Glypicans shunt the Wingless signal between local signalling and further transport

Extracellular ligands are typically degraded after engagement with their cognate receptors and activation of the signalling pathway. Degradation is usually initiated by receptor-mediated endocytosis. This raises a fundamental problem for ligands that must act over long distances within epithelia. High-affinity signalling receptors are expected to trap the ligand near the source and thus prevent transport to more distant cells(Chen and Struhl, 1996; Kerszberg and Wolpert, 1998). How is a locally produced ligand appropriately allocated to all the cells within a field, near and far? One possibility is that specialised,non-signalling receptors ensure long-range signalling. In wing imaginal discs of Drosophila, the secreted glycoprotein encoded by winglessis produced at the prospective wing margin and spread symmetrically along the dorsoventral axis (Zecca et al.,1996; Cadigan et al.,1998). One class of receptors that could modulate the distribution of Wingless are heparan sulphate proteoglycans (HSPGs). In particular,overexpression of the glypican Dally-like (Dlp) causes accumulation of Wingless at the surface of imaginal disc cells(Baeg et al., 2001). Somehow,such accumulation is accompanied by a reduction in Wingless signalling. Although this gain-of-function experiment suggests that Dally-like could regulate the distribution of Wingless, so far, this has not been confirmed by loss-of-function analysis. Indeed, the normal role of Dlp in Wingless function is still unknown.

Good evidence for the involvement of HSPGs in Wingless function comes from the isolation and analysis of mutants in genes encoding HSPG modifying enzymes. Drosophila embryos lacking sugarless, which encodes UDP-Glucose-Dehydrogenase, an enzyme required for heparan sulphate biosynthesis, are deficient in Wingless signalling(Binari et al., 1997; Hacker et al., 1997; Haerry et al., 1997). Likewise, embryos lacking sulfateless, which encodes heparan sulphate N-deacetylase/N-sulphotransferase (an enzyme needed for the modification of heparan sulphate) also resemble wingless mutants(Lin and Perrimon, 1999). Further genetic analysis showed that HSPG metabolising enzymes are also implicated in additional signalling pathways such as those activated by Hedgehog, FGF and Dpp (Bellaiche et al.,1998; Lin et al.,1999; Han et al.,2004a; Takei et al.,2004). To what extent HSPGs play distinct roles in the different signalling pathways remains to be determined. Practically, the involvement of HSPGs in both Wingless and Hedgehog signalling complicates genetic analysis in embryos because the terminal phenotype of wingless and hedgehog mutants look very similar to each other, making it difficult to distinguish roles in either or both signalling pathways. In imaginal discs,Wingless and Hedgehog signalling are easily distinguished experimentally and,in this tissue, it is clear that loss of sulfateless causes a reduction in Wingless signalling (Baeg et al., 2001). In particular, Wingless protein that normally accumulates at the surface of wild type Wingless-expressing cells does not do so at the surface of sulfateless mutant cells, suggesting a potential role of HSPGs in tethering Wingless at the cell surface.

Genetic analysis of HSPG modifying enzymes clearly implicates HSPGs in Wingless function. What are the relevant HSPGs involved? Four HSPGs are predicted to be encoded by the Drosophila genome(Nybakken and Perrimon, 2002). Of these, as mentioned above, Dlp is a good candidate on the basis of gain-of-function studies. However, RNAi injection in embryos has failed to confirm a role for Dlp in Wingless signalling. Instead, a clear role in Hedgehog signalling was demonstrated by RNAi(Desbordes and Sanson, 2003; Lum et al., 2003) and this was later confirmed with traditional genetic mutants(Han et al., 2004b). In these experiments, a possible role of Dally-like in Wingless signalling might have been masked by redundant activity from another glypican. Indeed, the other glypican found in Drosophila, Dally, has been implicated in Wingless function on the basis of RNAi-induced phenotypes in embryos(Lin and Perrimon, 1999)although this has been questioned by subsequent experiments(Desbordes and Sanson, 2003). Nevertheless, weak wingless-like phenotypes are seen in presumed weak allele of dally (Lin and Perrimon, 1999; Tsuda et al.,1999).

Overall, various experiments involving presumed weak alleles, RNAi injection or overexpression have implicated Dally and Dlp in Wingless function(Nakato et al., 1995; Lin and Perrimon, 1999; Tsuda et al., 1999; Baeg et al., 2001). However, a clear assessment of their role has been hampered by the involvement of HSPGs in other signalling pathways and also by the lack of suitable null mutants(Sanson, 2004). We have created null mutants in both dally and dlp. The analysis of single and double mutant embryos shows that both glypicans participate in normal Wingless function. Using a combination of gain- and loss-of-function approaches in imaginal discs, we have investigated the specific roles of the two glypicans in regulating the activity and distribution of the Wingless signal in wing imaginal discs.

To create an N-terminal FLAG tagged version of Dally, a double-stranded oligonucleotide containing three copies of the FLAG epitope was inserted in the unique AatII site (position 150 in Dally). The Dally-FLAG cDNA was subsequently inserted in pUAST as an EcoRI-XbaI fragment. DNA carrying UAS-Dally-HA and UAS-Dlp-HA as well as pMT-Gal4 used for the binding experiments with S2 cells were obtained from S. Cohen, EMBL. See Fig. 1B for a diagram of the constructs.

The following transgenic stocks were used for exogenous expression:UAS-Dally-FLAG (generated for this study), UAS-Dally-like-HA (from S. Cohen,EMBL). Both were driven with dpp-Gal4, engrailed-Gal4 or tubulin-Gal4. For loss-of-function studies, we generated mutants in dally and dlp by imprecise excision of a P-element inserted in the 5′ region (see Fig. 1A). For dlp, we obtained a deletion of nearly 500 nucleotides, which removes the 5′UTR and 16 nucleotides downstream of the ATG. For dally, we obtained a deletion of about 1.8 kb that removes the first exon. Both mutants are expected to produce no functional protein. Clones of mutant cells were induced by Flp-mediated mitotic recombination. The following additional stocks were used: GSV6 9608 and GSV6 11297 (P-insertions in dlp and dally, respectively, obtained from Toshiro Aigaki, Tokyo University); Df(3L) ED4413, a large deficiency that uncovers dally (generated by us with the Drosdel kit, www.drosdel.org.uk);and Df(3L) ED4543 (obtained from Drosdel), which uncovers dlp. w; FRT2A dlp^MH20, w; FRT2A dally^MH32 and w; FRT2A dally^MH32dlp^MH20 were generated for this study. We also used yw hsflp; FRT2A ubi-GFP (from the Bloomington Stock Center), wg^CX4 and hh^AC(described at flybase.bio.indiana.edu),UAS-GFP-Wingless, and sim-gal4 (from Benny Shilo, Weizmann Institute). Germline clones were induced with yw; FRT2A ovo^D (from the Bloomington Stock Center). To induce overexpression clones, the following stocks were used: yw hsflp; tubulin>>GAL4>>UAS lacZ, yw; FRT42D pwn and yw hsflp;FRT42D Gal-80; tubulin-Gal-4 (from S. Cohen, EMBL).

To visualise cuticle patterns, 24-hour-old embryos were mounted in Hoyer's and photographed under DIC or phase-contrast microscopy. In situ hybridisation was performed on fixed embryos hybridised with a digoxigenin labelled rhomboid probe (a gift from J. F. de Celis, Cambridge University).

Primary antibodies used were mouse anti-Wingless 4D4 (prepared from cells obtained from the DSHB), mouse M2 anti-Flag (Sigma; 1/1000), mouse anti-HA 1.1(Covance; 1/500), Alexa-488 labelled mouse anti-HA 1.1 (Covance; 1/500),rabbit anti-GFP (Abcam; 1/2500), mouse anti Distal-less (from I. Duncan,University of Wisconsin; 1/500) and guinea pig anti-Senseless (from Hugo Bellen, Baylor College of Medicine; 1/1000). Secondary antibodies (all from Molecular Probes) were Alexa488-conjugated goat anti-rabbit (1/200),Alexa488-conjugated goat anti-mouse (1/200), Alexa594-conjugated goat anti-mouse (1/200) and Alexa488-conjugated goat anti-guinea pig (1/200).

GFP-Wingless medium prepared from a stable line (S2-GFP-Wg) was applied to S2 cells grown on coverslips and transiently transfected with dallyor dlp constructs (HA-tagged) and induced overnight with 0.25 mM CuSO₄. GFP-Wingless binding was performed on ice as previously described (Bhanot et al.,1996). After fixation, GFP-Wingless was detected with an anti-GFP antibody and the HA tagged receptors with anti-HA.

In order to assess the function of Dally and Dlp, we have created, for each gene, a small deletion that removes the translation start and the signal peptide. These alleles are henceforth referred to as dally^MH32 and dlp^MH20 (see molecular description in Fig. 1). For both alleles, even if an opportunistic translation start were used, no signal peptide would be encoded. For dlp, the next possible ATG is 123 residues downstream of the normal translation start, while for dallythe mutation removes the first exon thus deleting at least the first 90 residues. Therefore, in both cases, any protein product would not be allowed to enter the secretory pathway and reach the cell surface even if a downstream translation start is used (i.e. no functional protein could be produced). On the basis of this evidence, we consider dlp^MH20 and dally^MH32 to be null. Previous dally alleles,including the strongest one, dally^P2 were thought to be hypomorphic. However, we find that the penetrance and expressivity of adult phenotypes is similar in dally^MH32 homozygotes as in dally^P2 homozygotes (as well as in dally^MH32/Df(3L) ED4413 animals). It appears therefore that dally^P2 is also a null allele.

As a first assessment of the role of these two proteoglycans in Wingless function, we analysed embryos that lack the maternal and zygotic contribution of either dally, dlp, or both. Embryos lacking dally appear normal, suggesting that this gene is not essential for embryonic development(data not shown) (see also Desbordes and Sanson, 2003). By contrast, and as previously shown, Dlp-deficient embryos die with a so-called denticle lawn phenotype, which has been attributed to a failure in transduction of the Hedgehog signal [early targets of Hedgehog signalling fail to be activated(Desbordes and Sanson, 2003; Lum et al., 2003)]. Assessing a possible, additional role in Wingless signalling has been difficult because the terminal phenotype of embryos lacking Hedgehog signalling is similar to that of wingless-deficient embryos. To investigate this, we used an epidermal molecular marker that differentiates between hedgehog,wingless and wingless hedgehog mutant embryos. As can be seen in Fig. 2A-C, the pattern of rhomboid expression is distinct in all three genetic conditions (see also Alexandre et al., 1999). In embryos lacking maternal and zygotic dlp, the pattern of rhomboid expression is variable with features of both hedgehog and wingless mutants, but falling short of phenocopying wingless hedgehog double mutants (not shown). This suggests a loss of Wingless activity, although an incomplete one, perhaps because of a redundant contribution from Dally. We therefore assessed the pattern of rhomboid expression in embryos lacking maternal and zygotic dally and dlp (dally dlp double mutants for short). As shown in Fig. 2D, it strongly resembles that seen in wingless hedgehog double mutants. This similarity is confirmed by close inspection of terminal cuticular phenotypes. The cuticle phenotype of dally dlp double mutants is more severe than that of embryos lacking maternal and zygotic dlp (not shown) and, importantly, it is characterised by a feature seen in wingless hedgehog double mutants, the presence of mid-ventral denticles whorls(Fig. 2E,F). We conclude that both glypicans participate in the normal activity of Wingless and Hedgehog.

Because dally dlp double mutant embryos are deficient in Hedgehog signal transduction, they express only a small amount of winglessduring a short period [wingless transcription is maintained by Hedgehog signalling (Lee et al.,1992)]. We therefore assayed the effect of artificially sustained Wingless expression in the absence of Dally and Dlp. As shown in Fig. 3, expression of UAS-GFP-Wingless under the control of sim-gal4 leads to local activation of Wingless signalling in dally dlp double mutants,as evidenced by the formation of naked cuticle in the mid ventral region. Therefore, dally and dlp are not absolutely essential for Wingless signal transduction. Their roles in the embryo could be to boost or sustain the signal, perhaps by ensuring sufficient retention/accumulation of Wingless at cellular interfaces. In any case, the role of Dlp in Wingless function seems to predominate because embryonic Wingless signalling appears normal in the absence of Dally. However, Dally must provide some activity that contributes to normal Wingless function because the phenotype of dally dlp mutants is stronger than that of dlp single mutants.

To further investigate the roles of Dally and Dlp in Wingless function, we analysed the phenotypes of adult mutants. As mentioned above, embryos lacking all dally function are normal. In fact, they give rise to viable adults. These have recognisable phenotypes in a variety of tissues, as reported previously for presumed weaker alleles. This includes reduced and rough eyes, disrupted antennae, loss of bristles and, in most males(occasionally in females), loss of external genital structures(Nakato et al., 1995). We concentrate on the wing because it allows a relatively simple assessment of Wingless function. In wing imaginal discs, dally transcription is upregulated near the Wingless source(Nakato et al., 1995),suggesting a role at the prospective margin. As shown in Fig. 4B, dally^MH32 homozygotes display notches in the wing margin at a low frequency (5%) (see also Nakato et al., 1995). Interestingly, the frequency and severity of these notches is significantly higher (41%) in adults arising from maternal null embryos (in an otherwise identical genetic background), suggesting that the maternal contribution of dally is long lasting. Because loss of margin tissue is indicative of insufficient Wingless signalling(Phillips and Whittle, 1993),we conclude that wild type Dally contributes positively to Wingless function.

Contrary to a previous report (Han et al., 2004b), we find that dlp is also not zygotically required for adult viability: dlp^MH20 homozygous flies(with wild-type maternal contribution) survive to adulthood, albeit at a low frequency. Therefore, the maternal contribution of dlp can support development to adulthood, thus enabling us to assess the phenotype in wings. Here, in situ hybridisation shows that dlp is uniformly expressed except at the Wingless source where is it downregulated (not shown). Surprisingly, dlp^MH20 wings display a weak but fully penetrant phenotype that is characteristic of apparent excess Wingless signalling. During normal wing development, specialised bristles form at the prospective wing margin, where Wingless signalling is highest during imaginal disc development (Couso et al.,1994). In the wings of surviving dlp^MH20homozygotes, supernumerary bristles form on either side of the margin(Fig. 4D), an indication of ectopic Wingless signalling (Couso et al.,1994). Therefore, loss of dlp activity leads to excess Wingless signalling, at least around the wing margin. It is intriguing that mutations in dally and dlp lead to qualitatively different phenotypes with respect to Wingless signalling at the wing margin, while these two genes appear to be redundant for viability (homozygous double mutants die at the end of embryogenesis, even in the presence of maternal contribution). One possible explanation is that the two glypicans could perform distinct cell biological functions that together contribute to optimal Wingless activity(see below).

In order to further investigate the specific roles of Dally and Dlp in Wingless activity, we turned to overexpression experiments. When Dally is overexpressed at 25°C with the dpp-gal4 driver, which is active in a broad group of cells along the anteroposterior boundary in the wing, no apparent effect on wing margin morphology can be seen (see also Strigini and Cohen, 2000). By contrast, overexpressed Dlp induces clear-cut phenotypes. Overexpression of Dlp in broad domains with engrailed-gal4 or dpp-gal4 at 25°C mainly causes loss of margin tissue, i.e. a reduction of Wingless signalling there (Fig. 5A,B)(see also Baeg et al., 2001). Likewise, scattered clones of misexpressing cells induced with the `Flp-on'system (Pignoni and Zipursky,1997) cause margin loss (not shown). Oddly, however, an opposite phenotype is also seen. As shown in Fig. 5C, the presence of random clones of cells that overexpress Dlp causes the formation of ectopic margin bristles. Under all these experimental conditions, Dlp overexpression induces a strong accumulation of Wingless at the cell surface (only shown here for overexpression in clones, in Fig. 5E,F) (see also Baeg et al., 2001). The induction of opposite phenotypes (loss and gain of Wingless signalling) by overexpression could be explained if Dlp had distinct cell-autonomous and non-cell-autonomous effects. To investigate this further, mitotic clones of Dlp-misexpressing cells were marked with the pwn mutation, which affects the morphology of hairs and bristles(Lawrence et al., 2002). As shown in Fig. 5D, ectopic margin bristles are formed by wild-type cells located next to misexpressing cells. This suggests that overexpressed Dlp causes non-cell-autonomous increase in Wingless signalling. At the same time, it appears that Dlp overexpression causes cell-autonomous reduction of Wingless signalling as overexpression in a broad band of cells leads to loss of margin tissue.

The results described above lead to a model whereby, at the surface of a given cell, Dlp would divert oncoming Wingless away from local signalling while at the same time favouring export or presentation to neighbouring cells(Fig. 6). According to such a model, overexpressed Dlp would be expected to extend the range towards more distant cells while at the same time reducing signalling near the source. This prediction is indeed borne out. Within cells that overexpress Dlp, senseless, a gene that responds to high level of Wingless(Nolo et al., 2000), is repressed as expected from the loss of margin tissue. For example, in discs overexpressing Dlp under the control of engrailed-gal4, senselessexpression is abolished in the posterior compartment (right side of the broken line in Fig. 7A). At the same time, the domain of expression of distal-less, a low level target(Zecca et al., 1996) is broadened (Fig. 7B), suggesting an extension of the range. A similar result is seen in large overexpression clones (Fig. 7C,D). In this figure, the slight upregulation of distal-less in cells flanking the misexpressing clone (arrowhead), consistent with the possibility that overexpressed Dlp favours presentation of Wingless to neighbouring cells.

If wild-type Dlp is indeed involved in shunting Wingless to neighbouring cells, one expects that loss of dlp activity would reduce the range of Wingless. This is indeed the case. Clones of dlp mutant cells contain reduced amount of Wingless in receiving tissue (arrowhead in Fig. 7F) without an apparent reduction at the source. A reduction of Wingless present at the surface of dlp mutant cells is also seen(Fig. 7G,H). To assess the functional consequence of reduced Wingless, we looked at the expression of distal-less in imaginal discs obtained from homozygous dlp^MH20 larvae. The domain of distal-lessexpression in dlp mutants is significantly narrower than that in wild-type discs, suggesting a reduction of the range(Fig. 7I,J). A concomitant increase in senseless expression is seen(Fig. 7K,L) suggesting increased signalling near the source (as expected from ectopic margin bristles, Fig. 4D). Thus, in dlp mutants, the balance between short-range and long-range signalling is upset, with short-range signalling being favoured at the expense of long-range signalling.

Because dally mutants have reduced Wingless signalling, Dally could act as a classical, though non-essential, co-receptor(Ruoslahti and Yamaguchi,1991), perhaps capturing the ligand and presenting it to the signalling receptors. To determine whether Dally does indeed promote binding of Wingless to the cell surface, we used a tissue culture assay. S2 cells were transfected with a plasmid encoding Dally. After a suitable time to allow expression, cells were transferred to 4°C to prevent endocytosis and treated with conditioned medium containing GFP-Wingless. As shown in Fig. 8A,B, Dally expression causes a significant increase in the accumulation of GFP-Wingless at the surface of S2 cells. Transfection of dlp also causes Wingless accumulation at the cell surface, possibly to a lesser extent(Fig. 8C,D). We suggest that in imaginal discs, after recruitment, Dally could present Wingless to signalling receptors expressed in the same cell. This would be followed by activation of signalling and rapid degradation. In imaginal discs, although exogenous Dlp accumulates almost exclusively at the cell surface, Dally is present both at the cell surface and in vesicles, perhaps because it is continuously endocytosed (insets in Fig. 8E).

The fact that mutations in dally and dlp cause different phenotypes suggests that, although they both underpin Wingless function, these two glypicans could perform distinct activities. It is likely that both Dally and Dlp are able to capture Wingless at the surface of imaginal disc cells. From the point of view of a given cell in vivo, Wingless captured by Dally would be mostly destined for `internal consumption', while Dlp-bound Wingless would be for export only. Subsequent long-range transport would occur by hopping from Dlp on one cell to Dlp on the next. Both glypicans would contribute to increasing the concentration of Wingless at the cell surface(Dally in cis and Dlp in trans). We suggest that in the embryo too, Dlp and Dally help in the presentation and reception of Wingless, respectively. However, in this system, little Wingless transport takes place(Pfeiffer et al., 2000), maybe because release of Wingless from Dlp is not allowed. It is interesting that,in embryos, dlp is highly expressed in cells that secrete Wingless. Therefore, the role of Dlp would mainly be to ensure that plenty of Wingless is retained at the surface of Wingless-expressing cells thus allowing sustained short-range signalling. In both the embryonic and disc systems, the genetic redundancy between dally and dlp could be viewed as follows: reduction of capturing activity in dally mutants would be compensated by the `presentation activity' of Dlp and vice versa. Further cell biological work will be needed to fully explore the specific activities of Dally and Dlp and also to discover how Wingless is transferred from one cell to another during transport, perhaps with the help of enzymes such as Notum/Wingful (Gerlitz and Basler,2002; Giraldez et al.,2002).

Xavier Franch-Marro is the recipient of a Wellcome Trust travelling fellowship, Oriane Marchand and Eugenia Piddini are funded by EMBO and Marie Curie fellowships, Sara Ricardo was funded by a Gulbenkian predoctoral fellowship, while the others are supported by the Medical Research Council. We acknowledge the Developmental Studies Hybridoma Bank, the Bloomington Stock Center and various colleagues (listed in the Materials and methods) for strains and antibodies. Special thanks to Toshiro Aigaki for the P-element stocks used to generate the mutants, to Steve Cohen for the generous provision of unpublished plasmids and strains, to Jose Casal for suggesting the pwn experiment (Fig. 4D) and providing the necessary materials, and to John Roote and Drosdel for stocks. David Wilkinson provided comments on the manuscript.