Full Design Automation of Multi-State RNA Devices to Program Gene Expression Using Energy-Based Optimization

Small RNAs (sRNAs) can operate as regulatory agents to control protein expression by interaction with the 5′ untranslated region of the mRNA. We have developed a physicochemical framework, relying on base pair interaction energies, to design multi-state sRNA devices by solving an optimization problem with an objective function accounting for the stability of the transition and final intermolecular states. Contrary to the analysis of the reaction kinetics of an ensemble of sRNAs, we solve the inverse problem of finding sequences satisfying targeted reactions. We show here that our objective function correlates well with measured riboregulatory activity of a set of mutants. This has enabled the application of the methodology for an extended design of RNA devices with specified behavior, assuming different molecular interaction models based on Watson-Crick interaction. We designed several YES, NOT, AND, and OR logic gates, including the design of combinatorial riboregulators. In sum, our de novo approach provides a new paradigm in synthetic biology to design molecular interaction mechanisms facilitating future high-throughput functional sRNA design.Author Summary: Is our current knowledge of in vivo RNA-RNA interactions and thermodynamics enough to perform the unsupervised computational design of fully synthetic sequences encoding functional RNAs in living cells? Recent work gave a positive answer for the challenging problem of designing activating riboregulators. This was done by integrating theory and computation to develop a physicochemical framework for the design of regulatory RNA systems, using Watson-Crick interactions and optimization algorithms. Still, the objective function was not directly validated, preventing using with confidence the methodology for other systems. We here validate experimentally an objective function relying on free energies of RNA complex activation and formation, which allows extending the framework to produce logic devices that can be implemented to program gene expression. We demonstrate that it is possible to design increasingly sophisticated and modular functions, pointing our results out that energy-based optimization methods can perform the large combinatorial search required for RNA design.