Identification of γ-Aminobutyric Acid Receptor-interacting Factor 1 (TRAK2) as a Trafficking Factor for the K+ Channel Kir2.1*

To identify proteins that regulate potassium channel activity and expression, we performed functional screening of mammalian cDNA libraries in yeast that express the mammalian K+ channel Kir2.1. Growth of Kir2.1-expressing yeast in media with low K+ concentration is a function of K+ uptake via Kir2.1 channels. Therefore, the host strain was transformed with a human cDNA library, and cDNA clones that rescued growth at low K+ concentration were selected. One of these clones was identical to the protein of unknown function isolated previously as γ-aminobutyric acid receptor-interacting factor 1 (GRIF-1) (Beck, M., Brickley, K., Wilkinson, H., Sharma, S., Smith, M., Chazot, P., Pollard, S., and Stephenson, F. (2002) J. Biol. Chem. 277, 30079-30090). GRIF-1 specifically enhanced Kir2.1-dependent growth in yeast and Kir2.1-mediated 86Rb+ efflux in HEK293 cells. Quantitative microscopy and flow cytometry analysis of immunolabeled surface Kir2.1 channel showed that GRIF-1 significantly increased the number of Kir2.1 channels in the plasma membrane of COS and HEK293 cells. Physical interaction of Kir2.1 channel and GRIF-1 was demonstrated by co-immunoprecipitation from HEK293 lysates and yeast two-hybrid assay. In vivo association of Kir2.1 and GRIF-1 was demonstrated by co-immunoprecipitation from brain lysate. Yeast two-hybrid assays showed that an N-terminal region of GRIF-1 interacts with a C-terminal region of Kir2.1. These results indicate that GRIF-1 binds to Kir2.1 and facilitates trafficking of this channel to the cell surface.

Electrical activity in excitable tissues largely relies on tight control of K ϩ channel expression. The physiological relevance of changes in K ϩ channel expression has been demonstrated in a variety of normal and pathological states in heart, brain, and pancreas. It has been found that K ϩ channel protein synthesis, assembly, trafficking, and turnover all contribute to the steady state density of K ϩ channels on the cell surface (reviewed in Ref. 1). The molecular mechanisms underlying this regulation are just beginning to emerge.
Mutation studies identified determinants in K ϩ channel proteins that affect their cell surface expression. Although some of these determinants arrest K ϩ channel surface delivery by protein misfolding (2), others encode information on K ϩ channel processing through the secretory pathway (reviewed in Ref. 3). The latter include endoplasmic reticulum retention (4), endoplasmic reticulum export (5,6), and Golgi export (7) signals identified in inwardly rectifying K ϩ channels. It is presumed that these determinants interact with their cognate protein trafficking receptors to facilitate delivery of K ϩ channels to the cell surface. For example, current candidates for such partners include the vesicle coat protein COPI and 14-3-3, which responds to arginine-based folding/trafficking signals in Kir channels (8,9). Further understanding of K ϩ channel trafficking signals critically depends upon identification and analysis of their specific binding partners.
The majority of proteins that govern K ϩ channel surface expression were discovered on the basis of their physical interaction with K ϩ channels. Examples include GM130 (10), calnexin (11), KChAP (12), PSD93 (13), and PSD95 (14), which were identified in yeast two-hybrid system, and MAGUK proteins, which were cloned using proteomics (15). However, a broader scope of K ϩ channel partners may be revealed by functional cloning. This approach can be based on the expression of functional mammalian Kir2.1 channels in yeast; the Kir2.1 channel complements the K ϩ transport defect in mutant yeast lacking the endogenous K ϩ transporters TRK1 and TRK2 and so allows growth at low K ϩ concentrations (16). Thus, Kir2.1dependent growth of the ⌬trk1 trk2 yeast strain at selective K ϩ concentrations can be used for screening for factors that upregulate Kir2.1-dependent K ϩ uptake. This approach was used for identification of structural (17,18), functional (19), and trafficking (9) determinants in K ϩ channels. Furthermore, we used Kir2.1 expression in yeast for screening of chemical libraries for K ϩ channel modulators (20). Here, using this system we screened the cDNA library for mammalian proteins to identify the ␥-aminobutyric acid (GABA) 3 receptor interacting factor-1 (GRIF-1) (21) as a regulator of Kir2.1 activity. We then show that GRIF-1 is a binding partner of Kir2.1 that facilitates trafficking of this channel to the cell surface in mammalian cells.
Yeast Cultivation-K ϩ -free medium contained 2% glucose, 0.16% yeast nitrogen base without amino acids, sodium, and ammonia, 8 mM H 3 PO 4 , 2 mM MgSO 4 , 0,2 mM CaCl 2 , 0.1% vitamin mix, 0.1% trace elements mix, 0.2% adenine, de-ionized water (adapted from Ref. 16). The media were adjusted to pH 6.5 with arginine base and sterilized by filtering. KCl was added to desired concentration. For preparation of plates with low [K ϩ ] agarose instead of agar was used. Yeast medium ingredients were from Qbiogene. Yeast transformation was performed using the alkali cation method (Qbiogene). For growth assay, logarithmic cultures with starting density of OD 600 ϭ 0.05 were incubated at 30°C in a shaker incubator for 48 h. Culture growth was monitored spectrophotometrically at ϭ 600 nm.
Screening of cDNA Library-Host strain iKir2.1 was transformed with a library of human brain cDNA in pGAD vector (Stratagene). Transformants were selected on medium lacking uracil and leucine supplemented to 100 mM KCl and replica plated on medium without uracil, leucine, and methionine supplemented to 0.5 mM KCl. cDNA clones from growing colonies were extracted and tested for a connection with the Kir2.1 channel.
Cell Lines-COS and HEK293 cells were grown in Dulbecco's modified Eagle's medium with 4.5 g/liter glucose and 0.02 glutamine, supplemented with 10% fetal bovine serum, 100 g/ml penicillin, 100 g/ml streptomycin at 37°C and 5% CO 2 . Kir2.1 and GRIF cDNA in pCDNA3 vector were transfected into COS and HEK293 cells using Lipofectamine reagent. HEK293 cell line stably expressing Kir2.1 channel tagged with the HA epitope was a gift of Dr. Min Li. Stable HEK293 transfectants were maintained in the presence of 200 g/ml G418. 86 Rb ϩ Flux Assay-HEK293 cells stably expressing Kir2.1-HA channels were plated on polylysine-coated plates and transformed with GRIF-containing and control plasmids. 24 -48 h after transfection cells were loaded with 86 Rb ϩ for 4 h by incubation with 1 Ci/ml 86 RbCl. Prior to 86 Rb ϩ measurements cells were washed two times with 86 Rb ϩ -free medium. Then medium was collected at 1-40 min time points, and the amount of released 86 Rb ϩ was measured by scintillation counting. At 40 min, the cells were lysed with 1% SDS, and residual radioactivity in cells was determined.
Quantitative Microscopy-COS cells were co-transfected with Kir2.1 channel tagged with GFP at the N terminus and extracellular HA epitope (GFP-Kir2.1-HA, gift from Dr. Nikolaj Klocker) and GRIF-containing or control plasmids. Twentyfour hours after transfection, live cells were blocked with 10% normal goat serum in PBS (NGS) and incubated with primary rat anti-HA antibody 3F10 (Roche Applied Science) diluted 1:500 in 10% NGS. Following wash with PBS cells were incubated with the goat anti-rat antibody conjugated with Texas Red (Jackson Immunoresearch Labs) diluted 1:1000 in 2% NGS. Incubations with antibodies were performed at room temperature for 30 min; PBS washings were performed three times for 5 min each. The cells were visualized at 40ϫ magnification with inverted microscope (Olympus IX70). The cells were illuminated with 488-and 543-nm light. GFP and Texas Red signals, corrected to background, were quantified in captured images of individual cells using SimplePCI software.
Flow Cytometry-HEK 293 cells stably expressing Kir2.1 channel tagged with extracellular HA epitope were transiently transfected with GRIF-containing or control plasmids. Twenty-four hours after transfection, the cells were washed with PBS and harvested by incubation with 0.5 mM EDTA in PBS for 5-10 min at room temperature. Following wash with Hanks' balanced salt solution supplemented with 5 mM Hepes, pH 7.3, and 2% fetal bovine serum, the cells were incubated with rat anti-HA monoclonal antibody 3F10 diluted 1:500 on ice for 1 h. The cells then were washed twice with Hanks' balanced salt solution and incubated with Cy2-labeled goat anti-rat antibody diluted 1:1000 (Jackson ImmunoResearch) for 15 min on ice. Finally, the cells were washed twice with Hanks' balanced salt solution, and fluorescence was measured by FACSCalibur (Becton Dickinson). The percentage of fluorescent cells was analyzed using CELLQUEST software (Becton Dickinson).
Immunoprecipitation and Western Blot-HEK293 cells were plated in 100-mm culture dish at the density of 3ϫ10 6 /dish and the next day co-transfected with Kir2.1 and FLAG-tagged GRIF constructs. After 24 h of incubation, the cells were washed with PBS and lysed with 500 l/dish radioimmune precipitation assay buffer (150 mM NaCl, 10 mM Tris, 0.1% SDS, 1% Nonidet P-40, 1% deoxycholate, 5 mM EDTA, protease inhibitor mixture, pH 7.2) for 30 min at 4°C. The cell lysates were clarified by centrifugation at 14,000 rpm for 10 min. The supernatants were incubated with 30 l anti-FLAG resin (Sigma) for 1 h at 4°C. The supernatants were aspirated, and the pellets were gently rinsed three times with 1 ml of lysis buffer. The immune complexes were eluted with 50 l of 100 g/ml FLAG peptide in lysis buffer, mixed with Laemmli buffer, boiled for 10 min, and analyzed by immunoblotting. Following electrophoresis in 10% SDS-polyacrylamide gel proteins were transferred to polyvinylidene difluoride membrane. The membranes were blocked with 5% nonfat dry milk in PBS with 0.5% Tween (PBST) for 1 h at room temperature and incubated overnight at 4°C with polyclonal anti-Kir2.1 antibody (Alomone) diluted 1:1000. After double washing with PBST, the membranes were incubated with horseradish peroxidase-conjugated secondary antibody. Western blots were developed using high sensitivity Pierce ECL kit. Complexes of GRIF-FLAG with the Kir2.1 C terminus tagged with Myc were precipitated using anti-FLAG resin as above and probed with anti-Myc antibody (Abcam) diluted 1:3000 for 1 h at 4°C. The cells transfected with GRIF 497 -Xpress and Kir2.1 C terminus tagged with Myc were lysed as above. Clarified lysates were incubated with 2 l of rabbit polyclonal anti-X-press antibody (Invitrogen) overnight at 4°C, and immunocomplexes were captured using 50 l/sample protein A/G Plus-agarose reagent (Santa Cruz Biotechnology) for 2 h at 4°C. The beads were washed four times with 1 ml of lysis buffer, mixed with Laemmli buffer, and incubated for 10 min at 95°C for elution. The eluates were subjected to 8% SDS-polyacrylamide gel electrophoresis and analyzed by immunoblotting with monoclonal anti-Myc antibody (Abcam) diluted 1:3000.
For immunoprecipitation of native Kir2.1 and GRIF proteins, the brain was harvested from two adult mice and immediately homogenized in 5 ml of Buffer D (20 mM Hepes, pH 7.6, 125 mM NaCl, 10% glycerol, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride) (15) supplemented with protease inhibitors (Pierce). Homogenate was lysed in buffer D supplemented with 1% Triton X-100 for 1 h at 4°C and clarified with centrifugation at 25,000 ϫ g for 20 min. For immunoprecipitation, the supernatant was incubated with 5 g of affinity-purified anti-Kir2.1/2.2 antibodies or control preimmune rabbit IgG (kindly provided by Dr. Carol Vandenberg) overnight at 4°C. Immune complexes captured on 30 l of protein A-agarose were collected by sedimentation and dried by aspiration. The pellets were gently washed three times in buffer D using centrifugation at 800 ϫ g at room temperature and subjected to immunoblotting. Anti-GRIF-1 antibody (a gift from Prof. Anne Stephenson) was used as in Ref. 21.
Immunofluorescent Cell Staining-COS cells were plated on glass coverslips and transfected with epitope-tagged GRIF-1-FLAG and Kir2.1-HA constructs. Twenty-four hours after transfection the cells were fixed in 4% paraformaldehyde in PBS for 30 min at room temperature, permeabilized by incubation in 0.5% Triton X-100 in PBS for 15 min at room temperature, and blocked with 20% NGS in PBS for 20 min at room temperature. The cells were incubated with rabbit anti-FLAG antibody (Cell Signaling) diluted 1:400 and mouse anti-HA 12CA5 antibody (Roche Applied Science) in 20% NGS overnight at 4°C. Following wash with PBS cells were incubated with the secondary goat anti-rabbit FITC-conjugated antibody (Upstate) diluted 1:500 in 5% NGS, washed with PBS, incubated with secondary goat anti-mouse antibody conjugated with Cy3 diluted 1:500 (Jackson Immunoresearch Labs) in 5% NGS, and washed with PBS. Incubations with secondary antibodies were performed for 30 min at room temperature. PBS washes were performed three times for 5 min each at room temperature. The specimens were mounted in mounting medium (Invitrogen) and observed with fluorescent inverted microscope Olympus under 40ϫ magnification in 488-and 543-nm light. No cross-reactivity was detected between 12CA5 antibody and anti-rabbit FITC conjugate or rabbit anti-FLAG antibody and anti-mouse Cy3 conjugate.

Isolation of GRIF-1 as an Enhancer of Kir2.1-dependent Yeast
Growth-Growth of a yeast ⌬trk1 trk2 double mutant at low K ϩ concentration is rescued by expression of mammalian Kir2.1 channels (Fig. 1A). Kir2.1 expressed from a single chromosomally integrated copy restores growth at 2 mM KCl but not at 0.5 mM KCl (Fig. 1A, sector 3). In contrast, transformation with a 2 micron-based multicopy Kir2.1 plasmid enables growth at

GRIF-1 Enhances Kir2.1 Surface Expression
0.5 mM KCl (Fig. 1A, sector 2). Using this dosage effect, we sought to identify mammalian cDNAs that enable single copy Kir2.1-dependent growth at 0.5 mM KCl and thus are likely to encode positive regulators of Kir2.1. The single copy Kir2.1 strain, iKir2.1, was transformed with a human cDNA yeast expression library, and transformants able to grow at 0.5 mM KCl were selected. To confirm cDNA dependence of the growth phenotype, library plasmids were rescued from yeast transformants in Escherichia coli, and their ability to confer yeast growth at 0.5 mM KCl was tested upon retransformation into the iKir2.1 strain. One of the confirmed cDNAs, KP39, promoted growth at 0.5 mM KCl (Fig. 1, A,  compare sector 4 with sector 3; and B, left panel). However, KP39 did not increase growth at 0.5 mM KCl of the ⌬trk1 trk2 strain in the absence of Kir2.1 (Fig. 1A, sector 5). Likewise, it did not improve growth at a high, nonselective K ϩ concentration (Fig. 1B), a condition in which Kir2.1 is not necessary for growth. Therefore, the KP39-encoded protein does not generally increase growth. Rather, the facilitation of growth by KP39 is coupled to Kir2.1 channel function.
Sequencing of KP39 identified a 1491-bp cDNA fragment exactly corresponding to the 5Ј half of the open reading frame of a known gene ALS2CR3. The ALS2CR3 gene maps to the 2q33-2q34 human chromosome locus associated with human juvenile amyotrophic lateral sclerosis (22). A causal relationship between ALS2CR3 and amyotrophic lateral sclerosis has not been established, and the function of this gene in humans remains unknown (22). A rat ortholog of ALS2CR3 protein has been identified in yeast two-hybrid system as GABA-receptor interacting factor-1 (GRIF-1, also named TRAK2) (21). KP39 contains the first 497 codons of the ALS2CR3 open reading frame, and its deduced amino acid sequence is 86.6% identical with the N-terminal half of rat GRIF-1. Thus, we now refer to KP39 as hGRIF 497 . Functionally, the full-length rat GRIF-1 cDNA was indistinguishable from hGRIF 497 in its ability to enhance Kir2.1-dependent yeast growth (Fig. 1, A, sector 6, and C).
GRIF-1 Increases Kir2.1-mediated 86 Rb ϩ Efflux in Mammalian Cells-To examine effects of GRIF-1 expression on Kir2.1 function in mammalian cells, we used the 86 Rb ϩ flux assay. Because hydrated Rb ϩ ion is similar in permeability to K ϩ , 86 Rb ϩ flux can be used as an indicator of K ϩ channel activity. HEK293 cells stably expressing Kir2.1 were transiently transfected with GRIF-1 constructs or corresponding empty vectors. The cells were then loaded with 86 Rb ϩ and washed, and the time course of 86 Rb ϩ release was determined. In these assays, channel activity is revealed in the initial rate of release. As seen in Fig. 2A, initial efflux of 86 Rb ϩ from cells transfected with full size GRIF-1 or hGRIF 497 fragment  occurs faster than from vector-transfected cells. Measuring 86 Rb ϩ efflux from multiple independent transfections proved the statistical significance of this difference (Fig. 2B). Overall, GRIF-1 expression in 40% of the cells increased total Kir2.1mediated 86 Rb ϩ flux by 12% Ϯ 1.7%. Thus, channel activity was increased by ϳ30% by GRIF-1. In contrast, transfection of GRIF-1 constructs into HEK293 cells that do not express Kir2.1 had no effect on the background 86 Rb ϩ efflux, which was ϳ10% of that in Kir2.1-transfected cells (data not shown). Therefore, we conclude that GRIF expression increases Kir2.1 channel activity.
GRIF-1 Increases Surface Expression of Kir2.1-To address the mechanism of GRIF-enhanced K ϩ flux, we examined surface expression of Kir2.1 with or without ectopic expression of GRIF constructs. HEK293 cells stably expressing Kir2.1 channel with HA epitope at the first extracellular loop were transiently transfected with full size GRIF-1, hGRIF 497 or empty vectors, immunostained for surface HA epitope, and analyzed by flow cytometry. Fig. 3A shows that GRIF-1 and hGRIF 497 each increase mean surface channel immunofluorescence. With both GRIF constructs, surface density of Kir2.1 channels was elevated by ϳ30% (Fig. 3B), which agrees with the increase of 86 Rb ϩ efflux in this cell line. Additionally, we examined the GRIF effect on Kir2.1 surface expression using quantitative microscopy. COS cells were transiently co-transfected with GRIF-1 or corresponding empty vector, as well as with Kir2.1 tagged with extracellular HA epitope and GFP at the N terminus. Live transfected cells were immunolabeled with anti-HA antibody and Texas Red-conjugated secondary antibody to mark surface channels. Red and green fluorescence corresponding to surface and total Kir2.1 signals, respectively, in individual cells were measured using quantitative fluorescence microscopy. The Texas Red signal was then normalized by the GFP signal to yield a measure of Kir2.1 surface expression. As shown in Fig. 3C, cells transfected with GRIF-1 had a significantly stronger surface Kir2.1 signal. Notably, total cellular Kir2.1 channel protein expression, measured by GFP-Kir2.1 fluorescence (Fig. 3D) and Western blot (Fig. 4A), was not affected by GRIF. Thus, full-length GRIF-1 and the N-terminal half of GRIF-1 each increase surface localization of Kir2.1 in two different cell lines without affecting total Kir2.1 protein expression. Taken together, these results demonstrate that  Kir2.1 (180 -428)-Myc was co-transfected with GRIF (1-497)-X-press. The lysates were precipitated using anti-X-press antibody and immunoblotted with anti-Myc antibody. D, yeast two-hybrid analysis of Kir2.1-GRIF interaction. Yeast strain YRG-2 was co-transformed with C terminus of Kir2.1 fused to GAL4 AD and GRIF-1 and hGRIF 497 fused to GAL4 BD. Control fusions SV40-AD and p53-BD interact in yeast two-hybrid assay. Growth of transformants on media without histidine indicates protein-protein interaction. E, co-immunoprecipitation of native brain Kir2 channels and GRIF-1. Detergent-solubilized protein from whole murine brain was precipitated with anti-Kir2.1/Kir2.2 antibody or control preimmune IgG and probed with anti-GRIF-antibody. GRIF-1 increases K ϩ transport through Kir2.1 by increasing surface expression of this channel.
GRIF-1 and Kir2.1 Associate Physically-The effects of GRIF-1 on Kir2.1-dependent yeast growth, Kir2.1 activity and surface expression may result from physical association between these two proteins. To test this possibility, we examined GRIF-1/Kir2.1 association with co-precipitation and yeast two-hybrid assays. For the co-precipitation assay, HEK293 cells were transiently transfected with pair-wise combinations of epitope-tagged Kir2.1 and GRIF-1 constructs. GRIF proteins were affinity-purified or immunoprecipitated from cell lysates, and the presence of Kir2.1 protein was examined by Western blotting. Fig. 4A shows that full-length Kir2.1 co-precipitated with full-length GRIF-1. The C-terminal cytoplasmic domain of Kir2.1 (amino acids 180 -428) was sufficient for this association (Fig. 4B). This domain also co-immunoprecipitated with the N-terminal half of GRIF-1 (Fig. 4C). Thus, immunoprecipitation experiments indicated that GRIF-1 associates with the C-terminal cytoplasmic domain of Kir2.1 and that the N-terminal half of GRIF-1 is sufficient for this association.
To confirm data obtained using immunoprecipitation, we used the yeast two-hybrid assay. A cytoplasmic domain of Kir2.1 (amino acids 180 -428) was inserted into the GAL4 tran-scription activation domain (AD) vector, and full-length GRIF-1 or its N-terminal half (amino acids 1-497) were inserted into the GAL4 DNA-binding domain (BD) vector. AD and BD constructs were then co-transformed into the YRG-2 yeast strain whose growth in the absence of histidine depends on functional Gal4 transcriptional activator. If Kir2.1 and GRIF-1 associate, this should bring AD and BD portions of Gal4 in close proximity, thus reconstructing functional Gal4 and allowing growth of YRG-2 in the absence of histidine. As shown in Fig. 4D, the cytoplasmic domain of Kir2.1 associates with full-size GRIF-1 as well as with its N-terminal half. This association is similar in strength to that between the positive control pair, SV40-AD and p53-BD. None of the GRIF constructs showed association with SV40-AD. Similarly, the Kir2.1 C terminus fused to AD did not confer histidine-independent growth alone or in combination with p53-BD, demonstrating the specificity of Kir2.1-GRIF-1 association (Fig. 4D). Thus, yeast two-hybrid results confirm biochemical results, and both sets of data indicate specific association of the C-terminal cytoplasmic domain of Kir2.1 with the N-terminal half of GRIF-1. Finally, to determine whether the Kir2.1-GRIF-1 interaction occurs in vivo, we attempted co-immunoprecipitation of these two proteins from murine brain. Fig. 4E (IP Kir2) shows that precipitate collected with anti-Kir2.1/Kir2.2 antibodies contains GRIF-1 protein. In contrast, control precipitate collected with nonspecific preimmune serum gave no signal (Fig. 4E,  mock IP). Thus, native Kir2 and GRIF proteins associate in vivo, indicating the physiologic relevance of the interaction originally discovered with heterologous expression.
Kir2.1 and GRIF-1 Co-localize in COS Cells-Co-immunoprecipitation and yeast two-hybrid interaction experiments indicate that Kir2.1 and GRIF-1 associate in vitro and in yeast cells. To examine association between these two proteins in mammalian cells, we transiently transfected COS cells with HA-tagged Kir2.1 and FLAG-tagged GRIF-1 and examined intracellular localization of both proteins by fluorescence microscopy with anti-HA and anti-FLAG antibodies. In cells transfected with GRIF-1 alone, this protein was diffusely localized throughout the cell, as reported previously by Beck et al. (21) (Fig. 6A). Kir2.1 immunofluorescence was observed routinely in perinuclear aggregates and punctate structures in the cytosol, consistent with localization to different compartments of the secretory pathway (Fig. 6B). Co-expression with GRIF-1 did not cause noticeable change in the pattern of Kir2.1 localization (data not shown). However, the subcellular distribution of GRIF-1 did change in the presence of Kir2.1. In addition to diffuse localization, GRIF-1 immunoreactivity was concentrated in punctate structures characteristic of Kir2.1 pattern (Fig. 6C). In these punctate structures, GRIF-1 often co-localized with Kir2.1 (Fig. 6D). Therefore, co-expression of Kir2.1 changed the pattern of subcellular distribution of GRIF-1. Moreover, Kir2.1 and GRIF-1 co-localized, consistent with association in mammalian cells.

DISCUSSION
Here we report a novel protein partner of the mammalian K ϩ inwardly rectifying channel Kir2.1. This protein is encoded by the ALS2CR3 gene mapping in the human juvenile amylotro-phic lateral sclerosis critical region, 2q33-2q34. The rat ortholog of human ALS2CR3 protein, GRIF-1, was identified recently as a binding partner of the ␤-subunit of GABA receptor (21). Based on GRIF-1 binding to kinesin and its homology with proteins implicated in vesicular and organelle motility, a role for GRIF-1 as a trafficking factor has been proposed (23). We identified a direct effect of GRIF-1 to promote K ϩ channel surface expression and activity.
Human and rat GRIF proteins enable growth at low K ϩ concentration of a ⌬trk1 trk2 yeast strain expressing a single copy of Kir2.1. Enhancement by GRIF-1 of Kir2.1-dependent K ϩ flow is not a yeast-specific phenomenon because similar enhancement occurs in mammalian cells (HEK293), as indicated by 86 Rb ϩ flux data. In mammalian cells, ectopic expression of full-length or the N-terminal half of GRIF-1 increases surface expression but not total Kir2.1 protein expression. This suggests that GRIF-1 increases K ϩ currents via increased delivery of Kir2.1 to the plasma membrane. Kir2.1 and GRIF-1 not only functionally interact, they also co-precipitate from mammalian cell lysates, associate in the yeast two-hybrid assay, and co-localize when expressed in mammalian cells, all indicating specific physical association. Furthermore, co-immunoprecipitation from brain indicates that native Kir2.1 and GRIF-1 proteins associate in vivo. The site responsible for association with GRIF-1 is located in the cytoplasmic tail of Kir2.1 between residues 348 and 396. This 49-amino acid-long region contains signals critically important for Kir2.1 trafficking from endoplasmic reticulum to plasma membrane, as well as putative coiled coil, amphipatic helix, and O-GlcNAc modification sites. Based on these findings, we propose that GRIF-1 acts as a binding partner of Kir2.1 that facilitates trafficking of this K ϩ channel through the secretory pathway.
GRIF-1 belongs to the TRAK family of coiled coil domaincontaining proteins implicated in trafficking of vesicles and mitochondria. The family includes the O-GlcNAc transferaseinteracting protein OIP (24); Milton, a protein transporting mitochondria to neuronal synapses in Drosophila (25); and TRAK1, a protein required for GABA receptor homeostasis in mice (26). All of these proteins specifically bind the heavy chain of the microtubule-associated motor protein kinesin (23,25). It has been proposed that GRIF/TRAK proteins provide a link between specific cargo proteins or organelles and kinesin motors (23,25,26). A trafficking adaptor role for GRIFs finds further support in the fact that the N-terminal half of GRIF-1 contains a stretch of 300 amino acids that is 47% homologous to the huntingtin-associated protein HAP-1 (27), which plays a role in trafficking of proteins and organelles in neurons by linking them to kinesin (28). Therefore, it is possible that GRIF-1 (TRAK2) links Kir2.1 to kinesin heavy chain, which is important for delivery of K ϩ channels to the cell surface (29).
The role of GRIF-1 as a trafficking factor for Kir2.1 is also supported by its binding to a segment of Kir2.1 that contains structural features relevant for protein-protein interaction and Kir2.1 trafficking signals (Fig. 5C). First, the 350 -380 region of Kir2.1 contains regularly spaced hydrophobic amino acids that may form an amphipatic ␣-helix, and the 376 -389 region is predicted to assume a coiled coil structure (29). A coiled coil structure is also located in the N-terminal half of GRIF-1 and is essential for interaction with GABA receptor (21). Thus, it is possible that coiled coil domain in Kir2.1 interacts with a similar structure in the N-terminal half of GRIF-1 and so contributes to Kir2.1-GRIF-1 association. Second, threonine-353 in the GRIF-binding region of Kir2.1 is a putative site of O-Glc-NAc modification (30). The addition of O-GlcNAc to serine or threonine residues is a labile modification that plays signaling role and may be an alternative to phosphorylation at the same site (reviewed in Ref. 31). Several proteins known to be modified by O-GlcNAc have functions that are relevant to the secretory pathway, including Golgi reassembly stacking protein 2 (GRASP55), cytoskeletal proteins, and microtubule assembly proteins MAP1B and MAP2B (32). This suggests that O-Glc-Nac signaling plays a role in protein trafficking. In light of the reported association of GRIF-1 with O-GlcNAc-transferase (24), it cannot be excluded that GRIF-1 facilitates the addition of O-GlcNac to Kir2.1 and thus provides a dynamic signal for the channel trafficking. Finally, the GRIF-1-binding segment of Kir2.1 contains at least two signals required for the efficient trafficking of this K ϩ channel to the plasma membrane. These signals, FCYENE and EEEEDSE, direct export from the endoplasmic reticulum and post-Golgi vesicles, respectively, and therefore are likely binding sites for distinct trafficking factors (6,7). FCYENE and EEEEDSE are necessary, but not sufficient for Kir2.1 surface expression because a larger segment (amino acids 369 -410), which incorporates both signals and overlaps with GRIF-1-binding site (amino acids 348 -396), is required to confer surface expression in a domain swap assay (6). Because FCYENE and EEEEDSE are parts of the essential GRIF-1-binding segment, it is possible that GRIF-1 plays a role in recognition of these signals. If GRIF-1 is a trafficking factor for Kir2.1, it is likely to act in concert with other factors that bind to the cytoplasmic tail of Kir2.1, such as actin-binding protein filamin A (33), vesicle coat proteins adaptins (34), or proteins with PDZ motifs (13,15). Preferential expression of GRIF-1 in heart, brain, and skeletal muscle (21) is also consistent with its role as a trafficking partner for Kir2.1, which is specifically expressed in these tissues.