Small silencing RNAs repress gene expression by a set of related mechanisms collectively called RNA-silencing pathways. In the RNA interference (RNAi) pathway, small interfering mRNA (siRNAs) defend cells from invasion by foreign nucleic acids, such as those produced by viruses. In contrast, microRNAs (miRNAs) sculpt endogenous mRNA expression. A third class of small RNAs, Piwi-interacting RNAs (piRNAs), defends the genome from transposons. This study reports that Drosophila piRNAs contain a 2'-O-methyl group on their 3' termini; this is a modification previously reported for plant miRNAs and siRNAs and mouse and rat piRNAs. Plant small-RNA methylation is catalyzed by the protein HEN1 (Yang, 2006; Li, 2005; Yu, 2005). DmHen1, the Drosophila homolog of HEN1, termed Pimet (piRNA methyltransferase) by Saito (2007) in a parallel study, methylates the termini of siRNAs and piRNAs. Without DmHen1, the length and abundance of piRNAs are decreased, and piRNA function is perturbed. Unlike plant HEN1, DmHen1 acts on single strands, not duplexes, explaining how it can use as substrates both siRNAs, which derive from double-stranded precursors, and piRNAs which do not. 2'-O-methylation of siRNAs may be the final step in assembly of the RNAi-enzyme complex, RISC, occurring after the Argonaute-bound siRNA duplex is converted to single-stranded RNA (Horwich, 2007; Saito, 2007).

In flies, both piRNAs (also known as repeat-associated siRNAs, rasiRNAs) and siRNAs, but not miRNAs, are modified at their 3' termini. The terminal nucleotide of Drosophila 0-2 hr embryo and mouse and bull testicular piRNAs was selectively labelled. The resulting ³²P-radiolabeled nucleoside 2' or 3'-monophosphates were resolved by 2D thin-layer chromatography (2D TLC) with a solvent system that can resolve nucleoside 2' monophosphates, nucleoside 3' monophosphates, and 2'-O-methyl nucleoside 3' monophosphates. Modified nucleoside monophosphates derived from the 3' termini of piRNAs were identified by comparison to modified and unmodified nucleoside 2' and 3' monophosphate standards. The terminal nucleotide of the piRNAs of all three animals comigrate with 2'-O-methyl nucleoside 3' monophosphate standards but not with any unmodified nucleoside monophosphate standard. Because mouse piRNAs were previously shown to contain 2'-O-methyl modified 3' termini by both mass spectrometry and a 2D TLC system, it is concluded that Drosophila and bull piRNAs also contain a 2'-O-methyl group at their 3' termini (Horwich, 2007).

In Arabidopsis, the RNA methyltransferase, HEN1, modifies the terminal 2' hydroxyl group of small silencing RNAs. In Drosophila, predicted gene CG12367, whose 1559 nucleotide mRNA encodes a 391 amino acid protein with a 220 amino acid evolutionarily conserved methyltransferase domain, most closely resembles Arabidopsis HEN1 (Park, 2002; Tkaczuk, 2006). For simplicity, this gene has been called Drosophila melanogaster (Dm) hen1. When homozygous, a piggyBac transposon insertion (PBac{WH}CG12367[f00810]) within the first intron of the fly hen1 gene reduces the accumulation of hen1 mRNA by 1000-fold in testes and by more than 40,000-fold in ovaries and can therefore be considered a null mutation, which is referred to as hen1^f00810 (Horwich, 2007).

The 3' termini of two types of highly abundant piRNAs were examined in the germline of flies heterozygous or homozygous for hen1^f00810. In testes, the Suppressor of Stellate [Su(Ste)] locus produces 24-27 nucleotide rasiRNAs, a subclass of piRNAs that directs silencing of the selfish genetic element Stellate. Su(Ste) rasiRNAs, like other Drosophila piRNAs, are modified at their 3' termini and therefore do not react with NaIO₄. In contrast, Su(Ste) rasiRNAs from hen1^f00810/hen1^f00810 mutant testes reacted with NaIO₄ and could therefore be β-eliminated to remove the last nucleotide of the RNA, thereby increasing their gel mobility and indicating that in the absence of DmHen1 protein, they are not modified. Similarly, rasiRNAs that guide silencing of roo, the most abundant retrotransposon in Drosophila melanogaster, were not modified in hen1^f00810 homozygous ovaries. The Su(Ste) and roo rasiRNAs were also shorter in the hen1^f00810 homozygotes. In contrast, the length and amount of miR-8, which is expressed in both the male and female germline, was unaltered in hen1^f00810 homozygotes. For both Su(Ste) and roo, rasiRNAs were on average shorter and less modified even in hen1^f00810 heterozygotes, compared to the wild-type, suggesting that the abundance of DmHen1 protein limits the stability or production of piRNAs in flies (Horwich, 2007).

Modification of the termini of Drosophila piRNAs plays a role in their function: mRNA expression from HeT-A, the element whose expression is most sensitive to mutations that disrupt piRNA-directed silencing in the female germline, quadrupled in hen1^f00810 heterozygotes and was increased by more than 11-fold in homozygotes, relative to wild-type tissue. It is concluded that Hen1 protein is required for piRNA-directed silencing in the Drosophila germline (Horwich, 2007).

To test whether DmHen1 is required for modification of the 3' termini of siRNAs, Hen1 was depleted by RNAi in cultured Drosophila S2 cells. The cells were transfected with long double-stranded RNA (dsRNA) targeting hen1 on day 1 and day 5, then cotransfected with both GFP dsRNA and hen1 dsRNA on day 8. Total RNA was harvested on day 9, probed for modification with NaIO₄/β-elimination, and analyzed by Northern hybridization with a 5' ³²P-radiolabeled DNA probe complementary to the most abundant GFP-derived siRNA. DsRNAs targeting two different regions of the fly hen1 mRNA both reduced the amount of GFP siRNA modified at its 3' terminus, whereas all the GFP siRNA remained modified when a control dsRNA was used (Horwich, 2007).

Surprisingly, RNAi-mediated depletion of Ago2, but not Ago1, prevented the GFP siRNA from being modified. This result suggests that Ago2, but not Ago1, plays a role in the modification of siRNAs by DmHen1. To test this idea, the modification status of the 3' terminus of miR-277, which partitions into both Ago1 and Ago2 complexes in vivo, was examined. Drosophila miRNAs associate predominantly or exclusively with Ago1 and have unmodified 3' termini. In contrast, approximately half the miR-277 in cultured S2 cells failed to react with NaIO₄, suggesting that approximately half of miR-277 is modified at its 3' terminus. The fraction of miR-277 that was modified was reduced when two different dsRNAs were used to deplete DmHen1 by RNAi. When the cells were treated with dsRNA targeting ago1, all detectable miR-277 was modified, whereas all miR-277 became unmodified when dsRNA targeting ago2 was used. In contrast, bantam, a miRNA that associates nearly exclusively with Ago1, was unmodified under all conditions (Horwich, 2007).

siRNA modification can be recapitulated in lysates of embryos, ovaries, or cultured S2 cells. Modification of siRNA in vitro was inhibited by S-adenosyl homocysteine, but not by S-adenosyl methionine, consistent with DmHen1 transferring a methyl group from S-adenosyl methionine to the terminal 2' hydroxyl group of the RNA, thereby generating S-adenosyl homocysteine as a product (Horwich, 2007).

Data from cultured S2 cells suggested that DmHen1 modifies that portion of miR-277 that enters the Ago2-RISC-assembly pathway, but not the population of miR-277 that assembles into Ago1-RISC. To further test the idea that small-RNA modification requires both Hen1 and the Ago2-RISC-assembly pathway, cytoplasmic lysates were prepared from dsRNA-treated cultured S2 cells. Lysate from control-treated cells modified the 3' terminus of a 5' ³²P-radiolabeled synthetic siRNA duplex but not lysate from hen1-depleted cells. The addition of either of two different preparations of purified, recombinant DmHen1, expressed in E. coli as a ~74 kDa glutathione S-transferase fusion protein (GST-DmHen1), restored the ability of the lysates to modify the siRNA, indicating that loss of DmHen1 caused the loss of siRNA modification. Moreover, lysates depleted for Ago2, but not Ago1, could not modify the ³²P-siRNA in vitro. These in vitro data, together with S2-cell experiments, suggest that modification of the 3' terminus of siRNAs and miRNAs is coupled to assembly into Ago2-RISC (Horwich, 2007).

Dcr-2 and R2D2 act to load double-stranded siRNAs into Ago2. Lysates were prepared from ovaries homozygous mutant for hen1, dcr-2, r2d2, and ago2 by using alleles that were unable to produce the corresponding protein. A 5' ³²P-radiolabeled siRNA duplex was incubated in each lysate to assemble RISC. At each time point, whether the siRNA was 3' terminally modified was determined by assessing its reactivity with NaIO₄. No modified siRNA accumulated when the duplex was incubated in hen1^f00810, dcr-2^L811fsX, r2d2¹, or ago2⁴¹⁴ mutant lysate. Adding 250 nM purified, recombinant GST-DmHen1 restored siRNA modification to the hen1^f00810 but not the ago2⁴¹⁴ lysate. It is concluded that the defect in ago2⁴¹⁴ reflects a requirement for Ago2 in small-RNA modification by DmHen1, rather than an indirect effect such as destabilization of DmHen1 in the absence of Ago2. GST-DmHen1 similarly rescued lysate from hen1(RNAi) but not ago2(RNAi)-treated S2 cells. Together, the results of experiments using cultured S2 cells -- a somatic-cell line -- and ovaries, which comprise mainly germline tissue, suggest that a functional Ago2-RISC-assembly pathway is required for siRNA modification in Drosophila (Horwich, 2007).

To test at which step in the Ago2-RISC-assembly pathway siRNAs become modified, whether siRNAs are 2'-O-methylated by DmHen1 as single strands or as duplexes was determined. In vitro, assembly of siRNAs into Ago2-RISC follows an ordered pathway in which the siRNA duplex first binds the Dicer-2/R2D2 heterodimer to form the RISC-loading complex (RLC). The RLC determines which of the two siRNA strands will become the guide for Ago2 and which will be destroyed (the passenger strand). The siRNA is then loaded into Ago2 as a duplex. In this pre-RISC complex, the passenger strand occupies the same position as future target RNAs. Cleavage of the passenger strand by the Ago2 endonuclease domain converts pre-RISC to mature RISC. No single-stranded guide or passenger RNA is produced prior to this maturation step. Thus, all single-stranded siRNA produced in vitro or in vivo corresponds to mature RISC (Horwich, 2007).

Ago2-RISC was assembled in vitro by using an siRNA designed to load only one of its two strands into Ago2. Then the reaction was sampled over time, isolating the 5' ³²P-radiolabeled siRNA under conditions previously demonstrated to preserve its structure, and single- from double-stranded siRNA was separated by native gel electrophoresis. The RNAs were then isolated from the gel and tested for reactivity with NaIO₄ to determine the presence of modification at their 3' termini. At each time, total siRNA was analyzed in parallel. 3' terminal modification increased over the course of RISC assembly and, at all times, was restricted to single-stranded siRNA: Within the limits of detection, all double-stranded siRNA was unmodified, even after 3 hr. It is concluded that siRNA modification is coupled to RISC assembly and occurs only after the conversion of pre-RISC to mature RISC (Horwich, 2007).

Whereas Arabidopsis HEN1 contains an N-terminal double-stranded RNA-binding motif, DmHen1 does not. To test whether DmHen1 modifies double-stranded small RNAs, purified, recombinant GST-DmHen1 was incubated with either single-stranded or double-stranded siRNAs. Modification, evidenced by loss of reactivity with NaIO₄, was detected only for the single-stranded RNA, suggesting that DmHen1 modifies single-stranded substrates, but not siRNAs or blunt RNA duplexes. A preference for single-stranded RNA would explain how DmHen1 could act on both siRNAs, which are born double stranded, and piRNAs, which are not. It is noted that the purified, recombinant GST-DmHen1 protein was more than 50-fold less active on its own than when supplemented with ovary lysate from hen1^f00810 homozygous flies. It is speculated that the Ago2-RISC machinery is required for Hen1 function in flies, although the possibility cannot be excluded that the lysate contains a factor (e.g., a kinase) required for activating Hen1 (Horwich, 2007).

Modification of single-stranded siRNAs -- that is, those loaded in fully mature Ago2-RISC but not double-stranded siRNAs might allow cells to distinguish siRNAs loaded successfully into functional complexes from those that fail to assemble. For example, if a 3'-to-5' nuclease acts to degrade single-stranded siRNAs, 2'-O-methylation of single-stranded siRNAs in Ago2 RISC may protect them from destruction. Moreover, such a nuclease might trim the 3' end of piRNAs. 2'-O-methylation of the piRNA 3' terminus may occur only when the length of RNA extending beyond the Piwi-family protein is short enough to permit the simultaneous binding of the final ribose sugar to the active site of DmHen1 and the interaction of DmHen1 with the Piwi protein itself. Modification of the terminus of the trimmed piRNA would then block further 3'-to-5' trimming of the small RNA, generating its Piwi-, Aubergine-, or Ago3-specific length. The observation that piRNAs are shorter in hen1^f00810 mutants supports this model (Horwich, 2007).

It is noted that all 2'-O-methyl-modified small RNAs identified thus far are associated with RISC complexes that efficiently cleave their RNA targets, i.e., Ago1-associated plant miRNAs, animal piRNAs, and Ago2-associated siRNAs in flies, whereas Drosophila miRNAs are typically both unmodified and associated with Ago1 RISC, which does not catalyze mRNA target cleavage in vivo. It is speculated that DmHen1 is recruited to RISC complexes containing single-stranded small silencing RNAs according to the identity of their Argonaute protein. This model predicts that DmHen1 will bind only to complexes containing fly Ago2 or the three fly Piwi proteins, Piwi, Aubergine, and Ago3, but not Ago1. Clearly, future experiments will need to test this hypothesis (Horwich, 2007).