Wingless promotes EGFR signaling in follicle stem cells to maintain self-renewal

Adult epithelial stem cells divide to produce differentiated progeny that replace lost or damaged cells in the tissue, or self-renewing progeny to replenish the population of stem cells. This asymmetry of cell fates is enforced by the adult stem cell niche microenvironment, which is essential for maintenance and self-renewal of the cell within that environment, ensuring that a consistent population of stem cells is available for regeneration of differentiated daughter cells. The mechanisms by which these signaling ligands are communicated between niche and epithelial stem cells with specificity, as well as how they maintain self-renewal robustly, are not fully understood.

The follicle epithelium of the Drosophila ovary is a tractable model that can be genetically manipulated to study epithelial stem cells in their native tissue environment (Sahai-Hernandez et al., 2012). The Drosophila ovary is composed of long strands of developing follicles, called ovarioles, and oogenesis begins at the anterior tip of each ovariole in a structure called the germarium. The germarium is divided into four regions (Regions 1, 2a, 2b and 3) that correspond to distinct stages of germ cell development (Fig. 1A). In Region 1, two somatic cell types, the cap cells and terminal filament cells, provide cues that regulate the proliferation and self-renewal of germline stem cells (GSCs) (Chen et al., 2011). GSC divisions produce cystoblasts that undergo four rounds of division with incomplete cytokinesis as they move downstream through the germarium to become 16-cell germline cysts. At this stage, referred to as Region 2a, two clearly identifiable 16-cell cysts are arranged side by side across the width of the germarium. In Regions 1 and 2a, the germ cells are surrounded by a population of somatic inner germarial sheath cells (IGS cells, also referred to as escort cells) that wrap around each cyst with long cytoplasmic processes and provide important germ cell differentiation cues. As germ cell cysts move from Region 2a to 2b, they shed the IGS cell layer, widen to span the entire width of the germarium, and become encapsulated by the follicle cells. Next, as the germ cell cyst moves further downstream into Region 3, it becomes more circular and the follicle cells organize into a single-layered epithelium. Many studies have confirmed the existence of follicle stem cells (FSCs) at the Region 2a/2b border (Chang et al., 2013; Margolis and Spradling, 1995; Nystul and Spradling, 2007; Reilein et al., 2017; Song and Xie, 2002), demarcated as the boundary between the two adjacent cysts in Region 2a and the first single-file cyst in Region 2b. A recent study suggested that additional FSCs or their progeny may also reside in Region 2a (Reilein et al., 2017), but we are focusing here on those at the Region 2a/2b border for consistency with previous studies on Wnt signaling in FSCs (Dai et al., 2017; Sahai-Hernandez and Nystul, 2013; Wang and Page-McCaw, 2014). FSC divisions give rise to prefollicle cells (pFCs) that go on to differentiate into main body follicle cells, which encapsulate each germline cyst to produce the follicle; polar cells, which provide signals to pattern the follicle; or stalk cells, which form the connections between consecutive follicles.

The Wnt and EGFR pathways function as necessary and specific FSC niche signals. Both pathways are active in FSCs and required for self-renewal, but must be downregulated in pFCs that have moved downstream from the niche to allow for differentiation (Castanieto et al., 2014; Sahai-Hernandez and Nystul, 2013; Song and Xie, 2003; Wang and Page-McCaw, 2014). A recent study demonstrated that Wnt signaling is also active in early pFCs at the Region 2a/2b border, where it may function to preserve the capacity of these cells to differentiate toward the polar and stalk cell fates (Dai et al., 2017). Taken together, these findings demonstrate that there is a steep gradient of Wnt and EGFR signaling at the Region 2a/2b border, and that this pattern is important to specify the FSC and pFC fates. However, it is unclear how this pattern of signaling is maintained.

Here, we demonstrate that the specificity of Wnt signaling in the FSC niche at the Region 2a/2b border is due to spatial restriction of the Wnt ligand, Wingless (Wg), to the FSC niche. In addition, we find that Wnt signaling is required for EGFR pathway activity and that constitutive activation of EGFR partially rescues the self-renewal defect caused by loss of Wnt signaling. Lastly, we show that Wnt signaling is required for expression of the EGFR ligand, Spitz (Spi), in the early FSC lineage, and that Spitz promotes EGFR signaling in FSCs. Collectively, these results support a model in which Wg and EGFR are activated specifically in stem cells by short-range signaling, and function in a signaling hierarchy to promote FSC self-renewal.

In canonical Wnt signaling, binding of a Wnt ligand to a Frizzled receptor results in recruitment of the positive regulator, disheveled (dsh), and the inhibition of the destruction complex, which limits pathway activity by targeting β-catenin for ubiquitin-mediated degradation. Thus, inhibition of the destruction complex allows β-catenin to accumulate in the nucleus, where it interacts with TCF (Pan – FlyBase) and regulates target gene expression in a cell type-specific manner (Clevers et al., 2014; van Amerongen and Nusse, 2009). Several reporter lines have been constructed that allow for monitoring of Wnt signaling in vivo. Two lines, notum-lacZ and frizzled3-RFP (fz3-RFP), that report the expression of a canonical Wnt pathway target gene as an indicator of pathway activity have been shown to have similar expression patterns in the germarium, although fz3-RFP is expressed more broadly and consistently (Dai et al., 2017; Sahai-Hernandez and Nystul, 2013; Wang and Page-McCaw, 2014). Both reporters are expressed strongly in FSCs and then rapidly downregulated in pFCs that have moved downstream from the FSC niche. However, whereas fz3-RFP is expressed consistently throughout all IGS cells but not in cap or terminal filament cells (Fig. S1A,B), notum-lacZ is expressed only sporadically in terminal filament cells, cap cells, outer muscle sheath cells and IGS cells (Dai et al., 2017; Sahai-Hernandez and Nystul, 2013; Wang and Page-McCaw, 2014). To confirm the pattern of Wnt signaling in the FSC lineage, we investigated the expression pattern of a recently developed synthetic reporter, 3×GRH-4TH-GFP (Zhang et al., 2014). This reporter contains four TCF binding sites (4TH) that facilitate TCF-driven GFP expression upon pathway activation in combination with three grainyhead binding sites (3×GRH) that repress GFP expression in the absence of Wnt pathway activity. This combination makes the reporter both highly sensitive to, and specific for, activation by TCF. We found that 3×GRH-4TH-GFP is detectable in posterior IGS cells and FSCs at the Region 2a/2b border, identified as either the anteriormost FasIII⁺ cells (Fig. 1B) or the anteriormost cell in a negatively marked FSC clone (Fig. S1C), but not in cap or terminal filament cells, germ cells or pFCs in Region 2b or beyond. Thus, all three reporters indicate that there is a sharp drop in Wnt signaling at the Region 2a/2b border, with clearly detectable levels in the FSCs at the anterior edge of the FasIII border, and little or no signal in the pFCs immediately downstream.

We demonstrated previously that IGS cells are the predominant source of Wg for Wnt signaling in FSCs (Sahai-Hernandez and Nystul, 2013). To confirm this observation with the other two reporters, we combined 3×GRH-4TH-GFP or fz3-RFP with UAS-wg RNAi, tub-Gal80^ts, and either 13C06-Gal4, which is expressed in IGS cells, FSCs and pFCs near the Region 2a/2b border, or the follicle cell driver 109-30-Gal4 (Hartman et al., 2010). We grew the flies at 18°C to inhibit Gal4 activity during development, and then shifted adult flies to 29°C for 7 days to inhibit Gal80 and drive expression of wg small interfering RNA (siRNA). We found that expression of wg siRNA with 109-30-Gal4 caused only a mild reduction in the level of 3×GRH-4TH-GFP, and had no effect on fz3-RFP expression (Fig. S2A-B′). In contrast, expression of wg siRNA with 13C06-Gal4 eliminated expression of 3×GRH-4TH-GFP and fz3-RFP in the FSC niche region (Fig. 1C, Fig. S2C-D′). Interestingly, expression of wg siRNA with 13C06-Gal4 substantially reduced fz3-RFP expression but did not eliminate the signal entirely, consistent with reports that there are other Wnt ligands that promote Wnt signaling in IGS cells (Upadhyay et al., 2016, 2018; Wang and Page-McCaw, 2018; Wang et al., 2015). These observations indicate that the 3×GRH-4TH-GFP reporter is responsive to Wnt signaling in the germarium, as expected, and provide confirmation of our previous finding that Wg produced by IGS cells is required for Wnt signaling in FSCs.

We considered several possibilities for how a steep gradient of Wnt signaling could be maintained at the Region 2a/2b border. First, we investigated whether pFCs express Wnt pathway components and are capable of transducing a Wnt signal. β-catenin is ubiquitously expressed in epithelial cells, and several studies have reported the expression of Armadillo (Arm, the Drosophila homolog of β-catenin) protein in follicle cells (Kronen et al., 2014; Song and Xie, 2002). We detected transcripts of dsh, frizzled2 (fz2) and arm by fluorescence in situ hybridization (FISH), as well as Fz2 protein in the FSCs and all follicle cells in the germarium (Fig. 1D-G′, Fig. S1D). Thus, both FSCs and pFCs express key positive regulators of Wnt signaling. In addition, we confirmed that pFC clones that were homozygous mutant for an essential component of the destruction complex, Axin (Axn), exhibited a multilayering phenotype (Fig. 1H) (Dai et al., 2017; Song and Xie, 2003). As discussed further below, we found that overexpression of wg causes a similar multilayering phenotype and the activation of the Wnt pathway reporter, 3×GRH-4TH-GFP, in pFCs (Fig. 3H-I′). Taken together, these findings indicate that the specificity of Wnt signaling for the FSC niche is not due to an inability of pFCs to transduce a Wnt signal.

Second, we investigated whether the pattern of Wnt signaling in the FSC lineage is caused by interactions with other pathways or by the expression of other Wnt ligands, which could have either activating or inhibitory effects. Both EGFR and Notch signaling are known to interact with Wnt signaling in other tissues (Cordero et al., 2012; Freeman and Bienz, 2001; Hing et al., 1994; Hu and Li, 2010; Nagaraj and Banerjee, 2009; Szüts et al., 1997) and have spatially restricted activity in the FSC lineage (Castanieto et al., 2014; Sahai-Hernandez and Nystul, 2013). In addition, EGFR signaling is required for self-renewal (Castanieto et al., 2014), and Notch signaling cooperates with Wnt signaling to pattern pFC differentiation (Dai et al., 2017). Thus, we tested whether perturbations of either pathway altered the pattern of Wnt signaling. We found that expression of EGFR siRNA with 109-30-Gal4 substantially reduced EGFR protein levels (Fig. S6B), but did not change the pattern of 3×GRH-4TH-GFP expression (Fig. 2A-B′). Likewise, RNAi knockdown of Notch or constitutive activation of EGFR in follicle cells caused the morphological defects expected from previous reports (Castanieto et al., 2014; Johnston et al., 2016; Lopez-Schier and St Johnston, 2001), but did not affect the pattern of 3×GRH-4TH-GFP expression (Fig. 2C-D′).

There are seven known Wnt ligands in the Drosophila genome (Swarup and Verheyen, 2012), including Wg, and three (Wnt2, Wnt4 and Wnt6) have been recently found to be important for Wnt signaling in anterior IGS cells (Hamada-Kawaguchi et al., 2014; Upadhyay et al., 2016; Wang et al., 2015). Of these, Wnt2 and Wnt4 are required in IGS cells (Upadhyay et al., 2016, 2018; Wang et al., 2015), and Wnt6 is required in cap and terminal filament cells (Wang and Page-McCaw, 2018). To test whether any other Wnt ligand besides Wg affects Wnt signaling in the FSC, we performed RNAi knockdown of each ligand using 13C06-Gal4. RNAi knockdown of Wnt3/5, Wnt6, Wnt8 or Wnt10 did not cause an observable follicle cell phenotype (Fig. S3A-D), although we could not confirm that the knockdown was effective in these cases. Knockdown of Wnt2 or Wnt4 caused germ cell differentiation defects, as reported previously, but did not cause an observable follicle cell phenotype (Fig. S3E-I). Interestingly, we found that Wnt4 knockdown, but not Wnt2 knockdown, also diminished expression of 3×GRH-4TH-GFP (Fig. S3J-L′), indicating that wg and Wnt4 have nonredundant roles in activating Wnt signaling to a level that is detectable with this reporter. However, because Wnt4 knockdown did not cause a follicle cell phenotype, it is likely that Wnt signaling in FSCs was not abrogated entirely. These data indicate that the pattern of Wnt signaling in the early FSC lineage is not dependent upon EGFR or Notch pathway activity, and suggest that Wnt pathway is activated in FSCs primarily by Wg, with more minor contributions from Wnt4.

Third, we investigated whether the pattern of Wnt signaling in the FSC lineage is due to the restriction of Wg movement through the tissue. wg transcript is detectable in terminal filament cells, cap cells and IGS cells, but protein levels, as assayed by immunofluorescence, are highest at the anterior tip of the germarium and taper off toward the posterior, becoming undetectable by Region 2b (Fig. S4A) (Sahai-Hernandez and Nystul, 2013; Wang and Page-McCaw, 2014). To determine whether Wg produced in the anterior half of the germarium can move beyond the Region 2a/2b border, we simultaneously overexpressed hemagglutinin (HA)-tagged Wg (HA::wg) and CD8::GFP with 13C06-Gal4 using tub-Gal80^ts to restrict expression to adulthood. We co-stained ovaries from control and experimental groups for HA and GFP, and found that, whereas the HA signal was diffuse and nonspecific in control ovaries without a Gal4 driver (Fig. 3A), bright HA::Wg signal was clearly detectable in the experimental ovaries and almost entirely confined to the 13C06-Gal4-expressing cells, as identified by CD8::GFP expression, with little or no signal visible beyond the boundary of this domain (Fig. 3D, Fig. S4B). In addition, similar to control ovaries (Fig. 3B-C′), overexpression of HA::wg with 13C06-Gal4 did not cause pFC overproliferation or differentiation defects (Fig. 3E, Fig. S4C), and 3×GRH-4TH-GFP was not activated in pFCs (Fig. 3F, Fig. S4B). In contrast, overexpression of HA::wg during adulthood with 109-30-Gal4 produced detectable levels of HA::Wg specifically within the 109-30-Gal4 expression region (Fig. 3G), and resulted in severe morphological phenotypes in the follicle epithelium that phenocopied the loss of components of the destruction complex (Fig. 1H) (Song and Xie, 2003). Specifically, overexpression of HA::wg in follicle cells caused the germaria to become substantially larger as follicles fused together and failed to bud properly, and follicle cells to accumulate in disorganized clusters (Fig. 3H, Fig. S4C). In addition, the 3×GRH-4TH-GFP reporter was ectopically activated in pFCs (Fig. 3I,I′). These observations provide further evidence that pFCs are capable of activating Wnt signaling and demonstrate that HA::Wg is functional. Taken together, these data indicate that Wg ligand is spatially restricted, such that, even in the overexpression case, Wg produced in IGS cells does not activate Wnt signaling beyond the Region 2a/2b border.

A recent study proposed that cap and terminal filament cells are an important source of Wg for the FSC niche (Wang and Page-McCaw, 2014). This conclusion was based on the observation that overexpression of wg using the cap and terminal filament cell driver, bab1-Gal4, resulted in an overproduction of stalk cells, whereas expression of wg siRNA with heat shock-Gal4 (hs-Gal4) caused a fused follicle phenotype that is suggestive of an underproduction of follicle cells. However, bab1-Gal4 is also expressed during ovarian development, and hs-Gal4 is expressed in all cells upon heat shock, so these experiments lacked either spatial or temporal precision. In addition, this conclusion is inconsistent with our previous finding that RNAi knockdown of wg from cap and terminal filament cells specifically during adulthood using bab1-Gal4 and tub-Gal80^ts is not sufficient to cause a follicle cell phenotype or eliminate notum-lacZ expression (Sahai-Hernandez and Nystul, 2013). Therefore, to investigate further, we repeated the knockdown or overexpression of wg using bab1-Gal4 and tub-Gal80^ts and looked for the phenotypes described in this study. We found that overexpression of HA::wg with bab1-Gal4 specifically during adulthood produced clearly detectable levels of HA::Wg at the anterior tip of the germarium, indicating that the protein was expressed in the expected region, but did not cause an overproduction of stalk cells or any other detectable follicle cell phenotype (Fig. 4A-B′). In addition, RNAi knockdown of wg specifically during adulthood using bab1-Gal4 did not cause follicle formation defects, and did not affect fz3-RFP expression in IGS cells or FSCs (Fig. 4C).

As an additional test of the range of Wg in the germarium, we examined the ovaries of Nrt::wg flies, in which the endogenous wg gene has been replaced with a gene encoding for a fusion of the transmembrane protein Neurotactin (Nrt) and Wg (Alexandre et al., 2014). This targets the fusion protein to the plasma membrane so that its ability to activate Wnt signaling in nonadjacent cells is substantially reduced or eliminated. Nrt::wg homozygous flies are viable and grow to adulthood with normally patterned wings, although they are smaller and substantially less fit than their wild-type siblings (Alexandre et al., 2014). In addition, we found that the ovaries of Nrt::wg homozygotes were much smaller than those of their heterozygous siblings, and that homozygotes laid fewer eggs, many of which did not hatch into larvae. However, the germaria were morphologically normal, with pERK⁺ (Rl⁺) FSCs (Fig. 4D) and a normal pattern of expression of the differentiation factors castor (cas) and eyes absent (eya) (Fig. 4E) (Chang et al., 2013). In addition, we observed 5-ethynyl-2′-deoxyuridine (EdU) incorporation into FSCs and pFCs, indicating that these cells are proliferative (Fig. 4F). These findings confirm that the cap and terminal filament cells are not a major source of Wg for the FSC niche. Taken together, our data strongly suggest that the specificity of Wnt signaling in FSCs is due to a short-range Wg signal from 13C06-Gal4-expressing cells, combined with spatial restriction of Wg movement across this region.

We found previously that EGFR signaling is necessary for FSC maintenance, and that pERK, a marker of EGFR signaling, is detectable in IGS cells, FSCs and pFCs along the Region 2a/2b border, but rarely in pFCs in Region 2b (Castanieto et al., 2014). In addition, we have since noticed that pERK is also detectable sporadically in Region 3 follicle cells. To determine whether Wnt signaling is required for EGFR signaling, we generated FSC clones that are homozygous mutant for arm⁸, arm^F or dsh³. arm⁸ is a hypomorph that impairs Wnt signaling but not cell-cell adhesion (Llimargas, 2000), and arm^F and dsh³ are null alleles (Haelterman et al., 2014; Perrimon and Mahowald, 1987; Yamamoto et al., 2014). We found that the pERK signal was diffuse and indistinguishable from background in most arm or dsh mutant (GFP^–) FSCs, whereas a nuclear pERK signal was clearly detectable in nearly all of the heterozygous (GFP⁺) FSCs in the same germaria (Fig. 5A-B‴,D, Fig. S5). FSCs that are homozygous mutant for arm or dsh are lost from the niche at an increased rate (Song and Xie, 2003), so it is possible that the loss of pERK in these cells is due to a general decrease in receptivity to niche signals prior to their exit from the FSC niche at the Region 2a/2b border. However, homozygosity for Axn^S044230, which causes constitutive activation of Wnt signaling, also causes an increased rate of FSC loss (Song and Xie, 2003), and yet we found that these mutant FSCs typically remained pERK⁺ (Fig. 5C-D). This suggests that the loss of pERK in arm or dsh mutant FSCs is not a general feature of mutations that increase the rate of FSC loss from the niche, but is instead specifically caused by decreased Wnt signaling. As an additional test, we assayed for pERK in germaria upon RNAi knockdown of arm in the FSCs and all follicle cells in the germarium using 109-30-Gal4, and found that pERK was undetectable in FSCs in the majority of germaria (Fig. 5E-F). Collectively, these data strongly suggest that Wnt signaling is required for EGFR signaling in FSCs.

Because both EGFR and Wnt signaling are required for FSC self-renewal (Castanieto et al., 2014; Song and Xie, 2003), we considered whether reduced EGFR signaling contributes to the loss of self-renewal in Wnt pathway mutants. To test this possibility, we performed standard FSC self-renewal assays (Kronen et al., 2014; Song and Xie, 2003), in which sparse wild-type or mutant FSC clones are induced during adulthood, and the change in clone frequency is quantified over time. We used the mosaic analysis with a repressible cell marker (MARCM) system (Lee and Luo, 2001) to produce clones that are positively marked by the expression of CD8::GFP, and quantified the change in clone frequency at 7, 14 and 21 days after clone induction (Fig. 6A). Consistent with previous reports (Song and Xie, 2003), the proportion of germaria with dsh³ FSC clones decreased at significantly higher rates compared with the wild-type control (Fig. 6B). Interestingly, overexpression of the constitutively active Egfr^λtop allele in dsh³ clones restored the pERK signal (Fig. 6C) and significantly decreased the rate of FSC clone loss (Fig. 6B), although the frequencies of Egfr^λtop, dsh³ double mutant clones were still lower than those of the wild-type control. These findings indicate that the signaling steps between EGFR activation and ERK phosphorylation do not require Wnt signaling, and that restoration of EGFR signaling can partially compensate for the self-renewal defect in dsh³ mutant FSCs.

To investigate how Wnt signaling functions genetically upstream of EGFR signaling, we first tested whether Wnt signaling is required for expression of EGFR. We found that EGFR is expressed throughout the FSC lineage in wild-type germaria (Fig. S6A-B′) and that this expression pattern was not affected by RNAi knockdown of arm or dsh (Fig. S6C-D′). Thus, Wnt signaling is not required for the expression of EGFR in FSCs or pFCs.

Next, we investigated whether Wnt signaling is required for the expression of the EGFR ligand, spitz. spitz is known to be expressed in the follicle cell lineage (Wasserman and Freeman, 1998), and we confirmed that spitz is expressed in follicle cells by FISH and with a spitz-lacZ enhancer trap (Fig. 7C, Fig. S7A-B′). In addition, we found that expression of spitz siRNA with 109-30-Gal4 eliminated pERK staining in a majority of FSCs and pFCs within Region 2b (Fig. 7A-B′,E), although it did not affect the pERK signal in follicle cells in Region 3 and later, where EGFR is likely activated by another EGFR ligand, such as gurken from the germline (Wasserman and Freeman, 1998). These results confirm that spitz is produced by cells in the FSC lineage and indicate that it is important for EGFR pathway activity in FSCs. In the intestine, Wnt operates via dMyc (Myc) to activate spitz (Cordero, et al., 2012). However, we found that RNAi knockdown of dMyc did not eliminate pERK in FSCs (Fig. S7C-D′), suggesting that this mechanism is not operating in the FSC lineage. Instead, we noticed that the chromatin immunoprecipitation with DNA sequencing (ChIPseq) data available on ModMINE (Contrino et al., 2012) indicate that there is a high confidence TCF binding site in the spitz gene that overlaps with the transcription start site of two spitz isoforms. We searched this sequence and identified three putative TCF-binding sites that are highly conserved across dipteran insect species (Fig. S7E,F), suggesting that spitz might be a direct target of TCF. Consistent with this, we found that expression of arm siRNA with 109-30-Gal4 substantially reduced spitz-lacZ expression in the FSCs and pFCs of most germaria (Fig. 7C-D′,F). Taken together, these results suggest a hierarchical structure of niche signals in which Wnt signaling promotes the activation of EGFR signaling in FSCs by inducing expression of spitz in FSCs and early pFCs.

In this study, we investigated the relationship between the Wnt and EGFR pathways in FSCs. Both pathways are known to be essential for FSC self-renewal and to inhibit pFC differentiation (Castanieto et al., 2014; Dai et al., 2017; Song and Xie, 2003), but how these pathways are regulated in the early FSC lineage is not well understood. Our data support a model in which the Wnt and EGFR pathways function in a signaling hierarchy, beginning with a short-range Wg signal from IGS cells that activates Wnt signaling in FSCs, which promotes expression of spitz in the early FSC lineage. This, in turn, leads to the activation of EGFR signaling in FSCs, which cooperates with Wnt signaling to promote FSC self-renewal. Several lines of evidence support this model. First, we have now shown with three separate Wnt reporters, two in this study and one previously (Sahai-Hernandez and Nystul, 2013), that Wg production by IGS cells is required for Wnt signaling in FSCs. As this is in contrast to a recent study that proposed a distant source of Wg for the FSC niche (Wang and Page-McCaw, 2014), we performed several additional tests of the range of Wg signaling in the germarium. We found that FSC niche function is not disrupted when Wg is tethered to the membrane, and that, even upon overexpression of HA::wg, both the pattern of Wnt pathway activity and the location of HA-tagged Wg are only detected within or very close to the expression domain of the Gal4 driver used. In contrast, overexpression of Wg in follicle cells causes significant pFC differentiation defects. This indicates that Wg movement in the germarium is substantially restricted and suggests that a short-range Wg signal is both necessary to allow for pFC differentiation and sufficient to promote FSC self-renewal. Although further study will be required to fully elucidate the mechanism by which the steep gradient of Wnt signaling at the Region 2a/2b border is maintained, our findings suggest that interactions with the EGFR and Notch pathways, or differences in the capacity of FSCs and downstream pFCs to transduce a Wnt signal, are not major factors in the patterning of Wnt activity within the early FSC lineage. Instead, recent studies suggest that proteins that interact with Wnt ligands in the extracellular space restrict the movement of Wnt ligands into the follicle epithelium. Specifically, RNAi knockdown of Matrix metalloproteinase 2 during adulthood in cap, terminal filament and IGS cells causes follicle cell phenotypes (Wang and Page-McCaw, 2014), and RNAi knockdown of heparan sulfate proteoglycans using 109-30-Gal4 reduces fz3-RFP expression in FSCs (Su et al., 2018).

Second, we found that effectors of the Wnt pathway are required for EGFR signaling in FSCs. Our data suggest that this is due, in part, to a role for Wnt signaling in promoting the expression of spitz in the early FSC lineage. Indeed, we found that knockdown of arm substantially reduced spitz-lacZ expression in the follicle cells within the germarium, and knockdown of spitz eliminated the pERK signal in most FSCs. However, the failure of spitz-RNAi expression to eliminate pERK in all FSCs suggests that either the RNAi knockdown was incomplete or that there are other sources of EGFR ligand for the FSC niche. It is interesting that knockdown of arm decreased spitz-lacZ expression in follicle cells throughout the germarium, even though Wnt signaling is not normally active downstream of the Region 2a/2b border. This suggests that Wnt signaling is required in FSCs for both the activation of spitz expression in FSCs, as well as the continued spitz expression during the early stages of pFC differentiation. Consistent with this possibility, we and others have identified signaling genes that are required in FSCs, but not pFCs, for pFC differentiation (Castanieto et al., 2014; Li et al., 2010), suggesting that some signals that act on FSCs have effects that persist into the daughter cells. Alternatively, it could be that other functions of arm, such as its role in cell-cell adhesion, contribute to the regulation of spitz in pFCs. However, knockdown of arm did not affect spitz-lacZ expression in budded follicles, indicating that spitz expression becomes independent of arm function at that point.

Third, we tested the functional relationship between Wnt and EGFR signaling in FSC self-renewal and found that constitutive activation of EGFR signaling significantly decreases the rate at which dsh³ FSCs are lost from the niche. Although we cannot rule out that activation of EGFR signaling circumvents the loss of Wnt signaling in a manner that is unrelated to the relationship between these two pathways in wild-type FSCs, the most straightforward interpretation of this result is that Wnt signaling functions upstream of EGFR signaling to promote FSC self-renewal. However, Wnt signaling is likely to be important for other aspects of self-renewal as well. For example, a previous study found that overexpression of a constitutively active allele of the BMP receptor thickveins decreases the rate of dsh³ FSC loss (Kirilly et al., 2005), suggesting that Wnt signaling might also function upstream of BMP signaling to promote self-renewal.

Our findings here and previous studies (Wang and Page-McCaw, 2014) indicate that fz3-RFP expression is broader and more consistent than notum-lacZ or 3×GRH-4TH-GFP expression in the germarium. Given the importance of Wnt signaling as an upstream activator of the FSC self-renewal program, it is likely that Wnt signaling is, in fact, active in all FSCs, as suggested by the fz3-RFP reporter. However, RFP has a relatively long half-life (Mirabella et al., 2004), so the variation in notum-lacZ and 3×GRH-4TH-GFP expression could be revealing fluctuations in pathway activity that are not easily detectable with fz3-RFP. Interestingly, constitutive activation of Wnt signaling actually impairs FSC self-renewal (Song and Xie, 2003), suggesting that a dynamic pattern of Wnt signaling might be required to maintain the FSC self-renewal program. In addition, fluctuations in Wnt signaling within FSCs might also be important for specifying the fate of the pFCs as they are produced (Dai et al., 2017). It is currently unclear what causes this dynamic pattern of Wnt pathway activation, but one possibility suggested by our finding that Wnt signaling in FSCs is activated by Wg from IGS cells, which undergo continuous rearrangements to facilitate the passage of germ cell cysts through the anterior half of the germarium (Morris and Spradling, 2011), is that the movement of IGS cells influences the amount of Wg ligand available to the FSC niche. This might be important for coordinating FSC niche output with demand from germ cells in the early stages of oogenesis. At the same time, the hierarchy of niche signals we have discovered here could help to ensure that FSC identity is maintained during brief periods of low Wnt signaling. Further studies into the fluctuation of Wnt signaling in relation to other signals in FSCs and the movement of IGS cells will continue to provide insights into how this dynamic niche functions.

Stocks were maintained on standard molasses food at 25°C and adults were given fresh wet yeast daily. All progeny containing tub-Gal80^ts were kept at 18°C until eclosion and then shifted to 29°C for 7-10 days for high UAS expression, unless indicated otherwise. The following stocks were used, listed by figure according to the first occurrence (stocks with ‘BL’ numbers were provided by Bloomington Drosophila Stock Center):

Fig. 1: 3×GRH-4TH-EGFP (from Dr Ken Cadigan, University of Michigan, Ann Arbor, MI, USA); w1118; P{GMR13C06-Gal4}attP2 (BL#47860); UAS-wg RNAi (TRiP.HMS00844, BL#33902); FRT82B AxnS044230/TM3 Sb (from Dr Ting Xie, University of Kansas, Lawrence, KS, USA); hsFlp; FRT82B, Ubi-GFP/TM3, Sb (Nystul Laboratory, University of California, San Francisco, CA, USA); and Canton S (BL#64349). Fig. 2: UAS-EGFR RNAi (TRiP.JF01696, BL#31183); UAS-Notch RNAi (TRiP.JF01356, BL#31383); and UAS-EGFR^λtop/TM6C, Sb (BL#59843). Fig. 3: y1, w*; P{GawB}109-30/CyO; (BL#7023) and UAS-HA::wg (BL#5918). Fig. 4: w*; P{GawB}bab1[Pgal4-2]/Tm6b, Tb1 (BL#6803); fz3-RFP (from Dr Ramanuj DasGupta, Genome Institute of Singapore, Singapore); and Nrt::wg [in Wg KO]/CyO; (from Dr Jean Paul Vincent, The Francis Crick Institute, London, UK). Fig. 5: arm^F FRT19a/FM7c; (BL#57054); w*, dsh³ P{neoFRT}19A/FM7a; (BL#6331); w122 (hsFlp), Ubi-GFP, FRT19a; MKRS/TM3 (from Dr Ben Ohlstein, Columbia University, New York, NY, USA); and UAS-arm RNAi (TRiP.HMS01414, BL#35004). Fig. 6: hsFlp, tsGal80 FRT19a/FM7; Act-Gal4, UAS-CD8 GFP/cyo (Nystul Laboratory) and P{neoFRT}19A; (BL#1709). Fig. 7: P{lacW}spi^s3547/cyo; (BL#10462) and UAS-spi RNAi; (BL#34645). Supplementary Figs: UAS-Wnt2 RNAi (TRiP.JF03377, BL#29441); UAS-Wnt4 RNAi (TRiP.JF03378, BL#29442); UAS-Wnt3/5 RNAi (TRiP.HMS0119, BL#34644); UAS-Wnt6 RNAi (TRiP.HM05236, BL#30493); UAS-Wnt8 (TRiP.HM05158, BL#28947); UAS-Wnt10 (TRiP.JF03423, BL#31989); UAS-dsh RNAi (TRiP.JF01253, BL#31306); UAS-dMyc RNAi (TRiPJF01761, BL#25783); Arm⁸, FRT101/FM7a; (BL#8557); bab1[Agal4-5]/TM3 (sb) (BL#6802); and tsGal80; TM2/TM6 (Nystul Laboratory, University of California, USA).

Ovaries were dissected in 1× phosphate buffered saline (PBS), fixed in 1× PBS+4% formaldehyde for 15 min, rinsed with 1× PBS+0.2% Triton X-100 (PBST) and blocked for 1 h with 1× PBST containing 0.5% bovine serum albumin (BSA). Samples were incubated with primary antibodies diluted in blocking solution overnight at 4°C. Next, samples were rinsed with PBST and blocked for 1 h before incubating with secondary antibodies for 4 h at room temperature. Samples were rinsed twice with PBST and once with PBS before a final 30 min wash with PBS. For the actin stain in Fig. 1E, Alexa Fluor 488-conjugated phalloidin (Thermo Fisher Scientific, A12379) was added to the last PBS wash at 1:1000 for 30 min, removed, and then the samples were washed a second time with PBS for 30 min. Samples were mounted on glass slides in Vectashield (Vector Laboratories) with 4′,6-diamidino-2-phenylindole (DAPI). For the anti-Wg stain in Fig. S4A, ovaries were dissected in cold 1× PBS and incubated with anti-Wg antibody [1:4; Developmental Studies Hybridoma Bank (DSHB), 4D4], fixed in 1× PBS+4% formaldehyde for 15 min, permeabilized with PBST and blocked for 1 h. Next, samples were incubated with secondary antibody at 1:1000 in block for 4 h at room temperature. Samples were rinsed twice with PBST, blocked for ∼15 min and washed a final time with PBS for 15 min. Samples were mounted on glass slides in Vectashield with DAPI.

The following primary antibodies were used: guinea pig anti-GFP (1:1000; Synaptic Systems, 132005), mouse anti-beta-Galactosidase (1:1000; Promega, Z3781), rabbit anti-Vasa (1:1000; Santa Cruz Biotechnology, sc-30210), mouse anti-Fz2 (1:50; DSHB, 12A7), mouse anti-FasIII (1:50; DSHB, 7G10), rabbit anti-RFP (1:1000; MBL International, PM005S), rat anti-RFP (1:1000; ChromoTek, 5F8), mouse anti-Lamin C (1:100; DSHB, LC28.26), mouse anti-Hts (1:50; DSHB, 1B1) mouse anti-Arm (1:4; DSHB, N27A1), mouse anti-Eya (1:50; DSHB, 10H6); rabbit anti-pERK (1:100; Cell Signaling Technology, 4370), rabbit anti-Castor [1:5000; from Ward Odenwald (Kambadur et al., 1998)], rabbit anti-HA (1:1000; Invitrogen, 71-5500) and mouse anti-EGFR (1:1000; Sigma Aldrich, E2906). The following secondary antibodies were purchased from Thermo Fisher Scientific and used at 1:1000: goat anti-guinea pig 488 (A-11073), goat anti-rabbit 488 (A-11008), goat anti-rabbit 555 (A-21428), goat anti-mouse 488 (A-11029), goat anti-mouse 555 (A-21424) and goat anti-rat 555 (A-21434).

All fixed images were acquired using a Zeiss M2 Axioimager with Apotome unit. For multicolor fluorescence images, each channel was acquired separately. Postacquisition processing, such as image rotation, cropping, and brightness or contrast adjustment, were performed using ImageJ (National Institutes of Health) and Adobe Photoshop. Acquisition settings and any brightness/contrast adjustments were kept constant across conditions within an experiment.

For EdU incorporation experiments, ovaries were dissected and incubated in Schneider's Medium containing 20 µM EdU (Click-it EdU Alexa Fluor 555 Imaging Kit, Life Technologies, C10338) at room temperature for 1 h. Ovaries were then fixed in 4% paraformaldehyde (PFA) for 15 min and washed twice in 1× block (0.5% BSA in PBST). Wash solution was removed and samples were permeabilized with 1× PBST (0.2% Triton X-100 in 1× PBS) for 20 min. Samples were then washed twice with block and then incubated in reaction cocktail (1× Click-iT EdU reaction buffer, CuSO₄, Alexa Fluor azide, 1× Click-iT EdU buffer additive) for 30 min at room temperature, protected from light. Finally, tissues were rinsed with 1× PBS and then stained with anti-FasIII primary antibody as described above before mounting on glass slides in Vectashield.

Flies were fed wet yeast for at least 2 days prior to dissection. The following protocol was performed in RNAse-free conditions. At least 10-20 well-fed flies were dissected in cold 1× Ephrussi-Beadle Ringer's solution (EBR) on a chilled glass plate on ice. Muscle sheath was removed and ovarioles separated while keeping the entire ovary intact. Ovaries were moved to sterile RNAse-free tubes containing 4% PFA in PBS and fixed for 1 h at room temperature. After discarding fix, samples were washed in 1× PBS, then in 50% MeOH in PBS, then 100% MeOH. Samples were then stored in 100% MeOH at −20°C.

Dehydrated samples were taken out of the freezer and rinsed with 100% MeOH, washed with 50% MeOH in EtOH for 5 min, then rinsed with 100% EtOH. Leaving about 100 µl EtOH, the EtOH was removed and replaced with 900 µl xylene. Samples were permeabilized with xylene for 1 h at room temperature. Samples were rinsed twice with EtOH, once with 50% MeOH/EtOH and twice with 100% MeOH. Samples were incubated in 50% MeOH/Fix (5% PFA in PBST) for 5 min, rinsed once with Fix solution and then fixed for 1 h at room temperature. Samples were rinsed three times with PBST, washed for 10 min in PBST and then washed for 10 min in 50% PBST with HYB solution (50% vol/vol formamide, 5× saline-sodium citrate, 0.1% vol/vol of 50 µg/ml heparin, 100 µg/ml transfer RNA and 100 µg/ml sheared, boiled salmon sperm DNA). Samples were incubated with 100% HYB solution in a 60°C water bath overnight.

Then, 400 ng RNA probe was diluted into 50 µl HYB. Probes were heated to 83°C for 2-3 min. Tubes were spun down and put on ice for 5 min, after which 50 µl probe was applied to samples and incubated overnight (up to 24 h) at 60°C overnight.

Samples were washed with preheated HYB over the course of 2 h, switching out washing solution at regular intervals. Samples were rocked at room temperature with 50% HYB/PBST for 5 min. Samples were washed with PBST over the course of 1 h, switching out washing solution at regular intervals. Samples were blocked with 1× Roche blocking reagent in PBST for 1 h at room temperature. Samples were then incubated with pre-adsorbed antibody against Dig diluted 1:500 in 1× block at 4°C overnight.

Primary antibody was removed and samples were rinsed twice with PBST. Samples were then washed over the course of 1 h, switching out washing solution at regular intervals. Samples were blocked for 1 h and then incubated with goat anti-sheep 555 secondary antibody diluted 1:500 in block for 1.5 h at room temperature away from light. Secondary antibody was removed and samples rinsed twice with PBST. Samples were washed in the dark over the course of 2 h, switching out washing solution at regular intervals. The last couple of washes were done with PBS instead of PBST. Samples were removed from wash, mounted on slides in Vectashield containing DAPI and imaged as for the immunostaining protocol.

The following probes were used (lower-case letters indicate the primers used to amplify the complementary DNA template in the antiparallel direction): armadillo, 5′-ggaacgttgcctctctgtgtCTTGGAAGGCGCGCATAAGCAGTCGCACCAGATGGTGGATGGCCCCGTGCTCCCGCAACGGGGCGTGATTGGCCGGACAGAGGGCCAAATTGCGTATGAGTCCAATGACGGCCTTGATCAAGGGCCAGCGTGATGGTGGATGCAATAGCTTTACAATCACCGATAGTCCGTAGTTTAAGCGTACGGCATTCTGGGCCAACTCAGAGTCCACATGACGCGAGGTCAAGTGACGCAGGGCACATACAGCCGGCTCGGTAATCTCTTCGCGATCTCCAGCATTGATAATAGTACGGACGAGGGCGTCCACACCGCCCACCTGACAAACGGTGGCCTTGTTGCGCTGATTGTTGCACGTCAGATTTGAAAGGATACCGGCGGCACAGGTGACCACGTTGACATCGGTCGAGCCCAGAACCTGGACGAGAGATTGGAGCAAAGCTTCAAGGCCCTCCACCTTAGTGGctgcatccgaaagattgcgg-3′; dishevelled, 5′-cgtcaattcggattcggtgcCAAAGAGACTGGTCGAGTCCAGATCGCTGGAGAGCACGGATGCCGATTGATATGTGAGCGGTGGCGGCTGCAGCAGCGGATTGCCCATCATCTTCTGGTGATGGAGCACCTGCTGCTGTTGCTGCTGCGCCAGCTGGACAGGCTGCACCTGCTGCTGCTGTTGTTGTTGCTGTTGCTGGTGCTGCTGCTGCTGTTGTTGTTGCTGCTGCTGATGCTGCTGCTGTTGCTGCTGTAGCTTCCTGTTGGTCAGACCCATGCCCAATTCACACTCACTGGTTGGCAGCTCGGAGCAGTTATCTGACTGATTGGTACCATCGGCGGAGACCAGCCAGGACACCACTCGCCCATTGAAGCAGGGCAGTATGGTGGAGTCGTCGGCTATTTCCTCTTTGACCACACCGAAATCGGCGTCCATTGACTTGAAGAAGTACTTGTAGTTGTTGTTCTGCTTGTTCAGCACCAGCTTGAAATCGCGCAGCGTCACCTGGGCGGATGGAATGGGGATCTTCACCAGATACGGCGTCGTCTCATCGTCGATGTGGTATATCACCTTCGTCTCCTGCCCGCCGCCCCTGTCCGCGTCCATTTTAGCTGCGCCGACTGCGGAAATCCACGGGAAAACTCCTCGACGCGATAAAACGCGATCAAaacgaaatgcgatggatgcg-3′; frizzled 2, 5′-gcgggtagtgtttccttccaGACACAGCGCAGCGTCAAATCGTCAACGCTCGACAAACGGCGTGGTCTAACATTTTTTCCATTCCGTTGGCTTTGCTGCTATCTGGTTGATTTTCTGGTTCATTCGATTCCGCCCAGCTGGGGAGCTGGTCCAACTTTCCGCCGTTAATCCCGAAAACTCTTTGGTGGTAATTAAAAAACGCCGTGCGTCTGTCAAAGTTAATCCAATTTCACGTCTGCTTACATCAAACACTCACACTGGCGCACGAACACCTGGAGATGCAGGCACACACGGATACAGCAAGTATCTGACGGATACACAGACACACATATAAATACGTGCGCCGGTTCTTATAGAAACTTTTTGCCGCTGCAAGTAATTTCTACTTTATTTTCGGACAGATTTTTTAATTAATATTTCATGTAGCGCCAAATGTTGTCGCTATCTTTCAGGTGCTACGGTGTGGGATGCTTGTTTTCCTACAGATATATTTAAAGGTGAGCCCCGAGCCCCGAAAATCCAATTTCAGTTGACGCTTGTCGATCAGCGATCACAGATCGGAATAGCAACAATAACAACGAACAGCAAATCTTGTTTCTTTATAACAATTTCCAACACACATTCTCACATGGCACGGCCACAATTTGTATTTTCATAACTCGTTTTCTGTATTTTCTGTCGCTTTTTCGAGGACGCCACTCTACGGTGTTAGCTTGAAGAAACTTCTTTTATTCACGAATGCTTACGATTGATTGTCACtgctttcgatttgcgcacaa-3′; spitz (antisense), 5′-atagttcgctccgctcacagCTTGTTGCTGCGTCGAATGGCGAAGGACATTAGCGCGTGCTCCAGCCGCATGTGGTAGGGTAGCTTGCGCTCCAGAATGACTGGCTCCTGGCCATCTGGACAGCACCGGTTGCGGCAGCACTCGCACTGGCCGTCGTCATCGTCGTATTCCTGCTGCAGTTCCTGCTCCAGTTCGTAGGCCTTCTTGGCAGCCCGCTGCTcgaagcgcaaatagaaggcc-3′; spitz (sense), 5′-ggccttctatttgcgcttcgAGCAGCGGGCTGCCAAGAAGGCCTACGAACTGGAGCAGGAACTGCAGCAGGAATACGACGATGACGACGGCCAGTGCGAGTGCTGCCGCAACCGGTGCTGTCCAGATGGCCAGGAGCCAGTCATTCTGGAGCGCAAGCTACCCTACCACATGCGGCTGGAGCACGCGCTAATGTCCTTCGCCATTCGACGCAGCAACAAGctgtgagcggagcgaactat-3′.

Flies of the appropriate genotype were cultured and collected upon eclosion. Heat shock was performed by transferring flies to empty plastic vials and immersing them in a 37°C water bath for 1 h. Flies were then allowed to recover at 25°C in vials containing food for at least 5 h. This process was repeated 2× daily for 2 days for a total of 4×1 h heat shocks. Flies were then maintained at 25°C and fed wet yeast daily until dissection.

MARCM clones of various genotypes were induced and the frequency of fully labeled, mosaic or no labeled germaria was measured at 7, 14 or 21 days after clone induction. Stem cell labeling counts were analyzed as described previously (Kronen et al., 2014). Replacement events can be measured within the subset of mosaic germaria at the first time point. Over time, mosaic germaria can become unlabeled or fully labeled, indicating replacement. An increase in the proportion of fully labeled germaria is related to the rate of clone expansion, whereas an increase in the proportion of unlabeled germaria is related to the rate of clone extinction. In wild-type tissue, the rates of extinction and expansion are equal. Homozygosity for a mutation that impairs FSC retention in the niche or self-renewal will cause the rate of clone extinction to be higher than the rate of clone expansion.

Statistics were performed using Microsoft Excel or R Studio. N numbers are provided in the figure legends. P-values for figures are as follows (R script and data are provided in the Supplementary Data): Fig. 5: (D) Control versus arm⁸, P=3.41×10⁻⁴; control versus dsh³, P=1.91×10⁻³; control versus axn^S044230, P=0.356; (F) P=0.021. Fig. 6: (B) Control all time points versus dsh³ all time points, 2.68×10⁻³⁰; dsh³ all time points versus dsh³ with Egfr^ltop all time points, 4.51×10⁻⁷; control 7 days after heat shock (dphs) versus dsh³ 7 dphs, 2.69×10⁻¹¹; control 14 dphs versus dsh³ 14 dphs, 4.88×10⁻¹⁴; control 21 dphs versus dsh³ 21 dphs, 3.03×10⁻¹¹; dsh³ 7 dphs versus dsh³ with Egfr^ltop 7 dphs, 0.004; dsh³ 14 dphs versus dsh³ with Egfr^ltop 14 dphs, 0.046; dsh³ 21 dphs versus dsh³ with Egfr^ltop 21 dphs, 0.022. Fig. 7: (E) P=0.001; (F) P=0.012. Fig. S4: (C) Control versus UAS-Wg, 13C06-Gal4, P=0.271; control versus UAS-Wg, 109-30-Gal4, P=8.94×10⁻²⁰. Fig. S5: (D) P=9.82×10⁻⁴.

Alignments of HMG sites in the spitz gene were performed using the UCSC Genome Browser and tools available on the website (https://genome.ucsc.edu; Kent, 2002; Kent et al., 2002)

We are grateful to the Bloomington Drosophila Stock Center, Ken Cadigan, Ting Xie, Jean Paul Vincent, Ben Ohlstein and Ramanuj DasGupta for stocks and reagents. We also thank FlyBase for curating data into a very useful resource and for providing the ChIPseq data we referenced in this study. Lastly, we thank Katja Rust and Sumitra Tatapudy for helpful comments on the manuscript.