Sorting of Drosophila small silencing RNAs partitions
microRNA* strands into the RNA interference pathway
MEGHA GHILDIYAL,1,4JIA XU,2,4HERVE´SEITZ,1,5,6ZHIPING WENG,3and PHILLIP D. ZAMORE1
1Howard Hughes Medical Institute, Department of Biochemistry and Molecular Pharmacology, University of Massachusetts Medical School,
Worcester, Massachusetts 01605, USA
2Department of Biomedical Engineering, Boston University, Boston, Massachusetts 02215, USA
3Program in Bioinformatics and Integrative Biology, University of Massachusetts Medical School, Worcester, Massachusetts 01605, USA
In flies, small silencing RNAs are sorted between Argonaute1 (Ago1), the central protein component of the microRNA (miRNA)
pathway, and Argonaute2 (Ago2), which mediates RNA interference. Extensive double-stranded character—as is found in small
interfering RNAs (siRNAs)—directs duplexes into Ago2, whereas central mismatches, like those found in miRNA/miRNA*
duplexes, direct duplexes into Ago1. Central to this sorting decision is the affinity of the small RNA duplex for the Dcr-2/R2D2
heterodimer, which loads small RNAs into Ago2. Here, we show that while most Drosophila miRNAs are bound to Ago1,
miRNA* strands accumulate bound to Ago2. Like siRNA loading, efficient loading of miRNA* strands in Ago2 favors duplexes
with a paired central region and requires both Dcr-2 and R2D2. Those miRNA and miRNA* sequences bound to Ago2, like
siRNAs diced in vivo from long double-stranded RNA, typically begin with cytidine, whereas Ago1-bound miRNA and miRNA*
disproportionately begin with uridine. Consequently, some pre-miRNA generate two or more isoforms from the same side of the
stem that differentially partition between Ago1 and Ago2. Our findings provide the first genome-wide test for the idea that
Drosophila small RNAs are sorted between Ago1 and Ago2 according to their duplex structure and the identity of their first
Keywords: microRNA; miRNA; siRNA; Argonaute; RNA interference; RNAi; small RNA sorting
In animals, microRNAs (miRNAs) regulate the stability
and rate of translation of mRNAs, whereas small interfering
RNAs (siRNAs) silence transposons, defend against viral
pathogens, and regulate mRNA expression (Ghildiyal and
Zamore 2009). Both miRNAs and siRNAs derive from
longer double-stranded RNA (dsRNA) precursors, which
are cleaved by RNase III dsRNA-specific endonucleases.
miRNA production begins in the nucleus, where long
primary miRNAs, transcribed by RNA polymerase II, are
converted into z65 nucleotides (nt) pre-miRNA hairpins
by the RNase III ribonuclease, Drosha, aided by a double-
stranded RNA-binding domain (dsRBD) partner protein
(Lee et al. 2003; Denli et al. 2004; Gregory et al. 2004; Han
et al. 2004; Zeng et al. 2005; Han et al. 2006). A minority of
pre-miRNAs—mirtrons—correspond to entire introns and
are excised from their primary transcripts by the pre-
mRNA splicing pathway (Okamura et al. 2007; Ruby et al.
2007a). Pre-miRNAs are then exported to the cytoplasm
(Yi et al. 2003; Bohnsack et al. 2004; Lund et al. 2004),
where they are processed by a second RNase III enzyme,
Dicer, together with its dsRBD partner protein, into z22-
nt-long miRNA/miRNA* duplexes (Grishok et al. 2001;
Hutva ´gner et al. 2001; Ketting et al. 2001; Chendrimada
et al. 2005; Fo ¨rstemann et al. 2005; Jiang et al. 2005;
Maniataki and Mourelatos 2005; Saito et al. 2005; Lee et al.
2006; Melo et al. 2009).
siRNA production also requires Dicer, which excises 21-nt
siRNA duplexes, comprising a guide and passenger strand,
from long dsRNA formed by the base pairing of comple-
mentary sense and antisense transcripts, convergently tran-
scribed mRNAs, or by the intramolecular base pairing of
long, self-complementary RNAs. Such endogenous dsRNAs
4These authors contributed equally to this work.
(LBME), Universite ´ de Toulouse, UPS, F-31000 Toulouse, France; and
6CNRS, Laboratoire de Biologie Mole ´culaire Eucaryote (LBME), F-31000
Reprint requests to: Phillip D. Zamore, Howard Hughes Medical Institute,
University of Massachusetts Medical School, LRB 822, 364 Plantation
Street, Worcester, MA 01605, USA; e-mail: phillip.zamore@umassmed.
edu; fax: (508) 856-2003.
Article published online ahead of print. Article and publication date are at
5Laboratoire de Biologie Mole ´culaire Eucaryote
RNA (2010), 16:43–56. Published by Cold Spring Harbor Laboratory Press. Copyright ? 2010 RNA Society.
yield endo-siRNAs. Similarly, exogenous dsRNA, introduced
experimentally or by viral infection, are converted by Dicer
In Drosophila melanogaster, Dicer-1 (Dcr-1) converts
pre-miRNAs into miRNA/miRNA* duplexes; Dicer-2
(Dcr-2) converts long dsRNA into 21-nt siRNA duplexes
(Zamore et al. 2000; Bernstein et al. 2001; Lee et al. 2004).
The use of different Dicer proteins to generate miRNAs and
siRNAs may minimize competition between the two
pathways, so that an RNAi defense to viral infection does
not perturb miRNA production.
All small silencing RNAs function bound to Argonaute
proteins. Argonaute proteins display nucleotides 2–8 of the
small RNA guide in a prehelical geometry that confers on
this region special importance in target recognition: the
majority of the binding energy for target binding is con-
tributed by this ‘‘seed’’ sequence (Jackson et al. 2003; Lewis
et al. 2003; Haley and Zamore 2004; Rajewsky and Socci
2004; Grun et al. 2005; Krek et al. 2005; Lewis et al. 2005;
Lim et al. 2005; Jackson et al. 2006; Ameres et al. 2007;
Grimson et al. 2007; Wang et al. 2008; Parker et al. 2009).
In flies, miRNAs are loaded from miRNA/miRNA* du-
plexes into Argonaute1 (Ago1), whereas siRNAs are loaded
from guide/passenger duplexes into Argonaute2 (Ago2)
(Hammond et al. 2000; Okamura et al. 2004; Rand et al.
2004; Matranga et al. 2005; Miyoshi et al. 2005; Rand et al.
2005; Kim et al. 2006; Leuschner et al. 2006). Two binary
choices accompany loading of small RNAs into Argonaute
proteins in Drosophila: the choice of Ago1 versus Ago2 and
the selection of one of the two strands of the duplex as
a miRNA or guide strand (Khvorova et al. 2003; Schwarz
et al. 2003).
Although fly miRNAs are overwhelmingly associated with
Ago1 and siRNAs with Ago2, small RNA production and
Argonaute loading are uncoupled (Fo ¨rstemann et al. 2007;
Tomari et al. 2007). Instead, miRNA and siRNA duplexes
are actively partitioned between Ago1 and Ago2 according
to their structure. Extensive double-stranded character di-
rects duplexes, such as siRNAs, into Ago2, which mediates
RNAi, whereas bulges and mismatches, like those found in
miRNA/miRNA* duplexes, are sorted into Ago1 (Kawamata
et al. 2009). Central to this sorting decision is the affinity
of the small RNA duplex for the Dcr-2/R2D2 heterodimer,
which loads small RNAs into Ago2 (Liu et al. 2003; Tomari
et al. 2004a,b, 2007). Central mismatches reduce binding of
small RNA duplexes by the Dcr-2/R2D2 heterodimer, an-
tagonizing Ago2 loading and promoting loading into Ago1
(Fo ¨rstemann et al. 2007; Tomari et al. 2007; Kawamata et al.
2009). The function of the Dcr-2/R2D2 heterodimer in
Ago2 loading is separate and distinct from its role in dicing
siRNAs from long dsRNA: Dcr-2 bearing a glycine to argi-
nine substitution (G31R) in its helicase domain cannot dice,
but can still load siRNA into Ago2 (Lee et al. 2004).
Increasingly, this simple picture of small RNA strand
choice is at odds with the intracellular abundance, process-
ing accuracy, and evolutionary conservation of miRNA*
strands. First, some evolutionarily conserved miRNAs are
less abundant than their miRNA* strands, which appear to
be evolving regulatory functions (Ruby et al. 2007b).
Second, miRNA* 59 ends are far more precisely defined
than their 39 ends, suggesting selective pressure to generate
an accurate seed region—implying that they have regula-
tory targets (Ruby et al. 2007b; Okamura et al. 2008c; Seitz
et al. 2008). Third, there is mounting evidence that some
miRNA*s may have regulatory potential (Ro et al. 2007;
Okamura et al. 2008c; Lin et al. 2009), and fly miRNA*
strands are evolutionarily conserved, albeit not to the same
extent as miRNAs (Okamura et al. 2008c). Thus, miRNA*
strands may regulate gene expression, rather than serve
merely as carriers for loading the miRNA strand. Such a
mechanism would make small RNA biogenesis more ef-
ficient, with each pre-miRNA producing two different reg-
ulatory small RNAs. Nonetheless, miRNAs are typically far
more abundant than their miRNA* counterparts, and reg-
ulation by low abundance Ago1–small RNA complexes has
not been reported in flies.
Here, we show that while most Drosophila miRNAs are
bound to Ago1 in vivo, most miRNA* strands accumulate
bound to Ago2. Partitioning of miRNAs into Ago1 and
Ago2 provides a wide-scale in vivo test for the previously
proposed principles for small RNA sorting in flies: miRNAs
and miRNA* strands are sorted between the two Argonaute
proteins according to the structure of their small RNA
duplex, a process that requires both Dcr-2 and R2D2. Like
the exo-siRNAs that direct RNAi, miRNA* strands bound
to Ago2 typically begin with cytidine, whereas Ago1-bound
miRNAs usually begin with uridine. Thus, the identity of
the first nucleotide of a small RNA plays a role in its sorting
in flies, as previously reported for plants. Finally, miRNA*s
bound to Ago2 are more abundant than siRNAs that direct
RNAi, suggesting that they function to silence target RNAs.
miRNAs and miRNA*s partition differentially
between Ago1 and Ago2
We used high throughput sequencing of 18–29-nt RNA
from fly heads to determine the small RNA profile and
distribution of small RNAs between Ago1 and Ago2 in this
complex somatic structure (Supplemental Table 1). Unlike
other fly tissues, heads express little, if any, Piwi-interacting
RNA, allowing us to focus on small RNAs bound to Ago1
or Ago2 (Ghildiyal et al. 2008). Of the z1.6 million
genome-matching small RNAs sequenced (excluding an-
notated noncoding RNAs such as 2S ribosomal RNA),
90.2% were derived from pre-miRNAs (Fig. 1A). In par-
allel, we used an Ago1 monoclonal antibody (Miyoshi et al.
2005) to immunoprecipitate Ago1-associated small RNAs
from fly head extracts. Nearly 97% of the >5.03 million
Ghildiyal et al.
RNA, Vol. 16, No. 1
small RNA reads associated with Ago1 were miRNAs; only
2.2% were miRNA* strands (Fig. 1A).
Ago2-loaded guide strands acquire a 39 terminal 29-O-
methyl modification after their corresponding passenger
strand is discarded (Horwich et al. 2007; Saito et al. 2007).
To enrich for Ago2-loaded small RNAs, we oxidized the
18–29 nt RNAs prior to library preparation, a treatment
that excludes from the library most Ago1-loaded small
RNAs, which bear 29,39 hydroxyl termini, but allows
sequencing of Ago2-loaded small RNAs, because their
29-O-methyl modification protects them from reaction
with NaIO4(Ghildiyal et al. 2008; Seitz et al. 2008). In
general, the pre-miRNA-derived small RNAs associated
with Ago1 correlated well with the total small RNA profile
(r = 0.91 for miRNAs; r = 0.70 for miRNA* strands),
supporting the view that the majority of small RNAs in fly
heads accumulate because they are bound to Ago1. How-
ever, a global fit of the sum of the miRNA and miRNA*
species detected in the Ago1 immunoprecipitation and the
miRNA and miRNA* species detected in the library pre-
pared from oxidized RNA more closely recapitulated the
total small RNA profile (r = 0.91 for miRNAs; r = 0.85 for
miRNA* strands), suggesting that Ago2-bound miRNA
and/or miRNA* species are a significant component of
the total pre-miRNA–derived small RNA population.
siRNAs were previously identified as the major class of
Ago2-associated endogenous small RNAs in flies (Chung
et al. 2008; Czech et al. 2008; Ghildiyal et al. 2008;
Kawamura et al. 2008; Okamura et al. 2008a,b). Yet, the
population of Ago2-associated small RNAs contained more
miRNA plus miRNA* combined (53.2%) than endo-
siRNAs (33.2%) (Fig. 1A). Thus, the identity of the Dicer
paralog that generates a small RNA does not determine the
Argonaute protein into which it is loaded. Compared to
the total small RNA population—where miRNAs repre-
sented z87.5% of all small RNAs, but miRNA* reads were
just 2.6%—miRNAs were underrepresented (39.4%) and
miRNA*s (13.8%) were overrepresented among the Ago2-
associated small RNA sequences. The abundance of pre-
miRNA–derived small RNAs associated with Ago2 calls
into question the prevailing view that Ago2 is restricted to
the RNAi pathway.
In general, Ago2 was significantly depleted of miRNAs
and enriched for miRNA* sequences (P # 2.2 3 10?16).
Conversely, Ago1 was significantly depleted of miRNA*
sequences and enriched for miRNAs (P # 2.2 3 10?16).
For some of these—especially miRNAs—more of a partic-
ular small RNA was present in Ago1 than in Ago2, but
more of that small RNA was associated with Ago2 than
would be expected by chance. In all, 26 miRNAs and 49
miRNA*s were significantly (P # 0.01) enriched in Ago2,
whereas 71 miRNAs and 9 miRNA*s were significantly
(P # 0.01) enriched in Ago1 (Fig. 1B). Of the 49 miRNA*s
enriched in Ago2, 32 had their corresponding miRNA
enriched in Ago1, while 15 had their miRNA enriched in
Ago2. Among the examples illustrated in Figure 2, the
miRNAs bantam and miR-308 were enriched in Ago1,
whereas bantam* and miR-308* were enriched in Ago2.
Supplemental Table 2 reports the enrichment or depletion
of individual miRNAs and miRNA* species between the
two Drosophila Argonaute proteins.
Although generally less abundant than miRNAs bound
to Ago1, miRNA* isoforms (i.e., all of the species derived
from the same side of the stem of a single pre-miRNA and
sharing a common seed) bound to Ago2 were equally or
more abundant than other small RNAs that exert their reg-
ulatory functions through Ago2, including the well-studied
FIGURE 1. miRNA*s are loaded in Ago2. (A) Relative abundance of
miRNA, miRNA*, and endo-siRNAs among total fly head small RNA,
Ago1-bound small RNAs—inferred from co-immunoprecipitation
with Ago1, and Ago2-bound small RNAs—inferred from their
presence in an oxidized small RNA library. (B) Box plots illustrating
the enrichment scores for all miRNA and miRNA associated with
Ago1 (i.e., in the Ago1 immunoprecipitate) or Ago2 (i.e., in the
oxidized library) and for miRNA and miRNA* that were significantly
(P # 0.01) associated with Ago1 or Ago2. For miRNA* enriched in
Ago2, six outliers with enrichment scores greater than 150 are not
shown: miR-92a* (score = 1206), miR-308* (score = 649), miR-998*
(score = 598), miR-315* (score = 514), miR-2a-2* (score = 309), and
miR-33* (score = 304). (C) Box plots illustrating the abundance of
Ago2-enriched miRNA* and white exo-siRNAs in the total RNA
library. For miRNA* enriched in Ago2, 18 outliers with abundance
greater than 250 ppm are not shown, including miR-8* (2748 ppm)
and miR-34* (1747 ppm).
Drosophila miRNA* strands loaded in Ago2
exo-siRNAs derived from an inverted repeat transgene that
fully silences the white gene via the RNAi pathway (Lee and
Carthew 2003). The median abundance for miRNA* iso-
forms enriched in Ago2 was more than twice that of the
median abundance for white exo-siRNAs bound to Ago2,
and 18 miRNA* were more abundant than the single most
abundant white exo-siRNA detected in the same fly heads.
(These 18 miRNA*s are outliers whose abundance was too
large to display on the box plot in Fig. 1C.) In fact, the
abundance of a single miR-8* isoform alone (2748 parts per
million [ppm]), was nearly two-thirds of the aggregate
abundance of all antisense white exo-siRNAs (4273 ppm),
whose concentration in heads is sufficient to phenocopy a
strong loss-of-function white mutation. Summing the iso-
forms of each miRNA*, 25 miRNA*s were more abundant
than all antisense white exo-siRNAs combined.
The siRNA-loading machinery sorts miRNA* strands
Apart from its function in producing siRNAs, Dcr-2 acts
with its double-stranded RNA-binding domain protein
partner, R2D2, to both load small RNA duplexes into
Ago2 and determine the identity of guide and passenger
strands. Thus, both Dcr-2 and R2D2 are required to load
Ago2 with siRNAs derived from exogenous dsRNA (exo-
siRNAs), such as those derived from a long inverted repeat
transcript designed to silence white mRNA expression (Lee
et al. 2004; Fo ¨rstemann et al. 2005). At least one Drosophila
miRNA, miR-277, which associates equally with Ago1 and
Ago2 in cultured S2 cells, requires Dcr-2 and R2D2 to load
it into Ago2, even though miR-277 requires Dcr-1 to
liberate it from pre-miR-277 (Fo ¨rstemann et al. 2007).
Likewise, those miRNA and miRNA* sequences that
were enriched in Ago2 required Dcr-2 and R2D2 for their
loading (Fig. 3A). The median extent of Ago2 loading of
these miRNAs declined 2.7-fold in dcr-2L811fsXand 3.3-fold
in r2d21heads, compared to wild-type; loading of miRNA*
into Ago2 declined 2.1-fold in dcr-2L811fsXand 3.1-fold in
r2d21. In contrast, the overall abundance of the miRNA se-
quences that were enriched in Ago1 was unaltered in dcr-2
or r2d2 mutant heads.
R2D2 is stabilized by its association with Dcr-2 (Liu et al.
2003; Fo ¨rstemann et al. 2007). Consequently, dcr-2L811fsX
flies are also deficient in R2D2. For miRNA and miRNA*
that were preferentially loaded into Ago2, the effect of the
absence of Dcr-2 and R2D2 on Ago2 loading were well
correlated (r = 0.828) (Fig. 3B,C). As expected, the abun-
dance of miRNA and miRNA* that were preferentially loaded
into Ago1 were largely unchanged in these two mutants.
The median abundance of Ago2-enriched miRNA*
sequences in the total RNA library declined z2.1-fold in
FIGURE 2. Exemplary miRNA and miRNA* duplexes. Typical miRNA/miRNA* duplexes load their miRNA strands into Ago1 and their
miRNA* strands into Ago2. The examples here correspond to duplexes whose miRNA strand was significantly (P # 0.01) enriched in Ago1 and
whose miRNA* strand was enriched in Ago2. These duplexes present different structures to the Ago1 and Ago2 sorting machinery, as the
prospective guide strand occupies a unique position during Argonaute loading. When viewed with the miRNA strand as the guide and the
miRNA* strand as the passenger, the duplex presents a structure with central bulges, mismatches, and G:U wobbles, but when the miRNA* strand
will become the guide and the miRNA strand serves as the passenger, the duplexes present more stably paired central regions. The duplexes are
drawn using the guide isoform that was most abundant for the specific Argonaute protein paired to the most abundant passenger sequence
detected in the total small RNA library. Red text, seed sequence; shaded bars highlight position that are significantly different between Ago1- and
Ago2-loaded guides (see Fig. 4).
Ghildiyal et al.
RNA, Vol. 16, No. 1
FIGURE 3. Association of miRNA* with Ago2 relies on the Ago2-loading machinery. (A) Efficient loading into Ago2 of miRNA and miRNA*
strands—measured by their abundance in an oxidized small RNA library—was diminished in heads from dcr-2L811fsXand r2d21mutants for miRNA
and miRNA* normally enriched in Ago2, but the abundance of Ago1-enriched miRNAs was unaltered, as measured in the total small RNA library.
Box plots illustrate the fold-change between mutant and wild-type. (B,C) The requirement for Dcr-2 and R2D2 for Ago2 loading was well
correlated for miRNA and miRNA* strands preferentially loaded into Ago2. (D) The overall abundance of Ago2-enriched miRNA and
miRNA*—measured in the total small RNA library—decline in ago2 mutant heads. Box plots illustrate the fold-change between mutant and wild-
type in total small RNA libraries.
Drosophila miRNA* strands loaded in Ago2
the absence of Ago2 (Fig. 3D). In contrast, the median
abundance of miRNA enriched in Ago1 was unaltered in
ago2414mutants heads, compared to wild-type (median
fold change = 1.0), a significant difference from the Ago2-
enriched miRNA* (P # 3.1 3 10?8). These data suggest
that in the ago2 mutant, those miRNA* species that nor-
mally are loaded into Ago2 become less stable when that
Argonaute protein is not available. We envision that these
miRNA*/miRNA duplexes, while good substrates for the
Ago2-loading machinery, are poor loading substrates for
the Ago1-loading machinery. In the absence of Ago2,
miRNA*/miRNA duplexes from which the Ago2-enriched
miRNA* are normally loaded into Ago2 can no longer be
used for this purpose. Instead, they are now used as
miRNA/miRNA* duplexes—whose structure typically fa-
vors Ago1 loading—to load their miRNA strand into Ago1.
The observation that abundance of Ago2-enriched miRNA*
sequences declines in ago2414heads supports the earlier
proposal that the duplex features that promote Ago2-
loading are anti-determinants for Ago1 loading (Tomari
et al. 2007; Kawamata et al. 2009).
miRNA/miRNA* duplex structure determines
The Dcr-2/R2D2 heterodimer interprets the structure of a
small RNA duplex, sorting centrally paired duplexes into
Ago2 and leaving duplexes with an unpaired region
centered on guide nucleotide 9 to enter the Ago1 loading
pathway (Fo ¨rstemann et al. 2007; Tomari et al. 2007). Each
small RNA duplex presents two distinct duplexes to the fly
sorting machinery. For example, bantam/bantam* displays
mismatches at guide positions 9 and 10 when viewed from
the 59 end of the miRNA, but these positions are paired
when viewed from the 59 end of the miRNA* strand
(Fig. 2). That is, the bantam/bantam* and bantam*/bantam
duplexes are not equivalent.
To evaluate if miRNA/miRNA* duplexes and miRNA*/
miRNA duplexes generally present distinct structures to the
Drosophila Argonaute loading machineries, we calculated
the pairing probability for each nucleotide in each miRNA/
miRNA* duplex that loads an Ago1- or Ago2-enriched
miRNA or miRNA*/miRNA duplex that loads an Ago1- or
Ago2-enriched miRNA* (Fig. 4A). Viewed in this way, two
significant (p < 0.01) structural differences emerge that
distinguish duplexes that load Ago1 from those that load
Ago2 (Fig. 4B,C): from the perspective of the loaded strand,
Ago1-loading duplexes are more likely to have an unpaired
59 end and a central unpaired region that spans nucleotide
positions 8–11. Conversely, Ago2-loading duplexes more
likely have a paired 59 end and a central region with greater
double-stranded character. Ago2-loading duplexes are also
more likely to have an unpaired guide 39 end (Fig. 4D).
Remarkably, these differences reflect the ‘‘rules’’ for sorting
small RNA duplexes between Ago1 and Ago2 that were
inferred previously from biochemical studies (Tomari et al.
2007; Kawamata et al. 2009). Thus, they provide in vivo
validation of the hypothesis that Drosophila small RNA
duplex structure determines its partitioning between Ago1
The 59 terminal nucleotide of a small RNA reflects
its partitioning between Ago1 and Ago2
Arabidopsis thaliana produces ten distinct AGO proteins,
and small RNAs are sorted among them according to their
first nucleotide. Of the 187 annotated miRNAs in Arabi-
dopsis, z76% begin with uridine, consistent with the idea
that a 59 U steers a small RNA into plant Ago1 (Mi et al.
2008; Montgomery et al. 2008). Arabidopsis Ago2 and Ago4
preferentially load small RNAs that begin with an adeno-
sine, whereas Ago5 favors small RNAs that begin with
cytidine (Mi et al. 2008; Montgomery et al. 2008). Small
RNAs in flies partition between Ago1 and Ago2 according
to the structure of the duplex from which they are loaded,
yet, as in plants, Drosophila miRNAs overwhelmingly begin
with U, whereas U is not overrepresented as the first
nucleotide of siRNAs (Ghildiyal et al. 2008).
We analyzed the sequence composition of Ago1- and
Ago2-loaded miRNA and miRNA* strands present in our
small RNA libraries from fly heads. To prevent differential
rates of transcription or miRNA precursor processing from
skewing our analysis, for each set of small RNAs derived
from a common precursor, we weighted the sequence bias
of each miRNA or miRNA* isoform by its relative abun-
dance, then averaged the sequence bias among all miRNAs
or miRNA* strands, weighting each locus equally (Fig. 5).
Our analysis suggests that the first nucleotide of a fly
small RNA reflects its sorting between Ago1 and Ago2.
miRNAs expressed in fly heads generally began with U
(72%) rather than A (15.2%), C (7.6%), or G (5.2%); for
miRNAs bound to Ago1, as judged by their copurification
with immunoprecipitated Ago1, 73.5% began with U,
whereas 7.1% began with C. Among the miRNA and
miRNA* species that were significantly (p # 0.01) enriched
in Ago1 relative to the total small RNA pool of fly heads,
83.9% began with U; just 3.4% began with C. In contrast,
49% of miRNAs that were enriched in Ago2 began with U;
21.6% began with C and 21.8% began with A, indicating
a selection against a 59 U.
miRNA* strands showed a distinctly different 59 se-
quence bias. The miRNA* detected in fly heads typically
began with A (28.2%), C (32.1%), or G (22.1%), rather
than U (17.6%). In contrast to this overall 59 sequence bias,
those miRNA* that were significantly enriched in Ago1
began either with A (56.3%) or U (29.2%); the population
of miRNA* loaded into Ago1 was depleted of miRNA*
isoforms that begin with C.
Ago2-load miRNA* strands showed the opposite bias:
they typically began with C. Nearly 58% of miRNA* strands
Ghildiyal et al.
RNA, Vol. 16, No. 1
enriched in Ago2 and detected in the oxidized library began
with C, 15.2% began with A, and just 7.7% began with U,
a sequence bias significantly different from the composition
of nucleotides 2–18 of the same small RNAs (P # 6.7 3
10?10, Fisher’s exact test) and from the first nucleotide bias
of miRNA* overall (P # 6.6 3 10?7) and of those miRNA*s
FIGURE 4. Pairing profiles of Ago1- and Ago2-loaded small RNA guides. (A) Box plots illustrate the predicted double-stranded character of each
nucleotide position, 1–19, for all Ago1- or Ago2-enriched miRNA or miRNA* strands. (B) The Wilcoxon test P-value for each comparison was
used to identify nucleotide positions that were significantly different between Ago1-enriched miRNA plus miRNA* compared with Ago2-enriched
miRNA plus miRNA*. The red line indicates P = 0.01. Gray circles, nonsignificant; black circles, significant. (C) Box plots illustrate the differences
in double-stranded character for each position that was significantly different in double-stranded character between Ago1- and Ago2-loaded
miRNA plus miRNA* in B. (D) The data in A–C suggest that miRNA duplexes with less stable 59 ends and central mismatches act as guides for
Ago1 and miRNA duplexes with less stable 39 ends act as guides for Ago2.
Drosophila miRNA* strands loaded in Ago2
loaded into Ago1 (P # 0.017). Overall, 40% of the Ago1-
enriched miRNA or miRNA* species began with U,
whereas 23% of the Ago2-enriched miRNA or miRNA*
species began with C.
Essentially identical sequence biases for both miRNA and
miRNA* were present in independent small RNA libraries
from male and female heads, in libraries prepared from
three distinct genetic backgrounds (Oregon R, dcr-2L811fsX/
CyO, or r2d21/CyO), in libraries of Ago2-associated small
RNAs that were prepared using either oxidation or oxi-
dation followed by b-elimination, and in libraries pro-
cessed and sequenced using two different high throughput
technologies: pyrosequencing (‘‘454’’) or sequencing-by-
synthesis (Illumina Genome Analyzer). Together, these
data suggest that, in flies, a 59 terminal U promotes Ago1
loading, but discourages association with Ago2, whereas
a 59 terminal C directs a small RNA away from Ago1 and
To further test this hypothesis, we analyzed the 59
nucleotide composition of exo-siRNAs derived from a
P-element transgene expressing a long inverted repeat
corresponding to exon 3 of the white mRNA. We compared
the overall population of white exo-siRNAs with those
white exo-siRNAs bound to Ago2, as inferred from their
presence in an oxidized small RNA library. Because the
white exo-siRNA species are transcribed and diced from a
common transcript, differences in their steady-state abun-
dance likely reflect, at least in part, their different pro-
pensities to load into an Argonaute protein. Supporting
this view, white exo-siRNAs levels decline >10-fold in vivo
in a dcr-2L811fsX, r2d21, or ago2414mutant (T Du and PD
Zamore, unpubl.). We therefore weighted each first nucle-
otide according to the abundance of the corresponding
white exo-siRNA species.
Like Ago2-enriched miRNA*, exo-siRNAs isolated from
fly heads typically began with C (39.8%), rather than A
(17.3%), G (20.5%), or U (22.4%), a sequence bias
significantly different from that of the corresponding
strand of the dsRNA from which they are derived (P #
1.8 3 10?9). Among the white exo-siRNAs in the library
prepared from oxidized small RNA—i.e., small RNAs
bound to Ago2—47% began with C. Supporting the view
that the strong C-bias of exo-siRNAs reflects their associ-
ation with Ago2, the 59 C bias was not observed among the
17-fold lower amount of exo-siRNAs that remained in an
r2d21mutant. r2d21mutant flies are defective in loading
exo-siRNAs into Ago2 and do not silence white expression
(Fo ¨rstemann et al. 2005).
To further test the idea that the first nucleotide of a small
RNA duplex influences its sorting between Ago1 and Ago2
in flies, we examined the loading of small RNA duplexes in
vitro, using a previously described UV cross-linking assay
(Tomari et al. 2007). We synthesized two miRNA duplexes,
one corresponding to the authentic Drosophila let-7
miRNA/miRNA* duplex, which begins with a 59 U, and
a second in which the initial U of let-7 was changed to
a 59 C (Fig. 6A). In parallel, we also synthesized two siRNA
duplexes in which the guide strand was either authentic
let-7 (paired to its reverse complement) or let-7 bearing a
59 C instead of a U (Fig. 6A). Each miRNA or guide strand
the proteins, including Ago1 and Ago2, to which it bound
when incubated in Drosophila embryo lysate.
The miRNA/miRNA* duplex containing authentic let-7
strand—i.e., let-7 that began with a 59 U—cross-linked to
Ago1 more efficiently than the let-7 variant that began with
a 59 C (Fig. 6A,B); neither duplex detectably loaded its
miRNA strand into Ago2. Moreover, when we performed
cross-linking in Ago1 immunodepleted lysate, not only was
Ago1 cross-linking absent, but no Ago2 cross-link ap-
peared. We conclude that the structure of a miRNA duplex
not only favors Ago1 loading, but actively prevents loading
of the miRNA into Ago2. Moreover, the interplay between
the structure of the miRNA duplex and its 59 nucleotide
determines its distribution between Ago1 and Ago2.
In contrast to the miRNA/miRNA* duplexes, the siRNA
duplexes cross-linked mainly to Ago2, although some Ago1
cross-linking was clearly detected. For the siRNA duplexes,
32P-radiolabeled, so that cross-linking identified
FIGURE 5. miRNAs and miRNA*s show an Argonaute-specific first
nucleotide bias. miRNAs and miRNA*s associated with Ago1 or Ago2
differ in the bias of their first nucleotide. miRNAs generally begin with
uridine; this bias increased for the subset of miRNA that were Ago1-
bound (measured in the Ago1 immunoprecipitate library), and
increased further for the subset of Ago1-enriched miRNAs (measured
in the total small RNA library). In contrast, Ago2-enrcihed miRNAs
were depleted of 59 uridine in the oxidized small RNA library.
miRNA* strands generally began with adenosine or cytidine. All
miRNA* strands detected in the oxidized library (i.e., loaded in Ago2)
or those enriched in Ago2, were significantly more likely to begin with
cytidine, whereas those miRNA*s enriched in Ago1 were depleted of a
59 cytidine. A 59 cytidine bias was also observed for white exo-siRNAs
and was diminished in r2d21, a mutant defective in Ago2 loading.
Ghildiyal et al.
RNA, Vol. 16, No. 1
the influence of first nucleotide identity on the efficiency
of Ago2 loading was opposite of that observed for the
miRNAs: the siRNA duplex whose guide strand began with
a 59 C loaded more efficiently into Ago2 than the siRNA
duplex whose guide began with U (Fig. 6A,C). Together,
these in vitro data provide strong support for the hypothesis
that the enrichment for a 59 U among Ago1-loaded miRNAs
and for a 59 C among Ago2-loaded miRNA, miRNA*, and
siRNAs reflects a direct role for 59 nucleotide identity in
small RNA sorting between Ago1 and Ago2 in Drosophila.
For some miRNA and miRNA*, distinct isoforms load
into Ago1 and Ago2
At least nine Drosophila pre-miRNA produce from one side
of their stem two small RNAs that partition differentially
between Ago1 and Ago2. Such differentially partitioning
miRNA or miRNA* isoforms differ at their 59 ends and
therefore present subtly different duplexes to the Argonaute-
loading machinery. Moreover, the differentially sorting iso-
forms have different seed sequences, which would allow them
to regulate distinct repertoires of target mRNAs. Figure 7
presents these ‘‘seed switching’’ miRNA and miRNA*
isoforms in the context of the duplexes from which they
are presumed to be loaded into Ago1 or Ago2. Pre-miR-
193 provides a particularly stunning example of such
isoform-specific Argonaute loading. This pre-miRNA
generates two miR-193 isoforms: one begins with a U
and loads into Ago1, whereas a miR-193 isoform that
begins at the next nucleotide, an A, loads into Ago2. Pre-
miR-193 also generates two miR-193* isoforms. Again, the
one that begins with a U loads into Ago1, whereas a less
FIGURE 6. Ago1 prefers to load miRNAs that begin with a 59 uridine, while Ago2 prefers siRNAs that begin with a 59 cytidine. (A) Four small
RNA duplexes were incubated with embryo lysate and then cross-linked with shortwave UV to identify small RNA-bound proteins. Representative
data are shown. (B) Kinetic analysis of miRNA association with Ago1, monitored by UV cross-linking. (C) Kinetic analysis of siRNA association
with Ago2, monitored by UV cross-linking. In B and C, each data point represents the average 6 standard deviation for three trials.
Drosophila miRNA* strands loaded in Ago2
abundant isoform that begins at the G that lies immediately
59 to the U loads into Ago2. This small collection of seed
switching miRNA and miRNA* gives the impression that
the sorting of imperfectly paired small RNA duplexes
between Ago1 and Ago2 reflects a complex interplay
between structural determinants or anti-determinants and
first nucleotide preferences and dislikes.
Historically, miRNAs were defined as the more abundant of
the small RNAs derived from the two sides of a pre-miRNA
stem (Lagos-Quintana et al. 2001; Lau et al. 2001; Lee and
Ambros 2001). The miRNA* strand has been proposed to
be destroyed during Argonaute loading, explaining its
considerably lower abundance (Khvorova et al. 2003;
Schwarz et al. 2003). Yet, high depth sequencing has re-
vealed that many miRNA* species are more abundant than
some miRNA species, and miRNA/miRNA* ratios may
vary dramatically among developmental stages (Ro et al.
2007; Okamura et al. 2008c).
In fly heads and ovaries, several miRNA* strands are
more abundant than their annotated miRNA counterparts
(Table 1). In our data sets, miR-92a was more abundant
than miR-92a* in ovaries (3240 ppm miRNA versus 15 ppm
miRNA*), while its miR-92a* was more abundant than
miR-92a in heads (24 ppm miRNA versus 106 ppm
miRNA*). Likewise, miR-988 (260 ppm miRNA versus
300 ppm miRNA* in heads, but 124 ppm miRNA versus
49 ppm miRNA* in ovaries) and miR-284 (4993 ppm
miRNA versus 915 ppm miRNA* in heads; 49 ppm miRNA
versus 72 ppm miRNA* in ovaries) showed distinctly
different miRNA/miRNA* ratios in ovaries and heads. Such
altered ratios may reflect different concentrations of Ago1
and Ago2 or of components of their respective Argonaute-
loading machineries in the two organs.
Sorting combines structure and sequence information
In general, miRNAs associate with Ago1 and miRNA*
strands associate with Ago2 in Drosophila. It is important
to note that our data argues strongly against a model in
which miRNA* strands bind Ago2 as a consequence of the
corresponding miRNA binding Ago1. First, we can identify
FIGURE 7. miRNA and miRNA* can switch seeds between Ago1 and
Ago2. Depicted are miRNA/miRNA* duplexes that load distinct
isoforms of their miRNA or miRNA* between Ago1 and Ago2,
resulting in seed switching between Argonautes. The duplexes are
drawn pairing the most abundant guide isoform associated with the
particular Argonaute to the most abundant passenger strand isoform
in total head small RNA library. Reads in parts per million represent
the sum of all isoforms that share the same seed as detected in the
total small RNA library. Ratio reports the relative number of reads
for the isoform enriched in Ago1: the number of reads for the
isoform enriched in Ago2 as detected within either the library pre-
pared from Ago1 immunoprecipitated small RNAs (Ago1 ratio) or
oxidized small RNAs (Ago2 ratio). Red text, seed sequence; shaded
bars, determinative positions for small RNA sorting between Ago1
and Ago2; N.D., detected in wild-type, but not detected in the ago2414
TABLE 1. Pre-miRNAs whose miRNA* strands were more
abundant than their miRNAs among small RNAs isolated from
fly heads and fly ovaries.
Ghildiyal et al.
RNA, Vol. 16, No. 1
six miRNAs in which both the miRNA
and the miRNA* strand are enriched in
Ago1 complexes in fly heads. Second,
we find 15 miRNAs for which both the
enriched in Ago2 complexes. Our data
suggest that each miRNA/miRNA* du-
plex presents two distinct structures to
the sorting machinery: One in which the
miRNA is the presumptive guide and
one in which the miRNA* assumes that
position. Evolution appears to have se-
lected for miRNA/miRNA* duplexes that
present sequence and structural features
appropriate for loading Ago1, simulta-
neously favoring Ago2 loading when the
of the miRNA*. Consequently, miRNAs
generally load into Ago1, whereas miRNA*s load into Ago2,
an Argonaute protein previously thought to act only in the
RNAi pathway. miRNA*s are therefore the first class of
Drosophila small silencing RNAs produced by Dicer-1, but
preferentially loaded into Ago2 (Fig. 8).
miRNA/miRNA* duplexes that preferentially load Ago1
are typically less stably paired at their 59 ends and contain
central mismatches, bulges, or G:U wobble pairs, whereas
miRNA*/miRNA duplexes that preferentially load Ago2
possess more stably paired 59 ends and center, but have less
stably paired 39 ends. In addition to structure, sequence
also plays a role in small RNA sorting in flies. Ago1-bound
miRNAs begin overwhelmingly with uridine, whereas
Ago2-bound miRNA, miRNA*, and siRNA tend to begin
with cytidine. Moreover, our in vitro cross-linking exper-
iments show that a 59 U increased the efficiency of miRNA
loading into Ago1, relative to a 59 C, whereas a 59 C—in the
context of an siRNA duplex—increased the efficiency of
Ago2 loading, relative to a 59 U.
The 59 terminal nucleotide of a small RNA is anchored in
the phosphate-binding pocket of Argonaute proteins and
unavailable for base pairing with its RNA target (Ma et al.
2005; Parker et al. 2005). We speculate that the structures
of Ago1 and Ago2 discriminate between U and C by
making specific hydrogen-bonding contacts with the edges
of the first base of a small RNA guide.
The fate of a miRNA/miRNA* duplex, therefore, depends
on multiple factors: structure of its duplex, thermodynamic
stability of the ends of the duplex and the identity of its 59
terminal nucleotide. We do not yet know to what extent
each factor weighs in the sorting decision.
miRNA loci appear to generate an extraordinary di-
versity of functional small RNAs. Some miRNA genes are
transcribed from both DNA strands, producing two differ-
ent hairpins from a single genomic locus (Stark et al. 2008;
Tyler et al. 2008). A few miRNA have been annotated as
miRNA-3p—from a single pre-miRNA, a phenomenon
that we suggest may be the rule rather than the exception.
Our data argue that the two small RNAs, typically anno-
tated as miRNA and miRNA*, from a single pre-miRNA
partition into distinct effector proteins, with the miRNA
loading into Ago1 and the miRNA* loading into Ago2.
These Ago2-loaded miRNA*s are present at levels compa-
rable to exo-siRNAs. Moreover, Ago2-loaded small RNAs
can guide either target cleavage or translational repression
(Iwasaki et al. 2009), suggesting that Ago2-loaded miRNA*s
function to regulate as yet to be identified target RNAs.
Finally, we find that a single arm of a single pre-miRNA
hairpin can give rise to several functional RNA isoforms
that possess different seed sequence and that associate with
different Argonaute proteins that have distinct biological
activities. These three layers of functional diversification—
multiple small RNAs that partition differently from the two
sides of the stem of a single pre-miRNA, different seed
isoforms from a single side of a pre-miRNA stem, and
distinct partitioning of these RNA seed isoforms—allows
a single, compact genomic locus, the miRNA gene, to
produce multiple riboregulators, each with a distinct bi-
ological activity and target repertoire.
MATERIALS AND METHODS
Fly strains were wild-type Oregon R, dcr-2L811fsX, r2d21, and
ago2414. Fly heads were isolated by vigorous shaking of liquid
nitrogen-frozen flies in nested, prechilled sieves (U.S.A. standard
sieve, Humboldt MFG), allowing the heads to pass through the top
sieve (No. 25), and collecting them on the bottom sieve (No. 40).
Small RNA sequencing
Total RNA was extracted with the mirVana kit (Ambion), then
18–30-nt-long RNA was gel purified. 2S rRNA was depleted as
FIGURE 8. A model for small RNA sorting. Sorting of small RNA into an Argonaute is
governed by structure and first nucleotide identity. Consequently, a single miRNA/miRNA*
duplex derived from a single pre-miRNA can present two distinct structures to the Argonaute-
loading machinery. From one end, the duplex can act as a favorable substrate for loading Ago1,
while from the other end, its structure and sequence can favor entry into the RNAi—i.e., the
Drosophila miRNA* strands loaded in Ago2
described (Seitz et al. 2008). A part of the sample was then
oxidized using sodium periodate (Horwich et al. 2007) without
b-elimination step. Size-selected RNA derived from at least 68 mg
total RNA for oxidation; and size-selected RNA derived from at
least 7 mg total RNA for untreated. Library preparation was as
described previously (Ghildiyal et al. 2008). High-throughput
sequencing was by Genome Analyzer II (Illumina). Raw high
throughput sequencing data sets have been deposited at the
National Center for Biotechnology Information Short Read Archive
(www.ncbi.nlm.nih.gov/sites/sra) as GSE18806.
Preparation of fly head extract
Isolated fly heads were transferred to 1.5 mL microcentrifuge
tubes, prechilled in liquid nitrogen, and homogenized using
a plastic ‘‘pellet pestle’’ (Kontes) in 1 mL ice-cold Lysis buffer
(100 mM potassium acetate, 30 mM HEPES-KOH at pH 7.4,
2 mM magnesium acetate) containing 5 mM DTT and 1 mg/mL
complete ‘‘mini’’ EDTA-free protease inhibitor tablets (Roche
Applied Science) per gram of heads. Lysate was clarified by cen-
trifugation at 14,000g for 30 min at 4°C. The supernatant was
dispensed into prechilled microcentrifuge tubes, flash frozen in
liquid nitrogen, and stored at –80°C. Total protein concentration
was determined by Bradford assay.
For small RNA cloning, immunoprecipitation of Ago1 protein
was essentially as described (Miyoshi et al. 2005). Briefly, 40 mL
GammaBind beads (GE Healthcare: #17-0885-01) were washed
four times with 1 mL of Lysis buffer with DTT and protease
inhibitors and containing 0.5% v/v NP-40, then incubated with
40 mL monoclonal anti-Ago1 antibody (Miyoshi et al. 2005) in 1 mL
Lysis buffer at 4°C for 3 h. After washing five times with 1 mL of
Lysis-IP buffer, the antibody-bound beads were incubated with
910 mL fly head lysate (z4.55 mg total protein) at 4°C for 16 h,
and then the supernatant collected and the beads washed five
times with 1 mL of RIPA buffer (50 mM Tris [pH 8.0], 1.0% v/v
NP-40, 150 mM NaCl, 0.5% v/v DOC, 0.1% v/v SDS, 1X Complete-
EDTA-free protease inhibitor cocktail tablet). Immunoprecipitation
efficiency was confirmed by Western blotting.
UV cross-linking was performed in embryo lysates prepared as
described (Tuschl et al. 1999). Embryo lysates were immunode-
pleted for Ago1 as described above. UV cross-linking was as
previously described (Tomari et al. 2007), except that the samples
were z0.5 cm from the UV lamp.
For each sequence read, the first occurrence of the hexamer
perfectly matching the 59 end of the 39 linker was identified.
Sequences without a linker match were discarded. The extracted
inserts for sequences that contained the 39 linker were then
mapped to the Drosophila melanogaster genome (Release R5.5).
Inserts that matched perfectly and completely to the genome were
collected using Bowtie (Langmead et al. 2009), and the corre-
sponding genomic coordinates were determined for downstream
functional analysis. Sequences corresponding to pre-miRNA hair-
pins (miRBase, 13.0) or noncoding RNAs (ncRNAs; Supplemental
Table S3) were identified using the same suffix tree-based
software. Genes were retrieved from FlyBase (R5.5). We manually
curated mature miRNA*. Mature miRNA annotations were
obtained from miRBase (13.0). We allowed sequence reads to
differ in 59 and 39 ends from mature miRNA or miRNA* for up to
9 nt. Endogenous siRNA (endo-siRNA) were defined as genome
mapping 21-mers detected in the oxidized library and that did not
map to ncRNA or miRNA hairpins. Exogenous siRNA (exo-
siRNA) were 21-mers detected in the oxidized library and that
mapped perfectly to the white inverted repeat. Except for Fisher’s
exact test, which requires raw sequence reads, all sequence reads
are reported in parts per million reads of sequencing depth, with
the sequencing depth defined as total number of linker contain-
ing, genome-matching reads excluding ncRNAs.
Fisher’s exact test was applied to each miRNA or miRNA* to
identify those that are enriched in Ago1 or Ago2. Take miR-1 as
an example, the 2 3 2 contingency table includes the following
cells: number of reads of miR-1 detected in the library prepared
from the Ago1 immunoprecipitate, number of reads of all other
miRNA or miRNA* in this library, number of reads of miR-1 in
the library prepared from oxidized small RNA, and number of
reads of all other miRNA or miRNA* in the oxidized library.
P-values # 0.01 were deemed significant. Furthermore, we re-
quired a miRNA or miRNA* enriched in an Argonaute protein to
be at least 10 ppm in that Ago. Enrichment score (Fig. 1B) was
defined as the number of reads of a particular miRNA or miRNA*
in one Argonaute versus the other. A pseudocount (or an
uninformed prior in Bayesian statistics) of 10 ppm was used to
control noise arising from extremely low abundance. For example,
for miR-1 the enrichment score was [(number of miR-1 reads in
Ago1 + 10)/(total number of all miRNA reads in Ago1 + 10)]/
[(number of miR-1 in Ago2 + 10)/(total number of all miRNA
reads in Ago2 + 10)]. Similarly, the fold change in a mutant
compared with the wild type, again using miR-1 as an example,
was defined as (number of miR-1 reads in the mutant + 10)/(total
number of miR-1 reads in the wild-type + 10), where 10 ppm was
Pairing probabilities were calculated using RNAcofold
(ViennaRNA-1.8.3, http://www.tbi.univie.ac.at/RNA/). For each
Argonaute-enriched miRNA or miRNA*, the most abundant
isoform for that miRNA or miRNA* was chosen to be the guide
strand and the corresponding passenger was taken to be the most
abundant isoform of the miRNA* or miRNA from the wild-type
untreated experiment (see Supplemental Discussion for empirical
support for this approach). Both the guide and passenger were
required to pass the aforementioned 10 ppm threshold. The
probability per position was the sum of the pairing probabilities
for that position. Pairing probability for each position was
smoothed by the values of the two neighboring nucleotides. For
each position, we tested the significance of the difference between
all Ago1-enriched miRNA and miRNA* together and all Ago2-
enriched miRNA and miRNA* together using the two-sided
Wilcoxon ranked-sum test with 0.01 as the threshold for signif-
To compute first nucleotide bias, we used an egalitarian weighting
scheme to account for the difference in transcriptional and process-
ing efficiency for different miRNA and miRNA*. The isoforms for
a particular miRNA or miRNA* were weighted by their abundance
Ghildiyal et al.
RNA, Vol. 16, No. 1
in a data set, then all miRNA and miRNA* were weighted equally.
Because white exo-siRNAs are produced from the same transcript,
we weighted all exo-siRNA sequences by their abundance.
Supplemental material can be found at http://www.rnajournal.org.
We thank Alicia Boucher for assistance with fly husbandry, Gwen
Farley for technical assistance, and members of the Zamore and
Weng laboratories for advice, suggestions, and critical comments
on the text. This work was supported in part by grants from the
National Institutes of Health to P.D.Z. (GM62862 and GM65236).
Received October 12, 2009; accepted October 22, 2009.
Ameres SL, Martinez J, Schroeder R. 2007. Molecular basis for target
RNA recognition and cleavage by human RISC. Cell 130: 101–112.
Bernstein E, Caudy AA, Hammond SM, Hannon GJ. 2001. Role for
a bidentate ribonuclease in the initiation step of RNA interference.
Nature 409: 363–366.
Bohnsack MT, Czaplinski K, Gorlich D. 2004. Exportin 5 is a Ran
GTP-dependent dsRNA-binding protein that mediates nuclear
export of pre-miRNAs. RNA 10: 185–191.
Chendrimada TP, Gregory RI, Kumaraswamy E, Norman J, Cooch N,
Nishikura K, Shiekhattar R. 2005. TRBP recruits the Dicer
complex to Ago2 for microRNA processing and gene silencing.
Nature 436: 740–744.
Chung WJ, Okamura K, Martin R, Lai EC. 2008. Endogenous RNA
interference provides a somatic defense against Drosophila trans-
posons. Curr Biol 18: 795–802.
Czech B, Malone CD, Zhou R, Stark A, Schlingeheyde C, Dus M,
Perrimon N, Kellis M, Wohlschlegel JA, Sachidanandam R, et al.
2008. An endogenous small interfering RNA pathway in Drosoph-
ila. Nature 453: 798–802.
Denli AM, Tops BB, Plasterk RH, Ketting RF, Hannon GJ. 2004.
Processing of primary microRNAs by the Microprocessor com-
plex. Nature 432: 231–235.
Fo ¨rstemann K, Tomari Y, Du T, Vagin VV, Denli AM, Bratu DP,
Klattenhoff C, Theurkauf WE, Zamore PD. 2005. Normal micro-
RNA maturation and germ-line stem cell maintenance requires
Loquacious, a double-stranded RNA-binding domain protein.
PLoS Biol 3: e236. doi: 10.1016/j.cell.2007.05.056.
Fo ¨rstemann K, Horwich MD, Wee L, Tomari Y, Zamore PD. 2007.
Drosophila microRNAs are sorted into functionally distinct argo-
naute complexes after production by dicer-1. Cell 130: 287–297.
Ghildiyal M, Zamore PD. 2009. Small silencing RNAs: An expanding
universe. Nat Rev Genet 10: 94–108.
Ghildiyal M, Seitz H, Horwich MD, Li C, Du T, Lee S, Xu J, Kittler EL,
Zapp ML, Weng Z, et al. 2008. Endogenous siRNAs derived from
transposons and mRNAs in Drosophila somatic cells. Science 320:
Gregory RI, Yan KP, Amuthan G, Chendrimada T, Doratotaj B,
Cooch N, Shiekhattar R. 2004. The Microprocessor complex
mediates the genesis of microRNAs. Nature 432: 235–240.
Grimson A, Farh KK, Johnston WK, Garrett-Engele P, Lim LP,
Bartel DP. 2007. MicroRNA targeting specificity in mammals:
Determinants beyond seed pairing. Mol Cell 27: 91–105.
Grishok A, Pasquinelli AE, Conte D, Li N, Parrish S, Ha I, Baillie DL,
Fire A, Ruvkun G, Mello CC. 2001. Genes and mechanisms related
to RNA interference regulate expression of the small temporal RNAs
that control C. elegans developmental timing. Cell 106: 23–34.
Grun D, Wang YL, Langenberger D, Gunsalus KC, Rajewsky N. 2005.
microRNA target predictions across seven Drosophila species and
comparison to mammalian targets. PLoS Comput Biol 1: e13. doi:
Haley B, Zamore PD. 2004. Kinetic analysis of the RNAi enzyme
complex. Nat Struct Mol Biol 11: 599–606.
Hammond SM, Bernstein E, Beach D, Hannon GJ. 2000. An RNA-
directed nuclease mediates post-transcriptional gene silencing in
Drosophila cells. Nature 404: 293–296.
Han J, Lee Y, Yeom KH, Kim YK, Jin H, Kim VN. 2004. The Drosha–
DGCR8 complex in primary microRNA processing. Genes & Dev
Han J, Lee Y, Yeom KH, Nam JW, Heo I, Rhee JK, Sohn SY, Cho Y,
Zhang BT, Kim VN. 2006. Molecular basis for the recognition of
primary microRNAs by the Drosha–DGCR8 complex. Cell 125:
Horwich MD, Li C, Matranga C, Vagin V, Farley G, Wang P,
Zamore PD. 2007. The Drosophila RNA methyltransferase,
DmHen1, modifies germline piRNAs and single-stranded siRNAs
in RISC. Curr Biol 17: 1265–1272.
Hutva ´gner G, McLachlan J, Pasquinelli AE, Balint E´, Tuschl T,
Zamore PD. 2001. A cellular function for the RNA-interference
enzyme Dicer in the maturation of the let-7 small temporal RNA.
Science 293: 834–838.
Iwasaki S, Kawamata T, Tomari Y. 2009. Drosophila argonaute1 and
argonaute2 employ distinct mechanisms for translational repres-
sion. Mol Cell 34: 58–67.
Jackson AL, Bartz SR, Schelter J, Kobayashi SV, Burchard J, Mao M,
Li B, Cavet G, Linsley PS. 2003. Expression profiling reveals off-
target gene regulation by RNAi. Nat Biotechnol 21: 635–637.
Jackson AL, Burchard J, Schelter J, Chau BN, Cleary M, Lim L,
Linsley PS. 2006. Widespread siRNA ‘off-target’ transcript silenc-
ing mediated by seed region sequence complementarity. RNA 12:
Jiang F, Ye X, Liu X, Fincher L, McKearin D, Liu Q. 2005. Dicer-1 and
R3D1-L catalyze microRNA maturation in Drosophila. Genes &
Dev 19: 1674–1679.
Kawamata T, Seitz H, Tomari Y. 2009. Structural determinants of
miRNAs for RISC loading and slicer-independent unwinding. Nat
Struct Mol Biol 16: 953–960.
Kawamura Y, Saito K, Kin T, Ono Y, Asai K, Sunohara T, Okada TN,
Siomi MC, Siomi H. 2008. Drosophila endogenous small RNAs
bind to Argonaute 2 in somatic cells. Nature 453: 793–797.
Ketting RF, Fischer SE, Bernstein E, Sijen T, Hannon GJ, Plasterk RH.
2001. Dicer functions in RNA interference and in synthesis of
small RNA involved in developmental timing in C. elegans. Genes
& Dev 15: 2654–2659.
Khvorova A, Reynolds A, Jayasena SD. 2003. Functional siRNAs and
miRNAs exhibit strand bias. Cell 115: 209–216.
Kim K, Lee YS, Carthew RW. 2006. Conversion of pre-RISC to holo-
RISC by Ago2 during assembly of RNAi complexes. RNA 13: 22–
Krek A, Grun D, Poy MN, Wolf R, Rosenberg L, Epstein EJ,
MacMenamin P, da Piedade I, Gunsalus KC, Stoffel M, et al.
2005. Combinatorial microRNA target predictions. Nat Genet 37:
Lagos-Quintana M, Rauhut R, Lendeckel W, Tuschl T. 2001.
Identification of novel genes coding for small expressed RNAs.
Science 294: 853–858.
Langmead B, Trapnell C, Pop M, Salzberg SL. 2009. Ultrafast and
memory-efficient alignment of short DNA sequences to the human
genome. Genome Biol 10: R25. doi: 10.1186/gb-2009-10-3-r25.
Lau NC, Lim LP, Weinstein EG, Bartel DP. 2001. An abundant class of
tiny RNAs with probable regulatory roles in Caenorhabditis elegans.
Science 294: 858–862.
Lee RC, Ambros V. 2001. An extensive class of small RNAs in
Caenorhabditis elegans. Science 294: 862–864.
Lee YS, Carthew RW. 2003. Making a better RNAi vector for
Drosophila: Use of intron spacers. Methods 30: 322–329.
Drosophila miRNA* strands loaded in Ago2
Lee Y, Ahn C, Han J, Choi H, Kim J, Yim J, Lee J, Provost P, Download full-text
Radmark O, Kim S, et al. 2003. The nuclear RNase III Drosha
initiates microRNA processing. Nature 425: 415–419.
Lee YS, Nakahara K, Pham JW, Kim K, He Z, Sontheimer EJ,
Carthew RW. 2004. Distinct roles for Drosophila Dicer-1 and
Dicer-2 in the siRNA/miRNA silencing pathways. Cell 117: 69–81.
Lee Y, Hur I, Park SY, Kim YK, Suh MR, Kim VN. 2006. The role of
PACT in the RNA silencing pathway. EMBO J 25: 522–532.
Leuschner PJ, Ameres SL, Kueng S, Martinez J. 2006. Cleavage of the
siRNA passenger strand during RISC assembly in human cells.
EMBO Rep 7: 314–320.
Lewis BP, Shih IH, Jones-Rhoades MW, Bartel DP, Burge CB. 2003.
Prediction of mammalian microRNA targets. Cell 115: 787–798.
Lewis BP, Burge CB, Bartel DP. 2005. Conserved seed pairing, often
flanked by adenosines, indicates that thousands of human genes
are microRNA targets. Cell 120: 15–20.
Lim LP, Lau NC, Garrett-Engele P, Grimson A, Schelter JM, Castle J,
Bartel DP, Linsley PS, Johnson JM. 2005. Microarray analysis
shows that some microRNAs downregulate large numbers of
target mRNAs. Nature 433: 769–773.
Lin EA, Kong L, Bai XH, Luan Y, Liu CJ. 2009. miR-199a, a bone
morphogenic protein 2-responsive MicroRNA, regulates chondro-
genesis via direct targeting to Smad1. J Biol Chem 284: 11326–11335.
Liu Q, Rand TA, Kalidas S, Du F, Kim HE, Smith DP, Wang X. 2003.
R2D2, a bridge between the initiation and effector steps of the
Drosophila RNAi pathway. Science 301: 1921–1925.
Lund E, Guttinger S, Calado A, Dahlberg JE, Kutay U. 2004. Nuclear
export of microRNA precursors. Science 303: 95–98.
Ma JB, Yuan YR, Meister G, Pei Y, Tuschl T, Patel DJ. 2005. Structural
basis for 59-end-specific recognition of guide RNA by the
A. fulgidus Piwi protein. Nature 434: 666–670.
Maniataki E, Mourelatos Z. 2005. A human, ATP-independent, RISC
assembly machine fueled by pre-miRNA. Genes & Dev 19: 2979–2990.
Matranga C, Tomari Y, Shin C, Bartel DP, Zamore PD. 2005.
Passenger-strand cleavage facilitates assembly of siRNA into
Ago2-containing RNAi enzyme complexes. Cell 123: 607–620.
Melo SA, Ropero S, Moutinho C, Aaltonen LA, Yamamoto H,
Calin GA, Rossi S, Fernandez AF, Carneiro F, Oliveira C, et al.
2009. A TARBP2 mutation in human cancer impairs microRNA
processing and DICER1 function. Nat Genet 41: 365–370.
Mi S, Cai T, Hu Y, Chen Y, Hodges E, Ni F, Wu L, Li S, Zhou H, Long C,
et al. 2008. Sorting of small RNAs into Arabidopsis argonaute
complexes is directed by the 59 terminal nucleotide. Cell 133: 116–127.
Miyoshi K, Tsukumo H, Nagami T, Siomi H, Siomi MC. 2005. Slicer
function of Drosophila Argonautes and its involvement in RISC
formation. Genes & Dev 19: 2837–2848.
Montgomery TA, Howell MD, Cuperus JT, Li D, Hansen JE,
Alexander AL, Chapman EJ, Fahlgren N, Allen E, Carrington JC.
2008. Specificity of ARGONAUTE7-miR390 interaction and dual
functionality in TAS3 trans-acting siRNA formation. Cell 133:
Okamura K, Ishizuka A, Siomi H, Siomi MC. 2004. Distinct roles for
Argonaute proteins in small RNA-directed RNA cleavage path-
ways. Genes & Dev 18: 1655–1666.
Okamura K, Hagen JW, Duan H, Tyler DM, Lai EC. 2007. The
Mirtron Pathway Generates microRNA-Class Regulatory RNAs in
Drosophila. Cell 130: 89–100.
Okamura K, Balla S,MartinR,Liu N, LaiEC. 2008a. Two distinct mech-
anisms generate endogenous siRNAs from bidirectional transcrip-
tion in Drosophila melanogaster. Nat Struct Mol Biol 15: 581–590.
Okamura K, Chung WJ, Ruby JG, Guo H, Bartel DP, Lai EC. 2008b.
The Drosophila hairpin RNA pathway generates endogenous short
interfering RNAs. Nature 453: 803–806.
Okamura K, Phillips MD, Tyler DM, Duan H, Chou YT, Lai EC.
2008c. The regulatory activity of microRNA* species has sub-
stantial influence on microRNA and 39 UTR evolution. Nat Struct
Mol Biol 15: 354–363.
Parker JS, Roe SM, Barford D. 2005. Structural insights into mRNA
recognition from a PIWI domain-siRNA guide complex. Nature
Parker JS, Parizotto EA, Wang M, Roe SM, Barford D. 2009.
Enhancement of the seed-target recognition step in RNA silencing
by a PIWI/MID domain protein. Mol Cell 33: 204–214.
Rajewsky N, Socci ND. 2004. Computational identification of micro-
RNA targets. Dev Biol 267: 529–535.
Rand TA, Ginalski K, Grishin NV, Wang X. 2004. Biochemical
identification of Argonaute 2 as the sole protein required for
RNA-induced silencing complex activity. Proc Natl Acad Sci 101:
Rand TA, Petersen S, Du F, Wang X. 2005. Argonaute2 cleaves the
anti-guide strand of siRNA during RISC activation. Cell 123: 621–
Ro S, Park C, Young D, Sanders KM, Yan W. 2007. Tissue-dependent
paired expression of miRNAs. Nucleic Acids Res 35: 5944–5953.
Ruby JG, Jan CH, Bartel DP. 2007a. Intronic microRNA precursors
that bypass Drosha processing. Nature 448: 83–86.
Ruby JG, Stark A, Johnston WK, Kellis M, Bartel DP, Lai EC. 2007b.
Evolution, biogenesis, expression, and target predictions of a sub-
stantially expanded set of Drosophila microRNAs. Genome Res 17:
Saito K, Ishizuka A, Siomi H, Siomi MC. 2005. Processing of pre-
microRNAs by the Dicer-1-Loquacious complex in Drosophila
cells. PLoS Biol 3: e235. doi: 10.1371/journal.pbio.0030235.
Saito K, Sakaguchi Y, Suzuki T, Suzuki T, Siomi H, Siomi MC. 2007.
Pimet, the Drosophila homolog of HEN1, mediates 29-O-methyl-
ation of Piwi- interacting RNAs at their 39 ends. Genes & Dev 21:
Schwarz DS, Hutva ´gner G, Du T, Xu Z, Aronin N, Zamore PD. 2003.
Asymmetry in the assembly of the RNAi enzyme complex. Cell
Seitz H, Ghildiyal M, Zamore PD. 2008. Argonaute loading improves
the 59 precision of both microRNAs and their miRNA* strands in
flies. Curr Biol 18: 147–151.
Stark A, Bushati N, Jan CH, Kheradpour P, Hodges E, Brennecke J,
Bartel DP, Cohen SM, Kellis M. 2008. A single Hox locus in
Drosophila produces functional microRNAs from opposite DNA
strands. Genes & Dev 22: 8–13.
Tomari Y, Du T, Haley B, Schwarz DS, Bennett R, Cook HA,
Koppetsch BS, Theurkauf WE, Zamore PD. 2004a. RISC assembly
defects in the Drosophila RNAi mutant armitage. Cell 116: 831–
Tomari Y, Matranga C, Haley B, Martinez N, Zamore PD. 2004b. A
protein sensor for siRNA asymmetry. Science 306: 1377–1380.
Tomari Y, Du T, Zamore PD. 2007. Sorting of Drosophila small
silencing RNAs. Cell 130: 299–308.
Tuschl T, Zamore PD, Lehmann R, Bartel DP, Sharp PA. 1999.
Targeted mRNA degradation by double-stranded RNA in vitro.
Genes & Dev 13: 3191–3197.
Tyler DM, Okamura K, Chung WJ, Hagen JW, Berezikov E,
Hannon GJ, Lai EC. 2008. Functionally distinct regulatory
RNAs generated by bidirectional transcription and processing of
microRNA loci. Genes & Dev 22: 26–36.
Wang Y, Juranek S, Li H, Sheng G, Tuschl T, Patel DJ. 2008. Structure
of an argonaute silencing complex with a seed-containing guide
DNA and target RNA duplex. Nature 456: 921–926.
Yi R, Qin Y, Macara IG, Cullen BR. 2003. Exportin-5 mediates the
nuclear export of pre-microRNAs and short hairpin RNAs. Genes
& Dev 17: 3011–3016.
Zamore PD, Tuschl T, Sharp PA, Bartel DP. 2000. RNAi: Double-
stranded RNA directs the ATP-dependent cleavage of mRNA at 21
to 23 nucleotide intervals. Cell 101: 25–33.
Zeng Y, Yi R, Cullen BR. 2005. Recognition and cleavage of primary
microRNA precursors by the nuclear processing enzyme Drosha.
EMBO J 24: 38–48.
Ghildiyal et al.
RNA, Vol. 16, No. 1