Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Identification of genes differentially expressed in glioblastoma versus pilocytic astrocytoma using Suppression Subtractive Hybridization


Glioblastoma (GBM) is a highly malignant glioma, which has the propensity to infiltrate throughout the brain in contrast to pilocytic astrocytoma (PA) of the posterior fossa, which does not spread and can be cured by surgery. We have used Suppression Subtractive Hybridization to define markers that better delineate the molecular basis of brain invasion and distinguish these tumor groups. We have identified 106 genes expressed in PA versus GBM and 80 genes expressed in GBM versus PA. Subsequent analysis identified a subset of 20 transcripts showing a common differential expression pattern for the two groups. GBM differs from PA by the expression of five genes involved in invasion and angiogenesis: fibronectin, osteopontin, chitinase-3-like-1 (YKL-40), keratoepithelin and fibromodulin. PA differs from GBM by the expression of genes related to metabolism (apolipoprotein D), proteolysis (protease-serine-11), receptor and signal transduction (PLEKHB1 for Pleckstrin-Homology-domain-containing-protein-family-B-member-1), transcription/translation (eukaryotic-translation-elongation-factor-1-α1) processes and cell adhesion (SPOCK1 for SPARC/Osteonectin-CWCV-kazal-like-domains-proteoglycan). The expression of these genes was confirmed by real-time quantitative RT-PCR and immunohistochemistry. This study highlights the crucial role of brain invasion in GBM and identifies specific molecules involved in this process. In addition, it offers a restricted list of markers that accurately distinguish PA from GBM.


Gliomas are a heterogeneous group of intracranial neoplasias of glial origin. They are the most common type of primary tumor (40%) in the central nervous system. Different subtypes can be distinguished, among them pilocytic astrocytomas (PA), oligodendrogliomas, oligoastrocytomas and glioblastomas (GBM) (Kleihues and Cavenee, 2000).

GBM (WHO grade IV) is the most frequent and malignant primary neoplasm of the human central nervous system. GBM have the propensity to infiltrate throughout the brain. They carry complex genetic aberrations resulting in the inactivation of various tumor suppressor genes, as well as the amplification of different protooncogenes (Kleihues and Cavenee, 2000). Despite progress in surgery and adjuvant therapy, patients with GBM have a very poor prognosis. In contrast, PA (WHO grade I), the most common glioma in children, has a favorable prognosis. Its slow and circumscribed growth pattern with almost no infiltrative behavior enables it to be cured, in most cases, by total surgical excision.

The heterogeneity of astrocytic tumors has made their pathological classification rather difficult. Although several critical genes have been identified as being associated with tumorigenesis and anaplastic progression of gliomas (Caskey et al., 2000), their contribution to the molecular classification of astrocytic tumors has been limited.

New genomic technologies such as DNA microarrays may facilitate the molecular profiling of human tumors. Some studies have used microarrays to compare the expression pattern of high-grade tumor with low-grade tumor or nontumoral brain tissue (Sallinen et al., 2000; Rickman et al., 2001; Gutmann et al., 2002; Khatua et al., 2003; van den Boom et al., 2003; Mischel et al., 2004; Oncomine database). They identified cell cycle and growth regulators, invasion and growth factors and cytokines that probably play a role in astrocytoma progression.

We chose to compare GBM and PA in order to find a restricted list of genes involved in tumor brain invasion. Owing to their putative common origin (astrocytic cells or glial precursor cells), but their radically different behavior, we found more judicious to compare these two types of gliomas rather than comparing GBM with normal brain. We chose to use Suppression Subtractive Hybridization (SSH) (Diatchenko et al., 1996) to identify genes expressed differentially between these two types of tumors. We were able to identify a number of mostly novel potential candidate genes. Our data indicate that GBM differ from PA by the expression of subset of genes involved in invasion and angiogenesis, whereas PA are characterized by the expression of genes related to metabolism, proteolysis, receptor and signal transduction, transcription/translation processes and cell adhesion.


Identification of differentially expressed genes by SSH

To identify differentially expressed genes in the two groups of tumors, SSH was performed. Two subtractive cDNA libraries were generated and a total of 1296 clones (648 for GBM and 648 for PA) were further subjected to dot blot screening in order to eliminate false positives. In all, 186 selected clones (80 for GBM and 106 for PA) were sequenced and found to represent, respectively, 36 different known genes for PA (plus eight unknown and 20 unexploitable) and 29 for GBM (plus three unknown and 10 unexploitable) (Tables 1 and 2). They were functionally categorized on the basis of known or inferred biological function of their protein product. In PA versus GBM, the 15 genes recorded at least twice and representing 64/106 of the clones sequenced encoded proteins that were related to metabolism (apolipoprotein D: five clones), proteolysis (protease-serine-11: six clones), receptor and signal transduction (PLEKHB1: five clones; DNER: three clones), transcription/translation processes (eukaryotic-translation-elongation-factor-1-α1: seven clones), and cell adhesion and invasion (SPOCK1: seven clones). Interestingly, in GBM versus PA, the five genes recorded at least twice and representing 43/80 (>50%) of the clones sequenced encoded proteins that were all related to cell adhesion and invasion. They included fibronectin (FN1: 34 clones), chitinase-3-like-1 (three clones), osteopontin (OPN), keratoepithelin and fibromodulin (two clones each). Four additional genes, including TIMP3, were recorded once.

Table 1 Results of genes expressed in pilocytic astrocytomas after screening by SSH against glioblastomas
Table 2 Results of genes expressed in glioblastomas after screening by SSH against pilocytic astrocytomas

Validation using QPCR: genes expressed in PA versus GBM

Quantification of PLEKHB1 and DNER mRNA transcripts revealed higher mRNA levels in PA compared with GBM samples (P=0.0009 and P=0.004, respectively). The mean level of PLEKHB1 mRNA was 192.71±107.21 in PA versus 67.13±76.65 in GBM samples and the mean level of DNER was 2.19±1.04 in PA versus 0.70±0.86 in GBM samples. SPOCK1 mRNA levels were on average 5.7 times higher in PA (mean=3.76±2.35) than in GBM samples (mean=0.65±0.77) (P=0.0017) (Figure 1a).

Figure 1

Scatter plots for the relative expression (R) of SPOCK1 (a), OPN (b), OPN/SPOCK1 (c), FN1 (e), and TIMP3/SPOCK1 (f) in GBM and PA. ROC (receiver operator characteristic) curve for differentiating GBM from PA by the OPN/SPOCK1 ratio (d). The cutoff point was set at 4.64 and the OPN/SPOCK1 ratio showed good agreement with the pathological diagnosis (specificity of 100% and sensitivity of 93.3%).

Validation using QPCR: genes expressed in GBM versus PA

The mean level of DEC1 mRNA expression was 1.42±1.05 in GBM versus 0.43±0.19 in PA samples (P=0.0032). Quantification of OPN mRNA confirmed higher OPN mRNA levels in GBM compared with PA samples (Figure 1b). The mean level of OPN mRNA expression was 14.77±16.14 in GBM, versus 1.61±1.55 in PA samples (P=0.0002). Interestingly, we observed that a ratio of OPN/SPOCK1>4.64 was strongly associated with the diagnosis of GBM (P<0.0001), with specificity of 100% and sensitivity of 93.3% (Figure 1c and d). Quantification of FN1 mRNA levels also confirmed the data obtained with the SSH technique. The mean level of FN1 expression was 2.64±7.12 in GBM versus 0.039±0.021 in PA samples (P=0.0002) (Figure 1e). Finally, the role of TIMP3 in GBM is unclear as this metalloproteinase inhibitor is a well known tumor suppressor and apoptosis inducing protein (Lam et al., 2005). We did not find any significant difference between GBM and PA regarding to the level of TIMP3 mRNA (P=0.85) (data not shown). However, TIMP3/SPOCK1>2.083 ratio could be used to distinguish GBM from PA (P=0.0032), with specificity of 93.3% and sensitivity of 80% (Figure 1f).

Validation at protein level using immunohistochemistry

FN1, OPN and TIMP3 protein expression was analysed by immunohistochemistry in series of formalin-fixed and paraffin-embedded GBM (Figure 2a) and PA (Figure 2b). FN1 showed strong staining in the 20 GBM analysed. In all specimens, microvascular proliferation and some tumor cells located near the neoplastic vessels were deeply stained (Figure 2c). In contrast, neither vessels nor tumor cells were immunolabelled in PA samples (Figure 2d) except for three of them exhibiting weak immunostaining in less than 5% of tumor cells (data not shown). Anti-OPN immunoreactivity was recorded in all GBM cases. The percentage of immunoreactive tumor cells ranged from 10 to 50%. The staining had a dot-like appearance within tumor cell cytoplasm (Figure 2e). In most cases, areas with numerous immunoreactive cells contrasted with others devoid of immunostaining. No staining was recorded in PA samples (Figure 2f) except in one case. Finally, TIMP3 expression was observed in 13/20 and 3/21 in GBM and PA samples, respectively (Figure 2g and h). The immunostaining was intracytoplasmic, weak and ranged from 10 to 50% of cells.

Figure 2

Immunohistochemical analysis of the expression of FN1, OPN and TIMP3 in GBM and PA. Microphotographs of immunohistochemical stained tumor sections for hematoxylin and eosin (a and b). FN1 immunostaining was observed in microvascular proliferation and in tumor cells usually located near the neoplastic vessels (c). Neither vessels nor tumor cells were immunolabeled in PA (d). The OPN immunostaining had a dot-like appearance, within the cytoplasm of GBM (e) but no staining could be detected in PA (f). Finally, TIMP3 immunostaining was intracytoplasmic and weak in GBM (g) and majority of PA studied did not express TIMP3 (h). (a-c, g) 50X; (d-f, h) 150X.


Using SSH approach, we identified 66 genes differentially expressed in PA and GBM tumors. They were functionally categorized in six groups (see Tables 1 and 2). Our results are in agreement with two previous microarray analyses using cDNA or oligonucleotide chips, identifying candidate genes expressed in PA but not in infiltrating anaplastic astrocytoma (Hunter et al., 2002) and in PA but not in GBM (Rickman et al., 2001). These studies identified apolipoprotein D, SPOCK1, protease-serine-11, glia-derived-nexin-1 and protein-tyrosine-kinase-3 as strong markers of PA versus anaplastic astrocytoma or GBM. We also confirmed the expression of a large number of genes involved in brain development in PA (Gutmann et al., 2002; Wong et al., 2005), suggesting that these tumors share gene expression profile with fetal brain and oligodendroglial precursors. Moreover, we reported in PA and confirmed by QPCR, strong expression of two novel genes, PLEKHB1 and DNER, which might be of diagnostic value. On the other hand, the expression of DEC1, a cAMP inducible bHLH transcription factor involved in embryo chondrocyte and neuron proliferation and/or differentiation, characterizes GBM (Shen et al., 1997).

As invasive behavior is a hallmark of GBM, we have focused our study on genes involved in this process. GBM spread throughout the brain as single cells, but usually, they do not metastasize outside the brain (Kleihues and Cavenee, 2000). Our study shows that GBM differ from PA by their expression of FN1, OPN, keratoepithelin, fibromodulin, chitinase-3-like-1 (YKL-40) and to a lesser extent TIMP3, suggesting that these proteins generate a neoplastic extracellular matrix (ECM) providing a permissive environment critical for tumor progression and brain invasion.

FN1, a glycoprotein of about 220 kDa which forms ECM when secreted, has also been identified in GBM by microarray strategy using cDNA or oligonucleotide chips (Sallinen et al., 2000; Tanwar et al., 2002; Freije et al., 2004). The FN1 mRNA has three alternative splicing sites termed EDA, EDB (B-FN) and IIICS in human (Kumazaki et al., 1999). B-FN is a marker of angiogenesis that accumulates in blood vessels in high-grade astrocytomas but not in PA (Castellani et al., 2002). Therefore, GBM cells known to express at least three different FN1 receptors α5β1, αvβ3 or CD44 (Bellail et al., 2004) also secrete FN1, providing both a permissive environment for tumor progression and angiogenesis.

OPN glycoprotein is expressed and secreted by numerous human cancers including malignant astrocytomas (Saitoh et al., 1995; Coppola et al., 2004). The OPN protein contains important functional domains including a central region that binds to seven different integrins (including α5β1, αvβ3 and αvβ5), a calcium-binding domain and an heparan-binding domain which mediates CD44V3 binding (Wai and Kuo, 2004). OPN increases invasiveness by inducing proteinases, particularly uPA (Tuck et al., 1999) and acts in concert with several growth factors including EGF (Tuck et al., 2003). OPN is a high-affinity ligand for the αvβ3 integrin and has also been implicated in angiogenesis (Rittling and Chambers, 2004). Therefore, it is likely that OPN increased GBM tumorigenicity by different mechanisms: degradation of ECM, angiogenesis and cell spreading on enriched fibronectin ECM.

GBM expressed YKL-40. It is a growth factor for connective tissue cells and stimulates migration of endothelial cells. YKL-40 has been shown to be strongly elevated in serum and biopsy material from GBM patients by ELISA assay, Western blot and oligonucleotide array analysis (Lal et al., 1999; Markert et al., 2001; Tanwar et al., 2002). These independent methods of determining gene expression levels additionally validate our findings. High YKL-40 mRNA expression was significantly associated with poorer survival; moreover, they demonstrated that YKL-40-transfected immortalized human astrocytes consistently showed increased resistance to radiation and a fivefold induction in invasive potential as compared with control cells (Nigro et al., 2005).

In addition, SPOCK1, another gene specifically involved in brain cell adhesion and invasion (Bonnet et al., 1992; Charbonnier et al., 1998), was highly expressed in PA. The purified gene product has been shown to inhibit cell attachment and neurite extension in culture and to inhibit membrane type 1 and 2 matrix metalloproteinases (Edgell et al., 2004). It remains to be determined whether SPOCK1 expressed by PA contributes to their noninvasive behavior. Interestingly, we observed that a ratio of OPN/SPOCK1>4.64 within a tumor was highly predictive of GBM.

In conclusion, our study provides cues to the identity of molecules likely to play a crucial role in GBM brain invasion. Moreover, it offers a restricted list of markers that accurately distinguish GBM from PA. Diagnosis of PA might be difficult in some cases (more than 60% interobserver discordance according to Mittler et al. (1996)) and finding reliable markers is critical because of the marked differences in the therapy and prognosis of these two tumors.

Materials and methods

Human tissues

The present study was undertaken after informed consent was obtained from each patient or relatives. Brain tumors were classified according to WHO CNS tumor classification (Kleihues and Cavenee, 2000). SSH was performed using four PA of the posterior fossa (one male and three females, mean age 5 years) and four grade IV GBM (three males and one female, mean age 67 years). PA were cured by total surgical excision and no recurrence was observed 5 years later. In contrast, patients with GBM died within 1 year. Samples were fixed in 10% formalin for routine histology and immunohistochemistry and snap frozen and stored at −80°C for transcriptome analysis. Before RNA isolation, a cryostat section was taken from each block of tissue in order to confirm that it was representative of the tumor and did not contain normal parenchyma or more than 20% of necrosis. For QPCR, the same specimens and additional specimens of six PA of the posterior fossa and 11 GBM were obtained and processed in the same way as above. Immunohistochemistry was carried out on formalin-fixed paraffin-embedded specimens of 5/6 PA and 4/11 GBM. In total, 16 additional PA of the posterior fossa and 16 additional GBM were also studied by immunohistochemistry.

RNA preparation

Total RNA extraction was performed using the acid guanidinium isothiocyanate/phenol/chloroform procedure (Chomczynski and Sacchi, 1987), and analysed on both spectrophotometer and Agilent 2100 bioanalyzer (Agilent Technologies, France). Only samples with 28S/18S ratio of >1.5 and no evidence of ribosomal degradation were included. Before use, RNA samples were treated with 1 U ribonuclease-free deoxyribonuclease (Promega, France) at 37°C for 1 h.

Suppression Subtractive Hybridization

Poly(A)+ mRNA were isolated from PA and GBM total RNA using an Oligotex mRNA isolation kit (Qiagen, France) and comparison of gene expression between these two RNA population was carried out by SSH using a PCR-Select cDNA subtraction kit (Clontech, Heidelberg, Germany).

Cloning and analysis of subtracted clones

Products from the final PCR amplification were cloned into a pGEM-T vector (Promega) and electro-transferred into XL1BlueMRF. Differential screening was performed to eliminate false positives. Hybridizations were performed in duplicate following classic procedures. Specific clones were prepared using a Qiagen plasmid mini-kit and sequenced (MWG-Biotech, Ebersberg, Germany). Nucleic acid homology searches were carried out using the BLAST program at the NCBI, USA.

Real-time quantitative RT-PCR

cDNA synthesis, real-time quantitative RT-PCR (QPCR) conditions and data analysis were as previously described (Rocchi et al., 2004). Forward and reverse primers for each gene are listed below: PLEKHB1: IndexTerm5′-GCTGACTGTGAACCTACGGGA-3′, IndexTerm5′-TCCAGCAGTGCTGTCTTCCA-3′; DNER: IndexTerm5′-CCTGGTATCCTTTGAAGTGCCA-3′, IndexTerm5′-TCCAGAGCAAAATCAGTGAGG-3′; DEC1: IndexTerm5′-CCAAGTGTACAAGTCAAGACGGG-3′, IndexTerm5′-GGTGCGGCAATTTGTAGGTC-3′; SPOCK1: IndexTerm5′-GGACAAGTACTGGAACCGCTTT-3′, IndexTerm5′-GGTAGTCCTGGGTCACACACACT-3′; OPN: IndexTerm5′-GCCGAGGTGATAGTGTGGTTTA-3′, IndexTerm5′-TGATGTCCTCGTCTGTAGCATC-3′; TIMP3: IndexTerm5′-AAGATGCCCCATGTGCAGT-3′, IndexTerm5′-CCCACCTCTCCACGAAGTT-3′. PCR conditions using the Light Cycler Fast Start DNA Master Sybr Green I kit (Roche Applied Science) were as follow: 5 min at 94°C, followed by 40 cycles of 10 s at 94°C, 10 s at 65, 66, 67, 64, 65, and 61°C for PLEKHB1, DNER, DEC1, SPOCK1, OPN, and TIMP3, respectively, and 15 s at 72°C to generate a 89-bp PCR-product for PLEKHB1, 82-bp for DNER, 72-bp for DEC1, 148-bp for SPOCK1, 95-bp for OPN, and 150-bp for TIMP3. Efficiencies (E) were calculated from the given slopes in the Light Cycler software using dilutions of total RNA prepared from U87 cell line and shown to be 1.82 for PLEKHB1 assay, 1.81 for DNER assay, 1.96 for DEC1 assay, 1.87 for SPOCK1 and TIMP3 assays, and 1.91 for OPN assay. The QPCR conditions for FN and 18S RNA levels have been previously described (Eikmans et al., 2003; Rocchi et al., 2004). The relative expression ratio (R) of the target mRNA and reference rRNA (18S) was calculated using QPCR efficiencies (E) and the crossing point (Cp) deviation of a tumor sample versus the U87 RNA cell line used as control as previously described (Pfaffl, 2001).


After steam-heat-induced antigen retrieval, 5 μm sections of formalin-fixed paraffin-embedded samples of 20 GBM and 21 PA were tested for the presence of fibronectin (FN1, A0245, 1:1000 dilution, Dako, France), OPN (1B20, 1:50 dilution, Assay Designs Inc., Ann Arbor, USA) and tissue inhibitor of metalloproteinase 3 (TIMP3, RB 9253, 1:50 dilution, Lab Vision corporation, USA). An avidin-biotine enzyme complex kit (Zymed) was used for the detection of OPN whereas the detection of FN1 and TIMP3 was performed using a Ventana automate (Ventana Medical Systems SA, Illkirch, France). Positive staining was defined as staining of more than 5% of tumor cells.

Statistical analysis

The association of the results of QPCR with the diagnosis (PA versus GBM) was assessed by the Kolmogorov–Smirnov test performed using the Statview software. Threshold values of OPN/SPOCK1 and TIMP3/SPOCK1 ratios were determined by ROC (Receiver Operator Characteristic) curves using the Analyse-It software (Excel, Microsoft). Sensitivity and specificity of the tests were deduced.


  1. Bellail AC, Hunter SB, Brat DJ, Tan C, Van Meir EG . (2004). Int J Biochem Cell Biol 36: 1046–1069.

  2. Bonnet F, Perin JP, Maillet P, Jolles P, Alliel PM . (1992). Biochem J 288 (Part 2): 565–569.

  3. Caskey LS, Fuller GN, Bruner JM, Yung WK, Sawaya RE, Holland EC et al. (2000). Histol Histopathol 15: 971–981.

  4. Castellani P, Borsi L, Carnemolla B, Biro A, Dorcaratto A, Viale GL et al. (2002). Am J Pathol 161: 1695–1700.

  5. Charbonnier F, Perin JP, Mattei MG, Camuzat A, Bonnet F, Gressin L et al. (1998). Genomics 48: 377–380.

  6. Chomczynski P, Sacchi N . (1987). Anal Biochem 162: 156–159.

  7. Coppola D, Szabo M, Boulware D, Muraca P, Alsarraj M, Chambers AF et al. (2004). Clin Cancer Res 10: 184–190.

  8. Diatchenko L, Lau YF, Campbell AP, Chenchik A, Moqadam F, Huang B et al. (1996). Proc Natl Acad Sci USA 93: 6025–6030.

  9. Edgell CJ, BaSalamah MA, Marr HS . (2004). Int Rev Cytol 236: 101–122.

  10. Eikmans M, Baelde HJ, Hagen EC, Paul LC, Eilers PH, De Heer E et al. (2003). J Am Soc Nephrol 14: 899–907.

  11. Freije WA, Castro-Vargas FE, Fang Z, Horvath S, Cloughesy T, Liau LM et al. (2004). Cancer Res 64: 6503–6510.

  12. Gutmann DH, Hedrick NM, Li J, Nagarajan R, Perry A, Watson MA . (2002). Cancer Res 62: 2085–2091.

  13. Hunter S, Young A, Olson J, Brat DJ, Bowers G, Wilcox JN et al. (2002). J Neuropathol Exp Neurol 61: 275–281.

  14. Khatua S, Peterson KM, Brown KM, Lawlor C, Santi MR, LaFleur B et al. (2003). Cancer Res 63: 1865–1870.

  15. Kleihues P, Cavenee WK . (2000). World Health Organization classification of tumors of the nervous system. IARC/WHO: Lyon.

    Google Scholar 

  16. Kumazaki T, Mitsui Y, Hamada K, Sumida H, Nishiyama M . (1999). J Cell Sci 112 (Part 10): 1449–1453.

  17. Lal A, Lash AE, Altschul SF, Velculescu V, Zhang L, McLendon RE et al. (1999). Cancer Res 59: 5403–5407.

  18. Lam P, Sian Lim K, Mei Wang S, Hui KM . (2005). Mol Ther 12: 144–152.

  19. Markert JM, Fuller CM, Gillespie GY, Bubien JK, McLean LA, Hong RL et al. (2001). Physiol Genomics 5: 21–33.

  20. Mischel PS, Cloughesy TF, Nelson SF . (2004). Nat Rev Neurosci 5: 782–792.

  21. Mittler MA, Walters BC, Stopa EG . (1996). J Neurosurg 85: 1091–1094.

  22. Nigro JM, Misra A, Zhang L, Smirnov I, Colman H, Griffin C et al. (2005). Cancer Res 65: 1678–1686.

  23. Pfaffl MW . (2001). Nucleic Acids Res 29: e45.

  24. Rickman DS, Bobek MP, Misek DE, Kuick R, Blaivas M, Kurnit DM et al. (2001). Cancer Res 61: 6885–6891.

  25. Rittling SR, Chambers AF . (2004). Br J Cancer 90: 1877–1881.

  26. Rocchi P, Muracciole X, Fina F, Mulholland DJ, Karsenty G, Palmari J et al. (2004). Oncogene 23: 9111–9119.

  27. Saitoh Y, Kuratsu J, Takeshima H, Yamamoto S, Ushio Y . (1995). Lab Invest 72: 55–63.

  28. Sallinen SL, Sallinen PK, Haapasalo HK, Helin HJ, Helen PT, Schraml P et al. (2000). Cancer Res 60: 6617–6622.

  29. Shen M, Kawamoto T, Yan W, Nakamasu K, Tamagami M, Koyano Y et al. (1997). Biochem Biophys Res Commun 236: 294–298.

  30. Tanwar MK, Gilbert MR, Holland EC . (2002). Cancer Res 62: 4364–4368.

  31. Tuck AB, Arsenault DM, O'Malley FP, Hota C, Ling MC, Wilson SM et al. (1999). Oncogene 18: 4237–4246.

  32. Tuck AB, Hota C, Wilson SM, Chambers AF . (2003). Oncogene 22: 1198–1205.

  33. van den Boom J, Wolter M, Kuick R, Misek DE, Youkilis AS, Wechsler DS et al. (2003). Am J Pathol 163: 1033–1043.

  34. Wai PY, Kuo PC . (2004). J Surg Res 121: 228–241.

  35. Wong KK, Chang YM, Tsang YT, Perlaky L, Su J, Adesina A et al. (2005). Cancer Res 65: 76–84.

Download references


This work was supported by the ARC, by the GEFLUC to DFB, by the ‘Cancéropôle PACA’, by the INCA (Institut National du Cancer, Grant No. RS019) and by institutional Grants from EA 3281 and a Grant from the Fondation Nélia and Amadeo Barletta (FNAB). We thank Dr G Rougon for helpful comments on this manuscript. We are grateful to the neurosurgeons, Professors F Grisoli, H Dufour, JC Peragut, Drs G Lena, D Scavarda, B Fuentes and P Metellus for providing tumor samples and to S Tong for technical assistance.

Author information



Corresponding author

Correspondence to D Figarella-Branger.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Colin, C., Baeza, N., Bartoli, C. et al. Identification of genes differentially expressed in glioblastoma versus pilocytic astrocytoma using Suppression Subtractive Hybridization. Oncogene 25, 2818–2826 (2006).

Download citation


  • Human gliomas
  • SSH (Suppression Subtractive Hybridization)
  • real-time quantitative RT-PCR
  • SPOCK1
  • fibronectin

Further reading


Quick links