Interaction between the Cardiac Rapidly (IKr) and Slowly (IKs) Activating Delayed Rectifier Potassium Channels Revealed by Low K+-induced hERG Endocytic Degradation*

Background: A reduction in either IKr or IKs can cause long QT syndrome. Results: Enhanced endocytic degradation of IKr decreases the expression of both IKr and IKs in the plasma membrane. Conclusion: IKr and IKs form a macrocomplex at the plasma membrane. Significance: Elucidation of IKr-IKs interaction is important for understanding the pathology of cardiac arrhythmias and designing anti-arrhythmic strategies. Cardiac repolarization is controlled by the rapidly (IKr) and slowly (IKs) activating delayed rectifier potassium channels. The human ether-a-go-go-related gene (hERG) encodes IKr, whereas KCNQ1 and KCNE1 together encode IKs. Decreases in IKr or IKs cause long QT syndrome (LQTS), a cardiac disorder with a high risk of sudden death. A reduction in extracellular K+ concentration ([K+]o) induces LQTS and selectively causes endocytic degradation of mature hERG channels from the plasma membrane. In the present study, we investigated whether IKs compensates for the reduced IKr under low K+ conditions. Our data show that when hERG and KCNQ1 were expressed separately in human embryonic kidney (HEK) cells, exposure to 0 mm K+ for 6 h completely eliminated the mature hERG channel expression but had no effect on KCNQ1. When hERG and KCNQ1 were co-expressed, KCNQ1 significantly delayed 0 mm K+-induced hERG reduction. Also, hERG degradation led to a significant reduction in KCNQ1 in 0 mm K+ conditions. An interaction between hERG and KCNQ1 was identified in hERG+KCNQ1-expressing HEK cells. Furthermore, KCNQ1 preferentially co-immunoprecipitated with mature hERG channels that are localized in the plasma membrane. Biophysical and pharmacological analyses indicate that although hERG and KCNQ1 closely interact with each other, they form distinct hERG and KCNQ1 channels. These data extend our understanding of delayed rectifier potassium channel trafficking and regulation, as well as the pathology of LQTS.

The delayed rectifier potassium current, I K , 2 plays an important role in the repolarization of cardiac action potentials (1).
Whereas I K was originally considered to be mediated by a single type of channel (1,2), it is now clear that I K is mediated by two distinct types of channels, the rapidly (I Kr ) and the slowly activating delayed rectifier potassium channels (I Ks ) (3)(4)(5)(6). I Kr is encoded by the human ether-a-go-go-related gene (hERG, also known as KCNH2) (3,4). I Ks is encoded by both KCNQ1 and KCNE1. KCNQ1 (also known as KvLQT1) encodes the poreforming ␣ subunit, and KCNE1 (also known as minK) encodes the regulatory ␤ subunit of I Ks (5,6). Both I Kr and I Ks are critical for cardiac repolarization. Naturally occurring mutations in KCNQ1 cause type 1 long QT syndrome (LQT1). Similarly, mutations in hERG cause LQT2 and mutations in KCNE1 cause LQT5. These mutations impair the function of either I Ks or I Kr and account for the majority (Ͼ90%) of inherited long QT syndromes (7). Furthermore, a number of medications can interfere with proper hERG function, which results in acquired long QT syndrome (8).
A reduction in extracellular K ϩ concentration ([K ϩ ] o ), clinically known as hypokalemia, also causes long QT syndrome (9). We previously demonstrated that a reduction in [K ϩ ] o prolongs rabbit QT intervals on the electrocardiogram (ECG) and decreases cell surface density of both I Kr in rabbit hearts and hERG channels in stable cell lines (10,11). We further showed that low K ϩ exposure induces rapid endocytic degradation of mature hERG channels, leading to a decreased hERG channel density at the plasma membrane (10 -12). In contrast, low K ϩ exposure does not affect the expression level of either the EAG or Kv1.5 potassium channels and only moderately decreases the KCNQ1 ϩ KCNE1 current (10).
The repolarization of the cardiac action potential is under joint control of I Kr and I Ks (8). In most physiological systems, when one component declines in function or abundance, another component with similar function compensates by promoting itself to higher activity levels, a phenomenon known as functional compensation (13). It would therefore be expected that a decrease in I Kr may lead to an increase in I Ks under hypokalemic conditions. In the present study, using electrophysiology, Western blot analysis, and immunocytochemistry, we investigated the interactions between I Kr and I Ks in a * This work was supported by Heart and Stroke Foundation of Ontario Grant T 6612 and Canadian Institutes of Health Research Grant MOP 72911 (to S. Z.). 1
hypokalemia rabbit model and in the HEK 293 cell lines. Our data demonstrated that physical interactions exist between hERG and KCNQ1 proteins at the plasma membrane. Although KCNQ1 delays low K ϩ -induced hERG degradation, the endocytic degradation of hERG channels subsequently promotes KCNQ1 degradation. As a result, low [K ϩ ] o reduces the density of both hERG and KCNQ1 in the plasma membrane.

EXPERIMENTAL PROCEDURES
Hypokalemia Rabbit Model-New Zealand White rabbits (2.5-3.0 kg) were divided into two groups (nine in each group) and fed a normal or a low K ϩ diet (TestDiet) for 6 weeks. To determine the earliest experimental end point and to ensure manageability of the electrophysiological experiments on isolated cardiac myocytes, the starting time of the rabbit experiments was staggered so that one rabbit was added to each group (control and low K ϩ diet) every week. The compositions of both diets were otherwise identical except for K ϩ content (0.62% versus 0.1%). Blood samples were taken weekly to monitor serum electrolyte levels at Kingston General Hospital Clinical Laboratory (Kingston, Canada). For each rabbit, a 9-min ECG recording was taken once a week on a lightly anesthetized condition, in which 1-2% isoflurane was administered via mask using a vaporizer and veterinary anesthesia machine (Queen's University Animal Care Service). ECG was recorded with a differential AC amplifier (A-M Systems Model 1700, Carlsborg, WA), digitized using a CED Micro 1401 and stored on a computer using Spike2 software for analysis (Cambridge Electronic Design, Cambridge, UK). Six ECG signals, one in every 90 s of the recording, were used to obtain the average QT and RR intervals to generate the data points.
Ventricular myocytes were isolated from rabbits after 6 weeks on the control or the low K ϩ diet. Hearts were excised from anesthetized rabbits, mounted onto a Langendorff apparatus, and flushed at 20 ml/min with Ca 2ϩ -free Kreb's solution, which contained 110 mM NaCl, 2.6 mM KCl, 1. After transfection, the cells were cultured in 10% FBS-supplemented minimum essential medium (MEM) containing 1 mg/ml G418 for selection of transfected cells. Twenty-four single-cell derived colonies were selected for electrophysiological screening. The colony with characteristic I Ks current (presence of both KCNQ1 and KCNE1) was selected to establish the KCNQ1ϩKCNE1 stable cell line. Similarly, the colony with both characteristic I Ks current (slowly activating, presence of both KCNQ1 and KCNE1) and hERG current (presence of the hERG-specific tail current) was selected to establish the hERGϩKCNQ1ϩKCNE1 stable cell line. The selected cell clones were amplified, confirmed for the expression of hERG, KCNQ1, and KCNE1 using Western blot analysis, and stored in liquid N 2 . To maintain the stable cell lines, the cells were cultured in MEM supplemented with 10% fetal bovine serum and 0.4 mg/ml G418. For the hERGϩKCNQ1ϩKCNE1 stable cell line, all genes were maintained within 40 passages as confirmed by the presence of both hERG and I Ks (KCNQ1ϩKCNE1) currents. For transient transfection, 2 g of KCNQ1, 2 of g KCNE1, or 2 g of KCNQ1 plus 2 g of KCNE1 plasmids were transfected into hERG-HEK cells growing in a 60-mm dish at 60 -70% confluence using Lipofectamine 2000 (Invitrogen). A GFP plasmid (1 g, pIRES2-EGFP, Clontech) was co-expressed to identify transfected cells in electrophysiological studies. 24 -36 h after transfection, the cells were cultured in a custom made 0 mM K ϩ MEM-based medium or standard (5 mM K ϩ ) MEM-based medium for various periods. The cells were then harvested for Western blot, immunocytochemistry, and electrophysiological analysis. For electrophysiological studies, the cells were harvested from the culture dish by trypsinization with 0.05% trypsin (Invitrogen) and stored in standard MEMbased medium at room temperature. The cells were studied within 8 h of harvest.
Electrophysiological Recordings-The whole cell patch clamp method was used. The standard bath solution contained 135 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, and 10 mM HEPES. This standard bath solution was used for recording I hERG , I KCNQ1ϩKCNE1 in cell lines, and the transient outward K ϩ current (I to ), inwardly rectifying K ϩ current (I K1 ), resting membrane potentials and action potentials in rabbit ventricular myocytes. The bath solution for recording I Ks in rabbit ventricular myocytes contained (in mM): 140 NMG, 1 MgCl 2 , 1 CaCl 2 , 10 glucose, 10 HEPES, 0.01 nifedipine and 0.005 E4031. The pipette solution for recording I hERG contained (in mM): 135 KCl, 5 EGTA, 1 MgCl 2 , and 10 HEPES. The pipette solution for recording I KCNQ1ϩKCNE1 contained (in mM): 135 KCl, 5 EGTA, 5 K 2 ATP, 10 HEPES. The pipette solution for recording I Ks , the transient outward K ϩ current (I to ), inwardly rectifying K ϩ current (I K1 ), resting membrane potentials and action potentials contained 135 mM KCl, 10 mM EGTA, 1 mM MgCl 2 , 5 mM MgATP, and 10 mM HEPES. For I Kr recordings in rabbit ventricular myocytes, Cs ϩ -rich solutions were used to isolate I Kr from all other currents (14). The bath solution contained 135 mM CsCl, 1 mM MgCl 2 , 10 mM glucose, 10 mM HEPES, and 0.01 mM nifedipine. The pipette solution contained 135 mM CsCl, 10 mM EGTA, 5 mM MgATP, and 10 mM HEPES. To record the Ba 2ϩ -mediated L-type Ca 2ϩ currents in rabbit ventricular myocytes, the bath solution contained 140 mM TEACl, 5.4 mM BaCl 2 , 1 mM MgCl 2 , 10 mM glucose, 10 mM HEPES, and the pipette solution contained 135 mM CsCl, 10 mM EGTA, 1 mM MgCl 2 , 5 mM MgATP, and 10 mM HEPES. The pH of all bath solutions was adjusted to 7.4, and that of all pipette solutions was adjusted to 7.2 using appropriate hydroxide salts or HCl. Patch clamp experiments were performed at room temperature (22 Ϯ 1°C).
Western Blot Analysis and Co-immunoprecipitation (co-IP)-Whole cell proteins from HEK 293 cells expressing various channels were used for analysis (10 -12). Proteins were separated on 8 or 12% SDS-polyacrylamide electrophoresis gels, transferred onto PVDF membrane, and blocked for 1 h with 5% nonfat milk. The blots were incubated with the primary antibody for 1 h at room temperature and then incubated with a horseradish peroxidase-conjugated secondary antibody. Actin expression was used for loading controls. The blots were visualized with Fujifilm using the ECL detection kit (GE Healthcare).
For immunoprecipitation, whole cell protein (0.5 mg) or cell surface protein was incubated with the appropriate primary antibody overnight at 4°C and then precipitated with protein A/G plus agarose beads (Santa Cruz) for 4 h at 4°C. The beads were washed three times with ice-cold radioimmune precipitation assay lysis buffer, resuspended in 2ϫ Laemmli sample buffer, and boiled for 5 min. The samples were centrifuged at 20,000 ϫ g for 5 min, and the supernatants were collected and analyzed using Western blot.
Isolation of Cell Surface Protein-A cell surface protein isolation kit (Pierce) was used. The hERGϩKCNQ1ϩKCNE1 stably expressing HEK cells were prepared in 100-mm cell culture plates and grown to 90% confluence. The cells were labeled with 10 ml of membrane-impermeant biotinylating reagent, Sulfo-NHS-SS-biotin, for 30 min at 4°C. The quenching solution (0.5 ml) was then added to quench the reaction. The cells were then lysed with 0.5 ml of lysis buffer containing a 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 (agarose beads). The bound proteins were eluted by incubating the resin in a Tris buffer (62.5 mM Tris-HCl, pH 6.8, 1% SDS, 10% glycerol) containing 50 mM DTT. The cell surface protein was then subjected to co-IP analysis to determine hERG-KCNQ1 interactions.
Cleavage of Cell Surface Proteins-To confirm that the 155-kDa hERG protein is localized at the plasma membrane, the hERG expressions between control (treated with the buffer solution, see below) and proteinase K-treated hERG-HEK cells were compared. Live cells were washed with PBS and treated with 200 g/ml proteinase K (Sigma) in a physiological buffer (10 mM HEPES, 150 mM NaCl, and 2 mM CaCl 2 , pH 7.4) at 37°C for 30 min to cleave cell surface proteins. The reaction was terminated by adding ice-cold PBS containing 6 mM phenylmethylsulfonyl fluoride and 25 mM EDTA. The whole cell pro-teins were then extracted from the control and proteinase K-treated cells for Western blot analysis.
Immunofluorescence Microscopy-hERG-HEK cells were transfected with an empty vector (control), KCNQ1, KCNE1, or KCNQ1ϩKCNE1. Thirty-six hours after transfection, the cell surface hERG channels were labeled by incubating the live cells with a rabbit anti-hERG primary antibody (Sigma). The cells were then exposed to 0 mM K ϩ medium for 4 h and fixed with freshly prepared 4% paraformaldehyde for 15 min. The fixed cells were permeabilized with 0.1% Triton X-100 for 10 min and blocked with 5% BSA for 1 h. The permeabilized cells were treated with a goat anti-KCNQ1 or goat anti-KCNE1 primary antibody (Santa Cruz) and Alexa Fluor 594-conjugated donkey anti-goat secondary antibody to detect KCNQ1 or KCNE1. After washing off excess secondary antibody with PBS (pH 7.4), cell surface hERG channels bound with rabbit anti-hERG (Kv11.1) primary antibody (Sigma) were detected using Alexa Fluor 488 goat anti-rabbit secondary antibody. The nuclei were stained using Hoechst 33342 (0.2 g/ml; Sigma). Images were acquired using a Leica TCS SP2 Multi Photon confocal microscope (Leica, Germany).
All of the data are expressed as the means Ϯ S.E. A one-way analysis of variance was used to test for statistical significance between the control and test groups. A p value of 0.05 or less was considered significant.

Hypokalemia Prolongs the QT Interval on ECG and Decreases Both I Kr and I Ks in Rabbits-
We have previously shown that lowering [K ϩ ] o prolongs the QT interval on the ECG and decreases I Kr in a rabbit model (10). However, the effects of low [K ϩ ] o on I Ks were not well defined. Although our previous data showed that I Ks was reduced in rabbits after 4 weeks on a low K ϩ diet, the reduction did not reach statistical significance (10). Furthermore, a compensatory increase in I Ks was not found. To investigate the role of I Ks and the potential interactions between I Kr and I Ks under hypokalemic conditions, we studied I Ks in rabbits with hypokalemia induced by a low K ϩ diet for 6 weeks. Nine rabbits in each group were included in the study.
As shown in Fig. 1A, prior to feeding, there was no difference in serum [K ϩ ] between the two groups of rabbits. Serum [K ϩ ] remained stable in rabbits fed the control diet during the 6-week experimental period (3.6 Ϯ 0.1 mM at week 6 versus 3.5 Ϯ 0.3 mM at week 0, n ϭ 9, p Ͼ 0.05). However, serum [K ϩ ] in rabbits on the low K ϩ diet decreased significantly after 1 week and continued to decrease during the 6-week period ( Fig.  1A). At week 6, serum [K ϩ ] was reduced to 1.2 Ϯ 0.1 mM (n ϭ 8, p Ͻ 0.01 compared with control), and sudden death began to occur (1 of 9 rabbits). This point represents an extreme hypokalemic condition in our experiments.
The effects of hypokalemia on the QT intervals in rabbits were studied. The QT interval is dependent on the heart rate; the faster the heart rate (or the shorter the RR interval), the shorter the QT interval. To correct for the effect of RR interval on QT interval, the Bazett's formula, QTc ϭ QT/[RR] 1 ⁄ 2 , has been widely used in clinical settings (15). Recently, a linear regression method has been used to correct the QT interval in rabbit ECGs. Bruner et al. (16) identified clear genotype differences in the QT/RR slope steepness in free-moving rabbits between wild-type littermate and LQT1/LQT2 animals. Using the genotype-specific heart rate correction formula, an expected QT interval (QT exp) at a given RR interval can be calculated (16,17). Odening et al. (17) expressed the observed QT interval during anesthesia as a percentage of the expected QT (QT index) to determine the effects of anesthetic agents on QT intervals. We used isoflurane (1-2%) delivered via gas mask to sedate rabbits for ECG recordings. We constructed QT index by dividing the observed QT by the expected QT calculated using the formula (QTexp ϭ 86 ϩ 0.22*RR) generated by Bruner et al. (16) in nonanesthesized rabbits. As shown in Fig.  1B, the QT index in control diet rabbits was slightly greater than 1.0, reflecting the fact that isoflurane prolongs QT intervals in rabbits (17). However, compared with a previous study by Odening et al. (17), the effects of isoflurane on QT interval in our study was small, probably because of the low dose of isoflurane. Importantly, QT index remained constant during the 6-week experimental period in control rabbits but significantly increased in rabbits fed on low K ϩ diet (n ϭ 8, p Ͻ 0.01; Fig. 1B). These data are consistent with the fact that the reduction in serum [K ϩ ] prolongs QT intervals in our previous study, in which the Bazett's formula was used to correct QT intervals (10). The heart rate did not change significantly during 6 weeks of experiments and was not different between the control and low K ϩ diet groups (at week 6, 240 Ϯ 4/min, n ϭ 9, in control versus 255 Ϯ 14/min, n ϭ 8, in low K ϩ diet rabbits, p Ͼ 0.05).
Representative ECG recordings from rabbits on low K ϩ diet at weeks 1 and 6 are shown in Fig. 1C. Because sudden death occurs beyond week 6, the experiments on hypokalemic rabbits were terminated at week 6. Ventricular myocytes from rabbits on the control and low K ϩ diet were isolated, and electrophysiological experiments were performed on the isolated cells using the patch clamp method.
After 6 weeks on the normal or low K ϩ diet, the resting membrane potentials of rabbit ventricular myocytes were not different between control and low K ϩ groups (control, 79.7 Ϯ 1.1 mV, n ϭ 7; low K ϩ , 77.2 Ϯ 1.3 mV, n ϭ 8; p Ͼ 0.05). For recording I K1 and I to , cells were held at Ϫ40 mV. To analyze I K1 , the current at the end of a 1-s hyperpolarizing step to Ϫ120 mV was measured. I K1 was not different between control and hypokalemic rabbit ventricular myocytes (21.9 Ϯ 3.4 pA/pF in control, n ϭ 6; 22.8 Ϯ 2.3 pA/pF in hypokalemic rabbits, n ϭ 9; p Ͼ 0.05). For I to analysis, the peak current upon a 200-ms depolarizing step to 50 mV was measured. I to was 3.7 Ϯ 0.4 pA/pF in control (n ϭ 6) and 3.9 Ϯ 0.7 pA/pF in hypokalemic rabbit ventricular myocytes (n ϭ 8, p Ͼ 0.05). To record the Ba 2ϩ -mediated L-type Ca 2ϩ currents, ventricular myocytes were held at Ϫ40 mV to inactivate Na ϩ current. The peak Ba 2ϩ FIGURE 1. Reduction in serum K ؉ concentration prolongs the QT interval on ECG and decreases both I Kr and I Ks in rabbit ventricular myocytes. A, serum K ϩ concentrations of rabbits on control or the low K ϩ diet for 6 weeks. B, QT index of rabbits on control or the low K ϩ diet for 6 weeks. QT index was generated by dividing the observed QT interval on ECG by the expected QT calculated using the formula (QTexp ϭ 86 ϩ 0.22 * RR). C, representative ECG tracings from control or low K ϩ diet rabbits. D, action potentials recorded in ventricular myocytes from control or low K ϩ diet rabbits. Summarized action potential durations at 90% repolarization (APD 90 ) are shown in the right panel. E, Cs ϩ -mediated I Kr recorded in ventricular myocytes from control or low K ϩ diet rabbits. The cells were depolarized to voltages between Ϫ70 and 70 mV in 10-mV increments for 0.4 s and repolarized to the holding potential of Ϫ80 mV. The summarized inward tail currents are shown in the right panel. F, K ϩ -mediated I Ks recorded in ventricular myocytes from control or low K ϩ diet rabbits. The cells were depolarized to voltages between Ϫ70 and 50 mV in 10-mV increments for 4 s and repolarized to Ϫ50 mV before returning to a holding potential of Ϫ80 mV. The summarized pulse currents at the end of the depolarizing steps to 50 mV are shown in the right panel. *, p Ͻ 0.05; **, p Ͻ 0.01 versus control. For each of the action potential, I Kr , or I Ks experiments, 6 -11 cells from at least three independent control or low K ϩ diet rabbits were analyzed. current was observed upon a depolarizing step to 0 mV. The densities of the Ba 2ϩ -mediated L-type Ca 2ϩ currents were 15.0 Ϯ 2.9 pA/pF in control (n ϭ 7) and 14.6 Ϯ 2.7 pA/pF in low K ϩ diet rabbit ventricular myocytes (n ϭ 8, p Ͼ 0.05). These data are consistent with our previous study (10).
Action potential duration at 90% repolarization (APD 90 ) was significantly prolonged in ventricular myocytes from rabbits on the low K ϩ diet compared with that from control rabbits (Fig.  1D). Because I Kr and I Ks are jointly responsible for the repolarization of ventricular myocytes, they were the primary focus in the present study. To isolate I Kr in ventricular myocytes for analysis, we used Cs ϩ as the charge carrier (14). Cs ϩ blocks other cardiac K ϩ channels, such as I K1 , I to , and I Ks , but uniquely permeates through the native I Kr and cloned hERG channels (14). Thus, recording Cs ϩ -carried I Kr (I Kr-Cs ) using symmetrical Cs ϩ solutions represents an effective way to record pure I Kr (14,18). The I Kr-Cs in ventricular myocytes from hypokalemic rabbits was significantly smaller than that from control rabbits (Fig. 1E). This conclusion is consistent with our previous data (10). I Ks was activated by depolarizing steps to voltages between Ϫ70 and 50 mV in 10-mV increments for 4 s. I Ks was significantly decreased in ventricular myocytes of low K ϩ diet rabbits compared with that of control rabbits (Fig. 1F). Thus, I Ks did not compensate for the decreased I Kr but also decreased under hypokalemic conditions.
Co-expression of KCNQ1 ϩ KCNE1 with hERG Alters the Response of hERG Channels to Reduced [K ϩ ] o -Studies have suggested that hERG interacts with KCNQ1ϩKCNE1 (I Ks ) (16, 19 -22). We hypothesize that because of a physical association between I Kr and I Ks , endocytic degradation of hERG may promote KCNQ1ϩKCNE1 degradation and thus simultaneously decrease I Ks . To address this possibility, we created a stable cell line that expresses both hERG and KCNQ1ϩKCNE1 channels. Fig. 2A shows families of currents recorded from HEK 293 cell lines stably expressing hERG, KCNQ1ϩKCNE1, or hERGϩKCNQ1ϩKCNE1. As shown in the top panel, hERG displayed its unique fast, voltage-dependent recovery from inactivation. This unique hERG property did not exist in the KCNQ1ϩKCNE1 stable cell line (middle panel) but did exist in the hERGϩ KCNQ1ϩKCNE1 stable cell line (bottom panel). The hERGϩKCNQ1ϩKCNE1 stable cell line also displayed the time-dependent, slow activation property of the KCNQ1ϩ KCNE1 channel, which was not seen in the hERG-HEK cell line. Fig. 2B shows the presence or absence of the unique, fast, voltage-dependent inactivation of hERG channels in each of the stable cell lines. In short, both hERG and KCNQ1ϩKCNE1 currents were present in the hERGϩKCNQ1ϩKCNE1 stable cell line, whereas the individual channels were expressed alone in their respective cell lines.
We then examined the effects of 0 mM K ϩ exposure on the currents recorded from the hERG, KCNQ1ϩKCNE1, or hERGϩKCNQ1ϩKCNE1 stable cell lines. We previously showed that overnight incubation in 0 mM K ϩ medium completely eliminated I hERG and reduced I KCNQ1ϩKCNE1 by 30.3 Ϯ 8.0% (10). Our subsequent study on the time course of 0 mM K ϩ exposure-induced reduction in the 155-kDa hERG band showed that the protein decrease by 2 h and essentially disappear by 6 h (11). Thus, a 6-h exposure to 0 mM K ϩ would be expected to minimally affect the KCNQ1ϩKCNE1 current. Thus, in the present study, we exposed cells to 0 mM K ϩ medium for 6 h to take advantage of the disparity between hERG and KCNQ1ϩKCNE1 in response to 0 mM K ϩ exposure. Indeed, exposure to 0 mM K ϩ medium for 6 h completely eliminated I hERG (n ϭ 7) and decreased I KCNQ1ϩKCNE1 by 19.2 Ϯ 2.3% (n ϭ 11 for 0 mM K ϩ , n ϭ 9 for control; Fig. 3, A and C). Interestingly, in contrast to the hERG or KCNQ1ϩKCNE1 separate cell lines, the hERGϩKCNQ1ϩKCNE1 stable cell line displayed a different response to 6-h 0 mM K ϩ exposure; I hERG was reduced by 55.5 Ϯ 7.9% (n ϭ 14 for 0 mM K ϩ , and n ϭ 17 for control), and I KCNQ1ϩKCNE1 was reduced by 76.4 Ϯ 4.9% (n ϭ 11 for 0 mM K ϩ , and n ϭ 15 for control; Fig. 3, B and C).
The effects of 6-h exposure to 0 mM K ϩ on expression levels of hERG and KCNQ1 were examined using Western blot analysis of HEK 293 cells stably expressing hERG, KCNQ1ϩKCNE1, or hERGϩKCNQ1ϩKCNE1. Exposure to 0 mM K ϩ for 6 h eliminated the 155-kDa, fully glycosylated, mature form of hERG channels in the hERG stable cell line (Fig.  4, A and E) and had no significant effect on the KCNQ1 protein expression in the KCNQ1ϩKCNE1 stable cell line (Fig. 4, B and  E). However, the same treatment reduced the hERG 155-kDa form to a lesser extent and KCNQ1 to a greater extent in the hERGϩKCNQ1ϩKCNE1 stable cell line (Fig. 4, C-E). These results are consistent with the electrophysiological data and indicate that the presence of KCNQ1ϩKCNE1 retains some of the hERG channels in the plasma membrane during the 6-h exposure to 0 mM K ϩ . Conversely, internalization of hERG pro- tein in 0 mM K ϩ resulted in the simultaneous internalization of some KCNQ1 proteins.
KCNQ1, but Not KCNE1, Delays Endocytic Degradation of hERG Channels Induced by 0 mM K ϩ -Both KCNQ1 and KCNE1 are required to generate the functional I Ks current (5,6). Expression of KCNQ1 alone does not generate I Ks but instead produces currents with amplitudes within hundreds of pA and fast activation properties. Similarly, expression of KCNE1 alone does not produce any current (5, 6). When either KCNQ1 or KCNE1 was expressed independently in HEK cells, neither protein's expression level was affected by a 6-h exposure to 0 mM K ϩ medium (Fig. 5A). We have shown that KCNQ1ϩKCNE1 can retain hERG in the plasma membrane under low K ϩ conditions (Figs. 3 and 4). To study whether either KCNQ1 or KCNE1 alone is sufficient to prevent mature hERG channels from degrading in 0 mM K ϩ , we transfected hERG-HEK cells with empty vector (pcDNA3, control), KCNQ1, or KCNE1 plasmids. Thirty-six hours after transfection, the cells were exposed to 5 or 0 mM K ϩ medium for 6 h. Compared with the control cells, co-expression of KCNQ1, but not KCNE1, effectively retained the 155-kDa hERG band in cells cultured in 0 mM K ϩ medium (Fig. 5B).
To investigate whether the retained 155-kDa band indeed represents the mature hERG channels in the plasma membrane, we performed immunocytochemistry to examine the localization of hERG channels in hERG-HEK cells under 0 mM K ϩ culture conditions with or without co-expression of KCNQ1 or KCNE1. To this end, KCNQ1, KCNE1, or KCNQ1ϩKCNE1 were transiently transfected into hERG-HEK cells. Thirty-six hours after transfection, the cell surface hERG channels were labeled by incubating live cells with an anti-hERG antibody. The cells were then exposed to 0 mM K ϩ medium for 4 h, fixed, and permeabilized. KCNQ1 and KCNE1 were labeled using appropriate primary antibodies. The cells were then incubated using appropriate Alexa Fluor-conjugated secondary antibodies to stain either KCNQ1 or KCNE1 and hERG. As shown in Fig. 5C, although every cell expresses hERG FIGURE 3. Co-expression of KCNQ1؉KCNE1 with hERG changes hERG response to 0 mM K ؉ exposure. A, hERG and KCNQ1ϩKCNE1 currents from their respective stable cell lines in 5 or 0 mM K ϩ culture for 6 h. B, hERG and KCNQ1ϩKCNE1 currents in the hERGϩKCNQ1ϩKCNE1 stable cell line in 5 or 0 mM K ϩ culture for 6 h. C, the relative hERG or KCNQ1ϩKCNE1 current from each cell line in 0 mM K ϩ compared with the respective current in 5 mM K ϩ conditions. For recording I hERG , the cell membrane was depolarized to 60 mV for 200 ms followed by a repolarizing step to Ϫ100 mV for 10 ms to recover the inactivated channel to its open state. The membrane voltage was then changed to between Ϫ60 and 80 mV in 10-mV increments to observe the time-and voltage-dependent inactivation. The peak current at 50 mV was used for analysis. For recording I KCNQ1ϩKCNE1 , the cells were depolarized to voltages between Ϫ70 and 70 mV for 4 s. The cells were then repolarized to Ϫ50 mV to observe the tail current. The current amplitude at the end of a 4-s depolarization to 50 mV was used for analysis. For both current recordings, the holding potential was Ϫ80 mV. **, p Ͻ 0.01 versus separate cell lines. We further examined the effects of KCNQ1 or KCNE1 coexpression on hERG response to 0 mM K ϩ exposure by recording I hERG . After 6 h of culture in 0 mM K ϩ medium, hERG-HEK cells transfected with empty pcDNA3 vector, KCNQ1, or KCNE1 were collected, and I hERG was recorded with the 135 mM K ϩ -containing pipette solution and the 5 mM K ϩ -containing bath solution. Although I hERG in hERG-HEK cells transfected with pcDNA3 (control) or KCNE1 was completely eliminated, 40% of I hERG remained in hERG-HEK cells transfected with KCNQ1 under the same treatment (Fig. 6). Thus, consistent with the Western blots and immunocytochemical data (Fig. 5, B and C), KCNQ1 decreased the sensitivity of hERG channels to 0 mM K ϩ exposure, whereas KCNE1 did not affect hERG sensitivity to 0 mM K ϩ exposure (Fig. 6).
Co-expression of KCNQ1 Delays 0 mM K ϩ Exposure-induced Acute Reduction of I hERG -We have previously shown that upon exposure to 0 mM K ϩ , the hERG channel enters into a nonconducting state, which triggers channel internalization and degradation. To investigate the role of KCNQ1 in the conductance loss of hERG channels induced by 0 mM K ϩ exposure, we recorded I hERG from control hERG-HEK cells (transfected with the empty pcDNA3 vector) or hERG-HEK cells transfected with KCNQ1. GFP was co-transfected to identify the transfected cells for electrophysiological analysis. Consistent with our previous finding (11), when I hERG from hERG-HEK cells was repetitively evoked by the depolarizing steps (shown above Fig. 7A) at a pulse-interval of 15 s, exposure to 0 mM K ϩ Tyrode solution led to a progressive decline of the current by more than 80% in 16 pulses (n ϭ 6; Fig. 7, A and C). However, when I hERG from hERGϩKCNQ1-expressing cells was recorded, the same treatment only decreased the current by 26 Ϯ 6% (n ϭ 9; Fig. 7, B and C). Thus, co-expression of KCNQ1 significantly slowed the acute I hERG reduction upon exposure to 0 mM K ϩ solutions, which may contribute to the effects of KCNQ1 on retaining hERG in the plasma membrane.  3). B, KCNQ1, but not KCNE1, interrupts 0 mM K ϩ -induced reduction of the 155-kDa hERG band intensity. The relative reduction in the 155-kDa band intensity under each condition is summarized underneath the Western blots. **, p Ͻ 0.01 versus hERG expression alone (n ϭ 4 -5). C, KCNQ1, but not KCNE1, retains hERG membrane expression in cells cultured in 0 mM K ϩ medium for 4 h. Cell surface hERG channels were stained green. KCNQ1 or KCNE1 proteins were stained red. Nuclei were stained blue. Scale bar, 10 m. FIGURE 6. KCNQ1, but not KCNE1, decreases hERG sensitivity to 0 mM K ؉ exposure. A, current traces from hERG-HEK cells co-transfected with empty pcDNA3 vector (control), KCNQ1, or KCNE1 after 6 h of culture in 5 or 0 mM K ϩ medium. The whole cell currents were recorded in the 5 mM K ϩ containing bath solution and the 135 mM K ϩ containing pipette solution. B, relative current amplitudes from cells in 0 mM K ϩ culture compared with those in 5 mM K ϩ culture. n ϭ 11-15 cells. **, p Ͻ 0.01 versus control hERG-HEK cells.

Co-expression of KCNQ1 Does Not Change Either the Biophysical Properties or Drug Sensitivity of hERG Channels-To
investigate the effects of KCNQ1 expression on the gating kinetics of the hERG current, the current-voltage (I-V) relationships, as well as the activation curves of currents from hERG-HEK cells transfected with empty vector (control) or KCNQ1, were compared. GFP was co-transfected to identify transfected cells for electrophysiological analysis. I-V relationships were constructed by plotting the current amplitudes measured at the end of 4-s depolarizing steps (Fig. 8A). Also, the hERG activation curves were obtained by plotting the hERG tail currents at Ϫ50 mV against the depolarizing voltages and fitting the data to the Boltzmann equation (Fig. 8B). Co-expression of KCNQ1 did not affect the hERG activation curves. The V1 ⁄ 2 and the slope factor were Ϫ1.7 Ϯ 0.4 mV and 9.5 Ϯ 0.4, respectively, for hERG, and Ϫ2.9 Ϯ 0.2 mV and 9.4 Ϯ 0.1, respectively, for hERGϩKCNQ1 (n ϭ 5-7, p Ͼ 0.05). To determine the effects of KCNQ1 co-expression on hERG deactivation, the tail current at Ϫ50 mV after full channel activation (4-s depolarizing step to 50 mV) was fitted to double exponential functions. The fast and slow deactivation time constants were 334 Ϯ 16 and 1680 Ϯ 56 ms, respectively, with a relative amplitude of the slow component of 0.71 for hERG expression alone (n ϭ 9). The fast and slow deactivation time constants were 297 Ϯ 25 and 1826 Ϯ 155 ms, respectively, with a relative amplitude of slow component of 0.68 for hERGϩKCNQ1 co-expression (n ϭ 8, p Ͼ 0.05). Thus, the deactivation rate of the hERG current was not significantly affected by KCNQ1 co-expression.
We also studied the voltage-dependent inactivation of the currents from cells expressing hERG or hERGϩKCNQ1. For this analysis, the channel was activated and inactivated by a depolarizing step to 60 mV for 200 ms. Voltage was then changed to Ϫ100 ms for 10 ms, a period that is sufficient to allow inactivated hERG channels to recover to the open state but too short for channel deactivation. The cell membrane was then depolarized to various voltages to induce voltage-dependent inactivation (Fig. 8C). The current traces were fitted to a single exponential function to obtain the time constants of current inactivation, which were plotted against voltages (Fig. 8D, n ϭ 4 -6). Co-expression of KCNQ1 did not affect the voltagedependent inactivation of the hERG channel (Fig. 8D).
As can be seen from Fig. 8 (A and B), co-expression of KCNQ1 with hERG did not significantly change either the pulse or tail current amplitudes of hERG channels. Compared with the current from cells expressing hERG alone, the current from cells expressing hERGϩKCNQ1 also displayed the unique, "bell-shaped" I-V relationship. However, the pulse current amplitudes upon depolarizing voltages between 50 and 70 mV were slightly bigger in hERGϩKCNQ1-expressing cells. This may reflect the overlap with the KCNQ1 current, which does not display voltage-dependent inactivation properties. These data, and those showing that KCNQ1 co-expression did not affect the biophysical property of hERG channels, suggest that KCNQ1 and hERG may not form heterologous channels. Instead, they form distinct hERG and KCNQ1 channels. Our pharmacological data described below directly support this notion. The hERG channel, but not KCNQ1, displays uniquely FIGURE 7. KCNQ1 inhibits the reduction of I hERG upon an acute exposure to 0 mM K ؉ solution during whole cell patch clamp recordings. A and B, the first and last hERG current traces of the same HEK 293 cell expressing either hERG (A) or hERGϩKCNQ1 (B) during a 4-min exposure to 0 mM K ϩ solution. I hERG was evoked by the voltage protocol shown above the traces every 15 s. C, time-dependent reduction of I hERG in 0 mM K ϩ solution recorded from HEK 293 cells expressing either hERG or hERGϩKCNQ1 (n ϭ 6 -9 cells). The amplitude of hERG tail current at Ϫ50 mV upon each pulse was normalized to the value upon the first pulse and plotted against pulse number.  OCTOBER 7, 2011 • VOLUME 286 • NUMBER 40 high sensitivity to a wide spectrum of drugs (8,23). Astemizole is one of the most potent hERG antagonists with an IC 50 (halfmaximal inhibitory concentration) in the nanomolar range (24). Although 100 nM astemizole completely eliminated the tail current from hERGϩKCNQ1-expressing cells, it only slightly inhibited the pulse current (Fig. 8, E-H). When hERG current is specifically blocked in hERG-HEK cells, no endogenous current is apparent (24 -26) (also see Fig. 6). Thus, in the presence of astemizole, the current of hERGϩKCNQ1-expressing cells represents KCNQ1 current. These data indicate that KCNQ1 and hERG form distinct channels in HEK cells where they are co-expressed.

Interaction between I Kr and I Ks
KCNQ1 Interacts with hERG at the Cell Surface-We hypothesized that mature hERG channels physically interact with KCNQ1 and that such an interaction stabilizes hERG channels in the plasma membrane, thus delaying 0 mM K ϩ -induced hERG endocytic degradation. To study the hERG-KCNQ1 association, we performed co-IP experiments using whole cell proteins extracted from the hERGϩKCNQ1ϩKCNE1 stable cell line. When the extracted proteins were precipitated with an anti-hERG antibody, the KCNQ1 protein was detected in the precipitated proteins (Fig. 9A). Inversely, when the whole cell protein was precipitated with an anti-KCNQ1 antibody and detected with anti-hERG antibody, the mature 155-kDa hERG band was detected (Fig. 9B). It has been shown that the 155-kDa hERG is localized at the plasma membrane (18). To confirm the 155-kDa band is localized in the plasma membrane, we applied the membrane-impermeant proteinase K to the culture solution to digest the cell surface proteins in live hERG-HEK cells. The cells were then collected, and the whole cell protein was subject to Western blot analysis. Extracellularly applied proteinase K selectively digested the 155-kDa hERG and did not affect the intracellularly localized 135-kDa band (Fig. 9C). Thus, the 155-kDa hERG is indeed localized at the plasma membrane. On the other hand, the trafficking-deficient mutant hERG channel G601S is known to display only the premature 135-kDa band without the mature 155-kDa band (and is thus nonfunctional) (27)(28)(29). As shown in Fig. 9B, whereas WT hERG displayed both the 155-and 135-kDa bands, G601S only displayed the 135-kDa band. The hERG band in the anti-KCNQ1 antibody precipitated proteins corresponds to the mature 155-kDa hERG band, suggesting that KCNQ1-hERG interaction takes place at the plasma membrane. To confirm this, we isolated cell surface protein using the biotinylation method (18). We then performed KCNQ1-hERG co-IP analyses on cell surface proteins. Our data show that KCNQ1 interacts with the 155-kDa hERG band in proteins isolated from cell surface (Fig. 9D).

DISCUSSION
We have recently demonstrated that cell surface hERG channels are sensitive to [K ϩ ] o and undergo rapid endocytic degradation upon exposure to a reduced [K ϩ ] o . In contrast, KCNQ1ϩKCNE1 (I Ks ) channels are relatively insensitive to [K ϩ ] o reduction (10). In the present study, using HEK cells expressing each type of channel either separately or in combination, we have demonstrated an interaction between hERG and KCNQ1 and obtained mechanistic insight into this interaction.
The delayed rectifier potassium current, I K , plays a pivotal role in the repolarization of cardiac action potentials (8). Although I K was originally described as a single current (1), it has become evident that I K is composed of two distinct currents, I Kr and I Ks , whose channel proteins are encoded by different genes and display currents with distinct biophysical and pharmacological characteristics (3)(4)(5)(6). Although they are separate channels, I Kr and I Ks have important functional interactions during cardiac action potential repolarization. Reductions in I Kr prolong the action potential, which consequently activate more I Ks to prevent excess repolarization delay (30,31). In addition to the functional interactions, a direct physical interaction between I Kr and I Ks was first reported by Ehrlich et al. (19) and confirmed by several recent studies (20 -22). Expression of a dominant-negative hERG or KCNQ1 mutant in transgenic rabbits led to the down-regulation of the reciprocal current, indi-  4). B, detection of hERG in proteins precipitated with an anti-KCNQ1 antibody from whole cell proteins extracted from hERGϩKCNQ1ϩKCNE1 stably expressing cells (n ϭ 6). Whole cell proteins from WT hERG-HEK (displaying both the 155-and 135-kDa bands) and the trafficking-deficient hERG mutant G601S (displaying only a premature, nonfunctional 135-kDa band) were run on the right. C, extracellularly applied proteinase K selectively digests the 155-kDa hERG band. Live hERG-HEK cells were treated with proteinase K (200 g/ml) in a physiological solution at 37°C for 30 min to digest cell surface proteins. After washing, the treated cells were collected, and their protein was extracted for Western blot analysis. D, detection of hERG in proteins precipitated with an anti-KCNQ1 antibody from hERGϩKCNQ1ϩKCNE1 stably expressing cell surface protein isolated by biotinylation. Whole cell protein from hERG-HEK cells was run on the left (n ϭ 3). IB, immunoblot; IP, immunoprecipitation.
cating that an interaction between I Kr and I Ks also occurs in animals in vivo (16).
Despite these observations, the nature of the interaction between I Kr and I Ks is not well understood, and the reported results are conflicting. Although it was reported that the presence of KCNQ1 increases hERG current and enhances hERG current deactivation (19), another study showed that KCNQ1 does not affect hERG deactivation gating (21). Moreover, a more recent study reported that overexpression of either KCNQ1 or hERG significantly decreases the reciprocal current (22).
In the present study, we investigated hERG-KCNQ1 interactions using the unique extracellular K ϩ sensitivity of hERG channels. Because low K ϩ exposure only triggers hERG internalization but has a much weaker effect on KCNQ1ϩKCNE1 (10), this approach enables us to study the hERG and KCNQ1 interaction at a robust level. Our data demonstrated that hERG and KCNQ1 physically associate at the plasma membrane (Fig.  9). As a result, KCNQ1 stabilizes hERG membrane localization under low K ϩ conditions. On the other hand, low K ϩ -induced hERG endocytic degradation drags KCNQ1, as well as KCNE1, into the cell for degradation.
Neither KCNQ1 nor KCNE1 displayed a significant sensitivity to 0 mM K ϩ exposure (Figs. 4B and 5A). KCNE1 has been shown to interact with both hERG and KCNQ1 (5,6,32). In fact, KCNE1 is required for KCNQ1 to form the functional I Ks channel (5,6). Also the interaction between KCNE1 and hERG is supported by our data. When KCNE1 was expressed alone in HEK cells, it was not affected by 6-h exposure to 0 mM K ϩ conditions (Fig. 5A). However, when KCNE1 was co-expressed with hERG, it was significantly reduced by the same treatment (Fig. 5B). Although it is possible that KCNE1 serves as a linker between hERG and KCNQ1, our data showed that overexpression of KCNQ1 alone is sufficient to delay the degradation of mature hERG channels under low K ϩ conditions (Fig. 5). Furthermore, our co-IP data indicate a direct interaction between hERG and KCNQ1 in hERGϩKCNQ1-expressing HEK cells (Fig. 9). This conclusion is also in line with the previous reports (19 -21).
KCNQ1, but not KCNE1, stabilizes hERG membrane expression under low K ϩ conditions (Fig. 5). The mechanism for this difference is not known and may be related to the size of the molecule. Although KCNQ1 has six transmembrane segments, KCNE1 is a small molecule with only one transmembrane segment (33,34). Furthermore, it has been shown that KCNQ1 interacts with the protein kinase A-anchoring protein Yotiao, which may enable KCNQ1 to retain hERG channels in the plasma membrane (35).
Different from previous reports that KCNQ1 either increases or decreases hERG current amplitude (19 -22), our data show that under normal (5 mM K ϩ ) culture conditions, KCNQ1 did not significantly affect the hERG current amplitude (Fig. 8). On the other hand, under conditions where hERG membrane stability is compromised, such as hypokalemia, KCNQ1 stabilizes hERG in the plasma membrane. Our data also show that KCNQ1 did not affect the biophysical or pharmacological properties of hERG currents. The unchanged hERG biophysical and pharmacological properties in hERGϩKCNQ1-expressing cells (Fig. 8) suggest that hERG and KCNQ1 do not coassemble to form heterologous channels. Also, after hERG channel antagonist astemizole was applied, the KCNQ1 current upon depolarizing steps was observed in hERGϩKCNQ1 cells. Thus, although hERG and KCNQ1 channels associate with each other, they may only form a macromolecular channel complex consisting of distinct hERG and KCNQ1 channels. Then again, KCNQ1 must have a close interaction with hERG because the 0 mM K ϩ -induced acute hERG conductance loss was significantly weakened by the KCNQ1 co-expression (Fig. 7).
Previously, we have illustrated that [K ϩ ] o is a prerequisite for the function and membrane stability of cell surface hERG channels (10,11). Upon removal of [K ϩ ] o , hERG channels enter into a nonconducting state within minutes, reflecting a conformational change of the hERG channel. This conformational change triggers subsequent internalization and degradation of the channel (10,11). Thus, whereas the conductance loss induced by acute [K ϩ ] o depletion is reversible, prolonged exposure of hERG-expressing cells to 0 mM K ϩ medium causes progressive reduction in the expression level of hERG channels (11,12,36). Under 0 mM K ϩ culture conditions, the surface density of the mature hERG protein begins to decrease by 2 h and is essentially eliminated by 6 h (10, 11). In the present study, our data show that co-expression of KCNQ1 slowed the conductance loss of hERG channels, which suggests that KCNQ1 interferes with the hERG protein conformational change upon 0 mM K ϩ exposure, and this may contribute to the enhanced membrane stability of hERG channels.
Our data show that only the plasma membrane-localized mature form (155-kDa) of hERG co-immunoprecipitated with KCNQ1 ( Fig. 9). This observation prompts us to propose that the KCNQ1-hERG interaction occurs at the plasma membrane, and such a physical interaction stabilizes mature hERG channels under low K ϩ conditions (10 -12). Our data on cell surface proteins from hERGϩKCNQ1ϩKCNE1-expressing cells directly support this notion (Fig. 9C).
To investigate the physiological relevance of potential hERG-KCNQ1 interaction, we used a hypokalemia rabbit model. Our data show that both I Kr and I Ks are decreased in hypokalemic rabbits. Because a certain concentration of serum K ϩ is required for rabbits to survive, the extent of reduction of I Kr and I Ks in the rabbit model is less than the reduction of hERG and KCNQ1ϩKCNE1 current seen in cell lines that were exposed to 0 mM K ϩ culture. Culturing adult cardiac myocytes and studying native I Kr and I Ks trafficking are difficult tasks. We therefore cultured hERG and/or KCNQ1-expressing cell lines in 0 mM K ϩ conditions to enhance the mechanistic investigation of I Kr -I Ks interactions. Although this extreme condition may never occur in humans or animals in vivo, we have previously demonstrated that a reduction in [K ϩ ] o decreases hERG expression in the plasma membrane in a concentration-dependent manner (10). Thus, our data regarding the hERG-KCNQ1 interaction under low [K ϩ ] o conditions provide an explanation for the reduced I Ks in the hypokalemic rabbits. Under hypokalemic conditions with reduced I Kr , I Ks is critical for cardiac repolarization. Its dysfunction would result in a loss of compensatory potential and increase the risk of long QT syndrome, arrhythmias, and sudden death (30,31).