A physics-based approach to predict protein phase separation under heat stress

A key question in understanding intracellular organization is how environmental stimuli affect behavior of biomolecules. When certain proteins inside the cell are subject to temperature perturbations, they phase separate to form condensates. Some proteins exhibit upper critical solution temperature (UCST) behavior, forming condensates at low temperatures, whereas others exhibit lower critical solution (LCST) behavior, condensing at higher temperatures. Computational approaches, specifically coarse-grained molecular dynamics simulations have proven useful in elucidating molecular underpinnings of condensate dynamics. While some approaches have recapitulated UCST behavior, capturing LCST behavior quantitatively has been significantly less well studied. To gain accurate and efficient insight into temperature-driven condensate behavior, we require a coarse-grained model that is trained to capture LCST behavior, for both single and multiple systems of protein chains. We have developed Mpipi-T, a quantitatively accurate residue-level coarse-grained model to predict temperature-driven protein condensation. By optimizing Mpipi-T using atomistic simulations and experimental data, we are able to capture LCST and UCST behavior of proteins. Furthermore, we have used Mpipi-T to elucidate thermodynamic properties of heat stress induced protein phase separation. Our model can now be used as a tool to uncover the biophysics of how cellular systems may respond to heat stress.