Demixing of distinct ribonucleoprotein condensates is under dynamical control.

Abstract

Whi3 is an RNA binding protein from Ashbya gossypii that phase separates to form RNA-protein condensates. Observations in cells show that certain RNA molecules co-localize with Whi3 and other RNA molecules into a single condensate, while some others form apparently demixed condensates. Is a there a thermodynamic basis for different ternary combinations of Whi3 and RNA molecules to form demixed condensates? Simulations show that RNA molecules must have strong repulsive interactions for one another, or strong homotypic interactions if demixing is to be thermodynamically feasible. We tested this using in vitro reconstitutions to quantify the phase behaviors of binary and ternary mixtures. A novel analysis of the measured phase boundaries show that heterotypic protein-RNA interactions are the main drivers of phase separation. This was further tested by deletion of cognate binding sites on RNA molecules, showing that there are several ways for Whi3 to associate with RNA molecules. Investigations of phase behaviors internary mixtures show that demixed condensates are dynamically arrested phases. Therefore, in vitro studies suggest that demixing is likely to be under dynamical rather than thermodynamic control. To test for this in live cells, we generated a mutant where differentially expressed RNAs are constitutively expressed under a single promoter. We find that RNA molecules that demix in a wildtype background formed colocalized condensates when they are expressed concomitantly. Our studies show that the non-equilibrium aspects of phase separation, specifically the ability to form dynamically arrested phases, can give rise to demixed condensates that achieve long-livedspatial segregation of RNA molecules within distinct condensates.