Wingless/Wnt signaling in Drosophila: the pattern and the pathway

Summary

Wnt signaling generates pattern in all embryos, from flies and worms to humans, and promotes the undifferentiated, proliferative state critical for stem cells in adult tissues. Inappropriate Wnt pathway activation is the major cause of colorectal cancers, a leading cause of cancer death. Although this pathway has been studied extensively for years, large gaps remain in our understanding of how it switches on and off, and how its activation changes cellular behaviors. Much of what is known about the pathway comes from genetic studies in Drosophila, where a single Wnt molecule, encoded by wingless (wg), directs an array of cell fate decisions similar to those made by the combined activities of all 19 Wnt family members in vertebrates. Although Wg specifies fate in many tissues, including the brain, limbs and major organs, the fly embryonic epidermis has proven to be a very powerful system for dissecting pathway activity. It is a simple, accessible tissue, with a pattern that is highly sensitive to small changes in Wg pathway activity. This review discusses what we have learned about Wnt signaling from studying mutations that disrupt epidermal pattern in the fly embryo, highlights recent advances and controversies in the field, and sets these issues in the context of questions that remain about how this essential signaling pathway functions.

Introduction

Thirty-six years have elapsed since the Drosophila wingless (wg) gene was cloned (Baker, 1987) and found to match the mouse oncogene, int-1 (Rijsewijk et al., 1987). These homologous molecules became the founding members of the Wnt class of growth factors (Nusse et al., 1991). Over the decades, we have learned many things about how the pathway functions and how remarkably different it is from other growth factor signaling pathways. Unlike most secreted peptide growth factors, Wnts are not freely soluble. They are membrane-associated because of a lipid attachment that is essential for their function (Nusse, 2003; Willert et al., 2003). Unlike many signal transduction pathways which involve an intracellular cascade of kinases, the Wnt pathway depends critically on alterations in protein stability, primarily of the effector molecule, beta-catenin. Beta-catenin is continually synthesized in epithelial tissues, where it associates with adherens junctions and mediates cell-cell adhesion. Cytosolic levels of beta-catenin are kept low by a complex of proteins, called the “destruction complex”, which targets beta-catenin for degradation. Wg/Wnt signaling somehow blocks this process, allowing beta-catenin to accumulate and translocate to the nucleus. There, it binds the transcription factor Tcf and acts as a transcriptional co-activator in a role completely independent of adhesion. The details of this mechanism came to light through analysis of mutations in armadillo (arm), the fly beta-catenin. These were among the mutations identified in the famous Heidelberg screen for embryonic pattern defects, which led to the 1995 Nobel Prize in Physiology or Medicine for Wieschaus and Nüsslein-Volhard (Jürgens et al., 1984; Nüsslein-Volhard et al., 1984; Wieschaus et al., 1984). This screen, and others that focussed on maternally deposited gene products (Perrimon et al., 1989; Schupbach and Wieschaus, 1989), identified many components of the Wg/Wnt pathway through their mutant effects on epidermal pattern in the fly embryo. Such work continues today, with genetic screens for pattern defects yielding new mutations that affect Wg/Wnt activity, which are then positioned in the pathway through genetic epistasis experiments with known components.

Epidermal Pattern in the Embryo

The cuticle pattern secreted by embryonic epidermal cells is exquisitely sensitive to Wg signaling levels. The wg gene is expressed in just a single row of epidermal cells in each segment (Fig. 1A, Fig. 2B) (Baker, 1988), but Wg activity influences cell fates across the segment. The segmental expression is turned on by the cascade of transcription factors that generates anterior-posterior positional information in the early embryo (reviewed in Akam, 1987). Wg protein forms a punctate, vesicular distribution over many cell diameters away from the wg-expressing row (van den Heuvel et al., 1989). Wg movement to neighboring rows of cells generates two very different aspects of the segmental pattern of the fly epidermis (reviewed in Peifer and Bejsovec, 1992). On the ventral surface, 6 rows of epidermal cells produce one or more hook-shaped structures called denticles, forming a trapezoidal belt of denticles in each abdominal segment (Fig. 1A). Each row of denticles within a belt has a characteristic size, shape and polarity. These actin-based structures become part of the tough exoskeleton when the epidermal cells secrete cuticle at the end of embryogenesis. Denticle belts are separated by a second aspect of the pattern: an expanse of smooth, or naked, cuticle secreted by the more posterior cells in each segment.

Figure 1.

Cuticle pattern of wg loss and gain of function mutants compared with wild-type. (A) Wild-type embryos secrete a repeating pattern of ventral denticle belts in the abdominal segments. Position of wg expression within one segment shown as gray line, schematic diagram of a segment below. Wg (pebbled gray) is secreted and moves away from wg-expressing cells, correlating with high accumulation of Arm (black/gray nuclei). Adapted from Jones and Bejsovec, 2003. (B) wg loss of function results in loss of naked cuticle. (C) Overexpression of a wild-type wg transgene, using the uniform embryonic E22c-Gal4 driver, results in excess naked cuticle. (D) Embryos mutant for Tcf, also known as pangolin, show a weak wg-like pattern, lacking the naked cuticle that should separate denticle belts. (E) Maternal/zygotic loss of Apc2 results in uniform naked cuticle. Adapted from Bejsovec, 2006. Anterior is to the left.

Figure 2.

svb expression defines the domain of denticles and Crinkled (Ck) is required for their proper morphology. (A) Schematic diagram of epidermal cells in embryonic segment showing expression of wg, svb, engrailed (en), hedgehog (hh), rhomboid, and Serrate with respect to the denticle pattern. (B) wg-Gal4 driving UAS-GFP in live embryos shows that the wg-expressing rows underlie the naked cuticle portion of the pattern in each abdominal segment. (C) Higher magnification of live wg>GFP embryo in profile. (D) svb7.3-Gal4 (McGregor et al., 2007) drives UAS-GFP in denticle-producing cells in each segment. (E) Higher magnification shows that each denticle-producing cell in the svb>GFP embryo expresses GFP. (F) Driving ectopic svb in wg-expressing cells leads to ectopic denticle formation (after Payre et al., 1999). Environmental scanning electron microscopy shows that (G) wild-type and (H) ck mutant embryos produce a robust, diverse array of denticles, whereas wg mutants (I) show reduced denticle diversity with robust denticle size and pointed shape. (J) Denticles are reduced and rounded in wg ck double mutants (from Bejsovec and Chao, 2012).

After hatching from the egg, larvae use their segmentally-repeating denticle belts for traction while crawling. Fly geneticists use them as markers for cell fates: denticle type versus naked cuticle indicate distinct positional values within the field of cells, which then must be interpreted by each cell to produce a specific structural outcome. Loss of function mutations in wg, or in any gene required for Wg pathway activation, result in loss of the naked cuticle expanse as well as loss of diversity among denticle morphologies within the belt (Fig. 1B, D) (reviewed in Bejsovec, 2006). Conversely, uniform Wg signaling – either through overexpression of a wg transgene (Fig. 1C) or by mutating a negative regulator of the pathway (Fig. 1E) – converts all of the ventral epidermis to naked cell fate. Thus Wg signaling activity is required for specifying naked cuticle cell fate and also for generating the diverse array of cell fates that produce different types of denticle within each belt.

Specification of naked cuticle depends on Wg-mediated regulation of ovo/shaven-baby (svb) gene expression: svb is expressed in epidermal cells furthest from the source of Wg in each segment (Fig. 2). As the name suggests, loss of function svb mutations result in absence of denticles on the ventral surface of mutant larvae (Wieschaus et al., 1984). The svb gene encodes a transcription factor required for turning on genes that produce structural elements, such as actin-bundling proteins, which construct the denticles (Payre et al., 1999; Delon et al., 2003). All cells that will produce a denticle express svb and the target genes that it activates (Fig. 2D, E). Wg signaling represses svb and thereby defines the domain of naked cuticle. In a wg mutant, svb expression was detected uniformly across the ventral epidermis, correlating with the excess formation of denticles (Payre et al., 1999). This effect was also observed in wild-type embryos where svb was driven artificially: transgenic svb expression produced ectopic denticles within the naked cuticle domain (Fig. 2F).

How Wg signaling generates the diverse array of denticles within each belt is less clearly understood. Work with a temperature sensitive allele of wg (wg^IL114) revealed that Wg signaling at early stages of development, between 4 and 6 hours after egg-laying (AEL), was required for denticle diversity while later signaling specified the naked cuticle part of the pattern (Bejsovec and Martinez Arias, 1991). Denticle diversification may result from a relayed induction of other signaling pathways. The early phase of Wg signaling is required for stabilizing engrailed (en) expression in the row of cells just posterior to the wg-expressing row (Bejsovec and Martinez Arias, 1991; Heemskerk et al., 1991). En in turn controls expression of hedgehog, providing a source of Hh signal emanating from the most posterior rows of cells in each segment (Lee et al., 1992). Wg signaling also influences both the Notch and the EGF signaling pathways, regulating expression of key pathway components in defined domains within each segment (Fig. 2A). Wg promotes epidermal expression of Serrate, a Notch ligand, and restricts expression of rhomboid, which is required for producing the soluble form of EGF ligand (Alexandre et al., 1999). Expression of these molecules in specific rows of cells underlying the denticle belt correlates with proper orientation and type of denticle (Alexandre et al., 1999; Wiellette and McGinnis, 1999; Walters et al., 2005). In addition, EGF signaling is required in the naked cuticle-secreting epidermal cells to promote their survival (Urban et al., 2004), and the Fat/Dachsous protocadherins contribute to denticle polarity (Donoughe and Dinardo, 2011). The tendon-specifying transcription factor Stripe (Becker et al., 1997) also contributes to denticle polarity. Its expression at the segment boundaries is controlled directly by Wg and Hh signaling (Piepenburg et al., 2000) and it is essential for the correct orientation of denticles in rows 1 and 4 of the belt, which point in an anterior direction (Dilks and DiNardo, 2010). Thus many different molecules may contribute to coordinating and shaping the denticle construction elements, encoded by Svb target genes, as they build individual denticle types to produce the final pattern.

Recent work suggests a continuing direct role for Wg signaling in shaping the denticle types, a requirement that is buffered by the cytoplasmic myosin, Crinkled (Ck) (Bejsovec and Chao, 2012). Ck is broadly expressed in the embryo, but is not essential, with loss of function mutants showing mild phenotypes and hypomorphic mutants occasionally surviving to adulthood (Kiehart et al., 2004). Loss of function ck mutants have normal ventral denticle structure (Fig. 2G, H), but when Wg activity was also removed, the denticles were reduced to rounded lumps (Fig. 2I, J). This Wg-specific effect correlates with aberrant Svb regulation, because gain or loss of Svb function caused similar structural defects in the ventral denticles (Bejsovec and Chao, 2012). This effect was also produced by gain or loss of Miniature (Bejsovec and Chao, 2012), a Svb target (Chanut-Delalande et al., 2006) which encodes a transmembrane protein thought to link the plasma membrane with the overlying cuticle. Thus, imbalances in Svb activity and in its target gene products can be tolerated when Ck activity is normal but not when Ck is absent. Since Ck myosin is an actin-based motor protein, it may increase the efficiency of denticle formation by positioning or organizing denticle construction elements, and this activity somehow buffers against fluctuations in the relative amounts of these elements. Moreover, experiments using the temperature sensitive wg mutation revealed that in the sensitized ck mutant background, loss of Wg activity even at late stages resulted in rounded denticles (Bejsovec and Chao, 2012). This indicates that, without the buffering myosin activity, the requirement for Wg signaling is extended, and denticle-producing cells continue to respond to it and use that information as they shape the mature structure of denticles.

Wg Production and Distribution

The mechanism by which Wg moves across the segment to influence epidermal cell fates remains a subject of debate. The lipid modification performed by porcupine (porc), which encodes an acyltransferase (Kadowaki et al., 1996; Tanaka et al., 2002), is essential for proper glycosylation, secretion and function of Wnt molecules. This acylation accounts for early observations that Wg protein localized internally in membrane-bound vesicles and was enriched at the plasma membrane (Gonzalez et al., 1991). Several models have been proposed for how lipid-modified Wg molecules might be distributed beyond the immediate neighbors of wg-expressing cells. Work on this problem has focussed mostly on Drosophila wing imaginal discs, where Wg moves dozens of cell diameters away from its source. Small extracellular vesicles called argosomes were observed when a glycosylphosphatidylinositol(GPI)-linked Green Fluorescent Protein (GFP) was expressed in a portion of the wing disc (Greco et al., 2001). These have since been correlated with lipoprotein particles, which appear to mediate the movement of both Wg and Hh across the developing wing imaginal disc (Panakova et al., 2005). The formation of lipid rafts, through the action of the reggie/flotillin protein family, enhances this process (Zhai et al., 2004; Katanaev et al., 2008; Solis et al., 2013). In addition, a secreted Wg-interacting molecule (Swim), related to the vertebrate Lipocalin family, was identified by co-purification with Wg and found to be required for long-range Wg signaling in the wing imaginal disc (Mulligan et al., 2012). An alternative mechanism for Wg movement involving long filopodia-like cell protrusions called cytonemes (Ramirez-Weber and Kornberg, 1999) has also been proposed. These actin-rich structures reach from the disc periphery to the central region where many signaling molecules are produced. Vesicles enriched in Wg, as well as the TGFbeta family member Decapentaplegic (Dpp), have been detected within cytonemes. However these signal-packed vesicles may not be relevant to patterning within the disc: the only signaling function attributed to cytonemes so far has been in interactions between the disc epithelium and overlying tracheal tube primordia (Roy et al., 2011).

It is not clear whether these proposed mechanisms could contribute to Wg protein distribution during embryogenesis. Swim is not present during, nor required for, embryonic phases of Wg signaling (Mulligan et al., 2012); nor is lipophorin, the main component of lipoprotein particles, expressed at significant levels in the fly embryo during the stages at which Wg signaling specifies epidermal pattern (Flybase: Graveley et al., 2011; Marygold et al., 2013). Likewise, cytonemes have been observed only in imaginal discs. Thus none of these mechanisms seems likely to account for embryonic Wg distribution. Movement of Wg from cell to cell in the embryonic epidermis is known to require active endocytosis (Bejsovec and Wieschaus, 1995). Blocking endocytosis in the engrailed-expressing cells, just posterior to the stripe of wg-expressing cells (Fig. 2A), resulted in pattern disruptions on the far side of the restricted zone (Moline et al., 1999). These experiments suggested that in the embryo, Wg transits through cells in an active process that may differ significantly from the extracellular routes used to pattern the much larger imaginal disc epithelium. If so, there may be two possible outcomes when a cell receives Wg: either the protein signals and is degraded or it is moved intact across the cell and re-exported. Consistent with this idea, mutations that diminish the signaling activity of Wg increase its stability (van den Heuvel et al., 1993) and in some cases increase the distance over which it can spread (Bejsovec and Wieschaus, 1995; Moline et al., 2000).

Certain mutations in wg also differentially disrupt protein distribution versus signaling activity, indicating that these functions are genetically separable within the Wg ligand. Three “transport-defective” missense alleles were found to restrict Wg distribution without blocking signal to immediate neighbors (Dierick and Bejsovec, 1998). Mutant Wg protein accumulated in and around the wg-expressing cells and stabilized en target gene expression in the neighboring row of cells. These transport-defective mutants also displayed a small expanse of naked cuticle in each segment at the end of embryogenesis, corresponding to the restricted distribution of mutant protein. All three mutations changed amino acid residues that are highly conserved throughout the Wnt family. Since these molecules were able to transmit signal locally, the conservation of the affected amino acids cannot be due to a requirement in signaling. Rather the conservation may indicate that these residues are required for an intercellular transport function in all Wnt molecules. A wg mutation that showed the opposite effect was also isolated: this transport-normal mutant allele produced a punctate, vesicular distribution indistinguishable from wild-type Wg protein, but with no detectable signaling activity (Bejsovec and Wieschaus, 1995). Thus it is possible for mutant Wg protein to be endocytosed and to move from cell to cell without activating its cell surface receptor complex. This raises the possibility that some other cell surface complex or chaperone molecule handles Wg/Wnt during its distribution across tissues.

Wg/Wnt Protein Structure

In a recent breakthrough, the structure for a Wnt protein was solved using X-ray crystallography (Janda et al., 2012). In order to solubilize a functional lipid-modified Wnt protein, the Xenopus Wnt8 was co-expressed in Drosophila S2 cells with the extracellular domain of its cognate receptor, Frizzled8. XWnt8 and the cysteine-rich domain (CRD) of Fz8 were then purified together and co-crystallized. The complex was soluble because, as the solved structure shows, the palmitate lipid moiety was completely buried in a hydrophobic cleft of the Fz8-CRD. This interaction shields the palmitate group (attached at S239 in Wg, S209 in Wnt3a) but raises the question of how it might occur when Wnt and Fz are not expressed in the same cell, as they were for the co-crystallization experiment. Although there is evidence for an autocrine signaling loop for Wg (Hooper, 1994), much of Wnt signaling occurs in a paracrine fashion in both flies and vertebrates. That is, Frizzled receptors act in cells that receive Wnt signal but do not express wg/Wnt. Thus it is not clear how the lipid moiety of Wnt would be shielded during transit to the responding cell and then unleashed to interact with Frizzled receptor.

A chaperone molecule, called Evi/Sprinter/Wntless (Wls), is required in Wg/Wnt-producing cells for proper secretion of ligand (Banziger et al., 2006; Bartscherer et al., 2006; Goodman et al., 2006). This multipass transmembrane protein was found predominantly in the endoplasmic reticulum and colocalized with Wg/Wnt throughout the secretory pathway. Wls may interact with the palmitate group of Wnt: Porc-mediated lipidation was found to be required for interaction between Wg/Wnt and Wls (Coombs et al., 2010; Herr and Basler, 2012). Wls then is a good candidate for the delivery mechanism by which Wg/Wnt arrives at the plasma membrane, but it does not explain how Wg/Wnt moves from the surface of secreting cells to that of responding cells, where it contacts the Fz receptor.

A second question that arises from the crystal structure is whether the Wnt-Fz complex represents the interaction that triggers signal transduction. Experiments with chimeric and mutated Frizzled molecules showed that the Fz-CRD was not absolutely required for signaling, although it did contribute to efficiency of signaling (Chen et al., 2004; Wu et al., 2004; Povelones and Nusse, 2005). These studies suggested that the ligand binding domain of Frizzleds may reside in the 6 exoplasmic loops of the 7 transmembrane-spanning region, as it does for the larger class of 7 transmembrane domain proteins, G-protein coupled receptors (reviewed in Venkatakrishnan et al., 2013). Thus the role of the Fz-CRD may be to concentrate Wnt ligand close to the ligand binding pocket, thereby improving transmission of signal. A similar role has been proposed for sulfated polysaccharides associated with cell surface proteoglycans, such as the fly Glypican encoded by Dally. Fly mutations blocking the synthesis of these molecules mimic loss of function wg phenotypes, but these pattern disruptions were rescued by overexpressing wg (Hacker et al., 1997; Lin and Perrimon, 1999). Thus sulfated sugars and proteoglycans are dispensable when Wg ligand levels are not limiting, and so must have an auxiliary rather than an essential role in triggering signal.

Understanding the basic structure of a Wnt molecule provides new insight into function. We now know which cysteine residues form disulfide bond pairs, and that the N-terminal and C-terminal domains fold separately and are not linked by disulfide bonds (Fig. 3A). The temperature sensitive fly mutation (wg^IL114), which has proven so useful in dissecting in vivo Wg activity, changes Cys104 to Ser in the N-terminal region (van den Heuvel et al., 1993); the corresponding position in XWnt8 disrupts the first disulfide bond and would destabilize interaction between the first alpha-helical region and the linker joining it to the second alpha helix (Fig. 3A). Two of the three transport-defective missense mutations (wg^PE6 A136V, wg^NE1 G258D) alter amino acids within predicted alpha-helical regions of the N-terminal helical bundle domain. The third (wg^NE2 C242Y) would disrupt the disulfide bond that stabilizes the base of the “thumb”, which attaches to the palmitate moiety. All three of these then might change the geometry of the lipid projection, suggesting that proper presentation of the lipid influences distribution of Wnt. Indeed, this portion of the molecule is intact in the mutant form of Wg that moved normally but did not signal: the wg^CE7 mutation introduces a premature stop codon that would truncate Wg at the end of the N-terminal helical bundle. Conversely, a missense mutation (wg^PE2 V453E) in the C-terminal region of Wg diminished signaling activity, in a fashion that did not restrict protein distribution (Bejsovec and Wieschaus, 1995; Moline et al., 2000). The corresponding position in XWnt8 lies between two disulfide bridges that stabilize the “index finger”, which contacts a different portion of the Fz CRD than does the “thumb” (Fig. 3C). The behavior of this mutant Wg molecule suggested that it had lower affinity for the receptor when proteoglycans were present: the phenotype was suppressed by overexpressing Fz2 receptor and it was enhanced by overexpressing Dally (Moline et al., 2000). It therefore seems likely that the C-terminal domain of Wg is essential for activating the receptor complex.

Figure 3.

Schematic diagrams of Wingless and Arm structure, and pathway relationships. (A) Wingless structure based on the crystal structure for XWnt8 (Janda et al., 2012). Wingless has a longer N-terminal region and an 85 amino acid insert (red lines) relative to vertebrate Wnt molecules; the 85 amino acid region can be deleted without disrupting Wg activity (Hays et al., 1997). Approximate positions of wg mutations (van den Heuvel et al., 1993; Bejsovec and Wieschaus, 1995; Dierick and Bejsovec, 1998) are indicated in red. Cylinders represent alpha helices, block arrows indicate beta-sheets and dotted lines indicate positions of disulfide bonds (after Janda et al., 2012). (B) Schematic diagram of Armadillo/beta-catenin. The N-terminal region contains phosphorylation sites for CK1 and Zw3/GSK3beta kinase (dark starbursts), which target it for degradation. The central Arm repeats provide protein interaction domains for a variety of binding partners. Regions of Arm/beta-catenin where binding partners interact are shown as lines, with partners that favor signaling shown above and those that oppose it shown below. Phosphorylation sites that correlate with activation of Arm/beta-catenin, rather than destruction, are depicted as light starbursts. The C-terminal region is dispensable for Arm/beta-catenin’s adhesion function but is essential for its Wg/Wnt signaling activity. (C) Schematic diagram of the Wingless signaling pathway. Wg structure is based on a space-filling model (from Janda et al., 2012), showing glycosylation modifications (light gray) and palmitate group (white) projecting into Frizzed CRD. Wg binding to Fz and Arrow brings them together and this recruits Dsh to the membrane. The PDZ and DIX domains are protein-interaction motifs that are essential for Wg signaling. Dsh somehow inactivates the destruction complex, perhaps by bringing Axin to the membrane where it is degraded. The destruction complex is composed of the Axin and Apc scaffolding molecules, which bring CK1 and Zw3/GSK3beta kinase together with Arm/beta-catenin. Apc has multiple protein interaction motifs including Arm repeats (light green) and Arm/beta-catenin binding domain repeats (dark green), as well as an oligomerization domain at the N-terminus and domains that allow microtubule association at the C-terminus. Phosphorylation of Arm/beta-catenin’s N-terminal domain (P) targets it for ubiquitination and degradation by the proteasome. When the destruction complex is inactivated by receptor/Dsh activity, Arm/beta-catenin is stabilized and translocates to the nucleus where it binds the N-terminus of Tcf, displacing the Groucho co-repressor and recruiting activators to drive target gene expression.

The Wg/Wnt Signal Transduction Pathway

Wnt binding to the Fz receptor appears to cause clustering with a coreceptor, the single transmembrane-spanning Arrow/LDL-related receptor protein (LRP) (Tamai et al., 2000; Wehrli et al., 2000). Although there is no direct evidence that these proteins interact (Wu and Nusse, 2002), a chimeric molecule that artificially fused the extracellular and transmembrane portions of Fz with the cytosolic domain of Arrow produced constitutive pathway activity (Tolwinski et al., 2003). This suggested that formation of Fz-Arr heterodimer is a driving force in transducing signal across the membrane. Events on the inside of the plasma membrane are less clear, although the consensus opinion is that receptor activation somehow inactivates the destruction complex to allow accumulation of Arm/beta-catenin (Fig. 3C). Two essential pathway components interact with the cytosolic face of the receptors: Dishevelled (Dsh) binds to intracellular loops and the cytosolic region of Fz (Wong et al., 2003), while Axin, a destruction complex component, binds to the cytosolic domain of Arrow/LRP (Mao et al., 2001). The interaction of Dsh with Fz is known to be important for planar cell polarity (reviewed in Seifert and Mlodzik, 2007), a non-canonical Wnt signaling process. Since Dsh also binds to Axin (Kishida et al., 1999; Smalley et al., 1999), Dsh-Fz interaction may facilitate recruitment of Axin to the Arrow/LRP cytosolic domain. This Axin movement could be a crucial event in pathway activation (Cliffe et al., 2003; Lee et al., 2003). Axin binds Apc to form a scaffold that facilitates phosphorylation of Arm/beta-catenin by bringing it together with kinases. Casein kinase 1 phosphorylates a residue that “primes” the system for subsequent phosphorylations by Zeste white 3/Glycogen synthase kinase 3 beta (Yanagawa et al., 2002). These phosphorylations make Arm/beta-catenin a target for ubiquitin ligase, leading to degradation by the proteasome. In the fly embryo, this gives rise to a dramatic segmentally-striped pattern of low and high levels of cytosolic Arm, with highest Arm levels centered over the wg-expressing row of cells (Peifer et al., 1994).

Regulation of Axin stability may explain how Arm/beta-catenin degradation is blocked in cells responding to Wg. Axin, of all the Wnt pathway components, is thought to be at limiting levels within the cell (Salic et al., 2000; Lee et al., 2003). Axin protein levels drop in response to Wg signaling in fly embryos, producing a striping effect that precisely complements the striping pattern of Arm protein levels (Tolwinski et al., 2003). This result led to the satisfying idea that Wnt receptor activation recruits Axin to the membrane where it is degraded, thereby inactivating the destruction complex. However, this result depended on overexpressed Axin; the Axin antibody used in this work was unable to detect endogenous low levels of Axin, so we do not know if the endogenous Axin behaves in this way. Furthermore, if Axin levels are limiting, then overexpressing Axin in the embryonic epidermis should have caused patterning defects. The high levels used in this experiment, which revealed the Wg-mediated striping, did not do so. Controversy over this issue, and the putative role of Dsh in the process, still rages within the field, as evidenced by conflicting reports at the most recent Wnt meeting (Verkaar et al., 2012). In particular, new data show that Wnt might act by blocking ubiquitination, causing Arm/beta-catenin to remain associated with the Axin/Apc/kinase destruction complex (Li et al., 2012). This would saturate the complex, effectively shutting it down while de novo synthesized Arm/beta-catenin translocates to the nucleus and drives target gene expression. This idea contrasts sharply with other work showing that Wnt response acts upstream of Arm/beta-catenin phosphorylation and involves inactivation of destruction complex function (Fiedler et al., 2011; Hernandez et al., 2012).

Even the dogma that Arm stabilization drives pathway activity is not without controversy. Stabilization correlates with activation but is not always required; some pathway activity can be observed in the absence of high fluxes of Arm protein accumulation. For example, the wg^PE2 V453E mutant form of Wg was found to induce en target gene activation in neighboring cells but did not produce the expected Arm stripes in fly embryos (Moline et al., 2000). Therefore, Arm protein below the level of detection is sufficient for signaling. Arm/beta-catenin also undergoes phosphorylation events that are unrelated to its stabilization, and these may be important for its activity (reviewed in Valenta et al., 2011). Arm/beta-catenin is a large molecule with multiple domains that mediate its essential roles in cell adhesion and signal transduction (Fig. 3B). The GSK3beta phosphorylation sites that negatively regulate stability are located in the N-terminus (Morin et al., 1997; Pai et al., 1997). The main portion of the molecule consists of 12 repeats of a 40 amino acid motif, termed an Arm repeat (Peifer et al., 1992); the Arm repeats fold into bundles of alpha helices that form an extended groove (Huber et al., 1997). Many binding partners interact with Arm along this groove. The binding sites for adhesion components such as E-cadherin and alpha-catenin, for the destruction complex components Apc and Axin, and for the transcriptional regulators Tcf and Legless/Bcl9, all fall within the Arm repeat region. Phosphorylation sites in the Arm repeats and C-terminus may regulate the switch between Arm’s cell adhesion versus signaling functions. The carboxy terminal domain of Arm has long been known to be essential for its role in Wg signaling (Peifer et al., 1991; Orsulic and Peifer, 1996). Phosphorylations in this region appear to shift the binding affinity of Arm for E-cadherin, perhaps by modulating a self-inhibitory folded conformation where the C-terminus blocks the E-cadherin binding portion of the Arm repeats (Gottardi and Gumbiner, 2004). This then favors interaction with Tcf and other transcriptional machinery and thereby promotes Wnt target gene expression. Other phosphorylation sites may also regulate binding partner affinity. For example, phosphorylation of Tyr142, in the first Arm repeat (line in Fig. 3B), is required for vertebrate beta-catenin to recruit the transcriptional co-activator Bcl9 (Brembeck et al., 2004). However, the corresponding Tyr phosphorylation in Arm was shown to be dispensable for Legless binding in Drosophila (Hoffmans and Basler, 2004), so this does not appear to be a conserved mechanism for switching between the adhesive and signaling function.

The Arm repeats also provide the means by which Arm translocates into the nucleus. Arm/beta-catenin lacks classical nuclear localization and export sequences, but Arm repeats 10 – 12 have been shown to be sufficient for nuclear import and export (Sharma et al., 2012). The conventional nuclear import chaperone, Importin-beta, possesses similar Arm repeats, and both molecules interact with the same nuclear pore complex proteins to transit through the nuclear pore (Sharma et al., 2012). Once inside the nucleus, Arm/beta-catenin binds to Tcf, an HMG-box DNA-binding protein. In the absence of Wg signaling, Tcf represses Wg/Wnt target genes by forming a complex with Groucho/TLE, a transcriptional co-repressor (Cavallo et al., 1998). When Arm enters the nucleus, it displaces Groucho (Daniels and Weis, 2005) and recruits Lgl/Bcl9 to convert the Tcf complex into one that activates target gene expression (Kramps et al., 2002). Thus pathway activation first “derepresses” target genes, and this can contribute to patterning even when full activation of target genes is not achieved. The Tcf loss of function phenotype (van de Wetering et al., 1997) (Fig. 1D) is less severe than that of the wg loss of function (Fig. 1B): both cuticle patterns lack naked cuticle, which requires full Wg activity levels, but Tcf mutants have substantially greater denticle diversity and showed more expression of engrailed, a Wg target gene (Cavallo et al., 1998). The wg; Tcf double mutant pattern looked similar to the Tcf single mutant pattern (Cavallo et al., 1998), meaning that the severity of the wg cuticle pattern depends on gene repression by wild-type Tcf activity. In other words, the greater denticle diversity observed in Tcf mutants is due to derepression of Wg target genes.

The identity of Wg/Wnt target genes, and their function in determining epidermal patterning, is another unsolved part of the puzzle. Both wg itself and the segment polarity gene engrailed have long been known to require Wg signaling for continued expression in the epidermis (DiNardo et al., 1988; Bejsovec and Martinez Arias, 1991; Hooper, 1994), and stripe requires both Tcf (Wg pathway) and Ci (Hh pathway) input for correct expression at points of muscle attachment in the epidermal segment borders (Piepenburg et al., 2000). decapentaplegic (dpp) as well as the homeotic gene Ultrabithorax are directly controlled by Wg signaling in the embryonic gut (Hoppler and Bienz, 1995; Yu et al., 1996). More recent biochemical studies established that Notum (Hoffmans et al., 2005; Parker et al., 2008) and the segment polarity gene naked (Fang et al., 2006; Li et al., 2007) are activated directly by Wg through Tcf binding sites in their regulatory regions. Both gene products negatively regulate the Wg pathway and so serve as feedback inhibitors for the pathway. Although their function is regulatory rather than morphogenetic, their analysis may lead to discovery of the morphogenetic targets that Wg signaling regulates during pattern formation. Careful study of Wg regulatory elements in the Notum and naked promoters revealed a “helper” site in addition to the known HMG-box recognition sequences (Chang et al., 2008). The helper sequence is bound by a region in Tcf C-terminal to the HMG domain, termed the C clamp. This discovery has led to a better understanding of the sequence requirements for Tcf binding, which in turn has led to a growing list of validated targets (Chang et al., 2008).

Prospects for the Future

The question of how Wg signaling is translated into epidermal cell fates, which generate the pattern secreted in the cuticle, is only partially answered. The naked cuticle portion of the pattern appears to be a default state adopted by epidermal cells when svb is repressed by Wg. Within the svb-expressing domain, Wg signaling somehow generates diversity among cell fates to specify the different denticle types. Presumably this reflects the coordination of Svb target gene products to produce distinct denticle morphologies, in a process that is modulated by the cytoplasmic Ck myosin. This aspect of the pattern may involve integration with the Hh, Notch and EGF signaling pathways, and perhaps also Dpp signaling as there is a dorsal-ventral modulation in denticle shapes within the belts. Identifying all target genes of these signaling pathways will be an important step toward full understanding of the process, just as identifying genetic components required for Wg-mediated patterning has revealed the inner workings of the pathway.

A second open question is whether we have in fact identified all of the pathway components. Perhaps the continuing controversies over pathway mechanics indicate that some critical piece of the puzzle is still missing. Zygotic and maternal genetic screens based on the fly cuticle pattern have been tremendously successful, but are limited in two ways. First, they will miss genes with mutant phenotypes that disrupt embryogenesis at an earlier point in a way that prevents epidermal maturation and cuticle deposition. In fact, arm is such a gene, but because of beta-catenin’s modular nature it was possible to recover arm mutations that disrupt signaling without affecting adhesion. Truncations that remove the C-terminal transactivation region (Fig. 3B), but not the Arm repeats, retain adhesive function which is essential for egg formation (Peifer and Wieschaus, 1990; Peifer et al., 1993). There may be other pleiotropic pathway components that do not have such a modular structure and so cannot be disrupted without dire consequences for the cell. Second, genetic screens cannot readily detect mutations in gene functions for which other, related molecules can compensate. This was the case for the Wg/Wnt receptor, Frizzled: two closely related family members, fz and Dfz2, have sufficient overlap in their functions that both must be knocked out to produce a wg mutant phenocopy in the fly (Bhanot et al., 1996; Bhanot et al., 1999). Because of this genetic redundancy, biochemical and cell culture experiments were crucial in making the connection between Wg ligand and its receptor. Current efforts to map protein-protein interactions on a genome-wide scale (Guruharsha et al., 2011) may yield similar break-throughs in finding missing links in the pathway. In addition, genetic screens can be made even more powerful by performing mutageneses in a sensitized background, that is, in a fly strain that carries a weak mutation in some signaling pathway component. These approaches combined may finally yield answers to long-standing questions about how Wg signal is transduced and then translated into cell morphologies and body pattern. And almost certainly, they will raise more questions.

Conclusions

The epidermal pattern of the fly embryo has been an important model system for elucidating the mechanism of Wg signal transduction. Mutations that mimic wg gain and loss of function phenotypes, called phenocopies, have identified many of the players in the pathway and have provided reagents for analyzing function of the pathway. However, the picture is still incomplete and there may be missing components that, once found, will resolve current confusion over how the pathway is activated. Mutations within the wg gene itself have provided insight into the function of the signal, and in combination with the recently-solved Wnt crystal structure, reveal which positions are essential for ligand movement versus signaling activity. But the molecules that translate Wg pathway activity into epidermal pattern elements have remained elusive: Wg-mediated repression of Svb sculpts the naked cuticle expanses that separate denticle belts, but the forces that shape denticle morphologies within the belt are still under investigation. Solving this puzzle will reveal how positional information generated by cell to cell signaling is converted into body form.

The cuticle pattern secreted by the fruitfly embryonic epidermis is exquisitely sensitive to Wingless/Wnt signaling and has proven to be an excellent model system for genetic dissection of this important cancer pathway.

Abbreviations

Wg

Wingless

Arm

Armadillo

Svb

ovo/shaven-baby

En

Engrailed

Dpp

Decapentaplegic

Hh

Hedgehog

Ck

Crinkled

Fz

Frizzled

Arr

Arrow

LRP

LDL-receptor related protein

Dsh

Dishevelled

Zw3

Zeste white 3

GSK3beta

Glycogen Synthase Kinase 3 beta

CK1

Casein Kinase 1

Apc

Adenomatous polyposis coli

EGF

Epidermal growth factor

HMG

High mobility group

N-terminal/terminus

amino-terminal/terminus

C-terminal/terminus

carboxy-terminal/terminus