Translation of circular RNAs

Abstract

Circular RNAs (circRNAs) are covalently closed RNAs that are present in all eukaryotes tested. Recent RNA sequencing (RNA-seq) analyses indicate that although generally less abundant than messenger RNAs (mRNAs), over 1.8 million circRNA isoforms exist in humans, much more than the number of currently known mRNA isoforms. Most circRNAs are generated through backsplicing that depends on pre-mRNA structures, which are influenced by intronic elements, for example, primate-specific Alu elements, leading to species-specific circRNAs. CircRNAs are mostly cytosolic, stable and some were shown to influence cells by sequestering miRNAs and RNA-binding proteins. We review the increasing evidence that circRNAs are translated into proteins using several cap-independent translational mechanisms, that include internal ribosomal entry sites, N6-methyladenosine RNA modification, adenosine to inosine RNA editing and interaction with the eIF4A3 component of the exon junction complex. CircRNAs are translated under conditions that favor cap-independent translation, notably in cancer and generate proteins that are shorter than mRNA-encoded proteins, which can acquire new functions relevant in diseases.

Graphical Abstract

Figure 1.

Generation and overall functions of mRNAs and circRNAs. (A) A pre-mRNA containing three exons and two introns flanking the central exon is processed into a mRNA containing a 5′ cap and a poly adenosine tail using the ‘linear’ pre-mRNA pathway, dashed lines. The invariant AG-GT nucleotides of the 3′ and 5′ splice sites are indicated. mRNAs are predominantly exported by the TREX (transcription and RNA export) complex from the nucleus to the cytosol where they are translated. (B) A small fraction of the pre-mRNA can undergo backsplicing when the 5′ and 3′ splice site are brought into close proximity, which is promoted by RNA secondary structures or mediated by proteins. In humans, pre-mRNA structures are often introduced by Alu elements that form long complementary sequences, which are the major substrate for ADAR enzymes. CircRNAs are exported via an exportin-2-IGFBP2-RanGTP complex. In the cytosol they can sequester miRNAs, RNA binding proteins or act as translational templates.

Figure 2.

Translation initiation mechanisms used by mRNAs and circRNAs. (A) Cap-dependent translational initiation: Schematic illustration of the cap-dependent translational initiation complex. Eukaryotic initiation factors and RNA modification readers are indicated by color. (B) Cap-dependent translational initiation of m6A-methylated mRNA: M6A methylation promotes formation of the cap-dependent translational initiation complex by stabilizing the interaction of the 40S subunit through interaction of the m6A reader YTHDF1 with eIF3. (C) m6A-methylation-dependent translation of circRNAs: The complex of methyltransferases 3 and 14 catalyzes m6A methylation of adenosines. M6A binds to the YTHDF3, which binds to eIF3G that recruits the 40S rRNA subunit. (D) RNA editing-dependent translation of circRNAs: ADAR activity (ADAR1, ADAR2, adenosine deaminase acting on RNA) modifies adenosines to inosines that bind to an unknown reader, which binds to the 40S rRNA subunit, initiating translation at the AUG indicated (arrow). (E) Exon-junction-dependent translation of circRNAs: eIF4A3 is part of the exon junction complex, which binds to the border of former exons spliced together (arrow). eIF4A3 interacts with eIF3G that initiates translation. For clarity, eIF4A3 is shown only on the backsplice junction, but is likely present in all exon junctions.

Figure 3.

Protein variants generated from circRNAs. (A) Schematic structures of proteins made from circRNAs, shown in Table 2. Sequences are in Supplementary Table S1. (B) Rolling circle translation. CircRNA-encoded proteins are translated from a methionine usually, but not aways, in the same reading frame as in the linear mRNA until a stop codon is encountered. If there is a change in reading frame between exon 4 and exon 2, a circRNA-specific N- and C-terminus is generated. Adenosine to inosine RNA editing can generate novel start codons by changing a AUA (Ile) into an AUI start codon. In the absence of a stop codon, rolling circle translation generates a infinite ORF, where protein production stops after several rounds of translation via an unknown mechanism. (C) CircRNAs can encode proteins lacking a transmembrane domain. Most transmembrane domains are encoded by the first exon that lacks a 3′ splice site and is not included in the circRNAs. Thus, the resulting circRNA-encoded protein lacks the transmembrane domain and is likely cytosolic.