ABSTRACT
Biomolecular condensates are compositionally-diverse organelles that reversibly assemble in the cytoplasm or nucleoplasm to localize cellular functions. They form by liquid-liquid phase separation and have no bounding lipid membrane. Experiments have shown that biomolecular condensates have a wide range of functions, and loss of cellular control is associated with chronic neurodegenerative diseases. Their main constituents are intrinsically-disordered proteins that are conformationally flexible and possess weak binding sites for proteins or RNA. Although the composition of many experimental condensates is increasingly clear, the quantitative connection between their structure and the molecular properties of their constituent proteins is still obscure. We use coarse-grained molecular simulations to explore the phase behaviour of a model biomolecular condensate and its dependence on its constituent intrinsically-disordered proteins. The proteins are represented as semi-flexible polymers with attractive end-caps. They spontaneously condense into fluid networks in which their end-caps reversibly bind at junctions. The spatial separation of the junctions scales with the polymer backbone length as a self-avoiding random walk. The network stability and structure are more sensitive to the separation of the end-caps than their affinity. This sensitivity to the binding site separation suggests that post-translational modifications or interactions with other proteins that modify the conformational fluctuations of a disordered protein will regulate its transition into a condensed phase. Additional proteins will be recruited only if their conformational fluctuations allow them to fluctuate between the existing junctions of the network. Cells may use this sensitivity to regulate assembly and composition of biomolecular condensates, and it provides a promising route towards therapeutic interventions