Arrow (LRP6) and Frizzled2 cooperate to degrade Wingless in Drosophila imaginal discs

Embryonic development as well as a healthy adult life requires cell signalling to be finely controlled. Cell fate specification during development requires exquisite spatial and temporal control of signalling(Martinez Arias and Steward,2002) and, in adult tissues, excess signalling is often associated with cancer (Bache et al.,2004). Degradation of ligand-receptor complexes has emerged as an important cell biological process that downregulates the activity of various signalling pathways (Piddini and Vincent,2003; von Zastrow,2003). Classic examples of receptors that are downregulated by degradation are the EGFR and β-adrenergic receptors. More recent work with developing tissues has shown that degradation ensures the proper spatial distribution of extracellular ligands. In Drosophila embryos, the range of Wingless, a secreted protein of the Wnt family, is developmentally controlled by degradation in lysosomes(Dubois et al., 2001). Likewise, in Drosophila imaginal discs, endocytosis and subsequent targeting to lysosomes affect the distribution of the two signals encoded by hedgehog and decapentaplegic(Chen and Struhl, 1996; Entchev et al., 2000). Here,we investigate the degradation of Wingless in wing imaginal discs of Drosophila. In this tissue, Wingless is produced by a narrow strip of cells located at the dorsoventral boundary (the prospective wing margin) and spreads symmetrically over a range of 20-25 cells(Cadigan et al., 1998; Strigini and Cohen, 2000).

Typically, the degradation of extracellular ligands is initiated by receptor-mediated endocytosis, which is then followed by targeting to multi-vesicular bodies and lysosomes. Two classes of receptors have been implicated in Wnt signalling, the Frizzled class of seven-transmembrane receptors and the LRP (LDL receptor-related protein) family. According to the current model (Huelsken and Behrens,2002), Wingless signalling is initiated by the binding of Wingless to Frizzled or Frizzled2. This leads to the recruitment of Dishevelled at the plasma membrane and the association of the Frizzled-Wingless complex with Arrow, the Drosophila homologue of LRP5/6. Arrow is then able to recruit Axin, thus allowing Armadillo (the Drosophila homologue ofβ-catenin) to accumulate. Genetic analysis clearly shows that both classes of receptors (Frizzled and Arrow/LRP) are essential for signal transduction (Kennerdell and Carthew,1998; Chen and Struhl,1999; Wehrli et al.,2000). Binding assays suggest that ligand capture is primarily performed by a Frizzled family member: Frizzled receptors are able to recruit Wingless at the cell surface (Bhanot et al., 1996; Wu and Nusse,2002), while similar experiments have failed to show that Arrow has such activity (Wu and Nusse,2002).

As shown by Cadigan et al. (Cadigan et al., 1998), overexpression of Frizzled2 causes Wingless stabilization in wing discs. Indeed, it has been suggested that Frizzled2 could protect Wingless from degradation(Cadigan et al., 1998),possibly by titrating a putative extracellular protease(Eldar et al., 2003). The stabilizing effect of overexpressed Frizzled2 is somewhat surprising because ligand degradation is usually initiated by receptor-mediated endocytosis and subsequent targeting to lysosomes. In this paper, we investigate and compare the role of the two signalling receptors Frizzled2 and Arrow in Wingless degradation. We show that, while Frizzled2 internalizes Wingless, Arrow targets internalized Wingless/Frizzled2 complexes to a degradative compartment. Therefore, the two receptors contribute specific features that together trigger Wingless degradation. The contribution of Arrow to degradation explains why overexpression of Frizzled2 alone leads to stabilization.

Arrow was tagged at its C terminus with a glycine linker and an HA tag(Arrow-HA). To generate ArrowΔC-HA, amino acids 1477 to 1602 (the N terminal half of the intracellular region) were replaced with the glycine linker and the HA tag. Frizzled2-FLAG carries the FLAG tag at the C terminus and Frizzled2ΔC-FLAG lacks residues 621 to 694 retaining only 14 residues after the 7th transmembrane domain (with the FLAG tag at the C terminus). The above constructs were introduced in pUAST for transgenic expression and pMT/V5-HisA (Invitrogen) for expression in cell culture. hs-Frizzled2-FLAG was made by subcloning Frizzled2-FLAG from pUAST into CaSpeR-hs. Frizzled2-HA was generated by replacing the FLAG-tag sequence with a HA-tag sequence in Frizzled2-FLAG. UAS-Frizzled2GPI (myc-tagged) was obtained from R. Nusse (Stanford University). UAS-Shi[ts] was a gift from T. Kitamoto. Expression of UAS constructs in cultured cells was induced by co-transfecting pMTGal4 (driving Gal4 expression under the metallothionein promoter), a gift from A. Ephrussi (EMBL). Additional details are available upon request.

The following transgenic stocks were used for exogenous expression: UAS-HRP-wingless (Dubois et al.,2001), UAS-Frizzled2-FLAG/CyO,UAS-Frizzled2ΔC-FLAG/CyO, UAS-Arrow-HA,UAS-ArrowΔC-HA/CyO, dpp-Gal4 UAS-Arrow-HA/TM6B (all generated for this study), UAS-Dally-like (from S. Cohen, EMBL),UAS-Armadillo[S10] (Pai et al.,1997), UAS-Armadillo (Marygold and Vincent, 2003), UAS-Shibire[ts](Kitamoto, 2001) and UAS-Frizzled2GPI (Cadigan et al.,1998). They were driven with dpp-Gal4, ap-Gal4 or en-Gal4. hs-Frizzled2-FLAG was generated for this study while hs-Patched was from Isabel Guerrero(Sampedro and Guerrero,1991).

The following genotypes are depicted: dpp-Gal4/UAS-HRP-wingless(Fig. 1A), dpp-Gal4/UAS-Frizzled2-FLAG (Fig. 2A-A′′′, Fig. 3A, Fig. S1A in the supplementary material), dpp-Gal4/UAS-Frizzled2GPI (Fig. 2B-B′′′, Fig. 4B, Fig. S1C in the supplementary material), UAS-Frizzled2-FLAG/+; dpp-Gal4/UAS-Arrow-HA(Fig. 3B), UAS-Armadillo[S10]/+; dpp-Gal4/UAS-Frizzled2-FLAG(Fig. 3D), dpp-Gal4/UAS-Dally-like (Fig. 4A), UAS-Arrow-HA/+; dpp-Gal4/UAS-Dally-like(Fig. 4A′), UAS-Arrow-HA/+; dpp-Gal4/UAS-Frizzled2GPI(Fig. 4B′), dpp-Gal4/UAS-Frizzled2ΔC-FLAG(Fig. 4C), UAS-Frizzled2ΔC-FLAG/+; dpp-Gal4/UAS-Arrow-HA(Fig. 4C′), en-Gal4/UAS-ArrowΔC-HA; hs-Frizzled2-FLAG/+(Fig. 4D-D′) shi[ts]/Y; dpp-Gal4/UAS-Arrow-HA(Fig. 5D), hs-Frizzled2-FLAG/+ (Fig. 6A-D), hs-Patched/+(Fig. 6E-F) and UAS-Frizzled2-FLAG/+; dpp-Gal4/UAS-Frizzled2-FLAG (see Fig. S1B in the supplementary material).

For the shibire[ts] experiments, the following genotypes are depicted: dpp-Gal4/UAS-Shibire[ts](Fig. 1D), ap-Gal4/+;UAS-Shibire[ts]/+ (Fig. 1F), shibire[ts]/+; en-Gal4 UAS-Armadillo(Fig. 1G) and shibire[ts]/Y; en-Gal4 UAS-Armadillo/+.

For loss-of-function studies, clones of mutant cells were induced by Flp-mediated mitotic recombination by heat-shocking larvae 2 days after egg laying for 1 hour at 37°C.

For arrow mutant clones, the following stocks were used: FRT42D pwn arr[2]/Gla Bc (obtained from R. Mann, Columbia University), FRT42D pcna ubi-GFP/CyO-GFP; lama-Gal4 UAS-Flp (from I. Salecker, NIMR, London), UAS-CD8-GFP hs-Flp; FRT42D tub-Gal80; tubulin-Gal4 (from S. Cohen,EMBL), FRTG13 arr[2]/CyO and y w hs-Flp; FRT G13 M(2)58F ubi-GFP/CyO (from X. Lin, Cincinnati Children's Hospital).

The following genotypes are depicted: y w hs-flp/+; FRT G13 M(2) 58F ubi-GFP/FRTG13 arr[2] (Fig. 7B-B′), FRT42D pcna/FRT42D pwn arr[2]; lama-Gal4 UAS-Flp (see Fig. S2 in the supplementary material) and UAS-CD8-GFP hs-Flp; FRT42D tub-Gal80/FRT42D pwn arr[2]; tubulin-Gal4 (see Fig. 3A-A′ in the supplementary material).

For frizzled frizzled2 mutant clones, the following stocks were used: y w hs-Flp; Sp/CyO; fz[P21] Dfz2[C2] ri FRT2A/TM2 (from G. Struhl, Columbia University) and y; M(3)I55 FRT2A; p[nls-GFP]/TM6B(from F. Schweisguth, Ecole Normale Superieure, Paris). The genotype depicted in Fig. 7A-A′ is y w hs-Flp/+; fz[P21] Dfz2[C2] ri FRT2A/M(3)I55 FRT2A.

For the deep-orange mutant clones the following stocks were used: dor[8] FRT18A/FM6 and w arm-LacZ FRT18A; hs-Flp(Sevrioukov et al., 1999). The genotype depicted in Fig. 1B,Cis dor[8] FRT18A/w arm-LacZ FRT18A; hs-Flp/+.

For hrs (l(2)23AdD28) mutant clones, the following stocks were used: hrs FRT40A/CyO and y w hs-Flp; ubi-GFP FRT40A. For the arrow rescue, the following stocks were used: arr[2]UAS-ArrowΔC/Gla Bc and arr[2] arm-Gal4/Gla Bc.

The following primary antibodies were used: mouse anti-Wingless 4D4(prepared from cells obtained from the DSHB), mouse M2 anti-FLAG (Sigma;1/15,000), rabbit anti-FLAG (Abcam; 1/3000), mouse anti-HA 1.1 (Babco;1/3000), Alexa-488 labelled mouse anti-HA 1.1 (Covance; 1/500), mouse anti-Myc(Roche; 1/600), rabbit anti-Myc (Santa Cruz; 1/500), rabbit anti-GFP (Abcam;1/2500), rabbit anti-β-galactosidase (Cappel; 1/12,000), mouse anti-Distalless (a gift from Ian Duncan; 1/500), mouse-anti-Armadillo N2 7A1(DSHB; 1/150), mouse anti-Engrailed 4D9 (DSHB;1/200) and mouse anti-Patched(a gift from Isabel Guererro; 1/50). Rabbit anti-Arrow antibody was raised for this study against peptide 892-908 (1/5000). Rabbit anti-Frizzled2 antibody was raised against peptide 232-251 (1/2000). The following reagents were used for secondary detection: Alexa488-conjugated goat anti-rabbit (Molecular Probes; 1/200), Alexa488-conjugated goat anti-mouse (Molecular Probes; 1/200),Alexa594-conjugated goat anti-mouse (Molecular Probes; 1/200), Cy5-conjugated goat anti-rabbit (Jackson; 1/200) and Alexa555 Zenon mouse IgG1 labelling kit(Invitrogen).

Heat-shock induced expression of Frizzled2 and Patched was obtained by incubating intact larvae at 37°C for 35 minutes and chasing at room temperature for the indicated times before fixation. Localized (UAS-Shi[ts])or uniform (shi[ts] hemizygous) temporary inhibition of shibire activity was obtained by incubating intact third instar larvae at 32°C for the indicated times before dissection.

Dextran labelling was carried out as described(Entchev et al., 2000) with a 10-minute pulse of Texas Red dextran (lysine fixable, M_r3000; Molecular Probes) followed by a 20-minute chase to label the endocytic compartment. Unless otherwise indicated, a standard antibody staining technique was used for wing imaginal disc labelling. The original protocol for extracellular staining (Strigini and Cohen, 2000) involves incubating live discs at 4°C to allow primary antibody binding in the absence of endocytosis(Strigini and Cohen, 2000). As endocytic blockade itself affects the spread of Wingless (see Fig. 1D), we sought to avoid the 4°C step and devised a post-fixation extracellular staining protocol. After fixation in 4% PFA/PBS for 20 minutes at room temperature (in the absence of detergent), discs were incubated for 30 minutes in 0.1% BSA/PBS. Primary antibody incubation was performed in S2 medium overnight at 4°C. Discs were then washed extensively in PBS and incubated for 2 hours at room temperature with the secondary antibody diluted in 0.1% BSA/PBS. Following extensive PBS washes, samples were post-fixed for 10 minutes in 4% PFA/PBS to avoid dissociation of antibody complexes from the antigen. This step was found to be crucial if the samples were further processed for intracellular staining. Control antibody staining showed that lumenal markers of intracellular organelles are not accessible by this staining protocol (see also Chen et al., 2004). In order to label exclusively intracellular Wingless, discs were first subjected to an extracellular staining with saturating amounts of anti-Wingless antibody as described above. This ensured that when another primary antibody incubation was performed, only intracellular Wingless molecules would be available for antibody binding. Extracellularly stained discs were then permeabilized for 20 minutes in 0.05% Triton-X-100/PBS, incubated in blocking solution (0.1% BSA,0.05% Triton-X-100/PBS) for 30 minutes and left overnight at 4°C in blocking solution containing 20 mg/ml of nonspecific mouse IgG1 to saturate remaining free sites on the secondary antibody used for extracellular labelling. Discs were then stained for 2 hours at room temperature with anti-Wg that had been fluorescently prelabelled using the Zenon antibody labelling kit (Invitrogen) according to manufacturer's instructions. Following a 30-minute wash in 0.05% Triton-X-100/PBS, discs were again postfixed and mounted.

A standard calcium-phosphate transfection protocol was used both for transient transfections and for the generation of stably transfected cell lines. GFP-Wingless medium was prepared from S2 cells stably transfected with a pMK33 construct containing a GFP-Wingless cDNA insert under the control of a metallothionein promoter (S2-GFPWg cells). GFP-Wingless- and control-conditioned media were prepared inducing S2-GFPWg and wild type S2 cells, respectively, for 2 days in serum free medium (SFM, Gibco) containing 0.375 mM CuSO₄. The medium was cleared by centrifugation at 38,000 g for 25 minutes, and then concentrated 10× using Centriprep YM-30 (Millipore). The concentrated medium was then supplemented with 2% BSA (final concentration) and filtered (filter unit 0.2 μM,Sartorius). Biological activity of GFP-Wingless conditioned medium was confirmed by a standard Armadillo stabilization assay. GFP-Wingless binding was carried out on ice as previously described(Bhanot et al., 1996) on S2 cells or S2R+ cells grown on coverslips and transiently transfected with the various receptor constructs and induced over night with 0.25 mM CuSO₄. After binding, internalization was performed by shifting cells to room temperature in normal growth medium for 30-40 minutes after a quick ice-cold PBS wash. Uninternalized, cell surface bound GFP-Wingless was stripped by a 30-second acid wash (0.5%Acetic acid, 0.5 M NaCl; pH 3) before fixation and antibody staining. GFP-Wingless was detected with an anti-GFP antibody. The transfected receptors were detected with antibodies against the relevant epitope tag. To allow for comparative binding/internalization analysis samples were processed in parallel.

All quantitation was performed using the freeware ImageJ 1.33u(http://rsb.info.nih.gov/ij/). For quantitation of GFP-Wg binding to S2 cells, total fluorescence of individual transfected and untransfected cells was calculated by applying an identical threshold to all images to eliminate intracellular background and summing the fluorescence intensity from a stack of 11 sections. In each experiment, the fluorescence intensity of individual overexpressing cells was corrected by subtracting the mean of the total cell fluorescence from the population of untransfected cells.

RNAi against arrow was performed by co-transfecting arrowdouble-stranded RNA into S2R+ cells together with pMT-Frizzled2-FLAG using the Effectene reagent (Qiagen) 2 days before the binding/internalization assay was carried out. Functional ablation of the Arrow protein was confirmed by the loss of signalling activity, as assayed by a modified superTOP FLASH luciferase reporter assay (Takemaru and Moon, 2000) (C. Alexandre, unpublished). To generate an arrow cDNA fragment suitable for in vitro transcription, the following PCR primers were used:5′ACGTTTAATACGACTCACTATAGGGAGAAAGATTGAGCGAGCCAGCAT and 5′TGCATTAATACGACTCACTATAGGGAGACTCGGCTCCTCCAAA.

In situ hybridization was performed according to standard protocols with a digoxigenin-labelled wingless probe (C. Alexandre, NIMR, London).

We have previously used a HRP-Wingless fusion protein expressed from a transgene to follow Wingless trafficking in Drosophila embryos(Dubois et al., 2001). As in embryos, this fusion protein is found in recognizable lysosomes in wing imaginal disc cells (Fig. 1A). To assess the importance of lysosomal activity in Wingless degradation in this tissue, we generated clones of cells that are mutant for deep-orange(dor), which encodes the homologue of yeast VPS18 and is required for trafficking to lysosomes (Sevrioukov et al., 1999). As shown in Fig. 1B,C, mutant cells accumulate Wingless. A similar result is seen in mutants of hrs (not shown), which encodes a ubiquitin-binding protein involved in targeting receptors to multivesicular bodies(Lloyd et al., 2002). Moreover, transient inhibition of endocytosis with a temperature-sensitive dominant-negative form of dynamin(Kitamoto, 2001) in dpp-Gal4 UAS-shi[ts] also leads to an expansion of the range(Fig. 1D). Overall, these results show that cells continuously degrade Wingless in a lysosomal compartment and that endosomal trafficking contributes to the gradient normally seen in wild-type discs.

No significant ectopic Wingless signalling is seen in dor or hrs mutant clones, as judged by target gene expression levels (not shown). Presumably, this is because in these cells, Wingless accumulates in a compartment where the receptors can no longer engage with the cytoplasmic signal transduction machinery. To determine whether endocytic trafficking contributes to signal downregulation, we assessed the effect of an earlier block in the endocytic pathway. The temperature-sensitive dominant-negative dynamin was expressed in the dorsal compartment (with the ventral compartment as a control). After 5 hours at 32°C (restrictive temperature), expression of distalless, a target of Wingless signalling, is noticeably depressed (Fig. 1E), suggesting a reduction in signalling, the opposite of our expectation. As a control, we assessed the expression of engrailed, a gene that is not known to be regulated by Wingless or another signal, and that is uniformly expressed in the posterior compartment at this stage. This too was specifically reduced in the dorsal compartment (Fig. 1F), suggesting that reduction of dynamin activity has a non-specific effect on gene expression, perhaps in response to stress upon reduction of endocytosis. No effect on dll expression was seen after 3 hours at restrictive temperature, suggesting that a transcriptional target is not appropriate to assess the effect of Dynamin loss of function on signalling. We therefore sought to measure Wingless signalling by looking at an earlier event in the signalling cascade. This was accomplished by obtaining an estimate of the rate of Armadillo degradation. Armadillo is stabilized in response to Wingless signalling and this is particularly evident when armadillo is overexpressed(Marygold and Vincent, 2003). In discs expressing armadillo under the control of the engrailed-Gal4 driver, Armadillo accumulation is confined to the cells near the source of Wingless within the pouch(Fig. 1G), even though the RNA is produced uniformly throughout the posterior compartment. If Dynamin activity is inhibited in this context (hemizygous shi[ts];engrailed-Gal4/UAS-armadillo at 32°C for 3 hours), Armadillo accumulates throughout the posterior of the wing pouch(Fig. 1H), but not in the hinge region where Wingless is absent. This is consistent with enhanced signalling activity in response to a block in endocytosis (although we cannot exclude a Wingless-independent effect of shi loss-of-function on Armadillo stability). This suggests that endocytosis (and presumably subsequent trafficking) normally participates in signal downregulation.

Before being targeted to lysosomes, Wingless is most probably internalized by specific receptors. Although Frizzled and Frizzled2 both participate in signal transduction, Frizzled2 is considered to be the main receptor because of its high affinity for Wingless (Wu and Nusse, 2002). Overexpression of Frizzled2 causes Wingless accumulation in wing imaginal discs, not degradation(Cadigan et al., 1998). Indeed,it has been suggested that Frizzled2 could protect Wingless from degradation by titrating out a putative extracellular protease(Eldar et al., 2003). Although the contribution of extracellular proteases cannot be excluded, it is clear that substantial degradation occurs in an endosomal compartment (see Fig. 1). In light of this finding, an alternative possibility is that Frizzled2 could protect Wingless from degradation by blocking Wingless endocytosis, thus preventing subsequent routing to lysosomes. We tested this possibility using imaginal discs and cultured S2 cells.

First, we looked at the relative subcellular distributions of Wingless,Frizzled2 and an endosomal marker in wing imaginal discs. Discs expressing FLAG-tagged Frizzled2 were briefly bathed in fluorescent dextran, which is internalized by fluid phase endocytosis and labels the endocytic compartment(Entchev et al., 2000). Colocalization of dextran, the FLAG tag and endogenous Wingless was assessed in cells that receive (that do not express) Wingless. Many vesicles were found to contain the three markers (57% of dextran positive structures contain Wingless and Frizzled2; Fig. 2A′-A′′′), suggesting that Wingless and Frizzled2 could co-internalize. If Frizzled2 can mediate Wingless internalization, one expects that an endocytosis-defective Frizzled2 would prevent internalization. A transgenic strain expressing the N-terminal extracellular domain of Frizzled2 (which binds Wingless) linked to the membrane by a GPI anchor has been reported (Frizzled2GPI)(Cadigan et al., 1998). Because this construct lacks all intracellular residues, putative endocytic signals are expected to be deleted and indeed, Frizzled2GPI accumulates at the cell surface (Cadigan et al., 1998). Importantly, expression of Frizzled2GPI in imaginal discs strongly reduces Wingless internalization (8.5% of dextran positive structures contain Wingless and Frizzled2GPI; Fig. 2B′-B′′′) (see also Cadigan et al., 1998). Residual internalization could be mediated by a different receptor or by Frizzled2GPI(as GPI-anchored proteins can undergo endocytosis)(Sabharanjak et al., 2002). Strong reduction of endocytosis following expression of Frizzled2GPI is consistent with the notion that Frizzled2 can mediate Wingless internalization.

To assess whether Frizzled2 stimulates Wingless endocytosis, we turned to a cell culture assay. Conditioned medium from GFP-Wingless-expressing cells was applied (at 4°C to prevent endocytosis) to S2R+ cells transfected with Frizzled2. As expected from previous work(Bhanot et al., 1996), the surface of transfected cells becomes decorated with GFP-Wingless(Fig. 2C), confirming that GFP-Wingless binds to Frizzled2. Upon switching the cells to room temperature,which is permissive for endocytosis, GFP-Wingless rapidly accumulates in intracellular Frizzled2-positive vesicles (shown in Fig. 2D-D′). This suggests that Frizzled2 can mediate Wingless internalization. Formation of these vesicles requires endocytosis as it does not occur at 4°C nor in cells that are co-transfected with temperature-sensitive dominant-negative dynamin and kept at the restrictive temperature (not shown). As intracellular accumulation of GFP-Wingless occurs to a significantly higher extent in transfected cells than in untransfected cells (asterisk in Fig. 2D-D′), we also conclude that increasing the level of Frizzled2 leads to increased Wingless endocytosis. By contrast, in agreement with our results in imaginal discs,cells expressing high levels of Frizzled2GPI internalize Wingless poorly(Fig. 2F-F′), even though they bind Wingless very efficiently (Fig. 2E). Overall, these data suggest that Frizzled2 harbours an internalization signal and that, by virtue of its ability to capture Wingless,contributes to the targeting of Wingless into an endocytic compartment. Clearly, then, overexpressed Frizzled2 does not stabilize Wingless by preventing internalization.

Expression of Frizzled2GPI leads to a dramatic extension of the distribution of Wingless: compare the spread of Wingless (double headed arrows) in Fig. 2B with that in a disc expressing full-length Frizzled2 shown in Fig. 2A. We ruled out the possibility that the stronger Wingless stabilization achieved by Frizzled2GPI could be due to a higher expression level (see Fig. S1 in the supplementary material). Therefore, a reduction in endocytosis (caused by Frizzled2GPI) is associated with increased stabilization, consistent with the possibility that Frizzled2-mediated endocytosis could be the first step towards the targeting of Wingless to lysosomes. Yet, the puzzle remains that overexpression of full-length Frizzled2, which does internalize Wingless, stabilizes Wingless to a significant extent (Fig. 2A). One possible resolution of this paradox is that another limiting factor is required for trafficking to lysosomes after endocytosis. One factor that could function with Frizzled2 in Wingless degradation is the single pass receptor encoded by arrow. To assess the role of Arrow, we asked whether it could suppress the stabilizing effect of overexpressed Frizzled2. Indeed,overexpression of Arrow abolishes the stabilization induced by Frizzled2. Strikingly the Wingless range is even shortened in cells that co-express Arrow and Frizzled2 (Fig. 3A,B). Overexpressed Arrow on its own has no noticeable effect on the range of Wingless (not shown). The `degrading' activity of Arrow requires endocytosis since a temporary inhibition of dynamin function (shi[ts] hemizygous at 32°C for 3 hours) in an Arrow and Frizzled2 co-expression experiment is sufficient to restore Wingless accumulation(Fig. 3C). Reduction of Frizzled2-stabilized Wingless could be due to Arrow itself or to a downstream effect of increased signalling caused by co-expression of the receptors. Activating the pathway with an activated form of Armadillo does not bring down Frizzled2-stabilized Wingless. Therefore, activation of target genes is not sufficient to direct Frizzled2-stabilized Wingless to lysosomes. Therefore, we suggest that, by virtue of its ability to associate with the Wingless-Frizzled2 complex, Arrow earmarks the ternary complex for lysosomal targeting.

To ask whether Arrow has indiscriminate Wingless-degrading activity or whether it acts specifically on the Frizzled2-associated Wingless, we assessed its effect on Wingless stabilized by overexpression of the glypican Dally-like(stabilization of Wingless by Dally-like was initially shown by Baeg et al., 2001; see also Franch-Marro et al., 2005). As shown in Fig. 4A,A′,imaginal discs co-overexpressing Arrow and Dally-like continue to accumulate excess Wingless, although at a slightly reduced level. In other words,Dally-like-stabilized Wingless is resistant to Arrow. As another test for specificity, we assessed the effect of Arrow on Wingless stabilized by Frizzled2GPI. We find that Wingless captured by Frizzled2GPI is not degraded by Arrow (Fig. 4B,B′). We conclude therefore that cooperation between Arrow and Frizzled2 is specific and requires parts of Frizzled2 beyond the extracellular cysteine rich domain that binds Wingless (Wu and Nusse,2002). Interestingly, Arrow can bring about the degradation of Wingless stabilized by a form of Frizzled2 that lacks most of the C-terminal tail (Frizzled2ΔC) (Fig. 4C,C′) and, as a result, reduces Wingless internalization(inset in Fig. 4C). Therefore,Frizzled2 residues located between the first and last transmembrane domains are required for the coordinated action of Arrow and Frizzled2 on Wingless degradation. The ability of Arrow to cooperate with a form of Frizzled2 with apparent reduced endocytic ability suggests that Arrow also carries its own internalization signal. In support of this suggestion, we find that overexpression of Arrow in S2R+ cells leads to a significant increase in the internalization of exogenously added GFP-Wingless (data not shown).

A possible interpretation of our results is that Arrow could harbour, in its intracellular domain, a signal that targets the Wingless-Frizzled2 complex to degradation. This signal could be located between residues 1477 and 1612,as overexpression of a truncated form of Arrow that lacks these residues(ArrΔC) on its own causes mild but reproducible accumulation of endogenous Wingless in otherwise wild-type imaginal discs (not shown). Moreover, ArrΔC potentiates the stabilizing effect of Frizzled2. Expression of ArrΔC in the posterior compartment by the en-Gal4 driver,followed by a uniform pulse of Frizzled2 expression throughout the disc,causes an extension of the Wingless gradient in the posterior compartment(Fig. 4D,D′), suggesting stabilization. By contrast, in the same experimental protocol, full-length Arrow causes a reduction of the gradient (not shown), consistent with increased degradation. Unfortunately, ArrΔC is less potent in signalling than wild-type Arrow (weak signalling is restored in arrow mutants overexpressing ArrΔC) and therefore does not allow a clean uncoupling of degradation and signalling. Nevertheless, we conclude that residues 1477 to 1612 contribute to improving the efficiency of Wingless degradation.

The results described in Figs 3 and 4 show that Arrow and Frizzled2 cooperate to bring about the degradation of Wingless, and that Arrow alone is not sufficient as it has no or little effect on Frizzled2GPI- or Dally-like-stabilized Wingless, respectively.

This could be due to the inability of Arrow to bind Wingless in the absence of Frizzled2, as initial attempts failed to show Arrow-Wingless interaction in cell-based binding assays or immunoprecipitation(Wu and Nusse, 2002) [however,also see Cong et al. (Cong et al.,2004)]. Thus, we performed binding experiments in S2 cells expressing comparable amounts of Arrow or Frizzled2. S2 cells expressing Frizzled2 strongly accumulate GFP-Wingless from externally applied conditioned medium (Fig. 5A-A′; Fig. 5C, top panel). By contrast, transfected Arrow has a limited effect on GFP-Wingless accumulation(Fig. 5B-B′; Fig. 5C, bottom panel)suggesting that Arrow can only weakly contribute to capture.

Despite its capturing activity in cells, overexpressed Arrow alone does not cause Wingless accumulation in imaginal discs as Frizzled2 does. This could be because Wingless captured by Arrow would be promptly targeted to degradation. To assess the capturing activity of Arrow in the absence of degradation, Arrow was overexpressed in shi[ts] hemizygotes at 32°C. Mild Wingless accumulation in the domain of Arrow overexpression was seen(Fig. 5D), consistent with the weak capturing activity apparent in cell culture. From these results and from the observation that Arrow overexpression alone does not reduce the endogenous Wingless gradient in otherwise wild-type discs, we conclude that the contribution of Arrow to Wingless capture must be minor.

In principle, Frizzled2 could accompany Wingless all the way to a degradation compartment. Alternatively, it could release Wingless earlier in the endocytic pathway and be recycled to the cell surface. According to the former possibility, Frizzled2 stability would be affected by the presence of Wingless. We decided to monitor receptor endocytosis and decay following a pulse of uniform exogenous expression of tagged protein. This method has two benefits. First, it focuses on post-transcriptional control because the pulse of transcription is uniform. Second, it allows clearance of biosynthetic receptor thus enabling endocytosed receptors to be more readily recognized than in steady state preparations. Transgenic larvae expressing FLAG-tagged Frizzled2 under the control of the heat shock promoter were heat shocked for 35 minutes to induce uniform high-level expression. Levels of FLAG-Frizzled2 and Wingless were then monitored at subsequent time points. Fig. 6A,B show that after a 1.5 hours chase, the number of Frizzled2-containing vesicles is higher around the source of Wingless. These vesicles largely colocalize with Wingless (insets in Fig. 6A,B), suggesting that Wingless stimulates Frizzled2 internalization. After 3 hours, FLAG-Frizzled2 staining becomes depressed preferentially around the normal source of Wingless(Fig. 6D, compared with 6C). To verify that these effects are specific to Frizzled2 and that not all receptors are preferentially trafficked in this region of the disc, an analogous assay was performed with Patched using a heat shock-patched strain(Sampedro and Guerrero, 1991). In contrast to the situation with Frizzled2, Patched is seen to decay rapidly in the posterior compartment where its ligand, Hedgehog, is produced(Fig. 6E,F). Preferential degradation of Frizzled2 near the source of Wingless suggests that Wingless could stimulate Frizzled2 degradation and that Wingless and Frizzled2 could be targeted together to lysosomes. Because the decay of Frizzled2 in the Wingless domain is not as pronounced as that of Patched in the Hedgehog domain, it is likely that additional, non-Wingless-dependent mechanisms of degradation are at play or that the Wingless-dependent mechanism is inefficient. Overall, our data suggest a model whereby Arrow and Frizzled2 contribute different, though overlapping, trafficking activities that, together, lead to targeting of Wingless to a degradation compartment (Fig. 6G,H). Frizzled2 functions predominantly in capture, while Arrow(and possibly downstream signalling) would be essential for targeting of internalized Wingless to a degradative compartment.

Because both receptor classes contribute to degradation, we expect the loss of either Arrow or of Frizzed and Frizzled2 to cause Wingless accumulation in vivo. Accumulation of extracellular Wingless either in arrow mutant clones or in frizzled frizzled2 mutant clones has indeed been reported recently (Baeg, et al.,2004; Han et al.,2005). Using various methods to generate mutant clones, we have confirmed this and found that such accumulation is not solely due to ectopic transcription because accumulation occurs up to 10 cell diameters away from the source (see Fig. S2A in the supplementary material) while ectopic transcription is confined to three cell diameters (see Fig. S2B,B′ in the supplementary material) (Rulifson et al., 1996). Accumulation has been suggested to be a consequence of Dlp upregulation in the mutant cells (arrow or frizzled frizzled2) (Han et al.,2005). We suggest that Wingless accumulation in the absence of the signalling receptors results from both the upregulation of Dlp and the removal of degrading receptors. In wild-type cells, Wingless would have access to both receptor systems and thus be allocated appropriately to degradation and transport (Franch-Marro et al.,2005).

Dextran labelling of imaginal discs shows that Wingless is still internalized by arrow mutant cells (see Fig. S3 in the supplementary material), suggesting that Frizzled and Frizzled2 may not require Arrow for internalization. Indeed, S2R+ cells pretreated with Arrow-RNAi and transfected with Frizzled2 internalize externally applied GFP-Wingless (see Fig. S3B,B′ in the supplementary material). We conclude that Frizzled2 can capture and internalize Wingless independently of Arrow and hence also of signalling. Conversely, clones of cells lacking Frizzled Frizzled2 have been shown to internalize Wingless (Baeg et al.,2004), suggesting that these receptors are also dispensable for Wingless internalization (although the frizzled [H51] allele used in this study encodes a receptor that is truncated only after the sixth transmembrane domain and thus retains both the Wingless-binding domain and several intracellular residues). In order to compare the trafficking effects of removing Arrow on the one hand with those of removing Frizzled Frizzled2 on the other hand, we developed a method that specifically reveals intracellular Wingless (and used alleles that are not expected to produce plasma membrane-tethered receptors). The same preparations were also stained using a`post-fixation extracellular only protocol' adapted from Strigini and Cohen(see details in the Materials and methods). In both types of mutant clones,intracellular staining is detected, confirming that Wingless internalization can occur in the absence of these two receptor classes. Interestingly, large Wingless-containing vesicles (arrowhead, Fig. 7A′) are relatively depleted in frizzled frizzled2 mutant cells (when compared with neighbouring wild-type tissue; Fig. 7A′) although fine-grained staining is still seen. By contrast, arrow mutant cells contain such vesicles at the same density as wild-type cells (Fig. 7B′). Therefore, frizzled frizzled2 mutant cells appear to be deficient in a trafficking step required to deliver or maintain Wingless in these vesicles.

The main conclusion of our work is that two receptors contribute distinct though overlapping trafficking activities that, together, lead to degradation of Wingless (see model in Fig. 6). Our binding data support the earlier suggestion that normally Wingless is primarily captured by a Frizzled family member and that this facilitates subsequent binding to Arrow(Cong et al., 2004; Wu and Nusse, 2002). As we have shown, Wingless is internalized by Frizzled2 in the absence of Arrow. This result extends and complements recent evidence that mammalian Frizzled4 is endocytosed upon stimulation by Wnt5a(Chen et al., 2003). Moreover,Wingless internalization in the absence of Arrow also shows that Wingless signalling is not required for endocytosis. However, in the absence of further targeting to a lysosomal compartment, endocytosis would clearly be insufficient for degradation.

Using gain-of-function experiments, we showed that Arrow contributes to the targeting of Wingless, maybe as a complex with Frizzled2, to a degradative compartment. As expected, loss of either Arrow or Frizzled and Frizzled2 leads to extracellular accumulation of Wingless(Baeg et al., 2004; Han et al., 2005) (our own observations). Frizzled and Frizzled2 are clearly redundant in this respect(as in signalling) because removal of either receptor has no noticeable effect on Wingless distribution (Han et al.,2005). Interestingly, large intracellular vesicles are lost in the absence of Frizzled D-Frizzled 2 but not in the absence of Arrow. We suggest that Frizzled-mediated endocytosis is sufficient to generate these large vesicles in the absence of Arrow. The fine-grained Wingless staining seen in the absence of Frizzled D-Frizzled 2 could be internalized by Arrow or by another receptor, such as Dally or Dally-like. The distinct intracellular distribution of Wingless in the absence of Frizzled D-Frizzled 2 when compared with that in Arrow-deficient cells is consistent with the suggestion that the two receptor classes have distinct trafficking activities.

It is unclear at this point whether the degrading activity of Arrow is regulated by post-translational modification or by the recruitment of other factors. Either process could be impaired in ArrowΔC. Work in Xenopus has identified negative regulators of Wnt signalling,Kremens, which operate by triggering LRP6 endocytosis and possibly degradation(Mao et al., 2002). It remains to be seen whether this leads to degradation of a Wnt during frog embryogenesis. Moreover, there is no Kremen homologue encoded by the fly genome. Clearly further work will be needed to understand the genetic control of Wnt/Wingless degradation both in flies and other systems. Our data provide a simple explanation of why overexpression of Frizzled2, a receptor that mediates Wingless internalization, causes Wingless stabilization. Under such experimental conditions, Arrow becomes limiting and in the absence of an effective degradation signal, Wingless accumulates.

Comparison of Fig. 1C with Fig. S2 in the supplementary material shows that Wingless accumulates to a lesser extent in arrow mutant cells than in deep-orangemutant cells. Therefore, it is likely that additional Wingless degradation pathways could operate, possibly mediated by different receptors. Likewise,although Frizzled2 is preferentially degraded in cells that are within the range of Wingless, this is superimposed on a general degradation mechanism as Frizzled2 eventually decays throughout the disc following a pulse of uniform expression (not shown).

Because the receptors involved in Wingless degradation are those required for signalling, Wingless degradation cannot be initiated before a signalling-competent complex is assembled. Even though signalling downstream of Armadillo is not sufficient to activate the degradation of Frizzled2-Wingless complexes (Fig. 3D), we do not know yet whether downstream signalling is necessary for degradation. In the case of EGF receptor signalling, ubiquitination, the first step towards degradation of the ligand, is contingent on the tyrosine phosphorylation that accompanies receptor activation(Shtiegman and Yarden, 2003). However, in this case, a single receptor type is involved. In the case of TGFβ signalling, two receptor types are required for signal transduction. Type 2 receptor is believed to capture the ligand and this is followed by the formation of a tripartite complex with type 1 receptor(Massague, 1998). Interestingly, like Arrow, type 1 receptor brings a degradation signal such that the two types of receptor cooperate to direct the ligand towards degradation and signalling pathways appropriately(Anders et al., 1998; Di Guglielmo et al., 2003; Ebisawa et al., 2001; Kavsak et al., 2000). Sharing of trafficking duties by distinct receptors may provide cells with increased flexibility as expression or turnover of the two receptors could be independently modulated. It may not be a coincidence that both Dpp (the fly TGF-β) and Wingless, which can act over a relatively long distance, use two receptors for signalling and degradation. Maybe separation of capture and degradation is a feature required for long-range signalling, perhaps by allowing modulation of local relative receptor levels.

Further work will be needed to identify the relevant trafficking signals in Arrow and Frizzled2, as well as the mechanisms that control relative receptor levels in order to obtain a full understanding of how degradation of Wingless is tuned to generate a reliable concentration gradient.

Eugenia Piddini was supported by postdoctoral fellowships from the EU and from EMBO. All others authors are supported by the Medical Research Council(UK). We thank Cyrille Alexandre for generating TOPFLASH for expression in S2 cells, Oriane Marchand for help with arrow RNAi, Iris Salecker for help with confocal imaging and Rafael Carazo-Salas for help with Image Analysis. We also thank colleagues (listed in the Materials and methods), the Bloomington Stock Center and the Developmental Studies Hybridoma Bank for reagents. Cyrille Alexandre, James Briscoe, Rafael Carazo-Salas and Bruno Glise provided helpful comments on the manuscript.