A Molecular and Computational Toolbox to Enable Therapeutic A-to-I RNA Editing

Abstract: Antisense RNA therapies are currently capable of transcript level precision through modulation of expression and splicing. To advance RNA therapeutics to single nucleotide level precision, we have designed antisense RNAs that recruit endogenous ADAR for targeted A-to-I RNA editing. Our scalable and modular discovery pipeline engineers DNA-encoded guide RNAs (gRNAs) that overcome many design challenges: editing efficiency and specificity; gRNA expression and localization; and targeted delivery. Biochemical assessment of ADAR editing efficiency and selectivity across the secondary landscape of >100,000 gRNA:mRNA secondary structures was empirically determined and used to build a machine learning (ML) model capable of de novo, in silico generation of high-performing gRNAs for any target RNA of interest. Next, ML generated secondary structures were built into a larger gRNA architecture that is further engineered for efficient editing in cells. This requires high expression, stability, localization, target RNA accessibility, and delivery to the target tissue. By leveraging and engineering regulatory elements from endogenous snRNAs, we improved gRNA expression and promoted nuclear localization. Additionally, we engineered a suite of AAV capsids that provide selective tissue targeting. The combination of these technological advancements significantly enhances the therapeutic potential of ADAR mediated RNA editing.