Structure of a prokaryotic sodium channel pore reveals essential gating elements and an outer ion binding site common to eukaryotic channels

Abstract

Voltage-gated sodium channels (Na_Vs) are central elements of cellular excitation. Notwithstanding advances from recent bacterial Na_V (BacNa_V) structures, key questions about gating and ion selectivity remain. Here, we present a closed conformation of Na_VAe1p, a pore-only BacNa_V derived from Na_VAe1, a BacNa_V from the arsenite oxidizer Alkalilimnicola ehrlichei found in Mono Lake, California, that provides insight into both fundamental properties. The structure reveals a pore domain in which the pore-lining S6 helix connects to a helical cytoplasmic tail. Electrophysiological studies of full-length BacNa_Vs show that two elements defined by the Na_VAe1p structure, an S6 activation gate position and the cytoplasmic tail ‘neck’, are central to BacNa_V gating. The structure also reveals the selectivity filter ion entry site, termed the ‘outer ion’ site. Comparison with mammalian voltage-gated calcium channel (Ca_V) selectivity filters, together with functional studies shows that this site forms a previously unknown determinant of Ca_V high affinity calcium binding. Our findings underscore commonalities between BacNa_Vs and eukaryotic voltage-gated channels and provide a framework for understanding gating and ion permeation in this superfamily.

Introduction

Voltage-gated sodium channels (Na_Vs) are large, multipass membrane proteins that are critical for cellular excitation^{1; 2}. These channels are targets for drugs directed at neuropathic pain, migraine, arrhythmias, and epilepsy^{3; 4}, as well as environmental toxins⁵. Na_Vs belong to the voltage-gated ion channel (VGIC) superfamily and are most closely related to voltage-gated calcium channels (Ca_Vs)^{6; 7}. Despite ion selectivity differences, mutational studies^{8; 9; 10} and sequence similarities^{6; 7} have suggested that Na_Vs and Ca_Vs share similar selectivity filter architectures². However, details of this presumed commonality are unknown.

Discovery of a large family of bacterial Na_Vs (BacNa_Vs)^{11; 12; 13} that may be ancestors of eukaryotic Na_Vs and Ca_Vs¹⁴ has enabled delineation of structural principles shared by this VGIC superfamily branch. BacNa_Vs are tetramers. Each subunit has six transmembrane segments that comprise a voltage-sensing domain (VSD) composed of the S1-S4 segments and a pore domain (PD) formed from the S5-S6 segments^{15; 16; 17}. This subunit architecture is recapitulated in eukaryotic Na_Vs and Ca_Vs where four homologous six transmembrane repeats occur in a single polypeptide^{2; 6; 7}. Protein dissection studies have demonstrated a further modular aspect of BacNa_V architecture within the membrane domains. BacNa_V ‘pore-only’ constructs lacking the VSD have been demonstrated to fold^{18; 19; 20, assemble18; 19; 20, and form functional, selective ion channels 19. These demonstrations of BacNa_V modularity are in accord with various lines of evidence that support the independence of the VSDs and PDs. These include: the fact that within the VGIC family, potassium channels occur in forms that encompass a PD alone (Kir and K_2P channels) and forms having a VSD attached to the PD6; 7}, results from VSD-PD chimera studies^{21; 22; 23; 24}, and structural evidence indicating that VSDs and PDs lack extensive contacts^{15; 16; 17; 25; 26; 27}. Although recent BacNa_V structures have revealed the basic transmembrane architecture^{15; 16; 17; 20}, fundamental questions about gating, ion permeation, and ion selectivity have remained unanswered.

BacNa_Vs have a conserved ~40 residue C-terminal cytoplasmic tail^{28; 29} that is important for assembly ²⁸ and function^{29; 30}. However, this domain is either unresolved^{15; 16} or absent from the crystallized constructs^{17; 20} of prior BacNa_V structures. Hence, its structure, relationship to the PD, and key functional elements have remained enigmatic.

Ion permeation is fundamental ion channel property². Original descriptions of the BacNa_V Na_VAb suggested a single ion pore model¹⁵. In contrast, functional studies of Na_Vs^{2; 31} and Ca_Vs^{2; 32} support the presence of multi-ion pores as a means to affect ion selectivity and permeation^{33; 34}. To date, only a single BacNa_V ion binding site has been observed at the inner vestibule of the Na_VRh selectivity filter¹⁷. Recent computational studies have suggested the possibility of other ion binding sites^{35; 36}, but the absence of experimental data have left unresolved questions regarding the existence of such sites, their exact locations, and residues involved in ion binding.

Here, we present the structure of Na_VAe1p, a pore-only sodium channel derived from the Alkalilimnicola ehrlichei BacNa_V Na_VAe1¹⁹. The structure shows a closed conformation of a complete PD and cytoplasmic tail. Functional tests of key structural elements suggest that BacNa_V opening involves changes at an S6 activation gate residue and a structural rearrangement in the neck region of the cytoplasmic tail. The structure also reveals an ion binding site in the selectivity filter that we term the ‘outer ion’ site. We demonstrate that the ion coordination residue comprising this site has a previously unrecognized counterpart in mammalian Ca_Vs that is crucial to high-affinity calcium binding, a result that lends support to long-standing proposals regarding the presence of multiple ion binding sites in Ca_Vs^{32; 33; 34}. Together, our results emphasize the deep evolutionary links between BacNa_Vs and mammalian channels and suggest that channels sharing this selectivity filter architecture have multiple ion binding sites.

Results

Structure of the pore-only channel Na_VAe1p

We determined the structure of Na_VAe1p, a ‘pore-only’ protein bearing the PD and cytoplasmic tail of the BacNa_V Na_VAe1¹⁹ (Supplementary Table S1) using X-ray diffraction data obtained from a I222 crystal that diffracted to a resolution of 4.00Å based on traditional measures such as R_pim or Iσ/I. However, as it has been shown recently that adding weak high-resolution data beyond the commonly arbitrarily defined cutoffs used to judge resolution limits may be beneficial ^{37; 38; 39} we used data to a resolution of 3.46Å based on correlation coefficient (CC) evaluation (CC1/₂>0.1)³⁷ Molecular replacement using a PD ensemble from the Acrobacter butzleri (Na_VAb)^{15; 16} and Rickettsia sp. (Na_VRh)¹⁷ BacNa_Vs revealed electron density spanning from the beginning of the transmembrane segment S5 (Ile150) through the end of the cytoplasmic coiled-coil domain (Ser285) (Supplementary Figure S1A). Na_VAe1p shows the funnel-shaped architecture found in other BacNa_Vs^{15; 16; 17; 20} (Figure 1A, B). Each tetramer subunit is comprised of S5 and S6 transmembrane helices that form the outer and inner parts of the PD, respectively. The P1 and P2 pore helices bridge S5 and S6 and are connected by the selectivity filter (Figure 1).

Figure 1. Structure of the Alkalilimnicola ehrlichei pore-only sodium channel protein Na_VAe1p.

A, Side view showing two transmembrane region subunits and four cytoplasmic tail subunits. Transmembrane helices S5 and S6 are colored in green and blue, respectively. P1 and P2 pore-helices are colored in olive. Selectivity filter is violet. Neck and coiled-coil are tan and orange, respectively. Calcium ion is a white sphere. Grey lines show approximate lipid bilayer boundaries. B, Na_VAe1p tetramer extracellular view. Colors are as in ‘A’. C, Na_VAe1p secondary structure and alignment with BacNa_V pore domain sequences. Na_VAe1p secondary structure elements are indicated and colored as in ‘A’. Selectivity filter position numbers are indicated relative to position ‘0’, Na_VAe1p Glu197. ‘+’ indicates the activation gate residue. ‘*’ indicates position of the Na_VAe1p His245 bend. Vertical black dashed line indicates division between ‘Neck’ and ‘Coiled-coil’. Colored bars indicate: selectivity filter (+1) position, red; S6 pore lining, blue; and coiled-coil core residues, orange. Positions of 3Gly and 7Gly neck mutants are indicated in brown. Grey letters show Na_VSulP portion used in the NaK chimera ²⁹. Red vertical lines show crystallized Na_VMs pore-only construct boundaries ²⁰. Other sequences are: Na_VSp1p, Silicibacter pomeroyi ¹⁹; Na_VAb1p, Alcanivorax borkumensis ¹⁹; Na_VBh1 (NaChBac), Bacillus halodurans ¹¹; Na_VAb Acrobacter butzleri (GI 157737984)¹⁵; Na_VRh, Rickettsiales sp. (GI: 262276647)¹⁵, Na_VSulP Sulfitobacter pontiacus (GI ZP_00961826.1)²⁹; and Na_VMs, Magnetococcus sp. (Uniprot ID A0L5S6)²⁰.

Initial electron density maps (Supplementary Figure S1A) revealed an element absent from prior BacNaV structures, a long helical cytoplasmic tail (Figures 1A and 2A). The Na_VAe1p tetramer forms the asymmetric unit and is packed in the crystal lattice such that the crystallographic axes are not coincident with the channel four-fold symmetry axis (Supplementary Figure S1B, Supplementary Table S1). This arrangement leads to four, independent but similar channel protomers arranged around the channel central axis in the asymmetric unit. Model building and refinement (R_work/R_free 22.4 / 26.8, Table S1) showed that the pore-lining S6 helix continues for one turn after Met241, the intracellular pore constriction point (Figure 2A). A 40° bend at His245 follows and leads to a continuous helix that terminates with a four-stranded, parallel coiled-coil encompassing residues Leu265-Ser285 (Supplementary Table S2). The clear quality differences between electron density maps calculated using data resolution cutoffs based on traditional (Supplementary Figure S1C) versus CC_1/2 metrics (Supplementary Figure S1D) support the choice of resolution cutoff based on CC_1/2 values. The obvious differences in map quality reinforce the assertion that adherence to traditional metrics for defining resolution limits can result in the omission of useful diffraction data^{37; 38}

Figure 2. Na_VAe1p structure comparison and analysis.

A, Side view of a single Na_VAe1p monomer. Secondary structure elements are labeled and colored as in Figure 1. Select residue positions are indicated. B, Cα superposition of the tetrameric Na_VSulP region of the NaK-NaVSulP chimera ²⁹ (Na_VSulP portion, red; NaK portion, wheat) with equivalent parts of the Na_VAe1p neck and coiled-coil (colored as in Figure 1). One Na_VAe1p and NaK-Na_VSulP chimera monomer is shown in cartoon representation in front of the surface of three remaining Na_VAe1p subunits. C, Cα superposition of the Na_VAe1p coiled-coil tetramer with the equivalent KcsA residues. One Na_VAe1p (colored as in Fig. 1) and KcsA (wheat) monomer is shown in cartoon representation in front of the surface of three remaining Na_VAe1p subunits. D, Plot of Na_VAe1p coiled-coil superhelix radius versus residue number. Neck and coiled-coil elements are shaded as in ‘A. E, Close-up of the coiled-coil regions of Na_VAe1p (orange) and Na_VSulP (red) used for superposition in ‘B’. F, Close-up of the cytoplasmic parts of KcsA (left panel, wheat) and Na_VAe1p (neck, sand; coiled-coil, orange). G, Na_VAe1p electrostatic surface potential (+4kT (blue) to −4kT (red)) mapped on the channel van der Waals surface. Grey lines show approximate lipid bilayer boundaries.

The cytoplasmic tail is considerably longer than the pore-forming region (~65Å vs. ~40Å, respectively). Consequently, Na_VAe1p spans ~110Å in the axial dimension and resembles the general architecture of a NaK-Na_VSulP chimera ²⁹ and the full-length KcsA potassium channel⁴⁰ (Figure 2B and 2C, respectively). The coiled-coil is common among BacNa_Vs²⁸ and is thought to participate in channel assembly ^{28; 30}. Its location, C-terminal to a segment that trails S6, is reminiscent of similar domains from eukaryotic Kv7 (KCNQ)^{41; 42} and TRP channels ⁴³ and agrees with predictions ²⁸ and a similar structure in the NaK-Na_VSulP chimera²⁹ (Figure 2B).

The cytoplasmic tail arrangement resembles the stems of a flower bouquet. Individual helices interact extensively at the C-terminal base throughout the eighteen-residue coiled-coil but splay apart above Met267 into individual helical stems that connect the coiled-coil to S6 (Figure 2D). This region, termed the ‘neck’ (Figures 1A and 2A), extends over six helical turns. The distance between the Cα positions and the superhelix axis widens from ~7Å in the coiled-coil to ~9Å in the neck (Figure 2D). The neck showed another unexpected feature, a large electron density that anomalous scattering indicated as a metal ion (Supplementary Figure S2A). B-factors indicate that the neck is as well ordered as other parts of the structure with the exception of the region near the neck ion (Supplementary Figure S2B).

Comparison of Na_VAe1p with a chimera between the nonselective NaK channel and the BacNaV Na_VSulP coiled-coil²⁹ shows good agreement in the coiled-coils (Figure 2B and E, Table S3) (RMSD Cα = 1.2 Å for the tetramer). Contrastingly, the corresponding KcsA region, where there is no superhelical coil (Supplementary Table S2), poorly matches the Na_VAe1p coiled-coil (Figure 2C and 2F, Supplementary Table S3) and reveals an unexpected diversity in how seemingly similar cytoplasmic domains can assemble. The essentially continuous helical conformation from S6 to the coiled-coil is contrary to predictions from circular dichroism and sequence analysis suggesting that this BacNa_V region is disordered²⁸. Notably, the neck has an abundance of charged and polar residues (15/20 residues) (Figures 1C and 2G). This density of hydrophilic residues may be important for neck function.

Comparison of Na_VAe1p pore region with other BacNa_V structures

Na_VAe1p conforms to expectations for a closed conformation as the selectivity filter is not collapsed and the intracellular gate is closed (Figure 3A–B, Figure S2C). Overall, the PD superposes well with other BacNa_Vs (Figure 3C, Table S3). Despite the VSD absence, S5 has position similar to that seen in Na_VAb and is only substantially different than S5 of Na_VRh, which is the outlier of currently known BacNa_V structures. The main variations from other BacNa_Vs lie in the C-terminal ends of S6 from the putative inactivated Na_VAb conformation¹⁶ and Na_VMs pore²⁰ (Figure 3C, and Figure S3).

Figure 3. Na_VAe1p pore region analysis.

A, Na_VAe1p Pore volume represented as a teal mesh calculated by CAVER ⁶⁶. In order to show the central cavity, dummy atoms closed the side vestibules. S6 residues forming the narrow constriction are shown as sticks and labeled. Na_VAe1p elements are labeled as in Figure 1. B, Plot of channel radius versus distance along the central channel axis for closed Na_VAb ¹⁵ (black), inactivated Na_VAbA/_B ¹⁶ (light grey), Na_VRh ¹⁵ (wheat), Na_VMs (dark red) ²⁰, Na_VMs_OM (orange) open state model following ²⁰, and Na_VAe1p (blue). Na_VAe1p constriction points caused by selectivity filter residue Ser198 and S6 residue Met241 are indicated. C, Backbone superposition of PDs of Na_VAb ¹⁵ (black), NaVAbA/B ¹⁶ (light grey), Na_VAb_C/D ¹⁶ (medium grey), Na_VRh ¹⁷ (wheat), Na_VMs ²⁰ (dark red), and Na_VAe1p (colored as in ‘A’). Outer ion from Na_VAe1p and inner ion from Na_VRh are shown as white and wheat spheres, respectively. Two subunits are shown. D, View from the intracellular side showing M241 sidechain Fo-Fc density. E, Outer ion site side view. Ca²⁺ (white sphere) is surrounded by four water molecules (red spheres), and Ser198 (+1) shown in sticks with 2F_O-F_C density surrounding the Ca²⁺ ion as a 1.0σ contoured blue mesh. F, Side view of outer ion anomalous difference density (blue mesh) calculated at 6.5 Å and contoured at 6.5σ. Glu197 (0) and Ser198 (+1) are indicated. Ca²⁺ is shown as a white sphere. The front channel monomer is removed for clarity.

Na_VAe1p has a wide extracellular funnel that connects through the selectivity filter to the central cavity (Figure 3A). Side chain oxygens of selectivity filter residue Ser198 form the narrowest extracellular constriction (pore radius, 1.6Å) (Figure 3B). This is larger than that seen at the analogous Na_VRh position (pore radius, 1.1Å), where the filter is closed¹⁷, but is not as wide as in Na_VAb (pore radius, 2.3Å)¹⁵ (Figure 3B, Figure S4). It is important to point out that the structure of Na_VMs has a diameter that is close to that of the other BacNa_Vs (Figure 3B) due the similarity of its backbone positions with the other BacNa_V structures (Supplementary Table S3). The remaining differences for Na_VMs are largely due to the fact that the Na_VMs structure has incomplete sidechains at seven out of eight of the positions that correspond to the Na_VAe1p residues that constrict the inner cavity, Phe233 and Ile 237 (Figure 3A), and truncates in three of the subunits before the portion that corresponds to Na_VAe1p Met241. The inner diameter of all of the solved structures contrasts to that of the open state model Na_VMs_OM that is generated from the most deviant subunit in the Na_VMs structure (Figure 3B)²⁰.

The inside of the selectivity filter abuts an aqueous cavity that includes lateral openings to the membrane and that ends in a constriction formed by S6 residues Phe223, Ile237, and Met241 (Figure 3A–B), positions largely conserved among BacNa_Vs (Figure 1C). Both the presence of Fo-Fc electron density (Figure 3D) and a sidechain sulfur difference anomalous signal (Supplementary Figure S2A) support the placement of Met241 as the site of intracellular pore closure. This position corresponds to the suggested Na_VAb activation gate¹⁵ and strikingly, forms a closure point further along S6 than in all but the initial Na_VAb structure (Figure 2B). It is notable that evaluation of possible boundaries of the lipid bilayer hydrophobic portion⁴⁴ suggests that unlike other BacNa_V structures, the Na_VAe1p activation gate region protrudes from the bilayer core into the zone comprising phospholipid head groups (Supplementary Figure S5).

Identification of the ‘outer ion’ binding site

We found a large positive electron density peak perched on the four-fold axis of the channel at the selectivity filter outer mouth (Figure 3E). Calculation of anomalous difference maps at 6.5Å revealed a strong peak (11a) indicating the presence of a non-protein anomalous scatterer (Figure 3F, Supplementary Figure S6A–S6D). Recognizing there could be challenges in identifying this peak due to the data resolution, we searched for other evidence that it represented an ion and not noise or some other possibility. We found a similar non-protein anomalous scatterer at the exact same location in a second lower resolution (3.80Å) data set (Na_VAe1p, Crystal II, Supplementary Table S1, Supplementary Figure S6C) obtained from a crystal grown using the same high calcium (200 mM CaCl₂) conditions as the crystal that yielded the 3.46Å resolution data set. By contrast, there was no anomalous peak at this location in maps calculated from crystals grown without calcium (Supplementary Figure S1E and S6C). In all three cases, the maps, which are all calculated at the same resolution (6.5Å), showed strong evidence for the neck ion. Finally, structural studies of the H245G mutant, in which crystals were grown in the absence of calcium and in the identical spacegroup as Na_VAe1p Crystal I and Crystal II (Supplementary Figure S6C), gave no evidence for the selectivity filter ion. Given these multiple lines of evidence that the anomalous peak at the selectivity filter outer mouth depends on the presence of calcium, we assigned this density as a calcium ion.

Inclusion of the ion alone in the refinement left substantial unaccounted electron density. Taking into account the ion position and likelihood that it is partly solvated, we modeled four waters using standard calcium coordination geometry (Figure 3E). We also found additional positive difference (F_O-F_C) electron density extending from the outer ion position through the selectivity filter along the four-fold channel axis; however, we were unable to model whether this arises from ions or solvent (Supplementary Figure S6E). The refined structure shows a partially hydrated calcium ion coordinated by four Na_VAe1p serine oxygens and four water molecules, giving a coordination number common to protein-Ca²⁺ complexes⁴⁵ and serine oxygen-Ca²⁺ distances (2.9Å −3.5Å) that are within those for calcium ions partially coordinated by protein ligands^{45; 46}

We denote the calcium ion position as the ‘outer ion’ because it is separated from the previously reported selectivity filter inner ion position^{17; 35} by 10.7Å (Figure 3C). The outer ion site also does not correspond to the position of the unassigned selectivity filter density reported for Na_VMs²⁰. Observation of the outer ion binding site, together with the strong structural (Figure 3C) and sequence similarity (Figure 1C) of the region and previous identification of an inner ion site establishes that BacNa_V selectivity filters have more than one ion binding site. Sodium and calcium ions have similar radii (0.95Å vs. 0.99Å, respectively) and coordination geometries². Hence, the outer ion position appears to mark the site of entry and partial dehydration as the ion passes into the selectivity filter.

To facilitate comparison among Na_V, Ca_V, and BacNa_V selectivity filters, we denote the residue corresponding to the mammalian Na_V ‘DEKA’ motif² and the conserved glutamates in Ca_Vs^{2; 33} and BacNa_Vs (Figure 1C) as position ‘0’. Other residues are numbered positively or negatively relative to this residue (Figure 4A). Hence, the Na_VAe1p Ser198 (+1) sidechain oxygens coordinate the outer ion (Figure 3E). Comparison of BacNa_V and Ca_V selectivity filter sequences revealed that the (+1) position is strictly conserved as an acidic residue in Ca_V domain II selectivity filters (Figure 4B). This (+1) position had not been previously implicated in Ca_V selectivity. As it plays a role in BacNa_V calcium selectivity^{19; 47}, the compelling similarities together with the observation of a bound calcium ion prompted us to examine the role of the (+1) position in a mammalian Ca_V.

Figure 4. Selectivity filter (+1) position is conserved in mammalian Ca_Vs and important for ion binding.

A, Cartoon depiction of a single Na_VAe1p subunit selectivity filter colored as in Figure 1. Selectivity filter residues are shown and indicated relative to the (0) position glutamate. Dashed line shows the central pore axis approximate position. B, Selectivity filter and pore helices sequence alignment for selected BacNa_Vs, mammalian Ca_V subtype exemplars, and mammalian Na_V1.4. Ca_VBh1 and Ca_VSp1p are calcium selective mutants of Na_VBh1 (NaChBac) ⁴⁷ and Na_VSp1p ¹⁹ respectively. Selectivity filter numbering is indicated. C, Two-electrode voltage-clamp recordings from Xenopus oocyte expressing wild-type Ca_V1.2 or the indicated mutants recorded in a buffer containing 100 mM Li⁺ and either 3 nM or 10 μM free Ca²⁺ and normalized to the 3 nM trace. Currents were elicited by a voltage step from −90 mV to −20 mV. D, Dose-response curves for calcium block of lithium currents for Ca_V1.2 (grey), E1115Q (black), D707N (blue), D707A (green), and D707G (red). Each data point at each calcium concentration is normalized to the current at 3 nM Ca²⁺ and averaged for n = 5–7 oocytes. Error bars are s.e.m.

Outer ion site is important for mammalian Ca_V ion binding

Calcium selectivity in Ca_Vs is thought to arise from the interaction of permeant calcium ions with a ring of selectivity filter (0) position glutamates^{33; 48}. Following experiments that first demonstrated the importance of the (0) position glutamates³³, we used two-electrode voltage clamp to measure calcium block of lithium currents through human cardiac Ca_V1.2 channels expressed in Xenopus oocytes and examined how (+1) position mutations in domain II affect the high affinity calcium site. In contrast to previous reports³³, we found that D707N channels exhibited a reduction in the apparent calcium affinity relative to wild type (IC₅0 11.3 ± 2.0 µM, and 1.9 ± 0.2µM, D707N and wild-type, respectively) (Figure 4C–D). This change was equivalent to that caused by the charge neutralization E1115Q (IC₅₀ 15.1 ± 1.9 µM), the (0) position glutamate neutralization having the largest reported impact on the high affinity site^{33; 48} Changing D707 to alanine to mimic the corresponding Ca_V domain IV position caused a reduction in apparent affinity similar to D707N (IC50 13.6 ± 2.1 µM). Complete removal of the D707 sidechain by D707G to mimic the equivalent position of Ca_V domains I and III caused an even greater reduction in calcium binding (IC50 22.8 ± 6.7 µM). The magnitude of the effects of the D707 neutralizations are striking as this position is much more exposed to bulk solvent than the (0) position glutamate and strongly suggest that the effects of the D707 mutation are through direct interaction with Ca²⁺ rather than a indirect consequence of electrostatic environment alteration. These data demonstrate the importance of a previously unrecognized calcium binding determinant of mammalian Ca_Vs. These findings underscore the similarities between selectivity filters of homomeric BacNa_Vs and their more distant eukaryotic relatives, which have four non-identical selectivity filter repeats.

S6 activation gate residue and neck are important for BacNa_V gating

Observation of a complete BacNa_V cytoplasmic domain connected to a closed pore prompted us to test how the newly described channel elements (Figure 5A) contribute to function. In line with the low success rate of BacNa_V functional expression^{12; 17}, our initial attempts to measure currents from full-length Na_VAe1 using transfected mammalian cells or mRNA injected Xenopus oocytes failed. Therefore, we turned to Na_VSp1, a previously characterized Silicibacter pomeroyi homolog¹² (Figure 1C and Figure 5B).

Figure 5. Functional studies of Na_VSp1 structure-based mutants.

A, Na_VAe1p cartoon depicting positions investigated by patch-clamp. S6, bend, and neck positions are colored yellow, red, and light blue, respectively and indicated using the corresponding Na_VSp1 residues. B, Left panel, exemplar Na_VSp1 Na⁺ currents in response to the activation protocol shown in the middle panel (300 ms depolarizations to −60 to +70 mV in 10 mV steps from a holding potential of −90 mV, sweep to sweep interval = 10 s). Right panel, protocol for examining steady state inactivation voltage dependence (5 s pre-pulse depolarization from 0 to −130 mV in 10 mV steps, followed by a 300 ms step to +30 mV, and repolarization to the holding potential, −90 mV. Sweep to sweep interval = 5 s). C, Left panel, exemplar Na_VSp1 L212A Na⁺ currents. Activation (middle panel) and inactivation curves (right panel) of: Na_VSp1 (open circles), L212A (black triangle) and I216A (inverted black triangle). D, Left panel, exemplar Na_VSp1 M220A Na⁺ currents. Activation (middle panel) and inactivation curves (right panel) of: Na_VSp1 (open circles) and M220A (black squares). E, V_1/2 act and V_1/2 inact for Na_VSp1 and indicated neck mutants. F, Left panel, exemplar Na_VSp1 3Gly Na⁺ currents. Activation (middle panel) and inactivation curves (right panel) of: Na_VSp1 (open circles), 2Gly (open diamonds), 3Gly (open squares), 7Gly (filled open squares), 3Ala (black diamonds), and 7Ala (black triangles). G, Left panel, exemplar Na_VSp1 H224G Na⁺ currents. Activation (middle panel) and inactivation curve (right panel) of: NavSp1 (open circles), H224G (black circles), and H224A (black squares). Activation curves are obtained by normalizing maximal amplitudes divided by the driving force. Inactivation curves are obtained by normalizing maximum amplitudes upon second pulse. NavSp1 activation and inactivation curve Boltzmann fits are shown in ‘C’, ‘D’, ‘F’, and ‘G’.

As prior structural studies have not achieved consensus regarding which S6 residues close the intracellular side of the pore (Figure 3B), we first examined the Na_VSp1 S6 positions equivalent to those that narrow the Na_VAe1p intracellular side (Figure 3A). Alanine substitution in each of the two helical turns above the constriction site, Na_VAe1p F233 and I237 (Na_VSp1 L212A and I216A) (Figure 3A and Figure 5A, Supplementary Figure S7A), did not affect the voltage-dependence of activation (V_1/2,act) (ΔV_1/2,act = −1.0 ± 4.5 and −0.4 ± 4.8 mV, respectively) (Figure 5C, Table 1). However, these mutants did shift the voltage-dependence of inactivation (V_1/2,inact) to more negative potentials (ΔV_1/2,inact = −14.8 ± 4.3 and −14.0 ± 3.8 mV, respectively) (Figure 5C, Table 1), decreased the inactivation time constants, and accelerated recovery from inactivation (Supplementary Figure S7B and S7C). By contrast, alanine substitution of the position equivalent to the pore occlusion point, Na_VAe1p Met241 (Na_VSp1, M220A) (Figure 3A and Figure 5A) caused dramatic negative shifts in both V_1/2,act and V_1/2,inact (ΔV_1/2,act = −49.8 ± 3.3 mV and ΔV_1/2,inact = −40.0 ± 3.5 mV)(Figure 5D, Table 1), but left the inactivation time constants and recovery from inactivation unaltered (Supplementary Figure S7B–C). The major effect of Na_VSp1 M220A on V_1/2,act contrasted against the absence of V_1/2,act changes from mutation of residues further into the central cavity suggests a critical role for this position in stabilizing the closed state of the channel and indicates that the more distal positions are important for inactivation only.

Table 1.

Activation and inactivation properties of BacNa_Vs and mutants

	V1/2, act	ΔV1/2, act	n	P-Value	V1/2, inact	ΔV1/2, inact	n	P-Value
NaVSp1	27.4 ± 1.1	−	13	n.a.	−42.7 ± 1.1	−	12	n.a.
S6
L212A	26.4 ± 3.1	−1.0 ± 4.5	5	n.s.	−57.5 ± 2.7	−14.8 ± 4.3	6	***
I216A	27.0 ± 3.4	−0.4 ± 4.8	3	n.s.	−56.7 ± 1.6	−14.0 ± 3.8	4	***
M220A	−22.4 ± 1.6	−49.8 ± 3.3	7	***	−82.7 ± 1.8	−40.0 ± 3.5	7	***
Bend
H224A	30.2 ± 1.1	2.8 ± 4.2	3	n.s.	−48.1 ± 3.5	−5.4 ± 4.8	3	n.s.
H224G	22.9 ± 1.7	−4.5 ± 3.9	4	n.s.	−67.7 ± 2.3	−25.0 ± 4.0	4	***
Neck
A226G	11.1 ± 0.2	−16.3 ± 4.2	3	***	−51.4 ± 2.8	−8.7 ± 4.3	4	***
E227G	15.2 ± 1.5	−12.2 ± 3.3	6	***	−46.8 ± 1.1	−4.1 ± 3.6	4	n.s.
E227A	25.1 ± 2.3	−2.3 ± 4.1	4	n.s.	−47.1 ± 1.3	−4.4 ± 3.2	6	*
D228G	5.7 ± 1.9	−21.7 ± 3.9	4	***	−51.7 ± 1.2	−9.0 ± 3.6	4	***
D228A	28.4 ± 1.3	1.0 ± 4.3	3	n.s.	−45.7 ± 2.5	−3.0 ± 4.2	4	n.s.
2G	2.1 ± 2.4	−25.3 ± 2.6	6	***	−47.2 ± 2.6	−4.5 ± 3.2	4	n.s.
2A	11.7 ± 1.3	−15.7 ± 3.4	5	***	−52.8 ± 2.0	−10.1 ± 3.9	4	***
3G	−7.6 ± 2.1	−35.0 ± 3.8	5	***	−60.8 ± 1.9	−18.1 ± 3.6	10	***
7G	−13.3 ± 1.9	−40.7 ± 3.9	4	***	−64.8 ± 4.1	−22.1 ± 5.2	4	***
7A	32.1 ± 2.8	4.7 ± 4.4	4	n.s.	−67.3 ± 4.2	−24.6 ± 5.2	4	***
Coiled-coil
R242E	7.8 ± 0.8	−19.6 ± 3.3	5	***	−44.4 ± 2.5	−1.7 ± 4.2	4	n.s.
NaVAe1
3G	32.1 ± 1.1	n.a.	6		5.3 ± 3.1		4

Data are mean ± s.e.m.

^ΔV_1/2 ‘±’ denotes 90% confidence interval P values are calculated relative to wild-type Na_VSp1 n.s., Not significant, P > 0.05

^* indicates 0.001 < P < 0.05

^***indicates, P< 0.001

Because of its direct structural connection to the activation gate, we next examined how the helical structure of the neck affected channel behavior by using a strategy of glycine substitutions to destabilize the helical conformation^{49; 50; 51}. Because BacNa_V necks have variable lengths, we focused on the most conserved region (Figure 1C). Similar to the S6 mutant Na_VSp1 M220A, single glycine substitutions at each of the three residues below the bend Na_VAe1p Ala248, Glu249, Asp250 (Na_VSp1, A226G, E227G, and D228G) caused a significant negative shift in V_1/2,act (Figure 5E, Table 1, Figure S7D–E). By contrast, alanine substitutions at Na_VSp1 E227 and D228 yielded essentially wild-type channels (Figure 5E, Table 1, Supplementary Figure S7D–E), indicating that the glycine impact comes from increased flexibility rather than sidechain deletion. Increasing the consecutive numbers of Na_VSp1 neck glycines to two (residues 227–228), three (residues 226–228), and seven (residues 226–232) further facilitated activation gate opening that was coupled to negative shifts in V_1/2,inact (ΔV_1/2,act = −25.3 ± 2.6, −35.0 ± 3.8, and −40.7 ± 3.9 mV and ΔV_1/2,inact = −4.5 ± 3.2, −18.1 ± 3.6 and −22.1 ± 5.2 mV, for 2Gly, 3Gly and 7Gly, respectively)(Figure 5E and F, Table 1). The 2Ala mutant revealed that part of the shift caused by the 2Gly mutant arises from a synergistic effect of removing Na_VSp1 E227 and D228 simultaneously (ΔV_1/2,act = −15.7 ± 3.4 versus −25.3 ± 2.6 mV, for 2Ala and 2Gly, respectively (Figure 5E, Table 1) that indicates an additional role for these charges. Nevertheless, the 7Ala mutant had activation properties similar to wild-type and that strongly contrasted the large negative shifts in V_1/,act and V_1/2,inact caused by 7Gly (ΔV_1/2,act = 4.7 ± 4.4 versus −40.7 ± 3.9 mV for 7Ala and 7Gly, respectively). The major negative shifts in V_1/2,act produced by neck substitutions that are detrimental to helix formation strongly supports the idea that a structured neck is critical for closed state stabilization.

The majority of neck mutants causing negative shifts in V_1/2,act also elicited negative shifts in V_1/2,inact (Figure 5E, Table 1). However, unlike previously reported mutations in the Na_VSulP coiled-coil that slowed τ_inact by an order of magnitude or more²⁹ most of the neck mutants left τ_inact unperturbed (Supplementary Figure S7F–G). Of the few that did not (E227G, D228G, 2G, 7G, P< 0.001), none caused a perturbation larger than a factor of three. To examine this discrepancy further, we characterized Na_VSp1 R242E (Supplementary Figure S8), a substitution at a conserved position at the N-terminal end of the coiled-coil region that was reported cause a negative shift in V_1/2,act and slow Na_VSulP inactivation by ~37-fold²⁹. Although Na_VSp1 R242E did cause a negative shift in V_1/2,act (Table 1), unlike its Na_VSulP counterpart, this change caused only a modest (~2.7-fold, P = 0.04) slowing of inactivation (Supplementary Figure S8D).

Finally, we tested the consequences of alanine and glycine substitutions at the bend, Na_VAe1p His245 (Na_VSp1, H224) (Figure 5G, Supplementary Figure S9A). Na_VSp1 H224A did not affect V_1/2,act or V_1/2,inact (ΔV_1/2,act = 2.8 ± 4.2, ΔV_1/2,inact = −5.4 ± 4.8 mV) (Figure 5G, Table 1), but did slow inactivation (~2-fold, P< 0.001) (Supplementary Figure S9B). In contrast, Na_VSp1 H224G spared V_1/2,act and caused a large negative shift in V_1/2,inact (ΔV_1/2,act = −4.5 ± 3.9 mV, ΔV_1/2,inact = −25.0 ± 4.0 mV) (Figure 5G, Table 1). Although local disorder (Supplementary Figure S2B) precluded us from modeling the Na_VAe1p His245 sidechain, this residue could co-ordinate the neck ion. To test the structural consequence of loss of this potential ligand, we determined a 5.8Å resolution structure of Na_VAe1p H245G crystallized from a low calcium condition (Supplementary Table S1). The structure showed no major changes from wild-type except for the loss of anomalous density for the neck ion and outer ion (Supplementary Figure S8C)(RMSD Cα = 0.5 Å for the tetramer relative to wild-type). Hence, the bend residue appears to be important for neck ion coordination but the neck ion is not essential for the bend structure or helical character of the neck.

Our structure-based mutational studies uncovered two functional phenotypes. Mutations in the activation gate and neck having negative shifts in V_1/2,act ≥−20 mV also caused negative shifts in V_1/2,inact, suggesting that the two processes are strongly coupled. Mutations at S6 residues above the activation gate and bend residue H224, selectively impacted V_1/2,inact (Table 1). Together, these data support the ideas that (i) the Na_VAe1p structure represents a closed state, (ii) destabilization of the neck facilitates channel opening, and (iii) residues in S6 above the constriction site and at the bend are important for the molecular transitions underlying inactivation.

Neck destabilization allows Na_VAe1functional characterization

Having established the importance of the neck helical structure for closed state stabilization, we revisited Na_VAe1 functional studies to test whether glycine substitution in the neck would permit us to record from full-length channels. Indeed, Na_VAe1 channels bearing the 3Gly mutation (residues 248–250) produced voltage-dependent channels (Figure 6A–B). These had a V_1/2,act similar to Na_VSp1 but ~40 mV more positive than the equivalent Na_VSp1 3Gly mutant (V_1/2,act = 32.1 ± 2.8, 27.4 ± 1.1, and −7.6 ± 2.1 mV for Na_VAe1 3Gly, Na_VSp1, and Na_VSp1 3Gly, respectively). These results further support the idea that the neck helical structure is important for closed state stabilization and suggest that wild-type Na_VAe1 has a very positive V_1/2,act of activation that had prevented functional characterization.

Figure 6. Na_VAe1 functional properties.

A, Left panel, exemplar NavAe1 3Gly Na⁺ currents in response to an activation protocol, right panel (600 ms depolarizations to 0 to +70 mV in 10 mV steps from a holding potential of −90 mV, sweep to sweep interval = 5 s). B, Left panel, protocol for examining steady state inactivation voltage dependence (5 s pre-pulse depolarization from to 50 to −30 mV in 10 mV steps, followed by a 300 ms step to +30 mV, and repolarization to the holding potential, −90 mV. Sweep to sweep interval = 10 s). Right panel, Na_VAe1 3Gly activation and inactivation curves. Boltzmann fits are indicated. C, Left panel, exemplar Na_VAe1 3Gly currents in presence of 130 mM intracellular Na⁺ and 107.5 mM extracellular Ca²⁺, in response to an activation protocol, right panel (600 ms depolarizations to 10 to +100 mV in variable steps (10 mV and then 5 mV after 70 mV) from a holding potential of −90 mV, sweep to sweep interval = 5 s). D, Normalized current-voltage curve from ‘C’. Reversal potential can be obtained by linear regression, as indicated. The averaged value obtained by this method (n=5) gives E_rev = −52.1 ± 10.3 mV, which corresponds to P_Ca/P_Na = 0.07 ± 0.02 (n=5) when corrected for the liquid junction potential (-17mV).

The ability to record from Na_VAe1 3Gly allowed us to test the functional properties of the Na_VAe1 selectivity filter defined by our structure. In agreement with the strong selectivity filter conservation (Figure 1C), biionic recording experiments (Figure 6C–D) showed that Na_VAe1 has selectivity properties similar to other BacNa_Vs including Na_VSp1p¹⁹, the ‘pore-only’ version of Na_VSp1 (P_Ca/P_Na = 0.07 ± 0.02 and 0.08 ± 0.01, for Na_VAe1 3Gly and Na_VSp1p, respectively), and Na_VBh1 (NachBac)⁴⁶. Due the relatively low current amplitude of Na_VAe1 3Gly, we were restricted to this extrapolation method of examining the permeability ratio. Hence, we validated the measurement by determining the ion selectivity of full-length Na_VSp1 3Gly, which expresses much better than Na_VAe1 3Gly, by two methods, the extrapolation method used to examine Na_VAe1 3Gly and a tail current protocol (Supplementary Figure S10). Both methods gave the same P_Ca/P_Na ratio (0.05 ± 0.02 and 0.07 ± 0.02, respectively) and agree with the values for Na_VAe1 3Gly and the ‘pore only’ Na_VSp1p. These results support the idea that the 3Gly mutation has minimal influence on selectivity and that Na_VAe1 is a sodium selective channel.

Discussion

Structure determination of the ‘pore-only’ BacNa_V Na_VAe1p revealed previously uncharacterized BacNa_V architectural features that are important for function. The presence of the long, helical intracellular domain allowed the complete definition of S6, which extends into the intracellular side of the membrane (Supplementary Figure S5). Relative to all but the initial Na_VAb structure¹⁵, which contained pore-lining cysteine mutants near the activation gate that may have influenced S6 positioning, this structure places the intracellular gate more towards the S6 C-terminus (Figure 3B). It seems likely that the absence of a consensus among the prior BacNa_V structures in defining the site of this important channel element arises from the fact that they lack the intracellular domain. The structural plasticity of the intracellular gate seems a likely consequence of it being the central point of structural changes required for gating and inactivation. Our observations are not unlike those described for KcsA in which the full-length structure⁴⁰ showed that the activation gate closure point was more intracellular than defined from a structure lacking the intracellular domain⁵². The Na_VAe1p activation gate and subsequent helical extension of S6 should protrude beyond the boundaries of the hydrophobic portion of the lipid bilayer into the lipid headgroups (Supplementary Figure S5). This location may have important consequences as, based on voltage-gated potassium channel studies ²⁶, it could allow the C-terminal portion of S6 to interact directly with the phospholipid headgroups in a way that could influence function. It might also permit interactions with the proposed S4-S5 linker closed state pose²⁶.

Our studies demonstrate that destabilizing the helical structure of the neck causes negative shifts in V_1/2,act. The largest perturbation, 7Gly in which two full helical turns of the neck are glycines, has effects that are of the same magnitude as activation gate disruption (Table 1). In both cases, there are parallel shifts in V_1/2,inact that indicate that activation and inactivation are tightly coupled. Previous work with Na_VSulP showed that coiled-coil disruption slowed inactivation kinetics by more than an order of magnitude, suggesting that this structure as a role in accelerating inactivation²⁹. Unlike these effects, which were caused by disruption further from the pore, destabilization of the Na_VSp1 neck had minimal impact on inactivation time constants (Supplementary Figure S7F–G). Moreover, examination of a coiled-coil mutation, R242E, equivalent to one from Na_VSulP that caused a dramatic slowing of inactivation, negatively shifted V_1/2,act but failed to produce a similar effect on Na_VSp1 inactivation kinetics (Figure S8D). Thus, the major role of the neck is to stabilize the channel closed state.

Our studies lead us to propose the following model for BacNa_V gating (Figure 7). In the closed state, represented by the Na_VAe1p structure (Figure 1), the intracellular side of the channel central pore is occluded by the activation gate residue constriction (Na_VAe1p M241). Opening would proceed with a radial expansion of this region¹⁵ accompanied by an order→disorder transition in the neck. The neck region is a site of potential disorder²⁸ The abundance of polar and charged neck residues (Figure 2G) may aid the transition to this state and assist in permeant ion escape into the cytoplasm (Figure 7). Whether such a state resembles the proposed BacNa_V open state model²⁰ is unclear, as much of the end of S6, including the activation gate equivalent of Na_VAe1p Met241 is absent from the Na_VMs structure. Eukaryotic Na_Vs and Ca_Vs lack an equivalent of the C-terminal tail; however, the prevalence of similarly located C-terminal coiled-coil domains among diverse eukaryotic VGICs^{41; 42; 43} and the importance of the intervening region that connects S6 to the coiled-coils for channel regulation by a diverse factors^{53; 54} suggests that the essence of this proposed BacNa_V mechanism has parallels in eukaryotic VGICs.

Figure 7. Cartoon model of BacNa_V gating.

Activation of BacNa_Vs is proposed to involve the expansion of the pore at the activation gate (Met 241) and an order→disorder transition in the neck region. Ions can escape into the cytoplasm through spaces created by the neck region. Channel elements are colored as in Figure 1. VSDs are not shown.

The ionic radii and coordination geometries of sodium and calcium ions are similar ². Hence, the Na_VAe1p outer ion position, revealed by calcium, appears to mark the site of entry and partial dehydration as the permeant ion interacts with the (+1) residues and passes into the selectivity filter. This role in ion coordination agrees with the observation that a single S^D change at the NaChBac (Na_VBh1) (+1) position alters selectivity for calcium over sodium by ~200 fold⁴⁷ and with simulations suggesting that residues forming the outer ion site may be involved in ion recruitment³⁵. The outer ion site may also participate in the divalent ion block described for NaChBac (NaVBh1)^{11; 17} and the NaChBac/Na_VRh selectivity filter chimera¹⁷ as it corresponds well with the predicted ‘site 1’ blocking site from molecular dynamics simulations ³⁵ It is also striking that structural changes at the (+1) serine cause substantial alterations to the size of the selectivity filter entrance in the putative inactivated conformation of Na_VAb¹⁶ and in Na_VRh, where the (+1) serine occludes the pore (Figure S4). Thus, this outer ion site may not only be important for engagement of permeant and blocking ions but may participate in rearrangements leading to slow inactivation⁵⁵

Observation of an outer ion binding site in the selectivity filter together with the prior discovery of a selectivity filter inner ion site ¹⁷ (Figure 3C) and strong sequence and structural conservation of this region (Figure 1C and Figure 3C) shows that BacNa_V selectivity filters posses more than one ion binding site. It seems possible, especially given the ~10Å separation between outer and inner ion sites, that multiple ions may occupy the selectivity filter simultaneously. This situation would be tantalizingly close to accepted ideas regarding multi-ion pores in mammalian Na_V and Ca_V counterparts^{2; 31; 32; 34}, not unlike that of potassium channels⁵⁶, in line with recent computational studies of BacNa_V filters⁵⁷, and argues against the single ion pore model suggested in the initial BacNa_V structure analysis¹⁵.

Identification of the outer ion binding site uncovered a previously unknown role for the analogous conserved (+1) position in mammalian Ca_Vs as an important determinant for calcium selectivity. Notably, despite its more exposed location, the impact of neutralization of the (+1) position is equivalent to that of neutralization of the (0) position glutamate that resides deeper in the selectivity filter (Figure 4D), strongly suggesting that it may interact directly with the permeant ion. These results demonstrate a deep commonality between BacNa_Vs and eukaryotic voltage-gated channels that should facilitate understanding ion permeation and gating in the superfamily.

Materials and methods

Crystallization

Na_VAe1p was expressed and purified in β-dodecyl maltoside (DDM) as described ¹⁹. For high calcium-condition crystals, purified protein was concentrated to 15 mg ml⁻¹ by centrifugal filtration (Amicon® Ultra-15 100-kDa MW cutoff, Millipore) and mixed with 5M trimethylamine oxide (TMAO^{58; 59}) creating a solution of 13.5 mg ml⁻¹ protein, 0.25 mM DDM, 0.5M TMAO, 200 mM NaCl, 20 mM Na-HEPES, pH8.0. The protein was crystallized using hanging-drop vapor diffusion at 4°C over a reservoir of 200 mM CaCl₂, 30% PEG 400, 100 mM Na-acetate, pH 5.0. 0.7 µl of protein-TMAO solution was laid on a 0.7 µl drop of mother liquor that had been mixed with agarose to a final concentration of 0.25% and left to solidify at room temperature for 1 min. I222 crystals grew from precipitate to a final size of ~200× 70 x 15 µm after three weeks.

For low calcium condition crystals, protein was concentrated to 13.5 mg ml⁻¹ (Amicon® Ultra-15 100-kDa MW cutoff, Millipore) following exchange into 0.25 mM DDM, 200 mM NaCl, 20 mM Na-HEPES, pH 8.0, during the last size exclusion chromatography purification step. P4₂ crystals were grown by hanging drop vapor diffusion at 4°C from equal volumes of protein reservoir solution of 200 mM MgCl₂, 30% polyethylene glycol (PEG) 400, 100 mM MES, pH 6.5. Crystals appeared in 2 days, grew to ~200 × 50 x 50 µm in ~two weeks.

Na_VAe1p H245G was expressed and was purified as Na_VAe1p ¹⁹ using a final size-exclusion chromatography buffer of 0.3 mM DDM, 200 mM NaCl, 2mM MgCl₂, 20 mM Na-HEPES, pH 8.0. Purified protein was concentrated to 13.5 mg ml⁻¹ (Amicon® Ultra-15 100-kDa MW cutoff, Millipore). 1 µl hanging drops were set-up at 4°C using a 24 well VDX Plate™ (Hampton Research) over a reservoir of 28% PEG 400, 100 mM Na acetate, pH 4.5. I222 crystals grew overnight and reached maximal proportions of ~300 × 75 X 15 µm after two weeks.

Data collection, structure determination and refinement

Crystals were frozen directly into liquid N₂ for data collection. Diffraction data were collected at Advanced Light Source Beamline 8.3.1, Lawrence Berkeley National Laboratory, integrated with MOSFLM 7.0.4⁶⁰ and scaled with SCALA (3.3.20)⁶¹. Phase information was obtained by molecular replacement with an ensemble model based on the 3RVY, 4DXW, and 4EKW pore domains using Phaser (2.1.4)⁶². Model was improved using cycles of manual rebuilding, Coot (0.7) ⁶³, and refinement, Refmac (5.7.32)⁶⁴. NCS-averaged maps improved apparent electron density and allowed placing of most of the side chains. For the high calcium structure, initial tight NCS restraints were employed and later relaxed for all segments except for residues 183– 208 and 214–220. For both the low calcium structure and the Na_VAe1p H245G mutant tight NCS restraints were employed throughout.

Two electrode voltage-clamp electrophysiology

Human Ca_V1.2 (α₁C77, GenBank CAA84346), rat Ca_Vβ_2a (GenBank NP 446303), and Ca_Vα₂δ-1(GenBank NM_00182276) were used for two-electrode voltage clamp experiments in Xenopus oocytes. Mutations were introduced using Quikchange (Stratagene). Linearized cDNA was translated into capped mRNA using the T7 mMessenger kit (Ambion). 50 nl of Ca_V1.2α₁, Ca_Vβ_2a, and Ca_Vα₂δ-1 mRNA at a 1:1:1 molar ratio were injected into Xenopus oocytes. Two-electrode voltage-clamp experiments were performed 2 to 3 days post-injection. Oocytes were injected with 50 nl of 100 mM BAPTA 4’ before recording to minimize calcium-activated chloride currents. Recording solutions contained 100mM LiOH, Ca(NO₃)₂ at the concentration indicated, and 10 mM HEPES, adjusted to pH 7.4 using HNO₃. Ca²⁺ concentrations were verified using a Ca²⁺ electrode. The solution with a nominal free Ca2+ concentration of 3nM contained 170 µM Ca(NO₃)2 and 15 mM EGTA. Buffered solutions with nominal free Ca²⁺ concentrations of 100nM, 300 nM, and 1 µM were also tested and gave results similar to the corresponding solutions in which Ca²⁺ was not buffered by EGTA. Electrodes were filled with 3M KCl and had resistances of 0.3–1.0 MΩ. Recordings were conducted at room temperature from a holding potential of −90 mV. Leak currents were subtracted using a P/4 protocol. Currents were analyzed with Clampfit 8.2 (Axon Instruments). All results are from at least two independent oocyte batches. Dose-response curves were calculated as follows: Ix/I_3nMCa = 1/(1+x/IC₅₀), where I_x is current at the Ca²⁺ concentration x and IC50 is the half-maximal inhibitory concentration.

Patch-Clamp electrophysiology

BacNaVs from Silicibacter pomeroyi, NavSp1, and Alkalilimnicola ehrlichei, Na_VAe1, were cloned into the pIRES2-EGFP vector (Clontech, Mountain View, CA). All the Na_VSp1 mutants were made using the QuikChange® Site-Directed Mutagenesis Kit (Stratagene, La Jolla, USA) and sequenced before recordings. Human embryonic kidney cells (HEK 293) were grown at 37 °C under 5% CO₂, in a Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum, 10% L-glutamine and antibiotics (100 IU/ml penicillin and 100 mg/ml streptomycin), (UCSF Cell Culture Facility). HEK 293 cells were transfected (in 35 mm diameter wells) with LipofectAMINE™ 2000 (Invitrogen, Carlsbad, CA, USA and plated onto coverslips coated with Matrigel (BD Biosciences, San Diego, CA, USA). 2 µg of DNA was used except for the L212A and I216A for which 4 µg of DNA was used to increase current amplitude.

Transfected cells were identified visually enhanced green fluorescent protein (EGFP) expression. Whole cell patch-clamp ⁶⁵ was used to record Na⁺ current at room temperature (23 ± 2 °C) 48–72 h post-transfection. Acquisition and analysis were performed using pCLAMP 9 (Molecular Devices, Sunnyvale, CA, USA) and an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA, USA). Pipettes were pulled from borosilicate glass capillaries (TW150F-3, WPI, Sarasota, USA) and polished (MF-900 microforge, Narishige, Tokyo, Japan) to obtain 2–3 MΩ resistances. Sixty to eighty percent of the voltage error due to the series resistance was compensated. Unless stated otherwise, pipette solution contained, in mM: 120 Cs-methane sulfonate, 8 NaCl, 10 EGTA, 2 Mg-ATP, and 20 HEPES (pH 7.4 with CsOH). Bath solution contained, in mM: 155 NaCl, 1 CaCl2, 1 MgCl₂, 5 KCl, 10 HEPES, and 10 glucose (pH 7.4 with NaOH).

For reversal potential measurements to determine the Na⁺ and Ca²⁺ relative permeabilities, pipette solution contained, in mM: 100 Na-Gluconate, 10 NaCl, 10 EGTA, 20 HEPES (pH 7.4 adjusted with NaOH, [Na] total 130). For tail protocol experiments on Na_VSp1 3Gly the pipette solution contained in mM: 30 Na-Gluconate, 10 NaCl, 10 EGTA, 20 HEPES, 70 NMDG-Cl (pH 7.4 adjusted with NaOH, [Na] total 45). External solution in both was, in mM: 5 NMDG-Cl, 100 CaCl2, 20 HEPES (pH 7.4 adjusted with CaOH, [Ca] total 107.5). The permeability ratio of Ca²⁺ over Na⁺ was estimated using the following equation:

$${\mathrm{P}}_{\text{Ca}}/{\mathrm{P}}_{\text{Na}}={\mathrm{a}}_{\text{Na},\text{int}}[\text{exp}({\mathrm{E}}_{\text{rev}}\mathrm{F}/\text{RT})][\text{exp}({\mathrm{E}}_{\text{rev}}\mathrm{F}/\text{RT})+1]/(4{\mathrm{a}}_{\text{Ca ext}}),$$

where R, T, F, and E_rev are the gas constant, absolute temperature, Faraday constant, and reversal potential, respectively (int, internal; ext, external)² Ca²⁺ and Na⁺ activity coefficients were estimated as follows:

$${\mathrm{a}}_{\mathrm{s}}={\mathrm{\gamma }}_{\mathrm{s}}[{\mathrm{X}}_{\mathrm{s}}],$$

where activity, a_s, is the effective concentration of an ion in solution, s related to the nominal concentration [X_s] by the activity coefficient γ_s. γ_s was calculated from the Davies equation. The calculated activity coefficients were γ_(Na)int = 0.76 and γ_(Ca)ext = 0.33. The liquid junction potentials were calculated by the JPCalc program (P. Barry) within Clampex (Molecular Devices) and taken into account to determine E_rev.

Highlights.

• Bacterial voltage-gated sodium channels (BacNa_Vs) neck region is involved in gating

• BacNaV neck region undergoes a helix-→disorder transition upon channel opening

• BacNaV selectivity filter has an outer ion binding site

• Outer ion binding site is shared with mammalian voltage-gated calcium channels