Hedgehog lipid modifications are required for Hedgehog stabilization in the extracellular matrix

The proteins of the Hedgehog (Hh) family are powerful signaling molecules that act as morphogens, controlling growth and patterning in certain tissues of a range of different organisms at different developmental stages(Ingham and McMahon, 2001). Although the Hh signal molecule was first described as a morphogen in the segmental patterning of the Drosophila larva cuticle(Heemskerk and DiNardo, 1994),the Hh morphogen gradient across the wing imaginal disc has provided a unique Hh signaling read out (reviewed by Torroja et al., 2005). In mammals, the function of Sonic Hedgehog (Shh) -one of the three members of the Hh protein family (Shh, Indian Hh and Desert Hh) - as a morphogen has been thoroughly characterized(Echelard et al., 1993)(reviewed by McMahon et al.,2003). Shh has been shown to act as a morphogen in patterning the ventral neural tube (Briscoe et al.,2001) (reviewed by McMahon et al., 2003) and in specifying vertebrate digit identities(Ahn and Joyner, 2004; Harfe et al., 2004). Recent works have also revealed an unexpected role of Shh in the control of axon guidance in the developing nervous system(Charron et al., 2003)(reviewed by Charron and Tessier-Lavigne,2005).

Establishing the molecular mechanisms that generate the Hh gradient is essential for our understanding of how the Hh signal elicits multiple responses in a temporally and spatially specific manner. The Hh spreading mechanism is especially intriguing, because Hh is a lipid-modified molecule. Normally, lipid-modified peptides appear firmly tethered to membranes(Peters et al., 2004). Both in vertebrates and in Drosophila, Hh is synthesized as a precursor protein that undergoes a series of post-translational modifications within the secretory pathway that lead to the presentation at the cell surface of the mature and signaling-active Hh (reviewed by Mann and Beachy, 2004). Following cleavage of a N-terminal signal peptide upon entering the secretory pathway, the Hh protein undergoes an autocatalytic processing reaction that involves internal cleavage between Gly-Cys-specific residues(Lee et al., 1994; Tabata and Kornberg, 1994;Bumcrot, 1995). A cholesteryl adduct then covalently binds to the N-terminal product of this cleavage (Lee et al.,1994; Tabata and Kornberg,1994; Bumcrot, 1995; Marti et al., 1995; Porter et al.,1996a; Porter et al.,1996b). The C-terminal domain of the Hh precursor mediates this autoprocessing reaction between Gly-Cys-specific residues(Lee et al., 1994; Tabata and Kornberg, 1994;Bumcrot, 1995). The second lipid adduct that modifies the Hh protein is palmitic acid, which attaches to the N-terminal cysteine exposed after signal peptide cleavage (Pepinsky et al.,1998). This acylation is catalyzed by the product of the sightless gene, also designated skinny hedgehog, central missing or raspberry (Amanai and Jiang, 2001; Chamoun et al., 2001; Lee and Treisman,2001; Micchelli et al.,2002). The doubly lipid-modified Hh is the fully active signaling molecule (Lee et al., 2001). Although much of the Hh maturation process has been determined in Drosophila, all the metazoan species examined so far show the same biochemical and functional mechanisms.

The importance of lipid modifications has been demonstrated by evaluating the diffusion and signaling properties of different forms of Hh that lack lipid modifications in several animal models. These forms include HhN or ShhN(without the cholesterol moiety) and Hhc85s or Shhc25s (lacking the palmitate modification). Although the loss of cholesterol reduces the signaling capabilities in vertebrate models (Porter et al., 1996a; Lewis et al.,2001; Zeng et al.,2001), minor effects have been reported in Drosophilawing disc (Burke et al.,1999). However, a detailed analysis in the embryo revealed an essential role for the activation of a specific subset of Hh targets in the embryo (Gallet et al., 2003). The lack of N-terminal acylation dramatically reduces signaling potential,both in Drosophila and vertebrates(Chamoun et al., 2001; Kohtz et al., 2001; Lee et al., 2001; Lewis and Eisen, 2001; Gallet et al., 2003; Chen et al., 2004; Tian et al., 2005). The diffusion of Drosophila HhN, however, differs from that of their vertebrate counterparts. Although cholesterol-free Hh is able to diffuse longer distances in Drosophila wing imaginal disc(Burke et al., 1999), but not in Drosophila embryos (Gallet et al., 2003), both forms of lipid-unmodified Shh show reduced activity and diffusion properties as measured by analyzing target gene activation and protein distribution (Lewis et al., 2001; Chen et al.,2004; Tian et al.,2005). These differences between Drosophila and vertebrates, and also in Drosophila between the wing imaginal disc and the embryonic epidermis are not well understood, though the effects of lipids on Hh have not yet been systematically explored in Drosophilawing imaginal disc cells. The molecular mechanisms through which a lipid-modified Hh protein is able to diffuse long distances are also unclear.

Extracellular matrix proteins such as heparan sulfate proteoglycans (HSPGs)have been attributed a regulatory role in the signaling activity of secreted morphogen molecules. Thus, the Drosophila EXT family of proteins,encoded by the genes tout velu (ttv), brother of tout velu and sister of tout velu, which are essential for the synthesis of HSPGs, are required for the diffusion of lipid-modified Hh(Bellaiche et al., 1998; Bornemann et al., 2004; Takei et al., 2004), yet do not affect the spread of cholesterol-free Hh(Bellaiche et al., 1998; Gallet et al., 2003). EXT proteins are glycosyl transferases that catalyze the formation of heparan sulfate glycosylaminoglycan chains, which are attached to a core protein. Recently, the glypican (a type of HSPG attached to the cell membrane by a GPI-anchor) proteins Dally and Dally-like, were found to be required for Hh diffusion (Han et al., 2004),and Dally-like was also shown to be needed for the reception of Hh in cultured cells and embryos of Drosophila(Desbordes and Sanson, 2003; Lum et al., 2003). These data indicate that HSPGs are important for the formation of morphogen gradients and Hh signal reception.

In this work, we analyze the role of Hh lipid modifications (cholesterol and palmitic acid) in spreading, internalization and signaling during Hh gradient formation in the wing imaginal disc. We also examine the function of extracellular matrix HSPGs in the spreading and signaling of lipid-modified and unmodified forms of Hh. Our results indicate that Hh lipid modifications are essential for Hh/HSPG interaction. This interaction retains and stabilizes Hh within the epithelium to control spreading and proper signaling of Hh. We propose a conserved role of lipids in Hh signaling in Drosophila and vertebrates.

A description of the mutations, insertions and transgenes in FlyBase is available at http://gin.ebi.ac.uk:7081.

The mutants used were: dor⁸, a null allele(Shestopal et al., 1997); shi^ts1 (Grigliatti et al., 1973) and hh^ts2(Ma et al., 1993), two temperature-sensitive alleles with restrictive temperatures at 32°C and 29°C, respectively; and ttv^l(2)00681(Bellaiche et al., 1998).

The reporter genes used were dpp-LacZ^{BS 3.0}(Blackman et al., 1991), dpp¹⁰⁶³⁸ (Zecca et al., 1995) and ptc-lacZ (a gift from C. Goodman).

The following Gal4 drivers were used in the ectopic expression experiments using the Gal4/UAS system(Brand and Perrimon, 1993): hh-Gal4 (Tanimoto et al.,2000), AB1-Gal4(Munro and Freeman, 2000), ap-Gal4 (Calleja et al.,1996) and Ubx-Gal4(Pallavi and Shashidhara,2003).

The hh transgenes were: UAS-hh-GFP(Torroja et al., 2004), UAS-hhN-GFP and UAS-hhc85s-GFP(Gorfinkiel et al., 2005). After testing the protein expression levels of several lines of GFP-tagged Hh forms by western blot, we selected the lines that showed the same expression levels for each Hh variant (see Fig. S1 in the supplementary material). UAS-hhN was prepared for the present study. UAS-hhN(Burke et al., 1999), UAS-hhN (Tabata and Kornberg,1994) and UAS- hhc85s(Lee et al., 2001) were also used as controls.

Clones were generated by FLP-mediated mitotic recombination. Larvae of the corresponding genotypes were incubated at 37°C for 1 hour 24-48 hours after egg laying (AEL), or for 45 minutes 48-72 hours AEL. The genotypes of the flies for clone induction were: shi^ts1 FRT18A/ arm-lacZ FRT18A; FLP/+; hh-Gal4 / Hh-GFP(or HhN-GFP or Hhc85s-GFP);dor⁸ FRT18A / arm-lacZ FRT18A; FLP/+; hh-Gal4 / Hh-GFP(or HhN-GFP or Hhc85s-GFP); dor⁸ FRT18A / arm-lacZ FRT18A; ap-Gal4 / Hh-GPF(or HhN-GFP or Hhc85s-GFP); FLP/+; FLP; FRT 42D ttv^681b / FRT 42D arm-lacZ; hhGal4 / Hh-GFP(or HhN-GFP or Hhc85s-GFP).

To generate random clones of Hh-GFP, Hhc85s-GFP and HhN-GFP, the transgenes actin>CD2>Gal4(Pignoni and Zipursky, 1997)and ubx>f⁺>Gal4, UAS-βgal(de Celis, 1998) were used. Larvae of the corresponding genotypes were incubated at 37°C for 10 minutes to induce HS-FLP-mediated recombinant clones.

To generate random clones of Hh-GFP, Hhc85s-GFP and HhN-GFP in an hh^ts2 background, larvae of the genotypes Act>CD2>GAL4 / HS-FLP122; UAS-Hh-GFP(or HhN-GFP or Hhc85s-GFP), hh^ts2 / hh^ts2 and y, w, HS-FLP122;ubx>f⁺>Gal4, UAS-βgal; UAS-Hh-GFP(or HhN-GFP or Hhc85s-GFP), hh^ts2 / hh^ts2 were incubated at 37°C for 10 minutes to induce recombinant clones and incubated at the restrictive temperature (29°C) for 18 hours before dissection.

To generate random clone mutants for ttv that ectopically expressed Hh and Hh-N under Gal4 control, we used the MARCM technique(Lee and Luo, 1999). Larvae of the following genotypes: UAS-Hh / y, w, FLP, Tub Gal4 UAS-GFP; FRT 42D ttv^681b / FRT 42D tub Gal80 and y,w, tubGal4 UAS-GFP; FRT 42D Tub Gal80 / FRT 42D ttv^681b; UAS-Hh were incubated at 37°C for 1 hour 24-48 hours AEL, or for 45 minutes 48-72 hours AEL.

Salivary glands were collected in lysis buffer [Tris-HCl 50 mM (pH 8.0),DOC 0.4%, NaCl 140 mM, Triton X-100 1%, EDTA 1 mM (pH 8.0) containing a cocktail of pepstatin, aprotinin and leupeptin proteases inhibitors] from flies that expressed different forms of Hedgehog-GFP proteins induced with the AB1 Gal4 driver. The lysates (0.1 ml) were loaded on a 12 ml linear glycerol gradient (15-50%). The tubes were then centrifuged for 18 hours at 100,000 g in a SW40 rotor (Beckman) at 4°C. The gradient was fractionated in 21-23 fractions from the bottom (50 % glycerol) to the top(15% glycerol). An identical gradient was prepared using the standard protein horse ferritin, which has an estimated molecular weight of 465 kDa for the multimer. Aliquots were precipitated with TCA, using 10 μg insulin as carrier, and analyzed by western blotting using rat anti-Hh 1/200(Guillen et al., 1995) and rabbit anti-Hh 1/500 (Tabata and Kornberg,1994) in the ECL method of immunodetection. The standard protein was developed with its specific antibody.

Immunostaining was performed according to standard protocols. Antibodies were used at the following dilutions: rabbit polyclonal anti-Hh(Takei et al., 2004), 1:500;mouse monoclonal anti-Ptc (Apa 1.3)(Capdevila et al., 1994),1/50; rabbit polyclonal anti-β-gal (from Jackson laboratories), 1/1000;rabbit polyclonal anti-GFP (Molecular Probes, A-6455), 1/100; rabbit polyclonal anti-Col antibody (Vervoort et al., 1999), 1/200; mouse monoclonal anti-En(Patel et al., 1989), 1/100;rat monoclonal anti-Ci (Motzny and Holmgren, 1995), 1/5; rat polyclonal anti-Caupolican(Diez del Corral et al.,1999), 1/100; and mouse monoclonal anti-DE-Cad (Iowa University Hybridoma Bank), 1:50.

For the western blots, protein extracts from salivary glands of AB1-Gal4/UAS-Hh, UAS-Hh-GFP flies were prepared in Laemmli buffer,resolved by SDS-PAGE, immunoblotted and then analyzed using anti-Hh (1/500)(Tabata and Kornberg, 1994),rat polyclonal anti-Hh (1/200) (Guillen et al., 1995) and anti-GFP (1/1000) (Molecular Probes, A-6455)antibodies. The signal was developed using the ECL Western Blotting Analysis System (Amersham Pharmacia). The presence of horse ferritin was established using a specific antibody (1:1000, Jackson ImmunoResearch Laboratories). The amount of protein extract loaded for each sample was control by Coomasie staining of the SDS-PAGE gel.

The protocols used for labeling the endocytic compartment and extracellular labeling of Hh-GFP are described elsewhere(Strigini and Cohen, 2000; Entchev et al., 2000; Torroja et al., 2004).

Bright-field microscopy imaging was performed using an Axioskop 2 plus(Zeiss) microscope, coupled to a CCD camera. A laser-scanning confocal microscope (LSM 510 META; MicroRadiance and Radiance 2000) was used for confocal fluorescence imaging. Metamorph and Image J software were used for image processing and determining fluorescence levels.

In larvae, the wing imaginal disc is formed by epithelial cells is a sac-like structure. One surface of this structure comprises a monolayer of columnar epithelial cells with their apical membranes orientated towards the disc lumen. This specialized group of cells will give rise to the wing (and notum). The other overlying surface is a squamous epithelium, called the peripodial membrane, which does not develop into any cuticular structure in the adult fly (see Fig. 5). Two populations of cells with different adhesion affinities divide the epithelium of columnar cells into posterior (P) and anterior (A) cells(García-Bellido et al.,1973). Hh is synthesized in P cells and the processed HhNp 20 kDa lipid-modified molecule is able to diffuse several cell diameters away,despite its potential to tether to membranes. However, the spreading of Hh is limited by interaction with the Patched (Ptc) receptor that is expressed in the A cells. When Hh reaches Ptc, a concentration gradient of Hh is formed at the boundary that separates the anterior and posterior (AP) compartments(reviewed by Ingham and McMahon,2001; Torroja et al.,2005). Here, we explored the role of lipid moieties in the spreading and signaling mechanisms of the Hh molecule in the wing imaginal disc. To this end, we engineered GFP-tagged versions of the different Hh variants: HhN (lacking cholesterol), Hhc85s (lacking palmitic acid) and Hhc85sN (lacking cholesterol and palmitic acid), and then compared their functions with that of fully lipid-modified Hh-GFP. The correct processing of Hh-GFP and of variants lacking lipid-modifications was checked in western blots, which showed bands of expected molecular weight for each protein (see Fig. S1A in the supplementary material).

Biochemical studies in tissue cultured cells have shown that Shh, the vertebrate ortholog of Drosophila Hh, can form multimeric complexes in which hydrophobic moieties are thought to be sequestered inside the multimer to make the complex soluble and diffusible(Zeng et al., 2001; Chen et al., 2004). This multimeric complex shows stronger signaling activity than Shh monomers and both lipid modifications are essential for its formation(Chen et al., 2004). Hh multimers have been isolated from S2 cells, a Drosophila cell line,but surprisingly, palmitic acid modification was not essential for Hh to form a multimeric complex in S2 cells (Chen et al., 2004). Thus, we decided to establish whether the lipid-modified and unmodified forms of Hh-GFP expressed in fly tissues could form a multimeric complex. To avoid a possible interaction with endogenous Hh,we ectopically expressed the different Hh-GFP forms in salivary gland cells,where there is no endogenous Hh expression(Zhu et al., 2003). Fig. 1A shows that only the lipid-modified Hh-GFP was able to form multimers. As a control, we repeated the experiment using GFP-untagged versions of Hh and obtained similar results(data not shown). These data are in agreement with vertebrate Shh multimerization but not with previous results, which sugggested that the palmitic modification is not required for the formation of a multimeric complex in Drosophila tissue culture cells(Chen et al., 2004). This discrepancy might be due to the experimental approach employed (see Materials and methods).

In Hh-GFP/hhGal4 wing discs, dots of Hh-GFP, which we previously identified as endocytic vesicles of Hh internalized by Ptc(Torroja et al., 2004), were observed at a distance of five, six or even more cell diameters in the A compartment close to the AP border. Discs expressing HhN-GFP, Hhc85s-GFP(Fig. 1C,D) and even Hhc85sN-GFP (data not shown) also showed these dots distributed throughout the whole A compartment. These GFP accumulations observed for HhN-GFP and Hhc85s-GFP were also detected using several anti-Hh antibodies (see Fig. S1B,part b; C, part c in the supplementary material).

We then tried to establish if these punctate structures were vesicles of the endocytic compartment or extracellular aggregates - it has been suggested that Hh multimers could form extracellular aggregates, called large punctuated structures (LPS), in the embryo (Gallet, 2003; Gallet, 2005). Discs expressing different Hh-GFP variants induced with Hh-Gal4 were incubated `in vivo' with dextran-red to label the endocytic compartment. We found extensive co-localization of Hh-GFP, Hhc85s-GFP and HhN-GFP (as well as Hhc85sN-GFP)with the internalized dextran-red (Fig. 1B-D, parts b-d, arrowheads; data not shown). To check whether the dots of Hh-GFP that did not co-localize with dextrans(Fig. 1B-D, arrows) were intracellular or extracellular, we simultaneously labeled the extracellular Hh-GFP with α-GFP antibody and the endocytic vesicles with dextrans. Fig. S1D shows that the punctate structures of Hh-GFP that did not co-localize with dextrans (arrows) are indeed intracellular vesicles. This indicates that all Hh variants dots correspond to accumulations in the endocytic compartment.

We and others previously reported that Ptc internalizes Hh through a dynamin-dependent mechanism, because there was build-up and co-localization of Hh and Ptc when endocytosis was blocked in the temperature-sensitive dynamin mutants, shibire^ts1(shi^ts1), both in the fly embryo and in imaginal disc cells (Capdevila et al., 1994; Han et al., 2004; Torroja et al., 2004; Gallet and Therond, 2005). We observed that the endocytosed Hh was targeted by the lysosomal protein degradation pathway because Hh and Ptc also accumulated in clones of dor^- mutant cells induced in imaginal discs(Torroja et al., 2004), a mutation that blocks the ubiquitin pathway(Sevrioukov et al., 1999). As the lipid-unmodified forms of Hh do not form a gradient in the wing disc(Fig. 1C,D), we then tried to determine whether the mechanisms of internalization and degradation of HhN-GFP and Hhc85s-GFP were affected.

We induced shi^ts1 mutant clones in discs expressing the variant forms of Hh in the P compartment and observed that Hhc85s-GFP and HhN-GFP accumulated at the apical membrane of shi^ts1mutant clones at the restrictive temperature(Fig. 2B,C,E), while Hh-GFP built up at the basolateral membrane (Fig. 2A,D). This distinct apicobasal accumulation in Hh and HhN or Hhc85s in shi^ts1 clones suggests that the mutant forms diffuse through the apical surface of the epithelium, while wild-type Hh diffuses through the basolateral surface. To visualize the diffusion of Hh, we compared the extracellular staining of Hh-GFP with that of the mutant forms in wing imaginal discs expressing the GFP fusion proteins in the P compartment. Fig. 3 shows the presence of the three proteins on the apical surface in the producing cells(Fig. 3A-C asterisks). Surprisingly, only the mutant forms were able to diffuse through the apical surface towards the A compartment, whereas no apical labeling was found for Hh-GFP in the A compartment (Fig. 3A-C apical arrows). However, Hh-GFP diffused and formed an extracellular gradient in the basolateral space(Fig. 3A, basolateral arrowhead), while the mutant forms of Hh were much less represented and no extracellular gradient or build up were observed(Fig. 3B,C, basolateral arrowheads).

Intracellular accumulation of Hh was also visualized when degradation was blocked in dor⁸ clones. In the A compartment clones,massive accretion of Hh-GFP, HhN-GFP and Hhc85s-GFP was observed(Fig. 4A-C). Remarkably,important differences were found between wild-type and mutant Hh forms. The amount of wild-type Hh deposition was found to be directly related to the amount of Ptc (see graph in Fig. 4A). However, there was no such relationship for HhN or for Hhc85s, given that there was an accumulation of Hh but not of Ptc in these clones (see Fig. 4B,C and graphs). This uncoupling between Ptc and Hh build up suggests distinctive Hh internalization and degradation mechanisms independent of Ptc. To explore this hypothesis, we induced dor⁸ clones in a wing disc expressing the various forms of Hh-GFP fusion proteins in the dorsal compartment. In this model, we can examine the internalization and degradation of Hh-GFP forms in the P compartment where there is no Ptc. As predicted, both HhN-GFP and Hhc85s-GFP were targeted for degradation both in the A compartment and P compartment (Fig. 4E,F),whereas wild type Hh-GFP is only degraded when Ptc is present(Fig. 4D)(Torroja et al., 2004). A small amount of Ptc accumulation was still observed in dor⁸ clones expressing HhN-GFP but not Hhc85s-GFP(Fig. 4E,F, arrows). This last result indicates a lower ability of Hhc85s to interact with Ptc and could explain the much lower response exerted by this form of Hh in terms of target activation compared with HhN.

Several conclusions can be derived from the above results. First, the non-palmitoylated and the non-cholesterolated forms of Hh synthesized in the P compartment (or in the dorsal compartment) are freely distributed and internalized throughout the disc. Second, internalization of wild-type Hh occurs through the basolateral membrane while internalization of the lipid-unmodified forms takes place through the apical plasma membrane; in all cases via a dynamin-dependent mechanism. Third, Ptc independent internalization of the lipid-unmodified forms of Hh also induces the Ptc independent degradation of Hh. This Ptc-independent mechanism of Hh internalization was previously suggested both in imaginal disc cells(Torroja et al., 2004) and in the embryo (Gallet and Therond,2005).

As the internalization of HhN-GFP and Hhc85s-GFP occurred throughout the disc epithelium via the apical plasma membrane, even for areas of the disc far away from the Hh-producing cells, our interpretation was that these secreted forms of Hh could not be properly retained by the extracellular matrix. Hence,it was very likely that Hh forms not modified by lipids would concentrate in the disc lumen after their secretion, escaping the spatial restriction imposed by the extracellular matrix. To test this hypothesis, we expressed Hh-GFP,HhN-GFP and Hhc85s-GFP only in peripodial membrane cells using the ubx-Gal4 line (Pallavi and Shashidhara, 2003) and determined whether these forms of Hh can travel through the lumen and be internalized by the disc proper cells. When we expressed HhN-GFP and Hhc85s-GFP with the ubx-Gal4 driver, we observed Hh vesicles in the disc proper cells(Fig. 5B-c′), whereas no vesicles were observed in the case of wild-type Hh(Fig. 5A,a,A′,a′). HhN endocytosis was more evident in the A than in the P compartment(Fig. 5B′,b′),indicating that although there is Ptc-independent internalization of HhN, Ptc is still able to interact with HhN, as is also reflected by its signaling properties. Thus, HhN-GFP expressed from the peripodial membrane was able to induce the activation of gene targets, such us dpp-LacZ in the disc proper cells (Fig. 5B′). However, Hh-GFP was unable to induce this distant activation of response genes(Fig. 5A′). This non-autonomous effect of HhN-GFP from peripodial to disc proper cells through the disc lumen was also patently manifested by the enlargement of the A compartment (Fig. 5B′)and by the adult phenotypes (Fig. 5E,G). Flies expressing HhN-GFP in the peripodial membrane (but not those expressing Hh-GFP) showed extra veins, enlargement of the costa region of the wing (Fig. 5E)and extra bristles in the notum (Fig. 5G). These mutant phenotypes are similar to those observed as a consequence of ectopic activation of the Hh pathway(Basler and Struhl, 1994; Tabata and Kornberg, 1994) and in hypomorph Ptc alleles (Phillips et al.,1990; Capdevila et al.,1994).

The fact that lipid unmodified Hh can diffuse from the peripodial membrane raises the question of whether the extended diffusion of these forms, seen when expressed in its own domain with the hh-Gal4 driver, is produced as a consequence of its release solely by the peripodial cells. However, these forms still generate more extended gradient than wild-type Hh when expressed with the Ap-Gal4 driver (that is expressed solely in the disc proper cells, Fig. 4) or when single clones are generated in the cells of the wing disc proper (see Fig. S2 in the supplementary material).

These results indicate that lipid-unmodified Hh is poorly retained by the extracellular matrix both in the peripodial cells and in the disc proper cells and is released to the lumen where it can freely diffuse.

We have previously reported that Hh-GFP is able to signal as wild-type Hh(Torroja et al., 2004). We next analyzed the behavior of lipid-free Hh-GFP molecules in clones of cells ectopically expressing each form of Hh in the wing imaginal disc. For all the experiments, we chose fly lines that expressed similar levels of the various forms of Hh-GFP. As a control, we also induced ectopic expression clones of GFP-untagged versions and observed the same responses as with the GFP-tagged forms (data not shown). Ectopic Hh-GFP-expressing clones in the A compartment of the wing imaginal disc induced all Hh responses, both autonomously (arrows)and non-autonomously (arrowheads), indicating that Hh-GFP is able to signal at the same distance from its source as the wild-type Hh protein(Fig. 6C-F)(Torroja et al., 2004). Although Hh is not expressed in A compartment cells, we have previously reported that ectopic Hh in the A compartment is able to activate endogenous Hh expression (Guillen et al.,1995). Hence, to avoid possible interaction of ectopic Hh with endogenous Hh, we also induced clones of all Hh forms in a hh^ts2 background at the restrictive temperature and obtained the same results. Ectopic Hhc85s-GFP clones in the A compartment did not activate en (Fig. 6K) but activated other Hh targets such as ptc (at very low levels and only cell autonomously) and dpp (also at low levels),only autonomously (Fig. 6L-N). HhN-GFP, however, was able to activate all the Hh targets, although in the case of en and ptc the non-autonomous activation was over a very short range (only one cell diameter outside the clone) (see arrowheads in Fig. 6G,H). Surprisingly, low threshold responses, such as dpp, Iro and the cytoplasmic stabilization of Ci, were activated throughout the disc, even though HhN-GFP induction occurred in a restricted clone(Fig. 6I,J; data not shown). This differential non-autonomous response between low- and high-threshold response target genes with cholesterol-unmodified Hh was unexpected (see Fig. 6O,P,Q).

There is compelling evidence that Hh moves through the extracellular matrix and it has been argued that HSPGs are required for the stable retention of Hh on the cell surface. This results in restricted movement through the surface of the epithelium (Bellaiche et al.,1998; The et al.,1999; Bornemann et al.,2004; Han et al.,2004; Tabata and Takei,2004; Takei et al.,2004; Glise et al.,2005; Gorfinkiel et al.,2005). As we observed greater Hh-N spreading and lower signaling than the wild-type Hh, we then analyzed the role of HSPGs in the movement and signaling of HhN-GFP compared with Hh-GFP. As previously published(Bellaiche et al., 1998),lipid-modified Hh was able only to activate its targets in the first row of cells touching the P compartment, reflecting problems in Hh stabilization and diffusion in ttv^- clones(Fig. 7A-D). Accordingly, we observed Hh-GFP endocytic vesicles only in this first row of cells of a ttv^- mutant clone (Fig. 7A arrowheads). These vesicles co-localized with Ptc, indicating that Ptc mediates the internalization of Hh-GFP. This finding was reinforced by the fact that Hh-GFP vesicles disappeared in ttv, ptc double mutant clones (see Fig. S3B in the supplementary material). HhN-GFP and Hhc85s-GFP, however, were able to move through a ttv^-clone, as shown by the GFP punctate structures(Fig. 7E,I arrows). In terms of signaling, HhN-GFP was able to rescue, in all cells inside the clone, Hh low-threshold responses [such as high levels of Ci (data not shown) or the activation of iro (Fig. 7G arrowhead, see also clone 1 in H)] but not high-threshold targets of Hh signaling [such as Ptc or Col(Fig. 7E,F)]. We observed,however, that ttv^- clones located distant to the AP border were not able to activate iro(Fig. 7G, arrow, clone 2 in 7H), indicating that some interaction still exists between HSPGs and HhN. In conclusion, these results suggest that the diffusion of lipid-unmodified forms of Hh and the activation of their targets are less dependent on HSPG function than in the case of fully lipid-modified Hh.

As HhN-GFP was able to move across a ttv^- mutant clone and activate low-threshold Hh targets, we wondered whether HSPG function was required for maximal activation of the Hh pathway. HSPGs have also been previously implicated in the reception of Hh by responsive cells. Thus, the glypican Dally-like is known to be required for Hh reception in embryos and tissue culture cells (Desbordes and Sanson, 2003; Lum et al.,2003; Lin, 2004). We used the MARCM technique (Lee and Luo,1999) to generate clones of ttv^- cells that simultaneously express high levels of the fully lipid-modified Hh. Target genes that respond to high levels of Hh were not expressed inside these clones, although activation in the first row of cells within the clone was observed (Fig. 8A,A′,B,B′, arrowheads). This activation of high-threshold response genes in the first row of cells inside the clone was also shown in ttv^- clones (that did not simultaneously express Hh) abutting the AP border (Fig. 7A,B). This suggests a possible non-autonomous interaction between the Ptc receptor expressed in the cells of the clone and the HSPGs of wild-type cells surrounding the clone. Interestingly, target genes responsive to low levels of Hh were still activated despite the absence of ttvfunction (Fig. 8C,C′,D,D′). When we analyzed ttv^-clones that expressed HhN, the same autonomous responses were observed as when we used fully lipid-modified Hh (data not shown). All these results indicate that HSPG function is required for maximal activation of the Hh pathway at the reception level.

This work indicates that interaction between lipid-modified Hh and HSPGs is required to restrict the movement of Hh to the extracellular matrix and promote optimal activation of the Hh pathway. This finding is based, first, on the analysis of the spreading and signaling properties of the lipid-unmodified forms of Hh and, second, on the responses shown by cells mutant for the genes that synthesize HSPGs to lipid-modified and unmodified forms of Hh.

One of the first observations made when expressing the lipid-unmodified forms of Hh is an extended gradient compared with that elicited by wild-type Hh (Fig. 1B-D). It is known that Hh gradient formation depends on the presence of its receptor Ptc, which is responsible for the internalization and degradation of Hh (reviewed by Torroja et al., 2005). Thus,we analyzed the internalization and degradation of lipid-unmodified forms of Hh, in the search for an explanation for the vast expansion of the gradient. We found that these mutant forms of Hh were efficiently internalized and degraded throughout the disc. So, why is there an extended gradient? One clue emerged from our analysis of Hh internalization in shi mutant cells. We found that the lipid-unmodified forms of Hh were internalized through the apical side of the epithelium, while wild-type Hh is internalized mainly through the basolateral surface. This differential internalization matches the preferential localization of lipid-unmodified Hh at the apical surface plasma membrane of A compartment cells. We also found that the lipid-unmodified forms of Hh can be internalized and degraded through a Ptc-independent mechanism. Thus, it is possible that this Ptc-independent mechanism would have no positive feedback from Hh, and would therefore not work as efficiently as when internalization was mediated by Ptc (see Lander et al., 2002) or by just a lower rate of internalization and degradation of Ptc. Alternatively or in addition, a reduced restriction of spreading through the extracellular matrix could explain why the gradient is extended. In agreement with this, we observed that the localization of lipid-unmodified Hh is not affected in mutants for HSPGs (data not shown). Furthermore, Hh is less represented at the basolateral membrane in the absence of HSPGs(Gorfinkiel et al., 2005),indicating an active role of HSPGs in anchoring lipid-modified Hh in the lateral cell region. This conclusion is further supported by the localization of the glypican Dally-like at the basolateral membrane of wing imaginal disc cells (Baeg et al., 2004; Han et al., 2005).

If the interaction of dual lipid modified Hh with HSPGs is important for Hh retention in the extracellular matrix, then we would expect to see the same phenotype if we removed the lipids from Hh as ttv mutants, i.e. the extended movement of Hh. This possibility could only be demonstrated using a whole disc mutant for ttv, which is difficult to obtain because ttv mutants are larval lethal. A further possibility would be to use a hypomorphic mutation of ttv, which is not available for Drosophila. In mice, in which an Ext1 hypomorphic mutant already exists, it has been demonstrated that this is indeed the case and Ihh signaling during embryonic chondrocyte differentiation shows an extended range(Koziel et al., 2004).

In the formation of the Hh morphogenetic gradient throughout the extracellular matrix, we have to consider not only the function of HSPGs,rendering a restricted space suitable for the spreading of Hh (reviewed by Lin, 2004; Tabata and Takei, 2004) but also a possible role of HSPGs in signal reception. Thus, if a specific HSPG acts as a co-receptor for Hh together with Ptc, binding the ligand to the HSPG and to the receptor, it could limit the range of ligand movement and, at the same time, boost the signaling process. In effect, compelling evidence in the fly embryo and in fly tissue culture cells indicates that the presentation of Hh to Ptc might require a specific HSPG such as Dally-like(Desbordes and Sanson, 2003; Lum et al., 2003). In this sense, we show here that HSPG function is also required for Hh reception in the imaginal disc, but only to trigger Hh high-threshold response genes (Figs 7, 8). Thus, it is plausible that Hh coupled to a specific HSPG could form a high-affinity complex together with Ptc to allow these high-threshold responses. However, we observed that in the wing imaginal disc, Hh lacking cholesterol was able to induce the same low responses as fully lipid-modified Hh, in both cases in an HSPG-independent manner. This suggests that the low-level response of the pathway might involve a different reception complex in which HSPGs are not necessary.

Two different mechanisms of Hh target gene activation have been also proposed in the ectoderm of Drosophila embryos. These mechanisms correlate with asymmetric cellular responses to Hh signaling. Hh requires the cholesterol modification for a maximal response of the pathway, i.e. for the activation of Wingless (Wg) and Ptc in anteriorly located cells (compartment border), but cholesterol is dispensable for the activation of Rhomboid (Rho)and Ptc in posteriorly located cells (segmental border)(Gallet et al., 2003). Moreover, ttv function is needed only for activating target genes in anterior cells (Wg) (The et al.,1999) and not for target genes posterior to the Hh source (Rho)(Gallet et al., 2003). This result was interpreted as a requirement for Hh movement towards anteriorly located cells to activate Wg. Based on the requirement of ttv for Hh high threshold responses in wing imaginal disc cells, we anticipate that ttv is required in the embryonic epidermis for the reception of Hh in the activation of specific Hh target genes but not others. This differential activation of Hh targets in the embryonic ventral epidermis and wing imaginal disc cells is known to be dependent on fused (fu) activity(Sanchez-Herrero et al., 1996; Ohlmeyer and Kalderon, 1998; Therond et al., 1999; Mullor and Guerrero, 2000). Hence, it is possible that this differential activation of the Hh pathway could be exerted through different structures of the Hh receptor complex. Thus, Ptc plus a glypican, possibly Dlp, could act as a high-affinity receptor whose activation would then be mediated by Fu kinase. Ptc without Dlp could act as a low-affinity receptor whose activation was independent of Fu.

Also in the embryonic epidermis, it has been proposed that apically distributed Hh LPSs are needed to activate Wg in anterior cells(Gallet et al., 2003). However, in the wing disc, we observed that all forms of Hh produced punctate structures, both apical and basolateral, which are also observed using different Hh antibodies (see Fig. S1 in the supplementary material). These structures are endocytic vesicles and are the result of the accumulation of Hh in the endocytic compartment rather than the visualization of multimeric lipid-modified Hh complexes moving from one cell to another. The high accretion of Hh observed in the dor^- clones indicates that these large endocytic vesicles are targeted to the degradation pathway.

Hh lipidation seems to confer a specific conformation to the Hh molecule so that it is targeted to specific locations in the receiving cells for signaling. Hence, the Ptc receptor might be located in sterol-rich membrane microdomains or lipid rafts in Drosophila, which function as platforms for intracellular sorting and signal transduction(Rietveld et al., 1999)(reviewed by Incardona and Eaton,2000). Hh without lipids might not recognize these platforms,resulting in less efficient signaling. Interestingly, we observed that the internalization of the lipid-unmodified Hh that was not mediated by Ptc, which is more striking in the case of Hhc85s. HhN has less potency to activate the Hh pathway than wild-type Hh. As previously reported(Lee et al., 2001; Gallet et al., 2003), HhN has less potency to activate the Hh pathway than wild-type Hh. Hhc85s is much less potent than HhN (Fig. 6) and Hhc85sN is even less potent (data not shown). As we have shown, this low response in terms of target activation of unlipidated forms of Hh could be explained by their diminished access to Ptc.

The literature contains several contradictory conclusions regarding the signaling functions of Hh lipids in the mouse and Drosophila. Thus,the elimination of cholesterol has been reported to have a major effect on Hh signaling in vertebrates but only minor effects in Drosophila(Porter et al., 1996a; Burke et al., 1999). However,the diffusion of Drosophila HhN and Hhc85s differs from that of their vertebrate counterparts. In vertebrates, forms of Shh lacking cholesterol or palmitic acid showed restricted signaling and diffusion(Lewis et al., 2001; Chen et al., 2004), while in Drosophila, Hh without lipids is able to diffuse longer distances(Porter et al., 1996a; Burke et al., 1999). It is likely that the different structural characteristics of the target tissues where Hh acts could account for differences in the lipid requirements of signaling and diffusion between Drosophila and vertebrates. The wing imaginal disc consists of a single-layered sac of polarized epithelial cells with their apical surfaces orientated towards the disc lumen. If the extracellular matrix were not able to retain lipid-unmodified Hh, this molecule would be delivered to the disc lumen. The peculiar structure of the wing disc concentrates the lipid-unmodified forms of Hh in the lumen and this is likely to promote the activation of low-threshold target genes in cells far away from the Hh-producing cells. A role for the luminal transmission of ligands has already been described for Dpp signaling in Drosophilawing disc (Gibson et al.,2002). By contrast, we show here that only lipid-unmodified Hh can travel from the peripodial membrane to the disc proper cells(Fig. 5). Our results indicate that wild-type Hh is not traveling from the peripodial membrane towards the wing disc. In other systems, Hh without lipids might be lost because it is not retained and stabilised by the extracellular matrix; therefore, it would be expected that very restricted diffusion and signaling of non lipid-modified forms of Hh occurs, as demonstrated by previously(Chamoun et al., 2001; Kohtz et al., 2001; Lee et al., 2001; Lewis and Eisen, 2001; Chen et al., 2004). Our results indicate that the role of lipids in Hh signaling is similar in Drosophila and vertebrates, and corroborate the previous notion that lipid-modified forms of Hh are predominantly membrane associated and that Hh mutated forms lacking lipid adducts dissociate from cells after secretion(Lee et al., 1994; Porter et al., 1995; Pepinsky et al., 1998; Peters et al., 2004). In summary, we conclude that dual lipid modification, by cholesterol and palmitic acid, appears to be crucial for interaction between Hh and HSPGs, as well as the Ptc receptor, and that these interactions are important both for a precise Hh spreading through the epithelium surface and for proper Hh reception.

We are very grateful to Nicole Gorfinkiel, Kiki Lalioti and Ignacio Sandoval for their help and discussions, to Carmen Ibañez for her excellent technical assistance and to Ernesto Sánchez-Herrero and Aphrodite Bilioni for comments on the manuscript. We also thank T. Tabata for the anti-Hh antibody, A. Vincent for the anti-Col antibody, R. Holmgren for the anti-Ci antibody and J. Modolell for the anti-caupolican antibody. Fly strains were provided by X. Lin, L. S. Shashidhara, J. Treisman, S. Zipursky,G. Morata, K. Basler, T. Kornberg, H. Kramers, S. Bray, T. Tabata and the Bloomington stock center. This study was financed by the Spanish D.G.I.C.Y.T.,grants BFU2005-04183/BMC and MCYT GEN2001-4846-C05-01, the Comunidad Autónoma de Madrid, grant GR/SAL/0185/2004, and by an institutional grant from the Fundación Areces. L.Q. is supported by an I3P contract of the CSIC.