Regulated Diversity of Heparan Sulfate*

Large numbers of proteins in animal tissues occur immobilized in the extracellular space, at cell surfaces or in the extracellular matrix. Some are anchored through interactions with other proteins. However, current research increasingly implicates proteoglycans as scaffold structures, designed to accommodate proteins through noncovalent binding to their glycosaminoglycan side chains (1). In particular, heparan sulfate (HS) proteoglycans are recognized as ubiquitous protein ligands. Binding of proteins to HS chains may serve a variety of functional purposes, from simple immobilization or protection against proteolytic degradation to distinct modulation of biological activity. Because of such interactions HS proteoglycans are critically involved in a variety of biological phenomena at various levels of complexity, including organogenesis in embryonic development, angiogenesis, regulation of blood coagulation and growth factor/cytokine action, cell adhesion, lipid metabolism, etc. (2–4). In this review we outline the evidence for regulated expression of specific domains in HS chains and its relation to selective protein binding. Further, we consider the current information on HS biosynthesis, with the aim of understanding the mechanisms in control of generating specific saccharide sequences. Contrary to heparin, which is synthesized exclusively as serglycin proteoglycan in connective tissue-type mast cells (1), the structurally related HS is produced by most mammalian (and many other) cell types. A variety of core proteins have been shown to carry HS chains, including agrin and perlecan in the extracellular matrix (5) and the syndecanand glypicantype species at the cell surface (6). There is so far no evidence of any correlation between the structures of core proteins and those of the attached HS chains, whereas, conversely, the fine structure of the HS chains on a given proteoglycan species may differ between cell types (7).

are predominantly N-sulfated, whereas those in HS show a more varied N-substitution pattern with appreciable proportions of both N-sulfated and N-acetylated and a smaller amount of N-unsubstituted GlcN units (3,8,9). These structures are generated through the formation of a [GlcA-GlcNAc] n polymer that is subsequently modified by partial N-deacetylation/N-sulfation of GlcNAc units, C-5 epimerization of GlcA to IdoA residues, and, finally, incorporation of O-sulfate groups at various positions (Fig. 1).
Compositional analysis of several HS preparations derived from different bovine organs revealed certain domain characteristics of unexpected regularity (10). Thus 20 -30% of the disaccharide units in all preparations were consistently arranged in alternating N-acetylated and N-sulfated sequence (NA/NS domain in Fig. 1). Moreover, although the proportions of consecutive N-sulfated disaccharide units (NS domain in Fig.  1) varied between preparations, those of IdoA residues within the NS domains and the extent of 2-O-sulfation of these units were remarkably constant. Very few 2-O-sulfate groups were located outside the NS domains. By contrast, more than half of the other major O-sulfate substituent, at C-6 of GlcN units, along with an appreciable proportion of IdoA residues occurred in NA/NS domains (Fig. 1). The trisulfated disaccharide unit, -IdoA(2-OSO 3 )-GlcNSO 3 (6-OSO 3 )-, abundant in heparin, is restricted to NS domains in HS (10,11), where it provides a key element in protein recognition. One route toward sequence variability in HS obviously involves modulation of 6-O-sulfation.
What are the potentials and limitations in the design of specific protein-binding domains in a HS chain? Clearly, major constraint is because of the substrate specificities of the enzymes that catalyze the various polymer modification reactions in HS biosynthesis (12). The internal portions of extended NA domains thus are devoid of IdoA units and sulfate groups and so far have not been implicated in direct interactions with proteins. Conversely, the seemingly nonselective interaction of many proteins with the heavily sulfated NS domains of heparin has tended to obscure the notion of selectivity in protein binding to HS (3). Such selectivity may be accomplished in various ways. The involvement of a "rare" component is conceptually appealing, because it accounts for the "unique" quality that may be recognized by some proteins but not by others. In fact, the early recognition of the specific AT-binding region in heparin and HS may be largely ascribed to the key position in this domain of a "unique" O-sulfate group (at C-3 of a single Nsulfated GlcN unit) (3, 4) (see Fig. 1). However, less conspicuous structures may have equal potential for selective ligand binding. The generation of a unique binding epitope may be based on the N-substituent pattern, in that the overall length or relative positioning of individual NS, NA/NS, or NA domains may be of importance, or on unusual combinations of N-substituents and other modifications, such as the occasional occurrence of 2-O-sulfated IdoA units in NA/NS domains (see Ref. 10). Indeed, selective protein binding may depend on the precise distribution of sugar units and O-sulfate groups within, for example, extended N-sulfated but otherwise sparingly modified domains, features not readily detectable upon routine structural analysis of HS preparations.
Variation in Composition of Heparan Sulfate-To what extent does the structural variability of HS preparations reflect regulated rather than random biosynthetic polymer modifica-tion? Some insight has been gained through compositional analysis of HS preparations from different bovine (10) or human (13) organs (see also Ref. 14). Remarkably, although HS from human cerebral cortex clearly differed in composition from HSs of several other tissues (liver, aorta, kidney), cerebral HS samples from different individuals appeared indistinguishable (13). The conclusion from these findings, that the structures of HSs from different sources are strictly regulated, was underlined by the observation that 2-O-sulfated GlcA units were abundant in HS from adult brain but lacking in HS from neonatal cerebral cortex (13).
Immunohistochemical application of monoclonal anti-HS antibodies revealed additional complexity (9,15,16). A panel of such antibodies that recognized different, albeit yet poorly defined, HS epitopes thus yielded markedly different staining patterns when applied to sections of rat kidney. Apparently, the composite kidney tissue displays several HS species, with distinct topology in relation to different cell types. The pronounced interindividual similarity in overall composition of HS from a given organ (13,17) suggests that any HS subspecies present must be expressed in remarkably constant proportion from one individual to another.
Modulation of Heparan Sulfate Structure-Changes in HS composition have been observed in association with development/aging as well as with certain pathological processes. During development of the embryonic brain, HS-mediated growth factor activity is switched between members of the fibroblast growth factor (FGF) family, from FGF-2 to FGF-1 (18). This change, which correlates with a transition from proliferation of neural precursor cells to neuronal differentiation, is accompanied by alterations in patterns of 6-O-sulfation, total chain length, and the number of sulfated domains of the predominant HS species. A somewhat related phenomenon, albeit on a different time scale, was observed on analysis of HS from human aorta (19). An age-dependent increase in GlcN 6-O-sulfation was demonstrated, resulting in increased abundance of the trisulfated -IdoA(2-OSO 3 )-GlcNSO 3 (6-OSO 3 )-disaccharide unit. Concomitantly enhanced binding of the HS was noted to isoforms of platelet-derived growth factor (PDGF) A and B chains containing polybasic cell retention sequences. By contrast, the binding to FGF-2 was affected to a much lesser extent.
Several reports have described changes in the sulfation pattern of HS in response to "differentiation" or "transformation" of cells in culture (see e.g. Refs. 20 and 21, and references therein). These data do not provide any unified picture; for instance, the expression of a given sulfate substituent may be up-or down-regulated because of transformation, depending on cell type. Nevertheless, the data demonstrate that the structure of HS produced by a single cell line may be modulated because of distinct stimuli. Also other diseases may be associated with changes in HS structure. Experimentally induced diabetes in rats thus resulted in decreased N-sulfation of hepatic HS (22). Further, recent analysis of the HS accumulated in internal organs in AA amyloidosis indicated a switch in O-sulfation, from the distinct distributions characteristic of each individual organ (liver and spleen) to a novel pattern that was common to the amyloid-associated HS irrespective of organ source (17).

Protein-binding Domains
Although only a few protein-binding domains in HS chains have been defined, sufficient information has been accumulated to implicate some of the potential domain types outlined above, including complex, composite binding sites. Oligomeric cytokines such as interferon-␥ (23), platelet factor 4 (24), or interleukin-8 (25) thus bind to HS through interactions of the monomer components with different NS domains within a single polysaccharide chain. In the interleukin-8-binding region such NS domains may be separated by an NA domain composed of up to ϳ10 monosaccharide units. The internal NA domain of the interferon-␥-binding region may encompass up to ϳ30 sugar units. However, most HS-binding proteins studied so far appear to interact with saccharide sequences of ϳ5-15 units in single domains. The AT-binding pentasaccharide region is one example of such a sequence that is functionally dependent on the occurrence of a "rare" component, in this case the 3-O-sulfated GlcNSO 3 unit (Fig. 1). In the AT-binding re- The arrows within parentheses indicate (arbitrary) sites of variable polymer modification, i.e. residues that satisfy the substrate specificity of the indicated reactions but nevertheless escape target selection. Two NDST isoforms are indicated, along with their potential target residues. For further information regarding isoforms of other enzymes (not indicated in the scheme) see the text. The reducing terminus is to the right. gion the GlcN(3-OSO 3 ) moiety is linked at C-4 to a GlcA unit (3). Rat glomerular HS (26) and, remarkably, "heparin" from a marine clam, Anomalocardia brasiliana (27), also contain the same 3-O-sulfated GlcN unit linked at C-4 instead to IdoA, presumably required for specific binding to other, still unidentified, proteins. Other unusual components found in both heparin and HS are the 2-O-sulfated GlcA and N-unsubstituted GlcN units. The GlcA(2-OSO 3 ) residues are preferentially expressed in certain HS subspecies (13, 28) but have not been implicated with any functional roles. N-Unsubstituted GlcN residues are enriched in HS species capable of binding L-and P-selectin and thus potentially involved in regulating leukocyte traffic (29).
Most protein-binding HS sequences investigated so far do not seem to depend on the presence of unique components but are composed of the commonly occurring major HS building blocks. Such an arrangement does not exclude selectivity in binding. For instance, FGF-2 requires an N-sulfated sequence with a single IdoA 2-O-sulfate group for binding to HS (30,31). GlcN 6-O-sulfate residues are tolerated but not necessary for binding. By contrast, interactions with PDGF (32), hepatocyte growth factor (33), lipoprotein lipase (34), and herpes simplex gC glycoprotein (35) all depend on the presence of one or more GlcN 6-O-sulfate groups. This requirement was generally expressed through the preferential binding of oligosaccharides containing the trisulfated disaccharide unit, -IdoA(2-OSO 3 )-GlcNSO 3 (6-OSO 3 )-. Although effects of nonspecific charge interaction cannot be excluded, it is noted that the implicated disaccharide unit constitutes the main portion of heparin chains but is highly variable in HS species, where it occurs primarily in internal portions of NS domains (10,11). Presumably, the formation of the trisulfated disaccharide units in HS occurs by tightly regulated 6-O-sulfation of previously 2-Osulfated NS sequences (Fig. 1). Such regulation would nicely explain the differential effect of age on binding of PDGF versus FGF-2 to human aortic HS, as described above. Finally, certain biological effects of HS chains may require simultaneous binding of two proteins, e.g. FGF-2 and its receptor (36) or thrombin and antithrombin (4), to adjacent but structurally different saccharide domains. Such a requirement poses special demands on long range regulation of saccharide biosynthesis.
The lack of information concerning the role of NA/NS domains in protein binding presumably reflects the technical problems associated with the isolation of oligosaccharides containing such sequences.

Biosynthesis of Heparan Sulfate
The Enzymes-A GlcA-Gal-Gal-Xyl oligosaccharide sequence links the polysaccharide chain to a serine residue in the core protein. So far, only one of the enzymes engaged in the build-up of this region, the GlcA transferase I, has been cloned (37). This enzyme is distinct from the GlcA/GlcNAc copolymerase (38) that catalyzes the subsequent polymerization of the polysaccharide chain, just as the transfer of the first GlcNAc residue, i.e. the committing step toward HS biosynthesis, requires a separate enzyme (39). Although the precise relation between polysaccharide chain elongation and modification has not yet been defined, the processes appear to be at least partly coupled (3). A dramatic development of the research in this area may be expected, because the enzymes implicated with polymer modification in HS biosynthesis were all recently cloned. Importantly, a number of these enzymes appear to occur as multiple isoforms (products of different genes or splice variants of the same gene). The first modification enzyme, i.e. the N-deacetylase/N-sulfotransferase (NDST), which replaces the N-bound acetyl group of the GlcNAc residue with a sulfate group, presents at least three genetically distinct isoforms (40 -43), and the same applies to the GlcN 6-O-sulfotransferase (6-OST) (44). 2 Also the GlcN 3-O-sulfotransferase (3-OST), which catalyzes the concluding modification reactions (Fig. 1), occurs in different forms that display distinct substrate recognition properties (4,45). So far, only single forms of the GlcA C-5 epimerase (46) and the IdoA 2-O-sulfotransferase (2-OST) (47) have been described.
How do these results help to explain the "regulated diversity" displayed by HS synthesized by different cells and tissues? Studies involving the two first encountered NDST forms (encoded by separate genes (48)) may provide a clue. Although most tissues express both NDST-1 and -2 in significant amounts, heparin-producing mast cells show a predominance of NDST-2 (48), thus suggesting a role for the latter enzyme in the generation of NS domains (see Fig. 1). Accordingly, transfection of HS-producing cells with NDST-2 led to the formation of an almost exclusively N-sulfated polysaccharide (49), whereas overexpression of NDST-1 had little effect on the N-substitution pattern (50). In Fig. 1, NDST-1 is tentatively implicated with the formation of the NA/NS domains typically found in HS but not in heparin. The potential of enzyme polymorphism in controlling the NDST reaction as well as further downstream modifications, hence the structural diversity of HS chains, remain to be explored. Intriguing possibilities include populations of distinct OSTs capable of introducing O-sulfate groups at predetermined positions in NS or NA/NS domains ( Fig. 1) (see also Ref. 4). The amounts/proportions of different isoforms in a given cell, as controlled by gene expression and/or posttranslational degradation, thus may be of critical importance to the structure of the polysaccharide synthesized. Further progress in this area will require experiments with recombinant enzymes, including the generation of mice with targeted mutations in selected genes.
The Concerted Process-Understanding HS biosynthesis and its regulation will require further insight into the organization of the process beyond the level of individual enzyme reactions. The mode of interaction of the various biosynthetic enzymes with each other as well as their positioning within the Golgi complex also need to be elucidated, bearing in mind that substrate recognition by most of the enzymes generally depends on structural modifications introduced in previous reactions (3,12). The NDST, the C-5 epimerase, the 2-OST, and the 6-OST enzymes all appear to be type II membrane-bound proteins, contrary to the 3-OSTs that lack a hydrophobic sequence of sufficient length (45). NDST-1 and NDST-2 may be located in different Golgi compartments because their N-terminal regions, including the cytoplasmic tail and the transmembrane domain, show little homology (48), and this region is important for protein retention and localization (51).
Additional factors of potential importance include the availability of the sulfate donor, PAPS, as determined by the combined capacity of the sulfate-activating enzyme system and the PAPS transmembrane transport system (52). The sulfation patterns may change because of modulations of the intra-Golgi PAPS concentration as determined by the K m of the various sulfotransferases. Unoccupied sulfation sites have been arbitrarily indicated in Fig. 1, in recognition of such variability. Still uncharacterized inhibitors and activators may further influence the enzymes of the biosynthetic machinery, hence the structure of the HS product. NDST-2 thus depends on a cationic protein cofactor for expression of N-deacetylase activity (42), whereas NDST-1 apparently does not (53).
The overall kinetics of HS assembly, as determined by the membrane-bound state of the biosynthetic machinery, remains poorly understood. The fundamental importance of this aspect may be illustrated by observations relating to the GlcA C-5 epimerization reaction. It is commonly held that this reaction, freely reversible in the solubilized state, is promoted by Osulfation at C-2 of the IdoA product. However, studies using a heparin-producing mouse mastocytoma microsomal fraction suggested that extended polysaccharide chains were formed within less than a minute; moreover, each conversion of GlcA to IdoA was in effect irreversible, because no back epimerization could be demonstrated (12,54). Related to our current notion of HS structure this information would apply to the generation of an extended NS domain, presumably occurring within a single Golgi compartment. Does the formation of whole HS chains, with their various domains, involve different enzyme machineries in separate Golgi compartments of a single cell? The answer to this question will require further studies on membrane-bound enzyme systems in isolated defined Golgi fractions as well as in artificial membrane assemblies.
Role in Development-Uncovering the gene structures for the biosynthetic enzymes has recently provided striking evidence for the role of HS in embryonic development, as revealed through genetic screens in Drosophila. Most of these findings converge on growth factors belonging to the Wnt family and the role of HS chains attached to a defined core protein (uncovered as the dally mutant) in promoting their interaction with Frizzled type receptors. Mutants associated with distinct malformations (for references see Ref. 55) have been found to reside in genes encoding key enzymes in HS biosynthesis, including sugarless (defective UDP-glucose dehydrogenase, required to generate the UDP-GlcA precursor) and sulfateless (defective NDST). Moreover, the Drosophila segregation distorter protein (56) shows appreciable sequence homology with the mammalian IdoA 2-OST (47). Future work along this line will undoubtedly promote our knowledge not only of the biological functions of HS but also of its biosynthesis, in particular the roles of enzyme isoforms.