A Chemical Biology Toolkit to Understand and Target Biomolecular Condensates

Abstract

Cells organize functional processes into compartments to survive. Biomolecular condensates are unique compartments in that they can locally concentrate components to execute specific functions while lacking a surrounding membrane. As such, condensates dynamically respond to changes caused by stress, disease states, cell cycle stages, and drug treatments. These responses are reflected in microscopically-visible changes in condensate morphology; however, the link between a condensate's (dys)function and its morphology change is not well understood. The aim of this work was to understand the structure-function relationship of the nucleolus, the cell's largest condensate and center of ribosome biogenesis. Specifically, we used a set of small molecule drugs to selectively inhibit different steps of ribosome assembly and observed distinct changes in nucleolar architecture by fluorescence microscopy. We then utilized this data as a training set to build a first-in-class deep neural network that accurately classifies drug-induced nucleolar morphology changes. Importantly, we demonstrated that the extent of nucleolar morphology disruption caused by these drugs can be quantified and used to predict the degree to which they perturb specific nucleolar processes. Next, we conducted a pilot drug screen to identify novel nucleolar phenotypes and nucleolar interaction networks. Given that nucleolar morphology and ribosome production are dysregulated in disease, current efforts are focused on clinically relevant applications of this technology. Together, this work demonstrates that automated imaging and deep learning-assisted analysis of condensate perturbations by small molecules can be used as a powerful discovery platform for new biology as well as for novel diagnostic and therapeutic development.