Piwi Is Required in Multiple Cell Types to Control Germline Stem Cell Lineage Development in the Drosophila Ovary

The piRNA pathway plays an important role in maintaining genome stability in the germ line by silencing transposable elements (TEs) from fly to mammals. As a highly conserved piRNA pathway component, Piwi is widely expressed in both germ cells and somatic cells in the Drosophila ovary and is required for piRNA production in both cell types. In addition to its known role in somatic cap cells to maintain germline stem cells (GSCs), this study has demonstrated that Piwi has novel functions in somatic cells and germ cells of the Drosophila ovary to promote germ cell differentiation. Piwi knockdown in escort cells causes a reduction in escort cell (EC) number and accumulation of undifferentiated germ cells, some of which show active BMP signaling, indicating that Piwi is required to maintain ECs and promote germ cell differentiation. Simultaneous knockdown of dpp, encoding a BMP, in ECs can partially rescue the germ cell differentiation defect, indicating that Piwi is required in ECs to repress dpp. Consistent with its key role in piRNA production, TE transcripts increase significantly and DNA damage is also elevated in the piwi knockdown somatic cells. Germ cell-specific knockdown of piwi surprisingly causes depletion of germ cells before adulthood, suggesting that Piwi might control primordial germ cell maintenance or GSC establishment. Finally, Piwi inactivation in the germ line of the adult ovary leads to gradual GSC loss and germ cell differentiation defects, indicating the intrinsic role of Piwi in adult GSC maintenance and differentiation. This study has revealed new germline requirement of Piwi in controlling GSC maintenance and lineage differentiation as well as its new somatic function in promoting germ cell differentiation. Therefore, Piwi is required in multiple cell types to control GSC lineage development in the Drosophila ovary.


Introduction
Small RNAs have received much attention in recent years because of their important and diverse roles in the regulation of various biological processes [1,2,3,4,5]. In contrast to other small RNAs, Piwi-associated small RNAs, also known as piRNAs, are abundantly expressed in germ cells of organisms ranging from C. elegans to human, and have emerged as an important class of small RNAs for maintaining genome stability in germ cells [6,7,8,9]. Recent studies have shown that piRNAs also function in somatic cells to regulate gene expression and repress TEs [10,11,12,13,14,15]. However, biological functions of piRNAs still remain poorly defined.
The Drosophila ovary is an attractive system for studying stem cell lineage development [16]. Two types of stem cells, germline stem cells (GSCs) and follicular stem cells (FSCs), are responsible for continuously producing differentiated germ cell cysts and follicle cells, respectively, which are assembled into egg chambers that eventually develop into mature oocytes. Two or three GSCs are situated in the tip of each ovariole, known as the germarium, and can be easily identified by their direct contact with cap cells and presence of an anteriorly localized spectrosome (Fig. 1A). Immediate GSC daughters, also known as cystoblasts (CBs), move away from cap cells and undergo four rounds of synchronized cell division to form 2-cell, 4-cell, 8-cell and 16-cell cysts. CBs and cysts are tightly encased by cellular processes of escort cells (ECs), also known as inner germarial sheath cells (Fig. 1A). Genetic and cell biological studies have shown that terminal filament (TF)/cap cells and anterior ECs form the self-renewing niche for GSCs, which provides the essential BMP signal for repressing GSC differentiation and thereby maintaining their self-renewal [16].
Based on recent studies from us and others [17,18], we have recently proposed that posterior ECs function as the microenvironment or niche for promoting germ cell differentiation [18]. One of the key functions of ECs is to prevent BMP signaling via two distinct strategies. First, EGFR-MAPK signaling has been proposed to directly repress expression of dally, encoding a proteoglycan facilitating BMP signal transduction and diffusion [17]. Rho signaling and Eggless have been shown to repress dally expression in ECs, thus promoting germ cell differentiation, but it remains unclear how they might regulate dally expression [18,19]. The second strategy is direct repression of transcription of dpp, which encodes a BMP ligand essential for GSC self-renewal in Drosophila. Histone lysine-specific demethylase 1 (Lsd1, a chromatin regulator) and Rho signaling have been shown to repress dpp transcription in ECs [18,20]. dpp knockdown can partially rescue the germ cell differentiation defects caused by inactivation of Lsd1 and Rho signaling in ECs, indicating that dpp upregulation contributes to the germ cell differentiation defects. Therefore, ECs have so far been demonstrated to promote germ cell differentiation by preventing the spreading of BMP signaling.
It is the Drosophila ovary in which the first piRNA regulator, piwi, was identified for its critical role in maintaining GSCs [21,22]. Although it is expressed in all germ cells and somatic cells of the Drosophila ovary, it has been suggested to function in TF/cap cells for maintaining GSCs [22,23]. In addition, Piwi is also required intrinsically to promote GSC division and primordial germ cell formation [24,25]. In Drosophila ovarian somatic cells, Yb works with Piwi to control primary piRNA biogenesis [12,14,26], and is also suggested to work in TF/cap cells to maintain GSC selfrenewal [27]. In addition, recent studies have shown that Armitage (Armi), Vreteno (Vret) and Tdrd12 are also required in somatic cells to control primary piRNA biogenesis [28,29,30]. Inactivation of histone H3K9 trimethylase Eggless function in ECs leads to defective piRNA biogenesis, upregulation of transposons and germ cell differentiation defects, indicating that piRNAs are important for maintaining EC function by repressing transposons [19,31]. Consistently, vret mutants also have a germ cell differentiation defect, which can be rescued by somatic cell-specific expression of vret [28,29]. In this study, we show that piwi is required in ECs and germline to control germ cell differentiation.

Piwi is required in ECs to control germ cell differentiation and EC survival
To identify the genes that are required in ECs for controlling germ cell differentiation, we carried out a genetic screen using an EC-expressing gal4 driver c587 and transgenic UAS-RNAi lines from the Vienna Drosophila RNAi Center (VDRC). These VDRC RNAi transgenic lines were designed based on the production of a long double-stranded RNA structure that can be further processed into small double-stranded RNAs degrading target mRNAs, and have been used to carry out genetic screens in various Drosophila tissue types [32,33,34]. c587 is expressed specifically in ECs and early follicle cell progenitors based on the expression of UAS-GFP [35] (Fig. 1A). In our screen, piwi was identified for its requirement in ECs to control germ cell differentiation as c587-mediated knockdown of piwi (piwiKD) causes the accumulation of spectrosome-containing ill-differentiated single germ cells (SGCs) located distantly from cap cells in the knockdown germaria, which is in great contrast with the control germaria (Fig. 1B). Although GSCs and cystoblasts (CBs, immediate GSC progeny) contain a spherical spectrosome, only GSCs directly contact cap cells. Differentiated 2-cell, 4-cell, 8-cell and 16-cell cysts contain a branched fusome, and can be easily distinguished from GSCs and CBs. The spectrosome and fusome are the same germ cell-specific intracellular organelle with different morphologies, which can be reliably labeled with antibodies against their components such as Hu li-tai shao (Hts) [36] (Fig. 1A and 1B). To verify the RNAi-mediated Piwi knockdown efficiency, we used polyclonal anti-Piwi antibodies to examine Piwi protein expression in the control and piwiKD germaria, in which cap cells and ECs are marked by the PZ1444 enhancer trap line [37]. Two additional microRNA-based UAS-RNAi transgenic strains were also used in this study: one RNAi strain was generated by the Perrimon Laboratory at Harvard Medical School, HMS [38,39], and the other RNAi line, THU, was generated to target a different piwi sequence by the Ni laboratory at Tsinghua University. The PZ1444 enhancer trap line expresses nuclear b-galactosidase protein in cap cells and ECs, which can be reliably distinguished by their location and morphology [37,40] (Fig. 1C). As previously reported [23,41], Piwi is generally expressed in both somatic cells and germ cells of the control germaria, but ECs express higher levels of Piwi than cap cells (Fig. 1C). Indeed, all the RNAi lines can efficiently eliminate Piwi expression in PZ1444-positive ECs and cap cells, while Piwi expression in germ cells including GSCs remains unchanged in the piwiKD germaria ( Fig. 1D-1F). c587-mediated expression of the HMS and THU RNAi lines can also cause the accumulation of illdifferentiated SGCs outside the GSC niche, similar to the VDRC line ( Fig. 1G and 1H). Because a wild-type germarium normally contains one or two CBs, a germarium containing three or more SGCs is considered to exhibit the germ cell differentiation defect [42]. To determine the severity of the germ cell differentiation defects, we classified the piwi knockdown germaria into three categories: normal (SGCs#2), moderate differentiation defect (3#SGCs#10) and severe differentiation defect (SGCs$11). Among the germaria in which piwi is knocked down in ECs by three independent RNAi lines, 40-60% of them have 3 or more SGCs, and approximately 20% of them have 11 or more SGCs (Fig. 1I). In addition, egg chambers are also often filled with undifferentiated spectrosome-containing SGCs ( Fig. 1B and 1H). Quantitatively, the three RNAi lines produce similar degrees of germ cell differentiation defects (Fig. 1I). As shown in Fig. 1I, there are some variations on SGC numbers in the knockdown germaria by different RNAi strains, and Fig. 1B, 1G and 1H reflect the variations among the three RNAi strains. In addition, we have observed that 10-50% of the piwiKD germaria by the three independent RNAi lines completely lose GSCs and become agametic, suggesting that Piwi is required in somatic cells for maintaining GSCs (Fig. 1I-L). These results indicate that Piwi is indeed required in ECs to promote germ cell differentiation and in ECs, cap cells or both to maintain GSCs.
The germarial region of the piwiKD ovaries appears to be reduced in size, suggesting that the EC number may also be reduced as well ( Fig. 1B and 1H). Our previous studies have suggested that the severity of EC loss is positively correlated with the severity of the germ cell differentiation defects [18,19]. We then quantified EC numbers in the control and piwiKD germaria. In contrast with the control germarium containing an average of 35 ECs (Fig. 1M and 1N), the piwiKD germarium contains significantly fewer ECs ( Fig. 1N-P). Because ECs rarely proliferate, the reduction in EC number could be due to apoptosis. To directly test this idea, we used TUNEL labeling to detect dying ECs identified by PZ1444 expression. Indeed, there is a consistent increase in apoptotic piwiKD ECs by the three independent RNAi lines (Fig. S1). These results demonstrate that Piwi is also required for maintaining EC survival.

Piwi functions in adult ECs to promote germ cell differentiation
Since c587 is known to be expressed by most, if not all, somatic precursor cells in the female gonad [43], the differentiation defects and the GSC loss phenotype caused by piwi knockdown could be due to its early requirement in somatic gonadal precursors. To definitively determine if Piwi is required in adult ECs to control germ cell differentiation, we carried out temperature shift experiments to inactivate Piwi function specifically in adult ECs. When the genetic crosses were carried out at 18uC, which lowers piwi RNAi expression and thus its knockdown effect, the germaria show almost normal germ cell differentiation and GSC maintenance because all the piwiKD germaria have normal SGC numbers and still retain two or three GSCs ( Fig. 2A-D). After the adult females emerged at 18uC, they were cultured at 29uC for a week to increase RNAi expression and piwi knockdown efficiency and thus inactivate Piwi function in adult ECs. Interestingly, the number of the germaria carrying three or more SGCs drastically increases, indicating that Piwi is indeed required in adult ECs to promote germ cell differentiation (Fig. 2E-H). Similarly, the numbers of the germaria containing no GSCs also increase following Piwi knockdown by the three RNAi lines (Fig. 2H-K). In addition, Piwi expression is still reduced in cap cells, suggesting that c587 is likely expressed at low levels in cap cells (Fig. S2). Since the previous findings have shown that ECs also contribute to GSC maintenance [18,19,44], the GSC loss phenotype could come either from Piwi knockdown in cap cells, ECs or both. Taken together, these results indicate that Piwi is required in adult ECs to promote germ cell differentiation as well as in adult ECs, adult cap cells or both for GSC maintenance.

Piwi is required in ECs to prevent BMP signaling in differentiated germ cells
Previous studies have shown that the germ cell differentiation defects caused by defective EC function result from elevated BMP signaling [17,18,19]. To determine if BMP signaling activity is augmented in the germ cells of the piwiKD germaria, we examined the expression of pMad, Dad-lacZ and bam-GFP, three BMP signaling activity reporters in Drosophila, in the control and piwiKD germaria. Activation of BMP receptors (Tkv and Sax) upon BMP ligand binding leads to production of phosphorylated Mad (pMad), which translocates into the nucleus with Medea, a SMAD4 homolog, to activate Dad expression and repress bam expression in GSCs [35,45,46]. In contrast with the control germarium in which pMad accumulates primarily in GSCs (Fig. 3A), pMad is also expressed in some, but not all, SGCs outside the GSC niche of the piwiKD germaria, indicating that BMP signaling activity indeed spreads outside the GSC niche ( Fig. 3B, 3C and S3A). bam-GFP and Dad-lacZ can recapitulate the expression patterns of bam and Dad in the control germarium: bam-GFP is normally expressed in differentiated germ cells but is absent from GSCs [47], while Dad-lacZ is normally expressed in GSCs but is absent in differentiated germ cells [35,45,46]  Previous studies have revealed that the elevated transcription of dpp, which encodes a BMP ligand, in ECs can contribute to increased BMP signaling in differentiated germ cells [18,19,20]. In Drosophila, another BMP-encoding gene, gbb, is also expressed in the germarium and is required for maintaining GSCs [35]. In addition, dally upregulation in ECs has also been shown to be responsible for BMP signaling activity elevation [17,48,49]. To determine if the elevated BMP signaling activity in SGCs is due to upregulation of dpp, gbb or dally in ECs, we sequenced the mRNAs isolated from the purified GFP-labeled control and piwiKD ECs. Based on RNA sequencing and qRT-PCR results, dpp is significantly upregulated in the piwiKD ECs compared to the control ECs ( Fig. 3H and 3I). Although RNA sequencing results show that gbb and dally are slightly upregulated in the piwiKD ECs ( Fig. 3H), qRT-PCR results fail to confirm the finding (Fig. 3I). These results suggest that dpp upregulation in the piwiKD ECs might be responsible for germ cell differentiation defects.
To determine if dpp upregulation in the piwiKD ECs contributes to germ cell differentiation defects, we quantified SGCs outside the GSC niche in the germaria in which piwi and dpp are simultaneously knocked down in ECs. Here we used two different piwi (VDRC and HMS) and dpp RNAi (TRP and HMS) lines to knockdown piwi and dpp expression in ECs, respectively. Based on the numbers of the germaria carrying 3 or more SGCs, c587driven dpp knockdown (TRP) can partially rescue the germ cell differentiation defects caused by piwi knockdown (VDRC) (Fig. 3J-L). c587-driven piwi knockdown by the HMS line yields stronger germ cell differentiation defects, which can be slightly and moderately repressed by c587-driven expression of TRP and HMS dpp RNAi lines, respectively ( Fig. 3L and Fig. S4). c587driven expression of the dpp HMS line causes partial GSC loss, but the expression of the dpp TRP line does not, suggesting that the HMS line might be stronger than the TRP line in knocking down dpp expression (Fig. S4). Based on the finding that the germ cell differentiation defects caused by piwi knockdown can only be partially repressed by c587-mediated dpp knockdown, the germ cell differentiation defects cannot be solely attributed to upregulated dpp expression in ECs (Fig. S4). Taken together, we propose that dpp upregulation in piwiKD ECs contributes to, but is not one of the major causing factors, for germ cell differentiation defects.
Defective EGFR-MAPK signaling in ECs causes germ cell differentiation defects by upregulating dally expression and thus increasing BMP signaling, and also prevents the formation of long cellular processes [17,50]. Although our results show that Piwi knockdown does not lead to dally upregulation ( Fig. 3H and 3I), we wanted to confirm if Piwi is required to maintain EGFR-MAPK signaling in ECs by examining the expression of pERK, a phosphorylated and active form of MAPK, in the piwiKD ECs. In the control, pERK is strongly and specifically expressed in all ECs, but not in cap cells and follicle cells (Fig. S5A). pERK is expressed at low levels in the remaining piwiKD ECs (Fig. S5A-E). Although pERK immunofluorescence intensity in the piwiKD ECs decreases by 25-65% in comparison with the control ECs (Fig. S5E), overall pERK levels might increase instead because the piwiKD ECs are often larger (Fig. S5A-D). To determine if increasing MAPK activity affects GSC maintenance and differentiation, we used c587 to drive the expression of a kinase-active rolled (rl, encoding MAPK in Drosophila) mutant, rl SEM , in ECs [51]. Increasing MAPK activity does not have any obvious effect on GSC maintenance and CB differentiation (Fig. S5). Interestingly, following c587mediated piwi knockdown by the three RNAi lines, ECs lose their long cellular processes (Fig. S6). These results align well with our earlier finding of no expression changes for dally in the piwiKD ECs, and also suggest that EGFR signaling is not the only pathway for maintaining EC cellular processes.
Piwi is required in ECs for repressing TE activity and preventing DNA damage piRNAs have been shown to be required for silencing TE activity in both germ cells and somatic cells [6,7,9,52]. One of the outcomes for elevated TE activity is DNA damage. Thus, we examined the expression of phosphorylated H2Av (c-H2Av), a Drosophila equivalent of mammalian H2AX [53], in the control and piwiKD ECs, and quantified c-H2Av-positive ECs. c-H2Av has been shown to be associated with DNA double-strand breaks in Drosophila cells [53]. In the control germaria, less than 5% of the ECs are positive for c-H2Av (Fig. 4A and 4C). In contrast, 8%-25% of the piwiKD ECs are positive for c-H2Av depending on the RNAi lines ( Fig. 4B and 4C). These results indicate that Piwi is required in ECs to prevent DNA damage.
To further determine if Piwi is required in ECs for silencing TE activity, we sequenced the RNAs from the purified GFP-labeled control and piwiKD ECs by fluorescence-activated cell sorting (FACS). In this study, we chose to examine two common somatic cell-specific TEs, gypsy and zam, and a germline-specific TE, tart [54]. Both gypsy and zam transcripts are drastically and significantly upregulated in the piwiKD germaria in comparison with the control (Fig. 4D and 4E). As expected, the germline-specific tart transcripts are not changed dramatically in the piwiKD germaria in comparison with the control (Fig. 4F). In addition, we also used the gypsy-lacZ reporter to verify the qRT-PCR results. In the control germaria, gypsy-lacZ is not expressed (Fig. 4G). In contrast, it is dramatically upregulated in the piwiKD ECs by the three RNAi lines (Fig. 4H-J). These results further support the idea that Piwi is required in ECs to repress TE activity and prevent DNA damage.
Yb has been shown to regulate Piwi expression in TF and cap cells [27]. Indeed, in the c587-mediated YbKD germaria, Piwi protein expression in ECs and follicle cells is consistently downregulated (Fig. S7A-C9). However, Yb protein expression in somatic cells, including cap cells, ECs and early follicle cells, remains unchanged in the c587-mediated piwiKD germaria (Fig.  S7D-G). To further determine if Yb is also required in ECs to repress TE activity, we examined the expression of gypsy-lacZ in the YbKD germarium. As previously reported, Yb is also expressed in all ovarian somatic cells, including ECs (Fig. 4K). c587-driven expression of two independent Yb RNAi lines can efficiently eliminate Yb expression in cap cells, ECs and early follicle cells ( Fig. 4L and 4M). Interestingly, gypsy-lacZ expression is upregulated in the YbKD cap cells and ECs, indicating that Yb is also required in somatic cells to silence TEs (Fig. 4N and 4O). However, gypsy-lacZ expression appears to be lower in the YbKD ECs than in the piwiKD ECs ( Fig. 4H-J, 4N and 4O). Although most of the YbKD germaria contain normal numbers of GSCs and SGCs ( Fig. 4N and 4O), approximately 25% of the YbKD germaria carry 3 or more SGCs (Fig. S7H-K). These results suggest that Yb is also required in ECs to repress TEs and promote germ cell differentiation.

Piwi is required intrinsically to maintain germ cells before adulthood
Since Piwi is expressed in all the germ cells, including GSCs, we then used nanos-gal4VP16 (nos-gal4) to specifically knock down piwi in germ cells to determine if Piwi is also required intrinsically for GSC maintenance. The nos-gal4 driver is expressed specifically in germ cells from primordial germ cells (PGCs) to adult germ cells, including GSCs [55]. In contrast with the control third-instar female gonad (Fig. 5A), nos-gal4 driven expression of the two independent piwi RNAi strains, HMS and THU, leads to a reduction in PGC numbers in the female gonads, indicating that Piwi is required for PGC proliferation, maintenance or both (Fig. 5B-D). Furthermore, germ cell-specific piwi knockdown germaria in newly emerged adults show a complete loss of all germ cells, including GSCs ( Fig. 5E and 5F). The GSC establishment takes place at the transitional period from the third instar-larval stage to the pupal stage. These results indicate that Piwi is required intrinsically to control PGC maintenance and/or GSC establishment.
To further explore whether other piRNA components are also required for GSC maintenance before adulthood, we used nos-gal4 driven expression of RNAi against armi and aub to inactivate their function throughout germ cell development. Germ cell-specific armi or aub knockdown by two independent RNAi lines for each gene leads to a dramatic reduction in nuclear Piwi protein expression in germ cells, but does not affect nuclear Piwi expression in somatic cells (Fig. 5G-I). In addition, germlinespecific armi or aub knockdown also causes the full penetrance of female sterility. These results suggest that both of them are efficiently knocked down in the germline because they are required for Piwi nuclear localization and to prevent the activation of meiotic cell cycle checkpoints caused by transposon-induced DNA damage [56]. However, the armi or aub knockdown germaria from newly eclosed females still contain two or three GSCs in their germaria, indicating that they are not required intrinsically for early germ cell development and GSC formation (Fig. 5G-I). Consistently, newly eclosed armi 72.1 /armi 1 and aub HN2 /aub QC42 mutant females also maintain two or three GSCs in their germaria, and dramatically decrease nuclear Piwi expression in GSCs and their progeny (Fig. 5J-L). Since these mutants carry strong loss-offunction mutations in armi or aub [56,57], these results further support that Armi and Aub are dispensable for germ cell development before adulthood. Taken together, our results indicate that Piwi controls early germ cell development, GSC formation or both independently of Armi and Aub. Piwi is required intrinsically to control GSC maintenance and germ cell differentiation in the adult ovary To determine if Piwi is required in the adult germline to maintain GSCs, we used the flip-out system, in which a transcriptional stop sequence flanked by two FRT sites is inserted between the nanos promoter and gal4VP16, to activate the expression of RNAi lines along with the GFP reporter specifically in germ cells after heatshock treatments of adult females (Fig. 6A). In the control ovaries, GFP-positive GSCs detected 1 day after heatshock (1 d AHS) remain in the niche for additional three weeks ( Fig. 6B and 6C). The GFP-marked piwiKD GSCs can be readily detected in the germaria 1 d AHS (Fig. 6D and 6E). In contrast, most of the GFP-marked piwiKD GSCs are lost three weeks AHS, and consequently over 30% of the piwiKD germaria have completely lost GSCs (Fig. 6F-H). In addition, more undifferentiated SGCs also accumulate in the piwiKD germaria three weeks AHS, indicative of germ cell differentiation defects ( Fig. 6I and 6J). Interestingly, some SGCs outside the GSC niche are GFP-negative and also Piwi-negative, which is caused by the failure in nos-gal4-driven GFP expression due to an unknown reason ( Fig. 6I and 6J). These results demonstrate that Piwi is required in adult germline for GSC maintenance and germ cell differentiation.

Discussion
Although the primary piRNA pathway is known to operate in Drosophila ovarian somatic cells to repress TE activity, its biological importance in Drosophila oogenesis is not well understood. Piwi, one of the key components in the primary piRNA pathway, has been shown to function in TF/cap cells to control GSC maintenance [22,23,24]. In this study, we have revealed a novel role of Piwi in ECs to control GSC lineage differentiation and additional roles in the germline for PGC and GSC maintenance (Fig. 6K). piwi knockdown in somatic cells results in defective GSC maintenance, defective germ cell differentiation as well as increased TE activity and DNA damage. dpp upregulation contributes to the germ cell differentiation defects caused by somatic piwi knockdown. A recent study has also come to a similar conclusion that Piwi is required in somatic cells for promoting germ cell differentiation by repressing BMP signaling [58]. In addition, we have also shown that Piwi is required in PGCs to control PGC maintenance, GSC formation or both, and is also required in adult germline to maintain GSCs and promote germ cell differentiation (Fig. 6K). Therefore, our genetic results argue strongly that Piwi functions in germline to maintain PGCs and GSCs as well as to promote germ cell differentiation. Therefore, we have revealed new functions of Piwi in multiple cell types to maintain GSCs and promote germ cell differentiation (Fig. 6K).

Piwi is required in ECs and germ cells to promote germ cell differentiation
Recent studies have shown that ECs play an important role in promoting germ cell differentiation by repressing BMP signaling [16]. Thus far, genes identified to be important in ECs for germ cell differentiation repress the expression of either dally or dpp, thereby preventing BMP signaling in ECs. EGFR signaling has been proposed to be responsible for directly repressing dally expression in ECs, but is dispensable for EC survival [17]. In addition, recent studies have also shown that Rho signaling and Eggless are also required in ECs for repression of dally expression, and are also required for EC survival and the maintenance of long EC cellular processes [18,19]. Lsd1, Rho signaling and Eggless have been shown to be required to repress dpp transcription [18,19,20]. In this study, we have also shown that Piwi is required in ECs for dpp repression but is dispensable for dally repression. In addition, it is also required in ECs for maintaining their survival and long cellular processes. Our genetic results suggest that dpp upregulation partially contributes to the germ cell differentiation defects caused by Piwi knockdown in ECs. In contrast with the recent study claiming that dpp upregulation is responsible for the germ cell differentiation defects caused by defective Piwi in ECs [58], our study shows that dpp upregulation is not the main cause for the germ cell differentiation defects caused by defective ECs. In our study, we used two independent RNAi lines against piwi and dpp, respectively, to show that dpp knockdown does not drastically rescue the germ cell differentiation defects caused by piwi knockdown ECs. Furthermore, the dpp levels remain low in piwi knockdown ECs even after upregulation based on RNA sequencing results. Thus, we conclude that dpp upregulation in piwi knockdown ECs is only partially responsible for the germ cell differentiation defects. In addition, Piwi is required in somatic ovarian cells to repress TE activity and prevent transposoninduced DNA damage. However, it remains unclear if dpp upregulation and the loss of ECs and their long cellular processes are caused by DNA damage, and how Piwi is involved in repressing BMP signaling activity in differentiated germ cells via repression of dpp expression in ECs.
Piwi has previously been demonstrated to be required intrinsically for promoting GSC division [27]. Piwi is expressed in GSCs and their differentiated progeny [27]. This study has shown that germline-specific knockdown of Piwi function in the adult ovary leads to the accumulation of undifferentiated single germ cells, revealing a new intrinsic role of Piwi in controlling germ cell differentiation (Fig. 6K). Piwi has been shown to be involved in the piRNA pathway and epigenetic regulation. In the future, it will be important to determine if the piRNA pathway, epigenetics or both play a role in the regulation of germ cell differentiation.

Piwi is required in both somatic cells and germ cells to maintain GSC lineage
Although Piwi is generally expressed in almost all somatic cells and germ cells of the Drosophila ovary, the previous studies proposed that Piwi acts in TF/cap cells to control GSC selfrenewal [22,23,24]. In this study, we have confirmed the somatic role of Piwi in GSC maintenance, and have also revealed new roles of Piwi to maintain PGCs before adulthood and GSCs after adulthood. In addition, our temperature shift experiments have shown that Piwi is also required in adult somatic cells, TF/cap cells, ECs or both, to maintain GSCs. Interestingly, RNAimediated knockdown of Piwi function in adult somatic cells only produces a moderate GSC loss phenotype in comparison with the severe GSC loss phenotype of piwi mutants, suggesting that Piwi might also function in other cell types to maintain GSCs. In the future, it will be important to determine how Piwi functions in TF/ cap cells and ECs to maintain GSC self-renewal. Therefore, our study has not only confirmed the somatic role of Piwi in controlling GSC self-renewal but also has suggested its function in additional cell types to maintain GSCs (Fig. 6K).
Although Piwi has been shown to be required to control PGC formation and GSC division [25,27], it remains unclear if Piwi is required intrinsically to maintain PGCs and GSCs. In this study, we have revealed critical roles of Piwi in different developmental stages of germ cells. First, Piwi is required in the developing female gonad to control PGC proliferation, survival or both because germline-specific Piwi knockdown leads to a reduction in PGC number in third-instar female larval gonads. Second, Piwi is required in PGCs to control their survival or GSC formation because germline-specific piwi knockdown leads to a complete elimination of germ cells including GSCs in newly eclosed adult females. Interestingly, germline-specific knockdown of either armi or aub, two of which work with piwi to control piRNA biogenesis, fails to produce any GSC loss phenotype in newly eclosed adult females, suggesting that Piwi controls PGC proliferation and survival or GSC formation possibly independently of Armi and Aub, possibly piRNAs. Piwi has been shown to physically interact with HP1a to epigenetically control gene expression in somatic tissues [41]. In addition, piwi genetically interacts with certo, encoding a chromodomain-containing protein, to control GSC maintenance [59]. Our findings are consistent with the notion that Piwi controls early germ cell development perhaps via epigenetics. Third, Piwi is required in adult GSCs for their maintenance because germline-specific knockdown in the adult ovary also causes a moderate GSC loss phenotype. Therefore, we propose that Piwi functions in multiple stages of germline development to control PGC proliferation and survival, and GSC maintenance (Fig. 6K).

Construction of UAS-RNAi and nos.mCherry SV40 polyA.gal4VP16 Strains
The new THU UAS-RNAi line targeting piwi was constructed using the pVALIUM20 vector according to the published procedure [60]. The targeting sequence for piwi is CCCGGTCATGCTGCAGACGAA, which was designed based on the algorithm of DSIR.

Immunohistochemistry
Immunohistochemistry was performed according to our previously published procedures [62,63]. The following antibodies were used in this study: rabbit polyclonal anti-bgalactosidase antibody (1: For qRT-PCR, total RNAs were first treated by DNase I, and were then used for synthesis of cDNAs using mixed oligo dT and random primers and SuperScript III Reverse Transcriptase (Life Technologies). Fluorescence-based quantitative real-time PCR (qPCR) was performed to quantify gypsy, zam, TART, gbb, dpp and dally with tbp, gapdh and rpl32 as internal controls using primers shown in Table S1. After cDNAs were diluted at 1:100, 2 ml aliquots of each cDNA sample were added to 5 ul of 26 power SYBR Green PCR Master Mix (Applied Biosysterms part No.: 4367659, Lot No. :1305403), 0.5 ml each of 10 nm Forward & Reverse primer and 2 ul of water in a 384-well plate. The resulting reactions were run on an ABI 7900HT according to the instruments standard protocol. Analysis of the fluorescence curves was done using ABI's SDS2.4 software. The Ct values were analyzed using the Biogazelle qBase Plus version 2.4 software to generate normalized relative quantities using assays for endogenous controls.