M E C H A N I S M S O F D E V E L O P M E N T 1 2 5 ( 2 0 0 8 ) 4 7 5 – 4 8 5

. sc iencedi rec t . com

ava i lab le a t www

journal homepage: www.elsevier .com/ locate /modo

Drosophila R2D2 mediates follicle formation in somatic tissuesthrough interactions with Dicer-1

Savitha Kalidasa,b, Charcacia Sandersa,b, Xuecheng Yed, Tamara Straussd, Mary Kuhnc,Qinghua Liud, Dean P. Smitha,b,*

aDepartment of Pharmacology, University of Texas Southwestern Medical Center, 5323 Harry Hines Boulevard Dallas,

TX 75390-9111, United StatesbDepartment of Neuroscience, University of Texas Southwestern Medical Center, 5323 Harry Hines Boulevard Dallas,

TX 75390-9111, United StatescDepartment of Molecular Biology, University of Texas Southwestern Medical Center, 5323 Harry Hines Boulevard Dallas,

TX 73390-9033, United StatesdDepartment of Biochemistry, University of Texas Southwestern Medical Center, 5323 Harry Hines Boulevard Dallas,

TX 75390-9038, United States

A R T I C L E I N F O

Article history:

Received 11 April 2007

Received in revised form

10 January 2008

Accepted 14 January 2008

Available online 24 January 2008

Keywords:

RNAi

Oogenesis

Somatic development

Female fertility

0925-4773/$ - see front matter � 2008 Elsevidoi:10.1016/j.mod.2008.01.006

* Corresponding author. Tel.: +1 214 648 16E-mail address: Dean.Smith@utsouthwes

A B S T R A C T

The miRNA pathway has been shown to regulate developmentally important genes. Dicer-1

is required to cleave endogenously encoded microRNA (miRNA) precursors into mature

miRNAs that regulate endogenous gene expression. RNA interference (RNAi) is a gene

silencing mechanism triggered by double-stranded RNA (dsRNA) that protects organisms

from parasitic nucleic acids. In Drosophila, Dicer-2 cleaves dsRNA into 21 base-pair small

interfering RNA (siRNA) that are loaded into RISC (RNA induced silencing complex) that

in turn cleaves mRNAs homologous to the siRNAs. Dicer-2 co-purifies with R2D2, a low-

molecular weight protein that loads siRNA onto Ago-2 in RISC. Loss of R2D2 results in

defective RNAi. However, unlike mutants in other RNAi components like Dicer-2 or Ago-

2, we report here that r2d21 mutants have striking developmental defects. r2d21 mutants

have reduced female fertility, producing less than 1/10 the normal number of progeny.

These escapers have normal morphology. We show R2D2 functions in the ovary, specifically

in the somatic tissues giving rise to the stalk and other follicle cells critical for establishing

the cellular architecture of the oocyte. Most interestingly, the female fertility defects are

dramatically enhanced when one copy of the dcr-1 gene is missing and Dicer-1 protein

co-immunoprecipitates with R2D2 antisera. These data show that r2d21 mutants have

reduced viability and defective female fertility that stems from abnormal follicle cell func-

tion, and Dicer-1 impacts this process. We conclude that R2D2 functions beyond its role in

RNA interference to include ovarian development in Drosophila.

� 2008 Elsevier Ireland Ltd. All rights reserved.

1. Introduction

Small RNAs perform diverse regulatory roles in cells and

organisms that are just beginning to be elucidated (Nilsen,

2007; Zamore and Haley, 2006). One role is gene silencing.

er Ireland Ltd. All rights

50; fax: +1 214 648 1801.tern.edu (D.P. Smith).

RNA interference (RNAi) is a cellular defense mechanism that

suppresses expression of abnormal or parasitic nucleic acids

(reviewed in Tomari and Zamore, 2005; Meister and Tuschl,

2004; Hannon, 2002). Silencing is initiated by cleavage of dou-

ble-stranded RNAs into 21-nucleotide small interfering RNAs

reserved.

476 M E C H A N I S M S O F D E V E L O P M E N T 1 2 5 ( 2 0 0 8 ) 4 7 5 – 4 8 5

(siRNAs, Elbashir et al., 2001) by the nuclease Dicer-2 (Lee

et al., 2004; Carmell and Hannon, 2004). Once produced, siR-

NAs are incorporated into a nuclease complex called RISC

(RNA induced silencing complex) that contains Ago-2, a

member of the Argonaut family (Carmell et al., 2002; Liu

et al., 2003). The transfer of siRNA to Ago-2 is mediated by

R2D2 (Liu et al., 2003; Tomari and Zamore, 2005), a protein

associated with Dicer-2 in vivo (Liu et al., 2003). Ago-2 is the

effector component of RISC, binding siRNA and cleaving

mRNA homologous to the siRNA (Liu et al., 2003; Rand et al.,

2004; Pham et al., 2004). Dicer-2, Ago-2 and R2D2 are all re-

quired for effective RNA interference in vivo (Lee et al., 2004;

Okamura et al., 2004; Liu et al., 2003). R2D2 is required for

RNAi in vivo, probably because in its absence, siRNAs are not

efficiently incorporated into RISC and are degraded (Liu

et al., 2003, 2006).

Genome-encoded microRNAs regulate many genes and are

processed by an RNAi-like pathway. Dicer-1 cleaves stem-loop

precursor miRNAs (pre-microRNAs) into mature miRNAs.

miRNAs are 21-22 nucleotide, single stranded RNAs that are

transferred to Ago-1 which then targets homologous cellular

mRNAs and blocks their translation or initiates their cleavage

(Ambros, 2004; Zeng et al., 2003; Lau et al., 2001). A protein re-

lated to R2D2, called Loquacious or R3D1, interacts with Dicer-

1 and augments its catalytic activity, but does not appear to

load miRNA into Ago-1 (Jiang et al., 2005; Forstemann et al.,

2005; Saito et al., 2005). Therefore, in Drosophila, distinct path-

ways are thought to mediate RNA interference and miRNA

silencing. Dicer-1, Loquacious and Ago-1 mediate miRNA

silencing while Dicer-2, R2D2 and Ago-2 mediate RNAi. Con-

sistent with a central role in miRNA processing and gene reg-

ulation, dcr-1 and ago-1 are lethal when mutated (Lee et al.,

2004; Kataoka et al., 2001). By contrast, mutations in the genes

encoding the RNAi factors Dicer-2 and Ago-2 are viable and

fertile, but defective for RNAi (Lee et al., 2004; Okamura

et al., 2004).

Several genes encoding RNAi-like components, including

the Argonaut homologues aubergine, piwi and the R2D2 homo-

logue loquacious produce oocyte defects when mutated. These

gene products function in the germline cells and have been

best studied in the ovary (Cox et al., 1998; Schmidt et al.,

1999; Wilson et al., 1996; Megosh et al., 2006; Jiang et al.,

2005; Forstemann et al., 2005). The ovary is divided into folli-

cles, each composed of 16 germ cells surrounded by somatic

epithelium of approximately 800 follicle cells (reviewed in

Spradling, 1993; Horne-Badovinac and Bilder, 2005; Berg,

2005). When a newly formed follicle buds from stem cells lo-

cated at the terminal filament it moves posterior to the vitel-

larium and joins the 6 or 7 more mature follicles that were

formed earlier. Each follicle matures sequentially through 14

stages of development defined by morphological features un-

til a mature oocyte is formed (Spradling, 1993). Follicle cells

are derived from stem cells located in the germarium, and

encapsulate the germ cells at the 16-cell cyst stage. Two small

subgroups of follicle cells, the polar cells and stalk cells, dif-

ferentiate early during follicle development. Each follicle

has two pairs of polar cells located at the anterior and poster-

ior poles. Stalk cells form a linear group of cells that separate

the follicles. The stalk cells are induced by polar cells that re-

lease the diffusible ligand unpaired (upd) that activates the

Jak/STAT pathway in stalk cell precursors (McGregor et al.,

2002). Notch signaling is also involved in stalk cell differenti-

ation (Lopez-Schier and St Johnston, 2001). The follicle cells

ultimately cover the oocyte in a single cell epithelial layer, se-

crete the eggshell and yolk proteins and help pattern the oo-

cyte before degenerating at the end of oocyte maturation.

Here we show that r2d21 mutants have reduced viability and

defective female fertility that stems from abnormal follicle

cell function.

2. Results

2.1. Generation of r2d21 mutants

To gain insight into R2D2 function in vivo, we created mu-

tant flies defective for expression of R2D2 protein. P-element

strain p[EP]2450 contains a single P element integrated 0.6 kb

upstream of the r2d2 gene (Drysdale and Crosby, 2005). To cre-

ate mutants defective for r2d2, we mobilized this transposon

and recovered chromosomes lacking the white-eye color mar-

ker carried by the transposon and screened these chromo-

somes for loss of the r2d2 gene (see Section 4). We identified

r2d21 mutants out of approximately 300 excised chromo-

somes. Fig. 1A shows results of a Western blot containing Dro-

sophila embryo extracts from wild type, r2d21 mutants, and

r2d21 mutants carrying a wild type rescuing transgene (res-

cue). Anti-R2D2 antiserum raised to the C-terminal of the

R2D2 protein detects a 35 kD protein lacking in the r2d21 mu-

tant. This protein null mutant was confirmed to be genetically

null, as r2d21/df(2L)XE-3801 is indistinguishable from r2d21/

r2d21 (see Fig. 2E). We have recently mapped this mutation,

and the entire r2d21 ORF is absent (Liu et al., 2006). R2D2

expression was restored in the r2d21 mutant by germline

transformation with a wild type r2d2 genomic construct

introduced into the mutant background (Fig. 1A, rescue; see

Section 4). r2d2 is the only functional gene in this rescuing

genomic DNA fragment, and reverts all defects observed in

r2d21 (see below). Therefore, the r2d21 mutant is a null allele

and the rescuing transgene contains all necessary regulatory

and structural elements for wild type R2D2 function.

2.2. R2D2 is required for embryonic development

Like other mutations in genes required for RNAi like ago-2

and dcr-2, r2d21 mutants that survive to adulthood are viable

with no obvious defects in morphology or behavior. However,

unlike dcr-2 or Ago-2 mutants, the r2d21 stock is difficult to

maintain due to a great reduction in viable offspring. This im-

plies a role for R2D2 in Drosophila fertility or early viability in

addition to its previously reported role in RNA interference

(Liu et al., 2003). Therefore, we set out to understand the role

of R2D2 in early development.

To determine when R2D2 function is required in the course

of development, we first compared the ability of r2d21 mu-

tants and wild type animals to undergo developmental transi-

tions. Zero- to 2-hour-old embryos from wild type, r2d21

mutants and r2d21 mutant-rescued flies were collected and

assayed for hatching. Fig. 1B depicts the percent of larvae that

hatched in each of these genotypes. While 79 ± 0.88% of lar-

vae hatched in the wild type flies, only 14 ± 1.5% of the r2d21

Fig. 1 – r2d21 mutants have defective fertility. (A) Embryo

extracts prepared from w1118, r2d21 homozygotes or flies

carrying a wild type, genomic copy of r2d2 integrated into

the r2d21 mutant background (rescue) resolved by 10% SDS–

PAGE and probed with anti-R2D2 antiserum. Blots were re-

probed with anti-a-tubulin to control for loading. (B) Percent

of embryos that develop to hatching. There was no

significant difference between wild type control and rescue

flies (79 ± 0.88% and 83 ± 2.6% of larvae hatched,

respectively), while only 14 ± 1.5% of r2d21 mutant larvae

hatched (significantly different from controls at P < 0.01 by

ANOVA). (C) Representative 0- to 4-hour-old embryos from

wild type or r2d21 mutants stained with anti-sperm tail

antibody and counterstained with Hoechst stain. Arrow

depicts sperm tail. (D) Female-specific fertility defects in

homozygous r2d21 flies are revealed by matings between

r2d21 mutants and wild type flies. Three mating vials were

set up for each cross. Fertility defect correlates with females,

although zygotic R2D2 from wild type sperm can rescue a

small fraction of these embryos. Significant differences are

observed between bars 1 or 2 compared to 3, and between

bars 3 and 4. All data are presented as means ± SE. For each

data set the statistical significance was tested using ANOVA

for independent observations. ** represents significant

differences at P < 0.01. (E) Anti-R2D2 staining (green) reveals

abundant R2D2 expression in the cytoplasm of stalk cells in

a stage 10 follicle. Nuclei are stained with DAPI. (F) R2D2 co-

localizes with DAPI stained nuclei in follicle epithelial cells.

Anterior is to the left, dorsal at the top.

M E C H A N I S M S O F D E V E L O P M E N T 1 2 5 ( 2 0 0 8 ) 4 7 5 – 4 8 5 477

mutant larvae hatched. The r2d21 mutants carrying a normal

r2d2 gene (rescue flies) showed normal hatching rates,

83 ± 2.6%, similar to controls. These data demonstrate that

there is an early developmental requirement for R2D2

function.

To establish whether the early loss of viable progeny re-

sults primarily from defects in fertilization, we stained 0–4 h

embryos of wild type, r2d21 and r2d21 rescued flies with Hoe-

chst and an antibody directed to the sperm tail. In wild type

controls, virtually all eggs were fertilized (Fig. 1C and Table

1) and these embryos survived through gastrulation. In r2d21

mutants we observed a significant reduction of sperm tail

antibody-positive embryos to 76%. Furthermore, of the fertil-

ized r2d21 embryos, only 19% survived through gastrulation

stage with many showing no signs of nuclear divisions. Thus

the total percentage of fertilized or unfertilized embryos sur-

viving to gastrulation in r2d21 mutants was only 14%. These

data reveal that R2D2 is important for normal early develop-

ment in most embryos, in part due to reduced fertilization.

However, it is not clear if these defects result from female de-

fects or if male fertility is also affected in r2d21.

2.3. r2d21 mutants have defective female fertility

We tested whether the fertility of males, females or both

are affected by loss of R2D2. We compared the number of via-

ble adults produced by control and r2d21 mutant virgin fe-

males mated to either wild type or r2d21 mutant males.

Fig. 1D shows the results of these crosses. r2d21 mutant fe-

males mated with wild type males produce significantly fewer

progeny than controls (P < 0.01, ANOVA). The number of prog-

eny produced by r2d21 males mated to wild type females is

not different from controls. Therefore, the fertility defect in

r2d21 appears to result primarily or exclusively from loss of

R2D2 protein in females. However, zygotic R2D2 contributes

to embryonic viability in addition to the maternally derived

R2D2. Wild type males mated to r2d21 mutant females pro-

duced significantly more viable progeny than r2d21 males ma-

ted to r2d21 females (Fig. 1D, compare last two bars).

Our fertility assay implicates a role for R2D2 in female fer-

tility and early embryonic development. To gain insight into

what cells may require R2D2, we stained wild type ovaries

with anti-R2D2 monoclonal antiserum. We found the stron-

gest expression of R2D2 protein was observed in the stalk

cells (Fig. 1E). R2D2 is also expressed in the follicle cell epithe-

lium (Fig. 1F). Interestingly, the expression of R2D2 in the fol-

licle epithelium appears to be associated primarily with the

nucleus.

2.4. r2d21 mutant females have defective oogenesis

To gain insight into the nature of the female fertility de-

fects in r2d21 mutants, we dissected and examined the ova-

ries from r2d21 mutant females and compared them to wild

type controls. Fig. 2 shows typical examples of the morpho-

logical defects we observed in r2d21 mutant ovaries. There

are several striking defects observed in a large fraction of ova-

ries from r2d21 females (Fig. 2C and D and Table 2). First, many

cysts lack stalks separating them from adjacent follicles.

Fig. 2C shows a typical example where two follicles are di-

rectly apposed, lacking stalks. This phenotype was observed

in 35% of r2d21 follicles. A second phenotype commonly ob-

served in r2d21 ovaries is the presence of large gaps in the fol-

Fig. 2 – r2d21 mutants have ovarian defects. A. Schematic representation of an adult ovariole colored to indicate the somatic

cell identities. The germarium (G), vitellarium (V), interfollicular stalk cells (SC), polar cells (PC) epithelial cells (EC) and

terminal filament (TF) are indicated. (B–H) Ovaries stained with Hoechst. Anterior is to left. (B) Wild type control. Arrow

indicates a typical stalk between follicles. (C) r2d21 mutant ovarioles. (C) Stalks are absent between follicles (arrow). (D)

Degenerating follicle (asterix) with apoptotic germ cell nuclei (arrow head). (E) r2d21/df(2L)4956 ovarioles showing similar

defects observed in r2d21 homozygous mutants. (F) Transgenic rescue ovariole. Defects observed in r2d21 mutants are

reverted by a wild type r2d2 transgene. Arrow depicts a stalk. (G) Wild type stage 10 follicle showing a symmetric array of

follicle cells ensheathing the oocyte. (H) r2d21 mutant stage 10 follicle at exaggerated contrast to show typical gaps in the

follicle cell epithelium.

Table 1 – Fertilization is reduced in r2d21 mutants

w1118 r2d21

Total number of embryos 100 100

Number fertilized 100 76 ± 3**

Number developed to gastrulation 100 24 ± 1.5**

(All data are presented as mean ± SEM, n = 3. Asterisks represent significant differences between genotypes at P < 0.01 using ANOVA for

independent samples).

Table 2 – Fraction of r2d21 ovaries with abnormal chambers

Total no. egg chambers Normal chambers Abnormal chambers

w1118 (n = 1223) 93.6% (n = 1145) 6.3% (n = 78)

w1118; r2d21/CyO (n = 1120) 91.6% (n = 1026) 8.4% (n = 94)

w1118; r2d21 (n = 462) 20.7% (n = 96) 79.3% (n = 366)

w1118; r2d21; dcr-1/+ (n = 112) 1.7% (n = 2) 98.2% (n = 110)

Egg chamber morphology was scored as either normal [wild type] or abnormal [adjacently apposed (35%), compound (15%) or degenerating

(29.3%)].

478 M E C H A N I S M S O F D E V E L O P M E N T 1 2 5 ( 2 0 0 8 ) 4 7 5 – 4 8 5

licle epithelium. Disruptions in the normally uniform distri-

bution of follicle cells covering stage 9 follicle were observed

in more than 50% of follicles (Fig. 2C and D, compare 2G and

H). A third phenotype in r2d21 follicles is that a large fraction

degenerate (Fig. 2D). Finally, Hoescht stain also revealed the

presence of compound follicles where two follicles are fused

M E C H A N I S M S O F D E V E L O P M E N T 1 2 5 ( 2 0 0 8 ) 4 7 5 – 4 8 5 479

together (15% of follicles). In total, over 80% of r2d21 mutant

follicles displayed one or more of these defects, while less

than 20% appeared normal. These defects correlate well with

the fertility defects above.

To further substantiate the defects observed with Hoe-

chst stain, we also examined ovaries stained for specific

markers. To clearly establish that stalk cells are missing in

many r2d21 mutant ovarioles and not just mis-localized,

we introduced a genetically encoded Lac Z marker previ-

ously reported to label these cells. ET93F is a transposable

element containing a Lac Z gene that is expressed exclu-

sively in stalk cells and the terminal filament (Tworoger

et al., 1999). We stained wild type and r2d21 mutants carry-

ing this marker for Lac Z expression. Fig. 3A and C shows

that wild type follicles are typically separated by six Lac Z-

positive cells, consistent with the normal distribution of

stalk cells. By contrast, in r2d21 mutants there is an average

of two Lac Z-positive stalk cells between follicles. These re-

sults are summarized in the graph in Fig. 3. Many follicles

have no Lac Z-positive cells separating the follicles, and

those that do fail to form stalks (Fig. 3B and D). Fig. 4 shows

the defects in the follicle cell epithelium ensheathing the

follicle in r2d21. Fig. 4A and B shows wild type follicles with

normal stalks, while C and D show r2d21 mutants with jux-

taposed follicles. Rhodamine–phalloidin staining with Sytox

green nuclear labeling reveals that there are often two layers

of follicle cells in regions of the epithelium, while other re-

gions lack follicle cells (Fig. 4C and D). The gaps in the fol-

licular epithelium are sometimes large enough for germ

cells to extrude out of the chamber (Fig. 4E). TUNEL labeling

confirms the nuclear fragmentation we observed in many

degenerating follicles by Hoechst staining is in fact TUNEL

positive, indicating these follicles are undergoing apoptosis

(Fig. 4J). Therefore, follicles lacking R2D2 have defects in

their follicle cells that often lead to degeneration of the fol-

licle. Indeed, we noted that r2d21 mutants produce approxi-

mately 1% of the number of eggs compared to wild type

(data not shown). This follicle cell defect is consistent with

the expression of R2D2 in these cells. However, R2D2 could

be affecting these cells in a non-cell-autonomous fashion.

Therefore, we set out to determine whether these defects

result from loss of R2D2 in the germline, the somatic tissues

surrounding the follicles, or both.

Fig. 3 – r2d21 mutants are defective for stalk cells. The expressio

stalk cells in wild type (A, C) and r2d21 mutant (B, D) ovarioles. In

and the stalk cells (arrow heads). Wild type ovarioles have 5–7

terminal filament staining is normal (arrow), but stalk cells are

cells (B, D). The number of stalk cells is significantly reduced in

2.5. R2D2 is required in somatic tissues for ovariandevelopment

To determine whether germ cells or follicle cells require

R2D2 function for female fertility and ovarian development,

we carried out a clonal analysis of R2D2 function. Patches of

r2d21 mutant tissue were generated using FLP-mediated so-

matic recombination (see Section 4) and the effects of differ-

ent mutant cell groups on follicle development were assayed

(Fig. 5). When R2D2 is missing from the follicle cells but pres-

ent in the germ cells, defects in stalk formation are observed

(Fig. 5D–F). When R2D2 is missing from the germ cells but

present in the follicle cells the follicles have stalks (Fig. 5A–

C). These findings were consistently observed in over 10 inde-

pendent clones (not shown). These results indicate that R2D2

is required in the somatic cells for normal stalk cell forma-

tion, and germline expression of R2D2 in the absence of so-

matic expression fails to rescue this defect.

2.6. A genetic interaction between R2D2 and Dicer-1

R2D2 associates with Dicer-2 in Drosophila S2 cells, and it is

required in embryos for RNAi triggered both by long dsRNA

and siRNA (Liu et al., 2003, 2006). If R2D2 acts with Dicer-2

during development, we expect Dicer-2 mutants to have sim-

ilar defects as those observed in r2d21. However, mutants

lacking Dicer-2 have no fertility defects (Lee et al., 2004). Alter-

natively, R2D2 may function with Dicer-1 in embryonic devel-

opment. No previous reports have implicated functional links

between Dicer-1 and R2D2. If R2D2 functions in a pathway

with Dicer-1 we might observe a genetic interaction between

R2D2 and Dicer-1. Therefore we examined whether r2d21 and

dcr-1 have genetic interactions on female fertility.

We generated r2d21 flies that also lack one copy of dcr-1

(Lee et al., 2004). We compared the ovarian phenotype and fer-

tility of these flies with flies with one copy of the mutant dcr-1

allele dcr-1Q1147X alone or missing r2d21 alone. We found that

the female fertility defect we observe in r2d21 mutants is dra-

matically enhanced by loss of a single copy of dcr-1. Indeed,

female flies lacking r2d21 and one copy of dcr-1 are completely

sterile (Fig. 6I). Fig. 6D–F shows that ovaries homozygous for

r2d21 and heterozygous for dcr-1Q1147X are severely disorga-

nized compared to r2d21 homozygotes (Fig. 6C). These

n of the lacZ enhancer trap line, 93F, was used to mark the

wild type, 93F strongly labels the terminal filament (arrow)

stalk cells between follicles (see graph). In r2d21 mutants,

either absent or contain fewer b-galactosidase-positive stalk

r2d21. Graph: *** indicates significant at P < 0.001.

Fig. 4 – Defects in r2d21 follicle revealed by actin staining. Wild type ovary stained with Sytox green nuclear stain (A) and co-

labeled rhodamine–phalloidin (red) for F-actin (B). Note stalk (arrowheads). (C–H) r2d21 mutants show defects in stalk

formation and follicle cell patterning. (C, D) Lack of stalks is apparent (large arrowhead). Large arrow indicates double layers

of follicle epithelium. Small arrow indicates discontinuities in the follicle epithelial layer. (E) Follicle with nurse cells

extruding from the stage 8–10 cyst. Small arrows indicate discontinuities in the follicle cell epithelium where germ cells are

extruded. (F) Fused stage 6–8 follicle with partial follicle cell sheath (arrow). (G) Nuclear fragmentation of germ cells in r2d21

mutant. (H) Degenerating follicle. Arrow indicates cytoskeletal elements condensed into a spherical mass. (I) TUNEL labeling

of a stage 5 wild type ovary. (J) TUNEL staining of a similarly staged r2d21 mutant. Arrows indicate TUNEL-positive

degenerating follicle never observed in the controls.

480 M E C H A N I S M S O F D E V E L O P M E N T 1 2 5 ( 2 0 0 8 ) 4 7 5 – 4 8 5

mutants show completely fused follicles and form no viable

eggs. Missing one copy of dcr-1Q1147X in a wild type r2d2 back-

ground results in normal ovary morphology (Fig. 6A) and nor-

mal fertility (Fig. 6I), indicating these defects do not arise from

the loss of one copy of dcr-1 or from some dominant mutation

on the dcr-1 chromosome. This datum strongly indicates there

is a functional interaction between r2d2 and dcr-1 in the ovary.

2.7. Female fertility defects are not due to reduced overallDicer

One possible explanation for the genetic interaction be-

tween dcr-1 and r2d21 is that there is a specific function served

by a Dicer-1/R2D2 complex in the ovary. For example, a subset

of small RNAs might be processed by Dicer-1/R2D2 that are

important for somatic function in the ovary. However, as

noted previously, mutants lacking R2D2 have reductions in

Dicer-2 protein levels (Liu et al., 2006). Therefore, it is also pos-

sible that the genetic interaction we observe reflects a reduc-

tion in the total Dicer levels in the ovaries. While dcr-1 and dcr-

2 are thought to have distinct roles, perhaps in the ovary their

combined function is required. Thus, the genetic interaction

we observe between r2d2 and dcr-1 may simply reflect an

overall loss of total Dicer activity (Dicer-1 + Dicer-2) in the

r2d21 mutant ovary.

To determine whether the fertility phenotypes are pro-

duced by loss of overall Dicer function or reflect a specific

requirement for R2D2, we compared ovaries from dcr-2R416X;

dcr-1Q1147X/+ flies (lacking 3 of 4 copies of Dicer) with r2d21;

dcr-1Q1147X/+, and dcr-1Q1147X/+. If the effects on ovarian devel-

opment stem from loss of overall Dicer we should see abnor-

malities in the flies missing 3 of 4 copies of dcr-1 and dcr-2 that

are similar or worse than r2d21; dcr-1Q1147X/+. Fig. 6B and I

shows that flies lacking both copies of dcr-2 and one copy of

dcr-1 have normal ovaries and normal fertility. Therefore,

the ovarian defects observed in r2d21 do not result from a

Fig. 5 – R2D2 is required in somatic cells for stalk formation. Mosaic clones of r2d21 mutant tissues (lacking green label) were

generated using hsFLP (Golic and Lindquist, 1989). r2d2+ cells express GFP and are stained green. Blue stain is DAPI stain

labeling nuclei. Cells lacking GFP are homozygous mutant for r2d21. Panels (A–C) shows an example of normal stalks between

chambers lacking R2D2 in the germ line, but wild type for R2D2 in the follicle cells. Panels (D–F) show an example of reduced

stalks between chambers that are wild type for R2D2 in the germ line, but mutant for R2D2 in the follicle cells. (A) The arrow

depicts a chamber with R2D2-positive (GFP+) follicle cells, but is mutant for r2d2 in the germ cells. (B) DAPI staining of the

same follicle. (C) Merged image of A and B. (D) Arrowhead indicates two adjacent follicles mutant for r2d21 in the follicle cells

but wild type for r2d2 in the germ tissues. Note the absence of stalk in E between these two chambers. Arrow indicates a

normal stalk between chambers that are wild type for R2D2 in the follicle cells. (F) Merged image of D and E.

M E C H A N I S M S O F D E V E L O P M E N T 1 2 5 ( 2 0 0 8 ) 4 7 5 – 4 8 5 481

simple loss of overall Dicer activity. The genetic interaction

between r2d21 and dcr-1 reflects, therefore, a specific effect

of missing R2D2 combined with a reduction in Dicer-1.

Having established a genetic interaction between r2d2 and

dcr-1, we set out to determine if R2D2 and Dicer-1 proteins

interact together in a protein complex. We were able to

immunoprecipitate Dicer-1 using R2D2 antiserum (Fig. 6J).

Furthermore, R2D2 antiserum could immunoprecipitate Di-

cer-1. These results indicate that R2D2 and Dicer-1 are present

in a protein complex. We suggest R2D2 functions with Dicer-1

in somatic tissues to form normal follicles in ovarian develop-

ment in addition to its role in RNA interference.

3. Discussion

3.1. R2D2 is required for RNAi, female fertility and earlydevelopment

R2D2 has a well characterized role with Dicer-2 and Ago-2

in RNA interference (Liu et al., 2003). R2D2 forms a stable

heterodimeric complex with Dicer-2 in vivo and in vitro.

R2D2 does not affect the enzymatic activity of Dicer-2, but in-

stead orchestrates the transfer of siRNAs produced by Dicer-2

to Ago-2 (Liu et al., 2003, 2006; Tomari et al., 2004). Dicer-2 and

Ago-2 physically interact in the same complex with R2D2 dur-

ing RNA interference. dcr-2 mutants are defective for RNAi but

have normal fertility (Lee et al., 2004). ago-2 mutants are also

viable and fertile (Okamura et al., 2004; Xu et al., 2003), but

have been recently reported to have defects in nuclear divi-

sions and migration in early embryonic development, with a

reduction in the number of pole cells that give ultimately give

rise to the germline. However, these subtle developmental de-

fects in ago-2 mutants are compensated for during develop-

ment so they have little effect on overall fertility, and were

not reported in the initial characterization of the null mutants

(Okamura et al., 2004). By contrast, r2d21 mutants have strik-

ing fertility defects. These abnormalities are not observed in

dcr-2 and have much higher penetrance and are qualitatively

different from ago-2 mutants, suggesting there is a Dicer-2

and Ago-2-independent role for R2D2 in vivo. The genetic

interaction we observe with Dicer-1 provides the first insights

into this role.

The hatching frequency of r2d21 null mutant embryos was

approximately 14% of that of wild type embryos. This drop in

hatching is due in part to defective fertilization. However, fer-

tilized r2d21 mutant embryos generally failed to develop,

showing no signs of nuclear divisions and were usually ar-

rested before the blastoderm stage. Viability was improved

in embryos fertilized by wild type sperm, indicating R2D2 is

important both maternally and zygotically. Therefore, our

Fig. 6 – r2d21 interacts genetically and physically with dicer-1. Ovaries from different genotypes stained with Hoechst. (A) dcr-

1/+ have normal ovarioles. Arrow indicates normal stalk cells. (B) dcr-2;dcr-1/+ mutant ovarioles lack 3 of 4 copies of the dicer

genes and appear normal. (C) r2d21 mutant ovariole with absence of stalks between adjacent follicles (arrow). (D–F)

r2d21;dicer-1/+ ovariole shows severe defects compared to r2d21 homozygotes (C). The asterisk indicates two completely

fused follicles. (G, H) Rhodamine–phalloidin (G) and Sytox green (I) stained ovarioles missing r2d21 and one copy of dcr-1. (I)

Graph depicting the fertility of females of different genotypes. In each case, females of the indicated genotype were crossed

wild type males. r2d21; dicer-1/+ female flies are completely sterile, while the dicer-1 heterozygotes or dicer-2; dicer-1/+ have

normal fertility. (J) R2D2 antiserum pulls down Dicer-1 and Dicer-1 antiserum pulls down R2D2 from embryo extracts. Lysate

lane is a fraction of the total protein used for the immunoprecipitation. Anti-GST is used as a control.

482 M E C H A N I S M S O F D E V E L O P M E N T 1 2 5 ( 2 0 0 8 ) 4 7 5 – 4 8 5

data indicates that R2D2 contributes to viability early in

development and embryogenesis, in addition to its estab-

lished role in RNA interference.

r2d21 mutants have defective ovaries and are partially ster-

ile. The ovaries from these mutants show a range of pheno-

typic defects, all of which are completely rescued by

introduction of an r2d2 transgene into the mutant back-

ground, clearly demonstrating loss of R2D2 produces these

defects in oogenesis. Clonal analysis reveals a requirement

for R2D2 in the somatic follicle cells, but our data indicates

there is not a germline requirement for R2D2. Removal of

R2D2 specifically from somatic cells results in the typical

ovarian architecture defects including loss of stalks observed

in r2d21, while loss of R2D2 in germ cells did not affect follicle

morphology. The follicle cell phenotypes we observe in r2d21,

while broader in scope, are similar to those previously de-

scribed for defective polar cell specification and formation

(Lopez-Schier and St Johnston, 2001). Once specified, polar

cells are known to secrete a variety of signals that pattern

the follicular epithelium (McGregor et al., 2002). The robust

expression of R2D2 protein in the stalk cells will focus our fu-

ture efforts on candidate signaling pathways expressed in

these cells.

Several genes related to RNAi components have defects in

ovarian development. These include piwi, aubergine and Spin-

dle-E. However, the phenotypes associated with these other

mutants are quite different from those reported here for

r2d21. piwi is an Argonaut homologue involved in stem cell

maintenance, co-suppression, heterochromatin formation,

and transposon silencing (Megosh et al., 2006; Kalmykova

et al., 2005; Pal-Bhadra et al., 2004, 2002; Sarot et al., 2004;

Cox et al., 1998). However piwi, while expressed in both so-

matic and germline cells, is required in the terminal filament

for stem cell maintenance but is also needed for cell division

in the germline (Cox et al., 2000; Szakmary et al., 2005). Auber-

gine and spindle-E are also involved in heterochromatin forma-

tion and aubergine plays a role in suppression of stellate

required for male fertility (Aravin et al., 2004). However, auber-

gine and spindle-E are required in the germline (Cox et al.,

1998; Kennerdell et al., 2002). Most recently Loquacious was

M E C H A N I S M S O F D E V E L O P M E N T 1 2 5 ( 2 0 0 8 ) 4 7 5 – 4 8 5 483

reported to have a role in germline stem cell maintenance

and stellate suppression (Aravin et al., 2004; Forstemann

et al., 2005). R2D2 appears unique among the RNAi-related

components identified to date because it is required in the fol-

licle tissue and not in the germline, and has no effect on male

fertility. Therefore, the function of R2D2 in female fertility ap-

pears to be distinct from these other RNAi related proteins.

3.2. R2D2 interacts with a Dicer-1-dependent processduring ovarian development

The genetic and physical interaction between Dicer-1 and

R2D2 implies these gene products are operating in the same

pathway to orchestrate female follicle cell patterning. Dicer-

1 mutants are lethal, likely due to loss of mature miRNAs

and subsequent regulation of endogenous gene function.

r2d21 flies can survive to adulthood and are morphologically

normal, so R2D2 is unlikely to have a significant role in pro-

cessing most miRNAs. One possibility is that Dicer-1 and

R2D2 are partnered in the ovary to process one or a small

number of miRNAs in the follicle cells. Dicer-1 has three splic-

ing variants, so specific splicing variants of Dicer-1 could

potentially partner with R2D2 (Kataoka et al., 2001). Alterna-

tively, R2D2 and Dicer-1 may perform an unknown function

independent of miRNA processing. Screens to identify sup-

pressor mutations of r2d21 with increased fertility will prove

valuable in providing future insights into R2D2-dependent

follicle formation.

4. Experimental procedures

4.1. Fly stocks and generation of r2d21 null mutants

Control strain was w1118 (Lindsley and Zimm, 1992). Df(2L)XE-3801 covers the r2d21 region and was obtained from the Bloom-ington Stock Center. r2d21 was derived by mobilization of the Pelement from stock p[EP]2450. r2d21 mutants were generated bycrossing p[EP]2450 to a stock expressing transposase (Robertsonet al., 1988) and recovering chromosomes lacking the eye colormarker carried by the P element. Candidate mutants were madehomozygous and screened for lack of a PCR product using PCRprimers corresponding to the 5 0 part of the R2D2 coding region.DNA from r2d21 failed to amplify this fragment and was con-firmed to be defective for R2D2 expression by Western blotting.r2d21 is a 5 kb deletion that has subsequently been mapped andremoves the entire open reading frame of r2d2 (Liu et al., 2006).FRT82B dcr-1Q1147X/TM3 and FRT42D dcr-2R416X were described pre-viously (Lee et al., 2004).

4.2. Transgenic rescue

To construct the r2d2 rescue transgene, a 5.6 kb genomic frag-ment containing 3 kb upstream of the starting methionine of r2d21

and 1 kb downstream of the polyadenylation signal was clonedinto pCasper4 (Pirrotta, 1988). This fragment contains a large por-tion of an uncharacterized gene (CG14536) with similarity to ubiq-uitin ligase that could complicate interpretation of the R2D2rescue phenotype. Therefore a frameshift mutation was intro-duced into the CG14536 reading frame at amino-acid 12 by digest-ing the rescue fragment at a unique SacII site, filling in withKlenow DNA polymerase, and religating. Disruption of this read-ing frame was confirmed by sequencing. Germline transformants

were generated by standard procedures. A transgenic strain withthe rescuing P element integrated into the third chromosomewas crossed into the r2d21 mutant background.

4.3. Embryo viability assay

Fertilized eggs were transferred by hand to apple juice–agaroseplates and monitored until hatched. Three replicates were as-sayed for each genotype. For each data set the statistical signifi-cance was tested using ANOVA for independent observations. Inall cases P < 0.05 was considered significant.

4.4. Western immunoblots and immuno- cytochemistry

Protein samples were analyzed on 10% SDS–polyacrylamidegels (Laemmli, 1970) and transferred to nitrocellulose membrane(Optitran BA-S 83, S&S, Dassel, Germany). R2D2 was detectedwith affinity purified anti-R2D2 antiserum (Liu et al., 2003).Anti-a-tubulin antibody was used to control for loading (Sigma,Saint Louis, MO). Dicer-1 antibody (generated by Q. L.) was usedat 1:2000 dilution. For immunocytochemistry, embryos were col-lected from apple juice–agar plates every 2 h, dechorionated in50% Clorox bleach for 1–2 min, rinsed with PBT (1% Triton X-100 in PBS), and fixed in a 1:1 mixture of fixative (97% metha-nol + 3% 0.5 M EGTA, pH 7.4) and heptane, with gentle rockingfor 5 min. Subsequently the embryos were shaken vigorouslyfor 30–60 s to remove vitelline membranes. The heptane layerwas removed and replaced by an equal volume of cold metha-nol fix, and the embryos were rocked at 4� for 2 h, after whichthey were gradually rehydrated into PBT (phosphate bufferedsaline with 0.1% Tween-20) and antibody staining was carriedout. Anti-sperm tail antibody (gift of D. McKearin) was used ata dilution of 1:1000 (Karr, 1991). The embryos were mountedin PBS–glycerol on slides and under a glass coverslip sealed withnail polish for viewing. Ovary dissections, fixation and immuno-histochemistry were performed as described previously (Ohl-stein and McKearin, 1997). Mouse anti-b-galactosidase(Promega, Madison, WI), and rabbit anti-b-galactosidase (Cappel,Durham, NC) were used at a dilution of 1:2500 and 1:10,000respectively. Monoclonal mouse anti-R2D2 was generatedagainst full-length R2D2 protein. The hybridoma cell culturesupernatant was used at a dilution of 1:1. Secondary antibodiesincluded anti-mouse Alexa 568 diluted at 1:300 (Jackson, WestGrove, PA) and anti-rabbit Alexa 488 used at the same dilution.In most cases, Hoechst staining (1:10,000; Molecular Probes, Eu-gene, OR) was performed to stain nuclei except for confocalimages where Sytox Green was substituted (1:50,000 MolecularProbes, Eugene, OR). TUNEL labeling was performed using Apop-Tag apoptosis detection kit according to the manufacturer(Chemicon, Temecula, CA) Actin visualization was performedby incubating ovaries in 1–2 units of rhodamine–phalloidin(Molecular Probes, Carlsbad, CA) Fluorescent images were ob-tained using a Zeiss LSM410 confocal microscope.

4.5. Immunoprecipitation

Embryo lysate was used for immunoprecipitation and was per-formed as previously described in Sevrioukov et al. (1999). Dicer-1antiserum was used at 1:1000 dilution.

4.6. Generation of mutant clones

r2d21 loss-of-function clones were generated by the FRT/FLPrecombination technique (Golic and Lindquist, 1989; Xu and Ru-bin, 1993). r2d21 was recombined onto a FRT40A chromosomeand females were crossed to males carrying hsFLP and FRT40A

484 M E C H A N I S M S O F D E V E L O P M E N T 1 2 5 ( 2 0 0 8 ) 4 7 5 – 4 8 5

GFP. Homozygous mutant clones in follicle cells were produced bya 1-hour heat shock of adult females at 37 �C. The heat shock wasrepeated two times on the first day and the same animals wereheat shocked again for another two days to achieve higher fre-quency of mitotic clones. Animals were then examined for mor-phological alterations by immunostaining in ovaries 2–5 dayspost heat shock. Cells made homozygous for r2d21 were recog-nized as GFP-negative against a background of GFP-positive, wildtype cells. Germline clones were generated by heat shocking thirdinstar larvae and pupae using the same regime described above.

Acknowledgements

We thank Tim Rand for helpful discussions and especially

Dennis McKearin for reagents and consultation over the ovar-

ian defects presented here. This work was supported by the

National Institutes of Health (GM 070684) to D.S. and (GM

078163) to Q.L. Q. L. is a W. A. ‘‘Tex’’ Moncrief Jr. Scholar in

Medical Research and a Damon Runyon Scholar.

R E F E R E N C E S

Ambros, V., 2004. The functions of animal microRNAs. Nature 431,350–355.

Aravin, A.A., Klenov, M.S., Vagin, V.V., Bantignies, F., Cavalli, G.,Gvozdev, V.A., 2004. Dissection of a natural RNA silencingprocess in the Drosophila melanogaster germ line. Mol. Cell. Biol.24, 6742–6750.

Berg, C.A., 2005. The Drosophila shell game: patterning genes andmorphological change. Trends Genet. 21, 346–355.

Carmell, M.A., Hannon, G.J., 2004. RNAse III enzymes and theinitiation of gene silencing. Nat. Struct. Mol. Biol. 11, 214–218.

Carmell, M.A., Xuan, Z., Zhang, M.Q., Hannon, G.J., 2002. TheArgonaut family: tentacles that reach into RNAi,developmental control, stem cell maintenance, andtumorigenesis. Genes Dev. 16, 2733–2742.

Cox, D., Chao, A., Baker, J., Chang, L., Qiao, D., Lin, H., 1998. Anovel class of evolutionary conserved genes defined by piwiare essential for stem cell self renewal. Genes Dev. 12, 3715–3727.

Cox, D.N., Chao, A., Lin, H., 2000. Piwi encodes a nucleoplasmicfactor whose activity modulates the number and division rateof germline stem cells. Development 127, 503–514.

Drysdale, R.A., Crosby, M.A., 2005. FlyBase: genes and genemodels. Nucleic Acids Res. 33, D390–D395.

Elbashir, S.M., Harborth, J., Lendeckel, W., Yalcin, A.W.K., Tuschl,T., 2001. Duplexes of 21-nucleotide RNAs mediate RNAinterference in cultured mammalian cells. Nature 411, 494–498.

Forstemann, K., Tomari, Y., Du, T., Vagin, V.V., Denli, A.M., Bratu,D.P., Klattenhoff, C., Theurkauf, W.E., Zamore, P.D., 2005.Normal microRNA maturation and germ-line stem cellmaintenance requires Loquacious, a double-stranded RNA-binding domain protein. PLoS Biol. 3, e236.

Golic, K.G., Lindquist, S., 1989. The FLP recombinase of yeastcatalyzes site-specific recombination in the Drosophilagenome. Cell 59, 499–509.

Hannon, G.J., 2002. RNA interference. Nature 418, 244–251.Horne-Badovinac, S., Bilder, D., 2005. Mass transit: epithelial

morphogenesis in the Drosophila egg chamber. Dev. Dyn. 232,559–574.

Jiang, F., Ye, X., Liu, X., Fincher, L., McKearin, D., Liu, Q., 2005.Dicer-1 and R3D1-L catalyze microRNA maturation inDrosophila. Genes Dev. 19, 1674–1679.

Kalmykova, A.I., Klenov, M.S., Gvozdev, V.A., 2005. Argonauteprotein PIWI controls mobilization of retrotransposons in theDrosophila male germline. Nucleic Acids Res. 33, 2052–2059.

Karr, T.L., 1991. Intracellular sperm/egg interactions in Drosophila:a three-dimensional structural analysis of a paternal productin the developing egg. Mech. Dev. 34, 101–111.

Kataoka, Y., Takeichi, M., Uemura, T., 2001. Developmental rolesand molecular characterization of a Drosophila homologue ofArabidopsis Argonaute 1, the founder of a novel super family.Genes Cells 6, 313–325.

Kennerdell, J.R., Yamaguchi, S., Carthew, R.W., 2002. RNAi isactivated during Drosophila oocyte maturation in a mannerdependent on aubergine and spindle-E. Genes Dev. 16, 1884–1889.

Laemmli, U.K., 1970. Cleavage of structural proteins duringassembly of the head of bacteriophage T4. Nature 227, 680–685.

Lau, N.C., Lim, L.P., Weinstein, E.G., Bartel, D.P., 2001. An abundantclass of tiny RNAs with probable regulatory roles inCaenorhabditis elegans. Science 294, 858–862.

Lee, Y.S., Nakahara, K., Pham, J.W., Kim, K., He, Z., Sontheimer,E.J., Carthew, R.W., 2004. Distinct roles for Drosophila Dicer-1and Dicer-2 in the siRNA/miRNA silencing pathways. Cell 117,69–81.

Lindsley, D.L., Zimm, G., 1992. The Genome of Drosophilamelanogaster. Academic Press, San Diego.

Liu, Q., Rand, T.A., Kalidas, S., Du, F., Kim, H.-E., Smith, D.P., Wang,X.-D., 2003. R2D2, a bridge between the initiation and effectorsteps of the Drosophila RNAi pathway. Science 301, 1921–1925.

Liu, X., Jiang, F., Kalidas, S., Smith, D., Liu, Q., 2006. Dicer-2 andR2D2 coordinately bind siRNA to promote assembly of thesiRISC complexes. RNA 12, 1514–1520.

Lopez-Schier, H., St Johnston, D., 2001. Delta signaling from thegerm line controls the proliferation and differentiation of thesomatic follicle cells during Drosophila oogenesis. Genes Dev.15, 1393–1405.

McGregor, J.R., Xi, R., Harrison, D.A., 2002. JAK signaling issomatically required for follicle cell differentiation inDrosophila. Development 129, 705–717.

Megosh, H.B., Cox, D.N., Campbell, C., Lin, H., 2006. The role ofPIWI and the miRNA machinery in Drosophila germlinedetermination. Curr. Biol. 16, 1884–1894.

Meister, G., Tuschl, T., 2004. Mechanisms of gene silencing bydouble-stranded RNA. Nature 431, 343–349.

Nilsen, T.W., 2007. RNA 1997-2007: a remarkable decade ofdiscovery. Mol. Cell 28, 715–720.

Ohlstein, B., McKearin, D., 1997. Ectopic expression of theDrosophila Bam protein eliminates oogenic germline stemcells. Development 124, 3651–3662.

Okamura, K., Ishizuka, A., Siomi, H., Siomi, M.C., 2004. Distinctroles for Argonaute proteins in small RNA-directed RNAcleavage pathways. Genes Dev. 18, 1655–1666.

Pal-Bhadra, M., Bhadra, U., Birchler, J.A., 2002. RNAi relatedmechanisms affect both transcriptional andposttranscriptional transgene silencing in Drosophila. Mol. Cell9, 315–327.

Pal-Bhadra, M., Bhadra, U., Birchler, J.A., 2004. Interrelationship ofRNA interference and transcriptional gene silencing inDrosophila. Cold Spring Harb. Symp. Quant. Biol. 69, 433–438.

Pham, J.W., Pellino, J.L., Lee, Y.S., Carthew, R.W., Sontheimer, E.J.,2004. A Dicer-2 dependent 80s complex cleaves targetedmRNAs during RNAi in Drosophila. Cell 117, 83–94.

Pirrotta, V., 1988. Vectors for P-mediated transformation inDrosophila. In: Rodriguez, R.L., Denhart, D.T. (Eds.), Vectors: ASurvey of Molecular Cloning Vectors and Their Uses.Butterworths, Boston, pp. 437–456.

Rand, T.A., Ginalski, K., Grishin, N.V., Wang, X., 2004. Biochemicalidentification of Argonaute 2 as the sole protein required for

M E C H A N I S M S O F D E V E L O P M E N T 1 2 5 ( 2 0 0 8 ) 4 7 5 – 4 8 5 485

RNA-induced silencing complex activity. Proc. Natl. Acad. Sci.USA 101, 14385–14389.

Robertson, H.M., Preston, C.R., Phillis, R.W., Johnson-Schlitz, D.M.,Benz, W.K., Engles, W.R., 1988. A stable source of P elementtransposase in Drosophila melanogaster. Genetics 118, 461–470.

Saito, K., Ishizuka, A., Siomi, H., Siomi, M.C., 2005. Processing ofpre-microRNAs by the Dicer-1-Loquacious complex inDrosophila cells. PLoS Biol. 3, e235.

Sarot, E., Payen-Groschene, G., Bucheton, A., Pelisson, A., 2004.Evidence for a piwi-dependent RNA silencing of the gypsyendogenous retrovirus by the Drosophila melanogaster flamencogene. Genetics 166, 1313–1321.

Schmidt, A., Palumbo, G., Bozzetti, M., Tritto, P., Pimpinelli, S.,Schafer, U., 1999. Genetic and molecular characterization ofsting, a gene involved in crystal formation and meiotic drive inthe male germ line of Drosophila melanogaster. Genetics 151,749–760.

Sevrioukov, E.A., He, J.P., Moghrabi, N., Sunio, A., Kramer, H., 1999.A role for the deep orange and carnation eye color genes inlysosomal delivery in Drosophila. Mol. Cell 4, 479–486.

Spradling, A., 1993. The Development of Drosophila melanogaster.Cold Spring Harbor Press, Cold Spring Harbor.

Szakmary, A., Cox, D.N., Wang, Z., Lin, H., 2005. Regulatoryrelationship among piwi, pumilio, and bag-of-marbles in

Drosophila germline stem cell self-renewal and differentiation.Curr. Biol. 15, 171–178.

Tomari, Y., Matranga, C., Haley, B., Zamore, P.D., 2004. A proteinsensor for siRNA asymmetry. Science 306, 1377–1380.

Tomari, Y., Zamore, P.D., 2005. Perspective: machines for RNAi.Genes Dev. 19, 517–529.

Tworoger, M., Larkin, M.K., Bryant, Z., Ruohola-Baker, H., 1999.Mosaic analysis in the Drosophila ovary reveals a commonhedgehog-inducible precursor stage for stalk and polar cells.Genetics 151, 739–748.

Wilson, J.E., Connell, J.E., Macdonald, P.M., 1996. Aubergineenhances oskar translation in the Drosophila ovary.Development 122, 1631–1639.

Xu, C.B., Pusateri, L., Barbarosa, V., Lehmann, R., 2003. l(3)malignant brain tumor and three novel genes are requiredfor Drosophila germ cells formation. Genetics 165, 1889–1900.

Xu, T., Rubin, G., 1993. Analysis of genetic mosaics in developingand adult Drosophila tissues. Development 117, 1223–1237.

Zamore, P.D., Haley, B., 2006. Ribo-genome: the big world of smallRNAs. Science 309, 1519–1524.

Zeng, Y., Yi, R., Cullen, B.R., 2003. microRNAs and smallinterfering RNAs can inhibit mRNA expression by similarmechanims. Proc. Natl. Acad. Sci. USA 100, 9779–9784.