The structural dynamics and molecular coupling in the slow inactivation of a prokaryotic voltage-gated sodium channel

Abstract

Slow inactivation in voltage-gated sodium channels (Navs) is crucial for regulating cell excitability. Many studies indicate that the selectivity filter (SF) serves as the inactivation gate, but the structural mechanism and coupling between the distal primary and inactivation gates remain poorly understood. Our single-molecule FRET studies reveal that the NavAb SF transits among three conformational states, with depolarization enriching a high FRET state, potentially the slow inactivated conformation. We show L176 in the SF and T206 in the S6 helix as key residues that form a steric-hindrance coupler. Crystallographic analysis reveals that bulkier L176F overrides the opening tendency of an S6 C-terminal deletion (ΔC230) and closes the primary gate, while smaller T206A partially uncouples the slow inactivation gate. SmFRET data confirm that L176F restores the high FRET population depleted by ΔC230 or the open-pore blocker lidocaine. Our results indicate that ‘conformational talk’ through L176-T206 underlies slow inactivation in prokaryotic Navs.

Slow inactivation of voltage-gated sodium channels is important for regulating cellular excitability. Here the authors probe the conformational dynamics in the selectivity filter using smFRET and reveal the critical residue pair that mediates conformational communication between the primary and slow inactivation gates

Introduction

Voltage-gated sodium channels (Navs) mediate Na⁺ diffusion across cell membranes to initiate action potentials, which are essential for nerve firing or the contraction of skeletal or cardiac muscle^1,2. Dysfunction in Nav causes numerous channelopathies, including epilepsy, chronic pain syndromes and cardiac arrhythmias^3,4. Immediately after activation, Navs quickly enter the fast inactivation state, allowing the cell membrane potential to be repolarized and initiating another action potential⁵. The ‘IFM(Ile-Phe-Met)’ motif between subdomains III and IV of Nav channels was identified to be involved in fast inactivation^5,6. Nav channels also undergo a distinct slow inactivation, which occurs over seconds to minutes, that serves as a critical mechanism to prevent persistent nerve firings or muscle contractions⁷.

Prokaryotic Nav channels and eukaryotic orthologs share similar core structures^8,9. Due to their homomeric subunit arrangement and amenability to crystallographic and cryo-EM analyses, studies on prokaryotic channels have provided invaluable mechanistic insights into the ion selectivity, voltage gating and inactivation of Nav channels more broadly^10–13. Prokaryotic Nav channels lack the IFM motif and do not exhibit the fast inactivation typically seen in eukaryotic Nav channels, but their conductance to sodium, upon repetitive activation, does gradually decrease in the time scale of seconds to minutes, a behavior similar to slow inactivation in eukaryotic Nav channels^9,14,15. Electrophysiological analyses have suggested that the collapse of the selectivity filter (SF) pore perhaps serves as a shared structural mechanism for slow inactivation in both prokaryotic and eukaryotic Nav channels^12,15–17. However, the structural basis for slow inactivation remains elusive. While electrophysiological studies and analogies to C-type inactivation in potassium channels have led to the hypothesis of SF pore collapse, direct structural evidence for such a transition in Nav channels is currently absent. Notably, all available crystal and cryo-EM structures of NavAb show almost identical SF conformations across different gating states, leaving the physical reality of SF-mediated inactivation as a subject of speculation. But the structural stability of NavAb has been demonstrated in various physicochemical analyses, including smFRET¹⁸ and high-speed atomic force microscopy (HS-AFM)¹⁹. It has been previously established that mutation at the T206 residue in the S6 helix or deletion of the S6 C-terminus, and open-pore blocker lidocaine attenuate slow inactivation^16,20. This wealth of existing data provides an essential foundation for implementing and interpreting smFRET measurements.

Here, we address this critical gap by using smFRET to capture the real-time dynamics of the NavAb SF and the changes induced by activating voltages, thereby informing molecular mechanisms driving slow inactivation. Our smFRET measurements on NavAb channels reconstituted into liposomes show that the SF spontaneously transits among 3 distinct conformational states, with activating voltages enriching the high FRET conformation. Moreover, we show L176 as a critical residue in coupling the primary and slow inactivation gates. Our smFRET studies, combined with electrophysiological and crystallographic analyses, consistently indicate that the high FRET, constricted conformation in the SF represents the slow inactivated conformation, with physical hindrance between L176 and T206 mediating conformational crosstalk between the primary gate at bundle crossing and the slow inactivation gate at the SF.

Results

Activating voltages alter the conformational landscapes of the NavAb SF

Although collapse of the SF has been a long-proposed structural model for slow inactivation, it has remained largely speculative in Nav channels. Despite extensive structural studies, static snapshots have failed to detect distinct inactivated SF conformations. Therefore, our aim was not merely to reaffirm the expected model, but to provide direct, dynamic observations of SF structural rearrangements during slow inactivation, leveraging the strength of smFRET to resolve transitions that static methods have previously missed. To carry out these studies, we constructed and purified functional tandem dimeric NavAb channel proteins¹⁸, which carry cysteine mutations at either E189 or V190 residues at the P2 helix, so that the two cysteines in a tetrameric channel are located in diagonally apposed subunits (Fig. 1a). To ensure that the cysteine mutations for FRET labeling did not perturb channel function, we characterized the electrophysiological properties of the tandem dimeric constructs. Both the full-length and C-terminal deletion (ΔC230) tandem dimers carrying the V190C mutation exhibited I-V relationships, steady-state inactivation, and tail current activation curves comparable to those of their respective non-labeled backgrounds (Supplementary Fig. S1). These results confirm that any specific structural transitions captured by smFRET occur within a functionally intact channel framework. The Cy3 and Cy5 FRET fluorophore pair carrying maleimide reactive groups were conjugated to the two cysteine residues, so their FRET changes will report the real-time conformational changes of the NavAb SF. We applied 4 different electrical potentials (−200, −85, 0, and 120 mV) on NavAb channel proteins in liposomes, generated by intra- and extraliposomal K⁺ gradients in the presence of valinomycin (Fig. 1b). This method has been successfully used in our previous work to examine voltage-dependent conformational changes in voltage-gated proton channels^21,22. The wild-type NavAb is inactivated even at a very negative potential of approximately −120 mV, but the N49K mutation shifts the activation curve by ~+75 mV to a half-activation voltage of −22 mV^10,14. In this study, the N49K background mutation (named BG in all figures if no additional mutations) was introduced into all mutants, so they were in deactivated, partially activated, or fully activated states at −200, −85, 0, or 120 mV, respectively. Importantly, tandem dimeric NavAb constructs carrying the V190C mutation expressed in Sf9 cells produced functional channels (Supplementary Fig. S1b, e), and purified tandem dimeric NavAb proteins were also functional, as reported in our previous study¹⁸. Our smFRET measurements clearly showed that FRET between Cy3/Cy5 fluorophores labeled at V190C residues at the diagonal subunits exhibits spontaneous transitions between 3 distinct states under all voltages tested, suggesting that the NavAb SF pores may undergo marked conformational changes (Fig. 1c, Supplementary Fig. S2). Importantly, similar changes were also observed by switching the fluorophore labeling sites to the adjacent E189C (Fig. 1c and Supplementary Fig. S3a). FRET histograms from both the E189C and V190C labeling sites are fitted well with 3 Gaussian components, which resulted in over 99% decreases in the sum of squared errors in comparison to that fitted with one (Fig. 1d). The 3 Gaussian peak centers determined from the E189C labeling sites are slightly higher than those from the V190C sites, consistent with the diagonal E189C residues being closer than diagonal V190C residues (Fig. 1a, e). The high FRET peaks revealed by our smFRET measurements are very close to those predicted from NavAb structures¹¹, and FRET changes between high and medium FRET peaks also match well with those predicted by cryo-EM structures of a eukaryotic NavEh channel at presumptive conductive and slow inactivation states¹². We included NavEh for this comparison because structural insights into the slow-inactivated state of Nav channels remain extremely scarce. Notably, currently available NavAb structures show almost identical SF conformations, which cannot account for the large FRET changes observed in our study. Therefore, NavEh provides a crucial proxy for the structural rearrangements that accompany slow inactivation. More importantly, switching labeling sites to G153, the extracellular end of the S5 helix away from the SF, no voltage-dependent FRET changes were observed (Supplementary Fig. S3c, d), consistent with smFRET data from V190 and E189 sites reflecting conformational changes in the NavAb SF, rather than fluorophore photophysics caused by voltage changes. To analyze the structural dynamics using smFRET data from both E189C and V190C labeling sites together, we used a 3-state kinetic model with FRET peak centers fixed at 0.25, 0.55, and 0.8, named as low (L), medium (M), and high (H) FRET states (Fig. 1f). From both labeling sites, our smFRET contour maps and histograms indicated that the low FRET conformations are prodominant at a resting voltage of −200 mV, while high FRET conformations were enriched by more positive, activating voltages suggesting that high FRET conformations represent the NavAb SF in the inactivated state (Fig. 1g, h). Importantly, voltage-dependent changes in FRET populations closely match the steady-state current-voltage relationship of slow inactivation (Fig. 2a), further supporting the association of conformational changes in the selectivity filter with slow inactivation in the NavAb channel.

Fig. 1. Activating voltage promotes the constricted conformation at the NavAb selectivity filter.

a Cartoon of the NavAb pore domain, with 2 subunits removed for clarity (PDB: 5VB2). The pore-forming helices are colored cyan, and the P1/P2 helices of the selectivity filter are colored pink. The C-terminal helices deleted in the ΔC230 mutant are colored yellow, and the L176 and T206 residues mutated are highlighted as red spheres. The Cα carbon atoms of E189C and V190C, at the diagonal subunits for fluorophore labeling, were highlighted by orange and blue spheres, respectively. b The experimental setup and sequence for examining the conformational dynamics of the NavAb selectivity filter using smFRET. The NavAb proteoliposomes were immobilized on the PEGylated surface (1) through neutravidin (2) and biotinylated anti-histag or anti-FLAG tag antibodies (3), so only proteoliposomes (step 4) with the NavAb C-terminus facing outside were retained for smFRET imaging. Transliposomal K⁺ gradients generate liposome voltages using valinomycin (Val). c smFRET traces showed that FRET between Cy3/Cy5 fluorophores conjugated at E189C or V190C sites in the NavAb selectivity filter exhibits transitions among 3 distinctive FRET states (−85 mV). d SSE of E189C and V190C FRET histograms fitting with 1–5 Gaussian components. Error bars indicate uncertainties in the SSE estimates from bootstrapping (N = 1213 and 1577 for E189C and V190C sites, respectively). e The centers of FRET states from Gaussian fitting of smFRET histograms from the diagonal E189C and V190C labeling sites (N = 1213 and 1577, respectively). Error bars indicate the width of each Gaussian peak at 95% confidence level. The FRET efficiencies of the diagonal E189 and V190 predicted from structures of NavAb (6P6W) or diagonal E318 from the structures of NavEh (8J7F and 8J7M) were also included for comparison. f The kinetic model for analyzing smFRET traces collected from E189C and V190C labeling sites. Contour maps and histograms of smFRET traces collected from E189C (g) and V190C (h) labeling sites. The FRET state occupancy data (right panels), calculated as fractional areas from histograms, with error bars indicating uncertainties estimated as the fractional occupancy of each FRET state from the analyzed smFRET traces, obtained by bootstrapping the traces 1000 times.

Fig. 2. SmFRET measurements revealed conformational changes in the selectivity filter induced by mutations that alter slow inactivation in the NavAb channel.

a Steady-state inactivation of NavAb BG (N = 4), L176F (N = 4), L176W (N = 4), T206A (N = 4) and ΔC230 (N = 5) mutants. Normalized currents (I/I_max) are shown with standard errors. Imax values were obtained from sigmoidal fitting. b The contour maps and histograms of smFRET traces from E189C labeling sites collected without or with the L176F and T206A background mutations under −85 mV membrane voltages. c FRET state occupancies of smFRET traces from E189C labeling sites without or with the L176F and T206A background mutations under −85 mV and 120 mV membrane voltages. Error bars indicate uncertainties, estimated as the fractional occupancy of each FRET state from the analyzed smFRET traces, obtained by bootstrapping the traces 1000 times. Two-sided t-tests were performed to examine significance levels relative to BG, with * denoting p < 0.05 and ** denoting p < 0.001. d The contour maps and histograms of smFRET traces from V190C labeling sites collected without or with the L176F C-terminal deletion ΔC230 and both background mutations under −85 mV membrane voltages. e FRET state occupancies of smFRET traces from V190C labeling sites without or with the L176F, C-terminal deletion ΔC230 and both background mutations under −85 mV membrane voltages. Error bars indicate uncertainties, estimated as the fractional occupancy of each FRET state from the analyzed smFRET traces, obtained by bootstrapping the traces 1000 times. T-tests were performed to examine significance levels relative to BG, with * denoting p < 0.05 and ** denoting p < 0.001.

SmFRET measurements revealed conformational changes in the selectivity filter induced by mutations that alter slow inactivation in NavAb channels

The S6 helix is a long, uninterrupted α-helix that is involved in both activation and slow inactivation of the NavAb channel²³. The cytosolic helix bundle formed by four S6 helices of a tetrameric NavAb channel is a structural feature shared among prokaryotic Navs²⁴. It has been established that the S6 C-terminal helical bundle enhances the early phase of slow inactivation¹⁶. Consistently, in the present study, we found that deletion of this bundle from Q230 (Fig. 1a), named ΔC230, shifts the steady-state inactivation curve toward more positive potentials (Fig. 2a and Table 1). These results indicate that removing the S6 bundle attenuates the overall development of slow inactivation in the NavAb channel. In addition, substitutions at the S6 residue T206, which is close to the SF, significantly alter the channel’s slow inactivation kinetics; for instance, T206A and T206G mutations attenuate inactivation, while T206V potentiates it in the NavAb channel^16,24. The T206A mutation also shifts the steady-state inactivation curve toward more positive potentials than that of BG (Fig. 2a and Table 1). Furthermore, alanine substitutions of glycine residues at the intracellular helix bundle crossing have also been shown to slow the inactivation rate in various prokaryotic channels²². L176 at the NavAb SF is very close to T206; thus it may couple the conformational changes between the primary helix bundle crossing gate and the SF slow-inactivation gate. Indeed, our electrophysiological data indicated that mutations to bulky side chain residues Phe (L176F) and Trp (L176W) significantly altered the slow inactivation of NavAb channels (Fig. 2a, Supplementary Fig. S1a, c, g and Table 1). The voltage dependencies of L176F and L176W activation were almost unchanged, but their half-inactivation voltages were shifted negatively by over 30 mV (Table 1). We observed opposite changes in slow inactivation from the ΔC230 and T206A mutant (Fig. 2a, Supplementary Fig. S1a, d, h and Table 1)¹⁶. Electrophysiological analysis suggests that most channels are predominantly in the inactivated state at −85 mV (Fig. 2a). SmFRET measurements reveal the dynamic conformational landscape of individual molecules. Thus, the shifts in FRET state occupancies reflect the changes in the relative stability or the propensity of the SF to enter the inactivated conformation, rather than the molecules being locked in a static state.

Table 1.

Half-inactivation and activation voltages of NavAb mutants

	Inactivation				Activation
	Vh (mV)	slope factor	Imax	Sample N	Vh (mV)	slope factor	Gmax	Sample N
BG(N49K)	−111 ± 2	−11.8 ± 1.6	1.00 ± 0.03	4	−45.3 ± 1.5	3.73 ± 1.10	0.90 ± 0.03	4
V190C-BG dimer	−102 ± 2	−12.3 ± 1.9	1.02 ± 0.04	4	−38.2 ± 0.8	3.23 ± 0.88	0.85 ± 0.02	3
L176F	−160 ± 3	−11.6 ± 1.4	1.14 ± 0.10	4	−30.9 ± 3.2	10.5 ± 2.8	1.08 ± 0.05	7
ΔC230	−99.7 ± 1.0	−6.53 ± 0.93	0.99 ± 0.04	5	−35.3 ± 0.9	4.35 ± 0.70	0.87 ± 0.02	3
ΔC230/V190C-ΔC230 dimer	−118 ± 1	−10.7 ± 1.1	1.01 ± 0.02	6	−34.2 ± 0.8	4.66 ± 0.61	0.94 ± 0.01	6
ΔC230/L176F	−128 ± 2	−9.22 ± 1.57	0.96 ± 0.04	3	−35.8 ± 1.4	4.49 ± 1.08	0.84 ± 0.03	6
L176W	−187 ± 31	−22.1 ± 5.8	2.41 ± 1.90	4	−33.7 ± 2.0	7.96 ± 1.70	0.82 ± 0.02	4
T206A	−90.9 ± 2.1	−12.0 ± 1.9	1.05 ± 0.03	4	−29.4 ± 2.3	10.3 ± 1.9	1.07 ± 0.04	4
L176F/T206A	−109 ± 2.4	−11.9 ± 2.0	1.00 ± 0.04	6	−27.9 ± 1.6	9.98 ± 1.40	1.15 ± 0.03	6

Half-inactivation voltage (V_h) is the membrane potential of 50% steady-state inactivation; Half-activation voltage (V_h) is the membrane potential of 50% tail-current activation. All results are expressed as mean ± SEM.

We examined the effects of the T206A, L176F, and ΔC230 mutations on the NavAb SF conformation using smFRET. Due to the absence of measurable currents in the L176W tandem dimer, this mutant was excluded from FRET analysis. The T206A mutation significantly depletes the high-FRET population at −85 mV, consistent with its functional effects to attenuate slow inactivation in NavAb channels. However, it had no effect on FRET distributions under strong activating voltages (Fig. 2b, c, Supplementary Fig. S3b). Dramatic changes, however, were observed from the ΔC230 mutant, where the high FRET population was almost absent, even under strong activating voltages, in sharp contrast to the T206A mutant (Fig. 2d, e, Supplementary Fig. S2b). Surprisingly, the L176F mutation does not cause significant changes in the SF conformation, as indicated by smFRET collected from both the E189C and V190C labeling sites (Fig. 2b–e, Supplementary Fig. S2b, Fig. 3a, b). However, it strongly attenuates the ΔC230 effects by enriching the high FRET state and diminishing the low FRET state, consistent with its role in promoting slow inactivation (Fig. 2d, e). Collectively, our smFRET data from NavAb mutants support the idea that conformational changes are associated with slow inactivation, and that high FRET conformations may reflect the slow inactivation state in the NavAb SF.

Fig. 3. The molecular coupler mediates conformational changes between the selectivity filter and the S6 helix bundle crossing gate.

a Superimposition of the selectivity filter of crystal structures of the BG, L176F, ΔC230 and ΔC230/L176F mutants. The T206 and L176 residues, which couple the selectivity filter to the S6 helix, and V190 for fluorophore labeling, are highlighted as sticks. b Root Mean Square Deviation (RMSD) of pore domain residues of mutant channels, L176F, ΔC230 and ΔC230/L176F, against NavAb BG mutant carrying only the N49K background mutation. c Pore diameter profiles of NavAb BG, L176F, ΔC230 and ΔC230/L176F mutants. The baseline of the z-axis is the z-axis values of the main chain carbonyl oxygen atoms of the 176th residue. d Bottom views of the helix bundle cross gates of NavAb channels without (BG) or with the L176F, ΔC230, ΔC230/L176F and ΔC230/T206A background mutations. The protein is represented as a cartoon ribbon model. Side chains are displayed only for V214, which exhibited well-defined electron density. For M221, the corresponding bundle crossing layers are highlighted in magenta, and the distances between Cα atoms (both diagonal and adjacent) are indicated to illustrate the expansion or closure of the gate. They clearly indicated that the L176F mutation stabilizes the ΔC230 opening mutant at the closed state. Superimposition of the selectivity filter and the S6 helix of BG and ΔC230 (e), BG and L176F (f), BG and ΔC230/L176F (g), ΔC230 and ΔC230/L176F (h), ΔC230/L176F and ΔC230/L176F/T206A (i), viewing from figure1a right direction. The L176F mutation pushes the S6 helix toward the closed state conformation.

T206 and L176 mediate conformational talk between the slow inactivation and primary gates

To directly visualize conformational state associated with slow inactivation, we further obtained crystal structures of the NavAb channel carrying the BG mutant and channels containing additional mutations, including L176F, ΔC230, and T206A (Supplementary Tables S1, S2). Structural alignments indicated that, despite changes in their slow inactivation behaviors, the crystallographic SFs of L176F, T206A, and ΔC230 mutants were indistinguishable (Fig. 3a and Supplementary Fig. S4). The overall RMSD of all structures before the S6 helix is below 1 Å, indicating there is very minimal overall structural change in all mutants (Fig. 3b).

However, the RMSD of the ΔC230 mutant around the S6 helix was greater than 1 Å, suggesting a significant difference in the primary gate conformation (Fig. 3b). This structural divergence is clearly supported by the S6-helix omitted electron density maps (Supplementary Figs. S5, S6). The electron density map reveals that the primary gate of the ΔC230 mutant is strikingly different from that of the full-length channel (BG), adopting an opening state (Fig. 3c, d). To ensure a rigorous comparison of these gating states, we analyzed the distances between Cα atoms of M221. Interestingly, the introduction of the L176F mutation effectively stabilized the ΔC230 mutant in a clearly closed conformation (Fig. 3c, d). This finding is particularly significant because the S6 C-terminal deletion (ΔC230) normally acts as an ‘opening mutant,’ as evidenced by our structural and RMSD analysis. The fact that the L176F mutation—a structural modification at the SF (part of the slow-inactivation gate)-can override this inherent opening tendency and stabilize the distant primary gate in a closed state provides direct evidence of the structural communication, or ‘conformational talk,’ between these two gates. The electron density for the S6 C-terminus in the ΔC230/L176F mutant is more continuous than the full-length BG channel (Supplementary Fig. S6) and better resolved than the ΔC230 channel, providing visual evidence for the robust stabilization of the closed state by the L176F coupler.

While the L176F mutation effectively stabilized the primary gate in the closed state, the T206A mutation did not exhibit such a closing effect. Instead, the T206A substitution likely created a structural gap that increased the mobility of the S6 helix, making its C-terminus too flexible to be resolved in the crystal structure of ΔC230/T206A mutant (Fig. 3d and Supplementary Figs. S5, S6). T206 is one of the residues where the pore radius undergoes significant narrowing in the ΔC230 mutant compared to the full-length channel (Fig. 3c). The T206A mutation is likely to create a gap that increases the mobility of the S6 helix, resulting in the disappearance of the electron density of the S6 C-terminus in the ΔC230/T206A mutant. Our previous studies also showed that the T206A mutation tends to decrease structural resolution²⁵, a finding that we may now ascribe to this high mobility of the S6 helix.

Notably, the L176F effects on the primary gate were also observable in the crystal structure of the ΔC230/L176F/T206A triple mutant (Fig. 3d). In all these structures, residues at the 176 position were very close to T206 in the S6 helix (Fig. 3e–i, Supplementary Figs. S5, S6). In the ΔC230 mutant, T206 moves toward L176 (Fig. 3e), whereas in the L176F and ΔC230/L176F structures, T206 superimposes well with the BG structure (Fig. 3f, g). Their proximity to L176 was similar, except for the ΔC230 mutant, as indicated by the distances from the T206 Cβ carbon to the L176 Cγ, Cβ, and Cα carbons (Table 2). Therefore, in the ΔC230/L176F structure, the T206 is pushed away from L176 by more than 1 Å through steric hindrance with the T206, which repositions the S6 helix, closing the primary helix bundle crossing gate (Fig. 3e, h). A similar effect was observed in the ΔC230/L176F/T206A triple mutant (Fig. 3i). A similar, even stronger effect pushing the S6 helix was observed with L176W (Table 2), consistent with our electrophysiological data (Fig. 2a). However, the increased B-factor of the crystal structure and our inability to measure the L176W tandem dimer suggest that excessive steric hindrance may destabilize the overall channel structure.

Table 2.

Root mean square deviation (RMSD) and distance between the 176 residues and T206 of the NavAb channels

	RMSD with		Distance from T206 Cβ to
	BG mutant (Å)		Nearest atom	Cγ	Cβ	Cα
	176 Residues	Thr206	of 176 residues (Å)
BG	–	–	6.6	6.9	8.4	9.0
L176F	0.30	0.68	4.1	6.8	8.3	8.9
ΔC230	0.23	2.31	6.2	6.2	7.6	8.1
ΔC230/L176F	0.15	0.50	4.3	6.8	8.3	8.7
T206A	0.33	0.28	6.8	7.0	8.5	9.2
L176W	0.70	0.79	4.8	7.8	9.2	9.9

To further probe the correlated SF conformation and primary gate closure, we first examined whether the L176F mutation could reverse the effects of the ΔC230 mutation, which normally attenuates slow inactivation. Electrophysiological analysis showed that the ΔC230/L176F double mutant exhibited a significant negative shift in steady-state inactivation (SSI) compared to ΔC230 alone, effectively restoring the inactivation phenotype (Supplementary Fig. S1d, f). This is consistent with our smFRET and crystallographic data showing that L176F enriches high-FRET conformations and stabilizes the primary gate in a closed state. To further probe the physical basis of this coupling, we characterized the T206A/L176F double mutant. We hypothesized that if the interaction is mediated by steric hindrance between side chains, reducing the volume at position 206 (Thr to Ala) would weaken the coupling. Indeed, while L176F still promoted inactivation in the T206A background, the shift was markedly attenuated compared to the BG background (Supplementary Fig. S1h, i and Table 1). This observation, combined with the fact that the ΔC230/L176F/T206A triple mutant remains structurally closed, suggests that the L176-T206 interface functions as a volume-sensitive physical coupler, in which the bulky L176F side chain sterically constrains the S6 helix by pushing against T206, thereby coordinating the conformational state of the SF with closure of the primary gate.

This concept of a volume-sensitive physical coupler is supported by the attenuation of electron density in the omit maps for the distal S6 helix in both ΔC230/T206A and ΔC230/L176F/T206A mutants (Supplementary Fig. S6). In the structure of the full-length T206A mutant, continuous electron density is observed up to L224, similar to other structures, because the C-terminal helix bundle anchors the distal S6 helix (the primary gate). However, in ΔC230/T206A, the loss of C-terminal bundle fixation combined with the T206A mutation markedly increases distal S6 mobility, leading to density being attenuated from T206. Similarly, in ΔC230/L176F/T206A, although L176F provides partial stabilization, the absence of the native-counterpart interaction due to the alanine mutation weakens this effect, leaving substantial structural flexibility from I216 onward. Therefore, the volume of the space facing L176 and T206 affects the mobility of the S6 helix, which is thought to regulate slow inactivation properties.

Open-channel blocker lidocaine alters the NavAb selectivity filter dynamics in the same way as the C-terminal deletion open mutation

As a classical open-channel blocker of eukaryotic Nav channels, lidocaine prevents Nav1.5 and Nav1.7 channels from entering the slow inactivation state^20,26. Lidocaine has also been shown to block prokaryotic NavAb channels with high affinities at the inner cavity immediately beneath the SF pore²⁷. In the present work, we examined the effects of lidocaine on the SF conformation using smFRET. Lidocaine decreased the fractional occupancy of the high FRET conformation in a clear concentration-dependent manner, reaching saturation at 0.5 mM (Fig. 4a, b), with no significant changes in FRET state occupancies being observed among different voltages in the presence of 5 mM lidocaine (Supplementary Fig. S2c). The effects of lidocaine on NavAb SF conformation matched well with its functional effect of preventing channels from entering the slow-inactivation state, again consistent with the high FRET conformation reflecting the SF in the inactivated state. Interestingly, introducing the L176F mutation partly restored the high FRET population, in the presence of 5 mM lidocaine, an effect similar to that observed with the ΔC230 mutant (Fig. 4a, b). Transition kinetics between L, M, and H FRET states, assessed from FRET state occupancy data, indicated that lidocaine promotes transitions from L to H and M to H FRET states, with a half-effective concentration close to 0.05 mM, very close to the half-inhibition concentration (IC₅₀) of 0.1 mM reported by electrophysiological studies²⁷ (Fig. 4c). The transition kinetics data collectively suggest that high FRET conformation is the slow-inactivate state of the NavAb channel, while the medium FRET could be an intermediate conformation in the process of slow inactivation.

Fig. 4. Lidocaine modifies the conformation of the NavAb selectivity filter in a dose-dependent manner.

a, b FRET contour maps and histograms (a), state occupancies of smFRET data collected from V190C labeling sites, without (BG) or with the L176F background mutation, in the presence of different concentrations of lidocaine under 0 mV. Lidocaine depleted the high FRET population in a dose-dependent manner, while the L176F mutation, which enhances slow inactivation, restored the high FRET conformation of the selectivity filter. Error bars represent uncertainties, which were estimated as the fractional occupancy of each FRET state from the analyzed smFRET traces, obtained by bootstrapping the traces 1000 times. c Changes in equilibrium constants for transitions between L and M (K_LM), M and H (K_MH), and L and H (K_LH) FRET states with lidocaine concentration were fitted to the Hill equation with a Hill coefficient of n = 1. Error bars indicate uncertainties in the equilibrium constant K, calculated from state occupancy data of panel b by error propagation. Only transition equilibrium constants between M and H and L and H states are altered by lidocaine, with a half-inhibition concentration (IC₅₀) very close to 0.1 mM, close to that determined by electrophysiological studies.

Discussion

Slow inactivation is a key mechanism by which mammalian Navs modulate electrical excitability in nerves and muscles⁷, but the underlying structural mechanisms remain unclear. The prokaryotic NavAb channel also exhibits a slow inactivation behavior^9,16. Its structural stability for biophysical analyses, including HS-AFM¹⁸ and smFRET²⁰, has been well-established, providing an essential foundation for capturing real-time conformational transitions. In the present work, we conjugated Cy3/Cy5 FRET to the diagonal E189C or V190C residues at the P2 helix of the SF to examine conformational dynamics in the NavAb SF. Our smFRET results indicated that NavAb SF conformations are different from those seen in potassium channels, which exhibit a very stable high FRET conformation under normal ionic conditions^28,29. SmFRET data from both E189C and V190C labeling sites consistently showed 3 distinct FRET states. High FRET state occupancy was (a) enriched by strong activating voltages, (b) reduced by the S6 helix T206A and ΔC230 mutations, that attenuate slow inactivation, (c) diminished by lidocaine, which has been found to prevent Nav channels from entering the slow-inactivation state, and (d) partly restored by the SF mutation L176F, which promotes slow inactivation. We further calculated the equilibrium constants (Kij) for transitions between L and M (LM), L and H (LH), M and H (MH) using FRET state occupancy data. Our results indicated that activating voltages significantly promote increased occupancy of high FRET states; T206A and ΔC230 mutations, and lidocaine, promote increased occupancy of medium and low FRET states, with the L176F mutation reversing these effects (Fig. 5a, b). Our results support the idea that the high FRET constricted SF conformation may be associated with the slow-inactivation state of the NavAb channel, while lower FRET conformations may represent dilated – and potentially conductive - SF pore conformations (Fig. 5c). Since the Cy3/Cy5 fluorophore pair actually reports apparent distances between two diagonal P2 helices, the medium FRET state could represent an asymmetric selectivity filter pore, with one subunit adopting the high-FRET inactivated conformation and the other adopting the low-FRET conductive conformation.

Fig. 5. Selectivity filter conformational dynamics underlying slow inactivation in the NavAb channel.

Effects of voltage (a, upper panel), mutations under −85 mV (a, lower panel) or lidocaine under 0 mV (b) on equilibrium constants Kij of transitions between L and M (LM), M and H (MH), L and H (LH) FRET states. The transition equilibrium constants Kij were calculated from FRET state occupancy data in Fig. 1g, h, Fig. 2c, e and Fig. 4b, obtained from either the E189C or V190C labeling sites, as described in the materials and methods section. Error bars indicate uncertainties of Kij, calculated from state occupancy data using error propagation. c Structural basis underlying slow inactivation in the NavAb channel. The low and high FRET populations correspond to the conductive and slow inactivation states, respectively. T206A, lidocaine (lido), and C-terminal deletion (ΔC230) promote transitions from high to medium or low FRET conductive state conformations, while depolarized voltages (depo) promote transitions from low and medium to high FRET inactivated state conformations. Only the pore domains of two subunits were shown, with the front and rear subunits removed for clarity.

In assigning the observed FRET states to corresponding structural states, we must consider that NavAb channels, whether open or closed, also exhibit almost identical selectivity filter conformations in crystals^11,23. Only very small changes have been observed in NavAb crystal structures with unusual 2-fold symmetry, but their roles in slow inactivation remain to be established¹³. So far, no coupled conformational changes have been identified between the primary and inactivated gates in any available NavAb structures, a hallmark feature of slow inactivation^7,17. One possibility is that the primary gates of all ‘open’ NavAb structures are not open sufficiently to stabilize the SF in the slow inactivated conformation. This appears to be the case for the KcsA potassium channel, in which the SF does not transit to the inactivated conformation until the primary gate opening is >20 angstroms^30,31. Thus, to capture the inactivated NavAb selectivity filter, mutants carrying a reinforced coupler, such as L176F, plus mutations that promote opening of the primary gate (equivalent to glutamate mutations at the helix bundle crossing gate in potassium channels^32,33), may be needed. The structures of the eukaryotic homotetrameric NavEh channel have suggested that the SF dilates during slow inactivation¹². This differs from the classical constriction model. However, it is crucial to note that NavEh possesses extracellular loops (ECLs) that are present in eukaryotic Navs but absent in the NavAb channel. These ECLs are thought to stabilize the dilated inactivated conformation by providing external structural support. In the absence of such constraints, the SF of NavAb may follow a different energetic path during inactivation.

Despite these difficulties in assigning FRET signals to structural states, real-time tracking of conformational rearrangements via FRET signals has emerged as a powerful approach, contributing to our understanding of the slow inactivation process. Many studies have suggested that the slow inactivation mechanism of Navs is quite similar to C-type inactivation in potassium channels^30,31. There is a long-standing idea that C-type inactivation is caused by the collapse of the selectivity filter pore³⁰. However, recent structural studies on eukaryotic Kv channels have suggested that the SF dilates during slow inactivation, placing the field in the midst of a major paradigm shift^25,34–38. Our approach, rather than adhering to a static, binary ‘constriction vs. dilation’ framework, shifts the focus toward capturing the conformational transitions of the slow inactivation. Strikingly, our smFRET data consistently captured a real-time shift toward a high FRET state at E189 and V190, reporting a decrease in the specific diagonal distance between P2 helices during inactivation. Thus, the high FRET population we observed serves as a sensitive physical readout of the NavAb SF’s tendency to enter the inactivated state, reflecting its underlying structural dynamics.

Furthermore, it is noteworthy that recent high-resolution structures of eukaryotic Nav channels have also redefined the molecular basis of fast inactivation. The hydrophobic IFM motif is now understood not as a direct physical “plug” for the pore, but as an allosteric trigger that binds to a distal pocket and induces the closure of an internal gate at the intracellular end of the S6 helices^39–41. This paradigm shift from simple pore occlusion to a complex, multi-step structural transition may also introduce a useful conceptual framework for understanding slow inactivation. Our smFRET observations of transitions between multiple FRET states suggest that slow inactivation is not a static transition to a single ‘collapsed’ structure. Instead, it likely involves a shift in the channel’s conformational ensemble where the propensity of the SF to enter non-conductive arrangements—characterized by the diagonal approach of P2 helices—is allosterically regulated by the state of the voltage sensors and the primary gate. It is worth noting that the L176F mutation stabilizes the S6 helices in the ΔC230 crystal structure in the closed state, suggesting that slow inactivation might also result from closure of the primary helix-bundle crossing gate, as in fast inactivation^39–41. Although most available structural and functional evidence supports an association between slow inactivation and the selectivity filter, more careful studies are needed to rule out the above, less popular but still possible mechanism.

By integrating static protein structures with dynamic smFRET insights, we redefine slow inactivation not as a single structural transition, but as an allosteric shift in a conformational ensemble. Crucially, because slow inactivation represents a non-conducting state where electrical measurements like patch-clamp recordings remain functionally blind to structural changes, smFRET emerges as an indispensable tool capable of directly visualizing these hidden, real-time molecular movements. Our approach successfully captured multiple dynamic transitions associated with slow inactivation using smFRET measurements, demonstrating that the SF exists in a dynamic thermal equilibrium even when functionally inactivated, rather than being locked in a static state. We identified a volume-sensitive molecular coupler, consisting of L176 in the SF and T206 in the S6 helix, that coordinates ‘conformational talk’ between the distal gates. Crystallographic analysis provided direct evidence for this coupling, showing that the L176F mutation can override the opening tendency of the ΔC230 mutant and stabilize the primary gate in a closed conformation. By integrating static protein structure with dynamic smFRET insights, we redefine slow inactivation not as a single structural transition, but as an allosteric shift in a conformational ensemble. This framework, mirroring the complex, multi-step nature of eukaryotic inactivation, paves the way for a comprehensive and unified understanding of the molecular basis of slow inactivation across the voltage-gated sodium channel family.

Methods

Site-directed mutagenesis and construction of NavAb mutants

All primers used in this study were custom synthesized by Eurofins Genomics or IDT. A complete list of oligonucleotide sequences is provided as Supplementary Table S3 in the Supplementary information. All newly generated materials are maintained by the corresponding authors and are available upon reasonable request to K.I and S.W. The NavAb mutated DNAs were subcloned into the modified pBiEX-1 vector (Novagen) that was modified by replacing the fragment from NcoI site (5′-CCATGG-3′) to SalI site (5′-GTCGAC-3′) in the multicloning site with the sequence 5′-CCA TGG GCA GCA GCC ATC ATC ATC ATC ATC ACA GCA GCG GCC TGG TGC CGC GCG GCA GCC ATA TGC TCG AGC TGG TGC CGC GCG GCA GCG GAT CCT AAG TCG AC-3′^42,43. The polymerase chain reaction was used to perform site-directed mutagenesis with PrimeSTAR® Max DNA Polymerase (Takara Bio). All clones were confirmed by DNA sequencing. The plasmids expressing tandem dimeric NavAb proteins were constructed by inserting 2 copies of NavAb cDNAs into the pET28a vector at NdeI-BamHI and BamHI-SalI sites, with a flexible linker sequence encoding double ‘GGGS’ and the FLAG tag¹⁸.

Electrophysiological measurement in insect cells

The recordings were performed using Spodoptera frugiperda Sf9 cells (ATCC#CRL-1711). Sf9 cells were purchased from Sigma-Aldrich (Cat. No. 71104-3CN). The identity of the Sf9 cells was authenticated using the manufacturer’s Certificate of Analysis. Sf9 cells were grown in Sf-900™ II medium (Gibco) complemented with 1% 100× Antibiotic-Antimycotic (Fujifilm-Wako) at 27 °C. Cells were transfected with target channel-cloned pBiEX vectors and enhanced green fluorescent protein (EGFP)-cloned pBiEX vectors using polyethyleneimine (PEI) reagent^42,43. PEI was solubilized into distilled water, and the pH of the 1 mg/mL PEI solution was adjusted to 7.0 with NaOH. First, the channel-cloned vector (0.8–6.6 μg) was mixed with 0.5-4.1 μg of the EGFP-cloned vector in 150 μL of the culture medium. Next, 15 μL of a 1 mg/mL PEI solution was added, and the mixture was incubated for 10 min before the transfection mixture was gently dropped onto the cultured cells. After transfection, approximately 5% of the 300,000 cells fluoresced, and data were collected from those with enough fluorescence intensity. After 24–48 h incubation, the cells were used for electrophysiological measurements.　Pipette solution (105 mM CsF, 35 mM NaCl, 10 mM EGTA and 10 mM HEPES, pH 7.4 adjusted by CsOH) and bath solution (150 mM NaCl, 2.5 mM CaCl₂, 1 mM MgCl₂, 10 mM HEPES, 10 mM glucose, pH 7.4 adjusted by NaOH) were used for the current measurement. Cancellation of the capacitance transients and leak subtraction were performed using a programmed P/10 protocol delivered at each holding potential. The bath solution was changed using the Dynaflow® Resolve system. All experiments were conducted at 25 ± 1 °C using a whole-cell patch-clamp recording mode with a HEKA EPC-10 amplifier and PatchMaster data acquisition software (v2x73). Data export was done using Igor Pro 9.05 and NeuroMatic (version 3.0b). All sample numbers represent the number of individual cells used for each measurement. Cells with leak currents less than 1 nA were used for measurements. When outliers were encountered, they were excluded if abnormalities were found in other measurement environments and included if no abnormalities were found. These criteria were pre-established before conducting the experiments, and no data meeting the quality criteria were arbitrarily excluded. All results are presented as mean ± SE. The graph data were plotted using SigmaPlot 14.

Protein expression and purification

Proteins were expressed in the Escherichia coli KRX strain (Promega). Cells were grown at 37 °C to an OD₆₀₀ of 0.6, induced with 0.1% rhamnose (Fujifilm-Wako), and grown for 16 h at 20 °C. The cells were suspended in TBS buffer (20 mM Tris-HCl pH 8.0, 150 mM NaCl) and lysed using LAB1000 (SMT Co., LTD.) at 12,000 psi. Low-speed centrifugation removed cell debris (12,000 × g, 30 min, 4 °C). Membranes were collected by centrifugation (100,000 × g, 1 h, 4 °C) and solubilized by homogenization in TBS buffer containing 30 mM n-dodecyl-β-D-maltoside (DDM, DOJINDO). After centrifugation (40,000 × g, 30 min, 4 °C), the supernatant was loaded onto a HIS-Select® Cobalt Affinity Gel column (Sigma). The protein bound to the cobalt affinity column was washed with 10 mM imidazole in TBS buffer containing 0.05% lauryl maltose neopentyl glycol (LMNG, Anatrace) instead of DDM. After washing, the protein was eluted with 300 mM imidazole, and the His tag of the protein samples for crystallization was removed by thrombin digestion (overnight, 4 °C). The eluted protein was purified on a TSKgel G4000SWXL (TOSOH Corp.) column in TBS buffer containing 0.05% LMNG.

Crystallization and structural determination

Before crystallization, the purified protein was concentrated to ~10mg ml-1 and reconstituted into a bicelle solution containing a 10% bicelle mixture at 2.8:1 (1,2-dimyristoyl-sn-glycero-3-phosphorylcholine [DMPC, Anatrace]: 3-[(3-cholamidopropyl) dimethylammonio]−2-hydroxy propane sulfonate [CHAPSO, DOJINDO]). The 10 mg/ml protein solution of NavAb and 10% bicelle were mixed in a 20:1 ratio. Prepared proteins of the full-length NavAb mutants were crystallized by sitting-drop vapor diffusion at 20 °C by mixing 300-nl volumes of the protein solution (8–10 mg/ml) and the reservoir solution (9%–11% polyethylene glycol monomethyl ether [PEG MME] 2000, 100 mM sodium chloride, 100 mM magnesium nitrate, 25 mM cadmium chloride and 100 mM Tris-HCl, pH 8.4) with mosquito LCP (STP Labtech). The crystals were grown for 1 to 3 weeks, and after growing, the crystals were transferred into the reservoir solution without cadmium nitrate, which was replaced with the following cryoprotectant solutions. The cryoprotectant solution contains 11% PEG MME 2000, 100 mM Tris-HCl pH 8.4, 2.5 M sodium chloride, 100 mM magnesium nitrate, 0.05% LMNG, and 20% (v/v) DMSO.

Prepared proteins of C-terminal deletion NavAb mutants were crystallized by sitting-drop vapor diffusion at 20 °C by mixing 300-nl volumes of the protein solution (8–10 mg/ml) and the reservoir solution (7%–9% PEG 6000, 100 mM sodium chloride, 100 mM magnesium nitrate, 100 mM cadmium chloride, 10 mM copper chloride and 100 mM Tris-HCl, pH 8.4) with mosquito LCP (STP Labtech). The crystals were grown for 1 to 3 weeks, and after growing, the crystals were transferred into the reservoir solution without cadmium nitrate, which was replaced with the following cryoprotectant solutions. The cryoprotectant solution contains 11% PEG 6000, 100 mM Tris-HCl pH 8.4, 2.5 M sodium chloride, 100 mM magnesium nitrate, 0.05% LMNG, and 20% (v/v) DMSO.

All crystallography data were automatically collected at 100 K using X-rays with a wavelength of 1 Å at BL41XU and BL45XU of SPring-8 and BL09U of Nano Terasu, and merged using the ZOO system⁴⁴. The data were processed by the KAMO system⁴⁵ with XDS (version 20220110)⁴⁶. The data sets of the NavAb BG were obtained from a single crystal. Datasets other than BG were collected using a small-wedge strategy (30° oscillation range per dataset) from multiple crystals. To obtain a complete dataset, individual 30° datasets were integrated and scaled, and then selected datasets were merged using the KAMO system. Analyses of the data with the STARANISO server (Global Phasing Ltd)⁴⁷ revealed severely anisotropic crystals. Therefore, the data sets were ellipsoidally truncated and rescaled to minimize the inclusion of poor diffraction data.

A molecular replacement method with PHASER⁴⁸ provided the initial phase using the structure of NavAb N49K mutant (pdb code: 5yuc) as the initial model. The final model was constructed in COOT (version 0.9.2)⁴⁹ and refined in Phenix (version 1.18)⁵⁰. CCP4 package (version 7.0.078)⁵¹ was used for Structural analysis. After structure refinement, Ramachandran plots for all structures showed no outliers, with over 98% in the favored region. Supplementary Tables 1, 2 summarizes data collection and refinement statistics for all crystals. All figures in the present paper were prepared using the program PyMOL 2.5.4 (Schrödinger, LLC, 2015:The PyMOL Molecular Graphics System, Version 1.8). The pore radii were calculated using the HOLE program⁵².

Fluorophore labeling, liposome reconstitution, smFRET imaging and statistics

NavAb proteins carrying either E189C or V190C mutations were conjugated with photostable Cy3/Cy5 c5 maleimide (1:1)^18,53. The fluorophore-labeled NavAb proteins were then reconstituted into liposomes containing POPE/POPG (3:1, w/w, Avanti Polar Lipids Inc.) with a protein lipid ratio of 1:4000 (w/w) using Bio-beads SM2 (Bioad Inc.) to remove detergents, and the NavAb proteoliposomes were either immediately used or stored in a −80 °C freezer. Immediately before smFRET imaging, the NavAb proteoliposomes were extruded through a 200 nm polycarbonate membrane >30 times to obtain unilamellar, uniform-size liposomes, then immobilized on PEG/biotin-PEG (98:2, w/w) passivated coverslip surfaces for imaging through biotinylated anti-Histag or anti-FLAG antibodies (ThermoFisher, MA1-21315-BTIN, MA1-91878-BTIN)⁵⁴. SmFRET imaging was performed with an objective-based TIRF microscope, and liposomal electrical voltages were generated by transliposomal K⁺ gradients (K⁺_in/K⁺_out, mM, 0.05/150 for −200 mV, 5/150 for −85 mV, 5/5 for 0 mV and 5/0.04 for 120 mV) in the presence of 0.45 µM valinomycin²². SmFRET movies were collected by an exciting laser of 532 nm (~1.0 W/cm²) with a time resolution of 100 ms. The PCD/PCA oxygen scavenger system and ~3 mM Trolox were included in the imaging buffer to enhance the photostability of the fluorophores⁵⁵. Raw smFRET movies without any corrections were processed using the SPARTAN software with a point spread function (PSF) window size of 7 pixels. SmFRET traces were then selected using the Autotrace function with criteria setting as FRET Lifetime >50 frames, donor/acceptor correlation coefficient between −1.1 and 0.5, signal-to-noise ratio >8, background noise level <70, Cy3 blinks <4, and overlap molecules removed, and further manually inspected to ensure that donor and acceptor fluorescence intensities show a clear anticorrelation parttern without more than 1 bleaching step and total fluorescence intensities are constant before bleaching^22,56.

After correction for crosstalk and gamma, smFRET traces were idealized using the Maximum Point Likelihood algorithm in SPARTAN with a 4-FRET-state model, including the low, medium and high FRET states, and one additional zero-FRET state accounting for blinking and bleaching FRET events⁵⁷. To fairly calculate and compare FRET state occupancies, smFRET traces collected from E189C and V190C sites were idealized into low, medium, and high FRET states, with peak centers fixed at 0.25, 0.55, and 0.8, respectively. FRET state occupancies were calculated as fractional areas from histograms of all idealized smFRET traces for each experimental condition, with uncertainties reported as standard errors. Steady-state occupancies of the three conformational states (L, M, H) were obtained from FRET histograms and expressed as fractional populations $p_L,p_M,$ and $p_H$. Pairwise equilibrium constants between states were calculated directly from the occupancies as in Equation (1):

$$K_{ij}=\frac{p_j}{p_i}$$

and converted to an energetic representation using Equation (2):

$$ln{K}_{ij}=ln{p}_j-ln{p}_i$$

Uncertainties in $\ln K$ were estimated by propagation of errors from the measured standard errors of the state occupancies. Thermodynamic consistency was verified by confirming closure of the three-state cycle using Equation (3):

$$ln{K}_{LH}=ln{K}_{LM}+ln{K}_{MH}$$

$\ln K$ values were plotted as a function of lidocaine concentration and fitted with the Hill equation with Hill coefficient n = 1, using Equation (4):

$$lnK\left(x\right)=a+b\frac{x}{K_d+x}$$

where $a$ is the baseline equilibrium, $b$ is the ligand-induced change in $\ln K$, and $K_d$ is the apparent dissociation constant. Curve fitting was performed by nonlinear least-squares regression to obtain the apparent $K_d$. Data are reported as mean ± propagated standard error.

Statistics & reproducibility

Electrophysiological data were analyzed using SigmaPlot 14. All results are presented as the mean with standard error (SEM). We measured the currents from three or more cells where clear currents were observed even among mutants that were difficult to measure. For smFRET measurements, each experimental condition included smFRET traces from at least three independently prepared proteoliposome samples and at least 30 movies, which were pooled to generate FRET histograms and contour maps, with the total number of traces (N) indicated in each figure. No statistical method was used to predetermine sample size. The experiments were not randomized.

Reporting summary

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

Author contributions

S.W. and K.I. conceived the studies. S.W. and S.H. designed and performed smFRET studies and analyzed smFRET data with the help of S.A. and J.V. K.I. designed and performed electrophysiological and crystallographic studies and analyzed the resulting data. K.I. and Y.M. collected and analyzed electrophysiological data. S.W. and K.I. prepared the manuscript with editing input from other authors.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was funded by NIH grant 1R01GM142816 (S.W.). This work was supported by Grants-in-Aid for Scientific Research (17K17795, 21K19332, 20K09193, and 24K02168), SEI Group CSR Foundation, Takeda Science Foundation, and the Institute for Fermentation (K.I.).

Data availability

The structural data generated in this study have been deposited in the Protein Data Bank under accession codes 9VDQ: Crystal structure of voltage-gated sodium channel NavAb N49K mutant, 9VDR: Crystal structure of voltage-gated sodium channel NavAb N49K/L176F mutant, 9VDS: Crystal structure of voltage-gated sodium channel NavAb N49K/L176W mutant, 9VDT: Crystal structure of voltage-gated sodium channel NavAb ΔC230/N49K mutant, 9VDU: Crystal structure of voltage-gated sodium channel NavAb ΔC230/N49K/L176F mutant, 24JS: Crystal structure of voltage-gated sodium channel NavAb N49K/T206A mutant, 45AM: Crystal structure of voltage-gated sodium channel NavAb ΔC230/N49K/T206A mutant, 45AL: Crystal structure of voltage-gated sodium channel NavAb ΔC230/N49K/L176F/T206A mutant. The structure of the NavAb N49K mutant for the initial model of molecular replacement was available in the Protein Data Bank under accession code 5YUC. The data that support the findings of this study are openly available in figshare [https://doi.org/10.6084/m9.figshare.32970413], reference number 32970413. Additionally, source data are provided within the Source Data File. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-78235-z.