Functions of Chondroitin Sulfate/Dermatan Sulfate Chains in Brain Development

Chondroitin sulfate (CS) and dermatan sulfate (DS) have been implicated in the processes of neural development in the brain. In this study, we characterized developmentally regulated brain CS/DS chains using a single chain antibody, GD3G7, produced by the phage display technique. Evaluation of the specificity of GD3G7 toward various glycosaminoglycan preparations showed that this antibody specifically reacted with squid CS-E (rich in the GlcUAβ1–3GalNAc(4,6-O-sulfate) disaccharide unit E), hagfish CS-H (rich in the IdoUAα1–3GalNAc(4,6-O-sulfate) unit iE), and shark skin DS (rich in both E and iE units). In situ hybridization for the expression of N-acetylgalac-tosamine-4-sulfate 6-O-sulfotransferase in the postnatal mouse brain, which is involved in the biosynthesis of CS/DS-E, showed a widespread expression of the transcript in the developing brain except at postnatal day 7, where strong expression was observed in the external granule cell layer in the cerebellum. The expression switched from the external to internal granule cell layer with development. Immunohistochemical localization of GD3G7 in the mouse brain showed that the epitope was relatively abundant in the cerebellum, hippocampus, and olfactory bulb. GD3G7 suppressed the growth of neurites in embryonic hippocampal neurons mediated by CS-E, suggesting that the epitope is embedded in the neurite outgrowth-promoting motif of CS-E. In addition, a CS-E decasaccharide fraction was found to be the critical minimal structure needed for recognition by GD3G7. Four discrete decasaccharide epitopic sequences were identified. The antibody GD3G7 has broad applications in investigations of CS/DS chains during the central nervous system's development and under various pathological conditions.


identified. The antibody GD3G7 has broad applications in investigations of CS/DS chains during the central nervous system's development and under various pathological conditions.
Chondroitin sulfate (CS) 3 and dermatan sulfate (DS) chains are covalently linked to a wide range of core proteins, forming proteoglycans (PGs) with a widespread distribution in mammalian tissues (1). CS/DS-PGs are increasingly implicated as important regulators of many biological processes, such as cell proliferation and recognition (2,3), cell adhesion and migration (4 -6), neurite outgrowth (6,7), and wound repair and anticoagulation processes (8 -10). CS-PGs, which constitute the major population of PGs in the central nervous system, influence the formation of neuronal nuclei and the establishment of boundaries for axonal growth and act as modulators of neuronal outgrowth (11). The CS chain's backbone consists of repetitive disaccharide units containing D-glucuronic acid (GlcUA) and N-acetylgalactosamine (GalNAc) residues, whereas DS is a stereoisomeric variant of CS with varying proportions of L-iduronic acid (IdoUA) in place of GlcUA. In mammalian tissues, these chains are often found as CS/DS hybrid structures, and further structural variability of these chains during chain elongation is produced by divergent sulfation in the repeating disaccharide units by various sulfotransferases. Sulfation occurs at C-2 of GlcUA/IdoUA and/or C-4 and/or C-6 of GalNAc in various combinations, thereby producing characteristic sulfation patterns and enormous structural diversity. Recent studies have shown that proportions of these disaccharide units in chick, mouse, and pig brain change with development (12,13), suggesting that CS/DS chains differing in the degree and profile of sulfation may be involved in the functional diversity of neu-rons during brain development. Moreover, oversulfated CS/DS chains from various marine animals showed growth factorbinding and neurite outgrowth-promoting activities in vitro (14 -17).
However, the reported functions of brain CS/DS chains in neuritogenesis are controversial. CS/DS chains act as neuritogenic molecules (for reviews, see Refs. 6, 18, and 19) and also as inhibitors of axonal regeneration in the injured central nervous system (20 -23). Such apparently contradictory functions are probably attributed to the structural diversity of CS/DS chains. Significant and various proportions of disulfated disaccharide units, such as the D unit (GlcUA(2S)-GalNAc(6S)) and E unit (GlcUA-GalNAc(4S,6S)), are detected in the brain of cattle (24), embryonic day 13 (E13) mice (25) and E18 rats (26) (where 2S, 4S, and 6S represent 2-O, 4-O, and 6-O sulfate groups, respectively). Recently, we demonstrated that the rare oversulfated disaccharide units D and/or iD (IdoUA(2S)-GalNAc(6S)) and E present in embryonic pig brain-derived CS/DS chains are critical elements for neuritogenesis (27). In addition, CS chains from appican, the amyloid precursor protein expressed by rat C6 glioma cells, contained significant proportions of D and E units (28). CS-H, isolated from the hagfish notochord, has a unique oversulfated structure characterized by a major disaccharide H (or iE) unit (IdoUA-GalNAc(4S,6S)) (6,18,29,30) and therefore is actually DS. The term DS-E for this glycosaminoglycan (GAG) and iE (where "i" represents IdoUA) for the H disaccharide unit have been proposed (6,18). Further, exogenous DS-E/CS-H exhibited appreciable inhibitory activity in midkine (MK)-mediated neuronal cell adhesion similar to CS-E (26). However, the presence of oversulfated CS-E and/or iE structures in the mammalian brain has not been rigorously characterized.
Hence, we performed in situ hybridization to probe the distribution of GalNAc4S-6ST in postnatal mouse brain and also studied the possible functions of E/iE-containing structures in brain development using a new antibody, GD3G7, generated by the phage display technique (39). Interestingly, a good correlation between GalNAc4S-6ST expression and the GD3G7 epitope was observed in the mammalian brain.

EXPERIMENTAL PROCEDURES
Materials-The ScFv antibody GD3G7 was selected for reactivity with rat embryo-derived GAGs by phage display as described by ten Dam et al. 4 DNA sequence analysis revealed that antibody GD3G7 belongs to the V h 3 family, has a DP 38 germ line gene segment, and contains the heavy chain complementarily determining region 3 amino acid sequence GRWTQMT. CS-A from whale cartilage, CS-B from porcine skin, CS-C and CS-D from shark cartilage, CS-E from squid cartilage, and chondroitinases ABC (EC 4.2.2.4), AC-I (EC 4.2.2.5), and B (EC 4.2.2) were purchased from Seikagaku Corp. (Tokyo, Japan). CS-H from hagfish notochord was a gift from the late Prof. Nobuko Seno (Ochanomizu University, Tokyo, Japan), and DS from shark skin was prepared as described earlier (17). Postnatal day 7 (P7), P14, P21, 7-week-old male and pregnant ddY mice were purchased from SLC Inc. (Shizuoka, Japan). The monoclonal anti-VSV tag antibody P5D4 and porcine intestinal mucosal heparin were obtained from Sigma, and alkaline phosphatase-linked goat anti-mouse Ig (G ϩ M) was purchased from StressGen Biotechnol Corp. (San Diego, CA). A Superdex TM Peptide HR 10/30 column was obtained from GE Healthcare Bio-Sciences (Tokyo, Japan). EZ-Link TM biotin-LC-hydrazide was obtained from Pierce. Evennumbered, saturated oligosaccharide (tetra-to hexadecasaccharide) fractions were prepared by the partial enzymatic digestion of a commercial CS-E from squid cartilage with sheep testicular hyaluronidase as described previously (40), and structurally defined CS-E octa-and decasaccharides were prepared as described by Deepa et al. (41). Poly-DL-ornithine and antineurofilament antibody were purchased from Sigma, and MAP2 (anti-microtubule-associated protein 2) was from Leico Technologies Inc. (St. Louis, MO). Laminin isolated from Engelbreth-Holm-Swarm mouse tumor was from Invitrogen. A Vectastain ABC kit was obtained from Vector Laboratories. 35 S-Labeled 3Ј-phosphoadenosine 5Ј-phosphosulfate (PAPS) was purchased from PerkinElmer Life Sciences.
Enzyme-linked Immunosorbent Assay (ELISA)-For the evaluation of the specificity of the antibody in antigen recognition, streptavidin-coated plates were used. The reactivity of the antibody with various GAG subtypes was tested by ELISA as described previously (42). Briefly, the GAGs were individually biotinylated and immobilized on the plates. Wells were blocked with 1% bovine serum albumin (BSA) and incubated with 10-fold diluted primary antibody GD3G7. Since this antibody contains a VSV tag, the tag was used to detect binding of the antibody. A hybridoma supernatant of the anti-VSV tag mouse monoclonal antibody P5D4 in 0.1% BSA, phosphate-buffered saline (PBS) (diluted 5,000-fold) was used, followed by alkaline phosphatase-linked goat anti-mouse Ig (G ϩ M). Enzymatic activity was detected using p-nitrophenylphosphate, and the absorbance was measured at 415 nm.
For the inhibitory ELISA, a certain amount of GAG, evennumbered oligosaccharide, or structurally defined oligosaccharide was incubated with the antibody GD3G7 (periplasmic frac-tion, diluted 10-fold) in a total volume of 50 l at room temperature for 1 h before being applied to CS-E and streptavidin-coated plates. The inhibition was calculated from the reduced absorbance relative to that obtained from a control incubation without GAG.
Construction of a Soluble Form of GalNAc4S-6ST-A zebrafish GalNAc4S-6ST was identified in silico by a TBLASTN search using as a query sequence the mouse counterpart (GenBank TM accession number AB 187269). The cDNA encoding a truncated form of zebrafish GalNAc4S-6ST lacking the first NH 2 -terminal 81 amino acid residues was amplified by PCR with the pGEM-T Easy Vector containing the full coding sequence of the protein using a 5Ј-primer containing an inframe BamHI site (5Ј-GCGGATCCGGGCTCTTTTTAA-CACC-3Ј) and a 3Ј-primer containing a BamHI site located 77 bp downstream from the stop codon (5Ј-GCGGATCCAC-CAAGCATCGGCCT-3Ј). PCR was carried out with KODplus-DNA polymerase (TOYOBO, Tokyo) by 30 cycles at 94°C for 30 s, 59°C for 42 s, and 68°C for 90 s. The PCR products of the expected size were digested with BamHI and cloned into the BamHI site of an expression vector, pEF-BOS/IP (43). The resultant vector contained the insulin leader peptide and a protein A IgG-binding domain followed by the truncated form of zebrafish GalNAc4S-6ST.
Expression of a Soluble Form of Putative GalNAc4S-6ST and Enzyme Assay-The expression plasmid was introduced into COS-1 cells in 100-mm plates using FuGENE TM 6 (Roche Applied Science) according to the manufacturer's instructions. Two days after the transfection, a 1-ml aliquot of the culture medium was incubated with 10 l of IgG-Sepharose beads (GE Healthcare Bio-Sciences) for 1 h at 4°C. The enzyme-bound beads were washed with and resuspended in the assay buffer described below and used as an enzyme source for the sulfotransferase assay.
Sulfotransferase activity toward CS/DS was assayed by a method described previously (44). Briefly, the standard reaction mixture (60 l) contained 10 l of the resuspended beads, 50 mM imidazole-HCl, pH 6.8, 2 mM dithiothreitol, 10 M [ 35 S]PAPS (1 or 3 ϫ 10 5 dpm), and CS-B from porcine skin, which had been pretreated with nitrous acid as the acceptor polysaccharide (2 g as GlcUA). The reaction mixture was incubated at 37°C for 1 h and subjected to gel filtration using a syringe column packed with Sephadex G-25 (superfine) (45). The incorporation of [ 35 S]sulfate into the polysaccharide acceptor was quantified by determination of the radioactivity in the flow-through fractions by liquid scintillation counting.
Identification of the Transferase Reaction Products-The 35 Slabeled CS-B was isolated by gel filtration as described above, dried, and subjected to exhaustive digestion with chondroitinase ABC, AC-I, or B (46). The digest was analyzed by anion exchange HPLC on a column of amine-bound silica PA03 (47) or by gel filtration on a column of Superdex TM peptide (GE Healthcare Bio-Sciences) equilibrated with 0.2 M NH 4 HCO 3 as described previously (48).
In Situ Hybridization-Brains were quickly removed from P7, P14, P21, and 7-week-old ddY mice and frozen in powdered dry ice. Consecutive brain sections were cut at a thickness of 16 m with a cryostat (Leica Microsystems, Tokyo, Japan), thaw-mounted onto 3-aminopropyltriethoxysilane-precoated glass slides, and stored at Ϫ80°C prior to use.
The pGEM-T Easy Vector containing a cDNA fragment (ϳ2.0 kbp) of mouse GalNAc4S-6ST (UniGene Cluster Mm. 213582) was linearized with a restriction enzyme, NcoI or SalI, for synthesis of antisense or sense RNA probes, respectively. After enzymatic digestion, each digest was treated with a mixture of phenol/chloroform/isoamyl alcohol (25:24:1) and purified by ethanol precipitation. 35 S-Labeled riboprobes were transcribed using a MAXIscript TM T7 kit (Ambion, Austin, TX) with 5Ј-␣-[ 35 S]thiotriphosphate (ϳ30 TBq/mmol) (GE Healthcare Bio-Sciences). Before being loaded on the slides, the radiolabeled probes were mixed individually into a hybridization buffer (50% (v/v) formamide, 10% (v/v) dextran sulfate, 2.5ϫ Denhardt's solution, 4ϫ saline-sodium citrate (SSC), 5 mM EDTA, pH 8.0, 0.5 mg/ml tRNA from brewers' yeast (Roche Applied Science), and 20 mM dithiothreitol) at a concentration of 10 7 cpm/ml. In situ hybridization was performed as described previously (49). Briefly, the brain sections were fixed in 4% (v/v) formaldehyde in PBS at room temperature for 15 min. This was followed by digestion with proteinase K (2 mg/ml) at 37°C for 10 min, acetylation in 0.25% (v/v) acetic anhydride in 0.1 M triethanolamine at room temperature for 10 min, and dehydration through graded concentrations of ethanol. After drying, the tissue sections were pretreated with the hybridization buffer without 10% (w/v) dextran sulfate at 55°C for 1 h and then with the hybridization buffer containing the RNA probe prepared as above and incubated in a moisture chamber at 55°C for 18 h. The sections were then rinsed twice in 2ϫ SSC containing 10 mM 2-mercaptoethanol at room temperature for 5 min; digested with RNase A (50 g/ml) at 37°C for 30 min; rinsed twice in 50% formamide, 2ϫ SSC, 10 mM 2-mercaptoethanol at 55°C for 30 min and subsequently in 2ϫ SSC, 10 mM 2-mercaptoethanol at room temperature for 10 min; and dehydrated through graded concentrations of ethanol. The processed sections were exposed to x-ray films (Bio Max MR; Eastman Kodak Co.) for 1 week for signal detection.
Immunohistochemistry-Immunostaining was done as described previously (27). Briefly, cryosections (18 m thick) of mouse brain were fixed with acetone/methanol (1:1) and rehydrated with distilled water. The sections were then treated sequentially with the following solutions: 1) 2.5% hydrogen peroxidase in PBS (10 mM, pH 7.4) for 30 min; 2) 1% BSA, 4% normal horse serum in PBS for 60 min; 3) primary phage display antibody in 1% BSA, PBS (diluted 25-fold) overnight at 4°C; 4) anti-VSV tag mouse monoclonal IgG antibody, P5D4, in 1% BSA, PBS (1,000-fold) for 60 min; 5) anti-mouse IgG biotinylated antibody in 1% BSA, PBS (200-fold) for 60 min; 6) Vectastain ABC solution in PBS (200-fold) for 60 min; 7) 0.06% diaminobenzidine in PBS for 10 min; 8) 0.06% diaminobenzidine, 0.01% hydrogen peroxide in PBS for 5 min. Finally, sections were fixed with a series of ethanol solutions and xylene and then mounted with a xylene-based mounting medium. To confirm the specificity of staining with the antibody, brain sections were treated with chondroitinase ABC protease-free (5 mIU/section) to remove CS and DS. To confirm the minimal size for GD3G7, 1 nmol of the CS-E octa-or decasaccharide fraction was preincubated with the primary antibody at 37°C for 60 min before being added to the sections. In this case, polyester wax sections were used as follows. Brains were fixed overnight in a 3.7% formaldehyde solution and then dehydrated through several changes of ice-cold 70%, 90%, and absolute ethanol for 15 min at room temperature. Subsequently, brains were placed in 50% (v/v) polyester wax in absolute ethanol overnight at 37°C followed by 90% polyester wax in absolute ethanol at 37°C overnight. Brains were embedded in a 90% wax solution, and the blocks were cut serially at 5 m with a cryostat and were mounted onto glass slides. The slides were dried overnight at 4°C and then processed for immunostaining as described above.
Cell Culture-Hippocampal cells were cultured using E16.5 mouse brains as described previously (16,50). The hippocampi were obtained by microdissection and dissociated with a brief trypsin treatment. Dissociated cells were suspended in Eagle's modified essential medium containing an N2 supplement, seeded on poly-DL-ornithine-coated coverslips, and cultured at 37°C in a humidified atmosphere containing 5% CO 2 .
Neurite Growth Assay-Neurite growth of hippocampal neurons from E16.5 mouse brains was assayed as described previously (16,50). Briefly, plastic coverslips (10 ϫ 10 mm) were treated with 1.5 g/ml of poly-DL-ornithine in 0.1 M borate buffer, pH 8.2, for 2 h at room temperature and then further coated with CS-E or laminin at a dose of 2 or 10 g/cm 2 per coverslip, respectively, in PBS at 4°C overnight, washed with PBS three times, and incubated with the antibody GD3G7 (10 and 100 g/ml). After 2 h of incubation at room temperature, the coverslips were washed with PBS three times and then seeded with hippocampal cells at 10,000 cells/cm 2 . An irrelevant scFv antibody, MPB49V (100 g/ml), without reactivity toward GAGs was run as a negative control.
After a 24-h culture, cells were fixed with 4% (w/v) paraformaldehyde and then immunostained with anti-MAP2 and antineurofilament antibodies. The antibodies were detected using a Vectastain ABC kit with 3,3Ј-diaminobenzidine as a chromagen as described above. Neurite outgrowth was evaluated by determining the mean length of the longest neurite per cell. One hundred clearly isolated cells with at least one neurite longer than the cell body were chosen at random per coverslip. The results were expressed as the mean Ϯ S.E., and the significance of differences between means was evaluated with Student's t test.

Evaluation of the Specificity of Antibody GD3G7 by ELISA-
The reactivity of GD3G7 toward various GAG preparations was tested using ELISA, in which biotinylated GAGs were individually immobilized onto streptavidin-coated plates. GD3G7 specifically reacted with CS-E, CS-H (oversulfated DS from hagfish notochord, which contains iE as the predominant disaccharide unit), and also with SS-DS (DS from shark skin, which contains an appreciable proportion of E and iE units) and not with any other GAGs tested, including CS-A, CS-B, CS-C, CS-D, and heparin ( Fig. 1). It should be emphasized that GD3G7 could recognize both E (GlcUA-GalNAc(4S,6S)) and iE (IdoUA-GalNAc(4S,6S)) units. To further investigate the specificity of GD3G7 toward iE, an inhibitory ELISA was performed, in which CS-B treated for enzymatic sulfation with or without GalNAc4S-6ST was used as the inhibitor.
Human GalNAc4S-6ST transfers sulfate from PAPS to the C-6 hydroxyl group of the GalNAc-4-O-sulfate residues of CS-A and DS to form GlcUA-GalNAc(4S,6S) and/or IdoUA-GalNAc(4S,6S) units (38). To facilitate the functional analysis of GalNAc4S-6ST, the activity of a zebrafish counterpart was assayed using CS-B as the acceptor and PAPS as the donor substrate. To identify the reaction products, first gel filtration was conducted using a Sephadex G-25 (superfine) column, and then an aliquot of the flow-through fraction was digested exhaustively with chondroitinase ABC, chondroitinase AC-I, or chondroitinase B. It should be noted that the sulfation of GalNAc residues flanking a target IdoUA residue is essential for the action of chondroitinase B (46), although chondroitinase AC-I reactions are little influenced by the sulfation pattern of neighboring sugar residues. One half of each digest was analyzed by gel filtration on a Superdex TM peptide column (Fig. 2, A-C), and the other half of each digest was analyzed by anion exchange HPLC on an amine-bound silica PA-03 column using a linear NaH 2 PO 4 gradient from 16 to 530 mM over a 1-h period (Fig. 2, D-F).
The gel filtration chromatograms revealed that the 35 S-labeled products after digestion with chondroitinase ABC or B eluted at the position corresponding to disaccharides (Fig. 2, A and C), whereas no detectable disaccharide products were eluted after digestion with chondroitinase AC-I (Fig. 2B). The integrated radioactivity of the disaccharide peak observed on gel filtration was well consistent with that of ⌬Di-diS E (⌬E)  Hep)) and the DS preparations from hagfish notochord (CS-H) and shark skin (SS-DS) were included. Biotinylated GAGs (2 g each) were individually immobilized to wells of a streptavidin-coated plastic plate, and processed for incubation with the primary antibody GD3G7 (diluted 10-fold). Bound antibodies were visualized by subsequent incubation with mouse anti-VSV antibody P5D4 followed by alkaline phosphatase-linked goat anti-mouse Ig (G ϩ M) (diluted 5,000-fold). Enzymatic activity was measured using p-nitrophenylphosphate as a substrate at 415 nm. Negative controls received no primary antibody. Bars, mean Ϯ S.D. (n ϭ 3). JULY 6, 2007 • VOLUME 282 • NUMBER 27 detected by anion exchange HPLC analysis using the other half of each digest, demonstrating that the majority of 35 S-labeled ⌬Di-diS E was released from the -IdoUA-GalNAc(4S,6S)-IdoUA-sequence by digestion with chondroitinase ABC and chondroitinase B (Fig. 2, D and F) and not from the -GlcUA-GalNAc(4S,6S)-GlcUA-sequence, since no products were detected after treatment with chondroitinase AC-I (Fig. 2E). These results suggest that when CS-B was used as the acceptor molecule, GalNAc4S-6ST catalyzed the sulfation of GalNAc(4S) residues adjacent to IdoUA rather than GlcUA, thereby generating iE disaccharide units. Under the same assay condition, a soluble form of mouse GalNAc4S-6ST also catalyzed the formation of iE units, although only a relatively small proportion (29%) of the total radioactivity was detected at the position of ⌬E for the chondroitinase B digest (data not shown). Hence, to produce iE units in vitro, the soluble form of zebrafish GalNAc4S-6ST was used for enzymatic synthesis of iE unitcontaining polysaccharides as described below.

Single Chain Antibody GD3G7 Recognizing CS-E and DS-E
Evaluation of Specificity of GD3G7 toward iE-Zebrafish GalNAc4S-6ST preferentially transferred sulfate to the 6-Oposition of GalNAc(4S) residues adjacent to IdoUA residues in CS-B, which contains ϳ95% IdoUA-GalNAc(4S), thereby generating iE (IdoUA-GalNAc(4S,6S)) disaccharide units. To fur-ther evaluate the specificity of GD3G7 toward iE, we assessed the ability of CS-B treated with or without GalNAc4S-6ST to inhibit the binding of GD3G7 to immobilized CS-E. As shown in Fig. 3, CS-B treated with GalNAc4S-6ST showed significant inhibition for the binding of GD3G7 to CS-E, whereas CS-B not treated with GalNAc4S-6ST showed no inhibition against the binding of GD3G7 to CS-E, suggesting that GD3G7 specifically reacted with iE. Inhibitory activity was also exhibited by CS-H but not CS-D. The current data thus support the conclusion that GD3G7 recognizes both E and iE disaccharide units.
Expression of GalNAc-4-sulfate 6-O-Sulfotransferase in the Mouse Brain during Postnatal Development-CS/DS chains regulate the development of the central nervous system and play various roles as neuritogenic molecules. It has been previously demonstrated that IdoUA-containing structures in CS/DS hybrid chains from embryonic pig brain are essential for neuritogenic and growth factor-binding activity (51). Hence, to gain insights into the distribution of CS/DS hybrid chains containing E/iE units in the mouse brain during postnatal development, expression of the mouse GalNAc4S-6ST was examined by in situ hybridization with the 35 S-labeled antisense probe. Fig. 4 presents film autoradiograms of sagittal sections from the brains of P7 (Fig. 4A), P14 (Fig. 4B), and P21 (Fig. 4C) mice. The GalNAc4S-6ST transcript was observed in many regions of  the developing brain including the olfactory bulb, cerebral cortex, hippocampus, cerebellum, and striatum. In a negative control experiment with the 35 S-labeled sense probe, no labeling was observed in any region of brain sections of a P21 mouse (Fig. 4E) or in a P7 or P14 mouse (data not shown), confirming the specific hybridization of the antisense probe. The expression of this gene was confined to the cerebellum in P7 mice, with strong expression observed in the external granule cell layer (Fig. 4A). The cerebellar expression switched from the external granule cell layer to internal granule cell layer and appeared to decrease slightly with development (Fig. 4, A-D), probably reflecting the migration and/or maturation of granu-lar cells. This gene expression suggests that CS/DS chains containing E and/or iE units are relatively abundant in the cerebellum of P7 mice and gradually decrease during postnatal development. It is interesting to note that GalNAc4S-6ST message was also highly expressed in the dentate gyrus of hippocampus and olfactory bulb (Fig. 4, A-D), where neurons are continuously replaced throughout life (52), and in the striatum (Fig. 4, A-D). The gene expression profile provides insight into the possible involvement of the E/iE unit-containing CS/DS chains in neurogenesis and in the organization of the nigrostriatal system.
Immunohistochemical Detection and Distribution of GD3G7 Epitope in the Mouse Brain-To study the expression of the unique GD3G7 epitope in the mouse brain, immunohistochemical expression of P7 brain sections was performed using the antibody GD3G7. Consecutive sagittal sections from P7 mouse brain were incubated with GD3G7, and the bound antibodies were detected by anti-VSV-tag P5D4 antibodies and biotinylated anti-mouse IgG antibodies tagged with peroxidase and visualized by 3,3Ј-diaminobenzidine (the substrate for peroxidase). The staining pattern of the sagittal sections of P7 mice is shown in Fig. 5. GD3G7 specifically stained the P7 cerebellum, hippocampus, and olfactory bulb. The staining was stronger in the cerebellum, where GD3G7 stained most intensely the developing molecular layer at its border with the external granular layer. During the neonatal development of the cerebellum, the generation of granule cells occurs in the proliferative zone of the external granular layer (53), where GalNAc4S-6ST was strongly expressed (Fig. 4A). Granule cells start to exit the cell cycle after birth, and as part of their differentiation program they migrate internally, passing through the molecular layer and Purkinje cells to form the inner granular layer (54). Since the staining of the molecular layer was reduced in the P21 sections (data not shown), the staining of the molecular layer with GD3G7 at P7 may be, in part, due to the close apposition of internally migrating granule cells. In the hippocampus, the GD3G7 reactivity was observed in the areas surrounding the pyramidal cell layer of the CA1-CA3 fields and the granular cell layer of the dentate gyrus. In the olfactory bulb, the GD3G7 epitope was expressed in the glomerular layer as well as the external and plexiform layers. These staining patterns in the hippocampus and olfactory bulb are indicative of a role for the epitope in axon tract formation. The staining of the brain was completely abolished upon the treatment of the tissue sections with chondroitinase ABC, confirming the specificity of the staining. These results suggest that CS-E is highly expressed in the cerebellum and also in the hippocampus and olfactory bulb of P7 mouse brain.   35 S-labeled antisense (A-D) or sense (E) cRNA probes for GalNAc4S-6ST and exposed to an x-ray film for 1 week. No labeling was observed in the respective neighboring sections hybridized with the sense probe (E and data not shown). The GalNAc4S-6ST transcript showed a widespread expression in the developing brain except at P7, when strong expression was observed in the external granule cell layer (arrowhead) in the cerebellum. The expression in the granule cell layer switched to the internal region and appeared to decrease slightly with development. Strong expression of GalNAc4S-6ST was also observed in the olfactory bulb, hippocampus, cerebral cortex, and striatum of the developing brain. OB, olfactory bulb; Cx, cerebral cortex; Hi, hippocampus; Cb, cerebellum; St, striatum. Scale bar, 5 mm. FIGURE 5. Immunohistochemical localization of GD3G7 epitope in mouse brain. Consecutive sagittal sections from P7 mouse brain were stained using the single chain variable fragment anti-CS antibody GD3G7 as described under "Experimental Procedures." Immunoreactivity to GD3G7 was observed in the P7 cerebellum, hippocampus, and olfactory bulb. The staining completely disappeared upon treatment of the tissue sections with chondroitinase ABC. Scale bars, 2 mm. OB, olfactory bulb; Cx, cerebral cortex; Hi, hippocampus; Cb, cerebellum.

Single Chain Antibody GD3G7 Recognizing CS-E and DS-E
structure embedded in CS-E. To investigate whether the GD3G7 epitope in CS-E is involved in the process of neurite outgrowth in hippocampal neurons, the antibody was added to the culture medium at two different concentrations (10 and 100 g/ml) 2 h prior to the seeding of the neuronal cells on the substratum coated with CS-E. The antibody significantly suppressed the growth of neurites in the neurons grown on the substratum containing CS-E (Fig. 6, A and B). This observation was confirmed by a statistical morphometric analysis, which revealed that GD3G7 markedly inhibited the formation of neurites and decreased the total length of neurites per cell at a concentration of 100 g/ml (Fig. 6C). However, GD3G7 showed no significant inhibition against the neurite outgrowth induced by a laminin substrate (supplemental Fig. S1), suggesting that the inhibition was specific to CS-E-stimulated neurite outgrowth. Control scFv antibody (MPB49V, 100 g/ml) showed no neutralizing activity either. These results demonstrated that the epitope of GD3G7 is embedded in the neurite outgrowth-promoting motif of CS-E.
Characterization of the Minimal Structure Required for Recognition by GD3G7-In view of the finding that the epitope of GD3G7 in CS-E may be involved in the neurite outgrowthpromoting activity, we characterized the minimal structure of CS-E required for recognition by this antibody. Even-numbered CS-E oligosaccharide fractions, which were prepared by partial digestion of CS-E from squid cartilage with sheep testicular hyaluronidase (40), were used as inhibitors for the binding of GD3G7 to immobilized CS-E. Fig. 7A shows that decasaccharides and larger oligosaccharides exhibited inhibitory activity, and the activity increased with molecular size, although the FIGURE 6. Effect of a single chain variable fragment antibody GD3G7 on the neurite growth of hippocampal neurons in culture. Antibody GD3G7 was added to the culture medium at two different concentrations (10 and 100 g/ml) 2 h prior to the seeding of hippocampal cells from E16.5 mice on a substratum coated with CS-E. After a 24-h culture, the cells were immunostained with anti-MAP2 and anti-neurofilament antibodies. Representative images of the cell morphology obtained in the absence (A) or presence (B) of 100 g/ml GD3G7 are shown. C, effects of GD3G7 on neurite growth were evaluated by measuring the total length of the neurites per cell. An irrelevant scFv antibody MPB49V (100 g/ml) was run as a negative control. The values represent the mean Ϯ S.D. of those obtained from two independent experiments in triplicate. **, 0.001 Ͻ p Ͻ 0.01, significant difference from the values obtained in the experiments without exogenous antibody. FIGURE 7. Determination of the minimum size of CS-E oligosaccharides for the recognition by GD3G7. A, even-numbered oligosaccharide fractions (0.5 g) derived from CS-E of squid cartilage were tested by ELISA as inhibitors of the reactivity of GD3G7 with immobilized CS-E. Note that the decasaccharide fraction was the smallest fraction to show significant inhibitory activity, although the octasaccharide fraction seemed to react slightly with GD3G7 compared with the hexasaccharide fraction. B, different doses of CS-E octaand decasaccharide fractions (0, 0.5, 1, and 2 g) were tested by ELISA as an inhibitor of the reactivity of GD3G7 with immobilized CS-E. Note that GD3G7 reacted preferentially with the decasaccharide fraction from CS-E. Values were obtained from the average of two separate experiments.
octasaccharide fraction seemed to react only slightly with GD3G7 compared with the hexasaccharide fraction. Next, to precisely determine the minimal structure of CS-E recognized by GD3G7, varying doses of CS-E octa-and decasaccharides were used as inhibitors against the binding of GD3G7 to CS-E. Fig. 7B shows that GD3G7 reacted preferentially with the decasaccharide fraction from CS-E but not at all with the octasaccharide fraction. The degree of binding of GD3G7 to the CS-E decasaccharide fraction increased in a dosedependent manner, whereas there was no binding to the octasaccharide fraction even at higher doses. These results were consistent with the above finding that the decasaccharide fraction derived from CS-E meets the critical minimal size needed for recognition by GD3G7.
We sought to corroborate the ELISA data by assessing the immunoreactivity of the antibody in brain sections using CS-E oligosaccharides as inhibitors. CS-E-derived octa-and decasaccharide fractions were used as inhibitors for the immunohistochemical staining of mouse brain sections from postnatal day 7. GD3G7 was incubated with 1 nmol each of the CS-E octaand decasaccharide fractions at room temperature for 1 h prior to the incubation with the brain sections. The staining pattern of the sagittal sections from the brain of P7 mice is shown in Fig. 8. Preincubation of GD3G7 with the CS-E decasaccharide prevented the antibody from staining the sections, whereas immunostaining in the cerebellum, hippocampus, and olfactory bulb was still visible when the CS-E octasaccharide was used as an inhibitor. The finding that the CS-E decasaccharide fraction was the most efficient inhibitor of GD3G7 indicates that a CS-E-derived decasaccharide fraction is the minimal structural requirement for the binding of GD3G7.
To further characterize the epitope structure of GD3G7, four structurally defined decasaccharide fractions and one octasaccharide fraction derived from CS-E were used in an inhibitory ELISA experiment for the binding of GD3G7 to CS-E. It was found that all four decasaccharide fractions inhibited the binding of GD3G7 to immobilized CS-E, but the octasaccharide fraction did not (Fig. 9), which is basically in agreement with the size-dependent reactivities observed in Fig. 7B. The structure of the oligosaccharides is summarized in Table 1. GD3G7 reacted preferentially with all of the structurally defined decasaccharides, whereas the octasaccharide with four E units was not bound by this antibody. The observed binding preference suggests that this antibody can recognize decasaccharides with a minimum of three consecutive E units.

DISCUSSION
In this study, the characteristic spatiotemporal distribution of CS-E/DS-E hybrid chains in the postnatal mouse brain was examined in detail using an antibody, GD3G7, produced by the phage display technique. This antibody reacted specifically with CS-E, but also with CS-H, which consists of characteristic iE units (29,30,55), and with SS-DS, where E (or iE) units account for 10% of all the disaccharide (17), as evaluated by ELISA, suggesting that the essential recognition unit is a GalNAc(4S,6S) residue, irrespective of whether GlcUA or IdoUA is attached to it. CS-H is derived from the hagfish notochord (29), and the corresponding mammalian tissue produces the soluble Sonic Hedgehog protein, implicated in the generation of ventral neurons and the differentiation of motor neurons (56). CS-H is structurally similar to CS-E but contains IdoUA instead of GlcUA as a major uronic acid and is regarded as an oversulfated DS (26). In situ hybridization for GalNAc4S-6ST, which is involved in the biosynthesis of CS-E/DS-E, FIGURE 8. Effect of CS-E oligosaccharides on the immunohistochemical staining of GD3G7 epitope in mouse brain. CS-E deca-and octasaccharides (1 nmol) were tested as inhibitors of the staining of GD3G7 in P7 mouse brain. Note that preincubation of GD3G7 with decasaccharides abolished the staining of the mouse brain sections, whereas the staining was still observed after using octasaccharides as inhibitors.

Single Chain Antibody GD3G7 Recognizing CS-E and DS-E
showed that CS-E/DS-E hybrid chains are more concentrated in the cerebellum, providing insight into the importance of these hybrid chains during the brain's development. GalNAc4S-6ST transfers sulfate from PAPS to position 6 of GalNAc(4S) in CS and DS (38). In vitro synthesis of DS-E from CS-B was carried out using purified GalNAc4S-6ST, which preferentially generated IdoUA␣1-3GalNAc(4S,6S). Interestingly, the product inhibited the binding of GD3G7 to immobilized CS-E, indicating that the antibody GD3G7 recognized an iE-enriched region in CS-B. Thus, GD3G7 recognizes both E and iE units and will be a useful tool to study the distribution of E-and iE-containing CS/DS structures.
The importance of the differential expression of sulfotransferases has not been fully evaluated. Kitagawa et al. (13) demonstrated that the sulfation profile of CS chains and the ratio of chondroitin 4-O-sulfotransferase and chondroitin 6-O-sulfotransferase activities forming the specific sulfation profile changed markedly with development in the chick embryonic brain, and these alterations were precisely coordinated. Here, we have shown for the first time the distribution of GalNAc4S-6ST in the developing brain. The strong expression of GalNAc4S-6ST was confined in the external granule cell layer in the cerebellum (Fig. 4A) of P7 mouse brain. Its restricted expression switched to the internal granule cell layer and probably the Purkinje cell layer and decreased slightly with development, demonstrating that E/iE-containing CS/DS chains play critical roles during brain development. GalNAc4S-6ST was also found to be strongly expressed in the olfactory bulb, hippocampus, cerebral cortex, and striatum of the developing brain, suggesting the involvement of E/iE-containing CS/DS chains in neurogenesis, axon guidance, and/or neuronal survival in these regions. Recently, it has been demonstrated that appican, the CS-PG form of the Alzheimer's amyloid precursor protein, produced by glioma cells, contains 14.3% CS-E (28). The CS chain in appican is responsible for the adhesion of neural cells and for promoting the outgrowth of neurites in primary rat hippocampal cultures, because the core of the Alzheimer amyloid precursor protein is less potent in promoting adhesion and neurite outgrowth than is appican PG (57, 58). The pres-ence of E units in appican PG may explain the neurotrophic activities of appican. The hypothesis that oversulfated disaccharides are mainly responsible for the neurotrophic activities shown by specific CS-PGs is based on the pioneering observations that oversulfated squid CS-E as well as shark CS-D promotes neurite outgrowth in cultures of primary rat hippocampal neurons (15,59). By contrast, CS-PGs lacking oversulfated units either fail to promote neurite outgrowth or show inhibitory activity (60). Our observation that the postnatal brain contains oversulfated E units supports this hypothesis and suggests that CS-E/DS-E may function in the brain to promote neurite outgrowth. Recently, it has been demonstrated that CS-E can bind various growth factors (61), suggesting that its neuroregulatory activities are exerted through signal transduction mediated by various growth factors. In fact, it was recently demonstrated that CS/DS chains derived from embryonic pig brain recruited endogenous PTN to promote neuritogenesis in a hippocampal cell culture system (42), although the core proteins of the CS/DS chains have not been identified. Previously, a transmembrane PG, syndecan-1 from mouse mammary gland epithelial cells, was demonstrated to bear E-containing CS chains (62), although it is unknown whether syndecan-1, expressed in the rat central nervous system, which binds MK and PTN (63), contains CS-E.
The E unit has been demonstrated in appreciable proportions in CS chains from bovine brain (24) and chick brain, where it exhibits a developmentally regulated expression (13). Interestingly, the antibody GD3G7 stained restricted regions in the P7 mouse brain, indicating the expression of CS-E/DS-Econtaining structures in the mammalian brain. However, it remains to be determined whether the GD3G7 epitope in the mouse brain is embedded in the DS domains that contain iE units. Oversulfated disaccharides, such as D/iD, B/iB, and E/iE, have been implicated in the development of the brain (6,16,18). Recently, Bao et al. (27) demonstrated the presence of functional iD-containing domains in the developing mouse brain using a monoclonal antibody, 2A12, against the DS from ascidian. Maeda et al. (12) have reported the systematic immunohistochemical localization of several CS epitopes, including CS-56 and MO-225 epitopes in the developing postnatal mouse brain. The CS-56 epitope is highly expressed in the cerebral cortex early in the postnatal period, and its expression is markedly decreased 3 weeks after birth (12). In contrast, the expression of the MO-225 epitope is weak in the cerebral cortex but stronger in other regions, including the cerebellum, during the developing stages. MO-225 reactivity in the cerebellum reaches a peak at postnatal day 14 and decreases thereafter (12). The original study by Yamagata et al. (64) showed the reactivity of MO-225 with the tetrasaccharides ⌬A-D and ⌬E-D. Recently, Ito et al. (65) have shown that MO-225 could strongly react with octasaccharide and larger oligosaccharide fractions derived from CS-D of shark cartilage, and Deepa et al. (41) showed that MO-225 reacted with a decasaccharide sequence rich in E units, prepared from squid cartilage CS-E.
CS-E promoted the outgrowth of axon-like neurities toward neurons in a hippocampal cell culture (15). GD3G7 antibody inhibited the neuritogenic activity mediated by CS-E, suggest- a See Ref. 41. b Plus (ϩ) and minus (Ϫ) indicate positive and negative inhibitory activity, respectively, against the binding of GD3G7 to immobilized CS-E in ELISA as described in the legend to Fig. 9. c E, A, and C represent GlcUA-GalNAc(4S,6S), GlcUA-GalNAc(4S), and GlcUA-GalNAc(6S), respectively.
ing that the functional domain for the neuritogenic activity of the CS-E chains also includes the GD3G7 epitope. Ueoka et al. (26) have shown the presence of CS-E in E18 rat brains and also that the oversulfated E motif appears to be involved in the neuroregulatory activities of MK. It remains to be investigated, however, which CS-PGs in the brain contain the E motif. Despite the identification of soluble CS-PGs, such as brevican (66) and neurocan (67), as well as transmembrane CS-PGs, such as NG2 (68), PTP, and neuroglycan C (69), the CS chains attached to the core proteins have not been rigorously investigated. Maurel et al. (70) have shown that phosphacan, an extracellular variant of PTP, a receptor-type protein-tyrosine phosphatase that is abundantly expressed in the brain, binds PTN, which binds to the same gene family as MK, and the binding of PTN to phosphacan/PTP was dependent on the CS portion of this proteoglycan, and various CS preparations competitively inhibited this binding (71,72). Notably, squid cartilage CS-E, which has a high proportion of E units, inhibited the binding very efficiently, suggesting that oversulfated CS structures play important roles in the binding of phosphacan/PTP to PTN (71,72). It is possible that PTP and/or phosphacan bear CS chains with distinctive sulfation profiles during different developmental stages. PTP/phosphacan-PTN signaling is involved in the neurite extension, neuronal migration, glioma migration, and morphogenesis of Purkinje cells (for reviews, see Refs. 71-75). Deepa et al. (61) have shown that PTN and MK interact with CS-E. In view of these reports, the possibility cannot be excluded that specific oligosaccharide sequences in the brain containing E/iE units are responsible for the high affinity binding of MK and PTN to PTP and thereby mediating the neuritogenesis of cerebellar neurons. We also demonstrated that a CS-E decasaccharide fraction was the smallest oligosaccharide fraction to inhibit the binding of GD3G7 to CS-E, and larger oligosaccharide fractions showed stronger inhibition. The CS-E decasaccharide fraction also had an inhibitory effect on the reactivity of GD3G7 with sections of mouse brain, confirming the notion that the decasaccharide was the minimal size required for the binding of GD3G7. The reactivity of GD3G7 with structurally defined octa-and decasaccharides suggests that a decasaccharide with three consecutive E units may be a possible epitope structure for GD3G7 but not an octasaccharide with four consecutive E units. Thus, it was concluded that not only the E units but also size is important for recognition by this antibody. Also, the possibility cannot be excluded that the GD3G7 epitope in the brain may be a decasaccharide with core E and or iE units or a mixture of E/iE units scattered along the functional domain. The isolation of GD3G7-reactive structures from brain-derived CS/DS chains will help clarify the structures of this unique epitope.
The results altogether demonstrate that CS-E/DS-E expressed in the brain plays vital roles during brain development. The human scFv antibody described here can be used as a tool to visualize structural diversity in CS as illustrated for other antibodies (76,77). Since the cDNA of the single chain antibody is available, antibodies can be produced and purified in large quantities. It will be interesting to identify the PGs carrying such a unique E/iE-containing GD3G7 epitope, which would provide valuable information on the functions of CS-PGs during the development of the central nervous system.