Structural and Functional Characterization of Oversulfated Chondroitin Sulfate/Dermatan Sulfate Hybrid Chains from the Notochord of Hagfish

Oversulfated chondroitin sulfate (CS)/dermatan sulfate (DS) hybrid chains were purified from the notochord of hagfish. The chains (previously named CS-H for hagfish) have an average molecular mass of 18 kDa. Composition analysis using various chondroitinases demonstrated a variety of d-glucuronic acid (GlcUA)- and l-iduronic acid (IdoUA)-containing disaccharides variably sulfated with a higher proportion of GlcUA/IdoUA-GalNAc 4,6-O-disulfate, revealing complex CS/DS hybrid features. The hybrid chains showed neurite outgrowth-promoting activity of an axonic nature, which resembled the activity of squid cartilage CS-E and which was abolished fully by chondroitinase ABC digestion and partially by chondroitinase AC-I or B digestion, suggesting the involvement of both GlcUA and IdoUA in neuritogenic activity. Purified CS-H exhibited interactions in a BIAcore system with various heparin-binding proteins and neurotrophic factors (viz. fibroblast growth factor-2, -10, -16, and -18; midkine; pleiotrophin; heparin-binding epidermal growth factor-like growth factor; vascular endothelial growth factor; brain-derived neurotrophic factor; and glial cell line-derived neurotrophic factor), most of which are expressed in the brain, although fibroblast growth factor-1 and ciliary neurotrophic factor showed no binding. Kinetic analysis revealed high affinity binding of these growth factors and, for the first time, of the neurotrophic factors. Competitive inhibition revealed the involvement of both IdoUA and GlcUA in the binding of these growth factors, suggesting the importance of the hybrid nature of CS-H for the efficient binding of these growth factors. These findings, together with those from the recent analysis of brain CS/DS chains from neonatal mouse and embryonic pig (Bao, X., Nishimura, S., Mikami, T., Yamada, S., Itoh, N., and Sugahara, K. (2004) J. Biol. Chem. 279, 9765–9776), suggest physiological roles of the hybrid chains in the development of the brain.

Pivotal functions of glycosaminoglycan (GAG) 1 side chains of proteoglycans (PGs), especially heparan sulfate (HS), have been implicated in biological processes such as cell adhesion, proliferation, and differentiation and tissue morphogenesis (1). Chondroitin sulfate (CS) and dermatan sulfate (DS) are also found ubiquitously in the extracellular matrices and at cell surfaces, being main components of cartilage and skin, respectively. CS consists of repeating disaccharide units of -4Gl-cUA␤1-3GalNAc␤1-, whereas DS is an isomeric form of CS and is formed from precursor CS through the action of glucuronyl C5 epimerase, thus consisting of disaccharide units of -4Gl-cUA␤1-3GalNAc␤1-and -4IdoUA␣1-3GalNAc␤1-in varying proportions (2). These disaccharide units are modified during chain elongation by specific sulfotransferases at C-2 of GlcUA/ IdoUA and/or C-4 and/or C-6 of GalNAc in various combinations, producing characteristic sulfation patterns critical for binding to various functional proteins displaying enormous structural diversity, comparable with that of HS, by embedding multiple overlapping functional sequences (3). Sulfation profiles of GAGs change during development (4,5). Growing evidence indicates the involvement of CS and DS in the signaling of various heparin-binding growth factors and cytokines (6 -8).
We and others have shown the importance of this class of molecule, from simple chondroitin involved in cell division of a nematode (9) to differentially oversulfated CS-D and CS-E involved in neuroregulatory functions (10 -12) and the binding of growth factors in mammalian systems (13). While pursuing the critical structural elements in the CS variants, we became * This work was supported in part by the Science Research Promotion Fund of the Japan Private School Promotion Foundation and Grant-inaid for Exploratory Research 15659021 and the National Project on Functional Glycoconjugate Research Aimed at Developing New Industry of the Ministry of Education, Science, Sports, and Culture of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This article is dedicated to the memory of Prof. Nobuko Seno. ** Present address: School of Health Sciences, University of Occupational and Environmental Health, Kita-Kyushu 807-8555, Japan.
aware of vital contributions of IdoUA to biological activities. Neuritogenic activities were shown for various IdoUA-rich oversulfated DS chains from marine animals (14), and not only neuritogenic but also growth factor binding properties were revealed for the CS/DS hybrid chains attached to phosphacan/ DSD-1-PG from neonatal mouse brains (14,15) and for the free hybrid chains isolated from embryonic pig brains (16), both of which contain an appreciable proportion of functional IdoUA.
The neurite outgrowth-promoting activities of the CS/DS hybrid chains and oversulfated CS/DS chains are in contrast to the conventional concept that CS chains are intrinsic inhibitory components for axonal growth and path finding of various neurons (17,18). The seemingly discrepant observations are most likely attributable to the structural changes during development as demonstrated for CS/DS hybrid chains of embryonic and adult pig brains (16). Enzymatic removal of CS chains permits axonal regeneration after nigrostriatal tract axotomy and spinal cord injury (19,20) and reactivation of ocular dominance plasticity in the adult visual cortex (21) and is therefore a promising clinical strategy. We are searching for CS, DS, or hybrid chains with neuritogenic activities that have potential for therapeutic applications to neuronal diseases and injuries.
Here, we characterized GAGs from hagfish notochord, originally termed CS-H for hagfish (22). Notochord is a connective tissue that supports the neural tube and that induces the formation of the central nervous system during the development of vertebrates. CS-H contains the characteristic oversulfated units IdoUA␣1-3GalNAc(4S,6S) and IdoUA(2S)␣1-3GalNAc(4S,6S), where 2S, 4S, and 6S represent sulfate at C-2, C-4, and C-6, respectively (22,23), hence regarded as DS, and exhibits axonic and dendritic neuritogenic activities for mouse hippocampal neurons (14). In view of the importance of the hybrid nature of the functional mammalian CS/DS chains in mouse (14,15) and pig (16) brains, we searched for CS/DS hybrid chains in CS-H and demonstrated axonic neuritogenic and binding activities for a number of growth factors and a few major neurotrophic factors expressed in the brain.
Extraction and Purification of CS-H-CS-H was isolated from the notochord of hagfish (Eptatretus burgeri) as detailed previously (22). Briefly, acetone-dried notochord was subjected to Pronase digestion. The proteins were removed by precipitation with 20% trichloroacetic acid, and GAGs were precipitated with 2 volumes of ethanol containing 2.5% calcium acetate and 0.25 M acetic acid. The GAG mixture was fractionated on a Dowex 1-Cl Ϫ column and eluted stepwise by increasing the NaCl concentration of the eluents. The fraction obtained by elution with 2.0 M NaCl was used for further characterization and subjected to treatment with freshly prepared nitrous acid (pH 1.5) according to the procedure of Shively and Conrad (26) to remove HS. After the treatment, nitrous acid was neutralized by addition of 0.5 M Na 2 CO 3 , and the resultant HS fragments were separated from CS-H by passing the treated sample through a Sephadex G-50 column (56 ϫ 1 cm) with 50 mM pyridine acetate buffer (pH 5.0) as eluent at a flow rate of 0.6 ml/min. Finally, the CS-H preparation was freed of hydrophobic peptides by passing it through a C 18 cartridge.
Molecular Mass Determination-Molecular mass was determined using a Superdex 200 column (10 ϫ 300 mm) calibrated with known molecular mass markers, including dextran preparations (average molecular masses of 65.5, 37.5, and 18.1 kDa), HS from bovine intestinal mucosa (average molecular mass of 7.5 kDa), and heparin from porcine intestinal mucosa (average molecular mass of 6 kDa) (27). V 0 and V t were determined using dextran (molecular mass of 170 -200 kDa) and NaCl, respectively. Dextrans were monitored by the orcinol method for neutral sugars (28). The purified CS-H preparation (13 g as uronic acid) was loaded onto the column and eluted with 0.2 M ammonium acetate at a flow rate of 0.3 ml/min; the fractions were collected at 3-min intervals, evaporated to dryness, and reconstituted in 100 l of water. An aliquot was taken for estimating GAG using DMMB according to the procedure of Chandrasekhar et al. (29), except that the absorbance was measured at 525 nm.
Determination of the Digestibility of the Purified CS-H Preparation by Various Chondroitinases-The purified CS-H preparation (1.3 g as uronic acid) was digested with 10 mIU of chondroitinase ABC (30), 5 mIU of chondroitinase AC-I (31), 5 mIU of chondroitinase AC-II, or 2 mIU of chondroitinase B (32). After each enzymatic treatment, the digest was reconstituted in 20 or 50 l of distilled water, and an aliquot was taken to estimate the resistant structure of CS-H by complexation with the metachromatic dye DMMB, which complexes with sulfated GAGs and long oligosaccharides, but not with short oligosaccharides. Briefly, 35 l of 0.05 M acetate buffer (pH 6.8) and 200 l of DMMB solution were added to a 5-10-l aliquot of the above digest, and the absorbance was measured at 525 nm within 1 min (29). The loss of reactivity toward the dye was checked after each digestion, and the amount remaining was calculated based on the absorbance value using the calibration curve obtained with varying amounts of standard commercial DS (0.2-1.0 g) derived from porcine skin. The amount of GAG before digestion was taken as 100%.
Neurite Outgrowth Promotion Assay-The neurite outgrowth-promoting activity of the purified CS-H preparation was assayed as reported previously (14). Briefly, the CS-H preparation (0.7 g as uronic acid) or an equivalent amount of its digest with chondroitinase ABC, B, or AC-I diluted with phosphate-buffered saline was coated on plastic coverslips, which had been precoated with poly-DL-ornithine at 37°C overnight. Hippocampal neuronal cells established from embryonic day 18 rat brain were plated at a density of 10,000 cells/cm 2 in Eagle's minimal essential medium containing N2 supplements (Invitrogen, Tokyo), 0.1 mM pyruvate, 0.1% (w/v) ovalbumin, 0.29% L-glutamine, 0.2% sodium hydrogen carbonate, and 5 mM HEPES. The cultures were maintained in a humidified atmosphere at 37°C with 5% CO 2 for 24 h, after which the cells were fixed using 4% (w/v) paraformaldehyde, and the neurites were visualized by immunochemical staining using antibodies directed against microtubule-associated protein-2 and neurofilament. The immunostained cells on each coverslip were scanned and digitalized with a ϫ20 objective lens on an Olympus BX 51 optical microscope equipped with an Olympus HC-300Z/OL digital camera. The photographs were analyzed using morphological analysis software (Mac SCOPE, Mitany Corp., Tokyo). The length of the longest neurite, with at least one process chosen at random being longer than the cell body, was determined for 100 cells. At least three independent experiments per parameter or condition were carried out.
Biotinylation of the Purified CS-H Preparation and HS-Purified CS-H or HS was dissolved in 100 mM MES (pH 5.5) at a concentration of 2 mg/ml. To this solution were added 50 mM biotin-LC-hydrazide freshly dissolved in dimethylsulfoxide and 0.5 M 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (Pierce) dissolved in MES. The mixture was incubated overnight at room temperature with continuous shaking. Excess biotinylating reagents were removed by dialysis in Spectrapor molecular porous membrane tubing (molecular mass cutoff of 3.5 kDa; Spectrum Medical Industries, Inc., Laguna Hills, CA) against several changes of phosphate-buffered saline.
Immobilization of Purified CS-H and HS on a Sensor Chip-A streptavidin-coated sensor chip (BIAcore AB) was conditioned using 1 M NaCl and 50 mM NaOH for 1 min three times according to the manufacturer's instructions. Biotinylated CS-H or HS (3.3 g as uronic acid) in phosphate-buffered saline was perfused and allowed to interact with the sensor's surface for 3 min in phosphate-buffered saline. Immobilization was confirmed by the increase in response units.
Interaction Analysis-An interaction analysis was carried out in the BIAcore J system using the CS-H-immobilized sensor chip. Growth factors (200 ng each) were checked individually for binding to immobilized CS-H in running buffer (10 mM HEPES-NaOH (pH 7.4), 0.15 M NaCl, 3 mM EDTA, and 0.005% Tween 20). The flow rate was kept at a medium pace of 30 l/min according to the manufacturer's protocol. Each growth factor was allowed to interact with immobilized CS-H for 3 min, which constituted the association phase. Dissociation of the growth factor, constituting the dissociation phase, was allowed to go on for an additional 2 min, after which the sensor chip was regenerated by injecting 1 M NaCl for 2 min. Neurotrophic factors (100 ng each) were tested for binding to immobilized CS-H or HS at a high flow rate of 60 l/min, and the time allowed for the association and dissociation phases was 3 and 2 min, respectively.
To determine the kinetics of the binding of various growth factors to CS-H, each growth factor in varying concentrations was allowed to interact with immobilized CS-H. Kinetic parameters (k a , k d , and K d ) were determined by collectively fitting the overlaid sensorgrams locally using the 1:1 Languimuir binding model with mass transfer of the BIAevaluation 3.1 software. Interaction analyses of CS-H and HS with BDNF and GDNF for determination of kinetic parameters were carried out by global fitting of the sensorgrams obtained by employing high flow rates using the model mentioned above. High flow rates were used because of the high mass transfer effects observed at medium flow rates, due to which overlaid sensorgrams could not be fit with the 1:1 Languimuir binding model with mass transfer.
Disaccharide Composition Analysis-A composition analysis of purified CS-H was carried out after digesting 0.5 g (as uronic acid) with 10 mIU of chondroitinase ABC or 1.0 g (as uronic acid) with 5 mIU of chondroitinase AC-I or 1 mIU of chondroitinase B for 1 h at 37 or 30°C (for chondroitinase B) as described above. The digested products were then derivatized with 2-AB (33). Excess 2-AB was removed by repeated extraction with a 1:1 (v/v) water/chloroform mixture, and the water phase was dried. The residue was then reconstituted in 400 l of 16 mM NaH 2 PO 4 solution, and a 200-l aliquot was analyzed by anion-exchange HPLC on an amine-bound silica PA-03 column (33).

Gel Filtration Analysis of the Chondroitinase Digests of the CS-H Preparation on a Superdex
Peptide Column-The purified CS-H preparation (0.17 g as uronic acid) was digested with chondroitinase ABC, AC-I, or B, and an equal amount was digested sequentially with chondroitinase B followed by chondroitinase AC-I as described above. The digests were individually labeled with the fluorophore 2-AB and processed as described above. Each digest was made up with 400 l of 0.2 M ammonium bicarbonate containing 7% 1-propanol, and a 200-l aliquot was analyzed by gel filtration on a Superdex peptide column using the above-mentioned solvent as eluent at a flow rate of 0.4 ml/min.

Effects of Digestion of CS-H on Its Ability to Bind Growth Factors-
The binding of various growth factors to immobilized CS-H was evaluated based on the ability of exogenously added CS-H or its digested products to competitively inhibit the binding. Purified CS-H (1.7 g as uronic acid) was digested with chondroitinase ABC (10 mIU), AC-I (5.0 mIU), or B (2 mIU), and a 0.12-g equivalent (as uronic acid) of each digest was dispensed into individual Eppendorf tubes and dried. Control tubes without CS-H were prepared in a similar way to normalize any nonspecific inhibition. Before injection, the tubes were reconstituted in running buffer and incubated with 50 ng of the growth factor to be tested for 10 min at room temperature. Injection was carried out for 3 min at a medium flow rate (30 l/min), and the response during the association phase was compared with that exhibited by the growth factor only.

Isolation and Purification of CS-H from Hagfish
Notochord-In vertebrates, the notochordal primodium plays a central role in morphogenetic movement during gastrulation to induce the central nervous system. In cyclostomes, to which hagfish belong, the notochord is not atrophied during evolution and is available in a sufficient amount for preparing GAG chains, although PGs in notochord have not been well characterized in terms of the core proteins. CS-H from hagfish notochord was isolated as described (22). Briefly, it was obtained by Pronase digestion of the acetone-dried notochord, followed by trichloroacetic acid precipitation for removal of proteins and ethanol precipitation of GAGs. It was then fractionated on a Dowex 1-Cl Ϫ column by stepwise elution at varying concentrations of NaCl. A unique predominant disaccharide unit (IdoUA␣1-3GalNAc(4S,6S)), representing 68% of all disaccharides, was previously revealed in the major fraction eluted at 3.0 M NaCl from a Dowex column, in addition to the minor units IdoUA␣1-3GalNAc(6S) and IdoUA␣1-3GalNAc(4S) (22), and was named the H or iE unit (where "i" stands for iduronic acid) (7). In this study, the 2 M NaCl-eluted fraction, representing ϳ35% of total CS-H, was characterized as having less sulfated CS/DS hybrid chains. Because the preparation showed trace amounts of HS disaccharides upon HPLC analysis after heparitinase digestion (data not shown), it was purified further by subjecting it to nitrous acid treatment (26) to remove HS. Peptides were removed by passing through a Sep-Pak C 18 cartridge with water as eluent. The purified final preparation gave a single band between pig skin CS-B and shark cartilage CS-C upon electrophoresis on a cellulose acetate membrane (data not shown), which disappeared completely after digestion with chondroitinase ABC, but only partially after digestion with chondroitinase AC-I or B, suggesting the presence of both CS and DS chains or CS/DS hybrid chains in the CS-H preparation.
Determination of the Molecular Mass of CS-H-The molecular size of purified CS-H was determined by gel filtration HPLC using a calibrated Superdex 200 column (27). The CS-H preparation eluted as a single fairly symmetrical peak when monitored using the dye DMMB (Fig. 1), giving an average molecular mass of ϳ18 kDa. This is similar to the molecular masses of porcine skin DS (19 kDa) and porcine intestine DS (21 kDa), but slightly larger than the molecular mass of DS from eel skin (14 kDa) (34).
The CS/DS Hybrid Nature of CS-H-The digestibility of the purified CS-H fraction was evaluated using bacterial chondroitinases ABC, AC-I, AC-II, and B by taking advantage of the ability of DMMB to form a complex with the GAGs (29). DMMB reportedly forms a complex with intact GAGs or with sulfated oligosaccharides long enough to be recognized by the dye, but not with shorter oligosaccharides. The digestibility of GAGs was therefore measured in terms of the elimination of reactivity with DMMB as a result of fragmentation of CS-H by the action of the respective chondroitinases (35). Chondroitinases AC-I and AC-II specifically cleave N-acetylgalactosaminidic linkages with GlcUA, whereas chondroitinase B cleaves Nacetylgalactosaminidic linkages with IdoUA. Chondroitinase ABC splits both types of linkages. It should be noted that, unlike the others, chondroitinase AC-II exhibits an exolytic action (36). As shown in Fig. 2, the CS-H preparation was fragmented almost completely by chondroitinase ABC and partially by chondroitinase AC-I or B (58 and 75%, respectively). It is likely that chondroitinase B produces disaccharides and nonretainable short oligosaccharides to a greater extent than chondroitinase AC-I. The oligosaccharides produced by the former and latter enzymes likely contain internal GlcUA and IdoUA residue(s) as a predominant uronic acid, respectively. Treatment with chondroitinase AC-II resulted in only 15% digestion, probably reflecting its exolytic action.
Neurite Outgrowth-promoting Activity of the CS-H Preparation-To evaluate the biological activity of the purified CS/DS hybrid CS-H preparation, the neurite outgrowth-promoting activity was tested in embryonic day 18 rat hippocampal neuronal cells (37). The CS-H preparation exhibited activity that was principally axonic in nature (Fig. 3A) and weaker than that shown by CS-E derived from squid cartilage (Fig. 3C). Control cells (cultured on coverslips coated with poly-DL-ornithine alone) showed no significant promotion of neurite outgrowth (Fig. 3B). Digestion of the CS-H preparation with chondroitinase ABC abolished the neurite outgrowth-promoting activity by 97%, whereas digestion with chondroitinase B or AC-I resulted in a loss of the neuritogenic activity of ϳ65 and 60%, respectively, as calculated based on the residual activity remaining after subtraction of the background activity due to poly-DL-ornithine, with the mean length of the neurites obtained using the undigested CS-H preparation taken as 100% (Fig. 3C). Analysis of the data was also carried out by enumerating the number of primary neurites/cell for 100 randomly selected cells, and the results show that the neuronal cells cultured on the CS-H substratum had, on average, 1.5 neurites/ cell compared with 3 or 1.9 neurites observed for cells grown on the substratum coated with poly-DL-ornithine alone or CS-E, respectively (Fig. 3D), which is in contrast to the formation of multiple dendritic neurites in addition to a single axonic neurite observed for the CS-H preparation eluted at 3.0 M NaCl from a Dowex column (14), which contained almost exclusively IdoUA and little GlcUA (23).
Specific Interactions of CS-H with Various Heparin-binding Growth Factors-Molecular interaction experiments were car-ried out using the BIAcore system. The purified CS-H preparation was biotinylated using the carboxyl groups of the uronic acid moieties. Approximately 2% of the uronic acid residues  (Fig. 4). All growth factors tested bound to the immobilized CS-H preparation to some extent, except FGF-1, which did not show any significant binding. Among the growth factors, FGF-18 exhibited the greatest response, followed by FGF-16 and PTN.
A kinetic analysis to determine association, dissociation, and dissociation equilibrium constants (k a , k d , and K d , respectively) was carried out for all growth factors except FGF-1. Growth factors in varying concentrations were perfused onto the CS-H-immobilized sensor chip, and analysis was carried out by fitting the overlaid sensorgrams with the 1:1 Languimuir binding model with mass transfer of the BIAevaluation 3.1 software (Fig. 5). All growth factors tested bound to CS-H, with FGF-16, FGF-18, and PTN having the greatest affinity as evident from the responses observed at equivalent concentrations, reflecting the different stoichiometry of the binding of the growth factors (Fig. 5, C, D, and F, respectively; see also "Discussion" for the stoichiometry). The affinity of the binding was particularly strong for FGF-10, FGF-18, MK, and PTN (Fig. 5, B, D, E, and F, respectively), and the K d values for these four growth factors could not be determined as accurately because of the negligible dissociation of the bound growth factors, which is characteristic of such high affinity binders and may reflect the possibility that CS/DS-PGs act as receptors or coreceptors like HS-PGs to present them to the cell-surface receptors. Alternatively, growth factor-oligosaccharide complexes may be released enzymatically from the CS/DS-PGs (see "Discussion"). FGF-2, FGF-16, HB-EGF, and VEGF 165 (Fig. 5, A, C, G, and H, respectively) exhibited higher association and dissociation rates than the other growth factors tested (Table I), as shown in the sensorgrams. The different characteristics of the kinetics of binding suggest possible distinct roles for CS/DS chains in the regulation of growth factor signaling (see "Discussion"). DS from porcine skin shows negligible binding to immobilized MK compared with CS-H (12), although the 3 M NaCl-eluted fraction was used previously. On the other hand, DS from porcine skin binds PTN with a K d of 51 nM when DS is immobilized on the sensor's surface (39).

Specific Interactions of the CS-H Preparation with Neurotrophic Factors-Neurotrophic factors play important roles
as key regulators of cell fate and cell shape in the vertebrate nervous system (40). The possibility that DS acts as a binding partner for neurotrophic factors was also examined. A neurotrophin family member (BDNF), a transforming growth factor-␤ superfamily member (GDNF), and a neuropoietic cytokine (CNTF) were used. It has been reported that GDNF signaling requires HS (41). In the present study, the ability of these factors to bind CS-H was evaluated compared with their binding to HS from bovine intestinal mucosa, the structure of which has been characterized previously to have a sulfate/disaccharide ratio of 0.41 and an IdoUA/GlcUA ratio of 35:65 analogous to HS derived from aorta, lung, and kidney (42), hence representing such common structural features. BDNF and GDNF showed significant binding to immobilized HS as well as CS-H, whereas the binding of CNTF to both HS and CS-H was negligible (Fig. 6).
The overlaid sensorgrams for the binding of BDNF and GDNF to immobilized HS and CS-H are shown in Fig. 7, and they illustrate high affinity interactions. A kinetic analysis was carried out to calculate the k a , k d , and K d values for these interactions (Table II). The affinity of BDNF for immobilized CS-H was 360-fold higher than that for HS, whereas the affinity of GDNF for CS-H was 10-fold higher than that for HS, suggesting that the interactions are biologically significant and suggesting the possible involvement of CS, DS, or CS/DS hybrid chains in the functional expression of neurotrophic factors. Notably, the fast dissociation of these neurotrophic factors from HS and the negligible dissociation from CS-H suggest distinct roles for these glycans in the regulation of protein functions, as in the case of growth factors (see "Discussion").
To identify the isomer (GlcUA or IdoUA) of the hexuronic acid of the parent disaccharides from which the ⌬Di-diS E unit was derived, a composition analysis was carried out by anionexchange HPLC after digestion with either chondroitinase AC-I or B, which specifically targets the N-acetylgalactosaminidic linkage with GlcUA or IdoUA, respectively. Digestion by both enzymes gave rise to ⌬Di-diS E , suggesting the presence of both disaccharide units, GlcUA␤1-3GalNAc(4S,6S) (E unit) and IdoUA␣1-3GalNAc(4S,6S) (iE unit), revealing the intrinsic microheterogeneity in this complex molecule. Digestion by chondroitinase AC-I or B resulted in ⌬Di-0S, ⌬Di-4S, and ⌬Di-6S and a number of oligosaccharides as well (Fig. 8, B and  C), indicating the existence of GlcUA-GalNAc (O unit), GlcUA-GalNAc(4S) (A unit), IdoUA-GalNAc(4S) (iA unit), GlcUA-GalNAc(6S) (C unit), and IdoUA-GalNAc(6S) (iC unit), respectively (3,7,14). The presence of ⌬Di-0S derived from the iO unit in the chondroitinase B digest was unexpected because the enzyme has been considered to recognize certain sulfation patterns (32). It apparently originated from ⌬Di-4S by the action of 4-sulfatase contaminating the enzyme preparation, as was determined later using another chondroitinase B preparation free of 4-sulfatase (see the legend to Fig. 8). The multiple oligosaccharides formed by digestion with chondroitinase AC-I or B indicates the distribution of both GlcUA and IdoUA along the chain, which displays the complex CS/DS hybrid structure. The oligosaccharides formed as a result of chondroitinase AC-I digestion would represent the IdoUA-containing moieties, whereas those formed by chondroitinase B digestion would contain GlcUA to a larger extent and IdoUA to a smaller extent, reflecting the specificity of chondroitinase B, which appears to require hitherto uncharacterized sulfation patterns of GalNAc and IdoUA residues adjacent to the cleavage site (32). The exact IdoUA content in the CS-H polymer chains remains to be determined.
To determine the distribution of the IdoUA-and GlcUA-   containing moieties in the purified CS-H preparation, the oligosaccharides formed as a result of digestion with chondroitinase AC-I or B or by sequential digestion with chondroitinases B and AC-I were individually labeled with 2-AB and analyzed by gel filtration on a Superdex peptide column. Digestion with either one of the above-mentioned enzymes resulted in disaccharides and oligosaccharides (Fig. 9, A and B). Sequential digestion by chondroitinase B followed by chondroitinase AC-I resulted in cleavage of many (but not all) oligosaccharides to disaccharides ( Fig. 9C; note that the vertical axis is different from those in Fig. 9, A and B), giving credence to the proposal that the digested oligosaccharides harbor GlcUA-containing disaccharide units. Some hexa-and octasaccharides could not, however, be broken down through sequential digestion, in contrast to the complete digestion achieved with chondroitinase ABC (Fig. 9D), presumably reflecting the resistant structure of certain uniquely sulfated domains. It should be noted that, because the oligosaccharides were labeled at the reducing ends with 2-AB, the peak areas reflect the molar ratios of the oligo-saccharides formed, but do not correspond to their amounts in terms of weight, with larger ones appearing smaller.
Inhibition of the Binding of Growth Factors to Immobilized CS-H Using Chondroitinase-digested Products of CS-H-Because the purified CS-H preparation showed significant digestibility by chondroitinase AC-I or B (Fig. 2), each enzyme digest was used to examine the effects of the digestion of CS-H on its growth factor binding ability, which would suggest whether either the CS-or DS-like moiety or both are responsible for efficient binding of the growth factors. This was carried out by investigating the binding of each growth factor in the presence of exogenously added CS-H or CS-H products digested with various chondroitinases. Preliminary experiments were first performed to determine the amount of exogenous CS-H required for achieving ϳ20 -50% inhibition of the binding of a given growth factor, and 0.12 g of the purified CS-H preparation (as uronic acid) was found to be suitable (data not shown). The same amount of the CS-H preparation and growth factors was used for comparative inhibition studies. Thus, binding to immobilized CS-H was tested for the growth factors listed in Fig. 4 in the absence and presence of undigested soluble CS-H or an equivalent amount of CS-H digested with chondroitinase ABC, AC-I, or B. FGF-1 and VEGF 165 were not tested because of comparatively weak binding (Fig. 4). The results from representative binding assays with HB-EGF on immobilized CS-H in the presence of soluble CS-H or its digested products are given in Fig. 10. The cumulative results for all growth factors tested are shown as bar graphs in Fig. 11. The result of the inhibition assay with FGF-16 was excluded from Fig. 11, but the pattern was similar to that of FGF-18.
Exogenous intact CS-H inhibited various growth factors from binding to immobilized CS-H to some extent, presumably reflecting a difference in the stoichiometry of the binding of soluble CS-H to growth factors in this assay system (see "Dis-

DISCUSSION
In this study, CS-H from hagfish notochord was structurally and functionally characterized. The CS/DS hybrid chains of CS-H exhibited marked yet somewhat weaker neuritogenic activity than a more highly sulfated CS-H preparation exclusive of GlcUA (14), supporting the structure-function relationship. Digestions with various chondroitinases revealed the involvement of both GlcUA and IdoUA. CS/DS chains bind heparin-binding growth factors (13,16) and may act as receptors or coreceptors in addition to conventional HS chains, and such growth factors may, in turn, be involved in the neuritogenic mechanism of CS/DS chains in vivo. Indeed, Bao et al. 2 recently identified PTN as an endogenous ligand for embryonic pig brain CS/DS chains and observed stimulation of the neuritogenic activity of these chains by exogenously added PTN. However, no such synergistic effects were observed for CS-H with PTN or MK, which was added at different concentrations in both soluble and immobilized forms to the CS-H-coated substrate. In addition, anti-PTN antibody inhibited the activity of the brain CS/DS chains, 2 but not that of CS-H or CS-E (data not shown). Notably, antibody 473HD, raised against the CS/DS chains of phosphacan/DSD-1-PG, neutralizes the dendritic neuritogenic activity of DSD-1-PG and CS-D, but not the axonic neuritogenic activity of CS-E (11). These findings together suggest that the mechanism of neuritogenesis by CS-H and CS-E is distinct from that by CS-D or pig brain CS/DS. The possible involvement of growth factors other than PTN and MK as well as neurotrophic factors remains to be examined.
CS-H bound various growth factors and two major neurotrophic factors with high affinity, all of which (except for FGF-16) are expressed in the brain. The binding of growth factors (including FGF-2) to HS and DS requires IdoUA residues (43,44). However, CS-E exclusive of IdoUA also exhibits specific interactions of high affinity with various heparin-binding growth factors (13), implying the importance of the E unit and the sulfation patterns as well. CS/DS hybrid chains from embryonic pig brain bind various growth factors (16), and hybrid chains of endocan DS-PG on the endothelial surface are also instrumental in bringing about hepatocyte growth factor-me-  b Peak areas of the respective disaccharides were integrated to calculate the molar proportion.
c The amounts of ⌬Di-0S obtained were added to ⌬Di-4S since we found that ⌬Di-0S observed after chondroitinase B digestion was generated through conversion of ⌬Di-4S by the action of 4-sulfatase in the chondroitinase B preparation (see "Results" and the legend to Fig. 8).
d Total disaccharides obtained are given in picomoles, and the values in parentheses represent percentages, with the total disaccharides obtained by chondroitinase ABC digestion taken as 100%. diated mitogenic activity of human embryonic kidney cells (8).
The present study provides supportive evidence for the physiological importance of CS/DS hybrid chains in the brain and also other tissues.
Inhibition assays using chondroitinase digests of CS-H showed that IdoUA-containing domains play a greater role in the binding of growth factors, with significant contributions by GlcUA-containing domains as well. Analysis of the GlcUA/ IdoUA distribution by digestion with various chondroitinases showed that these residues are not clustered, but are scattered, and higher oligosaccharides in the chondroitinase AC-I and B digests suggest the hybrid nature of CS-H with seemingly jumbled GlcUA and IdoUA residues. Further studies are needed to investigate the GlcUA/IdoUA distribution pattern. The purified CS-H preparation contained substantial proportions of both E and iE units, the importance of which has been suggested in the growth factor-binding (13,45) and neuritogenic (11,14,46) activities of CS and DS and which needs to be proven in such activities of CS-H. The iE unit has been identified in DS from mammalian liver (47), bovine kidney (48), and embryonic and adult pig brains (16).
All growth factors tested (except FGF-1) bound to CS-H with high affinity (Table I). Among them, MK requires the DS domain for the migration of MK-induced osteoblast-like cells (49), but does not bind to DS from porcine skin (12), probably due to differences in the sulfation pattern. A high affinity interaction (K d ϭ 34 nM) has been shown between immobilized PTN and soluble DS using the BIAcore system (15). The interaction between soluble PTN and immobilized DS (K d ϭ 111 nM) has also been evaluated using the IAsys system (39), with the result contrasting with the K d of 0.17 nM obtained in this study using immobilized CS-H and soluble PTN. The differences could be attributed to differences in the methodology or system as well as the sulfation patterns of the glycan chains. The high affinity of MK and PTN for CS-H and the negligible dissociation from CS-H prompt us to speculate that they are not released as free signaling molecules, but rather are enzymatically released as growth factor-oligosaccharide complexes from a parent PG molecule by endoglycosidases such as hyaluronidase and endo-␤-glucuronidase (50) and transferred to the cellsurface receptor. Alternatively, CS/DS-PGs may keep holding these growth factors when presenting them as coreceptors to the cell-surface receptors or may act as signal transducing receptors. Such possibilities may also be applicable to FGF-10 and FGF-18, which had sensorgrams similar to those of MK and PTN (Fig. 5).
VEGF 165 is a heparin-binding angiogenic growth factor highly specific for endothelial cells (51). VEGF is also expressed FIG. 9. Analysis of chondroitinase digests of the CS-H preparation on a Superdex peptide column. The purified CS-H preparation (0.17 g as uronic acid) was digested with 5, 1, or 10 mIU of chondroitinase AC-I (A), B (B), or ABC (D), respectively. The same amount of CS-H was also sequentially digested with chondroitinase B followed by chondroitinase AC-I (C). All digests were individually labeled with 2-AB (for details, see "Experimental Procedures") and analyzed by gel filtration on a Superdex peptide column using 0.2 M ammonium bicarbonate containing 7% 1-propanol as eluent. The elution positions of the authentic unsaturated CS-derived disaccharides/oligosaccharides are indicated as follows; arrow 1, decasaccharides; arrow 2, octasaccharides; arrow 3, hexasaccharides; arrow 4, tetrasaccharides; arrow 5, disulfated disaccharides; arrow 6, monosulfated disaccharides; arrow 7, unsulfated disaccharides. V 0 represents the void volume, and the total volume (V t ) was 24 ml. The peaks formed represent the molar ratios of the respective oligosaccharides labeled with 2-AB.

FIG. 10. Effects of digestions with chondroitinases on the binding activity of CS-H for HB-EGF.
Competitive inhibition experiments were carried out on the binding of HB-EGF as a representative growth factor to immobilized CS-H using various enzyme digests of the CS-H preparation as competitive inhibitors to investigate the inhibitory domains in CS-H. The CS-H preparation (1.7 g as uronic acid) was digested with chondroitinase ABC, AC-I, or B. Each digest (0.12 g as uronic acid) or an equivalent amount of intact CS-H was mixed with HB-EGF (50 ng) and then perfused onto a sensor chip on which CS-H (0.9 ng as GAG) had been immobilized (for details, see "Experimental Procedures"). HB-EGF binding was monitored in the BIAcore system as described in the legend to Fig. 4 in subpopulations of neurons in the developing and mature central nervous systems (52). Extracellular matrix-associated HS-PGs appear to serve as an extracellular storage reservoir for the heparin-binding VEGF types (53). Interestingly, such VEGF types bind to cell-surface and extracellular matrix-associated HS-PGs, releasing angiogenic factors such as FGF-2, which are stored on the heparin-like molecules in the extracellular matrix (54). Notably, DS released after injury is a potent promoter of FGF-2 activity (43). The sensorgrams obtained in this study for the binding of FGF-2, FGF-16, HB-EGF, and VEGF 165 to CS-H were similar to each other, showing a faster association and dissociation compared with the other four growth factors tested ( Fig. 5 and Table I). These binding features, together with the high affinity for CS-H as evidenced by the K d values (Table I), would facilitate the efficient association of these growth factors with CS/DS chains and their transfer to the respective cell-surface receptors. This study has demonstrated, for the first time, the high affinity binding of HB-EGF and VEGF to CS/DS chains. The possibility that these factors indeed use CS/DS in addition to HS as a physiological coreceptor in vivo remains to be explored.
BDNF influences myelin formation during nerve regeneration (55), and GDNF prevents neurodegeneration in Parkinson's disease (56). Recombinant human GDNF binds to GAGs showing high affinity for heparin, which is dependent on 2-Osulfate and inhibited by DS (57). Our interaction analysis of BDNF and GDNF with HS and CS-H revealed that the affinity exhibited by CS-H was 360-and 10-fold higher, respectively. The high affinity for CS-H and the negligible dissociation from CS-H are in contrast to their interactions with HS ( Fig. 7 and Table II) and are similar to the interactions of FGF-10, FGF-18, MK, and PTN with CS-H. These findings suggest, for the first time, that not only HS but also CS/DS may be involved in regulating the functions of these neurotrophic factors and that HS and CS/DS act in different ways. HS may transmit the factors as coreceptors to the cell-surface receptors (57,58), whereas CS/DS may keep holding the factors when presenting them to these receptors or may be enzymatically released as a functional neurotrophic factor-oligosaccharide complex from a parent PG. The domains responsible for high affinity binding to the growth factors and neurotrophic factors remain to be investigated.
Based on the binding of the growth factors to immobilized CS-H, the amounts of growth factors that bound to 1 mol of CS-H (18 kDa) were calculated as 0.5, 0.2, 1.1, 1.3, 1, 0.6, 0.7, and 0.07 mol for FGF-2, FGF-10, FGF-16, FGF-18, PTN, MK, HB-EGF, and VEGF 165 , respectively, using molecular masses of 17.5, 24, 21, 28, 15.4, 13.4, 12, and 45 kDa, respectively. The stoichiometry of the binding of BDNF and GDNF was also calculated to be 1.4 and 0.6 mol/mol of CS-H, using molecular masses of 13.6 and 20 kDa, respectively. These values indicate that the amounts of bound protein factors are not the same, but vary from one factor to another, thereby indirectly reflecting the differences in structure and average number of respective binding sequences on the GAG chains, only some of which appear to have binding sites. The possible in vivo implications of the observed interactions between the CS/DS chains and growth factors or neurotrophic factors remain to be explored to develop therapeutic agents for neuronal diseases and brain injury.