Chondroitin-4-sulfation negatively regulates axonal guidance and growth

Glycosaminoglycans (GAGs) are a widely distributed, structurally diverse family of sulfated, unbranched polysaccharides that are expressed abundantly on the surface of cells and incorporated into extracellular matrix (ECM) (Bishop et al., 2007). GAGs have emerged as important regulators of the signaling involved in cell growth, tumorigenesis and inflammation (Iozzo, 2005; Parish, 2006; Taylor and Gallo, 2006). One species of GAG that is uniquely important in morphogenesis, cell division and cartilage development is chondroitin sulfate, the carbohydrate component of chondroitin sulfate proteoglycans (CSPGs), molecules that are spatiotemporally regulated during brain development (Hwang et al., 2003; Knudson and Knudson, 2001; Laabs et al., 2005; Sirko et al., 2007) and upregulated after injury in the central nervous system (CNS) (Chung et al., 2000; Silver and Miller, 2004).

During development, several different CSPGs have been localized to specific regions, such as the optic chiasm, where they appear to provide chemorepulsive signals to guide axonal growth (Bandtlow and Zimmermann, 2000; Chung et al., 2000; Ichijo and Kawabata, 2001). In the adult nervous system, high levels of CSPGs are found in perineuronal nets, where they are thought to stabilize synaptic connections. Removal of chondroitin sulfate GAG chains with chondroitinase ABC (cABC) restores ocular dominance plasticity in the adult visual cortex of rats (Pizzorusso et al., 2002). Even higher levels of CSPGs are found after injuries to the adult mammalian CNS, where CSPGs are a major component of the glial scar that impedes axonal regeneration (Silver and Miller, 2004). cABC treatment enhances axonal growth and functional recovery after spinal cord injury (Bradbury et al., 2002). However, the distinctive features of chondroitin sulfate GAG chains involved in these processes have not been fully identified.

Chondroitin sulfate GAG chains are complex unbranched polysaccharides of variable length with a backbone structure composed of a repeating disaccharide unit consisting of D-glucuronic acid (GlcA) and N-acetyl-D-galactosamine (GalNAc). This simple repetitive structure then can undergo extensive modification by sulfation at the C2 position of GlcA and/or the C4 or C6 position of GalNAc residues during biosynthesis. The location of modifications by distinct sulfotransferases is not template-driven, leading to a huge number of potential combinations of sulfation along the carbohydrate backbone. Whether specific sulfation in chondroitin sulfate regulates biological events is a matter of conjecture.

In this study, we use axonal guidance/growth and specific modifications of the sulfation of chondroitin sulfate GAG chains as a model to decipher the nature and importance of specific sulfation and the mechanisms by which it coordinates biological events. We present evidence that small changes in 4-sulfation of chondroitin sulfate GAG chains have major effects on the potency of CSPGs to impart guidance cues to neurons. These results support the concept that distinct sulfation along the carbohydrate backbone carries instructions to regulate neuronal function.

We performed axonal guidance spot assays (Meiners et al., 1999) to determine the behavior of axons as they encounter immobilized CSPGs. Axonal behavior of cultured mouse cerebellar granule neurons (CGNs) was analyzed near a defined region of chicken CSPGs immobilized onto poly-L-lysine (PLL)-coated coverslips. As observed previously (Laabs et al., 2007), most axons were deflected and few crossed onto the CSPG-rich area of the coverslip (Fig. 1A). Time-lapse imaging with adult mouse dorsal root ganglion neurons showed that filopodia dynamically sampled the CSPG spot (red), and that the growing axons turned at the interface between PLL and CSPG, and continued to extend along the interface, which is in contrast to growth cone collapse (Supplemental material Movie 1). Removal of the chondroitin sulfate GAG chains by cABC abolished this negative axonal guidance cue, indicating that the repellant activity of CSPGs is specifically mediated by the chondroitin sulfate GAG chains (Fig. 1B).

We examined whether chondroitin sulfate GAG chains alone could repel axons. When 4-sulfate-enriched (84%, determined by HPLC; supplementary material Fig. S1) CS-A was spotted onto PLL, axons were repelled at the interface in a manner comparable to native CSPGs (Fig. 1C,G). This repellent activity of CS-A was found to be very sensitive to cABC treatment. CS-A partially digested with cABC was precipitated with ethanol to remove disaccharides from the fractions, immobilized as spots and subjected to axonal guidance spot assays. Digestion by the enzyme of less than 2% of the total GAG abolished its activity (Fig. 1D). This indicates that a small portion of chondroitin sulfate GAG chains is essential for neuronal guidance activity. More surprising is that 6-sulfate-enriched (84%) CS-C had no inhibitory activity, because axons and cell bodies grew well on immobilized CS-C (Fig. 1E). These results suggest that sulfation at the C4 position of the GalNAc moiety presents a specific negative guidance cue to axons. Axons of dissociated embryonic mouse cortical neurons showed the same behavior as immobilized CS-A and CS-C (supplementary material Fig. S2).

The role of chondroitin sulfate 4-sulfation in axonal guidance was further confirmed by the observation that chondro-4-sulfatase treatment of CS-A totally abolished the axon-repellant action (Fig. 1F,G), despite only a modest reduction in 4-sulfation (Table 1). This result strengthens the idea that subsets of the sulfation are crucial for its biological activity. To exclude the possibility that the presence of cABC in the chondro-4-sulfatase preparation was responsible for the drastic change in its biological activity, we conducted axonal guidance spot assays with CSPGs after treatment with chondro-4-sulfatase. Note that chondroitin sulfate disaccharides are good substrates for chondro-4-sulfatase, but intact CSPGs are not (Yamagata et al., 1968). Axons were repelled by sulfatase-treated CSPGs at comparable levels to non-treated CSPGs (data not shown). CS-A was then extensively digested with chondro-4-sulfatase at 37°C for 16 hours and subjected to fluorescent labeling without cABC treatment. Since the appearance of chondroitin sulfate disaccharides is dependent upon cABC activity, the presence or absence of fluorescent signals derived from chondroitin sulfate disaccharides allows us to determine whether cABC is a contaminant of the chondro-4-sulfatase. Although we observed clear fluorescent signals in HPLC analysis with cABC treatment (Table 1), there was no signal derived from chondroitin sulfate disaccharides when cABC treatment was skipped after chondro-4-sulfatase digestion (data not shown). Both of these experiments strongly suggest that it is the chondro-4-sulfatase that alters the biological activity of CS-A.

To explore the functional consequences of sulfation of chondroitin sulfate GAG chains on axonal behavior, we examined neuron-astrocyte interactions using a co-culture system, which is more physiologically relevant than studying cells in isolation. In the adult CNS, astrocytes are generally supportive of neuronal function. However, injuries to the CNS induce a gliotic reaction characterized by the presence of reactive astrocytes, which are major components of the glial scar, which is considered to be detrimental to axonal regeneration. TGFβ is rapidly upregulated after CNS injury in vivo and is important both as a soluble regulator of ECM formation and in inducing reactive astrocytes (Flanders et al., 1998; Smith and Strunz, 2005). Confluent cultures of astrocytes were pretreated with TGFβ1 for 7 days; dissociated CGNs were plated onto these monolayers and co-cultured in fresh medium without TGFβ1 for 2 days, followed by measurement of axonal length. Whereas axons of CGNs growing on untreated astrocytes elaborated long and thin processes (Fig. 2A) (process length: 93±4 μm, mean ± s.d.), the axons of neurons cultured on TGFβ1-treated astrocytes had significantly shorter processes (54±2 μm, P<0.01 compared with untreated astrocytes; Student's t-test). This reduction in axonal growth was also observed when neurons alone were cultured in conditioned medium derived from TGFβ1-treated astrocytes (Fig. 2B). To exclude the possibility that TGFβ1 directly affects axonal growth, a potent TGFβ type I receptor inhibitor, SB-431542, was added to conditioned medium derived from TGFβ1-treated astrocytes. SB-431542 addition failed to restore neuronal growth, confirming that TGFβ1-dependent axonal growth inhibition is mediated through its action on astrocytes and not neurons.

Consistent with axonal growth inhibition, CSPG production was increased in TGFβ1-treated astrocytes as determined biochemically (Fig. 2C) and cytochemically (supplementary material Fig. S3) using an antibody recognizing 4- and 6-sulfated chondroitin sulfate. Increased production of CSPGs in conditioned medium and cell lysates was observed after 3 days of treatment with TGFβ1. It should be noted that CS-56-positive bands were sensitive to cABC treatment and migrated faster and less diffusely on SDS-PAGE under reducing conditions than non-reducing conditions (supplementary material Fig. S3). However, production of laminin, a major growth-permissive component of the ECM, was not altered in response to TGFβ1 treatment (data not shown). More quantitatively, accumulation of CSPGs by reactive astrocytes was detected in conditioned medium using an ELISA as early as 1 day after TGFβ1 treatment (Fig. 2D). Quantitative RT-PCR revealed that mRNA levels of genes encoding neurocan and versican were upregulated after TGFβ1 treatment (Asher et al., 2000). These data indicate that the increased production of CSPGs by reactive astrocytes is likely to be responsible for inhibition of axonal growth.

To firmly establish the involvement of CSPGs in this inhibition, we performed axonal guidance spot assays with immobilized conditioned medium derived from astrocytes (Fig. 3). Axons favored growth on PLL compared with the spot where concentrated TGFβ1-treated conditioned medium was immobilized, and this preference was abolished by cABC treatment (Fig. 3A,B), demonstrating that it is the chondroitin sulfate GAG chains in the conditioned medium that impart neuronal guidance cues. Next, we examined the effect of GAG synthesis inhibitors on axonal growth. Astrocytes were pretreated with TGFβ1 together with xyloside or sodium chlorate, and neurons were cultured on the monolayers (Fig. 3C). Reduction of axonal growth by TGFβ1 treatment was prevented when the covalent attachment of GAG chains to the core protein was competitively inhibited by treatment of astrocytes with xylosides, or when sulfation was blocked by sodium chlorate. Together, these data provide substantial evidence that chondroitin sulfate GAG chains produced by reactive astrocytes mediate axonal growth inhibition.

We next determined whether TGFβ1 treatment regulates the sulfation of chondroitin sulfate GAG chains. Immunoblot analyses of conditioned medium with monoclonal antibodies 2B6 and 3B3 (specific for 4-sulfated and 6-sulfated chondroitin sulfate GAG chains, respectively) showed substantial increases in 4-sulfation and a slight increase in 6-sulfation 3 days after TGFβ1 addition (Fig. 4A). This was confirmed quantitatively by an ELISA with another set of sulfation-specific monoclonal antibodies (MAB2030 and 2035) (Fig. 4B). It is noteworthy that only 4-sulfated chondroitin sulfate was acutely induced within 24 hours of TGFβ1 exposure, and that accumulation rates of 4-sulfated and 6-sulfated chondroitin sulfate thereafter were similar.

The finding of this dramatic change in 4-sulfation led us to examine chondroitin sulfotransferases that are responsible for the sulfation in GalNAc. Consistent with our ELISA data, quantitative RT-PCR revealed a 3.8-fold induction in chondroitin 4-O-sulfotransferase 1 (C4ST1, official gene symbol Chst11) mRNA as early as 8 hours after TGFβ1 treatment that endured for 48 hours (Fig. 4C). By contrast, levels of chondroitin 6-O-sulfotransferase 1 (C6ST1, official gene symbol Chst3) mRNA remained unchanged (Fig. 4D). Other chondroitin sulfotransferases [chondroitin 4-O-sulfotransferase 2 (C4ST2), chondroitin 6-O-sulfotransferase 2 (C6ST2), and GalNAc 4-sulfonate 6-O-sulfotransferase (GalNAc4S-6ST)] were not altered upon TGFβ1 treatment (data not shown). Upregulation of C4ST1 protein in reactive astrocytes was also confirmed using an anti-C4ST1 peptide antibody (Fig. 4E). The fact that the increase in C4ST1 mRNA upon TGFβ1 treatment was not observed in Smad3-null astrocytes (supplementary material Fig. S3) demonstrates that the rapid change in sulfation of chondroitin sulfate is mediated by TGFβ signaling through the Smad pathway.

To investigate whether 4-sulfation of chondroitin sulfate GAG chains is crucial for the regulation of axonal growth, loss- and gain-of-function experiments were performed. Introduction of siRNA against C4ST1 into astrocytes decreased levels of C4ST1 protein in whole cell lysates, and correspondingly reduced the accumulation of 4-sulfated chondroitin sulfate in conditioned medium (Fig. 5A-C). Importantly, the TGFβ1-mediated increase in 4-sulfated chondroitin sulfate and C4ST1 was blocked by C4ST1 siRNA. We then performed axonal guidance spot assays with conditioned medium from astrocytes treated with combinations of TGFβ1 and C4ST1 siRNA. Conditioned medium from astrocytes treated with the combination of C4ST1 siRNA and TGFβ1 was significantly less potent than conditioned medium from astrocytes treated with TGFβ1 alone (Fig. 5D). Transfection of C4ST1 siRNA did not affect the induction of neurocan mRNA by TGFβ1 (data not shown).

Because 6-sulfated GAG chains have also been suggested to be involved in the brain-injury response (Properzi et al., 2005), we similarly examined the effects of alteration of C6ST1 and 6-sulfated GAG chains. Astrocytes treated with siRNA directed against C6ST1 showed a reduction of both mRNA level and the production of 6-sulfated chondroitin sulfate (Fig. 6B,D). By contrast, treatment with C6ST1 siRNA did not alter the levels of C4ST1 transcript nor 4-sulfated CS. Furthermore, TGFβ1 treatment still elicited an increase in C4ST1 mRNA and 4-sulfated chondroitin sulfate (Fig. 6A,C). More importantly, depletion of C6ST1 did not alter the inhibitory properties of TGFβ1-treated conditioned medium in the axonal guidance assays (Fig. 6E). Moreover, C4ST1 siRNA treatment did not alter the level of the mRNA encoding C4ST2, C6ST2 or GalNAc4S-6ST (data not shown). These results demonstrate the essential roles of C4ST1 and 4-sulfated chondroitin sulfate GAG chains in the induction of repellent activity of astrocytes by TGFβ1 treatment.

In gain-of-function experiments (Fig. 7), astrocytes were transfected with either an empty vector, wild-type C4ST1 or a mutated form of C4ST1 that fails to bind 3′-phosphoadenosine 5′-phosphosulfate and CGNs were plated on the monolayers of these astrocytes. Although similar levels of exogenous proteins were expressed in astrocytes (Fig. 7A), cells expressing wild-type C4ST1 produced more 4-sulfated CSPG in the conditioned medium compared with vector-transfected astrocytes (Fig. 7B), whereas astrocytes expressing the mutated form showed lower levels of 4-sulfated GAG chains. Neurons growing on astrocytes expressing C4ST1 had shorter axons than those growing on either vector-transfected astrocytes or those expressing the mutant C4ST1 (Fig. 7D). HPLC analysis of conditioned medium confirmed that the production of 4-sulfated GAG was highly correlated with perturbations in C4ST1 expression (Fig. 5C, Fig. 7C). We were unable to demonstrate an effect of overexperssion of C6ST1 because when exogenous C6ST1 was expressed in astrocytes, the cells looked unhealthy and survival was compromised (data not shown). Taken together, the loss- and gain-of function experiments establish a pivotal role of 4-sulfated GAG in axonal growth regulation by reactive astrocytes.

In vivo experiments confirmed that 4-sulfated chondroitin sulfate GAG chains might be a critical determinant of CNS regenerative failure (Fig. 8). A dorsal overhemisection of the spinal cord was made in mice and we examined expressions levels of 4-sulfated and 6-sulfated GAG chains, as well as glial fibrillary acidic protein (GFAP), a well-established marker of reactive astrocytes (Lemons et al., 1999; Pekny and Pekna, 2004) with specific antibodies. Although we observed very low levels of immunoreactivity for 6-sulfated chondroitin sulfate GAG chains, we found substantial staining for 4-sulfated chondroitin sulfate GAG chains proximal to the lesion as early as 1 day post injury (Fig. 8A,D). A similar increase in GFAP immunoreactivity was observed with similar proximity to and specificity for the lesion site. Colocalization of 4-sulfated GAG and GFAP was also apparent microscopically (Fig. 8B,C). Upregulation of 4-sulfated GAG chains upon spinal cord injury was also apparent when chondroitin sulfate disaccharides were extracted from uninjured/injured tissues and analyzed by HPLC (Fig. 8E). Thus, these data confirm a specific upregulation and deposition of 4-sulfated GAG by reactive astrocytes after CNS injury in an animal model.

CSPGs are ECM molecules that have a critical role in modulating axonal growth and guidance during development and also after nervous system injury. Although much evidence has accumulated suggesting that it is the GAG chain moieties of CSPGs that are recognized by neurons, the particular features of GAG chains that signal to growing axons are still a matter of contention. In this manuscript, we present compelling evidence that this signaling is mediated through specific sulfation, specifically 4-sulfation, of the chondroitin sulfate GAG chains. First, CS-A, but not CS-C, exhibits negative guidance cues to axons in a 4-sulfation-dependent manner, with comparable efficacy to native CSPGs. Second, reactive astrocytes in culture produce more 4-sulfated chondroitin sulfate GAG chains and knockdown of C4ST1 protein reduces the level of 4-sulfation in chondroitin sulfate GAG chains, resulting in a less inhibitory ECM. Third, overexpression of C4ST1 in cultured astrocytes increases 4-sulfation and reduces their ability to support neuronal growth. Finally, 4-sulfated chondroitin sulfate GAG chains are acutely upregulated and deposited by reactive astrocytes in an animal model of spinal cord injury. This combination of biochemical and physiological approaches synergistically demonstrate the major role of 4-sulfated GAG chains in astrocyte/neuron interactions.

The fact that CS-A, but not CS-C, repels axons highlights the exquisite structural specificity for signaling by the sulfated disaccharides that comprise chondroitin sulfate chains. Both CS-A and CS-C carry a similar charge distribution, demonstrating that these effects are not simply mediated by negative charge carried by the sulfate groups. Although 6-sulfation of chondroitin sulfate GAG chains has been reported to correlate with axonal inhibition (Properzi et al., 2005), we did not find any inhibitory action of CS-C in our axonal guidance assays, and siRNA-based depletion of C6ST1 in reactive astrocytes showed no effect on axonal guidance.

Only a small change in 4-sulfation significantly alters the ability of CS-A to impart neuronal guidance, suggesting that subsets of sulfation are critical determinants of function. This notion is supported by the finding that the biological activity of CS-A is eliminated after only a short duration of treatment with cABC, which digests as little as 2% of the GAG. Conversely, only a small percentage of 4-sulfation was reported to increase in vivo following injury, even though 4-sulfated disaccharides are the predominant species in the normal brain (Gris et al., 2007; Mitsunaga et al., 2006; Properzi et al., 2005). These data suggest that it is not the level of 4-sulfation per se that contributes to GAG chain signaling. It has been proposed that distinct motifs of sulfation (a `sulfation code') along the polysaccharide chain in heparan sulfate encode information required for substrate binding and growth regulation (Bülow and Hobert, 2004; Holt and Dickson, 2005). Although heparan sulfate and chondroitin sulfate are structurally different, our findings might suggest the presence of a sulfation code in chondroitin sulfate GAG chains that exhibits negative guidance cues to axons and inhibit axonal growth.

The direction and rate of axonal extension can be independently modulated by the ECM (Powell et al., 1997). The axonal guidance spot assays used in this study focus simply upon axonal guidance: axons growing on the PLL substrate turn as they encounter CS-A, and continue to extend along the interface (data not shown). Similar behavior is observed in vivo when growing axons encounter the CSPG-rich glial scar (Davies et al., 1997). By contrast, axonal growth depends upon both cell adhesion and neurite initiation/extension, and alterations in either of these conditions will result in measurable changes. It is intriguing that 4-sulfation of chondroitin sulfate GAG chains both alters axonal direction and limits the rate of axonal extension.

Paradoxically, tissues that express CSPGs do not always exclude the entry of axons, and in some cases, CSPG staining coincides with developing and regenerating axon pathways (Bicknese et al., 1994; McAdams and McLoon, 1995). Axonal extension during development and after injury to the adult CNS is a balance of inhibitory and promotional cues in the local environment consisting of several ECM molecules, cell adhesion molecules and growth factors (Lu et al., 2007; McKeon et al., 1995; Walsh and Doherty, 1996). In addition, changes in sulfation of chondroitin sulfate GAG chains are likely to contribute to the determination of the success or failure of axonal regeneration. Several in vitro studies suggest that CSPGs can promote rather than inhibit neurite outgrowth (Faissner et al., 1994; Fernaud-Espinosa et al., 1994; Garwood et al., 1999). These promotional effects have been attributed to the `oversulfated' chondroitin sulfates: CS-D (disulfated at the C2 position of GlcA and C6 position of GalNAc) and CS-E (disulfated at the C4 and C6 positions of GalNAc), both of which stimulate neurite growth in culture (Deepa et al., 2002; Gama and Hsieh-Wilson, 2005; Gama et al., 2006; Nadanaka et al., 1998). Axonal growth promotion has also been observed with an artificial tetrasaccharide with 4,6-sulfation, suggesting that a short stretch of sulfated GAG chains is sufficient to promote neurite outgrowth (Gama et al., 2006). Interestingly, when we used CS-D and CS-E in our axonal guidance assays, we did not observe any positive haptotactic effects of these sugars. Because oversulfated chondroitin sulfate chains have been shown to bind several different growth-promoting factors and cytokines (Deepa et al., 2002; Shipp and Hsieh-Wilson, 2007), the growth-promotional actions of these chondroitin sulfate sugars may be indirect.

In the developing brain, astrocytes are a preferred substrate for axonal growth and neuronal migration, whereas reactive astrocytes in the injured brain are detrimental to neuronal regeneration. The major difference in this functional shift is the increased production of sulfated proteoglycans by reactive astrocytes. Using a physiologically relevant system, we found that modulation of the sulfation in astrocytic CSPGs changes the interaction between astrocytes and neurons in vitro. Combined with our observation that 4-sulfated CSPGs are robustly and rapidly deposited within CNS lesions in animals, these findings suggest that modulation of sulfation in CSPGs serves as a signal to restrict axonal regrowth and may be an important new therapeutic direction for regenerative biomedicine.

Cultures of dissociated mouse CGNs were prepared from C57BL/6 mice (P5-8) as described previously (Levi et al., 1984; Romero et al., 2003). Dissociated cells were cultured in Neurobasal-A medium containing B27 supplement and 25 mM KCl. In co-culture experiments, dissociated CGNs were plated at a density of 6×10⁴ cells/well onto a confluent monolayer of astrocytes in 24-well plates (see below). When neurons were cultured in conditioned medium, 2% (v/v) of B27 supplement was added. Primary cortical neuron cultures were prepared from E16-E18 mouse embryos as previously described (Dulabon et al., 2000).

Primary cultures of cerebral cortical astrocytes were prepared from newborn C57BL/6 mice and Smad3-null mice (Wang et al., 2007) (P1-2) as previously described (Petroski et al., 1991). Confluent cultures of astrocytes were treated with TGFβ1 (10 ng/ml, R&D Systems) in the absence of serum for 7 days and dissociated CGNs were plated on the monolayers. Two days after plating, cells were fixed and stained with anti βIII-tubulin antibody (Sigma), followed by the incubation with FITC-anti-mouse IgG antibody (Jackson ImmunoResearch Laboratories). When astrocytes were treated with inhibitors, monolayers of astrocytes were incubated with TGFβ1 in combination with those inhibitors for 72 hours, after which neurons were plated on top and allowed to grow for 48 hours before analysis.

Axonal guidance spot assays were performed as described previously (Meiners et al., 1999). To quantify the behavior of axons, an interface between PLL and sample was created by placing a 5 μl drop of chicken CSPG (Millipore, 12.5 ng/spot) or chondroitin sulfate GAG chains (Seikagaku, Japan) with Texas Red in the center of a PLL-coated glass coverslip. Texas Red was used to visualize the interface and was used alone for negative control experiments. Dissociated CGNs were seeded onto the coverslips at a density of 6×10⁴ cells/well and cultured for 2 days. Cells were fixed and stained with anti-βIII-tubulin antibody (Sigma), followed by the incubation with FITC-anti-mouse IgG antibody (Jackson ImmunoResearch Laboratories). Fluorescence images were acquired on a Nikon TE2000 inverted microscope equipped with a CCD camera (Orca-ER, Hamamatsu) driven by Metamorph imaging software (Universal Imaging). Only single, non-fasciculated axons within 10 μm of the protein-PLL interface were considered for the analysis. In addition, only axons growing toward the immobilized sample were counted and no axon whose soma was sitting on the interface was scored. Each experiment was performed in triplicate. For treatment with cABC (Seikagaku), samples were digested with 10 mU of the enzyme at 37°C for 3 hours. Chondro-4-sulfatase (Seikagaku) digestion was carried out with 8 mU of the enzyme in 0.1 M ammonium acetate buffer (pH 7.0) at 37°C for 4 hours, followed by the inactivation of the enzyme at 95°C for 5 minutes.

Axonal outgrowth assays were performed as described previously (Meiners et al., 1999). Axonal length was measured using the ImageJ program (available at: http://rsb.info.nih.gov/nih-image/). A sample of 100 neurons with processes equal to or greater than one cell soma was considered for each condition. The total length of each primary process was measured for each neuron. For some experiments, relative axonal length was also obtained using a stereological technique from a large sample of neurons (Rønn et al., 2000). When neurons were co-cultured with astrocytes overexpressing C4ST1, astrocytes were first transfected with appropriate DNA constructs, followed by replacement of the media 1 day after transfection and an additional 1 day culture, and neurons were then plated onto the monolayer of astrocytes. Cells were fixed, stained with anti-βIII-tubulin antibody, and relative axonal length was measured as described above. Astrocyte-derived conditioned media and cell lysates were collected 2 days after transfection for immunoblot and disaccharide composition analysis.

Computed values were compared between the different conditions using either Student's t-test or one-way ANOVA, as appropriate.

The cDNA for chondroitin 4-sulfotransferase 1 (C4ST1) was obtained by RT-PCR from mouse astrocyte RNA using the following primers; 5′-TAGAATTCAC TAGTATGAAG CCGGCGCTGC TGGAAG-3′ and 5′-ATGAATTCCA CTCGAGTCCA ACTTCAGGTA GTTTGG-3′. PCR product was digested with EcoRI, followed by subcloning into the EcoRI site of pDsRed2-N1 (BD Biosciences) and pTracer-EF/V5His (Invitrogen). An inactive form of C4ST1 was generated by the introduction of mutations (R186A, S194A) into the putative PAPS binding site with QuikChange (Stratagene, CA) using the following primers; 5′-GTTCCTGTTC GTGGCTGAGC CCTTCGAGAG G-3′ and 5′-GAGCCCTTCG AGAGACTAGT GGCTGCCTAC CGCAAC-3′. The cDNA for mouse chondroitin 6-sulfotransferase 1 (C6ST1) was obtained by RT-PCR using the following primers; 5′-ATGAATTCAC TAGTATGGAG AAAGGACTCG CTTTGC-3′ and 5′-AAAAGCTTCT ACGTGACCCA GAAGGTGC-3′. PCR product was digested with EcoRI/HindIII, followed by subcloning into the EcoRI/HindIII sites of pDRed2-N1 (BD Biosciences).

Transient transfection was performed using the Nucleofector (Amaxa, Cologna, Germany) with a protocol specifically designed for mouse astrocytes. After transfection, astrocytes were plated on a 35 mm dish and grown to confluency.

CSPGs in conditioned medium and cell lysates derived from astrocyte cultures were separated by SDS-PAGE under reducing conditions and examined with immunoblot analyses as described previously (Katagiri et al., 2000). When conditioned medium was concentrated with a Centricon 100 (Millipore), a protease inhibitor cocktail (Calbiochem) was added to prevent protein degradation. Samples to be incubated with the 2B6, 3B3 (Seikagaku), MAB2030 or MAB2035 (Millipore) antibodies required prior digestion with cABC (10 mU/ml, 37°C for 3 hours) to expose the antigen. For immunoblotting of C4ST1, we generated a custom chicken anti-C4ST1 peptide antibody (Gallus Immunotech) against the peptide sequence RRQRKNATQEALRKGDDVKC. HeLa cell lysates expressing recombinant C4ST1 with a V5 epitope tag (pTracer-C4ST1) were used as positive controls for immunoblotting. Chicken CSPGs (Millipore) used in this study contained neurocan and phosphocan, confirmed by cABC digestion and tryptic digestion, followed by mass spectrometry, but we did not exclude the presence of other core proteins (data not shown).

TGFβ1-treated astrocytes cultured for 7 days on glass coverslips were rinsed with DMEM and incubated with CS-56 (Sigma) for 30 minutes at 4°C, followed by incubation with FITC-anti-mouse IgM antibody (Jackson ImmunoResearch Laboratories). Cells were then fixed and incubated with rabbit anti-GFAP antibody followed by Rhodamine anti-rabbit IgG antibody.

Chemically synthesized siRNA targeting C4ST1, C6ST1 and scrambled siRNA (as a negative control) were obtained from Dharmacon (siGENOME™ SMARTpool®). Transient transfection into primary cultured astrocytes with 50 nM siRNA was carried out using the mouse astrocyte Nucleofector kit (Amaxa). Transfection efficiency was more than 80% based the simultaneous transfection of pmaxGFP™. Medium was replaced with DMEM 1 day after transfection and the cells were cultured for two further days. The cells were then treated with TGFβ1 for 24 hours and conditioned medium was collected for axonal guidance spot assays, immunoblotting and ELISA.

Total RNA was isolated from cultured astrocytes with the Absolutely RNA purification kit (Stratagene). Genomic DNA was removed by DNaseI treatment following the manufacturer's protocol. RNA was reverse transcribed using SuperScript III (Invitrogen) and real-time PCR was performed on a Chromo4 (MJ Research) with DyNAmoT™ HS SYBR Green qPCR kit (MJ Research). PCR conditions consisted of a 15 minutes hot start at 95°C, followed by 45 cycles of 15 seconds at 94°C, 15 seconds at 57°C and 25 seconds at 72°C. All samples were run in triplicate and results were normalized to the level of GAPDH. The primer sequences are as follows: C4ST1; 5′-GAAGAGGCTC ATGATGGTCC-3′ and 5′-GAGAGAGTAG ACCGTCTG CC-3′, C6ST1; 5′-GGATTCCACC TTTTCCCATCTG-3′ and 5′-TGCCCTGCTG GTTGAAGAAC-3′, and GAPDH; 5′-AAGGTGGTGA AGCAGGCATC TG-3′ and 5′-TGGGTGGTCC AGGGTTTCTT AC-3′. cDNAs for C4ST1, C6ST1 and GAPDH were used as a template for PCR to obtain standard curves.

Relative CSPG amounts were measured by ELISA. Briefly, 96-well microtiter plates (Immulon 4; Dynex Technologies) pretreated with poly-L-lysine were coated with conditioned medium of astrocytes treated with or without TGFβ1. After blocking, appropriate antibodies were incubated, followed by incubation with anti-mouse antibody F(ab′)2 fragment conjugated with HRP (Abcam). Binding was measured with a microplate reader (Labsystems Multiskan, MCC/340) using SureBlue TMB Microwell Peroxidase Substrate (KPL) as a substrate.

All experiments strictly adhered to the NIH guidelines on the care and use of animals in research. Adult mice (8-12 weeks old) were deeply anesthetized with ketamine-xylazine (100 and 14 mg/kg, respectively). A laminectomy was performed at the level of T12-L1 and the spinal cord was exposed. A dorsal overhemisection was made at T12. After the injury, the subcutaneous tissue and skin were sutured in layers. One day after surgery, animals were anesthetized then perfused intracardially with PBS, followed by 4% paraformaldehyde. Spinal cords were removed and frozen. Sagittal serial sections were cut on a cryostat (15 μm) and processed for histological analyses. Monoclonal antibodies LY111 and MC21C (both Seikagaku), were used to detect changes in 4-sulfation and 6-sulfation in intact chondroitin sulfate GAG chains, respectively, and polyclonal anti-GFAP antibody (Dako) was used to visualize reactive astrocytes, followed by incubation with FITC-conjugated anti-mouse μ chain antibody (Abcam) and Alexa Fluor 633-conjugated anti-rabbit Ig antibody (Molecular Probes). Fluorescent images were acquired with a confocal laser-scanning microscope (Leica SP1, Leica, Germany).

Disaccharide composition analysis was performed essentially as described previously (Kinoshita and Sugahara, 1999) with minor modification. Briefly, chondroitin sulfate oligosaccharides in 0.1M ammonium acetate buffer (pH 7.0) were treated with cABC as described above and lyophilized. Derivatization of the oligosaccharides with 2-aminobenzamide (2AB, Sigma) was carried out with 5 μl of 0.35 M 2AB, 1.0 M NaCNBH₄, 30% acetic acid in DMSO at 65°C for 2 hours. Fluorescently tagged oligosaccharides were separated by HPLC on an amine-bound silica PA03 column (Waters). The HPLC system was equilibrated with solvent A (15 mM ammonium phosphate containing 5% methanol) and solvent B (1.5 M ammonium phosphate containing 5% methanol). At a uniform flow rate of 0.75 ml/minute, a gradient was developed by holding solvent B at 0% for 5 minutes, then increasing from 0 to 18% over 14 minutes and changing from 18% to 50% over 11 minutes. Separation was monitored using a L-7485 fluorescence detector (Hitachi, Japan) with excitation and emission wavelengths of 330 nm and 420 nm, respectively. When conditioned medium was used as a source for CSPGs, concentrated samples were digested with proteinase K extensively, followed by cABC treatment. 3′-Sialyl-N-acetyllactosamine (Dextra Laboratories, UK) was added to GAG-containing fraction as an internal control. Disaccharide composition analysis was performed as described above.

Disaccharide composition of chondroitin sulfate chains in the spinal cord sections on the glass slides were determined as described previously (Mitsunaga et al., 2006). Briefly, coronal cryosections of spinal cords were treated with cABC (25 mU/ml) for 16 hours at 37°C together with 3′-Sialyl-N-acetyllactosamine on the glass slides. The solution was recovered and disaccharide fractions were enriched by size exclusion chromatography (Superdex Peptide10/300 GL, GE Healthcare) in 150 mM ammonium bicarbonate at a flow rate of 0.3 ml/minute. Separation was monitored using a L-7400 UV detector (Hitachi, Japan) with absorbance of 232 nm. Purified disaccharides were derivatized with 2-AB.

Partial digestion of CS-A was performed with cABC (10 mU/ml) at room temperature and the digestion was monitored by measuring the absorbance at 232 nm. The digestion degree was calculated based on the comparison of the absorbance measured at 232 nm at the time of aliquot removal and the one at the time of reaction completion. After heat inactivation of the enzyme, chondroitinase-treated CS-A was precipitated with ethanol and subjected to axonal guidance spot assays.

Adult mouse dorsal root ganglia (DRG) prepared from C57/BL6 were used for axon guidance spot assay and images of the cells were acquired every minute over a period of 2 hours. Images were acquired on a TE-2000 Nikon and transferred to a computer using Metamorph software.

We thank M.V. Sofroniew for spinal cord tissues, and R. Adelstein, J. Sellers, N. Epstein, and Z. H. Sheng for critical comments; S. Wen, T. Laabs, K. Vartanian, and Y. Tailor for technical support. We are grateful to R. F. Shen and C. A. Combs for help with data collection at Proteomics Core Facility and Light Microscopy Core Facility in National Heart, Lung, and Blood Institute, NIH.