Genetic evidence that heparin-like glycosaminoglycans are involved in wingless signaling

The Wnt genes encode secreted glycoproteins which are required for intercellular communication in a variety of diverse organisms (Nusse and Varmus, 1992; McMahon, 1992). Although the Wnt proteins are too large to be classified as growth factors, their various roles do qualify their classifica-tion as such (discussed by Klingensmith and Nusse, 1994). Two well-studied members of the Wnt gene family are the Drosophila segment polarity gene wingless (wg) and the murine Wnt-1 gene. Studies on these genes have focused on elucidating the mechanism of signal transduction elicited by Wnts (reviewed by Klingensmith and Nusse, 1994). Biochem-ical studies on Wnt-1 have demonstrated the paracrine nature of Wnt protein function and have shown a close association of Wnt-1 with the extracellular matrix (ECM) (Papkoff and Schryver, 1990; Bradley and Brown, 1990; Jue et al., 1992). These studies have also implicated a close association of Wnts with glycosaminoglycans (GAGs) found in the ECM (Bradley and Brown, 1990; Jue et al., 1992; Burrus and McMahon, 1995). Genetic studies on Wnt-1 in mice have also shown the importance of this gene during development (McMahon and Bradley, 1990; Thomas and Capecchi, 1990). Recently, bio-chemical studies on wg protein (Wg) have led to the identifi-cation of a receptor molecule for Wg (Bhanot et al., 1996). However, most information regarding the mechanism of Wnt signaling has come from genetic studies on wg in Drosophila.

It has been possible to identify additional genes in Drosophila which are involved in wg function by focusing on mutations that have phenotypes similar to that of wg. This category includes the dishevelled (dsh), porcupine (porc) and armadillo (arm) genes (Perrimon et al., 1989). The phenotype of the zeste-white 3 (zw3) gene appears to be opposite to that of wg, leading to its inclusion as a component of wg signaling (Siegfried et al., 1992). Genetic epistatis experiments have led to a model whereby porc acts upstream of wg and dsh acts downstream of wg (Noordermeer et al., 1994), while dsh acts upstream of zw3 and arm acts downstream of zw3 (Siegfried et al., 1994). These genes have been cloned in Drosophila and all have been shown to be well conserved throughout evolution. For example, zw3 is a homolog of the Glycogen synthase kinase-3 (GSK-3) family, while arm is a homolog of vertebrate β-catenin (McCrea et al., 1991). Therefore it is apparent that the Wnt signaling mechanism is well conserved in evolution (reviewed in Klingensmith and Nusse, 1994).

One biochemical aspect of Wnt function has been its close association with the ECM. It has been documented that Wnt-1 is tightly bound to the cell surface and/or the ECM but can be released readily via the exogenous addition of the GAG, heparin (Bradley and Brown, 1990; Papkoff and Schryver, 1990). The ECM consists of a variety of complex carbohy-drates such as GAGs, which are frequently linked covalently to a distinct class of glycoproteins called proteoglycans (Kjellan and Lindahl, 1991). Proteoglycans can bind to a variety of growth factors and may even function as co-receptors during signaling (Klagsburn and Baird, 1991). Despite these results, genetic studies linking GAGs and/or proteoglycans to Wnt or other growth factor function have been lacking. In fact, the importance of proteoglycans for Wnt signaling has come into question recently (Schryver et al., 1996) and it has been demonstrated that not all Wnts can bind GAGs (Burrus and McMahon, 1995).

In this study, we provide the first genetic demonstration of a functional role of GAGs in signal transduction. We report the identification and molecular analysis of a novel gene, kiwi, whose activity is critical for wg signaling. We show that kiwi encodes the Drosophila homolog of UDP-glucose dehydro-genase (UDP-GlcDH), an enzyme responsible for the produc-tion of UDP-glucuronate (Feingold and Franzen, 1981). UDP-glucuronate is an uronic acid which is a precursor in GAG biosynthesis (Lennarz, 1980). We also show that kiwi activity is required for efficient wg function and that the injection of either UDP-glucuronate or heparan sulfate (HS) into kiwi mutant embryos can rescue their phenotype. We further show that the injection of purified bacterial heparinase (an enzyme degrading heparin-like GAGs) into wild-type embryos can phenocopy the kiwi and wg mutant phenotype. Our results provide the first genetic evidence for the involvement of heparan sulfate in wg/Wnt function.

l(3)08310 or P1731 (P{ry[+t7.2]=PZ} l(3)08310 ry⁵⁰⁶ / TM3, ry [RK]) is from the Spradling collection of P-element lethals and maps cytologically to 65D4-5 (Spradling et al., 1995). P1731 P-element excisions were generated by selecting single dysgenic males of genotype P1731/ry e Sb P[ Δ2,3] from the cross between P1731/TM3 females and males of genotype w/Y; ry e Sb P[ Δ2,3]/TM6, Ubx e. Stocks were then established and scored for viability and loss of ry⁺. Df(3R)W5.4, A31, N71, XAJ36 and EX34 were generated in a previ-ously reported screen (Anderson et al., 1995). A31 and N71 are EMS-induced, while XAJ36 and EX34 are X-ray-induced alleles of l(3)08310, since they fail to complement the lethality of l(3)08310. Complementation was confirmed by crossing GLC females of kiwi^P1731 to Df(3R)W5.4, A31, N71, XAJ36 and EX34 in order to generate the kiwi GLC phenotype. In order to perform GLC analysis, FRT^2A was recombined on each mutant chromosome (P1731, A31 and N71). We were unable to make such a recombinant with XAJ36 whereas GLC analysis with EX34 produced only sterile females. GLC analysis with FRT^2A was performed as described by Chou and Perrimon (1996). Initial experiments were performed with marked embryos and cuticles to determine the extent of zygotic rescue of the kiwi maternal phenotype. Embryos were marked by mating kiwi GLC females to males carrying mutant alleles over the TM3 (evelacZ) balancer chromosome. Cuticles were generated in a yellow (y) mutant background, and the balancer chromosome was TM3, y⁺. prdGAL4 insertion on the X chromosome was generated by mobilizing RG1 (Yoffe et al., 1995) using the w; Sb P[ Δ2,3, ry+] (99B) e chromo-some (Robertson et al., 1988). For epistasis experiments with kiwi, prdGAL4/+; kiwi GLC females were mated to UASwg; kiwi/TM3 males. These epistasis experiments were performed using P1731, A31 and N71 with all experiments yielding identical results. To generate wg^IG22 (prd-wg) embryos, prdGAL4; wg^IG22/CyO females were mated to UASwg, wg^IG22/CyO males. wg^IG22 is a null allele of wg and has been used before in similar experiments (Manoukian et al., 1995). For injection experiments, y kiwi GLC embryos were generated from the following cross: y; kiwi^A31 GLC females × y; kiwi^A31/TM3, y⁺ males.

For the generation of adult somatic clones in the wing, the following cross was used: yw HsFLPase; kiwi^P1731FRT^2A/TM3 × yw; P[y⁺] FRT^2A/TM3. Larvae from this cross were heat-shocked for 2 hours at 37°C at early crawling stage (72 hours after hatching). Adult wings were dehydrated and then mounted in Euparol^TM for observation.

After collection on apple juice-agar plates, embryos were plated onto a fresh plate and aged at 25°C. Unhatched embryos were dechorion-ated and then cleared in Hoyer’s/15% lactic acid medium for 24 hours (Wieschaus and Nüsslein-Volhard, 1986). Preparations were then flattened before analysis and photography. To process injected cuticles, preparations were briefly rinsed in n-heptane before mounting in Hoyer’s/15% lactic acid.

In situ hybridization with digoxigenin (DIG)-labeled en and wg probes was performed as previously described (Manoukian and Krause, 1992). DIG-labeled kiwi probe was generated from clone 51b using the same method.

Overlapping phage were isolated from a Drosophila genomic library made in bacteriophage EMBL3 (Blackman et al., 1987). Plaque hybridizations, DNA purification and cloning, and Southern blot analysis were done as described by Sambrook et al. (1989). Probe DNAs were labelled with ³²P by the random priming method (Feinberg and Vogelstein, 1983), and unincorporated nucleotides were removed using a G-50 Sephadex column. The probe used as an entry into the region was a 120 bp PCR fragment amplified from wild-type genomic DNA using the following primers:

5′CGCTCGGAGAAATCTGACT3′

5′CAGCCTTGTTGTTGTTGTTTT3′.

The sequence of these primers was deduced from that of the STS (sequence tagged site) associated with the P-element insert in the kiwi allele P1731, which was provided by the Berkeley Drosophila Genome Project. The PCR conditions for amplification were 35 cycles of: 94°C, 45 seconds; 60°C, 45 seconds; 72°C, 2 minutes.

Putative kiwi cDNAs were isolated from a 0-to 4-hour Drosophila embryonic cDNA library (Brown and Kafatos, 1988) using the 3.5 kb genomic HindIII fragment as probe (Fig. 4). Five positives were isolated from an initial screening of approximately 500,000 colonies. The longest clone (51b) was chosen and mapped back to genomic DNA from the walk phage to ensure that it derived from the correct region of the genome. This clone was then subjected to DNA sequence analysis (Amgen ETS program).

Clone 51b was linearized with NotI, and this template was then used to generate a capped sense transcript with the SP6 mMessage mMachine Kit (Ambion) as per the manufacturer’s instructions. Briefly, approximately 1 μg of linearized 51b was transcribed in vitro with SP6 polymerase at 37°C. After termination of the reaction and ethanol precipitation, the total resulting RNA was resuspended in 50 μl of injection buffer (5 mM KCl, 0.1 mM NaPO₄, pH 6.8) and stored at −70°C until use.

For injection experiments, precellular blastoderm embryos (0-to 1.5-hour old embryos) were dechorionated and lined up on double-sided tape and covered with halocarbon oil. At least 200 embryos were injected for each experiment. For 51b mRNA injection, the RNA was injected without further dilution. For protein or glucuronate injections, the volume of injection was held constant while the concentration of the material being injected was varied. In these manipulations, a com-promise was sought where most injected embryos developed to cuticle deposition. These conditions were arrived at for each experiment. Following injections, embryos were allowed to develop at 25°C and were then lifted off the tape and onto apple juice-agar plates. At this point, non-developed embryos were discarded and all other embryos were first rinsed with n-heptane (to remove excess halocarbon oil) and then mounted into Hoyer’s/15% lactic acid medium for analysis and observation. Each experiment was repeated at least three times and the results were reproduced every time. The results of single experi-ments are depicted below.

All material to be injected was dissolved and/or diluted in injection buffer (5 mM KCl, 0.1 mM NaPO₄, pH 6.8). kiwi mutant embryos were marked with y and rescue was deemed successful when signifi-cant naked cuticle was added back into y kiwi mutant embryos – partial rescue was scored as negative. The injection of 51b effectively rescued the kiwi mutant embryos at a frequency of 19/38. UDP-glu-curonate injection was effective at 1 mg/ml concentration resulting in a rescue frequency of 12/33. Heparan sulfate rescue was also effective at 1 mg/ml concentration, but we did notice that most of the cuticles were becoming converted to the naked phenotype (see below). As we focused on the y embryos, we had to remove a significant population of embryos which had the fully penetrant naked phenotype and thus could not be scored. Therefore this rescue frequency was only 8/17.

Expression and purification of heparinase I and III was as has been previously published (Sasisekharan et al., 1993; Ernst et al., 1996). Chondroitinase ABC from Proteus vulgaris was from Seikagaku Corporation (Ijamsville, MD). This enzyme can degrade chondroitin 4-sulfate, chondroitin 6-sulfate, dermatan sulfate and hyaluronic acid (manufacturer’s specifications). Heparinase I, III, chondroitinase ABC and heparan sulfate were injected into w¹¹¹⁸ embryos. Heparinase I was at a concentration of 0.5 mg/ml. With injection of heparinase, 23/44 embryos showed significant fusion of denticles and 11 of the 23 showed the severe wg phenotype. The injection of heparinase III was effective at 0.55 mg/ml and resulted in a wg-like phenotype at a frequency of 27/49. Injection of chondroitinase ABC was varied all the way up to a concentration of 1 mg/ml. Injection of heparan sulfate into w¹¹¹⁸ embryos was at 1 mg/ml (as in above) and denticles were converted to naked cuticle at a frequency of 11/36, with 7 embryos showing severe and fully penetrant effects. The data from these injection experiments are tabulated in Table 1 and depicted in Figs 7 and 8.

Although the wg phenotype is zygotic, other wg signaling pathway genes are also expressed maternally and thus their phenotypes have to be generated via the production of germ line clone (GLC) females (Perrimon et al., 1989). We have identified P1731 or l(3)08310 (Spradling et al., 1995) as having a GLC cuticle phenotype which resembles that of wg (Fig. 1B) and we have named this locus kiwi. This GLC phenotype is zygotically (paternally) rescuable, but not to wild-type. Zygot-ically rescued kiwi^P1731 embryos have weak segmental fusions in the ventral portion of the embryo and are often missing dorsal cuticle pattern elements (Fig. 7B). Excision experiments of P1731 were used to lose the rosy⁺ phenotype and generated viability at a frequency of 10/40. The results of this experiment convinced us that the lethal mutation in l(3)08310 was a result of P-element insertion. We have identified four additional kiwi alleles (Anderson et al., 1995), two of which (A31 and N71) we have also used in GLC experiments. In all experiments kiwi^P1731 behaved as the strongest allele, although the severity of the phenotype compared to A31 and N71 was rather com-parable (see Materials and Methods). The GLC phenotype of kiwi^P1731 is identical to the wg phenotype (Fig. 1B) and was not intensified over a deficiency of the region (Df (3R) W5.4) (not shown); we therefore believe that kiwi^P1731 is a null allele.

The expression of the engrailed (en) gene has been shown to be under the control of wg activity in the embryo (DiNardo et al., 1988; Martinez-Arias et al., 1988). In wg mutant embryos, en transcripts fade during stages 9-10 (Manoukian et al., 1995). In kiwi GLC embryos, en expression is lost during stages 10-11 (Fig. 1D). It appears that, although wg and kiwi GLC embryos have identical cuticle phenotypes, en expression persists longer in kiwi GLC embryos. Therefore, en expression is more resistant to the loss of kiwi activity than wg activity. In wg mutants, wg transcription is lost during stages 9-10 before the loss of en expression in the same embryos (Manoukian et al., 1995). In kiwi GLC embryos, wg expression is also lost before en during stage 10 (Fig. 1F). However, as in the case of en transcription, wg transcripts are lost earlier in wg mutants compared to kiwi mutants.

As wg expression is gone relatively early (stage 10) in kiwi mutants, we decided to test the activity of ectopic wg in these mutants by using the prdGAL4 transgene to activate UASwg in alternate segments (Brand and Perrimon, 1993; Yoffe et al., 1995; Manoukian et al., 1995). In wg mutant embryos, prdGAL4/UASwg (prd-wg) can rescue wg activity (naked cuticle and en expression) in alternate segments (Fig. 2A,C). In kiwi mutants, prd-wg results in a minor rescue of the kiwi cuticular phenotype but cannot specify naked cuticle (Fig. 2B). en expression is also rescued by prd-wg in kiwi mutants (Fig. 2D), although this effect is weak. These experiments suggest that kiwi is not upstream of wg, but do suggest that wg can have some activity in the absence of kiwi. Since hedgehog (hh) mutants have a phenotype similar to that of kiwi and wg, this opens up the possibility that kiwi functions in hh signaling and thus affects wg indirectly (Manoukian et al., 1995). As ectopic wg is fully functional in the absence of hh and smoothened activity (van den Heuvel and Ingham, 1996), our epistasis results with prd-wg suggest that kiwi is involved in wg signaling directly as opposed to being an exclusive component of hh signaling.

Further evidence to support a direct role for kiwi in wg signaling comes from the fact that kiwi somatic clones result in loss of anterior margin bristles of the adult wing (Fig. 3B), which have been shown to be specified by wg (Couso et al., 1994). This wing margin phenotype also gives us the opportu-nity to test the autonomy of kiwi function. In the wg pathway, wg and its upstream component porc function non cell-autonomously, whereas the downstream components (dsh, zw3 and arm) are cell-autonomous in mosaics (Klingensmith and Nusse, 1994). Making yellow marked clones of kiwi demon-strates that kiwi functions non cell-autonomously (Fig. 3C), suggesting that kiwi is at least genetically upstream of dsh.

To clone the kiwi locus, a chromosomal walk in the 65D region was initiated using as an entry point the STS (sequence tagged site) generated from the kiwi P-element-induced allele P1731 by the Berkeley Drosophila Genome Project. Overlapping phage clones were isolated that span approx. 30 kb of wild-type genomic DNA. Genomic fragments that hybridized to the STS were sequenced and used as probes to screen a 0-to 4-hour embryonic cDNA library. The longest clone (designated 51b) was chosen for further analysis. Southern blot analysis indicates that the cDNA extends over approx. 5 kb of genomic DNA (Fig. 4). Sequence analysis of genomic DNA and com-parison to that of the cDNA indicates that the P element asso-ciated with P1731 has inserted into the 5′ untranslated region of the transcription unit. A developmental northern analysis using 51b as probe detects a single transcript of approx. 2.4 kb present at most stages of development, including 0-to 2 hours (data not shown). In situ hybridization to wild-type whole-mount embryos with 51b as probe shows that kiwi transcripts are present uniformly throughout the embryo at blastoderm stage (Fig. 5A), and subsequently become localized to specific cephalic regions and to the presumptive amnioserosa beginning at stage 9 (Fig. 5B). This expression in the amnioserosa persists well into stage 14 of embryogenesis (not shown). To demon-strate that 51b encodes kiwi function, an SP6-generated tran-script corresponding to the sense strand of 51b was injected into kiwi germline clone-derived embryos, where it led to sig-nificant rescue of the kiwi phenotype (Table 1; Fig. 7C).

The nucleotide sequence of 51b (2343 bp) encodes a con-ceptualized open reading frame beginning with an ATG at position 315 and extending 1428 nucleotides to position 1742, thus encoding a putative protein of 476 amino acids. Stop codons in all three reading frames follow the open reading frame. The nucleotides immediately upstream of the putative initiation codon (CGAA) are a reasonably good match to the consensus sequence for Drosophila translation start sites [(C/A)AA(C/A)] (Cavener, 1987). The open reading frame is preceded by an untranslated region of 215 nucleotides and followed by an untranslated region of 601 nucleotides containing the consensus polyadenylation recog-nition sequence AATAAA (Birnstiel et al., 1985) 16 bases from the 3′ end of the cDNA.

A homology search of the available databases using the BLAST algorithm (Altschul et al., 1990) indicates that kiwi encodes the Drosophila homolog of uridine diphosphate glucose dehydrogenase (UDP-GlcDH) (Dougherty and van de Rijn, 1993; Hempel et al., 1994). The sequence with maximum homology is bovine UDP-GLcDH with an overall identity of 68% (see Fig. 6). UDP-GlcDH catalyzes the reaction UDP-glucose + 2NAD⁺ + H2O to yield UDP-glucuronic acid + 2NADH + 2H⁺ (Dougherty and van de Rijn, 1993). To show that the kiwi GLC phenotype is due to the loss of UDP-GlcDH activity, we attempted to rescue the kiwi phenotype by injecting UDP-glucuronic acid into kiwi GLC embryos. Since this leads to the rescue of the kiwi phenotype (Table 1; Fig. 7D), we suggest that UDP-glucuronic acid (or UDP-glucuronate) pro-duction in the embryo is essential for wg signaling.

UDP-glucuronate is one of the uronic acids incorporated into several complex polysaccharides or glycosaminoglycans (GAGs), including hyaluronic acid, chondroitin sulfate, heparin and heparan sulfate. Reichsman et al. (1996) have recently shown the involvement of GAGs in wg signaling in tissue culture cells, consistent with studies on Wnt-1 which had implicated the involvement of heparin in Wnt localization and function (Bradley and Brown, 1990; Jue et al., 1992; Burrus and McMahon, 1995). Given this background information of the potential roles for heparin-like and/or chondroitin sulfate GAGs in wg function, we injected separately, purified hepari-nases I and III (degrading heparin-like GAGs) and chondroiti-nase ABC (degrading chondroitin sulfates, hyaluronic acid and dermatan sulfate) into wild-type embryos to determine if either or both enzymes had any effect on wg signaling. We found that injection of heparinases I (Table 1; not shown) and III (Table 1; Fig. 8A) into wild-type embryos can mimic the kiwi and wg phenotypes, whereas injecting chondroitinase ABC does not mimic the wg phenotype (unpublished observations; see Materials and Methods). This suggests that heparin-like (heparin and heparan sulfate) GAGs are involved in wg signaling. As heparinase III is much more specific to heparan sulfate (Ernst et al., 1995), we attempted to rescue the kiwi phenotype by injecting exogenous heparan sulfate into kiwi GLC embryos. Since heparan sulfate was able to rescue the kiwi phenotype (Table 1; Fig. 8B), we suggest that the loss of UDP-GlcDH activity in kiwi embryos reflects a requirement of heparin-like GAGs and, perhaps more specifically, heparan sulfate in wg function. Furthermore we find that injection of excess heparan sulfate into wild-type embryos results in lethality and a ‘naked’ cuticle phenotype (Table 1; Fig. 8C), which is identical to that of Wg over-expression in embryos (Noordermeer et al., 1992). Thus it appears that excess heparan sulfate in embryos results in hyperactivity of Wg, consistent with the in vitro results obtained by Reichsman et al. (1996).

We have shown that the kiwi gene is required for full wg function in embryos and at the wing margin. We have also shown that kiwi encodes the Drosophila homolog of UDP-GlcDH, an enzyme involved in the biosynthesis of UDP-glu-curonate. UDP-glucuronate is further utilized in the biosyn-thesis of complex polysaccharides such as hyaluronic acid, chondroitin sulfate, heparin and heparan sulfate. This fact implicates all of these polysaccharides as having a role in wg function. Since chondroitinase ABC can degrade a wide range of GAGs such as hyaluronic acid, chondroitin sulfate and dermatan sulfate (Ernst et al., 1995), our inability to generate a wg phenotype by injecting chondroitinase ABC (not shown) suggests that the loss of only heparin-like GAGs results in the wg-like phenotype observed in kiwi mutants. This idea is fortified by our ability to rescue the kiwi phenotype with excess heparan sulfate.

We have demonstrated a crucial role for heparin-like GAGs in wg function by four criteria. First, the injection of heparinase III (Fig. 8A) and not chondroitinase ABC (not shown) leads to phenocopy of the wg (and also kiwi) phenotype. Second, and more specifically, injection of excess heparan sulfate leads to rescue of the kiwi phenotype (Fig. 7). Third, the effect of expression of ectopic wg is partially suppressed in kiwi embryos (Fig. 2). Finally, injection of heparan sulfate into wild-type embryos results in excess Wg signaling and a ‘naked’ phenotype (Fig. 8C). Our results are therefore consis-tent with those of Reichsman et al., (1996) which demonstrated a role for GAGs in wg signaling in tissue culture cells. In this study, we have been able to provide the first genetic evidence for the requirement of GAGs in Wnt function. Furthermore, we have specifically established a role for heparan-like GAGs in wg signaling. Our ability to rescue kiwi with exogenous heparan sulfate suggests that heparan sulfate is the crucial moiety in wg function. Heparan sulfate is similar to heparin except for a lower degree of sulfation and it contains a higher percentage of glucuronate residues consistent with the role of UDP-GlcDH (Ernst et al., 1996). Another important fact to note is that heparin is largely concentrated in the intracellular portion of cells, whereas heparan sulfate is extracellular.

Genetic mosaic experiments show that kiwi acts non cell-autonomously (Fig. 3), consistent with kiwi encoding UDP-GlcDH. As heparan sulfate is localized to the ECM, the UDP-GlcDH function would reflect non cell-autonomy. Since kiwi has a GLC phenotype which is zygotically rescuable (Fig. 7B), it is logical to assume that maternal kiwi expression (Fig. 5) has a crucial role in wg signaling. kiwi expression is at first uniform and persists until cellularization of blastoderm, gradually becoming restricted during gastrulation (Fig. 5). We therefore propose that maternal and pre-cellular blastoderm expression of kiwi generates enough UDP-glucuronate to be incorporated into the ECM as it is synthesized before cellular-ization. Once the ECM is formed, kiwi expression is not uniform and becomes restricted to specific tissues where it may be required for similar or perhaps different functions. Interest-ingly, the zygotic phenotype of kiwi shows extensive loss of dorsal cuticle structures and perhaps the amnioserosa (not shown), and may reflect the high level of kiwi expression in these tissues during gastrulation. The GLC phenotype of kiwi is identical to the wg phenotype and may only reflect the earliest crucial role for kiwi function. UDP-glucuronate may be required for many other processes which are either required later on in development or whose phenotypes are masked by the severe ‘wg-like’ phenotype apparent in kiwi GLC embryos.

Heparin-like GAGs are complex acidic polysaccharides that are important soluble components of the ECM. They provide the necessary hydration to the ECM scaffold and hence solu-bilize and modulate the function of several of the transient components of the matrix such as growth factors (Kjellen and Lindahl, 1991). The role of heparin-like GAGs in regulating the biological activities of several heparin-binding growth factors has garnered attention for some time. For example, heparin-like GAGs are thought to tether fibroblast growth factor (FGF) in the ECM so as to create a reservoir of FGF (reviewed by Jackson et al., 1991). In one model for FGF-GAG interaction, stable FGF-heparin complexes are released, diffuse and activate FGF receptor at a site distant from the release site (Jackson et al., 1991). A different model (the dual receptor model) suggests that heparin-like GAGs bind FGF and ‘present’ this ligand to the FGF receptor. Specifically, it is thought that heparin-like GAGs participate in a dual receptor system as low-affinity receptors to facilitate the interaction of ligands with their high-affinity receptors (Klagsbrun and Baird, 1991; Schlessinger et al., 1995). It has also been suggested that heparin binding to FGF causes FGF oligomerization leading to clustering of FGF receptors for signal transduction (Ornitz et al., 1992).

wg protein (Wg) has been shown to bind GAGs in vitro, suggesting that GAG moieties in the ECM sequester Wg (Reichsman et al., 1996). This interaction appears to be crucial for cells that receive Wg and not those that produce it (Reichsman et al., 1996). We propose that the extracellular GAG, heparan sulfate, physically interacts with Wg in cells which receive the Wg signal. The tight binding of Wg to heparin-like GAGs could regulate the stability, diffusion and effective concentration of the Wg reservoir in the ECM, while generating a gradient by limiting the diffusion of Wg (Lawrence and Struhl, 1996). This aspect is particularly intriguing in light of the finding that different threshold concentrations of Wg elicit distinct responses to organize cellular pattern during mor-phogenesis (Zecca et al., 1996). Alternatively, reception of the Wg signal could be regulated by heparan sulfate. With the recent identification of a Wg receptor (Bhanot et al., 1996), testing of models for regulation of Wg-receptor interactions by heparan sulfate can now be undertaken.

We thank the Berkeley Drosophila Genome Project for providing us with the STS data. We also thank Fabrizio Mastronardi for help with our graphic figures and Vuk Stambolic for comments on the man-uscript. We thank Ron Blackman for the EMBL 3 genomic library, and Nick Brown for his 0-4 cDNA library. Howard Lipshitz gener-ously allowed us to use his micro-injection apparatus and Sam Scanga provided us with a developmental northern blot. We also thank Mike O’Connor for communication of results. The sequencing for this project was performed by the Amgen ETS program and we thank Tak Mak for his generous support. R. B. and B. S. were Amgen/OCI post-doctoral fellows and this work was supported by grants from the Connaught Research Fund and the Medical Research Council of Canada to A. M.

Note: The sequence of 51b has been submitted to GenBank; accession number AF 000570.