Summary of my work at EMBL in May 2012

J. Medina July 9, 2012

Introduction

The nal goal of these simulations and analyses is the understanding of the forces that play a role in the separation of the centrosomes during prometaphase in yeast cells. Firstly, a simulation of the motors coupling to the bers in an aster is developed, taking into account the dynamic instability of the bers too. These results give us information of the density of the motors an the bers in the cell as a function of the distance to the center when the steady state is reached. In the second part we will use those conclusions to study how the two centrosomes pull each other. In order to calculate the forces in the centrosomes we will need to apply motion equations for very viscous media. These conditions and the geometry of the problem along with the probability density will show the dependence of the separation forces as a function of the properties of the aster and the motors.

2
2.1

One aster simulation

Description of the algorithm

The movement of the bers and the motors in the one aster problem is simulated in C++. The aster is formed by nf radial bers that have dynamic instability. Fibers do not have an upper limit for their length; on the other hand if they shrink until zero length another ber starts to grow in another random direction, so the number of bers is always constant. The movement of each ber is ruled by its growth and shrink rates and the catastrophe and rescue frequencies. From those values we can compute theoretically the distribution of the lengths of the wires and we can check it numerically in the simulation. Motor proteins are represented by many small point particles which follow a random brownian movement. Particles may bind to the bers with a nite probability if they are close enough to them, in the so-called evaluation zone. When they are bound to a ber they start their movement towards the center of the aster, where they stop if they get to there. Bound particles also have a nite probability to unbind and go back to the free brownian movement, even when they are already in the center of the aster. If the ber in which a particle is attached shrinks faster than their speed towards the center, then that particle automatically unbinds when gets to be out of the ber. After some time the steady state is reached, and there is a distribution of the density of the particles as a function of their distance r to the center that we can measure. That distribution will depend on the number of bers, the temperature (represented by the size of the random steps for the particles in their brownian movement), the size of the evaluation zone, the probabilities of binding and unbinding and the speed of the bound particles towards the center. At time 0, all the nf bers have a random initial length around 2 and they are randomly assigned whether they are initially growing or shrinking. In each iteration, every ber grows or shrinks one step and then checks if it is the time to change the state. The time in which a growing ber will have a catastrophe (or a shrinking one to have a rescue) is selected on an exponential probability distribution following Gillespies algorithm (Appendix A). In each iteration every ber checks if that time has passed. When it has, the ber changes its state for the next iterations and picks a time among the distribution of the state it turns in. When a ber shrinks to zero and starts to grow in some other direction it also picks the time in which it will shrink again from the catastrophe time distribution. Most of the particles in each iteration will just move a random step in the x and y directions. The exceptions are the following:

 For every particle, an algorithm evaluates whether it is in any of the regions that are close to the bers or not (evaluation zones). If the particle is close a random number between 0 and 1 is generated. A threshold with the binding probability of the particles in the closer zones is previously set. If the random number of the particle lies below that threshold it moves to the closer point of the ber and it attaches to it. Then they also pick the time in another exponential distribution in which they will unbind. Particles that are already bound to a ber move a step to the center of the aster with a previously dened speed. If they get to the center they just stop there, so there is always accumulation of them in the middle in the steady state. As the bers, they also check for each iteration of it is their time to change the state (unbind, in this case). If it actually is, then they unbind and become free particles in one of the evaluation zones and may or may not bind again to the ber.

2.2

Results

The distribution of the length of the bers and the density of particles in the steady state are interesting. For the bers, it is known that it is exponential with an expected value L is: v+ v (1) L = v+ f v f as long as it takes a nite positive value, in which v+ and v are the growing and shrinking velocities and f and f are the catastrophe and rescue frequencies. That exponential distribution has to be normalized. Therefore, the probability of nding the tip of a ber at a distance r and in the angle is: Pt (r) = 1 e L
L r

dr

(2)

For the nal goal of the problem it is more interesting to compute the probability that there is a piece of the ber at a distance r instead of where is the tip. Therefore:

Pf (r) =
r

Pt (r ) = e

L r

(3)

Some simulations without particles were done and they agreed with the predicted formula, as shown in the plot (Fig. 1). The distribution of the density of particles as a function of the distance in the steady state is also simulated (Fig. 2). As expected, it has a sharp peak closer to the 0 when nf , the evaluation zone, the probability of binding and the speed of the particles to the center are high and a softer and further peak when the unbinding probability, the movement of the free particles and the expected length of the bers increase. The theoretical formula is not easy to infer as other factors might play a role in a secondary order. This is the case of the shrinking speed that may produce that more particles unbind in their way to the middle. If temperature raised in the system the change in the value of some factors as the movement of the free particles or the unbinding probability might be related. In this simulation they are considered independent.

Figure 1: Histogram of the number of ber tips as a function of the distance to the center.

Figure 2: Histogram of the density of particles as a function of the distance to the center.

3
3.1

Two asters problem

Theoretical calculations

Lets analyze the case in which there are two asters. Two bers from dierent asters may bind with each other and produce forces mutually that will separate the centrosomes. Consider the static problem in which both centrosomes are separated by a distance 2d. A xed force F between the bers is produced at a negligible distance from the crossing point of the bers (x0 , y0 ). The longitudinal part of the force will move the centrosomes in the direction of the ber and the perpendicular component will produce a torque that rotates the ber around the center (Fig. 3).

Figure 3: Diagram of the problem.

As the medium is very viscous we can consider negligible the acceleration. The total longitudinal force that the left centrosome would suer is: F = F cos t v1 = 0 v1 = F cos t (4)

where t is the viscosity coecient for a moving sphere. The variables are dened as: t = 6R =
1 +2 2

(5) (6)

being the viscosity of the medium. The viscous force on the ber is negligible compared to the force on the sphere. The torque produces the centrosome to turn at an angular speed 1 dened as: | | r1 F sin r 1 = 0 1 = r1 F sin r (7)

in which r = 8R3 . All these calculations are analogous for the centrosome 2.

3.2

Computation and results

For any set of xed parameters we can represent these v1 and 1 in color plots as in Figs. 4, 5.

Figure 4: Plot of the modulus of the velocity of the centrosome when the binding of the bers occurs in the pixel (x, y) in arbitrary units. The direction would be always longitudinal to the ber, i.e., the line between the pixel and the center of each centrosomes.

Figure 5: Plot of the modulus of the angular velocity of the centrosome when the binding of the bers occurs in the pixel (x, y) in arbitrary units.

For any of the points in the plots there is a nite probability that a piece of a ber from the rst centrosome and a piece of a ber from the second one are in that point and then might bind. For this 2-dimensional geometry, this probability is proportional to: nf 1 e P (both pieces at (x, y)) r1
r1 R L

nf 2 e r2

r2 R L

r1 +r2 2R L

r1 r2

(8)

setting the number of bers of each aster to 1. We compute its value on every point of the rectangular grid and normalize (Fig. 6).

Figure 6: Normalized probability of nding two pieces of ber from the dierent asters in evry point. This is computed for a much larger area for the assumption that we considered all the space.

Therefore for a set of xed conditions we can compute the expected velocity v of the centrosome by integrating the product of the velocity and the probability in the whole grid, in the supposition that they will actually bind and just once. This number can be plotted as a function of the dierent parameters of the cell and the centrosomes in order to nd the best conditions for the centrosomes to separate (Figs. 7, 8).

A Computation of the time to change the state based on Gillespies algorithm

One way of determining when will an stochastic process may happen would be to compute a random number in each iteration and compare it with a previously set threshold. This method has several problems, as the computational eort to compute those numbers for every particle and iteration, the discretization of the model and the arbitrariness of the threshold, which cannot be compared with any frequency.

Figure 7: Expected value of v as a function of the expected length of the ber L for 2d = 2.

Figure 8: Expected value of v as a function of the separation of the centrosomes 2d for L = 2.

Using the Gillespies algorithm for problems of changing between binary states gives a better solution. Let be a reaction in which a particle can be in a state either 0 or 1, and the frequency of changing is f01 . The distribution of times in which the particles go from 0 to 1 is: f (t) = ef01 (9)

The range of this distribution is (0, 1]. In order to compute random numbers that follow it, the inverse function for is computed and a random number n from 0 to 1 is generated: = log n ; n (0, 1] f01 (10)

After this time the particle will enter in the state 1 and another will be generated with the same algorithm but for the frequency f10 . This algorithm is used for the time of unbinding of bound particles and the catastrophe and rescue times of the bers.

References