Structured elements drive extensive circular RNA translation

Abstract

The human genome encodes tens of thousands circular RNAs (circRNAs) with mostly unknown functions. Circular RNAs require internal ribosome entry sites (IRES) if they are to undergo translation without a 5' cap. Here, we develop a high-throughput screen to systematically discover RNA sequences that can direct circRNA translation in human cells. We identify more than 17,000 endogenous and synthetic sequences as candidate circRNA IRES. 18S rRNA complementarity and a structured RNA element positioned on the IRES are important for driving circRNA translation. Ribosome profiling and peptidomic analyses show extensive IRES-ribosome association, hundreds of circRNA-encoded proteins with tissue-specific distribution, and antigen presentation. We find that circFGFR1p, a protein encoded by circFGFR1 that is downregulated in cancer, functions as a negative regulator of FGFR1 oncoprotein to suppress cell growth during stress. Systematic identification of circRNA IRES elements may provide important links among circRNA regulation, biological function, and disease.

Keywords: 18S complementarity; FGFR1; cap-independent translation; circFGFR1p; circRNA-encoded protein; circular RNA; internal ribosome entry site; structured RNA element.

Figure 1.. High-throughput identification of RNA sequences that can facilitate circRNA cap-independent translation.

(A) A schematic overview of the high-throughput split-eGFP circRNA reporter screening assay for identifying circRNA IRES. (B) The eGFP expression distribution of captured synthetic oligos. Pie chart: the composition of different categories among eGFP(+) oligos. (C) Quantification of the percentage of captured eGFP(+) oligo sequences from different orgin. (D) Normalized eGFP expression of each captured oligo in the screening assay performed on the circular RNA or the linear RNA system. Red dashed lines: normalized eGFP expression threshold.

Figure 2.. CircRNAs containing eGFP(+) oligo have higher cap-independent translation activity.

(A) A schematic of the circRNA polysome profiling for translating circRNAs. (B) (Poly)ribosome fractionations of cells transfected with oligo library split-eGFP circRNA reporter followed by CHX treatment. (C) Quantification of the percentage of (poly)ribosome-enriched oligos of captured eGFP(−) or eGFP(+) oligos. (D) Sequencing reads from Ribo-seq and QTI-seq plotted on the genes showing eGFP(+) oligos harboring aTIS, nTIS, and dTIS with overlapping annotated circRNAs (brown segments). (E) Quantification of the percentage of eGFP(−) or eGFP(+) oligos harboring no TIS (TIS(−)) (left) or more than one TIS (TIS(+)) (right), and the percentage of aTIS, nTIS, or dTIS oligos among eGFP(+)/TIS(+) oligos.

Figure 3.. 18S rRNA complementary sequence on the IRES facilitates circRNA translation.

(A) A schematic of the sliding-window approach for mapping the active regions on the human 18S rRNA. (B) Quantification of the mean eGFP expression of the oligos overlapping with the corresponding position across the human 18S rRNA. Dashed line: background eGFP expression. Green shaded: active regions on the 18S rRNA. (C) An illustration of the secondary structure of human 18S rRNA showing the active regions (green) and reported mRNA (red) or IRES RNA (orange) contact regions. (D) Quantification of the number of the 18S rRNA active 7-mers or the random 7-mers harbored by eGFP(+) or eGFP(−) oligos plotted on a Tukey box-plot. (E) Quantification of the IRES activity for the oligo with higher or lower 18S rRNA complementarity. Error bar: SEM. (F) A schematic of the synthetic oligos for systematic scanning mutagenesis. (G) The eGFP expression of each oligo containing the random substitution mutation at the corresponding position on HCV IRES. The eGFP expression for each oligo was normalized to the mean eGFP expression of all the oligos on the HCV IRES. (H) Examples of circRNA IRES with local and global sensitivity identified by scanning mutagenesis. Blue shaded: the identified essential elements on the IRES. (I) The mean eGFP expression of all the circRNA IRES oligos with global sensitivity at each mutation position across the IRES (blue: 5–15 nt and 135–165 nt; red: 40–60 nt). (J) Quantification of the local MFE in a 15 nt sliding window on the IRES (blue: 5–15 nt and 135–165 nt; red: 40–60 nt).

Figure 4.. 40–60 nt SuRE on the IRES can facilitate circular IRES activity.

(A)-(H) The secondary structure of the mutated IRESs determined by M2-seq. Red arrowheads: 40–60 nt SuRE. CircIRES-dis: circular IRES with the SuRE disrupted by sequence substitution. CircIRES-relocate: circular IRES with the SuRE relocated to 90–110 nt region. CircIRES-single and circIRES-comp: circular IRES with single complementary mutations and compensatory double complementary mutations, respectively. circIRES-BoxB: circular IRES with the SuRE substituted by BoxB stem-loop. linearIRES-add: linear IRES with 40–60 nt region substituted by the 40–60 nt SuRE on the circular IRES. (I) Quantification of the IRES activity for each mutated IRES normalized to the linear IRES. An unpaired two-sample t-test relative to the linear IRES was performed. Error bar: SEM. (J) Quantification of the percentage of the eGFP(+) oligos (left) and endogenous translated circRNAs (right) harboring 18S rRNA complementarity or the SuRE element. (K) An illustration of two key regulatory elements facilitating circRNA translation.

Figure 5.. IRES elements facilitate translation initiation of endogenous circRNAs.

(A) A schematic of disrupting the key regulatory elements on the IRES by anti-sense LNAs targeting specific regions. (B) Quantification of the normalized eGFP fluorescence intensity of the cells co-transfected with the corresponding LNA and the reporter plasmid carrying the corresponding IRES. The number represents the index number of the oligo. (C) A schematic of QTI-qRT-PCR quantification of the level of translation-initiating endogenous circRNAs. (D) Quantification of the translation-initiating RNA level of the human endogenous circRNAs containing the corresponding IRES upon corresponding LNA disruption. The circRNA level was normalized to the GAPDH mRNA. (B) and (D): An unpaired two-sample t-test relative to mock transfection was performed. Error bar: SEM.

Figure 6.. Identification of putative endogenous circRNA-encoded proteins.

(A) Quantification of the percentage of IRES-mapped human endogenous circRNAs. (B) Top 12 represented biological processes from GO term analysis that are enriched in the parent genes of IRES(+) circRNAs. (C) A schematic of generating the putative endogenous circORF list. (D) Top 15 represented conserved motifs from Pfam analysis that are enriched in the predicted circORFs. (E) A schematic of peptidomic validation of the putative circORFs. (F) The heat map showing the number of the unique MS/MS peptides detected in different peptidomic datasets for each of the peptidomic detected circORF. The number on the right indicates the total number of different circORFs detected in the corresponding dataset. (G) The MS1 and MS2 spectra of a representative tryptic BSJ peptide captured from circORF_575.

Figure 7.. CircRNA-encoded circFGFR1p suppresses cell proliferation under stress conditions.

(A) A schematic of FGFR1 and circFGFR1 transcript. (B) Sanger sequencing results detecting the back-splicing junction (yellow box) of circFGFR1. (C) A schematic of the conserved motifs on FGFR1 and circFGFR1p. Ab (both): the antibody which can detect both FGFR1 and circFGFR1p. Ab-circFGFR1p: custom circFGFR1p antibody. (D) A schematic of the peptides captured by IP-MS (underline) that matched the circFGFR1p unique region (red) and the region overlaping with FGFR1 (black). The antigen peptide of the circFGFR1p custom antibody was labeled in bold. Red box: the extracted region for IP-LC-MS/MS. (E) The MS2 and the top 3 rank PRM-MS transition ions spectra of the spike-in heavy isotope labeled peptide (top) and the BJ tryptic peptide (bottom) of circFGFR1p. [L]: heavy isotope labeled leucine. (F) Western blots showing circFGFR1p and FGFR1 protein level (Ab-both), and the quantification of FGFR1 and circFGFR1 RNA level by qRT-PCR of cells transfected with siRNA or LNA. An unpaired two-sample t-test relative to siCtrl was performed. (G) Quantification of cell proliferation in cells with the knockdown of siCircFGFR1 or circFGFR1-LNA with FGF1 addition. An unpaired two-sample t-test relative to siCtrl was performed. (H) Western blots showing FGFR1 protein and circFGFR1p level with or without the heat-shock. (I) Quantification of the Western blot of circFGFR1p protein level relative to FGFR1 (all isoforms) under normal (WT) and the heat-shock (HS) condition. An unpaired two-sample t-test relative to WT was performed. (J) Quantification of the Western blot showing the change of the protein level of FGFR1 and circFGFR1p under the heat-shock condition. Protein level is normalized to the GAPDH. A one-sample t-test relative to 1 was performed. The error bars in (F), (G), (I), and (J) represent SEM.