In-Depth Analysis Reveals Production of Circular RNAs from Non-Coding Sequences

Abstract

The sequencing of total RNA depleted for ribosomal sequences remains the method of choice for the study of circRNAs. Our objective was to characterize non-canonical circRNAs, namely not originating from back splicing and circRNA produced by non-coding genes. To this end, we analyzed a dataset from porcine testis known to contain about 100 intron-derived circRNAs. Labelling reads containing a circular junction and originating from back splicing provided information on the very small contribution of long non-coding genes to the production of canonical circRNAs. Analyses of the other reads revealed two origins for non-canonical circRNAs: (1) Intronic sequences for lariat-derived intronic circRNAs and intron circles, (2) Mono-exonic genes (mostly non-coding) for either a new type of circRNA (including only part of the exon: sub-exonic circRNAs) or, even more rarely, mono-exonic canonical circRNAs. The most complex set of sub-exonic circRNAs was produced by RNase_MRP (ribozyme RNA). We specifically investigated the intronic circRNA of ATXN2L, which is probably an independently transcribed sisRNA (stable intronic sequence RNA). We may be witnessing the emergence of a new non-coding gene in the porcine genome. Our results are evidence that most non-canonical circRNAs originate from non-coding sequences.

Beyond Back Splicing, a Still Poorly Explored World: Non-Canonical Circular RNAs

“…We greatly appreciated to note [ 24 ] that CIRCexplorer2 [ 34 ] considers two models of circRNA (exonic and lariat-derived intronic). In addition to tools developed to detect “all circRNAs”, we would like to mention a tool designed to work on intronic reads [ 35 ] and the development of a dedicated approach to characterize intron-derived circRNAs [ 20 , 21 ]. In 2013, Zhang et al [ 6 ] developed circFinder specifically to study circRNAs derived from introns.…”

“…(2) Talhouarne and Gall proposed lists of intron-derived circRNAs, particularly from mouse and human red blood cells [ 23 ]. (3) CircRNAs deriving from intron lariats were identified by our group when analyzing datasets of porcine pubertal testes [ 20 , 21 ]. (4) A recent study of neuronal cells (neuronal projections and synapses) in mice identified 278 introns as able to produce sisRNA, and reads containing circular junctions were detected in 14 of them [ 22 ].…”

“…Thanks to studies on human branch points, we know a distance of 9–400 bp between the branch point and the 3′ end of the intron is possible [ 37 ], even though 90% of branch points occur within 19 to 37 nucleotides upstream of the 3′ splice site [ 16 ]. After examining published lariat-derived circRNAs [ 20 , 21 , 23 , 24 ], we have no objection considering a distance between the branch point and the 3′ end of the intron ranging from 12 to 60 nt for lariat-derived circRNAs. Moreover, when examining the sequence of reads spanning the circular junction of a lariat derived intronic circRNA, heterogeneity can sometimes be observed, as if several branch points are possible [ 20 , 21 , 24 ].…”

“…After examining published lariat-derived circRNAs [ 20 , 21 , 23 , 24 ], we have no objection considering a distance between the branch point and the 3′ end of the intron ranging from 12 to 60 nt for lariat-derived circRNAs. Moreover, when examining the sequence of reads spanning the circular junction of a lariat derived intronic circRNA, heterogeneity can sometimes be observed, as if several branch points are possible [ 20 , 21 , 24 ]. Several studies showed that lariat-derived circRNAs that originated from introns did not use an adenosine at the branch point [ 6 , 20 , 21 , 22 , 23 , 24 ], but sometimes an ‘A’ can be observed.…”

“…To become circular sisRNA, the lariat escapes the action of the debranching enzyme ( Figure 1 ). Several studies showed that lariat-derived circRNAs originate from introns that do not use an adenosine at the branch point but rather a ‘C’, or more rarely, a ‘G’ or ‘T’ [ 6 , 20 , 21 , 22 , 23 , 24 , 25 ]. The lariat debranching enzyme is hypothesized to be incapable of hydrolyzing the 2′−5′ bond at these residues [ 15 ].…”