## Introduction

Ion channels play a fundamental role in electrical signaling, which is crucial for numerous physiological processes including neuronal excitability, muscle contraction, secretion and perception of environmental changes1,2,3,4. Voltage-gated sodium (NaV) and potassium channels work in concert to generate action potentials in electrical excitable cells2. These channels activate in response to depolarizing stimuli and inactivate rapidly to terminate ion flux. Both activation and inactivation are critical for tuning cellular excitability1. Dysfunction of either process causes abnormal channel function and leads to life-threatening diseases5,6,7.

The four-domain NaV channel from eukaryotes usually inactivates within a few milliseconds8. Extensive studies had identified a three-residue hydrophobic motif, Ile-Phe-Met (IFM), located in the intracellular linker between domain III (DIII) and DIV, which is responsible for the fast inactivation4,8,9,10. By contrast, prokaryotic NaV channels lack the IFM-motif and have slow inactivation over hundreds of milliseconds11,12. The fast N-type inactivation of potassium channels is controlled by their N-terminus or by a cytoplasmic auxiliary subunit, which are thought to work through a “ball-and-chain” mechanism13,14,15,16,17. Recent structural advances revealed that the IFM-motif of NaV channels serves as a hydrophobic latch that allosterically closes the activation gate18,19,20,21. The IFM-motif binds to a hydrophobic pocket adjacent to the activation gate, and its binding shifts the pore-lining S6 helices to close the gate. A recent structural study revealed that the N-terminus of the potassium channel MthK functions as a tethered “ball” physically blocks the open activation gate13. Those structural observations highlighted the distinct mechanisms for fast inactivation of eukaryotic metazoan NaV and potassium channels.

Surprisingly, despite lacking the signature IFM-motif, a family of eukaryotic protozoan homotetrameric NaV channels from the ubiquitous marine plants coccolithophores Emiliania huxleyi and Scyphosphaera apsteinii, exhibit fast inactivation property on the millisecond timescale similar to human NaV channels22,23, suggesting an unknown alternative mechanism for NaV channel fast inactivation. Here, we employed cryo-electron microscopy (cryo-EM) and electrophysiological voltage clamp approaches to investigate the molecular mechanism for fast inactivation of the sodium channel NaVEh from the coccolithophore Emiliania huxleyi. Our results reveal an unexpected structural basis for the N-type fast inactivation of an NaV channel, mediated by its N-terminus and unrelated to the IFM-motif.

## Results

### Functional analysis of NaVEh and structure determination

The sodium channel NaVEh from Emiliania huxleyi is composed of a N-terminal helix (N-helix), 6 transmembrane segments (S1-S6) and a C-terminal EF-hand like domain (EF-L) with a total of 542 amino acid residues (Fig. 1a), suggesting the channel is formed in a homotetrameric fashion. It shares amino acid sequence identity of 27% with the bacterial NaV channel NaVAb and 21% with human NaV1.7, respectively (Supplementary Fig. 1), indicating NaVEh is more closely related to bacterial NaV channels. The gene encoding NaVEh was subcloned into a HEK293-F cell expression vector fused with a green fluorescent protein (GFP) at the C-terminus to facilitate tracing protein expression. We examined the functional characteristics of NaVEh expressed in HEK293T cell by whole-cell voltage clamp. NaVEh generated rapid inward currents in response to depolarizing pulses and became inactivated within 5 ms (Fig. 1b). The V1/2 for voltage-dependent activation and steady-state fast inactivation are −61.5 ± 2.0 mV (n = 15) and −94.4 ± 2.1 mV (n = 9), respectively (Fig. 1b). NaVEh displayed fast inactivation that closely resembles the asymmetric four-domain eukaryotic NaV channels20,24,25, and differs markedly from the homotetrameric NaV channels from prokaryotes that lack fast inactivation11,12. Surprisingly, no IFM-motif like sequence was found in the NaVEh protein sequence (Supplementary Fig. 1).

To investigate the molecular mechanism underlying fast inactivation of NaVEh independent of the IFM-motif, we purified a homogeneous sample of NaVEh in detergents (Supplementary Fig. 2a), and performed cryo-EM single-particle analysis of the purified sample, yielding a final reconstruction map at an overall resolution of 2.8 Å (Fig. 1c and Supplementary Fig. 2b). The excellent density map allowed accurate de novo model building of the region W69-D358 (Fig. 1d and Supplementary Fig. 3). Four blobs of globular density located in the cytosol may belong to the C-terminal EF-L domain (Fig. 1c); however, the map quality of this part was not sufficient to build a reliable model for the EF-L domain. Strikingly, a strong tubular density of ~26 Å in length was observed half-embedded in the intracellular activation gate of NaVEh (Fig. 1c). The tube-like density consistently emerged when map reconstruction was performed with either C1- or C4-symmetry imposed (Fig. 1c and Supplementary Fig. 3c), indicating that the intrinsic density may belong to a helix that extends into the activation gate.

### Architecture of NaVEh and its open activation gate

The NaVEh structure is assembled by four identical subunits in domain-swapped organization (Fig. 1d). Each subunit is composed of a voltage-sensing domain (VSD, S1-S4) and a pore module (PM, S5-S6). The transmembrane core region of NaVEh resembles the bacterial NaVAb26 and human NaV1.525 structures with root mean square deviation (RMSD) of 2.4 Å and 2.9 Å respectively, highlighting the conserved architecture of NaV channels across a wide range of species. Distinct from NaVAb, NaVEh has an extracellular loop (ECL) between S5 helix and pore-helix 1 (P1), extending the vestibule ~25 Å tall above the ion selectivity filter (SF) (Fig. 1a, d). Each ECL consists of a pair of anti-parallel beta sheets and two short helices projecting above the PM, which form extensive interactions with adjacent ECL to stabilize the vestibule (Fig. 2a, b). For example, carbonyl oxygen atoms of G237 and G280 form polar interactions with adjacent R254 and G271, respectively (Fig. 2b). A blob of strong density for a solvent molecule, mediates electrostatic interaction between E285 and N251 of neighboring subunit (Fig. 2b, red sphere). Notably, the ECL is in rich of acidic residues, which generate a strong electronegative surface and provide additional anionic coordination sites for cations (Fig. 2c). The four ECLs form a funnel-shaped vestibule with a diameter of 23.3 Å on the extracellular end, which narrows to 7.3 Å at the SF (Fig. 2d). Interestingly, such ECLs are not found in prokaryotic NaV channels (Fig. 3a), though significant differences between the ECLs of NaVEh and mammalian NaV channels (Fig. 3b, c) remain. At the bottom of the vestibule, a short loop of 303TGESWSE309 between P1 and P2 helices constitute the SF of NaVEh. The SF loop is highly conserved with the 175TLESWSM181 loop of its prokaryotic analog NaVAb26. Four Glu residues at the +3 position mainly determine sodium selectivity by forming a high-field strength site for dehydrating Na+ ions26 (Figs. 2c, d and 3d). However, the symmetric SF of NaVEh differs from the asymmetric SF of metazoan NaV channels, which feature the signature DEKA sequence27. Structure superposition revealed that the square SF of NaVEh is nearly identical to that of NaVAb with van der Waals diameter of 4.6 Å (Figs. 2e and 3d), indicating the SF of NaVEh is more closely related to NaVAb than to heterotetrameric eukaryotic Nav channels. Because the Lys from DIII of the DEKA sequence of heterotetrameric Nav channels was consistently found pointing inside the SF18,19,20,25,28, the asymmetric SF of human NaV channel is shorter than NaVEh in one dimension (Fig. 3e). Despite the conformational and compositional differences, all three SFs confer Na+ selectivity11,22,27. Although possible models have been proposed to explain the sodium selectivity20,26,29,30,31,32, the exact structural mechanisms for discriminating Na+ need further investigation.

The VSD of NaVEh was resolved in an activated conformation, similar to other NaV channel structures19,20,26. The gating charge carrier S4 helix adopts 310 helical conformation, with three of its five gating charges in the activated “up” conformation above the hydrophobic-constriction site (HCS) (Fig. 3f). Two gating charges are stabilized by intracellular negatively charged clusters (INC) below the HCS, suggesting the VSD of NaVEh is less fully activated than that of NaVAb or VSDI, VSDII, and VSDIII of human NaV channels (Fig. 4f–i). Activation of the VSDs generally causes pore opening or channel transition to a non-conductive inactivated state. We next calculated the pore radius of NaVEh at the intracellular activation gate, excluding the N-helix. As illustrated in Fig. 2d, the van der Waals diameter for the activation gate of NaVEh is ~8 Å, wider than the size of hydrated Na+ (7.2 Å)33, indicating that the gate of NaVEh is fully open. A closer look at the gate revealed that the gate is constricted by four L350 residues, whose distance is 11.5 Å measured from the side-chain distal carbon atoms of opposing residues (Fig. 2f). In fact, structural superposition suggests the gate of NaVEh is slightly wider than the open gate of NaVAb34 and rat NaV1.521 (Fig. 2e, f). Collectively, the activated VSD, open activation gate and the presence of a helix blocking the open gate indicate the NaVEh structure was captured in its open-inactivated state.

### The N-helix mediates fast inactivation of NaVEh

The helical density blocking the open gate coincides with the C4-symmetric axis of NaVEh; consequently, the density was averaged during the final refinement with C4-symmetry imposed (Fig. 1c and Supplementary Fig. 3c). To define which part of NaVEh actually forms the observed density, we performed electrophysiological studies of NaVEh variants with deletion or mutation. Deletion of the C-terminal P518-V542 (NaVEhΔ518-542) caused negligible effect on fast inactivation or activation compared to the wild-type (NaVEhWT) (Supplementary Fig. 4a, b). By sharp contrast, deletion of N-terminal I2-R13 (NaVEhΔ2–13) completely abolished the fast inactivation (Fig. 4a and Supplementary Fig. 4b). In addition, we found that the fast inactivation of the NaVEhΔ2–13 can be partially restored when intracellularly applying a synthetic polypeptide of the N-helix (peptide2–13) in a concentration-dependent manner (Supplementary Fig. 4b). At high concentration of 200 μM, the peptide2–13 can restore the fast inactivation of ~88.6 ± 6.1% (n = 9) when test pulse was held at −35 mV (Fig. 4a). These results strongly indicate that the N-helix is responsible for the fast inactivation of NaVEh. To confirm that the N-helix blocks the gate, we purified a NaVEhΔ2–13 protein sample and solved its cryo-EM structure at 4.0 Å resolution (Supplementary Fig. 5). The EM map of NaVEhΔ2–13 clearly showed a hollow gate without any visible density (Supplementary Fig. 5c, d), which confirms that the N-helix indeed binds in the activation gate and blocks it.

Sequence analysis shows that the first two helical-turns of the N-helix are composed of hydrophobic or small side-chain residues followed by five consecutive positively charged residues of Arg9-Arg13 (Arg-cluster) (Fig. 4b). We found that the N-helix can be neatly fitted into the density (Supplementary Fig. 3c). In particular, the first two helical-turns are embedded inside the open activation gate without clashing with the gate (Fig. 4c). Furthermore, we noticed that the outer mouth of the activation gate is rich in negatively charged residues (Fig. 4b, d); therefore, the Arg-cluster can form multiple electrostatic interactions with the negatively charged residues on the four S6 helices (Fig. 4c, d). We hypothesize that the Arg-cluster in the N-helix forms a pre-docking complex for fast inactivation by interacting loosely with the negatively charged outer mouth of the gate and moving the N-helix close to it. In this pre-docking position, the N-helix can plug the gate rapidly after gate opening. To validate this hypothesis, we mutated the Arg9-Arg13 to five Glu (NaVEhnE5) and examined the fast inactivation property of the mutants. Strikingly, the fast inactivation of NaVEhnE5 was completely removed by the mutations (Fig. 4a). These results clearly demonstrate that the electrostatic interactions play critical role in the fast inactivation process.

The N-terminus mediated inactivation in potassium channels is well-studied and often termed the “ball-and-chain” mechanism13,15,17,35,36,37. The fast inactivation of potassium channels is removed after deletion of the N-terminus15,17,37 and restored by intracellular application of the N-terminal polypeptide17,37,38, similar to the results we found for the NaV channel NaVEh (Fig. 4a). Sequence alignment reveals that the N-terminus of the three coccolithophore NaV channels, the potassium channels and their accessory subunits feature a conserved sequence motif with a short region of hydrophobic residues followed by a positively charged cluster poised to enter and block the activation gate (Supplementary Fig. 6a, b). Recent progress from the cryo-EM structure of a calcium-gated prokaryotic potassium channel MthK revealed the structural basis for its N-type inactivation13. Even though the density map was resolved at medium resolution that did not allow side-chain assignment, it clearly showed an N-terminal helical-like density inserted into the open activation gate. Superposition of the open gates of our NaVEh and the MthK (PDB code 6U68)13 shows that both gates have a helix inserted in the middle; however, the gate of MthK is ~1.5 Å wider than that of NaVEh (Supplementary Fig. 6c, d). The smaller gate of NaVEh is caused by the tight corral formed by four S4-S5 linker helices, which is absent in the MthK channel (Fig. 1d and Supplementary Fig. 6c). Nevertheless, these observations indicate that the N-type fast inactivation of NaVEh is similar to the N-type inactivation of potassium channels in mechanism.

### Unexpected N-type fast inactivation of NaV channels

Fast inactivation is the hallmark feature of eukaryotic NaV channels. Structures of eukaryotic NaV channels have established structural basis for understanding the molecular mechanism of fast inactivation mediated by the IFM-motif18,19,20,25,28. The IFM-motif folds into the channel structure and binds tightly to a hydrophobic receptor site adjacent to the S6IV helix, which shifts the S6IV helix and allosterically closes the activation gate (Fig. 5a). Release of the IFM-motif from its receptor site leads to pore opening21. This unique local allosteric inactivation mechanism for NaV channels not only can rapidly terminate sodium influx to prevent overactivity (Fig. 5c, τinact = 2.8 ± 1.4 msec at −10 mV), but can also ensure that the channels can quickly recover from fast inactivation in order to permit repetitive firing in nerve and muscle cells39,40 (Fig. 5e, τfast = 10.1 ± 2.3 msec). Unlike the canonical IFM-motif mediated inactivation, our NaVEh structure demonstrates an alternative mechanism for fast inactivation of NaV channels that is fundamentally different. The N-helix of NaVEh plugs into its open activation gate and physically blocks it (Fig. 5b), similar to the “ball-and-chain” mechanism observed in potassium channels13. Interestingly, the fast inactivation time course of NaVEh is comparable to the mammalian sodium channels (Fig. 5d, τinact = 1.3 ± 0.1 msec at −10 mV), indicating that the IFM-motif is not a prerequisite for the fast kinetics of NaV channel inactivation. Because IFM-motif mediated fast inactivation depends on activation of VSDIV19,20,41, its time course is voltage-dependent (Fig. 5c). In contrast, the time course of fast inactivation for NaVEh is independent of voltage (Fig. 5d), suggesting that the N-helix mediated fast inactivation is open-state inactivation. Strikingly, the recovery of NaVEh from fast inactivation is about 157-fold slower than human NaV1.7 (Fig. 5f, τfast = 1584 ± 473 msec). The recovery rate is even slower than human NaV1.7 and NaV1.8 recovery from slow inactivation42, which were reported to be less than 1 s. Our NaVEh structure provides key structural information that explains the dramatic differences in recovery rate between NaVEh and NaV1.7 (Fig. 5a, b). The binding of the IFM-motif to its receptor site buries a total of 866-Å2 solvent accessible surface. However, the N-helix embedded inside the activation gate buries a total surface of 1688-Å2, almost 2-fold greater than the IFM-motif. In addition, the multiple electrostatic interactions between the Arg-cluster on the N-helix and the negative charges on the S6 helices further strengthen the binding of the N-helix (Fig. 4c, d). The stronger binding interactions of the N-helix indicate that the energy barrier for releasing the N-helix from the open gate would be much higher than releasing the IFM-motif from its receptor site.

## Discussion

In this study, we presented high-resolution cryo-EM structure of the eukaryotic sodium channel NaVEh from the unicellular phytoplankton Emiliania huxleyi. The NaVEh structure shares a conserved core region with NaV channels from bacteria and mammals19,20,25,26,28, but it is more closely related to the bacterial NaV channels, especially in its homotetrameric assembly and selectivity filter. However, NaVEh possesses an additional ECL domain and intracellular EF-L domain that might regulate channel function compared to NaVAb. More importantly, NaV channels from the unicellular phytoplankton like NaVEh have gained the ability of fast inactivation similar to metazoan NaV channels, unlike its prokaryotic homologs that are inactivated at much slower pace, possibly through a C-type slow inactivation mechanism43,44,45,46. We identified the N-helix of NaVEh as the key determinant for its fast inactivation, revealing the N-type like fast inactivation for NaV channels that is similar to the “ball-and-chain” mechanism for potassium channels13,15,17. Our high-resolution structure, complemented with electrophysiological results, provides detailed mechanistic insights into N-helix mediated inactivation, and structural information that potentially explains its similar fast inactivation but much slower recovery from fast inactivation compared with the canonical IFM-motif mediated fast inactivation. The fascinating marine plants coccolithophores are critical for the marine ecology and are highly relevant to climate change47,48. The fast inactivation of NaVEh may be important for the unicellular phytoplankton to tolerate the high concentration of sodium in the living environment22, but the slow recovery may prevent use of this mechanism in metazoan that require high-frequency electrical signaling. It will be of great interest to determine the precise physiological role of NaVEh and its unique fast-inactivation process in coccolithophores and other single-celled organisms.

## Methods

### Whole-cell voltage-clamp recordings of NaVEh in HEK293 T cells

HEK293 T cells were cultured with Dulbecco’s Modified Eagle Medium (DMEM) (Gibco, USA) supplemented with 10% (v/v) fetal bovine serum (FBS, PAN-Biotech, Germany) at 37 °C with 5% CO2. HEK293 T cells were transfected with plasmids of NaVEh WT or mutants using Lipofectamine 2000 Reagent (Thermo Fisher Scientific, USA) for 12 h. Experiments were performed 12–24 h post transfection at room temperature (22–25 °C). In brief, cells were placed on a glass chamber containing 140 mM NaCl, 3 mM KCl, 10 mM HEPES, 10 mM D-Glucose, 1 mM MgCl2, 1 mM CaCl2, (pH = 7.3 with NaOH and osmolarity of ~310 mOsm/L). Whole-cell voltage-clamp recordings were made from isolated, GFP-positive cells using 1.5 ~ 2.5 MΩ fire polished pipettes (Sutter Instrument, USA) filled with standard internal solution, containing 140 mM CsF, 10 mM HEPES, 1 mM EGTA, 10 mM NaCl, (pH = 7.3 with CsOH and osmolarity of ~300 mOsm/L). Whole-cell currents were recorded using an EPC-10 amplifier (HEKA Elektronik, Germany) at 20 kHz sample rate and was low pass filtered at 10 kHz. The series resistance was 2~6 MΩ and was compensated 70~90%. The data was acquired by PatchMaster program (HEKA Elektronik, Germany).

To characterize the activation properties of NaVEh channels, cells were held at −150 mV and then a series of 100 ms voltage steps from −100 mV to +20 mV (5 mV increments) were applied. The time constant (τ) of fast inactivation was from single exponential fits of NaVEh activation in response to depolarization using this voltage protocol. The fast inactivation properties of NaVEh channels were assessed with a 500 ms holding-voltages ranging from −140 mV to −20 mV (5 mV increments) followed by a 50 ms test pulse at −50 mV. The recovery properties were assessed by a double-pulse protocol using a varying interval between the two voltage pulses. Holding potential was −150 mV and prepulse was −50 mV for 20 ms, followed by a recovery test pulse of −50 mV for 5 ms at 32–4096 ms. The currents elicited by the test pulse were normalized to construct the recovery curve.

As for the voltage-clamp recording analyses, all data were reported as mean ± SEM. Data analyses were performed using Origin 2019b (OriginLab, USA), Excel 2016 (Microsoft, USA), and GraphPad Prism 8.0.2 (GraphPad Software, USA).

Steady-state activation curves were generated using a Boltzmann equation.

$$\frac{G}{{G}_{{\max }}}=\frac{1}{1+{{\exp }}[(V-{V}_{0.5})/k]}$$
(1)

where G is the conductance, Gmax is the maximal conductance of NaVEh during the protocol, V is the test potential, V0.5 is the half-maximal activation potential and k is the slope

Fast inactivation curves were generated using a Boltzmann equation.

$$\frac{I}{{I}_{{\max }}}=\frac{1}{1+{{\exp }}[(V-{V}_{0.5})/k]}$$
(2)

where I is the current at indicated test pulse, Imax is the maximal current of NaVEh activation during test-pulse, V is the test potential, V0.5 is the half-maximal inactivation potential and k is the slope factor.

Recovery curves from fast inactivation were fit using a single exponential of the following equation.

$$\frac{{I}_{{test}}}{{I}_{{pre}}}=\left({y}_{0}-1\right)* {{\exp }}\left(-\frac{t}{\tau }\right)+1$$
(3)

where Ipre is the current at prepulse, Itest is the current at test pulse, y0 is the non-inactivated current at the first pulse, t is the delay time between prepulse and test-pulse, and τ is the time constant of recovery from fast inactivation.

### Expression and purification of NaVEh

The codon-optimized gene encoding NavEh (Emiliania huxleyi) was synthesized and was subcloned into the modified pEG BacMam vector49 (Supplementary Table 1). In order to monitor protein expression and purification, a green fluorescent protein (GFP) and a Twin-Strep tag were fused to the C-terminus of NavEh. All constructs were confirmed by DNA sequencing. HEK293-F cells were cultured with Freestyle 293 medium at 37 °C, supplied with 5% (v/v) CO2. When the cell density reached 2.5 × 106 cells/mL, a mixture (3:1) of expression plasmid and polyethylenimine (Polysciences) was added to the cell culture following a standard transfection protocol. After 12 h, sodium butyrate (Sigma, USA) was added to the culture at a final concentration of 10 mM, and the cells were incubated for another 48 h before harvesting.

For purification, the cell pellets were resuspended in Buffer A containing 20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM β-mercaptoethanol (β-ME), and a protease inhibitor cocktail including 1 mM phenylmethylsulfonic acid Acyl fluoride (PMSF), 0.8 μM pepstatin, 2 μM leupeptin, 2 μM aprotinin, and 1 mM benzamidine. Then cells were disrupted with a Dounce homogenizer and membrane fractions were enriched by ultracentrifugation at 36,900 rpm for 40 min. Subsequently, the membrane protein fraction was resuspended in buffer B (buffer A supplemented with 1% (w/v) n-dodecyl-β-D-maltoside (DDM), 0.15% (w/v) cholesterol Hemisuccinate (CHS), 5 mM MgCl2, and 1 mM ATP), and agitated at 4 °C for 2 h. The insoluble membrane fraction was removed by ultracentrifugation at 36,900 rpm for 40 min. Then the supernatant was incubated with Strep-Tactin beads (Smart-Lifesciences), which was pre-equilibrated with buffer C (buffer A supplemented with 5 mM MgCl2, 5 mM ATP, and 0.06% (w/v) Glyco-diosgenin (GDN) (Anatrace)). Subsequently, the Streptactin beads were washed with 10 column volumes of buffer C and buffer D (buffer C without 5 mM MgCl2 and 5 mM ATP), respectively. The protein was eluted by 5 ml buffer E (buffer D plus 5 mM desthiobiotin). The elution was concentrated and loaded onto Superose Increase 10/300 GL (GE Healthcare, USA) pre-equilibrated with 20 mM HEPES, 150 mM NaCl, 0.007% GDN (w/v), and 2 mM β-mercaptoethanol (β-ME), pH 7.5. Peak fractions were collected and concentrated to 7.8 mg/mL.

### Cryo-EM sample preparation and data collection

Aliquots of 2.5 μL purified sample was placed on glow-discharged holey copper grids (Quantifoil Cu R1.2/1.3, 300 mesh), which were blotted for 2.5–3.5 s and plunge-frozen in liquid ethane cooled by liquid nitrogen using a FEI Mark IV Vitrobot at 4 °C with 100% humidity. All data were acquired using a Titan Krios transmission electron microscope operated at 300 kV, a Gatan K2 Summit direct detector and Gatan Quantum GIF energy filter with a slit width of 20 eV. All movie stacks were automatically collected using SerialEM at a calibrated magnification of 105,000× with a physical pixel size of 1.04 Å (super-resolution mode). The defocus values were set from −1.2 to −2.2 μm. The dose rate was adjusted to 10 counts/pixel/s. A total of 1014 and 1119 movie stacks were collected for NaVEhWT and NaVEhΔ2–13, respectively. Each movie stack was exposed for 6.4 s fractionated into 32 frames with a total dose of 60 e2.

### Data processing

All the movie stacks were motion-corrected, binned by 2-fold and dose-weighted using MotionCorr250, yielding a pixel size of 1.04 Å. Defocus values of each summed micrographs were estimated with Gctf51. A total of 299,062 and 579,023 particles were auto-picked for NaVEhWT and NaVEhΔ2–13, respectively. All 2D classification, 3D classification, polishing, and CTF refinement were carried out in RELION3.052. The detailed data processing flow was shown in Supplementary Figs. 2 and 5. The best class containing 61,065 and 64,407 particles for NaVEhWT and NaVEhΔ2–13 were refined using cryoSPARC53 to 2.83 Å and 4.02 Å resolution, respectively.

### Model building

The predicted AlphaFold model of NaVEh was fitted into the cryo-EM density map of NaVEh using Chimera54, manually checked, and corrected in COOT55. Then the resulting model were refined in Phenix56. The model vs. map FSC curve was calculated by Phenix.mtrage. The statistics of cryo-EM data collection and model refinement were summarized in Supplementary Table 2.

All figures were prepared with PyMOL (Schrödinger, LLC), and Prism 8.0.1 (GraphPad Software) and ChimeraX57.

### Reporting summary

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