Extended Data Fig. 1: Optimization of lipid density and leaflet equilibration in CG-MD simulations.
From: Brominated lipid probes expose structural asymmetries in constricted membranes
(a–f) Final frame of tubule from side (left) and top-down (right) from simulations tuning total lipid count. Solvent and protein removed for clarity. Lipid composition was composition 1, with total lipid amounts (a) 1017, (b) 1187, (c) 1271, (d) 1314, (e) 1356, and (f) 1695. Cholesterol in light gray, lipid tails in dark gray, glycerol in light pink, phosphates in orange, SDPC headgroups in beige, POPS headgroups in yellow, and PIP2 headgroups in bright pink. Insufficient lipids result in vesicle formation (a) or tubule with varying degrees of neck narrowing away from the protein (b, c); whereas an over-packed tubule bulges outward away from the protein (e) or fills the tubule lumen with lipids (f). A lipid count of 1300 (d) was used for production simulations since this resulted in a consistent tubule radius along the z axis with a clearly defined and solvated tubule lumen. These simulations employed a protein coat (PDB ID 3JC1) with opposite handedness to the coat used for all simulations reported in the main text (PDB ID 6TZ5), but nearly identical dimensions. These simulations used an isotropic Berendsen barostat for pressure coupling and were run with a 0.03 ps timestep for 6 μs, while all other parameters were identical to those described elsewhere in Methods). (g, h) Leaflet compositions over time from one replicate of WT CHMP1B-IST1 composition 1 simulation during leaflet equilibration (procedure described in SI Section 13). Leaflet assignment was determined using a simple r cutoff (r < 42 Å inner, else outer) applied to the lipid center of mass, since the clustering algorithm performed poorly in the presence of the equilibration pore due to the connection between the two leaflets. PIP2 was found to typically require the longest time to equilibrate, likely due to its low total number in the lipid mix and strong electrostatic attraction to the protein coat. Equilibration was run for 9 μs for composition 1 simulations and 6 μs for composition 17 simulations, as composition 17 replicates more quickly converged to similar leaflet compositions. Data for graphs in g, h are available as source data.
