Research Article

Structural basis for the modulation of voltage-gated sodium channels by animal toxins

See allHide authors and affiliations

Science  19 Oct 2018:
Vol. 362, Issue 6412, eaau2596
DOI: 10.1126/science.aau2596

Structures of voltage-gated sodium channels

In “excitable” cells, like neurons and muscle cells, a difference in electrical potential is used to transmit signals across the cell membrane. This difference is regulated by opening or closing ion channels in the cell membrane. For example, mutations in human voltage-gated sodium (Nav) channels are associated with disorders such as chronic pain, epilepsy, and cardiac arrhythmia. Pan et al. report the high-resolution structure of a human Nav channel, and Shen et al. report the structures of an insect Nav channel bound to the toxins that cause pufferfish and shellfish poisoning in humans. Together, the structures give insight into the molecular basis of sodium ion permeation and provide a path toward structure-based drug discovery.

Science, this issue p. eaau2486, p. eaau2596

Structured Abstract

INTRODUCTION

Almost all venoms contain toxins that modulate the activity of voltage-gated sodium (Nav) channels in order to incapacitate prey or predators. The single-chain eukaryotic Nav channels comprise four homologous repeats. The central pore domain is constituted by the carboxyl-terminal segments from all four repeats, and each repeat also has a voltage-sensing domain (VSD). Toxins are broadly divided into two categories—pore blockers that physically occlude the channel pore and gating modifiers that alter channel gating by interfering with the VSDs. Whereas small-molecule neurotoxins such as tetrodotoxin (TTX) and saxitoxin (STX) function as pore blockers, most peptidic Nav channel toxins are gating modifiers that trap the channel in a particular stage of the gating cycle through interactions with one or more VSDs. In neither case is the structural basis of channel modulation fully understood.

RATIONALE

Dc1a is a peptidic gating modifier toxin (GMT) from venom of the desert bush spider Diguetia canities that specifically binds to VSDII of insect Nav channels to promote channel opening. We showed through biochemical analysis that Dc1a interacts with NavPaS, a Nav channel from the American cockroach Periplaneta americana, for which a cryo–electron microscopy (cryo-EM) structure was recently determined at 3.8-Å resolution. We therefore sought to solve the structure of the complex between NavPaS and Dc1a. As Dc1a occupies a distinctly different channel binding site to pore blockers, we also attempted to supplement the complex with TTX or STX to obtain structures of the ternary complexes.

RESULTS

The cryo-EM structure of NavPaS-Dc1a was determined to an overall resolution of 2.8 Å in the presence of 300 mM NaCl, whereas those of NavPaS-Dc1a-TTX and NavPaS-Dc1a-STX were resolved at 2.6 Å and 3.2 Å, respectively, in the presence of 150 mM NaCl.

VSDII constitutes the primary docking site for Dc1a, which undergoes considerable structural rearrangement upon binding to the channel. The toxin inserts into the cleft between VSDII and the pore region, making intimate contacts with both domains. The network of intermolecular interactions seen in the cryo-EM structure was validated through examination of the effect of toxin and channel mutations using the orthologous NavBg channel from the German cockroach Blattella germanica.

Four residues, Asp/Glu/Lys/Ala (DEKA), at a corresponding locus in the selectivity filter (SF) of each repeat confer Na+ selectivity. A Na+ ion was observed in the same position in the structures of NavPaS-Dc1a and NavPaS-Dc1a-TTX, coordinated by the Asp and Glu residues in the DEKA motif of the SF, and an invariant Glu on the P2 helix in repeat II, a helix in the entryway to the SF on the extracellular side. Both TTX and STX form extensive electrostatic interactions with residues in the outer electronegative ring that attracts cations into the SF and Asp and Glu in the DEKA motif, completely blocking access of Na+ ions to the SF.

CONCLUSION

The structure of the NavPaS-Dc1a complex suggests that the network of interactions between Nav channels and GMTs is more complex than previously anticipated. Therefore, caution has to be applied when using isolated Nav channel VSDs for drug discovery or for understanding the molecular basis of GMT action. The current structures elucidate the molecular basis for the insect selectivity of Dc1a and the subtype-specific binding of TTX or STX to Nav channels. Unambiguous structural elucidation of the bound TTX and STX, whose molecular weights are both around 300 Da, showcases the power of cryo-EM and its potential for structure-aided drug discovery.

Structural basis for specific binding of GMT Dc1a and guanidinium pore blockers TTX and STX by NavPaS.

(A) Dc1a inserts into the extracellular cavity between VSDII and the pore elements of repeat III. (B) Molecular mechanism for pore blockade by TTX and STX. Top: The carboxylate groups of Asp (D) and Glu (E) residues in the DEKA motif and an invariant Glu on P2II together constitute a potential Na+ binding site (designated the DEE site). Bottom: TTX and STX block access of Na+ to the DEE site from the extracellular side. A semitransparent presentation of the electrostatic surface potential of the entrance to the SF viewed from the extracellular side is shown. CTD, C-terminal domain; R, Arg; L, Leu; Y, Tyr; K, Lys.

Abstract

Animal toxins that modulate the activity of voltage-gated sodium (Nav) channels are broadly divided into two categories—pore blockers and gating modifiers. The pore blockers tetrodotoxin (TTX) and saxitoxin (STX) are responsible for puffer fish and shellfish poisoning in humans, respectively. Here, we present structures of the insect Nav channel NavPaS bound to a gating modifier toxin Dc1a at 2.8 angstrom-resolution and in the presence of TTX or STX at 2.6-Å and 3.2-Å resolution, respectively. Dc1a inserts into the cleft between VSDII and the pore of NavPaS, making key contacts with both domains. The structures with bound TTX or STX reveal the molecular details for the specific blockade of Na+ access to the selectivity filter from the extracellular side by these guanidinium toxins. The structures shed light on structure-based development of Nav channel drugs.

Voltage-gated sodium (Nav) channels play a critical role in generating membrane excitability (1) and are targeted by numerous chemical insecticides and human drugs. Nav channels are also the most common target of venom neurotoxins. Nav channels comprise one single polypeptide chain that folds to four homologous repeats (repeats I to IV), each containing six transmembrane helices designated S1 to S6. The S1 to S4 segments in each repeat constitute the voltage-sensing domain (VSD), and the S5, S6, and their intervening segments from the four repeats together enclose the ion-conducting pore domain.

Although small-molecule neurotoxins such as tetrodotoxin (TTX) and saxitoxin (STX) function as pore blockers, the vast majority of peptidic Nav channel toxins are gating modifiers that trap the channel in a particular stage of the gating cycle through interactions with one or more VSDs (2). In contrast to pore blockers, gating modifier toxins (GMTs) have more complex allosteric effects on Nav channel function, and they can inhibit (3) or agonize (4) the channel. GMTs, which generally have greater selectivity than pore blockers, are valuable leads for the development of subtype-selective Nav channel drugs (3, 5, 6).

Despite extensive studies of the molecular basis by which GMTs modulate Nav channel function, no consensus model of this interaction has emerged. Early studies suggested a dominant role for the extracellular S3-S4 loop in GMT binding (7, 8), but subsequent studies have revealed a key role for the S1-S2 loop in many GMT interactions (4, 9, 10). More recent studies suggest that GMTs nestle into an extracellular-facing cavity between the S1 and S4 helices, enabling them to act as a wedge that impedes voltage sensor movement (5, 11). It has been suggested that large GMTs such as those found in scorpion venom might be able to simultaneously contact the VSD and the extracellular loop connecting the pore helix P2 and the S6 segment in pore domain (12), but no studies to date have predicted a role for any of the pore-domain membrane helices in GMT binding.

Small molecules that occlude the pore of Nav channels are rare in animal venoms, but TTX and STX are exceptions. As the name indicates, TTX was originally found in tetrodontoid fish exemplified by the puffer fish (fugu). Puffer fish poisoning, resulting from consumption of TTX-containing fish, was documented thousands of years ago in China and Egypt and later in Japan and Mexico (13, 14). TTX was subsequently shown to be present in venom of the deadly blue-ringed octopus, in the poisonous secretions of frogs and newts, and in predatory moon snails; these animals do not synthesize TTX but rather acquire it from endosymbiotic bacteria (15). It was discovered in the mid–20th century that the potent toxicity of TTX is due to suppression of action potential generation through specific inhibition of Na+ influx (14, 1618). STX is a related guanidinium neurotoxin, produced by marine dinoflagellates and cyanobacteria, that competes with TTX for binding to Nav channels (15). The term saxitoxin is also used to refer to a class of >50 structurally related toxins that are responsible for paralytic shellfish poisoning (19, 20).

Because of their stringent specificity for Nav channels, TTX and STX are widely used for pharmacological characterization of Nav channels (2124). The nine subtypes of mammalian Nav channels are classified as TTX resistant or TTX sensitive based on their susceptibility to TTX. The latter are inhibited by nanomolar concentrations of TTX, whereas the TTX-resistant subtypes Nav1.5, Nav1.8, and Nav1.9 only respond to micromolar concentrations of the toxin (24, 25). Despite comprehensive studies over the past six decades (2629), our molecular understanding of the mechanism of action of these toxins has been impeded by the lack of structural information. Crystal structures of several bacterial Nav channels have been elucidated (3032), but these homotetrameric prokaryotic orthologs are insensitive to TTX and STX because they lack the receptor site found in their single-chain, asymmetric eukaryotic counterparts (33).

We recently elucidated the structure of the eukaryotic Nav channel NavPaS from the American cockroach Periplaneta americana at 3.8-Å resolution (34). Here, we present a 2.8-Å resolution cryo–electron microscopy (cryo-EM) structure of this channel in complex with Dc1a, a peptidic GMT from venom of the desert bush spider Diguetia canities that promotes channel opening (10). We also report cryo-EM structures of the NavPaS-Dc1a complex in the presence of the pore blockers TTX and STX at 2.6 Å and 3.2 Å, respectively. A Na+ binding site in the selectivity filter (SF) constituted by three carboxylate groups is observed. The structures elucidate the molecular basis for pore blockade by TTX and STX.

Results

Structural determination of NavPaS in complex with Dc1a, TTX, and STX

Details of cryo-sample preparation, image acquisition, data processing, model building, and structure refinement can be found in the materials and methods. Briefly, micrographs collected on a Titan Krios electron microscope equipped with Gatan K2 Summit detector, GIF Quantum energy filter, and spherical aberration (Cs) image corrector were used to reconstruct a three-dimensional (3D) EM map for the NavPaS-Dc1a complex purified in the presence of 300 mM NaCl to an overall resolution of 2.8 Å. Following a similar protocol, the structures of NavPaS-Dc1a bound to TTX and STX were obtained at 2.6 and 3.2 Å, respectively. The central region of NavPaS exhibits higher resolution in all three structures. Application of a mask for the central region during postprocessing further improved the resolution of this region to 2.7 Å for NavPaS-Dc1a and 3.1 Å for NavPaS-Dc1a-STX, whereas that for NavPaS-Dc1a-TTX remained at 2.6 Å (Fig. 1, A and B; figs. S1 to S4; and table S1). The excellent quality of the EM maps ensured reliable assignment of the ligands and surrounding residues. All residues of the SF, including the invariant residues from the four repeats, Asp/Glu/Lys/Ala (DEKA), and surrounding segments are unambiguously resolved in the high-resolution EM reconstructions (Fig. 1C and fig. S4).

Fig. 1 Structures of the complex between NavPaS and the peptide toxin Dc1a with or without TTX or STX.

(A) Gold standard Fourier shell correlation (FSC) curves for the 3D EM reconstructions of the NavPaS-Dc1a complex in the absence or presence of TTX or STX. Left: FSC curves for the overall structures. Right: FSC curves for the pore domains that were masked during postprocessing. (B) Local resolution map of the NavPaS-Dc1a-TTX complex. The map was estimated with RELION 2.0 and generated in Chimera. CTD, C-terminal domain. (C) Overall structure of the NavPaS-Dc1a-TTX complex. Side view and top view are shown. Because the three overall structures are nearly identical, only one is shown as a representative. The four repeats in NavPaS are shown in different colors, and Dc1a is colored orange. The sugar moieties are shown as black sticks. TTX, shown as black ball and sticks, is highlighted by the pink shade. The putative Na+ ion is shown as purple sphere. All structure figures were prepared with PyMOL (67).

VSDII and the pore domain together accommodate Dc1a

The structure of the Dc1a-NavPaS complex (Fig. 2, A to C) confirms that VSDII constitutes the primary docking site for Dc1a, as we reported previously (10). Comparison with the ligand-free NavPaS structure (34) reveals minor conformational changes in the channel upon Dc1a binding, mainly affecting VSDII (fig. S5). In contrast, the structure of Dc1a undergoes considerable rearrangement (Fig. 2B). The nuclear magnetic resonance (NMR) structure of Dc1a alone contains five short β strands that are organized into an N-terminal β sheet and a C-terminal inhibitor cystine knot (knottin) motif (10). In the complex, however, the two C-terminal β strands extend into the previously unstructured connecting loop region to form an elongated β hairpin that inserts deeply into the extracellular cavity enclosed by the four segments in VSDII and the adjacent pore-forming S5 segment from repeat III (S5III; Fig. 2, A and B).

Fig. 2 The interaction between Dc1a and NavPaS.

(A) Dc1a inserts into the extracellular cavity between VSDII and the pore elements of repeat III. The four disulfide bonds in Dc1a are shown as ball and sticks. EαIII: Extracellular α helix in repeat III. Inset: VSDII is shown as surface electrostatic potential calculated in PyMOL. (B) Conformational changes of Dc1a upon binding to NavPaS. The NMR-determined structure of free Dc1a (cyan) contains five short β strands, with β4 and β5 connected by a flexible linker. When binding to NavPaS, the segments containing β3-β5 (labeled as β3′-β5′ to be distinguished from those in the complex structure) become rigidified to form an elongated β hairpin. (C) Specific interactions between Dc1a and NavPaS. Electrostatic interactions are shown as red dashed lines. The three panels illustrate the contacts from top to bottom. Left: Interactions between Dc1a and the extracellular segments above the pore domain in repeat III of NavPaS. Middle: Interactions between Dc1a and the L1-2II loop (the loop that connects the S1 and S2 segments in VSDII). Asp542 and Arg549, which are not conserved in mammalian Nav channels, are highlighted with red labels. Right: Interactions of Dc1a with S4II and S5III. (D and E) Structure-guided mutagenesis characterizations corroborate the structural observations. (D) Shift in G-V curve (ΔG-V) for NavBg induced by wild-type (WT) and mutant Dc1a peptides (1 μM). WT Dc1a is shown in gray, whereas mutants are colored orange (n = 5 to 7). (E) Shift in G-V curve (ΔG-V) for WT and mutant NavBg channels induced by WT Dc1a (1 μM). Mutants are labeled according to the sequence of WT NavBg, with corresponding NavPaS numbering below. WT channel is shown in gray, whereas residues located in VSDII and S5-S6III are colored yellow and green, respectively (n = 5 to 6). All data are means ± SEM. Please refer to figs. S6 to S9 for experimental details. Single-letter abbreviations for amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

Dc1a makes extensive polar and hydrophobic interactions with NavPaS that are more complex than predicted by any model of GMT binding, involving interactions with the S1-S2 loop (the loop that connects S1 and S2), the extracellular pore loops, and the S5 pore-domain helix of repeat III. The toxin makes no interactions with the S3-S4 loop. The edge of the β sheet of Dc1a interacts with the short extracellular helix in repeat III above the pore domain (designated EαIII; Fig. 2B); specifically, Tyr33 and Asp56 on Dc1a interact with His1032 and Arg1027 on NavPaS, respectively (Fig. 2C, left). On one side of the VSDII cavity, the toxin interacts extensively with the S1-S2 loop (designated L1-2II); in particular, the guanidinium group of Dc1a-Arg41 interacts with the main-chain carbonyl oxygen and the side-chain carboxyl group of Asp542 (Fig. 2C, middle).

The β3-β4 hairpin of Dc1a inserts deeply into the VSDII cavity, with Phe47 and Phe48 at the tip of the hairpin surrounded by hydrophobic residues from S1II and the side wall of the pore domain involving S5III. Meanwhile, the aromatic ring of Dc1a-Phe48 makes a π-cation interaction with the gating-charge residue Arg613 (R3; Fig. 2C, right). Gln1002 on S5III makes extensive polar interactions with the side chain of Dc1a-Lys44 and the backbone amide of Dc1a-Phe48. These specific interactions with both VSDII and the pore domain collectively stabilize VSDII in the “up” state, consistent with Dc1a inducing opening of the channel (10).

We investigated the importance of these intermolecular interactions by examining the effect of toxin and channel mutations using the orthologous NavBg channel from the German cockroach Blattella germanica (Fig. 2, D and E). NavBg is potently activated by Dc1a, but unlike NavPaS, it is amenable to electrophysiological analysis (10). Modulation of NavBg activity by Dc1a was almost abolished when residues Asp21, Tyr33, Arg41, Lys44, and Asp56 were mutated to Ala, whereas Ala substitutions of Phe47, Phe48, and Ser49 severely diminished but did not completely abrogate Dc1a activity (Fig. 2D and fig. S6). On the basis of 1H NMR chemical shifts, none of these mutations perturb the structure of Dc1a (figs. S7 and S8); thus, we conclude that they all contribute to Dc1a modulation of insect Nav channels.

Mutation of NavBg residues involved in Dc1a binding caused minor shifts in the conductance-voltage (G-V) relationship for the channel (fig. S9). Thus, for each channel mutant, we quantified the previously noted ability of Dc1a to induce a hyperpolarizing shift (ΔG-V) in the G-V relationship (10), thereby allowing each mutant channel to serve as its own control (Fig. 2E and fig. S9). Both conservative (D→E) and harsher (D→A) mutations of Asp805 and Asp808 in L1-2II (corresponding to Asp539 and Asp542 in NavPaS) greatly reduced ΔG-V, highlighting the critical importance of these residues to Dc1a binding. Notably, neither residue is conserved in mammalian Nav channels (34), providing a molecular rationale for the insect selectivity of Dc1a (10). Mutation of Arg1447 to Lys in EαIII (Arg1027 in NavPaS) reduced Dc1a activity, providing support for this unexpected toxin-channel interaction, whereas mutation of His1452 in EαIII (His1032 in NavPaS) had minimal impact. Consistent with the cryo-EM structure (Fig. 2C), mutation of Arg876 (Arg610 in NavPaS), one of the uppermost gating-charge residues, had minimal impact on Dc1a activity, indicating that this residue is not crucial for the Dc1a-NavPaS interaction. Last, a conservative mutation of Gln1422 (Gln1002 in NavPaS) on S5III to Asn greatly reduced toxin activity (Fig. 2E), consistent with the interactions observed between this pore domain residue and residues Lys44 and Phe48 in Dc1a (Fig. 2C). Notably, the corresponding residue is Asn in human Nav1.1–1.8 and Tyr in Nav1.9, again consistent with the insect selectivity of Dc1a (10).

In summary, the mutagenesis data provide strong support for the physiological relevance of the complex network of intermolecular interactions observed in the Dc1a-NavPaS structure.

Recognition of TTX

TTX and STX, both of which have a molecular weight of ~300 Da, are clearly resolved in the cryo-EM reconstructions (Figs. 3 and 4 and fig. S4). TTX contains one guanidinium, two ether bonds, one oxygen anion, and multiple hydroxyl groups (Fig. 3A and fig. S4A). At lower pH, the oxygen anion is protonated. The map also reveals a density that likely belongs to a coordinated Na+ (fig. S4C), which we discuss further below.

Fig. 3 Specific interactions between NavPaS and TTX.

(A) Structure of TTX. Top: Chemical structure of TTX. Middle: The density for TTX, shown as blue mesh, is contoured at 10 σ. Bottom: Resolved 3D structure of TTX bound to NavPaS. (B) TTX is specifically coordinated by acidic residues and backbone amides at the outer vestibule of the SF. A stereo view from the extracellular side is shown. The putative Na+ ion is shown as purple sphere. (C) Detailed coordination of TTX by residues from the diagonal repeats shown in side views. (D) Sequence alignment of the SF elements and P2 helices in the four channel repeats. The panel is adapted from the reported sequence alignment (34). The residues whose side chains are involved in TTX coordination via polar interactions are shaded yellow. The residues whose backbone amides bind to the oxygen anion are shaded gray. In the TTX-resistant Nav subtypes, the equivalent of Tyr376, which appears to form π-cation interaction with the 1,2,3-guanidinium group of TTX, is Cys (Nav1.5) or Ser (Nav1.8 and Nav1.9).

Fig. 4 Specific interactions between NavPaS and STX.

(A) STX is well resolved in the 3.1-Å resolution EM reconstruction of the central domain of NavPaS. Left: An extracellular view of the overall EM map. The densities corresponding to STX and Dc1a are colored red and cyan, respectively. Right: Chemical structure (top) and 3D structure (bottom) of the NavPaS-bound STX. Middle: Density (blue mesh) for the bound STX at 5 σ. (B) STX is specifically coordinated by charged and polar residues from the four repeats at the outer vestibule of the SF. Left: STX tightly blocks the entrance to the SF. An extracellular view is shown. Middle and right: Side views of the coordination of STX by residues from the diagonal repeats. (C) Sequence alignment of the SF elements and P2 helices in the four repeats. Similar to Fig. 3D, the residues that participate in STX coordination via polar interactions are shaded yellow. Tyr376, which forms a π-cation interaction with the 1,2,3-guanidinium group of STX, is shaded gray. The locus corresponding to NavPaS-Gln1065 is Ile in human Nav1.7, which has a much lower affinity for STX than other subtypes of human Nav channels in which this locus is occupied by Asp.

TTX blocks the entrance to the SF vestibule through an extensive network of electrostatic interactions. The invariant acidic residues on the corresponding locus of the P2 segment in repeats I, II, and IV each form multiple hydrogen bonds or salt bridges with the polar groups of TTX (Fig. 3, B to D). Asp375 and Glu701 in the DEKA motif also directly contribute to TTX binding (Fig. 3C). Three consecutive backbone amides of the residues that demarcate the P2 helix from the preceding SF loop in repeat III simultaneously bind to C10-OH and the oxygen atom of the ether bond between C7 and C10, whereas Trp1063 and the carbonyl oxygen of Phe1060 coordinate C9-OH (Fig. 3, B and C).

All of the aforementioned residues are invariant in human Nav channels (Fig. 3D). Tyr376 on repeat I is positioned adjacent to the guanidinium group in TTX, enabling it to make π-cation interactions with the toxin (Fig. 3, B and C). The corresponding locus is occupied by either Phe or Tyr in TTX-sensitive Nav channel subtypes but replaced by Cys or Ser in the TTX-resistant subtypes Nav1.5, Nav1.8, and Nav1.9 (Fig. 3D). The structure therefore explains why substitution of Cys with Tyr at this position in Nav1.5 confers TTX sensitivity (24, 3537). The position after the invariant Glu on the first helical turn of P2I is occupied by Arg or Lys in TTX-resistant subtypes, whereas an Asn residue occupies this locus in TTX-sensitive channels (Fig. 3D). Although this residue is too far away to directly participate in TTX coordination, a basic residue at this site may reduce the local electronegativity and further lower channel affinity for TTX.

Recognition of STX

The functional groups of STX include the 1,2,3- and 7,8,9-guanidinium groups, the C12 hydroxyls (C12-OHs), and the 13-carbamoyl group. The distinctive shape of this small molecule allows reliable structural assignment (Fig. 4A and fig. S4, B and D). Polar and charged residues from all four repeats that are positioned at the outer entrance to the SF form extensive interactions with the functional groups of STX (Fig. 4, A and B).

The acidic residues on the first helical turn of the P2 helix in each repeat, which together constitute the outer electronegative ring, provide the primary docking site for STX. The 7,8,9- and 1,2,3-guanidinium groups are respectively bound to the invariant Glu residues on the P2 helix in repeats I and II, whereas the carbamoyl and C12-OH groups are coordinated by polar groups in repeat III and the invariant Asp in repeat IV, respectively. Tyr376 in repeat I contributes to coordination of the toxin through π-cation interaction with the 1,2,3-guanidinium group of STX (Fig. 4B). The carbonyl oxygen in the carbamoyl group and one adjacent C12-OH are hydrogen bonded to the backbone amide of the invariant Gly (Gly1062 in NavPaS) in repeat III. The DEKA-motif residue Glu701 in repeat II forms a hydrogen bond with N7 of STX (Fig. 4B, right).

All of the STX-coordinating residues in NavPaS, except for Tyr376 and Gln1065, are conserved in mammalian Nav channels (Fig. 4C). The corresponding residue for NavPaS-Gln1065 is Asp in all human Nav subtypes except Nav1.7, where this position is occupied by Ile. The NavPaS-STX structure provides a molecular basis for the lower affinity of STX for Nav1.7 than for Nav1.4 (38), as replacement of Asp with a hydrophobic Ile at this locus would lead to a loss of electrostatic interactions with the carbamoyl amine of STX (Fig. 4, B and C).

Molecular mechanism for pore blockade by TTX and STX

The cryo-EM structures of NavPaS in complex with TTX or STX reveal the details of their interaction with the channel. However, elucidation of their mechanism of action also requires a molecular understanding of Na+ permeation through the SF. Our recent molecular dynamics (MD) simulation of the pore domain of NavPaS suggests a preferred path involving the acidic residues from repeats I and II (fig. S10) (39). Examination of the EM map for the NavPaS-Dc1a complex, which was purified in the presence of 300 mM NaCl, identified a strong density encaged by three acidic residues, Asp375 and Glu701 from the DEKA motif and Glu704 on P2II, which is positioned above Glu701 (Fig. 5A). Coordination of this density is nearly identical to that observed in the 2.6-Å reconstruction of the NavPaS-Dc1a-TTX complex purified in 150 mM NaCl (fig. S4C). The stable conformation of the three carboxylate side chains suggests that they may be stabilized by a cation. The density therefore likely belongs to a Na+ ion rather than a water molecule. In addition, this site coincides with the energetic minimum observed in the MD simulation of Na+ penetration through the SF of NavPaS. We therefore assigned a Na+ ion to this density and refer to this Na+ binding site as the “DEE site” (Fig. 5A).

Fig. 5 Molecular mechanism for pore blockade by TTX/STX.

(A) Potential Na+ binding site within SF. The carboxylate groups of DE and an invariant Glu on P2II together constitute a potential Na+ binding site (designated the DEE site). Left: A density surrounded by three carboxylate groups may belong to a bound Na+ ion in the structure of NavPaS-Dc1a, which was purified in the presence of 300 mM NaCl. A similar density is observed in the EM map for NavPaS-Dc1a-TTX. Please refer to fig. S4C. Middle and right: Asymmetric coordination of the Na+ ion within the SF vestibule. Side and top views of the SF vestibule are shown. Repeat IV is omitted in the side view for visual clarity. The carbonyl oxygens that constitute the potential inner site for Na+ within the SF are shown as thin sticks. The structural observation is consistent with a recent MD simulation analysis of the Na+ permeation path (39). See fig. S10 for details. (B) TTX or STX completely blocks access of Na+ to the DEE site from the extracellular side. Left: The asymmetric chemical composition of the SF outer vestibule determines the permeation path for Na+. A semitransparent presentation of the electrostatic surface potential of the entrance to the SF viewed from the extracellular side is shown. The residues that give rise to the surface feature are shown in sticks. TTX and STX are shown as silver and black sticks, respectively. Middle and right: Relative position of TTX or STX with respect to the bound Na+ ion. Placement of TTX or STX at the outer mouth to SF prevents the access of Na+ to the DEE site from the extracellular side.

The preference of Na+ for the DEE site can be explained by the distinct chemical compositions of the four repeats (Fig. 5B, left). The invariant Arg on P2II (Arg696 in NavPaS), the Lys in the DEKA motif on repeat III, and a hydrophobic residue on P2III (Leu1064 in NavPaS and Met in human Nav channels) may together repel cations to the DEE site (Fig. 5B and fig. S10). Placement of TTX or STX at the entrance to the SF vestibule preserves the configuration of the DEE site but completely blocks access of Na+ to this site from the extracellular milieu (Fig. 5B).

SUMMARY

In this study, we report the structures of a eukaryotic Nav channel, NavPaS, in complex with three natural toxins. The structures of NavPaS in complex with the well-characterized neurotoxins TTX and STX provide a molecular explanation for a wealth of functional studies (3948). It is noteworthy that the molecular weights of TTX and STX are both around 300 Da. The clear resolution of these small molecules bound to a Nav channel with datasets collected in just a few days showcases the power of cryo-EM, which is likely to play an increasingly important role in structure-aided drug discovery.

The structure of the NavPaS-Dc1a complex confirmed the important role of VSDII in binding this GMT. However, it also revealed that the network of intermolecular interactions is much more complex than previously anticipated, with key interactions between the toxin and both the S5III pore-domain helix and the extracellular dome above the pore. Thus, one has to apply caution when using isolated Nav channel VSDs for drug discovery or for understanding the molecular basis of GMT action. Last, the recently determined structure of the EeNav1.4-β1 complex revealed that the extracellular dome provides a docking site for β subunits (49), which, together with the current structure, might explain why the sensitivity of Nav channels to some GMTs is modulated by the presence of an accessory β subunit (50).

Materials and methods

Purification of NavPaS in complex with toxins

Recombinant NavPaS and Dc1a proteins were expressed and purified as reported (10, 34). To assemble the complex, Dc1a (40 μM) was added to the concentrated NavPaS solution and incubated at 4°C for 0.5 hours before size exclusion chromatography (SEC, Superose 6 10/300 GL GE Healthcare). For NavPaS-Dc1a complexes bound to TTX or STX, TTX (50 μM) or STX (4 μM) were respectively added to the concentrated NavPaS solution 15 min before adding Dc1a. The peak fractions of size exclusion chromatography were pooled and concentrated to approximately 2 mg/ml. For the sample without TTX or STX, 300 mM NaCl was used during purification while for the sample with TTX or STX, 150 mM NaCl was used.

Production of Dc1a analogs

Plasmids encoding Dc1a mutants were generated via PCR-based mutagenesis using a plasmid encoding wild-type Dc1a [pLIC-NSB3; (10)] as template. The DNA sequence of all mutants was confirmed by Sanger sequencing. Peptide concentrations were determined by calculating the area under the RP-HPLC peak (at 214 nm) of all analogs, then comparing these to the peak area obtained from a Dc1a standard whose concentration had been determined by amino acid analysis.

Cryo-EM data acquisition

Cryo-EM samples were prepared as described (34). In brief, aliquots (3.5 μl) of freshly purified NavPaS complex were placed on glow-discharged holey carbon grids (Quantifoil Cu R1.2/1.3), which were blotted for 3.5 s and flash-frozen in liquid ethane cooled by liquid nitrogen with a Vitrobot Mark IV (Thermo Fisher Scientific Inc.). The grids were subsequently transferred to a Titan Krios (Thermo Fisher Scientific Inc.) electron microscope operating at 300 kV equipped with Cs-corrector (Thermo Fisher Scientific Inc.), Gatan K2 Summit detector and GIF Quantum energy filter. A total of 2764, 3050 or 4539 movie stacks, for NavPaS-Dc1a complex, TTX or STX-supplemented samples respectively, were automatically collected using AutoEMation (51) with a slit width of 20 eV on the energy filter and a preset defocus range from −1.8 μm to −1.5 μm in superresolution mode at a nominal magnification of 105,000×. Each stack was exposed for 5.6 s with an exposing time of 0.175 s per frame, resulting in a total of 32 frames per stack. The total dose rate was about 48 e/Å2 for each stack. The stacks were motion corrected with MotionCor2 (52) and binned two fold, resulting in a pixel size of 1.091 Å/pixel. Meanwhile, dose weighting was performed (53). The defocus values were estimated with Gctf (54).

Image processing

The procedure for data processing is summarized in fig. S2. A total of 895,227, 1,506,774, or 1,211,302 particles, for NavPaS-Dc1a complex, TTX or STX-supplemented samples respectively, were automatically picked using RELION (5558) and Gautomatch (K. Zhang, www.mrc-lmb.cam.ac.uk/kzhang/). After 2D classification, a total of 838,471, 1,042,430 or 447,993 particles were selected for the NavPaS-Dc1a complex, TTX or STX-supplemented samples and subjected to global angular searching 3D classification with only one class. For each of the last several iterations of the global angular searching 3D classification, a local angular searching 3D classification was performed, during which the particles were classified into 4 classes. A total of nonduplicated 595,020, 742,093 or 378,538 particles were selected from the local angular searching 3D classification for NavPaS-Dc1a complex, TTX or STX-supplemented samples, respectively. For the NavPaS-Dc1a complex and STX-supplemented samples, the particles were subjected to multi-reference classification to remove bad particles. Then the good particles were subjected to 3D auto-refinement with 255,265, 742,093, or 166,805 particles for NavPaS-Dc1a complex, TTX or STX-supplemented samples, respectively. The 3D auto-refinements were further optimized with a larger box size of 320 pixels and local defocus values determined with Gctf. Finally, a local mask was applied during postprocessing to improve the local resolution of the pore domain. 2D classification, 3D classification and auto-refinement were performed with RELION 2.1. The resolution was estimated with the gold-standard Fourier shell correlation 0.143 criterion (59) with high resolution noise substitution (60).

Model building and structure refinement

Model building was first carried out based on the 2.8-Å reconstruction map of the NavPaS-Dc1a complex. The structures of NavPaS and Dc1a (PDB accession codes: 5X0M and 2MI5, respectively) were fitted into the EM map by CHIMERA (61). Afterwards, the fitted models were manually adjusted in COOT (62).

In total, 1380 residues were built with 1272 side chains assigned for the structure. In addition, 7 sugar and 2 lipid moieties were assigned. The intracellular I-II linker, II-III linker, the N-terminal sequence preceding the NTD, and the C-terminal segment following the CTD were not modeled due to the lack of corresponding densities.

The model of NavPaS-Dc1a-STX complex was generated using the structure of the NavPaS-Dc1a complex as the starting model, which was fitted into the 3.2-Å 3D reconstruction map. 3D conformer of STX (PubChem CID: 37165) was processed with phenix.elbow application in PHENIX (63) and the resulted structure can be fitted into the map unambiguously in COOT. The docked models and residues were manually adjusted in COOT.

The model of NavPaS-Dc1a-TTX complex was generated using the structure of the NavPaS-Dc1a-STX complex as a starting model, which was fitted into the 2.6-Å 3D reconstruction map. The 3D structure of TTX (PubChem CID: 11174599) was used to replace STX. Every residue was manually checked in COOT.

Structure refinement was performed using the phenix.real_space_refine application in PHENIX in real space with secondary structure and geometry restraints to prevent structure over-fitting. Over-fitting 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 (64) (fig. S1D). Statistics of the map reconstruction and model refinement can be found in table S1.

Electrophysiology

Channel mutants were generated using PCR-based mutagenesis with NavBg (65) as template, then confirmed by DNA sequencing. Fragments from these mutant clones were excised and cloned back into the original NavBg containing plasmid to produce final mutant constructs. The DNA sequence of all constructs was confirmed by Sanger sequencing and cRNA synthesized using T7 polymerase (mMessage mMachine kit, Life Technologies, USA) after linearizing the DNA with Not I restriction enzyme. Xenopus laevis oocytes were injected with cRNA (0.5–4 ng depending on the channel) encoding wild-type or mutant NavBg together with the TipE subunit (66) (1:5 molar ratio), then they were incubated at 17°C in ND96 solution (in mM: 96 NaCl, 2 KCl, 1.8 CaCl2, 2 MgCl2 and 5 HEPES; pH 7.6) supplemented with 5 mM pyruvic acid, 50 μg/ml gentamicin sulfate, and 2.5% horse serum. Currents were recorded 1 to 3 days after injections using the two-electrode voltage-clamp technique (Axoclamp 900A, Molecular Devices, USA) with a 30 μl recording chamber. Microelectrodes were filled with 3 M KCl, and resistances were 0.5 to 1 MΩ. All experiments were performed at room temperature (~21°C) in ND96 solution containing 0.1% fatty acid–free bovine serum albumin to prevent adsorption of peptides to plastic. After addition of peptides to the recording chamber, equilibration between peptide and channel was monitored using weak depolarizations elicited at 5 s intervals. For all recordings, voltage-activation relationships were recorded in the absence and presence of peptide. To determine conductance-voltage relationships, oocytes were held at −90 mV and depolarized in 5-mV steps from −90 mV to +30 mV for 50 ms. Data were digitized at 20 kHz; leak and background conductance were identified by blocking channels with TTX and subtracting background currents. Data analyses were performed using Clampfit 10.5 (Molecular Devices, USA) and Prism 7 (GraphPad Software, USA).

SUPPLEMENTARY MATERIALS

REFERENCES AND NOTES

Acknowledgments: We thank X. Li (Tsinghua University) for technical support during EM image acquisition. We thank K. Dong (Michigan State University) for sharing BgNav1 (NavBg) and TipE constructs. Funding: This work was funded by the National Key Basic Research (973) Program (2015CB910101 to N.Y.) and the National Key R&D Program (2016YFA0500402 to N.Y. and 2016YFA0501100 to J.L.) from Ministry of Science and Technology of China, the National Natural Science Foundation of China (projects 31621092, 31630017, and 31611130036 to N.Y.), the Australian Research Council (DP160104411), and the Australian National Health and Medical Research Council (Principal Research Fellowship and Program Grant APP1072113 to G.F.K.). 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. N.Y. is supported by the Shirley M. Tilghman endowed professorship from Princeton University. Ethics statement: This study was carried out in accordance with the recommendations in the Australian code of practice for the care and use of animals for scientific purposes (8th ed., 2013). The protocol for Xenopus laevis studies was approved by the Animal Ethics Committee of The University of Queensland (approval number QBI/059/13/ARC/NHMRC). Surgeries for harvesting X. laevis oocytes with recovery endpoint were performed under anesthesia [animals exposed to tank filled with MS-222 (1.3 mg/ml)], with a minimum of 3 months between surgeries. Author contributions: N.Y., G.F.K., and Q.Z. conceived the project. H.S., Q.Z., Z.L., Y.J., X.P., J.W., B.C.-A., J.J.S., Y.K.Y.C., and J.L. performed the experiments. All authors contributed to data analysis. N.Y., G.F.K., Q.Z., and H.S. wrote the manuscript. Competing interests: The authors declare no competing interests. Data and materials availability: The atomic coordinates for NavPaS-Dc1a complex, NavPaS-Dc1a-TTX complex, and NavPaS-Dc1a-STX complex have been deposited in the Protein Data Bank (www.rcsb.org) with accession codes 6A90, 6A95, and 6A91, respectively. The EM maps have been deposited in EMDB (www.ebi.ac.uk/pdbe/emdb/) with accession codes EMD-6995, EMD-6997, and EMD-6996.
View Abstract

Navigate This Article