Rab11-dependent Recycling of the Human Ether-a-go-go-related Gene (hERG) Channel*

Background: The hERG-encoded potassium channel IKr is important for cardiac repolarization. Results: Internalized hERG channels are recycled back to the plasma membrane through a Rab11-associated pathway. Conclusion: Recycling plays an important role in the homeostasis of hERG channels. Significance: Identification of hERG recycling is useful for understanding hERG dysfunction and for developing new strategies to rescue hERG function. The human ether-a-go-go-related gene (hERG) encodes the pore-forming subunit of the rapidly activating delayed rectifier potassium channel (IKr). A reduction in the hERG current causes long QT syndrome, which predisposes affected individuals to ventricular arrhythmias and sudden death. We reported previously that hERG channels in the plasma membrane undergo vigorous internalization under low K+ conditions. In the present study, we addressed whether hERG internalization occurs under normal K+ conditions and whether/how internalized channels are recycled back to the plasma membrane. Using patch clamp, Western blot, and confocal imaging analyses, we demonstrated that internalized hERG channels can effectively recycle back to the plasma membrane. Low K+-enhanced hERG internalization is accompanied by an increased rate of hERG recovery in the plasma membrane upon reculture following proteinase K-mediated clearance of cell-surface proteins. The increased recovery rate is not due to enhanced protein synthesis, as hERG mRNA expression was not altered by low K+ exposure, and the increased recovery was observed in the presence of the protein biosynthesis inhibitor cycloheximide. GTPase Rab11, but not Rab4, is involved in the recycling of hERG channels. Interfering with Rab11 function not only delayed hERG recovery in cells after exposure to low K+ medium but also decreased hERG expression and function in cells under normal culture conditions. We concluded that the recycling pathway plays an important role in the homeostasis of plasma membrane-bound hERG channels.

The human ether-a-go-go-related gene (hERG) encodes the pore-forming subunit of the rapidly activating delayed rectifier potassium channel (I Kr ), which plays an important role in the repolarization of cardiac action potential (1). Medications or hERG mutations can reduce hERG 2 current (I hERG ) and cause long QT syndrome (LQTS), which predisposes affected individuals to ventricular arrhythmias and sudden death (2)(3)(4)(5). While drugs can interfere with hERG function by blocking the channel (6) or decreasing channel density in the plasma membrane (7,8), most LQTS-causing mutations in hERG impair channel function by decreasing hERG protein expression in the plasma membrane (9). The mechanisms that regulate hERG homeostasis in the plasma membrane are not well understood.
The density of hERG channels in the plasma membrane is controlled by a balance between anterograde and retrograde trafficking. Our recent works reveal that hypokalemia (a reduced extracellular K ϩ concentration) enhances the internalization and degradation of mature hERG channels (10). However, whether internalized hERG channels can recycle back to the plasma membrane has not been addressed.
In the present study, we investigated the recycling of internalized hERG channels under low K ϩ as well as normal culture conditions. Our data reveal that Rab11-mediated recycling plays an important role in the homeostasis of hERG channels in the plasma membrane.

Experimental Procedures
Molecular Biology-A human embryonic kidney (HEK) 293 cell line stably expressing hERG channels (hERG-HEK cells) was provided by Dr. Craig January (University of Wisconsin-Madison); hERG cDNA was provided by Dr. Gail Robertson (University of Wisconsin-Madison). GFP-Rab11, GFP-Rab11 dominant-negative mutant S25N GFP-Rab11, GFP-Rab4, and GFP Rab4 dominant-negative mutant N121I GFP-Rab4 plasmids were obtained from Addgene as well as from Dr. Terry Hébert (McGill University, Montreal). Cells were maintained in minimum essential medium (MEM) supplemented with 10% fetal bovine serum (FBS), 1% non-essential amino acids, and 1% sodium pyruvate (Invitrogen). For 0 mM K ϩ culture-induced hERG internalization, we used a custom 0 mM K ϩ MEM that lacks K ϩ in any form but contains all other components of standard MEM (Invitrogen). Because FBS contains K ϩ , FBS was not included in the 0 mM K ϩ or 5 mM K ϩ (control) culture medium. Lipofectamine 2000 (Invitrogen) was used for transfection of plasmids into hERG-HEK cells. For immunofluorescence staining of cell-surface hERG channels in live cells, a HA epitope tag with the sequence 436 TEEGPPATNSEHYPYDVP-DYAVTFEECGY 447 (bold indicates an insertion, and underlined indicates HA epitope) was inserted into the extracellularly localized S1-S2 loop of hERG to generate hERG-HAex via overlap extension PCR (24). The hERG-HAex plasmid was transfected into HEK293 cells, and a stable hERG-HAex cell line (hERG-HAex-HEK) was created using G418 for selection (1 mg/ml) and maintenance (0.4 mg/ml). As reported previously by others and us, inserting HA into hERG in this manner does not change the electrophysiological or trafficking properties of the hERG channel (8,25).
RNA Extraction and Quantitative Real-time PCR-Total cellular RNA was extracted from hERG-HEK cells cultured for 12 h in 5 or 0 mM K ϩ medium using a total RNA mini kit (catalog No. RB050, Geneaid Biotech Ltd., Taiwan). After treatment with DNase I (catalog No. M0303S, New England Biolabs), the RNA concentration and the 260/280 nm absorbance ratio were assessed using a spectrophotometer (SpectraMax Plus, Molecular Devices).
Total RNA (1 g) was reverse-transcribed to cDNA using the Omniscript RT kit (catalog No. 205111, Qiagen). Quantitative real-time PCR was performed using a thermal cycler (model 7500, Applied Biosystems, Foster City, CA) with TaqMan Gene Expression Master Mix (catalog No. 4369016, Life Technologies). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as a control housekeeping gene. Oligonucleotide primers were acquired from Life Technologies (hERG assay ID: Hs04234270_g1; GAPDH assay ID: Hs03929097_g1). The PCR conditions were as follows: 2 min at 50°C and 10 min at 95°C followed by 40 cycles at 95°C for 15 s and 60°C for 1 min. Data were calculated using the 2 Ϫ⌬⌬CT method and are presented as the fold induction of hERG transcripts normalized to GAPDH from hERG-HEK cells cultured in 5 or 0 mM K ϩ conditions.
Patch Clamp Recording Method-I hERG was recorded using the whole-cell patch clamp method. The bath solution contained 135 mM NaCl, 5 mM KCl, 10 mM HEPES, 10 mM glucose, 1 mM MgCl 2 , and 2 mM CaCl 2 (pH 7.4 with NaOH). The pipette solution contained 135 mM KCl, 5 mM EGTA, 1 mM MgCl 2 , and 10 mM HEPES (pH 7.2 with KOH). I hERG was recorded by depolarizing steps to voltages between Ϫ70 and 70 mV in 10-mV increments from a holding potential of Ϫ80 mV. A repolarizing step to Ϫ50 mV was used to record the tail currents. For recordings from hERG-HEK cells transfected with GFP, GFP-tagged Rab4 mutant N121I, or Rab11 mutant S25N, GFP-positive cells were used. Patch clamp experiments were conducted at room temperature (22 Ϯ 1°C).
Western Blot Analysis and Cell Biology Assays-Whole-cell protein lysates from hERG-HEK cells were separated on 5.0 or 8.0% polyacrylamide gels and electroblotted overnight at 4°C onto PVDF membranes. Membranes were blocked for 1 h using 5% skim milk in Tris-buffered saline (TBS) containing 0.1% Tween 20. Blocked membranes were immunoblotted for 1 h using the appropriate primary antibodies. Immunoblotted membranes were then incubated with the corresponding horseradish peroxidase-conjugated secondary antibodies. Protein signals were detected using an ECL detection kit (GE Healthcare). The Precision Plus Protein dual color standard was used as the protein ladder (Bio-Rad). For quantification of Western blot data, the intensities of proteins of interest in each gel were first normalized to their respective actin intensities; the normalized intensities were then expressed as values relative to their controls.
For each co-immunoprecipitation analysis, 0.5 mg of wholecell protein in 0.5 ml of lysis buffer was incubated with the appropriate primary antibody at 4°C overnight. The protein complexes were mixed with protein A/G Plus-agarose beads at 4°C for 4 h prior to precipitation by centrifugation at 10,000 ϫ g for 1 min. The immunoprecipitate was washed three times with ice-cold radioimmunoprecipitation assay lysis buffer. Then 2ϫ Laemmli sample loading buffer was added to the pelleted immunoprecipitate prior to boiling for 5 min. The sample was centrifuged at 20,000 ϫ g for 5 min, and the supernatant was collected for Western blot analysis to detect proteins associated with the pulldown protein.
For analysis of cell-surface proteins, a cell-surface protein isolation kit (catalog No. 89881, Pierce, Thermo Scientific) was used. hERG-HEK cells were cultured in 100-mm dishes and grown to 90% confluence. The cells were labeled with 10 ml of 0.25 mg/ml membrane-impermeant biotinylating reagent, sulfo-NHS-SS-biotin, for 30 min at 4°C. The quenching solution (0.5 ml) was added to stop the reaction. Cells were then lysed with 0.5 ml of lysis buffer containing 1% protease inhibitor mixture. After centrifugation at 10,000 ϫ g for 2 min at 4°C, the cell lysate was precipitated with immobilized NeutrAvidin gel. The bound proteins were eluted by incubating the resin in a sample buffer (62.5 mM Tris-HCl (pH 6.8), 1% SDS, and 10% glycerol) containing 50 mM DTT. Eluted cell-surface protein was then analyzed using Western blot analysis.
To eliminate the 155-kDa hERG protein on the cell surface, hERG-HEK cells cultured in 5 or 0 mM K ϩ medium were treated with proteinase K (PK). Specifically, intact live cells were washed with PBS and treated with 200 g/ml PK (Sigma-Aldrich) in a buffer (10 mM HEPES, 150 mM NaCl, and 2 mM CaCl 2 (pH 7.4)) at 37°C for 20 min (8,26). The reaction was terminated by adding ice-cold PBS containing 6 mM phenylmethylsulfonyl fluoride and 25 mM EDTA.
To examine the sensitivity of hERG proteins to trypsin digestion, 20 g of whole-cell protein from hERG-HEK cells cultured in 5 or 0 mM K ϩ medium for 12 h was treated with various concentrations of trypsin (ϳ10,000 benzoyl-L-arginine ethyl ester units/mg protein; Sigma-Aldrich) for 5 min at room temperature. The reaction was stopped with lima bean trypsin inhibitor (Sigma-Aldrich).
Sulfo-NHS-SS-biotin was used to track the endocytosed hERG channels. hERG-HEK cells were cultured in 5 or 0 mM K ϩ medium for 12 h. Sulfo-NHS-SS-biotin reacts with amine groups to form amide bonds (Pierce cell-surface protein isolation kit). Because the culture medium contains single amino acids, which may interfere with the reaction between sulfo-NHS-SS-biotin and cell-surface proteins, we labeled cell-surface proteins in 0 mM K ϩ Tyrode's solution with sulfo-NHS-SSbiotin (0.25 mg/ml) during a 2-h cell culture period. An additional 4-h culture in 0 mM K ϩ medium was conducted to induce internalization of biotinylated cell-surface hERG channels. Whole-cell protein lysates were collected. Biotinylated proteins were isolated using NeutrAvidin beads and assessed using Western blot analysis.
To remove N-linked oligosaccharides from the glycosylated hERG channels, peptide-N-glycosidase F (PNGase F) treatment was performed as per the protocol provided by the supplier (New England Biolabs). In brief, cells were lysed in a phosphate buffer (pH 7.4) containing 2 mM EDTA, 1% Triton X-100, 1% sodium deoxycholate, 1% protease inhibitor mixture, 1% phenylmethylsulfonyl fluoride, and 0.1% SDS. 20 g of the protein lysate was denatured by the addition of 1 l of 10ϫ denaturing buffer to a total volume of 10 l in water, and boiled (at 100°C) for 10 min. A final volume of 20 l was made by adding 2 l of 10ϫ G7 reaction buffer, 2 l of 10% Nonidet P-40, 1 l of PNGase F (500 NEB units or 7.6 IUB milliunits), and water. Samples were incubated at 37°C for 1 h and then subjected to Western blot analysis.
Endoglycosidase H (Endo H) treatment was performed as per the protocol provided by the supplier (New England Biolabs). In brief, cells were lysed in a phosphate buffer containing 2 mM EDTA, 1% Triton-X-100, 1% sodium deoxycholate, 1% protease inhibitor mixture, 1% phenylmethylsulfonyl fluoride, and 0.1% SDS (pH 7.4). 20 g of protein lysate was denatured by the addition of 1 l of 10ϫ denaturing buffer to a total volume of 10 l in water and boiled (100°C) for 10 min. A final volume of 20 l was made by adding 2 l of 10ϫ GlycoBuffer 3, 0.5 l of Endo H (250 NEB or 25 IUB milliunits), and water. Samples were incubated at 37°C for 1 h and then subjected to Western blot analysis.
Immunofluorescence Microscopy-To determine the co-localization between hERG and Rab4 or Rab11, hERG-HEK cells were grown on glass coverslips for 24 h. Cells were then cultured in 5 or 0 mM K ϩ medium for 3 h, fixed with 4% paraformaldehyde for 15 min, permeabilized with 0.1% Triton X-100 for 10 min, and blocked with 5% bovine serum albumin (BSA) for 1 h. hERG channels were stained with goat anti-hERG primary (C-20) and Alexa Fluor 488-conjugated donkey anti-goat secondary antibodies. Endogenous Rab4 or Rab11 was detected with rabbit anti-Rab4 or anti-Rab11 primary and Alexa Fluor 594-conjugated donkey anti-rabbit secondary antibodies. Images were acquired using a Leica TCS SP2 Multi Photon confocal microscope (Leica Microsystems, Wetzlar, Germany).
To visualize the recycling of hERG channels, hERG-HAex-HEK cells grown for 24 h on glass-bottom plates were incubated with anti-HA-FITC antibody at 4°C for 20 min. Unbound antibody was then washed away with PBS. Cells were cultured in 0 mM K ϩ medium at 37°C for 3 h to induce hERG endocytosis.
After images were taken, the cells were cultured under normal (5 mM K ϩ ) conditions for an additional 3 h to observe recycling of the labeled channels. Images were acquired using a Zeiss Observer.Z1 inverted fluorescence microscope (Zeiss).
To examine the recycling of hERG channels in cells under normal culture conditions, hERG-HAex-HEK cells grown on glass coverslips for 24 h were labeled with mouse anti-HA primary antibody for 30 min at room temperature (22 Ϯ 1°C). Cells were cultured for 2 h to allow internalization of labeled HA-hERG under normal culture conditions. Non-internalized HA-labeled hERG channels on the cell-surface were stained with Alexa Fluor 594-conjugated donkey anti-mouse antibodies for 30 min. Cells were then cultured for an additional 2 h under normal conditions to allow previously internalized and labeled channels to recycle back to the plasma membrane. Recycled channels (HA-labeled, secondary antibody-free) in the cell surface were then labeled with Alexa Fluor 488-conjugated goat anti-mouse antibodies for 30 min. As the non-internalized HA-labeled hERG channels on the cell surface had already been labeled with Alexa Fluor 594 (red)-conjugated donkey anti-mouse antibodies, Alexa Fluor 488 (green)-conjugated goat anti-mouse antibodies could only stain HA-labeled recycled hERG channels. Cells were fixed at various stages with 4% paraformaldehyde for 15 min. Images were acquired using a Leica TCS SP2 Multi Photon confocal microscope.
All data are expressed as the mean Ϯ S.E. A one-way analysis of variance or two-tailed Student's t test was used to determine statistical significance between the control and test groups. A p value of 0.05 or less was considered significant.

After Clearance by PK Treatment, Cell-surface hERG Channels Recover Much Faster in Cells Precultured in 0 mM K ϩ Medium Than in Cells Precultured in 5 mM K ϩ Medium-We
reported previously that culturing hERG-HEK cells in 0 mM K ϩ medium leads to a complete internalization of cell-surface hERG channels within 6 h (10, 27, 28). On Western blots, hERG proteins from whole-cell lysates of cells cultured under normal conditions (5 mM K ϩ ) display two distinct bands: a mature fully-glycosylated form in the plasma membrane with a molecular mass of 155 kDa and an immature core-glycosylated form, presumably in the ER, with a molecular mass of 135 kDa (8,29). Compared with cells cultured in 5 mM K ϩ MEM, cells cultured in 0 mM K ϩ medium for 6 h resulted in a disappearance of the 155-kDa hERG band and a 36 Ϯ 9% (n ϭ 11, p Ͻ 0.01) increase in the intensity of the 135-kDa band (Fig. 1A, upper panel). To demonstrate that the 155-kDa hERG is localized in the plasma membrane and that the 135-kDa hERG protein is localized intracellularly, we treated intact cells, precultured in 5 or 0 mM K ϩ medium for 6 h, with PK (200 g/ml, 20 min) to cleave cell-surface hERG channels (8,30). After treatment, whole-cell protein was extracted for Western blot analysis. As shown in the upper panel of Fig. 1A, treatment with PK selectively removed the 155-kDa band but did not affect the 135-kDa band, indicating that the 155-kDa hERG protein is localized in the plasma membrane. Whole-cell patch clamp analysis revealed that the 155-kDa hERG proteins represent the functional channels, as cleavage of the 155-kDa hERG by PK resulted in a complete disappearance of I hERG (Fig. 1A, lower panel).
It is generally believed that under normal culture conditions the 135-kDa protein represents the immature hERG channels localized in the ER (8,29). However, it may be more complex during hypokalemic conditions. As shown in Fig. 1A, culturing hERG-HEK cells in 0 mM K ϩ medium for 6 h led to a disappearance of the 155-kDa band and a concomitant increase in the 135-kDa band. We had demonstrated previously that 0 mM K ϩ culture-induced internalized hERG channels undergo multivesicular body and lysosomal degradation (31). Here, we propose that the internalized 155-kDa hERG channels may have been modified into a form close to 135 kDa and are initially stored in the recycling reservoir prior to degradation. The internalized hERG channels may recycle back to the plasma membrane in a route faster than new biosynthesis. Thus, following clearance of cell-surface hERG channels by PK treatment, hERG-HEK cells with enhanced hERG internalization would display an accelerated recovery of hERG channels to the cell surface upon reculture in normal K ϩ (5 mM) medium. The data shown in Fig. 1, B-D, support this notion. Specifically, we precultured hERG-HEK cells in 5 (control) or 0 mM K ϩ medium (to promote hERG internalization) overnight (12 h). We then used PK treatments to clear cell-surface hERG channels. Afterward, we cultured the cells in 5 mM K ϩ medium for various periods and monitored the recovery of I hERG using whole-cell patch clamp. We reasoned that compared with cells precultured in 5 mM K ϩ medium, cells precultured in 0 mM K ϩ medium would contain more internalized hERG channels due to the accelerated internalization. Thus, if the internalized hERG channels can recycle back to the cell surface upon reculture of PK-treated cells in normal conditions (5 mM K ϩ MEM), cells precultured in 0 mM K ϩ medium would display a faster recovery of hERG channels than cells precultured in 5 mM K ϩ medium. Our experimental data indeed support this notion. The I hERG in 0 mM K ϩ -precultured cells recovered at a rate much faster than I hERG in 5 mM K ϩ -precultured cells (Fig. 1B).
Next, we studied hERG recovery from PK treatment using Western blot analysis. The greatest difference in the recovered I hERG between the two preculture conditions was observed at 4 -8 h after PK cleavage ( Fig. 1B). Thus, we compared the recovery of hERG protein expression after a 6-h culture in 5 mM K ϩ medium following PK treatments of hERG-HEK cells precultured in 5 or 0 mM K ϩ for 12 h. As shown in Fig. 1C, the 0 mM K ϩ -precultured cells displayed a more robust expression of the recovered 155-kDa hERG band than the 5 mM K ϩ -precultured cells. The greater recovery of the 155-kDa hERG band was accompanied by a diminution of the accumulated 135-kDa hERG proteins in 0 mM K ϩ -precultured cells (differed by 4 Ϯ 5% compared with 5 mM K ϩ -precultured cells; p Ͼ 0.05, n ϭ 8). These data suggest that the accumulated pool of 135-kDa hERG derived from internalization have been recycled and converted back into the 155-kDa hERG channels, accounting for the faster and I hERG (bottom panel) in hERG-HEK cells cultured for 6 h in 5 or 0 mM K ϩ under control (Ctl) conditions or after treatment with PK (200 g/ml). The disappearance of the 155-kDa band in PK-treated cells was accompanied by the appearance of a new band with a molecular mass of 65 kDa, which represents the C-terminal hERG degradation products, as we used an anti-hERG antibody targeting an epitope in the C terminus of the channel. B, recovery of I hERG in hERG-HEK cells following the PK-mediated clearance of cell-surface hERG channels in cells precultured in 5 or 0 mM K ϩ for 12 h. After PK treatment, cells were cultured under normal conditions (5 mM K ϩ ) for various periods during which I hERG was recorded and summarized from 9 -12 cells from five independent trials. Representative hERG current traces after a 4-h recovery are shown above the summarized recovery time course. C and D, hERG expression in whole-cell protein (C, n ϭ 5) or biotin-isolated cell-surface protein (D, n ϭ 3) after a 6-h recovery from PK treatments in cells precultured in 5 or 0 mM K ϩ for 12 h. Actin and Na ϩ /K ϩ ATPase were used as the loading controls in C and D, respectively. **, p Ͻ 0.01 versus data from cells precultured in 5 mM K ϩ . recovery of mature hERG in 0 mM K ϩ -precultured cells than in 5 mM K ϩ -precultured cells.
We also isolated cell-surface proteins using a biotinylation method and assessed hERG expression using Western blot analysis. As shown in Fig. 1D, the recovery of cell-surface 155-kDa hERG expression in cells precultured in 0 mM K ϩ medium was significantly greater than that in cells precultured in 5 mM K ϩ medium. Thus, after clearance of the cell-surface hERG channels, hERG-HEK cells precultured in the 0 mM K ϩ condition, which promotes internalization, displayed a faster recovery of plasma membrane-bound hERG channels.
Rapid Recovery of Plasma Membrane-bound hERG following Vigorous Internalization Is Not due to Increased hERG Synthesis-To investigate whether an enhanced protein biosynthesis is responsible for the accelerated hERG recovery in 0 mM K ϩ -precultured cells, we first analyzed whether 0 mM K ϩ culture affects hERG mRNA levels using real-time PCR. As shown in Fig. 2A, no difference in hERG mRNA expression was observed between cells cultured in 0 and 5 mM K ϩ medium for 12 h. This suggests that hERG mRNA transcript levels are not affected by 0 mM K ϩ medium culture.
Next, we used CHX (10 g/ml) to block protein synthesis and then studied hERG recovery in cells precultured in 5 or 0 mM K ϩ medium. As shown in Fig. 2B, CHX effectively blocked hERG protein synthesis; incubation with CHX (10 g/ml) for 6, 12, and 18 h led to a 67, 82, and 91% reduction in the 135-kDa band expression, respectively. However, inhibition of protein synthesis by CHX did not disrupt the enhanced hERG recovery in 0 mM K ϩ -precultured cells. We cultured cells in 5 or 0 mM K ϩ medium for 12 h. We then cleared the cell-surface hERG channels with PK. After that, cells were cultured in normal (5 mM K ϩ ) medium containing 10 g/ml CHX for 6 h. As shown in Fig. 2, C and D, both I hERG and the 155-kDa protein expression levels were greater in 0 mM K ϩ -precultured cells. These data suggest that the fast recovery of functional hERG channels in 0 mM K ϩ -precultured cells is likely due to recycling of the internalized hERG channels.
To further determine the role of recycling in the recovery of hERG channels, we used the ER export inhibitor, FLI-06 (32). As distinct from BFA, which blocks protein transport from the ER to the Golgi and disrupts the trans-Golgi network (33), FLI-06 is a small molecule that blocks protein trafficking from the ER without disrupting the trans-Golgi network (32). As shown in Fig. 3A, FLI-06 (10 M) completely blocked hERG maturation from the 135-kDa immature form to the 155-kDa mature form. We then studied hERG recovery in cells treated with FLI-06. We precultured hERG-HEK cells in 5 or 0 mM K ϩ medium for 12 h. We then cleared the cell-surface hERG channels with PK (200 g/ml, 20 min). After PK treatment, cells were cultured in normal (5 mM K ϩ ) medium containing 10 M FLI-06 for 6 h. As shown in Fig. 3B, although there was limited recovery of I hERG in cells precultured in 5 mM K ϩ medium, recovery of I hERG in cells precultured in 0 mM K ϩ medium was significantly greater. Because FLI-06 blocks ER export (Fig. 3A), the recovered hERG current shown in Fig. 3B most likely results from the recycling of internalized hERG channels to the membrane. The faster recovery of hERG channels from PK treatment in cells precultured in 0 mM K ؉ is not a result of increased biosynthesis of hERG proteins. A, the expression levels of hERG mRNA do not differ between hERG-HEK cells cultured in 5 mM K ϩ medium and those in 0 mM K ϩ medium. After culturing cells for 12 h in 5 or 0 mM K ϩ medium, RNA was extracted, and real-time PCR was performed. The fold induction of transcripts for hERG was normalized to GAPDH in cells cultured in 5 or 0 mM K ϩ (n ϭ 4). B, hERG expression following blockade of protein biosynthesis using CHX (10 g/ml) for various periods. The intensities of the 135-kDa hERG bands were normalized to the control and plotted versus time of CHX treatment (n ϭ 4). C and D, blockade of protein synthesis with CHX (10 g/ml) does not affect the faster recovery of I hERG (C) and hERG expression (D) in 0 mM K ϩ -precultured cells. The numbers in parentheses above the bars in C indicate the number of cells examined from three independent trials. In D, the relative band intensity of the 155-kDa hERG bands from cells precultured in 0 mM K ϩ was normalized to the value from control cells (precultured in 5 mM K ϩ ) in each gel and summarized (n ϭ 3). **, p Ͻ 0.01 versus data from cells precultured in 5 mM K ϩ medium.  3). B, I hERG after a 6-h recovery culture in 5 mM K ϩ medium with FLI-06 from PK-treated cells precultured in 5 or 0 mM K ϩ medium for 12 h. The numbers in parentheses above the bars indicate the number of cells examined from three independent trials. **, p Ͻ 0.01 versus data from cells precultured in 5 mM K ϩ medium.

Visualizing the Recycling of Internalized hERG Channels
Back to the Plasma Membrane-Our data suggest that the enhanced hERG recovery rate in 0 mM K ϩ -precultured cells is due to the enlarged pool of internalized hERG channels that can recycle back to the plasma membrane. To visualize hERG recycling, we performed live cell imaging experiments. We first labeled cell-surface HA-tagged hERG channels with anti-HA-FITC antibody in hERG-HAex-HEK cells (HEK cells stably expressing HA-tagged hERG channels). After labeling, excess antibodies were removed. As shown in the left panel of Fig. 4A, only cell-surface HA-hERG channels were labeled. The cells were then cultured in 0 mM K ϩ medium for 3 h to induce internalization of the labeled HA-hERG channels (Fig. 4A, middle  panel). After taking images, the cells were restored to 5 mM K ϩ medium for an additional 3 h to allow for recycling of previously labeled and internalized hERG channels. As shown in Fig. 4A, right panel, a significant portion of the previously labeled and internalized HA-hERG channels had recycled back to the plasma membrane. Thus, it appears very likely that the internalized hERG channels can return to the cell surface within a span of hours.
Modification Occurs during Internalization and Recycling of hERG Channels-Our data indicate that internalized hERG can recycle back to the membrane. However, our Western blot data show that following a 6-h culture in 0 mM K ϩ MEM, the 155-kDa mature band was no longer present and the intensity of the 135-kDa band was increased (Fig. 1A). We hypothesized that the mature 155-kDa hERG channels undergo modification after internalization, which reduces the molecular mass to 135 kDa. To address this hypothesis, we performed an endocytosis assay using sulfo-NHS-SS-biotin to track the fate of internalized mature (155-kDa) hERG proteins. To ensure that only hERG proteins that were internalized from the cell surface were labeled with biotin, we first performed a control experiment. We precultured hERG-HEK cells in 0 mM K ϩ for 12 h to eliminate cell-surface hERG channels (intracellular hERG proteins are still abundant, see Fig. 1A). We subsequently labeled cellsurface proteins with sulfo-NHS-SS-biotin and then exposed them to 0 mM K ϩ medium for an additional 6 h. Biotinylated proteins were isolated using NeutrAvidin precipitation. Western blot analysis revealed that no detectable biotin-labeled hERG channels were present in these cells (Fig. 4B, right lane), Cell-surface HA-tagged hERG channels labeled with anti-HA-FITC antibody are internalized after a 3-h culture in 0 mM K ϩ . The internalized proteins return to the cell surface after the 0 mM K ϩ medium is switched back to 5 mM K ϩ medium for an additional 3 h. Cells were shown by taking differential interference contrast (DIC) images. B, internalized cell-surface channels are detected as the 135-kDa band. hERG-HEK cells precultured in 5 or 0 mM K ϩ for 12 h were exposed to 0 mM K ϩ Tyrode's solution containing sulfo-NHS-SS-biotin (0.25 mg/ml) for 2 h and then cultured in 0 mM K ϩ medium for an additional 4 h. The biotinylated proteins were isolated with NeutrAvidin and analyzed using Western blots (n ϭ 3). C, the 135-kDa hERG protein in 0 mM K ϩ -precultured hERG-HEK cells is less sensitive to trypsin digestion. Whole-cell lysates from hERG-HEK cells cultured in 5 or 0 mM K ϩ for 12 h were digested with different concentrations of trypsin and subjected to Western blot analysis. The intensities of the 135-kDa hERG band at each concentration of trypsin were normalized to the control (absence of trypsin) and summarized as relative values (n ϭ 4). **, p Ͻ 0.01 versus data from cells precultured in 5 mM K ϩ .
indicating that extracellularly applied sulfo-NHS-SS-biotin does not label intracellular hERG proteins. We then tracked the fate of the internalized cell-surface hERG channels. As shown in our previous work (34) as well as in Fig. 1D, cell-surface hERG with a molecular mass of 155 kDa is labeled by sulfo-NHS-SS-biotin in hERG-HEK cells cultured under normal (5 mM K ϩ ) conditions. After labeling, we cultured cells in 0 mM K ϩ medium for 6 h to induce hERG internalization. When biotinlabeled internalized proteins were isolated and analyzed by Western blot analysis, a hERG band with a molecular mass of 135 kDa was displayed (Fig. 4B, middle lane). These data indicate that mature cell-surface hERG channels are modified during internalization, decreasing their molecular mass from 155 to 135 kDa.
Therefore, after a 6-h culture in 0 mM K ϩ medium, the 135-kDa band would represent both the immature core glycosylated form and the internalized form, which may still retain its mature conformation. To test this possibility, we performed a trypsin digestion assay. It has been shown that the mature and immature forms of hERG proteins display different sensitivities to trypsin digestion; immature hERG is more sensitive to trypsin than mature hERG (35). It is thought that the complex folding structure shields certain vulnerable digestion sites from trypsin, making the mature form less susceptible to trypsin (35). We compared the sensitivity of the 135-kDa protein to trypsin in cells cultured in 5 or 0 mM K ϩ medium for 12 h. In cells cultured in 5 mM K ϩ , the 135-kDa hERG protein was indeed more sensitive to trypsin digestion than the 155-kDa hERG protein (Fig. 4C) (35). In contrast, the 135-kDa hERG protein in cells cultured in 0 mM K ϩ was much less sensitive to trypsin digestion (Fig. 4C). These data support the notion that the 135-kDa band in cells cultured in 0 mM K ϩ medium likely contains internalized mature hERG channels.
The hERG channel protein is modified by N-linked glycosylation at position Asn-598 (36). Core glycosylation of newly synthesized proteins occurs in the ER, whereas complex glycosylation occurs in the Golgi during channel maturation. On Western blots, nonglycosylated hERG channels have a molecular mass of 132 kDa (36). The mature (155 kDa) and immature (135 kDa) forms of hERG differ in the extent of N-linked complex glycosylation (36). We propose that 20-kDa oligosaccharides, which result from complex glycosylation, are trimmed off from mature channels during internalization. Thus, the internalized hERG retains its core glycosylation. As depicted in Fig.  5A, inhibition of glycosylation with tunicamycin (10 g/ml) led to a hERG band with a molecular mass of 132 kDa, which is 3 kDa less than the biotinylated and internalized form of hERG from cells cultured in 0 mM K ϩ . For further confirmation, we used PNGase F to remove oligosaccharides. Consistent with previous studies (29), PNGase F converted both the 155-kDa and 135-kDa hERG into a 132-kDa form, which is the same size as the nonglycosylated form resulting from tunicamycin treatment (Fig. 5B). We then used PNGase F to treat proteins extracted from hERG-HEK cells cultured in 0 mM K ϩ . We reasoned that if the 20-kDa loss observed in internalized hERG was not a result of glycan trimming, PNGase F treatment would further reduce the internalized hERG by 23 kDa (to ϳ112-kDa). However, our data showed that PNGase F only shifted the 135-kDa hERG proteins to the 132-kDa form (Fig. 5B, lower panel). These data support the notion that glycan trimming may occur during internalization so that the 155-kDa mature hERG channels are converted to a 135-kDa form. The internalization-derived 135-kDa channels may retain their mature structure, making them different from newly synthesized core-glycosylated hERG, which also displays a molecular mass of 135 kDa. To distinguish between these two forms of 135-kDa hERG, we used Endo H to treat hERG proteins from cells cultured in 5 or 0 mM K ϩ . Consistent with previous reports (30,36), Endo H shifted the 135-kDa immature hERG, but not the mature 155-kDa hERG, from cells cultured in normal (5 mM K ϩ ) medium (Fig. 5C). However, 135-kDa hERG proteins from cells cultured in 0 mM K ϩ medium were much less sensitive to Endo H (Fig.  5C). This suggests that the 135-kDa form in 0 mM K ϩ -cultured cells contained mature hERG channels that internalized from the cell-surface.
On the other hand, because recovered hERG channels displayed a molecular mass of 155 kDa (Figs. 1, C and D, and 2D), we believe that reglycosylation of glycan-trimmed hERG also occurs during recycling (from 135 to 155 kDa). To confirm that the recycled 155-kDa hERG is present on the cell surface, we performed a PK cut assay. As shown in Fig. 5D, extracellularly applied PK completely eliminated the 155-kDa hERG proteins, indicating their cell-surface localization. After PK clearance, the 155-kDa hERG channel during the culture in 5 mM K ϩ FIGURE 5. Deglycosylation (glycan trimming) and reglycosylation are involved in the internalization and recycling of hERG channels. A, Western blot from cells cultured in 5 or 0 mM K ϩ or 5 mM K ϩ medium with tunicamycin (Tuni, 10 g/ml) to inhibit glycosylation. Internalized hERG channels under 0 mM K ϩ culture are detected as the 135-kDa form, which is 3 kDa larger than nonglycosylated hERG channels (132 kDa) in tunicamycin-treated cells. B, upper panel, PNGase F converts both 155-and 135-kDa hERG to a 132-kDa form, which is similar to nonglycosylated hERG proteins in tunicamycintreated cells. Lower panel, PNGase F shifted both 155-and 135-kDa hERG in 5 and 0 mM K ϩ -cultured cells to 132 kDa. C, although the 135-kDa form, but not the 155-kDa form, of hERG in 5 mM K ϩ cultured cells is sensitive to Endo H digestion (converted to 132 kDa), the 135-kDa form in 0 mM K ϩ -cultured cells is less sensitive to Endo H digestion. D, recovered 155-kDa hERG channels are localized to the cell surface. After clearance of cell-surface channels by PK, upon reculture in 5 mM K ϩ medium, cells precultured in 0 mM K ϩ displayed a much faster recovery of the 155-kDa hERG channels than cells precultured in 5 mM K ϩ . The recovered 155-kDa channels are sensitive to PK cut, indicating their cell-surface localization. In A-D, consistent data were obtained in 4 -6 independent experiments for each panel.
medium recovered much faster in cells precultured in 0 mM K ϩ than in those precultured in 5 mM K ϩ . The recovered 155-kDa form is located on the cell surface because it can be completely cleared by PK digestion (Fig. 5D). These data are also consistent with our finding that the recovery of 155-kDa hERG protein parallels the recovery of I hERG (Fig. 2, C and D).
Core glycosylation of newly synthesized hERG channels occurs in the ER, and further complex glycosylation occurs in the Golgi where multiple enzyme families extend the core glycan with galactose, N-acetylglucosamine, and sialic acid (36,37). We propose that reglycosylation of internalized hERG occurs in the trans-Golgi network. To address this proposition, we used BFA (10 M) to disrupt the trans-Golgi network (33). As we reported previously (10), treatment of cells with BFA completely blocks hERG maturation (Fig. 6). More importantly, BFA also completely blocked the recovery of hERG channels in hERG-HEK cells precultured in 0 mM K ϩ (Fig. 6), where hERG recovery occurred through both the maturation and recycling of internalized channels. The complete block of hERG recovery by BFA in cells precultured in 0 mM K ϩ indicates that protein recycling passes through the trans-Golgi network where reglycosylation of hERG occurs.

Rab11 Is Involved in the Recycling of Internalized hERG Channels-Rab GTPases play important roles in ion channel
trafficking (18 -23). In particular, Rab4 and Rab11 participate in the recycling of various membrane proteins (15)(16)(17)(18)(21)(22)(23). To determine whether Rab4 and Rab11 are involved in the recycling of hERG channels, we performed co-immunoprecipitation experiments to investigate their interactions with hERG. hERG-HEK cells were transfected with GFP-Rab4 or GFP-Rab11. Twenty-four hours after transfection, whole-cell proteins were extracted. Anti-GFP antibody and protein A/G Plusagarose beads were used to pull down GFP-Rab4 or GFP-Rab11 and their associated proteins. hERG expression in the immunoprecipitates was examined using an anti-hERG antibody. As shown in Fig. 7, A and B, although an interaction between hERG and Rab11 was observed, no interaction was found between hERG and Rab4, an observation that is in line with our previous report (22). We further studied the localization of hERG and endogenous Rab4 and Rab11 in hERG-HEK cells cultured in 5 or 0 mM K ϩ medium for 3 h using immunocytochemical analysis. As shown in Fig. 7C, Rab11 co-localized with hERG under both 5 and 0 mM K ϩ culture conditions. Although hERG and Rab11 were localized primarily on the periphery of cells cultured in 5 mM K ϩ medium, they were localized within the internal vesicles of cells cultured in 0 mM K ϩ medium. In contrast, Rab4 did not co-localize with hERG (Fig. 7C). Thus, Rab11, but not Rab4, interacts with hERG, and this interaction appears to be intensified during 0 mM K ϩ medium-induced hERG internalization.
To investigate the role of Rab11 in the recycling of internalized hERG channels, we overexpressed the dominant negative mutants of Rab4 and Rab11 in hERG-HEK cells to disrupt their endogenous function. After transfection, cells were cultured in 0 mM K ϩ medium overnight (12 h) to induce hERG internalization. The cells were then restored to 5 mM K ϩ culture conditions for 6 h to allow hERG recovery. As shown in Fig. 8A, disruption of Rab11, but not Rab4, significantly delayed the recovery of I hERG . Consistent with the patch clamp data, Western blot analysis shows that disruption of Rab11, but not Rab4, delayed the recovery of the 155-kDa hERG channel expression (Fig. 8B).
Recycling Plays an Important Role in the Homeostasis of hERG Channels-Our data have demonstrated a vigorous recycling of internalized hERG channels following 0 mM K ϩ culture (Figs. 1-4). To address whether recycling also occurs in cells under normal culture conditions, we conducted a recycling assay in hERG-HAex-HEK cells under normal (5 mM K ϩ ) culture conditions. The intact cell-surface hERG-HAex channels were labeled with mouse anti-HA antibody for 30 min at room temperature, and excess antibody was removed. The cells were then cultured at 37°C for 2 h to allow internalization of HA antibody-labeled cell-surface hERG channels. Non-internalized HA antibody-labeled cell-surface hERG channels were saturated with donkey anti-mouse Alexa Fluor 594 (red)-conjugated antibody (time 0 h). At this point, an additional application of goat anti-mouse Alexa Fluor 488 (green)-conjugated antibody did not detect any signal, because HA antibodylabeled cell-surface hERG channels were preoccupied by Alexa Fluor 594 (red)-conjugated secondary antibody (time 0 h, top FIGURE 6. BFA disrupts recycling of hERG channels. BFA completely blocks the recovery of I hERG (A) and the 155-kDa band (B) following PK clearance of cell-surface proteins in hERG-HEK cells precultured in 5 or 0 mM K ϩ for 12 h. Following treatment with PK, cells were cultured in normal (5 mM K ϩ ) medium with 10 M BFA for 6 h. Patch clamp recording (A) and Western blot analysis (B) were then performed. In the bar graph in A, the numbers in parentheses indicate the number of cells tested from three independent trials. In the bar graph in B, the intensities of the 155-kDa hERG bands from cells in each condition were normalized to the value from the recovery in the absence of BFA in cells precultured in 5 mM K ϩ (n ϭ 4). **, p Ͻ 0.01 versus data from cells precultured in 5 mM K ϩ following a 6-h recovery from PK in 5 mM K ϩ . panels, Fig. 9A). After unbound antibodies were washed away, cells were then cultured for an additional 2 h to allow previously internalized HA antibody-labeled hERG channels to recycle back to the plasma membrane. Because the recycled HA antibody-labeled hERG channels were not occupied by any secondary antibody, they could be stained with Alexa Fluor 488 (green)-conjugated secondary antibody and thus were distinguishable from the yet uninternalized cell-surface hERG channels that had been stained with Alexa Fluor 594 (red)-conjugated antibody. Cells were fixed at each step, and confocal images were obtained. As shown in Fig. 9A, recycling of internalized hERG channels also occurred vigorously in cells under normal culture conditions. In other words, the recycling of hERG channels occurs not only after an enhanced internalization when cells are cultured in 0 mM K ϩ medium but also takes place when cells are cultured under normal conditions (5 mM K ϩ ).
Our data show that Rab11 was involved in the recovery of hERG channels following 0 mM K ϩ culture-induced channel internalization (Fig. 8). To determine whether Rab11 is involved in the recycling of hERG channels under normal conditions, we used the recycling assay with hERG-HAex-HEK cells, as described in Fig. 9A, to examine the effects of disrupting endogenous Rab11 by overexpressing the Rab11 dominant negative mutant. Disruption of Rab11 eliminated the recycling of hERG channels (data not shown). We further examined the impact of disrupting endogenous Rab4 and Rab11 by overexpressing their dominant negative mutants on the function and expression of hERG channels. Twenty-four to 48 h after transfection, hERG channel function was examined using the wholecell patch clamp method, and the expression levels were exam-ined using Western blot analysis. Disruption of Rab11, but not Rab4, significantly decreased I hERG and the 155-kDa hERG expression level (Fig. 9, B and C). Thus, we concluded that endogenous recycling contributes significantly to the maintenance of hERG channel density at the plasma membrane.

Discussion
The density of channels in the plasma membrane is an essential determinant of hERG function, which is critical for repolarization of the cardiac action potential (5). Although it is obvious that the density of hERG channels in the plasma membrane is controlled by both channel synthesis and degradation, the role of recycling of the internalized channels in the maintenance of hERG homeostasis is not clear. Our data in the present study show that the functional expression of cell-surface hERG channels is regulated by a recycling mechanism. Using immunocytochemical methods, we visualized the recycling of internalized hERG channels in cells exposed to 0 mM K ϩ culture conditions (Fig. 4A) as well as under normal culture (5 mM K ϩ ) conditions (Fig. 9A).
Previously, we showed that overexpression of Rab4 decreases hERG expression by enhancing the expression of the ubiquitin ligase Nedd4 -2, which mediates hERG degradation (23). In the present study, our data show that Rab4 does not directly mediate hERG recycling; Rab4 is not associated with hERG in the co-immunoprecipitation analysis, and overexpression of the dominant negative Rab4 mutant does not affect the recycling of internalized hERG channels (Figs. 7-9). On the other hand, Rab11 is involved in the recycling of internalized hERG channels; Rab11 associated with hERG based on co-immunoprecipitation analysis, and overexpression of the dominant negative Rab11 mutant significantly impaired the recycling of internalized hERG channels (Figs. 7-9).
The recovery of I hERG in hERG-HEK cells following 0 mM K ϩ culture was accompanied by the recovery of mature (155-kDa) hERG proteins (Fig. 1). The faster recovery following 0 mM K ϩ culture may not be a result of enhanced protein synthesis, as the hERG mRNA level was not affected by the 0 mM K ϩ culture ( Fig. 2A). Moreover, recovery still occurred when protein synthesis was inhibited by CHX (Fig. 2, C and D) and when ER export was blocked by FLI-06 (Fig. 3B). Culturing hERG-HEK cells in 0 mM K ϩ for 6 h led to a complete elimination of the 155-kDa hERG protein and an increase in the 135-kDa hERG protein (Fig. 1A). Therefore, we propose that following internalization, mature cell-surface hERG protein is modified from 155 to 135 kDa due to glycan trimming. The internalized mature hERG channels are still core-glycosylated, because they have a molecular mass that is 3 kDa larger than nonglycosylated hERG channels (Fig. 5A). Thus, following a 6-h culture in 0 mM K ϩ , the 135-kDa hERG proteins in whole-cell lysates are composed of two populations: the immature newly synthesized hERG proteins in the ER and the modified, internalized mature hERG proteins in the recycling endosomes. This notion is supported by our data shown in Fig. 4B; when the cell-surface 155-kDa hERG channels were tracked with sulfo-NHS-SS-biotin, FIGURE 8. Transfection of dominant negative Rab11 mutant impairs hERG recovery in cells precultured in 0 mM K ؉ medium for 12 h. Overexpression of S25N Rab11, but not N121I Rab4, slows hERG recovery from 0 mM K ϩ -induced I hERG elimination. hERG-HEK cells were transfected with GFP, GFPtagged N121I Rab4, or GFP-tagged S25N Rab11 for 24 h. Cells were then cultured in 0 mM K ϩ medium for 12 h to completely eliminate I hERG . Afterward, cells were recultured in 5 mM K ϩ medium for 6 h. A, I hERG recorded in cells from various experimental conditions. The numbers in parentheses above each bar indicate the number of cells examined from three independent trials. B, Western blots showing hERG expression in whole-cell lysates from each group of cells (n ϭ 5). **, p Ͻ 0.01 versus data from cells transfected with GFP (A) or pcDNA3 (B). FIGURE 9. Recycling of hERG channels in cells under normal (5 mM K ؉ ) culture conditions. A, confocal images showing a recycling assay of hERG channels. HA-tagged hERG-HEK cells were labeled with anti-HA primary antibody and incubated for 2 h to allow internalization. Non-internalized HA-labeled cell-surface hERG channels were stained with a saturating amount of Alexa Fluor 594-conjugated (red) secondary antibody (0 h). Cells were then incubated for another 2 h to allow the recycling of HA-labeled internalized hERG channels. Recycled channels on the surface were labeled with Alexa Fluor 488-conjugated (green) secondary antibody (2 h). B, overexpression of the dominant negative Rab11 mutant, S25N Rab11, decreases I hERG . hERG-HEK cells were transfected with GFP, GFP-tagged N121I Rab4, or GFP-tagged S25N Rab11 for 24 -48 h. I hERG was recorded using whole-cell patch clamp. The numbers in parentheses indicate the number of cells tested from three independent trials. C, overexpression of the dominant negative Rab11 mutant, S25N Rab11, decreases the 155-kDa hERG expression. The relative intensities of 155-kDa hERG bands from cells transfected with GFP-tagged N121I Rab4 or GFP-tagged S25N Rab11 were normalized to the value of control cells (transfected with pcDNA3) in each gel and summarized (n ϭ 3). **, p Ͻ 0.01 versus data from cells transfected with GFP (B) or pcDNA3 (C). the biotinylated hERG proteins reverted to 135 kDa within 6 h during 0 mM K ϩ culture. In addition, the 135-kDa hERG protein in cells cultured in 0 mM K ϩ medium was less sensitive to trypsin digestion than that of cells cultured in 5 mM K ϩ medium (Fig. 4C). Because internalized and modified hERG (also 135 kDa) in 0 mM K ϩ cultured cells may maintain the mature folding conformation that shields certain trypsin digestion sites, it is less sensitive to trypsin than the 135-kDa immature hERG proteins (35). Similarly, it has been shown that mature hERG is insensitive but immature hERG is sensitive to Endo H digestion (30,36). Our data show that Endo H digestion did not affect the 155-kDa hERG band but shifted the immature 135-kDa hERG band by 3 kDa in proteins from 5 mM K ϩ cultured cells. However, it had less effect on the 135-kDa hERG band from 0 mM K ϩ -cultured cells, which represents both mature (internalized and glycan-trimmed) and immature hERG channels (Fig. 5C).
We propose that glycan trimming and glycan addition occurs during hERG internalization and recycling. The 20-kDa difference in size between mature (155 kDa) and immature (135 kDa) hERG proteins is due to the extent of N-linked complex glycosylation (36). Our data with PNGase F support the notion that internalized 155-kDa hERG is converted to the 135-kDa form due to glycan trimming. Consistent with previous reports (29), PNGase F converted the 155-and 135-kDa forms of hERG to 132 kDa by removing all saccharides (Fig. 5B, lower panel). The fact that PNGase F shifted the 135-kDa hERG in 0 mM K ϩ medium-cultured cells to 132 kDa indicates that a 20-kDa glycan trimming occurred during hERG internalization. If the 20-kDa decrease in internalized channels was not a result of glycan trimming, PNGase F treatment would have further decreased the size of the internalized channels by 23 kDa.
The fast recovery of I hERG in 0 mM K ϩ -precultured cells is always accompanied by the reappearance of 155-kDa hERG channels, which are sensitive to PK digestion (Fig. 5D). These data indicate that glycan addition is involved in recycling. Complex glycosylation occurs in the Golgi, and our data show that disruption of the trans-Golgi network (and thus reglycosylation of internalized, glycan-trimmed hERG channels) completely prevented hERG recycling (Fig. 6). Thus, glycan trimming and glycan addition are likely involved in the internalization and recycling of cell-surface hERG channels.
Deglycosylation and reglycosylation of cell-surface glycoproteins during endocytosis and recycling has been reported previously (43)(44)(45). It has been proposed that cell-surface glycoproteins may undergo several rounds of de-and reglycosylation, and that reglycosylation might serve as a repair mechanism for cell-surface glycoproteins that are trimmed by glycosidases during endocytosis and recycling (43)(44)(45). Specifically, Kreisel et al. (43) have demonstrated that deglycosylation and reglycosylation of the glycoprotein serine peptidase dipeptidylpeptidase IV (DPPIV) occurs during recycling in rat hepatocytes. Volz et al. (44) have shown that hepatocyte glycoproteins (e.g. DPPIV) recycle through the trans-Golgi network and the trans-Golgi but not through the cis-Golgi or any earlier part of the biosynthetic pathway. Using rat hepatocytes and hepatoma cell lines, Porwoll et al. (45) have shown that postsynthetic glycans can be trimmed by mannosidases and that glycoproteins can return to the cell surface after endocytosis. Concur-ring with the results of Volz et al. (44), we found that blocking the early biosynthetic pathway using cycloheximide (Fig. 2) or blocking ER export using FLI-06 (Fig. 3) does not prevent the recycling of hERG channels. Our data suggest that internalized hERG channels recycle to the plasma membrane through the trans-Golgi network, where reglycosylation occurs. This notion is supported by the following observations. First, recovered I hERG is always accompanied by recovered 155-kDa hERG protein ( Figs. 1 and 2), which typically results from full glycosylation of hERG in the Golgi. Second, our data show that Rab11 is involved in hERG recycling (Figs. 7-9), and it has been shown that Rab11 docks at the trans-Golgi network in addition to residing in the recycling endosomes (17,46). Third, BFA directly disrupts the trans-Golgi network (33) and thus the ability to reglycosylate proteins. Our data show that BFA completely blocked recycling of internalized hERG channels (Fig.  6). The transition through the Golgi (where reglycosylation occurs) likely accounts for the slow nature of Rab11-mediated hERG recycling relative to other Rab-mediated recycling pathways of other membrane proteins (15).
Although we demonstrated that internalized hERG undergoes reglycosylation upon recycling back to the plasma membrane, whether reglycosylation is a necessity for reinsertion of hERG into the plasma membrane requires further investigation. In this regard, Porwoll et al. (45) have demonstrated that de-glycosylated internalized glycoproteins can recycle back to the plasma membrane without glycan re-addition. Gong et al. (36) have shown that glycosylation is not a requirement for hERG trafficking to the plasma membrane, but it does stabilize hERG expression on the cell surface.
The present study revealed a novel hERG channel recycling pathway, which involves Rab11 interaction as well as protein modifications during internalization and recycling. Although we used 0 mM K ϩ culture to increase internalization and thus amplify the recycling process for mechanistic investigations, recycling also occurs under normal culture conditions (Fig. 9A). Interfering with the recycling pathway by disrupting Rab11 function led to a diminished hERG function and expression in the plasma membrane (Fig. 9, B and C). Thus, the recycling of hERG channels plays an important role in maintaining the homeostasis of cell-surface hERG expression.
The hERG channel is unusually vulnerable to various conditions, such as hypokalemia, and to certain drugs that cause hERG internalization from the surface membrane. Thus, recycling represents an important energy-efficient means to rapidly restore cell-surface hERG channel expression. The significance of the hERG channel recycling pathway is highlighted by the relatively slow rate of hERG protein synthesis; after clearing cell-surface hERG channels with PK digestion, it takes almost 24 h for hERG channels to recover its initial homeostatic state (26). Conversely, hERG recovery is much faster following conditions in which hERG internalization is enhanced (e.g. low K ϩ culture; see Fig. 1B).
Most LQTS-causing mutations in hERG decrease hERG expression at the plasma membrane (9). Similarly, various conditions and drugs have been found to decrease cell-surface hERG expression, thus causing LQTS (7, 8, 10, 47, 48). Our findings raise the possibility that certain mutations and drugs may interfere with recycling, thereby disrupting hERG function. Consequently, strategies that enhance recycling may be useful in rescuing disrupted hERG function for the treatment of certain LQTS patients.
Author Contributions-J. C., J. .G, and W. L. designed, performed, and analyzed the experiments shown in Figs. 1-4, 6, 7, A and B, 8,  and 9, B and C. T. Y. designed, performed, and analyzed the experiments shown in Figs