Human circular RNAs: design, development and functional analysis of protein sponges
Circular RNAs (circRNAs) are a relatively new class of stable, non-coding RNAs often derived from protein-coding genes. They are generated by the so-called backsplicing mechanism, whereby one single or several adjacent exons are excised in circular configuration from a pre mRNA. The functional roles of the vast majority of circRNAs are largely ... unknown and likely diverse. While several putative functions have been suggested for this class of RNAs, such as miRNA sponging, templates for translation and RBP (RNA binding protein) sponging. Only a few of these proposed functions have been thoroughly validated. This work explores RBP sponging as a novel function of circRNAs. The primary focus of this work was on employing artificially designed circRNAs as molecular tools for RBP sequestration, with a view to influence key cellular processes such as splicing. Firstly, we investigated the combinatorial recognition of RNA motifs by RBPs using SELEX (Systematic Evolution of Ligands by EXponential enrichment) coupled with high-throughput sequencing and motif analysis, to identify highly specific and tailored RBP-binding sequences. These were implemented into a circular RNA, which functioned as an RBP sponge. As a proof of principle, we utilised this strategy for two RBPs regulating alternative splicing - hnRNP L (heterogeneous nuclear RiboNucleoProtein L) and RBM24. HnRNP L is a global regulator of alternative splicing, binding preferentially to CA-rich RNA sequences, while RBM24 is a muscle-specific alternative splice regulator that binds to GU-rich stretches on target mRNAs. We observed that circular RNAs have a clear advantage over their conventional linear counterparts due to their higher stability and resistance to degradation by cellular RNases. Secondly, the binding affinity of the SELEX-derived, highly specific consensus sequences and the protein-RNA interaction was validated by electrophoretic mobility shift assay and in vitro pulldown assays. Surprisingly, our findings revealed that the binding preference of hnRNP L with its cognate RNA not only depends on sequence specificity, but also on the spacing of the motifs within the RNA. The four domains of hnRNP L appear to prefer specific nucleotide spacing for binding. Finally, to demonstrate in vivo sponging of hnRNP L and its effects on splicing we focused on circRNA overexpression and applied a tRNA-based, ribozyme-driven Tornado circRNA expression system, as well as in vitro-generated circRNAs. Alternative splicing of hnRNP L target genes was analysed by RT-PCR. Interestingly, as a result of sponging, we observed hnRNP L translocation from the predominant nuclear localisation to a more cytoplasmic distribution in HeLa cells. This work demonstrates the considerable potential of designer circRNAs for modulating activities of RBPs or for alterations of particular alternative splicing events. As a promising alternative to pharmacological inhibition of proteins, they can be applied in a way similar to antisense splice-switching oligonucleotides targeting individual splicing events. This forms the basis for the design of optimal circRNA-protein sponges, which can be applied to any RBP of clinical importance.