Activin signaling balances proliferation and differentiation of ovarian niche precursors and enables adjustment of niche numbers

Stem cells and their niches constitute functional units that support organ homeostasis and regeneration. How stem cell unit numbers are determined during development remains an open question (O'Brien and Bilder, 2013). In the ovary of Drosophila melanogaster, 16-20 stem cell units form during larval development. These units are composed of somatic terminal filament (TF), escort and cap cells, which attach to two or three germline stem cells (GSCs) (Chen et al., 2011; Eliazer and Buszczak, 2011; Losick et al., 2011; Spradling et al., 2011). Both the somatic and germline components of the GSC unit originate from precursor cells. During the first 4 days after egg laying (AEL; for timing of larval development see Materials and Methods), primordial germ cells (PGCs, the precursors of GSCs) and the somatic precursors proliferate (Fig. 1A). At mid-third larval instar (ML3, 96 h AEL), cells that are positive for TF markers first appear (Gancz et al., 2011; Godt and Laski, 1995; Sahut-Barnola et al., 1995; Zhu and Xie, 2003). During the last day of larval development (96-120 h AEL), TFs gradually form, a process that occurs simultaneously with somatic precursor cell proliferation. Once formed, TFs induce cap cell formation by Notch signaling (Song et al., 2007). PGCs then attach to the cap cells via E-Cadherin (Shotgun – FlyBase) to form the completed stem cell unit (Song et al., 2002).

Previous work identified several transcription factors and signaling pathways that contribute to TF formation and numbers, including Engrailed (En), Bric a brac and Pipsqueak (Bartoletti et al., 2012; Bolivar et al., 2006; Godt and Laski, 1995; Sahut-Barnola et al., 1995). Nutritional cues, such as the Insulin and Tor pathways, also affect TF numbers (Gancz and Gilboa, 2013b; Green and Extavour, 2012,, 2014; Sarikaya et al., 2012).

Ecdysone is another hormonal cue that promotes TF formation (Gancz and Gilboa, 2013a; Gancz et al., 2011; Hodin and Riddiford, 1998; Konig et al., 2011). At early larval stages, ecdysone receptors act as repressors of TF differentiation, whereas in the second part of the third instar they become activators of TF differentiation. Ecdysone receptors thus act as a switch between the proliferation of somatic precursor cells and TF differentiation (Gancz et al., 2011). One known ecdysone target that is upregulated at the latter half of the third instar is the zinc-finger and BTB domain transcription factor Broad (Br) (Chao and Guild, 1986; Karim et al., 1993). The broad locus encodes four isoforms (Bayer et al., 1996; DiBello et al., 1991), and the Br-Z1 isoform is specifically upregulated in ovarian somatic cells following ecdysone activity (Gancz et al., 2011).

Since Br-Z1 plays a central role in GSC unit formation, it might also provide a hub for controlling the proliferation and differentiation processes that create them. Our present data show that such control occurs through Activin signaling. In Drosophila, Activin signaling is transmitted by Baboon (Babo), a type I receptor of the TGF-β superfamily (Massague, 2012; Miyazono et al., 2000). Babo phosphorylates the R-Smad Smox (Smad on X), which, together with the Smad4 homolog Medea, translocates to the nucleus and represses or activates various target genes (Brummel et al., 1999). Activin has been associated with proliferation, differentiation and morphogenetic events occurring in the nervous system, immune system, muscle and during imaginal disc development (Bai et al., 2013; Brummel et al., 1999; Clark et al., 2011; Ellis et al., 2010; Ng, 2008; Parker et al., 2006; Peterson and O'Connor, 2013; Ting et al., 2014; Yu et al., 2013; Zheng et al., 2003; Zhu et al., 2008). Activin signaling also acts systemically to control metamorphosis by promoting the expression of several ecdysone biosynthetic enzymes (Gibbens et al., 2011). Interestingly, Activin interacts with the ecdysone pathway also in the nervous system, where it promotes the expression of the ecdysone receptor EcR-B1 in mushroom body γ-neurons, thus potentiating their response to ecdysone (Yu et al., 2013; Zheng et al., 2003).

Our work now uncovers a third mode of interaction between the ecdysone and Activin pathways. We show that Activin is required in the somatic cells of larval ovaries for the correct accumulation of Br-Z1 following ecdysone receptor activation. In addition, Activin promotes somatic precursor cell proliferation independently of ecdysone. Thus, Activin signaling acts as a coordinator of proliferation and differentiation and balances these two opposing forces.

One of the attributes of TF cells is their inability to self-renew during adulthood (Xie and Spradling, 2000). Indeed, BrdU feeding and mosaic analysis experiments revealed that progeny of proliferating cells could be incorporated into TFs from the end of second larval instar until ∼105 h AEL with decreasing efficiency (Godt and Laski, 1995; Sahut-Barnola et al., 1996). To understand how proliferation and differentiation of somatic precursors may modulate niche numbers, we monitored somatic cell proliferation at different times during larval development. We used anti-β-gal or anti-En antibodies to mark TF cells of hh-lacZ and Oregon Red (OR) wild-type (WT) flies, respectively; Alexa 488-phalloidin delineated cell shape and anti-phospho-Histone 3 (pH3) detected mitotic cells. At the end of third instar (120 h AEL), pH3-positive cells were observed throughout the organ (Fig. 1B, arrowheads), but not in TF cells, which were stacked in filaments (Fig. 1B, inset).

We next examined whether TF cells could proliferate prior to their stacking in a filament. At 96 h AEL, TF stacking has only just begun, and En-positive cells that were not in stacks could be analyzed (Gancz et al., 2011; Godt and Laski, 1995; Sahut-Barnola et al., 1995). None of these cells was pH3 positive (Fig. 1C,C′). Therefore, TF cells cease proliferating prior to their stacking in filaments.

Interestingly, at 96 h AEL, the region where TFs were forming contained three groups of cells that were shaped differently to other anterior cells (Fig. 1C,C′, between white lines; supplementary material Movie 1). The first group consisted of mature, disc-shaped cells expressing the TF marker En (Fig. 1C, chevrons). Cells in the second group also expressed TF markers, but were cuboidal or box-shaped (Fig. 1C′, arrowheads). The majority of cells in this region belonged to the third group, with the same box-shape but without En expression (Fig. 1C′, arrows). We hypothesize that these three groups might represent a sequence of TF cell formation, whereby cells at the TF region first change their shape, then gain TF marker expression, followed by stacking into filaments and a second shape change to mature disc-like cells.

To determine the proliferation status of the different cells in the TF region, we used antibodies directed against mitotic Cyclin B (CycB). In line with gonad growth at 96 h AEL, many gonadal cells expressed CycB. Strikingly, a band in the region where TF cells formed showed a marked reduction in CycB expression (Fig. 1D,E, between white lines). This suggests that cells in the TF region reduce their proliferation or exit the cell cycle entirely prior to acquiring TF markers. To support this conclusion, we also analyzed EdU incorporation, which marks cells undergoing S phase. The TF region mostly lacked EdU-labeled cells. The few EdU-positive cells that were observed in the CycB-reduced region were located at its periphery (Fig. 1E, arrowhead). This strengthens the notion that cells in the TF region rarely divide. Reduction in CycB was already apparent at 68 h AEL, prior to entry into third instar (Fig. 1F). At this stage, neither hh-lacZ nor En-positive cells were observed. Therefore, lack of CycB is currently our earliest marker for TF differentiation. As cells in the TF region have not yet acquired any known TF cell markers, but differ from anterior proliferating precursors, we term them TF precursors.

To find pathways that might affect TF precursor accumulation, we conducted an RNAi screen in larval ovaries using the driver traffic jam (tj)-Gal4, which is expressed in the somatic cells of the ovary but not in germ cells (Li et al., 2003). Four different RNAi constructs directed against the Activin receptor and three that are directed against its effector Smox produced markedly smaller gonads at the end of larval development (Fig. 2A,B; supplementary material Table S1). This size reduction could not be ascribed to cell death (supplementary material Table S2). Conversely, somatic overexpression of the activated Activin receptor (Babo^QD) resulted in a marked increase in gonad size (Fig. 2C; supplementary material Table S1). Similar results were obtained with the somatic driver c587-Gal4 (supplementary material Fig. S1). Activin signaling affected ovary size even prior to TF formation, as both total ovary size and that of the TF precursor region were affected prior to ML3 (Fig. 2D-I; supplementary material Tables S3, S4).

Activin signaling affected cell proliferation, as Smox-RNAi ovaries contained fewer pH3-positive cells, even after correcting for the reduced gonad size (supplementary material Table S3). Interestingly, we could not detect a corresponding increase in cell proliferation upon Babo^QD expression (supplementary material Table S5). A similar effect of Babo on imaginal disc size (a change of 20-30%), without detectable changes in the fraction of dividing cells, was previously reported (Brummel et al., 1999), suggesting that the effects of Activin on the cell cycle in some organs might be evenly distributed among the different cell cycle phases.

To analyze the effect of Activin on cell size, we measured both anterior cells and intermingled cells (ICs) and found that cell size changed little in Smox-RNAi or Babo^QD ovaries (supplementary material Table S6). Cell proliferation and cell growth could be regulated either independently or by the same signals (Johnston and Gallant, 2002; Jorgensen and Tyers, 2004; Potter and Xu, 2001). Had cell growth been completely unaffected by Activin signaling, cells in Smox-RNAi ovaries, which divide less often, would have continued growing and would have been larger. However, cell size changed only slightly upon modulating Activin signaling, suggesting that it promotes both cell proliferation and growth.

In addition to increasing precursor cell proliferation/growth, Activin signaling promoted niche differentiation. At ML3, when TFs begin to form, a significant increase in TF stack numbers was observed in Babo^QD ovaries as compared with WT (Fig. 3A,B,H; supplementary material Table S7). At 101 h AEL, these stacks were not only more numerous, but also longer (Fig. 3C,D,H; supplementary material Table S7). Thus, once TFs start forming, increased Activin signaling accelerates the process. By 120 h AEL, the anterior region of Babo^QD ovaries contained TFs that were unevenly spaced (Fig. 3F, arrowheads, compare with 3E). This phenotype is reminiscent of EcR-RNAi and usp-RNAi ovaries, in which derepression of ecdysone target genes accelerates TF formation and results in uneven spacing of TFs (Gancz et al., 2011). The final TF number in Babo^QD ovaries did not significantly differ from that of control ovaries (Fig. 3H; supplementary material Table S7); by the end of larval development (120 h AEL), the correct, adult numbers of TF stacks had already formed (Fig. 3H; supplementary material Table S7). By contrast, Smox-RNAi and babo-RNAi ovaries had fewer, shorter, TF stacks at 120 h AEL (Fig. 3G,H; supplementary material Table S7). Importantly, unlike WT, TFs in Smox-RNAi and babo-RNAi ovaries continued forming in the pupa, and ovariole numbers in the adult were higher than at 120 h (Fig. 3H; supplementary material Table S7). Thus, some mechanism allows a partial rescue of ovariole numbers following modulation of Activin signaling.

We have previously shown that, in some cases, TF differentiation is co-regulated with PGC differentiation (Gancz and Gilboa, 2013b; Gancz et al., 2011). Since Activin signaling promoted niche development, we tested its effects on PGC differentiation. At late third instar, PGCs that are located away from the niche lose BMP signaling and upregulate the major differentiation gene bag of marbles (bam). This developmental step can be monitored by the bamP-GFP reporter (Chen and McKearin, 2003; Zhu and Xie, 2003) (Fig. 4A). Reduction of somatic Smox by RNAi reduced bamP-GFP expression only marginally (Fig. 4B-D). However, babo reduction by four different RNAi constructs did result in a significant decrease in bamP-GFP levels (Fig. 4E-I). Conversely, Babo^QD expression greatly enhanced PGC differentiation in a non-autonomous manner. As noted previously, at 96 h AEL, PGCs in WT ovaries are still proliferating and do not express bamP-GFP (Fig. 4J) (Gancz et al., 2011). By contrast, Babo^QD-overexpressing ovaries were filled with GFP-expressing PGCs, indicating that PGCs initiate their differentiation precociously (Fig. 4K). pMad labeling in PGCs that are close to niches remained normal in Smox-RNAi, babo-RNAi and Babo^QD ovaries, suggesting that BMP signaling could still maintain the prospective GSCs (supplementary material Fig. S2).

Precocious germline cyst formation was also observed following Babo^QD expression. We monitored cyst formation by the morphology of the fusome, an intracellular organelle within germ cells. Fusomes are round in PGCs, GSCs and their immediate daughter cells but branched in differentiating cysts (de Cuevas and Spradling, 1998). At 120 h AEL, WT bam-expressing PGCs still harbored a spherical fusome, indicating that cyst development has not yet occurred (Fig. 4L,M, arrowheads). By contrast, branched fusomes were prevalent in Babo^QD ovaries, showing that cyst development was underway (Fig. 4N,O, arrowheads). Combined, these data suggest that Activin signaling promotes both niche and PGC differentiation.

Co-development of ovarian niches and PGCs is controlled by ecdysone signaling (Gancz et al., 2011). We therefore examined a possible association between the ovarian Activin and ecdysone pathways. In larval brains, Activin signaling has been shown to control the response to ecdysone by promoting the expression of the ecdysone receptor EcR-B1 (Yu et al., 2013; Zheng et al., 2003). However, whereas EcR-B1 expression in γ-neurons is developmentally regulated, ovarian EcR is continually expressed throughout third instar (Gancz et al., 2011). Indeed, staining with an antibody directed against EcR-B1 did not reveal a significant change in the expression of this isoform when either one of the three Smox-RNAi or four babo-RNAi constructs was expressed in the somatic cells of larval ovaries (Fig. 5A-H). Likewise, anti-EcR-A and anti EcR-C staining – directed against the A isoform and the common region of EcR, respectively – was not significantly changed (supplementary material Figs S3 and S4). Thus, ecdysone receptor expression in larval ovaries does not positively correlate with the status of Activin signaling.

To further explore a possible connection between the ecdysone and Activin pathways, we measured the effects of Activin on the transcription of various ecdysone target genes. ftz-f1 and Eip74EF expression changed only a little (Fig. 5I). The Eip75B mRNA level was significantly increased in Babo^QD ovaries (Fig. 5I). However, only one Smox-RNAi construct resulted in a significant reduction of Eip75B mRNA levels (Fig. 5I). Thus, Eip75B might be somewhat affected by Activin signaling.

Our previous studies showed that the ecdysone target broad promotes niche and PGC differentiation and that the Br-Z1 isoform is specifically upregulated by ecdysone at the appropriate time (Gancz et al., 2011). Indeed, a consistent effect of Activin on br-Z1 transcription was observed in larval ovaries; expression of Babo^QD increased br-Z1 levels, whereas all three Smox-RNAi constructs reduced its levels significantly (Fig. 5J). By contrast, br-Z2 did not respond strongly to changes in Activin signaling (Fig. 5K). br-Z4 expression was significantly elevated by Babo^QD expression, but was only slightly affected by Smox-RNAi, suggesting that it might not constitute a major target of Activin in the gonad.

The specific effects of Activin on only some ecdysone targets indicate that, in the ovary, Activin does not interact with the ecdysone pathway at the receptor level, but at the level of specific target genes. The consistent effect that we observed on Br-Z1, which is the same isoform that is upregulated by ecdysone, suggests that at least part of the effects of Activin on the ovary occur by controlling Br-Z1.

To determine the effect of Activin signaling on Br-Z1 protein levels, we stained late third instar ovaries with anti-Br-Z1 antibody. At 120 h AEL, somatic nuclei of WT ovaries expressed high Br-Z1 levels (Fig. 6A,A′). Br-Z1 levels were even higher at 125 h AEL, in line with the high ecdysone levels at pupariation (Fig. 6B). By contrast, Br-Z1 levels were significantly lower in Smox-RNAi ovaries (Fig. 6C,C′). The reduction in Br-Z1 levels was stronger in ICs, which express Tj, and where the driver tj-Gal4 is strongly expressed. At the anterior of the gonad, where Tj and tj-Gal4 are expressed weakly, some Br-Z1 expression was observed (Fig. 6C′). Br-Z1 reduction in Smox-RNAi ovaries was less pronounced at pupariation, when high levels of ecdysone are present (compare Fig. 6D with 6B). Similar kinetics and localization of Br-Z1 expression were also observed in babo-RNAi ovaries (Fig. 6E-F).

To better quantify the modulation of Br-Z1 expression by Activin signaling, and to determine whether it acts in an autonomous or non-autonomous manner, we used mosaic analysis of the strong alleles Smox^MB388 and babo³². Br-Z1 levels in mutant and neighboring WT cells were compared at specific time points during the last 24 h of larval development and at pupariation (Fig. 6G,H). For each time point, the percentage of clones showing 0-25%, 25-50%, 50-75% or >75% of WT Br-Z1 levels was calculated. The results show progressive Br-Z1 accumulation in both WT and mutant nuclei. However, Br-Z1 accumulated at a slower rate in mutant nuclei. These results are consistent with a role for Activin in enhancing Br-Z1 expression during the latter half of third instar.

To further test the idea of Br-Z1 modulation by Activin signaling, we analyzed its accumulation in ovaries expressing Babo^QD. At 91 h AEL, Br-Z1 was not observed in WT or Babo^QD ovaries (Fig. 7A,B). Thus, activation of the pathway could not induce Br-Z1 expression prior to the timing specified by ecdysone signaling. At 96 h AEL, Br-Z1 is only beginning to be expressed in WT ovaries. However, its expression was clearly observed in Babo^QD ovaries (Fig. 7C,D, arrowheads). Five hours later, at 101 h AEL, the elevation of Br-Z1 protein in Babo^QD nuclei was still significant when compared with wild type (Fig. 7E,F). However, by 120 h AEL, when Br-Z1 was already strongly induced in all somatic WT nuclei, little difference was observed between control and Babo^QD ovaries (Fig. 7G,H).

Combined, these results suggest that, as long as ecdysone receptors repress Br-Z1, Activin signaling cannot induce its expression. However, once Br-Z1 expression is induced by the hormone, the rate of its accumulation depends on Activin signaling.

If indeed Activin signaling cannot induce Br-Z1 expression independently of ecdysone signaling, then expression of the repressive dominant-negative EcRA.W650A form (EcR^DN) should be epistatic to Babo^QD expression in the ovary. To test this, we co-expressed EcR^DN with either control lacZ or with Babo^QD. As expected of EcR^DN ovaries, no Br-Z1 was observed in Tj-expressing cells, even at 120 h AEL (Gancz et al., 2011). By contrast, cells at the anterior, which express low levels of tj-Gal4, did express Br-Z1 (Fig. 8A,A′). Co-expression of Babo^QD with EcR^DN did not result in Br-Z1 expression (Fig. 8B,B′).

Consistent with its inability to induce Br-Z1, Babo^QD could not rescue PGC differentiation defects in EcR^DN ovaries. Whereas WT ovaries at prepupal stages exhibit germline cysts, indicating normal germ cell differentiation (Gancz and Gilboa, 2013b), no cysts were observed in either control EcR^DN or the experimental EcR^DN; Babo^QD ovaries. In both cases only spherical fusomes were observed (Fig. 8C,D). Thus, activation of Activin signaling could not rescue Br-Z1 expression or the differentiation processes that associate with it.

Although Babo^QD could not ameliorate the EcR^DN-induced block in differentiation, change was noted in gonad size. EcR^DN; Babo^QD ovaries were distinctly larger than lacZ; EcR^DN ovaries (compare Fig. 8A with 8B, and 8C with 8D,E; supplementary material Table S8). This suggests that Activin promotes two separate activities: cell proliferation and cell differentiation. Cell differentiation is executed, at least in part, by controlling Br-Z1 accumulation downstream of ecdysone signaling. However, control of cell proliferation is independent of ecdysone signals and could be executed even in EcR^DN-expressing ovaries.

Consistent with this interpretation, removal of broad resulted in smaller ovaries (compare Fig. 8F with 8G) and additional removal of babo exacerbated the phenotype. Importantly, while little germ cell differentiation was observed in babo-RNAi ovaries (Fig. 4E-H), overexpression of Br-Z1 from a UAS promoter was sufficient to induce strong bam expression, as determined by bam-GFP levels (Fig. 8I). These results further strengthen the notion of Activin signaling having an ecdysone-dependent role in niche and PGC differentiation and an independent role in niche precursor proliferation.

We show that, during Drosophila ovarian formation, Activin signaling determines precursor cell proliferation in an ecdysone-independent manner and TF and PGC differentiation through modulating Br-Z1 accumulation. This coordination of the rates of cell proliferation and differentiation forms the basis of a correction mechanism for niche numbers.

We show that cells with reduced proliferation capacity already appear at 68 h AEL, prior to entry into third instar (Fig. 1F). The exact time at which TF precursors first arise is unclear, since at earlier time points (48 h AEL) the small gonad size did not allow us to distinguish between a random group of cells that happen to be in G1/S phase and the clear linear structure of CycB-negative cells observed at 68 h AEL. The identification of a positive marker for TF precursors would help to resolve this issue.

Our observations also indicate that between 68 h and 96 h AEL the region of cuboidal cells that are devoid of CycB increases significantly. This corroborates a previous pulse-chase study that indicated a first wave of proliferation at 68-72 h AEL, in which BrdU-retaining somatic cells contributed ∼50% of all TF cells throughout the larval ovary. A second wave of proliferation during third instar occurred gradually and resulted in labeled TF cells that appeared from the medial to the lateral side of the ovary (Sahut-Barnola et al., 1996), matching the topology of mature filament appearance (Gancz et al., 2011; Godt and Laski, 1995; Hodin and Riddiford, 1998; Sahut-Barnola et al., 1996,, 1995).

Combined, these studies suggest a sequence of events in which somatic precursors continually proliferate, giving rise to non- or rarely dividing box-shaped TF precursors as early as second instar. TF precursors continue to accumulate through early third instar. During the latter half of third instar, TF precursors express TF markers such as En and hh-lacZ, organize into filaments and attain the mature disc shape of terminally differentiated TF cells (Fig. 8J). We term the proliferating cells somatic precursors, rather than TF precursors, since they give rise to more than the TF lineage (Sahut-Barnola et al., 1996). We reserve the term TF precursors for the non-proliferating cells that give rise only to mature TF cells. Further understanding of the lineage awaits the identification of additional markers and TF determinants.

The sequence of events leading to TF differentiation suggests that niche numbers could be determined by the rate of somatic precursor cell proliferation, which affects the pool size from which TF precursors are drawn. Insulin signaling, which greatly increases ovarian cell numbers, acts at this level (Gancz and Gilboa, 2013b). A second factor is the timing of TF precursor cell differentiation. Once a cell becomes a TF precursor, it effectively reduces the pool of cells that could proliferate to give rise to more TF cells in the future. We have previously shown that Br-Z1 overexpression and precocious expression results in smaller gonads with fewer TFs (Gancz et al., 2011). This fits a model in which the proliferating pool size is decreased by precocious differentiation and removal of cells from the mitotic to the non-mitotic pool.

We now show that the rate of Br-Z1 accumulation in somatic nuclei can be modulated by Activin signaling (Figs 6 and 7). Three lines of evidence suggest that the differentiation function of Activin depends on modulating the ecdysone response. First, we show that Br-Z1, which is a major target of the ecdysone pathway, is also a target of the Activin pathway. Second, only after induction of Br-Z1 by ecdysone can Babo^QD increase Br-Z1 accumulation. Third, we show that ecdysone is epistatic to Activin function in promoting Br-Z1 expression, as well as niche and PGC differentiation (Fig. 8).

Importantly, Activin signaling also controls ovarian cell proliferation. Our epistasis analysis suggests that, unlike its role in niche differentiation, Activin promotes proliferation in an ecdysone-independent fashion. This places Activin signaling as an important balancer of proliferation and differentiation. Upon activation of the pathway, increased somatic precursor cell proliferation enlarges the pool of TF precursors. However, the rate of TF cell differentiation is also increased. This removes more cells from the effective proliferating population, resulting in normal niche numbers (Fig. 8K). Conversely, when Activin signaling is reduced, cell proliferation diminishes and the precursor pool is smaller. To balance this, the rate of TF differentiation is also reduced. This allows the slowly proliferating precursors additional time to increase their numbers prior to TF differentiation. Collectively, ecdysone signaling, which controls the major switch from proliferation to differentiation, and Activin signaling, which fine-tunes these processes, demonstrate how co-regulation of cell proliferation and differentiation may be used to adjust niche numbers during development.

We have previously shown that Broad is an essential target for the ovarian ecdysone response, which includes niche and PGC differentiation (Gancz et al., 2011). However, other ecdysone targets might also participate in niche/PGC differentiation. We now show that both Br-Z1 accumulation and TF formation are slower in Smox-RNAi and babo-RNAi ovaries. Interestingly, bamP-GFP, which is indicative of PGC differentiation, accumulates almost normally in Smox-RNAi but not in babo-RNAi ovaries (Fig. 4). Since Br-Z1 levels are reduced in both Smox-RNAi and babo-RNAi ovaries, an exciting possibility is that additional ecdysone targets are involved in the somatic control of PGC differentiation. What these targets might be is currently unknown. Moreover, the fact that PGCs fail to differentiate when the Activin receptors are reduced but not when the downstream effector is knocked down suggests that at least part of the function of Activin in the ovary might be Smox independent (Derynck and Zhang, 2003; Moustakas and Heldin, 2005; Ng, 2008; Ozdamar et al., 2005). If so, how such targets induce PGC differentiation remains to be investigated. Our analysis of Activin ligands suggests that activation of the pathway is mostly, if not entirely, dependent on somatically produced ligands (supplementary material Table S1). The Activin pathway therefore provides another example of how the ovarian soma controls germline differentiation.

The Activin and ecdysone pathways have previously been shown to associate in the nervous system and in the prothoracic gland (Gibbens et al., 2011; Zheng et al., 2003,, 2006). Interestingly, in each case, the Activin pathway serves to potentiate the function of ecdysone, and in each case Activin affects the ecdysone response at a different level. In the prothoracic gland, Activin is required for the expression of receptors for Prothoracicotropic hormone and Insulin, which control the expression of several ecdysone biosynthetic enzymes. It is thus placed upstream of the ecdysone response (Gibbens et al., 2011). In mushroom body γ neurons, Activin potentiates axonal pruning by promoting the expression of EcR-B1 (Zheng et al., 2003). A similar molecular function for Activin was identified in dorsal cluster neurons, where it was suggested that Activin is required to modulate the rate of neuronal terminal differentiation (Zheng et al., 2006).

In the ovary, Activin interacts with the ecdysone pathway at the level of the target gene br-Z1. Unlike in the nervous system, the ecdysone receptor is continually expressed in larval ovaries, since it fulfills the dual role of an early repressor and a late activator of stem cell unit differentiation (Gancz et al., 2011). This explains why the phenotypes of Babo^QD overexpression in larval ovaries are similar to EcR-RNAi phenotypes (Gancz et al., 2011); in both cases, br-Z1 is upregulated. Since ecdysone receptors are required as early repressors and late activators of Br-Z1 expression and GSC unit differentiation, there is more logic in placing a potentiator of the pathway at the level of the target gene and not the receptor. How the two pathways were connected at different nodes in different tissues during evolution remains an open question.

The following stocks were from the Bloomington Stock Center: Oregon Red (OR), UAS-EcRA.W650A (EcR^DN), FRT19A, arm-lacZ, Smox^MB388, babo³², and RNAi lines directed against Smox (HMS02203, JF02320). From the Vienna Drosophila RNAi Center (VDRC): RNAi lines against babo (KK108186, GD51, GD2547). From NIF-Fly: RNAi line against babo (8224R3) and against Smox (2262R2). From the Drosophila Genetic Resource Center: tj-Gal4 (P{GawB}NP1624). bamP-GFP located on the X chromosome was obtained from Dr Dennis McKearin (HHMI). UAS-lacZ was provided by Dr Jessica Treisman (NYU School of Medicine, USA). UAS-Babo^QD was from Dr Theodor Haerry. c587-Gal4 was from Prof. Ting Xie (Stowers Institute, USA). Smox^MB388 clones were generated using the line hs-Flp¹²², FRT19A, arm-lacZ and induced by heat shock at 48 h AEL for 35 min at 37°C. babo³² clones were generated using the line c587-Gal4, UAS-flp; FRT42D.

To obtain flies at similar developmental stages, care was taken to work with undercrowded cultures. Flies were transferred into fresh bottles (rather than vials) to lay eggs for 2 hours, and were then removed. Bottles were left at 25°C for 68-125 h; 68 h (late second instar), 96 h (mid-larval third instar), 120 h (late larval third instar) and 125 h (prepupa). Under these conditions the development of WT gonads is uniform. The terminology we use is according to Ashburner et al. (2005).

The following monoclonal antibodies were obtained from the Developmental Studies Hybridoma Bank, developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biology: anti-Hts (1B1, developed by Dr Howard Lipshitz), 1:20; anti-Broad-Z1 (Z1.3C11.OA1, developed by Dr Greg Guild), 1:10; anti-Engrailed (4D9, developed by Dr Corey Goodman), 1:20; anti-EcRA (15G1a; 1:10), anti-EcRB1 (AD4.4; 1:10) and anti-EcRC (AG10.2; 1:10) developed by Drs Carl Thummel and David Hogness; anti-CycB (F2F4, developed by Dr O'Farrell), 1:15. Rabbit anti-Vasa (1:5000) was a gift from Dr Ruth Lehmann (HHMI, New York University). Guinea pig anti-Tj (1:7000) was a gift from Dr Dorothea Godt (University of Toronto). Rabbit anti-β-galactosidase (β-gal) (1:15,000) was from Cappel (08559762). Rabbit anti-GFP (1:1000) was from Invitrogen (A11122). Rabbit anti-phospho-Histone H3 (1:1000) was from Millipore (06-570). Alexa 488-phalloidin was from Life Technologies. Secondary antibodies were from Jackson ImmunoResearch or from Invitrogen and used according to the manufacturer's specifications. The staining protocol was as previously described (Gancz et al., 2011).

EdU (Invitrogen, C10337) was used to label cells in S phase, propidium iodide (Sigma, P4864) to label dead cells and DAPI to label nuclei of adult ovaries, all according to the manufacturer's specifications.

Confocal imaging was with a Zeiss LSM 710 on a Zeiss Observer Z1. For analysis of bamP-GFP, GFP-positive PGCs located at the mid-section of an ovary were marked and the Measure tool in ImageJ (NIH) was applied. For Broad-Z1 staining intensity in Smox^MB388 and babo³² clones, the middle sections of a mutant and an adjacent WT cell were chosen and the Measure tool used for the selected cell area. For cell size and the size of the TF precursor region, the Measure tool was also used. For ovary volume, the Surface tool in Imaris software (Bitplane) was used.

15-20 ovaries were collected from early L3 larvae (112 h AEL). Tissue was disrupted using QIAshredder (Qiagen) and RNA isolated using the RNeasy Kit (Qiagen) according to the manufacturer's instructions. Reverse transcription was performed with the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Q-PCR employed SYBR Green (Invitrogen) with the following primers (forward and reverse): CAAGATTGCCGGCTATGTCA and CCTGCAACTTGATGGAGATA-CCA for RpS17; ACAATCCGCACCACAGAAAC and GCCGGAACTG-GAGCTGTC for Eip74EF; CTCCTACTCCATGCCCACAC and GCTGC-GAGAACTCCTGCT for Eip75B; AGTAGACTGGGCACGGAACA and CAGGTGATCCAGAACAAGCA for ftz-f1 (all from Sigma-Aldrich); or TaqMan assays: RpL32 (Dm02151827), br-Z1 (Dm01837161_m1), br-Z2 (Dm01821011_m1), br-Z4 (Dm01821013_m1).

Q-PCR was performed in a StepOne real-time PCR system (Applied Biosystems) and analyzed by ΔΔCT and normalized to RpS17 or RpL32.

For statistical analyses, two-tailed Student's t-tests were performed. P-values are indicated.

We thank Drs Michael B. O'Connor and Theodor E. Haerry for generously providing fly stocks and Dr Dorothea Godt for anti-Tj antibodies.