Defects in glucuronate biosynthesis disrupt Wingless signaling in Drosophila

Glucuronic acid is an essential building block of several classes of glycosaminoglycans (GAGs) that are found in both vertebrates (reviewed in Bernfield et al., 1992; Jackson et al., 1991; Kjellen and Lindahl, 1991) and invertebrates (Cassaro and Dietrich, 1977; Cambiazo and Inestrosa, 1990). GAGs are linear carbohydrate chains composed of a repeating disaccharide that is attached to serine resides of a core protein to produce a proteoglycan. Seven classes of GAG chains are recognized depending on the type of disaccharide repeat that they contain and the manner in which the chains are modified by sulfation and sugar epimerization (Kornfeld and Kornfeld, 1980). These glycans include heparin, heparan sulfate, hyaluronic acid, chondroitin sulfate, dermatan sulfate and keratan sulfate.

The biological functions ascribed to GAGs are quite varied and include biomechanical, structural and regulatory roles (David, 1993; Jackson et al., 1991; Kjellen and Lindahl, 1991; Yanagishita and Hascall, 1992). Of particular interest, GAG chains can concentrate and assemble various proteins on the cell surface including proteases, antiproteases, lipolytic enzymes (Griffith, 1986; Shimada et al., 1981) and, most importantly, growth factors (David, 1993; Jackson et al., 1991). Many growth factors will bind heparin (e.g. heparin-binding-EGF (HB-EGF), PDGF, BMP-2, members of the Wnt and Hedgehog (Hh) families, fibroblast growth factor (FGF) and transforming growth factor β (TGF-β), and several mechanisms have been proposed for the regulation of signal transduction by proteoglycans (Bradley and Brown, 1995; Kelly et al., 1993; Lee et al., 1994; Reichsman et al., 1996; Ruppert et al., 1996). These include a coreceptor function involving low affinity binding of ligand (Klagsbrun and Baird, 1991; Schlessinger et al., 1995), and possibly sequestering of growth factors in the matrix for later release (Yanagishita et al., 1992).

Despite the abundance of information on GAG chains in vertebrate tissues, little is known about the roles that these polymers play during development. In this report, we describe the phenotypic affects of mutations in the Drosophila suppenkasper (ska) gene. We show that ska codes for the enzyme UDP-glucose dehydrogenase, which makes UDP-glucuronate from UDP-glucose. In the absence of a known salvage or bypass pathway, ska mutants are expected to lack all glucuronate-containing GAG carbohydrate chains. ska mutants thus provide the first opportunity to investigate the function of the GAGs in vivo. Surprisingly, ska mutants do not exhibit highly pleiotropic phenotypes. Developmental defects appear to be primarily the result of reduced Wg signaling. These results are discussed in relation to some of the proposed roles of GAG polymers.

In this study, the ska alleles SG9 (Shearn and Garen, 1974), A31, N71 (Anderson et al., 1995), l(3)08310 and l(3)10503 (Bloomington stock center) and the deficiencies Df(3L)W5.4 and Df(3L)XAS96 (Anderson et al., 1995) were used. ska^SG9 and ska^P8310 were recombined with FRT sites at 79 D/E and balanced over TM3, Ubx-lacZ (TM3 P{w^+mC, Ubx-lacZ}). The P{w^+mc, Hs-wg} and P{w^+mc, Hs-hh} transgenes, abbreviated hs-wg and hs-hh (supplied by A. Bejsovec) and the P{w^+mc, Kr-lacZ} Kr-lacZ transgenes were crossed into the ska^P8310 line and homozygous stocks were established. The 2.3 kb cDNA was cloned into the ubiquitin-hs70-Casper (P{w^+mc, ubi-ska}) (Brummel et al., 1994) and UAST vectors (Brand and Perrimon, 1993) and injected into w embryos. The P{w^+mc, ubi-ska} transformants were used to rescue ska^SG9 and ska^P8310 mutants. The CyO-wglacZ chromosome (CyO, P{lacZ^en.A, ry^+t7.2=en1}wg^en11) (Kassis et al., 1992) was used to make lines heterozygous for wg.

To obtain germ-line clones, hs-FLP/+, ska^P8310-FRT79/ovo^D-FRT79 females were heat shocked for 2 hours in a 37°C warm room. After recovering from heat shock for several hours, they were crossed to the ska^P8310 / TM3,Ubx-lacZ stock or to the hs-wg and hs-hh versions of that stock. For ubiquitous wg and hh expression by heat shock, embryos were collected for 2-5 hours after egg laying and heat shocked for 20-30 minutes in a water bath at 37°C. They were transferred into a 25°C incubator and fixed 1.5-2 hours later with 7% formaldehyde in PBS for 20 minutes.

To produce mitotic clones in adult tissues, females of the following genotype, y w hsFLP1; mwh³ska^P8310P[ry⁺; hs-neo; FRT]80B/P[ry⁺;y⁺]66E,P[ry⁺; hs-neo; FRT]80B, were heat shocked during first and second instar larva stages at 37°C for 1 hour (Xu and Rubin, 1993). The resulting ska^P8310 clones were marked with mwh and y. Adult structures were dissected in 70% EtOH, dehydrated briefly in isopropanol, placed under a coverslip in Gary’s magic mountant (Reichsman et al., 1996) and photographed using a Nikon Optiphot.

wg and dpp expression were monitored by whole-mount in situ hybridization using digoxigenin-labeled antisense RNA probes. Plasmids (all in Bluescript) used as templates for probes were F3b (ska 2.3 kb cDNA); wg651 (a kind gift of B. Cohen), a 3 kb wg cDNA; dppE55, a 4 kb dpp cDNA (Padgett et al., 1987); hhC11, a 2.3 kb cDNA (Lee et al., 1992). The probes were prepared according to the manufacturer’s directions (Boehringer Mannheim 1277 073). Unin-corporated nucleotides were removed by LiCl precipitation and the RNA from 1 μg of template was resuspended in 100 μl DEPC water; 20 μl of each probe were hydrolyzed, LiCl precipitated, resuspended in 100 μl hybridization solution and diluted 1/50 μl for the hybridization reaction. The prehybridization procedure and hybridization conditions used are based on the protocol of (Tautz and Pfeifle, 1989) with previously described modifications (Mason et al., 1994).

Antibody stainings were carried out following the protocol published by Frasch et al. (1987), using the Vectastain reagents (anti-mousebiotin-streptavidin-HRP), AP-anti-rabbit antibodies from Promega, or FITC- and Cy3-conjugated anti-rabbit and anti-rat from Jackson labratories. Antiserum against En (a kind gift from N. Patel) was used at a 1:2 dilution. Antisera against β-galactosidase was obtained from Sigma (rabbit) and from Promega (mouse) and used at dilution of 1:5000 and 1:1000, respectively. Embryos were stained in 0.5 mg/ml 3,3-DAB in 0.1 M Tris-Cl, pH 7.5 (HRP) or with NBT/BCIP (AP). Antiserum against Arm (rat), a kind gift from M. Peifer, was used at 1:250. Embryos were mounted in either Canada balsam or 75% glycerol, 50 mg/ml n-propyl gallate (immunofluorescence) and observed with a Bio-Rad MRC 1024 scanning confocal microscope. Antiserum against Drosophila Syndecan (Spring et al., 1994), a gift from Stephenie Paine-Saunders, was used at 1:200. Antiserum against Mad (guinea pig) was a kind gift from Stuart Newfeld and used at 1:1000 (Newfeld et al., 1996). Protein bands that are recognized by the anti-syndecan (rabbit) and anti-Mad antisera were visualized using the ECL reagents from Amersham. Proteins that contain heparan sulfate side chains were detected by AP-staining using a monoclonal antiserum obtained from Seikagaku, America, Inc. at a dilution of 1: 2000.

Approximately 900 ska^P8310 germ-line mutant embryos were selected by their phenotype. Wild-type embryos were collected as a control. Both batches of embryos were homogenized in 2 M urea/150 mM NaCl/50 mM sodium acetate pH 4.5/0.1% Triton X-100, and protein concentrations were measured by Bradford assay. 42 mg of wild-type and mutant protein extracts were run on a 7% denaturing acrylamide gel and transferred to a nitrocellulose membrane in a wet transfer chamber (Harlow and Lane, 1988).

Genomic DNA was obtained from l(3)08310 and l(3)10503 flies by plasmid rescue. Briefly, the DNA was digested with XbaI or XbaI and SpeI, and then ligated and electroporated into DH5α-bacterial cells. After outgrowth for 1 hour at 37°C, the cells were plated on kanamycin-containing medium and resistant colonies picked for analysis. The P-element insertion sites were identified by sequencing using P-end and 5′ specific primers. A 4 kb HindIII fragment was used to screen genomic (λ-fix Stratagene) and cDNA (λ-zap, a kind gift from C. Thummel) libraries. Several positive cDNA clones were obtained. The longest, a 2.3 kb cDNA (F3b), was isolated and both strands were sequenced.

We have devised several sensitized genetic screens to identify genes that modulate either Wg or Dpp signaling. In one screen, an constitutive active Dpp type I receptor encoded by the thick veins (tkv) gene is ectopically expressed in wing discs resulting in a highly blistered wing phenotype (Hoodless et al., 1996). Dominant suppressors of this phenotype are recognized by their ability to reduce blistering. In a second screen, modulators of Wg signaling are identified by their ability to enhance a weak dishevelled (dshw) adult phenotype (Theisen et al., 1994). Dsh is a downstream component of the Wg signaling pathway (Klingensmith et al., 1994; Theisen et al., 1994; Yanagawa et al., 1995). Since both dpp and wg are required for proper growth and patterning of imaginal discs, we used our sensitized backgrounds to screen a collection of small disc mutants (Shearn and Garen, 1974) to determine whether loss of one allele of any of these genes would either enhance the weak dsh phenotype or suppress the activated tkv wing phenotype. One mutation of the collection, SG9, which defines the locus that we now refer to as ska, significantly enhances the weak dsh phenotype (Fig. 1A) when heterozygous. On average, 5% of the weak dsh mutant animals show adult defects consisting primarily of deleted or duplicated antennae and/or deleted wings with a concurrent duplication of notal structures (Fig. 1B,C). When heterozygous for skaSG9, the frequency of adult defects increases to 19%. The skaSG9 allele also shows a mild suppression of the activated tkv phenotype (data not shown). Homozygous and hemizygous skaSG9 animals die at the larval and pupal stage and contain small imaginal discs. Some homozygotes produce pharate adults with reduced and partially differentiated adult tissues that are derived from the imaginal discs while the size of tissues derived from histoblasts is normal (Fig. 2B).

To further characterize this locus, we sought additional alleles of ska. The original SG9 mutation was mapped by recombination to the left arm of the third chromosome (Shearn and Garen, 1974), corresponding roughly to the 65 cytological interval. Several deficiencies for this region fail to complement skaSG9. Complementation tests for lethal mutations reveal that the EMS mutations AE31 and N71 are allelic to skaSG9 (Anderson et al., 1995). In addition, two P element inserts obtained from the Drosophila genome project, l(3)08310 and l(3)10503, also fail to complement skaSG9. All these alleles appear to be stronger than SG9 since, when heterozygous to a deficiency for the region, they die between the first and second instar larval stages while SG9 dies as a pharate adult. The skaP8310 allele shows a higher frequency of adult defects (41%) in a weak dsh background than SG9 (Fig. 1A). The enhancement of the dshw phenotype by the strong ska alleles is similar to that by strong wg alleles suggesting that ska may be invloved in mediating Wg signaling.

To determine whether ska is maternally contributed, we produced germ-line clones of both the skaSG9 hypomorphic allele as well as the stronger skaP8310 allele. Cuticles prepared from mutant animals derived from germ-line clones (henceforth referred to simply as ska mutant embryos) show a loss of the naked cuticle and a mirror-image duplication of denticle belts (Fig. 2A). This phenotype is strikingly similar to that produced by loss-of-function mutations in genes of the Wg signaling pathway (Bejsovec and Martinez Arias, 1991). In contrast, mutant animals from germ-line clones that inherit a wild-type copy of ska from the father (identified in all our experiments by marked balancer chromosomes) are fully rescued and show no defects.

We employed several experiments to address the role of ska in Wg signaling. In the first set of experiments, we examined the expression of engrailed (en) and wg in ska mutant animals. Transcription of en and wg is initiated at the blastoderm stage through the action of pair-rule genes in stripes of neighboring cells that form the parasegment compartment boundary (Lawrence and Johnston, 1989; Lawrence and Struhl, 1996). At germ-band extension, wg and en mutually support each other’s expression as the result of a feedback loop involving hedgehog (hh) (reviewed in Martinez Arias, 1993). In the absence of wg signaling, expression of both wg and en fade during late germband extension (DiNardo et al., 1988; Ingham, 1993; Martinez Arias et al., 1988). We observed similar effects in ska germ-line mutant embryos. En and wg stripes are present at early germband extension (Fig. 2C,E), but start to fade at late germ-band extension and during germ-band retraction (Fig. 2D,F). During the time that wg expression is fading, partial En stripes are seen in many embryos even after germ-band retraction (Fig. 2D). Since similar results have been observed in temperature-sensitive mutants of wg (wg-ts), our results suggests that Wg signaling is reduced but not eliminated in strong ska mutants. In wild-type embryos and wg-ts mutants, En is expressed in stripes that are two cells wide. In contrast, in some of the ska mutant embryos, both early and late En stripes are up to four cells wide. One possible explanation for this observation is that Wg protein may migrate further from source cells in ska mutants than in wild type, but the overall local concentration that a cell perceives is lower and thus signaling is ultimately attenuated.

Since there is a feedback loop that requires Hh to maintain wg expression, these effects could also be interpreted as the result of loss of Hh rather than Wg signaling. Since accumulation of Arm in early germ-band-extended embryos is the most direct assay that we know for Wg signaling, we examined the distribution of Arm protein in ska embryos. Arm protein accumulates in the cytoplasm only in cells that receive the Wg signal (Riggleman et al., 1990). In paternally rescued ska germ-bandextended mutant embryos, the Arm protein is enhanced in stripes of cells that receive the Wg signal (Fig. 2G). However, in ska mutant embryos, the intensity of the Arm stripes are significantly reduced (Fig. 2H). This result indicates that Wg signaling is reduced but not completely elminated in ska mutants at a time when wg expression is still independent of Hh.

As a further test for the function of ska in Wg signaling, we ectopically expressed wg in ska germ-line mutants using a heat-shock promoter. In wild-type embryos, hs driving wg leads to an expansion of the En stripes from roughly two cells to four cells (Noordermeer et al., 1992). In contrast, loss of ska both maternally and zygotically blocks the ability of hs-wg to expand the En stripes and leads instead to reduced rather than expanded En stripes (Fig. 2J). Paternal rescue of ska embryos restores the ability of hs-wg to cause an expansion of the En stripes after heat shock (Fig. 2I). These results suggest that the Wg signal is not properly received or transmitted in ska mutants.

A similar experiment using hs-hh (Ingham, 1993) leads to expansion of the wg stripes in both wild-type or paternally rescued ska embryos (Fig. 2K). However, in ska mutant embryos, the wg stripes fade (Fig. 2L) suggesting that Hh is not able to induce wg expression in ska mutant background. However, since Wg autocrine signaling is required for wg expression (Hooper, 1994) and Wg signaling is reduced in ska mutants, it is difficult to determine whether the observed effect is primarily the result of reduced Wg signaling or the effect of a combined reduction of Wg and Hh signaling.

Since the ska mutations did show weak suppression of an constitutive active Tkv phenotype in wings, we examined whether the ska germ-line mutants exhibited any evidence of dorsal-ventral patterning defects by asking whether the ventral denticle belts are expanded in ska mutant embryos as they are in embryos mutant for dpp or its effectors. We find no evidence for expansion of denticle belt width as is typically produced by mutations in genes of the Dpp signaling pathway (Arora and Nüsslein-Volhard, 1992). Further, formation of the amnioserosa, the dorsal-most tissue that is the most sensitive indicator of loss of Dpp signaling (Wharton et al., 1993) appears normal in ska mutant embryos when monitored by expression of the Krüppel (Kr) gene as a marker for amnioserosa (Fig. 3E). Finally, ska embryos gastrulate properly. These results suggest that Dpp signaling is not substantially affected in the early ska mutant embryos.

To explore the role of ska in adult pattering, we generated clones of the strong skaP8310 mutation in imaginal discs using FRT-mediated recombination (Xu et al., 1993). Induction of clones during the first or second instar produced large clones in adult tissues but with no patterning defects even when the clones involved regions in which Wg signaling is required as wing margin or ventral legs (Fig. 3A,B).

We also analyzed the zygotic phenotypes of several different ska mutants. As described, skaSG9 is the only allele that survives to the third instar larval stage. skaSG9 homozygous larvae have smaller but normal-looking imaginal discs (data not shown). In skaSG9/skaA31 transheterozygotes, the eye disc is proportionally much smaller compared to the other discs. When stained with probes against wg, hh and dpp, normal wg and hh expression is observed but very low dpp expression was detected in the morphogenetic furrow compared to wild type (Fig. 3F-H). The expression of dpp in the antennal portion of this disc appears normal. Since dpp expression in the furrow is dependent on hh signaling, this result raises the possibility that hh signaling may be affected to some degree by the ska mutation.

To clone the ska locus, we used plasmid rescue to isolate genomic DNA flanking the P8310 and P10503 insertion sites. Hybridization probes prepared from the genomic sequences were then used to screen Drosophila cDNA libraries. The largest clone identified, a 2.3 kb cDNA was sequenced (Fig. 4A). The cDNA contains an open reading frame beginning 223 bp from the 5′ end and extending to bp 1532, 752 bp from the 3′ end. The open reading frame encodes a predicted protein of 53×103 Mr. Comparison of this clone to the GenBank database revealed that the Drosophila sequence is approximately 66% identical at the amino acid level to the C. elegans, bovine and soy bean (Glycine max) UDP-glucose dehydrogenase enzymes (Fig. 4B). The clone showed much lower similarity to enzymes of different bacteria as well as to other related dehydrogenases (not shown).

To confirm that this sequence corresponds to the ska locus, we determined the precise insertion point for the two P elements. These two transposons were located 38 bp from each other in the 5′ untranslated region (Fig. 4A). To show conclusively that the transcript represents the ska locus, we expressed the 2.3 kb cDNA in transgenic animals using a ubiquitin transformation vector P{w+mC, ubi-ska} or ubi-ska (Brummel et al., 1994). We found that animals homozygous for skaSG9 as well as skaP8310 were completely rescued in the presence of most transgenes. These results indicate that the ska phenotype is caused by the lack of UDP-glucose dehydrogenase activity.

Interestingly, some transgenic lines show only partial rescue, possibly due to position effects on the ubiquitin promoter. One of these transgenes gives partial rescue and produces a batshaped wing (Fig. 3C) with margin defects characteristic of loss of Wg signaling (Couso et al., 1994). Other transgenes produce distal deletions of veins (Fig. 3D) that are characteristic of mutants in the EGF- and Dpp-signaling pathways (Spencer et al., 1982; Sturtevant and Bier, 1995; Sutherland et al., 1996).

To further investigate the role of ska in Wg signaling at later stages of development, we crossed a wg mutant allele into the background of a skaP8310 mutant that was completely rescued by a single copy of the ubi-ska transgene. Animals heterozygous for a wg mutation died as second instar larvae whereas animals with two wild-type copies of wg were completely rescued. This result shows that a 50% reduction in the Wg signal is able to reverse the rescuing effect of the transgene suggesting that Wg signaling is compromised by ska mutations not only in embryos but also in larvae.

To determine the pattern of ska expression during development, we hybridized digoxigenin-labeled RNA probes to whole-mount embryos and imaginal discs (Fig. 5). As expected from the germ-line clonal analysis, significant maternally deposited ska message is found in precellularized embryos (Fig. 5A). At the germ-band-extended stage, expression is enriched in the mesoderm (Fig. 5B) followed by strong staining in the midgut after germ-band retraction (Fig. 5D). After dorsal closure, expression is enriched in the hindgut and pharynx (Fig. 5D,E). In third instar larvae, ubiquitous expression is observed in most imaginal discs (Fig. 5F-H). However, low levels of expression are seen in the eye disc with enriched levels in the morphogenetic furrow (Fig. 5F).

UDP-glucose dehydrogenase is required for the synthesis of UDP-glucuronate (GlcUA), a precursor for the production of GAG chains. Since both ska P-elements are inserted into the 5′ leader region of ska, we expected that these strong alleles should produce very little of this enzyme and hence should have little or no GAG modification on proteoglycans. To obtain physical evidence that GAG chain biosynthesis is indeed blocked in ska mutants, we monitored GAG-containing proteoglycans on immunoblots of extracts from wild-type and ska germ-line mutant embryos using a monoclonal antiserum against heparan sulfate. Heparan sulfate consists of a repetitive unit of GlcUA and GlcNAc that is attached via two galactose and a xylose molecule to Ser/Thr residues of proteoglycans. Since the UDP-xylose that is attached first to the Ser/Thr residue is also synthesized from UDP-GlcUA, we expect that no GAG side chains should be formed in the absence of GlcUA. As shown in Fig. 6A, the wild-type lane shows multiple bands that strongly interact with the antiserum against heparan sulfate. These bands do not appear in the ska mutant lane. The same blot was also monitored for the migration properties of Drosophila Syndecan (Kan et al., 1993; Spring et al., 1994). Syndecan is a heparan-sulfate-containing transmembrane proteoglycan present in both vertebrate and invertebrates (Bernfield et al., 1992; Couchman and Woods, 1996). On western blots, Syndecan from wild-type embryos migrates as a smear of two or three major bands above 200×103 Mr (Fig. 6B). In extracts from ska germ-line clones, the high molecular mass smear is missing while a novel band migrating around 125×103 Mr appears. Control staining using an antiserum against the mothers against dpp gene product (Mad) show that equal amount of protein extract were loaded, and that no protein degradation occured (Fig. 6C). These results suggest that GAG chain biosynthesis is inhibited in ska mutants in Drosophila.

In this report, we show that the Drosophila ska locus codes for UDP-glucose dehydrogenase. This enzyme is required for the conversion of UDP glucose to UDP-glucuronic acid. Our results suggest that the most crucial role for this compound during Drosophila development is to permit the synthesis of GAGs that help mediate Wg and possibly other growth factor signaling.

In vertebrates, UDP-glucuronic acid serves several functions. In the liver, UDP-glucuronate helps detoxify nonpolar molecules by converting them into more easily secreted polar derivatives (Tephyl and Burchell, 1990). In another pathway, UDP-glucuronate is converted into L-ascorbic acid (vitamin C; Lehninger et al., 1993). Several vertebrate species including humans lack the last enzyme in the vitamin C biosynthetic pathway and therefore require vitamin C in the diet. It seems unlikely that these two pathways could account for the ska mutant phenotype in Drosophila. First, problems caused by the failure to detoxify ingested compounds are not likely to occur in embryonic epidermal cells, a tissue whose patterning is dramatically affected by ska mutants. Second, it can be assumed that the lack of vitamin C synthesis, at least in larva, can be compensated by uptake of this compound from the media during feeding and therefore the larval defects that we observe in partial loss-of-function ska mutants are not likely to be caused by a vitamin C deficiency.

The most likely explanation for the ska mutant phenotype is a block in the synthesis of the glycosaminoglycan side chains of proteoglycans. As we have shown in Fig. 6A, extracts from ska germ-line mutant embryos lack the robust bands of several proteoglycans that are recognized by a monoclonal antiserum against heparan sulfate. In addition, the high molecular bands of the heparan sulfate-containing protein Syndecan are missing in extracts from mutant embryos whereas a novel band of about 125×103 Mr is recognized (Fig. 6B). Control staining against the Mad protein shows that there is no protein degradation (Fig. 6C) in our samples leading us to conclude that GAG chain synthesis is severely effected in ska mutants and that there is no bypass or salvage pathway that permits GAG chain synthesis in the absence of ska.

Consistent with our finding that GAG chain removal severely reduces Wg signaling is the recent observation that Wg signaling in Drosophila tissue culture cells can be inhibited by the removal of GAGs from the surface of cells receiving the Wg signal and restored by the addition of exogenous heparin (Reichsman et al., 1996). In addition, heparan sulfates are impli-cated in Wnt -11 autocrine signaling in the ureter epithelium of the mouse (Kispert et al., 1996).

Several mechanisms have been proposed whereby GAGs might influence cytokine signaling. One proposal is that cell surface proteoglycans act as low affinity coreceptors for growth factors such as FGF and TGF-β(Schlessinger et al., 1995). By this model, proteoglycans would reduce ligand diffusion from three to two dimensions (Bernfield et al., 1992). Binding to the cell surface via interaction with a low affinity proteoglycan would increase the local concentration of ligand, thus enhancing the probability of a productive interaction with a high affinity receptor. The heparan sulfate side chains of cell surface Syndecan may perform this function in mediating FGF signaling (Bernfield and Hooper, 1991, 1993; Steinfeld et al., 1996). For TGF-β, betaglycan represents the low affinity type III receptor (Lopez-Casillas et al., 1993). In this case however, TGF-βappears to bind directly to the core protein itself rather than the carbohydrate side chains (Lopez-Casillas et al., 1994).

A second model of GAG action is to oligomerize ligands thus inducing receptor clustering. In the case of FGF, as well as several other heparan sulfate binding growth factors, the high affinity receptors have intracellular tyrosine kinase domains which are activated by transphosphorylation as a result of ligand-induced receptor dimerization. Since FGF binds to the extracellular domain of its high affinity receptor as a monomer (Spivak-Kroizman et al., 1994) that is not capable of inducing receptor dimerization on its own, it has been proposed that a multimeric heparan sulfate FGF complex is required in order to produce a biologically active signal (Mason, 1994; Ornitz et al., 1992).

The phenotypic effects of ska mutations are consistent with a role for GAGs in mediating Wg signaling. There are two major effects on Wg signaling seen in ska mutants. First, there are initially more cells that express the En protein suggesting that Wg is diffusing further than in wild-type embryos. Second, the level of Arm that accumulates in the Wg-receiving cells is much lower in ska mutant background and eventually the En staining also fades. These observations suggest that GAG chains in the extracellular matrix may help to concentrate Wg perhaps by restricting its diffusion. GAG chains may also modulate reception of the Wg signal. The recently identified candidate Wg receptor (DFrizzled 2) is a member of the serpentine class of seven transmembrane G-protein-coupled receptors (Bhanot et al., 1996). There is no known requirement for receptors of this class to oligomerize in order to signal. Thus, the primary role of GAGs in Wg signaling may be to increase surface concentration of Wg on receiving cells as opposed to mediating receptor oligomerization. Alternative models such as stabilization of the ligand from potential proteolytic degradation could also contribute to an increased local concentration of Wg and these can not be ruled out at the present time. It will also be of interest to determine if there is one particular proteoglycan that acts as a Wg co-receptor or whether several different molecules are involved. In this regard, we note that syndecan mutants enhance the weak dsh phenotype similar to ska mutants while dally mutants (a Drosophila glypican homolog) do not (T. Heslip and J. L. Marsh unpublished). Thus, individual classes of proteoglycans may each have distinct roles in mediating signaling by different types of growth factors. Further study will be required to investigate which proteins are involved in the molecular mechanism of this effect.

Many vertebrate growth factors bind heparan sulfates and Drosophila expresses homologs of several of these including FGF, EGF, BMPs and Hh (Arora et al., 1994; Lee et al., 1992; Neuman-Silberberg and Schupbach, 1993; Padgett et al., 1987; Rijsewijk et al., 1987; Rutledge et al., 1992; Sutherland et al., 1996). It is therefore surprising that we do not see a more pleiotropic phenotype in ska mutants. The Wg pathway may be more sensitive to perturbations in GAG synthesis than other signaling pathways in Drosophila. An alternative explanation may be that Wg in the embryo is the earliest pathway to be affected and therefore involvement of GAGs in mediating later signaling events by other factors might be masked. Certainly this could be the case for Drosophila FGF signaling which is required for tracheal development (Sutherland et al., 1996). Tracheal branching and outgrowth in response to FGF signaling begins relatively late during the extended germ-band stage and continues through germ-band retraction and dorsal closure. By this stage, defects resulting from a lack of Wg signaling are quite pronounced and it is difficult to assess whether FGF is impaired or not.

We were also surprised by the absence of dpp-like pheno-types in ska mutants (e.g. no dorsal-ventral cuticle defects and normal amnioserosa). Likewise, expression of the Dpp responsive transcription factor spalt (de Celis et al., 1996) is normal in skaSG9/skaA31 mutant wing imaginal discs (data not shown) suggesting that Dpp signaling either does not require ska function or is much less sensitive to its loss. The recent demonstration that dally mutants interfere with Dpp signaling in the wing (Selleck, personal communication) suggests that the effect of dally on Dpp signaling may be mediated through the core protein rather than the GAG side chains as has been found to be the case for the effects of betaglycan on TGF-βsignaling (Lopez-Casillas et al., 1994).

Even for cytokine genes expressed early in embryogenesis, the requirement for ska and GAGs may differ. For example, EGF receptor mutants lack ventral ectoderm (Raz and Shilo, 1992, 1993) but strong ska mutants make ventral ectoderm suggesting that EGF signaling in the embryo may not be sensitive to loss of ska. A possible requirement for GAGs in later EGF signaling events in the imaginal discs involving the two EGF-like homologs, vein and spitz (Mason et al., 1994; Schweitzer et al., 1995) has not been directly examined. However, we do find that certain ubiquitin-ska transgenic lines only partially rescue ska mutants and also show wing phenotypes similar to those produced by mutations in the EGF signaling pathways (Fig. 2E).

The reason why ska mutations act as moderate suppressers of the activated Tkv phenotype is not clear. However, heterozygosity for EGF-receptor mutations can suppress the activated Tkv phenotype similar to the suppression seen with ska mutations (unpublished observations). Thus, one explanation for the suppression of the activated Tkv phenotype could be that EGF signaling is reduced in ska mutants sufficiently to suppress the activated thick veins phenotype but not enough to cause an EGF-like phenotype. Another possibility is that GAGs could directly influence Dpp receptor function. Consistent with this view, BMP-2 signaling is enhanced in a limb bud assay by addition of exogenous heparan sulfates even when the BMP-2 ligand is deleted for the heparan sulfate binding sequence (Ruppert et al., 1996). These authors suggest that the exogenous heparin might potentiate receptor activity perhaps by stabilizing receptor complexes or conformations. A third possibility is that GAG breakdown products have been implicated in direct modulation of some transcription factors and, in one recent case, in directly modulating a cell cycle control protein by binding to it (Grammatikakis et al., 1995). Thus, GAGs may be able to exert their functions inside the cell as well as outside. If so, lack of GAGs could potentially directly modulate intracellular signaling events.

Because of a mutually supportive positive feedback loop in the late germ-band-extended embryo, it is difficult to distinguish the role of GAGs in Wg versus Hh signaling at this stage (Martinez Arias, 1993). By-passing this loop by ectopic expression of one or the other factor does not resolve the two events because Wg signaling seems also to be required for Hh-induced expansion in wg expression (Hooper, 1995). Our studies of Arm accumulation and the late appearance of partial En stripes in ska mutants suggest that Wg signaling is greatly reduced but still partially functional in ska loss-of-function animals. Therefore, the lack of wg expansion in ska mutants after hh overexpression could be explained by the absence of sufficient Wg signaling to provide the autocatalytic requirement for Wg, or by a role for GAGs in Hh signaling, or by an additive effect of partial reduction in both Wg and Hh signaling. The reduction of dpp expression in the morphogenetic furrow of the eye disc while hh expression is readily detectable suggests some reduction of Hh signaling may be occuring in ska mutants. Interestingly, ska expression in the morphogenetic furrow is stronger than in the rest of the eye disc.

Although our data suggest a role for GAG chains in modulating growth factor signaling, other functions for GAGs during Drosophila development are also possible. GAG chains have been implicated in mediating numerous other biological processes most notably cell-cell and cell matrix interactions (Jackson et al., 1991). Remodeling of the matrix has long been speculated to be important for patterning since, depending on the components, the matrix could limit or enhance differential cell migration. The fact that gastrulation and germ-band extension appear normal in ska germ-line mutant embryos implies that, at least in Drosophila, these early cell movements do not require GAG polymers.

The lack of ska phenotypic effects in imaginal clones was surprising since we imagined that GAG chains might be required for Wg signaling. Several possibilities might account for this observation. Firstly, it is possible that the enzyme or the proteoglycans and/or GAGs are relatively stable and thus perdure long after clone induction. An alternative possibility is that UDP-glucuronate (or a downstream metabolite) can be exchanged between cells perhaps via gap junctions. Finally, it is formally possible that a proteoglycan coreceptor or recycled oligosaccarides could be passed between cells. Whatever the mechanism, it is striking that a defect in a general and fundamental step of GAG chain biosynthesis can have such specific effects on growth-factor-dependent developmental events in vivo.

The authors thank S. Park for injection of the P-element constructs and A. Bejsovec for the hs-wg and hs-hh lines. We are grateful to Udo Häcker, Xinhua Lin, Norbert Perrimon, Rich Binari and Armen Manokian for exchanging sequence and other data prior to publication. We note that ska is also called sugerless (Lin, Häcker and Perimon) and kiwi (Binari and Manokian) by these other groups. We thank members of the Marsh and O’Connor labs for helpful discussions and Heidi Theisen for critical reading of the manuscript. We thank Stephenie Paine-Saunders for the Syndecan antibody and advice on Syndecan blots. Use of the Optical Biology Shared Resource of the Cancer Center Support Grant (CA-62203) at University of California, Irvine is gratefully acknowledged. This work was supported by an NIH Research Program Project PO1 HD27173 to J. L. M. (P. J. Bryant director), by grants GM00599 and GM47462 from PHS to M. B. O., by a CRCC grant from the University of California and a grant from the Ciba-Geigy Jubiläumsstiftung to T. E. H. and by a Full-Time Fellowship from the Alberta Heritage Foundation for Medical Research to T.R.H. The authors gratefully acknowledge K. Matthews and the resources of the National Drosophila Stock Center in Bloomington, IN and appreciate the stocks and information on ska alleles received from Wayne Johnson.