A Possible Mechanism for Phase Separation during Cell Signaling

Dr. James L. ThomasDepartment of Physics and Astronomy

G. Matthew FrickeDepartment of Computer Science

Gold Group, May 17th, 2004

University of New MexicoFunded by the Center for Advanced Studies

Immune System Signaling (abridged)

Invader

Invader Invader

Invader

Macrophage

Macrophage encounters an invader

Immune System Signaling (abridged)

InvaderInvader

Invader

Macrophage

Macrophage destroys the invader

Immune System Signaling (abridged)

InvaderInvader

Invader

Macrophage

Macrophage displays proteins from the invader

Immune System Signaling (abridged)

Macrophage T-Cell

T-Cell

T-Cell

T-Cell are activated by the macrophage if their receptors match the displayed proteins

we care about this…

Immune System Signaling (abridged)

Macrophage T-Cell B-Cell(Memory)

B-Cells confirm that the proteins are from a pathogen encountered previously.

…and we care about this

B-Cell(Plasma)

Immune System Signaling (abridged)

T-Cell

InvaderInvader

Invader

B-Cells change state and emit antibodies that bind to the invader

Immune System Signaling (abridged)

Macrophage

Macrophage

Macrophage

Macrophage

Macrophage

InvaderInvader

Invader

Invader

Invader

Macrophages are recruited to the antibodies

Immune System Signaling (abridged)

Macrophage

Macrophage

Macrophage

Macrophage

Macrophage

Invader

Recruited macrophages destroy the invaders and display their proteins.

The cycle repeats and the immune response increases exponentially.

Foreign protein snippets are displayed by macrophage

Receptors that match particular protein snippets are displayed by T-cells.

In vivo, T cells are stimulated by monovalent binding to ligandson “antigen presenting cells” (e.g. Macrophages and B-Cells)

In vivo, T cells are stimulated by monovalent binding to ligandson “antigen presenting cells” (e.g. Macrophages and B-Cells)

The macrophage comes into contact with various T-cells.

In vivo, T cells are stimulated by monovalent binding to ligandson “antigen presenting cells” (e.g. Macrophages and B-Cells)

If the T-cell receptors match the protein snippet held by the ligands then they bind…

… and the ligands cluster together and the receptors cluster together. How?

Our hypothesis: Cross Membrane binding between ligand-receptor pairs serves to combine the attractive forces between proteins in their own membranes. This would allow receptor or ligand groups that by themselves are do not cluster to “sum” the attractive forces and cluster.

Sounds simple but we can’t predict how strong the inter-membrane force needs to be in relation to the intra-membrane forces to cause phase separation. So model it!

Our Approach� Write a model of phase separation on a single

membrane

� Confirm that our results match those of previous phase transition models

� Implement two copies of the single membrane model and bring them into contact

� Add a cross-membrane binding force

� Under what circumstances do we get phase separation?

Protein Random Direction

Metropolis Monte Carlo• n x n toroidal lattice

• Each site on the lattice can hold a single protein

• At each discrete time-step all proteins choose a random direction to move

• If the energy is reduced the motion is accepted.

• Otherwise the motion is accepted with probability .

• Repeat until we are confident that the system is in equilibrium= Favorable contact energy (in kT)

between neighboring proteins.

ε)/( kTe ∆Ε−

“Freeze Out:” the problem that isn’t?

Measuring Phase Separation – Spatial

Autocorrelation

• Autocorrelation Function g(d)

• Choose a protein and count the number of proteins at distance d (then 4 .)

d

÷

0

50

100

150

200

250

1 3 5 7 9 11 13 15 17 19 21 23 25

d (0...width/2)

g(d)

00.61.3

proteins are highly correlated ( supercritical)

proteins are randomlydistributed ( subcritical)

Correlation Functions for Three Values of

80

60

40

20

20151050

Fit Correlation Functions to an Exponential (fit deteriorates as critical epsilon reached)

120

100

80

60

40

20

20151050

30

25

20

15

10

5

20151050

80

60

40

20

0

20151050

d

g(d)

= 0.3 = 0.7

= 1.0 = 1.2

xCeCy 21=

Fit exponential length constants to 21 C

C

c

+−εε

*

*Gould H., and J. Tobochnik An Introduction to Computer Simulation Methods: Applications to Physical Systems, 1996

3 .5

3 .0

2 .5

2 .0

1 .5

1 .0

0 .9 50 .9 00 .8 50 .8 0

Epsilon

Leng

th c

onst

ant

This successful fit confirms that our model matches previous work on phase separation

Finite Size Effects

Two Membrane Model

Calculating the autocorrelation, exponents, and cri tical exponents is too slow

Instead: calculate the protein density for all over lapping 3x3 squares on the lattice

Standard deviation is a measure of phase separation

Low σσσσ, one phase High σσσσ, two phase

Density variance as a measure of phase separation

ϕεε === zyx c ,,

Complete phase separation occurs at 0.27

Random protein distributions have been observed to have z values of between 0.09 and 0.105

0 2 4 6 8 10 12 14 160.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.305

0.2590.259

0.259

0.2120.212

0.212

0.1660.166

0.166

0.119

Contour Plot of Phase Separation

,,x y zCross membrane εεεε

Intr

amem

bran

eεε εε

The Monte Carlo model confirms our hypothesis that binding between ligand-receptor pairs can cause two groups of proteins to phase separate even when their attractive forces are too low individually.

The model has also provided a picture of how strong the cross membrane binding needs to be in relation to the intra-membrane forces.

We have provided the simplest possible physical explanation for protein phase separation. (Satisfying from a “Occam’s Razor” point of view).

Conclusions

Further Work� The model was written in an object oriented

style to facilitate prototyping.

� In order to map the 2-d inter intra epsilon space in detail the model needs to be faster

� The predictions about ligand-receptor binding strength need to be grounded against observations of real biological systems.