Circular RNAs (circRNAs) are covalently closed RNA structures found across all life. Initially characterised in the 1990s, the existence of many circRNAs was recently proven with the advent of high-throughput RNA sequencing (RNA-seq). Different types of circRNAs exist, but circRNAs resulting from exon back-splicing may be of particular biological importance as they can be found in the cytoplasm and contain a subset of coding and regulatory sequences of their cognate linear counterparts. While some influences on the back-splicing rate of exons were determined (such as presence of complementary sequences or interacting RNA-binding proteins in the introns surrounding the back-spliced exons), many factors potentially affecting circRNA production have not yet been explored systematically. Similarly, although functions were identified for a few circRNAs (such as 'sponging' of microRNAs and regulation of specific RNA binding protein abundance), the presence of most circRNAs in the cell remains enigmatic.

To address this, we identified multiple circRNAs from a range of human and mouse RNA-seq datasets and also investigated attributes of their cognate primary un-spliced mRNA precursors. We find that the exon circularization rate at a given gene correlates positively with the high variance and abundance of linear mRNA isoforms produced. Moreover, circRNA-producing primary transcripts possess distinct patterns of RNA polymerase II elongation rates compared to non-circRNA producing transcripts. These findings suggest a kinetic link between circRNA production yield and the ratio between transcription elongation and splicing speeds, with circRNAs viewed as one of the possible RNA isoforms. By analysing available CLIP-seq datasets and using microRNA target prediction, we find evidence for the potential of circRNAs to interact with microRNAs and RNA binding proteins. Most intriguingly, we detect multiple circRNAs in the polysomal fractions of the cytoplasm and suggesting differential translational involvement across the circRNA range synthesised in the cell. Many of the translated circRNAs can result in an infinite loop protein synthesis due to the in-frame back-splicing. This feature as well as a distinctively different metabolism of circRNAs due to the lack of 5' cap, 3' poly(A) and any of the 5' or 3' ends, place them in a unique functional niche in the cell.