|Home | About | Journals | Submit | Contact Us | Français|
Voltage-gated ion channels generate electrical signals in species from bacteria to man. Their voltage-sensing modules are responsible for initiation of action potentials and graded membrane potential changes in response to synaptic input and other physiological stimuli. Extensive structure-function studies, structure determination, and molecular modeling are now converging on a sliding-helix mechanism for electromechanical coupling in which outward movement of gating charges in the S4 transmembrane segments catalyzed by sequential formation of ion pairs pulls the S4-S5 linker, bends the S6 segment, and opens the pore. Impairment of voltage-sensor function by mutations in Na+ channels contributes to several ion channelopathies, and gating pore current conducted by mutant voltage sensors in NaV1.4 channels is the primary pathophysiological mechanism in Hypokalemic Periodic Paralysis. The emerging structural model for voltage sensor function opens the way to development of a new generation of ionchannel drugs that act on voltage sensors rather than blocking the pore.
Electrical signaling in biology, from bacteria to man, depends on the rapid, highly sensitive response of voltage-gated ion channels to small changes in membrane potential. Voltage-gated ion channels derive their steep voltage dependence of activation from electrically driven movement of positively charged amino acid residues outward across the membrane in response to depolarization. In their landmark papers on voltage clamp analysis of Na+ currents in the squid giant axon, Hodgkin and Huxley predicted that activation of the Na+ conductance must involve movement of charged particles, now termed ‘gating charges’, across the membrane electrical field (Hodgkin and Huxley, 1952). The gating charge movements predicted by Hodgkin and Huxley were first directly measured for voltage-gated Na+ channels (Armstrong, 1981; Armstrong and Bezanilla, 1973; Keynes and Rojas, 1974), and present estimates indicate movement of 12–16 positive charges outward across the electric field during gating of Na+ or K+ channels (Bezanilla, 2000; Kuzmenkin et al., 2004). The essence of understanding the molecular and structural basis for voltage-dependent gating is defining the electromechanical coupling mechanism through which these gating charges move across the membrane and initiate channel activation. This review will address four main questions:
Biochemical studies using neurotoxins as molecular probes led to discovery of the voltage-gated Na+ channel protein and reconstitution of its voltage-dependent gating and ion conductance from purified protein and phospholipid components (Catterall, 1984; Hartshorne and Catterall, 1984; Hartshorne et al., 1985; Tamkun et al., 1984). Determination of the primary structure of the voltage-gated Na+ channel from Electrophorus electricus electroplax revealed a protein of more than 1800 amino acid residues in length, arranged in four repeated domains (I-IV) (Noda et al., 1984). These four domains are arrayed around a central pore, as observed in more recent cryoelectron microscopic images (Figure 1A, left; (Sato et al., 2001)). Each homologous domain was predicted to contain six alpha-helical segments (S1-S6; (Noda et al., 1984). Four of the six alpha-helical segments in each domain were proposed to be in transmembrane orientation, whereas the S3-S4 helical hairpin was proposed to project into the cytosol because of its hydrophilic character (Noda et al., 1984). The S4 segment in each domain was shown to contain 4 to 7 repeated three-residue motifs of a positive charge (usually arginine) followed by two hydrophobic amino acid residues (Noda et al., 1984). The striking concentration of positive charge in this alpha-helical segment led to the suggestion that it may be involved in voltage-dependent gating (Noda et al., 1984), but its placement in the cytosol outside the membrane electric field would not allow these positively charged residues to serve as gating charges.
The primary structure of the Na+ channel led directly to development of models for voltage sensor function and eventually to structure-function studies to test those models. Two models proposed that the S4 segments have a transmembrane orientation and that the positively charged residues within them serve as the gating charges, moving outward across the membrane in response to depolarization and thereby initiating the activation process (Catterall, 1986a, b; Guy and Seetharamulu, 1986). These proposals presaged the idea that the S1-S4 segments serve as the voltage-sensing module while the S5 and S6 segments serve as the pore-forming module and eventually led to the now-familiar six-transmembrane-segment structural model for the domains of voltage-gated Na+ channels (Figure 1A, right). Determination of the primary structure of the Shaker K+ channel from Drosophila revealed that it is analogous in structure to one domain of a Na+ channel (Tempel et al., 1987), further solidifying the concept that this six-transmembrane-segment structure is the functional unit for the voltage-gated ion channel superfamily.
How can an S4 segment containing 4 to 7 positive charges at three-residue intervals be stabilized in a transmembrane environment and move outward to translocate the gating charges across the membrane electric field? Relying on thermodynamic and structural considerations, respectively, the ‘Sliding Helix’ (Catterall, 1986a, b) and ‘Helical Screw’ (Guy and Seetharamulu, 1986) models for voltage sensor function arrived at similar solutions to this conceptual problem. The charged residues in the S4 segments were proposed to form ion pairs with negatively charged amino acid residues in the neighboring S1, S2, and/or S3 segments (Figure 1B, C). In this configuration, the positively charged residues in the S4 segment are drawn inward by the electrostatic force of the negative internal resting membrane potential. Upon depolarization this electrostatic force is relieved and the S4 segments move outward along a spiral path such that each positively charged amino acid residue in the S4 segment makes a series of ion pairs with negative charges (Figure 1C). This arrangement resolves the thermodynamic dilemma presented by placement of the gating charges in the S4 segments in a transmembrane position, and the isoenergetic exchange of ion pair partners provides a low-energy pathway for gating charge movement. This proposed model for gating charge movement, hereinafter termed the sliding-helix model for brevity, makes four testable predictions:
Structure-function studies have now rigorously tested these predictions of the sliding-helix model and provided a functional template to correlate with newly emerging structural data. To simplify discussion of structure-function data on different ion channels, the four most highly conserved arginine gating charges in the S4 segments are designated R1-R4 and two key negatively charged in the S2 segment are designated Anion 1 (An1) and Anion 2 (An2) (Figure 1B).
If the arginine and lysine residues in the S4 segments are the primary gating charges, neutralization of these charges by site-directed mutagenesis should reduce the steepness of voltage-dependent gating and alter its position on the voltage axis. Neutralization of single S4 gating charges of NaV1.2 channels and Shaker K+ channels reduced the steepness of voltage-dependent activation and substantially shifted its voltage dependence (Logothetis et al., 1992; Papazian et al., 1991; Stuhmer et al., 1989). Subsequent studies of K+ channels, in which the effects of mutation of gating charges on gating current were measured directly, provided definitive evidence that the S4 positive charges are the primary charges and also suggested an important role for a highly conserved negatively charged residue (An2) in the S2 segment (Aggarwal and MacKinnon, 1996; Seoh et al., 1996).
A crucial tenet of the sliding-helix model is that the S4 segment remains in a transmembrane position as it translocates the gating charges across the membrane electric field. The intracellular end of the S4 segment is restricted to an intracellular position because of its tight covalent connection to the N-terminal end of the S5 segment, which forms the outer circumference of the pore-forming module. Therefore, the primary question concerning the transmembrane position of the S4 segment is the location of its extracellular end.
α-Scorpion toxins are hydrophilic proteins of approximately 70 amino acid residues, which bind to a specific receptor site on the extracellular surface of Na+ channels and prevent coupling of activation to fast inactivation (Catterall, 1980). They bind with high affinity to the resting state of Na+ channels, and they are displaced from their binding site by depolarization to membrane potentials that cause channel activation (Catterall, 1979). Site-directed mutagenesis studies show that these toxins bind to the S3-S4 linker in domain IV of Na+ channels (Figure 2B; (Rogers et al., 1996)) and reduce the gating charge movement of this voltage sensor (Sheets et al., 1999). Therefore, this site must be available on the extracellular side of the membrane in the resting state. β-scorpion toxins also bind to the Na+ channel in the resting state, through a receptor site including the S3-S4 linker in domain II, and they enhance activation by trapping the voltage sensor in domain II in the activated state (Cestèle et al., 1998; Cestele et al., 2006). Protoxin II binds to the S3-S4 linker in domain II of Na+ channels in the resting state and opposes activation and gating charge movement (Schmalhofer et al., 2008; Sokolov et al., 2008a). ω-Agatoxin IVA binds to the S3-S4 linker in domain IV of CaV2.1 channels in the resting state and opposes activation of the channel (Bourinet et al., 1999; Winterfield and Swartz, 2000). Hanatoxin and related cysteine-knot toxins bind to the S3-S4 linker of K+ channels in the resting state and oppose activation (Lee et al., 2003; Li-Smerin et al., 2000; Phillips et al., 2005; Swartz and MacKinnon, 1997). These results with a range of toxins and channels strengthen the conclusion that the S3-S4 linker and the N-terminal end of S4 segment are located on the extracellular side of the membrane in the resting state of the voltage sensor.
The extracellular location of the S3-S4 linker in the resting state is also supported by other experimental approaches. Insertion of antibody epitopes in the intracellular and extracellular loops of the Shaker K+ channel followed by immunocytochemical localization of the labeled channels expressed in Xenopus oocytes showed that the S3-S4 linker is extracellular in the resting state (Shih and Goldin, 1997). In addition, covalent labeling studies discussed in the next section (Larsson et al., 1996; Yang et al., 1996; Yusaf et al., 1996) also support a transmembrane localization of the S4 segment in both resting and activated states.
The conclusion that neurotoxins bind to the S3-S4 linker at the extracellular surface of the Na+, Ca2+, and K+ channels in the resting state has been challenged, based on the idea that the toxins may insert into the lipid bilayer and approach their binding site from the hydrophobic phase of the membrane (Lee and MacKinnon, 2004). Their experiments show that the cysteine-knot toxin VsTx-1 from tarantula binds to phospholipid vesicles in vitro, at least in buffers with 150 mM K+ and no Na+ or divalent cations (Lee and MacKinnon, 2004). This mechanism seems unlikely for scorpion toxins because of their size and hydrophilicity, and evidence against lipid partitioning by purified scorpion toxins has been presented (Cohen et al., 2006). Moreover, in reconstitution of purified Na+ channels, it was found that phospholipid vesicles did not bind α-scorpion toxin with high affinity, whereas addition of the purified Na+ channel to the vesicles effectively reconstituted high-affinity, voltage-dependent toxin binding (Tamkun et al., 1984). These results strongly support specific binding of α-scorpion toxins to their receptor site on the Na+ channel protein with no evidence of high-affinity binding to phospholipids. On the other hand, hanatoxin and related tarantula toxins have a hydrophobic surface that partitions part way into the lipid phase and may allow the toxin to approach its binding site by diffusion in the lipid bilayer (Milescu et al., 2007). This characteristic may contribute to toxin-binding affinity by increasing the local concentration of toxin at the surface of the membrane, but it does not account for the high-affinity component of toxin binding and action when physiological salt concentrations are present (Milescu et al., 2007).
The motion of the S4 segment has been probed by chemical labeling studies and by real-time measurements of fluorescence quenching and energy transfer. In the first experiments on movement of the S4 segment, a cysteine residue was substituted for the outermost gating-charge-carrying arginine residue (R1) in the S4 segment of domain IV of the skeletal muscle NaV1.4 channel (Yang and Horn, 1995). At the resting membrane potential, perfusion of cysteine-reactive methanethiosulfonate (MTS) reagents outside of the cell had no effect on the Na+ current. In contrast, depolarization of the cell allowed reaction of the substituted cysteine with hydrophilic MTS reagents, which was revealed as an increase in the rate of inactivation (Yang and Horn, 1995). Further studies using this ‘substituted cysteine accessibility method’ (Karlin and Akabas, 1998) on Na+ channels, showed that the R2 and R3 gating charges were accessible to modification by hydrophilic MTS reagents from the inside of the cell in the resting state but became accessible to modification by the same MTS reagents from the outside of the cell after depolarization (Yang et al., 1996). The substituted cysteine accessibility method was also applied to movement of voltage sensors of voltage-gated K+ channels (Larsson et al., 1996; Yusaf et al., 1996). As for Na+ channels, these studies showed that arginine gating charges move from being accessible on the intracellular side of the membrane to accessible on the extracellular side of the membrane during activation, and only a short section of S4 is occluded from reaction at any one time. Together, these results provided clear evidence for outward movement of the S4 segment across the membrane permeability barrier during activation of the channel and led to the concept that only a narrow waist of the S4 segment is protected from reaction with hydrophilic MTS reagents at one time. These results strongly suggested that there are hydrophilic vestibules at either end of the voltage sensor, which accommodate the outer and inner gating charges, while the membrane electric field primarily drops across a short central ‘gating pore’ through which the S4 segment translocates. A revised version of the sliding-helix model accommodates these results (Figure 1D; (Catterall, 2000; Yang et al., 1997)).
The position of the S3 segment relative to the S4 segment has also been probed by the substituted cysteine accessibility method (Nguyen and Horn, 2002). Substitution of amino acid residues sequentially from the outer end of the S3 segment in domain IV of the NaV1.4 channel with cysteine, followed by reaction with hydrophilic MTS reagents, showed that the outermost four amino acid residues in S3 were accessible from outside the cell in both resting and activated states (Nguyen and Horn, 2002). The S3 segment was also shown to move little, if at all, relative to the S5 segment during activation of Shaker K+ channels, as assessed by disulfide crosslinking of substituted cysteine residues (Gandhi et al., 2003). These results support the concept that the S3 segment does not move substantially outward or inward during activation of the voltage sensor and imply that the S4 segment must move with respect to the S3 segment during activation.
The exposure of the S4 segments to the hydrophilic environment of the extracellular solution and the rotation of the S4 segment have been recorded in real time by fluorescent labeling experiments (Figure 2C; (Cha and Bezanilla, 1997; Mannuzzu et al., 1996)). These fluorescent labeling studies showed that several amino acid residues within the S4 segment move outward into a hydrophilic environment during activation of the voltage sensors of Shaker K+ channels on the millisecond time scale of channel activation. More detailed studies of channels labeled at multiple positions along the S4 helix provided evidence for both outward translocation and rotation of the S4 segment up to 180° during activation (Cha et al., 1999; Glauner et al., 1999). These results supported both outward translocation and substantial rotation of the S4 segment on the millisecond time scale of activation, as predicted in the sliding-helix model, but they did not define the distance or mechanism of S4 movement.
The energetic cost of placing the four positive gating charges of the S4 segment in a transmembrane position is enormous, and outward movement of these gating charges through a hydrophobic environment would be energetically prohibitive. Therefore, a key question is how the gating charges are stabilized in their transmembrane position and how their outward movement is catalyzed. The sliding-helix model of gating addresses these requirements through ion pair formation. The gating charges are proposed to be stabilized in the transmembrane environment by ion pair formation, and their outward movement is proposed to be catalyzed by sequential, isoenergetic exchange of ion pair partners. The first evidence in favor of ion pair formation came from studies of charge reversal mutations in the Shaker K+ channel (Papazian et al., 1995; Tiwari-Woodruff et al., 2000). Neutralization of the S4 gating charges prevented functional expression of the mutant channels, apparently due to misfolding, retention in the endoplasmic reticulum, and degradation. However, paired mutations in which gating charges were converted to negative charges and An1 or An2 in the S2 segment were converted to positive charges rescued functional expression. Paired charge reversal mutations of An1 with R3 impaired activation while pairing with R4 enhanced activation, consistent with outward movement of the S4 segment relative to An1 during activation. These results show that paired charge reversal mutations are required for stable folding and expression of K+ channels and support the conclusion that the gating charges must be paired with a negative charge to stabilize their transmembrane position.
Recent work from a different experimental approach demonstrates outward helical movement of the S4 transmembrane segment and sequential formation of ion pairs on the millisecond time scale of channel activation. Disulfide bond formation between substituted cysteine residues requires their sulfur atoms to approach within 2 Å, providing a high-resolution method of analysis of intramolecular interactions. Experiments using this method demonstrated sequential formation of ion pairs between the R3 and R4 gating charges and both An1 and An2 in the S2 segment during activation of the bacterial Na+ channel NaChBac (DeCaen et al., 2009; DeCaen et al., 2008). Analysis of the time course of disulfide locking showed that disulfide bond formation occurs on the millisecond time scale at nearly the rate of channel activation. R4 interacts first with An2 and then with An1 during activation of the channel, and R3 interacts with An1 at essentially the same time and voltage as R4 interacts with An2. These results require an outward helical movement of the S4 segment to place the substituted cysteine residues in position to form ion pairs sequentially with An2 and then An1 during activation.
In contrast to these results with the R3 and R4 gating charges, paired cysteine mutations of the R1 and R2 gating charges form disulfide bonds with cysteine residues substituted for An1, but the resulting channels are recognized as misfolded and degraded (DeCaen et al., Biophys Soc Abst., 2010). In contrast, disulfide bonding of a cysteine substituted for a hydrophobic residue adjacent to R1 with An1 locks the channel in a resting state, from which it can be released by disulfide-reducing agents (DeCaen et al., Biophys. Abst. 2010). These results provide further support for formation of an ion pair between R1 and An1 in a resting state. Evidently, the S4 segment moves outward during activation from an inward position in which R1 interacts with An1 to an outward position in which R4 interacts with An1, and the R1-R4 gating charges interact sequentially with An2 and then An1 to form ion pairs as S4 moves.
In the sliding-helix model, outward movement of gating charges is mediated by sequential formation of ion pairs, which provides a low energy pathway through the voltage sensor. In this sense, the voltage sensor acts as a catalyst whose substrates are the arginine gating charges. Remarkably, mutations of the R1 and R2 gating charges to smaller, uncharged residues, thereby removing the substrate of the voltage sensor catalytic site, induce a leak current through the mutant voltage sensor, termed omega current or gating pore current. In the Shaker K+ channel, mutations of gating charges to histidines caused a proton leak current through the voltage sensor (Starace and Bezanilla, 2004), and substitution of the R1 gating charge with smaller hydrophilic residues caused a nonselective cationic gating pore current in the resting state (Tombola et al., 2007; Tombola et al., 2005). In the brain NaV1.2 channel, mutation of both R1 and R2 gating charges to glutamine was required to observe substantial gating pore current in the resting state (Figure 3A, red; (Sokolov et al., 2005)). In both Na+ and K+ channels, depolarization of the membrane to cause activation of the voltage sensor blocked the gating pore current caused by mutations of R1 and R2, consistent with the model that outward movement of the S4 segment moves the defective R1 and R2 gating charges out of the voltage sensor and plugs the leak.
For NaV1.2 channels, conversion of the R2 and R3 gating charges to glutamine induces gating pore current in the activated state, which appears upon depolarization and activation of the channel and is blocked by repolarization to return the channel to the resting state (Figure 3A, blue; (Sokolov et al., 2005)). These results indicate that outward movement of the defective R2 and R3 gating charges into the gating pore upon activation causes an ionic leak through the voltage sensor, which is blocked by repolarization to return them to their resting position. Similarly, Gamal El-Din et al. found that paired mutations of two arginine gating charges in Shaker K+ channels to smaller residues are generally required to induce gating pore current (Gamal El-Din et al., 2009). The previously observed gating pore current measured for mutations of the R1 gating charge alone (Tombola et al., 2005) depends on the location of a small, uncharged amino acid residue (alanine) in the −3 position (R0) in the Shaker amino acid sequence (Gamal El-Din et al., 2009). Substitution of paired small residues at positions R0-R1, R1-R2, and R2-R3 all induce gating pore current, and progressively stronger depolarizations are required to generate gating pore current from the more inward paired mutants (Gamal El-Din et al., 2009). Altogether, these studies of gating pore currents argue persuasively that depolarization forces the S4 segment to move outward along a helical pathway, sequentially placing paired mutant gating charges in the gating pore and generating gating pore current charges. Only an outward helical motion of the S4 segment through the gating pore can easily accommodate these results (Gamal El-Din et al., 2009; Sokolov et al., 2005).
Analysis of ion channel structure and function has been revolutionized by the availability of high-resolution structural information from x-ray crystallography and by homology and ab-initio molecular modeling based on those structures.
The first high-resolution structure of a voltage-gated ion channel was the small bacterial K+ channel KVAP (Jiang et al., 2003a). This structure was surprising in several respects. The state of the channel depicted in the crystal structure had the voltage-sensing module lying on its side, nearly parallel to the surface of a hypothetical lipid bilayer. This position of the S3-S4 helical hairpin, lying on its side near the inner surface of the membrane, implied that the voltage sensor was in the resting state, yet the pore-forming module was in the open state. The position of the S3-S4 helical hairpin suggested a novel gating mechanism in which the S3-S4 helical hairpin would move through the phospholipid bilayer with a ‘paddle’ motion to translocate gating charges across the membrane, reaching a transmembrane position only in the activated state (Jiang et al., 2003b). This paddle model was supported by a structure of the separate voltage-sensing module without the pore-forming module attached (Jiang et al., 2003a; Jiang et al., 2003b), in which the innermost gating charges formed ion pairs with the An1 in the S2 segment, but this structure was also consistent with the proposed activated state in the sliding-helix gating model.
Although the paddle model of voltage sensor function fit the KVAP structure, it seemed incompatible with other experimental evidence. First, the voltage sensor was in the resting state while the pore was open, suggesting a non-native state of the channel in the crystal structure. Second, the position of the S3-S4 helical hairpin was inconsistent with strong evidence that the S3-S4 linkers form the receptor sites for scorpion toxins on Na+ channels and hanatoxin on K+ channels, which are available for toxin binding at the extracellular surface in the resting state of these channels. Moreover, the paddle mechanism also seemed inconsistent with chemical labeling studies showing that at least two of the S4 gating charges are protected from labeling from the intracellular side of the membrane in the resting state of Na+ and K+ channels (Larsson et al., 1996; Yang et al., 1996). Subsequent structural work has led to the probable conclusion that the KVAP structure did indeed represent a non-native state of the channel and its voltage sensor (Lee et al., 2005).
In a landmark advance for this field, Long et al. succeeded in determining the high-resolution structure of the KV1.2 channel by x-ray crystallography (Long et al., 2005a, b) and followed up that work with a structure of a chimera of the KV1.2/KV2.1 channels at even higher resolution (Long et al., 2007). In these structures, the S1-S6 segments are all in a transmembrane position, the S6 segments are bent, the pore is open, and the S4 segment is located in a transmembrane position with its innermost gating charges in a gating pore in the center of the voltage sensor (Figure 4A), as expected for an activated state with an open pore. An unexpected feature of the structure of the KV1.2 channel was separation of the voltage-sensing and pore-forming modules of the individual subunits (Long et al., 2005a). Each voltage-sensing module interacts most closely with the pore-forming module of the adjacent subunit in the clockwise direction as viewed from the extracellular side (Figure 4A). This surprising structure implies that voltage-dependent conformational changes in the voltage-sensing module may be communicated to the pore-forming module of the neighboring subunit, which could provide the molecular basis for concerted opening of the pore to a single fullconductance state by a linked conformational change in all four subunits, as predicted by detailed gating models (Zagotta et al., 1994).
An essential role for phospholipids in the structure and function of voltage-gated Na+ channels was suggested by early biochemical experiments in which specific phospholipid combinations were required for reconstitution of ion conductance, voltage-dependent toxin binding, and voltage-dependent gating of purified Na+ channels (Hartshorne et al., 1985; Tamkun et al., 1984). Consistent with this requirement for specific lipids, the high-resolution structure of a KV1.2/2.1 chimera revealed intrinsically bound phospholipid molecules (Long et al., 2007). These bound phospholipids are seen around the waist of the protein where it contacts the phospholipid bilayer and also in the internal and external vestibules in the voltage sensor (Long et al., 2007). The phospholipid head groups are in position to serve as ion pair partners of gating charges at the intracellular and extracellular surfaces of the membrane, but not within the core of the voltage-sensing module. Evidence that phospholipid head groups can affect voltage-dependent gating supports an important role for these bound lipids in voltage sensor function (Cohen et al., 2006; Schmidt et al., 2006).
Only the structure of the activated voltage sensor has been determined to date (Long et al., 2005, 2007), although these activated voltage-sensor structures may have been captured in the context of the open state or the inactivated state of the channel as a whole. It is not surprising that only the activated state of the voltage sensor has been visualized in these structures, because the activated state is stable in the absence of a membrane potential, as in a protein crystal, whereas generation of the resting state requires a negative internal membrane potential in the range of −80 mV. Consequently, structural modeling has been used to develop high-resolution models for the resting state. The Rosetta Membrane ab initio structural modeling program, which successfully predicts the structures of many complex membrane proteins (Yarov-Yarovoy et al., 2006b), was used to predict the resting state structure of KV1.2 (Yarov-Yarovoy et al., 2006a). The pore was constrained to be in the closed conformation, based on the structure of the bacterial K+ channel KcsA (Doyle et al., 1998), and the position of the S4 segment was constrained by the well-established interaction between the R1 gating charge and An1 in the S2 segment, which is required for folding and functional expression (Papazian et al., 1995). The resulting molecular model shows the S4 segment in a transmembrane position, with the R1 gating charge forming an ion pair with An1 (Figure 4B; (Yarov-Yarovoy et al., 2006a)). Additional molecular modeling revealed a second resting state structure with the S4 segment drawn even farther toward the intracellular side of the membrane (Pathak et al., 2007). Similar structural models of the resting states of NaChBac and the plant KAT channel have been derived from homology modeling methods (Shafrir et al., 2008a, b). The structure of the activated state of KV1.2 plus these models of resting states of KV1.2 and related channels provide the starting point for structural models of the voltage-sensing mechanism. Movies depicting the gating transition from resting to activated states are available as Supporting Information for Yarov-Yarovoy (2006a).
In the sliding-helix model, outward movement of the gating charges in the S4 segment proceeds as a stepwise exchange of ion pair partners, and sequential formation of ion pairs during activation of NaChBac has been directly demonstrated in disulfide-locking experiments (DeCaen et al., 2009; DeCaen et al., 2008). Therefore, molecular models for each step of the outward movement of the gating charges can be developed using formation of ion pairs as a structural constraint ((DeCaen et al., 2009); Yarov-Yarovoy et al., Biophys. Soc. Abst., 2010). These structural models show the stepwise movement of the S4 segment outward through the gating pore with sequential formation of ion pairs at each step (Figure 5A).
The results of disulfide-locking experiments indicate that the R3 gating charge interacts with An1 with the same kinetics and voltage dependence as R4 interacts with An2, suggesting that these two ion pair interactions occur simultaneously in the gating pore (Figure 2D; (DeCaen et al., 2009)). The structure of the activated voltage sensor of the KV1.2/KV2.1 chimera contains a short stretch of the S4 segment in the gating pore in the 310 helical conformation (Long et al., 2007), and both molecular dynamics simulations and EPR studies of cysteine-substituted mutants suggest that the S4 segment forms 310 helix in the resting state as well (Chakrapani et al., 2010; Khalili-Araghi et al., 2010). Placement of this short section of 310 helical conformation in the gating pore allows the dual interaction of R3/An1 and R4/An2 by stretching the two gating charges farther apart and placing them on exactly the same side of the S4 helix to facilitate interaction with An1 and A2 (Figure 5B). Each of the pairs of gating charges (R1-R2, R2-R3, and R3-R4) is hypothesized to adopt this stretched 310 helical conformation transiently as the S4 segment moves outward through the gating pore, allowing each gating charge to “stretch out” to make its next ion pair interaction. In the inner and outer vestibules on either side of the gating pore, the S4 segment retains its alpha-helical conformation and the two ends of the S4 segment rotate as it moves outward. Therefore, in this structural version of the sliding-helix model, the outward movement is not entirely screw-like; instead, the two ends of the helix move in a screwlike manner while the center section, about 9 Å in length, unwinds to a 310 helix and moves outward linearly. This form of outward movement diminishes the total rotation of the S4 segment required for gating charge movement because a section of the S4 segment remains in the 310 conformation within the gating pore in the fully activated state and never adopts the alpha-helical conformation on the extracellular side of the gating pore. Importantly, this processive, transient formation of 310 helix also allows productive ion pair interaction of two gating charges in the gating pore simultaneously while the intervening hydrophobic residues point away from An1 and An2 (Figure 5B). Although a 310 helix is not as stable as an alpha helix, the extent of 310 helix remains the same in all states in this model so there is no net energetic cost of breaking alpha helix and making 310 helix. This stretching of the S4 segment serves an analogous purpose to stretching of covalent bonds in an enzyme catalytic site— provision of a low energy pathway between the initial state and the final state. Thus, the catalytic site of the voltage sensor (ie., the gating pore) neutralizes the gating charges and unwinds and stretches the S4 segment as it moves outward, providing an isoenergetic, hydrophilic pathway for gating charge translocation across the focused membrane electric field.
Recent structure-function studies of Shaker K+ channels provide additional insight into the mechanism of S4 movement by the voltage sensor (Tao et al., 2010). The S2 segment of nearly all voltage sensors has a conserved phenylalanine residue positioned between the two negative charges that form ion pairs with gating charges on the same face of the alpha helix (underlined in Figure 1B; (Tao et al., 2010)). Mutation of this phenylalanine to other natural and unnatural amino acid residues shows that a rigid aromatic or an aliphatic cyclohexane ring in this position is essential for normal voltage sensor function in Shaker K+ channels (Tao et al., 2010). Other substitutions shift the voltage dependence of activation to more positive voltages, out of the physiological range. Because the cyclohexane ring is not well-suited for direct interaction with gating charges, these results suggest that the native phenylalanine residue serves a structural role as part of the seal to assure smooth S4 movement through the gating pore without ionic leakage.
How long is the narrow section of the gating pore that protects the S4 gating charges from their hydrophobic environment? Chemical labeling studies showed that only two gating charges are protected from the aqueous environment simultaneously, as assessed by reaction of substituted cysteine residues with MTS reagents (Glauner et al., 1999; Yang et al., 1996). These results led to the proposal that the S4 segment moves through a short gating pore that connects external and internal hydrophilic vestibules (Figure 1D; (Yang et al., 1997)). This proposal derives further support from measurements of a focused electrical field in the center of the gating pore, in which the membrane electric field drops from near maximum to near zero over a distance of 4–10 Å (Ahern and Horn, 2005; Campos et al., 2007; Starace and Bezanilla, 2004). Thus, these structure-function studies limit the length of the functional gating pore to 2–3 turns of an alpha helix.
The idea of a focused field in a short section of the gating pore fits closely with the requirement for paired substitutions of two gating charges by small amino acid residues in order to induce large gating pore currents (Gamal El-Din et al., 2010; Sokolov et al., 2005). In the 310 conformation, the side chains of two small hydrophilic residues substituted for two gating charges would face the lumen of the gating pore across from the An1 and An2 residues of the S2 segment, while the two intervening hydrophobic residues would face away from the lumen of the gating pore (Figure 5C). In this conformation, the hydrophilic side chains of the amino acid residues substituted for the mutated gating charges and the side chains of the An1 and An2 residues would form an artificial (that is, unnatural) ion selectivity filter that would mediate gating pore current only in the state of the mutant channels that places the two mutant gating charges in the gating pore. The placement of two gating charges within the gating pore may catalyze the outward movement of the S4 segment by stabilizing two gating charges simultaneously through ion pair interactions.
The sliding-helix model requires a substantial vertical movement of the S4 segment through the gating pore in order to translocate 3–4 gating charges. In the structural version of the sliding-helix model presented above, the gating pore is defined by An1 and An2 in the S2 segment and therefore is 10.5 Å in length. Because 3–4 gating charges must move all the way through the electric field to account for the measured gating charge movement of 12–16, the outward movement of the S4 segment must begin with R1 in the gating pore and continue at least until R4 reaches the gating pore. In this case, 9 amino acid residues in alpha helical conformation would move outward past An1 in the S2 segment, requiring an outward movement of the R1 gating charge of at least 13.5 Å with respect to this position in S2. Because the S4 segment moves outward at an oblique angle with respect to the membrane surface (Figure 4), the distance of movement perpendicular to the membrane surface would be less.
Studies of the purified KVAP channel in phospholipid bilayers with biotin-avidin labeling methods provided evidence for a substantial outward movement of the S4 segment of the bacterial K+ channel KVAP, in the range of 15 to 20 Å relative to the surface of the lipid bilayer (Ruta et al., 2005). Thus, this direct chemical-labeling evidence supports a large movement of the S4 segment relative to the membrane surface. Two complementary disulfide-locking studies also argue for a large outward movement of the S4 segment during activation. Disulfide-locking of cysteine residues substituted for the S4 gating charges with cysteines substituted at different positions along the S3 segment, showed that the S4 segment moves 12 Å with respect to the S3 segment (Broomand and Elinder, 2008). Disulfide locking cysteine residues in the S4 segment to cysteines substituted for An1 and An2 in the S2 segment showed that R1 is disulfide-locked to An1 in a resting state, whereas R4 is disulfide locked to An1 in an activated state. This requires outward movement of nine residues with respect to the position of An1, or a movement of 13.5 Å for the S4 segment ((DeCaen et al., 2009); DeCaen et al., Biophys. Soc. Abst., 2010). These studies are consistent with the conclusions of chemical labeling experiments and with the expectations of the structural version of the sliding-helix model for S4 movement through a short gating pore of approximately 10.5 Å in length (Figure 5).
The structure of the KV1.2 channel shows limited contacts between the pore-forming and voltage-sensing modules (Long et al., 2005a), consistent with the idea that these modules function in a largely independent manner. The presence of voltage sensors without pore-forming modules in the voltage-sensitive phosphatase Ci-VSP (Murata et al., 2005) and in HV1 channels (Ramsey et al., 2006; Sasaki et al., 2006) supports this view. The most definite contact between the voltage-sensing and pore-forming modules is the S4-S5 linker, which covalently connects them. Both site-directed mutagenesis studies and structural studies support the idea that the S4-S5 linker communicates the conformational change in the voltage sensor to the pore-forming domain through a pulling force on the S5 segment (Long et al., 2005b). However, a single point of contact is not sufficient for effective pulling. Just as a person engaged in a tug-of-war must have a firm footing on the ground, there must be another point of leverage for the voltage sensor on the pore. Disulfide crosslinking studies show that new interactions form between the extracellular end of the S4 segment and the extracellular end of the S5 segment upon activation (Broomand et al., 2003; Gandhi et al., 2003). Moreover, recent structural studies point to a highly conserved interaction between an extracellular site in the S1 segment of the voltage-sensing module, and the S5 segment as a second point of contact (Lee et al., 2009). It is likely that these points of contact act as a fulcrum for the voltage sensor to pull on the S4-S5 linker and open the pore.
Multiple lines of evidence point to bending the S6 segment and opening of the helical bundle at its intracellular end as the mechanism of pore opening for the voltage-gated ion channels. In structural studies of bacterial K+ channels having two transmembrane segments (TM1 and TM2), the closed state has the S6 segments in a straight conformation that leads to their crossing in a bundle to close the pore at its intracellular end (Doyle et al., 1998), whereas the open state has a bend in the S6 segments to open the intracellular mouth of the pore (Jiang et al., 2002a, b). This bend occurs at a highly conserved glycine residue, which also serves as a glycine hinge and enhances the opening of NaChBac channels (Zhao et al., 2004). On the other hand, a proline-valine-proline motif in a more inward position in the S6 segment appears to serve this function in KV channels (Webster et al., 2004). Nevertheless, in both cases, it seems that pulling on the S4-S5 linker leads to bending of the S6 segment and opening of the pore. Together with the emerging view of the mechanism of action of the voltage sensor, these studies define the complete voltage-dependent gating process required for activation of a ion channel—electrically driven outward movement and rotation of the S4 segment, catalyzed by a uniquely designed gating pore, pulling on the S4-S5 linker, and bending the S6 segments to open the bundle crossing at their intracellular ends.
Voltage-gated ion channels are the targets for a large number of mutations that cause inherited diseases. Mutations in multiple Na+ channel genes expressed in different tissues cause distinct inherited channelopathies. Almost all of these neurological diseases have dominant inheritance and are caused by gain-of-function mutations. These gain of function effects arise primarily from enhanced activation or impaired inactivation of ion channels, which can cause different diseases through mutations of the same gene. Mutations that impair fast and/or slow inactivation of NaV1.4 channels cause paramyotonia congenita or hyperkalemic periodic paralysis (Jurkat-Rott and Lehmann-Horn, 2006; Venance et al., 2006). Mutations that enhance activation of NaV1.7 channels cause inherited erythromelalgia, characterized by episodes of burning pain, erythema and mild swelling in the hands and feet triggered by mild warmth or exercise; whereas mutations that impair fast inactivation cause paroxysmal extreme pain disorder, characterized by intense rectal pain (Dib-Hajj et al., 2007). Mutations that enhance activation of CaV2.1 channels cause familial hemiplegic migraine by increasing the efficiency of synaptic transmission at excitatory synapses in the brain (Pietrobon, 2007). Although these disease mutations often have specific effects on activation or inactivation, they are spread throughout the ion channel protein in both voltage-sensing and pore-forming modules and in the inactivation gate, reflecting the functional coupling of voltage sensor function to pore-opening and inactivation. Therefore, no single molecular mechanism has been proposed that can explain the defect in ion channel function in any of these diseases. Nevertheless, although well-defined molecular mechanisms for the effects of these disease mutations have not been defined yet, all of these ion channelopathies result from failure of ion channel voltage-sensing and its coupling to activation and/or inactivation gating.
Hypokalemic periodic paralysis presents a different picture for its molecular mechanism. In contrast to the prominent effects of mutations that cause paramyotonia congenita and hyperkalemic periodic paralysis on inactivation of NaV1.4 channels, the mutations in NaV1.4 channels that cause hypokalemic periodic paralysis (HypoPP) do not have large or consistent effects on channel function as assessed in standard voltage clamp experiments (Jurkat-Rott and Lehmann-Horn, 2006; Venance et al., 2006). Nevertheless, these HypoPP mutations act in a dominant manner to cause episodic flaccid paralysis associated with low serum K+ levels and progressive cytopathological changes in the skeletal muscles of affected individuals who have only a single mutant allele (Jurkat-Rott and Lehmann-Horn, 2006; Venance et al., 2006). The combination of dominant periodic paralysis and cytopathology with lack of evident functional effects led to a search for unconventional disease mechanisms.
Remarkably, all of the originally described mutations in HypoPP neutralize the R1 or R2 gating charges in domain II of NaV1.4 by substitution of Gly, Cys, Ser, or His (Venance et al., 2006). These mutations correspond precisely in location to the paired mutations of the R1 and R2 gating charges in brain NaV1.2 channels that cause large gating pore currents (approximately 9% of peak Na+ current; (Sokolov et al., 2005)). Following up this lead, measurements of leak currents of these single-mutant NaV1.4 channels expressed in the cut-open Xenopus oocyte preparation revealed a small gating pore current, approximately 1% of peak Na+ current for the mutant R2G (Figure 3B; (Sokolov et al., 2007)). Like the corresponding mutations in NaV1.2 channels (Sokolov et al., 2005), mutations of the R1 and R2 gating charges in NaV1.4 conducted gating pore current in the resting state, which was blocked by depolarizations that activate the voltage sensor (Figure 3B; (Sokolov et al., 2007; Struyk and Cannon, 2007)). Evidently, gating pore current is conducted only when the defective R1 or R2 gating charge is located in the gating pore in the resting state and is blocked when the normal R3 and/or R4 gating charges replace the defective gating charge in the gating pore in the activated state.
The gating pore current conducted by the R2G mutant is weakly selective among inorganic monovalent cations (Cs+~K+>Na+~Li+) but conducts large organic cations such as tetramethylammonium or N-methyl-D-glucamine very poorly, if at all (Sokolov et al., 2007). In contrast, the R2H mutant is selective for protons, presumably because of the specific interaction of histidine with protons (Struyk and Cannon, 2007), as previously observed for Shaker K+ channels with substitutions of histidine for gating charges (Starace and Bezanilla, 2004). Surprisingly, both R2G and R2H mutants conduct much larger (>10-fold) gating pore currents when guanidine is the current carrier than when Na+ is the permeant ion (Sokolov et al., 2010). This selective permeability of guanidine suggests that it passes precisely through the position of the missing guanidine moiety from the wild-type arginine side chain in the gating pore. In contrast to the nonselective conductance of monovalent cations, the R2G gating pore is blocked by a range of divalent and trivalent cations in the concentration range from 100 µM up to 10 mM ((Sokolov et al., 2007); Sokolov et al., 2010). Ca2+ is not permeant through the gating pore and is a weak blocker in the mM concentration range (Sokolov et al., 2007).
Recent human genetic studies have uncovered additional mutations that cause periodic paralysis with some characteristics of HypoPP. One particularly interesting example is provided by studies of families with mutations in the R3 gating charge (Vicart et al., 2004). The affected individuals have Normokalemic Periodic Paralysis (NormoPP), which resembles HypoPP but paralysis occurs at normal levels of serum K+ (Fontaine, 2008; Vicart et al., 2004). As expected from previous structure-function studies of NaV1.2 channels (Sokolov et al., 2005), these NormoPP mutations in the R3 gating charge cause gating pore current in the activated state, which is lost on repolarization to return NaV1.4 channels to the resting state (Figure 3B; (Sokolov et al., 2008a)). These results fit closely with the idea that gating pore current is generated when the defective R3 gating charge in present in the gating pore in the activated state but is blocked when that defective gating charge moves out of the gating pore in the resting state.
Because gating pore currents in HypoPP and NormoPP are in the range of 1% or less of the peak Na+ current conducted through the central pore of NaV1.4 channels, it is important to consider how this small leak current could cause pathophysiology. The key point is that the leak current is conducted continuously, whereas the peak Na+ current only occurs for a millisecond during an action potential. In the case of the R2G mutant (Sokolov et al., 2007), calculations suggest that the rate of entry of Na+ at the resting membrane potential would be increased as much as 20-fold by the gating pore current. Such a large increase in Na+ entry would be sufficient to cause depolarization and Na+ overload, dominant impairment of action potential generation by the wild-type NaV1.4 channels encoded by the unaffected allele, and cytopathology owing to osmotic and ionic imbalance and the increased energetic cost of pumping Na+ out of the cell, as observed in patient muscle fibers (Jurkat-Rott and Lehmann-Horn, 2006; Jurkat-Rott et al., 2009). In the case of R2H and other histidine mutants, it is likely that a similar pathophysiological situation arises indirectly, as protons leak into the cell, activate Na+/H+ exchange, and bring excess Na+ into the cell (Jurkat-Rott et al., 2009; Struyk and Cannon, 2007). These ionic changes may also create bistable eletrophysiological conditions and thereby impair normal action potential generation (Jurkat-Rott et al., 2009; Struyk and Cannon, 2008). Finally, in the case of NormoPP mutations, it is likely that the slow-inactivated state plays an important role in pathophysiology (Sokolov et al., 2008b). Conductance of gating pore current in only the activated state would be unlikely to cause pathophysiology because of its short duration. However, the R3 gating charge mutants that cause NormoPP also conduct gating pore current in the slow-inactivated state (Sokolov et al., 2008b). Since slow inactivation occurs progressively during the long, high-frequency trains of action potentials that drive forceful contractions and reverses only very slowly (Ruff, 2008), conductance of gating pore current in the slow inactivated state would be sufficient to cause disease pathology by Na+ overload, as for R1 and R2 gating charge mutants.
HypoPP is also caused by mutations in the skeletal muscle CaV1.1 channel (Venance et al., 2006), which is responsible for excitation-contraction coupling. Here again, only mutations in the R1 or R2 gating charges have been found to cause the disease (Venance et al., 2006). Therefore, although difficulties of high-level expression of CaV1.1 channels have so far thwarted attempts to detect gating pore current, it is very likely that this alternative form of HypoPP is caused by gating pore current through the mutant voltage sensors of CaV1.1 channels.
There are no other ion channelopathies described to date in which all of the mutations are in gating charges in voltage sensors, so there may be no other diseases in which the primary pathophysiology is caused by gating pore current. However, a survey of mutations in the OMIM database revealed several mutations in a variety of channelopathies (including other periodic paralyses, inherited migraine, and inherited chronic pain) in which the R1, R3, or R3 gating charges are converted to small hydrophilic residues (Sokolov et al., 2007). If these mutants do indeed conduct gating pore current, it may lead to allele-specific aspects of the pathophysiology of these diseases that have not yet been accounted for by other pathophysiologic mechanisms. Thus, the role of gating pore current in ion channelopathies may extend beyond the HypoPP and NormoPP.
Research on the structure and function of voltage sensors has important implications for pathophysiology, drug discovery, and translational research. As outlined above, the role of gating pore current as a pathophysiological mechanism in HypoPP and NormoPP followed directly from studies of voltage sensor structure and function. Understanding voltage sensor structure and function will also open novel opportunities in ion channel pharmacology. Ion channel blocking drugs used most frequently in current therapy (local anesthetics, antiarrhythmics, antiepileptics) nearly all bind in pore-forming modules and are not highly selective among multiple related members of their ion channel family. These drugs derive their therapeutic usefulness from their characteristic frequency- and voltage-dependent block, arising from selective binding to activated and inactivated channels according to the modulated receptor model (Hille, 1977). This mode of binding endows them with a degree of specificity for blocking ion channels in depolarized and rapidly firing cells, which are frequently the source of pain signals and pathophysiology. However, increased specificity based on the amino acid sequence differences among ion channels would enhance the utility and safety of these drugs.
A major reason why ion channel blocking drugs are not selective is that they target the pore-forming modules that are highly conserved among channels with the same ion selectivity. In contrast, amino acid sequence variation among voltage-sensing modules is greater, and many natural toxins bind to voltage sensors of individual voltage sensors in a highly specific manner, as exemplified by the scorpion toxins and tarantula toxins that have been used as probes in the structure-function studies of voltage sensors in experiments discussed above. These toxins have highly state-dependent binding, which could be captured to great advantage in small molecules targeted to voltage sensors. Such drug candidates could potentially retain frequency- and voltage-dependent binding as a mechanism of drug specificity, along with added selectivity from recognition of unique amino acid sequences in their voltage sensor targets. The emerging knowledge of voltage sensor structure and function reviewed here will help to define the molecular basis for rational drug discovery efforts aimed at voltage sensors as drug targets.
The discovery that diseases are caused by voltage-sensor mutations that generate gating pore current also opens up therapeutic opportunities. Surprisingly, block of gating pore current by divalent cations and by a substituted guanidine derivative does not alter normal gating and function of NaV1.4 channels (Sokolov et al., J. Gen. Physiol., in press). Therefore, small molecules that block gating pore current with high affinity and specificity could be therapeutically useful in HypoPP, NormoPP, and other channelopathies in which gating pore current may contribute to allele-specific pathophysiology. Design of such agents will benefit from the emerging knowledge of the structure of the normal gating pore and the effects that mutations of gating charges may have on it. The horizon seems bright for development of a new generation of highly selective drugs that alter voltage sensor function and/or block gating pore current in a state-dependent manner as a novel therapeutic mechanism.
The author thanks Dr. Vladimir Yarov-Yarovoy, Dr. Stanislav Sokolov, Dr. Jian Payandeh, Mr. Paul DeCaen, Ms. Ellen Coyle, and Ms. Sharon Haywood for assistance with illustrations. Work in the author’s laboratory was supported by research grants from the National Institutes of Health (R01 NS15751 and U01 NS058039) and the Muscular Dystrophy Association.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.