Research Article

Structural basis of α-scorpion toxin action on Nav channels

See allHide authors and affiliations

Science  22 Mar 2019:
Vol. 363, Issue 6433, eaav8573
DOI: 10.1126/science.aav8573

How activation leads to gating

Voltage-gated sodium (Nav) channels are key players in electrical signaling. Central to their function is fast inactivation, and mutants that impede this cause conditions such as epilepsy and pain syndromes. The channels have four voltage-sensing domains (VSDs), with VSD4 playing an important role in fast inactivation. Clairfeuille et al. determined the structures of a chimera in which VSD4 of the cockroach channel NavPaS is replaced with VSD4 from human Nav1.7, both in the apo state and bound to a scorpion toxin that impedes fast activation (see the Perspective by Chowdhury and Chanda). The toxin traps VSD4 in a deactivated state. Comparison with the apo structure shows how interactions between VSD4 and the carboxyl-terminal region change as VSD4 activates and suggests how this would lead to fast inactivation.

Science, this issue p. eaav8573; see also p. 1278

Structured Abstract


Members of the voltage-gated sodium (Nav) channel family are critical contributors to electrical signaling. Accordingly, they are targets of drugs, toxins, and mutations that lead to disorders such as epilepsy (Nav1.1 to Nav1.3 and Nav1.6), pain syndromes (Nav1.7 to Nav1.9), and muscle paralysis (Nav1.4 and Nav1.5). Nav channels contain four peripheral voltage-sensing domains (VSD1 to VSD4), which regulate the functional state of a central ion-conducting pore. Fast inactivation is an essential process that rapidly terminates Na+ conductance, allowing excitable cells to repolarize and Nav channels to become available for reopening. Mutations that disrupt fast inactivation can cause devastating disease. Although the intracellular domain III-IV (DIII-DIV) linker and voltage-dependent conformational changes in VSD4 are known to be important for fast inactivation, structural details underlying the mechanism remain unclear owing to technical challenges. In this study, we used a potent α-scorpion neurotoxin, AaH2, that is known to target VSD4 to impede fast inactivation. We present cryo–electron microscopy (cryo-EM) structures of a hybrid Nav1.7-NavPaS (human-cockroach) channel with and without AaH2 bound to illuminate the pharmacology of α-scorpion toxin action on Nav channels and gain insights into fast inactivation.


For structural studies, we grafted the α-scorpion toxin receptor site from Nav1.7 onto the cockroach NavPaS channel chassis to ease challenges of producing human Nav channels. Specifically, we replaced VSD4 and a portion of the DI pore of NavPaS with related sequences from the human Nav1.7 channel. This protein engineering strategy permitted robust expression, purification, and complex formation between AaH2 and the Nav1.7-NavPaS chimeric channel. After cryo-EM structure determination of AaH2-bound and apo-Nav1.7-NavPaS channels to 3.5-Å resolution, we utilized traditional electrophysiological techniques to probe structure-function relationships in the related BgNav1 (cockroach), human Nav1.5 (cardiac subtype), and human Nav1.7 (peripheral nervous system) channels.


AaH2 wedges into the extracellular cleft of VSD4 to trap a deactivated state, analogous to a molecular stopper. Pharmacological trapping of VSD4 reveals state-dependent interactions of gating charges from the S4 helix and S4-S5 linker that bridge to acidic residues on the intracellular C-terminal domain (CTD). Our apo-Nav1.7-NavPaS channel structure uncovers a large S4 translation (~13 Å) during VSD4 activation as a key molecular event leading to unlatching of the CTD and the fast-inactivation gating machinery. Analyses of structure-guided mutations in the BgNav1, Nav1.5, and Nav1.7 channels recapitulate human disease-causing mutations and suggest that AaH2 has stabilized the fast-inactivation machinery of the Nav1.7-NavPaS channel in a potential resting state.


Cryo-EM was used to visualize AaH2 in complex with the classic neurotoxin receptor site 3 on a hybrid eukaryotic Nav channel. Mechanistically, AaH2 traps VSD4 in a deactivated state, revealing an unanticipated interface through which DIV gating charges can couple to the CTD, DIII-DIV linker, and fast-inactivation gating machinery. We outline a structural framework that sheds light on the distinctive functional specialization of VSD4 and provides a deeper understanding of voltage sensing, electromechanical coupling, fast inactivation, and pathogenic mutations in human Nav channels. The pharmacology of α-scorpion toxins is further illuminated through an unexpected receptor site on VSD1 and pore-glycan interaction adjacent to VSD4.

Cryo-EM structures of a human-cockroach hybrid Nav channel in the presence and absence of the α-scorpion toxin AaH2.

(A) View of the AaH2-Nav1.7-NavPaS channel complex highlighting AaH2 (purple), VSD4 (green), gating charges (blue), the DIII-DIV linker (teal), CTD acidic residues (red), and the DI pore glycan (white). (B) Alternate view of the AaH2-channel complex [colored as in (A)] with the apo-Nav1.7-NavPaS channel structure (orange) superimposed. In the magnified view, the VSD4-based superposition highlights the extent of AaH2-induced translation of the S4 helix (AaH2 omitted for clarity).


Fast inactivation of voltage-gated sodium (Nav) channels is essential for electrical signaling, but its mechanism remains poorly understood. Here we determined the structures of a eukaryotic Nav channel alone and in complex with a lethal α-scorpion toxin, AaH2, by electron microscopy, both at 3.5-angstrom resolution. AaH2 wedges into voltage-sensing domain IV (VSD4) to impede fast activation by trapping a deactivated state in which gating charge interactions bridge to the acidic intracellular carboxyl-terminal domain. In the absence of AaH2, the S4 helix of VSD4 undergoes a ~13-angstrom translation to unlatch the intracellular fast-inactivation gating machinery. Highlighting the polypharmacology of α-scorpion toxins, AaH2 also targets an unanticipated receptor site on VSD1 and a pore glycan adjacent to VSD4. Overall, this work provides key insights into fast inactivation, electromechanical coupling, and pathogenic mutations in Nav channels.

Voltage-gated sodium (Nav) channels initiate and propagate action potentials in excitable cells (13). After voltage-dependent opening, Nav channels undergo fast inactivation on the millisecond time scale to terminate Na+ conductance (3, 4). Fast inactivation is an essential hallmark of Nav channel function that allows cells to repolarize and Nav channels to become readily available for reactivation (1, 3, 5). Venomous animals have evolved an arsenal of toxins that disrupt this process to immobilize prey or predators, highlighting the importance of fast inactivation across the animal kingdom (6, 7). In humans, mutations that even subtly affect fast inactivation can cause epilepsy, cardiac arrhythmias, muscle disorders, or pain syndromes (2, 3).

In 1952, Hodgkin and Huxley postulated that Nav channels contain three “m” gating particles responsible for channel opening and one slower “h” gating particle associated with inactivation (8). Nav channels are now known to contain four homologous repeat domains (DI to DIV) in which four voltage-sensing domains (VSDs) surround a central pore module (PM) in a domain-swapped arrangement (912). Voltage-sensing domain IV (VSD4) has been established to play a specialized role in initiating fast inactivation (1317), and toxins that target the extracellular surface of VSD4 modify this process (1820). Mutagenesis studies identified an isoleucine-phenylalanine-methionine (IFM) motif within the intracellular DIII-DIV linker as an important part of the fast-inactivation gating machinery, leading to a conceptual hinged-lid model of fast inactivation (21, 22). Fast inactivation remains enigmatic, however, because the structural mechanism of coupling between VSD4 and the DIII-DIV linker has not been described; therefore, how toxins or disease mutations might modify this process also remains unknown.

Nav channels share a conserved architecture with other voltage-gated ion channels (23, 24). The VSDs are four-helix bundles (S1 to S4) that sense membrane potential using positively charged residues found in a repeating arginine-X-X (RXX) motif along the S4 helix (9, 25, 26). The S5 and S6 helices from each homologous domain together form the central PM (10, 24). Upon membrane depolarization, VSDs activate through the outward movement of S4-gating charges across a narrow hydrophobic constriction (25, 27, 28), which couples through intracellular S4-S5 linkers to dilate the S6 PM bundle crossing and initiate Na+ conductance (2931). Studies tracking S4 movements indicate that rapid VSD1 to VSD3 activation is requisite to open the central PM gate, where VSD4 activation is slower but necessary and sufficient to initiate fast inactivation (15, 17). Sequence asymmetry has permitted the functional specialization of Nav channel VSDs, but the basis for their asynchronous activation remains unclear (1517, 32, 33). Indeed, direct observation of the structural transitions correlated with voltage sensing, electromechanical coupling, and fast inactivation remains technically challenging owing to the negative membrane potentials required to maintain VSDs in a deactivated state.

Peptide toxins that target VSDs have been useful tools for elucidating the complex gating properties of Nav channels because they can alter the stability of open, closed, or inactivated states (3436). In particular, α-scorpion toxins impede Nav channel fast inactivation to sustain sodium influx, causing prolonged action potentials and decreased firing frequency in vivo, which lead to paralysis, cardiac arrhythmia, or death (6, 7, 37, 38). Only minute quantities of toxin II (AaH2) from the Androctonus australis Hector “man killer” scorpion are required to impart a lethal dose (39, 40). AaH2 is a 64–amino acid peptide stapled by four disulfide bonds into a compact β1-α1-β2-β3 scaffold that can modify multiple mammalian Nav channel subtypes with high potency (40, 41). Historically called neurotoxin receptor site 3 (42), the binding site for α-scorpion toxins has been mapped across the extracellular surface of VSD4 and onto the adjacent DI PM (4345). Functionally, α-scorpion toxins delay ~30% of total gating charge movement in Nav channels, presumably by trapping VSD4 in a deactivated state (14, 18). Thus, clarifying the details of AaH2 action on Nav channels should provide insights into fast inactivation and routes to design new channel modulators.

Structure determination of eukaryotic Nav channels has become possible with breakthroughs in cryo–electron microscopy (cryo-EM) (46). The cockroach NavPaS and eel Nav1.4 channels provided the first high-resolution structural templates (10, 11); however, NavPaS remains recalcitrant to functional recordings, and eel Nav1.4 was obtained from native sources. A recent human Nav1.4 channel structure has confirmed the utility of NavPaS and eel Nav1.4 as surrogate structural models but also highlighted the challenges of producing neuronal Nav channel subtypes, like Nav1.7 (12, 47). Human genetic studies have identified loss-of-function mutations in Nav1.7 that result in congenital insensitivity to pain (48, 49). These findings have motivated efforts to develop Nav1.7-selective inhibitors that could overcome the liabilities of opioid analgesics (50). However, clinically available Nav channel inhibitors lack molecular selectivity, and the discovery of subtype-selective antagonists remains challenging (51). Focusing on Nav1.7 as a model system, we humanized VSD4 of NavPaS to capture an AaH2 toxin–channel complex at high resolution. Our data reveal the structural basis of α-scorpion toxin action on Nav channels, which provides insight into the mechanisms of voltage sensing, electromechanical coupling, and fast inactivation operating in Nav channels.

AaH2 action on Nav channels

To assess the activity of AaH2 purified from the venom of the Androctonus australis Hector scorpion, we first measured its effect (median effective concentration, EC50) on human Nav1.2 and Nav1.7 channels at a membrane potential that stabilizes the closed state (−100 mV). AaH2 impaired fast inactivation at doses expected for α-scorpion toxins, for which the potency of modulation highlights the potential of AaH2 to target neurotoxin receptor site 3 across multiple Nav channel subtypes (Fig. 1, A and B; EC50 of 0.72 ± 0.59 and 51.7 ± 1.5 nM for Nav1.2 and Nav1.7, respectively). AaH2 started to inhibit peak currents of Nav1.2 and Nav1.7 at higher toxin concentrations (Fig. 1, A and B). AaH2 action on Nav1.7 was state-dependent, as expected, with the toxin being ~100-fold less potent at holding potentials that stabilize VSD4 in an activated conformation (Fig. 1C).

Fig. 1 Characterization and cryo-EM structure of the AaH2-VSD4-NavPaS channel complex.

(A) Sodium current traces from human Nav1.7 measured during a 0-mV pulse from a holding potential of −100 mV (black) with different doses of AaH2 (purple and green). (B) EC50 measurements of AaH2 from human Nav1.2 (gray) and human Nav1.7 (black) channels measured at a holding potential of −100 mV. (C) EC50 measurements of AaH2 from human Nav1.7 channels measured at a holding potential of −120 mV (black) or −40 mV (gray). (D) Schematic of the Nav1.7 VSD4–DI-S5–NavPaS channel. The DIII-DIV linker is shown in cyan, and portions humanized to the Nav1.7 sequence are shown in green. N-terminal domain (NTD) and CTD are indicted. (E) Differential scanning fluorimetry of purified WT-NavPaS and VSD4-NavPaS channels (shown here are data for the VSD4–DI-S5–NavPaS construct) in the absence or presence of GX-936 or AaH2. dλ330/350, first derivative of the fluorescence ratio change. (F) Side view (sectioned) of the single-particle cryo-EM reconstruction of the AaH2–VSD4–DI-S5–NavPaS channel complex (hereafter called the VSD4-NavPaS channel). (G) Top view of the single-particle cryo-EM reconstruction of the AaH2-VSD4-NavPaS channel complex.

Generation of an AaH2-Nav1.7-NavPaS channel complex

Despite our interest in Nav1.7-selective modulators (5055), studies of AaH2 in complex with VSD4 were initially unsuccessful because purification of full-length human Nav1.7 remains challenging (12, 47). We considered a protein engineering approach because VSDs that are transferred onto related channels can retain their pharmacological and structural properties (23, 56, 57). Indeed, we previously used this approach to visualize VSD4 in complex with the Nav1.7-selective antagonist GX-936 in the context of a human-bacterial Nav channel chimera (52). We reasoned that the cockroach NavPaS channel (10) might facilitate the production of a Nav1.7-NavPaS hybrid channel suitable to allow visualization of an AaH2 complex.

Replacing VSD4 of NavPaS with VSD4 from human Nav1.7 enabled high-yield production of a chimeric channel (Fig. 1D and fig. S1, A to C). Because Na+ currents could not be recorded from this construct, we used differential scanning fluorometry and the potent small-molecule inhibitor GX-936 to confirm reconstitution of a relevant Nav1.7 receptor site (Fig. 1E and fig. S1D) (52, 58). However, in contrast to GX-936, AaH2 did not substantially thermostabilize this VSD4-NavPaS chimeric channel (fig. S1D). On the basis of neurotoxin receptor site 3 mapping studies and the domain-swapped architecture of Nav channels (1012, 4345), we grafted the DI-S5 helix from human Nav1.7 onto a second chimeric construct (Fig. 1D). AaH2 appreciably thermostabilized this optimized VSD4–DI-S5–NavPaS channel construct and coeluted with the purified protein as a complex (Fig. 1E and fig. S1D). We purified this VSD4–DI-S5–NavPaS chimeric channel (hereafter called VSD4-NavPaS) in the detergent digitonin supplemented with cholesteryl hemisuccinate and AaH2 and obtained a single-particle cryo-EM reconstruction of the toxin-channel complex with a nominal resolution of ~3.5 Å at the AaH2-VSD4 receptor site (figs. S2, A and B, and S3 and table S1). To enable a direct comparison, we also analyzed single-particle cryo-EM reconstruction of a similarly purified apo-VSD4-NavPaS chimeric channel, in the absence of AaH2 toxin, at a nominal resolution of ~3.4 Å at the VSD4 receptor site (fig. S2, C and D, and table S1).

Overall structure of the AaH2-VSD4-NavPaS channel complex

The AaH2-VSD4-NavPaS channel complex resembles a four-point star, in which two tips of the star are occupied by toxin (Fig. 1, F and G, and Movie 1). AaH2 is wedged into an extracellular cleft formed by VSD4 and the DI PM (Fig. 1F), where it binds to VSD4 through contacts across the S1-S2 and S3-S4 loops (Fig. 2, A and B). AaH2 utilizes C-terminal residues Arg62 and His64 to engage the DI PM, while N-terminal residues contact a glycan moiety that decorates the DI-pore turret loop structure (Fig. 3, A and B). The overall architecture of the AaH2-VSD4-NavPaS complex agrees with expectations from neurotoxin receptor site 3 mapping studies (4345), although participation of a PM glycan was not previously appreciated.

Movie 1. AaH2 receptor sites on the Nav1.7 VSD4-NavPaS channel.

Views of the molecular interactions between AaH2 and the Nav1.7 VSD4-NavPaS chimeric channel.

Fig. 2 Activation and gating-charge transfer in VSD4.

(A) Side view of VSD4 from the apo-VSD4-NavPaS channel showing the ENC (red), INC (red), HCS (orange), and gating charges (R1 to R5 and K6, blue). Shown on the right are the positions of S4 gating charges relative to the AaH2-VSD4-NavPaS channel complex, with approximate distances between “equivalent C atoms” indicated for reference. (B) Side view of VSD4 from the AaH2-VSD4-NavPaS channel showing AaH2 (pink) and gating charges (R1 to R5 and K6, blue). R5 gating charge interactions with the intracellular CTD (gray) are shown. (C) Top view of the superposition of apo-VSD4 (gray) and AaH2-VSD4 (green). The R1 to R5 and K6 gating charges are rendered as spheres (but only to Cβ for clarity); R3 in both models has been colored blue for reference. Phe1547 from the HCS is shown as sticks; S3-S4 loops and AaH2 have been omitted for clarity. (D) Side view with S3-S4 loops shown, otherwise similar to (C). (E) Close-up view from the area outlined by dashed lines in (D). Phe1583 (S3) and Glu1589 side chains are shown for reference, and AaH2 is omitted for clarity. Single-letter abbreviations for the 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.

Fig. 3 AaH2 in complex with neurotoxin receptor site 3.

(A) Side view of the AaH2-VSD4 receptor site complex. Structural elements of AaH2 are indicated, and Arg62 and His64 side chains are shown as sticks for reference. (B) AaH2 interface with the DI PM and PM glycan. Dashed lines indicate likely direct-bonding interactions. (C) AaH2 interactions with the S1-S2 loop region. (D) AaH2 interactions with the S3-S4 loop region. (E) Superposition of an unbound AaH2 crystal structure [orange; Protein Data Bank (PDB) 1PTX] with the AaH2-VSD4-NavPaS channel complex (purple) reveals differences in the CTS region upon Nav channel binding; here, the channel is omitted for clarity. (F) Structure of AaH2 bound to a neutralizing antibody 4C1 fragment (PDB 4AEI). Heavy and light chains are shown in blue and yellow surface coloring, respectively. (G) EC50 measurements of AaH2 from human Nav1.7 channels. Shown are WT Nav1.7 (gray), mutants in the DI PM (light orange), the S1-S2 loop region (blue), and the S3-S4 loop region (green), colored according to the structure model represented in (A). Error bars represent mean ± SEM. (H) EC50 measurements of synthetic WT and mutant AaH2 toxins tested on the WT Nav1.7 channel. Schematics of non-alanine side-chain chemistries are shown for reference. Error bars represent mean ± SEM.

Unexpectedly, a second AaH2 toxin is seen bound to VSD1 (Fig. 1G). Density for this AaH2 molecule is less defined than that for the toxin that engages VSD4 (Fig. 1G and fig. S2B), suggesting a lower-affinity AaH2-VSD1 interaction. Because α-scorpion toxin binding to VSD1 has not been previously reported, we carefully refined an AaH2 model into this density (fig. S3) and closely examined our apo-VSD4-NavPaS chimeric channel structure (figs. S1, A and B, and S2, C and D). The availability of toxin-bound and apo-VSD4-NavPaS channel structures provides a rare opportunity to evaluate the consequences of AaH2 binding, including the direct observation of VSD4 activation.

Activation and gating-charge transfer in VSD4

The structural basis of voltage sensing in VSD4 has remained elusive but can now be appreciated by comparing AaH2-bound and apo-VSD4-NavPaS channel structures (Fig. 2, A to E, and Movie 2). In the apo-VSD4-NavPaS channel, VSD4 is found in an activated conformation (Fig. 2A). Typical of most VSDs visualized at 0-mV potential (fig. S4, A to C) (1012, 23, 24, 52, 59), the R5 (i.e., fifth) gating charge is found just beneath the conserved S2 phenylalanine (Phe1547) within the intracellular vestibule (Fig. 2A and fig. S4B). Physiological measurements suggest that when R5 is housed in this location and the R1 to R4 gating charges reside above the hydrophobic constriction site (HCS), VSD4 is in an activated conformation (25, 52, 59). Superposition of the available GX-936–VSD4–NavAb structure corroborates this assignment because GX-936 is a potent antagonist known to stabilize Nav1.7 VSD4 in an activated state (fig. S4C) (52). By contrast, the AaH2-VSD4-NavPaS channel structure reveals VSD4 in a deactivated state because only the R1 and R2 gating charges are located above the HCS within the extracellular vestibule (Fig. 2, B to D, and fig. S4D). Successful pharmacological trapping of the deactivated conformation of VSD4 under solution conditions likely reflects the high potency of AaH2 for Nav1.7 and our chimeric channel construct (Fig. 1, C and E).

Movie 2. VSD4 activation in the Nav1.7 VSD4-NavPaS channel.

Trajectory (morph) between the AaH2-bound and AaH2-free (apo) structures of the VSD4-NavPaS channel. AaH2 is not shown for clarity.

Cataloging the S4 gating charge locations provides insight into voltage sensing because the S4 helix translates ~13 Å (equivalent Cα-backbone atoms) between the apo-VSD4–activated and AaH2-VSD4–deactivated states (Fig. 2, A to D, and Movie 2). Each S4 arginine side chain relocates two positions along the gating charge–transfer pathway, where R3 and R4 toggle between the extracellular negative-charge cluster (ENC) and the intracellular negative-charge cluster (INC) (Fig. 2, A to D). Porting of these two gating charges across the HCS is consistent with physiological measurements where α-scorpion toxins immobilize ~30% of total Nav channel gating currents (14, 18), equivalent to the movement of about two elementary charges across the membrane electric field. In the apo-VSD4–activated state (Fig. 2A), R1 extends toward the DI-S5 helix dipole; R2 and R3 engage Gln265 (DI-S5) and approach Glu1524 on the S1 helix; R4 interacts with Asn1517 (S1) and potentially Asn1540 (S2) above the HCS; R5 pairs with Asp1571 (S3) below the HCS; and gating charge K6 aligns along Trp1567 (S3), a side chain conserved in Nav channel VSDs. In comparison, in the AaH2-VSD4–deactivated state (Fig. 2B), R1 pairs with Asn1540 (S2); R2 bonds with Asn1517 (S1) above the HCS; R3 is coordinated to Glu1550 (S2) and Asp1571 (S3) below the HCS; R4 aligns between Asp1507 (S1) and Trp1567 (S3); R5 extends beyond the vestibule of VSD4 toward Glu1431-NavPaS on the intracellular C-terminal domain (CTD); and K6 similarly reaches to Gln1461-NavPaS on the CTD. Overall, the apo-VSD4–activated and AaH2-VSD4–deactivated state structures appear to represent physically reasonable models of activation and gating-charge transfer in VSD4 (Fig. 2, A to E, and fig. S4, A to D).

Superposition of apo-VSD4–activated and AaH2-VSD4–deactivated state structures illustrates the mechanics of VSD4 activation in Nav channels (Fig. 2, C to E, and Movie 2). In the AaH2-VSD4–deactivated state, the S4 helix exists mainly in a 310-helical conformation, which serves to align gating charges for activation through the central gating pore (Fig. 2, B and C). Notably, we find the R5 gating charge positioned to interact with an acidic surface patch on the intracellular CTD (Fig. 2B). In the apo-VSD4–activated state, the S4 remains in a 310-helical conformation across the gating pore but adopts an α-helical conformation above the HCS (Fig. 2, A, C, and D). When directly superimposed, small deviations over the S1 to S3 core region indicate that the membrane electric field works to translate S4 gating charges across the HCS with minimal global rearrangements of VSD4 during activation (Fig. 2, C to E). Phe1583 emerges as a hinge point on the S3 helix that allows the S3-S4 loop to undergo substantial displacement during VSD4 activation gating (Fig. 2, D and E, and fig. S4, A, C, and D). This S3-S4 loop displacement has clear implications for state-dependent binding of AaH2 to neurotoxin receptor site 3.

Determinants of AaH2 binding to neurotoxin receptor site 3

The structure of the AaH2-VSD4-NavPaS channel complex illuminates the molecular details of α-scorpion toxin binding, polypharmacology, and antagonism of Nav channels. Akin to a molecular doorstop, the triangular-shaped AaH2 toxin makes multipoint contacts across VSD4 and the neighboring DI PM (Fig. 3, A to D). The compact β1-α1-β2-β3 fold of AaH2 is required to wedge into the receptor site, rationalizing why its four disulfide bonds are essential for potency (60). High surface and electrostatic complementarities appear to dominate contributions to complex formation (Fig. 4, A to C), but AaH2 also targets potential subtype-selective Nav channel determinants (Fig. 3, B to D and G, and figs. S5 and S6). Consistent with a high-affinity complex, a considerable surface area of AaH2 (~712 Å2) is buried by the neurotoxin receptor site 3 interaction, where the channel interface can be divided into four regions: the DI PM, a PM glycan, the S1-S2 loop, and the S3-S4 loop (Fig. 3, A to D, and Movie 1).

Fig. 4 VSD4 voltage-sensor trapping by AaH2.

(A) Extracellular view of the AaH2-VSD4 complex, where VSD4 and the DI PM are shown in electrostatic surface representation. Arg62 from AaH2 is shown in purple stick representation for reference. (B) Same view as in (A), with an electrostatic surface of AaH2 rendered and VSD4 shown in cartoon. Select residues on VSD4 are shown in stick representation for reference. The magenta outline indicates the surface border of AaH2. (C) The apo-VSD4 structure is shown in electrostatic surface representation, in a similar view as in (A). (D) Side view of the AaH2-VSD4 complex. Only Arg62 of AaH2 and the R1 gating charge are shown in stick representations for clarity. (E) Side view of the apo-VSD4 structure with the AaH2-VSD4 complex superimposed (although VSD4-deactivated is removed for clarity). A clash between Arg62 (AaH2) and R1 (S4) side chains is indicated by dots. (F) Similar to (E), but a different view, and with AaH2 in purple surface representation. Arrows point to clashes between AaH2 and the VSD4-activated state (gray) structure.

AaH2 uses both its rigid core scaffold and loop elaborations to target neurotoxin receptor site 3. Structure-activity relationship studies have suggested that Arg62 and His64 on the C-terminal segment (CTS) of AaH2 are critical for potent modulation of Nav channels (40, 61, 62). In the AaH2-VSD4–deactivated state, Arg62 hydrogen bonds directly to the carbonyl of Gln265 (DI-S5 helix) and the side chain of Glu1589 (S3-S4 loop), while His64 forms a close interaction network with a constellation of DI PM side chains including Asn270, His273, and Gln345-NavPaS (Fig. 3, A and B). Because Gln265, Asn270, and His273 are residues grafted from human Nav1.7 onto the NavPaS channel chassis required for AaH2-mediated stabilization (fig. S1, A and D), these DI PM interactions likely lock AaH2 into a productive receptor site complex and together may represent determinants of toxin potency and selectivity (fig. S6). Comparison to available structures of AaH2 bound or unbound to a monoclonal antibody fragment further emphasizes a key role for Arg62 and His64 in targeting neurotoxin receptor site 3 (Fig. 3, E and F) (41, 63) and suggests that sequestering of Arg62 and His64 is sufficient for antibody-mediated neutralization of this lethal toxin (Fig. 3F).

The N-terminal reverse turn (RT) and CTS of α-scorpion toxins are known to be important for neurotoxin receptor site 3 binding (61, 64, 65). In AaH2, this region is wedged alongside the elaborated structure of the DI PM (Fig. 3, A and B). Asp9 (RT), Val10 (RT), and Arg56 (CTS) contact an extended PM glycan that shields a hydrophobic surface on the turret loop structure (Fig. 3, A and B, and figs. S5 and S6). This glycosylation emanates from the NavPaS channel chassis (Asn330-NavPaS) and is conserved in most mammalian Nav channel subtypes (fig. S6). Curiously, human Nav1.7 lacks this glycosylation locus (fig. S6), establishing that toxin contacts to the PM glycan are not strictly required for potent AaH2-mediated Nav channel modulation (Fig. 1B).

The S1-S2 loop region of VSD4 is contacted by the CTS and β2-β3 loop of AaH2 (Fig. 3, A and C). Tyr42 (β2-β3 loop) bonds to Gln1530 (S1-S2 loop), as Gly59-Pro60 (CTS) frames the Gln1196 side chain (Fig. 3C). The carbonyl of Ala39 (β2-β3 loop) engages Tyr1537 (S2), as Ser40-Pro41 covers the Tyr1203 side chain (Fig. 3C). Targeting of Tyr1537 by AaH2 is notable because this residue is a hotspot for Nav1.7-selective antagonists (GX-936) and Nav1.1-selective activators (Hm1a) (52, 66). However, limited side-chain engagement along the S1-S2 region also rationalizes the ability of AaH2 to modulate multiple Nav channel subtypes.

The S3-S4 loop region of VSD4 forms a state-dependent interface that is engaged by the RT, β2-β3 loop, and CTS of AaH2 (Fig. 3, A and D). Asn44 (β2-β3 loop) bonds directly to the carbonyl of Phe1583 (S3), as Phe15 and Trp38 (β2-β3 loop) make flanking hydrophobic and van der Waals interactions (Fig. 3D). Notably, Phe1583 demarks the extracellular end of the S3 helix in the deactivated state of VSD4 (Figs. 2D and 3D), a demarcation not seen in all VSD4-activated state structures (fig. S4, A and C) (11, 12, 52). These observations establish that Asn44 targets a key state-dependent interaction on VSD4. Similarly, Thr13 (RT) of AaH2 coordinates Asp1586 (S3-S4 loop), as Arg62 (CTS) and the Cys63 carbonyl (CTS) form a network with Glu1589 (S3-S4 loop), whereas Asp1586 and Glu1589 can form part of the S3 helix in available VSD4-activated state structures (fig. S4, A and C) (11, 12, 52). Overall, the structural determinants of AaH2 binding revealed here align with prior receptor site–mapping studies but provide a deeper understanding of the state dependence of toxin modulation.

AaH2 traps VSD4 in a deactivated state

The AaH2-VSD4–deactivated and apo-VSD4–activated state structures clarify the mechanism by which α-scorpion toxins impede VSD4 activation. In the deactivated state, AaH2 wedges into the extracellular cleft of VSD4 to make multipoint-bridging contacts across the VSD4 loops and DI PM (Figs. 3A and 4, A and B, and Movie 1). This invokes a simple physical mechanism to antagonize VSD4 activation: AaH2 acts as a molecular stopper to prevent outward movement of the S4 voltage sensor.

In the VSD4-deactivated state, the receptor site presents an intense electronegative surface that is matched by an electropositive surface on AaH2, indicating a role for electrostatic steering in complex formation (Fig. 4, A and B). The large footprint that AaH2 occupies across neurotoxin receptor site 3 (~712 Å2) must also drive toxin potency for the deactivated state. By contrast, the VSD4-activated state undergoes substantial conformational and electrostatic remolding upon S4 activation (Figs. 2, C to E, and 4C), rationalizing why AaH2 modulation is state-dependent (Fig. 1C).

The S4 helix undergoes considerable displacement upon VSD4 activation (Figs. 2, C to E, and 4, D to F). On superposition, a severe clash occurs between the activated S4 helix and AaH2, explaining why the S4 is translated downward (~13 Å) in the toxin-bound complex (Fig. 4, E and F). Indeed, Arg62 of AaH2 occupies a similar DI-S5 helix coordination site as the R1 gating charge in the apo-VSD4–activated state, demonstrating that the toxin physically occupies a locale visited by gating charges during VSD4 activation (Fig. 4, D and E). These observations rationalize why AaH2 action is state-dependent and raise speculation that the fat-tailed scorpion has evolved Arg62 in AaH2 as a gating charge mimetic.

The S3-S4 loop is a major binding determinant for α-scorpion toxins (Fig. 3, D and G); however, this loop undergoes marked rearrangements upon VSD4 activation (Fig. 4F and fig. S4, A, C, and D). Accordingly, AaH2 uses Thr13, Asn44, Arg62, and Cys63 to focus multipoint coordination across the S3-S4 loop while simultaneously targeting interactions that are only available in the deactivated state (Phe1583, Asp1586, and Glu1589) (Fig. 3D). AaH2 binding therefore sterically prevents the S4 helix and S3-S4 loop from undergoing the appreciable conformational changes and electrostatic remolding required for VSD4 to achieve an activated state (Fig. 4, D to F). These structural observations collectively explain the state-dependence of AaH2 and impose a voltage-sensor trapping mechanism.

Electrostatics affect toxin binding and VSD4 trapping

Guided by the AaH2-VSD4-NavPaS complex structure, we set out to probe details of AaH2 interaction with neurotoxin receptor site 3 in the human Nav1.7 channel. Alanine mutations were generated for the equivalent three DI PM and four VSD4 side chains found to contact AaH2 directly (Fig. 3, A to D and G). Only the Asp1586→Ala (Asp1586Ala) mutation appreciably shifted the EC50 (Fig. 3G and fig. S7, A and B), confirming that this acidic residue on the S3-S4 loop is a binding hotspot (45). However, the Asp1586Arg mutant Nav1.7 channel also returned a considerable EC50 for AaH2 (Fig. 3G and fig. S7B), consistent with the large surface area AaH2 uses for binding and suggesting that α-scorpion toxins have evolved to engage Nav channels robustly.

We next sought to interrogate the Nav1.7 channel receptor site through the study of AaH2 variants. The His64Ala-AaH2 toxin displayed a ~fourfold loss in potency (Fig. 3H and fig. S7C), in line with the PM single-point mutations that had only a minor impact (Fig. 3G). By contrast, the Arg62Ala-AaH2 mutant significantly reduced toxin potency, ~80-fold (Fig. 3H and fig. S7C). As citrulline (Cit) is a neutral amino acid that is nearly isosteric with arginine (67), we examined an Arg62Cit-AaH2 toxin analog and measured an intermediate ~40-fold decrease in potency (Fig. 3H and fig. S7C). Further substitutions at position 62 that increased (hArg, homoarginine) or decreased (nArg, norarginine; and Agp, 2-amino-3-guanidinopropionic acid) the span of the guanidino group by one or two methylene units relative to arginine returned nearly identical potencies to the wild-type (WT) toxin (Fig. 3H), suggesting that Arg62 in native AaH2 is well positioned to antagonize Nav channels. Overall, Arg62 is an important contributor to AaH2 potency and appears to impart an electrostatic component to help impede S4 activation and trap VSD4 in the deactivated state (Fig. 3H).

AaH2 can bind to VSD1

α-Scorpion toxin action on Nav channels has been the subject of investigation for more than 50 years (37, 38), so it was initially surprising to find a second AaH2 toxin molecule bound to the VSD4-NavPaS channel (Figs. 1G and 5, A to D, and Movie 1). The weaker density for the AaH2-VSD1 complex suggests a low-affinity interaction (Fig. 1G and fig. S2B), which may explain why it has previously gone undetected, and calls into question the physiological relevance of this primitive VSD1 receptor site. Still, understanding the determinants of AaH2 binding to VSD1 will illuminate the structural basis of toxin polypharmacology and may reveal principles to design new Nav channel modulators.

Fig. 5 AaH2 bound to a VSD1 receptor site.

(A) Side view of the AaH2-VSD1 complex. Arg62 and His64 from AaH2 are shown as sticks for reference. (B) Side view of the activated VSD1, with AaH2 removed for clarity. ENC (red), INC (red), HCS (orange), and gating charges (R1 to R4, purple) are indicated. (C) AaH2-VSD1 (orange-blue) and AaH2-VSD4 (purple-green) complexes are superimposed based only on the VSD scaffold. (D) Close-up view of the VSD1 receptor site highlighting AaH2 interactions with Tyr166 (S2) and Asp219 (S3-S4 loop). (E) AaH2 was applied at 500 nM to WT, VSD1-Asp220Arg, and VSD4-Asp1670Arg BgNav1 channels. Representative current traces from Xenopus oocytes at 0.2 Hz (upper) and 1.0 Hz (lower) shown at 35 s (orange) and 150 s (blue) after toxin application.

AaH2 binds to an activated conformation of VSD1 under our experimental conditions (Fig. 5, A and B). In the AaH2-VSD1 complex, R4 forms salt bridges with acidic residues below the HCS, permitting the R1 to R3 gating charges to project into the vestibule to engage typical ENC interactions (Fig. 5B). Global changes in VSD1 structure are not observed compared with the toxin-free apo-VSD1 structure (fig. S4E). Intriguingly, the AaH2-VSD1 complex does share some resemblance to the AaH2-VSD4–deactivated state structure, particularly along the S3-S4 loop interface (Fig. 5C).

AaH2 targets only the S3-S4 loop and Tyr166-NavPaS (S2 helix) to bury a limited surface area (~432 Å2) upon binding to VSD1 (Fig. 5, A and D). The NC domain of AaH2 remains unengaged at the VSD1 receptor site, and consequently, Arg62 and His64 do not form securing interactions to the DII PM (Fig. 5A). The lack of a robust binding interface between AaH2 and the compact DII PM structure may rationalize the evolution of α-scorpion toxins to more proficiently target Nav channels at neurotoxin receptor site 3. Beyond this speculation, whether AaH2 can lock VSD1 into an activated or deactivated state remains unknown, but AaH2 may lack sufficient multipoint contacts to impose voltage-sensor trapping at the VSD1 receptor site (Fig. 5, C and D).

To investigate the potential relevance of the VSD1 receptor site, because NavPaS remains recalcitrant to functional recordings (10, 68), we turned to the sequence-related Blatella germanica cockroach BgNav1 channel. Exemplar S3-S4 loop mutations (45) expected to perturb AaH2 binding at VSD1 and VSD4 were introduced in the BgNav1 channel and assayed in Xenopus oocytes (Figs. 3, D and G, and 5D). AaH2 (100 nM) inhibited fast inactivation of WT-BgNav1 and VSD1-Asp220Arg-BgNav1 channels in a similar manner, but the VSD4-Asp1670Arg-BgNav1 mutation eliminated the effect (fig. S8, A to D). These results align with the expected action of α-scorpion toxins on Nav channels and demonstrate that an offending mutation can abolish the VSD4 interaction.

Examined at 500 nM AaH2, a slow-onset effect of current inhibition was observed in the WT-BgNav1 channel, in addition to inhibition of fast inactivation (Fig. 5E and fig. S8, E and F). The VSD4-Asp1670Arg-BgNav1 mutation eliminated the toxin effect on fast inactivation, but the slow onset of current inhibition was still present at 500 nM AaH2 (Fig. 5E). Notably, the VSD1-Asp220Arg-BgNav1 mutation showed the opposite effect: Fast inactivation was inhibited, but the slow onset of current inhibition was abolished at 500 nM AaH2 (Fig. 5E). When the test pulse frequency was increased from 0.2 to 1 Hz, the onset of current inhibition was faster in WT-BgNav1 and VSD4-Asp1670Arg-BgNav1 channels at 500 nM AaH2 (Fig. 5E), suggesting a toxin interaction with an activated state. These physiological data substantiate the biophysical relevance of the AaH2 interactions observed in the cryo-EM structure, including the unanticipated AaH2-VSD1 receptor site complex.

DIV gating charges couple to the CTD

Fast inactivation is a hallmark of Nav channel function, but the structural basis of this process has remained uncertain. In the AaH2-VSD4-NavPaS channel structure, AaH2 has trapped VSD4 in a deactivated conformation (Figs. 2B and 6A), providing a new template to evaluate the potential electromechanical coupling mechanism of fast inactivation in Nav channels (Movie 3). In the AaH2-VSD4–deactivated state, the R5 gating charge joins with K7 and R8 from the S4-S5 linker to form an electrostatic bridge to conserved acidic residues on the α1 helix of the CTD (Fig. 6A and fig. S6). We refer to this unpredicted molecular interface as “switch 1.” If this electrostatic bridge represents a physiologically relevant interaction, then a simplistic two-step structural model for the fast-inactivation mechanism emerges (Fig. 6, A to C, and Movie 3). We first must hypothesize that the electrostatic bridge physically restrains the CTD, which in turn promotes binding of the DIII-DIV linker across the CTD and DIV-S6. We tentatively refer to these DIII-DIV linker interactions as the electrostatic latch, or “switch 2.” During step 1, VSD4 activation severs the electrostatic bridge (switch 1) to release a key physical constraint on the CTD (Fig. 6, A, B, and D), likely straining the DIII-DIV linker interactions (Fig. 6E). During step 2, increased positional dynamics of the CTD will release the electrostatic latch (switch 2), allowing the fast-inactivation gating particle (i.e., IFM motif) to bind the PM and terminate Na+ conductance (Fig. 6, B, C, and E).

Fig. 6 Proposed structural model of fast-inactivation gating in Nav channels.

(A) Side view of the AaH2-VSD4-NavPaS complex structure with R1 to R8 gating charges shown in blue, conserved acidic switch 1 (CTD) and switch 2 (DIV S6) residues in red, DIII-DIV switch 2 residues in cyan, and the IFM-like motif residues in pink. AaH2 is omitted for clarity. (B) Side view of the apo-VSD4-NavPaS structure with residues highlighted as in (A). (C) Side view of the human Nav1.4 channel cryo-EM structure (PDB 6AGF) is shown with residues highlighted as in (A). The CTD appears disordered in this structure, and the β1 subunit is omitted for clarity. (D) Close-up view of the switch 1 interactions, comparing the VSD4-deactivated and apo-VSD4–activated state structures. (E) Close-up view of the switch 2 interactions, comparing the apo-VSD4–activated and human Nav1.4 channel structures. (F) Electrophysiological characterization of human Nav1.5. Shown at the top are example traces of WT and mutant Nav1.5 channels expressed in Xenopus oocytes. Oocytes were held at −120 mV and pulsed from −80 to 40 mV for 30 ms. The graph on the bottom left shows steady-state inactivation as a function of voltage, where oocytes were subjected to a 500-ms conditioning pulse at the indicated voltage, followed by a 1-ms step to −100 mV and a 20-ms test pulse at −20 mV. The graph on the bottom right shows the rate (τ) of fast inactivation from single exponential fits of Nav activation in response to depolarization using the voltage protocol described above. Error bars represent mean ± SEM. (G) Structural model of human Nav1.5 based on the AaH2-VSD4-NavPaS channel complex, showing switch 1 and switch 2 residues (sticks) and side chains conserved in human Nav channels that are mutated in disease (spheres). Side chains of pathogenic mutations discussed in the text are indicated (R5, R8, Lys1505, and Glu1784). S1N, pre-S1 helix.

Movie 3. A potential structural model of fast inactivation in Nav channels.

Trajectory (morph) between the AaH2-bound and apo-VSD4-NavPaS channel structures combined with the human Nav1.4 channel structure (PDB 6AGF). AaH2 and β subunits are omitted for clarity, and the CTD in Nav1.4 was not modeled or depicted here.

Our AaH2-bound VSD4-NavPaS channel structure suggests that the intracellular DIV gating charges are connected to the CTD through electrostatic bridging interactions in a resting state (Fig. 6, A and D). To assess a link between this putative electrostatic bridge (switch 1) and fast inactivation, we generated mutations of conserved acidic CTD residues in the related BgNav1 channel (fig. S6) and indeed observed enhanced steady-state inactivation (SSI) and accelerated fast inactivation, among other phenotypes (fig. S9, A to C, and table S2). We next tested the functional role of this potential switch 1 interface in the human Nav1.5 cardiac channel, noting that long QT and Brugada syndrome mutations are commonly found in this region (Fig. 6, F and G) (69, 70). Charge reversal of conserved acidic CTD residues to lysine (K) (CTD-3K mutant: Asp1789Lys, Asp1792Lys, and Glu1796Lys) produced an ~8-mV left shift in the midpoint of SSI and an accelerated time constant for fast inactivation (Fig. 6F, fig. S9D, and table S2), suggesting that breaking the electrostatic bridge is associated with entry into the fast-inactivated state. Conversely, reversal of the K7 and R8 gating charges to glutamic acid (E) (KR-2E mutant: Lys1641Glu and Arg1644Glu) also caused a left shift in SSI, with a shallowed slope expected for such severe gating charge manipulation (Fig. 6F, fig. S9D, and table S2). Notably, when the gating charge and CTD mutations were combined into a single Nav1.5 construct (2E+3K mutant), we observed that SSI was right-shifted in this “charge-swapped” mutant channel relative to either single alteration alone (Fig. 6F, fig. S9D, and table S2). This result supports a functional coupling interface across the electrostatic bridging interactions observed within the AaH2-VSD4-NavPaS channel structure (Fig. 6, A and D). To explore the generality of these observations further, an analogous CTD-2K mutation (Glu1773Lys and Glu1776Lys) and R8D (i.e., R8→D) gating charge reversal were examined in the human Nav1.7 channel and found to produce similar phenotypes (fig. S9, F and G, and table S2). Interestingly, a subtle depolarizing shift in the midpoint for activation was commonly observed (fig. S9, A, E, and G, and table S2), which might arise as a result of accelerated fast inactivation (17, 71, 72). Overall, our physiological data support a model in which DIV gating charges participate in a state-dependent interface with the acidic CTD and that these electrostatic interactions appreciably modulate Nav channel activation and fast-inactivation properties.


We used single-particle cryo-EM methods to visualize the α-scorpion toxin AaH2 in complex with a chimeric Nav1.7 VSD4-NavPaS channel. AaH2 traps VSD4 in a deactivated state by wedging into the extracellular cleft to form multipoint contacts across neurotoxin receptor site 3. Arg62 and His64 of AaH2 target a constellation of residues on the DI PM that help guide the toxin into a stable binding pose (Fig. 3B). Mutagenesis studies and a neutralizing monoclonal antibody confirm an important role for these basic residues in Nav channel modulation (Fig. 3, F to H) (63). Comparison to an apo-VSD4-NavPaS channel structure establishes that the S4 helix and S3-S4 loop undergo substantial conformational changes during VSD4 activation (Fig. 2, A to E). Consequently, AaH2 sterically occludes VSD4 activation by forming a number of interactions that serve to pin the S3-S4 loop and S4 helix into a deactivated conformation (Figs. 3, A to D, and 4F). These AaH2-mediated interactions rationalize the high potency and state dependence of Nav channel modulation (Fig. 1C).

The long-chain α-scorpion toxin AaH2 is a proficient gating modifier of Nav channels. AaH2 modulates multiple Nav channel subtypes despite contacting VSD4 residues that are selectivity determinants for an emerging class of small-molecule antagonists (Figs. 1B and 5E, fig. S6, and Movie 1) (5052, 54). AaH2 embraces a highly conserved PM glycan on the VSD4-NavPaS channel (Fig. 3B), which may in part explain the increased potency of AaH2 for Nav1.2 over Nav1.7 (Fig. 1B and fig. S6). A previously unknown activated-VSD1 receptor site highlights the intrinsic affinity of AaH2 for Nav channel VSDs (Fig. 5, A and B), where AaH2 appears competent to affect BgNav1 channel gating at increased toxin concentrations (Fig. 5E). By contrast, AaH2 can potently and robustly antagonize VSD4 by trapping a deactivated state (Figs. 3G and 4, D to F), underscoring the state dependence required by α-scorpion toxins to modulate fast inactivation. Mechanistically, AaH2 is distinct from other structurally characterized gating modifiers like GX-936 and Dc1a, which directly contact S4 gating charges to trap activated conformations of their VSD targets (fig. S10, A to C) (52, 68). The multipoint contacts made by AaH2 across VSD4, however, are reminiscent of the transmembrane-linking interactions used by divalent cations to antagonize VSD activation in TPC1 and Hv1 channels (fig. S10, D and E) (7375). The details of AaH2 action and polypharmacology uncovered here provide insight into α-scorpion toxin evolution and clues to design new Nav channel modulators, although understanding the influence of β subunits will require further study (fig. S10F) (5, 76, 77).

The structural basis of voltage sensing and electromechanical coupling in Nav channels that leads to fast inactivation has remained enigmatic owing to technical challenges. Here, our use of a potent α-scorpion toxin has helped to elucidate the VSD4 activation pathway (Movie 2). The S4 helix translates ~13 Å with two gating charges (R3 and R4) transported across the central HCS (Fig. 2, A to D), consistent with prior electrophysiological measurements (14, 18, 25, 52). The overall translation of the S4 can be described as a sliding helix, although the S4 itself remains in a 310-helical conformation across the HCS, and the S1 to S3 core region undergoes a notable rigid body displacement (fig. S11A). Studies have indicated that α-scorpion toxins do not trap a true resting state of VSD4 (18), raising the possibility that we have visualized a potential deactivated intermediate. Still, in the AaH2-VSD4–deactivated state structure, the R5, K7, and R8 gating charges form a dense network of electrostatic interactions with acidic residues on the proximal surface of the intracellular CTD (Fig. 6, A and D). These previously unseen state-dependent bridging interactions may represent a view of the elusive DIV-coupling interface to the fast inactivation gating machinery.

Although it is tenuous to assign functional states to Nav channel structures (fig. S11, B and C), we postulate that the available eukaryotic Nav channel structures outline a provisional two-step structural model of the fast-inactivation mechanism, as summarized in Fig. 6, A to E, and Movie 3. We propose that the AaH2-VSD4–deactivated state structure represents an early resting state of the fast-inactivation machinery (Fig. 6, A and D). Here, the R5, K7, and R8 gating charges form state-dependent electrostatic bridging (switch 1) interactions with acidic residues on the α1 helix of the CTD, while basic residues from the DIII-DIV linker form electrostatic latch (switch 2) interactions with acidic residues on the CTD and DIV-S6 helix (Fig. 6, A and E, and fig. S6). In step 1, voltage-dependent VSD4 activation will sever switch 1 interactions to release a key physical constraint on the CTD, producing an intermediate state represented by the apo-VSD4-NavPaS channel structure (Fig. 6, A, B, and D). In step 2, VSD4-releasing of the CTD will lead to increased strain on switch 2 interactions until the IFM motif is liberated to find its PM-receptor site, as seen in the human Nav1.4 channel structure (Fig. 6, B, C, and E) (12), which will ultimately terminate Na+ conductance. Why the NavPaS channel has allowed us to capture a potential gating intermediate remains unknown (Fig. 6B), but partial activation of VSD3 or other sequence differences might provide justification (figs. S6 and S11B). Nevertheless, the NavPaS channel chassis serves to highlight the CTD as an electronegative nexus that may link VSD4 activation status to the fast-inactivation IFM-motif gating particle (Fig. 6, A, B, and D). The apo-VSD4-NavPaS and WT-NavPaS channel structures are highly similar (10), as are the human and eel Nav1.4 channel structures (11, 12), further suggesting that physiologically relevant states may be depicted in Fig. 6 and Movie 3. Overall, we have outlined a basic structural framework that may begin to describe molecular events occurring during fast inactivation in Nav channels.

We recognize that our structure-based model does not attempt to account for known differences in fast-inactivation properties among Nav channel subtypes, nor the differential impact of cell type or auxiliary subunits (e.g., β subunits or, calmodulin) (7883). However, our simplistic fast-inactivation model does begin to rationalize important literature on Nav channels. Hodgkin and Huxley first noted the slower “h” inactivation gating-particle (8), but the basis for this essential feature of electromechanical coupling in Nav channels has remained unknown. Here, the requirement for VSD4 gating charges (e.g., R5) to first break electrostatic bridging interactions with the CTD (switch 1) may begin to account for the slower activation kinetics of VSD4 relative to VSD1 to VSD3 (Fig. 6, A, B, and D) (15, 17). Since Noda, Numa, and co-workers first cloned the Nav channel (9), it has also remained unknown why DIV, uniquely among VSDs, contains eight gating charges (fig. S6) compared with the four to six gating charges found in DI to DIII. The additional K7 and R8 gating charges of DIV are now seen to participate in a dense network of electrostatic bridging contacts to the CTD (switch 1) (Fig. 6D), where these S4-S5 linker–CTD interactions may earmark DIV specialization during eukaryotic Nav channel evolution.

Numerous pathogenic mutations map to the electrostatic bridge (switch 1) and electrostatic latch (switch 2) regions defined by our fast-inactivation model (Fig. 6G), supporting the physiological relevance of these interfaces. An epilepsy syndrome (GEFS+) mutation in Nav1.1 targeting R5 (Arg1648His) predicted to weaken switch 1 interactions (Fig. 6G) produces persistent current and accelerates recovery from fast inactivation (80, 84). Another GEFS+ mutation targeting R8 (Arg1657Cys) at the switch 1 interface produces a depolarizing shift in activation and reduces channel current density (85). Long QT and Brugada syndrome mutations bombard switch 1 and switch 2 regions in Nav1.5 (69), exemplified by the DIV-S6 helix Glu1784Lys mutation predicted to destabilize the switch 2 interface (Fig. 6G). Glu1784Lys produces a hyperpolarizing shift in SSI and persistent current in Nav1.5 (83, 86), and similar alterations have been reported when acidic α1-helix residues along the CTD are neutralized (83). A detailed study of Glu1784Lys demonstrated a hyperpolarizing shift of VSD4 gating-charge movement, providing a potential integrated view of the switch 1 and switch 2 coupling interfaces defined by our fast-inactivation model (71). Mutations that perturb the DIII-DIV linker–CTD interactions also cause disease (Fig. 6G), including the ΔK1505PQ insult that produces persistent current in Nav1.5 (69, 87) (Fig. 6G). Charge reversal or neutralization of Lys1505 in Nav1.5 and Nav1.3 produces similar effects on fast inactivation (78, 88), whereas phosphorylation of Ser1506 by protein kinase C slows inactivation in Nav1.2 (89) and causes a hyperpolarizing shift of SSI in Nav1.5 (90). Although the phenotypes and magnitude of alterations may depend on the Nav channel subtype or auxiliary subunits present, the perturbations that we observe with targeted switch 1 mutations in the BgNav1, Nav1.5, and Nav1.7 channels (Fig. 6F; fig. S9, A to G; and table S2) overlap with the defects of disease-causing mutations identified in human patients.

This study details the first high-resolution snapshots of α-scorpion toxin action on Nav channels, including pharmacological trapping of VSD4 in a deactivated state, and the resolution of an unpredicted VSD1 receptor site. When combined with other recent eukaryotic Nav channel structures, our results outline a structural framework to understand the mechanisms of voltage sensing, electromechanical coupling, and fast inactivation operating in Nav channels. We have proposed a simplistic fast-inactivation model that sheds light on the functional and evolutionary specialization of DIV and also advances our understanding of many pathogenic Nav channel mutations. Overall, our study provides templates for the design of new Nav channel modulators and a foundation to demystify the structural basis of fast inactivation.

Materials and methods

Expression and purification of chimeric and wild-type NavPaS channels

Similar to the original report (10), an optimized coding DNA for NavPaS and Nav1.7 VSD4-NavPaS chimeras containing a StrepII and FLAG tag in tandem at the N terminus was cloned into a pRK vector behind a CMV promoter. HEK293 cells in suspension were cultured in SMM 293T-I medium under 5% CO2 at 37°C and transfected using PEI when the cell density reached 4 × 106 cells per ml. Transfected cells were cultured for 48 hours before harvesting.

Fifty grams of harvested cell pellet was resuspended in 250 ml of 25 mM Tris pH 7.5, 50 mM NaCl, 1 μg/ml benzonase, 1 mM PMSF and Roche protease inhibitor tablets. Cells were lysed by gentle sonication, and NavPaS channel proteins were subsequently solubilized by addition of 1% digitonin supplemented with 0.1% cholesteryl hemisuccinate (CHS) for 2 hours at 4°C under gentle agitation. Insoluble debris was pelleted by ultracentrifugation at 40,000 rpm for 45 min, and the supernatant containing the solubilized protein was collected for affinity purification by batch-binding to 5 ml of M2-agarose FLAG resin (Sigma) for 1 hour at 4°C. Unbound proteins were washed with nine column volumes (CV) of purification buffer [25 mM Tris pH 7.5, 50 mM NaCl, 0.1% (wt/v) digitonin, 0.01% CHS], followed by three CV supplemented with 500 mM NaCl, and eluted with six CV of purification buffer supplemented with 300 μg/ml FLAG peptide (Sigma). The eluent was collected and applied to 3 ml Streptactin resin by gravity (six times). Unbound proteins were washed with 10 CV of purification buffer and eluted with five CV of purification buffer supplemented with 2.5 mM desthiobiotin. NavPaS proteins were then concentrated with 100-kDa MWCO concentrators to ~6 mg/ml and injected onto a Superose 6 3.2/300 column attached to an AKTA system (GE Healthcare) for size-exclusion chromatography into purification buffer.

Purification of AaH2

Native AaH2 was purified as previously described (91). Briefly, the Androctonus australis Hector venom (10 g) was extracted at 4°C with water, centrifuged 40 min, 13,000g at 4°C. The supernatant was adjusted with 0.1 M in ammonium acetate, pH 8.5, and immediately filtered through a set of four Sephadex G-50 columns connected in series. The fraction containing AaH2 was further submitted to cation exchange chromatography on Amberlite CG-50 in 0.2 M ammonium acetate, pH 6.7. Quantification and homogeneity of the pure toxin was achieved by UV spectrum, electrophoresis on 15% polyacrylamide gel in native conditions, amino acid analysis, and by mass spectrometry analysis (7244 Da). Standard biological tests were performed to confirm biological activity of the toxin.

Differential scanning fluorimetry

Melting experiments were conducted on a Prometheus NT48 (NanoTemper technologies) by measuring the tryptophan fluorescence 330/350 nm ratio of protein samples concentrated at 0.3 mg/ml in a standard capillary. AaH2 and GX-936 were mixed with purified NavPaS proteins 30 min prior to performing the experiment.

Cryo-EM sample preparation and data acquisition

Native AaH2 was added to the VSD4-NavPaS channel at 4:1 molar ratio prior to the size-exclusion step at a final concentration of 200/50 μM. Grids were prepared in the following manner. Holey carbon grids (C-flat, R 2/1 200 mesh Cu; Protochips) were plasma etched using the Solarus plasma cleaner (Gatan) in the hydrogen-oxygen setting. Grids were etched for 4 min on each side to remove burrs from hole edges. The grids were then coated on both sides with 5 nm of Au/Pd, which was plasma deposited using the Leica ACE600 (Leica). Three microliters of the peak fraction was applied to a glow-discharged holey Au/Pd grid, incubated for 60 s, then hand blotted and another 3 μl of the peak fraction was applied. Grids were then blotted in Vitribot Mark IV (Thermofisher) using 5-s blotting time with 100% humidity, and plunge-frozen in liquid ethane cooled by liquid nitrogen. Movie stacks were collected using SerialEM (92) on a Titan Krios operated at 300 keV with bioquantum energy filter equipped with a K2 Summit direct electron detector camera (Gatan). Images were recorded at 165000× magnification corresponding to 0.849 Å/pixel, using a 20-eV energy slit. Each image stack contains 40 frames recorded every 0.25 s for an accumulated dose of ~40 e/Å2 and a total exposure time of 10 s. Images were recorded with a set defocus range of 1 to 2.5 μm.

Cryo-EM data processing

Image stacks were processed using cisTEM (93). Frame motion was corrected and contrast-transfer function (CTF) parameters were fit from movies using the 30-4 Å band of the spectrum. Images were resampled to 1 Å per pixel, and those with CTF fits to 5 Å or better were selected for particle picking using a soft-edge disc as a template (4 σ peak threshold, with maximum and characteristic radii of 90 and 65 Å, respectively). Three rounds of 2D classification with 100 classes and a box size of 250 pixels were performed to sort AaH2-VSD4-NavPaS protein particles from debris and other false positives. The remaining, 504,252 particles were subjected to a global angular search and classification into four classes, using the published NavPaS EM map (EMD-6698) low-pass filtered to 40 Å as a reference volume. Three classes, containing 433,112 particles, were selected for a local angular search and classification into three classes with the best output from the previous global search as a template. One class with the best structural features containing 251,714 particles was selected for 3D refinement. To avoid overfitting of the detergent micelle, a protein-only mask was created from the published NavPaS model and used to low-pass filter nonprotein regions of the map at 20 Å between refinement iterations (94). Iterative global and local refinement improved the proteinaceous features of the map, at which point a new mask was created from the latest map, for a final refinement run using frequencies up to 1/5.0 Å−1 and yielding a map at 3.5-Å resolution (FSC = 0.143). For model building and figure preparation, maps were sharpened in cisTEM by (i) applying a B-factor of −90 Å2 from the origin of reciprocal space to 1/10 Å−1, (ii) flattening their rotational power spectrum from 1/10 Å−1 onwards, (iii) applying a B-factor of −30 Å2 in that same range, and (iv) applying a figure-of-merit filter (95). Maps were subsequently filtered according to the local resolution, using reimplementation of blocres (96) within cisTEM and a modified resolution criterion (Rohou, in preparation). The same workflow was followed for the apo-VSD4-NavPaS dataset, and relevant numbers of particles can be found in fig. S2.

Model building

The structure of WT-NavPaS (PDB 5X0M) was used as a template for modeling the Nav1.7-VSD4-NavPaS coordinates in SwissMODEL (97). The resulting model and the high-resolution crystal structure of AaH2 (PDB 1PTX) were fit as rigid bodies into the cryo-EM map. After manual adjustments to the channel-toxin model, a single round of real space refinement using the phenix.real_space_refinement tool with tight secondary structure restraints (98) was used to correct global structural differences between the initial model and the map. The model was further manually adjusted in Coot (99) through iterative rounds of model building and real space refinement. Phenix.mtriage was used to compute model vs map FSC curves (fig. S2) that are consistent with our resolution estimate from the half-map FSC measurement.

A definitive binding pose for the VSD1-bound AaH2 could be determined, although density for many side chains does not allow for unambiguous assignment. Several attempts at 3D classification, including using approaches focused on the VSD1 binding site, did not yield improved resolution or separation between particles with bound and unbound AaH2 on VSD1. Our confidence in the AaH2-VSD1 complex is supported by the observations that (i) the NavPaS channel is highly asymmetric, which makes it unlikely there were frequent misalignments by 90° around the pore axis, and (ii) no such extra densities can be seen for other major features protruding from the membrane, e.g., the extended DI S5-S6 turret loop.

For the apo-VSD4-NavPaS channel structure, the model was built and refined similarly as described above.

Xenopus oocyte recordings (BgNav1)

All Blattella germanica Nav (BgNav1) mutants were generated using the Gibson assembly method. For each mutant, four linear DNA fragments with approximately 20-bp homology arms were generated using PCR (Phusion High-Fidelity DNA Polymerase, New England Biolabs, USA). Fragments were purified (MinElute PCR Purification Kit, Qiagen, USA) and subsequently incubated for 2 hours at 55°C with Gibson Assembly Master Mix (New England Biolabs, USA). “Assembled” plasmids were then transformed (CopyCutter EPI400, Epicentre, USA) and mutagenesis was confirmed by automated Sanger sequencing. cRNA was synthesized using T7 polymerase (mMessage mMachine kit, Life Technologies, USA) after linearizing the DNA with appropriate restriction enzymes.

BgNav1 control and mutant constructs were expressed in Xenopus laevis oocytes (sourced from Xenopus one, USA), and electrophysiological recordings were taken after 2 to 4 days post cRNA injection. Oocytes were incubated at 17°C in Barth’s medium (96 mM NaCl, 2 mM KCl, 5 mM HEPES, 1 mM MgCl2, and 1.8 mM CaCl2, 50 μg/ml gentamycin, pH 7.6 with NaOH) and studied using the two-electrode voltage-clamp recording technique (OC-725C, Warner Instruments) with a 150-μl recording chamber. All data were filtered at 4 kHz and digitized at 20 kHz using pClamp 10 software (Molecular Devices, USA). The external recording solution used was ND-100 (100 mM NaCl, 5 mM HEPES, 1 mM MgCl2, and 1.8 mM CaCl2, pH 7.6 with NaOH) and microelectrode resistances were 0.5 to 1.0 MΩ when filled with 3 M KCl. All experiments were performed at room temperature (~22°C), and AaH2 samples were diluted in ND-100 with 0.1% BSA. Leak and background conductances were identified and subtracted by blocking the channel with TTX (Alomone Labs, Israel). After adding toxin to the recording chamber, equilibration between channel and toxin was monitored using weak depolarizations (~20) at 5-s intervals unless otherwise noted. Typically, voltage-activation relationships were recorded before and after toxin addition. Off-line data analysis was performed using Clampfit 10 (Molecular Devices, USA), Microsoft Excel (Microsoft, USA) and Prism 7 (GraphPad, USA). For data presented in figs. S8 and S9, significance of all normalized conductance-voltage (G-V) and steady-state inactivation (I-V) relationships was analyzed using two-way analysis of variance (ANOVA) with post hoc Tukey’s test. Individual time point values for fast-inactivation time constants (t), persistent current, peak current, and recovery from fast inactivation (RFI) were analyzed using Student’s t test. Values in all cases reflect the mean, and error bars reflect SEM, P < 0.05 (*) or 0.001 (**). For all data presented in table S2, Student’s t (unpaired) against wild-type was used.

Xenopus oocyte recordings (human Nav1.5)

Mutations in the wild-type human Nav1.5 gene in the pcDNA 3.1 vector were introduced via Gibson assembly in conjunction with gblock synthesis (IDT, Coralville Iowa) and verified with automated Sanger sequencing. In vitro transcription of cRNA was accomplished with the mMessage mMachine T7 ultra kit, following the manufacturer’s instructions.

For two-electrode voltage-clamp recordings, RNA encoding wild-type and mutant Nav1.5 channels were injected (50 to 100 ng) into Xenopus oocytes purchased from Ecocyte, Inc. (Austin, TX). Two-electrode voltage-clamp recordings were performed 1 to 3 days postinjection in oocyte Ringer’s solution containing (in mM): 116 NaCl, 2 KCl, 1.8 CaCl2, 2 MgCl2, 5 HEPES, pH 7.4. Electrodes had resistances of 0.3 to 0.6 MΩ when filled with 3 M KCl. The NPI Turbo-TEC 03X amplifier was used. Signals were filtered at 20 kHz by the amplifier and digitized at 100 kHz by pClamp software (Molecular Devices, San Jose, CA, USA). Traces were filtered at 10 kHz for display in figures. Where indicated, we co-injected cRNA encoding the β1 with Nav1.5 at a 2:1 mass ratio of alpha to β. Although neither the activation nor the rate of inactivation of Nav1.5 are affected by coexpression of the β1 subunit with Nav1.5 in Xenopus oocytes (100, 101), in our hands, co-injection with β1 caused a small (~3 mV) depolarizing shift in steady-state inactivation that was consistent among variants and did not affect their relationship relative to one another (table S2). Offline analysis was performed using Clampfit 10 (Molecular Devices, San Jose, CA, USA) for peak current quantification and Origin Pro (Northampton, MA) for fitting.

Whole-cell patch-clamp recordings (human Nav1.2 and Nav1.7)

Wild-type Nav1.7 voltage-clamp recordings were obtained from HEK293 cells constitutively expressing human Nav1.7 (GenBank Accession: NM_002977) and the human β2 subunit. HEK-TetOn cells were transfected with 2 μg of mutant Nav1.7 DNA in the pBi vector containing the human β2 subunit using Lipofectamine LTX (Invitrogen) 15 to 20 hours prior to recording. Wild-type Nav1.2 voltage-clamp recordings were obtained from HEK293 cells heterologously expressing human Nav1.2 (GenBank Acession: NM_021007) and the human β1 subunit. Cells expressing Nav1.2 were induced 15 to 20 hours prior to recording using doxycyclin. Whole-cell patch-clamp recordings were obtained using a Molecular Devices Axopatch 200B patch-clamp amplifier. The recording pipette intracellular solution contained (in mM): 140 CsF, 10 NaCl, 1.5 MgCl2, 10 HEPES, 5 EGTA adjusted to pH 7.3 with CsOH, osmolarity 300. The extracellular recording solution contained (in mM): 80 NaCl, 60 NMDG, 4 KCl, 2 CaCl2, 1 MgCl2, 5 glucose, 10 HEPES, and containing 0.1% BSA (pH 7.4, osmolarity 300 mOsm).

Lyophilized powder of both native and wild-type and toxin mutants (supplied by Smartox Biotechnology) were resuspended in extracellular recording solution, and the concentration checked with A280. Currents were recorded at 20-kHz sampling frequency and filtered at 5 kHz. Series resistance compensation was applied at >80%. A stable baseline was established prior to toxin perfusion, after which increasing concentrations of toxin were perfused using the Dynaflow Resolve System (Fluicell) and resulting currents were measured after equilibrium reached. Data was extracted using Clampfit (Molecular Devices), analyzed using the tidyverse package in Rstudio (Hadley Wickham), and plotted in Origin (OriginLab Corporation). A 15-ms window after channel opening was used to derive dose responses, the process of which is pictorially demonstrated in fig. S7A. EC50 measurements on Nav1.2 and Nav1.7 were made while holding the membrane voltage at −100 mV. A 100-ms pulse to 0 mV was used to open the channel. Using this protocol at a 0.5-Hz pulse rate, the membrane voltage is maintained at −100 mV 95% of the time. For state-dependence experiments, the membrane voltage was held at either −40 or −120 mV. A 20-ms prepulse to −150 mV was used to partially recover channels from inactivation, followed by a 20-ms pulse to 0 mV to open the channel. Using this protocol at a 0.2-Hz pulse rate, the membrane voltage is maintained at −40 or −120 mV 99.2% of the time.

Supplementary Materials

Figs. S1 to S11

Tables S1 and S2


References and Notes

Acknowledgments: We thank our Genentech colleagues in the Neuroscience, Structural Biology, BioMolecular Resources, Early Discovery Biochemistry, and Early Stage Cell Culture departments for their support of this project and, in particular, A. Estevez, H. Xu, M. Dourado, L. Deng, J. Fuhrmann, C. Koth, and S. Hymowitz. We thank our colleagues at Smartox Biotechnology for generating the AaH2 derivatives used in this study. C.A.A. acknowledges R. Kass for the wild-type Nav1.5 clone, and F.B. acknowledges K. Dong for the wild-type BgNav1 clone. Funding: A.C. was supported by an MRC Industrial iCASE Ph.D. studentship award (MR/N017927/1). D.T.I. was supported by the Cystic Fibrosis Foundation (INFIEL17F0). J.P.L. was supported by the James H. Gilliam Fellowships for Advance Study program through the Howard Hughes Medical Institute. M.-F.M.-E. acknowledges support from the Institut national de la santé et de la recherche médicale (Inserm). P.E.B. was supported by the Centre national de la recherche scientifique (CNRS: PEPS-2009/PAGAIE). C.A.A. was supported by GM122420 and is an American Heart Association established investigator (5EIA22180002). F.B. was supported by the National Institute of Neurological Disorders and Stroke of the National Institutes of Health (R01 NS091352). Author contributions: T.C. established conditions for protein purification and toxin complex formation. T.C. and C.P.A. optimized cryo-EM samples and, with assistance from C.C., collected cryo-EM data. T.C. and A.R. processed and determined cryo-EM structures and, with J.P., performed structural analyses. Z.R.L. and Y.J. generated essential molecular biology reagents. M.-F.M.-E. and P.E.B. performed native AaH2 purifications and provided guidance on toxin pharmacology. A.C. and D.H.H. performed electrophysiological analyses on human Nav1.2 and Nav1.7 channels. D.T.I. and C.A.A. performed electrophysiological analyses on the human Nav1.5 channel. J.P.L. and F.B. performed electrophysiological analyses on the cockroach BgNav1 channel. T.C. and J.P. prepared the manuscript with assistance and input from all other authors. C.A.A., F.B., D.H.H., A.R., and J.P. are senior co-authors, and J.P. supervised the project. Competing interests: T.C., A.C., C.P.A., Z.R.L., Y.J., C.C., D.H.H., A.R., and J.P. were all employees of Genentech, Inc., at the time of this study; all other authors declare no competing financial interests. Data and materials availability: Materials will be made available upon request and material transfer agreement with Genentech or the appropriate party. Accession numbers for the apo-VSD4-NavPaS (VSD4-activated) cryo-EM model and maps are Protein Data Bank (PDB) 6NT3 and EMD-0500, respectively; accession numbers for the AaH2-VSD4-NavPaS (toxin-bound, VSD4-deactivated) cryo-EM model and maps are PDB 6NT4 and EMD-0501, respectively.

Stay Connected to Science

Navigate This Article