Assembly and trafficking of a multiprotein ROMK (Kir 1.1) channel complex by PDZ interactions.

The ROMK subtypes of inward rectifier K+ channels (Kir 1.1, KCNJ1) mediate potassium secretion and regulate NaCl reabsorption in the kidney. In the present study, the role of the PDZ binding motif in ROMK function is explored. Here we identify the Na/H exchange regulatory factors, NHERF-1 and NHERF-2, as PDZ domain interaction partners of the ROMK channel. Characterization of the basis and consequences of NHERF association with ROMK reveals a PDZ interaction-dependent trafficking process and a coupling mechanism for linking ROMK to a channel modifier protein, the cystic fibrosis transmembrane regulator (CFTR). As measured by antibody binding of external epitope-tagged forms of Kir 1.1 in intact cells, NHERF-1 or NHERF-2 coexpression increased cell surface expression of ROMK. Channel interaction with NHERF proteins and effects of NHERF on ROMK localization were dependent on the presence of the PDZ domain binding motif in ROMK. Both NHERF proteins contain two PDZ domains; recombinant protein-protein binding assays and yeast-two-hybrid studies revealed that ROMK preferentially associates with the second PDZ domain of NHERF-1 and with the first PDZ domain of NHERF-2, precisely opposite of what has been reported for CFTR. Consistent with the scaffolding capacity of the NHERF proteins, coexpression of NHERF-2 with ROMK and CFTR dramatically increases the amount of ROMK protein that coimmunopurifies and functionally interacts with CFTR. Thus NHERF facilitates assembly of a ternary complex containing ROMK and CFTR. These observations raise the possibility that PDZ-based interactions may underscore physiological regulation and membrane targeting of ROMK in the kidney.

The ROMK (Kir 1.1 or KCNJ1) subtypes of weakly inward rectifying potassium channels (1) play critical roles in salt and water homeostasis. Chiefly localized on apical membrane of specific epithelial cells in the kidney (2)(3)(4), each of the products of the ROMK gene (amino-terminal splice variants are termed ROMK1 (Kir 1.1a), ROMK2 (Kir 1.1b), ROMK3 (Kir 1.1c) (5)) functions as an exquisitely regulated channel for the transport of potassium into the renal tubule lumen. ROMK1, ROMK2, and ROMK3 exhibit nearly identical functional properties but are expressed differentially along the nephron (6) for different physiological duties (7). ROMK2 channels in the thick ascending limb of Henle's loop are responsible for recycling potassium across the apical membrane to maintain avid NaCl reabsorption through the Na ϩ /K ϩ /2Cl Ϫ cotransporter, important for the urinary concentrating mechanism. ROMK1 and ROMK3 channels in the distal nephron, on the other hand, are thought to constitute the final regulated component of the potassium secretory machinery of the kidney, essential for controlling renal potassium excretion and maintaining potassium balance (8). The physiologic significance of these channels is underscored by the link to human disease. Loss-of-function mutations in the ROMK gene cause Bartter's syndrome, a familial salt-wasting nephropathy (9,10). Mice, lacking the ROMK gene, manifest a similar disorder (11).
Although recombinant ROMK channels share many functional features of the small conductance apical membrane potassium channels in the thick ascending limb and collecting duct, the absence of sensitivity to cytoplasmic ATP has suggested that the native channel might be more complex. Reminiscent of K ATP channels in excitable tissues, which are heteromultimeric proteins complexes comprised regulatory sulfonylurea receptor (SUR) 1 ATP-binding cassette gene products and pore-forming inward rectifier, Kir 6 subunits (12,13), present evidence from molecular reconstitution studies indicates that ROMK channels also require a ATP-binding cassette protein cofactor to manifest native channel properties. We and others found that coexpression of CFTR with ROMK in Xenopus oocytes leads to the formation of weakly inward rectifying channels that have acquired sensitivity to sulfonylurea agents (14) and ATP-dependent gating properties (15) like the native channel (16,17). With observations that the expression patterns of ROMK (2)(3)(4) and CFTR (18) overlap along the thick ascending limb and collecting duct apical membrane, it seems plausible that the native ROMK secretory channel may be regulated by CFTR in the kidney. Certainly, the concept has precedent with an ever growing body of data, demonstrating that CFTR not only functions as a ClϪ channel but also acts as a "conductance regulator," modulating a plethora of different epithelial transport proteins (19).
Although the mechanisms by which CFTR modulates ion channels and other transport molecules are poorly understood, observations that CFTR interacts directly with variety of different PSD95/Dlg (Disc large)/and ZO-1 (zona occludes) (PDZ)-domain containing proteins (20 -23) suggest a potential coupling mechanism. PDZ domains are 80 -90-amino acid residue protein interaction modules that generally bind to short motifs (type I PDZ binding motif is recognized by the sequence S/T-X-I/V/L/M) found at the extreme COOH-terminal tail of certain membrane proteins and cytoplasmic signal transduction molecules to organize multiprotein complex formation on specific membrane domains (24,25). PDZ proteins that bind to the COOH terminus of CFTR, such as the Na/H exchange regulator factor (NHERF (26,27)), contain multiple protein-protein interaction domains, allowing them to act as molecular scaffolds, which assemble CFTR and signaling molecules into multimeric complexes. In fact, recent studies strongly suggest that intracellular trafficking (28); intermolecular domain cross-linking (22,29), macromolecular signaling complex formation and phosphorylation-dependent regulation of CFTR (23,31) are coordinated by direct interactions with different PDZ proteins.
Interestingly, the COOH-terminal domain of ROMK1-3 channels also contains a canonical type I PDZ binding motif (TQM-COOH). We recently reported that the PDZ binding sequence is a determinant of ROMK channel function, being required for efficient expression of active ROMK channels on the plasma membrane (32). These observations raise the possibility that trafficking of ROMK and interaction with CFTR are linked by a common PDZ domain-based scaffold. The goal of the present study was to test this hypothesis. Here we show that ROMK associates directly with NHERF-1 and NHERF-2 through a PDZ binding interaction to facilitate expression of ROMK on the plasmalemma and to coordinate the assembly ROMK and CFTR into a ternary complex.

EXPERIMENTAL PROCEDURES
DNA Constructs-cDNAs encoding ROMK2 cytoplasmic COOH-terminal regions (amino acids 307-372 or 307-369) were amplified by PCR from a full-length template (GenBank NM017023) and cloned in-frame with GST in the fusion expression vector pGEX-5x (Amersham Biosciences) or with LexA in pJK202 for GST and yeast two-hybrid studies, respectively. The cDNAs encoding PDZ domains of NHERF-1 (NM004252) and NHERF-2 (AF035771), as indicated under "Results," were amplified by PCR from full-length templates and cloned in-frame with either a hexahistidine sequence in the pRSET plasmid (Invitrogen) or with a trans-activation domain in the pJG4-5 plasmid for generation of recombinant His-tagged NHERF PDZ proteins and prey proteins for yeast two-hybrid studies, respectively. All constructs used for studies in Xenopus oocytes were subcloned between the 5Ј-and 3Ј-untranslated region of the Xenopus ␤-globin gene in the modified pSD64 vector to increase expression efficiency (33). This vector also contains a polyadenylate sequence in the 3Ј-untranslated region (dA23dC30). With the exception of EGFP-ROMK, all constructs used for mammalian expression were subcloned into pcDNA3.1ϩ (Invitrogen). EGFP was engineered onto the NH 2 terminus of wild-type or HA-ROMK2 by subcloning the wild-type or epitope-tagged ROMK2 inframe with EGFP in the pEGFP-Cl vector (Clontech Clontech). The sequence of all amplified or modified cDNAs was confirmed by dye termination DNA sequencing (University of Maryland School of Medicine Biopolymer Core).
Cell Culture and Transfections-COS-7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, 10 mM HEPES, and 2 mM L-glutamine. Cells were transfected with 1-2 g of plasmid using LipofectAMINE (Invitrogen) or Gene Juice (Novagen) according to the manufacturer's specifications.
GST Affinity Chromatography-GST-ROMK COOH-terminal fusion proteins were produced in Escherichia coli (Top Ten, Stratagene), and purified under nondenaturing conditions using glutathione-Sepharose 4B affinity chromatography as recommended by the manufacturer (Amersham Biosciences). Rat kidney or COS-7 cell extracts were pre-pared by homogenization in cold lysis buffer (150 mM NaCl, 20 mM Tris, pH 7.5, 5 mM EDTA), containing a protease inhibitor mixture (10 g/ml antipain, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, and 10 g/ml pepstatin A) and either 1% Triton X-100 or 1% CHAPS. After centrifugation of the detergent-solubilized lysates at 15,000 ϫ g, the supernatant was collected and incubated with the GST fusion proteins (ϳ25 g) bound to glutathione-Sepharose beads. After incubation at room temperature for 1 h with gentle rocking, beads were washed four times with phosphate-buffered saline and four times with TEE (50 mM Tris, pH 7.5, 1 mM EDTA, 1 mM EGTA). Protein bound to beads was eluted in SDS-sample buffer, resolved by SDS-PAGE, and then processed for silver staining (Bio-Rad, Silver Stain Plus) or for immunoblotting. For the latter, proteins were transferred electrophoretically to nitrocellulose (Hybond, Amersham Biosciences) and probed with either a rabbit anti-NHERF antibody (34) at 1:250) or a rat monoclonal anti-HA (Roche Applied Science) antibody followed by horseradish peroxidase-conjugated secondary antibody (either donkey anti-rabbit IgG or goat-anti mouse IgG at 1:10,000 (Jackson Laboratory). An enhanced chemiluminescence system (Amersham Biosciences) was used as recommended by the manufacturer to detect bound antibodies.
Immunoprecipitation and Immunoblotting-48 h post-transfection, COS cells were washed once with ice-cold phosphate-buffered saline, harvested in cold lysis buffer (150 mM NaCl, 20 mM Tris, pH 7.5, 5, mM EDTA), pelleted (2, 000 ϫ g for 5 min), and resuspended (ϳ5 times the cell pellet volume) in lysis buffer containing 1% Triton and a protease inhibitor mixture (10 g/ml antipain, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, and 10 g/ml pepstatin A). Cells were then passed though a 27-gauge needle, rotated at 4°C for 1 h, and then centrifuged at 15, 000 ϫ g for 15 min at 4°C. Soluble fractions from COS cells were precleared using 100 l of a 50% Sepharose B slurry at 4°C for 2 h with rotation. Precleared COS cell lysates were then rotated overnight with 100 l of a 10% protein A/G slurry with either 2 g of rabbit anti-HA antibody (Santa Cruz Biotechnology, Santa Cruz, CA) or mouse monoclonal anti-CFTR R (R & D Systems) or an unrelated IgG. After washing three times with lysis buffer containing 0.1% Triton X-100, the immunoprecipitates from COS cells were eluted for 30 min at room temperature with SDS-sample buffer. Eluates were separated on a 10% SDS-polyacrylamide gel and transferred to nitrocellulose. For Western blotting, a rabbit anti-ROMK antibody was used as described before (35) followed by horseradish peroxidase-conjugated goat antimouse or anti-rabbit antibodies (Jackson Laboratory) for detection with the enhanced chemiluminescence system.
Immunofluorescence and Confocal Microscopy-Cells, grown on glass coverslips, were transfected as above. After 2 days, cells were washed in ice-cold modified Ringer's solution (144 mM NaCl, 5 mM KCl, 1 mM CaCl 2 , 5.5 mM CaCl 2 , 5.5 mM glucose, 1.2 mM NaH 2 PO 4 , 5 mM HEPES, 7.4) and fixed with 4% formaldehyde in Ringer's solution for 15 min at 4°C. For colocalization studies of EGFP-ROMK2 and HA-tagged NHERF, cells were permeabilized with 0.1% Triton X-100 for 30 min at room temperature and washed three times in the modified Ringer's solution before blocking. For the external HA epitope EGFP-tagged ROMK surface expression studies, cells were not treated with Triton. Both groups were blocked with 1% bovine serum albumin in the modified Ringer's solution for 30 min at room temperature and labeled with primary antibody in 0.1% bovine serum albumin (rat anti-HA (Roche Applied Science) at 10 ng/ml) for 3 h at room temperature, then washed in modified Ringer's solution three times. Next, cells were incubated with secondary antibody conjugated to either ALEXA 568 (goat anti-rat 1:100, Molecular Probes) in 0.1% bovine serum albumin for 1 h at room temperature in the dark. Slides were then washed and mounted onto slides in VectaShield and sealed with nail polish. Cells were visualized using a 410 Zeiss laser scanning confocal microscope under a 63ϫ oil immersion lens. Images were acquired at zoom of 2 and pinhole size of 18 using 16-frame averaging and processed with Adobe Photoshop. Images of anti-HA cell surface labeling were collected at a constant gain and contrast for all samples.
Cell Surface Expression Assay-The surface luminescence of COS-7 cells was measured as described previously by Sharma et al. (36) with slight modifications. Briefly, transfected COS-7 cells were washed twice in ice-cold Ringer's solution and then labeled with 0.2 g/ml rat anti-HA monoclonal antibody in complete medium (Dulbecco's modified Eagle's medium, 10% fetal bovine serum, 5 mM HEPES) for 2 h at 4°C. Cells were then washed three or four times in ice-cold Tris-buffered saline containing 1 mM CaCl 2 and 1 mM MgCl 2 followed by incubation in horseradish peroxidase-coupled goat anti-rat antibody at 0.8 g/ml. Cells were washed once in TBS containing Ca 2ϩ and Mg2ϩ then twice in TBS containing 2 mM EGTA. Once washed, the cells were detached, pelleted, and resuspended in Ringer's solution. Aliquots of the resus-Assembly and Trafficking of ROMK (Kir 1.1) Channel Complex pended cells were added to 100 l of Super Signal enzyme-linked immunosorbent assay substrate (Pierce), incubated for 30 s at room temperature, and luminescence was measured using a Sirius luminometer.
Gel Overlays-Interaction between the NHERF-1 PDZ domains and the ROMK2 COOH terminus was determined in vitro by performing an overlay assay of recombinant protein fragments following the methods of Wyszynski and Sheng (37). His-tagged NHERF protein fragments were produced in E. coli, purified to homogeneity under nondenaturing conditions using nickel-nitrilotriacetic acid spin columns (Qiagen), and used to probe for interaction with immobilized GST-ROMK2 COOHterminal fusion proteins. GST and the GST-ROMK2 proteins (ϳ10 g) were resolved by SDS-PAGE (10% SDS-acrylamide) and transferred electrophoretically onto a nitrocellulose membrane. Proteins, immobilized on the filter, were renatured in gradually decreasing concentrations of guanidine HCl (from 6 M to zero), blocked with 5% milk in buffer B (120 mM KCl, 25 mM HEPES, 1 mM EDTA, and 0.2% Triton X-100), and then overlaid with His-tagged NHERF domains (5 g/ml) in buffer B containing 5% milk. After incubation with the probes at room temperature for 30 min, membranes were washed three times with buffer B to remove spurious binding as described previously (37). Blots were then probed with anti-His antibody followed by donkey anti-rabbit IgG conjugated to horseradish peroxidase at 1:10,000 (Amersham Biosciences). An enhanced chemiluminescence system (Amersham Biosciences) was used as recommended by the manufacturer to detect bound His-tag probe.
Yeast Two-hybrid Interaction Studies-The yeast two-hybrid interaction trap system was used according to established methods (38,39) to test for interactions between the ROMK2 COOH terminus and the PDZ domains in NHERF-1 and NHERF-2. In these studies, a LexA fusion protein of either a wild-type ROMK COOH terminus (amino acids 307-372) or a truncated mutant ROMK COOH terminus (aminoacids 307-369X) was used as bait (in pJK202 containing a His selectable marker). Conditionally expressed NHERF PDZ domains in the pJG4-5 plasmid (Trp selectable marker) were used as prey. Saccharomyces cerevisiae (EGY48, Mat a-ura3, his3, trp1, ura3, 3lexAop-leu2) was transfected with the wild-type and mutant baits and the PDZ domains and then plated onto selection plates to screen for yeast that were transfected with prey, baits, and reporter constructs. In this system, the interaction of library and bait proteins causes the transcriptional activation of lacZ. To determine the relative strength of the interactions, ␤-galactosidase activity was measured at V max using the o-nitrophenyl ␤-D-galactopyranoside (ONPG) solution assay according to established methods (40). Because the interaction trap uses Galinducible promoter to express the prey protein conditionally, yeast were grown (A 600 of 0.9 -1.2) under conditions that induce (2% galactose and 1% raffinose) or repress (2% glucose) the promoter. Yeast cultures were pelleted (2,000 ϫ g) and resuspended in 750 l of Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , and 50 mM ␤-mercaptoethanol). For the assay, 100 l of the Z buffer cell suspension was added to 900 l of Z buffer in glass culture tubes containing 10 l of 0.1% SDS and 20 l of chloroform. Tubes were vortexed for 15 s and equilibrated at 30°C for 15 min. After equilibration, 200 l of stock 4 mg/ml ONPG was added to the samples and vortexed for 5 s. Reactions were stopped at 30 s-4 min by adding 0.5 ml of 1 M Na 2 CO 3 , centrifuged to pellet debris, and the reaction product was measured by spectrofluorometery. Miller units were calculated as described previously (41). Results represent the average of at least four independent clones, repeated in duplicate.
cRNA Synthesis-Complementary RNA was transcribed in vitro in the presence of capping analogue G(5Ј)ppp(5Ј) from linearized plasmids containing the cDNA of interest using SP6 RNA polymerase (mMessage Machine, Ambion Inc.). cRNA was purified by phenol-chloroform extraction and precipitated with ammonium acetate/isopropyl alcohol. Yield was quantified spectrophotometrically and confirmed by agarose gel electrophoresis.
Oocyte Isolation and Injection-Oocytes from selected female Xenopus laevis (Xenopus Express, Homosassa, FL) were isolated and maintained using the standard procedures as described previously (42). Although the recombinant ROMK channel, as expressed in Xenopus oocytes, has been reported to by some investigators to exhibit sensitivity to the sulfonylurea K ATP channel blocker glibenclamide (43), we and others (44) have found that the response is modest, highly variable, and dependent on oocyte donor, suggesting the involvement of inconsistently expressed factors in the oocyte (44). Consequently, to study the effects of CFTR and NHERF on ROMK in the present study, we used a stock of frogs that was prescreened for lack of glibenclamide effects on expressed ROMK. Briefly, frogs were anesthetized with 0.15% 3-aminobenzoate and a partial oophorectomy was performed through an abdominal incision. Oocyte aggregates were manually dissected from the ovarian lobes and then incubated in OR-2 medium (82.5 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , and 5 mM HEPES, pH 7.5) containing collagenase (type 3, Worthington) for 2 h at room temperature to remove the follicular layer. After extensive washing with collagenase-free OR-2, oocytes were stored at 19°C in OR-3 medium (50% Leibovitz medium, 10 mM HEPES, pH 7.4). 12-24 h later, healthy looking Dumont stage V-VI oocytes were injected pneumatically with 50 nl of diethyl pyrocarbonate-treated water containing ROMK2, CFTR, and/or NHERF-2 cRNA, and then stored in OR-3 medium at 19°C for 2-6 days.
Electrophysiology-Whole cell currents in Xenopus oocytes were monitored using a two-microelectrode voltage clamp as described previously (42,45) under conditions in which potassium currents, carried through ROMK, could be isolated. In all studies, we purposely excluded endogenous cAMP analogs or activators to prevent CFTR Cl Ϫ channel activation. Briefly, oocytes were bathed in a 5 mM potassium, low Cl solution (5 mM KCl, 45 mM sodium gluconate, 1 mM MgCl 2 . 1 mM CaCl 2 , 5 mM HEPES, pH 7.4). Voltage sensing and current injecting microelectrodes had resistances of 0.5-1.5 megohms when back-filled with 3 M KCl. Once a stable membrane potential was attained, oocytes were clamped to a holding potential of Ϫ80 mV, and currents were recorded over a continuous 0.2-Hz train of 500-ms voltage steps to O mV, permitting a constant monitor of the outward potassium conductance. Periodically, the train was interrupted briefly to determine an IV relationship, monitoring the steady-state current over voltage 500-ms steps ranging from Ϫ120 mV to ϩ80 mV in 20-mV increments. Data were collected using an ITC16 analog to digital, digital to analog converter (Instrutech Corp.), filtered at 1 kHz, and digitized on line at 2 kHz using Pulse software (HEKA Electronik) for later analysis. Barium-sensitive outward current (10 mM barium acetate) was measured before and after the addition glibenclamide. The barium-sensitive current in the absence of glibenclamide was considered an estimate of the total ROMK current. The fraction of this current that was glibenclamide-sensitive was considered to represent the fraction of active ROMK channels that functionally interact with CFTR.

RESULTS
Identification of NHERF as a ROMK Binding Partner-As a first step toward testing the hypothesis about ROMK-PDZ protein interactions, we sought to identify PDZ proteins in the kidney which specially interact with the ROMK channel. In these studies GST fusion proteins of either the extreme COOH terminus of ROMK2 (326 -372) or a mutant ROMK COOH terminus, lacking the PDZ binding motif (326 -369X), were constructed. Binding proteins specific for the GST-ROMK COOH terminus affinity column were purified from rat kidney extracts, resolved by SDS-PAGE, and visualized by silver staining (Fig. 1). The wild-type COOH terminus but not the mutant COOH terminus (369X) uniquely bound several species, indicative of PDZ proteins. The most predominant and consistently observed protein migrated at ϳ50 kDa (n ϭ 3). Based on the molecular mass, we suspected that this protein could be NHERF, a PDZ protein known to interact with CFTR (20,23). Western blot analysis proved that this was indeed the case; as shown in Fig. 1C, an antibody raised against NHERF (34) specifically detected a ϳ50-kDa protein purified on the wild-type COOH terminus but not the mutant COOH terminus (369X).
To validate that the ROMK COOH terminus interacts with NHERF-1 or NHERF-2 rather than an immunologically related protein, GST affinity chromatography studies as above were conducted with extracts from COS-7 cells transfected with either HA-tagged NHERF-1 or HA-tagged NHERF-2. As detected by anti-HA antibodies in Western blot analysis, the wild-type COOH terminus but not the mutant COOH terminus (369X) bound to HA-tagged NHERF-1. Similar observations were made with HA-tagged NHERF-2 (not shown). Thus, the COOH terminus of ROMK, containing a type I PDZ binding motif, is capable of binding directly to NHERF-1 or NHERF-2.
To authenticate further ROMK-NHERF interaction, coim-munoprecipitation studies were performed with full-length ROMK and NHERF expressed in a cellular environment. In these studies, COS-7 cells were cotransfected with HA-tagged NHERF (either NHERF-1 or -2) and ROMK or transfected separately with either construct alone. Recovered immunoprecipitates on anti-HA antibody-bound beads were resolved by SDS-PAGE and immunoblotted with an anti-ROMK antibody (35). As shown in Fig. 2, ROMK copurified with NHERF-1 or NHERF-2 in cells transfected with ROMK and NHERF. ROMK could not be immunoprecipitated with an unrelated IgG, and coprecipitation required cotransfection of HA-tagged NHERF and ROMK, providing evidence of specific immunoprecipitation of ROMK and NHERF. Collectively, these studies reveal that ROMK is capable of interacting directly with either NHERF-1 or NHERF-2 in cells.

NHERF Facilitates Cell Surface Expression of ROMK-
The ROMK channel is not efficiently expressed on the plasma membrane in mammalian expression systems, accumulating largely within the endoplasmic reticulum (46,47), and possibility the Golgi apparatus (Fig. 3). Although reminiscent of Kir 6.2 channels, which are retained in the endoplasmic reticulum in the absence of the SUR binding partner (48,49), ROMK is different in that exit from the endoplasmic reticulum is not facilitated by SUR (43) or CFTR, suggesting a distinct processing mechanism. Because intracellular trafficking of membrane proteins can depend on PDZ interactions, we determined whether NHERF coordinates plasmalemma expression of ROMK. In initial studies, COS-7 cells were cotransfected with GFPtagged ROMK2 and NHERF-1 or NHERF-2 and visualized by fluorescent confocal microscopy (Fig. 3). In most cells, NHERF-1 and NHERF-2 colocalized with ROMK2 within the endoplasmic reticulum and Golgi. In some cells, ROMK2 also colocalized with NHERF in dense clusters or patches along segments of the cell periphery (Fig. 3). The redistribution of ROMK to the outer cell border upon NHERF cotransfection is FIG. 3. ROMK2 colocalizes with NHERF-1 and NHERF-2 in a PDZ-dependent manner. Immunolocalization of ROMK2 (green) and either NHERF-1 or NHERF-2 (red) in COS-7 cells transfected with wild-type ROMK2, ROMK2, lacking the PDZ binding motif, 369X, and NHERF-1/2. Comparable results were observed in three separate transfections. consistent with NHERF-dependent recruitment or retention of the channel on the plasmalemma. To verify that this observation actually represents an authentic increase in cell surface expression, COS-7 cells were cotransfected with NHERF and an extracellular HA-epitope tagged ROMK channel so that channels at the cell surface could be more accurately detected by external HA antibody binding (46). As assessed by immunofluorescent confocal microscopy of nonpermeabilized COS cells, cotransfection of either NHERF-1 or NHERF-2 dramatically increased the number of cells that express the external HA epitope-tagged EGFP-ROMK2 on the cell surface (Fig. 4). Although the majority of cells expressed ROMK on the cell surface in the presence of NHERF, the amount of ROMK on the cell surface was variable. To verify the NHERF-dependent plasmalemma expression of ROMK using an independent biochemical test, we quantified the amount of HA epitope on the cell surface by analytical luminometry. These studies revealed that NHERF expression modestly stimulated plasmalemma expression of ROMK2 above background. Removal of the PDZ binding motif in ROMK2 (369X) prevented colocalization of ROMK2 with NHERF and completely abrogated the NHERFdependent expression of ROMK2 on the plasma membrane. Thus, NHERF-1 or -2 increases cell surface expression of ROMK2 by a process dependent on a PDZ interaction.

ROMK Interacts Differentially with the PDZ Domains in NHERF-1 and NHERF-2-Both
NHERF-1 and NHERF-2 have tandem PDZ domains. Present evidence strongly suggests the two domains have different binding specificities (50), providing a potential mechanism to link disparate proteins that preferentially interact with different domains. For instance, the PDZ ligand at the COOH terminus of CFTR binds to the first PDZ domain of NHERF-1 with higher affinity than the second (29), possibility freeing the second PDZ domain to interact with other proteins, such as ROMK. To test this hypothesis, we first measured the capacity of ROMK to interact with each PDZ domain in NHERF-1 using an in vitro protein-protein interaction assay and by yeast two-hybrid. Fig. 5 illustrates the results of the in vitro interaction assay. In this study, purified GST-ROMK COOH-terminal fusion proteins were resolved by SDS-PAGE, immobilized on nitrocellulose, renatured, and probed for interaction with purified, Histagged NHERF-1 protein fragments. These included the NH 2 -terminal two-thirds of NHERF containing the two PDZ domains (PDZ1_PDZ2 amino acids 1-247), the first PDZ domain alone (PDZ1, amino acids 1-102) and the second PDZ domain alone (PDZ2, amino acids 144 -247). After hybridization, the GST fusions were washed and blotted with an anti-His antibody to detect bound probe. Significant interaction was only observed between the wild-type GST-ROMK protein and the PDZ 1-2 or PDZ 2 protein fragment, consistent with specific interaction between the ROMK COOH terminus and the second PDZ domain of NHERF-1. Binding was lost when the last three amino acids of ROMK (369X) were removed, consistent with a canonical PDZ interaction.
The binding preference of ROMK for the second PDZ domain in NHERF-1 was corroborated independently using the yeast two-hybrid system. In these studies, yeast were transfected with activation-tagged fusions of the NHERF-1 PDZ domains and LexA fusions of either the wild-type ROMK COOH terminus or the 369X mutant. As assessed by the transcriptional activation of the ␤-galactosidase gene reporter and quantified using the ONPG solution assay, interaction was observed only between the wild-type ROMK COOH terminus and either PDZ1-2 or PDZ2 (Fig. 6). No reporter activity was detected in yeast grown under conditions that repress expression of the NHERF domain. Furthermore, removal of the last three amino acids of ROMK (369X) completely abrogated the interaction. Thus, ROMK preferentially binds to the second PDZ domain of NHERF-1 through a canonical type-I PDZ interaction.
Although the PDZ domains in NHERF-1 and NHERF-2 are highly conserved, studies on NHERF-1 interaction with CFTR (29) combined with the observation of Sun et al. (23) that CFTR binds to the second PDZ domain better than first PDZ domain of NHERF-2 suggest that the PDZ domains in NHERF-1 and NHERF-2 may exhibit subtle functional differences. Accordingly, we compared the relative strength of interaction between the ROMK COOH terminus and the different PDZ domains in NHERF-1 and NHERF-2 using the yeast two-hybrid system (Fig. 7). As assessed by the transcriptional activation of the ␤-galactosidase gene reporter and quantified using the ONPG solution assay, the ROMK COOH terminus bound to the second PDZ domain of NHERF-2 with the same affinity as the second PDZ domain of NHERF-1. Surprisingly, however, the first PDZ domain of NHERF-2 also interacted with the COOH terminus FIG. 4. Cotransfection of NHERF-1 or NHERF-2 with ROMK2 increases cell surface expression of ROMK2. A, fraction of nonpermeabilized, EGFP-positive COS that exhibited extracellular anti-HA staining as assessed by confocal microscopy. COS cells were cotransfected with a double tagged wild-type or 369X mutant ROMK2, containing an EGFP on the cytoplasmic NH 2 terminus and a extracellular HA epitope, and vector alone or NHERF-1 or NHERF-2. B, exposed HA epitope on the cell surface of COS cells that were cotransfected either with wild-type or 369X mutant ROMK channels, containing an extracellular HA epitope, and vector alone, or NHERF-1 or NHERF-2. Cell surface ROMK expression was assessed by HA antibody binding and quantified by analytical luminometry. Background was assessed in COS cells transfected with wild-type ROMK that did not contain a HA epitope. Surface expression is reported in luminescence units (RFU)/g of total cell protein. *, p Ͻ 0.05. of ROMK. In fact, ϳ4-fold greater reporter activity was induced by the interaction between ROMK and NHERF-2 PDZ1 than with NHERF-2 PDZ2 or NHERF-1 PDZ2.
NHERF Facilitates Physical and Functional Interaction between ROMK and CFTR-The studies above, demonstrating that ROMK has a different binding preference for the PDZ domains in NHERF-1 and NHERF-2 than those reported for CFTR, indicate that NHERF has the capacity to act as a scaffolding molecule, which could potentially assemble ROMK and CFTR into a ternary complex. To determine whether this is actually the case, we tested whether NHERF increases the physical interaction between ROMK and CFTR. Because NHERF-2 appears to have a higher binding affinity for ROMK than NHERF-1, and our previous studies indicate that NHERF-2 colocalizes with ROMK2 in the kidney (51), we focused on the effects of NHERF-2. Physical interaction was examined biochemically by coimmunoprecipitation in COS cells transfected with different combinations of ROMK2, CFTR, and NHERF-2 (Fig. 8). In the absence of NHERF-2, a small amount of ROMK2 protein copurified with CFTR, as immunoprecipitated with anti-R CFTR antibodies, consistent with a weak intrinsic interaction between ROMK and CFTR. More importantly, and as predicted by our hypothesis, coexpression of NHERF-2 increased the physical interaction of CFTR with ROMK, as evidenced by the dramatic increase in the amount of ROMK2 that coimmunoprecipitates on anti- CFTR-bound beads. Thus, NHERF-2 facilitates physical interaction between ROMK and CFTR.
We and others (14,15) have observed that ROMK channels expressed in Xenopus oocytes develop a low affinity sensitivity to sulfonylurea agents when coexpressed with relatively large amounts of CFTR. Consequently, we quantified the extent to which CFTR modifies ROMK in the presence and absence of NHERF-2 by measuring the fraction of active ROMK channels that acquire glibenclamide sensitivity (Fig. 9). In these studies outward ROMK potassium currents were isolated in Xenopus oocytes under a two-microelectrode voltage clamp following the coinjection of combinations of CFTR, ROMK, and NHERF-2 cRNA. Glibenclamide sensitivity was assessed at a dose that is approximately 1 order of magnitude greater than the microscopic K i (15). Consequently, the fraction of outward potassium current (10 mM barium-sensitive current at 0 mV) that was also glibenclamide-sensitive was taken as an estimate of the portion of the total active ROMK population that interacts at a functional level with CFTR. When equal molar CFTR and ROMK cRNA are expressed in the absence of NHERF-2, only a small fraction of the active ROMK current was found to be glibenclamide-sensitive. Coexpression of NHERF-2 dramatically increased the fraction of ROMK current that is glibenclamidesensitive, indicating that NHERF-2 facilitates functional coupling between CFTR and ROMK, consistent with a molecular scaffold. DISCUSSION In the present study, we identify the Na/H exchange regulatory factors, NHERF-1 and NHERF-2, as PDZ domain binding partners of the ROMK channel. Characterization of the biochemical basis and functional consequences of NHERF-ROMK interaction reveals a coupling mechanism for linking ROMK to modifier proteins, expanding the role of the NHERF family.

FIG. 6. ROMK interacts differentially with the NHERF-1 PDZ domains in a yeast two-hybrid assay.
Yeast were cotransfected with activation domain-tagged NHERF-1 PDZ domains or an unrelated protein, SHC, and either a LexA fusion of the wild-type ROMK COOH terminus or a LexA fusion of a mutant ROMK COOH terminus, lacking the PDZ binding motif (369X). The strength of the ONPG ␤-galactosidase solution assay is reported in Miller units.
The coregulatory role appears to be widespread with a growing body of data demonstrating that the NHERF family of proteins facilitate multiprotein signaling complex assembly for efficient phosphorylation and regulation of a variety of transporters, channels, and receptors (for review, see Refs. 26 and 27). The "transducsome" organization function is supported by the domain architecture of these proteins. Both NHERF-1 and NHERF-2 contain two PDZ domains and a COOH-terminal Merlin/Ezrin/Radixin/Mosein (MERM) binding domain, allowing recruitment and local organization of MERM actin-binding proteins and kinase scaffolds (56,57), the regulatory subunit of protein kinase A II (23, 58 -60) and other kinases (61,62) with specific PDZ domain-binding protein receptors (31,63), phosphoacceptors, and other signal transduction molecules (64 -66). Recent work has suggested the possibility that the tandem PDZ domains in the NHERF proteins might facilitate another scaffolding function, in which regulatory or channel proteins interact with one PDZ domain and modulate the activity of similar proteins that simultaneously bind to the other PDZ domain (67,68). Our study provides evidence for such a mechanism, revealing that NHERF proteins can coordinate the assembly of a ternary complex, containing the ROMK channel and CFTR.
The heterophilic tethering function is presumably affected by the tandem nature and differential binding properties of the PDZ domains in the NHERF proteins. Characterization of consensus binding sequences of isolated NHERF-1 PDZ domains by phage display, affinity selection techniques revealed that the two PDZ domains have different ligand binding specificities (50) with distinct preferences for residues at the 0, Ϫ1, and Ϫ3 positions (the COOH-terminal residue is defined as the 0 position) of type 1 PDZ ligands (recognized by the X-S/T-X-I/V/L/ M-COOH motif). Our observations that the ROMK COOH terminus, containing a type 1 PDZ binding motif, preferentially binds to the second PDZ domain in NHERF-1 and to the first PDZ domain in NHERF-2 are consistent with this concept and suggest that a PDZ domain might be available for concurrent interaction with a disparate binding partner. Of note, the carboxyl-terminal position in ROMK is occupied by a methionine residue; whereas generally atypical for a type 1 binding sequence, phage display selection studies indicate that second PDZ domain (but not the first PDZ domain) in NHERF-1 prefers peptide binding sequences with a methionine (or leucine) at the zero position. Importantly, the relative binding preference of ROMK is precisely opposite of what has been reported for CFTR. Although the COOH terminus of CFTR is capable of interacting with both PDZ domains in the NHERF proteins (29), it more favorably interacts with PDZ1 in NHERF-1 (29,69) and PDZ2 in NHERF-2 (23). Collectively these observations indicate that both NHERF forms have the capacity to coordinate ROMK-CFTR assembly.
The PDZ domains in NHERF-1 and NHERF-2 are highly conserved, exhibiting more similarity among equally positioned domains in the two NHERF proteins than with domains in the same NHERF form (75% identity in the first PDZ domains of NHERF-1 and NHERF-2 and 78% identity in the second PDZ domain). Consequently, it is curious that our observations with ROMK along with previous studies of CFTR-NHERF interaction (23,29) suggest that the like-PDZ domains in the two NHERF proteins have subtle differences in binding preference. Although the precise basis for high affinity binding to each of the NHERF PDZ domains remains largely unknown, the crystal structures of the first NHERF-1 PDZ domain in complex with the COOH-terminal regions of CFTR (70), the ␤-adrenergic receptor and the platelet-derived growth factor receptor (71) provides some clues. Crystallographic analyses of these structures reveal that the penultimate residues of each ligand are engaged in numerous interactions with residues in PDZ1. Although each different ligand interacts with nearly the identical set of residues, several key interacting residues can undergo large ligand-dependent conformational changes to accommodate the ligand for favorable binding. As a consequence, the ordered water molecules and hydrogen bond networks that stabilize the PDZ1-ligand interactions are different for each ligand (72). Based on these observations, it seems likely that differences in ligand binding preferences among the different NHERF family PDZ domains are determined, at least in part, by the differences in the flexibility of the binding pockets, particularly because the key ligand-interacting residues in NHERF-1 PDZ1 are highly conserved across other NHERF PDZ domains.
Our observation that coexpression of NHERF-2 with ROMK and CFTR dramatically increases the amount of ROMK protein that coimmunopurifies with CFTR is consistent with the scaffolding capacity of NHERF-2 and strongly suggests that NHERF-2 facilitates assembly of a ternary complex containing ROMK and CFTR. In principle, formation of such a complex is determined by the relative abundance of the three components, absolute affinities of ROMK and CFTR for the NHERF PDZ domains, and the number of binding sites in ROMK and CFTR which are available for interaction with NHERF. Like other FIG. 9. NHERF-2 increases CFTR-dependent modification of ROMK. Glibenclamide-sensitive outward potassium currents were measured in Xenopus oocytes injected with ROMK2 and CFTR and ROMK2, CFTR and NHERF-2. A, currents were measured during a train of continuous 0.2-Hz train of 500-ms voltage steps from Ϫ80 to 0 mV. Periodically, the train was interrupted briefly to determine an IV relationship (*), monitoring the steady-state current over voltage 500-ms steps ranging from Ϫ120 to ϩ80 mV in 20-mV increments. B, the amount of ROMK current that functionally interacts with CFTR was measured as the total barium-sensitive current that acquires glibenclamide sensitivity.
Kir channels (73), ROMK channels are produced by the tetrameric arrangement of ROMK subunits, providing four PDZ domain docking sites/functional potassium channel. Thus, the complex could potentially assemble as a dodecamer with up to four CFTR molecules interacting with each functional potassium channel through four NHERF molecules. NHERF proteins can dimerize (74,75) as well as interact with the actin cytoskeleton through the MERM actin-binding proteins (61), increasing the potential for complexity. On the other hand, because ROMK can interact with both PDZ domains in NHERF-2, there is a possibility, subject to spatial constraints, for NHERF-2 to cross-link ROMK subunits and/or tetramers and thereby limit the formation of a complex containing all there components. Recent observations by Raghuram et al. (77) that interaction of CFTR with the PDZ2 of NHERF-1 is negatively regulated by phosphorylation of residues in the PDZ2 domain raise the intriguing possibility that the composition and stoichiometry of NHERF complexes are dynamic. Because cross-linking two CFTR molecules through bivalent PDZ domain interactions stimulates CFTR channel gating (22,29), regulated or competitive replacement of one CFTR with ROMK on the NHERF scaffold could switch the prevailing function of CFTR from a Cl Ϫ channel to a conductance regulator.
The involvement of NHERF proteins in CFTR-ROMK assembly might be regarded as a dramatic departure from the mechanism by which the ATP-binding cassette proteins, SUR1 and SUR2A/B, interact with Kir 6.1 and Kir 6.2 to form pancreatic islet ␤-cell and cardiac K ATP channels. In these cases, assembly is solely governed by direct interactions between the transmembrane and cytoplasmic domains of the ATP-binding cassette protein and the potassium Kir channel subunits (48,49). It should be pointed out, however, that we cannot exclude the possibility that similar direct interactions do not occur between ROMK and CFTR. In fact, previous studies, documenting functional modification of ROMK by CFTR (14,78,79) even in the excised patch-clamp configuration (15), coupled with present observations that ROMK coimmunoprecipitates with CFTR in the absence of exogenous NHERF, imply that ROMK is capable of interacting directly with CFTR. Perhaps by simultaneously binding ROMK and CFTR, NHERF proteins simply bring channel components into close proximity to promote low affinity intersubunit interactions.
The requirements for proper trafficking of ROMK channels to the plasma membrane appear to be much different from the prototypical K ATP channels, however. SUR and Kir 6.2 subunits contain endoplasmic reticulum retention/retrieval signals, which are masked upon full assembly of the octameric (4 SUR⅐4 Kir 6.2) KATP channel complex, providing a quality control mechanism that ensures that only completely assembled channels are delivered to the plasma membrane (48). Although ROMK largely accumulates within the endoplasmic reticulum in mammalian expression systems (46), reminiscent of Kir 6.2 channels, exit from the ER is not facilitated by coexpression of its ATP-binding cassette partners, CFTR or SUR (80). Instead, our work indicates that either NHERF-1 or NHERF-2 can stimulate cell surface expression of ROMK. With observations that anterograde trafficking of the protransforming growth factor in the secretory pathway is dependent on interaction with the PDZ protein, syntenin (81), and studies suggesting that plasma membrane-directed traffic of NR1 NMDA receptors in the biosynthetic trafficking pathway requires PDZ-protein interactions to conceal ER retention signals (82,83), it is tempting to speculate that a similar process is at play with ROMK-NHERF. However, other mechanisms must be seriously considered. Certainly, present evidence indicates that NHERF proteins orchestrate downstream trafficking steps such as postendocytic recycling (84,85) and anchoring target proteins on the plasma membrane (20,86). Taken together with our observations that NHERF-dependent trafficking of ROMK is remarkably variable and that other factors, such as direct phosphorylation of the channel by SGK-1, protein kinase A (46) and c-Src (87,88), also modulate cell surface expression of ROMK, it seems likely that efficient expression of ROMK on the plasmalemma involves a series of disparate trafficking processes. For instance, our previous studies with a related channel, Kir 2.3, indicate that interaction with the PDZ protein complex, Lin 7/CASK, specifies a plasma membrane retention step (89) in a hierarchical plasma membrane trafficking program (76). Obviously, further studies are required to determine whether and how NHERF interaction colludes with other intracellular trafficking processes to facilitate efficient expression of ROMK to the cell surface.
Previously we reported that NHERF-2, rather than NHERF-1, colocalizes with ROMK in the rat kidney (51), providing important evidence that this particular NHERF isoform is relevant to ROMK renal physiology. NHERF-2 is expressed predominantly with ROMK in the collecting duct, but not the thick ascending limb of Henle's loop. Cell type-specific expression of ROMK modifier proteins and adaptor molecules, like NHERF-2, may play an important role in the differential regulation of ROMK channels along the nephron. Importantly, the activity and surface density of functional ROMK channels in the collecting duct, but not the thick ascending limb, is exquisitely regulated by aldosterone and other factors to maintain renal potassium secretion in concert with the demands of potassium homeostasis (8). Germane to the observations in the present study with CFTR, studies in CFTR knockout mice indicate the native ROMK channels in the kidney maintain constitutive gating properties but lose sensitivity to glibenclamide and cytoplasmic ATP (30). It will be interesting to determine whether genetic ablation of NHERF-2 produces a similar phenotype.
In conclusion, we have provided evidence that ROMK binds to NHERF-1 and NHERF-2 through a canonical type 1 PDZ interaction to facilitate association with CFTR and to coordinate cell surface expression of the channel. Our observations raise the possibility that similar PDZ-based interactions may recruit other channel modifiers and accessory proteins into multimeric ROMK complexes for physiological regulation and targeting of the channel on specific plasma membrane domains.