Properties of Stress Granule and P-Body Proteomes

Abstract

Stress granules and P-bodies are cytosolic biomolecular condensates that dynamically form by the phase separation of RNAs and proteins. They participate in translational control and buffer the proteome. Upon stress, global translation halts and mRNAs bound to the translational machinery and other proteins coalesce to form stress granules (SGs). Similarly, translationally stalled mRNAs devoid of translation initiation factors shuttle to P-bodies (PBs). Here, we review the cumulative progress made in defining the protein components that associate with mammalian SGs and PBs. We discuss the composition of SG and PB proteomes, supported by a new user-friendly database (http://rnagranuledb.lunenfeld.ca/) that curates current literature evidence for genes or proteins associated with SGs or PBs. As previously observed, the SG and PB proteomes are biased toward intrinsically disordered regions and have a high propensity to contain primary sequence features favoring phase separation. We also provide an outlook on how the various components of SGs and PBs may cooperate to organize and form membraneless organelles.

Integrated multi-omics reveals common properties underlying stress granule and P-body formation

“…A number of studies have highlighted that PBs and SGs have different components, leading to the suggestion that they are functionally distinct (Youn et al, 2019). SGs act to store mRNAs that can re-enter the translationally active pool of mRNAs, whilst PBs were originally thought to play a key role in mRNA decay.…”

“…Low complexity protein domains and RNA serve as key determinants of the intracellular phase separation necessary for the formation of PBs and SGs (Weber & Brangwynne, 2012). Hence, both PBs and SGs are known to be enriched for proteins containing RNA-binding domains and intrinsically disordered regions (IDRs) (Youn, Dyakov et al, 2019). IDRs are protein domains usually containing stretches of low sequence complexity which have been implicated in the formation of stress-induced RNP granules, as well as in the aggregation of unproductive and cytotoxic protein forms, such as amyloid protein (Gilks, Kedersha et al, 2004).…”

Prediction of liquid-liquid phase separation proteins using machine learning

“…published a comprehensive database containing the proteins related to the formation of stress granules and P-body condensates. Protein was each assigned to tier 1-4 according to the degree of confidence in evidence supporting protein localization to stress granules or P-body (4385 protein sequences) [21]. We predicted these data by using PSPredictor and compared the results with those of the reported PSP prediction tools.…”

“…Considering the uncertainty of the negative dataset and that most of the first generation tools only returns prediction scores, it is necessary to set a reasonable PSP/Non-PSP separation threshold Previous tools set this threshold by using the whole proteome as a background, for example, top 1.8% [13], 2% [21] and 20% [13], scored proteins in whole proteome were regarded as PSPs and were regarded as Non-PSPs if not. In this work, for the positive dataset F1+, we used a recall-receive curve to analyze the recalls under different threshold of top n percent of the proteome scores (n was between 0 and 25%).…”

“…Development of PSP prediction tools by using machine learning have been hampered due to lacking the collection of experimentally determined PSPs data. Recently, a number of PSP databases have been published [21][22][23], they set the ground for the birth of more general PSP prediction tools. In particular, Li et al curated a LLPS database (LLPSDB) [24] by extracting LLPS experimental results from literatures.…”