Ras-activated Dsor1 promotes Wnt signaling in Drosophila development

The Wnt signaling pathway [Wingless (Wg) in Drosophila] plays a crucial role in all metazoans during growth and proliferation, cell fate determination and tissue homeostasis (Clevers and Nusse, 2012). Wnt signaling does not control all these processes independently, rather this is achieved by extensive interaction with other signaling pathways, leading to synergistic or antagonistic effects normally resulting in desirable biological outcomes (reviewed in Zeller et al., 2013; Kim and Jho, 2014; Itasaki and Hoppler, 2010; Collu et al., 2014). Determining how signaling pathways can influence each other through either antagonism or promotion at different levels is essential to understanding disease progression and basic cellular functions.

The canonical Wnt/Wg signaling pathway revolves around the stabilization and localization of the key transducer, β-catenin [Armadillo (Arm) in Drosophila]. β-catenin is a continuously produced multifunctional protein essential for the maintenance of the adherens junctions as well as for Wnt signaling (Valenta et al., 2012). In the absence of Wnt ligand, cytoplasmic β-catenin is continuously targeted for degradation by a destruction complex consisting of Axin, Adenomatous polyposis coli (APC) and the kinases Casein kinase 1α (CK1α), and Glycogen synthase kinase 3β (GSK3β, also known as Shaggy in Drosophila). CK1α and GSK3β phosphorylate β-catenin, targeting it for ubiquitination and subsequent proteasomal digestion. Upon Wnt/Wg binding to its coreceptors Frizzled (Fz) and LRP/Arrow (Arr), Dishevelled (Dvl/Dsh) is recruited to the receptors and then mediates recruitment of the destruction complex, along with Arm, to Arr, which is localized at the membrane. This results in the disruption of the destruction complex, allowing accumulation of newly synthesized β-catenin, which can then translocate to the nucleus, acting as a co-activator with T-cell factor (TCF) and lymphoid enhancer factor (LEF) to initiate expression of target genes (Daniels and Weis, 2005). Perturbations to many of these components or additional regulatory proteins have been implicated in a multitude of cancers and developmental disorders (Clevers and Nusse, 2012).

One signaling pathway that is also crucial for growth, cell fate and tissue homeostasis, and which has been implicated in many of the same diseases as aberrant Wnt signaling, is the receptor tyrosine kinase (RTK)–Ras–mitogen-activated protein kinase (MAPK) pathway, in particular that using the RTK epidermal growth factor receptor (EGFR) (Guturi et al., 2012; Lee et al., 2010; Paul et al., 2013). Pathway activation occurs upon ligand binding to the receptor, leading to the recruitment of Grb2 and the guanine exchange factor, Son of sevenless (Sos). Sos converts the G-protein Ras85D (hereafter referred to as Ras) to an active state, and thereby activates the MAPK cascade. The MAPK signaling cascade consists of three kinases, Raf (also known as Polehole in Drosophila), a MAPK-extracellular signal-regulated kinase family protein (MEK1/2, also known as MAP2K1/2) [Downstream of Raf1 (Dsor1) in Drosophila], and an extracellular signal-regulated kinase (ERK1/2, also known as MAPK3 and MAPK1) family protein [Rolled (Rl) in Drosophila]. Ras first activates Polehole, which in turn phosphorylates Dsor1 to activate it. Dsor1 can then dually phosphorylate and activate Rl, which in turn can then phosphorylate and modulate a wide range of substrates throughout the cell, including transcription factors to affect target gene expression (Courcelles et al., 2013). Several proteins from the EGFR–Ras–MAPK pathway have been identified to directly interact with Wnt signaling components, in particular ERK family proteins (Červenka et al., 2011; Hoschuetzky et al., 1994; Krejci et al., 2012; Ding et al., 2005).

Previous studies of Wnt and Ras–MAPK have focused on interactions in various disease states and cancers (Zeller et al., 2013). There is a gap in understanding of how these two key signaling pathways interact in a normal context for proper organismal development. In a kinome and phosphatome RNA interference (RNAi) screen in the developing Drosophila larva, multiple components of the EGFR–Ras–MAPK signaling cascade were found to affect Wg signaling (Swarup et al., 2015). Here, we show that Dsor1 plays a crucial conserved function in mammalian cells and Drosophila in promoting Wg signaling. We find this is independent of activation of Rolled, and instead that it rather occurs through a new mechanism that involves recruitment and subsequent inhibition of the destruction complex at the membrane. In addition, we show that activation of Dsor1 by Ras in this context is not through EGFR, but most likely the insulin-like growth factor receptor. Taken together, these studies identify a new conserved interaction between RTK–Ras–MAPK and Wg pathways that is mediated by Dsor1 and MEK. This crosstalk is crucial for full Wnt signal propagation for normal development.

Several kinases and phosphatases from the EGFR–Ras–MAPK pathway have been previously identified in an RNAi screen based on their ability to modulate Wg target gene expression (Swarup et al., 2015). We focused our attention on Dsor1, as it was the most terminal protein in the EGFR signaling cascade identified in the screen. In the developing wing imaginal disc, Wg is secreted from a small stripe of cells marking the dorsoventral boundary, which will eventually form the adult wing margin (Couso et al., 1994). Wg triggers pathway activation and transcriptional response in a concentration-dependent manner in the surrounding cells of the wing pouch (Neumann and Cohen, 1997; Zecca et al., 1996). Sens is only transcriptionally activated in cells flanking the Wg-secreting cells, which have the highest levels of pathway activation (Fig. 1A,A″) (Parker et al., 2002). Dll is expressed in a much broader domain within the wing pouch (Fig. 1B), as it requires lower levels of active Wg signaling to induce its transcription (Zecca et al., 1996). The expression of two independent dsor1-RNAi lines in the posterior domain of the disc using engrailed (en)-Gal4 (marked by GFP; Fig. 1C,C′), resulted in a loss of Sens and strong reduction in Dll transcription (Fig. 1C″,D), suggesting that Dsor1 acts a positive regulator of the Wg pathway.

Wg is a target gene of Notch signaling at the dorsoventral boundary in the wing disc (Klein and Arias, 1998), and crosstalk has been described between Notch and EGFR signaling in Drosophila development (Doroquez and Rebay, 2006). To determine whether loss of Dsor1 was affecting Notch signaling upstream of Wg signaling, a wg transcriptional reporter, wg-lacZ was utilized (Fig. 1E). Expression of dsor1-RNAi in the anterior domain of the wing (using ci-Gal4) had no effect on wg transcription (Fig. 1F). Next we assessed whether Wg processing and stability were being affected. Wg is a lipid-modified glycoprotein, and must undergo post translational modifications before being secreted (Franch-Marro et al., 2008; Tanaka et al., 2002). Analyzing total Wg protein levels is a good indicator of defects in Wg stability or processing. Total Wg protein did not differ in control (Fig. 1G) and dsor1-RNAi-expressing tissue (Fig. 1H), suggesting that Dsor1 does not act to regulate wg transcription or processing in the secreting cells.

To confirm that reductions in Dsor1 did not critically affect cell viability, and were indeed directly downregulating Wg signaling, we utilized the apoptotic marker cleaved caspase-3 (Casp-3). Small patches of cells within the dsor1-RNAi expressing domain did have elevated levels of cleaved Casp-3, indicating that apoptosis was occurring in some areas (Fig. S1A,A′). To rule out the effects of apoptosis on Wg signaling, the baculoviral P35 anti-apoptotic protein was co-expressed with dsor1-RNAi (Fig. S1B,B′) (Hay et al., 1994). Co-expression of p35 and dsor1-RNAi still resulted in loss of Sens (Fig. S1C,C′), suggesting that the Wg pathway effects are not solely due to cell death. To validate that the effects of the RNAi lines were a result of a direct reduction in Dsor1, and not due to off target knockdown, expression of a dsor1 transgene was utilized to rescue the dsor1-RNAi phenotype. Expression of dsor1-RNAi in adult wings resulted in a loss of vein material, indicative of loss of EGFR signaling (Fig. S1F) (Diaz-Benjumea and Hafen, 1994). Expression of dsor1 alone had no effect on the adult wing (Fig. S1E) or Dll expression in the wing imaginal disc (Fig. S1H,I), as Dsor1 is inert unless activated by the rate-limiting kinase Raf (Fujioka et al., 2006). UAS-dsor1 was able to rescue the vein defect and reduction of Dll caused by dsor1-RNAi, confirming the phenotype seen from RNAi was directly a result of a loss of Dsor1 (Fig. S1G,J,K).

To further confirm RNAi results, mitotic clones were generated for a dsor1 loss-of-function allele (LOF). Clones of dsor1^LOF, marked by the presence of GFP (Fig. 1I,I′), also resulted in a loss of Sens (Fig. 1I″, arrow). In addition, the adult wings of flies harboring small clones at the wing margin exhibited notches, a phenotypic hallmark of reduced Wg signaling (Fig. 1J, inset). Taken together, these results suggest that Dsor1 is required for Wg target gene transcription.

To determine at which point Dsor1 acts on the Wg pathway, we first focused on Arm stability as its regulation is key for pathway activation. In the wing imaginal disc, Arm is seen in highest concentrations in two bands flanking the Wg-producing cells at the dorsoventral boundary (Fig. 2A) as a result of Arm stabilization allowing high levels in the cytoplasm and nucleus of these cells (Marygold and Vincent, 2003). The expression of dsor1-RNAi resulted in a reduction of stabilized Arm, but Arm located in the adherens junction appeared unchanged (Fig. 2B,B′). These results were also seen with co-expression of P35 to inhibit any potential cell death (Fig. S1C″). Clones of dsor1^LOF, marked by the presence of GFP (Fig. 2C,C′), resulted in a similar loss of stabilized Arm (Fig. 2C″). These results suggest that, in the absence of Dsor1, Arm is still targeted for proteasomal degradation, even under conditions of active Wg signaling, where the destruction complex would normally be inactivated.

To investigate a link between Dsor1 and the destruction complex, we generated axin^null clones and tested for effects on Arm. Axin is a scaffolding protein which acts as the backbone holding the destruction complex together (Zeng et al., 1997). Clones lacking Axin, marked by GFP (Fig. 2D,D‴), showed massive accumulations of Arm and resulted in ectopic production of Sens within the clone (Fig. 2D′,D″). The introduction of dsor1-RNAi into an axin^null background had no effect on Arm or ectopic Sens (Fig. 2E–E‴). Thus, although under normal conditions reduction of Dsor1 caused loss of Arm, in the absence of the destruction complex Dsor1 had no effect. Taken together, our findings suggest Dsor1 functions upstream or at the level of the destruction complex.

Regulation of the destruction complex can involve its recruitment and inhibition at the cell membrane following receptor–ligand interaction (Tamai et al., 2004). To investigate whether Dsor1 affects the Wg pathway at the membrane, we looked for effects on protein localization following dsor1 knockdown. The wing imaginal disc epithelium is composed of tightly packed columnar cells that make it difficult to identify protein localization beyond apical or basal positioning. The Drosophila salivary gland, with its giant polyploid cells, offers a much better opportunity to study protein localization in vivo, although information about the activity of Wg in this tissue is minimal.

It has been previously shown that Wg is present in the salivary gland where it induces graded ploidy levels (Taniue et al., 2010). We thus set out to characterize Wg pathway activity in this tissue and to use this to study the effects of Dsor1 on Wg signaling. Wg staining revealed that the protein was produced and secreted by cells within the imaginal ring, marked by small nuclei (Fig. S2B,B′), and formed a gradient in the proximal cells (Fig. S2A). This resulted in the highest levels of cytoplasmic and nuclear Arm within the proximal cells, and transcription of Wg target gene fz3 (Fig. 3B,C; Fig. S2C,E,E′). As no suitable anti-Axin antibody was available, we obtained a fly strain that is heterozygous for a mutation in axin, while simultaneously expressing FLAG-tagged Axin under the ubiquitous tubulin promoter (Petersen-Nedry et al., 2008). In this strain, FLAG–Axin was found at high levels at the membrane in areas with active Wg signaling, and was reduced gradually outside of the zone of active Wg signaling (Fig. S2G–G″). Membranous Axin localization can be used as an indicator of destruction complex membrane recruitment and disruption (Tamai et al., 2004). In the proximal Wg-receiving cells, a Dsh–GFP fusion protein revealed that Dsh was mainly cytoplasmic, with puncta scattered throughout the cytoplasm and at the membrane (Fig. 3A,D; Fig. S2I,I′, arrow) (Axelrod, 2001). Puncta formation is due to the Dsh DIX domain, which allows Dsh polymerization and anchoring of Axin to the membrane (Bilic et al., 2007; Metcalfe et al., 2010; Schwarz-Romond et al., 2007). As Wg signaling diminishes in the distal cells, cytoplasmic levels of Dsh–GFP dropped dramatically and the protein became mainly nuclear, with moderate levels remaining at the cell membrane (Fig. S2I″). The model given in Fig. S2K represents expression domains of Wg and several key proteins in the pathway in the salivary gland.

Having better characterized the distribution of Wg pathway components in the salivary gland, we were able to focus on the effect of Dsor1 on their localization and functionality. A strong potential target for Dsor1 to interact with is Dsh, as it must be phosphorylated to be recruited to Fz and to, in turn, recruit the destruction complex to the membrane (Rothbächer et al., 2000; Umbhauer et al., 2000). The use of dpp-GAL4 to express dsor1-RNAi throughout the salivary gland (Fig. S2B″) caused no change in Dsh–GFP puncta and overall localization in proximal cells compared to wild type (Fig. 3D,D′,G,G′). dsor1 knockdown resulted in a complete absence of the Wg target gene fz3 (Fig. 3E–F′, Fig. S2D), coinciding with a notable reduction in nuclear Arm levels (Fig. S2F,F′). From these results, it appears that the role of Dsor1 in the pathway is likely after Dsh phosphorylation and recruitment to Fz, given that Dsh recruitment is normal without Dsor1, but target gene activation is inhibited.

As a reduction in Dsor1 did not perturb Dsh localization, we speculated that Dsor1 interacts with the destruction complex itself. The destruction complex had constitutive activity with the loss of Dsor1, as seen by loss of Arm even in the presence of Wg signaling. The scaffolding protein Axin has diverse roles in the cell in many different contexts and locations (Luo and Lin, 2004). Under active Wg signaling conditions, Axin is found throughout the cell, but with enrichment at the cell membrane (Fig. 3H–I′), as a result of destruction complex recruitment (Tamai et al., 2004). This phenotype was also seen upon Dsor1 overexpression (Fig. S2L-L‴). Reduction of Dsor1 by RNAi resulted in a striking loss of membranous Axin and a build up of nuclear levels (Fig. 3J–K′; Fig. S2M–M‴). This affect could be rescued by the overexpression of the dsor1 transgene (Fig. S2N–N‴). We propose that the loss of membranous Axin and nuclear accumulation are two independent events. First, the reduction of membranous Axin within the active Wg-receiving cells suggests that Dsor1 is required to recruit the destruction complex to the membrane. Without Dsor1, Axin and the destruction complex remain active in the cytoplasm where they mediate the degradation of Arm and thus prevent transcription of target genes. Second, we have shown that dsor1-RNAi can critically impair cell viability and induce patches of apoptotic cells (Fig. S1A). Axin has been shown to translocate to the nucleus under conditions of cell stress, where it plays a crucial role with p53 in cell viability to initiate cell cycle arrest or apoptosis (Li et al., 2009). In addition, Lui et al. (2011) have identified a nuclear role for Axin in promoting Wnt signaling post β-catenin stabilization. Given that we observed reduced Wg signaling due to loss of dsor1, our findings suggest that the nuclear Axin associated with Dsor1 knockdown is independent of Wg signaling. Thus, we propose that loss of Dsor1 causes both the Wnt-specific loss of Axin membrane recruitment and the unrelated nuclear localization linked to cell stress. This interpretation is bolstered by the appearance of nuclear Axin in every cell in the salivary gland, including the proximal cells distant from active Wg signaling (Fig. S2H). Taken together, these results show that the role of Dsor1 in the Wg pathway occurs after Dsh recruitment, but appears to be critical for the subsequent recruitment of Axin and the destruction complex.

Having determined that Dsor1 interacted with the Wg pathway at the level of recruitment of the destruction complex, it became imperative to confirm that loss of Wg signaling was directly due to loss of Dsor1, and not simply the failure to activate the downstream MAPK Rolled (Rl). An interaction between Wnt components and the Rl ortholog ERK has been identified in several cases (Červenka et al., 2011; Ding et al., 2005; Krejci et al., 2012), including in Drosophila (Freeman and Bienz, 2001). Knockdown of rl by RNAi showed no effect on Dll expression and did not induce cell death in the wing disc (Fig. S3A–A″,B,B″), but was able to disrupt patterning and expression of the EGFR target gene argos (Fig. S3B,B′,G). In addition rl-RNAi expressed along the anterior–posterior section of the wing resulted in a loss of vein tissue (Fig. S3D, arrowhead), a hallmark of reduced Rl activity (Brunner et al., 1994), demonstrating the effectiveness of rl-RNAi.

To address the possibility that rl-RNAi was unable to reduce Rl levels enough to directly affect Wg targets, we used mutant alleles of rl, namely the rl hypomorphic allele rl¹ and the strong hypermorphic allele rl^{sevenmaker (sem)} (Brunner et al., 1994). Homozygous rl¹ adult wings were smaller, creased and droopy (Fig. 4D), compared to wild type (Fig. 4A,B) indicative of reduced rl activity (Brunner et al., 1994), yet discs displayed normal Sens and stabilized Arm expression (Fig. 4C–C″). Heterozygous rl^sem adult wings contained excess vein material indicative of hyperactivated EGFR and Rl activity (Fig. 4F) (Brunner et al., 1994). Analysis of rl^sem/+ wing discs revealed undisrupted Sens and Arm expression compared to wild type (Fig. 4A–A″,E–E″).

To further confirm a Dsor1-specific role in the Wg pathway and rule out the predominant target of Dsor1 (i.e. the MAPK Rl) as an effector in this role, we expressed dsor1-RNAi in the posterior domain of the wing disc in an activated Rl genetic background (rl^sem/+). Adult wings exhibited a mix of ectopic and lost vein tissue in the posterior domain (Fig. 4H, asterisk), indicating that hyperactive Rl was still partially functional in the tissue even with reduced Dsor1. Additionally wings did not show creases, or the ‘droopy’ phenotype, suggesting that Rl activity was not reduced as a result of Dsor1 reduction. The wings also exhibited a notched wing phenotype, a hallmark of reduced Wg signaling (Fig. 4H). Staining of the imaginal disc revealed a complete loss of Sens and reduced Arm levels within the posterior domain, indicating that this is a Dsor1-specific effect (Fig. 4G–G″). We also examined interactions between UAS-Rl^sem and Dsor1. The co-expression of Rl^sem and dsor1-RNAi still caused reduced Sens and a loss of stabilized Arm in the wing disc (Fig. S3E–E″). Taken together, these results support the model that Dsor1 is acting in a Rl-independent manner to modulate Wg signaling.

As part of the MAPK signaling cascade, Dsor1 activation requires phosphorylation by upstream components (Gardner et al., 1994). It is essentially inert without upstream activation, as seen by the observation that overexpression of Dsor1 alone results in a completely wild-type phenotype (Figs S1E,I and S2L–L‴). This is supported by Fujioka et al. (2006), who found that Raf is the rate-limiting protein in the MAPK activation cascade, and MEK family proteins (the Dsor1 equivalent) are of highest concentration, with only a subset of them activated at any time.

In the developing larva, the most prominent activator of the MAPK cascade in the imaginal discs is EGFR (Shilo, 2003). EGFR has been shown to interact with Engrailed in early wing disc development to restrict wg expression to the wing pouch (Baonza et al., 2000) before its final well-known role in patterning the wing veins in late third-instar larvae (Sturtevant et al., 1993). To avoid complications from possibly disrupting early patterning in the wing disc, we focused on the leg imaginal disc, which has simpler gene expression patterns. In the developing leg, wg is expressed in the ventral domain (Fig. 5A″) and Decapentaplegic (Dpp) signaling in the dorsal, which together initiate patterning of the leg through distinct wedge domains. Wg and Dpp activate dachshund (dac) in the medial cells, followed by EGFR signaling in the distal cells in the early third instar, through inducing expression of the EGFR ligand vein and the protease rhomboid (Campbell, 2002; Galindo et al., 2002). The EGFR target gene Bar, in turn, represses dac expression in the distal cells (Giorgianni and Mann, 2011), resulting in refinement of Dac to a distinct domain of medial cells and its absence in the active distal EGFR domain (Fig. 5A‴, arrow). We utilized the Wg-specific reporter, Fz3–dsRed (which is expressed in a wider wedge domain than Wg in the dorsal domain; Fig. 5A′), to monitor the effect of Dsor1 on Wg signaling. Using wg-Gal4 to express dsor1-RNAi (marked by GFP; Fig. 5B″), we observed a strong reduction in Fz3–dsRed within the wg expression domain, yet surrounding Wg-receiving cells still expressed Fz3–dsRed (Fig. 5B,B′). This cell autonomous effect bolstered previous results showing that Dsor1 acts in the Wg-receiving cells and does not affect its secretion or processing. In addition, this loss of Fz3–dsRed was found in the Dac domain, which is outside of the zone of EGFR activity (Fig. 5B‴). This demonstrated that the effect of Dsor1 on Wg signaling is independent of endogenous EGFR signaling.

To extend these results we further examined the interaction between Wg and EGFR in the wing disc. In the late third-instar stage, active EGFR signaling is refined to the primordial wing veins (Martín-Blanco et al., 1999), as seen by transcription of the target gene and pathway inhibitor argos (Fig. S3G). Blistered (bs) expression is initiated in the inter-vein cells of the wing disc in third-instar larvae, in a pattern complementary to the domain of active EGFR signaling (Fig. S3G,H, arrows) (Fristrom et al., 1994; Grenier and Carroll, 2000). The use of bs-GAL4 to drive dsor1-RNAi resulted in a notable reduction in Dll transcription in the inter-vein domain (marked by arrowheads) compared to wild type (Fig. S3I,J). Furthermore a role for Dsor1 that is independent of EGFR was seen in the salivary glands. Dsor1 knockdown in the polyploid cells reduced Wg signaling, and EGFR signaling is not active in that cell type (Kuo et al., 1996). Taken together, these results show that in multiple contexts Dsor1 can promote Wg signaling independent of upstream EGFR activation.

Having ruled out the EGFR as the mechanism for Dsor1 activation, we wanted to determine whether traditional components of the MAPK cascade downstream of the receptor, like Ras, were still required for Dsor1 activation and affected the Wg pathway. Using a dominant-negative variant of Ras, ras^N17 (Feig and Cooper, 1988), we generated GFP-positive mitotic flip-out clones expressing ras^N17 (Fig. 5C″). Cells expressing ras^N17 displayed a loss of stabilized Arm and a strong reduction of Sens, phenocopying Dsor1 loss (Fig. 5C′,C‴). These experiments show that Dsor1 activity in Wg signaling is independent of EGFR, but appears to still require Ras-mediated activation of the MAPK cascade.

Given that Dsor1 activity was independent of EGFR activation, but still appeared to require Ras, we wanted to identify the Ras-activating source. Ras-MAPK signaling can be initiated by a myriad of sources in development, including other RTKs, G-protein-coupled receptors (GPCRs) and integrins (Chen et al., 2011; Crampton et al., 2009; Sopko and Perrimon, 2013). To refine our search, we focused on proteins that were expressed ubiquitously during development and could therefore serve as activators of Dsor1 in multiple contexts. We screened for potential activators of the Ras–Dsor1–Wg pathway using transgenic RNAi and dominant-negative variants to knockdown or inhibit different receptors, while monitoring Dll expression in the wing disc. RNAi and dominant-negative expression of the Drosophila insulin-like growth factor receptor (InR^DN) resulted in a strong reduction in Dll expression (Fig. 6A). InR has been identified to activate Ras–MAPK signaling to induce proliferation during Drosophila development, making it a strong candidate for the signal source (Oldham et al., 2002; Yenush and White, 1997). InR^DN did not affect total Wg levels, but did cause a reduction of stabilized Arm within the disc (Fig. 6B–C′). In addition, the reduction in active Wg signaling was not accompanied by any elevated levels of apoptosis, as seen with cleaved Casp-3 expression (Fig. 6C″). These results phenocopy what is seen in tissue with a reduction of active Dsor1.

To understand how Dsor1 might be required for the recruitment of Axin to the cell membrane, we focused on Dsor1 interactions with destruction complex components and the Wg receptors, Fz and Arrow. We utilized proximity ligation assays (PLA) to determine whether Dsor1 interacts with Wg components in vivo. In the PLA assay, probe fluorescence will only occur if two proteins of interest are <40 nm apart, only a slightly larger distance than with FRET (<10 nm), inferring strong evidence of protein interaction within cells. As there is currently no anti-Dsor1 antibody available for immunofluorescence, en-Gal4 was utilized to express a HA-tagged Dsor1 protein. To validate that ectopically expressed Dsor1 did not interact with non-specific proteins due to cellular saturation, GFP was co-expressed and probed for interaction. PLA against HA-tagged Dsor1 and GFP failed to produce any signal over background control tissue (Fig. S4A,A′). We next confirmed that Dsor1–HA could interact with a known target protein. PLA against Dsor1–HA and endogenous Rl (using an anti-ERK1/2 antibody) produced a strong signal exclusively in cells expressing Dsor1–HA, confirming that Dsor1–HA was capable of binding with endogenous proteins (Fig. S4B,B′).

With validation of our assay conditions, we next tested Dsor1–HA for interactions with Wg pathway components. Dsor1–HA did not produce a significant signal with Axin, Arrow or Dsh (Fig. S4C–E). Surprisingly, Dsor1–HA and endogenous Arm gave a strong signal, indicating an interaction (Fig. 7A). Lateral view analysis revealed that the Dsor1–Arm interaction occurred solely at the apical and basal surfaces of the wing disc (Fig. 7C′). The wing disc consists of two layers of cells. DAPI-stained nuclei can be used to mark the tightly packed columnar cells, where Wg signaling occurs, and the thin layer of squamous peripodial cells above the apical surface (Fig. 7B,B′). The majority of Arm found within the columnar cells of the wing disc is located near the apical surface at the adherens junctions, and is not involved in Wg signaling (reviewed in Valenta et al., 2012). It is difficult to distinguish whether the Dsor1–Arm PLA signal at the apical surface is occurring in the peripodial cells, at the adherens junction or near the Wg receptors on the apical surface. However, signal at the basal surface, which lacks adherens junctions, gives strong support that Dsor1 and Arm are in close contact with one another due to membrane recruitment of the destruction complex mediated by Wg signaling.

To determine whether the promotion of Wg signaling by Dsor1 is conserved, we examined the effects of MAPK components on Wnt signaling in HEK-293 cells. Wnt pathway activity was stimulated by Wnt3a and the cellular response was measured by using a TCF-responsive TOPFLASH transcriptional reporter (Korinek et al., 1997) (Fig. 7D, columns 1, 2). Transfection of validated catalytically inactive dominant-negative MEK1 (Mansour et al., 1994) (a Dsor1 homolog) resulted in a greater than threefold reduction in transcriptional activity (Fig. 7D, column 3). Although our studies thus far have indicated that the effect of Dsor1 in flies was not transduced through the MAPK Rl, we tested whether a validated constitutively active ERK2 (Emrick et al., 2001) (a Rl homolog) could rescue the transcriptional repression by dominant-negative MEK1, given that vertebrate ERK proteins have been previously identified as having a role in promoting Wnt signaling (Červenka et al., 2011; Ding et al., 2005; Krejci et al., 2012). Co-transfection of dominant-negative MEK1 and constitutively active ERK2 provided only a partial rescue (Fig. 7D, columns 2, 3 and 4), suggesting that MEK proteins provide a direct role for Wnt signaling independently of activation of ERK. This was further supported by expression of dominant-negative ERK2 (Emrick et al., 2006), as it caused a reduced TCF-mediated response, but only roughly half that seen with dominant-negative MEK1 (Fig. 7D, columns 2, 3, 5). Taken together, these results demonstrate the conservation of a MEK-specific role for Wnt target gene activation.

Next, epistasis experiments were performed to determine whether MEK functions similarly in Wnt signaling in mammalian cells to how Dsor1 does in Drosophila. Transfection of a stabilized β-catenin, with its GSK3β and CK1α phosphorylation sites mutated to alanine (S33A, S37A, T41A and S45A) (β-catenin^AAAA), and which is therefore resistant to proteasomal degradation, induced much stronger TCF-transcriptional activity than Wnt3a alone (Fig. 7D, columns 2, 6). Upon co-transfection of dominant-negative MEK1 with β-catenin^AAAA, there was no significant change in reporter activity (Fig. 7D, column 7). This indicates that, like Dsor1, the role of MEK within the Wnt pathway does not occur after β-catenin stabilization.

Because the promotion of the Wnt pathway mediated by MEK is conserved in HEK-293 cells, we further examined whether MEK could also regulate β-catenin. HEK-293 cells were transfected with dominant-negative MEK1, and the levels of β-catenin were assessed. First, using an antibody that recognizes active non-phosphorylated β-catenin (i.e. no phosphorylation on Ser33, Ser37 or Thr41) (active β-catenin), we observed dramatically reduced protein levels following expression of dominant-negative MEK1 (Fig. 7E). We next blocked proteasomal degradation using MG132 and examined active β-catenin, GSK3-phosphorylated β-catenin and total β-catenin. Western blotting of lysates showed no change in the levels of GSK3-mediated phospho-β-catenin (Ser33, Ser37 and Thr41). These phospho-sites serve to target β-catenin for proteasomal digestion, which suggests that the destruction complex, including GSK3, is functional in the absence of MEK1 (Fig. 7E). In addition, total β-catenin levels were not significantly disrupted (Fig. 7E). Protein levels of active β-catenin were dramatically reduced in the presence of dominant-negative MEK1, further suggesting a role for Wnt propagation by promoting β-catenin stability (Fig. 7E).

In this study, we have demonstrated that Dsor1 and MEK are new regulators of Wg/Wnt signaling. We have shown that Dsor1 and the mammalian ortholog MEK are required for full activation of Wg/Wnt signal transduction. We have also demonstrated that their role in Wg/Wnt signaling is likely to be independent of MAPK phosphorylation, challenging the current dogma of the MAPK signaling cascade. Our study has also identified that the currently known activator of the Ras–Dsor1 pathway in imaginal discs, EGFR, is non-essential for a role in Wg signal propagation. Activation of this new crosstalk mechanism might be initiated through insulin-like growth factors (IGFs) or Drosophila insulin-like peptides (DILPs) and the InR (Fig. 8).

Our findings have highlighted a new function for components of the MAPK signaling cascade in Wnt activation. These two crucial signaling pathways have been found to influence each other in many carcinomas (Zeller et al., 2013), yet their interaction has not been well established in normal developmental contexts. Our results uncover a role of Dsor1 in Wg signaling in the developing larval imaginal discs and salivary glands. Epistasis experiments have identified that Dsor1 acts upon the Wg pathway in the receiving cells after Dsh recruitment to the co-receptors Fz and Arrow, and its presence is needed for the recruitment of the destruction complex components to the membrane surface. As Dsor1 showed a close physical interaction with Arm itself at both the apical and basal cell surface, it suggests that Dsor1 acts to promote and retain the destruction complex near the receptor complex. Future experiments will need to clarify whether Dsor1 is utilized as a linker to promote destruction complex retention, or if its ability to phosphorylate specific target proteins within the complex is key. However, the utilization of catalytically inactive MEK1 resulted in reduced TCF reporter activity and significantly reduced active β-catenin, suggesting the phosphorylation activity of MEK is crucial for Wnt signaling in mammalian cells.

Quite surprisingly our results identified that Dsor1 and MEK activity was independent of its well-known function in Rl and ERK activation. The MAPK signaling cascade is one of the best characterized and understood phosphorylation pathways to date. Its central dogma has revolved around its simple and exclusively linear signaling series for the activation of Rl/ERK. A small number of previous studies have questioned this model, demonstrating that MEK is capable of phosphorylating other targets (Jo et al., 2011; Tang et al., 2015), even GSK3 (Takahashi-Yanaga et al., 2004). Our results are consistent with this alternative ‘MEK multiple substrate model’. We demonstrated, through RNAi knockdown, ectopic expression and genetic interaction studies, that Rl does not influence Wg activity, and that the effect we observe is specific to Dsor1. Our mammalian cell culture experiment does support the previous finding that vertebrate ERK can promote Wnt activity (Červenka et al., 2011; Ding et al., 2005; Hoschuetzky et al., 1994; Krejci et al., 2012), suggesting the mechanism might diverge slightly between flies and vertebrates. Moreover, we demonstrate a new MEK function for direct promotion of Wnt signaling.

Our findings reveal that the predominant larval MAPK signaling cascade initiated by EGF is not required for the activation of this new Ras–Dsor1–Wg interaction. Endogenous EGFR signaling is the only identified activator of di-phospho-Rolled in the developing wing (Martín-Blanco et al., 1999). Our results do not contradict the current understanding of MAPK activity in the developing wing, but reveal a new pathway using a subset of MAPK cascade components. It was surprising to identify that inhibition of the InR resulted in a striking phenocopy of Dsor1 disruption, suggesting that InR might be the upstream activator of the Ras–Dsor1 pathway. InR activation of Ras–MAPK signaling for a proliferation response has been previously identified (Oldham et al., 2002). Our results suggest that DILPs might also play a role in patterning through Wg signaling. It has been identified that insulin and IGF1 can promote Wnt activity by increasing β-catenin stability and nuclear accumulation, as well as by upregulating other pathway components through multiple distinct mechanisms (Sun and Jin, 2008; Sun et al., 2009,, 2010). In future studies it will be interesting to identify if InR–Ras–MEK–Wnt crosstalk can also be elucidated in mammalian cells, as well as trying to distinguish whether IGF or DILPs promote Dsor1 activity for Wg signaling.

Flies and crosses were raised on standard medium at 25°C unless stated otherwise. w¹¹¹⁸ was used as wild type. The following fly strains were used in this study: UAS-GFP, UAS-lacZ/TM6B,UAS-rl-RNAi (BL34855), rl¹(BL386), rl^sem/CyO (BL108365), UAS-p35 (BL5072, 5073), dpp-Gal4 (BL1553), ptc-Gal4 (BL56807), bs¹³⁴⁸-Gal4/CyO (also referred to as 1348-gal; BL25753), UAS-ras^N17(BL4845), UAS-inr^K1409A (InR^DN, BL8252/8253), Dll-lacZ (BL10981) (obtained from the Bloomington Drosophila Stock Center), dpp-Gal4,UAS-gfp,dpp-Gal4,UAS-dsor1-HA, UAS-dsor1-RNAi (VDRC 107276, 40026) [obtained from the Vienna Drosophila Resource Center (Dietzl et al., 2007)], yw,dsor1^K8C13,FRT19A/FM7i (a gift from Jessica Treisman, The Helen L. and Martin S. Kimmel Center for Biology and Medicine at the Skirball Institute of Biomolecular Medicine, NY, NY, USA, utilized as the dsor1^LOF allele), UAS-dsor1/TM3,Sb, UAS-dsor1-HA, hh-Gal4/TM6B (Port et al., 2011),wg-Gal4(ND382) and en-Gal4,UAS-gfp (gifts from Konrad Basler, Institute of Molecular Life Sciences University of Zurich, Switzerland), axin^S044230,FRT82/TM3, Sb (a gift from Yashi Ahmed, Department of Genetics, Geisel School of Medicine at Dartmouth, USA), MARCM82B and hs-flp; Act>CD2>UAS-Gal4, UAS-gfp/SM6∼TM6 (gifts from Bruce Edgar, Zentrum für Molekulare Biologie der Universität Heidelberg, Germany), yw;tub>FLAG-axin/CyO;axin^S044230,FRT82 (a gift from Marcel Wehrli, Oregon Health & Science University, USA), MARCM19A (a gift from Ronwen Xi, National Institute of Biological Sciences, Beijing, China), ci-GAL4,UAS-Dcr2 (a gift from Ken Irvine, Dept. of Molecular Biology and Biochemistry, Waksman Institute, Rutgers University, USA), dsh-gfp/CyO (a gift from Jeffrey Axelrod, Dept. of Pathology, Stanford University School of Medicine, USA), wg-lacZ/CyO (Kassis et al., 1992), fz3-lacZ/FM7a, fz3-dsRed/TM6B (a gift from Ramanuj Dasgupta, NYU Cancer Institute and Department of Biochemistry and Molecular Pharmacology, USA), argos-lacZ/TM6B.

UAS-dsor1-RNAi (VDRC 107276, 40026) strains target regions of dsor1s second exon. Specific targeting sequences for these two lines and additional information can be found on the Vienna Drosophila Resource Center website at: http://stockcenter.vdrc.at/control/product/~VIEW_INDEX=0/~VIEW_SIZE=100/~product_id=107276, and http://stockcenter.vdrc.at/control/product/~VIEW_INDEX=0/~VIEW_SIZE=100/~product_id=40026, respectively.

In assays examining the interaction between two UAS transgenes, control crosses were performed with UAS-GFP or UAS-lacZ, to rule out suppressive effects due to titration of Gal4. Mosaic analysis with a repressible cell marker (MARCM) clones were generated by crossing yw,hs-flp, tub-Gal4,UAS-GFP;; tub-Gal80, FRT82B (MARCM82B) and hs-flp¹²², tub-Gal80, FRT19A; act-Gal4,UAS-GFP (MARCM19A) to corresponding lines and heat-shocking first-instar larvae at 37°C for 2 h and incubating them at 29°C until dissection. Heat-shock inducible flip-out clones were generated by crossing hs-flp;;Act>CD2>UAS-GAL4,UAS-GFP/SM6∼TM6 to corresponding lines and heat-shocking first-instar larvae at 37°C for 15 min and incubating them at 29°C until dissection.

The following plasmids were used in this study: pMCL-HA-MAPKK1-8E (K97M) (dominant-negative MEK1; Addgene plasmid #40811) (Mansour et al., 1994), pCMV5-rat ERK2-L73P/S151D (constitutively active ERK2; Addgene plasmid #40819) (Emrick et al., 2001), pCMV5-rat ERK2-K52R (dominant-negative ERK2; Addgene plasmid #40813) (Emrick et al., 2006), all a gift from Natalie Ahn (Dept. of Chemistry and Biochemistry, Univ. of Colorado, Boulder, USA), pCMV-Myc (empty vector; Clontech), TOPFLASH (Korinek et al., 1997), FOPFLASH (Korinek et al., 1997), pRL-CMV (Renilla luciferase; Promega), pcDNA-Wnt3A and pcDNA-β-catenin^AAAA a gift from Cara Gottardi (Department of Pulmonary and Critical Care Medicine Northwestern University Feinberg School of Medicine, USA).

Wandering third-instar larvae were dissected in phosphate-buffered saline (PBS). Imaginal discs and salivary glands were fixed in 4% paraformaldehyde at room temperature for 20 min followed by three washes in PBS. Tissue was blocked [2% BSA diluted in PBS 0.1% Triton X-100 (PBT)] for 45 min at room temperature, followed by incubation of primary antibodies overnight at 4°C. Tissue was then washed three times with PBT and incubated with secondary antibodies at room temperature for 1.5 h. A final series of three PBT washes were performed to the tissue followed by mounting in 70% glycerol solution, or VECTASHIELD Mounting Medium with DAPI (Vector Labs). The following primary antibodies were used in this study: mouse anti-Wg 4D4 (1:100, 1:500 salivary glands, DSHB), mouse anti-Arm (1:50, 1:200 salivary glands, DSHB), mouse anti-β-galactosidase (1:2000, Promega), rabbit anti-cleaved Casp3 (1:100, Cell Signaling), guinea pig anti-Sens (1:1000, a gift from Hugo Bellen, Dept. of Molecular and Human Genetics, Baylor College of Medicine, USA), mouse anti-HA (1:500, Sigma), rabbit anti-HA (1:1600, Cell Signaling), rabbit anti-ERK1/2 (1:200, Cell Signaling), rabbit anti-FLAG (1:200, Sigma), mouse anti-GFP (1:500, Cell Signaling), mouse anti-Dac2-3 (1:75, DSHB), rabbit anti-Arrow (1:15,000, a gift from Stephen DiNardo, Dept. of Cell and Developmental Biology, Perelman School of Medicine, University of Pennsylvania, USA) and mouse anti-Fz2 (1:50, DSHB) antibodies. All secondary antibodies (Jackson ImmunoResearch) were used at a 1:200 dilution. Proximity ligation assays were performed using Duolink In Situ Red Starter Kit Mouse/Rabbit, following the manufacturer's protocol. Adult wings were dissected in 95% ethanol and mounted in Aquatex (EMD Chemicals Inc.).

Microscopy images were taken with a Nikon A1R laser scanning confocal microscope and processed using Adobe Photoshop CS6. Adult wings were imaged with an Axioplan 2 microscope.

HEK-293 cells were cultured in six-well plates at 37°C in 5% CO₂ with Dulbecco's modified Eagle's medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (FBS; Invitrogen). Transient transfection was performed with 2 µg total DNA with Polyfect transfection reagent (Qiagen) according to the manufacturer's instructions. When required, the final amount of DNA used for transfection was kept constant by the addition of empty vector DNA. All cells were harvested and lysed at 36 h post transfection with lysis buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM Na₂-EDTA, 1 mM EGTA, 1% Triton, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM Na₃VO₄, 1 µg/ml leupeptin; Cell Signaling Technology), supplemented with protease inhibitors (Roche). For analysis of degradation-bound β-catenin (phospho-β-catenin), MG132 (Calbiochem) was used to treat cells at a concentration of 25 μM for 6 h, prior to harvesting to prevent proteasomal digestion.

HEK-293 cells transfected with Wnt3A and empty vector, or Wnt3a and dominant-negative MEK1 were treated with lysis buffer. Lysates were then sonicated for several seconds on ice, followed by a 16,300 g centrifugation for 10 min at 4°C. The supernatant was removed and boiled for 10 min with Laemmli buffer, then separated on 10% SDS-PAGE gels. Proteins were transferred onto nitrocellulose membranes, and probed with the following primary antibodies: anti-phospho-β-catenin (Ser33, Ser37 and Thr41) (1:1000, Cell Signaling), anti-β-catenin (1:1000, Cell Signaling), anti-non-phospho (Active) β-catenin (Ser33, Ser37 and Thr41) (1:1000, Cell Signaling) and anti-β-actin (1:1000, ABM) antibodies. Membranes were visualized using the enhanced chemiluminescence (ECL) western blotting substrate (Pierce) with a LAS4000 luminescent imager (Fujifilm). The protein levels were determined using ImageJ software to perform densitometry. Each sample was normalized to β-actin levels. Transfections and western blotting was performed in triplicate.

Luciferase assays were performed in HEK-293 cells with the dual luciferase reporter assay system (Promega) according to manufacturer's instructions. TOPFLASH or FOPFLASH reporter gene plasmids with control reporter plasmid encoding Renilla luciferase (to normalize transfection efficiencies and for monitoring cell viability) were transfected with each expression vector as indicated, to determine overall Wnt pathway activity through TCF/LEF reporter activity. The values shown represent the average and standard deviation of three biological replicate transfections, performed in triplicate. Values were normalized to FOPFLASH reporter activity equal to 1. Significance between groups was assessed by one-way analysis of variance (ANOVA), and P<0.01 was considered significant.

We thank the following individuals and stock centers for reagents and fly strains: Konrad Basler, Jessica Treisman, Marcel Wehrli, Bruce Edgar, Ramanuj Dasgupta, Yashi Ahmed, Steve DiNardo, Meghan Maher, Cara Gottardi, Natalie Ahn, Hugo Bellen, Jeffrey Axelrod, Ronwen Xi, Ken Irvine, the Bloomington Drosophila Stock Center, The Vienna Drosophila RNAi Center and the Developmental Studies Hybridoma Bank. Thanks to Sharan Swarup and Dipa Pradhan-Sundd for initial guidance on the project, as well as past and present members of the Verheyen lab for discussions, and Nick Harden for comments on the study.