Volatile anaesthetics modulate voltage-gated sodium channel function at a site directly linked to channel gating

Abstract

Voltage-gated sodium channels (VGSCs) mediate neuronal excitability and synaptic transmission and are functionally relevant targets for volatile anaesthetic (VA) actions. Here, we show that multiple VAs at clinically relevant concentrations share binding sites on NavMs, a prokaryotic VGSC. Sevoflurane, a representative VA, interacts with NavMs and NaChBac with functional effects paralleling those on human VGSCs, including modulation of channel inactivation. X-ray crystallography of purified NavMs reveals an atomic-resolution VA binding site in a VGSC, in which sevoflurane displaces lipid to occupy a membrane-embedded hydrophobic pocket. Alanine substitution of an invariant tyrosine within this binding pocket abolishes sevoflurane binding and eliminates the sevoflurane-induced hyperpolarising shift of steady-state inactivation. Sevoflurane modulates both fast and slow inactivation of human Nav1.1, demonstrating VA modulation of steady-state slow inactivation in a neuronal VGSC. Supporting evidence shows that VAs interact with homologous sites in human VGSCs. These findings define a VA binding site in VGSCs that supports a membrane-assisted pathway for modulating channel gating and neuronal activity in general anaesthesia.

Volatile anaesthetics modulate sodium channels critical for neuronal activity, yet structural details are lacking. This study presents an X-ray structure of sevoflurane bound to a sodium channel and identifies a critical gating residue.

Introduction

Since the discovery of diethyl ether as the first general anaesthetic in the 1840s, an important milestone in modern medicine, volatile anaesthetics (VAs) have become established as essential medicines for enabling painful procedures by producing amnesia, unconsciousness and immobility. Nonflammable halogenated ethers have since replaced diethyl ether, with sevoflurane, desflurane and isoflurane currently in widespread use^1,2. Despite their clinical importance, the mechanisms by which VAs produce anaesthesia remain poorly understood³, although the positive correlation between VA potency and lipophilicity implicates a role for hydrophobic interactions⁴.

The principal neuronal targets of VAs are membrane proteins, in particular ligand-gated and voltage-gated ion channels^3–6. Relatively low-affinity interactions of VAs with relevant target proteins modulate the neural networks responsible for consciousness, memory and nocifensive responses through suppression of excitatory and potentiation of inhibitory synaptic transmission⁷. Interactions of VAs with neuronal and cardiac voltage-gated sodium channels (VGSCs) are also implicated in the serious adverse side effects of VAs, including neurotoxicity, cognitive dysfunction, respiratory depression and cardiovascular complications^8,9. Identification of functionally-important anaesthetic binding sites in the molecular targets of VAs is crucial for understanding their effects, defining their molecular pharmacology, and thus enabling rational structure-based design of more selective VAs with fewer side effects.

Voltage-gated sodium channels are critical to cellular excitability, and are important targets for the neurophysiological effects of VAs^3,10–16. The nine human VGSC isoforms (hNav1.1–hNav1.9) are expressed primarily in excitable tissues¹⁷, where they drive action potentials and shape excitability by transitioning between resting (non-conducting, available to open), open (conducting) and inactivated (non-conducting, unavailable to reopen until recovery) conformational states in a voltage-dependent manner¹⁷. The haloether VAs sevoflurane, desflurane, isoflurane and halothane, a haloalkane, exert similar overall effects on VGSCs at clinically relevant concentrations by inhibiting peak Na⁺ current (I_Na), accelerating current decay, modulating inactivation kinetics and delaying recovery from inactivation^{13–16,18–20}. Preferential interactions with inactivated states^{11,16,18,21,22} contribute to reduced cellular excitability^21,23,24, synaptic vesicle exocytosis^25,26 and neurotransmitter release^25,27, all neurophysiological targets of the well-known effects of VAs on synaptic transmission.

Understanding the molecular mechanisms underlying these effects of VAs on VGSCs has been limited by the lack of atomic-resolution structural data to define their functionally-important binding sites. Obtaining high-resolution structural data for eukaryotic VGSCs is challenging, as these structurally complex membrane proteins are refractory to X-ray crystallography. Although cryo-electron microscopy (cryo-EM) structures of eukaryotic VGSCs are available, no cryo-EM structure of a VA bound to any protein has been reported. This is likely due to the inherent challenges of detecting small, volatile and low-affinity ligands using this technique.

Bacterial homologues provide effective models of eukaryotic ion channels for structural studies of ion channel-drug interactions to identify the molecular basis for drug effects. For example, X-ray crystallography has been used to identify and characterise binding sites for isoflurane and desflurane in bacterial homologues of pentameric ligand-gated ion channels, including of GABA_A receptors, another major general anaesthetic target^28,29. Bacterial VGSCs (BacNavs) share major structural and functional properties with their eukaryotic counterparts^30,31. Although BacNavs lack the machinery for fast inactivation and produce only slow inactivation, some drugs that modulate inactivation of BacNavs similarly affect fast inactivation of human VGSCs at equivalent concentrations^32,33. This suggests that allosteric binding sites common to all VGSC subtypes can potentially modulate one or both inactivation processes. While the BacNav NaChBac from Bacillus halodurans is inhibited by VAs at clinical concentrations^13,34,35, and binding sites have been predicted using molecular dynamics simulations and nuclear magnetic resonance (NMR) spectroscopy^34,36,37, NaChBac has never been crystallised, and no atomic-resolution structures exist to substantiate the putative VA binding sites. In contrast, NavMs from Magnetococcus marinus crystallises to high resolution and provides an excellent BacNav model for understanding drug actions on eukaryotic VGSCs^32,38,39.

NavMs has the typical BacNav structure of four identical ~30 kDa monomeric subunits that assemble to form a homotetrameric channel⁴⁰. Each subunit consists of six transmembrane helices (S1–S6, labelled 1–6 in Fig. 1a) that form channels with four peripheral voltage-sensing domains (VSDs; helices S1–S4), each connected by a linker helix (the S4–S5 linker helix) to a Na⁺-conducting pore module (PM; helices S5–S6 and their interconnecting loops from all four subunits) organised in a domain-swapped arrangement where the VSD from one subunit packs against the PM from the adjacent subunit³¹ (Fig. 1b). Outward movement of the S4 helix within the VSD is transmitted via the S4–S5 linker to the PM, inducing conformational changes that drive the transition between closed and open channel states. Lateral fenestrations in the PM form intramembrane openings to the channel pore, providing a hydrophobic membrane-accessible pathway for drugs that can block channels from their closed, resting state⁴¹. BacNavs share their fundamental tetrameric architecture with eukaryotic VGSCs, which consist of a single amino acid chain (~210 kDa) with the four nonidentical domains (DI–DIV) replacing the four identical BacNav subunits to create a pseudo-tetrameric structure.

Fig. 1. The homotetrameric bacterial voltage-gated Na⁺ channel NavMs interacts directly with multiple volatile anaesthetics.

a Membrane topology of NavMs; the six transmembrane helices (1–6) make up the monomeric subunit. b NavMs tetramers form the functional domain-swapped channel with one subunit having coloured helices as in (a). c Sevoflurane chemical structure with labelling of the three sevoflurane protons (H1–H3, top) and in stick representation (bottom). d Stacked NMR spectra show STD (top, green) produced after 4 s saturation at the OFF (bottom, blue) and ON (middle, red) resonances, showing that Saturation Transfer Difference (STD) is produced at all sevoflurane proton peaks. P1/P1’ is a single split peak formed by splitting of the peak for the equivalent H1 and H2 protons shown in (c) to give a doublet peak due to ²J_HF coupling with the fluorine of the CH₂F-group, and P2 corresponds to the H3 proton shown in (c) and is a broad septet peak due to ³J_HF coupling with the six fluorine atoms of the two neighbouring CF₃-groups. e Saturation time-dependent build-up of STD at the P1/P1’ doublet resonance continues until plateau (STDmax, ~33%). STD(%) is calculated as (I_off – I_on) / I_off × 100, where I_off and I_on are the integrals of the sevoflurane proton peaks in the OFF RES (irradiation −20 ppm) and the ON RES (−0.5 ppm) spectra at the indicated saturation times. Plot shows STD vs. time at P1/P1’ with the experimental values fitted to Eq. 1 (see “Methods”). f Similar stacked NMR STD spectra as shown in (d) for desflurane (left), isoflurane (middle) and halothane (right) show STD for all three VAs, indicating direct interactions (a stick drawing of each VA is provided at the top of each panel). g VAs compete for binding sites on NavMs. Stacked spectra shown at the sevoflurane P1/P1’ doublet peak at STDmax show that the STD produced without competitor VA (bottom spectrum) is reduced by saturating concentrations of desflurane (pink) and isoflurane (purple), and abolished by halothane (light green). Source data are provided as a Source Data file.

Here we show that the VAs sevoflurane, desflurane, isoflurane and halothane compete for common binding sites on VGSCs to produce their functional effects. We present an atomic-resolution crystal structure of a VA bound to a VGSC, revealing a membrane-accessible hydrophobic pocket as a VA binding site. Alanine substitution of a conserved BacNav tyrosine in the sevoflurane binding site (Y143 in NavMs) abolishes sevoflurane binding at this site and eliminates the sevoflurane-induced hyperpolarising shift in steady-state inactivation observed in wild-type (WT) BacNavs, a process considered analogous to slow inactivation in hNavs. We extend these findings to two homologous sites in human VGSCs, where phenylalanine occupies the position corresponding to the conserved tyrosine by showing that the functional effects of sevoflurane on NavMs are preserved in the “humanised” NavMs Y143F mutant. Moreover, we show that sevoflurane stabilises inactivated states of neuronal hNav1.1 by hyperpolarising the voltage dependence of both steady-state fast and slow inactivation, while in silico docking studies support interaction at the corresponding pockets in human Nav.

Results

Volatile anaesthetics interact directly with NavMs

We explored whether direct interactions between VAs and NavMs occur using saturation transfer difference nuclear magnetic resonance (¹H STD NMR) spectroscopy⁴². This is an ideal technique for identifying weak protein-ligand interactions such as those reported for VA binding to ion channels, including NaChBac³⁶. These interactions are reported through reduced intensities of ligand ¹H NMR signals caused by internuclear saturation transfer from a selectively irradiated protein site to the ligand when in close (< 7 Å) and repetitive contact⁴³. We probed interactions between NavMs and the widely used VA sevoflurane, which contains 3 protons (Fig. 1c), using a clinically relevant aqueous sevoflurane concentration of 0.57 mM (equivalent to 2 times MAC [minimum alveolar concentration, the 50% effective dose in vivo], the concentration used in all subsequent functional experiments).

Selective irradiation of the NavMs ¹H signal at −0.5 ppm (ON resonance) resulted in saturation transfer from NavMs to sevoflurane, reducing the intensities of all sevoflurane ¹H spectral signals (Fig. 1d, middle); this did not occur when irradiation was performed at a control resonance outside the NavMs ¹H spectrum (−20 ppm, OFF resonance, Fig. 1d, bottom). The difference (OFF resonance – ON resonance) resulted in the saturation transfer difference spectrum (STD, Fig. 1d, top), which increased at all sevoflurane proton resonance signals in a time-dependent manner until STDmax was achieved at plateau (Fig. 1e, showing STD build-up to STDmax of ~33% at the P1/P1’ doublet resonance that corresponds to the equivalent H1/H2 protons, see figure legend). No STD signals were detected at sevoflurane proton resonances in the absence of NavMs (Supplementary Fig. 1). Together, these data identify direct interaction between NavMs and sevoflurane.

We investigated whether other VAs also directly interact with NavMs at the same sites as sevoflurane despite their structural differences. We performed competition STD experiments by introducing buffer-saturating concentrations of a competitor VA to a buffer already containing 1 μM NavMs and 2 MAC sevoflurane. High saturating concentrations of desflurane, isoflurane or halothane were used to observe their maximum achievable effects on the sevoflurane STD signals, since this technique relies on fast-exchanging ligands. STD was produced at the proton resonances of each of the competitor VAs (shown at STDmax in Fig. 1f), indicating direct interactions with NavMs, with ratios of competing VA to sevoflurane of ~35:1 for desflurane, ~40:1 for isoflurane and ~80:1 for halothane. In the absence of a competitor VA, sevoflurane produced time-dependent STD up to a plateau (STDmax of ~8% after 8 s saturation time shown for the P1/P1’ doublet peak; Fig. 1g, blue). The sevoflurane STD at STDmax was reduced by ~50% in the presence of desflurane (Fig. 1g, pink) and by ~71% in the presence of isoflurane (Fig. 1g, purple), while halothane completely abolished the sevoflurane STD (Fig. 1g, light green). This demonstrates that these three VAs compete with sevoflurane for shared binding sites on NavMs. Halothane, the most chemically distinct VA of those investigated (a haloalkane rather than a haloether), fully displaces sevoflurane binding to NavMs, which is consistent with its much greater aqueous solubility compared with desflurane and isoflurane, which would require experimentally unattainable concentrations to eliminate the sevoflurane STD signal.

Sevoflurane has comparable effects on bacterial and mammalian VGSC function

At clinically relevant concentrations, VAs affect mammalian, including human, VGSCs by inhibiting peak Na⁺ current (I_Na), accelerating current decay, producing a characteristic hyperpolarising shift in the voltage-dependence of steady-state inactivation and slowing channel recovery from inactivation^13,18, with minimal effects on activation. In NavMs heterologously expressed in HEK293T cells, sevoflurane reduced peak I_Na from the resting state (Fig. 2a, b, d), accelerated current decay (Fig. 2c), shifted the voltage-dependence of steady-state inactivation toward more hyperpolarised potentials without altering the voltage-dependence of activation (Fig. 2d) and slowed recovery from inactivation (Fig. 2e, f, Supplementary Table 1). Overall, these functional effects on NavMs are comparable to those of VAs on human VGSCs, and are consistent with multisite VA binding to VGSCs as predicted by molecular dynamics studies^34,36,37.

Fig. 2. Sevoflurane effects on NavMs function.

a NavMs was expressed in HEK293T cells, and Na⁺ current (I_Na) was measured by whole-cell patch-clamp. Representative families of whole-cell inward I_Na (protocol, inset) in the absence (CTL; left, black traces) or presence of a clinical concentration of sevoflurane (0.57 mM (2 MAC) SEVO; right, orange traces). b Sevoflurane inhibited peak I_Na. Representative traces of peak I_Na evoked by a single pulse (inset) in the absence (CTL; black) or presence of sevoflurane (SEVO; orange). Normalised peak I_Na was reduced to 0.48 ± 0.12 by 0.57 mM (2 MAC) sevoflurane (P = 8.2 × 10⁻¹¹, n = 15). c Sevoflurane accelerated current decay. Normalised traces from (b) were fitted to a mono-exponential equation (green curves) to yield τ_inact. Sevoflurane reduced τ_inact by ~two-fold from 21.3 ± 7.9 ms (CTL; white circles, right panel) to 10.6 ± 3.4 ms (SEVO; orange circles, P = 5.8 × 10⁻⁶, n = 15). d I–V curve (left), normalised conductance (G/G_max) and inactivation (I_Na/I_Namax; right). Sevoflurane did not significantly affect activation (V_½-act −71.5 ± 7.6 mV [CTL] vs. −66.7 ± 6.6 mV [SEVO], P = 0.0695, n = 8). Sevoflurane shifted inactivation (V_½-inact) by −19.4 ± 6.0 mV (−106.1 ± 6.4 mV [CTL] to −125.5 ± 8.2 mV [SEVO], P = 4.0 × 10⁻⁵, n = 8). e, f Sevoflurane slowed recovery from inactivation. A two-pulse protocol was used (inset). Peak I_Na of the second test pulse was normalised to the first pulse (Pulse₂/Pulse₁). Data were fitted to a mono-exponential equation to yield τ_recovery. e Normalised I_Na in CTL (black) or SEVO (orange). The sevoflurane traces were plotted overlaid with an x-axis offset of +30 ms to allow visual comparison between the current dynamics. f Fitted data in the absence (CTL; white circles) or presence (SEVO; orange circles) of sevoflurane. Inset shows τ_recovery derived from the fitted data, which increased from 432 ± 158 ms to 714 ± 59 ms in the presence of sevoflurane, slowing recovery from inactivation (P = 0.0023, n = 6). Data shown as mean ± SD; drug effects vs. control were tested by paired two-tailed Student’s t-test (*P < 0.05; **P < 0.01; ****P < 0.0001). n indicates the number of individual cells from ≥2 independent transfections. Source data are provided as a Source Data file.

Identification of a sevoflurane binding site in NavMs at atomic resolution

No atomic-resolution structure of a VA bound to any voltage-gated ion channel has been reported, so the binding site(s) underlying VA-VGSC interactions inferred from NMR, MD simulations, and electrophysiological studies have not been identified. To address this critical gap in anaesthetic pharmacology, we used X-ray crystallography to identify sevoflurane binding sites in NavMs F208L, which is functionally indistinguishable from WT NavMs but crystallises more reliably and has been used in previous structure-function studies of VGSC-drug interactions^32,38,39. Apo-NavMs F208L crystals were grown and incubated with sevoflurane-saturated mother liquor, then harvested after increasing incubation times (denoted as NavMs-SEVO crystals). X-ray diffraction data were collected for both NavMs-SEVO crystals and apo-NavMs crystals (apo-NavMs).

Prolonged incubation led to reduced diffraction quality, culminating in a complete loss of diffraction at extended times, but shorter incubation times (<20 min) resulted in some crystals that diffracted to the high resolution of apo-NavMs crystals. Comparison of multiple datasets from control and sevoflurane-exposed crystals showed that while the protein structures were identical (rmsd <0.3 Å), a marked change in the shape of non-protein electron density occurred in a hydrophobic pocket below the side chain of residue Y143, located on the membrane-facing side of the S5 helix of the PM. Polder-OMIT maps⁴⁴, generated to isolate these ligand electron densities minimised for interference from bulk solvent effects, showed that apo-NavMs crystals had a short, elongated stretch of density attributed to part of the alkyl chain of lipid or detergent (Fig. 3a, left), while the electron density at this location in NavMs-SEVO crystals was ellipsoidal (Fig. 3a, middle) with a shape into which a single sevoflurane molecule could be fit in several orientations.

Fig. 3. Sevoflurane binding to NavMs at a membrane-exposed site.

a Polder OMIT map density showing the morphologies of electron density (grey) below the helix S5 residue Y143 in apo-NavMs F208L crystals (left, contoured @2.5σ), NavMs-SEVO crystals (middle, contoured @4σ) and chloroSEVO crystals (right, contoured @4σ). The anomalous signal for the chlorine atom in chloroSEVO crystals is also shown (red density, contoured @4σ). b Two views of chloroSEVO (stick representation in grey electron density contoured at 1σ) in its binding site located between the S1 and S4 helices of one subunit (yellow) and the S5 helix from the adjacent subunit (magenta). Anomalous chlorine electron density (red, 4σ) allows accurate positioning of chloroSEVO into the density. c Space-filling model (left, coloured by subunit) showing sevoflurane (sphere representation) in its intramembrane binding site below Y143 (bilayer boundaries indicated with black lines on left side of image). Close-up of the sevoflurane binding site (right inset) showing residues (coloured by subunit) involved in binding to sevoflurane (stick) with sevoflurane electron density contoured at 1σ. d Hydrophobic surface representation of sevoflurane (stick) bound in the hydrophobic intramembrane pocket created by the binding site residues (coloured by hydrophobicity, red - highest to blue - lowest). e ¹⁹F-¹⁹F STD build-up curve at the sevoflurane fluorine resonance (−74.7 ppm) confirms the pharmacological relevance of the site for sevoflurane binding at the clinically relevant concentration of 0.57 mM (2 MAC). STD (%) is calculated as (I_off – I_on) / I_off × 100, where I_off and I_on are the integrals of the sevoflurane fluorine peak at −74.7 ppm in the OFF RES spectrum (where irradiation is at 0 ppm, empty of fluorine signal) and the ON RES spectrum (where irradiation is at −83.8 ppm, which contains the unique fluorine signal from BTFA-labelled NavMs T19C) at the indicated saturation times. The curve produced from experimental values was fit using Eq. 1 (see “Methods”). Source data are provided as a Source Data file.

Orientation and interactions of sevoflurane in the binding pocket

To identify this electron density unambiguously as sevoflurane and to orientate sevoflurane into this binding site, we carried out the same crystal exposure experiments replacing sevoflurane (2-(fluoromethoxy)-1,1,1,3,3,3-hexafluoropropane) with its monochloro-substituted analogue 2-(chloromethoxy)-1,1,1,3,3,3-hexafluoropropane (chloroSEVO). This analogue retains anaesthetic properties, but contains a fluorine-to-chlorine substitution on the fluoromethyl moiety of sevoflurane⁴⁵. Chlorine is a weak anomalous X-ray scatterer that can be detected using long-wavelength X-ray crystallography⁴⁶. Analysis of data collected on three NavMs-chloroSEVO crystals showed the same ellipsoidal density found below Y143 in the electron density map of NavMs-SEVO, with the anomalous data containing a signal within the density (Fig. 3a, right) that overlapped across all three crystal datasets (Supplementary Fig. 2), but which was not present in the anomalous data collected from apo-NavMs crystals. These data confirmed sevoflurane binding to this site in NavMs-SEVO crystals, with the anomalous chlorine signal allowing orientation of chloroSEVO, and by inference sevoflurane, into this binding site (Fig. 3b, c).

Sevoflurane binds within a preformed, intramembranous pocket at the interface between the VSD of one NavMs subunit and the PM of the adjacent subunit (Fig. 3c, d). Binding is mediated predominantly by long-range hydrophobic packing interactions consistent with van der Waals force-dominated interactions involving residues V23, G26, and A27 from helix S1 and V107 and L110 from helix S4 of the VSD, and with T139, V140 and Y143 in the S5 helix of the PM (Fig. 3c, boxed). One trifluoromethyl group of sevoflurane and its fluoromethyl group interact with VSD residues and Y143 in S5, while the other trifluoromethyl group interacts with the PM only. Atoms from all three groups make multiple contacts with Y143, identifying it as the primary binding residue. Sevoflurane displaces the alkyl tail of a lipid (or HEGA-10 detergent, which can replace lipid molecules during protein purification) to fill a hydrophobic cavity, thus disrupting native protein-lipid interactions, which can be essential for normal channel function⁴⁷. This binding pocket would be dynamic during channel gating, as the S4 helix and the PM move in response to membrane potential changes between polarised and depolarised channel states. Consequently, this sevoflurane binding site identified in the inactivated channel state of the NavMs structure would not be present in the resting (polarised) channel state. Previous MD simulations predicted multiple potential VA binding sites within VGSCs^34,37. The Y143 binding site was not identified in these studies.

The highest-resolution structures of apo-NavMs and NavMs-SEVO collected, both solved up to a resolution of 2.2 Å, have been deposited in the Protein Data Bank (apo-NavMs PDB ID: 9GTQ, NavMs-SEVO PDB ID: 9GV1).

Sevoflurane interacts with the Y143 binding site at a clinically relevant concentration

The crystal structure of sevoflurane bound to NavMs was determined at a saturating concentration of sevoflurane. To determine sevoflurane interactions with the Y143 binding site at the clinically relevant concentration used in our functional studies, we employed a fluorine-modified STD NMR technique (¹⁹F-¹⁹F STD NMR) in which proteins are selectively fluorinated at cysteine residues introduced near predicted binding sites to report on fluorinated ligand interactions with those sites³⁶. NavMs contains one native cysteine residue, which was mutated to alanine (C52A) prior to incorporating a cysteine at T19 (T19C) to make NavMs C52A/T19C. We chose to mutate and modify T19 because the fluorinated probe introduced at this position would be too far away to report on any of the potential VGSC-VA binding sites previously proposed from MD simulations using the BacNav NaChBac^36,37. This allows selective reporting of sevoflurane binding to the Y143 binding site, where fluorine atoms in sevoflurane are in close proximity (~ 6 Å) to probe fluorines.

NavMs C52A/T19C expression was ~15 times lower than that of WT NavMs, but gel filtration showed a protein peak that matched that of WT NavMs, indicative of an equally well-folded protein. Fluorine tagging was performed using the cysteine-specific alkylating agent 3-bromo-1,1,1-trifluoroacetone (BTFA), which conjugates a trifluoromethyl-containing group onto free cysteines. Selective irradiation at the trifluoromethyl resonance of BTFA-labelled NavMs C52A/T19C (at −83.8 ppm) incubated with 0.57 mM sevoflurane produced time-dependent STD build up at the two sevoflurane fluorine resonance peaks present in the 1D ¹⁹F spectrum at −74.7 ppm, representing the six-equivalent fluorines in the hexafluoroisopropyl group of sevoflurane (Fig. 3e), and at −152.2 ppm, corresponding to the single fluorine in its fluoromethyl group, indicating that sevoflurane interacts with the Y143 binding site at this clinical sevoflurane concentration. Relative STDmax produced at the ¹⁹F resonance of the hexafluoroisopropyl group fluorine resonance was greater (17.5%, Fig. 3e) than that produced at the fluoromethyl group fluorine resonance (3.5%, Supplementary Fig. 3). This supports the pose of sevoflurane in the Y143 binding site, inferred from the anomalous difference peak corresponding to the chlorine atom of chloroSEVO, in which fluorine atoms of the hexafluoroisopropyl group of sevoflurane would be >3 Å nearer to the position of the T19 fluorine probe compared to the fluoromethyl fluorine, which is located near the detection limit of the technique.

Y143 is necessary for both sevoflurane binding to NavMs and normal channel function

Our structural studies of NavMs-SEVO predict that the Y143A mutation would abolish sevoflurane binding at the Y143 binding site by disrupting its binding interactions. This would facilitate characterisation of the specific contribution of sevoflurane binding at this site to the overall functional effects of sevoflurane binding to NavMs. The approach of introducing Y143A into the double mutant (NavMs C52A T19C Y143A) and using ¹⁹F-¹⁹F STD NMR to explore binding at this site was not possible, given its low protein expression. We therefore tested for the effect of this mutation using ¹H STD NMR of NavMs Y143A, which had good expression and eluted on gel filtration with the same retention time as WT NavMs. ¹H STD NMR of NavMs Y143A and WT NavMs with 0.57 mM sevoflurane showed that NavMs Y143A produced lower STD compared to WT NavMs (STDmax, Y143A = 29.6 ± 2.3, WT = 34.2 ± 0.9, *P < 0.034, n = 3, at the P1-P1’ doublet peak in Fig. 4a). This is consistent with reduced STD arising from loss of the sevoflurane interaction at the Y143A binding site while the other sevoflurane interactions with NavMs remain.

Fig. 4. Functional effects of tyrosine to alanine substitution in the Y143 binding site of NavMs (Y143A) and NaChBac (Y156A).

a STD build-up curve produced at the proton P1/P1’ doublet resonance peak by 0.57 mM (2 MAC) sevoflurane for WT NavMs (red) and NavMs Y143A (blue). Y143A had a reduced STDmax (%) compared to WT (inset, Y143A 29.6 ± 2.3, WT 34.2 ± 0.93, P = 0.0288, unpaired two-tailed Student’s t-test, n = 3 independent experiments), consistent with reduced sevoflurane interaction with NavMs Y143A. Curves fit using Eq. 1 (see “Methods”). b Functional effects of sevoflurane on NavMs Y143A. Representative families of whole-cell inward Na⁺ currents (I_Na) evoked from a holding potential (V_h) of −180 mV in the absence (CTL; left, black traces) or presence of a reduced concentration of 0.28 mM sevoflurane (1 MAC) (SEVO; right, orange traces); stimulation protocol shown in the inset. The Y143A mutation caused a major leftward shift in channel gating, requiring a more hyperpolarised holding potential for recording that exceeded the operational limits of the amplifier. This mutant also expressed poorly with small currents, which were further reduced by sevoflurane, precluding reliable determination of sevoflurane effects on activation and inactivation properties. c NaChBac Y156A mutation had more favourable gating properties from a holding potential (V_h) of −160 mV, shown in the absence (CTL; left, black traces) or presence (SEVO; right, orange traces) of 0.57 mM (2 MAC) sevoflurane; stimulation protocol shown in the inset. d, e Comparison of voltage-dependence of activation and inactivation for WT NaChBac (d) and NaChBac Y156A (e) in the absence (white symbols) or presence (orange symbols) of sevoflurane. Current-voltage relationship of channel activation (normalised conductance; G/G_max) and inactivation (I_Na/I_Namax). Sevoflurane shifted the voltage-dependence of half-maximal inactivation (V_½-inact) by −13.4 ± 3.1 mV for WT NaChBac (d, f; P = 2.9 × 10⁻⁷, n = 10; Supplementary Table 1). A hyperpolarising shift in V_½-inact was not seen for NaChBac Y156A (e, f; P = 0.140, n = 8; Supplementary Table 1). Data shown as mean ± SD; drug effects vs. control were tested by paired two-tailed Student’s t-test; (*P < 0.05; ****P < 0.0001). n indicates the number of individual cells from ≥2 independent transfections. Source data are provided as a Source Data file.

We examined the functional impact of the Y143A mutation using whole-cell patch clamp electrophysiology. The NavMs Y143A mutation had much smaller I_Na compared to WT NavMs (Fig. 4b, left). These reduced currents were almost fully inhibited by sevoflurane (Fig. 4b, right), such that characterisation of the consequences of sevoflurane binding to the Y143 binding site on channel function was not possible.

Functional characterisation of the sevoflurane-Y143 interaction revealed in NaChBac

NavMs Y143A showed marked alterations in gating compared to WT NavMs (Supplementary Table 1), including a large hyperpolarising shift in the voltage-dependence of inactivation. This made current recordings difficult by further lowering the holding potential needed to return channels to the resting state, which exceeded the technical capabilities of the amplifier used and resulted in a significant decrease in measurable I_Na. All known BacNavs have a conserved tyrosine at the position equivalent to Y143 in NavMs, including NaChBac (Y156), which has previously been used in functional studies involving VAs^34,35. Compared to NavMs, NaChBac requires less extreme hyperpolarisation to return to its resting state, which occurs at membrane potentials closer to those for mammalian VGSCs^34,35,48.

WT NaChBac produced large currents with less negative inactivation gating compared to WT NavMs (Fig. 4d; Supplementary Table 1). Sevoflurane (0.57 mM; 2 MAC) shifted the voltage dependence of steady-state inactivation towards more hyperpolarised potentials, inhibited peak I_Na (Supplementary Table 1) and accelerated current decay. NaChBac Y156A produced robust currents with negative shifts in the voltage dependence of both steady-state activation and inactivation and shallower slopes compared to WT NaChBac (Fig. 4e; Supplementary Table 1). The predominant effects of sevoflurane on WT NaChBac were maintained in NaChBac Y156A, except that the sevoflurane-induced negative shift in steady-state inactivation seen in WT channels was selectively eliminated (Fig. 4e, f). This property of VGSC modulation has been proposed to underlie the reduced neuronal excitability and neurotransmitter release produced by VAs in cellular systems⁴⁹.

Sevoflurane accelerated activation kinetics in both WT and Y156A mutant channels, as evidenced by reductions in the time to peak current and activation time constant (Supplementary Table 1). These effects are thus independent of sevoflurane interactions with the Y156 binding site in NaChBac.

Homologous volatile anaesthetic binding sites are conserved in human VGSCs

All hNav isoforms have a conserved phenylalanine residue at positions equivalent to Y143 in the S5 helices of domains II and IV (i.e., F902 and F1682 in Nav1.1, Supplementary Fig. 4), supporting the existence of related VA-binding sites. The cryo-EM structure of human Nav1.1 (PDB ID: 7DTD⁵⁰, inactivated state) confirms the presence of these residues in conserved hydrophobic pockets. We analysed the effects of the Y143F substitution in NavMs on channel function and sevoflurane binding. NavMs F208L/Y143F expressed and purified similarly to NavMs F208L. It produced crystals under the same conditions that diffracted to high resolution, producing an electron density map and an atomic structure identical to that of NavMs F208L, except for the absence of density attributable to the Y143 hydroxyl group. X-ray diffraction data obtained from NavMs F208L/Y143F crystals following exposure to sevoflurane produced electron density maps that showed the same ellipsoidal density below F143 as observed below Y143 in NavMs-SEVO crystals, confirming sevoflurane occupancy (Fig. 5a). NavMs Y143F expressed in HEK293T cells displayed electrophysiological properties comparable to those of WT NavMs both with and without sevoflurane (Fig. 5b–e), indicating that the Y143F substitution preserves channel function and susceptibility to sevoflurane.

Fig. 5. Substitution of the human residue in the NavMs volatile anaesthetic binding pocket (NavMs Y143F) revealed similar function and sevoflurane sensitivity to WT NavMs.

a Crystal structures of NavMs Y143F exposed to sevoflurane showed a sevoflurane density identical to that found in NavMs-SEVO crystals. b Representative families of whole-cell inward Na⁺ currents (I_Na) in HEK293T cells expressing NavMs Y143F (inset shows stimulation protocol) from a holding potential (V_h) of −180 mV in the absence (CTL; black traces) or presence of 0.57 mM (2 MAC) sevoflurane (SEVO; orange traces). c, d Current-voltage relationships of channel activation (I-V curve; c), normalised conductance (G/G_max) and inactivation (I_Na/I_Namax; d). Sevoflurane did not affect the voltage-dependence of activation (−77 ± 9 mV for CTL; white circles, G/G_max; vs−80 ± 9 mV for SEVO; orange circles, P = 0.544, n = 5). Sevoflurane shifted the voltage-dependence of half-maximal inactivation (V_½-inact) by −20 ± 11 mV (−107 ± 8 mV for CTL; white circles, I_Na/I_Namax; vs. −126 ± 8 mV for SEVO; orange circles, P = 0.0305, n = 5). e Sevoflurane inhibited NavMs Y143F peak I_Na. Representative overlaid traces of I_Na evoked by a single pulse (inset shows stimulation protocol) from V_h − 180 mV in the absence (CTL; black trace, left panel) or presence of 0.57 mM (2 MAC) sevoflurane (SEVO; orange trace). Normalised peak I_Na was reduced to 0.41 ± 0.09 of control by 0.57 mM (2 MAC) sevoflurane (right panel; P = 0.00014, n = 5). WT NavMs data from Fig. 2b are included for comparison (right panel). f Sevoflurane increased the time course of current decay (τ_inact) of channel inactivation in NavMs Y143F, as in WT NavMs. The traces in (e) were normalised, and data for current decay were fitted to a mono-exponential equation (green curves) to calculate τ_inact. Sevoflurane accelerated current inactivation τ_inact from 26.8 ± 11.2 ms (CTL; white circles; right panel) to 9.6 ± 1.8 ms (SEVO; orange circles, P = 0.0172, n = 5). Data shown as mean ± SD; drug effects vs. control were tested by paired two-tailed Student’s t-test; (*P < 0.05; ***P < 0.001, ****P < 0.0001). n indicates the number of individual cells from ≥2 independent transfections. Source data are provided as a Source Data file.

Sevoflurane modulation of hNav1.1 implicates homologous binding pockets

To characterise sevoflurane modulation of hNav1.1, we expressed WT hNav1.1 in HEK293T cells and recorded macroscopic Na⁺ currents by whole-cell patch clamp electrophysiology in the absence or presence of 0.57 mM (2 MAC) sevoflurane. In addition to fast inactivation, attention was given to steady-state slow inactivation, which has not previously been examined for a volatile anaesthetic in a neuronal Na⁺ channel, given its relevance to BacNav function. Using an alternating stimulation protocol to assess voltage-dependence, sevoflurane did not inhibit peak I_Na when channels were activated from a holding potential of −120 mV, at which channels occupy closed resting states (V₀), but significantly reduced peak I_Na following a prepulse to V_½, at which half the channels are fast-inactivated (Fig. 6a). This is consistent with preferential interaction with inactivated states. Sevoflurane also significantly shifted steady-state fast inactivation to more hyperpolarised potentials and produced a smaller but significant hyperpolarising shift in activation (Fig. 6b, c; Supplementary Table 1). Using a 15 s conditioning prepulse protocol to induce slow inactivation (Fig. 6d, e), we found that sevoflurane caused a pronounced hyperpolarising shift in the voltage-dependence of slow inactivation, approximately twice that observed for fast inactivation, consistent with stabilisation of slow-inactivated hNav1.1 conformations.

Fig. 6. Sevoflurane interaction with hNav1.1.

a Wild-type hNav1.1 was expressed in HEK293T cells and voltage-dependent inhibition of peak I_Na was tested using a periodic alternating stimulation protocol (inset) in CTL (black) or 0.57 mM (2 MAC) SEVO (orange). Sevoflurane did not inhibit peak I_Na from a holding potential of −120 mV (V₀; solid traces; normalised peak I_Na of 0.98 ± 0.07; P = 0.4543, n = 11). However, sevoflurane significantly inhibited peak I_Na with a prepulse to V_½, a potential at which half of the channels were in a fast-inactivated state (dashed traces; normalised peak I_Na 0.76 ± 0.06; P = 5.7 × 10⁻⁸, n = 11). b Representative families of I_Na evoked (inset) in CTL (left, black traces) or SEVO (right, orange traces). c Activation and steady-state fast inactivation of hNav1.1 in the absence (white) or presence (orange) of sevoflurane. Current-voltage relationship of channel activation (normalised conductance; G/G_max) and inactivation (I_Na/I_Namax). Sevoflurane significantly shifted inactivation (V_½-inact) by −4.6 ± 1.1 mV (P = 5.6 × 10⁻⁸, n = 11; Supplementary Table 1). Sevoflurane also caused a small but significant shift in activation (Supplementary Table 1). d Representative overlaid traces of slow inactivation test pulses in CTL (black) or SEVO (orange), assessed using a three-pulse protocol (inset). Peak I_Na of P3 was normalised to P1 to isolate the slow-inactivated fraction (P3/P1). Notably, a non-inactivating fraction was observed at depolarised prepulse potentials, consistent with that reported for hNav1.1 using 30-s conditioning prepulses⁷⁸. e Normalised I_Na fitted with a Boltzmann function. Sevoflurane significantly shifted V_{½-slow-inact} by −9.3 ± 1.9 mV (P = 9.0 × 10⁻⁸, n = 10). f Molecular docking of sevoflurane to the NavMs Y143 binding pocket. Top pose (stick, coloured by heteroatom) overlays closely with sevoflurane in the NavMs-SEVO crystal structure (magenta). g Top docking pose of sevoflurane (stick, coloured by heteroatom) in the DI-DII pocket of hNav1.1 (space-filled model). h Top docking pose of sevoflurane in the DIII-DIV pocket of hNav1.1 (space-filled model). i F902A (left) in DII and F1682A (right) in DIV mutations introduced into hNav1.1 abolished channel function. Data shown as mean ± SD; drug effects vs. control were tested by paired two-tailed Student’s t-test (****P < 0.0001). n indicates the number of individual cells from ≥2 independent transfections. Source data are provided as a Source Data file.

The state-dependent effects of sevoflurane on hNav1.1, together with the preservation of sevoflurane binding in NavMs Y143F, suggested that homologous phenylalanine-containing pockets in hNav1.1 could contribute to the structural basis of this modulation. We therefore used in silico docking to examine whether the corresponding pockets in hNav1.1 could accommodate sevoflurane. As a control, sevoflurane was docked around the NavMs Y143 binding site, where the top-scoring pose closely matched that observed in the crystal structure (Fig. 6f), with a docking score of −4.5 kcal/mol, consistent with a weak VGSC–VA interaction. We then docked sevoflurane into the two predicted pockets in hNav1.1, at the DI–DII interface containing F902 and the DIII–DIV interface containing F1682. In both cases, the top-scoring poses occupied these pockets (Fig. 6g, h) and overlapped with the sevoflurane pose in the NavMs-SEVO structure (Supplementary Fig. 5). The corresponding docking scores were −4.1 kcal/mol for the DI–DII pocket and −5.3 kcal/mol for the DIII–DIV pocket. Together, these data identify hNav1.1 as a functional target of sevoflurane and suggest that homologous hydrophobic pockets to those found in NavMs and NaChBac contribute to specific aspects of its modulation.

Alanine substitution of predicted binding-site residues F902 and F1682 abolishes hNav1.1 currents

Finally, we probed for the functional effects of the individual F902A and F1682A mutations in hNav1.1 expressed in mammalian HEK293T cells. Unlike WT hNav1.1, expression of the F902A and F1682A channels produced no measurable Na⁺ current, indicating that each mutation abolished hNav1.1 expression or function (Fig. 6i, left and right). In contrast, mutation of S259 (S259A) in DI and L1359 (L1359A) in DIII, residues at equivalent positions to the phenylalanine residues in domains DII and DIV in hNav1.1, had minimal effect on Na⁺ currents compared to WT hNav1.1 (Supplementary Fig. 6), suggesting that the loss of current in the F902A and F1682A mutants is more indicative of impaired gating than of a general defect in channel expression or localisation.

Discussion

We employed BacNavs as established models to identify and characterise atomic-resolution structures of a volatile anaesthetic bound to a voltage-gated Na⁺ channel. Sevoflurane occupied an intramembrane pocket formed by residues from the VSD and the PM of adjacent subunits when in the inactivated state. Alanine substitution that abolished sevoflurane binding at this site selectively eliminated the sevoflurane-induced hyperpolarising shift in the voltage dependence of steady-state inactivation observed in WT channels. Sevoflurane produced comparable hyperpolarising shifts in steady-state fast and slow inactivation in the human neuronal isoform hNav1.1, which contains homologous hydrophobic pockets that accommodated sevoflurane in molecular docking simulations. However, alanine mutation of the corresponding binding site residues in hNav1.1 yielded non-conducting channels, demonstrating the functional importance of these sites but preventing direct functional analysis of their contributions to VA modulation.

Physiological resting membrane potentials in neurons maintain VGSCs in a dynamic equilibrium between their functional states. Negative shifts in the voltage-dependence of steady-state inactivation reduce the fraction of resting channels available for activation at the resting potential by increasing the number of channels in an inactivated state. This reduces overall neuronal excitability, thereby reducing stimulation-evoked synaptic vesicle exocytosis and synaptic transmission. One contributing mechanism to reduced exocytosis is impaired action potential propagation and presynaptic depolarisation due to VGSC inhibition, which reduces subsequent Ca²⁺ influx and thereby suppresses synaptic vesicle release, a major mechanism of VA effects^{18,49,51–54}.

BacNavs only populate slow inactivated states, but their relevance in this study is not necessarily restricted to slow inactivation, as the conserved VSD-PM pocket contains energetic determinants also relevant to fast inactivation in hNav. Evidence from hNav disease mutations indicates that the S5 helix, which forms part of the sevoflurane binding pocket and is not involved in the isoleucine, phenylalanine, methionine (IFM)-mediated interaction that stabilises the fast inactivated state, acts as an energetic coupling element capable of biasing channels towards both fast and slow inactivated states^55,56, while disease mutations in the VSD and PM of hNav1.1 primarily affect fast inactivation⁵⁰. This is supported by a NavMs-riluzole co-crystal structure that identifies a PM binding site required for riluzole modulation of steady-state inactivation in NavMs and fast inactivation in hNav1.4³². Also, the available cryoEM structure of hNav1.1 used here as the template for VA docking has the IFM motif engaged, consistent with our simulation being performed on the fast-inactivated state. Our demonstration that sevoflurane shifts the voltage-dependence of both fast and slow inactivation in hNav1.1 provides direct experimental support for this model and VA modulation of steady-state slow inactivation in a neuronal Nav channel. Together, these observations support a model in which intramembrane pockets at the VSD-PM interface act as conserved energetic sites through which small hydrophobic ligands can bias occupancy of fast- and slow-inactivated states, even when the underlying inactivation mechanisms differ across channel families.

The interaction between sevoflurane and NavMs at the Y143 site involves binding within a membrane-embedded hydrophobic pocket on the lipid-exposed surface, which would require access via a membrane-assisted pathway. Consistent with this, VAs are highly lipophilic and partition (>99%) from the aqueous phase to the membrane phase in simulations of ion channels in model membranes^34,37,57, facilitating access to intramembrane protein surfaces³⁷ and hydrophobic intramembrane pockets^32,36,57. In simulations involving VGSCs, such membrane partitioning enables VA molecules to reach membrane-embedded fenestrations, through which they can access and block the channel pore^34,37, a route functionally supported by electrophysiological studies showing a tonic component to inhibition of Nav isoforms by VAs from closed/resting states¹³. Access of sevoflurane to the Y143 site is therefore likely to occur by a similar membrane-mediated route. Any functional consequences of membrane-assisted binding linked to modulation of ion channel functions related to anaesthesia³ provide a molecular basis for the positive correlation between VA potency and lipophilicity as evident in the Meyer-Overton correlation^58–60. Additionally, our structure reveals that sevoflurane must displace a lipid molecule in order to bind at the Y143 site. Because lipid interactions can have modulatory effects on VGSCs⁶¹, we cannot exclude an indirect effect in which VAs perturb channel activity through the displacement of lipids necessary for normal channel function. This possibility could help explain the structural diversity of compounds that can produce the same functional effects on VGSCs by interaction at a common site, which in our study ranged from the 2-carbon haloalkane halothane to the 4-carbon haloether sevoflurane.

While the Y143 binding site is the only site identified at atomic resolution, our data support the existence of additional binding sites, possibly of lower affinity and less amenable to identification by high-resolution structural approaches. In ¹H STD-NMR experiments, removal of sevoflurane binding at this site in NavMs Y143A produced only a modest reduction in ligand binding, consistent with loss of one of multiple interaction sites, and in functional studies with NaChBac, only one sevoflurane effect was abolished in Y156A relative to WT. This is in agreement with previous MD and NMR studies on BacNavs that propose multiple VA-binding sites, including those near the activation gate and within the PM^34,36,37. Thus, we hypothesise that the Y143 site plays a key role in the stabilisation of the inactivated state by VAs, whereas other VA-induced effects on channel function arise from interactions at distinct sites, resulting in multiple site-dependent functional effects.

We applied a combined structure-function approach to match specific binding sites to the functional effects of VAs on VGSCs, thereby elucidating molecular interactions that contribute to anaesthetic actions. Identification of these sites through structural pharmacology is critical for further improvements in the design and development of novel anaesthetics using high-throughput screening and structure-activity studies of VGSC anaesthetic binding sites^62,63. Identification and characterisation of function-modulating anaesthetic binding sites in VGSCs could provide an approach to precision anaesthesia, and also be leveraged for drug development to treat VGSC channelopathies, including epilepsies, arrhythmias, myotonias and neuropathic pain, where modulatory sites affecting single-channel properties are considered desirable drug targets^64,65.

Eukaryotic VGSCs are essential to physiological processes that require electrical signalling. Given their sensitivity to VAs at clinical concentrations, it is likely that these ion channels are affected by VAs during general anaesthesia, as supported by many pharmacological studies^62,63. Stabilisation of fast- and slow-inactivated states by VA binding to human Navs at binding sites equivalent to the Y143 site found in BacNavs provides a structure-based mechanism for the alterations in synaptic transmission, neuronal excitability and network interactions that underlie general anaesthesia.

Methods

Molecular biology

Site-directed mutagenesis of the NavMs DNA sequence contained within the pET-15b plasmid vector was performed to create the mutants described in this study (Quikchange, Agilent), with the primers listed in Supplementary Table 3. Plasmids generated were chemically transformed into NEB® 5-alpha competent E. coli (New England Biolabs) with mutations verified by automated DNA sequencing (Eurofins Genomics) before transformation of correct plasmids into Overexpress™ C41(DE3) Chemically Competent E. coli cells (Lucigen) for protein expression.

Protein expression and purification

NavMs proteins were expressed in C41(DE3) E. coli cells grown in Luria Broth containing 100 µg/ml ampicillin, in shaking flasks at 37 °C. To induce NavMs protein expression, when the cells reached an optical density (OD₆₀₀) of 0.8, 0.5 mM IPTG was added, followed by further incubation for 3.5 h. Cell pellets were resuspended in 20 mM Tris, pH 7.5, 150 mM NaCl, 1 mM MgCl₂ and cells were broken by pressure. Cell debris was removed by centrifugation at 20,000 x g for 30 min, and the resulting supernatant was spun at 195,000 x g for 2 h to pellet membranes. Protein was extracted from the membranes using a buffer containing 20 mM Tris, pH 7.5, 300 mM NaCl, 20 mM imidazole and 1.5% DDM (Anatrace) on a rotating shaker at 4 °C for 2 h. Solubilised proteins were loaded onto a 1 ml HisTrap HP column (Cytiva Life Sciences), which was washed, and detergent was exchanged on the column with buffer containing 20 mM Tris–HCl, pH 7.5, 300 mM NaCl, 50 mM imidazole and 0.52% HEGA-10 (Anatrace) prior to protein elution using the same buffer with 1 M imidazole. The eluate volume was reduced, and imidazole concentration lowered to ~50 mM by sequential adding of buffer without imidazole using a 100 kDa cut-off concentrator (Amicon). NavMs proteins were treated with thrombin protease (Merck) for 16 h at 4 °C, then purified by size exclusion chromatography using a Superdex 200 10/300 column (Cytiva Life Sciences). NavMs was concentrated to 10 mg/ml and either used directly or stored at −80 °C. NavMs mutants were concentrated to 10 mg/ml, accounting for the different extinction coefficients caused by the mutations.

BTFA-labelling

Following purification, the NavMs C52A/T19C protein was diluted to 10 µM. 5 mM BTFA (Merck) was added, and the mixture was incubated at 4 °C for 24 h with gentle rotation. Free BTFA was removed from protein by gel filtration using a Superdex 200 10/300 column, and gave a protein peak identical in elution volume when compared to untreated protein. Complete removal of the probe was confirmed by the lack of a sharp peak at the BTFA fluorine resonance in the 1D ¹⁹F NMR spectrum of the NavMs-BTFA sample.

Sevoflurane preparation

Sevoflurane was used at an aqueous concentration of 0.57 mM, corresponding to twice the minimum alveolar concentration (2 MAC)⁶⁶, where MAC (in vol%) is the concentration needed to prevent movement in 50% of subjects in response to a painful stimulus (comparable to the ED₅₀).

Saturation Transfer Difference Nuclear Magnetic Resonance (STD-NMR)

All NMR spectra were acquired at 277 K. ¹H Experiments were performed with 0.57 mM (2 MAC) sevoflurane in buffer (20 mM Tris-Cl, 300 mM NaCl, 0.52% HEGA-10) +/− 5 µM NavMs protein and 5% D2O. ¹⁹F experiments we performed in the same conditions, but using 50 μM fluorine-labelled NavMs protein. All STD NMR experiments were performed on a Bruker Biospin Avance IIIHD 700 spectrometer, equipped with a 5 mm TCI cryoprobe, using the acquisition software Topspin 3.5 (Bruker). NMR spectra were acquired by collecting alternating on- and off-resonance spectra with saturation achieved using a train of Gaussian-shaped pulses of 50 ms duration with a peak B1 field of 500 Hz determined over randomised collections at the saturation time point specified. ¹H STD spectra were acquired with a spectral width of 16 ppm (~11,160 Hz), 16,384 data points, and a recycle delay of 10 or 12 s, depending on acquisition time. 64 scans were averaged per saturation condition (ON and OFF). ¹⁹F STD spectra were acquired similarly but with a spectral width of 140 ppm (92,592 Hz), 320 averages, 32,768 data points, and a 10 s recycle delay. Data were processed in Topspin 3.5 and fitted to the mono-exponential equation:

$$\mathrm{STD}\left(\%\right)=\mathrm{STD}\max \left[\left(\right)1-\exp \left(-\mathrm{ksatt}\right)\right]$$ 1

where STD (%) = (Voff – Von)/Voff x 100. Voff and Von are the peak integrals from the off and on resonance saturation transfer NMR spectra. STDmax represents the STD at the maximal plateaued value.

¹H STD NMR competition experiments were performed similarly to those described above, but with 0.57 mM sevoflurane and 1 µM NavMs in mixtures containing competing concentrations of desflurane (~20 mM), isoflurane (~23 mM) or halothane (~46 mM).

Crystallography

Crystallisation was undertaken using sitting drops at 4 °C containing 75 nl of crystallisation medium and 75 nl of stock NavMs F208L or mutants (10 mg/ml) dispensed using a mosquito® crystal robot (SPT Labtech). Apo-crystals were grown in crystallisation media containing 30–35% PEG 300, 100 mM HEPES (pH 7) and 0–100 mM NaCl. For NavMs-SEVO crystals, sevoflurane was saturated into reservoir solutions of crystal-containing wells and incubated with the crystals for up to 1 h with periodic harvesting and flash freezing of individual crystals in liquid nitrogen.

Data collection and processing

Crystal optimisation was required in this study. X-ray diffraction data were collected for crystals using the beamlines P13 (EMBL, Hamburg, Germany), ID23-1 and ID30A-1 (ESRF, Grenoble, France) and I24 and I04 (Diamond Light Source, Oxford, UK) working at wavelengths between 0.886 and 0.9795 Å. Indexing and integration were performed with the XIA2-DIALS pipeline^67,68, followed by scaling and merging with AIMLESS⁶⁹. Molecular replacement was performed with PHASER⁷⁰ using the search model (PDB ID: 6SX5), followed by model building in COOT⁷¹ and structure refinement using REFMAC⁷². Polder Omit⁴⁴ maps were created in Phenix⁷³. Graphic illustrations were produced using PYMOL⁷⁴ and UCSF Chimera⁷⁵.

Long wavelength X-ray crystallography

Data were collected on the long-wavelength MX beamline I23 at Diamond Light Source at a wavelength of 2.755 Å (4.5 keV). This wavelength, although not optimal for chlorine (K-edge 4.4 Å, 2.82 keV), was selected as a compromise between the increasing signal from chlorine as it reaches its K-edge and the decreasing data quality that results from increasing X-ray absorption at longer wavelengths. Each dataset consisted of a 360° rotation with an exposure time of 0.1 s per 0.1° image. Crystallographic data processing was performed as above. Anomalous difference maps were generated using the programme Anode⁷⁶.

NavMs and NaChBac cDNA constructs and cell culture

The NavMs-pTracer-CMV2 and NaChBac-pTracer-CMV3 plasmids carrying the separately expressed eGFP gene were used for whole-cell electrophysiology^35,77. Standard site-directed mutagenesis protocols using the QuikChange Lightning kit (Qiagen, Germantown, MD) were used to create mutants in NavMs (Y143A and Y143F) and NaChBac (Y156A). The entire open reading frames of successful cDNA clones were confirmed by Sanger sequencing.

NavMs and NaChBac were expressed in human embryonic kidney 293 T (HEK293T) cells to minimise interference from endogenous Na⁺ currents and facilitate electrophysiological isolation. HEK293T cells (CRL-3216 from ATCC, Manassas, VA) were maintained in high-glucose Dulbecco’s modified Eagle medium supplemented with 10% (vol/vol) foetal bovine serum (FBS), 2 mM GlutaMAX (Life Sciences, Waltham, MA) and 1% penicillin-streptomycin (Life Sciences). Passage numbers between 4 and 30 were used. Cells were seeded into 35-mm dishes and transfected with the respective NavMs cDNA after 1–3 days using Lipofectamine 2000 (Life Sciences) according to the manufacturer’s protocol. One day after transfection, cells were released with TrypsinLE (Life Sciences) and replated at a lower density onto 12-mm round #1.5 glass coverslips (Warner Instruments), a minimum of 3 h before recording isolated adherent cells with eGFP fluorescence.

Human Nav1.1 cDNA constructs and cell culture

Na⁺ currents from human Nav1.1 (hNav1.1) were recorded following transient expression in HEK293T cells as previously described⁷⁸. HEK293T cells were maintained and plated using the same workflow as described for the bacterial channel expression experiments. hNav1.1 (accession number NM_001165963) was kindly provided by A. L. George Jr. (Northwestern University) and obtained from AddGene⁷⁹. Nav1.1 was co‑transfected with human β1- and β2-subunits (10:1:1 ratio⁷⁸) using Lipofectamine 2000. Standard site-directed mutagenesis protocols using the QuikChange Lightning kit (Qiagen) were used to create the hNav1.1 mutants (for primer details see Supplementary Table 3). The entire open reading frames of successful cDNA clones were confirmed by Sanger sequencing.

Whole-cell patch-clamp electrophysiology

Pipettes were pulled from standard borosilicate glass (1.5 mm OD/0.86 mm ID; Sutter Instrument) to a resistance of 3–4 MΩ (when filled) for NavMs/NaChBac or 1.5–2.5 MΩ for hNav1.1 using a P-1000 puller (Sutter Instrument). Whole-cell voltage-clamp electrophysiology was performed⁸⁰ using an AxoPatch 200B amplifier (Molecular Devices) connected to a DigiData 1550B analogue-to-digital converter (Molecular Devices). Signals were sampled at 20 kHz and filtered at 5 kHz, respectively. Series resistance was corrected by 85–90%. Capacitive current transients were cancelled by the internal amplifier circuitry, and leak currents were subtracted using a standard P/4 protocol applied after the desired stimulus. Input resistance (R_in) was determined from the steady-state current response to a 10 mV voltage step from the holding potential, calculated as R_in = ΔV/ΔI, where ΔI is the difference between the baseline holding current and the steady-state current during the step. R_in values for all recordings are reported in Supplementary Table 1. For NavMs and NaChBac, cells were continuously perfused with extracellular solution at room temperature (24 °C) containing (mM) 150 NaCl, 10 Hepes, 1.8 CaCl₂ and 1 MgCl₂, adjusted to pH 7.4 with NaOH. Pipette solutions contained (mM) 110 CsF, 30 NaCl, 10 Hepes, 5 ethylene glycol tetraacetic acid (EGTA), adjusted to pH 7.30 with CsOH. The osmolality of all solutions was balanced to 300 ± 3 mOsm/kg H₂O with sucrose. For hNav1.1, cells were continuously perfused with extracellular solution at room temperature (24 °C) containing (mM) 145 NaCl, 10 Hepes, 1 MgCl₂, 4 KCl, 1.8 CaCl₂, adjusted to pH 7.35 with NaOH. Pipette solutions for hNav1.1 contained (mM) 110 CsF, 10 NaF, 10 Hepes, 2 ethylene glycol tetraacetic acid (EGTA), 20 CsCl, adjusted to pH 7.30 with CsOH. The osmolality of all solutions was balanced to 310 ± 3 mOsm/kg H₂O with sucrose. Saturated sevoflurane (Abbott Laboratories) stock solutions were prepared in extracellular solution in gas-tight glass vials and further diluted in gas-tight Hamilton syringes (Hamilton) to the desired working concentration. Solutions in gas-tight syringes were delivered using a pressurised perfusion system (ALA Scientific Instruments) with Teflon tubing via a 200 μm diameter manifold tip positioned $~$ 200–300 μm from the recorded cell. After control recordings, sevoflurane was perfused before subsequent recordings and continuously thereafter until washout. Solutions were delivered through a pressurised perfusion system to minimise mechanical disturbance of cells during sevoflurane superfusion. Sevoflurane concentrations equivalent to clinically effective concentrations in mammals were 0.28 mM and 0.57 mM, equivalent to 1 or 2 times the minimum alveolar concentration (MAC)⁸¹ after correction to room temperature, and were confirmed by gas chromatography⁸².

Molecular docking

Molecular docking was performed using the docking platform Autodock Vina⁸³ in UCSF Chimera⁷⁵. Sevoflurane was docked into a box surrounding the sevoflurane binding site identified in NavMs (PDB ID: 6SX5) and the predicted cognate binding sites in human Nav1.1 (PDB ID: 7DTD) using a receptor search volume of 20 × 20 × 20 Å and selecting for generation of the top 10 binding modes for each search. Only the top binding pose and associated docking score produced at the sites in NavMs and Nav1.1 are reported. The reported scores (in kcal/mol) represent empirical estimates (docking scores) from the programme’s scoring function, and are used to rank ligands based on predicted binding affinity. These values do not represent rigorous free energy calculations and should not be interpreted as absolute binding free energies. No post-docking free energy refinements (e.g., MM/GBSA or FEP) were performed.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

H.C.H., B.A.W., D.H. and K.F.H. conceived the project; D.H. and K.F.H. designed the research; D.H. performed the molecular biology, protein expression and purification and chemical modifications for the structural studies; G.K. and D.H. performed and analysed NMR spectroscopy; D.H. performed and analysed X-ray crystallography; V.B.M. and D.H. performed and analysed the long-wavelength X-ray crystallography; K.F.H. performed the molecular biology, cell culture and electrophysiology using NavMs, NaChBac and human Nav1.1; D.H. and K.F.H. wrote the paper under the supervision of B.A.W. and H.C.H.

Peer review

Peer review information

Nature Communications thanks Joseph Ransdell, who co-reviewed with Zahra Hosseini Farjam; Vincenzo Carnevale, Pei Tang and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

US National Institutes of Health grant R01 GM58055 to H.C.H. and B.A.W., a British Journal of Anaesthesia International Collaborative Grant to H.C.H., B.A.W. and K.F.H., and a Rosetrees Trust grant (ref: CF2\100001) to B.A.W. This work was also supported by the Francis Crick Institute through the provision of access for D.H. and B.A.W. to the MRC Biomedical NMR Centre, which is funded by Cancer Research UK (CC1078), the UK Medical Research Council (CC1078), and the Wellcome Trust (CC1078).

Data availability

The atomic coordinates and crystallographic structure factors generated for Apo-NavMs and NavMs complexed with sevoflurane (NavMs-SEVO) have been deposited in the protein data bank (PDB) under the accession codes 9GTQ for Apo-NavMs and 9GV1 for NavMs-SEVO, with data collection and refinement statistics details provided in Supplementary Table 2. The PDB codes of the previously published structures used in this study are for NavMs 6SX5 and for Nav1.1 7DTD. Source data are provided with this paper as a separate data file.

Competing interests

H.C.H. is editor-in-chief of the British Journal of Anaesthesia. The other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74518-7.