The Drosophila hairpin RNA pathway generates endogenous short interfering RNAs

Abstract

In contrast to microRNAs and Piwi-associated RNAs, short interfering RNAs (siRNAs) are seemingly dispensable for host-directed gene regulation in Drosophila. This notion is based on the fact that mutants lacking the core siRNA-generating enzyme Dicer-2 or the predominant siRNA effector Argonaute 2 are viable, fertile and of relatively normal morphology. Moreover, endogenous Drosophila siRNAs have not yet been identified. Here we report that siRNAs derived from long hairpin RNA genes (hpRNAs) programme Slicer complexes that can repress endogenous target transcripts. The Drosophila hpRNA pathway is a hybrid mechanism that combines canonical RNA interference factors (Dicer-2, Hen1 (known as CG12367) and Argonaute 2) with a canonical microRNA factor (Loquacious) to generate approximately 21-nucleotide siRNAs. These novel regulatory RNAs reveal unexpected complexity in the sorting of small RNAs, and open a window onto the biological usage of endogenous RNA interference in Drosophila.

Figure 1. Examples of Drosophila hpRNA transcripts

a, hp-CG18854 contains a >400-bp duplex separated by a large loop; the enlarged region highlights the phased nature of small RNA duplexes. Northern probes were designed against the RNAs labelled in blue. b, The hp-CG4068 locus consists of 20 tandem repeats that partially overlap the 3′ UTR of CG4068 (Supplementary Fig. 3) and generate phased small RNA duplexes. Each repeat adopts a hairpin structure, but higher-order hairpins are possible because repeats are complementary to each other; ‘1-repeat’ and ‘2-repeat’ isoforms are shown. Distinct small RNAs were cloned from related repeats with minor sequence differences (for example, RNAs highlighted in red and green).

Figure 2. Distinct biogenesis pathways for miRNAs and hpRNAs

a, Unlike miRNAs (for example, Bantam), hpRNA biogenesis in S2 cells is highly dependent on Dcr-2 and AGO2; Loqs and AGO1 suppression affect both miRNA and hpRNA biogenesis. b, miRNA and hpRNA biogenesis in pharate adult Drosophila. miR-8 was affected only in the loqs mutant, whereas hpRNA products were strongly decreased in the Dcr-2 and loqs mutants (<10%), and significantly affected in the AGO2 and hen1 mutants. c, Modification of hpRNA-derived small RNAs is mediated by Hen1. Note that hpRNA products from hen1 mutants run as a range of faster-migrating species after β-elimination. CS, Canton S (a strain of fruitfly); KO, knockout deletion strain.

Figure 3. hpRNAs generate regulatory RNAs that can repress endogenous targets

a, In S2 cells, both single-repeat (1×) and double-repeat (2×) hp-CG4068 constructs specifically repressed the hp-CG4068B, hp-CG4068C and mus308 sensors; the hp-CG4068D sensor was unaffected. b, Both genomic (g) and cDNA (c) hp-CG18854 expression constructs specifically repressed the hp-CG18854B sensor. c, The 2′-O-methyl ASO against hp-CG18854B specifically derepressed the hp-CG18854B sensor, whereas ASO-CG4068B specifically relieved endogenous repression of the hp-CG4068B sensor (d). e, Compared to a perfect hp-CG4068B sensor, mutant sensors with a central bulge or central bulge plus seed mutations exhibited the same level of derepression. Error bars depict the standard deviation of eight transfections; statistical comparisons were performed with the unequal variance Students t-test; **P < 6 × 10⁻⁵, ***P < 1 × 10⁻⁸. f, Compensatory covariation (pink shaded box) between hp-CG4068B and mus308 target sites of D. melanogaster (Dm) and D. sechellia (Dsec). Red font, nucleotides that have evolved; sens, sensor. g, Endogenous RNA-induced silencing complex from S2R cleaved both perfect hp-CG4068B and mus308 sensors, and this activity was specifically competed away by cognate ASO.