terminus segment of ROMK1 appeared to alter the structure or entrance of the pore.
Significance for Topology-Our results show that the GYG signature pore sequence and part of the putative intracellular COOH terminus are extracellular, suggesting two additional transmembrane segments (Fig. 7). The present results, together with our previous demonstration that positions 117, 128, 133, 139, and 153 could be glycosylated (14), places the signature GYG sequence and the linker between M1 and M2 (EL1) as extracellular (Fig. 7). In contrast to earlier models ( Fig. 1), we also provide evidence of glycosylation at position 257, indicating that part of the putative cytoplasmic COOH terminus is extracellular. Furthermore the lack of utilized sites at positions 199, 222, 274, 298, 377, and 384 (14) and the presence of protein kinase A phosphorylation sites at positions 219 and 313 (34) would indicate that there are at least two transmembrane segments in the putative cytoplasmic COOH terminus. Based on these findings and glycosylation studies, which have indicated that a utilized N-glycosylation sequon is a minimum of 11 residues from the nearest transmembrane segment and in an exofacial loop of at least 30 residues (29, 30), we predict that the 10 -20 residues, depending on whether the transmembrane segments are ␣-helices or ␤-strands, that make up M3 would range from 222 to 246 and those for M4 from 270 to 308 (Fig. 7). Other experimental results that would be consistent with our proposed model are the following: NH 2 terminus involved in regulating intracellular pH sensitivity (35,36); M1-H5 region implicated in external pH regulation (37), Ser 44 site for cAMP-dependent kinase phosphorylation (34); the Walker A-type motif, running from Gly 23 to Lys 229 (38); and position 171 involved in Mg 2ϩ and polyamine binding (20,21).
Our study has also shown that mutating residues Val 199 , Ile 222 , and Asp 298 with Asn produced nonselective channels and a time-independent barium block. In our newly revised model these residues are localized in or near the membrane where they may contribute to the selectivity filter. Thus, residue Asp 298 may be located near M4. Ile 222 may be near the membrane or may be part of M3. In our model Val 199 is part of IL1, because Ser 219 was shown to be phosphorylated. However, this segment may either interact with the membrane or may loop into the cytoplasmic entrance of the pore. Clearly, future studies are needed to determine the exact placement of these residues with respect to the plasma membrane and to determine whether these residues are directly involved in forming the selectivity filter and barium binding site/s or whether they are indirectly involved in stabilizing the pore structure.  Fig. 1B. The present revision shows that the putative intracellular COOH terminus contributes two additional transmembrane segments, M3 and M4, a second extracellular loop, EL2, and an intracellular loop, IL1. M3 ranges from residue 222 to residue 246 and M4 from 270 to 308, based on glycosylation at position 257 and phosphorylation at Ser residues 219 and 313. The underlined numbers represent the first residues of the introduced N-glycosylation sequons from this study and other numbers are from our previous reports. The branched structure denotes glycosylation. The Ser residues are cAMP-dependent kinase sites (34).