Mutations within the S4–S5 Linker Alter Voltage Sensor Constraints in hERG K⁺ Channels

Abstract

Human ether-a-go-go related gene (hERG) channel gating is associated with slow activation, yet the mechanistic basis for this is unclear. Here, we examine the effects of mutation of a unique glycine residue (G546) in the S4–S5 linker on voltage sensor movement and its coupling to pore gating. Substitution of G546 with residues possessing different physicochemical properties shifted activation gating by ∼−50 mV (with the exception of G546C). With the activation shift taken into account, the time constant of activation was also accelerated, suggesting a stabilization of the closed state by ∼1.6–4.3 kcal/mol (the energy equivalent of one to two hydrogen bonds). Predictions of the α-helical content of the S4–S5 linker suggest that the presence of G546 in wild-type hERG provides flexibility to the helix. Deactivation gating was affected differentially by the G546 substitutions. G546V induced a pronounced slow component of closing that was voltage-independent. Fluorescence measurements of voltage sensor movement in G546V revealed a slow component of voltage sensor return that was uncoupled from charge movement, suggesting a direct effect of the mutation on voltage sensor movement. These data suggest that G546 plays a critical role in channel gating and that hERG channel closing involves at least two independently modifiable reconfigurations of the voltage sensor.

Introduction

The human ether-a-go-go related gene (hERG) encodes the pore-forming subunit of the potassium channel underlying the cardiac delayed rectifier current, I_Kr (1,2). hERG channel structure resembles that of Shaker voltage-gated K⁺ (Kv) channels (3) possessing six (S1–S6) transmembrane domains that comprise voltage sensor (S1–S4) and ion conduction pore (S5–S6) units. Despite their sequence homology with the Kv channel superfamily, hERG channels exhibit distinct functional properties: hERG activation and deactivation gating kinetics are slow, whereas inactivation and the recovery from inactivation are rapid and voltage-dependent (4).

The molecular mechanisms underlying the distinct gating properties of hERG channels are not clearly understood. The S4 helix has been proposed to act as the primary voltage sensor for both activation and inactivation in hERG (5). The extent of S4 movement in hERG upon depolarization was recently shown to be similar to that in Shaker and KvAP (6). However, voltage clamp fluorimetry (VCF) and gating current measurements suggest that the slow rate-dependent movement of S4 upon depolarization is responsible for the slow activation kinetics of hERG channels (7,8). A kinetic scheme of the hERG activation pathway that includes both voltage-dependent and -independent transitions was previously described (9) as follows (Scheme 1):The additional constraints upon S4 movement have been suggested to arise from stabilization of closed states afforded by the outermost S4 positive charges, K525 and R528 (10,11), and from extensive salt bridge formation between S4 positive charges and S1–S3 domain negative charges (12–14). Slow deactivation, on the other hand, is due in part at least to stabilization of channels in the open state by the N-terminus (15–22). It has been proposed that the N-terminus may interact with the S4–S5 linker, since modification of G546C with a bulky adjunct group apparently impedes the action of the N-terminus (17).

(Scheme 1).

Electromechanical coupling of voltage sensor activation with opening of the intracellular pore gate has been proposed to occur via the α-helical S4–S5 linker region in hERG and other Kv channels (23–27). In Shaker, the S4–S5 linker is thought to directly interact with distal portions of S6, causing the opening of a gate that is formed by the highly conserved Pro-X-Pro motif (27–29). Due to the absence of the Pro-X-Pro motif, the intracellular gate remains poorly defined in hERG. However, evidence suggests that residues in the lower portion of S6 may form a gating ring (30), and that the gate is controlled by a direct electrostatic interaction between charged residues in the S4–S5 linker and S6 (23,24,26). It is reasonable to suggest that electromechanical coupling via the S4–S5 linker may contribute to the slow activation and deactivation properties of hERG channels. In this work, we investigated the effect of hERG mutations at a site in the S4–S5 linker that is known to critically regulate the coupling of voltage sensor activation to opening of the Shaker channel pore. This site, hERG G546, disrupts a highly conserved leucine zipper motif, and here we demonstrate its importance in the gating of hERG.

Materials and Methods

Molecular biology

A pBluescriptSKII vector was used to express hERG (a kind gift from Dr. D. Fedida, University of British Columbia) in Xenopus oocytes. The L520C mutation in the S3–S4 linker was introduced as a site for fluorophore labeling. Labeling of the two native extracellular cysteine residues (C445 and C449 in the S1–S2 linker) did not produce voltage-dependent fluorescence changes with the fluorophore used (see Fig. S4 A in the Supporting Material) and were not removed. Mutant constructs were generated by means of overlap extension polymerase chain reaction with primers synthesized by Sigma Genosys (Oakville, Canada). All constructs were sequenced with the use of Macrogen (Seoul, Korea) or Eurofins MWG Operon (Huntsville, AL). Constructs were linearized with XbaI restriction endonuclease, and cRNA was synthesized with the mMessage mMachine T7 Ultra cRNA transcription kit (Ambion, Austin, TX).

Oocyte preparation and injection

Xenopus laevis oocytes were isolated from frogs during recoverable surgical procedures. After 20 min of anesthetization in 0.8% tricaine solution, three to four lobes of oocytes were removed through a 1–2 cm incision made in the lower abdomen. Stage V–VI oocytes were isolated and defolliculated using a combination of collagenase treatment (1 h in 1 mg/mL collagenase type 1A; Sigma-Aldrich) and manual defolliculation. After defolliculation, the oocytes were incubated in SOS+ media (in mM: 96 NaCl, 2 KCl, 1.8 CaCl₂, 1 MgCl₂, 5 HEPES, 5% horse serum, 2.5 sodium pyruvate, 100 mg/L gentamicin sulfate, pH 7.4) for 5–24 h before injection. The oocytes were injected with 50 nL (5–10 ng) cRNA via a Drummond digital microdispenser (Fisher Scientific, Nepean, Canada) and incubated in SOS+ media at 19°C for 1–10 days. All chemicals were purchased from Sigma-Aldrich (Mississauga, Canada).

Data acquisition and analysis

Currents were recorded from wild-type (WT) and mutant hERG channels by means of a two-electrode voltage clamp. Membrane currents were recorded with an Axoclamp 900A amplifier (Axon Instruments, Foster City, CA) using computer-driven voltage protocols (pClamp 10 software and Digidata 1440 interface; Axon Instruments) while the cells were bathed in ND96 solution (in mM: 96 NaCl, 3 KCl, 1 MgCl₂, 0.5 CaCl₂, and 5 HEPES, pH 7.4). The microelectrodes had a resistance of 0.2–2.0 MΩ when filled with 3 M KCl. Signals were acquired at a sampling rate of 10 kHz with a 4 kHz low-pass filter. Experiments were performed at 20–22°C.

Conductance-voltage (G-V) relations were calculated from peak tail current amplitudes. Curves were fitted with a single Boltzmann function (y = 1/1(1+exp(V − V_1/2)/k), where y is the conductance normalized with respect to the maximal conductance, V_1/2 is the half-activation potential, V is the test voltage, and k is the slope factor). Data throughout the text and figures are shown as the mean ± SE. In the figures, arrows mark the zero current level, and dotted lines mark the baseline current to guide the eye; r-values describing the correlations represent Pearson product moment correlation values.

Voltage clamp fluorimetry

L520C was labeled with the sulphydryl-reactive fluorescent tag, tetramethylrhodamine-5-maleimide (TMRM; Invitrogen). Oocytes were labeled with 5 μM TMRM in a depolarizing solution (in mM: 98 KCl, 1 MgCl₂, 2 CaCl₂, and 5 HEPES, pH 7.4) for 30 min at 10°C in the dark. Fluorimetry experiments were performed with a Nikon TE2000S inverted microscope (Nikon, Mississauga, Canada) with an epifluorescence attachment and photomultiplier tube detection module (Cairn Research, Kent, UK). Fluorophore molecules were excited by light through a 525 nm band-pass (45 nm) filter (Omega Optical, Brattleboro, VT) and focused onto the oocyte via a 0.75 NA 20× objective lens. Fluorescence emission collected via the same objective was filtered with a 565 nm long-pass emission filter and detected with a bialkali photocathode PMT recording module. During VCF experiments, fluorescence signals were acquired simultaneously with ionic current and voltage signals. Unless otherwise stated, the fluorescence signals were not averaged. In some cases, to account for the majority of bleaching of the fluorescence signal during activation and deactivation protocols, the fluorescence recorded during potentials at which there was no channel opening was subtracted.

Results

Mutation of a critical site in the S4–S5 linker helix alters activation properties

Fig. 1 A shows a sequence alignment of the S4–S5 linker of Shaker and hERG channels and highlights residues that contribute to a leucine zipper motif in Shaker (31). It is known that mutation of Shaker L382 (the V2 mutation) exerts dramatic effects on gating by destabilizing the open state and uncoupling voltage sensor activation from pore opening (31–35). Of interest, the sequence alignment shows that the leucine/isoleucine zipper is disrupted in hERG by the presence of G546 at the equivalent site to L382 in Shaker. We reasoned that the presence of glycine at this critical site in hERG might contribute to the characteristic slow gating kinetics by altering either the electromechanical coupling between the voltage sensor and the pore gate or the constraints on voltage sensor movement.

Figure 1.

Restoration of the leucine/isoleucine zipper in hERG channels dramatically alters activation properties. (A) Sequence alignment of the S4–S5 linker of Shaker and hERG channels. Boxes highlight the leucine/isoleucine zipper motif that is disrupted in hERG by G546. (B) Typical WT and G546L currents recorded during the voltage protocols shown. (C) WT and G546L G-V relations constructed from peak tail currents. Data were fitted with a Boltzmann function. (D) WT and G546L current at the end of depolarizing steps normalized to the peak current.

We first examined the effects of the hERG mutation G546L. Fig. 1 B shows representative current traces from WT and G546L in response to changes in the membrane potential. Fig. 1 C shows G-V relations for the two channels calculated from peak tail current amplitudes. The G546L mutation shifted the V_1/2 of activation from −25.7 ± 0.1 mV in WT to −70.9 ± 1.2 mV in G546L. Fig. 1 D shows the fractional current amplitude measured at the end of the depolarizing pulse. These data indicate that rectification in G546L was not different from that in WT. Values for the V_1/2 of inactivation measured using a conventional triple pulse protocol were −55.4 ± 3.1 mV and −55.6 ± 7.2 mV (n = 2–4) for WT and G546L, respectively (Fig. S1). This suggests that activation and inactivation processes are not directly coupled in hERG channels.

The G546L mutation also accelerated the time course of channel activation (Fig. 2, A–D) even when the shift of activation was taken into account. To compare activation τ-values, we calculated the total energy of activation at each potential (Table S1). With comparable total driving forces for activation (29.2 KJ/mol at +60 mV in WT and 29.6 KJ/mol at +10 mV in G546L), the activation τ-value in G546L was 43.2 ± 7.7 ms compared to 70.9 ± 6.9 ms in WT (n = 4–6; p < 0.01). Fig. 2 D (inset) plots the τ-value of the late phase of activation of WT and G546L (9), which decreased exponentially with voltage. Wang et al. (9) described this dependence on voltage as evidence for a voltage-independent transition in the activation pathway that becomes rate-limiting at depolarized potentials (Scheme 1). WT and G546L reached asymptotic values of 37.3 and 31.5 ms, respectively (Fig. 2 D, inset), suggesting that G546L did not alter the voltage-independent transition but did alter other transitions in the activation pathway.

Figure 2.

The G546L mutation destabilizes the closed state of hERG channels. (A and C) WT and G546L currents recorded during an envelope of tails protocol to measure the time course of channel opening. The peak tail current at −110 mV was measured after a step to +60 mV of varying duration (10–500 ms in 10 ms increments). The holding potential was −80 mV (WT) or −120 mV (G546L). To account for the shift in activation, WT and G546L traces are compared at +20 and −30 mV. (B and D) Normalized WT and G546L peak tail current after depolarizing steps to a range of potentials. (Inset) Plot of the time constant of activation of WT and G546L at each potential. Data points show the τ-value from single exponential fits of the late phase of activation from the data in B (from t = 100 ms) and D (from t = 50 ms). The voltage dependence of the τ-value of the late phase of activation was fitted with a single exponential function. (E and F) Typical fluorescence reports from TMRM-labeled L520C and G546L L520C during voltage steps to +60 or +10 mV from a holding potential of −80 or −120 mV, respectively. The insets show the same recordings on a faster timescale. Single exponential fits are shown as gray lines. (G) Mean G-V and F-V relations for L520C and G546L L520C. Data were fitted with a Boltzmann function. V_1/2- and k-values for G-V relations were −17.4 ± 4.5 mV and 10.9 ± 1.1 mV for L520C (n = 4), and −55.8 ± 1.7 mV and 9.1 ± 0.7 mV for G546L L520C (n = 7), respectively. V_1/2- and k-values for F-V relations were −15.2 ± 2.0 mV and 15.4 ± 1.0 mV for L520C (n = 4), and −62.4 ± 10.2 mV and 12.8 ± 1.2 mV for G546L L520C (n = 4), respectively.

To understand the mechanism by which G546L affects gating, we measured voltage sensor movement using VCF. Previous fluorimetric analyses of conformational dynamics have shed light on gating mechanisms in a range of channel types (7,36–38). Typical fluorescence reports of hERG voltage sensor dynamics reported by TMRM attached at L520C in the S3–S4 linker are shown in Fig. 2, E and F. Fluorescence records from L520C and G546L L520C show that voltage sensor activation was accelerated in G546L; the τ-value of the fluorescence deflection upon depolarization was 43.1 ± 3.4 ms (at +60 mV) and 21.4 ± 1.0 ms (at +10 mV) in L520C and G546L L520C, respectively (n = 6; p < 0.01). The τ-value of the fluorescence report of voltage sensor movement in L520C was faster than WT ionic current activation (43.1 vs. 70.9 ms) because the TMRM label accelerated channel activation (Fig. S3). Fig. 2 G shows G-V and fluorescence-voltage (F-V) relations for L520C and G546L L520C. As reported previously (7), the F-V relation overlaid the G-V relation in hERG L520C, suggesting that voltage sensor movement is the rate-limiting step for channel opening (TMRM labeling had no effect on the G-V curve (Fig. S3 A)), and we found that this was unchanged in G546L. Taken together, the data in Figs. 1 and 2 suggest that G546L destabilizes the closed state of the channel by altering voltage sensor movement.

Reducing the flexibility of the S4–S5 linker helix destabilizes the closed state of the channel

To better understand the mechanism by which mutation of G546 alters channel activation gating, we made a number of substitutions with amino acids possessing a range of physicochemical properties. Fig. 3 A shows that of nine mutations tested, all but G546C induced large hyperpolarizing shifts of the voltage dependence of activation (see also Table 1). Given that the S4–S5 linker has been reported to be α-helical (4,28), these data suggest that the flexibility of the linker afforded by the native glycine residue in WT channels constrains movement of the voltage sensor and therefore channel opening. To test this, we used the predictive algorithm AGADIR to investigate the α-helical propensity of the S4–S5 linker region in hERG channels. The AGADIR algorithm predictions are based on a helix/coil transition theory that includes potential side-chain-side-chain interactions, electrostatic effects, and comparison with a large database of experimentally measured helix contents in different peptides (39). Fig. 3 B shows the predicted α-helical propensity of each residue in the S4, S4–S5, and S5 regions of the hERG channel. The predictions for the S4–S5 linker alone are shown in Fig. 3 C. Substitutions at G546 are predicted to have significant effects on the α-helical content, particularly at the N-terminal end of the helix. Those mutations that destabilized the closed state of the channel dramatically increased the α-helical propensity of the S4–S5 linker. In contrast, G546C, which produced only a modest shift of the G-V relation (Fig. 3 A), likewise was predicted to only modestly alter the α-helical content. Taken together, these data are consistent with the notion that the reduced α-helical propensity of the S4–S5 linker afforded by the native glycine stabilizes the closed state of WT hERG channels and contributes to the slow gating kinetics.

Figure 3.

Flexibility of the S4–S5 α-helical linker afforded by G546 stabilizes the closed state of WT hERG channels. (A) G-V relations recorded from WT and G546 mutants. Data were fitted with a Boltzmann function. Values for V_1/2 and k are shown in Table 1. (B and C) AGADIR algorithm predictions of the α-helical content of the S4, S4–S5, and S5 regions (B) and the S4–S5 region only (C).

Table 1.

Activation properties of WT and mutant hERG channels

	Activation parameters
	V1/2 (mV)	k (mV)	ΔG (Kcal mol−1)	ΔΔG	n
WT	−25.7 ± 0.1	7.2 ± 1.3	−2.1 ± 0.2		5
G546Q	−67.2 ± 1.5	7.3 ± 0.3	−5.4 ± 0.3	−3.3	5
G546I	−56.6 ± 1.0	8.4 ± 0.2	−4.0 ± 0.1	−1.9	5
G546L	−70.9 ± 1.2	6.7 ± 0.4	−6.4 ± 0.4	−4.3	5
G546R	−55.8 ± 2.7	8.3 ± 0.3	−4.0 ± 0.3	−1.9	6
G546E	−74.6 ± 4.1	8.8 ± 0.8	−5.1 ± 0.5	−3.0	5
G546Y	−67.7 ± 1.1	8.1 ± 0.1	−4.9 ± 0.1	−2.8	6
G546V	−71.8 ± 0.5	8.9 ± 0.5	−4.8 ± 0.2	−2.6	5
G546C	−27.1 ± 1.2	10.1 ± 0.3	−1.6 ± 0.1	−0.5	5
G546A	−58.5 ± 3.7	9.7 ± 1.0	−3.7 ± 0.5	−1.6	4

Values for V_1/2 and k were obtained from Boltzmann fits of the G-V relations of each channel. Free-energy changes (ΔG) were calculated using −zFV_1/2, and perturbation energies (ΔΔG) were calculated using ΔΔG = ΔG_mutant − ΔG_WT.

Mutations at position G546 differentially affect deactivation gating

In addition to the effects of G546 mutations on activation, the series of mutations had profound, but different, effects on deactivation. Fig. 4, A–C, shows deactivation data recorded from WT channels. As previously observed (1,40), current decay over a range of potentials was best described by a double exponential (Fig. 4 A). Fig. 4, B and C, show time constants and relative amplitudes of the fast and slow phases of current decay. Fast and slow τ-values had a similar dependence on voltage; however, whereas deactivation was dominated by the fast phase at negative potentials, the slower phase became more dominant at less hyperpolarized potentials. Fig. 4, D–F, shows deactivation properties of G546L channels. As a consequence of the left-shifted voltage dependence of activation in G546 mutants, deactivation was studied from −70 to −150 mV. G546L deactivating currents were well described by a single exponential function at all potentials tested, and there was no evidence of a slow phase of deactivation (Fig. 4, D–F). Conversely, in another mutant, G546V, the slow phase of deactivation was more pronounced and showed no dependence on voltage (Fig. 4, G–I). Table 2 shows the ratio of the slow phase τ at moderately and strongly hyperpolarized potentials. The τ_{slow(−50mV)}/τ_{slow(−100mV)} ratio in WT was 10.5 ± 1.8, demonstrating a strong dependence of the rate of deactivation upon voltage. In contrast, the equivalent ratio (accounting for the shift in the total activation energy; see Table S1), τ_{slow(−100mV)}/τ_slow(-150mV), was reduced to 1.2 ± 0.1 in G546V, indicative of a weak (if any) dependence of the slow phase upon voltage. The data in Table 2 reveal that the different G546 mutations fell into one of three categories with respect to their effects on deactivation: 1), no effect; 2), slow phase abolished (as with G546L); or 3), slow phase enhanced and voltage-independent (as with G546V). These observations suggest that the nature of the amino acid at position 546 plays an important role in regulating deactivation, but in a different way compared to its role in regulating activation. We observed no clear correlation between S4–S5 α-helical propensity, or the size, charge, or hydrophobicity of the amino acid side chain at 546 and the effect on deactivation. However, a significant correlation was clear if we considered only smaller (<150 ų) side chains when correlating deactivation with hydrophobicity (Fig. S2). To better understand how the G546 mutations affect deactivation gating, and in particular, how mutations preferentially affect the slow component of deactivation, we further investigated the gating of G546V using VCF.

Figure 4.

G546 mutations affect deactivation gating independently of activation gating. (A, D, and G) Typical currents recorded from WT, G546L, and G546V during the voltage protocols shown. (B, E, and H) Deactivation τ-voltage relations on a log scale. WT deactivation was best described by a double exponential function. G546L deactivation was well described by a single exponential function at all potentials tested. G546V deactivation displayed an unusual prominent and voltage-independent slow phase of current decay. (C, F, and I) Relative amplitudes of the phases of deactivation in each of the channel types.

Table 2.

Deactivation properties of WT and mutant hERG channels

	Deactivation parameters
	τfast (ms)		τslow (ms)		Afast	Aslow	τslow ratio	n
	−100 mV	−150 mV	−100 mV	−150 mV
WT	105 ± 4		493 ± 62		0.85 ± 0.03	0.15 ± 0.03	10.5 ± 1.8	7
G546Q		61 ± 9		156 ± 20	0.88 ± 0.05	0.12 ± 0.05	8.2 ± 1.7	5
G546I		99 ± 4			1.0			6
G546L		334 ± 89			1.0			5
G546R		38 ± 8			1.0			6
G546E		43 ± 7			1.0			5
G546Y		71 ± 2		980 ± 28	0.44 ± 0.02	0.56 ± 0.02	1.7 ± 0.1	5
G546V		69 ± 2		1471 ± 98	0.56 ± 0.02	0.44 ± 0.02	1.2 ± 0.1	8
G546C	142 ± 10		1871 ± 203		0.68 ± 0.04	0.32 ± 0.04	1.0 ± 0.1	5
G546A		31 ± 7		623 ± 65	0.38 ± 0.03	0.15 ± 0.02	1.8 ± 0.5	5

Time constants and relative amplitudes from double exponential fits of WT and mutant deactivating currents. Values are shown at −100 mV for WT and G546C and −150 mV for all other mutants to account for the shift in total activation energy. G546I, G546L, G546R, and G546E deactivation could be fitted with a single exponential function at all potentials. The τ_slow ratio values represent the ratio of τ_{slow(−50mV)}/τ_{slow(−100mV)} (WT and G546C) or τ_{slow(−100mV)}/τ_{slow(−150mV)} (all other mutants) to highlight the voltage independence of the slow phase of deactivation observed in some mutant channels.

G546L and G546V mutations directly affect voltage sensor return during deactivation

Mutation of G546 may alter gating by modifying voltage sensor movement, or the way in which S4 movement is coupled to the pore gate. To distinguish between these possibilities, we conducted a fluorimetric analysis of voltage sensor movement during deactivation of WT and mutant channels. Fig. 5 A shows example fluorescence records from TMRM-labeled hERG L520C channels during deactivation. The fluorescence report of voltage sensor movement was best fitted with a double exponential and was accelerated at more hyperpolarized potentials. Fig. 5 B shows the relative amplitudes of fast and slow fluorescence changes, and that each correlates well with the fast and slow phases of ionic current decay. This suggests that the two phases of deactivation of WT hERG channels are the result of at least two different conformational changes of the voltage sensor that occur upon repolarization. Fig. 5, C and E, show example fluorescence reports from G546L L520C and G546V L520C during deactivation. For G546L L520C, fluorescence deflections were best fitted with a single exponential function at all potentials tested. In contrast, the fluorescence report of voltage sensor return in G546V L520C was biphasic even at strongly hyperpolarized potentials, such as −140 mV. Moreover, the slow phase of voltage sensor return remained prominent at all potentials tested, and the relative amplitudes of the fast and slow components of fluorescence decay were voltage-independent (Fig. 5 F). These data suggest that mutation of G546 directly alters voltage sensor movement. Moreover, the data are consistent with the conclusion that deactivation is associated with two reconfigurations of the voltage sensor that are independently modifiable.

Figure 5.

The G546V mutation directly affects voltage sensor return during deactivation. (A, C, and E) Typical fluorescence records from L520C, G546L L520C, and G546V L520C during the voltage protocols shown. Fluorescence traces were fitted with a double exponential function for L520C and G546V L520C, and a single exponential function for G546L L520C (gray lines). The interpulse interval was 30 s to allow for relaxation of fluorescence signals to baseline levels. (B, D, and F) Mean relative amplitudes of the fast and slow components of fluorescence change during deactivation (n = 5–8). Dotted lines represent the relative amplitudes of fast and slow components of ionic current decay from Fig. 4.

The G546V phenotype is not the result of altered N-terminal interactions

Several lines of evidence suggest that the N-terminus modulates deactivation of hERG channels (15–22), perhaps via interactions with the S4–S5 linker (17). We therefore examined deactivation in channels lacking the majority of the N-terminus (Δ2-354) to determine whether the prominent slow phase of voltage sensor return in G546V resulted from altered interaction with the N-terminus. Fig. 6, A and D, show example current records in response to hyperpolarizing voltage pulses. As reported previously, the Δ2-354 N-terminal deletion accelerated deactivation (Fig. 6, A–C). The fast and slow phases of deactivation were faster in Δ2-354 than in WT, but the contribution of the fast phase in Δ2-354 was dominant at all potentials tested (Fig. 6 C), unlike in WT (Fig. 4 C). For G546V Δ2-354, the fast phase of deactivation was accelerated, suggesting the N-terminus remained able to interact with the G546V channel. However, the prominent voltage-independent slow phase of deactivation remained unaltered (Fig. 6, D–F). The fact that the effect of the G546V mutation was evident regardless of the presence or absence of the N-terminus suggests that the slow voltage-independent component of deactivation associated with the S4–S5 linker mutation was not mediated via the N-terminus, but rather by a direct effect on voltage sensor movement. To test this further, we measured the fluorimetric report of voltage sensor movement in Δ2-354 channels in the absence and presence of the G546V mutation (Fig. 6, G and H). The N-terminal deletion accelerated the fluorescence report of voltage sensor return (Fig. 6 G), as shown previously (7). Introduction of G546V (Fig. 6 H) produced a fluorescence report that retained the slow voltage-independent component observed in G546V channels with an intact N-terminus. Since deletion of the N-terminus did not alter the slow reconfiguration of the voltage sensor in G546V channels, these data suggest that G546V alters deactivation by directly affecting voltage sensor movement in a manner that is independent of the N-terminus.

Figure 6.

The G546V phenotype is not the result of altered N-terminal interactions. (A and D) Typical current records from Δ2-354 and Δ2-354 G546V during the voltage protocols shown. (B and E) Deactivation τ-voltage relations on a log scale (n = 6–10). Values for Δ2-354 G546V τ_slow are not shown because the slow phase was too slow to accurately measure. (C and F) Relative amplitudes of the phases of deactivation in the two channel types. Values for A_{slow + residual} are plotted for Δ2-354 G546V rather than A_slow because of the slow decay of current. (G and H) Typical fluorescence records from Δ2-354 L520C and Δ2-354 G546V L520C during the voltage protocol shown. Relative amplitudes of the fast and slow phases of deactivation were 91.4 ± 4.8 and 8.6 ± 4.8% for Δ2-354 L520C (−110 mV) and 32.4 ± 3.4 and 67.6 ± 3.4% for Δ2-354 G546V L520C (−140 mV).

hERG channel gating modifiers act by altering voltage sensor movement

The data so far suggest that voltage sensor return during hERG channel deactivation is complex and undergoes at least two configurations that can be independently modified. Given that voltage sensor movement has been reported to be rate-limiting for hERG channel pore opening (7,8) (see also Fig. 2, E and G), we hypothesized that agents or interacting partners known to modify hERG gating might do so by directly altering voltage sensor movement. To test this, we investigated the effects of extracellular acidic pH, a well-studied hERG channel gating modifier (41,42) whose mechanism of action remains unclear.

Fig. 7 shows the effect of low extracellular pH on the ionic current and fluorescence report of voltage sensor movement during deactivation. As reported previously (41,42), acidic pH accelerated deactivation of WT hERG (Fig. 7, A and B). This is highlighted by the superimposed traces at −70 mV (Fig. 7 B, inset) and the time constants and relative amplitudes of the fast and slow phases of deactivation at pH 7.4 and 5.5 (Fig. 7 C). Fig. 7, D and E, show example fluorescence records from TMRM labeling L520C at pH 7.4 and 5.5. The superimposed traces (Fig. 7 E, inset) show that extracellular protons accelerated voltage sensor return. As with the effect of low pH on ionic current, both the fast and slow phases of the fluorescence report of voltage sensor return were accelerated at low pH (Fig. 7 F). These observations suggest that protons alter hERG channel deactivation by directly altering voltage sensor movement.

Figure 7.

Low acidic pH acts by directly altering voltage sensor movement. (A, B, D, and E) Typical currents and fluorescence records from WT during 4 s steps to a range of potentials after a depolarizing step to +60 mV during perfusion of solution at either pH 7.4 or pH 5.5. (Insets) Scaled current and fluorescence signals with pH 7.4 and 5.5 at −70 mV to highlight the effect of acidic pH. (C and F) Plots of τ_fast and τ_slow deactivation-voltage relations on a log scale from double exponential fits of deactivating currents and fluorescence signals (n = 5–9). Relative amplitudes of the fast and slow phases as a fraction of the total amplitude of the decaying current and fluorescence signal are expressed in C and F, respectively (lower panel). ● τ_fast or A_fast pH 7.4; ○ τ_fast or A_fast pH 5.5; ▾ τ_slow or A_slow pH 7.4; ∇ τ_slow or A_slow pH 5.5.

Discussion

Despite a recent focus on the mechanism by which voltage sensor activation leads to opening of the ion conducting pore gate in hERG channels, this mechanism remains incompletely defined. Structural and functional evidence suggests that in other Kv channels, mechanoelectrical coupling is achieved by translation of S4 voltage sensor movement to the opening of a pore gate formed by a conserved Pro-X-Pro motif in the lower portion of S6. However, hERG channels lack this motif and channel activation appears to be constrained by slow rate-limiting movement of the S4 domain. Our findings suggest that G546 in the S4–S5 linker of hERG channels contributes to the stabilization of the voltage sensor in its resting state by increasing the flexibility of the S4–S5 α-helical linker. In addition, our VCF reports suggest that S4–S5 linker mutations directly alter voltage sensor return upon repolarization, and that direct modulation of S4 movement offers a mechanism by which interacting partners may modulate hERG channel gating.

Effects on activation

In many Kv channels, as well as Na⁺ and Ca²⁺ channels, a leucine zipper spans the lower portion of S4, the S4–S5 linker and the lower portion of the S5 helix. The leucine zipper is a common motif in proteins that acts to stabilize and orient interacting surfaces of the protein. In Shaker channels, disruption of the zipper motif by the V2 mutation shifts activation in the depolarizing direction, slows channel opening, and uncouples pore opening from gating charge translocation (31–35). In hERG channels, the leucine/isoleucine heptad is naturally disrupted by G546 at the second leucine position. Our functional data show that substitution of this glycine with leucine, or any of a number of amino acid resides besides cysteine, destabilizes the closed state of the channel (Figs. 1–3). We calculated the alteration in the free energy of activation (ΔG) associated with each of these mutations (Table 1) from −zFV_1/2, where z is the effective valence derived from the slope of the G-V relations (k = RT/zF). The change in free energy (ΔΔG = ΔG mutant − ΔG WT) induced by the different substitutions is also shown in Table 1. Whereas the G546C mutation caused only minor perturbations in the energy required for activation (as previously suggested (17,21)), all other mutations reduced the activation energy in the range of −1.6–4.3 kcal/mol. This corresponds to the energy of one to two hydrogen bonds (assuming 1.9 kcal/mol per hydrogen bond). Although AGADIR predictions of α-helical content in a complex membrane protein require some caution, the close correlation of the predictions with our electrophysiological measurements lead us to propose that G546 stabilizes the closed state of WT hERG channels by reducing the α-helical content of the S4–S5 linker, thereby increasing its flexibility. It is tempting to speculate further that interacting partners, such as the N-terminal domain and accessory subunits (e.g., KCNE1/2), may alter hERG channel gating by forming protein-protein interactions with the S4–S5 α-helix, altering its flexibility and therefore movement of the voltage sensor; however, further experiments are clearly required to test this hypothesis.

Effects on deactivation

Our findings indicate that substitutions at G546 had separable effects on activation and deactivation (Figs. 3 and 4). All substitutions (except cysteine) shifted activation gating similarly, but deactivation was affected differently by different substitutions. Clearly, the changes in the α-helical content at G546 are not responsible for the observed changes in deactivation. The biphasic fluorescence report of voltage sensor return recorded from WT and G546V demonstrates the presence of two reconfigurations of the voltage sensor during deactivation that occurred with different time courses (Fig. 5). This is consistent with hERG gating current records that displayed a biphasic return of charge upon repolarization (8), and suggests that the voltage sensor experiences a different environment upon deactivation than activation. We propose that G546 mutations directly alter these reconfigurations. For example, G546V alters deactivation by retarding the slower reconfiguration of the voltage sensor and reducing its dependence on voltage, whereas G546L presents a monoexponential fluorescence report of voltage sensor movement. Of interest, the fluorescence signal recorded in both G546V and G546L overshot the baseline at hyperpolarized potentials (Fig. 5, C and E). This overshoot was not observed in L520C (Fig. 5 A). We interpret this overshoot of the fluorescence as evidence that the voltage sensor exhibits movement at negative potentials in the G546 mutants. This is consistent with the conclusion that the closed state is destabilized by these mutations, reducing the constraints on voltage sensor movement over this voltage range.

We found no clear correlation between deactivation and the size, charge or hydrophobicity of the amino acid at 546. However, we did observe a positive correlation with hydrophobicity when larger bulky amino acids were not considered (Fig. S2). This suggests that hydrophobic side chains at position 546 slow deactivation gating, and that bulky amino acids have an additional effect in disrupting local interactions. Since activation was similar among the mutant channels, these interactions must form when the voltage sensor is in the activated configuration and dictate the ability of the voltage sensor to return to its resting configuration upon repolarization.

It is well known that the presence of the N-terminus slows deactivation gating, stabilizing channels in the open state (15–22). Indeed, modification of G546C with N-ethylmaleimide has been shown to impede the modulation of deactivation by the N-terminus (17), suggesting that the N-terminus comes into close proximity to G546C. However, our data argue against altered interaction with the N-terminus as a mechanism for the effects of G546 mutations on deactivation, since the G546V phenotype was observed whether the N-terminus was present or absent (Fig. 6). The acceleration of the fast phase of deactivation in Δ2-354 channels demonstrates that the N-terminus still interacts with the G546V channel. However, the introduction of the slow voltage-independent phase of deactivation was clearly not dependent on the N-terminus.

Our VCF measurements in hERG channels are in general agreement with those reported by Smith and Yellen (7), although we have expanded the voltage range studied, particularly in relation to deactivation. We interpret our results as a direct report of voltage sensor movement. However, it could be argued that the slow fluorescence report from TMRM attached at L520C during depolarization is not due to voltage sensor movement, but rather to K⁺ flux through the channel or gating at the intracellular pore, as recently suggested (43). Several observations argue against this: 1), fluorescence reports from TMRM labeling the equivalent residue to L520C in Shaker, Kv1.2, Kv1.4, and Kv1.5 channels have been shown to provide a faithful report of voltage sensor movement; 2), in our hands, labeling of the native C445 and C449 in the S1–S2 linker produced no voltage-dependent fluorescence deflections (Fig. S4 A); and 3), 4-aminopyridine (4-AP) intracellular pore block of hERG L520C channels is associated with a reduction of the ionic current but no change in the fluorescence report (Fig. S4, B and C).

Our findings show that low external pH accelerates hERG channel deactivation by altering both fast and slow reconfigurations of the voltage sensor. This suggests that protons modify hERG channel gating by altering the way the voltage sensor moves. This raises the possibility that direct modification of voltage sensor movement, rather than effects on electromechanical coupling or pore gate dynamics, may represent a general mechanism by which hERG channels are modulated by interacting partners. We have demonstrated an effect on voltage sensor movement by extracellular protons, but it is interesting to speculate that intracellular interacting partners may also modify voltage sensor movement.