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

Abstract

Most of the circRNAs reported to date originate from back splicing of a pre-mRNA, and these exonic circRNAs are termed canonical circRNAs. Our objective was to provide an overview of all other (non-canonical) circRNAs that do not originate from the junction of two exons and to characterize their common properties. Those generated through a failure of intron lariat debranching are the best known, even though studies on them are rare. These circRNAs retain the 2'-5' bond derived from the intron lariat, and this feature probably explains the difficulties in obtaining efficient reverse transcription through the circular junction. Here, we provide an unprecedented overview of non-canonical circRNAs (lariat-derived intronic circRNAs, sub-exonic circRNAs, intron circles, tricRNAs), which all derive from non-coding sequences. As there are few data suggesting their involvement in cellular regulatory processes, we believe that it is early to propose a general function for circRNAs, even for lariat-derived circRNAs. We suggest that their small size and probably strong secondary structures could be major obstacles to their reliable detection. Nevertheless, we believe there are still several possible ways to advance our knowledge of this class of non-coding RNA.

Keywords: annotation of circRNAs; back splicing; intron circle; intron lariat; lariat-derived circRNA; sisRNA; sub-exonic circRNA.

Figure 1

Genesis of lariat-derived circRNA. When intron lariats escape degradation through failure of intron debranching, they can become circRNA precursors. The intron excision from pre-mRNA releases a lariat molecule, in which the branch point nucleotide, predominantly an adenosine, is linked 2′–5′ to the 5′ end of the intron [15,16]. The lariat debranching enzyme is hypothesized to be incapable of hydrolyzing the 2′–5′ bond when the branch point nucleotide is not an ‘A’ [15].

Figure 2

Detection and annotation of circRNAs. (A) In this region, several circRNAs (4× blue, 2× brown, 1× pink, and 1× purple) are detected. (B) Three genes are described in this region: two on the forward strand (brown gene and blue gene) and one on the reverse strand (pink gene). (C) Annotation of circRNAs. The circRNA indicated in pink can be annotated as exonic circRNA (canonical circRNA). In the circular junction sequenced, an exonic downstream donor 3′ splice site (indicated in green in (B)) is covalently joined to an exonic upstream 5′ splice site (indicated in dark-green in (B)). The parent gene of this exonic circRNA is the pink gene. For the two circRNAs indicated in brown, they are mapped on the forward strand inside intronic sequences of the brown gene located on the same strand. As the position ‘Start-circRNA’ (see (A)) is identical to the position ‘Start-intron’ (see (B)) and the distance between the position ‘End-circRNA’(see (A)) and the position ‘End-intron’ (see (B)) is less than 60 nt, we can annotate these two circRNAs as lariat-derived circRNAs from the brown gene. The circRNA indicated in purple is mapped on the reverse strand inside an intron of the pink gene located on the same strand. We can note identical genomic coordinates for ‘Start-circRNA’ and ‘Start-intron’ and for ‘End-circRNA’ and ‘End-intron’. Thus, we can annotate this circRNA as an intron circle produced by the pink gene. When we examine the characteristics of the four circRNAs indicated in blue, we observe that they are compatible with a set of sub-exonic circRNAs produced by the blue gene, which is a non-coding gene. Indeed, their genomic coordinates are located inside the single exon of this gene and the transcription strand is identical.