Differential Recruitment of Kv1.4 and Kv4.2 to Lipid Rafts by PSD-95*

The activity of voltage-gated potassium (Kv) channels, and consequently their influence on cellular functions, can be substantially altered by phosphorylation. Several protein kinases that modulate Kv channel activity are found in membrane subdomains known as lipid rafts, which are thought to organize signaling complexes in the cell. Thus, we asked whether Kv1.4 and Kv4.2, two channels with critical roles in excitable cells, are found in lipid rafts. Acylation can target proteins to raft regions; however, Kv channels are not acylated, and therefore, a different mechanism must exist to bring them into these membrane subdomains. Because both Kv1.4 and Kv4.2 interact with postsynaptic density protein 95 (PSD-95), which is acylated (specifically, palmitoylated), we examined whether PSD-95 can recruit these channels to lipid rafts. We found that a portion of Kv1.4 and Kv4.2 protein in rat brain membranes is raft-associated. Lipid raft patching and immunostaining confirmed that some Kv4.2 is in Thy-1-containing rafts in rat hippocampal neurons. Using a heterologous expression system, we determined that palmitoylation of PSD-95 was crucial to its localization to lipid rafts. We then assessed the contribution of PSD-95 to the raft association of these channels. Co-expression of PSD-95 increased the amount of Kv1.4, but not Kv4.2, in lipid rafts. Deleting the PSD-95 binding motif of Kv1.4 eliminated this recruitment, as did substituting a palmitoylation-deficient PSD-95 mutant. This work represents the first evidence that PSD-95 binding can recruit Kv channels into lipid rafts, a process that could facilitate interactions with the protein kinases that affect channel activity.

cient PSD-95 mutant eliminated this recruitment. Surprisingly, a Kv1.4 deletion mutant lacking the C-terminal PDZbinding motif was largely excluded from lipid rafts, whether or not PSD-95 was co-expressed. This suggests that binding to an endogenous PDZ domain protein in tsA201 cells allowed raft localization of Kv1. 4. This work represents the first evidence that PSD-95 binding can recruit Kv channels to lipid rafts, a process that could facilitate interactions with the protein kinases that affect Kv channel activity.
Preparation of Sucrose Density Gradients-To assess the lipid raft localization of Kv1.4 and Kv4.2 in native tissue, sucrose density gradients were prepared (18) from rat brain membranes (as in Ref. 19). Rat brain membrane protein (1 mg total) was solubilized in 2 ml of 1% Triton X-100 (Sigma) in MBS (MES-buffered saline; 25 mM MES pH 6.5, 150 mM NaCl). After homogenizing with 10 up-and-down strokes of a tight-fitting Dounce homogenizer, the extract was adjusted to 45% sucrose and overlaid with 4 ml of 30% sucrose in MBS and 4 ml of 5% sucrose in MBS. The sucrose gradient was formed by centrifuging at ϳ200,000 ϫ g for 18 h at 4°C using a Beckman SW41 rotor. Twelve 1-ml aliquots were removed beginning at the top.
To assess the contribution of PSD-95 to the lipid raft localization of the Kv channels, ECFP-Kv1.4, ECFP-Kv4.2 or their respective deletion mutants were transfected alone or co-transfected with PSD-95 or C3,5S-PSD-95 into tsA201 cells (also known as HEK293T cells) using calcium phosphate precipitation (three 100-mm plates per transfection condition), as described previously (15). Forty-eight hours after transfection, cells were washed twice with Dulbecco's phosphate-buffered saline (D-PBS; 139 mM NaCl, 2.7 mM KCl, 8.8 mM Na 2 HPO 4 , 1.48 mM KH 2 PO 4 ) and solubilized in 2 ml 1% Triton X-100 in MBS. Lipid raft isolation was carried out as above although 20 up-and-down strokes were used to homogenize the lysates, which were then adjusted to 40% sucrose. As before, twelve 1-ml aliquots were removed beginning at the top.
Cholesterol Depletion and Repletion-tsA201 cells were transfected with ECFP-Kv1.4, ECFP-Kv4.2 or PSD-95 as described above. To deplete cholesterol, cells were washed once with D-PBS 48 h after transfection, and incubated with 2% 2-hydroxypropyl-␤-cyclodextrin (cyclodextrin; Sigma) in serum-free medium for 1.5 h at 37°C. Where indicated, cells were repleted with cholesterol by incubating them in a mixture of 80 g/ml cholesterol (Sigma) and 2% cyclodextrin for 1.5 h at 37°C (20). Sucrose density gradients were then prepared as above.
Slot Blotting of Sucrose Density Gradients-Ganglioside M 1 (GM 1 ) was used as a marker of lipid rafts in tsA201 cells. In order to pinpoint the location of lipid rafts in sucrose density gradients prepared from mock-transfected tsA201 cells, equal aliquots of each fraction were diluted 1:100 and applied to nitrocellulose membrane with a slot blot apparatus (Invitrogen). After extensive washing with PBS and blocking with 5% Carnation powdered skim milk, the membrane was then incubated with 1:1000 cholera toxin B subunit conjugated to horseradish peroxidase (Sigma) for 1 h at room temperature, followed by ECL.
Detection of Cellular Free Cholesterol with Filipin-Cholesterol depletion disrupts the integrity of lipid rafts, which is often demonstrated by re-location of GM 1 into higher density fractions (21,22). We proved that cholesterol depletion and repletion were successful in tsA201 cells by detecting free cellular cholesterol with filipin (Sigma) as previously described (23). Cells were plated on coverslips coated with poly-L-lysine (Sigma) and transfected with salmon sperm DNA (mock-transfected) the following day. Cholesterol depletion and repletion were carried out as described above, and results were compared with control cells incubated with serum-free DMEM for the same period. Cells were fixed for 1 h with 2% paraformaldehyde and then stained with 50 g/ml filipin in PBS for 1 h at 37°C. The coverslips were mounted with Slowfade (Molecule Probes), sealed with nail polish and imaged with a Zeiss Axioplan 2 Imaging microscope (Carl Zeiss Inc., Thornwood, NY) equipped with a digital camera (Zeiss AxioCam) and a 40ϫ objective, using identical exposure settings.
Lipid Raft Patching in Neurons-Postnatal rat hippocampal neurons were prepared as previously described (24) with some modifications. Briefly, hippocampi were dissected from 2-3-day-old rats, incubated in 2 mg/ml papain (Worthington Biochemical, Lakewood, NJ) and then mechanically triturated in Neurobasal A/B27 (Neurobasal A medium with 2% B27 supplement, 2 mM L-glutamine and 0.05 mg/ml gentamycin; all from Invitrogen, Burlington, ON, Canada). Neurons were separated from other brain cells with an Optiprep (Invitrogen) density gradient. The neuron-enriched layer was plated on poly-D-lysine (Sigma)-coated German glass coverslips (Bellco Glass, Inc., Vineland, NJ) at a density of 4 ϫ 10 4 cells/coverslip and maintained in Neurobasal A/B27. Cultures were re-fed every 4 days by removing half of the conditioned medium and replacing it with an equal volume of fresh medium. Cells were used at 21 days in culture for the lipid raft patching procedure (25). After chilling on ice for 15 min, neurons were washed with ice-cold D-PBS supplemented with 0.7 mM CaCl 2 and 0.48 mM MgCl 2 (D-PBS/CM). They were subsequently incubated with 1:50 anti-Thy-1 (the same antibody used to immunoblot rat brain membranes) in 0.1% BSA in D-PBS/CM for 1 h on ice, followed by 1:200 goat anti-mouse conjugated to Alexa Fluor 594 in 0.1% BSA in D-PBS/CM for 1 h on ice. The neurons were then fixed with 4% paraformaldehyde for 15 min on ice and methanol for 5 min at Ϫ20°C. After blocking with 5% BSA, Kv4.2 was detected with 1:100 anti-Kv4.2 (Sigma), followed by a 1 h incubation with 1:200 goat anti-rabbit conjugated to Alexa Fluor 488 (both antibodies were diluted in 0.1% BSA/D-PBS/CM). The coverslips were mounted with Slowfade and sealed with nail polish. Image stacks spanning 0.25 m (with images spaced 25 nm apart) were collected and deconvolved in AxioVision 3.1 (Carl Zeiss Inc.) using the nearest neighbors algorithm.
Statistical Analysis-Student's t test was used to analyze immunoblotting data following arcsine transformation to produce data with a normal distribution (26). Values of p Ͻ 0.05 were considered significant.

Kv1.4, Kv4.2, and PSD-95 Are Found in Lipid Rafts in Rat
Brain Membranes-Immunoblots of sucrose density gradients prepared from rat brain membranes show a portion of Kv1.4 and Kv4.2 protein in low density membrane fractions, with the majority in high density membrane fractions at the bottom of the gradient (Fig. 1). These results were replicated in five independent gradient preparations. We confirmed the quality of our preparation by immunoblotting the sucrose density gradients for a non-raft membrane protein, Na ϩ /K ϩ -ATPase (27), and a membrane protein marker of lipid rafts, Thy-1 (28). As expected, Na ϩ /K ϩ -ATPase was found exclusively in the high density membrane fractions at the bottom of the gradient, and Thy-1 was found exclusively in a low density membrane fraction ( Fig. 1).
An earlier study did not detect Kv4.2 in lipid rafts (immunoreactivity was confined to the bottom of the gradient) (29). Since that study used a lower detergent:protein ratio, which has been shown to increase the amount of raft protein extracted (30), it is surprising that a raft association for Kv4.2 was not detected. However, it should be noted that positive and negative controls for raft-and non-raft proteins were not provided in that study, and the reproducibility of the results was not discussed. Although our Kv4.2 antibody has been extensively employed to detect Kv4.2 in immunoblots of heart tissue (31, 32), brain (29,31) and transfected cells (29), as well as for immunohistochemistry of heart tissue (32), we also probed a sucrose density gradient with a commercially available antibody against Kv4.2 (6,9,33,34) and obtained similar results (data not shown).
In agreement with previous studies (35-40), we found PSD-95 in both low and high density membrane fractions, essentially mirroring the distributions of Kv1.4 and Kv4.2. Unfortunately, three of the earlier studies did not use negative controls to delineate non-raft fractions (35,36,40). However, the other three studies used both positive and negative markers for raft association (37)(38)(39), thus providing strong evidence that a portion of PSD-95 protein is localized to lipid rafts.
Palmitoylation Is Necessary to Localize PSD-95 to Lipid Rafts in tsA201 Cells-Sucrose gradients prepared from mocktransfected tsA201 cells were probed for GM 1 , which was used as a positive marker of lipid rafts because these cells do not express Thy-1 or caveolin (41). Band intensities for GM 1 peaked in fractions 4 and 5 ( Fig. 2A); therefore, these fractions were designated as the lipid raft fractions. To verify that fractions 4 and 5 were free of non-raft proteins, we also immunoblotted for Na ϩ /K ϩ -ATPase and ␤-actin. Immunoreactivity for these proteins was confined to the high-density fractions at the bottom of the gradients, confirming our ability to separate lipid rafts from other membrane and cellular components.
We then prepared density gradients from tsA201 cells transfected with wild-type PSD-95 or a palmitoylation-deficient mutant, C3,5S-PSD-95 (42). Dual palmitoylation at the N terminus is necessary for the localization of growth-associated protein-43 (GAP-43) (43), linker for activation of T cells (LAT) (44) and regulator of G-protein signaling (RGS) (45) to raft microdomains. We therefore predicted that a PSD-95 mutant lacking the requisite cysteine residues at positions 3 and 5 would not localize to lipid rafts. Wild-type PSD-95 distributed between raft and non-raft fractions, with most of the protein in the bottom of the gradient (non-raft fractions) ( Fig. 2A). In contrast, C3,5S-PSD-95 was primarily concentrated at the bottom of the gradient ( Fig. 2A), although very faint bands were detected in the low density raft fractions after overnight exposure (data not shown). This demonstrates that palmitoylation assists in bringing PSD-95 into lipid rafts in tsA201 cells. These results were replicated using three independently transfected cell batches. However, they differ from a previous study which reported that an identical PSD-95 mutant remained significantly raft-associated and that removing the N terminus and three PDZ domains was required to abolish immunoreactivity in the low density raft fractions (36). A possible explanation arises from the different cell types used for heterologous expression: tsA201 cells in our study, COS-7 cells in (36). Although both are derived from kidney tissue, they likely exhibit different patterns of protein expression. COS-7 cells may express an endogenous protein that binds to the SH3 and/or guanylate kinase domains in the C terminus of PSD-95 and recruits it to lipid rafts.
As a control, cells expressing wild-type PSD-95 were treated with cyclodextrin to deplete cholesterol levels prior to detergent extraction and density ultracentrifugation. This treatment eliminated PSD-95 immunoreactivity from the low density fractions, as would be expected for a protein localized to cholesterol-rich lipid rafts, and yielded a distribution pattern similar to that seen for C3,5S-PSD-95 (i.e. concentrated primarily in the high-density fractions) ( Fig. 2A). Cholesterol repletion of PSD-95-transfected cells previously treated with cyclodextrin restored the raft localization of PSD-95 ( Fig. 2A).
To demonstrate that cholesterol depletion and repletion were successful, mock-transfected cells whose cholesterol levels had been manipulated were fixed, and treated with filipin to detect free cellular cholesterol. Filipin is a fluorescent polyene antibiotic that complexes with sterols possessing a free 3␤-hydroxyl group, such as cholesterol (46). Fluorescence images made with identical exposures (Fig. 2B) Fig. 3, A and B). We used ECFP-Kv4.2 because it is expressed more robustly than non-tagged Kv4.2 (9,15). Similar to results obtained from the sucrose density gradients of rat brain membranes, the raft fractions (fractions 4 and 5; identified by GM 1 immunoreactivity in Fig. 2) contained a portion of ECFP-Kv1.4 and ECFP-Kv4.2 protein, while the non-raft fractions (fractions 8Ϫ12; identified by Na ϩ /K ϩ -ATPase and ␤-actin immunoreactivity in Fig. 2) contained most of these proteins. The presence of either Kv channel in the low density fractions was independent of PSD-95 co-expression (middle blots). Furthermore, Kv channel localization to the lipid raft fraction was not due to the addition of the ECFP tag, since the distribution of this protein was restricted to the high-density, non-raft fractions (Fig. 3C). We verified that ECFP-Kv1.4 and ECFP-Kv4.2 in the low density fractions were located in lipid rafts by depleting cholesterol with cyclodextrin before detergent extraction. As expected, this treatment eliminated the immunoreactivity for both Kv channels in the low density fractions (Fig. 3, A and B). The raft localization of both channels was restored with cholesterol repletion (Fig. 3, A and B).
Our finding that heterologously expressed ECFP-Kv4.2 is partially distributed to lipid rafts is consistent with the raft association of native Kv4.2 in rat brain membranes (Fig. 1). However, our results differ from a previous report by Martens et al. (29) that failed to detect raft association for Kv4.2 in transfected mouse L-cells. Although the detergent:protein ratio was not explicitly calculated in that report, it was likely much lower than in our study (ϳ2:1), since it appears that Martens et al. used more cells for each sucrose density gradient. It is therefore puzzling that they were unable to detect raft association for heterologously expressed Kv4.2. However, a plausible explanation is that rafts isolated from different cell lines exhibit distinct lipid and protein profiles, as has been recently demonstrated (47). This could lead to a higher affinity of Kv4.2 for lipid rafts in tsA201 cells compared with L-cells.
To determine if PSD-95 co-expression affected the level of ECFP-Kv1.4 or ECFP-Kv4.2 in lipid rafts, densitometry was used to quantify the intensities of the bands in lipid rafts (fractions 4 and 5) and in the lysates. The ratio of band intensities in raft fractions to those in lysates was analyzed after arcsine transformation with the Student's t test. Co-expression with PSD-95 increased the amount in the low density gradient fractions by 3-fold for ECFP-Kv1.4 (p Ͻ 0.05), but had no effect on the level of raft-associated ECFP-Kv4.2 (p Ͼ 0.5) (Fig. 3D). This was unexpected because binding to PDZ domain-containing proteins appears to not be necessary for the localization of Kv1.5 to rafts (48).

The Increased Localization of ECFP-Kv1.4 to Rafts Requires Its PDZ-binding Motif and Palmitoylation of PSD-95-Having
demonstrated that PSD-95 increases the raft localization of ECFP-Kv1.4, we were interested in determining the motifs required for this process. To test the requirement of PSD-95 palmitoylation, tsA201 cells were co-transfected with fulllength ECFP-Kv1.4 and the palmitoylation-deficient PSD-95 mutant. To test the requirement of a PDZ-binding motif in Kv1.4, cells were co-transfected with wild-type PSD-95 and a deletion mutant of ECFP-Kv1.4 lacking the C-terminal ETDV motif (ECFP-Kv1.4⌬ETDV). Sucrose density gradients were prepared from detergent-extracted cells (three independently transfected cell batches) as indicated above. Some full-length ECFP-Kv1.4 distributed to the low density raft fractions when C3,5S-PSD-95 was co-expressed (Fig. 4), as it did when expressed alone (Fig. 3A). Densitometry indicated that the amounts of ECFP-Kv1.4 in lipid rafts did not increase in the presence of C3,5S-PSD-95 (p Ͼ 0.5). Thus, palmitoylation of PSD-95 is needed for its ability to recruit ECFP-Kv1.4 to raft FIG. 2. Palmitoylation is required for the lipid raft localization of PSD-95. A, sucrose density gradients were prepared from tsA201 cells transfected with wild-type PSD-95, a palmitoylation-deficient mutant PSD-95 (C3,5S-PSD-95) or salmon sperm DNA (mock-transfected; probed with reagents or antibodies to GM 1 , Na ϩ /K ϩ -ATPase, and ␤-actin). Wild-type PSD-95 is in both the low density, lipid raft fractions (fractions 4 and 5, as indicated by the presence of GM 1 ) and in the high density, non-raft fractions (fractions 8Ϫ12, as indicated by the presence of Na ϩ /K ϩ -ATPase and ␤-actin). In contrast, C3,5S-PSD-95 is restricted to the high-density fractions. Immunoreactivity in the low density fractions was eliminated after cholesterol depletion with cyclodextrin (PSD-95Ϫchol.) and restored with cholesterol repletion (PSD-95 ϩchol.). B, mock-transfected cells were subjected to cholesterol depletion (cyclodextrin treatment; middle panel), cholesterol repletion after depletion (cyclodextrin treatment, then addition of cholesterol and cyclodextrin; right) or no treatment (serum-free Dulbecco's modified Eagle's medium alone; left). Cells were then stained with filipin to assess whether the treatments had the expected effect on cholesterol levels. Cells subjected to cholesterol depletion exhibited much lower cholesterol levels, as indicated by filipin fluorescence, than untreated cells. Cholesterol repletion restored filipin fluorescence to levels displayed by untreated cells. Scale bar ϭ 100 m and applies to all panels. microdomains in tsA201 cells. Deleting the C-terminal ETDV motif from ECFP-Kv1.4 also prevented wild-type PSD-95 from increasing the amount of channel protein in raft fractions (Fig.  4). In fact, when ECFP-Kv1.4⌬ETDV was co-expressed with PSD-95, levels of raft-associated channel protein were dramatically lower than when full-length ECFP-Kv1.4 was expressed alone. It was not possible to quantify the reduction because the range of band intensities in the raft fractions and lysates exceeded the linear range of the film.
Deletion of the PDZ-binding Motif Reduces or Nearly Eliminates Channel Raft Association-We next determined whether the PDZ-binding motifs of Kv1.4 and Kv4.2 are needed for channel localization to lipid rafts when PSD-95 is not co-transfected. tsA201 cells were transfected with ECFP-Kv1.4, ECFP-Kv1.4⌬ETDV, ECFP-Kv4.2 or ECFP-Kv4.2⌬VSAL and sucrose density gradients were prepared from detergent-extracted cells in parallel. ECFP-Kv1.4⌬ETDV was almost completely confined to the high density fractions when expressed alone (Fig.  5A), since only very faint bands in the low density fractions were visible, and only after overnight exposure (data not shown). This result contrasts with the distribution of full-length ECFP-Kv1.4 into both low and high density fractions (Fig. 5A), and suggests that an endogenous PDZ domain protein normally brings Kv1.4 constructs into lipid rafts in tsA201 cells. Kv4.2 may be recruited to rafts by an endogenous PDZ domain protein since levels of ECFP-Kv4.2⌬VSAL in raft fractions appeared to be greatly reduced compared with full-length ECFP-Kv4.2 (Fig. 5B). The reduction for the Kv4.2 deletion mutant was not as dramatic as for the Kv1.4 deletion mutant, since standard exposure conditions usually yielded more obvious bands in the lipid raft fractions for ECFP-Kv4.2⌬VSAL than for ECFP-Kv1.4⌬ETDV. In addition, overnight exposure (data not shown) revealed readily detectable bands for ECFP-Kv4.2⌬VSAL in the low density fractions, in contrast to ECFPKv1.4⌬ETDV (see above). The extremely large differences in band intensities in low density fractions compared with the lysate precluded accurate quantification of the magnitude of this reduction. Nevertheless, this provides an alternate basis for the lack of raft association of Kv4.2 in L-cells (29), since these cells may not express the PDZ domain protein endogenous to tsA201 cells. Conversely, Kv channels can interact with a wide array of proteins, and Kv4.2 in particular can FIG. 3. Kv1.4 and Kv4.2 are found in lipid rafts; PSD-95 recruits Kv1.4 to rafts. A, sucrose density gradients were prepared from tsA201 cells expressing ECFP-Kv1.4 under the following conditions: channel alone (alone), with PSD-95 (ϩPSD-95), after acute cholesterol depletion before detergent extraction (Ϫcholesterol), or after acute cholesterol depletion followed by repletion before detergent extraction (ϩcholesterol). Western blots were probed with anti-Kv1.4 antibody. B, same treatment regime as in part A, but using tsA201 cells expressing ECFP-Kv4.2. Western blots were probed with anti-Kv4.2 antibody. There is some ECFP-Kv1.4 (A) and ECFP-Kv4.2 (B) protein in the lipid raft fractions (fractions 4 and 5) but most of the protein is in the non-raft fractions for both channels (fractions 8Ϫ12). ECFP-Kv1.4 and ECFP-Kv4.2 distributions were independent of PSD-95 co-expression, but both channels were eliminated from the lipid raft fractions by depleting cholesterol. Cholesterol repletion restored raft association. C, sucrose density gradients were prepared from tsA201 cells expressing ECFP alone. Western blots probed with anti-GFP antibody show that immunoreactivity is limited to the high density fractions. D, densitometry was performed on the immunoblots in parts A and B, and from two other independent experiments. Results are expressed as the ratio of intensities of the bands in the lipid raft fractions to those in the lysates. PSD-95 co-expression increased the amount of ECFP-Kv1.4 in the lipid raft fraction (p Ͻ 0.05), but had no effect on ECFP-Kv4.2 (p Ͼ 0.5).
The endogenous PDZ domain protein is not PSD-95, since immunoblots of sucrose density gradients prepared from mocktransfected tsA201 were negative for PSD-95 (data not shown), which is consistent with previous results (17). However, tsA201 cells express hDlg (54,55), a PDZ domain protein that is the human homolog of Drosophila Dlg (56) and rat SAP97 (57). Because hDlg is highly homologous to PSD-95, and because SAP97 can bind to Kv1.4 (58), we assessed whether it could be the endogenous protein in tsA201 cells responsible for the raft association of ECFP-Kv1.4. Immunoblots of sucrose density gradients prepared from mock-transfected cells detected hDlg only in the high-density, non-raft fractions (Fig. 5C), indicating that it does not bring ECFP-Kv1.4 into rafts. The lack of hDlg/ SAP97 association with rafts may not be surprising since these proteins interact strongly with the cortical cytoskeleton in epithelial cells (59 -61).
Kv4.2 Co-localizes with Thy-1, a Raft Marker, in Postnatal Rat Hippocampal Neurons-Our results showing that Kv4.2 partially distributes into low density raft fractions in both rat brain membranes and transfected cells conflicts with an earlier study that reported no raft association for this Kv channel (29). Accordingly, we wished to confirm our findings by co-localizing Kv4.2 with Thy-1, a raft marker in neurons. While sizes of lipid rafts have not been conclusively determined, two estimates place them well below the resolution of light microscopy (62,63). Techniques to create "patches" of lipid rafts using multivalent toxins and/or antibodies can make them large enough to visualize by standard immunofluorescence microscopy (25, 64 -66), and are frequently used to complement and support biochemical studies. Lipid raft patching has allowed immunofluorescent detection of co-localized proteins and raft markers, including PKC-in T cells (67), MHC class II molecules in B cells (68,69), Syk tyrosine kinase in mast cells (70), and GAP-43 in rat hippocampal neurons (25). We patched rafts in the membranes of live rat hippocampal neurons using a monoclonal Thy-1 antibody that recognizes an external epitope and produced a single band when immunoblotting rat brain membranes (Fig. 1) and a fluorescent secondary antibody. The neurons were then fixed and permeabilized with methanol, rather than Triton X-100, to prevent raft solubilization, and subsequently processed for Kv4.2 immunolabeling using standard techniques. The antibody patching procedure caused Thy-1 to redistribute into discrete clusters on the membrane (Fig. 6, red) similar to those seen in an earlier report (25). Kv4.2 also displayed a punctate distribution (Fig. 6, green), which partially overlapped with Thy-1 (Fig. 6, inset, yellow; representative of 6 images). Most of the Kv4.2 puncta did not co-localize with Thy-1, which is consistent with the results obtained from the sucrose density gradients (Figs. 1 and 3B). Localization of Thy-1 and Kv4.2 in the same neuronal compartment in postnatal hippocampal neurons was contrary to some studies of embryonic hippocampal neurons, where Thy-1 displayed an axonal distribution (71,72). However, another study of embryonic rat hippocampal neurons (25) found a strikingly similar pattern of Thy-1 staining to that shown here. Thus, the complexities underlying Thy-1 localization in neurons have not been fully resolved, and may be affected by the development stage of the rats from which the neurons were isolated (i.e. postnatal versus embryonic), culturing method (compare Ref. 24 with 73) or time in culture. Imaging Kv1.4 in Thy-1-labeled lipid rafts was not possible since Thy-1 was primarily dendritic in these postnatal hippocampal neurons (discussed above), while, as expected, Kv1.4 displayed an axonal distribution (data not shown) (74 -76). DISCUSSION Using well established biochemical techniques to isolate lipid rafts, we demonstrated that a portion of Kv1.4, Kv4.2, and PSD-95 proteins localize to these membrane microdomains in both neurons and a heterologous expression system. Raft fractions were clearly marked by Thy-1 in brain and GM 1 in tsA201 cells, while non-raft fractions were identified by by Na ϩ /K ϩ -ATPase and/or ␤-actin. Acute cholesterol depletion of transfected cells prior to detergent extraction abolished localization to lipid rafts and cholesterol repletion restored raft association. Palmitoylation of the N terminus of PSD-95 was required for its distribution into raft regions. Kv1.4 and Kv4.2 both partitioned into raft fractions without PSD-95 co-expression; however, PSD-95 increased the level of Kv1.4, but not Kv4.2, in raft fractions. Recruitment of Kv1.4 to rafts required palmitoylation of PSD-95 and the PDZ-binding motif at the C terminus of Kv1.4. Removing the PDZ-binding motif of either channel decreased its raft association, more markedly for Kv1.4. Finally, we employed an extensively used antibody-patching procedure that increases the size of rafts and confirmed that Kv4.2 partially distributes into Thy-1-containing lipid rafts in cultured rat hippocampal neurons.
One possible explanation for the low amount of Kv1.4, Kv4.2, and PSD-95 proteins in lipid rafts in both native and heterologous systems is that the association of these proteins with rafts may be transient or unstable. Since the major biochemical method for isolating lipid rafts uses detergent extraction and overnight ultracentrifugation through a density gradient, it captures a static profile of raft-associated proteins. Thus, it may not detect associations if they are dynamic, or if easily disrupted protein-protein or protein-lipid interactions are required for association. Moreover, several surface receptors are normally excluded from lipid rafts to prevent the activation of signaling pathways in the absence of antigen or ligand binding. They include the T cell receptor (77), B cell receptor (78), IgE Fc⑀RI receptor in mast cells (79), insulin receptor in liverderived cells (80) and neuregulin ErbB4 receptor in neurons (40). Their recruitment to or enrichment in lipid rafts is triggered by receptor dimerization following antigen binding (immune cell receptors) or ligand binding (insulin and neuregulin receptors). If lipid rafts were isolated from these cell types in the absence of antigen or ligand binding, very low levels of receptor protein would be detected in raft fractions. Thus, another possibility is that the levels of Kv channels and PSD-95 in rafts might increase in response to an unknown stimulus. Alternatively, Kv channels may be located in non-raft regions to physically segregate them from raft-localized enzymes that modify their activity. By locating only a small proportion of Kv channels in rafts at a given time, subtle changes in current by channel modification would be possible.
The trafficking of Kv channels to lipid rafts has not been extensively characterized. Of the Shaker (Kv1) type channels, Kv1.5 is known to be significantly associated with caveolin (48), which is both a major structural protein and a distinguishing feature of caveolae, a specialized subpopulation of lipid rafts. Different mechanisms might bring different Kv channels into lipid rafts or even into distinct raft subtypes (i.e. caveolar versus non-caveolar). Like Kv1.4 in the present study, Kv1.5 distributed into both raft and non-raft regions; however, truncation of the last 57 C-terminal amino acids of Kv1.5 had no effect on raft association (36). One possibility is that Kv1.5 is brought into caveolae through its interaction with Src tyrosine kinase (81), which can bind to caveolin (82). Since interaction between Kv1.5 and Src requires a proline-rich domain in the N terminus of Kv1.5 (that is not present in Kv1.4) (81), it should not be affected by deletions in the C terminus.
Our results also imply a role for an endogenous PDZ domain protein in tsA201 cells that brings transfected Kv1.4 to lipid rafts, since Kv1.4 was present in the raft fraction even in the absence of co-expressed PSD-95. Several earlier observations on Kv1.4 are intriguing, even though they are not proven to reflect raft localization. Many raft-associated proteins and lipids are less mobile than their counterparts outside rafts (63,83,84). Kv1.4 displayed low mobility when channel clustering was induced by PSD-95 in COS-7 cells (85), suggesting that the Kv1.4 clusters might be in lipid rafts. In agreement with our results, there is evidence for interaction between Kv1.4 and an endogenous PDZ domain protein in HEK cells (from which tsA201 cells were derived). In the absence of PSD-95 co-expression, wild-type Kv1.4 channels displayed less lateral mobility than Kv1.4 mutants with an altered or missing PDZ binding motif (85). Such an unidentified PDZ domain protein may be one factor underlying normal recruitment of channel proteins to lipid rafts. Removing the PDZ domain-binding motif reduced raft association of Kv1.4 much more than Kv4.2. Thus, it appears that an endogenous PDZ domain protein helps bring these Kv channels into lipid rafts in tsA201 cells, but is more important for Kv1.4 than Kv4.2. PSD-95 increased the level of Kv1.4, but not Kv4.2 in rafts. One possible explanation is that PSD-95 and Kv1.4 might be targeted to a different subpopulation of lipid rafts from Kv4.2 in tsA201 cells. Distinct populations of lipid rafts within the same cell (or cell type) have been identified by differential detergent solubility (86 -88) and by differential lipid (89,90) and/or protein composition (48, 86 -88, 91). However, this result could also arise from different binding affinities between PDZ domain proteins and these Kv channels. First, PSD-95 may bind more avidly to Kv1.4. Indeed, in coimmunoprecipitations from lysates of transfected tsA201 cells, PSD-95 consistently pulled down more Kv1.4 than Kv4.2 (data not shown). Second, the interaction of Kv4.2 with the endogenous PDZ domain protein may be stronger than its interaction with PSD-95. Third, the interaction of Kv1.4 with PSD-95 may be stronger than its interaction with the endogenous protein.
Collectively, these results suggest that an endogenous PDZ domain protein in tsA201 cells influences the raft localization of heterologously expressed Kv1.4 and Kv4.2 to differing degrees. While the identity of the endogenous PDZ domain protein in tsA201 cells is not known, previous results rule out PSD-95, since it is not expressed by tsA201 cells (17), and our results now rule out hDlg, since it was not significantly associated with lipid rafts. In the future, a proteomics approach will undoubtedly prove useful and it will be intriguing to determine if this protein recruits its native binding partners to lipid rafts, since this would be a novel function for PDZ domain proteins.
Because lipid rafts are enriched in specific components of signaling pathways, notably protein kinases, we speculate that Kv1.4, Kv4.2, and their modulators may be co-localized in lipid rafts. For instance, both channels are regulated by PKC, several isoforms of which have been found in lipid rafts (92). Kv4.2 is modified by ERK2 (6), which is enriched in lipid rafts (35) and becomes hyperactive under cholesterol-depleting conditions (20). Thus, targeting a portion of Kv1.4 and Kv4.2 to lipid rafts might allow the cell to harbor differentially phosphorylated populations of specific Kv channels and traffic them to specific subcellular locations.
By increasing the levels of raft-associated Kv1.4, PSD-95 might facilitate the interaction of this channel with signal transduction cascades localized to lipid rafts. In principle, the role of PSD-95 in localizing Kv channels or other binding partners to lipid rafts could be assessed by eliminating PSD-95 expression through treatments such as antisense or RNA interference. However, interpreting data from knockdown experiments would be complicated by the activity-dependent, dynamic cycling of palmitates on PSD-95 (93). According to our results, depalmitoylation should dissociate PSD-95 from lipid rafts, disrupt the interaction between Kv1.4 and PSD-95, and therefore sever the links between Kv1.4 and signaling molecules that are directly connected to PSD-95, such as PKA (94) and Src family tyrosine kinases (95). Alternatively, the detachment of PSD-95 from raft microdomains might also move interacting Kv1.4 channels away from rafts. Since depalmitoylated PSD-95 proteins would encounter a different set of signaling molecules outside raft microdomains, this might serve as a mechanism to couple PSD-95 ligands to different signal transduction cascades. A similar model has been proposed for AMPA receptors (96).
Collectively, our results show that two Kv channels that are important for functions of excitable cells are present in lipid rafts and demonstrate a novel role for PSD-95 in recruiting Kv channels to rafts. Future experiments will be needed to reveal the complexity and significance of raft associations for Kv channels.