A Multiprotein Trafficking Complex Composed of SAP97, CASK, Veli, and Mint1 Is Associated with Inward Rectifier Kir2 Potassium Channels*

Strong inward rectifier potassium (Kir2) channels are important in the control of cell excitability, and their functions are modulated by interactions with intracellular proteins. Here we identified a complex of scaffolding/ trafficking proteins in brain that associate with Kir2.1, Kir2.2, and Kir2.3 channels. By using a combination of affinity interaction pulldown assays and co-immunopre-cipitations from brain and transfected cells, we demonstrated that a complex composed of SAP97, CASK, Veli, and Mint1 associates with Kir2 channels via the C-termi-nal PDZ-binding motif. We further demonstrated by using in vitro protein interaction assays that SAP97, Veli-1, or Veli-3 binds directly to the Kir2.2 C terminus and recruits CASK. Co-immunoprecipitations indicated that specific Veli isoforms participate in forming distinct protein complexes in brain, where Veli-1 stably associates with CASK and SAP97, Veli-2 associates with CASK and Mint1, and Veli-3 associates with CASK, SAP97, and Mint1. Addition-ally, immunocytochemistry of rat cerebellum revealed overlapping expression of Kir2.2, SAP97, CASK, Mint1, with Veli-1 in the granule cell layer and Veli-3 in the molecular layer. We propose a model whereby Kir2.2 associates

Strong inward rectifier potassium (Kir2) 1 channels are a widely expressed family of ion channel proteins distinguished by their ability to pass K ϩ current in the inward direction more readily than outward. Kir2 channels are key components in control of neuronal excitability in brain, electrical activity in heart, vascular tone, and glial buffering of potassium (1,2). Kir2 channels are important in the modulation of cell excitability, repolarization of the action potential, and determination of the cellular resting potential. Furthermore, Kir2 mutations are implicated in at least one genetic disease causing periodic paralysis and heart arrhythmias (Andersen's syndrome (3)).
The function, localization, and trafficking behavior of many ion channels are regulated by interactions of their intracellular domains with members of the MAGUK protein family. MAGUK proteins have a characteristic domain structure containing up to three PDZ domains, a Src homology 3 (SH3) domain, and a catalytically inactive guanylate kinase-like domain (GK). These domains interact with numerous other proteins allowing MAGUK proteins to participate in the assembly of macromolecular signaling or transport complexes (4,5). Kir2.1, Kir2.2, and Kir2.3 bind to the PDZ domains of selected members of the MAGUK family via their PDZ-binding motif ((T/S)X(V/I)) found at the extreme C terminus (6,7). Kir2 channels have been demonstrated to interact with the MAGUK proteins PSD-95/SAP90 (6,8) and SAP97 in a PDZ-dependent manner (7). Whereas PSD-95 is known to cause Kir2 channel clustering in the plasma membrane (8), the functional consequences of the interaction of Kir2 channels with SAP97 in native tissues are still unclear.
Because SAP97 is involved in the formation of macromolecular signaling complexes (9), we hypothesized that one function of SAP97 might be to assemble larger protein complexes to recruit effector molecules to Kir2 channels. SAP97 has been shown to interact with a number of proteins including another MAGUK molecule known as CASK/Lin-2 (10,11). CASK contains a unique N terminus that interacts with Veli (also known as MALS and Lin-7) and Mint1 (also known as X11␣ and Lin-10), forming a tightly bound, evolutionarily conserved tripartite complex implicated in basolateral receptor localization and dendritic vesicle/receptor trafficking (12)(13)(14)(15). Recently, Olsen et al. (16) demonstrated that the Veli-2 (hLin7-b) isoform associates with Kir2.3 in epithelial cells, suggesting that Veli-2 directly interacts with Kir2 channels. Additionally, by using immunocytochemical methods they showed that Kir2.3, Veli-2, and CASK proteins are all expressed at the basolateral membrane of renal cortical collecting duct epithelial cells (16) appropriate for complex formation. These data suggest that Kir2 channels may form a multiprotein complex with SAP97, CASK, Veli, and Mint1.
The strong association of CASK with SAP97 led us to hypothesize that the CASK-Veli-Mint1 complex is recruited to Kir2 channels by SAP97 and possibly participates in channel trafficking or assembly of a signaling complex. Additionally, we identified SAP97, CASK, Veli, and Mint1 in association with Kir2 channels in a parallel proteomic study (17). Here we address this hypothesis by demonstrating that a complex composed of CASK, Veli, Mint1, and SAP97 indeed associates with Kir2 channels in brain by a combination of co-immunoprecipitations, affinity interaction assays, and co-localization studies. When expressed in polarized epithelial cells, Kir2.2 channels are targeted to the basolateral membrane. As a test of the functional role of this complex, we expressed a dominant-interfering form of CASK. The resulting mislocalization of Kir2.2 suggests that a functional role of the CASK-containing complex is in Kir2.2 trafficking and localization in the plasma membrane.
Cerebellum, Whole Brain, and Heart Extract Preparation-Tissues were harvested from adult rats and immediately frozen in liquid nitrogen and stored at Ϫ80°C. Tissues were homogenized in 10 volumes of Buffer D (20 mM Hepes, pH 7.6, 125 mM NaCl, 10% glycerol, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride) containing 1ϫ protease inhibitor mixture (Roche Applied Science). Triton X-100 was added to a final concentration of 1% and incubated at 4°C for 1 h. The homogenate was centrifuged (16,000 ϫ g, 10 min at 4°C), and the supernatant was centrifuged once more (100,000 ϫ g, 30 min at 4°C) to remove insoluble material and was used for affinity chromatography and immunoprecipitations.
GST Affinity Pulldown Assay-Fusion proteins were expressed in Escherichia coli and purified on glutathione-agarose (Sigma) as described previously (7). Cerebellum, whole brain, or COS-1 extracts were precleared with GST (20 g/pulldown) bound to glutathione-agarose. GST fusion protein bound to glutathione-agarose (5 g/tissue extract pulldown) was added to the precleared solubilized protein and incubated overnight with rotation. Beads were collected by centrifugation and washed four times with 25 mM Tris, pH 7.4, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100. Beads were resuspended in SDS-PAGE loading buffer and the eluted proteins analyzed by immunoblotting.
Co-immunoprecipitation-Solubilized protein was precleared by the addition of protein A-agarose (Pierce) (1 h). 5 g of affinity-purified anti-Kir2.1/2.2 antibodies or rabbit IgG (control) was added to the precleared solubilized protein and incubated for 1 h. To collect immune complexes, 25 l of protein A-agarose was added and incubated overnight with rotation. Precipitated proteins were washed four times with 25 mM Tris, pH 7.4, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, and co-immunoprecipitated proteins were analyzed by immunoblotting.
In Vitro Protein Interaction Assays-To determine whether CASK binds directly or indirectly with Kir2 channels, in vitro binding assays were performed. GST-SAP97 was purified as described above, and purified, full-length SAP97 was released from GST by cleavage with biotinylated thrombin followed by thrombin removal with streptavidinconjugated agarose (Novagen). Full-length CASK was subcloned into the bacterial expression vector pET23a and expressed in BL21 E. coli. Bacteria were sonicated on ice in HEN20 IP buffer and centrifuged at 10,000 ϫ g for 10 min at 4°C, and the supernatant containing CASK was used for the interaction assay. 2 g of GST, GST-Kir2.2, or GST-Kir2.2⌬3 fusion protein attached to glutathione-agarose was incubated with or without 2 g of purified SAP97 at room temperature for 10 min followed by addition of CASK extract and incubation overnight at 4°C. Beads were washed twice with HEN20 IP buffer with 500 mM NaCl and twice with HEN20 IP buffer with 150 mM NaCl. Bound proteins were eluted in SDS sample buffer and processed for immunoblotting.
Immunofluorescence Microscopy-Rats were euthanized with CO 2 . Cerebellums were removed and immediately fixed in 60% methanol, 30% chloroform, 10% acetic acid at 4°C for 24 h and then transferred to 100% methanol for 24 h, followed by PBS for 24 h, then 20% sucrose in 100 mM phosphate buffer, pH 7.4, for 24 h. Cerebellums were embedded in 300 bloom gelatin and frozen at Ϫ70°C. 16-m frozen sections were obtained with a cryostat (Leica CM 1850) and transferred to charged microscope slides (Fisher). Sections were incubated in blocking buffer (3% bovine serum albumin, 1% donkey serum, 0.1% Triton X-100 in 1ϫ PBS) overnight at 4°C. Sections then were incubated with primary antibodies diluted in blocking buffer overnight at 4°C. Sections were washed 5 times for 30 min with PBS. Sections were incubated with secondary antibodies diluted in blocking buffer (1:200 donkey antimouse CY3 (Jackson ImmunoResearch); 1:50 donkey anti-rabbit FITC (Molecular Probes) for 1 h at 4°C, followed by washing 5 times for 30 min with PBS. Coverslips were then mounted on sections with Prolong (Molecular Probes). Images were captured using a Bio-Rad MRC 1024 laser-scanning confocal microscope.

RESULTS
The C-terminal PDZ Binding Domain of Kir2 Channels Mediates Association with the Tripartite Protein Complex Composed of CASK, Veli, and Mint1-Previously, we identified SAP97 as a Kir2-interacting protein in the brain and heart (7). Because SAP97 is a multifunctional protein that acts as a scaffold to link signaling complexes, we sought to identify other potential members of a Kir2 channel-associated complex (9,19). Recently, it was reported that the MAGUK protein CASK interacts with the N terminus and SH3 domains of SAP97 (10,11). Additionally, CASK is known to interact with the proteins Mint1 and Veli in the brain. This raised the possibility that a role for SAP97 may be to recruit CASK and its associated proteins Mint1 and Veli to Kir2 channels.
To determine whether both CASK and Kir2.2 are expressed in the same tissues, a multiple tissue Western blot with anti-Kir2.2 and anti-CASK antibodies was performed. We observed expression of CASK and Kir2.2 in brain, kidney, lung, liver, pancreas, and spleen ( Fig. 1). In the cell lines tested, three cell lines derived from kidney (COS, MDCK, and tSA201) expressed CASK but not Kir2.2, whereas the neuronal-like PC12 cell line expressed Kir2.2 but not CASK (Fig. 1). Previous protein expression analysis also indicated expression of SAP97 in these tissues (7). The Veli proteins are expressed in brain, heart, kidney, lung, liver, and spleen as well (14,20). Mint1 is expressed in neuronal tissues (21) and heart (17). Therefore, a complex composed of Kir2.2, SAP97, CASK, Veli, and Mint1 is possible in brain and heart, whereas complexes lacking Mint1 may form in other tissues.
To determine whether CASK can associate with the C-terminal portion of Kir2 channels, we performed GST fusion protein affinity chromatography with C termini of Kir2.1, Kir2.2, and Kir2.3. GST was fused to ϳ50 C-terminal amino acids of Kir2.1, Kir2.2, and Kir2.3 ( Fig. 2A). GST fusion proteins were incubated with either cerebellum or heart extracts, and bound proteins were resolved by SDS-PAGE, transferred to nitrocellulose, and immunoblotted with an anti-CASK antibody. The C termini of all of these Kir2 channels associated with CASK from cerebellum and heart extracts ( Fig. 2B, upper panels, lanes 3, 4, and 6). Association with the Kir2.1 construct could be observed with longer film exposure times (data not shown). The interaction with CASK was specific for the C-terminal amino acids of Kir2.1, Kir2.2, and Kir2.3 because neither GST nor Kir6.2, which lacks a PDZ-binding motif ( Fig. 2A), interacted with CASK ( Fig. 2B, upper panels, lanes 2 and 8). To test whether the presence of a C-terminal PDZ-binding motif was sufficient for CASK association, we constructed a C-terminal fusion protein of TASK (TASK), a tandem pore potassium channel containing a similar putative PDZ-binding motif (SSV) ( Fig.  2A). CASK was not detected when cerebellum or heart extracts were incubated with the TASK fusion protein, despite the presence of an intact consensus PDZ-binding motif (Fig. 2B, upper panels, lane 7).
The PDZ domain binding (T/S)X(V/I) motif is present in several ion channels, including Kir2 channels, and is necessary for interactions with the MAGUK proteins PSD-95 (6,22,23) and SAP97 (7). If CASK is binding directly to the channel via its PDZ binding domain or via another protein utilizing a PDZ domain, we predicted that deletion of the C-terminal PDZinteracting motif (SEI) of Kir2.2 should abolish the ability of Kir2.2 to interact with CASK. Deletion of the three C-terminal amino acids of the GST-Kir2.2 C-terminal fusion construct (Kir2.2⌬3, Fig. 2A) resulted in a complete loss of CASK association, demonstrating the absolute requirement for this Cterminal PDZ binding domain for CASK association with the channel (Fig. 2B, upper panels, lane 5). It was recently demonstrated that CASK can directly associate with SAP97 (10, 11); therefore, a full-length SAP97 construct with GST fused to the N terminus of SAP97 was used as a positive control for CASK interaction. The SAP97 fusion protein interacted with CASK from both cerebellum and heart (Fig. 2B, upper panels, lane 9). Staining the immunoblot with the protein stain Ponceau S shows approximately equal amounts of GST fusion proteins were used and recovered (Fig. 2B, lower panel).
To determine whether CASK, Veli, and Mint1 are recruited to Kir2 channels, scaled up affinity interactions were performed by using GST-Kir2.2 and GST-Kir2.2⌬3 fusion proteins linked to Affi-Gel-10 matrix. SAP97, CASK, Mint1, and Veli (using a pan-Veli antibody; most likely the Veli-1 protein based on the apparent molecular weight) were all detected on immunoblots of eluates from these columns but were not observed in eluates from the control Affi-Gel-Kir2.2⌬3 columns (Fig. 2C). This shows that the association of these proteins is specific and occurs via the PDZ binding domain at the C terminus of Kir2.2. The ␣-amino-3-hydroxy-5-hydroxy-5-methyl-4-isoxazoleproprionate glutamate receptor GluR1, which also interacts with SAP97 (24), was not isolated with the Kir2.2 column, indicating that it is probably not a component of a macromolecular complex with Kir2.2. Additionally, we have performed mass spectrometry analysis on samples from these columns, and we have confirmed the presence or absence of the proteins described above (17). These data show that CASK, Veli, and Mint1 can associate with the Kir2.2 C terminus via interaction with the channel PDZ binding motif.
To identify the Veli isoform(s) associated with the Kir2 chan- nels, pulldown experiments were performed, and immunoblots were probed with isoform-specific anti-Veli antibodies. GST pulldown experiments with rat cerebellum extracts with Kir2.2 and Kir2.3 fusion proteins (scaled down 50-fold from the affinity columns) readily indicated that SAP97, PSD-95, CASK, Veli-1, and Veli-3 can associate with the Kir2.2 and Kir2.3 C termini (Fig. 2D, lanes 3 and 5) but not with the GST alone or the Kir2.2⌬3 construct lacking the three C-terminal amino acids (Fig. 2D, lanes 2 and 4). Veli-2 was not detected in these GST pulldown assays even though it was present in the extract (Fig. 2D, lane 1), possibly due to lower overall expression levels in this tissue. These data agree with mass spectrometry anal-ysis, where Veli-1 and Veli-3 associated with the channel C terminus but Veli-2 was not detected (17). Mint1 was not detected in these GST pulldown assays with the Kir2 channel complex possibly due to weaker interactions or lower overall expression levels in the starting material (data not shown). These affinity interaction experiments indicate that brain-expressed SAP97, CASK, and at least two isoforms of Veli (Veli-1 and Veli-3), and Mint1 can all associate specifically with the C-terminal PDZ-binding motif of Kir2 channels.
CASK Associates with Kir2.2 When Co-expressed in a Cellular Environment-To verify that Kir2.2 interacts with CASK in a cellular environment, we expressed both proteins alone and FIG. 2. The CASK-Mint1-Veli tripartite complex specifically associates with the C termini of Kir2 channels. A, schematic representation of Kir2 channel subunit topology. For use in pulldown assays, GST was fused to the C termini of Kir2.1, Kir2.2, Kir2.3, Kir2.2⌬3, TASK, and Kir6.2. The boxed region indicates the ϳ50 C-terminal amino acids that were used in constructing the GST fusion proteins used in pulldown assay. Numbers represent starting amino acids of each construct, and the single letter code for the extreme C-terminal amino acids is shown. GST fused to full-length SAP97 is shown with modular domains schematically depicted. NT, N terminus. B, upper panels, glutathione-agarose charged with the respective GST fusion protein was incubated with detergent-solubilized rat cerebellum or heart proteins as indicated at the left. Bound proteins were analyzed by probing the immunoblots with anti-CASK antibodies. The Input lanes contain 1% of protein used in each binding reaction. Lower panel, to confirm that equal amounts of fusion protein were used and recovered, proteins were stained with Ponceau S. C, GST-Kir2.2 and GST-Kir2.2⌬3 fusion proteins were cross-linked to Affi-Gel-10 matrix. 50 l of affinity matrix was incubated with 100 mg of detergent-solubilized brain extract and eluted with an excess of peptide corresponding to the last 10 amino acids of Kir2.2 (see "Experimental Procedures"). Eluates from these columns were analyzed by probing individual immunoblots with anti-SAP97 (monoclonal which also recognizes PSD-95), anti-CASK, anti-pan-Veli, anti-Mint1, or anti-GluR1 antibodies. The Input lane contains 25 g (0.025%) of the protein loaded onto the column. D, upper panels, glutathione-agarose charged with the respective GST fusion protein was incubated with detergent-solubilized rat cerebellum extracts. Bound proteins were analyzed by probing individual immunoblots with the antibodies indicated to the right of each panel. The Input lanes contain 1% of protein used in each binding reaction. Lower panel, to confirm that equal amounts of fusion protein were used and recovered, proteins transferred to immunoblots were stained with Ponceau S. in combination in COS-1 cells. These cells also express endogenous SAP97 and CASK. By using an antibody against the N terminus of Kir2.1/2.2, we immunoprecipitated Kir2.2 from COS-1 cell extracts. Endogenous CASK as well as transfected CASK co-immunoprecipitated with Kir2.2 (Fig. 3A, lanes 2 and  5). When CASK was co-expressed with a mutant Kir2.2 construct with the last three C-terminal amino acids deleted (Kir2.2⌬3), no CASK co-immunoprecipitation could be detected (Fig. 3A, lane 4). Additionally, CASK did not co-immunoprecipitate if Kir2.2 was not transfected or if non-immune rabbit IgG was used as the precipitating antibody (Fig. 3A, lanes 1, 3,  and 6). Expression levels of Kir2.2, Kir2.2⌬3, and CASK were comparable between transfections as indicated by immunoblotting a fraction of the immunoprecipitation inputs (Fig. 3A, lower panels). This shows that full-length Kir2.2 can associate with CASK in cells, and the C terminus of Kir2.2 encoding the PDZ-binding motif is required for the association with CASK.
Kir2.2 Associates with CASK in the Cerebellum and Heart-Do Kir2 channels and CASK associate in brain tissue? Because both Kir2.2 and CASK are abundantly expressed in the cerebellum (25)(26)(27)(28)(29), we performed co-immunoprecipitations with the anti-Kir2.1/2.2 antibody from rat cerebellum extracts. Because there is very little Kir2.1 expressed in the cerebellum (25,26), most of the primary precipitating protein should be Kir2.2. Co-immunoprecipitation of CASK was detected with the anti-Kir2.1/2.2 immunoprecipitation but not with control rabbit IgG (Fig. 3C, upper panel). Stripping and reprobing the same blot showed that SAP97 co-immunoprecipitated in these extracts, as shown previously (Fig. 3C, upper panel) (7). This demonstrates that Kir2.2 is associated with CASK in the cerebellum and provides evidence that CASK and SAP97 may be associated together with Kir2 channels.
We also performed an identical co-immunoprecipitation with the anti-Kir2.1/2.2 antibody from rat heart extracts. Co-immunoprecipitation of CASK was detected with the anti-Kir2.1/2.2 immunoprecipitation but not with control rabbit IgG (Fig. 3C,  lower panel). These data demonstrate that the Kir2 channel association with CASK is not restricted to brain but may be important in many different tissues. Furthermore, the data suggest that a similar Kir2 channel complex containing CASK and SAP97 also forms in heart tissue.
The C Terminus of Kir2.2 Interacts with CASK through Its Direct Interaction with SAP97-There are two possibilities for how CASK associates with Kir2.2. It is either associated directly with the channel or binds through interactions with other proteins. We demonstrated previously that the second PDZ domain of SAP97 binds to the Kir2.2 C terminus (7) and recent reports (10,11) have detected the interaction of SAP97 with CASK in epithelial cells. This leads to the hypothesis that SAP97 binds to both CASK and Kir2.2 simultaneously, thus acting as a bridging scaffold. To determine whether SAP97 and CASK associate in brain, extracts were immunoprecipitated with anti-SAP97 antibodies followed by immunoblotting for CASK. Here we show that CASK robustly co-immunoprecipitates with SAP97 in brain using two different anti-SAP97 antibodies (Fig. 3D). Immunoprecipitations with non-immune rabbit IgG control antibodies showed no co-precipitating CASK. These data suggest that Kir2.2, SAP97, and CASK form a complex in cerebellum.
PDZ domains have been classified into three different subfamilies based on their ligand specificity. Class I PDZ domains have the consensus ligand binding sequence (S/T)X⌽-COOH; class II PDZ domains have the consensus ligand sequence ⌽X⌽-COOH; and class III PDZ domains have the consensus ligand sequence (E/D)XW(C/S)-COOH (where ⌽ indicates a hydrophobic residue) (30,31). CASK contains a class II PDZ domain, which makes it an unlikely candidate for direct interaction with Kir2 channels. Class II PDZ domains have a preference for tyrosine or hydrophobic amino acids at the C-terminal Ϫ2 position of binding polypeptides (31), whereas the Kir2 channel C termini contain a serine at this position (class I ligand binding sequence SEI). This led to the prediction that CASK does not interact directly with Kir2 channels but is tethered via its association with SAP97, which directly interacts with the class I PDZ-binding motif of Kir2 channels. To address this hypothesis, we tested the ability of bacterially expressed recombinant proteins to interact with the C terminus of Kir2.2 in vitro. GST, GST-Kir2.2, or GST-Kir2.2⌬3 fusion proteins attached to glutathione-agarose beads were incubated with or without bacterially expressed and purified full-length SAP97 followed by incubation with bacterially expressed CASK extracts. Unbound protein was washed away, and then bound proteins were separated by SDS-PAGE and immunoblotted with anti-CASK antibodies. CASK only was detected in the pulldown with the intact Kir2.2 C terminus with the addition of SAP97 (Fig. 4, lane 6). The Kir2.2 fusion construct did not associate with CASK without added SAP97 (Fig. 4, lane 3). Neither GST nor the Kir2.2⌬3 construct associated with CASK in the presence or absence of SAP97 (Fig. 4,  lanes 2, 4, 5, and 7), which demonstrates that association of CASK with the Kir2.2 C terminus requires the intact PDZbinding motif. As a positive control, a GST-SAP97 construct lacking the GK domain (SAP97⌬GK; shown to interact with CASK more robustly than full-length SAP97 (11)) efficiently pulled down CASK (Fig. 4, lane 8). These data demonstrate that CASK does not bind directly the Kir2 channel C terminus but is associated with the potassium channel via its interaction with SAP97.
SAP97 Mediates the Association of Kir2.2 with the CASK-Veli-Mint1 Complex-Previous data demonstrated that the N terminus of SAP97 is critical for interaction with CASK, with the SH3 domain also contributing to the association in epithelial cells (10,11). To confirm that these interactions occur in brain, we performed affinity pulldowns using GST fusion proteins encoding different domains of SAP97 (Fig. 5A). CASK was detected in all the pulldowns using constructs containing the SAP97 N terminus (Fig. 5B, lanes 3 and 5-7) including a SAP97NT construct that encodes only the N-terminal amino acids up to but not including PDZ1 (Fig. 5A). This is consistent with a previous report indicating that CASK interacts with SAP97 via its N terminus in epithelial cells (10). No association was observed with GST alone or when the N terminus was deleted (SAP97PDZ1-2) (Fig. 5B, lanes 2 and 4). An identical interaction pattern to CASK was observed for Veli-1 and Veli-3 with the N terminus of SAP97 being sufficient for association. Mint1 was also found to associate with SAP97 (Fig. 5B, lanes 6  and 7), but only the full-length SAP97 and SAP97⌬GK constructs were able to associate, suggesting that the SH3 domain of SAP97 may participate in strengthening the association with Mint1. This further supports the idea that Mint1 associates with Kir2 channels via interactions mediated by SAP97. Importantly, we were also able to pull down Kir2.2 from these brain extracts with SAP97 constructs that contain the PDZ1-2 domains (Fig. 5, lanes 4 -7) (7), further supporting the hypothesis that Kir2.2, SAP97, CASK, Veli, and Mint1 form a macromolecular complex. These data indicate that the N terminus of SAP97 is critical for association with CASK and Veli, but Mint1 may require additional interactions with SAP97 to form a stable Kir2.2-SAP97-CASK-Veli-Mint1 complex.
Veli-1, Veli-2, and Veli-3 Can Directly Associate with the C Terminus of Kir2.2-To determine whether Kir2.2 and individual isoforms of Veli do indeed interact in cells, we co-transfected Kir2.2 pairwise with FLAG-tagged Veli-1, Veli-2, and Veli-3. Immunoprecipitations were performed on detergentsolubilized COS-1 cell extracts with an anti-FLAG antibody followed by immunoblot analysis with the Kir2.2 antibody. Veli-1, Veli-2, and Veli-3 all associated with Kir2.2 (Fig. 6A,  lanes 1, 3 and 5, upper panel). When the control Kir2.2⌬3 construct was co-transfected with any of the Veli isoforms, no co-immunoprecipitation with Kir2.2⌬3 could be detected (Fig.  6A, lanes 2, 4, and 6, upper panel). Equal amounts of Kir2.2 and Veli proteins were expressed in the cells used in these experiments (Fig. 6A, Input, lower panels). These data indicate that all three isoforms of Veli can associate with Kir2.2 in a cellular environment, and this interaction requires the three C-terminal amino acids of the channel protein.
Because Veli proteins contain a single class I PDZ domain that has been shown to interact with NMDA receptors in brain (20) and Kir2.3 in epithelial cells (16), we hypothesized that at least the Veli-1 and Veli-3 isoforms, shown to associate with used in each immunoprecipitation. D, detergent-solubilized protein from rat cerebellum was used for co-immunoprecipitations with anti-SAP97-A antibody, anti-SAP97-B antibody, or control rabbit IgG. Immunoprecipitated CASK was detected by immunoblotting with a monoclonal anti-CASK antibody. The Input lane contains 1% of the protein used in each immunoprecipitation.

FIG. 3. CASK associates with Kir2 channels in vivo.
A, upper panel, detergent-solubilized protein from COS-1 cells transiently transfected with the constructs indicated (ϩ) was immunoprecipitated (IP) with antibodies to the N terminus of Kir2.1/2.2 or control IgG. CASK was detected on immunoblots with an anti-CASK antibody. Lower panels, 2% of the protein used in each immunoprecipitation was immunoblotted with the antibodies indicated at the right. B, upper panels, detergent-solubilized protein from COS-1 cells transiently transfected with CASK alone or with CASK plus GFP-Kir2.1, GFP-Kir2.2, and GFP-Kir2.3 fusion proteins was immunoprecipitated with rabbit polyclonal anti-GFP antibodies. Immunoprecipitated proteins were detected on immunoblots with an anti-CASK antibody or monoclonal anti-GFP antibody. Lower panel, 2% of the protein used in each immunoprecipitation was immunoblotted with the anti-CASK antibody. C, upper panels, detergent-solubilized protein from rat cerebellum was used for co-immunoprecipitations with antibodies to the N terminus of Kir2.1/ 2.2 or control rabbit IgG. Immunoprecipitated CASK was detected by immunoblotting with a monoclonal anti-CASK antibody. The immunoblot was stripped and reprobed with anti-SAP97 antibody. Lower panel, detergent-solubilized rat heart protein was used for co-immunoprecipitations with the anti-Kir2.1/2.2 antibody or control rabbit IgG, and CASK was detected as above. The Input lanes contain 1% of the protein Kir2.2 C terminus from brain extracts (Fig. 2, C and D), may interact with the C terminus of Kir2.2 directly. To determine a direct interaction with the channel C terminus, we expressed all three isoforms of Veli as His 6 -tagged proteins in bacteria and purified the proteins with nickel-agarose affinity chromatography. We then incubated these recombinant proteins with GST, Kir2.2, Kir2.2⌬3, or SAP97 fusion proteins. The Kir2.2 construct interacted directly with Veli-1, Veli-2, and Veli-3 (Fig. 6B, lane 3), but the recombinant Veli proteins did not interact with GST alone, Kir2.2⌬3, or SAP97 (Fig. 6B, lanes 2,  4, and 5). These data demonstrate that Veli-1, Veli-2, and Veli-3 are capable of directly interacting with the Kir2.2 C terminus. In addition, these proteins do not interact directly with SAP97, indicating that Veli association with SAP97, in vivo must occur via another linking protein, most likely CASK.
Veli-1 and Veli-3 but Not Veli-2 Recruit CASK to the C Terminus of Kir2.2-We have already demonstrated that SAP97 can link CASK to the Kir2.2 C terminus (Fig. 4). Because the Veli proteins bind directly to the C-terminal Kir2.2 construct (Fig. 6B) and have been shown to interact with the L27C domain of CASK (10), we postulated that all three Veli A, GST fused to full-length SAP97 as well as various domains or deletion constructs is shown with modular domains schematically depicted. B, glutathione-agarose charged with the respective GST fusion protein was incubated with detergent-solubilized rat brain protein extracts. Bound proteins were analyzed by probing individual immunoblots with the antibodies indicated to the right of each panel. The Input lanes contain 0.5% of protein used in each binding reaction. C, detergent-solubilized protein from rat cerebellum was used for co-immunoprecipitations with the anti-SAP97-B antibody or control rabbit IgG. Immunoprecipitated proteins were analyzed by probing individual immunoblots with the antibodies indicated below each panel. * indicates upper band above CASK that likely represents incompletely stripped anti-SAP97 antibody. The Input lane contains 0.5% of the protein used in each immunoprecipitation.
FIG. 6. Veli-1, Veli-2, and Veli-3 directly associate with the C terminus of Kir2.2 and recruit CASK. A, COS-1 cells were cotransfected with FLAG-tagged Veli-1, Veli-2, or Veli-3 and either Kir2.2 or Kir2.2⌬3. Upper panel, detergent-solubilized extracts from these cells were used in co-immunoprecipitations (IP) with anti-FLAG antibody followed by immunoblotting with anti-Kir2.2 antibody. Lower panels, inputs are equal to 10% of the protein used in co-immunoprecipitation. B, upper panels, glutathione-agarose charged with the respective GST fusion protein was incubated with recombinant purified Veli-1, Veli-2, or Veli-3. Bound proteins were analyzed by probing individual immunoblots with the antibodies indicated to the right of each panel. The Input lanes contain 25% of protein used in each binding reaction. Lower panel, to confirm that equal amounts of fusion protein were used and recovered, proteins transferred to immunoblots were stained with Ponceau S. isoforms could also recruit CASK to Kir2.2. Kir2.2 fusion proteins attached to glutathione-agarose beads were combined with recombinant CASK extracts with and without recombinant Veli-1, Veli-2 or Veli-3. Bound proteins were separated by SDS-PAGE and probed with an anti-CASK antibody. The control protein constructs (GST alone or Kir2.2⌬3) showed no detectable CASK interaction with or without added Veli proteins (Fig. 7, lanes 2, 4, 6, and 8). CASK did not interact with Kir2.2 directly (Fig. 7, lane 3) but could be detected when either Veli-1 (Fig. 7, lane 7, upper panels) or Veli-3 (Fig. 7, lane 7, lower panels) were added. Surprisingly, the addition of the recombinant Veli-2 did not facilitate recruitment of CASK (Fig.  7, lane 7, middle panels). The same immunoblots were stripped and reprobed for Veli-1, Veli-2, or Veli-3, revealing that each Veli protein associated with the Kir2.2 construct in these assays (Fig. 7, lane 7). This demonstrates the ability of Veli-1 and Veli-3, like SAP97, to bind directly to the Kir2.2 C terminus and recruit CASK to the channel. The SAP97⌬GK construct was used as a positive control for interaction with CASK, which bound CASK both in the absence and presence of Veli (Fig. 7,  lanes 5 and 9). Surprisingly, the Veli isoforms were not recruited to SAP97⌬GK in the presence of bound CASK (Fig. 7,  lane 9). These data indicate that Veli-1 and Veli-3 bind directly to the Kir2.2 C terminus and participate in linking CASK to the channel.
Different Isoforms of Veli Participate in Distinct Multiprotein Complexes-To determine which isoforms of Veli are likely to be involved in complex with CASK, SAP97, Mint1, and Kir2.2 in brain, co-immunoprecipitation experiments using rat cerebellum or brain extracts were performed with isoform-specific anti-Veli antibodies. Fig. 8 shows results from co-immunoprecipitations from cerebellar extracts (Fig. 8A) and whole brain extracts (Fig. 8B). Immunoblots indicate that CASK associates with all three Veli proteins (Fig. 8, A and B, lanes 3-5). However, only anti-Veli-1 and anti-Veli-3 antibodies co-immunoprecipitated SAP97 (Fig. 8, A and B, lanes 3 and 5), with much more SAP97 detected in conjunction with Veli-3. SAP97 could not be detected in co-immunoprecipitations with anti-Veli-2 antibodies either from cerebellum or whole brain (where Veli-2 is more abundant) ( Fig. 8A and B, lane 4). Mint1 co-immunoprecipitated with both anti-Veli-2 and anti-Veli-3 antibodies (Fig. 8, A and B, lanes 4 and 5); however, Mint1 could not be detected in immunoprecipitates with anti-Veli-1 antibody (Fig.  8, A and B, lane 3). This indicates that Mint1 probably only associates with Veli-2 or Veli-3 in brain. The data from these co-immunoprecipitations suggest that the most likely Kir2.2-

FIG. 7. Veli-1 and Veli-3 but not Veli-2 recruits CASK to the C terminus of Kir2.2.
Glutathione-agarose charged with 2 g of the respective GST fusion protein was incubated with detergent-solubilized CASK bacterial lysate with (ϩ) or without (Ϫ) 2 g of added purified Veli-1 or Veli-3. The immunoblot of the bound protein was probed with the anti-CASK antibody followed by stripping and reprobing with anti-Veli antibodies. The Input lane contains 1% (CASK extract) or 0.5 g (purified Veli) of the protein used in each affinity pulldown. Lower panel, to confirm that equal amounts of fusion protein were used and recovered, proteins transferred to immunoblots were stained with Ponceau S.
FIG. 8. Distinct protein complexes co-immunoprecipitate with isoform-specific anti-Veli antibodies in brain. Detergent-solubilized protein from rat cerebellum (A) or whole brain (B) was used for co-immunoprecipitations with isoform-specific anti-Veli antibodies (lanes [3][4][5] or control rabbit IgG (lane 2). Immunoprecipitated proteins were analyzed by probing individual immunoblots with the antibodies indicated below each panel. * indicates position of band that most likely represents incompletely stripped anti-SAP97 antibody. The Input lanes contain 0.5% of the protein used in each immunoprecipitation. associated complexes in brain are composed of Kir2.2, SAP97, CASK, Veli-3, with Mint1 and a complex with Kir2.2, SAP97, CASK, and Veli-1 without Mint1. Additionally, a complex composed of Veli-2, CASK, and Mint1 excluding SAP97 and Kir2 channels is possible. We could not directly probe for the presence of Kir2.2 in these immunoprecipitates because the Kir2.2 antibody was made in the same species as the anti-Veli antibodies (rabbit). In attempts to use a secondary antibody directed against rabbit IgG, the detection of Kir2.2 at ϳ50 kDa was occluded by the secondary antibody recognition of the immunoprecipitating anti-Veli antibodies.
Kir2.2 Displays an Overlapping Expression Pattern with SAP97, CASK, Veli-1, Veli-2, Veli-3, and Mint1 in the Cerebellum-Kir2.2, SAP97, CASK, Veli-1, Veli-2, Veli-3, and Mint1 have all been reported to be expressed in the cerebellum (7,26,27,29,(32)(33)(34). Our GST fusion protein affinity pulldown and co-immunoprecipitation experiments from cerebellar extracts provide biochemical evidence for the association of these proteins in a macromolecular complex. Because this complex of proteins was identified in the cerebellum, we predicted that their expression patterns would overlap. To address this hypothesis, we performed immunofluorescence confocal microscopy on fixed rat cerebellar tissue sections.
The cerebellum, known for its role in motor control and learning, has an invariant structure with well characterized anatomy in normal animals (35). We focused on the outer layers of the cerebellum where strong Kir2.2 expression was observed previously (7,26). Kir2.2 shows a striking punctate labeling pattern along radial processes extending from the Purkinje cell layer to the molecular layer along with punctate labeling in the granule cell layer (Fig. 9A) (7, 26). Previously, we demonstrated that the Kir2.2 expression was co-localized with the glial cell marker GFAP in Bergmann glia in the molecular layer and astrocytes in the granule cell layer, thus indicating the presence of Kir2.2 in glial cells of the cerebellum (7). CASK immunostaining appeared to be ubiquitous, present in all three cerebellar layers uniformly (Fig. 9B). As observed previously (7,34), SAP97 showed intense labeling in the molecular layer as well as expression in the granule cell layer (Fig.  9C). Co-localization of CASK with SAP97 showed overlapping distribution throughout the cerebellum (Fig. 9, B and C), consistent with the intimate association of these two proteins supported by the robust co-immunoprecipitation data (Fig. 3D) (10,11). Mint1 appeared to be expressed in all three cerebellar layers, including the molecular layer and granule cell layer (Fig. 9G), both areas that express Kir2.2 (Fig. 9A) (7, 25). Veli-1 expression was clearly seen in the Purkinje cell bodies with prominent labeling in the granule cell layer (Fig. 9D). Staining with the anti-Veli-2 antibody was relatively weak, with very little staining in the molecular layer and only slightly more in the granule cell layer (Fig. 9E). Veli-3 labeling appeared to be highly expressed in the molecular layer (Fig. 9F) in an inverse pattern to Veli-1. These Veli immunostaining patterns are in agreement with previous studies (33). Background staining of the cerebellum was negligible when the primary antibodies were omitted (secondary antibodies: anti-rabbit FITC (Fig. 9H, FITC) or anti-mouse CY3 (Fig. 9I, CY3)). Comparison of the immunoprecipitation data ( Fig. 8) with the immunocytochemical distribution suggests that Kir2.2, CASK, and SAP97 associate with Veli-1 primarily in the granule cell layer and with Veli-3 primarily in the molecular layer. Because Mint1 was only detected in association with SAP97, CASK, and Veli-3 (Fig. 8, A and B, lane 5), it is likely that a complex with Kir2.2, SAP97, CASK, Veli-3, and Mint1 would primarily be found in the molecular layer.
Dominant Negative CASK Alters Basolateral Localization of Kir2.2 in Polarized Epithelial Cells-To address the role of CASK in Kir2 channel trafficking, we focused on the polarized kidney epithelial MDCK cell line, in which membrane proteins containing PDZ-binding motifs often specifically target and localize to the basolateral membrane (16,36,37). To determine the membrane localization of Kir2.2, we imaged transfected, polarized MDCK cells in both the XY and XZ focal planes by immunofluorescence confocal microscopy. Indeed, stable transfection of the GFP-Kir2.2 construct in MDCK cells resulted in basolateral expression of the channel (Fig. 10, A and D) in agreement with Kir2.3 localization in these cells (16). The GFP-Kir2.2⌬3 construct, which lacks the PDZ-binding motif, displayed incomplete basolateral localization, with much of the channel expressed intracellularly and mislocalized to the apical surface (Fig. 10, G and J). This suggests that the PDZbinding motif of Kir2.2 is important for targeting and/or stabilization of the channels to the basolateral plasma membrane. The targeting and stabilization of GFP-Kir2.2 to the basolateral membrane suggests that PDZ domain-containing interacting proteins may play a role in these functions. Because both of the Kir2.2-associated proteins SAP97 and CASK are expressed at the basolateral membrane in MDCK cells (Fig. 10, C, F, I, and L) (10,16), these proteins could possibly play roles in channel targeting and membrane stabilization. Lee et al. (10) has demonstrated that a mutant form of CASK that lacks the final 297 amino acids (CASK-(1-612)) acts in a dominant negative fashion and causes the mislocalization of SAP97 in MDCK cells. Because SAP97 directly interacts with Kir2.2 (Fig. 4) (Fig. 10P). As a control, overexpression of SAP97 construct (Fig. 10, U and X) did not affect the GFP-Kir2.2 localization (Fig. 10, S and V), indicating that the effect of the dominant negative CASK was not simply an effect of protein overexpression. These data indicate that a functional role for the association of Kir2.2 with the CASK-SAP97 complex in polarized epithelial cells is in subcellular trafficking and/or localization of channels at the plasma membrane.

DISCUSSION
In this study we demonstrate for the first time that the inward rectifier potassium channels Kir2.1, Kir2.2, and Kir2.3 associate in brain with a complex of proteins that include SAP97, CASK, Veli-1, Veli-3, and Mint1. Our results indicate that the complex of associated proteins is recruited to Kir2 channels via direct binding of either SAP97 or Veli. Furthermore, our results show that the three isoforms of Veli participate in the formation of distinct protein complexes in brain. Veli-1 associates with Kir2 channels, SAP97, and CASK, whereas Veli-3 associates with Kir2 channels, SAP97, CASK, and Mint1. Veli-2 does not appear to form detectable amounts of complex with Kir2 channels or SAP97. A dominant interfer-ing CASK construct resulted in the redistribution of Kir2.2 from the basolateral membrane of MDCK cells and an accumulation of intracellular Kir2.2, indicating an essential role for CASK and its interacting proteins in Kir2 channel trafficking.
A Kir2 Channel Complex with SAP97, CASK, Veli, and Mint1-Affinity pulldown assays and co-immunoprecipitations demonstrated that the three C-terminal amino acids of the Kir2.2 channel that encode a PDZ interaction motif (SEI) are required for the association of any of the identified complex proteins (SAP97, CASK, Veli-1, Veli-3, and Mint1) with the channel (Figs. 2, 3A, 4, 6, and 7). Co-immunoprecipitation experiments using detergent-solubilized cerebellum extracts showed that the previously identified Kir2 channel-interacting protein SAP97 interacts with CASK, and both of these proteins associate with Kir2.2 in native tissue (Fig. 3, C and D). Further experiments utilizing GST-SAP97 pulldown assays as well as co-immunoprecipitations with anti-SAP97 antibodies demonstrated that the CASK-Veli-Mint complex associates with SAP97 as well as Kir2.2 (Fig. 5). In vitro protein interaction assays showed that CASK cannot bind to the C terminus of Kir2 channels directly but associates with the channel through its direct interaction with SAP97 (Fig. 4). We also investigated the association of the three Veli protein isoforms with Kir2.2, and we demonstrated that the proteins co-immunoprecipitate when co-expressed in heterologous cells as well as directly interact with the Kir2.2 C terminus using in vitro interaction assays (Fig. 6). This interaction is most likely mediated via its PDZ domain and does not require any other scaffolding pro- teins. Like SAP97, we also demonstrated that both Veli-1 and Veli-3 can recruit CASK to the Kir2 channel C terminus (Fig. 7).
From these data, we hypothesize that two alternate complexes of Kir2 with SAP97-CASK-Veli-Mint1 are present in brain, with different proteins linking Kir2 to the channel-associated proteins (Fig. 11). In one complex, SAP97 forms the link (Fig. 11A); our data support a model in which the PDZ2 domain of SAP97 interacts directly with the Kir2 channel C terminus and recruits CASK (Fig. 4) (7). This interaction is likely to occur through binding of the N-terminal domain of SAP97 to the N-terminal L27C domain of CASK (10,13,14,32). Coimmunoprecipitation data (Figs. 5C and 8) and affinity interaction assays (Figs. 2C and 5B) suggest that Veli and Mint1 associate with this complex, most likely by Veli interacting with the L27C domain of CASK (10) and Mint1 interacting with the calmodulin kinase II-like domain of CASK (13, 14, 32) (Fig. 11A). In the second complex, isoforms of Veli bind directly to the channel; our data support a model (Fig. 11B) in which the PDZ domain of Veli-1 or Veli-3 interact directly with the Kir2 channel C terminus and recruit CASK (Fig. 7), presumably through binding of the Veli L27 domain to the CASK C-terminal L27C domain (10,14). In this alternative complex, SAP97 and Mint1 are also recruited to the Kir2 channel (Fig.  11B), via interactions with the CASK N terminus (10).
Kir2 channels associated with both Veli-1 and Veli-3 in the brain, and these Veli isoforms formed distinct protein complexes. Co-immunoprecipitations with brain extracts using isoform-specific anti-Veli antibodies indicate that only Veli-1 and Veli-3 associate with SAP97 (Fig. 8). Of these two isoforms, only Veli-3 stably associated with Mint1. These data suggest that Kir2 channel complexes that contain Veli-1 exclude Mint1, whereas Kir2 complexes that contain Veli-3 include Mint1 (Fig.  11). We could not detect Veli-2 association with the C terminus of Kir2.2 or Kir2.3 (Fig. 2D) or complex formation with SAP97 ( Fig. 5B and Fig. 8) in GST pulldown assays using brain extract. This indicates that a complex containing Kir2 channels and SAP97 is unlikely to contain Veli-2 in the brain. Overlapping expression in adult rat cerebellum of Kir2.2, SAP97, CASK, Mint1, and Veli-1, in the granule cell layer, and of Kir2.2, SAP97, CASK, Mint1 and Veli-3, in the molecular layer, was established using immunofluorescence microscopy. These tissue distributions suggest that distinct Kir2.2-associated complexes occur in specific layers of the cerebellum.
Functional Implications of Kir2-associated SAP97-CASK-Veli-Mint1 Complexes-The MAGUK proteins SAP97 and CASK seem to have similar and/or complementary functions in trafficking and targeting ion channels and receptors in cells. Both of these proteins have been studied in epithelial cells where they are targeted and localized to the basolateral membrane at sites of cell-cell contact. The first 65 amino acids of SAP97 were found to be critical for this targeting (38). Most interesting, this is the same region of SAP97 that interacts with the L27N domain of CASK (10). Lee et al. (10) also used a CASK deletion mutant lacking the GK domain to observe its effects on SAP97 localization. This dominant negative CASK construct caused SAP97 to deviate from its normal basolateral location in MDCK epithelial cells (10). Functional genetic analyses of CASK and SAP97 in Drosophila (39,40) and Caenorhabditis elegans (12,41,42) also implicate these proteins in targeting and localization in polarized cells. Taken together, it is apparent that both SAP97 and CASK have functional roles in the trafficking and targeting of ion channels and receptors.
We expressed a GFP-Kir2.2 construct in MDCK cells and, like Kir2.3 (16), found the channels to traffic and localize to the basolateral membrane (Fig. 10, A and D). Expression of GFP-Kir2.2⌬3 construct lacking the PDZ-binding motif displayed a lack of specific membrane localization with much of the channel expressed intracellularly (Fig. 10, G and J). This suggests that the PDZ-binding motif of Kir2.2 plays a role in targeting or stabilization of the channels in the basolateral membrane. To examine the possible role of CASK, the central linking protein in this Kir2 complex, we co-expressed a dominant negative CASK construct in MDCK cells stably expressing GFP-Kir2.2. This experiment resulted in the mislocalization and redistribution of Kir2.2 from the basolateral membrane to intracellular and apical locations. Thus, it appears that the proper assembly of Kir2.2 with CASK-containing complexes is required for appropriate subcellular trafficking Kir2 channels.
We anticipate that the CASK-Veli-Mint1 complex functions in conjunction with SAP97 as a unit in trafficking of Kir2 channels. CASK may be involved in targeting the complex, Mint1 tethering the complex to motor proteins, and Veli stabilizing the complex at the plasma membrane. Data from other receptors and channels support this role of the complex. In C. elegans, the CASK/Veli/Mint1 homologs Lin-2/Lin-7/Lin-10 are essential proteins in the basolateral targeting and localization for the Let-23 tyrosine kinase receptor (12) and GLR-1 glutamate receptor in vulval epithelial cells (42) as well as serving a key role in vulval development (43). Involvement of Mint1 in trafficking has been demonstrated in neurons by Setou et al. (15), who identified the kinesin motor protein KIF17 as a binding partner for Mint1, showed association of KIF17, Mint1, CASK, Veli, and NMDA receptor, and demonstrated that vesicles containing NMDA receptors could be transported along microtubules by KIF17. Additionally, Veli has been shown to facilitate the retention of the ␥-aminobutyric acid transporter at the basolateral surface of epithelial cells (36), and recent studies demonstrated increased plasma membrane stabilization by Veli-2 of Kir2.3 expressed in Xenopus oocytes (16). These studies suggest that for Kir2 channels expressed in brain, Velis could function in a similar manner, acting to stabilize channels once they are trafficked to the plasma membrane. Because we could not detect Veli-2 associated with Kir2 channels or SAP97 in brain, this may reflect a preference of the channels for Veli-1 and Veli-3 in brain.
The interaction of Kir2 channels with the SAP97-CASK-Veli-Mint1 complex may serve to attach channels to cytoskeletal elements. Many MAGUK proteins, including SAP97 and CASK have an alternatively spliced short peptide sequence (known as the protein 4.1 binding domain or HOOK domain) between their SH3 and GK domains with the capability of interacting with protein 4.1 (44 -46). Protein 4.1 is known for its role in linking plasma membrane components to the actin/spectrin cytoskeleton (47). In a recent study, CASK was demonstrated to bind to a neuronal form of protein 4.1 and nucleate filamentous actin on the C termini of neurexin membrane proteins (46). In addition, motor and motor-like proteins have been found to associate with SAP97 including GAKIN (48), myosin VI (49), and dynein light chain (50). These data support the idea that Kir2 channels may be linked via SAP97, CASK, and associated proteins to either the actin-or microtubule-based cytoskeleton and implicate the complex of channel-associated proteins in intracellular channel movement or channel attachment to cytoskeletal elements at the plasma membrane.
Another role of the Kir2-associated SAP97-CASK-Veli-Mint1 complex may be to form a molecular scaffold that recruits signaling molecules to Kir2 channels, which may then modulate the channel activity. MAGUKs have been demonstrated to associate with a number of signaling molecules including protein kinase A (9), Src family tyrosine kinases (51,52), and GTPase-associated proteins (53,54).
How might this Kir2 channel-associated complex be regulated? An intriguing possibility is that the Kir2-associated complex is regulated by phosphorylation. It was shown previously that protein kinase A phosphorylation of the serine within the PDZ-binding motif at the C terminus of Kir2 channels abolishes the ability of the channels to interact with MAGUK proteins (6,7). An interesting possible scenario is that Kir2 channels are trafficked to the plasma membrane by the SAP97-CASK-Veli-Mint1 complex, transiently phosphorylated by protein kinase A to release Kir2 channels from the complex, followed by interaction with a membrane-associated MAGUK such as PSD-95, which would stabilize, anchor, and cluster the channels. Additionally, it is likely that this complex may be regulated by intramolecular and/or intermolecular interactions as well as post-translational modifications and cytoplasmic calcium levels (29,32,(55)(56)(57)(58)(59).
In summary, we have identified novel macromolecular complexes containing SAP97, CASK, Veli-1, Veli-3, and Mint1 that are assembled at the C termini of Kir2 channels in brain. These complexes are capable of binding to Kir2 channels through two possible mechanisms: one bound to the C terminus of Kir2 channels through SAP97, and the other with the complex associated via recruitment by Veli. In addition, distinct protein complexes are recruited by different Veli isoforms. Veli-1 is associated with CASK, SAP97, and Kir2 channels, whereas Veli-3 is associated with Kir2 channels, CASK, SAP97, and Mint1. These Kir2-associated proteins have been suggested to be involved in ion channel targeting, trafficking, and localization (15,16,20,60). CASK is a key member of this complex possibly functioning in conjunction with other complex proteins to traffic and stabilize channels at the plasma membrane. These studies have important implications in the regulation of Kir2 channels, and hence a host of physiological processes from neuronal excitability to cardiac arrhythmia. Further studies such as knockdown of individual genes using RNA interference should provide insight into the precise roles of SAP97, CASK, Veli, and Mint1 in trafficking and function of Kir2 channels.