Engrailed, Suppressor of fused and Roadkill modulate the Drosophila GLI transcription factor Cubitus interruptus at multiple levels

The Hedgehog (Hh) pathway is one of a small number of signaling cascades essential for the proper development of most animals. It plays a key role in patterning arthropod segments and numerous tissues in mammals (Lee et al., 2016). Defects in Hh signaling lead to a number of congenital abnormalities including holoprosencephaly, whereas inappropriate activation of the pathway is involved in several important human cancers including basal cell carcinoma and medulloblastoma (Briscoe and Thérond, 2013).

Hh reception and early transduction involves several transmembrane proteins, including its co-receptor Patched (Ptc) and the G Protein Coupled Receptor (GPCR) Smoothened (Smo). In the absence of Hh, Ptc negatively controls Smo, likely by controlling its access to accessible cholesterol (Radhakrishnan et al., 2020; Kinnebrew et al., 2021), leading to the internalization and degradation of Smo. Binding of Hh to Ptc, however, relieves Ptc repression of Smo (Taipale et al., 2002) and promotes its hyperphosphorylation by multiple kinases (Chen and Jiang, 2013). Activated Smo recruits a complex composed of Costal2 (Cos2; also known as Cos), the Fused (Fu) kinase and Cubitus interruptus (Ci) (Robbins et al., 1997; Sisson et al., 1997). Activation of the Fu kinase is required for high level Hh signaling (Alves et al., 1998; Zhou et al., 2006; Ranieri et al., 2012; Sanial et al., 2017; Giordano et al., 2018; Han et al., 2019) and the release of Ci for translocation into the nucleus (Lefers et al., 2001).

As in other signaling systems, the Hh pathway is subject to a number of regulatory loops to ensure that it is activated in the appropriate locations and that responses are robust and precise. One of the best known feedback loops is that involving ptc. The ptc gene is a transcriptional target of Hh signal transduction, and its activation by Hh signaling leads to increased sequestration and degradation of Hh, limiting its range of action (Chen and Struhl, 1996). Another feedback loop that acts positively involves the interplay between Smo and the Fu kinase (Claret et al., 2007; Sanial et al., 2017).

A second level of regulation is the processing of the Ci transcription factor into a repressor in the absence of Hh signaling (Aza-Blanc et al., 1997). The production of the Ci repressor ensures that Hh target genes are actively repressed in the absence of signal and generates a reciprocal gradient to that of the full-length Ci transcriptional activator (Parker et al., 2011). A surprising consequence of these reciprocal gradients is that genes such as ptc that contain enhancers with consensus Ci binding sites are only expressed in response to high level Hh signaling, whereas genes such as decapentaplegic (dpp) that contain enhancers with imperfect Ci binding sites are expressed in response to modest level Hh signaling. Parker et al. (2011) demonstrated that replacing the imperfect Ci binding sites in a dpp enhancer with perfect Ci binding sites caused the expression of this enhancer to shift from the domain of modest level Hh signaling to the domain of high level Hh signaling along the compartment boundary. The authors concluded that the Ci repressor was able to outcompete the full-length Ci activator form for binding to these perfect sites.

To explore other potential feedback loops involving Hh target genes, we examined the roles of the transcription factor engrailed (en) in Hh signaling. It was originally thought that, in Drosophila, expression of en defines the posterior compartment where Hh is produced and is required for hh expression (Tabata et al., 1992), whereas expression of ci defines the anterior compartment (Hh-receiving cells). In the wing disc pouch, the en gene is also activated in the anterior compartment of the wing pouch in response to the highest levels of Hh signaling (Blair, 1992). The role of this anterior expression of en is not well understood, though it does downregulate the expression of dpp (Strigini and Cohen, 1997), and its region of expression corresponds to that identified as a region currently considered to contain a labile high-activity form of Ci (Ohlmeyer and Kalderon, 1998). Here, we show that anterior expression of en attenuates the expression of ci, leading to decreased levels of ptc expression. This modifies the Hh gradient and results in an expansion of the domain between longitudinal wing veins three and four.

We also examine the role of a second Hh target gene, roadkill (rdx) (also known as hib). rdx was identified as a gene activated in response to Hh signaling. rdx encodes an E-3 ligase that functions with Cul3 to target proteins for degradation by the proteasome (Kent et al., 2006; Zhang et al., 2006). Unlike Slmb and Cul1, which mediate Ci processing into the repressor, Rdx causes the complete degradation of Ci from both the N and C termini (Zhang et al., 2009). Rdx has also been implicated in regulating the nuclear import of Ci (Seong et al., 2010). In overexpression experiments, it has been shown that the Ci-interacting protein Suppressor of fused [Su(fu)] competes with Rdx for binding to Ci and partially protects Ci from Rdx-mediated degradation (Zhang et al., 2006) and regulates this process in vivo (Seong and Ishii, 2013). Notably, Rdx plays an important role in the degradation of Ci posterior to the morphogenetic furrow in the eye disc, where loss of rdx leads to a disruption in ommatidial packing (Kent et al., 2006; Zhang et al., 2006). The eye disc is unique in that the domain between Hh signaling and Hh receiving cells shifts with the movement of the morphogenetic furrow (Strutt and Mlodzik, 1997). As a consequence, cells must rapidly transition from expressing high levels of Ci to shutting off its expression.

Here, by engineering competition between full-length Ci and a truncated repressor-like form of Ci, Ci^Ce2, we show that Rdx preferentially targets full-length Ci. Moreover, we demonstrate that Su(fu) specifically protects full-length Ci from nuclear-localized Rdx (Zhang et al., 2006; Seong and Ishii, 2013; Liu et al., 2014), allowing it to successfully compete with Ci^Ce2. This reveals an as yet undescribed positive role for Su(fu). Together, our data reveal previously unreported regulatory processes that control Hh signaling via finely tuning Ci levels and activity.

In the Drosophila wing, the domain between longitudinal wing veins three (LV3) and four (LV4) is directly patterned by Hh signaling (Strigini and Cohen, 1997) and the location of the third wing vein primordium is determined by the range of Hh signaling (Biehs et al., 1998). This signaling process needs to be accurately regulated as the positioning of the third wing vein is precise to within one cell diameter (Abouchar et al., 2014).

To examine the role of anterior en expression on wing morphogenesis, we used a ptc-GAL4 driver to express two UAS-RNAi constructs targeting en and invected (inv) (inv is an inverted gene duplication of en) in the region of the anterior compartment adjacent to the compartment boundary (Fig. 1). As can be seen in Fig. 1C, the distance between LV3 and 4 is significantly reduced in animals in which the expression of en and inv has been attenuated in the cells of the anterior compartment that abut the anterior/posterior (A/P) boundary. A corresponding posterior shift in the position of the third wing vein primordium is also observed in the developing wing disc using antibodies to the Blistered (Bs) protein, which marks cells destined to make the intervein region (Fig. 2A1-A3,B1-B3,C,D). We also observed an analogous shift in the domain of dpp expression (Fig. 2A2,B2,C,D).

A decrease in the distance between LV3 and LV4 is generally considered to be a consequence of a reduced Hh gradient or to decreased Hh signaling. To test whether Hh signaling had been altered, clones mutant for en^E, a deletion mutation that removes both the en and inv genes, were generated, and the expression of ptc assayed. Surprisingly, the expression of ptc-lacZ is increased in clones mutant for en^E (Fig. 3A,B). ptc expression depends on Ci, and we also observed elevated levels of full-length Ci protein in the en mutant clones that are directly adjacent to the compartment boundary (Fig. 3A,B). This effect is associated with an increase in the expression of a transcriptional reporter of ci, ci^Dplac (Fig. 3C,D). Moreover, the domain of attenuated Ci corresponds precisely with the domain of anterior en expression, and shifts with it when en expression expands in response to activated Smo and is eliminated when Smo function is attenuated (Fig. S1).

Anterior expression of en is not observed in early third instar larvae but becomes prominent by late third instar (Blair, 1992). To examine the consequences of this temporal change in anterior en expression, we used a dual-color fluorescent transcriptional timer, UAS-TransTimer (UAS-TT) (He et al., 2019). In this system, a GAL4 UAS controls the expression of a bicistronic message encoding a destabilized GFP (half-life ∼2.1 h) followed by RFP (half-life ∼18.5 h). We assessed the dynamics of ci expression by using ci-GAL4 to drive UAS-TT and followed the GFP/RFP ratio in the domain of ci expression. Mid-third instar larvae with ci-GAL4 driving the expression of UAS-TT showed comparatively uniform distributions of destabilized GFP and RFP throughout the whole anterior compartment, indicating stable expression of ci (Fig. 4A,C). In late-third instar larvae, the distribution shifts, with lower levels of destabilized GFP relative to RFP close to the compartment boundary where en expression is activated (Fig. 4B,D). This late shift reflects decreased ci expression along the compartment boundary where anterior en expression is activated.

Together, these results show that loss of anterior compartment expression of en both increases the expression of ci and of its target ptc and reduces the LV3-LV4 spacing. They suggest that the control of ci expression along the A/P boundary by the anterior Hh-dependent expression of en finely tunes the Hh gradient and/or transduction by controlling the levels of its receptor Ptc.

Next, we examined whether anterior compartment expression of rdx along the compartment boundary could also affect wing patterning and Hh signaling. A UAS-RNAi line targeting rdx was expressed in the anterior cells adjacent to the compartment boundary using ptc-GAL4. Similar to what was suggested with unmarked rdx⁵ wing clones (Kent et al., 2006), downregulation of rdx along the compartment boundary has little effect on the positioning of LV3 (compare Fig. 1D with Fig. 1A,B) or its primordium (Fig. S2A,B). Note that expression of this UAS-RNAi-rdx line was able to knock down rdx expression, as it suppressed the phenotype induced by overexpression of UAS-rdx-myc (Fig. S2C,D,E).

Previous studies in mouse have shown that the mammalian homologue of Rdx, SPOP, specifically targeted the full-length forms of Gli2 and Gli3 (Ci homologues) but not the Gli3 repressor, and that loss of Su(fu) destabilizes full-length Gli2 and Gli3 (Wang et al., 2010). Therefore, we examined whether Rdx affected the competition between the full-length Ci and a repressor-like form of Ci, and what role Su(fu) might have in modulating those effects. For this purpose we used the ci^Ce2 mutation, which encodes a truncated Ci protein of 975 amino acids that mimics the Ci repressor as it is missing the CBP binding site for transactivation (Chen et al., 2000) and the C-terminal Su(fu) binding site (Croker et al., 2006; Han et al., 2015; Oh et al., 2015). In heterozygous flies, Ci^Ce2 protein is present throughout the anterior compartment, including the domain of high-level Hh signaling where it can compete with full-length Ci. As a consequence, enhancers with perfect Ci binding sites, as in the reporter 4bs-lacZ (Hepker et al., 1999), are silenced even in the domain of high Hh signaling adjacent to the compartment boundary (Fig. S3B). Nonetheless, ci^Ce2/+ mutants usually have normal wing patterning (compare Fig. 1E with Fig. 1A), but overexpression of rdx in the ci^Ce2/+ sensitized background leads to a phenocopy of a moderately strong reduction in Hh signaling with a complete fusion between LV3 and LV4 (Fig. 1F). This effect correlates with a dramatic decrease in the expression of Hh target genes, as can be seen with the expression ptc-lacZ when rdx is overexpressed in the dorsal compartment of ci^Ce2/+ wing discs (Fig. S4A).

An opposite effect on wing vein patterning is observed when rdx is overexpressed in a wild-type background, where it leads to a mild expansion of the domain between LV3 and LV4 (Fig. S2D). These opposite outcomes are presumably a consequence of competition between full-length Ci and the repressor-like Ci^Ce2 in the ci^Ce2 heterozygotes. Overexpression of rdx in a wild-type background results in a modest reduction of ptc (Kent et al., 2006; Zhang et al., 2006), allowing Hh to reach further into the anterior compartment, expanding the domain between LV3 and LV4 while also attenuating dpp expression, leading to a smaller wing and a decrease in the domains anterior to LV3 and posterior to LV4. In contrast, overexpression of Rdx in a ci^Ce2/+ background results in a dramatic decrease in Hh target gene expression, eliminating the domain between LV3 and LV4 (Fig. 1F, Fig. S4).

Because previous overexpression experiments had shown that Su(fu), which binds Ci, could partially protect it from Rdx-mediated degradation (Zhang et al., 2006), we examined whether there was a genetic interaction between Su(fu) and ci^Ce2. As seen in ci^Ce2/+, Su(fu) mutants have normal wings (Préat, 1992). In Su(fu)/+; ci^Ce2/+ animals, the distance between LV3 and LV4 is reduced and the wing veins fuse at the site of the anterior cross vein (Fig. 1G). In Su(fu); ci^Ce2/+ animals, the wings are reduced and show a strong/total fusion between LV3 and LV4 (Fig. 1H). These changes in wing patterning are also observed in the developing wing disc in Su(fu)/+; ci^Ce2/+ animals, where the LV3/4 intervein region is less distinct, and in Su(fu); ci^Ce2/+ animals, where it is missing (Fig. S5).

To examine the effects of loss of Su(fu) on the competition between full-length Ci and Ci^Ce2, clones lacking Su(fu) were generated in a ci^Ce2/+ background. As can be seen in Fig. 5A-D, loss of Su(fu) in this genetic context leads to a dramatic reduction in the expression of both ptc-lacZ and dpp-lacZ. By contrast, loss-of-function Su(fu) clones in a wild-type background had little effect on the expression of these target genes (Fig. S6A,B).

In summary, these data reveal that overexpression of rdx or loss of Su(fu) interact genetically with ci^Ce2 in a similar fashion, likely by shifting the competition between wild-type full-length Ci and the truncated Ci^Ce2 protein in favor of Ci^Ce2.

The similar effects of Su(fu) loss-of-function and rdx overexpression in ci^Ce2 heterozygotes suggest that one role of Su(fu) is to protect full-length Ci from Rdx. To test this hypothesis, clones double mutant for Su(fu)^LP and rdx⁵ were generated in a ci^Ce2/+ background and the expression of dpp-lacZ assayed. As can be seen in Fig. 5E, clones double mutant for Su(fu) and rdx show little to no effect on dpp expression, which corresponds to the phenotype of rdx mutant clones. This contrasts with the strong effect of single mutant Su(fu) clones and shows that loss of rdx can suppress the effect of Su(fu) in a ci^Ce2/+ background. In conclusion, these results show that Su(fu) and Rdx can modulate the competition between the full-length and repressor forms of Ci.

Previous work has shown that there are both N-terminal and C-terminal domain binding sites for Su(fu) in the Ci protein (Croker et al., 2006; Han et al., 2015; Oh et al., 2015). We have also shown that full-length Ci brings Su(fu) with it into the nucleus and that the Ci repressor does not (Sisson et al., 2006). We decided to revisit this experiment and specifically compare the nuclear import of Su(fu) in relation to that of full-length Ci and Ci^Ce2.

To increase the levels of endogenous Ci in the region directly adjacent to the compartment boundary for better visualization, we downregulated en and inv along the compartment boundary. As full-length Ci contains both nuclear localization and nuclear export signals and shuttles in and out of the nucleus, it is necessary to block CRM1 (Emb)-mediated nuclear export with Leptomycin B (LMB) to observe nuclear accumulation of full-length Ci in response to Hh signaling. Under these conditions, full-length Ci accumulates in the nucleus, as expected, along with Su(fu) in the entire domain of Hh signaling, including the domain of high level Hh signaling directly adjacent to the compartment boundary (Fig. S7A). Away from the domain of Hh signaling where full-length Ci is cytoplasmic, but Ci repressor is nuclear, Su(fu) remains primarily cytoplasmic (Fig. S7A). In animals heterozygous for ci^Ce2, Ci and Su(fu) are nuclear in the domain of Hh signaling as both full-length Ci and Ci^Ce2 enter the nucleus. However, away from the boundary, Ci^Ce2 is still able to enter the nucleus in the absence of Hh, whereas Su(fu) remains cytoplasmic (Fig. S7B). This nuclear localization of Ci^Ce2 away from Hh signaling is presumably due to failure of Cos2, which is known to normally retain Ci in the cytoplasm in the absence of Hh (Chen et al., 1999; Wang and Holmgren, 2000), to sequester Ci^Ce2 in the cytoplasm due to the absence of the C-terminal Cos2 binding site (called CORD) in Ci^Ce2 (Wang and Jiang, 2004).

In summary, these results show that Su(fu) is in the nucleus with full-length Ci throughout the domain of Hh signaling, but not with Ci^Ce2 nor presumably the repressor form. This raises the possibility that Su(fu) could protect full-length Ci, but not Ci^Ce2, nor presumably the repressor form, from Rdx.

One example in which full-length Ci must be cleared from the nucleus is in the developing eye disc. With the passage of the morphogenetic furrow, cells must shut off ci for proper patterning (Lee and Treisman, 2002). Clearance of Ci posterior to the furrow is mediated by Cul3 (Ou et al., 2002) and Rdx (Kent et al., 2006; Zhang et al., 2006).

To examine whether a similar situation occurs in the wing disc, we again made use of the UAS-TT transcriptional timer (He et al., 2019), but in this case monitored the dynamics of ptc expression using ptc-GAL4. In comparison with the cells just along the compartment boundary, cells further away from the compartment boundary have a higher ratio of RFP relative to destabilized GFP, indicating that ptc expression is shutting off in these cells (Fig. 6A,C). This effect can be seen in individual cells at the anterior edge of the ptc expression stripe (Fig. 6E), where the boxed cell has a higher level of RFP expression versus a boxed cell abutting the compartment boundary that has higher levels of destabilized GFP (Fig. 6F). To assess the role of rdx in the dynamics of ptc downregulation away from the compartment boundary, ptc-GAL4 was used to drive the expression of both UAS-TT and UAS-RNAi-rdx. Unlike the wild-type situation, RNAi knockdown of rdx leads to a more persistent expression of destabilized GFP in cells away from the compartment boundary (Fig. 6B,D).

In this study we examined the roles of three potential negative regulators of Hh signal transduction, two of which are themselves encoded by Hh target genes. In each case we discovered interesting new aspects about the pathway's regulation.

Anterior expression of en in the wing imaginal disc was first observed 30 years ago and en is the Hh target gene requiring the highest level of Hh signaling. Its domain of expression exactly correlates with a region of lower full-length Ci protein levels. It had been proposed that the lower Ci protein levels are a consequence of Ci being particularly active and labile in this region (Ohlmeyer and Kalderon, 1998). We show here that the lower levels of Ci are not primarily due to it being particularly labile, but rather are a consequence of negative transcriptional regulation by En. The role of this negative feedback loop appears to be to modulate the Hh gradient by downregulating the expression of ptc in addition to its effects on dpp (Strigini and Cohen, 1997). This leads to Hh signaling extending further into the anterior compartment, with a corresponding anterior shift in the location of LV3 and the expression of dpp. We prefer a model in which the attenuation of ptc expression by anterior En is indirect via Ci (Fig. 7), but in principle En could also directly negatively regulate ptc. We feel this is less likely as, ‘flip-out’ clones expressing Ci activate high levels of ptc in the posterior compartment in the presence of En (Hepker et al., 1997). The anterior expression of en occurs late in third instar larvae (Blair, 1992), which correlates with the downregulation of ci expression as visualized using the UAS-TT transcriptional timer (Fig. 4B,D) and the refinement of wing vein specification (Blair, 2007).

Why did this mechanism evolve to modulate the Hh gradient? Morphogen gradients, by virtue of their central roles in the development of multiple tissues, must be robust and resistant to perturbation. Therefore, to specifically expand the range of the Hh gradient in the wing disc a new component was added, anterior expression of the ci repressor En.

The lack of the C-terminal domain in the Ci repressor has multiple consequences. It loses the binding site for the co-activator CBP (Chen et al., 2000), and it loses C-terminal binding sites for Su(fu) (Croker et al., 2006; Han et al., 2015; Oh et al., 2015), Cos2 (Wang and Jiang, 2004) and Rdx (Zhang et al., 2009). As a consequence, the Ci repressor is not sequestered in the cytoplasm by Cos2 in the absence of Hh signaling and enters the nucleus without Su(fu), whereas full-length Ci enters the nucleus only in the presence of Hh signaling and as a complex with Su(fu).

In order to better understand the roles of Su(fu) and Rdx, we examined animals heterozygous for the ci^Ce2 mutation. In this context, overexpression of rdx or loss of Su(fu) function leads to a complete fusion between LV3 and LV4. In addition, clones mutant for Su(fu) show dramatic reduction in the expression of the Hh target genes ptc and dpp. These results show that Su(fu) has a potential novel positive role in Hh signal transduction, improving the ability of full-length Ci to compete with the repressor form (Fig. 7). A positive role for Su(fu) has also been found in mammals where Su(fu) appears to function as a chaperone for the full-length Gli proteins, but not the repressor forms, and is required for full activation of Gli target genes (Oh et al., 2015; Zhang et al., 2017). The requirement for Drosophila Su(fu) is obviated in the absence of Rdx, suggesting that Rdx primarily targets full-length Ci and not Ci repressor, even though the repressor is not protected by Su(fu) (Fig. 7). These results are analogous to what is seen with the mammalian homologue of Rdx, SPOP, indicating that this mechanism has been conserved during evolution. SPOP is opposed by Su(fu) and degrades the full-length forms of the mammalian GLI2 and GLI3 but not the GLI3 repressor form (Wang et al., 2010). The competition between Rdx and Su(fu) appears to be rather finely balanced as either increasing the expression of rdx or reducing the expression of Su(fu) enhances the ability of Ci^Ce2 to compete with full-length Ci. This function of protecting full-length Ci from Rdx presumably takes place in the nucleus, as this is where the Rdx protein primarily localizes (Zhang et al., 2006; Seong and Ishii, 2013; Liu et al., 2014).

However, the functional relevance of rdx being an Hh target gene has been unclear. Zygotic loss of rdx in the embryo has no visible effect on segmental patterning of the cuticle (Kent et al., 2006) and, unlike en, knockdown of rdx along the compartment boundary in the wing disc has little effect on wing patterning. Perhaps its role is to clear full-length Ci from cells that were once within the domain of Hh signaling and have moved outside the domain of Hh signaling. Perdurance of Rdx could target full-length Ci in the nucleus allowing the Ci repressor to shut off Hh target genes. This is the situation in the eye disc with the progression of the morphogenetic furrow. Cells that recently received high level Hh signaling and activated Ci must now downregulate Ci to allow proper differentiation of the ommatidia. Rdx appears to be important for this process, as loss of rdx leads to defects in the eye (Kent et al., 2006; Zhang et al., 2006). A similar situation may exist in other tissues. Looking at the temporal regulation of ptc expression with UAS-TT, cells removed from the compartment boundary in the wing disc have lower levels of destabilized GFP relative to RFP and appear to be in the process of shutting off ptc. This distinction is lost following downregulation of rdx by RNAi.

In the domain of modest level Hh signaling (in which dpp is expressed), both full-length Ci and Ci repressor must be present in some form of reciprocal gradients. In this domain, enhancers with perfect Ci consensus binding sites are silent due to binding of Ci repressor. The dpp enhancer with imperfect Ci binding sites is expressed, and for it to be completely active, full-length Ci must be bound (Müller and Basler, 2000). Why is full-length Ci able to better compete with Ci repressor for the imperfect binding sites? Full-length Ci and the Ci repressor share the same DNA binding domain, and it would be expected that the repressor would outcompete full-length Ci for binding to target sites because the repressor is primarily nuclear, whereas full-length Ci is primarily cytoplasmic, even in the presence of Hh signaling, due to a strong nuclear export signal (NES). In Parker et al., 2011, they suggest that cooperativity between Ci repressor proteins at perfect Ci binding sites can account for this distinction. Another potential mechanism for preferentially recruiting full-length Ci to imperfect binding sites might be suggested by the different protein interactions observed with full-length Ci and Ci^Ce2. Full-length Ci enters the nucleus with Su(fu) while the Ci repressor is not bound to Su(fu). In addition, the Ci repressor is missing the CBP binding site. As a consequence, full-length Ci could engage in protein-protein interactions with other transcription factors that are not available to the Ci repressor. This added affinity to other proteins within the enhanceosome could allow the preferential recruitment of full-length Ci to enhancers with imperfect Ci binding sites. Differential protein-protein interactions may also explain why full-length Ci is still able to activate ptc-lacZ expression along the compartment boundary in ci^Ce2/+ heterozygotes (Fig. S4) but not the artificial enhancer 4bs-lacZ (Fig. S3). The ptc-lacZ enhancer is a bona fide Drosophila enhancer and is likely to recruit a constellation of proteins that could interact with full-length Ci, whereas protein-protein interactions are likely to be much less robust at 4bs.

In conclusion, these results highlight the complexity of Hh signal transduction and its modulation. Expressing en in the anterior compartment of the wing pouch modulates the Hh gradient, whereas Su(fu) has a surprising positive role in the pathway, acting to partially protect full-length Ci from the E-3 ligase Rdx that Ci activates.

Most of the Drosophila lines were obtained from the Bloomington Drosophila Stock Center. The RNAi lines to en, inv and rdx were obtained from the Vienna Drosophila Resource Center. en^E and ci^Dplac were obtained from T. Kornberg (Cardiovascular Research Insttiute, University of California San Francisco, CA, USA); ptc-lacZ, UAS-rdx-myc, FRT82B rdx⁵ e, and FRT82B Su(fu)^LP rdx⁵ were obtained from J. Hooper (Department of Cell and Developmental Biology, University of Colorado Denver, CO, USA); and UAS-TT was obtained from N. Perrimon (Department of Genetics, Harvard Medical School, MA, USA). 4bs-lacZ and ci-GAL4 (Croker et al., 2006) were generated in our lab.

Mouse anti-GFP was obtained from Roche (1:750; 11 814 460 001); mouse anti-Blistered/DSRF was provided by S. Blair (Department of Integrative Biology, University of Wisconsin Madison, WI, USA) (1:750; 61385, Active Motif); mouse anti-Su(fu) (1:5; 25H3), mouse anti-Myc (1:10; 9E10), and mouse anti-En (1:10; 4D9) were obtained from the Developmental Studies Hybridoma Bank; and rat anti-Ci (2A1) (1:1000) (Motzny and Holmgren, 1995) and rabbit anti-β-Galactosidase (1:75,000) (Hepker et al., 1999) were generated in our laboratory. Note that the 2A1 anti-Ci antibody recognizes an epitope present in full-length Ci and Ci^Ce2, but that epitope is missing from the Ci repressor.

Adult flies were collected in ethanol and the wings dissected. After two washes in isopropyl alcohol, wings were mounted in a mixture of Canada balsam and methyl salicylate and incubated overnight at 65°C. Wings were imaged with a Nikon Optiphot using brightfield illumination.

Unless otherwise noted, eggs were collected for 24 h, allowed to develop for an additional 48 h and then heat shocked for 1 h at 35°C. Antibody stainings were performed as in Carroll and Whyte (1989) and image analysis was carried out using ImageJ. In order to recover Su(fu) rdx double mutant clones (Fig. 5E), it was necessary to give them a growth advantage (the Minute technique) by including a mutation [M(3)w] on the homologous chromosome arm that disrupts the gene encoding the S3 ribosomal protein.

Discs expressing UAS-TT were dissected in PBS, fixed on ice for 10 min in PBS+10% glycerol and 0.37% formaldehyde, rinsed in PBS+10% glycerol and mounted in Vectashield. Imaginal discs were imaged using either a Nikon Yokogawa spinning disk system or a Zeiss LSM 710 confocal microscope. Image analysis was carried out using ImageJ.

We thank Tom Kornberg, Joan Hooper, Seth Blair and Norbert Perrimon for Drosophila stocks and reagents, the Biological Imaging Facility at Northwestern University and the ImagoSeine core facility of Institut Jacques Monod, member of France-BioImaging (ANR-10-INBS-04) for access and support of microscopy. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study. Antibodies from the Developmental Studies Hybridoma Bank were developed under the auspices of the Eunice Kennedy Shriver National Institute of Child Health and Human Development and maintained by the University of Iowa.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200159.