A liquid-to-solid phase transition of biomolecular condensates drives in vivo formation of yeast amyloids and prions

Liquid-liquid phase separation (LLPS) and liquid-solid phase transition (LSPT) of amyloidogenic proteins are now being intensively studied as a potentially widespread mechanism of pathological amyloids formation. However, the possibility and importance of such a mechanism in living systems is still questionable. Here, we investigated the possibility of such LSPT for a series of yeast prion proteins-based constructs overproduced in yeast cells lacking any pre-existing amyloid template. By combining fluorescence and electron microscopy with biochemical and genetic approaches, we have shown that three such constructs (containing the prion domains (PDs) of either Sup35, Rnq1 or Mot3 proteins) form amyloid fibrils via the intermediate stage of liquid-like condensates, that age over time into the more solid-like hydrogels and amyloid bodies. In turn, LSPT of these constructs triggers prion conversion of the corresponding wild-type protein. Two other constructs studied (Ure2- and Sap30-based) are unable to phase separate in vivo and their amyloidogenesis is therefore strongly suppressed. Using PrK-resistant amyloid core mapping, we showed that Sup35PD amyloids formed via LSPT have a different molecular architecture compared to those formed via amyloid cross-seeding. Finally, we showed that physiological LLPS of wild-type Sup35 protein can increase its prion conversion in yeast.