The rapid, repolarizing K+ current in cardiomyocytes (I Kr) has unique inwardly rectifying properties that contribute importantly to the downstroke of the cardiac action potential. The humanether-à-go-go-related gene (HERG) expresses a macroscopic current virtually identical toI Kr, but a description of the single-channel properties that cause rectification is lacking. For this reason we measured single-channel and macropatch currents heterologously expressed byHERG inXenopus oocytes. Our experiments had two main findings. First, the single-channel current-voltage relation showed inward rectification, and conductance was 9.7 pS at −100 mV and 3.9 pS at 100 mV when measured in symmetrical 100 mM K+ solutions. Second, single channels frequently showed no openings during depolarization but nevertheless revealed bursts of openings during repolarization. This type of gating may explain the inward rectification of HERG currents. To test this hypothesis, we used a three-closed state kinetics model and obtained rate constants from fits to macropatch data. Results from the model are consistent with rapid inactivation from closed states as a significant source of HERG rectification.
- rapid repolarizing cardiac potassium current
the potassium ion(K+) current expressed heterologously in Xenopus oocytes by the human ether-à-go-go-related gene (HERG) (15, 23) has many properties ofI Kr, the rapid component of the repolarizing K+current in heart (16). For example HERG current, likeI Kr, activates slowly and at positive potentials displays the inward rectification so critical to the role ofI Kr in producing the downstroke of the cardiac action potential. HERG current, likeI Kr in human cardiomyocytes (28) and ferret cardiomyocytes (13), has a transient peak at positive potentials (11, 17). Pharmacologically, HERG is sensitive to block by the class III methanesulfonanilides (11, 20, 23) that block I Kr(8, 12, 16). Mutations in the HERGgene cause one form of hereditary long Q-T syndrome (3, 9) that is associated with the potentially lethal arrhythmia torsade de pointes, and class III methanesulfonanilide blockers ofI Kr produce similar phenomena.
Recently whole oocyte and macropatch measurements have suggested that C-type inactivation is the primary cause of inward rectification of HERG current (17, 19, 20). To understand further how HERG might produce rectification and a peak transient current, we measured its elementary currents. We found that single HERG channels may fail to open on depolarization but will open on repolarization. We attribute this result to inactivation from a closed state and note that this type of gating will contribute to inward rectification. We demonstrate that single HERG channels accumulate in inactivated states during depolarization and reopen or open for the first time during repolarization, confirming that, once repolarization has been initiated, HERG is uniquely suited to produce the downstroke of the cardiac action potential.
MATERIALS AND METHODS
Xenopus oocyte measurements were performed using standard two-microelectrode voltage-clamp techniques with a Dagan 8500 voltage clamp (Dagan, Minneapolis, MN) (11). Macropatch currents were recorded from oocytes using patch pipettes made from borosilicate glass with tip openings of 10–15 μm and resistances of 150–250 kΩ with an Axopatch 1D patch-clamp amplifier (Axon Instruments, Foster City, CA). After a gigaseal was achieved, recordings were made in the cell-attached mode of the conventional patch-clamp technique (4). Single-channel recordings were performed using pipettes pulled from hard borosilicate glass with resistances of 5–10 MΩ. The seal resistance was 50–500 GΩ. Pipettes were coated with Sylgard and fire polished immediately before use.
Solutions and drug administration.
Two-microelectrode voltage-clamp measurements ofXenopus oocytes were performed in a bath solution (low-K+ solution) containing (in mmol/l) 5 KCl, 100 NaCl, 1.5 CaCl2, 2 MgCl2, and 10 HEPES (pH 7.3).
For macropatch and single-channel recordings, the 100 mM K+ pipette solution contained (in mM) 100 KCl, 2 MgCl2, and 10 HEPES (pH adjusted to 7.3 with KOH). The 5 mM K+ solution contained (in mM) 5 KCl, 100 NaCl, 1.5 CaCl2, 2 MgCl2, and 10 HEPES (pH adjusted to 7.3 with NaOH). The bath solution in all single-channel and macropatch measurements contained (in mM) 100 KCl, 1 MgCl2, and 10 HEPES (pH adjusted to 7.3 with KOH). All measurements were done at room temperature (22°C).
Data were low-pass filtered at 1–2 kHz (−3 dB, 4-pole Bessel filter) before digitization at 5–10 kHz. pCLAMP software (Axon Instruments) was used for generation of the voltage-pulse protocols and for data acquisition. The single-channel measurements were corrected for leak and capacitance current by subtracting the average current of 5–10 null recordings. Single-channel kinetics were analyzed using Transit software (24). This resulted in histograms for amplitudes, open time, closed times, and burst duration. Probability density function parameter estimates were obtained with the maximum-likelihood method and gave values for the exponential components for open time (τopen), closed times (τclosed,1, τclosed,2, and τclosed,3), and burst duration (τburst). Transit software uses a statistic based on the maximum-likelihood ratio to determine the minimum number of exponential components in dwell-time distributions. A second, slow open-time component (29) was not statistically validated in our data. For calculation of burst duration, we used a critical closed time calculated so that equal proportions of short and long closed intervals are misclassified (1). Statistical data are given as means ± SD. In single-channel data,n refers to the number of patches analyzed.
Single-channel detection error.
We have calibrated the response of our channel-detection system with simulated data to calculate the error for detection of single channels by Transit under our experimental conditions. Transit idealizes channels by detecting the instantaneous transitions among channel open and closed states rather than the crossing of an amplitude threshold. Transit only requires that the transitions have a derivative (slope) that exceeds the standard deviation of the baseline noise derivative by a user-specified multiple. Details of the algorithm have been published recently (24). The maximum likelihood-optimization procedure in Transit is based on the variable metric Davidon-Fletcher-Powell method and includes a correction for intervals shorter than the minimum and longer than the maximum observable intervals (2). We set the minimum observable time as the filter rise time and the maximum observable interval as the pulse length for estimation of dwell-time parameters. For our experiments these were 0.3 and 398 ms, the closest values permitted by Transit.
Single-channel simulation software was provided by Dr. Antonius VanDongen. We simulated data with a simple two-state model with a mean open time of 2 ms and a mean closed time of 8 ms. Under these conditions at a 20-kHz sampling rate without filtering or added noise, the maximum-likelihood parameter estimate for the mean open time was 2.004 ms, the average open time was 2.00 ms, and the number of openings detected was 780. The same simulation sampled at 5 kHz with 1-kHz filtering and the addition of 0.07 standard deviation noise resulted in a maximum-likelihood parameter estimate for the mean open time of 2.042 ms, an average open time of 2.30 ms, and 714 detected openings. There was negligible error in estimating the mean open time, and the number of missed openings corresponds to 8% of the number of openings detected at 20 kHz in the absence of filtering and noise. With these model parameters, 2% of openings should be missed as the result of concatenation of open times due to missed brief closures. We conclude that, at 5-kHz sampling and 1-kHz filtering, Transit misses 6% of single openings with a true mean of 2 ms because the openings are too brief. As an aid to channel detection, because the HERG channels burst, failure to detect all openings in a burst would require that all the openings in the burst have an open time less than the detection limit.
Nonlinear least-squares fitting of kinetics model transition rates to current recordings was accomplished with the Solver add-in to Microsoft Excel 97 and Windows NT 4. Experimental current data at each potential were fit to the functionI(t) =N ⋅ P o(t) ⋅ i, whereI(t) indicates the macroscopic current, Nis the number of channels,P o is the probability of occupancy of the open state, andi is the single-channel current amplitude. Values ofP o(t) were generated in Excel by Euler integration of the kinetics equations for each model. Initial parameter estimates for a model were obtained with a step size of 0.025 ms. When a consistent set of parameters was obtained that was relatively insensitive to variation of initial parameter values, the step size was increased to 0.25 ms to permit more rapid exploration of the model. Solutions were checked by reducing the step size to 0.025 ms. Reported solutions were stable when the step size was decreased. Values for N were optimized, and values for i were obtained from experimental data. Initial state occupancies were fixed, with all channels occupying the closed state farthest from the open state. Allowing the optimization procedure to vary initial state occupancies produced negligible occupancies of other closed states in the models evaluated. Currents from the deactivating voltage step were used to generate simultaneous fits to the tail currents. Initial state occupancies for the deactivating step were set equal to state occupancies at the end of the activating step. We constrained the sum of the transition rates for leaving the open state to be equal to the reciprocal of the experimentally determined open time and constrained the sum of the transitions leaving the proximal closed state to be equal to the experimentally observed fast closed time at each potential. Microscopic reversibility was maintained in models with transitions among states forming closed loops by making one of the transition rates in each loop take on appropriate values calculated as a function of all the other independent transition rates in the loop.
The HERG clone was a gift from Dr. M. T. Keating (3). The pSP64 construct containingHERG was linearized withEcoR I (Boehringer Mannheim, Indianapolis, IN) and transcribed into cRNA with the mMESSAGE mMACHINE in vitro transcription kit (Ambion, Austin, TX) using SP6 polymerase. RNA (50–500 ng/μl) was injected intoXenopus oocytes, and measurements were performed 2–8 days after injection.
In whole oocyte recordings HERG currents had an activation threshold of −40 mV. The current amplitude became larger at −20 and 0 mV and then decreased at 20-, 40-, and 60-mV membrane potentials because of inward rectification (Fig.1 A). At 40 and 60 mV the currents showed a transient peak. Tail currents were outward at −70 mV and had a characteristic rising phase (hook), reached a peak, and then deactivated more slowly. The rising phase in the tails is attributed to recovery from inactivation that is faster than deactivation (18). The current kinetics in cell-attached macropatches were identical to the whole oocyte measurements [5 mM extracellular K+ concentration ([K]o) cell-attached macropatch currents not shown].
In cell-attached macropatches in 100 mM [K]o, the activation threshold was also −40 mV and currents were small and inward at −40 and −20 mV. From 20 to 100 mV, outward currents that had transient peaks were observed (Fig. 1,B andinset). During the steps back to −120 mV, tail currents were large and inward and showed the rising phase associated with rapid recovery from inactivation preceding slower deactivation.
Nature of peak transient current.
To investigate the peak transient at positive potentials, we did single-channel measurements in 100 mM [K]o. The single-channel conductance of HERG depends strongly on [K]o and is 2 pS at 5 mM [K]o and 10 pS at 100 mM [K]o (11). From a holding potential of −80 mV, we made measurements at 0.2 Hz to a test potential of 100 mV for 300 ms to activate channels and then stepped back to −120 mV for 100 ms to rapidly remove inactivation (voltage protocol shown in Fig.2 A,bottom). We used 100 mM [K]o to make detection of HERG-channel activity possible at potentials of both 100 and −120 mV. Figure 2 A shows typical leak and capacitance current-subtracted unitary currents from one HERG channel. A striking feature is the occurrence of openings only during repolarization (Fig. 2 A, 1st through 4th recordings). During the step to −120 mV, the channel recovered from inactivation, produced bursts of openings with a mean burst duration of 15 ms, and finally entered a resting closed state. A possible interpretation of this pattern is that during depolarization the inactivated state may be entered directly from a closed state, whereas during repolarization the inactivated channel closes after visiting an open state. We refer to this gating pattern as closed-state inactivation. Our estimate of one active channel in the patch is based on the absence of any overlapping openings at −120 mV from all recordings, in this case, 160 recordings. We refer to recordings with single-channel openings present at −120 mV (84% of all recordings) as “repolarization active” (RA) recordings.
We observed three different types of channel activities. In 55% of RA recordings there were no openings during the activating pulse (Fig.2 A). Twenty-four percent of RA recordings showed early openings (<30-ms latency) during the activating pulse to +100 mV. These openings produce the initial peak transient current observed in the whole oocyte and macropatch recordings (Fig. 2 A, 5th and 6th recordings; Fig. 2 B, arrow). In 21% of RA recordings, we observed late first openings during the activating pulse (Fig. 2 A, 7th and 8th recordings). The averaged current of all 160 recordings had the same kinetics as cell-attached macropatch recordings (cf. Figs.2 B and1 B). Similar results were obtained in five patches with only one channel in each patch.
We analyzed the latency to first opening of the single-channel currents and found that the cumulative distribution could be best fitted with a biexponential function (n = 5) (Fig.2 C). The fast time constant of 14.3 ± 6.3 ms represents early openings that produce the transient peak. The second time constant of 105 ± 22 ms represents late openings that contribute to a small steady-state current at 100 mV. The weights of the rapid and slow components in the cumulative first-latency distribution were 46 ± 11 and 54 ± 11%, respectively. First latencies in Fig.2 C were calculated only for recordings with openings resulting in a maximal probability value of 1. Including depolarizations with no openings makes the maximal probability value equal to 0.375.
We analyzed recovery from inactivation in the tail currents and found that it could be fitted monoexponentially with a time constant of 8.9 ± 2.1 ms (n = 5) (Fig.2 D). A similar value was obtained from macropatches at −120 mV (6.6 ± 1.8 ms) (n = 6).
At −40 mV we observed inward single-channel openings (Fig.3 A). Patches with only one channel were not useful because openings were too rare to collect enough data for evaluation. Therefore, we used patches with approximately 5–10 channels. We analyzed patches that, in the majority of recordings, had no overlapping openings during activation. The averaged single-channel current showed slowly activating inward current identical to the macropatch recordings (Fig.3 B). HERG channels show bursting behavior (Fig. 3 A, 2nd, 3rd, 5th, 7th, and 8th recordings).
Resting inactivation at holding potential of −80 mV.
If significant channel inactivation exists at a holding potential of −80 mV, the number of channels undergoing closed-state inactivation could be vanishingly small. We evaluated the amount of steady-state inactivation at a holding potential of −80 mV by monitoring the amplitude of the initial peak transient current at 100 mV. Channels producing the peak transient current activate directly from resting closed states. The fast component of the first-latency distribution (Fig. 2 C) is similar to the inactivation rate at 100 mV in solutions with high [K+]o(27), so contributions from channels recovering from inactivation during the transient should be minimal. We used average currents from multichannel recordings to obtain sufficient data at several potentials in the same patch. Under our solution conditions, average peak currents were identical at potentials of −120 to −80 mV, indicating the absence of inactivation at −80 mV in our recordings (Fig.4). Inactivation does appear when the patch is held at −60 mV (Fig. 4,inset). We obtained similar data from three other multichannel patches. Therefore, channels that fail to open at 100 mV but that do open at −120 mV may have occupied resting closed states before depolarization, entered an inactivated state directly from a closed state during depolarization, and opened only after recovering from inactivation at −120 mV.
Comparison of single-channel currents near threshold and at 100 mV.
The single-channel currents at 100 mV (n = 11) had a mean amplitude of 0.38 ± 0.09 pA (Fig.5 A). The open probability (Fig. 5 C) showed a pattern similar to that of the macropatch measurements (Fig.1 B) with a prominent initial transient component. The open-time distribution could be fitted monoexponentially (Fig. 5 E), and we obtained an average open time of 2.5 ± 0.49 ms (n = 11) from all our data. The closed-time distribution was fitted by a sum of three exponentials with τclosed,1 = 0.78 ± 0.26 ms, τclosed,2 = 6.7 ± 4.3 ms, and τclosed,3 = 66 ± 28 ms (Fig. 5 G). The burst-duration distribution gave τburst = 5.2 ± 1.7 ms (Fig. 5 I).
Because we used multichannel patches at −40 mV to increase the frequency of channel openings, we analyzed burst data with no overlapping openings. Single-channel parameter analysis at −40 mV gave a mean amplitude of 0.32 ± 0.07 pA (n = 5) (Fig.5 B). The open-probability distribution at this potential showed a slow sigmoidal rising phase with no peak transient and was similar to that of macropatch recordings (Fig. 5 D). The open-time distribution could be fitted with a monoexponential function (Fig.5 F), and we obtained an average open time of 3.2 ± 0.53 ms (n = 5) from all our data. The closed-time distribution was triexponential with τclosed,1 = 0.95 ± 0.20 ms, τclosed,2 = 3.7 ± 0.8 ms, and τclosed,3 = 45 ± 14 ms (n = 5) (Fig.5 H). The values for τclosed,2 and τclosed,3, but not for τclosed,1, depend on the number of channels in the patch and underestimate the true values of the slower closed times. Nonetheless, three closed-time components were detected in agreement with the data at 100 mV from patches with only one active channel. The burst-duration distribution gave τburst = 14.8 ± 2.9 ms (n = 5) (Fig.5 J).
Single-channel currents between −120 and 100 mV.
We evaluated single-channel openings at several voltages in our patches. The analysis includes data from multichannel patches with few recordings displaying overlapping openings. By analyzing bursts without overlapping openings, it is possible to evaluate the amplitude, open time, rapid closed time, and burst duration (2). We found that the single-channel current-voltage relation (i-V) is linear in the inward direction and shows inward rectification in the outward direction (Fig.6 A). The calculated slope conductance in the inward direction was 9.7 pS, and the reversal potential was −7 mV. The calculated chord conductance at 100 mV was 3.5 pS. The open times were nearly voltage independent (Fig. 6 B). The rapid closed time was voltage independent (Fig.6 C). Burst duration was strongly voltage dependent with a bell shape characteristic (Fig.6 D). τburst became shorter with more depolarized potentials. During hyperpolarization, the burst duration in the tail currents was 20.2 ± 0.6 ms (n = 2) at −60 mV, reached a maximum at −80 mV [23.8 ± 4.3 (n = 3)] and −100 mV [24.6 ± 3.5 (n = 5)], and became shorter at −120 mV [14.0 ± 3.7 (n = 7)].
When patches (n = 4) were excised in a solution containing 100 mM KCl, 10 mM EDTA, and 0 mM Mg2+, the single-channel amplitude at 100 mV was not significantly altered (i = 0.39 ± 0.03 pA). In addition, the single-channel kinetics parameters at 100 mV were not significantly changed [τopen = 3.0 ± 0.77 ms, τclosed,1 = 0.92 ± 0.32 ms, τburst = 5.5 ± 0.9 ms].
Rectification of instantaneous macropatch currents.
Inward rectification of the single-channel conductance was confirmed in macropatch measurements when we measured the instantaneous macroscopic current-voltage relation (I-V) after removal of inactivation (19). Measurements were done in a 100 mM [K+]osolution in the cell-attached mode with a two-step protocol. We stepped first from a holding potential of 0 mV to a potential of −80 mV to remove inactivation and then, in a second step, to various test pulses to measure the instantaneousI-V curve (Fig.7 A). The instantaneous I-V showed that the conductance in the inward direction is virtually linear but is rectified inwardly in the outward direction (n = 7) (Fig.7 B), similar to the single-channel measurements (Fig. 6 A). With this protocol we could also measure the rate of inactivation that became faster at positive potentials, producing a crossover of the currents and indicating that inactivation is voltage dependent.
Modeling HERG currents.
We fitted kinetic models to the macropatch currents at a 100-mV test pulse potential and −120-mV tail current potential (Fig.1 B) with the nonlinear least-squares method. Activating and tail currents could be fitted simultaneously using appropriate values fori from our single-channel data. Models were constrained to have the experimentally determined open time and fast closed time. We found that the transient peak at 100 mV was the most difficult kinetics property to reproduce with the different models and could be used to evaluate the models. Therefore, we show best fits of each of the models to this transient peak (Fig.8,A–D). All of the models produced visually indistinguishable fits to the tail current data (Fig.8 E).
Our single-channel measurements suggest that HERG channels exhibit closed-state inactivation. Therefore, we tested linear models with an additional transition pathway from the fast closed state to inactivation (C1 → I) to model closed-state inactivation. A four-state model with two closed states produced a poor fit of the current data (Fig.8 A). A five-state model with three closed states produced a good fit to the transient peak at 100 mV (Fig.8 B). We then tested modified versions of this model. When we removed the C1 → I transition (Fig.8 C), the fit to the transient peak was poor. However, the five-state model without the transition pathway from the open state to inactivation (O → I) produced an indistinguishably good fit to the transient peak (Fig.8 D). This model was tested further at different test-pulse potentials with the experimentally obtained values for open and rapid closed times and the single-channel amplitudei. Initial state occupancies were specified so that all channels occupied the closed state farthest from the open state at rest. Model parameters were most sensitive to scatter in the data for the fast closed time. Because the fast closed time is voltage independent (Fig. 6 C), we used its average value calculated from data obtained at all potentials. This average value was the constraint for the fast closed time (duration of a sojourn in C1) at each potential. The number of channels was optimized for the depolarization record for 100 mV and kept constant for all other potentials. Transition rates for deactivation at −120 mV were optimized with the depolarization recording for 100 mV and kept constant for all other activation potentials. Figure9 A shows the macropatch recordings overlaid with simulated model currents at 40-, 60-, 80-, and 100-mV test pulses at which direct single-channel values were obtained. The transition rates are displayed in Table1.
Simulated traces and original measurements are very similar during activation and in the tail currents at −120 mV. As an example, the transition rates are given in an arrow diagram for the model at 40 mV (Fig. 9 C). The transition rate for closed-state inactivation (C1→ I; 0.568 ms−1) is similar to the opening rate (C1 → O; 0.588 ms−1). The probability of the channel leaving C1 to I is 48%, that of leaving C1 to O is 50%, and that of leaving C1 to C2 is 2%. This agrees with our experimentally observed frequency of closed-state inactivation of 55%. At the end of the 40-mV, 400-ms test pulse, 94% of channels have accumulated in the inactivated state I (Table2). During repolarization at −120 mV, the majority of channels open before returning to resting closed states (Fig. 9 D). The probability of leaving C1 to O is 94%, that of leaving C1 to C2 is 6%, and that of leaving C1 to I is insignificant. Channels accumulate in the C2 state during deactivation because the tail current is non-zero at the end of the pulse. We examined the ability of this model to predict currents with a different voltage protocol. The model was used to fit the macropatch data in Fig. 9 B (same as in Fig. 7 A). The voltage protocol (described in legend to Fig. 7) generates an instantaneousI-V after removal of inactivation. Model fits to the data are shown superimposed on the original recordings in Fig. 9 B. At some potentials good fits of the data to the instantaneous current immediately after the voltage step from −80 mV to positive potentials could not be obtained using mean values from our single-channel amplitude measurements. We obtained satisfactory fits to the instantaneous current by optimizing the single-channel amplitudes within the standard deviation of the amplitude measurements at a given potential. The values for the channel amplitudei, N, and transition rates for the step to −80 mV were optimized for the second voltage step to 100 mV and fixed at these values for second voltage steps to 80, 60, 40, and −40 mV. The fits overlay the currents.
Efficiency of HERG channels in repolarizing cardiac action potential.
In this paper we provide direct experimental evidence that HERG inactivates from a closed state during depolarization and opens during repolarization. A two-step protocol revealed the frequent occurrence of failures during the test pulse at positive potentials followed by openings during the tail pulse at negative potentials. To make this statement, it is important to determine precisely the number of missed events under our recording conditions. We calculated that we missed only 8% of single openings during the 100-mV pulse when we simulated single-channel currents with a 2-ms mean open time. Moreover, there is no significant steady-state inactivation of HERG at −80 mV. This indicates that many of the channels used the closed-to-inactivated state transition. In the data in Fig.2 A, 55% of recordings showed no openings during the depolarization but did show openings during the following repolarization. We estimate that 92% of these recordings were produced by failure of the channel to open during the depolarization because it inactivated from a closed state. Cloned neuronal A-type potassium channels are also known to reopen during recovery from inactivation (14). Closed-state inactivation and opening during recovery from inactivation make HERG the ideal channel to perform repolarization without shortening the plateau phase of the cardiac action potential. If inactivation were strongly coupled to opening (linear model in Fig.8 C), outward current would markedly affect early repolarization and the plateau phase of the cardiac action potential.
The trigger to begin repolarization is probably not HERG, because the outward current it contributes is small and relatively steady at this phase of the plateau. It is likely that the sum of all outward currents triggers repolarization, and once the process has been initiated, HERG performs it efficiently.
Basis of peak transient current.
HERG currents produced a small, peak transient at positive potentials, which has also been reported forI Kr recorded in human atrial (28) and ferret cardiomyocytes (13). The peak probably arises from the fraction of channels with brief first latencies that transit the open state before they inactivate. The persistent, small current (Figs. 1 B and 8) may result from a fraction of slowly activating HERG channels at 100 mV or incomplete inactivation, or a combination of both. In support of incomplete inactivation of HERG-channel activity, on rare occasions we observed second openings during depolarizations after pauses too long to be associated with bursting, suggesting that the channel inactivated and then reopened (see Fig. 2 A, 8th recording).
Steady-state inactivation of HERG channels.
We used a holding potential of −80 mV for our single-channel studies. Because channels already inactivated at −80 mV will fail to open during the step to 100 mV but will recover in the step to −120 mV, they will be indistinguishable from channels that inactivated before opening during activation. This will result in an overestimation of the frequency of closed-to-inactivated transitions. In our analysis, we assume that channels are not inactivated at the holding potential of −80 mV. Previous measurements of HERG andI Kr steady-state inactivation in 98 mM K+ are consistent with no significant steady-state inactivation at −80 mV (27), but other measurements have found a much more negative midpoint for steady-state inactivation (19). In fact, most of our evidence for 55% probability of closed-state inactivation could be accounted for by steady-state inactivation at the holding potential of −80 mV in the vicinity of 50%. This possibility has to be considered, because Smith at al. (19) obtained a midpoint for inactivation of HERG channels expressed in HEK293 cells of −90 mV.
However, the observations made by Smith et al. (19) are not directly comparable to our measurements, because they used 10 mM [K+]oand we used 100 mM [K+]o. High [K+]oshifts the midpoint of the inactivation-voltage relation ∼20–30 mV (27, 30), which is consistent with our results. In addition, measurements of HERG inactivation-voltage relations are complicated by rapid channel closing at potentials less than about −60 mV, requiring short-duration voltage steps. Two protocols have been used to measure steady-state inactivation (19, 27). Because of the small conditioning-pulse durations employed and HERG-inactivation time constants that can be similar to the pulse duration in solutions with high [K+]o, the inactivation-voltage relations may deviate significantly from steady state.
The presence of inactivation at −80 mV can be critically tested by measuring availability at more negative conditioning potentials. We focused on the initial peak transient current at 100 mV as an index of channel availability because long (many seconds) holding potentials can be used and because the experiments are the same as those for our single-channel recordings, except for the holding potential. Although the current at 100 mV is small in amplitude, it is generated primarily by channels that occupied resting closed states before the step to 100 mV (see Resting inactivation at holding potential of −80 mV). These channels open for the first time and then inactivate during the time of the transient (∼30 ms). At 100 mV, channels are unlikely to reopen from inactivated states during the time of the peak transient, so peak current amplitude should be proportional to the number of resting channels available for activation at a particular potential. We found that the peak transient was unchanged by varying holding potential from −120 to −80 mV but did become reduced at a holding potential of −60 mV. This result is in agreement with the steady-state inactivation-voltage relation described by Wang et al. (27) for HERG channels measured with 98 mM [K+]oin oocytes. Kiehn et al. (10) measured the steady-state inactivation-voltage relation with a protocol similar to that of Smith et al. (19) and found with 5 mM [K+]othat the midpoint of inactivation was −68 mV. At the same [K+]oused by Smith et al. (19), Taglialatela et al. (22) obtained a midpoint of −62 mV for HERG channels expressed in oocytes. Because we expect a positive shift of the inactivation-voltage relation at higher [K+]o, in our solutions the midpoint of the inactivation-voltage relation will be positive to these values by at least 10–20 mV. The difference in the midpoint values measured with similar protocols suggests that HERG channels expressed in oocytes are not equivalent to HERG channels expressed in HEK293 cells.
HERG and IKr have similar properties.
Single-channel kinetics parameters and conductance of HERG were similar to reported values forI Kr.I Kr unitary currents measured in sinoatrial nodal cells of the rabbit heart have (in 150 mM [K]o) an inward single-channel conductance of 11.1 pS (18) or inward/outward conductances of 10.8/7.8 pS (7), with values of 10/3 pS in guinea pig atrial myocytes (5) and 10.8 pS (inward) in rabbit ventricular myocytes (25). I Kr in human ventricular myocytes has an inward single-channel conductance of 12.9 pS in 140 mM [K]o(26). Measurements of single HERG channels expressed in oocytes found an inward/outward conductance of 12.1/5.1 pS in 120 mM [K]o (29). Our value for HERG unitary inward/outward current was 9.7/3.5 pS in 100 mM [K]o. The conductance properties for HERG andI Kr are in good agreement.
I Krsingle-channel kinetics analysis shows evidence for a single open state and two closed states in rabbit sinoatrial nodal cells (18) and guinea pig atrial myocytes (5). The open state in 150 mM [K]o in sinoatrial nodal cells is short lived (τopen = 2.5 ms at −60 mV) relative to the value for guinea pig atrial myocytes (τopen = 9 ms at −100 mV). In guinea pig atrial myocytes the relatively large value of τopen reported forI Kr may actually represent a mean burst duration (τburst), because the authors describe the distribution used to generate τopen as the distribution of burst durations and used a relatively slow sampling rate for data acquisition (5).
HERG channels expressed in oocytes are reported to have open dwell-time constants similar to those forI Kr. However, one report has found kinetics evidence from HERG recordings for two open states. HERG open-time distributions in 120 mM [K]o were best fit with biexponential distributions with mean values at −90 mV of 2.9 and 11.8 ms for the fast and slow components, respectively (29). We did not find evidence for a statistically significant second, kinetically distinct open state in our HERG data and obtained τopen = 2.8 ms at −120 mV and τopen = 2.5 ms at 100 mV. Mean open times in our data were weakly voltage dependent, becoming shorter with increasing depolarization as reported previously for HERG (29) and I Kr(18).
I Kr values in 150 mM [K]o for mean fast and slow closed times in rabbit sinoatrial nodal cells are biexponential with values of 0.7 and 17.6 ms at −60 mV (18), and comparable values of 1.2 and 37 ms at −100 mV were obtained from guinea pig atrial myocytes (5). Closed-time distributions from HERG expressed in oocytes are also biexponential and yield estimates for the mean fast and slow closed times of 0.54 and 14.5 ms, respectively, at −90 mV that are similar to the values forI Kr (29). Our closed-time distributions are distinguished from previous measurements in observing three closed-time distribution components. The additional closed-time component in our data has a value (6.7 ms at 100 mV) intermediate to the fast and slow components reported for HERG andI Kr. The slowest component in our closed-time data has a time constant (66 ms at 100 mV) that is larger than that in previous reports. The mean fast closed time was voltage independent in our data, as in data forI Kr (18), and was not significantly voltage dependent for HERG (29). Previous data from HERG and I Krunitary currents are consistent with burst gating of the channels. Our data shows that burst duration has a bell-shaped voltage dependence and suggests that bursts may be organized into clusters.
Inward rectification of HERG conductance.
The single-channel conductance of HERG rectifies inwardly (Fig.6 A), as does the instantaneous tail current I-V (Fig.7 B) in our macropatch recordings. When patches were excised in a solution containing zero Mg2+ and zero Na+, the single-channel rectification was not changed, similar to that forI Kr (7). This rectification is therefore not produced by soluble internal blocking particles (30) and may be intrinsic to the ion-conduction pathway itself. However, rectification of macroscopic currents appears to be due primarily to voltage-dependent gating of the channel, resulting in reduced open probability at depolarized potentials (19, 21). Rectification of the single-channel conductance only appears at much more positive potentials. Therefore, rectification produced by HERG pore structures appears to be more of biophysical than of physiological interest. Our instantaneous tail current measurements are similar to other results (19, 21) that showed a linear instantaneousI-V up to 40 mV. These authors have invoked C-type inactivation as an explanation.
A model of HERG kinetics.
Others have successfully used a linear five-state kinetics scheme to quantitatively model macroscopic HERG currents expressed in oocytes without single-channel constraints (27). From our data a minimum five-state constrained model was also required to adequately fit all our experimental data (Figs. 8 and 9). Variations of this model showed that inactivation exclusively from the proximal closed state (Fig.8 D) was better able to fit the data than inactivation exclusively from the open state (Fig.8 C). This supports a role for closed-state inactivation during channel activation and is consistent with our single-channel data. We chose to analyze data with the model without open-state inactivation (Fig.8 D) because it fit the data as well as the more general model (Fig. 8 B) with fewer parameters. We have no experimental data demonstrating the absence of open-state inactivation, so we cannot exclude the more complex general model (Fig. 8 B), especially at potentials that fail to produce a transient current. All of the tested models can produce good fits to noninactivating plateau currents during depolarization and the large tail current on repolarization. In the absence of experimental constraints, simpler four-state models can also produce good fits of the data. The open time and the rapid closed time of HERG are nearly voltage independent (Fig.6, B andC). The same has been reported for cloned ShakerK+ channels in the absence of N-type inactivation (6). However, in the models with closed-state inactivation, there is an inverse relationship in the voltage dependence of the rates from the fast closed state to the inactivated state and the open state such that their sum is voltage independent, as required by the voltage independence of the fast closed time constraint. In these models this is the mechanism responsible for strong inward rectification of HERG currents.
In summary, our data provide more evidence thatHERG encodesI Kr in cardiomyocytes. Its exceptional kinetic features make HERG an efficient channel for producing the downstroke of the cardiac action potential (21). Because I Krin cardiomyocytes is difficult to separate from other currents,HERG expressed inXenopus oocytes is a satisfactory system in which to study the kinetics ofI Kr.
J. Kiehn and A. E. Lacerda contributed equally to this work.
Address for reprint requests and other correspondence: A. E. Lacerda, Rammelkamp Center, 2500 MetroHealth Drive, Cleveland, OH 44109-1998 (E-mail:).
We thank Dr. G. Kirsch for comments on the manuscript, P. Kiehn and Dr. W. Q. Dong for technical assistance, and Dr. M. Keating for providing the HERG clone.
This study was supported by a Deutsche Forschungsgemeinschaft Grant (to J. Kiehn) and National Heart, Lung, and Blood Institute Grants HL-37044 and HL-36930 (to A. M. Brown).
Present address of J. Kiehn: Dept. of Cardiology, Medical Univ. Hospital, Bergheimerstr. 58, 69115 Heidelberg, Germany.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.
- Copyright © 1999 the American Physiological Society