Machine-Learning Aided Design of Avidity-Based Biosensors of Ubiquitin Signaling

PROJECT SUMMARY This proposal is to develop a generalizable method to make genetically-encoded biosensors to monitor the dynamics, abundance, and positions of specific post-translational modifications (PTMs) of proteins. The strategy is based on our previous success in generating avidity-driven biosensors for distinct types of ubiquitin (Ub)- modified nucleosomes; each sensor is comprised of a nucleosome-recognizing Anchor, a Ub-binding UBD domain, and a Linker developed to maximize affinity and specificity. For Ub-modified proteins in particular, development of antibodies or other site-specific detection reagents has been notoriously difficult. Our goal is to extend the avidity-based strategy to engineer sensors able to detect heretofore intractable molecular targets such as the multiple and functionally distinct Ub signals found on nucleosomes. To accomplish this, our team will develop rational protein design technology that embraces multi-valent binding with tunable molecular flexibility. In Aim 1, we will design and test Linkers that deliver tunable domain geometry and motion. We will adapt new machine-learning algorithms to design fusion proteins that fix constituent Anchor and UBD domains in conformational space to maximize avidity and specificity. Furthermore, we will explore two new approaches to expand the Anchor repertoire and broaden applicability of the avidity strategy. In Aim 2, we will adopt the splitFAST fluorogen system to install a transferrable Anchor-binding tag on the substrate, and In Aim 3 we will develop customized Anchors from a yeast-display DARPin library. The efficacy and utility of the new sensors will be evaluated in vitro and in cells where they will be used to probe signaling associated with DNA damage repair pathways. In Aim 4, we will test and optimize conditions to use the sensors in genomic applications such as CUT&RUN assays. The reagents we develop in this project will allow researchers to probe otherwise invisible live-cell processes that are difficult or impossible to image with existing technologies. Our innovative approach directly addresses the challenge of binding to a highly flexible multi-domain protein target. As such, the resulting technology and design workflow will find application for diverse ubiquitinated targets, and more generally for binding targets that would otherwise be inaccessible due to high flexibility.