Comprehensive identification of internal structure and alternative splicing events in circular RNAs

Abstract

Although previous studies demonstrated circular RNAs (circRNAs) does not exclusively comprise mRNA exons, no study has extensively explored their internal structure. By combining an algorithm with long-read sequencing data and experimental validation, we, for the first time, comprehensively investigate internal components of circRNAs in 10 human cell lines and 62 fruit fly samples, and reveal the prevalence of alternative splicing (AS) events within circRNAs. Significantly, a large proportion of circRNA AS exons can hardly be detected in mRNAs and are enriched with binding sites of distinct splicing factors from those enriched in mRNA exons. We find that AS events in circRNAs have a preference towards nucleus localization and exhibit tissue- and developmental stage-specific expression patterns. This study suggests an independent regulation on the biogenesis or decay of AS events in circRNAs and the identified circular AS isoforms provide targets for future studies on circRNA formation and function.

Figure 1. The algorithm and workflow of CIRI-AS.

(a) The workflow of cirexon and AS detection. BSJ, back-spliced junction; FSJ, forward-spliced junction. (b) BSJ read pairs are peculiar to circRNAs and are thus used to identify splice junctions, indicating boundaries of cirexons. Red and blue bold lines indicate BSJ read pairs, which are connected by black dashed lines. Curved black lines indicate the FSJs present within circRNAs. (c) All possible routes are constructed using identified FSJs and cirexons to detect AS events within circRNAs. Red and grey nodes indicate alternatively spliced and constitutive cirexons, respectively. Dashed lines and numbers represent FSJs and supporting read pairs, respectively, in which grey numbers indicated that the FSJs are involved in multiple routes and supporting read pairs for each route are summed.

Figure 2. Cirexon and AS detection in three human cell lines.

(a) BSJs, cirexons and ICFs detected in the three samples treated by RNase R. (b) Rarefaction analysis of detected cirexons, as well as corresponding BSJs and alternatively spliced cirexons in Hs68. (c) Overlap of cirexon detection of CIRI-AS using 100 bp paired-end reads and long reads. The bottom circular plots indicate two circRNAs and their cirexons validated by long-read sequencing (grey and orange spiral lines). Inner coloured solid-line curves indicate the cirexons detected by CIRI-AS, whereas the dashed-line curves represent the cirexons missed by CIRI-AS. (d) Percentage of circRNAs (≥20 BSJ reads) containing four types of alternatively spliced cirexons in the three samples: ES, alternative 3′-splicing site, alternative 5′-splicing site and IR. (e) Comparison of Ψ value for ES events between circRNA and mRNA in HeLa and HEK293 samples. (f–j) An experimentally validated example of alternatively spliced cirexons: ES and alternative 5′-splicing site within circRNA chr1:231,090,079|231,097,049. (f) Cirexons and splice junctions, as well as sequencing depth and corresponding BSJ reads, within the circRNA detected by CIRI-AS in RNase R-treated sample of HeLa cells. (g) Exons and splice junctions, as well as sequencing depth and corresponding sequencing read pairs, in poly(A)-selected sample of HeLa cells. (h) Positions of outward-facing primers and critical splice junctions on the three circular transcipts sharing the same BSJs. (i) The amplified fragments by RT–PCR corresponding to the three circular transcripts. (j) Sequencing chromatograms across the critical splice junctions of the PCR products.

Figure 3. Binding site density of splicing factors in transcribed exons of HeLa cells.

(a) Binding site density of splicing factors in skipped/constitutive cirexons, skipped/constitutive mRNA exons and randomly selected annotated exons, as well as their adjacent regions. The average binding density was calculated as the percentage of binding sites in the whole length of the sequence for prediction and shown in the heatmap after Z-score normalization. (b) Comparison for the above density between skipped cirexons and skipped mRNA exons with the corresponding statistical significances of Mann–Whitney U-test. *P-value <0.05; **P-value <0.01; ***P-value <0.001.

Figure 4. Features of four types of AS events in circRNAs.

(a–b) Four types of AS events are enriched in the nucleus. Transcriptomic data sets (poly(A)− and poly(A)+) from cytoplasm and nucleus RNA samples of seven cell lines in ENCODE were used to predict AS events. (a) Percentage of alternatively spliced cirexons in all detected cirexons within circRNA. (b) Percentage of alternatively spliced exons in all detected exons of mRNA. (c) Overlap of IR, ES and alternative 3′- or 5′-splicing site in HeLa, HEK293 and Hs68 samples treated by RNase R. (d) Comparison of retained intron length with other intron length within circRNA, as well as those in mRNA. (e) Sorted Ψ values of IR and other AS events in HeLa, HEK293 and Hs68. (f) An experimentally validated example of alternatively spliced cirexons: IR within circRNA chr10:105,197,772|105,198,565. Cirexons and splice junctions, as well as sequencing depth and corresponding BSJ read pairs, within the circRNA detected by CIRI-AS in RNase R-treated sample of HeLa cells. Exons and splice junctions, as well as sequencing depth and corresponding sequencing reads, in poly(A)-selected sample of HeLa cells. Different outward primers were designed and RT–PCR was performed separately for each isoform. Sanger sequencing was used to validate one BSJ junction and two FSJ junctions.

Figure 5. CircRNA AS events in 62 samples of D. melanogaster.

(a) Relative abundance of detected AS cirexons in all of the 62 samples. The cirexons can be clustered into three distinct groups according to their Ψ values in the tissues and cell lines. Yellow denotes cirexons with Ψ value equal to 1 and dark brown denotes cirexon with Ψ value equal to zero or no expression/detection. (b) Gene Ontology enrichment analysis for parental genes of alternatively spliced circRNAs in adult heads and laval/pupal CNS. BP, biological process; CC, cellular component; MF, molecular function. (c) Principal components analysis based on the Ψ values of the AS cirexons in each sample. Each dot represents a sample and its colour corresponds to the sample type on the top of a.

Figure 6. Both AC and AS contribute to the diversity of circRNAs.

(a) Back-splicing donors differ between the two AC isoforms, which can be mediated by complementary sequences within flanking introns of exon a/c and exon a/b, respectively. (b) The two AS isoforms share the same back-splicing acceptor and donor but differ in their internal structure.