Cell Surface Heparan Sulfate Proteoglycans: Selective Regulators of Ligand-Receptor Encounters*

Cell surface heparan sulfate proteoglycans (HSPGs), substantially more abundant than most receptors, modulate encounters of extracellular protein ligands with their receptors by forming HSprotein complexes. Two gene families account for most cell surface HSPGs. Both consist of discrete core proteins covalently attached to two or three chains of HS, an Nand O-sulfated linear polysaccharide of repeating disaccharides containing N-acetylglucosamine (GlcNAc) and uronic acid (glucuronic acid (GlcA) or iduronic acid (IdoUA)). The syndecan family was the first discovered, which in mammals contains four gene products with distinctive extracellular domains (ectodomains) and highly conserved short cytoplasmic domains. These apparently extended proteins place the HS chains distal from the plasma membrane (1, 2). The syndecan family contrasts with the glypican family, which in mammals contains six gene products that are covalently linked to plasma membrane lipid by glycosylphosphatidylinositol anchor (1, 3). The glypican core proteins contain six invariant disulfide bonds, are likely to be globular, and place HS chains adjacent to the plasma membrane. Expression of both the syndecans and glypicans is extensively regulated during mouse embryogenesis and results in discrete adult expression patterns for each HSPG such that every adherent cell exhibits a distinct repertoire of cell surface HSPGs. Binding to HS chains is remarkably widespread among extracellular proteins, especially matrix proteins, proteases and their inhibitors, lipases, lipoproteins, growth factors and their binding proteins, cytokines, chemokines, collectins, and antimicrobial peptides. These proteins are involved in morphogenesis, tissue repair, energy balance, and host defense (Fig. 1). Additionally, numerous pathogens (e.g. herpes simplex virus, Neisseria, Plasmodium) bind to the cell surface via HS (4). Importantly, many of these ligand-HS interactions are essentially identical from Drosophila to the mouse, including those involved in generation of the basic metazoan body plan, e.g. dpp (bone morphogenetic proteins 2–4), wg (Wnt-1), and sog (chordin). Formation of the complexes can enhance or reduce receptor activation, often depending on the concentrations of ligand, receptor, and HSPG. The HS chains catalyze encounters between ligand and signaling receptor by bringing them together. Because binding to the HS chain reduces the dimensionality of this interaction from three (when the ligand is soluble) to two (when the ligand is bound to the HS chain), interaction could result from a localized increase in ligand concentration at optimal HS concentrations (5). However, at HS levels lower or higher than optimal, the effective ligand concentration for engaging the receptor will fall, potentially accounting for the bell-shaped activity curve typically seen experimentally when HSPG (or heparin) concentrations are varied. The curve may be concave or convex depending on whether ligand binding to the HSPG is inhibitory or stimulatory (6). Furthermore, the cytoplasmic domains of the syndecans also form complexes with cytoplasmic enzymes and scaffolding proteins, adding to the modulating influence of these proteoglycans (7). The HS chains are structurally diverse by virtue of their biosynthesis. A non-sulfated repeating disaccharide precursor is generated while attached to a core protein and is then sequentially modified by a variety of enzymes in reactions that do not go to completion. The details of this biosynthetic scheme have been recently reviewed in this series (8). Three major characteristics of this scheme produce HS chains with selectivity for protein binding. First, the process yields an extraordinary variety of saccharide sequences. Second, clustering of the modifications along the HS chain yields highly N-sulfated domains (NS domains) of approximately 12–20 residues that alternate with typically larger sized, relatively unmodified N-acetyl-rich domains (NA domains). The NS domains are rich in IdoUA, which can assume several different conformations and thus influence the orientation of the sulfate residues in space. This domain organization places relatively flexible NA domains adjacent to relatively rigid NS domains, thus facilitating protein interactions with the sulfate residues. Finally, this microsequence diversity and macro-organization are cell typespecific and do not appear to be core protein-specific (e.g. HS chains of syndecan-1 and -4 from mammary epithelia are indistinguishable), presumably the result of cell type-specific repertoires of the HS chain-modifying enzymes. Distinct oligosaccharide sequences in HS chains are recognized by the various proteins whose function depends on this interaction. The best characterized of these interactions is the recognition of a specific pentasaccharide sequence by antithrombin III (9). FGF-2 binds most tightly to a specific hexasaccharide sequence, but an additional 4–6 sugar residues are required to activate the receptor (10). Specific oligosaccharide binding sequences are known for multiple ligands (11–14); however, there is no universal consensus amino acid sequence for protein binding to HS chains. Most studies suggest that multiple arginine and/or lysine residues aligned on the protein surface accommodate a distinctive array of anionic sites on the HS chain (14).

tor, and HSPG. The HS chains catalyze encounters between ligand and signaling receptor by bringing them together. Because binding to the HS chain reduces the dimensionality of this interaction from three (when the ligand is soluble) to two (when the ligand is bound to the HS chain), interaction could result from a localized increase in ligand concentration at optimal HS concentrations (5). However, at HS levels lower or higher than optimal, the effective ligand concentration for engaging the receptor will fall, potentially accounting for the bell-shaped activity curve typically seen experimentally when HSPG (or heparin) concentrations are varied. The curve may be concave or convex depending on whether ligand binding to the HSPG is inhibitory or stimulatory (6). Furthermore, the cytoplasmic domains of the syndecans also form complexes with cytoplasmic enzymes and scaffolding proteins, adding to the modulating influence of these proteoglycans (7).
The HS chains are structurally diverse by virtue of their biosynthesis. A non-sulfated repeating disaccharide precursor is generated while attached to a core protein and is then sequentially modified by a variety of enzymes in reactions that do not go to completion. The details of this biosynthetic scheme have been recently reviewed in this series (8). Three major characteristics of this scheme produce HS chains with selectivity for protein binding. First, the process yields an extraordinary variety of saccharide sequences. Second, clustering of the modifications along the HS chain yields highly N-sulfated domains (NS domains) of approximately 12-20 residues that alternate with typically larger sized, relatively unmodified N-acetyl-rich domains (NA domains). The NS domains are rich in IdoUA, which can assume several different conformations and thus influence the orientation of the sulfate residues in space. This domain organization places relatively flexible NA domains adjacent to relatively rigid NS domains, thus facilitating protein interactions with the sulfate residues. Finally, this microsequence diversity and macro-organization are cell typespecific and do not appear to be core protein-specific (e.g. HS chains of syndecan-1 and -4 from mammary epithelia are indistinguishable), presumably the result of cell type-specific repertoires of the HS chain-modifying enzymes.
Distinct oligosaccharide sequences in HS chains are recognized by the various proteins whose function depends on this interaction. The best characterized of these interactions is the recognition of a specific pentasaccharide sequence by antithrombin III (9). FGF-2 binds most tightly to a specific hexasaccharide sequence, but an additional 4 -6 sugar residues are required to activate the receptor (10). Specific oligosaccharide binding sequences are known for multiple ligands (11)(12)(13)(14); however, there is no universal consensus amino acid sequence for protein binding to HS chains. Most studies suggest that multiple arginine and/or lysine residues aligned on the protein surface accommodate a distinctive array of anionic sites on the HS chain (14).

HSPGs Regulate Ligand-Receptor Encounters as Coreceptors
Cell surface HSPGs bind to a large number of proteins; thus, some of their molecular interactions have been considered nonspecific. This perception is challenged by both in vivo evidence showing that cell surface HSPGs are required for specific morphogenetic events and by a growing body of data demonstrating that HS interactions with ligands depend on specific HS sequences. These interactions are mostly modulating or regulatory and do not typically result in intracellular signaling from the HSPG. The studies reviewed below suggest that when HSPGs accelerate ligand-receptor encounters, the subsequent receptor action depends on whether the ligand is soluble or insoluble.
Soluble Ligands-Extensive research has focused on HSPGs as coreceptors for a large variety of soluble ligands, including FGFs, transforming growth factors-␤1 and -␤2, vascular endothelial growth factor (VEGF 165,189 ), CC and CXC chemokines, and various cytokines (1). The diverse cognate receptors for these ligands include receptor tyrosine kinases and seven pass G-protein-coupled receptors. In this role, the HSPGs can modulate ligand-receptor encounters by altering ligand concentrations, stability, or conformation and by ligand or receptor oligomerization. Indeed, there seem to be several distinct mechanisms by which the HSPGs act as coreceptors.
One of the best studied examples is that of the fibroblast growth factor receptor-1 (FGFR-1), a receptor tyrosine kinase, and its ligands FGF-1 and -2. Rapraeger et al. (15) first demonstrated that in the absence of cell surface HS, FGF-2 interacts poorly with its high affinity FGFR-1 and does not activate downstream intracellular signaling. Subsequent studies showed that HS, FGF-2, and FGFR-1 form a ternary complex at the cell surface, where HS can be provided by either syndecans or glypicans (16). The HS in the complex solely accelerates the ligand-receptor interactions, as is evident from the finding that its absence can be overcome by increasing either reactant concentration. Because FGFR-1 must dimerize to signal, the cell surface HS is thought to bind monomeric FGF-2 and form oligomers that, in turn, dimerize the receptor (5,6). Despite its simplicity, this model is controversial. For example, it was recently shown that the minimal subunit for FGFR-1 activation is a monomer, suggesting that the HS provides a template for the FGF-2 and FGFR-1 interaction and does not act by oligomerizing either ligand or receptor (17). Similarly, for the FGF-7 receptor, heparin or HSPG-induced oligomerization of either FGF-1 or -7 does not correlate with biological activity, supporting the model that HSPG-induced ligand oligomerization is not critical in formation of the receptor-ligand complex (18).
Insoluble Ligands-Binding to insoluble ligands immobilizes the HSPG in the plane of the membrane, which causes the syndecan cytoplasmic domain to interact with the actin cytoskeleton and form more stable adhesions. Because glypicans are linked only to the outer membrane leaflet, their direct interactions with cytoplasmic elements are limited. Individual syndecan cytoplasmic domains do not produce soluble cytoplasmic signals, but they can oligomerize, be phosphorylated, and interact with scaffolding and signaling molecules (19 -21).
The short syndecan cytoplasmic domains (28 -34 amino acids) contain three functional subdomains, C 1 , V, and C 2 , at the C terminus. The highly conserved C 1 subdomain contains binding sites for Src protein tyrosine kinases (21). The variable V subdomain of syndecan-4 can interact with phosphatidylinositol bisphosphate and the catalytic domain of protein kinase C␣, interactions thought to result in oligomerization of the cytoplasmic domain and localization of syndecan-4 to focal adhesion complexes (22). A recently identified intracellular protein, syndesmos, also binds specifically to the cytoplasmic domain of syndecan-4 via the C 1 and V subdomains (23). This interaction can regulate cell spreading and actin stress fiber organization.
The C 2 subdomain contains a conserved amino acid sequence, FYA, which is recognized by type II PDZ domain binding proteins. Syntenin and CASK, PDZ domain proteins that form subplasmalemmal scaffolding, bind immobilized syndecan cytoplasmic domains in vitro, colocalize with syndecans at the plasma membrane in cells, and appear to link the syndecans and the cytoskeleton. Syntenin is a widely distributed protein comprising an N-terminal region of unknown function followed by two tandem PDZ domains, resulting in a 1:2, syntenin:syndecan molecular stoichiometry (24,25).
CASK binds to the C 2 subdomain in the syndecans with its single PDZ domain (26). CASK also contains a Ca 2ϩ /calmodulin kinase and an enzymatically inactive guanylate kinase domain (27) that can bind to the T-box transcription factor, Tbr-1. Importantly, when CASK and Tbr-1 are coexpressed, CASK translocates to the nucleus where it acts as a coactivator for this transcription factor (28). However, overexpression of syndecan-3 results in retention of CASK with syndecan-3 in the perinuclear endoplasmic reticulum, reducing its translocation to the nucleus. This result was unexpected because syndecan-3 would be expected to localize to the cell surface. Nonetheless, these findings imply a link between syndecan interactions at the cell surface and transcriptional activity in the nucleus.
These interactions of the cytoplasmic domains and possibly others yet to be established are responsible for the effects of syndecans on cell shape and adhesion. Syndecan-4 participates with integrins and a variety of kinase-based signaling systems in the formation of focal adhesions (22, 29). Syndecan-1, which polarizes to the basolateral surfaces of epithelial cells, is required for the maintenance of epithelial morphology and organization (30).
HSPGs as Coreceptors in Vivo-The major cell surface HSPGs are also found in Drosophila, including a single syndecan (Dsyndecan) and two glypicans (Dally and Dally-like protein) (Table I). Genetic analyses implicate dally in signaling by both Wingless (wg, a Wnt family member) and Decapentaplegic (dpp, a transforming growth factor-␤ family member). The data are consistent with Dally acting as a coreceptor for these morphogens, yet there is unexplained selectivity in the effect of the mutation; Dally affects Wg activity in the embryo, but Dpp activity is affected solely in imaginal discs and not earlier during development. The basis of this selectivity is unclear.
Other Drosophila mutations validate the role of HSPGs as coreceptors in the action of FGFs (46). Two identified FGFRs are associated with alterations in mesodermal migration and cardiogenesis (Heartless, Htl) and in branch formation in the tracheal system (Breathless, Btl) (35,36). Mutations in the FGF ligand (Branchless, Bnl) recapitulate the Btl mutation (37). Importantly, sgl and sfl show similar phenotypes (38). As expected, FGFR-dependent MAP kinase activation, the downstream target of Btl and Htl, is reduced in sfl and sgl mutants. Furthermore, consistent with the role of HSPGs as FGF coreceptors, a constitutively active form of Htl can partially rescue mutations in both sgl and sfl, and overexpression of Bnl can partially overcome the requirement of sgl and sfl in Btl mutants.
The selective coreceptor functions of HSPGs are conserved in mammals. Transgenic mice expressing the wnt-1 oncogene driven by the mouse mammary tumor virus long terminal repeat develop mammary alveolar hyperplasia and frequent mammary adenocarcinomas. The hyperplasia and tumorigenesis are abolished when these mice are made syndecan-1-deficient by crossing with syndecan-1 null mice (39). Thus, syndecan-1, the predominant HSPG in mammary epithelial cells, apparently acts as a coreceptor for Wnt-1. However, the syndecan-1 null mice show normal mammary morphogenesis, an FGF-dependent process. The basis for this selectivity of coreceptor activity is unknown.

Shed, Soluble HSPGs Regulate Molecular Encounters in
the Extracellular Space Cell surface HSPGs are released from the cell surface in a process commonly known as shedding (40,41). Shedding of cell surface HSPGs was thought to occur only as part of normal turnover for these molecules. However, recent data indicate that shedding can also be a highly regulated cellular response to biological cues and that shed ectodomains themselves act as regulators of molecular encounters.
Mechanism of Shedding-Cell surface molecules are shed when they are cleaved by a family of enzymes known collectively as sheddases or secretases, and their ectodomains are released relatively intact from the cell surface (42,43). Shedding is an important mechanism of activation and secretion for approximately 1% of cell surface proteins, including growth factors, cytokines, cell adhesion molecules, and enzymes, among others. Because the shed ectodomains of these molecules perform critical functions in various pathophysiological events such as septic shock, cell proliferation, and host defense, the process of shedding has been targeted for prophylactic and therapeutic interventions, and the mechanism has been under intense investigation (43).
Each of the four mammalian syndecans is shed rapidly from the cell surface by proteolytic cleavage of the core protein (1). The site of cleavage for syndecan-1 has been localized to within 9 amino acids adjacent to the extracellular face of the plasma membrane (44). Although the identity of the syndecan cleaving enzyme is unknown, its activity is cell surface-associated and can be inhibited by peptide hydroxamates and tissue inhibitor of metalloproteinase-3 (TIMP-3) but not by TIMP-1 or -2 or by inhibitors of aspartic acid, cysteine, and serine proteinases (44). These properties suggest that the enzyme is a cell surface zinc metalloproteinase that belongs to the TIMP-3-sensitive ADAM (a disintegrin and metalloproteinase) family. Interestingly, this mechanism appears to be responsible for the shedding of all four mammalian syndecans despite the lack of sequence similarity in the juxtamembrane region, implicating conservation of cleavage site secondary structures or the contribution by other regions of the proteoglycans for substrate recognition by the shedding enzyme. The glypicans can be shed from the cell surface (at least from cells in culture) possibly through the action of glycosylphosphatidylinositol-specific phospholipases (3).
Shedding of syndecans can be regulated by various external stimuli and intracellular signaling pathways. Shedding of syndecan-1 and -4 can be enhanced by growth factors (45), cellular stress (44), and soluble virulence factors of microbial pathogens (46). These agonists activate specific signaling pathways to stimulate shedding. For example, epidermal growth factor family members act on the shedding mechanism via the extracellular signal-regulated kinase (ERK) MAP kinase pathway, whereas stress-related agonists work through the c-Jun NH 2 -terminal kinase (JNK) MAP kinase pathway (44). Other signaling pathways also appear to regulate shedding because of direct stimulation of protein kinase C by phorbol esters, and incubation of cells with membrane-permeable second messengers (e.g. ceramide), membrane-acting sphingomyelinase, and the protein tyrosine phosphatase inhibitor, pervanadate, all result in acceleration of syndecan-1 and -4 ectodomain shedding (44,45). These findings suggest that distinct signaling pathways converge to activate the cell surface-associated metalloproteinase that cleaves the syndecans. How they converge to stimulate a common mechanism is not known. However, available evidence indicates that all regulated syndecan shedding is inhibited by protein tyrosine kinase inhibitors such as genistein and tyrphostin, suggesting that a key regulatory component is a protein tyrosine kinase(s) acting downstream of other signaling pathways.
Functions of Shed Cell Surface HSPGs-The shedding of cell surface HSPGs (i) generates an intact, soluble HSPG ectodomain, (ii) rapidly reduces the amount of cell surface HS, and (iii) is a host response to tissue injury. Because shed ectodomains have lost only their membrane attachment, they retain the ligand binding activities of their cell surface counterparts and can act as soluble, biological effectors. For example, purified syndecan-1 ectodomains bind tightly to neutrophil elastase and cathepsin G and reduce the affinity of these proteases for their physiological inhibitors, thereby enhancing their activities (47). Purified syndecan-1 ectodomains can also function as potent inhibitors of heparin-mediated FGF-2 mitogenicity via the NA domains of their HS chains (48). However, when the same syndecan-1 ectodomains are treated with platelet heparanase, as would happen during tissue injury, the enzyme digests the NA domains. This releases the enzyme-resistant NS domains that now activate FGF-2 mitogenicity (48). Indeed, mitogenically active concentrations of the heparin-like NS domains are found in acute wound fluids. Thus, the macro-organization of HS chains has a functional significance. Inhibition of cell proliferation by shed syndecan-1 ectodomains has also been observed with carcinoma cell lines (49).
Shedding can also potentially regulate receptor signaling events by rapidly reducing the amount of cell surface HSPGs. The biological importance of this is evident from data showing reduced proliferation in response to FGF-2 in cells treated with either HS digesting enzymes or agents that interfere with HS sulfation (e.g. chlorate ion). Syndecan shedding is accelerated by tissue injury mediators such as growth factors, stress-related agonists, proteases, and microbial infections (1), and shed syndecan ectodomains are elevated in tissue injury fluids such as in tracheal aspirates of intubated infants and in acute dermal wound fluid (45,48). Importantly, the shed ectodomains themselves can bind and modify the activity of some shedding agonists, indicating that syndecan ectodomains are soluble regulators of tissue injury.

Evolutionary and Future Perspectives
The syndecans apparently arose during the major metazoan radiation of the Cambrian period because their core protein genes are similarly organized and present in conserved forms in nematodes, arthropods, and chordates. Metazoans evolved, in large part, because of their ability to establish epithelia that generate an intercellular environment segregated from the outside world. Syndecans are required to maintain such epithelia. Once an extracellular space was segregated, which emerged with gastrulation and the appearance of HS in Cnidaria, the evolution of extracellular signaling and matrix proteins was possible. Thus, HS-protein interactions seem to be based on an evolutionarily conserved extracellular code. This idea is consistent with (i) the remarkable prevalence of such proteins that depend for their function on binding HS, (ii) the many distinct protein structures responsible for this binding, (iii) the involvement of most of these proteins in processes that are fundamental to an organism's survival, and (iv) the apparent lack of major evolutionary changes in HS-protein interactions. Like the RNA code, this HS code is degenerate, redundant, and sequence-specific; its fidelity is based on interactions with proteins and its significance is in directing cellular behavior. However, unlike the RNA code, this code is not based on a direct template mechanism, is restricted to metazoans, and appears to depend on cellular differentiation. Despite the seemingly indiscriminant binding properties of HS chains, as we learn more most HSPG-protein interactions prove to be physiologically specific. For example, the initial HS chain-modifying enzyme, NDST-1, is the most widely distributed NDST isoform involved in HS biosynthesis, yet mice made null for this enzyme show incomplete maturation of type II pneumocytes, a strikingly limited defect (50,51). Another example is the absence of kidney development in mice with a disrupted HS-2-O-sulfotransferase gene (52). Human genetic diseases are also revealing unexpectedly specific functions for HSPGs. For example, the formation of benign bone tumors (hereditary multiple exostoses) is a consequence of mutations in HS copolymerases (53), and limitations on pre-and postnatal somatic growth are imposed by glypican-3 (Simpson-Golabi-Behmel syndrome) (54). Technologies recently introduced for the sequencing of HS chains (55) and for the chemical synthesis of specific HS sequences (56) will be available in the near future to help elucidate the molecular mechanisms underlying such distinct developmental and physiological processes.