Genome-wide identification of phospho-regulators of Wnt signaling in Drosophila

The canonical Wnt signaling pathway is evolutionarily conserved and regulates essential biological processes such as cell fate specification, proliferation and migration during metazoan development. As a consequence, aberrant Wnt signaling can result in diverse human developmental disorders and cancers (MacDonald et al., 2009; Wodarz and Nusse, 1998). Cells use the reversible phosphorylation of proteins to control the state and amplitude of signaling pathways (Cohen, 1992; Hunter, 1995; Salazar and Höfer, 2009). In the silent state of the Wnt pathway, the transcriptional effector β-catenin is phosphorylated within a cytosolic Axin (Axn) complex by Casein Kinase 1α (CK1α) and Glycogen Synthase Kinase 3β (GSK3β). This modification facilitates its poly-ubiquitination and degradation (Aberle et al., 1997; Amit et al., 2002; Kitagawa et al., 1999; Liu et al., 1999, 2002). In the absence of stabilized β-catenin, the DNA-binding protein TCF represses target gene expression (Cavallo et al., 1998; Roose et al., 1998). Binding of secreted Wnt ligand to its transmembrane Frizzled (Fz) receptor and LRP co-receptor initiates pathway activity through the recruitment of the Axn complex to the cell surface (Cliffe et al., 2003; Tamai et al., 2004; Umbhauer et al., 2000; Wong et al., 2003; Yang-Snyder et al., 1996). This induces the phosphorylation of LRP by CK1α, GSK3β and the plasma membrane-associated CK1γ (Davidson et al., 2005; Tamai et al., 2004; Zeng et al., 2005), which then reciprocally facilitates the disassembly of the Axn complex to prevent β-catenin degradation (Cselenyi et al., 2008; Kim et al., 2013; Mi et al., 2006; Willert et al., 1999). Stabilized β-catenin translocates to the nucleus to form a transcriptional complex with TCF to direct expression of target genes (Behrens et al., 1996; Molenaar et al., 1996; van de Wetering et al., 1997). Preceding these events in signal-receiving cells, the secretion, diffusion and reception of Wnt itself is phospho-regulated in ligand-producing cells and in the extracellular environment (Buechling and Boutros, 2011).

Although several kinases and phosphatases are known to regulate Wnt signaling (Gao et al., 2014; Verheyen and Gottardi, 2010), our current knowledge of these enzymes remains incomplete. In fact, differential phospho-proteome analyses have identified both novel phospho-regulators and novel phosphorylation sites on known regulators of the pathway (Bodenmiller et al., 2007; Tang et al., 2007). It is unlikely that the reversible phosphorylation of all of these newly identified phospho-epitopes is catalyzed by the current subset of known kinases and phosphatases of Wnt signaling, thereby suggesting that there are additional unidentified phospho-regulators of the pathway.

Several large-scale in vitro screening analyses have been performed in various Drosophila and mammalian cell lines to identify regulators of the Wnt pathway. These high-throughput loss- and gain-of-function studies quantitated the response of exogenous Wnt pathway reporters under conditions of elevated signaling (Anton et al., 2011; Buechling et al., 2011; Caspi and Rosin-Arbesfeld, 2008; DasGupta et al., 2005; Firestein et al., 2008; Groenendyk and Michalak, 2011; Jacob et al., 2011; James et al., 2009; Kategaya et al., 2009; Major et al., 2008; Miller et al., 2009; Port et al., 2011; Tang et al., 2008). In this study, we have performed a comprehensive genome-wide in vivo RNAi screen for kinases and phosphatases in Drosophila to build a phospho-regulatory network of the Wnt pathway. Drosophila has significantly contributed to our understanding of the molecular mechanism of the Wnt pathway (Bejsovec, 2006), and has low functional redundancy but high functional conservation of the genes with humans (Fortini et al., 2000; Reiter et al., 2001). The strength of our approach is that we assayed the effects of putative phospho-regulators of the Wnt pathway on endogenous targets under physiological levels of signaling in an intact tissue. Wing discs from third instar larvae were immunostained against direct high- and low-threshold targets of the pathway, allowing us to evaluate the entire ligand-induced gradient of signaling (Barolo, 2006). As the wing disc comprises distinct cells that either produce or respond to the ligand, we could identify phospho-regulators at discrete positions in the pathway. Our analyses of the Wnt pathway have yielded a large subset of high-confidence kinases and phosphatases, including 22 previously unidentified regulators of signaling. Among candidate Wnt regulators, we determined which acted indirectly through the Notch pathway to modulate wg transcription. Furthermore, we established those shared between the Wnt and Hh pathways. Last, we validated two dual regulators of the Wnt and Hh pathways using biochemical assays to demonstrate that their roles are evolutionarily conserved from Drosophila to mammalian cells.

In the wing imaginal disc, Wingless (Wg) (Drosophila Wnt) produced from cells at the dorsal/ventral (D/V) compartment boundary induces the nested expression of pathway target genes senseless (sens) (high-threshold) and Distal-less (Dll) (low-threshold) in adjacent non-boundary cells that receive the ligand (Neumann and Cohen, 1997; Zecca et al., 1996) (Fig. 1). We compiled a list of all protein kinases and phosphatases, non-protein kinases and phosphatases, as well as factors that associate with these enzymes, such as cyclins and regulatory subunits, which we will collectively refer to as the kinome and phosphatome (supplementary material Figs S1, S2). Using transgenic RNAi libraries of UAS-driven inverted repeats (IRs) and tissue-specific Gal4 drivers, we knocked down in a spatially restricted fashion the expression of each of 385 and 205 genes present in the Drosophila kinome and phosphatome, respectively, to assay their effect on Wnt signaling (Fig. 1). At least two non-overlapping IRs per gene were independently tested to minimize positive and negative false discovery.

The primary screen was performed using a combination of decapentaplegic (dpp)-Gal4 and hedgehog (hh)-Gal4 that are expressed along the anterior/posterior compartment boundary and in the posterior compartment of the wing disc, respectively. These Gal4 drivers were used to knock down gene function in both the ligand-producing and ligand-receiving cells (Fig. 1). A UAS-dicer-2 transgene was used in combination with the Gal4 drivers for all screening analyses to enhance the efficiency of RNAi-mediated gene knockdown (Dietzl et al., 2007). Wing discs from every genotype were immunostained against Sens, Dll and Wg to assay for pathway activity and status of the ligand. A subset of known regulators of the Wnt pathway was knocked down in a pilot screen to validate our experimental design (supplementary material Fig. S3). We consistently observed reproducible and expected defects in pathway targets (and Wg protein in certain cases) with known components of signaling, although some effects were subtle, yet highly penetrant and reproducible (e.g. supplementary material Fig. S3).

The wing disc is specified by Wnt and other signaling pathways (Couso et al., 1994; Rulifson et al., 1996), and disruption of this process may induce compensatory mechanisms that form a relatively normal adult wing (Herrera et al., 2013; Ryoo et al., 2004; Wells et al., 2006). Discs were therefore immunostained from every genotype irrespective of the presence of an adult wing or other phenotype. A gene that modified the levels of Sens and/or Dll when knocked down independently with at least two non-overlapping IRs using any combination of Gal4 drivers was classified as a candidate of the Wnt pathway (examples of modifiers are shown in Fig. 2). We identified 90 candidates from the primary screen (Fig. 3). Scoring of all crosses with multiple RNAi lines is provided in supplementary material Figs S4, S5. No gene, when knocked down, modified the levels of only Sens or Dll, although in some cases one pathway target was affected to a greater extent than the other (Figs 2, 3). Of note, most candidates had mild to moderate effects on targets, consistent with results obtained in our pilot screen with known Wnt pathway components, suggesting that under these assay conditions relatively subtle effects are valid, as confirmed by further analyses.

Nineteen of the 90 candidates had no observable effect on Wg levels or distribution, and thus do not function upstream of or at the level of the ligand-receptor interaction to affect the secretion, diffusion or reception of Wg. These candidates were classified as Group I high-confidence regulators of the Wnt pathway that function downstream of the ligand-receptor interaction in the ligand-receiving cells (Figs 2, 3). Group I includes known regulators of signaling such as Wnk (Serysheva et al., 2013) and novel regulators such as Myosin binding subunit (Mbs), Cyclin A (CycA), MYPT-75D, corkscrew (csw) and Cdc37 (Fig. 2A-G).

Seventy-one of the 90 candidates modified the levels of Wg (Fig. 2; supplementary material Figs S4, S5). The Notch pathway signals from non-boundary cells to induce the expression of wg in cells at the D/V boundary of the wing disc (Diaz-Benjumea and Cohen, 1995; Neumann and Cohen, 1996; Rulifson and Blair, 1995; supplementary material Fig. S6). To test whether the candidates that affect Wg levels do so as a result of their regulation of the ligand after translation or by affecting expression of the ligand via the Notch pathway, they were re-analyzed in a secondary screen. To address transcriptional regulation by Notch signaling, we determined whether they also modified the levels of another Notch pathway target, Cut (Ct) (Micchelli et al., 1997; Fig. 1). Only three of the 71 candidates when knocked down with either dpp-Gal4 or hh-Gal4 had no effect on Ct levels. We inferred that these three candidates have no effect on the Notch pathway and wg expression, but rather modify the secretion, diffusion or reception of Wg, which was detected as a change in its levels. These candidates were classified as Group II high-confidence regulators of the Wnt pathway that function upstream of or at the level of the ligand-receptor interaction in the ligand-producing cells or extracellular environment, respectively (Figs 2, 3). Group II includes a known regulator of signaling, myopic (mop) (Miura et al., 2008; Pradhan-Sundd and Verheyen, 2014), and a novel regulator, wunen2 (wun2) (Fig. 2H-I). As further evidence of the functional role of Group II candidates, we and others have confirmed that two of the three Group II candidates (CG3530, Mop) have no effect on wg-lacZ but do affect the Wg protein (Silhankova et al., 2010; Pradhan-Sundd and Verheyen, 2014).

The remaining 68 of the 71 candidates modified Ct levels and were inferred to regulate wg expression as a result of their effect on multiple Notch targets (Fig. 3). These candidates were further tested to determine whether they also affected the Wnt pathway independently of their regulation of wg expression. The majority of regulators of developmental pathways function in the ligand-receiving cells and not the ligand-producing cells. Therefore, if a gene independently affects the Wnt and Notch pathways in the wing disc, it would likely do so in the ligand-receiving cells of both pathways, regulating Wnt signaling in the non-boundary cells and Notch signaling in the boundary cells. We knocked down the 68 candidates that we identified as regulators of the Notch pathway using the C5-Gal4 driver, which is expressed in only non-boundary cells adjacent to the D/V compartment boundary (Fig. 1). By knocking down gene function in only the ligand-receiving cells of the Wnt pathway, which correspond to the ligand-producing cells of the Notch pathway (supplementary material Fig. S6), we could distinguish candidates that independently regulate both Wnt and Notch signaling from those that regulate only Notch signaling in the wing disc. When knocked down in the non-boundary cells, 32 of the 68 candidates modified the levels of the Wnt pathway targets Sens and Dll, but not Wg, and thus did not affect the Notch pathway in these cells to regulate wg expression at the D/V compartment boundary. These 32 candidates were classified as Group III high-confidence regulators of the Wnt pathway that function downstream of the ligand-receptor interaction in the signal-receiving cells and that affect the Notch pathway independently (Figs 2, 3). Group III includes known regulators of signaling, such as string (stg) (Davidson et al., 2009) and gilgamesh (gish) (Davidson et al., 2005; Zhang et al., 2006), and a novel regulator, CG8878 (Fig. 2J-O). The remaining 36 candidates when knocked down in the non-boundary cells had no effect on the levels of Sens, Dll or Wg, and thus do not regulate the Wnt or Notch pathway in these cells (Fig. 3). Although these 36 candidates (supplementary material Fig. S7) regulate the Notch pathway and wg expression in the boundary cells, through our analyses we could not determine whether any of these candidates, albeit unlikely, also independently regulate Wg secretion from these cells to affect the Wnt pathway.

In summary, the 590 phospho-regulators screened in the wing disc yielded 90 candidates that were classified into 54 high-confidence regulators of the Wnt pathway and 36 candidates that regulate wg expression to indirectly affect the Wnt pathway. The high-confidence regulators were further classified into three phenotypic categories, referred to as Groups I, II and III, each of which comprise known and novel regulators of signaling (Fig. 4). Of the high-confidence regulators, 33 are kinases (and associated factors) and 21 are phosphatases (and associated factors). Forty-five of the 54 high-confidence regulators promote (blue in Fig. 4A), while the remaining nine inhibit (yellow in Fig. 4A), signaling.

Remarkably, 32 of the 90 candidates are known regulators of the Wnt pathway (indicated by gray shading in Fig. 4). This large subset of known regulators is reflective of the robustness and low false-negative error rate of our screen design. Moreover, based on our unbiased genetic analyses, all 32 of these known regulators cluster together in the category of high-confidence Groups I, II and III regulators. This strongly suggests that at least some of the 22 novel high-confidence regulators identified are bona fide regulators of the Wnt pathway. We classified the high-confidence regulators of the Wnt pathway according to their respective kinase and phosphatase groups. The Drosophila kinome comprises 13 groups (defined in supplementary material Fig. S1), of which 12 are represented by the high-confidence regulators of the Wnt pathway (Fig. 4B; supplementary material Fig. S1). The Drosophila phosphatome comprises nine groups (defined in supplementary material Fig. S2), of which six are represented by the high-confidence phosphatases (and associated factors) of the Wnt pathway (Fig. 4C; supplementary material Fig. S2).

The Hh pathway has diverse functions during metazoan development, such as the regulation of organogenesis and stem cell homeostasis (Varjosalo and Taipale, 2008). Similar to the Wnt pathway, the Hh pathway is also subject to reversible phosphorylation in its silent and active states. In the absence of signaling, the transcriptional effector GLI is phosphorylated by Protein Kinase A (PKA), GSK3β and CK1α within a cytosolic Kif7 complex (Chen et al., 1998; Jia et al., 2002, 2005; Price and Kalderon, 1999, 2002; Zhang et al., 2005). The phosphorylation of GLI triggers its poly-ubiquitylation and partial degradation to yield a truncated form of the protein that represses target gene expression (Aza-Blanc et al., 1997; Jia et al., 2005; Jiang and Struhl, 1998; Méthot and Basler, 1999; Smelkinson and Kalderon, 2006; Smelkinson et al., 2007; Tempé et al., 2006). Binding of the Hh ligand to its receptor Patched (Ptc) recruits the Kif7 complex to the transmembrane signal transducer Smoothened (Smo) (Lum et al., 2003; Ogden et al., 2003; Ruel et al., 2003; Zheng et al., 2010). This induces Smo phosphorylation by PKA and CK1α, which promotes its accumulation at the cell surface (Apionishev et al., 2005; Denef et al., 2000; Jia et al., 2004; Zhang et al., 2004; Zhu et al., 2003). The interaction between Kif7 and phospho-Smo disassembles the complex to stabilize full-length GLI that directs target gene expression (Aza-Blanc et al., 1997; Jia et al., 2003; Liu et al., 2007; Ohlmeyer and Kalderon, 1998). Although the evolutionary relationship between Wnt and Hh signaling remains unclear, these pathways have a similar phospho-regulatory mechanism of signal transduction and comprise similar or identical regulators that exert the same effect on signaling by functioning at analogous levels of the relays (Kalderon, 2002).

We evaluated the ability of all 90 candidates of the Wnt pathway to regulate the Hh pathway in vivo. We performed this counter-screen to identify shared phospho-regulators that exert the same effect to either promote or inhibit signaling, and thereby potentially function at analogous levels of these pathways. Hh signaling stabilizes full-length Cubitus interruptus (Ci) (Drosophila GLI) to regulate expression of the target gene ptc along the anterior/posterior (A/P) boundary in the anterior compartment of the wing disc (Strigini and Cohen, 1997) (Fig. 5A). Each candidate was knocked down using MS1096-Gal4 (with UAS-dicer-2) and wing discs were immunostained to detect Ci and Ptc. The MS1096-Gal4 domain is in the center of the wing disc with stronger expression in the dorsal half relative to the ventral half (supplementary material Fig. S8A). The enhancer of ptc, unlike that of other Hh pathway targets, responds in a cooperative manner to both the levels and active state of Ci; thus, a change in the levels of Ci that modulates the amplitude of Hh signaling might not necessarily result in an effect on the expression of ptc (Parker et al., 2011). Any gene that modified the levels of Ci, but not necessarily Ptc, when independently knocked down with at least two non-overlapping IRs was classified as a regulator of the Hh pathway (Fig. 5B; supplementary material Fig. S8B). A limitation of this approach is that the counter-screen cannot identify regulators of the Hh pathway that do not affect the levels of Ci, but function downstream of its stabilization to modulate the expression of targets other than ptc. Sixty-six of the 90 candidates of the Wnt pathway modified the levels of Ci (and Ptc in most cases) to regulate the Hh pathway (supplementary material Figs S4, S5). The remaining 24 candidates of the Wnt pathway had no effect on the levels of Ci or Ptc (supplementary material Figs S4, S5). While the majority of Wnt regulators promoted signaling, 45 of 66 regulators of the Hh pathway inhibited signaling (supplementary material Fig. S8B). Nevertheless, 29 Wnt candidates exerted the same effect on the Hh pathway to either promote or inhibit signaling in the wing disc (Fig. 5B). Of these 29 dual regulators that we propose function at analogous levels of these pathways, 25 are high-confidence regulators and eight are novel regulators of the Wnt pathway. As proof of principle, we recovered hipk in our screen, which we have previously shown to be a dual regulator of Wnt and Hh acting on the E3 ubiquitin ligase Supernumerary limbs (Slimb) at analogous levels of the pathways (Swarup and Verheyen, 2011).

We identified Cdc37 and gish as novel negative and positive regulators, respectively, of the Hh pathway. Knockdown of a negative control, lacZ (supplementary material Fig. S8A), displayed normal levels of Ci and Ptc, as seen in wild-type tissue (Fig. 5C). Knockdown of Cdc37 resulted in the robust enhancement of Ci levels, both within and away from the signaling domain in the anterior compartment of the wing disc (Fig. 5D). Knockdown of Cdc37 consistently led to the distortion of the morphology of the disc, which precluded an accurate evaluation of Ptc levels. Based on our data, Cdc37 knockdown does not seem to appreciably alter Ptc levels, but we cannot definitively rule out this possibility. Cdc37 knockdown phenocopies loss-of-function mutants of other negative regulators of the Hh pathway that strongly enhance the levels of Ci but have minimal or no effect on Ptc levels, such as sgg (Drosophila GSK3β) (Jia et al., 2002; Price and Kalderon, 2002) and slimb (Wang et al., 1999). Knockdown of gish resulted in the reduction of Ci and a reduction in the levels of Ptc, as indicated by the width of its expression domain in the central region of the wing disc (Fig. 5E).

In our screening analyses, all 90 candidates of the Wnt pathway were evaluated against Notch (as part of the secondary screen) and Hh (as part of the counter-screen) pathways in the wing disc (Fig. 6A). This allowed us to distinguish between candidates that are specific to the Wnt pathway from those that are shared between the Wnt and other signaling pathways (Fig. 6B; Tabata and Takei, 2004). These analyses also allowed us to evaluate whether a candidate of the Wnt pathway does so only indirectly as a result of its effect on cell death, cell proliferation or non-specific gene transcription. For example, if knockdown of a candidate decreases Wnt signaling but does not affect or increases Notch and/or Hh signaling, we inferred that this candidate does not affect cell death or non-specific gene transcription. However, if knockdown of a candidate decreases Wnt, Notch and Hh signaling, it is possible that it indirectly regulates the Wnt pathway due to an effect on cell death or non-specific gene transcription. Twelve of the 90 candidates of the Wnt pathway when knocked down had the same effect (i.e. positive or negative regulator of both pathways) on both the Notch and Hh pathways (Fig. 6C). This does not necessarily imply that these 12 candidates indirectly regulate the Wnt pathway, but that further analysis of these candidates is required. Coincidentally, 11 of these 12 candidates have been previously described to regulate the Wnt pathway (Fig. 6C).

As proof of concept for dual regulators that operate at analogous levels of the Wnt and Hh pathways, we further characterized the functions of cdc37 and gish. Cdc37 is a chaperone that functions to regulate the folding and biogenesis of diverse kinases (Caplan et al., 2007). We identified cdc37 as a novel negative regulator whose knockdown resulted in increased pathway activity within the signaling domain and ectopic pathway activity outside the signaling domain (Fig. 2G). The Drosophila eye has frequently been used to screen for regulators of the Wnt pathway using a sevenless (sev)>wg (or equivalent) gain-of-function genetic background in which eye development is impaired (Fig. 7A; Greaves et al., 1999; Port et al., 2011). Heterozygosity for a loss-of-function allele of cdc37 caused an enhancement of the sev>wg phenotype (Fig. 7B), suggesting that reduction of cdc37 enhances the aberrant eye phenotype caused by Wg signaling. We examined embryos homozygous for the cdc37^eD4 allele. Wg signaling is required for specification of regions of naked cuticle between denticle belts in the cuticle (Fig. 7C). Enhanced Wg signaling caused loss of denticle belts, such as was seen in cdc37 mutant embryos, suggesting loss of cdc37 promotes Wg signaling (Fig. 7D,E). Somatic loss-of-function clones of cdc37 do not survive in the wing disc due to its requirement for cell viability (Lange et al., 2002). We therefore generated MARCM loss-of-function clones of cdc37 (positively marked with nuclear GFP) that overexpress the apoptosis inhibitor p35. These MARCM clones, although small in size, displayed a cell-autonomous upregulation of Wnt and Hh signaling, as indicated by an increase in Dll and Ci levels, respectively. Similar to the effect seen with RNAi (Figs 2G, 5D), ectopic Dll (Fig. 7F) and Ci (Fig. 7G) were observed in MARCM clones away from the signaling domains of these pathways. This result suggests that Cdc37 may normally destabilize the effectors to function as a negative regulator of signaling in the silent and active states of the Wnt and Hh pathways. Indeed, when we assayed for levels of stabilized Armadillo (Arm) (Drosophila β-catenin), we found more in wing discs reduced for cdc37 function compared with wild type (Fig. 7I). Although cdc37 RNAi did not significantly affect ptc expression, it did result in increased expression of another target gene of the Hh pathway, dpp-lacZ (supplementary material Fig. S9A). Although loss of cdc37 clearly stabilizes Ci, studies suggest that there is an additional step that regulates Ci import into nucleus. We propose that only a small amount of Ci enters the nucleus in cdc37 clones, which may explain why there are no apparent changes in ptc (which requires high-level Hh for expression), although we do observe a change in dpp, which is a low level target. Knockdown of cdc37 caused increased cell proliferation in discs, as indicated by increased levels of phospho-Histone H3 (PH3; supplementary material Fig. S9B), but this proliferation cannot solely explain the expression of targets in regions outside the normal domain of signaling. Thus, these results indicate that loss of cdc37 enhances both Wg and Hh signaling outputs.

To examine whether the function of Cdc37 is evolutionarily conserved, we carried out biochemical studies in mammalian cells. Knockdown of Cdc37 with siRNA in 3T3 cells strongly enhanced levels of stabilized β-catenin (Fig. 7J). Furthermore, knockdown of Cdc37 also induced signaling in unstimulated 3T3 cells, as measured using the Wnt pathway-specific TOPFLASH reporter, compared with mock-transfected 3T3 cells (Fig. 7K). Knockdown of Cdc37 also enhanced the levels of full-length GLI3 (Fig. 7L). We were unable to detect the truncated form of GLI3 in our assay. These effects of Cdc37 in mammalian cells mimic those of negative regulators of the Wnt and Hh pathways, such as GSK3β and CK1α. As Cdc37 is a kinase-associated chaperone, we propose that it functions to promote the stability of GSK3β and/or CK1α, which both constitutively destabilize the effectors of the Wnt and Hh pathways. Accordingly, cycloheximide-treated 3T3 cells reduced for Cdc37 displayed progressively lower levels of GSK3β due to its shorter half-life, compared with mock-transfected cells (Fig. 7M,N). We did not evaluate whether Cdc37 regulates the stability of CK1α in our assay. Thus, our analyses suggest that Cdc37 has a novel evolutionarily conserved function from Drosophila to mammalian cells to promote the stability of GSK3β and inhibit both the Wnt and Hh pathways.

Gish (Drosophila CK1γ) is a plasma membrane-associated kinase that has been described to promote the Wnt pathway by phosphorylating the co-receptor LRP (Davidson et al., 2005; Zhang et al., 2006). Consistently, it was recovered as a high-confidence Wnt regulator (Fig. 2L,O). Thus far, no role has been ascribed to this kinase in the regulation of the Hh pathway. Although knockdown of gish decreased the expression of targets of the Wnt, Notch and Hh pathways in the wing disc (Fig. 6A), the levels of Delta (Dl) and cleaved-Caspase 3 (Casp 3) were unaffected (supplementary material Fig. S9C). Thus, gish does not have a non-specific effect on gene transcription or cell death. Compared with wild type, a gish mutant wing disc displayed lower levels of Ci (Fig. 8A). Hh signaling stabilizes Ci to regulate expression of dpp-lacZ within the morphogenetic furrow (MF) of the eye disc (Fig. 8B). A somatic clone of gish in the eye disc had decreased levels of Ci and correspondingly dpp-lacZ expression (Fig. 8C), thereby confirming its role as a positive regulator of Hh signaling across multiple tissues.

We propose that Gish/CK1γ regulates the phosphorylation of the transmembrane protein Smo to promote the Hh pathway, analogous to its role in regulating the phosphorylation of LRP in the Wnt pathway. We tested this hypothesis using protein lysates from wing discs in a gel mobility shift assay (Fig. 8D). Knockdown of gish throughout the wing blade using 71B-Gal4 reduced the phosphorylation of Smo in disc protein lysates (as detected by a faster migrating band). As Smo phosphorylation is a prerequisite to its accumulation, knockdown of gish also decreased Smo levels. Conversely, overexpression of the Hh ligand, hh-N, expectedly enhanced the phosphorylation (as detected by a slower migrating band in lane 3) and levels of Smo (Fig. 8D). The simultaneous knockdown of gish in the presence of ectopic hh-N reduced the phosphorylation (migration) and levels of Smo, compared with ectopic hh-N alone (determined by densitometry), thus indicating that Gish acts downstream of the Hh ligand.

To determine whether the effect of Gish on the Hh pathway is evolutionarily conserved, the effect of CK1γ on signaling was assessed in mammalian cells. 3T3 cells transfected with CK1γ displayed increased levels of full-length GLI3, compared with mock-transfected cells (Fig. 8E). When exogenous CK1γ and Smo were co-transfected into 3T3 cells, they were detected in a complex in a co-immunoprecipitation assay (Fig. 8F). Furthermore, CK1γ robustly phosphorylated Smo at one or more serine and/or threonine residues (as detected by phospho-specific antibodies) in the presence of ATP in an in vitro kinase assay (Fig. 8G). Thus, Gish/CK1γ regulates the phosphorylation of Smo to promote the Hh pathway in Drosophila and mammalian cells.

Divergent disease states have been attributed to be a cause or consequence of aberrant protein phosphorylation (Reiter et al., 2001). Wnt signaling is phosphor-regulated both in its silent and active states, but thus far our understanding of kinases, phosphatases and associated factors of the pathway has been limited. In this study, we performed the first genome-wide in vivo screen under physiological conditions in the Drosophila wing disc for phospho-regulators of the Wnt pathway. We identified 54 high-confidence regulators, 22 of which are novel. The results of our analyses do not indicate whether a high-confidence regulator has a direct or indirect effect on signaling. However, as ∼60% of the high-confidence regulators identified have been previously validated to have a direct effect on Wnt signaling, we predict that at least some of the novel high-confidence regulators identified would also have a direct effect on the pathway. Indeed, subsequent analyses of Myopic revealed a novel role in regulating Wg secretion (Pradhan-Sundd and Verheyen, 2014). Although the mechanism and components of the Wnt pathway are for the most part conserved between Drosophila and humans, there are possibly vertebrate-specific phospho-regulators of signaling that would not have been identified in our analyses. Our dataset represents the largest list of putative phospho-regulators of the Wnt pathway identified to date, almost all of which have identified human orthologs (supplementary material Figs S1, S2) and are therefore likely to be functionally conserved.

As part of this study, we also established previously unknown relationships between the Wnt and Hh pathways in vivo by identifying 12 novel dual regulators that we propose function at analogous levels of signaling (Fig. 5B). As proof of concept, we biochemically characterized the roles of Cdc37 and Gish/CK1γ to demonstrate that their functions are conserved from Drosophila to mammalian cells. We also describe an initial analysis of candidate regulators of Notch signaling during wing disc development. Although these findings are preliminary, they highlight an emerging theme of phospho-regulation of Notch that likely hold parallels in vertebrate biology. The comparison of signaling pathways in vivo and the identification of specific versus shared phospho-regulators facilitate our understanding of human development and disease states.

The following Drosophila strains were used: w¹¹¹⁸ (wild type), dpp-Gal4/TM6B, C5-Gal4, MS1096-Gal4, omb-Gal4, 71B-Gal4, UAS-flp, UAS-dicer-2, UAS-p35, dpp-lacZ/CyO, Cdc37^eD4 FRT79/TM6B, FRT82,GFP/TM6B, MARCM79, hs-flp;;FRT82,GFP/TM6B and gish^EY15206 (Bloomington Drosophila Stock Center); hh-Gal4/TM6B and eyFlp;ey-Gal4,GMR-Gal4;sev>y+>wg (Port et al., 2011); UAS-hhN (Su et al., 2011); and FRT82 gish^e01759/TM6B (Gault et al., 2012). The transgenic RNAi strains used for the screens were obtained from the Vienna Drosophila RNAi Center (Dietzl et al., 2007), National Institute of Genetics and Harvard Transgenic RNAi Project (supplementary material Fig. S4, Fig. 5). The percentage of inverted repeats with predicted sequence-dependent off-target effects are as follows: 65% (0), 27% (1-2) and 8% (>2). All genes (at least two independent RNAi lines per gene) were first tested with the dpp-Gal4 driver. If a phenotype was only observed with one RNAi line for dpp-Gal4 or with two lines targeting the same region of the mRNA, the gene was re-tested with hh-Gal4. The one or more RNAi lines that were re-tested with hh-Gal4 were either the same lines that were used with dpp-Gal4 or additional lines (if new ones were available). If a RNAi line displayed a phenotype with both the dpp-Gal4 and hh-Gal4 drivers in the primary screen, the hh-Gal4 driver was used in the secondary screen. Wing discs from 20 larvae of each genotype were immunostained (30 larvae for genotypes that were non-homozygous with a balancer chromosome). Knockdown of no gene precluded analysis due to lethality before the third instar stage. Twenty flies of each genotype were scored for adult phenotypes. The penetrance of all phenotypes indicated is between 80 and 100%. False-positive results due to cell death, cell proliferation or non-specific gene transcription were evaluated through testing the ability of candidate regulators to affect different signaling pathways in the wing disc. cdc37 MARCM clones were generated by crossing the UAS-p35; Cdc37^eD4 FRT79/TM6B and MARCM79 strains, and progeny were heat-shocked 48 h after egg laying (AEL) for 2 h at 38°C. gish somatic clones in the wing disc were generated by crossing the UAS-flp;FRT82 gish^e01759/TM6B and omb-Gal4;;FRT82,GFP/TM6B strains. gish somatic clones in the eye disc were generated by crossing the hs-flp;;FRT82,GFP/TM6B and FRT82 gish^e01759/TM6B strains, and progeny were heat-shocked 48 h AEL for 2 h at 38°C.

Drosophila wing and eye discs from third instar larvae were dissected, immunostained and mounted according to standard procedures (Swarup and Verheyen, 2011). The following primary antibodies were used: anti-Ci 2A1 (1:50), anti-Ptc (1:50), anti-Wg 4D4 (1:100), anti-Ct 2B10 (1:75), anti-Delta C594.9B (1:50) (Developmental Studies Hybridoma Bank), anti-Phospho Histone H3 (1:100), anti-β-galactosidase (1:1500), anti-Cleaved Caspase 3 (1:100) (Cell Signaling Technology), anti-Sens (1:1000) (Nolo et al., 2000) and anti-Dll (1:400) (Duncan et al., 1998; Panganiban et al., 1995). Fluorescent secondary antibodies (1:400) were from Jackson Immunolabs. Wing disc images were obtained on a Nikon A1R laser scanning confocal microscope and all images are derived from stacked z-series. A fixed-size box (width=width of stabilized Ci domain in wild type, height=bottom of disc to top of wing pouch) was used to compare the integrated density of wild type and gish mutant. For quantification of immunostained discs, integrated density (pixel area multiplied by baseline subtracted intensity) was calculated.

NIH-3T3 cells (American Type Culture Collection) were cultured at 37°C in DMEM supplemented with 10% FBS (Invitrogen). 3T3 cells were transfected with CK1γ-FLAG (Davidson et al., 2005), Smo-MYC (Chen et al., 2011), TOPFLASH, FOPFLASH, or Renilla luciferase (Ishitani et al., 2003) using Polyfect Reagent (Qiagen) and harvested 36-48 h post-transfection. Knockdown of Cdc37 was performed with siRNA and Lipofectamine RNAiMAX Reagent (Invitrogen). As a negative control, cells were mock-transfected with empty vector. Cells were treated with cycloheximide (25 µg/ml) (Sigma) 24 h after transfection with Cdc37 siRNA [a mixture of two distinct siRNAs targeted to different parts of the genes from Ambion (136227 and 136228)] and harvested at the indicated time points.

Protein lysates were prepared from 3T3 cells and wing discs. Proteins of interest were immunoprecipitated from lysates using antibodies and Protein-G Sepharose beads (Sigma). The kinase assay with non-radiolabeled ATP/kinase assay buffer (Cell Signaling Technology) was performed with immunoprecipitated proteins at 30°C for 30 min. The de-phosphorylation assay was performed on protein lysate using lambda protein phosphatase at 30°C for 1 h (New England Biolabs).

Western blot analyses were performed using the following primary antibodies: anti-Tubulin (1:1000), anti-Myc (1:1000), anti-FLAG (1:1000) (Sigma), anti-GSK3β (1:500) (Abcam), anti-GLI3 N19 (1:200) (Santa Cruz Biotechnology), anti-Phosphoserine (1:500), anti-Phosphothreonine (1:500) (Sigma), anti-Smo 20C6 (1:10) and anti-Armadillo N27A1 (1:200) (Developmental Studies Hybridoma Bank). For quantification of western blot bands, integrated density (pixel area multiplied by baseline subtracted intensity) of each band was calculated within an appropriate exposure range and then normalized to the loading control.

Transcriptional assays were performed using TOPFLASH, FOPFLASH and Renilla (control) luciferase reporter plasmids. The experiment was performed using the Dual Luciferase Reporter Assay System (Promega). The value for each data point is the average of three individual experiments. The s.d. was calculated and Student's t-test was performed for statistical significance.

We are grateful to the numerous people who provided fly strains, plasmids and antibodies, especially Hugo Bellen, Ian Duncan, and Sean Carroll. We are also grateful to the RNAi stock centers (Vienna Drosophila RNAi Center, National Institute of Genetics and Harvard Transgenic RNAi Project) for generating the strains used in this study. Thanks to past and present members of the Verheyen lab for discussions, as well as to Hans Clevers, Roel Nusse and Nick Harden for comments on the study.