Differential Targeting of Shaker-like Potassium Channels to Lipid Rafts*

Ion channel targeting within neuronal and muscle membranes is an important determinant of electrical excitability. Recent evidence suggests that there exists within the membrane specialized microdomains com-monly referred to as lipid rafts. These domains are en-riched in cholesterol and sphingolipids and concentrate a number of signal transduction proteins such as nitric-oxide synthase, ligand-gated receptors, and multiple protein kinases. Here, we demonstrate that the voltage-gated K 1 channel Kv2.1, but not Kv4.2, targets to lipid rafts in both heterologous expression systems and rat brain. The Kv2.1 association with lipid rafts does not appear to involve caveolin. Depletion of cellular cholesterol alters the buoyancy of the Kv2.1 associated rafts and shifts the midpoint of Kv2.1 inactivation by nearly 40 mV without affecting peak current density or channel activation. The differential targeting of Kv channels to lipid rafts represents a novel mechanism both for the subcellular sorting of K 1 channels to regions of the membrane rich in signaling complexes and for modulat-ing channel properties via alterations in lipid content. The subcellular localization of ion channels is necessary for proper electrical signaling. In cardiac and skeletal myocytes, ion channels show a differential surface distribution (1, 2). Within the brain, voltage-gated K 1 (Kv) channels often show not only polarized sorting to either axons or dendrites, but also isoform-specific localization within dendrites alone. Thus, there exists specific sorting mechanisms for restricting lateral distribution within a given membrane domain (3). One physi-ological consequence for such specific nals were collected using a Nikon E800 microscope equipped with standard epifluorescence and a Princeton Instruments CCD camera. Electrophysiology— Electrophysiological recordings and data analy- sis were made as described previously using the whole-cell configura-tion of the patch clamp technique (14). The steady-state inactivation was measured using a 10 s. conditioning pulse to various potentials followed by 500-ms test steps to 1 40 mV. Kv2.1 exhibits complex inactivation kinetics including a U-shaped voltage dependence and excessive cumulative inactivation (17). Therefore, we measured only the voltage dependence of inactivation over the range of potentials ( 2 100 to 1 10 mV), which show maximal inactivation. Additional details are presented in the figure legends.

Ion channel targeting within neuronal and muscle membranes is an important determinant of electrical excitability. Recent evidence suggests that there exists within the membrane specialized microdomains commonly referred to as lipid rafts. These domains are enriched in cholesterol and sphingolipids and concentrate a number of signal transduction proteins such as nitricoxide synthase, ligand-gated receptors, and multiple protein kinases. Here, we demonstrate that the voltagegated K ؉ channel Kv2.1, but not Kv4.2, targets to lipid rafts in both heterologous expression systems and rat brain. The Kv2.1 association with lipid rafts does not appear to involve caveolin. Depletion of cellular cholesterol alters the buoyancy of the Kv2.1 associated rafts and shifts the midpoint of Kv2.1 inactivation by nearly 40 mV without affecting peak current density or channel activation. The differential targeting of Kv channels to lipid rafts represents a novel mechanism both for the subcellular sorting of K ؉ channels to regions of the membrane rich in signaling complexes and for modulating channel properties via alterations in lipid content.
The subcellular localization of ion channels is necessary for proper electrical signaling. In cardiac and skeletal myocytes, ion channels show a differential surface distribution (1,2). Within the brain, voltage-gated K ϩ (Kv) channels often show not only polarized sorting to either axons or dendrites, but also isoform-specific localization within dendrites alone. Thus, there exists specific sorting mechanisms for restricting lateral distribution within a given membrane domain (3). One physiological consequence for such specific localization is that it places various signal transduction molecules near their ion channel substrates (4). Several families of intracellular proteins, PDZs and AKAPs, have been shown to cluster both ion channels and modulatory signaling enzymes. Indeed, great emphasis has been placed on the role of PDZ proteins such as PSD-95 in the targeting and localization of ion channels and neurotransmitter receptors (5). In contrast, the role of membrane lipids in differential targeting and integration of ion channels within the plane of the plasma membrane has not been addressed.
Recent advances in the study of cell membrane structure have led to the emerging idea that microdomains exist within the fluid bilayer of the plasma membrane. These dynamic structures, termed lipid rafts, are rich in tightly packed sphingolipids and cholesterol (6). The rafts, which are present in both excitable and non-excitable cells, localize a number of membrane proteins, including multiple signal transduction molecules, while excluding others (7). Different types of rafts are likely to exist based on the presence of specific marker proteins and ultrastructure data (8). Caveolae represent one well studied subpopulation of lipid raft having an invaginated morphology and containing the scaffolding protein, caveolin, which interacts directly with several intracellular proteins (7, 9 -12). Here, we demonstrate that Kv2.1 K ϩ channels target to a non-caveolar lipid raft in both transfected cells and brain, whereas the Kv4.2 channel does not. In addition, cholesterol depletion dramatically alters Kv2.1 inactivation, while having no effect on Kv4.2. Thus, lipid raft association represents a new targeting mechanism for Kv channel localization that is based on protein-lipid interactions.
Raft Isolation-Low density, Triton-insoluble complexes were isolated as described by Lisanti and co-workers (13) from mouse L-cells stably expressing either rat Kv2.1 or Kv4.2 channels (14,15). Briefly, cells from 10 100-mm near confluent culture dishes were homogenized in 1 ml of 1% Triton X-100 and sucrose added to a final concentration of 40%. A 5-30% linear sucrose gradient was layered on top of this detergent extract followed by ultracentrifugation (39,000 rpm) for 18 -20 h at 4°C in a Beckman SW41 rotor. Gradient fractions (600 l) were collected from the top and analyzed by Western blot. Rat brain raft isolation used one brain (approximately 0.6 g) homogenized in 10 ml Mes 1 -buffered saline (25 mmol/liter Mes, pH 6.5, 0.15 mol/liter NaCl) containing 1% Triton X-100, 1 mmol/liter phenylmethylsulfonyl fluoride, 2 g/ml pepstatin, and leupeptin using 18 strokes of a loose fitting dounce homogenizer. This homogenate was then centrifuged at 3000 rpm, 4°C for 15 min to sediment debris. The supernatant was processed as described above (13). The detergent-free raft isolations were performed as described previously (16). Basically, this protocol introduces a sonication step to disrupt the cellular membranes followed by subcellular fraction on a discontinuous 5-35% sucrose gradient in a buffer containing sodium carbonate, pH 11.0. Triton solubility experiments were performed on cells homogenized in Mes-buffered saline containing 1% Triton X-100 with or without a 0.5% saponin pretreatment for 30 min at 4°C. The detergent-soluble (lysate) and -insoluble fractions (pellet) were separated by centrifugation (14,000 rpm, for 30 min at 4°C) and analyzed by immunoblotting.
Immunostaining-Immunostaining of cells was performed using anti-Kv2.1 polyclonal antibodies and/or anti-caveolin antibodies as described previously (1). Demecolcine (0.1 g/ml; Sigma) was dissolved in the culture medium. The bound anti-caveolin antibody was detected with BODIPY-conjugated streptavidin (green), and the Kv2.1 antibody was detected with CY3-conjugated streptavidin (red). Fluorescent sig-nals were collected using a Nikon E800 microscope equipped with standard epifluorescence and a Princeton Instruments CCD camera.
Electrophysiology-Electrophysiological recordings and data analysis were made as described previously using the whole-cell configuration of the patch clamp technique (14). The steady-state inactivation was measured using a 10 s. conditioning pulse to various potentials followed by 500-ms test steps to ϩ40 mV. Kv2.1 exhibits complex inactivation kinetics including a U-shaped voltage dependence and excessive cumulative inactivation (17). Therefore, we measured only the voltage dependence of inactivation over the range of potentials (Ϫ100 to ϩ10 mV), which show maximal inactivation.
Additional details are presented in the figure legends.

RESULTS AND DISCUSSION
Kv2.1, but Not Kv4.2, Is Localized to Lipid Rafts in Transfected Fibroblasts-Kv2.1 is a delayed rectifier K ϩ current that inactivates very slowly over a period of seconds. In contrast, Kv4.2 encodes for a transient current that activates and inactivates within milliseconds. Kv2.1 and Kv4.2 stably transfected mouse LtkϪ cells (L-cells) showed high levels of cell surface expression as evidenced by the nanoampere current recordings (Fig. 1A). Kv2.1 is Triton X-100-insoluble when transfected in Madin-Darby canine kidney cells, suggesting that an interaction with the detergent-insoluble cytoskeleton may be necessary for proper Kv2.1 localization (3). As shown in Fig. 1B, Kv2.1 stably expressed in mouse L-cells was also insoluble in 1% Triton X-100. However, when these cells were pretreated with 0.5% saponin, a detergent that complexes with membrane cholesterol, Kv2.1 was nearly completely solubilized in Triton X-100, suggesting that cholesterol is required for the detergent resistance of Kv2.1. Similar results were obtained for the lipid raft marker protein, caveolin (Fig. 1B). Saponin alone did not solubilize either Kv2.1 or caveolin (data not shown). Next, we isolated low density, Triton X-100-insoluble complexes from cells stably expressing Kv2.1 and Kv4.2. Western blot analysis of sucrose gradient fractions probed with anti-Kv2.1 antibodies demonstrated that a majority of Kv2.1 floats in a low density, Triton-insoluble fraction together with caveolin. The small percentage of Kv2.1 in the non-raft fractions most likely represents overexpressed intracellular protein. The percentage of non-raft channel varied between clonal cell lines and correlated with the amount of intracellular protein as determined by immunostaining (data not shown). In contrast to Kv2.1, Kv4.2 was excluded from the low density fractions and found exclusively at the bottom of the gradient (Fig. 1C). Additionally, the endogenous Na ϩ /K ϩ ATPase, actin and tubulin, were all found in non-raft fractions (Fig. 1C). However, with very long exposure times, a small percentage of the immunoreactive actin appeared in the low density fractions. The differential association of Kv2.1 and Kv4.2 with the light membrane fractions was also found to be true in transfected human embryonic kidney cells (HEK), demonstrating that this is not limited to fibroblasts (data not shown). The localization of Kv2.1 and caveolin to lipid rafts in the transfected L-cells was also confirmed using a detergent-free method of raft isolation (Fig. 1D) (16). This technique uses sonication to disrupt the membrane followed by alkaline extraction and avoids the potential issue of a detergent-induced artifactual association of Kv2.1 and light membrane fractions.
Kv2.1 Is Localized in Non-caveolar Lipid Rafts-Dual immunostaining of L-cells expressing Kv2.1 with anti-channel and anti-caveolin antibodies revealed only partial overlap of the two proteins (Fig. 2). Although both caveolin (Fig. 2B) and Kv2.1 (Fig. 2C) display punctate cell surface staining, there is incomplete colocalization (Fig. 2, D and E), suggesting that Kv2.1 is in a lipid raft that does not contain caveolin. To better resolve this issue, a pharmacological intervention was used to disrupt the normal cell surface distribution of caveolae. Previous reports have suggested a role for microtubules in the maintenance and transport of caveolae with microtubule disruption resulting in caveolin internalization (18,19). L-cells stably expressing Kv2.1 were immunostained with anti-Kv2.1 and caveolin antibodies before (Fig. 3, B and F) and after (Fig. 3, D and H) treatment with demecolcine, to disrupt microtubule organization. In the treated, and thus rounded cells (compare Fig. 3, A and C and E and G), caveolin was retracted from the cell surface (compare Fig. 3, F and H), consistent with previous reports (19). However, demecolcine treatment did not alter the cell surface localization of Kv2.1 channel protein (Fig. 3D). Therefore, Kv2.1 channel protein does not follow caveolin re- Differential Targeting of Shaker-like Potassium Channels 7444 distribution after microtubule disruption. Consistent with these results, immunoprecipitation of caveolin failed to demonstrate an association between Kv2.1 and caveolin in either the L-cell system or rat heart, which expresses both Kv2.1 and caveolin (data not shown). Taken together, these data strongly suggest that Kv2.1 is associated with a non-caveolar lipid raft.
Kv2.1, but Not Kv4.2, Is Localized to Lipid Rafts in Rat Brain-The issue of Kv channel targeting to lipid rafts in vivo was addressed in the brain where the subcellular distribution and polarized sorting of voltage-gated K ϩ channels is critical for neural excitability. In addition, both Kv2.1 and Kv4.2 have previously been localized to neurons where they display a very different and restricted distribution within the lateral plane of the neuronal surface (20 -22). Sucrose density gradient fractionation of Triton X-100 extracted rat brain lysates revealed that nearly all of the Kv2.1 was found in low density membrane fractions (Fig. 4). In contrast, Kv4.2 was detected only in the high density fractions at the bottom of the gradient. Thus, differential targeting to lipid rafts is not an artifact of either the L-cell or HEK 293 expression systems. In agreement with an association with non-caveolar rafts, caveolin expression is lacking in neurons where Kv2.1 is concentrated (20,23).
Depletion of Membrane Cholesterol Alters Kv2.1-containing Raft Buoyancy and Channel Function-Given the unique organization of membrane sphingolipids and cholesterol within lipid rafts, these structures are sensitive to cholesterol modifying agents (7,24). Treatment of cells stably expressing Kv2.1 with 2% 2-hydroxypropyl-␤-cyclodextrin for 1 h altered the buoyancy of the Kv2.1 channel containing rafts (Fig. 5A), with the Kv2.1 protein shifted toward higher density fractions following treatment with the cholesterol-binding drug. These cholesterol-depleted cells were assayed by voltage clamp to assess Kv2.1 channel function. The current density and activation kinetics were not affected by cyclodextrin treatment (compare Fig. 5, B and C). Voltage sensitivity was also unaltered as shown in Fig. 5D. However, cyclodextrin significantly altered the steady-state inactivation of Kv2.1 as evidenced by a 36-mV hyperpolarizing shift in the inactivation curve (Fig. 5E). The V1 ⁄2 for inactivation of control and cyclodextrin-treated cells was Ϫ15.7 Ϯ 0.59 and Ϫ51.6 Ϯ 0.44, respectively. The drug effect was not due to a direct interaction with channel protein because acute application of cyclodextrin in the bath solution did Differential Targeting of Shaker-like Potassium Channels 7445 not affect channel function. Therefore, cyclodextrin treatment did not non-selectively modify channel gating or cell surface expression but rather specifically altered inactivation. Treatment of cells stably expressing Kv4.2 showed no observable effect (data not shown). These data show that altering raft structure significantly affects the function of raft-associated channels. Such a large shift in steady-state inactivation can lead to a dramatic shift in both resting potential and/or action potential duration. Possible Functions of the Lipid Raft-Channel Complex-Since rafts often localize signaling proteins such as protein kinase C, nitric-oxide synthase, tyrosine kinases, Ha-Ras, mitogen-activated protein kinase, glycosylphosphatidylinositolanchored proteins and G-proteins, channel/raft association could serve primarily to cluster signaling molecules with ion channels (9 -12). Kv channels are known to be modulated by activation of various signal transduction pathways and often contain multiple consensus phosphorylation sites (4). In fact, Kv2.1 is constitutively tyrosine phosphorylated and physically associates with tyrosine kinases in Schwann cells (25). Multiple reports have localized tyrosine kinases to lipid raft microdomains including those from neuronal plasma membranes (26). Differential targeting to various lipid raft subpopulations may serve to organize signaling molecules and their K ϩ channel substrates. It is possible that the functional effects of cyclodextrin treatment are a consequence of kinase disruption as opposed to direct effects of altered lipid on channel activity.
It is also tempting to hypothesize a possible role for lipidprotein rafts in the polarized sorting of K ϩ channels in the brain. Certainly, this appears to be a mechanism for the polarized sorting of other neuronal proteins (27)(28)(29). Although both channels target to dendrites, it is clear that even within the dendritic region multiple plasma membrane subdomains exist (29), for the Kv2.1 and Kv4.2 channels segregate to somatodendritic and distal regions, respectively (22).
Potential Mechanisms of Channel-Raft Association-One obvious question deals with the mechanism of channel targeting to rafts. The channel could bind raft-associated proteins or it could directly target to, or interact with, the raft lipid. The cytoplasmic COOH-terminal domain of Kv2.1 has been implicated in the polarized sorting and clustering of Kv2.1 in Madin-Darby canine kidney cells (3). However, preliminary results based on truncation mutants of Kv2.1 suggest that neither the amino or carboxyl terminus is necessary for targeting to lipid rafts (data not shown). This finding is consistent with reports that transmembrane regions of integral membrane proteins may contain the information that determine raft association (30).
An alternative mechanism for channel-raft association may involve the channel binding to other raft-associated proteins. One candidate is PSD-95, which has been reported to associate with low density lipid rafts in both COS cells and rat brain (31). In addition, protein-lipid interactions are necessary for clustering of this PDZ protein at the synapse (32). Therefore, raftbound PDZ proteins could localize ion channels to raft domains. However, Kv2.1 does not contain standard PDZ binding sequences nor does it interact with PSD-95 (3). Even if it did contain a PDZ binding site, the Kv2.1 truncation mentioned above still targets to rafts. In addition, glutamate receptors bind PDZ proteins but are not raft associated (26). Thus, PDZ proteins are not likely to be responsible for raft localization. Recent work from the Clapham laboratory (33) indicates that PSD-95, when expressed in slices of rat cortex, only targets to axons in the presence of Kv1.4 channel coexpression. Thus, the ion channel, or another protein which recognizes the PSD-95channel complex, is responsible for localizing PSD-95 to the axon. It is tempting to speculate that localization of channels to rafts is part of a mechanism by which PDZ proteins are themselves localized. Another candidate protein for targeting K ϩ channels to rafts is the K ϩ channel beta subunit. This protein has been implicated in channel association with the cytoskeleton. However, Kv2.1 does not associate with the Kvbeta 2.1 subunit present in the L-cell expression system (data not shown), making the ␤ subunit a poor candidate (34). Additional mutagenesis experiments are necessary to determine a potential raft association signal within the channel.
Conclusion-This report is the first description of ion channels localizing to lipid microdomains and provides a dramatic example of the differential targeting of protein isoforms to lipid rafts. Although progress has been made in identifying elements involved in channel targeting, clustering, and anchoring, it is not yet clear how the number and location of channel complexes within the plane of the membrane are determined (32).
Our data indicate that protein-lipid interactions should be considered as a mechanism of Kv channel localization. The finding that cyclodextrin treatment shifts steady-state inactivation in the hyperpolarizing direction by more than 30 mV suggests that alteration of membrane lipid, either by disease (35) or the clinical use of lipid-lowering drugs, can affect membrane excitability by altering the function of raft-localized channels.