Structural basis of human Nav1.5 gating mechanisms

Voltage-gated Nav1.5 channels are central to the generation and propagation of cardiac action potentials1. Aberrations in their function are associated with a wide spectrum of cardiac diseases including arrhythmias and heart failure2–5. Despite decades of progress in Nav1.5 biology6–8, the lack of structural insights into intracellular regions has hampered our understanding of its gating mechanisms. Here we present three cryo-EM structures of human Nav1.5 in previously unanticipated open states, revealing sequential conformational changes in gating charges of the voltage-sensing domains (VSDs) and several intracellular regions. Despite the channel being in the open state, these structures show the IFM motif repositioned in the receptor site but not dislodged. In particular, our structural findings highlight a dynamic C-terminal domain (CTD) and III-IV linker interaction, which regulates the conformation of VSDs and pore opening. Electrophysiological studies confirm that disrupting this interaction results in the fast inactivation of Nav1.5. Together, our structure-function studies establish a foundation for understanding the gating mechanisms of Nav1.5 and the mechanisms underlying CTD-related channelopathies.

the proposed 'door wedge' model for fast inactivation 8,[16][17][18][19] .It was also shown that mutations that hinder the interactions between the IFM motif and its receptor binding pocket allosterically propagate to the S6 segments 8,20 .Another model suggests that residues located at the intracellular end of the poreforming S6 segment, rather than the IFM motif, cause fast inactivation 21 .Published work also implied a role for the CTD in regulating sodium channel function 2,[22][23][24][25] .Indeed, the replacement of the native CTD of Na v 1.5 with that of Na v 1.4 confers inactivation kinetics that mimics Na v 1.4 and vice versa 23,24,26 .
Additionally, the CTD has been shown to be a hotspot for LQT3 mutations that affect fast inactivation 27 .Although electrophysiological and structural studies have identi ed critical functional elements, the molecular mechanisms underlying fast inactivation and the role of the CTD have yet to be established in the framework of full-length structures of Na v 1.5, since these intracellular domains have not been resolved in published structures.
Here, we report three high-resolution cryo-electron microscopy (cryo-EM) structures of human full-length Na v 1.5 (hNa v 1.5) in the open state.Although there has been signi cant advancement in the understanding of Na v 1.5 function over the past three decades, our insights into fast inactivation mechanisms remain limited due to incomplete structural knowledge of the intracellular regions.Our structures show previously unseen intracellular regions of Na v 1.5 including the NTD and the CTD.The open state structures feature signi cant conformational changes in the VSDs and exhibit high plasticity in the positioning of the CTD and the III-IV linker.Site-directed mutagenesis coupled with electrophysiological measurements indicate that disrupting the interaction between the CTD and the III-IV linker can result in compromised fast inactivation mechanisms.In summary, our ndings elucidate a mechanism for fast inactivation and the role of CTD in maintaining the open state conformation.
The overall architecture of full-length hNa v 1.5   We puri ed recombinant full-length hNa v 1.5 from Expi293 cells (Extended Data Fig. 1a).Negative stain electron microscopy and SDS-PAGE demonstrated the purity of our protein preparations (Extended Data Fig. 1b,c).We used optimal screened grids for cryo-EM data collection.Analysis of our cryo-EM data showed signi cant heterogeneity owing to the exible intracellular regions of hNa v 1.5.We assessed the structural heterogeneity by establishing a data processing work ow where particles were clustered into three classes (classes 1-3) based on the presence of intracellular features (Extended Data Fig. 2a).Class-1 contained high-resolution features of the core sodium channel.Class-2 and class-3 additionally contained the intracellular regions of hNa v 1.5 (Extended Data Fig. 2a).The processed data yielded structures of hNa v 1.5 with overall resolutions of ~3.2 Å for class-1 (Model-I), ~3.3 Å for class-2 (Model-II), and ~3.6 Å for class-3 (Model-III) (Extended Data Figs.2b and 3a-d).Model-I comprises a total of 1353 residues and contains the transmembrane core, extracellular regions, the III-IV linker, portions of the I-II and II-III linker, and the NTD (Fig. 1a).Model-II and Model-III consist of 1418 residues each and contain the CTD in addition to all structural features observed in Model-I (Fig. 1b,c).The cryo-EM reconstructions show unambiguous densities for the VSDs, the selectivity lter, the pore-lining S6 segments, and the NTD (Extended Data Figs.4-7).The CTD and other intracellular regions exhibit lower local resolutions due to their exibility (Extended Data Figs. 2 and 7).
The a-subunit of the hNa v 1.5 shows a pseudo-tetrameric structure consisting of the D I to D IV domains in a domain-swapped manner (Fig. 1a-c), consistent with the overall architecture observed in published structures of Na v channels 6,8,[16][17][18]28,29 . The ND and the CTD emanate from the transmembrane portion of the channel on the intracellular side.The NTD is located at the base of VSD I .The CTD is a compact domain that is connected to the S6-helix of the D IV via a exible linker.The position of the CTD is variable with respect to the core transmembrane domain (Fig. 1b,c).Model-I exhibits a root-mean-square deviation (RMSD) of 0.6 Å to 0.8 Å over about 1050 Cα residues when compared to Model-II and Model-III.The RMSD between Model-II and Model-III is 0.7 Å for 1113 Cα residues.The transmembrane cores of Models-I to -III have an RMSD of 1.6 Å to 1.8 Å over nearly 1100 C-α residues compared to the human Na v 1.5-E1784K structure (PDB ID: 7DTC) and show an outward movement of ~2 Å of the individual VSDs (Fig. 2a and Supplementary Video 1).This structural change is accompanied by the expansion of the pore domain (Fig. 2a and Supplementary Video 1).

Sequential activation of gating charges is coupled with fast inactivation
Capturing VSDs in different con rmations is necessary to understand the precise activation and inactivation mechanisms of Na v 1.5.The VSDs in the previous open and inactive state structures of Na v 1.5 show similar conformations (Extended Data Fig. 8) 17,20,30 .The superposition of our models showed a gradual upward movement of gating charges (GCs) in VSD II and VSD IV caused by altered sidechain conformations (Extended Data Figs. 9 and 10).Speci cally, multiple salt bridge interactions were rearranged, and two unique p-cation interactions were established between occluding residue (OR) and R4 of Model-III (Extended Data Fig. 9).Only three GCs in VSD II and VSD IV with minimal outward translocation are positioned above the OR, indicating a partially depolarized state.VSD I and VSD III are in the depolarized conformation with three and four GCs positioned above the OR, respectively.This suggests a hierarchical pattern (S4 III > S4 I > S4 II > S4 IV ) of VSD activation states as predicted in previous functional studies 31,32 .
Classi cation of our open state structures based on the conformational spectrum of the VSDs revealed that Model-III represents a late activated state, while Models II and I represent early inactivated states of hNa v 1.5.Comparison of our structures with that of Na v 1.5-E1784K (PDB ID: 7DTC; intermediate inactivated state) revealed a sequential shift in the side chains of GCs as they transition from the open to the fast inactivated state (Fig. 2a and Extended Data Figs. 10 and 11).This observation provides a structural framework for fast inactivation in the order of Model-III > Model-II > Model-I > Na v 1.5-E1784K.

The III-IV linker and interaction with CTD
The positioning of the CTD with respect to the III-IV linker and the transmembrane core is crucial for the working mechanisms of hNa v 1.5.The IFM motif is loosely docked into the hydrophobic receptor pocket formed by the S4 III -S5 III linker and the intracellular ends of S5 IV and S6 IV (Fig. 2c).This positioning differs from other mammalian Na v 1.5 structures as follows.First, the short α-helix that immediately follows the IFM motif and the IFM receptor exhibits a downward displacement (Fig. 2c).The IFM motif is mostly engaged in hydrophobic interactions within the receptor binding pocket (Fig. 2d).The stability of the pocket is primarily maintained by a cluster of hydrophobic residues and polar contacts (Fig. 2d,e).
Altered side chain conformations of N1765, F1473, and Q1476 cause a downward shift in the IFM motif without fully displacing it from the receptor (Fig. 2f).This structural rearrangement is further stabilized by the interaction between D1484 and K1492, located in the short α-helix (Fig. 2e).Second, we observed a signi cant change in the conformation of the S0 IV helix and the connecting loop of the III-IV linker.The N-terminal end of the S0 IV moved ~4 to 6 Å outward which caused a signi cant displacement of the exible loop of the III-IV linker (Fig. 3a,b).A consequence of this transition is that the IFM motif is repositioned but not displaced from the receptor despite the channel being in the open state (Supplementary Video 1).
Notably, we have resolved 100 residues of the previously unseen CTD in Model-II and Model-III in two different conformations.The CTD is positioned closer to the III-IV linker in Model III and away from the III-IV linker in Model II.The superimposition of Model-II and Model-III shows that the position of the CTD differs by ~9° (Fig. 3c,d and Supplementary Video 2).To understand the direction and magnitude of CTD exibility, we performed normal mode analysis (NMA) on Model-II and Model-III.This analysis revealed similar dynamics of the CTD in both structures (Supplementary Video 3).The comparison with the CTD in Na v PaS and Na v Pas-Na v 1.7 chimera revealed signi cant differences in the positioning of the III-IV linker.In our structures, the CTD interacts with the exible loop of the III-IV linker rather than being positioned on the short α-helix of the III-IV linker.This results in the bending of the exible loop of the III-IV linker at residues K1504 and K1505 and causes an inward movement of the S0 IV helix and the short αhelix of the III-IV linker (Fig. 3b,e and Supplementary Video 2).In Model III, both residues are positioned proximate to the negatively charged surface residues, speci cally E1788 on the αI helix and E1867 on the αV helix, within the CTD (Fig. 3b).Site-directed mutagenesis coupled with electrophysiological measurements were used to assess the importance of this interface in the inactivation kinetics of hNa v 1.5.Our data revealed that charge reversal mutations K1540E and K1505E in the III-IV linker and E1788K and E1867K in the CTD altered hNa v 1.5 inactivation parameters.Mutations K1504E, K1505E and E1867K resulted in a faster time course of inactivation compared to WT (Fig. 3f, Extended Data Fig. 12a and , and Extended Data Table 2).Additionally, all four mutants produced destabilization in inactivation as illustrated by a signi cant hyperpolarized shift in steady-state inactivation (Fig. 3g and Extended Data Table 2).We also assessed the recovery from inactivation and one of the mutants, E1788K, displayed a slower recovery (Fig. 3h and Extended Data Table 2).While none of the mutants signi cantly affected current densities (Extended Data Fig. 12b), K1504E and E1867K exhibited a depolarized shift in the conductance curve (Fig. 3g and Extended Data Table 2), suggesting delayed activation.Our results show that the dynamics between the CTD and the III-IV linker are crucial for the transition of hNa v 1.5   from the open state to the inactivated state during the kinetic cycle.

Conformation of the two-tier hydrophobic activation gate in the open state
The opening and closing of the activation gate regulate the in ux of sodium ions into the cells.In our structures, the activation gate is open.We used the position of two hydrophobic rings at the lower part of the S6 helices as a reference for the comparison of the activation gate in our three structures with the Na v 1.5-E1784K (PDB ID: 7DTC, intermediate inactive state) and rNa v 1.5c/QQQ (PDB ID: 7FBS, open state) structures 21 .The residues on the top layer are L409, L935, I1466, and I1768.The bottom layer residues are A413, L938, I1470, and I1771.In the open state structure of rNa v 1.5c/QQQ, the average pore diameter is 10.2 Å at the top layer and 10.3 Å at the bottom layer (Fig. 4, A and C).The average pore diameter in the upper and lower layer of Na v 1.5-E1784K is 8.4 Å and 8.7 Å (Fig. 4a,c) indicative of an intermediate state.Comparison with Model-I showed that our average pore diameters at the top and bottom layer are 10.5 Å and 10.2 Å and closely resemble those of rNa v 1.5c/QQQ (Fig. 4a-c) 20 .In Model-II, the diameter of the top layer decreased to 10.2 Å and the diameter of the bottom layer increased to 10.8 Å in Model-II (Fig. 4a and Extended Data Fig. 13).We observed an even larger activation gate diameter in Model-III, measuring 10.9 Å at the top layer and 11.0 Å at the bottom layer (Fig. 4a-c).Thus, we propose that Models I to III represent open state conformation of hNa V 1.5.Notably, in all our structures, we observed a synchronized movement of the S6 helices, leading to an equivalent pore diameter of the activation gate (Supplementary Video 1).A van der Waals space-lling model shows that hydrated Na + ts through the ori ce of the activation gate in our open state structures (Extended Data Fig. 14).In summary, our structures show activation gates with average pore diameters that exceed those of published Na v 1.5 structures.We attribute this characteristic to the presence of untruncated, native intracellular regions.

Distributions of disease-causing mutations in the NTD and CTD
Previous Na v 1.5 structures have not resolved intracellular regions that contain the majority of Na v 1.5 disease-causing mutations.However, our cryo-EM structures include the previously unresolved NTD and the CTD of hNa v 1.5.This allowed us to systematically map the disease-causing mutations linked to LQT3 and BrS in these regions.Analysis of mutations in the NTD and its interacting S6 I helix, as well as in the CTD and its interacting III-IV linker, showed a distinct pattern of distribution in the structures (Fig. 5a and Extended Data Fig. 15).We found that mutations associated with LQT3 and BrS are dispersed evenly in the NTD (Extended Data Fig. 15) while a cluster of mutations is present in the III-IV linker and the adjacent region of the CTD (Fig. 5a,b).
Mutations in the bending region of the III-IV linker, speci cally at K1505 and P1506 of the KPQ sequence, can lead to LQT3 and BrS (Figs. 3b and 5b) [3][4][5]33,34 . P1506is crucial in shaping the curvature of the III-IV linker, which enables the interaction between K1504-K1505 and the CTD (Supplementary Video 2).Our structures further revealed an important mutational hotspot at the C-terminal end of the III-IV linker and its connecting S0 IV helix where residue R1512 engages in a p-cation interaction with F1522 in Model-III (Fig. 5c and Supplementary Videos 1 and 2).This interaction is disrupted in both Model-II and Model-I, suggesting that the p-cation interaction is transient and reliant on the conformation of the S0 IV helix.
Mutation R1512W can lead to a robust p-stacking interaction with F1522 that limits the exibility of the S0 IV helix and its connecting III-IV linker and leads to the reported changes in hNa v 1.5 kinetics in BrS 35,36 .It is well-established that mutations leading to LQT3, found in regions with inactivation, often present with an increase in persistent current.Here, we observed that the introduction of mutant E1867K resulted in a signi cant increase in late current (Fig. 5d), further supporting the contribution of this region to cardiac channelopathies.

Conclusion
Our cryo-EM studies resolved novel structural features of hNa v 1.5, highlighting conformational dynamics around the VSDs and CTD.Our native, untruncated structures of hNa v 1.5 are captured in the open state.
Importantly, the IFM motif in our structures is repositioned within its receptor but not displaced, providing crucial insights into the fast inactivation mechanism.The consequence of this repositioning of the IFM and the III-IV linker are transduced allosterically to the S6 segments thereby facilitating the opening of the activation gate.These allosteric conformational changes may also be attributed to the overall increased lateral dilation of the structures.Together, we demonstrated how the expansion of the activation gate and pore domain are interconnected, along with the overall expansion of the entire structure through concerted movements of the pore-forming segments.Additionally, in the open state structure of hNa v 1.5, we identi ed a partially activated GC conformation in VSD II and VSD IV that is coupled with the interaction between the III-IV linker and the CTD.We suggest that this interaction stabilizes the open state of the activation gate.Disruption of this interaction leads to the accelerated closing of the activation gate, underscoring the essential role of CTD in fast inactivation regulation (Fig. 6).
Our structures capture the unanticipated kinetic states of the hNa v 1.5, showing the conformational changes from the open state to the inactivated state.Notably, these structures further enabled us to pinpoint several key residues related to LQT3 and BrS in the NTD, the III-IV linker, and the CTD.Overall, our study sheds light on the electromechanical coupling among VSDs, the III-IV linker, and the S5-S6 segments, and their previously unresolved relationship with the CTD.In conclusion, our ndings reveal the potential mechanism of fast inactivation and provide an essential roadmap for studying inactivation mechanisms and regulation of the channel by its numerous protein partners which interact predominantly with the intracellular regions of Na v 1.5.

Methods
Cloning of the full-length human Na v 1.5 expression vector Codon-optimized full-length human Na v 1.5 (hNa v 1.5) cDNA was synthesized (GenScript, NJ, USA) and cloned into a modi ed pCDNA3.1 vector containing a twin-Strep tag and a TEV site at the N-terminus and an HRV 3C site and a FLAG tag at the C-terminus of the channel cDNA.Mutations K1504E, K1505E, E1788K, and E1867K were introduced in the Na v 1.5 cDNA via site-directed mutagenesis (Quick change II mutagenesis kit, Agilent, CA, USA).All vectors were veri ed by Sanger sequencing.

Puri cation of the full-length human cardiac sodium channel
Pellets from 6 litres of cells were used for the puri cation of hNa v 1.5.The cell pellet was rst washed with 1X PBS and subsequently suspended in lysis buffer containing 25 mM HEPES (pH 7.4), 150 mM NaCl, 0.1 mM EGTA, and 10% glycerol (buffer A).A protease inhibitor cocktail was added to the suspension before homogenization.The membrane fraction was collected through ultracentrifugation, then resuspended and dissolved in buffer A supplemented with protease inhibitors, 1% (w/v) n-dodecylβ-D maltopyranoside (DDM, Anatrace), and 0.1% (w/v) cholesteryl hemisuccinate (CHS, Anatrace).The suspension was gently stirred at 4°C for 2 hours.The cell lysate was subjected to ultracentrifugation (30,000 RPM, 4° C, 30 min).The supernatant was incubated for 2 hours with 5 mL of anti-Flag M2 a nity gel that had been equilibrated in advance with buffer B (buffer A supplemented with 0.06% (w/v) glycoldiosgenin (GDN, Anatrace) and protease inhibitor cocktail).The protein-bound FLAG M2 a nity gel was washed with 10-column volumes (CV) of buffer B. The bound protein was eluted using buffer B supplemented with 200 μg/ml of FLAG peptide.The eluent was incubated for 1.5 hours with 3 mL of Strep-Tactin XT 4Flow (IBA) high-capacity resin that had been pre-equilibrated with buffer B. The resin was rinsed with 5 CVs of buffer B, followed by elution of the protein using buffer B containing 50 mM biotin (IBA).The eluted protein was concentrated using a 100-kDa cut-off concentrator (Millipore) and subsequently puri ed using a Superose 6 increase 10/300 gl column (Cytiva) in buffer C (25 mM HEPES at pH 7.4, 150 mM NaCl, 0.1mM EGTA and 0.06% GDN).The peak fractions of the puri ed protein were pooled and concentrated to approximately 8 µM for cryo-EM analysis.
Cell culture and expression of hNa v 1.5 HEK293 cells were grown in DMEM (Gibco) high glucose supplemented with 10% fetal bovine serum (Gibco) and 1% penicillin-streptomycin (Sigma) at 37°C in 5% CO2.The optimized Nav1.5-WT and mutations were stably transfected into HEK293 cells.The transient transfection was performed by electroporation method for maximal transfection e ciency with ATx from MaxCyte (Gaithersburg, MD) according to the manufacturer's instructions.Brie y, cells were dissociated at about 70% con uency and mixed with the target plasmid and transfection buffer.The concentration of the plasmid transfected was 200ng/ml in the electroporation buffer.Then, the cells were electroporated, followed by a 20 min recovery at 37°C and 5% CO2.Cells were then transferred to the maintenance medium for about 48 hours until electrophysiological recordings.

Whole-cell electrophysiology
High-throughput automated patch-clamp experiments were performed using the SyncroPatch 384i (Nanion, Munich, Germany) as previously described 37  The current-voltage relationships of the sodium currents were recorded by holding the cells to -120mV and stepping from -80mV to +60mV in 5mV intervals (each step held for 30ms).For activation, the G/V curve was determined by tting the linear part before the peak of the current-voltage curve with a Boltzmann function: The time course of inactivation was obtained by tting the current traces obtained from the currentvoltage relationship with a single exponential tting of each current trace from peak current to the end of the pulse (30ms) using the following equation: where A is the amplitude, τ is the time constant, I is the current, and t is the time.
The recovery from inactivation was recorded with a two-pulse protocol.The pre-pulse and the test-pulse duration are 30ms, stepping from -120mV to -30mV.The interval between the two pulses ranges from 1ms to 250ms.Currents from the recovery from inactivation were tted to the following equation: The steady-state inactivation was studied with a 500 ms pre-pulse ranging from -140mV to -30mV, followed by a 30ms test pulse stepping from -120mV to -30mV.The currents for the steady-state inactivation were t to a Boltzmann distribution using the following equation: The late Na current was recorded with a 300ms pulse ranging from -120mV to -30mV.Late current was calculated from the percentage of the current measured at 250ms to the peak current.
The tting curves for steady-state inactivation and recovery from inactivation analysis were generated with Origin 10.1.1 software (OriginLab Corporation, Northampton, MA).The leak subtraction protocol was used.

Statistical analysis
Statistical analyses for electrophysiology data were performed using the standard statistical package in Origin 10.1.1 (OriginLab Corporation, Northampton, MA).The student's t-test was performed at a signi cance level of p < 0.05 for a single comparison after a normality test with the Shapiro-Wilk method for sample sizes 7-50.Two-sided p-values less than 0.05 were considered statistically signi cant.
Multiple comparisons with the different mutants were performed using one-way ANOVA, and p-values less than 0.05 were considered statistically signi cant.Results were presented as mean ± SEM.

Sample preparation and cryo-EM data collection
Aliquots of 3.5 μL freshly puri ed hNa v 1.5 was applied onto glow-discharged UltrAuFoil 300 mesh, R1.2/1.3 grids (Quantifoil).Grids were blotted for 3.5 s and plunge frozen in liquid ethane using Vitrobot Mark IV (Thermo Fisher).To evaluate the stability of the samples and the distribution of particles, the grids were screened using Glacios (Thermo Fisher Scienti c) at 200 kV and equipped with a Falcon 3EC direct electron detector.Optimal grids based on the ice thickness and sample distribution were transferred to a Titan Krios G3i microscope (Thermo Fisher Scienti c) operated at 300 kV and equipped with a K3 direct electron detector, a BioQuantum energy lter, and a Cs image corrector.A total of 4,967 movies were captured in super-resolution mode (pixel size of 0.4495 Å) at a magni cation of 81,000x.
The defocus range during collection varied from −0.5 to −2.5 μm.Each movie was exposed to a total electron dose of 60 e−/Å 2 .The data collection was performed using EPU software (Thermo Fisher Scienti c).The summed and dose-weighted micrographs were binned to a pixel size of 0.899 Å/pixel for further data processing.Data collection statistics are reported in Extended Data Table 1.

Image processing and 3D reconstruction
All raw movies were aligned, drift-corrected, and dose-weighed using Patch Motion Correction in cryoSPARC v4.4.1 38 .Defocus and contrast transfer function (CTF) parameters were estimated using Patch CTF in cryoSPARC.The blob-picking module in cryoSPARC was used to generate templates for particle picking.A small set of particles for reference-free two-dimensional (2D) classi cations was selected and subsequently used as templates for particle picking.After two rounds of iterative 2D classi cations, selected particles were used for training in Topaz.The extracted particles were subjected to another two rounds of iterative 2D classi cations.Sixteen well-aligned 2D classes were used as templates for ab initio 3D reconstruction with 281,571 particles in cryoSPARC.After one round of homogeneous re nement and non-uniform re nement, a 3D classi cation was performed to split the particles into four classes.Class-I with 100,809 particles showed clear features for the transmembrane region and extracellular domains.It was selected for non-uniform re nement and after optimizing and re ning the defocus of particles, local re nement was performed by applying a soft mask to mask out the noisy regions and achieved a global resolution of 3.2 Å at a Fourier shell correlation (FSC) at 0.143.Class-II (75,907 particles) and Class-III (65,347 particles) showed extra density for the intracellular domains.Both classes were processed similarly to Class-1 with the exception that a soft binary mask which included the intracellular density was applied during the local re nement.A global resolution of 3.5 Å (Class-II) and 3.9 Å (Class-III) at a Fourier shell correlation (FSC) at 0.143 was achieved in cryoSPARC.Additionally, density modi cations of the Class-II and Class-III maps were performed with the Resolve density modi cation tool in Phenix 39 .The nal resolution after density modi cation was improved to 3.3 Å for Class-II and 3.6 Å for Class-III based on the gold-standard FSC=0.143criteria.All the 3D maps were sharpened with the Autosharpen tool in Phenix and/or DeepEMhancer 40,41 .
The 3DFSC server was used to calculate the 3D Fourier shell correlation and sphericity of each map 42 .The local resolution of the map was calculated using MonoRes in cryoSPARC.Data collection statistics and image processing summary are shown in Extended Data Table 1 and Extended Data Figs. 2 and 3.
Model building, re nement, and validation Three models of hNa v 1.5 were built utilizing the sharpened maps and the AlphaFold2 model of hNa v 1.5 as an initial template 43 .The full-length model was initially docked into the Class-I map using the Phenix Dock in Map tool.Subsequently, the un tted regions of the docked model were removed in Coot.Iterative model building was performed using real-space re nement in Phenix and Coot to remove the outliers and improve the model re nement statistics.The model re ned into the Class-I map has a total of 1353 residues.Additionally, 15 NAG and 11 lipid molecules were modeled into densities.The C-terminal sequences beyond A1778 were left unmodeled.For the maps corresponding to Classes-II and III, an additional 104 residues corresponding to the C-terminal domain (CTD) were built by using PDB ID: 4OVN as a template for rigid-body re nement of the backbone.It is important to note that the positioning of the side chain orientations are ambiguous at lower resolutions in the CTD.Thus, CTD residues were tted and rotamers minimized using Namdinator 44 and further re ned in Phenix using real-space re nement with secondary structure and geometry restraints.