Research organism

elife

eLife

2050-084X

eLife Sciences Publications, Ltd

97696

10.7554/eLife.97696

version of record

Research Article

Structural Biology and Molecular Biophysics

Isoleucine gate blocks K⁺ conduction in C-type inactivation

Treptow

Werner

https://orcid.org/0000-0003-4564-3205

treptow@unb.br12†Liu

Yichen

https://orcid.org/0000-0003-0774-6932

3†Bassetto

Carlos AZ

https://orcid.org/0000-0002-7012-5699

2†‡Pinto

Bernardo I

https://orcid.org/0000-0003-0200-1069

2Alves Nunes

Joao Antonio

1Uriarte

Ramon Mendoza

2Chipot

Christophe J

https://orcid.org/0000-0002-9122-1698

245Bezanilla

Francisco

https://orcid.org/0000-0002-6663-7931

26Roux

Benoit

https://orcid.org/0000-0002-5254-2712

roux@uchicago.edu21

https://ror.org/02xfp8v59

Laboratório de Biologia Teórica e Computacional (LBTC), Universidade de Brasília

Brasilia

Brazil2

https://ror.org/024mw5h28

Department of Biochemistry and Molecular Biology, The University of Chicago

Chicago

United States3

https://ror.org/024mw5h28

Department of Neurobiology, The University of Chicago

Chicago

United States4

https://ror.org/04vfs2w97

Laboratoire International Associé Centre National de la Recherche Scientifique et University of Illinois at Urbana−Champaign, Unité Mixte de Recherche No. 7019, Université de Lorraine, Université de Lorraine

Vandœuvre-lès-Nancy

France5

https://ror.org/047426m28

NIH Center for Macromolecular Modeling and Bioinformatics, Beckman Institute for Advanced Science and Technology, and Department of Physics, University of Illinois at Urbana−Champaign

Urbana

United States6

https://ror.org/00h9jrb69

Centro Interdisciplinario de Neurociencia de Valparaíso, Facultad de Ciencias, Universidad de Valparaíso

Valparaíso

Chile

Islas

Leon D

Reviewing Editor

https://ror.org/01tmp8f25

Universidad Nacional Autónoma de México

MexicoMaduke

Merritt

Senior Editor

https://ror.org/00f54p054

Stanford University

United States

†

These authors contributed equally to this work

‡

Department of Physics and Astronomy, The University of Texas at San Antonio, San Antonio, United States

12112024

e97696

0803202428102024

This manuscript was published as a preprint at bioRxiv.

24032024

2024

Treptow, Liu, Bassetto et al

http://creativecommons.org/licenses/by/4.0/

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Many voltage-gated potassium (Kv) channels display a time-dependent phenomenon called C-type inactivation, whereby prolonged activation by voltage leads to the inhibition of ionic conduction, a process that involves a conformational change at the selectivity filter toward a non-conductive state. Recently, a high-resolution structure of a strongly inactivated triple-mutant channel kv1.2-kv2.1-3m revealed a novel conformation of the selectivity filter that is dilated at its outer end, distinct from the well-characterized conductive state. While the experimental structure was interpreted as the elusive non-conductive state, our molecular dynamics simulations and electrophysiological measurements show that the dilated filter of kv1.2-kv2.1-3m is conductive and, as such, cannot completely account for the inactivation of the channel observed in the structural experiments. The simulation shows that an additional conformational change, implicating isoleucine residues at position 398 along the pore lining segment S6, is required to effectively block ion conduction. The I398 residues from the four subunits act as a state-dependent hydrophobic gate located immediately beneath the selectivity filter. These observations are corroborated by electrophysiological experiments showing that ion permeation can be resumed in the kv1.2-kv2.1-3m channel when I398 is mutated to an asparagine—a mutation that does not abolish C-type inactivation since digitoxin (AgTxII) fails to block the ionic permeation of kv1.2-kv2.1-3m_I398N. As a critical piece of the C-type inactivation machinery, this structural feature is the potential target of a broad class of quaternary ammonium (QA) blockers and negatively charged activators thus opening new research directions toward the development of drugs that specifically modulate gating states of Kv channels.

ion conductionmembranefree energyconformationelectrophysiolobymolecular dynamics

Research organismHuman

http://dx.doi.org/10.13039/100000002

National Institute of Health Sciences

R35-GM152124

Roux

Benoit

http://dx.doi.org/10.13039/100000002

National Institute of Health Sciences

R01GM030376

Bezanilla

Francisco

http://dx.doi.org/10.13039/100000001

National Science Foundation

OMA-2121044

Bezanilla

Francisco

http://dx.doi.org/10.13039/100000875

Pew Charitable Trusts

Fellow

Pinto

Bernardo I

http://dx.doi.org/10.13039/501100003593

National Council for Scientific and Technological Development

302089/2019-5

Treptow

Werner

http://dx.doi.org/10.13039/501100003593

National Council for Scientific and Technological Development

200114/2020-4

Treptow

Werner

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Author impact statement

The conformation of an isoleucine gate located along the TM6 segment on the intracellular side below the selectivity filter is a critical component leading to a non-conductive state in the C-type inactivation process of K⁺ channels.

Introduction

The selectivity filter of voltage-gated potassium (Kv) channels is a specialized molecular structure, responsible for the fast and selective conduction of potassium (K⁺) over other ionic species. It is well established that the selectivity filter of Kv channels can exist in both conductive and non-conductive states (Figure 1). Known as the process of (slow) C-type inactivation (Hoshi et al., 1990; Ostmeyer et al., 2013), the conductive to non-conductive conformational transition is of great physiological importance as it contributes to fine-tune long-term activity of Kv channels. The canonical conductive conformational state of the filter was first revealed with the crystallographic structure of the prototypical bacterial K⁺ channel KcsA at high resolution (e.g. PDB id 1BL8 [Doyle et al., 1998] or 1K4C [Zhou et al., 2001]). Molecular dynamics (MD) studies confirm that ion conduction along this filter conformation is possible and unopposed by large free-energy barriers (Bernèche and Roux, 2001; Kopec et al., 2019). Since, additional structures have been resolved for other Kv channels of the (eukaryotic) Shaker family, broadening our knowledge of the conductive state of the filter (Long et al., 2007). Recently, high-resolution structures of Kv channels revealed a novel conformation of the selectivity filter that is partially dilated at its outer end and constricted near its internal face (Tan et al., 2022; Reddi et al., 2022; Yangyu et al., 2024). Because a number of mutations known to strongly enhance the process of C-type inactivation were introduced in the construct used in structural determination (Perozo et al., 1993), this ‘dilated’ conformation has been interpreted as the non-conductive conformational state of the selectivity filter, accounting for the phenomenon of C-type inactivation. Reinforcing that notion, the most recent structure of the wild-type Shaker B channel displays the same dilated conformation of the filter at low concentration of external potassium (Stix et al., 2023).

Figure 1.Comparative analysis of Kv channel structures.

(A) Schematic representation of a voltage-gated K⁺ channel undergoing C-type inactivation, whereby prolonged activation by an external voltage V leads to blockage of ionic conduction across the selectivity filter of the open channel. The voltage-sensor positively charged S4 helix (blue), the S4S5 linker (red), and main-pore S6 helix (silver) are highlighted. (B) Structural models of the selectivity filter in the conductive and dilated conformations. The conductive and dilated conformations derive respectively from the high-resolution X-ray structures of the wild-type (Long et al., 2007) and triple-mutant (Reddi et al., 2022) kv1.2-kv2.1 channel (PDB codes 2R9R and 7SIT). Major structural deviations (root-mean-square deviation [RMSD]) between the selectivity filter conformations are primarily accounted for side-chain rearrangements of Y373 and D375. Despite the side-chain rearrangements of Y373 and D375, the profile of oxygen-oxygen distances between opposing subunits of the selectivity filter indicates that the geometry of sites S₄ and S₃ in the dilated conformation closely resembles that of the conductive state. For comparison purposes, the profile of oxygen-oxygen distances is also shown along the selectivity filter of the experimental structures of the conductive state of Shaker B (Tan et al., 2022) (PDB code 7SIP) and the dilated conformations of Shaker-W434F (Tan et al., 2022) (PDB code 7SJI) and Kv1.3 (Selvakumar et al., 2022) (PDB code 7SJ1). (C) Electrostatic properties of the conductive and dilated conformations of the selectivity filter. Shown are molecular representations of the main-pore S6 segments of the channel, highlighting the permeation pathway along the intracellular entrance (iN), central cavity (cC), and selectivity-filter sites (S₄, S₃, S₂, S₁, S₀, S_ext). Hydrated cavities (red) give ionic access to the selectivity filter from the intracellular and extracellular milieu. The dielectric morphology of the protein and waters accounts for a significant voltage drop (ɸ) across the selectivity filter. The voltage-drop profile along the permeation pathway was computed as described elsewhere (Souza et al., 2014), following the charge-imbalance protocol which is a variant of the linear field method (Roux, 2008).

The ‘dilated’ and ‘conductive’ filter conformations differ markedly. The all-atom root-mean-square deviations (RMSD) of the selectivity filter between the dilated and the conductive conformation is about 3.0 Å, arising mostly from rearrangements of the side chain of the highly conserved tyrosine and aspartic acid along the signature of the selectivity filter TVGYGD (Figure 1). Yet, despite the considerable conformational differences, a simple inspection reveals no apparent physical barrier opposing ion conduction along the permeation pathway of the ‘dilated’ structure, challenging its postulated non-conductivity. The selectivity filter in the dilated conformation is accessible to the intracellular solution via a large open and hydrated vestibular cavity. Incoming ions can bind to sites S₄ and S₃ and translocate to the extracellular side via a hydrated crevice corresponding to the widened sites S₂, S₁, and S₀. The open and fully hydrated vestibular cavity is clearly conductive in the dilated conformation, which disagrees with previous cysteine modification and blocker-protection assays indicating that the cavity of the channel is changed in the slow-inactivated state compared with the open state (Panyi and Deutsch, 2007). Besides, the voltage drop across the selectivity filter through the conductive and dilated conformations is similar (Figure 1C). In contrast to classic chemical and peptide blockers or the constricted filter associated with the inactivation of the KcsA bacterial channel that typically inhibit conduction by providing a physical barrier along the permeation pathway (Ostmeyer et al., 2013; Yangyu et al., 2024; Eriksson and Roux, 2002; Banerjee et al., 2013; Cuello et al., 2010), the dilated conformation does not appear to impose any structural impediment to ionic conduction. These observations beg for a quantitative assessment of the ion conduction properties of the dilated conformation, and the relation of the latter to the functional C-type inactivated state identified in electrophysiological experiments.

Results

To address these questions, detailed all-atom MD simulations were carried out on the basis of the high-resolution X-ray structure of the kv1.2-kv2.1-3m chimera channel containing the triple mutation W362F, S367T, and V377T (Reddi et al., 2022). These mutations enhance C-type inactivation in the kv1.2-kv2.1 chimera, making the pore similar to the pore of the Shaker channel. To circumvent the uncertainties inherent with MD and achieve robust conclusions, simulations based on two established force fields, AMBER (Maier et al., 2015; Joung and Cheatham, 2008) and CHARMM36m (Huang et al., 2017), were considered. Additional simulations were also carried out using a version of CHARMM36m force field with a few modified interactions to explain the difference in ion conduction with the AMBER force field (Materials and methods, Supplementary files 1 and 2) In all the simulations, the channel was embedded in a fully hydrated phospholipid bilayer at 150 mM KCl and simulated with an applied transmembrane (TM) of +200 mV.

The kv1.2-kv2.1-3m channel with the dilated selectivity filter appears remarkably stable during a 10 μs simulation using the AMBER force field (Figure 2). The structural RMSD of the selectivity filter relative to the initial X-ray structure (Reddi et al., 2022) is less than 2.8 Å, indicative of structural stability (Figure 2B, structural deviation R). Simulation of kv1.2-kv2.1-3m with the CHARMM36m force field also illustrates the stability of the dilated conformation over the microsecond timescale, with structural deviation of the selectivity filter less than 2.5 Å (Figure 2C, structural deviation R). The average density of the K⁺ ions along the permeation pathway from the AMBER trajectory is consistent with a predominant occupancy of the sites S₄, S₃, and S_ext. Not resolved in the X-ray structure of kv1.2-kv2.1-3m (Reddi et al., 2022), the density at the external site S_ext results from close interactions of K⁺ and the carboxylate group of D375. The acidic side chain adopts a relaxed conformation throughout the simulation, in which the carboxylate moiety is fully exposed to the external solution in an orientation similar to that previously reported in the cryogenic electron microscopy (cryo-EM) structure of the homologous Kv1.3 channel with a dilated selectivity filter (Selvakumar et al., 2022). According to a hard-knock mechanism at the level of binding sites S₄ and S₃ (Figure 2—figure supplement 1; Stix et al., 2023), there is an early voltage-driven K⁺ conduction event across the selectivity filter within the first ~2 μs of the simulation (Figure 2B, trajectory of K⁺ along the permeation pathway Z). Conduction increases the residence time of the ion in the dilated region of the selectivity filter and accounts for a minor density peak at the level of site S₁ (Figure 2B, linear density of K⁺), which is consistent with the resolved electron density of the cation in the experimental structures (Tan et al., 2022; Reddi et al., 2022; Yangyu et al., 2024; Stix et al., 2023).

Figure 2.Molecular dynamics (MD) simulation of kv1.2-kv2.1-3m at +200 mV.

(A) Molecular representation of the main pore of the channel, highlighting the initial configuration of the selectivity filter, I398 (red) and V402 (light gray). (B and C) Analysis of AMBER and CHARMM36m trajectories. Shown is the structural deviation of the selectivity filter (R), the trajectory of K⁺ ions along the permeation pathway (Z), and the intersubunit C_β-C_β separation distance of I398 and V402 (L) as a function of simulation time. Inset shows instantaneous configurations of I398 (red arrows). Time averages are the linear density of K⁺ ions, the linear density of water oxygen, and the pore radius profile (r) along the permeation pathway Z.

Figure 2—figure supplement 1.Ion conduction across the dilated conformation of the selectivity filter of kv1.2-kv2.1-3m.

(A, B) Shown are representative single-ion conduction events across the selectivity filter of the triple-mutant channel along AMBER and CHARMM36m simulations. (C) Closure of the isoleucine gate blocks ion conduction across the the selectivity filter.

Figure 2—figure supplement 2.Molecular dynamics (MD) simulation of kv1.2-kv2.1-3m at +200mV.

(A) Molecular representation of the main pore of the channel, highlighting the initial configuration of the selectivity filter, I398 (red) and V402 (light gray). (B and C) Analysis of AMBER* (restrained to keep I398 open) and CHARMM36m-NBFIX trajectories. Shown are the structural deviation of the selectivity filter (R), trajectory of K⁺ along the permeation pathway (Z), and the intersubunit C_β-C_β separation distance of I398 and V402 (L) as a function of simulation time. Inset shows instantaneous configurations of I398 (red arrows). Time averages are the linear density of K⁺ ions, the linear density of waters, and the pore radius profile (r) along the permeation pathway Z.

The knock-on mechanism depends on the concentration of incoming ions in the central cavity of the channel. Although the last conduction event takes place around ~2 μs of simulation, note that only after ~4 μs of simulation, the knock-on mechanism is persistently disrupted as ions are excluded from the central cavity of the channel by the isoleucine side chains at position 398. More specifically, ion access to the central cavity of the channel and their subsequent conduction across the selectivity filter completely cease after ~4 μs of simulation time when the side chain of isoleucine at position 398 from all four S6 segments twist toward the central axis of the channel, dehydrating and blocking the permeation pathway immediately beneath the selectivity filter (Figure 2B, density of water along the pore axis). The C_β-C_β distance between I398 in opposing subunits decreases over a period of 3–4 μs, reaching a value of about 9 Å—in stark contrast with the initial distance of 15 Å in the X-ray structure (Reddi et al., 2022; Figure 2, separation distance L of I398). As a consequence, the pore radius becomes locally constricted (≤2 Å), blocking ion conduction (Figure 2B, pore radius r) (Treptow and Tarek, 2006).

In contrast, ion conduction events are observed in the simulation with the CHARMM36m force field. This observation is in sharp contrast with the conclusion drawn from previous simulations of the dilated conformation of Shaker B also based on the CHARMM36m force field, where no ion conduction event was observed even with an applied membrane potential of +300 mV (Tan et al., 2022; Stix et al., 2023). Importantly, dihedral restraints were applied in these simulations to preclude deviations from the X-ray structure; it is possible that those restraints apparently prevented ion conduction through the open-dilated conformation. Because this is the same conformation of the selectivity filter and the same CHARMM36m force field, suggesting that the key difference in the simulation results is more likely due to different MD simulation conditions used than the very minor differences in the pore domain between Kv1.2-2.1-3m and Shaker. Because the CHARMM36m force field favors the open configuration of the isoleucine gate, the simulation shows that the dilated conformation is highly conductive under membrane depolarization. The distribution of ions along the permeation pathway in the CHARMM36m simulation is distinct from that inferred from the AMBER simulations, as the increased mobility of K⁺ in the selectivity filter accounts for a more pronounced reallocation of the ionic density from sites S₄/S₃ to sites S₂/S₁.

To further clarify the ion conduction properties of the dilated filter and understand the role of isoleucine 398, we simulate the channel with the AMBER force field in the presence of harmonic distance restraint to keep the isoleucine gate in the open configuration (Figure 2—figure supplement 2). Compared to the unrestrained simulation with the AMBER force field, a larger number of ions access the central cavity of the channel, triggering two spontaneous conduction events across the selectivity filter in the early ~4 μs of the simulation. The density profile of the K⁺ ion along the channel axis is distinct from that of the unconstrained simulation. Partial reallocation of the ionic density from sites S₄/S₃ to sites S₂/S₁ reflects the increased mobility of K⁺ in the selectivity filter, thereby corroborating the hypothesis that the dilated conformation of the selectivity filter is conductive when the isoleucine gate is open. This conclusion is further confirmed from additional simulations based on a hybrid CHARMM36m force field referred to as CHARMM36m-NBFIX in which only the ion-carbonyl, ion-water, and water-carbonyl interactions were adjusted to mimic the values from the AMBER force field (Figure 2—figure supplement 2, Supplementary file 2). While the changes are fairly small and reproduce known behavior for the conductive state of the selectivity filter (Materials and methods, Supplementary file 3), the simulations recapitulate the ion conduction properties of the dilated conformation from AMBER, demonstrating how these three key interactions are directly responsible for the observed differences with CHARMM36m. In the CHARMM36m-NBFIX simulation, the local concentration of ions in the central cavity of the channel is significantly smaller than that of the original trajectory. Pronounced density peaks at sites S₄/S₃ indicates that K⁺ binds more strongly to the selectivity filter. The combined effect yields a conductive AMBER-like channel, characterized by fewer conduction events per simulation time.

These simulations strongly suggest that there is nothing about the dilated conformation of the filter that inherently impedes ion conduction, and that the conformational motion of I398 is necessary to truly block conduction. Thus, it is the conformational change implicating the isoleucine gate that leads the channel toward a true non-conductive state. As a significant modification of the channel structure, the local rearrangement of I398 seems to be coupled to motions of other regions of the main pore, including V402 at the highly conserved PVP motif (Figure 2). Analysis indicates that CHARMM36m favors the open and fully hydrated state of the I398 gate. The average intersubunit C_β-C_β distance of the isoleucine side chains (~16 Å) is close to the reference value in the X-ray structure (~15 Å) most of the simulation time. Gating fluctuations of I398 are, however, clearly observed in the late stages of the simulation and correlate well with the reduction of ions in the central cavity of the channel and with the conduction across the selectivity filter. Particularly important, the CHARMM36m simulation adds support to the assumption that the dilated conformation of the selectivity filter is conductive, and that closure of the isoleucine gate is required to shut down ion transport across the channel.

Despite intrinsic force-field differences with respect to channel conductivity, all three atomistic models support the conclusion that the dilated conformation of the selectivity filter is, by itself, conductive and the isoleucine gate seems to be important to block K⁺ current across the channel. According to the voltage-driven MD trajectories in which the isoleucine gate is open, the total number of conduction events across the dilated conformation of the selectivity filter over the total simulation time is 44 (ions)/30 μs (Supplementary file 4). Based on the number of crossing events per simulation time, the ionic current of a macroscopic population of ion channels in the open-gate dilated state is ~0.2 pA at the TM voltage of +200 mV (single-channel conductance of ~1.17 pS in symmetric 150 mM KCl). In all likelihood underestimation of the channel conductance as a consequence of the well-documented force-field limitations in reproducing the ionic current in Kv channels at low voltages (<300 mV) (Jensen et al., 2013), the estimate of ~0.2 pA is still orders of magnitude larger than the measured current in the triple-mutant channel upon C-type inactivation (vide infra), and, therefore, the conductivity properties of the ‘dilated’ conformation of the selectivity filter cannot explain alone the inactivation of kv1.2-kv2.1-3m under membrane depolarization. Consistent with single-channel measurements (Yang et al., 1997), the estimate of ~1.17 pS is actually more comparable to simulation predictions of the single-channel conductance of the conductive selectivity filter i.e., ~3.5 pS in symmetric 300 mM KCl (Stix et al., 2023). Beside the dilated conformation of the selectivity filter, the isoleucine gate then appears to be a critical molecular element of the channel machinery, largely implicated in C-type inactivation.

To validate and corroborate the key role of the isoleucine gate in C-type inactivation inferred from the MD simulations, electrophysiology experiments were carried out (Materials and methods, Figure 3). In stark contrast with recordings of the kv1.2-kv2.1 chimera channel (Figure 3A), C-type inactivation is greatly enhanced in the triple-mutant channel kv1.2-kv2.1-3m, as is evidenced by the fast decay of the ionic current in the ms timescale and the macroscopic current appearing on similar scales as gating current (Figure 3B, Figure 3—figure supplement 1). However, substitution of the isoleucine by the polar amino acid asparagine, with similar side-chain volume, restores the conductivity of the triple-mutant channel (Figure 3C). While not disturbing the current-voltage relationship of the triple-mutant channel, mutation I398N drastically increases ionic conduction without any apparent time-dependent inactivation (Figure 3D and E). Importantly, restoration of ion conduction with I398N is not caused by inadvertently stabilizing the selectivity filter in the conductive state, as demonstrated by pore-blocking toxin assays. Agitoxin-II (AgTxII), dendrotoxin, and charybdotoxin (CTX) are potent toxin blockers of Kv channels (Takacs et al., 2009), binding to the outer mouth of the channel (Yangyu et al., 2024; Eriksson and Roux, 2002; Banerjee et al., 2013). Such pore-blocking toxins preferentially bind and occlude the conductive conformation of the selectivity filter of Kv channels and are not effective in blocking channels that have W434F-like selectivity filter, as indicated by previous functional studies (Kitaguchi et al., 2004) as well as implicit solvent binding free-energy calculations (Materials and methods, Figure 3F, Figure 3—figure supplement 2, Supplementary file 5). Whereas AgTxII blocks the ionic current across the conductive kv1.2-kv2.1 channel by binding to the outer mouth of the filter in the conductive conformation (Figure 3G), it fails to bind and occlude the kv1.2-kv2.1-3m channel with its filter mostly in the dilated conformation (Figure 3H). Similar effect is observed in Kv1.2-kv2.1-3m_I398N, suggesting the resumed ion permeation is not caused by any stabilization effect of I398N on the selectivity filter and most likely the ions are conducting through the dilated filter (Figure 3I). Because kv1.2-kv2.1-3m and kv1.2-kv2.1-3m_I398N are not affected by the toxin (Figure 3J), the ion conduction in kv1.2-kv2.1-3m_I398N mutant indicates that the permeation pathway must be altered at some location along the permeation pathway other than the selectivity filter. Such structural modifications of the permeation pathway are specific of C-type inactivation since a double mutant (S367T/V377T) of the chimera channel kv1.2-kv2.1-2m that does not inactivate and is blocked by AgTxII, irrespective of the I398N mutation (Figure 3—figure supplement 3), suggesting that I398N mutation neither affect the selectivity filter of kv1.2-kv2.1-2m nor kv1.2-kv2.1-3m. The critical C-type inactivation mutation W362F (Perozo et al., 1993) is missing in the kv1.2-kv2.1-2m channel, hence its inability to slow-inactivate.

Figure 3.Electrophysiology measurements of I398N substitution.

(A, B, C, and D) Macroscopic current recorded from: (A) kv1.2-kv2.1 chimera, (B) triple-mutant kv1.2-kv2.1-3m, (C) triple-mutant kv1.2-kv2.1-3m with I398N substitution, and (D) their respective comparison at +80 mV (the line colors correspond to the colors indicated in A, B, and C of the three mutants at +80 mV). The triple mutation W362F, S367T, and V377T in kv1.2-kv2.1-3m significantly speed up the inactivation process and the gating current could be seen simultaneously with ionic current (shown in inset, with gating current highlighted in yellow and ionic current highlighted in blue). Note the effect of the triple mutation cancel out with the I398N substitution. (E) Current-voltage relationship (IV curve). Due to the mixture of gating current and ionic current in kv1.2-kv2.1-3m, only curves shown are taken from the peak current. Voltage-dependent activation relationship for kv1.2-kv2.1, kv1.2-2.1-2m, and kv1.2–2.1-2m_I398N are shown in Figure 3—figure supplement 3. (F) Net free-energy difference (ᐃᐃG) involved in the binding of the charybdotoxin (CTX) to the conductive (Banerjee et al., 2013) and dilated conformations of the selectivity filter of the kv1.2-kv2.1 chimera channel. ᐃᐃG indicates a strong preference of CTX to the conductive conformation of the selectivity filter. The same binding preference is found between CTX and kv1.2-kv2.1-3m (Supplementary file 5). (G, H, I, and J) Effects of agitoxin-II, a more potent CTX analogous (Takacs et al., 2009), on kv1.2-kv2.1 chimera, triple-mutant kv1.2-kv2.1-3m, triple-mutant kv1.2-kv2.1-3m with I398N substitution. Clearly, agitoxin binds and blocks the chimera channel while shows minimal influence on triple and triple_I398N, suggesting the selectivity filter in triple_I398N likely also adopts a dilated conformation.

Figure 3—figure supplement 1.Decay of the ionic current of the triple-mutant kv1.2-kv2.1-3m channel.

(A) Ionic traces from the triple-mutant channel and (B) its time constants fitted with a one exponential decay. When extrapolated to +200 mV where the simulation was performed, the time constant (tau) was in the microsecond range (gray dashed line indicates 1 ms), suggesting the timescale of the simulation and the conformational changes seen were relevant in physiological terms. Data plotted with mean and SEM.

Figure 3—figure supplement 2.Estimation of the binding free energy of charybdotoxin (CTX).

(A) Docking of CTX to the selectivity filter of the kv1.2-kv2.1 chimera channel. (B) Docking solutions (light black traces) best reproducing the experimentally resolved bound state of the toxin (tick black trace, PDB 4JTD). (C) Dielectric (∈) map of the toxin-protein complex considered in the Poisson-Boltzmann calculation. (D, E, and F) Energy distribution of best docking solutions shown in (B). Poisson-Boltzmann (ᐃG_LP) and van der Waals (ᐃE_LP) electrostatic contributions to the binding free energy of CTX are shown. The van der Waals component was scaled by an empirical factor λ=0.17, intended to resolve the protein-solvent interaction absent in the implicit solvent representation. The same procedure was adopted to estimate the binding energy of CTX to all channel constructs and states (Supplementary file 5).

Figure 3—figure supplement 3.State dependence of I398N effects.

(A) Macroscopic current from the double-mutant chimera channel kv1.2-kv2.1-2m (S367T/V377T). (B) Macroscopic current from kv1.2-kv2.1-2m with I398N substitution. Mutation I398N in the absence of W362F does not significantly alter the phenotype of the channel. (C) G-V curves for kv1.2-kv2.1, kv1.2–2.1-2m, and kv1.2–2.1-2m_I398N. The G-V curves were measured from the tail currents and fitted using a two-state model given by: G(V)=[1+exp(zFRT(V−V12))]−1, where z is the apparent charge expressed in units of elementary charge e0, V is the voltage and V1/2 is the voltage of half maximal conductance. R, T, and F have their usual meanings. All three mutants share similar voltage dependency in activation: V_1/2 (mV): kv1.2-kv2.1 (–4.6±0.7), kv1.2-2.1-2m (–15.9±0.6) and kv1.2-2.1-2m_I398N (–8.1±0.6) and z: kv1.2-kv2.1 (2.4±0.1), kv1.2-2.1-2m (2.7±0.2) and kv1.2-2.1-2m_I398N (2.5±0.1). All three mutants share similar voltage dependency in activation. (D, E, and F) Effects of 100 nM agitoxin-II (AgTxII) on the double-mutant kv1.2-kv2.1-2m and double-mutant kv1.2-kv2.1-2m with I398N substitution. AgTxII binds and blocks efficiently both channel constructs, suggesting the effect of I398N depends on the C-type inactivation induced by the W362F mutation.

Figure 3—figure supplement 4.Molecular dynamics (MD) simulation of kv1.2-kv2.1-3m with I398N at +200mV.

(A) Molecular representation of the main pore of the channel, highlighting the initial configuration of the selectivity filter, N398 (green) and V402 (light gray). (B and C) Analysis of AMBER and CHARMM36m trajectories. Shown are the structural deviation of the selectivity filter (R), trajectory of K⁺ along the permeation pathway (Z), and the intersubunit C_β-C_β separation distance of N398 and V402 (L) as a function of simulation time. Inset shows instantaneous configurations of N398 (green arrows). Time averages are the linear density of K⁺ ions, the linear density of waters, and the pore radius profile (r) along the permeation pathway Z.

Figure 3—figure supplement 5.Primary-sequence conservation throughout the main-pore segments PH, SF, and S6.

(A) Logos conservation across K⁺ channels. (B) Multiple sequence alignment of most studied K⁺ channels.

Consistent with experiment, a polar amino acid allows ion conduction by keeping the permeation pathway constitutively hydrated and open. Additional simulations of the triple-mutant channel support this view by revealing that I398N prevents closure of the gate at position 398 (Supplementary file 1). Figure 3—figure supplement 4 shows the AMBER simulation of the triple-mutant channel with the I398N mutation. The conformation of the dilated selectivity filter is stable in presence of the mutation. The average intersubunit C_β-C_β distance of I398N fluctuates between ~11 Å and ~15 Å in the beginning of the AMBER trajectory, before reaching the value of ~9 Å in the final stage of the simulation. Compared to the wild-type simulation, the local pore radius and water density are significantly enhanced under I398N mutation (Supplementary file 6). The open and hydrated configuration of the gate in the I398N mutant allows ion conduction across the dilated selectivity filter as long as structural fluctuations of the PVP motif (V402) do not obstruct the permeation pathway, as expected because that motif is the main gate of the channel. Similar conclusions can be drawn from CHARM36m simulation in which long-lived fluctuations of the mutant gate I398N allows intermittent conduction of ions across the dilated conformation of the selectivity filter over the microsecond timescale.

Discussion

Taken together, the present computational and experimental results demonstrate that the dilated conformation of the selectivity filter of kv1.2-kv2.1-3m is conductive and that an isoleucine gate is critical to block K⁺ currents during C-type inactivation of the channel. Judged by the primary sequence conservation of I398 (Figure 3—figure supplement 5), the isoleucine gate seems to be relevant for potassium channels that undergo C-type inactivation in general, and, potentially, for other voltage-gated Na⁺ channels possessing distinct selectivity filters (Liu et al., 2023). The action of the isoleucine gate in C-type inactivation is averted by the I398N mutation because the pathway remains hydrated with the polar asparagine side. In the homologous Shaker B channel, the single mutation I470C affects the rate of inactivation (Holmgren et al., 1997; Peters et al., 2013) while the double mutation T449V/I470C (Olcese et al., 2001) converts the slow inactivated state into a conductive state. Both mutations alter the slow-inactivation phenotype under long depolarizations, strongly corroborating our findings. Closing of the I398 gate in the dilated conformation of kv1.2-kv2.1-3m involves occlusion of the binding site of internally applied quaternary ammonium (QA) blockers (Lenaeus et al., 2005), explaining the previously reported 20-fold decreased affinity of TEA for the inactivated state of Shaker-IR compared with that of the open state (Panyi and Deutsch, 2007). On the other hand, the mechanism whereby I398N renders the gate constitutively open, not occluding the binding site of QA blockers, is also consistent with the demonstration that I470C in Shaker-IR morphs the channel that does not trap QA blockers into one that does (Holmgren et al., 1997) and with the fact that MTSEA modifications on the inactivated state are sixfold slower than in the open state of the T449K/I470C Shaker-IR (Panyi and Deutsch, 2007). Across all these measured effects, the modus operandi of the isoleucine gate is expected to be coupled to the PVP motif, and, therefore, to reflect to some extent the conformational allostery between the selectivity filter and the bundle-crossing region previously reported for C-type inactivation (Panyi and Deutsch, 2007; Cuello et al., 2017; Labro et al., 2018). While these experimental results were indicative of the role of the isoleucine gate its mechanistic significance with regards to the non-conductive C-type inactivation seems to have been overlooked (Tan et al., 2022; Reddi et al., 2022; Yangyu et al., 2024; Stix et al., 2023).

It is important to address the approximate and imperfect nature of the atomic models used in the present MD simulations. Even with well-established force fields like AMBER and CHARMM36m, ion conduction through the selectivity filter in the canonical ‘conductive’ conformation tends to be too small compared to experimental values, especially at physiological voltages (less 50 mV) (Jensen et al., 2013). With AMBER, there is observable ion conduction via the hard-knock mechanism (no water molecules between the ions) but a fairly high voltage (200–300 mV) is required to generate a significant ionic current (Köpfer et al., 2014). With CHARMM36m, there is also observable conduction via the hard-knock mechanism at high voltages, albeit less so than with AMBER. Thus, the simulated conductance through K⁺ channels is generally too small, and it is relative to this baseline that conduction through the open dilated conformation of the filter based on the of kv1.2-kv2.1-3m structure must be critically assessed. Here, along the dilated conformation of the selectivity filter, we find that there is no substantial ion conduction with the AMBER force field because the ion binds strongly to the selectivity filter sites S₄ and S_3. This behavior is reproduced by the CHARMM36m-NBFIX force field, with its AMBER-like ion-carbonyl interactions. However, there is noticeable ion conduction with the CHARMM36m force field because the ion does not bind strongly to the sites S₄ and S₃. In fact, multiple conduction events across the open-dilated conformation of the triple-mutant chimera channel were observed in the simulation with the CHARMM36m force field, with no applied restraints.

Our physiological experiments demonstrate that by mutating I398 to a polar residue, ion permeation can be resumed in the triple mutation. More importantly, this effect is not a result of reverting C-type inactivation and stabilizing the filter in the conductive state, as shown by the toxin blocking experiments (Figure 3, Figure 3—figure supplement 3). This implies that the potassium ions are permeating through the dilated filter in kv1.2-kv2.1-3m_I398N, providing strong evidence that I398 is in fact the residue that is responsible for blocking the ion conduction during C-type inactivation instead of the dilated filter. We note that the activation and deactivation become extremely fast in the kv1.2-kv2.1-3m_I398N construct which maybe the consequence of rendering the slow-inactivated state conductive, a fact that should be explored further but is beyond the scope of this study. However, the main conclusion still stands: I398 most likely form the barrier for conduction in C-type inactivated channels and the dilated filter itself is not sufficient.

Based on our findings, we propose that the structures of kv1.2-kv2.1-3m (Reddi et al., 2022), Shaker-W434F (Tan et al., 2022), Kv1.2-W366F (Yangyu et al., 2024), and Shaker-lowK (Stix et al., 2023) corresponds to a conductive metastable intermediate on the path toward the true non-conductive C-type inactivated state wherein the permeation pathway is blocked by the isoleucine gate. The free-energy landscape associated with the with C-type inactivation process comprises two sequential transitions between three metastable states: conductive → dilated-intermediate → deeply inactivated. Each transition was characterized by a separate potential of mean force (PMF) calculation (Materials and methods, Figure 4, Figure 4—figure supplements 1 and 2). The first PMF monitoring the conformation of the selectivity filter while the isoleucine gate is in the open conformation shows two metastable states: conductive and dilated. The transition, which favors the latter state with a downhill free-energy change of –12 kcal/mol, involves a significant conformational change of Y373 from its buried configuration next to T367 in the conductive state toward its externally exposed orientation in the intermediate state resulting from unfavorable close interactions between the polar side chain of Y373 and F362. This is actually a metastable intermediate state that remains conductive. As shown by the second PMF monitoring the closure of the isoleucine gate toward a non-conductive deep inactivated state while the filter remains in dilated conformation shows a downhill change of –6.0 kcal/mol. Notwithstanding the limited accuracies of these computational estimates, the forward and backward free-energy barriers for these two transitions are in qualitative agreement with C-type inactivation of the triple-mutant channel in the sense that kv1.2-kv2.1-3m inactivates substantially faster than recovers from it—at +200 mV, the time constant of C-type inactivation is predicted to be in the same microsecond range of the simulated deep inactivation of the channel (Figure 3, Figure 3—figure supplement 1). The overall free-energy landscapes pointing toward the greater stability of the ‘deep inactivated’ state is qualitatively correct.

Figure 4.Mechanism of C-type inactivation of the triple-mutant channel kv1.2-kv2.1-3m.

(A) Free-energy profile w(s) along the conformational transition path s connecting the conductive (s=0.2) and dilated (s=1) states of the selectivity filter. The free-energy profile is conditional to the open configuration of the isoleucine gate. (B) Free-energy profile w(d) associated to closure of the isoleucine gate. The reaction coordinate d corresponds to the inter-subunit separation distance between β-carbon atoms of I398. The free-energy profile is conditional to the dilated conformation of the selectivity filter. (C) C-type inactivation mechanism of the triple-mutant channel inferred from the representative structures of the selectivity filter and isoleucine gate along the free-energy profiles (A) and (B).

Figure 4—figure supplement 1.Conformational path between the conductive (<bold>O</bold>) and dilated (<bold>D</bold>) states of the selectivity filter.

(A, B) Time evolution of the instantaneous (red) and target (black) root-mean-square deviation (RMSD) of the selectivity filter between states O and D. Reference structures of the dilated and conductive states of the selectivity filter were respectively defined on basis of the high-resolution X-ray structures of the triple-mutant kv1.2-kv2.1-3m and kv1.2-kv2.1 (PDB codes 7SIT and 2R9R). Targeted molecular dynamics (TMD) was carried out for 50 ns with an applied constant force of 400 kcal/mol/Å², corresponding to a per atom constant force of 2.38 kcal/mol/Å². (C) A total of 22 atomic distances were used for definition of the conformational path between states D and O (red dashes): 14 main-chain carbonyl distances between two opposing subunits of the channel, 4 side-chain distances between TYR373(OH) and THR367(O), and 4 side-chain distances between ASP375(OD2) and PHE362(CD1). Atomic distances were symmetrized across the channel to support convergence of the calculation. (D) Time evolution of a representative coordinate of the transformation path. The coordinate was smoothed out following a running average procedure of the data. The running average length was 10% of the total number of data points, i.e., 500 (red). (E) The resulting string z consisted of 111 equidistant points i satisfying the monotonic condition zi−zi−1+zi+1−zi≤zi+1−zi−1 at a distance interval of 0.25 Å.

Figure 4—figure supplement 2.Convergence analysis of free-energy calculations.

(A, B) Respectively shown is the root-mean-square deviation (RMSD) of the free-energy profiles w(s) and w(d) as a function of simulation time t.

Figure 4—figure supplement 3.Molecular dynamics (MD) simulation of kv1.2-kv2.1-3m at +200 mV and 150 mM NaCl.

(A) Molecular representation of the main pore of the channel, highlighting the initial configuration of the selectivity filter, I398 (red) and V402 (light gray). Two sodium (Na⁺) ions (yellow) are shown in the selectivity filter at the ion binding sites S₃ and S₄. (B and C) Analysis of AMBER and CHARMM36m trajectories. Shown are the structural deviation of the selectivity filter (R), trajectory of Na⁺ along the permeation pathway (Z), and the intersubunit C_β-C_β separation distance of I398 and V402 (L) as a function of simulation time. Inset shows instantaneous configurations of I398 (red arrows). Time averages are the linear density of Na⁺ ions, the linear density of waters, and the pore radius profile (r) along the permeation pathway Z.

The new structural insights into the intricate mechanism of C-type inactivation suggests new research directions in the field. Worth of investigation is the experimental observation that Shaker B leaks Na⁺ in the absence or low concentration of K⁺ in the C-type inactivation state (Starkus et al., 1997). AMBER and CHARM36m simulations of kv1.2-kv2.1-3m in presence of 150 mM NaCl show that one or two Na⁺ ions can stably bind the dilated conformation of the selectivity filter at sites S₃, S₄, and S_ext, while favoring the open configuration of the isoleucine gate (Materials and methods, Figure 4—figure supplement 3 and Supplementary file 1). The internal [Na⁺]/[K⁺] concentration ratio seems to affect the closing of the isoleucine gate with functional implications for sodium leak in C-type inactivation. Also important, it is the more extensive investigation of the structure-function relationship of the inactivation gate according to the amino acid sequence within the pore domain. In particular, note that except for hERG, most studied potassium channels including, Kv1.2, Shaker-B, KcsA, and MthK, display either isoleucine, leucine, valine, or phenylalanine at position 398 (Figure 3—figure supplement 5). The hERG channel has a tyrosine at position 398 which in the cryo-EM structure (Wang and MacKinnon, 2017) is twisted toward the central cavity in the open state. Because tyrosine is bulkier than asparagine, there is a chance that C-type inactivation in hERG also involves constriction of the permeation pathway at position 398 when the selectivity filter is dilated—another fascinating assumption worth of investigation in structural studies. As a piece of the molecular machinery implicated in multiple states, the isoleucine gate might interfere in a state-dependent manner with the mechanism of action of a broad class of QA blockers (Panyi and Deutsch, 2007) and negatively charged activators that bind beneath the selectivity filter and operate as master keys to open a variety of K⁺ channels (Schewe et al., 2019). Of particular importance, the understanding of this last aspect could offer a distinct advantage in the development of drugs that modulate gating states of potassium channels more specifically.

Materials and methodsMolecular dynamics

The high-resolution X-ray structure of the triple-mutant Kv1.2 channel (kv1.2-kv2.1-3m) was obtained from the Protein Data Bank (PDB code 7SIT) (Reddi et al., 2022). The channel structure was embedded in a (POPC) phospholipid bilayer, hydrated by a symmetric 150 mM KCl solution. Using the purpose-built Anton2 supercomputer (Shaw et al., 2014), the system was simulated with Desmond at constant temperature 300 K and pressure 1 atm, neutral pH and with applied TM electrostatic potential (Supplementary file 1). MD simulations were performed. Equations of motion were integrated using a time step of 2.5 fs and van der Waals interactions were truncated at 12 Å. Ionic currents were driven by application of a constant electric field E across the simulation box to mimic a voltage clamp experiment at the depolarized voltage of +200 mV (corresponding to a TM electric field of 0.043 kcal/mol/Å/e) (Roux, 2008). Simulations were performed with three distinct all-atom force fields: AMBER, CHARMM36m, and CHARMM36m-NBFIX. ff14SB version of the AMBER force field (Maier et al., 2015) was used in combination with ion parameters by Joung and Cheatham, 2008, CHARMM36m (Huang et al., 2017) was used with standard ion parameters and CHARMM36m-NBFIX was used with modified ion parameters in which ion-carbonyl interactions were made more attractive to mimic the Joung and Cheatham model. Specifically, all parameters of CHARMM36m are preserved in CHARMM36m-NBFIX except for three critical interactions. In CHARMM36m-NBFIX, the nonbonded pairwise ion-carbonyl, ion-water, and water-carbonyl Lennard-Jones parameters E_min and R_min are adjusted (NBFIX) to mimic the value of these three interactions in the AMBER force field. Supplementary file 2 shows the optimized energy E and distance R for the ion-carbonyl, ion-water, and water-carbonyl interactions in CHARMM36m-NBFIX. Water molecules were described by the TIP3P model (Jorgensen et al., 1983). Setup and analysis of the MD trajectories was performed in visual MD (VMD) (Humphrey et al., 1996).

Site-directed mutagenesis and RNA synthesis

Kv1.2-kv2.1 chimera (kindly provided by Eduardo Perozo) was cloned into pMax vector flanked by Xenopus β-globin sequence. Mutagenesis was performed utilizing the QuickChange techniques. All the clones were verified with full-length sequencing (Plasmidsaurus). DNA was linearized at the unique PmeI restriction site and then transcribed in vitro using T7 transcription kit (Ambion).

Channel expression in <italic>Xenopus</italic> oocytes and electrophysiology

Ovaries of Xenopus laevis were purchased from XENOPUS1. The follicular membrane was digested by collagenase type II (Worthington Biochemical Corporation) 2 mg/ml supplemented with bovine serum albumin 1 mg/ml. Oocytes were incubated in standard oocytes solution (SOS) containing in mM: 96 NaCl, 2 KCl, 1.8 CaCl₂, 1 MgCl₂, 0.1 EDTA, 10 HEPES, and pH set to 7.4 with NaOH. SOS was supplemented with 50 mg/ml gentamycin. Stage V-VI oocytes were then selected and microinjected with 50–150 ng of cRNA. Injected oocytes were maintained in SOS solution and kept at 18°C for 1–4 days prior to recordings. Ionic currents were recorded using the cut-open voltage-clamp technique (Stefani and Bezanilla, 1998). Voltage-measuring pipettes were pulled using a horizontal puller (P-87 Model, Sutter Instruments, Novato, CA, USA) with resistance between 0.3 and 0.8 MΩ were used to impale the oocytes. Currents were acquired by a setup comprising a Dagan CA-1B amplifier (Dagan, Minneapolis, MN, USA) with a built-in low-pass four-pole Bessel filter for a cutoff frequency of 20 kHz. Using a 16-bit A/D converter (USB-1604, Measurement Computing, Norton, MA, USA) for acquisition and controlled by an in-house software (GPatch64MC), data were sampled at 1 MHz, digitally filtered at Nyquist frequency and decimated for a storage acquisition rate of 100 kHz. Capacitive transient currents were compensated using a dedicated circuit. The voltage clamp was controlled by GPatch64MC and we used the USB-1604 16-bit as the D/A converter. Transient capacitive current was compensated by a dedicated circuit and in some cases, the transients were further minimized by an online P/N protocol holding at –80 mV (Armstrong and Bezanilla, 1973). All experiments were performed at room temperature (~17–18°C) in external solution containing: (in mM) 120 potassium methylsulfonate (KMES), 2 calcium hydroxide, 0.1 EDTA, and 10 HEPES, pH = 7.40 (with MES). Internal solution was composed by (in mM): 120 KMES, 10 HEPES, and 2 EGTA, pH = 7.40 (with MES). AgTxII was obtained from Alomone Labs and was titrated to 100 nM in the external solution prior to experiments. The current were elicited prior to the external application of the toxin. The blockage effects were assessed by series of 50 depolarizing pulses (from –120 to +60 mV) every 5 or 10 s. Between experiments, 1% albumin solution (in water) was used to clean the chamber and the bridges. All chemicals used were purchased from Sigma-Aldrich (St. Louis, MO, USA). GraphPad 9 (Prism) and in-house software (Analysis) were used to analyze the data.

Binding free energy of CTX

The high-resolution X-ray structures of kv1.2-kv2.1 (PDB code 2R9R) (Long et al., 2007) and kv1.2-kv2.1-3m (PDB code 7SIT) (Reddi et al., 2022) were used as molecular templates for modeling (Webb and Sali, 2014) the pore domain of the wild-type, double-mutant (S367T/V377T) and triple-mutant (W362F/S367T/V377T) constructs of the channel in the conductive and ‘dilated’ conformational states, respectively. Binding of the molecular structure of CTX (Bontems et al., 1992) to each of the channel constructs was investigated with HDOCK (Yan et al., 2020), according to the condition that Lys27 of CTX is in close proximity to the external entrance of the selectivity filter. The RMSD between docking poses and the X-ray bound configuration of CTX (Banerjee et al., 2013) was considered as the structural criterion (RMSD≤5 Å) to select docking solutions best reproducing the bound state of the toxin. At least 10 independent docking solutions were selected for computation of the net free-energy difference (ᐃᐃG) involved in the binding of CTX to each of the channel constructs and states.

ᐃᐃG was evaluated according to the continuous implicit solvent calculations of the PB-VDW model used in previous studies (Eriksson and Roux, 2002). The Poisson-Boltzmann (PB) solvation energy of the ligand-protein bound complex (ᐃG_LP) was calculated using the Adaptive Poisson-Boltzmann Solver 1.4.1 (APBS) (Baker et al., 2001) through a finite-difference scheme, by considering a 240 Å cubed box and a grid of 1.0×1.0×1.0 Å³. By representing explicitly the protein atoms without any charges, a dummy run was first carried out with APBS to generate dielectric, charge, and accessibility maps for the molecule in solution. Following the molecular surface definition, the internal dielectric constant of the protein was set to 15. The electrolyte solution was represented with a dielectric constant of 80 and salt concentration of 100 mM. These maps were then modified for the inclusion of a low-dielectric (∈=2) lipid surrogate. Input files and maps for APBS were generated with APBSmem (Callenberg et al., 2010). The van der Waals component of the bare electrostatic energy of the bound complex (ᐃE_LP) was computed with the CHARMM36m force field (Huang et al., 2017) by using the namdenergy plugin linked to VMD (Humphrey et al., 1996). The van der Waals component was scaled by an empirical factor (λ=0.17), intended to resolve the protein-solvent interaction, absent in the implicit solvent representation (Nandigrami et al., 2022). Calculations included the two experimentally resolved bound potassium ions at sites S₂/S₄ and S₃/S₄ of the conductive and ‘dilated’ conformation of the selectivity filter, respectively. Solvent accessible surface area and entropic contributions (Gilson et al., 1997) associated to the binding energy of CTX were assumed to be similar in both conformations of the channel and as such, they were not included in the calculation of the net binding free-energy difference ᐃᐃG.

Primary sequence analysis

Primary sequence logos conservation throughout the main-pore segments PH, SF, and S6 were generated with Weblogo3 (Crooks et al., 2004) by taking into consideration an HMMER3.0 (Finn et al., 2011) generated multiple sequence alignment of 657 unique UniProt sequences.

Energetics of C-type inactivation

The energetics of C-type inactivation was investigated by means of two PMFs. The first PMF reports the free-energy profile associated to the conformational transition of the selectivity filter between the conductive and dilated states, conditional to an open isoleucine gate. The second PMF reports the free-energy profile associated to closure of the isoleucine gate under the condition of a dilated conformation of the selectivity filter. Both conditions were imposed in the free-energy calculations via soft harmonic restraints of 0.5 kcal/mol/Å² respectively applied to α-carbon atoms of residue I398 and the selectivity filter. PMFs were determined employing the NAMD (Phillips et al., 2020) implementation of the well-tempered metadynamics extended adaptive biasing force algorithm (Fu et al., 2018; Fu et al., 2019), with the corrected z-averaged restraint estimator (Lesage et al., 2017). Calculations were respectively carried out with CHARMM36m (Huang et al., 2017) and the ff14SB version of the AMBER force field (Maier et al., 2015).

The free-energy profile associated to the conformational transition of the selectivity filter between the conductive and dilated states was computed with two path-collective variables (PCVs) (Branduardi et al., 2007), formed by 22 internal atomic distances. The first PCV, s, corresponds to the path connecting the two end states of the transformation, i.e., a string of discrete intermediate values inferred from an independent targeted MD simulation, whereas the second, orthogonal one, σ, represents the width of the tube embracing the path. The gradient of the free energy was measured along s, while a soft harmonic potential with a force constant of 5 kcal/mol Å² was applied on ѡ. No time-dependent bias was applied until a threshold of 50,000 samples was reached.

For the free-energy profile underlying the constriction of the pore domain, the collective variable (CV), d=d₁+d₂, was defined as the sum of two Euclidean distances separating the β-carbon atom of residue I398 of subunits KCH1 and KCH3, on the one hand, and of subunits KCH2 and KCH4, on the other hand. The reaction pathway, 18≤d≤36 Å, was discretized in bins 0.1 Å wide, wherein samples of the local force acting along the CV were accrued. To minimize nonequilibrium effects, no time-dependent bias was applied until a threshold of 10,000 samples was reached.

MD simulation of the MthK channel with CHARMM36m, AMBER, and CHARMM36m-NBFIX

Ion conduction through the canonical ‘conductive’ conformation of the selectivity filter K⁺ channels was examined for the well-established CHARMM36m and AMBER force fields, as well as the CHARMM36m-NBFIX modified force field. The simulations were carried out on the basis of the very accurate X-ray crystal structure of the MthK channel at 1.45 Å resolution (PDB code 3LDC) (Ye et al., 2010). For each force field, a 2 μs simulation was generated. In the simulated system, the channel structure was embedded in a (POPC) phospholipid bilayer, hydrated by a symmetric 400 mM KCl solution. The system was simulated in the NVT ensemble at constant temperature 320 K and pressure 1 atm, neutral pH, and with applied TM electrostatic potential (300 mV). Restraints were applied to the dihedral angels of the backbone of the selectivity filter with a flat-bottom restraint with a 0.159 kcal/mol/degree² force constant. The following angels were allowed to vary ±10 degrees from the crystal structure dihedral angels, viz. ±10 degrees from Thr59 (ϕ: 77, ψ: 9), Val60 (ϕ: –63, ψ: –45), Gly61 (ϕ: 48, ψ: 52), Tyr62 (ϕ: –52, ψ: –36), and Gly63 (ϕ: 84, ψ: 8). Beyond these ranges the restraints push the dihedrals back. In addition, to the selectivity restraints, harmonic distance restraints were applied on the level of the C-alpha distances of Pro19 and Phe97 to keep the inner gate open during the simulations, with a force constant of 2.39 kcal/mol/Å² and applied to both adjacent and opposing subunits, viz., Pro19: 25.5 Å for adjacent and 36 Å for opposing subunits, and Phe97: 26 Å for adjacent and 36.6 Å for opposing subunits. The CHARMM36m and CHARMM36m-NBFIX simulations were performed on OpenMM 7.7 (Eastman et al., 2017) for 2 µs with frames recorded every 100 ps. A real-space cutoff of 12 Å was used, and the potential was smoothly truncated at the cutoff with a switching function starting at 10 Å. The long-range electrostatics were treated with PME. The constant temperature was maintained with the Langevin dynamics thermostat with 1/ps friction coefficient. The AMBER simulation followed a similar route as the CHARMM36m simulations with the following changes. The simulation was performed with AMBER20 using the PMEMD GPU accelerated MD simulation engine. A real-space cutoff of 9 Å was used.

Our simulations at 300 mV indicate that the MthK structure is conductive at the level of the selectivity filter sites S₄ through S₁, with an average conduction rate per voltage of 48.24 pS in symmetric 400 mM KCl— significantly below the recorded single-channel conductance of MthK, i.e., ~170 pS in symmetric 150 mM KCl (Li et al., 2007) or ~96 pS in symmetric 200 mM KCl (Zadek and Nimigean, 2006). Force-field differences in the simulated conduction rate and mechanism of MthK resume are as follows: (1) conduction is accelerated in the AMBER force field via the hard-knock mechanism; (2) conduction is slower in the CHARMM36 force field via the hard-knock mechanism; and (3) conduction is also accelerated in the AMBER-like CHARMM-NBFIX force field via the hard-knock mechanism. Independently, these findings support that CHARMM36m-NBFIX with only modified ion-carbonyl, ion-water, and water-carbonyl interactions is able to reproduce the known AMBER-like ion conduction behavior, including the accelerated conduction rates and hard-knock mechanism.

Additional information

Competing interests

No competing interests declared

Author contributions

Conceptualization, Formal analysis, Investigation, Visualization, Methodology, Writing – original draft

Formal analysis, Methodology, Writing – original draft

Investigation, Methodology, Writing – original draft

performed new multi-microseconds simulations

Conceptualization, Methodology, Writing – original draft

Conceptualization, Formal analysis, Supervision, Methodology, Writing – original draft

Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Methodology, Writing – original draft, Project administration

Ethics

Electrophysiology using Xenopus oocytes, in compliance with protocol at University of Chicago.

Additional files

Supplementary file 1.Molecular dynamics (MD) simulations of triple-mutant channel kv1.2-kv2.1-3m.

Supplementary file 2.NBFixes for potassium, carbonyl, and water interactions.

Supplementary file 3.Number of conduction events along simulations of the open-conductive MthK.

Supplementary file 4.Number of conduction events along molecular dynamics (MD) simulations in which the isoleucine gate is open.

Supplementary file 5.Binding free-energy difference of charybdotoxin (CTX).

Supplementary file 6.Comparative analysis of the average properties of the isoleucine gate with and without mutation (I398N).

MDAR checklist

Data availability

All data considered in the study, including molecular configurations and scripts for MD simulations, MD trajectories, docking configurations and electrophysiology data, can be downloaded from the Zenodo repository https://doi.org/10.5281/zenodo.10938041.

The following dataset was generated:

WernerT2024Isoleucine gate blocks K+ conduction in C-type inactivationZenodo10.5281/zenodo.10938041

Acknowledgements

Helpful discussions with Eduardo Perozo and Leticia Stock are gratefully acknowledged. We thank Gethiely Gasparini for technical assistance with site-directed mutagenesis and RNA synthesis. Anton 2 computer time was provided by the Pittsburgh Supercomputing Center (PSC) through Grant MCB100018P from the National Institutes of Health. The Anton 2 machine at PSC was generously made available by DE Shaw Research. The work was supported by National Council for Scientific and Technological Development CNPq (WT grant number 302089/2019-5 and 200114/2020-4), by the National Institutes of Health Award R01GM030376 (FB) and R35-GM152124 (BR), National Science Foundation Award QuBBE QLCI (NSF OMA-2121044) (FB). BP is a PEW Latin American Fellow (2019).

References

Armstrong

Bezanilla

1973

Currents related to movement of the gating particles of the sodium channels

Nature242459461

10.1038/242459a0

4700900

Baker

Sept

Joseph

Holst

McCammon

2001

Electrostatics of nanosystems: application to microtubules and the ribosome

PNAS981003710041

10.1073/pnas.181342398

11517324

Banerjee

Lee

Campbell

Mackinnon

2013

Structure of a pore-blocking toxin in complex with a eukaryotic voltage-dependent K(+) channel

eLife2

e00594

10.7554/eLife.00594

23705070

Bernèche

Roux

2001

Energetics of ion conduction through the K+ channel

Nature4147377

10.1038/35102067

11689945

Bontems

Gilquin

Roumestand

Ménez

Toma

1992

Analysis of side-chain organization on a refined model of charybdotoxin: structural and functional implications

Biochemistry3177567764

10.1021/bi00149a003

1380828

Branduardi

Gervasio

Parrinello

2007

From A to B in free energy space

The Journal of Chemical Physics126

054103

10.1063/1.2432340

17302470

Callenberg

Choudhary

de Forest

Gohara

Baker

Grabe

2010

APBSmem: A graphical interface for electrostatic calculations at the membrane

PLOS ONE5

e12722

10.1371/journal.pone.0012722

20949122

Crooks

Hon

Chandonia

Brenner

2004

WebLogo: a sequence logo generator

Genome Research1411881190

10.1101/gr.849004

15173120

Cuello

Jogini

Cortes

Pan

Gagnon

Dalmas

Cordero-Morales

Chakrapani

Roux

Perozo

2010

Structural basis for the coupling between activation and inactivation gates in K(+) channels

Nature466272275

10.1038/nature09136

20613845

Cuello

Cortes

Perozo

2017

The gating cycle of a K+ channel at atomic resolution

eLife6

e28032

10.7554/eLife.28032

Doyle

Morais Cabral

Pfuetzner

Kuo

Gulbis

Cohen

Chait

MacKinnon

1998

The structure of the potassium channel: molecular basis of K+ conduction and selectivity

Science2806977

10.1126/science.280.5360.69

9525859

Eastman

Swails

Chodera

McGibbon

Zhao

Beauchamp

Wang

Simmonett

Harrigan

Stern

Wiewiora

Brooks

Pande

2017

OpenMM 7: Rapid development of high performance algorithms for molecular dynamics

PLOS Computational Biology13

e1005659

10.1371/journal.pcbi.1005659

28746339

Eriksson

MAL

Roux

2002

Modeling the structure of agitoxin in complex with the Shaker K+ channel: a computational approach based on experimental distance restraints extracted from thermodynamic mutant cycles

Biophysical Journal8325952609

10.1016/S0006-3495(02)75270-3

12414693

Finn

Clements

Eddy

2011

HMMER web server: interactive sequence similarity searching

Nucleic Acids Research39W29W37

10.1093/nar/gkr367

21593126

Zhang

Chen

Shao

Chipot

Cai

2018

Zooming across the free-energy landscape: Shaving barriers, and flooding valleys

The Journal of Physical Chemistry Letters947384745

10.1021/acs.jpclett.8b01994

30074802

Shao

Cai

Chipot

2019

Taming rugged free energy landscapes using an average force

Accounts of Chemical Research5232543264

10.1021/acs.accounts.9b00473

31680510

Gilson

Given

Bush

McCammon

1997

The statistical-thermodynamic basis for computation of binding affinities: a critical review

Biophysical Journal7210471069

10.1016/S0006-3495(97)78756-3

9138555

Holmgren

Smith

Yellen

1997

Trapping of organic blockers by closing of voltage-dependent K+ channels: evidence for a trap door mechanism of activation gating

The Journal of General Physiology109527535

10.1085/jgp.109.5.527

9154902

Hoshi

Zagotta

Aldrich

1990

Biophysical and molecular mechanisms of Shaker potassium channel inactivation

Science250533538

10.1126/science.2122519

2122519

Huang

Rauscher

Nawrocki

Ran

Feig

de Groot

Grubmüller

MacKerell

2017

CHARMM36m: an improved force field for folded and intrinsically disordered proteins

Nature Methods147173

10.1038/nmeth.4067

27819658

Humphrey

Dalke

Schulten

1996

VMD: Visual molecular dynamics

Journal of Molecular Graphics143338

10.1016/0263-7855(96)00018-5

Jensen

MØ

Jogini

Eastwood

Shaw

2013

Atomic-level simulation of current-voltage relationships in single-file ion channels

The Journal of General Physiology141619632

10.1085/jgp.201210820

23589581

Jorgensen

Chandrasekhar

Madura

Impey

Klein

1983

Comparison of simple potential functions for simulating liquid water

The Journal of Chemical Physics79926935

10.1063/1.445869

Joung

Cheatham

TEI

2008

Determination of alkali and halide monovalent ion parameters for use in explicitly solvated biomolecular simulations

The Journal of Physical Chemistry. B11290209041

10.1021/jp8001614

18593145

Kitaguchi

Sukhareva

Swartz

2004

Stabilizing the closed S6 gate in the Shaker Kv channel through modification of a hydrophobic seal

The Journal of General Physiology124319332

10.1085/jgp.200409098

15365093

Kopec

Rothberg

de Groot

2019

Molecular mechanism of a potassium channel gating through activation gate-selectivity filter coupling

Nature Communications10

5366

10.1038/s41467-019-13227-w

31772184

Köpfer

Song

Gruene

Sheldrick

Zachariae

de Groot

2014

Ion permeation in K

Science346352355

10.1126/science.1254840

25324389

Labro

Cortes

Tilegenova

Cuello

2018

Inverted allosteric coupling between activation and inactivation gates in K⁺ channels

PNAS11554265431

10.1073/pnas.1800559115

29735651

Lenaeus

Vamvouka

Focia

Gross

2005

Structural basis of TEA blockade in a model potassium channel

Nature Structural & Molecular Biology12454459

10.1038/nsmb929

15852022

Lesage

Lelièvre

Stoltz

Hénin

2017

Smoothed biasing forces yield unbiased free energies with the extended-system adaptive biasing force method

The Journal of Physical Chemistry. B12136763685

10.1021/acs.jpcb.6b10055

27959559

Berke

Chen

Jiang

2007

Gating and inward rectifying properties of the MthK K+ channel with and without the gating ring

The Journal of General Physiology129109120

10.1085/jgp.200609655

17261840

Liu

Bassetto

CAZ

Pinto

Bezanilla

2023

A mechanistic reinterpretation of fast inactivation in voltage-gated Na+ channels

Nature Communications14

5072

10.1038/s41467-023-40514-4

Long

Tao

Campbell

MacKinnon

2007

Atomic structure of a voltage-dependent K+ channel in a lipid membrane-like environment

Nature450376382

10.1038/nature06265

18004376

Maier

Martinez

Kasavajhala

Wickstrom

Hauser

Simmerling

2015

ff14SB: Improving the accuracy of protein side chain and backbone parameters from ff99SB

Journal of Chemical Theory and Computation1136963713

10.1021/acs.jctc.5b00255

Nandigrami

Szczepaniak

Boughter

Dehez

Chipot

Roux

2022

Computational assessment of protein-protein binding specificity within a family of synaptic surface receptors

The Journal of Physical Chemistry. B12675107527

10.1021/acs.jpcb.2c02173

35787023

Olcese

Sigg

Latorre

Bezanilla

Stefani

2001

A conducting state with properties of A slow inactivated state in A shaker K(+) channel mutant

The Journal of General Physiology117149163

10.1085/jgp.117.2.149

11158167

Ostmeyer

Chakrapani

Pan

Perozo

Roux

2013

Recovery from slow inactivation in K+ channels is controlled by water molecules

Nature501121124

10.1038/nature12395

23892782

Panyi

Deutsch

2007

Probing the cavity of the slow inactivated conformation of shaker potassium channels

The Journal of General Physiology129403418

10.1085/jgp.200709758

17438120

Perozo

MacKinnon

Bezanilla

Stefani

1993

Gating currents from a nonconducting mutant reveal open-closed conformations in Shaker K+ channels

Neuron11353358

10.1016/0896-6273(93)90190-3

8352943

Peters

Fedida

Accili

2013

Allosteric coupling of the inner activation gate to the outer pore of a potassium channel

Scientific Reports3

3025

10.1038/srep03025

24149575

Phillips

Hardy

Maia

JDC

Stone

Ribeiro

Bernardi

Buch

Fiorin

Hénin

Jiang

McGreevy

Melo

MCR

Radak

Skeel

Singharoy

Wang

Roux

Aksimentiev

Luthey-Schulten

Kalé

Schulten

Chipot

Tajkhorshid

2020

Scalable molecular dynamics on CPU and GPU architectures with NAMD

The Journal of Chemical Physics153

044130

10.1063/5.0014475

32752662

Reddi

Matulef

Riederer

Whorton

Valiyaveetil

2022

Structural basis for C-type inactivation in a Shaker family voltage-gated K⁺ channel

Science Advances8

eabm8804

10.1126/sciadv.abm8804

35452285

Roux

2008

The membrane potential and its representation by a constant electric field in computer simulations

Biophysical Journal9542054216

10.1529/biophysj.108.136499

18641071

Schewe

Sun

Mert

Mackenzie

Pike

ACW

Schulz

Constantin

Vowinkel

Conrad

Kiper

Gonzalez

Musinszki

Tegtmeier

Pryde

Belabed

Nazare

de Groot

Decher

Fakler

Carpenter

Tucker

Baukrowitz

2019

A pharmacological master key mechanism that unlocks the selectivity filter gate in K⁺ channels

Science363875880

10.1126/science.aav0569

30792303

Selvakumar

Fernández-Mariño

Khanra

Paquette

Wang

Huang

Smider

Rice

Swartz

Meyerson

2022

Structures of the T cell potassium channel Kv1.3 with immunoglobulin modulators

Nature Communications13

3854

10.1038/s41467-022-31285-5

35788586

Shaw

Grossman

Bank

Batson

Butts

Chao

Deneroff

Dror

Even

Fenton

Forte

Gagliardo

Gill

Greskamp

Ierardi

Iserovich

Kuskin

Larson

Layman

Lee

Lerer

Killebrew

Mackenzie

Mok

SYH

Moraes

Mueller

Nociolo

Peticolas

Quan

Ramot

Salmon

Scarpazza

Schafer

Siddique

Snyder

2014

Anton 2: Raising the Bar for Performance and Programmability in a Special-Purpose Molecular Dynamics Supercomputer” in SC ’14

Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis

4153

10.1109/SC.2014.9

Souza

Amaral

Treptow

2014

Electric fingerprint of voltage sensor domains

PNAS1111751017515

10.1073/pnas.1413971111

25422443

Starkus

Kuschel

Rayner

Heinemann

1997

Ion conduction through C-type inactivated Shaker channels

The Journal of General Physiology110539550

10.1085/jgp.110.5.539

9348326

Stefani

Bezanilla

1998

Cut-open oocyte voltage-clamp technique

Methods in Enzymology293300318

10.1016/s0076-6879(98)93020-8

9711615

Stix

Tan

Bae

Fernández-Mariño

Swartz

Faraldo-Gómez

2023

Eukaryotic Kv channel Shaker inactivates through selectivity filter dilation rather than collapse

Science Advances9

eadj5539

10.1126/sciadv.adj5539

38064553

Takacs

Toups

Kollewe

Johnson

Cuello

Driessens

Biancalana

Koide

Ponte

Perozo

Gajewski

Suarez-Kurtz

Koide

Goldstein

SAN

2009

A designer ligand specific for Kv1.3 channels from A scorpion neurotoxin-based library

PNAS1062221122216

10.1073/pnas.0910123106

20007782

Tan

Bae

Stix

Fernández-Mariño

Huffer

Chang

Jiang

Faraldo-Gómez

Swartz

2022

Structure of the Shaker Kv channel and mechanism of slow C-type inactivation

Science Advances8

eabm7814

10.1126/sciadv.abm7814

35302848

Treptow

Tarek

2006

Molecular restraints in the permeation pathway of ion channels

Biophysical Journal91L26L28

10.1529/biophysj.106.087437

16751240

Wang

MacKinnon

2017

Cryo-EM structure of the open human Ether-à-go-go-related K+ channel hERG

Cell169422430

10.1016/j.cell.2017.03.048

Webb

Sali

2014

Comparative protein structure modeling using MODELLER

Current Protocols in Bioinformatics47

10.1002/0471250953.bi0506s47

25199792

Yan

Tao

Huang

2020

The HDOCK server for integrated protein-protein docking

Nature Protocols1518291852

10.1038/s41596-020-0312-x

32269383

Yang

Yan

Sigworth

1997

How does the W434F mutation block current in Shaker potassium channels?

The Journal of General Physiology109779789

10.1085/jgp.109.6.779

9222903

Yangyu

Yangyang

Youshan

Shumin

Alberto

Ken

Fred

2024

Cryo-EM structures of Kv1.2 potassium channels, conducting and non-conducting

eLife12

e89459

10.7554/eLife.89459

Jiang

2010

Novel insights into K+ selectivity from high-resolution structures of an open K+ channel pore

Nature Structural & Molecular Biology1710191023

10.1038/nsmb.1865

20676101

Zadek

Nimigean

2006

Calcium-dependent gating of MthK, a prokaryotic potassium channel

The Journal of General Physiology127673685

10.1085/jgp.200609534

16735753

Zhou

Morais-Cabral

Kaufman

MacKinnon

2001

Chemistry of ion coordination and hydration revealed by a K+ channel-FAB complex at 2.0 A resolution

Nature4144348

10.1038/35102009

11689936

10.7554/eLife.97696.sa0

Editor's evaluation

Islas

Leon D

Reviewing Editor

https://ror.org/01tmp8f25

Universidad Nacional Autónoma de México

Mexico

This manuscript addresses the molecular mechanism of C-type inactivation observed in a mutant of the Kv2.1-1.2 (Shaker-like) chimeric voltage-gated potassium channel. Previous structural studies using a triple mutant of this channel, which enhance slow inactivation, have demonstrated that inactivation involves a dilation at the outer mouth of the selectivity filter of the channel, leading to a non-conductive state. Here, based on solid molecular dynamics simulations, corroborated by electrophysiological experiments, the authors conclude that the dilated state on its own is conductive, and that an additional conformational change involving occlusion of the pore by I398 is critical to halt conduction. This important conclusion is thought-provoking and motivates further exploration to evaluate pore dilation and I398 in other Kv channels.

10.7554/eLife.97696.sa1

Decision letter

Islas

Leon D

Reviewing Editor

https://ror.org/01tmp8f25

Universidad Nacional Autónoma de México

Mexico

Carnevale

Vincenzo

Reviewer

https://ror.org/00kx1jb78

Temple University

United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

Thank you for submitting your article "Isoleucine gate blocks K⁺ conduction in C-type inactivation" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Merritt Maduke as the Senior Editor.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

During collaborative discussion, three reviewers and the reviewing editor identified the following main concerns that the authors should respond to.

1) It is well established that K-selective channels support potassium permeation through tight coordination by the selectivity filter. It is not clear how a dilated pore might allow selective permeation of this ion. The authors need to clarify possible mechanisms regardless of the MD simulation results.

2) The main conclusion of this manuscript is based on the results of long MD simulations with three force fields. Two of the force fields give inconsistent results and this is fixed by simulations with an ad-hoc correction of the CHARM36 force field. There are no simulations carried out with the corrected CHARMM36m-NBFIX force field that indicate that it continues to reproduce known behavior in WT channels. Authors suggest that the difference in simulation results might be due to the application of harmonic restrains in previous simulations. It should be clarified with simulations if the key differences in results are due to properties of the force fields or the setup of the simulation system.

3) In the AMBER simulation presented in Figure 2, permeation cessation does not seem to be correlated with the movement of the I398 as stated in the text. Please clarify.

4) It is stated that in the AMBER simulations, the distance between β carbons of I398 residues is reduced to 9 angstroms and this leads to a less than 2 angstroms constriction. What are the expected dimensions for the I398N mutant channel? Do simulations of this mutant channel show a WT-like permeation behavior or a dilated selectivity filter.

5) The authors claim that the simulations indicate a similar rate of potassium flow for the kv1.2/2.1-3m channels in the dilated state as for WT Shaker channels. However, it has been known for a while that Shaker W434F channels, which are thought to be permanently c-type inactivated, still allow permeation at the same rate as WT, ~13 pS, just with extremely low open probabilities and very short duration (Yang, Yan and Sigworth, 1997). The simulations presented here, and elsewhere, seem to suggest that the permeability of the slow inactivated state(s) is just significantly reduced. This important discrepancy needs clarification.

6) The 3m channel and the 3m-I398N mutant seem to activate are much more positive voltages than the kv1.2/2.1 channel (Figure 3 E) however in the comparison of time courses (Figure 3d) the I398N mutant activates way faster at the same voltage, this seems to be inconsistent. Also, the voltage pulses employed are too short. Kv1.2 channels slow-inactivate over a time course of seconds. It is possible that the I398N mutant still inactivates over seconds. In fact, Figure 3c shows an indication of slow inactivation as compared with kv1.2/2.1 channels. Given that the I398N mutation activates at more positive voltages, inactivation should be assessed at voltages that saturate the open probability for each channel.

Reviewer #1:

In the present manuscript, Treptow, Liu, Bassetto Jr and colleagues propose a novel mechanism for how C-type inactivation diminishes ion conduction in voltage-gated potassium (Kv) channels. C-type inactivation is a time-dependent mechanism that manifests as a decrease in the ionic current within the second timescale until it reaches a steady-state with minimal conductance. Although the mechanism of C-type inactivation was originally studied in the Shaker Kv channel (Hoshi et al. 1991 Neuron), the structural basis for a related mechanism of slow inactivation was first studied in the KcsA channel (Chakrapani et al. 2007 JGP, Cordero-Morales et al. 2006 NSMB, Imai et al. 2010 PNAS, Kim et al. 2016 JGP, Maffeo et al. 2012 Chem.Rev., Piasta et al. 2011 JGP, Tilegenova et al. 2017 PNAS, Varga et al. 2007 Biochim Biophys Acta, Cuello et al. 2010 Nature). A consensus based on the large amount of experimental data collected on KcsA pointed to the selectivity filter (SF) as the responsible part of the channel for C-type inactivation, which was proposed to collapse during inactivation. This collapse of the filter was the accepted working model prior to recent cryo-EM structures of Kv channels in C-type inactivated states showing that the filter dilates during inactivation (Reddi et al. 2022 Sci.Adv, Tan et al. 2022 Sci Adv, Selvakumar et al. 2022 Nat Comm, Wu Y et al. 2024 pre-print, Stix et al. 2023 Sci Adv,). As the mechanism of selective K⁺ permeation at high rates has long been established to result from multiple occupancy of the SF (see below), dilation of the filter to remove the outer K⁺ binding sites would be expected to diminish ion permeation during inactivation, an expectation borne out in Molecular Dynamics simulations (Tan et al. 2022 Sci Adv, Stix et al., 2023Sci Adv). The mechanism proposed by the authors in the present paper, however, challenges our current understanding of the general mechanism for C-type inactivation in Kv1 channels by proposing a new residue, an Ile below the SF as the 'true' gate that impedes ion conduction in C-type inactivated channels. The proposed novel mechanism arises from an MD simulation using the AMBER forcefield and the cryo-EM structure of the Kv1.2-2.1 chimera with 3-point mutations (3m). The 3m mutation (W362F, S367T and V377T) aims to render the channel in a non-conducting C-type inactivated state by speeding its inactivation in a fashion similar to W434F in Shaker (Perozo et al. 1993 Neuron, Yang et al. 1997 JGP) but failing to do so, the Kv1.2/2.1-3m channel is still able to conduct ions; a transient, fast inactivating macroscopic current can be seen upon depolarization. The reason for this is that the chimeric channel contains the sequence of Kv1.2 in the SF, which is known to be particularly resistant to inactivation and requiring more than one mutation to achieve a fast-inactivating phenotype (Suarez-Delgado et al. 2020 JGP, Wu et al. 2022 JGP, Reddi et al. 2022 Sci.Advances).

The fact that the authors chose this particular chimeric channel with the 3 point mutations (Kv1.2/2.1-3m) to study the mechanism of ion permeation during C-type inactivation is an odd choice given that block of ion permeation is incomplete in this mutant (Figure 3B) (Reddi et al. 2022 Sci.Advances). Moreover, the data presented in this manuscript do not support the authors' conclusions.

Strengths:

The new idea presented by the authors is provocative and both MD simulations and electrophysiological techniques are appropriate to explore the mechanism of C-type inactivation.

Weaknesses:

1) The authors seem confused why a dilated filter would be less conductive than one containing 4 ion binding sites. Although details about the mechanism of K⁺ permeation across the SF remain incompletely understood (including whether there is 'hard' or 'soft knock-on' between ions to promote permeation), it would seem to be established that K⁺ channels are exquisitely K⁺ selective because the backbone carbonyls replace waters of hydration and that multiple ions bound within the filter repulse each other to promote rapid throughput (Doyle et al. 1998 and Zhou et al. 2001, Morais-Cabral et al. 2001, Zhou and Mackinnon 2003).

2) The authors proposal that I398 is the gate for C-type inactivation is inconsistent with over multiple structures of KcsA and Kv channels with varying propensities to inactivate because that residue never occludes ion permeation in any of those structures, solved by both X-ray crystallography or cryo-EM. The bar should be high for overturning the weight of evidence that KcsA collapses during inactivation or that Kv channels dilate, and would logically require new structures to support the key conclusions in this study.

3) The MD simulations and the permeation events.

The mechanism proposed by the authors arises from the sole finding of Ile398 twisting during MD simulations produced with the Kv1.2/2.1-3m structure and the AMBER forcefield. The authors propose that the twisting of the Ile creates a gate under the SF that blocks ion permeation during C-type inactivation. This MD result was observed using the unrestrained structure and the AMBER forcefield is the hypothesis generator and the only condition where the authors see this conformational change (Figure 2B). The MD simulations using the CHARMM36 forcefield, on the contrary, show permeation events for half the 10 µs simulation (Figure 2C) and no twisting on the I398. This discrepancy is presented by the authors as a property of the different forcefields used, so they use CHARMM36m-NBFIX to approximate the CHARMM36m force field parameters to AMBER, the permeation events reduce, however, the Ile does not flip in this simulation. These inconsistencies are problematic and not adequately justified by the authors.

In addition, to contextualize these permeation events, it is necessary to see how these simulations, with the exact same conditions and force fields would describe ion permeation for the WT channel in a conducting conformation. The WT MD simulations would likely show many more permeation events without flipping of the I398. This direct comparison will help understand which simulation/force field is more representative of the functional state of the channel and put in context how conductive/nonconductive the Kv1.2/2.1-3m channel is. It seems likely that the present results would be qualitatively consistent with simulations of Shaker performed while constraining the structure and showing that dilation diminishes ion permeation The identity of the I398 as a gate

If the I398 residue is a gate and the residue responsible for diminishing ionic flow in the C-type inactivated state, it is rather curious that it has never been seen before given how many simulations have been run on KcsA and Kv channels. Can the authors provide a rationale for supporting their conclusions in light of what has already been done? How many times did the authors observe the conformational change of this residue relative to the amount of MD runs? If the authors constrain the structure as in previous simulations on Shaker, would I398 no longer adopt a conformation that blocks ion permeation? Also, the results in Figure 2 and FigS2 using the AMBER forcefield, seem to disagree. When the authors repeat the MD simulations using the AMBER forcefield restricting the I398 movement and making it 'permeable' the permeation stops after 2 µs with a long residency of K⁺ ions in the pore. This result alone would seem to challenge the authors hypothesis and clearly suggests that the twist of the I398 is not required to stop ion permeation events because the dilated structure alone seems to be doing that. The CHARMM36m-NBFIX simulations show only 3 permeation events during the whole simulation. How do the authors reconcile these results with their conclusions?

4) Functional consequences of the I398N mutation

It is known that mutations at I398 have a strong functional effect in other channels like Shaker (I470) or KcsA (F103), where previous studies have provided support for a key role of this residue in coupling opening of the inner gate with conformational changes in the SF during inactivation. Those studies are considerably more detailed than the present functional studies and would seem to be inconsistent with I398 functioning as a gate. Val substitutes well for Ile, both Cys and Phe are slower and Leu is faster, and no mutations completely disrupt C-type inactivation (Holmgren et al. 1997, Peter CJ et al. Sci Reports 2013, Cuello et al. 2010 Nature 466 203-8, Cuello et al.2010 Nature 466 272-5). How can the authors reconcile their new ideas with these earlier studies and mechanistic ideas about the role of I398? In addition, introducing an Asn, introduces a polar side chain in a hydrophobic region, a radical change that can affect more than just the C-type inactivation of the channel. In order to understand the effect of I398N it is necessary to study the effect of that mutant in isolation without the 3m mutations since it is a new mutation in the context of Kv1.2/2.1. For instance, does it also impair C-type inactivation in the absence of the 3m mutations? Does it affect the permeation of potassium? Does it shift the G-V curve? Without this information it's not possible to fully understand the results presented in Figure 3. Might introducing a polar residue in this region impact ion binding within the SF? Might the mutation alter inactivation by increasing the affinity of K⁺ for the filter? Finally, the MD simulation results with an Asn at position 398 disagree with the proposed mechanisms and the functional data. The electrophysiological experiments show a channel that conducts ions (Figure 3C) however the MD simulations using the AMBER forcefield do not, and the CHARMM36m only shows permeation events for 3 µs out of 10. How do the authors explain these results?

5) Toxin binding to the outer pore

The results presented with AgTxII seem quite preliminary and it's hard to understand how they support the proposal of I398 functioning as a gate. Only a few traces are shown at one toxin concentration rather than time courses to demonstrate that equilibrium has been achieved. It seems that the 3m mutations are somehow altering toxin binding regardless of whether the I398N mutations rescue ion conduction, but how this supports I398 functioning as a gate is unclear. How do the AgTxII results relate to the model proposed for CTX in Fig3F? Are the authors proposing that the outer pore of the SF changes its architecture when the I398N mutation is introduced?

Recommendations for the authors

1) The statement "These simulations strongly suggest that the filter in the dilated conformation can conduct K⁺ ions, and that the conformational motion of I398 is necessary to truly block conduction." As well as "Despite intrinsic force-field differences with respect to channel conductivity, all three atomistic models support that the dilated conformation of the selectivity filter is, by itself, conductive and the isoleucine gate is required to effectively block K⁺ current across the channel." are not fully supported by the MD results since restricting the conformational motion of I398 also blocks conductions as shown in S2, please review these statements.

2) "Gating fluctuations of I398 are, however, clearly observed in the late stages of the simulation and, correlate well with the reduction of ions in the central cavity of the channel and with the conduction across the selectivity filter. Particularly important, the CHARMM36m simulation adds support to the assumption that the dilated conformation of the selectivity filter is conductive, and that closure of the isoleucine gate is required to shut down ion transport across the channel." Please reference these results in the figure, the CHARMM36m forcefield did not show the Ile changing its conformation in the Figures presented.

3) In observance of the result "the estimate of ~0.2pA is still orders of magnitude larger than the measured current in the triple-mutant channel upon C-type inactivation (vide infra), and, therefore, the conductivity properties of the "dilated" conformation of the selectivity filter cannot explain alone the inactivation of kv1.2-kv2.1-3m under membrane depolarization." There is another possible explanation for this discrepancy related to the configuration of the MD simulations. Taken together the variability of the results restraining vs non-restraining the I398, there is a possibility that the calculations obtained from the MD simulations are not representing the C-type inactivated state of the Kv1.2/2.1-3m triple mutant.

4) "Recently, high-resolution structures of Kv channels revealed a novel conformation of the selectivity filter that is partially dilated at its outer end and constricted near its internal face (8-10)". The internal face of the SF architecture of all the cited structures (S3 and S4) can still solve densities for coordinated K ions in their internal face and as stated in the legend of Figure 1 resembles that on the conductive state, arguing against the constricted conformation stated by the authors please review.

5) This statement in the Abstract could be misleading for the reader, if I understand it correctly, it reads as if the electrophysiology measurements demonstrate that the Kv1.2-2.1-3m mutant is conducting, but then is stated that functional experiments show inactivation, please review: "While the experimental structure was interpreted as the elusive non-conductive state, molecular dynamics simulations and electrophysiology measurements demonstrate that the dilated filter of kv1.2-kv2.1-3m, however, is conductive and, as such, cannot completely account for the inactivation of the channel observed in functional experiments".

Reviewer #2:

Based on computational analysis of structures of the conductive WT Shaker B and Kv1.2-2.1 chimera and the pore-dilated Shaker-W434F and triple-mutant Kv1.2-2.1 chimera channels, the authors hypothesize that pore-dilation alone cannot account for the non-conductive tendency of these channels in the C-type inactivated state. The authors then go on to analyze the Kv1.2-2.1 triple mutant (kv1.2-kv2.1-3m) by simulation with AMBER and CHARMM36m force fields, and find that under conditions where the pore-lining residue I398 is allowed to relax, the I398 side chains from all four subunits rapidly twist to occlude K⁺ conduction, whereas K⁺ conduction is maintained under conditions where I398 does not occlude the pore, as in the kv1.2-kv2.1-3m crystal structure.

To validate the role of I398 in controlling conduction in pore-dilated channels, the authors introduce the mutation I398N in kv1.2-kv2.1-3m channels and find that the substitution with the hydrophilic asparagine residue effectively abrogates C-type inactivation behavior. The addition of I398N does not appear to act by preventing the pore-dilated conformation, as Agitoxin-II, which strongly blocks the open-conducting but not pore-dilated channels, does not block the kv1.2-kv2.1-3m-I398N channels.

The manuscript follows a logical series of experiments and thoughtful, rigorous analysis. The mechanism presented is supported by computational and electrophysiological data, and underscores a potential role for conformational changes in pore-lining residues in inactivation that may occur in other K⁺ channels.

Results presented in the manuscript make the strong prediction that an asparagine at position 398 should not occlude the pore in the triple-mutant background, and should stabilize conduction even when the pore is "dilated". It should be possible to show this directly with a simulation, and I think such a demonstration would greatly strengthen the manuscript.

Reviewer #3:

This manuscript reports on an investigation by Treptow et al. of C-type inactivation in voltage-gated potassium (Kv) channels, a process whereby prolonged voltage activation leads to a nonconductive state. They examined a triple-mutant Kv1.2-Kv2.1 channel to provide a detailed characterization of the dilated conformation of the selectivity filter. This structure was initially thought to represent the nonconductive state. However, molecular dynamics simulations and electrophysiology showed that this dilated state is actually conductive. The study found that effective inactivation involves an additional conformational change at isoleucine residues (I398) in the pore-lining segment S6, which acts as a hydrophobic gate just below the selectivity filter. This mechanism is critical for C-type inactivation and presents new targets for drug development to modulate Kv channel gating states. This work constitutes a significantly novel contribution to our understanding of the mechanism of conduction of potassium ions. As such, I strongly recommend publication.

I have a few comments that the authors could consider for improving the presentation of their results:

1) The conformational free energy landscape is a crucial piece of the story, which gives quantitative substance to the hypotheses tested throughout the work. However, the description of these results appears surprisingly only in the Discussion section as an afterthought. The authors should make an effort to incorporate these results in the body of the results.

2) Related to the previous point is a general lack of details concerning these calculations. For instance, it is imperative to have an idea of the error associated to the estimated free energies.

3) The differential affinity of the toxin for the two selectivity filter conformations is another crucial piece of the puzzle as it enables to unambiguously interpret the effect of the mutation of isoleucine into asparagine. However, the docking and binding affinity calculations are buried in the supplementary information. The authors should consider giving greater space and emphasis to these results in the main text

4) I am intrigued by the massively different behavior shown by the charmm force-field. What is the reason for this? I wonder if the greater stability of the hydrated configuration of the isoleucine side chain is an artifact due to the water model (tip3p does not reproduce water's surface tension, so wetting/dewetting transitions are not expected to be correctly described).

10.7554/eLife.97696.sa2

Author response

Essential revisions:

During collaborative discussion, three reviewers and the reviewing editor identified the following main concerns that the authors should respond to.

The net conduction through the “dilated” conformation of the channel is greatly affected by the I398 gate that is located along the TM6 segment below the filter on the intracellular side. When the I398 gate is open, conduction can occur through the dilated filter. It is only when the I398 gate is closed that there is zero conduction. We have not estimated the selectivity of ion conduction through the dilated filter, but it seems likely that the selectivity for K⁺ over Na⁺ would be affected. Supporting that notion, AMBER and CHARMM36m simulations of kv1.2-kv2.1-3m in presence of 150mM NaCl show indeed that one or two Na⁺ ions can stably bind the dilated conformation of the selectivity filter at sites S3, S4 and Sext (Figure 4—figure supplement 3). Furthermore, Jiang and co-workers have shown that the multi-ion occupancy nature of the filter affects selectivity in K⁺ channels. By examining the properties of MthK and NaK mutants, they showed that the channel becomes selective only if there are four consecutive binding sites along the filter [S. Ye, Y. Li and Y. Jiang. Novel insights into K⁺ selectivity from high-resolution structures of an open K⁺ channel pore. Nature structural & molecular biology. 17(8):1019-23 (2010); M.G. Derebe, D.B. Sauer, W. Zeng, A. Alam, N. Shi and Y. Jiang. Tuning the ion selectivity of tetrameric cation channels by changing the number of ion binding sites. Proceedings of the National Academy of Sciences of the United States of America. 108(2):598-602 (2011); D.B. Sauer, W. Zeng, S. Raghunathan and Y. Jiang. Protein interactions central to stabilizing the K⁺ channel selectivity filter in a four-sited configuration for selective K⁺ permeation. Proceedings of the National Academy of Sciences of the United States of America. 108(40):16634-9 (2011)].

First, we would like to emphasize that the experimental validation of the computational observations using site-directed mutagenesis and electrophysiology is a critical part of the present study. Results from long MD simulations with different force fields were used to generate mechanistic hypotheses that were subsequently tested with experiments. The experiments would not have been carried out without the information generated by the simulations.

In summary, the main points about ion conduction from the MD simulations are:

1) Ion conduction through the open-dilated conformation of kv1.2-kv2.1-3m is possible in the standard CHARMM36m force field because the ion does not bind strongly to the sites S4 and S3. Multiple conduction events across the open-dilated conformation of the triple-mutant chimera channel were observed in the simulation with the CHARMM36m force field, with no applied restraints.

2) Ion conduction through the open-dilated conformation of kv1.2-kv2.1-3m is possible but slow in the AMBER force field because the ion binds strongly to the selectivity filter sites S4 and S3;

3) Ion conduction through the open-dilated conformation of kv1.2-kv2.1-3m is also possible but slow in the CHARMM36m-NBFIX force field because it has AMBER-ized ion interactions, and the ion binds also strongly in S4 and S3. These results clarify the observations from points #1 and #2.

The observation of a conductive filter in the dilated conformation with the CHARMM36m force field (#1 above) is in sharp contrast to previously reported simulations of Shaker B in which the same force field did not display any ion conduction at +300mV (Tan et al. 2022 and Stix et al. 2023). This previous simulation was carried out with dihedral restraints to preclude structural distortions of the channel; those restraints apparently rendered the open-dilated conformation non-conductive. Because this is the same structure of Shaker B and the same CHARMM36m force field, one can conclude with confidence that the key difference in the results is caused by the restraints applied during the simulation. This aspect was clarified in the revised manuscript.

While it is true that “Two of the force fields give inconsistent results” (point #1 and #2 above), this statement gives only a partial account of reality. All force fields are approximate and imperfect. We exploit the simulations to attract our attention to the molecular factors that are of functional relevance with respect to C-type inactivation in K⁺ channels. The ultimate goal is to get a better understanding of this process. Sometimes MD simulations give results that are in quantitative agreement with experiments, and sometimes they don’t. For example, it is well known that ion conduction calculated from MD through the selectivity filter in the canonical “conductive” conformation (as in 1K4C) tends to be too small compared to experimental values, especially at physiological voltages (less 50 mV). This is true for all force fields. With the AMBER force field, there is observable conduction via the hard-knock mechanism but at fairly high voltages (200-300 mV). The reasons for this situation are unknown, likely involving complex polarization effects that are not included in any of the classical force fields used in MD simulations (Jensen et al., 2013). So, simulated conductances are generally too small with current force fields, and the present work is not intended to devise an improved force field to simulate ion conduction through Kv channels. Nonetheless, despite the limitations of current force fields, it is possible to use them to gain insight by comparing the conductance of different conformations of the selectivity filter. For instance, using the AMBER force field or the CHARMM36m-NBFIX force field with AMBER-ized ion interactions (point #3 above), the rate of conduction through the “dilated” filter is not significantly smaller than that of the “conductive” filter. Based on this observation, it is reasonable to propose that the “dilated” conformation should indeed be conductive. This isn’t surprising. After all, there is no obvious physical reason supporting the notion that the dilated conformation should be non-conductive (there is no occlusion or constriction of the pre lumen).

In this context, the purpose and significance of the CHARMM36m-NBFIX force field seems to have been misunderstood. We are sorry for the confusion. It is incorrect to think of this as an effort to “correct” the CHARMM36m force field. The purpose of CHARMM36m-NBFIX is not to fix or improve CHARMM36m. Instead, this force field is only meant to serve as a tool to help reveal how a few key interactions (ion-carbonyl, ion-water, water-carbonyl) are directly responsible for the main observed difference in ion conduction between the AMBER and CHARMM36m force fields. Specifically, all parameters of CHARMM36m are kept unchanged in CHARMM36m-NBFIX except for these 3 key interactions, which are modified to mimic the values from the AMBER force field. The changes are fairly small, at most ~2.0 kcal/mol. In particular, the interaction of K⁺ with the backbone carbonyl group in CHARMM36m-NBFIX is slightly stronger than in CHARMM36m by about -1.8 kcal/mol, to match the AMBER force field. But with these simple modifications, the ion conduction properties of K⁺ through the conductive selectivity filter from CHARMM36m-NBFIX essentially recapitulate that from AMBER. Supplementary file 2 was revised to show the minimized energy E and distance R for these specific interactions in the CHARMM36m-NBFIX force field, with the fine-tuned Lennard-Jones (LJ) parameters Emin and Rmin.

Regarding the dynamical fluctuations of the I398 gate, our observations from MD are that the gate can close with the AMBER force field, but less so with the standard CHARMM36m force field. Again, this shows that force fields are approximate and imperfect. We suspect that the reasons for this situation are likely to involve subttle differences in the backbone and side chain torsion potentials. Here, in contrast to ion conduction, we have not been able to pin-point the origin of the difference between the AMBER and CHARMM36m force fields that is directly responsible for these observations by creating a slightly modified CHARMM36m force field that would replicate the AMBER behavior. However, to resolve these inconsistencies and make the most of the imperfect information provided by the simulations, the functional importance of the residue at position 398 was then verified experimentally through site-directed mutagenesis and electrophysiology. The experiments supersede the uncertainty from the MD simulations. What matters at the end is the genuine knowledge remaining when considering the totality of the work.

In summary, our MD simulation study explores the conduction of K⁺ across the dilated conformation of the selectivity filter of the triple-mutant chimera channel kv1.2-kv2.1-3m under two distinct configurations of the isoleucine gate I398: (i) conductive (open) or (ii) non-conductive (closed). Whenever the isoleucine gate is non-conductive (closed), there is no conduction of ions across the dilated conformation of the selectivity filter. Main-text Figure 2 provides us with most compelling structural evidence for this conclusion. In contrast, we do observe conduction of ions across the dilated conformation of the selectivity filter whenever the isoleucine gate is conductive (open). Ion conduction follows a hard-knock mechanism at the level of the selectivity filter binding sites S4 and S3 (cf. Figure 2—figure supplement 1), with an average conduction rate per voltage of ~1.17pS at 150mM KCl (cf. Supplementary file 4) that compares well to the simulated conductance of the conductive state ~3.5pS at 300mM KCl (Stix et al. 2023). Based on these findings, two consensual conclusions were drawn from the simulations: (1) the dilated conformation of the selectivity filter is, by itself, conductive and (2) the I398 gate is required to effectively block K⁺ current across the triple-mutant channel (cf. Figure 2—figure supplement 1B).

Additional information about ion conduction with the CHARMM36m-NBFIX force field

As requested by the reviewers, additional information was provided about ion conduction with the CHARMM36m-NBFIX. We focused on the conductive state of K⁺ channels and compared to CHARMM36m and AMBER. For each force field, an additional simulation of 2μs was performed with the MthK channel. This K⁺ channel is chosen because a very high 1.45Å resolution x-ray structure (PDB code 3LDC) is available. The filter is in the canonical “conductive” conformation as in the 1K4C structure of the KcsA channel. The MthK channel was embedded in a (POPC) phospholipid bilayer, hydrated by a symmetric 400mM KCl solution. The system was simulated in the NVT ensemble at constant temperature 320K and pressure 1atm, neutral pH and with applied transmembrane (TM) electrostatic potential (300mV). Restraints were applied to the dihedral angles of the backbone of the selectivity filter with a flat-bottom restraint with a 0.159 kcal/mol/degree^2 force constant. The following angels were allowed to vary +/- 10 degrees from the crystal structure dihedral angles, viz. +/-10 degrees from Thr59 (ϕ:77, ψ: 9), Val60 (ϕ: -63, ψ: -45), Gly61 (ϕ: 48, ψ: 52), Tyr62 (ϕ: -52, ψ: -36), and Gly63 (ϕ: 84, ψ: 8). In addition, to the selectivity restraints, harmonic distance restraints were applied on the level of the C-α distances of Pro19 and Phe97 to keep the inner gate open during the simulations, with a force constant of 2.39 kcal/mol/Å2 and applied to both adjacent and opposing subunits, viz., Pro19: 25.5Å for adjacent and 36Å for opposing subunits, and Phe97: 26Å for adjacent and 36.6Å for opposing subunits. The CHARMM36m and CHARMM36m-NBFIX simulations were performed on OpenMM 7.7 for 2 microseconds with frames recorded every 100 ps. A real-space cutoff of 12 angstroms was used, and the potential was smoothly truncated at the cutoff with a switching function starting at 10 angstroms. The long-range electrostatics were treated with PME. The constant temperature was maintained with the Langevin dynamics thermostat with 1/ps friction coefficient. The AMBER simulation followed a similar route as the CHARMM36m simulations with the following changes. The simulation was performed on AMBER20 using the PMEMD GPU accelerated MD simulation engine. A real-space cutoff of 9 angstroms was used.

The simulations indicate that the MthK structure is conductive at the level of the selectivity filter sites S4 through S1, with an average conductance of 48.24pS in symmetric 400mM KCl—significantly below the recorded single-channel conductance of MthK ie., ~170pS in symmetric 150mM KCl (Li et al., 2007) or ~96pS in symmetric 200mM KCl (Zadek and Nimigean, 2006). At such high concentration (400mM KCl), one would expect a single-channel conductance on the order of 400-500 pS. Force-field differences in the simulated conduction rate and mechanism resume as following: (1) conduction is accelerated in the AMBER force field via the hard-knock mechanism; (2) conduction is slightly slower in the CHARMM36 force field via the hard-knock mechanism; and, (3) conduction is also accelerated in the AMBER-like CHARMM-NBFIX force field via the hard-knock mechanism. Independently, these findings thus support that the CHARMM36m-NBFIX with AMBER-ized ion interactions reproduces the known AMBER behavior of conductive channels, including the accelerated conduction rates and hard-knock mechanism (Köpfer et al., 2014). Details of the MthK simulations and Supplementary file 3 were added and discussed in the revised manuscript.

3) In the AMBER simulation presented in Figure 2, permeation cessation does not seem to be correlated with the movement of the I398 as stated in the text. Please clarify.

The conduction mechanism depends on the concentration of incoming K⁺ ions in the central cavity of the channel. Although the last conduction event takes place around ~2μs of simulation, note that only after ~4μs of simulation, the knock-on mechanism is persistently disrupted as ions are excluded from the central cavity of the channel by isoleucine I398. There is also a phenomenon of dewetting of the central cavity when the I398 side chain starts to block ion conduction through the pore. The main text was revised accordingly.

Figure S6 shows the AMBER simulation of the triple-mutant channel with the I398N mutation. The conformation of the dilated selectivity filter is stable in presence of the mutation. The average intersubunit C_β-C_β distance of I398N fluctuates between ~11Å and ~15Å in the beginning of the AMBER trajectory, before reaching the value of ~9Å in the final stage of the simulation. Compared to the wild-type simulation, the local pore radius and water density are significantly enhanced under I398N mutation—as clearly indicated in the revised Supplementary file 6, showing a comparative analysis of the average properties of the isoleucine gate with and without mutation (I398N). The open and hydrated configuration of the gate in the I398N mutant allows ion conduction across the dilated selectivity filter as long as structural fluctuations of the PVP motif (V402) do not obstruct the permeation pathway. Similar conclusions can be drawn from CHARMM36m simulation in which long-lived fluctuations of the mutant gate I398N allows intermittent conduction of ions across the dilated conformation of the selectivity filter over the microsecond timescale (Figure 3-supplement 4 and Supplementary file 6).

The fact that the conductance of the dilated or conductive conformations calculated from MD are both smaller than the experimental values should not cause confusion. All force fields are imperfect and tend to underestimate the rate of ion conduction at physiological voltages. The key result from MD is that the dilated filter is able to conduct, at a rate similar to that of the conductive conformation of the filter, and what truly blocks conduction completely is the conformational change involving the I398 gate that occludes the pore. Indeed, the concept of a conductive dilated conformation together with a dynamical I398 gate deduced from the present MD simulations provides a simple and compelling explanation for the observation of permeation in Shaker W434F at the same rate as WT with extremely low open probabilities and very short duration that was reported by Sigworth and co-workers (1997).

It is correct that the 3m channel and the 3m_I398N mutant activate at more positive voltages compared to the kv1.2/2.1 channel. We believe this is most likely due to the triple mutation, especially the W to F mutation (Positions W362F) (see Figure 3—figure supplement 3). The activation kinetics is similar between 2m and 2m_I398N, which we believe is the fair comparison to evaluate the effect of I398N. The 3m channel has 3 mutations compared to WT (W362F, S367T and V377T). A fair comparison to evaluate the effects of I398N is to compare with the double mutant without the W362F mutation. The double mutant (2m) and the double mutant (2m_ I398N) have similar biophysical features as to the WT, G-V, kinetics of activation and inactivation. Moreover, the fact that AgTx is able to bind to and block more efficiently the I398N_2m, but not the I398N_3m demonstrates that I398N does not affect the dilated state of the selectivity filter induced by the W362F mutation. These results point to two main conclusions: (1) the W362F mutation affects the channel displacing the G-V to the right, and (2) the I398N does not affect the channel. Same goes for the slow inactivation comparison. While 3m channel inactivates completely within 10ms of depolarization, the addition of I398N completely changes the channel behavior: after 100ms of depolarization, more than 95% of current is sustained. What we tried to demonstrate here is that a dilated filter, facilitated by the triple mutation, itself doesn’t stop the ion permeation. Only with the additional conformational changes at isoleucine 398 can the channel enter into inactivated state and ceases to conduct. We acknowledge that the kinetics of the 3m_I398N channels are faster than the 3m, 2m, 2m_I398N and the Chimera WT. We do not know exactly the mechanism for this. It could be that I398N combined with the 3m affects the coupling between the activation gate and the selectivity filter. A paragraph discussion is added to address this issue.