Ntrons in Drosophila are often longer than other introns, the properties of long flanking introns and 5 exon bias are potentially linked. However, some prime examples of circular species with long flanking introns did not involve second exons. To test if we dissociate these features, we performed the analyses above using 684 circles that were not adjacent to first introns. We still observed progressive increases in flanking intron lengths with higher circular RNA accumulation, for both upstream and downstream introns. Although increased 
flanking intron length was observed in mammalian circular RNAs, these data suggest that long flanking introns are an intrinsic determinant for circularization in Drosophila. Relevance of Drosophila circular RNAs to miRNA regulation We assessed the extent to which miRNA sites Oleandrin reside on Drosophila circular RNAs, which predominantly involve 5 UTRs and coding regions. A relevant fact is that Drosophila coding regions exhibit much greater utilization of conserved miRNA target sites than mammals. We typically consider miRNAs that are conserved throughout Sophophora to be well-conserved, and catalogued 28 or 6merA seed matches to such miRNAs. We also implemented a stricter criterion for coding sites, in order to surpass simple coding constraints, by requiring putative sites of panDrosophilid miRNAs to be present in 11/12 genomes. As well, since miRNA strands frequently have conserved regulatory activity, we similarly catalogued matches to conserved miRNA strands. However, so as not to inflate site numbers, we kept their tallies separate, and also filtered their matches for lowcomplexity motifs. These analyses indicate that Drosophila circular RNAs bear a substantial population of constrained miRNA sites. From the most conservative view, there are ~150 deeply Cell Rep. Author manuscript; available in PMC 2015 December 11. NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript Westholm et al. Page 9 conserved 28 seed sites for pan-Drosophilid mature miRNAs strictly within UTRs. This rises to >800 for Sophophoran-conserved mature or star miRNAs of both target site definitions, again considering only UTRs. While miRNA sites are generally considered to reside in 3 UTRs, as circular RNAs frequently involve 5 UTRs, there are more conserved miRNA binding sites in circularized 5 UTRs than 3 UTRs. Thus, circular RNAs may explain the atypical location of some conserved miRNA sites. For example, the highestexpressed circular RNA derives from a 5 UTR/coding exon of muscleblind. The mbl circle is preferentially expressed in the nervous system, and contains highly conserved 5 UTR sites for several miRNAs, including neural miR-279 and miR-315. The numbers of putative miRNA binding sites on circular RNAs increase dramatically when considering coding regions. Even in the most cautious interpretation, focusing only on the most deeply-conserved 28 seed matches of pan-Drosophilid mature miRNAs, there are still ~1000 such sites. Many-fold higher numbers of coding sites were recorded with slightly less stringent parameters. To control for the possibility that coding regions might intrinsically encode amino acids that preferentially overlap miRNA seeds, we compared the numbers of deeply conserved seed matches on SAR 405 manufacturer pubmed ID:http://www.ncbi.nlm.nih.gov/pubmed/19846406 the antisense strands of circular RNA coding regions. The miRNAs with the highest number of strictly-conserved coding region seed matches had far fewer matches to the antisense strands of these circles. A s.Ntrons in Drosophila are often longer than other introns, the properties of long flanking introns and 5 exon bias are potentially linked. However, some prime examples of circular species with long flanking introns did not involve second exons. To test if we dissociate these features, we performed the analyses above using 684 circles that were not adjacent to first introns. We still observed progressive increases in flanking intron lengths with higher circular RNA accumulation, for both upstream and downstream introns. Although increased flanking intron length was observed in mammalian circular RNAs, these data suggest that long flanking introns are an intrinsic determinant for circularization in Drosophila. Relevance of Drosophila circular RNAs to miRNA regulation We assessed the extent to which miRNA sites reside on Drosophila circular RNAs, which predominantly involve 5 UTRs and coding regions. A relevant fact is that Drosophila coding regions exhibit much greater utilization of conserved miRNA target sites than mammals. We typically consider miRNAs that are conserved throughout Sophophora to be well-conserved, and catalogued 28 or 6merA seed matches to such miRNAs. We also implemented a stricter criterion for coding sites, in order to surpass simple coding constraints, by requiring putative sites of panDrosophilid miRNAs to be present in 11/12 genomes. As well, since miRNA strands frequently have conserved regulatory activity, we similarly catalogued matches to conserved miRNA strands. However, so as not to inflate site numbers, we kept their tallies separate, and also filtered their matches for lowcomplexity motifs. These analyses indicate that Drosophila circular RNAs bear a substantial population of constrained miRNA sites. From the most conservative view, there are ~150 deeply Cell Rep. Author manuscript; available in PMC 2015 December 11. NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript Westholm et al. Page 9 conserved 28 seed sites for pan-Drosophilid mature miRNAs strictly within UTRs. This rises to >800 for Sophophoran-conserved mature or star miRNAs of both target site definitions, again considering only UTRs. While miRNA sites are generally considered to reside in 3 UTRs, as circular RNAs frequently involve 5 UTRs, there are more conserved miRNA binding sites in circularized 5 UTRs than 3 UTRs. Thus, circular RNAs may explain the atypical location of some conserved miRNA sites. For example, the highestexpressed circular RNA derives from a 5 UTR/coding exon of muscleblind. The mbl circle is preferentially expressed in the nervous system, and contains highly conserved 5 UTR sites for several miRNAs, including neural miR-279 and miR-315. The numbers of putative 
miRNA binding sites on circular RNAs increase dramatically when considering coding regions. Even in the most cautious interpretation, focusing only on the most deeply-conserved 28 seed matches of pan-Drosophilid mature miRNAs, there are still ~1000 such sites. Many-fold higher numbers of coding sites were recorded with slightly less stringent parameters. To control for the possibility that coding regions might intrinsically encode amino acids that preferentially overlap miRNA seeds, we compared the numbers of deeply conserved seed matches on PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19846406 the antisense strands of circular RNA coding regions. The miRNAs with the highest number of strictly-conserved coding region seed matches had far fewer matches to the antisense strands of these circles. A s.
