COMPARISON AND EVOLUTIONARY ANALYSIS OF PATTERNS CONTROLLING LIQUID-LIQUID PHASE SEPARATION AND PRION FORMATION BY A FUNGAL PROTEIN

Some proteins, including yeast translation termination factor Sup35 (eRF3) are capable of both stress-induced liquid-liquid phase separation (LLPS) and formation of solid fibrous aggregates (amyloids). Fragmentation and propagation of amyloid fibrils generates transmissible (in yeast, heritable) self-perpetuating protein agents, termed prions. Relationships between these processes are still poorly understood. Previous literature data suggested that the ability of Sup35 orthologs to form a prion is sporadically distributed in fungal evolution, and depends on amino acid composition of Sup35 prion domain (PrD), rather than on a evolutionarily variable specific sequence.

We have studied two groups of proteins: 1) fungal Sup35 PrDs of various evolutionary origins, and 2) artificially synthesized “scrambled” variants of Saccharomyces cerevisiae Sup35 PrD, that possess identical amino acid composition but different sequences. These proteins were fused to fluorophores and expressed in S. cerevisiae cells. LLPS and amyloid/prion formation were assessed by fluorescence microscopy and biochemical approaches. Amino acid sequences were analyzed by various computational algorithms.

Our data indicates that propagation of prion state strongly depends on the evolutionary distance from the host. In contrast, majority of proteins studied are capable of both LLPS and ability to form amyloid fibrils. These capabilities are associated with specific patterns of PrD amino acid distribution, that are broadly conserved among fungi. Notably, PrDs of different sequences differ from each other by their ability to convert from liquid condensates to amyloids, and relationship between these processes is apparently optimized in evolution. Moreover, heterotypic PrDs are can colocalize with each other within liquid condensates and influence amyloid conversion by each other.

To conclude, LLPS and amyloid properties depend on specific evolutionarily conserved sequence patterns, indicating possible important biological roles for these processes. These patterns could potentially be used to predict LLPS and prion potential in other sequence contexts.

This work was supported by NSF grant 2345660.