**6. Conclusion: Repeats are a general characteristics of prion-like domains**

In this study, we have analyzed RNA-binding proteins with prion-like domains, such as TAF15, FUS, and EWS. Using the FoldAmyloid program [77], we revealed the existence of amyloidogenic regions in the RRM domain in all three proteins of the FET family. This allows us to suggest that when the binding to RNA is violated, the proteins can aggregate spontane‐ ously forming amyloid fibrils. In this case, protein FUS can aggregate both upon the removal of RRM and at point mutations in this domain [65, 66]. Further studies should be conducted to clarify whether aggregation of protein FUS in the presence of RRM does not lead to the formation of irreversible pathologic aggregations and functions only by the mechanism of assembly and possible disassembly under favorable conditions analogous to the functioning of SGs. We have also found that in all members of the FET/TET family the zinc-finger motif does not correspond to the classical one because it concerns both the location of tandem repeats and the amino acid sequence. Moreover, according to our prediction, in the FUS and TAF15 proteins, it does not even form any structure. Based on this, we can propose that in proteins of the FET/TET family the zinc-finger motif is not responsible for the DNA binding, but performs other functions determination of which requires further investigations.

We have also predicted unstructured regions corresponding both to prion-like domains and to additional regions, where we have revealed several recurrent amino acid motifs—tandem repeats. In addition, these proteins are characterized by the occurrence of homorepeats when one amino acid is multiply recurrent. The length of homorepeats can affect not only the capacity to aggregate but also the toxicity in an aggregated state. Correspondingly, the larger is the number of repeats, the higher is the probability that the aggregated protein containing them is associated with the development of the disease. We have established that protein FUS contains five disordered patterns from the pattern library obtained from the protein data bank: GSHM, GGGGSGG, GGGGG, GGSGGGGSGGG, and RGGGGSG, which are not located in the prion-like domain predicted for FUS. For other members of this family, these patterns are not characteristic and, as known, are less toxic when compared to FUS. When a larger part of the prion-like domain in protein FUS was removed (100 residues of the 165), it retained its capacity to aggregate [65], but lost completely the ability to form gel and display toxicity. These results allow us to suggest that the presence of short repeats in the unstructured prion-like domain of RNA-binding proteins is required for fast formation of a dynamic cross-beta structure of SGs [84].
