Tandem RNA binding sites induce self-association of the stress granule marker protein TIA-1.

TIA-1 is an RNA-binding protein that sequesters target RNA into stress granules under conditions of cellular stress. Promotion of stress granule formation by TIA-1 depends upon self-association of its prion-like domain that facilitates liquid-liquid phase separation and is thought to be enhanced via RNA binding. However, the mechanisms underlying the influence of RNA on TIA-1 self-association have not been previously demonstrated. Here we have investigated the self-associating properties of full-length TIA-1 in the presence of designed and native TIA-1 nucleic acid binding sites in vitro, monitoring phase separation, fibril formation and shape. We show that single stranded RNA and DNA induce liquid-liquid phase separation of TIA-1 in a multisite, sequence-specific manner and also efficiently promote formation of amyloid-like fibrils. Although RNA binding to a single site induces a small conformational change in TIA-1, this alone does not enhance phase separation of TIA-1. Tandem binding sites are required to enhance phase separation of TIA-1 and this is finely tuned by the protein:binding site stoichiometry rather than nucleic acid length. Native tandem TIA-1 binding sites within the 3' UTR of p53 mRNA also efficiently enhance phase separation of TIA-1 and thus may potentially act as potent nucleation sites for stress granule assembly.

Micellar TIA1 with folded RNA binding domains as a model for reversible stress granule formation.

TIA1, a protein critical for eukaryotic stress response and stress granule formation, is structurally characterized in full-length form. TIA1 contains three RNA recognition motifs (RRMs) and a C-terminal low-complexity domain, sometimes referred to as a "prion-related domain" or associated with amyloid formation. Under mild conditions, full-length (fl) mouse TIA1 spontaneously oligomerizes to form a metastable colloid-like suspension. RRM2 and RRM3, known to be critical for function, are folded similarly in excised domains and this oligomeric form of apo fl TIA1, based on NMR chemical shifts. By contrast, the termini were not detected by NMR and are unlikely to be amyloid-like. We were able to assign the NMR shifts with the aid of previously assigned solution-state shifts for the RRM2,3 isolated domains and homology modeling. We present a micellar model of fl TIA1 wherein RRM2 and RRM3 are colocalized, ordered, hydrated, and available for nucleotide binding. At the same time, the termini are disordered and phase separated, reminiscent of stress granule substructure or nanoscale liquid droplets.