Voltage-gated sodium channels (Navs) are responsible for the initiation of the action potential in excitable cells. Several prokaryotic sodium channels, most notably NavMs from Magnetococcus marinus and NavAb from Arcobacter butzleri, have been shown to be good models for human sodium channels based on their sequence homologies and high levels of functional similarities, including ion flux, and functional consequences of critical mutations. The complete full-length crystal structures of these prokaryotic sodium channels captured in different functional states have now revealed the molecular natures of changes associated with the gating process. These include the structures of the intracellular gate, the selectivity filter, the voltage sensors, the intra-membrane fenestrations, and the transmembrane (TM) pore. Here we have identified for the first time how changes in the fenestrations in the hydrophobic TM region associated with the opening of the intracellular gate could modulate the state-dependent ingress and binding of drugs in the TM cavity, in a way that could be exploited for rational drug design.
In excitable cells, the rising phase of the action potential is driven by the coordinated opening of voltage-gated sodium channels (Navs). Nine different Nav isoforms have been identified in humans; they are differentially expressed across the heart, skeletal muscle, central and peripheral nervous systems, and are responsible for every heartbeat, movement, sensation, and thought . Depending upon the isoform, mutations or dysregulation in Navs (called channelopathies) give rise to conditions such as epilepsy, cardiac arrhythmia, and chronic pain . Consequently, they are important targets for the development of novel pharmaceuticals .
Electrophysiological data reveal that Navs cycle through distinct functional states, including resting, activated, and inactivated. While the mechanism of drug binding to Navs has not been elucidated in detail, various drugs have been shown to have increased affinity for one of the functional states (known as state-dependent block) . Hence, designing novel painkillers or anti-epileptics requires not only structural information about all of the isoforms (to prevent cross-reactivity of drugs between channels), but also a detailed understanding of the structural changes that take place during their functional cycle.
All eukaryotic Navs share a common architecture; they are pseudo-tetrameric, with each of the four domains comprised of six transmembrane (TM) helices, four of which (designated S1–S4) bundle to form a voltage sensor (VS), whilst the remaining two TM helices (S5–S6) form the sodium-conducting pore (P) (Figure 1A, top). Sodium ions are coordinated by the selectivity filter (SF) residues surrounded by the selectivity filter helices (P1 and P2) at the extracellular side of the pore. From there, they pass through the TM central cavity, lined by the S5 and S6 helices, which ends at the pore gate at the intracellular surface of the membrane. In the 1970s, Hille proposed  that hydrophobic drugs would enter the channel via TM openings in the wall of the pore, which have since been termed fenestrations. To date, no structures of human sodium channels have yet been determined (either by crystallography or cryo-electron microscopy), owing to difficulties in isolating significant amounts of pure, folded, and active channels. Recently, structures of two eukaryotic Nav channels have been determined by cryo-electron microscopy, which show they share the same pseudo-tetrameric architecture as proposed for human channels. However, those samples did not exhibit functional properties, and/or their structures do not show the channel in a single identifiable functional state [6,7].
Overview of the channel structures.
Hence, most structural studies on sodium channels to date have focused on prokaryotic channels (Figure 1A, bottom). They share the same 6TM arrangement of helices, but are true tetramers instead of single polypeptide chains, with one monomer corresponding to one domain of a human protein. They are considered to be good models for human sodium channels , as they exhibit between 25 and 30% sequence identities (depending on the homologue) with the different human Navs, several prokaryotic Navs have similar sodium ion translocation activities  and one (NavMs from Magnetococcus marinus) has been shown to be blocked by human sodium channel blockers with similar affinities and kinetics as the human Nav1.1 isoform . These structures have revealed the domain-swapped arrangement of the channel, with each VS leading to its S4–S5 linker, which wraps around the pore domain of the next monomer, culminating in a coiled-coil C-terminal domain (CTD) (Figure 1A, bottom left), which has been shown to be important for inactivation.
Two recently determined crystal structures of full-length prokaryotic Nav channels [8,10] have revealed the structural differences in the intracellular gate regions associated with the open and closed states (Figure 1B). The present paper analyses and discusses their structural features, including associated differences in their TM ion pathways and in their intra-membrane fenestrations, and hypothesizes how such structural differences could be exploited in the design of new state-dependent drugs for the treatment of sodium channelopathies.
Crystal structures of NavMs and NavAb channels
In the present study, the crystal structures of ‘open’ NavMs (from Magnetococcus marinus)  and ‘closed’ NavAb (from Arcobacter butzleri)  were compared (Figure 2). These are the first full-length sodium channel structures with both their VSs and their pore domains, as well as their entire CTDs, visible. Both of these channels are functional, exhibiting strong sodium flux and selectivity. These two channels are closely related phylogenetically (with 44% sequence identity and 64% sequence similarity overall). More importantly, they exhibit 66% sequence identity in their pore domains (including the S5, P1, P2, and S6 helices) (Figure 3) thus facilitating structural comparisons. Their SF sequences (TLESW) are 100% conserved.
Comparisons of the open NavMs structure (PDBID = 5HVX) and the closed NavAb structure (PDBID = 5VB2).
Sequence alignments of the NavMs and NavAb channels.
The open forms of the NavMs channel are as yet the only crystal structures of a sodium channel with sodium ions visible in the SF [8,11]; the open NavMs structure has also provided us with a view of drug binding sites in the hydrophobic membrane cavity . NavMs and NavAb exhibit slightly different ion flux properties, with NavMs having a slightly more positive-shifted voltage dependence of activation [8,12,13]. This, and the presence of the sodium ions, may be why it is more likely to crystallise in the open state.
Overall comparisons of NavMs and NavAb structures
NavMs (Figures 4A, 5A and 6A) was crystallised in the presence of the non-ionic detergent HEGA10, in the I422 space group with one molecule in the asymmetric unit, and diffracted to 2.45 Å, the highest resolution structure of any sodium channel determined thus far. NavAb (Figures 4B and 5B) was crystallised in the presence of lipids and the zwitterionic CHAPSs detergent, in the P21221 space group with four molecules in the asymmetric unit, and diffracted to 3.2 Å resolution.
Comparisons of the transmembrane cavities of the channels.
Comparisons of fenestrations in the open NavMs and closed NavAb structures.
Comparisons of the fenestrations in the open NavMs and open (truncated) NavAb (PDBID = 5VB8) structures.
The overall RMSD of the Ca atoms between the open NavMs and closed NavAb structures is 3.3 Å. The main difference between them is a bend in the middle of the S6 helix, starting at the hinge residue T206, in the NavMs structure. This bend is ∼20° from straight (Figure 2) and ultimately enables the opening of the gate at the intra-membrane surface. The pore regions of the NavMs and NavAb structures (including the S5, helices, the SF, and the first nine residues of the S6 helices) superpose with an RMSD of less than 0.4 Å. However, the bend starting at residue T206 in NavMs means that the equivalent regions from T206 to the pore ‘exit residue’ (M222 in NavMs) differ by more than 6 Å RMSD and reflect a significant conformational change between the open and closed states.
More recently, a truncated version of NavAb  (Figure 6B) was created by the removal of the bottom of the S6 helix (below C226) and the entire CTD domain, which also produced an ‘open gate’ structure, albeit not in a full-length channel. The structure of its S6 helix is very similar to that of the open NavMs structure, with the hinge starting with residue V205 (the equivalent residue to T206 in NavMs). However, it exhibits a smaller bend angle of 16° and an overall RMSD of 2.0 Å in the region beyond the ‘hinge’ residue, when compared with the NavMs structure. Interestingly, there are detergent molecules bound at the bottom of S6 in the closed NavAb structure, but not in the open NavAb structure; these detergents could have prevented the S6 helix from adopting an open conformation. Indeed, it was when this lipid binding pocket region in the full-length NavAb structure was omitted from the construct , that NavAb crystallised in an open conformation.
Open and closed pore dimensions and ion permeability pathway
The dimensions of the ion-conducting TM pores calculated for the NavMs and NavAb structures (Figure 4) give insight into the features important for maintaining open and closed structures. Whilst the closed NavAb structure is slightly narrower than the open NavMs structure in the SF region (possibly due to the lack of sodium ions in the SF), overall the pores are roughly the same shape and size until they reach the hinge residue. The open NavMs structure exhibits a continuous central hole of at least ∼5 Å diameter (Figure 4), which is wide enough enable passage of at least a partially hydrated sodium ion throughout , until it reaches the intracellular surface of the membrane. In the open NavMs structure, the region of S6 beyond the hinge is stabilized by an extensive array of hydrogen bonds and salt bridges formed in the ‘interaction motif’ [8,14] (composed of residues R118, R119, and Q122 in the S4–S5 linker, plus other adjacent residues in S3) (Figure 1B, left); this appears to enable its gate to be in an open conformation. The closed NavAb structure, on the other hand, widens starting at the hinge region, but then narrows dramatically (<2.8 Å diameter, far less than the estimated diameter (∼5–6 Å) for a fully hydrated sodium ion ) as it reaches the coiled-coil helix of the CTD (Figure 1B, right), a feature which restricts the separation between the monomers at their C-termini.
The intracellular (or pore) gates of sodium channels are located at the intracellular edges of the lipid bilayer, at the lower end of the TM regions (Figure 4A,B), and provide the site for the opening of the channel into the cell interior. Hence, when open, they must be wide enough for a sodium ion to pass through them, if the channel is to be considered in the conducting, open state. The designations ‘open’ and closed’ thus arise from NavMs having a pore gate diameter which is greater than ∼4.8 Å wide at its narrowest dimension, whilst the pore gate diameter in the closed NavAb structure is <2.8 Å (Figure 4C), and hence unable to allow passage of hydrated (or even ‘naked’ sodium ions) across the membrane.
Fenestration sizes and key residues
To explore what other potential functional differences result from the opening of the gate, the size and shape of the intra-membrane fenestrations were compared (Figures 5 and 6). These fenestrations are openings to the surrounding hydrophobic region of the lipid bilayer in the four sides of the TM regions of the channel, with one fenestration present in each subunit. They have been proposed to provide the hydrophobic TM passageways through which hydrophobic molecules such as anaesthetics could enter into the central hydrophobic cavity of the channel, potentially blocking the passage of ions through the TM pore, and thereby producing modulation of channel function. Molecular dynamics calculations  have suggested that hydrophobic molecules (even lipid fatty acid chains) would be capable of entering into the TM region of the channel via these fenestrations. Hence, the new structures were examined in order to determine if the fenestrations in the open and closed state structures described above would provide comparable passageways or passageways with different dimensions and shapes.
The size and the shape of hydrophobic cavities lying perpendicular to the pore regions in the open NavMs and closed NavAb structures were compared in order to reveal any structural difference (Figure 5). The open NavMs structure (Figure 5A) has a narrow constriction at the entrance to the hydrophobic cavity (starting from the hydrophobic TM region), corresponding to the position of M204, one of the key residues lining the fenestration. In this region, ∼12 Å from the centre of the tetramer, the size of the cavity is at a minimum (∼2 Å, yellow arrow in Figure 5C). Notably, this occlusion is not present in the fenestration of the NavAb structure (Figure 5B), despite the presence at the equivalent position of the even bulkier residue, F203. Furthermore, the fenestration cavities of NavAb not only lack this constriction, but they are also much wider overall compared with those in NavMs, with the largest region underlining F203 (∼3.5 Å, yellow circle in Figure 5C), a bulge that is not seen in the open NavMs cavities. This difference in size is directly connected to the different architecture of the intracellular gate and the CTD in the open and closed structures. Indeed, in the open conformation, the bend of the S6 helices would not sterically allow a large hydrophobic cavity, such as the one present in the closed NavAb structure. This would also explain the restriction observed at the entrance of the open NavMs fenestration, even when compared with the larger F203 residues in the closed NavAb structure. These results further strengthen our hypothesis that the characteristics of the fenestrations change between the open and the closed conformation, getting narrower when the channel opens and wider when it closes. These observations then point to a novel opportunity for exploiting such size and shape differences to design specific state-dependent drugs.
On the basis of the above results, we propose that the sizes of the fenestrations change between functional (open and closed) states. However, one final question must be considered: a previous study  based on molecular dynamics simulations comparing the pore-only open NavMs open structure  and an earlier closed NavAb structure without a visible CTD  had suggested that differences in the fenestration sizes were due to differences in the sequences of the two homologues rather than due to open versus closed states. This was hypothesized because NavMs and NavAb have different size residues in two of the fenestration-impinging positions: the residue designated M79 in the pore-only NavMs structure (which is the same as residue M204 in the full-length NavMs structure) corresponds in position to F203, which has a bulkier side chain, in the full-length NavAb structure); additionally residue V85 in the pore-only NavMs (which corresponds to the larger M209 in the NavAb structure). Now, however, with the new full-length open NavMs structure (Figure 6A) and the truncated but open NavAb structure (Figure 6B) available, we have an opportunity to directly test the hypothesis that it is residue size rather than conformational state that dictates the fenestration size: direct comparisons of the narrowest fenestration constrictions in the open NavAb structure (Figure 6C) and those in the closed NavAb structure (Figure 5C) [which obviously both have the same sequences] at sites roughly 12 Å from the cavity centre, confirm that the closed structure has a much more extensive and wider fenestration at this location than does the open structure. Hence, the differences described above for the two intact full-length structures (NavMs open and NavAb closed) must not be due to the sequence differences of residues surrounding the fenestrations, but rather due to the states of the intracellular gates of the channels.
Qualitative and quantitative comparisons of the full-length open NavMs and closed NavAb structures have revealed the molecular nature of the opening and closing of the sodium channel intracellular gate, an important step of the action potential cycle, and how this affects other features of the channel structure such as the TM cavity where hydrophobic channel-blocking drugs bind, and the intra-membrane fenestrations, originally proposed to provide pathways for drugs into the TM cavity. The present study has also provided a potential new clue for the design of hydrophobic state-dependent drugs for the treatment of channelopathies, notably that the entry pathway for hydrophobic drugs is both larger and a different shape in the closed form than in the open form. Drugs designed with appropriate shape and dimensions could thus exploit these differences, producing state-dependent blockage of ion translocations and thus modulations of the function of aberrant disease-related neuron firings.
G.M., J.B., and A.S. all contributed equally to this manuscript, including analyses, writing, and production of figures. B.A.W. contributed to the writing and analyses and supervised the project. All authors approved of the final version of the paper.
This work was supported by grants R001294 and L006790 from the UK Biotechnology and Biological Sciences Research Council (to B.A.W.) and, previously, by a Pfizer Neusentis PhD studentship (to G.M.). J.B. was supported by a PhD studentship from the BBSRC LiDO programme.
The Authors declare that there are no competing interests associated with the manuscript.
These authors contributed equally to this work.