3D-aligned tetrameric ion channels with universal residue labels for comparative structural analysis

Abstract

Despite their large functional diversity and poor sequence similarity, tetrameric and pseudotetrameric potassium, sodium, calcium, and cyclic-nucleotide gated channels, as well as two-pore channels, transient receptor potential channels, and ionotropic glutamate receptor channels, share a common folding pattern of the transmembrane (TM) helices in the pore domain. In each subunit or repeat, two TM helices connected by a membrane-reentering P-loop contribute a quarter to the pore domain. The P-loop includes a membrane-descending helix, P1, which is structurally the most conserved element of these channels, and residues that contribute to the selectivity-filter region at the constriction of the ion-permeating pathway. In 24-TM channels, the pore domain is surrounded by four voltage-sensing domains, each with conserved folding of four TM helices. Hundreds of atomic-scale structures of these channels, referred to as “P-loop channels,” have been obtained through x-ray crystallography or cryoelectron microscopy. The number of experimental structures of P-loop channels deposited in the PDB is rapidly increasing. AlphaFold3, RoseTTAFold, and other computational tools can be used to generate three-dimensional (3D) models of P-loop channels that lack experimental structures. While comparative structural analysis of P-loop channels is desirable, it is hindered by variations in residue numbers and 3D orientations of the channels. To address this problem, we have developed a universal residue-labeling scheme for TM helices and P-loops. We further created a database of P-loop ion channels, PLIC: www.plic3da.com, which currently includes over 400 3D-aligned structures with relabeled residues. We use this database to compare multiple 3D structures of channels from different subfamilies. The comparison, which for the first time employs statistical methods, highlights conserved and variable elements in the channels' folding, reveals irregularities, and identifies outliers that warrant further analysis.

Significance

Comparative structural analysis of voltage-gated and other tetrameric and pseudotetrameric potassium, sodium, and calcium channels with membrane reentering P-loops (P-loop channels) is hindered by variations in residue numbers and three-dimensional (3D) orientations of the channels. Here, we present an approach to compare the 3D structures of these channels that involves universal residue labels and an algorithm of 3D alignment. This approach, along with the database of 3D-aligned and relabeled P-loop channels: plic3da.com, is expected to facilitate the analysis of disease mutations and assist in the development of new selective drugs.

Introduction

Ion channels are a diverse class of proteins that vary in subunit composition (trimeric, tetrameric, pentameric, and hexameric), transmembrane topologies of subunits, domain organization, and activation mechanisms. Classifications of channel-forming proteins based on functional characteristics, such as activation mechanisms or primary permeant ions, often show poor correlation with their architecture and evolution due to convergent and divergent processes. From an evolutionary and structural perspective, it seems reasonable to categorize channels according to the spatial architecture of the pore domain (PD), which is the major component of all these proteins.

Following the publication of the crystal structure of the KcsA prokaryotic potassium channel from the soil bacterium Streptomyces lividans (1), advances in x-ray crystallography and high-resolution cryoelectron microscopy (cryoEM) have significantly facilitated both experimental and computational structural studies of various ion channels in different functional states and in complexes with various ligands, including important drugs and naturally occurring toxins (2,3). Multiple atomic-scale structures have clearly demonstrated that many channels share significant structural similarities despite having poor sequence similarity. In particular, it has been found that a structurally conserved tetrameric or pseudotetrameric PD is present in various voltage-gated ion channels (VGICs), including potassium (Kv) channels (4,5,6,7,8), sodium (Nav) channels (9,10,11,12), calcium (Cav) channels (13,14,15), transient receptor potential (TRP) channels (16,17,18,19), two-pore channels (TPCs) (20,21), and cyclic-nucleotide-gated channels (22,23,24), as well as in numerous non-voltage-activated tetrameric channels. These include but are not limited to ionotropic glutamate receptor channels (iGluR channels) (25,26,27,28,29), potassium inward-rectifying channels (4,30,31), pH-sensing and other potassium (K) channels (1,4,32), and sodium and potassium permeating (NaK) channels (33,34).

Tetrameric cation channels vary greatly in their functional characteristics, including ion selectivity and permeation, gating kinetics, modulation, and regulation. However, in all these proteins, the PD contains eight transmembrane (TM) helices, which are connected by four membrane-reentrant loops (P-loops). Each P-loop includes a membrane-descending pore helix P1. Residues that are C-terminal to P1 converge to the pore axis to form a narrow selectivity filter, which discriminates among various cations and divides the ion permeation pathway into two parts. Each P-loop in Nav, Cav, and TRP channels and some TPCs also includes a membrane-ascending helix, P2, which was first observed in the crystal structure of the NavAb channel (12). The bundle of four P-loops is surrounded by four inner helices (S6 in VGICs) and four outer helices (S5 in VGICs). Residues at the C-ends of the inner helices form the activation gate, which opens in response to various stimuli to allow ion permeation. These conserved features support our proposition to classify numerous channels with vastly different functional properties into a large superfamily of P-loop channels. This designation reflects the distinguishing feature of the channel’s architecture: an indispensable membrane-reentering P-loop in the pore domain.

Kv, Nav, Cav, and TRP channels possess voltage-sensing domains (VSDs) that contain TM helical segments S1–S4, although in some channels, the VSDs may be not functional. However, many channels, including iGluR channels, potassium inward-rectifying channels, KcsA, and NaK channels, lack VSDs. Therefore, the term "voltage-gated channels" or VGICs is not applicable to the entire superfamily. "P-loop ion channels," abbreviated as PLICs, seems to be a more appropriate designation. Indeed, every channel in the superfamily features four P-loops, with the membrane-descending P1 helices being the most structurally conserved elements. To some extent, the name P-loop channels parallels the term "Cys-loop channels" (35,36), which refers to another major superfamily of ion channels that includes pentameric ligand-gated channels.

Potassium, TRP, and iGluR channels are tetramers. TPCs are dimers of dimers, where two repeats resembling a subunit in a tetrameric channel are linked to form a single chain. In eukaryotic Nav and Cav channels, the pore-forming α subunit is composed of a single polypeptide chain that folds into four homologous but not identical repeats. In the majority of P-loop channels, P-loops reenter the cytoplasmic membrane from the extracellular side. However, iGluR channels have the opposite orientation relative to the membrane, with P-loops reentering from the cytoplasm. P-loop channels usually feature a domain-swap architecture, where a VSD forms multiple contacts with the PD quarter from the adjacent subunit/repeat. Nevertheless, some potassium channels adopt nonswapping domain architecture, with a VSD forming contacts with the PD quarter from the same subunit (37,38). Additionally, intra- and extracellular loops between the TM segments vary greatly in sequences and geometry and may contain functional domains.

Despite the generally common architecture of the pore domains, their comparison reveals significant structural variations, particularly in the ion selectivity filter region, which differentiates between various cations. Large variations are also observed in the C-terminal parts of the inner helices, which move during activation gating. Of particular interest are interfaces between the pore-lining inner helices. These interfaces are relatively narrow in potassium channels but much wider in Nav and Cav channels. In the latter channels, the interfaces are referred to as "fenestrations" and are proposed to contribute to hydrophobic access pathways for the ingress and egress of small-molecule ligands into the inner pore (12). The existence of such hydrophobic pathway has long been proposed by Bertil Hille (39), and theoretical studies involving molecular modeling of eukaryotic Nav channels with drugs have predicted the exact location of this pathway between the S6 helices of repeats III and IV (S6_III and S6_IV) as the "sidewalk" (40,41).

In addition to differences between channel families, there are significant structural variations between the functional states within each family. Channels exist in three major groups of functionally and structurally different states: the open ion-permeable states, resting nonpermeable states, and nonpermeable inactivated states. Moreover, specific ligands can induce structural deformations in P-loop channels. As a result, numerous nonidentical three-dimensional (3D) structures can exist for a single channel.

Comparative structural analysis of multiple structures is essential for identifying conserved and specific characteristics within each channel subfamily and across subfamilies. However, systematic analyses of numerous structures is challenging due to varying orientations of the 3D structures deposited in the PDB: https://www.rcsb.org/pdb and the AlphaFold database: https://alphafold.ebi.ac.uk, inconsistent labeling of protein chains, and different sequential numbers of residues in matching positions of sequence alignments. While a "manual" comparison of individual structures is possible, it does not allow for the application of high-throughput techniques and machine-learning algorithms in structural analyses. Additionally, the potential to translate experimental data from one subfamily to another would be significantly enhanced if the structures were 3D aligned and their residues had universal labels. The possibility of designing such a universal labeling scheme further supports the unification of different ion channel subfamilies with generally similar PD folding into the P-loop channels superfamily.

A similar issue with residue numbering has been known for the superfamily of G-protein-coupled receptors (GPCRs), which have only seven TM segments, whereas voltage-gated P-loop channels have 24 TM segments. The problem for GPCRs was resolved by Ballesteros and Weinstein, who developed a universal numbering scheme for TM segments (42). In this scheme, the most conserved residue in each TM helix is assigned the relative number 50 (e.g., Asn^2.50 in the second TM helix), with other residues numbered relative to this position. This system is now widely used in publications involving GPCRs. The paper introducing this scheme, which greatly facilitates comparison between different GPCRs, has been cited over 3000 times.

We proposed a universal labeling scheme for the pore domain of various P-loop channels in 2004 (43) and have used it in numerous publications, e.g., (44,45). In August 2024, a versatile residue-numbering scheme for Nav and Cav channels, based on an unambiguous sequence alignment of these highly homologous channels, was introduced (46). In the present work, we tackle the more challenging task of developing universal residue labels for all P-loop channel subfamilies (henceforth referred to as PLIC labels), despite the poor sequence similarity between some subfamilies. We also developed an algorithm of 3D alignment of the channels to facilitate comparative studies. Furthermore, the labeling scheme encompasses not only TM segments but also extracellular and intracellular loops. Based on the PLIC-labeling scheme proposed here, we created a PLIC database: www.plic3da.com. The database currently includes over 400 3D-aligned structures with PLIC labels (henceforth referred to as the PLIC-3DA database) where coordinate files are in the PDB format and can be used with standard software. Using this database, we provide several examples of structural comparisons across P-loop channels from different subfamilies. The results demonstrate that simultaneous analysis of multiple P-loop channels from various subfamilies is productive.

Methods

Reference residues

We employed the following approach to develop a universal residue-labeling scheme that accommodates all P-loop channels with conserved PD folding. First, we established a general pattern of residue labels applicable to all P-loop channels, regardless of their diverse domain organizations, while having the conserved folding of the PD. Second, we step-by-step defined labels for the TM segments, beginning with the most conserved key elements in the superfamily: the P-loops. We analyzed numerous structures from various subfamilies to select a "reference residue" for each segment, defined by its 3D-conserved position and orientation. Choosing reference residues would be straightforward if the amino acid sequences could be unambiguously aligned. However, different subfamilies have poor sequence similarity and diverse experimental 3D structures. To resolve this problem, we selected reference residues to ensure that in the 3D-aligned structures, CA–CB bonds in the same-label residues of individual helical segments (S1–S6, P1, and P2) have parallel orientations. Notably, the assignment of a reference residue within a given helical segment is unambiguous because even a one-position shift would result in an ∼100° reorientation of the CA–CB bonds across all residues of the respective helix. We assigned reference residues once and only once for each subfamily to achieve similar orientations of the same-label residues in all 3D-aligned P-loop channels.

General organization of residue labels

In tetrameric channels, the four subunits are assigned chain labels A, B, C, and D, as commonly used in many PDB structures (Fig. 1 A). In pseudotetrameric Nav and Cav channels, repeats I, II, III, and IV of the pore-forming α subunit are similarly labeled A, B, C, and D, respectively. In all tetrameric channels, chains A, B, C, and D are arranged clockwise when viewed from the extracellular space (Fig. 1 A) to match the arrangement of repeats I, II, III, and IV, respectively, in Nav channels (47).

Figure 1.

Universal labeling scheme of P-loop channels. (A) The extracellular view of a 24-TM channel (PDB: 2R9R). Note a clockwise arrangement of subunits I, II, III, and IV labeled as A, B, C, and D, respectively. Transmembrane helices S1, S2, S3, and S4 (red, green, yellow, and blue, respectively) form voltage-sensing domains. Helices S5 and S6 along with the P-loop (magenta, cyan, and blue, respectively) contribute one-quarter to the PD. (B) Membrane topology of subunit/repeat "A" in a 24-TM channel. TM segments and P-loops are colored as in (A). Circles indicate the approximate positions of reference residues Ax.550 in the TM helices and residue A5.850 in the P-loop. Arrows show approximate positions from which residues in the loops are counted.

A PLIC label has two parts delimited by a period (Fig. 1 B). The first part refers to the chain (A, B, C, or D), which may be omitted, and a number from 1 to 6 that indicates respective segment (S1 to S6). Number 7 refers to residues located C-terminal to S6 (e.g., helix M4 in iGluR channels). The second part of the label has three digits representing the relative residue number counted from the reference residue. In helices S1 to S6, the reference residues are numbered 1.550 to 6.550, respectively (Fig. 1 B). There are two reasons for using four-digit numbers of reference residues. First, in some experimental structures, e.g., NavAb (5VB2), the long S6 helices extend to the cytoplasm, leading to large residue numbers at their C-terminal ends (S^6.609 in NavAb). Second, some loops, e.g., the S5-S6_I loop in channel hCav3.3 (7WLI), contain over 100 residues.

A particular problem was defining common boundaries in the TM segments. In 3D-aligned channels from different subfamilies, the N- and C-ends of helices often do not coincide. Due to poor sequence similarity, we were unable to unambiguously assign common borders between the TM segments and their adjacent loops. This inconsistency in segment boundary definitions complicates comparative analysis, as some channels may lack residues with corresponding labels. For example, in the NavPaS channel (PDB: 6A90), the shortest loop between helices S1 and S2 in repeat I consists of only one residue, and the shortest S2-S3 loop has six residues. Furthermore, repeat I in Nav channels lacks the S3-S4 loop. In most VGICs, the sliding voltage-sensing helix S4 is directly connected to the linker helix S4-S5 without an intervening loop. Therefore, we considered the S4-S5 linker helices as the N-terminal parts of S5 helices and assigned them corresponding labels. In TPCs, the shortest loop between helices S5 and P1 contains only three residues, and in some channels, such as iGluR channels, the P-loop and S6 helix are directly connected. Given that loop lengths, amino acid sequences, and 3D structures vary widely, we did not attempt to align their sequences and instead treated loops as extensions of their respective TM helices. The N-ends of the intra- and extracellular loops S1-S2, S2-S3, S3-S4, S5-P1, and P2-S6 are assigned numbers 1.600, 2.600, 3.600, 5.600, and 5.900, respectively (Fig. 1 B), as we did not find any P-loop channel in which the respective helices have residues more than 50 positions away from the reference residue 550 or 850. To date, we are unaware of any P-loop channel whose PLIC labels do not conform to this general scheme.

Results

Reference position in P-loops and their 3D alignment

The cone-shaped bundle of four membrane-descending P1 helices is the most 3D conserved feature across all P-loop channel subfamilies. P-loops are not TM segments; in 24-TM and 8-TM channels, they are located, respectively, between helices S5 and S6 and S1 and S2. The C-ends of P1 helices converge toward the pore axis, while their N-ends form tight contacts with the inner and outer TM helices. For potassium channels, we selected valine in the signature sequence motif TVGYG as the reference residue for P-loops, assigning it the PLIC number 5.850 (V^5.850). While the first digit “5” typically designates S5 helices, P-loops are not part of S5 helices. Since the digit “6” is reserved for S6 helices, we assigned the reference residue in P-loops a large number (850) to prevent overlap with residue numbers in the S5-P1 linkers, even in cases where the linkers are unusually long.

For Nav and Cav channels, achieving the same 3D orientation of P-loop residues that matches those in potassium channels is possible only if the selectivity filter residues (the DEKA and EEEE rings) are assigned the label 5.850. In iGluR channels, the orientation of the CA–CB bonds in P1 residues, which matches those in Nav, Cav, and potassium channels, is only achievable if the residue located one position C-terminal to the selectivity filter (the N/Q/R site) is assigned label 5.850. Consequently, the selectivity filter residues in iGluR channels have been labeled 5.849. In TRPV channels, the TIGMGD motif is homologous to the TVGYGD motif in potassium channels. However, assigning label 5.580 to the isoleucine in the TIGMGD sequence resulted in an opposite orientation of the P1 residues compared to potassium channels. To ensure matching orientations of the same-labeled P1 residues, label 5.580 was assigned to the methionine in TIGMGD, which is located two positions upstream of the isoleucine (48). Thus, the same-label P1 residues in different subfamilies do not necessarily occur at matching positions of the sequence alignment (Table S1), but they do have parallel orientations of the CA–CB bonds in 3D-aligned structures.

Assigning labels to P helices allows unambiguous 3D alignment of structures. For these alignments, we selected the Kv1.2–2.1 chimeric potassium channel in the open state (PDB: 2R9R) as the template. This is the first eukaryotic P-loop channel whose crystal structure was determined with a resolution below 2.5 Å (8). The crystal structure was reoriented so that the pore axis coincided with the extracellularly directed z axis. The four backbone oxygen atoms O_V^5.850 were positioned in the xOy plane, with one of these atoms, designated O_V^A5.850, located at the Ox axis.

We initially compiled a dataset of 380 structures, including 88 Nav channels, 48 Cav channels, 45 TPCs, 30 iGluR channels, 64 TRP channels, 44 Kv channels, and 61 other potassium and NaK channels. The original structures were downloaded from the PDB (www.rcsb.org) as “biological assembly” files. These structures were 3D aligned to the reference structure (PDB: 2R9R) by minimizing root mean-square deviations between the CA atoms of residues 5.838–5.847 in the P1 helices. Residues with these labels are present in nearly all P-loop channels, and the P1 helix bundle is the most structurally conserved region in all P-loop channels.

The 3D-aligned 2R9R and 3RBY structures are shown in Fig. 2 A. Fig. 2 B shows P1 helices of these structures with bonds CA–CB at position 5.842 in the middle of the helices represented by sticks for many representative structures. The PLIC labels ensure 3D alignment, providing a similar orientation for residues with the same label. Despite the parallel orientation of bonds CA–CB^5.842, there are notable variations in the distances between the pore axis and the P1 helices and, to some extent, in the slope of the P1 helices relative to the pore axis. For instance, the CA^5.842 atoms in different channels can be as far as 5 Å apart. Most importantly, these deviations of the P1 helices from the pore axis are not random but strongly subfamily dependent (Fig. 2 B). Differences between individual structures within a subfamily are much smaller than those between subfamilies (Fig. 2 C). In various K and Kv channels, as well as in iGluR channels, the P1 helix is positioned close to the pore axis, while in Cav and Nav channels, it is more distant. TPCs distinctly fall into two categories: in 24-TM channels, the P1 helices are positioned even farther from the pore axis than in Navs, whereas in 8-TM channels, they are positioned similarly to Kv channels (orange sticks in Fig. 2 B correspond to structures 4BW5 and 6E1M). In most TRP structures, the P1 helices are situated at intermediate distances from the pore axis (between Kv and Nav/Cav channels). However, in some TRPV2 and TRPM structures (e.g., 6NR2, 6NR3, 6NR4, 6DRJ, 6DRK, 6O6A, and 6BWJ), the P1 helices are most distant from the pore axis. It is unclear whether the diversity of the P1 positions in TRP channels is related to their gating at the selectivity filter. This interesting feature deserves further investigation.

Figure 2.

Conservation and variability of P1 helices. (A) 3D-aligned channels Kv1.2–2.1 (red, 2R9R) and NavAb (blue, 3RBY). Side view of the pore domain with two nonadjacent subunits. P1 helices shown as thick ribbons. (B) P1 helices in 3D-aligned structures 2R9R and 3RBY. The CA–CB bonds at positions 5.842 and 5.850 are shown as sticks and colored according to the channel subfamily: Kv and KcsA (red), other K channels (magenta), Nav channels (blue), Cav channels (cyan), iGluR channels (yellow), TRP channels (green), and TPCs (orange). Note parallel orientations of the CA–CB bonds at position 5.842 but their varying distances from the pore axis. The enlargement shows the CA–CB bonds at position 5.842 in several channels: NaK (2Q68, magenta), Kv12-Kv21 (2R9R, red), 2TM-TPC (4BW5, gray), AMPA (6QKC, yellow), TRPV3 (6UW8, green), NavAb (3RVY, blue), Cav1.1 (7JPL, cyan), and 6TM-TPC (6E1M, orange). (C) Distances of the CA atoms at position 5.842 from the pore axis in different channel subfamilies (mean ± SD). In the K, Kv, and iGluR channels and TPCs, the average distances are significantly smaller compared to those in the Nav, Cav, and TRP channels.

By utilizing the proposed PLIC-labeling scheme and associated 3D alignment, we were able to, for the first time, systematically compare 3D arrangements of the P1 helices across all subfamilies of P-loop channels. The comparison revealed that the P1 helices have rather similar inclinations relative to the pore axis, but their distances from the axis vary between subfamilies. Given the large number of 3D-aligned structures, these conclusions remain consistent regardless of the resolution and specific characteristics of individual experimental structures.

Labeling residues in TM helices of the pore domain

The 3D alignment of the PLIC channels, based on the structural conservation of their P1 helices, allowed us to select reference positions in TM segments and assign them numbers x.550. In each TM helix, we considered various candidates for the reference position and have chosen one where the CA–CB bonds in multiple 3D-aligned structures are parallel and close to each other. S6 helices are structurally conserved in their N-terminal ends, which are tightly packed against the P1 helices and C-ends of S5 helices. In contrast, the C-terminal parts of S6 helices, which move significantly during activation gating, exhibit variability in 3D-aligned structures. For the inner helices in potassium channels, we used the conserved glycine gating hinge as a reference residue and assigned it label 6.550. For other subfamilies of P-loop channels, we selected reference positions 6.550 to ensure that the CA–CB bonds in the N-terminal half of the S6 helices were parallel and close to the same-label CA–CB bonds of the Kv1.2–2.1 channel (Table S2).

To select reference positions for the outer helices, we considered that in 3D-aligned P-loop channels, the C-ends of the S5 helices are structurally conserved and form tight contacts with the N-ends of the S6 and P1 helices, whereas the N-terminal parts of the S5 helices are highly variable. Most S5 helices contain a small residue (G/A/S) at the C-terminal end. We selected these as the reference residues 5.550 (Table S3). In the majority of P-loop channels, these G/A/S residues participate in intrasubunit (intrarepeat) contacts with residues at the N-terminal part of the P1 helices.

Fig. 3, A and B, show the extracellular and membrane views, respectively, of the pore domain in 3D-aligned channels Kv1.2–2.1 and NavAb. In nearly all 3D-aligned P-loop channels, CA–CB bonds in the C-terminal part of the S5 helices exhibit parallel orientations and similar positions, and this is also true for the N-terminal part of the S6 helices (Fig. 3, C and D). However, atypical orientations of the S5 helices are observed in structures 3PJS and 1ORQ, where the CA–CB bonds deviate significantly from those in other channels. Given the variation in distances of the P1 helices from the pore axis across different subfamilies (Fig. 2), we anticipated subfamily specific clustering of the S5 and S6 helices. Indeed, clustering was observed when visualizing the CA–CB bonds at positions 5.554 and 6.546 (Fig. 3, C and D). In Kv and KcsA structures, the N-terminal parts of the S6 helices are generally closer to the P1 helix compared to Nav and Cav channels (Fig. 3, E and F). Unexpectedly, residues at the N-end of S6 helices in Kv7.1 and Kv7.2 channels are positioned similarly to those in Nav channels, whereas Kv1.2 and KcsA structures form a separate cluster (Fig. 3 C). The reason for this previously unrecognized diversity is unknown.

Figure 3.

Conservation and variability of helices S5 and S6. (A and B) Extracellular and membrane views of the pore domain in 3D-aligned channels Kv1.2–2.1 (PDB: 2R9R, magenta) and NavAb (PDB: 3RVY, violet). (C and D) Extracellular and membrane views of helices S5, S6, and P1. Bonds CA–CB in positions 5.554 (C-terminal part of S5) and 6.546 (N-terminal part of S6) are shown as sticks in multiple 3D-aligned channels. CB atoms are colored red for Kv channels, blue for Nav channels, and cyan for Cav channels. Interestingly, in Kv7.1 and Kv7.2 channels, position 6.546 is close to that in Nav and Cav channels but distant from Kv1.2 and KcsA. (E and F) Distances of CA atoms at positions 5.554 and 6.546 from the pore axis across different channel subfamilies (mean ± SD). In K, Kv, and iGluR channels, the average distances are significantly smaller than those in TPCs and Nav, Cav, and TRP channels.

The 3D clustering of segments in the densely packed regions of the pore domain raises questions about specific interactions that determine the folding similarities and variations between different channel subfamilies. These interactions likely involve specific intersegment contacts. However, the structural determinants responsible for these folding variations remain unknown. Previously, we identified distinct folding-stabilizing interactions in some Kv, Nav, and TRP channels (48). However, our earlier study was based on a small number of manually aligned structures. Using multiple 3D-aligned structures would allow applying high-throughput techniques and machine-learning algorithms to identify folding determinants in large datasets of P-loop channels.

Comparison of the activation-gate region

In crystal and cryoEM structures of various P-loop channels, the pore domain is captured in different functional states. Consequently, we did not observe any subfamily specific dimensions of the pore lumen or other 3D characteristics in the C-terminal parts of the inner helices. The 3D alignment of the channels allowed us to analyze their structural features by visualizing CA–CB bonds of the same-label residues. Fig. 4 shows the CA–CB bonds at position 6.556. As expected, the distances of CA^6.556 atoms from the pore axis vary significantly and do not form subfamily specific categories. However, structural diversity is evident not only in different openings of the activation gates but also in the orientations of the CA–CB^6.556 bonds, which are directed either away from the pore axis or toward a neighboring subunit/repeat (Fig. 4 A). This qualitative difference is due to the presence or absence of π-bulges in the S6 helices near position 6.550. A π-bulge introduces an additional residue per helical turn, causing an ∼100° reorientation of residues at the C-terminal half of the S6 helix.

Figure 4.

π-Bulges in the inner helices affect orientation of residues at the activation gate region. (A) The pore domain in channels Kv1.2–2.1 (2R9R, red) and TRPV1 (3J9J, green). Spheres show side chains in position 6.565. Due to the π-bulged S6s in TRPV1 and a fully α-helical S6 in Kv1.2–2.1, the side chains have different orientations. (B and C) Extracellular view of bonds CA–CB^6.565 with CB atoms colored as in Fig. 2. In potassium and iGluR channels (B), the orientation of these bonds is relatively conserved despite a significant dispersion of their positions. In Nav, Cav, and TRP channels and TPCs (C), which have asparagine in position 6.556, the orientations of bonds CA–CB^6.565 are divided into two categories based on whether or not a π-bulge is present in the middle of S6. In the fully α-helical structures, the CA–CB^6.565 bonds point away from the pore axis, whereas in structures with π-bulges, the bonds are directed toward neighboring subunits or repeats. (D) The definition of angle ɸ, which quantifies the orientation of the CA–CB^6.556 bond. (E and F) Distribution of cos(ϕ) for the channel subfamilies depicted in (B) and (C), respectively. For K, Kv, and iGluR channels (B), the values indicate a small angle. In contrast, for Nav, Cav, and TRP channels and TPCs, there is a clear maximum at ϕ ≈ 90° (π-bulges).

Analysis of the 380 structures representing all the channel subtypes in the PLIC-3DA database revealed that in NaK, iGluR, and potassium channels, CA–CB^6.556 bonds are oriented away from the pore axis, even though individual structures are very diverse (Fig. 4 B). No π-bulged S6 structures were observed in any of these subfamilies, regardless of the channel gate state. Even in exceptional cases like 4GX1, the orientation of the CA–CB^6.556 bonds is similar to that in most other channels (Fig. 4 B). Only in some Nav, Cav, and TRP channels and 24-TM TPCs are the CA–CB^6.556 bonds oriented toward neighboring subunits, indicative of a π-bulge (Fig. 4 C). These channels feature a highly conserved asparagine (N^6.556), which is not present in K, Kv, NaK, and iGluR channels or 8-TM TPCs (Table S7). To quantify these differences, we measured angle ɸ between projections of line CA–CB^6.556 and the line from CB^6.556 to the pore axis at a plane perpendicular to the axis (Fig. 4 D) and analyzed the distribution of cos(ɸ) across different channel subfamilies (Fig. 4, E and F). For K, Kv, and iGluR channels, all cos(ɸ) values reflect an outward direction for the CA–CB^6.556 vector. In contrast, π-bulged S6s show a second peak at low cos(ɸ) values, corresponding to ɸ ≈ 90° (Fig. 4 F).

The above observation highlights the significance of the asparagine N^6.556 side chains in channel structures. These side chains can donate H-bonds to the backbone carbonyl groups in positions 6.551 or 6.552, which do not accept H-bonds from the backbone amide groups in π-bulged S6 helices. The presence of an H-bond between the N^6.556 side chain and a "bachelor" backbone carbonyl would help stabilize a π-bulge. This explains why π-bulges are absent in channels that lack N^6.556. It is likely that the conserved asparagines evolved as insertions into the S6 helices, and the resulting π-bulges initially provided structural stabilization for these helices (49).

Fig. 4C and Table 1 indicate that π-bulges in S6 are absent in some channels that possess asparagine at position 6.556. For example, all prokaryotic Nav channels exhibit entirely α-helical S6s across the four repeats (pattern αααα). In contrast, 24-TM TPCs feature π-bulges in nonadjacent S6 helices (παπα), the pattern also observed in most eukaryotic Nav and Cav structures. Different patterns can also emerge due to π-bulges induced by pore-bound ligands. These observations suggest that π-bulges may dynamically appear and disappear. A ligand interacting with residues facing the pore can alter its "native" orientation toward the pore lumen, thereby inducing a π-bulge. This reorientation of S6 residues significantly influences patterns of intersegment interactions. The dynamic nature of π-bulges in the inner helices may be linked to channel gating. A π-helical bulge could play a crucial role in the gating by disrupting the continuous α-helical H-bonding network, potentially causing bending or kinking of the helix and thereby opening the S6 bundle at the activation gate. This potential role of π-bulging in gating is particularly discussed for TRP channels (50). However, the relationships between π-helical bulging or deflating in the inner helices and the activation gating are not yet fully understood (3).

Table 1.

π-Helical bulges in S6 segments of Nav, Cav, and TRP channels and TPCs, which contain asparagine in position 6.556

Channel	Structure	Patterna
NavAb	3RVY, …	α α α α
NavMs	4OXS, …	α α α α
NaChBac	4DXW, 6VWX, 6VX3, 6VXO, 6W6O	α α α α
NavAe	4LTQ, 5HK7, 5IWN, 5IWO	α α α α
NavPaS	5X0M, 6A90, 6A91, 6A95	π π π π
NavAb-Nav1.7	5EK0, 7K48, 6N4I, 6N4Q, 6N4R	α α α α
NavPaS-Nav1.7	6NT3 6NT4	π π π π
Nav1.1	7DTD	π α π α
Nav1.2	6J8E	π α α α
Nav1.3	7W7F 7W77	π α π α
Nav1.4	6AGF	π α π α
	5XSY (EEL)	α α π α
Nav1.5	7DTC, 6LQA, 6UZ3,6UZ0	π α π α
	7FBS, 7K18	π α π α
Nav1.7	7W9P, 7W9M	π α π π
	7XMG	π π π α
	7W9K, 7W9L, 7W9T, 6J8H, 6J8G, 6J8J, 6J8I, 7XMF	π α π α
	7XM9	π α π π
	7XVE, 7XVF	π π π π
Nav1.8	7WE4, 7WEL, 7WFR, 7WFW	π α π α
CavAb	5KLB, ….	α α α α
Cav1.1	3JBR, 5GJV,5GJW, 6BYO	α α α α
	8E56,8E57, 8E58, 6JP8 6JPB 6JP5, 7JPX, 7JPW, 7JPK, 7JPL, 7JPV	π α π α
	6JPA	π α α α
Cav1.3	8E59, 8E5B, 7UHG, 8E5A	π α π α
	7UHF	π α α α
Cav2.2	7MIY, 7MIX	π α π α
Cav3.1	6KZP	π π α α
	6KZO	π α α α
TPC	6E1K, 6E1L, 6E1M, 6E1P	π α π α
	5DQQ, 5E1J	π π π π
	6C9A, 6C96	π α π α
TRPA1	3J9P	π π π π
TRPV1	3J5R	π π π π
	3J5P	π π π π
	5IRZ, 5IS0,	π π π π
	3J9J	π π π π
TRPV2	5AN8, 5HI9, 6BWM, 6BWJ, 6OO4, 6U88, 6U8A, 6U84, 6U86, 6OO3, 6OO5	α α α α
	6BO4, 6BO5	π π π π
	6OO	π α π α
TRPV3	6UW9, 6PVL, 6DVW, 6DVY, 6OT2, 6MHO, 6MHV, 6MHW, 6MHX	α α α α
	6UW4, 6UW6, 6UW8, 6LGP, 6OT5, 6PVP, 6PVN, 6PVO, 6PVM, 6DVZ,	π π π π
TRPV6	5IWK	α α α α
	6BO8, 6E2G, 6E2F	π π π π
TRPM2	6D73, 6PKV, 6PKW, 6PKX	α α α α
	6MIX, 6MIZ, 6MJ2, 6CO7	π π π π
	6DRK, 6DRJ	π π π π
TRPM4	6BQR, 6BQV, 6BWI, 5WP6	π π π π
	6BCO, 6BCQ, 6BCL, 6BCJ	π π π π
TRPM8	6NR2, 6NR4, 6O6A	α α α α

^aLetters "π" or "α" in position 1, 2, 3, and 4 designate the presence (π) or absence (α) of the π-bulge in S6 helices of repeat/subunit I, II, III, or IV, respectively.

Residue labels in helices S1–S4 of VSDs

The four VSDs surrounding the PD are connected to it via the S4-S5 linker helices. An intriguing feature, observable only in 3D-aligned multiple structures, is the variation in the relative orientation of the VSDs with respect to the PD in different subfamilies (Fig. 5). For example, in the structures 7CR0 (KCNQ2 in the apo state) and 6BO5 (TRPV2), the PDs are well aligned in 3D, but the VSDs are positioned entirely differently relative to the PD (Fig. 5 A). The primary structural difference accounting for the specific VSD-PD orientations lies in the hinges between the S4 helix and the S4-S5 linker helix. This intriguing difference between potassium and TRP channels was described previously (51). Fig. 5, B and C, illustrates subfamily specific arrangements of VSD helices relative to the PD in TRP, Nav, Cav, and Kv channels. The largest variations occur in helix S2, which is the most distant from the PD (Fig. 5 B). Despite these significant variations in orientation of VSDs, the overall folding of VSDs (i.e., the relative positioning of helices S1–S4) remains highly conserved, with only a few exceptions (1ORQ, 3JBR, 4DXW, 6E1P). This local structural conservation made it possible to identify reference positions in helices S1–S4 across the different channels.

Figure 5.

Relative disposition of VSDs and the pore domain. (A) The intracellular view of 3D-aligned TRPV2 (6BO5, green) and KCNQ2 (7CRO, red) demonstrates well-aligned PDs, while VSDs have completely different orientations. The enlargement highlights the distinct conformations of the hinge between helices S4 and S4-S5. (B) The CA atoms in helices S2 are shown as dots and colored according to the channel subfamily: Kv and KcsA (red), other K channels (magenta), Nav channels (blue), Cav channels (cyan), iGluR channels (yellow), TRP channels (green), and TPCs (orange). The CA tracing of the PD in Kv1.2–2.1 (2R9R) is shown in the center. A distinct, family-specific orientation of S2 helices, which are the most distant from the PD, is evident. (C) The distribution of sin(ψ), the angle defined in (B), is shown as mean ± SD. The value of sin(ψ) is significantly smaller for Kv channels compared to those of Nav, Cav, and TRP channels, indicating a family-specific difference in VSD orientation relative to the PD.

The selection of the reference position in helix S1 was somewhat challenging due to the low sequence similarity across different subfamilies (Table S4). Most of the channels have a hydrophobic residue at this position. For other VSD helices, we selected reference positions where channels have rather conserved residues that participate in strong intersegment contacts. In most VGICs, a glutamate is found in helix S2, which we designated as the reference position 2.550 (Table S5). For instance, in the activated VSD_I of the hNav1.5 channel, glutamate E^A2.550 forms salt bridges with lysines K^A4.559 and K^A2.554. In many other channels, this glutamate also participates in an intersegment salt bridge, reinforcing its importance across different channel families.

Most of the VGICs have an aspartate in helix S3, which we selected as the reference position 3.550 (Table S6). In cryo-EM structures of the hNav1.5 channel with activated VSDs, as well as in several other channels, aspartate D^3.550 forms salt bridges with basic residues in positions 4.559 and 4.562 of the sliding helix S4. For this helix, we selected arginine R^4.550 as the reference residue, which is the second basic residue from the extracellular side in most of VSDs (Table S7). The majority of Kv, Nav, and Cav channels have voltage-sensing basic residues in positions 4.547, 4.550, 4.553, 4.556, and 4.559. Comparing the structures of helices S1-S4 in P-loop channels, which were 3D aligned by minimizing root mean-square deviations of CA atoms in the P1 helices from the same-labeled atoms in the reference structure (2R9R), reveals that the CA–CB bonds of same-label residues in individual VSDs across each subfamily exhibit parallel orientation and close spatial disposition.

The universal labeling system enabled us to quantify the different orientations of the VSDs relative to the PD (Fig. 5 A). To do this, we measured the angle ψ between the Ox axis of the Cartesian coordinate system and atom CA^2.550 (Fig. 5 B). This analysis revealed a significant difference in orientation between Kv channels and other members of the PLIC superfamily (Fig. 5 C).

The PLIC-3D alignment facilitates the analysis of both large structural dissimilarities between P-loop channel subfamilies and detection of subtle variations within individual subfamilies. This is demonstrated by visualizing the CA–CB bonds at reference positions in the four VSDs of Nav channels (Fig. 6). While the folding of the pore domain (PD) is well conserved across these channels, the structures of the VSDs show less conservation. In experimental structures, S4 helices are captured in various sliding states. As a result, the positions and orientations of the CA–CB^4.560 bonds are notably scattered (Fig. 6, C and D).

Figure 6.

Diversity of VSDs in sodium channels. (A and B) Extracellular (A) and membrane (B) views of the four VSD backbones in the NavAb channel (3RVY). Helices S1, S2, S3, and S4 are colored red, green, yellow, and blue, respectively. (C and D) Enlarged views of VSDs I–IV. Sticks show the CA–CB bonds in positions 1.550, 2.550, 3.550, and 4.560 in multiple 3D-aligned Nav structures. In most structures, the CA–CB bonds of reference residues in helices S1, S2, and S3 tend to form one or more clusters, with outliers labeled by their respective PDB codes. The bonds CA–CB^4.560 in the voltage-sensing helix S4 (shown as blue sticks) are dispersed at the membrane view (D) but remain clustered in the extracellular view (C). This dispersion in the membrane view reflects the different sliding positions of helix S4 captured in the respective structures. In contrast, CA–CB^2.550 bonds are dispersed in both extracellular and membrane views.

Additional structures

Following the suggestion of an anonymous reviewer, we expanded the PLIC database by including cryo-EM structures of 35 additional channels. These include four Cav channels, 11 Kv channels, and 20 TRP channels. Although these structures were not involved in the initial analysis of 380 channels, they provide additional support for our main structural conclusions. Specifically, the additional data reinforce the findings on 1) the family-specific distances between P-loops and the pore axis, 2) the family-specific presence of π-bulges in the inner helices, and 3) the family-specific orientations of VSDs relative to the PD.

Discussion

In this study, we developed a universal residue-labeling scheme (PLIC labels) for the P-loop channel superfamily and an algorithm for their 3D alignment. In the 3D-aligned structures of highly diverse channels, the CA–CB bonds of the same-label residues generally have parallel orientations and similar spatial dispositions. This confirms the applicability of PLIC labels across the entire P-loop channels superfamily even though the sequence similarity between channels from different subfamilies may be poor. We have created the PLIC-3DA database: www.plic3da.com that currently contains over 400 3D-aligned experimental structures with PLIC-labeled residues. This database will support the analysis of structural conservation and variability within specific segments of P-loop channel subfamilies, as well as comparisons between subfamilies. We demonstrated the potential of the PLIC-3DA database by comparing various geometric features of the PD and VSDs. This analysis revealed previously unrecognized structural variations, including subtle but significant alterations in protein folding. Specifically, the comparison of multiple structures revealed family-specific arrangements of P-loops and S5 and S6 segments, as well as differing orientations of VSDs relative to the PD. These findings may provide insights into the unique functional characteristics of specific channels.

The PLIC labeling is based on 3D structural alignment rather than on sequence alignment. Within individual subfamilies, the PLIC sequence alignments do not contradict the conventional genetic-based sequence alignments obtained, e.g., with ClustalW (52) that maximizes identities and similarities of the aligned residues. However, the PLIC sequence alignments across different subfamilies, which are derived from structural alignments, do not always parallel genetic-based sequence alignments. For example, as shown in Table S8, the alignment of TM segments S1–S3 in TRPV and TRPC channels differs from that in TRPM channels. This discrepancy may arise from limitations in sequence alignment procedures or reflect different evolutionary origins of these segments in various TRP channels. The methodological difference underlying this discrepancy is that PLIC labels is designed to ensure close spatial positioning of same-label residues in 3D-aligned channels (structural analogy) rather than focusing solely on evolutionary homology.

Genetic-based sequence alignment of helices S4 and S5 in TRP channels is ambiguous because the S4 helices lack some basic residues that sense voltage in other P-loop channels. To address this, we selected reference positions in helices S4 and S5 of TRP channels (Fig. 7 A) to ensure parallel orientations and close proximity of the CA–CB bonds in same-label residues. To achieve a good 3D alignment of S5 helices in TRPV and TRPM channels (Fig. 7, B and C), we introduced an insertion/deletion at position 4.600 between helices S4 and S4-S5 (Fig. 7 A). We avoided an alternative insertion/deletion at position 5.530 (C-terminal to the sequences QK/QR) to ensure continuous PLIC labels within linker helices S4-S5. Other examples of variation between PLIC-based and genetics-based sequence alignments include insertions/deletions at position 4.600 in TRPV channels (Fig. 7 A) and the shift of motif TIGMGD in TRPV channels compared to the motif TVGYGD in potassium channels (Table S6). The importance of structure-based corrections in the sequence alignment across different channel families is mentioned, for instance, in the study describing the design principles of voltage-dependent gating (53). We expect that our work can facilitate such studies by providing a large structural dataset for comparative analysis.

Figure 7.

PLIC labels in helices S4 and S5 of TRP channels. (A) PLIC-based sequence alignment. To ensure identity/similarity of residues in matched positions of the alignment and parallel orientations of the same-label bonds CA–CB in helices S4 and S5, an insertion/deletion is introduced in position 4.600 between helices S4 and S4-S5. (B) Extracellular view of the 6UW9 structure with helices S4, S4-S5, S5, and P1 highlighted in one subunit. (C) Membrane view at helices S4, S4-S5, S5, and P1 in 3D-aligned TRP channels 3J9P, 6CO7, 5WP6, 3J9J, 6MHO, and 6BO8. Spherical CA atoms are shown in same-label positions.

The PLIC-3DA database opens a possibility to utilize high-throughput bioinformatics tools for analyzing disease mutations. Many thousands of channelopathy mutations in Nav, Cav, Kv, and other P-loop channels are reported in the ClinVar database (54) and other public databases. However, the clinical significance of many reported mutations remains uncertain. This presents a serious medical problem: genotyping of a patient may reveal a mutation in a channel, but whether this mutation causes the disease is often unknown, and appropriate treatment remains unclear. The application of various bioinformatics tools, including the paralog annotation method (55), has helped reclassify mutations of unknown clinical significance as pathogenic or likely pathogenic for the Nav1.5, Cav1.2, and TRPM4 channels (56,57,58). The PLIC-3DA database can facilitate similar analyses for other medically relevant P-loop channels. The PLIC-3DA database would aid in the detection of correlated mutations (59,60), which may advance our understanding of the general physiology and pathology of ion channels. In pharmacology and toxicology, comparative structural analysis using PLIC labels of 3D-aligned channels could help explain the varying sensitivities of related channels to drugs and toxins, thereby aiding in the development of selective drugs targeting specific subtypes of P-loop channels. The PLIC-3DA database is continually expanding, and we welcome suggestions of new structures to include. Furthermore, the database has a utility that allows users to renumber and 3D-align new structures of P-loop channels.

Data and code availability

We compiled a PLIC-3DA database: www.plic3da.com with hundreds of relabeled and 3D-aligned structures of P-loop channels. The structures are free to download. The database also contains file PLIC.xlsx with multiple sequence alignment of representative P-loop channels (Front page/Download/link)

Acknowledgments

This work was supported by grant no. 075-15-2024-548 from the Ministry of Science and Higher Education of the Russian Federation. B.S.Z. acknowledges support from the Natural Sciences and Engineering Research Council of Canada (RGPIN-2020-07100).

Author contributions

Conceptualization, D.B.T. and B.S.Z.; investigation, D.B.T. and B.S.Z.; methodology, D.B.T. and B.S.Z.; collection and 3D alignments of the channel structures, D.B.T. and B.S.Z.; data curation, D.B.T. and B.S.Z.; formal analysis, D.B.T. and B.S.Z.; validation, D.B.T. and B.S.Z.; visualization, D.B.T. and B.S.Z.; writing – original draft, D.B.T. and B.S.Z.; writing – review & editing, D.B.T. and B.S.Z.; creation of the PLIC database, V.S.K.

Declaration of interests

The authors declare no competing interests.

Editor: Carlos Villalba-Galea.