Probing α-3₁₀ Transitions in a Voltage-Sensing S4 Helix

Abstract

The S4 helix of voltage sensor domains (VSDs) transfers its gating charges across the membrane electrical field in response to changes of the membrane potential. Recent studies suggest that this process may occur via the helical conversion of the entire S4 between α and 3₁₀ conformations. Here, using LRET and FRET, we tested this hypothesis by measuring dynamic changes in the transmembrane length of S4 from engineered VSDs expressed in Xenopus oocytes. Our results suggest that the native S4 from the Ciona intestinalis voltage-sensitive phosphatase (Ci-VSP) does not exhibit extended and long-lived 3₁₀ conformations and remains mostly α-helical. Although the S4 of NavAb displays a fully extended 3₁₀ conformation in x-ray structures, its transplantation in the Ci-VSP VSD scaffold yielded similar results as the native Ci-VSP S4. Taken together, our study does not support the presence of long-lived extended α-to-3₁₀ helical conversions of the S4 in Ci-VSP associated with voltage activation.

Introduction

Voltage-gated ion channels (VGICs) play essential physiological roles in excitable cells such as neurons and myocytes but also in some endocrine and immune cells. VGICs sense changes of the cell’s membrane potential and modulate their ionic permeability (gating) accordingly, thereby regulating a variety of vital biological tasks such as the generation and propagation of electrical impulses (1). Most VGICs are formed by four related domains (voltage-gated sodium channels, Nav, and voltage-gated calcium channels, Cav) or four individual subunits (voltage-gated potassium channels, Kv, and prokaryotic Nav channels), each harboring six transmembrane segments (S1–S6). Segments S1–S4 form four voltage-sensor domains (VSDs) located at the periphery whereas segments S5–S6 form a unique central pore (2, 3, 4).

VSDs are also present in voltage-sensing phosphatases (VSPs) (5) and voltage-gated proton channels (Hv channels) (6, 7). Hv channels are believed to form dimers (8, 9), whereas Ci-VSP, the prototypical VSP cloned from Ciona intenstinalis (5), seems to be present as monomers in the membrane (10).

VSDs respond to changes in the membrane potential and control the conformation of a separate pore (VGIC) or phosphatase domain (VSP). The voltage sensitivity of the VSD relies on the transfer of charged residues (gating charges), mainly arginines or lysines located in the S4-helix, across the electric field (2). Despite decades of studies, the mechanism by which the S4 transports its gating charges remains unsolved.

One unsolved issue in mechanisms of voltage sensing is the actual structural conformation of the S4-helix, which varies widely between different x-ray structures of VGICs and VSPs. The depolarized paddle-chimera Kv1.2/2.1 structure displays an S4 with a hybrid helical conformation, having an α-helical N-terminal end and a 3₁₀ helical C-terminal end (11). In contrast, in the crystal structure of MlotiK (a cyclic nucleotide gated ion channel lacking the positively charged residues in S4), the S4 exhibits a more extended 3₁₀-helix (12). The S4 segment of NavAb, a prokaryotic Nav channel from Arcobacter butzlen, is mostly made of a 3₁₀-helix in two distinct conformations (13, 14).

The recently solved crystal structure of the VSD of Ci-VSP (Ci-VSD) shows its S4 in a pure α-helix, even in different states (15). In addition, recent NMR studies of Kv7.1 reveal a full α-helix S4 (16). These divergent results led several groups to propose that the S4 undergoes intramolecular α-3₁₀ conversions to align its gating charges and thus facilitate their transfer (17, 18). The idea that the S4-helix could adopt a 3₁₀ conformation was originally postulated when the first amino-acid sequence of a Nav channel was deduced from its cDNA (19), and is supported by results obtained doing periodic perturbations of S4 mutations in a Shaker K channel (18). However, the physical detection of dynamic α-3₁₀ conversions has not yet been established and will be needed to validate this hypothesis.

In this study, we explored the existence of such α-3₁₀ transitions of S4 in functional VSDs derived from Ci-VSP. This was done by measuring the length of the S4 segment in an attempt at determining whether a significant portion of the helix switches between α- and 3₁₀-conformations. The length of the S4 segment from Ci-VSP and from NavAb was measured using the lanthanide-based resonance energy transfer (LRET) technique (20), which allows steady-state atomic-scale distance measurements between probes attached to the N-terminal and C-terminal ends of the S4 while the VSD is functionally expressed in voltage-clamped Xenopus oocytes. The existence of short-lived α-3₁₀ transitions was also investigated using time-resolved FRET experiments. Our results show that in the Ci-VSD, the S4 does not undergo extended α-3₁₀ conversions. The results suggest that the extended 3₁₀ conformation of the S4 seen in some VGIC x-ray structures may originate from specific interactions of the S4 with other regions of the channel protein or from short-lived 3₁₀ incursions captured in the crystals.

Materials and Methods

Molecular biology, expression in Xenopus oocytes and labeling

The cDNA of Ci-VSP was a kind gift from Dr. Y. Okamura and the cDNA of NavAb was a generous gift from Dr. W. A. Catterall. Mutations and insertions were made using Quik-Change (Stratagene, La Jolla, CA). Plasmids were transcribed using in vitro transcription kits, mMESSAGE mMACHINE SP6 (Ambion, Austin, TX). Freshly isolated oocytes were injected with 50 ng of cRNA and kept in SOS (Standard Oocyte Saline) solution (96 mM NaCl, 2 mM KCl, 1 mM MgCl₂, 1.8 mM CaCl₂, and 10 mM HEPES, pH 7.4) for 1–5 days at 18°C. For ATTO 425 labeling (ATTO-TEC, Siegen, Germany), we pretreated oocytes expressing the proteins of interest with 1 mM DTT (dithiothreitol) for 25 min at room temperature. After washing them briefly in fresh SOS solution, they were incubated on ice for 25 min in SOS containing 20 μM ATTO 425 maleimide (ATTO-TEC). Before use, oocytes were washed again with dye-free SOS to remove dye attached nonspecifically.

LRET recording in voltage-clamp configuration

LRET is a fluorescence resonance energy transfer (FRET) technique that takes advantage of the properties of lanthanides, in particular, Tb³⁺ which has a long lifetime, dark regions between emission peaks in its spectrum, and, most importantly, isotropic emission (20). The donor is Tb³⁺ bound to the genetically encoded lanthanide-binding-tag motif (LBT),

$$\text{YIDTNNDGWYEGDELLA},$$

which binds Tb³⁺ with high affinity (K_d = 57 nM) (20, 21). The optical setup for LRET measurements has been described in Hyde et al. (22) and Sandtner et al. (23). The extracellular solution was used as recording solution (see next subsection), normally with 2.5 μM Tb³⁺. During measurements, oocytes were under voltage-clamp with a two electrode voltage-clamp (Dagan, Minneapolis, MN). For details, see Results, below; Fig. 4, C and D (inset); and the Supporting Material.

Figure 4.

Measurement of S4 length of LBT-CiVS-mC under voltage-clamp conditions. (A) Representative gating currents of LBT-CiVS-mC during activation from the closed state (Q_−120mV) and after prolonged depolarization at +40 mV (Q_+40mV) used to induce the relaxed state. On the left, respective protocols used for gating current measurements. (Dashed square) Acquisition period for the currents shown to the right. (B) Q-V relations and voltage-dependent kinetics of activation obtained from the Q_-120mV protocol (solid circles for Q-V and solid triangles for time constants) or Q_+40mV protocol (open circles for the Q-V and open triangles for time constants). (C and D) LRET measurements of distance c in LBT-CiVS-mC for several voltages. Respective protocols, V_Pulse and V_Hold, are shown above each plot. (Dotted squares) Acquisition period. (Solid squares) Measured distances (mean ± SE). (Horizontal lines) Estimated value of c in the case that segment A215-F234 in Fig. 1A was entirely α-helical (fine dotted line) and in the case it adopted a whole 3₁₀-helix conformation (long dashed line).

The emission spectrum of Tb³⁺ ions bound to LBT was measured in our laboratory. The absorption and emission spectra of mCherry were obtained from http://www.tsienlab.ucsd.edu. ATTO 425-maleimide absorption and emission spectra were obtained from ATTO-TEC.

Cut-open oocyte epifluorescence recording

The experimental setup is composed of both cut-open oocyte voltage-clamp (COVC) and optical setups (24). Gating currents were recorded in an extracellular solution of NMG-MS (115 mM n-methylglucamine-methylsufonate, 2 mM Ca(OH)₂, and 10 mM HEPES, pH 7.4) and an internal solution (115 mM N-methyl-D-glucamine methanesulfonate, 2 mM EGTA, and 10 mM HEPES, pH 7.4).

The optical configuration has been described in Cha and Bezanilla (24). Filters and dichroic mirror were chosen as follows: model No. D425/40X as the excitation filter (Chroma Technology, Bellows Falls, VT); model No. 505DRLP as the dichroic mirror (Omega Optical, Brattleboro, VT); model No. D500/20m (Chroma Technology) as the emission filters for donor channel; and model No. HQ665LP (Chroma Technology) as the emission filter for the acceptor channel. Gating currents and fluorescence were measured simultaneously averaging 16 recordings each.

Anisotropy and polarization measurements

The anisotropy and polarization of both ATTO 425 (ATTO-TEC) and mCherry were measured as described in Cha and Bezanilla (24). We measured the fluorescence of ATTO 425 from oocytes labeled by ATTO 425 expressing LBT-A222C-CiVS-mC-R99A. The fluorescence of mCherry was measured in nonlabeled oocytes expressing LBT-A222C-CiVS-mC. The calibration process was done by measuring the anisotropy of a known system. We measured, in our setup, ATTO 425 dissolved in glycerol for the correction of labeling with ATTO 425 and TMRM (tetramethylrhodamine-6-maleimide) dissolved in glycerol for the correction of mCherry. Then correction factors were calculated assuming that the anisotropies of ATTO 425 and TMRM in glycerol are 0.38. Using the correction factor, we calculated the anisotropy of ATTO 425 and mCherry. For details, see the Supporting Material.

Data collection and analysis

Fluorescence and electrophysiological recordings were done using separate 14-bit A/D converters that were acquired with an in-house acquisition system based on a model No. SBC6711 board (Innovative Integration, Westlake Village, CA) running its own operating system and communicating via a USB to a computer running a Windows operating system. Gating and fluorescence signals were acquired at 50 μs/point and filtered through a 5-kHz low-pass filter. In LRET measurements, lifetime data acquisition is started by gating on the photomultiplier tube 10 μs after the excitation laser pulse. Photomultiplier tube photocurrents are amplified by a custom high-bandwidth I/V converter/amplifier system. Lifetime data are sampled at 500 kHz and collected as averages of ∼20 consecutive traces. Data were analyzed using an in-house software (GpatchM for data acquisition and Analysis and DecayAnalysis for data analysis) and the commercial softwares MATLAB (The MathWorks, Natick, MA), Excel (Microsoft, Redmond, WA), and ORIGIN (OriginLab, Northampton, MA), and are presented as mean ± SE.

Error propagation

We obtained the time constant values through LRET measurements and computed distances according to the Förster relation applied to LRET as

$$R=R_0\times {\left[{\tau }_{DA}/\left({\tau }_D-{\tau }_{DA}\right)\right]}^{1/6}.$$

Errors of computed distances were estimated from error propagation using simple average error methods where R₀ has no error, but the errors in τ_D and τ_DA are propagated. The results obtained by error propagation are shown in Table S1 in the Supporting Material.

Results

LRET as a tool for detecting long-lived α-3₁₀ S4 transitions

LRET measures energy transfer between a luminescent lanthanide (LRET donor, here a terbium ion, Tb³⁺) and a compatible spectroscopic acceptor (LRET acceptor) (Fig. 1). Because the interpretation of luminescence decay in LRET experiments becomes difficult if multiple donors and acceptors are present (22, 25, 26), we opted for the VSD of Ci-VSP, which is believed to be a monomer (10), to study S4 length changes. To perform LRET measurements in Ci-VSP, we inserted an LBT (21, 23) at the top of the S4 segment (between residues L215 and G216), and we replaced the phosphatase domain (residues 237–576) by the red fluorescent protein mCherry without the linker of N-terminus (residues 8–235) (27) (LBT-CiVS-mC; Fig. 1).

Figure 1.

LRET construct and estimated lengths. (A) Amino-acid sequence of the S4 region of LBT-CiVS-mC. (Green loop) LBT. (Dashed rectangle) S4 portion, which corresponds to that in the crystal structure of Ci-VSP. (Red rectangle) mCherry (8-235). Assuming that segment F234 to N (first residue of mCherry) is an α-helix, its length would be 4.5 Å. In this model, length c (A215, last residue of the LBT, to F234) would be between 28.5 Å (whole α-helix) and 38 Å (whole 3₁₀-helix). (B) Schematic representation of LBT-CiVS-mC in the membrane. Based on its listing in the Protein Data Bank (http://www.rcsb.org/), the size of the LBT is 12.1 Å and the distance from the beginning (asparagine) of the mCherry to the chromophore is 23.2Å. The distance R is measured with LRET and the distance c is obtained by subtracting 12.1, 23.2, and 4.5 Å from R. (C) Hypothetical schematics of α- and 3₁₀-helices in this study. Estimated distances are shown as α-helix (28.5 Å) or 3₁₀-helix (38 Å) for a segment spanning the region from A215 to F234.

The donor-acceptor distance, R (Fig. 1 B) is obtained from the Förster relation

$$R=R_0{\left[{\tau }_{DA}/\left({\tau }_D-{\tau }_{DA}\right)\right]}^{1/6},$$

where R₀ is the distance for 50% energy transfer, and τ_DA and τ_D are the lifetimes of Tb³⁺ luminescence decay in presence and absence of an acceptor, respectively. To measure the donor lifetime in absence of acceptor, we made an additional construct in which the mCherry chromophore is impaired by the point mutation R99A (LBT-CiVS-mC-R99A) homologous to the chromophore-disrupting mutation R96A identified in the green fluorescence protein (28).

In our LRET constructs, the whole S4 segment encompasses 20 residues (from A215, the last amino acid (aa) of the LBT, to F234). Its length is thus expected to range from 28.5 Å (19 × 1.5 Å increment per residue) for a pure α-helix to 38 Å (19 × 2 Å increment per residue) for a pure 3₁₀-helix. Hence a complete α-3₁₀ conversion is expected to stretch the S4 length by ∼9.5 Å (Fig. 1 C). Given the known structures of the LBT (PDB:1TJB) (21) and of mCherry (PDB:2H5Q) (29), and assuming that the LBT and mCherry are aligned with the S4 segment, we estimate the average distance from the N-terminal end of S4 (A215) to the LBT-bound Tb³⁺ to be 12.1 Å (distance a in Fig. 1 B, and see Fig. S1 in the Supporting Material) and the average distance from the C-terminal end of the S4 (F234) to the mCherry chromophore to be 27.7 Å (distance b in Fig. 1 B, and see Fig. S1). Therefore, an extended conversion of the whole S4-helix from α to 3₁₀ would translate into a change of the measured LRET distance from 68.3 to 77.8 Å, a Δ of 9.5 Å (Fig. 1 C).

Typical long-lifetime Tb³⁺ decays were detected from both our LRET constructs. The time constant of the Tb³⁺ signal measured in LBT-CiVS-mC-R99A was 2.41 ± 0.01 ms (τ_D) and 1.98 ± 0.01 ms (τ_DA) in LBT-CiVS-mC, indicating effective energy transfer between the donor and the acceptor (Fig. 2 A). This τ_D value is in good agreement with previous studies (22, 23). In addition, we were able to detect mCherry fluorescence in the LBT-CiVS-mC construct, but not in the control construct, thus confirming the disruption of the chromophore by the R99A mutation (data not shown).

Figure 2.

LRET as a tool for detecting long-lived α-3₁₀ S4 transitions. (A) Normalized decays of Tb³⁺ emission signals (unclamped oocyte) in the absence (solid trace, D-signal) and presence of acceptor (shaded trace, DA-signal). Time constants of decay are τ_D = 2.41 ± 0.01 ms (n = 15) and τ_DA = 1.98 ± 0.01 ms (n = 14). (B) Overlap between the Tb³⁺ emission spectrum (solid line) and the mCherry absorption spectrum (dotted line). (C) Energy transfer efficiency as a function of distance. The plot shows the theoretical curve for the transfer efficiency between the Tb³⁺ donor and the mCherry acceptor. R₀ is the distance for 50% energy transfer from donor to acceptor.

In LRET, R₀ depends on the spectral overlap between the donor emission and acceptor absorption, the extinction coefficient of the acceptor, the refraction index of the medium, and the relative orientation between donor and acceptor (κ²). Because Tb³⁺ emission is isotropic, it is safe to assume κ² = 2/3 (20). The calculated R₀ value for our donor-acceptor pair was 54 Å (Fig. 2, B and C). Based on the measured values of τ_D and τ_DA, the estimated distance between Tb³⁺ and the mCherry chromophore is 69.6 ± 0.9 Å (R, in Fig. 1 B). Thus, the calculated S4 length of LBT-CiVS-mC in unclamped cells was 29.8 ± 0.9 Å (Fig. 3 B), which is within the theoretical range expected, between 28.5 and 38 Å, and is in good agreement with a mostly α-helical S4 conformation (28.5 Å). Two-electrode voltage-clamp and COVC measurements indicated that oocytes expressing LBT-CiVS-mC or LBT-CiVS-mC-R99A have resting potentials between −10 and −20 mV. Therefore, LRET measurements in unclamped cells here would represent a mixed population of conformations present at said voltage range.

Figure 3.

LRET experimental conditions are apt to detect expected differences in length. (A) Sequences of LRET constructs: wild-type Ci-VSP, LBT-CiVS-mC, LBT-CiVS-(+6aa)-mC, and LBT-CiVS-(+14aa)-mC. (Shaded loops) LBTs. (Dashed rectangles) S4 portions of CiVS. (Solid rectangles) mCherry (8-235). (B) Measurement of distance c (Fig. 1) in LBT-CiVS-mC, LBT-CiVS-(+6aa)-mC, and LBT-CiVS-(+14aa)-mC using LRET. (Solid circles) Measured mean distances; error bars are mean ± SE.

Limitations of distance measurements

To evaluate whether we could detect a distance change of the expected magnitude, we made two pairs of constructs with insertions between residue S236 of CiVS and the beginning of the mCherry (Fig. 3 A).

The first pair, LBT-CiVS-(+6aa)-mC and LBT-CiVS-(+6aa)-mC-R99A, had the additional six residues, HQQMKA, of the original Ci-VSP (H237-A242) reinserted. The second pair of constructs, LBT-CiVS-(+14aa)-mC and LBT-CiVS-(+14aa)-mC-R99A, had a total of 14 amino acids: the 6aa, HQQMKA, from Ci-VSP plus the 8aa, MVSKGEED, from the original mCherry (M0-D7). Tb³⁺ luminescence was detected from both sets of constructs expressed in oocytes and recorded under unclamped conditions. The time constants of the Tb³⁺ signals were 2.42 ± 0.02 ms in LBT-CiVS-(+6aa)-mC-R99A, accelerating to 2.03 ± 0.01 ms in LBT-CiVS-(+6aa)-mC for the first pair, and 2.49 ± 0.01 ms in LBT-CiVS-(+14aa)-mC-R99A, accelerating to 2.26 ± 0.02 ms in LBT-CiVS-(+14aa)-mC for the second pair (see Table S1).

Based on these values, the estimated distances R between the Tb³⁺ and the mCherry chromophore are 71.3 ± 1.0 Å in LBT-CiVS-(+6aa)-mC and 78.7 ± 1.6 Å in LBT-CiVS-(+14aa)-mC. Hence, the distance c, which is obtained by subtracting 12.1 + 27.7 Å from distance R (Fig. 1, A and B, and see Fig. S1), is 31.5 ± 1.0 Å for LBT-CiVS-(+6aa)-mC and 38.9 ± 1.6 Å for LBT-CiVS-(+14aa)-mC (Fig. 3 B).

This result shows that our LRET measurements can detect changes in the Tb³⁺ signal caused by extension of the segment between the Tb³⁺ and the mCherry chromophore, even though the relation between the distance measured and the number of inserted residues was not linear (Fig. 3 B). Although we do not know the conformation adopted by the +6aa in our construct, in one of the Ci-VSD structures, that stretch appears helical and bent, making a ∼90° turn (15). Thus, if the 6aa stretch adopts the same configuration in our construct as in Ci-VSD, it is not unreasonable to see an extension of only ∼1–2 Å. Although the conformation of the 8aa native to mCherry is unknown, both the 8aa and the 6aa are not expected to be buried within the membrane, therefore the +14aa extension should be more mobile but will not necessarily adopt a helical configuration. Thus, it is not unreasonable to deviate from linearity in this experimental context.

For LRET, one needs a pair of time constants from Tb³⁺ decay signals: one time constant in the absence, and one in the presence, of an acceptor. In this study, we obtained the values of these pairs from independent experiments because they had to be measured from different constructs: the wild-type mCherry and the inactive mCherry. Most lifetime measurements exhibited errors in the range between ±0.01 and ±0.03 ms. The maximum error of ±0.03 ms in the measurements in the absence and in the presence of acceptor, respectively, can create a ±1.7 Å error in the distance through error propagation (see Materials and Methods), which means that a difference smaller than 3.4 Å cannot be discriminated in our experimental conditions. However, we would be able to detect distance changes of ∼9.5 Å, which is the predicted extension if the entire S4-helix were to convert from α to 3₁₀-helix. As we showed in Fig. 3 B, our LRET experimental setup could detect 9 Å differences in length.

LBT-CiVS-mCherry is functional

The gating properties of LBT-CiVS-mC were studied using the COVC technique (30). Gating currents during activation and after prolonged depolarization were measured using the pulse protocols shown in the insets of Fig. 4 A. Also in Fig. 4 A, representative gating currents recorded with each protocol, Q_−120mV and Q_+40mV, are shown. The Q-V relation and the voltage-dependence of gating current kinetics for Q_−120mV (solid circles for Q-V and solid triangles for kinetics) and for Q_+40mV (open circles for Q-V and open triangles for kinetics) are shown in Fig. 4 B. These results confirmed that LBT-CiVS-mC is functionally expressed in oocytes. Although the general gating properties of LBT-CiVS-mC were comparable to those reported for Ci-VSP-C363S (the wild-type Ci-VSP with an inactivated phosphatase (5, 18)), the Q-V relation was leftward-shifted and the time constants of gating current were faster (see Fig. S2). Interestingly, the kinetics of Q_+40mV in this construct did not become slower nor did the Q-V relation shift further in the direction of hyperpolarization (Fig. 4 B) (18); this indicated that LBT-CiVS-mC either does not reach the relaxed state or that the active state coincides with the relaxed state.

LRET measurements of CiVS-S4 length under voltage-clamp conditions

Using LRET, τ_D and τ_DA were measured under several voltage-clamp conditions (insets in Fig. 4, C and D). To begin, τ_D and τ_DA were measured using a pulse protocol (V_Pulse) in which the laser pulse was applied 50 ms after the beginning of the depolarizing pulse, a time when gating charges are expected to be fully activated judging from the time course of the gating currents (Fig. 4, A and B). The τ_D and τ_DA were also measured in steady state (V_Hold protocol) while the membrane was clamped at −100, 0, or 80 mV (Fig. 4 D). Based on the values measured, the estimated S4-helix length (c in Fig. 1 B) is plotted in Fig. 4, C and D as a function of voltage. In these plots, the fine dotted lines indicate the estimated length of an entire S4 segment as an α-helix, whereas the long dashed lines indicate the value when the entire S4 is a 3₁₀-helix. These plots show that the distance c is close to the length that corresponds to an S4 segment entirely in α-helical conformation for all the voltages tested during pulses and in steady state. Although there is variation in the length between points (3–4 Å), these results show that the S4-helix in LBT-CiVS-mC does not undergo an extended long-lived α-3₁₀ transition. Even though there is uncertainty on the relative alignment of the LBT and the mCherry with S4 (which we assume to be aligned), the absolute distance determined here suggests that the secondary conformation of the CiVS-S4 remains mostly α-helical, which is in agreement with recent crystallographic data showing the Ci-VSD S4 as a pure α-helix (15).

LRET measurements of the length of the NavAb S4-helix in Ci-VSP scaffold under voltage-clamp conditions

So far, we have shown that the S4-helix of LBT-CiVS-mC does not undergo an extended long-lived α-3₁₀ transition during voltage-sensor operation. Our results raise the question of whether this finding is specific to Ci-VSP because, in the case of NavAb, S4 appears mostly in a 3₁₀-helix in the available crystal structures (13, 14). To test whether the absence of a 3₁₀-helix in CiVS-S4 reflects a unique characteristic of Ci-VSP, we performed similar experiments using a chimeric construct in which the S4 segment of NavAb was introduced into the Ci-VSD scaffold. Comparison of the sequences of the Ci-VSP and NavAb S4s led to two possible alignments, depending on which arginine is taken as the first S4 charge. Hence, we made two different chimeras: NavAb-S4 Ver.1, in which the segment G216 to S236 in LBT-CiVS-mC was substituted by the segment F95 to Q115 in NavAb; and, NavAb-S4 Ver.2, in which segment G216 to S236 in LBT-CiVS-mC was substituted by segment S92 to A112 in NavAb (Fig. 5 A).

Figure 5.

Measurement of the S4-helix length of NavAb embedded in the Ci-VSD under voltage-clamp conditions. (A) Chimeras harboring the NavAb S4 in the Ci-VSD. (Left panel) Q-V curves for NavAb Ver. 1 (open triangles) and Ver. 2 (solid triangles). (Right panel) Amino-acid sequences of LBT-CiVS-mC; NavAb-S4 Ver.1; and NavAb-S4 Ver. 2. (B–E) LRET measurements of distance c in NavAb-S4 Ver. 1 (open triangles) using the V_Pulse protocol (B) and the V_Hold protocol (C), and in NavAb-S4 Ver. 2 (solid triangles) using the V_Pulse protocol (D) and the V_Hold protocol (E). Data are mean ± SE. (Horizontal lines) As in Fig. 4, C and D.

The gating properties of these chimeras were studied using the COVC technique (see Table S2). Gating currents during activation were measured using the same pulse protocol as shown in Fig. 4 A. The corresponding Q-V curves are shown in Fig. 5 A. Both chimeras exhibited saturated charge-versus-voltage relations, but the slope factor in the Q-V curve of NavAb-S4 Ver.1 (open triangles) was shallower than that of NavAb-S4 Ver.2 (solid triangles).

We measured τ_D and τ_DA of these two chimeras using the same V_Hold and V_Pulse protocols shown in Fig. 4, C and D, and estimated the S4-helix length for each NavAb chimera (Fig. 5, B–E). The results indicate that, similarly to the native CiVS-S4, the NavAb-S4 transplanted in the Ci-VSD does not undergo long-lived transitions from fully α to fully 3₁₀ conformations, or vice versa.

Time-resolved FRET experiment: strategy to detect short-lived α-3₁₀ helix transitions of the S4 segment

Because of the relatively long lifetime decay of Tb³⁺ luminescence, the LRET experiments cannot rule out the existence of reversible submillisecond α-3₁₀ transitions. To explore the existence of such short-lived transitions, we performed time-resolved FRET experiments using the cut-open oocyte epifluorescence technique (24). We mutated residue A222 of the LBT-CiVS-mC construct to a cysteine residue, creating the FRET construct LBT-A222C-CiVS-mC, and the control construct LBT-A222C-CiVS-mC-R99A. We then conjugated ATTO 425 maleimide to the introduced cysteine to become the donor while mCherry was kept as the acceptor (Fig. 6, A and B).

Figure 6.

Time-resolved FRET constructs. (A) Amino-acid sequences of LBT-CiVS-mC, of Ci-VSP and of LBT-A222C-CiVS-mC S4 regions. If the segment A222C to F234 in LBT-A222C-CiVS-mC changed from entire α to entire 3₁₀-helix, its length would go from 18 to 24 Å. Assuming a linker length of ATTO 425 of 15 Å, and also the length between F234 and the chromophore of mCherry of 27.7 Å (Fig. 1A), the distance between donor and acceptor would be between 60.7 and 66.7 Å. (B) Schematic representation of the FRET construct: LBT-A222C-CiVS-mC. (Star, S4, top) Position of A222C, which is labeled with ATTO 425 maleimide. (C) Energy transfer efficiency as a function of distance. Representative theoretical curve for transfer efficiency between ATTO 425 and mCherry, assuming an orientation factor, κ², of 2/3. R₀ is the distance for 50% energy transfer. If the whole S4-helix had a transition from α to 3₁₀-helix during gating, energy transfer efficiency should decrease from ∼25.3% (60.7 Å, α-helix) to ∼16.2% (66.7 Å, 3₁₀-helix). (D and E) LRET measurements (mean ± SE) of distance c in FRET construct using the V_Pulse protocol (D) or the V_Hold protocol (E). (Dotted lines) Estimated values in the case segment A215-F234 in Fig. 6A was a whole α-helix (fine dotted line) or a whole 3₁₀-helix (long dashed line).

To predict whether FRET is possible, we computed the distance between ATTO425 and mCherry. Making the assumption that both donor and acceptor emit isotropically, we used κ² = 2/3 to compute the R_o for the pair ATTO 425-mCherry, found to be 50.7 Å (Fig. 6 C). Based on our LRET results and assuming the linker between chromophore and maleimide-base of ATTO 425 to be 15 Å long, this R_o value should be long enough to detect energy transfer between ATTO 425 at A222C and the mCherry chromophore (Fig. 6, A and B). In this case, if the whole S4-helix had a transition from α to 3₁₀-helix during gating, FRET efficiency should decrease from ∼25.3% to ∼16.2% (Fig. 6 C), i.e., within the region in which a change in distance can be detected.

Time-resolved FRET measurement of FRET constructs

We first confirmed with LRET that the FRET constructs do not undergo long-lived α-3₁₀ transitions (Fig. 6, D and E) mirroring the results with LBT-CiVS-mC. Fluorescent signals were then recorded under COVC conditions (Fig. 7, A and B) from Xenopus oocytes expressing each FRET construct (24). The ΔF/F (ratio of voltage-dependent signal changes to the total fluorescence) increased in a voltage-dependent manner (Fig. 7, B and E).

Figure 7.

Time-resolved FRET measurements. (A) Schematic representation of FRET experiment. ATTO 425 in the FRET construct is excited by 425-nm light (blue) and the emission is detected through a 665-nm emission filter (Dark red). In the lower panel, excitation (blue) and emission (dark red) filter spectra are shown superimposed on the spectra of donor (ATTO 425, absorption in dotted green line and emission in solid green line) and acceptor (mCherry, absorption in dotted red line and emission in solid red line). (B) Voltage-dependent FRET signals (ΔF/F) detected with the pulse protocol shown (inset). (Dotted square) Acquisition period. (C and D) Comparison between the kinetics of gating charge and those of the FRET signals. Data for pulses to 80 mV and −20 mV are shown in panels C and D, respectively. (Gray lines) Fits of FRET signals to single exponential functions; (black lines) gating charge moved. FRET signals contain a slow component not seen in the gating charge movement (see also Table S3 in the Supporting Material). (E) Q-V relation of the FRET construct after labeling with ATTO 425 (blue circles) and voltage-dependence of FRET signals (red bars). Error bars are mean ± SE. FRET efficiency increases in a voltage-dependent manner. (F) Anisotropy of ATTO 425 (green) and of mCherry (red). Neither shows voltage-dependent changes. ATTO 425 in position A222C and mCherry exhibits a relatively high anisotropy (>0.1).

As controls, we recorded signals concomitant with voltage changes from oocytes expressing the control FRET construct, LBT-A222C-CiVS-mC-R99A, labeled with ATTO 425 (Fig. 8 A) and from oocytes expressing the unlabeled FRET construct (Fig. 8 B).

Figure 8.

Controls for FRET experiments. Donor and acceptor emission signals obtained by direct excitation at 425 nm filtered through the 665LP filter, the acceptor emission channel (A and B); and through the 500/20 filter, the donor emission channel (C and D). (A) Illustration of LBT-A222C-CiVS-mC-R99A labeled with ATTO 425 (left panel) and the filter set transmittances superimposed on the ATTO 425 spectrum (middle panel). Voltage-dependent donor emission signals from LBT-A222C-CiVS-mC-R99A, labeled with ATTO 425 observed through the 665LP filter (right panel). Donor excitation at 425 nm does not result in fluorescence changes of ATTO 425 at the acceptor emission channel during depolarization. (B) Illustration of unlabeled LBT-A222C-CiVS-mC (left panel) and the filter set transmittances superimposed on the spectra of mCherry (middle panel). No detectable voltage-dependent acceptor emission signals through the 665LP filter from unlabeled LBT-A222C-CiVS-mC when the acceptor is excited directly at 425 nm (right panel). (C) Illustration of LBT-A222C-CiVS-mC-R99A labeled with ATTO 425 (left panel), and the filter set transmittances superimposed on the spectrum of ATTO 425 (middle panel). Recordings of voltage-dependent donor emission signals observed through the 500/20 filter from LBT-A222C-CiVS-mC-R99A labeled with ATTO 425 and with inactivated mCherry (right panel). ΔF/F shows a small voltage-dependent increment, possibly by a change in local quenching. (D) Illustration of LBT-A222C-CiVS-mC labeled with ATTO 425 (left panel), and filter set transmittances superimposed over the spectra of ATTO 425 in presence of mCherry (middle panel). Voltage-dependent donor emission signals observed through the 500/20 filter from LBT-A222C-CiVS-mC labeled with ATTO 425 (right panel). Signals show a small voltage-dependent decrease in ΔF/F that likely results from a large decrease of donor fluorescence due to an increase in energy transfer to mCherry during the pulse (see Fig. 7B), subtracted by an increase in fluorescence due to a change in local quenching of ATTO 425.

No changes in fluorescence were detected in either case. However, when we compared total donor emission in the presence of acceptor, we found that it was decreased relative to that obtained in the absence of acceptor (Fig. 8, C and D). Taken together, these results indicate that the voltage-induced change in ΔF/F seen with the FRET construct labeled with ATTO 425 is the result of a change in energy transfer from ATTO 425 to mCherry and that FRET efficiency increases in a steady-state manner in response to membrane depolarization (Fig. 7 E). Moreover, the kinetics of these FRET signals were found to contain a slower component when compared to that of gating charge movement, especially during activation (Fig. 7, C and D, and see Table S3). Interestingly, there was no transient change in the ΔF/F signals, indicating that a short-lived α to 3₁₀ interconversion of the whole S4-helix in response to a voltage pulse is unlikely.

The fact that some S4 segments display a significant 3₁₀ conformation in depolarized crystal structures led to the hypothesis that the S4 may extend from α-helix in the hyperpolarized conformation (resting state) to a 3₁₀-helix in its active state. In our experiments this would translate into a decrease in FRET efficiency (increase of the length of the S4). However, this expectation is based on the assumption that the relative orientation of the dipole moments of the donor and acceptor chromophores are randomized, i.e., both donor and acceptor molecules undergo rapid and unrestricted tumbling (isotropic emission and absorption with κ² = 2/3). To test whether this assumption was correct in this FRET case, we measured the anisotropy of ATTO 425 labeled at position A222C and of mCherry inserted at the intracellular end of S4.

As shown in Fig. 7 F, neither exhibited voltage-dependent changes in anisotropy, which means that changing the voltage does not significantly change the way each chromophore tumbles (see Table S4). Unexpectedly, the absolute anisotropy of both donor and acceptor is relatively high (0.12 ± 0.01 for mCherry and 0.19 ± 0.01 for ATTO 425), indicating that the tumbling of the dipole moments is somewhat restricted. Therefore, it is quite likely that what we see with FRET is the product of a rather large change in the relative orientation of the dipoles without a change in the longitudinal distance between ATTO 425 and mCherry probes. The relative reorientation of probe dipoles in the S4 segment, for instance by a rotation or tilt from the top or bottom of the S4, could explain the depolarization-induced increases in FRET efficiency (see Fig. S4 and Fig. S5).

Discussion

In this study, we report that the native S4-helix of the isolated VSD from Ci-VSP does not exhibit long-lived transmembrane length changes upon voltage stimulation. The transplantation of the NavAb-S4 inserted into the Ci-VSD scaffold provided similar results. No short-lived changes of the whole S4 length in Ci-VSD could be detected using time-resolved FRET, indicating the unlikelihood of rapid and extended α-3₁₀ transitions. Hence, our study does not support the existence of α- to 3₁₀-helical extensions of the whole Ci-VSD S4, and the hypothesis that all the sensing charges must be aligned to physically be transferred across the voltage-sensor hydrophobic plug. However, our results are not incompatible with the concertina effect, which proposes that only a small S4 region moving across the electric field transiently adopts a 3₁₀-helical conformation (11, 31).

We designed LBT-CiVS-mC so that the positions of the LBT and the mCherry were brought as close as possible to improve energy transfer efficiency. Although we scanned the positions of LBT and mCherry to obtain shorter distances between them, most constructs did not express in oocytes (see Fig. S6). Other constructs, even if they expressed, showed gating currents contaminated with large leak currents or that the Tb³⁺ signal emission was very weak. According to our preliminary tests of these different constructs, the length of the segment between LBT and mCherry in LBT-CiVS-mC is the minimum to function properly, and this segment is supposed to be buried mostly within the membrane. Interestingly, the Ci-VSD crystal structure showed that S236, which is the residue connecting mCherry in LBT-CiVS-mC, located at the end of S4. In addition, mCherry (8–235) in LBT-CiVS-mC has almost no N-terminus linker and directly attaches to S236 in Ci-VSD. Therefore, the length between the end of S4 and mCherry is quite short. Although this might restrict the movement of mCherry somewhat, LBT-CiVS-mC showed robust gating currents and Tb³⁺ signals. Consequently, we decided that LBT-CiVS-mC was the best construct, having both an appropriate voltage-dependent function and enough robust Tb³⁺ emission signals through all voltages tested.

In LBT-CiVS-mC, we estimated the length of the S4 portion shown as c in Fig. 1 B by subtracting the length of the LBT and the mCherry (based on crystal structural data) from the distance measured by LRET experiments under the assumption that LBT and mCherry are aligned. In our measurements, there could be uncertainty in the estimate of the actual values of S4 length because we do not know the actual orientations of LBT and mCherry. However, it is unlikely that changes in their orientation would completely dominate the expected distance change (∼9.5 Å) because the LBT and, especially, the mCherry do not have much freedom to move, inasmuch as the joining segment is buried in the bilayer and, as stated above, the linkers are short. This strengthens our view that the S4 in the Ci-VSD scaffold does not extend fully during VSD activation. Furthermore, the estimated distance predicts an entirely α-helical S4 conformation for Ci-VSD, which is precisely what was recently observed from two Ci-VSD crystal structures trapped in different states (15).

To date, a large body of structural and functional data has allowed the characterization of three main conformations of the VSD: resting, active, and relaxed states (18, 32, 33). The relaxed state develops upon prolonged depolarization. However, our construct does not show the expected shift of the Q-V, indicative of the relaxed state, as shown in Fig. 4 B. Published data in Shaker LBT constructs showed that the S4 position in the relaxed state is higher than that in the active state relative to the membrane (22). Therefore, we speculate that mCherry may physically prevent the necessary outwardly-directed movement to populate the relaxed state from the active state. In LBT-CiVS-mC, we showed that the S4-helix does not undergo 3₁₀-helical conversions. However, because LBT-CiVS-mC does not seem to reach the relaxed state (unless it transitions to it very quickly so that there is no apparent shift of the Q-V curve) (33), we cannot exclude the possibility that the S4-helix in wild-type Ci-VSP could adopt a 3₁₀-helix when populating the relaxed state.

We also showed that the length of the NavAb-S4 in Ci-VSD does not exhibit long-lived α-3₁₀ transitions, even though the NavAb-S4 helix is mostly in a 3₁₀ conformation in the crystal structure. We propose possible scenarios to explain why NavAb-S4 in the Ci-VSD does not display a 3₁₀-helix.

First scenario: the NavAb S1-S3 segments may create a specific stereochemical environment required for the S4 to adopt its 3₁₀ structure, so that our NavAb-S4 chimeras may not reflect the real potential of NavAb-S4 conformation. It is well established that gating charges tethered to the S4 interact with negatively charged residues (counter charges) in the S1-S3 segments (34, 35, 36, 37, 38). These residues are generally located in two clusters on each side of the central hydrophobic layer, but the positions of the counter charges are not necessarily conserved between VGICs and VSPs. Hence, when attempting to align secondary structures of VSD among Kv1.2/2.1, NavAb, and CiVS-mC, there are misalignments between some countercharges (see Fig. S3). These differences may influence the S4 conformations.

Second scenario: in the NavAb crystal structure, the gating charges create salt-bridges with chemical groups from the VSD backbone that could be necessary to stabilize the S4 3₁₀ structure.

In line with this reasoning, our NavAb-S4 chimeras showed differences in their electrophysiological properties, predictions based on NavAb crystal structure, and LRET measurements. With respect to the electrophysiological properties, to our knowledge, gating currents from wild-type NavAb have not yet been measured, although the voltage-dependence of its ionic conductance has been reported (39). Pore opening in NavAb was observed at very negative voltages from ∼−130 mV with a G-V midpoint of ∼−98 mV. From these data, one could speculate that gating charge movement in NavAb would occur at voltages that are either similar or slightly more negative, as is the case for the NavAb homolog NaChBac (40), which in either case is different from our NavAb chimeras. If that was the case, the influence of the Ci-VSD scaffold environment on the NavAb S4 would be to stabilize the resting state, moving the whole activation voltage range to more positive voltages.

Per the crystal structure, the transplanted S4 of NavAb-S4 Ver. 1 would fit well into the Ci-VSD scaffold because that segment is mostly helical, as is the native S4. However, NavAb-S4 Ver.1 showed a Q-V relation with shallower slope factor than the rest of the constructs and does not seem to reach a fully closed state even at −120 mV. It is thus possible that NavAb-S4 Ver. 1 may not reflect the real NavAb-S4 conformation. On the other hand, to maintain the same number of residues in the constructs, the transplanted S4 in NavAb-S4 Ver. 2 contained three residues of the NavAb S3-S4 linker (S92–G94), which made Ver. 2 a construct with a longer flexible S3-S4 linker and a shorter helical region. Although NavAb-S4 Ver.2 showed good voltage-dependent function, the flexibility of LBT position may make the LRET measurement underestimate the real longitudinal distance of the transplanted S4 even if it formed a 3₁₀-helix.

The structure of the native S4 in NavAb may also be influenced by the pore domain through the S4-S5 linker, extending it to a 3₁₀-helix. Consequently, in terms of absolute values of our LRET measurement in the NavAb-S4 chimeras, our data does not warrant a strong conclusion on their actual conformation, α- or 3₁₀-helix. In addition, our results may not be generalized to other VSDs, because they do not exclude the possibility of the transition happening, or being short-lived, in other S4-based voltage sensors with different structures. However, our results do show little change in S4 length by voltage, indicating that the transition from full α to full 3₁₀ transition (or vice versa) is not necessary for gating to occur in our constructs.

The FRET results showed no evidence of a reduction, not even transiently, of ΔF/F upon depolarization, in contrast to the prediction of an increase in the S4 length upon VSD activation. The ΔF/F, in fact, increased with depolarization, which would indicate a reduction in donor-acceptor distance. FRET efficiency increased in a voltage-dependent manner but our LRET data does not show voltage-dependent distance changes. One could posit that the difference between LRET and FRET measurements may arise from the respective donors (LBT between L215 and G216, and ATTO 425 at A222) not being located at the same site. However, per the crystal structure, the distance between these two donor insertion points is small (<2 helix turns) relative to their distance to the acceptor, ∼70 Å. Also, the linker between the chromophore and maleimide-base of ATTO 425 contributes ∼15 Å, which positions the donors closer together. Although the voltage-dependent change of S4 length in FRET may be smaller than in LRET, a decrement in ΔF/F, not an increment, should have been detected if FRET could follow the distance changes as well as LRET (Fig. 6 C).

Thus, to interpret the discrepancy, we need to account for the difference between LRET and FRET. In LRET, Tb³⁺ emission is isotropic whereas the inserted ATTO 425 has a relatively high anisotropy (Fig. 7 F), which can influence the energy transfer efficiency in FRET (see Fig. S4). It is quite likely that the steady-state FRET signals are indicative of a conformational change produced by a change in the dipole-dipole orientation between the ATTO 425 and mCherry. The kinetics of the FRET signals contains a slow component not seen in the gating charge movement (Fig. 7, C and D). This result suggests that the FRET signals may also be tracking a conformational change occurring after the gating charges have traversed the electric field, produced by a rotation and/or a tilt of the S4-helix (see Fig. S5). However, we cannot exclude the possibility of a transient α/3₁₀ conversion of part of the S4 (not the whole S4), which would produce a decrease in energy transfer that would be dominated by the change in the alignment of the dipoles.

Conclusion

Our study indicates that in the voltage-sensor of Ci-VSP, the transport of the sensing charges does not require dynamic conformational changes of the whole S4 segment between α and 3₁₀ helical conformations. Our conclusion may not be generalized to VSDs present in VGICs, for which further investigations are needed.

Editor: Michael Pusch.

Supporting Citation

Reference (41) appears in the Supporting Material.