Ion permeation through a narrow cavity constriction in KCNQ1 channels: Mechanism and implications for pathogenic variants Authors: Liana Mkrtchyan, Harutyun Sahakyan, Jodene Eldstrom, Tatev Karapetyan, Astghik Abrahamyan, Karen Nazaryan, Jürgen R. Schwarz, Matthias Kneussel, David Fedida, Vitya Vardanyan Published: 2024 ABSTRACT SignificanceThe latest atomic resolution structures of KCNQ channels feature a narrow cavity region not sufficiently wide to accommodate fully hydrated K+ ions. The mechanism of K+ permeation through this constriction and the functional impact of the narrow cavity region remains unknown. Here, we demonstrate that the partial dehydration of K+ near the constriction allows permeation of ions. Our findings indicate that constriction hinders the ion conduction of KCNQ1 channels and provides a plausible rationale to explain clinically relevant loss-of-function IKs mutations in patients with long QT syndrome. The KCNQ1 protein, also known as Kv7.1 or KvLQT1, is a voltage-activated potassium channel expressed in several human excitable and epithelial tissues. Numerous mutations in KCNQ1 have been associated with diseases of the human heart (1, 2), inner ear (3, 4), and endocrine system (5, 6). For the development of pharmacological tools against human pathologies associated with KCNQ1 dysfunction, understanding the biophysical and pharmacological properties of the channel, its subunit composition, and mechanisms of its regulation is essential. KCNQ1 channels are regulated by various cellular factors including intracellular free Ca2+ (7), extracellular K+ (8) adenosine nucleotides (9), membrane lipids (10, 11), and protein phosphatases (12). These features, along with the regulatory impact of auxiliary subunits, enable KCNQ1 channels to fulfill versatile physiological tasks in various cell types (13, 14). KCNQ1 is known to associate with KCNE1-5 β subunits to build heteromeric channel complexes (15, 16). KCNE subunits are single-transmembrane domain proteins that play a crucial role in regulating the gating, permeability, and pharmacological characteristics of KCNQ1. The scale of KCNQ1 modulation is dependent on the specific type (17) and α to β ratio of the associated KCNE subunits (18). Coexpression of KCNE1 with KCNQ1, for instance, results in a noticeable shift of the conductance–voltage curve of the channel toward positive potentials (15), causes a severe deceleration of the channel’s activation and deactivation kinetics (19), eliminates fast inactivation (20), and increases unitary conductance of the channel (21). In the human heart, KCNQ1/KCNE1 complexes mediate the slowly activating delayed-rectifier IKs current strictly required for heart rhythm regulation. It is no wonder that many KCNQ1 mutations have been identified in patients suffering from severe heart rhythm abnormalities. Mutations in KCNQ1 have been also linked to congenital hearing loss in humans (22), as these channels play a critical role in K+ recycling in the inner ear and the generation of endocochlear potential (23). KCNE3 renders the KCNQ1 constitutively open at the physiological range of potentials (16) making the channel an efficient regulator of water and electrolyte balance in nonexcitable tissues of the gastrointestinal tract, kidney, and airways (24, 25). Recent cryo-EM studies have uncovered insights into the structural arrangement of KCNQ1 channels, providing a comprehensive explanation for the known functional and pharmacological traits of channels at the atomic scale (26–30). In these structures, KCNQ1 adopts a domain-swapped tetrameric conformation characteristic of voltage-dependent K+ channels with a central pore surrounded by four peripheral voltage-sensor domains (26–30). KCNE3 auxiliary subunits bind in the cleft formed between the pore domain and voltage sensor of KCNQ1 suggesting a similar orientation for other KCNE subunits in KCNQ1/KCNE complexes (28). The cellular Ca2+-sensor protein calmodulin (CaM), previously shown to be an integral part of the KCNQ1 channel complex (31–33), undergoes significant rearrangements in conformation upon channel gating (26, 28). Remarkably, the high-resolution structures of KCNQ1 captured in the presence of KCNE3, the activator ML277, and PIP2 molecules, which render the channel constitutively open in functional studies (16, 27), are thought to represent the open state of the channel and feature a 4.5 to 5.1 Å wide constricted region in the cavity (Fig. 1 and SI Appendix, Table S1). This raises important questions as i) whether the open channel conformation is correctly reflected in the recent cryo-EM structures, and if so, ii) how this region allows for the passage of hydrated K+ with an average diameter of ~6.8 Å (34). Sun and MacKinnon (28) propose two possibilities: i) the pore may dilate more than observed in static cryo-EM structures and ii) the flexibility of side chains of the residues adjacent to the constricted region may allow further widening of the aperture during the dynamic ion conduction process. The latest studies on inward rectifier K+ channels (35–37) point to the potential role of partial dehydration of ions as a third alternative for ion permeation through the KCNQ1 cavity constriction. To understand the process of K+ permeation through the constriction of the KCNQ1 cavity and address several lingering questions, we utilized molecular dynamic (MD) simulation based on the KCNQ1/KCNE3 cryo-EM structure, conducted biophysical and biochemical analysis of various mutants at the G345 residue situated at the narrowest part of the cavity, and measured the unitary conductance of IKs channels carrying G345 mutations. Cavity constriction in open cryo-EM structures of the KCNQ1 channel. (A) Putative open conformations of KCNQ1 observed in single-particle cryo-EM structures i) when in association with KCNE3 subunit (shown in green with red KCNE3) (28), ii) upon binding of channel activator ML277 (26) in the presence (colored cyan and blue) and absence of PIP2 molecule (magenta). ML277 is indicated by red in these structures. Calmodulin molecules are shown in yellow. For clarity of view, only two subunits are shown. Below the structures, the corresponding constricted regions are shown. The differences in structure of the cytoplasmic domain reflect two different binding configurations of Calmodulin, referred to as “attached” and “detached” (26). (B) Overlay of the pore regions for structures shown in (A) with highlighted residue G345 located at the narrowest point of the cavity, as well as residues S349 and L353 forming the cytoplasmic gate. (C) Cavity dimensions and constriction width near the G345 region are almost the same for the structures depicted in (A). Location of residues forming the gate region is also shown. (D) The cavity dimensions of KCNQ1/KCNE3 are compared to the size of the fully hydrated K+. HOLE software was used to generate cavity contours in (C and D). The green line represents the boundary of the solvent-accessible surfaces based on van der Waals’ atomic radii. Results KCNQ1 Pore with Narrow Cavity Is Conductive. To test whether the KCNQ1 conformation believed to represent an open channel is capable of conducting K+, we employed the Computational Electrophysiology (CompEl) method of MD simulation that was successfully used in recent years to describe the ion permeation process through prototypic K+ channels (38, 39). We constructed a similar molecular system based on the cryo-EM structure of KCNQ1/KCNE3 heteromers (PDB ID: 6V01) as schematically illustrated in SI Appendix, Fig. S1A. Briefly, the pore region of KCNQ1 corresponding to amino acids G245-V354 was embedded into membrane bilayers composed of POPC lipids in a parallel orientation. The addition of distinct Na+, K+, and Cl− ions to the water-filled compartments between membranes (SI Appendix, Supplementary Methods) led to a corresponding extracellular K+ concentration of 5 mM and an intracellular K+ concentration of 150 mM. A transmembrane potential of approximately 300 mV was created in this system due to an imbalance of 2 Cl− ions between compartments. This potential acts as a driving force for the transfer of ions from lower compartment A to B, imitating the intracellular and extracellular environments respectively (SI Appendix, Fig. S1 A, Right). The selection of this system aimed to replicate the experimental set-up and simulation conditions as specified for prototypic K+ channels in earlier studies (40). The KCNE3 subunit was removed in order to ensure the utmost simplicity of the molecular system enabling us to perform microsecond-scale MD simulations. SI Appendix, Supplementary Methods and our recent study give an in-depth description of the methodology (41). After system equilibration, we proceeded with an MD simulation, where the terminal amino acids in the lower pore region were not constrained. This was done to test the hypothesis that the narrow region of the cavity could expand further allowing ion conduction (28). Under these experimental conditions, the channel pore quickly (<10 ns) adopted a conformation that was consistent with the closed state of the channel observed in recent cryo-EM studies (SI Appendix, Fig. S2). In particular, the lower regions of S6 helixes from all four subunits converged to form a hydrophobic seal at the pore’s cytoplasmic side involving the side chains of S349 and L353 residues (Movie S1) (26–28). Throughout three individual simulations, each lasting 320 ns, we did not observe any entrance of K+ ions from the cytoplasmic compartment into the channel cavity. By implementing position restraints on the backbone of amino acids G245-T247 in S5 and A352-K354 in S6 to keep the cytoplasmatic gate in an open conformation, we observed 126 permeation events during 10 independent 500 ns simulations (SI Appendix, Fig. S1 B and C and Table S2). The movement of K+ in the selectivity filter of KCNQ1 occurs via a direct Knock-on mechanism (38). A value of 14.5 ± 0.9 pS (median ± SEM) for channel conductance at 300 mV, noted in MD simulations, was fivefold higher than the cord conductivity of KCNQ1/KCNE1 (~2.9 pS). This discrepancy is attributable to the use of high voltages in MD simulations and truncation of the channel to a pore-only structure. Even so, the ratios of conductance between mutants and the wild type observed in MD simulations are in good agreement with the results from electrophysiological measurements discussed below. Taken together, the data confirm that the KCNQ1 pore structures shown in Fig. 1A, featuring a narrow constriction in their cavity, represent the open channel conformation and effectively allow ion permeation. The Narrow Cavity Opening Maintains Smaller Dimensions Relative to Fully Hydrated K+. To understand whether the constricted cavity widened during MD simulations allows ion permeation, we analyzed the internuclear distance between G345Cα – G345Cα atoms of opposing subunits denoted as d1 and d2 in this study. G345Cα corresponds to the narrowest point of the cavity in the open cryo-EM structures (Fig. 1), with interatomic distances between the opposite subunits of about 8 Å (Fig. 2A and SI Appendix, Table S1). This corresponds to solvent-accessible distance of 4.6 Å between subunits at the constriction point (SI Appendix, Table S1). After the insertion of the channel pore into the membrane and subsequent system equilibration, the d2 distance increased by ~1.6 Å, whereas d1 was only slightly (~0.2 Å) changed (Fig. 2B). This implies that the cavity opening remained approximately the same diameter as observed in the cryo-EM structures—meaning about 2 Å smaller than the effective size of the fully hydrated K+ of ~6.8 Å (34). Analysis of the distribution of d1 and d2 during MD simulations indicated that in the vast majority of the frames, the cavity opening did not change considerably throughout entire simulations (Fig. 2C). To understand whether K+ permeations occurred due to short temporary enlargement of the constriction, we extracted frames in which permeating K+ ions were localized closest to the narrowest point of the cavity (Fig. 2D). Then, d1-d2 values for these frames were overlaid onto a hexagonal plot representing d1-d2 pairs of entire simulations (Fig. 2E). Results demonstrate that the width of the constriction was always smaller than the size of fully hydrated K+ when ions permeated through the constriction. Thus, the minor asymmetrical enlargement of the cavity at the constriction and the dynamic changes during the ion conduction process could not explain the permeation of K+ through the constriction. The dimensions of the constricted cavity as ions pass through. (A) Internuclear distance between alpha carbonyls of G345 of opposite subunits (d1 and d2) observed in the cryo-EM structure of KCNQ1/KCNE3 (28). (B) After the KCNQ1 pore was embedded in POPC lipids and the simulation system was equilibrated, there was a slight asymmetrical increase in dimensions. (C) The distribution of d1 and d2 values across all simulation frames. The data were fitted to Gaussian functions shown as solid lines. Parameters of fit were as follows: d1; A = 1,764; Xo = 8.07; width = 0.7; Yo = 8.7 and for d2: A = 1,902; Xo = 9.76; width = 0.65; Yo = 8.3. Dashed lines indicate the position of the values observed in the initial frame of the simulation. (D) A snapshot from simulations showing the permeating K+ closest to the narrowest point of the constriction. G345 atoms are shown as sticks, the K+ ion is represented as a pink sphere. (E) The d1-d2 values (shown as red circles) for frames corresponding to the position of permeating potassium ions as indicated in (D). In the background, the hexagonal plot displays the values of d1-d2 pairs for all simulation frames. Dashed lines indicate the boundaries of the cavity constriction that would allow the fully hydrated K+ with a diameter of 6.8 Å to pass. The red rectangle highlights these limited frames. The secondary axes represent the corresponding solvent-accessible spaces between alpha carbon atoms of G345 in Angstroms. K+ Ions Pass the Cavity Constriction in a Partially Hydrated State. In order to check the third potential scenario, which assumes ions lose water molecules of their hydration shell before going through the constriction, we assessed the hydrated state of K+ ions at the narrowest point of constriction—illustrated in Fig. 2D. The hydration of K+ was analyzed by counting the water oxygen atoms within a 3 Å center-to-center distance from the K+ ions. At the narrowest point of the cavity, the hydration for ions was decreased compared to the water compartment (Fig. 3A and Movie S2). According to the hydration diagram in Fig. 3A, about 40% of cases showed permeating ions with 5 water molecules, and 15% of cases showed ions with 6 water molecules at the narrowest point of the cavity. Inspection of trajectories, nevertheless, indicated that one or two of these molecules originate from the cavity compartment adjacent to the selectivity filter (Movie S3). In particular, ions reach the constricted area in a more dehydrated state than shown in Fig. 3A, after which water molecules from the upper cavity compartment—adjacent to the selectivity filter—replenish their hydration shell. Remarkably, a larger number of ions that entered the narrow area of the cavity did not pass through the constriction and instead moved back toward the cytoplasm (Movie S4). These ions showed a markedly higher level of hydration at the peak of their trajectory (highest Z coordinate) toward the selectivity filter (Fig. 3A). To get an idea of the proportion of ions that did not make it through the constriction, we analyzed the ratio of the permeated and repelled ions from a cylindrical volume placed at the constricted region with dimensions shown in Fig. 3C. The dimensions of this volume were selected based on the dynamics of G345Cα atoms of all four subunits (Fig. 3B). This revealed that a significant fraction, more than 80%, of ions that reached this cylindrical space did not cross the constriction (Fig. 3D). In summary, the findings indicated that i) ions crossed the cavity constriction in a partly hydrated state and ii) the constriction may impede the conduction of ions. The partial dehydration of K+ at the narrowest point of the cavity constriction. (A) Count of the water molecules within 3 Å distance of the K+ ions at the narrowest point of the cavity for permeating ions (K+- passed) compared to those located in the cytoplasmatic compartment (at Z coordinate of −40 Å relative to COM SF). Hydration of ions that reached the highest point of their trajectory in the constricted region but did not traverse the constriction is shown as “K+-repelled”. Results represent n = 119 passing ions as well as equal number of ions returned back from the virtual cylindrical contour shown in (C), randomly selected from all simulations. (B) The range of Z coordinate changes of G345Cα atoms for four subunits. Values correspond to frames where the permeating K+ ions were at the narrowest point of the cavity as shown in Fig. 2D). (C) The definition of the cylindrical volume with Z boundaries for calculation of the proportion of ions that passed. An ion was considered as inside the cylinder based on the location of its atomic center. (D) The percentage of ions that passed the constriction compared to those that were expelled from the constricted region toward the cytoplasmic compartment. Each point represents a separate simulation. Rectangles represent the average of each dataset. Transient Interaction of K+ with Carbonyl Oxygens of G345 Residues. We hypothesized that protein oxygen atoms could potentially substitute for water oxygens associated with permeating K+ ions, thereby facilitating the passage of ions through the narrow constriction. The cavity of channels, and likely the environment of the constricted regions, contains exposed backbone carbonyl oxygens and polar/negatively charged side chains of amino acids that may interact with moving K+ ions substituting the water oxygen(s) of their hydration shell. In contrast, in the presence of positively charged elements or a hydrophobic environment, ion–protein interactions may be impeded and the ligand exchange might be unfavorable. The position of carbonyl oxygen atoms of G345 (G345 CO) immediately below the constricted area led us to consider a potential scenario where the dehydration of K+ is induced by electrostatic interactions between the ion and G345 CO. Examination of the simulation trajectories revealed that passing K+ ions transiently enter into electrostatic interactions with one or more carboxyl oxygens before crossing the constriction (Fig. 4 A and C and Movie S5). We used the internuclear distance between the ion and the G345 CO less than or equal to 3 Å as a criterion for the existence of electrostatic interactions, which revealed that 79% of permeating ions interact with the carbonyl oxygens of G345 (94 ions out of 119) before they crossed the constriction. As we observed that most interactions lasted between 20 and 540 ps (140 ± 60 ps, mean ± SEM), we suspected that interactions lasting less than 20 ps may not have been captured due to the 20 ps/frame resolution of the simulations. Therefore, we conducted 4 additional simulations with a 4 ps/frame trajectory interval, which showed that every permeating ion (n = 54, SI Appendix, Table S2) transiently interacts electrostatically with G345 CO of one or more subunits prior to its translocation across the constriction (Fig. 4C). Interaction of K+ with carbonyl oxygens led to a marked decrease in the number of water molecules within a 3 Å distance from K+ (Fig. 4B). This indicates that carbonyl oxygens of G345 mimic the water oxygens of the hydration shell of K+, which increases the probability of partial dehydration of K+ near the constriction. Nevertheless, in line with our findings that most ions did not pass through the narrow region, we noted a significantly higher number of K+–G345 CO interactions without subsequent ion permeation (Fig. 4 C, Lower). Thus, while the interaction between K+ and G345 CO is important in decreasing the hydration level of ions, it may not be the sole factor responsible for permeation through the constriction. Electrostatic interaction of K+ with the carbonyl oxygen of G345 may increase the probability of partial dehydration of ions near the constricted area. (A) A snapshot from simulations showing electrostatic interaction of K+ with G345 CO. K+ is shown as a pink sphere. The water molecules are depicted as sticks in red and white. Only two opposite subunits (cyan helices) are shown for clarity of the view. (B) When interacting with G345 CO, the K+ hydration level is shown in comparison to both the cytoplasmic compartment (Comp. A) and when located at the narrowest point of constriction. The quantification of total oxygen atoms within 3 Å distance of K+ is also displayed for comparison. (C) The Top diagram displays the Z coordinate of permeating K+ ions, each uniquely colored to highlight the electrostatic interactions preceding the permeation (compare to the Lower panel). For the purpose of clarity, the trajectories of ions that did not permeate the channel are omitted. In the Lower panel, the K+–G345 CO interactions are displayed based on a 3 Å center-to-center distance criterion. Note that ions can briefly interact with oxygen from one or multiple subunits, as denoted by four different symbols on the Right. Modulation of KCNQ1 Currents through G345 Mutations. We next aimed to incorporate side chains of varying sizes and hydropathic properties at the G345 position in order to change the dimensions and chemical characteristics of the cavity constriction and explore how K+ permeation relates to the features of the constriction at the functional level. Through Ala, Val, Leu, and Trp mutations, we gradually enhanced the side chain volume while preserving its hydrophobic nature, while the polar (Ser, Thr, Tyr, Asn, and Gln) or charged (Asp, Glu, Lys, and Arg) replacements (Fig. 5A) allowed us to test how similar-sized side chains with distinct chemistry influence K+ permeability. It is important to note that the G345 residue is part of the PAG motif of KCNQ channels that plays an imperative role in channel gating (42). Specifically, the PAG motif acts as a gating hinge for sideward movement of the C-terminal half of the S6 helixes, which underlies the opening and closure of the cytoplasmatic gate (26, 28). For this reason, we anticipated that mutations will not only impact the properties of the constriction but also may significantly modify the channel’s gating features. Indeed, the majority of homomeric G345 mutants markedly altered the steady-state gating (SI Appendix, Figs. S3–S6 and Table S3) and conferred macroscopic inactivation to the channel (SI Appendix, Fig. S3A), whereas the positively charged side chains of Lys or Arg at G345 position resulted in nonselective channels (SI Appendix, Fig. S6). To eliminate or reduce the mutational effect on gating, we coexpressed these mutants with the KCNE3 auxiliary subunit, which locks KCNQ1 in an open state (16) at physiological voltages. These locked-open channels mimic MD simulation experiments in which restraint was applied to the gate region. KCNQ1/KCNE3 channels carrying G345 mutations showed marked reductions of current when compared to the wild type (Fig. 5 A and B). A comparable decrease in current was seen in homomeric channels as well (SI Appendix, Fig. S3), apart from G345N and G345Q mutants. The current amplitude of several KCNQ1/KCNE3 mutants exhibited little difference to oocytes injected with KCNE3 mRNA alone used as control (Fig. 5 A and B). Nonetheless, a difference in expression was evident when we measured the membrane potentials of these oocytes in the 8 to 14 d after injection and compared to those of water-injected oocytes (Fig. 5C) indicating that these channels are conductive as well. Analysis of the membrane protein levels via surface biotinylation and western blot indicated that membrane expression was largely unaffected by the G345 mutations (Fig. 5 D and E), indicating that mutations alter conduction of the channel. Nevertheless, the observed current deficit of the G345 mutants could be attributable, at least in part, to the macroscopic inactivation of channels at positive potentials, which was more pronounced in G345L, G345N, G345Q, G345K, and G345R. Taking advantage of previous findings demonstrating the effectiveness of KCNE1 coexpression in preventing the macroscopic inactivation in the KCNQ1-L279F mutant (43), we next studied the mutant KCNQ1/KCNE1 heteromers. With the exception of the G345K and G345R channels, no macroscopic inactivation was detected in these mutants (SI Appendix, Fig. S7A). We observed a noticeable decrease in current for the KCNQ1/KCNE1 heteromeric mutants except for KCNQ1-G345N/KCNE1, which exhibited a current level comparable to the wild type (SI Appendix, Fig. S7 A and C). The substitution of the amide group with a carboxyl group in the G345D and G345E mutants resulted in a significant reduction in current amplitude, indicating the importance of the chemical properties of the side chains (SI Appendix, Fig. S7). The presence of large hydrophobic side chains at G345 resulted in a marked decrease in current, even lower than those mediated by channels formed by endogenous Xenopus KCNQ1 and exogenously expressed KCNE1 (44), that we used as a control (SI Appendix, Fig. S7 A and C). A similar reduction has been observed in heteromeric mutants with negatively charged side chains at the G345 position. To confirm the dominant negative influence of these mutations, we coinjected oocytes with equal amounts of the wild type and mutant mRNA, along with KCNE1 auxiliary subunits in a 1:1:2 molecular ratio. The results clearly showed a dominant-negative influence of mutants on wild-type channels (SI Appendix, Fig. S8). We also observed alterations in the steady-state activation curves for some mutants in relation to the wild type (SI Appendix, Fig. S7 and Table S3). In summary, most G345 mutants showed a considerable decrease in current compared to the wild-type forms, regardless of whether they were in their voltage-gated or locked-open states. This decrease cannot be merely attributed to changes in steady-state gating, macroscopic inactivation, or protein expression and trafficking (SI Appendix, Fig. S7 C and D). The absence of a mutational effect at the G345 position on protein expression and trafficking is in accordance with earlier analyses of G345R (45) and G345A channels (42). Thus, results point to the mutational influence on the channel’s conductive properties, corroborating our MD simulation findings (Fig. 3D). We next carried out a single-channel conductance measurement on the G345T and G345N mutants coexpressed with KCNE1 in mouse ltk- fibroblast cells to check this hypothesis. Current reduction in KCNQ1/KCNE3 channels carrying mutation at the G345 position. (A) Sample current traces of WT and G345-KCNQ1/KCNE3 mutant channels in response to step-depolarizing pulses as indicated. The fold change of side-chain volume compared to Alanine at 345 position as well as the principal chemical groups are shown below the traces. Note the 10-fold amplitude scale for the wild type compared to mutants. (B) Relative change in current amplitudes measured at +60 mV and normalized (black circles) to the mean of the WT. G345V, G345S, G345Y, G345D, and G345E channels were recorded in the time period between 8 and 14 d following injection, and statistical analysis refers to water(E3)-injected oocytes measured on the same day. No significant difference in current amplitude was observed between these mutants and oocytes injected with water+E3 within the standard 3 to 7 d postinjection measurements. The number of individual cells measured for mutants is displayed in parentheses. Student’s t test was used to evaluate statistical significance compared to water(E3)-injected oocytes. (C) The membrane potential of cells injected with mutant mRNA as indicated. Potentials were measured in the time period between 8 and 14 d. Student’s t test was used to evaluate statistical significance; *P < 0.05, **P < 0.01; ***P < 0.001. ****P < 0.0001. (D) Sample western blot showing expression of mutant KCNQ1/KCNE3 channels on the cell surface. The Na-K ATPase was used as a membrane reference. β-Actin staining indicates the purity of the membrane protein fraction. (E) Quantification of the KCNQ1/KCNE3 proteins level relative to Na-K ATPase from three independent experiments. The Reduction of the Unitary Conductance in G345T and G345N Mutants. KCNQ1-345T/KCNE1 was selected for single-channel measurements due to its steady-state activation being the most similar to the wild type (SI Appendix, Fig. S7 and Table S3) and its activation kinetics being almost identical to the wild type (SI Appendix, Fig. S9). Furthermore, KCNQ1-345N/KCNE1 was included in this set of experiments to examine the reasons for the higher current amplitude relative to other mutants, whether in its homomeric form or in association with the KCNE1 subunit (SI Appendix, Figs. S3 and S7). We measured single KCNQ1/KCNE1 heteromeric tandem channels with a fixed 4:4 of α to β stoichiometry termed EQ-G345T and EQ-G345N alongside the corresponding wild type (EQ), as previously detailed (18). Typical single-channel records are shown in Fig. 6A. The data analysis revealed a substantial decrease in unitary conductance for both EQ-G345T and EQ-G345N channels compared to the wild type (Fig. 6 B and C). The G345T mutation leads to a 10-fold decrease in conductance, while the EQ-G345N variant shows a fourfold reduction. It is important to acknowledge that the exact quantification is hampered by uncertainties regarding the precision of localizing the peak in Gaussian fits due to the small amplitudes of openings (Fig. 6 A and B). The records, however, clearly showed delayed openings in mutants (Fig. 6A) which corresponds to the similar delay in the wild type described previously (21). This was in line with comparable activation gating observed in macroscopic recordings of KCNQ1-G345T/KCNE1 heteromers in oocytes (SI Appendix, Figs. S7 and S9). We aimed also to record the single EQ-G345Y heteromers, but despite considerable efforts, no discernable openings could be detected. It is likely that the current flowing through a single G345Y mutant channel, as well as other mutants with similar amplitudes, falls below the resolution capabilities of our recording system, taking into consideration the almost 10-fold decrease of macroscopic current in KCNQ1-345Y/KCNE1 relative to the threonine mutant (SI Appendix, Fig. S7C). In summary, the G345T and G345N mutations lead to attenuation of unitary conductance of KCNQ1/KCNE1 heteromers likely attributable to the mutation-induced narrowing of the cavity constriction. The greater conduction value of G345N in comparison to G345T is in line with whole-cell recordings indicating a larger current, a shift in steady-state gating, and a constitutively open component (SI Appendix, Fig. S7 and Table S3). Narrowing of the constriction leads to reduction of the unitary conductance. (A) Sample single-channel records of WT-EQ, as well as EQ-G345T and EQ-G345N mutant channels. Open and closed levels indicated correspond to peak of the Gaussian fits shown in (B). (B) Amplitude-histograms corresponding to +60 mV of the recordings shown in (A). Curves correspond to fit of the Gaussian function to histogram with two separate components shown with green and blue curves for G345T. (C) Current-voltage relationship of single KCNQ1-G345T/KCNE1 and KCNQ1-G345N/KCNE1 mutant channels along with WT. Slope-conductance values are—EQ-WT: 2.90 pS; EQ-G345T: 0.30 pS, EQ-G345N: 0.77 pS, X-intercepts, EQ-WT: −91.2 mV; EQ-G345T: − 73.5 mV, EQ-G345N: −82.3 mV. (D) The reduction in permeation events through the constricted region for the G345T mutant compared to WT observed in MD simulations. Empty circles on the graph show permeations per 0.5 µs; the average is shown by filled rectangles with error bars representing ±SEM. (E) Exchange of water oxygens with Threonine’s hydroxyl oxygens for K+ positioned at the constriction of G345T. (F) In a simulation snapshot, the K+ (pink sphere) at the constricted region of the G345T mutant is coordinated by the hydroxyl oxygens of the threonine side chains. Two water molecules are visible within a 3 Å proximity to the K+, shown as red-white sticks. The threonine side chains of four subunits are depicted as sticks. This structure was detected in 3.4% of frames during 5 µs simulations in total. (G) Configuration of the narrow constriction similar to that shown in (F) observed in 0.11% of frames, where one of the threonines has flipped orientation. To explore further how the narrowing of the cavity constriction by the G345T mutation leads to diminution of conductance, we introduced this mutation into the pore in our molecular model, allowed the systems to equilibrate, and performed MD simulations. After the equilibration, the constriction opening decreased significantly relative to the wild type reaching a dimension that is comparable to the ionic diameter of K+ (SI Appendix, Fig. S10A). During five independent MD simulations, each lasting 1 µs, the opening width remained relatively unaffected (SI Appendix, Fig. S10D) except in 11% of the frames where the alignment of the threonine methyl groups led to a reduction in the constriction opening that fell below the ionic size of K+. We observed overall 8 permeation events through the narrow cavity region (Fig. 6D and SI Appendix, Table S2), which is equivalent to 7.1% of the permeations through the wild type corrected for the 11% of frames that exhibited an occluded constriction. The observed decrease in permeations reflected the reduced current amplitudes of the locked-open G345T channel observed in electrophysiological recordings (Fig. 5B). We observed significant narrowing of the constriction in the 345T mutants as evidenced by a marked shift of the corresponding hexagonal plot compared to the wild type (SI Appendix, Fig. S10D). In order to pass through the cavity constriction, ions exchanged water oxygens in their hydration shell for hydroxyl oxygen atoms of the threonine (Fig. 6E). The coordination of the ion by these oxygens at the narrowest point of the cavity resembles the way K+ is coordinated in the selectivity filter of K+ channels (Fig. 6F). In a small fraction of frames (0.11%), we observed a quite similar K+ coordination with one of the threonine oxygens flipped downward (Fig. 6G). In addition, two water molecules were bound to ions at the constriction on average as shown in Fig. 6 F and G. Analysis reveals that the water molecule in the ion’s hydration shell enters the constricted area first, followed by the ion, which is coordinated by the hydroxyl oxygens for a certain period of time before moving toward the selectivity filter (Movie S6). Taken together, the G345T mutation leads to a narrowing of the constriction comparable to the ionic diameter of K+, which results in a significant reduction in unitary conductance of the channel. Clinically Relevant G345 Variants. G345A, G345E, G345R, and G345V missense variants have been identified in patients suffering from Long QT syndrome (SI Appendix, Table S4). Out of the 66 individuals identified with these specific mutations till now, 50 were reported to have moderate to severe heart rhythm abnormalities commonly associated with congenital Long QT syndrome with sudden cardiac arrest and even death (SI Appendix, Table S4). It is noteworthy that the G345E missense variant was one of the first among clinical mutations identified during the positional cloning of the KCNQ1 gene almost three decades ago (1). It has been shown that G345E exerts a dominant negative effect on wild-type IKs channels (46), similarly to G345R variant, for which no changes in the membrane fraction of protein were shown in transfected CHO cells (45). The G345A homomeric mutant’s surface expression was found to be unaffected as well (42) but functional analysis of corresponding IKs channels has not yet been conducted. Thus, we next investigated the IKs channels that mimic the situation in patients by coexpression of the wild type with the G345 mutants together with KCNE1 in a 1:1:2 mRNA molecular ratio. The results demonstrate a prominent decrease in current amplitudes among all mutants (Fig. 7 A and C), which is the most likely reason for the manifestation of pathological conditions in patients. Suppression of Iks current by coexpression with G345 clinical variants. (A) Examples of recorded current traces from WT and mutant channels, coexpressed with the KCNE1 auxiliary subunit. Shown on the Right is a diagram of the applied voltage protocol. (B) The normalized conductance–voltage curves for the channels shown in (A). The data were fitted to a Boltzmann function with an offset, as represented by the lines. Fit parameters are shown in the SI Appendix, Table S3. (C) The current amplitude of IKs channels containing G345 variants normalized to the mean of the WT. The number of individual cells recorded is shown in parentheses. (D) A reduction of ion permeations through WT/345A heteromeric channels when compared to the WT. Each point (empty circle) represents the number of permeations per 0.5 µs of consequent simulation period. n = 14 for WT; n = 10 for AAGG and AGAG. (E) Comparison of K+ hydration levels at constriction between the wild type and WT/345A heteromers by counting water oxygens within 3 Å of permeating ions. The histogram that represents the total count of oxygen atoms is almost indistinguishable and has been omitted. (F) Current amplitudes mediated by KCNQ1/KCNE3 wild type and KCNQ1/345A/KCNE3 heteromers normalized to the mean of the WT. The number of cells recorded for each channel is indicated in parentheses. Student’s unpaired t test was applied to calculate the statistical significance with P values as follows: *P < 0.01; ***P < 0.001; ****P < 0.0001. Heteromeric IKs channels carrying G345A and G345E mutations also exhibited ~10 mV shift in steady-state gating toward positive potentials (Fig. 7B and SI Appendix, Table S3). Yet, this minor shift cannot account for the large current deficit observed (Fig. 7C). In order to explore the mechanism of ion permeation in heteromeric channels, we conducted MD simulations on the largest fraction of WT/345A heteromers that consist of two wild-type and two mutated subunits arranged in two different (AAGG and AGAG) orientations as shown in SI Appendix, Fig. S11A. The Alanine heteromers were chosen because Alanine substitution induces the smallest current reduction (Fig. 7 A and C) among disease-causing variants, potentially allowing for the detection of a sufficient number of permeation events for analysis. We observed 69 and 52 permeations through constriction of AAGG and AGAG heteromers, respectively, during 10 µs total simulation time for each channel (SI Appendix, Fig. S11 B and C and Table S2). Analysis of permeations per unit of time revealed a reduction of conduction to roughly one-fourth of the wild type in both heteromers (Fig. 7D). The aperture size of both heteromers was mildly reduced (SI Appendix, Fig. S11 D and E), and there was a slight decrease in the quantity of oxygen atoms within 3 Å of the permeating K+ at the constriction (Fig. 7E). Since MD simulations represent the locked-open pore conformation of channel, we also coexpressed the wild type with the 345A mutant subunits along with KCNE3 in Xenopus oocytes and studied the resultant constitutively open channels (SI Appendix, Fig. S12). The findings indicate that the current reduction in KCNQ1/345A/KCNE3 heteromers is similar to that observed in MD simulations of the AGAG and AAGG channels (Fig. 7D vs. Fig. 7F), with a dominant negative effect of the mutant subunit. In summary, the results indicate that the cause of the pathology in patients with G345A clinical variant is the current deficit resulting from the dominant-negative effect of mutant subunits attributable to the narrowing of the cavity constriction. The clinical mutants with bulkier side chains might affect the heteromeric channels in a similar way but in addition, marked alterations in gating properties, for instance in WT/345R heteromers, may lead to the more pronounced channel dysfunction and a more severe disease phenotypes as observed (45). Discussion For an extended period, it was generally agreed upon that the ion remains in its fully hydrated state within the K+ channel cavity until it reaches the selectivity filter, where an exchange of the water oxygens of its hydration shell with the backbone oxygens of the SF underlies the fast and selective conduction (47, 48). Recent atomic resolution structures have revealed that the cavities of many K+ channels have constrictions with opening widths that are smaller than the hydrated K+ diameter, leaving scientists puzzled. Such a narrow cavity constriction was found in atomic resolution structures of inward rectifying Kir2.2 and KirBac3.2 channels (49–51), small-conductance Ca2+-activated K+ (SK) channels (52), and in the closely related KCNQ2 (53, 54) and KCNQ4 channels (55). A key question raised by these studies is the mechanism of K+ permeation through cavity constrictions, which presents an intriguing biophysical problem and has the potential for finding of drugs with unconventional mechanisms of action. Recent research, primarily focused on inward rectifier channels, suggests that the passage of K+ through the cavity constrictions occurs after the partial dehydration of ions (35–37). Interestingly, it has been reported that the cavity constriction of Kirbac3.1 with a van der Waals accessible opening only slightly larger than the ionic diameter of K+ allows unrestrained ion permeation (36). Cation–pi interactions at the tyrosine side chain are considered important for ion passage through the narrow constriction (36) and the channel’s activity was reduced when tyrosine was substituted with hydrophobic isoleucine in a liposome flux assay (35). The process of ion permeation through Kir3.2, another inward rectifier channel, is also attributed to the partial dehydration of K+ (37). However, in this case, it is believed that the interaction between K+ and the carbonyl oxygens of glycine residue (G318) at the G-loop gate plays a role (37). Our MD simulations demonstrate that every permeating ion transiently interacts with the carbonyl oxygen of G345 prior to its permeation (Fig. 4). However, the majority of ions that interacted with G345’s carbonyl oxygens do not cross the constriction and are instead repelled back to the cytoplasmic compartment (Fig. 4 and Movie S4). This implies that interaction alone perhaps is a necessary but not sufficient factor for K+ to traverse the constriction in KCNQ1. The conformation (or the shape) of the constricted region as the ions (partially dehydrated) come into that contact might be essential. The configuration of water-K+ molecular assembly when it reaches the narrow area could also play a role. The functional analysis of G345 mutants with the side chains of similar size but different chemical properties emphasizes the importance of chemical features of the constriction—a phenomenon observed in both homomeric and heteromeric KCNQ1/KCNE1 mutants (SI Appendix, Figs. S3 and S7). This suggests that the passage of K+ through narrow regions of cavities is a more intricate process than presently understood (36, 37). In KCNQ1, the larger side-chain volumes at G345 has a negative impact on current amplitude for hydrophobic or polar side chains. It should be noted that the narrow cavity constriction in KCNQ1 is created by the bent S6 alpha helices, resulting in a stiffer constricted area compared to those formed by the tyrosine or methionine side chains in inward rectifier channels. In the course of voltage-dependent gating, the flexibility of this region could be enhanced by the converging and diverging movements of the four S6 segments, explaining the higher current in some mutants, such as those with amide groups (SI Appendix, Figs. S3 and S7) compared to the locked-open situation of the same mutants (Fig. 5). Through the resolution of the atomic structure of key mutants studied here with MDs simulations with higher-resolution one can delve deeper into the permeation process and gain a better understanding of the dehydration mechanism of K+ ions at the constriction site. Narrow Constricted Region in Other KCNQ Channels and Its Clinical Significance. Absolute conservation of the PAG motif and a high degree of sequence conservation in the pore region of KCNQ channels point to a common narrow cavity structure and potassium permeation mechanism in all KCNQ family members. Indeed, recent cryo-EM structures of KCNQ2 (53, 54) and KCNQ4 (55) demonstrate a strikingly similar cavity constriction to KCNQ1. The narrowest region in KCNQ2 and KCNQ4 channels’ cavity is formed by alpha carbons of G310 and G316 residues, correspondingly, with a comparable accessible van der Waals opening to KCNQ1 (SI Appendix, Table S1). This indicates that the constriction existing in other KCNQ channel members may impede the transfer of K+ as well. Consistent with this idea, the G310S mutation in KCNQ2, identified in a patient with a severe case of developmental epileptic encephalopathy (DEE) (56), significantly decreases the current amplitude of the M-channel (57). Analysis of KCNQ2/KCNQ3 heteromers has shown that the mutation-induced current reduction is related neither to changes in plasma membrane expression nor to voltage-dependent properties of the channel (57). M-current deficit could be a result of reduced channel conduction similar to the analogous G345S mutant in KCNQ1. A missense variant G316V in KCNQ4 has been found in a patient suffering from autosomal dominant deafness DFNA-2A (58). Although the functional consequences of the KCNQ4-G316V mutation are not yet known, marked reduction of current observed in the homologous KCNQ1-G345V in our study (SI Appendix, Fig. S3) suggests a similar mechanism of action. Analysis of literature, HGMD, and ClinVar databases revealed multiple clinical variants near the constricted cavity region in every member of the KCNQ channel family (SI Appendix, Table S5). While the exact pathological mechanism for the majority of these variants remains to be investigated, it is likely that mutations impact the physical and/or chemical properties of the constricted region of the cavity in these channels. These mutations could markedly influence channel gating as well, emphasizing the significance of investigation of the gating effects when evaluating the overall pathological consequences of mutations. Due to the marked decrease in channel conduction we observed, restoring current levels in patients carrying mutations at or near the constriction site in KCNQ channels may pose a challenge, since currently approved drugs, or those in clinical trials, usually target the voltage-gating properties of these channels. Supporting this notion, patients suffering from DEE due to the G310S mutation in KCNQ2 showed resistance to conventional medication in therapeutic trials (57). However, patients carrying these mutations in KCNQ channels may benefit from drugs that target the constricted region rather than the voltage-dependent properties. Our study suggests that a new therapeutic approach for these patients could involve discovery of drugs that can widen the constricted opening or the entire pore region of KCNQ channels, as a recent study demonstrates (53). Our mutational analysis of the KCNQ1/KCNE1 heteromers indicates that the effect of mutation on channel conductance is much greater than the effect of gating modification, highlighting the potential effectiveness of this approach. Materials and Methods The CompEL (computational electrophysiology) method (39) with dual membrane configuration implemented in GROMACS was used to investigate ion permeations through the pore region of the KCNQ1 channel as previously described (41). A transmembrane voltage, caused by an uneven distribution of 2Cl− ions between compartments, drove the permeation of ions. Simulations were based on the cryo-EM structure of KCNQ1 (28), where the voltage-sensing domains, KCNE3 subunits, and cytoplasmic domains were removed. In order to maintain the open state of the intracellular gate, the terminal amino acids of the S5 (G245-T247) and S6 (A352-K354) were constrained by application of 1,000 kJ mol−1 nm−2 force to the backbone atoms. The protein-membrane system was assembled via CHARMM-GUI and embedded into a POPC lipid bilayer surrounded by water and K+/Na+/Cl− ions. Systems were minimized and equilibrated gradually releasing position restraints ending with 10 ns of NTP ensemble equilibrations. We used the TIP3P water model (59) and Joung and Cheatham ion parameters for TIP3P type of water (60). The SI Appendix section provides a comprehensive overview of mutagenesis, channel expression in Xenopus oocytes, current recordings using the two-electrode voltage clamp method, single-channel recordings procedure, and protein biotinylation and western blotting. Supplementary Material Appendix 01 (PDF) Pore closure in MD simulations without constraints applied to terminal amino acids of S5 and S6 segments. Kv7.1 pore is shown as silver ribbons in tetrameric form, residues S349 and L353 are shown as spheres. Ion permeation through the narrow constriction in partially hydrated state. Only water molecules within 3 Å distance from permeating K+ ion is shown. G345 residue is shown as spheres, the backbone of the selectivity filter is shown as sticks. Only two subunits are shown for simplicity of the view. Lipid molecules are omitted. Replenishment of the hydration shell of K+ by water molecules from the cavity compartment adjacent to the selectivity filter. the Water molecules interacting with permeating K+ ion that originates from the cavity compartment adjacent to the selectivity filter is shown as spheres. Ion repelled back from the narrow cavity region. Water molecules within 3 Å center-to-center distance from K+ ion are shown as sticks. The G345 residue is shown as spheres for two opposite subunits. Interaction of K+ with carbonyl oxygens of the G345 residues. The ion permeating the cavity constriction is shown as a pink sphere. The carbonyl oxygen of G345 residue is represented red, as a part of G345 residue. Ion permeation through the constricted region of the G345T mutant. The mutated Threonine residues are visualized as spheres. Water molecules located 3 Å distance away for permeating ions are depicted as sticks. To simplify the illustration, two opposing subunits and lipid molecules have been excluded. Ion permeation through the constriction of WT/345A heteromers in AAGG configuration. The mutated Alanine is visualized as spheres. Water molecules within 3 Å distance from permeating ions are shown as sticks. For a clearer presentation, two opposing subunits and lipid molecules have been left out. Ion permeation through the constriction of WT/345A heteromers in AGAG configuration. The mutated Alanine and the opposite glycine residues are visualized as spheres. Water molecules close to permeating ions are depicted as sticks within a 3 Å range, with two other opposing subunits and lipid molecules excluded for better view.