S4-based voltage sensors have three major conformations

Abstract

Voltage sensors containing the charged S4 membrane segment display a gating charge vs. voltage (Q–V) curve that depends on the initial voltage. The voltage-dependent phosphatase (Ci-VSP), which does not have a conducting pore, shows the same phenomenon and the Q–V recorded with a depolarized initial voltage is more stable by at least 3RT. The leftward shift of the Q–V curve under prolonged depolarization was studied in the Ci-VSP by using electrophysiological and site-directed fluorescence measurements. The fluorescence shows two components: one that traces the time course of the charge movement between the resting and active states and a slower component that traces the transition between the active state and a more stable state we call the relaxed state. Temperature dependence shows a large negative enthalpic change when going from the active to the relaxed state that is almost compensated by a large negative entropic change. The Q–V curve midpoint measured for pulses that move the sensor between the resting and active states, but not long enough to evolve into the relaxed states, show a periodicity of 120°, indicating a 3₁₀ secondary structure of the S4 segment when determined under histidine scanning. We hypothesize that the S4 segment moves as a 3₁₀ helix between the resting and active states and that it converts to an α-helix when evolving into the relaxed state, which is most likely to be the state captured in the crystal structures.

A number of membrane proteins respond to changes in the membrane electric field via an intrinsic voltage sensor in the protein structure (1). The classical examples, first described by Hodgkin and Huxley (2), are the voltage-dependent sodium and potassium conductances, which are crucial players in the generation and propagation of the nerve impulse. In the voltage-gated ion channels that generate these conductances, the movement of the voltage sensor generates a transient current that has been traditionally called gating current, because in the original recordings it was correlated with the opening of the conduction pathway in Na channels (3, 4). Gating currents are transient because the sensing or gating charges are tethered to the protein and so their movement is restricted within the membrane electric field. Most of the moving charges that produce gating currents have been identified as the four most extracellular basic residues of the fourth transmembrane segment (S4) in voltage-gated channels (5, 6).

At extremely negative and positive membrane potentials the gating charges are driven to extreme positions so that a plot of the transported charge as a function of the membrane potential (the Q–V curve) has a sigmoid shape that saturates at extreme potentials. The voltage dependence of this charge movement is an expression of the amount of charge involved and the number of states that the sensing charge populates when moving between the extreme positions. The steepness of the Q–V curve in going from one extreme position to the other increases by increasing the moving charge or by decreasing the number of intermediate stable states. Typically, in voltage-gated channels the gating charge starts moving at about −100 mV and saturates at ≈ +20 mV and the resultant Q–V curve can be fitted approximately with a two-state Boltzmann distribution having a half-maximal voltage or center point at ≈ −55 mV (7, 8). However, the voltage dependence of many channels exhibits hysteresis because it has been found that the center point of the Q–V curve depends on the value of the initial potential and how long the channel has been at that potential. This phenomenon was first described for the sodium channel gating currents of the squid giant axon (9) where it was found that holding the membrane potential for several minutes at 0 mV shifted the Q–V curve by −70 mV relative to Q–V curves obtained at −90 mV holding potentials, whereas the total transported charge (as assessed by the charge moving between the two asymptotes) was not changed by the holding potential. The shift of the Q–V curve was found to correlate with the development of slow inactivation of the sodium conductance. Although the same type of shift was later found in L-type Ca channels (10), the Shaker K channel (8), the HERG K channel (11), the bacterial Na channel (NaChBac) (12), and channels in the hyperpolarization-activated cyclic nucleotide-gated (HCN) family (13–15), it could not, in every case, be associated with entry into the slow-inactivated state of the channel. For example, the inactivation-deficient mutant S631A of HERG showed the same Q–V shift as the wild type (11). Also, a double mutant of Shaker eliminated its slow inactivation without modifying the shift of the Q–V while it induced a second open state with the characteristics of the slow inactivation of the channel (16). In addition, HCN channels deactivate at positive potentials rather than activate, and open at negative potentials, yet the direction of the Q–V shift is not reversed. Furthermore, EAG K channels (C. Y. Tang, F.B., and D. M. Papazian, unpublished work) do not exhibit conductance inactivation at all, but their Q–V curve shifts with initial conditions in the same manner as that associated with slow inactivation in other channels. These observations indicate that the Q–V shift and the inactivation of the conductance are not strictly tied and suggest that the shift of the Q–V by depolarizing the membrane for long times is an intrinsic property of the voltage sensor.

The discovery of the voltage-dependent phosphatase, Ci-VSP (17), showed that the voltage sensor motif found in voltage-gated ion channels is conserved and used in a membrane protein that has no conducting pathway. The Ci-VSP responds to membrane potential changes with a voltage sensor that has high homology to the sensor in voltage-gated channels. There is no crystal structure of the Ci-VSP, but its sequence indicates four transmembrane segments S1 through S4. S4 contains the canonical pattern of several basic residues, each separated by two hydrophobic residues; S2 and S3 segments contain acidic residues; and the intracellular C-terminal domain contains the phosphatase. On membrane potential depolarization, the Ci-VSP generates robust transient gating (or sensing) currents that increase the phosphatase activity. This protein is then another excellent model to study voltage sensing in S4-based voltage-dependent proteins, independent of its interaction with the conduction pathway of voltage-gated channels.

In this article we report that the Q–V curve of the voltage sensor in the Ci-VSP is also shifted to more negative potentials when the membrane is maintained depolarized for long periods of time, confirming the idea that the hysteresis in the Q–V curve is an intrinsic property of S4-based voltage sensors. These results are consistent with our proposal that, in S4-based voltage sensors, a depolarization moves the sensor rapidly from its resting to active position. The active position is not stable, and after some short period it relaxes into a lower energy state that has been traditionally correlated with the inactivated state. We call that state the “relaxed” state and propose it as a third major conformation of the voltage sensor in addition to the classical resting and active states.

Results

Operation of the Voltage Sensor in the Voltage-Dependent Phosphatase, Ci-VSP.

When the membrane potential is changed, the voltage sensor responds by moving electric charges in the field. Experimentally, this can be seen as a gating or sensing current when the voltage is suddenly changed from one value to another. The experiments shown in Fig. 1 A and B were done under voltage-clamp conditions on an oocyte expressing the Ci-VSP [see Methods and detailed pulse procedure in supporting information (SI) Fig. S1]. In Fig. 1A, the membrane was held steadily at V_i = −60 mV and a series of pulses of 800-ms duration to voltages ranging from −120 to 140 mV in 10-mV increments (indicated as a superimposed family of black traces) were applied sequentially. The sensing currents elicited by each of these voltage pulses were recorded and shown as the superimposed family of blue traces. At all potentials the charge moves in the outward direction during the pulses.

Fig. 1.

Shift of Q-V in Ci-VSP. (A) Family of sensing currents of Ci-VSP (C363S) (blue traces) in response to a series of test pulses (shown in black below) ranging from −120 to 140 mV from a holding potential of −60 mV (V_i = −60 mV). In the superimposed sensing currents, the voltage of the pulse is indicated for two of the traces. (B) Family of sensing currents for the same oocyte (red traces) for a series of test pulses (shown in black below) ranging from −120 to 100 mV with a 5-s conditioning prepulse to +80 mV (V_i = 80 mV) precedes each pulse. In the superimposed sensing currents, the voltage of the pulse is indicated for two of the traces. (C) Relationship between total charge movement and test pulse amplitude (Q–V curves) from an initial potential, V_i = −60 mV (blue squares) and V_i = +80 mV (red filled circles). The Q–V for V_i = +80 mV was displaced by the total amount of charge moved and plotted as the open red circles. The charge at each potential was determined by the time integral of the sensing current for each test pulse (see Methods) in A and B. Unsubtracted records.

Our objective is to determine the amount of sensing charge (Q) that moves at each membrane potential (V) and this was done by computing the time integral of the sensing current for each pulse. The Q–V curve thus obtained is plotted as the blue squares in Fig. 1C. When comparing the Q–V curve of the Ci-VSP (17) with the Q–V curve of voltage-gated channels, it is immediately apparent that it is shifted to more positive potentials. The Ci-VSP charge movement is not observed at potentials more negative than −20 mV and saturates at potentials positive to +100 mV (blue squares, Fig. 1C). Therefore, to study the hysteresis in the voltage dependence of this protein, the Q–V shift should be maximal between an initial potential, V_i, more negative than −20mV and one near +100mV. Because we found that a large nonspecific current develops in oocytes at such positive and maintained potentials, instead of holding the oocyte at high potentials, we used a conditioning prepulse to positive potentials before recording the sensing currents in response to test pulses. We previously determined that a 5-s prepulse was enough to establish the same effect on the sensing currents as the holding potential. Similar to the experiment shown in Fig. 1A, the pulses and the currents are shown in Fig. 1B. In this case, the sensing currents (shown as a family of red traces) for pulses ranging from −120 to 100 mV were applied sequentially (shown as the family of black traces), each preceded by a 5-s prepulse to +80 mV, making the effective initial potential, V_i = +80 mV. In this case, all sensing currents are in the inward direction during the pulses. The Q–V curve from V_i = +80 mV was obtained by integrating the pulse-evoked sensing currents and is shown in Fig. 1C as filled red circles. To facilitate the comparison of the Q–V curve for V_i = −60 with the Q–V at V_i = 80, the Q–V for V_i = 80 (red filled circles) was displaced along the charge axis by the total charge displaced between the asymptotes of the curve. The result is plotted as open red circles in Fig. 1C, where it is apparent that the Q–V curve recorded at V_i = +80 mV is shifted toward more negative voltages when compared with the Q–V curve recorded at a V_i = −60 mV, whereas the total transported charge is not affected by the starting potential.

A simple interpretation of these results is that a prolonged depolarization drives the voltage sensor to a new conformation that is more stable. This interpretation can be represented by a simple four-state model (R, A, AL, and RL), although we prefer to add one more intermediate state (L) because the resultant five-state model shown in Fig. 2A (see Methods) fits the data better. In this model the transition rate constants, α's and β's, are voltage-dependent but γ and δ are not. The “normal” movement of the sensor is voltage-dependent between the resting (R) state and the active state (A), which eventually evolves through a slow voltage-independent transition to the inactivated-like or relaxed (AL) state. The transition rates between the normal and relaxed states, γ and δ, are not voltage-dependent and therefore no gating charge is expected to move between those states. The backward rate constant δ is larger than γ. Therefore, in the normal initial condition of V_i = −60 mV, where β > α, most of the sensors will be in the R state. At depolarized potentials α becomes much larger than β and the sensor evolves to the A state. If the depolarization is maintained, the sensor evolves to the AL state because this state is stabilized with respect to the A state by a free energy w. One possible way to implement this stabilization is to decrease the rates leaving the AL state, δ and β₂, by the factor exp(−w/RT). As a first approximation, we considered the case in which the rates γ and δ are much slower than the rates α and β, so that we could simplify the model by separating the upper and lower pathways and fit them separately to the data obtained at two initial conditions.

Fig. 2.

Kinetic model of the voltage sensor. (A) Simple kinetic model of the major states of the voltage sensor. For details of the model, see Methods. (B and C) Symbols plot the normalized Q–V (B) and τ–V (C) curves obtained from initial hyperpolarized (open circles) and depolarized potentials (closed squares) from the pooled data of six experiments. The experimental time constants are from the decay of the gating currents and are the weighted average of a major fast and a minor slow component. The continuous lines are the fits to the model in A. Fitted parameter are in ms⁻¹: α₁₀ = 6.64, α₂₀ = 3.18, β₁₀ = 8.78, β₂₀ = 42.5 with w = 3.7 RT, z = 1.41 e₀, and x₁ = 0.143, x₂ = 0.252, x₃ = 0.25, x₄ = 0.385.

The model of Fig. 2A was simultaneously fitted to Q–V curves (Fig. 2B) and time constants of gating current decays vs. voltage (Fig. 2C), each taken from different starting potentials. The Q–V and τ–V curves are from the pooled data of six experiments. The parameters from the best fit are listed in the legend of Fig. 2. The salient features of the fit are that the apparent valence of the sensor is 1.4 e₀ and the stabilization energy was ω = 3.7 RT. These results indicate that a simple model with a stabilization energy of moderate magnitude for the transition into the AL state gives a reasonable fit to the data and provides a simple explanation for the shift of the Q–V curve by prolonged depolarization. In this model, we have assumed that the transition between the normal and relaxed pathways is much slower than the voltage-dependent transitions. We turn next to studying this voltage-independent transition with the aid of site-directed fluorescence labeling.

Site-Directed Fluorescence Labeling Indicates Two Conformational Changes of the Voltage Sensor.

Replacing glycine 214 at the top of the S4 segment by a cysteine (G214C) and labeling it with the cysteine-reactive fluorescent probe TMRM, Kohout et al. (18) recorded voltage-dependent changes in fluorescence in Ci-VSP. Fig. 3 shows voltage-dependent sensing currents (Fig. 3A) and fluorescence changes of the G214C mutant of Ci-VSP labeled with TMRM for V_i = −60mV (Fig. 3B). The fluorescence is quenched during the depolarizing steps that generate outward-sensing currents. The fluorescence changes reflect the time course of the sensor position. If we wish to compare the electrical recordings with the fluorescence changes, we cannot use directly the sensing current because the sensing current shows the rate of change of the charge position. Therefore, we compared the time course of the fluorescence change for each pulse with the charge position as a function of time, which was computed as the running integral of the sensing current during the same pulse.

Fig. 3.

Simultaneous electrophysiological and optical recordings of Ci-VSP G214C. Sensing currents (A) and fluorescence changes (B) recorded in response to test pulses from V_i = −60 mV. The test pulses ranged from −60 mV to +120 mV. (C) The fluorescence signal (black trace) in response to a test pulse of 80 mV from V_i = −60mV has been normalized and inverted to compare it with the integral of the corresponding scaled sensing current (red trace) and displayed on a logarithmic time scale to visually enhance the difference between the two time components. (D) By using the same normalization factors as in C, fluorescence (black traces) and corresponding sensing current integrals (red traces) are superimposed for all test pulses on a voltage range from −20 to 120 for V_i = −60 mV. (E) Time constants of the integral of the sensing current (black symbols) and the fluorescence (red symbols) are plotted against the membrane potential applied during a test pulse from a V_i of −60 mV. Filled symbols, fast component; open symbols, slow components. Average of six experiments. Compared with the Ci-VSP C363S, in Ci-VSP-G214C-C363S, the τ–V curves showed a shift toward negative potentials.

The fluorescence signals exhibit two distinct components of kinetics during the test pulse when V_i was −60 mV. By superimposing the time course of the charge position in response to a pulse to 80 mV with the time course of the normalized fluorescence change it was found that the fast component of the fluorescence change coincides with the time course of the charge position (Fig. 3C). This was observed for all depolarization steps above +20 mV, but there is a clear divergence of the two signals at longer times (Fig. 3D). This can be seen in more detail in Fig. 3C where the time scale has been plotted logarithmically to enhance the difference of time courses for short and long times.

It is possible to fit the charge position and the fluorescence changes with a sum of two decaying exponential functions. In Fig. 3E the fast component of the charge position (black squares) overlaps with the fast component of the fluorescence signal (red squares) over the whole voltage range, except for a small deviation at 0 mV. However, the slow component of the fluorescence change (red circles) was at least three times slower than the slow component of the charge position (black circles). These observations indicate that the fast component of the fluorescence is tracking the fast component of the charge movement, but the slow component of the fluorescence is tracking some other conformational change that is not displayed in the gating currents. One possibility is that the slow component of the fluorescence tracks the conformational change that occurs when the voltage sensor evolves from the A state to the AL state, which is the transition into the relaxed state.

The Kinetics of the Transitions Between the Normal and Relaxed States.

If the slow component of voltage-dependent fluorescence changes in Ci-VSP (G214C) is tracking the transition from A to AL, it should be possible to correlate the time course of the V_i-dependent shift of the Q–V curve with the time course of the fluorescence changes. To this end we measured Q–V curves as a function of the duration of the +80-mV conditioning prepulse. From a holding potential of −60mV, TMRM-labeled Ci-VSP (G214C) gating currents were recorded in response to a series of voltage pulses preceded by conditioning pulse to +80 mV for a variable period Δt, as indicated by the pulse protocol shown in Fig. 4A Inset. The voltage at which half of the charge has moved (V_1/2) for each of these Q–V curves was plotted as a function of the prepulse duration Δt (Fig. 4A, circles). We measured the shift with the labeled mutant G214C because it has a different voltage dependence and kinetics than the Ci-VSP-C363S shown in Fig. 1. Comparison with the fluorescence changes was done by pulsing the membrane from V_i = −60 to +80 mV for 2.5 s and the recorded fluorescence change was scaled and superimposed on the same graph (Fig. 4A, continuous line). There is excellent agreement between the time course of the Q–V shift, represented by the circles, and the fluorescence change in position 214, represented by the continuous trace (Fig. 4A). This result strongly suggests that the slow fluorescence change is tracking the conformational change between the A and the AL states.

Fig. 4.

Correlation between the time course of the fluorescence changes (continuous traces) and the time course of the V_i-dependent shift of the Q–V curves (open symbols) in the labeled Ci-VSP (G214C). (A) The V_1/2 of the Q–V curves were obtained by fitting each curve to a Boltzmann distribution and plotted as a function of the +80mV prepulse duration. Each point is from three experiments. The continuous trace is the fluorescence recorded for a pulse to +80 mV from V_i = −60 mV. (B) The V_1/2–t curves were obtained from the fit of Q–V curves from sensing current recordings when a prepulse of variable duration to −60 mV (recovery pulse) followed a 5-s conditioning pulse to +80 mV. The continuous trace is the scaled fluorescence recording for a pulse of 800 ms to −60 mV after the 5-s conditioning pulse to +80 mV.

It is expected that the fluorescence also tracks the recovery from the relaxed state. This assertion was tested by comparing the time course of recovery of the midpoint of the Q–V after being shifted by a depolarization as shown in Fig. 4B. The pulse protocol to measure the midpoint of the Q–V curve is shown in Fig. 4B Inset and the results are shown as the circles in Fig. 4B. The fluorescence was obtained from a single pulse of 0.8-s duration from to −60 mV, preceded by a 5-s conditioning pulse to +80 mV from −60 mV, which was the holding potential. As found previously for the case of V_i = −60, the recovery of the Q–V shift (circles) is in excellent agreement with the fluorescence change (Fig. 4B, continuous trace), confirming that the slow fluorescence change in position 214 tracks the transitions to and from the relaxed states.

Energetics of the Voltage Sensor.

The enthalpy and entropy changes between conformational states can be extracted from the effects of temperature on the transitions between states. We first measured the Q–V of Ci-VSP when V_i = −60 mV at different temperatures and we found no effect on the V_1/2 of the Q–V curve when the temperature was varied between 15 and 30°C (data not shown). This means that when the sensor evolves from the R to the A states the energy cost of that transition is entirely enthalpic with no entropic change (19, 20). In contrast, when we studied the transition from the active to the relaxed state, the results were quite different. We first estimated the temperature dependence of the decay rates as

when evolving from A to AL (forward transition) and when returning from RL to R (backward transition), respectively, by fitting the reciprocal of the slow time constants of the fluorescence signals as a function of temperature. The forward and backward time constants were measured between 15 and 30° and the plots of the reciprocal of the time constants are shown as Arrhenius plots in Fig. 5. From this plot we estimated the enthalpic barrier (see Methods) of the forward transition as 3.7 kcal/mol and the backward transition as 9 Kcal/mol.

Fig. 5.

Arrhenius plots that show the temperature dependence of the reciprocal of the slow time constant of the ON phase (squares) and the OFF phase (circles) of the fluorescence for V_i = −60 mV and a pulse to +80 mV. The enthalpic change ΔH was computed as ΔE − RT, where ΔE, the activation energy, was obtained from the slope of the plots.

To compute the overall entropic and enthalpic change when going from A to AL states we used the measured temperature dependence of the forward and backward rates and the value of the stabilization energy according to the procedure shown in Methods. The value of the stabilization energy w is 6 RT when the fluorophore is attached to the Ci-VSP, because the shift of the Q–V is even larger than in the wt Ci-VSP. We found (see Methods) that the transition from A to AL involved a large negative enthalpic change of −26.7 kcal/mol and a large negative entropic change of −23.4 kcal/mol (at 20°C). This means that the very large thermal stabilization of the AL state is destabilized by an entropic penalty that increases the order of the AL state making the total free-energy change only −3.3 kcal/mol.

The Shift of the Q–V Occurs with the Voltage Sensor Only.

It is important to know whether the V_i-dependent Q–V shift is an intrinsic property of the voltage sensor alone, independent of any interaction with a load. In the case of the Ci-VSP, the load is the phosphatase activity that is turned on by the conformational change of the sensor. To assess the influence of the phosphatase on the sensor properties, we made a mutant where the phosphatase part of the Ci-VSP was deleted, leaving only the voltage sensor (see Methods). We found that the off sensing currents elicited by depolarizing pulses from V_i = −60 mV became much faster, as if the phosphatase imposed a mechanical load that makes it difficult for the sensor to move back to its original position. Most importantly, we found that the Q–V curve is still shifted to more negative potentials when the protein was initialized with prolonged membrane depolarizations (see Fig. S2A). In fact, the shift of the phosphatase-less Ci-VSP was even larger than in the wt Ci-VSP. Taken together, these results are consistent with the idea that the voltage sensor itself undergoes the conformational change into the relaxed state and that the rest of the protein may modify its extent but it is not responsible for it.

The Q–V curve of the Ci-VSP is shifted to more positive potentials relative to the Q–V curves of many voltage-gated channels. However, the Q–V curve of Ci-VSP can be changed to the range observed in voltage-gated channels by a simple neutralization (of the gating charge), R217 to Q. We investigated the effect of initializing the R217Q mutant with prolonged depolarizations and found that the Q–V is also shifted to the left, and the Q–V shift is also increased relative to wt Ci-VSP and the initial potential required for the shift is less positive (see Fig. S2B). This is expected, because the transition to the relaxed state occurs from the fully active state that is reached at ≈0 mV in the R217Q mutant.

Discussion

The Normal and Relaxed States and the Structure of the Voltage Sensor.

The results described above show that the voltage sensor undergoes extra conformational changes in addition to the generally accepted voltage-dependent transitions between the resting and active states. We have presented evidence here that the active state, which is normally recognized as giving origin to the open state of voltage-gated channels, is a transient state that only lasts hundreds of milliseconds to a few seconds before converting into a relaxed state that is more stable. This means that at depolarized potentials, when all of the gating charge has moved, the voltage sensor will spontaneously drift into the relaxed state. Therefore, at 0 mV many of the K⁺ and Na⁺ channels will be naturally in the relaxed state. This has important consequences when relating structure to function because all of the crystal structures of voltage-gated channels have been obtained at zero membrane potential. These structures, therefore, most likely represent the relaxed state and not the active state. Because the available structures (21, 22) show the bundle-crossing gate open, they have been identified with the open state of the channel. However, it is known that after long depolarizations the voltage sensor in K⁺ channel is relaxed and the conductance is inactivated. This discrepancy between what the crystal structure and the functional data show is explained by the results of Cordero-Morales et al. (23) who found that inactivation in KcsA is the result of a conformational change in the selectivity filter region which blocks ion conduction even though the bundle-crossing gate is open. Several features of slow inactivation found in KcsA have counterparts in eukaryotic voltage-gated channels, making the Cordero-Morales explanation a viable one for the molecular basis of slow inactivation in voltage-gated channels (24). Therefore, the most likely functional state that correlates to the crystal structure of both Kv1.2 and the chimera of Kv1.2 with Kv2.1 is the open-inactivated state that has the sensor in the relaxed state, the bundle crossing is in the open state, but the pore is nonconducting because the selectivity filter is in the inactivated state.

The Origin of the Normal and Relaxed States.

The results presented in this article indicate that the voltage sensor may exist in two major modes: the normal mode (the upper path in Fig. 2A) and the relaxed mode (the lower path in Fig. 2A) giving origin to at least three major conformational states of the sensor: resting, active, and relaxed. The functional distinction between the states has been detected with electrophysiological methods, and the structural differences between them have been revealed by site-directed fluorescence changes that show local conformational changes and allow kinetic determinations. An immediate consequence of these results is that the active state, which is normally associated with channel opening in voltage-gated channels, is a transient state. This transient state has a lifetime of only milliseconds to a few seconds, thus imposing an important restriction when classical structural techniques are used to study conformational states of voltage sensors.

The energetics of the Ci-VSP shows that the energy cost in going between the resting and active states is purely enthalpic and it is driven by the membrane potential. This is similar to what has been found in studying Shaker gating currents, where only one component of the Q–V curve has an entropic change that is quite small (20). The lack of entropy change in the gating charge movement should not be confused with the large negative entropic change found in pore opening for both Na and K channels (19, 20). In contrast, the evolution from the active to the relaxed state of the voltage sensor has a large negative enthalpic factor that would make it very difficult to leave the AL state if the total energy difference was not reduced to ≈ −3 kcal/mol by a large negative entropic change. The transition rate from the A to the AL state has a low activation energy (4.3 kcal/mol), but because the actual measured rate is quite slow (in the seconds time scale) there must be a negative entropic component in the energy barrier between A and AL. Taken together, these results point to a decrease in its degrees of freedom or an ordering of the molecule during the transition to the relaxed state.

The differences between the normal and relaxed pathways must be the result of different conformational structures in the sensor and, in particular, it could be a change in the secondary structure of the S4 segment. In the crystal structures available, the S4 segments are mostly α-helical but there are one to two turns of the helix that are closer to a 3₁₀ conformation (21, 22). The possibility that the whole length of S4 that contains the gating charges is a 3₁₀ helix was suggested just after the first sequence of the eel Na channel was published (25, 26).

One way to explore the secondary structure of a transmembrane segment is to scan the entire length of the segment and study the impact of the scanning on the function of the protein (27). A direct way to test the impact of the scan on the voltage sensor function is to measure gating currents as the functional assay of the scan. To be useful, the amino acid used for such a scan must be tolerated by the majority of the sites tested. We did a partial histidine scanning of the S4 segment of Shaker K channel to understand the changes in exposure of the charged residues during voltage gating by replacing the charges (arginines or lysines) of the S4 segment one by one by histidine. This procedure revealed proton pores and proton transporters driven by voltage sensor movement between internal and external solutions (28–30). Because histidine was well tolerated, we extended this histidine scan to the noncharged groups of the S4 segment between the first and the fourth gating charges and we found that the voltage dependence of the Q–V curves of the histidine mutants underwent a periodic shift (Fig. 6A). The analysis of the periodicity by Fourier transformation according to the method of Cornette et al. (31) gives an angle of 120°, which is indicative of a 3₁₀ helix (Fig. 6B). At the time these experiments were done, there was no support in the data bank for such a long stretch of 3₁₀ helix and, although we suggested the possibility that the S4 was indeed a 3₁₀ helix, we also gave alternative explanations that could be satisfied by an α-helix structure (32). The recent article on MlotiK1, a non-voltage-gated cyclic nucleotide-regulated channel, shows an uncharged S4 segment that is all in a 3₁₀ conformation (33). This crystal structure gives strong support to the possibility that the S4 in voltage-gated channels is indeed a 3₁₀ helix. But then why do the crystal structures of the voltage-gated channels not show such a conformation? The results presented in Fig. 6 are obtained from Q–V curves measured from gating currents in response to short voltage pulses (100 ms) from a holding potential of −90 mV. Therefore, those Q–V curves represent transitions between resting and active states with no participation of the relaxed states, which require much longer pulses. This interpretation suggests that the 3₁₀ conformation indicated by the periodicity in the Shaker S4 voltage sensor is restricted to the normal pathway between resting and active states of the voltage sensor. The hypothesis that emerges from these results (see Fig. 6C) is that during movement of the S4 segment between the resting and activated states (the normal pathway) it may be in its 3₁₀ conformation and that it relaxes into a α-helical conformation only after prolonged depolarization, which is the structure shown in the crystal structures. The relaxed state is in a lower energy well than the active state and this is consistent with the fact that a α-helical configuration is more stable than a 3₁₀ helix. An enthalpic stabilization (in this case, −26 kcal/mol) is expected from a transition between a long stretch of 3₁₀ helix to an α-helix, but the actual free-energy difference measured in Fig. 6A is much less due to a negative entropic change, indicating a more disordered configuration when in the 3₁₀ conformation. The origin of the change in entropy between the two conformations is unknown and may involve side-chain rearrangements and interactions with the other segments of the protein or may even involve some ordering of the water molecules in the crevices accessible to the sensor. The conversion of α-helix to 3₁₀ helix may be aided by the extremely strong electric field where the charges of the S4 are embedded during the gating process (30) and, in this case, the 3₁₀ conformation may be short-lived and the conversion extremely dynamic. Future experiments using time-resolved spectroscopic techniques addressing the i to i + 3 or i to i + 4 differences between the two types of helices may provide an answer to this question that may be central to the operation of the voltage sensor.

Fig. 6.

Histidine scanning of the S4 segment and its impact on the Q–V curve in Shaker K⁺ W434F channel. Each residue of the S4 segment from R363 to R371 was mutated one at a time to histidine and the Q–V curves were measured in response to 100-ms pulses from a holding potential of −90 mV in neutral external pH. Position 369 did not express gating currents. All curves were fit with single Boltzmann distributions to determine the V_1/2 and valence, z, and these values were used to determine ΔG for each mutant. (A) Plot of ΔΔG (the difference between ΔG in each mutant and ΔG of the wt Shaker W434F channel) as a function of the residue number. (B) Plot of the Fourier transform of the periodic changes in ΔΔG analyzed according to the method of Cornette et al. (31). The peak of P(ω) is at 125° and the periodicity index (PI) is 3.57. The PI is 3.35 at 120°. (C) Hypothetical secondary structures involved in the operation of the voltage sensor: 3₁₀ helix in R–A path and α-helical in the RL–AL path.

Conclusions

Voltage sensors based on the movement of the charged S4 membrane segment have three major states: resting, active, and relaxed. In voltage-gated ion channels, the resting state is associated with the closed conductance, the active state is associated with the open conductance (and the opposite for HCN channels), and, in most cases, the relaxed state is associated with the slow-inactivated state of the conductance. The voltage-dependent phosphatase shows the same three major states and confirms that these states are a property of the voltage sensor itself. The transition from the active to the relaxed state seems to be voltage-independent and has been shown to involve a conformational change by site-directed fluorescence quenching. We hypothesize that the S4 segment moves as a 3₁₀ helix between the resting and active states and it converts into an α-helix when going to the relaxed state, which is the state observed in the available crystal structures.

Methods

Mutagenesis, Expression of Ci-VSP, and Oocytes Labeling.

Most of the experiments, unless noted, were done with the phosphatase-inactivated mutant C363S of Ci-VSP (17, 34) which was expressed in Xenopus laevis oocytes. The phosphatase-deleted mutant was done by deleting the C terminus of Ci-VSP from position 244. The DNA was linearized with NotI (New England Biolabs). The linearized DNA was transcribed by using T7 RNA polymerase (Ambion). Injected into each oocyte was 50 nl of 0.5–1 μg/μl RNA. The oocytes were incubated for several days at 18°C in a solution containing 96 mM NaCl, 2 mM KCl, 1 mM MgCl₂, 1.8 mM CaCl₂, and 5 mM Hepes, pH 7.5. For fluorescence recording, the cysteine engineered at position 214 (18) was labeled with tetramethylrhodamine-5-maleimide (TMRM) (Invitrogen). To do the labeling, Ci-VSP (G214C) expressing oocytes were incubated for 1 h in the incubation solution with 0.1 mM DTT. Subsequently they were washed with depolarizing solution [of 120 mM K-Mes (methylsulfonate), 10 mM Mops, and 50 μM EDTA] and incubated for 20 min into an ice-chilled depolarizing solution containing 20 μM TMRM. Finally, the oocytes were washed with incubation solution.

Electrophysiology and Fluorescence Recordings.

Sensing currents were measured 2–3 days after injection. Currents were recorded at room temperature (or at the indicated temperatures) with the cut-open oocyte voltage-clamp technique (35). The external recording solutions for experiments contained 120 mM NMG-Mes, 10 mM Hepes, and 2 mM CaCl₂, pH 7.3. The internal solutions contained 120 mM NMG-Mes, 10 mM Hepes, and 2 mM EGTA, pH 7.3. There was no subtraction of leak components during the acquisition; the subtraction, when needed, was done off-line. The charge (Q) was estimated as the time integral of the sensing currents at each potential after subtracting the leak baseline estimated at the end of the pulse. The time constants (τ) of the sensing currents were estimated from fittings of the sensing current decays as single- or double-exponential decays. In some cases a mean time constant τ was obtained from the two time constants of decay τ₁, τ₂ as τ = (A₁τ₁ + A₂τ₂)/(A₁ + A₂), where A₁ and A₂ are the coefficients of the two exponential decays. Electrophysiological and fluorescence data were filtered at 1–5 kHz and sampled at 2–20 kHz. The currents were measured from a holding potential of −60 mV and +80 mV in response to various test potentials (unless otherwise stated). Test pulses were at least 800 ms long and were applied with a 10-s interval. Data acquisition and analysis were carried out with in-house programs.

The experimental setup was as described previously (36) with some modifications. In brief, the setup consisted of a BX51WI microscope (Olympus) with optical filters suitable for TMRM (Invitrogen) excitation and emission acquisition. A 150-W tungsten-halogen microscope lamp, powered by a regulated power supply (Condor F15-15-A+) was used as excitation source. The excitation and light collection was done with a LUMPlanFl 40× water-immersion objective with N.A. 0.80. Light measurements were performed by using a PhotoMax-201-PIN photodiode (Dagan) controlled by a PhotoMax 200 amplifier (Dagan). A amplifier Dagan CA-1B was used for electrophysiological recording by using the cut-open oocyte voltage-clamp technique.

Kinetic Modeling.

A simple four-state model was initially used to test the data but it did not fit well the Q–V and τ–V curves simultaneously in the whole voltage range. We therefore added one more state (L) in the relaxed pathway as shown in Fig. 2A with a barrier separating L from RL and another barrier separating L from AL. To simplify, we assumed that there was no accumulation in state L (equilibrium condition) and solved for the resultant probability of being in AL (P_AL) and the equivalent time constant (τ) of the transition between RL and AL which are given by

Although there are other approaches to represent the voltage dependence of the rate constants, we used simple barrier models for the rates: α₁ = α₁₀ exp(zx₁FV/RT); α₂ = α₂₀ exp(z(x₂ − x₃)FV/RT); α₃ = α₂₀ exp(zx₄FV/RT); β₁ = β₁₀ exp(−z(x₃ − x₁)FV/RT); β₂ = β₂₀ exp(−z(1 − x₂)FV/RT). β₃ was computed from microscopic reversibility. V is the membrane potential, z is the total valence, and x₁, x₂, and x₄ are the fraction of the field where the peaks of the barriers are located and x₃ is the fraction of the field where the well of L is located. F is the faraday, R is the gas constant, and T is the absolute temperature.

Estimation of Entropic and Enthalpic Changes in the A to AL Transition.

The forward and backward rates of the A to AL transition given by

were estimated from the slow time constant of the fluorescence. By measuring these rates (as the reciprocal of the time constant of the fluorescence) as a function of temperature, we created Arrhenius plots that give a direct estimation of the activation energy ΔE for k_on and k_off. The enthalpic change is obtained from the relation ΔH = ΔE − RT. It is not possible to estimate the entropic change of the rates because we do not know the preexponential factor. However, we can estimate the overall entropic and enthalpic changes of the transition as follows.

By using the stabilization energy w, obtained from the Q–V shift, and the rates k_on and k_off (estimated as the reciprocal of the time constants of the slow component of the fluorescence) we solved for γ and δ as a function of temperature, T, and computed the equilibrium constants of the two transitions as

We define the elementary rate constants γ and δ as

where the subindices f and b in the enthalpic (ΔH) and entropic (ΔS) components correspond to forward and backward directions, respectively. We can replace γ(T) and δ(T) in the equations for K_on and K_off and, assuming that the preexponential factors (γ₀ and δ₀) are the same, extract the values of the overall enthalpic and entropic changes of the A to AL transition as