The Bric-à-Brac BTB/POZ transcription factors are necessary in niche cells for germline stem cells establishment and homeostasis through control of BMP/DPP signaling in the Drosophila melanogaster ovary

Summary

Many studies have focused on the mechanisms of stem cell maintenance via their interaction with a particular niche or microenvironment in adult tissues, but how formation of a functional niche is initiated, including how stem cells within a niche are established, is less well understood. Adult Drosophila melanogaster ovary Germline Stem Cell (GSC) niches are comprised of somatic cells forming a stack called a Terminal Filament (TF) and associated Cap and Escort Cells (CCs and ECs, respectively), which are in direct contact with GSCs. In the adult ovary, the transcription factor Engrailed is specifically expressed in niche cells where it directly controls expression of the decapentaplegic (dpp) gene encoding a member of the Bone Morphogenetic Protein (BMP) family of secreted signaling molecules, which are key factors for GSC maintenance. In larval ovaries, in response to BMP signaling from newly formed niches, adjacent primordial germ cells become GSCs. The bric-à-brac paralogs (bab1 and bab2) encode BTB/POZ domain-containing transcription factors that are expressed in developing niches of larval ovaries. We show here that their functions are necessary specifically within precursor cells for TF formation during these stages. We also identify a new function for Bab1 and Bab2 within developing niches for GSC establishment in the larval ovary and for robust GSC maintenance in the adult. Moreover, we show that the presence of Bab proteins in niche cells is necessary for activation of transgenes reporting dpp expression as of larval stages in otherwise correctly specified Cap Cells, independently of Engrailed and its paralog Invected (En/Inv). Moreover, strong reduction of engrailed/invected expression during larval stages does not impair TF formation and only partially reduces GSC numbers. In the adult ovary, Bab proteins are also required for dpp reporter expression in CCs. Finally, when bab2 was overexpressed at this stage in somatic cells outside of the niche, there were no detectable levels of ectopic En/Inv, but ectopic expression of a dpp transgene was found in these cells and BMP signaling activation was induced in adjacent germ cells, which produced GSC-like tumors. Together, these results indicate that Bab transcription factors are positive regulators of BMP signaling in niche cells for establishment and homeostasis of GSCs in the Drosophila ovary.

Keywords:

Cell differentiation – RNA interference – Transcription factors – Ovaries – Germ cells – Larvae – BMP signaling – Stem cell niche

Introduction

A stem cell niche allows, first, the establishment of stem cells, and second, the maintenance of a balance between stem cell self-renewal and differentiation. Much more is known about stem cell maintenance than about initial stem cell establishment. The interactions between niche and stem cells need to be strictly controlled for maintaining homeostasis of adult tissues. In fact, a defect in stem cell homeostasis can be pathological in humans, producing for example cancer stem cells [1–3], and may also be an important part of premature aging when stem cell populations lose their potential to self-renew [4]. The discovery of pre-metastatic niches in cancer [5,6] also makes the study of the properties of stem cell niches a key for gaining headway in cancer biology.

The Drosophila melanogaster adult ovary has proven to be an excellent model for understanding how interaction with adjacent somatic niche cells allows for maintenance of Germline Stem Cell (GSC) status [7,8]. Approximately 20 individual GSC niches, each associated with a small number of GSCs (2–3), are present in the Drosophila adult ovary at the tip of structures called germaria (Fig 1A). Each GSC niche is composed of several types of somatic cells: Terminal Filament (TF) cells, a triangularly-shaped transition cell (TC), Cap Cells (CCs) and the anterior Escort Cells (ECs) (Fig 1A) [9–11]. Both CCs and anterior ECs are in direct contact with GSCs. CCs are considered to be the key component of GSC niches, their number correlating closely with the number of GSCs [12,13]. CCs anchor GSCs to the niche by DE-cadherin-mediated adhesion [14] and produce two Bone Morphogenetic Proteins (BMPs), Decapentaplegic (Dpp) and Glass Bottom Boat (Gbb), acting as short-range secreted signals required for GSC maintenance [8,13,15–19]. Indeed, Dpp/Gbb signals are transduced in GSCs, which leads to the phosphorylation of the transcription factor Mad (pMad), its translocation into the nucleus and transcriptional repression of germline differentiation genes such as bag-of-marbles (bam) [18,20]. The homeobox transcription factor Engrailed (En), which is only expressed in niche cells in the ovary, binds dpp cis-regulatory sequences in vitro and activates dpp transcription in CCs allowing for GSC maintenance in the adult ovary [17]. Several other signaling pathways, such as the JAK/STAT [19,21] and Hedgehog [16,22,23] pathways, are active in different niche cells types and have also been implicated in regulation of dpp expression in these cells. In addition, ectopic expression of dpp or engrailed and ectopic activation of JAK/STAT signaling in non-niche somatic cells of the germarium, lead to a germarial GSC-like tumorous phenotype further supporting the implication of these factors in GSC homeostasis [8,18,19,24,25].

Morphogenesis of GSC niches occurs in the third instar larval ovary beginning with TF formation which involves flattening, sorting, intercalation and stacking of somatic TF cell precursors, initiating at the medial side of the ovary and progressing as a wave laterally (Fig 1A) [26,27]. Individualization of each TF is accomplished by the migration of apical somatic cells between TFs (Sheath Cells, Fig 1A) [28]. The number of TFs that form in the larval ovary (18–20) corresponds to the number of GSC niches at the adult stage [29–31]. At the prepupal stage, at the base of each newly formed TF, the anterior-most Intermingled Cells (ICs) differentiate into CCs adopting a cuboidal shape clearly distinguishable from that of TF cells (Fig 1A–1B') [10,32]. At this stage, the transition cell is already distinguishable (Fig 1A–1B') [10]. Posterior-most ICs will give rise to ECs in adult ovaries (Fig 1A) [33].

Only a few genes have been reported to be implicated in TF formation, notably, the bric-à-brac paralogs (bab1/bab2) [26,27,29,34,35] and the engrailed/invected paralogs (en/inv) [36,37] (Fig 1B–1D). The bab1 and bab2 genes encode proteins sharing evolutionarily conserved domains: a BTB (Broad-Complex, Tramtrack and Bric-à-brac)/POZ (POx virus and Zinc finger) domain involved in homodimeric and heterodimeric interactions [38] and a Bab-Conserved Domain (BabCD) involved in Protein-DNA interactions [34,39]. In the larval ovary, Bab1 has been reported to be present only in niche cells and Bab2 in all somatic cells, however at higher levels in niche cells [34] (Fig 1C and 1C’). Heterozygosity for strong or null alleles of both bab genes leads to a dominant phenotype characterized by an excess of TFs, resulting in an excess of GSC niches in adults [29], and by a recessive phenotype characterized by a defect in TF formation associated with production of atrophied ovaries with few germ cells in adults and sterility [26,27]. The en/inv paralogs are only expressed in GSC niche cells in larval and adult ovaries [31,40–42]. Induction of TF cell clones homozygous for a deletion encompassing both paralogs identified a function for these genes within TF cells for their correct alignment to form straight TFs [36]. CC specification has been shown to require the combined action of Notch signaling and the large Maf transcription factor Traffic Jam (Tj) [10,12,40,43,44].

The newly formed GSC niches become functional as of the prepupal stage [32]. Before niche formation, all Primordial germ Cells (PGCs) have been shown to exhibit BMP signaling activation [45,46], which is correlated with detection of Dpp in all somatic cells of the ovary [47]. During larval stages, Dpp signaling prevents PGC differentiation into germline cysts [32,45,48] and promotes PGC proliferation [32,47]. Upon GSC niche formation starting at the late third instar larval stage, dpp expression [32,48] and Dpp protein accumulation [47], become unevenly distributed in the ovary, with highest levels found in niche cells. Among PGCs, only those in contact with the Dpp-secreting niche cells retain BMP signaling activation as evidenced by the presence of pMad (Fig 1D–1D") [40] and become functional GSCs [32]. The more posterior PGCs, not in contact with niche cells, lose Dpp signaling activation (Fig 1D") and proceed to differentiate into cystoblasts, whereupon they produce the first germline cysts [32,48,49].

In this study, we addressed the roles of bab1 and bab2 specifically within niche cells for several aspects of functional GSC niche formation. Indeed, it has previously been proposed that the small size of adult germaria in bab mutants in which Bab proteins are depleted from all somatic ovarian cells might indicate a decrease in the number of GSCs produced [27]. We used a strong hypomorphic allele of bab1 (bab^A128) and Gal4-targeted RNA interference (RNAi) for efficient knockdown of each or both of the two paralogs specifically in GSC niches during their formation in larvae. This approach allowed us to demonstrate that drastic reduction of both Bab1 and Bab2 levels only within precursor niche cells inhibits TF formation. Surprisingly, cells exhibiting several CC characteristics were nonetheless present upon depletion of Bab proteins in developing niches. In addition, we have identified a new essential role for bab genes in prepupal niches for initial establishment of GSCs correlated with a role in ensuring dpp expression in CCs, as assessed by two dpp reporter transgenes. This function of bab in CCs is unlikely to require that of the en/inv genes since their expression levels were not affected when Bab1 and Bab2 were depleted. We also have evidence suggesting that En/Inv may not be essential for activation of BMP signaling and GSC establishment in the larval ovary, contrasting with their known essential functions in adult ovaries for BMP-mediated GSC maintenance. In adult ovaries, we show that Bab proteins contribute to GSC maintenance and activate a transgene reporting dpp expression in CCs. Finally, when bab2 was overexpressed outside of niche cells in somatic ovarian cells at the adult stage, germaria with a large excess of GSC-like cells forming tumors were produced. This was associated with ectopic somatic activation of a transgene reporting dpp expression and BMP signaling activation in GCs without ectopic expression of en/inv in the bab2-overexpressing somatic ovarian cells. Together, our results indicate that bab gene functions play a major role in niche cells for GSC establishment in the larval ovary and contribute to GSC maintenance in the adult, both likely via positive regulation of dpp expression in CCs.

Results

Reduction of Bab1 and Bab2 specifically in GSC niche cells during larval stages leads to abnormal TF formation

Although the two bab genes have been shown to be necessary for TF formation [26,27], we aimed at addressing their roles specifically within niche cells for this process. We thus tested the effect of depletion of each of the Bab proteins individually or together in TFs and CCs during niche formation. As a control, we used the hedgehog-Gal4 (hhG) driver combined with a UAS-GFP reporter (hhG>GFP) in order to indicate the cells in which the driver is active. We found specific expression of GFP in a subset of anterior somatic cells close to PGCs in L2 stage ovaries (S1A Fig) through to late L3 ovaries whereupon GFP was expressed in fully-formed TFs [33,50] (S1B–S1D Fig). In these prepupal ovaries, much higher GFP accumulation in medial vs. lateral niches was noted (Fig 2A–2A”’). Inside each niche, hhG>GFP expression levels were mosaic between TF cells and always low in CCs (Fig 2A and 2A').

In these control hhG ovaries, Bab2 was detected in all somatic cells at L2 and L3 stages (S1E–S1H Fig), and at higher levels in TFs and CCs in prepupae (Fig 1C and [34]). When the hhG driver was used to express a bab2-specific RNAi construct (hhG>bab2^IR), efficient Bab2 depletion was observed in most, but not all, hhG+ cells of L2 and midL3 ovaries (S1I–S1I" and S1J–S1J" Fig). By the prepupal stage, Bab2 depletion was observed in all medial TF cells, but not in all CCs, nor in lateral hhG+ cells consistent with low hhG expression in CCs and lateral TFs (Fig 2B', and S4B and S4B’ Fig). In this context, the formation of TFs is not affected (Fig 2B, 2B’, 2F and 2G). In contrast, bab2 knockdown in all ovarian somatic cells during niche formation, using the bab-Gal4 driver (hereafter named babG, S2 Fig), led to severe overall ovarian morphogenetic and growth defects (S3A Fig and [51]) indicating an essential role for bab2 in non-niche cells for ovarian development and rendering it impossible to study niche formation in this context.

Bab1 accumulation in control hhG>GFP ovaries was only detected as of mid L3 ovaries in cells beginning to form TFs (S1A–S1D’ Fig). By the prepupal stage Bab1 was detected at high levels in both TF cells and CCs (Fig 1C') as previously reported [34]. In addition, we report that low levels of Bab1 are also detected in ICs at this stage (Fig 1D'). For bab1 depletion, we used bab^A128 reported as a bab1 strong hypomorphic allele [34,52]. Females homozygous for bab^A128 had undetectable levels of Bab1 in prepupal ovaries (Fig 2C'). Like for Bab2, Bab1 depletion did not affect several parameters for normal TF formation (Fig 2C, 2C’, 2F and 2G). Importantly, we did not detect any cross-regulation between bab genes at the prepupal stage (Fig 2A’–2C’). From these results, we conclude that efficient depletion of either Bab1 or Bab2 in niche cells throughout ovary development does not affect niche morphogenesis.

Next, to test the effect of knocking down bab1 and bab2 at the same time in niche cells, we used four genetic contexts: (1) bab^A128,hhG>GFP,bab2^IR, (2) hhG>GFP,bab1^IR,bab2^IR, (3) bab^A128,hhG>GFP,bab1-bab2^shmiR (chained shmiR allow to co-express bab1 and bab2 shmiRs, [53]) and (4) hhG>GFP,bab1-bab2^shmiR. In prepupal ovaries from females of the first three genotypes, the medial part of the prepupal ovary contained a large cluster of hhG+ cells strongly depleted of both Bab1 and Bab2, which failed to form TFs (Fig 2D–2E' and S4C and S4C’ Fig). The fourth genotype, hhG>GFP,bab1-bab2^shmiR did not produce a clear mutant prepupal ovary phenotype likely due to inefficient extinction of bab1 (S4B and S4B’ Fig). The defects observed on niche morphogenesis upon depletion of both Bab proteins during larval stages are consistent with the previously-described phenotypes of bab gene mutants [26,27,34]. We quantified the number of TFs per ovary (Fig 2F), the number of cells per TF (Fig 2G) and the extent of flattening (ratio of the width to height of the cell) of the medial hhG+ cells since flattening is a hallmark of TF cell differentiation (Figs 1B’ and 2H) and found that these parameters were significantly lower than those in the control, while the overall number of hhG+ cells per ovary was not affected (Fig 2I). In the lateral-most part of the ovary, where hhG driver expression was low, low depletion of Bab proteins was consequently obtained and hhG+ cells formed apparently normal TFs (Fig 2D, 2D’, 2E and 2E’).

Therefore, efficient reduction of both Bab1 and Bab2 specifically in niche cells, using different genetic tools, impaired TF formation significantly, while reduction of each one individually did not. Our results also indicate that reduction of both Bab proteins did not affect the overall number of precursor niche cells. Together, these results indicate that Bab proteins are necessary specifically in these cells for the morphogenetic processes involved in TF formation during larval stages.

Reduction of Bab1 and Bab2 levels specifically in GSC niche cells during larval stages leads to loss of GSC establishment

We used the same genetic approaches as those described in the previous section to deplete Bab1 and/or Bab2 proteins from niche cells, and then asked whether the resulting disorganized hhG+ cells were able to recruit the initial GSCs during the larval to prepupal transition. To test for functional niche activity of hhG+ cells, BMP pathway activation in adjacent GCs was used as a characteristic of GSC identity. For this, accumulation of a direct downstream component of BMP pathway activation, phosphorylated Mad (pMad), was monitored in GCs. In control prepupal ovaries, pMad was detected in anterior-most GCs adjacent to niches at either high or low levels (Figs 1D', 1D", 3A and 3A' and S4D and S4D’ Fig). In the bab^A128 mutant context, pMad+ GSCs were present at the base of TFs (Fig 3B and 3B') and at the same mean number per ovary as that in the control (Fig 3F). In hhG>GFP,dicer2,bab2^IR ovaries, only very rare niches in which all CCs presented undetectable levels of Bab2 were obtained (Fig 3C"), and in these cases, adjacent GCs were always positive for pMad just as in Bab2+ niches (Fig 3A, 3A’, 3C and 3C’). These results indicate that strong reduction of either Bab1 or Bab2 individually in niche cells did not affect GSC establishment.

In prepupal ovaries, upon efficient reduction of both Bab1 and Bab2 in hhG+ cells, GCs were found in close contact with medial hhG+ cells, but almost all of these GCs were devoid of pMad staining (Fig 3D–3E', and S4D–S4E’ Fig). In contrast, in the lateral region where normal TFs were formed, pMad+ GCs (i.e. GSCs) were found at the base of each niche, as in control ovaries (Fig 3A, 3A’ and 3D–3E', and S4D–S4E’ Fig). In the Bab1 and Bab2 depleted ovaries, the mean number of GSCs per ovary was significantly lower than in the control and, when present, these GSCs were associated with lateral TFs in almost all cases (Fig 3F). Thus, when Bab1 and Bab2 levels were efficiently reduced in hhG+ niche cells using three different genetic contexts, GCs were present next to hhG+ cells, but the BMP pathway was, for the most part, not activated in these cells, indicating that GSC status was compromised.

Close contact between CCs and GSCs via E-Cadherin (E-Cadh) junctions has been shown to be important for GSC maintenance, and E-Cadh accumulation between CCs and GSCs was shown to begin at the late L3 stage [18]. We thus analyzed the levels of E-Cadh between hhG+ cells and adjacent GCs in hhG>GFP,bab1^IR,bab2^IR prepupal ovaries. In the normally-formed lateral niches, E-Cadh was observed between niche cells and underlying GCs, as is also the case in the medial niche region, even though properly structured niches were not present (S5 Fig). Together, our results indicate that bab1 and bab2 functions are necessary in the developing niche for GSC establishment by the prepupal stage and suggest that these functions are not likely related to deregulation of E-cadherin-mediated adherens junction contact between niche cells and GSCs.

We next asked whether GCs next to Bab1- and Bab2-depleted niche cells devoid of pMad are driven to differentiate precociously. Germline cell differentiation was monitored by the expression of a GFP transcriptional reporter for bam expression (bamP-GFP) [54]. In control prepupal ovaries (bamP-GFP, hhG>lacZ), GCs that differentiate (GFP+/Vasa+) were almost never detected among GCs contacting hhG+ (β-Gal+) niche cells (Fig 4A, 4A' and 4C), whereas all GCs located one-cell diameter away from these niche cells expressed this differentiation marker (Fig 4A'). In contrast, differentiating GCs were found significantly more frequently in direct contact with clusters of disorganized hhG+ (β-Gal+) cells with undetectable Bab1 and Bab2 in the medial zone of prepupal ovaries (Fig 4B'). Abnormal differentiation of some GCs adjacent to the cluster of medial hhG+ Bab1 and Bab2 depleted cells thus correlated with the absence of the GSC marker pMad in the same region. Taken together these results indicate that the functions of bab1 and bab2 are necessary in niches for acquisition of GSC status by adjacent PGCs as marked by sustained activation of the BMP pathway and absence of expression of the differentiation reporter bamP-GFP.

Reduction of Bab1 and Bab2 levels specifically in GSC niche cells during larval stages does not alter expression of several Cap Cell markers at the prepupal stage

We next tested whether Bab proteins are necessary for specification of CCs, which are known to be essential for GSC establishment and maintenance [10,12,13,55]. We characterized the nature of the hhG+ cells upon efficient reduction of Bab1 and Bab2 in these cells using a combination of markers that allow the distinction to be made between TF cells and CCs in prepupal niches: Traffic Jam (Tj) [56], P1444-LacZ [10], Delta [10,12] and E(spl)mβ-CD2 [57].

Within normal niches, high levels of Tj protein and P1444-lacZ expression designate CCs, while low levels of P1444-lacZ expression without any Tj distinguishes TF cells [10](Fig 5A–5A”’ and 5C’). Upon efficient reduction of Bab1 and Bab2 by RNAi specifically in niche cell precursors during larval stages, the disorganized clustered medial hhG+ cells were composed of two cell populations in prepupal ovaries. One of the populations was posteriorly-positioned, adjacent to GCs, and presented Tj nuclear accumulation (Fig 5B’–5B”’ and 5D’), associated with the expression of P1444-lacZ (Fig 5B’ and 5B”), like control CCs, except that the latter expressed higher levels of P1444-lacZ. Therefore, in posterior hhG+ cells, these CC markers were present even though Bab proteins were reduced to undetectable levels. The second population of cells within the medial hhG+ cell cluster, which was positioned anteriorly, did not accumulate Tj, like control TF cells (Fig 5B’ and 5D’), but unlike control TF cells, did not express even low levels of P1444-lacZ (Fig 5B’ and 5B”). The identity of this second population of cells is thus not that of normal TF cells.

We next tested for the presence of Notch pathway-associated niche markers, i.e. the Notch ligand, Delta (Fig 5C–5D”), and a Notch transcriptional reporter, E(spl)mβ-CD2 (Fig 5E–5F”). In hhG+ cells knocked down for bab1 and bab2, no Delta was observed at the plasma membrane, unlike in control TF cells (Fig 5C’, 5C”, 5D’ and 5D”). Instead, in anterior-most hhG+ cells, Delta was detected only in its cytoplasmic vesicular form, while posterior-most hhG+/Tj+ cells did not present any Delta at the plasma membrane nor in vesicles as for most of the control CC population (Fig 5D’ and 5D”). Despite this, all medial hhG+ cells strongly knocked down for bab1 and bab2 expressed the E(spl)mβ-CD2 reporter indicative of active Notch signaling, as control niche cells (Fig 5E’, 5E”, 5F’ and 5F”). Therefore, strong reduction of Bab proteins in niche cells during larval stages did not prevent Notch signaling activation in these cells. This activation may be due to vesicular Delta present in the most anterior Tj-/hhG+ cells (Fig 5D’ and 5D”) since they resemble control transition cells which have been shown to be Delta-sending cells presenting predominantly vesicular Delta and activating Notch signaling in adjacent CC precursors (Fig 5C’ and 5C” and [44]). Altogether, these results indicate that Bab1 and Bab2 are necessary for expression of some TF markers, which could be linked to the inability of Bab1 and Bab2 depleted hhG+ cells to form proper TFs. On the other hand, Bab1 and Bab2 do not seem to be necessary for expression of four different CC markers. Therefore, the almost complete absence of GSC establishment next to hhG+ cells deficient for Bab1 and Bab2 cannot be attributed to failure in CC specification as assessed by four different markers.

Reduction of Bab1 and Bab2 levels specifically in GSC niche cells during larval stages leads to loss of expression of two dpp reporter transgenes in these cells

The absence of GSC establishment in the larval ovary upon efficient reduction of Bab proteins in niche cells could be due to a specific defect in dpp expression in CCs. Therefore, we tested whether bab gene functions control dpp expression in CCs using two transgenic lines reporting dpp expression. The first one, hereafter called dpp-nlsGFP, contains a construction with a large genomic region (about 44 kb) covering both the dpp regulatory and complete coding sequences into which a C-terminal GFP tag was introduced [58]. To test this line as a valid dpp reporter, we characterized its expression pattern in germaria of the adult ovary since endogenous dpp expression has been more clearly described at this stage than at the prepupal stage [13,15–17,19], and found that it reflects the endogenous dpp expression pattern in CCs and prefollicle cells (S6A Fig). The second transgene, dppP4-lacZ, contains a small region of dpp cis-regulatory sequences (1.5kb) that controls expression of the lacZ coding sequence [59]. This dpp transcriptional reporter was shown to be expressed in CCs, like endogenous dpp, but, unlike endogenous dpp, it was also expressed in TF cells and absent from prefollicle cells, thereby not fully recapitulating the endogenous dpp expression pattern.

At the prepupal stage in the medial niche region of control ovaries containing either the dpp-nlsGFP or the dpp-P4lacZ transgene, GFP and β-Galactosidase (β-Gal) were detected in CCs (Fig 6A’, 6A”, 6C’ and 6C”). Dpp-nlsGFP was not found in TF cells, and rarely in transition cells (Fig 6A”), while β-Gal was slightly detectable in TF cells and often detected in transition cells (Fig 6C”). The CCs expressing the dppP4-lacZ transgene were in contact with GSCs (Fig 6C’). In striking contrast, upon efficient reduction of Bab1 and Bab2 in hhG+ cells, dpp-nlsGFP and dpp-P4lacZ were not or only very faintly expressed in posterior most-hhG+ cells of the medial region of the ovary considered as CC-like (Fig 6B–6B” and 6D–6D”) and this correlated with the absence of pMad in adjacent GCs (Fig 6B and 6D’). Altogether, these results indicate that, in niche cells, Bab1 and Bab2 are necessary for expression of two dpp transgenes in CCs at the prepupal stage.

Reduction of Bab proteins in niche cells during larval stages leads to reduced levels of Engrailed/Invected in TF cells, but not in CCs

Like the bab genes, engrailed and its paralog invected (en/inv) have also been shown to be involved in proper TF formation [36]. In addition, evidence strongly suggests that En directly controls dpp expression in adult CCs [17]. In order to explore possible functional interactions between Bab and En/Inv proteins, we tested whether en/inv expression in niche cells of prepupal ovaries depends on Bab proteins. In control prepupal ovaries, we found that en/inv are specifically expressed in fully formed TFs and CCs (Fig 7A, 7A’, 7B and 7B’), as previously reported [31,40,41]. In addition, we observed a significant difference in En/Inv levels between TF cells and CCs (Fig 7A, 7A’, 7B and 7B’), with a three-fold higher level in control TF cells than in CCs (Fig 7D). Mitotic TF cell clones homozygous for bab^AR07, a null allele for both bab1 and bab2 [34], showed a more than two-fold reduction in En/Inv levels when compared to control TF cells in the same prepupal ovaries (Fig 7A, 7A’ and 7D). In contrast, En/Inv levels in bab^AR07 CC clones were not significantly different from those in wild-type CCs in the same ovary.

Using RNAi to knockdown bab gene functions in hhG+ niche cells (Fig 7B–7C’), similar results were obtained. En/Inv levels were significantly lower in anterior-most hhG+ cells in hhG>GFP, bab1^IR, bab2^IR ovaries when compared to the hhG>GFP control TFs and CCs, while present at the same levels in posterior-most hhG+ Bab depleted cells in contact with GCs (Fig 7B’, 7C’ and 7E). Taken together these results show that bab gene functions are required in TF cells to ensure high En/Inv accumulation at the prepupal stage but are not required in CCs to ensure normal low levels of En/Inv. Therefore, the function of Bab proteins for expression of dpp in CCs, as assayed with two dpp regulatory sequence-containing transgenes (see previous section), cannot be linked to En/Inv, since Bab proteins do not control En/Inv accumulation in these cells.

Reduction of Engrailed/Invected levels in developing niches of larval ovaries does not prevent either TF formation or initial GSC establishment

To explore the implication of en/inv in niche formation using the same experimental approaches as for bab, we tested whether knockdown of en/inv in all niche cells would phenocopy the defect in prepupal TF morphogenesis provoked upon strong reduction of Bab1 and Bab2 in these same cells. To achieve efficient reduction of en/inv, we used the babG driver and RNAi transgenes directed against en and inv (babG>en^IR,inv^IR). Although this led to undetectable En/Inv in niche cells (Fig 8A and 8B), no abnormalities in TF formation were observed (Fig 8A–8E). The only mutant phenotype observed in babG>en^IR,inv^IR ovaries was a defect in TF individualization (Fig 8B–8B”), normally resulting from the migration of anterior somatic cells between TF stacks. These results strongly suggest that en/inv function is not essential for TF formation. Interestingly, these results also indicate that the function of bab in TF formation most likely does not depend on en/inv function.

Since in adult ovary GSC niches en is known to be necessary for GSC maintenance through the direct regulation of dpp expression in CCs [17,23], we tested whether en/inv gene functions are also necessary for GSC establishment in ovaries by the prepupal stage. Upon env/inv knockdown throughout larval development, prepupal ovaries displayed niches with high and low pMad levels in GSCs as in control (Fig 9A’ and 9B’), with a normal proportion of TFs associated with GSCs (Fig 9C). The average number of GSCs per ovary, however, was about 30% lower in en/inv RNAi ovaries than in control ovaries (Fig 9D). This difference corresponds to a lower proportion of high pMad+ GSCs per ovary (about half as many, Fig 9D). It is possible that under RNAi conditions enough En/Inv is present during larval ovary development to contribute to GSC establishment. However, the phenotype obtained is much weaker than that using RNAi to deplete Bab1 and Bab2 (Fig 3F). Therefore, it is possible to propose that GSCs with high pMad levels may be normally recruited at the end of the L3 stage in En/Inv RNAi-depleted niches, but that these niches may not have the capacity to maintain GSC status efficiently up to the prepupal stage. In order to assess whether GSC maintenance depends on en/inv function as of the prepupal stage, we used the Gal80^TS system (hereafter named G80^TS), which allows Gal4 activity at 29°C and impedes it at 18°C. We tested the fate of GSCs in 1-day-old adult females of the same genotype (G80^TS, babG> en^IR, inv^IR) but raised at different temperatures (Fig 9E–9G). Using 18°C as the rearing temperature, Env/Inv were present normally in niches (Fig 9F–9F”) and GSCs were present in every germaria (Fig 9F”’ and 9I). In contrast, GSC niches of females maintained at 29°C throughout development were efficiently knocked down for en/inv (Fig 9E–9E”) and 100% of corresponding germaria lacked GSCs (Fig 9E”’ and 9I), with a majority of germaria even devoid of any GCs (Fig 9H and 9I). Thus, all newly established GSCs in prepupal ovaries depleted of En/Inv were no longer present at adulthood. This result is consistent with the hypothesis that when En/Inv are efficiently reduced in larval stages, GSC establishment occurs but GSCs are not maintained if En/Inv continues to be reduced. This is further supported by the fact that partial re-expression of en/inv in niche cells from the onset of pupariation (29°C to 18°C switch) substantially rescued the adult loss of GSCs phenotype from 0% to more than 60% GSCs present (Fig 9G–9G”’ and 9I). Altogether, these results strongly suggest that en/inv functions are not essential for initial GSC establishment within newly formed niches, but would be essential for GSC maintenance in niches starting from at least the beginning of the pupal stage, and possibly even earlier, as soon as GSCs are established in larval stages.

Knockdown of bab genes in the adult ovary extinguishes dpp transgene expression and reduces GSC number

We next tested whether, beyond their function in GSC niches during larval stages, Bab proteins also play a role in adult GSC niches. In order to knockdown bab genes in niche cells only during the adult stage, we used the G80^TS system and analyzed ovaries from 7-day-old females. To set up the experimental design, we raised G80^TS, hhG>UAS-GFP flies at 18°C until pupal stage and shifted to the restrictive temperature (29°C, 31°C) at different stages. We found that proper reactivation of hhG in adult ovaries occurred when these flies were shifted to 29°C as of 24h after puparium formation (a stage where GSC niches are already formed and GSCs already recruited), followed by a shift to 31°C upon eclosion (S7A–S7A” and S7B–S7B” Fig). With these experimental conditions, adult ovaries of G80^TS,hhG>bab2^IR 7-day old females, though displaying a strong reduction of Bab2 levels in niche cells (S8A–S8B”’ Fig), were normal regarding the structure of the germarium and the mean number of GSCs per germarium (S8C–S8E Fig). For Bab1, it has been reported that adult ovaries homozygous for the strong hypomorphic bab1^A128 allele were normal [34]. Together, these results indicate that the depletion of Bab1 throughout development and in the adult or of Bab2 only as of the pupal stage does not affect adult niche formation and function. In contrast, in adult ovaries from G80^TS, bab1^A128, hhG>bab2^IR females, although a few normal germaria and ovarioles were present, the majority of the ovaries were rudimentary and largely devoid of clearly distinguishable ovarioles and GCs (S7C–S7F” Fig). In contrast, prepupal ovaries from females of the same genotype reared at 18°C, that displayed normal levels of Bab2 and no detectable Bab1, presented a normal morphology and niches containing GSCs (S7G–S7G" and S7J–S7J” Fig). Therefore, efficient reduction of Bab1 throughout development, along with depletion of Bab2 only as of the pupal stage, leads to loss of GSCs that were initially present at the prepupal stage. This mutant phenotype was too strong to allow analysis of possible defects in GSC maintenance and dpp transgene expression specifically. Thus, we used the protocol described above to reduce both Bab1 and Bab2 levels with RNAi constructs only as of pupal stages and were able to recover less altered ovaries. In adult germaria of G80^TS; hhG>bab1^IR,bab2^IR females also carrying the dpp-nlsGFP transgene, Bab1 and Bab2 were undetectable in CCs (Fig 10A–10D”) and strikingly, these CCs were also almost or completely devoid of Dpp-nlsGFP signal (Fig 10A”’, 10B”’, 10E”, 10E”’, 10F”, 10F”’, 10G”, 10G”’, 10H” and 10H”’). Therefore, in adult ovaries, Bab proteins are necessary for expression of this dpp transgene in CCs, as we showed is also the case in larval ovaries. Furthermore, in these adult ovaries, only 74.5% of germaria had at least one GSCs (Fig 10E–10F”’ and 10I(a)), while other germaria had either GCs with no detectable pMad in the niche region (Fig 10G–10G”’ and 10I(a)) or no GCs at all (Fig 10H–10H”’ and 10I(a)). When GSCs were present, their average number was significantly lower than in control adult ovaries (Fig 10J). Therefore, Bab proteins are important in CCs specifically during pupal and/or adult stages for GSC maintenance. Adult niches depleted of Bab1 and Bab2 and containing GCs displayed an overall normal structure of the anterior part of the germarium, with the presence of TF cells and CCs, although a small decrease in CC number was observed (Fig 10K).

To determine whether loss of GSCs in the adult ovary was progressive upon Bab protein depletion, we performed the analysis on the same G80^TS; hhG>bab1^IR,bab2^IR females but at 16-days of age. We observed a decline with age in the average number of GSCs in control females, as previously described [60], associated with a decrease in the mean number of GSCs with high pMad levels (Fig 10J). There was a smaller, but nonetheless significant, difference in GSC number between control and bab knockdown conditions in germaria at 16 days when compared to that of 7-day old females. However, the proportion of germaria with at least one GSC did not decrease between 7 and 16 days (Fig 10I(a,b)). Altogether, these results strongly suggest that Bab proteins are implicated in GSC maintenance through positive control of dpp expression in CCs and, possibly also through a small effect on establishing a full set of CCs per niche.

Overexpression of bab2 in germaria leads to GSC-like tumors

Since Bab proteins are necessary for TF morphogenesis and initial GSC establishment, we then asked whether these proteins could be sufficient to induce formation of ectopic TFs and/or GSCs. With this aim, we expressed UAS-bab1 or UAS-bab2 constructs with the C587-Gal4 driver (called C587G hereafter) at either 25°C or 29°C. C587G is expressed in most, but not all, somatic cells of developing female gonads [32]. We obtained prepupal ovaries exhibiting higher levels of Bab1 or Bab2 than in control ovaries, particularly in ICs, and ectopic Bab1 in basal somatic cells (S9A–S9C’ Fig). We found that ICs with higher Bab1 or Bab2 levels than in the control did not exhibit ectopic En/Inv accumulation (S9A”, S9B” and S9C” Fig). In addition, we did not detect any obvious change in Bab1 levels in cells overexpressing Bab2, and reciprocally (S9B, S9B’, S9C and S9C’ Fig). Finally, neither of these conditions appeared to affect the organization of pupal ovaries.

In adult germaria, C587G has been shown to be expressed only in ECs and early prefollicle cells [18]. In order to test the effect of expressing bab1 ectopically or over-expressing bab2 in adult ovaries, we analyzed ovaries from 10-day old females raised at 29°C as of eclosion. Under these conditions, Bab1 and Bab2 accumulated strongly in ECs and prefollicle cells (S10 Fig and Fig 11H' and 11I'). We did not observe any cross-regulation between the two genes (S10 Fig). Strikingly, in ovaries overexpressing bab2, all germaria contained more than the normal number of GSCs, some in continuity with the niche and others at ectopic locations (Fig 11A, 11A’, 11C, 11C’ and 11D). Among these germaria, 74% exhibited more than 20 GSCs and 26% between 5 and 20 GSCs (n = 137), when compared to 0 and 13% (n = 144), respectively, for control germaria (S1 Dataset). In many cases, overexpression of bab2 led to the formation of huge germaria with GSC-like tumors as marked by the presence of spectrosomes (Hts+ spherical cytoplasmic structures, Fig 11E). In these tumorous germaria, no ectopic En/Inv was detected in somatic cells outside of the endogenous niche (Fig 11D’). Upon ectopic expression of bab1, only 4 germaria out of 133 exhibited ectopic GCs in which the BMP pathway was activated (Fig 11B and 11B’), suggesting that Bab1 does not have the equivalent capacity as Bab2, or is not expressed at a high enough level in the condition tested to affect GSC number significantly. Altogether, these results show that bab2 overexpression in somatic cells outside of the adult GSC niche is sufficient to induce activation of the BMP pathway in adjacent GCs, thereby producing numerous ectopic GSCs and tumorous germaria.

Since Dpp is known to be one of the niche associated signals promoting GSC self-renewal, we next investigated whether bab2 overexpression in somatic cells outside of the adult GSC niche could induce expression of dpp reporters. We did not detect any ectopic lacZ expression when using dppP4-lacZ, which, although expressed in adult CCs like dpp, does not faithfully reproduce endogenous dpp expression as it is expressed in TF cells and not in prefollicle cells [13]. On the other hand, presence of the Dpp-nlsGFP fusion protein reflects endogenous dpp expression in the adult ovary, as it is found in CCs and prefollicle cells and not in TF cells (S6 Fig and Fig 11F and 11G) [13,15–17,19]. Upon overexpression of Bab2, Dpp-nlsGFP was found to be present ectopically in somatic cells between the GSC niche and the prefollicle cells along with Bab2 (Fig 11F–11I’ and corresponding merges). In addition, towards the middle of the germaria, somatic cells with both high Dpp-nlsGFP and Bab2 seemed to be more numerous than in the control (Fig 11F–11I’). Quantification showed that germaria of females overexpressing bab2 contained significantly more GFP-positive cells than control germaria (Fig 11J). Together, these results indicate that Bab2 somatic overexpression in adult germaria leads to expanded expression of a dpp-GFP genomic transgene and production of tumors of GSC-like cells in which the BMP/Dpp pathway is activated, independently of En/Inv expression.

Discussion

The presence of either Bab1 or Bab2 in precursor niche cells during larval ovary development is necessary for functional GSC niche formation

The two bab paralogs encode transcription factors that have been shown to be required for ovary morphogenesis, leg proximal-distal differentiation, and sexually-dimorphic abdomen pigmentation [26,27,29,52,61]. bab1 and bab2 display at least partially overlapping expression patterns in these three organs [34,61,62]. Indeed, Bab1 is always present with Bab2, while Bab2 is present alone in additional territories in the leg and ovary. The question of the individual functions between Bab paralogs has been hampered in part by their overlapping expression (at least during leg and ovary development), but especially by the lack of any known bab2 null mutation not affecting bab1. For bab1, the bab^A128 allele has been shown to be associated with loss of detectable Bab1 protein without affecting Bab2 levels (confirmed in the ovary in the present study), but no equivalent allele has been characterized for bab2 [34,63]. Roeske and co-workers [53] addressed the question of individual necessities of bab1 and bab2 in sex-specific abdomen pigmentation using tissue-specific expression of bab1 and bab2 shmiR constructs. They showed that the presence of both bab1 and bab2 (therefore four doses of Bab proteins) is necessary for efficient repression of pigmentation of female abdomen segments A5 and A6. In addition, their study and that of Couderc and co-workers [34] indicate equivalent capacities for Bab1 and Bab2 in this system since ectopic expression of either one at the same level leads to a similar decrease in pigmentation of segments A5 and A6 in males. Taken together, these results suggest functional equivalence between Bab1 and Bab2 proteins for sexually-dimorphic abdominal pigmentation and indicate that this process is particularly sensitive to Bab protein levels. Finally, one of the only direct targets for Bab protein transcriptional regulation, the yellow gene, has been identified in this system [53].

Our results indicate that efficient depletion of either Bab1 or Bab2 in developing ovary niche cells during larval stages leads to formation of normal TFs and correct establishment of initial GSCs by the prepupal stage. However, we cannot exclude that use of a strong hypomorphic bab1 allele (bab^A128) or RNAi-mediated depletion of bab1 or bab2, which lead to undetectable levels of each protein in developing GSC niches, nonetheless allows enough remaining Bab1 or Bab2 to carry out their functions in these two processes. On the other hand, efficiently reducing both Bab proteins at the same time using the same tools led to very dramatic defects in TF formation and GSC recruitment. Therefore, in larval ovaries, two normal doses of either Bab1 or Bab2 proteins are sufficient for efficient TF formation and GSC establishment. In addition, in double heterozygous females for amorphic or strong hypomorphic bab1 and bab2 alleles, we have previously shown that TF formation occurs and that an excess in TF number is even produced [29]. Therefore, one dose of each of Bab1 and Bab2 allows for TF formation. In contrast, ectopic expression or overexpression of bab1 or bab2, does not always produce the same effects. In this study, we found that somatic overexpression of bab2 in germaria led to production of GSC-like tumors, while under similar conditions bab1 ectopic expression did not affect GSC number significantly. Thus, either Bab1 does not have the equivalent capacity as Bab2, or it is not expressed at a high enough level, perhaps because of a difference in the expression levels of the UAS-bab1 and UAS-bab2 transgenes. In other contexts, Bab protein activities do not seem fully equivalent. In fact, bab mutant rescue experiments conducted with respect to leg and ovary developmental mutant phenotypes have shown that expression of either bab1 or bab2 is able to rescue the mutant phenotypes, but that Bab2 does so more efficiently [64]. In addition, bab2 has a much larger expression domain than bab1 in particular in the ovary, and removing bab2 from both GSC niches and the rest of the somatic ovarian cells leads to a much stronger phenotype of atrophied ovaries than removing bab2 only from niche cells precursors ([34] and this study). Thus, the role of Bab proteins in different tissues seems to require various doses of these proteins suggesting that they may have specific transcriptional targets in different tissues.

Newly identified roles for Bab proteins in niches for dpp reporter expression in Cap Cells and initial establishment of adjacent GSCs

The BMP family member Dpp, known to be emitted principally by CCs within germarial niches, has been shown to be essential for maintenance of GSC status in the adult ovary [8,17,18]. During the larval to pupal transition, Dpp produced by niches has also been shown to be required for GSC establishment [32,45]. We showed that when both bab1 and bab2 functions were removed from precursor niche cells during larval development, these cells were not organized into TFs and GSC establishment, as evidenced by activated Dpp signal transduction, was largely absent by the prepupal stage.

Using several markers, we showed that TF cell specification seemed particularly perturbed when Bab1 and Bab2 were both efficiently depleted in niche cell precursors. Though Tj expression was absent as in control TF cells, En/Inv levels were two-fold lower, and P1444-lacZ expression and membranous Delta were undetectable unlike in control TF cells. It is therefore not possible to attribute a clear identity to these cells. Incorrect TF cell specification could explain why these cells are unable to undergo proper TF morphogenesis. In contrast, CC specification was not affected according to the expression pattern of the same markers in Bab1 and Bab2-depleted developing niche cells suggesting that several aspects of this process do not depend on bab gene functions.

However, we found that two transgenes containing dpp transcriptional regulatory sequences were expressed normally in control CCs, but were not, or only very faintly, expressed in bab deficient CCs in prepupae and adult ovaries. Since hhG+ niche cells with undetectable levels of Bab proteins express some CC markers but not these dpp transgenes, these hhG+ cells are partially defective CCs. These results indicate that Bab proteins may thus be necessary for normal dpp expression in CCs, which is essential for the function of these cells in GSC establishment [32,45]. The control of dpp expression by Bab proteins may be direct or indirect. One of the dpp expression reporters used, dppP4-lacZ, only contains a 1.5kbp regulatory sequence from the 5' region of the dpp gene [59]. We found that Bab1 and Bab2 are necessary for expression of this reporter in niche cells at the prepupal stage, but Bab2 is not sufficient for ectopic expression of dppP4-lacZ in adult germaria. In contrast, the second dpp transgene, containing a much larger genomic region including the dpp coding region and surrounding sequences (about 44kbp, [58]), was sensitive both to reduced levels of Bab proteins in larval and adult niches, and to overexpression of Bab2 in outside of niches in adult germaria. Bab proteins have been shown to bind A/T rich sequences with TA or TAA repeats [39] and the dppP4-lacZ 1.5kbp regulatory sequence, though containing A/T rich sequences which may bind Bab proteins, do not contain the consensus sequence (TAAATATAATTG), nor the in vitro determined optimal binding sequences of 3 or 4 TAA motifs in a row. The larger 44kbp genomic dpp-nls-GFP transgene, of course, contains many potential Bab protein-binding sites so it is probably not surprising that these two transgenes do not respond exactly the same to changes in Bab levels. Finally, a study in the embryo has shown enrichment of a Bab2:GFP fusion protein on two fragments of about 400bp in the 5' region of the dpp gene [65], and these do not overlap the dppP4-lacZ fragment which is nonetheless also in this region of the gene. Therefore, Bab proteins may regulate dpp expression directly.

Complementary roles for Bab1/Bab2 and En/Inv in larval and adult GSC niches

In the developing ovary, en/inv are expressed specifically in niches during their formation. Our results show that their functions, however, are not essential for niche formation. Indeed, we found that when En/Inv were depleted efficiently by RNAi in developing niche cells during larval stages, the same number of correctly formed TFs, and TF cells and CCs per TF were found as in controls. This is consistent with the observation of larval ovaries upon En/Inv depletion in Allbee et al. [66], using a different niche cell Gal4 driver and the same en/inv RNAi transgenes used in the present study. In contrast, another study, based on induction of clones of ovarian somatic cells homozygous for a null allele for both en/inv, concluded that these genes are involved specifically for proper TF cell alignment within individual stacks in larval ovaries [36]. The difference with the results presented here may be related to the different approaches used to abolish en/inv expression. With the RNAi approach used here, all niche cells are depleted of En/Inv, while en/inv null mutant clone induction used in Bolívar et al. [36] led to the formation of stacks with both wild-type and mutant TF cells. Possible heterogeneity in the identity between the two populations of cells may have impeded normal interaction between TF cells during the intercalation process leading to TF formation. We cannot exclude that the RNAi approach, though leading to undetectable levels of En/Inv proteins, may nonetheless not have depleted En/Inv proteins completely. However, the babG driver used for en/inv RNAi induction is expressed early in L3 stage, before niche formation, and is strongly and homogeneously expressed in all niche cells, contrary to hhG which, nevertheless, gave a very strong TF formation defect with bab1/bab2 RNAi transgenes.

Our results also indicate that En/Inv are not essential for initial GSC establishment. In fact, when their levels were efficiently reduced, the proportion of prepupal niches associated with at least one GSC was the same as in the control, but the overall number of GSCs per ovary was about 30% lower than in the control at this stage. We cannot exclude that RNAi knockdown of en/inv in niches was not sufficient to produce a stronger decrease in initial GSC numbers. However, the fact that we were able to show that all of the GSCs present at the prepupal stage in this en/inv RNAi context were subsequently completely lost between pupal stages and the first day after eclosion suggests a defect in GSC maintenance rather than in GSC establishment. In addition, when expression of en/inv was restored in niche cells from the beginning of pupariation, at least one GSC was present in more than 60% of adult germaria. These results suggest that the known function of En/Inv for GSC maintenance in the adult [17,23], may be necessary for this process even earlier, as soon as functional niches are formed. In contrast, upon strong reduction of Bab proteins in larval niche cells, the number of GSCs was about 90% lower than in the control at the prepupal stage, indicating a crucial role for Bab in initial GSC establishment.

In the adult ovary, a role for En/Inv in GSC maintenance through direct regulation of dpp expression has been shown [17], but this had not been tested for Bab proteins until now. Using the temperature-sensitive Gal80^TS/UAS/Gal4 system to knockdown bab1 and bab2 only after GSC establishment at the pupal stage, we found that ovaries of 7-day old females present a significantly lower mean number of GSCs per germarium than control ovaries, as well as germaria with no GSCs and even no GCs at all, none of which was found in the control. This could not be attributed to a loss of en/inv expression in these cells. There was however, a slight but significant, lower mean number of CCs per niche when compared to the control (6.2 and 6.8, respectively). Nonetheless, the almost undetectable expression of the dpp-nlsGFP transgene in adult CCs strongly depleted of Bab proteins indicates that the loss of GSCs is most likely due to loss of dpp expression. The adult GSC loss phenotype is however much stronger when En/Inv proteins are depleted (this study and [17,23]). These results indicate that Bab1 and Bab2 contribute to GSC maintenance in adult ovaries likely via control of dpp expression in CCs.

Therefore, we propose that Bab proteins are the major players for initial GSC establishment through control of dpp expression in CCs during larval stages, and that, as of this point, GSC maintenance is mainly supported by En/Inv. Bab function for initial GSC establishment may be due solely to the direct regulation of dpp expression levels by these proteins, as shown for En for adult GSC maintenance, or to indirect regulation through specific targets of other transcription factors. Indeed, Lmx1a, encoding a LIM-homeodomain transcription factor expressed in somatic apical cells, TF cells and CCs in the prepupal ovary, is necessary for ovary morphogenesis similar to Bab proteins [66]. In addition, Bab proteins were shown to be necessary for Lmx1a expression. Thus, Lmx1a is a Bab target, either direct or indirect, and should be tested in the future for mediation of the role of Bab on dpp expression.

Bab proteins are involved in homeostasis of GSCs in adult ovaries

We have shown that during ovary development, Bab proteins are necessary in niches for BMP pathway activation in initial GSCs, while they also contribute to GSC maintenance in the adult ovary. In both cases, this is associated with Bab positive control of expression of transgenes containing dpp transcriptional regulatory sequences, and independent of En/Inv. We also found that Bab2 overexpression in ECs and prefollicle cells in adult germaria was sufficient to produce excess GSCs with activated BMP/Dpp pathway, leading to the formation of tumorous germaria. This GSC-like tumorous phenotype was similar to that previously described for germarial overexpression of dpp [8,18] and en [17,24], and also upon JAK/STAT signaling ectopic activation [19,21]. Importantly, the GSC-like tumorous phenotype observed in germaria overexpressing bab2 was independent of En/Inv. Our results indicate that a dpp genomic transgene is expressed in a significantly greater number of somatic germarial cells upon bab2 overexpression than in the control and some of these are positioned in the region between the niche and prefollicle cells where dpp and this transgene are not expressed normally. This study thus identifies a new role for Bab proteins in GSC niches in regulating GSC establishment and homeostasis through activation of the BMP/Dpp pathway, thereby adding another level of complexity to be integrated into this system.

Materials and methods

Fly stocks

We used hedgehog-Gal4 (gift from P. Therond and hereafter called hhG), a Gal4-expressing enhancer trap insertion in the hh locus, to target niche cells during their differentiation, and bab^PGal4-2 (gift from J.L. Couderc and hereafter called babG), a Gal4-expressing enhancer trap insertion in the bab locus, to drive expression in all somatic cells of the larval ovary ([51] and present study). These drivers were combined with UAS-dicer2 and tub-Gal80^TS (gift from J. Montagne) when indicated. Two different UAS-GFP transgenes (one associated with a Nuclear Localization Signal sequence, Bloomington Drosophila Stock Center, BDSC 4476, and one without, gift from A. Boivin), were used in combination with the hhG driver to mark niche cells. The RNAi lines UAS-bab1^IR (Vienna Drosophila Stock Center, VDRC 50285) and UAS-bab2^IR (VDRC 49042), and the bab1 strong hypomorphic allele bab^A128 [52] (gift from D.Godt) were used for the bab loss-of-function analysis. bab^AR07,FRT80B/TM6B (gift from M. Boube) and hs-FLP;; ubi-GFP, FRT80B (gift from J. Montagne) lines were used to generate bab null mitotic cell clones. bab^AR07 is a deletion mutation inactivating both bab1 and bab2 [34]. UAS-en^IR (VDRC 35697) and UAS-inv^IR (BDSC 41675) were used for the engrailed/invected knockdown analysis. The enhancer reporter line P{PZ}1444 (P1444-lacZ) [67] and bamP-GFP, a GFP transcriptional reporter for bam expression ([20], Kyoto DGRC 118177) were used as cell-type specific markers. E(spl)mβ-CD2, in which the sequence encoding rat CD2 protein is inserted downstream of the Enhancer of Split [E(spl)mβ] promoter that is activated by Notch signaling, was used as a readout of Notch pathway activity [57] (gift from A. Bardin). dpp-nlsGFP (VDRC 318414) [58] and P4-lacZ [59] (gift from R. Xi) were used as dpp reporters. dpp-nlsGFP contains 43766 bp of the dpp genomic locus. P4-lacZ contains 1494bp of a dpp enhancer region. To drive ectopic and over-expression of bab1 and bab2, we used C587-Gal4 [32] (gift from T. Xie), UAS-bab1 (BDSC, 6939) and UAS-bab2^4-66 [61] (gift from A. Kopp).

Experimental conditions

For all the experiments, flies were raised on standard cornmeal medium under uncrowded conditions. Prepupal ovaries correspond to ovaries extracted from 0–2 hour old white and slightly older (<3 hours) yellow prepupae.

For pupal analysis of the effect of bab and engrailed/invected (en/inv) depletion in niche cells with bab^A128, hhG, babG, and RNAi lines, crosses were started at 25°C, parents removed 24-to-48h later, descendants were then transferred to 29°C and ovaries from prepupae were dissected.

To generate bab mosaic mutant prepupal ovaries, bab^AR07, FRT80B/TM6B flies were crossed to hs-FLP;; ubi-GFP, FRT80B flies at 25°C. Clones marked by absence of GFP were induced by one 1-hour heat shock at 38°C at the end of the second instar larval stage, and prepupae were dissected for ovary analysis 50h after heat shock.

Transient expression of the UAS-engrailed^IR and UAS-invected^IR transgenes was achieved using tub-Gal80^TS and babG. For analysis of adult (1-day old) ovaries upon RNAi-mediated knockdown of engrailed/invected (en/inv), two conditions were used: 18°C (from crosses to dissection) for the control and 29°C (crosses at 25°C and transferred 24-to-48h later to 29°C) for RNAi activation. To test the effect of the depletion only during niche formation, en/inv were reexpressed at the prepupal stage by shifting prepupae from 29°C to 18°C for the rest of development.

Pupal and adult knockdown of bab1 and bab2 in G80^TS; hhG/UAS-bab1^IR, UAS-bab2^IR flies for comparison with in G80^TS; hhG/+ control flies was achieved raising females at 18°C to the young pupal stage (24h after puparium formation, APF), shifted to 29°C until eclosion and then shifted to 31°C for 7 days.

To analyze the effect on prepupal ovaries of bab1 ectopic expression or bab2 overexpression in somatic cells of larval ovaries, crosses were conducted at 25°C for 24-to-48h and descendants were then transferred either to 29°C or maintained at 25°C until dissection of prepupal ovaries. For bab1 and bab2 gain-of-function experiments in the adult, individuals were raised at 18°C and then shifted to 29°C or to 31°C just after eclosion of females as indicated in the figures legends. Ovaries from 10 day-old females were dissected.

Immunostaining

Prepupal ovaries were dissected in PBS medium, fixed in 4% formaldehyde (R1026, Agar Scientific) for 20 to 30 minutes, and washed 3 x 10 minutes in PBT (PBS supplemented with 0.3% tritonX-100 –Sigma T8787). Ovaries were then incubated in a blocking solution (PBTA: PBT supplemented with 1% BSA–Sigma A3059) for a minimum of 20 minutes. Primary antibodies were diluted in PBTA and ovaries were incubated in this solution for 6 hours at room temperature or overnight at 4°C. The following primary antibodies were used: rabbit anti-Smad3 (1:200, ab52903, Abcam); rabbit anti-Bab1 (1:4000, gift from T. Williams); rabbit anti-GFP (1:200, FP-37151B, Interchim); rat anti-Bab2 (1:4000, gift from J-L. Couderc); rat IgM anti-Vasa (1:500, Developmental Studies Hybridoma Bank-DSHB); mouse anti-En/Inv recognizing both paralogs (1:200, 4D9, DSHB); guinea-pig anti GFP (1:200, 132 002, Synaptic System); mouse anti-β-Gal (1:400, 40-1a, DSHB); guinea pig anti-Traffic Jam (1:5000, gift from D. Godt); mouse anti-Delta (1:200, C594.9B, DSHB), mouse anti-rat CD2 (1:100, MCA154GA, Biorad); mouse anti Hts (1:200, DSHB). After primary antibody incubation, tissues were washed in PBT 3x10min, and incubated in PBTA for at least 30 minutes before incubation in secondary antibodies in PBTA. The following secondary antibodies were used: donkey anti-mouse Cy3 (715-165-151- Jackson Laboratories); and the following ones from Thermo Fisher Scientific, chicken anti-rabbit 488 (A21441); goat anti-rabbit 568 (A11011); goat anti-guinea pig 488 (A11073), anti-mouse 488 (A11029); anti-rat IgM 647 (A21248); anti-rat IgG 647 (A21247). Nuclei were detected with DAPI (1:200, 1mg/ml) and F-actin with 647nm-fluorescent dye conjugated phalloidin (1:200, 65906, Sigma). After secondary antibody incubation, tissues were washed 3x10 min in PBT, and mounted between slide and coverslip in DABCO (D27602, Sigma) with 70% glycerol. For larval ovaries, a spacer was positioned between the slide and the coverslip to avoid ovary squashing. The lateral side of the prepupal ovary was identified by its contact with the fat body.

Fluorescence microscopy, image analysis and statistics

Images were acquired with a 40x or 63x objective on a confocal laser-scanning microscope Leica TCS SP8 equipped with 405nm, 488nm, 552nm, and 638nm emission diodes. Same settings were used between the different genotypes in each experiment. Prepupal ovaries were acquired with 1μm steps between each section and AOTF/EOF compensation was used to increase diode percentage during acquisition using the LasX software. Adult ovary images were acquired with 1μm steps between each section.

Images represent either a projection of 2–4 consecutive sections, full tissue stack, or a 3D reconstruction of the entire tissue. Images were processed with Fiji [68].

Counting TF number in prepupal ovaries was achieved through immunodetection of niche cell markers, such as GFP (hhG+ cells), Bab1 or En/Inv, or DAPI labeling of flattened TF cell nuclei. Both lateral views and 3D reconstructions from the anterior pole of ovaries were used for this quantification. To distinguish TF cells from CCs within niches, labeling of F-actin, which delimits cell contours and/or DAPI labeling of TF cell nuclei was used. Niche cells with contours covering the entire width of the TF and with flattened nuclei were considered as TF cells, and more basal cells contacting GCs that did not cover the width of the TF and had round nuclei, as CCs (See Fig 1B, compare green and yellow brackets, respectively).

hhG+ cell flattening of control prepupal ovaries (hhG>GFP) or upon depletion of Bab proteins (hhG>GFP, bab1^IR, bab2^IR) was quantified with GFP immunostaining marking hhG+ cells and fluorescent phalloidin-bound F-actin marking cell contours in projections of 5 consecutive sections separated by 1μm. The width and height between cell contours of individual hhG+ cells were measured, and the degree of flattening corresponded to the ratio between the width and height.

For comparison of En/Inv levels between control niche cells and niche cells depleted of Bab proteins (bab^AR07 null allele and RNAi-mediated depletion), the confocal section presenting the biggest area for individual nuclei was used to quantify fluorescence intensity of En/Inv levels using the RawIntDen tool in Fiji.

To quantify GSC number in both prepupal ovaries and adult niches, we used nuclear pMad staining in GCs as a readout of Dpp pathway activity. pMad staining levels in nuclei were very variable between GCs of same ovary and between ovaries of a same experiment ranging from very high to very low when detected with Royal LUT. GCs presenting relatively low but unambiguous nuclear pMad staining were considered separately from GCs with high pMad+ GCs but both were considered as GSCs since pMad detection reflects high or moderate activity of Dpp/BMP pathway (see Fig 1D’ and 11D”).

Prism 7 (GraphPad Software) was used for statistical tests and for the generation of graphs. Values are presented as means + s.d., p-value calculated using a two-tailed Student's t-test or Mann Whitney test when two conditions were compared, or an ANOVA one-way test or Kruskal-Wallis test for comparisons between multiple conditions. Results are indicated with: NS (Not Significant—p>0.05); * (0.05>p>0.01); ** (0.01>p>0.001); *** (0.001>p>0.0001) and **** (p<0.0001). The data used to make all the figures and statistical analyses in this study can be found in S1 Dataset.

Supporting information

S1 Fig [tf]
expression patterns in developing niches and Bab1 and Bab2 depletion using UAS-RNAi transgenes and the driver.

S2 Fig [green]
Expression pattern of the driver during L3 stages.

S3 Fig [a]
Bab2 is necessary for ovary development during larval stages.

S4 Fig [green]
Almost complete reduction of Bab1 and partial reduction of Bab2 levels using shmiRNAs impedes TF formation and GSC establishment.

S5 Fig [tfs]
High levels of E-Cadherin are found between Germ Cells and niche cells depleted of Bab1 and Bab2.

S6 Fig [a]
Expression pattern of the transgene construct in adult germaria.

S7 Fig [a]
Reduction of Bab proteins in niche cells only from pupal to adult stages leads to loss of GSCs.

S8 Fig [green]
Reduction of Bab2 in niche cells from the early pupal stage onwards does not affect GSC numbers in adult germaria.

S9 Fig [lut]
The ectopic expression of or the overexpression of does not affect ovary morphogenesis.

S10 Fig [green]
Adult ectopic expression and overexpression using the driver.

S1 Dataset [xlsx]
Data used to make all the figures and statistical analyses.

S1 Text [docx]
Supplemental materials and experimental procedures.