Interactions between hERG and KCNQ1 α-subunits are mediated by their COOH termini and modulated by cAMP

Abstract

KCNQ1 and hERG encode the voltage-gated potassium channel α-subunits of the cardiac repolarizing currents I_Ks and I_Kr, respectively. These currents function in vivo with some redundancy to maintain appropriate action potential durations (APDs), and loss-of-function mutations in these channels manifest clinically as long QT syndrome, characterized by the prolongation of the QT interval, polymorphic ventricular tachycardia, and sudden cardiac death. Previous cellular electrophysiology experiments in transgenic rabbit cardiomyocytes and heterologous cell lines demonstrated functional downregulation of complementary repolarizing currents. Biochemical assays indicated direct, protein-protein interactions between KCNQ1 and hERG may underlie the interplay between I_Ks and I_Kr. Our objective was to investigate hERG-KCNQ1 interactions in the intact cellular environment primarily through acceptor photobleach FRET (apFRET) experiments. We quantitatively assessed the extent of interactions based on fluorophore location and the potential regulation of interactions by physiologically relevant signals. apFRET experiments established specific hERG-KCNQ1 associations in both heterologous and primary cardiomyocytes. The largest FRET efficiency (E_f; 12.0 ± 5.2%) was seen between ion channels with GFP variants fused to the COOH termini. Acute treatment with forskolin + IBMX or a membrane-permeable cAMP analog significantly and specifically reduced the extent of hERG-KCNQ1 interactions (by 41 and 38%, respectively). Our results demonstrate direct interactions between KCNQ1 and hERG occur in both intact heterologous cells and primary cardiomyocytes and are mediated by their COOH termini. Furthermore, this interplay between channel proteins is regulated by intracellular cAMP.

i_kr and i_ks are two distinct currents critically involved in ventricular cardiomyocyte repolarization (32). Their respective pore-forming α-subunits, hERG and KCNQ1 (also called KVLQT1), are encoded for by the genes KCNH2 and KCNQ1. To develop I_Ks current, KCNQ1 coassembles with KCNE1 (minK) (4, 30), whereas hERG may associate with KCNE1 or KCNE2 β-subunits (1). Aberrant channel structure or function can disrupt currents through multiple cellular and biophysical mechanisms, leading to long QT syndrome (LQTS), which can result in syncope, polymorphic ventricular tachycardia, and sudden cardiac death (SCD) (10, 31, 34).

In vivo, I_Kr and I_Ks function together, with some redundancy, to maintain appropriate action potential durations (APDs) under various physiological conditions, following the concept of repolarization reserve (28). The complex interplay between the two channels plays a significant role during times of exertion or stress when the compensatory increase in I_Ks following elevated sympathetic tone becomes critical for cardiomyocyte repolarization (18, 23, 29). Our recent cellular electrophysiological studies in transgenic LQT rabbit cardiomyocytes and stable, heterologous cell lines demonstrated a reciprocal, functional downregulation between I_Kr and I_Ks (7, 27). Co-expression of wild-type (wt) or dominant-negative pore mutants of KCNQ1 dramatically reduced hERG currents and vice versa (7, 27).

Data supporting hERG-KCNQ1 interactions have been previously reported (5, 11, 16) and highlighted the capability of KCNQ1 to facilitate and enhance membrane expression of hERG channels. Co-expression of KCNQ1 increased the current density of hERG, and biochemical assays suggested a direct interaction between KCNQ1 and the COOH terminus of hERG (11). Follow-up studies focused on the role of KCNQ1 in rescuing several trafficking-deficient hERG mutants through direct interactions between the α-subunits (5, 16). Evidence for direct, reciprocal hERG-KCNQ1 interactions was further strengthened by co-immunoprecipitation (co-IP) assays showing hERG pulls down KCNQ1 and the reciprocal, KCNQ1, pulls down hERG in heterologous systems (27). Surface plasmon resonance (SPR) experiments identified a COOH-terminal portion of hERG [amino acid (aa) 771–870] capable of binding the full length COOH terminus of KCNQ1 (aa 349–676) (27). Notably, this specific region of hERG aligns with the putative cyclic nucleotide-binding domain (CNBD).

Although there is potential for physiologically relevant, direct interactions between KCNQ1 and hERG, little is known about the involved mechanisms that regulate these interactions. Previous experimental approaches were also limited, since electrophysiology examines the interplay between the currents but not the channel proteins per se. Additionally, standard biochemical assays require removal of channel polypeptides from their native plasma membrane environment, rendering them unsuitable for quantitative, dynamic, physiologically relevant studies of protein-protein interactions. Our objective was to investigate hERG-KCNQ1 interactions in the intact cellular environment primarily through acceptor photobleach FRET (apFRET) experiments. Our results suggest that KCNQ1 and hERG form direct interactions mediated by their COOH termini and modulated by intracellular cAMP. Because interplay between K⁺ currents is critical in maintaining APDs, alterations in hERG-KCNQ1 interactions could be important in controlling repolarization reserve, which may prove significant in a variety of clinical phenotypes including LQT syndrome, heart failure, and sudden cardiac death.

MATERIALS AND METHODS

Fluorescent protein fusions.

Fusions of hERG, KCNQ1, and Kir2.1 α-subunits to green fluorescent protein variants (XFPs) were created via PCR-based cloning techniques using Expand High-Fidelity PCR system (Roche Applied Science, Indianapolis, IN) and human cDNA templates. Products were subcloned in to the multiple cloning site of pEXFP-N1 or -C1 vectors (Clontech, Mountain View, CA) using the restriction sites HindIII/BamHI (KCNQ1 and hERG) or SacI/BamHI (Kir2.1). Before subcloning, the amino acid mutation A206K was introduced into the XFP sequence to create monomeric XFPs (mXFP), thus eliminating interactions between fluorophores that may contribute to FRET signals (36). Amino acid (aa) linkers of between 14 and 18 aa connect the ion channels and XFP proteins. Positive control constructs consisted of a cytosolic CFP-YFP fusion (22-aa linker) and a membrane-targeted KCNQ1-CFP-YFP fusion (3- and 5-aa linkers, respectively). The entire regions encompassing ion channel subunits, linkers, and XFPs were sequence verified.

Cell culture and transfection.

Human embryonic kidney 293 (HEK) and Chinese hamster ovary (CHO) cells were cultured in 1:1 DMEM/F-12 with 10% FBS, and 1% pen-strep under standard conditions. For imaging, HEK cells were plated 24 h before transfection on coverslips in tissue culture-treated six-well plates at a density of 250,000/well. For patch clamping, cells were plated in 35-mm dishes at 200,000 (HEK) or 125,000 (CHO) cells/dish. Transient transfection was accomplished following product literature using a FuGENE6:DNA ratio (Roche Applied Science) of 3 μl:1 μg. Co-transfections for FRET or co-expression of KCNE1 + KCNQ1 used equal amounts (1 μg) of each plasmid. For unlabeled ion channel constructs, cytosolic GFP vectors were used as a transfection marker in a 1:10 ratio of GFP to channel protein.

Intracellular cAMP was elevated through the application of 100 μM forskolin (stock solution in DMSO; Sigma, St. Louis, MO) with 100 μM IBMX (stock solution in dH₂O; Sigma) in cell media or via exposure to 500 μM pCPT-cAMP [8-(4-chlorophenylthio)adenosine 3′,5′-cyclic monophosphate sodium; stock solution in dH₂O; Sigma], a membrane-permeable cAMP analog, with 100 μM IBMX in cell media. HEK cells were incubated at 37° in treatment media for either 5 or 30 min before fixation.

To obtain primary cardiomyocytes, neonatal New Zealand White rabbit kits (both sexes, 3–5 days old) were administered pentobarbital sodium (65 mg/kg) and heparin (1,000 U/kg). All animal experiments and procedures were approved by the Rhode Island Hospital Institutional Animal Care and Use Committee. After anesthesia, the heart was excised, mounted on a Langendorff system, and perfused with a Ca²⁺-free solution containing (in mM) 140 NaCl, 4.4 KCl, 1.5 MgCl₂, 0.33 NaH₂PO₄, 16 taurine, 5 HEPES, 5 pyruvic acid, and 7.5 glucose for 5 to 7 min. Then, 0.3 mg/ml collagenase I (Worthington Biochemical, Lakewood, NJ) was added to the solution and perfused for 4–5 min at a flow rate of 3 ml/min. Ventricles were separated and minced with forceps in prewarmed KB-EDTA (Boston Bioproducts, Ashland, MA) containing (in mM) 45 KCl, 65 K-glutamate, 3 MgSO₄, 15 KH₂PO₄, 16 taurine, 10 HEPES, 0.5 EDTA, 10 glucose, and 1% BSA (Sigma) (pH 7.3). The cell suspension was filtered through a 100-μm cell strainer, and cells were allowed to settle for 1 h at room temperature, yielding ∼30 million cells/heart. Then, 75% of the supernatant was removed, and the remaining suspension was centrifuged at 300 rpm for 1 min. The cell pellet was resuspended in Ca²⁺-free MEM (Lonza, Hopkinton, MA) and gradually acclimated to a final Ca²⁺ concentration of 1.8 mM over the course of 1 h. Cells were plated on coverslips coated with laminin (10 μg/ml) at ∼100,000 cells/cm². Three hours postplating, media was replaced with culture media (DMEM, 7% FBS, l-glutamine, with or without 1% pen-strep, Life Technologies, Carlsbad, CA), and 100 μM BRDU (Sigma) was added. Cells were maintained under standard culture conditions and were fed with culture media + BRDU every other day. Transient transfections were performed in antibiotic-free media after 6 days in culture using Lipofectamine 2000 (Life Technologies) and OptiMEM (Life Technologies) following product literature.

End-point analyses occurred 48 h posttransfection. For imaging studies, transfected cells were rinsed in PBS and fixed in 4% paraformaldehyde for 15 min. For immunolabeling of α-sarcomeric actinin or vimentin in neonatal rabbit cardiomyocytes (NRbCMs), samples were blocked and permeabilized for 30 min in 3% BSA PBS with 0.1% Triton X. Mouse (actinin) or Rat (vimentin) monoclonal antibody was applied for 1 h (1:1,500 actinin, or 1:50 vimentin, in 3% BSA PBS), and Alexa Fluor-labeled secondary antibodies were used (1:1,000 in 3% BSA PBS; Life Technologies). Coverslips were then rinsed three times in PBS and mounted in ProLong Gold antifade reagent (Life Technologies).

Biochemical analysis.

For immunoblotting, cells were lysed and harvested using standard protocols (27). In brief, 48-h posttransfection cells were harvested on ice in cold lysis buffer containing radioimmunoprecipitation assay buffer (RIPA; Boston BioProducts, Ashland, MA), Complete Mini proteinase inhibitor tablets (Roche Applied Science), 1 mM PMSF, and 5 mM DTT. Lysis was allowed to proceed for 15 min at 4° under nutation. Tubes were spun at 10,000 g for 15 min at 4°, and the supernatant was reserved. Protein extract (25 μg) was run in 10% acrylamide gels via SDS-PAGE, transferred onto nitrocellulose membranes, and blocked in 5% nonfat milk in PBS. Primary antibodies for KCNQ1 (rabbit polyclonal, Alomone Labs, Jerusalem, Israel), hERG (rabbit polyclonal, Alomone Labs), and GAPDH (mouse monoclonal, Sigma) were diluted in 5% milk-PBS. Membranes were washed five times over 30 min in Tween-PBS before exposure to HRP-linked secondary antibodies and signal detection using chemiluminescent reagents (HyGLO, Denville Scientific, Metuchen, NJ).

Cellular electrophysiology.

Whole-cell patch-clamp recordings were performed in CHO and HEK cells using an Axopatch-200B amplifier (Axon Instruments, Foster City, CA). Pipette resistances were kept between 2 and 4 MΩ. Pipette solution contained (in mM) 50 KCl, 65 glutamate-K, 5 MgCl₂, 5 EGTA, 10 HEPES, 5 K2-ATP, and 0.25 Na-GTP, pH = 7.2 with KOH. Extracellular bath solution contained (in mM) 140 NaCl, 5.4 KCl, 1 MgCl₂, 1 CaCl₂, 0.33 NaH₂PO₄, 5 HEPES, and 7.5 glucose, pH = 7.3 with NaOH. Recordings were taken at room temperature and from cell lines with low (below 20) passage numbers. Membrane potential was held at −80 mV, with test potentials from −70 to +40 mV for hERG channels and from −50 to +70 mV for I_Ks channels. Tail currents for all cells were recorded at a test potential of −60 mV. Average I/V relationships were used to compare current densities following normalization of current traces to individual cell capacitance measurements (Fig. 1C). Steady-state I/V relationships were measured at the end of the test pulse.

Fig. 1.

Expression of potassium channel fusion constructs. A: representative confocal images of constructs expressed in fixed human embryonic kidney (HEK) cells (top). The dashed square on the left is magnified in the right image. Arrows indicate the intracellular aggregation of mYFP-KCNQ1 not seen with other fusion constructs. Bottom: immunoblots of total protein from HEK cell lysates show expression levels for mXFP-labeled and -unlabeled channel constructs. Lanes 1–2: untransfected cells. Lanes 3–4: unlabeled, wild-type channels. Lanes 5–6: mXFP-fusion constructs. B: representative current traces from electrophysiological characterization of hERG and KCNQ1 fusion constructs. hERG currents were assessed in HEK cells, and I_Ks currents (KCNQ1 and KCNE1 expression) were measured in Chinese hamster ovary cells (CHOs). C: steady-state current and peak tail current I/V relationships for wild-type and mXFP fusions of hERG (top) and KCNQ1 + KCNE1 (bottom).

Confocal imaging and acceptor photobleach FRET.

Images of fusion construct expression and localization were taken confocally (1 Airy unit) on a Zeiss LSM 510 Meta microscope using a ×63 Plan-Apochromat, NA 1.4 objective. Acceptor Photobleach FRET (apFRET) experiments were performed on the same system as described previously (12), with empirical modifications. In brief, mCFP and mYFP were excited with 458- and 514-nm light, and transmission percentages were held constant for each laser line. Twelve-bit, 512 × 512 images were obtained with a pinhole of 5 Airy units, corresponding to an image depth of <3.2 μm. Following the method of Karpova (19), a 14.3-μm² region of interest (ROI) was monitored throughout a time series of 100 scans (14.8 s). Within this area, we monitored circular ROIs, each 2.9 μm in diameter, representing bleach, control, and background ROIs. Photobleaching of the acceptor molecule (mYFP) was accomplished using the 514 laser line at 100% transmission for 250 rapid scans of only the bleach ROI, requiring 5.3 s. In cells expressing mYFP, the bleach depth averaged 74%. The Zeiss AIM physiology software quantified the mean intensity in each region over time, which was plotted as raw data curves. The efficiency of energy transfer (E_f) was calculated as E_f = (I_postbleach-I_prebleach) × (100%/I_postbleach) (Eq. 1), where I is the mCFP (donor) intensity from the raw data in the bleach ROI for a given time point. Analogous equations used data from the control ROI to provide an internal control for false-positive FRET efficiencies (C_f) in each cell.

Statistical analysis.

Cell pools were constructed across multiple transfections with individual treatments completed in triplicate. Imaging for each sample set was also pooled across several sessions. All data are presented as means ± SD. Statistical significance (P < 0.05) was determined by Student's t-test when comparing two groups and by ANOVA with Tukey's honestly significantly difference post hoc testing for multiple comparisons using SPSS 19 (IBM, Armonk, NY).

RESULTS

Expression and function of ion channel fusion constructs.

Confocal imaging in fixed HEK cells demonstrated adequate expression of hERG and KCNQ1 fused to mGFP variants at either the NH₂ or COOH terminus. Notably, in a significant number of cells, mYFP-KCNQ1 constructs clustered perinuclearly, possibly indicative of poor trafficking (Fig. 1A). Immunoblots assessed expression levels of hERG or KCNQ1 in cells expressing wild-type (wt) or XFP-fusion constructs. Fusion constructs showed the anticipated shift in molecular weight due to the addition of the ∼28-kDa mXFP.

Complementary cellular electrophysiology showed that all mXFP-fusion constructs form functional channels, which demonstrate the hallmarks of I_Kr or I_Ks currents in heterologous cells (Fig. 1B). Qualitatively, the biophysical properties of the mXFP-fusion channels are similar to wt, demonstrating that gating of the channels is largely intact (Fig. 1C). We did notice some variation in current densities likely due to transfection and/or expression efficiency and potential changes in hERG deactivation kinetics with mXFP insertion as previously reported (13, 14, 17).

FRET-based assessments of potassium channel interactions.

apFRET experiments examined changes in CFP emission in a subcellular region where YFP was photobleached relative to a corresponding control region where no bleaching was performed (Fig. 2A). Initial experiments utilized control constructs to empirically establish the potential range of FRET efficiencies (Table 1). Positive control data demonstrated intramolecular FRET between donor (CFP) and acceptor (YFP) fluorophores directly linked by a short aa chain (5 or 22 residues), which is reflected in representative raw data showing an increase in CFP emission following bleaching of YFP in the KCNQ1-CFP-YFP positive control (Fig. 2B). No significant change in CFP emission was measured in the absence of the YFP acceptor molecule. Histograms of individual E_f and C_f values (Eq. 1) show distinct population separation in FRET efficiencies calculated from bleach and control regions for positive control samples, which is absent for donor-only samples (Fig. 2C). In cells expressing cytosolic CFP-YFP or the positive control fusion KCNQ1-CFP-YFP, E_f values were 22.0 ± 4.0% and 21.4 ± 5.1%, respectively. Cells solely expressing negative, donor-only control CFP fusions provided an assessment of false-positive FRET. For mCFP-hERG and hERG-mCFP, FRET efficiencies were 5.1 ± 4.8% and 6.4 ± 5.6%, respectively. The largest C_f value for all controls was 1.1 ± 5.4% (Fig. 2D).

Fig. 2.

Acceptor photobleach FRET (apFRET) methodology and results from control experiments. A: representative cell expressing a cytosolic CFP–YFP fusion. A square region of interest (ROI) (top) is imaged during apFRET experiments (bottom). Two circles represent the experimental, bleach region and also an internal, control region. B: representative raw data for positive (top) and negative (bottom) control experiments. Each data point is the mean intensity per frame of either the bleach or control region of interest (ROI). The increase in CFP intensity in the bleach ROI reflects the FRET efficiency calculated using Eq. 1. C: histograms of FRET efficiency values in bleach and control regions for two positive and two negative control constructs. D: average E_f values demonstrate negative control conditions with no acceptor present are statistically reduced from positive controls as indicated by an asterisk (*P < 0.001). Efficiencies measured in control ROIs (C_f) were similar across all groups (P > 0.8). Values are means ± SD.

Table 1.

Efficiency percentages for all samples evaluated by apFRET

Donor	Acceptor	Treatment	Ef	Cf	n
Positive controls
CFP—YFP			21.96 ± 3.97	0.28 ± 2.84	10
KCNQ1-CFP-YFP			21.43 ± 5.11	−0.30 ± 8.12	39
KCNQ1-CFP-YFP		5 min; vehicle	19.19 ± 5.03	1.84 ± 8.99	29
KCNQ1-CFP-YFP		5 min; forksolin + IBMX	20.81 ± 5.60	2.79 ± 7.28	43
KCNQ1-CFP-YFP		5 min; pCPT-cAMP + IBMX	20.18 ± 5.89	2.23 ± 7.89	29
Negative controls
mCFP-hERG			5.11 ± 4.81	1.05 ± 5.37	39
hERG-mCFP			6.44 ± 5.61	0.08 ± 6.22	49
mCFP	mYFP-hERG		5.51 ± 2.82	2.39 ± 2.94	10
mCFP	mYFP-KCNQ1		5.41 ± 7.35	−0.12 ± 5.69	10
Overexpression controls
hERG-mCFP	mYFP-Kir2.1		6.09 ± 6.12	0.03 ± 7.95	43
hERG-mCFP	Kir2.1-mYFP		5.10 ± 6.55	1.41 ± 7.30	40
Experimental FRET pairs
mCFP-hERG	mYFP-hERG		11.86 ± 5.94	2.28 ± 5.63	35
hERG-mCFP	KCNQ1-mYFP		12.04 ± 5.24	0.30 ± 6.41	39
hERG-mCFP	KCNQ1-mYFP	NRbCM	10.07 ± 5.21	−0.34 ± 7.5	29
hERG-mCFP	KCNQ1-mYFP	5 min; vehicle	9.97 ± 4.92	2.49 ± 8.07	44
hERG-mCFP	KCNQ1-mYFP	5 min; forksolin + IBMX	7.10 ± 7.15	1.63 ± 6.06	45
hERG-mCFP	KCNQ1-mYFP	30 min; forksolin + IBMX	7.89 ± 4.27	0.58 ± 5.90	32
hERG-mCFP	KCNQ1-mYFP	5 min; pCPT-cAMP+IBMX	7.47 ± 4.85	0.10 ± 5.57	30
mCFP-hERG	mYFP-KCNQ1		5.40 ± 5.88	2.43 ± 7.40	32
hERG-mCFP	mYFP-KCNQ1		9.62 ± 5.91	0.66 ± 8.42	33
mCFP-hERG	KCNQ1-mYFP		8.81 ± 3.91	1.85 ± 5.27	37

Values are means ± SD. apFRET, acceptor photobleach FRET; E_f, FRET efficiency; C_f, control for false-positive FRET efficiency.

Qualitatively, relatively uniform overlap was seen when KCNQ1 and hERG FRET pairs were co-expressed (Fig. 3), except when both channels were labeled on the NH₂ terminus (mCFP-hERG + mYFP-KCNQ1; Fig. 3B), corresponding to the perinuclear accumulation seen when mYFP-KCNQ1 was expressed alone (Fig. 1). This observation also held in raw and analyzed FRET data (Figs. 3 and 4). Importantly, a large, significant increase in CFP intensity following YFP bleaching was apparent when both KCNQ1 and hERG were COOH-terminally labeled (Fig. 3A). Increases were also seen in raw traces when one channel was COOH-terminally labeled and the other NH₂-terminally labeled (Fig. 3, C and D).

Fig. 3.

Co-expression of paired KCNQ1 and hERG constructs and raw apFRET data demonstrate differences based on label placement. A–D: FRET pairs are indicated, representing the four possible combinations of fluorophore placement. Top: representative images of HEK cells co-expressing indicated FRET pair. Pseudocoloring is nonstandard (mYFP/KCNQ1 in green and mCFP/hERG in red) to accentuate differences in localization. Of particular note was a relative lack of overlap in B, indicated by arrows. Bottom: representative raw data from single apFRET experiments.

Fig. 4.

apFRET demonstrates significant interactions between KCNQ1 and hERG fusion constructs. A: average FRET efficiencies for overexpression and positive controls as well as the four pairs varying KCNQ1 and hERG fluorophore placement. C and Y indicate mCFP or mYFP fluorophores. Values are means ± SD. *Statistically significant differences in E_f compared with the COOH-terminal-labeled FRET pair (hERG-mCFP + KCNQ1-mYFP) (P < 0.001). †Statistical significance from both overexpression control E_f values (P < 0.001). C_f values were similar among all groups (P > 0.6). B: histograms displaying all analyzed E_f and C_f efficiency values.

Pooled, quantitative FRET analysis revealed the NH₂-terminally labeled pair, mCFP-hERG + mYFP-KCNQ1, had the lowest E_f value (5.4 ± 5.9%) of the four combinations of label positioning (Figs. 3B and 4). This value was commensurate and statistically indistinguishable (P > 0.4) from those of negative controls and of overexpression controls, which measured nonspecific FRET in samples co-expressing two non-interacting channels, hERG and Kir2.1 (Fig. 4A). The largest FRET efficiency (12.0 ± 5.2%) was seen between COOH-terminally labeled channels, hERG-mCFP + KCNQ1-mYFP. This was statistically significant from the NH₂-terminally labeled FRET pair and from overexpression controls (P < 0.001). An additional control evaluated FRET in cells only transfected with hERG but expressing both mCFP and mYFP fusions. This sample predominantly reflects interactions between hERG α-subunits within the same tetrameric channels, and the E_f value (11.9.0 ± 5.9%) was equivalent to that measured between COOH-terminally labeled KCNQ1 and hERG (Fig. 4A). As seen for positive and negative controls (Fig. 2C), histogram analyses showed distinct differences in E_f and C_f populations only for mCFP-hERG + mYFP-hERG and hERG-mCFP + KCNQ1-mYFP samples (Fig. 4B).

FRET demonstrates hERG-KCNQ1 interactions in primary cells.

To test whether hERG-KCNQ1 interactions are present in a more physiologically native environment, we performed apFRET experiments in cultured neonatal rabbit cardiomyocytes (NRbCM). To demonstrate the utility of these cultures in assessing hERG-KCNQ1 interactions, we have extensively characterized this model system (Fig. 5). NRbCM cultures demonstrated hERG-KCNQ1 interactions occurred at a similar level in this primary cell type as in the HEK heterologous system. Electrophysiological measurements in acutely isolated cells measure I_Kr currents in all cells tested, and a subset of cells also express I_Ks current (data not shown). Four to five days postisolation, cultured NRbCMs have a myocyte purity of 80%, as evidenced by double immunolabeling for vimentin and α-sarcomeric actinin (Fig. 5A). Immunoblots of whole cell lysates collected after 6 days in culture show expression proteins indicative of a cardiomyocyte phenotype, including calcium handling proteins, hERG, Na_v1.5, α-sarcomeric actinin, and connexin 43 (Fig. 5B). NRbCMs were successfully transiently transfected with the COOH-terminally labeled FRET pair, which had the strongest interaction levels in HEK cells (Fig. 5C). Both α-sarcomeric actinin positive and negative cells expressed ion channel fusions, but FRET experiments were only performed in cells with defined sarcomeric structures. apFRET experiments in this primary culture system demonstrated an average FRET efficiency between hERG-mCFP and KCNQ1-mYFP of 10.1 ± 5.2% (Fig. 5, D and E).

Fig. 5.

hERG-mCFP + KCNQ1-mYFP expression and interaction in cultured neonatal rabbit cardiomyocytes (NRbCM). A: 4–5 days postisolation, cultures have a myocyte purity of 80%, as evidenced by double immunofluorescent labeling for vimentin (fibroblasts, green) and α-sarcomeric actinin (CMs, red). B: immunoblots of whole cell lysates collected after 6 days in culture demonstrate expression of calcium handling and other proteins indicative of a cardiomyocyte phenotype. CSQ, calsequestrin; Cxn 43, connexin 43; Cav 2.1, L-type Ca²⁺ channel; Nav1.5, SCN5A sodium channel; NCX, sodium-calcium exchanger; SERCA2a, sarco-/endoplasmic reticulum Ca²⁺-ATPase type 2a; PLN, phospholamban. C: representative image of hERG-mCFP and KCNQ1-mYFP transient expression 48 h posttransfection. Before FRET experiments, cells were immunofluorescently labeled with anti-α-sarcomeric actinin, indicating myocyte phenotype. D: raw data from a representative apFRET experiment measuring interactions between COOH-terminally labeled constructs in NRbCMs. E: average E_f values were commensurate between HEKs and NRbCMs (P > 0.1). Values are means ± SD.

Increased intracellular cAMP modulate FRET efficiencies.

Since our FRET data supported previous studies identifying the COOH terminus of hERG and the CNBD in particular as an important motif in hERG-KCNQ1 interactions (11, 27), we questioned whether cAMP may affect the interplay between hERG and KCNQ1 channels. Such regulation could demonstrate the physiological relevance of hERG-KCNQ1 interactions, so we also performed FRET experiments in HEK cells following treatment with 100 μM forskolin + 100 μM IBMX to increase intracellular cAMP as a mimic of elevated sympathetic tone. Cells expressing the positive control construct KCNQ1-CFP-YFP showed no changes in measured FRET efficiencies after either treatment with vehicle (DMSO + H₂O) or after a 5-min treatment with forskolin and IBMX (Fig. 6). However, in cells expressing the COOH-terminally labeled FRET pair hERG-mCFP + KCNQ1-mYFP, elevated cAMP resulted in a significant 35–41% reduction in FRET efficiency. E_f values decreased to 7.10 ± 7.15% and 7.89 ± 4.27% for 5- and 30-min treatment times, respectively. To further demonstrate the potential of cAMP to regulate hERG-KCNQ1 interactions, apFRET experiments were performed after 5 min of treatment with 500 μM pCPT-cAMP (a membrane-permeable cAMP analog) + 100 μM IBMX (Fig. 6B). No change in FRET efficiencies was measured in cells expressing the FRET-positive control construct, but in cells expressing the COOH-terminally labeled FRET pair hERG-mCFP + KCNQ1-mYFP, membrane-permeable cAMP exposure resulted in a significant 38% reduction in E_f (to 7.47 ± 4.85%).

Fig. 6.

Increased intracellular cAMP reduces hERG-KCNQ1 interactions. A: average E_f values for the positive FRET construct KCNQ1-CFP-YFP are unchanged following treatment with forskolin + IBMX (P > 0.6). For the COOH-terminally labeled FRET pair (hERG-mCFP + KCNQ1-mYFP) both 5 and 30 min exposure to forskolin + IBMX significantly reduced E_f efficiencies. *Statistically significant reduction from untreated cells (P < 0.03). C_f values were similar among all groups (P > 0.4). B: average E_f values showed a similar response following 5 min of treatment with a membrane-permeable cAMP analog (pCPT-cAMP) + IBMX. Values for the positive control construct KCNQ1-CFP-YFP were unchanged (P > 0.7), whereas increased cAMP significantly reduced the FRET efficiency measured between the COOH-terminally labeled FRET pair. *Statistically significant reduction from untreated cells (P < 0.01). C_f values were similar among all groups (P > 0.4). Values are means ± SD normalized to their respective untreated average E_f values.

DISCUSSION

Direct, functional interactions between KCNQ1 and hERG have been previously discussed based predominantly on results from electrophysiology experiments and biochemical assays. Initial work suggested coexpression of KCNQ1 increased the current density of hERG by upregulating hERG surface expression (11). KCNQ1 has been proposed to act as a chaperone, aiding hERG trafficking to the membrane through an interaction involving the hERG COOH terminus (aa 681–1,159) (5, 11). This was supported by evidence that trafficking-competent KCNQ1 can rescue trafficking-deficient hERG mutants but not necessarily other mutant types (16). Our previous studies, including cellular electrophysiology data from transgenic long-QT rabbit cardiomyocytes and heterologous cells as well as biochemical assays, demonstrated functional downregulation of the complementary repolarizing channels hERG and KCNQ1 (7, 27). However, other groups, including recent results from pluripotent-induced stem cell models of a trafficking-deficient LQT1 mutation, did not identify reciprocal interactions, possibly because evaluating such interactions was tangential to the primary scientific objectives (11, 23).

In this work, fluorescent microscopy, immunoblot assays, and cellular electrophysiology experiments demonstrated mXFP-fusion channels were functionally expressed in heterologous cells (Fig. 1). Representative traces along with average steady-state and peak tail current I/V curves (Fig. 1C) did reveal some differences in current density and kinetics due to label placement. For hERG channels, fluorophore placement on the NH₂ terminus resulted in increased deactivation rates, whereas a COOH-terminal fluorophore slowed deactivation kinetics. However, interactions between domains in both the NH₂- and COOH-terminal regions have been implicated in the mechanism of characteristic, slow deactivation kinetics in hERG channels, and manipulations of these protein regions have been previously reported to alter the biophysical properties of hERG channels (13, 14, 17). Variations in the magnitude of peak current are likely due to expression variability in transient transfection experiments. For KCNQ1 channels, the β-subunit KCNE1 was co-expressed, allowing for the evaluation of I_Ks current. Similar kinetics and peak currents were seen for both NH₂- and COOH-terminally fused mXFPs, but both peak current and peak tail current were reduced compared with the wild-type channel. This is potentially due to variability in the interaction or stoichiometry between KCNQ1 and KCNE1 subunits as the proteins were concatenated in the wild-type constructs but transiently co-expressed for mXFP-KCNQ1 and KCNQ1-mXFP constructs.

Initial apFRET experiments showed that hERG-KCNQ1 interactions measured in fixed HEK cells were significantly larger than those of negative controls where no acceptor was present and were higher than overexpression controls where two non-interacting proteins (hERG and Kir2.1) were co-expressed. Although our FRET experiments were not designed for structural analyses of protein configurations (22), the largest experimental FRET efficiency measured was between KCNQ1 and hERG subunits with fluorophore fusion at the COOH terminus. Our previously published work biochemically demonstrated a direct interaction between a portion of the hERG COOH terminus encompassing the CNBD and the full length COOH terminus of KCNQ1(27), and the apFRET data lends support to COOH-to-COOH interactions predominating in intact cells. hERG-KCNQ1 interactions mediated by COOH termini are also physiologically interesting since there are numerous mutations in the COOH terminus of both channels that result in human pathophysiologies. For hERG, in particular, the CNBD has been suggested to be critical for proper channel trafficking, and a large percentage of COOH-terminal mutations are trafficking deficient (2, 3). In our FRET data, the lowest experimental E_f value was measured between NH₂-terminally labeled channels (mCFP-hERG + mYFP-KCNQ1). However, it should be noted that NH₂-terminal fusions are prone to trafficking issues, and, for KCNQ1 in particular, confocal imaging demonstrated differences in protein location based on fluorophore placement (Figs. 1 and 3).

Although the absolute measured efficiency values for experimental FRET pairs were low (9–12%), they were within the range previously reported for this experimental design (12, 19). Additionally, the theoretical maximal FRET efficiency possible between two XFPs is <50% due to the chromophore location and physical structure of fluorescent proteins (24, 25), and our optimal positive control KCNQ1-CFP-YFP empirically suggests a maximal E_f of ≈20%.

Our investigations also suggest protein-protein interactions between KCNQ1 and hERG have physiological relevance. We first demonstrated that interactions between COOH-terminally labeled constructs also develop in primary cells. Interactions between KCNQ1 and hERG were measured with apFRET in neonatal rabbit cardiomyocytes, the native environment for the expression and function of these channels. Although measured E_f values were lower (not statistically significantly) in cardiomyocytes compared with HEK cells, this is potentially due to interactions with endogenous channels, since NRbCMs functionally express I_Kr and biochemically demonstrate expression of both core (135 kDa) and fully glycosylated (155 kDa) hERG (Fig. 5B). Interestingly, compared with humans and guinea pigs, rabbits have reduced expression levels of KCNE1 and thus decreased I_Ks currents, making them highly susceptible to I_Kr mutations and blockade (37). Hence, in this model system, the interplay between hERG and KCNQ1 channels is likely complex. However, it is important to note the initial observations of reciprocal, functional downregulation were made in a transgenic rabbit model (7) and were then recapitulated in several in vitro heterologous cell systems (27). Additionally, the establishment of hERG-KCNQ1 interactions in cells with a cardiomyocyte phenotype (sarcomeric structure and expression of cardiomyocyte-specific proteins; Fig. 5B) is an essential first step toward understanding the potential implications of these interactions in vivo.

Next, we examined whether increased cAMP might alter hERG-KCNQ1 interactions. Several lines of evidence led us to select this signaling molecule for initial perturbations in FRET experiments. First, intracellular cAMP levels are a critical component of the β-adrenergic cascade in cardiomyocytes. It is well established that KCNQ1 is affected by sympathetic stimulation through cAMP-/PKA-dependent pathways (21, 26). This results in augmented I_Ks via increased open probability, slower deactivation kinetics, and a shift of the activation curve (15). Although I_Kr was originally thought to be unaffected, more recent data suggest that β-adrenergic stimulation alters the biophysical properties of hERG through both PKA-mediated pathways and direct binding of cAMP to the putative cyclic nucleotide-binding domain (CNBD) in the hERG COOH terminus (8, 9). Our prior biochemical data suggested the hERG CNBD is involved in interactions with KCNQ1 (27), which is interesting in the context of the nominal response of I_Kr to changes in sympathetic tone. The FRET results reported here revealed a statistically significant reduction in interaction efficiencies after 5 or 30 min of treatment with forskolin + IBMX to activate adenylate cyclase and prevent inactivation of cAMP signals. A similar, significant reduction was determined following 5 min of exposure to a membrane-permeable cAMP analog (pCPT-cAMP) + IBMX. From these data, we now postulate that direct binding of cAMP to the hERG CNBD destabilizes hERG-KCNQ1 interactions. Abrogation of interactions may release KCNQ1 from a state of functional downregulation, thus contributing to the increase in I_Ks current density following adrenergic stimulation. However, further studies are needed to delineate the contribution of PKA-mediated effects on hERG-KCNQ1 interactions from direct cAMP-binding, particularly as recent reports of the crystal structures of COOH-terminal domains in other EAG-family channels contend that cAMP does not directly bind to the CNBD (6, 20). It is also possible that both pathways are involved in the mechanisms regulating repolarization reserve. Additional investigations are also necessary to elucidate the functional consequences of cAMP-regulated hERG-KCNQ1 interactions. The interplay between I_Ks and I_Kr currents following increased cAMP may not directly correlate with the destabilization of interactions between the COOH termini of the pore-forming α-subunit proteins. Detailed electrophysiological investigations may find that the effects of cAMP on the functional interactions between KCNQ1- and hERG-encoded currents are more subtle and/or more complex. Nevertheless, the protein-protein interactions described in this work represent a first step in characterizing and understanding the interplay between KCNQ1 and hERG channels.

In summary, we have quantified direct interactions between functional KCNQ1 and hERG α-subunits using apFRET-based microscopy. This approach allows for assessment of direct interactions at the protein level within intact cells, providing an advantage over electrophysiological and biochemical techniques. These experiments demonstrated that hERG-KCNQ1 interactions mediated by the COOH-terminal regions of both channels do, in fact, predominate in the cellular environment of heterologous cells and primary cardiomyocytes. Moreover, we have demonstrated hERG-KCNQ1 interactions are sensitive to a ubiquitous biological signal, intracellular cAMP, as elevation of intracellular cAMP results in a reduction of the measured FRET efficiencies between hERG and KCNQ1. This data suggests that the interplay between the channels may be modulated by cAMP, which could play a role in regulating repolarization reserve. We speculate that cAMP could abolish inhibitory interactions between hERG and KCNQ1, perhaps removing a functional interference from KCNQ1 or enabling enhanced trafficking of KCNQ1 polypeptides to the membrane. Further investigations of this phenomenon may elucidate potential molecular mechanisms that underlie spontaneous arrhythmias and sudden cardiac death across a variety of patient populations and also have the potential to describe a novel, regulated interaction between members of two distinct potassium channel families.

GRANTS

This work was supported by National Heart, Lung, and Blood Institute Grants 5R01 HL-046005-19 (G. Koren) and 5T32 HL-094300-04 (L. E. Organ-Darling and W. Li).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: L.E.O.-D., K.R., W.L., and G.K. conception and design of research; L.E.O.-D., A.N.V., J.R.G., Y.L., K.M., and W.L. performed experiments; L.E.O.-D., A.N.V., Y.L., K.M., and W.L. analyzed data; L.E.O.-D., K.M., K.R., W.L., and G.K. interpreted results of experiments; L.E.O.-D., A.N.V., and K.M. prepared figures; L.E.O.-D. drafted manuscript; L.E.O.-D., W.L., and G.K. edited and revised manuscript; L.E.O.-D., A.N.V., J.R.G., Y.L., K.M., K.R., W.L., and G.K. approved final version of manuscript.