1. Computational and Systems Biology
  2. Structural Biology and Molecular Biophysics
Download icon

The push-to-open mechanism of the tethered mechanosensitive ion channel NompC

  1. Yang Wang
  2. Yifeng Guo
  3. Guanluan Li
  4. Chunhong Liu
  5. Lei Wang
  6. Aihua Zhang
  7. Zhiqiang Yan  Is a corresponding author
  8. Chen Song  Is a corresponding author
  1. Center for Quantitative Biology, Academy for Advanced Interdisciplinary Studies, Peking University, China
  2. Peking-Tsinghua Center for Life Sciences, Academy for Advanced Interdisciplinary Studies, Peking University, China
  3. State Key Laboratory of Medical Neurobiology and MOE Frontiers Center for Brain Science, Institute of Brain Science, School of Life Sciences, Fudan University, China
  4. Institute of Molecular Physiology, Shenzhen Bay Laboratory, China
Research Article
  • Cited 0
  • Views 722
  • Annotations
Cite this article as: eLife 2021;10:e58388 doi: 10.7554/eLife.58388

Abstract

NompC is a mechanosensitive ion channel responsible for the sensation of touch and balance in Drosophila melanogaster. Based on a resolved cryo-EM structure, we performed all-atom molecular dynamics simulations and electrophysiological experiments to study the atomistic details of NompC gating. Our results showed that NompC could be opened by compression of the intracellular ankyrin repeat domain but not by a stretch, and a number of hydrogen bonds along the force convey pathway are important for the mechanosensitivity. Under intracellular compression, the bundled ankyrin repeat region acts like a spring with a spring constant of ~13 pN nm−1 by transferring forces at a rate of ~1.8 nm ps−1. The linker helix region acts as a bridge between the ankyrin repeats and the transient receptor potential (TRP) domain, which passes on the pushing force to the TRP domain to undergo a clockwise rotation, resulting in the opening of the channel. This could be the universal gating mechanism of similar tethered mechanosensitive TRP channels, which enable cells to feel compression and shrinkage.

Introduction

Many types of sensations initiate from the gating of transient receptor potential (TRP) ion channels, which regulate the intracellular cation concentration that triggers downstream signaling pathways (Montell et al., 2002; Mutai and Heller, 2003; Pedersen et al., 2005; Basbaum et al., 2009; Cheng et al., 2010a; Fowler and Montell, 2013). NompC is one of the earliest identified mechanosensitive ion channels belonging to the TRP family, which plays crucial roles in the sensation of light touch, hearing, balance, and locomotion of Drosophila melanogaster (Walker et al., 2000; Göpfert et al., 2006; Yan et al., 2013; Zhang et al., 2013; Zanini et al., 2018). NompC is structurally unique as it has the largest number of ankyrin repeats (ARs) among the known TRP channels (Montell, 2005), 29 in total. The AR region is associated with microtubules, and it has been proposed to act as a gating spring to regulate the channel gating according to the so-called ‘tethered gating model’ (Albert et al., 2007; Cheng et al., 2010b; Zhang et al., 2015). Although NompC orthologs have not been found in mammals (Venkatachalam and Montell, 2007; Knecht et al., 2015), it was shown to function in mechanosensation of Caenorhabditis elegans and Danio rerio as well (Kang et al., 2010; Sidi et al., 2003). It therefore serves as a useful model for studying the molecular mechanism of the tethered mechano-gating. The cryo-EM structure of NompC has been resolved (Jin et al., 2017), showing that four AR chains form an ~15-nm-long supercoiled helix and connect to the transmembrane (TM) pore domain via a linker helix (LH) region (Figure 1A).

Figure 1 with 13 supplements see all
The tethered NompC channel was opened by compression of the intracellular ankyrin repeat domain.

(A) The simulation systems. The NompC was divided into two subsystems, denoted by the cyan and red rectangular boxes, for the molecular dynamics (MD) simulations. (B−D) The transmembrane (TM) pore size evolution for the force-free (B), pulling/stretch (C), and pushing/compression (D) simulations, calculated from the MD trajectories FI0, SI0, and CI0 (Supplementary file 1a), respectively. (E) A schematic figure of the cell-attached patch-clamp electrophysiological experiment for NompC. (F) Representative traces of the electrophysiological measurements for the S2 blank cell and NompC-expressed cell, showing that there are significantly larger signals under positive pressure (PP) in the presence of NompC. (G) The mean and standard deviation (SD) of the mechano-gated currents in the S2 blank and NompC expressing cells under positive (PP) and negative pressure (NP) in the cell-attached patch-clamp experiments (S2_NP: n = 6; S2_PP: n = 6; NompC_NP: n = 5; NompC_PP: n = 7). All of the error bars denote ± SD.

The new structure confirmed that the AR helices probably act as a spring to conduct forces to the TM pore when the neuron cells deform. However, what kind of forces (or what type of cell deformation) can open the NompC channel, and how the force is transduced from ARs to the TM region to finally open the pore, are still elusive. In previous studies, it has been suggested that pulling the AR spring may open the channel (Zhang et al., 2015; Gaudet, 2008). In contrast, there are other models indicating that a pushing force may be required to open the channel (Howard and Bechstedt, 2004; Argudo et al., 2019). Therefore, the detailed gating mechanism of this unique tethered ion channel requires clarification. Additionally, the membrane surface tension-induced ion channel gating provides a mechanism by which cells can respond to volume expansion (Martinac et al., 1990; Sukharev, 2002; Nomura et al., 2012; Zhang et al., 2018; Martinac et al., 2018). However, there is no obvious mechano-gating mechanism that can respond to cell compression or volume shrinkage. In this study, we combined molecular dynamics (MD) simulations and electrophysiological experiments to study the detailed gating mechanism of NompC. We provide a plausible push-to-open mechanism for the tethered ion channels, which may be used by cells to sense and respond to compression and shrinkage.

Results

TM pore opens under an intracellular pushing force

To study the atomistic details of how mechanical stimuli can lead to the gating of the tethered NompC channel, we used a divide-and-conquer protocol. We performed all-atom MD simulations on the transmembrane and linker helix (TM + LH) domains, and the linker helix and ankyrin repeat (LH + AR) domains of the cryo-EM NompC structure, respectively (Figure 1A, Figure 1—figure supplement 1). We considered two forms of the most essential forces on the AR helices: pulling and pushing. For the TM + LH system, we applied forces that are normal to the membrane surface on the AR29, which directly connects to the LH region, and we monitored how the TM domain responds by calculating the radius of the TM pore. We observed that the channel remains closed (with a very narrow constriction site, radius <1.0 Å, around the residue I1554) throughout the simulations without any external forces (Figure 1B and Figure 1—figure supplement 2A, Video 1), indicating that the closed-state cryo-EM structure was stable in our ‘force-free’ MD simulations. When the direction of the pulling force was away from the membrane surface, the TM channel also remained closed in our MD simulations (Figure 1C and Figure 1—figure supplement 2B, Video 1). In fact, the narrow region with a radius of less than 1 Å expanded in the latter part of the trajectory compared to the ‘force-free’ simulations, indicating that the channel was actually more closed than the free NompC in our ‘pulling’ simulations. In contrast, when applying a proper pushing force (toward the membrane) on the AR29, we observed that the narrowest constriction site of the channel was significantly dilated in the latter part of the ‘pushing’ simulation (Figure 1D and Figure 1—figure supplement 2C, Video 1). The above-mentioned analysis was carried out based on the all-atom structure of NompC in the MD trajectories. We also conducted pore radius analysis based only on the backbone structures, and we observed the same trend as mentioned above (Figure 1—figure supplement 3 and Figure 1—figure supplement 4), which confirmed that the pore dilation was due to the global conformational change of the backbone rather than merely a side-chain movement.

Video 1
The transmembrane (TM) pore size evolution during the 250 ns simulation trajectories (FI0, SI0, and CI0) as shown in Figure 1B–D.

This video shows the TM pore size evolution during the 250 ns ‘free’, ‘pulling’, and ‘pushing’ simulations.

The number of water molecules at the gate region should be increased when the pore is dilated, which is often viewed as an additional indicator for the channel opening. Our analysis showed that the number of water molecules was indeed evidently increased in the ‘pushing’ simulations, as compared to the ‘free’ and ‘pulling’ simulations (Figure 1—figure supplement 5). In the absence of TM potential, we observed two spontaneous ion permeation events only when the pore was dilated under a pushing force (Figure 1—figure supplement 6). When applying a membrane potential of ±300 mV, we observed continuous ion permeation through the dilated pore caused by the pushing force in our MD simulations (Figure 1—figure supplement 7 and 8, Video 2). Therefore, our simulation results indicated that the NompC channel may be opened by a pushing force from the intracellular side but not by a pulling force.

Video 2
Ion permeation through the partially opened NompC channel under transmembrane potential.

To validate these findings, we did cell-attached patch-clamp experiments (Figure 1E), in which positive or negative pressure with a 20 mm Hg increment was applied. Since the AR region is associated with microtubules (Liang et al., 2013), it is conceivable that positive pressure will result in a slight compression of the AR region and thus a pushing force on the TM domain, whereas a negative pressure will generate a slight stretch of the AR region and a pulling force on the TM domain in the cell-attached patch-clamp experiments. As shown in Figure 1F and G, the reference Drosophila S2 cells without NompC expressed showed no response to the positive and negative pressure stimuli, while we detected a clear electrical signal through the NompC-expressed S2 cells under positive pressure, whereas the signal under negative pressure was nearly negligible. Similarly, we can detect a clear signal through the NompC-expressed S2 cells from the outside-out patch clamp under negative pressure which corresponds to the cell-attached patch clamp under positive pressure. On the other hand, the signal under inside-out patch clamp with negative pressure was nearly negligible (Figure 1—figure supplement 9). The electrical signals from cell-attached patch clamp under positive pressure and outside-out patch clamp under negative pressure were nearly completely abolished after adding GdCl3 (a blocker for NompC) to the bath (Figure 1—figure supplement 10), confirming that the detected signal was indeed due to the ion permeation through NompC. Our results are consistent with a previous study showing that NompC can be activated by mechanical forces, and the AR regions are crucial for the mechano-gating of NompC (Zhang et al., 2015). Therefore, our experimental results indicated that it is the compression of the AR region and the resulted pushing force that opens the channel, which is consistent with MD simulations.

Conformational changes of the TM domain associated with gating

We investigated how a pushing force from the AR region can open the TM pore by analyzing the TM + LH simulations. The free, pulling, and pushing trajectories were concatenated, and principal component analysis (PCA) was performed to visualize the collective motion of the NompC pore domain. As shown in Figure 2A, the second PCA eigenvector can distinguish the conformations of the free, pushing, and pulling simulations very well, with the larger values corresponding to the more dilated states. We extracted the two extreme conformations along the second PCA eigenvector, and we overlaid them to visualize the most significant conformational changes of the TM domain under the three mechanical stimuli (Figure 2B). We observed an evident clockwise rotation (looking from the intracellular side, Figure 2B, Figure 2—figure supplements 12, and Video 3) and an obvious upward tilt (looking from lateral side of membrane, Figure 2—figure supplements 12, and Video 4) of the TRP domain when a pushing force was applied to the AR29. This clockwise rotation and upward tilt of the TRP domain may be associated with the opening of the TRP channels (Liao et al., 2013; Cao et al., 2013; Gao et al., 2016; Zheng and Qin, 2015). The overlaid structures in Figure 2B show that the clockwise rotation of the TRP domain induced the S6 helices (which are directly linked to the TRP domain) to rotate clockwise as well, albeit to a lesser extent. The gating constriction site is located at I1554 of the S6 helix (Jin et al., 2017), and in our simulations, they were pulled away from the channel axis when the S6 helices rotate clockwise together with the TRP domain, leading to the dilation of the pore (Figure 2B and C). Thus, consistent with previous structural studies of TRPV1 (Figure 2—figure supplement 3; Liao et al., 2013; Cao et al., 2013; Gao et al., 2016), our simulations showed that the clockwise rotation of the TRP domain (as well as the S6 helices) may lead to the opening of the NompC pore. It is the pushing force (compression of the intracellular domain) that leads to this collective gating motion.

Figure 2 with 6 supplements see all
Conformational changes of the transient receptor potential (TRP) and transmembrane (TM) domains during gating.

(A) Principal component analysis (PCA) of the molecular dynamics (MD) simulation trajectories (FI0, SI0, and CI0 in Supplementary file 1a). The projections on the second eigenvector can distinguish the conformations under pulling (red) or pushing (blue). (B) The overlaid extreme structures along the second eigenvector of the PCA. The most closed conformation (silver) and open conformation (cyan) showed the global changes of the TRP domain during gating: a clockwise rotation. (C) The orientation and position of the gate residue, I1554, in the most closed (silver) and open (cyan) conformations in the simulations. (D, E) The residues forming four stable hydrogen bonds between the TRP and LH domains. (F) The mean and standard deviation (SD) of the mechano-gated current of the wild-type NompC, as well as the mutants W1572A, S1421A, Q1253A, S1577A, K1244A, E1571A, D1236A, and R1581A, under negative pressure in the outside-out patch-clamp experiments (wild type: n = 13; W1572A: n = 7; S1421A: n = 6; Q1253A: n = 6; S1577A: n = 5; K1244A: n = 6; E1571A: n = 6; D1236A: n = 9; R1581A: n = 6). All of the error bars denote ± SD. Hydrogen bonds are indicated by dashed lines (D, E).

Video 3
The rotation of the transient receptor potential (TRP) domain in the steered molecular dynamics (SMD) simulations (SI0 and CI0 in Supplementary file 1a).
Video 4
The tilt of the transient receptor potential (TRP) domain in the steered molecular dynamics (SMD) simulations (SI0 and CI0 in Supplement 1a).

Key residues around the TRP domain for gating

We analyzed the hydrogen bonding network around the TRP domain, and we attempted to locate the key residues ensuring the clockwise rotation of the TRP domain in response to the pushing force from AR. We identified four stable hydrogen bonds throughout the MD simulations (Figure 2D and E and Supplementary file 1d). Three of the four hydrogen bonds can also be directly identified in the cryo-EM structure, except for the one between Q1253 and S1577, which was more stable only in the presence of a pushing force in the MD simulations (Supplementary file 1d). These stable hydrogen bonds indicate a conservative interaction network as well as a stable local configuration during the gating process. We then did mutations on the residues forming these hydrogen bonds and performed electrophysiological experiments to determine if any of them play crucial roles in the gating of NompC. Figure 2F shows that mutations of most of the eight residues led to some loss-of-function. In the meantime, the mutants showed normal membrane targeting (Figure 2—figure supplement 5). This indicated that most of the mutations changed the mechanosensitivity of NompC. In particular, the W1572A mutation completely abolished the gating behavior of the channel, consistent with the work of Jin et al., 2017. Interestingly, we found that W1572 may be the rotation pivot of the TRP domain in our MD trajectory, which forms a stable hydrogen bond with the backbone of S1421 on the S4-S5 linker. This highlighted the importance of the interactions between the TRP domain and the S4-S5 linker in the gating process. Notably, this hydrogen bond does not involve the side chain of S1421, so the mutations at S1421 would not be expected to alter the above hydrogen bond and would not lead to significant loss-of-function of NompC. This was confirmed for S1421A, as shown by the dashed line in Figure 2F. In addition, the mutations S1577A and R1581A on the TRP domain, and K1244A and Q1253A on the LH domain, all resulted in significant loss-of-function, indicating the essential roles of these residues in conveying the forces from the AR region to the TRP domain. Thus, five out of seven residues, whose side chains form hydrogen bonds between the TRP and LH domains bonds as identified in our MD simulations, were crucial for the proper gating behavior of the NompC channel. The other two residues, D1236 and E1571, which are also involved in the hydrogen bonding between the TRP and LH domains, were found to be replaceable by adjacent residues in stabilizing the local conformation (Figure 2—figure supplement 6). These data show that W1572 acts as a rotation pivot by interacting with the S4-S5 linker, while the TRP domain senses a pushing force from the LHs upon AR compression. The force was stabilized by at least four hydrogen bonds, resulting in a clockwise rotation of the TRP domain around W1572. This is consistent with previous findings that the TRP domain (Jin et al., 2017; Gao et al., 2016), as well as the S4-S5 linker (Jin et al., 2017; Cox et al., 2019), play crucial roles in the gating of TRP or mechanosensitive channels. These results also confirm that a force/conformational change has to be transferred from the AR region to the TRP domain through the LHs when the NompC channel is opening in response to a mechanical stimulus.

Mechanical properties of the AR region

To study how a pushing/pulling force is transferred to the LHs from the ARs, we performed multiple MD simulations on the truncated LH and AR domains (Figure 3A). We applied position restraints on the LHs (orange) and ran simulations with, or without, external forces applied to the terminal AR1 (Figure 3A). Several mechanical properties were obtained from these simulations. First, we analyzed the reaction forces of the position restraints on the LHs, which were identical, in magnitude and direction, to the forces acting on the LHs by ARs. The analysis indicated that when pushing the four AR1 toward the membrane with a total force of 20 pN (5 pN of force on each chain), the four AR29 apply a total torque of ~13 pN·nm on the LH domain pointing to the extracellular side, in addition to a dominant pushing force. This torque would help to rotate the TRP domain clockwise and drive the channel to open (Figure 3B, Figure 3—figure supplement 1). This is consistent with a continuum mechanics study by Argudo et al., 2019. Second, we calculated the force constant of the AR spring by k=Fz, where F is the force we applied on AR1 and (Sotomayor et al., 2005) z is the distortion of the AR region (Figure 3C, Videos 5 and 6). The spring constant of each AR helix was estimated to be 3.3 ± 0.9 pN nm−1 in the supercoiled helix bundle formed by the four AR chains. Thus, the whole AR helix bundle has a force constant of ~13 pN nm−1. For comparison, previous atomic force microscopy measurements determined a force constant of 1.87 ± 0.31 pN nm−1 for a single AR chain (Lee et al., 2006), and previous steered MD (SMD) simulations obtained a value of ~4.0 pN nm−1 (Sotomayor et al., 2005). However, our calculation was performed for the supercoiled AR helix complex, while the previous study evaluated a single 24-AR spring. For comparison, we performed SMD on a single 29-AR spring and estimated the spring constant to be 2.5 ± 0.4 pN nm−1 (Figure 3—figure supplement 2). The close agreement of the values from the single AR and one AR in the supercoiled complex indicated that the four AR helices are not tightly coupled. We also performed simulations with weaker mechanical forces, ranging from 1 to 4 pN, and the resulting average force constants were all ~3 pN nm−1 for each AR chain in the helix bundle (Figure 3—figure supplement 3). Therefore, the AR supercoiled spring remained in its elastic limits in the study, although in the SMD trajectories with larger forces, the ARs showed a degree of distortion (Video 5). Third, we analyzed how fast the forces can be transferred from the AR1 to the LH. The deviation of the directions of the forces exerted on the LH regions when the AR region was stretched or compressed occurred after about 7-8 ps (Figure 3D, Figure 3—figure supplement 4). Considering that the length of the relaxed AR region is about 15 nm, we estimated that the force was transferred through the AR region at a speed of 1.8 ± 0.2 nm ps−1. A recent study showed that forces are propagated via membranes at a speed of 1.4 ± 0.5 nm ps−1 (Aponte-Santamaría et al., 2017). Therefore, it appears that the force transfer speed in the tethered NompC channel is comparable to, or slightly faster, than that in the membranes.

Video 5
The conformational changes of the AR domain in the “free”, “pulling”, and “pushing” MD simulations as shown in Figure 3C.
Video 6
The conformational changes of the AR domain in the “free”, “pulling”, and “pushing” MD simulations, in which a 2-pN force was applied to each AR chain.

Key residues at the interface between the AR and LH regions

We also found two stable hydrogen bonds between the ARs and LHs in the MD trajectories, between W1115 and D1142, and R1127 and E1163, respectively (Figure 3E and Supplementary file 1d). Mutations of D1142A, R1127A, and E1163A, which break the hydrogen bonds but show normal membrane targeting, led to a significant loss-of-function in the electrophysiology experiment (Figure 3F, Figure 3—figure supplement 5). However, W1115A does little to alter the mechanosensing behavior (Figure 3F, Figure 3—figure supplement 5), probably because its hydrogen bonding and stabilizing role can be replaced by the adjacent Y1109, which can form a stable hydrogen bond with D1142 in the presence of the W1115A mutation as observed in our MD simulations (Figure 3—figure supplement 6). These results indicate that the interface between the AR and LH regions is crucial for the force transduction, further supporting the tethered spring model for NompC.

Figure 3 with 7 supplements see all
Mechanical properties of the ankyrin repeat (AR) region.

(A) The simulation system in which the linker helix (LH) domain (orange) was restrained and a pushing or pulling force was applied to the first AR (gray). (B) Projection of the reaction forces of the restraints on the LH domain (same as the forces exerted on the transient receptor potential [TRP] domain by the LH domain) on the plane parallel to the membrane surface, showing that a torque is generated that will drive the LH and TRP domain to rotate clockwise (looking from the intracellular side). The calculation was based on the molecular dynamics (MD) trajectories CII1 and symmetrized from the original data as shown in Figure 3—figure supplement 1. (C) The AR region was compressed/stretched by a pushing/pulling force of 5 pN and reached its equilibrium length within 40 ns simulations (from MD trajectories SII1 and CII1 with respect to FII1). (D) The evolution of the net average reaction forces of the restraints on the LH domain when pushing (blue) or pulling forces (red) were applied to AR1, calculated from the MD trajectories FII2-6, CII2-6, and SII2-6 (Supplementary file 1b). A clear deviation occurred at around 7-8 ps during the simulation time, indicating that the forces applied to AR1 have reached LH at the time. (E) The residues forming two stable hydrogen bonds between the LH domain and AR29. (F) The mean and standard deviation (SD) of the mechano-gated current of the wild-type NompC and the mutants W1115A, D1142A, R1127A, and E1163A, under negative pressure in the outside-out patch-clamp experiments (wild type: n = 13; W1115A: n = 5; D1142A: n = 4; R1127A: n = 5; E1163A: n = 6). All of the error bars denote ± SD. Hydrogen bonds are indicated by dashed lines (E).

The hydrogen bonds between W1115-D1142 and R1127-E1163 were not observed in the cryo-EM structure. Our analysis showed that the distances between the side chains of the above two pairs of hydrogen bond-forming residues obtained from the MD trajectories were closer than those in the cryo-EM structure, while the distances between the α-carbon atoms were nearly identical (Figure 3—figure supplement 7). This indicated that the hydrogen bonding difference in the MD simulations and the cryo-EM structure was due to the side-chain adjustments in the simulations. As the resolution of the cryo-EM structure was insufficient to identify the exact locations of the side chains and the MD simulations accounted for all of the atomistic interaction details when dynamically evolving the systems, we believe that the MD simulations may have presented a better equilibrated local conformation that allowed identification of the two additional hydrogen bonds.

Discussion

A combination study of MD simulations and electrophysiological experiments produced a clear ‘push-to-open’ gating model of the NompC channel. As illustrated in Figure 4, compression or shrinkage of cells can compress the AR spring, which generates a pushing force and also a torque on the TRP domain with a component pointing to the extracellular side and perpendicular to the membrane surface. The torque is generated by the specific supercoiled structure of the AR region, as demonstrated by the mechanics study of Argudo et al., 2019. This torque helps drive the TRP domain to rotate clockwise. Our simulations indicate that the pushing force alone may be sufficient to generate a clockwise motion of the TRP domain, which in turn pulls the S6 helices to open the NompC gate. Critical residues between the TRP domain and LH, including R1581, W1572, Q1253, S1577, and K1244, as well as the presence of the S4-S5 linker above the TRP domain, ensure that the TRP domain will rotate clockwise around the pivot W1572 when a pushing force is applied to the LH region. This is consistent with a study showing that a TRPV1 mutant, which has only two ARs, can be mechanically opened by a pushing force (Prager-Khoutorsky et al., 2014). We believe that TRP channels similar to NompC, with a certain number of ARs, can be tethered to microtubules and use the push (-AR)-to-open mechanism to sense and respond to cell shrinkage or compression. This sensing mechanism can be complementary to the well-studied stretch(-membrane)-to-open mechanism that responds to cell expansion.

A gating model of NompC.

(A) The compression of the ankyrin repeat (AR) region will generate a pushing force and a torque on the linker helix (LH) domain, pointing to the extracellular side. (B) The LH domain further pushes the transient receptor potential (TRP) domain, leading to a tilt (side view), and (C) a clockwise rotation of the TRP domain (looking from the intracellular side). The motion of the TRP domain pulls the S6 helices to slightly tilt and rotate, which dilates the constriction site of the pore.

In this study, we focused on the intrinsic mechanical properties of NompC. To be more comprehensive, the possible effect of ‘force-from-lipids’ should be considered. Previous studies showed that lipid molecules may be involved in NompC gating and a stable lipid molecule has been found near H1423 in the cryo-EM map (Jin et al., 2017). Our analysis on MD simulation trajectories also showed that H1423 can stably interact with a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) lipid molecule. This lipid molecule was not in the interacting pocket in the initial simulation system but rapidly diffused to a location similar to that in the cryo-EM structure and acted as a bridge to link H1423 and S1268 (Figure 5A–D). This suggested that lipid molecules can help stabilize the local structure of NompC around H1423 and/or convey forces from the membrane.

The interaction between H1423 and lipids and the effect of adding 1-oleoyl-2-acetyl-sn-glycerol (OAG) on the NompC opening.

(A) The bottom view and (B) the side view of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) molecules moving around H1423. The transient receptor potential (TRP) domain is shown in blue, the S4S5 linker is shown in cyan, and the POPC density around H1423 is shown with silver transparent surfaces. The analysis was performed on the molecular dynamics (MD) trajectory FI0. (C) The initial (salmon) and final (violet) locations of a POPC molecule in the MD simulation trajectory FI0. (D) A lipid molecule was observed in the cryo-EM structure of NompC (PDB ID: 5vkq). The pocket between H1423 and S1268 can stably accommodate a lipid molecule in both the MD simulations and the cryo-EM structure. (E) The representative traces of the spontaneous NompC current before/after adding OAG to the intracellular monolayer of the membrane, and the corresponding average open probabilities (N = 5, paired Student’s t-test, p=0.0208). (F) The representative traces of the spontaneous NompC current before/after adding OAG to the extracellular monolayer of the membrane, and the corresponding average open probabilities (N = 5, paired Student’s t-test, p=0.0047).

A pushing/pulling force would lead to membrane curvature and a resulting asymmetrical TM stress profile that might activate mechanosensitive channels as demonstrated by Cox et al., 2019; Bavi et al., 2016 Also, previous studies showed that the addition of OAG tends to activate TRPC6 by introducing an asymmetrical TM stress profile (Spassova et al., 2006; Nikolaev et al., 2019). Therefore, we conducted additional patch-clamping recordings to examine whether NompC is sensitive to the asymmetrical TM stress profile created by OAG. The experiments showed that the addition of OAG to the intracellular monolayer tends to activate NompC, while adding OAG to the extracellular monolayer had the opposite effect (Figure 5E–F). This finding was consistent with previous studies on TRPC6 (Nikolaev et al., 2019). The unidirectional OAG-induced activation suggests a synergetic gating mechanism. Pushing the AR spring would lead to the intrinsic gating of NompC and could simultaneously generate a membrane curvature and an asymmetrical stress profile that may also facilitate the channel gating. The synergetic gating mechanism between the force-from-tether and force-from-lipids warrants future research.

Due to limitations of the simulation timescale, we were unable to observe the full gating process of the NompC. Therefore, the ion conductance in the MD simulations was smaller than the experimental results for the fully open state. Nonetheless, a clear and detailed opening trend of the channel in the presence of pushing forces was seen in the MD simulations and this was supported by electrophysiological experiments. The combination of the two methods provides a plausible ‘push-to-open’ gating model for tethered mechanosensitive ion channels.

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional information
Cell line (D. mel)Schneider 2 (S2) cellsCCTCC (China Center for Type Culture Collection)Serial# GDC0138Cell species report and Mycoplasma contamination test reports provided
AntibodyRabbit anti-αNOMPC-EC (polyclonoal)Ref. (Zhang et al., 2015)Immunostaining dilution (1: 500), primary antibody
AntibodyAlexa Fluor 594 AffiniPure Donkey Anti Rabbit IgG(H + L)YeasonCat# 34212ES60Immunostaining dilution (1: 100), secondary antibody
Recombinant DNA reagentpActin-Gal4 (plasmid)Ref. (Yan et al., 2013)Plasmid for driving Gal4 expression under actin promoter in S2 cells
Recombinant DNA reagentpUAST-NOMPC-EGFP (plasmid)Ref. (Yan et al., 2013)Plamid for Gal4-driven NompC expression in S2 cells
Recombinant DNA reagentpUAST-NOMPC-EGFP (del-miniwhite,dm) (plasmid)This paperPlamid for Gal4-driven WT NompC expression in S2 cells, no miniwhite sequence
Recombinant DNA reagentpUAST-NOMPC(D1236A)-EGFP(dm) (plasmid)This paperContains Drosophila NOMPC CDS with alanine substitution on D1236
Recombinant DNA reagentpUAST-NOMPC(R1581A)-EGFP(dm) (plasmid)This paperContains Drosophila NOMPC CDS with alanine substitution on R1581
Recombinant DNA reagentpUAST-NOMPC(K1244A)-EGFP(dm) (plasmid)This paperContains Drosophila NOMPC CDS with alanine substitution on K1244
Recombinant DNA reagentpUAST-NOMPC(E1571A)-EGFP(dm) (plasmid)This paperContains Drosophila NOMPC CDS with alanine substitution on E1571
Recombinant DNA reagentpUAST-NOMPC(Q1253A)-EGFP(dm) (plasmid)This paperContains Drosophila NOMPC CDS with alanine substitution on Q1253
Recombinant DNA reagentpUAST-NOMPC(S1577A)-EGFP(dm) (plasmid)This paperContains Drosophila NOMPC CDS with alanine substitution on S1577
Recombinant DNA reagentpUAST-NOMPC(S1421A)-EGFP(dm) (plasmid)This paperContains Drosophila NOMPC CDS with alanine substitution on S1421
Recombinant DNA reagentpUAST-NOMPC(W1572A)-EGFP(dm) (plasmid)This paperContains Drosophila NOMPC CDS with alanine substitution on W1572
Recombinant DNA reagentpUAST-NOMPC(W1115A)-EGFP(dm) (plasmid)This paperContains Drosophila NOMPC CDS with alanine substitution on W1115
Recombinant DNA reagentpUAST-NOMPC(D1142A)-EGFP(dm) (plasmid)This paperContains Drosophila NOMPC CDS with alanine substitution on D1142
Recombinant DNA reagentpUAST-NOMPC(R1127A)-EGFP(dm) (plasmid)This paperContains Drosophila NOMPC CDS with alanine substitution on R1127
Recombinant DNA reagentpUAST-NOMPC(E1163A)-EGFP(dm) (plasmid)This paperContains Drosophila NOMPC CDS with alanine substitution on E1163
Chemical compound, drug1-Oleoyl-2-acetyl-sn-glycerol (OAG)Sigma-AldrichCat# O6754DAG analogue
Chemical compound, drugGdCl3Sigma-AldrichCat# 439770NOMPC blocker
Chemical compound, drugConcanavalin A (Con A)Sigma-AldrichCat# C5275Cell adhesion
Chemical compound, drugClonExpress II One-step Cloning KitVazymeSerial# C112Site-directed mutagenesis
Chemical compound, drugTransIT-Insect Transfection ReagentMirusCat# MIR 6100S2 cell transfection

The simulation systems

Request a detailed protocol

We adopted a ‘divide-and-conquer’ strategy for the MD simulations and simulated two systems separately. System I included the TM region, the LH region, and the AR 29 of NompC (refer to Supplementary file 1e and Figure 1—figure supplement 1 for the details of the residue range). The PPM server was used to reorient the NompC structure to ensure that the TM domain of NompC was well located in a lipid bilayer (Lomize et al., 2012). The protein was embedded in a POPC bilayer and then solvated in a water box of 150 × 150 × 150 Å3. CHARMM-GUI was used to generate the configuration and topology of the simulation system, as well as the parameter files with the CHARMM36m force field (Wu et al., 2014; Lee et al., 2016; Jo et al., 2009). There were 492 POPC molecules, 72,000 water molecules, and sodium and chloride ions corresponding to a concentration of 150 mM in the setup, resulting in a system of 314,352 atoms in total.

System II included the LH domain and the AR domain of NompC (refer to Supplementary file 1e and Figure 1—figure supplement 1 for residue range details). The protein was solvated in a water box of 200 × 200 × 200 Å3. CHARMM-GUI was used to generate the configuration, topology, and parameter files with CHARMM36m force fields. In addition to the protein, 354,567 water molecules were added and sodium chloride ions were added to maintain an ion concentration of 150 mM. The simulation system II contained 1,134,213 atoms in total.

MD simulations

Request a detailed protocol

All of the MD simulations were performed with GROMACS 5.1.2 (Hess et al., 2008). The REDUCE program in AMBER was used to add hydrogens to the original PDB files and determine the protonation state of the histidine residues (Word et al., 1999; Case et al., 2005). For system I, energy minimization was achieved using the steepest descent algorithm, followed by a two-stage equilibration, a 0.4 ns NVT (constant particle number, volume, and temperature) equilibration simulation with harmonic restraint applied to the protein molecules (a force constant of 4000 kJ mol−1 nm−2 on the backbone and 2000 kJ mol−1 nm−2 on the side chains), and a 20 ns NPT equilibration simulation with gradually decreased restraint (from 2000 to 100 kJ mol−1 nm−2 on the backbone and from 1000 to 50 kJ mol−1 nm−2 on the side chains). During the equilibration processes, harmonic restraints were applied to heavy atoms of the protein, and planar restraints were used to keep the positions of lipid head groups along the normal direction of the membranes. The simulation temperature of the system was set to 300 K. After all of the equilibration steps were completed, the restraints were removed and the production simulations were performed in the NPT ensemble. The time step was 2 fs. The cubic periodic boundary condition was used during the simulations and the van der Waals interaction was switched off from 10 to 12 Å. The long-range electrostatic interactions were calculated with the particle mesh Ewald (PME) method (Darden et al., 1993).

For system II, the steepest descent algorithm was used to achieve initial energy minimizations, and then it was followed by a two-stage equilibration, a 0.2 ns NVT equilibration simulation with harmonic restraint forces applied to the protein (force constants of 400 kJ mol−1 nm−2 on the backbone and 40 kJ mol−1 nm−2 on the side chains), and a 10 ns NPT equilibration simulation with restraints on the protein backbone (force constant of 400 kJ mol−1 nm−2) and side chains (force constant of 40 kJ mol−1 nm−2). The temperature was set to 300 K. In the production simulations of system II, 1000 kJ mol −1 nm−2 harmonic restraints were applied to the heavy atoms of the LH domain while the restraints on the AR region were removed. The time step was set to 2 fs, and the trajectories were saved every 10 ps. The long-range electrostatic interactions were calculated using the PME method (Darden et al., 1993).

SMD simulations

Request a detailed protocol

For system I, after equilibration, SMD simulations were utilized to pull AR29 to simulate the mechanical stimuli from the AR region (spring) (Izrailev, 1999; Isralewitz et al., 2001). The TM regions of the four chains of NompC were treated as the reference group, and the AR29 of the four chains were treated as the pulling group. In addition to the force-free simulations (no pulling forces on the AR region), we considered the two most essential mechanical stimuli: the pulling and pushing forces on the AR29 along the direction normal to the membrane surface (the z-axis in our simulations), where pulling meaning that the force is pointing to the intracellular side (stretch of the AR spring) and pushing meaning that the force is pointing to the extracellular side (compression of the AR spring) along the z-axis. We tested a series of harmonic force constants as well as pulling speeds, and a harmonic force constant of 100 kJ mol−1 nm−2 and a pulling speed of 0.1 Å ns−1 were found to be reasonable for the gating simulations, where the opening of the pore was observed and the global protein structure was not disrupted in 200 ns. As further validations, a series of weaker SMD simulations with smaller force constants of 50 kJ mol−1 nm−2 and a slower pulling speed of 0.05 Å ns−1 were performed for each condition (Supplementary file 1a). During the MD/SMD simulations, the distances between the TM region of NompC and AR29, and the driving forces that act on the four chains of AR29 were recorded. The frames from MD/SMD trajectories were saved every 1 ns. All of the MD/SMD trajectories of system I are listed in Supplementary file 1a.

For system II, starting from the equilibrated structure, the LH domain was position-restrained and SMD simulations were performed to pull AR1, simulating the compression and stretch of the AR region (Izrailev, 1999; Isralewitz et al., 2001). The LH domains of the four chains of NompC were treated as the reference group, and the AR1 of the four chains were treated as the pulling groups. Constant pulling forces of 5 pN were applied on AR1 of each chain along the z-axis. Again, we considered the two most essential mechanical stimuli here: the pulling and pushing forces on the AR1, where pulling means the force is pointing to the intracellular side (a stretch of the AR spring), and pushing means the force is pointing to the extracellular side (compression of the AR spring) along the z-axis. In addition, we performed many 100 ns MD/SMD trajectories with a series of pushing/pulling forces ranging from 0 to 5 pN (Supplementary file 1b). During the MD/SMD simulations, the distances between the LH domain and the AR1 of each chain were recorded. To take into account the position restraints applied to the AR1 of the four chains by microtubules, an additional flat bottom potential of 100 kJ mol−1 nm−2 with a 3 nm radius was added on the four chains of AR1 on the x-y plane, to restrain each AR1 to move within a cylinder parallel to the z-axis. To estimate the mechanical property of a single chain of AR, the same protocol was applied to the single chain A of system II.

To estimate the speed of the forces that are conveyed along the AR spring, five 40 ps trajectories (FII2-6, CII2-6, and SII2-6 in Supplementary file 1b) were generated for each condition (free, compress/push, and stretch/pull) with a constant force of 5 pN on the AR1 of each chain. These trajectories were saved every 10 fs, a frequency high enough for the force transfer analysis. All of the MD/SMD trajectories of system II are listed in Supplementary file 1b.

The ion permeation simulations

Request a detailed protocol

After 200 ns pushing SMD simulations for system I, the TM pore of NompC was partially opened (Figure 1D). This partially open state was then stimulated by an umbrella pushing potential with a force constant of 100 kJ mol−1 nm−2 and an initial force of 50 pN on the AR29 of each chain for more than 500 ns, and a transient structure with the pore radius of the lower constriction more than 2.0 Å was exacted at 545 ns for the ion permeation simulations. The gate region, which includes S5, S6, the selectivity filter, and the TRP region, was position-restrained by the harmonic potential with a force constant of 1000 kJ mol−1 nm−2, while a TM potential of 300 mV was applied by setting a uniform electric field along the z-direction. Three independent 200 ns MD trajectories were generated with 150 mM KCl or NaCl in the systems, respectively. All of the ion permeation simulations of the partially open NompC are listed in Supplementary file 1c. The ion permeation events were analyzed in the simulations, from which we calculated the current by I=Δq/Δt, which was then used to calculate the conductance by C=I/U, where U was the TM potential 300 mV. The estimated conductance of the channel was about 7-15 pS.

Mutation MD simulations

Request a detailed protocol

To determine why the single mutations of critical residues D1236A, E1571A, and W1115A did not significantly impact the gating of NompC, three 500 ns all-atom MD simulations were performed with the mutations D1236A, E1571A, and W1115A incorporated into the system, respectively (Supplementary file 1a).

Principal component analysis

Request a detailed protocol

The distances between the centers of mass of the TM domain and AR29 (TM-AR29 distance) were monitored in all of the MD simulations for system I (Figure 1—figure supplements 1112). The data after a sharp change of the distance were discarded for further analysis, where the global conformation was distorted due to the strong pulling forces (gray areas in Figure 1—figure supplement 11). On the other hand, the overlaid initial and 200 ns conformations from the MD trajectories FI0, SI0, and CI0 indicated that the global conformation remains undistorted before the sharp change of distance (Figure 1—figure supplement 13); 950 protein structures (500 frames from FI0, the first 250 frames from SI0, and the first 200 frames from CI0) were concatenated for the PCA.

Analysis of the motion of the TRP domain

Request a detailed protocol

The motion of the TRP domain was characterized by the tilt angle and the rotation angle. The tilt angle of the TRP domain was defined as the variation of the angle between the z-axis and the axis of the TRP domain with respect to that of the initial conformation. The rotation angle was defined as the angle between the XY-plane projection of the axis of the TRP domain with that of the initial conformation.

Analyzing the role of the AR region in the force convey

Request a detailed protocol

In the simulations of system II, we calculated the reaction forces of the restraint on the LH domain of each chain of the tetramer, whose average magnitude and direction should be the same as the forces exerted on LH by the ARs. After that, these forces were projected on the x-y plane and z-direction (Figure 3—figure supplement 1). The calculation was performed from 20 to 40 ns in the trajectories CII1 and SII1, and from 50 to 100 ns in the trajectories CII8-17 and SII8-17.

For the calculation of the force constant of the ARs, the distance between the center of the LH and AR1 (the length of the AR region) was monitored in the MD trajectories. After the length of the AR region became stable, the force constant was calculated by using the formula k=F/Δz, where F was the applied force on each chain, and Δz was the variation of the length of the AR region with respect to the same value in the force-free simulations.

To estimate the force transfer speed through the AR region, we analyzed how long it took for the force applied to the AR1 to impact the LH domain (Figure 3—figure supplement 4). We generated five short trajectories (40 ps each) for the free/pushing/pulling simulations of system II with high output frequency (10 fs per frame). In the first stage of each trajectory, the force on the LH region could not be distinguished by the simulation conditions, and then, at some point, the force values on the LH domain started to deviate among the free/pushing/pulling simulations (Figure 3—figure supplement 4, where gray areas end). This represents the forces applied to AR1 starting to impact the LH domain. With this, we estimated that the forces applied on AR1 required about 6-12 ps to arrive at the LH domain, and the speed of force transfer was estimated to be 1.8 ± 0.2 nm ps−1 along the AR region.

Cell lines

Request a detailed protocol

S2 cells were purchased from China Center for Type Culture Collection (CCTCC), Serial number: GDC0138. The S2 cell line authentication was confirmed by COI authenticate. The mycoplasma contamination was tested negative by fluorescence quantitative PCR. The above cell tests were performed by a third party, Jiangsu Micro Spectrum Detection Technology Co., Ltd, and the cell test report Number was WJS-21046354-HJ-01-ER1.

Electrophysiological recording

Request a detailed protocol

Drosophila S2 cells were cultured in Schneider’s Insect medium supplied with 10% FBS at 27°C. TransIT-Insect Transfection Reagent (Mirus) was used to transfect cells according to the product protocol. The miniwhite region was deleted from all of the pUAST-NompC-EGFP plasmids (Supplementary file 1f and Figure 2—figure supplement 4) to enable site-directed mutagenesis. pUAST-NompC-EGFP (wild-type or mutants) constructs were co-transfected with pGal4 (Yan et al., 2013). Recordings were carried out 36–48 hr after transfection. Cells were transferred onto glass slides, pre-coated with Con A 30 minutes prior to recording.

Electrophysiological recordings were conducted under an Olympus CKX41 microscope equipped with a 40× water immersion lens. Transfected cells were identified by green fluorescence. The sample rate was 10 kHz and filtered at 1 kHz (low-pass). Patch electrodes with 12–20 MΩ resistance were used. The bath solution contained 140 mM NaMES (sodium methanesulfonate) and 10 mM HEPES. For cell-attach mode recording, the pipette solution was the same as the bath solution. For inside-out and outside-out mode recording, the pipette solution contained 140 mM potassium D-gluconate (CsMES in OAG application experiment) and 10 mM HEPES. All of the solutions were adjusted to 285 mOsm and pH 7.2.

After forming a specific recording mode (cell-attach mode, inside-out mode, or outside-out mode), negative pressure or positive pressure was applied to the excised membrane via a high-speed pressure clamp (HSPC, ALA-Scientific). Signals generated from pClamp software were sent to HSPC to control the timing and intensity of the pressure.

To record the dose-response curve of the mechanosensitive current, pressure steps of 500 ms with 10 mm Hg increment (for inside-out and outside-out recording) or 20 mm Hg increment (for cell-attach recording) were applied to the membrane patch through the recording pipette. The inside-out and outside-out patch-clamp traces under different pressure are shown in Figure 1—figure supplement 9A while the mean currents under different pressure are shown in Figure 1—figure supplement 9B.

Mutation generation

Request a detailed protocol

All of the point mutations on pUAST-NompC-EGFP plasmid were introduced by site-directed mutagenesis using a CloneExpress II One-step Cloning kit (Vazyme) and confirmed via sequencing of the mutation region. Further experiments were performed the same as outside-out and inside-out patch clamp in the wild-type NompC described in the electrophysiological recording.

Non-permeablized immunostaining of membranous NompC

Request a detailed protocol

For non-permeablized staining, the transfected cells were fixed and incubated with 4% paraformaldehyde at room temperature for 10 min. The cells were then washed with PBS three times and blocked with 4% BSA at 37°C for 100 min. The primary antibody (rabbit anti-αNOMPC-EC, 1:500; used in Zhang et al., 2015) was diluted in PBS and incubated with transfected cells at 4°C overnight. Cells were then washed with PBS three times and incubated with secondary antibody (Alexa Fluor 594 AffiniPure Donkey Anti-Rabbit IgG(H + L), 34212ES60, Yeasen) for 90 min at room temperature. After being washed briefly with PBS, cells were mounted on a coverslip for imaging.

Drug application

Request a detailed protocol

GdCl3 was dissolved in the bath solution (both bath and electrode solutions under cell-attached mode) to a final concentration of 100 µM. OAG was dissolved in DMSO and diluted in bath solution to a final concentration of 30 µM. The time window of 10 s was used to calculate the spontaneous open probability of NompC (NP0).

Data availability

All data generated or analysed during this study are included in the manuscript and supporting files. Numeric data files have been provided for Figure 1 B C D F G, Figure 1—figure supplements 2, 3, 4, 5, 8, 9, 10, 11, 12, Figure 2 A F, Figure 2—figure supplements 1, 2, 5, 6, Figure 3 C D F, Figure 3—figure supplements 1, 2, 3, 4, 5, 6, 7 and Figure 5 E, F.

References

    1. Montell C
    (2005) Drosophila TRP channels
    Pflügers Archiv - European Journal of Physiology 451:19–28.
    https://doi.org/10.1007/s00424-005-1426-2

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
  3. Boris Martinac
    Reviewer; Victor Chang Cardiac Research Institute, Australia
  4. Rachelle Gaudet
    Reviewer; Harvard University, United States

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

Acceptance summary:

The manuscript by Wang et al. describes a worthwhile study proposing a novel mechanism for the NompC ion channel gating by mechanical stimuli, demonstrating that NompC can only be open by application of mechanical force by pushing the bundle of ankyrin repeats outwards, causing the clockwise rotation of the pore helices and thus opening the channel. The model is interesting and appealing and given the rapid pace at which the mechanotransduction field is currently growing, the reviewers judge that this study will be of interest to a broad readership interested in molecular mechanobiology.

Decision letter after peer review:

Thank you for submitting your article "The push to open mechanism of the tethered mechanosensitive Ion channel NompC" 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 following individuals involved in review of your submission have agreed to reveal their identity: Boris Martinac (Reviewer #1); Rachelle Gaudet (Reviewer #3).

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

As the editors have judged that your manuscript is of potential interest, but as described below that several additional experiments and simulations are required before the conclusions' credibility can be assessed, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Summary:

The manuscript by Wang et al. describes a worthwhile study proposing a novel mechanism for the NompC ion channel gating by mechanical stimuli. NompC has to date been the only TRP-type ion channel that has convincingly been shown to be gated by the tethered mechanism. In this study, the authors used a combination of molecular dynamics simulations and patch clamp recording to demonstrate that NompC can only be open by application of mechanical force by pushing the bundle of ankyrin repeats outwards (with respect to the cell interior). This is consistent with what the 'rough' model of opening that Argudo et al. published in 2017 based on a combination of their continuum mechanics calculations and comparisons to the open state of the TRPV1 channel structure. The NompC ankyrin repeat bundle is proposed to act as a spring, which when pushed (compression of the spring) should cause the clockwise rotation of the pore helices and thus opening the channel. Pulling on the spring (stretching the spring) has a negligible effect on the channel opening. Simulation results are complemented by electrophysiology experiments in Drosophila S2 cells carried out to show that positive pressure (corresponding to compression) causes channel gating, but that much less current is evoked from negative compression. The model suggested is interesting and appealing and given the rapid pace at which the mechanotransduction field is currently growing, the reviewers judges that this study will be of interest to a broad readership interested in molecular mechanobiology.

There are, however, significant scientific issues and issues of presentation that made the manuscript as presented challenging to evaluate. The reviewers have therefore listed several aspects that need to be revised before the paper can be accepted for publication.

Essential revisions:

The majority of the conclusion stem from the interpretation of the MD simulations. The reviewers, however, expressed concern over the length and small number of repeats. Indeed, the simulations are very short for such large deformations, and the reviewers were unsure if it was possible to extract the true behavior of protein to a mechanical force on the 100 ns time scale, when the true force is applied over the microsecond to millisecond timescale. It is noteworthy that mechanical responses of proteins tend to have a huge dependence on time scale. Given the system size, the reviewers judge that it should be easy enough to double the time scale and collect many replicates to ensure that the results are robust. We note that given the issues related to presentation (see below), it is possible that the reviewers have misinterpreted the results and repeats were indeed carried out. For example, there were simulations of System I with different force constants, but those don't seem to be discussed anywhere in the text. Related to the interpretation of the MD simulations, the reviewers had the following comments:

The conformational changes observed in Figure 2b,c (and Supplemental Video S1C) are very small (especially panel c). Does this really open the channel? More analysis (with ions moving through the constriction zone for instance or water wetting/dewetting changes) must be done to support this claim. How much water is seen moving back and forth? What about free ion movement into and out of the cavity. Is unbiased ion entry to the cavity observed? Moreover, what about the local pore structure in the pushing experiment shown in Supp Video S1C really starts to open, is it side chain movement or all backbone? What are the backbone radius changes?

Figure 2d,e – is this based on gazing at the structure itself, or did were these residues identified from changes in the simulations? The authors mention that the values come from the simulations, but one wonders if they could have been guessed from the structure themselves.

Similarly, are Figure 3e residues identified from simulation or sequence gazing at the structure? Additionally, the authors state 'we found two stable hydrogen bonds between the ARs and LHs in our MD trajectories, between W1115 and D1142, and R1127 and E1163, respectively'. Cheng and Jan did not find potential hydrogen bonds directly from the atomic model of NOMPC at these two sites structure. The two pairs are either too far to establish a hydrogen bond (D1142-W1115: 5.4 Å) or on the borderline (E1163-R1127: 3.9 Å). The authors should comment on this point – perhaps this points to an important role in the simulations, but why do the simulations show h-bonding while the structure does not?

Looking at the Video S5: in all three simulations illustrated, the AR3-AR4 packing is disrupted. This is not something expected, yet it is not addressed anywhere in the manuscript and should be addressed. This will affect how the forces are transmitted through the protein.

The simulations of System II seem to suggest that pushing and pulling on AR1 while restraining the LH domain lead to the LH domain experiencing rotational forces (e.g. Figures 3b and S7). It seems like these are the forces that should be applied to System I, rather than pushing or pulling perpendicularly to the membrane plane.

The authors attempt to provide some experimental evidence that NompC is only pushed open by showing NompC failed to respond to negative pressure using cell-attached patch-clamp. However, previously the Jan lab performed the same experiment (Figure 5F, Zhang et al., Cell 2015), and their data showed NompC in fact could respond to negative pressure. The authors need to explain the inconsistency between their results and what the Jan lab has shown.

The authors state on p. 5, l. 10-102 that "Previous studies have shown that the AR regions are necessary for the mechano-gating of NompC, and the force from lipids cannot open the channel in the absence of the AR region (15)." However, this statement is incorrect: on p. 1401 of ref. 15, the authors of the Zhang et al., 2015 write: "We wish to emphasize that while our results strongly support the notion that ARs function as a tether for mechanogating of NOMPC, our results do not exclude the potential role of interactions between NOMPC protein and the membrane lipids nearby." And indeed, in the 2017 Nature paper by Jin et al. there is evidence that His1423 in the S4-S5 linker of NompC coordinates the lipid polar head group and is required for the channel mechanosensitivity because "the His1423Ala mutant is no longer responsive to mechanical stimulation" (p. 2). And further on p. 4 of this paper: "While it is unlikely that membrane deformation alone is the driving force in NOMPC gating, as in the case of other mechanosensitive channels such as Piezo and TRAAK channels, the proximity of a lipid molecule interacting with the functionally important His1423 at the S4-S5 linker suggests that lipid-protein interactions may have an important role in channel activity." The question is how to reconcile the views of Zhang et al., Jin et al. and the authors of this study, who seem to exclude the possibility of the bilayer involvement in NompC mechanogating? To answer this question, the reviewers have the following suggestion, which can be tested experimentally (and possibly using MD as well) by the authors of this study. Given the unidirectionality of the force causing the channel to open, it is likely that pushing the ankyrin bundle spring would locally curve the membrane bilayer. The local membrane curvature can activate a mechanosensitive channel by changing the asymmetry of the transbilayer pressure profile favouring the channel opening (Cox et al., Cell Reports 29(1): 1-12, 2019). In contrast, the force pulling on the spring would change the local membrane curvature in the opposite direction favouring the closed conformation of the channel. Given that the lipid bilayer is much more compliant compared to a membrane protein the displacement of the NompC channel with respect to the lipid bilayer by pushing the ankyrin spring could trigger the channel open conformation (Kung, Nature 436: 647-654, 2005). As shown by FE modelling (Bavi et al., Membranes 6: 14, 2016) a local curvature corresponding to a radius of less than 50 nm results in membrane tension of several mN/m, which is sufficient to open a mechanosensitive channel. The unidirectionality of the activation force reflecting the pushing vs. pulling the spring in the NompC case can be tested experimentally by inserting an amphipathic molecule, such as OAG (1-oleoyl-2-acetyl-sn-glycerol), a commercially available diacylglycerol (DAG) analogue, which like other amphipathic molecules can bend the cell membrane and was shown to activate the TRPC6 channel, which could not be activated by membrane stretch (Nikolaev et al. (2019) J Cell Sci 132: jcs238360). By adding OAG into the extracellular monolayer of the membrane bilayer it may be possible to open NompC (corresponding to pushing force), whereas adding it to the intracellular monolayer (corresponding to pulling force) would not open the channel. Thus the request from the reviewers is to perform patch clamp experiments using inside-out patch configuration as described in their manuscript. By adding OAG to the extracellular monolayer of the membrane patch accessible from the bath they should be able to open the NompC channel, whereas adding OAG from the pipette they would not be able to do it. Provided the authors could demonstrate this effect of OAG on NompC gating, these experiments together with the results of their study described in this manuscript would reconcile the tethered mechanism with the force-from-lipids paradigm in the case of NompC and indicate that "sharing the force" between the membrane bilayer and cytoskeleton is the likely mechanism of NompC activation by mechanical force.

Relatedly, P. 10, l. 205-206: "These results indicate that the interface between the AR and LH regions is also crucial for the force transduction, further confirming the tethered spring model for NompC." – Further to this statement the authors omitted to mention that His1423 residue within the S4-S5 linker was reported to be functionally important for NompC mechanosensitivity.

On P. 6, l. 125-128, the authors write: "Thus, consistent with previous structural studies of TRPV1.… it is the pushing force (compression of the intracellular domain) that leads to such a collective gating motion." – This is correct. However, it does not completely explain the mechanism of opening the channel. As stated above, the force opening the channel can also result from a change in the local membrane curvature causing the channel to open given that the force pushing on the ankyrin bundle spring could generate positive curvature and consequently change asymmetry of the transbilayer pressure profile as indicated by the FE modelling study (Bavi et al., Membranes 6: 14, 2016).

Additionally, the reviewers judged that many of the conclusions are over-interpreted:

Many of the mutations resulted in non-responsive channels. These non-responsive channels could be non-responsive for many reasons, not only gating defects (e.g. folding defects). The authors need to show that the membrane localization and quantity of NompC at the membrane is not affected by the mutations at the least.

The ion conductance in the simulation being lower than experimentally measured is interpreted as the channel conformation being only "partially open". But many other reasons could explain the observed conductance. Are there properties of these currents that you can extract that are not dependent on current or, more importantly, can single channel recordings be performed to confirm the impact on channel function?

Hypotheses about compensatory interactions in the hydrogen-bonding networks, etc, are presented as conclusions.

P. 9, L. 191-196: "Considering the fact that the length of the relaxed AR region is about 15 nm, we estimated that the force was transferred through the AR region at a speed of 1.8 ± 0.2 nm/ps.… the force is propagated via the membrane at a speed of 1.4 ± 0.5 nm/ps (37). Therefore, it appears that the force transfer speed in the tethered NompC channel is faster than that in the membranes." This statement requires statistical testing to compare quantitatively the two speeds of force propagation. As presented here, the force propagation through AR region does not appear to be significantly faster from the force propagation via the membrane bilayer. In contrary, they seem quite comparable.

P. 11, l. 232-234: "Nonetheless, a clear and detailed opening process of the channel in the presence of pushing forces was revealed by our MD simulations and validated by electrophysiological experiments." – MD simulations have not been validated experimentally because in outside-out patches CSK connections to the membrane bilayer/channels are largely disrupted, which has been shown to happen upon gigaohm seal formation (see for example Suchyna et al., Biophys J. 97(3): 738-747, 2009). Consequently, opening NompC by pushing the ankyrin spring is not the likely mechanism of the channel opening in membrane patches as suggested by the model shown in Figure 4A of the manuscript.

The reviewers identified presentation issues that obscured the message of the paper:

The methods and Table S1 indicate that three separate simulations with three different force constants were performed. But the authors do not indicate which of these simulations they used in each the figures or analyses. This lack of clarity as to which simulation is used in the analyses is pervasive throughout the manuscript, making a lot of the data difficult to interpret and evaluate. In general, all figure legends or specific mentions of simulations in the main or supplemental text should include a reference to the trajectory labels provided in Tables S1-S3.

The concatenation for PCA analysis is fine, but the data should be "unconcatenated" clearly in the figures. This is done in Figure 2A (although using saturation gradients to indicate progression through each of the three simulations would be useful). I.e. in Figure S3, for example, the three different simulations should clearly be demarcated.

The methods section reads more like a mixture of results and methods. The manuscript should be reorganized to separate the two. The methods are also inadequately detailed: Residue numbers should be used to indicate the residues included in various simulations, residues restrained in simulations, residue ranges used as AR1, AR29, LH, TM, etc. The published structure, 5VKQ, is missing residues 1603-1669 between the TRP box and the C-terminal helix that completes the LH domain, how was this handled in building the simulated systems?

The methods should provide enough detail that the exact full sequence of the expression vectors (including the codons used for the mutations) can be reconstructed.

https://doi.org/10.7554/eLife.58388.sa1

Author response

Essential revisions:

The majority of the conclusion stem from the interpretation of the MD simulations. The reviewers, however, expressed concern over the length and small number of repeats. Indeed, the simulations are very short for such large deformations, and the reviewers were unsure if it was possible to extract the true behavior of protein to a mechanical force on the 100 ns time scale, when the true force is applied over the microsecond to millisecond timescale. It is noteworthy that mechanical responses of proteins tend to have a huge dependence on time scale. Given the system size, the reviewers judge that it should be easy enough to double the time scale and collect many replicates to ensure that the results are robust. We note that given the issues related to presentation (see below), it is possible that the reviewers have misinterpreted the results and repeats were indeed carried out. For example, there were simulations of System I with different force constants, but those don't seem to be discussed anywhere in the text.

We are sorry for the confusion brought by the original supplementary tables, which contained multiple trial and validation simulations that were not discussed in the main text. We have revised these tables (Supplement File 1a – c), and only kept the production simulations.

In the original manuscript, the simulation time scale was determined by the pulling speed in the MD simulations, and longer simulation time would lead to the disruption of the global protein structure. Following the reviewers’ suggestions, we doubled the simulation time scale by slowing the pulling speed, and performed nine additional 500-ns simulations for the system I (three replicates for each condition, please refer to the revised Supplement File 1a). The simulation results were indeed very robust and consistent with our original observations (Figure 1—figure supplement 2 ).

Also, for system II, which contained more than 1,100,000 atoms, we performed extensive additional simulations with various pulling/pushing forces (2.2μs in total) to obtain more convincing results regarding the force constant of the ankyrin repeats region without disrupting the global structure (please refer to the revised Supplement File 1b and Figure 1—figure supplement 3).

The new simulations for the revision are highlighted in Supplement File 1a – b. We have also revised the corresponding method section and discussed these new simulations in the main text.

Related to the interpretation of the MD simulations, the reviewers had the following comments:

The conformational changes observed in Figure 2b,c (and Supplemental Video S1C) are very small (especially panel c). Does this really open the channel? More analysis (with ions moving through the constriction zone for instance or water wetting/dewetting changes) must be done to support this claim. How much water is seen moving back and forth? What about free ion movement into and out of the cavity. Is unbiased ion entry to the cavity observed? Moreover, what about the local pore structure in the pushing experiment shown in Supp Video S1C really starts to open, is it side chain movement or all backbone? What are the backbone radius changes?

Although the conformational changes were not large enough to fully open the channel, the observation of pore dilation was consistent in all of the pushing simulations (Figure 1D and new Figure 1 —figure supplement 2C ). Following the reviewer’s instructive suggestions, we have:

1. Analyzed the water distribution around the narrowest constriction zone, and observed an evident increase of the number of water molecules in the pushing simulations (new Figure 1 —figure supplement 5 ), supporting the notion that a pushing force tends to open the gate and thereby make it easier for water molecules to move back and forth around the gate of the channel.

2. Analyzed ion movement around the gate, and observed two Na+ spontaneously passing through the partially opened gate in the pushing simulations without any transmembrane potential (new new Figure 1—figure supplement 6).

3. Analyzed the backbone radius changes, which confirmed that the backbone pore was also dilated under a pushing force (new Figure 1—figure supplement 3 – 4). Thus, the observed pore dilation was not merely a side-chain movement but a global conformational change.

Figure 2d,e – is this based on gazing at the structure itself, or did were these residues identified from changes in the simulations? The authors mention that the values come from the simulations, but one wonders if they could have been guessed from the structure themselves.

Similarly, are Figure 3e residues identified from simulation or sequence gazing at the structure? Additionally, the authors state 'we found two stable hydrogen bonds between the ARs and LHs in our MD trajectories, between W1115 and D1142, and R1127 and E1163, respectively'. Cheng and Jan did not find potential hydrogen bonds directly from the atomic model of NOMPC at these two sites structure. The two pairs are either too far to establish a hydrogen bond (D1142-W1115: 5.4 Å) or on the borderline (E1163-R1127: 3.9 Å). The authors should comment on this point – perhaps this points to an important role in the simulations, but why do the simulations show h-bonding while the structure does not?

The H-bonds shown in Figure 2D-2E and Figure 3E were identified from our MD trajectories (Supplement File 1d), most of which were also present in the cryo-EM structure. The agreement between the MD simulations and cryoEM structure was reasonable and can be viewed as cross-validation between the structure and the simulations. The exception was Q1253-S1577, which formed a hydrogen bond for ~50% of our simulation time but not in the cryo-EM structure (Supplement File 1d). It was noticeable that this H-bond was more stable when the AR region was being pushed, while less stable when the AR region was being pulled or free (Supplement File 1b ). Therefore, the formation of this pair of hydrogen bond was stabilized by the pushing force, partially explaining why it was observed in our MD simulation but not in the “force-free” cryo-EM structure.

In the meantime, it was also likely that the MD simulations optimized the local structures. As the reviewer pointed out, we also observed another two stable pairs of H-bonds in our MD simulations that were not present in the cryo-EM structure (D1142-W1115 and E1163-R1127, Figure 3E and Supplement File 1d ), and we think this was due to the fact that MD simulations can dynamically explore the free energy surface around the cryo-EM structure, and thus further optimize the local conformation of the structural model. Considering the resolution of the cryo-EM structure (3.55 Å) was not high enough to see the exact side-chain locations, it was possible that our MD simulations further optimized the model to get a better sampling of the side-chain conformations. As can be seen in the newFigure 3—figure supplement 7. , the distances between the side chains of the above two pairs of hydrogen bond-forming residues obtained from the MD simulations trajectories were closer than those in the cryo-EM structure, while the distances between the α carbon atoms were nearly identical, indicating that the differences observed in the MD simulations and cryo-EM structure were due to the side-chain movements in the simulations. As the MD simulations took into account all the atomistic interactions, including water molecules, ions, lipids, and hydrogen atoms of the protein, we believe the MD simulation result may represent a more optimized model. We have discussed this in the relevant result section.

Looking at the Video S5: in all three simulations illustrated, the AR3-AR4 packing is disrupted. This is not something expected, yet it is not addressed anywhere in the manuscript and should be addressed. This will affect how the forces are transmitted through the protein.

This disruption of the ARs packing was probably caused by the strong and fast pulling forces applied to AR1. To validate, we performed many additional simulations with smaller forces applied to AR1 (Supplement File 1b). In the weaker-force simulations, the AR region was not disrupted, and we obtained a similar force constant of around 3 pN/nm, although with larger error bars (new Figure 3—figure supplement 3 ). We have added a new Video 6 and discussed it in the revised text. That being said, it was also possible that the disruption was due to the absence of the microtubule, which may assist to stabilize the terminal ARs.

The simulations of System II seem to suggest that pushing and pulling on AR1 while restraining the LH domain lead to the LH domain experiencing rotational forces (e.g. Figures 3b and S7). It seems like these are the forces that should be applied to System I, rather than pushing or pulling perpendicularly to the membrane plane.

We projected the forces on the Z-axis and XY plane, and the results showed that the pushing force is still the dominant part of the mechanical stimuli appreciated by the LH domain (new Figure 3—figure supplement 1 ). In the meantime, our results showed that a perpendicular pushing force was enough to drive the channel to open (and the TRP domain to rotate), so we mainly focused on these forces in this work.

The reviewer did raise a very interesting question, and we agree that the rotational forces on the XY plane may assist to drive the TRP domain to move clockwise. This kind of simulation is much more complicated to implement, and we may need to develop a new simulation method to systematically investigate this in future work.

The authors attempt to provide some experimental evidence that NompC is only pushed open by showing NompC failed to respond to negative pressure using cell-attached patch-clamp. However, previously the Jan lab performed the same experiment (Figure5F, Zhang et al., Cell 2015), and their data showed NompC in fact could respond to negative pressure. The authors need to explain the inconsistency between their results and what the Jan lab has shown.

Our results of cell-attached (and inside-out) recordings were different from Zhang et al. (Ref 15, Zhang et al., Cell 2015), while our outside-out recordings showed consistent results with Zhang et al. (Ref 15, Zhang et al., Cell 2015), which confirmed that NompC can be activated by pressure (mechanical forces). We noticed that the resistance of the electrodes used by Zhang et al. was 5-7 MΩ, while ours was 12-20 MΩ. The resistance of our electrode is 2-3 times larger compared to the previous work, which might explain the different observations. In such a case, our electrodes were much thinner, meaning the patched area sucked into the electrodes in our recordings was much smaller than in the previous work. Such a big difference in the patched area may cause different physical changes of the membranes when the pressure is applied. For instance, if the patched area is large, the patched membrane may tend to bend rather than move upwards or downwards along the electrode. On the other hand, if the patched area is small, the patched area can be harder to bend, and therefore it may move upwards and downwards fiercely instead, which may cause direct compressing or stretching of the ankyrin repeat domain. Therefore, we believe our recordings with the smaller patched area may better minimize the effect of membrane curvature, and therefore better represent the effects of stretching or compressing of the ankyrin repeat domains.

The authors state on p. 5, l. 10-102 that "Previous studies have shown that the AR regions are necessary for the mechano-gating of NompC, and the force from lipids cannot open the channel in the absence of the AR region (15)." However, this statement is incorrect: on p. 1401 of ref. 15, the authors of the Zhang et al., 2015 write: "We wish to emphasize that while our results strongly support the notion that ARs function as a tether for mechanogating of NOMPC, our results do not exclude the potential role of interactions between NOMPC protein and the membrane lipids nearby." And indeed, in the 2017 Nature paper by Jin et al. there is evidence that His1423 in the S4-S5 linker of NompC coordinates the lipid polar head group and is required for the channel mechanosensitivity because "the His1423Ala mutant is no longer responsive to mechanical stimulation" (p. 2). And further on p. 4 of this paper: "While it is unlikely that membrane deformation alone is the driving force in NOMPC gating, as in the case of other mechanosensitive channels such as Piezo and TRAAK channels, the proximity of a lipid molecule interacting with the functionally important His1423 at the S4-S5 linker suggests that lipid-protein interactions may have an important role in channel activity." The question is how to reconcile the views of Zhang et al., Jin et al. and the authors of this study, who seem to exclude the possibility of the bilayer involvement in NompC mechanogating? To answer this question, the reviewers have the following suggestion, which can be tested experimentally (and possibly using MD as well) by the authors of this study. Given the unidirectionality of the force causing the channel to open, it is likely that pushing the ankyrin bundle spring would locally curve the membrane bilayer. The local membrane curvature can activate a mechanosensitive channel by changing the asymmetry of the transbilayer pressure profile favouring the channel opening (Cox et al., Cell Reports 29(1): 1-12, 2019). In contrast, the force pulling on the spring would change the local membrane curvature in the opposite direction favouring the closed conformation of the channel. Given that the lipid bilayer is much more compliant compared to a membrane protein the displacement of the NompC channel with respect to the lipid bilayer by pushing the ankyrin spring could trigger the channel open conformation (Kung, Nature 436: 647-654, 2005). As shown by FE modelling (Bavi et al., Membranes 6: 14, 2016) a local curvature corresponding to a radius of less than 50 nm results in membrane tension of several mN/m, which is sufficient to open a mechanosensitive channel. The unidirectionality of the activation force reflecting the pushing vs. pulling the spring in the NompC case can be tested experimentally by inserting an amphipathic molecule, such as OAG (1-oleoyl-2-acetyl-sn-glycerol), a commercially available diacylglycerol (DAG) analogue, which like other amphipathic molecules can bend the cell membrane and was shown to activate the TRPC6 channel, which could not be activated by membrane stretch (Nikolaev et al. (2019) J Cell Sci 132: jcs238360). By adding OAG into the extracellular monolayer of the membrane bilayer it may be possible to open NompC (corresponding to pushing force), whereas adding it to the intracellular monolayer (corresponding to pulling force) would not open the channel. Thus the request from the reviewers is to perform patch clamp experiments using inside-out patch configuration as described in their manuscript. By adding OAG to the extracellular monolayer of the membrane patch accessible from the bath they should be able to open the NompC channel, whereas adding OAG from the pipette they would not be able to do it. Provided the authors could demonstrate this effect of OAG on NompC gating, these experiments together with the results of their study described in this manuscript would reconcile the tethered mechanism with the force-from-lipids paradigm in the case of NompC and indicate that "sharing the force" between the membrane bilayer and cytoskeleton is the likely mechanism of NompC activation by mechanical force.

We thank the reviewers for pointing out this important question. We focused on the mechanical properties of NompC itself and omitted the possible effect of membrane curvature and lateral pressure changes on the gating of NompC, which was indeed flawed. Following the reviewers’ suggestion, we conducted additional patch-clamp recordings with OAG added. Our results showed that the activation was indeed unidirectional: by adding OAG to the intracellular monolayer, the NompC was more likely to be activated; while adding OAG to the extracellular monolayer showed the opposite effect (new Figure 5 E -F ). This is also consistent with the previous study on TRPC6 (Ref 41, Nikolaev et al., J Cell Sci 2019). Given both NompC and TRPC6 belonging to the TRP family, this may imply a common mechanism of the family.

We think this is also consistent with the push-to-open mechanism. Given a relaxed and planar bilayer containing NompC, pushing ARs from the intracellular side would lead to larger lateral stress in the intracellular monolayer than the extracellular monolayer. Adding OAG to the intracellular side would make the intracellular monolayer more crowded, and thus a larger lateral pressure than the extracellular monolayer as well. Therefore, adding OAG to the intracellular monolayer would generate a similar effect on the transmembrane stress profile to applying a pushing force on the AR domain toward the extracellular side. Indeed, it is possible that the pushing force would drive the pore to open and at the same time generate an asymmetry of the transbilayer stress profile that can facilitate gating. It is not clear which is more important, but it is highly likely that both effects work together to open the channel.

We have deleted the statement “the force from lipids cannot open the channel in the absence of the AR region” and edited the text accordingly. Also, we added the new experimental data as well as new text and citations in the Discussion section. We thank the reviewers very much for directing us to give a more comprehensive discussion.

Relatedly, P. 10, l. 205-206: "These results indicate that the interface between the AR and LH regions is also crucial for the force transduction, further confirming the tethered spring model for NompC." – Further to this statement the authors omitted to mention that His1423 residue within the S4-S5 linker was reported to be functionally important for NompC mechanosensitivity.

On P. 6, l. 125-128, the authors write: "Thus, consistent with previous structural studies of TRPV1.… it is the pushing force (compression of the intracellular domain) that leads to such a collective gating motion." – This is correct. However, it does not completely explain the mechanism of opening the channel. As stated above, the force opening the channel can also result from a change in the local membrane curvature causing the channel to open given that the force pushing on the ankyrin bundle spring could generate positive curvature and consequently change asymmetry of the transbilayer pressure profile as indicated by the FE modelling study (Bavi et al., Membranes 6: 14, 2016).

Following the reviewer’s suggestion, we further analyzed the interaction between the H1423 and lipid molecules in our MD trajectories. Indeed, it was interesting to see that H1423 can stably interact with a lipid (POPC) molecule in the simulation trajectories, and the lipid molecule was not located at the interaction site in the initial setup but diffused to the right place to act as a bridge to link H1423 and S1268 during the simulations (new Figure 5 A – D ). Therefore, it seems that lipid molecules stabilized the local structure by interacting with H1423, which may be necessary for gating. Also, as discussed above, membrane curvature-induced asymmetrical stress profile may play a role to facilitate gating, and H1423 is probably a key residue for conveying forces from lipids to the NompC channel. We have added the relevant data and rephrased the gating mechanism in the Discussion section to make it more comprehensive.

Additionally, the reviewers judged that many of the conclusions are over-interpreted:

Many of the mutations resulted in non-responsive channels. These non-responsive channels could be non-responsive for many reasons, not only gating defects (e.g. folding defects). The authors need to show that the membrane localization and quantity of NompC at the membrane is not affected by the mutations at the least.

Thanks for the suggestion. We performed non-permeablized staining with an antibody targeting the pore-helix (α-NompC-EC) (same as in Ref 15, Zhang et al., Cell 2015) to quantify membranous NompC.

Our staining results showed that for the mutations that led to significant loss-of-function, there were no significant quantity decreases on membranous NompC, except for S1577A and W1572A (Figure 2—figure supplement 5. ). Notably, these two mutations led to a moderate degree of quantity decreases of membranous NompC, but complete loss-of-function in the outside-out recordings, indicating that the two mutations altered the mechnosensitivity of NompC in addition to the quantity of NompC on membranes. Therefore, the results suggested that the point mutations discussed in our work did influence the intrinsic mechanoactivity of NompC.

The other special case is S1421A. As discussed in the text for the case of S1421A, the hydrogen bonding with S1421 is on its backbone rather than side chain, so we expected that the activity of S1421A would not be significantly changed if it can still get into the membrane. Therefore, the observation that the abundance of the membranous S1421A (Figure 2—figure supplement 5. ) and the activity of S1421A (Figure 2F) both decreased was reasonable.

We have added the data in the SI and briefly discussed this in the main text.

The ion conductance in the simulation being lower than experimentally measured is interpreted as the channel conformation being only "partially open". But many other reasons could explain the observed conductance. Are there properties of these currents that you can extract that are not dependent on current or, more importantly, can single channel recordings be performed to confirm the impact on channel function?

We think the difference of the conductances measured from our simulations and experiments is due to the limitation of the simulation time scale, which is a common problem in the MD simulations: the full gating processes of ion channels usually occur on a time scale of millisecondwhile current MD simulations fall on the timescale of hundreds to thousands of nanoseconds, which is not long enough to obtain a fully opened structure, and therefore the ion conductance in the simulations should be lower than that in the fully open state.

Nevertheless, the trend of pore opening caused by the pushing force was very clear and well reproducible in the MD simulations, so we believe the push-to-open gating mechanism is solid; and we conducted further qualitative validations that were not dependent on the specific current values:

1. We conducted cell-attached patch-clamp experiments: from our MD simulation results, we suspected that a positive pressure would activate the NompC channel but not a negative one, and that was what we observed in the cell-attached patch-clamp experiments; adding Gd3+ to the bath killed the currents (Figure 1—figure supplement 10), which further confirmed the currents were due to the ion permeation through NompC.

2. By analyzing the interaction networks in MD trajectories, we proposed several potentially important residue pairs involved in the hydrogen bonding that should be important for stabilizing the structure and mechanosensitivity of NompC, and the mutagenesis and electrophysiology experiments validated that these mutations did suppress or abolish the mechanosensitivity of the NompC channel without significantly changing their localization and abundance on membranes ( Figure 2 —figure supplement 5 and Figure 3 – —figure supplement 5).

The single-channel recordings and conductance measurements were performed thoroughly in previous studies by one of our corresponding authors and his colleagues (Ref 9, Yan et al., Nature 2013; Ref 15, Zhang et al., Cell 2015), and they obtained a single-channel conductance of ~150 pS for NompC (~10 pA under 60 mV). This was also observed in our electrophysiology experiments (Figure 5 E- F ).

Taken together, we believe that the difference of conductance values observed in the MD simulations and electrophysiology was due to the common problem of insufficient sampling of MD simulations, but our qualitative results and validations were robust to support the push-to-open gating model.

Hypotheses about compensatory interactions in the hydrogen-bonding networks, etc, are presented as conclusions.

We thank the reviewer for pointing this out. We should have included the analysis to show that the compensatory H-bonding was observed in our simulations. In the revision, we conducted quantitative analysis on these hydrogen bond formations in the trajectories. The results are in the Figure 2—figure supplement 6 and Figure 3—figure supplement 6 and clearly show new H-bonds forming in the mutants during the simulations.

P. 9, L. 191-196: "Considering the fact that the length of the relaxed AR region is about 15 nm, we estimated that the force was transferred through the AR region at a speed of 1.8 ± 0.2 nm/ps.… the force is propagated via the membrane at a speed of 1.4 ± 0.5 nm/ps (37). Therefore, it appears that the force transfer speed in the tethered NompC channel is faster than that in the membranes." This statement requires statistical testing to compare quantitatively the two speeds of force propagation. As presented here, the force propagation through AR region does not appear to be significantly faster from the force propagation via the membrane bilayer. In contrary, they seem quite comparable.

We have revised the text: “Therefore, it appears that the force transfer speed in the tethered NompC channel is comparable to, or slightly faster, than that in the membranes.”

P. 11, l. 232-234: "Nonetheless, a clear and detailed opening process of the channel in the presence of pushing forces was revealed by our MD simulations and validated by electrophysiological experiments." – MD simulations have not been validated experimentally because in outside-out patches CSK connections to the membrane bilayer/channels are largely disrupted, which has been shown to happen upon gigaohm seal formation (see for example Suchyna et al., Biophys J. 97(3): 738-747, 2009). Consequently, opening NompC by pushing the ankyrin spring is not the likely mechanism of the channel opening in membrane patches as suggested by the model shown in Figure 4A of the manuscript.

Here we mainly meant that the cell-attached patch-clamp experiments supported our MD simulation results. We do agree that there are more uncertainties in the outside-out patch-clamp experiments, but the agreement between the MD simulation results, the cell-attached patch-clamp, and the outside-out patch-clamp was reassuring.

We have changed our wording in the text: ‘Nonetheless, a clear and detailed opening trend of the channel in the presence of pushing forces was seen in the MD simulations and this was supported by electrophysiological experiments. The combination of the two methods provides a plausible “push-to-open” gating model for tethered mechanosensitive ion channels.’

The reviewers identified presentation issues that obscured the message of the paper:

The methods and Table S1 indicate that three separate simulations with three different force constants were performed. But the authors do not indicate which of these simulations they used in each the figures or analyses. This lack of clarity as to which simulation is used in the analyses is pervasive throughout the manuscript, making a lot of the data difficult to interpret and evaluate. In general, all figure legends or specific mentions of simulations in the main or supplemental text should include a reference to the trajectory labels provided in Tables S1-S3.

Sorry for the confusion and thanks for the suggestions. We have done the following to solve the presentation issues:

1. We revised Supplement File 1a – c to include the production simulations only and labeled all the trajectories.

2. We have included references to the above trajectory labels for all the figure legends. Some of the specific mentions of simulations in the main or supplemental text have also been referenced to the labels when necessary. The other mentions in the text refer to the figures that already have the references, so these are not labeled again.

The concatenation for PCA analysis is fine, but the data should be "unconcatenated" clearly in the figures. This is done in Figure 2A (although using saturation gradients to indicate progression through each of the three simulations would be useful). I.e. in Figure S3, for example, the three different simulations should clearly be demarcated.

We have updated the Figures (Figure 2A and Figure 2 – supplement figure 1 in the revision) to demarcate the trajectories.

The methods section reads more like a mixture of results and methods. The manuscript should be reorganized to separate the two. The methods are also inadequately detailed: Residue numbers should be used to indicate the residues included in various simulations, residues restrained in simulations, residue ranges used as AR1, AR29, LH, TM, etc. The published structure, 5VKQ, is missing residues 1603-1669 between the TRP box and the C-terminal helix that completes the LH domain, how was this handled in building the simulated systems?

The methods should provide enough detail that the exact full sequence of the expression vectors (including the codons used for the mutations) can be reconstructed.

Many thanks for these suggestions. we have rewritten the method section by removing the redundant text and moving the relevant results to the main text, and added a new Supplement File 1e and a new Figure 1 – supplement figure 1 to present the residue ranges used in our MD simulations.

The missing residues 1603-1669 were ignored in our MD simulations, as modelling such a large unstable structure would introduce too many uncertainties. The C-terminal helix was removed from the simulation system, as it is located in the large gap between adjacent AR chains, not between the AR and LH domains, and therefore it is expected to play a minor role in the force convey. This was not a perfect solution, but we think the artifacts would be minor.

https://doi.org/10.7554/eLife.58388.sa2

Article and author information

Author details

  1. Yang Wang

    1. Center for Quantitative Biology, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China
    2. Peking-Tsinghua Center for Life Sciences, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China
    Contribution
    Data curation, Formal analysis, Validation, Investigation, Visualization, Writing - original draft, Writing - review and editing
    Contributed equally with
    Yifeng Guo
    Competing interests
    No competing interests declared
    Additional information
    Affiliations 1, 2 and 3 contributed equally to this work
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5504-9800
  2. Yifeng Guo

    State Key Laboratory of Medical Neurobiology and MOE Frontiers Center for Brain Science, Institute of Brain Science, School of Life Sciences, Fudan University, Shanghai, China
    Contribution
    Data curation, Formal analysis, Validation, Investigation, Writing - original draft, Writing - review and editing
    Contributed equally with
    Yang Wang
    Competing interests
    No competing interests declared
    Additional information
    Affiliations 1, 2 and 3 contributed equally to this work
  3. Guanluan Li

    1. State Key Laboratory of Medical Neurobiology and MOE Frontiers Center for Brain Science, Institute of Brain Science, School of Life Sciences, Fudan University, Shanghai, China
    2. Institute of Molecular Physiology, Shenzhen Bay Laboratory, Shenzhen, China
    Contribution
    Data curation, Validation
    Competing interests
    No competing interests declared
  4. Chunhong Liu

    Center for Quantitative Biology, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China
    Contribution
    Formal analysis, Validation, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3666-2343
  5. Lei Wang

    Center for Quantitative Biology, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  6. Aihua Zhang

    1. Center for Quantitative Biology, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China
    2. Peking-Tsinghua Center for Life Sciences, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China
    Contribution
    Software, Formal analysis, Validation
    Competing interests
    No competing interests declared
  7. Zhiqiang Yan

    1. State Key Laboratory of Medical Neurobiology and MOE Frontiers Center for Brain Science, Institute of Brain Science, School of Life Sciences, Fudan University, Shanghai, China
    2. Institute of Molecular Physiology, Shenzhen Bay Laboratory, Shenzhen, China
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Validation, Investigation, Methodology, Project administration, Writing - review and editing
    For correspondence
    zqyan@szbl.ac.cn
    Competing interests
    No competing interests declared
  8. Chen Song

    1. Center for Quantitative Biology, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China
    2. Peking-Tsinghua Center for Life Sciences, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Investigation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    c.song@pku.edu.cn
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9730-3216

Funding

National Natural Science Foundation of China (32071251)

  • Chen Song

National Natural Science Foundation of China (21873006)

  • Chen Song

National Natural Science Foundation of China (31571083)

  • Zhiqiang Yan

National Natural Science Foundation of China (31970931)

  • Zhiqiang Yan

Ministry of Science and Technology of the People's Republic of China (2016YFA0500401)

  • Chen Song

Ministry of Science and Technology of the People's Republic of China (2017YFA0103900)

  • Zhiqiang Yan

Ministry of Science and Technology of the People's Republic of China (2016YFA0502800)

  • Zhiqiang Yan

Program for Professor of Special Appointment, Eastern Scholar of Shanghai (TP2014008)

  • Zhiqiang Yan

Shanghai Municipal Science and Technology Major Project (No.2017SHZDZX01)

  • Zhiqiang Yan

Shanghai Rising-Star Program (14QA1400800)

  • Zhiqiang Yan

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

Acknowledgements

We thank Prof Wei Zhang at Tsinghua University for his generous sharing the antibody with us. The research was supported by the National Natural Science Foundation of China (32071251 and 21873006 to CS; 31571083 and 31970931 to ZY), and the National Key Research and Development Program of the Ministry of Science and Technology of China (2016YFA0500401 to CS; 2017YFA0103900 and 2016YFA0502800 to ZY). ZY was supported by the Program for Professor of Special Appointment (Eastern Scholar of Shanghai, TP2014008), the Shanghai Municipal Science and Technology Major Project (No. 2017SHZDZX01) and ZJLab, and the Shanghai Rising-Star Program (14QA1400800). Part of the MD simulation was performed on the Computing Platform of the Center for Life Sciences at Peking University, and part of the MD simulation was performed on the Tianhe II supercomputer in the National Supercomputing Center in Tianjin.

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

Reviewers

  1. Boris Martinac, Victor Chang Cardiac Research Institute, Australia
  2. Rachelle Gaudet, Harvard University, United States

Publication history

  1. Received: April 29, 2020
  2. Accepted: May 16, 2021
  3. Version of Record published: June 8, 2021 (version 1)

Copyright

© 2021, Wang et al.

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

Metrics

  • 722
    Page views
  • 65
    Downloads
  • 0
    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)

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

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

  1. Further reading

Further reading

    1. Computational and Systems Biology
    2. Neuroscience
    Jack Goffinet et al.
    Research Article Updated

    Increases in the scale and complexity of behavioral data pose an increasing challenge for data analysis. A common strategy involves replacing entire behaviors with small numbers of handpicked, domain-specific features, but this approach suffers from several crucial limitations. For example, handpicked features may miss important dimensions of variability, and correlations among them complicate statistical testing. Here, by contrast, we apply the variational autoencoder (VAE), an unsupervised learning method, to learn features directly from data and quantify the vocal behavior of two model species: the laboratory mouse and the zebra finch. The VAE converges on a parsimonious representation that outperforms handpicked features on a variety of common analysis tasks, enables the measurement of moment-by-moment vocal variability on the timescale of tens of milliseconds in the zebra finch, provides strong evidence that mouse ultrasonic vocalizations do not cluster as is commonly believed, and captures the similarity of tutor and pupil birdsong with qualitatively higher fidelity than previous approaches. In all, we demonstrate the utility of modern unsupervised learning approaches to the quantification of complex and high-dimensional vocal behavior.

    1. Computational and Systems Biology
    2. Evolutionary Biology
    Rossana Droghetti et al.
    Research Article Updated

    Recent results comparing the temporal program of genome replication of yeast species belonging to the Lachancea clade support the scenario that the evolution of the replication timing program could be mainly driven by correlated acquisition and loss events of active replication origins. Using these results as a benchmark, we develop an evolutionary model defined as birth-death process for replication origins and use it to identify the evolutionary biases that shape the replication timing profiles. Comparing different evolutionary models with data, we find that replication origin birth and death events are mainly driven by two evolutionary pressures, the first imposes that events leading to higher double-stall probability of replication forks are penalized, while the second makes less efficient origins more prone to evolutionary loss. This analysis provides an empirically grounded predictive framework for quantitative evolutionary studies of the replication timing program.