PIWI-interacting RNAs (piRNAs) are small, 21–30 nt
single-stranded RNAs that associate with PIWI proteins
in various organisms. Typically, multiple PIWI
paralogues are encoded in a given organism, and the
family is named after the initially identified member,
Piwi, in Drosophila melanogaster1,2
. PIWI proteins
are a clade within the larger family of Argonaute
proteins that is mostly specifically expressed in the
germ line3–5
. In general, Argonaute proteins use their
bound small RNA molecules to identify relevant RNA
targets through base-pairing interactions. The precise
function of an Argonaute at a specified target can be
quite diverse, but usually the end result is silencing of
the target. The silencing by Argonaute proteins can
proceed through target RNA degradation (using the
RNaseH‑likedomainofArgonauteproteins4–9
),theinhibition
of translation or the recruitment of chromatinmodifying
activities10–14
. Generally, to exert a function
on targeted RNAs, Argonaute proteins recruit
additional cofactors3
.
In most animal species studied, PIWI proteins have
been shown to repress ‘non-self’ sequences, such as
transposable elements15
. These elements have the ability
to move and/or to multiply themselves to new positions
in the genome, thereby posing a threat to the genomic
stability of an organism16
. However, non-transposonrelated
PIWI targets and piRNAs derived from regular
mRNAs have been described as well, including early
developmental genes and genes involved in the establishment
of memory in neurons17–21
. These aspects
will not be discussed here further. We focus on recent
developments on the molecular mechanisms of the
PIWI pathway, including piRNA biogenesis and target
silencing.
The biogenesis of piRNAs is quite different from that
of other small-RNA pathways, such as the microRNA
(miRNA) pathway. In the miRNA pathway, the endonuclease
Dicer, which cleaves double-stranded RNA,
is essential, whereas for piRNAs, Dicer is dispensa-
ble22,23
. Consistently, instead of double-stranded RNA
molecules, single-stranded RNA molecules are used as
precursors for piRNAs24,25
. These transcripts are derived
from specific loci named piRNA loci or piRNA clusters.
The transcription factors involved are currently largely
unknown. The piRNA precursors are then believed
to be processed into mature piRNAs in the context of
an electron-dense cytoplasmic material — known as
nuage or inter-mitochondrial cement26,27
— that can be
found at the nuclear pores of germ cell nuclei. piRNA
biogenesis occurs through both primary and secondary
pathways3,12,15
, which are distinguished by the mechanism
by which the 5′ ends of piRNAs are generated.
The mechanism of piRNA precursor cleavage in the
primary pathway has long been — and to some extent
still is (see the discussion below) — an open question,
whereas in the secondary pathway, it is mediated by
the PIWI proteins themselves24,28,29
. Functionally, primary
biogenesis is thought to initiate PIWI pathways,
whereas secondary biogenesis (sometimes also referred
to as ping-pong amplification) is thought to ensure that
the total pool of piRNAs relates well to the expression
of the actual targets of primary piRNAs. This idea has
1
Hubrecht Institute-KNAW
and University Medical
Centre Utrecht,
Uppsalalaan 8, 3584 CT
Utrecht, The Netherlands.
2
Institute for Molecular
Biology, Ackermannweg 4,
D-55128 Mainz, Germany.
Correspondence to R.F.K.
e-mail:
r.ketting@imb-mainz.de
doi:10.1038/nrg3495
Published online 25 June 2013
PIWI-interacting RNAs: from
generation to transgenerational
epigenetics
Maartje J. Luteijn1
and René F. Ketting1,2
Abstract | Small-RNA-guided gene regulation is a recurring theme in biology. Animal
germ cells are characterized by an intriguing small-RNA-mediated gene-silencing
mechanism known as the PIWI pathway. For a long time, both the biogenesis of
PIWI-interacting RNAs (piRNAs) as well as their mode of gene silencing has remained
elusive. A recent body of work is shedding more light on both aspects and implicates PIWI
in the establishment of transgenerational epigenetic states. In fact, the epigenetic states
imposed by PIWI on targets may actually drive piRNA production itself. These findings
start to couple small RNA biogenesis with small-RNA-mediated epigenetics.
NON-CODING RNA
REVIEWS
NATURE REVIEWS | GENETICS	 VOLUME 14 | AUGUST 2013 | 523
© 2013 Macmillan Publishers Limited. All rights reserved
21U RNAs
Short RNA molecules found in
Caenorhabditis elegans that
are characterized by a length
of mostly 21 nt and
a uracil at their 5ʹ end. 21U
RNAs are bound by the Piwi
proteins PRG‑1 and PRG‑2.
been best documented in D. melanogaster, in which
mutation of a gene encoding a Piwi protein that acts
only in the secondary pathway leads to a collapse of
the whole piRNA pool30
, and maternally inherited
piRNAs have been shown to set‑up zygotic piRNA
populations31,32
. For further reading on the background
of these processes, see REFS 3,15,33,34.
At the level of target silencing by PIWI proteins, many
mysteries remain. Some PIWI proteins act in the cytoplasm
and may trigger degradation of mRNAs22,24,29,30,35
.
However, some PIWI proteins also translocate to the
nucleus on piRNA loading, suggesting they may have a
role in silencing their targets at the transcriptional level.
Such phenomena are well described at the molecular
level in non-piRNA pathways in, for example, fission
yeast and plants18,36–38
. However, until recently, it was
unclear to what extent these molecular details were
relevant for nuclear PIWI activity.
Recent publications have shed considerable light on
the mechanisms of piRNA biogenesis and PIWI target
silencing. In this Review, we focus on these recent
advances and follow piRNA biogenesis from the transcription
of precursor piRNA transcripts through to
their nuclear export and processing into mature piRNAs
and finally back into the nucleus, where PIWI proteins
exert their functions relating to transcriptional gene
silencing. We will see that piRNA-mediated silencing
at genomic loci can remain stable through organismal
generations and thus might represent a mechanism of
transgenerational epigenetic inheritance. Furthermore,
the resulting chromatin at silenced PIWI targets seems
closely related to the chromatin found at piRNAencoding
loci, raising the intriguing possibility that
the silencing activities of PIWI proteins may act both
to silence targets as well as to specify loci for precursor
piRNA transcription.
piRNA biogenesis
Transcription and transport of piRNA precursor RNAs.
An important question is how are precursor piRNA
transcripts selected to enter the piRNA-processing
pathways? One scenario could be that transcription
of precursor piRNAs has special features that lead to
the specific uptake of these transcripts into piRNA biogenesis
mechanisms. This scenario might operate in
Caenorhabditis elegans. In this animal, most piRNA species
(known as 21U RNAs)39,40
are individually encoded
by separate genes41
, and each is characterized by a
conserved motif that is found ~40 bp upstream of the
piRNA-encoding sequence42
(FIG. 1a). The motif is an
octamer sequence that is recognized by Forkhead family
(FKH) transcription factors, which are required for
the expression of 21U RNAs43
. Transcription of these
precursors starts two bases upstream of the mature
piRNA 5′ end. The transcripts are then 5′ capped43
and are mostly 26 nt in length44
. Additionally, so‑called
‘type 2’ 21U RNAs have been described in C. elegans;
these 21U RNA species relate to the 5′ ends of fulllength
mRNAs44
, suggesting that 21U RNA biogenesis
may also involve the generation of short 5′ transcripts
from full-length genes. The molecular details behind
the generation of both types of 21U precursors are currently
unclear, but possibly they relate to mechanisms
involved in premature transcriptional termination near
the 5′ ends of genes45
.
However, the generation of single piRNAs from precursor
transcripts seems not to operate in many other
species. In most studied organisms, piRNAs are derived
from precursor transcripts that are a few kilobases to
more than 100 kb in length and are processed into
many different piRNAs22,24,46–49
. The promoters and
transcription factors involved are starting to be identified:
for example, in D. melanogaster, promoter
mutations have been retrieved in a piRNA cluster that
drives very long single-stranded transcripts24
, and in
mice the transcription factor MYB-related protein A
(A‑MYB) has been shown to drive transcription of specific
piRNA clusters25
. How are such piRNA precursor
transcripts distinguished from other long transcripts
and sent into piRNA biogenesis routes? The answer to
this question is currently not known, but one hypothesis
is that a particular chromatin feature of piRNAproducing
loci may have a role. This thought is largely
based on the finding that in D. melanogaster, a particularly
fast evolving homologue of heterochromatin
protein 1 (HP1) called Rhino binds to many piRNA
clusters and is required for their piRNA production50
.
But how can such a heterochromatin-related protein be
required for the production of any RNA species, as it
seems to contradict the generally accepted role for HP1
proteins in transcriptional silencing51
? The answer may
come from a recent study showing that HP1‑bound loci
in fission yeast can still be transcribed52
. In that system,
HP1 not only functions by binding histone H3
trimethylated on lysine 9 (H3K9me3) but also acts as
an RNA-binding protein that accepts transcripts from
its bound loci and transports these to sites at which
they will be degraded. Thus, HP1 shuttles between
chromatin-bound and chromatin-free states. Such a
shuttling role for Rhino or other functionally related
proteins could play a part in piRNA pathways: instead
of targeting transcripts for degradation, Rhino might
escort precursor piRNA transcripts to cellular sites of
piRNA processing (FIG. 1b).
In fact, Rhino has been shown to colocalize with the
DEAD box protein UAP56 in nuclear speckles bound to
piRNA precursors, and it has been proposed that both
proteins have important roles in specifying piRNA
precursors and sending them to the right subcellular
compartments for piRNA processing23,24,35,46,48,49,53
. This
processing is believed to occur in specialized nuage
domains that flank nuclear pores on the cytoplasmic
side. Indeed, UAP56–Rhino foci on the nuclear side of
nuclear pores are often flanked on the cytoplasmic side
by the nuage marker Vasa, which is another DEAD box
protein26,53
. UAP56 is proposed to act on the nuclear side
to hold on to piRNA precursors until it is accepted by
Vasa on the other side of the nuclear membrane (FIG. 1b).
In support of this, mutations in UAP56 that are presumably
defective in RNA binding disrupt its colocalization
with Rhino, disturb subcellular Vasa localization and
inhibit piRNA production.
REVIEWS
524 | AUGUST 2013 | VOLUME 14	 www.nature.com/reviews/genetics
© 2013 Macmillan Publishers Limited. All rights reserved
Although it is tempting to speculate that the previously
described processes ensure that only piRNA
precursors, rather than normal mRNAs, are delivered
to sites of piRNA processing, clearly many predictions
coming from this model still have to be experimentally
tested. For example, why do some genic mRNAs
give rise to piRNAs19
? Would the tethering of Rhino or
UAP56 to ‘normal’ transcripts turn them into piRNA
precursors? And can Rhino homologues be identified
in other species? Whatever the precise roles of
Rhino and UAP56 are, another unresolved issue is
how Rhino initially finds its target piRNA-encoding
loci in the genome in order to serve as a molecular tag
for these loci. Although this is currently unknown,
there may be an intriguing link to the gene-silencing
activities of the PIWI pathway itself, as will be
discussed later.
5′ end formation. The next step in piRNA biogenesis
is 5′ end specification. How does this work? In the
secondary piRNA pathway, the 5′ end of piRNAs is
generated by PIWI-induced cleavage as a part of the
ping-pong mechanism (FIG. 2). However, in the primary
pathway, this activity has long been disputed.
A strong candidate to execute this step is a protein
named Zucchini (Zuc) because in D. melanogaster,
Zuc mutants display primary biogenesis defects54
,
and on the basis of primary sequence, Zuc is a protein
with potential nuclease activity. One of the reasons
why the role of Zuc has been difficult to pin down is
because besides its predicted nuclease domain, its primary
amino acid sequence shows features that predict
phospholipase activity.
One proposed function of Zuc in germ cells is to generate
phosphatidic acid on the mitochondrial surface,
Figure 1 | Transcription and selection of piRNA precursors.  a | In Caenorhabditis elegans, PIWI-interacting RNAs
(piRNAs) are individually encoded by small transcription units characterized by consensus motifs (red DNA region)
upstream of the piRNA-encoding DNA42
(green DNA region). Forkhead transcription factors (TFs) have been shown
to transcribe these loci43
, and the primary transcripts from these loci are mostly 26 nt in length44
. How these small
transcripts are terminated and how they are specified for 21U RNA production is unclear. Nevertheless, because the
26 nt precursors are two bases longer at their 5′ end than are the mature 21U RNA, we anticipate an endonucleolytic
event that removes these two bases. b | In Drosophila melanogaster, loci that give rise to piRNAs may have a particular
chromatin structure. In particular, a heterochromatin protein 1 (HP1)‑like protein named Rhino has been identified to
bind to many (but not all) piRNA-producing loci50
. Rhino further interacts with the RNA helicase UAP56 (REF. 53) and
with piRNA precursor transcripts. These transcripts then presumably channel through the nuclear pores and are
captured on the cytoplasmic side by the nuage component Vasa, which is another RNA helicase. Transcripts coming
from loci that lack Rhino-like proteins may not be able to follow the same route and hence will not be processed
into piRNAs.
UAP56
Nature Reviews | Genetics
Precursor piRNA
transcript
5′ cap
TF
21U gene
Upstream motif
Coding region
of 21U gene
5′ processing
3′ processing into
21 nt 21U RNA
21U RNA
Shuttling of Rhino
between chromatin
and RNA
RNA
Pol II
piRNA cluster
UAP56
Rhino
Rhino
a C. elegans piRNA precursor production b D. melanogaster piRNA precursor production
ZucZuc
?
5′ cleavage
Vasa binding
Cytoplasmic nuage
Nucleus
Cytoplasmic nuage
5′ cap
26 nt precursor RNA
PRG-1
PRG-1
Vasa
Rhino
Nucleus
Selection by UAP56
and Rhino
RNA
Pol II
REVIEWS
NATURE REVIEWS | GENETICS	 VOLUME 14 | AUGUST 2013 | 525
© 2013 Macmillan Publishers Limited. All rights reserved
which would induce the recruitment or activation
of Piwi-related nuage components. Consistent with
this, Zuc is present on the membranes of mitochondria,
and in mice that are deficient for ZUC (which is
also known as PLD6 in mice), mitochondrial defects
are observed24,28,55–57
. Furthermore, studies based on
in vivo relocation experiments using sensors that
detect phosphatidic acid suggest that mouse ZUC is
indeed a phospholipase3,15,33,34,55,56
, and one study showed
such activity in vitro58
. However, the crystal structures
of mouse ZUC and D. melanogaster Zuc have now
been determined, and their biochemical properties are
more robustly characterized. These results clearly show
that both proteins have the three-dimensional fold of
Figure 2 | Primary and secondary piRNA biogenesis.  Here we depict the general features of both primary and
secondary Piwi-interacting RNA (piRNA) biogenesis, using the Piwi pathways of Drosophila melanogaster as an example.
In the primary pathway, the precursors are cleaved by an endonuclease, most likely Zucchini (Zuc)59–61
. Next, cleaved
fragments are incorporated into Piwi or Aubergine (Aub) with the help of Shutdown (Shu) and Heat shock protein 83
(Hsp83, which is the fly homologue of HSP90)69
. Fragments with a 5′U may be heavily selected for at this step. After
loading into the Piwi protein, the unknown Trimmer enzyme trims the 3′ end63
to fit the Piwi protein, after which Hen1
(REFS 74,75) methylates the mature 3′ end. This completes primary biogenesis. Aub, but not Piwi, can then enter the
secondary pathway. Aub can recognize a cognate transcript and cleave it. The 3′ cleavage fragment of the targeted RNA
can then be taken up by another Piwi protein named Argonaute 3 (Ago3), again with the help of Shu69
. In addition, on the
basis of work in mice72
, Shu may assist in removing 5′ cleavage fragments following Aub cleavage. Further downstream,
steps are probably identical to primary biogenesis. In turn, Ago3 may assist in loading more Aub protein with secondary
piRNAs. Very similar processes take place during piRNA biogenesis in other species, such as mice and fish, although
many details appear to be quite flexible. For example, in mice, primary piRNAs have sense polarity.
Nature Reviews | Genetics
piRNA cluster Active transposon
Sense transcript
Nucleus
3′ truncation
?
CH3
A
U
3′ trimming?
Loading by
chaperone
complex?
Zuc Zuc Zuc
Aub
CH3U
Aub
Sense transcript5′3′
Antisense piRNA 5′
3′
3′
CH3
A
A
U
C
G
U
Aub
5′
5′
5′ formation
A 5′
Trimmer
3′
3′
Ago3
CH3 A 5′
Ago3
Hen1
Hen1
3′ methylation
5′ formation
CH3 A
U
5′3′
3′5′
Ago3
Antisense transcript
Sense piRNACH3 A
U
5′3′
3′5′
5′3′
3′5′
5′3′
Ago3
3′ methylation
3′ truncation
Trimmer
Hen1
Trimmer
CH3U
Aub or Piwi
U
Aub or Piwi
Shu
Hsp83
Shu
Hsp83
piRNA precursor transcript
Cleavage by Zuc?
Selection by Piwi or Aub?
5′
3′5′
Transposon transcripts
piRNA precursor transcripts
Primary piRNA pathway CytoplasmSecondary piRNA pathway
Aub
3′5′ U
5′
Shu
Hsp83
Aub or Piwi
Sense transcript
Antisense transcript
REVIEWS
526 | AUGUST 2013 | VOLUME 14	 www.nature.com/reviews/genetics
© 2013 Macmillan Publishers Limited. All rights reserved
a nuclease, and in vitro they are capable of cleaving
single-stranded DNA and RNA59–61
. Although these
data still leave space for additional roles of Zuc in the
piRNA pathway, a role as an enzyme that cleaves singlestranded
RNA seems evident. Whether this indeed is 5′
end generation remains to be definitively proven, but
in the absence of stronger candidates, Zuc is the prime
suspect to act as the 5′ nuclease during primary piRNA
biogenesis.
However, in mice that are deficient for ZUC, low
levels of processed piRNAs can still be detected56
. How
can this be reconciled with the idea that ZUC generates
the 5′ ends, which is a seemingly essential step? The
answer may simply be that ZUC merely boosts precursor
cleavage without being absolutely essential for it.
In the absence of ZUC, other nucleases may generate
sufficient cleavages in piRNA precursors to generate the
observed piRNAs. In this respect, it is noteworthy that
in vitro this enzyme does not specifically seem to generate
fragments with a 5′U base, which is the signature of
primary piRNAs. Rather, this signature may be imposed
later, through selective binding of PIWI proteins to
piRNA intermediates that contain a 5′U (FIG. 2), just as
Argonaute proteins in Arabidopsis thaliana have been
shown to select small RNA species with a strong preference
for specific 5′ bases62
. We note that C. elegans does
not appear to have a Zuc homologue, whereas 21U precursors
are still in need of some 5′ end processing activity.
This activity has so far remained totally unexplored.
Finally, a potential piRNA biogenesis intermediate carrying
a 5′ hydroxyl group has been described to accumulate
in D. melanogaster mutants that are defective in
piRNA loading. This species was also shown to carry a
2′–3′ cyclic phosphate57
. How this species relates to the
above and to the below is currently unclear.
PIWI loading. Following 5′ cleavage and preceding the
generation of the mature 3′ end, it is thought that
the piRNA intermediates are loaded into PIWI proteins
(FIG. 2). This is based on in vitro piRNA loading data63
as well as on the characteristics of piRNA 3′ ends (see
below). Several studies in both animals as well as plants
have indicated that the loading of RNA molecules into
Argonaute proteins is facilitated through the action of
chaperone proteins, most notably heat shock protein 90
(HSP90) homologues64–68
. A study in D. melanogaster
has now shown that a protein named Shutdown (Shu)
— a chaperone of the immunophilin class that contains
a tetratricopeptide repeat (TPR) and a prolyl-isomerase
domain — is required to load piRNA intermediates into
PIWI proteins as well69
. In order to do so, Shu requires its
TPR domain to mediate interaction with Hsp83 (the fly
orthologue of HSP90), which was previously shown to
affect piRNA biogenesis and transposon silencing70
. In
D. melanogaster, the primary piRNA biogenesis pathway
can be cleanly separated from the secondary simply by
analysing the germline follicle cells that use only the primary
pathway, hence it could be demonstrated that Shu
is essential for both piRNA biogenesis pathways69
. This
shows that following the distinct events during 5′ end
generation, primary and secondary piRNA biogenesis
pathways seem to merge into a common mechanism
during loading of piRNA intermediates.
Shu is a conserved protein, and the mouse homologue
FKBP6 has also been found to be associated with
PIWI proteins71,72
. Furthermore, FKBP6‑deficient mice
display similar fertility defects to those observed in
mice that are deficient for the PIWI proteins MILI and
MIWI2. However, on molecular dissection of defects
in FKBP6‑deficient mice, no defect in primary piRNA
biogenesis was observed72
. Rather, the authors of this
work suggest that FKBP6 is required only for the biogenesis
of MIWI2‑bound piRNAs, which arise solely
through secondary biogenesis. Still, like D. melanogaster
Shu, mouse FKBP6 was found to interact with HSP90
via the TPR domain, suggesting that at least some of its
roles in the PIWI pathway are conserved. On the basis
of experiments with Bombyx mori ovary-derived cells in
culture and pharmacological inhibition of Hsp90 with
geldanamycin, it was suggested that the Hsp90
machinery is required to remove one of the cleavage
fragments from Piwi–target RNA complexes following
target cleavage72
.
Thus, seemingly, two different roles for FKBP6–
HSP90 complexes in the PIWI pathway have been put
forward. One is a general role during piRNA loading,
and one entails a function specific to secondary biogenesis.
How can these be reconciled? One explanation is
that Shu and FKBP6 execute both functions but that the
role of FKBP6 in secondary biogenesis can be seen only
in specific circumstances. Such conditions could be that
the role of FKBP6 during piRNA loading in mice may be
redundant, whereas its role during secondary biogenesis
is not. Likewise, pharmacological inhibition of HSP90
may not have a major impact on primary biogenesis if
we assume that at the time of HSP90 inhibition, a sufficient
pool of piRNA-loaded PIWI protein is present,
such that only the role of FKBP6 during secondary biogenesis
can be revealed. Finally, how exactly FKBP6 or
Shu functions is not known. For example, mammalian
MOV10L1 (known as Armi in D. melanogaster) is an
RNA helicase that has also been implicated in piRNA
loading, but it is still unclear how FKBP6 and HSP90
interact with this protein.
3′ end formation. The last step in piRNA biogenesis
involves the formation of its 3′ end (FIG. 2). An important
clue to the formation of the piRNA 3′ end came
from analyses of piRNA length profiles. In most species,
the length distribution of piRNAs is relatively broad, and
there is greater variability at the 3′ end than at the 5′
end. So far, only C. elegans piRNAs deviate from this
picture, as for unknown reasons 21U RNAs display a
much more strict size distribution42
. In all other species
studied, the typical size profile of piRNA populations
shows bell-shaped distributions with a maximum length
that is characteristic of specific PIWI paralogues. For
example, the three Piwi proteins in D. melanogaster —
Piwi, Aubergine and Argonaute 3 (Ago3) — bind piRNA
populations with a length that peaks at 26, 25 and 24 nt,
respectively. It is assumed that these size profiles are a
signature of a 3′−5′ exonuclease that trims the 3′ ends of
REVIEWS
NATURE REVIEWS | GENETICS	 VOLUME 14 | AUGUST 2013 | 527
© 2013 Macmillan Publishers Limited. All rights reserved
Pachytene
The third stage of meiotic
prophase. Homologous
chromosomes are tightly held
together by the synaptonemal
complex, and homologous
recombination (‘crossing over’)
begins.
RNA-induced transcriptional
silencing complex
(RITS complex). An RNA
interference (RNAi) effector
complex required for
heterochromatin assembly in
fission yeast. It targets
centromeric transcripts to
induce both histone H3 at
lysine 9 (H3K9) methylation
and small interfering RNA
amplification.
22G RNAs
Short RNA molecules in
Caenorhabditis elegans made
by RNA-dependent RNA
polymerase (RdRP) activity on
an RNA template. 22G RNAs
are characterized by a
triphosphate group at their 5ʹ
end and a strong preference
for a G at their 5ʹ end. They are
bound by Argonaute proteins.
RNA-dependent RNA
polymerases
(RdRP). RNA polymerases that
use single-stranded RNA as a
template to synthesize
double-stranded RNA.
RNA-induced epigenetic
silencing
(RNAe). A form of epigenetic
silencing that can be induced
by various forms of RNA
interference (RNAi)-related
pathways, resulting in
trimethylation of histone H3 at
lysine 9 (H3K9) residues at the
silenced locus. The silencing is
maintained through both
mitosis and meiosis and can
last for tens of generations.
piRNA intermediates that are already loaded into a Piwi
protein, and all of the different Piwi paralogues leave
a different footprint on the maturing piRNA. Indeed,
3′−5′ exonucleolytic activity has been demonstrated to
act on longer RNA molecules loaded into PIWI proteins
in cell extracts in vitro63
. The enzyme responsible for this
activity has not yet been identified but has been named
Trimmer. Coupled with Trimmer activity is the methylation
activity of Hen1, which adds a methyl group to
the piRNA at the 2′OH position of the 3′ end base63,73–75
,
similarly to the plant enzyme HEN1 during miRNA bio-
genesis76
. This methylation protects piRNAs from the
addition of non-templated uridylation and destabiliza-
tion73–75,77
. Such uridylation seems to be a general feature
of small RNAs, including miRNAs and small interfering
RNAs (siRNAs), whenever (near) perfect base pairing
to target RNA molecules is involved78,79
. Hen1 finalizes
piRNA biogenesis and piRNA target silencing can
commence.
piRNA target silencing
Although piRNA-independent target recognition by
PIWI proteins has been proposed80
, we will focus on
piRNA-mediated target silencing. One still largely open
question regarding piRNA-mediated target recognition
is how strong the interactions between the piRNA and
the target RNA have to be to trigger a response. In the
case of piRNAs targeting transposons, many perfectly
base-pairing target RNAs are produced in the germ line,
and this has led to the idea that, in general, perfect base
pairing is the rule for piRNA–target RNA interactions.
However, in C. elegans, it has now been demonstrated
that mismatched base-pairing interactions between
piRNAs and target RNAs can be sufficient to generate a
response as well81,82
. This result begs for a closer examination
of piRNA targets in systems in which targets have
not yet been identified, such as for the piRNAs expressed
during pachytene stages in mammals46,48,49
. In this stage
of meiosis, the homologous chromosomes completely
pair and chromosomal crossover occurs. These pachytene
piRNAs are expressed in tremendous quantities
and seem to be required for proper germ cell maturation,
as male mice that lack their PIWI protein partner
MIWI are sterile owing to meiotic defects. Yet for most
pachytene piRNA species, no perfectly base-pairing
target RNA (other than potential antisense transcripts
from pachytene piRNA-encoding loci themselves) is
produced by the host genome. Applying the mismatch
rules found in C. elegans may partly solve this issue of
pachytene piRNA target identification, although we
note that it has been reported that MIWI-catalysed
cleavage requires near complete target base pairing83
and that D. melanogaster Piwi is also quite intolerant to
mismatches between piRNA and target RNA84
.
Regardless of the manner of base pairing, target
silencing should follow target recognition. Some PIWI
proteins silence their targets in the cytoplasm in a posttranscriptional
manner. As discussed before, PIWI
proteins can cleave their target RNAs24,29
, but effects of
PIWI proteins on translation have also been reported35
.
However, it has long been suspected that PIWI proteins
may also act transcriptionally because some PIWI
protein homologues go into the nucleus after they
have been loaded with a piRNA57,85–89
. Consistent with
nuclear PIWI protein localization, the mouse PIWI
proteins MILI and MIWI2 together lead to DNA methylation
at piRNA target loci28,86,90
. Unfortunately, more
detailed information on nuclear PIWI activity has
long been lacking, and hence the well-studied nuclear
RNA interference (RNAi) pathway of the fission yeast
Schizosaccharomyces pombe is often used as a template
for nuclear PIWI protein activity. In S. pombe, a nuclear
RNA-induced transcriptional silencing (RITS) complex is
involved in transcriptional silencing processes. RITS
contains an Argonaute protein (namely, Ago1, which
is loaded with siRNAs) that recognizes nascent pericentromeric
transcripts. This binding results in the
recruitment of another complex containing the histone
methyltransferase Clr4 that can methylate H3K9. This
in turn leads to the binding of Swi6, a homologue of
HP1, resulting in heterochromatin formation. In addition,
a short interfering RNA (siRNA)-amplifying
machinery is recruited to these targets, establishing
a potent silencing response. For reviews that specifically
focus on this topic see, for example, REFS 18,37,38.
Despite these ideas, causal links between PIWI protein
activity and transcriptional effects have long
remained unclear, but several recent studies in both
D. melanogaster and C. elegans now shed more light on
nuclear roles for piRNAs.
PIWI-mediated heterochromatin and germline immortality
in C. elegans. Recent data from several laboratories
have shown that the C. elegans PIWI pathway can affect
the chromatin structure of target loci91–93
. Interestingly,
PRG‑1 (which is the main C. elegans PIWI protein)
itself does not enter the nucleus. Rather, on PRG‑1–
target interaction in the cytoplasm, a class of small RNAs
(namely, 22G RNAs) is synthesized by RNA-dependent RNA
polymerases (RdRPs) using the PRG‑1‑targeted
RNA as a template39,40
. PRG‑1 endonucleolytic activity
does not seem to be required for this39,40
. These
22G RNAs are then loaded into a distinct set of wormspecific
Argonaute proteins, one of them being WAGO‑9
(also known as HRDE‑1). Presumably following loading
with a 22G RNA, WAGO‑9 goes into the nucleus92–94
. In
the nucleus, WAGO‑9, with help of previously identified
nuclear RNAi factors such as NRDE‑1, NRDE‑2 and
NRDE‑4 (REFS 94–96), triggers transcriptional silencing
(FIG. 3a). This is accompanied by trimethylation of H3K9
(H3K9me3) at the genomic loci that encode the target
RNAs91–93
. Related to these findings, the HP1 homologues
hpl‑1 and hpl‑2 have been shown to have a role
in the silencing, just like the histone methyltransferase
enzymes SET‑25 and SET‑32 (REFS 91–93).
Interestingly, the silenced state thus established can,
through some unknown mechanism, be converted into
a state that is stably inherited across generations, and at
this stage the silencing becomes independent of PRG‑1.
This phenomenon is called RNA-induced epigenetic
silencing (RNAe). Notably, RNAe does not occur on
PRG‑1 targets by default, as transgenes engineered to
REVIEWS
528 | AUGUST 2013 | VOLUME 14	 www.nature.com/reviews/genetics
© 2013 Macmillan Publishers Limited. All rights reserved
HPL-1
and
HPL-2
Initiation
22G RNAs
Maintenance
‘Non-self’ RNA‘Self’ RNA
b D. melanogaster target silencing in the nucleusa C. elegans target silencing in the nucleus
HP1
shuttling?
CH3
Piwi
Nucleus
Cytoplasm
Nucleus
Cytoplasm
Mael
NRDE-1
NRDE-2
Nature Reviews | Genetics
H3K9me3
H3K9me3
Transposon locus
?
?
?
?
?
Mael targeting
transposon RNA?
Transposon RNA
CH3
Piwi
RNA Pol II
HP1 HP1 HP1
HP1
H3K9me3
spreading
HPL-1
and
HPL-2
G
WAGO-9
G
WAGO-9
U
PRG-1
PRG-1CSR-1
NRDE-2
G
WAGO-9
Mael
NRDE-4
NRDE-4 HP1
shuttling?
?
G
G
G
UG
G
CH3
21U RNA
SET-25 or SET-32
RNA Pol II
22G RNA generation
PRG-1
RdRP
RdRP
RdRP
PRG-1-mediated
recruitment of RdRP?
HPL-1
and
HPL-2
Figure 3 | Effects of PIWI pathways on chromatin.  a | In Caenorhabditis elegans, the PIWI protein PRG‑1 triggers an
RNA-dependent RNA polymerase (RdRP) enzyme to produce 22G RNAs using PRG‑1‑targeted RNA molecules in the
cytoplasm as a template81,82
. The Argonaute protein WAGO‑9 accepts these 22G RNAs and shuttles into the nucleus,
where it triggers transcriptional gene silencing, resulting in trimethylation of histone H3 at lysine 9 (H3K9) residues by
histone methyltransferases91–94
. Two such methylases have been identified — SET‑25 and SET‑32 (REFS 92,93) — but their
specificities still have to be determined. Consistently, heterochromatin protein 1 (HP1)‑like proteins HPL‑1 and HPL‑2
have been shown to be required for the silencing effects. Analogously to the nuclear Argonaute protein NRDE‑3, it is
presumed that WAGO‑9 interacts with target loci via nascent RNA transcripts and leads to inhibition of RNA
polymerase II (RNA Pol II) elongation via NRDE‑4 (REFS 94,96). We show the shuttling of HPL‑1 and HPL‑2 between
chromatin and RNA-bound states, combined with RNA destabilization in analogy to the work from Keller et al.52
in
Schizosaccharomyces pombe. Through an unknown mechanism, the nuclear silencing can become independent of
PRG‑1. Hence, the steps including PRG‑1 are referred to as initiation steps, whereas further downstream, steps are
considered to be maintenance steps. We note that this maintenance thus involves the generation of 22G RNAs in the
absence of PRG‑1 and thus may involve a novel mechanism through which RdRP enzymes are recruited to (possibly
nascent) transcripts. The CSR‑1 pathway is depicted downstream of PRG‑1, reflecting the idea that CSR‑1 may be
responsible for the observed capacity of some loci to overrule PRG‑1‑mediated silencing, perhaps by cleaving PRG‑1
targets to prevent RdRP activity. However, direct experimental support for these steps is lacking. b | In Drosophila
melanogaster, Piwi has been shown to be responsible for H3K9 trimethylation of euchromatic transposon copies84,104–106
.
Although it has been suggested that this interaction may indeed proceed via nascent transcript84
, this still needs to be
further resolved. HP1 then binds these H3K9me3 marks. Following Piwi targeting, RNA Pol II occupancy at promoters is
reduced, suggesting that this pathway affects initiation of transcription. However, more molecular detail is required to
make strong statements. Maelstrom (Mael) is one of the factors that mediates the effects of Piwi-induced silencing
downstream of H3K9 trimethylation. Its exact roles are, however, unclear. We depict two possibilities that may both
occur: direct inhibition of some step during RNA Pol II transcription and interaction with transcripts coming from
silenced loci. Again, we depict hypothetical shuttling of HP1 proteins.
REVIEWS
NATURE REVIEWS | GENETICS	 VOLUME 14 | AUGUST 2013 | 529
© 2013 Macmillan Publishers Limited. All rights reserved
be targeted by PRG‑1 can maintain their dependence
on PRG‑1 for continued silencing in wild-type ani-
mals81,91
. The distinction and transition between the
PRG‑1‑dependent and PRG‑1‑independent silent states
are currently unclear.
Mutational analysis of this transgenerational silencing
effect has indicated that the transmission of the
silenced state through meiosis requires WAGO‑9,
the NRDE factors and other RNAi-related factors,
such as MUT‑7 and RDE‑3 (REFS 91–93). Interestingly,
WAGO‑9 mutants display an additional phenotype: their
germ cells lose their characteristic immortality, such that
after several generations, WAGO‑9 mutants become
sterile. This is accompanied by loss of H3K9me3 marks
at target loci94,95
. The molecular reasons behind this
phenotype are not clear, but it suggests a requirement in
C. elegans for PRG‑1 to act through WAGO‑9 to stably
maintain proper gene expression patterns in the germ
cells over generations. Although PRG‑1 may initiate
fairly stable forms of gene silencing, the stability is probably
less than 100%, leading to sporadic activation of
germline-incompatible genes. If they are not re‑silenced
through RNAe, such events may eventually mount up
to sterility. Whether similar effects are at play in other
systems is hard to judge, given the immediate sterility
of PIWI mutants in most other animals studied1,2,22,97
.
Although transgenerational silencing — for example,
maintenance of silencing through meiosis — requires
the various RNAi-related factors mentioned above, the
maintenance of silencing through mitosis within one
individual does not depend on these factors91–93
. How
can this difference be explained? Possibly, mitosis does
not require RNAi to maintain established RNAe-related
heterochromatin, just as other types of chromatin are
maintained independently of RNAi. By contrast, inheritance
through meiosis may require RNAi to store the
silencing information as, at least in mammals, meiosis
involves an epigenetic resetting process98
. The information
stored in the form of small RNAs would allow the
re‑establishment of heterochromatin after meiosis has
finished (FIG. 4). Such a system would closely mirror the
situation in mouse germ cells, in which it has been proposed
that the effect of Mili and Miwi2 mutations on the
remethylation of certain transposons reflects such a storage
of information on transposon silencing as well28,86,90
.
Finally, several observations suggest the existence
of a pathway that counteracts PRG‑1‑induced
silencing in C. elegans. These observations include the
capability of certain transgenes to re‑activate RNAesilenced
transgenes dominantly. It has been suggested
that another RNAi-related pathway mediated by the
Argonaute protein CSR‑1 may be responsible for this92,93
.
Figure 4 | Transgenerational inheritance of PIWI-induced heterochromatin.  a | Mitotic inheritance of a
PRG‑1‑mediated heterochromatic state (represented as the condensed structure flanked by other genomic DNA,
which is represented as a dashed line) in Caenorhabditis elegans. Presumably, the epigenetic state of the locus is not
reset during mitosis, and hence no RNA-induced epigenetic silencing (RNAe) pathway is required to maintain the
heterochromatic state. This mode probably operates during mitotic germline development within an individual.
b | A meiotic division during which the epigenetic state of a PRG‑1‑targeted locus may be reset (represented by the
change of the condensed structure to an open, linear structure of the locus; note that this has not been directly
demonstrated). To re‑establish the heterochromatic state after meiosis, RNAe, which involves WAGO‑9 bound with
22G RNAs that are homologous to the locus that needs to be re‑silenced, is required. The situation may be similar in
mice in which PIWI proteins have been shown to be required to re‑establish DNA methylation at PIWI target loci.
Nature Reviews | Genetics
a Mitosis
No RNAeRNAe
G
WAGO-9
G
WAGO-9
G
WAGO-9
G
WAGO-9
No RNAeRNAe
G
WAGO-9
G
WAGO-9
G
WAGO-9
b Meiosis
GG
GG
REVIEWS
530 | AUGUST 2013 | VOLUME 14	 www.nature.com/reviews/genetics
© 2013 Macmillan Publishers Limited. All rights reserved
This idea is based on the fact that CSR‑1 also acts in
the germ line and binds 22G RNAs derived from genes
expressed in the germ line99,100
and that loss of CSR‑1
does not activate its targets. As CSR‑1, like WAGO‑9,
binds 22G RNAs, there must be a factor that initiates
RdRP activity on CSR‑1 target RNAs. It is worthwhile
to consider that PRG‑1 may be this factor, just
as PRG‑1 drives WAGO‑9 22G RNA production. In
such a scenario, PRG‑1 may in some way set a balance
between silencing and activating pathways by providing
a proper assessment of the ‘selfness’ of a given
sequence. Obviously, these ideas need to be tested, but
they represent an interesting line of thought. Curiously,
a recent report indicates that the unicellular organism
Oxytrichia trifallax — which undergoes extensive deletion
of genomic sequences as part of its life cycle — uses
its Piwi pathway to protect particular sequences from
being deleted101
, and also D. melanogaster Piwi has at
least once been implicated in activation of target loci84,102
.
Piwi-mediated heterochromatin formation in D. melanogaster.
Although the Piwi proteins Aub and Ago3
are cytoplasmic, Piwi itself is nuclear. In fact, direct
interactions between Piwi and heterochromatin factors
such as HP1 have been described103
. However, whether
Piwi affects chromatin has long been unclear. A number
of recent studies shed more light on a nuclear role for
Piwi in transcriptional gene silencing84,104–106
. The general
picture that emerges is that loss of Piwi leads to a
strong reduction of histone H3 trimethylated at lysine 9
(H3K9me3) marks on Piwi targets in euchromatic
regions. This is accompanied by an increase in RNA
polymerase II (RNA Pol II) occupancy at the promoters
of Piwi targets and an increase in nascent RNA transcript
levels at target loci. Most probably, these effects do not
require Piwi cleavage of the target RNA107
. Despite this
inverse correlation between Piwi presence and chromatin
immunoprecipitation (ChIP) signals for RNA
Pol II occupancy, colocalization of Piwi and RNA Pol II
on nurse-cell-derived polytene chromosomes has been
demonstrated106
. Although the resolution of such studies
is low, it suggests that Piwi target loci in wild-type cells
still attract RNA Pol II. Interesting light has been shed on
the silencing mechanism through analysing Maelstrom
(Mael) mutants105
. Loss of Mael, which is another known
player in the Piwi pathway108
, results in transposon
activation, although piRNA levels and Piwi loading
are largely normal. Intriguingly, although transposons are
activated in Mael mutants, H3K9me3 levels at these
loci were almost unchanged, indicating that H3K9me3
marks are by themselves not sufficient to silence gene
expression and that Mael functions downstream of Piwi
and H3K9 trimethylation steps (FIG. 4b). Intriguingly,
in Mael mutants, RNA Pol II occupancy was increased
on Piwi target loci, and the H3K9me3 mark spread to
regions downstream of the original target. The authors
suggest a model in which the spreading of H3K9me3
marks in Mael mutants may in fact be a consequence of
the increase in RNA Pol II transcription, as an increase
of nascent transcripts at target loci may recruit more
Figure 5 | Conversion of a Piwi target locus into a piRNA-producing locus.  In Drosophila melanogaster, a
Piwi-interacting RNA (piRNA) target locus can be turned into a piRNA-producing locus109
. Although the inheritance of
this trait seems to be largely cytoplasmic, associated chromatin changes may still be at the basis of their conversion
into piRNA-producing loci. What is depicted here is a purely hypothetical process in which ‘regular’ heterochromatin
protein 1 (HP1) proteins that silence the Piwi target are replaced by Rhino-like HP1 molecules, channelling the
transcripts into piRNA processing rather than to destine them for destruction. Clearly, substantial experimentation is
needed to test this model in the various model systems.
Nature Reviews | Genetics
Transposon locus
To RNA
degradation
Transposon RNA
CH3
Piwi
RNA Pol II
HP1 HP1 HP1
HP1
H3K9me3
spreading
H3K9me3
To piRNA
processing
Rhino
Rhino
Rhino
HP1
piRNA locus
Transformation of a ‘regular’ piRNA target locus into a piRNA-producing locus.
Possible exchange of HP1-like molecules
UAP56
UAP56
Rhino
RNA Pol II
REVIEWS
NATURE REVIEWS | GENETICS	 VOLUME 14 | AUGUST 2013 | 531
© 2013 Macmillan Publishers Limited. All rights reserved
Paramutation
The interaction between two
alleles through which the
epigenetic state of one allele
can be imposed on the other
allele.
Piwi proteins to that locus. In agreement with this
notion, RNA Pol II transcription of individual transposon
insertion loci was indeed found to be required for
Piwi-mediated silencing and heterochromatinization105
.
The inheritance of the Piwi-mediated heterochromatin
chromatin has not been directly studied yet in D. mela‑
nogaster, but given the absence of secondary RNAi
pathways coupling to Piwi, as is the case in C. elegans, it
seems likely that D. melanogaster Piwi will be required at
each generation to maintain proper heterochromatin
at Piwi targets.
Finally, it was found that Piwi-induced silencing has
an impact on genes that are physically close to targeted
transposons105
. Apparently, the effects of the chromatin
marks established by the Piwi pathway are not restricted
to the primary targets but can extend to sequences that
are close‑by. Thus, this pathway may affect the transcription
of endogenous genes without directly targeting
them. It is perhaps because of this effect that the Piwi
pathway, at least in some organisms, may also act to sustain
the expression of essential genes that may otherwise
be potentially silenced (see above).
Concluding remarks
Going full circle: piRNAs from targeted loci? From the
advances described in the above sections, it has become
clear that both PIWI target silencing and piRNA precursor
specification depend on similar types of chromatin
that are characterized by H3K9me3 marks and
HP1‑like proteins. Intriguingly, there is now evidence
that the silencing activity of a Piwi pathway can turn
a target locus into a piRNA-generating locus. A recent
study has revealed that in D. melanogaster, a nonpiRNA-producing
locus (such as a protein-coding
gene) can be turned into a piRNA-producing one
simply by bringing it into contact with homologous
piRNAs109
. During and after this process, the locus
continues to be transcribed, but rather than leading to
protein expression, the transcripts are now processed
into piRNAs. Effectively, this silences the gene and
re‑routes the transcripts into the piRNA pathway. The
resulting piRNAs can in turn silence, in trans, other
loci of similar sequence109
. As the effect can be maintained
over generations as long as the involved Piwi
pathway is intact, the authors compare their observations
to paramutation, which is a phenomenon that
entails the change of protein-expressing loci into silent
loci that can in turn silence other homologous loci110,111
.
In this intriguing case of D. melanogaster paramuta-
tion109
, it remains to be tested to what extent the chromatin
structure of the transformed locus changes and
whether it starts to recruit proteins such as Rhino
and UAP56 that seem to be involved in trafficking precursor
piRNA transcripts (see above). Still, already at
this point in time, the remarkable similarities between
the chromatin state that is required at a piRNAproducing
locus and the chromatin state resulting from
the transcriptional silencing of a target locus by PIWI
proteins are striking. Everything considered, the findings
may fit a model in which piRNAs are amplified
not only post-transcriptionally — through ping-pong
amplification or RdRP activities — but perhaps also
transcriptionally by converting piRNA target loci into
piRNA-producing loci (FIG. 5). Clearly, these ideas are
in dire need of more experimental support, and most
probably the situation will be more complex than
sketched here. We may even find that in different species,
different strategies have been adopted to specify
piRNA transcripts and/or to silence Piwi protein targets
and that the generalizing statements made here are far
too simple. Screens such as those recently published for
D. melanogaster, which are aimed at the identification
of factors involved in the various steps of the Piwi pathway,
will be of big help in this respect112–114
. We are very
much looking forward to the studies that will contribute
to solving the many questions that remain in this field.
1.	Cox, D. N. et al. A novel class of evolutionarily
conserved genes defined by PIWI are essential for
stem cell self-renewal. Genes Dev. 12, 3715–3727
(1998).
2.	 Deng, W. & Lin, H. Miwi, a murine homolog of Piwi,
encodes a cytoplasmic protein essential for
spermatogenesis. Dev. Cell 2, 819–830 (2002).
3.	 Ghildiyal, M. & Zamore, P. D. Small silencing RNAs:
an expanding universe. Nature Rev. Genet. 10,
94–108 (2009).
4.	 Song, J. J. Crystal structure of argonaute and its
implications for RISC slicer activity. Science 305,
1434–1437 (2004).
5.	 Tolia, N. H. & Joshua-Tor, L. Slicer and the Argonautes.
Nature Chem. Biol. 3, 36–43 (2007).
6.	Wang, Y. et al. Nucleation, propagation and cleavage
of target RNAs in Ago silencing complexes. Nature
461, 754–761 (2009).
7.	 Schirle, N. T. & MacRae, I. J. The crystal structure of
human Argonaute2. Science 336, 1037–1040 (2012).
8.	Elkayam, E. et al. The structure of human argonaute‑2
in complex with miR‑20a. Cell 150, 100–110 (2012).
9.	Wang, Y. et al. Structure of an Argonaute silencing
complex with a seed-containing guide DNA and target
RNA duplex. Nature 456, 921–926 (2008).
10.	 van Wolfswinkel, J. C. & Ketting, R. F. The role of small
non-coding RNAs in genome stability and chromatin
organization. J. Cell Sci. 123, 1825–1839 (2010).
11.	 Czech, B. & Hannon, G. J. Small RNA sorting:
matchmaking for Argonautes. Nature Rev. Genet. 12,
19–31 (2010).
12.	 Siomi, H. & Siomi, M. C. On the road to reading the
RNA-interference code. Nature 457, 396–404
(2009).
13.	 Jinek, M. & Doudna, J. A. A three-dimensional view of
the molecular machinery of RNA interference. Nature
457, 405–412 (2009).
14.	 Rana, T. M. Illuminating the silence: understanding the
structure and function of small RNAs. Nature Rev.
Mol. Cell Biol. 8, 23–36 (2007).
15.	 Malone, C. D. & Hannon, G. J. Small RNAs as
guardians of the genome. Cell 136, 656–668 (2009).
16.	 Kazazian, H. H. Mobile elements: drivers of genome
evolution. Science 303, 1626–1632 (2004).
17.	Rouget, C. et al. Maternal mRNA deadenylation and
decay by the piRNA pathway in the early Drosophila
embryo. Nature 467, 1128–1132 (2010).
18.	 Castel, S. E. & Martienssen, R. A. RNA interference in
the nucleus: roles for small RNAs in transcription,
epigenetics and beyond. Nature Rev. Genet. 14,
100–112 (2013).
19.	Robine, N. et al. A broadly conserved pathway
generates 3′UTR-directed primary piRNAs. Curr. Biol.
19, 2066–2076 (2009).
20.	 Seto, A. G., Kingston, R. E. & Lau, N. C. The coming of
age for PIWI proteins. Mol. Cell 26, 603–609 (2007).
21.	Rajasethupathy, P. et al. A role for neuronal piRNAs in
the epigenetic control of memory-related synaptic
plasticity. Cell 149, 693–707 (2012).
22.	Houwing, S. et al. A role for Piwi and piRNAs in germ
cell maintenance and transposon silencing in
zebrafish. Cell 129, 69–82 (2007).
23.	 Vagin, V. V. A. Distinct small RNA pathway silences
selfish genetic elements in the germline. Science 313,
320–324 (2006).
24.	Brennecke, J. et al. Discrete small RNA-generating loci
as master regulators of transposon activity in
Drosophila. Cell 128, 1089–1103 (2007).
25.	Li, X. Z. et al. An ancient transcription factor initiates
the burst of piRNA production during early meiosis in
mouse testes. Mol. Cell 50, 67–81 (2013).
This paper describes the first transcription factor
that initiates the transcription of pachytene piRNA
precursors and genes involved in piRNA biogenesis,
thereby creating a feed-forward loop to boost
piRNA abundance.
26.	 Al‑Muktar, K. A. K. & Webb, A. C. An ultrastructural
study of primordial germ cells, oogonia and early
oocytes in Xenopus laevis. J. Embryol. Exp. Morphol.
26, 195–217 (1971).
27.	 Ketting, R. F. The many faces of RNAi. Dev. Cell 20,
148–161 (2011).
28.	 Aravin, A. A., Sachidanandam, R., Girard, A.,
Fejes-Toth, K. & Hannon, G. J. Developmentally
regulated piRNA clusters implicate MILI in transposon
control. Science 316, 744–747 (2007).
29.	Gunawardane, L. S. et al. A slicer-mediated
mechanism for repeat-associated siRNA 5′ end
formation in Drosophila. Science 315, 1587–1590
(2007).
30.	Li, C. et al. Collapse of germline piRNAs in the
absence of Argonaute3 reveals somatic piRNAs in
flies. Cell 137, 509–521 (2009).
REVIEWS
532 | AUGUST 2013 | VOLUME 14	 www.nature.com/reviews/genetics
© 2013 Macmillan Publishers Limited. All rights reserved
31.	Grentzinger, T. et al. piRNA-mediated
transgenerational inheritance of an acquired trait.
Genome Res. 22, 1877–1888 (2012).
32.	Brennecke, J. et al. An epigenetic role for maternally
inherited piRNAs in transposon silencing. Science
322, 1387–1392 (2008).
33.	 Klattenhoff, C. & Theurkauf, W. Biogenesis and
germline functions of piRNAs. Development 135,
3–9 (2007).
34.	 Siomi, M. C., Sato, K., Pezic, D. & Aravin, A. A.
PIWI-interacting small RNAs: the vanguard of
genome defence. Nature Rev. Mol. Cell Biol. 12,
246–258 (2011).
35.	 Grivna, S. T., Pyhtila, B. & Lin, H. MIWI associates
with translational machinery and PIWI-interacting
RNAs (piRNAs) in regulating spermatogenesis.
Proc. Natl Acad. Sci. USA 103, 13415–13420
(2006).
36.	 Molnar, A., Melnyk, C. & Baulcombe, D. C.
Silencing signals in plants: a long journey for small
RNAs. Genome Biol. 12, 215 (2011).
37.	 hler, M. B. U. & Gasser, S. M. Silent chromatin at the
middle and ends: lessons from yeasts. EMBO J. 28,
2149–2161 (2009).
38.	 Grewal, S. I. RNAi-dependent formation of
heterochromatin and its diverse functions. Curr. Opin.
Genet. Dev. 20, 134–141 (2010).
39.	Das, P. P. et al. Piwi and piRNAs act upstream of an
endogenous siRNA pathway to suppress Tc3
transposon mobility in the Caenorhabditis elegans
germline. Mol. Cell 31, 79–90 (2008).
40.	Batista, P. J. et al. PRG‑1 and 21U‑RNAs interact to
form the piRNA complex required for fertility in
C. elegans. Mol. Cell 31, 67–78 (2008).
41.	Billi, A. C. et al. A conserved upstream motif
orchestrates autonomous, germline-enriched
expression of Caenorhabditis elegans piRNAs.
PLoS Genet. 9, e1003392 (2013).
42.	Ruby, J. G. et al. Large-scale sequencing reveals
21U‑RNAs and additional microRNAs and
endogenous siRNAs in C. elegans. Cell 127,
1193–1207 (2006).
43.	 Cecere, G., Zheng, G. X. Y., Mansisidor, A. R.,
Klymko, K. E. & Grishok, A. Promoters recognized by
forkhead proteins exist for individual 21U‑RNAs.
Mol. Cell 47, 734–745 (2012).
44.	Gu, W. et al. CapSeq and CIP-TAP identify Pol II start
sites and reveal capped small RNAs as C. elegans
piRNA precursors. Cell 151, 1488–1500 (2012).
45.	Brannan, K. et al. mRNA decapping factors and the
exonuclease Xrn2 function in widespread premature
termination of RNA polymerase II transcription.
Mol. Cell 46, 311–324 (2012).
46.	Aravin, A. et al. A novel class of small RNAs bind to
MILI protein in mouse testes. Nature 442, 203–207
(2006).
47.	 Aravin, A. A., Hannon, G. J. & Brennecke, J.
The Piwi-piRNA pathway provides an adaptive
defense in the transposon arms race. Science 318,
761–764 (2007).
48.	Lau, N. C. et al. Characterization of the piRNA
complex from rat testes. Science 313, 363–367
(2006).
49.	 Girard, A., Sachidanandam, R., Hannon, G. J. &
Carmell, M. A. A germline-specific class of small
RNAs binds mammalian Piwi proteins. Nature 442,
199–202 (2006).
50.	Klattenhoff, C. et al. The Drosophila HP1 homolog
rhino is required for transposon silencing and piRNA
production by dual-strand clusters. Cell 138,
1137–1149 (2009).
51.	 Grewal, S. I. S. & Jia, S. Heterochromatin revisited.
Nature Rev. Genet. 8, 35–46 (2007).
52.	Keller, C. et al. HP1(Swi6) mediates the recognition
and destruction of heterochromatic RNA transcripts.
Mol. Cell 47, 215–227 (2012).
This work shows the RNA-binding activity of SWI6
is required for silencing but leaves heterochromatin
formation intact.
53.	Zhang, F. et al. UAP56 couples piRNA clusters to the
perinuclear transposon silencing machinery. Cell 151,
871–884 (2012).
This paper describes the roles of nuclear UAP56
and perinuclear Vasa proteins in forming a
piRNA-processing compartment that spans the
nuclear envelope and is essential for piRNA
production and transposon silencing.
54.	 Pane, A. Wehr, K. & Schüpbach, T. Zucchini and squash
encode two putative nucleases required for rasiRNA
production in the Drosophila germline. Dev. Cell 12,
851–862 (2007).
55.	Huang, H. et al. piRNA-associated germline nuage
formation and spermatogenesis require MitoPLD
profusogenic mitochondrial-surface lipid signaling.
Dev. Cell 20, 376–387 (2011).
56.	Watanabe, T. et al. MITOPLD is a mitochondrial
protein essential for nuage formation and piRNA
biogenesis in the mouse germline. Dev. Cell 20,
364–375 (2011).
57.	Saito, K. et al. Roles for the Yb body components
Armitage and Yb in primary piRNA biogenesis in
Drosophila. Genes Dev. 24, 2493–2498 (2010).
58.	Choi, S.‑Y. et al. A common lipid links Mfn-mediated
mitochondrial fusion and SNARE-regulated exocytosis.
Nature Cell Biol. 8, 1255–1262 (2006).
59.	Voigt, F. et al. Crystal structure of the primary piRNA
biogenesis factor Zucchini reveals similarity to
the bacterial PLD endonuclease Nuc. RNA 18,
2128–2134 (2012).
60.	Nishimasu, H. et al. Structure and function of Zucchini
endoribonuclease in piRNA biogenesis. Nature 491,
284–287 (2012).
61.	 Ipsaro, J. J., Haase, A. D., Knott, S. R., Joshua-Tor, L.
& Hannon, G. J. The structural biochemistry of
Zucchini implicates it as a nuclease in piRNA
biogenesis. Nature 491, 279–283 (2012).
References 59, 60 and 61 show crystal structures
of Zuc, showing that it has a nuclease domain and
possesses single-stranded nuclease activity. These
studies provide proof that Zuc is a nuclease that
probably acts to specify piRNA 5′ ends.
62.	Mi, S. et al. Sorting of small RNAs into Arabidopsis
Argonaute complexes is directed by the 5′ terminal
nucleotide. Cell 133, 116–127 (2008).
63.	 Kawaoka, S., Izumi, N., Katsuma, S. & Tomari, Y. 3′
end formation of PIWI-interacting RNAs in vitro.
Mol. Cell 43, 1015–1022 (2011).
64.	Iwasaki, S. et al. Hsc70/Hsp90 chaperone machinery
mediates ATP-dependent RISC loading of small RNA
duplexes. Mol. Cell 39, 292–299 (2010).
65.	 Johnston, M., Geoffroy, M.‑C., Sobala, A., Hay, R. &
Hutvagner, G. HSP90 protein stabilizes unloaded
Argonaute complexes and microscopic P‑bodies
in human cells. Mol. Biol. Cell 21, 1462–1469
(2010).
66.	 Miyoshi, T., Takeuchi, A., Siomi, H. & Siomi, M. C.
A direct role for Hsp90 in pre-RISC formation in
Drosophila. Nature Struct. Mol. Biol. 17, 1024–1026
(2010).
67.	 Iki, T., Yoshikawa, M., Meshi, T. & Ishikawa, M.
Cyclophilin 40 facilitates HSP90‑mediated
RISC assembly in plants. EMBO J. 31, 267–278
(2012).
68.	Gangaraju, V. K. et al. Drosophila Piwi functions in
Hsp90‑mediated suppression of phenotypic variation.
Nature Genet. 43, 153–158 (2010).
69.	 Olivieri, D., Senti, K.‑A., Subramanian, S.,
Sachidanandam, R. & Brennecke, J. The cochaperone
shutdown defines a group of biogenesis factors
essential for all piRNA populations in Drosophila.
Mol. Cell 47, 954–969 (2012).
70.	Specchia, V. et al. Hsp90 prevents phenotypic
variation by suppressing the mutagenic activity of
transposons. Nature 463, 662–665 (2010).
71.	Vagin, V. V. et al. Proteomic analysis of murine Piwi
proteins reveals a role for arginine methylation in
specifying interaction with Tudor family members.
Genes Dev. 23, 1749–1762 (2009).
72.	Xiol, J. et al. A role for Fkbp6 and the chaperone
machinery in piRNA amplification and transposon
silencing. Mol. Cell 47, 970–979 (2012).
73.	Kamminga, L. M. et al. Hen1 is required for oocyte
development and piRNA stability in zebrafish.
EMBO J. 29, 3688–3700 (2010).
74.	Saito, K. et al. Pimet, the Drosophila homolog of
HEN1, mediates 2′-O‑methylation of Piwi-interacting
RNAs at their 3′ ends. Genes Dev. 21, 1603–1608
(2007).
75.	Horwich, M. D. et al. The Drosophila RNA
methyltransferase, DmHen1, modifies germline
piRNAs and single-stranded siRNAs in RISC.
Curr. Biol. 17, 1265–1272 (2007).
76.	 Yu, B. Methylation as a crucial step in plant
microRNA biogenesis. Science 307, 932–935
(2005).
77.	 Kurth, H. M. & Mochizuki, K. 2′‑O‑methylation
stabilizes Piwi-associated small RNAs and ensures
DNA elimination in Tetrahymena. RNA 15,
675–685 (2009).
78.	Ameres, S. L. et al. Target RNA-directed trimming
and tailing of small silencing RNAs. Science 328,
1534–1539 (2010).
79.	 Ameres, S. L., Hung, J.‑H., Xu, J., Weng, Z. &
Zamore, P. D. Target RNA-directed tailing and
trimming purifies the sorting of endo-siRNAs between
the two Drosophila Argonaute proteins. RNA 17,
54–63 (2011).
80.	Vourekas, A. et al. Mili and Miwi target RNA
repertoire reveals piRNA biogenesis and function of
Miwi in spermiogenesis. Nature Struct. Mol. Biol. 19,
773–781 (2012).
81.	Bagijn, M. P. et al. Function, targets, and evolution
of Caenorhabditis elegans piRNAs. Science 337,
574–578 (2012).
82.	Lee, H.‑C. et al. C. elegans piRNAs mediate the
genome-wide surveillance of germline Transcripts. Cell
150, 78–87 (2012).
References 81 and 82 show that the
C. elegans-specific PIWI protein PRG‑1 binds its
targets by imperfectly base pairing and silences
independently of its endonuclease activity. Instead,
it recruits a downstream small RNA pathway that
generates endogenous siRNAs (endo-siRNAs) to
induce silencing.
83.	Reuter, M. et al. Miwi catalysis is required for piRNA
amplification-independent LINE1 transposon
silencing. Nature 480, 264–267 (2011).
84.	Huang, X. A. et al. A major epigenetic programming
mechanism guided by piRNAs. Dev. Cell 24, 502–516
(2013).
85.	 Cox, D. N., Chao, A. & Lin, H. piwi encodes a
nucleoplasmic factor whose activity modulates the
number and division rate of germline stem cells.
Development 127, 503–514 (2000).
86.	Aravin, A. A. et al. A piRNA pathway primed by
individual transposons is linked to de novo DNA
methylation in mice. Mol. Cell 31, 785–799 (2008).
87.	 Houwing, S., Berezikov, E. & Ketting, R. F. Zili is
required for germ cell differentiation and meiosis in
zebrafish. EMBO J. 27, 2702–2711 (2008).
88.	Qi, H. et al. The Yb body a major site for Piwiassociated
RNA biogenesis and a gateway for
Piwi expression and transport to the nucleus in
somatic cells. J. Biol. Chem. 286, 3789–3797
(2011).
89.	 Olivieri, D., Sykora, M. M., Sachidanandam, R.,
Mechtler, K. & Brennecke, J. An in vivo RNAi assay
identifies major genetic and cellular requirements for
primary piRNA biogenesis in Drosophila. EMBO J. 29,
3301–3317 (2010).
90.	Kuramochi-Miyagawa, S. et al. MVH in piRNA
processing and gene silencing of retrotransposons.
Genes Dev. 24, 887–892 (2010).
91.	Luteijn, M. J. et al. Extremely stable Piwi-induced
gene silencing in Caenorhabditis elegans. EMBO J.
31, 3422–3430 (2012).
92.	Ashe, A. et al. piRNAs can trigger a multigenerational
epigenetic memory in the germline of C. elegans. Cell
150, 88–99 (2012).
93.	Shirayama, M. et al. piRNAs initiate an epigenetic
memory of nonself RNA in the C. elegans germline.
Cell 150, 65–77 (2012).
94.	Buckley, B. A. et al. A nuclear Argonaute promotes
multigenerational epigenetic inheritance and germline
immortality. Nature 489, 447–451 (2012).
References 91, 92, 93 and 94 show a role for
PRG‑1 in inducing stably inherited silenced state by
recruiting a nuclear RNAi machinery that triggers
heterochromatin formation. Reference 94
demonstrates the nuclear activity of WAGO‑9 in
RNAi-induced transgenerational silencing and
reveals a role in germline mortality.
95.	 Burton, N. O., Burkhart, K. B. & Kennedy, S.
Nuclear RNAi maintains heritable gene silencing in
Caenorhabditis elegans. Proc. Natl Acad. Sci. 108,
19683–19688 (2011).
96.	Burkhart, K. B. et al. A pre-mRNA-associating factor
links endogenous siRNAs to chromatin regulation.
PLoS Genet. 7, e1002249 (2011).
97.	Kuramochi-Miyagawa, S. et al. Mili, a mammalian
member of Piwi family gene, is essential for
spermatogenesis. Development 131, 839–849
(2004).
98.	 Morgan, H. D., Santos, F., Green, K., Dean, W. &
Reik, W. Epigenetic reprogramming in mammals.
Hum. Mol. Genet. 14 (Suppl. 1), R47–R58 (2005).
99.	Claycomb, J. M. et al. The Argonaute CSR‑1 and
its 22G‑RNA cofactors are required for holocentric
chromosome segregation. Cell 139, 123–134
(2009).
100.	Wolfswinkel, J. C. V. et al. CDE‑1 affects chromosome
segregation through uridylation of CSR‑1‑bound
siRNAs. Cell 139, 135–148 (2009).
REVIEWS
NATURE REVIEWS | GENETICS	 VOLUME 14 | AUGUST 2013 | 533
© 2013 Macmillan Publishers Limited. All rights reserved
101.	Fang, W., Wang, X., Bracht, J. R., Nowacki, M. &
Landweber, L. F. Piwi-interacting RNAs protect DNA
against loss during oxytricha genome rearrangement.
Cell 151, 1243–1255 (2012).
102.	Yin, H. & Lin, H. An epigenetic activation role of Piwi
and a Piwi-associated piRNA in Drosophila
melanogaster. Nature 450, 304–308 (2007).
103.	Brower-Toland, B. et al. Drosophila PIWI associates
with chromatin and interacts directly with HP1a.
Genes Dev. 21, 2300–2311 (2007).
104.	Rozhkov, N. V., Hammell, M. & Hannon, G. J.
Multiple roles for Piwi in silencing Drosophila
transposons. Genes Dev. 27, 400–412 (2013).
105.	Sienski, G., Donertas, D. & Brennecke, J.
Transcriptional silencing of transposons by Piwi and
maelstrom and its impact on chromatin state and
gene expression. Cell 151, 964–980 (2012).
106.	Le Thomas, A. et al. Piwi induces piRNA-guided
transcriptional silencing and establishment of a
repressive chromatin state. Genes Dev. 27,
390–399 (2013).
References 104, 105 and 106, along with reference
82, show that D. melanogaster Piwi is required to
establish heterochromatic H3K9me3 marks on
transposons and their genomic surroundings.
Reference 105 also shows that the high-mobility
group (HMG) protein Mael is essential for
transposon silencing but not for heterochromatin
formation.
107.	Darricarrère, N., Liu, N., Watanabe, T. & Lin, H.
Function of Piwi, a nuclear Piwi/Argonaute protein, is
independent of its slicer activity. Proc. Natl Acad. Sci.
110, 1297–1302 (2013).
108.	Aravin, A. A. et al. Cytoplasmic compartmentalization
of the fetal piRNA pathway in mice. PLoS Genet. 5,
e1000764 (2009).
109.	de Vanssay, A. et al. Paramutation in Drosophila linked
to emergence of a piRNA-producing locus. Nature
490, 112–115 (2012).
This work shows that piRNA targets can be
converted into piRNA-producing loci, resulting in
genetic behavior that closely mirrors that of
paramutation.
110.	Erhard, K. F. Jr & Hollick, J. B. Paramutation:
a process for acquiring trans-generational
regulatory states. Curr. Opin. Plant Biol. 14,
210–216 (2011).
111.	 Pilu, R. Paramutation: just a curiosity or fine tuning of
gene expression in the next generation? Curr. Genom.
12, 298–306 (2011).
112.	 Muerdter, F. et al. A genome-wide RNAi screen draws a
genetic framework for transposon control and primary
piRNA biogenesis in Drosophila. Mol. Cell. http://dx.
doi.org/10.1016/j.molcel.2013.04.006 (2013).
113.	Czech, B., Preall, J. B., McGinn, J. & Hannon, G. J.
A transcriptome-wide RNAi screen in the Drosophila
ovary reveals factors of the germline piRNA pathway.
Mol. Cell http://dx.doi.org/10.1016/j.
molcel.2013.04.007 (2013).
114.	Handler, D. et al. The genetic makeup of the
Drosophila piRNA pathway. Mol. Cell http://dx.doi.
org/10.1016/j.molcel.2013.04.031 (2013).
Acknowledgements
The authors thank members of their laboratory for helpful
discussions. This work is supported by grants from the
European Research Council (ERC; StG 202819),
the Netherlands Organization for Scientific Research (NWO;
ECHO 700.57.006 and Vici 724.011.001) and the
Boehringer Ingelheim Stiftung. The authors apologize to
those colleagues whose work could not be cited owing
to space restrictions.
Competing interests statement
The authors declare no competing financial interests.
REVIEWS
534 | AUGUST 2013 | VOLUME 14	 www.nature.com/reviews/genetics
© 2013 Macmillan Publishers Limited. All rights reserved