Circular RNA is expressed across the eukaryotic tree of life

Abstract

An unexpectedly large fraction of genes in metazoans (human, mouse, zebrafish, worm, fruit fly) express high levels of circularized RNAs containing canonical exons. Here we report that circular RNA isoforms are found in diverse species whose most recent common ancestor existed more than one billion years ago: fungi (Schizosaccharomyces pombe and Saccharomyces cerevisiae), a plant (Arabidopsis thaliana), and protists (Plasmodium falciparum and Dictyostelium discoideum). For all species studied to date, including those in this report, only a small fraction of the theoretically possible circular RNA isoforms from a given gene are actually observed. Unlike metazoans, Arabidopsis, D. discoideum, P. falciparum, S. cerevisiae, and S. pombe have very short introns (∼ 100 nucleotides or shorter), yet they still produce circular RNAs. A minority of genes in S. pombe and P. falciparum have documented examples of canonical alternative splicing, making it unlikely that all circular RNAs are by-products of alternative splicing or 'piggyback' on signals used in alternative RNA processing. In S. pombe, the relative abundance of circular to linear transcript isoforms changed in a gene-specific pattern during nitrogen starvation. Circular RNA may be an ancient, conserved feature of eukaryotic gene expression programs.

Figure 1. Eukaryotic Tree of Life.

This shows the divergence between organisms studied in this report (in red) and metazoans where circular RNA expression has been previously reported. Adapted from Csuros et al. under Creative Commons license CC0. According to , “Branch widths are proportional to intron density which is shown next to terminal taxa and some deep ancestors, in units of the introns count per 1 kb coding sequence”.

Figure 2. Circle-specific PCR and relative RNase R resistance.

a) An example of circular and linear isoforms, in this case for the P. falciparum gene MAL13P1.337, and circle- and linear-specific PCR design. PCR is performed on cDNA from total RNA that was mock-treated or RNase R-treated, or on P. falciparum genomic DNA. Circle-specific PCR amplifies from RNA but not genomic DNA; it amplifies the candidate junction (177 bp band) but also an unexpected band corresponding to a 4-1 circle. b) Quantitation of RNase R resistance. Plotted here is the relative RNase R resistance of a circular isoform compared to its counterpart linear isoform (ΔΔCt): RNase R resistance(circle) – RNase R resistance(linear); gray bars are standard errors. RNase R resistance (the log₂ fold-change in RNA isoform abundance with RNase R treatment) was measured by quantitative RT-PCR and taken as ΔCt  =  Ct(mock – treatment) – Ct(RNase R – treatment). RNase R resistance values for circular and linear isoforms are separately shown in Figure S1. All linear isoforms were sensitive to RNase R, showing a greater than 32-fold drop in abundance after RNase R treatment (ΔCt <−5). Circular isoforms show no significant decrease in abundance with RNase R treatment, and in many cases the signal actually increases (see main text). The absolute Ct for mock-treated RNA is also given, as an indicator of the comparative abundance of circular and linear isoforms. For S. pombe, data shown here is for exponential growth. c) RNase R resistance of genes in two additional organisms, Dictyostelium and S. cerevisiae. Format is similar to b).

Figure 3. S. pombe and human genes producing circular RNAs.

The four S. pombe genes for which we validated circular isoforms are shown schematically. Exons are boxes, with untranslated regions in light green and coding regions in darker green, introns are indicated by a black line; the total size of the transcribed region is given in parentheses. The exons present in circular RNA are indicated by red boxes and their sizes indicated. For comparison, two human genes that produce circular isoforms are similarly presented, though at a very different scale since they are considerably larger and contain much larger introns.

Figure 4. S. pombe circular and linear RNA changes during nitrogen starvation.

S. pombe cultures were grown in complete minimal media to exponential phase. Time zero marks the switch to media lacking a nitrogen source. Relative RNA abundance (log₂ fold-change) of circular and linear isoforms was determined by quantitative RT-PCR (an equal mass of RNA for each timepoint was used as input), adjusted to a per cell basis, and expressed relative to time zero; the plotted value is Ct(time-zero) – Ct(time n) – log₂(RNA per cell at time n/RNA per cell at time-zero).