Gene-linked shift in ganglioside distribution influences growth and vascularity in a mouse astrocytoma

Abstract

Brain tumor growth and progression is dependent upon vascularity, and is associated with altered ganglioside composition and distribution. In this study, we examined the influence of gangliosides on growth and vascularity in a malignant mouse astrocytoma, CT-2A. Ganglioside distribution was altered in CT-2A tumor cells using an antisense construct to β-1,4-N-acetylgalactosaminyltransferase (GalNAc-T), a key enzyme that uses the simple ganglioside GM3 as a substrate for the synthesis of the more complex gangliosides, GM2, GM1 and GD1a. GalNAc-T gene expression was significantly lower in CT-2A cells stably transfected with the antisense GalNAc-T plasmid, pcDNA3.1/TNG (CT-2A/TNG) than in either non-transfected CT-2A or mock-transfected (CT-2A/V) control tumor cells. GM3 was elevated from 16% to 58% of the total ganglioside distribution, whereas GM1 and GD1a were reduced from 17% and 49% to 10% and 17%, respectively, in CT-2A/TNG tumor cells. Growth, vascularity (blood vessel density and Matrigel assay) and vascular endothelial growth factor (VEGF) expression was significantly less in CT-2A/TNG tumors than in control CT-2A brain tumors. In addition, the expression of VEGF, hypoxia-inducible factor 1α (HIF-1α) and neuropilin-1 (NP-1) was significantly lower in CT-2A/TNG tumor cells than in control CT-2A tumor cells. These data suggest that gene-linked changes in ganglioside composition influence the growth and angiogenic properties of the CT-2A astrocytoma.

Abbreviations used

DMEM

Dulbecco's modified Eagle's medium

ECM

extracellular matrix

EGFR

epidermal growth factor receptor

GalNAc-T

β-1,4-N-acetylgalactosaminyltransferase

HIF-1

hypoxia-inducible factor 1

HPTLC

high performance thin-layer chromatography

LacCer

lactosylceramide

MVD

microvessel density

M-MLV RT

Moloney murine leukemia virus reverse transcriptase

NP-1

neuropilin-1

PBS

phosphate-buffered saline

SCID

severe-combined immuno-deficient

VEGF

vascular endothelial growth factor

Gangliosides are sialic acid-containing glycosphingolipids present in the outer leaflet of plasma membranes in virtually all mammalian tissues, and are particularly enriched in the CNS. Aberrant distribution and concentration of gangliosides is associated with neoplastic development and may influence tumor progression in brain tumors through the modulation of cell growth, adhesion and angiogenesis (Berra et al. 1985; Manfredi et al. 1999; Zeng et al. 2000; Hakomori 2002; Birkle et al. 2003). Gangliosides influence tumor angiogenesis through interactions either at the cell surface (receptor and integrin function) or with the surrounding extracellular matrix (ECM), including other tumor and host (endothelial and macrophages) cells (Yohe et al. 1985; Ziche et al. 1989; Ziche et al. 1992; Koochekpour et al. 1996; Alessandri et al. 1997; Yates and Rampersaud 1998; Lang et al. 2001; Miljan and Bremer 2002). Gangliosides may therefore modulate various aspects of tumor progression, including angiogenesis.

Ganglioside biosynthesis occurs in the Golgi by the step-wise addition of monosaccharides to the ceramide moiety (Keenan et al. 1974; Yusuf et al. 1984; Yu et al. 2004). β-1,4-N-Acetylgalactosaminyltransferase or GM2/GD2 synthase (GalNAc-T, EC 2.4.1.92) is a key enzyme required for the synthesis of complex gangliosides (Fig. 1a). Specifically, GalNAc-T catalyzes the conversion of lactosylceramide (LacCer), GM3 and GD3 to asialo-GM2 (GA2), GM2 and GD2, respectively (Walton and Schnaar 1986; Sango et al. 1995; Furukawa and Takamiya 2002). The structurally simple monosialoganglioside GM3, containing a single-terminal sialic acid, inhibits tumor progression through reduced angiogenesis and cell proliferation (Alessandri et al. 1997; Manfredi et al. 1999; Noll et al. 2001). The more complex gangliosides such as GD3, GM2, GM1 and GD1a, containing longer oligosaccharide chains, enhance angiogenesis and proliferation (Ziche et al. 1992; Manfredi et al. 1999; Zeng et al. 2000; Lang et al. 2001; Liu et al. 2004). Hence, the relative distribution of simple and complex gangliosides may be important for tumor progression and malignancy.

Figure 1.  Complex ganglioside biosynthesis and the design of the antisense β-1,4-N-acetylgalactosaminyltransferase (GalNAc-T) plasmid, pcDNA3.1/TNG. (a) Synthesis of complex gangliosides through the ‘a’ metabolic pathway. Sialyltransferase I (ST1) adds a single sialic acid to the terminal galactose of lactosylceramide (LacCer) to form GM3, the precursor for the synthesis of complex gangliosides. GalNAc-T adds a β-linked N-acetylgalactosamine to the galactose of GM3 to form GM2. GM2 is converted to GM1 by the addition of a galactose, which is catalyzed by galactosyltransferase II (Gal T2). GM1 is converted to GD1a by the addition of a sialic acid to the terminal galactose of GM1, which is catalyzed by sialyltransferase IV (ST4). (b) A schematic representation of antisense GalNAc-T plasmid design. (c) An analysis of plasmid sequence by restriction enzyme digestion. Lane 1, DNA ladder; lane 2, pcDNA3.1/TNG digested with BamHI, which produced a single band at 5.9 kb; lane 3, pcDNA3.1/TNG double digested with BamHI and XbaI, which produced two bands at 5.0 and 0.9 kb; lane 4, pcDNA3.1 vector digested with XbaI, which produced a single band at 5.0 kb.

The malignant mouse astrocytoma CT-2A, which is fast growing and highly vascularized, expresses low levels of the simple ganglioside GM3 and high levels of the complex gangliosides GM2, GM1 and GD1a (Seyfried et al. 1992). Although CT-2A tumor cells express minor quantities of LacCer, they do not synthesize gangliosides of the ‘asialo’ metabolic pathway (Seyfried et al. 1996; Bai and Seyfried 1997). In addition, CT-2A tumor cells do not express sialyltransferase II (ST2 or GD3-synthase) and thus do not synthesize GD3 and gangliosides of the ‘b’ metabolic pathway (Seyfried et al. 1996; Bai and Seyfried 1997). Hence, the CT-2A tumor is a good model for studying the role of GM3 and complex gangliosides of the ‘a’ metabolic pathway in tumor growth and vascularity.

Brain tumor invasiveness and malignancy is positively associated with the degree of tumor vascularity and angiogenic potential (Leon et al. 1996; Chaudhry et al. 2001). Vascular endothelial growth factor (VEGF) is an endothelial cell-specific mitogen (Claffey et al. 1996; Ferrara and Davis Smyth 1997) and is a biomarker of angiogenesis in brain tumors (Chaudhry et al. 2001). VEGF is a secreted 40–45-kDa homodimer that exists in four major isoforms (VEGF₁₂₁, VEGF₁₆₅, VEGF₁₈₉ and VEGF₂₀₆) produced from alternative mRNA splicing. Although VEGF primarily enhances the survival and proliferation of endothelial cells, VEGF is also known to influence tumor cell growth (Podar et al. 2001). The transcription factor, hypoxia-inducible factor 1 (HIF-1), regulates gene expression in various biological processes to include angiogenesis, growth, cell survival and glycolysis (Lee et al. 2004). Under hypoxic conditions, HIF-1 binds to the hypoxia response element (HRE) upstream of the VEGF gene to initiate gene transcription. Although HIF-1 is a heterodimer of HIF-1α and HIF-1β subunits, HIF-1α expression is altered by hypoxia whereas HIF-1β is constitutively expressed. Both VEGF and HIF-1α are up-regulated in brain tumors and their expression is associated with tumor progression and angiogenesis (Zagzag et al. 2000; Chaudhry et al. 2001).

The biological functions of VEGF are mediated through two different tyrosine kinase receptors, VEGFR-1 (Flt-1) and VEGFR-2 (KDR/Flk-1) (Ferrara and Davis Smyth 1997). Both VEGFR-1 and VEGFR-2 bind VEGF with high affinity and are expressed predominately on endothelial cells, yet both receptors can be expressed on tumor cells (Herold-Mende et al. 1999). Recently neuropilin-1 (NP-1), a 130–135-kDa glycoprotein, was found to bind VEGF, suggesting a possible role in vascularity. In contrast to VEGFR-1 and VEGFR-2, NP-1 lacks an intracellular tyrosine kinase domain and acts as a co-receptor enhancing the binding of VEGF₁₆₅ to VEGFR-2 (Soker et al. 1998). NP-1 is associated with VEGF expression and tumor vascularity in human astrocytomas (Ding et al. 2000). Modulation of VEGF and/or VEGF receptor function may have therapeutic potential for brain cancer management.

Although previous studies have shown that gangliosides influence growth and angiogenesis, the mechanisms involved remain unclear (Manfredi et al. 1999; Zeng et al. 2000). In this study, we examined the effects of shifting the distribution of GM3 and complex gangliosides (GM1 and GD1a) on the growth and angiogenic properties of the CT-2A astrocytoma. An antisense GalNAc-T plasmid, pcDNA3.1/TNG, was produced and transfected into CT-2A tumor cells (CT-2A/TNG). This elevated GM3 while reducing GM1 and GD1a. Brain tumor growth and vascularity was significantly lower in CT-2A/TNG tumors than in control CT-2A and CT-2A/V (mock-transfected or vector alone) tumors. In addition, we observed a reduction in HIF-1α and NP-1 mRNA expression in the CT-2A/TNG tumor cells, indicating a possible mechanism by which gangliosides may modulate tumor angiogenesis.

Experimental procedures

Mice and experimental brain tumors

The C57BL/6J (B6) strain and the BALBc/J-severe combined immunodeficient (SCID) strain were obtained from the Jackson Laboratory (Bar Harbor, ME, USA). The mice were propagated in the animal care facility of the Department of Biology of Boston College, using the animal husbandry conditions described previously (Flavin et al. 1991). The syngeneic mouse brain tumor, CT-2A, was originally produced by implantation of a chemical carcinogen, 20-methylcholanthrene, into the brains of B6 mice (Zimmerman and Arnold 1941; Seyfried et al. 1992). The CT-2A tumor arose in the cerebral cortex and was characterized as a malignant anaplastic astrocytoma (Seyfried et al. 1992). The morphological, biochemical and growth characteristics of the CT-2A brain tumor have been previously described (Seyfried et al. 1992). Male B6 mice (8–12 weeks of age) were used as tumor recipients. All animal experiments were carried out with the ethical committee approval in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Institutional Care Committee.

Cell culture

The CT-2A cell line was established as previously described by Seyfried et al. (1992). All CT-2A tumor cell lines were maintained in Dulbecco's modified Eagle's medium (DMEM) (Sigma, St Louis, MO, USA) supplemented with 10% fetal bovine serum (Sigma) and 0.5% penicillin/streptomycin (Sigma). The cells were cultured in a CO₂ incubator with a humidified atmosphere containing 95% air and 5% CO₂ at 37°C.

Construction of antisense GalNAc-T plasmid and stable transfection

The expression vector pCMV containing a 1.6-kb mouse GalNAc-T cDNA insert (a gift from Dr Richard Proia, NIH, Bethesda, MD, USA) was digested to produce a 0.9-kb fragment with 5′XbaI and 3′BamHI overhangs. The expression vector pcDNA 3.1/Zeo (Invitrogen, San Diego, CA, USA) was digested to produce 5′BamHI and 3′XbaI overhangs. The restriction digests were electrophoresed, were gel purified using the CONCERT Matrix Gel Extraction System (Invitrogen), and were ligated at 4°C overnight. Competent Escherichia coli cells were transformed, ampilicin-resistant clones were selected and plasmid DNA was isolated using Qiagen Midi preps (Qiagen, Valencia, CA, USA). The plasmid containing the 0.9-kb GalNAc-T cDNA in the reverse orientation (from 3′ to 5′) was confirmed by sequence analysis and double digestion with XbaI and BamHI and named the antisense GalNAc-T plasmid, pcDNA3.1/TNG (GenBank DQ286394). CT-2A tumor cells were stably transfected either with the expression vector pcDNA3.1 alone (CT-2A/V) or with pcDNA3.1/TNG using FUGene 6 (Roche Molecular Biochemicals, Indianapolis, IN, USA) following the manufacturer's protocol. Cells that survived treatment with 400 µg/mL Zeocin (Invitrogen) for 2 weeks were expanded and analyzed for GalNAc-T expression using RT-PCR. Cells that expressed lower GalNAc-T levels were subcloned to ensure a single clonal population.

Semi-quantitative RT-PCR

Total RNA was isolated from either homogenized cell pellets or tumor tissues using TRIzol Reagent (Invitrogen) following the manufacturer's protocol. RNA concentration and purity was determined by spectrophotometric measurements at 260 nm and 280 nm. Single strand cDNA was synthesized from total RNA (3 µg) using oligo (dT) primers (Promega, Madison, WI, USA) in a 20 µL reaction with Moloney murine leukemia virus reverse transcriptase (M-MLV RT; Promega) according to the manufacturer's protocol. cDNA (3 µL) was used for PCR amplification of various genes (Table 1). Gradient PCR was performed to obtain optimal primer annealing temperatures. In order to determine the optimal linear range for the amplification reaction, PCR was performed at increasing cycle numbers. PCR amplification was performed with Taq DNA polymerase (Promega) using the following protocol: initial denaturation at 94°C for 2 min, followed by the previously determined optimal number of cycles of denaturation at 94°C for 1 min; annealing at the optimal primer annealing temperature for 45 s; extension at 72°C for 1 min (except for GalNAc-T amplification, extension at 72°C for 1 min adding 3 s every cycle); and a final extension at 72°C for 6 min following the last cycle. PCR products (5–15 µL) were separated on 0.8–1.5% agarose gels containing ethidium bromide, visualized with UV light, and analyzed using 1d Kodak Software (Eastman Kodak Co, Rochester, NY, USA). RT-PCR was performed on the total RNA of each sample in the absence of reverse transcriptase to control for possible DNA contamination.

Table 1.   Primer sequences and PCR conditions

Gene	GenBank Accession	Primer sequence	nt position^a	Annealing temperature (^oC)	Cycle	Product (bp)
a Primer nucleotide positions are based upon the CDS of the given GenBank accession number, except for VEGF. ^b Primer sequences for β-actin are based upon Kitakata, et al. 2002. ^c Primer sequences for VEGF are based upon Manfredi, et al. 1999.
β-actin^b	N_M007393	For 5′-TGT GAT GGT GGG AAT GGG TCA G-3′	126-147	58	22	514
		Rev 5′-TTT GAT GTC ACG CAC GAT TTC C-3′	639-618
GalNAc-T	L25885	For 5′-GAG ACT CGA TCT GCT AGA CCT TG-3′	141-163	63	30	1208
		Rev 5′-CCA CTT GGC GCA CTT TGT CTG-3′	1348-1328
Gal T2	AF082504	For 5′-TGC TCA TCC TGG TGT GTA CG-3′	218-237	69	30	550
		Rev 5′-GAG GCA TAG GGT GGA AAA GG-3′	767-748
HIF-1α	N_M010431	For 5′-GGT GCT AAC AGA TGA CGG CGA C-3′	303-324	67	28	365
		Rev 5′-TCA GCA CCA AGC ACG TCA TGG G-3′	667-646
NP-1	N_M008737	For 5′-GGT GTT CAT GAG GAA GTT CAA G-3′	1527-1548	65	28	551
		Rev 5′-TGA AGT TGC CAT CTC CTG TAT G-3′	2077-2056
ST1	AF119416	For 5′-TCC TGC ACG GAC TAG AGC TG-3′	440-459	69	30	496
		Rev 5′-GGT TGA CTG AGG TCG TAG CC-3′	935-916
ST4	N_M009178	For 5′-CAA CCG TTC TAG GGA ACA GC-3′	180-199	69	30	497
		Rev 5′-GGT TGA CAT CCC AGA TGA GG-3′	676-657
VEGF^c	BC061468	For 5′-TGG ATC CAT GAA CTT TCT GCT-3′	388-408	60	33	400–600
		Rev 5′-GAA TTC ACC GCC TCG GCT TGT-3′	971-951
VEGFR-1	N_M010228	For 5′-ACT GCG AAG CCA CCG TCA AC-3′	620-639	67	28	441
		Rev 5′-AGG ACC GTC TTC CTG CTG TG-3′	1060-1041
VEGFR-2	N_M010612	For 5′-CAG TGA CCA GCA TGG CAT CG-3′	393-412	66	32	555
		Rev 5′-ATC CGT CCA CTG GAC GCT AC-3′	947-928

Radiolabeling of cells

Either untransfected or stably transfected CT-2A tumor cells were seeded into 75-cm² tissue culture flasks at approximately equal densities and cultured until 50–60% confluent. Cells were then labeled with 0.3 µCi/mL of [¹⁴C]galactose (New England Nuclear, Boston, MA, USA) as described previously by Bai and Seyfried (1997). After 48 h, the culture medium was removed and the cells were washed with 10 mL of cold phosphate-buffered saline (PBS). The cells were removed from the flask with 0.25% trypsin containing 1 mm EDTA, harvested by centrifugation at 400 g, and counted using a hemocytometer.

Ganglioside isolation

Total lipids were extracted from the radiolabeled cells by adding 5 mL of chloroform/methanol (1 : 1 v/v). Polar and non-polar lipids were separated by Folch partitioning (Folch et al. 1957; Ecsedy et al. 1998). Gangliosides were isolated and purified by a modification of methods previously described (Seyfried et al. 1996; Bai and Seyfried 1997; Ecsedy et al. 1998). Briefly, the Folch ‘upper phase’ was converted to a chloroform : methanol : water ratio of 30 : 60 : 8 (v/v/v). The samples were added to a DEAE-Sephadex A-25 ion exchange column (Pharmacia Biotech, Uppsala, Sweden) that was equilibrated in 30 : 60 : 8 chloroform : methanol : water (v/v/v). Neutral lipids were eluted from the column with 30 mL of the same solvent. The gangliosides were eluted from the column with 30 mL of 30 : 60 : 8 chloroform : methanol : 0.8 m sodium acetate (v/v/v). The solvents were evaporated under vacuum and then treated with base 0.5 N NaOH at 37°C for 1.5 h and desalted as described previously (Seyfried et al. 1987; Ecsedy et al. 1998).

Ganglioside analysis

The level of radioactive incorporation into gangliosides was determined by placing 50 µL of sample into 6 mL of Ecoscint A scintillation solution (National Diagnostics, Atlanta, GA, USA) and counting dpm on a 1219 Rackbeta Counter (LKB Wallac, Turka, Finland). Ganglioside distribution was analyzed using high performance thin-layer chromatography (HPTLC), visualized with a Phosphorimager (Molecular Dynamics, Sunnyvale, CA, USA) and quantified using either Bioscanning or Phosphorimaging as described previously (El-Abbadi and Seyfried 1994; Seyfried et al. 1994; Ecsedy et al. 1998).

Cell proliferation

In vitro tumor cell proliferation was analyzed using the CellTiter 96 Non-Radioactive Cell Proliferation Assay (Promega) following the manufacturer's protocol. Briefly, 1 × 10³ cells/200 µL DMEM were seeded into 10 wells of a 96-well plate for each cell line (1 × 10³ cells/well). The plates were placed in a CO₂ incubator with a humidified atmosphere containing 95% air and 5% CO₂ at 37°C. After 24 h, 15 µL of tetrazolium salt dye solution was added to each well and the plate was returned to the incubator for 4 h. Stop solution (100 µL) was then added to solubilize the metabolite and to lyse the cells. The plate was returned to the incubator. After 24 h, the plate was shaken for 2–5 min and read in a microplate reader (Bio-Rad Laboratories, Hercules, CA, USA) at 595 nm. This procedure was repeated for each day of growth and the experiment was performed in triplicate.

Intracerebral tumor growth

Tumor cells were grown in culture, harvested with 0.25% trypsin containing 1 mm EDTA, and pelleted by centrifugation at 400 g. Cell pellets were washed twice with serum-free DMEM and implanted into the cerebral cortex of B6 mice using a trocar (Ranes et al. 2001). The resulting tumor became the donor tissue for the growth study, for histology, for RT-PCR and for western blot analysis. Briefly, mice were anesthetized with Avertin (0.1 mL/10 g body weight) intraperitoneally and their heads were shaved and swabbed with 70% ethyl alcohol under sterile conditions. Small tumor pieces (about 1 mm³, estimated using a 1-mm × 1-mm grid) from the donor tumor were implanted into the right cerebral hemisphere of anesthetized recipient mice as previously described (Ranes et al. 2001; Mukherjee et al. 2004). Intracerebral tumor growth was analyzed directly by measuring the total tumor weight 14 days after tumor implantation. The mice used for these studies were killed using isoflurane anesthesia (Halocarbon, River Edge, NJ, USA). Tumors were dissected from normal-appearing brain tissue, analyzed as wet weights, frozen at −80°C and used for RT-PCR and western blot analyses.

Immunohistochemistry and microvessel density

Fourteen days after tumor implantation, tumor samples were fixed in 10% neutral buffered formalin (Sigma), embedded in paraffin and sectioned. The sections (5 µm) were incubated with trypsin at 37°C for 30 min after deparaffinization, rehydration and washing as previously described (Mukherjee et al. 2004). Briefly, the sections were quenched with 0.3% H₂O₂-methanol for 30 min and then blocked with 10% normal goat serum in 100 mL of 0.01 m phosphate and 0.9% sodium chloride (pH 7.4) with 1.0 g of bovine serum albumin and 0.1 mL of Tween 20 [phosphate buffer plus albumin (PBA) buffer]. The sections were treated with rabbit polyclonal antibody directed against human factor VIII-related antigen (Dako Corp., Carpinteria, CA, USA; 1 : 100 dilution with PBA) followed by a biotinylated anti-rabbit IgG at 1 : 100 dilution (Vector Laboratories, Inc., Burlingame, CA, USA). The sections were then treated with avidin-biotin complex followed by 3,3′-diaminobenzidine as substrate for staining according to the manufacturer's directions (Vectastain Elite ABC kit; Vector Laboratories, Inc.). The sections were then rinsed three times with PBS (0.01 m phosphate buffer with 0.9% NaCl). Sections were counterstained with methyl green and mounted. Corresponding tissue sections without primary antibody served as negative controls. Microvessel density (MVD) was quantified by examining areas of vascular hot spots as described previously (Weidner et al. 1991; Mukherjee et al. 2004). Sections were scanned at low magnification (×40 and ×100) for the localization of vascular hot spots. The three most vascular areas of the tumor, not containing necrosis, were determined and then counted at higher magnification (×200). The values of the three sections were averaged for all tumors. Branching structures were counted as a single vessel as described previously (Mukherjee et al. 2004).

Western blot analysis

Tumor tissues were homogenized in ice-cold lysis buffer (Cell Signaling Technology, Beverly, MA, USA) containing 20 mm Tris-HCl (pH 7.5), 150 mm NaCl, 1 mm Na₂EDTA, 1 mm EGTA, 1% Triton, 2.5 mm NaPP_i, 1 mmα-glycerophosphate, 1 mm Na₃VO₄, 1 µg/mL leupeptin and 1 mm phenylmethylsufonyl fluoride. Lysates were transferred to Eppendorf tubes, mixed on a rocker for 1 h at 4°C, and then centrifuged at 8100 g for 20 min. Supernatants were collected and protein concentrations were estimated using the Bio-Rad DC protein assay. Approximately, 8 µg of total protein from each sample were loaded on a 12% sodium dodecyl sulfate–polyacrylamide gel (Bio-Rad) and electrophoresed. Proteins were transferred to a polyvinylidene difluoride (PVDF) immobilon TM-P membrane (Millipore) overnight at 4°C and blocked in 5% non-fat powdered milk in Tris-buffered saline with Tween 20 (pH 7.6) for 3 h. Blots were then probed with VEGF mouse monoclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA) overnight at 4°C with gentle shaking. The blots were then incubated with anti-mouse whole horseradish peroxidase-conjugated secondary antibody at room temperature (∼22°C). Bands were visualized using enhanced chemiluminescence plus system (Amersham Pharmacia Biotech, Piscataway, NJ, USA). Blots were reprobed with β-actin antibody (Novus Biologicals, Littleton, CO, USA) as a loading control, and the ratio of VEGF to actin was analyzed by scanning densitometry.

In vivo matrigel model of angiogenesis

The experiments were performed as previously described by Manfredi et al. (1999). Briefly, tumor cells were grown in culture and harvested with 0.25% trypsin containing 1 mm EDTA. The cells were washed twice, resuspended in serum-free DMEM, and then thoroughly mixed with Matrigel (Collaborative Biomedical, Bedford, MA, USA) 1 : 2 (v/v) at 4°C. Male BALB/c-SCID mice were anesthetized with Avertin (0.1 mL/10 g body weight) and then injected with 1 × 10⁶ cells in 300 µL of Matrigel subcutaneously (s.c.) in the dorsal midline using a pre-chilled tuberculin syringe (27-gauge needle). Seven days after implantation, Matrigel plugs with the surrounding skin were removed as previously described and vascularity was photographed (Manfredi et al. 1999; Mukherjee et al. 2002).

Statistical analysis

All measurements were analyzed by the one-way analysis of variance (anova) followed by Fisher's protected least significant difference (PLSD) to calculate two-sided pair-wise comparison among different test groups by use of StatView 5.0 (SAS Institute, Cary, NC, USA). For statistical analysis, the control CT-2A and CT-2A/V groups were pooled and compared with the CT-2A/TNG group.

Results

Stable transfection with the antisense GalNAc-T plasmid down-regulates GalNAc-T expression and influences ganglioside distribution

An expression vector containing an antisense cDNA sequence to a conserved region in the GalNAc-T gene, pcDNA3.1/TNG, was constructed and stably transfected into CT-2A tumor cells (CT-2A/TNG) (Figs 1b and c). CT-2A tumor cells were also stably transfected with the expression vector pcDNA3.1 alone as a control (CT-2A/V). Semi-quantitative RT-PCR was used to examine GalNAc-T expression in the control cells and the stable transfectants. Although GalNAc-T mRNA was detected in all CT-2A tumor cells, the relative expression of GalNAc-T was about 50% less in the CT-2A/TNG tumor cells than in the control CT-2A and CT-2A/V tumor cells (Figs 2a and b).

Figure 2.  Analysis of β-1,4-N-acetylgalactosaminyltransferase (GalNAc-T) expression and ganglioside distribution in CT-2A, CT-2A/V and CT-2A/TNG tumor cells. (a) Detection of GalNAc-T gene expression by RT-PCR. The 1208-bp amplification product was detected in CT-2A (C), CT-2A/V (V) and CT-2A/TNG (T) tumor cells. PCR was performed using β-actin primers as an internal control and PCR conditions as described in the ‘Experimental procedures’. (b) The ratio of GalNAc-T to β-actin. The values are expressed as the means of three independent samples; the error bars represent ± SE. *Significant compared with control CT-2A and CT-2A/V cells at the p < 0.01 level. (c) High performance thin-layer chromatography (HPTLC) of ¹⁴C-labeled gangliosides. Approximately 3000 dpm of labeled gangliosides were spotted for each sample. The plate was developed by one ascending run with chloroform : methanol : water containing 0.2% CaCl₂ 2H₂O (55 : 45 : 10). After visualizing radiolabeled lanes with a Phosphorimager, the plate was sprayed with resorcinol reagent and heated to 100°C to visualize standards (Stds). (d) The percentage distribution of gangliosides. The values are expressed as the means of either two (CT-2A) or three (CT-2A/V and CT-2A/TNG) independent experiments; the error bars represent ± SE. *Significant compared with control CT-2A and CT-2A/V cells at the p < 0.01 level.

Differential GalNAc-T expression was correlated with alterations in the neosynthesis of tumor gangliosides, as detected by the HPTLC of radiolabeled gangliosides (Fig. 2c). CT-2A and CT-2A/V tumor cells expressed GalNAc-T and synthesized the complex gangliosides GM2, GM1 and GD1a as major species, with lower levels of the simple ganglioside GM3. In contrast, CT-2A/TNG tumor cells synthesized GM3 as its major ganglioside with lower levels of GM1 and GD1a (Fig. 2c). GM3, GM2 and GM1 migrated as doublets, whereas GD1a migrated as a triplet on the HPTLC because of the heterogeneity in the ceramide structure, as previously described (Bai and Seyfried 1997; Manfredi et al. 1999). Although GM3 comprised 12–16% of the total radiolabeled gangliosides in the control CT-2A and CT-2A/V tumor cells, GM3 comprised 58% of the total in the CT-2A/TNG tumor cells. In addition, the complex gangliosides GM1 and GD1a, which comprised 16–17% and 49–54% of the total radiolabeled gangliosides in the control CT-2A and CT-2A/V tumor cells, comprised 10% and 17% of the total in the CT-2A/TNG tumor cells, respectively (Fig. 2d). GM2 levels in the CT-2A/TNG tumor cells were similar to those of control CT-2A and CT-2A/V tumor cells. Also, we examined whether the down-regulation of GalNAc-T affected the total ganglioside synthesis as well as the ganglioside distribution. Total ganglioside synthesis was analyzed by the incorporation of radiolabeled [¹⁴C]galactose into tumor cell gangliosides. The rate of total ganglioside synthesis, expressed as dpm/10⁶ cells, was similar in the CT-2A/TNG tumor cells (13,547, n = 2) and in the control CT-2A and CT-2A/V tumor cells (13,936, n = 5). These findings show that the down-regulation of GalNAc-T expression in CT-2A tumor cells elevated GM3 and reduced GM1 and GD1a without altering the total ganglioside concentration. In addition, we found that the down-regulation of GalNAc-T influenced the expression of other glycosyltransferases in the ‘a’ metabolic pathway (Fig. 1a). Specifically, sialyltransferase I (ST1) expression was elevated by 35%, whereas galactosyltransferase II (Gal T2) and ST4 expressions were reduced by 63% and 27%, respectively, in CT-2A/TNG tumor cells compared with control CT-2A and CT-2A/V tumor cells (data not shown). These changes in glycosyltransferase expression are consistent with the changes in ganglioside distribution.

A shift in ganglioside distribution inhibits brain tumor growth and vascularity

Brain tumor growth studies were performed to determine whether changes in ganglioside distribution altered the orthotopic growth rate of the CT-2A brain tumor. Tumor weights were approximately 48% less in the CT-2A/TNG tumors than in the control CT-2A and CT-2A/V tumors (Fig. 3). In addition, we examined blood vessel densities and VEGF mRNA and protein levels using factor VIII immunostaining, RT-PCR and western blot analysis, respectively, to determine the effects of gangliosides on tumor vascularity. The number of blood vessels was significantly less in the CT-2A/TNG tumors than in control CT-2A and CT-2A/V tumors (Fig. 4). Also, VEGF mRNA and protein levels were significantly reduced in the CT-2A/TNG tumors than in the control CT-2A and CT-2A/V tumors (Fig. 5). Tumor cells shed gangliosides into their surrounding microenvironment, which may influence tumor–host interactions such as angiogenesis (Ladisch et al. 1987; Ziche et al. 1989; Ziche et al. 1992; Olshefski and Ladisch 1996; Alessandri et al. 1997; Lang et al. 2001). The in vivo Matrigel angiogenesis model represents early events of angiogenesis and tumor progression, and is dependent on the activation and infiltration of host stromal cells, which include monocytes, macrophages and endothelial cell precursors (Claffey et al. 2001; Mukherjee et al. 2002). When tumor cells are implanted within Matrigel, their effect on the developing stroma is more selective for soluble factors, such as tumor-derived gangliosides (Manfredi et al. 1999). As CT-2A, CT-2A/V and CT-2A/TNG tumor cells shed gangliosides similar to their respective ganglioside distribution (data not shown), we determined whether changes in tumor cell ganglioside distribution could influence the early stages of angiogenesis using the in vivo Matrigel angiogenesis model. The number and dilation of blood vessels was less in the Matrigel plugs containing CT-2A/TNG tumor cells than in plugs containing either control CT-2A or CT-2A/V tumor cells (Fig. 6). These findings indicate that the shift in ganglioside distribution to elevate GM3 and reduce GM1 and GD1a significantly decreases CT-2A brain tumor growth and angiogenesis.

Figure 3.  The influence of the shift in ganglioside distribution on intracerebral CT-2A tumor growth. Tumor wet weight was analyzed 14 days after tumor implantation for CT-2A (n = 10), CT-2A/V (n = 12) and CT-2A/TNG (n = 14) tumors. The values are expressed as means; the error bars represent ± SE. *Significant compared with control CT-2A and CT-2A/V tumors at the p < 0.01 level.

Figure 4.  The influence of the shift in ganglioside distribution on CT-2A tumor vascularity. (a) Factor VIII-immunostained microvessels (arrows) in tumor tissue (×100, original magnification). The scale bar represents 200 μm. (b) The quantification of blood vessel density. The blood vessels were determined per ×200 field (hpf, high-powered field) from tissue sections. The values are expressed as the means of three independent samples; the error bars respresent ± SE. *Significant compared with control CT-2A and CT-2A/V tumors at the p < 0.01 level.

Figure 5.  The influence of the shift in ganglioside distribution on vascular endothelial growth factor (VEGF) expression in CT-2A brain tumors. (a) VEGF mRNA expression by RT-PCR. VEGF (multiple splice variants: 400–600 bp) amplification products were detected in CT-2A (C), CT-2A/V (V), and CT-2A/TNG (T) tumors. PCR was performed as described in Fig. 2(a). (b) The ratio of total VEGF to β-actin. The values are expressed as the means of three independent samples; the error bars represent ± SE. *Significant compared with control CT-2A and CT-2A/V tumors at the p < 0.01 level. (c) VEGF protein expression by western blot analysis. VEGF protein was detected as a major band at ∼27 kDa. The conditions were as described in the ‘Experimental procedures’. (d) The ratio of VEGF to β-actin. The values are expressed as the means of either two (CT-2A) or three (CT-2A/V and CT-2A/TNG) independent samples; the error bars represent ± SE. *Significant compared wth control CT-2A and CT-2A/V cells at the p < 0.05 level.

Figure 6.  Tumor cell-induced vascularity in the in vivo Matrigel plug assay. The tumor cells were injected subcutaneously (s.c.) in severe-combined immunodeficient (SCID) mice as described in the ‘Experimental procedures’. The Matrigel plugs, with the surrounding skin, were removed on day 7 after implantation. The plugs were photographed at the same magnification (×12.5, original magnification). The experiments were performed in triplicate. The results show that the number and size of blood vessels was less in the plugs containing CT-2A/TNG tumor cells than in the plugs containing either control CT-2A or CT-2A/V tumor cells.

A shift in ganglioside distribution reduces tumor cell proliferation and the expression of VEGF, HIF-1α and NP-1

We examined whether the shift in ganglioside distribution influenced CT-2A tumor cell proliferation. We observed a significant reduction in tumor cell proliferation in the CT-2A/TNG tumor cells compared with the control CT-2A and CT-2A/V tumor cells (Fig. 7). After 6 days of growth, proliferation (assessed by absorbance at 595 nm) was 40% less in the CT-2A/TNG tumor cells than in the control CT-2A and CT-2A/V tumor cells. These data show that the gene-linked elevation of GM3, together with corresponding reductions of GM1 and GD1a, reduces the proliferation of CT-2A tumor cells.

Figure 7.  The influence of the shift in ganglioside distribution on CT-2A tumor cell proliferation. Proliferation was measured each day for 6 days as described the ‘Experimental procedures’. The values are expressed as the means of either two (CT-2A/V) or three (CT-2A and CT-2A/TNG) independent experiments; the error bars represent ± SE. *Significant compared with control CT-2A and CT-2A/V cells at the p < 0.01 level. , CT-2A; □, CT-2A/V; , CT-2A/TNG.

To investigate whether the shift in ganglioside distribution influenced the angiogenic properties of the CT-2A brain tumor cells, we measured VEGF, HIF-1α and NP-1 expression using RT-PCR. VEGF expression was significantly lower in the CT-2A/TNG tumor cells than in control CT-2A and CT-2A/V tumor cells (Fig. 8). As VEGF expression was lower in the CT-2A/TNG tumor cells than in control CT-2A and CT-2A/V tumor cells, we examined whether HIF-1, a transcription factor that can influence the expression of various angiogenic genes to include VEGF, was also affected. HIF-1α expression was significantly lower in the CT-2A/TNG tumor cells than in control CT-2A and CT-2A/V tumor cells (Fig. 8). Moreover, we examined whether shifting the ganglioside distribution in CT-2A tumor cells influenced the expression of receptors for VEGF such as NP-1, VEGFR-1 and VEGFR-2. NP-1 expression was significantly lower in the CT-2A/TNG tumor cells than in control CT-2A and CT-2A/V tumor cells (Fig. 8). Also, VEGFR-1 expression was not significantly different and VEGFR-2 expression was non-detectable in CT-2A, CT-2A/V and CT-2A/TNG tumor cells (data not shown). These findings show that the gene-linked elevation of GM3, together with corresponding reductions of GM1 and GD1a, reduces the angiogenic potential of CT-2A tumor cells.

Figure 8.  The influence of the shift in ganglioside distribution on vascular endothelial growth factor (VEGF), hypoxia-inducible factor 1α (HIF-1α) and neuropilin-1 (NP-1) expression in CT-2A tumor cells. (a) VEGF, HIF-1α and NP-1 mRNA expression by RT-PCR. VEGF (multiple splice variants: 400–600 bp), HIF-1α (365 bp) and NP-1 (551 bp) amplification products were detected in CT-2A (C), CT-2A/V (V) and CT-2A/TNG (T) tumor cells. PCR was performed as described in Fig. 2(a). (b) The ratio of gene to β-actin. The values are expressed as the means of three independent samples; the error bars represent ± SE. *Significant compared with control CT-2A and CT-2A/V cells at the p < 0.01 level.

Discussion

We showed that a gene-linked shift in ganglioside distribution significantly influenced growth and vascularity in the rapidly growing, highly vascularized malignant mouse astrocytoma, CT-2A. Specifically, the down-regulation of GalNAc-T expression in the CT-2A tumor cells caused a shift in the distribution of gangliosides synthesized through the ‘a’ metabolic pathway. This resulted in elevation of the ganglioside GM3 with concomitant reductions of the complex gangliosides GM1 and GD1a. Also, down-regulation of GalNAc-T expression affected the expression of other glycosyltransferases in the ‘a’ metabolic pathway. As the end product of a ganglioside pathway regulates the expression of biosynthetic enzymes within the same pathway, it is possible that the changes in ST1, Gal T2 and ST4 expression resulted from lower levels of the ‘a’ pathway end product, GD1a (Yusuf et al. 1987). The shift in ganglioside distribution significantly reduced growth, VEGF expression, and blood vessel number in the CT-2A brain tumor grown in vivo. The reduced growth and vascularity seen in vivo was also associated with reductions in expression of angiogenic biomarkers (VEGF, HIF-1α and NP-1) and in cell proliferation in vitro. As asialogangliosides are largely undetectable in the CT-2A tumor cells (Seyfried et al. 1996; Bai and Seyfried 1997), the observed effects of GalNAc-T down-regulation are likely to arise from the shift in distribution of ‘a’ metabolic pathway gangliosides. We do not, however, exclude the possibility that alterations in glycosyltransferase expression may influence other glycolipids and glycoproteins. Our data also support previous findings in other tumor models indicating that genetic manipulation of ganglioside biosynthesis, which shifts the relative distribution of GM3 and complex gangliosides, influences tumor growth and angiogenesis (Manfredi et al. 1999; Zeng et al. 2000).

Although the relationship between gangliosides, brain tumor growth and angiogenesis is complicated, our findings in the CT-2A brain tumor can provide new insight on this phenomenon. The gene-linked shift in ganglioside distribution may reduce CT-2A growth and vascularity through direct effects on angiogenesis, on tumor cell proliferation or through indirect effects on cell surface receptors that modulate both angiogenesis and proliferation.

As brain tumor growth is correlated with the degree of tumor vascularity (Leon et al. 1996; Chaudhry et al. 2001; Mukherjee et al. 2004; Kieran 2005), the shift in CT-2A ganglioside distribution may reduce growth and vascularity through effects on the angiogenic properties of the microenvironment. For example, shed gangliosides from tumors in the ECM can either enhance or suppress angiogenic responses through autocrine and paracrine effects on tumor cells and tumor-associated host cells (endothelial and macrophages) (Yohe et al. 1985; Ziche et al. 1989; Ziche et al. 1992; Koochekpour et al. 1996; Olshefski and Ladisch 1996; Alessandri et al. 1997; Lang et al. 2001). Our results with the CT-2A/TNG tumor cells support these findings as we found that these cells shed gangliosides and that the number and size of blood vessels was less in the in vivo Matrigel plugs containing CT-2A/TNG tumor cells than in those containing control CT-2A tumor cells. Previous studies in the rabbit cornea model of angiogenesis showed that GM3 inhibits endothelial cell migration and proliferation, whereas complex gangliosides (GT1b, GM1, GD3, etc.) counteracted the inhibitory effects of GM3 (Ziche et al. 1989, 1992; Alessandri et al. 1997). Although GD3 is considered the most pro-angiogenic ganglioside, CT-2A tumors do not express GD3. GD1a also stimulates endothelial cell proliferation and migration in vitro (Lang et al. 2001). We suggest that the gene-linked elevation of GM3 and the reduction of complex gangliosides, reduces the level of CT-2A angiogenesis in part through direct effects on the tumor microenvironment. We do not, however, exclude the possibility that the shift in ganglioside distribution may influence angiogenesis in the CT-2A brain tumor through other mechanisms to include integrin function and cellular immune responses (Yohe et al. 1985; McKallip et al. 1999; Wang et al. 2002).

In addition to the modulation of tumor–host interactions in the microenvironment, the shift in CT-2A ganglioside distribution could also influence angiogenesis through effects on the angiogenic properties of the CT-2A tumor cell itself. Support for this possibility comes from our findings that the expression of angiogenic biomarkers (VEGF, HIF-1α and NP-1) was lower in the CT-2A/TNG tumor cells than in the control CT-2A cells. VEGF, HIF-1α and NP-1 expression is associated with the function of cell surface receptors (Semenza 2000; Parikh et al. 2003; Lee et al. 2004; Bos et al. 2005; Kaur et al. 2005). Gangliosides are known modulators of cell surface receptor function and signaling (Zhou et al. 1994; Yates and Rampersaud 1998; Meuillet et al. 2000; Miljan and Bremer 2002; Liu et al. 2004). It is therefore possible that the shift in CT-2A ganglioside distribution indirectly influences VEGF, HIF-1α and NP-1 expression through the modulation of cell surface receptors. As VEGF expression is also regulated by HIF-1α, the overall reduction in VEGF expression in the CT-2A/TNG cells may result from both lower HIF-1α expression and modulation of receptor function. We suggest that the gene-linked elevation of GM3 and the reduction of complex gangliosides, reduces CT-2A growth and vascularity in part by decreasing the angiogenic potential of the CT-2A tumor cell.

Besides influencing angiogenesis, the shift in CT-2A ganglioside distribution may also influence growth and vascularity through effects on tumor cell proliferation. Support for this possibility comes from our findings that proliferation was less in the CT-2A/TNG tumor cells than in the control CT-2A cells. GM3 inhibits epidermal growth factor receptor (EGFR) signaling, whereas GD1a enhances EGFR signaling through effects on receptor phosphorylation and dimerization (Zhou et al. 1994; Liu et al. 2004). Moreover, treatment of human glioma cells with GM3 inhibited cell proliferation and induced apoptosis, suggesting a role of GM3 as a growth regulator (Noll et al. 2001). It is therefore possible that the shift in CT-2A ganglioside distribution influences tumor cell proliferation in part through the modulation of growth factor receptor function.

Finally, the shift in CT-2A ganglioside distribution may influence growth and vascularity through indirect effects on cell surface receptors that modulate both angiogenesis and proliferation. For example, gangliosides influence the EGFR signaling that regulates both cell proliferation and VEGF expression (Goldman et al. 1993; Maity et al. 2000; Meuillet et al. 2000; Miljan and Bremer 2002; Liu et al. 2004). It is therefore possible that the ganglioside shift in the CT-2A brain tumor simultaneously modulates cell proliferation and angiogenesis through effects on EGFR function. Further studies will be needed to confirm these interesting possibilities.

Taken together, our findings show that a shift in the distribution of ‘a’ metabolic pathway gangliosides in the CT-2A brain tumor significantly influences growth and vascularity, and also provide new insight into the possible mechanisms underlying these effects. Moreover, we suggest that the relative distribution of simple and complex gangliosides is important for tumor progression and malignancy. As gangliosides influence numerous processes involved in tumor progression (proliferation, invasion and angiogenesis), gangliosides are potential targets for the management of brain tumors and other types of cancer.

Acknowledgements

We thank Michael Kiebish, Allison Timmons, Elizabeth Adams and Tiernan Mulrooney for their technical assistance. We also thank Kevin Claffey for comments and discussion. This work was supported in part from NIH grants (HD39722) and (CA102135), a grant from the American Institute of Cancer Research, and the Boston College Expense Fund.