Changes in ESCRT-III filament geometry drive membrane remodelling and fission in silico

2019-09-20T10:30:39Z (GMT) by Lena Kirschneck Andela Saric Buzz Baum
These are the data sets required to generate the figures in our paper 'Changes in ESCRT-III filament geometry drive membrane remodelling and fission in silico'.

ESCRT-III is a protein complex with the unique ability to deform cell membranes away from the cytoplasm. It plays a vital role in many important processes inside our cells such as cell division, membrane healing, budding of certain viruses (such as HIV and Ebola), formation of multi vesicular bodies and many more. The mechanism by which ESCRT-III deforms membranes is unknown.

We developed a coarse-grained biophysical model of the complex and studied its membrane deforming abilities in molecular dynamics simulations using lammps ( The aim was to identify the physical skeleton of this bio machine - the core elements that drive membrane deformation. We therefore built a pre-assembled spiral filament resembling ESCRT-III once it polymerised and had it interact with a simulated membrane.

The filament consists of identical subunits that are interconnected via very strong bonds and each experience the same target curvature. The resulting target geometry of the filament would hence be a ring - however, the filament is longer and it instead grows into a spiral, which has internal tension since not all subunits can be at their desired curvature. We found that a 3 bead subunit that is interconnected to the next subunit via 9 specific bonds is required to give the filament this chiral curvature (unique direction of curvature in 3D).

It was previously assumed that the inner tension due to its spiral shape (unsatisfied curvature) is what drives the filament to deform membranes. However, this would not explain why the deformations only occur away from the cytoplasm. We tested this hypothesis by placing our ESCRT-III spiral on a membrane and giving all subunits the same intrinsic curvature so that they would want to form a ring on the flat surface rather than a spiral. The result can be seen in 'Buckledepth_evolution.txt' (or Fig1 in the paper), which contains the time steps of our simulation and the corresponding buckle depth (measured by the smallest z-value of the filament at that time step). We can see that no buckle is forming despite the intrinsic tension of the spiral. So this tension can't be the driving force of the deformation.

2. & 3.
Instead we found that there is one essential parameter that we call the tilt angle which decides whether the spiral stays flat on the membrane or forms a helix. This angle tilts the membrane attracted surface of the spiral out of its 2D plane into a 3D cone surface.
Experimental evidence suggests that different ESCRT-III protein combinations can lead to very different deformations and that the transition from a flat spiral to a helix occurs because of such structural changes within the filament (e.g. depolymerisation or co-polymerisation).
We tested this by introducing a tilt angle change from 0 degrees (flat) to 60 degrees making the target geometry of our filament a tilted ring. 'Buckledepth_evolution_down.txt ' contains the time steps with corresponding buckle depth for this simulation (see Fig 2 in the paper). Now we can indeed observe a downward buckle forming. Similarly if we tilt the filament in the opposite direction (target geometry change from 0 degrees to -40 degrees) we observe a buckle forming in the opposite direction as can be seen in 'Buckleheight_evolution.txt' (see Fig. 2 in the paper).

We then studied the influence of the persistence length of the filament and the radius of the target geometry on the resulting buckle and found that a larger persistence length (stronger bonds between subunits) leads to more tubular (and hence deeper) deformations and a smaller radius results in narrower buckles (hence also deeper given we use the same filament length and it can form more loops in the helix). The data for this heat map is stored in Radius_persistencelength_depth.txt, which contains the target ring radius with matching persistence length and resulting buckle depth (compare to Fig.3 in the paper).

We also studied how the radius and tilt angle affect the forming buckle and stored this data in 'Radius_tilt_z.txt', which contains the target state radius with matching tilt angle tau and the resulting buckle depth or height. At tilt 0 we get no deformations, for positive tilt angles we get downward buckles that become more tubular with increasing angle and for negative tilt angles we get upward buckles that also become more tubular as the modulus of the angle increases.
These dataset are used to generate the heatmaps in Fig 3.

Finally we found that a series of target geometry changes from spiral (tilt 0 degrees) to helix (tilt 60 degrees) and back to flat spiral (tilt 0 degrees) around a large cargo particle that is attracted to receptors within the membrane can lead to scission of this cargo particle. The vesicle buds off enclosed by a piece of the original membrane. '' is the read-in file for lammps used to generate the simulation output '' (Fig 4.a) and '' is the read-in file for lammps used to generate the simulation output '' (Fig 4.b). The difference is the size of the cargo particle and the percentage of receptor beads used. Both result in vesicle formation but in case b (large cargo) the filament enters an already existing membrane neck, whereas in case a (small cargo) it forms the neck itself. The simulation output files contain the x y and z coordinates of all particles within the simulation at any time step as well as the atom number, molecule id and particle type (1 membrane, 2 and 3 membrane attracted filament beads, 4 non attracted filament bead, 5 membrane receptor, 6 cargo particle).