brinker levels regulated by a promoter proximal element support germ cell homeostasis

Maintenance of germline stem cell (GSC) homeostasis is regulated by numerous pathways that signal between the germline and somatic cells that comprise the stem cell niche, as well as by other external and long-range signals (Nelson et al., 2019; Zhang and Cai, 2020). In the Drosophila melanogaster model system, the ovary, in which the oocyte develops into a mature egg, contains about fifteen ovarioles composed of germline and somatic cells (Fig. 1B,B′). At the anterior-most tip of each ovariole lies the germarium; a tapered structure made up of several distinct cell types that support the differentiation of one GSC daughter into a cystoblast (CB) and maintenance of the GSC lineage by the other daughter (Fig. 1C,C′). The anterior-most region of the germarium contains the stem cell niche, which comprises three somatic cell populations – the terminal filament (TF), cap cells (CCs) and an anterior subset of escort cells [ECs, alternatively inner germarial sheath (IGS) cells] – and supports the maintenance of two or three GSCs throughout adulthood (Fig. 1C; Liu et al., 2015; Tan et al., 2018; Xie and Spradling, 2000). ECs located more posteriorly influence the differentiation of stem cells, forming what is considered the ‘differentiation niche’ (Kirilly et al., 2011). GSCs produce cystoblasts via asymmetric division aligned along the anterior-posterior axis of the germarium such that daughter cells that move out of the niche escape the self-renewal signal and begin to differentiate, whereas those that remain in contact with the CCs are maintained as GSCs (de Cuevas and Spradling, 1998; Deng and Lin, 1997). This is a complex but well-studied phenomenon that requires the precise localization and interaction of a number of cell signaling pathways (reviewed by Harris and Ashe, 2011; Hayashi et al., 2020; Hsu et al., 2019).

The most pivotal cue is arguably extracellular Decapentaplegic (Dpp), the main Drosophila BMP ligand, which is expressed at high levels in and secreted from CCs to promote self-renewal of GSCs within approximately one cell diameter (Eliazer and Buszczak, 2011). In many tissues, Dpp functions as a long-range morphogen to direct cell fate decisions in a concentration-dependent manner across tissues; however, in its role in GSC maintenance, the range of Dpp is limited by the expression of receptors and extracellular matrix components in the niche that serve as a sink for extracellular ligand (Guo and Wang, 2009; Liu et al., 2015; Wang et al., 2008; Wilcockson and Ashe, 2019). This limited signaling range is crucial for proper germline development as GSC division pushes daughter cells out of the range of Dpp and permits differentiation factors, such as the gene bag of marbles (bam), to be expressed (Song et al., 2004). It was also recently shown that dpp is expressed at low levels in ECs to maintain a population of partially differentiated germline cells that can de-differentiate to repopulate the germarium (Liu et al., 2015); a result suggesting that not only the presence of signals, but also their expression levels, can be interpreted by the germline to affect cell fate.

The transcription factor Brinker (Brk) encodes a canonical repressor of Bmp signaling and has been demonstrated to repress expression of both dpp (Hasson et al., 2001; Theisen et al., 2007) and Dpp-dependent target genes (Rushlow et al., 2001; Sivasankaran et al., 2000). Inversely, BMP signaling activates a complex that directly represses expression of brk (Marty et al., 2000). As a result of this mutual repression, brk is typically expressed in an obverse pattern to dpp, for example in the embryo (Jaźwińska et al., 1999a) and wing imaginal disc (Campbell and Tomlinson, 1999; Minami et al., 1999), which, in addition to other mechanisms, helps shape the Dpp gradient (Affolter et al., 2001; Müller et al., 2003; O'Connor et al., 2006). Brk acts similarly in the ovaries starting at stage 8 of oogenesis when it is important for establishing the anterior-posterior gradient of Dpp expression, which patterns the eggshell and is essential for dorsal appendage and operculum formation (Chen and Schüpbach, 2006). Despite extensive studies of Dpp expression in the germarium, however, the role of Brk in this tissue has not been previously described.

Here, we show that, not only is brk expressed in the germarium, but its expression coincides with and, remarkably, positively regulates dpp expression in somatic cells. We also demonstrate that a previously described promoter-proximal element (PPE), first characterized as supporting distal enhancer action in the early embryo (Dunipace et al., 2013), also has a role in the ovary.

To examine the role of the PPE in supporting ovary brk expression, we first used a set of large reporters in which the brk coding sequence is replaced with gfp in the context of ∼30 kb of flanking sequence (brkNFgfp). We examined wild-type reporter expression as well as that of reporters with deletions of the full-length PPE (PPE2kb) or its distal or proximal subdomains (PPEdist and PPEprox, respectively) (Fig. 1A; Dunipace et al., 2013). These reporters were used previously in the embryo to show that the PPE does not itself drive expression but instead serves to facilitate the action of other enhancers located at a distance (Dunipace et al., 2013). To identify the cell types expressing brk reporters, ovaries were co-stained with antibodies for Traffic jam (Tj) to mark all follicle cells except for the TFs, Lamin C (LamC) to mark CCs and TF, and/or α-Spectrin (Spec), which outlines all later-stage follicle cells as well as marking both spectrosomes (a specialized rounded organelle found in GSCs and cystoblasts) and fusomes (found on differentiating cysts) (de Cuevas et al., 1996; Li et al., 2003; Xie and Spradling, 2000). In the germarium, brkNFgfp was expressed in a number of cell types: TF, CCs, ECs and follicle stem cells (FSCs) (Fig. 1D-D″). Expression persisted in follicle cells throughout egg chamber development (Fig. S1A), becoming restricted after stage 7 to only the columnar follicle cells (CFCs) and the polar and border follicle cells (PCs, BCs) (Fig. 1C″,E,E′). Deletion of PPE2kb from brkNFgfp (brkNFgfp-ΔPPE2kb) abolished GFP reporter expression in the ovary, except for some low level expression in the FSC region (Fig. 1F-G′), indicating that the PPE is required for all brk expression in this tissue.

Previously, we found that the PPE could be divided into subunits that were largely redundant in their ability to support distal enhancer action in the early embryo (Dunipace et al., 2013). We therefore examined expression of the brkNFgfp reporter in which either the distal or proximal PPE domain is deleted (brkNFgfp-ΔPPEdist and brkNFgfp-ΔPPEprox, respectively). In the germarium, PPEdist and PPEprox appeared to be generally redundant in their ability to support brkNFgfp reporter expression (Fig. 1H-H″,J-J″). Similarly, each domain appeared to be sufficient to support CFC reporter expression (Fig. 1I,K). However, in brkNFgfp-ΔPPEdist specifically, GFP signal was lost in many follicle cells associated with mid-stage egg chambers (Fig. S1B), including stage 10 BCs/PCs (Fig. 1I′), indicating a distinct requirement for that element to support expression in a subset of cells.

In order to explore further the sufficiency of the PPE to drive expression in the ovary, we created direct-fusion nuclear-localized RFP transgenic reporters of the PPE and its distal and proximal subdomains (PPE2kb>, PPEdist> and PPEprox>NLS-mCherry, respectively; Fig. 1A; Table S1; see Materials and Methods). These small transgenic reporters revealed that the PPE acts as a traditional enhancer in the ovary as it is capable of driving expression in multiple cell types (Fig. 2A-F′). The PPE2kb reporter is active in most of the somatic tissues in the niche: the TF, CCs and ECs (Fig. 2A-A″), but was not detectably expressed in the germline, either in the germarium or in later-stage nurse cells (Fig. 2A,B). In stage 10 egg chambers, PPE2kb drove expression in the BCs and PCs but only weakly in the CFCs (Fig. 2B,B′).

When we examined direct reporter expression driven by PPE subdomains, we observed significant differences in levels of activity. The nuclear localization of these reporters permitted quantification of differences in levels of expression that were not readily observable from the cytoplasmic brkNFgfp signal, especially in ECs because these cells form extensive protrusions into the germline (Kirilly et al., 2011). Expression levels of small reporters were quantified in accessible cell types, the ECs and CCs in the germarium, but excluding TFs, which were variable as a result of mounting (Fig. 2G,H; Table S2; see Materials and Methods). In the germarium, PPEdist drove higher levels of expression in CCs and ECs than the full PPE2kb (Fig. 2C-C″,G,H). Inversely, PPEprox drove low level expression in CCs and a subset of ECs (Fig. 2E-E″,G,H). In late-stage egg chambers, PPEdist was expressed in both BCs and PCs, but drove little to no CFC expression (Fig. 2D,D′). By contrast, PPEprox supported weak expression in BC and CFCs, but did not support expression in PCs (Fig. 2F,F′). Finally, PPEdist was active in the late-stage germline where brkNFgfp and other brk reporters were not detected (Fig. 2D,D′; Fig. S1E). These results indicate that both halves of the PPE can drive reporter expression in both CCs and ECs in the germarium, but at different levels (distal high, proximal low) and that wild-type expression levels (i.e. those of the full-length PPE2kb) require both halves. In late-stage egg chambers, the two PPE domains act more additively with each supporting subsets of the full expression pattern, except in the case of the germline expression of PPEdist, which is repressed in the context of the full PPE2kb.

To provide insight into PPE function in CFCs, we examined a previously described distal brk cis-regulatory module (CRM), brkB (Charbonnier et al., 2015), which is not active in the germarium or mid-stage egg chambers (Fig. 2I-J′) but does drive strong expression exclusively in the CFCs (Fig. 2K,K′). The fact that the PPE itself is not a strong CFC driver (Fig. 2B), but brkNFgfp reporter expression is lost upon PPE deletion (Fig. 1G), indicates that the PPE is required for brkB activity in CFCs. Also, like the redundancy previously noted in the early embryo, neither distal nor proximal deletion affects PPE reporter expression in CFCs (Fig. 1I,K), indicating that either region is sufficient to support the action of distal enhancers, such as brkB. Taken together, this reporter analysis suggests that in the ovary the PPE has two functions: to facilitate the action of other enhancers and to serve as a direct driver of brk expression (Fig. 2L).

The fact that we observe brk expression in cells that comprise the germline stem cell niche (i.e. TF, CCs and ECs) suggests that Brk plays a role in regulating germline homeostasis. To test this, we generated deletions of the PPE and its distal and proximal subdomains in the context of the endogenous brk locus using CRISPR-Cas9 genome editing (ΔPPE2kb, ΔPPEdist and ΔPPEprox, respectively; see Materials and Methods, Tables S1 and S3). We also deleted brkB (Charbonnier et al., 2015) in the same manner (ΔbrkB). PPE, but not brkB, deletions had significant effects on germarium morphology, including germline differentiation, spectrosome number and distribution, and the overall organization of the germline, as well as expression pattern and level of Bam, which marks differentiating cystoblasts (Fig. 3). Specifically, whereas wild-type germaria contained two or three GSCs, which present rounded spectrosomes, contact the CCs, and can be labeled by phosphorylated Mothers Against Decapentaplegic (pMad) antibody staining (Fig. 3A,U; Song et al., 2004), germaria from PPE mutant females consistently contained significantly more pMad⁺ cells (Fig. 3E,I,M,U). This pMad⁺ cell population likely contains true GSCs (in contact with CCs) as well as dysregulated cystoblasts (located in proximity to, but not directly contacting the CCs) and will be referred to collectively hereafter as pMad⁺ cells. Counterintuitively, this increase in pMad⁺ cell number occurred in all PPE mutants (i.e. ΔPPE2kb, ΔPPEdist and ΔPPEprox), despite the fact that these deletions had varying effects on brk reporter expression levels (Fig. 2G,H). Furthermore, pMad⁺ cell number was unaffected in ΔbrkB germaria (Fig. 3Q,U), indicating that this is a PPE-specific effect. We also observed that the number of rounded spectrosomes, which mark GSCs and cystoblasts, increased correspondingly with pMad⁺ cell number, confirming that these changes represent a delay in the differentiation of pMad⁺ cells and cystoblasts into more mature cysts (Fig. 3B,B′,F,F′,J,J′,N,N′,R,R′).

These observations of effects on germline homeostasis in the niche are further supported by our findings of corresponding changes in overall morphology of the germline in mutant germaria. The germline in wild-type germaria can be divided into morphological regions whereby region 1 contains the GSCs, cystoblasts and 2- to 8-cell cysts; region 2a contains the 16-cell cysts and 2b the same cysts once they have adopted an elongated, lens-like shape that spans the width of the germarium; and region 3 contains a spherical cluster comprising 15 nurse cells and one oocyte completely enclosed by follicle cells (see Fig. 1C′; King, 1970; McKearin and Ohlstein, 1995). PPE mutants showed aberrant germline morphology with tumorous expansion of regions 1 and 2a with region 2b sometimes affected (Fig. 3H,L,P compared with 3D). Cyst organization appeared to recover by region 3, which was structured normally in nearly all samples. Germline organization appeared normal in ΔbrkB germaria (Fig. 3T). Strikingly, ΔPPEprox ovaries also lacked detectable Bam expression (Fig. 3O) whereas Bam was present in all other genotypes examined (Fig. 3C,G,K,S). This loss of the key differentiation factor, Bam, is likely to be a contributor to the elongated region 1 and lack of distinction of region 2 of ΔPPEprox germaria. However, the severity of this particular germline phenotype (i.e. loss of Bam) did not correlate with fertility: ΔPPEdist and ΔPPEprox mutants are both viable and fertile whereas ΔPPE2kb mutants are semi-lethal (few survivors) and female sterile, and the brkB mutants are healthy but female sterile. Because ΔPPEprox mutants are normally fertile, it is likely that the lack of detectable Bam is rescued by a counteracting effect on the gene network, such as repression of Pumillio (Pum) as pum, bam double mutants have been shown to rescue differentiation defects caused by loss of Bam alone (Chen and McKearin, 2005).

Our analysis of brk PPE reporters suggest a model where PPEprox functions to downregulate brk gene expression as PPEdist alone is a stronger driver of reporter expression than the full-length PPE2kb (Fig. 2G,H). To confirm and quantify brk expression in wild-type and PPE deletions, we used in situ hybridization chain reaction (HCR), which relies on signal amplification to permit highly sensitive labeling of RNA in fixed samples (Choi et al., 2018). We observed colocalization of brk HCR signal with Tj antibody staining, confirming expression of brk in all ECs and CCs in both wild-type and ΔPPEprox mutant germaria (Fig. 4A,B). Our analysis of the PPEdist-mCherry reporter raised the possibility that brk might be expressed in the germline (Fig. 2D); however, reporters have been previously shown to be regulated in unexpected ways specifically in the Drosophila germline (DeLuca and Spradling, 2018). We therefore used HCR to examine brk expression in the germline. Because HCR signal is non-nuclear, assigning cell identity to observed spots is challenging in the germarium, where ECs form projections that surround germline cells (Kirilly et al., 2011). To overcome this, we examined confocal stacks of germaria co-stained with Tj antibody and did indeed observe brk expression in the germline, i.e. only those cells in the center of germaria that lack Tj staining (Fig. 4A-A″, blue dotted outline). This germline brk expression was upregulated in ΔPPEprox mutants (Fig. 4B-B″, compare with 4A-A″). We quantified brk expression levels in the early germarium as a whole (stages 1-2a, follicle and germline cells representing ‘non-CCs’), as well as in CCs alone by counting pixels above background normalized to area in wild type as well as in ΔPPEdist and ΔPPEprox mutants (see Materials and Methods). Consistent with the trends observed from the H2A-mCherry reporter constructs, this HCR analysis confirmed that brk expression is lower in ΔPPEdist mutants and upregulated in ΔPPEprox mutants in both the CC and non-CC regions of the germarium (Fig. 4F). Taken together, both reporter and HCR results support a model in which PPEprox acts as a ‘damper’ to downregulate expression supported by PPEdist in all brk-expressing cell types of the germarium.

We had previously noted that brk PPE mutants have altered numbers of pMad⁺ cells and reasoned that this might be caused by changing expression of the BMP ligand Dpp, which provides a key self-renewal signal for GSCs and has a well-studied repressive regulatory relationship with brk. Again using HCR, we co-stained for brk and dpp transcripts in wild-type and PPE mutant ovarioles (Fig. 4C-E; Fig. S2A-D″). Surprisingly, we found that brk and dpp expression domains overlap in wild-type germaria, with dpp and brk co-expressed at high levels in CCs, as well as at lower levels throughout the germarium (Fig. 4C-C‴). Although this was an unexpected result (e.g. Jaźwińska et al., 1999b; Minami et al., 1999), our observation of co-expression of brk and dpp in the germarium is supported by two recent single-cell RNA-sequencing studies of the ovary (Rust et al., 2020; Slaidina et al., 2020). We also observed a corresponding increase in dpp levels in CCs as well as non-CC cell types in ΔPPEprox germaria (Fig. 4E-E‴,G) indicating that, in contrast to its canonical role as a repressor of dpp and its target genes (e.g. Jaźwińska et al., 1999a; Takaesu et al., 2008; Theisen et al., 2007), Brk upregulates dpp expression in this developmental context. In keeping with this trend, loss of brk in CCs of ΔPPEdist germaria led to loss of dpp expression, again indicating a positive regulatory relationship (Fig. 4D-D‴,G). In contrast, however, we observed an increase in dpp expression in non-CC cells of ΔPPEdist germaria despite a loss of brk in that region. Although this finding of increased non-CC dpp in ΔPPEdist mutants helps clarify the contradictory observation that all PPE mutants have increased pMad⁺ cell number despite varying effects on brk levels (Fig. 3U), these experiments do not shed light on why we did not observe a positive brk-dpp regulatory relationship in the context of the ΔPPEdist non-CCs. One explanation is that, because these are constitutive PPE mutants, they may have effects in multiple interacting cell types (including within the germline; see below) or at earlier time points.

Importantly, the positive relationship between brk and dpp observed in ΔPPEprox mutants and in the CCs of ΔPPEdist mutants is specific to the germarium, as these genes exhibited an obverse expression pattern in late-stage egg chambers, as would be expected from their canonical mutually repressive relationship (Fig. S2A-B″).

As the germarium phenotypes described above (Fig. 3) resulted from constitutive PPE mutants, we next sought to identify specific tissues in which regulation of brk levels contributes to germline homeostasis. We used two approaches based on the GAL4/UAS binary expression system to manipulate brk levels and to generate cell type-specific mutants (Brand and Perrimon, 1993; Duffy et al., 1998). Using a GAL4 driver active in either the soma (all follicle cells except TF and stalk cells), tj-GAL4 (Fig. 5A; Sahai-Hernandez and Nystul, 2013), or the germline, nos-GAL4 (Fig. 5F; Van Doren et al., 1998), we first performed tissue-specific brk overexpression and RNA interference (RNAi). We repeated these misexpression experiments in the presence of temperature-sensitive GAL80 (GAL80^TS) and with timed temperature shifts to constrain GAL4 activity to the adult, as constitutive mutants might produce effects at earlier developmental time points that could confound interpretation of phenotypes (McGuire et al., 2004; see Materials and Methods).

Second, we used brk null mutants with linked FRT sites to facilitate generation of cell type-specific brk⁻ mosaic clones using the same GAL4 drivers described above. In short, GAL4 expression in the presence of UAS-FLP recombinase drives mitotic recombination between brk mutant, FRT chromosomes and a wild-type FRT homolog generating mosaics only in the GAL4-expressing tissue (see Materials and Methods). These null clones served to validate our observations from brk RNAi and as a proxy for brk^ΔPPE2kb and brk^ΔPPEdist, as both these lesions functioned as mutants in that they led to a reduction in brk reporter expression across all germarium cell types (Figs 1F and 4F). We also generated corresponding ΔPPEprox, FRT recombinants to make cell type-specific brk^ΔPPEprox mosaic clones, which we predicted would cause clonal brk overexpression for comparison with parallel UAS-brk experiments.

Because we observed dpp upregulation as a result of increased brk in ΔPPEprox mutants and dpp is a key GSC maintenance signal, we reasoned that this phenomenon might explain the increased pMad⁺ cell phenotype observed in this genotype. We therefore first examined germaria in which brk was overexpressed in specific cell types. Indeed, somatic overexpression of brk using either UAS-brk or brk^ΔPPEprox-FRT with tj-GAL4 led to an increase in pMad⁺ cell number (Fig. 5D,E). Parallel adult-constrained overexpression experiments with GAL80^TS (see Materials and Methods) were not significantly different from non-GAL80 experiments, indicating that pupal tj-GAL4 activity does not contribute to the increased pMad⁺ cell number in the case of somatic brk ectopic expression (Fig. 5E). The tj-GAL4 driver was expressed in most follicle cells of the ovary, including weakly in CCs and stronger in all ECs, with stronger expression in the follicle cells associated with later-stage egg chambers (Fig. 5A; Li et al., 2003). In order to identify specific somatic domains in which brk overexpression affects pMad⁺ cell number, we next performed parallel overexpression experiments using additional GAL4 drivers active in subsets of follicle cells: c587-GAL4, a strong driver in all anterior follicle cells of the germarium except CCs (Jin et al., 2013), and GMR25A11-GAL4, a strong driver in a subset of region 1 and 2a ECs (Liu et al., 2015). As expected by their largely overlapping expression profiles in the anterior germarium, C587-driven, adult-constrained brk overexpression recapitulated tj-GAL4 overexpression (Fig. S3A). In addition, there was also a significant increase in pMad⁺ cell number when brk was overexpressed only in a subset of anterior ECs using GMR25A11-GAL4 (Fig. S3B). As the anterior ECs represent the intersection of these three drivers' expression domains, with GMR25A11 being highly specific to that cell type, these data collectively support the conclusion that increased brk in anterior ECs is sufficient to induce an increase in pMad⁺ cell number. In contrast, brk overexpression experiments using a driver specific to FSCs and later follicle cells, 109-30-GAL4 (Hartman et al., 2015), did not lead to an increase in pMad⁺ cells (Fig. S3C). Taken together, these overexpression experiments establish that increased brk levels in anterior ECs alone is sufficient to induce an increase in pMad⁺ cell number.

In parallel, these drivers were also used to assess the effects of loss of brk in somatic cells using brk RNAi or by generating tissue-specific brk null clones. In general, loss of brk in the somatic cells led to a loss of pMad⁺ cells, the opposite effect to that observed in overexpression experiments (Fig. 5C,E; Fig. S3A,B). Additionally, tj>brkRNAi ovaries often had collapsed germaria containing only one or no pMad⁺ cells and many spectrosomes (Fig. 5C). There was, however, no significant change in pMad⁺ cell number in either adult-constrained tj-GAL4 knockdown or in brk mutant mosaics using this driver (Fig. 5E), indicating that loss of somatic brk at an earlier stage in development may contribute to the low pMad⁺ cell number phenotype and the collapse of the differentiation niche. However, adult-constrained c587-GAL4 knockdown experiments did show a decrease in pMad⁺ cell number (Fig. S3A), indicating that adult knockdown of brk can also affect pMad⁺ cell number. These germaria also have more spectrosomes, similar to tj>brkRNAi experiments (Fig. S4B; compare with Fig. 5C). The discrepancies between tj and c587 adult-constrained experiments could be explained by the differences in the subsets of germarium cell types that express these drivers (i.e. TF by C587-GAL4 and CCs by tj-GAL4), but may also be due to a difference in driver strength. Indeed, when we compared c587- or tj-driven UAS-mcd8-GFP reporter expression, GFP signal was markedly higher in c587-GAL4 germaria (Fig. S4E,F). This difference in levels correlates with the severity of collective germarium phenotypes, with the weaker driver, tj-GAL4, producing perturbed germaria but a wild-type number of pMad⁺ cells, and the stronger driver, c587-GAL4, causing more severely perturbed germaria and a reduction or complete loss of pMad⁺ cells (Fig. S4A,B). The fact that experiments with a third somatic driver expressed in ECs, GMR25A11-GAL4, could produce either an increase or a decrease in pMad⁺ cell number from somatic brk overexpression or knockdown, respectively (Fig. S3B), lends further support to the conclusion that changes in expression of brk – in either direction – specifically in anterior ECs is sufficient to affect pMad⁺ cell number.

We again performed parallel experiments using an alternative somatic driver, 109-30-GAL4, which is specific to FSCs and later follicle cells (Hartman et al., 2015). Although experiments with this driver did not significantly affect germ cell number, except in ΔPPEprox mosaics (Fig. S3C), 109-30-driven brk knockdown did result in a significant reduction in the number of mid-stage egg chambers (Fig. S4C). This loss of mid-stage egg chambers was also observed after adult-constrained brk knockdown with either tj-GAL4 or c587-GAL4, both of which are active in FSCs (Fig. S4A,B), but not with GMR25A11-GAL4, which is not expressed in the FSCs (Fig. S4D). To quantify this phenomenon, we counted the number of egg chambers per ovariole for each of these brk knockdown genotypes. All drivers active in later FCs (tj-GAL4, c587-GAL4 and 109-30-GAL4) showed a significant reduction in egg chamber number, but the early EC driver (GMR25A11-GAL4), which is absent from FSCs and later FCs, did not (Fig. S4G). Taken together, these results indicate that Brk is required in FSCs and later-stage follicle cells for specification and/or proliferation of follicle cells outside the niche. The lack of effect on germ cell number in 109-30-GAL4 and adult-constrained tj-GAL4 brk RNAi experiments further suggests that this FSC role is separable from the contribution of Brk to germline differentiation.

Because both the tj-GAL4 and c587-GAL4 experiments produced significant changes in pMad⁺ cell number and both drivers showed some expression in either the CCs or TF, tissues that are known to contribute to germline homeostasis, we sought to identify whether brk expression levels in these tissues might also affect pMad⁺ cell number. To investigate this possibility, we performed brk overexpression and knockdown experiments using the bab1-GAL4 driver (constrained to the adult with GAL80^TS), which is specific to TF and CCs (Bolívar et al., 2006). Although bab1-driven brk overexpression did result in a significant increase in pMad⁺ cells, in agreement with other somatic driver experiments, parallel bab1-driven brk knockdown produced the same phenotype (Fig. S3D). This finding conflicts with results from all other anterior somatic drivers tested (Fig. 5E; Fig. S3A,B). One possible explanation is that our use of GAL80^TS was not sufficient to constrain bab1 activity to the adult and our observations were therefore confounded by leaky pupal GAL4 activity. Alternatively, brk levels in the TF and CCs may have different effects that conflict when brk is either up- or downregulated in both cell populations. However, further testing for a specific requirement for brk in the TF and CCs was not possible because FRT experiments are not appropriate for this driver, as they depend on mitosis to generate clones and TF and CCs do not divide in the adult (Bolívar et al., 2006). Although our HCR imaging suggests that constitutive ectopic brk does indeed lead to an increase in dpp specifically in CCs (Fig. 4G), we were not able to discern the contribution of CC-specific brk expression to germ cell differentiation.

In addition to its role in the soma, brk expression in the germline itself is also required for proper germ cell differentiation. Despite the fact that brk expression was not detectable in the germline using either the large (brkNFgfp) or small (PPE2kb-mCherry) reporter (Figs 1D and 2A), loss of germline brk (either misexpression or mosaic) generated using the germline-specific nos-GAL4 driver (Fig. 5F) led to increased GSC number (Fig. 5G-J). These results add support to our finding that brk is expressed in the germline in wild-type germaria and further suggest that its presence there is important to support germ cell homeostasis. In experiments without GAL80^TS, this effect was also observed in overexpression conditions (nos>brk and ΔPPEprox, FRT; Fig. 5J); however, addition of GAL80^TS and appropriate temperature shifts rescued this phenotype, indicating that pupal nos-GAL4 activity was responsible for the overexpression-pMad⁺ cell loss phenotype (Fig. 5J). Furthermore, experiments using bam-GAL4, which is active only in the cystoblast and the 2-8 cell cysts (Chen and McKearin, 2003), recapitulated the trends in pMad⁺ cell number observed in nos-GAL4, GAL80^TS experiments (Fig. S3E). This result underscores the important role that brk plays in the germline despite its relatively low levels of expression there, as knocking down brk even in very few cells significantly impacts germ cell differentiation. Unlike the experiments described above that alter expression in somatic cells, changing brk levels in the germline does not have noticeable effects on fertility or strong defects in later-stage egg chambers, indicating that germline brk expression is important for GSC homeostasis but not for later ovariole development.

The ΔPPEprox mutant supports increased levels of brk in both the soma and the germline, leaving open the possibility that the increased expression of dpp observed in this mutant is non-cell-autonomous. Therefore, we again performed brk overexpression and RNAi experiments using both tj-GAL4 and nos-GAL4 (constrained to the adult using GAL80^TS) and performed HCR analysis to assess the effects on dpp expression. In concordance with the increased pMad⁺ cell number we observed upon tj-driven brk overexpression (Fig. 5D,E), we also observed a corresponding increase in expression of dpp in the ECs of these germaria (Fig. 6E,E′, compared with A,A′, I), supporting the conclusion that increased brk upregulates dpp in the same cells (Fig. 6G-J). Notably, because tj-GAL4 is only expressed at low levels in the CC, we did not observe a significant increase in brk in the CCs upon tj-driven brk overexpression (Fig. 6E,H), yet there was still a significant increase in dpp expression in these cells (Fig. 6E′,J). Similarly, the loss of dpp in the CCs in the tj>brkRNAi (Fig. 6C,C′ compared with A,A′, J) is consistent with the loss of pMad⁺ cells in this genotype (Fig. 5C,E). As discussed in the preceding section, this change is likely at least partially due to an earlier role for brk in the pupal ovary.

To investigate further the mechanism for the pMad⁺ cell increase in the ΔPPEprox mutant, we used RNAi to knock down dpp in the mutant background with two different FC drivers (tj-GAL4 and GMR2A11-GAL4). In both cases, the number of pMad⁺ cells in the ΔPPEprox was decreased, mirroring the effect of the dpp RNAi in the wild-type background, indicating that tumorous germaria in the ΔPPEprox mutant are dependent on the levels of dpp in the ECs (Fig. S5). In contrast, neither upregulation nor knockdown of brk in the germline changed the expression of dpp in either the ECs or the CCs (Fig. 6B,B′,D,D′,F-J), indicating that the increase in pMad⁺ cells observed in nos>brkRNAi germaria (Fig. 5H,J) is independent of dpp levels.

To gain insight into the specific function of Brk in the germline, we again generated germline brk mutants, as well as performing RNAi and overexpression using both somatic and germline drivers, and examined readouts of key signaling pathways and established markers of germarium morphology. We found that increased brk expression, either ubiquitously (ΔPPEprox) or specifically in either the somatic cells (tj>brk and C587>brk, Gal80^TS) or the germline (nos>brk, GAL80^TS), leads to posterior expansion of the CC marker Engrailed (En) (Fig. 6N-P compared with 6K), with no such effect apparent upon RNAi (Fig. 6L,M). Additionally, we observed enrichment of the Drosophila E-Cadherin Shotgun (Shg), in ECs in all of the conditions in which brk is upregulated in the EC (i.e. ΔPPEprox mutants, tj>brk and C587>brk, Gal80^TS) compared with wild type (Fig. S6A-D′). Shg is a key component of adherens junctions, which are important in wild-type germaria to maintain GSC contact with cap cells so that they remain within range of the dpp self-renewal signal (Song et al., 2002). These results suggest that precise regulation of brk levels in the somatic cells of the germarium is important for EC fate (see Discussion).

Consideration of tissue-specific perturbation of brk levels and the resulting effect on germ cell number (Fig. 5) as well as on dpp levels (Fig. 6) sheds light on the seemingly contradictory observation that all constitutive PPE mutants have increased numbers of pMad⁺ cells (Fig. 3) despite having opposing effects on brk levels (Fig. 4F). The increase in pMad⁺ cells observed in the ΔPPE2kb and ΔPPEdist mutants can be explained by loss of germline Brk, resulting in de-repression of BMP signaling. This is supported by our germline-specific brk misexpression experiments (Fig. 5J; Fig. S3E) and represents the expected result for Brk acting in its canonical role as a repressor of dpp and its targets (e.g. Fig. 7C,E). Conversely, increased pMad⁺ cell numbers observed in ΔPPEprox mutants is caused by increased EC brk expression, which we show results in increased dpp (Fig. 4E,G). This phenomenon is recapitulated by our somatic brk misexpression experiments (Figs 5E; 6E,I,J; Fig. S3A,B) and represents a previously unappreciated regulatory relationship between brk and dpp (e.g. Fig. 7B,D).

brk expression in the ovary depends on a cis-regulatory PPE, which has pleiotropic and multimodal functions. Previously identified in the early embryo as required for the action of two distal enhancers (Dunipace et al., 2013), here we show that the PPE also performs that function in the ovary where it supports enhancers active in CFCs. In the germarium, however, the PPE also itself acts as a driver of expression, supporting expression of brk within the soma as well as the germline. Furthermore, the proximal and distal PPE domains interact to finely tune levels of brk output. This regulatory role is crucial as we find that perturbation of brk levels in either direction – in soma or germline – significantly impacts germ cell homeostasis. Somatic brk levels also have important implications for follicle cell development in later-stage egg chambers and for fertility. Analysis of brk PPE mutants as well as cell type- and developmental stage-specific perturbation of brk levels demonstrate that Brk is a near-ubiquitous factor in the ovary with broad impacts on development.

In other developmental contexts, Brk has been shown to act primarily as a repressor of the BMP ligand Dpp and its targets, often to shape morphogen gradients formed as a result of long-range Dpp signaling. We were therefore not surprised to find that brk is expressed in the ovary, where dpp is known to play an important role as a short-range signal that supports GSC maintenance. What was unexpected, however, was the extent to which brk and dpp expression domains overlap. Our results establish that Brk acts as a positive regulator of dpp in the germarium and itself has an important role in regulating germline homeostasis. This role reversal of Brk, from its canonical function as a repressor of dpp and its targets, to a positive regulator (Fig. 7A,B,D) appears to be confined to the germarium; in other ovary tissues where BMP signaling is important (e.g. egg chambers stage seven and older), dpp and brk domains are mutually exclusive (Fig. S2A-B″).

This unexpected brk and dpp co-expression in wild-type germaria, as well as our finding of dpp upregulation in response to brk overexpression, aligns with the phenotypes we observed in mutants that lead to somatic brk overexpression (i.e. ΔPPEprox and tj>brk), including increased GSC number. Taken together, these observations support a model in which somatically expressed Brk promotes escort cell dpp expression (Fig. 7B,D), which effectively expands the niche by providing self-renewal signals throughout the anterior germarium. This model is further supported by the expansion of the cap cell marker En into the escort cell domain in brk overexpression mutants (Fig. 6N,O). En has also been demonstrated to be necessary and sufficient for the expression of dpp in the germarium (Luo et al., 2017), indicating that the upregulation of dpp as a result of brk overexpression could also be an indirect effect mediated by En (Fig. 7B,D).

Enrichment of Cadherin in the escort cells in ΔPPEprox mutants, which overexpress brk, suggests that these cells are not only changing their gene expression programs as a result of higher brk but are also adopting morphological properties key to the cap cells' biophysical function in the niche (Fig. S6B). These abnormal escort cells present in ΔPPEprox mutants, which display mixed characteristics of both CCs and ECs, are similar to the phenotype observed in mutants for the Swi/Snf chromatin remodeling complex protein Osa, which exhibited a similar expansion of En and dpp expression into ECs, as well as an increased number of undifferentiated germ cells, yet continued to express EC markers such as the c587-GAL4 reporter (Hu et al., 2021). Epigenetic modifications in general are crucial for germ cell maintenance, and expansion of BMP signaling in the niche is a common phenotype of misregulation of these factors (e.g. Eliazer et al., 2014; Wang et al., 2011; Yang et al., 2017). These broad changes in gene expression suggest that Brk is required for EC differentiation and/or chromatin remodeling.

Our observation of increased pMad⁺ cell number in germline-specific brk mutants or RNAi was surprising given our findings that increased brk expression in the soma is correlated with increased pMad⁺ cell number, indicating that Brk has opposite effects in these closely connected tissues. This finding, however, aligns with our analysis of the constitutive PPE mutants in which deletion of the full-length, distal or proximal PPE domains all resulted in increases in pMad⁺ cell number, even though these deletions had opposing effects on brk expression levels. Taken together, these results indicate that proper regulation of brk in both the germline and soma is collectively important and that relative expression levels may be more important than the absolute expression in a single tissue. Brk and Dpp are known to be involved in cell-cell competition and survival in other tissues in which the relative expression level from the neighboring cell is important in determining the fate of a given cell (Moreno et al., 2002). Additionally, in the testis, ectopic Dpp signaling in cyst stem cells (CySCs) resulted in CySC-GSC competition and GSC loss (Lu et al., 2019), indicating that the signaling between the germline and soma can result in cell-cell competition in the opposing tissue. Based on our findings in this study, it is probable that the relative expression levels of brk between the germline and somatic cells is involved in stem cell competition and important for GSC homeostasis. Furthermore, if, as in the wing disc, the upregulation of brk is required for apoptosis of less-fit cells in a tumorigenic tissue (Moreno et al., 2002) then loss of brk in the germline could indeed lead to excess GSCs.

All flies used were strains of Drosophila melanogaster and reared at 25°C on standard fly media. See Table S1 for details regarding stocks used.

Large reporter constructs have been described by Dunipace et al. (2013). The H2A-mCherry small reporter construct was made by subcloning the hsp70 promoter from pUASBP (Horne-Badovinac and Bilder, 2008) into the attB vector (Bischof et al., 2007), and subsequently cloning the H2A-mCherry reporter with associated SV40 terminator (Kadam et al., 2012) downstream of the promoter. PPE small reporter fragments were amplified by PCR from brk-gfp (Dunipace, et al., 2013) and cloned into the H2A-mCherry vector. All reporter constructs were injected into y¹ M{vas-int.Dm}ZH-2A w*; M{3xP3-RFP.attP}ZH-86Fb flies [Bloomington Drosophila Stock Center (BDSC) #24749].

For deletions within the genome, gRNA constructs were created by modifying the pCFD4 plasmid (Addgene plasmid #49411) to target PAM sequences flanking the region to be deleted. flyCRISPR Optimal Target Finder (https://flycrispr.org/target-finder/) was used to identify the PAM sequences with no predicted off-target hits. The gRNA plasmids were then injected into y2 cho2 v1 P{nos-phiC31\int.NLS}X; attP2 (III) (NIG-Fly #TBX-0003) flies. Stable gRNA transgenic lines were created and then crossed to a Cas9-expressing line (y2 cho2 v1; Sp/CyO, P{nos-Cas9, y+, v+}2A, NIG-Fly #Cas-0004). Individuals from the next generation were screened by PCR for the deletions (see Table S3 for primers and genomic sequence of mutants).

Flies containing either P{UAS-brk.3PF3} (brkUAS.Tag:HA, BDSC #78350) or P{NIG.9653R} (UASt-dsRNA against brk, NIG-FLY #9653R-2) transgenes were crossed to flies containing the desired GAL4 driver (tj-, c587-, GMR25A11-, bab1-, 109-30-, nos- or bam-GAL4). F1 females of the appropriate genotype were collected and raised at 27°C on standard fly media supplemented with yeast paste together with males to promote robust egg production. For genotypes indicated in Figs 5, 6, S3 and S4, GAL80^TS was used to limit activity to the adult ovary by crossing flies at 18°C and then shifting to 29°C after eclosion. After 3-5 days, ovaries were dissected from these females as described below. Each GAL4 line alone or GAL4, GAL80^TS line (with parallel temperature shifts) was stained and used for control pMad⁺ cell counts (Fig. 5 and Fig. S3, ‘GAL4 controls’).

To generate cell type-specific mitotic clones of brk ΔPPEprox, we recombined CRISPR-Cas9 deletions with FRT18A (BDSC #5245). We then crossed this recombinant, as well as previously made brk-FRT recombinants (see Table S1), to a UAS-FLP fly strain (BDSC #4539 or 4540) to generate stocks of the generic genotypes brk* (either gene null mutant or PPEprox deletion), FRT18A; UAS-FLP and y*, w*, ubi-GFP (BDSC #5245 or #5624), FRT18A; X-GAL4 (where X indicates either tj, nos or 109-30). Appropriate combinations were crossed to obtain germaria mosaic for brk in GAL4-expressing tissues. Although two different GFP markers were used to visualize clones, neither was strongly expressed enough to observe mosaicism consistently in germaria. We confirmed that our system was working to generate mosaics by observing later-stage egg chambers where the GFP marker was clearly visible and clones could be observed, indicating that recombination was occurring at earlier stages. For brk mutant clones, one of two null alleles were used (brk^XA or brk^KO), except in the case of nos-GAL4 experiments for which data from both mutants was pooled, i.e. ‘brk⁻’, as cell count distributions did not differ between mutants. Control crosses were performed using identical genotypes but lacking UAS-FLP (Fig. 5 and Fig. S3, ‘FRT controls’).

Ovaries were collected from 3- to 5-day-old flies aged at 25°C on standard food supplemented with yeast. Ovaries were dissected in cold EBR (0.13 M NaCl, 4.7 mM KCl, 1.9 mM CaCl₂, 10 mM HEPES, pH 6.9) and then fixed for 20 min at room temperature in PBS with 0.1% Tween 80 (PBT), 4% paraformaldehyde and 1% DMSO. Tissues were washed three times, 5 min each, in PBT and then incubated for 1 h in 1% Triton X-100 in PBS. Three more five-minute washes with PBT were followed by incubation for 1 h in 1× Western Blocking Solution (Millipore/Sigma, 11921673001) in PBT (WB). The tissues were incubated overnight at 4°C with primary antibody in WB. The following day the samples were washed three times, 10 min each, and then blocked in WB for 30 min. The tissues were incubated with secondary antibodies overnight at 4°C in WB. Finally, the tissues were washed three times, 10 min each, with PBT, incubated with DAPI in PBT for 30 min, washed a final two times, 5 min each, with PBT and mounted in 70% glycerol/30% PBS. HCR was carried out as described by Slaidina et al. (2020) using brk 4049/E190 and dpp 4049/E192 probes from Molecular Technologies, and mounted in Slow Fade Gold (Thermo Fisher, S36936).

Primary antibodies used in this study were: α-Spectrin (1:100, 3A9), Lamin C (1:50, LC28.26), Lamin (1:100, ADL84.12), Bam (1:10), En (1:20, 4D9), E-Cadherin (1:50, DCAD2) and Vasa (1:50) from the Developmental Studies Hybridoma Bank; pSMAD1/5 (1:50, Cell Signaling Technology, 9516), RFP (1:1000, MBL International, PM005), GFP (1:5000, Rockland Immunochemicals, 600-101-215) and Tj (1:5000, kindly provided by the Godt Lab; Li et al., 2003). Secondary antibodies (1:400) were conjugated to Alexa Fluor 488, Alexa Fluor 555 or Alexa Fluor 647 (Molecular Probes, Life Technologies) and used with DAPI (Invitrogen, D3571). All antibodies used have been previously published and were validated by confirming expected expression patterns in wild-type tissues.

All images were captured using a Zeiss LSM 800 confocal microscope and associated Zeiss microscope software (ZEN blue). All images, except for HCR experiments, were captured using a 20× objective, N.A 0.75. HCR images were captured with 63 ×oil objective, N.A. 1.4, using a scan speed of 7 and 8 times averaging.

All female flies of the appropriate genotype that eclosed within a 2-day window were collected, dissected and stained. In all cases this was at least three individuals (i.e. biological replicates). From these collections, all unobstructed germaria in which relevant structures were visible on the slide were imaged and quantified. Power analysis performed using the G*Power tool (version 3.1; Faul et al., 2007) on preliminary data collected for pMad⁺ cell number in PPE mutants suggested that a sample size of ∼11 samples per group would be sufficient to detect the differences in means (effect size ∼0.8) with high statistical confidence (1−β err prob=0.95). As more subtle effects were likely with misexpression and mosaic tissue-specific experiments, we more than doubled this n for all experiments in which pMad⁺ cell numbers were counted. Similar analyses indicated that our sample sizes were more than sufficient for other phenotypes quantified: H2A intensity and egg chamber number. For HCR analysis, acquisition of sufficiently high-resolution scans of ovarioles is time consuming and therefore this constraint determined sample size for these experiments. Still, we detected statistically significant differences between samples that appeared visually different, indicating that our sample sizes were sufficient to capture real differences in HCR signal. Operators were not blinded to sample genotype.

PPE H2A-mCherry reporter expression was quantified using Imaris software using the spot detection tool to find all spots with an estimated 3 μm diameter. The mean intensity for each spot was determined and the cap cells were manually defined using LamC staining. Undifferentiated germ cell counts were obtained from maximum projections of confocal stacks of germaria stained for pMad, as well as Spectrin and Lamin C to mark the niche. HCR signal for dpp and brk probes was quantified using Zen software. A maximum intensity projection of 16 z-stacks (0.5 µm each) was made and a region of the germarium encompassing regions 1-2a, but excluding the CCs, was drawn and measured using the spline contour tool. All pixels of a given wavelength with a signal intensity above a threshold of 30 were counted and the total count was divided by the area to give the final measurement. The threshold was chosen empirically by examining ten images from different genetic backgrounds and choosing the lowest intensity at which a spot could be detected by eye. This process was repeated for the CC analysis.

All statistical tests and data visualizations were performed using Prism software (version 9.1.0, www.graphpad.com). Reporter intensity, GSC number and egg chamber number were compared between genotypes or experimental conditions using ordinary one-way ANOVA, and HCR signal with Brown–Forsythe ANOVA, whereby each experimental condition was compared with the matched control (see Table S2). P-values were corrected for multiple comparisons using Dunnett's test. HCR quantification for dpp in the non-CC domain of ΔPPEprox germaria was additionally compared with wild type using an unpaired, two-tailed t-test. These data show a clear increase in dpp signal intensity compared with wild type that was not captured as statistically significant with our stringent P-value correction approach using ANOVA (see Fig. 4G ‘ns’). We provide the significant P-value resulting from this t-test in Table S2 to support our assertions in the text that the ΔPPEprox mutant has elevated dpp. Statistical significance of P-values (GraphPad format) is abbreviated on plots as follows: not significant (ns) where P≥0.05, significant (*) where 0.01≤P<0.05, very significant (**) where 0.001≤P<0.01, extremely significant (*** and ****) where 0.0001≤P<0.001 and P<0.0001, respectively. Error bars represent mean ±s.d. Box plots extend from the 25th to 75th percentiles with the horizontal line indicating the median and with whiskers indicating minimum and maximum values (Figs 4 and 6). See Table S2 for n (number of nuclei, ovarioles or germaria, as appropriate) and corrected P-values for all genotypes and conditions analyzed.

We thank Gerard Campbel, Dorothea Godt and George Pyrowolakis for sharing fly stocks and antibodies, and Peiwei Chen for comments on the manuscript.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199890.