The hERG channel has a relatively slow activation process but an extremely fast and voltage-sensitive inactivation process. Direct measurement of hERG's gating current (Piper, D.R., A. Varghese, M.C. Sanguinetti, and M. Tristani-Firouzi. 2003. PNAS. 100:10534–10539) reveals two kinetic components of gating charge transfer that may originate from two channel domains. This study is designed to address three questions: (1) which of the six positive charges in hERG's major voltage sensor, S4, are responsible for gating charge transfer during activation, (2) whether a negative charge in the cytoplasmic half of S2 (D466) also contributes to gating charge transfer, and (3) whether S4 serves as the sole voltage sensor for hERG inactivation. We individually mutate S4's positive charges and D466 to cysteine, and examine (a) effects of mutations on the number of equivalent gating charges transferred during activation (za) and inactivation (zi), and (b) sidedness and state dependence of accessibility of introduced cysteine side chains to a membrane-impermeable thiol-modifying reagent (MTSET). Neutralizing the outer three positive charges in S4 and D466 in S2 reduces za, and cysteine side chains introduced into these positions experience state-dependent changes in MTSET accessibility. On the other hand, neutralizing the inner three positive charges in S4 does not affect za. None of the charge mutations affect zi. We propose that the scheme of gating charge transfer during hERG's activation process is similar to that described for the Shaker channel, although hERG has less gating charge in its S4 than in Shaker. Furthermore, channel domain other than S4 contributes to gating charge involved in hERG's inactivation process.
Human ether-a-go-go related gene encodes the pore-forming subunits (hERG) of rapid delayed rectifier (IKr) channels expressed in the heart and several other cell types (Sanguinetti et al., 1995; Zhou et al., 1998; Emmi et al., 2000; Rosati et al., 2000; Shoeb et al., 2003). The hERG/IKr channel plays an important role in maintaining the electrical stability of the heart, as is suggested by the linkage between inherited or acquired long QT syndrome and mutations in the hERG gene or drug suppression of IKr. Relative to other voltage-gated K (Kv) channels, hERG manifests two unique gating properties. First, its activation process is relatively slow. At +50 mV, the time constant (τ) of activation is >50 ms for hERG, but <2 ms for the Shaker channel (Zagotta et al., 1994). Second, the inactivation and recovery from inactivation processes of hERG are extremely fast and voltage sensitive. Inactivation of hERG is due to conformational changes around the outer mouth region (Smith et al., 1996), similar to the “C-type” inactivation process first described for the Shaker channel (Hoshi et al., 1991). The τ of inactivation in the hERG channel ranges from 6.5 to 2.2 ms in the voltage range of −10 to +30 mV (corresponding to a gating charge of ∼0.7 eo per channel), while τ of C-type inactivation in the Shaker channel in this voltage range is ∼2,000 ms and voltage insensitive (Hoshi et al., 1991). An important task in understanding the function of the hERG channel is to identify the structural basis for its unique gating properties.
hERG shares the basic structure design with other Kv channels (Fig. 1 A). It has four subunits arranged symmetrically around a central pore, with each subunit containing a voltage-sensing domain (S1–S4) and a pore domain (S5-P-loop-S6). Fig. 1 B aligns the amino acid sequences of transmembrane segments in the voltage-sensing domains of hERG, Shaker, bEAG (a close relative of hERG), and KvAP (a prokaryotic Kv channel whose crystal structure was solved) (Jiang et al., 2003a). Note that there is no consensus as to how the S4's positive charges in hERG or bEAG should be aligned with those in the Shaker (Schonherr et al., 2002; Smith and Yellen, 2002; Silverman et al., 2003), because a direct comparison of functional roles of these positive charges among the channels has not been available.
In terms of charge distribution in the voltage-sensing domain, hERG (as well as bEAG) differs from Shaker in two aspects. First, the S4 domain in hERG has one less positive charge than that of Shaker. Second, in addition to the three negative charges in S2 and S3 that are well conserved in all Kv channels including Shaker (labeled as −2, −4, and −5; Fig. 1), hERG has three extra negative charges in its voltage-sensing domain (−1, −3, and −6; Fig. 1). The functional role of the positive charges in Shaker's S4 segment have been extensively studied (Aggarwal and MacKinnon, 1996; Seoh et al., 1996; Baker et al., 1998; Bezanilla, 2000; Cohen et al., 2003; Starace and Bezanilla, 2004; Horn, 2004). The generally accepted scheme is that the first four positive charges sense the transmembrane voltage and change their positions relative to the membrane barrier during channel activation (Aggarwal and MacKinnon, 1996; Seoh et al., 1996; Starace and Bezanilla, 2004). Therefore, these four residues are the “gating charges” in the Shaker channel. It is not known which of the positive charges in hERG's S4 segment carry gating charges during channel activation. Presumably the number of gating charges in hERG's S4 segment can influence its voltage sensitivity and thus the rate of activation (Logothetis et al., 1992, Islas and Sigworth, 1999, Starace and Bezanilla, 2004). The functional role of negative charges in the transmembrane segments of voltage-sensing domain has been well described for the Shaker and EAG channels (Tiwari-Woodruff et al., 2000; Papazian et al., 2002). These negative charges stabilize S4's positive charges by forming salt bridges. As S4 moves relative to the other transmembrane segments during gating, different charge pairings can occur that serve to stabilize the channel in the closed or open state (Tiwari-Woodruff et al., 2000; Papazian et al., 2002). Furthermore, in EAG channels, the extra negative charges in S2 and S3 can form a divalent cation (M2+) binding site in the extracellular crevice around S4 (Silverman et al., 2000; Schonherr et al., 2002). It has been proposed that M2+o (Ca2+o or Mg2+o) binding here can restrict the crevice size and retard outward S4 movement during membrane depolarization, slowing down the channel's activation process (Silverman et al., 2000; Schonherr et al., 2002). Therefore, both positive and negative charges in the voltage-sensing domain contribute to determining the rate of channel activation.
Fig. 1 C aligns the amino acid sequences lining the outer mouth of hERG and other Kv channels. There are two unique features in hERG (and bEAG). First, hERG lacks some H bonds between residues around the outer mouth. Such H bonds have been proposed to exist in Shaker and other Kv channels (Doyle et al., 1998, Larsson and Elinder, 2000). Specifically, H bonds can be formed between the carboxylate side chain outside S5 (E418 in Shaker) and an amido nitrogen of the peptide backbone in the P-S6 linker (V450/G451 in Shaker) (Larsson and Elinder, 2000), and between nitrogens of two tryptophans and the hydroxyl group of a tyrosine at the two ends of the P-loop (W434/W435 and Y445 in Shaker) (asterisks connected by horizontal lines in Fig. 1 C) (Doyle et al., 1998). These H bonds serve to stabilize the outer mouth in the open state, slowing down the rate of C-type inactivation (Yang et al., 1997; Larsson and Elinder, 2000). In hERG, the corresponding residues cannot form H bonds (Fig. 1 C). A reduced H-bonding capability in hERG can make its outer mouth more flexible than in the Shaker channel. Second, the S5-P linker is much longer in hERG (and bEAG) than in other Kv channels. Previously, we have shown that the central segment of the S5-P linker in the hERG channel (positions 583–597) plays a critical role in determining the degree of inactivation and K+ selectivity (Liu et al., 2002). NMR spectroscopy showed that this segment may form an α-helix (termed “S5-P helix” in the following text) (Torres et al., 2003). We proposed that this S5-P helix can assume dynamic conformational changes during membrane depolarization, coming close to the pore entrance and participating in the outer mouth function (unpublished data). These features can contribute to the fast rate of inactivation in the hERG channel. However, one central question remains: what controls the voltage sensitivity of hERG's inactivation process? Arguments both for (Tristani-Firouzi, 2004) and against (Johnson et al., 1999) S4 as the voltage sensor for hERG's inactivation process have been made. In the latter case, no specific structural correlate has been proposed.
This study was designed to address the following questions: (1) which of the positive charges in hERG's S4 segment carry gating charges during channel activation, and (2) do they also serve the voltage sensor function for the inactivation process? In the Shaker channel, a negative charge at the cytoplasmic end of S2 (E293) may contribute to gating charge transfer (Seoh et al., 1996). Therefore, we also include the equivalent charge (D466) in our study. Fig. 1 D lists the hERG residues examined in this study and the corresponding residues in the Shaker channel. We apply two strategies in our experiments: (1) mutating charged residues and monitoring the effect on equivalent gating charges involved in activation and inactivation processes, and (2) testing the sidedness and state dependence of accessibility of cysteine side chains introduced into these charged-residue positions to a membrane-impermeable thiol-modifying reagent (MTSET). We discuss our results based on the “conventional gating model” (Ahern and Horn, 2004), instead of the recently proposed “paddle gating model” (Jiang et al., 2003b), because the latter model has been seriously challenged by experimental findings from Shaker and other Kv channels (Broomand et al., 2003; Cohen et al., 2003; Gandhi et al., 2003; Laine et al., 2003; Lee et al., 2003; Ahern and Horn, 2004; Horn, 2004; Starace and Bezanilla, 2004).
Materials And Methods
hERG in a vector, pGH19 (a gift from G.A. Robertson, University of Wisconsin, Madison, WI), was subcloned into the Kpn I/Xba I site of pAlterMax. Mutagenesis was performed using the oligonucleotide-directed method and a commercial kit (Alter site II in vitro Mutagenesis System; Promega). Mutations were confirmed by direct DNA sequencing around the mutation sites. For transcription, plasmids were linearized by Not I and transcribed using T7 RNA polymerase using a commercial kit (mMessage mMachine; Ambion). The cRNA products were run on denaturing RNA gel along with RNA size markers, and their quality and quantity were evaluated using densitometry (ChemiImager model 4400; α-Innotech Corp). The mutant is designated by wild-type (WT) amino acid, followed by position number and the substituting amino acid using one-letter code.
Oocytes were isolated as described before (Tseng-Crank et al., 1990) and incubated in an ND96-based medium (composition given below), supplemented with 10% horse serum and penicillin/streptomycin at 16°C. 2–4 h after isolation, each oocyte was injected with 40 nl of cRNA solution (containing cRNA of 4–18 ng) using a Drummond digital microdispenser. Oocytes were incubated in the above medium at 16°C and studied 2–4 d after cRNA injection.
Voltage Clamp Experiments
Oocytes expressing cysteine-substituted mutants were pretreated with DTT (5 mM) before recording. Membrane currents were recorded from whole oocytes using the “2-cushion pipette” voltage clamp method (Schreibmayer et al., 1994). Both current-passing and voltage-recording pipettes had tip resistance 0.1–0.3 MΩ. During recordings, the oocyte was continuously superfused with a low-Cl ND96 solution to reduce interference from endogenous Cl channels. Voltage clamp was done at room temperature (24–26°C) with OC-725B or OC-725C amplifier (Warner Instruments). Voltage clamp protocol generation and data acquisition were controlled by pClamp5.5 via a 12-bit D/A and A/D converter (DMA; Axon Instruments). Current data were low-pass filtered at 1 kHz (Frequency Devices) and stored on disks for off-line analysis.
The voltage clamp protocols and methods of data analysis are described in figure legends and text. The following software was used for data analysis: pClamp6 or 8, EXCEL (Microsoft), SigmaPlot, SigmaStat, and PeakFit (SPSS).
Cysteine Side Chain Modification
MTSET (Toronto Research Chemicals, Inc.) powder was dissolved in deionized water at 0.1 M shortly before experiments. The stock solution was stored on ice and used within 2 h. For extracellular application, after control data were obtained, the MTSET stock solution was diluted with bath solution to 1 mM and applied to the oocyte immediately. For intracellular application, a third fine-tipped pipette filled with MTSET stock solution was used to impale the oocyte. After confirming the stability of membrane currents, MTSET was injected into the oocyte. Assuming an oocyte volume of 0.5 μl (corresponding to a sphere of 1 mm diameter), injecting 10 nl of the stock solution would produce a cytoplasmic concentration of 2 mM immediately after the injection. The injection was controlled by the Drummond digital microdispenser as was used for cRNA injection.
The ND96 solution had the following composition (in mM): NaCl 96, KCl 2, CaCl2 1.8, MgCl2 1, HEPES 5, Na-pyruvate 2.5, pH 7.5. The low-Cl ND96 used during voltage clamp experiments was made with Cl− ions in ND96 replaced by methanesulfonate. In some experiments, [K] in low-Cl− ND96 was raised to 20 mM. Equimolar [Na] was reduced to maintain the osmolality.
Effects of Charge Mutations on the Voltage Dependence of hERG Activation and Pore Domain Function
We first characterize the effects of charge mutations on the hERG channel function. The purpose is to evaluate how serious these single charge mutations can affect the structure–function relationship of the hERG channel. This information is important for the following experiments where we examine the effects of charge mutations on the channel's gating charges and use substituted cysteine side chains to deduce the accessibility of native side chains in the WT channel. Fig. 2 A shows test pulse currents of WT hERG and mutant channels during depolarization to various levels of Vt, and tail currents upon repolarization to Vr (Vt range and Vr marked adjacent to current traces). For each cell, the peak amplitudes of tail currents are normalized by the maximum tail current induced by a strong positive Vt to the plateau level of channel activation. This gives an estimate of the macroscopic channel open probability (Po) at the end of the preceding test pulses. The relationship between Vt and Po is used to construct the activation curves as shown in Fig. 2 B. For WT-hERG and all the mutant channels, the activation curves can be well described by a simple Boltzmann function:
where V0.5 and k are the half-maximum activation voltage and slope factor. The numerical values of V0.5 and k for the WT and mutant channels are listed in Table I.
Neutralizing the positive charges at the two ends of S4, K525C, and K538C causes a prominent hyperpolarizing shift in the activation curve and induces channel opening at the negative holding voltage (Fig. 2 A, white arrows). R534C also causes a modest but statistically significant hyperpolarizing shift in the activation curve. Thus, these three mutations stabilize the open state, relative to the closed state, of the hERG channel. Neutralizing the other three positive charges in S4 (R528C, R531C, and R537C) and D466 in S2 causes differing degrees of depolarizing shift in the activation curve, indicating a stabilizing effect on the closed states relative to the open state of the hERG channel.
Fig. 3 A depicts current traces of WT-hERG and mutant channels recorded at different levels of Vr after channel activation and inactivation by a strong depolarization pulse to +60 mV. For each cell the peak or plateau levels of tail currents are normalized by the maximum outward tail current recorded at −50 to −30 mV, and the data are averaged over cells (current–voltage relationships shown in Fig. 3 B). In all cases, there is a prominent negative slope in the Vr range of −50 to +30 mV, i.e., smaller outward currents at more depolarization voltages. This “inward rectification” property reflects the fast and voltage-sensitive inactivation process of the hERG channel. Furthermore, except for K525C, the reversal potentials (Erev) for WT and all the other mutant channels are at about −100 mV. This is close to the calculated K+ equilibrium potential (EK) under our recording conditions (−105 mV, assuming [K]i of 125 mM) (Dascal, 1987). Since K525C has a sizable open probability at the oocytes' resting membrane potential (Fig. 2 A), the [K]i may be lower than 125 mM, causing a positive shift of EK to ∼−80 mV. This is supported by the observation that a total replacement of external Na+ ions with NMDG shifts Erev of K525C from −81.8 + 1.1 to −86.1 + 0.1 mV (n = 3 each), corresponding to a permeability ratio of Na+ to K+ of only 0.003 (calculated based on the constant field equation).
How serious is the perturbation of hERG's structure–function relationship by these single charge mutations? Mutation-induced changes in the free energy of channel activation (ΔΔGo) are listed Table I. The highest degree of perturbation is seen in K538C (ΔΔGo = −8.7 ± 1.0 kcal/mol). This change amounts to the energy of less than two H bonds in aqueous solution (4–5 kcal/mol per H bond), suggesting that the degrees of voltage shift shown in Fig. 2 B can be due to subtle changes in the channel's conformation. Fig. 3 shows that these charge mutations do not disrupt the function of the pore domain (fast inactivation process and high K+ selectivity). Therefore, these charge mutations do not drastically alter the structure–function relationship of the hERG channel, justifying their use in the following experiments.
Effects of Charge Mutations on the Number of Equivalent Gating Charges Involved in hERG Activation
The slope factors of Boltzmann fit to activation curves can be used to calculate the number of gating charges of channel activation (zg, Table I). Except for D466C, the zg values of the mutant channels are similar to, or even larger than (K538C), the zg value of WT-hERG. Since zg values are calculated based on a simple two-state channel gating model and can be influenced by cooperative interactions between subunits, they are not sensitive or accurate enough to deduce the number of gating charges involved in the activation process. Therefore, we use a less model-dependent “limiting slope” method to estimate the number of equivalent gating charges for hERG activation (Zagotta and Aldrich, 1990, Bezanilla, 2000):
where za is equivalent gating charges (in eo) transferred during channel activation, dLn(Po)/dV is the slope of relationship between Ln(Po) and depolarization voltage V, and k and T are the Boltzmann constant and absolute temperature. This method is valid for a sequential model with one open state, where the transition rates are exponentially dependent on the voltage.
A prerequisite in using the limiting slope method to deduce the za value is the ability to measure very low Po values (estimated by normalized peak tail current amplitudes, as described for Fig. 2 B). To facilitate the measurement of extremely small tail currents at negative voltages, we record tail currents in 20 mM [K]o (raising [K]o to increase single channel conductance) (Kiehn et al., 1996) and use a negative Vr of −120 mV (to increase the driving force for inward tail currents). Furthermore, two voltage clamp protocols are used. The first protocol applies 1-s pulses in 10-mV increments to sample the full range of channel activation, and thus to measure the maximum amplitude of tail current in the plateau range of channel activation. The second protocol applies long (10 s) depolarizing pulses in 2-mV increments, starting at Vt ∼ 20 mV below the apparent activation threshold to Vt 10–20 mV above the activation threshold (based on Fig. 2). The long (10 s) depolarization duration allows Po to fully develop at negative Vt, when channel activation is slow. The small (2 mV) steps increase the resolution of changes in Po around the activation threshold. Fig. 4 A depicts WT current traces elicited by these two voltage clamp protocols. Under these conditions, we can detect Po values at 10−3 or even lower for WT-hERG and mutant channels.
In Fig. 4 B, the Po values are plotted on a logarithmic scale against Vt for WT (gray open circles) and mutant channels (black open circles). Data are averaged from three to six measurements each. For each measurement, the relationship between Ln(Po) and Vt is subjected to linear regression analysis to estimate the slope at the lower limit of Po. This slope is then used to calculate za according to Eq. 2. The results for WT-hERG and mutant channels are plotted in Fig. 4 C. Neutralizing the first three positive charges in S4 (K525C, R528C, and R531C) and the negative charge in S2 (D466C) reduces za, supporting their role as “gating charges” for hERG activation. On the other hand, neutralizing the inner three positive charges in S4 (R534C, R537C, and K538C) has no effect on za.
MTSET Accessibility Test of Cysteine Side Chains Introduced Into Charged-residue Positions in S4 and S2
If a charge is within the membrane electrical field and thus capable of sensing changes in the membrane voltage, its position relative to the membrane barrier will change during channel gating. Cysteine introduced into its position is expected to change the thiol side chain accessibility to the surrounding aqueous crevice in a state-dependent manner (assuming that the cysteine substitution does not alter the native side chain orientation) (Wang et al., 1999). On the other hand, charged residues not involved in gating charge transfer do not change their positions relative to the membrane barrier. Their side chain accessibility to the aqueous crevice should not be state dependent. Therefore, to confirm the role of the first three positive charges in S4 and D466 in S2 as the “gating charges” in hERG's activation process, we test the accessibility of introduced cysteine side chains to extracellular and intracellular MTSET (MTSETo and MTSETi). Ideally, accessibility to MTSETi should be tested using the inside-out patch recording configuration and applying MTSET directly to the intracellular side of the membrane (Baker et al., 1998, Bell et al., 2004). Unfortunately, due to the low expression level of cysteine mutants in oocytes (in 2 mM [K]o maximal whole cell outward currents <10 μA), it is difficult to record large and stable currents from inside-out patches. Thus, we take an alternative approach: applying MTSETi using intra-oocyte injection.
Neither MTSETo nor MTSETi Affects the WT hERG Channel
There are 24 native cysteines in WT hERG. Previously, we have shown that MTSETo (1 mM) has no effects on WT hERG (Fan et al., 1999). Fig. 5 shows that MTSETi (estimated initial cytoplasmic concentration 2 mM) does not affect WT hERG either. The efficacy of intra-oocyte injection is confirmed by the observation that injecting TEA (estimated cytoplasmic concentration 2 mM) causes a prominent decrease in the WT hERG current amplitude (Smith et al., 1996). These data indicate that either native cysteines in hERG are not accessible to MTSETo and MTSETi or their modification by MTSET has no impact on channel function. Therefore, the effects of MTSETo and MTSETi on cysteine-substituted mutants can be attributed to modification of the introduced cysteine side chains.
Cysteine Side Chain at 525 is Accessible to MTSETo in a State-dependent Manner but Inaccessible to MTSETi
The most prominent effect of K525C modification by MTSET is a slowing of channel deactivation. Therefore, we use the half-time (T1/2) of decay of tail current at −120 mV to quantify the degree of K525C modification by MTSET. Three protocols are used to evaluate the state dependence of K525C modification by MTSETo. MTSETo is applied (1) while the oocyte membrane is held at Vh −80 mV to keep most of the channels in the closed states (Fig. 6 A), (2) while the membrane is held at 0 mV to keep the channels in the open/inactivated states (Fig. 6 B), or (3) while the membrane voltage is pulsing from Vh −90 mV to +40 mV for 1 s once every 30 s to cycle the channels between closed and open/inactivated states (Fig. 6 C). For the first two protocols, MTSETo (1 mM) is applied for 5 min (flow rate 5–6 ml/min, bath solution totally exchanged in <30 s), and the effect on T1/2 of channel deactivation is measured after 5 min washout of MTSETo. Fig. 6 B shows that applying MTSETo while holding the channels in open/inactivated states allows K525C modification by MTSET; T1/2 is markedly prolonged to 4.2 ± 0.7 of control (n = 8). Fig. 6 A shows that applying MTSETo while holding the channels in the closed state prevents K525C modification by MTSET (n = 4). Using the third protocol, MTSETo exposure causes a gradual slowing of K525C deactivation (T1/2 prolonged to 2.4 ± 0.2 of control, n = 7). Since K525C modification by MTSETo occurs in the open/inactivated state, the cumulative exposure time to MTSETo in the open/inactivated state can be calculated (additional abscissa below the data points during MTSETo exposure in Fig. 6 C). The time course of K525C modification by MTSETo can be well described by a single exponential function. The τ value is 4 mM·s, which translates into a modification rate of 250 M−1s−1. Note that T1/2 prolongation after exposure of K525C to MTSETo at Vh 0 mV is not sustained, but gradually declines (to 3.2 ± 0.4 of control at steady state, n = 8). This degree of T1/2 prolongation is not statistically different from that reached when MTSETo is applied during constant pulsing (P = 0.106). This suggests that some of the 525C residues might have lost their MTSET adduct when the channels cycle through the closed–open states, perhaps because the gating pore around S4 is too narrow for the MTSET-modified 525C side chain to go through easily.
Fig. 6 D shows that MTSET applied to the cytoplasmic side of the cell membrane does not affect the K525C tail kinetics even after a 20-min exposure. After MTSET injection, extracellular application of MTSETo induces a marked prolongation of tail current, supporting the modifiability of the K525C side chain. Similar observation was obtained in five experiments.
Cysteine Side Chains Introduced into the Other Positive-charge Positions in S4 Are Accessible to MTSETi but not to MTSETo
Fig. 7 summarizes the MTSET accessibility of cysteine side chains introduced into the other five positive-charge positions in S4: R528C, R531C, R534C, R537C, and K538C. The major effect of MTSET modification on these mutant channels is a suppression of current amplitude (Fig. 7 B, insets). Therefore, we use the decrease in the peak tail current amplitude as a measure of MTSET modification of introduced cysteine side chains at these positions.
The left panel of Fig. 7 A shows MTSETo application to R528C for 5 min while the membrane voltage is held at +20 mV (reaching an average of 80% channel activation, Fig. 2 B, or a total exposure time of 528C to MTSETo in the open/inactivated state of 240 mM·s). There is no sign of R528C modification by MTSETo under this condition. The remaining four panels of Fig. 7 A depict time courses of changes in the peak tail current amplitudes of R531C, R534C, R537C, and K538C before, during, and after MTSETo (1 mM) exposure (n = 4–7 each). MTSETo does not affect these mutant channels. These observations indicate that cysteine side chains introduced into these five positions in S4 are not accessible to MTSETo.
For all the five cysteine mutants, applying MTSET to the cytoplasmic side of the cell membrane causes a gradual decrease in the current amplitude, indicating that cysteine side chains at all five positions are accessible to MTSETi (Fig. 7 B). The state dependence of MTSETi accessibility is tested by comparing the degree of current reduction when MTSETi is applied under two voltage clamp protocols. For the first protocol, MTSET is injected into the oocyte while the membrane voltage is held at −80 mV with only brief (0.5 s) depolarization pulses applied once every 30 s to monitor the degree of current reduction (“closed state–preferred” exposure, represented by solid circles). For the second protocol, MTSET is injected into the oocyte while the membrane is held at a depolarized Vt for 20 s using a duty cycle of 30 s (“open/inactivated state–preferred” exposure, represented by open circles, details in Fig. 7 legend). For each of the mutant channels, the average time courses of MTSETi modification using these two protocols are plotted in the same graph to facilitate comparison (n = 3–8 each). When MTSETi is applied using the closed state–preferred exposure protocol, the current amplitude gradually declines following an apparent single exponential time course. The time constants are 5.6 min (R528C), 4.5 min (R531C), 3.3 min (R534C), 6.6 min (R537C), and 8.9 min (K538C). On the other hand, when MTSETi is applied using the open/inactivated state–preferred exposure, the degree of MTSETi modification is significantly reduced for R528C and R531C (P < 0.05 for data points following the first 2 min after MTSET injection). The observation that the apparent degree, instead of the rate, of current suppression by MTSETi is state dependent may be explained by (a) MTSET hydrolysis by the reducing environment of oocyte cytoplasm, and (b) slow rates of MTSETi modification of 528C and 531C. There is no statistically significant difference in R534C or R537C modification by MTSETi between the closed and open/inactivated state–preferred exposures. MTSETi modification of K538C proceeds more rapidly when MTSETi is applied using the open/inactivated state–preferred exposure than using the closed state–preferred exposure.
Cysteine Side Chain at Position 466 in S2 Is Accessible to MTSETo in a State-dependent Manner
Previously we showed that 466C is accessible to MTSETo, and the accessibility is higher when the channel is in the closed state than in the open/inactivated states (Liu et al., 2003). We now test whether 466C is accessible to MTSETi and whether this accessibility manifests a state dependence opposite to that of MTSETo accessibility. Fig. 8 (A and B) shows two representative time courses of change in D466C current amplitude (measured by the peak tail currents at −60 mV, insets) before and after impalement with a pipette containing 100 mM MTSET solution, and after MTSETi injection (estimated initial cytoplasmic concentration 2 mM). The membrane voltage is cycling through open/inactivated states (1 s to +60 mV, once every 30 s) and the closed state (Vh −80 mV). There is a very gradual and modest reduction in the D466C current amplitude (by ∼20% after 15 min exposure). On the other hand, when the oocyte is subsequently exposed to 1 mM MTSETo for 5 min while the membrane is held at −80 mV (closed state–preferred exposure) and then washed out, the D466C current is practically abolished (Fig. 8 A). In Fig. 8 B, the oocyte is exposed to 1 mM MTSETo for 5 min while the membrane is held at +20 mV (open/inactivated state–preferred exposure) followed by washout, there is a further but very modest modification of D466C by MTSETo. Similar observations are obtained in five experiments. These data confirm that D466C is more accessible to MTSETo in the closed state than in the open/inactivated states. However, D466C has only limited accessibility to MTSETi with no clear state dependence (Fig. 8 C).
Effects of S4 Charge Mutations on the Voltage Sensitivity of hERG's Inactivation Process
Fig. 3 shows that there are no major changes in the degree of hERG inactivation when S4's positive charges are substituted by cysteine individually. We further quantify the voltage sensitivity of inactivation in WT hERG and the cysteine mutants. A triple-pulse protocol is used to estimate the rate of channel inactivation at different voltages (Fig. 9 A, top). A long and strong depolarizing pulse to +60 mV (+80 or +100 mV for R531C) fully activates and inactivates the channels. This is followed by a short repolarizing step to −80 mV for 15 ms, during which the inactivated channels recover from inactivation (τ ∼ 3 ms) without appreciable deactivation (the fast and slow time constants of deactivation range 20–100 and 200–1,000 ms, respectively, at −80 mV; unpublished data). The membrane is then stepped to test voltages ranging from +40 to −20 mV, during which the channels reinactivate. For WT and mutant channels, this process can be well described by a single exponential function, as is shown by the curve fitting to representative current traces in Fig. 9 A. The values of τinactivation are plotted on a semilogarithmic scale versus the test voltages (Vtest) in Fig. 9 B. A linear regression analysis of the reciprocal of τinactivation vs. Vtest gives an estimate of the gating charge involved in the inactivation process according to Eq. 3.
where K(Vtest) and K(0) are the rate constants of inactivation (reciprocal of τinactivation) at voltages Vtest and 0 mV, respectively, and F/RT = 0.04 mV−1. The values of zi are listed in Table II. These charge mutations do not have statistically significant effects on the zi value (ANOVA, P > 0.05). The rate of K525C inactivation is significantly faster than that of WT (Fig. 9 B). Although MTSETo modification of 525C greatly slows the rate of deactivation (Fig. 6), this intervention has no effect on K525C inactivation (unpublished data), further supporting the notion that this S4 positive charge is not involved in channel inactivation.
The major findings in this study can be summarized as the following. (a) The number of gating charges transferred during hERG activation (za) is estimated to be 6.4, based on the limiting slope method. (b) Neutralizing the first three positive charges in hERG's S4 segment reduces za while neutralizing the following three positive charges does not, suggesting that the first three positive charges are involved in gating charge transfer. (c) Consistent with the above observations, cysteine side chains introduced into the first three, but not the following three, positive-charge positions in S4 manifest state-dependent changes in their MTSET accessibility, consistent with an outward S4 movement relative to the membrane barrier during depolarization. (d) Neutralizing D466 in S2 reduces za. Furthermore, cysteine side chain at 466 manifests a state-dependent change in MTSETo accessibility, consistent with an inward S2 movement relative to the membrane barrier during depolarization. (e) None of the charge mutations affect the number of gating charges involved in hERG's inactivation process (zi).
Comparison Between hERG and Shaker of Scheme of Gating Charge Transfer During Channel Activation
Based on the “conventional gating model” that results from extensive studies on the Shaker channel (Baker et al., 1998; Bezanilla, 2000; Cohen et al., 2003; Horn, 2004; Starace and Bezanilla, 2004), the S4 is oriented more or less perpendicular to the plane of the cell membrane, contacting the periphery of the pore domain on one side, and surrounded by S1, S2, and S3 on the other. Most of the S4's surface is exposed to water-filled crevices connected to the extracellular or intracellular aqueous compartment, where the positive charges on S4 are stabilized by negative charges on S1, S2, or S3, and by counter ions in the crevices. Only a narrow portion of the S4 is in a “gating pore,” which forms a barrier between the extracellular and intracellular compartments where the membrane electrical field is focused. Membrane depolarization exerts an electrostatic force on S4's positive charges that are within or near the membrane electrical field and thus can sense changes in the membrane voltage (gating charges). This causes S4 movements, resulting in a transfer of gating charges through the gating pore. The other positive charges on the S4 that cannot sense the membrane voltage are not involved in gating charge transfer, but can serve other functions (e.g., producing topogenic signals during folding of nascent channel proteins) (Papazian et al., 2002). In the Shaker channel, the first four positive charges carry gating charges: R362, R365, R368, and R371 (Starace and Bezanilla, 2004). In addition, a negative charge in the cytoplasmic half of S2 (E293, Fig. 1 D) may also contribute to the gating charges (Seoh et al., 1996).
Table II lists the equivalent gating charges (za, in eo) for WT hERG and cysteine-substituted mutants. It has been shown that za values obtained by the limiting slope method underestimate the total numbers of gating charges (ztotal) by ∼20% (Schoppa et al., 1992; Zagotta et al., 1994). Therefore, we further estimate the ztotal values of WT-hERG and mutant channels using the mean values of za: ztotal = za/0.8. The values of ztotal are also listed in Table II.
Our data support the alignment of S4's positive charges between hERG and Shaker shown in Fig. 1 B (also shown in Fig. 10, top). (a) The first three positive charges in hERG's S4 carry gating charges during channel activation. (b) The ztotal for WT hERG is ∼8 eo, less than that of the Shaker channel (12–13 eo) by 4–5 eo (Schoppa et al., 1992; Aggarwal and MacKinnon, 1996; Seoh et al., 1996). This difference is consistent with the notion that hERG has one less gating charge per subunit that traverses most of the membrane electrical field during activation. The first positive charge in Shaker's S4, R362, fulfills this criterion (Aggarwal and MacKinnon, 1996). (c) The pattern of state dependence of MTSET accessibility of cysteine side chains introduced into the S4 segment of the hERG channel is in general agreement with that described for the equivalent positions in the Shaker channel (Fig. 10) (Larsson et al., 1996; Baker et al., 1998). Furthermore, our data support a role of D466 in S2 in gating charge transfer, again similar to the putative role of E293 in S2 of the Shaker channel (Seoh et al., 1996). Thus the scheme of gating charge transfer during the activation process is similar between hERG and Shaker.
The state dependence of 528C and 531C accessibility to MTSETi is qualitatively similar to that described for the equivalent positions in the Shaker channel, 368C and 370C (Larsson et al., 1996; Baker et al., 1998). However, quantitatively the change in modification rate in hERG (open to closed state transition increases the MTSETi modification rate of 528C and 531C by 4- and 1.5-fold, respectively) is much less than that seen in Shaker (open to closed state transition increases the MTSETi modification rate of 368C and 370C by 100- and 10-fold). This suggests structural differences between hERG and Shaker. Perhaps the inner crevice around S4 is narrower in hERG than in Shaker, restricting MTSETi accessibility to cysteine side chains at these positions even in the closed state.
Why Is the Activation Process in the hERG Channel So Slow?
Despite the similarity in the scheme of gating charge transfer between the two channels, the rate of hERG activation is >30-fold slower than that of Shaker. Slow activation can be due to slow movement of the voltage sensor, inefficient coupling between the voltage sensor and the activation gate, slow opening of the activation gate, or a combination of the above. Two studies suggest that a slow movement of the voltage sensor is the major cause for the slow rate of activation in the hERG channel. First, Smith and Yellen (2002) used the technique of voltage clamp fluorometry to probe conformational changes at or around the outer end of hERG's S4 in response to changes in membrane voltage. They observed a fast and a slow component of changes in fluorescence (ΔF) at positions 518 and 519, and a slow ΔF at 520. The slow ΔF is correlated with the activation and deactivation processes in terms of kinetics, voltage dependence, and sensitivity to an NH2-terminal truncation mutation. Second, Piper et al. (2003) directly recorded gating currents (Ig) from the hERG channel. They also observed a fast and a slow component of Ig, and the slow Ig is associated with channel activation and deactivation. An S4 segment with fewer gating charges will be less sensitive to membrane depolarization, and thus may move more slowly than an S4 with more gating charges (Logothetis et al., 1992; Islas and Sigworth, 1999; Ahern and Horn, 2004). Thus, relative to the Shaker channel, the fewer gating charges in the S4 segments of hERG may contribute to reduced voltage sensitivity and a slower rate of activation.
However, other factors may also contribute to slowing the S4 movement in hERG. Negative charges in the other transmembrane segments (S1, S2, and S3) of the voltage-sensing domain may slow down S4 movement by two mechanisms. First, they may form salt bridges with S4's positive charges preferentially in the closed states (Papazian et al., 2002; Silverman et al., 2003). Our unpublished data suggest that in the hERG channel, negative charges in S1 and S2 may form salt bridges with positive charges at the two ends of S4 in the closed state (D411-K538 and D456-K525). These salt bridges can stabilize the hERG channel in the closed state and retard activation. This is consistent with the observations shown in Fig. 2 that neutralizing K525 or K538 leads to a hyperpolarizing shift in the voltage dependence of activation, which can be explained by a destabilizing effect on the closed state relative to the open state. Second, the extra negative charges in the voltage-sensing domain of hERG may form a divalent cation (M2+) binding site in the external crevice around S4, similar to that described for the EAG channels (Silverman et al., 2000; Schonherr et al., 2002). Ca2+o or Mg2+o bound here can restrict S4 outward movement and slow activation (more discussion below).
Where Is the Voltage Sensor for hERG's Inactivation Process?
The fast ΔF signals in the voltage clamp fluorometry experiments on hERG (Smith and Yellen, 2002) and the fast Ig component in the hERG's gating current (Piper et al., 2003) are similar in their kinetics and voltage dependence, suggesting that they originate from the same gating charge movement. This fast gating charge movement correlates well with the fast inactivation process of the hERG channel in terms of kinetics and voltage dependence. However, perturbing the fast inactivation process of the hERG channel, by mutating residues at the external pore entrance (G628C/S631C and S631A) or by applying external TEA (that interferes with conformational changes around the outer mouth), does not impact on the fast ΔF signals or the fast Ig component (Smith and Yellen, 2002; Piper et al., 2003). These suggest that the inactivation voltage sensor is well insulated from the inactivation gate at the outer mouth of the hERG channel (Smith et al., 1996; Herzberg et al., 1998). Where is this inactivation voltage sensor? Based on data from alanine scanning mutagenesis experiments on hERG's S4 segment, Tristani-Firouzi (2004) proposed that S4 is the voltage sensor for the inactivation process. On the other hand, based on the differential effects of changing [Ca2+]o on the activation and inactivation processes, Johnson et al. (1999) proposed that the inactivation process has a separate voltage sensor than S4, although no specific structural elements were identified.
Our data do not support a role of S4 as the voltage sensor for hERG's inactivation process. First, neutralizing positive charges in hERG's S4 segment does not alter the gating charge mediating inactivation (zi), although mutating the first three positive charges in S4 reduces the gating charges mediating activation (za) as expected. Note that our measurement of zi, based on the voltage dependence of τinactivation, matches the gating charge estimated from the fast ΔF and fast Ig experiments (zi ∼ 0.7). Second, the τinactivation and thus the zi value are measured when the S4 segment is maintained at its outermost (activated) position (triple-pulse protocol in Fig. 9), and the voltage dependence of τinactivation is observed in a voltage range when channels are fully activated. Therefore, zi cannot arise from the S4 movement that mediates channel activation.
What are the alternatives for hERG's inactivation voltage sensor? The structural element must fulfill two criteria: it must move rapidly in response to changes in membrane voltage, and it must carry only small amount of gating charge (∼0.7 eo per channel). This small gating charge can originate from negative charges in S1, S2, or S3, or from an α-helix dipole in the membrane electrical field (Bezanilla, 2000). The first possibility is not very likely, because although neutralizing the negative charges in hERG's S1, S2, and S3 can shift the voltage dependence of activation and alter the rate of activation/deactivation, there is no effect on the kinetics or the voltage dependence (zi) of hERG inactivation (Liu et al., 2003).
The P-helix of the hERG channel (positions 614–621 of the NH2-terminal half of the P-loop; Fig. 1 C) (Doyle et al., 1998) can be a candidate for the inactivation voltage sensor for the following reasons. (a) An α-helix carries a dipole moment due to inducible charge separation between amide nitrogens and carbonyl oxygens of the peptide backbone. The effect can be approximated by placing 0.5–0.7 positive eo near the NH2 terminus and 0.5–0.7 negative eo near the COOH terminus of such an α-helix (Hol, 1985). A calculation of the transmembrane potential profile along the axis of the KcsA channel has shown that the major portion of the transmembrane field drops across the selectivity filter (Roux et al., 2000). This is likely applicable to the hERG channel, given the well-conserved structure and function of the pore domains of all K channels. Therefore, the P-helix is within the membrane electrical field with its COOH-terminal negative dipole pointing in the inward direction. The helical dipole may sense membrane depolarization and change its position in the field. (b) The hERG channel lacks the H bond that bridges the two ends of the P-loop as in the Shaker channel (between W434/W435 and Y445, asterisks connected by a horizontal line above the Shaker sequence in Fig. 1 C) (Yang et al., 1997; Doyle et al., 1998). Therefore, the P-helix in the hERG channel may slip or rotate in response to membrane depolarization. (c) The inactivated state of the hERG channel confers a high sensitivity to drug molecules that bind to the inner cavity of the pore (Numaguchi et al., 2000). The movement of P-helices provides a logical link between conformational changes around the outer mouth (inactivation) and changes in the inner cavity (drug binding). This proposal awaits tests by direct experimental examination or by indirect theoretical calculations.
Why Are hERG and EAG So Different in their Gating Properties?
EAG channels share all the features of amino acid sequence that are proposed to be important for the unique gating properties of the hERG channel. This is exemplified by the bEAG sequence in Fig. 1. In the voltage-sensing domain, bEAG has the same extra negative charges in S1, S2, and S3, and the same positive charges in S4 as in hERG. In the pore domain, bEAG has an even longer S5-P linker than hERG. However, EAG channels do not share the unique gating properties of the hERG channel. In low mM [M2+]o (M2+ = Ca2+ and Mg2+), EAG channels exhibit a much faster activation rate than hERG; at +60 mV, τ of activation <10 ms for dEAG (Silverman et al., 2000), but >50 ms for hERG. EAG channels do not manifest a fast inactivation as in hERG.
Raising [Mg2+]o from 1 to 10 mM causes a pronounced slowing of EAG activation (Silverman et al., 2000; Schonherr et al., 2002). It is possible that the hERG channel operates in the M2+o-bound, slow gating mode even in normal range of [M2+]o (low mM). This is suggested by the observation that lowering [Ca2+]o from 1 to <0.1 mM induces a profound acceleration of hERG activation (Johnson et al., 2001), although elevating [Ca2+]o from 2 to 10 mM has very modest effect (Liu et al., 2003). Therefore, the difference in the activation rate and in the sensitivity of activation rate to changes in [M2+]o between hERG and EAG may be due to a difference in the set points of M2+o binding; hERG has a high affinity M2+o binding site (in <1 mM range) while EAG channels have a lower affinity M2+o binding site (in 1–10 mM range).
In terms of inactivation, Fig. 1 C reveals that several positions that are critical for hERG's inactivation process are not conserved in EAG. These are highlighted by boxes on the hERG and bEAG amino acid sequences in Fig. 1 C. For example, the equivalent of hERG's S620 is T in EAG channels, and the equivalent of hERG's S631 is A in EAG channels. It has been shown that both S620T and S631A mutations in hERG greatly disrupt the inactivation process (Herzberg et al., 1998; Piper et al., 2003). Other differences between hERG and EAG may also contribute to the absence of fast inactivation in the EAG channels.
The authors would like to thank Dr. H. Robert Guy (Laboratory of Experimental and Computational Biology, National Cancer Institute, National Institutes of Health [NIH], Bethesda, MD) for constructive discussion along the course of this project.
This study was supported by RO1 HL-46451 from National Heart, Lung, and Blood Institute, NIH, and a Grant-in-Aid Award from the American Heart Association/Mid-Atlantic Affiliate (to G.-N. Tseng).
Olaf S. Andersen served as editor.
J. Liu's current address is Department of Pathophysiology, First Military Medical University, People's Republic of China.
Abbreviation used in this paper: WT, wild type.