Conformational Mobility of Voltage-Gated Ion Channels

1. Introduction

Ion channels are a functionally important and extensive class of transmembrane proteins that play an important role in cell life. In excitable tissues, they are involved in the maintenance of homeostasis and resting potential, as well as in the development of action potential [1,2]. A number of studies demonstrate the regulatory role of ion channels in processes such as cell differentiation [3,4,5], cell growth [6,7,8], apoptosis [7], neurotransmitter release [9], and hormone release [8]. Diseases caused by the disfunction of ion channels can have consequences such as ataxia, epilepsy, hereditary migraines, heart rhythm abnormalities, hormone secretion disorders, osteopetrosis, development of cancerous tumors, etc [6,10,11,12,13,14,15,16,17]. Thus, the study of ion channel functioning is of vital importance for medicine, including personalized medicine. Interest in this group of proteins has led to the development of a number of biochemical and biophysical approaches for the study of ion channels.

In recent years, methods for studying ion channels have been developing rapidly. Purification to preserve the membrane environment, protocols for creating polarized vesicles to observe individual channel states are being developed, and structural biology approaches that allow the observation of conformational changes are being refined. In this review we address a number of solved problems related to ion channel performance and method improvement.

2. Topology and Classification of Cation Channels

Ion channels have a modular structure, with most ion channels comprising of three key functional modules [18,19]:

(i)

A pore domain (PD) with a narrow selective filter that conducts ions of a certain type;

(ii)

A gate that regulates the opening and closing of a channel;

(iii)

Sensors that respond to external signals. In cationic voltage-gated channels, four repetitive subunits are radially arranged around the pore [20]. The transmembrane α-subunit of Na_v and Ca_v channels consists of four transmembrane domains comprising of S1-S6 helices, encoded by a single polypeptide chain (Figure 1A). Helixes S1-S4 form the voltage-gated domain, while S4, the voltage sensor, being rich in positively charged amino acid residues, changes its position upon changing the membrane potential. The S5 and S6 helixes form a pore domain [2,21,22].

Potassium channels are tetramers, subdivided into four groups depending on the structural features of their monomers [23] (Figure 1B):

(i)

K_v (voltage-gated) possess six transmembrane helices (S1-S6), with S1-S4 forming a potential-sensing domain and S5-S6 – a pore domain;

(ii)

K_ir (internally rectifying) through which ions pass easily into the cell, but not out; they have two transmembrane helices,

(iii)

K_2P (bipore delayed rectification), with four transmembrane helices,

(iiii)

K_Ca (calcium-activated), have six transmembrane helices, similar to those of K_v.

K_v transmembrane helices are conserved, whereas cytoplasmic regulatory domains are structurally distinct in different families, fulfilling different functions [24].

Figure 1. Diagram of the structure of the α-subunit of voltage-gated ion channels. A – sodium channels, B – calcium channels, C – potassium channels [25].

Figure 1. Diagram of the structure of the α-subunit of voltage-gated ion channels. A – sodium channels, B – calcium channels, C – potassium channels [25].

Voltage-gated ion channels function following the "all-or-nothing" law, i.e., they are either in a state of total ionic conductance, or zero conductance [26]. Three basic functional states are known: closed (with the potential sensor lowered and the pore closed), open (with the potential sensor raised), and inactivated (closed even when the potential sensor is raised) (Figure 2) [27]. Several transition models describe open channel transitions to the inactivated (Figure 2A) or to the closed state (Figure 2B).

For a long time, experimental reconstructions of ion channels in intermediate conformational states have been lacking, however their presence was assumed. A highly conserved site (charge transfer center, CTC) of two negatively charged residues and phenylalanine was found in a voltage-gated domain, which presumably binds the positively charged amino acid of S4 and catalyzes the movement of the potential sensor [29]. For the K_v1.2 channel, five states: open, closed, and three intermediate ones, where different positively charged amino acids were bound to the CTC, have been suggested by a group of French researchers [30]. In addition, in 2011, a K_v channel deactivation model (MDM) was proposed to describe the intermediate conformations of the Shaker channel, among which an open, two intermediate, closed, and a deep closed state were identified [31]; also reviewed in [27].

To date, a number of intermediate states have been established for few channels. For example, PIP₂ binding is required for K_v7.1 activation [32,33]. For K_v7.1, a model with an intermediate S4 position was constructed based on experimental data [34]. For K_v10.2, reconstructions of several intermediate states were obtained, and, in addition, the closed state of K_v10.2 was able to be subdivided into several distinct sub-states [35].

Two topology variants have been described for voltage-gated ion channels [34,36]: (i) "domain-swapped" channels, in which the voltage-sensing domain (VSD) is spatially closer to the neighboring PD, and a long S4-S5 linker is present. These include K_v1-8, Na_v, Ca_v channels [37,38,39]. (ii) “non-domain-swapped" channels, where the VSD is spatially closer to the PD of the same polypeptide chain, with a relatively short S4-S5 linker. These include K_v7-12 channels. Attachment of low-molecular-weight ligands is required for the activation of these channels [34,40,41,42].

The development of methods that would allow the study of intermediate states of the ion channels, as well as continuous, rather than discrete, conformational changes based on experimental data, is of high relevancy.

3. Experimental Prerequisites for Studying the Structure of Ion Channels

Structural studies provide the most thorough information on the peculiarities of protein macromolecule functioning. Membranes of different cells are unique in composition; to study ion channels in a near native state, membrane-modulating media are used [28,43]. The selection of a suitable membrane mimetic is a separate non-trivial task that is solved for each specific protein. The main types of membrane mimetics used are detergents (micelles, bicelles), amphipoles, nanodiscs, and amphiphilic SMA polymers (lipodiscs) [43,44] (Figure 3). Recently, studies of channels incorporated directly into liposomes [45,46] have become widespread. All these methods will be reviewed below.

3.1. Purification Strategies

3.1.1. Detergents: Micelles and Bicelles

Detergents are amphiphilic substances including hydrophilic and hydrophobic parts that form micelles (Figure 3). According to their nature of action, they are distinguished as hard detergents that destroy protein-protein interactions (e.g., SDS) and soft detergents (e.g., TX-100, DDM, Tween, and OG). Zwitter-ionic, ionic, and non-ionic detergents are differentiated by their chemical nature. When selecting a detergent, general detergent properties: charge, fatty acid chain length, hydrophilic-hydrophobic balance, and other parameters, should be taken into account, as well as the selection of the optimal concentration, depending on the critical micelle concentration (CMC) [47,48,49]. Using detergents, the first reconstructions of the structures with the near atomic resolution were obtained for Kv7.1 [32], Kv10.1 [50], Kv11.1 [51], and a number of other ion channels [52].

The membrane-modulating properties of detergent micelles, which may differ significantly in spatial organization from the cell membrane, can be improved by the addition of phospholipids. The resulting particles, comprising of detergent and phospholipids, have been named bicelles (Figure 4) [53,54]. Bicelles have been used for various structural studies of ion channels [55,56,57], however they did not gain as much popularity as detergents, nanodiscs, and SMALPs, especially for cryoEM studies. The main problems in using bicelles are that the total lipid concentration can affect the size and geometry of a bicelle and that insufficient bilayer size may lead to the risk of membrane protein disruption [58].

3.1.2. Polymers: Amphipols, Nanodiscs, Lipodiscs

Amphipols (Figure 3) are artificially synthetized amphipathic polymers comprising of hydrophobic and hydrophilic parts, in which groups are randomly arranged, but their ratio is preserved. Amphipols embrace the hydrophobic regions of membrane proteins, making them hydrophilic [59]. Membrane proteins can be transferred to amphipols after solubilization in detergents, but this approach leads to heterogeneity of the samples [60].

Nanodiscs are proteolipid systems consisting of two copies of an amphipathic α-helical membrane scaffold protein (MSP) wrapped around a small disc of the lipid bilayer that they stabilize (Figure 3). In recent years, nanodiscs have become most popular for studying membrane proteins, due to their monodispersity, adjustable lipid composition, and high stability. Membrane proteins can be assembled into nanodiscs by co-dissolving the protein, lipid, and MSP in a single detergent [61,62,63,64,65].

The application of styrene-maleic acid (SMA) copolymers has been a methodological breakthrough in the development of membrane mimetics for the study of membrane proteins. This amphiphilic copolymer solubilizes biological membranes to form SMALPs (Styrene-Maleic Acid-Lipid Particles), a.k.a. lipodiscs, disc-shaped structures surrounding the membrane protein of interest; thus, preserving its native lipid environment (Figure 3)[66,67,68,69,70]. This recently introduced promising approach has already been actively applied into practice [68,70,71].

3.1.3. Lipids: Liposomes and Membrane Vesicles

Liposomes (Figure 3) are multilamellar membrane vesicles consisting of several concentric lipid bilayers [72,73]. A number of methods for their preparation are known, including dispersion of lipid films in aqueous solution, ultrasound treatment, and the extrusion of lipid dispersion through polycarbonate membranes [74]. Liposomes are diverse in bilayer number, size, and composition. By varying the lipid composition of liposomes, it is possible to approximate their properties to those of the cell membrane and place the protein under study in a close to native environment, which is a key advantage of liposomes [75].

Recently, studies on the structure of ion channels in vesicles obtained directly from cell membranes after ultrasonic treatment have emerged. The advantage of this approach is that the membrane protein remains in its native environment. The vesicles are purified by ion exchange chromatography. Both the proteins from intracellular membranes and the proteins from the plasma membrane are purified [76]. Recently, the Slo1 potassium channel embedded in membrane vesicles was examined by cryo-electron tomography [77], and reconstructions of Slo1 embedded into the intracellular and plasma membranes were obtained at 3.8 Å and 2.7 Å resolution, respectively. This procedure resulted in about 90% of the vesicles being derived from intracellular membranes. The proteins in the vesicles can be orientated in two ways, for the extracellular and intracellular parts different tags can be used for purification.

A method for studying ion channels in polarized proteoliposomes is currently being actively developed, which is discussed in detail in section 3.4.

3.2. Methods for Obtaining Ion Channels in Specific Functional States

For a long time, the study of ion channels by structural biology methods was hampered by difficulties in 3D crystallization for X-ray analysis [78]. A number of models describing the transition from one state of ion channels to another have been proposed, based on indirect studies; current models are based on structural data [27]. Studies of mutations leading to channelopathies have provided important information on the role of specific sites of ion channels for their functioning [79]. Other approaches, such as the introduction of cysteine substitutions [80,81], fluorescent methods (FRET, LRET, VCF) [82,83], and molecular modelling [84,85], have also provided a better understanding of the mechanisms of conformational transitions.

Pioneering structural studies of ion channels were performed by R. MacKinnon's group in late 1990-s by X-ray diffraction analysis [86]. A number of ion channel structures have been obtained through crystallization in the lipid cubic phase [87,88]. Now it is obvious that cryo-electron microscopy is best suited to work with membrane proteins, due to their large size, difficulty in crystallization, and requirement for lipid environment preservation [89].

Membrane disruption during cell lysis results in depolarization, leading to the voltage sensor S4 of the ion channel to be in a raised position, thus, many structures of the voltage-gated potassium channels were solved in open or semi-open states [34,51,90,91]. Therefore, for many ion channels, only structures for a fraction of the conformational states have been obtained to date [92,93,94,95], while homology modelling has been applied to study the mechanism of switching between states.

For the K_v10.2 channel, structures in different states (including several closed states) were recently obtained through the use of advanced data analysis algorithms for particles of proteins solubilized with detergents, however, the authors suggest the presence of other states, such as a deep closed state at -70 mV [35].

Various approaches can be used to obtain channel structures that are unlikely to occur when the membrane is destroyed: approaches based on the understanding of the biology of individual proteins and methods suitable for one family of channels, but that may be inapplicable for another. Let us consider some of these methods using specific ion channels as examples.

3.2.1. Application of Ion Channel Modulators

Calmodulin (CaM) was shown to bind to K_v10.1 (EAG1) in the presence of Ca²⁺ to inhibit ionic conductance, through сlosing the channel pore (Figure 4A) [96,97]. Each of the four EAG1 subunit has three CaM contacting sites that form two binding sites [97].

By binding simultaneously to cytoplasmic CMBD and PAS domains from neighboring subunits, CaM acts as a molecular clamp, pulling the two domains together and, thus, changing their orientation [50]. By using CaM-mediated channel inhibition, the group of R. MacKinnon obtained the structure of the K_v10.1 channel in a closed state with a resolution of 3.8 Å [50].

3.2.2. Peptide Binding Mimics the Functional States of Ion Channels

For a number of voltage-gated channels, for which the ligand-receptor model of action has been proposed, a method of fixing the channel in a particular state through peptides that mimic the S4-S5 linker ("ligand") or the C-terminal portion of the S6 ("receptor") can be used. With peptides mimicking the S4-S5 linker, the channel is closed in the presence of depolarization and voltage sensor bias. Similarly, with a peptide mimicking the C-terminus of S6, the channel can be secured in the open position without membrane depolarization (Figure 4B).

Experimentally, it was shown that it is possible to open K_v11.1 [37] and Na_v1.4 [98] and close K_v11.1 [37] and K_v10.1 [38] channels using chemical cross-linking to form disulfide bonds.

Interestingly, we have found a new mutation in the K_v7.1 channel [14] that disrupts the poly-Lys strip in the proximal part of the highly conserved cytoplasmic A–B linker of the channel, which was not shown before to be crucial for the correct functioning of K_v7 channels [99,100] and, thus, it was removed in several structural studies [101,102]. This mutation leads to the development of the LQT syndrome in the patient, demonstrating the importance of a flexible structural part.

3.2.3. Chemical Cross-Linking and Coordination of Metal Ions

Subsequently, a method to secure the channel in a closed state was proposed. When cysteines are introduced into the neighboring positions of the S4-S5 linker and the C-terminus of S6, disulfide bonds are formed under oxidative conditions to close the channel. Thus, for the D540C-L666C mutant variant of hERG, it was shown that, in the presence of an oxidizer, it does not conduct current when voltage is applied (Figure 4C) [37]. The same is applicable for the Na_v [98], K_v7.1 [39] and K_v10.2 channels [38].

Chemical cross-linking of PD and VSD cysteines has also been used for domain-swapped channels. The closed Na_vAb channel reconstruction with a resolution of 4.0 Å was obtained [103].

The coordination of the IIB-group of metal ions by cysteines has been used to reconstruct the resting state structure of K_v4.2 [28]. In this study, four channel states corresponding to the states of the inactivation model through the closed state were characterized and a possible mechanism of inactivation (through the disruption of channel symmetry) was described. The mechanism includes C4 symmetry breakdown while the channel becomes C2-symmetric. The C2 symmetric pore enables the creation of narrow constrictions along the ion conduction pathway. It is interesting that the closed-state inactivation of the K_v4.2 channel operates through a mechanism that differs from other voltage-gated ion channels [28].

3.3. Application of Toxins with Voltage-Gated Ion Channels

Some toxins, such as scorpion α-toxin, affect the voltage-gated domain in such a way that the potential sensor is blocked in the lowered state. When this toxin was exposed to the Na_v1.7-Na_vPaS chimeric channel (with the human Na_v1.7 scorpion α-toxin binding site transferred to the cockroach Na_vPaS channel to simplify expression), a structure was obtained with the potential sensor downregulated at near-atomic resolution, whereas in the apo state it was upregulated. Comparison of these reconstructions revealed displacements of the potential sensor during activation [104]. A similar effect is exerted by protoxin-II of the Peruvian green velvet tarantula. It was used to reconstruct the human Na_v1.7 in complex with a toxin with a voltage sensor in the ‘down’ position [105].

3.4. New Approach – Polarized Membrane Vesicles

To screen potential drugs acting on potassium ion channels, a method was developed using proteoliposomes incorporating the channel of interest, an ionophore (e.g. CCCP (carbonyl cyanide m-chlorophenylhydrazone) conducting protons) (Figure 5A). Vesicles prepared in a buffer rich in potassium ions are transferred to an isotonic buffer with sodium ions, resulting in a concentration gradient and an outflow of potassium ions from the vesicle through the potassium channel, which is balanced by an influx of protons through the ionophore. The fluorescent dye ACMA, whose fluorescence is quenched by protons, is used to visualize the process, with the proton passing through the lipid bilayer before binding and not after attachment (Figure 5A).

Thus, if the potassium channel is open, the medium inside the vesicle is acidified and the quenched dye accumulates inside the vesicle and the fluorescence signal drops, whereas if the channel is closed or blocked, no drop in fluorescence was observed. When a potassium ionophore is added to the system, even in the absence of potassium current through the channel, potassium can escape from the vesicle and the fluorescence drops in this case as well [106].

Subsequently, a similar system was used by R. MacKinnon's group to create polarized vesicles and to study ion channels within these vesicles during membrane polarization. In this approach, vesicles are formed in a buffer containing KCl and transferred to a buffer with isotonic NaCl solution on a column; polarization is achieved by the potassium current conducting ionophore valinomycin (Figure 5B). CCCP is also present in the system, making it possible to visualize the process using ACMA. Using this system allowed us to study the Eag channel with the potential sensor omitted and the potential sensor in the intermediate state by cryo-electron microscopy [107]. The same method was used to study Kv7.1, which allowed us to build a model of the activation of this channel upon PIP₂ binding. Models were constructed with the potential sensor omitted and the potential sensor in the intermediate state, which allowed to determine that, upon opening, the ligand binding site appears first, then PIP₂ attaches, and the pore is opened [34].

The described mechanism of channel closure is potentially applicable for structural analysis of ion channels in intermediate states, for which reconstructions have not yet been obtained. The structural analysis of the bacterial channel NaChBac under an electrochemical gradient was performed by cryoET and subtomogram averaging. This study explores the limits of studying small ion channels in polarized proteoliposomes [46].

4. Advanced Structural Methods for the Study of Voltage-Gated Ion Channel Conformational Changes

4.1. Cryo-Electron Microscopy

In the past decade, cryo-electron microscopy (cryoEM) has become one of the leading methods for studying the patterns of ion channel functioning [108,109]. With the revolution in cryoEM resolution, the number of structures obtained by this method has increased dramatically. Currently, more than 1000 ion channel structures are present in PDB: both cationic and anionic, from a variety of cellular compartments collected from a range of organisms using a variety of purification methods. Furthermore, the resolution obtained for ion channels using this method approaches atomic resolution [92,108,110]. Advances that have led to the resolution revolution in cryoEM include the advent of direct electron detection detectors and the development of data analysis algorithms [108,109]. Most of the published structures have a resolution in the range of 3 – 4 Å, the best resolution for ion channels is currently in the order of 1.8 Å. This result was obtained for the calcium-activated human chloride channel hBest2 [111].

The interaction of ion channels with their lipid environment is being actively investigated [112], aided by the development of biochemical methods for channel solubilization and purification (see Section 3.1). A number of studies demonstrate that lipids can be sequenced and can interact specifically with hydrophobic regions of ion channels [64,113,114]. For some cases, the regulatory role of lipids has been shown [115]. For example, phosphoinositide lipids serve as negative modulators of TRPV1, whose release from the binding pocket is a critical step towards activation [116]. PIP₂ binding is a prerequisite for the ability to activate K_v7.1 [34,41]. As another example, membrane lipids affect the conformation and function of the two-pore potassium channel TREK1 [117].

In addition to the structures of ion channels in different conformational states, cryo-electron microscopy was also used to study:

• chimeric channels [118,119,120],
• mutant channels [121,122,123,124,125,126],
• channels in complex with regulatory proteins, like calmodulin [41,125,127,128,129,130],
• channels in complex with low molecular weight ligands, both natural [131,132,133,134], and synthetic [33,135,136,137].

Obtaining ion channel structures in complex with ligands is not only of fundamental interest, but is also of practical significance for medicine [110].

Thus, at present, cryoEM is one of the most convenient, informative structural biology methods applicable to ion channel studies.

4.2. Development of New Algorithms for the Identifшсation of Distinct Conformational States of Ion Channels

There are a number of tools in cryoEM available to identify distinct conformational states during the analysis of cryo-electron microscopy data, including specifying multiple class types during initial model building, 2D and 3D classification, and heterogeneous refinement [138,139,140], which pertain to discrete state methods. Simultaneously, in recent years, new algorithms for analyzing conformational homogeneity that investigate continuous conformational changes have been developed [141,142,143]. For this purpose, PCA-based methods using data subsets [144], PPCA-based methods [145,146], and covariance estimation-based methods [147,148] were proposed.

In 2021, the developers of the cryoSPARC software package implemented the 3D Variability Analysis (3DVA) algorithm [146]. 3DVA allows the visualization of movements of individual structural elements of protein macromolecules, which offers new biological information based on cryoEM data. 3DVA is designed as a variant of the expectation-maximization algorithm for the probabilistic principal component method (PPCA). 3DVA resolves continuous conformational changes, allows the identification of new biologically relevant patterns from cryoEM data, and simplifies the analysis of conformational heterogeneity [146].

The 3DVA algorithm was used to analyze the conformational mobility of ion channels, including the complex of the potential-dependent sodium channel Na_v1.7 with the region of the immunoglobulin molecule that binds antigen (Na_v-Fab) [146]. A 3D analysis of variability was performed with a low-pass filter with a 3 Å resolution cut-off and six components of variability. Figure 7A shows part of the variability components. The first component describes the bending of the two transmembrane subunits and the movement of bound Fab’s. The outer transmembrane helices move left and right, while the Fab’s alternately approach and move away from each other. The second component reflects the lateral bending of the four α-helices of the cytoplasmic domain. The sixth component shows the up-and-down movements of the two subunits not associated with Fabs. High resolution of peripheral transmembrane helices could not be obtained in this work [149], 3DVA provides an explanation of what causes such limitations.

3DVA was also used to study potassium channels, namely the lysosomal channel TMEM175, which is evolutionarily distant from all known channels. The algorithm allowed for the characterization of the conformational heterogeneity and the identification of two states, which were further used for iterative rounds of heterogeneous refinement. Reconstructions at 2.6 and 3.0 Å resolution were constructed, providing a better understanding of the mechanism of channel selectivity and opening [150] (Figure 7B).

Methods that in the long term may allow a transition to energy computation are also being actively developed. Thus, the RECOVAR method [152], a PCA-based method computed using regularized covariance estimation, is similar in principle to 3DVA, but has a number of differences. For example, RECOVAR allows automatic regularization, which makes it more stable than 3DVA when choosing incorrect parameters, as well as faster and less computationally demanding [152].

Recently, the 3D Flexible Refinement (3D Flex) algorithm [153], a model of continuous heterogeneity based on a deep neural network, was implemented in cryoSPARC. 3DFlex directly exploits the knowledge that protein conformational variability is often the result of physical processes, given physical constraints. From 2D image data, the 3DFlex model learns a single canonical 3D map, latent coordinate vectors that define positions on the protein conformational landscape, and a flow generator that, given the latent position as input, outputs a 3D strain field. This deformation field transforms the canonical map into the corresponding conformations to explain the experimental images. When applied to experimental data, 3DFlex studies non-rigid motion spanning several orders of magnitude, while preserving high-resolution details of the secondary structure. In addition, 3DFlex resolves canonical maps, which are an improvement over conventional refinement methods. 3DFlex was used to analyze the TRPV1 ion channel (Figure 6С) [151].

Two types of flexible, coordinated motion among the four peripheral domains of the ion channel have been recorded. Along the first latent dimension, each pair of opposing subunits bends towards one another, while the other pair bends sideways. The second involves all four subunits twisting concentrically around the axis of the channel pore. In both cases, the most peripheral helices move about 6 Å. Both movements are non-rigid and involve the bending of substantial regions of protein density. Improved alignment using this algorithm improved the resolution of the peripheral regions of the protein from 4 to 3.2 Å [151].

The cryoDRGN method [143], based on the use of deep neural networks, also provides interesting results. This method has been applied in a number of studies of ion channels, including the characterization of the mobility of the N-terminal domain of the ligand-dependent DeCLIC channel and the role of this domain in the regulation of channel function [154].

Figure 6. A – 3DVA results with six variability components (three shown) on images of 300,759 Na_v1.7 channel particles [146], B – classes extracted for TMEM175 using 3D classification and 3DVA [150], С – 3DFlex results for TRPV1 [151].

Figure 6. A – 3DVA results with six variability components (three shown) on images of 300,759 Na_v1.7 channel particles [146], B – classes extracted for TMEM175 using 3D classification and 3DVA [150], С – 3DFlex results for TRPV1 [151].

4.3. Cryo-Electron Tomography

In cryo-electron tomography (cryoET), the sample is physically rotated around an axis perpendicular to the optical path of the microscope at different angles. The data obtained are used to construct a reconstruction – a 3D tomogram. Subtomogram averaging, a computational procedure that uses subvolumes extracted from the tomogram and that includes individual particles, is used to obtain a reconstruction of macromolecules [155,156]. Solutions for working with the membrane and membrane-associated proteins are also being developed, both experimental (e.g., using extracellular vesicles as a platform for structure determination [157]) and computational [158,159,160], as well as new algorithms and pipelines for data analysis [65,155,156,161,162,163,164].

In 2015, a reconstruction of the 5-HT3 receptor, a ligand-driven ion channel, was obtained using cryoET, with a resolution of 12 Å [165]. In 2020, a paper was published on the structure of the ryanodine receptor (RyR1) in native membranes, a resolution of the order of 12.6 Å was obtained. In this work, it was shown that, upon activation by ryanodine and calcium, a change in the receptor conformation led to a change in the membrane curvature, which may play an important role for channel opening [166]. More recently, for RyR1, the resolution was improved to 9.1 Å using a combination of subtomogram averaging and single particle analysis [167]. Regarding potential-dependent ion channels, a reconstruction of the bacterial NaChBac sodium channel in polarized proteoliposomes was obtained in 2023 with a resolution of ~16 Å [46].

Thus, the study of ion channels by cryoET provides an opportunity to analyze their function in native membranes, and, although the resolution has not yet reached the same orders of magnitude as in single particle cryoEM, new data analysis pipelines and algorithms that have emerged in recent years have a great potential.

5. Conclusions

Within the last decade, new approaches and algorithms have rapidly been developed to study the structure and conformational changes of voltage-gated ion channels. CryoEM is well suited for determining the 3D structures, but until recently its data analysis techniques were inefficient for reconstructing flexible domains and conformational changes in protein. After 2D or 3D averaging, the resolution of the mobile domain is usually lower than for stable regions of a protein [168]. Recently, a number of tools became available to identify distinct conformational states during the analysis of cryo-EM data, including 3D classification and heterogeneous refinement [138,139,140], as well as 3D Flexible Refinement [146]. In situ methods became more widespread, which will help identify unique structures of ion channels in the membrane.