Early history of circular RNAs, children of splicing

ABSTRACT

In this commentary we briefly summarize early work on circular RNAs derived from spliceosome mediated circularization. We highlight how this early work inspired work on the basic mechanisms of nuclear RNA splicing, the possible function of circular RNAs and the potential uses of circular RNAs as tools in biomedicine. Recent developments in the study of circular RNAs, summarized in this volume, have brought these questions back to the foreground.

The 1977 discovery of pre-mRNA splicing (Berget et al., 1977; Chow et al., ¹⁹⁷⁷) immediately led to the realization that there was an exon (or intron) definition problem: If an organism fails to define exons stringently, RNA processing could result in (1) exon skipping by ligation of a 5′ splice site to a far-downstream 3′ splice site and other forms of alternative splicing, (2) trans-splicing by ligation of a 5′ splice site to a 3′ splice site in another pre-mRNA, or (3) RNA circularization by ligation of a 5′ splice site to an upstream 3′ splice site. Indeed, all of these are now known to occur and to play critically important roles in providing plasticity to the coding capacity of genomes. Although much attention was given to alternative pre-mRNA splicing and some to trans-splicing, exon circularization was not appreciated immediately and indeed remained woefully understudied for decades. Acknowledging that several different mechanisms are utilized to form circular RNAs, here we review the early history of circular RNAs formed from exon circularization or ‘back splicing’: circular RNAs that are the children of splicing.

About a decade following the discovery of self-splicing introns (Cech et al., 1981), Puttaraju and Been (1992) showed that a self-splicing group I intron, permuted by placing the 5′ splice site downstream of the 3′ splice site, yielded circular RNAs in vitro. Indeed, this work confirmed observations by Price et al. (1987), who detected an RNA circle believed to be formed by the splicing of a cryptic 5′ splice site downstream from a canonical 3′ splice site in a group I splicing substrate. Ford and Ares (1994) showed that circular RNAs obtained by group I splicing can be expressed in vivo in E. coli and yeast. Circular RNAs were, and still are, believed to be more stable in vivo than linear RNAs due to their resistance to end-dependent degradation (Puttaraju et al., 1993). Importantly, by ¹⁹⁹⁶ Puttaraju and Been, using the permuted group I intron method, showed that trans-acting RNA enzymes (the hepatitis delta virus ribozyme and B. subtilis RNase P RNA) can be expressed in vivo as circular RNAs that retain catalytic activity.

The experiments involving RNA circularization by permuted group I introns prompted the prediction of similar mechanisms for group II and pre-mRNA exons. Consistent with these predictions, splicing-mediated exon circularization was observed with permuted group II introns (Jarrell, 1993). The existence of spliceosome mediated RNA circularization, however, went against the general view that splice sites were used in the order in which they appeared in nascent transcripts – the ‘first come, first served’ model (Aebi et al., 1986; Beyer and Osheim, ¹⁹⁸⁸). Nonetheless, RNAs with ‘scrambled exons’, consistent with either exon circularization or trans-splicing, were observed (Nigro et al., 1991; Cocquerelle et al., ¹⁹⁹²), although these events were rare. Importantly, Capel et al. (1993) showed that transcripts of the testis-determining gene Sry were efficiently circularized in mouse adult testis and that this circularization prevented their translation. Capel et al., also showed that unique inverted repeats in the Sry transcript approximated the splice sites in a long exon and posited that RNA circularization was mediated by the spliceosome.

Spliceosome-mediated RNA circularization was demonstrated with single exon substrates in yeast extracts (Schindewolf et al., 1996) and mammalian nuclear extracts (Pasman et al., 1996). One difference between the yeast and mammalian reactions was that the yeast system seemed to produce only the spliced product, whereas the mammalian reactions produced, in addition to the spliced circular exon, low levels of many other non-linear RNAs. The significance of these products remains unclear, but these could very well reflect the looser definition of splice sites in mammalian pre-mRNAs and be the products of cryptic reactions as proposed for group I introns by Price et al. (1987).

The discovery of spliceosome mediated exon circularization raised questions about the mechanism of splice site approximation and about how the splicing machinery commits to a specific pairs of splice sites in alternative splicing. Similar implications were suggested by earlier work of Konarska et al. (1985), Solnick (1985), and Chiara and Reed (1995), who showed that 5′ and 3′ splice sites on separate conventional pre-mRNAs can be spliced, producing the expected linear product. These splicing reactions, termed trans-splicing, were consistent with the idea that splicing does not require continuous intron sequences between the splice sites (although reaction kinetics were not yet addressed). These experiments suggested a 3-dimensional diffusion model of splice site approximation, which postulates that the splice sites are brought into proximity via interaction of complexes assembled at the splice sites sequences irrespective of the intron sequences separating them. A scanning model, in contrast, postulates that, once assembled, complexes at the splice sites are brought together in a process that involves one-dimensional scanning of intron sequences. The scanning model is reminiscent of the eukaryotic translation process involving 5′ RNA cap recognition followed by one-dimensional scanning to the translation start codon. These two models were tested directly by Pasman and Garcia-Blanco (1996). Using several lines of evidence, including splicing precursors in which portions of the intron were replaced with a flexible non-nucleic acid polymer, the results supported the 3 dimensional diffusion model. It should be noted that while the splicing machinery does not bring the splice sites together by scanning, subcomponents of this machinery piggyback on the DNA dependent RNA polymerase II (RNAP II) that scans splicing precursors as it synthesizes them (reviewed in Goldstrohm et al., ²⁰⁰¹). This may explain how splice sites that are large distances away can be brought into close spatial proximity, resulting in accurate and rapid splice site pairing via a 3-dimensional diffusion mechanism.

It is clear that early studies of circular RNAs contributed to the broad understanding of basic nuclear RNA splicing mechanisms. Indeed, some of the questions raised by these early experiments remain unanswered, particularly regarding the commitment to different splice site pairs during alternative splicing. Early studies also led to questions regarding the function of many circular RNAs – questions that awaited the development of novel technologies. New insights into the function of circular RNAs, which are eloquently presented in this volume, echo the very special function of Sry circularization in gene expression, but now in trans on the function of other RNAs. Finally, it should be pointed out that synthetic circular RNAs were proposed to have special uses for biomedical applications (Bohjanen et al., 1996) and this continues to be a promise today.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acnowledgement

M.A.G-B. acknowledges funding from The University of Texas Medical Branch, and from the Duke-NUS Signature Research Program in Emerging Infectious Diseases funded by the Agency for Science,Technology and Research, Singapore, and the Ministry of Health, Singapore.