Fig. 4

CircRNAs interact with RBPs in a cell-type-specific manner. a Percentage of the exonic part of circRNAs that is covered with RBP binding sites in HepG2 (right) and K562 (left). Only the top 1% highest expressed circRNAs are shown. Some circRNAs are depicted with their host gene name. Some circRNAs are highly covered with RBP binding sites in both cell lines, e.g., circCDYL and circRBM33. Colors indicate the number of exons constituting each circRNA. We only consider circRNAs with annotated exons. b Comparison of RBP coverage for individual highly expressed circRNAs and all internal non-circ-exons in expressed genes in HepG2 (red) and K562 (blue). Common circRNAs that are highly covered with RBP binding sites in one or both cell lines are depicted with their host gene name. Stars indicate an RBP coverage in the 5% tail. c RBP coverage enrichment for the top 1% highest expressed circRNAs. The RBP coverage enrichment was calculated by dividing circRNA-RBP coverage with the mean RBP coverage of internal non-circ-exons of the host gene in HepG2 (red) and K562 (blue). Most circRNAs contain an equal amount or slightly more RBP binding sites than non-circ-exons in their parent gene (≥ 1× enrichment), while some circRNAs are highly enriched with RBP binding sites, e.g., circCDYL (> 10×) and circRBM33 (100×). d RBP binding sites in exons of the CDYL gene. Exons of the five main CDYL transcripts (thick part indicates protein-coding regions) and all circRNAs (stipulated lines) supported by at least two reads by both pipelines are shown. The highly expressed circCDYL (chr6:4891946-4892613; supported by 5163 (HepG2) and 2272 (K562) reads based on CIRI2) is indicated by a red line. This exon is highly covered by RBP binding sites in both cells. Zoom-in at the top shows the binding sites and names of the individual RBPs in both cell lines. All other CDYL-circRNAs are supported by less than 22 reads. Bars show the percentage of each exon covered with RBP binding sites. e Validation of circRNA-RBP interactions from RNA immunoprecipitation. All circRNA-RBP interactions, except for one, were verified. Here, RNA immunoprecipitation (RIP) experiments confirmed that circCDYL interacts with several RBPs in K562, e.g., UCHL5, IGF2BP1, IGF2BP2, but not with GRWD1. In HepG2, RIP experiments confirmed strong circCDYL-GRWD1 interactions. Predicted interactions between circSMARCA5 and UCHL5 in K562 and GRWD1 in HepG2 were also validated. 50-bp markers were used