Research Articles

Structure of a eukaryotic voltage-gated sodium channel at near-atomic resolution

See allHide authors and affiliations

Science  03 Mar 2017:
Vol. 355, Issue 6328, eaal4326
DOI: 10.1126/science.aal4326

Navigating regulated cell excitation

Voltage-gated sodium (Nav) channels respond to a change in voltage potential by allowing sodium ions to move into cells, thus initiating electrical signaling. Mutations in Nav channels cause neurological and cardiovascular disorders, making the channels important therapeutic targets. Shen et al. determined a high-resolution structure of a Nav channel from the American cockroach by electron microscopy. The structure affords insight into voltage sensing and ion permeability and provides a foundation for understanding function and disease mechanism of Nav and the related Cav ion channels.

Science, this issue p. eaal4326

Structured Abstract

INTRODUCTION

Voltage-gated sodium (Nav) channels are responsible for the generation and propagation of action potentials in excitable cells. They undergo voltage-dependent activation to initiate electrical signaling at millisecond scale and inactivate by both fast and slow mechanisms. The eukaryotic Nav channels comprise a pore-forming α subunit and auxiliary β subunits that facilitate membrane localization and modulate channel properties. The α subunit is a single polypeptide chain that folds to four homologous repeats (domains I to IV), each containing six transmembrane segments, S1 to S6. The S5 and S6 segments enclose the central pore domain, and their intervening sequences constitute the selectivity filter (SF). One residue at the corresponding SF locus in each repeat, Asp/Glu/Lys/Ala (DEKA), determines Na+ selectivity. The S1 to S4 segments in each repeat form a voltage-sensing domain (VSD), wherein S4 carries repetitively occurring positive residues essential for voltage sensing.

More than 1000 mutations have been identified in human Nav channels associated with various neurological and cardiovascular disorders. Nav channels represent important targets for multiple pharmaceutical drugs and natural toxins. The atomic structure of a eukaryotic Nav channel is required to reveal the molecular basis for ion selectivity, voltage-dependent activation and inactivation, and recognition of toxins, agonists, and antagonists.

RATIONALE

The technological breakthrough of electron microscopy (EM) has offered unprecedented opportunity for structure elucidation of eukaryotic Nav channels, but the bottleneck exists in the generation of sufficient amounts of high-quality proteins. After extensive screening, we succeeded in obtaining homogeneous proteins of PaFPC1, a putative Nav channel from the American cockroach. Despite the lack of electrophysiological characterizations, PaFPC1 contains all the hallmarks of a Nav channel except for the fast inactivation motif. We designated the protein NavPaS.

RESULTS

The cryogenic EM (cryo-EM) structure of NavPaS was determined with an overall resolution of 3.8 Å, allowing side-chain assignment for the complete transmembrane fold, extracellular loops of the pore domain, the intact III-IV linker, and the carboxy-terminal domain (CTD). A poly-Ala backbone was modeled for an amino-terminal domain that is located below VSDI. In addition, 20 sugar moieties were built into seven glycosylation sites on the extracellular loops. Conserved disulfide bonds are observed within the extracellular loops and between P2II and S6II segments.

The asymmetric selectivity filter vestibule is constituted by the side chains of the signature DEKA residues and the carbonyl oxygens of the two preceding residues in each repeat. The closed pore domain has only one small fenestration constituted by the S6III and S6IV segments. The four VSDs exhibit distinct conformations, with the corresponding gating charges located at different heights relative to their respective charge transfer center. Extensive interactions are observed between the III-IV linker, CTD, VSDIV, S4-S5IV, S6IV, and S6III segments of NavPaS. Despite the sequence variations of the III-IV linker and the lack of Ile/Met/Phe/Thr motif, the structure provides an important clue to understanding the fast inactivation mechanism of Nav channels.

CONCLUSION

The structure of a single-chain eukaryotic Nav channel serves as the framework for elucidating function and disease mechanisms of Nav channels. It provides the molecular template for interpretation of a wealth of experimental observations accumulated over the past six decades. Structural comparison between the related NavPaS and Cav1.1 reveals conformational shifts that may shed light on the understanding of the electromechanical coupling mechanism of voltage-gated channels. Structure-guided protein engineering will facilitate future mechanistic investigations of Nav and Cav channels.

The cryo-EM structure of a eukaryotic Nav channel at 3.8-Å resolution.

(Left) The overall structure of NavPaS/PaFPC1, a putative Nav channel identified from American cockroach. The glycosyl moieties and disulfide bonds are shown as sticks and spheres, respectively. The structure is domain colored. (Right) The pore domain of NavPaS. The permeation path is illustrated by brown dots in the pore domain, and the corresponding pore radii along the conducting passage are tabulated on the right. The functional entities along the permeation path—including the selectivity filter, the central cavity, and the intracellular activation gate—are annotated.

Abstract

Voltage-gated sodium (Nav) channels are responsible for the initiation and propagation of action potentials. They are associated with a variety of channelopathies and are targeted by multiple pharmaceutical drugs and natural toxins. Here, we report the cryogenic electron microscopy structure of a putative Nav channel from American cockroach (designated NavPaS) at 3.8 angstrom resolution. The voltage-sensing domains (VSDs) of the four repeats exhibit distinct conformations. The entrance to the asymmetric selectivity filter vestibule is guarded by heavily glycosylated and disulfide bond–stabilized extracellular loops. On the cytoplasmic side, a conserved amino-terminal domain is placed below VSDI, and a carboxy-terminal domain binds to the III-IV linker. The structure of NavPaS establishes an important foundation for understanding function and disease mechanism of Nav and related voltage-gated calcium channels.

Voltage-gated sodium (Nav) channels are ubiquitously present in eukaryotes and responsible for the initiation of electrical signaling in excitable systems such as nerve and muscle. The Nav channels remain closed at resting potentials and activate upon membrane depolarization, allowing for the Na+ influx that corresponds to the rapid upstroke of an action potential. Nav channels inactivate rapidly; meanwhile, the activation of voltage-gated potassium (Kv) channels resets the membrane electric field to the resting state. Defects in Nav channels underlie a variety of neurological and cardiovascular disorders. More than 1000 mutations have been identified in human Nav channels that are associated with epilepsy, arrhythmia, muscle paralysis, chronic pain, and other syndromes (15). Nav channels are targeted by multiple toxins and pharmaceutical drugs (6, 7).

The eukaryotic Nav channels consist of a pore-forming α subunit and auxiliary β subunits (8). The α subunit is sufficient for voltage-dependent ion conductance, whereas the β subunits facilitate membrane localization and modulate channel properties (9). There are nine isoforms of Nav α subunits in humans, designated Nav1.1 to Nav1.9 (10). A sequence-related protein, now named NavX, was found to function in salt sensing (11). The α subunit, similar to the closely related Cav channels (12), is composed of a single polypeptide chain that folds to four homologous repeats (domains I to IV), each containing six transmembrane segments, S1 to S6 (Fig. 1A). The four sets of S5 and S6 segments and their intervening sequences enclose the ion-permeation pore domain, whereas the S1 to S4 segments in each repeat form a voltage-sensing domain (VSD) (13). The α subunit is subjected to multiple posttranslational modifications, such as glycosylation, phosphorylation, and palmitylation (14). Nav channels have also been found in bacteria, exemplified by NaChBac (15). Similar to Kv channels, bacterial Nav channels are tetramers of four identical subunits.

Fig. 1 The structure of NavPaS determined using single-particle cryo-EM.

(A) The general topology of eukaryotic Nav channels. The previously identified key residues in ion selectivity and voltage sensing are indicated. “F” represents cyclic hydrophobic residues that occlude the gating charges from either side of the membrane. “-” represents the negative or polar residues on S2 segments that are designated An1 and An2. (B) The EM reconstruction of NavPaS. The map was generated in Chimera (97). (C) The gold-standard FSC curve for the 3D reconstruction of the EM map. (D) The overall structure of NavPaS. The structure is domain colored. The glycosyl moieties and disulfide bonds are shown as sticks and spheres, respectively. All structure figures were prepared in PyMol (98).

The ion selectivity of a voltage-gated ion channel is determined by the selectivity filter (SF), a molecular sieve enclosed by the partial membrane penetration loops between the S5 and S6 segments (1618). The residues that determine ion selectivity are identical in each protomer in the homotetrameric bacterial channels (15, 19) and in the four homologous repeats of eukaryotic Cav channels (20, 21). In contrast, a different residue, Asp/Glu/Lys/Ala (DEKA), is present at the corresponding SF locus in each repeat of eukaryotic Nav channels (22, 23) (Fig. 1A).

The eukaryotic Nav channels can undergo rapid voltage-dependent transitions between closed and open states, a prerequisite for producing the regenerative wave of electrical signals. The inner tetrahelical bundle of the pore domain screws at the cytoplasmic boundary of the membrane, forming the intracellular activation gate. Essential for voltage gating, VSDs contain the “gating charges” (24), which are a set of highly conserved positively charged residues occurring at every third place along the S4 segment. Upon depolarization, according to measurements in Kv channels, approximately 12 gating charges per channel are transferred across the membrane from the intracellular side to the extracellular side (2527). During this process, the gating charges interact sequentially with conserved acidic or polar residues on the S2 (designated An1 and An2) and S3 segments (2830). A charge-transfer center consisting of An2 and a cyclic hydrophobic residue on S2 and an adjacent Asp on S3 was identified that facilitates charge transfer (31).

Although multiple models of voltage-dependent activation have been proposed, the precise mechanism remains to be elucidated. It is generally accepted that the S4-S5 linker helices translate the voltage-dependent shifts of S4 segments to pore openings in voltage-gated ion channels with the “canonical” domain swapped arrangement (32). However, the S4-S5 helices are missing in the recently resolved structures of Eag1, Slo1, HCN1, and CNG channels whose domains are not swapped, suggesting a potentially different electromechanical coupling mechanism for these channels (3336).

Although the activation mechanism has not been fully understood, at least as bewildering are the intricate inactivation mechanisms for voltage-gated channels. Fast inactivation, which takes place on a millisecond scale, is executed by a cytoplasmic moiety between repeats III and IV of Nav channels (37). During prolonged depolarization, slow inactivation takes place by a mechanism that is not well understood (38), although it is thought at least in some cases that a conformational change of the selectivity filter underlies this process in Nav and Kv channels (3943).

A detailed atomic model of a eukaryotic Nav channel is required to reveal the molecular basis for ion selectivity, voltage-dependent activation and inactivation, and recognition of toxins, agonists, and inhibitors. Previous structural interpretations of Nav channels have been mostly based on the crystal structures of Kv channels and then on the bacterial Nav channels (17, 30, 32, 44, 45). The technological breakthrough in electron microscopy (EM) has offered unprecedented opportunity for structure elucidation at near-atomic resolutions of macromolecules that were challenging by x-ray crystallography (4648). The recent structural elucidation of the Cav1.1 channel complex provides a more relevant template for homology modeling of Nav channels (49, 50).

The bottleneck for structural determination of a eukaryotic Nav channel exists in the generation of a sufficient amount of high-quality proteins (51). Unsuccessful in obtaining homogeneous proteins from either electric eel or mammalian Nav channel preparations, we turned to insect Nav channels (52). Similar to their mammalian relatives, the insect Nav channels have single-chain α subunit and auxiliary subunit exemplified by TipE (53). Among the insect homologs tested, we succeeded in purifying sufficient recombinant proteins for a putative Nav channel from the American cockroach Periplaneta americana (54). To facilitate description, the protein that was originally named PaFPC1 is designated NavPaS. Here, we report the structure of NavPaS at a nominal resolution of 3.8 Å, determined using single-particle cryogenic EM (cryo-EM).

Results

Structural determination of NavPaS

NavPaS shares 36 to 43% sequence identities with human Nav1.1 to 1.9, containing the signature DEKA residues on the selectivity filter and highly conserved transmembrane segments (fig. S1). The variations mainly exist in the intracellular linkers between repeats. NavPaS, consisting of 1553 amino acid residues, has considerably shorter I-II and II-III linkers than the mammalian and other insect homologs, whose sequence lengths range between 1800 and 2050 residues. The III-IV linker of NavPaS is of identical length with those of the human channels but lacks the Ile/Phe/Met/Thr (IFMT) or similar motif that is critical for fast inactivation (55) (fig. S1).

The electrophysiological properties of NavPaS have not been characterized. We attempted to record NavPaS in several expression systems, including human embryonic kidney 293 (HEK293) cells, Chinese hamster ovary cells, and Xenopus laevis oocytes, in the presence or absence of the β subunits PaTipE and PaTEH1 (56, 57). Despite the fact that the channel can be expressed on plasma membrane even in the absence of any β subunit (fig. S2A), no Na+ current could be unambiguously assigned to NavPaS in all three tested systems, a problem that is occasionally encountered for heterologously expressed channels (30, 56). Although the physiological function and biophysical properties of NavPaS remain to be investigated, it possesses all the hallmarks of a Nav channel except for the fast inactivation motif and displays good solution behavior (figs. S1 and S2B). We thereby proceeded with structural determination using cryo-EM.

Details of grid preparation, cryo-EM data acquisition, and structural determination of NavPaS are presented in the supplementary materials (Fig. 1, B and C, and figs. S2 and S3). We selected 1.3 million particles and calculated an EM map to 3.8 Å according to the gold-standard Fourier shell correlation (FSC) 0.143 criterion (Fig. 1, C and D, and figs. S2, C to G). The EM maps are well resolved for most of the extracellular and transmembrane sequences (figs. S4 and S5). In total, 1323 residues are structurally modeled with 1177 side chains assigned, covering the complete voltage-gated ion channel fold, extracellular loops of the pore domain, the intact III-IV linker, and the carboxy-terminal domain (CTD). A poly-Ala backbone was built for the amino- terminal domain (NTD) containing residues 47 to 120 that immediately precede the VSD of repeat I (VSDI).

The missing segments are all on the intracellular side, including the N-terminal 46 amino acids, the I-II linker (residues 436 to 501) and II-III linker (residues 747 to 832), and the C-terminal 32 residues. In addition to the polypeptide chain, 20 sugar moieties are built into seven glycosylation sites on the extracellular loops, namely the L5 loops between the S5 and P1 segments and the L6 loops between the P2 and S6 segments (Fig. 1D and table S1).

The overall structure resembles that of the α1 subunit of the rabbit Cav1.1 (49). The sequence for NTD, which is located below VSDI and between VSDI and S6I, is conserved in Nav channels and harbors a number of disease-related mutations (50) (Fig. 1D and figs. S1 and S6). We will refrain from detailed analysis of NTD due to the low resolution for this domain (fig. S2F).

The asymmetric selectivity filter

In NavPaS, the SF vestibule is enclosed by the side groups of the signature residues Asp375/Glu701/Lys1061/Ala1353 at the upper position and the carbonyl oxygen atoms of the two preceding residues in each repeat at a lower level (Fig. 2, A and B). Above DEKA, the outer negative ring composed of Glu378/Glu704/Gln1065/Asp1356 guards the entrance to the SF vestibule. In human Nav channels, a conserved Asp in repeat III occupies the position corresponding to Gln1065 (Fig. 2A and fig. S1). In EM reconstructions, the densities for the carboxylate groups of Asp and Glu residues are usually invisible due to radiation damage. Fortunately, the backbone and the alkyl groups are discernible for most of the negatively charged residues in SF and the P1 and P2 helices of NavPaS (fig. S5, A and B).

Fig. 2 The asymmetric selectivity filter (SF) of NavPaS.

(A) The SF vestibule is enclosed by the side groups of Asp/Glu/Lys/Ala (DEKA) and the carbonyl oxygen atoms of the two preceding residues in each repeat. The residues that constitute the negative ring above the SF vestibule are also shown. (B) Structure of the signature DEKA residues that are critical for Na+ selectivity. Two perpendicular views are shown. (C) Structural variations of the selectivity filter between one-chain NavPaS and homotetrameric NavAb. Whereas the inner site constituted by the carbonyl oxygen groups remains similar, the outer site formed by the side groups of DEKA in NavPaS is distinct from that formed by Ser/Glu in NavAb (PDB code: 3RVY) with respect to both chemical composition and structural conformation. (D) The deviated backbone conformations of the SF sequences in the four repeats. The SF sequence in repeat II (yellow) is one residue shorter than those in the other three. The superimposed SF sequences and pore helices P1 and P2 in the four repeats are presented in two opposite side views. The Cα atoms of DEKA are shown as spheres.

The SF vestibule of a bacterial Nav channel is also composed of side groups for the outer site and eight carbonyl oxygen atoms for the inner site (17, 30, 45). However, the SF vestibule of the eukaryotic Nav channel is distinct from that of the bacterial counterparts with respect to both chemical composition and structural conformation, providing the molecular basis to investigate the different selectivity mechanisms between eukaryotic and bacterial Nav channels (58) (Fig. 2C). The SF sequence in repeat II is one residue shorter than in the other three repeats. Consistently, the backbone conformation of SFII deviates from the other three, further adding to the asymmetry of the SF vestibule (Fig. 2D and fig. S1).

At the current resolution, we refrain from assessment of the precise dimension of the SF vestibule (59) or analysis of the interactions between D/E and K. Nevertheless, the side group of Lys1061 appears to adopt a downward conformation, with the amino group pointing away from Asp375 and Glu701 (Fig. 2B and fig. S5, B and C). The protein for data acquisition was prepared in the presence of 50 mM NaCl. Discontinuous densities are observed in the SF vestibule. However, the relatively weak signal and the slight positional variations between half-maps make it difficult to distinguish between the densities of Na+ ions and noises (fig. S5B).

A closed pore with one small fenestration

The pore domain of NavPaS is well resolved in the EM reconstruction, allowing accurate assignment of side groups (Fig. 3A and figs. S4 and S5). Unexpectedly, the complete extracellular L5 and L6 loops are also resolved, suggesting their structural stability even in the absence of binding partners. Indeed, four pairs of disulfide bonds are identified within the L5I, L5III, and L6IV loops. An additional disulfide bond is found between the P2 helix and S6 in repeat II (Fig. 3A and fig. S7A). All these disulfide-forming Cys residues are invariant in the human and insect Nav channels (fig. S1).

Fig. 3 A closed pore with only one narrow fenestration.

(A) The heavily glycosylated extracellular loops above the pore domain are stabilized by multiple disulfide bonds. (B) The extracellular loops provide the electronegative potential surrounding the outer mouth to the SF vestibule. The surface electrostatic potential was calculated in PyMol. (C) The intracellular gate is closed. The permeation path, calculated by HOLE (99), is illustrated by brown dots in the left panel. The pore radii along the conducting passage are tabulated in the middle panel. The two layers of the hydrophobic residues that form the intracellular gate are shown as sticks in two perpendicular views. A bottom view of surface electrostatic potential of the pore domain is shown in the upper panel. (D) The polar and charged residues within the central cavity of the pore domain. An extracellular view is shown. The highly conserved Asn409/Asn734/Lys1105/Asn1404 residues at corresponding locus on S6 point to the central cavity in the current conformation. (See fig. S1 for detailed sequence comparison and fig. S7B for side views of the central cavity.) (E) The pore domain is sealed from the lipid bilayer except for a small fenestration enclosed by the S6 segments in repeats III and IV. The residues that seal the central cavity from the lipid bilayer are shown, with the invariant ones between NavPaS and human Nav channels as spheres and the altered ones as red sticks.

The L5 loops in repeats I and III are particularly long, containing short helices and antiparallel β strands. These bulky extracellular structures provide the scaffold for binding with the auxiliary β subunits through covalent or noncovalent interactions (9). A recent characterization identified Cys910 in hNav1.2, a residue on the L5II loop, to be disulfide-bonded with the β2 subunit (60). Despite the fact that the specific Cys is not conserved in NavPaS, the similar length and general sequence conservation of the L5II loop between NavPaS and human Nav channels will facilitate structural modeling of complex formation between α and β subunits of Nav channels (figs. S1 and S6).

Compared with the extracellular loops in Cav1.1 that are involved in the interactions with the α2δ subunit (18, 49), the L5 and L6 loops of NavPaS have more folded secondary structural elements and are heavily glycosylated (Fig. 3, A and B, and fig. S7A). The extracellular loops of NavPaS leave a spacious opening to the outer mouth of the SF vestibule. The surfaces above SF are enriched in negatively charged residues distributed on the L5 loops in repeats I to III and the L6IV loop, a structural shield that would attract cations and exclude anions (Fig. 3B).

The S6 tetrahelical bundle encloses a distinctive permeation path that is sealed at the intracellular boundary by two layers of hydrophobic residues Leu, Ile, Val, and Ala (Fig. 3C). Interestingly, the central cavity between the SF vestibule and the intracellular gate appears to have a smaller volume than that of Cav1.1 (49). In particular, a number of polar and charged residues line up the wall of the central cavity, including the highly conserved Asn residue (Asn409/Asn729/Lys1105/Asn1404) on each S6 segment (Fig. 3D and fig. S7B).

Similar to Cav1.1, the four-fold pseudosymmetry of the pore domain is disrupted by the conformational variations of the S5 and S6 segments (fig. S7C). Unexpectedly, only one small pore fenestration is observed on the side constituted by repeats III and IV, whereas the other three sides are completely sealed from the lipid bilayer, a conformational state that differs from the structures of prokaryotic Nav channels and Cav1.1, wherein pore fenestrations are present on all four sides (17, 30, 45, 49) (Fig. 3E and fig. S7D).

The III-IV linker and CTD

Another important functional element revealed by the NavPaS structure is the III-IV linker and the CTD, which interact with each other below repeat IV, reminiscent of the Cav1.1 structure (49) (Figs. 1D and 4A). The III-IV linker is involved in the fast inactivation gating of Nav channels, with the hydrophobic cluster IFM being a critical motif (55, 6163). The solution structure of an isolated III-IV linker and the cryo-EM structure of Cav1.1 both revealed the presence of a short α helix, presumably following the IFM motif (49, 64). However, the residues corresponding to IFM are missing in Cav1.1, preventing modeling of this motif in the context of the overall structure.

Fig. 4 Extensive interactions between the III-IV linker, CTD, and adjacent structural elements.

(A) The α helix–containing III-IV linker interacts with the globular CTD. An enlarged view of the III-IV linker, the CTD, and the adjacent pore-forming segments is shown in the inset. Residues Ala/Thr/Asp (ATD), which correspond to the fast inactivation motif IMF, are shown as spheres. (B) The IFM motif–corresponding residues are located on a turn connecting S6III and the III-IV helix. (Inset) The ATD motif is distanced from the lower layer of the intracellular gate by ~15 Å. (C) Extensive interactions between the III-IV linker, CTD, VSDIV, S4-S5IV, and S6IV. The polar and charged residues that may contribute to the interactions are shown as sticks. The corresponding residues in human Nav channels, if different from those in NavPaS, are annotated in parentheses. (D) NavPaS-CTD is structurally similar to Nav1.5-CTD. The structures of NavPaS-CTD and Nav1.5-CTD (PDB code: 4DCK) can be superimposed with a root-mean-square deviation of 0.318 Å over 71 Cα atoms.

In the reconstruction of NavPaS, the complete III-IV linker is resolved, comprising a helix followed by a long unwound segment (Fig. 4, A and B). The IFM-corresponding residues are 1127Ala/Thr/Asp1129 (ATD) in NavPaS, which are located at a short turn connecting the C terminus of S6III and the III-IV helix (Fig. 4B). In the side view, the ATD and the following III-IV helix are sandwiched between the C-terminal tips of S6III and S6IV, with the ATD motif at similar height to the C termini of S6III and S6IV, ~15 Å away from the lower level of the intracellular gate (Fig. 4B). Because the adjacent structural elements, including S4-S5IV, S6III, S6IV, and CTD, are conserved between NavPaS and human Nav channels, the position of ATD may reflect those of the corresponding IFM in the human Nav channels (fig. S1). The structure implies that the short linker between the C terminus of S6III and the IFM motif may impede the latter from reaching the intracellular gate unless S6III and S6IV undergo a pronounced conformational rearrangement. Notwithstanding the structural analysis, it remains to be characterized whether NavPaS undergoes fast inactivation.

The polar interactions along the intracellular surface of repeat IV are extensive, involving the III-IV linker, CTD, VSDIV, S4-S5IV, S6IV, and S6III (Fig. 4C). In addition to the contacts with CTD, the III-IV linker also binds to some cytoplasm-facing residues in VSDIV, S6IV, and S6III. The first α helix of CTD interacts with VSDIV and S4-S5IV (Fig. 4, C and D). The structure of NavPaS-CTD is nearly identical to that of the Nav1.5-CTD (Fig. 4D) (65). The interface residues between CTD and the S4-S5IV linker, exemplified by the three consecutive negatively charged residues on the first α helix of CTD and the repetitively occurring positive residues on the S4-S5IV linker, are highly conserved (Fig. 4C and fig. S1). Therefore, the cytoplasmic interfaces observed in NavPaS are likely preserved in the human Nav channels. The extensive interactions among the functional elements—including VSDIV, S4-S5IV, S6III, S6IV, III-IV linker, and CTD—provide a molecular foundation to interpret the critical role of VSDIV in inactivation in the recently proposed “asynchronous gating model” (4, 6668) and the involvement of CTD in inactivation (69, 70).

The voltage-sensing domains in distinct conformations

In the structure of NavPaS, the four VSDs, including the connecting loops, are completely resolved (fig. S4A). Unlike the homotetrameric Kv channels and bacterial Nav channels, the relative angles between neighboring VSDs of NavPaS are all several degrees deviated from 90° (Fig. 5A). The lengths and sequences of VSDIV are similar between NavPaS and human Nav1.7 (fig. S1). However, the S3 and S4 segments in NavPaS VSDs are each one helical turn shorter than those in the chimeric VSD generated by fusing the extracellular halves of VSDIV from Nav1.7 to the scaffold of NavAb (designated as the NavAb-1.7 VSD4) (71) (Fig. 5B). The four VSDs have different numbers of gating-charge residues (fig. S1). Similar to the numbering system used in Cav1.1 (49), we define the Arg/Lys on the last helical turn of the S4 segment as R6. Thereby, S4I and S4III have R2-R5, whereas S4II and S4IV have R2 to R6. The invariant Lys (K1) on S4III in human Nav channels is replaced by Gln948 in NavPaS (fig. S1).

Fig. 5 The four VSDs of NavPaS exhibit distinct conformational states.

(A) All four VSDs are completely resolved. A bottom view is shown with the intracellular segments omitted for visual clarity. The relative angles between neighboring VSDs are estimated in chimera, wherein the respective and common centroids of the VSDs are determined and the relative angles are calculated based on the distances between these centroids. (B) The NavPaS VSDs have shorter S3 and S4 segments than the chimeric NavAb-1.7 VSD4. The structural comparison is made between VSDII and VSDIV of NavPaS with NavAb-1.7 VSD4 (PDB code: 5EK0). (C) The four VSDs exhibit distinct conformations. The four VSD structures are superimposed relative to the CTC residues (An2 and the occluding Phe/Tyr on S2 and a conserved Asp on S3) and An1. The same reference is used for all the domain comparisons shown in this figure. The shifts of the backbone of S4 and the ensuing S4-S5 linker are indicated by red arrows. Two opposite side views are shown. (D) Structures of the NavPaS VSDs. For visual clarity, the S1 segment in each VSD is omitted. The side chains of the gating charges on S4, the CTC residues, and An1 are shown. The gating charge residues that are coordinated by An1 and An2 are labeled blue and red, respectively. (E) Pairwise comparison of the VSDs. The Cα atoms of the gating charge residues are shown as spheres.

In mammalian Nav channels, the four repeats display distinct activation kinetics with the molecular basis unclear (67). In the structure of NavPaS, the corresponding gating charges are at different positions relative to the charge transfer center (CTC) and An1, with VSDII and VSDIII at the most and least activated conformations, respectively, among the four (Fig. 5, C, D, and E).

The S4 segments adopt 310 helical conformations in the VSDs of repeats I to III, whereas the two extracellular helical turns of S4IV relax to α helix (Fig. 5D). Only R2 in VSDIII is coordinated by An1, whereas those in the other three VSDs are out of reach of An1. R3 is coordinated by An1 in three VSDs other than VSDII, in which R4 is “up” enough to interact with An1. R5 is similarly coordinated by An2 and the conserved Asp on S3 in VSDI and VSDII. The Cα of R5 in VSDIV is at a further “down” position relative to CTC, but its side chain can still bind to the two CTC negative residues. VSDIII represents the least activated one, as its R4 is coordinated by the negative residues of CTC below the occluding residue, Phe901 (Fig. 5D).

It is noteworthy that the Cα atoms of all the gating charges are at similar heights in S4III and S4IV. It is the swing of the long side chain of Arg that places the guanidinium group of R4 below or above the occluding Phe in VSDIII and VSDIV, respectively (Fig. 5E). A more pronounced difference involves the backbone shift. The backbone of S4III moves down by one complete helical turn relative to S4II and by half a helical turn in height relative to S4I when the corresponding VSDs are aligned with respect to CTC and An1 (Fig. 5E). The molecular determinants for the heterogeneous conformations of the VSDs at 0 mV remain to be elucidated.

Structural comparison between NavPaS and Cav1.1

Among the prokaryotic and eukaryotic Nav and Cav channels whose structures are available, Cav1.1 is most closely related to NavPaS (fig. S8). Structural comparison of NavPaS and Cav1.1 reveals similarity in the general structural organization and, more interestingly, pronounced conformational changes of both the pore domain and VSDs (Fig. 6A and Movie 1).

Fig. 6 Structural changes between NavPaS and Cav1.1.

(A) Comparison of the overall structures of NavPaS and Cav1.1 (PDB code: 5GJV, chain A). The structures are superimposed relative to the funnel constituted by the P1-SF-P2 segments in the four repeats. The CTDs are omitted in the intracellular view. (Right) The S4-S5 constriction ring appears to be relaxed from NavPaS to Cav1.1. The red arrows indicate the shifts of the corresponding structural elements from NavPaS to Cav1.1 in the intracellular view. (B) The S6 segments undergo axial rotation between NavPaS and Cav1.1, placing the conserved Asn residues inside or outside the central cavity of the pore domain, respectively. Structural alignment suggests that the rotations initiate at the G(S/T)F motif or corresponding positions on S6. (C) Structural shifts of S5 and S6 are coupled to the opening or closure of the fenestrations. (Inset) The corresponding residues that are involved in the binding of local anesthetic drugs adopt distinct conformations in NavPaS and Cav1.1. The residues Cys1399 and Tyr1406 correspond to Phe1764 and Tyr1771 in Nav1.2, respectively (75). See Movies 1 to 3 for detailed analysis of the conformational changes between NavPaS and Cav1.1.

Movie 1 The overall structural changes between NavPaS and Cav1.1.

A homologous model of NavPaS derived from the structure of Cav1.1 was generated based on sequence alignment in fig. S8 using the SWISS-MODEL web site (https://www.swissmodel.expasy.org) (100103). The modeled structure was superimposed to the cryo-EM structure of NavPaS relative to the selectivity filter and the P1 and P2 helices (the P1-SF-P2 funnel). The cryo-EM structure of NavPaS and the homologous structural model of NavPaS were used as the initial and end frame, respectively, for morph generation. The intermediate morphs were generated using the Crystallography and NMR System (CNS) (104, 105). The movies were prepared in PyMOL. The NTD and extracellular loops are omitted. The structures are domain colored following the same scheme as in Fig. 1D.

When the structures of NavPaS and Cav1.1 are overlaid relative to the funnel constituted by the selectivity filter and pore helices (the P1-SF-P2 segments in the corresponding repeats), VSDII and VSDIV of NavPaS exhibit a slight clockwise rotation from the corresponding domains of Cav1.1 in the intracellular view, whereas the respective positions of VSDI and VSDIII are similar in the two structures (Fig. 6A).

Despite the fact that both channels are closed at the intracellular gate, the pore domain of NavPaS appears to be further tightened by an overall right-handed screw of the S5 and S6 segments, as well as the S4-S5 constriction ring around the central axis of the pore domain (Fig. 6A and Movie 1). Similar structural shifts also occur between NavPaS and bacterial Nav channels (fig. S9).

Accompanying the overall twisting of the S6 tetrahelical bundle, each S6 segment undergoes axial rotation starting at the relatively conserved G(S/T)F motif or corresponding positions (Fig. 6B, Movie 2, and figs. S1 and S8). Consequently, the aforementioned Asn residues point to the central cavity in NavPaS (Fig. 3D), whereas the corresponding ones in Cav1.1 are outside the cavity and able to contact residues in the neighboring S4-S5 segment (Fig. 6B, Movie 2, and figs. S1 and S8). A single point mutation of the conserved Asn on S6I or S6III alters slow inactivation in Nav1.2 and Nav1.4 (7274). The pronounced shifts of the corresponding residues between NavPaS and Cav1.1 provide a clue to the mechanistic understanding of their involvement in slow inactivation.

Movie 2 Conformational changes of the pore domain segments between NavPaS and Cav1.1.

An intracellular view is shown. The conserved Asn/Lys residues on the S6 segments are shown as sticks. The axial rotations of S6 segments between the two structures result in the placement of the conserved Asn/Lys residues inside or outside the central cavity.

The structural comparison between NavPaS and Cav1.1 suggests a potential mechanism for the gating of the pore domain fenestrations (Figs. 3E and 6C and Movie 3). From NavPaS to Cav1.1, the S5 and S6 helical bundles are relaxed through the overall irislike rotation. Accompanying this motion, all the closed pore fenestrations in the structure of NavPaS become open (Fig. 6C and Movie 3). The fenestration constituted by S6III and S6IV was suggested to be the receptor site for local anesthetic drugs, involving Phe1764 and Tyr1771 in Nav1.2 and Phe1579 in Nav1.4 (75, 76). The corresponding residue for Nav1.2-Phe1764 and Nav1.4-Phe1579 is Cys1399 in NavPaS, which sits on one side of the fenestration (Fig. 6C, inset). Tyr1406, which corresponds to Tyr1771 in Nav1.2, is on the opposite side of the fenestration in the structure of NavPaS but encloses the fenestration in the conformation of Cav1.1, owing to the axial rotation of S6IV (Fig. 6C, inset, and Movie 3). These structural observations may facilitate mechanistic elucidation of the state-dependent action of local anesthetics targeting Nav channels.

Movie 3 The structural shifts of S5 and S6 segments lead to the opening or closure of fenestrations.

The pore domain is shown as a cartoon under semitransparent surface. The S6III and S6IV residues corresponding to those in Nav1.2 and Nav1.4 that may be involved in binding to local anesthetics are shown as spheres (75, 76).

Discussion

Despite the fact that NavPaS and Cav1.1 are both captured at closed conformations and their functional states are yet to be defined, the concerted conformational shifts of pore domain elements and VSDs between the two structures may shed light on the understanding of the electromechanical coupling mechanism of voltage-gated ion channels (Fig. 7 and Movies 4 and 5).

Fig. 7 Implications on the electromechanical coupling mechanism of Nav and Cav channels.

(A) The conformational changes of the corresponding VSDs and neighboring S5 segments between NavPaS and Cav1.1. The two structures are superimposed as in Fig. 6A. The red arrows indicate the shifts of the corresponding segments from NavPaS to Cav1.1. (B) The electromechanical coupling of voltage-gated ion channels with “canonical” fold may involve intricate and extensive interactions between adjacent S4, S4-S5, S5, and S6 segments in the same and neighboring repeats. The potential force transmissions that are responsible for coupled motions between intra- and interrepeat elements are indicated by orange and red arrows, respectively. (C) A number of disease-related mutations of Nav channels mapped to the interface between S4 and S5 segments in the neighboring repeats. The Cα atoms of representative disease-related residues are shown as spheres, and the interface residues are shown as sticks. The residues are color-coded for the Nav subtypes in which the mutations were identified (50). When mutations occur to the same locus in different channels, the one in the channel with the higher number is shown. Please refer to Movies 4 and 5 for the conformational changes illustrated in this figure.

Movie 4 Coupled motions between VSDs and pore domain elements.

The shifts of S4 and S4-S5 linker in one repeat appear to be coupled to the concerted motion of S5 and S6 in the neighboring repeat.

Movie 5 Representative disease-related mutations mapped to the interface between S4 and S5 segments in the adjacent repeats.

It is noted that the interfaces between S4I and S5II and between S4III and S5IV host more disease mutations than those between S4II and S5III and between S4IV and S5I.

In the cryo-EM structure of Cav1.1, all four VSDs display similar conformations, with R5 coordinated by CTC and R4 close to An1 (49), a state reminiscent of NavPaS-VSDII. Therefore, VSDs I, III, and IV in Cav1.1 appear to be more activated than the corresponding ones in NavPaS (Fig. 5, D and E). From NavPaS to Cav1.1, the S4 segments in these three VSDs indeed display a tilting motion, which is a combination of upward shift of the helix and outward displacement of the intracellular tip of S4 from the central axis (Fig. 7A and Movie 4). The ensuing S4-S5 segments are pulled outward by the S4 segments accordingly. Due to the tight connection between the S4-S5 segment and the ensuing S5 segment, any shift of S4-S5 can be immediately translated to S5 in the same repeat.

The coupled motions between VSDs and pore domain are more evident for VSDI and VSDIII and their respective neighboring pore domain segments (Movie 4). S5II and S5IV appear to undergo concerted shifts with S4I and S4III, respectively, as the involved interfaces are largely preserved in NavPaS and Cav1.1 (Fig. 7A and Movie 4). The neighboring S4-triggered shift of an S5 segment may in turn drag its preceding S4-S5 segment to move accordingly. The motion of one S4-S5 segment would then be coupled to the neighboring S4-S5, S5, and S6 segments. The conformation and position of S6 can also be affected by both S5 in the same repeat and S6 from the neighboring repeat (Fig. 7B). To be brief, the electromechanical coupling of a voltage-gated ion channel, which is like a delicate gear, may involve extraordinarily intricate force transmissions among S4-S5, S5, and S6 segments in the same repeat and between neighboring repeats once the motion is triggered by the voltage-dependent conformational shifts of VSDs (Fig. 7B).

More than 1000 disease-associated mutations have been identified in human Nav channels (50). It is known that the selectivity filter, gating charge residues, and the III-IV linker are hot spots for disease mutations, although the disease-causing mechanism for the majority of the mutations remains enigmatic. The structure of NavPaS provides a template to map hundreds of disease mutations and serves as a framework to mechanistically investigate these mutations (fig. S10). For instance, a number of mutations are mapped to the interfaces between S4 and S5 of neighboring repeats. Alteration of the interface residues may affect electromechanical coupling as analyzed above (Fig. 7, B and C, and Movie 5).

Conclusions

The near–atomic resolution structure of a single-chain eukaryotic Nav channel establishes an important foundation for investigating the function and disease mechanisms of Nav channels and for structure-guided drug development. Despite the exciting features revealed by the structure, many important questions remain to be answered. The Na+ selectivity cannot be explained by a single structure, especially in the absence of the bound Na+ in the SF vestibule. The near–atomic resolution structure of NavPaS provides the template for molecular dynamics simulations analysis. The conformational changes between NavPaS and Cav1.1 may shed light on the molecular understanding of the electromechanical coupling mechanism of Nav and Cav channels. Nonetheless, we have to interpret the conformational changes with caution because the comparison is made for two channels with considerable sequence variations, and the functional states of the structures are yet to be defined. A series of structures of the same channel in distinct states is required to establish the structure-function correlation. The technological advances of cryo-EM and a protein that can be conveniently manipulated will make this premise possible.

Materials and methods

Transient expression of NavPaS

The optimized coding DNA for NavPaS (Uniprot: D0E0C2) with Twin-Strep-tag and FLAG tag in tandem at the amino terminus was cloned into the pCAG vector (77). HEK293F cells (Invitrogen) were cultured in SMM 293T-I medium (Sino Biological Inc.) under 5% CO2 in a Multitron-Pro shaker (Infors, 130 rpm) at 37°C. When the cell density reached 4 × 106 cells per ml, the plasmid was transfected. For one liter cell culture, 3 mg plasmids were pre-incubated with 6 mg 25-kDa linear polyethylenimines (PEIs) (Polysciences) in 25 ml fresh medium for 15–30 min. The transfection was initiated by adding the mixture into cell culture, followed by dilution of the cell culture to 2 × 106 cells per ml with fresh medium. Transfected cells were cultured for 48 hours before harvesting. The expression of target protein was examined by Western blot using monoclonal antibodies against the FLAG tag.

Immunostaining of HEK293 cells expressing NavPaS

HEK293 cells, cultured on poly-L-ornithine (Sigma)-coated 15 mm circular coverslips seated in 35 mm dishes, were cotransfected with the expression plasmid for NavPaS and an eGFP-coding plasmid when cell confluency reached 30-50%. After incubation at 37°C under 5% CO2 for 36 hours, the transfected cells were washed three times with PBS (Gibco) before being fixed by 4% paraformaldehyde (Electron Microscopy Sciences) in PBS at room temperature for 15 min. Rinsed twice, the cells were then blocked by 4% BSA (Sigma) in PBS plus 0.5% (w/v) Triton X-100 (PBS-TX) (Invitrogen) for one hour before the 4% BSA was replaced by anti-Strep-tag primary antibody (Biocompare). After incubation at 4°C overnight, the cells were washed three times with PBS-TX before treated with the goat anti-mouse IgG highly cross-adsorbed secondary antibody, Alexa Fluor 568 (Invitrogen) and phalloidin, Alexa Fluor 647 (Invitrogen). After being incubated at 22°C for two hours, the cells were washed three times with PBS-TX. The coverslips were covered by a layer of DABCO (Sigma) and sealed on an object slide by transparent nail polish. The sample was imaged under microscope LSM780 (Zeiss).

Purification of the NavPaS protein

For each batch of protein purification, eight liters of transfected HEK293F were harvested by centrifugation at 800 g and resuspended in lysis buffer containing 25 mM Tris-HCl, pH 7.4, and 50 mM NaCl. The suspension was supplemented with 1% (w/v) digitonin (Sigma), and protease inhibitor cocktail containing 2 mM phenylmethylsulfonyl fluoride (PMSF), 2.6 μg/ml aprotinin, 1.4 μg/ml pepstatin, and 10 μg/ml leupeptin, and incubated at 4°C for 2 hours. After ultra-centrifugation at 200,000 g for 25 min, the supernatant was applied to anti-Flag M2 affinity gel (Sigma) and flowed through by gravity at 4°C. The resin was rinsed three times with the wash buffer containing 25 mM Tris-HCl, pH 7.4, 50 mM NaCl, 0.1% digitonin (Sigma), and protease inhibitor cocktail. The target NavPaS protein was eluted with wash buffer plus 200 μg/ml FLAG peptide (Sigma). The eluent was then applied to the Strep-Tactin Sepharose (IBA) and flowed through by gravity. The resin was rinsed three times with the aforementioned wash buffer. The protein was eluted with the wash buffer plus 2.5 mM D-Desthiobiotin (IBA). The eluent was then concentrated using a 100-kDa cut-off Centricon (Millipore) and further purified by size exclusion chromatography (Superose-6, GE Healthcare). The peak fractions were pooled and concentrated to 50 μl at a concentration of approximately 1 mg/ml.

Cryo-EM data acquisition

Cryo-EM grids were prepared with Vitrobot Mark IV (FEI). Aliquots (3.5 μl) of freshly purified NavPaS were placed on glow-discharged holey carbon grids (Quantifoil Cu R1.2/1.3). Grids were blotted for 3.5 s and flash-frozen in liquid ethane cooled by liquid nitrogen. A total of 13,757 movie stacks were semi-automatically collected with UCSF Image4 in super-resolution mode on Titan Krios operating at 300 kV and equipped with Gatan K2 Summit detector at a nominal magnification of 22,500 X. The defocus range was set from –1.7 μm to –2.6 μm. Each stack was exposed for 8 s with an exposing time of 0.25 s per frame, resulting in a total of 32 frames per stack. The total dose rate was about 50 e-2 for each stack. The stacks were first motion corrected with MotionCorr (78) and binned 2 fold, resulting in a pixel size of 1.307 Å/pixel. The output stacks from MotionCorr were further motion corrected with MotionCor2 (79), meanwhile dose weighting was performed (80). The defocus values were estimated with Gctf (81).

Image processing

A diagram of the procedures for data processing is presented in fig. S3. A total of 11,843 good micrographs were manually selected, from which a total of 4,739,175 particles were automatically picked using RELION 1.4 (8284) or 2.0 (85). After two-dimensional (2D) classification with RELION 1.4 or 2.0, a total of 3,039,806 good particles were selected and subjected to global angular search 3D classification using RELION 2.0 with one class and step size of 7.5°. The trans-membrane domain and CTD of Cav1.1α1 that was low-pass filtered to 60 Å was used as the initial model. After global angular search 3D classification, the particles were further subjected to 3D classification with 3 to 10 classes and local angular search step of 3.75°. The local angular search 3D classification was performed for several times, with the output from different iterations of the global angular search 3D classification as input. A total of 1,962,964 good particles from the local angular search 3D classification were combined and re-extracted from the magnification distortion corrected micrographs (86). The pixel size after magnification distortion correction is 1.295 Å/pixel. A total of 1,958,925 non-outlier particles were subjected to several cycles of random-phase 3D classification (87) to remove bad particles. The handedness of the 3D reconstruction was checked and corrected. The particles selected from random-phase 3D classification were subjected to 3D auto-refinement with RELION 2.0. The final particle number for the 3D auto-refinement is 1,373,581. The resolution was estimated with the gold-standard Fourier shell correlation 0.143 criterion (88) with a high-resolution noise substitution method (89).

Model building and structure refinement

The 3.8 Å reconstruction map was used for model building. The structure of Cav1.1α1 was used as the starting model (PDB code: 5GJV, chain A), fitted into the EM map by CHIMERA. CHAINSAW (90) was used to keep the side chains of the conserved residues and to remove the non-conserved side chains based on the sequence alignment between Cav1.1 and NavPaS. Model building was performed in COOT (91). The assignment of the four repeats was based on the extracellular loops and the CTD. De novo model building was then performed for each repeat. Sequence assignment was guided mainly by bulky residues. The chemical properties of amino acids were considered during model building. The model building of the CTD was facilitated by docking of the crystal structures of CTDs of Nav1.2 and Nav1.5 (PDB codes: 4JPZ and 4DCK, respectively). After these assignments, a globular density near the intracellular side of VSDI remained to be modeled. Backbone trace suggested that it should belong to the NTD. A poly Ala model was built for this domain due to insufficient resolution for side chain assignment.

Structure refinement was performed using phenix.real_space_refine application in PHENIX (92) in real space with secondary structure and geometry restraints to prevent structure over-fitting. The final model was refined against the 3.8 Å map using REFMAC (93) in reciprocal space, using secondary structure restraints that were generated by ProSMART (94). Overfitting of the overall model was monitored by refining the model in one of the two independent maps from the gold-standard refinement approach and testing the refined model against the other map (95) (fig. S2D). Statistics of the 3D reconstruction and model refinement can be found in table S1.

For structural analysis, the conserved residues between human Nav1.4 and NavPaS were mapped to the structure of NavPaS using ConSurf (96).

Supplementary Materials

References and Notes

  1. Acknowledgments: We thank J. Lei, X. Li, and X. Li for technical support during EM image acquisition. We thank K. Wu, J. Chen, S. Fu, X. Zhu, H. Qi, W. Xiong, B. Xiao, C. Canessa, and Q. Tao at Tsinghua University for technical support and critical discussions. We thank the Tsinghua University Branch of China National Center for Protein Sciences (Beijing) for providing the cryo-EM facility support. We thank the computational facility support on the cluster of Bio-Computing Platform (Tsinghua University Branch of China National Center for Protein Sciences Beijing) and the “Explorer 100” cluster system of Tsinghua National Laboratory for Information Science and Technology. This work was supported by funds from the Ministry of Science and Technology of China (2015CB910101, 2016YFA0500402, and 2014ZX09507003-006) and the National Natural Science Foundation of China (projects 31621092, 31630017, and 31611130036). N.Y. was supported in part by an International Early Career Scientist grant from the Howard Hughes Medical Institute and an endowed professorship from Bayer Healthcare. The atomic coordinates have been deposited in the Protein Data Bank (PDB) with accession code 5X0M, and the EM map has been deposited in the Electron Microscopy Data Bank with accession code EMD-6698.
View Abstract

Navigate This Article