Molecular basis of force-from-lipids gating in the mechanosensitive channel MscS

Version of Record

Accepted for publication after peer review and revision.

Download
Cite
Share
CommentOpen annotations (there are currently 0 annotations on this page).

Version of Record published: January 24, 2020 (This version)
Accepted: December 27, 2019
Accepted Manuscript published: December 27, 2019 (Go to version)
Received: July 24, 2019

1. Of interest
A pH-sensitive switch activates virulence in Salmonella

Dasvit Shetty, Linda J Kenney

Research Article Sep 25, 2023
Further reading

Abstract
Introduction
Results
Discussion
Materials and methods
Data availability
References
Decision letter
Author response
Article and author information
Metrics

Abstract

Prokaryotic mechanosensitive (MS) channels open by sensing the physical state of the membrane. As such, lipid-protein interactions represent the defining molecular process underlying mechanotransduction. Here, we describe cryo-electron microscopy (cryo-EM) structures of the E. coli small-conductance mechanosensitive channel (MscS) in nanodiscs (ND). They reveal a novel membrane-anchoring fold that plays a significant role in channel activation and establish a new location for the lipid bilayer, shifted ~14 Å from previous consensus placements. Two types of lipid densities are explicitly observed. A phospholipid that ‘hooks’ the top of each TM2-TM3 hairpin and likely plays a role in force sensing, and a bundle of acyl chains occluding the permeation path above the L105 cuff. These observations reshape our understanding of force-from-lipids gating in MscS and highlight the key role of allosteric interactions between TM segments and phospholipids bound to key dynamic components of the channel.

Introduction

In principle, all molecules are mechanosensitive. Remarkably, some have evolved as force transducers, where they participate in a variety of fundamental biological functions, including turgor control in plants, development and morphogenesis, touch, hearing, proprioception, as well as osmoregulation in bacteria (Haswell et al., 2011; Katta et al., 2015; Kung, 2005; Ladoux and Mège, 2017; Murthy et al., 2017). Many of these functions are driven by the activity of mechanosensitive channels, switches that couple force sensing with the electrical activity of cells (Cox et al., 2018; Naismith and Booth, 2012; Perozo, 2006; Sukharev and Corey, 2004). When the membrane is stretched, the resulting change in the trans-bilayer pressure profile will drive the conformational equilibrium of membrane proteins (Cantor, 1999; Gullingsrud and Schulten, 2004). To explain this phenomenon, several physical properties of the lipid bilayer have been considered (Perozo, 2006). Nevertheless, understanding the nature of the lipid forces that drive MS channel gating continues to be one of the fundamental questions in biological mechanotransduction.

In prokaryotes, MscL and MscS are the foundational members of two structurally distinct classes of MS channels (Cox et al., 2018; Kung et al., 2010; Naismith and Booth, 2012). Sensitive to tension changes in the plane of the bilayer, they play a key role in the response to osmotic challenges and remain a de facto standard in the search for the molecular principles underlying membrane force transduction. The E. coli MscS crystal structure in detergent (Bass et al., 2002; Steinbacher et al., 2007) revealed a homoheptamer with three TM segments (TM1, TM2 and TM3a/b) and a large cytoplasmic C-terminal domain cradling a water-filled internal cavity. The narrowest region along the permeation path is located at L105 in TM3a, and while its van der Waals diameter can be larger than 7 Å, the current consensus is that the original crystal structure represents a non-conductive conformation (Anishkin et al., 2010). However, whether it corresponds to a closed or an inactivated state has been a matter of discussion (Anishkin and Sukharev, 2004; Anishkin et al., 2008a; Anishkin et al., 2008b). Structures believed to represent a conducting or expanded state have been obtained for the gain of function (GOF) mutant A106V (Wang et al., 2008) and in DDM-solubilized wt-MscS (Lai et al., 2013). A comparison with wt-MscS revealed that transition to this expanded state appears to be associated with rearrangements in the packing interface between TM3 helices. These crystallographically-derived conformational changes (in detergent) are in general agreement with lower resolution analyses of amphiphile-driven gating of MscS in a lipid bilayer (Vasquez et al., 2008).

Given the existing set of MscS structures, a few plausible mechanisms have been proposed to describe how bilayer forces lead to channel opening. Early hypotheses suggested that membrane lateral forces at the aqueous interface (Nomura et al., 2006) disrupt the hydrophobic interactions between TM1-TM2 hairpin, straightening the tilt angle of TM3a and TM3b and opening the channel (Anishkin et al., 2008b; Vasquez et al., 2008). Alternatively, at rest, internal elastic strain energy might simply be countered by the bilayer pressure profile, keeping the channel closed (Anishkin et al., 2008b; Malcolm et al., 2015). More recently, a unique mechanism has been proposed based on the idea that bilayer lipids can drive MscS conformation by acting as ligands while they freely exchange between the membrane and hydrophobic pockets between TM2 and TM3a (Pliotas et al., 2015). At rest, these pockets are saturated with phospholipids, preventing structural rearrangements in TM3a. But as tension is applied, lipids diffuse away, leading to TM3a movements and opening the channel. This model requires not only that the TM2/TM3a cavity must be located at the membrane/water interface but it also takes advantage of the presumed membrane deformations induced by the interaction of the ‘angled’ TM1-TM2 seen in the nonconductive MscS crystal structures (Phillips et al., 2009).

Understanding the fundamental role played by bilayer forces in MS channel gating requires the elucidation of channel structures in a lipid bilayer environment. Yet, until recently (Rasmussen et al., 2019), all available high-resolution structures have been determined in detergent micelles, either by crystallographic (Bass et al., 2002; Dong et al., 2015; Lai et al., 2013; Pliotas et al., 2015; Steinbacher et al., 2007; Wang et al., 2008) or single particle cryo-EM methods (Guo and MacKinnon, 2017; Saotome et al., 2018; Zhao et al., 2018). Solved in the absence of the lipid bilayer, these structures have been essential in understanding the structural basis of MS channel function, but cannot, on their own, elucidate the central role that lipid-protein interactions play in force-from-lipid mechanotransduction. However, in a recent EM structure of MscS in a nanodisc (Rasmussen et al., 2019), lipids were observed in the TM2-TM3 hairpin cavity, two lipids parallel to TM3b, and a putative lipid density was reported along the permeation pathway. Though not explicitly stated, this model also suggests a shift in the membrane footprint. Independently, we have solved the nanodisc-reconstituted MscS (MscS-ND) by single particle cryo-electron microscopy (cryoEM) under a variety of lipid compositions and protein constructs (Figure 1—figure supplement 1) as well as a DDM detergent structure. These structures, together with electrophysological and computational data highlight new membrane-interacting regions at MscS N-terminal end and define structurally bound lipids with potentially important roles in mechanotransduction, gating, and permeation.

Results

Structure of membrane-embedded MscS

We first determined the structure of nanodisc-reconstituted MscS (PC:PG, 4:1) with a histidine tag at its N-terminal end (6xHis-MscS-ND) to a resolution of 3.1 Å (Figure 1—figure supplement 2). While the EM structure (Figure 1A) recapitulates some of the major characteristic seen in the MscS crystal structure (Bass et al., 2002; Steinbacher et al., 2007), the channel shows a slightly different angle for the TM1-TM2 hairpin and displays additional density towards its periplasmic face. This is a consequence of a newly resolved N-terminal domain and additional ~3 turns of helix that further extends TM1 in MscS-ND (Figure 1A, right). This new density is also present in the absence of imposed symmetry (Figure 1—figure supplement 3). This domain displays some anisotropic behavior, and together with the lower end of the TM1-TM2 hairpin, they represent the most flexible regions of the channel (Figure 1—figure supplement 2).

Figure 1 with 5 supplements see all

Download asset Open asset

Structure of membrane-embedded MscS (MscS-ND) and its anchor domain.

(A) Left, the 3.1 Å resolution structure of the nanodisc-reconstituted (E3D1) MscS heptamer, shown in cartoon representation. Each subunit is shown in a different color. Bound lipids are shown as stick representation. The transparent EM density is shown overlapped to the cartoon of the protein. Right, cartoon diagram showing the MscS monomer. Colored grey are regions of the channel resolved in the crystal structure (2OAU), regions newly resolved in the MscS-ND structure are shown in cyan. The putative location of the lipid bilayer is shown as a pair of dashed lines. (B) Residue conservation and inter-subunit interactions stabilizing the anchor domain. Highly conserved sites are shown in blue, variable sites in red. Shown in sticks and balls representation are residues participating in inter-subunit interactions, either polar in nature (E2–Q11) or hydrophobic (van der Waals) packing (L23–L24, Y27–V29). On the left, a cartoon representation of the TM segments in two adjacent subunits, where the top box indicates the location of the diagram on the right.

The N-terminal domain sits atop TM1, forming a returning loop that projects away from the seven-fold symmetry axis and lines the periphery of an N-terminal ring at the periplasmic face of the channel (Figure 1B). Despite the overall lower resolution in the region, we were able to fit a Cα backbone and buildout sidechains using the existing density information and molecular dynamic flexible fitting (Croll, 2018). The N-terminal ring is potentially stabilized by both polar (E2-Q11) and hydrophobic (van der Waals) (L23-L24, Y27-V29) interacting pairs at the inter-subunit interface (Figure 1B). As will be described below, a key finding of the MscS-ND structure is that this newly resolved N-terminal domain is membrane-embedded and seems to dominate its interactions with the outer leaflet of the bilayer. Comparing the sequences of MscS homologs revealed that this particular structural motif is largely present in Enterobacteriales (Figure 1—figure supplement 4). Figure 1B shows the relative rate of evolution of individual residues mapped on the newly resolved domain. Conserved residues (in blue) are predominantly found at the subunit interface, while W16 sits deep in a pocket at the bottom of the N-terminal domain. Indeed, tryptophan residues have been shown to be enriched at the membrane interface and contribute about ~4 kcal/mol as ‘anchors’ of TM segments in membranes (de Jesus and Allen, 2013). Given its membrane placement and location of the conserved W16, we named the MscS N- terminal the ‘anchor’ domain.

Strikingly, other than the upper third of TM1 and anchor domain the structure of MscS-ND is rather similar to the MscS crystal structure in detergent (2OAU) (Steinbacher et al., 2007) (~1.4 Cα RMSD, Figure 1—figure supplement 5). Previous attempts to reconfigure the interactions between TM2 and TM3 to compensate for perceived low inter-helical packing (Anishkin et al., 2008a; Vásquez et al., 2008) are not supported by the present data. In fact, we reason that the present EM structure represents the physiological closed state: The channel is at rest, embedded in a lipid bilayer and ostensibly, in the absence of any applied tension. Furthermore, as MscS-ND includes a 6xHis tag at the N-terminus (with a 10 residue linker), clear density corresponding to the oligomeric assembly of individual Hisx6 tags is observed as a ‘crown’ on top of MscS (Figure 2A left, Figure 3, left and Figure 4—figure supplement 1). This crown leads to the formation of a strong stabilizing force, which precludes N-terminal movement in the closed state. Interestingly, the quality of the density of the anchor domain is significantly degraded in the absence of the N-terminal Hisx6 tag (Figure 3, left-center), presumably due to an increase in local dynamics in the absence of the stabilizing His-tag. Besides being a fortuitous result, the N-terminal Hisx6 serves as a useful probe of MscS conformational changes. 6xHis-MscS-ND appears insensitive to mechanical stimulation in HEK293 cells patch-clamp experiments under high-speed pressure clamp conditions (Figure 4—figure supplement 1C). This functional inhibition is fully relieved upon in situ thrombin treatment (Figure 4—figure supplement 1D), strongly suggesting that the anchor domain must move away from its current position upon channel opening, in a way consistent with the MscS expanded conformation (Lai et al., 2013; Wang et al., 2008).

Figure 2 with 4 supplements see all

Download asset Open asset

A new lipid-protein interface for membrane-embedded MscS.

(A) Close-up of MscS-ND EM density (in Chimera’s ‘solid’ representation). Left, Side view. The location of the bilayer in the nanodisc in indicated by dashed yellow likes (approximately 38 Å in diameter). Density for the putative heptameric histidine tag complex is shown by a dotted arrow. Right, Top view. The yellow circle represents the putative average size of the E3D1 nanodisc (~130 Å) in relation to the density, which points to a partial averaging of the density likely due to MscS lateral mobility. (B) Comparison between the location of the membrane interface in MscS-ND, the FC14 crystal structure (2OAU) and the ‘Cryst’ deletion construct. Black dashed lines depict the limits of the lipid bilayer based on the nanodisc EM density. Left, EM density for the protein (red) and the nanodisc (grey) for MscS-ND, the black ovals highlight the fact that the prominent cavities formed between the TM1-TM2 hairpin and TM3 are fully located outside the membrane. Center, relative positioning of 2OAU based on a rigid fit of the structure onto MscS-ND EM density. The gray rectangle in the background represents the previous consensus membrane location. Right, the low-resolution cryoEM structure of MscS ∆2–27 (‘Cryst’,~20 Å) shows an overall architecture for the nanodisc-embedded channel. In spite of the N-terminal deletion, the nanodisc is located at the same position as in MscS-ND. (C) Probing the energetics of the membrane interface. A Potential of mean force (PMF) calculation was carried out by relocating a lipid bilayer from a coordinate origin (0 Å) predicted by the CHARM-GUI server (Jo et al., 2008) and moved up to 16 Å (the thickness of a lipid monolayer) along the Z-axis coordinate (see Figure 2—figure supplement 2). Left, free energy as a function of Z-axis displacement. A global minima was found at ~25 Å (~8 Å above the prediction) and the free energy increases exponentially beyond this point. The energy minima coincides with the location if the interface as defined by the EM density of MscS-ND. Right, evolution of MD simulation starting at three membrane interface locations: predicted by CHARM-GUI (0 Å, red trace), at the cryo-EM density (+ 8 Å, black trace) and a further +16 Å (Higher placement, blue trace). After ~60 ns simulation all membrane interfaces converge to that defined by the cryo-EM density.

Figure 3 with 3 supplements see all

Download asset Open asset

Membrane interface location in nanodisc-reconstituted MscS is independent of lipid composition.

CryoEM structures for three additional lipid reconstitution/detergent conditions show a common membrane interface. Four independently determined structures are shown: Left, MscS-ND (in POPC:POPG 4:1) determined at 3.1 Å. This structure includes the N-terminal 6xHis (red rectangle). Center left, MscS-ND (in POPC:POPG 4:1) after thrombin proteolysis of the N-terminal 6xHis, determined at 4.1 Å. Center right, MscS-ND (in *E. coli* lipids) after thrombin proteolysis of the N-terminal 6xHis, determined at ~10 Å. Right, DDM-solubilized MscS after thrombin proteolysis of the N-terminal 6xHis, determined at 3.4 Å. In all cases, EM density is shown as Chimera’s ‘solid’ representation with the protein depicted in ribbon representation (cyan). The calculated location of the membrane is shown as a yellow slab with black dash lines, while the previous membrane interface consensus location is represented by the grey dashed lines. The red arror points to a partially unfolded region of the N-terminal loop region.

To evaluate the role of the anchor domain on MscS function, we designed a construct lacking the first 26 residues of the channel: Δ2–26 MscS (‘Cryst’), physically recapitulating the resolved regions crystal structure (2OAU) model, which has been the basis of many past experimental and computational insights (Cox et al., 2018; Naismith and Booth, 2012; Perozo, 2006; Sukharev and Corey, 2004). This construct displays a severe loss of function (LOF) phenotype and is unable to elicit any mechanically activated currents in either HEK293 cells patch clamp experiments (Figure 4A) or in downshock assays (Figure 4B). To further evaluate the functional role of the anchor domain, we carried out an alanine scan at residues 2–30 for in vivo analysis of MS channel activity (Figure 4B). The effect of alanine substitutions was dramatic. About 70% of the mutants displayed LOF phenotypes, 8 of those severe (E2A, V6A, S9A, G12A, N20A, Q21A, L24A, L25A). Some of these severe LOF mutations might be potential GOF mutations due to their survival being lower than the negative control of no MscS. A severe MscS GOF mutant will have dramatic effects on the survival of any cell, but final demonstration requires electrophysiological data under pressure clamp conditions. Remarkably, mapping the positions of the alanine LOF mutants (Figure 4C) show that they are located almost exclusively at the membrane or inter-subunit contact interfaces (Figure 4C, dotted line). These results not only highlight the importance of subunit-subunit interactions for MscS function, but point to a role of the anchor domain in MscS mechanotransduction. In fact, an earlier cysteine scan of the same region (Figure 4—figure supplement 2) (Vásquez et al., 2008), shows that almost a third of the mutants display a noticeable LOF phenotype (V7C, N11C, A13C, V18C, and L25C), two were severe (V6C and A19C) and most map to the subunit interface.

Figure 4 with 2 supplements see all

Download asset Open asset

Functional significance of the anchor domain.

(A) Functional consequences of deleting the anchor domain (MscS ∆1–27). High-speed pressure clamp recordings of co-expressed MscS and MscL in HEK296 cells under voltage clamp conditions. A family of macroscopic currents elicited by pressure ramps are shown for co-expressed wt-MscL and wt-MscS (left traces), wt-MscL and Cryst (center traces) or Cryst alone (right traces). (B) Osmotic downshock assays of N-terminal alanine scan mutants. Box plots are shown for nine independent experiments, where the central bar represents the median, the box 25%−75% quartiles and the individual data are depicted as black dots. Data were normalized to the wild-type behavior and the dotted lines represent the mean survival for the negative control (red, empty vector) and wt-MscS (green). Asterisks indicate residues located at the subunit interface. (C) Functional complementation of alanine scan mutants mapped onto the anchor domain structure. Left, side view of the anchor domain monomer. Right, top view of the anchor domain as a heptamer with data mapped on two adjacent subunits. The residues are colored by surivial relative to Wt (LOF Severity index) where 0 corresponds to no survival compared and one is equal or greater survival than wild-type. Black residues are wild-type alanines (not evaluated). The dashed line indicates the location of the subunit interface.

Redefining MscS interaction with the lipid bilayer

Based on the MscS crystal structure (2AOU) (Bass et al., 2002; Steinbacher et al., 2007), a de facto consensus had been reached where the bilayer interacted with the channel at the level of the resolved regions of the TM segments (see Booth and Blount, 2012; Cox et al., 2018; Edwards et al., 2004; Perozo, 2006; Pliotas and Naismith, 2017; Zhang et al., 2016). In fact, early MD simulations of membrane-embedded MscS showed unexpected instabilities in the putative transmembrane segments, even under different force fields (Anishkin et al., 2008a; Sotomayor and Schulten, 2004; Spronk et al., 2006). We note, however, that a partial shift in membrane location had been previously suggested after modeling the N-terminus of the channel (Anishkin et al., 2008a). The present MscS-ND EM density now redefines the nature and extent of MscS interactions with the lipid bilayer. Remarkably, our data show that the disc-like region corresponding to the nanodisc lipid bilayer (Figure 2A B left) is upwards shifted ~14 Å when compared to the previous consensus placement (Figure 2B, center). This is also highlighted on side views from 2D classes (Figure 1—figure supplement 2).

Given this location, the membrane not only interacts closely with the N terminal domain (Figures 1A,B,2A) but a large portion of the cytoplasmic end of the TM1-TM2 hairpin and most of TM3a (including the location of the vapor lock) now lies outside of the membrane (and displays considerable degree of conformational heterogeneity (Figure 1—figure supplement 2, Figure 3—figure supplement 1 and Figure 3—figure supplement 2). This arrangement is fully compatible with the distribution of MscS surface charged residues (Figure 2—figure supplement 1A). Furthermore, modern algorithmic predictions of membrane placement (Lomize et al., 2012; Newport et al., 2019) are in agreement with the present membrane location when the MscS-ND structure is evaluated (Figure 2—figure supplement 1B,C).

As a quantitative evaluation of the bilayer interaction energetics, we carried out a potential of mean force (PMF) calculation where MscS-ND was moved along the Z-axis in relation to a fixed bilayer (Figure 2—figure supplement 2). Calculation of the free energy as a function of linear displacement reveals a clear energy minima centered precisely at the location experimentally defined by the EM density (Figure 2C, left). Indeed, equilibration of three bilayer positions (upwards and downwards of MscS-ND placement) all converge to the EM density position within 60 ns (Figure 2C, right). When the free energy change based on our continuum calculation is compared to those obtained from PMF calculations, the contribution of hydrophobic mismatch dominates that of membrane curvature (Figure 2—figure supplement 3). Although the hydrophobic mismatch in our initial MD-PMF simulations was not systematically changed (the reaction coordinate was set to move the protein up/down across the bilayer thickness), the minima in the free energy obtained from our MD simulation is matched with that obtained from our mean-field calculations (at the reaction coordinate for the PMF calculations has been defined as the distance between the Z coordinate of the center mass of phosphate molecules of the lipid bilayer and the Z coordinate of center mass of the pore-forming helices of MscS (i.e. residue 105 to 115), Z distance ~25 Å). Hence, our PMF calculations confirm the most energetically favorable position of MscS in the bilayer (at Z distance ~25 Å), otherwise the hydrophobic mismatch between the membrane and protein would have excess energetic costs (Figure 2—figure supplement 3C). These results must be evaluated by taking into formal consideration potential issues of force/tension bias once MscS transitions from one conformation to another within a nanodisc. However, we believe these to be relatively minor.

To expand on this result, we pursued additional MscS EM structures under various lipid/detergent conditions. Two specific questions were addressed: Is the location of the bilayer related to its lipid composition? What is the EM structure of MscS in DDM and what is its micelle placement? The structure of MscS-ND was determined in nanodiscs containing PC:PG 4:1, yet E. coli membranes are composed mostly of PE, PG, and cardiolipin (Raetz and Dowhan, 1990). Figure 3 shows that the same bilayer placement is observed in the EM densities for two PC:PG MscS-ND structures (with and without N-terminal 6xHis, Figure 3—figure supplements 1,3) and in nanodiscs containing E. coli polar lipids (at 10 Å resolution), suggesting that MscS-bilayer placement is not affected by alternative lipid compositions. The EM density in DDM (at about 3.4 Å, Figure 3—figure supplements 2,3) appears to show a similar placement for the detergent micelle. However, the structure of the N-terminal domain has partially unraveled (Figure 3, red arrow), suggesting a rationale for why this region is unresolved under crystallographic conditions or in nanodiscs in the absence of the N-terminal 6xHis (Rasmussen et al., 2019).

Lipids bound to dynamic regions of MscS

At the present resolution, the MscS-ND maps provide an excellent template to evaluate the nature and extent of the interactions between the channel and the lipid bilayer. Notably, we find lipid-like densities both, in a newly defined cavity between subunits and at the center of the permeation pathway (Figure 5A). Seven clearly defined phospholipids appear to ‘hook’ the top of each of the TM2-TM3 loops, a region that putatively displays large conformational rearrangements during gating (Lai et al., 2013; Wang et al., 2008). The hook lipids are firmly embedded by threading their head group through an inter-subunit opening formed by the extended TM1 and the top of the TM2-TM3 hairpin (Figure 5—figure supplement 1A–B) and facing the permeation path. At the present resolution, the nature of the hook lipids head group was not defined, but was ultimately modeled as PC due to the nanodisc composition. But given the E. coli membrane composition, it is likely that the hook lipids are PE or PG.

Figure 5 with 3 supplements see all

Download asset Open asset

Bound lipid at the inner gate and the permeation pathway.

(A) Side (left) and top (right) views of EM density (transparent red surface) associated with putative lipid molecules bound to MscS-ND (shown in white cartoon representation). A ‘hook’ phospholipid is cradled at the subunit interface atop the TM2-TM3 hairpin, while seven individual acyl chains line the permeation pathway along TM3, above the narrowest portion of the gate. (B) Contact map and coordination of the hook lipid. Left, TM helices from two adjacent subunits (red and blue) are shown. Key interactions are highlighted for R88 (in the red subunit) with the head group nitrogen (PC or PE) and Y27 (in the blue subunit) with the phosphate group. Residues within van der Waals distances are shown in red. Right, a cartoon representation of the contact/coordination map. (C) Same as (B), but with the pore lipid acyl chains.

Given that hook lipids bind to a conformationally active region of MscS, we suggest they might participate in the transduction of bilayer forces that influence the conformation of the MscS gate (TM3a). Indeed, binding of the hook lipids is stabilized by R88 in one subunit and Y27 from the neighboring subunit (Figure 5B). Both residues are known to generate LOF phenotypes when mutated (Figure 4D; Rasmussen et al., 2015). Additional residues along TM1 and the TM2-TM3 linker participate via hydrophobic contacts (Figure 5B). A cluster of seven linear densities is also observed lining the patch of hydrophobic residues immediately above the narrowest region of the permeation pathway (L105) (Figure 5—figure supplement 1C–E). Although there is no clear indication of EM density associated with headgroups (whether it is PE or PG), we hypothesize these are either acyl chains from a fairly mobile (and not fully resolved) phospholipid or perhaps bound fatty acids trapped along the permeation pathway. Modeled as hexadecanes, the pore lipids are stabilized (likely weakly) via hydrophobic interactions along TM3 (Figure 5C). Though not currently resolved, a potential pore lipid head group might thread through an inter-subunit gap between the G104s in TM3a (Figure 5—figure supplement 1F). Interestingly, a tryptophan scan encompassing TM3a and the TM2-TM3 linker (Rasmussen et al., 2015) shows a remarkable correlation with LOF residues aligning at the putative hook lipid binding pocket and the pore lipids interaction surface (Figure 5—figure supplement 2B).

We believe the pore and hook lipids were carried over from the cell membrane during purification and reconstitution, as we find evidence of similarly placed density in our DDM MscS map (Figure 5—figure supplement 3A). Furthermore, an evaluation of the electron density map for the 2OAU crystal structure (Steinbacher et al., 2007) displays unassigned density that precisely corresponds to the location of the phosphate group in the hook lipid (Figure 5—figure supplement 3B), implying unresolved lipid-bound density in MscS crystals. However, the high resolution 5AJI expanded crystal structure does not show either lipids, which might be due to the lipid being solubilized away in the detergent environment when not protected by the closed conformational N-terminal loop. That MscS engages in close interactions with membrane lipids has in fact been suggested earlier based on non-denaturing mass spectroscopy (MS) (Pliotas et al., 2015). And while MS and chemical extraction helped identify at least five phospholipids per MscS (mostly PE) the predicted placement and interaction with MscS (Pliotas et al., 2015) appears to be incompatible with those observed in the present EM densities (Figure 5). Our EM density does not show explicit lipid density in the TM2-TM3 pockets. However, lipids bound to the TM pockets have been reported recently in ND-reconstituted MscS (Rasmussen et al., 2019), but whether or not these lipids play a functional role remains to be established.

Previous studies of closed state water permeation in MscS (Anishkin and Sukharev, 2004; Spronk et al., 2006) have shown that in spite of a wide (~7 Å) diameter at rest, the hydrophobic characteristics of the pore lead to a functional occlusion by a 'vapor lock' mechanism. Assuming that the pore lipids do occlude the permeation path in the closed state we set out to evaluate their influence on water dynamics along the permeation pathway. We consider three conditions: closed MscS pore with no associated lipids, with only the hook lipid, or with both hook and pore lipids (Figure 6A). As reported (Anishkin and Sukharev, 2004), the permeation path in MscS with no bound lipids fluctuates between a vapor locked state and a filled state where water is able to permeate. Figure 6A shows that during a 10 ns MD run the running averages for water permeation in the absence of bound lipid fluctuates around two water molecules in a 3 × 2 Å cylinder centered at L105. Inclusion of the hook lipid atop the TM2-TM3 hairpin reduces the running average about one water molecule at a time. However, the addition of both hook and pore lipids thoroughly eliminates any water permeability. This is illustrated from side views and cross sections of the pore in Figure 6B. The substantial effect of the pore lipid on water dynamics further suggests that under physiological conditions the pore lipid might be able to act as a low dielectric blocker, suggesting that the transition to the open conformation in MscS could be accompanied by a reduction in the occupancy of the pore lipid along the permeation pathway.

Figure 6

Download asset Open asset

Role of bound lipids on the permeation pathway.

(A) MD simulation of water permeation during 10 ns of equilibration under three lipid occupancy conditions: In the absence of bound lipids (black trace), with the hook lipid-bound (red trace) and with both hook and pore lipids bound (blue trace). Waters were counted in a 3 × 2 Å cylinder that includes the L105 residue (shown in the insets). (B) Close-up of the MscS permeation path under the three conditions described in (A). In each case the top image shows side views with water molecules in stick representation and the hook and bound lipids as VDW spheres. The cyan bar represents the slab thickness of the cross section in the bottom image (rotated 90 degrees).

Discussion

On the mechanism of force-from lipid gating in MscS

The new MscS structural features, membrane footprint and bound lipids all have important mechanistic consequences regarding MscS force transduction. For one, TM3b can no longer be considered an interfacial helix (Bavi et al., 2016b) as it is located almost 15 Å away from the membrane/water interface (Figures 1A,2B, left). This is confirmed by our five structures presented here and elsewhere (Rasmussen et al., 2019). The proposed gating mechanisms where lipids act as ligands are based on the assumption that the TM2/TM3a cavity is open and accessible to the lipid bilayer (Pliotas et al., 2015) are unlikely due to the location of the TM2/TM3a cavity (or TM pocket) in relation to the membrane annulus around MscS (Figure 2B) seems incompatible with a proposed phospholipid exchange between bilayer and TM pockets. Accordingly, the suggestion that lateral tension would 'pull' on lipid acyl chains located in TM pockets appears unrealistic. Although our data only shows marginal additional density within the TM pockets, a recent structure of MscS in nanodiscs does find density that has been interpreted as lipidic (Rasmussen et al., 2019). Because of the volume change observed in the TM pockets of the expanded MscS conformation (Lai et al., 2013; Pliotas et al., 2015) it is expected that this lipid density should change during gating. The relative contribution of all bound lipids to force transduction in MscS will require further investigation.

Our structure in nanodiscs reveals much of the anchor domain locked in the closed state. We suggest that the anchor domain must play a significant role in the mechanism of mechanotransduction in MscS. This is based on three primary observations. First, deletion of the anchor domain renders MscS unresponsive to tension, although it does not preclude its folding and oligomeric assembly (Figures 2A,4B); second, limiting the conformational flexibility of the anchor domain by the N-terminal His-x6 ‘bundle’ leads to non-functional MscS, and deleting the His-x6 ‘bundle’ via proteolysis restores activity to WT level (Figure 4—figure supplement 1B-D); finally, we show that the anchor domain is unusually sensitive to mutagenesis with up to 70% of its amino acids leading to LOF (or severe GOF) phenotypes upon mutation to ALA (Figure 4B). This is a much more severe effect than previous surface mutagenesis efforts on the basis of the ‘incomplete’ crystal structure. The significant functional and structural role of the anchor domain must now be considered in any force transduction model.

Earlier spectroscopic work (Vasquez et al., 2008) provides excellent clues regarding the types of conformational changes expected at the periplasmic side of the channel during its transition towards the open state. Changes in the NiEdda (water) accessibility between closed (at rest, in liposomes) and open MscS conformations (after LysoPC activation) show a massive reduction in water accessibility for the majority of the anchor domain (Figure 7A) when compared to the values at rest (Figure 7—figure supplement 1). This suggests that in the open state the anchor domain transitions to a deeper location in the outer leaflet of the bilayer (most likely as an extension of TM1), while it tilts and moves away from the permeation pathway (as seen in the MscS expanded conformations, Lai et al., 2013; Wang et al., 2008).

Figure 7 with 1 supplement see all

Download asset Open asset

Mechanisms of Force-from-Lipid gating in MscS.

(A) Extent and direction of environmental parameter changes upon MscS opening. Left, NiEdda accessibility (ΠNiEdda) mapped on MscS-ND transmembrane segments. Right, Changes in NiEdda accessibility (ΔΠNiEdda) mapped on MscS-ND transmembrane segments. Data from Vasquez et al. (2008). Note, however that the EPR data were obtained on an unconstrained (though lipid embedded) N-terminal domain and should be treated as a general guide. A decrease in NiEdda accessibility corresponds to a decrease in accessibility to the aqueous milieu. (B) A lipid-centric mechanism of force transduction in MscS. In the nominal absence of external forces, MscS populates the resting, closed state represented by the MscS-ND structure (Figure 1). The non-conductive nature of the closed conformation is determined by a narrow cuff of hydrophobic residues around L105 (black circles) and above that, a plug of acyl chains from the pore lipid. The inter-subunit hook lipid links the top of the TM2-TM3a hairpin and the hydrophobic core of the bilayer. Applying forces in the plane of the bilayer lead to two gating scenarios, depending on whether the hook lipid stays bound to the open channel or not. In the first case (top) the expansion of TM1 in the periphery of MscS is allosterically communicated to the TM2-TM3 hairpin through the hook lipid, leading to the expansion of TM3a and release of the pore lipid. In the second option (bottom), expansion of TM1 destabilizes the binding of the hook lipid which exchanges with the bulk of the bilayer. Acting as a ligand, the release of the hook lipid triggers a conformational rearrangement in TM3a, with subsequent release of the pore lipid, leading to conduction.

We find that at least two general gating models are consistent with both the data provided by the present MscS-ND structure and the conformational changes observed crystallographically (Figure 7B). In the first one (Open 1), the hook lipid remains bound to its pocket, allosterically coupling intramembrane forces with anchor domain rearrangements, TM1-TM2 hairpin reorientation and the expansion of the TM3a inner bundle. The second possibility (Open 2), would be reminiscent of the mechanism proposed to explain mechanosensitivity in the K2P channels TRAAK (Brohawn, 2015). As such, membrane stretch would trigger rearrangements at the anchor domain and TM1-TM2 hairpin, destabilizing the hook lipid pocket and leading to diffusion of the hook lipid into the bilayer. In turn, the release of the hook lipid triggers expansion of the TM3a inner bundle, opening the channel. Of course, in both gating models, channel opening ought to be accompanied by a release of the pore lipid, unblocking of the permeation path. Further research is needed to elucidate potential mechanisms of pore lipid dynamics.

The present results suggest that, for MscS, the energetic differences derived from tension changes in the plane of the lipid bilayer should be evaluated in the context of key lipids bound to mechanistically important regions of the channel (TM2 and TM3a). Functionally, MscS behaves as a lipo-protein complex, where the hook lipid may help transduce bilayer forces and the pore lipid is poised to influence ion and water fluxes at rest. These interactions, together with the revised location of the lipid-protein interface must be accounted for by mechanotransduction models which strong allosteric coupling between TM segments and the lipid-protein interface.

Materials and methods

Key resources table

Reagent type (species) or resource	Designation	Source or reference	Identifiers	Additional information
Gene (E. coli		Addgene# 7855		6x N-Terminal His-Tag
Gene (E. coli)		Addgene# 20066
Strain, strain background (E. coli)	MJF465	Ian Booth and Samantha Miller, University of Aberdeen
Other	POPC	Avanti Polar Lipids	850457C
Other	POPG	Avanti Polar Lipids	840457C
Other	E. coli Polar Lipids	Avanti Polar Lipids	100600C
Other	Bio-Beads SM-2 Resin	Bio Rad	1523920
Other	Quantifoil 2/2 Mesh 200	Quantifoil
Other	Quantifoil 1.2/1.3 Mesh 300	Quantifoil
Other	Octyl Maltoside, Fluorinated	Anatrace	O310F
Other	n-Dodecyl-β-D-Maltopyranoside	Anatrace	D310A
Other	Fos-Choline-14	Anatrace	F312S
Strain, strain background (E. coli)	Rosetta 2	Millipore Sigma	71400-3
Other	Thrombin	MP Biomedicals	154163	Bovine

MscS expression purification

Request a detailed protocol

Full-length E. coli MscS was expressed and purified as previously described (Vásquez et al., 2007). In brief, MscS was sub-cloned into pET28a containing a His₆ tag and a thrombin cleavage site on the N-termini. Rosetta 2 (Millipore Sigma) E. coli cells were transformed with MscS-pET28a vector and grown overnight in the presences of kanamycin and chloramphenicol. The cells were diluted 1:100 in LB medium and grown at 37°C to an OD₆₀₀ of 0.8-1.0. Before induction, the cell culture was supplemented to a final concentration of 0.4% glycerol and allowed to cool to 26°C, and protein expression was induced with 0.8mM IPTG. The cells were grown for 4h at 26°C and were harvested, and either were frozen at -80°C for later use or immediately resuspended in PBS pH 7.4 (Sigma), 10% glycerol, protease inhibitors, and homogenized (high-pressure homogenizer, EmulsiFlex-C3). The membranes were isolated via centrifugation at 100,000g for 30 min, and the pellet was resuspended in PBS and 10% glycerol. Solubilization was carried out in 1% Fos-Choline (Anatrace) 14 for 4-16h at 4°C. This resuspension was spun down at 100,000g for 30 min, and the supernatant supplemented with a final concentration of 5mM imidazole (Fisher) was incubated with cobalt resin(Clonetech) for 2-4h at 4 °C. The resin was washed with 20-bed volumes of 1 mM DDM(Anatrace), 10mM imidazole and 10% glycerol in PBS buffer. MscS was eluted in 1 mM DDM, 300mM imidazole, and 10% glycerol in PBS buffer. Unless explicitly stated MscS His, thrombin was added to cleave the his tag and incubated overnight. The final purification step was to run the protein on a Superdex 200 Increase 10/30 column (GE Healthcare) with 1 mM DDM and PBS buffer. The removal of glycerol is critical for EM grid preparation. The typical yield of MscS is about 5-8mg per liter of E. coli. For the MscS-Cryst construct, residues 2-26 residues were removed and subcloned into pQE70 and grown in MJF465 E. coli cells (to avoid co-assembly with chromosomal wt-MscS), a gift from Ian Booth (Levina et al., 1999). Typical yield of MscS-Cryst is less than 0.1mg per liter of MJF465 E. coli. Otherwise, the purification steps were the same. The MscS structure solved in DDM was solubilized in 1% DDM instead of Fos-Choline 14.

MscS nanodisc preparation

Request a detailed protocol

MscS nanodiscs (ND) were prepared following previously described protocol (Ritchie et al., 2009). Several variants of ND scaffold proteins were tested, and Msp1 E3D1 was deemed the most homogenous by size exclusion. The molar ratio of MscS:MSP1 E3D1:Lipids was 7:10:650, respectively, after extensive optimizations. Each lipid solution of mixed micelles contained 30-50mM DDM with a final lipid concentration of 10-17mM. The compositions of the mixed micelles were either (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) POPC and (1-palmitoyl-2-oleoylglycero-3-phosphoglycerol) POPG (4:1) or E. coli Polar Lipids (EPL). Nanodiscs were made by adding mix micelles to protein for 20 minutes on ice. MSP was added to the solution and incubated on ice for 5 minutes. The reconstitution mixture was incubated in activated bio beads (Biorad) overnight at 4°C. The detergent free mixture was run on a Superdex 200 Increase 10/30 column to separate the empty ND peak. The MscS ND peak was concentrated to ~2mg/ml and stored at 4°C.

EM data collection and structure determination

Request a detailed protocol

To help with orientation preferences and ice thickness, MscS ND was supplemented with Octyl Maltoside, Fluorinated (Anatrace) to a final concentration of 0.01%. MscS and was applied twice, with a filter paper blotting between each application, onto Mesh 200 2/1 or Mesh 300 1.2/1.3 Quantifoli holey carbon grids and flash frozen in a Vitrobot (Thermofisher) set at 3 seconds with a force of 3 with 100% humidity at 22°C. MscS His ND POPC:POPG, MscS No His ND POPC:POPG, and MscS DDM were collected on a Titan Krios with a K2 detector in counting mode with a GIF energy filter using Latitude S (Thermofisher). Movies were acquired at 1e^-/A² per frame for 50 frames. MscS ND EPL was collected on Titan Krios with a Falcon 3 detector in counting mode. MscS-Cryst ND POPC:POPG was collected on Talos Artica with a Falcon 3 detector in counting mode. Movies were acquired at 1e^-/A² per frame for 50 frames. Motion correction was performed using Motioncor2 (Zheng et al., 2017), and K2 movies were binned by 2. CTF estimation was done using CTFFIND4.1 (Rohou and Grigorieff, 2015). Initial particle picking was done using Eman’s (Tang et al., 2007) neural net particle picker or Relion’s built-in reference based auto picker and the coordinates were fed into Relion (Scheres, 2012) for particle extraction. Subsequent structure determination steps were done in Relion. An initial 2D refinement was done to remove non-particles and poor-quality classes, which were fed into 3D classification. 3D classification was performed using the MscS crystal structure as an initial model. After a subset of particles were identified for the final refinement, the particles underwent per particle CTF refinement followed by Bayesian polishing. The final 3D reconstruction used the classes with both top and side views and refined using a tight mask excluding the membrane and his-tag (when necessary) and C7 symmetry. Model building was based on the MscS crystal structure (PDBID: 2OAU) and used coot to build the remaining TM1, N-terminal domain, and the hook and pore lipids. EM density maps used in subsequent steps were not were not postprocessed or sharpened. While postprocessing and sharpening did improve the density for the most part, the N-terminal domain became much noiser. The initially built model was iteratively refined using Coot (Emsley et al., 2010), Chimera (Pettersen et al., 2004), MDFF (McGreevy et al., 2014) using VMD (Humphrey et al., 1996) and NAMD (Phillips et al., 2005) or ChimeraX (Goddard et al., 2018) with the ISOLDE (Croll, 2018) plugin, Arp/Warp (Langer et al., 2008), and Phenix’s (Adams et al., 2010) real space refine.

Downshock assay

Request a detailed protocol

Downshock assays were performed from a modified protocol from what was previously described (Batiza et al., 2002; Vásquez et al., 2007). MJF465 cells transformed with various MscS mutants in pEQ70 were grown modified Luria-Bertani (LB) medium with 500mM NaCl and 100µg/ml ampicillin(Fisher), 50 µg/ml kanamycin(Fisher), and 25 µg/ml chloramphenicol(Fisher) at 37°C to an OD₆₀₀ of 0.6. The cells were cooled to room temperature and induced with 1 mM IPTG (Fisher) for 2 hours at 25°C. The OD₆₀₀ was measured and downshocks were performed by diluting cells 1:50 into a modified LB medium at 50mM NaCl and 1:100 was plated on standard LB agar plates overnight at 37°C. The colonies on the LB agar plates were imaged and counted and normalized by the OD₆₀₀ readings. Additionally, to assess the expression of each mutant, a western blot was performed. The western blot of each MscS expressing mutant was from a pellet from the downshock experiment and resuspended in PBS and SDS to a final 1% solution. The lysate was then sonicated, ran on a 4-20% SDS-PAGE gel (Biorad), transferred to PVDF and probed with the Penta-HIS(Qiagen) primary and anti-mouse conjugated to Alexa 488 secondary.

Phylogeny analyses

Request a detailed protocol

Enterobacteriales and Vibrionales MscS protein sequences were extracted from the complete proteomes in the NCBI Assembly database. From each proteome, only one protein showing the highest BLAST bit score (Camacho et al., 2009) to the E. coli MscS protein query was extracted. Sequences were aligned using MUSCLE (v.3.5) (Edgar, 2004), and the ML phylogeny was inferred using RAxML (v.8.2.11) (Stamatakis, 2014) (best-fit model of evolution: LG+G+X). The schematic representation of the phylogeny was generated using iTOL (Letunic and Bork, 2019). The relative rate of evolution for each site was inferred from an alignment of Enterobacteriales MscS proteins using RAxML (v.8.2.11) (Stamatakis, 2014).The rate of evolution was mapped on protein structure using Chimera (Pettersen et al., 2004). The sequence logo was generated from an alignment of Enterobacteriales MscS proteins using WebLogo 3 (Crooks et al., 2004).

Proteoliposome preparation and patch clamp electrophysiology

Request a detailed protocol

Proto-liposomes were prepared using Dehydration Rehydration (D/R) method as fully described in previous studies (Nomura et al., 2015). Briefly, Avanti soybean lipid dissolved in chloroform were dried with nitrogen flow to create a thin lipid film on a glass tube. The film was suspended and vortexed with D/R buffer (200 mM KCl, 5 mM HEPES, adjusted to pH 7.2 with KOH) and was subjected to 15 min of sonication. MscS was added to the lipid at a protein to lipid ratio of 1:200 (w/w) and incubated at 4°C for 1 h. to remove detergent, Biobeads (BioRad,Hercules,CA, USA) were added and incubated at 4°C overnight (minimum 3 h). The proteoliposomes were collected by ultracentrifugation and resuspended in 50 ml of D/R buffer. Small aliquots were spotted onto the glass cover slips and dehydrated overnight under vacuum conditions and at 4°C. The dried proteoliposomes were then rehydrated with 20-25 µl D/R buffer. After 6 h incubation at 4°C, they are ready for electrophysiological experimentation. The channel activity was examined in excised (inside-out) configuration. An isotonic recording solution were used in the bath and pipette (200 mM KCl, 40 mM MgCl2, and 5 mM HEPES adjusted to pH 7.2 with KOH). Borosilicate glass pipettes were pulled using Sutter micropipette puller (P-1000, Flaming/Brown). The resistance of the capillary pipettes was from 2 to 4 mOhm. The current was amplified with an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA, USA), filtered at 2 kHz and the data acquired at 5 kHz with a Digidata 1322A (Axon instruments) interface using pCLAMP 10 acquisition software (Molecular Devices). Negative pressure was applied using High Speed Pressure Clamp-1 apparatus (ALA Scientific Instruments, Farmingdale, NY, USA).

All-atom molecular dynamics (MD) simulation

Request a detailed protocol

CHARMM GUI was used to embed MscS structure into a POPC:POPG (4:1) bilayer mix (to mimic our nanodisc lipid composition) (Jo et al., 2008). Different computational models have been generated as listed in Supplementary file 1 The equilibration steps were performed similarly to our previous MD simulation of EcMscL (Bavi et al., 2016a). TIP3P water molecule was used to solvate the system. The lipid and water molecules in close proximity to the channel (<0.5 Å and <0.5 Å respectively) were removed first. The system was ionized by 200 mM KCl. Short lipid tail randomization was done for 20 ps. Lipid and water were packed around the protein for 1 ns, while the Cα atoms in the protein were fixed. The restraint on the protein was released, and the equilibration run was performed for 60 ns. In order to simulate our system in an NPT ensemble, a modified Nosé-Hoover Langevin piston pressure control provided in NAMD (Phillips et al., 2005) was applied to control fluctuations in the barostat around the constant pressure of 1 atm, whereas the temperature was controlled at 298 K via Langevin dynamics. The Particle-Mesh Ewald (PME) method was used in all simulations to compute electrostatic interactions beyond a real-space cut-off of 1.2 nm using a Fourier grid spacing of 0.1 nm. van der Waals interactions were smoothly switched off at 8−10 Å. Periodic boundary conditions were applied in all three directions. The CHARMM c36 Force field was used for all MD calculations (Brooks et al., 2009). We used VMD and Chimera for visualization and illustration of our simulation results (Humphrey et al., 1996; Pettersen et al., 2004). We continued the equilibration until the RMSD values of the protein backbone over the equilibration time was plateaued (Figure 2—figure supplement 2B).

Pore hydration calculations

Request a detailed protocol

Water molecules were counted for the last 10 ns of equilibration for each model. A custom TCL script was used (Bavi, 2019 at https://github.com/Perozo-lab/PMF; copy archived at https://github.com/elifesciences-publications/PMF), which counts the number of hydrating water molecules that pass the central hydrophobic pore (i.e. L105) over the simulation period.

Potential mean force (PMF) energy calculations

Request a detailed protocol

PMF calculations were performed using a similar approach to previous studies (Corry and Thomas, 2012; Li et al., 2018; Shen and Guo, 2012). Free-energy values for different bilayer-to-protein position along the Z-axis (bilayer thickness) (Figure 2—figure supplement 2C) were calculated by umbrella sampling simulations (Egwolf and Roux, 2010; Torrie and Valleau, 1977). Three different models were built to explore the most energetically favorable position of the lipid bilayer with respect to MscS along the Z-axis. The first model is Model 4 (Supplementary file 1 ) where the MscS position in the bilayer has been determined based on the CRARMM-GUI potentials (Jo et al., 2008). We first defined our reaction coordinates as the distance between the Z coordinate of center mass of phosphate molecules of the lipid bilayer and Z coordinate of center mass of the pore-forming helices of MscS (i.e. TM3a, Figure 2—figure supplement 2D). The predicted model was shifted upward (in the +Z direction, i.e., along with the bilayer thickness) by ~ 8 Å to match the bilayer position based on our Cryo-EM structure determined in nanodisc (Model 3, Figure 4C). The third model was built such that it was 16 Å higher than the position predicted by CHARMM-GUI (Model 5, Figure 4C). The starting configurations for the umbrella sampling simulations were taken from the MD trajectory of Model 3, Model 4 and Model 5 (varying in 1 Å steps from reaction coordinates of 16 Å to 31 Å) (Figure 2—figure supplement 2B).

As there is sufficient overlap between the sampling windows (Figure 2—figure supplement 2D), the number of windows is, therefore, enough to have an acceptable evaluation of the free energy landscape. A biasing harmonic potential force of 5 kcal/mol/Å² was used to constrain the position of the bilayer with respect to the protein. Therefore, 16 simulations were performed, where each simulation consisted of 1ns equilibration (no harmonic force) followed by 10 ns of production run (in the presence of harmonic force). Data were unbiased and combined using the weighed histogram analysis method using WHAM package (Grossfield, 2010). The minimum-energy path connecting the free-energy minima with respect to the reaction coordinate (bilayer-to-protein position) was shown in Figure 4C. We are aware there are differences between the ND membrane crossectional area (~130Å diameter) vs. the membrane in MD (160Å x-y box), however we believe MscS is not restricted in either case as even at its widest predicted TM point, MscS is less than 75Å in diameter. Considering the crossectional area of a POPC lipid is ~67Å² (Bayburt and Sligar, 2010) this maintains lipids several layers deep before hitting the edge (Figure 4A).

Continuum Mean-Field calculations of the free energy change

Request a detailed protocol

Hydrophobic length of the protein, $d_{p}$ , was determined based on the average Z distance between the center of mass (COM) of residue W16 to I48 (Figure 2—figure supplement 3A). The hydrophobic length of lipid, $d_{l}$ , was calculated as the average Z distance between the COM of C1 atoms in the upper leaflet and the lower leaflet. Given the membrane thickness changes drastically from around the channel towards the boundary of our simulation box (Figure 2—figure supplement 3B left), we calculated the $d_{l}$ for lipids that are within r = 8Å of the protein (Figure 2—figure supplement 3B right). The hydrophobic mismatch length, $d_{H}$ , then can be estimated as:

d_{H} = d_{p} - d_{l}

The maximum radius of curvature (R) and curvature (C) were estimated by measuring the l and h values as following (Figure 2—figure supplement 3C),

R = \frac{h}{2} + \frac{l^{2}}{8 h} & C = \frac{1}{R}

Here we investigate whether the binomial curve seen in our free energy diagram (Figure 2—figure supplements 2,3D, E), is due to change in the membrane curvature or due to hydrophobic mismatch between the protein and lipid bilayer at different Z distances (reaction coordinates). For this aim, we have monitored and measured the average curvature (C) and hydrophobic mismatch ( $d_{H}$ ) values over the last 5 ns of each umbrella sampling window for the PMF calculation. The free energy contribution of curvature, Δ $G_{C}$ , and hydrophobic mismatch, Δ $G_{H}$ , can be described as the following phenomenological expressions (Marsh, 2007; Kralj-Iglič et al., 1999; Kralj-Iglic et al., 1996; Bavi et al., 2016b; Svetina, 2015).

Δ G_{C} = \frac{1}{2} K_{C} {(C_{1} + C_{2} - C_{0})}^{2} A + K_{G} C_{1} C_{2} A

Δ G_{H} = \sqrt{2} {(\frac{K_{A}^{3} K_{B}}{t^{6}})}^{0.25} \frac{{| d_{p} - d_{l} |}^{2}}{4} 2 π R_{a v e}

Here for simplicity, it is assumed that since the MscS is an isotropic inclusion (i.e. cylindrical or conical in both principal planar directions), its insertion in the membrane causes symmetric curvature (i.e., $C_{1} = C_{2} = C$ ) in the membrane planar directions. Moreover, the intrinsic curvature of the protein has been assumed zero due to its shape. Where $K_{A}$ , $K_{C}$ and $C_{0}$ are the area expansion and bending moduli and spontaneous curvature of the lipid bilayer which, based on previous and our current MD simulations are assumed to be 200 mN/m, 31 k_BT and ~0 Å^-1, respectively (Akitake et al., 2005; Feller and Pastor, 1999; Marrink and Mark, 2001). $C_{0}$ is assumed to be ~ 0 Å^-1 based on the flat shape of the bilayer (including the embedded MscS structure) at equilibrium (Figure 6 and Figure 2—figure supplement 3C). t is the global thickness and A is the surface area of each monolayer and $R_{a v e}$ is the average external radius of the transmembrane part of the protein which is assumed to be ~ 2 nm. $K_{G}$ is the elastic modulus for Gaussian curvature, which has experimentally approximated to be ~ - 0.1 $K_{C}$ (Venable et al., 2015; Siegel and Kozlov, 2004; Templer et al., 1998; Marsh, 2007; Raghunathan et al., 2012; Venable et al., 2015).

We also investigated whether variable C as a function of lipid bilayer area would change the final values of the free energy. To do this, we discretized the membrane into concentric ribbons, $a_{1}$ to $a_{n}$ with $c_{1}$ to $c_{n}$ being their corresponding curvature values (Figure 2—figure supplement 4). Each value was averaged across different frames of the simulation (i.e. last 20 ns). Then the integral below (Equation 4) was used for calculating the free energy due to curvature change.

Δ G_{C} = \frac{1}{2} K_{C} \int C^{2} d A = \frac{1}{2} K_{C} \sum_{i = 1}^{n} a_{i} c_{i}^{2}

The resulting trend is similar to the case where we assumed a constant curvature across the bilayer, where the contribution of free energy due to change in the curvature is still an order of magnitude smaller than that of the hydrophobic mismatch (Figure 2—figure supplement 3D-F).

Software used in this project was curated by SBGrid (Morin et al., 2013).

Data availability

EM maps and atomic models have been deposited at the Electron Microscopy Data Bank (accession numbers EMD-20508, EMD-20510, EMD-20509, and EMD-20148) and the Protein Data Back (entry codes 6PWN, 6PWP and 6PWO).

The following data sets were generated

1. Reddy BG
2. Perozo E
(2019) RCSB Protein Data Bank
ID 6PWN. MscS Nanodisc with N-terminal His-Tag.

https://www.rcsb.org/structure/6PWN
1. Reddy BG
2. Perozo E
(2019) RCSB Protein Data Bank
ID 6PWP. MscS Nanodisc.

https://www.rcsb.org/structure/6PWP
1. Reddy BG
2. Perozo E
(2019) RCSB Protein Data Bank
ID 6PWO. MscS DDM.

https://www.rcsb.org/structure/6PWO
1. Reddy BG
2. Perozo
(2019) Electron Microscopy Data Bank
ID EMD-20508. MscS Nanodisc with N-terminal His-Tag.

https://www.ebi.ac.uk/pdbe/entry/emdb/EMD-20508
1. Reddy BG
2. Perozo E
(2019) Electron Microscopy Data Bank
ID EMD-20510. MscS Nanodisc.

https://www.ebi.ac.uk/pdbe/entry/emdb/EMD-20510
1. Reddy BG
2. Perozo E
(2019) Electron Microscopy Data Bank
ID EMD-20509. MscS DDM.

https://www.ebi.ac.uk/pdbe/entry/emdb/EMD-20509

References

1. Adams PD
2. Afonine PV
3. Bunkóczi G
4. Chen VB
5. Davis IW
6. Echols N
7. Headd JJ
8. Hung LW
9. Kapral GJ
10. Grosse-Kunstleve RW
11. McCoy AJ
12. Moriarty NW
13. Oeffner R
14. Read RJ
15. Richardson DC
16. Richardson JS
17. Terwilliger TC
18. Zwart PH
(2010) PHENIX: a comprehensive Python-based system for macromolecular structure solution
Acta Crystallographica Section D Biological Crystallography 66:213–221.
https://doi.org/10.1107/S0907444909052925
- PubMed
- Google Scholar
(2005) The "dashpot" mechanism of stretch-dependent gating in MscS
The Journal of General Physiology 125:143–154.
https://doi.org/10.1085/jgp.200409198
- PubMed
- Google Scholar
(2008a) Characterization of the resting MscS: modeling and analysis of the closed bacterial mechanosensitive channel of small conductance
Biophysical Journal 94:1252–1266.
https://doi.org/10.1529/biophysj.107.110171
- PubMed
- Google Scholar
(2008b) Mechanosensitive channel MscS in the open state: modeling of the transition, explicit simulations, and experimental measurements of conductance
The Journal of General Physiology 132:67–83.
https://doi.org/10.1085/jgp.200810000
- PubMed
- Google Scholar
(2010) Hydration properties of mechanosensitive channel pores define the energetics of gating
Journal of Physics: Condensed Matter 22:454120.
https://doi.org/10.1088/0953-8984/22/45/454120
- Google Scholar
1. Anishkin A
2. Sukharev S
(2004) Water dynamics and dewetting transitions in the small mechanosensitive channel MscS
Biophysical Journal 86:2883–2895.
https://doi.org/10.1016/S0006-3495(04)74340-4
- PubMed
- Google Scholar
1. Bass RB
2. Strop P
3. Barclay M
4. Rees DC
(2002) Crystal structure of Escherichia coli MscS, a voltage-modulated and mechanosensitive channel
Science 298:1582–1587.
https://doi.org/10.1126/science.1077945
- PubMed
- Google Scholar
1. Batiza AF
2. Kuo MM
3. Yoshimura K
4. Kung C
(2002) Gating the bacterial mechanosensitive channel MscL invivo
PNAS 99:5643–5648.
https://doi.org/10.1073/pnas.082092599
- PubMed
- Google Scholar
1. Bavi N
2. Cortes DM
3. Cox CD
4. Rohde PR
5. Liu W
6. Deitmer JW
7. Bavi O
8. Strop P
9. Hill AP
10. Rees D
11. Corry B
12. Perozo E
13. Martinac B
(2016a) The role of MscL amphipathic N terminus indicates a blueprint for bilayer-mediated gating of mechanosensitive channels
Nature Communications 7:11984.
https://doi.org/10.1038/ncomms11984
- PubMed
- Google Scholar
(2016b) The combined effect of hydrophobic mismatch and bilayer local bending on the regulation of mechanosensitive ion channels
PLOS ONE 11:e0150578.
https://doi.org/10.1371/journal.pone.0150578
- PubMed
- Google Scholar
Software
1. Bavi N
(2019) VMD script to Calculate Potential Mean Force, version 785277f
GitHub.

https://github.com/Perozo-lab/PMF
1. Bayburt TH
2. Sligar SG
(2010) Membrane protein assembly into nanodiscs
FEBS Letters 584:1721–1727.
https://doi.org/10.1016/j.febslet.2009.10.024
- PubMed
- Google Scholar
1. Booth IR
2. Blount P
(2012) The MscS and MscL families of mechanosensitive channels act as microbial emergency release valves
Journal of Bacteriology 194:4802–4809.
https://doi.org/10.1128/JB.00576-12
- PubMed
- Google Scholar
1. Brohawn SG
(2015) How ion channels sense mechanical force: insights from mechanosensitive K2P channels TRAAK, TREK1, and TREK2
Annals of the New York Academy of Sciences 1352:20–32.
https://doi.org/10.1111/nyas.12874
- PubMed
- Google Scholar
1. Brooks BR
2. Brooks CLI
3. Mackerell ADJ
4. Nilsson L
5. Petrella RJ
6. Roux B
7. Won Y
8. Archontis G
9. Bartels C
10. Boresch S
11. Caflisch A
12. Caves L
13. Cui Q
14. Dinner AR
15. Feig M
16. Fischer S
17. Gao J
18. Hodoscek M
19. Im W
20. Kuczera K
21. Lazaridis T
22. Ma J
23. Ovchinnikov V
24. Paci E
25. Pastor RW
26. Post CB
27. Jz P
28. Schaefer M
29. Tidor B
30. Venable RM
31. Woodcock HL
32. Wu X
33. Yang W
34. York DM
35. Karplus M
(2009) CHARMM: the biomolecular simulation program
Journal of Computational Chemistry 30:1545–1614.
https://doi.org/10.1002/jcc.21287
- Google Scholar
1. Camacho C
2. Coulouris G
3. Avagyan V
4. Ma N
5. Papadopoulos J
6. Bealer K
7. Madden TL
(2009) BLAST+: architecture and applications
BMC Bioinformatics 10:421.
https://doi.org/10.1186/1471-2105-10-421
- PubMed
- Google Scholar
1. Cantor RS
(1999) The influence of membrane lateral pressures on simple geometric models of protein conformational equilibria
Chemistry and Physics of Lipids 101:45–56.
https://doi.org/10.1016/S0009-3084(99)00054-7
- PubMed
- Google Scholar
1. Corry B
2. Thomas M
(2012) Mechanism of ion permeation and selectivity in a voltage gated sodium channel
Journal of the American Chemical Society 134:1840–1846.
https://doi.org/10.1021/ja210020h
- PubMed
- Google Scholar
1. Cox CD
2. Bavi N
3. Martinac B
(2018) Bacterial mechanosensors
Annual Review of Physiology 80:71–93.
https://doi.org/10.1146/annurev-physiol-021317-121351
- PubMed
- Google Scholar
1. Croll TI
(2018) ISOLDE : a physically realistic environment for model building into low-resolution electron-density maps
Acta Crystallographica Section D Structural Biology 74:519–530.
https://doi.org/10.1107/S2059798318002425
- Google Scholar
(2004) WebLogo: a sequence logo generator
Genome Research 14:1188–1190.
https://doi.org/10.1101/gr.849004
- PubMed
- Google Scholar
1. de Jesus AJ
2. Allen TW
(2013) The role of tryptophan side chains in membrane protein anchoring and hydrophobic mismatch
Biochimica Et Biophysica Acta (BBA) - Biomembranes 1828:864–876.
https://doi.org/10.1016/j.bbamem.2012.09.009
- Google Scholar
1. Dong YY
2. Pike AC
3. Mackenzie A
4. McClenaghan C
5. Aryal P
6. Dong L
7. Quigley A
8. Grieben M
9. Goubin S
10. Mukhopadhyay S
11. Ruda GF
12. Clausen MV
13. Cao L
14. Brennan PE
15. Burgess-Brown NA
16. Sansom MS
17. Tucker SJ
18. Carpenter EP
(2015) K2P channel gating mechanisms revealed by structures of TREK-2 and a complex with prozac
Science 347:1256–1259.
https://doi.org/10.1126/science.1261512
- PubMed
- Google Scholar
1. Edgar RC
(2004) MUSCLE: multiple sequence alignment with high accuracy and high throughput
Nucleic Acids Research 32:1792–1797.
https://doi.org/10.1093/nar/gkh340
- PubMed
- Google Scholar
(2004) Gating the bacterial mechanosensitive channels: mscs a new paradigm?
Current Opinion in Microbiology 7:163–167.
https://doi.org/10.1016/j.mib.2004.02.006
- PubMed
- Google Scholar
1. Egwolf B
2. Roux B
(2010) Ion selectivity of the KcsA channel: a perspective from multi-ion free energy landscapes
Journal of Molecular Biology 401:831–842.
https://doi.org/10.1016/j.jmb.2010.07.006
- PubMed
- Google Scholar
1. Emsley P
2. Lohkamp B
3. Scott WG
4. Cowtan K
(2010) Features and development of coot
Acta Crystallographica. Section D, Biological Crystallography 66:486–501.
https://doi.org/10.1107/S0907444910007493
- PubMed
- Google Scholar
1. Feller SE
2. Pastor RW
(1999) Constant surface tension simulations of lipid bilayers: the sensitivity of surface Areas and compressibilities
The Journal of Chemical Physics 111:1281–1287.
https://doi.org/10.1063/1.479313
- Google Scholar
1. Goddard TD
2. Huang CC
3. Meng EC
4. Pettersen EF
5. Couch GS
6. Morris JH
7. Ferrin TE
(2018) UCSF ChimeraX: meeting modern challenges in visualization and analysis
Protein Science 27:14–25.
https://doi.org/10.1002/pro.3235
- PubMed
- Google Scholar
1. Grossfield A
(2010) WHAM: The Weighted Histogram Analysis Method
WHAM: The Weighted Histogram Analysis Method, http://membrane.urmc.rochester.edu/wordpress/?page_id=126.

http://membrane.urmc.rochester.edu/wordpress/?page_id=126
- Google Scholar
1. Gullingsrud J
2. Schulten K
(2004) Lipid bilayer pressure profiles and mechanosensitive channel gating
Biophysical Journal 86:3496–3509.
https://doi.org/10.1529/biophysj.103.034322
- PubMed
- Google Scholar
1. Guo YR
2. MacKinnon R
(2017) Structure-based membrane dome mechanism for piezo mechanosensitivity
eLife 6:e33660.
https://doi.org/10.7554/eLife.33660
- PubMed
- Google Scholar
(2011) Mechanosensitive channels: what can they do and how do they do it?
Structure 19:1356–1369.
https://doi.org/10.1016/j.str.2011.09.005
- PubMed
- Google Scholar
(1996) VMD: visual molecular dynamics
Journal of Molecular Graphics 14:33–38.
https://doi.org/10.1016/0263-7855(96)00018-5
- Google Scholar
1. Jo S
2. Kim T
3. Iyer VG
4. Im W
(2008) CHARMM-GUI: a web-based graphical user interface for CHARMM
Journal of Computational Chemistry 29:1859–1865.
https://doi.org/10.1002/jcc.20945
- PubMed
- Google Scholar
(2015) Feeling force: physical and physiological principles enabling sensory mechanotransduction
Annual Review of Cell and Developmental Biology 31:347–371.
https://doi.org/10.1146/annurev-cellbio-100913-013426
- PubMed
- Google Scholar
(1996) Shapes of bilayer vesicles with membrane embedded molecules
European Biophysics Journal 24:311–321.
https://doi.org/10.1007/BF00180372
- PubMed
- Google Scholar
(1999) Free energy of closed membrane with anisotropic inclusions
Physics of Condensed Matter 10:5–8.
https://doi.org/10.1007/s100510050822
- Google Scholar
1. Kung C
(2005) A possible unifying principle for mechanosensation
Nature 436:647–654.
https://doi.org/10.1038/nature03896
- PubMed
- Google Scholar
(2010) Mechanosensitive channels in microbes
Annual Review of Microbiology 64:313–329.
https://doi.org/10.1146/annurev.micro.112408.134106
- PubMed
- Google Scholar
1. Ladoux B
2. Mège RM
(2017) Mechanobiology of collective cell behaviours
Nature Reviews Molecular Cell Biology 18:743–757.
https://doi.org/10.1038/nrm.2017.98
- PubMed
- Google Scholar
1. Lai JY
2. Poon YS
3. Kaiser JT
4. Rees DC
(2013) Open and shut: crystal structures of the dodecylmaltoside solubilized mechanosensitive channel of small conductance from Escherichia coli and Helicobacter pylori at 4.4 å and 4.1 å resolutions
Protein Science 22:502–509.
https://doi.org/10.1002/pro.2222
- PubMed
- Google Scholar
(2008) Automated macromolecular model building for X-ray crystallography using ARP/wARP version 7
Nature Protocols 3:1171–1179.
https://doi.org/10.1038/nprot.2008.91
- PubMed
- Google Scholar
1. Letunic I
2. Bork P
(2019) Interactive tree of life (iTOL) v4: recent updates and new developments
Nucleic Acids Research 47:W256–W259.
https://doi.org/10.1093/nar/gkz239
- PubMed
- Google Scholar
1. Levina N
2. Tötemeyer S
3. Stokes NR
4. Louis P
5. Jones MA
6. Booth IR
(1999) Protection of Escherichia coli cells against extreme turgor by activation of MscS and MscL mechanosensitive channels: identification of genes required for MscS activity
The EMBO Journal 18:1730–1737.
https://doi.org/10.1093/emboj/18.7.1730
- PubMed
- Google Scholar
1. Li J
2. Ostmeyer J
3. Cuello LG
4. Perozo E
5. Roux B
(2018) Rapid constriction of the selectivity filter underlies C-type inactivation in the KcsA potassium channel
The Journal of General Physiology 150:1408–1420.
https://doi.org/10.1085/jgp.201812082
- PubMed
- Google Scholar
1. Lomize MA
2. Pogozheva ID
3. Joo H
4. Mosberg HI
5. Lomize AL
(2012) OPM database and PPM web server: resources for positioning of proteins in membranes
Nucleic Acids Research 40:D370–D376.
https://doi.org/10.1093/nar/gkr703
- PubMed
- Google Scholar
(2015) The mechanosensitive channel of small conductance (MscS) functions as a Jack-In-The box
Biochimica Et Biophysica Acta (BBA) - Biomembranes 1848:159–166.
https://doi.org/10.1016/j.bbamem.2014.10.022
- Google Scholar
1. Marrink SJ
2. Mark AE
(2001) Effect of undulations on surface tension in simulated bilayers
The Journal of Physical Chemistry B 105:6122–6127.
https://doi.org/10.1021/jp0103474
- Google Scholar
1. Marsh D
(2007) Lateral pressure profile, spontaneous curvature frustration, and the incorporation and conformation of proteins in membranes
Biophysical Journal 93:3884–3899.
https://doi.org/10.1529/biophysj.107.107938
- PubMed
- Google Scholar
1. McGreevy R
2. Singharoy A
3. Li Q
4. Zhang J
5. Xu D
6. Perozo E
7. Schulten K
(2014) xMDFF: molecular dynamics flexible fitting of low-resolution X-ray structures
Acta Crystallographica Section D Biological Crystallography 70:2344–2355.
https://doi.org/10.1107/S1399004714013856
- PubMed
- Google Scholar
1. Morin A
2. Eisenbraun B
3. Key J
4. Sanschagrin PC
5. Timony MA
6. Ottaviano M
7. Sliz P
(2013) Collaboration gets the most out of software
eLife 2:e01456.
https://doi.org/10.7554/eLife.01456
- PubMed
- Google Scholar
(2017) Piezos thrive under pressure: mechanically activated ion channels in health and disease
Nature Reviews Molecular Cell Biology 18:771–783.
https://doi.org/10.1038/nrm.2017.92
- PubMed
- Google Scholar
1. Naismith JH
2. Booth IR
(2012) Bacterial mechanosensitive channels--MscS: evolution's solution to creating sensitivity in function
Annual Review of Biophysics 41:157–177.
https://doi.org/10.1146/annurev-biophys-101211-113227
- PubMed
- Google Scholar
(2019) The MemProtMD database: a resource for membrane-embedded protein structures and their lipid interactions
Nucleic Acids Research 47:D390–D397.
https://doi.org/10.1093/nar/gky1047
- PubMed
- Google Scholar
(2006) Lipid-protein interaction of the MscS mechanosensitive channel examined by scanning mutagenesis
Biophysical Journal 91:2874–2881.
https://doi.org/10.1529/biophysj.106.084541
- PubMed
- Google Scholar
1. Nomura T
2. Cox CD
3. Bavi N
4. Sokabe M
5. Martinac B
(2015) Unidirectional incorporation of a bacterial mechanosensitive channel into liposomal membranes
The FASEB Journal 29:4334–4345.
https://doi.org/10.1096/fj.15-275198
- Google Scholar
1. Perozo E
(2006) Gating prokaryotic mechanosensitive channels
Nature Reviews Molecular Cell Biology 7:109–119.
https://doi.org/10.1038/nrm1833
- PubMed
- Google Scholar
1. Pettersen EF
2. Goddard TD
3. Huang CC
4. Couch GS
5. Greenblatt DM
6. Meng EC
7. Ferrin TE
(2004) UCSF chimera--a visualization system for exploratory research and analysis
Journal of Computational Chemistry 25:1605–1612.
https://doi.org/10.1002/jcc.20084
- PubMed
- Google Scholar
1. Phillips JC
2. Braun R
3. Wang W
4. Gumbart J
5. Tajkhorshid E
6. Villa E
7. Chipot C
8. Skeel RD
9. Kalé L
10. Schulten K
(2005) Scalable molecular dynamics with NAMD
Journal of Computational Chemistry 26:1781–1802.
https://doi.org/10.1002/jcc.20289
- Google Scholar
1. Phillips R
2. Ursell T
3. Wiggins P
4. Sens P
(2009) Emerging roles for lipids in shaping membrane-protein function
Nature 459:379–385.
https://doi.org/10.1038/nature08147
- PubMed
- Google Scholar
1. Pliotas C
2. Dahl AC
3. Rasmussen T
4. Mahendran KR
5. Smith TK
6. Marius P
7. Gault J
8. Banda T
9. Rasmussen A
10. Miller S
11. Robinson CV
12. Bayley H
13. Sansom MS
14. Booth IR
15. Naismith JH
(2015) The role of lipids in mechanosensation
Nature Structural & Molecular Biology 22:991–998.
https://doi.org/10.1038/nsmb.3120
- PubMed
- Google Scholar
1. Pliotas C
2. Naismith JH
(2017) Spectator no more, the role of the membrane in regulating ion channel function
Current Opinion in Structural Biology 45:59–66.
https://doi.org/10.1016/j.sbi.2016.10.017
- PubMed
- Google Scholar
1. Raetz CR
2. Dowhan W
(1990)
Biosynthesis and function of phospholipids in Escherichia coli

The Journal of Biological Chemistry 265:1235–1238.
- PubMed
- Google Scholar
(2012) Structure and elasticity of lipid membranes with genistein and daidzein bioflavinoids using X-ray scattering and MD simulations
The Journal of Physical Chemistry B 116:3918–3927.
https://doi.org/10.1021/jp211904j
- PubMed
- Google Scholar
1. Rasmussen T
2. Rasmussen A
3. Singh S
4. Galbiati H
5. Edwards MD
6. Miller S
7. Booth IR
(2015) Properties of the mechanosensitive channel MscS pore revealed by tryptophan scanning mutagenesis
Biochemistry 54:4519–4530.
https://doi.org/10.1021/acs.biochem.5b00294
- PubMed
- Google Scholar
(2019) Structure of the mechanosensitive channel MscS embedded in the membrane bilayer
Journal of Molecular Biology 431:3081–3090.
https://doi.org/10.1016/j.jmb.2019.07.006
- PubMed
- Google Scholar
(2009) Chapter 11 - Reconstitution of membrane proteins in phospholipid bilayer nanodiscs
Methods in Enzymology 464:211–231.
https://doi.org/10.1016/S0076-6879(09)64011-8
- PubMed
- Google Scholar
1. Rohou A
2. Grigorieff N
(2015) CTFFIND4: fast and accurate defocus estimation from electron micrographs
Journal of Structural Biology 192:216–221.
https://doi.org/10.1016/j.jsb.2015.08.008
- PubMed
- Google Scholar
(2018) Structure of the mechanically activated ion channel Piezo1
Nature 554:481–486.
https://doi.org/10.1038/nature25453
- PubMed
- Google Scholar
1. Scheres SH
(2012) RELION: implementation of a bayesian approach to cryo-EM structure determination
Journal of Structural Biology 180:519–530.
https://doi.org/10.1016/j.jsb.2012.09.006
- PubMed
- Google Scholar
1. Shen R
2. Guo W
(2012) Mechanism for variable selectivity and conductance in mutated NaK channels
The Journal of Physical Chemistry Letters 3:2887–2891.
https://doi.org/10.1021/jz301225d
- Google Scholar
1. Siegel DP
2. Kozlov MM
(2004) The gaussian curvature elastic Modulus of N-Monomethylated dioleoylphosphatidylethanolamine: relevance to membrane fusion and lipid phase behavior
Biophysical Journal 87:366–374.
https://doi.org/10.1529/biophysj.104.040782
- Google Scholar
1. Sotomayor M
2. Schulten K
(2004) Molecular dynamics study of gating in the mechanosensitive channel of small conductance MscS
Biophysical Journal 87:3050–3065.
https://doi.org/10.1529/biophysj.104.046045
- PubMed
- Google Scholar
(2006) Voltage-dependent hydration and conduction properties of the hydrophobic pore of the mechanosensitive channel of small conductance
Biophysical Journal 90:3555–3569.
https://doi.org/10.1529/biophysj.105.080432
- PubMed
- Google Scholar
1. Stamatakis A
(2014) RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies
Bioinformatics 30:1312–1313.
https://doi.org/10.1093/bioinformatics/btu033
- PubMed
- Google Scholar
1. Steinbacher S
2. Bass R
3. Strop P
4. Rees DC
(2007) Structures of the prokaryotic mechanosensitive channels MscL and MscS
Current Topics in Membranes 58:1–24.
https://doi.org/10.1016/S1063-5823(06)58001-9
- Google Scholar
1. Sukharev S
2. Corey DP
(2004) Mechanosensitive channels: multiplicity of families and gating paradigms
Science Signaling 2004:re4.
https://doi.org/10.1126/stke.2192004re4
- Google Scholar
1. Svetina S
(2015) Curvature-dependent protein-lipid bilayer interaction and cell mechanosensitivity
European Biophysics Journal 44:513–519.
https://doi.org/10.1007/s00249-015-1046-5
- PubMed
- Google Scholar
1. Tan YZ
2. Baldwin PR
3. Davis JH
4. Williamson JR
5. Potter CS
6. Carragher B
7. Lyumkis D
(2017) Addressing preferred specimen orientation in single-particle cryo-EM through tilting
Nature Methods 14:793–796.
https://doi.org/10.1038/nmeth.4347
- PubMed
- Google Scholar
1. Tang G
2. Peng L
3. Baldwin PR
4. Mann DS
5. Jiang W
6. Rees I
7. Ludtke SJ
(2007) EMAN2: an extensible image processing suite for electron microscopy
Journal of Structural Biology 157:38–46.
https://doi.org/10.1016/j.jsb.2006.05.009
- PubMed
- Google Scholar
(1998) Gaussian curvature Modulus of an amphiphilic monolayer
Langmuir 14:7427–7434.
https://doi.org/10.1021/la980701y
- Google Scholar
1. Torrie GM
2. Valleau JP
(1977) Nonphysical sampling distributions in monte carlo free-energy estimation: umbrella sampling
Journal of Computational Physics 23:187–199.
https://doi.org/10.1016/0021-9991(77)90121-8
- Google Scholar
(2007) An optimized purification and reconstitution method for the MscS channel: strategies for spectroscopical analysis
Biochemistry 46:6766–6773.
https://doi.org/10.1021/bi700322k
- PubMed
- Google Scholar
(2008) A structural mechanism for MscS gating in lipid bilayers
Science 321:1210–1214.
https://doi.org/10.1126/science.1159674
- Google Scholar
1. Vásquez V
2. Sotomayor M
3. Cortes DM
4. Roux B
5. Schulten K
6. Perozo E
(2008) Three-dimensional architecture of membrane-embedded MscS in the closed conformation
Journal of Molecular Biology 378:55–70.
https://doi.org/10.1016/j.jmb.2007.10.086
- PubMed
- Google Scholar
(2015) Mechanical properties of lipid bilayers from molecular dynamics simulation
Chemistry and Physics of Lipids 192:60–74.
https://doi.org/10.1016/j.chemphyslip.2015.07.014
- PubMed
- Google Scholar
1. Wang W
2. Black SS
3. Edwards MD
4. Miller S
5. Morrison EL
6. Bartlett W
7. Dong C
8. Naismith JH
9. Booth IR
(2008) The structure of an open form of an E. coli mechanosensitive channel at 3.45 A resolution
Science 321:1179–1183.
https://doi.org/10.1126/science.1159262
- PubMed
- Google Scholar
1. Zhang XC
2. Liu Z
3. Li J
(2016) From membrane tension to channel gating: a principal energy transfer mechanism for mechanosensitive channels
Protein Science 25:1954–1964.
https://doi.org/10.1002/pro.3017
- PubMed
- Google Scholar
1. Zhao Q
2. Zhou H
3. Chi S
4. Wang Y
5. Wang J
6. Geng J
7. Wu K
8. Liu W
9. Zhang T
10. Dong MQ
11. Wang J
12. Li X
13. Xiao B
(2018) Structure and mechanogating mechanism of the Piezo1 channel
Nature 554:487–492.
https://doi.org/10.1038/nature25743
- PubMed
- Google Scholar
1. Zheng SQ
2. Palovcak E
3. Armache JP
4. Verba KA
5. Cheng Y
6. Agard DA
(2017) MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy
Nature Methods 14:331–332.
https://doi.org/10.1038/nmeth.4193
- PubMed
- Google Scholar

Decision letter

José D Faraldo-Gómez

Reviewing Editor; National Heart, Lung and Blood Institute, National Institutes of Health, United States
Richard W Aldrich

Senior Editor; The University of Texas at Austin, United States
Roderick MacKinnon

Reviewer; Howard Hughes Medical Institute, The Rockefeller University, United States
Sergei Sukharev

Reviewer; University of Maryland, College Park, United States

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

Acceptance summary:

The ion channel MscS has been an important model system to study the physical principles that explain molecular mechanosensation – which remain to be completely understood. In this article, Perozo and co-workers report single-particle cryo-EM structures of the MscS in lipid nanodiscs. These structures reveal the position of the channel relative to the membrane, which had been previously misjudged based on then-available structures. Specifically, the new data shows the transmembrane region is shifted by ~14 Å, taking TM3a and much of TM2 out of the membrane, and placing more of TM1 inside the membrane. These insights are consistent with free-energy calculations based on MD simulations. The structures also reveal the fold of a previously unrecognized 'anchor domain' at the N-terminus (of TM1), as well as ordered lipid molecules seemingly bound along the periphery of the channel as well as inside the pore. These are important insights that will no doubt serve as the foundation for future studies of the mechanism of gating of MscS and related mechanosensitive channels.

Decision letter after peer review:

Thank you for submitting your article "Molecular basis of force-from-lipidsgating in the mechanosensitive channel MscS" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by José D. Faraldo-Gomez as the Reviewing Editor andRichard Aldrich as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Roderick MacKinnon (Reviewer #2); Sergei Sukharev (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.

The reviewers find that the central conclusions of the article are "well supported" and "convincing". By contrast, the computational examination of the mechanism of channel opening, and of the role that the bound lipids might play in this process, was perceived as too preliminary and inconclusive.

Essential revisions:

1) The reviewers agree that the computational section that examines the gating mechanism of the channel in response to a bilayer expansion and the potential role that bound lipids might have in this process is too preliminary and incomplete for publication in eLife. The reviewers are skeptical that these challenging questions can be addressed convincingly within the limited timeframe permitted to make revisions (2 months). Thus, we very strongly recommend that this element be removed from the article. The reviewers agree that this change will not diminish the value of the study, but rather the opposite. The authors are however encouraged to pursue this kind of studies in the future; some of the reviewers' comments are therefore provided below, as they might be useful to the authors:

– As the data stands, it is concerning that the process induced is clearly far from equilibrium, as a very large perturbation in membrane morphology is created in a very short time. To establish statistical significance on the basis of this kind of data, the authors ought to carry out and analyze a large number of repeats; alternatively, the authors could change their simulation protocol so that the system remains close to equilibrium throughout the perturbation of the membrane.

– The surface tension of the simulated membrane was increased from 50 mN/m at rest to 150 mN/m. The authors state that this is several times larger than the experimental value so that the opening can be observed in shorter times. But the lytic tension of ordinary lipid membranes is 10-15 mN/m, which means the tension in the simulation goes from about 5 times lytic to about 15 times lytic. What happens to the bilayer in the simulation? These simulated tensions deserve some careful thought, consideration and explanation for the reader.

– Forcing MscS opening by applying 150 mN/m of tension within 10 ns is a regime that imposes unnatural force and time scales on the transition. The bilayer gets much thinner than can be expected under MscS gating tension; hence the lipid-derived forces are acting at very different z levels. The bilayer thins so much that in videos it looks like the outer monolayer reaches the level of the "anchor" lipids that used to be close to the midplane. Also, why so much (50 mN/m) tension at rest? POPE seems to be pretty well parameterized in the latest release of CHARMM. Did the bilayer over-compact too much at zero tension or was the structure collapsing as in some previous simulations? This should be explained.

– The putative role of R88-associated lipids seems important and warrants a much more thorough MD study that would require several repetitions of channel opening under 'slow' regimes closer to equilibrium. Regarding ways to illustrate the conclusions, either cartoons should be presented (Figure 6B) or a real and convincing MD trajectory, not both. We suggest that in the next paper the authors devote special effort to the MD part which appears unfinished in the present form.

– The simulations do not seem to clearly reveal the role that either the 'hook lipids' or the 'pore lipids' might play in the mechanism of gating, or indeed whether they do play a role. For example, the authors propose that the pore lipids are a feature of the closed state, and that they must return the membrane upon channel opening. Simulations could in principle be used to evaluate the plausibility of this hypothesis, but the trajectories discussed in the manuscript are much too short to explore the dynamics of these lipids in a meaningful way. Similarly, much longer simulations (or simulations based on enhanced-sampling techniques) could be used to ascertain whether the sites occupied by lipid molecules in the cryo-EM structure indeed remain occupied in this closed state, as hypothesized, and to evaluate how this occupancy might depend on the membrane surface tension. In summary, this element of the study is promising but as it stands, it also appears inconclusive and not ready for publication.

2) The hypothesis that lipids are naturally present inside the pore in the resting state and move out during gating seems to be far-fetched. The lipid action in K2P channels, for instance, appears to be feasible due to clear fenestrations in the channel wall; similarly, a possible action of the 'hook' lipids here in MscS can be envisioned because there is an exchange path with the lipid bilayer. There is no obvious path that would connect the pore lipids with the bilayer. In this case, the detachment from the hydrophobic wall would imply a complete solvation by water, which would be energetically costly and slow. Because the headgroups of these lipids have not been resolved in any of the structures, would it be possible that these densities are fatty acids that came as products of lipid degradation? Fatty acid exchange is much more feasible due to higher solubility.

Given these concerns, we strongly recommend that the authors tone down their conclusions in regard to the observation of 'pore lipids' and that they discuss other plausible interpretations of the density signals. Otherwise, the authors are asked to provide an additional WT MscS structure obtained with a deliberate attempt to remove or minimize putative free-fatty acids, e.g. by incubating the nanodiscs with cyclodextrins or other type of fatty-acid absorbents. High-resolution structures might not be required to discern whether the densities inside the pore are still present or not.

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

Thank you for submitting the revised version of your article "Molecular basis of force-from-lipids gating in the mechanosensitive channel MscS" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by José D. Faraldo-Gómez as the Reviewing Editor and Richard Aldrich as the Senior Editor. The following individuals involved in review of your resubmission have agreed to reveal their identity: Roderick MacKinnon (Reviewer #2); Sergei Sukharev (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.

Summary:

The reviewers believe the most important issues addressed in the earlier evaluation have been addressed, and that the revised manuscript will be worthy of publication in eLife, pending the following revisions:

The quantity defined as DGc in paragraph three of subsection “Continuum-Mean filed calculations of the free energy change” is not a free energy but a free-energy density, defined only locally at a point where the membrane curvature is C. The free energy that the authors seek should be an integral of this density over the area of membrane; the resulting values will be therefore considerably larger than the values currently plotted in Figure 3—figure supplement 3D. In addition, although it is valid as a first approximation to assume that at a given point (X, Y) the curvature C is same in the both directions, C(X,Y) cannot be assumed to be constant everywhere – i.e. the value of the abovementioned integral must be finite. Thus, the statement that "Although the contribution of curvature in our PMF calculations are minimal compared to the effect of hydrophobic mismatch" is not evident based on the data provided.

https://doi.org/10.7554/eLife.50486.sa1

Author response

Essential revisions:

1) The reviewers agree that the computational section that examines the gating mechanism of the channel in response to a bilayer expansion and the potential role that bound lipids might have in this process is too preliminary and incomplete for publication in eLife. The reviewers are skeptical that these challenging questions can be addressed convincingly within the limited timeframe permitted to make revisions (2 months). Thus, we very strongly recommend that this element be removed from the article. The reviewers agree that this change will not diminish the value of the study, but rather the opposite. The authors are however encouraged to pursue this kind of studies in the future; some of the reviewers' comments are therefore provided below, as they might be useful to the authors:

Indeed. We had been aware that a surface tension of 150 mN/m (applied solely to better sample any putative transitions in MscS) led to a membrane deformation that was perhaps too sharp and fast (see below). The intention was to include those bilayer expansion calculations as placeholders for more precise and ongoing calculations closer to equilibrium conditions (less intense surface tensions at 75, and 100 and 125 mN/m). These calculations (in replicates) started running before the manuscript was originally submitted. While the results are encouraging and reproducible, our intent is not to further delay publication of the present contribution. We agree with the reviewers that a dedicated computational study should be addressed in a separate publication. Therefore, we have removed this section of the manuscript.

– As the data stands, it is concerning that the process induced is clearly far from equilibrium, as a very large perturbation in membrane morphology is created in a very short time. To establish statistical significance on the basis of this kind of data, the authors ought to carry out and analyze a large number of repeats; alternatively, the authors could change their simulation protocol so that the system remains close to equilibrium throughout the perturbation of the membrane.

Agreed. These precise simulations (as mentioned above) are ongoing and shall be reported in a separate manuscript. Also, see below.

– The surface tension of the simulated membrane was increased from 50 mN/m at rest to 150 mN/m. The authors state that this is several times larger than the experimental value so that the opening can be observed in shorter times. But the lytic tension of ordinary lipid membranes is 10-15 mN/m, which means the tension in the simulation goes from about 5 times lytic to about 15 times lytic. What happens to the bilayer in the simulation? These simulated tensions deserve some careful thought, consideration and explanation for the reader.

– Forcing MscS opening by applying 150 mN/m of tension within 10 ns is a regime that imposes unnatural force and time scales on the transition. The bilayer gets much thinner than can be expected under MscS gating tension; hence the lipid-derived forces are acting at very different z levels. The bilayer thins so much that in videos it looks like the outer monolayer reaches the level of the "anchor" lipids that used to be close to the midplane. Also, why so much (50 mN/m) tension at rest? POPE seems to be pretty well parameterized in the latest release of CHARMM. Did the bilayer over-compact too much at zero tension or was the structure collapsing as in some previous simulations? This should be explained.

These two comments touch on similar issues, so we combined our response here:

There might be a confusion between “surface tension” and “membrane tension”. The values we are referring to are of surface tension. We have included a figure (Author response image 1) from our previous publication to further clarify this point. Our original MD simulations used an NγP_zT ensemble (1-3) in which “target surface tension” refers to the area of the pressure profile of the lipid bilayer near the lipid-water interface (Author response image 1A). This is always a positive (attractive) tension of about 50 mN/m, which corresponds to the red area in the pressure profile below. In fact, without this level of surface tension the bilayer would otherwise shrink references [1,3,4].

The large surface tension values at the lipid-water interface are compensated to near zero by the compressive forces from entropic repulsion at 1), the core of the bilayer and 2) electrostatic repulsions between the headgroups (the blue areas in Author response image 1A). Therefore, the total forces (membrane tension) along the Z-axis of the membranes at rest are indeed close to zero.

Membrane tension, as used in high speed patch clamp electrophysiology or micropipette aspiration rather refers to the latter case, i.e. total external forces that have been applied to the bilayer at rest, which cause an almost uniform increase in the lipid bilayer pressure profile from the equilibrium and subsequently an increase in the area per lipid. With this distinction, if the surface tension is set to i.e. 75 mN/m in an MD simulation, compared to equilibrium (~ 50 mN/m), it means that the total additional force (membrane tension) that has been applied to our MD system is ~ 25 mN/m (Author response image 1B).

Ultimately, however, we fully agree with the reviewers that the force in our MD simulation is still several times higher than the experimental range (i.e., ~ 4 x when the surface tension is set to 75 mN/m). This is given that the midpoint of activation of MscS is ~ 5-6 mN/m (references 5-7) or as pointed out by the reviewers the lytic tension of lipid bilayers are usually <15 mN/m. Nonetheless, this is a common issue with these types of MD simulations and is a ‘necessary evil’ to be able to capture noticeable gating transitions within limited computational time scales (~200-300 ns).

Author response image 1

Download asset Open asset

Difference between the surface tension and overall membrane tension based on the lateral pressure profile of the bilayer.

(A) The area shaded with red shows where regarded as “surface tension”, and blue the repulsive forces at the tail region of the bilayer and between the headgroups. (B) Illustration of how the pressure profile of the lipid bilayer in the presence of an MS channel (MscL) changes at different surface tensions of 50 mN/m and 75 mN/m in an NγPzT ensemble (Adapted from Bavi et al., 2016 (reference 4)). As shown, when the target surface tension is 50 mN/m the pressure profile of the lipid shows little or no change in the presence of the channels. Changes in the pressure profile only start to be noticed at 75 mN/m, particularly at the lipid water interface. In this condition (target surface tension of 75 mN/m; blue line), compared to a non-stressed bilayer (equilibrium; dark green) and based on the area under the pressure profile we estimate a mean membrane tension of 25.5 mN/m has been applied on the membrane, which is about 4x the experimental value required to activate MscS in patch clamp experiments.

– The putative role of R88-associated lipids seems important and warrants a much more thorough MD study that would require several repetitions of channel opening under 'slow' regimes closer to equilibrium. Regarding ways to illustrate the conclusions, either cartoons should be presented (Figure 6B) or a real and convincing MD trajectory, not both. We suggest that in the next paper the authors devote special effort to the MD part which appears unfinished in the present form.

Fully agree. Also, see response to comments above. We have eliminated the explicit trajectories and summarize our conclusion in cartoon form.

– The simulations do not seem to clearly reveal the role that either the 'hook lipids' or the 'pore lipids' might play in the mechanism of gating, or indeed whether they do play a role. For example, the authors propose that the pore lipids are a feature of the closed state, and that they must return the membrane upon channel opening. Simulations could in principle be used to evaluate the plausibility of this hypothesis, but the trajectories discussed in the manuscript are much too short to explore the dynamics of these lipids in a meaningful way. Similarly, much longer simulations (or simulations based on enhanced-sampling techniques) could be used to ascertain whether the sites occupied by lipid molecules in the cryo-EM structure indeed remain occupied in this closed state, as hypothesized, and to evaluate how this occupancy might depend on the membrane surface tension. In summary, this element of the study is promising but as it stands, it also appears inconclusive and not ready for publication.

We agree with the gist of these comments. Longer and targeted simulations will likely help discriminate between the different possible roles of the hook and pore lipids on MscS gating. As stated throughout this overall response to reviewers comments, the gating mechanism simulations are ongoing and shall be reported in a separate publication. Specifically, at identical external forces and computational conditions, we compared the channel opening with and without the presence of the hooked and pore lipids. We are also comparing all these scenarios with the model based on “previously believed” position of the bilayer systems with respect to the protein.

We do however believe that the simple evaluation of the closed state water occupancy (with and without pore lipids) and its influence on “vapor lock” mechanism are still informative and will be somewhat toned down but remain in the manuscript. Also, see response below.

2) The hypothesis that lipids are naturally present inside the pore in the resting state and move out during gating seems to be far-fetched. The lipid action in K2P channels, for instance, appears to be feasible due to clear fenestrations in the channel wall; similarly, a possible action of the 'hook' lipids here in MscS can be envisioned because there is an exchange path with the lipid bilayer. There is no obvious path that would connect the pore lipids with the bilayer. In this case, the detachment from the hydrophobic wall would imply a complete solvation by water, which would be energetically costly and slow. Because the headgroups of these lipids have not been resolved in any of the structures, would it be possible that these densities are fatty acids that came as products of lipid degradation? Fatty acid exchange is much more feasible due to higher solubility.

Given these concerns, we strongly recommend that the authors tone down their conclusions in regard to the observation of 'pore lipids' and that they discuss other plausible interpretations of the density signals. Otherwise, the authors are asked to provide an additional WT MscS structure obtained with a deliberate attempt to remove or minimize putative free-fatty acids, e.g. by incubating the nanodiscs with cyclodextrins or other type of fatty-acid absorbents. High-resolution structures might not be required to discern whether the densities inside the pore are still present or not.

We have indeed toned down our interpretation and expanded our discussion regarding the identity of the “pore lipids”. The density we observe supports the likely presence of some type of fatty acid chains in the pore. Indeed, we were able to fit a hexadecamer fatty acid within the density. As understood by the reviewers, likely due to dynamics/resolution we were unable to resolve a putative head group. We see one of two possibilities of where the head group can come from. Either the fatty acid chain could potentially thread through a gap between G104 residues, where MOLEonline (https://mole.upol.cz) supports the presence of a pathway in our structure (Author response image 2); alternatively, the lipid could come from the monolayer on the periplasmic side. We stand by our finding: The density we associate to the “pore lipids” is a robust feature of the density signal, it is present in every single one of our high-res FC14-extracted ND structures (even in DDM) and is also prominently featured in the Rasmussen structure (using different biochemical approaches). Ultimately, whatever it this density represents (and we still hypothesize it is lipidic), we firmly argue that it is not a consequence of lipid “degradation” or poor biochemical handling of our sample. Further, we are not keen on carrying out the cyclodextrin experiment. Given the location of the pore lipids (closest extracellular access is through a ~23Å opening, ~30 Å from the aqueous solution, at its shortest) it is unlikely that any cyclodextrin would be able to effectively access and extract any of it. Thus, while we hypothesize that these putative pore lipids contribute to a more effective “vapor lock” in the closed state, there are still aspects of the gating mechanism that are both fascinating and mysterious. Along those lines, we have changed the manuscript accordingly: softening our conclusions in this regard while reporting on the presence of the pore lipid density right above the narrowest point of the MscS permeation pathway.

Author response image 2

Download asset Open asset

Predicted pathway connecting the location of the putative pore lipids in the permeation path with the intracellular TM2/TM3a cavity.

Volume predicted by MOLEonline (https://mole.upol.cz) is depicted as a blue transparent envelope. The tan ribbon corresponds to the TM3 a and TM3b helices, with G104 colored red. An arrow points to the potential connecting gap between adjacent subunits.

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

The quantity defined as DGc in paragraph three of subsection “Continuum-Mean filed calculations of the free energy change” is not a free energy but a free-energy density, defined only locally at a point where the membrane curvature is C. The free energy that the authors seek should be an integral of this density over the area of membrane; the resulting values will be therefore considerably larger than the values currently plotted in Figure 3—figure supplement 3D. In addition, although it is valid as a first approximation to assume that at a given point (X, Y) the curvature C is same in the both directions, C(X,Y) cannot be assumed to be constant everywhere – i.e. the value of the abovementioned integral must be finite. Thus, the statement that "Although the contribution of curvature in our PMF calculations are minimal compared to the effect of hydrophobic mismatch" is not evident based on the data provided.

The reviewer is correct. We have now recalculated this free energy density as the integral of a series of concentric area rings (sed to reduce frame-to-frame-variability in the MD simulation run). We include that as a modified Figure 3—figure supplement 3 and a new Figure 3—figure supplement 4, with corresponding text and full methodological description in subsection “Continuum Mean-Field calculations of the free energy change”. The new calculations supports the original idea and demonstrates that “the contribution of free energy due to change in the curvature is still an order of magnitude smaller than that of the hydrophobic mismatch”.

References

[1] Feller, Scott E., and Richard W. Pastor. (1999) J Chem Phys 111: 1281-1287.

[2] Marrink, S. J., and A. E. Mark. (2001) J Physical Chem B105: 6122-6127.

[3] Gullingsrud, Justin, and Klaus Schulten. (2004) Biophys J 86: 3496-3509.

[4] Bavi, Navid, et al. (2016) Nature Comm 7 (2016): 11984.

[5] Akitake, Bradley, Andriy Anishkin, and Sergei Sukharev. (2005) JGP 125: 143-154.

[6] Nomura et al. (2012) PNAS 109 (2012): 8770-8775.

[7] Nomura et al. (2015) FASEB J 29: 4334-4345.

https://doi.org/10.7554/eLife.50486.sa2

Article and author information

Author details

Bharat Reddy

Department of Biochemistry and Molecular Biology, The University of Chicago, Chicago, United States

Contribution
Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology

Competing interests
No competing interests declared
Navid Bavi

Department of Biochemistry and Molecular Biology, The University of Chicago, Chicago, United States

Contribution
Data curation, Formal analysis, Investigation

Competing interests
No competing interests declared
Allen Lu

Department of Biochemistry and Molecular Biology, The University of Chicago, Chicago, United States

Contribution
Data curation, Formal analysis, Investigation

Competing interests
No competing interests declared
Yeonwoo Park

Department of Ecology and Evolution, The University of Chicago, Chicago, United States

Contribution
Formal analysis, Investigation

Competing interests
No competing interests declared
Eduardo Perozo
1. Department of Biochemistry and Molecular Biology, The University of Chicago, Chicago, United States
2. Institute for Biophysical Dynamics, The University of Chicago, Chicago, United States
Contribution
Conceptualization, Formal analysis, Supervision, Funding acquisition, Investigation, Visualization

For correspondence
eperozo@uchicago.edu

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0001-7132-2793

Funding

National Institute of General Medical Sciences (R01GM131191)

Eduardo Perozo

National Institute of General Medical Sciences (U54GM087519)

Eduardo Perozo

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

Acknowledgements

We thank Drs. Charles Cox, Boris Martinac, Ray Hulse, Yamuna Krishnan, Engin Ozkan, Minglei Zhao, Valeria Vasquez, Radomir Slovchov and the members of the Perozo lab for a healthy exchange of ideas and comments on the manuscript. Dr. Joseph W Thornton for fundamental insights on molecular phylogeny. Dr. Pedro Rodriguez and Wieslawa Milewski carried out the initial screen and mutant generation, respectively. We thank Ian Booth and Samatha Miller for their contribution of plasmids and bacterial strains. Michael Clark inspired the first sentence of this work. We are grateful to Drs. Lorenzo Pesce, Abhishek Singharoy and Rong Shen for invaluable assistance. This research was, in part, supported by the National Cancer Institute’s National Cryo-EM Facility at the Frederick National Laboratory for Cancer Research under contract HSSN261200800001E. We thank Ulrich Baxa and Thomas J Edwards at NCEF for cryo-EM data collection, Pedro Rodriguez, Joe Austin II, and Tera Lavoie at the University of Chicago Advanced Electron Microscopy Facility for microscope maintenance and training. Dr. Mario Borgnia and Allen Hsu provided invaluable microscope time at the National Institute of Environmental Health Sciences. This work was supported by the Membrane Protein Structural Dynamics Consortium U54GM087519 and by NIGMS grant R01GM131191 to EP. AL was recipient of a BSCD Fellowship at the University of Chicago. NB is a Chicago Fellow. EM maps and atomic models have been deposited at the Electron Microscopy Data Bank (accession numbers EMD-20508, EMD-20510 and EMD-20509) and the Protein Data Back (entry codes 6PWN, 6PWP and 6PWO).

Senior Editor

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

Reviewing Editor

José D Faraldo-Gómez, National Heart, Lung and Blood Institute, National Institutes of Health, United States

Reviewers

Roderick MacKinnon, Howard Hughes Medical Institute, The Rockefeller University, United States
Sergei Sukharev, University of Maryland, College Park, United States

Version history

Received: July 24, 2019
Accepted: December 27, 2019
Accepted Manuscript published: December 27, 2019 (version 1)
Version of Record published: January 24, 2020 (version 2)

Copyright

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.