The I_Ks Channel Response to cAMP Is Modulated by the KCNE1:KCNQ1 Stoichiometry

Abstract

The delayed potassium rectifier current, I_Ks, is assembled from tetramers of KCNQ1 and varying numbers of KCNE1 accessory subunits in addition to calmodulin. This channel complex is important in the response of the cardiac action potential to sympathetic stimulation, during which I_Ks is enhanced. This is likely due to channels opening more quickly, more often, and to greater sublevel amplitudes during adrenergic stimulation. KCNQ1 alone is unresponsive to cyclic adenosine monophosphate (cAMP), and thus KCNE1 is required for a functional effect of protein kinase A phosphorylation. Here, we investigate the effect that KCNE1 has on the response to 8-4-chlorophenylthio (CPT)-cAMP, a membrane-permeable cAMP analog, by varying the number of KCNE1 subunits present using fusion constructs of I_Ks with either one (EQQQQ) or two (EQQ) KCNE1 subunits in the channel complex with KCNQ1. These experiments use both whole-cell and single-channel recording techniques. EQQ (2:4, E1:Q1) shows a significant shift in V_1/2 of activation from 10.4 mV ± 2.2 in control to −2.7 mV ± 1.2 (p-value: 0.0024). EQQQQ (1:4, E1:Q1) shows a smaller change in response to 8-CPT-cAMP, 6.3 mV ± 2.3 to −3.2 mV ± 3.0 (p-value: 0.0435). As the number of KCNE1 subunits is reduced, the shift in the V_1/2 of activation becomes smaller. At the single-channel level, a similar graded change in subconductance occupancy and channel activity is seen in response to 8-CPT-cAMP: the less E1, the smaller the response. However, both constructs show a significant reduction of a similar magnitude in the first latency to opening (EQQ control: 0.90 s ± 0.07 to 0.71 s ± 0.06, p-value: 0.0032 and EQQQQ control: 0.94 s ± 0.09 to 0.56 s ± 0.07, p-value < 0.0001). This suggests that there are both E1-dependent and E1-independent effects of 8-CPT-cAMP on the channel.

Introduction

Normal cardiac action potential repolarization and changes of duration in response to inotropic and dromotropic modulation depend on the integrated function of potassium (K) and sodium (Na) channels, most importantly, I_Kr, I_Ks, and Late I_Na (1). I_Ks has been identified in the heart of several mammalian species (2, 3), where outward K⁺ currents through voltage-gated potassium (Kv) channels contribute to the initiation of repolarization and the subsequent termination of the action potential. Functional loss of I_Ks and I_Kr due to inherited mutation(s) or acquired conditions or late I_Na gain of function account for ∼95% of genetically identified long QT interval syndrome (LQTS) (4), which may cause ventricular arrhythmias and sudden death in children and adults, as well as ∼30% of unexplained Sudden Infant Death Syndrome (5).

Although I_Ks is relevant at normal heart rates, its major role appears to be in shortening the action potential duration and the QT interval at elevated heart rates (6) via two main mechanisms. Firstly, although the activation kinetics of I_Ks are relatively slow compared to the duration of a single action potential, the slowed deactivation kinetics are thought to allow accumulation of I_Ks in the open state during repetitive high-frequency activity. Such accumulation of open channels allows greater outward K⁺ efflux, and this is referred to as a “repolarizing reserve,” which in turn allows the physiological abbreviation of the cardiac action potential, and therefore systole, at high heart rates (7). Secondly, β-adrenergic stimulation results in the phosphorylation of KCNQ1 channels and increases I_Ks by causing a hyperpolarizing shift in the voltage dependence of activation, accelerating activation and slowing deactivation kinetics, both of which contribute to action potential shortening to accommodate rapid heart rates (8). As a result, individuals with LQTS type I (deficiency of I_Ks) tend to suffer arrhythmic events during exercise or at times of high emotion when their heart rates and/or sympathetic drive to the heart is elevated.

The I_Ks channel in the heart is composed of a homotetramer of peptides encoded by the KCNQ1 gene. This tetramer forms the potassium-selective pore and has a voltage-sensitive domain and multiple regulatory regions, including two that are phosphorylated by PKA via cyclic adenosine monophosphate (cAMP) activation. An accessory peptide encoded by the gene KCNE1 (E1) is typically coassembled with KCNQ1 (Q1) and greatly modifies the ion channel kinetics (9, 10). It slows the rate of activation and deactivation and prevents inactivation of the channel (11). The number of accessory peptides that coassemble with the channel has been the topic of much debate over the years. Some groups believe that the stoichiometry of the channel exists in a strict 2:4 (E1:Q1) ratio (12, 13, 14, 15, 16, 17), whereas others propose that the number of E1 subunits may vary from one to four units per channel (18, 19, 20, 21, 22, 23, 24, 25, 26). Our lab has previously published a study in favor of a variable stoichiometry. However, the physiological stoichiometry in cardiomyocytes is still unknown. A study by Yu et al. investigated the composition in cardiomyocytes using ML277, a Q1 activator that reduces in efficacy as the number of E1 subunits increases (19). The cultured human cardiomyocytes’ response to ML277 suggested that there were unsaturated channel complexes present that did not have a full complement of E1 (19). Another study using human embryonic-stem-cell-derived cardiomyocytes produced biophysical characteristics consistent with an unsaturated channel complex. However, when the expression level of E1 was increased, they saw a change in the activation kinetics implying that more E1 was now present in the channel complex (24). This suggests that the stoichiometry is dependent on the level of E1 expression, which may provide another means to regulate I_Ks activity in the heart.

Despite the importance of I_Ks, the biophysics of its adrenergic regulation have not been fully investigated. We, as well as others, have previously shown that exogenous cAMP shifts the voltage dependence of activation to more negative potentials and increases the magnitude of I_Ks when the E1 subunit is present (8, 27, 28, 29). Using single-channel recording techniques to investigate changes in the I_Ks channel kinetics in the presence of 8-4-chlorophenylthio (CPT)-cAMP, we showed that the increase in I_Ks current seen with 8-CPT-cAMP addition is caused by channels that open faster, more often, and to higher-conducting open levels. We suggested that these effects were mediated through voltage-sensor activation (27). However, our studies were done under transfection conditions designed to ensure a fully saturated complex of Q1 and E1 (four Q1 to four E1 subunits). The experiments in the current study use both whole-cell and single-channel recording techniques to examine the effect of varying the ratio of E1 to Q1 on the response to cAMP using previously validated (18) fusion constructs that fix the ratio of E1 to Q1 in the channel complex at different stoichiometries. Our major findings are that the I_Ks response to 8-CPT-cAMP is graded depending on the number of E1 subunits present within the complex and that both E1-dependent and E1-independent components of the I_Ks response to 8-CPT-cAMP exist. Some of these findings were more obvious at the single-channel level.

Methods

Reagents

8-(4-chlorophenylthio)adenosine 3′,5′-cyclic monophosphate sodium salt (8-CPT-cAMP) (Sigma-Aldrich, St. Louis, MO), a membrane-permeable cAMP analog, was used at 200 μM to activate protein kinase A (PKA). Okadaic acid (OA) at 0.2 μM (EMD Millipore, Burlington, MA) was used to inhibit protein phosphatase 1 to sustain the response to the cAMP analog.

Molecular biology

Both KCNQ1 and KCNE1 were purchased from OriGene (Rockville, MD). The fusion constructs (EQ, EQQ, and EQQQQ) were made as described in Murray et al. (18). Yotiao (AKAP-9) was a gift from Dr. R. Kass of Columbia University (New York, NY).

Cell culture and transfections

For both single-channel and whole-cell recordings, ltk⁻ mouse fibroblast and tsA201 transformed human embryonic kidney 293 cells were used. Both were grown in Minimum Eagle Medium (Thermo Fisher Scientific, Waltham, MA) with 100 U/mL penicillin, 10% fetal bovine serum, and 100 μg/mL streptomycin (Thermo Fisher Scientific) added. Cells were kept at 37°C in a humid atmosphere containing 5% CO₂. The cells were exposed for 1 min to trypsin/EDTA to lift the cells and replated on 25 mm² glass coverslips one day before transfection. Constructs were overexpressed by transient transfection using Lipofectamine 2000 (Thermo Fisher Scientific). EQQQQ, EQQ, or EQ, yotiao (required cofactor for phosphorylation of Q1), and GFP (green fluorescent protein) were coexpressed in a 1.5:1.5:1 ratio. All recordings were performed 24–48 h after transfection at room temperature (18, 27, 30, 31).

Patch-clamp electrophysiology

Cells attached to a glass coverslip were placed in a chamber containing control bath solution. Cells selected for both whole-cell and single-channel recordings were GFP-fluorescent. Recordings were obtained using an Axopatch 200B amplifier (Molecular Devices, San Jose, CA) and a Digidata 1440A digitizer (Molecular Devices). The software used was pClamp 10.5 (Molecular Devices) (18, 27).

Whole-cell recording patch electrodes were made from borosilicate glass (World Precision Instruments, Sarasota, FL) and were pulled into electrodes using a linear multistage puller (Sutter Instruments, Novato, CA). Glass pipettes were then fire polished, leading to pipette resistances between 1 and 3 MΩ. Currents were filtered at 5 kHz and sampled at 10 kHz (18).

Single-channel electrodes were made of a thick-walled borosilicate glass (Sutter Instruments), fire polished and coated with Sylgard (Dow Corning, Midland, MI). Electrodes had resistances between ∼40 and 60 MΩ. Acquired data were passed through a low-pass filter at 2 kHz (−3 dB, four-pole Bessel filter) (18, 27, 30, 31).

Solutions

For single-channel recordings, the bath solution contained 135 mM KCl, 1 mM MgCl₂, 1 mM CaCl₂, and 10 mM Hepes and was adjusted to pH 7.4 with KOH. The pipette solution contained 6 mM NaCl, 129 mM 2-(N-morpholino)ethanesulfonic acid, 1 mM MgCl₂, 10 mM Hepes, 5 mM KCl, and 50 μM CaCl₂ and was adjusted to pH 7.4 with NaOH (18, 27).

For whole-cell recordings, the bath solution contained 135 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 2.8 mM NaAcetate, and 10 mM Hepes, pH adjusted to 7.4 with NaOH. The pipette solution contained 130 mM KCl, 5 mM EGTA, 1 mM MgCl₂, 4 mM Na₂-ATP, 0.1 mM GTP, and 10 mM Hepes, pH adjusted to 7.2 with KOH (18, 27).

Data analysis

Data analysis was performed using Prism 7 (GraphPad Software, La Jolla, CA). Using normalized tail current amplitudes, G-V relations were obtained. The data were fitted for each cell using a Boltzmann sigmoidal function:

GV = G_max/(1 + exp [−(V − V_1/2)/k]), where G_max is max conductance, V membrane voltage, V_1/2 the half-activation voltage and k the slope factor.

All single-channel data shown—sweeps and all-points histograms—were filtered at 200 Hz unless otherwise stated, and analysis of channel dwell times at different sublevels (Fig. 5) was done using data filtered at 500 Hz (27).

Figure 5.

Subconductance dwell time analysis of EQQ before and after 8-CPT-cAMP/OA. (A) Raw all-points histograms of 10 active control sweeps (black line) from an EQQ patch and after 200 μM 8-CPT-cAMP/0.2 μM OA (gray dashed line). (B) Idealized histograms of 14 active control sweeps (black line) and 14 8-CPT-cAMP sweeps (gray dashed line). The idealization process used five different thresholds: 0.145, 0.22, 0.33, 0.5, and 0.75 pA. (C) Data in the table show total and average dwell times (ms) for each of the thresholds, the percentage of time spent at each level, and the number of events at each threshold before and after 8-CPT-cAMP. The bin width is 0.01 pA. Only events longer than 1.5 ms were used. Data were filtered at 500 Hz.

The signal/noise ratio of single-channel data is an obstacle in the analysis, particularly for I_Ks, and filtering after acquisition is required to reduce the noise. 500 Hz filtering provides a better signal/noise ratio while keeping the amount of filtering to a minimum (Fig. S1). At this level of filtering, the dead time is 0.36 ms (0.54 × 0.66, rise time of the filter), and this means some brief events, both openings and closings, may be missed, thereby leading to an overestimation of the dwell times.

Sublevel analysis was performed in Clampfit using the single-channel search function. Sublevels are detected across many different filter frequencies, and therefore it seems unlikely that they are an artifact of filtering (Fig. S2). Levels were set to 0, 0.145, 0.22, 0.33, 0.5, and 0.75 pA following the 3/2 rule (32). The “update level automatically” function was disabled. This program detects each event, amplitude level, and length of the event. Each event is then added to a spreadsheet, from which data for the idealized amplitude histograms and dwell time analysis are extracted. An example of the idealization process is shown in Thompson et al. Fig. S8 (27). Because of the rise time of the system at 500 Hz, which was calculated to be 0.66 (0.332/0.5 kHz), only events over 1.5 ms (two times the rise time) in duration were included in the analysis (27, 30).

Statistics

Results shown here are mean values ± SE and are averages of at least three independent experiments. Both the median and the 95% confidence intervals are also shown for the first latency data. A nonparametric Mann-Whitney test (Prism 7; GraphPad Software) was used to compare first latencies of both control and 8-CPT-cAMP-treated groups. Unpaired t-tests were used to compare V_1/2 of control and 8-CPT-cAMP/OA groups as not all experiments were paired. Significant differences between means are considered as p-values < 0.05.

Results

The effect of cAMP on I_Ks in whole-cell recordings

In whole-cell recordings (Fig. 1), exogenous 8-CPT-cAMP had no effect on currents in cells transfected with Q1 and yotiao. Data show no overall change in the Q1 current amplitude or the waveform of whole-cell currents after 8-CPT-cAMP addition (Fig. 1, A–C) and no shift of the normalized conductance to hyperpolarizing potentials (Fig. 1 D). Q1 displayed an average V_1/2 in control of −22.0 mV and in the presence of 8-CPT-cAMP of −22.7 mV (Fig. 1 D; Table 1). However, during whole-cell recordings from cells transfected with both Q1 + E1 (transfected DNA ratio, 3:1 E1:Q1), 8-CPT-cAMP caused an increase in peak current and a large hyperpolarizing shift of ∼−18 mV in the G-V relationship, as we have previously shown in Thompson et al. (27) (Table 1).

Figure 1.

KCNQ1 does not respond to 8-CPT-cAMP in whole-cell recordings. Currents were recorded at room temperature. (A) Representative traces for Q1 alone with yotiao before (control, black line) and after 200 μM 8-CPT-cAMP/0.2 μM OA (gray line), pulsed to +60 mV for 2 s and then −40 mV for 900 ms. (B) A diary plot of the peak current of a 2 s +60 mV pulse over time. The addition of 200 μM 8-CPT-cAMP and 0.2 μM okadaic acid (OA) is marked by a solid bar. (C) Representative traces are shown from every other voltage starting from −80 to +60 mV for both control (top panel) and in the presence of 200 μM 8-CPT-cAMP and 0.2 μM OA (bottom panel). (D) G-V curves recorded before (black circles) and after 8-CPT-cAMP (gray squares). Cells were held at −90 mV and pulsed from −80 to +60 mV in 10-mV steps for 2 s. Tail currents were recorded at −40 mV for 900 ms. Control V_1/2: −22.0 mV ± 1.8 SE, n = 5 and 8-CPT-cAMP V_1/2: −22.7 mV ± 1.8 SE, n = 4, unpaired t-test p-value: 0.8038.

Table 1.

V_1/2 of Activation Before and After 8-CPT-cAMP/OA

Construct	Control			8-CPT-cAMP/OA				Unpaired t-Test
	V1/2	k-factor	n	V1/2	k-factor	n	Δ in V1/2	p-value
KCNQ1	−22.0 ± 1.8	10.8 ± 0.3	5	−22.7 ± 1.8	11.5 ± 0.2	4	−0.7 ± 2.6	0.8038
EQQQQ	6.3 ± 2.3	16.1 ± 1.2	4	−3.2 ± 3.0	13.7 ± 1.2	4	−9.5 ± 3.7	0.0435
EQQ	10.4 ± 2.2	16.1 ± 1.6	8	−2.7 ± 1.2	12.3 ± 0.4	4	−13.1 ± 3.3	0.0024
EQ	26.7 ± 2.5	18.9 ± 1.0	6	4.2 ± 6.0	16.0 ± 1.9	4	−22.5 ± 5.7	0.0042
KCNQ1 + KCNE1a	28.2 ± 5.4	21.3 ± 2.9	4	10.5 ± 2.6	18.5 ± 0.8	4	−17.7 ± 6.0	0.0261

V_1/2 of activation was obtained from −40 mV tail portion of activation protocols (−90 mV holding potential, pulsed from −90 up to +100 mV in 10 mV steps for 4 s, then −40 mV). p-values compare control and the 200 μM 8-CPT-cAMP and 0.2 μM okadaic acid (OA) group using an unpaired t-test.

^aData taken from Thompson et al. (27).

Cells transfected with an EQ linked construct, which forces a 4:4 E1:Q1 stoichiometry (Fig. S3), show shifts in current magnitude (Fig. S3, A and B) and V_1/2 that were similar to those observed when cells were transfected with Q1 and E1 separately. EQ had an average control V_1/2 of 26.7 mV that was hyperpolarized to 4.2 mV in the presence of 8-CPT-cAMP (Fig. S3). 8-CPT-cAMP displayed a significant average shift of ∼−23 mV (p-value: 0.0042, Table 1).

When the stoichiometric ratio of E1:Q1 was reduced to 2:4 using an EQQ construct, 8-CPT-cAMP caused only small changes in the overall waveform (Fig. 2 A) and a small increase in peak current at +60 mV (Fig. 2 B) but not as profound an increase as when fully saturated with E1 (Fig. S3). The waveform of both the control and 8-CPT-cAMP traces in Fig. 2 C suggests from the tail currents that saturation of the current upon activation is occurring at more negative potentials in the presence of cAMP. This steepening and hyperpolarizing shift of the G-V relationship is shown in Fig. 2 D, which was obtained from tail current analysis. The average V_1/2 in EQQ was 10.4 mV, and the V_1/2 in the presence of 8-CPT-cAMP was −2.7 mV (p-value = 0.0024). The shift in normalized conductance (∼−13 mV) is smaller than that seen with EQ (∼−23 mV) (Table 1).

Figure 2.

EQQ V_1/2 is hyperpolarized in response to 8-CPT-cAMP addition. Currents were recorded at room temperature. (A) Representative traces for EQQ with yotiao before (control, black line) and after 200 μM 8-CPT-cAMP/0.2 μM OA (gray line), pulsed to +60 mV for 2 s and then −40 mV for 900 ms. (B) A diary plot of the peak current of a 2 s +60 mV pulse over time. The addition of 200 μM 8-CPT-cAMP/0.2 μM OA is marked by a solid bar. (C) Representative traces are shown from every other voltage starting from −80 to +100 mV for both control (top panel) and in the presence of 200 μM 8-CPT-cAMP and 0.2 μM OA (bottom panel). (D) G-V curves recorded before (black circles) and after 8-CPT-cAMP/OA (gray squares). Cells were held at −90 mV and pulsed in 10-mV steps for 4 s from −80 to +100 mV in control and +60 mV in 8-CPT-cAMP/OA. Tail currents were recorded at −40 mV for 900 ms. Control V_1/2: 10.4 mV ± 2.2 SE, n = 8 and 8-CPT-cAMP V_1/2: −2.7 mV ± 1.2 SE, n = 4, unpaired t-test p-value: 0.0024. ΔV_1/2: −13.1 mV ± 3.3 SE.

Having shown a reduction in the response to cAMP of the I_Ks current with a half-saturated complement of E1 (EQQ 2:4) compared to a full complement of E1 (EQ 4:4), it was of interest to examine the response of an I_Ks complex with only one E1 (EQQQQ 1:4) to cAMP. EQQQQ was transfected along with yotiao similarly to EQ and EQQ. Representative +60 mV traces (Fig. 3 A) and a diary plot (Fig. 3 B) of EQQQQ do not show increases in the current amplitude over time with exposure to cAMP. The waveforms of both control and 8-CPT-cAMP exposed cells do not appear to be dissimilar to each other (Fig. 3 C). Upon further analysis of the tail currents before and after 8-CPT-cAMP, a significant hyperpolarization of the V_1/2 of activation was observed. The average V_1/2 in control cells was 6.3 mV and the V_1/2 of 8-CPT-cAMP/OA cells was −3.2 mV (p-value = 0.0435, Fig. 3 D; Table 1).

Figure 3.

EQQQQ V_1/2 is slightly hyperpolarized in response to 8-CPT-cAMP addition. Currents were recorded at room temperature. (A) Representative traces for EQQQQ with yotiao before (control, black line) and after 200 μM 8-CPT-cAMP/0.2 μM OA (gray line), pulsed to +60 mV for 4 s and then −40 mV for 900 ms. (B) A diary plot of the peak current of a 4 s +60 mV pulse over time. The addition of 200 μM 8-CPT-cAMP/0.2 μM OA is marked by a solid bar. (C) Representative traces are shown from every other voltage starting from −80 to +90 mV for both control (top panel) and in the presence of 200 μM 8-CPT-cAMP and 0.2 μM OA (bottom panel). (D) G-V curves recorded before (black circles) and after 8-CPT-cAMP/OA (gray triangles). Cells were held at −90 mV and pulsed from −80 to +90 mV in 10 mV steps for 4 s. Tail currents were recorded at −40 mV for 900 ms. Control V_1/2: 6.3 mV ± 2.3 SE, n = 4 and 8-CPT-cAMP V_1/2: −3.2 mV ± 3.0 SE, n = 4, unpaired t-test p-value: 0.0435. ΔV_1/2: −9.5 mV ± 3.7 SE.

The effect of cAMP on I_Ks single-channel openings

Data in Figs. 4, 5, and 6 demonstrate the effects of 8-CPT-cAMP on the single-channel properties of EQQ and EQQQQ. Single-channel recordings in control solutions have been reported previously (18). I_Ks characteristically has a long first latency to opening after depolarization and, once open, exhibits bursts of rapid brief openings. EQ, which has 4 E1 subunits, exhibits the longest first latency and more openings to high amplitudes (∼0.5 pA) (18, 27). When the number of E1 subunits within the I_Ks complex is reduced to a 2:4 ratio of E1:Q1, these properties change. The characteristic fast-flickering opening behavior is still intact, but the channel opens more quickly and visits the higher-amplitude levels less often (Figs. 4 A and 5 A). EQQQQ has the shortest first latency and the smallest amplitudes, never reaching the higher subconducting states (Fig. 6).

Figure 4.

8-CPT-cAMP shortens the first latency of EQQ single channels. Currents were recorded at room temperature. (A) Representative 4 s cell-attached single-channel recordings of control (left panel) and 8-CPT-cAMP/OA (right panel). Data were filtered at 200 Hz. (B) All-points event histograms of 10 active sweeps in control (left panel) and 8-CPT-cAMP/OA-treated EQQ (right panel). (C) A cumulative latency histogram showing that the mean first latency for the 107 active control sweeps (black line and arrow) of 259 was 0.90 ± 0.07 s. In 200 μM 8-CPT-cAMP/0.2 μM OA (gray line and arrow), mean first latency for the 129 active sweeps of 378 was 0.71 ± 0.06 s; p = 0.0032 (see Table 2). Sweeps without activity were given a first latency >4 s; arrows indicate sweep total in each case. Note split-scale ordinate.

Figure 6.

8-CPT-cAMP shortens the first latency of EQQQQ single channels. Currents were recorded at room temperature. (A) Representative 4 s single-channel sweeps of control (left panel) and 8-CPT-cAMP/OA (right panel). Data were filtered at 200 Hz. (B) All-point event histograms of 10 active sweeps in control (left panel) and 8-CPT-cAMP/OA (right panel). Inset panels show the enlarged foot of each histogram. (C) A cumulative latency histogram showing that the mean first latency for the 50 active control sweeps (black line and arrow) of 171 was 0.94 ± 0.09 s. In 200 μM 8-CPT-cAMP/0.2 μM OA (gray line and arrow), mean first latency for the 62 active sweeps of 314 was 0.56 ± 0.07 s; p-value = <0.0001 (see Table 2). Sweeps without activity were given a first latency >4 s; arrows indicate sweep total in each case. Note split-scale ordinate.

In Fig. 4 A, representative EQQ single-channel traces are shown for both control (left panel) and 8-CPT-cAMP (right panel) environments. The all-points histograms in Fig. 4 B display the number of events recorded in each 0.01 pA amplitude bin. In control sweeps, the peak of the main Gaussian distribution was at 0.17 pA, and in 8-CPT-cAMP, the peak was at 0.18 pA. There did not appear to be any great change in the distribution of events like that seen previously with the 4:4 stoichiometries of EQ and Q1 + E1 (27). There was a subtle decrease in the smaller subconductance events (0.08 pA) and an increase in events around 0.2 pA.

Although there was no significant change in amplitude, the channel opened significantly more quickly in the presence of 8-CPT-cAMP, which is also illustrated in the representative sweeps in Fig. 4 A. This was quantified as a significant reduction in the time to first opening of the channel in the presence of 8-CPT-cAMP (Fig. 4 C). The average first latencies of 107 active control sweeps was 0.90 s, whereas the average first latency of 129 active sweeps in 8-CPT-cAMP was 0.71 s (Fig. 4 C; Table 2, p-value = 0.0032).

Table 2.

First Latency Data for Q1 + E1, EQ, EQQ, and EQQQQ, Before and After 200 μM 8-CPT-cAMP/0.2 μM OA

Construct	Control					8-CPT-cAMP/OA					p-value	n (cells)
	Mean First Latency (s) ± SE	95% Confidence Interval of the Mean (Lower–Upper)	Median (s)	# Active Sweeps	# Total Sweeps	Mean First Latency (s) ± SE	95% Confidence Interval of the Mean (Lower–Upper)	Median (s)	# Active Sweeps	# Total Sweeps
Q1 + E1a	1.32 ± 0.13	1.06–1.57	0.90	68	278	0.79 ± 0.08	0.63–0.95	0.50	104	309	0.0002	3
EQa	1.61 ± 0.13	1.35–1.87	1.57	43	367	1.06 ± 0.11	0.84–1.27	0.85	69	345	0.0005	3
EQQ	0.90 ± 0.07	0.76–1.0	0.64	107	259	0.71 ± 0.06	0.58–0.84	0.46	129	378	0.0032	4–5
EQQQQ	0.94 ± 0.09	0.76–1.1	0.76	50	171	0.56 ± 0.07	0.41–0.70	0.38	62	314	<0.0001	4–5

First latencies were obtained from 4 s sweeps pulsed to +60 mV. Only active sweeps were used for average. p-values compare control and 8-CPT-cAMP/OA group using the Mann-Whitney test.

^aData taken from Thompson et al. (27).

The number of active sweeps did not increase in the presence of 8-CPT-cAMP. In control, there were 107 active sweeps out of 259 (∼41% active), and in 8-CPT-cAMP, there were 129/378 (∼34% active) (Fig. 4 C; Table 2). This is unlike the saturated 4:4 complex, EQ, which showed a large increase in the number of active sweeps from 43/367 to 69/345 (27).

An analysis of the dwell times in five identifiable open states of EQQ channels before and in the presence of 8-CPT-cAMP is shown in Fig. 5. These averaged data were obtained from 14 active sweeps in each case and show only small differences induced by the presence of cAMP. The all-points and idealized histograms clearly diverge at the higher open levels (Fig. 5), and this corresponds to an increased occupancy of the 0.5 and 0.75 pA levels (Fig. 5 C). At the lower levels, no differences were observed, and data are remarkably concordant before and after 8-CPT-cAMP.

EQQQQ single-channel responses to 8-CPT-cAMP are described in Fig. 6. Representative traces before (left panel) and after 8-CPT-cAMP (right panel) are shown in Fig. 6 A. The EQQQQ single-channel open amplitudes are very small in the presence of only one E1 subunit—so small that the main opening peak is hard to distinguish from the closed peak (Fig. 6 B). The all-points histograms do show a small decrease in the number of closed events and a small increase in the number of active events with 8-CPT-cAMP (Fig. 6 B), but unlike EQQ constructs, there was no increase in range of amplitudes visited between control (Fig. 6 B, left panel inset) and 8-CPT-cAMP (Fig. 6 B, right panel inset). In both control and 8-CPT-cAMP, EQQQQ opening events larger than 0.2 pA were very rarely seen. However, like EQQ, EQ, and Q1 + E1, EQQQQ channel first latency to opening was reduced upon 8-CPT-cAMP exposure (Fig. 6 C). The average first latency in 50 active control sweeps was 0.94 s compared to 0.56 s in 62 active sweeps in the presence of 8-CPT-cAMP (Fig. 6 C; Table 2, p < 0.0001). As for EQQ, the number of active sweeps after 8-CPT-cAMP did not increase (50/171 in control and 62/314 in 8-CPT-cAMP, Table 2).

Ensemble averages of EQ, EQQ, and EQQQQ single-channel sweeps are shown in Fig. S4 A. EQ, EQQ, and EQQQQ all show faster activation kinetics in the presence of 8-CPT-cAMP compared with control, which is consistent with a shortening of the first latency for all three constructs. Activation of whole-cell currents also appears to be faster in EQ, EQQ, and EQQQQ in the presence of 8-CPT-cAMP, as shown by the representative sweeps in Fig. S4 B.

Discussion

The functional effect of phosphorylation of Q1 in response to β-adrenergic stimulation has long been thought to require the E1 subunit to be present. Under sympathetic stimulation, the S27 and possibly S92 residues in the N-terminal region of the Q1 subunit become phosphorylated (28, 33, 34, 35). However, phosphorylation is not enough to produce a functional effect, as E1 is also a requirement (33). In Fig. 1, data show that there was no change in the V_1/2 or increase in current when Q1 alone was exposed to 8-CPT-cAMP, in agreement with previous studies that have also shown no functional response in the absence of E1 despite Q1 still being phosphorylated (33). There are several LQT missense mutations in the E1 C-terminal region that have an impaired β-adrenergic response, such as D76N, and deletions within the E1 C-terminus (33, 36, 37, 38, 39, 40) have similarly led to the conclusion that the E1 C-terminus is important in the sympathetic response of I_Ks to cAMP.

The present experiments confirm the importance of the E1 subunit in the response of I_Ks to 8-CPT-cAMP. As stated above, the Q1 tetramer alone is unresponsive to cAMP exposure (Fig. 1). It has also previously been shown that a 2:4 E1:Q1 stoichiometry showed a 34% reduction in the current amplitude response to cAMP in Chinese hamster ovary cells compared with a 4:4 E1:Q1 I_Ks construct or independently transfected Q1 + E1 (33). This result is comparable with some of the findings of this study. In data from whole-cell experiments, exposure to 8-CPT-cAMP also produced a ∼35% increase in current at +60 mV in the EQQ (2:4 E1:Q1) construct, although this was often not sustained (Fig. 2 B). This increase is rather less than that seen in the EQ construct (Fig. S3). In our experiments, the increased current in the EQQ construct was associated with a steepening of the G-V relationship (Fig. 2 D; Table 1) and a −13 mV hyperpolarization of the V_1/2. This may be compared with a −23 mV hyperpolarization in the EQ construct and −18 mV in independently expressed KCNQ1 and KCNE1 (Table 1).

It appears that E1 subunits depolarize the V_1/2 of activation of Q1 in a graded manner so that the more E1 subunits in the complex, the more depolarized the V_1/2 (Fig. S5 B). Phosphorylation by PKA antagonizes this shift, and the magnitude of the hyperpolarizing response to cAMP is reduced as the number of E1 subunits is reduced. This seems to imply that phosphorylation is directly acting to reverse the depolarizing action of E1 subunits on the V_1/2. However, in extremely rare longer-lasting cells of both EQ and EQQQQ, 8-CPT-cAMP hyperpolarized the V_1/2 of activation to ∼−10 mV, which suggests that perhaps all the constructs are able to converge around this value, upon phosphorylation. An example of this effect in EQ is shown in Fig. S6, in which an initial V_1/2 of activation of +36 mV shifts to −11 mV as time progresses. Q1 alone has a V_1/2 of activation of −22 mV, and a shift that negative was not seen in any of our E1-containing constructs after cAMP (Fig. S5 B). Such long time-course data are experimentally difficult to obtain, but we can speculate that the hyperpolarizing limit of the cAMP response may be at ∼−10 mV and might explain the apparent convergence of the first latencies between the different constructs (Fig. S5 A; Table 2). From the cumulated data, a linear relationship exists between the V_1/2-values of EQ, EQQ, and EQQQQ in the absence and presence of 8-CPT-cAMP (Fig. S5 A). But, this linear relationship does not extend to Q1 alone, as a gap of around 10 mV is seen between the Q1 V_1/2 and the most hyperpolarized V_1/2 seen in response to cAMP in the presence of E1 (Fig. S5 B). Given the discussion above, this suggests that although cAMP appears to reverse the depolarizing action of E1 subunits, the two phenomena are not directly related.

Experiments with the EQQQQ construct, in which only one E1 subunit is present in the I_Ks complex, did not show a significant effect on the current amplitude at the whole-cell level (Fig. 3, A and B) but did show a small but significant hyperpolarizing shift in the V_1/2 of activation. As the number of E1 subunits present in the I_Ks complex decreases, the amplitude response and the hyperpolarizing shift of the activation V_1/2 is reduced. These data all support the idea that the degree of response to cAMP is graded to the number of E1 accessory subunits present (although a 3:4 E1:Q1 construct was not tested) and that at least one E1 subunit is required to observe the effects of cAMP (Fig. S5).

Many of the whole-cell findings were mirrored in the single-channel data. One of the important characteristics of the I_Ks channel complex is its low open probability. However, EQ and E1 + Q1 expressed separately become more active in response to 8-CPT-cAMP and show an increase in the number of active sweeps per patch (27). This was not seen with either EQQ (Fig. 4 C) or EQQQQ (Fig. 6 C). In control EQQ, ∼41% of sweeps showed I_Ks activity, but in the presence of 8-CPT-cAMP, this decreased to ∼34% (Fig. 4 C; Table 2). EQQQQ showed a similar reduction in active sweeps from ∼29% in control to ∼20% in 8-CPT-cAMP (Fig. 5 C; Table 2). As previously shown (18, 27), EQ (4:4) is able to visit many different subconductance levels with the main peak around 0.45 pA, as also seen when Q1 and E1 are expressed independently. However, EQQ and EQQQQ both no longer consistently reach the main open amplitude, with EQQ rarely having events at 0.5 pA and EQQQQ not at all. The occupancy of the higher amplitudes was increased in EQQ in the presence of 8-CPT-cAMP but only reached 2.6% of the total dwell time, and there was no effect for EQQQQ. However, there was an increase in the number of events for both EQQ and EQQQQ due to the shorter first latencies (Figs. 4 and 5 B).

Interestingly, the response of the first latency to channel opening to 8-CPT-cAMP was not affected by subunit composition in these experiments. The first latency was reduced by cAMP in EQ by 35% and in Q1 and E1 expressed separately by 41% (Table 2), compared with 22% in EQQ and 39% in EQQQQ (Figs. 4 C and 6 C). The robust reduction in first latency seen in the single-channel data from the EQQ and EQQQQ constructs suggests that phosphorylation may have both KCNE1-dependent and independent actions on the gating of the I_Ks channel complex. Given the similar reduction in first latency across the constructs, this effect could be an all-or-none response to phosphorylation by PKA. Q1 alone may also be affected, but we are not able to reliably determine its first latency. When the I_Ks channel first opens, it characteristically does so to very small amplitudes, representing the lowest subconductance levels (30), as also clearly seen in Fig. 4 A. It is possible that the modulation by 8-CPT-cAMP of the earliest steps in channel activation that lead to the first subconductance level openings are not as dependent on the presence of E1 as openings to the higher subconductance levels. This finding is consistent with the idea that E1 itself does not affect the rate of voltage-sensor activation in the Q1 subunit (for review, see Liin et al. (41)).

Mechanisms for the E1-dependent action of 8-CPT-cAMP

Previous studies have shown that the C-termini of E1 and Q1 interact with each other to cause a conformational change in the Q1 subunit that brings the N- and C-termini closer together as the channel opens (23). This step has been thought to be crucial for the phosphorylation of serine residues in the N-terminus of Q1. However, these residues have been shown to still be phosphorylated when E1 is not present or when the E1 C-terminus is truncated (Δ109–129 (40)). This suggests that both the N- and C-termini can come close enough for phosphorylation to occur without E1. When the E1 C-terminus is truncated, not present, or has one of a few known point mutations, there is no longer an effect of this phosphorylation (33, 36, 40, 42). The interaction between the C-termini of Q1 and E1 may not be facilitating phosphorylation but rather allowing another conformational change to occur that allows the E1 subunit to modulate the interaction between the voltage-sensor domains and the pore domains of the channel complex, perhaps by changing the affinity of Q1 to phosphatidylinositol 4,5-bisphosphate (PIP₂) (43), among other mechanisms (44).

E1 D76N is an LQT mutation that does not respond to cAMP, has much faster deactivation, and causes a ∼35 mV depolarization in the V_1/2 of activation (39). Both the E1 C-terminus and Q1 can still associate with each other, and phosphorylation of Q1 is still observed, but there is no functional response to phosphorylation by PKA. The movement of the E1 and Q1 C-termini coming closer together as seen in the wild type is no longer observed with this mutant (39). Chen et al. suggested that the change from an acidic to a polar, uncharged R-group may result in an E1 C-terminus that cannot adequately slow deactivation or stabilize the open state of the channel (39).

An NMR study (12) of the distal E1 C-terminus (residues 106–129) suggests that it is quite disordered and flexible. Shortening of the E1 C-terminus (Δ69–77) was found to increase the binding of both E1 and Q1 C-termini compared to the wild type. Molecular modeling has suggested that the E1 C-terminus is able to fold back toward the membrane, which could allow it to interact with the S4-S5 linker and/or the pore (12). Using a cysteine scanning method, Lvov et al. (45) postulated that the E1 C-terminus could interact with the S6 activation gate and/or the S4-S5 linker of Q1. This flexible tail may allow E1 the freedom/ability to stabilize I_Ks complex open states or destabilize closed states. If E1 is able to cause a conformational change in response to phosphorylation of Q1, then this may explain the graded response to 8-CPT-cAMP as additional E1 subunits are present. What is clear is that voltage sensor to pore coupling in Q1 is strongly regulated by the presence of E1 (44), and thus, changes in Q1 after phosphorylation require E1 to be present to be transduced into higher occupancy of subconductance states later in the channel-activation process.

I_Ks in the heart

In normal human cardiac myocytes, the I_Ks current is clearly important because of the high number of LQTS mutations in Q1 (and E1), which have a phenotype, but as yet, even the physiological composition of the channel complex remains uncertain. One study recorded a current-voltage relationship for I_Ks from human left ventricular myocytes that appeared to have a V_1/2 of activation of ∼5 mV (46). Importantly, for our study, these data were collected from cells exposed to forskolin, which was required to increase the amplitude of I_Ks to enable the study. EQQ and EQQQQ exposed to 8-CPT-cAMP have V_1/2-values of activation of ∼−3 mV, and EQ is ∼5 mV at room temperature (Table 1) but could be as hyperpolarized as −10 mV (Fig. S3). An increase in recording temperature leads to an additional ∼10 mV shift in the hyperpolarizing direction (47). Together, these data suggest that the stoichiometry in the human left ventricle could well include a 4:4 ratio of Q1:E1.

Conclusions

At least one E1 subunit appears to be needed to observe a functional response to 8-CPT-cAMP. At the whole-cell level, with only one E1 subunit, the EQQQQ channel shows smaller changes in the V_1/2 of activation in response to PKA phosphorylation compared with the fully saturated EQ complex. The actions of 8-CPT-cAMP are graded depending on the number of E1s present, and this graded effect indicates that complexes with fewer E1s may show a blunted sympathetic response. At the single-channel level, changes in subconductance occupancy and channel activity mirror the whole-cell effects of reduced numbers of E1 in the I_Ks complex. However, changes in the latency to first opening of I_Ks, induced by cAMP, were not affected by the number of E1 subunits present, which has highlighted the possibility of E1-dependent and E1-independent actions of 8-CPT-cAMP on the gating of I_Ks channels.

Author Contributions

E.T., J.E., and D.F. designed the research and wrote the manuscript. E.T, J.E., M.W., and D.M. performed experiments. E.T., J.E., M.W., D.M. and D.F. analyzed the data.

Editor: Henry M. Colecraft.