Tomer Lapidot*, Saranya Santhosh Kumar*, Yevgeniy Serebrenik, Stephanie Sansbury and Ophir Shalem
Poster # 4
RNA binding proteins (RBPs) play a crucial role in the reorganization of cellular milieu upon stress. Most well-known is the formation of stress granules, transient ribonucleoprotein structures that sequester mRNAs to quickly inhibit translation during acute stress. While stress granule formation is considered a protective physiological process, impaired assembly and disassembly is linked to many neurodegenerative diseases. Despite the importance of this regulatory process, it is usually studied using a small number of marker proteins such as G3BP1 and TIA1, as there are currently no technologies that enable profiling of subcellular localization of hundreds of proteins. Previous studies employ proximity labelling techniques to identify RBPs that participate in stress granule formation with a specific protein marker. However, such techniques do not provide insight regarding the spatial localization of RBPs in the cell and are not designed to detect stress granules of RBPs that can form in the absence of the designated protein marker. In this work, we present scalable mapping of the subcellular spatial dynamics of RNA binding proteins using pooled gene tagging. High throughput imaging and in-situ sequencing are used to follow the dynamic localization of hundreds of RBPs across normal and stress conditions. Using this approach, we were able to accurately quantify the co-localization between G3BP1 and hundreds of RBPs to map components of canonical stress granules in intact cells. We further analyze images using deep learning variational autoencoders to classify RBP localization patterns into discrete localization classes which provide a bird's-eye-view of RBP dynamics across diverse stress conditions and identify non-canonical stress associated ribonucleoprotein granules.
Comentários