A Molecular View of Self-Assembly and Liquid-Liquid Phase Separation of Intrinsically Disordered Proteins

Intracellular compartmentalization of biomolecules into non-membrane-bound compartments, commonly referred to as membraneless organelles (MLOs), has been observed for over a century. The past decade has seen a massive surge of research interest on this topic due to evidence that a liquid-liquid phase separation (LLPS) process is responsible for the assembly of biomolecules into liquid-like compartments constituting MLOs. Since the initial discovery, dozens of cellular systems have been explored with this in mind, and have also been shown to have liquid-like properties, owing several unique functions to their liquid-like nature. MLOs such as stress granules may spontaneously form in response to cellular stress, while others such as the nucleolus may form multi-layer architectures that accelerate multi-step assembly processes, similar to an assembly line. The ability of biomolecules to undergo LLPS has been largely attributed to the presence of disordered proteins and nucleic acids. Intrinsically disordered proteins (IDPs) are proteins which lack a native, folded structure while remaining physiologically functional are able to promote LLPS due to their polymeric nature which allows for transient multivalent interactions between many amino acids. In this thesis, I work toward a greater understanding of the relationship between an IDP's sequence, and its ability to undergo LLPS. Using a combination of all-atom and coarse-grained simulations, I make several important contributions of significant and general interest to the field of IDP-driven LLPS. I start by developing a coarse-grained modelling framework which explicitly represents amino acid sequences, and is the first of its kind to directly simulate phase coexistence of IDPs at sequence resolution. I then leverage this model to demonstrate the relationship between a single IDP chain, and a condensed phase of the same IDP, showing that one can predict conditions where LLPS will be possible, simply by observing the single-chain behavior and infinitely-dilute two-chain binding affinity. I then provide a rationalization of the thermoresponsive behavior of some proteins which undergo lower critical solution temperature (LCST) phase transitions, and how amino acid composition can lead to different thermoresponsive behaviors. Finally, I present atomic-resolution simulations showing the different interaction modes responsible for driving LLPS of two particular IDPs.