Different modes of Notch activation and strength regulation in the spermathecal secretory lineage

The Notch signaling pathway is evolutionarily conserved and involved in cell-fate specification in almost every organ system across metazoans (Artavanis-Tsakonas et al., 1999; Lai, 2004). The canonical Notch-signaling pathway is activated when the membrane-tethered Delta/Serrate/Lag-2 (Delta and Serrate in Drosophila) ligands on the signal-sending cell bind to the Notch-transmembrane receptor on the signal-receiving cell. This ligand-receptor binding leads to a conformational change and a sequential proteolytic cleavage of Notch, which leads to the release of the Notch-intracellular domain (NICD). The NICD is translocated into the nucleus, binds to CBF-1/Su(H)/Lag-1 [Suppressor of Hairless or Su(H) in Drosophila] transcription factor, and activates downstream gene expression (Schweisguth, 2004; Kopan and Ilagan, 2009). This simple Notch-signaling pathway specifies a broad range of cell fates in different organ systems and at different steps during the cell-lineage progression; however, the underlying mechanism is not fully understood (Andersson et al., 2011; Bray, 2016). In addition to the context-dependent interaction of Notch signaling with other developmental signals, recent studies suggest that differential Notch signaling and strength can influence gene expression and cell fate (Nandagopal et al., 2018; Henrique and Schweisguth, 2019). For example, different Notch signaling strengths can specify distinct cell fates in multiple organ systems and organisms (Contreras et al., 2018; Basch et al., 2016; Gama-Norton et al., 2015; Ohlstein and Spradling, 2007). However, we do not fully understand how Notch signaling strength is precisely regulated and affects gene expression in various cellular contexts.

Endocytic trafficking of ligands and the Notch receptor also play crucial roles in modulating Notch signaling (Fortini and Bilder, 2009; Yamamoto et al., 2010; Schnute et al., 2018). Epsin-dependent ligand endocytosis generates a pulling force that is essential for the conformational change of the Notch receptor to expose the proteolytic cleavage site (Langridge and Struhl, 2017; Meloty-Kapella et al., 2012). Notch-receptor endocytosis and trafficking include multiple routes of internalization, all of which play different roles in signal regulation. For example, a minority of AP-2-independent but dynamin-, clathrin- and Rab5-dependent Notch endosomal entry is crucial for Notch activation (Lu and Bilder, 2005; Vaccari et al., 2008; Windler and Bilder, 2010). However, most of the Notch receptor is internalized via AP2-dependent endocytosis and trafficked to the lysosome for degradation. If this latter fraction of Notch is not incorporated into the intraluminal vesicles before lysosomal degradation [e.g. mutations affecting the endosomal sorting complex required for transport (ESCRT) complexes], this can result in ectopic activation of Notch signaling, which can be ligand independent (Thompson et al., 2005; Vaccari and Bilder, 2005; Wilkin et al., 2008; Vaccari et al., 2009; Troost et al., 2012). In contrast to general endosomal trafficking regulators, Deltex (Dx) and Suppressor of Deltex [Su(dx)] are two E3 ubiquitin ligases that specifically regulate Notch ubiquitylation and trafficking. Su(dx) promotes Notch poly-ubiquitylation and incorporation into intraluminal vesicles and negatively modulates Notch signaling; Dx promotes Notch mono-ubiquitylation, prevents its incorporation into intraluminal vesicles and is considered a positive regulator of Notch signaling (Wilkin et al., 2004, 2008; Hori et al., 2004, 2011; Shimizu et al., 2014). Numb is also involved in Notch trafficking and cell-fate determination in a tissue-specific manner (Schweisguth, 2015). Numb inhibits Notch recycling back to the cell membrane and leads to a Notch-off state during asymmetric cell division (Guo et al., 1996; Spana and Doe, 1996; Giebel and Wodarz, 2012; Couturier et al., 2013; Johnson et al., 2016). Therefore, Notch endocytic trafficking can result in either inhibiting or activating Notch signaling, depending on the fate of the endocytosed Notch. It is unclear whether endosomal trafficking of Notch contributes to the different Notch signaling strengths in physiological settings.

We have shown that dynamic Notch signaling regulates binary fate determination in Drosophila spermathecal lineage (Shen and Sun, 2017). Drosophila spermatheca is a class-III insect gland containing ∼80 secretory units that produce essential secretions for sperm storage and ovulation (Filosi and Perotti, 1975; Schnakenberg et al., 2011; Sun and Spradling, 2013; Cattenoz et al., 2016). Lineage-tracing experiments show that spermathecal gland precursors (P0) in the genital imaginal disc proliferate to give rise to one secretory unit precursor (SUP) and one lumen epithelial precursor (LEP) around 26 hours (h) after puparium formation (APF; Sun and Spradling, 2012; Fig. 1A). Delta (Dl) activates Notch in LEP cells to maintain the expression of runt-domain transcription factor Lozenge (Lz) and promote LEP differentiation into lumen epithelial cells to form spermathecal lumen. In contrast, SUP cells do not gain Notch signaling, lose Lz expression, but gain the expression of transcription factor Hindsight (Hnt) and Cut (Shen and Sun, 2017; Fig. 1A). The SUP divides again around 32 h APF and gives rise to a pIIa and pIIb cell. Notch signaling is activated in pIIb cells to promote differentiation into apical cells (ACs) with strong Cut expression. This facilitates the formation of the end apparatus: an acellular chitin structure that forms part of the adult secretory unit (Filosi and Perotti, 1975; Allen and Spradling, 2008). In contrast, pIIa cells do not have Notch signaling, lose Cut expression and divide again to give rise to a basal cell (BC) and secretory cell (SC). Whereas Notch signaling is active in SCs to promote Hnt expression and their differentiation into adult secretory cells, BCs do not have Notch signaling and lose Hnt expression (Shen and Sun, 2017). The AC, BC and SC wrap around each other in a concentric ring to build the complex secretory unit in adults; both AC and BC die after unit formation (Mayhew and Merritt, 2013; Sun and Spradling, 2012). Although we have observed asymmetric Notch signaling and know it is essential for binary fate determination, we do not fully understand how Notch signaling is activated and regulated in this lineage.

Here, we have used mosaic analysis to identify the core Notch pathway in each cell division in spermathecal secretory lineage. We found that Dl and Serrate (Ser) in BCs redundantly serve as ligands for Notch signaling in SCs. Dl in pIIa functions as the ligand for Notch signaling in pIIb/AC and is compensated by adjacent cells when Dl is mutated in the SUP lineage. We also found that Notch signaling in pIIb/AC requires a lower level of Su(H) than that in SCs, consistent with our previous report that Notch-signaling activity is lower in ACs than SCs (Shen and Sun, 2017). Furthermore, we found that Numb and Rab5 are essential for suppressing Notch signaling in pIIa cells but not BCs. Dx, a positive regulator of Notch signaling, suppresses Notch signaling in SCs but does little in pIIb/ACs when ectopically expressed. Altogether, our work suggests that cell fate specification in spermathecal lineage uses different strengths of Notch signaling, which is activated by different ligands and regulated by various endocytic components.

Using RNA interference (RNAi), we have previously shown that Notch (N) knockdown in SUP at 20 h APF transformed 25% of pIIb cells into pIIa cells, indicating that Notch is required for pIIb/AC specification (Shen and Sun, 2017). To test this using another method, we used the mosaic system (Xu and Rubin, 1993) to generate N^55e11 mutant SUP clones. After inducing clones at 20 h APF and examining them at 48 h APF, we observed four-cell clones (likely SUP clones) and eight-cell clones (likely P0 clones), none of which contained Cut⁺ ACs (Fig. 1B). This suggests that complete removal of Notch leads to 100% pIIb-to-pIIa transformation. None of the Notch mutant clones contained Hnt⁺ SCs at 48 h APF (Fig. 1C), suggesting that all Notch mutant cells had a BC-like fate. Together, these data strongly support our previous conclusion that Notch is essential for determining the fate of pIIb (versus pIIa) and SC (versus BC) in the spermathecal lineage.

Our previous data suggested that Dl functions in SUPs to activate Notch in LEPs (Shen and Sun, 2017). However, the ligand required for Notch signaling in the pIIb and SC remains a mystery as knockdown of either Dl or Ser did not affect pIIb or SC specification. We reasoned that RNAi knockdown may not completely disrupt Dl or Ser function. Therefore, we generated Dl or Ser mutant clones using the mosaic system. Perhaps surprisingly, neither Dl nor Ser mutant clones affected SC fate. When we induced Dl^rev10 clones at 20 h APF and examined them at 48 h APF, we observed either three- or six-cell clones. All three-cell clones contained a Hnt⁺ SC. The remaining two cells in the clones exhibited either faint Hnt expression (characteristic of AC) or no Hnt expression (characteristic of BC) (Fig. 2A). All six-cell clones contained two Hnt⁺ SCs, two ACs and two BCs, which were likely P0 clones that engendered two SUPs (Fig. S1A). These data are consistent with our previous RNAi analysis (Shen and Sun, 2017) and suggest that Dl is not required for SC fate. To test whether Dl mutant clones disrupt Dl function, we generated the same Dl^rev10 clones in the Drosophila ovary. As expected, Dl-mutant follicle-cell clones prematurely activated Hnt in stage-five egg chambers (Fig. S1B; Palmer et al., 2014). This indicates that Dl is indeed disrupted in the Dl^rev10 clones.

Similar to Dl mutant clones, both Ser^RX82 and Ser^RX106 SUP clones contain three cells with one Hnt⁺ SC at 48 h APF (Fig. 2B and Fig. S1C). Both Ser^RX82 and Ser^RX106 mutant lines have been used to disrupt Ser function (Lebestky et al., 2003; LeBon et al., 2014). These studies suggest that removing Ser alone will not disrupt Notch signaling in the SC.

Notch signaling can be activated by Dl and Ser either independently or redundantly (Palmer et al., 2014; Zeng et al., 1998). To test these two possibilities, we generated Dl^rev10 Ser^RX82 double-mutant SUP clones at 20 h APF. Strikingly, none of the Dl^rev10 Ser^RX82 SUP clones contained Hnt⁺ SCs (Fig. 2C), indicating a SC-to-BC transformation and Notch signaling disruption. When Dl^rev10 Ser^RX82 clones were generated in pIIa cells at 26 h APF, none of the two-cell clones at 48 h APF contained Hnt⁺ SCs (Fig. 2D). This result points to an efficient SC-to-BC transformation. We also observed that Dl Ser mutant BC clones disrupted SC specification in corresponding twin-spot clones (wild-type clones) when we induced clones at 28 h APF (Movie 1). Together, these data suggest that Dl and Ser function redundantly in BCs to activate Notch signaling in SCs.

We next tried to determine the ligand source for Notch signaling in pIIb cells, which differentiate into ACs. When Dl Ser mutant clones were induced at 20 h APF, we frequently observed three-cell clones containing two BCs and one cell with low Hnt expression (Fig. 2C), indicating an AC. To test this hypothesis, we stained the tissue with the Cut antibody, which specifically labels ACs at 48 h APF. We found that most three-cell clones of Dl^Rev10, Ser^RX82, Ser^RX106 and Dl^Rev10Ser^RX82 had one Cut⁺ AC (Fig. 3A-C, Fig. S2A). This suggests that removing both Dl and Ser in SUPs is not sufficient to disrupt Notch signaling in pIIb specification. Consistent with this conclusion, most of the generated clones were three-cell clones, unlike N^55e11 four-cell clones (Fig. 1B,C).

Several hypotheses can explain why Dl Ser double-mutant clones do not disrupt pIIb specification: (1) remnants of Dl or Ser protein in Dl Ser mutant pIIa cells are sufficient to activate Notch signaling in pIIb cells; (2) Dl or Ser from adjacent cells, such as epithelial cells, can activate Notch signaling in pIIb cells; or (3) Notch signaling in pIIb cells is activated in a ligand-independent manner. To test the first hypothesis, we generated Dl Ser double-mutant P0 clones at 8 h APF and examined the clone composition at 48 h APF. This way, the two divisions away from Notch activation in pIIb cells should yield enough time for Dl and Ser to be diluted or degraded. Surprisingly, Dl Ser double-mutant clones still contained normal Cut⁺ ACs. We examined clones with four or more cells, which were likely P0 clones. Ninety-seven percent of these clones contained one to three Cut⁺ ACs (Fig. 3D). The clone size varied as the lineage above SUP was not fixed. Therefore, we calculated the ratio of Cut⁺ ACs to BCs and BC-like cells (transformed from SCs), all of which can be differentiated from ECs by their faint Cut staining when overexposed during image acquisition. The ratio of ACs to BCs and BC-like cells in Dl Ser clones is 0.43, which is close to the expected ratio of 0.5 and does not significantly differ from the ratio of 0.48 in clones induced at 20 h APF (Fig. 3E). This result suggests that Dl Ser mutant P0 clones do not affect pIIb/AC specification. Therefore, it is unlikely that Dl or Ser protein remnants activate Notch signaling in pIIb cells in Dl Ser mutant clones.

To test the second and third hypotheses, we generated large Dl Ser mutant clones by heat shocking animals at the third instar larval stage. If Dl or Ser from adjacent cells can activate Notch signaling in pIIb cells, we do not expect to see Cut⁺ ACs at the center of the clone where surrounding cells are all Dl Ser mutant. Thus, we predict that the ratio of ACs to BCs/BC-like cells should decrease. On the other hand, if Notch signaling in pIIb cells is activated in a ligand-independent manner, we would expect the ratio of ACs to BCs/BC-like cells to remain at 0.5 and the ACs to be located everywhere in the clone. Most large Dl Ser mutant clones still contained Cut⁺ ACs at 48 h APF (Fig. 3F). However, all Cut⁺ ACs were located at the periphery of the clones adjacent to wild-type cells (Fig. 3F, Fig. S2B, and Movies 2,3). We failed to observe any Cut⁺ ACs in the region that was detaching from the lumen, which is likely due to the loss of ECs, as observed previously (Shen and Sun, 2017). The ratio of ACs to BCs/BC-like cells was 0.26, which was significantly lower than the ratio of 0.48 observed in clones induced at 20 h APF (Fig. 3E). These data suggest that Notch signaling in pIIb is ligand dependent and the observed AC specification in Dl Ser mutant clones is due to the ligand compensation from adjacent wild-type cells. We also found that four out of 63 clones contained no Cut⁺ ACs (Movie 4), which suggests that ligand compensation from adjacent wild-type cells may be inefficient in some cases.

To pinpoint which ligand is crucial for pIIb specification during normal development, we generated Dl and Ser mutant clones at the third instar larval stage and examined the clones at 48 h APF. Ser^RX82 clones contained Cut⁺ ACs with a ratio of 0.5 for AC to BC/BC-like cells (Fig. 3E,G). We did not observe any clones detaching from the lumen (Fig. 3G and Movie 5). In contrast, Dl^rev10 mutant clones showed a detachment similar to Dl Ser mutant clones (Fig. 3H and Movie 6). In addition, Dl^rev10 clones showed a ratio of 0.27 for AC to BC/BC-like cells, similar to Dl Ser mutant clones (Fig. 3E). Four out of 24 clones contained no Cut⁺ AC (Movie 7). Together, these data support the idea that Dl is the endogenous ligand for Notch signaling in pIIb cells.

Thus far, our data suggested that Notch activation in pIIb cells and SCs utilizes different molecular mechanisms. Our previous work showed that pIIb/ACs had much lower Notch activity than SCs according to a Gbe-Su(H)-lacZ reporter (Shen and Sun, 2017). In order to understand whether pIIb cells and SCs have different Su(H) requirements, we examined whether Su(H) is crucial for both pIIb and SC specification using Su(H)^SF8: a strong loss-of-function allele. We first validated the Su(H)^SF8 allele in follicle-cell clones, which disrupted Notch signaling and extended Cut expression in stage 7 egg chambers, as expected (Fig. S3A; Sun and Deng, 2005). Next, we generated Su(H)^SF8 SUP clones at 20 h APF and examined them at 48 h APF. To our surprise, we still detected Hnt⁺ SCs in these clones (Fig. 4A). In contrast, when we generated clones at 1 h APF and examined them at 48 h APF, we noticed drastically reduced Hnt expression in SCs (Fig. 4B and Movie 8). We observed a similar result in Su(H)^SF8 clones induced at the third instar larval stage (Fig. 4C). These data suggest that Su(H) is crucial for Notch signaling in SC fate specification. The fact that the defect was not observed in SUP clones induced at 20 h APF indicates that Su(H) protein is more stable than Dl, Ser or Notch protein in the spermathecal lineage.

Next, we examined the AC fate in Su(H)^SF8 clones with Cut staining. Interestingly, all P0 clones induced at 1 h APF contained Cut⁺ ACs (Fig. 4D), suggesting that Su(H)^SF8 clones generated in P0 do not affect the pIIb fate. As Su(H) is more stable, we generated Su(H)^SF8 clones at the third instar larval stage. Surprisingly, most Su(H)^SF8 clones still contained Cut⁺ ACs at 48 h APF (Fig. 4E), despite the fact that these clones showed no detectable Su(H) protein at 26 h APF (Fig. S3B). Su(H)^SF8 clones induced at the second instar larval stage still contained Cut⁺ ACs at 48 h APF (Fig. 4F and Movie 9). We also observed mutant cells detaching from the lumen as Dl Ser mutant clones (Fig. 4E). These data indicate that Su(H)^SF8 clones do not affect pIIb/AC specification.

The result from Su(H)^SF8 clones contradicted our previous findings that ectopic expression of a dominant-negative form of Su(H) [Su(H)^DN] can induce pIIb-to-pIIa transformation (Shen and Sun, 2017). There are two possible explanations: (1) our previous experiment with Su(H)^DN was incorrect; or (2) pIIb/AC specification requires only a low level of Su(H) as Su(H)^SF8 does not completely disrupt Su(H) function (Bray and Furriols, 2001). We tested the first possibility by generating flip-out clones with Su(H)^DN overexpression at 8 h APF and examined at 48 h APF. Most SUP clones were four-cell clones containing neither Hnt⁺ SCs nor Cut⁺ ACs (Fig. 5A,B). This result demonstrated the efficient pIIb-to-pIIa and SC-to-BC transformations in Su(H)^DN-overexpressing clones. Therefore, we rejected the first hypothesis.

We tested the second possibility by generating Su(H)-null clones with Su(H)^attP or Su(H)^Δ47 (Morel and Schweisguth, 2000; Praxenthaler et al., 2017). Similar to Su(H)^SF8 clones, Su(H)^attP clones did not affect SC fate when generated at 20 h AP (Fig. S4A) and did affect SC specification when generated at 0 h APF and the third instar larval stage (Fig. S4B and Fig. 5C). Unlike Su(H)^SF8 clones, all Su(H)^attP clones generated at the third instar larval stage did not contain Cut⁺ ACs when examined at 48 h APF (Fig. 5D and Movie 10), which indicates a pIIb-to-pIIa transformation. We observed the same results in Su(H)^Δ47 clones (Fig. S5). Together, these results suggest that a low level of Su(H) is sufficient for specifying the pIIb/AC fate. Therefore, pIIb/AC and SC specification uses not only different ligands for Notch activation but also different levels of Su(H) protein and Notch signaling.

Endosome sorting of Notch protein is crucial for regulating Notch signaling (Andersson et al., 2011; Fortini and Bilder, 2009). We sought to explore how different levels of Notch signaling are generated for pIIb/AC and SC specification. To that end, we first examined the role of Numb: a conserved endosomal adaptor protein involved in suppressing Notch recycling and preventing ectopic Notch signaling during asymmetric Notch activation (Frise et al., 1996; Spana and Doe, 1996; Uemura et al., 1989; Johnson et al., 2016). Ninety-four percent of numb² SUP clones induced at 20 h APF were made of two Cut⁺ ACs at 48 h APF, indicating a pIIa-to-pIIb transformation (Fig. 6A). The remaining clones were three-cell clones and contained one Cut⁺ AC, indicating normal pIIa/pIIb specification (Fig. 6B). Ninety-four percent of flip-out clones with numb knockdown also showed the pIIa-to-pIIb transformation (Fig. 6E), mirroring flip-out clones with NICD overexpression (Shen and Sun, 2017). These results suggest that pIIa cells need Numb for preventing Notch signaling. Therefore, Numb plays a key role in asymmetric Notch signaling during pIIa/pIIb specification.

By contrast, Numb played no role in SC/BC specification. First, all numb² SUP clones that had normal pIIa/pIIb specification when induced at 20 h APF still contained three cells with one Hnt⁺ SC (Fig. 6C). Second, when numb² clones were induced at 24 h APF, all three-cell SUP clones and two-cell pIIa clones still contained one Hnt⁺ SC (Fig. 6D). Finally, we observed the same phenomenon in flip-out clones with numb knockdown (Fig. 6F). Therefore, Numb is not required for SC/BC specification.

We also determined whether Numb is required for SUP/LEP specification. We knocked down numb in all gland precursors using the lz-Gal4 and 51B02-Gal4 drivers. At 27 h APF, we observed normal Hnt and Cut expression in basal SUPs and Lz expression in apical LEPs, indicating that Numb is not required for LEP/SUP specification (Fig. S6A-F). We also observed that spermathecae at 48 h APF had more Cut⁺ ACs but fewer Hnt⁺ SCs, which rendered fewer secretory cells in adult spermathecae (Fig. S6G-L). These data support our conclusion that Numb is required only during pIIa/pIIb specification but not in SC/BC specification.

The differential involvement of Numb in pIIb/pIIa and SC/BC specification prompted us to examine other endosome components. We first examined the function of GTPase Rab5, which is required for Notch endosome entry and activation (Fortini and Bilder, 2009; Wilkin et al., 2008). We used the flip-out clone system to overexpress a dominant-negative form of Rab5 (Rab5^S43N). Over 80% of the Rab5^S43N-overexpressing SUP clones induced at 20 h APF were composed of two Cut⁺ ACs with faint Hnt staining, indicating a pIIa-to-pIIb transformation (Fig. 7A-C). This suggests that Rab5 is required in pIIa cells to prevent Notch activation, contrasting with the known role of Rab5 in promoting Notch activation. We propose that Rab5 and Numb mediate Notch endocytosis and degradation to prevent Notch activation in pIIa cells. In pIIb cells, Notch activation occurs through a Rab5-independent mechanism.

To determine whether Rab5 regulates Notch signaling in SC/BC specification, we generated Rab5^S43N-overexpressing SUP clones at 23 h APF to reduce the rate of pIIa-to-pIIb transformation. Only 17% of SUP clones underwent pIIa-to-pIIb transformation, while 66% and 17% of clones contained SCs with normal and low level of Hnt, respectively (Fig. 7C,D and Fig. S7A,B). This suggests that Rab5 promotes Notch signaling during SC/BC specification, in contrast to inhibiting Notch signaling during pIIb/pIIa specification.

We tested the role of Deltex (Dx), an E3 ubiquitin ligase that promotes Notch activation when ectopically expressed (Matsuno et al., 1995). Surprisingly, overexpression of Dx in SUP clones did not affect pIIb/pIIa specification (Fig. 7E) but disrupted Notch signaling during SC/BC specification according to Hnt expression (Fig. 7F). Consistent with this, overexpression of Dx in the SUP lineage with 51B02-Gal4 did not affect Cut⁺ ACs but almost completely disrupted Hnt⁺ SCs at 48 h APF and in adult spermathecae (Fig. S7C-E). We conclude that pIIb/pIIa and SC/BC specification use different strengths of Notch signaling, which are activated by different ligands and influenced by different endosomal pathways (Fig. 7G).

We have used mosaic analysis to demonstrate the essential role of Notch signaling in specifying cell fate and how Notch signaling is differentially regulated in each division of the spermathecal lineage (Fig. 7G). Using the Gbe-Su(H)-lacZ reporter, a synthetic Notch-activity reporter with three Grainy head-binding sites and two Su(H)-binding sites fused with the lacZ gene (Furriols and Bray, 2001), we have previously demonstrated that Notch signaling in ACs is much lower than SCs at 48 h APF and LEPs at 26 h APF (Shen and Sun, 2017). Here, we show that Su(H)^attP or Su(H)^Δ47 mutant clones, but not Su(H)^SF8 mutant clones, affect AC fate. Both Su(H)^attP and Su(H)^Δ47 have a deletion in the Su(H)-coding region and therefore are molecular null alleles (Morel and Schweisguth, 2000; Praxenthaler et al., 2017). In contrast, Su(H)^SF8 is a strong loss-of-function allele but still has low Su(H) function (Bray and Furriols, 2001). Together, these data strongly suggest that a low level of Notch signaling is sufficient for specifying AC fate; a higher level is required for specifying SC and LEP fate. There are multiple examples of differential strengths of Notch signaling for specifying cell fates in Drosophila as well as mammals (Basch et al., 2016; Contreras et al., 2018; Gama-Norton et al., 2015; Ohlstein and Spradling, 2007). This is likely a conserved mechanism used by the Notch signaling to fulfill its pleiotropic functions in different cellular environments. Recent work showed that different dynamics of Notch signaling activated by different ligands can turn on different gene expression program (Nandagopal et al., 2018). Here, we showed that only Dl is involved in low Notch signaling during pIIa/pIIb specification, and both Dl and Ser are involved in high Notch signaling during SC/BC specification. An interesting next step would be to investigate whether the use of different ligands mediates different strength of Notch signaling in this case.

Having found low and high Notch signaling in the spermathecal lineage, we next explored the underlying regulatory mechanism. We first examined the role of Notch endosomal entry, which is not only required for normal physiological Notch activation but also for abnormal pathological Notch activation (Hori et al., 2004; Vaccari et al., 2008; Windler and Bilder, 2010). Overexpression of Rab5^DN, which disrupts the fusion of the internalized endocytic vesicles with the early endosome, did not affect pIIb/AC specification (Fig. 7A,B). This implies that low Notch signaling in pIIb cells does not rely on Rab5-dependent Notch endosome entry. In contrast, Rab5 seems to be essential for high Notch signaling in SCs (Fig. 7C,D). Therefore, we identified a Rab5-independent pathway for low Notch signaling and a Rab5-dependent pathway for high Notch signaling. In future, we would like to examine whether Rab5-dependent and -independent Notch activation generates high and low Notch signaling in other cellular contexts, respectively.

Instead of inhibiting Notch signaling, Rab5^DN overexpression promotes Notch activation in pIIa cells and transforms them into pIIb cells (Fig. 7A,B). This phenomenon mirrors numb mutant clones, where numb mutant pIIa cells transformed into pIIb cells (Fig. 6A,E). Given that asymmetrically localized Numb promotes internalized Notch for late endosome/lysosome targeting and degradation, we believe that Rab5 works upstream of Numb to promote endosomal targeting of Notch. Therefore, without Rab5, Notch cannot efficiently move into the endosome/lysosome for degradation and achieve Rab5-independent Notch activation. This was also observed in the bristle lineage of Drosophila (Johnson et al., 2016) and is likely conserved in other cellular contexts.

Another surprising finding relates to Dx, the E3 ubiquitin ligase that promotes Notch signaling (Matsuno et al., 1995). Dx overexpression leads to ligand-independent Notch activation in wing imaginal discs, which depends on Notch trafficking into the late endosome/lysosome (Hori et al., 2004, 2011; Wilkin et al., 2008). Here, we have found that Dx overexpression suppresses Hnt expression and SC fate, which depends on high Notch signaling (Fig. 7F and Fig. S7D). Therefore, ectopic Dx seems to inhibit Notch signaling in SCs, but does not affect Notch signaling during LEP/SUP and pIIb/pIIa specification. Depending on the cellular context, Dx can either promote or suppress Notch signaling. In most mammalian studies, Dx negatively regulates Notch signaling (Yamamoto et al., 2001; Zheng and Conner, 2018). Taken together, Dx likely plays a conserved role in suppressing Notch signaling.

We also examined the role of Su(dx), another E3 ubiquitin ligase that negatively regulates Notch signaling (Cornell et al., 1999). Overexpression of Su(dx) did not affect Notch signaling and cell-fate determination (Fig. S7F-H). Therefore, we do not know how ectopic Dx suppresses Notch signaling in SCs. One possibility is that Krz – the Drosophila non-visual β-arrestin that blocks Dx-induced Notch activation when ectopically expressed (Mukherjee et al., 2005) – is highly expressed in SCs but not pIIb or LEPs. However, genetically testing this possibility is difficult due to the technical challenges of working with pupal genital discs. Nonetheless, our analysis suggests that modulating Notch endocytosis and vesicle trafficking is likely conserved in cells so they can generate different Notch-signaling strengths for different gene expression and cell-fate determination.

The importance of endocytic regulation of Notch signaling has long been recognized (Fortini and Bilder, 2009; Fürthauer and González-Gaitán, 2009; Kandachar and Roegiers, 2012). Owing to the complexity of the endosome network, endocytic Notch regulation does not always yield the same result in terms of Notch activation or inhibition in different cellular contexts. Here, we present evidence that the spermathecal secretory lineage applies both low and high Notch signaling to regulate binary fate determination. In addition, low and high Notch signaling are subjected to different endocytic regulation, which has very distinct consequences on Notch activation or inhibition. Therefore, a comparative study of low and high Notch signaling in the spermathecal lineage presents an exciting model with which to study the endocytic regulation of Notch activation and Notch-signaling strength.

Flies were reared on standard cornmeal-molasses food at 25°C, unless otherwise indicated. The following mutant stocks were used: N^55ell, FRT19A/FM7 (Sun and Deng, 2005); Dl^rev10, FRT82B/TM6B (Heitzler and Simpson, 1991); Ser^RX82, FRT82B/TM6B (Lebestky et al., 2003); Ser^RX106, FRT82B/TM6B (LeBon et al., 2014); Dl^rev10 Ser^RX82, FRT82B/TM6B (Bloomington #6300); Su(H)^SF8, FRT40A/Cyo, Dfd-eYFP (Sun and Deng, 2007); Su(H)^attP, FRT40A/Cyo, Dfd-eYFP (Praxenthaler et al., 2017); Su(H)^Δ47, FRT40A/Cyo, Dfd-eYFP (Morel and Schweisguth, 2000); and numb², FRT40A/Cyo, Dfd-eYFP (Rhyu et al., 1994). To generate N^55ell clones, we crossed hsFLP, ubiGFP, FRT19A/Y; E(spl)-CD2/Cyo males with N^55e11, FRT19A/FM7 virgin females. To generate Dl^rev10, Ser^RX82, Ser^RX106 and Dl^rev10Ser^RX82 double-mutant clones, we used hsFLP; ubiRFP, FRT82B/TM6B as virgin females. To generate Su(H)^SF8, Su(H)^attP, Su(H)^Δ47 and numb² mutant clones, we used hsFLP; ubiRFP, FRT40A/Cyo, Dfd-eYFP as virgin females. We used the following transgenic lines for the flip-out clone analysis: UAS-numb^RNAi (Bloomington, 35045), UAS-Su(H)^DN (Mukherjee et al., 2011), UAS-Rab5^S43N (Bloomington, 42704), UAS-flag:Dx; MKRS/TM6B (Hori et al., 2011) and UAS-Su(dx) (Bloomington, 51664). To generate flip-out clones, hsFLP; act<CD2<Gal4, UAS-GFP virgin females were crossed with corresponding transgenic lines. We used lz-Gal4, UAS-GFP and 51B02-Gal4 (Shen and Sun, 2017; Sun and Spradling, 2013) to overexpress transgenes in gland precursors.

We collected and dissected pupae using previously described methods (Shen and Sun, 2017). In short, we collected white prepupae (designated as 0 h APF) over a 30 min window into a new food vial. We sexed prepupae according to the gonadal size, and aged them to the desired pupal stage for clone induction and dissection. We generated mutant clones and flip-out clones using FLP/FRT-based mitotic recombination and cis-recombination, respectively (Xu and Rubin, 1993; Pignoni and Zipursky, 1997). For mosaic-clone induction, we heat shocked larvae or pupae in a 37°C water bath for 1 h, which is optimal for generating mutant clones without disrupting gland formation. For flip-out clone induction, we heat shocked pupae in a 37°C water bath for 15-30 min to generate isolated clones in spermathecal glands. On average, we observed fewer than 10 interested clones per spermatheca (see details in Table S1). We captured z-stack images for each clone and identified the cells of each clone using molecular and morphological markers (Shen and Sun, 2017; Sun and Spradling, 2012). We identified the individual clone by the loss (mutant clones) or gain (flip-out clones) of fluorescent-protein markers, which were surrounded by non-clone cells in three dimensions. In the case of mutant-clone analysis, we also made sure that all cells in spermathecae without heat-shock treatment had fluorescent-protein markers (Fig. S8). Therefore, clone induction in spermathecae is strictly controlled by the heat-shock treatment. We excluded clones with only epithelial cells from our analysis.

Antibody staining was performed as previously described (Shen and Sun, 2017; Sun and Spradling, 2012). In short, we fixed the entire genital disc attached to the cuticle in 4% electron microscopy-grade paraformaldehyde for 15 min and blocked in PBTG (phosphate-buffered saline+0.2% Triton-X 100+2% normal goat serum+0.5% BSA) before carrying out primary and secondary antibody staining. The following primary antibodies were used: mouse anti-Cut (1:15), anti-Lz (1:15) and anti-Hnt (1:150) obtained from the Developmental Studies Hybridoma Bank (DSHB); rabbit anti-RFP (1:1000; MBL International, PM005); rabbit anti-GFP (1:4000; Life Technologies, A11122); and rabbit anti-Su(H) (1:1000; Santa Cruz Biotechnology, sc26761). We used the following secondary antibodies: Alexa-568 goat anti-rabbit, Alexa-568 goat anti-mouse, Alexa-488 goat anti-rabbit and Alexa-488 goat anti-mouse (1:800; Life Technologies, A11011, A11004, A11008, A11001, respectively). Nuclei were stained with DAPI (Sigma-Aldrich). We used a Leica SP8 confocal microscope to capture images, which we then processed with ImageJ (Schneider et al., 2012) or Fiji (Schindelin et al., 2012) software. We assembled the images in Adobe Photoshop.

We thank Drs Spyros Artavanis-Tsakonas, Utpal Banerjee, Wu-Min Deng, Yuh-Nung Jan, Hamed Jafar-Nejad, Dieter Maier and Francois Schweisguth, and the Bloomington Drosophila Stock Center for providing us fly stocks and the DSHB for antibodies. We appreciate the comments from anonymous reviewers. We thank the Sun lab members for technical support and productive discussions.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.184390.reviewer-comments.pdf