PRogRess

Primary and secondary nematode siRNAs

Endogenous small interfering Until recently, C. elegans was the only animal
for which endo-siRNAs had been well char-
acterized. Primary siRNAs that are processed
RNAs in animals by dicing dsRNA are exceedingly rare in this
organism22–24. Instead, the 3′ ends of targets
that have been cleaved by primary siRNAs
Katsutomo Okamura and Eric C. Lai
are recognized by an RNA-dependent RNA
Abstract | Until recently, only nematodes among animals had a well-defined polymerase (RdRP), which generates abun-
dant, untemplated, secondary siRNAs with
endogenous small interfering RNA (endo-siRNA) pathway. This has changed
distinctive 5′ triphosphates (FIG. 3). Secondary
dramatically with the recent discovery of diverse intramolecular and intermolecular siRNAs are then loaded into specialized
substrates that generate endo-siRNAs in Drosophila melanogaster and mice. These secondary Argonautes (SAGOs)25. Because
findings suggest broad and possibly conserved roles for endogenous RNA other animals do not seem to encode RdRP
interference in regulating host-gene expression and transposable element or SAGOs, it is not evident that the mecha-
transcripts. They also raise many questions regarding the biogenesis and function nism for worm siRNA biogenesis is broadly
conserved. Nematodes also lack conventional
of small regulatory RNAs in animals.
piRNAs, as the Piwi homologue PRG-1
contains ‘21U’ RNAs. The biogenesis of these
RNA interference (RNAi), the process by biological roles of siRNAs. We highlight these 21 nt RNAs does not seem to be related to
which double-stranded RNA (dsRNA) is new classes of endo-siRNAs and the pressing that of fly or vertebrate piRNAs22,26–28 (FIG. 2b).
processed into small interfering RNAs questions that are raised by their discovery. Therefore, fundamental aspects of conserved
(siRNAs) that silence homologous tran- animal small-RNA pathways have clearly
scripts, is both a fascinating cellular machin- Argonaute-bound small RNAs been altered in C. elegans.
ery and a powerful experimental technique. Argonaute proteins lie at the heart of related
Despite an avalanche of RNAi research over small-RNA pathways that operate in organ- Endo-siRNAs in flies and mice
the past decade, however, a nagging question isms as diverse as Archaea, plants and Recent work now reveals diverse sources
remained mostly unanswered: what good is animals14. They bind various small RNAs of endo-siRNAs in D. melanogaster and in
RNAi to the organism itself? that are <32 nucleotides (nt) in length which mouse. Most of these endo-siRNA classes
Substantial roles for RNAi in regulating guide the Argonaute complexes to their seem to be analogous between species, and
endogenous gene expression have been regulatory targets (FIG.1). include those derived from transposable ele-
difficult to ascertain because Drosophila Among animals, the AGO and Piwi ments, from complementary annealed trans-
melanogaster1,2 and Caenorhabditis elegans3,4 subclasses constitute two main classes of cripts, and from long ‘fold-back’ transcripts
mutants that selectively inactivate RNAi seem conserved Argonaute proteins. AGO proteins called hairpin RNAs (hpRNAs).
to be normal and fertile. These mutants are bind to microRNAs (miRNAs)14,15 — RNAs
hypersensitive to viruses, which suggests that of ~22 nt that derive from host transcripts siRNAs from transposable elements. Because
RNAi defends against selfish and invasive with short (usually <100 nt) inverted repeats. of the mutagenic consequences of transposa-
nucleic acids5. But if RNAi had an ancestral These repeats are processed by the RNase III ble elements (TEs), powerful mechanisms are
role in virus restriction it seems to have been enzymes Drosha (in the nucleus) and Dicer needed to restrict their activity. Such protec-
subsumed in vertebrates by the interferon (in the cytoplasm) (FIG. 1a). Specialized AGO tion is indispensable in the germ line to main-
pathway. In fact, the nonspecific capacity of proteins with efficient ‘slicing’ activity are tain faithful transmission of the genome. In
dsRNA to activate the interferon response, the carriers of 21 nt siRNAs2,16,17. Exogenous this context, piRNAs mediate a major defence
thereby leading to the general inhibition of dsRNAs are processed into siRNAs in the against TEs29. However, scattered reports in
cellular translation, was widely perceived to cytoplasm by Dicer, and therefore they do the literature indicated that canonical RNAi
preclude substantial roles for endogenous not require Drosha (FIG. 1b). Piwi-interacting also influences TEs. This was most clear in
RNAi in vertebrates. RNAs (piRNAs) are slightly longer RNAs C. elegans, because many RNAi-defective
Eight concurrent papers from the (~24–32 nt) that are bound by Piwi-family mutants also deregulate transposons3,30. It was
Zamore, Sasaki, Siomi, Lai and Hannon proteins, which also have slicer activity18–20. proposed that transcriptional read-through
laboratories recently described a rich diver- Although their biogenesis is not completely across Tc1 transposable elements might
sity of endogenous siRNAs (endo-siRNAs) understood, a major pathway for piRNA produce intramolecular dsRNA between the
in mice6,7 and D. melanogaster8–13. These production involves reciprocal cleavages of terminal inverted repeats, the processing of
studies introduce unanticipated complexity sense and antisense substrates by antisense which by RNAi could generate siRNAs that
in small-RNA sorting pathways and in the and sense piRNAs, respectively19,21 (FIG. 2a). silence Tc1 elements in trans31.

NATURE REvIEWS | molecular cell biology vOLUME 9 | SEPTEMBER 2008 | 673

Progress

a miRNA b exo-siRNA c endo-siRNA

Hairpin RNA
Drosha Pasha
Viral dsRNA
Exogenous dsRNA TEs
AAAAAAA

trans-NAT dsRNA
R2D2 DCR2
cis-NAT dsRNA
DCR1 LOQS ?
LOQS DCR2
exo-siRNA ? endo-siRNA

miRNA miRNA*
miRNA* R2D2 DCR2 ???
degradation
AGO1 AGO1
AGO2 2Ome AGO2 2Ome
AGO2 AGO1 AGO2 AGO1 AGO1
2Ome AGO1 2Ome AGO2 2Ome AGO2 2Ome

mRNA translational repression Virus defence Transposon silencing

mRNA deadenylation/degradation mRNA cleavage mRNA cleavage
(mRNA cleavage?)
Figure 1 | Small rNa pathways in Drosophila melanogaster. In Drosophila deadenylation by Ago1. miRNA* is the species
Nature on the
Reviews other sideCell
| Molecular of the hair-
Biology
melanogaster, micro (mi)RNAs are ~22 nucleotides (nt) long, have free hydroxy pin to the miRNA. b | In D. melanogaster, viral dsRNA or artificial dsRNA pro-
groups at their 3′ ends, and associate primarily with the Argonaute protein duce exogenous siRNAs (exo-siRNAs) that are mostly sorted to Ago2 and
Ago1. small interfering (si)RNAs are ~21 nt, are methylated at their 3′ ends, restrict viral replication or cleave designed targets. c | D. melanogaster cells
and associate primarily with Ago2. Three main protein families are denoted and mammalian oocytes produce several sources of endogenous dsRNA —
with RNase III enzymes (Drosha, Dicer-1 (DCR1) and DCR2; shown as hexa- transposable elements (Tes), cis-natural antisense transcripts (cis-NATs), trans-
gons), their dsRNA-binding domain (dsRBD) partners (Pasha, Loquacious NATs and hairpin RNA transcripts — that are processed into endo-siRNAs that
(LoQs) and R2D2; shown as squares) and Argonaute proteins (Ago1 and load mostly Ago2. These repress transposon transcripts or endogenous
Ago2; shown as ovals). a | miRNA pathway. endogenous transcripts that mRNAs. Note that a minority of miRNAs programme Ago2 and a small
contain short inverted repeats are processed into ~21–22 nt RNAs that mostly fraction of exo- and endo-siRNAs associate with Ago1, but the functional
function to repress endogenous targets by translational repression and significance of this is currently unknown. 2ome, 2′-O-methyl group.

A conundrum for mammalian piRNA In D. melanogaster, deep sequencing of can involve 5′, 3′ or internal exons (FIG. 4).
studies was that although multiple mouse the small RNAs that directly associate with Careful analysis of small-RNA sequences
Piwi-gene mutants exhibit testicular defects, AGO2 (the Argonaute that mediates RNAi) in mouse oocytes7 and D. melanogaster
transposon activation and sterility, cor- revealed that TEs are a substantial source of tissues and cultured cells8,9,11,12 revealed
responding mutant ovaries were normal and RNAs of precisely 21 nt11,12. Similar conclu- that cis-NAT overlaps are favourable for
functional32–34. Instead, Dicer-mutant ovaries sions were reached by sequencing RNAs that siRNA production. The extent of 21 nt
and oocytes exhibit higher levels of certain were β-eliminated — this prevents RNAs RNA production was limited to annotated
retrotransposon transcripts35,36. This is con- from being ligated on their 3' ends, unless exons that are transcribed bidirectionally,
sistent with either an miRNA-based system they bear a 3' modification, and thus enriches excluding adjacent introns. D. melanogaster
for TE control or perhaps the usage of endo- AGO2-loaded RNAs8 — or by analysing total cis-NAT-siRNAs are dependent on DCR2,
siRNAs. In fact, earlier small-scale sequencing head or cultured-cell RNAs13. Their accumu- and mouse cis-NAT-siRNAs are similarly
from mouse oocytes and testes revealed lation is dependent on DCR2 (one of the two Dicer-dependent. However, although virtu-
that some siRNAs derived from retrotrans- Dicers in D. melanogaster, and the one that ally all cis-NAT-siRNAs in flies derived from
posons37, which could silence long inter- generates exogenous siRNAs14; FIG. 1c), and 3′ untranslated region (UTR) overlaps, one
spersed nuclear elements (LINEs) in trans38. the depletion or mutation of either DCR2 or of the abundant mouse cis-NAT-siRNA loci
Newer large-scale cloning provided clearer AGO2 elevates TE transcript levels8,11–13. The involved Pdzd11/Kif4, whose transcripts
evidence for TE-siRNAs in mouse oocytes6,7. TE-siRNA response is extremely active in overlap on their 5′ UTRs (FIG. 4).
Many of these mapped to the same genomic various lines of cultured cells and correlates The levels of the 3′-overlapping tran-
locations as piRNA clusters, which raised the with the strong genomic amplification of scripts Pdzd11 and Kif4 increased modestly
possibility that these specialized ‘master loci’ specific LTR retrotransposons in these cells. in mouse Dicer mutants7, consistent with
are involved in both piRNA-mediated and Therefore, both TE-siRNAs and TE-piRNAs an autoregulatory activity of the siRNAs
siRNA-mediated TE control. However, some repress transposon transcripts in flies and generated by this cis-NAT. D. melanogaster
transposon classes were apparently targeted mammals (FIG. 4). cis-NAT-siRNAs specifically load AGO2,
by only one of these RNA classes, which sug- but evidence for changes in their progenitor
gested that piRNAs or siRNAs preferentially siRNAs from cis-natural antisense transcripts. transcripts on loss of DCR2 or AGO2 was
control certain TEs. For example, several Cis-natural antisense transcript (cis-NAT) equivocal. However, D. melanogaster cis-
long-terminal repeat (LTR) retrotransposons arrangements are genomic regions that NAT-siRNA genes (but not cis-NAT genes
were nearly exclusively targeted by siRNAs6,7. encode exons on both DNA strands, and in general) exhibited striking enrichment

674 | SEPTEMBER 2008 | vOLUME 9 www.nature.com/reviews/molcellbio

 Progress

for several nucleic-acid-based functions, Earlier functional tests showed that long a piRNA
including transcription cofactors, deoxyribo- dsRNA does not activate protein kinase R Active
nucleases and ribonucleases9. In addition, or the interferon response in oocytes, as it Piwi/AUB transposon
most co-expressed cis-NATs in D. mela- does in most other mammalian cells39,40.
nogaster S2 cells did not generate siRNAs. Therefore, oocytes might provide a favourable 3v trimming
These data indicate that only a subset of setting for the exploitation of endogenous
co-expressed cis-NAT pairs are selected for RNAi to regulate host transcripts. Genes
Piwi/AUB AGO3
siRNA production, presumably reflecting with complementary pseudogene siRNAs
an endogenous functional use. Intriguingly, are heavily enriched for microtubule-related
one of the most highly expressed cis-NAT- functions6. This suggests a regulatory focus 3v trimming
siRNA loci in the entire genome involves the to the trans-NAT-siRNA pathway.
CG7739/Ago2 gene pair8,9,12 — thus AGO2 piRNA AGO3
carries its own siRNAs. siRNAs from hpRNA transcripts. Although cluster
A special class of cis-NAT-siRNAs come animal miRNA hairpins are usually <100 nt, transcript
from the D. melanogaster klarsicht (klar) plant miRNA hairpins can be significantly
AUB Piwi
gene, which is involved in lipid-droplet longer41. Because of this property, the hairpin 2Ome 2Ome
transport and nuclear migration, and from precursors of some plant miRNAs were not AGO3
2Ome
the thickveins (tkv) gene, which is involved in initially recognized. Likewise, some ‘long’ Germline transposon silencing
transforming growth factor-β signalling9,12. miRNA hairpins that are double the length of
Although these loci produce 3′ modified, typical miRNAs were only recently identified b Nematode 21U RNA
21 nt, AGO2-bound RNAs from both DNA in D. melanogaster 42. Therefore, animal RNAs 21U RNA
strands, they seem to involve a specialized that map to inverted repeats might have
mechanism for extremely efficient cis-NAT- escaped conventional miRNA annotation. ?
siRNA production over extended genomic Bioinformatics studies in D. melanogaster
intervals that are 5–10 kb in length9,12. In revealed a number of candidate loci that PRG-1
2Ome?
addition, klar and tkv are not 3′ cis-NATs, produce small RNAs from extended inverted Regulation of spermatogenesis transcripts
but instead involve overlaps with 5′ exons, repeats that are termed hairpin RNAs Repression of Tc3 transposition
internal transcript exons and/or annotated (hpRNAs), the stems of which were up to 400 Figure 2 | Specialized small-rNa regulatory
intronic regions. Therefore, the strategy for base pairs in length10. At least seven distinct pathwaysNaturein theReviews
animal| Molecular
germ line. These
Cell are
Biology
klar and tkv siRNA production seems to differ loci generate siRNAs, and the hp-CG4068 mediated by Piwi-class Argonaute proteins
(ovals). a | The Piwi-interacting (pi)RNA pathway
from that of conventional cis-NAT-siRNAs. locus alone encodes 20 tandem hairpins10–12.
operates in the Drosophila melanogaster and
Despite their structural similarity to miRNAs, vertebrate germ line. A ‘ping-pong’ strategy
siRNAs from mammalian pseudogene– hpRNAs are processed by DCR2 instead of amplifies piRNAs from complementary trans-
gene pairs. Mammalian genomes encode DCR1, and generate 3′ blocked siRNAs that cripts, in which the slicer activity of Piwi proteins
large numbers of pseudogenes, which are load AGO2 (REFS 10–12) (FIG. 1). As with (Piwi, Aubergine (AUB) and Ago3 in D. melano­
presumed to be non-functional entities that siRNAs from artificial long-inverted repeats, gaster) reciprocally define piRNA 5′ ends. The
will eventually be lost. Small-RNA cloning the siRNA duplexes derived from hpRNAs mechanism that defines the 3′ ends of piRNAs is
from mouse oocytes revealed an unexpected are phased and direct AGO2 to cleave targets. not known. A conserved role of the piRNA path-
class of ‘functional’ pseudogenes. Multiple One of the hp-CG4068 siRNAs is highly way is to restrict transposon activity in the germ
genes with antisense-transcribed pseu- complementary to the coding region of line; however, there might be other roles for
abundant non-transposon-derived piRNAs that
dogenes were inferred to anneal with their mutagen-sensitive-308 (mus308), a DNA
are found in mammals. b | Nematode 21U RNAs
complementary progenitors (as trans-NATs) polymerase that is involved in the DNA- might be a functional analogue of piRNAs. These
and be diced into siRNAs6,7. The existence damage response, and can cleave this target 21-nucleotide RNAs begin with U and are pro-
of siRNAs that bridge exon–exon junctions site10,12. In this case, mus308 is the only obvi- duced from genomic loci with a characteristic
suggested that mature mRNAs constitute ous target of the many siRNAs that are gener- upstream motif (CTgTTTCA), and they are bound
the dsRNA substrate, as suggested for cis- ated by hp-CG4068. However, hp-CG18854 by the Piwi protein PRg-1. The details of 21U
NAT-siRNA pairs. Microarray profiling and is a pseudogene with substantial homology biogenesis and function are unclear, but 21Us are
quantitative PCR analysis of Dicer-mutant to CG8289, which encodes a chromodomain linked to spermatogenesis and control of Tc3
oocytes revealed substantial upregulation of protein, and elevated hp-CG18854 could transposition. 2ome, 2′-O-methyl group.
multiple genes with complementary siRNAs repress CG8289 in trans10. Curiously, several
(FIG. 4), indicating that this system regulates candidate hpRNA loci were identified in
endogenous gene expression6,7. It is unclear mouse, including a long-inverted repeat Fly endo-siRNAs require Loquacious
whether the dicing of targets during trans- pseudogene of the Ran GTPase-activating There are two Dicers in D. melanogaster
NAT-siRNA biogenesis accounts for target protein-1 (Rangap1) gene6,7. It is unclear — DCR1 cleaves pre-miRNA hairpins into
regulation, or whether pseudogene-derived whether these hpRNA pathways are miRNA duplexes, whereas DCR2 cleaves
antisense siRNAs actively slice sense-strand conserved or convergent, but they at least long dsRNA into siRNA duplexes14 (FIG. 1).
mRNAs (FIG. 1c). In at least one case — histone suggest that analogous systems operate in Each Dicer directly binds to a dsRNA-
deacetylase-1 (Hdac1) — siRNAs derived D. melanogaster and mammals. However, it is binding domain (dsRBD) partner that aids
exclusively from sense–antisense pseudogene clear that entry into an endo-siRNA pathway its function. DCR2 interacts with R2D2
duplexes, which were inferred to repress can endow pseudogenes in both species with (whose name derives from the fact that it
functional Hdac1 trancripts6. regulatory activity. contains two dsRNA-binding domains (R2)

NATURE REvIEWS | molecular cell biology vOLUME 9 | SEPTEMBER 2008 | 675

Progress

Exogenous dsRNA co-expressed cis-NATs in D. melanogaster9. Is

Endogenous trigger there active selection for entry into the RNAi
pathway, which could be mediated at the step
DRH-1/2 of dsRNA formation? Conversely, how do
RDE-4 DCR-1 co-expressed mammalian cis-NATs, and co-
Normal transcripts? expressed pseudogene–gene complementary
1° siRNAs m7G AAAAAA pairs, avoid triggering an interferon response
PPP outside of oocytes? Finally, although it seems
P RDE-1
RdRP
evident that cis-NAT and trans-NAT siRNAs
?
are generated from processed transcripts, it
m7G AAAAAA is not known whether the dsRNA substrate
P
RDE-1
Cleaved (or broken) forms in the nucleus or cytoplasm, nor is it
3v end 2° siRNAs PPP SAGOs
transcripts?
formation clear where the dsRNA encounters Dicer.
m7G PPP CSR-1
PPP PPP SAGOs
valuable lessons were taught by the
mRNA cleavage
RdRP Translational repression? transcriptional silencing? length and structure of primary hpRNA
Figure 3 | Nematode small interfering rNa pathways. Processing of double-stranded (ds) substrates transcripts. Their dsRNA character was
by a complex that includes an RNAse III enzyme (Dicer-1 (DCR-1), a dsRNA-binding partner (dsRBD; recognized only after genomic fragments
RDe-4) and Dicer-related helicases (DRH-1/2) generates primary (1°)Reviews
Nature small interfering
| Molecular(si)RNAs with
Cell Biology of sufficient length were examined, and
5′ monophosphates (P). These load into the RDe-1 Argonaute protein and can slice complementary consequently their siRNAs were prone to
transcripts. Cleaved transcripts, and possibly uncleaved transcripts, are substrates for an RNA- being misannotated as having derived from
dependent RNA polymerase (RdRP) complex that generates secondary (2°) siRNAs that have shorter, unstructured precursors8. The stems
5′ triphosphates (PPP). These are loaded into various secondary Argonautes (sAgos) that lack slicer of some plant miRNA hairpins are separated
activity, or into the Argonaute slicer CsR-1. by long, unstructured terminal loops and
even introns54, and we now recognize the
and is associated with DCR2 (D2)), which is some of the members of all of the siRNA same to be true for several hpRNAs10,12. It is
essential for the loading of siRNA into AGO2 classes — TE-siRNAs, cis-NAT-siRNAs therefore conceivable that the stems of some
(REFS 43,44). DCR1 interacts with Loquacious and hpRNA-siRNAs — are dependent on hpRNA precursors might be separated by
(LOQS), which promotes its ability to LOQS. Although previous tests did not kilobases or tens of kilobases. Do the struc-
cleave pre-miRNA hairpins, the products of reveal a physical interaction between LOQS tured precursors of any anonymous cloned
which are preferentially loaded into AGO1 and DCR2, proteomic analysis of DCR2 small RNAs that are currently deposited in
(REFS 45–47). complexes revealed that there is comparable public databases await discovery?
Although the attractive symmetry of coverage of LOQS and R2D2 peptides12. Endogenous sources of mammalian
RNase III, dsRBD and AGO partnerships in Therefore, LOQS is a component of both dsRNA remain to be recognized outside of
the RNAi and miRNA pathways lent support miRNA and RNAi pathways. oocytes. As is the case with oocytes, intro-
to the proposed division of these pathways, duction of long dsRNA into embryonic stem
genetic observations suggested that there Endo-siRNA biogenesis: open questions cells (ESCs) does not activate an interferon
are much more complex interactions among The recent papers on endo-siRNAs raise fun- response55,56. Might ESCs also harbour endo-
these factors. For example, unlike Dcr2 damental questions regarding the biogenesis siRNAs, the action of which is relevant for
mutants, r2d2 mutants reveal its requirement of small RNAs. Some of the most important maintaining pluripotency? Although endo-
for early development and female fertility. questions concern mechanistic aspects of siRNAs were not previously found in ESCs57
Moreover, r2d2 (but not Dcr2) phenotypes small-RNA sorting pathways. For example, this possibility might deserve further study.
are strongly enhanced on reduction of Dcr1 how does LOQS work with DCR2? And Finally, although small-RNA sorting
(REF. 48). Reciprocally, Dcr1 proved to be an given that R2D2 is needed to load exo- pathways have received little attention in
RNAi-defective mutant1. There is substantial siRNAs into AGO2 (REFS 43,44), to what mammalian systems, there is growing rec-
functional overlap between AGO1 and extent do endo-siRNAs require R2D2 for ognition of their importance to siRNA and
AGO2, as detected by double-mutant analy- loading? How are miRNA and hpRNA pre- miRNA function in plants58,59, worms60,61 and
sis49, and some miRNAs sort to both AGO1 cursors distinguished? Some ‘long’ miRNAs flies51,62. As only one of the four mammalian
and AGO2 (REFS 11,12,50–52). Finally, loqs and ‘short’ hpRNAs in D. melanogaster are AGO proteins (AGO2) has slicer activity16,17,
functions in inverted-repeat RNA-mediated indistinguishable in size and structure10,42. the directed sorting of mammalian siRNAs
silencing46. These findings indicate that there They are effectively sorted, however, as long is presumably important for their ability to
is substantial crosstalk between the RNAi miRNAs make only a single small-RNA slice complementary targets63. Consequently,
and miRNA pathways. duplex (as is typical for DCR1 substrates), the elucidation of mammalian siRNA sorting
Despite its original classification as a core whereas short hpRNAs produce multiple rules might have important implications
component of the miRNA pathway, loqs-null duplexes (as is typical for DCR2 substrates). for attempts to improve siRNA efficacy for
mutants have only modest defects in the How can the cell distinguish these hairpins? experimental and therapeutic purposes.
maturation of many miRNAs53. It seems The regulation of dsRNA formation is
that DCR1 can cleave pre-miRNAs without another mystery. For example, the cis-NAT- The biology of endo-siRNAs
LOQS, albeit with lowered efficiency that siRNA pathway accepts many substrates To return to the question posed at the begin-
varies between miRNAs53. Surprisingly, — at least 17 in mouse oocytes6,7 and at ning of this Perspective, what good is endo-
LOQS is essential for the accumulation of least 140 in D. melanogaster 8,9,12. However, genous RNAi to an organism? The necessity
many endo-siRNAs9,10,12,13 (FIG. 1c). At least cis-NAT-siRNA loci constitute only 25% of to preserve RNAi in mammals has been

676 | SEPTEMBER 2008 | vOLUME 9 www.nature.com/reviews/molcellbio

 Progress

Loci collected
siRNA Gene structure dsRNA structure Fly Mouse
TE-siRNA ? Many Many
Transposable element

cis-NAT-siRNA 3′ overlap
AAAAAAAA
140+
AAAAAAAA
Convergent
5′ overlap
AAAAAAAA Unknown
AAAAAAAA 17 (+28*)
Divergent

Unknown 2
Intronic
trans-NAT-siRNA AAAAAAAA
AAAAAAAA Unknown 15**

hpRNA Inverted repeat

AAAAAA 7 (+19?) 4

Figure 4 | Substrates for endo-sirNa production in flies and mouse. The trans-natural antisense transcript (trans-NAT) pairs. Trans-NAT dsRNAs
precise structure of the double-stranded (ds)RNA substrates of small inter- Nature Reviews
form between transcripts that are produced Molecular
from |distinct Cell Biology
genomic loca-
fering (si)RNAs derived from transposable elements (Tes) is unknown, but tions, and usually comprise an mRNA and an antisense-transcribed pseudo-
hundreds or thousands of Tes are inferred to directly generate siRNAs. gene. Based on the cumulative data of Tam et al.6 and Watanabe et al.7, 15
siRNAs derived from cis-natural antisense transcripts (cis-NATs) involve trans-NAT gene–pseudogene pairs generate siRNAs (**; some of the 28
bidirectional transcription across the same genomic DNA, and can be mRNAs listed in the cis-NAT-siRNA category might have antisense pseudo-
convergent, divergent or involve annotated introns and/or internal exons. genes that were not reported). siRNAs that are derived from hairpin RNA
Drosophila melanogaster cis-NAT-siRNAs derive almost exclusively from (hpRNAs) are long, inverted repeat transcripts whose double-stranded
3′-overlapping mRNAs, but two highly-expressed siRNA loci include anno- segment is typically much longer than that of miRNA precursors. In
tated introns. Watanabe and colleagues describe 17 cis-NAT-siRNA loci7, D. melanogaster, one of the 7 identified hpRNA loci encodes 20 hairpin
but their precise categorization is ambiguous as many of them lack an direct repeats, which can function autonomously or as components of
annotated overlapping transcript. Tam and colleagues describe another higher-order hairpins. The figure shows the numbers of loci that were col-
28 mRNAs (*) with siRNAs whose complementary transcript was not spe- lected from the recently published studies, but these numbers will probably
cifically described6. These might be cis-NATs, but some might represent increase with future studies.

somewhat of an enigma as they seem to have in D. melanogaster10,12, and such evidence demonstrated by specific knockouts of
mostly dispensed with siRNAs for antiviral would be desirable for other classes of hpRNAs or siRNA-generating pseudogenes.
defence, and some aspects of mammalian endo-siRNAs. Overall, the fact that core RNAi pathway
biology can be rescued by slicer-defective The established targets of D. melanogaster mutants in worms and flies are mostly nor-
AGO2 (REF. 64). However, in addition to a few hpRNAs encode DNA-binding proteins10,12. mal and fertile, whereas core miRNA path-
endogenous cleavage targets of miRNAs65, This seems reminiscent of the fact that way mutants are lethal, suggests that the role
and a role for AGO2 in the biogenesis of D. melanogaster cis-NAT-siRNA loci of endogenous RNAi is fundamentally dif-
select miRNAs66, the new studies suggest are significantly enriched for DNA and ferent than that of miRNA regulation. This
widespread usage of endo-siRNAs as endog- RNA-binding proteins9, raising this as a is further suggested by the fact that many
enous regulators of gene expression. substantial molecular axis for endo-siRNA miRNAs are deeply conserved but most
However, it is safe to say that we do not regulation. It is relevant to note, therefore, D. melanogaster hpRNA loci10,12 and most
understand the specific biological functions that D. melanogaster Dcr2 mutants exhibit mouse pseudogenes that generate siRNAs6,7
of endo-siRNAs well. Indeed, the question of abnormal nucleolar morphology69, whereas are poorly conserved. We must therefore
endo-siRNA function remains mostly Ago2 mutants were reported to have think more openly about their usage. Is the
unanswered in worms22,67,68, and the discov- chromosome segregation defects70. These usage of these RNAs a matter of fine-tuning
ery of abundant endo-siRNAs in flies and phenotypes are plausibly connected to gene expression, or perhaps a matter of
mammals only makes the understanding of the types of gene functions that are highly maintaining fitness in an ever-changing
this topic more pressing. The recent papers enriched in cis-NAT-siRNAs. The mouse environment? Is endogenous RNAi used for
do show deregulation of retrotransposon oocyte pseudogene–gene siRNA system robustness in gene regulation, perhaps to
transcripts, pseudogene-complementary seems to preferentially target genes that are canalize traits? Or is it a regulatory mecha-
transcripts and some cis-NAT pairs in Dicer involved in microtubule dynamics6, and this nism that generates species-specific charac-
and/or Ago mutants, and thus their regula- is plausibly connected to the observation ters during evolution? These are questions
tion by endo-siRNAs is plausible, although that Dicer loss in growing oocytes disrupts that remain for the future, but given the pace
this remains to be shown directly. Evidence spindle formation and chromosome segre- with which the field of endo-siRNAs has
for direct siRNA-mediated target regulation gation35,71. Nevertheless, the endogenous recently advanced, we might expect some
was only explicitly shown for some hpRNAs requirement of these systems remains to be answers to soon be forthcoming.

NATURE REvIEWS | molecular cell biology vOLUME 9 | SEPTEMBER 2008 | 677

Progress

Katsutomo Okamura and Eric C. Lai are at the 27. Batista, P. J. et al. PRG‑1 and 21U‑RNAs interact to 53. Liu, X. et al. Dicer‑1, but not Loquacious, is critical for
Sloan-Kettering Institute, form the piRNA complex required for fertility in assembly of miRNA‑induced silencing complexes. RNA
C. elegans. Mol. Cell 31, 67–78 (2008). 13, 2324–2329 (2007).
Department of Developmental Biology,
28. Das, P. P. et al. Piwi and piRNAs act upstream of an 54. Sunkar, R., Girke, T., Jain, P. K. & Zhu, J. K. Cloning and
521 Rockefeller Research Laboratories, 1275 York endogenous siRNA pathway to suppress Tc3 characterization of microRNAs from rice. Plant Cell 17,
Avenue, BOX 252, New York, New York 10065, USA. transposon mobility in the Caenorhabditis elegans 1397–1411 (2005).
germline. Mol. Cell 31, 79–90 (2008). 55. Paddison, P. J., Caudy, A. A. & Hannon, G. J. Stable
Correspondence to E.C.L.
29. Aravin, A. A., Hannon, G. J. & Brennecke, J. The Piwi‑ suppression of gene expression by RNAi in mammalian
e-mail: [email protected] piRNA pathway provides an adaptive defense in the cells. Proc. Natl Acad. Sci. USA 99, 1443–1448
doi:10.1038/nrm2479 transposon arms race. Science 318, 761–764 (2007). (2002).
30. Ketting, R. F., Haverkamp, T. H., van Luenen, H. G. & 56. Yang, S., Tutton, S., Pierce, E. & Yoon, K. Specific
1. Lee, Y. S. et al. Distinct roles for Drosophila Dicer‑1 and Plasterk, R. H. Mut‑7 of C. elegans, required for double‑stranded RNA interference in undifferentiated
Dicer‑2 in the siRNA/miRNA silencing pathways. Cell transposon silencing and RNA interference, is a mouse embryonic stem cells. Mol. Cell. Biol. 21,
117, 69–81 (2004). homolog of Werner syndrome helicase and RNaseD. 7807–7816 (2001).
2. Okamura, K., Ishizuka, A., Siomi, H. & Siomi, M. C. Cell 99, 133–141 (1999). 57. Calabrese, J. M., Seila, A. C., Yeo, G. W. & Sharp, P. A.
Distinct roles for Argonaute proteins in small RNA‑ 31. Sijen, T. & Plasterk, R. H. Transposon silencing in the RNA sequence analysis defines Dicer’s role in mouse
directed RNA cleavage pathways. Genes Dev. 18, Caenorhabditis elegans germ line by natural RNAi. embryonic stem cells. Proc. Natl Acad. Sci. USA 104,
1655–1666 (2004). Nature 426, 310–314 (2003). 18097–18102 (2007).
3. Tabara, H. et al. The rde-1 gene, RNA interference, and 32. Kuramochi‑Miyagawa, S. et al. Mili, a mammalian 58. Mi, S. et al. Sorting of small RNAs into Arabidopsis
transposon silencing in C. elegans. Cell 99, 123–132 member of piwi family gene, is essential for argonaute complexes is directed by the 5′ terminal
(1999). spermatogenesis. Development 131, 839–849 (2004). nucleotide. Cell 133, 116–127 (2008).
4. Tabara, H., Yigit, E., Siomi, H. & Mello, C. C. The dsRNA 33. Carmell, M. A. et al. MIWI2 is essential for 59. Montgomery, T. A. et al. Specificity of ARGONAUTE7–
binding protein RDE‑4 interacts with RDE‑1, DCR‑1, spermatogenesis and repression of transposons in the miR390 interaction and dual functionality in TAS3
and a DExH‑box helicase to direct RNAi in C. elegans. mouse male germline. Dev. Cell 12, 503–514 (2007). trans-acting siRNA formation. Cell 133, 128–141
Cell 109, 861–871 (2002). 34. Deng, W. & Lin, H. miwi, a murine homolog of piwi, (2008).
5. Ding, S. W. & Voinnet, O. Antiviral immunity directed by encodes a cytoplasmic protein essential for 60. Jannot, G., Boisvert, M. E., Banville, I. H. &
small RNAs. Cell 130, 413–426 (2007). spermatogenesis. Dev. Cell 2, 819–830 (2002). Simard, M. J. Two molecular features contribute to
6. Tam, O. H. et al. Pseudogene‑derived small interfering 35. Murchison, E. P. et al. Critical roles for Dicer in the the Argonaute specificity for the microRNA and
RNAs regulate gene expression in mouse oocytes. female germline. Genes Dev. 21, 682–693 (2007). RNAi pathways in C. elegans. RNA 14, 829–835
Nature 453, 534–538 (2008). 36. Svoboda, P. et al. RNAi and expression of (2008).
7. Watanabe, T. et al. Endogenous siRNAs from naturally retrotransposons MuERV‑L and IAP in preimplantation 61. Steiner, F. A. et al. Structural features of small RNA
formed dsRNAs regulate transcripts in mouse oocytes. mouse embryos. Dev. Biol. 269, 276–285 (2004). precursors determine Argonaute loading in
Nature 453, 539–543 (2008). 37. Watanabe, T. et al. Identification and characterization of Caenorhabditis elegans. Nature Struct. Mol. Biol.
8. Ghildiyal, M. et al. Endogenous siRNAs derived from two novel classes of small RNAs in the mouse germline: 14, 927–933 (2007).
transposons and mRNAs in Drosophila somatic cells. retrotransposon‑derived siRNAs in oocytes and 62. Tomari, Y., Du, T. & Zamore, P. D. Sorting of Drosophila
Science 320, 1077–1081 (2008). germline small RNAs in testes. Genes Dev. 20, small silencing RNAs. Cell 130, 299–308 (2007).
9. Okamura, K., Balla, S., Martin, R., Liu, N. & Lai, E. C. 1732–1743 (2006). 63. Azuma‑Mukai, A. et al. Characterization of endogenous
Two distinct mechanisms generate endogenous siRNAs 38. Yang, N. & Kazazian, H. H. Jr. L1 retrotransposition is human Argonautes and their miRNA partners in RNA
from bidirectional transcription in Drosophila. Nature suppressed by endogenously encoded small interfering silencing. Proc. Natl Acad. Sci. USA 105, 7964–7969
Struct. Mol. Biol. 15, 581–590 (2008). RNAs in human cultured cells. Nature Struct. Mol. Biol. (2008).
10. Okamura, K. et al. The Drosophila hairpin RNA 13, 763–771 (2006). 64. O’Carroll, D. et al. A slicer‑independent role for
pathway generates endogenous short interfering 39. Svoboda, P., Stein, P., Hayashi, H. & Schultz, R. M. Argonaute 2 in hematopoiesis and the microRNA
RNAs. Nature 453, 803–806 (2008). Selective reduction of dormant maternal mRNAs in pathway. Genes Dev. 21, 1999–2004 (2007).
11. Kawamura, Y. et al. Drosophila endogenous small mouse oocytes by RNA interference. Development 65. Yekta, S., Shih, I. H. & Bartel, D. P. MicroRNA‑directed
RNAs bind to Argonaute2 in somatic cells. Nature 127, 4147–4156 (2000). cleavage of HOXB8 mRNA. Science 304, 594–596
453, 793–797 (2008). 40. Stein, P., Zeng, F., Pan, H. & Schultz, R. M. Absence of (2004).
12. Czech, B. et al. An endogenous siRNA pathway in non‑specific effects of RNA interference triggered by 66. Diederichs, S. & Haber, D. A. Dual role for argonautes
Drosophila. Nature 453, 798–802 (2008). long double‑stranded RNA in mouse oocytes. Dev. Biol. in microRNA processing and posttranscriptional
13. Chung, W. J., Okamura, K., Martin, R. & Lai, E. C. 286, 464–471 (2005). regulation of microRNA expression. Cell 131,
Endogenous RNA interference provides a somatic 41. Reinhart, B. J., Weinstein, E. G., Rhoades, M. W., 1097–1108 (2007).
defense against Drosophila transposons. Current Bartel, B. & Bartel, D. P. MicroRNAs in plants. Genes 67. Spike, C. A., Bader, J., Reinke, V. & Strome, S. DEPS‑1
Biology 18, 795–802 (2008). Dev. 16, 1616–1626 (2002). promotes P‑granule assembly and RNA interference in
14. Farazi, T. A., Juranek, S. A. & Tuschl, T. The growing 42. Ruby, J. G. et al. Evolution, biogenesis, expression, and C. elegans germ cells. Development 135, 983–993
catalog of small RNAs and their association with target predictions of a substantially expanded set of (2008).
distinct Argonaute/Piwi family members. Development Drosophila microRNAs. Genome Res. 17, 1850–1864 68. Grishok, A. & Sharp, P. A. Negative regulation of
135, 1201–1214 (2008). (2007). nuclear divisions in Caenorhabditis elegans by
15. Mourelatos, Z. et al. miRNPs: a novel class of 43. Liu, Q. et al. R2D2, a bridge between the initiation and retinoblastoma and RNA interference‑related genes.
ribonucleoproteins containing numerous microRNAs. effector steps of the Drosophila RNAi pathway. Science Proc. Natl Acad. Sci. USA 102, 17360–17365 (2005).
Genes Dev. 16, 720–728 (2002). 301, 1921–5 (2003). 69. Peng, J. C. & Karpen, G. H. H3K9 methylation and RNA
16. Liu, J. et al. Argonaute2 is the catalytic engine of 44. Tomari, Y., Matranga, C., Haley, B., Martinez, N. & interference regulate nucleolar organization and
mammalian RNAi. Science 305, 1437–1441 (2004). Zamore, P. D. A protein sensor for siRNA asymmetry. repeated DNA stability. Nature Cell Biol. 9, 25–35
17. Meister, G. et al. Human Argonaute2 mediates RNA Science 306, 1377–1380 (2004). (2007).
cleavage targeted by miRNAs and siRNAs. Mol. Cell 45. Jiang, F. et al. Dicer‑1 and R3D1‑L catalyze microRNA 70. Deshpande, G., Calhoun, G. & Schedl, P. Drosophila
15, 185–197 (2004). maturation in Drosophila. Genes Dev. 19, 1674–1679 argonaute‑2 is required early in embryogenesis for the
18. Saito, K. et al. Specific association of Piwi with rasiRNAs (2005). assembly of centric/centromeric heterochromatin,
derived from retrotransposon and heterochromatic 46. Forstemann, K. et al. Normal microRNA maturation nuclear division, nuclear migration, and germ‑cell
regions in the Drosophila genome. Genes Dev. 20, and germ‑line stem cell maintenance requires formation. Genes Dev. 19, 1680–1685 (2005).
2214–2222 (2006). Loquacious, a double‑stranded RNA‑binding domain 71. Tang, F. et al. Maternal microRNAs are essential for
19. Gunawardane, L. S. et al. A slicer‑mediated mechanism protein. PLoS Biol. 3, e236 (2005). mouse zygotic development. Genes Dev. 21, 644–648
for repeat‑associated siRNA 5′ end formation in 47. Saito, K., Ishizuka, A., Siomi, H. & Siomi, M. C. (2007).
Drosophila. Science 315, 1587–1590 (2007). Processing of pre‑microRNAs by the Dicer‑1–
20. Lau, N. C. et al. Characterization of the piRNA complex Loquacious complex in Drosophila cells. PLoS Biol. Acknowledgements
from rat testes. Science 313, 363–367 (2006). 3, e235 (2005). K.O. was supported by the Charles Revson Foundation. E.C.L.
21. Brennecke, J. et al. Discrete small RNA‑generating loci 48. Kalidas, S. et al. Drosophila R2D2 mediates follicle was supported by the V Foundation for Cancer Research, the
as master regulators of transposon activity in formation in somatic tissues through interactions with Sidney Kimmel Foundation for Cancer Research, the Alfred
Drosophila. Cell 128, 1089–1103 (2007). Dicer‑1. Mech. Dev. 125, 475–85 (2008). Bressler Scholars Fund and the National Institutes of Health
22. Ruby, J. G. et al. Large‑scale sequencing reveals 49. Meyer, W. J. et al. Overlapping functions of argonaute (GM083300).
21U‑RNAs and additional microRNAs and endogenous proteins in patterning and morphogenesis of
siRNAs in C. elegans. Cell 127, 1193–207 (2006). Drosophila embryos. PLoS Genet. 2, e134 (2006).
23. Pak, J. & Fire, A. Distinct populations of primary and 50. Seitz, H., Ghildiyal, M. & Zamore, P. D. Argonaute DATABASES
secondary effectors during RNAi in C. elegans. Science loading improves the 5′ precision of both microRNAs entrez gene: http://www.ncbi.nlm.nih.gov/entrez/query.
315, 241–244 (2007). and their miRNA strands in flies. Curr. Biol. 18, fcgi?db=gene
24. Sijen, T., Steiner, F. A., Thijssen, K. L. & Plasterk, R. H. 147–151 (2008). Hdac1 | klar | mus308 | Rangap1 | tkv
Secondary siRNAs result from unprimed RNA synthesis 51. Forstemann, K., Horwich, M. D., Wee, L., Tomari, Y. & UniProtKB: http://www.uniprot.org
and form a distinct class. Science 315, 244–247 (2007). Zamore, P. D. Drosophila microRNAs are sorted into AGO1 | AGO2 | DCR1 | DCR2 | LOQS | PRG‑1 | R2D2
25. Yigit, E. et al. Analysis of the C. elegans Argonaute functionally distinct argonaute complexes after
family reveals that distinct Argonautes act sequentially production by dicer‑1. Cell 130, 287–297 (2007). FURTHER INFORMATION
during RNAi. Cell 127, 747–757 (2006). 52. Horwich, M. D. et al. The Drosophila RNA eric C. Lai’s homepage:
26. Wang, G. & Reinke, V. A C. elegans Piwi, PRG‑1, methyltransferase, DmHen1, modifies germline piRNAs http://www.mskcc.org/mskcc/html/52949.cfm
regulates 21U‑RNAs during spermatogenesis. Curr. and single‑stranded siRNAs in RISC. Curr. Biol. 17, all liNkS are active iN the oNliNe pdf
Biol. 18, 861–867 (2008). 1265–1272 (2007).

678 | SEPTEMBER 2008 | vOLUME 9 www.nature.com/reviews/molcellbio