[Ca2+]i Elevation and Oxidative Stress Induce KCNQ1 Protein Translocation from the Cytosol to the Cell Surface and Increase Slow Delayed Rectifier (IKs) in Cardiac Myocytes*

Background: The degree of KCNQ1/KCNE1 (IKs components) colocalization in cardiomyocytes is unclear. Results: In ventricular myocytes, KCNE1 is mainly on the cell surface, whereas KCNQ1 is mainly in the intracellular compartment. [Ca2+]i elevation or oxidative stress triggers KCNQ1 trafficking to the cell surface. Conclusion: Stress-induced KCNQ1 trafficking influences IKs amplitude. Significance: This is a novel mechanism for IKs function as a “ventricular repolarization reserve.” Our goals are to simultaneously determine the three-dimensional distribution patterns of KCNQ1 and KCNE1 in cardiac myocytes and to study the mechanism and functional implications for variations in KCNQ1/KCNE1 colocalization in myocytes. We monitored the distribution patterns of KCNQ1, KCNE1, and markers for subcellular compartments/organelles using immunofluorescence/confocal microscopy and confirmed the findings in ventricular myocytes by directly observing fluorescently tagged KCNQ1-GFP and KCNE1-dsRed expressed in these cells. We also monitored the effects of stress on KCNQ1-GFP and endoplasmic reticulum (ER) remodeling during live cell imaging. The data showed that 1) KCNE1 maintained a stable cell surface localization, whereas KCNQ1 exhibited variations in the cytosolic compartment (striations versus vesicles) and the degree of presence on the cell surface; 2) the degree of cell surface KCNQ1/KCNE1 colocalization was positively correlated with slow delayed rectifier (IKs) current density; 3) KCNQ1 and calnexin (an ER marker) shared a cytosolic compartment; and 4) in response to stress ([Ca2+]i elevation, oxidative overload, or AT1R stimulation), KCNQ1 exited the cytosolic compartment and trafficked to the cell periphery in vesicles. This was accompanied by partial ER fragmentation. We conclude that the cellular milieu regulates KCNQ1 distribution in cardiac myocytes and that stressful conditions can increase IKs by inducing KCNQ1 movement to the cell surface. This represents a hitherto unrecognized mechanism by which IKs fulfills its function as a repolarization reserve in ventricular myocytes.

Our goals are to simultaneously determine the three-dimensional distribution patterns of KCNQ1 and KCNE1 in cardiac myocytes and to study the mechanism and functional implications for variations in KCNQ1/KCNE1 colocalization in myocytes. We monitored the distribution patterns of KCNQ1, KCNE1, and markers for subcellular compartments/organelles using immunofluorescence/confocal microscopy and confirmed the findings in ventricular myocytes by directly observing fluorescently tagged KCNQ1-GFP and KCNE1-dsRed expressed in these cells. We also monitored the effects of stress on KCNQ1-GFP and endoplasmic reticulum (ER) remodeling during live cell imaging. The data showed that 1) KCNE1 maintained a stable cell surface localization, whereas KCNQ1 exhibited variations in the cytosolic compartment (striations versus vesicles) and the degree of presence on the cell surface; 2) the degree of cell surface KCNQ1/KCNE1 colocalization was positively correlated with slow delayed rectifier (I Ks ) current density; 3) KCNQ1 and calnexin (an ER marker) shared a cytosolic compartment; and 4) in response to stress ([Ca 2؉ ] i elevation, oxidative overload, or AT1R stimulation), KCNQ1 exited the cytosolic compartment and trafficked to the cell periphery in vesicles. This was accompanied by partial ER fragmentation. We conclude that the cellular milieu regulates KCNQ1 distribution in cardiac myocytes and that stressful conditions can increase I Ks by inducing KCNQ1 movement to the cell surface. This represents a hitherto unrecognized mechanism by which I Ks fulfills its function as a repolarization reserve in ventricular myocytes. KCNQ1 and KCNE1 associate to form slow delayed rectifier (I Ks ) channels (1). I Ks functions as a "repolarization reserve" in human ventricles by contributing outward currents to prevent undue action potential prolongation when the ␤-adrenergic tone is high or when other repolarizing currents are diminished (2). Loss-of-function mutations in KCNQ1 or KCNE1 have been linked to long QT syndrome (LQT1 and LQT5) (3), whereas gainof-function mutations have been linked to familial atrial fibrillation or short QT syndrome (SQT2) (4,5). These observations indicate that proper control of I Ks amplitude is required to maintain the electrical stability of both ventricles and atria. Situations of too little or too much I Ks are equally dangerous.
We have shown that, in COS-7 cells, the partnership between KCNQ1 and KCNE1 is dynamic (6). KCNE1 can dissociate from KCNQ1, leaving room in the KCNQ1 channel to pick up new KCNE partners. Whether these observations can be extrapolated to native KCNQ1 and KCNE1 in cardiac myocytes is an open and important question. Other members of the KCNE family (KCNE2-KCNE4) have been detected in the human heart (7)(8)(9). In heterologous expression systems, these KCNE subunits can each associate with KCNQ1 to alter its current amplitude and gating kinetics (10). The degree and stability of KCNQ1/KCNE1 assembly in cardiac myocytes will influence how likely it is that the other KCNE subunits can associate with KCNQ1 and modulate its function.
It is not clear how well and where native KCNQ1 and KCNE1 are colocalized in cardiac myocytes. The major distribution pattern of KCNQ1 in atrial and ventricular myocytes of several animal species is transverse striations (11)(12)(13)(14)(15). On the other hand, the distribution pattern of native KCNE1 in cardiac myocytes has been described as "diffuse" (16), clustered to the intercalated disc region (13) or localized in the z-line region (15,16). In this study, we used multiple approaches to simultaneously determine the distribution patterns of native KCNQ1 and KCNE1 in cardiac myocytes. We also explored the mechanism and functional implications for variations in the degree of KCNQ1/KCNE1 colocalization in cardiac myocytes. We used the guinea pig model because its atrial and ventricular myocytes express a robust I Ks .
Neonatal rat ventricular myocytes (NRVMs) were isolated from 1-or 2-day-old Wistar-Kyoto rats under sterile conditions using mild enzymatic (0.1% trypsin) digestion and gentle mechanical trituration as described previously (19). After preplating (36°C for 30 min) to remove contaminating fibroblasts, myocytes were plated on laminin-or fibronectin-coated coverslips in minimum Eagle's medium containing 10% FBS and cultured at 36°C in 5% CO 2 moist incubator. After being cultured for 3 days, cells were incubated with adenoviruses harboring Q1-GFP and E1-dsR (0.25 ϫ 10 7 plaque-forming units/ml each) overnight. After removing the viruses, culture continued for the specified amounts of time (1, 3, 6, and 12 h). At the end of the culture, cells were used for patch clamp experiments or live-or fixed-cell confocal imaging.
Patch Clamp Experiments-Currents were measured using the whole-cell recording configuration of the patch clamp technique with EPC-8 (HEKA, Germany) or Axiopatch 200B (Molecular Devices) amplifiers. Patch pipettes had tip resistances of 2.5-4.0 M⍀ (atrial myocytes, COS-7, and NRVM) or 1.5-2.5 M⍀ (ventricular myocytes). After forming the wholecell recording configuration, series resistance was compensated up to 95%. The pipette tip potential was zeroed before forming the whole-cell recording configuration, and the liquid junction potential (pipette versus bath solutions, estimated to be Ϫ10 mV) was subtracted during data analysis. To record the I Ks from GPA and GPV myocytes, after establishing the whole-cell recording configuration in normal Tyrode's, the superfusate was switched to sodium-, potassium-, and calcium-free solution with 3 M E-4031. These conditions isolate I Ks from overlapping I Na , I CaL , I K1 , I Kr , and I NCX . In all experiments, the bath solution was maintained at 35 Ϯ 1°C. The program suite pClamp10 was used for data acquisition and analysis. Specific voltage clamp protocols and data analysis are described in the figure legends. COS-7 Cell Culture and cDNA Transfection-COS-7 cells were maintained in DMEM (Invitrogen) supplemented with 10% FCS and penicillin/streptomycin in a 5% CO 2 moist incubator at 36°C. Cells were plated at a subconfluence level the day before transfection. Cells were transfected with cDNAs encoding Q1-GFP and E1-dsR (cDNA molar ratio 1:1) and, where noted, angiotensin type 1 receptor (AT1R) or ER-FRP. Transfection was facilitated by Lipofectamine 2000.
Immunofluorescence, Confocal Microscopy, and Image Analysis-Fixed cells attached to coverslips were permeabilized by 0.2% saponin (at 4°C overnight or at room temperature for 2 h). Cells were incubated with primary/Alexa Fluor-conjugated secondary antibody pairs (see below). Where specified, cells were incubated with Alexa Fluor-conjugated wheat germ agglutinin (WGA To quantify KCNQ1 immunofluorescence signals in the cell periphery (Figs. 1F and 9C), the z-stack images of each myocyte were projected into a summed image that was used to calculate the total KCNQ1 signals, Q total (within the cell contour), and periphery KCNQ1 signals, Q periphery (within a pace of Յ 2 m width from the cell contour). The percentage of KCNQ1 in the cell periphery was calculated as (Q periphery /Q total ) ϫ 100%.
For colocalization analysis (Fig. 5B), confocal data were analyzed using the Volocity program (PerkinElmer Life Sciences). We focused on the central z-optical slice. After image deconvolution to reduce out-of-focus pixels from adjacent z-sections, the region of interest (the lateral cell surface defined as a Յ 2 m-wide space within the cell contour) was specified. The threshold value for each fluorescence channel was set by the average pixel intensity in the cell-free area, and the degree of colocalization of KCNQ1/KCNE1 immunofluorescence signals was evaluated by the Pearson's correlation coefficient.
We used acceptor photobleach fluorescence resonance energy transfer (apFRET) to detect and quantify the functional assembly between native KCNQ1 and KCNE1 in GPA and GPV myocytes (Fig. 5, C and D) or between native ether-a-go-go related gene (ERG) and KCNE1 in GPV myocytes (Fig. 11). KCNQ1 or ERG was labeled by rabbit Ab/Alexa Fluor 488 antirabbit (as FRET donor), and KCNE1 was labeled by mouse Ab/Alexa568 anti-mouse (FRET acceptor). Each apFRET experiment had five scans, applied once every 16 s. Each scan contained sequential scans of Alexa Fluor 488 and Alexa Fluor 568 using 1% laser power (to minimize unintentional photobleaching). An optic thickness of 4 m was used to catch most of the fluorescence signals from cells. Scans 1 and 2 were prebleach scans. Between scans 2 and 3, laser 561 at 100% power was applied 100 times to selected regions of interest to bleach Alexa Fluor 568 to Ͻ 20% of the control. Cells in the same microscopic field whose Alexa Fluor 568 fluorescence was not bleached served as a reference, "REF", reporting changes in Alexa Fluor 488 intensity unrelated to apFRET. Scans 3-5 were post-bleach scans. The pixel contents of donor fluorescence from the whole cell contour were measured for all five scans, background-subtracted, and normalized by the scan 2 data. The and KCNE1 mAb (mouse) for detecting native proteins in GPA and GPV myocytes, respectively, using immunofluorescence (IF). ϩ, Ab preincubated with antigen (Ag) overnight; Ϫ, Ab preincubated overnight in buffer without antigen. B and C, simultaneously detecting three-dimensional distribution patterns of KCNQ1 and KCNE1 in ventricular and atrial myocytes using KCNQ1 goat pAb and KCNE1 mouse mAb. The main panels depict x-y plane images of KCNQ1 and KCNE1 immunofluorescence signals taken from the central z slice, with x-z plane images reconstructed from the z-stacks shown on top. In B, enlarged views of the areas boxed in yellow frames are shown below. D, KCNQ1 rabbit pAb detected similar variations in KCNQ1 distribution patterns in ventricular and atrial myocytes. In each panel, a cell or cell cluster is identified as #1 or #2, and the KCNQ1 distribution patterns are noted (the definition of KCNQ1 distribution patterns is described in the text). Scale bars in B-D ϭ 5 m. E, percentages of atrial (A) and ventricular (V) myocytes manifesting KCNQ1 distribution patterns 1, 2, and 3. F, degrees of KCNQ1 presence in periphery of atrial and ventricular myocytes of KCNQ1 distribution pattern 1 versus 2 or 3. ***, t test, p Ͻ 0.001. degree of apFRET was calculated as [(Donor post Ϫ Donor pre )/ Donor pre ] ϫ 100%. A FRET map was generated using the ImageJ plug-in "FRETcalc." Immunoblot Experiments-Whole-tissue or whole-cell lysate (WTL or WCL, respectively) was prepared from guinea pig heart or myocytes using the protocol described by O'Rourke et al. (20). Whole cell lysate was prepared from COS-7 cells using a method described previously (6). Protein concentration was measured with a Micro BCA TM protein assay kit (Pierce). Samples were loaded onto 10% (KCNQ1) or 16% (KCNE) denaturing (SDS)/reducing (mercaptoethanol) polyacrylamide gels. After electrophoresis, the proteins were blotted onto PVDF membranes and probed with suitable antibodies. Immunoreactivity was visualized using an ECL detection kit (Amersham Biosciences). To check loading variations, the same membranes were stripped and reprobed with actin mAb, or the remaining proteins in the gels were stained with Coomassie Blue. Immunoreactive band or Coomassie Blue intensities were measured by densitometry (ChemiImager model 4400, ␣-Innotech).
Ionomycin was dissolved in dimethyl sulfoxide to make 10 mM stock solutions, stored in aliquots at Ϫ20°C and diluted to 10 M in bath solution for experiments. Angiotensin II (Sigma) was dissolved in Millipore water to make 1 mM stock solutions, stored in aliquots at Ϫ20°C and diluted to 1 M in bath solution for experiments. H 2 O 2 (30% solution, Sigma) was diluted to 0.01-0.1% with bath solution for experiments.
Statistics-Data are presented as mean Ϯ S.E. Comparison between two groups was done by two-tailed Student's t test. Multiple group comparisons were assessed by one-way analysis of variance, followed by Dunn's test versus reference. SigmaStat (version 2) was used for statistical analysis.

Three-dimensional Distribution Patterns of Native KCNQ1 and KCNE1 in Adult Guinea Pig Atrial and Ventricular Myocytes
We simultaneously monitored the immunofluorescence signals of KCNQ1 and KCNE1 and used z-stack images to reconstruct their three-dimensional patterns. This allowed us to "see" where they were, both inside of the cells and on the cell surface. A prerequisite for the success of this approach was reliable Abs for immunofluorescence detection. Fig. 1A shows that preincubating primary Abs (anti-KCNQ1 goat and rabbit pAbs and anti-KCNE1 mouse mAb) with respective antigens abolished the immunofluorescence signals, supporting their specificity.
In both guinea pig atrial and ventricular myocytes, KCNE1 maintained a stable cell surface localization. This is best seen in the cross-section images reconstructed from the z-stacks (Fig. 1, B and C, top panels). On the other hand, the KCNQ1 immunofluorescence signals showed clear variations. There were three main patterns. In pattern 1, KCNQ1 was distributed as prominent transverse striations with a spacing of 1.9 Ϯ 0.1 m (Fig. 1, B (GPV #1) and C (GPA #1)). The cross-section images show that these striations were deep in the cytosol. In pattern 2, KCNQ1 was in prominent vesicles as well as in striations (Fig. 1D, GPV #1 and GPA #1). In pattern 3, KCNQ1 was mainly in vesicles, with no or only dim/ incomplete striations (Fig. 1D, GPV #2 and GPA #2).
On the basis of data from 172 atrial and 125 ventricular myocytes isolated from six guinea pigs and analyzed in seven independent confocal experiments, 53, 32, and 15% of atrial myocytes manifested KCNQ1 distribution patterns 1, 2, and 3, respectively (Fig. 1E). For ventricular myocytes, the numbers were comparable (62, 28, and 10%). Most atrial myocytes had a strong KCNQ1 presence on the lateral surface, irrespective of the cytosolic KCNQ1 distribution pattern (Fig. 1F). In "pattern 1" ventricular myocytes, KCNQ1 signals on the lateral surface were weak and uneven. Reconstructed three-dimensional images suggest a patchy KCNQ1 distribution pattern on the lateral surface. "Pattern 2" or "pattern 3" ventricular myocytes had a significantly higher KCNQ1 presence in the cell periphery (Fig. 1F). The nature of these variations in the KCNQ1 distribution pattern and their functional significance were explored in the following experiments.

Location of KCNQ1 Striations in Guinea Pig Atrial and Ventricular Myocytes
We compared the KCNQ1 striation pattern with those of markers for subcellular compartments or organelles. Alexa Fluor-conjugated WGA was used to stain surface membrane and t-tubules. In most ventricular myocytes, the KCNQ1 striations coincided with those of WGA, consistent with the notion that KCNQ1 is in t-tubules ( Fig. 2A, a) (21). However, in ventricular myocytes that had been partially detubulated (incubation for 30 min in 1.5 M formamide), we still saw clear Top panel, patch clamp recording from COS-7 cells expressing Q1-GFP alone or with E1-dsR. Currents were elicited by 2-s test pulses from Ϫ80 mV holding voltage to Ϫ70 to ϩ 100 mV in 10-mV increments, and tail currents were recorded after repolarization to Ϫ60 mV. The relationship between tail current amplitude (I tail ) and test pulse voltage (V t ) was fit with a simple Boltzmann function: to estimate the maximal I tail (I max ), half-maximum activation voltage (V 0.5 ) and slope factor (k). I tail values were normalized by I max (fraction activated) and plotted against V t on the right. The V 0. 5  KCNQ1 striations in areas where t-tubules were missing ( Fig.  2A, b). Guinea pig atrial myocytes did not have t-tubules (WGA stained the lateral surface without striations), yet KCNQ1 was in clear striations ( Fig. 2A, c). Finally, Fig. 2A, d, illustrates a ventricular myocyte in which KCNQ1 striations alternated with WGA striations. There was a minor KCNQ1 band between the major striation bands that coincided with the WGA band.
We further compared KCNQ1 striation patterns with the following cytosolic markers: ryanodine receptor (RyR2, marking the junctional sarcoplasmic reticulum (SR)), ␣-actinin (z-lines), and calnexin (ER/SR). Fig. 2B shows that, in both atrial and ventricular myocytes, KCNQ1 striations registered with those of RyR2, ␣-actinin, and calnexin. The cross-section images of merged immunofluorescence signals clearly show that KCNQ1 was distributed deep into the cytosol along with these cytosolic markers. Together, these data suggest that KCNQ1 striations do not necessarily represent t-tubule localization. Instead, they suggest that KCNQ1 resides in the z-line region, probably in an intracellular compartment.

Verification of KCNQ1 and KCNE1 Immunofluorescence Patterns in Guinea Pig Ventricular Myocytes
The above data suggest that, in guinea pig ventricular myocytes, KCNQ1 and KCNE1 are often not colocalized. If confirmed and extrapolated to human ventricular myocytes, this observation has far-reaching implications in terms of the control of I Ks current amplitude and KCNQ1 channel function. Although we have validated our Abs for immunofluorescence application (Fig. 1A), one remaining concern was that Abs might bind to proteins that were unrelated to KCNQ1 or KCNE1 but had three-dimensional conformations similar to the epitopes. Therefore, we further tested the distribution patterns of KCNQ1 and KCNE1 in guinea pig ventricular myocytes without the need for Abs. We used a KCNQ1 construct with enhanced GFP ligated to its C terminus (Q1-GFP (17)) and a KCNE1 construct with mono-dsRed ligated to its C terminus (E1-dsR). By using adenoviruses to introduce these constructs into guinea pig ventricular myocytes, we could directly and simultaneously monitor their distribution patterns with fluorescence microscopy.
Validating Fluorescent Protein-tagged KCNQ1 and KCNE1-Q1-GFP and E1-dsR expressed in both COS-7 cells (Fig. 3A) and NRVMs (data not shown) could assemble into functional I Ks channels. To test whether the attached enhanced GFP and mono-dsRed (225 and 239 amino acids, respectively) might interfere with their ability of trafficking, we compared the distribution patterns of E1-dsR and E1-HA (HA epitope attached to the C terminus of KCNE1) in the presence of KCNQ1. We also compared the distribution patterns of Q1-GFP and Q1-V5 (V5 epitope attached to the C terminus of KCNQ1) in the presence KCNE1. The small HA and V5 epitopes (9 and 14 amino acids) should have a minimal impact on their trafficking ability. In the COS-7 expression system, E1-dsR and E1-HA had the same distribution pattern. They were colocalized in vesicles and on the cell surface. Q1-GFP and Q1-V5 also had the same distribution pattern. They were colocalized in the ER and on the cell surface. These data confirm that attaching fluorescent pro-teins to the C termini of KCNQ1 and KCNE1 did not interfere with their ability to traffic in COS-7 cells.
However, the distribution patterns of KCNQ1 and KCNE1 in COS-7 cells were distinctly different from those in cardiac myocytes. To test whether this was due to differences in the cellular environment, e.g. sarcomere organization in cardiac myocytes but not in COS-7 cells, we further examined the distribution patterns of Q1-GFP and E1-dsR in NRVMs. NRVMs in culture gradually develop ordered sarcomeres (22), allowing us to compare the distribution patterns of Q1-GFP and E1-dsR in a myocyte environment before and after the development of sarcomere organization. NRVMs were incubated overnight with adenoviruses carrying Q1-GFP and E1-dsR. After removing the viruses, the cells were cultured for varying lengths of time before fixation for confocal experiments. In NRVMs without clear sarcomere organization (e.g. 3-h myocyte in Fig. 3B), Q1-GFP was detected in the ER tubular network (colocalization with the ER marker calnexin), whereas E1-dsR was detected in vesicles. Their patterns were similar to those of KCNQ1 and KCNE1 in COS-7 cells. In NRVMs that had developed ordered sarcomeres (e.g. 12-h myocyte in Fig. 3B), the Q1-GFP along with calnexin was partially organized into striations. E1-dsR and Q1-GFP were both detected on the surface of the 12-h myocyte. These observations suggest that the KCNQ1 distribution pattern depended on cellular architecture; i.e. it manifested striations in cardiac myocytes that had organized sarcomeres.
Confirming the KCNQ1 and KCNE1 Distribution Patterns in Guinea Pig Ventricular Myocytes-To reach a suitable expression level of Q1-GFP and E1-dsR in adult guinea pig ventricular myocytes, we needed to culture the myocytes for a total of 3 days. Many, although not all, cultured guinea pig ventricular myocytes maintained their architecture, as shown by the similar distribution patterns of subcellular compartment markers to those in freshly isolated myocytes (data not shown).
We examined myocytes from six animals (independent procedures of cell isolation, culture, and adenoviral infection), focusing on myocytes that had an elongated shape with striations. Similar results were obtained in a total of 65 myocytes, and representative images are shown in Fig. 4. Q1-GFP expressed in guinea pig ventricular myocytes clearly exhibited a transverse striation pattern with a spacing of 1.7 Ϯ 0.2 m, whereas E1-dsR was present on the lateral cell surface as well as in vesicles (Fig. 4A). The Q1-GFP striations registered with those of RyR2, ␣-actinin, and calnexin (Fig. 4B, a-c).
The distribution patterns of Q1-GFP and E1-dsR were reminiscent of the main features of native KCNQ1 and KCNE1 in guinea pig ventricular myocytes detected by immunofluorescence ( Figs. 1 and 2B). Q1-GFP and E1-dsR also showed a strong presence in the perinuclear region, where the rough ER and Golgi apparatus were. In addition, E1-dsR was detected in prominent vesicular compartments, similar to what was seen in COS-7 and NRVM expression (Fig. 3). These reflected the high level of protein translation and trafficking following adenovirus-mediated gene expression. We conclude that the antibodies faithfully detected native KCNQ1 and KCNE1 in guinea pig cardiac myocytes, and, indeed, KCNQ1 and KCNE1 are not well colocalized in the ventricular myocytes.

Functional Consequence of Variations in KCNQ1/KCNE1 Colocalization in Cardiac Myocytes
The degree of KCNQ1/KCNE1 colocalization in the cell periphery was higher in guinea pig atrial than ventricular myocytes (Fig. 5, A and B). To further test whether KCNQ1/KCNE1 colocalization was correlated with assembly into I Ks channels, we tested whether fluorescently labeled KCNQ1 and KCNE1 could engage in FRET. FRET between a fluorescent donor and acceptor occurs when the fluorophores are within 20 -140 Å from each other with proper orientation. The degree of FRET declines as the sixth power of the distance between the fluorophores (23). It has been reported that the fluorescent donor and acceptor attached to the C termini of KCNQ1 and KCNE1, respectively, can engage in FRET when the tagged KCNQ1 and KCNE1 form a functional I Ks channel (24). We labeled native KCNQ1 with Alexa Fluor 488 (as FRET donor) and native KCNE1 with Alexa Fluor 568 (FRET acceptor) and compared the Alexa Fluor 488 signal intensities before and after photobleaching Alexa Fluor 568. If there was FRET between the two, the Alexa Fluor 488 signal would increase after Alexa Fluor 568 had been bleached. Fig. 5C depicts examples of Alexa Fluor 488 and Alexa Fluor 568 fluorescence images from guinea pig atrial and ventricular myocytes before and after photobleaching. The time courses of changes in Alexa Fluor 488 fluorescence signals before and after photobleaching are shown in Fig. 5D. Photobleaching Alexa Fluor 568 led to an increase in Alexa Fluor 488 fluorescence intensity, and the degree of increase was significantly higher in atrial than ventricular myocytes (5.2 Ϯ 0.2% versus 2.9 Ϯ 0.2%, calculated for scan 3). The latter was higher than that seen in reference (REF, reporting changes in the Alexa Fluor 488 fluorescence signal without photobleaching Alexa Fluor 568, 0.8 Ϯ 0.2%). We conclude that a higher degree of cell periphery KCNQ1/KCNE1 colocalization corresponds to a higher degree of KCNQ1/KCNE1 assembly into I Ks channels in atrial than in ventricular myocytes.
We patch-clamped 24 atrial and 21 ventricular myocytes isolated from seven guinea pigs. To isolate I Ks from other overlapping currents, the cells were bathed in sodium-, potassium-, and calcium-free Tyrode's solution containing 3 M E4031 (Fig.  6A). The I Ks current density in atrial myocytes was significantly higher than that in ventricular myocytes (estimated maximal tail currents recorded at Ϫ50 mV, 5.4 Ϯ 0.4 versus 2.2 Ϯ 0.2 pA/pF, p Ͻ 0.05). Although the I Ks current densities differed markedly between atrial and ventricular myocytes, the I Ks gating kinetics was similar in terms of voltage-dependence of activation, rates of activation, and deactivation (Fig. 6, B-E).
We further quantified the KCNQ1 and KCNE1 protein levels in whole tissue lysates prepared from five regions of guinea pig For each myocyte, the relationship between tail current amplitude (I tail ) and test pulse voltage (V t ) was fit with a simple Boltzmann function as described in Fig. 3A. The fraction of channel activated (I tail /I max ) was plotted against V t . C, current traces elicited by an "envelope test" protocol. Test pulse durations ranged from 50 -3000 ms, and the interpulse interval was 10 s. nA, nanoampere. D, time course of I Ks activation constructed from the growth of tail currents during the envelope test protocol as shown in C. For each myocyte, the relationship between tail current amplitude (I tail ) and test pulse duration (t) was fit with a single exponential function with a delay, d: I tail ϭ I ss (1- DECEMBER  hearts. Fig. 7 shows that the KCNQ1 protein levels were modestly lower in ventricles than in atria. There were no statistically significant differences in the KCNE1 protein levels among the five regions. Together, these data suggest that I Ks current amplitude is determined not only by the expression levels of KCNQ1 and KCNE1 but also by the degree of their colocalization on the cell surface.

Mechanism for Dynamic KCNQ1 Distribution Patterns in Cardiac Myocytes
Ventricular myocytes isolated from the same guinea pig heart were incubated in KB (containing 0.5 mM EGTA without added calcium, essentially calcium-free) or medium 199 (containing 1.8 mM calcium) for 8 h before fixation and processing for confocal microscopy. Although KCNE1 maintained stable lateral surface localization under both conditions, KCNQ1 distribution switched from mainly pattern 1 in KB to mainly pattern 2 in medium 199 (Fig. 8A). In pattern 2 ventricular myocytes, most of the KCNQ1 vesicles appeared attached to the striations or in their vicinity, whereas in pattern 3 ventricular myocytes, the KCNQ1 vesicles were clustered to the cell periphery (Fig. 8B). To understand what might have caused the dynamic distribution patterns of KCNQ1, we probed the origin and destination of KCNQ1 vesicles in cardiac myocytes.
KCNQ1 and Endocytic/Recycling Vesicles in Cardiac Myocytes-We compared the immunofluorescence signals of KCNQ1 with those of markers for vesicles in the endocytic/ recycling pathways: early endosome antigen 1 (EEA1, a marker of early endosomes), M6PR (late endosomes), lysosome-associated membrane protein 1 (LAMP1, lysosomes), Rab5 (early endosomes), and Rab11 (recycling endosomes) (Fig. 8C). On the basis of analysis of the central z-plane images, we detected 4.1 Ϯ 0.7 KCNQ1 vesicles/guinea pig ventricular myocyte that were positive for EEA1, 2.2 Ϯ 0.3 KCNQ1 vesicles positive for M6PR, and 0.8 Ϯ 0.2 KCNQ1 vesicles positive for LAMP1 (n ϭ 29, 30, and 30 myocytes, respectively, from two animals). We could not detect any overlap between KCNQ1 vesicles and Rab5 or Rab11. Therefore, only a minute fraction of KCNQ1 vesicles represented endocytosed KCNQ1 in the early or late endosomal compartment. The majority of KCNQ1 vesicles originated from somewhere else.
KCNQ1 and SR Vesiculation in Cardiac Myocytes-In pattern 1 myocytes, KCNQ1 striations registered with those of the ER/SR marker calnexin (Fig. 2B, c). In pattern 2 ventricular myocytes, both KCNQ1 and calnexin were in prominent vesicles and dim in striations (Fig. 8D, left panel). In pattern 3 ventricular myocytes, both KCNQ1 and calnexin clustered to the cell periphery (Fig. 8D, right panel). It has been reported that, in rat basophilic leukemia (2H3) and mouse fibroblast (3T3) cells, elevating [Ca 2ϩ ] i with ionomycin causes reversible ER fragmentation (25). The striking correlation between KCNQ1 and calnexin in guinea pig ventricular myocytes provided a clue that KCNQ1 vesicles might originate from SR vesiculation and [Ca 2ϩ ] i elevation might be the trigger. It is possible that, during myocyte isolation, the cells experienced different degrees of "assault" that led to [Ca 2ϩ ] i elevation, producing KCNQ1 distribution patterns 1, 2, and 3. This scenario is consistent with the data shown in Fig. 8A.
[Ca 2ϩ ] i elevation might not be the only trigger of KCNQ1/SR vesiculation, as was suggested by live cell imaging of NRVMs expressing Q1-GFP. Fig. 9A and supplemental video 1 show that, in a beating NRVM (cyclic rise in [Ca 2ϩ ] i ), Q1-GFP was in the ER tubular network (first scan). However, Q1-GFP vesicu- Data analysis was the same as that described for KCNQ1. One-way analysis of variance, p ϭ 0.778. lated after repeated laser scans. This process was readily reversible (Q1-GFP resumed ER tubular network morphology) after stopping laser scanning for 3 min. Thus, in NRVMs, the cyclic rise in [Ca 2ϩ ] i was not sufficient to induce Q1-GFP vesiculation. Q1-GFP vesiculation occurred when the cell was further stressed by laser scanning, which produced reactive oxygen species.
Direct Observation of Simultaneous Q1-GFP and ER Vesiculation in Live COS-7 Cells-To test whether Q1-GFP and ER simultaneously vesiculated in response to stress ([Ca 2ϩ ] i elevation or oxidative stress), we switched to the COS-7 expression system because these robust cells could sustain repeated laser scans for long durations. Fig. 9B, a- (11) did not induce Q1-GFP or ER vesiculation: hypo-osmotic cell swelling (50% reduction of bath solution osmolarity) and ␤-adrenergic stimulation (isoproterenol, 1 M, on the coexpressed ␤1-adrenergic receptor).
KCNQ1 Vesiculation and Cell Surface Expression in Response to Angiotensin II Receptor (AT1R) Stimulation-AT1R stimulation leads to [Ca 2ϩ ] i elevation (through the inositol 1,4,5trisphosphate receptor) and production of reactive oxygen species (through NADPH oxidase) (26). Fig. 9B, d, and supplemental video 4 show that stimulating AT1R coexpressed with To test whether this occurred with native KCNQ1, we incubated guinea pig cardiac myocytes with angiotensin II (1 M, 60 min). This induced a translocation of KCNQ1 from cytosolic striations to the cell periphery in ventricular myocytes (Fig. 9C). On the other hand, in guinea pig atrial myocytes where native KCNQ1 was already high on the cell surface, incubation with angiotensin II did not induce a further increase.

KCNQ1 Striations and Endo(sarco)plasmic Reticulum-We
propose that, in most myocytes, KCNQ1 striations represent an intracellular compartment close to the Z-lines. On the basis of the tight correlation between KCNQ1 and calnexin, we further propose that this KCNQ1 striation compartment is part of the ER/SR. Calnexin has been used as a specific marker for the ER/SR. Colocalization between membrane channels and calnexin has been interpreted as channel retention in the ER because of misfolding or trafficking defects. However, the ER/SR is an extensive, highly dynamic structure and is compartmentalized to fulfill its many functions (27,28). In adult guinea pig ventricular myocytes, the striation pattern of calnexin likely reflects intimate contacts between the calnexinpositive ER/SR domain and the t-tubule membranes (27).
KCNQ1 Vesicles and Trafficking in Response to Stress-On the basis of the data presented in Figs. 8 and 9 (and in supplemental videos 1-4), we suggest the following scenario (schematic in Fig. 10). In unstressed guinea pig ventricular myocytes, KCNQ1 is mainly in the cytosolic ER/SR compartment. When ventricular myocytes are stressed ([Ca 2ϩ ] i elevation and/or oxidative overload), KCNQ1 vesiculates, along with partial and reversible ER/SR fragmentation. The KCNQ1 vesicles travel to the cell periphery, where KCNQ1 can be delivered to the cell surface and assemble with KCNE1 to form functional I Ks channels. Cell surface KCNQ1 can reenter the cytosolic compartment via vesicles, consistent with the observation that some of the KCNQ1 vesicles carried markers for early and late endosomes (EEA1 and M6PR).
Differential Response to Stress between Guinea Pig Atrial and Ventricular Myocytes-Although we observed similar variations in KCNQ1 distribution patterns in guinea pig atrial and ventricular myocytes (Fig. 1E), there was a much higher KCNQ1 presence on the surface of atrial than ventricular myocytes (F). In situ atrial myocytes may be chronically stressed by mechanical stretch (because of the thin atrial wall) and by oxidative overload (because of high oxygen tension in the left atrium). This may lead to a situation similar to pattern 3 ventricular myocytes. This is consistent with the observations that KCNQ1 distribution in atrial myocytes was relatively insensitive to Ang II incubation (Fig. 9C).
Comparison with Previous Studies-The major distribution pattern of KCNQ1 in atrial and ventricular myocytes is transverse striations (11)(12)(13)(14)(15), interpreted as KCNQ1 localization in t-tubules (21). In the majority (50 -60%) of guinea pig atrial and ventricular myocytes, we observed a KCNQ1 distribution pattern as transverse striations, with a spacing of 1.9 Ϯ 0.1 m. Our observation is consistent with the literature. However, the interpretation differs. Fig. 2A shows that KCNQ1 was in clear striations even when there was no t-tubule, indicating that the striation pattern does not necessarily mean t-tubule localization.
It has been shown that stress-induced activation of serumand glucocorticoid-inducible kinase 1 (SGK-1) can increase the I Ks amplitude by facilitating exocytosis of KCNQ1 vesicles (29), although such KCNQ1 vesicles have not been described for cardiac myocytes. Furthermore, elevating [Ca 2ϩ ] i can increase I Ks in guinea pig ventricular myocytes (30). One of the mechanisms is calcium binding to the calmodulin molecules associated with the cytoplasmic C-terminal domain of KCNQ1, facilitating KCNQ1 trafficking to the cell surface (31). The stress-induced transition of KCNQ1 distribution from pattern 1 to pattern 3 reported here may underlie the above observations.
There is a precedent for KCNQ1 redistribution in response to stimulus. In resting gastric parietal cells, KCNQ1 is mainly in an intracellular compartment. Cell activation leads to KCNQ1 translocation to the cell surface (32).
Is Cell Surface KCNE1 Partnered with ERG?-It has been shown in heterologous expression experiments that KCNE1 can associate with human ERG to modulate its current amplitude (33). To test this possibility in guinea pig ventricular myocytes, we simultaneously monitored the three-dimensional distribution patterns of guinea pig ERG and KCNE1 immunofluorescence along with WGA (surface and t-tubule marker). There was a partial overlap between ERG and KCNE1 on the lateral cell surface (Fig. 11A). Furthermore, Alexa Fluor 488labeled ERG and Alexa Fluor 568-labeled KCNE1 could engage in apFRET (Fig. 11B). These observations suggest that, in guinea pig ventricular myocytes, cell surface KCNE1 can assemble with KCNQ1 or ERG. Because KCNQ1 and ERG may be colocalized in cardiac myocytes (34), it is possible that KCNE1 may switch its partnership between these two K channels. Whether, and under what conditions, this happens in cardiac myocytes requires further investigation.
KCNQ1, KCNE1, other KCNE Subunits, and I Ks Function-I Ks functions as a repolarization reserve in human ventricular myocytes. Our study suggests a previously unrecognized mechanism whereby the I Ks amplitude in ventricular myocytes is maintained at a low level until it is needed. The I Ks function in atrial myocytes is less clear. Because of the atria-specific expression of ultra-rapid delayed rectifier (I Kur ) channels, atrial myocytes do not need a repolarization reserve. We observed a high cell surface KCNQ1/KCNE1 colocalization and high I Ks density in guinea pig atrial myocytes under basal conditions. If this can be extrapolated to human atrial myocytes, I Ks may be a risk factor for atrial fibrillation. This is in keeping with the observations that genetic variations leading to "gain of function" in I Ks and atrial action potential shortening are a common mechanism underlying diverse familial atrial fibrillation syndromes (35).
We propose a scenario of a dynamic relationship between native KCNQ1 and KCNE1 in ventricular myocytes. This is consistent with the transient nature of the KCNQ1/KCNE1 partnership observed in COS-7 experiments (6). This raises the possibility that KCNQ1 may associate with other members of the KCNE family in ventricular myocytes. In particular, KCNE2 is expressed in ventricles of humans and guinea pigs (7). In guinea pig ventricular myocytes, KCNE2 is distributed in cytosolic striations as well as on the lateral surface, i.e. it is better colocalized with KCNQ1 than KCNE1. It is possible that, in ventricular myocytes, KCNE2 can associate with KCNQ1 to suppress the I Ks amplitude (15).