Establishing a Multivariate Model for Predictable Antisense RNA-Mediated Repression

Recent advances in our understanding of RNA folding and functions have facilitated the use of regulatory RNAs such as synthetic antisense RNAs (asRNAs) to modulate gene expression. However, despite the simple and universal complementarity rule, predictable asRNA-mediated repression is still challenging due to the intrinsic complexity of native asRNA-mediated gene regulation. To address this issue, we present a multivariate model, based on the change in free energy of complex formation (ΔG CF) and percent mismatch of the target binding region, which can predict synthetic asRNA-mediated repression efficiency in diverse contexts. First, 69 asRNAs that bind to multiple target mRNAs were designed and tested to create the predictive model. Second, we showed that the same model is effective predicting repression of target genes in both plasmids and chromosomes. Third, using our model, we designed asRNAs that simultaneously modulated expression of a toxin and its antitoxin to demonstrate tunable control of cell growth. Fourth, we tested and validated the same model in two different biotechnologically important organisms: Escherichia coli Nissle 1917 and Bacillus subtilis 168. Last, multiple parameters, including target locations, the presence of an Hfq binding site, GC contents, and gene expression levels, were revisited to define the conditions under which the multivariate model should be used for accurate prediction. Together, 434 different strain-asRNA combinations were tested, validating the predictive model in a variety of contexts, including multiple target genes and organisms. The result presented in this study is an important step toward achieving predictable tunability of asRNA-mediated repression.