Phase Transitions in Biological Systems with Many Components

Abstract

Biological mixtures such as the cytosol may consist of thousands of distinct components. There is now a substantial body of evidence showing that, under physiological conditions, intracellular mixtures can phase separate into spatially distinct regions with differing compositions. In this article we present numerical evidence indicating that such spontaneous compartmentalization exploits general features of the phase diagram of a multicomponent biomolecular mixture. In particular, we show that demixed domains are likely to segregate when the variance in the intermolecular interaction strengths exceeds a well-defined threshold. Multiple distinct phases are likely to become stable under very similar conditions, which can then be tuned to achieve multiphase coexistence. As a result, only minor adjustments to the composition of the cytosol or the strengths of the intermolecular interactions are needed to regulate the formation of different domains with specific compositions, implying that phase separation is a robust mechanism for creating spatial organization. We further predict that this functionality is only weakly affected by increasing the number of components in the system. Our model therefore suggests that, for purely physico-chemical reasons, biological mixtures are naturally poised to undergo a small number of demixing phase transitions.

Figure 1

Representative phase diagrams and free-energy landscapes of two-component mixtures. In the constant-temperature and pressure phase diagrams drawn in (a)–(c), single phase regions are shown in white; two-phase coexistence regions in light gray; and a three-phase coexistence region in dark gray. The components have concentrations ${\rho }_1$ and ${\rho }_2$; dotted lines indicate constant compositions; and dashed lines indicate example tie lines connecting coexisting phases (circles). The red dotted line indicates an equimolar parent composition, while the red dashed line indicates the tie line at the boundary of the equimolar homogeneous phase. The angle of phase separation, θ, is the angle between the highlighted parent composition and the tie line at the cloud point (solid circles). For each phase diagram, the corresponding (d)–(f) depicts the free-energy landscape for a mixture with a parent concentration lying on the highlighted tie line. In these landscapes, the chemical potentials of the components are fixed, and the component concentrations fluctuate among two or more free-energy basins. In (c) and (f), the highlighted tie line is in a two-phase region, but the appearance of a metastable phase on the free-energy landscape indicates close proximity to a three-phase coexistence region. To see this figure in color, go online.

Figure 2

Histograms showing the bimodal distributions of the angle of phase separation in simulations with (a) 32 and (b) 64 components. The histogram for each random-mixture ensemble was constructed from all high-concentration free-energy basins at the cloud point (see text). Condensation into two phases with equal compositions is associated with a small angle of phase separation, θ, whereas demixing into phases with dissimilar compositions occurs when θ approaches ${\theta }_N$, the angle of phase separation corresponding to the demixing of a single component. In (b), the $\sigma =\overline{\mathit{ϵ}}/2$ ensemble has nearly equal probabilities of condensation and demixing. To see this figure in color, go online.

Figure 3

The probability of multiphase coexistence correlates with the angle of phase separation at the cloud point. (a) Increasing the number of components in a solution suppresses demixing and leads to a single condensation phase transition, regardless of the variance of the interactions. Each point represents an average over an ensemble of random mixtures; error bars indicate the standard deviation of the distribution of angles for each ensemble. The dashed line roughly indicates where the crossover from demixing to condensation occurs. (b) The probability of observing more than two free-energy basins at the cloud point, $p\left(n_{\text{basins}}>2\right)$, indicating close proximity to a multiphase coexistence region. Demixing transitions are associated with the presence of multiple coexisting phases. Vertical error bars indicate the error due to finite sampling. To see this figure in color, go online.

Figure 4

A mean-field theory predicts bimodal phase behavior for random mixtures. (a) An illustration of a 2×2 submatrix (red squares) selected by the vectors ${\mathit{\tau }}^{\left(\mathbf{2}\right)}$ and ${\mathit{\upsilon }}^{\left(\mathbf{2}\right)}$. (b) The mean-field free-energy difference between the most stable phase comprising $n^2$ distinct interaction energies and the equal-composition phase with N components, under the constraint that both phases have the same overall concentration ϕ. Numerical evaluations of Eq. 5 are averaged over many realizations of the random matrix $\mathbf{\Delta }\mathrm{ϵ}$; for this illustration, we have chosen ${\beta }^{\prime }\sigma =1.05$. The conditions for the crossover regime can be found by tuning N and ${\beta }^{\prime }\sigma$ to equalize the free energies of the most stable demixed phase $\left(n\simeq 1\right)$ and the condensed phase $\left(n\simeq N\right)$. (c) The mean-field phase diagram as a function of the control parameters N and ${\beta }^{\prime }\sigma$. The coexistence curve between condensation and demixing scales approximately as $\sqrt{\text{ln}N}\sim {\beta }^{\prime }\sigma$. To see this figure in color, go online.