I. INTRODUCTION

Top Previous Next References

Proteoglycans, a group of glycoproteins that carry covalently bound sulfated glycosaminoglycan (GAG) chains, are molecules that have "come of age." Recognized in the early 1960s as important structural components of the extracellular matrix of cartilage, proteoglycans were once thought to be specific to that tissue. By now it has become clear that they are found in the matrices of all tissues, including the brain. The diversity of proteoglycans is dependent on differential expression of genes encoding core proteins, different exon usage of these genes, as well as variations in the length and types of GAG chains. A number of comprehensive investigations to the occurrence of proteoglycans in the central nervous tissue indicate that a great variety of individual proteoglycans exist in the developing as well as in the mature brain. Because neurons are ectodermal in origin, it is not surprising that their fate is intimately determined by interactions with the extracellular matrix. In the past 15 years more than 30 primary structures of proteoglycan core proteins have been elucidated, and their GAGs have been characterized. In central nervous tissues, the majority of the proteoglycans carry either chondroitin sulfate or heparan sulfate side chains (Table 1). Some of them are constituents of the extracellular matrix (e.g., hyalectans); others are bound to the cell surface through a glycosylphosphatidylinositol (GPI) anchor (glypicans) or intercalated into the cell membrane by a transmembrane domain (e.g., syndecans, NG2). In vitro, proteoglycans display affinity to a variety of ligands, including growth factors, cell adhesion molecules, matrix components, enzymes, and enzyme inhibitors. It seems likely that these interactions play also an important role in vivo, since the proteoglycans and their putative binding partners frequently colocalize during histogenesis of the central nervous system (CNS). Thus proteoglycans appear not only to function as structural elements supporting cells and providing tissue turgor but may also mediate events that form crucial steps during brain development.

In this review we present the various proteoglycan families and discuss their putative function in neural development.

	II. THE GLYCOSAMINOGLYCAN CHAINS

Top Previous Next References

GAGs are large unbranched polymers composed of ~20-200 repeating disaccharide units, which are usually attached to the core proteins through a serine residue and characteristic carbohydrate linkage regions (108). Depending on the disaccharide structures, they can be grouped into chondroitin/dermatan sulfate (CS/DS), heparin/heparan sulfate (HS), and keratan sulfate (KS) side chains (Fig. 1). Hyaluronan (HA), which is in contrast to the aforementioned GAGs a nonsulfated polymer of glucosamine (GlcN) and glucuronic acid (GlcA), exists as a protein free polysaccharide on cell surfaces and in the extracellular matrix (278). CS/DS are composed of alternating units of galactosamine (GalN) and either glucuronic or iduronic acid (IdoA), HS of GlcN, and GlcA or IdoA and KS of disaccharide units of glucosamine and galactose (Gal), respectively.

View larger version (28K): [in this window] [in a new window]   Fig. 1. Structure of disaccharide monomers forming the basis of the different glycosaminoglycan chains. Arrows with dashed lines refer to position of potential epimerization of glucuronic acid (GlcA) to iduronic acid (IdoA). Potential sites for O- and N-sulfations are indicated. The extent of modification may greatly vary. In heparan sulfate, the sequence of the modification reactions is as follows: 1) N-deacetylation/N-sulfation of GlcNAc, 2) epimerization of the GlcA to IdoA, 3) 2-O-sulfation of IdoA, 4) 6-O-sulfation of GlcNSO3, and 5) 3-O-sulfation of GlcNSO3 (129).

Diverse modifications through epimerases (DS and HS), combined N-deacetylase/N-sulfotransferases (only HS), and several O-sulfotransferases are responsible for a great part of the structural heterogeneity of the core protein bound GAG (129). There is increasing evidence that the cell- and tissue-specific expression of the modifying enzymes and the formation of specialized enzyme complexes controls the generation of microstructural domains within the GAG chains. These particular microstructures may participate in the specific binding of certain heparan sulfates to growth factors, protease inhibitors, and extracellular matrix molecules and hence may greatly influence the overall function of a particular proteoglycan (129, 133).

	III. THE CORE PROTEIN STRUCTURES OF CENTRAL NERVOUS SYSTEM PROTEOGLYCANS

Top Previous Next References

Hyalectans (94) or lecticans (222) are large aggregating proteoglycans that carry mainly CS side chains. Members of the hyalectan family are currently versican (also called PG-M) (251, 309), aggrecan (44), neurocan (211), and brevican (212, 300) (Fig. 2). Hyalectans are except for a GPI-anchored brevican splice variant (246) secreted molecules that take part in the formation of extracellular matrices. The modular core proteins of hyalectans are composed of an array of structural motifs widely used in hyaluronan binding proteins, in selectins, in cell surface and extracellular matrix molecules, and in other proteins (93). The characteristic features of versican, aggrecan, neurocan, and brevican are highly similar NH₂- and COOH-terminal sets of modules that appear in rotary shadowing electron micrographs as globular structures (168, 219). Poorly conserved GAG-carrying middle portions of variable length separate these globular domains. The homologous NH₂-terminal regions of hyalectans consist of an immunoglobulin domain followed by a hyaluronan-binding tandem repeat. In aggrecan, this repeat element is duplicated, giving rise to an additional globular domain with presently unknown function. The COOH-terminal globular structure includes epidermal growth factor (EGF)-like repeats, a C-type lectin domain, and a sushi (or CRP) element.

View larger version (16K): [in this window] [in a new window]   Fig. 2. Structural models of hyalectans. Wavy lines indicate chondroitin sulfate side chains. Positions of potential glycosaminoglycan (GAG) substitution of versicans, brevicans, and neurocan are deduced from the primary structures. The number of chondroitin sulfate side chains present in brain aggrecan is currently not known but appears to be significantly lower than the >100 GAG chains attached to cartilage aggrecan. Because the versicans, neurocan, and brevican form an almost perfect array of functionally related but differently sized molecules (49, 211, 300, 307), it seems conceivable that the spatiotemporally regulated expression of these hyalectans fine tunes the inhibitory capacity of specialized hyaluronan-rich matrices in the developing and mature central nervous system. GPI, glycosylphosphatidylinositol.

The high similarities among hyalectans are also reflected in the virtually identical organization of the different genes encoding brevican (212), neurocan (210), aggrecan (45), and versican (173). The intron-exon boundaries of these genes are highly conserved and coincide with the divisions between the similarity domains. It appears therefore likely that the different hyalectans have arisen from a common ancestor gene, which in turn may have been assembled from other genes by exon duplication and shuffling in the course of evolution (93).

Alternative splicing and transcription-termination add greatly to the structural diversity of hyalectans. Different isoforms of versican (49, 251, 307), aggrecan (8, 46, 63), and brevican (246) have been identified (Fig. 2). The mammalian versican splice variants result from alternative usage of the two large exons encoding the central GAG carrying core protein portions named GAG- and GAG-. Versican V0, the longest splice variant, contains both of these GAG attachment domains, whereas the smaller V1 and V2 isoforms only include the GAG- or the GAG- domain, respectively (49). The fourth versican splice variant (V3) is solely composed of the NH₂- and the COOH-terminal globular domains lacking all potential GAG attachment sites (307). As a peculiarity, two extra splice forms of versican/PG-M have been identified in the chicken (306). This additional variability is based on the existence of a so-called PLUS domain, a chicken-specific stretch of 138 amino acids located at the NH₂-terminal end of the GAG- domain. This PLUS domain is present in all V0⁺- to V3⁺-variants, whereas it is absent in the V1- and V3-isoforms due to the usage of an alternative 5'-splice donor site in the GAG- encoding exon. Because the versican PLUS domain bears a considerable amino acid similarity with the KS domain of chicken aggrecan, it seems conceivable that this extra module may carry KS side chains.

In contrast to versican, the central GAG domain-encoding exons seem not to be affected by alternative splicing processes in other hyalectans. However, different COOH-terminal core protein regions have been described for aggrecans and brevicans. Aggrecan isoforms containing one, two, or no EGF-like elements and also splice forms lacking the sushi domain have been identified by RT-PCR (8, 46, 63). Only the lectinlike domain seems to be invariably present in the aggrecan isoforms. Interestingly, the existence of some splice variants appears to be species dependent, since a functional exon encoding the first EGF-like domain has only been found in the human aggrecan gene (45, 62). An even more striking difference between splice variants has been observed in the COOH-terminal region of the two brevican core proteins. Because of alternative transcription termination, the COOH-terminal globular structure can be replaced by a short peptide module carrying a GPI anchor (246) (Fig. 2). The structure of this membrane-bound brevican isoform greatly contrasts the core proteins of the other hyalectan family members and therefore may point to a unique function of this particular splice variant.

CS as well as O- and N-linked oligosaccharide side chains play an important role in the overall structure of hyalectans. In particular, they seem to be responsible for maintaining an extended conformation of the central GAG attachment domain(s) mainly through charge repulsion (168, 219). In versicans, neurocan, and brevican, the number of GAG side chains per length of the GAG-carrying domains are rather constant. Estimated numbers of CS side chains are 17-23, 12-15, and 5-8 in versicans V0, V1, and V2, respectively (49), an average of 3 in neurocan (209, 211) and 0-5 in brevican, which appears to be a "part-time" proteoglycan (300). Furthermore, it seems rather unlikely that the smallest hyalectan, versican V3, carries GAG. Most of the above numbers are estimates based on putative carbohydrate attachment sites in the primary structures of the hyalectan core proteins. Nevertheless, the number, size, and composition of the carbohydrate substitution is not only influenced by the core protein structure; it may also greatly vary depending on the cell and tissue origin of a particular hyalectan. This holds especially true for aggrecan, which bears in cartilage ~100 CS and 30 KS side chains. Brain aggrecan, in contrast, carries on the same core protein substantially fewer and probably shorter CS side chains and appears to lack KS (125, 243). In addition to the GAGs, hyalectans contain N-linked and large numbers of sialylated O-linked oligosaccharides (175, 184, 209, 240). Whereas few N-linked oligosaccharides are distributed along the entire core proteins, O-linked oligosaccharides appear to be predominantly localized in the central region, conferring a rather rigid mucinlike structure to these hyalectan domains (219).

There is evidence for a proteolytical processing of hyalectans in vivo. Specific cleavage sites have been identified in the GAG attachment domains of neurocan (146, 155, 209, 211), brevican (300, 301), and in the interglobular domain (G1-G2) of aggrecan (235, 236). Interestingly, a consensus sequence for cleavage [E(G/S)E>(A/S)RG] could be identified in aggrecan and brevican (235, 236, 301), whereas the site in neurocan seems to be distinct (211). Based on the similarity, the existence of a specific "aggrecanase" has been postulated. Furthermore, several 40- to 70-kDa proteolytical products of versicans have been described. These fragments, named hyaluronectin (40) or glial hyaluronate-binding protein (GHAP) (13), contain all the NH₂-terminal globular domain of versican (41, 194) and eventually portions of the GAG- domain of versican V1 (193). Currently it is unclear whether these fragments that are mostly smaller than the expected size for the versican splice variant V3 are products of a physiological degradation process.

Glypicans form one of two large families of cell membrane-bound HS proteoglycans (Fig. 3) that are widely expressed in the CNS (36, 118). To date, complete primary structures of six different mammalian glypicans (38, 58, 185, 238, 268, 287, 288, 296) and one Drosophila homolog (dally gene product) have been determined (172). All glypicans are synthesized as precursor peptides with a secretory signal sequence at the NH₂ terminus and a hydrophobic stretch at the COOH-terminal end, which is replaced in the mature polypeptide by a covalently linked GPI anchor (Fig. 3). The predominant portion of glypicans consists of a domain that bears 14 highly conserved cysteine residues and displays no significant similarity to other proteins. Because nonreduced glypicans migrate on SDS-PAGE much faster than expected from their calculated sizes, it seems likely that these cysteines form intramolecular disulfide bonds and confer a rather compact shape to the core proteins. The potential HS binding sites are localized at the COOH-terminal end of this domain in close vicinity to the cell membrane. Furthermore, glypican-1 and -2 may carry an additional GAG chain closer to the NH₂ terminus of the core proteins. Estimated numbers of HS side chains range between two and five depending on the particular glypican core protein. Interestingly, glypican-5 overexpressed in COS-7 cells carries both HS and CS chains, an observation which has now to be confirmed with ex vivo glypican-5 (238). Also, the Drosophila homolog dally appears to carry HS (248). All glypicans, except for glypican-2 and -6, include furthermore a few potential N-glycosylation but little or no O-glycosylation sites. Based on the sequence similarities among the different core proteins, one can group glypican-1, glypican-2 (also known as cerebroglycan), glypican-4 (K-glypican), and glypican-6 together, whereas glypican-3 (OCI-5, MXR7) and glypican-5 show the closest resemblance.

View larger version (21K): [in this window] [in a new window]   Fig. 3. Models of the cell membrane-associated heparan sulfate proteoglycans of the syndecan and the glypican families. The length of core proteins has been adjusted according to the number of amino acids in the mature polypeptide. Although the core proteins of glypicans have been drawn in an extended conformation, they may assume a more compact shape due to the formation of intramolecular disulfide bonds. Globules indicate conserved cytoplasmic domains in syndecans, zig-zag lines denote GPI anchors of glypicans, and wavy lines represent GAG chains.

The chromosomal localization of five of the six glypican genes in the human genome has been determined. GPC1 is present on the long arm of chromosome 2 (2q35-q37) (285), whereas GPC5 and GPC6 are grouped together on chromosome 13q32 (185, 287, 288). Finally, GPC3 and GPC4 form a tandem array at position q26 of the X chromosome (199, 289). This locus has recently been linked to the Simpson-Golabi-Behmel overgrowth syndrome (199). The juxtaposition of these two genes suggests that not only functional inactivation of glypican-3 (90) but eventually also of glypican-4 is involved in the etiology of the disease (289).

Syndecans are type I transmembrane proteins that carry predominantly HS side chains (11, 26, 207) (Fig. 3). All four mammalian syndecans are expressed in developing and/or adult brain tissues. These are syndecan-1 (141, 237), syndecan-2 (formerly called fibroglycan) (37, 143, 198), syndecan-3 (N-syndecan) (27, 28), and syndecan-4 (ryudocan, amphiglycan) (39, 110). Syndecans are like glypicans evolutionary ancient molecules. Homologs have been identified in Drosophila melanogaster (264), in Caenorhabditis elegans, and Xenopus laevis (220, 272). The core proteins of syndecans are significantly shorter than their glypican counterparts. Nevertheless, it appears that the cysteine-free extracellular structures of syndecans would be more extended and not as compact as in glypicans. Interestingly, syndecan core proteins migrate on SDS-PAGE significantly slower than predicted from the amino acid sequence. It is currently not clear whether this difference arises from poor SDS binding or from the formation of SDS-resistant multimers.

There is little similarity among the ectodomains of the different syndecan family members, whereas a rather high resemblance of the transmembrane domains has been observed. However, the hallmarks of the syndecan core proteins are their relatively short cytoplasmic domains (Fig. 3). They include a virtually identical stretch of 10 amino acids [RM(R/K)KKDEGSY] just adjacent to the transmembrane domain, followed by a region that is variable with regard to the different syndecans but conserved among the species homologs. Finally, an EFYA tetrapetide sequence, present in all currently identified vertebrate and invertebrate syndecans, forms the COOH-terminal end of the core protein (207).

Although the overall sequence similarity of the syndecans is low, comparisons still display a closer relationship between syndecan-1 and -3 and between syndecan-2 and -4. The differences between the two groups are also reflected in the location of the potential GAG attachment sites. In syndecan-1 and -3, these sites are clustered near the NH₂ terminus and in the vicinity of the transmembrane domain. Additional binding sites closer to the NH₂-terminal end of the ectodomain are only present in syndecan-2 and -4. The two GAG-binding clusters in syndecan-1 and -3 are separated by a proline- and threonine-rich sequence, which has in syndecan-3 a mucinlike character (27). Syndecan-2 and -3 seem to carry exclusively HS, whereas mixed HS/CS side chains may be present on syndecan-1 and -4 (204, 256). A certain preference of the membrane proximal sites for CS attachment has been reported for syndecan-1 (111).

The release of the syndecan ectodomains by membrane shedding may play an important role in the function of these proteoglycans. Shedding of syndecans has been mainly observed in cell culture experiments but seems to reflect a physiological process (27, 50, 103, 271).

The genes encoding syndecans are located on different chromosomes (263). Currently there are structural data available for the syndecan-1, -3, and -4 genes (7, 27, 83, 290). They all consist of five exons that are similarly organized advocating for a common ancestral gene. As expected, major differences are found in exon 3, the exon that encodes most of the core protein portion between the GAG-binding clusters in the ectodomain.

Apart from hyalectans, glypicans, and syndecans, there are a number of proteoglycans that either are structurally unique or belong to protein families that in general do not carry GAGs. One proteoglycan of the latter group is the receptor-type protein-tyrosine phosphatase RPTP (also known as RPTP) (72, 86, 142). Three isoforms of RPTP that arise from alternative exon usage have been identified (114, 124, 149). At least two of them carry CS chains (9, 10, 136, 149, 255). The largest RPTP isoform is a membrane protein composed of a carbonic anhydrase domain, a fibronectin type III repeat, and a structurally novel cysteine-free stretch in the extracellular portion, which most likely carries the GAG chains (114). Two intracellular tyrosine phosphatase domains follow the single transmembrane-spanning region. A shorter membrane-bound isoform of RPTP lacks 860 amino acids of the extracellular portion including most of the GAG attachment sites. There is evidence that this isoform is mostly devoid of CS side chains and hence has to be considered a part-time proteoglycan (124, 176, 228). The third RPTP splice variant, which has been named phosphacan, contains the entire set of extracellular domains including the GAG attachment region but terminates shortly before the transmembrane domain (149). Hence, phosphacan is a secreted proteoglycan. Two major glycoforms of phosphacan exist: one that carries exclusively CS side chains and the other that is modified with CS and KS chains (phosphacan-KS) (149). Also Lewis-X oligosaccharides appear to be present, since phosphacan is recognized in the postnatal and adult CNS by the Lewis-X-specific antibody FORSE-1 (1, 2).

NG2 is a unique transmembrane CS proteoglycan that has been identified on the surface of O2A glial progenitor cells during brain development (123). Its multidomain core protein is composed of a large extracellular portion, a single transmembrane domain, and a short cytoplasmic stretch (177). The ectodomain can be subdivided into a compact disulfide bonded element at the NH₂ terminus, an extended GAG attachment region, and a second cyteine-rich sequence close to the membrane-spanning domain. Four type A and two type B internal repeats have been identified. Similarities of the type A repetitive element with the putative Ca²⁺-binding region of the cadherins have been observed. NG2 may carry two to three GAG chains and bears a number of potential N-glycosylation sites.

Agrin is a large basement membrane-associated HS proteoglycan (43). Originally being detected at the neuromuscular junction, it has also been located in various nonmuscular tissues including the brain, where it seems to be predominantly expressed during embryonic development (66, 134, 183, 218, 221, 224). The agrin core protein has like perlecan (93), the other basement membrane proteoglycan, a complex modular structure. Agrin is composed of a NH₂-terminal laminin-binding domain, followed by eight follistatin-like (or Kazal-like) elements, two laminin EGF-like domains, another follistatin-like module, a serine/threonine-rich region, a SEA element (denotes sea urchin sperm protein, enterokinase, agrin domain), a second serine/threonine-rich region, one EGF-like element, a laminin G-like domain (G1), two EGF-like modules, the second laminin G-like domain (G2), another EGF-like module, and finally the third laminin G-like domain (G3) at the COOH-terminal end (42, 224, 281). There are at least three sites where alternative splicing may occur through insertion of small peptides (42, 55, 221, 223, 280). Two of them are localized within the G2 and the G3 domain. Variations at these sites denoted A and B in chicken and y and z in rodent agrin, respectively, give rise to multiple isoforms. It appears that only neurons express isoforms that include inserts at these splice sites (84, 134, 151, 221, 257, 269, 276). As mentioned above, agrin carries HS side chains. Potential GAG attachment domains are present between follistatin-like modules 7 and 8 and in the first serine/threonine-like region, whereas more COOH-terminal consensus sequences seem not to be substituted (42, 279).

Several isoforms of the Alzheimer amyloid precursor protein (APP) and two variants of the amyloid precursor-like protein 2 (APLP2) carry a single CS side chain (253, 274, 275). The multiple splice forms of these type I transmembrane proteins arise from alternative splicing of three exons in APP and two exons in APLP2 (107, 112, 200, 231). The CS attachment site is formed near the membrane-spanning domain by elimination of an 18 (APP) or 12 (APLP2) amino acid encoding exon, bringing a Gly-Ser-Gly sequence into close vicinity of acidic amino acid residues (186, 275). All other splice variants that include the exon lack the GAG side chain. The distribution of APP and APLP is rather ubiquitous. It appears, however, that the CS-carrying isoforms are predominantly expressed by glial cells in the CNS (231, 233).

Little is currently known about two other proteoglycans, named testican/SPOCK (18, 31) and neuroglycan C, that have been identified in brain tissues. Testican/SPOCK is a secreted mixed CS/HS proteoglycan (18, 19, 31). Its core protein consists of BM-40/osteonectin/SPARC-like sequences including a follistatin-like module and an extracellular calcium-binding domain followed by a thyroglobulin type I module (3, 18, 109, 167). The two GAG attachment sites carrying HS and CS chains are localized close to the COOH-teminal end of the core protein (18, 19). The testican/SPOCK cDNA clones have originally been isolated from a testicular cDNA library; it has, however, become evident that the predominant location of testican expression is the CNS (18). Neuroglycan C appears to be a brain-specific type I membrane proteoglycan that carries several CS side chains in the NH₂-terminal half of its core protein (295, 303). It displays some similarities to short peptide sequences in aggrecan, including an EGF-like sequence close to the transmembrane domain and portions of the GAG attachment regions. The expression of neuroglycan C in rat brain is associated with neuronal cells in the developing cerebral cortex.

	IV. INTERACTION OF PROTEOGLYCANS PRESENT IN THE DEVELOPING BRAIN WITH EXTERNAL LIGANDS

Top Previous Next References

HS GAGs in the nervous system are predominantly expressed on proteoglycans of either the transmembrane syndecan family (26) or the GPI-anchored forms, the glypican family (118). They are proposed to influence cell-environment interactions by binding to a heterogeneous number of HS binding molecules such as growth factors, matrix ligands, and cell surface molecules (Table 2). Molecules whose cellular effects on the development of neuronal connections are modulated by or dependent on HS include the neural cell adhesion molecule (NCAM), slit-1 and slit-2 (necessary for the development of the midline glia and commissural axon pathways), members of the fibroblast growth factor (FGF) family, Wingless/Wnt, transforming growth factor- (TGF-), and Hedgehog families (14, 34, 126, 181, 206, 217, 225, 304) and heparin-binding growth factors, such as pleiotrophin/HB-GAM (215).

The interaction of HS proteoglycan (HSPG) in FGF signaling is particularly well documented, and several mechanisms of action have been postulated. The biological activity of FGFs depends on their ability to bind cell surface or extracellular matrix HS (239). Cells that express the high-affinity FGF receptors (FGFR) but lack HS neither bind nor respond to their ligands (206, 304). HSPG can stabilize FGF by protecting them from proteolysis and thermal denaturation (226, 227, 291). Interaction with proteoglycans can lead to an increase in the local concentration of FGF, thus enhancing their affinities for their appropriate signaling receptors (205, 291). Another function of HS action is to oligomerize ligands and thereby facilitate receptor dimerization and subsequent signaling (188, 205, 239, 262). Because many growth factors such as FGF bind to the extracellular domain of their corresponding high-affinity receptors as monomers that are incapable to induce receptor dimerization, it has been proposed that a multimeric HS-FGF complex is required to mediate proper intracellular signaling (144). Furthermore, evidence was presented for a large amount of HSPG in brain tissue that was not attached to membranes, consistent with the shedding of HSPG that has been observed with cultured cells (38, 50, 296). Membrane shedding would provide a mechanism for terminating functional activity that is dependent on attachment to the plasma membrane. This could include cell-cell adhesion activity or binding events that were dependent on cytoskeletal attachment, or generated intracellular signals through the cytoplasmic domain. Taken together, HSPGs can be considered as essential coreceptors for certain factors interacting with a much lower affinity than the signaling receptors. Specific structural aspects, including the sulfation of HSPG, are required for the proper interaction of HPSG with FGF (74, 201). Thus the regulation of GAG biosynthesis may greatly influence the functions of HSPG. There is evidence that syndecans and glypicans not only stimulate FGF/receptor interactions and signaling, but depending on the cellular origin or the level of expression they can also inhibit the activities of FGF (6, 140, 267). The level of expression of HSPG core proteins in a given tissue and the ability of cells to synthesize HS side chains of a defined structure are developmentally regulated (20, 36, 181). Moreover, it was shown that one particular HSPG is able to rapidly switch its potentiating activity from FGF-2 to FGF-1 during the transition from neuronal precursor cell proliferation to neuronal differentiation (181), suggesting that a given HSPG can intricately regulate FGF signaling in the developing brain. The differential binding characteristics of specific HS structures might explain why some brain-derived HS potentiate the biological activity of FGF whereas others prevent the binding of FGF to its receptor.

Another class of soluble proteins with high affinity for HSPG is the newly identified family of developmentally regulated, secreted heparin-binding proteins. It comprises the heparin-binding growth-associated molecule (HB-GAM), also designated pleiotrophin (PTN), and the closely related molecule midkine (MK) (153, 214, 215). They are suggested to play important roles in the regulation of mitogenesis, angiogenesis, and neurite and glial process outgrowth activities in vitro, but their functional roles in vivo are as yet unknown.

Recently, genetic studies in Drosophila have demonstrated a role for heparin-like GAG in the reception of the Wingless (Wg) signal (14, 75, 76, 128, 282). Embryos that are either deficient in an enzyme catalyzing the production of a GAG precursor or that have been treated with an enzyme that specifically degrades HS GAG exhibit phenotypes reminiscent of loss-of-Wingless signaling. It has also been suggested that changes in HSPG expression may influence the distance over which a morphogen such as Wg can act, restricting it to a short-range signal in some circumstances but allowing it to serve as a more widespread gradient morphogen in others (75). To what extent these findings are comparable to developmental processes in the mammalian nervous system remains to be determined.

One of the most prominent ligands of hyalectans is hyaluronan, which binds with high affinity to the tandem repeat in the NH₂-terminal globular domain (79, 122, 209). Furthermore, neurocan and phosphacan bind tightly to a variety of immunoglobulin superfamily NCAM, including Ng-CAM/L1, N-CAM, Nr-CAM, and TAG-1/axonin-1 (60, 159, 161). There is also detectable but considerably less binding of neurocan and phosphacan to contactin (191). Other high-affinity ligands of these proteoglycans are tenascin-C (73, 158) and tenascin-R (4, 157). Binding of neurocan and phosphacan to both tenascins has the same apparent dissociation constant of ~3 nM, but the affinities are significantly lower than the corresponding values determined for their interactions with NCAM. However, the NCAM and tenascins represent the major ligands that have been identified up to now. No significant binding was observed to a number of other cell surface and extracellular matrix proteins such as N-cadherin, fibronectin, vitronectin, merosin, thrombospondin, EGF, and basic FGFR, and collagens I-VI and laminin (157). Recent findings suggest that there are presumably two binding sites, in the NH₂-and COOH-terminal portions of neurocan, that are involved in its interactions with tenascin-C and tenascin-R. Moreover, the high-affinity binding requires the presence of both sites in the full-length proteoglycan core protein (157, 208). In addition to neurocan and phosphacan, other members of the hyalectan family were also shown to bind to tenascin-R via their lectinlike domain (4) and are able to bind to sulfated glycolipids, such as sulfatides in a calcium-dependent manner (166).

Contrasting to HSPG, CS proteoglycans (CSPG) were not assumed to bind to soluble heparin-binding proteins such as pleiotrophin or FGF in earlier reports. However, studies by Maeda et al. (136) then demonstrated that the HB-GAM/PTN is a ligand of phosphacan. This interaction appeared to be mediated only in part by the CS chains. Later it was shown that not only phosphacan but also neurocan can interact with HB-GAM and amphoterin, another heparin-binding protein prominently expressed in embryonic brain (152, 157). As opposed to the data of Maeda et al. (136), binding of neurocan and phosphacan to HB-GAM and amphoterin was largely abolished by chondroitinase treatment in this study, indicating that CS is the likely mediator of their binding to these proteins (157). Furthermore, the core protein of phosphacan was shown to bind FGF-2 with high affinity and to potentiate its mitogenic effect (163). In contrast, neurocan seems to bind FGF-2 but has no modulatory effect on its activity (163). With the use of a solid-phase binding assay, the large transmembrane CSPG NG2 has been shown to bind to collagens V and VI through the central domain of its core protein (22, 266, 277) and to interact with platelet-derived growth factor receptor . Although most studies of growth factor interactions with proteoglycans have concerned HS proteoglycans and more specifically their GAG chains, it would appear that the interactions of CS proteoglycans with differentiation factors might also play a significant role in developmental processes of the nervous system.

	V. FUNCTIONAL IMPLICATIONS OF PROTEOGLYCANS DURING BRAIN DEVELOPMENT

Top Previous Next References

In the developing rodent nervous system, patterns of glypican expression correspond to areas of high mitotic activity, where heparin-binding proteins, such as FGF, are known to play important regulatory roles. During early rodent embryogenesis, glypican-1 is the most prominent family member. Its transcripts are primarily seen in ventricular zones of the developing forebrain, midbrain, and hindbrain areas, sites that contain proliferating neuronal and glial progenitors (130). Most neuronal precursor cells do not express glypicans-2, -4, or -5 except in the ventricular zone of the cerebral wall where glypican-4 has been detected (296) and in the ganglionic eminence where glypican-5 is expressed (238). Two mechanisms of how glypicans may regulate cell division and survival have been proposed: 1) glypicans may mediate growth factor interactions with their cell surface receptors and hereby initiate an intracellular signal cascade or 2) glypicans may directly act on the nuclear level as suggested by the nuclear localization of glypican-1 in adult spinal neurons (127).

Compelling evidence, however, that glypicans are involved in regulation of cell growth and differentiation in vivo has come from genetic studies in Drosophila. Mutations in the dally locus, a gene coding for a Drosophila glypican, cause alterations in specific patterns of cell division in the larval brain (172). One site where dally is expressed is the lamina precursor cells in the morphogenic furrow of the eye imaginal disc. These cells differentiate into lamina neurons upon input from innervating photoreceptor axons. Detailed analysis of the phenotype revealed that loss of functional dally protein delays lamina precursor cells to enter their final rounds of cell division. Similarly, the epithelium containing the anterior outer proliferative center and the eye disc are disordered. Several nonneuronal structures, such as the antennae and genitalia, are also morphologically abnormal. These observations imply that the absence of this glypican results in a delay or even arrest of precursor cell progression in specific stages of cell cycle. At present, however, little is known about the cellular and molecular mechanisms of action. Glypicans may predominantly perform coreceptor functions involving the presentation or delivery of soluble, heparin-binding proteins (e.g., growth factors) to their receptors. Recent studies show that dally is indeed part of a receptor complex that regulates certain signaling pathways that control proliferation and differentiation during Drosophila development (128, 282). One could therefore envision that loss of glypican function would disrupt one or presumably even more important growth factor signaling pathways. Evidence for this hypothesis comes again from studies on Drosophila. Dally mutants show defects in signaling mediated by Decapentaplegic (Dpp), a homolog of the bone morphogenic protein (BMP)-2/4 and a subfamily of the TGF- superfamily (98). Normal synchronization of the cell cycle of lamina precursor cells depends on the requirement of Dpp (192), consistent with a role for dally as a Dpp coreceptor. However, how cell surface HSPG affect the nature of the Dpp signal remains unknown.

Recently, the human X-linked Simpson-Golabi-Behmel syndrome (SGBS) was shown to result from loss-of-function mutations in the glypican-3 (GPC3) gene (199). This syndrome is characterized by pre- and postnatal overgrowth as well as by various other abnormalities, including increased risk of embryonic tumors such as Wilms tumors of the kidney. Furthermore, the high susceptibility for neuroblastomas indicates an involvement of glypican-3 in neurogenesis. SGBS shares many clinical features with the Beckwith-Wiedemann syndrome (284, 297). This disorder is thought to be in most cases the result of the overexpression of insulin-like growth factor (IGF)-II, a growth factor that can also act as a survival factor (78). It has been proposed, therefore, that glypican-3 is involved in the downregulation of IGF-II activity (90, 199). However, transfection studies of rat glypican-3 (OCI-5) into human cell lines did not confirm a direct interaction of IGF-II (261). On the other hand, recent studies demonstrated that glypican-3 can induce apoptosis in a cell line-specific manner, an effect that could be rescued by IGF-II, indicating that it is possible that glypican-3 regulates IGF-II activity by interacting with other molecules involved in IGF-II signaling (67). Induction of apoptosis required the anchoring of glypican-3 to the cell membrane, while GAG chains did not seem to be necessary, thus implying that shedded glypican (38, 296) is unlikely to cause apoptosis. Furthermore, the finding that the GAG chains are not required raises the question of whether or not the interaction with heparin-binding proteins is involved for the induction of apoptosis in this system, or if the glypican-3 core protein can exert these functions. Moreover, given the high abundance of the nonglycanated form of glypican-3 in several cell types, one possibility to be explored is that only the nonglycanated form of glypican-3 is responsible for this apoptosis-inducing activity. In addition, the findings on the possible functional interaction between glypicans and growth factors of the TGF- superfamily raise the possibility that SGBS patients may be defective in proper TGF- signaling. In contrast to Drosophila, where Dpp promotes cell division, TGF- is known as a negative growth factor for most vertebrate cells (145). It seems therefore reasonable to assume that the loss of GPC3 gene function in SGBS patients could compromise TGF-/BMP-mediated cell cycle arrest, leading to overgrowth and tumor progression.

Consistent with their proposed roles as modulators of cellular activities, syndecans exhibit a complex pattern of cell- and developmental-specific expression. Analysis of syndecan expression in a number of tissues and cell lines led to the conclusion that virtually all cells express at least one of the four forms of syndecan, with most cells expressing even multiple forms. The most abundant form found in the adult mammalian brain, spinal cord, and peripheral nervous system (PNS) is syndecan-3 (N-syndecan), reflecting a rapid increase in its expression around birth, with a peak on postnatal day 7, followed by a decline to low levels in the adult CNS (27) where it remains concentrated in axons (88). The period of maximal synthesis of syndecan-3 correlates with the period of glial cell differentiation, myelination, and formation of neuronal connections; however, the functional significance remains to be established.

During embryonic development of the rat, phosphacan mRNA is largely confined to areas of active cell proliferation such as the ventricular zone of the ganglionic eminence, the embryonic precursor of the basal ganglia, the ependymal layer surrounding the central canal of the spinal cord (51), and the ventricular zone of the cerebrum (163). Precursor cell division in proliferative zones of the developing CNS depends on cell-cell interactions and on external factors such as members of the FGF family. The presence of FGF-2 was recently shown to correlate with the expression of phosphacan, which can potentiate the mitogenic activity of FGF-2 (163). Thus the appearance of phosphacan in dividing precursor cells suggests that its interaction with FGF-2 may play a role in the regulation of proper cell division or differentiation.

The formation of the PNS relies on the spatiotemporally coordinated migration of neural crest (NC) cells, a process that appears to depend on the heterogeneity in the extracellular matrix that forms their migration substrate. It is well known that various embryonic tissues act as "barriers" to migrating NC cells such as the caudal part of the sclerotome, the dermomyotome, and the perinotochordal mesenchyme, thereby directing NC cells into specific migration pathways. Based on their spatially and temporally regulated expression pattern within such barrier tissues, CS proteoglycans have been suggested as possible candidates that inhibit NC migration (119, 182, 196). Furthermore, many in vitro studies have suggested that CSPG, CS, or the core proteins do not support neuronal migration when applied as a substrate. In chick embryos, immunohistochemistry revealed a restricted appearance of versican V1/V0 to the caudal sclerotome and the perinotochordal mesenchyme (119). Once NC cell migration is completed, expression of versican becomes associated with prechondrogeneic areas. In addition, purified versican does not support the attachment of NC cells in vitro (197), underlining the notion that the close correlation of versican expression within areas that are avoided by migrating NC cells may be of functional importance. Analysis of a number of mouse mutants showing abnormalities of NC migration as part of their phenotype, further support the importance of CSPG in controlling NC migration. Splotch mice, resembling the human Waardenburg syndrome phenotype, harbor mutations in the Pax3 gene and have been useful in addressing the functional importance of versican V0/V1 during development. These mice exhibit NC-related abnormalities including pigmentation defects, reduced or absent dorsal root ganglia, diminished Schwann cell numbers, and failure of NC cell colonization of the outflow tract of the heart, resulting in persistent truncus arteriosus (30). However, the cranial skeleton develops normally, indicating that Pax3 is affecting only a subset of NC cells. Interestingly, splotch NC cells fail to colonize target tissues, although they initiate migration in vivo and appear to migrate indistinguishable from wild-type NC cells in vitro. Thus neural crest abnormality in splotch may reside not in the NC cells themselves, but rather in the extracellular environment through which they migrate. Recent analysis revealed a marked overexpression of versican mRNA in the pathways of NC cell migration, but also at ectopic sites, suggesting that the overabundance of versican prevents NC cells in Splotch homozygotes from migrating to their final target areas (80). Abnormal accumulation of proteoglycans or other extracellular matrix constituents have also been described in other mouse mutants with NC cell migration defects, such as patch, white spotting, lethal spotting, or piebald lethal (97, 169, 190, 273). However, in many of these mutants, the precise types of proteoglycans have not been characterized, or the upregulation of proteoglycans as a possible secondary effect due to abnormal development of crest-derived structures cannot be excluded.

Neuronal migration is a prerequisite for the development of layered structures in the CNS. Migrations of CNS neurons are guided by a radial glial fiber system, in which postmitotic neurons generated in the proliferative ventricular zone translocate to their final positions along the radial glial fibers, which provide them with a scaffold for their oriented movement. The process of neuronal migration is a complex developmental program and depends on orchestration of multiple molecular events involving cell adhesion molecules, extracellular matrix molecules, ion channels, and cell surface receptors (202).

In the early cortical development, the proteoglycan-type protein tyrosine phosphatase RPTP- and its extracellular splice variant phosphacan have been localized along radial glial fibers and on migrating neurons, suggesting that this receptor-type phosphatase is involved in neuronal migration (23, 135). RPTP-/phosphacan has been shown to bind various cell adhesion and extracellular matrix molecules such as F3/contactin, N-CAM, L1, TAG1, and tenascin (73, 159, 191). Furthermore, one of the most prominent factors that interacts with RPTP- is pleiotrophin, a member of a newly identified family of developmentally regulated, secreted heparin-binding proteins that stimulates neurite outgrowth in vitro. The CS side chains and the protein portion of RPTP- together constitute the binding site of pleiotrophin, and various GAGs inhibit the interaction of RPTP- with pleiotrophin (137). In the early cerebral cortex, pleiotrophin is synthesized by radial glial cells and is deposited on radial glial fibers during neuronal migration (147, 148, 216). The localization of all RPTP- isoforms along radial glial fibers and on migrating neurons raises the possibility that the ligand-receptor mechanism between RPTP- and pleiotrophin plays a role in the neuronal migration. This notion was recently supported by a study showing that antibodies against RPTP- or addition of GAGs can perturb pleiotrophin-induced migration of neurons (139).

A large number of developmentally expressed proteoglycans of both neuronal and astrocytic origin have been implicated in the regulation of axonal pathfinding. Many of these studies rely on the fact that during development, the tightly regulated expression pattern of HSPG, such as glypican-1 and -2 or syndecans, are closely associated with neuritogenesis. Shortly after neurons become postmitotic they transiently express glypican-2 (cerebroglycan), which is exclusively present in brain. Glypican-2 is found at stages when axons are actively growing but disappears around the time axons arrive at their targets. Unlike other glypicans, it is completely absent in the adult brain (96, 268). Given that glypicans can interact in vitro with a variety of factors known to modulate axonal growth (Table 2), it is tempting to speculate that glypicans may act as coreceptors for these guidance cues, but specific functional evidence for this hypothesis remains to be seen. In support of this idea, it was demonstrated that enzymatic removal of endogenous HS in cockroach embryos perturbed the directed growth of pioneer fibers (293). Similarly, retinal axon elongation in the optic tract of embryonic Xenopus was retarded (292). Although a specific HSPG has not been identified in these systems as yet, the results suggest a role for HS, which exists in vivo only as HSPG. In vitro studies provided insights that HSPG can support neurite growth by activating growth-enhancing molecules, such as laminin, NCAM, or members of the MK family. They may all stimulate neurite growth when complexed with HSPG or by presenting growth factors such as FGF or hepatocyte growth factor to their appropriate receptors (239, 304). To provide directional information to a growing axon in vivo, a growth factor would have to be expressed in a spatially restricted fashion. In particular, FGF show extremely localized expression patterns during specific stages of the developing brain. Such a restriction could be accomplished by binding of the secreted growth factors to HSPG present on cell surfaces and in the ECM in the local environment. In regulating the diffusion range and the availability of growth factors, HSPG could determine the efficacy of growth factors as axonal guiding cues. Thus the steepness of a vectorized gradient would depend on the diffusion rate of the secreted factors. Factors that are immediately tethered after secretion would result in short-range guidance cues. Factors that diffuse for some distance before becoming tethered would produce shallow gradients providing more long-range guidance information.

Recently, syndecan-3 (N-syndecan) was isolated from brain neurons and perinatal rat brain (104, 213) and identified as a possible receptor for HB-GAM, an extracellular matrix-associated neurite outgrowth-promoting factor localized to basement membranes and axonal pathways in embryonic and early postnatal tissues (165, 216). N-syndecan was shown to be strongly expressed in the perinatal rat brain (27, 28), where the developmental expression pattern in Northern and Western blots appears almost identical to the expression pattern of HB-GAM (180). HB-GAM-dependent neurite outgrowth of syndecan-3-transfected cells is inhibited by selective inhibitors of the Src family kinases, and recent biochemical data imply that syndecan-3 mediates the neurite growth-promoting signal of HB-GAM to the cytoskeleton of neurites via activation of the cortactin-src kinase pathway (106). Whether syndecan-3 may also transmit signals on its own in neuronal systems has to be investigated (310).

In contrast to HSPGs, numerous authors have argued that CS and CSPGs are inhibitory to axon growth, whereas others, using different assays, have argued that these molecules are growth promoting. Immunostaining for CS often localizes to territories thought to act as barriers to migrating neurons or extending axons. These include the posterior sclerotome (119, 182, 196), the dorsal midline of the spinal cord and optic tectum (99), and regions of developing retina (21, 260). In vitro, CSPG (48, 60, 138, 174, 195, 240, 258), the isolated core proteins of CSPG (48, 138), and CS by itself (24, 286) have been shown to inhibit cell migration or neurite outgrowth on defined growth-promoting substrata. Data from organotypic cultures also suggest that inhibition of CSPG expression can allow neuronal migration or outgrowth into previously avoided territories (21, 56). Despite these data, it is clear that tissues that strongly express CS do not always exclude the entry of axons (182, 298), and in some cases, CS immunostaining coincides with developing axon pathways (12, 150, 250). Furthermore, several in vitro studies suggest that CSPG (53, 270), CS (54, 117), and isolated core proteins (91) promote rather than inhibit neurite outgrowth.

Good examples for the versatility of certain proteoglycans are the two CSPG neurocan and phosphacan. They inhibit neurite growth of embryonic chick neurons promoted by Ng-CAM (60, 159), even after removal of CS. However, in another type of assay, phosphacan was shown to promote neurite outgrowth of cortical neurons on fibronectin substrates (138). Moreover, the same molecules occur in anatomical regions such as the developing cortex, spinal cord, and cerebellum before and during their robust invasion by axonal fibers. Increasing evidence indicates that neurocan is involved in the delineation of efferent and afferent intracortical pathways during early development of the rat cortex. The most interesting feature of neurocan expression in the developing cortex appears in the subplate region (154, 164, 184). Previous experiments in fetal cat provided evidence that subplate neurons or molecules produced by these cells are important for determining the thalamocortical pathway (65). Whereas axons leaving the cortical plate of the cortical anlage cross the subplate region and extend within the underlying intermediate zone, a region that contains much less CSPG, thalamocortical afferents travel specifically within the subplate and form transient synapses before they enter the cortical plate (12, 249, 305). Although neurocan has been shown to inhibit axonal growth in vitro, it does not seem to act as a barrier for these fibers but may instead even serve as a guidance cue to distinguish between efferent and afferent pathways. Similarly, although phosphacan is not restricted to a certain area but more diffusely distributed in the cortical anlage (61, 135), it does not seem to affect axonal extension. These findings suggest that certain CSPG, such as neurocan and phosphacan, cannot be simply classified as neurite growth supportive or inhibitory, but that rather the spatial and temporal regulation of expression and the possible interaction with other molecules, such as cell adhesion molecules, may define their properties on neurite growth. Tissue culture experiments indicate that CSPG can bind to a variety of ECM constituents and cell adhesion molecules (see Table 2), resulting in a modification of their growth-promoting propensities. At early stages of cortical development, restricted localization of L1 immunoreactivity was found on thalamocortical afferent axons, whereas TAG-1 was localized in cortical neurons and fibers of the intermediate zone, but not the subplate (61). Binding of neurocan with the cell adhesion molecule L1 has been shown in vitro (60), thus supporting the hypothesis that heterophilic interaction of these two molecules may help to guide thalamocortical axons along their pathway in vivo. However, in vitro, using embryonic chick neurons, both neurocan and phosphacan inhibited neurite growth promoted by Ng-CAM/L1 (60, 159). Single matrix cell culture assays, however, provide poor models for the actual complexity of the extracellular environment of the growing neurites in vivo, where presumably multiple guidance signals impinge on an extending axon. Interestingly, expression of HB-GAM in the cortical subplate/intermediate zone is found before the arrival of the first syndecan-3-expressing thalamic afferents (105), whereas tenascin-C expression in the same areas coincides with the maturation of the afferent and efferent fiber tracts (69, 250). In addition to syndecan-3 binding, HB-GAM has been shown to bind to phosphacan and with somewhat lower affinity also to neurocan (137, 157), whereas both molecules bind tenascin-C with high affinity (162). Thus HB-GAM as well as tenascin-C could be candidate molecules that might influence the ability of CSPG-containing areas to support neurite extension. Hence, the relative concentration of various ECM constituents and adhesion molecules in the local environment and the expression of the corresponding receptor or binding molecules on a growing axon may determine its growth behavior.

Another region enriched in CSPG that was suggested to serve as barriers for extending axons is the roofplate in the developing spinal cord (21, 258). This region, containing specialized glial cells, is believed to prevent nerves from crossing the midline of the spinal cord. CS immunoreactivity at the spinal cord roofplate is high at a time when dorsal column fibers elongate near to but distinctly avoid the midline, and it decreases at a later stage, when axons of the dorsal spinal commissure grow across the midline (259). Phosphacan, but not neurocan, was found to be present in the roofplate glial cells (154). It is possible that the selective expression of molecules such as phosphacan in a region containing low levels of growth-promoting molecules may act as a barrier to axonal growth, preventing the invasion of certain fibers into the roofplate. A related set of observations has been made at the tectal midline in the developing optic tectum, a midline structure broadly analogous to the specialized glia seen in the roofplate. As developing retinal axons elongate into the superior colliculus, some of them course near the specialized raphe glia of the tectal midline, but normally do not cross the midline, unless the raphe glia have been previously removed or damaged (242). In contrast, at an earlier stage of development, intertectal afferent fibers freely cross the tectal midline (99). Interestingly, however, a recent study in the developing hamster did not reveal an increased production of specific CSPG at the midline of the developing tectum but suggested that the concentration of CS at the midline is a reflection of a higher rate of synthesis or sulfation of GAG by midline cells (85). Thus an increased GAG biosynthesis at the tectal midline could enable midline cells to concentrate and present various kinds of molecules that act as guidance cues.

The remarkable functional adaptability of the mammalian CNS resides in the ability of neurons to modify their synaptic efficacy in response to activity. The best studied system is the N-methyl-D-aspartate receptor-dependent long-term potentiation (LTP), a phenomenon suggested to contribute to the shaping of synaptic connectivity in the developing and adult CNS (15). Previous studies have shown that modulations that disturb cell-matrix interaction and thereby affect synapse morphology should interfere with the induction or maintenance of LTP. Indeed, antibodies against NCAM or removal of the associated carbohydrate polysialic acid prevents LTP (132, 170), whereas FGF enhances LTP (95, 247). In addition, the stabilization of LTP was suppressed in an early phase of LTP by recombinant HB-GAM in hippocampal slice cultures (121). Interestingly, these molecules are known to bind to HSPG, such as syndecan-2 and -3 (26) and also to the CSPG phosphacan (135, 157). Although the expression of phosphacan in adult hippocampus has not been studied, syndecan-2 (89), syndecan-3 (120), and agrin (33) were identified as prominent HSPG in adult hippocampus, thus suggesting that they could function as coreceptors for at least some of the aforementioned molecules. A recent study shows that addition of soluble syndecan-3 to the dendritic CA1 area prevents tetanus-induced LTP, suggesting that it is one of the molecules that helps to modulate cell-matrix interactions associated with synaptic plasticity (120). It appears that HS can regulate LTP through multiple ligand interactions and that the cellular consequences of a high-frequency stimulus might be critically dependent on the balance between different heparin-binding molecules available.

Interestingly, the COOH-terminal EFYA tetrapeptide motif common to all syndecans was found to interact with PDZ domain proteins, such as syntenin (71), and the postsynaptic protein CASK (32, 89). CASK/LIN-2 is a member of the membrane-associated guanylate kinase homologs (MAGUK) (87), a large family of proteins typically localized at cell junctions such as the synapse, and thought to coordinate the localization of multiple molecules (35). Thus synaptically localized syndecans may be involved in binding to certain growth factors, thereby potentiating their action on synaptic receptor kinases, and to induce cytoskeletal changes. Although syndecan-2 is specifically localized in synapses in the adult brain, its late appearance during synapse formation in vivo (88) and in vitro (52) suggests that syndecan-2 may be involved in morphological maturation of dendritic spines rather than in specifying the formation of spines. Evidence for this hypothesis was recently provided by a study showing that overexpression of syndecan-2 in hippocampal neurons at an early stage of culture (before endogenous snydecan-2 expression occurs) accelerated the maturation of dendritic spines but did not affect the number of synapses or spines. The mechanisms through which syndecan-2 induces the formation of morphologically mature spines is presently not known, but it seems likely that the interaction with PDZ domain proteins or CASK is directly involved.

In the PNS, the HS proteoglycan agrin is one of the most important molecules in the developing and regenerating neuromuscular junction (221). It is responsible for the formation of aggregates containing the ACh receptor (AChR) and other molecules. Multiple isoforms of agrin are generated by alternative splicing, and the presence of an 8-, 11-, or 19 (8 + 11)-amino acid insert at splice site B is required for agrin's AChR aggregation activity. Less clear is the role of agrin in the CNS. Besides motor neurons, agrin is expressed in many different types of neurons throughout the CNS, including CA1 and CA3 pyramidal cells in the hippocampus and granule cells in the dentate gyrus, where agrin expression is regulated during development and by neuronal activity (33). Despite its widespread expression in the nervous system, it seems unlikely that agrin plays in the CNS a similar role in synaptogenesis as in the PNS. Immunohistochemical data of the developing chick retina provide evidence that the different agrin splice variants are differentially distributed in the retina, but not restricted to the synapse layer (113). Moreover, the presence of distinct agrin isoforms around retinal ganglion cells before synapse formation implies that agrin may have additional yet unknown functions in the CNS.

More recent studies imply a possible role of CSPG in restricting structural plasticity in the adult brain. Inhibitory properties for neurite outgrowth of cerebellar granule cells were shown for purified brevican (299) and the brain specific splice variant of versican V2 (240) (Fig. 4). The rather late onset of expression of both molecules in the postnatal CNS (160, 299) suggests that their capacity to inhibit axonal growth may be linked to the stabilization of the mature neuronal network, rather than to axonal guidance processes. Studies of the developing rat cerebellum demonstrated the expression of brevican on the surface of astrocytes that form neuroglial sheaths surrounding glomeruli in the protoplasmic islets of the granule cell layer (299). As mossy fibers form synapses with other neurons of the cerebellum within such glomeruli, it was suggested that brevican may restrict the infiltration of axons and dendrites into maturing glomeruli (299). Immunohistochemical analysis demonstrated that brevican and hyaluronan are also present in perineuronal nets in a number of nuclei of postnatal rat brain, often colocalized with tenascin-R (77). Although the physiological relevance of such perineuronal nets is mainly unknown, their postnatal appearance suggests a role in preventing the establishment of new synaptic contacts. The assembly of a hyalectan-associated matrix on the surface of mature neurons may provide such a barrier against further synapse formations. Similarly, the expression of versican V2 in white matter areas, and its presence in CNS myelin (174, 240), suggests a role in restricting structural plasticity and regeneration of CNS fiber tracts. The capacity for plasticity and regeneration in the CNS decreases during postnatal development (59, 116), a process that coincides in time with the formation of myelin (101). Interestingly, prevention of myelin formation in the spinal cord results in the persistence of the sprouting capacity in adult rats (244, 283). Thus versican V2 may contribute to the limited structural plasticity of the adult mammalian CNS.

View larger version (111K): [in this window] [in a new window]   Fig. 4. A: schematic representation of neurite outgrowth from retinal explants of embryonic day 7 chick embryos in a stripe-choice assay. Test substrates and neurite-promoting substrates are coated in an alternating stripe pattern on glass coverslips. B: demonstration of neurite outgrowth inhibition by the chondroitin sulfate proteoglycan versican V2 in this assay. Neurites extend on the outgrowth-promoting substrate laminin-1 (20 µg/ml coating concentration) () but avoid the stripes containing a mixture of versican V2 (70 µg/ml) and laminin-1 (100 µg/ml) (+). C: alternating stripes of laminin-1 alone (20 µg/ml and 100 µg/ml) have no influence on the direction of neurite extension (for details, see Ref. 240).

	VI. CONCLUSION

Top Previous Next References

The identification and primary structure determination of proteoglycan core proteins have significantly contributed to elucidate their possible functions during brain development. In parallel, many biochemical and immunohistochemical studies have shown that proteoglycans participate in multifactorial mechanisms modulating cell-cell and cell-matrix interactions. It has become clear that most of these interactions are tightly controlled in a spatiotemporal expression of the core proteins and by the regulated biosynthesis of specific carbohydrate structures. Thus the functions of defined proteoglycans appear to be confined to specific areas and developmental stages of the nervous system. In particular, the interactions with multiple ligands conferring different properties provide a means by which proteoglycans can modulate a wide spectrum of regulatory processes, such as neurogenesis, neuronal migration, axonal pathfinding, and synaptic plasticity. Further insights into the physiological role of individual proteoglycans in brain development will be provided by the recently initiated analysis of transgenic and gene knockout animal models.