INTRODUCTION:

Inwardly rectifying potassium (Kir) channels are a functionallydiverse family of proteins characterized by their ability toconduct large inward currents at potentials negative to thepotassium equilibrium potential and smaller outward currentsat more positive membrane potentials. Kir channels contributeto a wide variety of physiological functions such as vasculartone, heart rate, glial buffering of potassium, renal salt flow andinsulin release (Isomoto et al., 1997; Nichols and Lopatin,1997; Vandenberg, 1994). The classical strong Kir channels,Kir2.1, Kir2.2 and Kir2.3, are important in the modulation ofcell excitability, repolarization of the action potential anddetermination of the cellular resting potential.

Recent studies reveal that the function and localization ofmany ion channels are in part regulated by interactions withmembers of the MAGUK protein family. MAGUK proteins arecharacterized by their modular domain structure, whichcontains up to three PDZ domains, a src homology 3 (SH3)domain and a catalytically inactive guanylate kinase-likedomain (GK). All of these domains are involved in protein-protein interactions allowing MAGUK proteins to act asintracellular scaffolding molecules in the formation ofmacromolecular signaling complexes (Craven and Bredt, 1998;

Kornau et al., 1997; Pawson and Scott, 1997; Sheng andWyszynski, 1997). Various ion channel families, includingKirs, bind to the PDZ domains of selected members of theMAGUK family via their PDZ binding motif (T/S-X-V/I),found at the extreme C terminus (Cohen et al., 1996; Horio etal., 1997; Kim et al., 1995; Kornau et al., 1995). Specifically,Kir2.1 and Kir2.3 have been demonstrated to interact with theMAGUK protein PSD-95/SAP90 (Cohen et al., 1996; Nehringet al., 2000). However, in contrast to the widespreaddistribution of Kir2 family ion channels in neuronal and non-neuronal cells, the distribution of PSD-95 is restricted to thebrain. To date, no other MAGUK proteins have been identifiedin brain or other tissues that interact with any of the Kir2.xchannels. Given that other ion channel families can interactwith more than one MAGUK family member (Garcia et al.,1998; Horio et al., 1997; Kim et al., 1995; Kornau et al., 1995),it is possible that other members of the MAGUK protein familycould interact with the C-terminal PDZ binding motifs ofKir2.1, Kir2.2 and Kir2.3 in brain as well as other tissue types.

A candidate MAGUK family member for the association ofKir2 proteins in non-neuronal tissues is the ubiquitouslyexpressed protein SAP97/hDlg (Lue et al., 1994; Matsumineet al., 1996; Muller et al., 1995). Although MAGUK familymembers like PSD-95 and SAP97 share common domain

987

The strong inwardly rectifying potassium channels Kir2.xare involved in maintenance and control of cell excitability.Recent studies reveal that the function and localization ofion channels are regulated by interactions with members ofthe membrane-associated guanylate kinase (MAGUK)protein family. To identify novel interacting MAGUKfamily members, we constructed GST-fusion proteins withthe C termini of Kir2.1, Kir2.2 and Kir2.3. GST affinity-pulldown assays from solubilized rat cerebellum and heartmembrane proteins revealed an interaction between allthree Kir2.x C-terminal fusion proteins and the MAGUKprotein synapse-associated protein 97 (SAP97). Atruncated form of the C-terminal GST-Kir2.2 fusionprotein indicated that the last three amino acids (S-E-I) areessential for association with SAP97. Affinity interactionsusing GST-fusion proteins containing the modular domainsof SAP97 demonstrate that the second PSD-95/Dlg/ZO-1(PDZ) domain is sufficient for interaction with Kir2.2.

Coimmunoprecipitations demonstrated that endogenousKir2.2 associates with SAP97 in rat cerebellum and heart.Additionally, phosphorylation of the Kir2.2 C terminus byprotein kinase A inhibited the association with SAP97. Inrat cardiac ventricular myocytes, Kir2.2 and SAP97colocalized in striated bands corresponding to T-tubules. Inrat cerebellum, Kir2.2 was present in a punctate patternalong SAP97-positive processes of Bergmann glia in themolecular layer, and colocalized with astrocytes andgranule cells in the granule cell layer. These results identifya direct association of Kir2.1, Kir2.2 and Kir2.3 with theMAGUK family member SAP97 that may form part of amacromolecular signaling complex in many differenttissues.

Key words: Inward rectifier potassium channel, SAP97, MAGUK,Scaffolding protein, Protein kinase A, Cardiac myocyte, Astrocyte

SUMMARY

Inward rectifier potassium channel Kir2.2 isassociated with Synapse-Associated Protein SAP97Dmitri Leonoudakis, William S. Mailliard, Kevin L. Wingerd, Dennis O. Clegg and Carol A. Vandenberg* Department of Molecular, Cellular and Developmental Biology, and Neuroscience Research Institute, University of California, Santa Barbara, CA93106, USA*Author for correspondence (e-mail: vandenbe@lifesci.ucsb.edu)

Accepted 23 December 2000Journal of Cell Science 114, 987-998 © The Company of Biologists Ltd

RESEARCH ARTICLE

988

structures and show amino acid similarity, their functions,cellular localizations and binding interactions are often distinct(Cho et al., 1992; Kim and Sheng, 1996; Leonard et al., 1998;Muller et al., 1995), and interaction with one member of thefamily does not imply interaction with another. Indeed, SAP97and PSD-95 differ in their ability to cluster proteins, theirinteraction with other scaffolding proteins, and their ability tobind to ion channels (Kim and Sheng, 1996; Tiffany et al.,2000). Potential clues to the function of SAP97 come from thestudy of its Drosophilahomolog DlgA, which plays a role inthe proper formation of neuromuscular junctions in neuronsand the establishment of proper apicobasal polarity in epithelialcells (Budnik et al., 1996; Lahey et al., 1994; Tejedor et al.,1997). Recently, SAP97 was shown to form a direct interactionwith the A-kinase anchoring protein AKAP79/150, whichrecruited protein kinase A to the AMPA receptor GluR1, andenhanced phosphorylation of the channel (Colledge et al.,2000). Thus, SAP97 also plays a role in assemblingmacromolecular signaling complexes.

To address the hypothesis that other MAGUK proteinsinteract with the C-terminal PDZ binding motif of the stronginwardly rectifying potassium channels, we specifically testedfor the association of SAP97 with C-terminal GST-fusionproteins of Kir2.1, Kir2.2 and Kir2.3. We report here thatthese C-terminal fusion proteins associated with SAP97when incubated with solubilized membrane protein from ratcerebellum and heart. The association required the PDZbinding motif (S-E-I) at the C terminus of Kir2.2 and thesecond PDZ domain of SAP97. Coimmunoprecipitationexperiments indicate that endogenous Kir2.2 and SAP97associate. Additionally, protein kinase A phosphorylation ofthe Kir2.2 C terminus inhibited SAP97 association.Immunofluorescence microscopy revealed colocalization ofKir2.2 and SAP97 in the T-tubules of cardiac ventricularmyocytes. In the cerebellum, Kir2.2 and SAP97 are expressedand colocalize in glial cells of the molecular and granularcell layers. These experiments identify SAP97 as a Kir2.2-interacting protein in both the cerebellum and the heart.

MATERIALS AND METHODS

GST fusion proteinscDNAs encoding potassium channel C-terminal amino acids andSAP97 domains were generated by PCR and cloned in-frame to the3′ end of glutathione-S-transferase (GST) in the bacterial expressionvector pGEX2T (Amersham Pharmacia Biotech). The sequences ofall constructs were confirmed by dideoxy sequencing. The C-terminalsegments of the fusion constructs coded for Kir2.1-CT (mouse Kir2.1amino acids 376-428), Kir2.2-CT (rat Kir2.2 amino acids 362-427),Kir2.2-CT∆3 (rat Kir2.2 amino acids 362-424), Kir2.3-CT (humanKir2.3 amino acids 390-445), Kir6.2-CT (rat Kir6.2 amino acids 333-390) and TASK-CT (rat TWIK-related acid-sensitive K+ channelamino acids 361-411). Rat SAP97 (Muller et al., 1995) domain fusionproteins coded for the N-terminal domain (amino acids 1-220), PDZ1(amino acids 220-313), PDZ1 and PDZ2 (amino acids 220-420),PDZ2 (amino acids 314-420), PDZ2 and PDZ3 (amino acids 314-584)and the guanylate kinase-like domain (amino acids 729-912).

Fusion proteins were expressed in E. coli and purified onglutathione-agarose (Sigma). GST-fusion protein expression wasinduced with 0.5 mM isopropyl β-D-thiogalactoside (IPTG) (4 hours,37°C), bacteria were collected by centrifugation (3000 g, 10 minutes)and were washed once in phosphate-buffered saline (PBS). All protein

manipulations were performed at 4°C unless otherwise indicated. Thebacterial pellet was resuspended in PBS containing 1% Triton X-100.Bacteria were lysed by sonication, the debris pelleted bycentrifugation (12,000 g, 10 minutes) and the resulting supernatantincubated with glutathione-agarose (30 minutes). Beads were washedthree times with HNTG (1% Triton X-100, 10% glycerol, 150 mMNaCl, 20 mM Hepes, pH 7.5) followed by one wash in HNG (10%glycerol, 150 mM NaCl, 20 mM Hepes, pH 7.5) and resuspended withHNG to make a 50/50 slurry of buffer/beads. Purity was determinedby SDS-PAGE followed by Coomassie Blue staining. Fusion proteinbound to beads was divided into portions, flash-frozen in liquidnitrogen and stored at −80°C.

Antibodies and antiseraAffinity-purified rabbit anti-rat Kir2.2 polyclonal antibodies werepreviously described (Raab-Graham and Vandenberg, 1998). Anti-Kir2.1/2.2 rabbit polyclonal antibodies were generated to apeptide consisting of amino acids 5-22 of mouse Kir2.1(RTNRYSIVSSEEDGMKLA) (Raab-Graham and Vandenberg,1998). Other primary antibodies used include mouse monoclonal anti-SAP97 antibody (Transduction Laboratories), rat monoclonal anti-ZO-1 antibody (Chemicon), rabbit polyclonal anti-GRIP1 antibody(generous gift of Dr Richard Huganir), mouse monoclonal anti-mycantibody and horseradish peroxidase-linked anti-GST antibody (SantaCruz Biotechnology). Anti-rabbit IgG, anti-mouse IgG (AmershamPharmacia Biotech) or anti-rat IgG (Sigma) secondary antibodiesconjugated to horseradish peroxidase were used where appropriate.Additional antibodies used for microscopy are described below.

Cerebellum and heart crude membrane isolationTissues were harvested from adult rats, immediately frozen in liquidnitrogen and stored at −80°C. Tissues were homogenized with aglass/teflon homogenizer in 10 volumes of Buffer A (320 mM sucrose,5 mM EDTA, 5 mM EGTA, 1 mM PMSF, 1× Protease InhibitorCocktail (Roche Molecular Biochemicals), 20 mM Hepes, pH 7.5),except for heart, which was ground with a mortar and pestle underliquid nitrogen prior to homogenization. The homogenate wascentrifuged (800 g, 10 minutes) and the supernatant was saved onice. The pellet was rehomogenized and centrifuged as before. Thesupernatants were combined, centrifuged in a Beckman SW-28(100,000 g, 1 hour) and the pellets resuspended in Buffer A.

Coexpression in COS-1 cellsFull-length cDNAs encoding Kir2.2 (a generous gift of Dr YoshihisaKurachi) and myc-tagged SAP97 (a generous gift from Dr CraigGarner) were subcloned into the mammalian expression vectorpcDNA1/Amp (Invitrogen). The cDNA encoding myc-tagged PSD-95 was obtained in the mammalian expression vector GW1 (agenerous gift from Dr Craig Garner). COS-1 cells were grown andmaintained (10% CO2, 37°C) in DMEM-supplemented with 10% fetalcalf serum. COS-1 cells were cotransfected with myc-SAP97 andKir2.2 constructs utilizing FuGene 6 transfection reagent (RocheMolecular Biochemicals). 2 days post-transfection, cells wereharvested, washed twice with PBS, resuspended in Buffer A and lysedby passing through a 22-gauge needle (20 passes, on ice). Nuclei, celldebris and unbroken cells were pelletted by centrifugation (800 g, 10minutes). Crude membranes from the low-speed supernatant wereisolated by centrifugation in a Sorvall S55-S rotor (100,000 g, 30minutes) and resuspended in Buffer A. For experiments to examinethe effects of protein kinase A (PKA) stimulation, cotransfected COS-1 cells were incubated in culture medium supplemented with 20 µMforskolin and 100 µM 3-isobutyl-1-methylxanthine (IBMX) for 20minutes prior to harvesting, or with vehicle alone (dimethyl sulfoxide,0.2% final concentration). Buffer A was supplemented with 50 mMsodium fluoride, 1 mM sodium orthovanadate and 10 mM sodiumpyrophosphate for membrane isolation in these experiments to inhibitphosphatase activity.

JOURNAL OF CELL SCIENCE 114 (5)

989SAP97 associates with Kir2 channels

GST affinity-pulldown assayCerebellum, heart or COS-1 membranes (0.5 mg, 1 mg or 75 µg/GSTpulldown assay, respectively) were pelleted by centrifugation in aSorvall S-100AT3 rotor (165,000 g, 15 minutes). The membrane pelletwas solubilized (1 hour) by resuspending in HNNS buffer (1%IGEPAL CA-630, 0.1% SDS, 150 mM NaCl, 20 mM Hepes, pH 7.5,1× protease inhibitor cocktail) to a final protein concentration of 1mg/ml. Insoluble material was removed by centrifugation (100,000 g,10 minutes) and the soluble protein precleared with GST (20µg/pulldown) bound to glutathione-agarose. GST-fusion proteinprebound to glutathione-agarose (5 µg/tissue extract pulldown; 2.5µg/COS-1 extract pulldown) was added to the precleared solubilizedprotein and incubated overnight with rotation. Beads were collectedby centrifugation and washed three times with 1 ml HNNS. Beadswere resuspended in SDS-PAGE loading buffer and the elutedproteins resolved on an 8.75% SDS-PAGE gel. Proteins weretransblotted to Hybond ECL nitrocellulose membranes (AmershamPharmacia Biotech) and incubated with primary antibodies followedby peroxidase-conjugated secondary antibodies. Immunoreactiveproteins were detected by Supersignal West Dura enhancedchemiluminescent substrate (Pierce).

CoimmunoprecipitationPrior to solubilization, membranes were pelleted by centrifugation ina Sorvall S-100AT3 rotor (165,000 g, 15 minutes). Membrane proteinsolubilized in HNNS buffer (tissues: 2 mg/ml, COS-1 cells: 25 µg/ml,1 hour) was prepared as described above. Solubilized protein wasprecleared with rabbit preimmune serum (1 hour), followed by theaddition of protein A-agarose (Pierce) (30 minutes). Anti-Kir2.1/2.2immune serum or preimmune serum was added to the preclearedsolubilized protein and incubated for 1 hour. To collect immunecomplexes, protein A-agarose was added and incubated overnightwith rotation. Precipitated proteins were washed four times with 1 mlHNNS. Coimmunoprecipitated proteins were analyzed byimmunoblotting as described above. The peptide competitionexperiment was performed by preincubating anti-Kir2.1/2.2 immuneserum with immunogenic peptide (1 mg/ml, 1 hour, 37°C) prior to itsuse in immunoprecipitation as described above.

PKA phosphorylation of Kir2.2-CT10 µg of fusion protein prebound to glutathione-agarose wasequilibrated in kinase buffer (10 mM MgCl2, 50 mM Tris, pH 7.5).The phosphorylation reaction was carried out in kinase buffersupplemented with 1 mM ATP and 5 ng/µl PKA catalytic subunit(Sigma) (2 hours, 30°C). Phosphorylated protein bound to beads waswashed three times with 1 ml HNNS buffer supplemented withphosphatase inhibitors (0.1 mM sodium vanadate, 10 mM sodiumpyrophosphate and 10 mM sodium fluoride). Phosphorylated fusionprotein was subsequently used for GST pulldowns. Proteins wereresolved on a 10% SDS-PAGE gel and immunoblotted as describedabove.

Fluorescence microscopyThe following procedures were carried out at 4°C. Adult female Long-Evens rats were euthanized, tissue was dissected, rinsed with coldPBS and fixed in PBS with 4% paraformaldehyde (Fisher). Tissueswere rinsed with PBS and incubated overnight in PBS with 20%sucrose. Tissue was embedded in 7.5% gelatin (Sigma, 300 bloom),15% sucrose, and 0.1 M phosphate buffer (pH 7.4). Tissue/gelatinblocks were frozen with crushed dry ice and mounted on a cryostatstage with histoprep (Fisher). 10 µm frozen sections were collected(Leica CM 1850 cryostat) on charged microscope slides (Fisher) andstored at 4°C until use. Tissue sections were incubated with blockingbuffer (1% normal goat serum, 3% bovine serum albumin, 0.01%Triton X-100 in PBS) for 1 hour. Primary antibodies were diluted inblocking buffer, (rabbit anti-Kir2.2, 1:200; mouse anti-SAP97, 1:200;rabbit anti-SAP97 (Affinity Bioreagents), 1:1000; mouse anti-GFAP

conjugated to CY3, 1:3000 (Sigma)) added to sections and incubatedovernight. Primary antibodies were aspirated and the slides rinsedthree times 5 minutes with PBS. Secondary antibodies were dilutedin blocking buffer, (goat anti-mouse IgG conjugated to CY3 (JacksonImmunoResearch), 1:400; goat anti-rabbit IgG conjugated to Alexa488 (Molecular Probes), 1:100) added to sections, and incubated for30 minutes. Slides were rinsed three times for 5 minutes with PBS.All liquid was aspirated off tissue sections and slides were mountedwith Prolong (Molecular Probes). Images were captured on anOlympus BX60 epifluorescence microscope equipped with anOptronics CCD-camera (DEI-750) or a Zeiss standard epifluorescencemicroscope and images captured with a Nikon FX 35A 35mm cameraon color slide film followed by digitization on a Canon Canoscan FS2710 slide scanner. Digitally stored images were combined anddisplayed with Adobe Photoshop.

Cerebellar cell cultureA single P6 Long-Evens rat cerebellum was dissociated by passingthrough a 22-gauge needle in DMEM supplemented with 10% fetalcalf serum. Dissociated cells were plated on poly-D-lysine coatedcoverslips and incubated at 37°C, 10% CO2 for 1 week. Cells werefixed with 4% paraformaldehyde in PBS for 10 minutes and stainedas described above. Images were captured on a Bio-Rad MRC 1024laser scanning confocal microscope.

RESULTS

SAP97 interacts with the C termini of Kir2 channelsGST was fused to the C-terminal amino acids of Kir2.1(Kir2.1-CT) (Kubo et al., 1993), Kir2.2 (Kir2.2-CT) (Koyamaet al., 1994) and Kir2.3 (Kir2.3-CT) (Périer et al., 1994) (Fig.1A). These fusion proteins were incubated with detergent-solubilized membrane proteins from rat cerebellum or heart inaffinity interaction GST pulldown assays. Immunoblot analysiswith an anti-SAP97 antibody detected a protein of approx. 140kDa corresponding to SAP97 (closed arrowheads) incerebellum and heart membranes (Fig. 1B,C, top). Theantibody also detected a protein of approx. 95 kDa associatedwith Kir2.2-CT in cerebellum (open arrowhead, Fig. 1B, top)but not heart (Fig. 1C, top). The approx. 95 kDa protein couldalso be detected in the input lane and associated with Kir2.1-CT and Kir2.3-CT in cerebellum membrane extracts withlonger exposures (data not shown). We identified this approx.95 kDa protein as the brain-restricted MAGUK PSD-95, whichcrossreacted with the mouse anti-SAP97 antibody (data notshown). We observed differences in the amount of SAP97associated with the different C-terminal fusion proteins in bothcerebellum and heart membranes with the strongest associationbeing with Kir2.2-CT (Fig. 1B,C, top). In the cerebellum, PSD-95 mirrored SAP97 having the strongest association withKir2.2-CT (Fig. 1B, top). Immunoblotting using an anti-GSTantibody confirmed that equal amounts of fusion protein wereused and recovered in each of the pulldowns (Fig. 1B,C,bottom). The interaction with SAP97 was specific for the C-terminal amino acids of Kir2.1, Kir2.2 and Kir2.3, sinceneither GST nor Kir6.2-CT, which lacks a PDZ binding motif(Inagaki et al., 1995) (Fig. 1A), interacted with SAP97 (Fig.1B,C, top).

To test whether the presence of a C-terminal PDZ bindingmotif was sufficient for SAP97 association, we constructed aC-terminal fusion protein of TASK (TASK-CT) (Duprat et al.,1997; Leonoudakis et al., 1998), a tandem pore potassium

990

channel containing a putative PDZ binding motif (S-S-V) (Fig.1A). SAP97 and PSD-95 were not detected when cerebellumor heart membrane proteins were incubated with the TASK-CTfusion protein, despite the presence of an intact putative PDZbinding motif (Fig. 1B,C, top). To determine whether otherPDZ domain-containing proteins could interact with the C-terminal fusion proteins, we probed fusion protein pulldownsfrom solubilized cerebellum membrane proteins for thepresence of the MAGUK ZO-1 (Willott et al., 1993) and themultiple PDZ domain-containing molecule, GRIP1 (Dong etal., 1997). ZO-1 and GRIP1 were present in the cerebellumextracts, but were not detected in the affinity-pulldowns by anyof the fusion protein constructs (Fig. 1B, middle). These datademonstrate that the C termini of Kir2.1, Kir2.2 and Kir2.3 arespecifically recognized by SAP97.

Previous studies demonstrated that the T/S-X-V/I motif ofseveral ion channels, including Kirs, is required for interactionswith the MAGUK protein PSD-95 (Cohen et al., 1996; Kim etal., 1995; Kornau et al., 1995). Therefore, we predicted thatdeletion of the C-terminal PDZ-interacting motif (S-E-I) ofKir2.2 should abolish the ability of Kir2.2-CT to interact withSAP97. A GST fusion protein lacking the three most C-terminal amino acids (S-E-I) of Kir2.2 was constructed(Kir2.2∆3-CT). No detectable SAP97 was pulled down withKir2.2∆3-CT from either cerebellum or heart membraneproteins (Fig. 1B,C) demonstrating the requirement of an intactKir2.2 PDZ-binding motif (S-E-I) for association with SAP97.

The second domain of SAP97 is sufficient forinteraction Kir2.2To map the domain(s) of SAP97 required for interaction withKir2.2, GST was fused to the domains of SAP97 (Fig. 2A).Detergent-solubilized membrane proteins prepared fromKir2.2-transfected COS-1 cells were incubated with variousGST-SAP97 domain fusion proteins. Immunoblot analysiswith an anti-Kir-2.2 antibody revealed that the fusion proteinscontaining the first and second PDZ domains (PDZ1+PDZ2),the second PDZ domain (PDZ2) and the second and third PDZdomains (PDZ2+PDZ3) strongly interacted with Kir2.2 (Fig.2B, lanes 5, 6 and 7), whereas a fusion protein containing thefirst PDZ domain (PDZ1) interacted weakly with Kir2.2 (Fig.2B, lane 4). No association with Kir2.2 was observed with GSTalone, the N terminus of SAP97 (NT) or the guanylate kinase-like domain (GK) (Fig. 2B, lanes 2, 3 and 8). These dataindicate that the second PDZ domain (PDZ2) is sufficient forassociation with Kir2.2; they do not, however, rule out aninteraction with the third PDZ domain.

Kir2.2 and SAP97 are coexpressed in a variety oftissuesTo further examine the interaction between Kir2.2 and SAP97,we assessed the tissue distribution of both proteins. Ratmultiple tissue immunoblots were probed with antibodies toKir2.2 and SAP97. SAP97 was observed in all tissues analyzedwith the highest expression levels being in the cerebellum,cortex, kidney, lung, liver and pancreas (Fig. 3A). Kir2.2 wasexpressed at relatively equal levels in all tissues analyzed (Fig.3B). These protein distributions are in agreement with themRNA expression patterns of SAP97 (Matsumine et al., 1996;Muller et al., 1995) and Kir2.2 (Karschin et al., 1996; Koyamaet al., 1994). The lower band seen in the SAP97 immunoblot

JOURNAL OF CELL SCIENCE 114 (5)

Fig. 1.SAP97 expressed in rat cerebellum and heart specificallyassociates with the C termini of Kir2.1, Kir2.2 and Kir2.3.(A) Schematic representation of the constructs used in the GSTpulldowns. GST was fused to the C-terminal (CT) of Kir2.1, Kir2.2,Kir2.3, Kir2.2∆3, TASK, and Kir6.2. Numbers represent startingamino acids of each construct and the single letter code for theextreme C-terminal amino acids are shown. (B) Glutathione-agarosecharged with the respective GST-fusion protein was incubated withdetergent-solubilized rat cerebellum membrane proteins. Boundproteins were analyzed by probing individual immunoblots with theantibodies indicated to the left of each panel. To confirm that equalamounts of fusion protein were used and recovered, the anti-SAP97immunoblot was stripped and reprobed with an anti-GST antibody(B and C, bottom). The closed arrowhead indicates a band of approx.140 kDa representing SAP97 and the open arrowhead a band ofapprox. 95 kDa representing PSD-95, which is also recognized bythe anti-SAP97 antibody. The input lane contains 4% of protein usedin each binding reaction. (C) The GST-fusion protein affinity-pulldown assay was performed as in B, utilizing rat heartmembranes. Note that SAP97 can often be resolved into doubletsthat may arise from differential splicing (Muller et al., 1996; Mulleret al., 1995). The input lane contains 10% of protein used in eachbinding reaction. Positions of protein markers in kDa are given to theleft of each panel.

991SAP97 associates with Kir2 channels

of the cortex and cerebellum corresponds to PSD-95 (Fig. 3A).As expected (Cho et al., 1992), PSD-95 was not detected innon-neuronal tissues (Fig. 3A).

Full-length Kir2.2 interacts with SAP97To examine the ability of full-length Kir2.2 to interact withSAP97 within a cellular environment, COS-1 cells werecotransfected with mammalian expression vectors containingeither full-length Kir2.2 and myc epitope-tagged full-lengthSAP97 (myc-SAP97) or myc-SAP97 alone. myc-SAP97coimmunoprecipitated with anti-Kir2.1/2.2 serum (Fig. 4A,lane 3), but not with preimmune serum (Fig. 4A, lane 2) oranti-Kir2.1/2.2 serum that had been preincubated with theimmunogenic peptide (Fig. 4A, lane 4). Anti-Kir2.1/2.2 serum

did not coimmunoprecipitate myc-SAP97 from COS-1 cellstransfected with myc-SAP97 alone (Fig. 4B). This wasexpected since COS-1 cells do not contain endogenous Kir2.1or Kir2.2, and indicates that the anti-Kir2.1/2.2 antiserum didnot interact nonspecifically with SAP97. Interestingly, anti-Kir2.1/2.2 serum did not coimmunoprecipitate endogenousCOS-1 SAP97 from cells transfected with Kir2.2 alone, butrequired cotransfection with exogenous SAP97 (data notshown). These data show that Kir2.2 specifically associateswith SAP97 in a cellular context.

Do Kir2.x channels interact with SAP97 to form acomplex in tissues where both proteins are coexpressed?Coimmunoprecipitation experiments utilizing both ratcerebellum and heart membrane proteins were carried out. Inboth tissues, SAP97 specifically coimmunoprecipitated withanti-Kir2.1/2.2 serum, but not with preimmune serum (Fig. 5).Although PSD-95 was identified in Kir2.1, Kir2.2 and Kir2.3C-terminal GST-fusion protein affinity-pulldowns (Fig. 1B,top) and associated with Kir2.2 in COS-1 cells, no PSD-95was detected in by coimmunoprecipitation from cerebellummembrane extracts (data not shown). The nonionic detergentutilized in these experiments solubilizes PSD-95 poorly(Leonard et al., 1998; Muller et al., 1995) and may account for

Fig. 2. The second PDZ domain of SAP97 is sufficient forinteraction with Kir2.2. (A) Schematic representation of the SAP97domains, which were fused to GST for affinity interactionpulldowns. (B) Kir2.2 was expressed in COS-1 cells and thesolubilized membrane proteins utilized for pulldown assays withGST-SAP97 domain fusion proteins. Kir2.2 was detected with ananti-Kir2.2 antibody specific to the C terminus of Kir2.2. Equalfusion protein loads and recovery were confirmed by immunoblottingwith anti-GST, as in Fig. 1. The input lane contains 20% of proteinused in each binding reaction.

Fig. 3. Kir2.2 and SAP97 are coexpressed in a wide variety oftissues. Membranes (50 µg each) isolated from the indicated rattissues were resolved by SDS-PAGE, transblotted to nitrocelluloseand immunoblotted. (A) Probed with an anti-SAP97 antibody, whichrecognizes a SAP97 doublet at approx.140 kDa. The band observedin cortex and cerebellum at approx. 95 kDa corresponds to PSD-95.(B) Probed with anti-Kir2.2 antibody, which recognizes a Kir2.2band of approx. 50 kDa.

Fig. 4. SAP97 associates with Kir2.2 when coexpressed in COS-1cells. Detergent-solubilized membrane protein from COS-1 cellstransiently transfected with (A) Kir2.2 and myc-SAP97 or (B) myc-SAP97 alone were immunoprecipitated with antiserum to the Nterminus of Kir2.1/2.2. Myc-tagged proteins were detected with amonoclonal antibody to the myc epitope. For peptide block,Kir2.1/2.2 antiserum was pre-absorbed with antigenic peptide priorto immunoprecipitation (lane 4). The input lane (lane 1) contains 7%of the protein present in each immunoprecipitation.

Fig. 5. Kir2.2 and SAP97 associate to form a complex in ratcerebellum and heart. Detergent-solubilized membrane proteins fromeither rat cerebellum or heart were used for coimmunoprecipitationswith antiserum to the N terminus of Kir2.1/2.2. ImmunoprecipitatedSAP97 was detected by immunoblotting with a monoclonal antibodyto SAP97. The input lane contains 1% (cerebellum) and 3.5% (heart)of the protein used in each immunoprecipitation.

992

the absence of PSD-95 detected in coimmunoprecipitationsfrom cerebellum. These data demonstrate that endogenousKir2.x channels associate with SAP97 to form a complex inheart and cerebellum tissue.

Protein kinase A phosphorylation of Kir2.2 Cterminus inhibits interaction with SAP97A previous study (Cohen et al., 1996) found that PKAphosphorylation of a critical serine (Ser440) in the PDZ domainbinding motif of Kir2.3 (R-R-E-S-A-I ) inhibited interactionwith PSD-95. Like Kir2.3, Kir2.2 contains a PKA consensusphosphorylation site at the C terminus that overlaps the PDZbinding motif (R-R-E-S-E-I). Therefore, we hypothesized thatthe interaction between Kir2.2 and SAP97 might be regulatedby PKA. Addition of exogenous PKA catalytic subunit andATP to Kir2.2-CT resulted in a molecular mass shift to a larger,phosphorylated form (Fig. 6A, lower panel, lane 4). The siteof PKA phosphorylation was confirmed by PKA incorporationof 32P from [γ-32P]ATP into Kir2.2-CT but not Kir2.2∆3-CT,which lacks the PKA consensus serine (Ser425) (data notshown). The PKA-phosphorylated Kir2.2-CT fusion proteinshowed no detectable association with SAP97 from solubilizedheart membranes (Fig. 6A, upper panel, lane 4), whereas theunphosphorylated Kir2.2-CT readily associated with SAP97(Fig. 6A, upper panel, lane 3). These data suggest that PKAphosphorylation of Ser425at the C terminus of Kir2.2 regulatesthe interaction of Kir2.2 and SAP97.

Regulation of the association of Kir2.2 with SAP97 by PKAphosphorylation was assessed in COS-1 cells. PKA activitywas stimulated in COS-1 cells cotransfected with Kir2.2 andSAP97 by treatment with forskolin and IBMX for 20 minutesprior to cell lysis and membrane preparation. Thesemembranes were then used in coimmunoprecipitations (Fig.

6C). Forskolin treatment diminished the amount of SAP97coimmunoprecipitated with Kir2.2 (Fig. 6C, lane 5) comparedto control cells (Fig. 6C, lane 4). The input lanes (lanes 1and 2) confirm that equal amounts of protein were usedin immunoprecipitations. Solubilized membranes fromtransfected COS-1 cells were also used in affinity-pulldownexperiments with the PDZ1+2 domain of SAP97 (Fig. 6B).Treatment of the cells with forskolin and IBMX for 20 minutessimilarly inhibited the association of Kir2.2 with PDZ1+2 (Fig.6B, lane 5) compared to control cells (Fig. 6B, lane 4). Theseexperiments demonstrate that the interaction of Kir2.2 withSAP97 in a cellular context is dynamically inhibited (within20 minutes) by reagents that stimulate PKA activity.

Kir2.2 and SAP97 are colocalized in the T-tubules ofventricular myocytesIndirect immunofluorescence microscopy was used todetermine the localization of Kir2.2 and SAP97 in rat cardiacmyocytes. Longitudinal sections of ventricular myocytesdouble-labeled with anti-Kir2.2 (Fig. 7a,d) and anti-SAP97

JOURNAL OF CELL SCIENCE 114 (5)

Fig. 6.Protein kinase A (PKA) phosphorylation of Kir2.2 inhibitsassociation with SAP97. (A) Kir2.2-CT was incubated with (+) orwithout (−) the catalytic subunit of PKA in the presence of ATP (1mM). (Upper panel) GST alone, PKA-untreated and PKA-treatedKir2.2-CT were incubated with detergent-solubilized rat heartmembrane protein. SAP97 was detected by immunoblotting with amonoclonal antibody to SAP97. (Lower panel) The immunoblot wasthen stripped and reprobed with a monoclonal horseradishperoxidase-linked anti-GST antibody to confirm equal fusion proteininput and recovery. The asterisk indicates the shift in molecular massresulting from phosphorylation of Kir2.2-CT by PKA (lane 4). Theinput lane (lane 1) contains 5% of protein used in each bindingreaction. (B) COS-1 cells transfected with Kir2.2 and SAP97 wereincubated with forskolin (20 µM) and IBMX (100 µM) (+), orvehicle alone (−). Membranes were solubilized and used in affinity-pulldown assays with GST alone, or GST-fusion protein containingthe PDZ1+2 domains of SAP97. (Upper panel) Bound Kir2.2 wasdetecting by immunoblotting. (Lower panel) The blot was stainedwith Ponceau S to confirm equal fusion protein input and recovery.Input lanes contain 1% of protein used in the corresponding bindingreaction. (C) COS-1 cells cotransfected with Kir2.2 and SAP97 wereincubated with forskolin (20 µM) and IBMX (100 µM) (+), orvehicle alone (−). Solubilized membranes were thenimmunoprecipitated with preimmune or antiserum to the N terminusof Kir2.1/2.2. Immunoprecipitated SAP97 was detected byimmunoblotting with a monoclonal antibody to SAP97. Exposuretime in C for lanes 1-2 was 15 times longer than lanes 3-5. Inputlanes contain 1% of the protein used in the correspondingimmunoprecipitation.

993SAP97 associates with Kir2 channels

(Fig. 7b,e) antibodies showed prominent immunoreactivityin transverse striations spaced at approx. 2 µm intervalsthroughout the length of the myocytes (Fig. 7a-f; arrowheads;overlap = yellow). The spacing of this staining pattern isconsistent with the known distribution of the T-tubule systemat the level of the Z-lines in ventricular myocytes (Kostin et al.,1998). Image analysis demonstrated a clear overlap of Kir2.2and SAP97 in these transverse striations, suggesting their closeassociation in T-tubules (Fig. 7c,f; arrowheads). Kir2.2 labelingwas more intense on the T-tubules compared to the plasmamembrane, and appeared to be absent from intercalated disks(Fig. 7a,d). SAP97 labeling was observed in the T-tubules (Fig.7b,e; arrowheads), at the plasma membrane, and at intercalateddisks (arrows). To further confirm that the striated pattern oflabeling of ventricular myocytes was due to labeling of T-tubules, we examined atrial myocytes, in which the T-tubularsystem is reduced or absent. Transverse striations were notdetected in atrial myocytes labeled with antibodies to Kir2.2or SAP97, consistent with their colocalization in T-tubules(data not shown). Staining with fluorescent secondaryantibodies alone showed no significant labeling.

Kir2.2 and SAP97 colocalize in cerebellar glial cellsIndirect immunofluorescence microscopy was used todetermine the localization of Kir2.2 and SAP97 in ratcerebellar cortex. Sagittal cerebellar tissue sections weredouble-labeled with both anti-Kir2.2 and anti-SAP97antibodies. Kir2.2 staining revealed a punctate distribution inboth the molecular layer (ML) and the granule cell layer (GCL)(Figs 8, 9). Kir2.2 immunostaining in the ML was present aspunctate labeling along radial tracks originating near the

Purkinje cell (marked with asterisks) border and extendingtoward the pial surface (Fig. 8a,g), which corresponds to thedistribution of the radial processes of Bergmann glia. PunctateKir2.2 staining colocalized with astrocytes and granule cells inthe GCL (Fig. 8a,d). SAP97 also stained both granule cells andastrocytes in the GCL (Fig. 8b,e); the ML showed intense,uniform SAP97 staining (Fig. 8b,h). This agrees with previouselectron microscopy studies showing widespread SAP97localization in the granule cell parallel fibers in the ML (Mulleret al., 1995). In addition, distinct radial fiber tracks positive forSAP97 were observed in the ML (Figs 8b,h, 9Ad, arrowheads),and these fiber tracks were decorated with punctate Kir2.2staining (Figs 8g-i, 9Ad-f, arrowheads). In the Purkinje celllayer, cells were labeled uniformly with anti-Kir2.2 antibody(Fig. 8a,g, asterisks), and SAP97 staining appeared along theedges of Purkinje cells (Figs 8h, 9Ad, asterisks). PSD-95 alsois found in cerebellum (Cho et al., 1992; Kistner et al., 1993),and may have contributed some of the immunolabelingobserved using the mouse anti-SAP97 antibody in Fig. 8.Similar immunostaining was observed using a rabbit anti-SAP97 antibody that does not crossreact with PSD-95,confirming the cerebellar localization of SAP97 (comparesimilar staining patterns in Figs 8h and 9Ad). Controlexperiments with fluorescent secondary antibodies aloneshowed no significant labeling.

We postulated that Kir2.2 and SAP97 were coexpressed inBergmann glia, since Kir2.2 staining appeared as punctateradial fiber tracks that colocalized with SAP97 in the ML.Bergmann glia are unipolar radial astrocytes that associate withdeveloping granule cells and with mature Purkinje cells. Thecell body is located along the Purkinje cell layer and the

Fig. 7.Kir2.2 and SAP97 colocalize in the T-tubules of cardiac myocytes. Indirect immunofluorescence images of longitudinal sections of adultrat cardiac tissue. (a-f) Ventricular myocytes double-labeled with anti-Kir2.2 (a,d) and monoclonal anti-SAP97 antibodies (b,e) show striatedlabeling spaced at regular intervals of approx. 2 µm (arrowheads). This labeling is consistent with the known distribution of the T-tubule systemat the level of the Z-lines in ventricular myocytes. Composites of the images on the left are shown on the right (c,f). Colocalization of Kir2.2and SAP97 is observed as yellow on the composite images. Only SAP97 stains intercalated disks (arrows). Bar, 10 µm.

994

process radiates outwards to the pial surface (Rakic, 1995). Toexamine this question in more detail, sagittal cerebellar tissuesections were double-labeled with anti-glial fibrillary acidicprotein (GFAP) antibody (a glial astrocyte cell marker) andeither anti-Kir2.2 or anti-SAP97 antibodies. The radial fibersof Bergmann glia in the ML were clearly identified by GFAPimmunoreactivity, and these radial fibers corresponded to thetracts of Kir2.2 immunofluorescence (Fig. 9Aa-c). SpecificSAP97 staining in the ML (Fig. 9Ad) included radial fibers thatcolocalized with GFAP-positive Bergmann glia (Fig. 9Ad-f,arrowheads) as well as astrocytes in the GCL (Fig. 9Ad-f,arrows). Since many other cell types in the cerebellum appearto stain positively for SAP97, it is difficult to distinguishSAP97 staining of glial cells from other cell types. Therefore,dissociated and isolated cerebellar glial cells were culturedand double-labeled with anti-GFAP and Kir2.2 or SAP97.

Confocal microscopy revealed that GFAP-positive glial cellswith an astrocyte-like morphology indeed express both Kir2.2(Fig. 9Ba-c) and SAP97 (Fig. 9Bd-f).

In summary, Kir2.2 was clustered on cerebellar glia(Bergmann glia in the ML and astrocytes in the GCL) andgranule cells. SAP97 colocalized with Kir2.2 on these cells,with punctate Kir2.2 expressed in an overlapping distributionto a more uniform SAP97 expression pattern. Thus, SAP97colocalizes with Kir2.2 and is in a position in vivo to form asignaling complex with the channel.

DISCUSSION

We have demonstrated an interaction between the stronginward rectifiers Kir2.1, Kir2.2 and Kir2.3 and the MAGUK

JOURNAL OF CELL SCIENCE 114 (5)

Fig. 8.Kir2.2 and SAP97 are colocalized in the cerebellum. Indirect immunofluorescence images of sagittal sections of adult rat cerebellarcortex were used in double-labeling experiments. Tissue was double-labeled with anti-Kir2.2 (a,d,g) and mouse anti-SAP97 antibodies (whichalso recognize PSD-95) (b,e, h). Composites of the images on the left are shown on the right (c,f, i). (a-c) Images show the granule cell layer(GCL) and molecular layer (ML) of the cerebellar cortex. (d-f) Higher power image of the GCL. Insets in d-f highlight the colocalization ofKir2.2 and SAP97 at higher magnification. (g-i) High power image of the ML. Anti-Kir2.2 immunoreactivity is punctate in the GCL (a,d) andalong radial tracks in the ML (a,g). SAP97 staining is present in both the GCL (b,e) and ML (b,h) and is more intense in the ML. Kir2.2 andSAP97 are colocalized with astrocytes and granule cells in the GCL (d-f) and along radial fibers in the ML (g-i, arrowheads). Purkinje cells areindicated with asterisks. Bars, 20 µm (a-c); 10 µm (d-i); 5 µm (inset).

995SAP97 associates with Kir2 channels

Fig. 9. Kir2.2 and SAP97 colocalize in cerebellar glial cells. (A) Sagittal sections of adult rat cerebellar cortex were double-labeled with ananti-GFAP-CY3 antibody (glial cell marker) and either anti-Kir2.2 (a-c) or rabbit anti-SAP97 antibodies (specific for SAP97, and notrecognizing PSD-95) (d-f). Composites of the images on the left are shown on the right (c,f). Punctate, radial tracks of Kir2.2 staining overlapGFAP-positive fiber tracks in the ML (a-c). These fibers are consistent with the radial orientation of Bergmann glia in a sagittal section throughthe ML. SAP97-positive staining colocalizes with GFAP-positive fibers in the ML (arrowheads), demonstrating the expression of both SAP97and Kir2.2 in Bergmann glia radial fibers (d-e). Additionally, SAP97 staining is observed in astrocytes of the GCL (arrows, d-f). Bars, 10 µm.(B) Cells from primary cerebellar cultures were double-labeled as in A. Both Kir2.2 (a) and SAP97 (b) are expressed in GFAP-positive cells(b,e) with astrocyte-like morphologies. Bar, 10 µm.

996

protein SAP97 using GST-fusion protein affinity-pulldownassays. We determined that the association of Kir2.2 andSAP97 requires the presence of the Kir2.2 C-terminal PDZbinding motif (S-E-I) and the second PDZ domain of SAP97.Coimmunoprecipitations from solubilized rat cerebellumand heart membrane protein demonstrate that endogenousKir2.2 is associated with SAP97 in native tissues.Immunocytochemistry identified the cellular and subcellulardistribution of Kir2.2 and SAP97 in these tissues. In the heartthey are colocalized in the T-tubules of cardiac ventricularmyocytes and in the cerebellar cortex in glial astrocytes andgranule cells. Finally, we suggest a role for phosphorylation inthe regulation of the association between Kir2.2 and SAP97 bydemonstrating that protein kinase A phosphorylation of theKir2.2 C terminus inhibits the association with SAP97.

Although SAP97 was colocalized with Kir2.2, it was alsopresent in regions of the same cells in which Kir2.2 expressionwas low or absent. This suggests that SAP97 is associated withKir2.2, but the subcellular localization of Kir2.2 does notappear to be determined by SAP97. Indeed, the punctatedistribution of Kir2.2 in cerebellum suggests that additionalchannel interactions, possibly with PSD-95-like MAGUKs,may be involved in Kir2.2 channel clustering in the brain. Thisis in agreement with the interaction of SAP97 with AMPAreceptors, in which SAP97 mediates signaling (Leonard et al.,1998), while other proteins appear to determine its synapticlocalization.

What is suggested by the subcellular localization of Kir2.2in heart and brain? Cardiac ventricular myocytes containseveral classes of K+ channels, as well as Na+ and Ca2+

channels, and each appears to be restricted to a distinctsubcellular location. Identification of the T-tubules as a site ofKir2.2 channels places them in close proximity to L-type Ca2+

channels, which are also located in T-tubules and are the majorsite of excitation-contraction coupling in ventricular cells (Gaoet al., 1997). This finding is consistent with the proposal thatmuch of the Kir current and ATP-sensitive K+ current islocalized to the T-tubule system, based on electrophysiologicalstudies of cultured cells (Christe, 1999). These results implythat Kir2.2 channels play a role in termination of the actionpotential at the level of the T-tubule. Under physiologicalconditions, this may result in the accumulation of K+ ionsin the T-tubule lumen at high heart rates. Thus, the Kir2.2inward rectifier channel (as well as the Kir2.1 channel; D.Leonoudakis and C. A. Vandenberg, unpublished data) appearsto serve a distinct role in K+ ion fluxes in the T-tubule.

In the cerebellar cortex, localization of Kir2.2 to radialBergmann glia in the molecular layer and astrocytes in thegranule cell layer implies that a major role of this inwardrectifier channel may be in K+ buffering in the cerebellum.Inward rectifier channels have been identified as key players inthe buffering of extracellular K+ by glial cells. In the centralnervous system, neuronal activity results in the release of K+

and its accumulation in extracellular space. Glial cells,particularly astrocytes, are believed to play a major role inbuffering the activity-dependent K+ increases via inwardlyrectifying K+ channels (Newman, 1995). However, the type ofinward rectifier channels responsible for K+ buffering has beenunclear. Previous studies have shown by in situ hybridizationthat the primary site of Kir2.2 expression is in the granule celllayer of cerebellum (Bassand et al., 1999; Isomoto et al., 1997;

Karschin and Karschin, 1997; Muller et al., 1995). Our dataare consistent with mRNA expression in the granule cell layer,and additionally demonstrate that Kir2.2 is expressed in glialastrocytes. It has been proposed that Kir4.1 channels areresponsible for K+ buffering in astrocytes, as they have beenidentified on retinal Müller cells and cerebellar Bergmann glia(Poopalasundaram et al., 2000; Takumi et al., 1995). Otherglial cells also contain inward rectifier channels: Kir2.3 ispresent in reactive astrocytes (Perillan et al., 2000), and Kir2isoforms have been reported in oligodendrocytes and Schwanncells (Mi et al., 1996; Stonehouse et al., 1999). Our datademonstrate that Kir2.2 is a prominent astrocyte channel thatmay play a major role in glial K+ buffering.

How might SAP97 regulate the channel? The ability ofSAP97 to interact with voltage-gated potassium channels (Kimand Sheng, 1996), the AMPA receptor GluR1 (Leonard et al.,1998) and GluR6 (Garcia et al., 1998), supports a role forSAP97 in the subcellular targeting of ion channels and theassembly of ion channel signaling complexes. Experimentswith the SAP97 homologue, Dlg, suggest that it is importantfor proper synaptic organization and localization of synapticproteins in Drosophila (Tejedor et al., 1997; Thomas et al.,1997; Zito et al., 1997). It has been proposed that MAGUKproteins may regulate the number of functional channels thatexist at the plasma membrane (Horio et al., 1997; Jugloff et al.,2000; Nehring et al., 2000; Tiffany et al., 2000). For example,SAP97 coexpressed with Kv1 channels decreased channelsurface expression (Tiffany et al., 2000); in contrast SAP97coexpressed with Kir4.1 increased whole cell currents andcurrent density (Horio et al., 1997). Due to its modular domainstructure, SAP97 may also bring a variety of channels andsignal transduction proteins together with Kir2.2 to form amacromolecular signaling complex. Recently, AMPA receptorregulation by protein kinase A via interaction with SAP97 andAKAP 79/150 has been described (Colledge et al., 2000). Theprotein kinase A regulation of SAP97 interaction with Kir2.2that we have demonstrated (Fig. 6) may contribute a furtherlayer of regulation to SAP97 signaling complexes. The Srcfamily tyrosine kinase, p56lck, also was identified as an SAP97interacting protein in T-lymphocytes (Hanada et al., 1997). Thepossibility of a Src tyrosine kinase-SAP97-Kir2.x channelcomplex is intriguing, since Kir2.1 and Kir2.2 each contain atyrosine kinase consensus phosphorylation site. However, theimportance of the association of SAP97 in regulating Kir2.2physiology remains unclear, since the physiological functionsof SAP97 are not yet well defined.

Although SAP97 and PSD-95 share >70% amino acididentity and associate with many of the same proteins, thereare important differences between them, indicating a differentfunctional role for SAP97. The PSD-95-like scaffoldingproteins (PSD-95, SAP102 and Chapsyn 110) are restricted tobrain, in particularly high concentrations at the post-synapticdensity, and their functions include clustering of proteins andthe formation of signaling complexes. In contrast, SAP97 iswidespread in both brain and peripheral tissues, where itassociates interacting proteins with signaling complexes, butvia interactions that do not form clusters on the cell membrane.These scaffolding proteins also differ in their selection ofbinding partners. Some proteins with a C-terminal PDZbinding motif interact with both SAP97 and PSD-95 (e.g.Kir2.1, Kir2.2, Kir2.3 (Fig. 1) (Cohen et al., 1996; Nehring et

JOURNAL OF CELL SCIENCE 114 (5)

997SAP97 associates with Kir2 channels

al., 2000), Kv1 channels (Kim and Sheng, 1996; Tiffany et al.,2000) and Kir4.1 (Horio et al., 1997), while other proteins suchas AMPA receptor, appear to exclusively select SAP97 and notbind PSD-95-like proteins (Leonard et al., 1998). In contrast,the NMDA receptor selectively bound PSD-95, SAP102 andChapsyn110, but was not immunoprecipitated with SAP97 inbrain (Leonard et al., 1998). Thus, these related MAGUKs mayinteract in functionally distinct ways with ion channels.

How is the association between SAP97 and Kir2.1, Kir2.2and Kir2.3 regulated? One potential regulatory mechanism isPKA phosphorylation of the serine located in the C-terminalPKA consensus site (R-R-E-S-E/A-I), which overlaps the PDZbinding motif (R-R-E-S-E/A-I ) in all three channels. In fact,PKA phosphorylation of this serine residue on Kir2.3 inhibitsits association with PSD-95 (Cohen et al., 1996). In theexperiments presented above, we demonstrate that PKAphosphorylation of Ser425 on the Kir2.2 C-terminal fusionprotein inhibits its association with SAP97 (Fig. 6). In addition,forskolin stimulation of PKA activity led to a decrease in theamount of coimmunoprecipitated SAP97 with Kir2.2. The factthat this decrease occurs in a relatively short period of time (20minutes) suggests that this interaction may be dynamicallyregulated. Phosphorylation of Ser425 on Kir2.2 by PKA mightalso directly modulate channel activity. Indeed, PKAphosphorylation of Kir2.1 on a corresponding site decreaseschannel currents (Wischmeyer and Karschin, 1996).

Association of SAP97 with the strong inward rectifiersKir2.1, Kir2.2 and Kir2.3 supports a role for SAP97 inrecruiting inwardly rectifying potassium channels into amacromolecular complex. Like PSD-95, additionalintracellular proteins interacting with the multiple modulardomains of SAP97 are likely to be important in subcellularlocalization and signal transduction. Further functional studiesof these and other SAP97 interacting proteins may help to shedlight on the role of SAP97 in regulating inwardly rectifyingpotassium channel activity.

We thank Dr K. Foltz for critical reading of the manuscript and DrB. Reese, L. Conti and C. Radeke, for insightful discussions. Thiswork was supported by the National Institutes of Health grantHL41656 (to C.A.V.), Tobacco Related Disease Research Programgrant 9RT-0212 (to D.O.C.), National Institutes of Health PostdoctoralFellowship IF32HL10239-01 (to W.S.M.) and the American HeartAssociation, Western States Affiliate Predoctoral Fellowship9910015Y (to D.L.).

REFERENCES

Bassand, P., Bernard, A., Rafiki, A., Gayet, D. and Khrestchatisky, M.(1999). Differential interaction of the tSXV motifs of the NR1 and NR2ANMDA receptor subunits with PSD-95 and SAP97. Eur. J. Neurosci.11,2031-2043.

Budnik, V., Koh, Y. H., Guan, B., Hartmann, B., Hough, C., Woods, D.and Gorczyca, M.(1996). Regulation of synapse structure and function bythe Drosophilatumor suppressor gene dlg. Neuron17, 627-640.

Cho, K. O., Hunt, C. A. and Kennedy, M. B. (1992). The rat brainpostsynaptic density fraction contains a homolog of the Drosophila discs-large tumor suppressor protein. Neuron9, 929-942.

Christe, G. (1999). Localization of K+ channels in the tubules ofcardiomyocytes as suggested by the parallel decay of membranecapacitance, IK1 and IKATP during culture and by delayed IK1 response tobarium. J. Mol. Cell. Cardiol.31, 2207-13.

Cohen, N. A., Brenman, J. E., Snyder, S. H. and Bredt, D. S.(1996).

Binding of the inward rectifier K+ channel Kir 2.3 to PSD-95 is regulatedby protein kinase A phosphorylation. Neuron17, 759-767.

Colledge, M., Dean, R. A., Scott, G. K., Langeberg, L. K., Huganir, R. L.and Scott, J. D.(2000). Targeting of PKA to glutamate receptors througha MAGUK-AKAP complex. Neuron27, 107-19.

Craven, S. E. and Bredt, D. S.(1998). PDZ proteins organize synapticsignaling pathways. Cell 93, 495-498.

Dong, H. L., Obrien, R. J., Fung, E. T., Lanahan, A. A., Worley, P. F. andHuganir, R. L. (1997). GRIP: A synaptic PDZ domain-containing proteinthat interacts with AMPA receptors. Nature386, 279-284.

Duprat, F., Lesage, F., Fink, M., Reyes, R., Heurteaux, C. and Lazdunski,M. (1997). TASK, a human background K+ channel to sense external pHvariations near physiological pH. EMBO J.16, 5464-5471.

Gao, T., Puri, T. S., Gerhardstein, B. L., Chien, A. J., Green, R. D. andHosey, M. M. (1997). Identification and subcellular localization of thesubunits of L-type calcium channels and adenylyl cyclase in cardiacmyocytes. J. Biol. Chem.272, 19401-7.

Garcia, E. P., Mehta, S., Blair, L. A. C., Wells, D. G., Shang, J., Fukushima,T., Fallon, J. R., Garner, C. C. and Marshall, J.(1998). SAP90 binds andclusters kainate receptors causing incomplete desensitization. Neuron21,727-739.

Hanada, T., Lin, L. H., Chandy, K. G., Oh, S. S. and Chishti, A. H.(1997).Human homologue of the Drosophiladiscs large tumor suppressor binds top56lck tyrosine kinase and Shaker type Kv1.3 potassium channel in Tlymphocytes. J. Biol. Chem.272, 26899-26904.

Horio, Y., Hibino, H., Inanobe, A., Yamada, M., Ishii, M., Tada, Y., Satoh,E., Hata, Y., Takai, Y. and Kurachi, Y. (1997). Clustering and enhancedactivity of an inwardly rectifying potassium channel, Kir4.1, by ananchoring protein, PSD-95/SAP90. J. Biol. Chem.272, 12885-12888.

Inagaki, N., Tsuura, Y., Namba, N., Masuda, K., Gonoi, T., Horie, M.,Seino, Y., Mizuta, M. and Seino, S.(1995). Cloning and functionalcharacterization of a novel ATP-sensitive potassium channel ubiquitouslyexpressed in rat tissues, including pancreatic islets, pituitary, skeletalmuscle, and heart. J. Biol. Chem.270, 5691-5694.

Isomoto, S., Kondo, C. and Kurachi, Y. (1997). Inwardly rectifyingpotassium channels: Their molecular heterogeneity and function. Jpn. J.Physiol.47, 11-39.

Jugloff, D. G. M., Khanna, R., Schlichter, L. C. and Jones, O. T.(2000).Internalization of the Kv1.4 potassium channel is suppressed by clusteringinteractions with PSD-95. J. Biol. Chem.275, 1357-1364.

Karschin, C., Dissmann, E., Stuhmer, W. and Karschin, A.(1996). IRK(1-3) and GIRK(1-4) inwardly rectifying K+ channel mRNAs are differentiallyexpressed in the adult rat brain. J. Neurosci.16, 3559-3570.

Karschin, C. and Karschin, A. (1997). Ontogeny of gene expression of Kirchannel subunits in the rat. Mol. Cell. Neurosci.10, 131-148.

Kim, E., Niethammer, M., Rothschild, A., Jan, Y. N. and Sheng, M.(1995).Clustering of Shaker-type K+ channels by interaction with a family ofmembrane-associated guanylate kinases. Nature378, 85-88.

Kim, E. and Sheng, M.(1996). Differential K+ channel clustering activity ofPSD-95 and SAP97, two related membrane-associated putative guanylatekinases. Neuropharmacology35, 993-1000.

Kistner, U., Wenzel, B. M., Veh, R. W., Caseslanghoff, C., Garner, A. M.,Appeltauer, U., Voss, B., Gundelfinger, E. D. and Garner, C. C.(1993).SAP90, a rat presynaptic protein related to the product of the Drosophilatumor suppressor gene dlg-A. J. Biol. Chem.268, 4580-4583.

Kornau, H. C., Schenker, L. T., Kennedy, M. B. and Seeburg, P. H.(1995).Domain interaction between NMDA receptor subunits and the postsynapticdensity protein PSD-95. Science269, 1737-1740.

Kornau, H. C., Seeburg, P. H. and Kennedy, M. B.(1997). Interaction ofion channels and receptors with PDZ domain proteins. Curr. Opin.Neurobiol.7, 368-373.

Kostin, S., Scholz, D., Shimada, T., Maeno, Y., Mollnau, H., Hein, S. andSchaper, J. (1998). The internal and external protein scaffold of the T-tubular system in cardiomyocytes. Cell Tissue Res.294, 449-60.

Koyama, H., Morishige, K., Takahashi, N., Zanelli, J. S., Fass, D. N. andKurachi, Y. (1994). Molecular cloning, functional expression andlocalization of a novel inward rectifier potassium channel in the rat brain.FEBS Lett.341, 303-307.

Kubo, Y., Baldwin, T. J., Jan, Y. N. and Jan, L. Y.(1993). Primary structureand functional expression of a mouse inward rectifier potassium channel.Nature362, 127-33.

Lahey, T., Gorczyca, M., Jia, X. X. and Budnik, V.(1994). The Drosophilatumor suppressor gene dlg is required for normal synaptic bouton structure.Neuron13, 823-835.

998

Leonard, A. S., Davare, M. A., Horne, M. C., Garner, C. C. and Hell, J.W. (1998). SAP97 is associated with the alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor GluR1 subunit. J. Biol. Chem.273, 19518-19524.

Leonoudakis, D., Gray, A. T., Winegar, B. D., Kindler, C. H., Harada, M.,Taylor, D. N., Chavez, R. A., Forsayeth, J. R. and Yost, C. S.(1998). Anopen rectifier potassium channel with two pore domains in tandem clonedfrom rat cerebellum. J. Neurosci.18, 868-877.

Lue, R. A., Marfatia, S. M., Branton, D. and Chishti, A. H.(1994). Cloningand characterization of hdlg – the human homologue of the Drosophila discslarge tumor suppressor binds to protein 4.1. Proc. Natl. Acad. Sci. USA91,9818-9822.

Matsumine, A., Ogai, A., Senda, T., Okumura, N., Satoh, K., Baeg, G. H.,Kawahara, T., Kobayashi, S., Okada, M., Toyoshima, K. et al. (1996).Binding of APC to the human homolog of the Drosophila discs large tumorsuppressor protein. Science272, 1020-1023.

Mi, H., Deerinck, T. J., Jones, M., Ellisman, M. H. and Schwarz, T. L.(1996). Inwardly rectifying K+ channels that may participate in K+ bufferingare localized in microvilli of Schwann cells. J. Neurosci.16, 2421-9.

Muller, B. M., Kistner, U., Kindler, S., Chung, W. J., Kuhlendahl, S.,Fenster, S. D., Lau, L. F., Veh, R. W., Huganir, R. L., Gundelfinger, E.D. et al. (1996). SAP102, a novel postsynaptic protein that interacts withNMDA receptor complexes in vivo. Neuron17, 255-265.

Muller, B. M., Kistner, U., Veh, R. W., Caseslanghoff, C., Becker, B.,Gundelfinger, E. D. and Garner, C. C.(1995). Molecular characterizationand spatial distribution of SAP97, a novel presynaptic protein homologousto SAP90 and the Drosophila discs-large tumor suppressor protein. J.Neurosci.15, 2354-2366.

Nehring, R. B., Wischmeyer, E., Doring, F., Veh, R. W., Sheng, M. andKarschin, A. (2000). Neuronal inwardly rectifying K+ channelsdifferentially couple to PDZ proteins of the PSD-95/SAP90 family. J.Neurosci.20, 156-162.

Newman, E. A. (1995). Glial cell regulation of extracellular potassium. InNeuroglia(ed. H. Kettenmann and B. R. Ransom), pp. 717-731. New York:Oxford University Press.

Nichols, C. G. and Lopatin, A. N. (1997). Inward rectifier potassiumchannels. Annu. Rev. Physiol.59, 171-91.

Pawson, T. and Scott, J. D.(1997). Signaling through scaffold, anchoring,and adaptor proteins. Science278, 2075-2080.

Périer, F., Radeke, C. M. and Vandenberg, C. A.(1994). Primary structureand characterization of a small-conductance inwardly rectifying potassiumchannel from human hippocampus. Proc. Natl. Acad. Sci. USA91, 6240-6244.

Perillan, P. R., Li, X., Potts, E. A., Chen, M., Bredt, D. S. and Simard, J.M. (2000). Inward rectifier K+ channel Kir2.3 (IRK3) in reactive astrocytesfrom adult rat brain. Glia 31, 181-92.

Poopalasundaram, S., Knott, C., Shamotienko, O. G., Foran, P. G., Dolly,J. O., Ghiani, C. A., Gallo, V. and Wilkin, G. P. (2000). Glialheterogeneity in expression of the inwardly rectifying K+ channel, Kir4.1,in adult rat CNS. Glia 30, 362-72.

Raab-Graham, K. F. and Vandenberg, C. A.(1998). Tetrameric subunitstructure of the native brain inwardly rectifying potassium channel Kir 2.2.J. Biol. Chem.273, 19699-19707.

Rakic, P. (1995). Radial glial cells: scaffolding for brain construction. InNeuroglia(ed. H. Kettenmann and B. R. Ransom), pp. 746-762. New York:Oxford University Press.

Sheng, M. and Wyszynski, M.(1997). Ion channel targeting in neurons.BioEssays19, 847-853.

Stonehouse, A. H., Pringle, J. H., Norman, R. I., Stanfield, P. R., Conley,E. C. and Brammar, W. J. (1999). Characterisation of Kir2.0 proteins inthe rat cerebellum and hippocampus by polyclonal antibodies. Histochem.Cell Biol. 112, 457-65.

Takumi, T., Ishii, T., Horio, Y., Morishige, K., Takahashi, N., Yamada, M.,Yamashita, T., Kiyama, H., Sohmiya, K., Nakanishi, S. et al. (1995). Anovel ATP-dependent inward rectifier potassium channel expressedpredominantly in glial cells. J. Biol. Chem.270, 16339-46.

Tejedor, F. J., Bokhari, A., Rogero, O., Gorczyca, M., Zhang, J. W., Kim,E., Sheng, M. and Budnik, V.(1997). Essential role for dlg in synapticclustering of Shaker K+ channels in vivo. J. Neurosci.17, 152-159.

Thomas, U., Phannavong, B., Muller, B., Garner, C. C. and Gundelfinger,E. D. (1997). Functional expression of rat synapse-associated proteinsSAP97 and SAP102 in Drosophila dlg-1 mutants: Effects on tumorsuppression and synaptic bouton structure. Mech. Dev.62, 161-174.

Tiffany, A. M., Manganas, L. N., Kim, E. J., Hsueh, Y. P., Sheng, M. andTrimmer, J. S. (2000). PSD-95 and SAP97 exhibit distinct mechanisms forregulating K+ channel surface expression and clustering. J. Cell Biol.148,147-157.

Vandenberg, C. A. (1994). Inwardly rectifying potassium channels. In IonChannels in the Cardiovascular System: Function and Dysfunction(ed. A.M. Brown, W. A. Caterall, G. J. Kaczorowski, P. M. Spooner and H. C.Strauss), pp. 145-167. Armonk: Futura Publishing Co.

Willott, E., Balda, M. S., Fanning, A. S., Jameson, B., Vanitallie, C. andAnderson, J. M. (1993). The tight junction protein ZO-1 is homologous tothe Drosophila discs-large tumor suppressor protein of septate junctions.Proc. Natl. Acad. Sci. USA90, 7834-7838.

Wischmeyer, E. and Karschin, A.(1996). Receptor stimulation causes slowinhibition of IRK1 inwardly rectifying K+ channels by direct proteinkinase A-mediated phosphorylation. Proc. Natl. Acad. Sci. USA93, 5819-5823.

Zito, K., Fetter, R. D., Goodman, C. S. and Isacoff, E. Y.(1997). Synapticclustering of Fasciclin II and Shaker: Essential targeting sequences and roleof Dlg. Neuron19, 1007-1016.

JOURNAL OF CELL SCIENCE 114 (5)