23MicroRNAs (miRNAs) are key mediators of post-transcriptional gene 24 expression silencing. Although Drosophila has been of critical importance for miRNA 25 discovery, biogenesis and function, there has been no comprehensive experimental 26 annotation of functional miRNA target sites. To close this gap, we generated the first 27 in vivo map of miRNA::mRNA interactions in Drosophila melanogaster, making use 28 of crosslinked nucleotides in Argonaute (AGO) crosslinking and immunoprecipitation 29 (CLIP) experiments that enable an unambiguous assignment of miRNAs to AGO 30 binding sites at much higher signal-to-noise ratio than computational predictions 31 alone.
32Absolute quantification of cellular miRNA levels showed the miRNA pool in
33Drosophila cell lines to be more diverse than previously reported. Benchmarking two 34 different CLIP approaches, we identified a similar predictive potential to 35 unambiguously assign thousands of miRNA::mRNA pairs from AGO1 interaction 36 data at unprecedented depth. Quantitative RNA-Seq and subcodon-resolution 37 ribosomal footprinting data upon AGO1 depletion enabled the determination of 38 miRNA-mediated effects on target expression and translation. We thus provide the 39 first comprehensive resource of miRNA target sites as well as their quantitative 40 functional impact in Drosophila. 41 42 43 65 have greatly enhanced our understanding about miRNA function, while our 66 understanding of fly miRNA function is lagging behind. In Drosophila, a recent 67 comparative study of small RNAs across 25 cell lines suggested that the miRNA 68 landscape in non-ovary cell lines showed little diversity and low complexity in terms 69 of relative expression levels of individual miRNAs, which would argue against fly cell 70 lines as a good model to study miRNA function 12 . 71 4 Although knowledge of mature miRNA sequences alone have enabled to 72 identify physiologically relevant targets 5,17 , computational methods have greatly 73 contributed to successful miRNA target prediction, especially after recognition of the 74 miRNA seed region (nt 2-7) 18-21 . To date, there is a plethora of computational 75 miRNA target predictions tools, including popular approaches such as TargetScan, 76 MIRZA and mirSVR 2,22,23 , which leverage conservation, target sequence context 77 feature information or RNA::RNA hybridization energies and other features to 78 improve prediction accuracy. Purely computational tools predict miRNA targets sites 79 across entire 3'UTRs, neglecting cell type specific miRNA expression level and target 80 site availability, which can lead to numerous and tightly spaced predictions. In vivo 81 AGO binding information generated from Crosslinking and Immunoprecipitation 82 (CLIP) followed by sequencing (CLIP-seq or HITSCLIP) methods has been used to 83 greatly decrease the search space from whole 3'UTRs to about 30-40nt per AGO 84 footprint 24 . AGO footprints do not directly reveal the identity of the miRNA engaged, 85 and in many cases, multiple possible miRNA seed matches overl...