Direct interaction between the actin-binding protein filamin-A and the inwardly rectifying potassium channel, Kir2.1.

The role of filamins in actin cross-linking and membrane stabilization is well established, but recently their ability to interact with a variety of transmembrane receptors and signaling proteins has led to speculation of additional roles in scaffolding and signal transduction. Here we report a direct interaction between filamin-A and Kir2.1, an isoform of inwardly rectifying potassium channel expressed in vascular smooth muscle and an important regulator of vascular tone. Yeast two-hybrid screening of a porcine coronary artery cDNA library using the carboxyl terminus of Kir2.1 as bait yielded cDNA encoding a fragment of filamin-A (residues 2481-2647). Interaction between filamin-A and Kir2.1 was confirmed by in vitro overlay assay of membrane-bound Kir2.1 with glutathione S-transferase fusion protein of the isolated filamin clone. Additionally, antibodies directed against Kir2.1 coimmunoprecipitated filamin-A from arterial smooth muscle cell lysates, and immunocytochemical analysis of individual arterial smooth muscle cells showed that Kir2.1 and filamin co-localize in "hotspots" at the cell membrane. Interaction with filamin-A was found to have no effect on Kir2.1 channel behavior but, rather, increased the number of functional channels resident within the membrane. We conclude that filamin-A is potentially an important regulator of Kir2.1 surface expression and location within vascular smooth muscle.

Filamins are large actin-binding phosphoproteins that crosslink actin filaments, stiffening or "gelling" the dispersed microfilament net immediately beneath the cell membrane (1,2). They also establish critical links between the submembrane actin gel and integral membrane proteins, which stabilize the membrane particularly during changes in cell shape associated with cell motility and migration (3,4). Indeed to date, more than 20 different protein partners have been identified for filamin. These include not only transmembrane proteins such as ␤-integrins (5, 6), D2 and D3 dopamine receptors (7), and K v 4.2 potassium channels (8) but also intracellular signaling molecules such as the Rho family of GTPases (9), the stressactivated kinase SEK-1 (also known as MKK-4 (10)), and protein kinase C␣ (11). This ability to aggregate cytoskeletal elements, transmembrane receptors, and cytoplasmic signaling proteins is potentially important not only in the stabilization of receptors at the cell surface but also in cell signal integration (1,2).
The role of filamin within the vasculature is incompletely understood. It almost certainly plays an important role in cellular migration during vascular development since mutations of the filamin-A gene that block filamin expression and cause human periventricular heterotopia, an X-linked disease characterized by recurrent epileptic seizures, also lead to unusually high incidences of congenital vascular abnormalities (12). Unlike the brain, where filamin-A is concentrated during periods of active neuronal migration then down-regulated in the adult, filamin-A persists at high levels in the adult vasculature (13). Within vascular smooth muscle cells, it is known to be involved in bundling actin into stress-fibers within the body of the cell and to associate with the plasma membrane at specialized adhesion sites known as membrane-associated dense plaques (14). These belong to a family of adherens type junctions that include the focal adhesions of cultured cells (15,16) and are the sites at which the cytoskeleton and contractile apparatus of smooth muscle cells anchor to the plasma membrane (17). In common with other focal adhesions, the dense plaque regions act as assemblage sites for a number of signaling molecules and are potentially important loci for the integration of signal transduction pathways controlling processes such as actin-remodeling and contractility (18). Whether filamin plays any role in this signal integration and whether it binds to proteins other than actin within smooth muscle cells has not been fully investigated.
Here we report the results of a yeast two-hybrid screen of a porcine coronary artery cDNA library that identify filamin-A (also known as actin-binding protein 280 and filamin-1) as a binding partner of the inwardly rectifying potassium channel, Kir2.1. Inwardly rectifying potassium (Kir) 1 channels constitute a functionally diverse family of channel proteins that allow K ϩ to move into the cell more easily than out (19,20). This asymmetry in the current-voltage relation results from the channel susceptibility to voltage-dependent block by intracellular polyamines and magnesium ions and allows for modulation of the electrical properties of cells without excessive K ϩ loss. Consequently, Kir channels are involved in setting and maintaining the resting membrane potential, modulation of prolonged action potentials in electrically excitable cells, and potassium homeostasis and secretion (19,20). Kir2.1 is the only isoform of the classical strong (Kir2) inward rectifier subfamily expressed in smooth muscle cells (21) and, crucially, has been shown to be an important regulator of smooth muscle membrane potential and, hence, vascular tone under conditions of metabolic stress (22). * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence should be addressed. We investigate the potential functional consequences of filamin-A binding to Kir2.1 and show through the use of human melanoma cells that lack filamin-A that the interaction does not affect the way the channel behaves but, instead, is important for its functional surface expression. We propose that filamin-A acts as a cytoskeletal anchoring protein for Kir2.1, ensuring its stable surface expression and potentially controlling its location and proximity to other proteins at the cell membrane. In smooth muscle cells this interaction may function to recruit Kir2.1 to signaling complexes within membrane specializations such as the dense plaque region.

EXPERIMENTAL PROCEDURES
Antibodies, Polyacrylamide Gel Electrophoresis, and Immunoblotting-The following primary antibodies were used: mouse monoclonal anti-filamin-A (Santa Cruz Biotechnology, Inc.); rabbit polyclonal anti-GST (Sigma-Aldrich); mouse monoclonal anti-GFP (Abcam Ltd.); rat monoclonal high affinity anti-HA (Roche Applied Science); rabbit polyclonal anti-Kir2.1 (a kind gift from Dr. R. Norman, Dept. of Medicine, University of Leicester (23)). Horseradish peroxidase-, Texas Red-, and fluorescein isothiocyanate (FITC)-conjugated anti-rabbit, anti-mouse, and anti-rat secondary antibodies were from Jackson ImmunoResearch Laboratories, Inc. Protein extracts were resolved by SDS-polyacrylamide gel electrophoresis on 12% polyacrylamide-Tris gels and transferred electrophoretically onto nitrocellulose membranes (Hybond ECL, Amersham Biosciences). Membranes were blocked overnight at 4°C in blocking solution containing 5% (w/v) skim milk powder and 0.1% Tween 20 in Tris-buffered saline. Primary antibodies were diluted in blocking solution containing 1% skim milk powder and 0.1% Tween 20 in Tris-buffered saline and incubated with the membranes for 2-3 h at room temperature. Membranes were washed in Tris-buffered saline then incubated with horseradish peroxidase-conjugated secondary antibodies for a further hour at room temperature. Labeled bands were visualized using enhanced chemiluminescence (ECL; Amersham Biosciences) according to manufacturer's protocol.
Plasmid Constructs-Plasmid construct filamin-A (repeats 19 -24) fused to the NH 2 -terminal HA epitope tag in pcDNA-HANII was a generous gift from Prof. Sonnenburg (The Netherlands Cancer Institute, Amsterdam, The Netherlands). Wild-type and eGFP-Kir2.1 in pcDNA3 were as described previously (24). Dual-tagged HA-and eGFP-Kir2.1 in pcDNA3 was generated by digestion of an internal cassette flanked by BsmB1 and XhoI restriction sites and containing the HA extracellular epitope from HA-Kir2.1 pcDNA3 (25) and subsequent ligation of the cassette into eGFP-Kir2.1 pcDNA3.
Cells and Cell Transfection-Human melanoma cell line, filamin Ϫ (lacking expression of filamin-A) and isogenic cell line, filamin ϩ (stably transfected with full-length filamin-A), were kind gifts from Dr. J. Hartwig (Harvard Medical School, Boston, MA). Cells were grown in ␣-minimal essential medium supplemented with 8% newborn calf serum, 2% fetal calf serum. Filamin ϩ cells were cultured in G418 (0.5 mg/ml) to maintain selection of cells harboring the transfected plasmid. HEK293 were grown in minimal essential medium supplemented with 1% nonessential amino acids and 10% (v/v) fetal bovine serum. All media and reagents were from Invitrogen. Cells were transiently transfected using LipofectAMINE transfection reagent (Invitrogen) according to the manufacturer's protocol. Transfections were performed in 6-well culture plates with cells at 80% confluence. Yeast Two-hybrid Library Screen-Yeast two-hybrid analysis was performed according to the Matchmaker library construction and screening kit (BD Biosciences). A bait construct comprising amino acids 307-428 of mouse Kir2.1 was amplified by PCR (Pfu polymerase) using the wild-type Kir2.1 expression construct in pcDNA3 as template. The PCR product was A-tailed and cloned into the pGEM-T vector (Promega) before subcloning into BamHI and EcoRI sites of the GAL4 binding domain plasmid, pGBKT7. The plasmid was transformed into yeast strain, Y187, harboring the lacZ reporter gene under the control of GAL4-binding sites.
Total RNA was isolated (SV total RNA isolation kit, Promega) from porcine coronary arteries dissected from hearts donated by a local abattoir (Joseph Morris & Sons, Leicestershire, UK) and reversedtranscribed using SMART (switching mechanism at 5Ј end of RNA transcript) technology. cDNA was co-transformed with SmaI-linearized GAL-4 activation domain vector, pGADT7, into yeast strain AH109 harboring reporter genes HIS3 and ADE2 under the control of GAL4binding sites. Library-carrying AH109 were mated with Y187-carrying bait plasmid pGBKT7-2.1C and plated onto synthetic quadruple dropout medium (SD/ϪAde/ϪHis/ϪTrp/ϪLeu). Positive colonies were screened for ␤-galactosidase activity using a filter-lift assay. Activation domain plasmids, pGADT7, were isolated from yeast colonies displaying positive phenotype and transformed into bacteria to obtain plasmids suitable for sequencing reactions. Plasmid inserts were sequenced (Protein and Nucleic Acid Chemistry Laboratory, University of Leicester, Leicester, UK) and compared against the GenBank TM data base by BLAST search analysis (National Institutes of Health).
Generation of Truncation Constructs and Two-hybrid Mapping Studies-Mapping studies of protein-protein interactions between the COOH-terminal filamin clone and Kir2.1 truncation constructs were analyzed by filter-lift and liquid culture ␤-galactosidase assay in yeast strain Y187. Truncation constructs of the Kir2.1 carboxyl terminus (amino acids: 307-366, 367-428, 307-326, 325-346, 347-366) were amplified by PCR (Pfu polymerase) using the yeast two-hybrid vector, 2647 (corresponding to part of repeat 23, the hinge 2 region, and repeat 24) isolated from the yeast two-hybrid screen. Filamins are composed of two identical rod-like subunits that associate head to head via a carboxyl-terminal self-association site to form dimers. The free amino-terminal region of each subunit contains an actin binding motif. The rest of the filamin chain forms a semi-rigid rod composed of 24 repeated sequences interrupted by two short flexible hinge regions (H1 and H2). B, aligned sequences of human filamin-A and cloned porcine protein isolated from yeast two-hybrid screen. Highlighted amino acids correspond to differences between the porcine protein and its human homologue.
Coimmunoprecipitation-Mammalian HEK293 cells were transiently transfected with plasmid cDNA encoding HA-tagged filamin (carboxyl-terminal repeats, 19 -24) and eGFP-tagged Kir2.1. Transfected cells were lysed in lysis buffer (20 mM Tris-HCl, 250 mM NaCl, 3 mM EDTA, 3 mM EGTA, pH 7.6) containing 1% Triton X-100 and protease inhibitors (1:100 dilution, Sigma Protease Inhibitor Mixture containing 4-(2-aminoethyl)benzenesulfonyl fluoride, aprotinin, bestatin, leupeptin, pepstatin A), and insoluble material was cleared by centrifugation. Soluble fractions were incubated with monoclonal anti-GFP antibody (5 g) or mouse non-immune control serum (5 g) overnight at 4°C. Rat arterial smooth muscle lysates were prepared by homogenization of tissue in ice-cold lysis buffer in a hand-held homogenizer. Insoluble material was cleared by centrifugation, and soluble Filamin Binds to Inward Rectifier K ϩ Channels fractions were incubated overnight with rabbit polyclonal anti-Kir2.1 or rabbit non-immune control serum as above. Antigen-antibody complexes were captured with protein-A-Sepharose (Amersham Biosciences; 4°C, 2 h). Beads were washed extensively before removal of bound proteins by boiling in SDS-sample buffer. Samples were resolved by SDS-PAGE, transferred onto nitrocellulose membrane, and analyzed by immunoblotting.
Cell Imaging-Cells were grown to 80% confluence on poly-L-lysinecoated glass coverslips before transfection with LipofectAMINE (Invitrogen). Before imaging of live cells, coverslips were sealed to the base of an imaging chamber with vacuum grease and maintained in PBS at room temperature. Alternatively, filamin ϩ and filamin Ϫ cells were fixed and permeabilized in ice-cold methanol. Fixed cells were incubated with anti-filamin (1:200) primary antibody diluted in PBS containing 10% (v/v) goat serum overnight at 4°C. Excess antibody was removed by extensive washing in PBS before incubation with FITC-conjugated antirabbit secondary antibody (1:200 in PBS containing 10% v/v goat serum) for 2 h at room temperature under dark conditions. After repeated washing (PBS), coverslips were mounted onto microscope slides using fluorescent mounting medium (Dako Ltd). Live and fixed cells were viewed using an inverted PerkinElmer Life Sciences Ultraview confocal laser scanning microscope equipped with a krypton/argon laser and ϫ60 oil immersion lens with numerical aperture of 1.0. Images of cells expressing the eGFP-HA-Kir2.1 construct were analyzed using a method similar to that described previously (26). Tagged image file format (TIFF) images were imported into NIH ImageJ, the cells were outlined, and the mean pixel values were obtained. Pixel values were on an 8-bit scale (2 8 ϭ 256; 0 -255). Laser levels were kept constant between 488-and 568-nm channels.
Smooth Muscle Immunocytochemistry-Smooth muscle cells were isolated enzymatically using a modification of methodology described previously (27), whereby secondary enzymatic digestion of arterial segments with collagenase and hyaluronidase was omitted. Cells were plated onto poly-L-lysine-coated coverslips before fixation and permeabilization in paraformaldehyde and Triton X-100 (0.1%), respectively. Antibody staining and cell imaging were as described for heterologous cells. Cells were incubated overnight at 4°C with mouse monoclonal anti-filamin (diluted 1:1000 in PBS with 10% v/v goat serum) and rabbit polyclonal anti-Kir2.1 (1:500) and visualized with anti-mouse Texas Red-and anti-rabbit FITC-conjugated secondary antibodies, respectively.
Electophysiology-Whole-cell currents were recorded from either filamin ϩ or filamin Ϫ cells typically 24 -48 h post-transfection using an Axopatch 200B amplifier (Axon Instruments). Currents recorded in response to voltage steps were filtered at 5kHz (Ϫ3 dB, 8-pole Bessel), digitized at 10 kHz using a DigiData 1320A interface (Axon Instruments), and analyzed using pCLAMP software. Electrodes were pulled from borosilicate glass (outer diameter 1.5 mm, inner diameter 1.17 mm; Clarke Electromedical, Pangbourne, UK) and fire-polished to give a final resistance of 5 megaohms when filled. The pipette-filling solution contained 140 mM KCl, 1 mM MgCl 2 , 10 mM EGTA, 10 mM HEPES, pH 7.2. The external solution contained 70 mM KCl, 70 mM NaCl, 2 mM MgCl 2 , 2 mM, CaCl 2 , 10 mM HEPES, pH 7.25. The junction potential between pipette and external solutions was sufficiently small (Ͻ1.5 mV) to be neglected. As far as possible analogue means were used to correct capacity transients. Up to 90% compensation was routinely used to correct for series resistance. Single channel currents recorded from either filamin ϩ or filamin Ϫ cells were filtered at 2 kHz and digitized at

Identification of Filamin-A as a Binding Partner for Kir2.1 by Yeast Two-hybrid Screening-To isolate binding partners of
Kir2.1, a carboxyl-terminal segment of the channel (aa 307-428) was subcloned into the GAL4 binding domain bait vector, pGBKT7. This was used as bait to screen a porcine coronary artery cDNA library incorporated into the GAL4 activation domain vector, pGADT7, by yeast two-hybrid analysis. A screen of 1 ϫ 10 6 yeast colonies resulted in a large number (1738) exhibiting positive phenotype (growth on synthetic quadruple drop-out medium SD/ϪAde/ϪHis/ϪTrp/ϪLeu). These colonies were screened by ␤-galactosidase filter assay, but only 10 demonstrated a positive reaction (development of blue color at ϳ40 min). One of the positive clones encoded a partial cDNA representing part of repeat 23, the hinge 2 region, and repeat 24 of filamin-A (actin-binding protein 280, aa 2481-2647; Fig. 1A). Translation of the sequence revealed that the isolated porcine clone exhibited a 94% homology with its human homologue (156 of 166 amino acids; Fig. 1B).
Mapping of the Region on Kir2.1 Responsible for Interacting with Filamin-To further define the carboxyl-terminal Kir2.1 amino acid motif/motives responsible for interaction with filamin, successive deletion constructs of Kir2.1 (aa 307-428) were tested by yeast two-hybrid analysis for interaction with the isolated filamin clone. Fig. 2Ai shows the truncation constructs of Kir2.1 that were cloned into the bait vector pGBKT7. These were cotransformed into yeast strain Y187 along with the filamin clone in prey vector pGADT7 for quantitative ␤-galactosidase assays. Activation of the reporter gene occurred with amino acids 307-367 of Kir2.1 as bait but not amino acids 367-428. The strength of interaction increased with successive deletions of the 307-367 amino acid channel segment such that amino acids 307-326 demonstrated the strongest interaction tested (Fig. 2Bi). A similar observation has been reported for the interaction of glycoprotein GPIb␣ with filamin-A, where stronger yeast two-hybrid signals were obtained with smaller protein fragments (28). Transfection of the Kir2.1 truncation constructs with empty prey plasmid, pGADT7, prevented the Filamin Binds to Inward Rectifier K ϩ Channels positive interactions observed in the ␤-galactosidase filter-lift and liquid assays.
Mutations in Kir2.1 can give rise to the development of Andersen's syndrome, a rare condition characterized by dysmorphic features, periodic paralysis, and cardiac arrhythmias (29). One genetic variant of the disease has been documented as an in-frame deletion of amino acids 314 -315. Because our results suggest that amino acids in the region 307-326 are involved in the interaction of Kir2.1 with filamin, we investigated the importance of amino acids 314 -315 to filamin binding. We constructed 314 -315 deletion mutants within the yeast two-hybrid bait vectors, pGBKT7-Kir2.1 aa 307-428 (⌬ 314 -315) and pGBKT7-Kir2.1 aa 307-326 (⌬ 314 -315) (Fig. 2Aii). Fig. 2, Aii and Bii, shows that deletion of the Andersen's syndrome amino acids 314 -315 had no effect on the ability of filamin to interact with Kir2.1 amino acids 307-326, as assayed by filter-lift and liquid ␤-galactosidase experiments. Transformation of the latter Kir2.1 aa 307-326 (⌬ 314 -315) deletion constructs together with an empty prey domain vector prevented the positive result observed for the ␤-galactosidase assays, demonstrating the specificity of the interactions.
Filamin Interacts in Vitro with Kir2.1-The interaction between filamin and the carboxyl terminus of Kir2.1 (Kir2.1C) was confirmed using an in vitro GST overlay assay. The filamin cDNA clone (aa 2482-2647) was excised from pGADT7 and subcloned into pGEX-6P-1 for expression as a GST fusion protein. Using a Far Western overlay assay, GST-filamin was shown to interact with membrane-bound His-tagged Kir2.1C, whereas GST alone failed to show any interaction (Fig. 3A).
Filamin-A Coimmunoprecipitates with Kir2.1-To establish an interaction between Kir2.1 and filamin within intact mammalian cells, co-immunoprecipitation experiments were performed. HEK293 cells were co-transfected with cDNA con-structs corresponding to eGFP-tagged Kir2.1 (entire protein) and HA-tagged filamin-A (COOH-terminal truncation of repeats 19 -24). Antibodies directed against the eGFP epitope on Kir2.1 were able to coimmunoprecipitate the HA-filamin COOH-terminal construct (Fig. 3B), suggesting that these two proteins form complexes within HEK293 cells. Both filamin and Kir2.1 are expressed within arterial smooth muscle (21,22,30), and we were interested in establishing whether such an interaction was likely to occur within native tissue. Fig. 4A shows that antibodies directed against Kir2.1 coimmunoprecipitate full-length (280 kDa) filamin-A from arterial smooth muscle lysates, consistent with a physical interaction between these two proteins within smooth muscle cells.

Kir2.1 and Filamin Colocalize in Arterial Smooth Muscle Cells-Because
Kir2.1 is an integral membrane protein and filamin organizes the actin cytoskeleton immediately beneath the membrane, these proteins would be expected to be in close proximity within cells. Fig. 4, B and C, shows confocal images of an isolated arterial smooth muscle stained with antibodies directed against filamin-A and Kir2.1 and visualized with secondary antibodies conjugated with the non-overlapping fluoroprobes Texas Red (Fig. 4B) and FITC (Fig. 4C), respectively. The images show both proteins localized predominantly to the plasma membrane, and more specifically, a slightly punctate distribution was observed for both filamin and Kir2.1, suggesting that both proteins are concentrated in certain areas on the membrane. Kir2.1 and filamin colocalize to some, but not all puncta (Fig. 4, B and C, inset), indicating the potential existence of membrane complexes made up of Kir2.1 and filamin within smooth muscle cells.
Kir2.1 Is Targeted to the Plasma Membrane in the Absence of Filamin-A-Interaction between Kir2.1 and filamin within cells could fulfil a number of different roles, including regula- FIG. 4. Kir2.1

and filamin-A coimmunoprecipitate from arterial smooth muscle lysates and colocalize in regions on the plasma membrane of arterial smooth muscle cells.
A, immunoprecipitation (IP) of proteins from rat arterial smooth muscle lysates was performed using rabbit polyclonal antibodies directed against Kir2.1 or rabbit non-immune control serum as described under "Experimental Procedures." Precipitated proteins were separated by SDSpolyacrylamide gel electrophoresis, transferred onto nitrocellulose membrane, and immunoblotted (IB) with anti-filamin-A. 25% of smooth muscle lysate was run in the extract lane. Full-length filamin-A runs as a single band of molecular mass 280 kDa. B and C, confocal images of rat arterial smooth muscle cell stained with mouse anti-filamin-A and rabbit anti-Kir2.1. The top end of the cell is out of the focal plane. The subcellular distribution of filamin-A (B) was visualized by the addition of a Texas Red-conjugated antimouse secondary antibody, and Kir2.1 was visualized by the addition of a FITCconjugated anti-rabbit secondary (C). In control experiments no cross-reactivity was observed between the secondary antibodies, and no staining was observed if the secondaries were applied alone. Both proteins were found predominantly in the plasma membrane and colocalized in discrete punctate regions (indicated inset and with arrows). Scale bar indicates 50 m.
Filamin Binds to Inward Rectifier K ϩ Channels tion of the surface expression and location of Kir2.1. Studies into the physiological roles of filamin have been aided greatly by the development of tumor cell lines from human malignant melanomas that lack filamin-A (filamin Ϫ cells; also called M2 cells (4)). These cells show extensive and continuous blebbing of the plasma membrane and, consistent with the role of filamin in membrane stabilization and cell motility, remain rounded and unable to extend membrane projections and spread normally (Fig. 5A). Normal cell morphology and motility can be restored by stable transfection with cDNA encoding filamin (filamin ϩ cells; also called M2A7 cells, Fig. 5B). Filamin can be seen in the rescued filamin ϩ cells concentrated at the cell periphery in close apposition to the membrane (Fig. 5D); no filamin-A immunoreactivity is observed in the filamin-deficient cells (Fig. 5C). Levels of actin and other actin-associated proteins such as gelsolin, profilin, and ␣-actinin are comparable between filamin ϩ and filamin Ϫ cells (4).
Filamin-deficient cells have been reported to have a reduced surface expression of many receptors and ion channels, suggesting that organization of the submembranous cytoskeleton may be important for the appropriate insertion or removal of proteins from the membrane. As an extreme example of this, expression of the D2 dopamine receptor in filamin-deficient cells results in an almost total loss of receptor expression in the membrane (7). To test if filamin had a similar influence on the trafficking of Kir2.1 we expressed eGFP-tagged Kir2.1 (24) in filamin-deficient and filamin-replenished cell lines, filamin ϩ and filamin Ϫ , respectively. Figs. 5, E and F, show confocal images of filamin Ϫ and filamin ϩ transiently transfected with eGFP-tagged Kir2.1. The expressed eGFP-Kir2.1 protein is clearly visible at or near the plasma membrane of both cell types. eGFP-Kir2.1 is also observed throughout the cell body of both cell types, harbored within intracellular compartments.
Effect of Kir2.1-Filamin Association on Whole-cell Kir2.1 Currents-To establish whether the Kir2.1 channels observed at the cell periphery in Figs. 5, E and F, are inserted into the membrane we used the conventional whole-cell clamp technique to record membrane currents from single filamin ϩ (Fig.  6A) and filamin Ϫ cells (Fig. 6B) that had been transiently transfected with cDNA encoding eGFP-Kir2.1. Currents were recorded in response to voltage steps from a holding potential of Ϫ17mV (the equilibrium potential for K ϩ (E K ) under these recording conditions) to test potentials ranging from ϩ60mV to Ϫ100mV in 10-mV increments. Voltage steps positive to E K elicited only small outward currents, whereas steps negative to E K produced substantial inward currents consistent with the membrane expression of Kir2.1. No significant whole-cell currents were recorded from non-transfected filamin ϩ or filamin Ϫ cells under these conditions. In the absence of filamin, wholecell Kir2.1 currents were significantly reduced (Fig. 6, C and  D). At Ϫ100 mV inward current in filamin ϩ cells was 1015 Ϯ 145 pA/pF (n ϭ 8), whereas it was only 598 Ϯ 122 pA/pF in filamin-deficient filamin Ϫ cells (n ϭ 8; p Ͻ 0.05).
Kir2.1 Single Channel Recordings from Filamin ϩ and filamin Ϫ Cells-The reduction in whole-cell Kir2.1 current in filamin-deficient cells may attributed to several factors: 1) a reduction in the number of functional channels at the cell surface; 2) a reduction in Kir2.1 single channel conductance; 3) a change in the probability of single channel opening; 4) reduced Kir2.1 expression levels in filamin Ϫ cells. To help distinguish between these possibilities we recorded single channel currents from filamin ϩ and filamin Ϫ cells transiently transfected with eGFP-Kir2.1. To ensure that we did not disrupt Kir2.1-filamin interactions by pulling excised patches, we used the cell-attached configuration of the patch-clamp technique. Typical Kir2.1 single channel recordings from filamin-deficient cells are shown in Fig. 7A; consistent with the expression of Kir2.1 we were only able to record significant single channel currents at potentials negative to the equilibrium potential for K ϩ (ϩ10 mV, under these recording conditions). The single channel conductance of Kir2.1 was identical in filamin ϩ and filamin Ϫ cells (Fig. 7B). Unitary conductance at Ϫ100 mV was 25.5 Ϯ 0.4 pS in filamin ϩ cells and 23.9 Ϯ 1.1 pS in filamin Ϫ cells (n ϭ 3 each). Open probabilities for Kir2.1 were also unaffected by the presence of filamin. At Ϫ100 mV p open was 0.65 Ϯ 0.07 for filamin ϩ cells and 0.68 Ϯ 0.05 for filamin Ϫ cells (n ϭ 3 each; Fig. 7C). These data suggest that the reduction in whole-cell Kir2.1 currents in filamin-deficient cells stems from either a reduction in the number of functional channels inserted into the cell membrane or a difference in the overall expression levels of Kir2.1 between filamin ϩ and filamin Ϫ cells.
Total Expression Versus Surface Expression-Only a percentage of the total number of channels expressed will be inserted in the membrane at any one time. To determine how many channels are in the membrane for a given level of total expression, we constructed a Kir2.1 channel with intracellular and extracellular epitope tags. This construct (eGFP-HA-Kir2.1) produces a Kir2.1 channel that has eGFP fused to the intracellular amino terminus and an HA epitope inserted into the extracellular loop between the first transmembrane segment (M1) and the pore-forming H5 region. Fig. 8, A-D, shows confocal images of non-permeabilized filamin ϩ and filamin Ϫ cells that have been transiently transfected with the eGFP-HA-Kir2.1. The cells have been stained with antibodies against the extracellular HA epitope, and distribution of the HA epitope has been visualized with secondary antibodies conjugated with the fluorophore Texas Red. No antibodies are required to visualize the distribution of the eGFP tag. When excited at 488 nm, the intensity of intracellular eGFP fluorescence gives an indication of the total level of eGFP-HA-Kir2.1 expression within filamin ϩ and filamin Ϫ cells (Fig. 8, A and C), whereas the intensity of the extracellular Texas Red signal at 568 nm gives an indication of the number of channels inserted into the membrane (Fig. 8, B and D). We found that the fraction of expressed eGFP-HA-Kir2.1 channels that were inserted into the membrane was significantly less in filamin-deficient cells as compared with filamin-containing cells (Fig. 8E), highlighting a potential role for filamin in regulating Kir2.1 surface expression. DISCUSSION Our findings suggest that the actin-binding protein filamin-A acts as a cytoskeletal anchoring protein for Kir2.1 within cells, stabilizing its surface expression and potentially recruiting the ion channel to signaling complexes within membrane specializations in arterial smooth muscle cells.
Kir2.1 is the predominant isoform of the classical strong inward rectifier to be expressed in smooth muscle (21). Targeted disruption of the Kir2.1 gene produces arteries that fail to dilate in response to the modest elevations in extracellular K ϩ that are typically associated with periods of hypoxia and ischemia, indicating the essential role for Kir2.1 in the regulation of vascular tone under conditions of metabolic stress (22). Little is known, however, about the mechanisms involved in regulating the surface expression or location of Kir2.1 in smooth muscle cells or indeed whether the channel complexes with other proteins that modulate its activity. To identify potential proteins that interact with Kir2.1 within the vasculature we screened a porcine coronary artery cDNA library using the yeast two-hybrid system and a carboxyl-terminal construct of Kir2.1 as bait. The screen yielded cDNA encoding a carboxylterminal fragment (residues 2481-2647 corresponding to part of repeat 23, hinge region 2, and repeat 24) of filamin-A, a cytoskeletal protein known to be involved in actin cross-linking and tethering of cell surface receptors to the actin cytoskeleton (1,2). The interaction between filamin-A and Kir2.1 was confirmed by overlay of His-tagged Kir2.1 with a GST fusion protein of the isolated filamin clone. The ability for Kir2.1 and Filamin Binds to Inward Rectifier K ϩ Channels filamin-A to form a complex within intact cells was verified by using antibodies directed against an epitope tag on Kir2.1 to coimmunoprecipitate filamin-A (COOH-terminal repeats 19 -24) from heterologous cells transfected with cDNAs encoding both proteins. Furthermore, antibodies directed against Kir2.1 were able to isolate full-length filamin-A from lysates of arterial smooth muscle, indicating a physical interaction between these proteins in native cells. Indeed, immunocytochemical analysis of individual vascular smooth muscle cells showed that Kir2.1 and filamin are co-localized in "hotspots" on the membrane, consistent with them existing together at membrane specializations. The nature of these specialized sites is speculative, but they may represent dense plaque regions on the smooth muscle. These are points of attachment between the cytoskeleton, the contractile apparatus, and the plasma membrane and are structurally similar to the focal adhesion sites of cultured cells (14 -16). In common with other focal adhesions, dense plaque regions act as assembly sites for a number of different signaling proteins and cytoskeletal elements, including filamin, integrins, focal adhesion kinase, (FAK), paxillin, and non-receptor tyrosine kinases (18). Any potential functional significance of recruiting Kir2.1 to these regions remains to be determined.
Members of the Kir2 subfamily of inward rectifying potassium channels have previously been shown to interact with the PDZ domains of selected members of the membrane-associated guanylate kinase protein family (31)(32)(33)(34). Interaction with membrane-associated guanylate kinases is believed to facilitate the subcellular targeting of ion channels and the formation of functional ion channel signaling complexes. Kir2.1 and Kir2.3 both interact with PSD-95/SAP90, a cytoskeletal component of postsynaptic densities, via a type I PDZ binding motif ((T/S)X(V/I)) located at the extreme carboxyl terminus of the channel (31). Kir2.3 colocalizes with PSD-95 at post-synaptic membranes in the rat forebrain (31,35). A likely binding partner for the strong inward rectifiers in non-neuronal tissues has been shown to be the ubiquitously expressed membrane-associated guanylate kinase, SAP-97/hDlg, which again interacts with Kir2.1, 2.2, and 2.3 via the carboxyl-terminal PDZ binding motif and colocalizes with Kir2.2 in T-tubules of cardiac ventricular myocytes (34). Recent evidence also suggests that SAP97 may mediate formation of large multi-protein complexes centered around Kir2 isoforms in both the heart and the brain (36).
The region of the carboxyl terminus of filamin that interacts with Kir2.1 is particularly rich in binding sites for transmembrane and signaling proteins and may also function as a protein scaffold. Filamins are composed of two identical rod-like subunits (each of 240 -280 kDa) that associate head to head via a carboxyl-terminal self-association site to form dimers that appear by electron microscopy as elongated V-shaped strands (37)(38)(39). The free amino-terminal region of each subunit con- tains an actin binding motif common to other actin-interacting proteins such as ␤-spectrin, dystrophin, and ␣-actinin. The rest of the filamin chain forms a semi-rigid rod composed of 24 repeated sequences interrupted by two short flexible hinge regions. The majority of filamin-interacting proteins bind to a region just upstream of the dimerization site. The calcitonin receptor (40), the voltage-gated K ϩ channel K v 4.2 (8), SEK-1 (also known as MKK-4, a member of the stress-activated Jun kinase cascade (10)), TRAF-2 (tumor necrosis factor receptorassociated factor-2 (41)), the small GTPases Rac, Rho, CDC42, and RalA (9), protein kinase C␣ (11), and now Kir2.1 all bind in the region between repeats 20 and 24 that includes the second hinge region.
We identified the reciprocal binding site of filamin on Kir2.1 as being primarily a stretch of 19 amino acids (307-326) on the carboxyl terminus of the channel. The much weaker interaction observed with filamin and amino acids 325-346 of Kir2.1 suggests that other regions of Kir2.1 may also contribute to binding in the final association complex. For the potassium channel K v 4.2, the binding site for filamin was mapped to a proline-rich PTPP motif (8). This motif does not appear in the filamin binding regions on Kir2.1 and suggests that different mechanisms might be employed for scaffolding different ion channels. Indeed, no obvious sequence homology exists between any of the protein domains that associate with the filamin-A repeats, indicating that secondary or tertiary structures may be important in defining binding specificity.
Interactions with the actin cytoskeleton through filamin-A may regulate not only the location of Kir2.1 on cell membranes and its vicinity to other proteins but also the stability of the channel at the cell surface. We found whole-cell Kir2.1 currents recorded from human melanoma cells that lack filamin-A to be significantly smaller than Kir2.1 currents recorded from stable transformants of these cells expressing filamin-A. Kinetic analysis of single Kir2.1 channel currents showed that the absence of filamin has no significant effect upon single channel amplitude or open probability but most likely diminishes the number of functional channels in the cell membrane. We confirmed this finding by use of a double-tagged Kir2.1 construct that allows us to monitor surface expression as a fraction of total expres-sion. A number of other filamin-interacting proteins including ␤-integrins (28), K v 4.2 voltage-gated potassium channels (8), D2 dopamine receptors (7), and calcitonin receptors (40) show markedly decreased levels of functional cell surface expression in the absence of filamin. The reason for this is unclear, although the presence of a correctly organized submembranous actin cytoskeleton may be essential for the insertion or removal of proteins into or out of the plasma membrane. In support of this, binding of filamin to the carboxyl tail of the calcitonin receptor has been shown to be important in the recycling of the receptors back to the membrane after internalization (40).
To conclude, the role of filamin within the vasculature is incompletely understood, but elsewhere its ability to create peripheral actin gel networks by orthogonal cross-linking of actin filaments is essential for membrane elasticity and cell motility, particularly during development (4,12). Additionally, its ability to link integral membrane proteins to the cytoskeleton and intracellular signaling molecules point to potential roles in scaffolding and the stabilization of receptor/ion channel surface expression. We identify Kir2.1 as belonging to an everincreasing number of transmembrane proteins interacting with cytoskeletal protein filamin-A and suggest that this interaction may be a method of targeting and regulating surface expression of Kir2.1 in vascular smooth muscle.