Differential ion dehydration energetics explains selectivity in the non-canonical lysosomal K+ channel TMEM175

  1. SeCheol Oh
  2. Fabrizio Marinelli
  3. Wenchang Zhou
  4. Jooyeon Lee
  5. Ho Jeong Choi
  6. Min Kim
  7. José D Faraldo-Gómez  Is a corresponding author
  8. Richard K Hite  Is a corresponding author
  1. Structural Biology Program, Memorial Sloan Kettering Cancer Center, United States
  2. Theoretical Molecular Biophysics Section, National Heart, Lung and Blood Institute, National Institutes of Health, United States
  3. Department of Chemistry, Chungbuk National University, Republic of Korea

Abstract

Structures of the human lysosomal K+ channel transmembrane protein 175 (TMEM175) in open and closed states revealed a novel architecture lacking the canonical K+ selectivity filter motif present in previously known K+ channel structures. A hydrophobic constriction composed of four isoleucine residues was resolved in the pore and proposed to serve as the gate in the closed state, and to confer ion selectivity in the open state. Here, we achieve higher-resolution structures of the open and closed states and employ molecular dynamics simulations to analyze the conducting properties of the putative open state, demonstrating that it is permeable to K+ and, to a lesser degree, also Na+. Both cations must dehydrate significantly to penetrate the narrow hydrophobic constriction, but ion flow is assisted by a favorable electrostatic field generated by the protein that spans the length of the pore. The balance of these opposing energetic factors explains why permeation is feasible, and why TMEM175 is selective for K+ over Na+, despite the absence of the canonical selectivity filter. Accordingly, mutagenesis experiments reveal an exquisite sensitivity of the channel to perturbations that mitigate the constriction. Together, these data reveal a novel mechanism for selective permeation of ions by TMEM175 that is unlike that of other K+ channels.

Editor's evaluation

This manuscript explores the mechanisms of permeation and selectivity in the unusual potassium-selective ion channel TMEM175, which lacks a canonical selectivity filter. The study is led by molecular dynamics simulations and free energy calculations, complemented by a cryoEM analysis and electrophysiological recordings. The authors propose a novel, single ion-based mechanism of permeation, together with a partial dehydration-driven selectivity mechanism. This study will appeal to readers interested in the structure and function of ion channels and in molecular mechanisms of ion translocation.

https://doi.org/10.7554/eLife.75122.sa0

Introduction

Transmembrane protein 175 (TMEM175) is a potassium (K+)-selective cation channel that is evolutionarily distinct from all other known ion channels. In mammals, TMEM175 is expressed in lysosomes, where it is critical for maintaining lysosomal homeostasis (Cang et al., 2015). As a K+-selective channel, TMEM175 contributes to establishing the membrane potential across the lysosomal membrane (Cang et al., 2015). Loss of TMEM175 function can lead to numerous defects in the endolysosomal system including dysregulation of lysosomal pH, defects in autophagy and mitophagy, reduced glucocerebrosidase activity, and an increased susceptibility to α-synuclein cytotoxicity (Jinn et al., 2019; Jinn et al., 2017; Krohn et al., 2020). Dysregulation of lysosomal function is commonly associated with neurodegenerative disease and point mutations in TMEM175 have been identified that are strongly associated with development of Parkinson’s disease (PD) (Blauwendraat et al., 2019; Iwaki et al., 2019; Jinn et al., 2019; Jinn et al., 2017; Krohn et al., 2020; Nalls et al., 2014; Wie et al., 2021). For example, the M393T point mutation is a loss-of-function mutation associated with the increased likelihood and early onset of PD and the Q65P mutant is a gain-of-function mutation associated with a reduced likelihood of developing of PD (Wie et al., 2021).

Consistent with being evolutionarily distinct from other ion channels, TMEM175 channels from prokaryotes and humans adopt a unique architecture relative to all known ion channel structures (Brunner et al., 2020; Lee et al., 2017; Oh et al., 2020). Structures of two prokaryotic TMEM175 homologs in closed states revealed a homotetrameric channel possessing a central ion conduction pathway lined by several layers of hydrophobic residues from the first transmembrane (TM) helix of each of the four protomers (Brunner et al., 2020; Lee et al., 2017). Despite similar overall structures, analysis of the two prokaryotic TMEM175 structures led to conflicting proposals for how TMEM175 channels selectively permeate K+ ions (Brunner et al., 2020; Lee et al., 2017). However, human TMEM175 (hTMEM175) channels are homodimers, rather than homotetramers, and are ~30-fold selective for K+ over Na+, whereas prokaryotic TMEM175 channels exhibit only 2- to 4-fold selectivity for K+/Na+ (Cang et al., 2015). Because of the notable structural and functional differences between human and prokaryotic TMEM175, it remains unclear if the mechanisms gleaned from analyses of prokaryotic TMEM175 channels are generalizable for hTMEM175.

To begin to understand the mechanisms that underlie TMEM175 function in lysosomes, we determined the structure of hTMEM175 by cryo-EM and identified two conformational states (Oh et al., 2020). hTMEM175 protomers consist of two homologous repeats of six TM helices, which assemble into a dimer to form the channel. A 2-fold symmetric ion conduction pathway runs along the center of the dimer, lined by two copies of TM1 and TM7, the first helices of each repeat. Unlike most cation channels, including canonical K+ channels such as KcsA and Shaker, the pore of TMEM175 contains no clusters of hydrophilic side chains or backbone atoms that control selectivity (Doyle et al., 1998; Long et al., 2005). Rather, the most striking feature in the pore of TMEM175 is a narrow hydrophobic constriction formed by the side chains of the conserved Ile46 from TM1 and the conserved Ile271 from TM7, which we termed the isoleucine constriction (Oh et al., 2020). In one of the two cryo-EM structures, the isoleucine constriction is too narrow to accommodate K+ ions and thus we assigned it as a closed state. In the other structure, however, the isoleucine constriction is sufficiently dilated to accommodate partially hydrated K+ ions, and therefore we proposed that this structure corresponds to a conductive state of the channel. Moreover, because ions would need to be partially dehydrated to traverse the pore, it was proposed that the relative energetics of K+ and Na+ dehydration through the isoleucine constriction contributes to the ion selectivity of TMEM175 (Lee et al., 2017; Oh et al., 2020). While plausible, this proposal relies on the assumption that the putative open state is indeed conductive; however, assigning functional states to cryo-EM structural snapshots is challenging. Here, we employ high-resolution cryo-EM structure determination, molecular dynamics (MD) simulations, and electrophysiological analyses to investigate the mechanisms of ion permeation and selectivity in TMEM175.

Results

Improved cryo-EM structures of hTMEM175

Taking advantage of recent advances in cryo-EM image analysis, we reprocessed our published cryo-EM images of TMEM175 in KCl to achieve higher resolution (Punjani and Fleet, 2021; Punjani et al., 2020; Zivanov et al., 2018). The resolution of the putative open state improved from 2.64 to 2.45 Å and the closed state improved from 3.03 to 2.61 Å (Figure 1—figure supplement 1 and Appendix 1—table 1). While the improved structures are similar to those previously determined with an all-atom RMSD of 0.3 Å for the open state and 0.4 Å for the closed state, the higher-resolution reconstructions allowed us to better define the ion conduction pathways (Figure 1A–D). In both states, the ion conduction pathway is 2-fold symmetric, narrow, and largely devoid of polar contacts, particularly near the center due to the presence of several hydrophobic side chains including Ile46, Val50, Leu53, Ile271, Leu275, and Leu278 (Figure 1C–D). In the closed state, the isoleucine constriction formed by Ile46 and Ile271 is too narrow to allow ion permeation (Figure 1D). The side chains of Ile271 point toward the channel axis, narrowing the pore to a minimum radius of 0.2 Å (Figure 1D and E). In the open state, by contrast, conformational changes in the pore-lining helices TM1 and TM7 helices dilate the isoleucine constriction to a minimum radius of 1.6 Å, which appears to be sufficient to accommodate a partially dehydrated K+ ion (Figure 1C and E). In its dilated conformation, the isoleucine constriction adopts a pseudo 4-fold symmetric configuration with the side chains of both Ile46 and Ile271 adopting the mt rotamer (Figure 1—figure supplement 2).

Figure 1 with 2 supplements see all
Structures of transmembrane protein 175 (TMEM175) in open and closed states.

(A–B) Structures of the human TMEM175 in open (A) and closed (B) states. The channel is orientated with the cytosolic side facing up and the luminal side down. Gray bars represent approximate width of the membrane. (C–D) Ion conduction pathways in open (C) and closed (D) states. Pore-lining residues are shown as sticks. Bound K+ ions and water molecules are shown as purple and red spheres, respectively. Density peaks corresponding to K+ and water molecules are shown as blue mesh and contoured at 12σ. Front and rear domains are removed for clarity. (E) Plot of pore radius as a function of position along the pore axis. The 0 position corresponds to the isoleucine constriction, positive values correspond to the luminal side of the pore, and negative values correspond to the cytosolic side of the pore.

The higher-resolution reconstructions also allowed an improved modeling of five ion-binding sites, which we call K1 through K5, and ordered water molecules in the pore of the open state. Two non-protein densities are now resolved at the K3 and K4 sites that flank the isoleucine constriction (Figure 1—figure supplement 2). We assigned these densities as K+ ions based on our previous comparison with a structure determined in the presence of Cs+ (Oh et al., 2020). The K3 and K4 ion-binding sites are positioned 3.7–4.0 Å away from the side chains of Ile46 and Ile271 and are coordinated exclusively by water molecules (Figure 1—figure supplement 2). Notably, the ions in the K3 and K4 sites are only partially hydrated. The K3-binding site on the cytoplasmic site of the constriction is coordinated by four ordered water molecules that are between 2.8 and 3.0 Å away, while the K4 site on the luminal side is coordinated by four water molecules that are between 2.9 and 3.5 Å away. The distance between the two sites is 2.8 Å and thus they are unlikely to be simultaneously occupied by ions.

For the closed state, we can now model three ion-binding sites and 25 water molecules into non-protein densities (Figure 1D). In the cytosolic region of the pore, where the channel undergoes minimal conformational changes during gating, the K1 and K2 ion-binding sites and water molecules occupy almost identical positions to those resolved in the open state (Figure 1—figure supplement 2). In contrast, there is little correspondence between the ion and water configurations at the isoleucine constriction or on the luminal side of the pore due to the conformational changes that occur to the channel during channel gating. The inward movement of Ile271 displaces the ion-binding sites on either side of the isoleucine constriction. Movements of Thr49 and Thr274, which both coordinate water molecules in the open state, result in a movement of the luminal K5 ion-binding site toward the luminal entrance of the pore. Thus, the higher-resolution structures reveal how changes in the protein structure alter the ion conduction pathway to prevent ion permeation in the closed state.

Energetics and mechanism of K+ permeation

To ascertain whether the structure of the putative open state is indeed permeable to K+, we turned to all-atom MD simulations of the channel in a model phospholipid bilayer (Figure 2—figure supplement 1). Specifically, using the enhanced-sampling methodology known as multiple-walker Metadynamics (Raiteri et al., 2006), we induced the permeation of K+ across the channel at 0 mV and symmetric 100 mM KCl and evaluated the associated potential of mean force (or free-energy profile). Importantly, these calculations used a newly formulated reaction coordinate (Materials and methods) designed to avoid a priori assumptions about the mechanism of permeation, such as the number of ions that reside within the pore at a given time.

The results from this analysis are summarized in Figures 2 and 3. We observe that K+ ions reach the isoleucine constriction readily from either side of the membrane, through a series of shallow free-energy barriers and transient-binding sites (Figure 2A). At the isoleucine constriction, however, the free energy increases steeply, peaking at about 7 kcal/mol (Figure 2A). These features recapitulate the pattern of non-protein densities resolved in the pore in the reconstruction of the putative open state; densities near Ser38 and Ala42 on the cytoplasmic side and near Val50 on the luminal side appear as metastable states in the calculated free-energy profile. Densities right below and above the central isoleucine constriction also appear as shoulders in the free-energy profile. Unlike in canonical K+ channels, however, the simulation data reveals no evidence of a multi-ion process involving concurrent occupancy of proximal-binding sites near the constriction (Figure 2—figure supplement 2). Instead, K+ ions approach and cross this constriction individually. It is apparent, too, that permeation requires a gradual but drastic depletion of the ion hydration shells. As K+ traverses the isoleucine constriction, the first two hydration shells are reduced to only ~9 molecules, down from 31 in the bulk; the first shell averages to about 4, down from 7 (Figures 2B, C ,, 3A).

Figure 2 with 3 supplements see all
Enhanced-sampling molecular dynamics (MD) simulations of K+ and Na+ permeation through transmembrane protein 175 (TMEM175).

(A) Potential of mean force (or free energy) as a function of the ion position along the pore axis (black), for either K+ (left) or Na+ (right). See also Figure 2—figure supplement 1. The difference between the free-energy peak for Na+ and that for K+ (ΔΔG) is indicated. Blue arrows indicate the approximate position of the density peaks assigned to K+ ions in Figure 1. For reference, the position of selected residues in helices TM1 and TM7 is also indicated under the free-energy curves. These positions are defined as the time average of the center-of-mass of side chain and Cα atoms, including equivalent protein subunits, projected along the pore axis. (B) Average number of water molecules in the first and second ion hydration shells, as a function of the ion position along the pore axis (black). The degree of depletion at the isoleucine constriction (ΔN), relative to bulk hydration, is indicated for each ion. (C) Same as (B), only for the first ion hydration shell (black). The average number of protein-oxygen atoms observed to coordinate an ion, as a function the ion position, is also shown alongside (blue), indicating the principal contributing residues. See also Figure 2—figure supplement 3. (D) Ion-protein electrostatic interaction energy as a function of the ion position along the pore axis. In all panels, gray profiles represent the same quantity shown in black/blue calculated using only the first or second half of the simulation data. Comparison of these two profiles provides a metric of the statistical error in these calculations. The mean difference between the free-energy profiles for K+ is 0.14 kcal/mol; for Na+, it is 0.22 kcal/mol. See Materials and methods for further details and definitions.

Figure 3 with 2 supplements see all
Molecular dynamics (MD) simulations of K+ and Na+ permeation through human transmembrane protein 175 (hTMEM175).

(A) Snapshots of one of the trajectories calculated with enhanced-sampling MD, at 0 mV and 100 mM KCl, showing a K+ permeation event across the isoleucine constriction. For clarity only TM7 is shown, in cartoon representation, alongside the side chain of I271. Red spheres represent the oxygen atoms of the water molecules in the first ion hydration shell, defined as in Figure 2. The number of water molecules in the first ion hydration shell is shown in parentheses. (B) Same as (A), for a Na+ permeation event. (C) Same as (A), from a trajectory calculated with conventional MD sampling, but at 500 mV and 400 mM KCl. (D) For the latter trajectory, time series of the position along the channel axis of the subset of the K+ ions in the simulation found within 10 Å of that axis. The trajectory revealed two permeation events in 1 µs.

Given the largely hydrophobic nature of the TMEM175 pore, it is logical that the free-energy barrier for permeation is significantly higher than those in canonical K+ channels. The finding that this barrier is only 7 kcal/mol is however intriguing, given that the cost of K+ dehydration is much larger than that (up to about 80 kcal/mol for full dehydration). This observation indicates that other factors must facilitate ion permeation in TMEM175. Indeed, further examination of the simulated trajectories using Poisson theory (Materials and methods) reveals that acidic side chains at both the luminal and cytosolic entrances of the pore generate a strongly favorable electrostatic field that spans the length of the pore (Figure 2D). Some of these residues also transiently coordinate the permeating ions as they enter and exit the pore, along with a few other carbonyl and carboxyl groups in side chains and backbone (Figure 2C, Figure 2—figure supplement 3). However, for a span of about 15 Å centered at the isoleucine constriction, direct protein contacts are minimal. That the barrier for ion conduction in TMEM175 is surmountable is thus a consequence of the degree of hydration retained by the permeating ions as well as the electrostatic field generated by the protein that attracts them to the pore interior.

Integration of the potential-of-mean-force profile obtained at 0 mV and 100 mM KCl translates into a single-channel conductance of approximately 0.23 pS (Materials and methods). At 500 mV, and assuming the structure of the channel is completely unresponsive to voltage, this conductance would in turn produce a 0.1 pA current, or a rate of K+ permeation of about 0.6 ions/μs. To assess these estimates, we calculated an independent 1-μs-long MD simulation using conventional sampling but under 500 mV (lumen positive) and 400 mM KCl. This trajectory revealed a stable pore conformation and two K+ permeation events, confirming that the open state is functionally conductive and validating the single-ion mechanism inferred from the free-energy simulations (Figure 3C and D; Figure 3—video 1 and Figure 3—video 2). K+ ions are seen to dwell at the shallow-binding site adjacent to the constriction, on the luminal side, before returning to the bulk or rapidly traversing the free-energy barrier, along with 4 water molecules in the first hydration shell (Figure 3C).

Differentials in ion dehydration relative to bulk explain selectivity for K+ over Na+

Like canonical K+ channels, TMEM175 channels are more permeable to K+ than Na+. However, in the absence of the selectivity filter seen in canonical K+ channels, it has been unclear how to explain their selectivity in structural terms. To begin to bridge this gap, we again turned to MD simulations. Specifically, we repeated the enhanced-sampling simulations carried out to examine K+ permeation after substituting the KCl solution with NaCl and examined the resulting data analogously (Figures 2 and 3). The potential-of-mean-force profile for Na+ is similar to that obtained for K+ in its overall features, including the rate-limiting barrier at the isoleucine constriction (Figure 2A) . The largest difference between the profiles is the magnitude of the barrier at the isoleucine constriction, which peaks at about 8 kcal/mol relative to bulk for Na+ compared to about 7 kcal/mol for K+. Integration of the potential-of-mean-force profile obtained for Na+ translates into a single-channel conductance of approximately 0.037 pS, that is, the calculated conductance for Na+ is 6-fold smaller than for K+. While available experimental estimates for human indicate TMEM175 a stronger preference for K+, the qualitative agreement between experimental and calculated values suggest the simulations capture the essence of the mechanism of K+ selectivity. As observed for K+, Na+ permeation proceeds via a single-ion mechanism (Figure 2—figure supplement 2, Figure 3B). Permeation is again opposed by the drastic dehydration of the ion (Figure 2B), and favored by its interaction with the electrostatic field created by the protein within the pore (Figure 2D), as well as by transient interactions with multiple residues on both cytoplasmic and luminal sides (Figure 2C). These stabilizing factors are not identical but appear comparable to those observed for K+. By contrast, the degree of dehydration required for permeation clearly differs. We observe that the depletion in the both the first and second solvation shells, relative to bulk numbers, is significantly smaller for Na+ than for K+. As will be further discussed below, this relative difference in dehydration energetics likely explains why TMEM175 is only ~30-fold more permeable to K+, even though the bulk water selectivity for Na+ is over a trillion-fold. In other words, like for K+ channels that feature a canonical selectivity filter, it can be said that the narrow pore of TMEM175 favors Na+ over K+, but not as much as bulk water. The channel is thus more permeable to K+ than Na+.

Role of the isoleucine constriction in ion selectivity and permeation

The isoleucine constriction constitutes the largest free-energy barrier to the permeation of K+ and Na+ in the pore of TMEM175 (Figure 2). This narrow constriction forces the ions to shed much of their hydration shells in order to traverse the pore. It was previously demonstrated that mutating both Ile46 and Ile271 to asparagine diminishes the selectivity of TMEM175, indicating that the hydrophobicity of the isoleucine constriction is crucial for selectivity (Lee et al., 2017). To determine if the dimensions of the constriction are also important, we used MD simulations to calculate the change in free energy that results when a K+ ion at the isoleucine constriction is exchanged with a Na+ ion, both for wild-type TMEM175 and for a I46V mutant with an expanded isoleucine constriction (Materials and methods). In both cases the exchange results in a free-energy gain (Figure 4A), consistent with the notion that Na+ is the favored species at the constriction. However, this gain is smaller than that observed for the exchange of a K+ ion in bulk water with a Na+ ion. As discussed above, it is relative favorability for K+ at the constriction compared to bulk solution that would be the expected result for a K+-selective channel (Figure 4A). Specifically, the differential relative to bulk is 1.2 kcal/mol in favor of K+, similar to the ~1 kcal/mol difference in the peak values of the free-energy curves discussed in the previous section (Figure 2A). Na+ is however favored to a larger extent in the I46V mutant, whose isoleucine constriction has a similar hydrophobicity but is expanded by the loss of a methyl group from the side chain of Ile46 in each protomer (Figure 4A); specifically, the calculations indicate that this mutant is about 2.5-fold (or 0.5 kcal/mol) less K+-selective than the wild-type channel (Figure 4A). These results again correlate with the degree of dehydration of each ion, relative to bulk values (Figure 4B).

Ion selectivity of the isoleucine constriction.

(A) Left, free-energy difference between hydrated K+ and Na+, in bulk water, calculated with molecular dynamics (MD) simulations using the free-energy perturbation (FEP) method. The free energy is plotted as a function of a varying parameter λ that determines the size of ion (λ=0 for K+ and λ=1 for Na+). Results are shown for a calculation wherein K+ is transformed into Na+ (forward, black line) and for another carried out in the opposite direction (backward, red circles). Middle, analogous free-energy differences, calculated for an ion that is confined within the isoleucine constriction of wild-type TMEM175. Right, same as above, for the I46V mutant. Comparison of forward and backward curves provides a metric of the statistical error in these calculations, which is at most 0.1 kcal/mol. (B) Snapshots of the MD/FEP simulations at the end of the forward and backward transformations, for wild-type TMEM175 and I46V mutant. For clarity, only TM1 is shown, in cartoon representation, alongside the side chain of I46/V46. Red spheres represent the oxygen atoms of the water molecules in the first ion hydration shell, defined as in Figure 2.

Using whole-cell electrical recordings, we experimentally assessed the effects of expanding the size of the isoleucine constriction on selectivity by measuring the selectivity of Cs+ over Na+ for comparison with the computational analyses. We choose Cs+ rather than K+ because Cs+ blocks most canonical K+ channels, thereby reducing endogenous currents. Nevertheless, whole-cell currents recorded from cells expressing the I46A, I46V, I271A, and I271V mutants in a bi-ionic Cs+/Na+ condition were much smaller than those recorded from wild-type TMEM175 and in some cases similar to currents recorded from non-transfected cells, which hampered accurate determination of ion permeability ratios via measuring reversal potentials (Erev) (Figure 5 and Figure 5—figure supplement 3). Thus, isolation of the TMEM175-specific currents from the endogenous HEK293T and background leak was necessary to evaluate the selectivity of the TMEM175 mutants. To this end, we first performed a background subtraction with the TMEM175 inhibitor 4-aminopyridine (4-AP) (Cang et al., 2015). However, while 4-AP can effectively inhibit wild-type TMEM175, it was ineffective against the mutants and thus not suitable for isolating the TMEM175-specific currents (Figure 5—figure supplements 3 and 4). We therefore repeated the background subtraction experiments with a novel TMEM175 inhibitor that we have developed, which we call AP-6 (Figure 5 and Figure 5—figure supplements 13). Following AP-6 background subtraction, I46V and I271V displayed Cs+-selective currents with Erev of –13 and –16 mV, corresponding to Cs+ over Na+ permeability ratios (PCs/Na) of 1.7 and 1.9, respectively (Figure 5). Using the same background subtraction approach, we estimated the PCs/Na of wild-type TMEM175 to be 20 (Erev of –76 mV). This 10-fold reduction in PCs/Na ion permeation ratio is in line with the computational analyses of the free-energy difference in resulting from K+ to Na+ ion exchange at the isoleucine constriction of I46V and wild type. We were not able to detect any AP-6-sensitive currents from the I46A or I271A mutants either due their insensitivity toward AP-6 or inactivity of the mutant channels. These data demonstrate that increasing the size of the isoleucine constriction leads to a clear reduction in K+ selectivity. Thus, the isoleucine constriction, which is universally conserved among eukaryotic TMEM175 channels, serves as a hydrophobic selectivity filter that is exquisitely sensitive to very subtle changes in its structure.

Figure 5 with 4 supplements see all
Electrophysiological analysis of human transmembrane protein 175 (hTMEM175)-transfected HEK293T cells.

Representative whole-cell electrical recordings and current-voltage relationships of hTMEM175-transfected (A), I46V-transfected (B), and I271V-transfected (C) HEK293T cells in the absence and presence of inhibitor AP-6, in bi-ionic conditions of 150 mM Cs+ (intracellular) and 150 mM Na+ (extracellular). Current amplitudes in the current-voltage relationship plots were gained by averaging current values in the shaded time window. (D) A schematic of whole-cell patch clamp performed in (A–C and E). (E) Current-voltage relationships of six independent whole-cell patch clamp recordings of hTMEM175 WT-, I46V-, and I271V-transfected cells in bi-ionic conditions of 150 mM Cs+ (intracellular) and 150 mM Na+ (extracellular).

Discussion

Permeation through selective ion channels arises from the interplay between channel-ion interactions, solvent-ion interactions, and in some cases, ion-ion interactions. The energetic cost associated with fully or partially dehydrating an ion such as K+ is so large that permeation would only very rarely occur without favorable interactions with the channel as the ion traverses the conduction pathway. In canonical K+ channels, backbone and side chain oxygen atoms provide a series of direct coordinating interactions that closely resemble those experienced by the ion in bulk water. In addition, these channels are often capable of attracting multiple ions to the pore concurrently. Thus, the cost of dehydration is almost entirely offset by channel-ion and ion-ion interactions, and the resulting free-energy barriers for permeation are small. In contrast, such interactions are lacking for much of the TMEM175 pore, and entirely absent near the isoleucine constriction. Lacking these interactions, it is unlikely that direct protein-ion interactions impart ion selectivity on hTMEM175 channels as was proposed based upon the structural analysis of a TMEM175 homolog from the prokaryote Marivirga tractuosa (Brunner et al., 2020). K+ ions crossing the pore must nevertheless shed much of their hydration shells; only ~9 of the 31 water molecules in its first two shells remain at the constriction. In addition, ions permeate TMEM175 one at a time, without direct or indirect knock-on interactions. Permeation is however strongly favored by a delocalized electrostatic field across the pore, which offsets much of the cost of ion dehydration, resulting in a free-energy barrier of ~7 kcal/mol (Figure 2). While this is a high barrier, it is not insurmountable. Indeed, considering that MD simulations using conventional force fields tend to underestimate ion conduction rates, in some cases up to 10-fold (Allen et al., 2006; Mironenko et al., 2021), we estimate the unitary conductance of hTMEM175 is 0.1–0.5 pS, which is well within the broad range of experimentally measured permeation rates for other selective cation channels such as ~100 pS for Slo K+ and 9–24 fS for CRAC channels (Tao et al., 2016; Zweifach and Lewis, 1993). Thus, despite the hydrophobicity of the isoleucine constriction, the unique architecture and amino acid make-up of the TMEM175 pore facilitates ion permeation.

Importantly, the balance between protein-ion interactions and solvent-ion interactions is dependent on the ion type, and so these opposing forces also govern ion selectivity in channels. For example, when rationalizing differences in permeability between K+ and Na+, it is useful to keep in mind that dehydration of Na+ is much more costly than that of K+; the penalty for full dehydration is ~18 kcal/mol greater, rendering it a trillion-fold less probable. However, the observed selectivity of biological K+ channels against Na+ is only 1000-fold or less, implying Na+ establishes much more favorable interactions with the selectivity filters of this kind of K+ channels, as noted in previous studies (Kim and Allen, 2011; Kopec et al., 2018). In other words, what ultimately determines the specificity of a channel is the sum of the relative cost of dehydration of the competing ions and the differential in their free energy of interaction with the channel. Remarkably, for TMEM175 these favorable protein-ion interactions are imparted by residues near the luminal and cytoplasmic entrances to the pore, quite far from the isoleucine constriction. Thus, we propose that the ion selectivity of TMEM175 primarily results from a differential in the degree of dehydration required for permeation, relative to bulk water. Incidentally, whereas recent studies have revealed differing results for simulations of ion channels performed with different force fields (Klesse et al., 2020; Ocello et al., 2020), a mechanism that relies on dehydration energetics, rather than specific ion-protein or ion-ion interactions, will be reasonably described by the force field employed in our simulations to reproduce the relative free energies of hydration (and thus dehydration) of the alkali cations (Figure 4A).

Together, our analyses provide a structural and energetic model for the unique mechanisms of ion permeation and selectivity in TMEM175 channels. By combining a narrow hydrophobic sieve with a favorable electrostatic field, hTMEM175 selectively permeates K+ ions across the lysosomal membrane, as was proposed by Lee and colleagues for the TMEM175 homolog from Chamaesiphon minutus (Lee et al., 2017). As the large conductance Slo1 K+ channel, whose K+ permeation rate is several orders of magnitude higher than TMEM175, has also been observed in lysosomes, it will be important to uncover the specific roles for TMEM175’s distinct conduction and selectivity properties in the lysosome and why mutations of TMEM175 can both increase and decrease the likelihood of developing PD (Wang et al., 2017; Zhong et al., 2016).

Materials and methods

Analysis of electron microscopic images

Request a detailed protocol

7907 images of TMEM175 in KCl previously acquired in two data sets as 40-frame super-resolution movies (0.544 Å/pixel) using a Gatan K2 (Oh et al., 2020) were gain corrected, Fourier cropped by two, and aligned using whole-frame and local motion correction algorithms by Motioncor2 (Zheng et al., 2017) (1.088 Å/pixel). Whole-frame CTF parameters were determined using CTFfind 4.1.14 (Rohou and Grigorieff, 2015). Particles were automatically selected in Relion 3.0 using templates previously generated from 2D classification, resulting in 4,153,614 particles (Zivanov et al., 2018). False-positive selections and contaminants were excluded from the data using multiple rounds of heterogeneous classification in cryoSPARC v3.2 using the open and closed states, as well as several decoy classes generated from noise particles via ab initio reconstruction in cryoSPARC v3.2, resulting in a stack of 636,148 particles (Punjani et al., 2017). After Bayesian polishing in Relion and local CTF estimation and higher-order aberration correction in cryoSPARC v3.2, a consensus reconstruction was determined at resolution of 2.8 Å (Zivanov et al., 2019). A second round of Bayesian polishing in Relion 3 using a pixel size of 0.85 Å and box size of 384 yielded an improved consensus reconstruction at 2.5 Å. Iterative supervised heterogenous classification using open and closed maps low pass filtered to 6 Å as references resulted in 261,536 particles in the open state and 163,651 particles in the closed state. A 2.45 Å reconstruction of the open state particles and a 2.61 Å reconstruction of the closed state particles were obtained using non-uniform refinement in cryoSPARC v3.2 employing global and per-particle CTF correction. The reconstructions were subjected to density modification using the two unfiltered half-maps with a soft mask in Phenix (Terwilliger et al., 2019).

Model building and coordinate refinement

Request a detailed protocol

The structures of open (PDB:6WC9) and closed (PDB:6WCA) hTMEM175 were docked into the density maps in COOT and manually adjusted to fit the density (Emsley et al., 2010). Densities corresponding to TM5 and TM6 (residues 174–251) were too poorly ordered and omitted from the model. The final models are composed of residues 30–173 and 254–476. Atomic coordinates were refined against the density modified map using phenix.real_space_refinement with geometric and Ramachandran restraints maintained throughout (Adams et al., 2010). Pore radius calculations were performed using HOLE (Smart et al., 1996).

Simulation systems and general specifications

Request a detailed protocol

All simulations were calculated with NAMD 2.12 using the CHARMM36 force field for protein and lipids (Best et al., 2012; Klauda et al., 2010; Pastor and Mackerell, 2011; Phillips et al., 2005). The simulations were carried out at constant temperature (298 K) and semi-isotropic pressure (1 atm), using periodic boundary conditions and an integration time step of 2 fs. Long-range electrostatic interactions were calculated using PME, with a real-space cut-off of 12 Å. Van der Waals interactions were computed with a Lennard-Jones potential, cut-off at 12 Å with a smooth switching function taking effect at 10 Å.

The simulations were based on the high-resolution cryo-EM structure of the hTMEM175 channel (PDB 6WC9). The specific construct studied includes residues 30–165 and residues 254–476. K+ ions and water molecules originally included in the cryo-EM structure were removed. All ionizable side chains were set in their default protonation state at pH 7, except for H57, which was protonated on account of its proximity to D279, E282, and D283. The protein construct was embedded in a pre-equilibrated hydrated palmitoyl-oleoyl-phosphatidyl-choline (POPC) lipid bilayer using GRIFFIN, and enclosed in a periodic orthorhombic box of ~100 × 100 × 111 Å3 in size (Staritzbichler et al., 2011). The two resulting simulation systems contain 222 POPC lipids, 24,342 water molecules, 49 Cl- ions, and either 43 K+ or 43 Na+ ions; that is, a salt concentration of 100 mM plus counterions to neutralize the protein net charge. The simulation systems were equilibrated following a staged protocol comprising a series of restrained simulations. The protocol consists of both positional and conformational restraints, gradually weakened over 100 ns. A third simulation system was prepared with 400 mM KCl, by adding 132 K+ and 132 Cl- ions to the 100 mM KCl system (replacing water), followed by a 100 ns equilibration.

Simulation of K+ permeation under voltage

Request a detailed protocol

To simulate the flow of K+ across TMEM175 from the luminal to the cytoplasmic side, a conventional MD trajectory of 1 μs was calculated for the 400 mM KCl condition under a constant electric field perpendicular to the membrane plane, of magnitude Ez = –0.1045 kcal/(mol Å e). Given the average length of the box in this direction (Lz = 110.5 Å), the corresponding TM potential is –500 mV (Φ=Ez × Lz × 0.0434 V e-1 (kcal/mol)-1) (Roux, 2008).

Enhanced-sampling simulation of K+ and Na+ permeation

Request a detailed protocol

To induce reversible ion permeation across TMEM175 at 0 mV and 100 mM KCl or 100 mM NaCl, we used the multiple-walker Metadynamics method (Raiteri et al., 2006). In Metadynamics, a biasing potential is introduced in an MD simulation to facilitate the exploration of configurational space, as defined by one or more reaction coordinates, also known as collective variables. This biasing potential consists of a series of Gaussian functions that expands as the simulation progresses, to foster the trajectory to visit high free-energy configurations. In the multiple-walker approach, several MD simulations are carried out in parallel, each sampling a different trajectory, but with a shared biasing potential that is also collectively constructed. To study K+ permeation, we used eight walkers, each producing a 600 ns trajectory. For Na+, we also used eight walkers, 700 ns each.

In biased-sampling simulations of channel permeation, it is common for the position of one or more ions along the pore axis to be used as collective variables. This choice might be reasonable when the number of ions involved in the mechanism is known. It was not known in our case, and hence, the type of collective variable used in these simulations was newly formulated to circumvent a priori assumptions. This variable was implemented in a modified version of PLUMED 1.3 (Bonomi et al., 2009). Specifically, the variable is a quantitative descriptor of the proximity between the ions in the simulation system and a virtual center within the pore. More precisely, the variable is:

(1) ζmin=βlogkexp{β/(|ΔZk|+C)}C

where the k index identifies each of the K+ or Na+ ions in the system and Zk is the Cartesian Z-component of the distance between ion k and a center-of-mass defined by a group of protein atoms. C is a positive constant (set to 2 Å) that avoids numerical instabilities when Zk~0 and β is a smoothing parameter (set to 100 Å). To successfully induce ion permeation events, we used two variables of the kind specified by Equation 1, namely ζminA and ζminB , each defined in reference to a center of atoms. Center A is defined by the Cα atoms of residues 44–47 and 268–271, while center B is defined by the Cα atoms of residues 45–48 and 269–272. Centers A and B are therefore along the pore axis, separated by 1 Å and flanking the isoleucine constriction. The Gaussian functions used to gradually construct the biasing potentials applied to ζminA and ζminB had a width 0.25 Å and were added in 4 ps intervals. To reduce systematic errors near the boundary ζmin=0, reflected and inverted Gaussians were added beyond this boundary for both variables as previously described (Crespo et al., 2010). In the simulations of K+ permeation, the Gaussians height was gradually raised from 0.0035 to 0.007 kcal/mol in the first 30 ns of simulation, and gradually diminished back to 0.0035 kcal/mol after 100 ns. For Na+, the Gaussians height was gradually raised from 0.0035 to 0.007 kcal/mol in the first 50 ns of simulation and diminished back to 0.0035 kcal/mol after 250 ns.

Derivation of free energies and other quantitative descriptors from biased-sampling trajectories

Request a detailed protocol

A post hoc reweighting procedure was used to derive unbiased averages and histograms from the MD trajectories enhanced by the Metadynamics biasing potential (Marinelli et al., 2009). In this approach, a time average of the Metadynamics biasing potential is calculated after it becomes approximately stationary across the range of (ζminA , ζminB) values of interest. (Here, the last 400 ns of simulation for K+ and the last 600 ns of simulation for Na+.) This effective potential, V¯(ζminA,ζminB) , is then used to adjust the statistical weight of each of the simulation snapshots considered in the analysis. Specifically, the statistical weight of a given snapshot Xi is:

(2) w(Xi)=exp{V¯(ζminA(Xi),ζminB(Xi))/kBT}jexp{V¯(ζminA(Xj),ζminB(Xj))/kBT}

where kB is the Boltzmann constant and T is the temperature. That is, snapshots Xi that fall in easily accessible regions of (ζminA,ζminB) where the accumulated Metadynamics bias is large are given a greater statistical weight, while those in unfavorable regions where the accumulated bias is less are also assigned a smaller weight. The effective potential for each simulation snapshot was derived using a 2D discretization across the space of (ζminA,ζminB), wherein each snapshot is assigned to a grid point and an associated value of V¯ using a recently developed tool for free-energy analysis (Marinelli and Faraldo-Gómez, 2021).

Following this approach, it is straightforward to derive unbiased estimates of the average ion occupancy along the length of the channel pore, and thus of the corresponding free-energy landscape. Specifically, let us define this length by an axis connecting the abovementioned centers A and B (which fluctuates with the channel but is approximately perpendicular to the membrane plane). For each ion k in the simulation system, let us also define zk(Xi) and Rk(Xi) as the projection of the ion coordinates along the pore axis and the distance to that axis, respectively, for a given snapshot Xi . If we divide the pore axis into a series of intervals or bins, the ion occupancy of a given bin α centered in zα is:

(3) ρ(zα)=ik(α,R0)w(Xi)

where the i index again denotes each trajectory snapshot included in the analysis, and the sum over k is restricted to ions whose zk(Xi) coordinate falls into bin α and that are also proximal to the pore axis (Rk(Xi)R0=10 Å). It follows that the potential of mean force (or free-energy profile) associated with this occupancy distribution along zα will be:

(4) F(zα)=kBT log ρ(zα)=kBT logikαw(Xi) +C

where C is a constant arbitrarily selected so that F0 in the bulk, where the profile is nearly flat.

We followed same reweighting approach to derive unbiased estimates for other descriptors, namely average ion coordination numbers and average ion-protein electrostatic interaction energies (definitions provided below). Specifically, the mean value of either of these observables at a given position along the pore axis is:

(5) O¯(zα)=1ρ(zα)ik(α, R0)Ok(Xi) w(Xi)

where i again denotes a trajectory snapshot, Ok(Xi) is the descriptor of interest of ion k, in that particular snapshot, and the sum over k is again restricted to ions within the pore and within bin α.

Calculation of ion coordination numbers

Request a detailed protocol

To quantify the coordination of K+ or Na+ by either water or protein oxygen atoms as a function of the ion position along the pore axis, we evaluated the following function of the ion-oxygen distances for each simulation snapshot Xi and each ion k in the system:

(6) Sk(Xi)=j1(rjk(Xi)/r0)1001(rjk(Xi)/r0)200

where the j index denotes all possible coordinating oxygens, and rjkXi denotes their distance to ion k in snapshot Xi. Note this function is virtually identical to a step function cut-off at distance equal to r0. For computational expediency, this evaluation was carried out with PLUMED 1.3 (Bonomi et al., 2009). For the evaluation of the number of water molecules in the first hydration shell, r0 was set to 3.5 Å for K+ and to 3.2 Å for Na+, which in each case corresponds to the position of the first minimum of the ion-oxygen radial distribution functions (RDF) in bulk water (data not shown). The same values of r0 were used for calculating the coordination with protein oxygens. Following the same criteria, the number of water molecules in the first and second shells was calculated by setting r0 to 6.0 and 5.7 Å, respectively, as these values reflect the position of the second minimum of the RDF in each case. The two exponents of the step function in Equation 6 were set to reproduce the integral of the RDF in bulk water within r0.

Calculation of ion-protein electrostatic interaction energies

Request a detailed protocol

To estimate the electrostatic interaction energy between each K+ or Na+ ion and the channel, we used a continuum electrostatic model based on the linearized form of Poisson’s equation. That is, for each ion k and snapshot Xi , we evaluated the electrostatic potential Φk generated at the position of the ion by the atomic charges in the protein structure (qp), in the context of a heterogenous dielectric-constant distribution defined by the protein (εp = 2 or 1), an implicit membrane surrounding the protein (εm = 2), the bulk water solvent (εs = 80), the water solvent inside the pore (εs` = 40), and the ion in question (εk = 2 or 1). (Note that both charge and dielectric-constant distributions are a function of Xi , as is the position of the ion.) The interaction between this electrostatic potential and the ion is given by:

(7) Ek(Xi) =qk Φk(qp,εp, εm,εs,εs,εk; Xi)

These calculations were carried out with the PBEQ numerical solver implemented in CHARMM c44b1 (Brooks et al., 2009). The set of atomic charges employed are those in the CHARMM36 force field; the set of atomic radii used to define the dielectric boundaries derives from an existing optimization for continuum-electrostatic calculations based on the CHARMM27 force field (Best et al., 2012; Nina et al., 1997). Electrostatic potential, charge, and dielectric-constant distributions were discretized on a lattice of 150 × 150 × 150 Å3, with lattice-point spacing of 1 Å. The thickness of the implicit membrane was 28.8 Å. The low-dielectric span of pore water was defined by a cylinder of radius 18 Å and height 28.8 Å. The protein surface was defined using the REEN method, using a probe radius of 1 Å.

Calculation of ion conductance

Request a detailed protocol

To obtain an approximate estimate of the K+ and Na+ conductance of TMEM175 channel, we used the theoretical formulation proposed by Zhu and Hummer, 2012, wherein the conductance γ is inferred from the free-energy (Fz) and diffusion (Dz) profiles as a function of the ion position along the pore axis:

(8) γ=q2CSkBTz1z2expFzkBT1Dzdz-1

where q (=1) is the charge of the permeant ion, C is the bulk ion concentration (100 mM), and S is the cross-sectional area (314 Å2) of the cylindrical region considered to evaluate Equations 3–5. As a first approximation, we assumed the diffusion profile be flat, and used Dz ~ 2 10–9 m2/s for K+ and Dz ~ 1.5 10–9 m2/s for Na+, following previous simulation studies (Zhou et al., 2017; Zhu and Hummer, 2012). Note that γ depends linearly with Dz , while the dependence on Fz is exponential; thus, while it is very plausible that the diffusion constant will vary by at least a factor of 2 as the ion traverses the pore (Zhou et al., 2017), the order of magnitude of the conductance is largely set by the features of the free-energy profile.

Calculation of free energies of selectivity

Request a detailed protocol

To evaluate free-energy differences between K+ and Na+ states, whether in bulk water or at the isoleucine constriction within the TMEM175 pore, we used the free-energy perturbation method as implemented in NAMD 2.12. To ascertain the magnitude of sampling errors, all transformations were carried out in the forward (K+ to Na+) and backward (K+ to Na+) directions. The transformations were carried out in 26 steps; each step included a 50 ps equilibration, excluded from analysis, followed by 500 ps of averaging time. For bulk water, the coupling parameter lambda was varied in increments of 0.04 in the [0, 0.8] interval and in increments of 0.02 in the [0.8, 1] interval. For the protein calculations, lambda was changed in increments of 0.04 in the [0, 0.96] interval and in increments of 0.02 in the [0.96, 1] interval. To ensure that the K+ and Na+ ions involved in the transformation are in the same position, the distance between the two particles is restrained to zero using a harmonic potential of force constant 2425.58 kcal/(mol Å2). In the protein calculations, the ion(s) are confined to remain within the isoleucine constriction using a flat-bottom distance restraint, defined relative to the center-of-mass of the backbone atoms of two groups of residues, namely 44–45, 269–270 (first group) and 47–49, 272–274 (second group). These two groups also define an axis; the restraint acts on the distance between the ion and the center-of-mass, as projected on that axis. The confining potential is flat-bottomed, permitting fluctuations of ±0.1 Å in the distance, but further deviations are suppressed with a harmonic function of force constant 100 kcal/(mol Å2).

Electrophysiological analysis

Request a detailed protocol

Electrophysiological recordings of TMEM175 constructs were performed in HEK293T cells (ATCC CRL-3216). HEK293T cells were recently purchased from ATCC and are negative for mycoplasma contamination. HEK293T cells cultured in DMEM supplemented with 10% FBS. To transfer cells in single dishes, cells were detached by trypsin treatment. The detached cells were transferred to poly-Lys-treated 35 mm single dishes (FluoroDish, World Precision Instruments) and incubated overnight at 37°C in fresh media. Cells in a single dish were transfected with 1.25 μg of c-term EGFP tagged hTMEM175 plasmid using 3.75 μg of PEI 25 k (Polysciences, Inc). Electrophysiological recordings were performed 48–72 hr after transfection. Prior to recording, media was replaced with a bath solution containing 145 mM Na-methanesulfonate (MS), 5 mM NaCl, 1 mM MgCl2, 1 mM CaCl2, 10 mM HEPES/Tris pH 7.4. Ten-cm-long borosilicate glasses were pulled and fire polished (Sutter instrument). Glass pipettes of resistances between 3 and 10 MΩ were filled with a pipette solution containing 150 mM Cs-MS, 5 mM MgCl2, 10 mM EGTA, 10 mM HEPES/Tris pH 7.4, and GΩ seals were formed after gentle suction. The recordings were performed in whole-cell patch clamp configuration using the following protocol: from a holding potential of 0 mV, the voltage was stepped to voltages between −100 and +100 mV, in 20 mV increments. To measure currents reduced by 4-AP or AP-6, bath solutions were perfused with a solution containing 1 mM 4-AP or 2 mM AP-6. DMSO concentration was maintained as 0.33% during the entire experiment. The currents were recorded using Axon Digidata 1550B digitizer and Clampex 10.6 (Molecular Devices, LLC) and analyzed using AxoGraph X 1.7.6 (AxoGraph Scientific). Each experiment was performed in a unique cell. To determine the IC50 of 6-AP, the dose-response curve for TMEM175 current at +100 mV in the presence of AP-6 was fit with the following equation:

(9) I=Imin+(Imax-Imin)×IC50IC50+[AP-6]

where I is the normalized current, Imax and Imin are maximum and minimum normalized current, respectively. [AP-6] is the concentration of AP-6 perfused.

General chemistry

Request a detailed protocol

Chemical reagents and materials were purchased from Sigma-Aldrich, Thermo-Fisher, and TCI, and used without further purifications. DCM (dichloromethane), MeOH, and n-hexane for column chromatography and recrystallization were used for HPLC grade without additional purifications. Thin layer chromatography (TLC) analysis was performed for reaction monitoring on the pre-coated silica gel 60 F254 glass plates. Both starting materials and the desired product were checked by UV light (254 nm). Flash column chromatography was carried out on silica gel (400–630 mesh) to separate the target molecule.

Synthesis of compound 1 2,2'-(1,3-phenylene)bis(pyridin-4-amine) (AP-6)

Request a detailed protocol

AP-6 was prepared through Suzuki-Miyaura cross-coupling reaction with palladium catalysis. In a round-bottom flask with a magnetic bar, a mixture of 2-bromopyridin-4-amine (5 mmol, 865 mg), 1,3-phenylenediboronic acid (0.75 equiv., 622 mg), K2CO3 (2 equiv., 1.38 g), and Pd(OAc)2 (7 mol%, 78.6 mg) were dissolved in H2O:EtOH solution mixture (8 mL: 32 mL), and the solution was stirred at 100°C for 24 hr under air. After completion (monitored by TLC), the reaction mixture was filtered through a Celite (after cooling to room temperature), and then the solid on the filter was washed with EtOAc. The mixture was added to brine and extracted with EtOAc for three times. The combined organic layer was dried over MgSO4, filtered, and concentrated under reduced pressure. The residue was purified by chromatography in silica gel (DCM/MeOH), and purified again with recrystallization (n-hexane/DCM) to give the desired product.

Proton nuclear magnetic resonance (1H NMR) spectra were recorded on Bruker AVANCE 500 (500 MHz). In addition, 13C{1H} NMR was measured on the same machine (125 MHz), and the spectra was fully decoupled with proton by broad band decoupling. Chemical shifts for NMR were quoted in parts per million referenced to the appropriate solvent peak (DMSO in DMSO-d6). The abbreviation codes were adopted to describe 1H NMR peak patterns; d = doublet, t = triplet, and br = broad. Coupling constants, J, were displayed in Hertz unit (Hz). Infrared (IR) spectra were recorded on Bruker Alpha FT-IR spectrometer. High-resolution mass spectra were acquired on a high-resolution Q-TOF mass spectrometer (ionization mode: ESI).

Appendix 1

Appendix 1—table 1
Cryo-EM data acquisition, reconstruction, and model refinement statistics.
TMEM175OpenTMEM175Closed
Cryo-EM acquisition and processing
EMDB accession #EMD-26626EMD-26627
Magnification22,500×22,500×
Voltage (kV)300300
Total electron exposure (e-/ Å2)6161
Exposure time (s)88
Defocus range (µM)−1.0 to −2.5−1.0 to −2.5
Pixel size (Å)1.0881.088
Final pixel size (Å)0.850.85
Symmetry imposedC2C2
Initial particles4,153,6144,153,614
Final particles261,536163,651
Resolution (masked, Å)2.452.61
Density modified CC (0.5, Å)2.432.65
Model refinement
PDB ID7UNL7UNM
Model resolution (Å)2.48/1.962.88/2.49
FSC threshold0.50/0.1430.50/0.143
Model refinement resolution300–2.4300–2.6
RMS deviations
 Bond length (Å)0.0030.002
 Bond angle (°)0.5590.433
Ramachandran plot
 Favored (%)98.999.45
 Allowed (%)1.10.55
 Disallowed (%)00
Rotamer outliers (%)1.940
Validation
 MolProbity score1.221.13
 Clashscore2.243.44

Data availability

Cryo-EM maps and atomic coordinates have been deposited with the EMDB and PDB under accession codes EMD-26626 and PDB 7UNL for open TMEM175 and codes EMD-26627 and PDB 7UNM for closed TMEM175. Source data have been provided for Figure 5.

The following data sets were generated
    1. Oh S
    2. Hite RK
    (2022) RCSB Protein Data Bank
    ID 7UNL. Open state TMEM175 model.
    1. Oh S
    2. Hite RK
    (2022) EMDataResource
    ID EMD-26626. Open state TMEM175 map.
    1. Oh S
    2. Hite RK
    (2022) RCSB Protein Data Bank
    ID 7UNM. Closed state TMEM175 model.
    1. Oh S
    2. Hite RK
    (2022) EMDataResource
    ID EMD-26627. Closed state TMEM175 map.
The following previously published data sets were used
    1. Oh S
    2. Hite RK
    (2020) RCSB Protein Data Bank
    ID 6WC9. Open state TMEM175 model.

References

    1. Emsley P
    2. Lohkamp B
    3. Scott WG
    4. Cowtan K
    (2010) Features and development of Coot
    Acta Crystallographica. Section D, Biological Crystallography 66:486–501.
    https://doi.org/10.1107/S0907444910007493

Decision letter

  1. Lucie Delemotte
    Reviewing Editor; KTH Royal Institute of Technology, Sweden
  2. Richard W Aldrich
    Senior Editor; The University of Texas at Austin, United States

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Differential ion dehydration energetics explains selectivity in the non-canonical lysosomal K+ channel TMEM175" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Richard Aldrich as the Senior Editor. The reviewers have opted to remain anonymous.

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:

1) Improved justification of methodological choices.

2) A characterization of the uncertainties on the free energies.

3) A demonstration that a reasonable choice of a different force field would give similar results.

4) A more thorough discussion of agreement (or lack thereof) with other (e.g., experimental) data.

Reviewer #1 (Recommendations for the authors):

1) The Introduction reads as if the authors were the first and only group of determining the structure of a TMEM175 channel. While they were the first for the human isoform, two bacterial isoforms have been structurally characterized and were found to have almost identical pore structure. Those works also formulated hypotheses about the mechanisms of selectivity and gating; both identified the isoleucine constriction. One of these articles is mentioned in passing (Lee et al., 2017, and it's not becoming clear that this was a structural work), while the other one is not cited at all in the entire manuscript (Brunner et al. 2020, eLife 9, the authors cite a preprint of this work in their 2020 article). Also, the only previous theoretical work on the TMEM175 (Rao et al. 2018 in Faraday Discuss. 209) is not cited. With only a low single-digit number papers available on the structure and mechanism of TMEM175, they should be given credit. The same applies for the discussion; a discussion of these new results in relation to the literature is lacking.

2) The section "Improved cryo-EM structures of hTMRM175" is a follow-up to the previous paper (Oh et al., 2020), reporting on a refined on the crystal structure data.

a) It is unclear to me how the new and improved structural model relates to that one from last year. The authors merely mention that it is similar.

b) Were there any changes in the structure or ion position due to the new refinement, or were the now better resolved densities in the expected places? Are the ion binding sites K1-K4 in the same position as in the previous work? Since the notation K1-K4 was dropped (why?), it is very hard for the reader to tell by comparison.

3) Sections "Energetics and mechanism of K+ permeation" and "Differentials in ion dehydration relative to bulk explain selectivity for K+ over Na+"

a) It is not mentioned whether the minima in the energy surfaces or the binding sites observed in the time traces coincide with the ion binding sites in the cryoEM structure. This is important in the light that the authors specifically point out that they made no a priori assumptions about this.

b) The comparison of K+ and Na+ comes a bit short. Although the degree of water depletion is different for K+ and Na+, they end up at the same number of 4 remaining waters. Coordination by the protein is also very similar and it stays unclear whether these two points are relevant for selectivity. Permeation simulations for both K+ and Na+ were performed, but the numerical results (number of permeation events, estimated conductivity) are only given for K+, so the comparison of both ions remains somewhat vague.

c) How appropriate is it to use a constant electric field for a channel with an hourglass pore, where most of the field drops likely over the narrow constriction?

d) The Supplement 2 to Figure 2 is not very helpful. It is very crowded. I was able to pick out some instances where one line seems to cross the entire gap in the middle, which I presume is an ion transition. But I see those both for K+ and Na+. In Figure 3, Na+ is also seen to permeate. How are the statistics of permeation events for K+ and Na+? What is the estimated conduction for 100 mM Na+? I am aware that permeation does not equal selectivity, but I am curious as to why this basic information was left out.

4) Section "Role of the isoleucine constriction in ion selectivity and permeation"

a) Brunner et al. (2020) showed experimentally that not only the isoleucine, but also a stack of polar residues strongly contribute to selectivity in both the bacterial and human isoforms. They also formulated a hypothesis on the gating mechanism. Does this contradict or complement the results of this paper? The hypothetical open state modelled by Rao et al. (2018) is also not mentioned.

b) The new blocker AP-6 seems to be indeed slightly more effective at blocking the mutants than 4-AP. But it seems do drop out of thin air. Neither the design rationale nor whether it is specific to TMEM175 is discussed.

Reviewer #2 (Recommendations for the authors):

1) Error estimates for the calculated free energies and free energy profiles should be provided to see if found differences of ca. 1 kcal/mol are statistically significant. Similarly, the applied field simulations show two permeation events, a rather low number, that should also be treated with care.

2) The differences of ca. 1 kcal/mol are around the accuracy of current force fields for free energy predictions, and the authors are using only one force field (CHARMM) for a somewhat difficult target (an ion in a hydrophobic environment) that was not explicitly parameterized in this force field. Given that in the field of canonical potassium channels force field inaccuracies led to some 20 years long discussion about the permeation mechanisms, I'd suggest to either include an investigation with another force field, or tone down the conclusions (for example that this is a 'proposed permeation mechanism for TMEM175') and discuss potential issues with the current methodology. A recent paper showing a likely too hydrophobic character of the TMEM175 cavity in nonpolarizable force fields (Lynch et al., ACS Nano 2021) will be also of interest here.

3) The authors state in the abstract that the channel is 'capable of permeating K+ ions at the expected rate' and later (page 12) their estimated conductance of human TMEM175 is 0.1-0.5 pS, which the authors comment as "well within the broad range of experimentally measured permeation rates". Can the authors actually provide some numbers here? For example, the bacterial TMEM175 shows a conductance of ca. 70 pS (Brunner et al. eLife 2020) and Slo1 K+ ca. 100 pS (Tao et al. Nature 2017).

4) For clarity, I suggest calculating the conductance for sodium from PMFs the way it was done for potassium; then, compare the K+/Na+ permeability ratios obtained from MD with the experiment. If possible, the error estimates for the permeability ratios should be included as well.

5) Throughout the manuscript, the authors sometimes use the thermodynamic description (i.e. differences in binding free energies between K+ and Na+ ions at a given site) to explain experimentally measured ion selectivity (i.e. permeation rates), that naturally also contains the kinetic part (height of free energy barriers). For example: 'it is useful to keep in mind that dehydration of Na + is much more costly than that of K + ; the penalty for full dehydration is ~18 kcal/mol greater, rendering it a trillion-fold less probable. However, the observed selectivity of biological K+channels against Na+is only 1000-fold or less, implying Na+ establishes much more favorable interactions with the selectivity filters of this kind of K+ channels.'

I'd be careful to directly connect Na+ dehydration penalty with experimentally observed permeation rates, and make a clear distinction in the manuscript between kinetic and thermodynamic selectivity derived from MD. Moreover 'Na + establishes much more favorable interactions with the selectivity filter of this kind of K+ channels' is not only implied but actually shown before, see free energy calculations in Kim et al., PNAS 2011 and Kopec et al. Nat. Chem. 2018.

6) The authors seem to ignore several interesting insights into ion selectivity of TMEM175 channels reported recently by Brunner et al. eLife 2020 – the paper is not even cited. It would be very beneficial for the whole field to include a discussion on how the authors' mechanism and findings agree (or not) with those of Brunner et al.

7) line 78 – table S1 is absent in the manuscript.

8) line 129 – "we detected no evidence of a multi-ion process (Figure 2—figure supplement 2)" – I assume it refers to the number of ions present at the same time at the level of the constriction, however it is not clear from the figure. I suggest defining a "multi-ion" permeation process in the text and put coordinates of SF on Figure 2—figure supplement 2 to make it clear.

9) line 203 – "specifically, the calculations indicate this mutant is about 2-fold less K+ selective than the wild-type channel" – it is not calculated in the manuscript at the moment.

10) lines 177-181 – "We observe that the depletion in the both the first and second solvation shells, relative to bulk numbers, is significantly smaller for Na+than for K+. As will be further discussed below, this relative difference in dehydration energetics likely explains whyTMEM175 is only~30-fold more permeable to K+, even though the bulk water selectivity for Na+is over a trillion-fold." – it may not be clear how the number of water molecules lost during dehydration affects selectivity by itself (rather than the free energy difference between the processes of going from the bulk to the constriction). If it's the case, it should be explained more clearly.

11) Figure 5 Supplementary Figure 3 – the points for the raw currents before adding AP-6 are much more scattered than before 4-AP. What could be the reason for this behavior?

6. line 492 – "To measure currents reduced by AP-6, bath solutions were perfused" – the method for measuring the currents reduced by 4-AP are not described.

12) The starting structure for the simulations is stated to be 6WC9, the previously published structure of the open TMEM175, even though structures with higher resolution were obtained in this study. The reasoning behind using a lower-resolution structure should be provided.

13) Even though the positions of ions as a function of time are shown in Figure 2 Supplementary Figure 2, it may not be enough to estimate the convergence of metadynamics simulations. Can the authors provide an example of the time dependence of the CVs, or the deposited potential, or some other suitable measure for convergence as well?

14) As another control for the metadynamics simulations, would it be possible to run MD of TMEM175 with the potassium ions and water molecules not removed from the initial structure? It could show if there are some energy minima not resolved by metadynamics, and if those ions/water molecules have any effect on the overall behavior of TMEM175. Also, would you expect any side effects from not modelling residues 164-253?

Reviewer #3 (Recommendations for the authors):

1) While the refined static structure of the open channel is discussed in detail, the conformational dynamics of the pore is not: was there any conformational isomerization of side chains? What is the extent of fluctuations in pore radius and relative helix arrangement? Do the apparent kink and tilt of the pore helices fluctuate?

2) There seems to be a discrepancy between the fully-hydrated state of ions in the cryoEM densities (lines 95-97, Figure 1 Suppl. Figure 2) and their partial dehydration in the simulations (Figure 2). This is not a trivial point, since the major finding of the paper is that selective K+ permeation arises from ion (de)solvation effects.

3) Systematic error analysis of the simulation results would strengthen the confidence in the numerical agreement noted above.

4) A broader discussion of the general significance of the findings, including the role of ion desolvation effects and the novelty that a hydrophobic locus controls both gating and selectivity in K+ channels and ion channels in general, would be welcome.

5) The introduction could provide a bit more background. What is the biological function of this protein? Only dysfunction is mentioned.

6) Please specify which atoms were used for the analysis shown in Figure 2 Suppl. Figure 1 (Calpha atoms, etc).

7) It would be useful to computational biophysicists if the authors could clarify the rationale for specific methodological choices. In particular, what are the considerations that presided over the choice of force field? Same question for the collective variable used in the metadynamics simulations.

8) Likewise, why did the authors use the particular functional form of Equation 6 to compute the ion coordination number rather than a simple cut-off distance? Please provide references for that choice if appropriate. Also, please note that brackets are missing in the summations in Eq. 6.

9) Error estimates should be provided systematically. In the current manuscript, they are sometimes vague or non-existent. In particular:

– Please provide data for estimating the convergence for metadynamics; and generally, error estimates as appropriate for all the numerical results.

– The penultimate sentence in the caption of Figure 2 is unclear: "gray profiles represent the same quantity shown in black/blue calculated using only the first or second half of the simulations data."

10) Please provide a reference for variations in the value of the diffusion constant (see lines 453-454).

11) I would recommend including a schematic figure of a thermodynamic cycle to the Discussion for clarity and to highlight the consistency between the two different pathways followed for the DeltaDeltaG calculation (see "Strengths" in public review above).

12) The statement in lines 236-237 seems to imply that the balance of channel-ion, ion-ion, and water-ion interactions controls selective ion permeation. The authors may consider including the effect of channel-water (and arguably also water-water and channel-channel) interactions for the sake of completeness and generality.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Differential ion dehydration energetics explains selectivity in the non-canonical lysosomal K+ channel TMEM175" for further consideration by eLife. Your revised article has been evaluated by Richard Aldrich (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below. We anticipate acceptance after these changes have been implemented.

Please address the comments by reviewers 1 and 3, as detailed below.

In addition, regarding the justification of the force field choice, we wish to highlight that claiming that default CHARMM ion parameters from 1994 (Beglov and Roux, 1994) reproduce the relative free-energies of hydration (and thus dehydration) of the alkali cations 'with excellent accuracy' somewhat invalidates next 20 years of ion force field development. When addressing the issue in the article's text, as requested by reviewer 3, please ensure the comment is accurate and well-substantiated, and do not forget to mention possible shortcomings of the chosen force field.

Reviewer #1 (Recommendations for the authors):

I thank the authors for their careful revision of the manuscript. Most of my previous comments have been addressed adequately, except for these two:

Regarding the position of the ion binding sites in the different figures created by different methods: Noting the residues in Figure 2 is certainly helpful (as has already been done in the original manuscript). However, indicating K1-K5 directly as I suggested in my initial review would spare the reader to flip back and forth the figures and comparing each single residue manually. There is still no indication of the binding sites in the time traces of the MD simulations, my request "All figures where this applies: please indicate the location of the binding sites from the structural data" has been ignored by the authors without any comment.

I still strongly suggest amending the figures accordingly as this would make them more accessible.

This comment of mine has maybe been overlooked by the authors: What is the meaning of the numbers in the bracket in Figure 3A? I presume the number of water molecules?

Reviewer #2 (Recommendations for the authors):

All points have been adequately addressed. I recommend the paper for publication.

Reviewer #3 (Recommendations for the authors):

The authors have addressed most of the comments appropriately.

In their response concerning the choice of forcefield, they write: "It could be convincingly argued that a mechanism that relies on dehydration energetics, rather than specific ion-protein interactions, will be reasonably described by this forcefield, as it is in fact parametrized, with excellent accuracy, to reproduce the relative free energies of hydration (and thus dehydration) of the alkali cations." That is the kind of justification that I was hoping for. Please make that convincing argument in the text, including references as needed.

https://doi.org/10.7554/eLife.75122.sa1

Author response

Essential revisions:

1) Improved justification of methodological choices.

We thank the reviewers for their questions. We have included the clarifications suggested by the reviewers below, and when appropriate, in the revised manuscript.

2) A characterization of the uncertainties on the free energies.

As requested, we now have clarified how we had estimated the uncertainties of the free energies. For the Metadynamics calculations described in Figure 2, we had plotted free-energy profiles based on either the first 50% or the second 50% of all trajectory data. Comparison of these two independent profiles provides a straightforward metric of convergence and hence statistical error. The statistical uncertainties of the free-energy barrier for permeation of K+ and Na+ are 0.14 kcal/mol and 0.22 kcal/mol, respectively. These data confirm that the statistical error of the calculations is smaller than the difference in the free energy between K+ and Na+.

In Figure 4, which reports on the FEP calculations, we had superposed the results for the forward and backward transformations. By comparing these plots, we can obtain a simple metric of convergence for this type of calculation. As in Figure 2., the statistical errors in these calculations (0.1 kcal/mol) are much smaller than any of the computed free-energy differences. The excellent agreement between Metadynamics calculations and FEP yield, despite being radically different methodologies, is further evidence that the statistical error in these calculations is minimal. We thank the reviewers for encouraging us to clarify these important analyses.

3) A demonstration that a reasonable choice of a different force field would give similar results.

We thank the editor and reviewers for their suggestion. However, for the reasons enumerated below we disagree that replicating the simulations using a different forcefield should be essential for publication of this work:

1. The reviews do not articulate what specific concern with the forcefield used in our simulations could, in the reviewers’ opinion, significantly alter the conclusions of our study. Reviewer #2 points that out that “an ion in a hydrophobic environment […] was not explicitly parameterized in [the CHARMM36] force field”. Indeed, no general forcefield is explicitly parameterized in any one specific context. The premise of all general forcefields used in molecular simulations of biomolecular systems is precisely that they will yield plausible results even when they are not explicitly parameterized for the structure being examined or for a specific problem of interest. Requiring that studies employ explicitly parameterized forcefields, which are rare, would greatly limit the use of simulations for analyzing biomolecular systems. Problem-dependent corrections to a general forcefield can and have been of course developed to improve the accuracy of specific elements when evidence exists that these elements are not only incorrectly captured but also critical. Two examples from our own laboratory are a comprehensive examination of anion-protein interactions (Orabi et al., JCTC 2021) as well as sodium- and calcium-protein interactions (Liao et al., NSMB 2016). Analogous corrections have been used to examine channels wherein K+ ions are directly multi-coordinated by carbonyls or carboxyl groups (Noskov et al., Nature 2001 among others). For the specific process examined in our study, however, we at this time have no reason to question the approximate validity of the CHARMM36 forcefield. Indeed, it could be convincingly argued that a mechanism that relies on dehydration energetics, rather than specific ion-protein interactions, will be reasonably described by this forcefield, as it is in fact parameterized, with excellent accuracy, to reproduce the relative free-energies of hydration (and thus dehydration) of the alkali cations.

2. This request appears to imply that the plausibility of a simulation study is questionable unless its conclusions can be replicated using different forcefields. Are all forcefields equally predictive, for any given problem? There is no reason to assume that they are. Then why should a result be the same for different forcefields? And how does one determine which forcefields must be compared, and how many? Should one compare forcefields that are similar in essence, like CHARMM and AMBER? Or would the critique then be that they are too similar? Or should one select forcefields that are fundamentally different – like fixed-charge vs polarizable – despite the fact they might be at radically different stages of development and testing? For example, our own recent attempts to calculate differences in the free energy of interaction between ions and proteins using the Drude/CHARMM36 polarizable forcefield, with NAMD, revealed critical flaws in the implementation of the algorithm, as this calculation had apparently not been attempted before. This situation illustrates the difficulties and potential pitfalls of this type of comparison. It is unquestionable that simulation studies are inherently hypothetical, regardless of forcefield, or the myriad other assumptions and simplifications adopted in these studies. Reviewer #2 points out that “in the field of canonical potassium channels force field inaccuracies led to some 20 years long discussion about the permeation mechanisms”. Well, that is the fundamental nature of the scientific enterprise; the same could be said about most important problems, whether examined with experimental or computational techniques. The way to deal with the uncertainties inherent to the scientific enterprise is careful presentation and interpretation. In the case of simulation work, it needs to be understood that it does not result in definitive answers to a given problem, but rather a working hypothesis based on a specific model. What makes a computational study merit publication is not that it can be replicated against all possible models – but that it presents results that are statistically significant and reproducible for a given set of conditions, obtained with state-of-the-art quantitative methods, and most importantly, that it provides a plausible – but necessarily tentative – interpretation to experiment, or a verifiable prediction.

3. It is not the editorial policy of eLife, or other reputable journals, to generically require that simulation studies be replicated with different forcefields in order to merit publication, without a specific concern or justification. As articulated above, there are good reasons that explain why that is the case.

4) A more thorough discussion of agreement (or lack thereof) with other (e.g., experimental) data.

We thank the reviewers for encouraging us to provide additional discussion of our work. We now add several additional sections to the text. In the introduction, we describe the previous structural studies on prokaryotic TMEM175 channels and highlight some of the important structural and functional differences between prokaryotic TMEM175 and human TMEM175 channels. In the discussion we state that our results are consistent with the suggestion of Lee and colleagues that a hydrophobic constriction imparts ion selectivity on TMEM175 channels but disagree with Brunner and colleagues who suggest that the hydrophobic constrictions serve exclusively as channel gates while direct protein-ion interactions impart ion selectivity.

In our previous work, we observed layers of ordered water molecules immediately above and below the isoleucine constriction in the putative open state that were partially coordinated by the side chains of polar residues including Ser45 and Thr274. Based on this observation, we mutated Ser45 conservatively to threonine and non-conservatively to alanine. We similarly mutated Thr274 to serine and valine. For both residues, the non-conservative mutations resulted in a loss of channel activity, while the conservative mutations had a lesser effect. The S45T and T274S mutations also resulted in a slight reduction in ion selectivity, which may have resulted from greater influence of the non-selective endogenous currents on the smaller currents from the mutant channels. We therefore proposed that Ser45 and Thr274 are critical for ion permeation due to their role in stabilizing water molecules near the isoleucine constriction; thus, the permeating ion retains its first hydration shell except when traversing the constriction.

In agreement with our work, Brunner and colleagues proposed that Ile46/Ile271 form the channel gate. However, they did not observe a similar loss of activity for the S45A mutant or a T49A/T274A double-mutant, but rather a loss of selectivity for K+ over Na+. We are currently unclear of the origin of the conflicting results between the groups regarding the effect of mutating Ser45 and Thr49/Thr274. However, as we observe only minimal interactions between Ser45, Thr49 or Thr274 and the K+ ions in the simulations (Figure 3C), the computational results are consistent with our previous conclusion that Ser45 or Thr49/Thr274 do not have a strong role in establishing ion selectivity. Future studies will be required to uncover the basis of the disagreements.

Reviewer #1 (Recommendations for the authors):

1) The Introduction reads as if the authors were the first and only group of determining the structure of a TMEM175 channel. While they were the first for the human isoform, two bacterial isoforms have been structurally characterized and were found to have almost identical pore structure. Those works also formulated hypotheses about the mechanisms of selectivity and gating; both identified the isoleucine constriction. One of these articles is mentioned in passing (Lee et al., 2017, and it's not becoming clear that this was a structural work), while the other one is not cited at all in the entire manuscript (Brunner et al. 2020, eLife 9, the authors cite a preprint of this work in their 2020 article). Also, the only previous theoretical work on the TMEM175 (Rao et al. 2018 in Faraday Discuss. 209) is not cited. With only a low single-digit number papers available on the structure and mechanism of TMEM175, they should be given credit. The same applies for the discussion; a discussion of these new results in relation to the literature is lacking.

We thank the reviewer for their suggestion. We have added a section to the introduction describing the previous structural studies on prokaryotic TMEM175 channels and the structural and functional differences between human TMEM175 and prokaryotic TMEM175 homologs, focusing on their different pore structures, oligomeric state and selectivity profiles. We have also added a section to the discussion comparing our results with the findings described by other groups.

2) The section "Improved cryo-EM structures of hTMRM175" is a follow-up to the previous paper (Oh et al., 2020), reporting on a refined on the crystal structure data.

a) It is unclear to me how the new and improved structural model relates to that one from last year. The authors merely mention that it is similar.

b) Were there any changes in the structure or ion position due to the new refinement, or were the now better resolved densities in the expected places? Are the ion binding sites K1-K4 in the same position as in the previous work? Since the notation K1-K4 was dropped (why?), it is very hard for the reader to tell by comparison.

We thank the reviewer for their suggestions. The RMSD between the protein atoms of the updated open structure and the previously published open structure (PDB:6WC9) is 0.3 Å and the and ion-binding sites are all located within 0.6 Å of their previously assigned positions. The RMSD between the protein atoms of the updated closed structure and the previously published closed structure (PDB:6WCA) is 0.4 Å and the ion-binding sites are all located within 1 Å of their previously assigned positions.

We have added labels to Figure 1 for the ion binding sites.

3) Sections "Energetics and mechanism of K+ permeation" and "Differentials in ion dehydration relative to bulk explain selectivity for K+ over Na+"

a) It is not mentioned whether the minima in the energy surfaces or the binding sites observed in the time traces coincide with the ion binding sites in the cryoEM structure. This is important in the light that the authors specifically point out that they made no a priori assumptions about this.

We thank the reviewer for their suggestion and have revised the manuscript to clarify this question and highlight the similarities of these two orthogonal approaches to identify ion binding sites in the pore of TMEM175. The position of residues lining the permeation pathway is indicated alongside the density in the revised version of Figure 1 and alongside the free-energy profile calculated for K+ in Figure 2. There is very good agreement between these observations. Densities near S38 and A42 on the cytoplasmic side and near V50 on the luminal side appear as metastable states in the calculated free-energy curve. Densities right below and above the central constriction also appear as shoulders in the free-energy profile.

b) The comparison of K+ and Na+ comes a bit short. Although the degree of water depletion is different for K+ and Na+, they end up at the same number of 4 remaining waters. Coordination by the protein is also very similar and it stays unclear whether these two points are relevant for selectivity. Permeation simulations for both K+ and Na+ were performed, but the numerical results (number of permeation events, estimated conductivity) are only given for K+, so the comparison of both ions remains somewhat vague.

We thank the reviewer for highlighting this important finding of our work. It is precisely because K+ and Na+ interact similarly with the protein (unlike other K+ or Na+ channels) that we argue that the selectivity of this channel primarily stems from differences in their dehydration energetics. That in both cases the hydration shells are depleted down to the same number of water molecules at and near the constriction in fact implies different energetic penalties relative to bulk water – which as we show are sufficient to explain the measured difference in permeability. A direct metric of these differences is provided by the free-energy calculation in which one ion is replaced by the other while residing at the constriction, relative to the same exchange in bulk water – see Figure 4. In summary, the different degree of depletion is not an inconsequential observation; it is actually a very important factor and one of the unique features of this channel.

c) How appropriate is it to use a constant electric field for a channel with an hourglass pore, where most of the field drops likely over the narrow constriction?

The physical quantity that drops over a narrow constriction in a transmembrane pore is not the applied electric field – it is the electrostatic potential. Specifically, it is the potential resulting not only from the applied field but also from the response of the molecular system to this applied field. Thus, an ion diffusing in bulk water or within water-accessible regions of a wide pore will experience essentially no change in the surrounding electrostatic potential, even if the applied electric field is constant across the same region. The reason is that water is highly dielectric, and the applied electric field is largely countered (or screened) by the field that results from the preferential orientation of water molecules in response to the applied field. By contrast, as a pore becomes narrower the orientational freedom of water and thus its screening effect are diminished, and so the net electric field within is non-zero; an ion traveling from a wider region into this narrower region will thus experience a change in electrostatic potential. How to correctly simulate transmembrane potentials has been discussed at length in the existing literature. See for example Roux, Biophys J 2008, or Gumbart et al., BBA Biomembranes 2012. These and other studies “demonstrate that the constant-field method is a simple and valid approach for accounting for the membrane potential in molecular dynamics studies of biomolecular systems”.

d) The Supplement 2 to Figure 2 is not very helpful. It is very crowded. I was able to pick out some instances where one line seems to cross the entire gap in the middle, which I presume is an ion transition. But I see those both for K+ and Na+.

Trajectories that cross the central gap are indeed crossing events. The figure captions have been revised to clarify this question. That we see crossings for both K+ and Na+ is expected as the channel is permeable to both ions. It should be however noted that the free-energy profiles shown in Figure 2 are based on the totality of these data, not only crossing events. It is for that reason that we have opted to preserve these plots in the revised version. For clarity the revised manuscript now includes two movies depicting ion permeation events (induced by voltage).

In Figure 3, Na+ is also seen to permeate. How are the statistics of permeation events for K+ and Na+? What is the estimated conduction for 100 mM Na+? I am aware that permeation does not equal selectivity, but I am curious as to why this basic information was left out.

We did not simulate Na+ permeation under voltage. Na+ permeation was examined only at 0 mV, using enhanced-sampling simulations – see Figure 2 and related figure supplements. From this data, the estimated conduction for Na+ at 100 mV is 0.04 pS, which is 5-fold smaller than what we estimate for K+. Both estimates are indicated in the revised manuscript.

4) Section "Role of the isoleucine constriction in ion selectivity and permeation"

a) Brunner et al. (2020) showed experimentally that not only the isoleucine, but also a stack of polar residues strongly contribute to selectivity in both the bacterial and human isoforms. They also formulated a hypothesis on the gating mechanism. Does this contradict or complement the results of this paper? The hypothetical open state modelled by Rao et al. (2018) is also not mentioned.

In our previous work, we observed layers of water molecules immediately above and below the isoleucine constriction that were partially coordinated by the side chains of polar residues including Ser45 and Thr274. Based on this observation, we hypothesized that the water molecules may contribute to ion permeation. To test the hypothesis, we mutated Ser45 conservatively to threonine and non-conservatively to alanine. We similarly mutated Thr274 to serine and valine. For both residues, the non-conservative mutations resulted in a loss of channel activity, while the conservative mutations had a lesser effect. The S45T and T274S mutations also resulted in a slight reduction in ion selectivity, which may have resulted from increased influence of the endogenous currents on the smaller currents. We therefore proposed that Ser45 and Thr274 are critical for ion permeation due to their role in stabilizing water molecules near the isoleucine constriction.

In agreement with our work, Brunner and colleagues proposed that Ile46/Ile271 form the channel gate. However, they did not observe a similar loss of activity for the S45A mutant or a T49A/T274A double-mutant, but rather a loss of selectivity for K+ over Na+. We are currently unclear of the origin of the conflicting results between the groups regarding the effect of mutating Ser45 and Thr49/Thr274. However, as we observe only minimal interactions between Ser45, Thr49 or Thr274 and the K+ ions in the simulations (Figure 3C), the computational results are consistent with our previous conclusion that Ser45 or Thr49/Thr274 do not have a strong role in establishing ion selectivity.

The hypothetical open state modelled by Rao and colleagues is a model of a prokaryotic TMEM175 homolog. As the pore structures between human and prokaryotic TMEM175 channels differ due to sequence divergence and oligomeric state, our work on human TMEM175 does not inform on the structure of the open state of a prokaryotic TMEM175 channel. Moreover, as the selectivity profiles of human TMEM175 and prokaryotic TMEM175 are distinct and the mechanisms of gating in prokaryotic TMEM175 channels are unknown, our analyses do not inform on their ion selectivity and gating mechanisms. Due to these fundamental differences between human and prokaryotic TMEM175 channels, we would prefer not to speculate on prokaryotic TMEM175 channels.

b) The new blocker AP-6 seems to be indeed slightly more effective at blocking the mutants than 4-AP. But it seems do drop out of thin air. Neither the design rationale nor whether it is specific to TMEM175 is discussed.

We are preparing a manuscript describing our studies into describing novel TMEM175 inhibitors, including AP-6 in which the rationale and specificity of AP-6 will be characterized.

Reviewer #2 (Recommendations for the authors):

1) Error estimates for the calculated free energies and free energy profiles should be provided to see if found differences of ca. 1 kcal/mol are statistically significant. Similarly, the applied field simulations show two permeation events, a rather low number, that should also be treated with care.

Please see our response to the editor’s summary above. We have revised the captions of Figure 2 and Figure 4 to describe how our error calculations were performed. The mean difference between PMF profiles obtained with the first or second half of the trajectory data for K+ is 0.14 kcal/mol; for Na+, it is 0.22 kcal/mol. The difference between forward and backward FEP transformations is 0.1 kcal/mol. Differences of ca. 1 kcal/mol between K+ and Na+ are therefore statistically significant.

2) The differences of ca. 1 kcal/mol are around the accuracy of current force fields for free energy predictions, and the authors are using only one force field (CHARMM) for a somewhat difficult target (an ion in a hydrophobic environment) that was not explicitly parameterized in this force field. Given that in the field of canonical potassium channels force field inaccuracies led to some 20 years long discussion about the permeation mechanisms, I'd suggest to either include an investigation with another force field, or tone down the conclusions (for example that this is a 'proposed permeation mechanism for TMEM175') and discuss potential issues with the current methodology. A recent paper showing a likely too hydrophobic character of the TMEM175 cavity in nonpolarizable force fields (Lynch et al., ACS Nano 2021) will be also of interest here.

Please see our response to the editor’s summary above. We have revised the text to underscore we propose a mechanism of selectivity and that the mechanism is supported by three approaches, but that future studies will be helpful in evaluating our proposed mechanism.

3) The authors state in the abstract that the channel is 'capable of permeating K+ ions at the expected rate' and later (page 12) their estimated conductance of human TMEM175 is 0.1-0.5 pS, which the authors comment as "well within the broad range of experimentally measured permeation rates". Can the authors actually provide some numbers here? For example, the bacterial TMEM175 shows a conductance of ca. 70 pS (Brunner et al. eLife 2020) and Slo1 K+ ca. 100 pS (Tao et al. Nature 2017).

We thank the reviewer for their suggestion and have revised the text to include several examples covering the wide range of conductance values measured for cation channels.

4) For clarity, I suggest calculating the conductance for sodium from PMFs the way it was done for potassium; then, compare the K+/Na+ permeability ratios obtained from MD with the experiment. If possible, the error estimates for the permeability ratios should be included as well.

We thank the reviewer for their suggestion. When examined at 0 mV using enhanced-sampling simulations, the estimated conduction for Na+ at 100 mV is 0.04 pS. The calculated K+/Na+ permeability ratio is thus 5-fold, which is similar to measurements that others have reported (10-36). We have included both estimates in the revised manuscript.

5) Throughout the manuscript, the authors sometimes use the thermodynamic description (i.e. differences in binding free energies between K+ and Na+ ions at a given site) to explain experimentally measured ion selectivity (i.e. permeation rates), that naturally also contains the kinetic part (height of free energy barriers). For example: 'it is useful to keep in mind that dehydration of Na + is much more costly than that of K + ; the penalty for full dehydration is ~18 kcal/mol greater, rendering it a trillion-fold less probable. However, the observed selectivity of biological K+channels against Na+is only 1000-fold or less, implying Na+ establishes much more favorable interactions with the selectivity filters of this kind of K+ channels.'

I'd be careful to directly connect Na+ dehydration penalty with experimentally observed permeation rates, and make a clear distinction in the manuscript between kinetic and thermodynamic selectivity derived from MD.

We thank the reviewer for highlighting this important distinction. We designed and carried out two very different calculations precisely to evaluate alternative potential explanations. PMF profiles derived from Metadynamics trajectories show that for both ions the rate permeation is controlled by a single free-energy barrier; the height of that barrier is greater for Na+ than K+, explaining why the channel is more permeable to K+. That difference is recapitulated by the FEP calculations, which by construction only probe the barrier top (relative to bulk), and which in this case reflect mostly differences in dehydration energetics. Taken together, these results support our working hypothesis that differences in dehydration energetics explain the selectivity of this channel.

Moreover 'Na + establishes much more favorable interactions with the selectivity filter of this kind of K+ channels' is not only implied but actually shown before, see free energy calculations in Kim et al., PNAS 2011 and Kopec et al. Nat. Chem. 2018.

These references have been added to the bibliography.

6) The authors seem to ignore several interesting insights into ion selectivity of TMEM175 channels reported recently by Brunner et al. eLife 2020 – the paper is not even cited. It would be very beneficial for the whole field to include a discussion on how the authors' mechanism and findings agree (or not) with those of Brunner et al.

Please see our response to reviewer 1 above regarding the work from Brunner and colleagues. While our work disagrees with the studies described by Brunner and colleagues, our work supports the hypothesis of Lee and colleagues that a narrow hydrophobic sieve and favorable electric field can together facilitate the selective permeation of K+ ions.

7) line 78 – table S1 is absent in the manuscript.

Table S1 was inadvertently left out of the original submission and has been added the revised version.

8) line 129 – "we detected no evidence of a multi-ion process (Figure 2—figure supplement 2)" – I assume it refers to the number of ions present at the same time at the level of the constriction, however it is not clear from the figure. I suggest defining a "multi-ion" permeation process in the text and put coordinates of SF on Figure 2—figure supplement 2 to make it clear.

We thank the reviewer for this suggestion. The statement has been clarified.

9) line 203 – "specifically, the calculations indicate this mutant is about 2-fold less K+ selective than the wild-type channel" – it is not calculated in the manuscript at the moment.

The stated 2-fold factor directly stems from the calculated free-energy differences in Figure 4: p = log (DG/kBT). This has been clarified.

10) lines 177-181 – "We observe that the depletion in the both the first and second solvation shells, relative to bulk numbers, is significantly smaller for Na+than for K+. As will be further discussed below, this relative difference in dehydration energetics likely explains whyTMEM175 is only~30-fold more permeable to K+, even though the bulk water selectivity for Na+is over a trillion-fold." – it may not be clear how the number of water molecules lost during dehydration affects selectivity by itself (rather than the free energy difference between the processes of going from the bulk to the constriction). If it's the case, it should be explained more clearly.

We thank the reviewer and have clarified this point in the manuscript. In Figure 4, we show that the change in free energy for a K+ ion moving from bulk to the constriction is 1.2 kcal/mol less costly than moving a Na+ ion from the bulk to the constriction. In Figure 2, we show that both ions enter the constriction with four water molecules in their primary hydration shell, and another four in the second shell. Thus, while the absolute number of water molecules lost during dehydration is greater for K+ (3 for K+ and 2 for Na+ in the first shell; and about 20 for K+ and 16+ for Na+ in the second shell), it is actually more costly for Na+ to achieve the same dehydration state at the constriction than it is for K+, because in the bulk Na+ interacts with its hydration shells much more strongly than K+ (by about 18 kcal/mol).

11) Figure 5 Supplementary Figure 3 – the points for the raw currents before adding AP-6 are much more scattered than before 4-AP. What could be the reason for this behavior?

6. line 492 – "To measure currents reduced by AP-6, bath solutions were perfused" – the method for measuring the currents reduced by 4-AP are not described.

We thank the reviewer for noting that we did not describe how the 4-AP currents were measured. We have revised the methods section to include a description. The difference between the current levels prior to AP-6 are due to the variation in channel expression from cell to cell. All cells were prepared and recorded from in an identical fashion.

12) The starting structure for the simulations is stated to be 6WC9, the previously published structure of the open TMEM175, even though structures with higher resolution were obtained in this study. The reasoning behind using a lower-resolution structure should be provided.

The simulations were initiated prior to obtaining the newer high-resolution structure of TMEM175 in the open state. As the structures are essentially identical (RMSD = 0.3 Å), we continued to use 6WC9 for consistency.

13) Even though the positions of ions as a function of time are shown in Figure 2 Supplementary Figure 2, it may not be enough to estimate the convergence of metadynamics simulations. Can the authors provide an example of the time dependence of the CVs, or the deposited potential, or some other suitable measure for convergence as well?

It was not our intention for the data in Figure 2 Supplementary Figure 2 to serve as a convergence metric. As discussed above, to evaluate the convergence of the Metadynamics data we compared PMF profiles derived with either the first or the second half of all trajectory data – these profiles were shown in Figure 2. We have clarified how we evaluated convergence in the revised figure captions.

14) As another control for the metadynamics simulations, would it be possible to run MD of TMEM175 with the potassium ions and water molecules not removed from the initial structure? It could show if there are some energy minima not resolved by metadynamics, and if those ions/water molecules have any effect on the overall behavior of TMEM175. Also, would you expect any side effects from not modelling residues 164-253?

We thank the reviewer for their suggestions and we will consider them in future studies.

Reviewer #3 (Recommendations for the authors):

1) While the refined static structure of the open channel is discussed in detail, the conformational dynamics of the pore is not: was there any conformational isomerization of side chains? What is the extent of fluctuations in pore radius and relative helix arrangement? Do the apparent kink and tilt of the pore helices fluctuate?

We thank the reviewer for highlighting the close correspondence between the conformation of the pore in the structures and during the simulations. As illustrated in Figure 2 Supplementary Figure 1 the structural fluctuations of the pore are very limited. No noticeable changes in tilt or kink angles in any of the helices flanking the pore were detected, or large changes in the pore radius. The only significant change concurrent with ion permeation across the constriction is the rotation of the terminal CD atom in the sidechain of Ile 271 around to the CB-CG bond. We have noted the minimal changes in the text and two supplementary movies with different time-resolution have been added to revised manuscript to document these observations.

2) There seems to be a discrepancy between the fully-hydrated state of ions in the cryoEM densities (lines 95-97, Figure 1 Suppl. Figure 2) and their partial dehydration in the simulations (Figure 2). This is not a trivial point, since the major finding of the paper is that selective K+ permeation arises from ion (de)solvation effects.

We thank the reviewer for their comment and have revised the manuscript to note that the ordered ions in the cryo-EM structures are not fully hydrated, extending their description from the original manuscript, which stated that “The binding site on the cytoplasmic site of the constriction is coordinated by four ordered water molecules […], while the site on the luminal side is coordinated by four water molecules […].” The partial dehydration of the ions is also shown in Figure 1 S2. These data thus reflect a depletion of the ion hydration shells, as observed in simulation and we have remarked on this finding in the revised manuscript.

3) Systematic error analysis of the simulation results would strengthen the confidence in the numerical agreement noted above.

Please see our responses to Editor’s summary and Reviewer #2.

4) A broader discussion of the general significance of the findings, including the role of ion desolvation effects and the novelty that a hydrophobic locus controls both gating and selectivity in K+ channels and ion channels in general, would be welcome.

We thank the reviewer for their suggestion. However, discussion on the unique (to our knowledge) of a hydrophobic constriction participating in ion selectivity as well as serving as the channel gate was included in our previous work. As the gating of TMEM175 remains poorly understood at the molecular level and this study is focused on the selectivity of the channel, such a discussion would be largely redundant if repeated in the current study. Therefore, we would prefer to focus our discussion on the findings that we present in the current study.

5) The introduction could provide a bit more background. What is the biological function of this protein? Only dysfunction is mentioned.

We thank the reviewer for their suggestion and have extended our introduction of the physiological roles of TMEM175.

6) Please specify which atoms were used for the analysis shown in Figure 2 Suppl. Figure 1 (Calpha atoms, etc).

Corrected.

7) It would be useful to computational biophysicists if the authors could clarify the rationale for specific methodological choices. In particular, what are the considerations that presided over the choice of force field? Same question for the collective variable used in the metadynamics simulations.

We thank the reviewer for the opportunity to describe our rationale. As indicated in the original version of the manuscript we designed a collective variable that does not presuppose a mechanism of permeation. In biased-sampling simulations of ion channel permeation it is common to use the coordinate of one or more ions along the pore axis as collective variables. This choice might be reasonable when the number of ions involved in the mechanism is known. It was not known in our case, and hence the variable we designed is a metric of the proximity between any ion in the simulation system and a virtual center within the pore; by using two of these centers, one at either side of the constriction, we were able to foster sampling of the length of the pore and the central barrier, as well as a sufficient number of crossing events. This reasoning has been further clarified in the revised version of the manuscript. Regarding the choice of forcefield, we opted for what we believe to be the best compromise between accuracy and performance, based on our own experience as well as the specific nature of the questions addressed in this study.

8) Likewise, why did the authors use the particular functional form of Equation 6 to compute the ion coordination number rather than a simple cut-off distance? Please provide references for that choice if appropriate. Also, please note that brackets are missing in the summations in Eq. 6.

We chose this form for computational convenience. As explained in the Methods section, to rigorously derive the profiles shown in Figure 2B from trajectories calculated under an applied bias requires calculation of coordination numbers for every ion in every snapshot in every replica trajectory. We found that PLUMED, which features the function defined in Equation 6 as a collective variable, was the most efficient tool to carry out this analysis. Note that the large exponents imply that the function is essentially the same as a square step function cut off at a certain distance. This point has been clarified in the revised manuscript.

9) Error estimates should be provided systematically. In the current manuscript, they are sometimes vague or non-existent. In particular:

– Please provide data for estimating the convergence for metadynamics; and generally, error estimates as appropriate for all the numerical results.

– The penultimate sentence in the caption of Figure 2 is unclear: "gray profiles represent the same quantity shown in black/blue calculated using only the first or second half of the simulations data."

Please see our responses to Editor’s summary and Reviewer #2 – as well as the revised captions of Figures 2 and 4.

10) Please provide a reference for variations in the value of the diffusion constant (see lines 453-454).

The reference has been added.

11) I would recommend including a schematic figure of a thermodynamic cycle to the Discussion for clarity and to highlight the consistency between the two different pathways followed for the DeltaDeltaG calculation (see "Strengths" in public review above).

We appreciate the reviewer’s suggestion but believe this figure is not essential.

12) The statement in lines 236-237 seems to imply that the balance of channel-ion, ion-ion, and water-ion interactions controls selective ion permeation. The authors may consider including the effect of channel-water (and arguably also water-water and channel-channel) interactions for the sake of completeness and generality.

We have no data that indicates that differentials in channel-water interactions are a determining factor for the ion selectivity of this channel.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below. We anticipate acceptance after these changes have been implemented.

Please address the comments by reviewers 1 and 3, as detailed below.

In addition, regarding the justification of the force field choice, we wish to highlight that claiming that default CHARMM ion parameters from 1994 (Beglov and Roux, 1994) reproduce the relative free-energies of hydration (and thus dehydration) of the alkali cations 'with excellent accuracy' somewhat invalidates next 20 years of ion force field development. When addressing the issue in the article's text, as requested by reviewer 3, please ensure the comment is accurate and well-substantiated, and do not forget to mention possible shortcomings of the chosen force field.

We agree that there have been important advances in force field development. As requested, our revised text now includes a discussion on our chosen force field along with citations of recent studies comparing force fields for simulations of ion channels.

Reviewer #1 (Recommendations for the authors):

I thank the authors for their careful revision of the manuscript. Most of my previous comments have been addressed adequately, except for these two:

We thank the reviewer for their comments that have greatly improved our manuscript. We have addressed these two points in the revised version.

Regarding the position of the ion binding sites in the different figures created by different methods: Noting the residues in Figure 2 is certainly helpful (as has already been done in the original manuscript). However, indicating K1-K5 directly as I suggested in my initial review would spare the reader to flip back and forth the figures and comparing each single residue manually. There is still no indication of the binding sites in the time traces of the MD simulations, my request "All figures where this applies: please indicate the location of the binding sites from the structural data" has been ignored by the authors without any comment.

I still strongly suggest amending the figures accordingly as this would make them more accessible.

We thank the reviewer for the suggestion. As requested, we now have added blue arrows to Figure 2 to denote the locations of the ion-binding sites resolved in the cryo-EM density maps. For the time traces shown in Figure 2—figure supplement 2 and Figure 3, the y-axis (ion position along pore axis) is too compressed to allow readers to distinguish the individual binding sites. We hope the reviewer understands why we cannot include the arrows in these panels.

This comment of mine has maybe been overlooked by the authors: What is the meaning of the numbers in the bracket in Figure 3A? I presume the number of water molecules?

The reviewer is correct: the number in the brackets correspond to the number of water molecules in the primary shell of the ion in the snapshot. The figure legend has been revised accordingly.

Reviewer #2 (Recommendations for the authors):

All points have been adequately addressed. I recommend the paper for publication.

We thank the reviewer for their helpful comments during the review process that have greatly improved this manuscript.

Reviewer #3 (Recommendations for the authors):

The authors have addressed most of the comments appropriately.

We thank the reviewer for their helpful comments during the review process and have addressed their remaining comment in the revised manuscript.

In their response concerning the choice of forcefield, they write: "It could be convincingly argued that a mechanism that relies on dehydration energetics, rather than specific ion-protein interactions, will be reasonably described by this forcefield, as it is in fact parametrized, with excellent accuracy, to reproduce the relative free energies of hydration (and thus dehydration) of the alkali cations." That is the kind of justification that I was hoping for. Please make that convincing argument in the text, including references as needed.

We thank the reviewer for the suggestion and have included a short discussion of the forcefield used in our work in the revised manuscript.

https://doi.org/10.7554/eLife.75122.sa2

Article and author information

Author details

  1. SeCheol Oh

    Structural Biology Program, Memorial Sloan Kettering Cancer Center, New York, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Validation, Visualization, Writing – original draft, Writing – review and editing
    Contributed equally with
    Fabrizio Marinelli and Wenchang Zhou
    Competing interests
    The authors declare no competing interests
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1685-5922
  2. Fabrizio Marinelli

    Theoretical Molecular Biophysics Section, National Heart, Lung and Blood Institute, National Institutes of Health, Bethesda, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review and editing
    Contributed equally with
    SeCheol Oh and Wenchang Zhou
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0044-6718
  3. Wenchang Zhou

    Theoretical Molecular Biophysics Section, National Heart, Lung and Blood Institute, National Institutes of Health, Bethesda, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review and editing
    Contributed equally with
    SeCheol Oh and Fabrizio Marinelli
    Competing interests
    The authors declare no competing interests
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0397-1032
  4. Jooyeon Lee

    Department of Chemistry, Chungbuk National University, Cheongju-si, Republic of Korea
    Contribution
    Investigation, Methodology, Resources, Validation
    Competing interests
    The authors declare no competing interests
  5. Ho Jeong Choi

    Department of Chemistry, Chungbuk National University, Cheongju-si, Republic of Korea
    Contribution
    Investigation, Methodology, Resources, Validation
    Competing interests
    The authors declare no competing interests
  6. Min Kim

    Department of Chemistry, Chungbuk National University, Cheongju-si, Republic of Korea
    Contribution
    Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – review and editing
    Competing interests
    The authors declare no competing interests
  7. José D Faraldo-Gómez

    Theoretical Molecular Biophysics Section, National Heart, Lung and Blood Institute, National Institutes of Health, Bethesda, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review and editing
    For correspondence
    jose.faraldo@nih.gov
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7224-7676
  8. Richard K Hite

    Structural Biology Program, Memorial Sloan Kettering Cancer Center, New York, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review and editing
    For correspondence
    hiter@mskcc.org
    Competing interests
    The authors declare no competing interests
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0496-0669

Funding

National Cancer Institute (P30 CA008748)

  • Richard K Hite

National Institute of General Medical Sciences (R01-GM141553)

  • Richard K Hite

Josie Robertson Investigators Program

  • Richard K Hite

Searle Scholars Program

  • Richard K Hite

Ministry of Education (2019H1A2A1076014)

  • Jooyeon Lee

National Heart, Lung, and Blood Institute (Division of Intramural Research)

  • José D Faraldo-Gómez

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

Acknowledgements

We thank the MSKCC HPC group for assistance with data processing and the members of the labs for comments on the manuscript. We also thank Dr Rahul Banerjee for early simulation studies of closed state TMEM175 not included in this manuscript. This work was supported by NIH-NCI Cancer Center Support Grant P30 CA008748, NIGMS R01-GM141553 (RKH), the Josie Robertson Investigators Program (RKH), the Searle Scholars Program (RKH), the NRF Global PhD. Fellowship program funded by the Republic of Korea Ministry of Education 2019H1A2A1076014 (JL) and the Division of Intramural Research of NHLBI-NIH (JDFG). Computational resources were in part provided by the NIH supercomputing center (Biowulf).

Senior Editor

  1. Richard W Aldrich, The University of Texas at Austin, United States

Reviewing Editor

  1. Lucie Delemotte, KTH Royal Institute of Technology, Sweden

Publication history

  1. Received: October 29, 2021
  2. Preprint posted: November 6, 2021 (view preprint)
  3. Accepted: May 1, 2022
  4. Version of Record published: May 24, 2022 (version 1)

Copyright

This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.

Metrics

  • 938
    Page views
  • 319
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. SeCheol Oh
  2. Fabrizio Marinelli
  3. Wenchang Zhou
  4. Jooyeon Lee
  5. Ho Jeong Choi
  6. Min Kim
  7. José D Faraldo-Gómez
  8. Richard K Hite
(2022)
Differential ion dehydration energetics explains selectivity in the non-canonical lysosomal K+ channel TMEM175
eLife 11:e75122.
https://doi.org/10.7554/eLife.75122
  1. Further reading

Further reading

    1. Structural Biology and Molecular Biophysics
    Stefan L Schaefer, Gerhard Hummer
    Research Article Updated

    Gasdermin-D (GSDMD) is the ultimate effector of pyroptosis, a form of programmed cell death associated with pathogen invasion and inflammation. After proteolytic cleavage by caspases, the GSDMD N-terminal domain (GSDMDNT) assembles on the inner leaflet of the plasma membrane and induces the formation of membrane pores. We use atomistic molecular dynamics simulations to study GSDMDNT monomers, oligomers, and rings in an asymmetric plasma membrane mimetic. We identify distinct interaction motifs of GSDMDNT with phosphatidylinositol-4,5-bisphosphate (PI(4,5)P2) and phosphatidylserine (PS) headgroups and describe their conformational dependence. Oligomers are stabilized by shared lipid binding sites between neighboring monomers acting akin to double-sided tape. We show that already small GSDMDNT oligomers support stable, water-filled, and ion-conducting membrane pores bounded by curled beta-sheets. In large-scale simulations, we resolve the process of pore formation from GSDMDNT arcs and lipid efflux from partial rings. We find that high-order GSDMDNT oligomers can crack under the line tension of 86 pN created by an open membrane edge to form the slit pores or closed GSDMDNT rings seen in atomic force microscopy experiments. Our simulations provide a detailed view of key steps in GSDMDNT-induced plasma membrane pore formation, including sublytic pores that explain nonselective ion flux during early pyroptosis.

    1. Neuroscience
    2. Structural Biology and Molecular Biophysics
    Alec R Nickolls, Gabrielle S O'Brien ... Alexander T Chesler
    Research Article Updated

    Piezo1 is a stretch-gated ion channel required for mechanosensation in many organ systems. Recent findings point to a new role for Piezo1 in the gut, suggesting that it is a sensor of microbial single-stranded RNA (ssRNA) rather than mechanical force. If true, this would redefine the scope of Piezo biology. Here, we sought to replicate the central finding that fecal ssRNA is a natural agonist of Piezo1. While we observe that fecal extracts and ssRNA can stimulate calcium influx in certain cell lines, this response is independent of Piezo1. Additionally, sterilized dietary extracts devoid of gut biome RNA show similar cell line-specific stimulatory activity to fecal extracts. Together, our data highlight potential confounds inherent to gut-derived extracts, exclude Piezo1 as a receptor for ssRNA in the gut, and support a dedicated role for Piezo channels in mechanosensing.