High-grade glial tumors (anaplastic astrocytoma and glioblastoma multiforme) are the most frequent brain tumors in adults.1, 2 Despite multimodal therapeutic efforts, the mean survival time is only 9–12 months for glioblastoma multiforme (WHO Grade IV) and 2 years for anaplastic astrocytoma (WHO Grade III). The ability of individual tumor cells to infiltrate brain tissue is the primary reason for this poor prognosis. Individual tumor cells (“guerrilla cells”) can migrate more than 4.0–7.0 cm from the gross tumor into the surrounding normal brain tissue.3
In part, the migratory and invasive behavior is a consequence of tumor cell interaction with specific components of the extracellular matrix (ECM). Brain ECM components include laminin, collagen IV, tenascin, fibronectin and proteoglycans.4 The glycosaminoglycan hyaluronic acid (HA) plays a central role in cellular proliferation, differentiation and migration, and may be a key element in tumor spreading.5 BEHAB (brain enriched hyaluronan binding)/brevican is a recently identified HA binding protein that belongs to the lectican family.6, 7, 8, 9 BEHAB/brevican expression is developmentally regulated in the central nervous system, in particular during periods of glial cell proliferation and migration.9 Accordingly, BEHAB/brevican is upregulated in the vast majority of surgical glioblastomas,9, 10, 11 and expression levels and proteolytic cleavage of BEHAB/brevican seem to correlate with tumor invasiveness in a rat glioblastoma model system.11, 12, 13
ADAMTS are a new family of ADAM (a disintegrin and metalloproteinase) related proteins that are characterized by a ADAM-like protease domain, a disintegrin-like and a cysteine-rich domain14 (compare MEROPS protease database [http://merops. sanger.ac.uk/]). A thrombospondin type 1 (TSP-1) repeat15 is found between the disintegrin-like and the cysteine-rich domains, followed by a varying number of TSP-1 repeats at the C-terminus. The TSP-1 motif is critical for substrate recognition and cleavage.16 ADAMTS lack a transmembrane domain and, therefore, are secreted into the ECM. Eighteen genetically different ADAMTS have been identified in human tissues. Functional analysis of ADAMTS has demonstrated their participation in a wide diversity of processes, e.g., ADAMTS1 (or METH-1) and ADAMTS8 (or METH-2) have angio inhibitory properties, ADAMTS2, ADAMTS3, and ADAMTS14 are procollagen N-proteinases, ADAMTS4 (= aggreganase-1) and ADAMTS5 (= aggreganase-2) are implicated in proteoglycan degradation and others are only structurally characterized.14, 15, 16, 17, 18 Because ADAMTS4 and 5 are effective ECM-degrading enzymes, they are of pivotal importance in tissue degradation/remodeling and cell infiltration.
According to recent studies, synthesis and cleavage of the BEHAB/brevican protein may play critical roles in the invasiveness of glioblastomas. The highly infiltrating rat CNS-1 glioblastoma cell line is able to produce and to cleave BEHAB/brevican by the action of ADAMTS4.19 In contrast to the findings on experimental rat glioblastoma cells, little is known about the occurrence of ADAMTS4 and 5 in human brain tumors, their regulation and their functional roles in brevican cleavage. We investigated their expression in glioblastoma cell lines, cells cultivated from solid human glioblastomas and different brain tumors in situ at the mRNA and protein level and their regulation under different inflammatory conditions. In addition, we localized these proteases in correlation to expression levels of BEHAB/brevican, the astroglial marker glial fibrillary acidic protein (GFAP), and proliferation status of glioblastoma cells by immunohistochemistry. We analyzed the cleavage activity of ADAMTS5 derived from glioblastoma cells on BEHAB/brevican.
ADAMTS, a disintegrin and metalloproteinase with thrombospondin motifs; CT, cycle of threshold; EGF, epidermal growth factor; GFAP, glial fibrillary acidic protein; Il-1β, interleukin-1β; RT-PCR, reverse transcriptase-polymerase chain reaction; TGF-β, transforming growth factor-β; TNF-α, tumor necrosis factor-α.
Material and methods
Recombinant human epidermal growth factor (EGF), interleukin-1β (Il-1β), transforming growth factor-β (TGF-β) and tumor necrosis factor-α (TNF-α) were obtained from Pepro Tech (Rocky Hill, NJ) and diluted from stock solutions (0.1 mg/ml) in bidistilled water into culture media with FCS.
Cell culture and tissue samples
Solid human brain tumors (classified as astrocytomas Grade III, glioblastomas WHO Grade IV, meningiomas and neurinomas) and normal brain tissue were surgical dissected tissues or autopsy material from the Departments of Neurosurgery and Legal Medicine (Kiel, Germany) and obtained in accordance with approved ethical standards of the responsible committee of the University of Kiel and with the Helsinki Declaration of 1975. The human glioblastoma cell lines U343, U373, U138 and U118 were obtained from Deutsches Krebsforschungszentrum (Heidelberg, Germany). Glioblastoma cells from solid tumors were obtained by dissociation and cultivation in DMEM plus 10% FCS as described previously.20 Subcultures from 3–10 were used. Purity of the culture was controlled routinely by immunostaining for the cell type specific markers glial fibrillary acidic protein (GFAP, astrocytes/glioblastoma cells; antibody from Boehringer, Mannheim Germany) and CD68 (contaminating microglial cells and macrophages).21 Contaminations by Mycoplasma were checked by staining with bisbenzimide (Merck, Darmstadt, Germany).
HEK293 cells stably transfected with brevican vector pRC-CMV/359H, eukaryotic selection marker G418/neomycin,7 were cultivated in DMEM with 10% FCS and 600 μg/ml G418/neomycin (Calbiochem, San Diego, CA). The human monocyte cell line THP-1, human dermal fibroblasts and human immortalized chondrocytes were obtained and cultivated as described.22, 23
RNA was isolated with the TRIZOL reagent (Invitrogen, Carlsbad, CA), digested by DNase (65°C for 10 min; Promega, Madison, WI), and cDNA synthesized with RevertAid™ H Minus M-muLV Reverse Transcriptase (Fermentas, Hanover, MD). RT-PCR was carried out in 3 replicates of each sample using an ABI 7700 Prism Sequence Detection System and TaqMan primer probes (assays on demand; Applied Biosystems, Foster City, CA) using a total reactive volume of 20 μl, that contained 1 μl of 20× Target Assay Mix, 10 μl of 2× TaqMan Universal Master Mix and 100 or 10 ng of cDNA template (diluted in RNase-free water to 9 μl). After 2 min at 50°C for UNG incubation and 10 min at 95°C for polymerase activation, 40 cycles of 15 sec at 95°C (denaturation) and 1 min at 60°C (annealing and extension) were run. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) in each sample was tested concurrently as intrinsic positive control and used as a normalizer. Each plate included at least 3 no-template controls (NTC). Cycle of threshold (CT) of each tumor sample was averaged and normalized to GAPDH according to tumor types yielding normalized CT-values. For each gene, logarithmic linear dependence of CT-values from the numbers of copies was verified by using different amounts of cDNA; one magnitude yielded a ΔCT of 3.33 (23.33 = 10). Relative expression values were calculated with 2(normalized CT non-stimulated − normalized CT stimulated) × 100 = % of control.
Northern blot analysis
RNA was separated by electrophoresis in 1% agarose gels (5–10 μg per lane) and blotted onto Hybond nylon membranes (Amersham Pharmacia Biotech, Buckinghamshire, UK) by capillary transfer. After purification of ADAMTS4, ADAMTS5 and GAPDH PCR products (449 bp/362 bp/983 bp) from 1% agarose gels with the NUCLEOTRAP® extraction kit (Macherey-Nagel, Düren, Germany), the products were labeled with 32P-dCTP using the Megaprime™ DNA labeling system and purified with Superdex G-25 MicroSpin™ columns (Amersham Pharmacia). After pre-hybridization at 63°C with ExpressHyp Hybridization Solution (Clontech, Palo Alto, CA) hybridization was carried out overnight at 63°C in the same solution with 106 cpm/ml denaturated probe. Blots were washed with 2× SSC and 0.1% SDS (2 × 10 min at room temperature) and 2 × 20 min at 63°C with 0.1× SSC and 0.1% SDS and exposed overnight on a Kodak BIOMAX MS Scientific Imaging Film (Eastman Kodak Company, Rochester, NY).
Stimulation of glioblastoma cells
U343 (or U118) cells (2 × 106) were grown for 3 days in 10% serum-supplemented DMEM, washed in 37°C-thermostatted 0.5% serum-supplemented DMEM (3 × 20 min), and stimulated in the same medium with interleukin-1β (IL-1β, 10 ng/ml), transforming growth factor-β (TGF-β, 10 ng/ml), tumor necrosis factor-α (TNF-α, 10 ng/ml), or epidermal growth factor (EGF, 10 ng/ml). After incubation at 37°C for 6 to 48 hr, medium was withdrawn, cells were lysed immediately and RNA was isolated as described below.
Tissue homogenates (aliquots with equal protein contents) were boiled and reduced in 50–200 μl SDS-PAGE sample buffer, proteins separated in NuPAGE pre-cast gel (NP0321, NuPAGE 4–12% Bis-Tris Gel, Invitrogen) or by 10% standard SDS-PAGE, transferred onto a polyvinylidene difluoride (PVDF) membranes that were blocked and incubated with anti-ADAMTS4 or 5 (C-terminal domains, rabbit polyclonal antibodies 1:250; Chemicon, Temecula, CA) or anti-brevican7 (1:5,000) followed (after washings) by horseradish peroxidase-labeled anti-rabbit IgG (1:30,000; DAKO, Glostrup, Denmark). Bound antibody was visualized by enhanced chemiluminescence (ECL system; Amersham). For semiquantitative analysis, same samples were separated by another gel-electrophoresis and stained with Coomassie blue for total protein amount.
Cryostat tissue sections were incubated with primary antibodies at 4°C overnight in the following dilutions: guinea pig or rabbit anti-brevican7 (both 1:500), rabbit anti-ADAMTS5 (RP1ADAMTS-5, Triple Point Biologics, Portland, OR; 1:100), monoclonal anti-MIB-1 (DAKO, 1:100), guinea pig anti-GFAP (Advanced Immunochemicals, 1:250). After blocking and washing,25 sections were incubated with corresponding fluorescent secondary antibodies (Cy3, Cy5, FITC; Jackson, WY; 1:250), and nuclei stained with by diamidino-2-phenylindole (Molecular Probes, USA, 1:10000). They were analyzed by confocal scanning laser microscopy (Zeiss, Germany), signals collected with LSM software, digitally color enhanced, and superimposed with Photoshop 6.0 (Adobe Photoshop).
Analysis of brevican degradation
For analyses of the catalytic activity of tumor cell-derived ADAMTS5, the protease was immunoprecipitated and incubated with recombinant proteoglycan. Brevican was enriched from cell supernatants (80 ml) of stably transfected HEK2293 cells (1 × 106 cells) by concentration with an YM-10 ultrafiltration (Amicon) membrane to 4 ml, followed by PBS washing steps (repeated ultrafiltration). For immunoprecipitation of ADAMTS5, 2 × 106 U343 tumor cells grown for 3 days in 10% serum-supplemented DMEM were incubated overnight in 0.5% serum-supplemented DMEM. Conditioned medium was withdrawn; cells were lysed with ice-cold 0.14 M NaCl in 10 mM HEPES 1% Triton X-100 buffer (pH 7.4) and scraped off by a rubber policeman. Lysate and medium were incubated with anti-ADAMTS5 (RP1ADAMTS-5, rabbit polyclonal, 1:1,000 for medium, 1:500 for lysate; Triple Point Biologics) overnight at 4°C. Protein A-Sepharose (Santa Cruz, CA) was added and samples were incubated for 2 hr under gentle shaking at room temperature. The beads were separated by centrifugation (5 min at 10,000g), washed 5× with 0.14 M NaCl 10 mM HEPES buffer, pH 7.4, and aliquots incubated in this buffer (50 μl) with 50 μl of brevican at 37°C under gentle shaking. At 0 hr, 1 hr and 24 hr, 20 μl aliquots were withdrawn, boiled and reduced, separated by SDS-PAGE and Western blot obtained as above that were probed with anti-brevican (guinea pig or rabbit polyclonal antibodies, 1:20).7
ADAMTS4 and 5 mRNA expressions in solid and cultivated glioblastomas
In an initial step we investigated the quantitative expression of mRNA for ADAMTS4 and 5 by real time RT-PCR (Fig. 1) in cultivated cells as well as in surgical samples from astrocytic tumors (that contain also other cell types like endothelial cells and microglial cells/macrophages). Cultivated brain tumor cells included glioblastoma cell lines (Fig. 1, gcl), astrocytoma cells derived from solid glioblastomas (cg, WHO Grade IV) or WHO Grade III astrocytomas (ca) and tumor cells from meningioma (cm) and neurinoma (cn). Cultivated human chondrocytes and dermal fibroblasts served as positive expression controls, samples of normal brain and the human monocyte cell line THP-1 for comparison. ADAMTS5 was detected in all samples, also ADAMTS4 except in THP-1 monocytes. The expression of mRNA for ADAMTS5 in glioma cells/cell lines was additionally sustained by Northern blots (Fig. 2). A single mRNA band of 5.5 kb was detected for ADAMTS5. Quantitative densitometric analysis showed similar GAPDH-normalized values as determined with the more accurate real time RT-PCR (not shown).
Surgical sections from solid glioblastomas (Fig. 1, sg) clearly showed the highest expression of both, ADAMTS4 and 5. Expression was higher than in chondrocytes or fibroblasts, a well known source for both proteases (“aggrecanases”). Normal brain tissue showed a lower, but somewhat variable expression for both proteases as compared to solid glioblastomas. Malignant cells (about 98% pure) derived from solid tumors and glioblastoma cell lines showed a considerable expression of both proteases, similar or partly higher than as chondrocytes or fibroblasts. Cultivated tumor cells exhibited a similar (ADAMTS4) or even a bit lower expression (ADAMTS5) as compared to homogenates of normal brain tissue. However, glioblastoma cell lines, cultivated meningioma and neurinoma were particularly rich in ADAMTS5. Cells derived from a WHO Grade III astrocytoma did not differ from the Grade IV glioblastoma cells.
To elucidate possible reasons for a higher expression of the proteases in solid glioblastomas as compared to cultivated glioblastoma cells, we investigated the influence of cytokines and epidermal growth factor (EGF) on their expression in the U343 glioblastoma cell line by quantitative RT-PCR experiments (Fig. 3). Stimulation with transforming growth factor-β (TGF-β) or interleukin-1β (Il-1β) resulted in about 2- to 2.5-fold higher expression. The effects of tumor necrosis factor-α (TNF-α) and EGF were lower. The effects differed between cytokine and protease. The highest and most robust effect was seen for Il-1β stimulation of ADAMTS5 that did not cease with time in contrast to other stimulators. ADAMTS4 was induced strongest by TGF-β after 6 hr. Il-1β showed only little effect. Similar results were obtained with another glioblastoma cell line (U118 cells after 6 hr stimulation, not shown).
In conclusion, glioblastoma cells and especially solid glioblastomas express mRNA for ADAMTS4 and 5. In stimulation experiments in vitro, their expression was differentially upregulated by cytokines, ADAMTS4 transiently by TGF-β, and ADAMTS5 by Il-1β.
ADAMTS4 and 5 proteins in normal and glioblastoma tissues
Protein structures of ADAMTS4 and 5 contain a pre- and a pro-sequence, both cleaved off during maturation before secretion. The mature, secreted protein is glycosylated and further processed proteolytically into active and inactive fragments.26 Accordingly, Western blots with C-terminally directed antibodies showed differently processed fragments of ADAMTS4 and 5 in homogenates of surgical glioblastoma and normal brain tissue (2 individual samples each, Fig. 4). The specificity of the immunoreaction was proved by omitting the primary antibodies (not shown).
For ADAMTS4 a 75-kDa band and several further bands (most prominent 53 and 47 kDa) were identified. The 75 kDa corresponds to the glycosylated active form found previously in transfected HEK293 cells26 where further degradation products occur. The pro-form (about 100 kDa) was not detectable in tissue homogenates. ADAMTS4 staining was intense in glioblastomas whereas normal brain yielded only weak bands (equal quantities of protein applied). Staining of ADAMTS5 yielded bands of the pro-form (about 105 kDa), the active form (75 kDa) and fragments at 60 and 50 kDa. Staining was less intense in homogenates from glioblastomas than in those from normal brain, but the smaller fragments could be well detected. Thus, ADAMTS4 and 5 proteins are found in differentially processed forms in situ.
In the next step, we analyzed the expression of ADAMTS5 protein in situ by confocal laser microscopy using different combinations of antibodies against the protease, the extracellular matrix component brevican, the proliferation marker MIB-1, the glial marker GFAP and a nuclear counterstain (Fig. 5). In accordance to the real time RT-PCR experiments, in glioblastomas ADAMTS5 was detected throughout the surgical samples; stronger staining was detected in the vicinity of proliferating glioblastoma cells (as stained with the proliferation marker MIB-1, Fig. 5top b–d). The possible substrate, brevican, was similarly distributed throughout the samples, and brevican was also co-localized with MIB-1-positive tumor regions (Fig. 5top g–i). Furthermore, brevican was distributed in ADAMTS5-positive regions (Fig. 5top a,b). Overall, ADAMTS5 could be co-stained in GFAP-positive, proliferating (MIB-positive) glioma cells (Fig. 5top k–n). These immunostainings show that ADAMTS5 and brevican are widely distributed in glioblastomas in situ.
In sections of normal brain tissue, brevican and ADAMTS5 were distributed evenly throughout the samples, but ADAMTS5 staining was clearly lower level compared to glioblastoma samples (Fig. 5bottom). Overlays showed that ADAMTS5 was localized mainly in GFAP-positive astroglial cells (Fig. 5bottom i) and ADAMTS5-positive cells were spotted in the brevican extracellular matrix (Fig. 5bottom d). Proliferating cells (MIB-1 positive, Fig. 5bottom c,h) were nearly undetectable in normal brain sections. Thus, ADAMTS5 is also expressed in normal astrocytes in situ, but with considerably lower intensity.
ADAMTS5 cleaves brevican in vitro
To elucidate whether brevican can be cleaved by human ADAMTS5, we incubated immunoprecipitates of lysates and culture supernatants obtained from U343 glioblastoma cells with recombinant brevican and monitored the time-dependent degradation by Western blots with antibodies to brevican (Fig. 6a). In particular, the ADAMTS5 immunoprecipitates from culture supernatants degraded the 160 kDa band of the proteoglycan to smaller fragments of 100 kDa and 90 kDa. A corresponding, but far less intense degradation of brevican was observed with immunoprecipitates from cell lysates indicating that the main pool of active ADAMTS5 is secreted and some traces are adhered to the cells. This experiment proves that active ADAMTS5 is secreted by glioblastoma cells and can cleave brevican.
Western blots of tissue homogenates from normal brain and glioblastomas for brevican yielded the 160 kDa brevican band, the 90 kDa fragment and several smaller fragments (Fig. 6b). Staining in gliomas, especially for intact brevican, was far more intense than in the normal brain samples. The observed fragments corresponded partially, but not totally to those observed in the in vitro experiment.
The tendency to infiltrate the nervous system is a characteristic feature of high grade glial tumors, and morbidity and mortality are directly related to their ability to invade and infiltrate surrounding tissue.1 Migration and invasion of tumor cells is considerably facilitated (but not entirely dependent) by the expression of proteases digesting the extracellular matrix.27 For glioblastomas, matrix metalloproteinases (MMP) and few serine proteases have drawn most attention.28 Secreted MMP-2 (gelatinase A), MMP-9 (gelatinase B), urokinase and other secreted proteases, as well as cell-surface proteases involved in their activation like MT1-MMP or MT2-MMP (membrane type matrix metalloproteinases-1 and -2) seem to play a critical role in this migratory process.28, 29, 30, 31, 32, 33 Beside the proteases identified so far, others may be of additional or even similar importance for the invasive process.
We investigated the expression of the secreted proteases ADAMTS4 and 5 in human brain tumors, in particular in glioblastomas. Furthermore, their regulation and the functional role of ADAMTS5 were followed in detail. We showed that surgical glioblastomas showed comparably high expression of both proteases. This was generally lower in cultivated glioblastoma cells. In situ, ADAMTS5 was confined to glioblastoma and astroglial cells by immunohistochemical double-labeling; therefore the difference is not due to the production by other (normal) cells in solid tumors. Alternatively, this divergence may result from their up-regulation by cytokines and growth factors that are produced either by the tumor cells or by microglial cells that contribute up to 30% of all cells in glioblastomas.24
In our studies with cultivated glioma cells, transcription of ADAMTS4, but not ADAMTS5 was upregulated by TGF-β whereas ADAMTS5 was induced by IL-1β. EGF and TNF-α had almost no effects. In fibroblast-like synoviocytes ADAMTS4, but not ADAMTS5, was about 10-fold induced by TGF-β, whereas Il-1β and TNF-α induced ADAMTS5 about 2-fold.34 TNF-α induced ADAMTS5, especially in a pigment epithelium-derived cell line.35 In chondrocytes, ADAMTS4 was induced by Il-1β.36 ADAMTS4 and 5 inductions in glioblastoma cells is partly similar, but not completely identical as in mesenchymal cells. It is well known that TGF-β is a major cytokine produced by high grade gliomas and is, therefore, an example for the regulation of these proteases in the tumor microenvironment.
Functionally, we could show that glioblastoma cell-derived ADAMTS5 was able to degrade the ECM proteoglycan brevican to smaller fragments. In previous studies it has been shown that recombinant human ADAMTS5 degrades aggrecan at mostly identical sites than recombinant ADAMTS4.37 Since, ADAMTS4 is able to cleave brevican at the Glu(395)-Ser(396) bond,32, 38 it can be anticipated that ADAMTS5 hydrolyses brevican at least at the same site. Brevican degradation may also be accomplished by other proteases. MMP-1, -2, -3, -7, -8, -10 and -13 have been shown to degrade brevican but at different bonds than ADAMTS4.32 We show the constitutive expression of ADAMTS5 in normal astrocytes and potential activities of other proteases may be responsible for dominant presence of brevican fragments in humans (as compared to rodents). It should be noted, however, that ADAMTS4/5 expressions are considerably higher in malignant than normal brain tissue.
Previous studies on brevican cleavage and its importance for glioblastoma invasion11, 13, 19 were only carried out in the rat system where brevican is particularly abundant.7 We show that in human glioblastomas and, to a lesser extent, in normal human brain brevican as well as the proteases responsible for its degradation are produced. This explains that part of the extracellular brevican is degraded, the lower amounts in normal brain even more. In humans, compared to the rat system, not only ADAMTS4, but also ADAMTS5 is expressed and responsible for brevican cleavage. Both proteases also occur in the normal adult brain but seem to be considerably upregulated in glioblastomas in situ by their stromal environment.
In conclusion, the secreted proteases ADAMTS4 and 5 are upregulated in brain tumors, especially glioblastomas in situ. ADAMTS5 is able to degrade the extracellular matrix proteoglycan brevican and therefore contributes to the invasive potential of glioblastoma cells.
We thank M. Burmester, D. Freier, M. Lemmer, B. Rehmke and J. Krause for expert technical assistance and C. Franke for drawing figures.