Functions of heparan sulfate proteoglycans in cell signaling during development

During metazoan development, the formation of complex body structures and patterns is governed by several secreted signaling molecules, including members of the Wnt/Wingless (Wg), Hedgehog (Hh), transforming growth factor-β (TGFβ) and fibroblast growth factor (FGF) families. Over the past decades, intensive biochemical and genetic studies have elucidated the central components of the signaling pathways that these molecules function in. More recently, attention has shifted to understanding the mechanisms by which the distributions of these signaling molecules are regulated in morphogenetic fields. Studies in Drosophila and vertebrates have demonstrated the crucial roles of heparan sulfate proteoglycans (HSPGs) in these signaling pathways during development. This review focuses on recent insights into the functions of HSPGs in regulating the activities and distributions of these signaling molecules. For detailed reviews of previous biochemical and genetic studies on the HSPGs, please see(Belting, 2003; Bernfield et al., 1999; Esko and Selleck, 2002; Lin and Perrimon, 2003; Nybakken and Perrimon, 2002; Princivalle and de Agostini,2002; Selleck,2001; Song and Filmus,2002).

HSPGs are cell-surface and extracellular matrix (ECM) macromolecules that comprise a core protein to which heparan sulfate (HS) glycosaminoglycan (GAG)chains are attached (Bernfield et al.,1999; Esko and Selleck,2002). HSPGs are classified into several families based on their core protein structure (Fig. 1). Glypicans and Syndecans are two major cell surface HSPGs, and are linked to the plasma membrane by a glycosylphosphatidylinositol (GPI)linkage or a transmembrane domain, respectively. Perlecans are secreted HSPGs that are mainly distributed in the ECM. Although Glypicans and Perlecans exclusively bear HS GAG chains, Syndecans are decorated with both HS and chondroitin sulfate (CS). All three families of HSPGs are evolutionarily conserved from vertebrates to Drosophila and C. elegans(Esko and Selleck, 2002; Nybakken and Perrimon, 2002). The Drosophila genome encodes four HSPG homologs: a single Syndecan(Sdc) (Johnson et al., 2004; Spring et al., 1994; Steigemann et al., 2004), two Glypicans [Division abnormally delayed (Dally) and Dally-like protein (Dlp)(Baeg et al., 2001; Khare and Baumgartner, 2000; Nakato et al., 1995)], and a Perlecan [Terribly reduced optic lobes (Trol)(Datta, 1995; Voigt et al., 2002)].

To date, most HSPG studies have demonstrated the importance of their HS chains. HS chains are polysaccharides synthesized in the Golgi apparatus and contain repeating disaccharide units of uronic acid linked to glucosamine(Esko and Selleck, 2002). HS chain biosynthesis is initiated at the GAG attachment site(s) of the core protein, which contains two to four Ser-Gly sequences. As depicted in Fig. 2, various glycosyltransferases and modification enzymes are involved in the polymerization and modification processes of HSPG biosynthesis. These enzymes are conserved in Drosophila and vertebrates(Esko and Selleck, 2002; Lin and Perrimon, 2003; Nybakken and Perrimon, 2002). In recent years, some of these enzymes have been genetically and biochemically characterized in Drosophila (see Table 1).

Wnts are secreted signaling molecules that function in numerous developmental processes in vertebrates and invertebrates(Wodarz and Nusse, 1998). Through the receptors of the Frizzled (Fz) family, Wnts can transduce their signals by a β-catenin/Armadillo-dependent pathway, called the`canonical' Wnt pathway (Wodarz and Nusse,1998). Wnts can also relay their signals via the planar cell polarity (PCP) pathway or Wnt/Ca<sup>2+</sup> pathway, the so-called`non-canonical' pathways (Veeman et al.,2003). Wnt proteins are lipid-modified, and are tightly associated with the cell surface and the ECM (Nusse,2003). Therefore, how Wnt proteins move from cell to cell to initiate intercellular signaling is of particular interest.

Drosophila wg encodes a homolog of vertebrate Wnt1, and its function has been well explored in the patterning of the fly embryonic epidermis (see Box 1) and of the larval wing imaginal disc (see Box 2), where Wg acts as both a short-range inducer and a long-range morphogen (Tabata and Takei,2004; Teleman et al.,2001; Vincent and Dubois,2002).

The function of HSPGs in Wg signaling was first revealed by the identification of sugarless (sgl)(Binari et al., 1997; Hacker et al., 1997; Haerry et al., 1997) and sulfateless (sfl) (Lin and Perrimon, 1999) as segment polarity genes in Drosophila. sgl and sfl mutant embryos develop cuticle defects that are similar to those in Wg or Hh signaling mutants(Box 1), indicating that they function in Wg and/or Hh signaling. Other evidence has also implicated the involvement of HSPGs in Wg signaling. First, several Wg-dependent embryonic developmental processes are abnormal in sgl or sfl mutant embryos (Hacker et al., 1997; Lin and Perrimon, 1999). Second, Wg target-gene expression and extracellular Wg levels are reduced in sfl mutant clones in the wing disc(Baeg et al., 2001; Lin and Perrimon, 1999). Third, in vitro biochemical experiments have supported these findings(Reichsman et al., 1996). Finally, other segment polarity genes, including fringe connection(frc) and slalom, are also involved in HS biosynthesis, and their studies have provided further evidence for the involvement of HS chains in Wg signaling and distribution (Goto et al., 2001; Luders et al.,2003; Selva et al.,2001) (see Table 1for more).

The Drosophila EXT proteins, Tout-velu (Ttv), Sister of ttv (Sotv)and Brother of ttv (Botv), are also involved in Wg signaling and distribution(Bornemann et al., 2004; Han et al., 2004a; Perrimon et al., 2004; Takei et al., 2004) (see Box 3, Fig. 2 and Table 1). Although previous studies had shown that Ttv is required specifically for Hh signaling(The et al., 1999) (as discussed later), the more recent careful analyses of the phenotypes of ttv, sotv and botv mutants have indicated that all three proteins are required for normal Hh, Wg and Dpp functions during wing development (Bornemann et al.,2004; Han et al.,2004a; Takei et al.,2004). Consistent with their expected roles in HS GAG biosynthesis, mutations in any of these genes lead to striking reductions in HS levels (Bornemann et al.,2004; Han et al.,2004a; Takei et al.,2004; Toyoda et al.,2000a; Toyoda et al.,2000b). These mutations also lead to reduced extracellular Wg protein levels and to the reduced expression of Distal-less(Dll), a Wg long-range target gene(Wodarz and Nusse, 1998). However, these proteins do not function redundantly in HS GAG biosynthesis, as ttv expression can only rescue the embryonic cuticle defect that is associated with the ttv mutant, and not those of the sotvand botv mutants (Han et al.,2004a). Together, these data suggest that all three EXT proteins are required for Wg long-range signaling, possibly by modulating extracellular Wg levels (Bornemann et al.,2004; Han et al.,2004a; Takei et al.,2004) (see Box 3).

Interestingly, Ttv, Sotv and Botv modulate Wg functions to varying degrees in the wing disc (Han et al.,2004a). Extracellular Wg levels are more greatly reduced in botv mutant clones than in either ttv or sotvclones. Moreover, the levels of senseless (sens), a Wg short-range target gene, are strikingly reduced in botv mutant clones and ttv-sotv double mutant clones, but not in ttv or sotv mutant clones. These data suggest that Ttv and Sotv act redundantly in Wg signaling, but that both are required for Wg distribution(Han et al., 2004a). Based on their observations, Han et al. have proposed that HSPGs have distinct roles in Wg distribution and signaling (Han et al.,2004a) (see Box 3for more).

However, another study has shown that both Ttv and Sotv are required for Wg short-range signaling (Bornemann et al.,2004). This study showed that the expression of achaete(ac), another Wg short-range target gene, is reduced in sotvor ttv mutant cells in the wing disc(Bornemann et al., 2004). The discrepancy between this and the findings of Han et al. could be due to the different sensitivity of ac and sens to Wg signaling. For example, higher levels of Wg signaling might be needed to activate acexpression than to activate sens expression, causing acexpression to be reduced in sotv or ttv mutant clones in the wing disc. Nevertheless, the effects of sotv and ttvmutations on Wg signaling and its distribution are relatively weak compared with those in ttv-sotv double or botv single mutants(Han et al., 2004a),indicating that some residual HS activity remains in the absence of ttv or sotv (see Box 3). Analyses of ttv, sotv and botv mutants in other developmental processes will further clarify this issue.

Which HSPG core proteins are involved in Wg signaling? Both Dally and Dlp,two Drosophila Glypicans, function in Wg signaling(Baeg et al., 2001; Lin and Perrimon, 1999; Tsuda et al., 1999). Disruption of Dally function by mutations in the dally locus or by its RNA interference (RNAi)-mediated knockdown results in weak segment polarity defects, suggesting that Dally may be required for Wg signaling in the embryonic epidermis (Lin and Perrimon,1999; Tsuda et al.,1999). However, embryonic cuticle defects associated with dally mutant or dally RNAi embryos are relatively weak compared with those of sgl or sfl mutants. Subsequent studies have demonstrated that dlp RNAi embryonic segment polarity defects are stronger than those of dally RNAi knockdown embryos. dally-dlp RNAi embryos have more severe segment polarity defects than dally RNAi or dlp RNAi embryos(Baeg et al., 2001). These data suggest that both Dally and Dlp are required for Wg signaling in the embryonic epidermis (Baeg et al.,2001).

However, a recent study by Desbordes and Sanson suggests that Dally and Dlp, separately or together, are not necessary for Wg signaling in the embryonic epidermis (Desbordes and Sanson,2003; Perrimon et al.,2004). Using dally RNAi against relatively small regions of dally, they showed that in dally knockdown embryos only very weak denticle fusion is present, similar to that seen in control embryos injected with buffer. Furthermore, the ectopic expression of wg in dally-dlp RNAi embryos can rescue engrailed (en)expression, a Wg target gene, and can generate naked cuticle, whereas ectopic expression of hh in dlp RNAi embryos fails to rescue wg expression, which is dependent on Hh signaling(Desbordes and Sanson, 2003). On the basis of these data, the authors argue that both Dally and Dlp are not required for Wg signalling, and that Dlp is strictly required for Hh signaling(Desbordes and Sanson, 2003; Perrimon et al., 2004). Whether Dlp is involved solely in Hh, or in both Hh and Wg, signaling needs to be resolved using dlp or dally-dlp null embryos. It is possible that Dally and Dlp may only be involved in controlling Wg protein retention or stability, but not in Wg signaling per se in the embryonic epidermis, as suggested by Pfeiffer et al.(Pfeiffer et al., 2002).

Relatively compelling evidence for the functioning of Dally and Dlp in Wg signaling has been demonstrated in the fly wing disc, where Hh and Wg signaling function independently from each other. Dally mutants exhibit wing margin defects and show genetic interactions with Wg signaling components, suggesting that Dally is required for Wg signaling in the wing disc (Lin and Perrimon, 1999). Interestingly, overexpressing dlp, but not dally, can strikingly increase extracellular Wg protein levels, suggesting that Dlp might bind to Wg with a higher affinity than Dally does(Baeg et al., 2001; Strigini and Cohen, 2000). However, overexpressing Dlp reduces, rather than enhances, Wg signaling. This could be because an excess amount of Dlp may compete with a limiting amount of Wg protein for its receptors Fz and Dfz2 (Fz2 – FlyBase), thereby inhibiting Wg signaling (Baeg et al.,2001). Whether Dlp plays a positive or a negative role in Wg signaling remains to be determined by clonal analysis in available dlp null mutants (Han et al.,2004b). Notum, which encodes a member of the alpha/beta-hydrolase superfamily, can also repress Wg signaling, possibly by modifying Dlp and/or Dally (Gerlitz and Basler,2002; Giraldez et al.,2002), thus providing further evidence for the involvement of Dally and Dlp in Wg signaling.

Several recent studies have demonstrated the essential role of HSPGs in Wnt signaling during vertebrate development(De Cat et al., 2003; Ohkawara et al., 2003; Topczewski et al., 2001). In Zebrafish, knypek encodes a glypican that is essential for convergent extension movement during gastrulation(Topczewski et al., 2001). Genetic analyses have demonstrated that Knypek is required for the signaling activity of Wnt11, which acts via the non-canonical PCP pathway during vertebrate gastrulation (Topczewski et al., 2001). Similarly, in Xenopus, reducing glypican 4(Xgly4) disrupt cell movements during gastrulation(Ohkawara et al., 2003). Xgly4 physically interacts with Wnt11, and might function in the Wnt/PCP pathway(Ohkawara et al., 2003). Interestingly, vertebrate glypican 3 can influence both canonical and non-canonical Wnt signaling in in vitro cell culture assays(De Cat et al., 2003), and blocking glypican 3 with antisense morpholino oligonucleotides disrupts gastrulation movements in zebrafish (De Cat et al., 2003). Glypican 3 is processed by a furin-like convertase (an enzyme involved in endoproteolytic processing), which is required for glypican 3 modulation of Wnt signaling both in vitro and in vivo(De Cat et al., 2003). Collectively, these data demonstrate that glypicans are required for Wnt signaling during vertebrate development. Whether other glypicans also require processing by a convertase(s), and whether a similar process exists in Drosophila, remains to be resolved.

One important finding from HSPG studies in vertebrates has been the role of sulfatases in Wnt signaling. Sulfatases modulate the sulfation states of HS chains by removing sulfates from them. As previously discussed, the sfl-encoded sulfotransferase in Drosophila is crucially required for proper Wg signaling (Lin and Perrimon, 1999). It is therefore surprising that QSulf1, the avian homolog of the HS-specific, N-acetyl glucosamine sulfatases, can promote Wnt signaling in myoblasts (Dhoot et al., 2001). In vertebrates, both Sulf1 and its highly related protein Sulf2 are secreted, HS-specific, 6-O endosulfatases(Ai et al., 2003; Morimoto-Tomita et al., 2002; Ohto et al., 2002). A recent study has shown that QSulf1 can function cell autonomously to remodel the sulfation state of cell-surface HS chains and can promote Wnt signaling when localized either at the cell surface or in the Golgi apparatus(Ai et al., 2003). Ai et al. propose that the HS 6-O desulfation activity of QSulf1 might convert cell-surface HSPGs to a low-affinity Wnt-binding state, thereby promoting the formation of low-affinity HS-Wnt complexes that can functionally interact with Frizzled receptors to initiate Wnt signal transduction(Ai et al., 2003; Ohto et al., 2002). As both Sulf1 and Sulf2 are secreted enzymes and are expressed in a variety of developmental contexts, they may participate in the signaling and distribution of Wnts, as well as that of other signaling molecules, such as the FGFs (see FGF signaling section). As Drosophila contains one QSulf1 homolog, it will be interesting to determine whether this protein is also required for Wg signaling.

Two models have been proposed to explain how HSPGs modulate Wg signaling. In the first model, HSPGs can either stabilize the Wg protein from being degraded or reduce the dimensionality of Wg ligand diffusion from three to two dimensions, thereby maintaining the local concentration of Wg ligand that is available for its receptor (Hacker et al.,1997; Pfeiffer et al.,2002) (Fig. 3C). This model is supported by data from several studies. For example, in the fly embryonic epidermis, Wg protein is tightly associated with the surface of the wg-expressing cells and is undetectable in sgl null embryos(Hacker et al., 1997; Pfeiffer et al., 2002). Similarly, in the wing imaginal disc, extracellular Wg protein levels are strikingly reduced in sfl mutants(Baeg et al., 2001), and in the ttv, sotv and botv single mutants(Bornemann et al., 2004; Han et al., 2004a; Takei et al., 2004). Finally,the overexpression of Wg protein can partially rescue the segmentation defects associated with sfl and sgl, thereby compensating for the loss of Wg signaling (Hacker et al.,1997). In the second model, HSPGs are proposed to act as co-receptors that directly facilitate the formation of Wnt/Wg-Fz signaling complexes (Lin and Perrimon,1999). Although most data support the first model, there is some evidence that HSPGs may directly regulate Wg signaling, possibly by acting as a co-receptor(s). The overexpression of Notum, for example, can extinguish the signaling activity of membrane-tethered Wg that does not move through tissue (Gerlitz and Basler,2002). And, as already discussed, analyses of EXT proteins in Wg signaling suggest that HSPGs might have two distinct roles in Wg distribution and signaling (Han et al.,2004a) (see also Box 3).

The Wg morphogen can be transported either by transcytosis through dynamin-mediated endocytosis or by passive diffusion through the extracellular space (Tabata and Takei, 2004; Teleman et al., 2001; Vincent and Dubois, 2002). Existing data support the latter mode of Wg movement in the wing disc. As discussed earlier, extracellular Wg protein is lost in clones of cells defective in HS biosynthesis, such as in the sfl or EXT(ttv, sotv and botv) mutants, suggesting that HSPGs are involved in Wg morphogen distribution (Baeg et al., 2001; Bornemann et al.,2004; Han et al.,2004a; Takei et al.,2004). If this is so, how do HSPGs regulate Wg movement? Several mechanisms have been proposed. Extracellular Wg could move through restricted diffusion by attaching to the GAG chains of HSPGs, as has been proposed for Hh and Dpp movement (Belenkaya et al.,2004; Han et al.,2004b) (Fig. 3A). In support of this view, extracellular Wg accumulates in the wild-type cells next to ttv-botv double mutant cells(Takei et al., 2004), while it is strikingly reduced in the mutant cells(Baeg et al., 2001; Bornemann et al., 2004; Han et al., 2004a; Takei et al., 2004). Alternatively, HSPGs could control the stability of Wg morphogen to ensure that it diffuses across a field of cells without being degraded(Bornemann et al., 2004; Han et al., 2004a; Takei et al., 2004). These two processes are not mutually exclusive and are most likely coupled during Wg gradient formation. It is currently unknown how HSPGs control the stability of Wg. HSPGs could prevent extracellular Wg from being degraded by extracellular proteinases (Bornemann et al.,2004), or they might bind to Wg and guide it to a different endocytosis route, thereby preventing it from being degraded by lysosomal-mediated degradation. In support of this, it has been shown that blocking Dynamin-mediated endocytosis can lead to enhanced extracellular Wg levels, suggesting that Wg is normally downregulated through this process(Strigini and Cohen,2000).

HSPGs might also transport Wg protein through vesicles called argosomes,which are generated from basolateral membranes and can travel from Wg-producing cells to various regions of the wing disc(Greco et al., 2001). Wg-expressing cells produce argosomes that contain high levels of Wg. Following the treatment of these cells with either heparinase I or III, Wg is no longer detected on the plasma membrane of the Wg-expressing cells, nor in the surrounding tissue. These and other data led Greco et al. to propose that argosomes are involved in Wg morphogen transport, and that this process is likely to be mediated by HSPGs (Greco et al., 2001).

Like Wnts, Hh signaling molecules act as key mediators in many developmental processes and essentially require HSPGs for their proper distribution and signaling activity(Ingham and McMahon, 2001; Lin and Perrimon, 2003).

The first evidence that HSPGs function in Hh distribution came from the genetic analysis of ttv (Bellaiche et al., 1998) (see Box 3 and Table 1). In the wing disc, Hh acts as a morphogen that forms a concentration gradient in an anterior strip of cells near the anteroposterior (AP) border of the wing disc (see Box 2). In wing discs containing ttv mutant clones, Hh can only be detected in the posterior-most ttv mutant cells that lie adjacent to wild-type cells. Furthermore, Hh can diffuse through a ptc mutant clone in the wing disc, but not through ptc-ttv double mutant cells(Bellaiche et al., 1998). Mainly based on these data, Bellaiche et al. proposed that a Ttv-modified HSPG is required for Hh to move from the cells where it is expressed to the anterior Hh-receiving cells (Bellaiche et al., 1998). Several recent studies have also shown similar defects in Hh distribution in wing clones mutant for sfl, sotv or botv (Han et al.,2004a; Takei et al.,2004). Importantly, Hh protein accumulates abnormally in the posterior compartment when the ttv-botv double mutant clones are made in the anterior compartment along the AP boundary, further suggesting that Hh fails to move into the HSPG mutant cells(Takei et al., 2004).

HSPGs are also required for Hh movement in the embryonic epidermis. The mature form of Hh, HhNp (see Box 4) is distributed as large punctate particles between Hh-expressing cells (Gallet et al.,2003). Cholesterol modification is required for HhNp to form these punctate particles (Gallet et al.,2003). In ttv mutant embryos, these punctate particles are not distributed between Hh-expressing cells(Gallet et al., 2003; The et al., 1999). Although Ttv is required for HhNp movement, it is not required for that of HhN (see Box 4)(Gallet et al., 2003; The et al., 1999). Together,these studies support the notion that HSPGs are required for the movement of cholesterol-modified HhNp.

The HSPG core proteins Dally and Dlp are also involved in Hh movement in the embryonic epidermis (Han et al.,2004b). dlp mutant embryos exhibit virtually identical defects to those of hh mutant embryos, and in their epidermis, Hh punctate particles can only be detected in Hh-expressing cells, and not in adjacent cells (Han et al.,2004b). However, in the wing disc, both Dally and Dlp act partially redundantly in Hh movement (Han et al., 2004b).

HSPGs might also control the stability of Hh by protecting it from degradation. Interestingly, Hh levels are reduced in Hh-producing cells mutant for sotv or ttv(Bornemann et al., 2004). Bornemann et al. argue that, by extension, Hh ligand instability could also contribute to reduced Hh levels and signaling in Hh-receiving cells lacking HSPGs (Bornemann et al., 2004). Although the evidence of Hh accumulation in front of HSPG-defective cells implicates HSPGs in Hh movement (Takei et al., 2004), current data suggest that HSPGs are likely to be involved in both Hh movement and stability.

Both vertebrate sonic hedgehog (Shh) and Drosophila Hh are secreted from cells as multimeric and monomeric forms(Chen et al., 2004; Zeng et al., 2001). The soluble Shh multimeric form is freely diffusive(Zeng et al., 2001) and is responsible for activating Shh-target genes(Chen et al., 2004). Do secreted Hh proteins freely diffuse to receiving cells through extracellular spaces, and what is the role of HSPGs in Hh movement? Han et al. have demonstrated that a narrow strip of sfl or ttv mutant cells in the fly wing disc is sufficient to completely block Hh signaling in the anterior wild-type cells adjacent to mutant cells, suggesting that Hh fails to move across these HSPG-deficient cells(Han et al., 2004a)(Fig. 3). Similar results are observed in clones mutant for both dally and dlp(Han et al., 2004a). Han et al. further showed that HSPG-mediated Hh movement is independent of dynamin-mediated endocytosis (Han et al.,2004a). On the basis of these and other data, Han et al. proposed that Hh movement is mediated by restricted diffusion involving Dally and Dlp(Han et al., 2004b)(Fig. 3A).

HSPGs are also required for Hh signaling. In tissue culture experiments,Dlp, but not Dally, Sdc or Trol, is required for Hh signaling(Lum et al., 2003). As mentioned earlier, Desbordes and Sanson have also argued that Dlp is specifically required for Hh signaling in the embryonic epidermis(Desbordes and Sanson, 2003). Although there are some arguments regarding whether Dlp is also involved in Wg signaling (Perrimon et al.,2004), the reduced bagpipe expression, which is normally activated by Hh signalling but is inhibited by Wg signalling in mesodermal cells of dlp null embryos, resemble that of hh mutant embryos, providing further evidence that Hh signaling requires Dlp during fly embryogenesis (Han et al.,2004a). Importantly, although previous studies have shown that ectopic expression of Hh can rescue the cuticle defects associated with HS GAG mutants, such as those seen in sgl, sfl, frc and slalommutants (Luders et al., 2003; Perrimon et al., 2004; Selva et al., 2001), ectopic expression of Hh fails to restore Hh signaling activity, as assayed by wg expression, in dly RNAi embryos(Desbordes and Sanson, 2003). These results suggest that the core protein of Dly is crucially required for Hh signaling, whereas the attached HS chains are required for optimal Hh signaling activity. Alternatively, the Dly core protein may be required for Hh processing in the embryonic epidermis. It is also important to note that although Dlp is required for Hh signaling during embryogenesis, Dlp is functionally redundant with Dally in Hh signaling in the wing disc(Han et al., 2004b),suggesting that the specificity of HSPG involvement in Hh signaling depends on the developmental context.

Dlp might regulate Hh signaling in several ways. It might modulate Hh levels at the cell surface and indirectly control the interaction of Hh with its receptor Ptc. Alternatively, it might act as a co-receptor, perhaps by transferring Hh to its receptor Ptc, or by forming a Hh-Dlp-Ptc ternary complex in which Dlp may function to facilitate the Hh-Ptc interaction or to stabilize a Hh-Ptc complex, as in the case of FGF signaling(Ornitz, 2000; Pellegrini, 2001). On the one hand, in tissue culture experiments, Lum et al. showed that Dlp acts cell autonomously upstream or at the level of Ptc to activate the expression of an Hh responsive-reporter, indicating that Dlp might deliver Hh to its receptor. However, the knockdown of Dlp did not block Hh signaling when Hh was expressed in responding cells (Lum et al.,2003). These results suggest that Dlp is not absolutely required for Hh signaling in the presence of relatively large amounts of Hh. On the other hand, ectopic expression of Hh or HhN fails to rescue Hh signaling defects in dlp-RNAi embryos(Desbordes and Sanson, 2003),and thus these data suggest that Dlp acts as a co-receptor for Hh signaling(Desbordes and Sanson, 2003). Studies in the dlp null mutant should help to resolve this issue.

Hh signaling might also be regulated by other HSPG core proteins in different tissues or developmental processes. For example, mutations in the gene encoding Trol, the Drosophila Perlecan that forms a complex with Hh, causes neuroblasts to undergo cell cycle arrest in the larval brain(Datta, 1995; Park et al., 2003; Voigt et al., 2002). Genetic interaction experiments also indicate that trol is required for Hh signaling during neuroblast division (Park et al., 2003).

The TGFβ superfamily of secreted proteins, such as the bone morphogenetic proteins (BMPs), act as morphogens in many developmental contexts to pattern tissues. Drosophila Decapentaplegic (Dpp), the functional ortholog of vertebrate BMP2 and BMP4, provides a paradigm for studying the mechanisms of BMP signaling(Raftery and Sutherland,1999). Dpp acts as a long-range morphogen during wing development(Box 2), and its activity and ligand gradients can be assayed by monitoring the activated form of Mothers against Dpp (pMAD), a cytoplasmic transducer of Dpp signaling(Tanimoto et al., 2000), and by GFP-Dpp fusion proteins that retain signaling activity(Entchev et al., 2000; Teleman and Cohen, 2000).

dally mutants exhibit various dpp-like patterning defects and these defects show genetic interactions with the dpp signaling pathway (Jackson et al., 1997; Nakato et al., 1995). Ectopic expression of Dally results in enhanced Dpp signaling in the fly wing disc(Fujise et al., 2003). These data suggest a positive role of Dally in Dpp signaling. Recent data suggest that Dally and Dlp are partially redundant in Dpp signaling, as the removal of both Dally and Dlp causes stronger Dpp signaling defects than those seen in either dally or dlp mutants(Belenkaya et al., 2004). Consistent with the involvement of Dally and Dlp in Dpp signaling, Dpp signaling is strikingly reduced in cells defective in HS biosynthesis such as in EXT mutants (Bornemann et al.,2004; Han et al.,2004a; Takei et al.,2004) or in the sfl mutant(Belenkaya et al., 2004).

What is the role of HSPGs in Dpp signaling? Elevated levels of Dally increase the sensitivity of cells to Dpp in a cell-autonomous fashion in the wing disc, leading the authors to propose that Dally might serve as a co-receptor for Dpp (Fujise et al.,2003). However, Dally overexpression might lead to enhanced Dpp signaling activity by increasing the levels of Dpp on Dally-expressing cells. Moreover, Dpp signaling is not defective in the first row of sflsingle or dally-dlp double mutant cells if they are adjacent to wild-type cells that face Dpp-producing cells(Belenkaya et al., 2004). Finally, extracellular Dpp levels are strikingly reduced in sfl or dally-dlp mutant cells (Belenkaya et al., 2004). These data suggest that the main function of HSPGs in Dpp signaling in the wing disc is to modulate levels of Dpp(Fig. 3C)(Han et al., 2004a; Takei et al., 2004; Belenkaya et al., 2004).

Biochemical studies have demonstrated that both Dpp and BMP2 are heparin-binding proteins (Groppe et al.,1998; Ruppert et al.,1996), and that the binding of BMP2 to heparin requires its N-terminal domain (Ruppert et al.,1996). Recently, the crystal structure of BMP2 alone and in complex with its receptor has revealed that the N-terminal domains of both BMP2 monomers are well placed for GAG binding(Kirsch et al., 2000; Scheufler et al., 1999). Moreover, the GAG-binding domain of BMP2 is not located in regions required for BMP2 dimerization and receptor interaction. These data suggest that HSPGs are unlikely to be directly involved in Dpp/BMP signaling as a co-receptor, as they are in FGF signaling (Ornitz,2000; Schlessinger et al.,1995), nor are they likely to be essential for Dpp signaling in tissues where Dpp is abundant. Indeed, in striking contrast to the requirement for HSPGs for proper Dpp signaling during wing development, in sgl, sfl,ttv, sotv and botv mutant embryos, embryonic dorsoventral (DV)patterning controlled by Dpp signaling is not defective(Haerry et al., 1997; Han et al., 2004a; Lin and Perrimon, 1999). During early embryogenesis, Dpp is highly expressed in the dorsal half of the embryo, and its activity gradient is regulated by an inverse gradient of its antagonist Short gastrulation (Srinivasan et al., 2002).

Dpp acts as a morphogen during wing disc development(Tabata and Takei, 2004; Vincent and Dubois, 2002). Do HSPGs control the movement of the Dpp morphogen in the wing disc? Previous findings have suggested that the Dpp morphogen moves across cells by a transcytosis mechanism through Dynamin-mediated endocytosis(Entchev et al., 2000). However, Lander et al. have proposed on theoretical grounds that diffusive mechanisms of Dpp transport are much more plausible than non-diffusive mechanisms (Lander et al.,2002). Several studies implicate the involvement of HSPGs in Dpp distribution. Dpp distribution is defective in wing disc cells mutant for HS biosyntheis, such as in the EXT or sfl mutants(Belenkaya et al., 2004; Han et al., 2004a; Takei et al., 2004). Both Dally and Dlp are required for Dpp distribution(Belenkaya et al., 2004; Fujise et al., 2003). Importantly, recent data also demonstrate that although Dynamin-mediated endocytosis is required for Dpp signaling it is not essential for Dpp morphogen movement (Belenkaya et al.,2004). Extracellular Dpp can move across endocytosis defective cells, but not across sfl or dally-dly mutant cells(Belenkaya et al., 2004). These data led Belenkaya et al. to propose that Dpp moves across cells through a restricted diffusion mechanism that involves Dally and Dlp(Belenkaya et al., 2004). Dally and Dlp have been proposed to have a similar role in Hh movement(Han et al., 2004b)(Fig. 3A).

FGFs are among the best-studied HSPG-binding proteins(Ornitz, 2000), and play many important roles in regulating cell proliferation, migration and differentiation during development(Coumoul and Deng, 2003; Ornitz and Itoh, 2001). Recent studies have provided new insights into the mechanisms by which HSPGs modulate FGF signaling in development.

As in the case of Wnt signaling, the first genetic evidence of a role for HSPGs in FGF signaling came from sgl and sfl mutant embryos,which show defective FGF signaling during Drosophila embryogenesis(Lin et al., 1999). In Drosophila, two FGF receptors, Heartlless (Htl) and Breathless (Btl),are required for the migration of mesodermal and tracheal cells, respectively. In sgl and sfl mutant embryos, both Htl and Btl signaling are defective, demonstrating the essential role of HSPGs in FGF signaling in development (Lin et al., 1999; Nybakken and Perrimon, 2002). It is currently unknown which HSPG core proteins are involved in Htl- and Btl-mediated FGF signaling during embryogenesis. However, Trol is not only required for Hh signaling, but is also involved in Btl-mediated FGF signaling during neuroblast division in the larval brain(Park et al., 2003). Study of available null mutants in sdc, dally and dly should reveal which core proteins are involved in FGF signaling in Drosophila.

Recent studies have provided in vivo evidence for the requirement of HSPGs in FGF signaling in vertebrates. First, lazy mesoderm(lzme), a mouse homolog of Drosophila sgl, arrests during gastrulation with defects in mesoderm and endoderm migration, which require FGF signaling (Garcia-Garcia and Anderson,2003). In lzme embryos, although Fgf8 expression is not affected, the expression of several Fgf8 downstream-target genes is defective, resembling the altered expression patterns seen in Fgf8^–/– embryos. These data strongly suggest that lzme is required for FGF8 signaling in mice(Garcia-Garcia and Anderson,2003). Second, in Xenopus embryos, depleting Glypican 4 transcripts with antisense morpholino oligonucleotides causes FGF signaling defects during neural tube closure (Galli et al., 2003).

Biochemical studies have demonstrated the importance of 6-Osulfation of HS for FGF signaling (Nakato and Kimata, 2002). Crystallographic studies have further suggested that the 6-O sulfation of HS is required for FGF1-FGFR2 and FGF2-FGFR1 interactions (Pellegrini et al., 2000; Schlessinger et al., 2000). The 6-O sulfation of HS is indeed essential for FGF signaling in development. In Drosophila, this 6-Osulfation is catalyzed by a single Heparan sulfate 6-O-sulfotransferase (Hs6st) gene(Kamimura et al., 2001). Interestingly, Hs6st is specifically expressed in mesodermal and tracheal cells. Knocking down this transcript by RNAi severely perturbs tracheal development and diminishes Btl-dependent MAPK activity(Kamimura et al., 2001),suggesting that 6-O sulfations in HS chains are essential for FGF signaling.

As mentioned earlier, QSulf1, a HS-specific 6-O endosulfatase, and its mammalian orthologs can promote Wnt signaling(Ai et al., 2003; Dhoot et al., 2001). By contrast, two recent studies have shown that Sulf1 inhibits FGF signaling. First, the expression of human SULF1 resulted in a striking reduction in FGF2 signaling activity in ovarian cancer cell lines(Lai et al., 2003). Second,QSulf1 was shown to block FGF signaling activity in both Xenopus and chicken embryos (Wang et al.,2004). These studies provide further evidence of the involvement of 6-O sulfation of HS in FGF signalling, and also indicate that Sulf1, and possibly Sulf2, has a dual regulatory function as a positive regulator of Wnt signaling and as a negative regulator of FGF signaling.

Biochemical studies have provided evidence that HSPGs participate in FGF signaling by directly interacting with FGFs and their receptors in a ternary complex on the cell surface (Ornitz,2000; Pellegrini,2001). The crystal structures of FGF-FGFR-heparin complexes have shown that one FGF, one heparin molecule and one FGFR chain constitute the 1:1:1 FGF-FGFR-heparin complex(Pellegrini, 2001; Pellegrini et al., 2000; Schlessinger et al., 2000). One possible role of heparin/HSPGs within the FGF signaling complex is to facilitate the FGF and FGFR interaction(Fig. 3B). Alternatively,heparin/HSPGs may enhance the stability of a binary FGF-FGFR complex. In support of this view, various biochemical studies have demonstrated that heparin can increase the affinity of FGF for its receptor(Ornitz, 2000).

The studies described above have illustrated that HSPGs have essential roles in the Wnt, Hh, TGFβ and FGF developmental signaling pathways. Do HSPGs have any specificity in regulating these pathways? How do HSPGs control these pathways in various tissues? Now studies in Drosophila and vertebrates are providing evidence that the specificity of HSPG function in these pathways is controlled by both the HSPG core proteins and their attached HS GAG chains. Furthermore, these HSPG functions are being discovered to be context dependent.

The nature of HSPG core proteins is likely to contribute to the specificity of HSPGs in cell signaling. The distribution of HS GAG chains solely depends on the expression of HSPG core proteins. The expression of HSPG core proteins is, in turn, regulated by Wg, Hh and Dpp signaling, as has been shown for Dally in the wing disc (Fujise et al.,2001; Fujise et al.,2003). HSPG core proteins can also contribute to the modification of HS GAG chains (Chen and Lander,2001; Esko and Zhang,1996). The GAG attachment sites and their surrounding amino acid sequence can determine the number of HS chains that attach to a specific HSPG core protein (Esko and Selleck,2002). Moreover, all of the Glypican core proteins share a unique,highly conserved N-terminal globular domain that contains 14 cysteine residues, which is a potent enhancer of preferential HS glycosylation(Chen and Lander, 2001). Finally, The HSPG core proteins may be directly involved in cell signaling. For example, Glypican proteins contain a GPI anchor. This anchor is enriched in the lipid raft domain, which has been implicated in protein sorting and in signaling complex assembly (Simons and Toomre, 2000). Other parts of the Glypican protein core could be directly involved in cell signaling. Indeed, while overexpression of Hh can rescue Hh signaling defects in sgl null embryos(Hacker et al., 1997), it fails to restore Hh signaling activity in dlp RNAi embryos(Desbordes and Sanson, 2003). Similarly, in Xenopus, the overexpression of full-length Xgly4 can inhibit Activin-induced elongation (Ohkawara et al., 2003). These data suggest direct roles for Glypican core proteins in cell signaling.

The specificity of HSPGs is also controlled by the biosynthesis of HS GAG chains. As shown in Fig. 2,several glycosyltransferases and modification enzymes are involved in this process. Genetic studies of the Drosophila EXT proteins Ttv, Botv and Sotv, have illustrated the involvement of HS biosynthesis in specific functions of HSPGs. For example, these proteins are required for Hh and Dpp signaling, but neither Wg nor FGF signaling is defective in ttvmutant embryos during embryogenesis (The et al., 1999). By contrast, Wg short-range signaling is defective in both botv single and ttv-sotv double mutant clones in the wing disc (Han et al., 2004a). Specific defects associated with distinct modifications of HS GAG chains have also been reported. Drosophila Hs6st is specifically expressed in mesodermal and tracheal cells during fly embryogenesis, and is required for FGF signaling (Kamimura et al.,2001). In the mouse, mutations in HS 2-O-sulfotransferase(Hs2st) (Bullock et al., 1998; Merry et al., 2001) and D-glucuronyl C5-epimerase (Li et al.,2003) are associated with renal agenesis and skeletal malformations, but major organ systems, including the brain, liver and gastrointestinal tract appear to be normal. Although specific signaling defects have not yet been identified in these two mouse mutants, the unique patterning defects observed indicate that Hs2st and the C5 epimerase have specific functions in development. Perhaps the disruption of the Drosophila homologs of these genes will shed light on the signaling pathways they act in.

The requirement for HSPGs for the activity of certain signaling pathways also depends on the developmental context. For example, Dpp signaling in DV patterning during embryogenesis is not defective in fly embryos mutant for HS biosynthesis enzymes (sgl, sfl or Exts) or HSPG core proteins (Dally and Dlp) (Haerry et al., 1997; Han et al., 2004a; Lin and Perrimon, 1999; The et al., 1999). By contrast, clones mutant for these genes show striking defects in Dpp distribution and signaling in the wing disc(Belenkaya et al., 2004; Bornemann et al., 2004; Fujise et al., 2003; Han et al., 2004a; Takei et al., 2004). One explanation for this is that dpp is highly expressed in the dorsal half of embryos, where the Dpp activity gradient is regulated by an inverse gradient of its antagonist Short gastrulation(Srinivasan et al., 2002),whereas in the wing disc, Dpp forms concentration gradients to transduce its signaling. As such, the main function of HSPGs in Dpp signaling in the wing disc is likely to be the modulation of extracellular Dpp levels by preventing it from being degraded. A recent study has also suggested that the role of proteoglycans in Drosophila Wg and Dpp signaling has not been conserved in vertebrates (Garcia-Garcia and Anderson, 2003), owing to the fact that mouse embryos with mutations in lzme are defective in FGF signaling, but not in Wnt3 or Nodal (a TGFβ member) signaling during gastrulation. These authors propose that many of the phenotypes of proteoglycan-related human disorders could be due to disrupted FGF signaling. However, in other developmental contexts, such as in developing limbs, HSPGs may play essential roles in Wnt and BMP signaling. For example, loss of mouse glpyican 3 in association with heterozygous Bmp4 deficiencies results in polydactyly and skeletal abnormalities that are not seen in either Gpc3 null or Bmp4heterozygous mice, suggesting a role for glypican 3 in BMP4 signaling(Paine-Saunders et al.,2000).

Developmental and biochemical studies in Drosophila and vertebrate model systems have revealed essential roles for HSPGs in regulating a number of developmental signaling pathways in a variety of developmental contexts. HSPGs appear to control these signaling pathways in various ways, such as by retaining and stablizing ligands, by transporting ligands and by facilitating ligand-receptor interactions. These roles are likely to be mediated by both the HSPG protein cores and the specific HS GAG modifications that are carried out by a combination of enzymes. Further detailed analyses of animals mutant for individual core proteins and HS GAG chain-modifying enzymes are needed to define the molecular mechanisms by which HSPGs regulate cell signaling in various developmental processes. Cell biology studies of live tissues are also needed to further define the cellular basis of the HSPG-mediated regulation of morphogen gradient formation during development. Finally, we also need to determine the HS structures of specific HSPGs by HS GAG sequencing and advanced mass spectroscopy to elucidate the molecular nature of HSPG-ligand interactions. Given their molecular complexity, a complete understanding of HSPG functions in development requires that we investigate them using combined genetic, cell biological and biochemical approaches.

Two recent papers have demonstrated that Dlp promotes low-level Wg activity far from the site of Wg production, but reduces high-level Wg signaling near the Wg source (Kreuger et al.,2004; Kirkpatrick et al.,2004). Kreuger et al. further show that Notum, a previously identified secreted enzyme, can cleave the GPI anchor of Dlp, releasing Dlp from the cell surface. Kirkpatrick et al. show through genetic interaction studies that Dlp and Notum cooperate to restrict high-level Wg signaling. As Notum is transcribed in response to high-level Wg signaling, these findings suggest that feedback confers spatial specificity on proteoglycan function by switching Dlp from a membrane-tethered cofactor to a secreted antagonist.

I thank Chun Han and Tatyana Belenkaya for suggestions and figures. Work in the author's laboratory is supported by grants from the NIH.