4. Ring modelling results and discussion 105
4.8. Discussion 168
4.8.1. Weaknesses of the original model 168
As previously mentioned, the models that we developed here were adapted from a previously developed ring model, the goal of which was to reproduce experimentally measured values of ring tension in silico. An average ring tension of 390 ± 150 pN was measured from ring sliding in fission yeast spheroplasts, while the simulation produced a ring tension of 340 ± 57 pN [35]. However, we initially found some aspects of the model to be unrealistic, particularly with the way that individual myosin clusters/nodes interact with multiple actin filaments, and this calls into question the validity of the results previously obtained with this model. Most importantly, we found that the value of the parameter maxInt, which controls how a myosin cluster/node interacts with multiple actin filaments, was vastly overestimated. In the original model, myosin clusters were assumed to contain 40 Myo2 molecules, and were given a maxInt value of 10, meaning that the authors were claiming that 4 Myo2 molecules were able to exert a time averaged force of 4 pN on an actin filament [35]. Type II myosins are known to be non-‐processive, making this behaviour unlikely [169]. Furthermore, when we implemented an equivalent description in our model (20 Myo2 molecules per node, maxInt set to 5) using a cylindrical geometry we found that large numbers of actin filaments peeled away from the ring as individual filaments (Figure 4.2C), and this was only resolved by reducing the value of maxInt to 1 (Figure 4.2D), so that the myosin pulling forces did not overpower the grabbing forces.
This would then suggest that a maxInt value of 2 or 3 would be more appropriate for the original simulations performed with this model, where there were 40 Myo2 molecules per cluster. These maxInt values correspond to 20 or ~13 Myo2 molecules per cluster, respectively, walking processively along a single actin filament. This would then reduce the tension generated by the model by a factor of ~5 or ~3, respectively, leading to approximate values of ~68 pN or ~110 pN, respectively. This is significantly less than what was observed experimentally, suggesting that the mechanism of tension generation in this model is not sufficient to produce the experimentally measured values, and/or the experimentally measured value is over estimated. It is also possible that rings in spheroplasts are able to recruit myosin at a greater density than rings in cells, meaning that they have a higher tension. Better understanding of this discrepancy will require further experimental work, including more measurements of AMR tension in fission yeast, and further mathematical modelling.
We also argue that the myosin grabbing forces were implemented in an unrealistic manner. Unlike the pulling forces, which were reduced when a myosin cluster/node interacts with multiple actin filaments, the grabbing forces are not reduced, meaning that a single node or cluster can hypothetically exert a finite force on an infinite number of actin filaments [35]. We first attempted to correct this by limiting the number of interactions with actin filaments that are available for each node. We set this to 2, meaning that each node can only exert pulling and grabbing forces on 2 actin filaments (Figure 4.3A). However, this prevented us from using the 3D cylindrical geometry for our simulations, as the pulling forces would now overpower the grabbing forces and lead to actin filaments being pulled away from the membrane. This was a problem that was present in subsequent iterations of our model, and which we did not find a way to solve.
One possible, and somewhat realistic, solution to this problem would be to increase the number of entities that are able to grab actin filaments, but not exert pulling forces on them. For example, the node
protein Rng2 contains an actin binding calponin homology domain (CHD) [170,171], whilst unpublished research from our lab has shown that Pxl1, which interacts with the SH3 domains of Cdc15 and Imp2 [75], is also able to bind to tropomyosin, which provides another mechanism whereby actin filaments can be linked to the membrane. Therefore, it seems likely that there are proteins other than Myo2 that are able to bind to actin filaments and link them to the membrane.
A recent model of the fission yeast AMR in a 3D cylindrical geometry also attempted to model individual heads of Myo2 that interact with individual actin filaments, in a similar way to the later iteration of our model, except with even greater detail in the description of the myosin molecules [133]. The authors simulated a number of potential AMR structures, but in all of these the only linker between actin filaments and the membrane was the simulated Myo2 molecules, and despite this they did not see the peeling away of actin filaments that we would expect to see from our simulations. In these simulations, they use an actin bead friction coefficient 30 times greater than that used in previous simulations, including our own [133]. This value is used without justification, and additionally the beads in actin filaments are now placed only 5.5 nm apart, which gives a friction/length around 500 times greater than that used in previous simulations (~1000 pN s/μm2 vs. 2 pN s/μm2). This will significantly slow down the speed with which actin filaments move, since the pulling force exerted by the myosin molecules was the same as in our simulations, which could mean that filaments get disassembled by cofilin before they are able to be pulled out of the ring. When we attempted to implement our own ring model where we explicitly include individual pairs of Myo2 heads, we found that very few actin filaments were captured and pulled on by the Myo2 heads, even when we biased the polymerisation direction of actin filaments to be parallel to the ring, and this lead to a very low ring tension (Figure 4.6E, Figure 4.6F). This is despite our observation that this description of the myosin clusters/nodes worked quite well in a toy model (Figure 4.6A– Figure 4.6D). In the toy model, the actin filament was placed at an
optimum height above the cluster, which may have helped the