Targeting of voltage-gated potassium channel isoforms to distinct cell surface microdomains

Ion channel regulation is an important determinant of electrical excitability in the nervous and cardiovascular systems, skeletal muscle, gastrointestinal tract and uterus. Voltage-gated potassium (Kv) channels play a key role in setting the resting membrane potential and shaping action potential repolarization in these functionally diverse systems. Kv channels comprise the largest and most diverse class of voltage-gated ion channels and most cells express multiple channel types belonging to one or more subfamilies. Structurally, Kv channel isoforms share a conserved core of six transmembrane domains, which includes the voltage sensor and the ion conducting pore. However, the intracellular N- and C-termini are divergent, even in Kv channel isoforms with nearly identical functional properties, giving rise to speculation that these domains regulate trafficking and localization in a manner unique to each channel. Consistent with these differences, some Kv1 family channels bind α-actinin at the N-terminus (Maruoka et al., 2000) and interact with PDZ proteins such as PSD-95 via binding domains within the C-terminus (Kim et al., 1995). However, such motifs are not present in all channels. For example, the C-terminus of Kv2.1 plays a role in the clustering of Kv2.1 in hippocampal neurons (Lim et al., 2000), despite the absence of any known interaction motifs, such as PDZ binding domains (Scannevin et al., 1996). A role for the N- and C-termini in channel trafficking has been shown for the differential trafficking of inward rectifier K (K_IR) channel isoforms (Bendahhou et al., 2003; Ma et al., 2002).

Kv channels often show a highly specific subcellular localization; for example, the A-type channel Kv1.4 is found only along axonal membranes (Sheng et al., 1992), whereas the delayed rectifier Kv2.1 is found only on the soma and proximal dendrites (Lim et al., 2000). Such compartmentalization of functionally distinct channels is significant, as it permits spatial regulation of electrical excitability as well as localizing signal transduction molecules near their ion channel substrates. We previously reported that Kv channels differentially target to distinct lipid raft microdomains within the plasma membrane (Martens et al., 2000; Martens et al., 2001), indicating that protein-lipid interactions must also be considered as a potential mechanism of Kv channel localization. Furthermore, many signal transduction pathways, including those known to modulate Kv channels, are often found in lipid rafts, including kinases, nitric oxide synthase, GPI-linked proteins and others (Martens et al., 2004; O'Connell et al., 2004).

From a clinical standpoint, mutations resulting in mistrafficked ion channels are responsible for human disease. For example, the most common mutation in cystic fibrosis, ΔF508 in the cystic fibrosis transmembrane conductance regulator (CFTR), results in a mistrafficked channel (Powell and Zeitlin, 2002). Mutations in the human ether-a-go-go related (HERG) K⁺ channel result in trafficking defects that are a common disease mechanism in the LQT2 form of Long QT syndrome (Delisle et al., 2004). Despite this clinical relevance, the mechanisms by which ion channels are trafficked remain poorly understood. Few studies have examined channel trafficking in live cells (see Burke et al., 1999), with most using fixed preparations or purely biochemical approaches. Thus, it is essential that we expand our basic understanding of Kv channel localization in live cells. To this end, we examined the trafficking of three Kv channel isoforms. Kv2.1, Kv1.4 and Kv1.3 exist in distinct membrane compartments that traffic via distinct mechanisms in HEK 293 cells and are differentially sensitive to the perturbation of raft lipid by depletion of membrane cholesterol. In addition, the behavior of Kv2.1 in these cells establishes HEK293 cells as a meaningful model system in which to explore the mechanisms underlying trafficking and cell surface localization of this delayed rectifier channel.

Standard PCR mutagenesis techniques were used to construct Kv2.1HA and Kv1.4myc. To make Kv2.1HA, the HA epitope sequence YPYDVPDYA was inserted in tandem after Gly217 of the rat Kv2.1 primary sequence, which places the epitope in the extracellular S1-S2 loop. To make Kv1.4myc, the myc epitope sequence EQKLISEEDL was inserted in tandem after Gly430 of the human Kv1.4 primary sequence, which places the epitope in the extracellular S3-S4 loop. For both constructs, glycines were inserted before and after the epitope sequence. Kv2.1HA, Kv1.4myc and Kv1.3 were also fused in frame to the C-terminus of fluorescent proteins using the pECFP, pEYFP or pEGFP-C1 vectors (Clontech, Palo Alto, CA) or PA-GFP-C1 (gift from J. Lippincott-Schwartz, NTH, Bethesda, MD).

For imaging and immunostaining, vectors containing fluorescently tagged channels were transfected into HEK293 cells (ATCC, Manassas, VA) using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) at 200 ng per 60 mm dish. For live cell imaging, HEK cells were grown in 60 mm culture dishes and, 24 hours post-transfection, were passed at a 1:4 dilution onto collagen coated glass bottom 35 mm dishes (MatTek, Ashland, MA). For some experiments, HEK cells were electroporated using a BioRad Genepulser Xcell (BioRad Laboratories, Hercules CA). Electroporation was done using a single pulse to 110 V for 25 milliseconds in a 0.2 cm gap cuvette. Cells were subsequently imaged within 48-72 hours following transfection or electroporation. All FRAP experiments were performed in phenol-red and serum-free media to decrease nonspecific fluorescence intensity.

Some FRAP studies were done using GFP-tagged channels on an Olympus FluoView1000 confocal microscope. A circular region of interest was photobleached in tornado scan mode with a 405 nm diode laser at 35% transmission for 1 second using the SIM scanner of the FV1000. The SIM scanner was synchronized with the main scanner during bleach and acquisition. Following bleach, imaging was performed by raster scanning with the 488 nm line of a 40 mW Ar laser at 1% transmission and the variable bandpass filter set at 505-530 nm emission. A 60×, 1.4 NA oil immersion objective was used with the pinhole set to 1 Airy Unit. Images were acquired every 10 seconds for 250 scans (41.6 minutes), except for some GFP-Kv1.4 FRAP experiments where images were acquired for 200 scans at the maximum frame rate for 512 ×512 resolution (1 frame every 1.1 seconds).

For photoactivation experiments, photoactivatable GFP-tagged channels were photoactivated by raster scanning a region of interest with a 405 nm laser at 10% transmission for 500 milliseconds using the SIM scanner. Cells were cotransfected with monomeric RFP to identify expressing cells. Following photoactivation, GFP fluorescence was imaged as described with images acquired every 10 seconds for 250 scans.

All offline image analysis was done using Zeiss LSM510 v3.2 software or Olympus FV1000 software and Sigmaplot 8.0. All data are presented as mean±s.e.m. unless otherwise indicated.

For cholesterol depletion, cells were washed once with serum-free DMEM, then incubated in 2% 2-hydroxypropyl-β-cyclodextrin (Sigma, St Louis, MO) in serum-free DMEM for 1.5 hours at 37°C (Martens et al., 2000). Following cyclodextrin treatment, cells were washed once with DMEM containing 10% FBS and imaged within 2 hours. Imaging was performed in phenol red- and serum-free media. Cluster sizes in control and cholesterol-depleted cells were measured by randomly selecting six clusters from the basal optical section and measuring the area of each cluster. No change in Kv2.1 distribution was seen in cells incubated in serum-free media with no cyclodextrin (data not shown).

The detection of cell surface YFP-Kv2.1HA and YFP-Kv1.4myc was performed by incubating live cells with either anti-HA monoclonal (1:500, Sigma) (St Louis, MO) or anti-myc monoclonal (1:500, Upstate, Waltham, MA) antibodies diluted in MEM + 10% horse serum for 1 hour at room temperature. Following antibody incubation, cells were washed twice with MEM+10% HS and then fixed with 4% paraformaldehyde. Alexa 594-conjugated goat anti-mouse secondary (Molecular Probes, Eugene, OR) was used for secondary detection. Fluorescence signals were collected using a Zeiss LSM 510 Meta confocal microscope. YFP fluorescence was imaged using 514 nm excitation and a LP530 emission filter. Alexa594 was imaged using 543 nm excitation and BP 585-615 emission filters. The YFP and Alexa594 signals were acquired using multitrack mode and no crosstalk between YFP and Alexa594 was observed. Three-dimensional imaging of cells was done by optically sectioning the cells with the pinhole optimized at 1 Airy unit on each channel. Zeiss LSM510 v.3.2 software was used to reconstruct and two-dimensional deconvolve the images.

Two 100 mm dishes containing HEK cells at 50% confluency were transfected with channel expressing vectors using Lipofectamine 2000 as described above and incubated for 18-24 hours. All subsequent steps were performed at 4°C. The media were replaced with 1.5 ml of PBS containing 20 mg/ml pefabloc and the cells were then scraped from the dish with a rubber policeman. Following centrifugation of the cell suspension at 1000 g the cells were resuspended in 0.1 ml of PBS/pefabloc and diluted 1:2 with 2% Triton X-100 and 60 mM octylglucoside in PBS. The detergent extract was then homogenized with a glass-glass dounce and incubated for 30 minutes at 4°C. Three hundred and fifty microliters of this detergent extract was then layered on top of a 4.6 ml 5-20% linear sucrose gradient (containing 150 mM NaCl, 10 mM HEPES, pH 7.4, 1 mM CaCl₂, 10 mM MgCl₂ and 0.1% Triton X-100) poured in a SW 50.1Ti Beckman rotor tube. The gradient was centrifuged for 2 hours at 180,000 g. Four hundred microliter fractions were collected starting from the top of the gradient and aliquots analyzed by western analysis using an anti-GFP monoclonal antibody (Chemicon, Temecula, CA) as previously described (Martens et al., 2001). The amount of channel present in each fraction was quantitated using a BioChemi CCD camera system from UVP (Upland, CA).

The studies undertaken in our present work were initiated to understand the differential dynamics of Kv channel isoform trafficking in live cells and thus lay the foundation for a detailed study of the cell biology of Kv channel intracellular trafficking. Because we are interested in the dynamic trafficking of these channels, we added fluorescent tags (CFP, YFP or GFP) to the N-terminus of each channel to permit visualization in living cells. Additionally, epitope tags were inserted into extracellular loops of Kv2.1 (HA) and Kv1.4 (myc) to permit detection of those channels present on the surface membrane. Both YFP-Kv2.1HA and YFP-Kv1.4myc generate nanoampere currents when expressed in HEK cells and display a subcellular distribution identical to their untagged counterparts as detected with anti-channel antibodies (data not shown), confirming that the addition of the N-terminal and loop tags does not interfere with channel biology. YFP-Kv1.4myc does display less inactivation than the non-YFP-tagged channel, most probably because the N-terminal YFP alters ball and chain inactivation (Burke et al., 1999).

To examine subcellular expression patterns, plasmid DNAs encoding YFP-tagged Kv2.1, Kv1.4 and Kv1.3 were transfected into HEK293 cells and examined by confocal fluorescence microscopy as shown in Fig. 1. Kv2.1 and Kv1.4 isoforms were readily detected on the cell surface by the immunostaining of live cells with either anti-HA (Fig. 1A) or anti-myc (Fig. 1B) antibodies. Comparison of this surface signal to YFP fluorescence indicated that both isoforms traffic efficiently to the surface, with only a small fraction of the channel in an intracellular compartment. Kv1.3, another member of the Kv1 family, is closely related to Kv1.4, so for comparison, we expressed YFP-tagged Kv1.3 in HEK cells, where it exhibited a distribution similar to that of YFP-Kv1.4myc. However, Kv1.3 tended to accumulate in an intracellular compartment to a greater extent than either YFP-Kv2.1HA or YFP-Kv1.4myc (Fig. 1C).

Kv2.1 and Kv1.4 do not form heteromeric channels when coexpressed, and given their distinct cell surface distributions, it seemed unlikely that the two channels would reside in similar plasma membrane compartments. However, both Kv2.1 and Kv1.4 are present in noncaveolar lipid raft domains (Martens et al., 2000; Wong and Schlichter, 2004). To address the possibility that Kv2.1 and Kv1.4 could reside in the same raft compartment, we coexpressed CFP-tagged Kv2.1HA and YFP-tagged Kv1.4myc in HEK cells. Coexpression of CFP-Kv2.1HA and YFP-Kv1.4myc did not alter the subcellular localization of either channel; CFP-Kv2.1HA still exhibited a punctate distribution (Fig. 2A), while YFP-Kv1.4myc remained evenly distributed over the entire cell surface (Fig. 2B). Moreover, despite the apparent overlap between Kv2.1 and Kv1.4, shown in Fig. 2C, detailed pixel-by-pixel analysis (Fig. 2D) suggests that they are unlikely to truly colocalize. Kv2.1 and Kv1.4 must reside in distinct raft domains.

In neurons, Kv2.1 displays a punctate expression pattern and tends to preferentially accumulate on the bottom surface of the cell body (Lim et al., 2000). This expression pattern was reproduced well in the HEK cells (Fig. 1A and Fig. 3A). Fig. 3A (top) shows an optical section of the basal surface of a representative YFP-Kv2.1HA-expressing cell, clearly illustrating the clustered distribution of Kv2.1 in HEK cells. An x-z image of the same cell is presented in Fig. 3A (bottom), illustrating the preferential distribution of Kv2.1 on the basal cell membrane. By comparison, YFP-Kv1.4myc is evenly distributed over the entire bottom surface of the cell, and shows no apparent polarization between top and bottom (Fig. 3B).

Because Kv2.1 formed punctate surface structures that may be indicative of channel anchored to the cytoskeleton, we investigated the mobility of the channel on the plasma membrane using two complementary approaches: photobleach (FRAP) and photoactivation. For photobleach experiments, a small region of surface membrane of a YFP-Kv2.1HA-expressing cell was bleached using the 514 nm laser line and recovery monitored by scanning once every 15 seconds for ∼21 minutes. Fluorescence recovery within the bleach region was normalized to the pre-bleach intensity and the data fit by Eqn 1 to obtain recovery time constants (τ) (see Materials and Methods). Following photobleach, YFP-Kv2.1 exhibited low recovery, with an average mobile fraction (M_f) of 0.41±0.04 (Fig. 4A,B). Even if a 10-15% cumulative photobleach during the recovery process is taken into account, no more than 50-55% of the channel probably recovers during the 20 minute recovery period. Additionally, recovery was not well fit by a single exponential, suggesting that YFP-Kv2.1 mobility is not a diffusion-limited process as would be expected if fluorescence recovery was due to lateral membrane diffusion (Fig. 4B). Further evidence for a more complex recovery scheme is that YFP fluorescence recovered evenly across the bleached ROI (see Fig. 4, insets and Fig. S1 in supplementary material) as opposed to first recovering at the ROI border. If YFP-Kv2.1 fluorescence was due to the lateral diffusion of nonbleached channels from adjacent membrane, recovery would occur initially at the edge of the ROI. Furthermore, in no experiment did fluorescence recover completely, and the slow component of the recovery kinetics has a relatively long time constant (τ_slow=4.9 minutes, compared with 36 seconds for the fast component). We were only able to follow recovery for ∼20 minutes due to cumulative photobleaching and cell movement, and not all cells reached steady state during this time frame, introducing some error in the long time constant.

While FRAP is useful for determining the mobility of a protein, it provides no information on protein trafficking pathways. However, the recent development of photoactivatable GFP, which becomes fluorescent only after irradiation by 405-413 nm light (Patterson and Lippincott-Schwartz, 2002), provides a means of following a specific photolabeled protein. The restricted lateral mobility of Kv2.1 is clearly illustrated through the use of photoactivatable GFP-tagged Kv2.1. A region of interest along the cell membrane was photoactivated (see Materials and Methods) and the movement of GFP fluorescence monitored by scanning once every 10 seconds for ∼41 minutes (see Materials and Methods). As seen in Fig. 5, even 2500 seconds after photoactivation, at least half the GFP fluorescence is restricted to the ROI, indicating that this Kv2.1 is probably anchored in place on the cell surface. Interestingly, approximately 10 minutes after photoactivation, GFP fluorescence can be seen on the opposite membrane from the activation ROI, suggesting that some channels are not anchored at the membrane and traffic through an intracellular pathway rather than diffusing though the membrane.

While Kv2.1 resides in distinct surface puncta, Kv1.4 and Kv1.3 are both evenly distributed over the surface membrane in HEK cells (Fig. 1B,C). However, like Kv2.1, both of these channels are also present in lipid raft domains, raising the question of whether raft association alone is sufficient to immobilize channels in the plasma membrane. We therefore also used photobleach and photoactivation to investigate the mobility of these channels. As shown in Fig. 6, following photobleach, both channels recovered to a greater extent than Kv2.1, with Kv1.4 and Kv1.3 having an M_f of 0.78±0.07 and 0.78±0.04, respectively. Because these data have not been corrected for 10-15% photobleach during recovery, the measured recovery is at least 90% and in reality probably complete. Unlike Kv2.1, Kv1.4 recovery was well fit by a single exponential (τ=133 seconds). Similarly Kv1.3 fluorescence recovery was also well fit with a single time constant of 118 seconds. Unlike the recovery observed with Kv2.1, Kv1.4 appeared to recover via lateral diffusion of nonbleached channel from adjacent membrane (Fig. 6 and Fig. S2 in supplementary material). YFP-Kv1.3 also appears to recover by lateral diffusion, similar to YFP-Kv1.4 (data not shown). The high degree of mobility exhibited by Kv1.4 is dramatically illustrated by photoactivation of PA-GFP-Kv1.4, as shown in Fig. 7. As for PA-GFP-Kv2.1, a region of cell membrane was photoactivated by brief scanning with a 405 nm laser and movement of GFP fluorescence monitored every 10 seconds for approximately 41 minutes. Unlike Kv2.1, PA-GFP-Kv1.4 rapidly diffuses out of the ROI, with green fluorescence apparent in membrane proximal to the ROI within 1 minute (Fig. 7, 69 seconds). Approximately 4 minutes after photoactivation, green channel appears to have diffused throughout the entire cell membrane (Fig. 7, 259 seconds).

Because nearly half of the surface Kv2.1 channel is immobile but most of Kv1.4 is readily diffusible, we asked whether a difference in macromolecular complex assembly could be detected between these two channel isoforms. Cells were homogenized in the presence of Triton X-100 and octylglucoside to fully disrupt lipid raft domains, and the entire extract, including nuclear material, was then loaded onto a 5-20% linear sucrose density gradient. As shown in Fig. 8, approximately 50% of the YFP-Kv2.1HA migrated only slightly heavier than the predicted tetrameric molecular weight of 520,000, while the remaining channel sedimented to the bottom of the sucrose gradient. Because the gradient was centrifuged for only 2 hours at 180,000 g, the Kv2.1 at the bottom of the tube represents a very large macromolecular complex, easily exceeding a molecular weight of several million. By contrast, 95% of the Kv1.4 channel was found at the top of the sucrose gradient near the expected molecular weight of 480,000 for the tetrameric complex. The transferrin receptor and caveolin remained, as expected, near the top of the gradient (data not shown). Lack of Kv1.4 association with a large macromolecular complex is consistent with its apparent high degree of mobility following FRAP. It is also tempting to speculate that the mobile fraction of Kv2.1 is represented by the Kv2.1 protein at the top of the sucrose gradient.

All three Kv channels studied here can be isolated in cholesterol-rich lipid raft domains (Bock et al., 2003; Martens et al., 2000; Wong and Schlichter, 2004). We have previously shown that treatment of cells stably expressing Kv2.1 with 2% 2-hydroxypropyl-cyclodextrin for 1 hour alters Kv2.1 channel function, shifting the voltage-dependence of inactivation to more hyperpolarized potentials (Martens et al., 2000). Therefore, we next asked whether Kv channel localization or mobility was altered following cholesterol depletion with cyclodextrin. Interestingly, cholesterol depletion dramatically altered Kv2.1 cluster size. Following cyclodextrin treatment, Kv2.1 redistributed into large patches, as opposed to small discrete clusters (Fig. 9). Panel A shows the typical pattern in control cells (cluster size equals 1.6±0.2 μm²), whereas panel B illustrates the altered distribution following cholesterol depletion (cluster area=4.4±0.6 μm²). The Kv2.1 cluster size following cholesterol depletion is an underestimate, as the boundaries of the clusters become less distinct after cyclodextrin treatment (see Fig. 9B) and only those clusters whose margins could be clearly identified were chosen for analysis. By contrast, the surface distribution of both YFP-Kv1.4 and YFP-Kv1.3, which are already evenly distributed over the cell membrane, was unaffected by cholesterol depletion (data not shown).

Given the dramatic increase in Kv2.1 cluster size following cholesterol depletion, we next used FRAP to determine whether channel mobility was also affected. As illustrated in Fig. 10A, YFP-Kv2.1 fluorescence recovers slightly less than control following cyclodextrin treatment (M_f=0.32±0.03); however, the kinetics of recovery were altered, as recovery appears to be slowed over the early part of the recovery curve. Recovery could not be fit by a single exponential, requiring a two exponential fit with time constants of 51 seconds and 2890 seconds as compared with 36 seconds and 307 seconds under control conditions. The fractional recovery of Kv1.4 was decreased also following cholesterol depletion but to a greater extent than Kv2.1 (M_f=0.58±0.07). By contrast, the kinetics of recovery were only minimally affected (Fig. 10B), with only a slight increase in the time constant (τ=196 seconds). This decreased recovery is not a common feature of Kv1 family members, as Kv1.3 recovery was slightly increased following cyclodextrin treatment (M_f=0.89±0.06 compared with 0.78±0.04 under control conditions). Furthermore, similar to what was observed for Kv2.1, the recovery kinetics were dramatically altered for Kv1.3. Following cholesterol depletion, a single exponential fit no longer adequately fits the data and two exponentials are required (Fig. 10C). Taken together, these FRAP data indicate that Kv2.1, 1.4 and 1.3 are differentially affected by cholesterol depletion as expected if these three channels traffic through distinct membrane compartments.

The degree to which cells go to segregate ion channels into discrete compartments (e.g. Kv2.1 on the soma, Kv4.2 in dendrites and Kv1.4 in axons) is striking. However, little is known about how these channels reach their ultimate destination in native tissues, or how the highly specific localization affects channel function and cellular excitability. Surprisingly, we also know little about how these channels are trafficked to the cell surface of HEK 293 cells, the most common system used for the heterologous expression of ion channels. We show here that even in the simplified HEK 293 system, Kv2.1, Kv1.4 and Kv1.3 show differential trafficking and distinct sensitivities to membrane cholesterol depletion. These channels share extensive sequence homology in the transmembrane domains (particularly Kv1.4 and Kv1.3) but are highly divergent within the N- and C-termini, supporting the hypothesis that it is these unique domains that play a primary role in intracellular trafficking and/or immobilization on the cell surface. Indeed, several groups have implicated these domains in either channel trafficking or localization for both Kv1 and Kv2 family channels (Antonucci et al., 2001; Li et al., 2000; Lim et al., 2000; Misonou et al., 2004; Zhu et al., 2003). Kv channel N- and C-termini most probably drive trafficking through different subcellular pathways via interaction with unique sets of endogenous proteins. However, the effects of cholesterol depletion reported here support some role for lipid rafts in Kv channel localization and trafficking.

At least half of the expressed Kv2.1 had a very limited mobility in HEK cells as evidenced by FRAP recovery. Channels that did recover appeared to do so from an intracellular compartment, as no apparent diffusion of channel from neighboring membrane was observed (see Fig. 4 insets and Fig. S1 in supplementary material). The existence of two distinct Kv2.1 channel populations is supported by the sucrose density gradient centrifugation experiment of Fig. 8 in which ∼50% of Kv2.1 sediments with a mass of at least several million while the rest sedimented as expected for a Kv channel tetramer. It is tempting to correlate the high molecular weight pool with the immobile fraction observed during FRAP, as low mobility is expected if the channel is part of a large protein complex. Consistent with this idea, Kv1.4 was freely mobile in the HEK cells and 95% of this channel was found in the lighter fractions of the sucrose density gradient (Fig. 8). Despite this argument, it should be kept in mind that while channel association with the cytoskeleton could result in the heavy fraction of Kv2.1, it is also plausible that Kv2.1 is interacting with a large cytoskeletal motor, making it both heavy and highly mobile. However, Kv2.1 remains associated with insoluble elements following Triton X-100 extraction of live cells (data not shown), which is highly suggestive of cytoskeletal attachment.

It was recently reported that TrpC channels are rapidly recruited to the membrane from a submembrane compartment (Bezzerides et al., 2004), and similar mechanisms may exist for inward rectifier K⁺ channels (Ma et al., 2002). If such a mechanism exists for Kv2.1 trafficking, it could explain why Kv2.1 exhibits two different mobilities. Perhaps the mobile Kv2.1 fraction represents such a submembrane, vesicular compartment while the immobile fraction represents cytoskeleton-linked channel on the cell surface. Electron microscopy studies have shown that surface Kv2.1 clusters are localized near membrane-endoplasmic reticulum (ER) junctions in hippocampal and cortical neurons and subsurface Kv2.1 clusters have been detected in spinal motoneurons (Du et al., 1998; Muennich and Fyffe, 2004). Thus, it is plausible that a subsurface pool of Kv2.1 exists, even in HEK cells, thus providing a means of rapidly delivering Kv2.1 to the surface. We have attempted to detect such an intracellular pool of Kv2.1 just below the cell surface. For example, YFP-Kv2.1-HA channel truly on the surface, in theory, can be distinguished from subsurface channel by comparing YFP fluorescence with that derived from fluorescent antibody labeling of the external HA epitope. Despite several attempts, we have been unable to consistently detect a significant YFP signal in the absence of a cell surface derived fluorescence signal (data not shown). However, channels present in a submembrane compartment or ER near the membrane may not be discernable from surface membrane at the resolution of confocal microscopy.

One of the most striking effects of cholesterol depletion was the dramatic redistribution of Kv2.1 from small clusters into large patches (Fig. 9). It is unknown what causes the assembly of Kv2.1 into small clusters, other than the long C-terminus seems to play a role. Expression of a C-terminal truncation in hippocampal neurons results in a diffuse pattern of expression of Kv2.1 as does de-phosphorylation of the C-terminus (Lim et al., 2000; Misonou et al., 2004). Because the precise molecular machinery involved in Kv2.1 clustering remains unknown, it is not clear whether depletion of membrane cholesterol in HEK cells is mechanistically linked to the effects of C-terminal truncation or de-phosphorylation. However, Kv2.1 is lipid raft-associated, and these membrane microdomains are thought to be platforms for phosphorylation pathways and intracellular trafficking.

Given the dramatic effect cholesterol depletion had on Kv2.1 localization, the more subtle effects on Kv2.1 mobility were surprising. While overall recovery at 20 minutes was decreased by 20%, the time course of recovery was significantly slowed, with the long time constant now approaching 1 hour (Fig. 10). Because 20 minutes is insufficient time to reach steady state, it is likely that recovery would have continued, eventually reaching the control value had it been possible to follow recovery for long enough. As cholesterol depletion is expected to alter raft structure, it is reasonable to assume that rafts play some role in Kv2.1 localization and/or trafficking. Perhaps cholesterol-depleted submembrane vesicles are not as mobile as their native counterparts. Interestingly, while Kv1.4 and Kv1.3 are also lipid raft-associated, their surface distribution was not altered following cholesterol depletion (data not shown).

One potential artifact is that cholesterol depletion is expected to change membrane fluidity and the overall health of the cells, potentially affecting Kv2.1 trafficking. If Kv2.1 is part of a macromolecular complex, then a decrease in membrane fluidity following cyclodextrin treatment could conceivably slow the mobility of a large complex. There are several arguments against this, however. First, Kv2.1 recovery is not due to lateral membrane diffusion. Second, the Kv1.4 mobile fraction is decreased, but with minimal effects on recovery kinetics. Third, Kv1.3 recovery is also significantly slowed, despite an increase in the mobile fraction. Together these data suggest that protein-lipid interactions may play a role in ion channel trafficking and that the different effects of cholesterol depletion observed with the three channels under study here may be due to differential targeting to distinct raft domains.

While photobleach techniques such as FRAP are a well-established means of determining protein mobility, FRAP suffers from inherent limitations, such as a low signal-to-noise ratio. Furthermore, FRAP provides little to no information about the origin or mechanism behind fluorescence recovery and protein mobility. For protein trafficking studies, it is more advantageous to selectively photolabel the protein of interest and visualize its movement through the cell. The recent introduction of photoactivatable fluorescent proteins, such as the PA-GFP used in this study, represents a significant advance for live-cell imaging of protein trafficking, allowing the monitoring of channel movement in real time in a way that real-time imaging with conventional fluorescent proteins cannot.

One consideration that must be kept in mind is the inherent limitations on the excitation focal plane imposed by single photon excitation. The point spread function of any single photon laser will result in the bleach or activation of a substantial volume of the cell (Fig. S3 in supplementary material). This is of course also a consideration during FRAP, confounding precise measurements of recovery, since bleach and recovery occur in three dimensions. For this reason, we chose to quantitate recovery by measuring the time constant of recovery rather than calculating a diffusion constant, which would have required knowing the volume of the bleach region. For future work with PA-GFP, it will be necessary to consider that the photolabeled channels exist in a three-dimensional volume rather than a two-dimensional plane. Nevertheless, PA-GFP is a powerful tool for imaging channel trafficking in real time, as evidenced by experiments such as those shown in Figs 5 and 7. We have been able to clearly visualize migration of Kv2.1 from the activation ROI to distal regions of the cell (Fig. 5), as well as visualize the rapid diffusion of Kv1.4 (Fig. 7). In the future, photoactivation of channels early in the biosynthetic pathway will allow us to follow channels on their way to the plasma membrane.

Despite extensive amino acid sequence and functional similarities, the Kv2.1, Kv1.4 and Kv1.3 channels display unique plasma membrane expression in HEK 293 cells, distinct FRAP kinetics and different responses to cholesterol depletion. These differences support the hypothesis that the variable N- and C-termini play a primary role in intracellular trafficking and/or immobilization of the cell surface, even in HEK 293 cells that normally express no Kv currents. The majority of Kv1.4 and Kv1.3 are highly mobile in the plane of the surface membrane. By contrast, only half of Kv2.1 is mobile and channel truly on the cell surface is probably tethered to a very large macromolecular complex. The FRAP kinetics that are observed for Kv2.1 are most probably due to diffusion from inside the cell to the surface as opposed to lateral movement within the membrane. Since cholesterol depletion dramatically alters Kv2.1 clustering, but not the localization of Kv1.4 or 1.3, its remains likely that Kv isoform association with distinct lipid raft domains is involved channel trafficking and localization.

The use of HEK cells is potentially problematic, as the Kv channels studied here are normally expressed in nerve and muscle and almost certainly interact with tissue-specific proteins not present in HEK cells. However, Kv2.1 retains its neuronal expression pattern in HEK cells, indicating that much of the machinery required for this distribution is common between the two cell types. Therefore, HEK cells are a suitable system for studying Kv2.1 mobility and will be invaluable in identifying protein partners for Kv2.1 by biochemical purifications, which are not feasible in a primary culture system.

We thank Patrick D. Sarmiere for discussion and critical reading of this manuscript and Roger Tsien for providing the monomeric RFP construct. This work was supported by grants from the National Institutes of Health to M.M.T. (HL49330 and NS41542) and an NIH NRSA to K.M.S.O. (HL77056).