Mechanism of Voltage Gating in Potassium Channels

See allHide authors and affiliations

Science  13 Apr 2012:
Vol. 336, Issue 6078, pp. 229-233
DOI: 10.1126/science.1216533

Open and Shut Case

Voltage-sensing domains (VSDs) control the activity of voltage-gated ion channels to regulate the ion flow that underlies nerve conduction. Structural and biophysical studies have provided insight into voltage gating; however, understanding has been hindered by the lack of a crystal structure of a fully closed state. Starting from a structure of an open conducting state, a voltage-gated K+ channel, Jensen et al. (p. 229) used all-atom molecular dynamics simulations to show the conformational changes involved in switching to the closed, nonconducting state. Additional simulations revealed the major steps of channel activation. The computational determination of a closed state may guide development of drugs to treat channelopathies associated with this resting state.


The mechanism of ion channel voltage gating—how channels open and close in response to voltage changes—has been debated since Hodgkin and Huxley’s seminal discovery that the crux of nerve conduction is ion flow across cellular membranes. Using all-atom molecular dynamics simulations, we show how a voltage-gated potassium channel (KV) switches between activated and deactivated states. On deactivation, pore hydrophobic collapse rapidly halts ion flow. Subsequent voltage-sensing domain (VSD) relaxation, including inward, 15-angstrom S4-helix motion, completes the transition. On activation, outward S4 motion tightens the VSD–pore linker, perturbing linker–S6-helix packing. Fluctuations allow water, then potassium ions, to reenter the pore; linker-S6 repacking stabilizes the open pore. We propose a mechanistic model for the sodium/potassium/calcium voltage-gated ion channel superfamily that reconciles apparently conflicting experimental data.

Hodgkin and Huxley discovered that voltage-regulated ion flow underlies nerve conduction (1). Only decades later were voltage-sensing domains (VSDs) identified as controlling the activity of voltage-gated K+, Na+, and Ca2+ channels (24), shifting these proteins between activated and deactivated states in response to changes in transmembrane voltage (57). Different mechanistic models have been proposed to describe how conserved arginine and lysine gating-charge residues on a VSD transmembrane helix (S4) couple with the electric field to gate ion channel conduction (810). Some experiments suggest substantial S4 motion during gating (11), others far less (12). Also unresolved has been whether S4 moves through a largely static electric field or whether the VSD instead reshapes the field around S4 during gating. Even less clear has been how S4 triggers the attached channel pore domain to gate conduction. Finally, it has been unknown whether other motions, either subtle or large-scale, are involved in voltage gating (6, 13), largely because, unlike the open state, no crystal structure of a fully deactivated, closed-state voltage-gated channel exists.

To study the voltage-gated transition at the atomic level, we subjected the open conformation of the KV1.2/KV2.1 “paddle chimera” (10, 14) voltage-gated K+ channel to molecular dynamics (MD) simulations at both hyperpolarizing (V < 0 mV) and depolarizing (V > 0 mV) voltages over experimentally determined channel-gating time scales. Our all-atom system comprises the channel, either with (T1+) or without (T1) the modulatory, but functionally nonessential cytoplasmic T1 domain (15, 16), in a symmetric, neutral phospholipid bilayer [omitting modulatory, negatively charged lipids (17)], hydrated with 0.5 M KCl (18). The simulations were performed using a special-purpose machine designed for high-speed MD simulations (19).

At depolarizing voltages, the channel exhibited steady outward conduction through a fully hydrated pore cavity (fig. S1); K+ current and H2O/K+ permeation ratio (~1) broadly agree with experiment (20) and pore domain–only simulations (21). No noticeable conformational changes occurred, suggesting that the crystal structure (10) represents a fully activated, open state.

In marked contrast, at hyperpolarizing voltages the channel went from an open, inwardly conducting state to a closed, nonconducting state—the resting state. The channel exhibited a transient inward “tail” current, but conduction halted upon water leaving the hydrophobic pore cavity (dewetting) and concurrent pore closure (cavity collapse) at ~20 μs (Fig. 1A and fig. S1); the observed dewetting, as also seen in pore-only simulations (21), explains the osmotic sensitivity of the overall gating process (21, 22). Subsequent inward S4 translation and lateral loosening of the VSDs from the pore domain completed the transition over ~100 to 200 μs (Fig. 1A). Overall, the gating transition thus consists of the channel moving from a VSD-up, (VSD–pore apposed) conformation to a VSD-down (VSD–pore loosened) conformation (Fig. 1A, fig. S2, and movies S1 to S3). The T1+ resting state exhibited less VSD–pore separation than T1.

Fig. 1

VSD motion during gating. (A) Intracellular views of activated (+V, conducting) and resting (−V, nonconducting) states; dashed lines separate two subunits each of T1 and T1+ resting states. Decrease in pore cavity water molecules (red/blue graph; initially ~40) indicates dewetting and pore hydrophobic closure after ~20 μs at −V. Magnified views: water-filled and empty (closed) cavities; PVP (hydrophobic constriction) motif is purple. Increase of the R1(Q)–Ala351 distance (arrows) explains lack of resting-state interdomain cross-linking (47); both states are compatible with Shaker tryptophan tolerance mutagenesis data (48) (see fig. S3). (B) Sequential inward movement of the VSD S4 gating-charge residues; colored traces track R2 to K5 Cα positions. R2 translates ~15 Å (T1: 15.4 ± 2.5 Å, simulations 5 and 6; T1+: 14.3 ± 0.9 Å, simulation 9), in agreement with KvAP (11, 41) and Shaker (42) accessibility data and resting state-specific cross-linking of Shaker Ile230 (S2) and R1(Q) (49). (C) Local S4 helix rotation of R2 to R4 (simulations 5, 6, and 9). (D) VSD gating-charge displacements, Q(t) = Σi qi [f(zi,t) − f(zi,0)]. f(z) is the transmembrane fractional potential drop [colored background in (B)]. zi,t is the z position of VSD atom i at time t, qi, partial charge. The total gating charge, 13.3 ± 0.4 e, was estimated as the difference between the final displacements at +V and −V. (Inset) Fractional potential drop across the VSD (18) and gating-charge residue z positions. (E) Cα RMSDs for the entire channel (T1 and T1+), the pore domain, and a single VSD decomposed into S1 to S3a (loops omitted), S3b to S4, and S4 alone. (Inset) Schematic of the full channel.

The activated-state VSDs delay channel closure by preventing pore hydrophobic collapse into the intrinsically more stable closed state (21, 23). The time at which pore closure occurred (determined by the pore cavity hydration level) and the tail current persistence time observed here—both ~20 μs—are in line with the 20 μs experimental tail current time constant [temperature-corrected Shaker data (18, 24) (see fig. S1 and table S1)]. This closure time is ~10 times as long as what we found in pore domain–only simulations (21), indicative of a VSD-imposed delay on the closure. A control simulation with a different water model (18) also exhibited dewetting, after ~30 μs, with little gating-charge transfer (fig. S4). The ~100 to 200 μs taken to complete the activated-to-resting transition—including full gating-charge displacement (Fig. 1)—compares well with the ~300 μs observed experimentally for the slow off-gating component (25).

The S4 helix—bearing gating-charge residues R1 [neutral Gln in the chimera: R1(Q)], R2, R3, R4, and K5—is the main VSD moving part. S4 translated ~15 Å overall across the membrane in sequential steps while rotating ~120°, moving in a groove formed by the largely stationary S1 to S3a helices (Fig. 1, B and C, and movies S4 to S6). S3b, while more mobile than S1 to S3a, did not translate inward to a notable extent. R4—centrally located in the activated state at the point of strongest transmembrane electric field (Fig. 1, B and D, inset)—initiated gating-charge movement; Phe233, a central hydrophobic residue, separated the VSD extracellular and intracellular hydrated lumens throughout. R3 and R2 moved in turn, and inward S4 motion typically stopped when R1(Q) reached Phe233. These observations support a recent gating model that emphasizes sequential motion of S4 gating-charge residues past Phe233 (26).

As the gating-charge residues filed past Phe233, their side chains faced the aqueous VSD lumens, not the hydrophobic membrane or VSD core; transient salt bridges formed by the gating-charge residues with acidic residues on S1 and S2 and with lipid negatively charged phosphodiester groups facilitated the transition. Relative to the activated state, the resting state had fewer intra-VSD salt bridges but more S4-phosphate interactions (fig. S2). Our findings are consistent with VSD–phosphodiester group interactions being functionally critical, whereas anionic lipids are modulatory, perhaps by interacting with gating charges or the T1 domain (17, 27).

In line with experimental gating-charge values of 12 to 14 elementary charges (e) (28), the complete transition into the fully relaxed state—all four VSDs “down” with all the gating-charge residues (R2 to R4) inward of Phe233—yielded a calculated gating charge of 13.3 ± 0.4 e (Fig. 1D and table S1). This VSD gating charge was solely accounted for by the ~15 Å translation of the S4 helices through an electric field, focused over ~15 Å, that we found to be similar in activated and resting VSD conformations (Fig. 1D and fig. S3). Our data indicate that limited motion of a single VSD (S4) toward the resting state suffices to close the pore. Charge displacements tied to early S4 motion—typically ~1 to 7 e with at least one R4 inward of Phe233 (table S1)—preceded pore closure, yet the pore always closed before all four VSDs were fully down. Experimentally, it is known that ~25% charge displacement suffices to close the channel (29).

Initial S4 inward motion disrupted the extracellular VSD–pore domain interface, resulting in domain separation: ~14 Å for T1 and ~5 Å for T1+, measured as the R1(Q)–Ala351 distance parallel to the membrane plane. VSD–pore separation contributed to a whole-protein root mean square deviation (RMSD) of >14 Å (Cα atoms, relative to the x-ray structure) for the T1 resting state and ~9 Å for that of T1+ (Fig. 1E); the linker between the T1 domain and the VSD—an additional constraint not present in T1—serves to reduce T1+ VSD mobility within the membrane plane (movies S5 and S6). By contrast, the tetrameric pore domain exhibited, relative to the open-state crystal structure, only a modest, ~3 Å RMSD increase, due to pore closure at ~20 to 30 μs. Translation of S4 alone, or as the main moving part of the S3b to S4 paddle, resulted in large, >10 Å RMSDs for S4 relative to the mostly stationary S1 to S3a helices (RMSD <3 Å).

Additional simulations revealed the major steps of channel activation (simulations 10 to 14) (table S1). We studied the first step of the resting-to-activated transition by subjecting the computationally determined T1+ resting state, obtained above (simulation 9), to depolarizing voltages. Experimentally, T1+ channels activate faster than T1 deletion mutant channels (16), presumably due to restraint of VSD mobility by the T1 domain.

Upon depolarization, helix S4 immediately moved ~5 to 10 Å outward, transferring ~50% of the total gating charge in ~75 μs (Fig. 2, A and B). Initially, gating-charge transfer is fast because most salt bridges between S4 and the rest of the VSD are disrupted in the resting state; as S4 moved outward, these salt bridges began transiently to re-form, leading to a gradual slowing of S4 motion and gating-charge transfer. As S4 movement neared completion, the VSDs reapproached the activated state, with RMSDs of ~3 Å and ~4 Å for S1 to S3a and S3b to S4 in the VSD returning closest to the crystal structure (Fig. 2A and movie S7). Experimentally, gating-charge transfer must be complete, in all four subunits, for the pore to open (8, 28). In line with these experimental observations, full S4 outward movement (and gating-charge transfer) in one or two VSDs was insufficient to open the pore in our simulations (table S1).

Fig. 2

Key steps of voltage-gated K+ channel activation. (A and B ) Deactivation (red), then early activation (blue; first ~120 μs). Paddle Cα RMSD relative to the initial structure, inward and outward S4 gating-charge residue movement, and total gating-charge transfer. (C to H) Late activation (final ~10 μs, after essentially complete gating-charge transfer). (C) Pore cavity rewetting, Leu331 (S5)–Pro405 (S6) side-chain interchange, and cumulative outward K+ permeation events. [(D) and (E)] K+ population of pore cavity and SF. (F) S4-S5 linker/S6 interaction energies. (Insets) S5 and S6 (ribbons), S4-S5 linker (spheres), and S6 (surface) before and after repacking. (G) S4-S5 linker/S6 contacts. (H) Upper gate (Ile402/SF site S5) lateral opening.

We simulated the second, and perhaps most enigmatic (6, 7, 13), activation step—the final cooperative transition (30, 31) to the open state—by starting from a computationally determined T1 configuration in which all VSDs, save one, were “up” (S4 helices fully outward, as in the activated state) but for which the pore remained fully or partially closed. [The partially closed pore cavity contains ~20 water molecules (table S1, simulations 10 to 12, and fig. S4); the fully closed state resembles that observed in NaVAb (32)]. These simulations, at depolarizing voltages, led to a fully activated and conducting state within a few tens of microseconds (Fig. 2, C to H).

Experiments have shown that a single, cooperative transition precedes conduction; this final transition, which contributes only ~5% of total gating charge, occurs after all VSDs have moved fully up (8). In our simulations, little or no additional gating-charge transfer occurred after the final outward motion of the S4 helix brought the S4-S5 linker into a tense conformation (table S1). This final motion perturbed the packing between the S4-S5 linker and the pore domain S6 helix, leading to packing fluctuations, as shown by the linker-S6 interaction energies (Fig. 2, F and G). This weakened and fluctuating packing enabled partial pore opening and rapid partial rehydration—water molecules reentered the pore cavity in less than 1 μs—allowing subsequent K+ entry and initial slow, outward conduction (Fig. 2, C to E).

The partially open, partially hydrated, and slowly conducting pore persisted for a few microseconds before reaching the fully conducting state (Fig. 2, C and E). The presence of ions in the cavity, which subsequently were driven outward through the selectivity filter (SF) by the depolarizing voltage, further increased pore rehydration (Fig. 2, C and D). As the hydrostatic pressure within the cavity thereby increased, the lower gate, at the pore Pro-Val-Pro (PVP) motif, fully opened, with the Leu331 (S5) and Pro405 (S6) side chains interchanging positions (21). This interchange caused S6 to kink at the PVP motif, widening the cavity and allowing it to finally become fully hydrated (Fig. 2C). Concurrent opening of the upper (hydrophobic) gate, at Ile402 (Fig. 2H), enabled SF site S5 to become populated with K+, thus increasing the cavity ion occupancy by one (Fig. 2, D and E). K+ presence in S5 allowed formation of the critical knock-on conduction mechanism intermediate [S5,(S4,S2)] (Fig. 2E) (21), causing the channel to assume a fully conducting state (20.7 ± 2.7 pA) (Fig. 2, C and E). Early conduction occasionally slowed as Leu331 and Pro405 transiently back-interchanged; after the final Leu331/Pro405 interchange took place, however, S6 and the S4-S5 linker settled into a closely packed configuration that stabilized the open pore, and thus the activated, conducting state (Fig. 2, C, F, and G).

Interaction of the S4-S5 linker and helix S6 is central to gating. In KV1.5, for instance, pairs of interacting linker and C-terminal S6 residues have been identified (33). Mapped onto KV1.2, these interactions are between residues Ile316 and Thr320 (linker) and Asn412, Phe413, Phe416, and Tyr417 (S6). These pairs, and linker residue Ala323, all stand out in residue contact maps (Fig. 2G), highlighting linker-S6 interaction as central to activation. Substitution of the KV2.1 linker and S6 C terminus into KV1.5, moreover, has been shown to confer KV2.1 activation and deactivation kinetics to this KV1.5 chimera, while preserving KV1.5 voltage dependency (33). This and similar findings (34) suggest, in line with our results, that the linker and S6 C terminus together ultimately govern pore gating, irrespective of the details of the VSD structural machinery that moves up or down during voltage sensing.

We propose a mechanistic model for voltage gating (Fig. 3) that integrates our analysis of atomic-level observations of both the activated-to-resting-state transition and key steps of the resting-to-activated-state transition (table S1). Our model incorporates Hodgkin and Huxley’s observations (1) that voltage-gated K+ channel opening requires the motion of four independent membrane-bound charged particles, whereas channel closing requires the motion of only one.

Fig. 3

Mechanistic model for voltage gating. Subjecting the activated state (1) to hyperpolarizing voltage initiates S4 inward movement and VSD–pore loosening. Ion depletion of the pore cavity (2)—coupled to inward motion of a single S4—leads to pore hydrophobic collapse. Closure of upper (Ile402 in Kv1.2) and lower gates [PVP motif; Leu331 (S5)−Pro405 (S6) side-chain interchange (21)] halts conduction (3). S4 continues inward; as S4 completes its inward motion, the S4-S5 linker helix moves fully down and the VSDs loosen from the pore, consolidating the resting state (4). Subjecting the resting state to depolarizing voltage drives S4 outward. When all four S4 and S4-S5 linker helices are fully up (5) and all VSDs repack against the pore, the lower gate becomes destabilized; the 45 transition constitutes the rate-limiting step in the activation process. Lower gate fluctuation triggers pore opening and partial pore rehydration (water molecules cooperatively enter the cavity) that allow ion entry and initial outward conduction (6); the 56 transition is essentially voltage-independent. The presence of ions drives complete pore rehydration that pushes fully open the upper and lower gates, returning the channel to the activated state (1). The lateral position of the VSDs (circles) relative to the pore domain (squares) is shown schematically (extracellular view).

Beginning with the activated state (Fig. 3, state 1), ion depletion, hydrophobic dewetting, and closure of the pore cavity, with concurrent early gating-charge inward movement, halt ionic conduction. Subsequent full VSD relaxation—inward S4 translation by ~15 Å relative to a largely rigid S1 to S3a VSD core coupled with ~120° S4 rotation that keeps the gating-charge residues pointed toward the VSD lumens, and lateral VSD–pore loosening due to VSD rotation and translation relative to the pore—permits the pore to remain closed (Fig. 3, state 4). Translation of S4 gating-charge residues accounts for the gating charge. Channel activation reverses these steps. The key difference is that all four VSDs must be up before the closed pore can reopen; a fully outward S4 perturbs the S4-S5 linker/S6 packing, thereby allowing water, and hence ion, reentry and subsequent conduction, stabilized by linker/S6 repacking.

Our observed gating mechanism and resting-state structure may help reconcile the many apparently conflicting measures of voltage-gated K+ channel conformation. Given natural sequence variability in functionally critical regions (S3b-S4 paddle and the interacting S4-S5 linker and S6), certain details of gating likely vary between channels, including between wild-type KV1.2 and the KV1.2/KV2.1 chimera that we used in our simulations. Yet, the fact that the S3b-S4 paddle can be swapped between channels to yield chimeras that retain voltage-gating function (35, 36) suggests that the key mechanistic aspects are broadly shared across the voltage-gated ion channel superfamily.

Our results suggest that channel opening and closing are energetically asymmetric processes. Because the intrinsically most stable state of the pore in isolation is dewetted and closed (21, 23), due to the hydrophobic lining of the pore cavity (21), S4 need not actively push the S4-S5 linker down to close the pore. Channel activation does, however, require depolarization-driven work—the forcing of S4 back across the membrane—to ultimately pull the S4-S5 linker tight, which perturbs its interaction with S6, leading to pore opening. Only when all gating-charge residues and S4-S5 linkers are fully up does the closed pore become sufficiently destabilized that fluctuations of the lower gate, through perturbed linker-S6 packing, allow partial and then—as the last conformational transition during activation—complete cavity rewetting (Figs. 2 and 3). The S4-S5 linker is tense in the activated state but relaxed in the resting state, perhaps explaining conservation of linker length: A shorter linker would inhibit channel closing, because S4 could not translate inward far enough, and a longer linker would inhibit opening, because even full S4 outward translation could not lead to an effective pull on the pore through the linker.

Our atomic-level computational determination of a resting-state conformation may be useful in guiding the development of drugs to treat those human channelopathies associated with the resting state (37). VSDs normally are impermeable to ions, but certain inherited S4 gating-charge mutations permit abnormal cation conduction through resting state omega pores (38), leading to, for example, cardiac long-QT syndromes, various paralyses, and migraine (37, 39, 40). VSD residues accessible from the intracellular and extracellular sides (41, 42) and residues thought to line the omega pore (43) map well onto our resting-state conformation (Fig. 4A). This resting-state VSD should thus exhibit omega currents, once an S4 gating-charge residue is appropriately mutated to a smaller, polar, uncharged residue (37, 44, 45). Such mutation of Shaker R1 to histidine or serine permits H+ or alkali cation and guanidinium currents (38, 46).

Fig. 4

The omega pore. (A and C) Resting-state conformations of chimera and R2Ser mutant VSDs. Spheres in (A) mark residues accessible to chemical modification from the extracellular (yellow) and intracellular (purple) sides in the resting state (42), consistent with ~15 Å S4 inward motion (Fig. 2A). (B) Water and K+ (R2Ser) densities (arbitrary units). The R2Ser mutation increases hydration at Phe233 by ~50%, facilitating K+ permeation. The ion permeation pathway (the omega pore) in the R2Ser mutant is shown as a gray mesh in (C). The green surface and spheres indicate predominant K+ sites and actual positions from a single permeation event; the blue surfaces represent VSD hydration.

We introduced an R2Ser mutation (37) into the resting-state conformation (R2 to R4 down) (18, 38). The mutant VSD exhibited significant inward K+ current (no Cl current) (movie S8, table S2, and fig. S5). This current arises because apposition of Phe233 with the mutated residue, which lacks the large, positive guanidinium group of the gating-charge residues, leads to increased hydration of the VSD hydrophobic constriction and thereby permits permeation of cations (Fig. 4, B and C). Depolarization halted the current and transferred ~2 e of gating charge (table S2).

The transition into the resting state, as well as the conformation of the state itself, demonstrates that the VSD omega and gating-permeation pathways are one and the same. Mutation of gating-charge residues enables pathological cation leaks through the VSD along the identical pathway taken by the physiological gating-charge guanidinium groups. We thus provide a structural explanation for hyperpolarization-induced (as well as depolarization-induced) cationic leak currents associated with channelopathies in certain human voltage-gated ion channels.

Supplementary Materials

Materials and Methods

Figs. S1 to S5

Tables S1 to S4

References (5066)

Movies S1 to S8

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank R. MacKinnon for useful discussions; K. Lindorff-Larsen, S. Piana, and K. Palmo for their help with force-field modifications; A. Philippsen and J. Gullingsrud for creating the animations; E. Chow, K. Mackenzie, and D. Scarpazza for developing simulation software; and R. Kastleman and M. Kirk for editorial assistance.
View Abstract

Navigate This Article