Different types of glioma cell lines and primary cultures were used for in vitro experiments. Human high-grade glioma cell lines U87MG and A172 were obtained from American Type Culture Collection (Manassas, VA, USA). The gliomas named as ‘HTZ' were primary tumour cell cultures derived from surgical specimens of human high-grade gliomas (WHO Grade IV) as described before (Bogdahn et al, 1989). All tumour cells were maintained as standard monolayer cultures in tumour growth medium at 37°C, 5% CO₂, 95% humidity in a standard tissue culture incubator. Growth medium was comprised of Dulbecco's modified Eagle's medium (Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal calf serum (FCS).

To elucidate the effect of exogenous TGF-β2 on the regulation of versican isoforms, we performed treatment assays with different concentrations of TGF-β2; glioma cells were seeded at an equal density in cell culture flasks containing growth medium as described above. After 24 h, triplicates of subconfluent cells were treated with four different concentrations (1, 5, 10 and 50 ng ml⁻¹) of activated rhTGF-β2 protein (R&D Systems, Minneapolis, MN, USA) and incubated for 72 h. Cells and supernatants were harvested to prepare mRNA or protein as described below. In time-point assays, cells were treated with 10 ng ml⁻¹ of TGF-β2 and harvested at three different time points: days 1, 3 and 5.

The TGF-β2-specific antisense phosphorothioate oligodeoxynucleotides (PTOs) (AS-11) as described previously (Nickl-Jockschat et al, 2007) was used with the sequence 5′-GTAGTGCATTTTTTAAAAAA-3′ (mRNA target region 171–190) (Sigma-Genosys, Steinheim, Germany). As a control PTO, we used NS (mis) with three mismatch bases (sequence: 5′-GTAATGAATGTTTTAAAAAA-3′).

Total RNA was extracted from tumour cells with the RNA purification system RNeasy Mini Kit (Qiagen, Hilden, Germany) following the manufacturer's instructions. RNA concentration and purity was determined by measuring optical density at wavelengths of 260 and 280 nm using a standard spectrophotometer. First-strand cDNA generated from 1 μg of total RNA samples by using a reverse transcription kit (Promega, Madison, WI) was used to amplify gene-specific cDNAs from expressed genes. Appropriate forward and reverse primers to detect transcripts of interest were used in reverse transcriptase polymerase chain reaction (RT–PCR) reactions for cDNA amplification. The primers used were as follows: TGF-β2 (forward: 5′-TCTAGGGTGGAAATGGATACACGAACC-3′; reverse: 5′-TGTTACAAGCATCATCGTTGTCGTCG-3′) resulting in a 314 bp fragment. Versican primers (forward: 5′-GTGACTATGGCTGGCACAAATTCC-3′; reverse: 5′-GGTTGGGTCTCCAATTCTCGTATTGC-3′) were designed to detect all known splice variants of the gene resulting in 229 bp fragment.

The specific primers for the PCR amplification of versican isoforms were used as described before (Cattaruzza et al, 2002). The sizes of the amplification products were 405 bp for the V0 splice form (forward: 5′-TCAACATCTCATGTTCCTCCC-3′ and reverse: 5′-TTCTTCACTGTG GGTATAGGTCTA-3′), 336 bp for V1 (forward: 5′-GGCTTTGACCAGTGCGATTAC-3′; reverse: 5′-TTCTTCACTGTGGGTATAGGTCTA-3′), 498 bp for V2 (forward: 5′-TCAACA TCTCATGTTCCTCCC-3′; reverse: 5′-CCAGCCATAGTCACA TGTCTC-3′) and 429 bp for V3 (forward: 5′-GGCTTTGACCAGTGCGATTAC-3′; reverse: 5′-CCAGCCATAGTCACATGTCTC-3′). Annealing temperatures were optimised for each primer pair using the following program: DNA polymerase was activated at 95°C for 5 min, amplified for 30 cycles (95°C for 45 s, 57–60°C for 1 min, 72°C for 45 s) and extended at 72°C for 5 min. RT–PCR products were analysed on 1% agarose gel and visualised with ethidium bromide staining. The housekeeping gene β-actin was used as a positive control to assess cDNA quality.

To precisely quantify mRNA expression, a real-time PCR system (ABI PRISM 7000 Sequence Detection System, CA, USA) that measures nucleic acid molecules based on the detection of a fluorescent reporter molecule (SYBR Green dye) was used. Target-cDNA-specific primers as described above were established. Briefly, five serial twofold dilutions of cDNA were amplified in triplicates to construct standard curves for both the target gene and the endogenous reference (β-actin). Standard curves generated by the software were used for extrapolation of expression levels for the unknown samples based on their threshold cycle (C_T) values. All amplifications of unknown samples were in the linear range. For each reaction melting curves and agarose gel electrophoresis of PCR products were used to verify the identity of the amplification products. Each probe was run in parallel with primers specific for β-actin as standard for quantification of target cDNA. The target gene amount was divided by the housekeeping gene amount to obtain a normalised target value. Each of the experimental normalised values was divided by the normalised control (untreated) sample value to generate the relative expression levels in fold.

Total cell lysates were prepared in radioimmuno precipitation assay lysis buffer (20 mM Tris, pH 7.4, 150 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate (SDS)), supplemented with Complete Protease Inhibitor Cocktail Tablets (Roche Molecular Biochemicals, Manheim, Germany). Equal amounts of total protein quantified in a Bicinchoninic acid (BCA) assay (Uptima, Montpellier, France) were incubated with protein G beads at 4°C for 6 h for pre-clearing. After centrifugation, the nonspecifically bound G beads were discarded and then supernatants were incubated with 2 μg ml⁻¹ versican V0/V1 Neo-rabit polyclonal antibody (ABR, Golden, CO) and incubated at 4°C overnight. After washing, the beads were boiled in 1 × protein loading dye for 5 min and loaded directly into pre-poured Tris-HCl-glycine SDS–polyacrylamide gel electrophoresis (PAGE) gels (10%) along with pre-stained molecular weight standards (Bio-rad Laboratories, Palo Alto, CA, USA). Electrophoresis was performed in Tris/glycine/SDS running buffer (Biorad Laboratories) at 125–150 V for a suitable migration period. Following transfer to polyvinylidene fluoride (PVDF) membranes (Biorad Laboratories) at 120 mA constant current for 1–2 h, blots were briefly washed in Tris buffered saline with Tween (TBST) (10 mM Tris, 150 mM NaCl, and 0.5% Tween-20, pH 8.0), blocked for 1 h at RT with 5% non-fat dry milk and then incubated with 1 μg ml⁻¹ of the versican V0/V1 Neo-antibody at 4°C overnight. Immunocomplexes were visualised using a horsedish peroxidase-conjugated antibody followed by chemoluminescence reagent (Pierce Biotechnology, Rockford, IL, USA) detection on photographic film.

Multi-cell tumour spheroids were initiated by seeding (3–8) × 10⁶ cells incubated in agar-coated wells in order to inhibit adhesion. Mature spheroids with a mean diameter of 200–250 μm were explanted to uncoated 96-well plates containing the corresponding treatment (TGF-β2, 10 ng ml⁻¹; versican V0/V1 Neo-antibody, 2–20 μg ml⁻¹; AS-11, 20 μM). Six spheroids were used for each experimental condition in each experiment. Spheroids were allowed to migrate for 5–7 days. Spreading of the spheroids was monitored by microscopic photographs of each spheroid after 0, 1, 3, 5 and 7 days. For quantification, the mean diameter of randomly selected glioma cells that had migrated from the tumour spheroid was measured by a blinded investigator and expressed in relation to the mean radial distance at time 0 h. Bovine serum albumin (BSA) was used as a control protein. As unrelated controls, normal rabit immunoglobulin G (IgG) (R&D Systems, Minneapolis, MN) and NS (mis) were used at the same concentrations, respectively. Assays were repeated at least twice.

After migration of cells from the spheroids, the spheroids were collected with a pipette tip. Total cell lysates of migrated cells and spheroids were prepared separately for protein expression. For versican expression, western blotting with versican V0/V1-specific antibody was used as described above.

The spreading and migration capabilities of HTZ-349 cells were assessed using a scratch (wound) assay measuring the expansion of a cell population on a given surface. The cells were seeded into uncoated six-well tissue culture dishes at a density of 2.5 × 10⁵ cells and cultured in medium containing 10% FCS to nearly confluent cell monolayers, which were then scratched using 1 ml sterile pipette tips. Any cellular debris was removed by washing with PBS. The wounded monolayers were then incubated in 10% FCS medium containing TGF-β2 (20 ng ml⁻¹) and BSA (20 ng ml⁻¹) as control for 48–72 h. The wound area in a marked field of view was inspected at different time intervals subsequently until closure to determine the distance migrated by the cells.

The wound areas were photographed under a light microscope by a blinded investigator. The width of the scratch was measured at 0 and 24, 48, 72 h after treatment to measure the distance traveled by the cells. The difference between the width of scratch (wound area) at 0 h and at a given time point represented the distance migrated by the cells. The quantification of the distance migrated by the cells expressed as percentages and comparison made with untreated control using statistical analysis. The experiments were repeated in duplicate wells at least three times.

The results (mean value+standard deviation) of control (untreated) vs stimulated (treated) cell samples were analysed using the Student's t-test for migration assays. The level of significance was set at ^*P<0.05.

To evaluate the profiles of TGF-β2 and versican in two human glioma cell lines (U87, A172) and in five primary cell cultures (HTZ-324, HTZ-349, HTZ-417, HTZ-419, HTZ-421), expression levels of TGF-β2 and versican isoforms were detected by RT–PCR at mRNA level using β-actin as control gene (Figure 1A). All cell lines expressed TGF-β2 in different amounts. Versican isoforms were found to be differentially expressed in high-grade glioma cells; among the four isoforms, V1 was the most prominent one and found to be expressed in all glioma cells. V0 was intermediately expressed in almost 30% of glioma cell lines. V3 had a heterogeneous expression pattern. The V2 isoform was detected as a faint signal only in HTZ-417 (Figure 1A).

Next, the modulation of versican expression by TGF-β2 was assessed. Treatment assays were performed with the glioma cell lines U87, A172, HTZ-417 and HTZ-349. These cell lines were chosen because they secrete different levels of TGF-β2 and also their growth characteristics provided consistent confluency in a short period of time.

First, we analysed versican expression at the mRNA level by quantitative PCR (qPCR) with a primer binding to the G3 domain which detects all isoforms. In these experiments, a significant increase of versican expression was detected in HTZ-349 cells treated with TGF-β2 in comparison to untreated cells, reaching a fold level of 2.3 with a maximum dose of TGF-β2 (50 ng ml⁻¹) (Figure 1B). Results with the other cell lines tested (U87, A172 and HTZ-417) were similar with a 1.9-, 1.7- and 2.1-fold increase, respectively (data not shown). In time-dependent assays, the upregulation of versican expression reached a peak after 72 h (2.2-fold) and then decreased gradually over time (Figure 1C).

Since versican V1 and V0 isoforms predominantly increase in tumours of different origin, suggesting that these isoforms are mainly involved in tumour development, we also elucidated whether TGF-β2 differentially modulates the expression of versican isoforms in high-grade glioma. We found that V1 was the most upregulated isoform with a 1.7-fold increase (Figure 1D). V0 was intermediately increased and the expression of V2 was not induced by TGF-β2 (data not shown).

Next, we examined the expression of versican at protein level with western blot and immunoprecipitation (IP) using a polyclonal antibody (Figure 2). The versican V0/V1 Neo-antibody (Affinity BioReagents, Golden, CO, USA) is a rabbit polyclonal antibody to the versican Asp-Pro-Glu-Ala-Ala-Glu (DPEAAE) neo-epitope at the GAG-β domain of human versican and recognises the N-terminal (G1 domain) neo-epitope cleavage products of versican after cleavage with ADAMTS at the Glu441–Ala442 bond in the V1 isoform; the corresponding peptide bond is Glu1428–Ala1429 in the V0 isoform (Sandy et al, 2001). As the antibody does not react with the DPEAAE sequence when it is present in intact versican (V0) (Sandy et al, 2001; Kern et al, 2007), no high molecular mass band representing intact versican (V0), which migrates at 350–400 kDa (data not shown), could be detected in low-density gels (4%) and after chondroitinase ABC digestion. However, V1, which runs around 280 kDa, could be detected with high versican V1 concentrations (Russell et al, 2003; Kern et al, 2006).

In IP without enzymatic digestion of chondroitin sulphate chains, the products had the characteristic smear of proteoglycans to which GAG chains of different sizes have been covalently attached as described before (Yang et al, 1999). We detected high molecular mass intact versican V1 at around 300 kDa, which is the expected molecular mass of V1 without chondroitinase digestion in glioma cell lines (Dours-Zimmermann and Zimmermann, 1994) verifying that the antibody reacts with the DPEAAE sequence in intact versican V1 (Figure 2A). The products migrating around 75 and 50 kDa represent the G1-DPEAAE-cleaved products of human versican V1 by ADAMTS-1 and ADAMTS-4 as described before with the same antibody (Sandy et al, 2001; Russell et al, 2003; Somerville et al, 2003; Kern et al, 2006). These signals representing intact versican V1 and its cleavage products were upregulated with exogenous TGF-β2 in a concentration-dependent manner. The highest signals were observed with 10 and 50 ng ml⁻¹ of TGF-β2, consistent with qPCR results. Densitometric analysis also allowed quantification of this upregulation, which showed almost 80% increase compared to untreated control (Figure 2B). Taken together, these results demonstrate that exogenous TGF-β2 induces not only versican V1 expression, but also cleavage of versican V1 via ADAMTS-1 and ADAMTS-4.

Next, we examined the effect of TGF-β2 on glioma migration. We generated tumour cell spheroids from the HTZ-349 cell line and allowed the cells to migrate from the spheroids in the absence or presence of TGF-β2 (20 ng ml⁻¹). Exposure to TGF-β2 significantly induced the migration rate of glioma cells in spheroid assays compared to untreated control and BSA (Figure 3A). Quantification of the migration rate in scratch assay showed significantly increased migration with TGF-β2 treatment, in parallel to spheroid assay results (Figure 3B; ^**P<0.01).

To elucidate the mechanisms involved in enhancement of glioma migration mediated by TGF-β2, we evaluated whether the upregulation of the versican isoforms V0/V1 by TGF-β2 might be involved in this process. We used versican V0/V1 Neo-antibody to block the GAG-β chain and IgG-unspecific controls. In these assays, the migration rate was significantly higher in TGF-β2 20 ng ml⁻¹-treated spheroids compared to untreated controls (^**P<0.01) (Figure 4B). As we detected previously that TGF-β2 had no effect on proliferation of HTZ-349 cells (unpublished data), proliferation is very likely to have no contribution to the increased area covered by cells migrating away from spheroids. When versican V0/V1 antibody was added to TGF-β2-treated spheroids, the enhancement of glioma migration by TGF-β2 was reversed in a dose-dependent manner. With 20 μg ml⁻¹ of versican V0/V1 antibody, migration was even inhibited compared to untreated control and control IgG (^**P<0.01). We also observed similar results with versican V0/V1 antibody in the absence of exogenous TGF-β2. These results might explain that, even without exogenous TGF-β2, the anti-GAG-β antibody interferes with the interaction of versican with endogenous TGF-β2 and inhibits glioma migration, while HTZ-349 expresses and secretes high levels of TGF-β2 (Figures 1A and 2). AS-11-treated cells showed significant decreased migration rate (28.6%) in comparison to untreated control (100%) and control oligonucleotide (87.2%) (P<0.001) (Figure 3C).

We also detected that cells migrating from spheroids express significantly higher levels of versican V1 compared to immobile glioma cells within spheroids (Figure 4A) which further supports the involvement of versican in glioma migration.