Discussion

Methods

References

Discussion

In the present work, we describe a fast, sensitive and easy-toimplement approach for the identification of artificial, ligandresponsive riboswitches in human cells in a high-throughput manner. By applying cDNA-amplicon-seq-based counting of conditionally expressed mRNAs, functional riboswitch candidates can be rapidly identified from complex libraries containing several thousand constructs, thereby enabling the exploration of large sequence spaces. Using this method, we identified tetracyclineand guanine-inducible hammerhead and HDV riboswitches, allosterically controllable Twister-riboswitches for use in human cells and finally, switches using a mode of action based on the conditional modulation of U1-snRNP-dependent polyadenylation. Our initial approach was based on the fact that functional aptazyme riboswitch sequences within a gene expression construct lead to the conditional, ligand-dependent cleavage and subsequent cellular degradation of the mRNA in which they are encoded. Due to the mRNA-intrinsic mode of action, the abundance and identity of riboswitch candidate sequences is directly linked to their function. A similar concept was previously applied to identify enhancer elements on a whole-genome scale, by cloning enhancer candidate libraries into reporter plasmids and quantifying the self-amplified reporter mRNAs by amplicon-seq. This method is known as “self-transcribing active regulatory region sequencing” (STARR-seq)22. Recently, efforts have also been undertaken in the field of ribozyme and riboswitch engineering, to establish high-throughput screening methods. Townshend and colleagues developed a method allowing screening of construct libraries in yeast10. However, this method relied on FACS-selection to generate several bins of low- to high-expressing constructs prior to the sequencing-based identification of functional sequences. Moreover, their method required expensive sequence barcoding and has not yet been shown to be transferrable to human cells. More recently, the Yokobayashi group published a series of papers, utilizing sequencing of in vitrotranscribed constructs to analyze cleavage of Twister-, HDV-like drz-Agam-2-1- and Pistol ribozyme variants23–۲۵٫ They further described a strategy for the identification of functional constructs in cells11. However, that method requires the isolated RNA to be separated into a non-cleaved and cleaved fraction by gel electrophoresis prior to sequencing, which again, is labor-intensive, more difficult to implement and potentially error-prone. In contrast, the method described in our work enables the rapid identification of conditionally cleaving aptazymes in a relevant cellular system, like e.g. the human cell line HEK-293, in a high-throughput manner, independent of additional pre-selection steps and costly sequence barcoding. The method relies on five core steps, namely library transfection, RNA extraction, cDNA synthesis, PCR amplification and sequencing, and enables direct screening for mRNA selfcleavage in libraries harboring up to approximately 18,000 constructs. Interestingly, while preparing our manuscript, the Smolke lab published a similar method that also successfully applies cDNA-sequencing to identify riboswitches26. In contrast to our method, which exploits the inherent self-barcoding nature of permutated sequence, the method described by Xiang et al. requires an additional sequence barcoding step26. Moreover, while both approaches underscore the applicability and value of cDNAamplicon-sequencing for the discovery of gene switches based on ribozyme platforms, the data presented in our study further extend the use of this method towards expression platforms beyond ribozymes, as demonstrated by the U1-snRNP library screen.