The FASEB Journal express article 10.1096/fj.01-0632fje. Published online February 12, 2002.
NG2 proteoglycan promotes angiogenesis-dependent tumor growth in the central nervous system by sequestering angiostatin Martha Chekenya,* Mari Hjelstuen,† Per Øyvind Enger,* Frits Thorsen,* Anne L. Jacob,* Beatrice Probst,* Olav Haraldseth,‡ Geoffrey Pilkington,§ Arthur Butt,¶ Joel M. Levine,** and Rolf Bjerkvig* *Department of Anatomy and Cell Biology, University of Bergen, N-5009 Bergen, Norway; † SINTEF Unimed MR Center, N-7465 Trondheim, Norway; ‡Department of Anesthesia and Medical Imaging, The Norwegian University of Science and Technology, Trondheim, Norway; § Experimental Neuro-Oncology, Department of Neuropathology, Institute of Psychiatry, King’s College, London SE5 8AF, UK; ¶Center for Neuroscience, Guys’, King’s and St. Thomas School of Biomedical Sciences, King’s College, London SE1 1UL, UK; and **Department of Neurobiology and Behavior, State University of New York, Stony Brook, New York 11794, USA Corresponding author: Professor Rolf Bjerkvig, Department of Anatomy and Cell Biology, University of Bergen, Årstadveien 19, N-5009 Bergen, Norway. E-mail:
[email protected] ABSTRACT During embryogenesis, the NG2 proteoglycan is expressed on immature capillary vessels, but as the vessels mature they lose this expression. NG2 is up-regulated in high-grade gliomas, but it is not clear to what extent it contributes to malignant progression. Using a combination of high spatial and temporal resolution functional magnetic resonance imaging and histopathological analyses, we show here that overexpression of NG2 increases tumor initiation and growth rates, neovascularization, and cellular proliferation, which predisposes to a poorer survival outcome. By confocal microscopy and cDNA gene array expression profiles, we also show that NG2 tumors express lower levels of hypoxia inducible factor-1α, vascular endothelial growth factor, and endogenous angiostatin in vivo compared with wild-type tumors. Moreover, we demonstrate that NG2-positive cells bind, internalize, and coimmunoprecipitate with angiostatin. These results indicate a unique role for NG2 in regulating the transition from small, poorly vascularized tumors to large, highly vascular gliomas in situ by sequestering angiostatin. Key words: angiogenesis • glioblastoma • pericytes
T
he periventricular region of the lateral ventricle in the mammalian central nervous system (CNS) is home to an extensive network of totipotent neuroepithelial stem cells capable of differentiating into neurons and both types of glial cells (1-4). Outside the CNS, these stem cells can give rise to ventral mesodermal cells such as cardiomyocytes, smooth muscle cells, and pericytes (5, 6, 6a). Lineage restriction of the stem cells gives rise to oligodendrocyte precursor cells that are reliably identified by expression of the membrane-spanning chondroitin sulfate proteoglycan, neuron-glia 2 (NG2) (7, 8).
Temporal and spatial distribution studies of NG2 expression during embryogenesis indicated that the earliest appearance was at embryonic day 12 (E12) on immature brain capillaries (5). NG2 expression continued throughout the period of rapid expansion of the brain vasculature and was subsequently down-regulated as the vessels terminally differentiated (5). In the adult CNS, oligodendroglial precursor cells also express NG2 (9, 10). Immunohistochemical and electron microscopic studies have suggested that expression of NG2 on neovasculature is restricted to vascular pericytes (6, 6a, 11, 12). However, the expression of NG2 by endothelial cells has also been reported in both normal and malignant brain tissue (5, 13, 14). In addition, NG2 is upregulated in the adult brain during neovascularization in granulating tissue of healing wounds (11, 15), which supports a role for NG2 during vascular proliferation and morphogenesis. Several highly vascular tumor types including glioblastoma multiforme (GBM), acute myeloid leukemias, and melanomas express NG2, both on the tumor cell surface and on the associated vasculature (11, 13, 16-18). Another prominent component of the vasculature is angiostatin, a proteolytic fragment of plasminogen that is thought to have important regulatory functions during both normal and pathological angiogenesis by inhibiting endothelial cell proliferation and migration (19). Evidence is emerging that NG2 binds specifically and saturably to angiostatin, resulting in the stimulation of endothelial cell proliferation and migration (20). METHODS Cell culture, transfection, and selection of NG2-expressing cells The previously described NG2 cDNA (21) was subcloned into the pcDNA 3.1 vector (Invitrogen, Leek, The Netherlands) and used to transfect the human GBM cell line U251N (American Type Culture Collection, Rockville, MD). The cells were propagated in Dulbecco’s modified Eagle’s medium (GIBCO Life Technologies, Paisley, Scotland), supplemented with 10% heat-inactivated newborn calf serum, four times the prescribed concentration of essential amino acids, 2% glutamine, penicillin (100 IU/ml), and streptomycin (100 µg/ml) (called complete growth medium). This was supplemented with geneticin disulfate (800 µg/ml) (Sigma, Dorset, UK) to select for neomycin-resistant clones. The sham control cells were transfected with the pcDNA 3.1 vector alone. Proliferation assays Tumor cells were seeded at a density of 2 × 104 cells/35-mm well (Nunc, Roskilde, Denmark) in growth medium and were counted in triplicate at 24, 48, 72, and 96 h using a hemocytometer. Immunoprecipitation and Western blot analysis Both 80% confluent wild-type (WT) and NG2-transfected cells were incubated at 37°C for 16 h with 5 µg/ml human angiostatin (Angiogenesis Research Industries, Chicago, IL) in serumreduced Optimem growth medium (Sigma). Control cells were not treated with angiostatin. After cells were washed with PBS, they were lysed in immunoprecipitation buffer (IP, 50 mM Tris, pH 8.0, 150 mM NaCl, 1% NP40, 1% phenylmethylsulfonyl fluoride, and 0.02 U of chondroitinase ABC (Sigma) for 30 min at 4°C. The lysate was centrifuged at 13,000g for 15 min at 4°C. The supernatant was incubated with 5 µg/ml monoclonal antibodies against angiostatin (American
Diagnostica, Greenwich, CT), after which the antibodies were captured by adding 100 µl of a 50% slurry of protein A-Sepharose beads (Pharmacia, Peapack, NJ). Immunopurified protein was solubilized with sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer and resolved on 4-15% gradient polyacrylamide gels (Bio-Rad, Burlington, MA). Western blot analyses of NG2 and angiostatin were carried out with antibodies to rabbit anti-rat NG2 (diluted 1:1000 in PBS) and mouse anti-human angiostatin (diluted 1:200 in PBS). Immunoreactive protein bands were visualized by using the enhanced chemiluminescence method, as described by the manufacturer (Pierce, Rockford, IL). GBM spheroid cultures We initiated cells into spheroids by using the method described previously (22). Spheroids 250 µm in diameter were selected for intracranial implantation. Animals Forty-one nude rats (Han:rnu/rnu Rowett) of both sexes were used in the experiments. All animal experiments were performed according to protocols approved by The National Animal Research Authority (Oslo, Norway). Intracranial implantation of spheroids We anesthetized the animals before surgical procedures with subcutaneous injections of fentanyl/ fluanisone/midazolam at a dose of 0.1-0.15 ml/100 g of body weight. The rats were immobilized in a sterotactic frame (David Kopf Instruments model 900, Tujunga, CA), and a burr hole was located 1.5 mm to the right of the sagittal suture and 0.5 mm posterior to the bregma. We injected 7 µl of PBS containing 10 spheroids into the forebrain to a depth of 1.5 mm from the brain surface. We killed the animals when neurological signs of passivity, clumsiness, and paresis were evident. At this point, by using ventricular puncture with a syringe, we sampled cerebrospinal fluid (CSF) for Western blot analysis of circulating angiostatin. The brains were embedded in Tissue-Tek O.C.T. (Miles Laboratories, Naperville, IL) and frozen in liquid N2. Serial frozen sections of 60 µm were cut on a Reichert Jung Cryomicrotome (Leica, Wetzlar, Germany). Tissue for histopathological examination was stained with Harris hematoxylin and eosin (H&E) (Merck, Darmstadt, Germany) and examined with a Nikon Diaphot light microscope. Images were captured with a Nikon CoolPix 990 digital camera (Nikon, Tokyo, Japan). The cell nuclei and mitotic figures from 10 random high-power fields in a section were quantified by using the Lab-count Denominator (Denominator, Woodbury, CT). Magnetic resonance imaging (MRI) : sequences and image analysis We performed MRI examinations at 2.35 T on a Bruker Biospec (Bruker Medisinteknik, Ettlingen, Germany), using a specially designed rat head coil, with an inner diameter of 4.5 cm. The rats were continuously anesthetized through a mask with 1-2% isoflurane in 70%/30% N2/O2. We inserted a polyethylene catheter into the right femoral vein for intravenous (i.v.) injection of the contrast agent gadodiamide (Gd) (Omniscan, Nycomed Imaging AS, Oslo, Norway). Tumor blood volume and vascular permeability
These measures were assessed by using a dynamic contrast-enhanced T1-weighted gradient echo sequence (TR/TE 91/5 ms, field of view [FOV] 35 mm, acquisition matrix 64 × 64, and 2 NEX). We administered an i.v. injection of Gd at 0.1 mmol/kg before recording the baseline images. Thereafter, we obtained 100 images from each transaxial slice with a temporal resolution of 12 s and selected a region of interest covering the viable tumor to determine the mean enhancement time curve. This curve was converted into a concentration-time curve by using the Gd quantification calibration curve (23). We then applied a modified two-compartment pharmacokinetic model and a nonlinear least squares fitting technique to derive the fitting parameters, the vascular volume fraction (D0) and the vascular permeability (K2) (24). Tumor volume After the dynamic contrast-enhanced imaging, another injection of Gd at 0.3 mmol/kg was administered. A T1-weighted spin echo sequence (TR/TE 524/13 ms, FOV 35 mm, acquisition matrix 256 × 256, and 4 NEX) was used to determine the tumor volume with high spatial resolution. The tumor volume in each slice was calculated as the number of pixels within a region of interest determined a priori as signal hyperintensity multiplied by the voxel size of 0.0224 mm3 (corrected for interslice gap). Immunohistopathology We performed indirect immunolocalization by using standard procedures (25). Primary antibodies used to label acetone-fixed or 4% paraformaldehyde-fixed monolayers or frozen sections were mouse anti-rat CD31 (diluted 1:10 in PBS; PharMingen, Franklin Lakes, NJ), rabbit anti-human von Willebrand factor (vWF) (diluted 1:100 in PBS; Dako, Glostrup, Denmark), mouse anti-human Ki67 (diluted 1:20 in PBS; Clone MIB-1, Dako), rabbit anti-rat NG2 (diluted 1:1000 in PBS; J. M. Levine), mouse anti-human α-smooth muscle actin (α-SMA) (diluted 1:25 in PBS; Dako), mouse anti-desmin (diluted 1:50 in PBS; Dako), mouse anti-human hypoxia-inducible factor 1α (HIF-1α) (diluted 1:10 in PBS; Transduction Laboratories, Lexington, KY), mouse anti-vascular endothelial growth factor (VEGF) (diluted 1:50 in PBS; C1, Santa Cruz Biotechnology, Santa Cruz, CA), and mouse anti-human angiostatin (diluted 1:200 in PBS; American Diagnostica). Fluorescein isothiocyanate (FITC)- or Cy3-conjugated antirabbit or anti-mouse antibodies (diluted 1:30 in PBS; Zymed, San Francisco, CA) were used to visualize antibody binding. A Leica TCS NT confocal laser-scanning microscope (Leica Lasertechnik, Heidelberg, Germany) was used to examine the tumor specimens. Frozen sections were stained by using the EnVision+ System, horseradish peroxidase, and 3',3'-diaminobenzidine (Dako) method, as described in the manufacturer’s protocol. Harris hematoxylin (Merck) was used for nuclear counterstaining. The volume (mm3) of vWF-positive, CD31-positive microvessel densities (MVDs), and Ki67 nuclear proliferation antigen-positive cells from four random fields in a section was quantified by using the AVS Express software, version 3.4 (Application Visualization System, Waltham, MA). RNA extraction and cDNA gene array procedures Tumor tissue was homogenized in Trizol LS Reagent (Gibco), and the total RNA was isolated according to the standard Trizol protocol. RNA integrity was judged by denaturing formaldehyde agarose gel electrophoresis. cDNA gene array analysis was done as recommended by the manufacturer’s GEarray protocol, available at http://www.superarray.com. Briefly, 10 µg of undegraded RNA was reverse transcribed with MMLV reverse transcriptase (50 U/µl) in the
presence of [α-32P]dCTP to produce 32P-labeled cDNA probes. After stringency washes, membranes were exposed for 16 h in a Phosphor Imager and thereafter were scanned to produce digital images (Fujifilm BAS-5000, Bundoora, Australia). The intensity of each duplicated signal was visualized and quantified by phosphor-imaging computer software. For relative “fold” estimation, the ratio between the signals obtained from VEGF was compared in both samples and standardized to the housekeeping β-actin gene. Statistical analyses Data were analyzed by linear regression and Student’s t test (two tailed) (Graphpad Instat Software version 3.05, San Diego, CA). The Kaplan-Meier method (26) and the log-rank test (27) were used to generate and analyze survival curves. RESULTS NG2 increases glioblastoma cell proliferation in vitro To investigate the effect of NG2 on the proliferation of GBM cells, we compared the growth of three NG2-transfected U251N clones against the WT and sham negative controls. Indirect immunocytochemistry (Fig. 1A, B) and Western blotting (Fig. 1C) were used to detect the expression of NG2 in the tumor sublines. All three NG2 clones showed significantly enhanced growth rates compared with both WT (P
