Figure 4. Closed-state MscS induces drastic perturbations of the lipid bilayer.

The figure summarizes the results from multiple simulations of the closed structure of MscS in different membrane compositions and using different forcefield representations (Table 1). The cryo-electron microscopy (EM) structure of MscS (blue cartoons) is overlaid with calculated 3D density distributions mapping the morphology of the alkyl chain bilayer in each of the molecular dynamics (MD) trajectories (gold volume), up to 50 Å from the protein surface. Protein and density maps are viewed along the membrane plane (top row) and along the pore axis, from the cytosolic side (bottom row); the latter includes only the transmembrane domain of the channel, for clarity. The calculated density maps derive from 20 µs of trajectory data for each of the coarse-grained systems and at least 8 µs of trajectory data for the all-atom systems.

Figure 4—figure supplement 1. Lipid solvation of hydrophobic cavities outside the membrane drives the formation of inner-leaflet protrusions in closed-state MscS.

The figure shows three different views of a fragment of the closed MscS structure comprising TM1, TM2, and TM3a (from left to right) in two alternative representations (top and bottom). Residues are colored according to type, as indicated. The location of the inner leaflet protrusions under the TM1–TM2 hairpin is indicated; the site where the so-called ‘hook’ lipids are observed is also indicated. The conformation of the channel that is represented is that in the all-atom snapshot shown in Figure 5A.

Figure 4—figure supplement 2. Relaxation of the closed-state MscS structure in all-atom simulations.

(A) To preclude large-scale changes in fold that might develop in the 10-μs timescale due to cumulative forcefield inaccuracies, a restraining potential was applied to all $\varphi$ and $\psi$ angles in the channel backbone, of the form ${\displaystyle U({\theta }_t)=k\sum _{m=1}^{m=6}(-1{)}^m[1+\mathrm{c}\mathrm{o}\mathrm{s}(\mathrm{m}{\theta }_{\mathrm{t}}-\mathrm{m}({\theta }_{\mathrm{e}\mathrm{x}\mathrm{p}\mathrm{t}}-180))]/\mathrm{m}!}$, where ${\theta }_t$ is the value of either dihedral angle at time $t$ in the simulation, ${\theta }_{expt}$ denotes the value in the experimental structure, and $k$ equals 4 kJ/mol (see Materials and methods). The black curve in the left plot in (A) exemplifies this potential for ${\theta }_{obs}$ = 180; for comparison, a harmonic potential $U`\left({\theta }_t\right)=k{({\theta }_t-{\theta }_{obs})}^2$ is superimposed (red line), where $k=$ 0.0025 kJ/mol/deg². The right plot in (A) shows each of the first four terms contributing to $U\left({\theta }_t\right)$. (B) For each residue in the channel, observed deflections in $\varphi$ and $\psi$ during the simulation relative to the corresponding values in the experimental structure, individually quantified as probability density distributions. The left plots show data for a single chain in the heptamer, selected at random; the right plots show a global analysis for all seven chains. (C) Root-mean-squared (RMS) difference between simulated and experimental channel structure, as a function of simulation time. Data is provided for the backbone atoms only and for backbone and sidechains, excluding hydrogen atoms. The RMS difference calculation was preceded by least-squares superposition of the protein backbone in the snapshot considered and in the experimental structure. The plot on the left shows data for the initial equilibration of the simulation system, carried out using NAMD; the plot on the right shows data for the ANTON2 simulation (see Materials and methods).