Intermediate states of the Kv1.2 voltage sensor from atomistic molecular dynamics simulations

Abstract

The response of a membrane-bound Kv1.2 ion channel to an applied transmembrane potential has been studied using molecular dynamics simulations. Channel deactivation is shown to involve three intermediate states of the voltage sensor domain (VSD), and concomitant movement of helix S4 charges 10–15 Å along the bilayer normal; the latter being enabled by zipper-like sequential pairing of S4 basic residues with neighboring VSD acidic residues and membrane-lipid head groups. During the observed sequential transitions S4 basic residues pass through the recently discovered charge transfer center with its conserved phenylalanine residue, F²³³. Analysis indicates that the local electric field within the VSD is focused near the F²³³ residue and that it remains essentially unaltered during the entire process. Overall, the present computations provide an atomistic description of VSD response to hyperpolarization, add support to the sliding helix model, and capture essential features inferred from a variety of recent experiments.

Voltage sensor domains (VSDs) are membrane-embedded constructs, which work as electrical devices responding to changes in the transmembrane (TM) voltage. They are ubiquitous to voltage-gated channels (VGCs) in which four of these units are attached to the main pore (1). During channel activation, the displacements of the charges tethered to the VSD give rise to transient “gating” currents, the time integral of which is the “gating charge” (GQR) translocated across the membrane capacitance. Phenomenological kinetic models devised to describe the time course of such currents are very diverse but all indicate that during VGC activation, the VSD undergoes a complex conformational change that encompasses many transitions (2–5).

Three main models have been proposed to rationalize the transfer of a large GQR across the low dielectric membrane in VGCs (6, 7). All are associated with the motion of S4, the conserved highly positively charged helix of the VSDs (8). In the sliding helix model (9, 10), the positively charged (basic) residues of the S4 segment form sequential ion pairs with acidic residues on neighboring TM segments and move a large distance perpendicular to the membrane plane. The transporter model derives from measurements of a focused electrical field within the membrane and suggests that during activation, the latter is reshaped. Accordingly, it is posited that S4 does not move its charges physically very far across the membrane (8). A third model was introduced following publication of the KvAP structure (11). Here, the position of the S3-S4 helical hairpin with respect to the pore domain suggested a gating mechanism in which the hairpin moves through the membrane in a paddle-like motion translocating S4 basic residues across the membrane, and reaching a TM position only in the activated state. Crystal structures of the Kv1.2 channel (12) and the Kv1.2-Kv2.1 paddle chimera (13) indicated later that the KvAP structure likely represented a nonnative state of the channel and its VSD (14). The newer structures also provided the opportunity to develop molecular models of the VSD response, most of which have converged toward the sliding helix model with S4 motion of 5–10 Å (15). Although the latter is much larger than proposed in the transporter model, this S4 displacement is below the estimate from avidin binding experiments, which suggested that KvAP S4 amino acids move as much as 15–20 Å across the membrane (16). Recently, charge reversal mutagenesis (17) and disulfide locking (18) were used to probe pair interactions within the VSD of different VGCs. Also, mutations with natural and unnatural amino acids, electrophysiological recordings, and X-ray crystallography were combined to identify a charge transfer center that facilitates the movement of positively charged amino acids across the membrane field (19). This important work enabled a dissection of VSD movements and their relation to ion channel opening. These experiments demonstrated the existence of the sequential ion pair formation involving the S4 basic residues, an essential feature of the sliding helix model.

Here, we employ molecular dynamics (MD) simulations in atomic detail to investigate the structure of the VSD transition states of the Kv1.2 channel embedded in a lipid bilayer subjected to a hyperpolarized potential. The structure of the channel in its open conformation (VSD up state, α) has been thoroughly examined in previous MD simulations performed in the absence of a depolarized TM potential (ΔV = 0 mV) (20). Here, an unconstrained MD simulation has been carried out on a Kv1.2 channel embedded in a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) bilayer while being subjected to a hyperpolarized TM voltage, ΔV, applied via a charge imbalance protocol (21, 22) (Fig. S1). This 2.2-μs MD trajectory starting from the up state, α, has uncovered the initial steps of the VSD response to ΔV, which involves two intermediate states (β, γ). Specifically, the VSD is observed to undergo transitions that involve a zipper-like motion of the six S4 basic residues (R1, R2, R3, R4, K5 and R6), in a sequential ion pairing with nearby VSD acidic residues and the membrane-lipid head groups. The final stages of the VSD response were uncovered by using biased-MD simulations, in which the S4 basic residues were constrained to move along the spontaneously initiated pathway until reaching the down state (ϵ) of the VSD. The combined MD simulations unveil and characterize five distinct states of the VSD that are involved in the deactivation process: the initial up state, α; three intermediate states, β, γ, δ; and the down state, ϵ.

Results

Electrical Response.

The electrical response of the system to a hyperpolarized potential ΔV was characterized by monitoring Q(t), the total GQR associated with the displacement of all charges in the system with respect to the membrane capacitor. To do so, and as in electrophysiology experiments, the ionic current through the main alpha pore was inhibited throughout the MD simulations by imposing harmonic constraints on the selectivity filter backbone residues. Fig. 1 reports the variation of the GQR and indicates a substantial channel electrical activity: Q(t) undergoes three major drops (I, II, and III) occurring approximately at t = 0.2 μs, 1.6 μs, and 2.0 μs, respectively, with associated gating charges Q(t) approximately - 1.4 e, -1.3 e, and -1.0 e (± 0.3 e), respectively.

Fig. 1.

Intermediate states of the VSD. (A) Evolution of the Kv1.2 channel gating charge, Q(t). along the 2.2-μs unbiased MD trajectory. Major Q(t) transitions are shown as I, II, and III. The standard error on each estimation of the TM voltage amounts to ± 50 mV. (B) Representative conformations (α, β, γ, δ, and ϵ) of the VSDs revealed during the unbiased and subsequent biased-MD simulations were characterized by monitoring the distances (cf. SI Text) between the S4 basic residues (numbered 1 to 6 for R1, R2, R3, R4, K5, and R6, respectively) and their binding sites (numbered 1 to 6 for top , E¹⁸³, E²²⁶, D²⁵⁹, E²³⁶, and bottom , respectively). The distances were calculated between the geometrical centers of the side chain atoms H₂N = C_ζ(NH₂)-N_εH-C_δH₂ (Arg), H₃N_ζ-C_εH₂ (Lys), HOOC_γ - C_βH₂ (Asp), HOOC_δ - C_γH₂ (Glu), and the lipid phosphate group from a representative conformation and averaged over all four subunits of the channel. (Bottom) The closest interacting pairs are shown. Note the zipper-like motion in which, for successive transitions, the pairs involving S4 basic residues are formed with lower countercharges. (C) Molecular views of the VSDs in the five key conformations highlighting the position of the S4 basic residues (blue sticks) and the salt bridges they form with the acidic residues (red sticks) of the other VSD segments or with the lipid head group moieties (yellow). The highly conserved residue F²³³ of S2 is shown as cyan spheres.

VSD Conformational States.

Throughout the MD simulation the Kv1.2 pore domain remained very stable, but the VSD undergoes substantial conformational changes involving zipper-like motion of the salt pairing interactions. The modifications accounting for the largest Q(t) variations (drops I to III in Fig. 1A) involved salt-bridge rearrangements within the VSDs as a result of the S4 basic residues moving from external to internal binding sites along the domain. These binding sites are specifically the acidic amino acids of segments S1 through S3 (E¹⁸³, E²²⁶, D²⁵⁹, and E²³⁶) and the moieties of the lipid head groups of the outer and inner bilayer leaflets (Fig. 1C). In order to correlate the changes in Q(t) during the unconstrained MD simulation with these ion pair rearrangements, we monitored the matrix of distances between the centers of the charged moieties of the S4 basic residues and those of the acidic countercharges (Fig. 1B), along with the rmsd of the salt pair distances for each conformation, calculated using Eq. S10 (Fig. S2). During the first Q(t) drop, the four VSDs relaxed from the initial α-state toward a metastable β-state. The latter bears a strong similarity with the intermediate kinetic state previously identified in independent work (22–24). In the subsequent Q(t) drop (II), the VSD of subunit 1 underwent further salt-bridge rearrangements, which led to yet another state, called here γ, lasting for over 0.5 μs. Then, in the third Q(t) drop (III), the same VSD underwent yet another conformational change.

The sequential conformational changes, α → β → γ, of the VSDs, and especially that of subunit 1, provide a reaction pathway for the first transition events occurring in the voltage-sensing process. In the latter, each conformation α, β, and γ is stabilized by a maximum number of salt bridges between the S4 basic residues and negative residues of the VSD or moieties of the lipids (Fig. 1C and Fig. S3). To uncover the complete VSD response from the up state (α) to the down state (ϵ), biased MD was used to simultaneously drag all the charged moieties of the basic residues from a given binding site to the next along the downstream path (see Materials and Methods). Along the biased-MD trajectory between the γ and ϵ states, a fourth VSD conformation was identified, called here the δ-state, which is also characterized by a specific network of salt bridges (Fig. 1 B and C). Equilibration MD runs, each spanning approximately 15 ns, confirmed the structural stability of these additional VSD structures, for which the distance matrix rmsd profile converged to a value < 2.4 Å (Fig. S2). Together with β and γ, the conformation δ was therefore considered as yet another possible intermediate state of the VSD. The electrostatic network (Table S1) stabilizing the VSDs in ϵ agrees with electrophysiology experiments (25). The conformation in state ϵ bears strong similarities with the molecular models of the VSD down state of Kv1.2 that have been proposed previously (15, 26–28). However, the present model places R1 in a lower position. Indeed, whereas R1 was initially thought to be engaged in a salt bridge with E²²⁶ in the resting state, recent evidence has appeared, showing that R1 is likely lower (19). To further validate the resting state of the VSD, we have evaluated the GQR associated with the transition from α to ϵ. The latter, most appropriate to compare with electrophysiology, is associated with the whole channel deactivation and amounts to 12.8 ± 0.3 e (Table S2), which is in good agreement with values obtained for Shaker-like channels (12–14 e) (29–31). The position of S4 basic residues in ϵ were further checked by probing the effect of mutations of the arginines, R1 and R2. In agreement with electrophysiology experiments (32), only the mutation of R1 into an uncharged homologue led to the destabilization of the VSD and to the appearance of an omega leak current through the latter under hyperpolarized voltages (33).

S4 Displacement.

The position of the geometrical center of the charged moieties [H₂N = C_ζ(NH₂)-N_εH-C_δH₂ (Arg), H₃N_ζ-C_εH₂ (Lys)] (Fig. 2A) and of their C_α backbone atoms (Fig. 2B) with respect to the center of the bilayer show that during the α → ϵ transition, the S4 backbone atoms undergo a substantial downward translation, estimated in the range of 10–15 Å, as the charged moieties move between 15–20 Å across the membrane to satisfy their electrostatic interactions with the counter charges. We monitored additionally the rigid body motion of the backbone atoms of the segment comprising residues R²⁹⁴ (R1) through R³⁰³ (R4) of S4. The center of mass of this segment moves downward (approximately 12 Å) as the VSD undergoes the four transitions, between the end states (Fig. 2C). When viewed from the extracellular face of the membrane, this S4 motion is accompanied by a slight helical tilt (approximately 15°); i.e., in state ϵ, S4 becomes slightly more perpendicular to the bilayer compared to state α, as described in early FRET and luminescence resonance energy transfer experiments (34, 35), a moderate clockwise helical precession (approximately 45°) and a significant counterclockwise helical twisting (approximately 90°) (Fig. 2 D–F). In agreement with other molecular models, significant overall translation of S4 does not occur before the rotation of its top (24, 26, 36). In spite of significant TM displacement, the S4 side chains move relative to a rather static transfer center during the process (Fig. 2H).

Fig. 2.

Extent of S4 motion. (A and B) Positions of the S4 basic residues R1 (black) through R6 (orange) with respect to the membrane center (z = 0). (A) Positions of the geometrical center of the charged moieties. (B) Positions of the main-chain C_α atoms. (C to G) Rigid-body movements of the S4 segment along the α → ε transition. Only the backbone atoms of the S4 residues R1 through R4 were included in the calculation. (C) Position of the segment center of mass with respect to the membrane center (z = 0). (D) Tilt Δϕ with respect to z, the bilayer normal. (E) Precession Δα around z, and (F) rotation Δω about the helical main axis Z^′ calculated as , Δα_i = α_i - α_α, and Δω_i = ω_i - ω_α, where i = {α,β,γ,δ,ε}. The S4 motions were determined after elimination of overall rotation and translation of the VSD structures, by fitting the gating charge binding sites (E¹⁸³, E²²⁶, D²⁵⁹, and E²³⁶) over the structures. (G) Coordinate systems and definition of various angles. (H) TM position (z) of the geometrical center of side chains of the charge transfer center residues (D²⁵⁹, E²³⁶, and F²³³) in each of the VSD conformational states. All the data were computed from the representative conformations considered in Fig. 1 and averaged over the four VSD subunits, with the error computed as the standard deviation from this average value.

Lipid Participation.

Recent studies have pointed to the crucial role the charges of lipid head groups may play in modulating the gating of VGCs. The presence or absence of the groups was shown to have a dramatic influence on VGC function: The activation of K⁺ VGCs may indeed be suppressed when the channels are embedded in bilayers formed by cationic lipids (37). Removal of the lipid head groups by enzymes also results in an immobilization of the VSD motion, thereby inhibiting VGC function (38, 39).

Here, as inferred from electrophysiology experiments, we show that lipids, and in particular their negatively charged head group moieties, provide countercharges for the S4 basic residues, during the gating process. Although earlier molecular models of the VSD (22, 24, 27) have shown that lipids from the upper and lower bilayer leaflets stabilize, respectively, the activated and resting states of the VSD, this study provides evidence that lipids also play this role in the intermediate states (Fig. 1C and Fig. S3).

Electrostatic Properties.

The TM-electric field is by essence the driving force for the response to membrane polarization (Fig. 3). In the presence of a TM field, a number of protein tethered charges cross the field giving rise to the experimentally measured gating charge, Q. In the following, we use the five states of the VSDs to gain further insights on the molecular/electric properties of the domain accounting for Q. For a structure undergoing a general α → β transition under ΔV, the gating charge Q may be linked (22, 40–43) to the variation of the free energy of the channel: , in which is the excess free energy in each conformation, λ due to the applied potential ΔV. is the so-called “electrical distance” of the residue, i. It accounts for the degree of coupling between the local electrostatic potential felt by q_i located at r_i and ΔV and is expressed as .

Fig. 3.

Electric field maps under a hyperpolarized TM potential in representative VSD conformations α and δ. (A) A representative Kv1.2 (α) VSD is located in the center of the panel and for clarity, only its backbone atoms (white ribbons), the conserved F²³³ residues (cyan spheres) and the basic (blue sticks) and acidic (red sticks) residues of its TM domain are shown. The local electric field direction is shown as black arrows. Closeup view of the charge transfer center region in the α (B) and δ (C) states highlighting the electric field pointing downward and rationalizing the downward movement of the basic residues of S4.

This formulation allows the identification of the specific molecular components that contribute to the gating charge (Fig. 4 A and B and Fig. S4). During the whole VSD transition from up to down states, 11.2 elementary charges were transported by the channel, of which approximately 10 e transported by the S4 helices. For the four consecutive transitions from α → ϵ, the cumulative GQRs measured for each VSD amounts to -0.52 e, -1.33 e, -2.17 e, and -2.8 e, respectively (Table S3). As shown in Fig. 4B and Fig. S4, substantial contributions arise from the S4 basic residues, which is consistent with experiments (4, 31, 44) and with recent MD simulations of the VSD (26, 27, 45). Fig. 4C reports the normalized GQRs transported by the S4 basic residues during each transition. It shows that the contributions from these residues are not uniform: During deactivation, the lower S4 residues carry more charges during the early transitions, whereas the upper ones transport more charges during the late transitions. Evidence from the raw data suggests therefore that the electric field in which the charges physically move is not uniform across the membrane.

Fig. 4.

Electrical properties of the VSD (A) Electrical distances, , for each TM residue in the α (black) and ϵ (orange) conformations and net charge per residue (green) along the Kv channel sequence (excluding the T1 domain). The position of the TM segments S1 to S6 is indicated by arrows. was normalized assuming and 0 for residues positioned, respectively, above 25 Å and below -25 Å from the bilayer center. The data were averaged over the four subunits of the channel. (B) Corresponding cumulative (orange line) and per-residue (bars) gating charges for the α → ϵ transition (basic residues in blue and acidic ones in red). (C) Contributions of each S4 basic residue to the normalized GQR associated with each transition, enabling the identification of the residue(s) transporting most of the gating charge (red arrow). Error bars correspond to the standard deviation from the average value over the four subunits. (D) Electrical distance through the VSD in each conformation as a function of z, the normal to the bilayer. (E) Gating charge (Q) and TM position (z) for the S4 basic residues in each of the VSD states α through ϵ. Circles represent the GQRs that were obtained with regular MD simulations, whereas triangles stand for the configurations that were obtained with biased MD. (F) Activation master curve, describing the dependence of Q with z.

The overall conformational change from the up to down states of the VSD necessitates a large physical TM displacement of the charged moieties of R1 to R6 amounting to approximately 15–20 Å (Fig. 2A). For the purpose of linking the physical displacement of these charges to the measured GQRs, it is interesting to consider , the electrical distance of the VSD residues located at the TM position z. In each channel conformation, these values collapse onto a single sigmoidal curve that provides a good estimate of the fraction of the potential the residues are sensing across the membrane. In Fig. 4D, we report this fraction of potential for the conformations α to ϵ of the VSD. Interestingly, seems not to vary substantially from one conformation to another. This is a clear indication that the local electric field within the VSD is not drastically reshaped during activation. Fig. 4E reports the cumulative gating charge evaluated for each S4 basic residue as a function of its TM position z. The data show clearly that residues R2, R3, and R4 located in the middle of S4 carry more charges than R1, K5, and R6 during the full α → ϵ transition. As expected from the previous analyses, these cumulative contributions all collapse onto a master curve, essentially , which itself describes the electrical activity of the VSD (Fig. 4F). The gating charges for any elementary transition are directly connected to the physical displacements of the S4 basic residues across the membrane through this master curve. Comparison of the latter to the voltage fraction across a bare bilayer underlines the reshaping of the electric field within the VSD due mainly to the distribution of charges and to the presence of solvent (Fig. S4). In support of the phenomenological models developed earlier, the electric field appears to be focused in the middle of the bilayer, in the sense that identical physical displacements along z result in larger GQRs when the translocation occurs around z = 0. The master curve describing the VSD electrical activity predicts that a total displacement of a given basic S4 residue of over 40 Å is necessary to carry a GQR of 1 e. However, over 70% of the latter can result from the displacement over a narrow window of 20 Å near the bilayer center.

Charge Transfer Center.

MacKinnon and coworkers (19) have recently identified an occluded site in the VSD of the K⁺ VGCs formed by two negatively charged residues (D²⁵⁹, E²³⁶ in Shaker) and the highly conserved F²³³, which appears to “catalyse” the transfer of each of the VSD basic residues across the membrane field. Here, the structural analyses fully corroborate this model; i.e., the site is shown to be occupied by the basic residues K5, R4, R3, R2, and R1 in the VSD conformations α, β, γ, δ, and ϵ, respectively. Furthermore, each of the four transitions α → β, β → γ, γ → δ, and δ → ϵ is shown to be accompanied by the translocation of a single residue (K5, R4, R3, and R2, respectively) through this charge transfer region (Figs. 1 and 3 and Fig. S5). Moreover, in each of the VSD conformations the TM position of the charge transfer center is near the center of the bilayer (Fig. 2H).

Discussion

Unbiased MD simulations of the membrane-bound Kv1.2 channel subject to a hyperpolarized potential have been used to reveal the initial steps of the VSDs participation in the channel deactivation mechanism. The MD simulations showed that the conformational changes taking place within the VS involve a zipper-like motion of the S4 basic residues in sequential ion pairing with nearby countercharges from the VSD and lipid head groups from both the upper and lower membrane leaflets. In addition to the two intermediate states (β, γ) arising from the unbiased MD trajectory, one more (δ) was identified using biased-MD simulations in which the imposed “reaction” pathway was consistent with the zipper-like motion. Importantly, the final down state (ϵ) arising from the biased MD is consistent with experimental constraints from the literature (24–26) and although the β- and γ-states have been generated by unbiased simulation, only the δ state required assuming a pathway that follows the sequential base pairing. The overall conformational change from the active to the resting states of the VSD results in a total gating charge in agreement with electrophysiology experiments and necessitates a large displacement of the S4 backbone amounting to 10–15 Å, which is in closer agreement with estimates from avidin-binding experiments (16) than previous models (15). The molecular conformations of the identified transition states are therefore consistent with the sliding helix model. Importantly, the states identified in the present MD simulations agree with the recent proposal that the VSD transitions involve sequential passage of the S4 basic residues through a catalytic center involving the conserved F²³³ residue and based on this mechanism, it was posited that the VSD can adopt five conformational states (19).

Perhaps the most interesting feature revealed by the present MD study is in the analysis of the electrical activity of the channel and the finding that the cumulative gating charge transported by the S4 basic residues can be described by a unique function, which defines the electromechanical coupling mechanism that the VSD charges undergo. Although in agreement with previous phenomenological models this coupling results from a focused electric field within the domain, the shape and intensity of which is hardly modified during deactivation. Although final conformations of the sensor have been generated using biased-MD simulations, the ones resulting from the unbiased MD trajectory contained enough information to reconstruct such a coupling (Fig. 4F), which strengthens confidence in our findings.

The molecular mechanism underlying voltage sensing has remained a matter of debate. The present computational study favors the sliding helix model and provides a molecular description of three intermediate states of the VSD. Though the present results are specific to the VSD of the Kv1.2 channel, they likely can be generalized to other VGCs. When combined with electrophysiology measurements, they should allow a better characterization of the molecular mechanisms implicated in the S4 residue mutations that underlie certain channelopathies.

Materials and Methods

Molecular Dynamics.

The MD simulations were carried out using the program NAMD2 (46), using now well-established simulation parameters (see SI Text for more information). The water molecules were described using the TIP3P model (47). The simulation used the CHARMM22-CMAP force field for the protein (48, 49) and CHARMM27 for the phospholipids (50). A united-atom representation was adopted for the acyl chains of the POPC lipid molecules (51). The TM voltage was imposed by explicit ion dynamics, and the creation of air/water interfaces (21, 22). The ionic current through the main alpha pore was inhibited during the simulation by imposing harmonic constraints on the backbone atoms of the selectivity filter.

Biased-MD Simulation—Generation of States δ and ϵ.

Further conformational change was triggered by applying “moving” harmonic constraints on the charged moieties of the S4 basic residues directed toward the charged group of the next binding sites (magnitude 1 kcal/mol/Ų and velocity of 0.00003 Å/fs) (cf. SI Text for details). The charged moieties correspond here to the side chain atoms H₂N = C_ζ(NH₂)-N_εH-C_δH₂ (Arg), H₃N_ζ-C_εH₂ (Lys), HOOC_γ-C_βH₂ (Asp), HOOC_δ - C_γH₂ (Glu) and the lipid phosphate group, . During this procedure, the C_α atom position of the protein’s negative binding site was fixed by “static” harmonic constraints. Moreover, to avoid spurious effects on the channel structure, this whole procedure was performed slowly, throughout successive cycles involving short MD runs (approximately 1 ns) with harmonic constraints turned on followed by equilibration (approximately 1 ns).