Isolation and sequencing of AGO-bound RNAs reveals characteristics of mammalian stem-loop processing in vivo

Ian M. Silverman; Sager J. Gosai; Nicholas Vrettos; Shawn W. Foley; Nathan D. Berkowitz; Zissimos Mourelatos; Brian D. Gregory

Given this observation, we examined our dataset for novel pre-miRNAs embedded in other mRNAs. To do this, we took a conservative approach, using pre-miRNA-seq reads that failed to map to miRBase, and filtering them by . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/294488 doi: bioRxiv preprint 1 mapping to small nuclear RNAs (snRNAs), snoRNAs, tRNAs, ribosomal RNAs (rRNAs), and repeat-masked sequences (Fig. 5A, see Methods) . Approximately 1.2 and 0.37 million clones passed our stringent filtering steps and mapped to human and mouse mRNAs, respectively (Table S1 ). To identify pre-miRNA-like AGO-associated stem-loops derived from mRNAs, we used a CLIP-seq peak calling approach to identify significant (modified false discovery rate (mFDR) < 0.01) read clusters in mRNAs (35) . We found a number of significant peaks that corresponded to nearly the full length of some highly expressed genes (e.g ACTB). Therefore, we further filtered significant clusters based on their size (< This resulted in 403 and 373 AGO-associated stem-loops derived from human and mouse mRNAs, respectively (Table S6) . We intersected our list of AGO-associated stem-loops derived from human mRNAs with recently identified miRNAs in humans from extensive smRNA-seq or from DICER photoactivatableribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP) (26, 55) . In fact, 12 of our AGO-associated stem-loops were annotated as novel miRNAs in these lists supporting the ability of our approach to identify novel miRNAs (Fig. 5A) . Furthermore, 34 and 37 AGO-associated stem-loops overlapped with DICER and DGCR8 binding sites respectively, revealing that a number of the AGO-associated stem-loops derived from mRNAs interact with other components of the canonical miRNA processing pathway (55, 56) . The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/294488 doi: bioRxiv preprint We next examined the distribution of AGO-associated stem-loops derived from mRNAs across mRNA regions and found that they were equally present in all regions of mRNAs and similarly distributed between human and mouse (Fig. 5B ). When we normalized the distribution of AGO-associated stem-loops derived from mRNAs by the relative genomic coverage of each mRNA region, we found that the CDS was underrepresented, whereas the 5' UTR was enriched 2.5 fold for AGO-associated stem-loops in both organisms (Fig. 5C) . We also examined the size distribution of reads mapping to AGO-associated stem-loops derived from mRNAs and found them to be similar in size in both mammals, between 55-75 nt in humans and 52-80 nt in mouse (Figs. 5D-E) . This size range was slightly broader than pre-miRNA-seq reads that mapped to miRBase (Figs. 1E-F) .