Different angiogenic phenotypes in primary and secondary glioblastomas



Primary and secondary glioblastomas (pGBM, sGBM) are supposed to evolve through different genetic pathways, including EGF receptor and PDGF and its receptor and thus genes that are involved in tumor-induced angiogenesis. However, whether other angiogenic cytokines are also differentially expressed in these glioblastoma subtypes is not known so far, but this knowledge might be important to optimize an antiangiogenic therapy. Therefore, we studied the expression of several angiogenic cytokines, including VEGF-A, HGF, bFGF, PDGF-AB, PDGF-BB, G-CSF and GM-CSF in pGBMs and sGBMs as well as in gliomas WHO III, the precursor lesions of sGBMs. In tumor tissues, expression of all cytokines was observed albeit with marked differences concerning intensity and distribution pattern. Quantification of the cytokines in the supernatant of 30 tissue-corresponding glioma cultures revealed a predominant expression of VEGF-A in pGBMs and significantly higher expression levels of PDGF-AB in sGBMs. HGF and bFGF were determined in nearly all tumor cultures but with no GBM subtype or malignancy-related differences. Interestingly, GM-CSF and especially G-CSF were produced less frequently by tumor cells. However, GM-CSF secretion occurred together with an increased number of simultaneously secreted cytokines and correlated with a worse patient prognosis and may thus represent a more aggressive angiogenic phenotype. Finally, we confirmed an independent contribution of each tumor-derived cytokine analyzed to tumor-induced vascularization. Our data indicate that an optimal antiangiogenic therapy may require targeting of multiple angiogenic pathways that seem to differ markedly in pGBMs and sGBMs. © 2005 Wiley-Liss, Inc.

Gliomas are the most common primary malignant brain tumors in adults.1, 2, 3 A grading scheme proposed by the World Health Organization (WHO) distinguishes 4 different grades of gliomas, of which glioblastoma multiforme WHO IV (GBM) is the most malignant one.4

GBM may develop rapidly without clinical and histopathological evidence of a less malignant precursor lesion as a de novo or primary glioblastoma (pGBM) or more slowly through progression from low-grade (WHO II) or anaplastic (WHO III) astrocytoma to a secondary glioblastoma (sGBM).5 Recent studies have shown that these glioblastoma subtypes evolve through different genetic pathways, affect patients at different ages and are quite likely to differ in their responses to therapy. pGBMs occur in elder patients and typically show overexpression of epidermal growth factor receptor (EGF receptor), PTEN mutations, p16 deletions, LOH on 10p and, less frequently, MDM2 amplification. sGBMs develop in younger patients and often contain p53 mutations, and overexpress the platelet-derived growth factor (PDGF) ligand and also the PDGF receptor as their earliest detectable alteration.5, 6, 7, 8 Interestingly, the EGF/EGF receptor as well as the PDGF/PDGF receptor pathway are not only supposed to stimulate tumor growth in an autocrine manner but also to act in a paracrine fashion by inducing neoangiogenesis.9, 10, 11, 12, 13

PDGF is composed of 2 disulfide-linked polypeptide chains designated as A and B,14 which dimerize in all possible combinations to PDGF-AA, PDGF-BB and PDGF-AB. PDGF is known to be stored by platelets and to be expressed by mononuclear phagocytes, endothelial cells, vascular smooth muscle cells and megakaryocytes, and also by tumor cells.11, 12, 15

Besides EGF and PDGF, a number of angiogenic cytokines have been identified, which are released by glioma cells and are able to induce endothelial cells to proliferate and migrate towards the tumor, such as vascular endothelial growth factor A (VEGF-A), hepatocyte growth factor (HGF), basic fibroblast growth factor (bFGF), granulocyte-colony-stimulating factor (G-CSF) and granulocyte-macrophage-colony-stimulating factor (GM-CSF,16, 17, 18, 19, 20, 21).

VEGF-A is one of the best studied angiogenic growth factors. It is considered to be a key regulator in tumor-induced neoangiogenesis.22 VEGF-A consists of 4 different monomers generated by differential splicing.23, 24 VEGF165, the major monomer of the VEGF-A family, is one of the most potent mitogens and chemoattractants for endothelial cells and increases blood vessel permeability.25, 26 VEGF-A is overexpressed in a variety of malignant human tumors, including breast cancer, lung cancer and gliomas.27, 28, 29 HGF is a heterodimeric molecule that has been shown to act as a mediator of tumor invasion30 and is also a powerful inducer of angiogenesis in vivo.31 An upregulation of HGF has been shown in different tumors, including malignant gliomas.20, 32 bFGF mediates important steps in angiogenesis by inducing endothelial cell proliferation and migration in vivo and in vitro and by regulating proteases like interstitial collagenase, urokinase type plasminogen activator (uPA) and plasminogen activator inhibitor (PAI-1), and of uPA receptor and adhesion molecules such as β1 integrin.33, 34 G-CSF and GM-CSF belong to a family of growth factors controlling the proliferation, maturation, and functional activity of granulocytes, macrophages and their precursors.35 Outside the hematopoietic system, they induce proliferation and migration of endothelial cells in vitro18 and angiogenesis in vivo.19

Although the last years distribution and expression levels of different angiogenic growth factors in gliomas have been subject of intensive research, there is only little knowledge regarding the simultaneous expression of different angiogenic factors and possible differences in certain GBM subtypes. However, such information may be essential for planning an antiangiogenic therapy, which targets angiogenic factors, e.g., this knowledge could help to avoid compensation by other simultaneously secreted angiogenic cytokines. Therefore, we analyzed protein expression of the above described angiogenic growth factors in native tumor tissues, quantified secretion of the cytokines in conditioned media of 30 corresponding glioma cultures and confirmed angiogenic activity of tumor cell-derived cytokines. Our goal was to assess expression profiles of angiogenic growth factors that possibly contribute to the angiogenic phenotype of primary and secondary GBMs.


bFGF, basic fibroblast growth factor; EGF, epidermal growth factor; GBM, glioblastoma; G-CSF, granulocyte-colony-stimulating factor; GM-CSF, granulocyte-macrophage-colony-stimulating factor; HGF, hepatocyte growth factor; PDGF, platelet derived growth factor; pGBM, primary glioblastoma; sGBM, secondary glioblastoma; VEGF-A, vascular endothelial growth factor; WHO, World Health Organization.

Material and methods


Tissue specimens of 22 GBMs WHO IV (8 sGBM and 14 pGBM) were obtained intraoperatively from central parts of the tumors. Tumors were histopathologically classified according to the classification of the WHO. In general, fresh tissues were divided into 2 parts; 1 part was used to establish primary tumor cultures, and the other part was snap-frozen immediately after surgery in liquid nitrogen and stored at −80°C until processing. However, in 1 case (sGBM NCH59), material was sufficient only to establish a tumor culture. In addition, tumor tissues were obtained from 8 patients suffering from gliomas WHO III (6 astrocytomas WHO III and 2 oligoastrocytomas WHO III) to establish primary tumor cell cultures. Clinical data of the respective patients are summarized in Table I.

Table I. Patient Data
PatientAgeSexPFS (months)OS (months)TherapyPrecur. lesionLocalization
  • Age, age in years at time of diagnosis; PFS, progression free survival in months; OS, overall survival in months after surgery; precur. lesion, precursor lesion; rad, radiation therapy; op, operation; chemo, chemotherapy.

  • 1

    Anaplastic astrocytoma WHO III.

  • 2

    Anaplastic oligoastrocytoma WHO III; M, male; F, female; no recur., no recurrence.

Primary GBM
 NCH3761M410op/rad/chemofrontal, right
 NCH7965F915op/radtemporal, right
 NCH8255M14op/radparietal, left
 NCH8940F26op/rad/chemotemporal, left
 NCH9271F1118op/radtemporal, left
 NCH12543M9152x op/radfrontal, right
 NCH14959M1217op/radtemporal, left
 NCH15654M7212x op/rad/chemopostcentral, right
 NCH18661F59op/radcentral, left
 NCH18755M513op/radtemporal, right
 NCH19059M37op/radtemporal, right
 NCH19464M55Alive (80)op/rad/chemofrontal, right
 NCH21060M1018op/rad/chemofrontal, left
 NCH30738M262x op/rad/chemotemporal, right
Median59 614   
Secondary GBM
 NCH3844M24482x op/rad/chemoWHO IIIparietal, right
 NCH5944F6172x op/rad/chemoWHO IImultilocular, right
 NCH17318M342x op/rad/chemoWHO IIItemporal, right
 NCH19344M10302x op/rad/chemoWHO IIfrontal, right
 NCH19938M2252x op/rad/chemoWHO IIbifrontal
 NCH20357F252x op/rad/chemoWHO IIItemporal, right
 NCH21440M8142x op/radWHO IIfrontal, right
 NCH24637M8192x op/rad/chemoWHO IIfrontal, left
Median42 718   
Gliomas WHO III
 NCH107137M16273x op/rad/chemoWHO IItemporal, right
 NCH133125M6373x op/rad/chemoWHO IIfronto-parietal, right
 NCH135131M32433x op/rad/chemoWHO IIfronto-parietal, left
 NCH192154F5592x op/radWHO IIfrontal, right
 NCH200233M5162x op/rad/chemoWHO IItemporal, right
 NCH218143F6202x op/rad/chemoWHO IIIfrontal, right
 NCH228238M28704x op/rad/chemoWHO IIparieto-occipital, left
 NCH273167M22222x op/rad/chemoWHO IIfrontal, left
Median37.5 1132   

Cell culture conditions

Primary cultures of glioma tissues were established by dissecting tissues in small pieces of about 1 mm3 and transferring them into 75 cm2 plastic tissue-culture flasks (Falcon, Becton Dickinson, Heidelberg, Germany). Cells were cultured routinely in DMEM supplemented with 10% fetal calf serum and antibiotics in a humidified incubator at 37°C, with 5% CO2 and 95% air. Cells were grown to confluence with medium changes twice a week and were harvested by a brief incubation of trypsin/EDTA solution (Viralex, PAA, Linz, Austria). Mycoplasma contamination of the cell cultures was excluded by DAPI staining (Roche Diagnostics, Mannheim, Germany).

Cell identification and characterization

Primary glioma cultures used were characterized as described earlier.36 In brief, glial origin of cultured cells was confirmed by staining with an α-glial fibrillary acidic protein (α-GFAP) antibody (Dako, Hamburg, Germany). Furthermore, endothelial and neuronal cell contamination of these cultures was excluded by staining with an α-platelet/endothelial cell adhesion molecule antibody (α-PECAM-1) directed against CD31 (Pharmingen, San Diego, Ca, USA), an α-factor VIII antibody (Dako, Hamburg, Germany), and antibodies against neurofilament protein (NFP) 70, 160 and 200 kDa (all Progen, Heidelberg, Germany).


Immunohistochemical staining was performed on glioma cells grown for 24 hr on slides (Histobond, Marienfeld, Bad Mergentheim, Germany) as well as on serial cryostat sections (5–7 μm) of the frozen biopsies mounted on 3-aminopropyl-triethoxysilan coated slides. Acetone (10 min) at −20°C and alternatively freshly prepared buffered 4% paraformaldehyde (20 min) were used as fixatives. In parallel, hematoxylin and esosin staining was performed on the first and last cryostat section to confirm that samples were obtained from vital tumor areas. Incubation with the primary and secondary antibodies and detection was carried out as described elsewhere,37 with Vectastain Laboratories Elite ABC Kit (Vector Laboratories, Burlingame, California). Primary antibodies used apart from the before mentioned were directed against PDGF (recognizing all PDGF subtypes), bFGF, G-CSF, GM-CSF (all Dianova, Hamburg, Germany), HGF (R&D Systems Wiesbaden, Germany) and VEGF-A (PromoCell, Heidelberg, Germany). Immunohistochemical analysis of angiogenic growth factors was carried out by 2 investigators independent from one another.

Conditioned media and quantification ELISA

For the preparation of conditioned media, 1.26 × 106 cells were seeded into 6 cm2 dishes in DMEM, with 10% FCS. After 24 hr, the medium was shifted to 2.5 ml DMEM and 0% FCS, and 2 days later, the conditioned medium was harvested, centrifuged for 10 min at 10,000g and stored in aliquots at −80°C until use. Viable cells were counted in a cell counter (Casy 1, Tuebingen, Germany). The growth factors VEGF165, HGF, PDGF-AB, PDGF-BB, GM-CSF, G-CSF and bFGF were measured quantitatively using commercially available ELISA Kits (R&D Systems, Wiesbaden, Germany) and performed according to the manufacturer's instructions. Data were referenced to 1 × 106 cells as determined on the harvesting day. Only data with intratest variation values below 10% were included in the present study. Final data are means of at least 2 independent measurements.

2D Coculture model of angiogenesis

Angiogenic effects of tumor cell-derived growth factors were tested using the commercially available AngioKit (TCS Cell Works Ltd, Buckingham, UK), an in vitro multicellular 2D model to determine angiogenic effects. According to Bishop et al.,38 this model contains human umbilical vein endothelial cells (HUVECs) cocultured with human dermal fibroblasts in a specially designed matrix. At the beginning of the experiment, endothelial cells form small islands within the culture matrix. They subsequently begin to proliferate and then enter a migratory phase, during which they move through the matrix to form threadlike tubule structures. These gradually join up (by 11 days) to form a network of anastomosing tubules, which closely resemble a microvascular bed. The AngioKit was supplied as a 24-well multi-well plate. All experimental conditions were performed in duplicates.

Each well received 0.5 ml optimized medium supplied by the manufacturer (supplemented with 2% FCS but with no addition of any of the growth factors analyzed), together with conditioned medium of NCH89 cells at a 1:1 dilution. In addition, in some of the wells, neutralizing antibodies directed to VEGF-A (AB1442, Chemicon, Hofheim, Germany), PDGF (Upstate Biotechnology, Lake Placid, NY), GM-CSF, G-CSF (both Dianova, Hamburg, Germany) and HGF (R&D Systems Wiesbaden, Germany) were applied at concentrations from 2 to 10 μg/ml, according to the manufacturers' instructions. Medium changes were performed on day 1, 4, 7 and 9. Cells were incubated at 37°C with 5% CO2 humidified atmosphere. On day 11, all cells were fixed in ice cold (−20°C) 70% ethanol. Subsequently, the capillary tubules were visualized by indirect immunohistochemical staining for the endothelial cell adhesion molecule CD31 (PECAM-1, Pharmingen, San Diego, Ca, USA). At a 40-fold magnification, 5 representative photographs of the tubule development were taken. Quantification of the tubules was performed using the image analysis system analySIS (Olympus, Hamburg, Germany), which allowed assessment of tubule length, number of tubules and number of junctions.

Statistical analysis

Correlation of quantitative growth factor expression with glioma subgroups was determined by Kruskal-Wallis Test, followed by a 2-sided Wilcoxon Test. Survival data were correlated with growth factor expression by Spearman Rank Order Correlation Test. A 2-sided t-test was used for correlation of changes of tubule length with the application of neutralizing antibodies. p-values < 0.05 were considered to be statistically significant.


Growth factor expression in tissues from pGBMs and sGBMs

Expression of the angiogenic growth factors VEGF-A, PDGF, HGF, GM-CSF, G-CSF and bFGF was analyzed by immunohistochemistry on cryostat sections of 14 pGBMs and 7 sGBMs (Table II). Of all of the growth factors analyzed VEGF-A, PDGF and bFGF were expressed in all GBM tissues. However, staining for VEGF-A seemed to be more extensive in pGBMs, while in the same tumors, PDGF staining was more heterogeneous and weaker compared to sGBMs (Fig. 1). Expression pattern of bFGF was similar in both GBM groups consisting of a predominant blood vessel staining and a focal expression in some tumor areas. HGF was found in almost all pGBMs (12/14) and all sGBMs (7/7), varying from a focal and weak staining to a strong staining. Similarly, GM-CSF expression was assessed in 12/14 pGBMs and 6/7 sGBMs. However, in the majority of tumors, staining was localized in single cells (Fig. 1) or missing completely. Only 6/14 pGBMs and 1/7 sGBMs showed a homogeneous distribution of GM-CSF in the whole tumor tissue. Regarding the growth factor G-CSF, expression was observed in 11/14 pGBMs and 2/7 sGBMs. In contrast to GM-CSF, staining was stronger in most of the G-CSF-positive tumors.

Figure 1.

Immunohistochemical analysis of angiogenic growth factors in tumor tissue of pGBM NCH156 and sGBM NCH203. Staining is presented for VEGF-A, HGF, PDGF, bFGF, G-CSF and GM-CSF. In pGBM, strongest staining can be observed for VEGF-A, while G-CSF and bFGF show the weakest and most heterogeneous staining. In sGBM, the most pronounced staining can be seen for PDGF followed by HGF. VEGF-A, bFGF and GM-CSF expression can only be dected in single cells, and staining for G-CSF is lacking completely. Scale bar = 100 μm.

Table II. Immunohistochemical Analysis of Angiogenic Growth Factors in Native GBM Tissues
  1. −, no staining; (+), focal staining; +, weak staining; ++, moderate staining; +++, strong staining; n.d., not determined.


Quantification of growth factors in pGBMs, sGBMs and gliomas WHO III

To exclude growth factor production by nontumor cells and to obtain quantitative data on tumor-cell-derived growth factor secretion, we established GBM cultures corresponding to the native tissues analyzed by immunohistochemistry. From these GBM cultures, conditioned cell culture supernatant was harvested, and the same panel of growth factors analyzed in the corresponding tumor tissues was quantified using sandwich ELISAs. In addition, we included tumor cell cultures of 8 gliomas WHO III representing the precursor lesions of sGBMs to assess whether the growth factor pattern of sGBM resembles to that seen in gliomas WHO III. Results are summarized in Table III.

Table III. Growth Factor Quantification in Cell Culture Supernatants of Glioma Cultures in ng/ml/106 Cells
Tumor culturesVEGF-APDGF-ABPDGF-BBHGFGM-CSFG-CSFbFGFSimult. expr. GF (n)PFS (months)OS (months)
  • Simult. expr. GF, simultaneously expressed growth factors; n, number; PFS, progression free survival in months; OS, overall survival in months after surgery.

  • 1

    Patient still alive.

Glioma WHO III

Secretion of considerable amounts of VEGF-A was observed in all GBMs as well as in all gliomas WHO III ranging from 0.171 to 90.742 ng/ml/106 cells. However, with the exception of the sGBM NCH199, highest VEGF-A levels were detected in pGBMs. PDGF-AB secretion could also be demonstrated in most of the tumor cultures. High levels of PDGF-AB (above 1 ng/ml/106 cells) were almost exclusively seen in sGBMs. Expression of GM-CSF and G-CSF was predominantly observed in pGBMs. Regarding the other cytokines, we saw no glioma subtype-related differences.

Simultaneous growth factor expression and characteristic growth factor profiles in pGBMs, sGBMs and gliomas WHO III

To further identify typical growth factor profiles that may contribute to the establishment of different GBM subtypes, we focused on the simultaneous growth factor secretion (Table III). All tumor cultures secreted at least 3 of the growth factors tested, but only 1/8 sGBM (13%) and 1/14 pGBM cultures (7%) secreted no more than these 3 growth factors. The vast majority of gliomas WHO III (75%) and nearly half of the GBM cultures analyzed produced 4 growth factors simultaneously (6/8 gliomas WHO III, 4/8 sGBMs, 6/14 pGBMs), while secretion of 6 growth factors at the same time was observed only in GBM-derived tumor cell cultures (3/14 pGBMs, 1/8 sGBMs).

In addition, we wanted to know whether the total amount of growth factors produced, being important for the efficacy of growth factor-related functions, differ among the malignant gliomas analyzed. Comparison of mean growth factor values in gliomas WHO III, sGBMs and pGBMs are presented in Figure 2. We identified a predominant secretion of high VEGF-A levels in pGBM cells with a mean value of 18.262 ng/ml/106 cells, while mean values for sGBMs and gliomas WHO III were much lower, with 3.162 and 0.89 ng/ml/106 cells, respectively. Although the case number is small, differences between pGBM and glioma WHO III values turned out to be statistically significant (p = 0.029). Comparison of VEGF-A amounts in pGBM and sGBM cultures revealed a p-value of 0.067. In contrast, mean values for PDGF-AB revealed highest concentrations in sGBM, with 3.593 ng/ml/106 cells, followed by gliomas WHO III and pGBMs, with 1.457 and 0.741 ng/ml/106 cells, respectively. Differences in PDGF-AB levels between pGBMs and sGBMs were significant (p = 0.029). Secretion levels of HGF and bFGF did not show any significant difference between all glioma subgroups (Fig. 2). Although GM-CSF and G-CSF were almost exclusively produced by pGBMs and sGBMs, a significant difference could only be determined for GM-CSF when comparing pGBMs and gliomas WHO III (p = 0.029), while values of pGBMs and sGBMs did not differ significantly (GM-CSF, p = 0.365 and G-CSF, p = 0.441).

Figure 2.

Quantification of angiogenic growth factors by ELISA in conditioned media of gliomas WHO III, sGBMs and pGBMs. Mean growth factor values in different glioma subgroups are illustrated. Data are based on results of ELISA experiments shown in Table III. pGBMs secrete highest mean levels of VEGF-A compared to sGBMs and gliomas WHO III, while PDGF-AB is produced in highest concentrations in sGBMs followed by gliomas WHO III. G-CSF and GM-CSF are mainly secreted in GBM. HGF and bFGF show a similar expression level in all glioma groups tested. * depicts significant differences with p-values < 0.05.

Finally, we wanted to find out whether the secretion of a specific growth factor is associated with a poorer clinical prognosis. We compared patient's progression-free survival and overall survival time with the quantitative growth factor amount expressed by the corresponding tumor cell culture. Results as determined by Spearman Rank Order Correlation Test revealed a significant inverse correlation between VEGF-A protein levels in vitro and progression-free survival of the respective glioma patients (r = −0.385, p < 0.05). Also, we determined a significantly poorer progression-free survival and overall survival time for patients bearing tumors, which produced GM-CSF (r = −0.604, p < 0.001 and r = −0.52, p < 0.01, respectively). In accordance with this observation, shortest survival of patients with gliomas WHO III was observed when their tumor cultures secreted GM-CSF. However, for the other growth factors, there was no significant correlation with patient prognosis.

Angiogenic activity of glioma-secreted growth factors

To confirm our hypothesis that the tumor-secreted growth factors analyzed are important for the angiogenic phenotype of gliomas, we determined the angiogenic activity of tumor-derived growth factors in a 2D coculture model of HUVECs and fibroblasts while blocking a specific cytokine. For 11 days, endothelial cells and fibroblasts were exposed in duplicates to conditioned supernatant of NCH89 cells, which have been shown to produce rather high amounts of VEGF-A, PDGF-AB, HGF, GM-CSF and G-CSF (Table III). These were compared to cocultures in which growth factor-neutralizing antibodies were added to the medium. During this period, endothelial cells formed tubule-like structures joining up to a vessel network. Development of tubules was assessed on day 11 by immunohistochemical staining with an α-CD31 antibody recognizing endothelial cells and subsequent image analysis of the tubule length (Figs. 3a and 3b). Application of all neutralizing antibodies revealed a significantly reduced total tubule length (Fig. 3b) with p-values < 0.01. This effect was most pronounced when inhibiting HGF. However, regarding the number of endothelial junctions that contribute to the vessel network, the most dramatic effect could be observed after inhibition of VEGF-A (Fig. 3a).

Figure 3.

Angiogenic activity of glioma-derived growth factors. A coculture of HUVECs and fibroblasts was exposed to conditioned medium of NCH89 cells secreting all factors analyzed in considerable amounts in the presence of growth-factor neutralizing antibodies. After an incubation period of 11 days, formation of tubule-like structures and a vessel network was assessed by staining of endothelial cells with α-CD31 antibodies. In (a), representative photographs for all experimental conditions analyzed are presented. (b) shows the evaluation of the photographs by using an image analysis system. In the presence of neutralizing antibodies, a significant reduction of total tubule length was determined compared to untreated cells. * depicts significant differences with p-values < 0.01.


The concept that cancer cells secrete distinct substances responsible for the stimulation of angiogenesis was initially proposed by Judah Folkman.39 Especially, these early steps of angiogenesis are therapeutical targets for several classes of antiangiogenic compounds.40 Among solid neoplasms, GBM is one of the most highly vascularized4; therefore, treatments that target the neovascularization process could be of great clinical importance to improve prognosis of these aggressive tumors. However, the formation and malignant progression of human gliomas are complex processes involving chromosomal multiploidy and genetic mutations of either the EGFR or the PDGF/PDGF receptor system in pGBMs or sGBMs, respectively, being known to stimulate angiogenesis.5, 7, 8 Whether other angiogenic factors may also be differentially expressed in pGBMs and sGBMs resulting in different angiogenic phenotypes has so far not been analyzed.

Therefore, we studied the expression of a panel of angiogenic growth factors in tissues from primary and secondary GBMs and corresponding tumor cell cultures. With the exception of G-CSF, expression of the other angiogenic cytokines was observed in all (VEGF-A, PDGF and bFGF) or nearly all (HGF and GM-CSF) GBM tissues analyzed. However, there were remarkable differences with respect to distribution pattern and staining intensity. Most interestingly, a strong and homogeneous staining for VEGF-A was more often observed in pGBMs, while PDGF showed a more intense staining mainly in sGBMs. Since cytokines are proteins that are secreted in the extracellular matrix, analysis of the tumor tissue does not allow us to identify the cellular source of the respective growth factors and to quantify the produced amount of a certain cytokine. To circumvent these problems, we measured tumor-secreted amounts of the same growth factors in conditioned cell culture supernatant of corresponding GBM cultures. In addition, analysis of glioma WHO III cultures representing the precursor lesions of sGBMs was included to assess similarities and malignancy-associated changes of growth factor profiles. While VEGF-A, bFGF PDGF-AB and HGF were determined in all or most glioma cultures, GM-CSF and especially G-CSF were secreted less frequently, consistent with the results obtained from the originating GBM tissues. PDGF-BB was not produced by any of the glioma cultures. Also, we were able to confirm tumor-specific differences. For VEGF-A, highest values were found in pGBMs, while most of the sGBM and all glioma WHO III cultures secreted levels below 2 ng/ml/106 cells. In contrast, significantly higher amounts of PDGF-AB were found in sGBMs. Even in gliomas WHO III, mean PDGF-AB values were higher as that seen for pGBMs. Noteworthy, G-CSF and GM-CSF were predominantly produced in pGBM cultures. Occasional discrepancies between no detection of a growth factor in tumor cell cultures via ELISA and positive staining in tumor tissues may be caused by an expression of these cytokines by different cells in tissues such as endothelial cells and immune cells or by the storage of the growth factors in the extracellular matrix.

It is of note that differences in VEGF-A and PDGF-AB expression between pGBMs and sGBMs have not been described before. This may be caused by the fact that most studies compared gliomas WHO III with GBMs without discriminating pGBMs and the less frequently occurring sGBMs. Nevertheless, our data fit well to the observations describing a PDGF-A chain and PDGF receptor overexpression as earliest detectable alterations of sGBMs and their precursor lesions.5, 7 Also, others postulated a possible contribution of either the PDGF-A or the PDGF-B chain to tumor growth, angiogenesis and tumor progression.41, 42, 43

In accordance with previous data describing a positive correlation between VEGF-A expression in tumor tissues and patient prognosis,44 we observed an inverse correlation between PFS of glioma patients and quantitative VEGF-A levels in vitro. Especially interesting, we were able to demonstrate a statistically significant correlation between GM-CSF secretion and a worse PFS and OS although the case number was small. The observation that nearly all GM-CSF-producing cultures secreted at least 5 or 6 cytokines simultaneously and the occurrence of GM-CSF in glioma WHO III patients with the shortest survival times strengthens the hypothesis that during malignant progression tumors become increasingly independent of the growth regulatory mechanisms of the surrounding tissue by producing growth factors themselves.

In previous studies, we and others have shown that all growth factors analyzed not only exert paracrine functions on endothelial cells but can also act in an autocrine manner on tumor cells.21, 32, 36, 45 To determine the specific contribution of these cytokines to the stimulation of blood vessel growth, we tested the angiogenic activity of conditioned medium of NCH89 cells secreting all growth factors analyzed in considerable amounts in a 2D angiogenesis assay by inhibiting one of the tumor-cell-produced growth factors at a time. This allowed us to mimic an antiangiogenic monotherapy by targeting only 1 angiogenic factor. Neutralization of every single cytokine revealed a significantly reduced tubule length, supporting the idea that the worse prognosis of patients secreting an increased number of growth factors might be caused by their elevated angiogenic activity and demonstrates an independent contribution of every angiogenic cytokine to the angiogenic phenotype. Thus, an antiangiogenic agent inhibiting only one or few growth factor pathways might not be effective in advanced tumors.

In summary, we have shown some marked differences in growth factor profiles of pGBMs and sGBMs. While pGBMs have a tendency to produce higher amounts of VEGF-A, we observed in sGBMs, and gliomas WHO III significantly increased levels of PDGF-AB, which further supports the concept of Kleihues et al.6 that pGBMs and sGBMs develop by different pathways. Most interestingly, in this small collection of tumors, production of GM-CSF correlated with a worse patient prognosis and may thus represent a more aggressive angiogenic phenotype by reflecting the stepwise acquired ability to escape from regulatory mechanisms of the tumor environment. Finally, our data on the contribution of the cytokines analyzed to tumor-induced vascularization strongly suggest that an antiangiogenic therapy against 1 or 2 angiogenic cytokines alone might be compensated in advanced tumors by other simultaneously produced angiogenic growth factors. Therefore, optimal cancer treatment may require targeting of multiple angiogenic pathways, and as Jain stated very recently46 “the challenge for the oncologist will be to formulate combinations of antiangiogenic agents specifically targeting the angiogenic profile of individual tumors”.


We gratefully acknowledge the help of all medical colleagues in gathering the tumor biopsies. We thank Ms. Heike Westphal, Ms. Melanie Greibich and Ms. Renate Steinle for expert technical assistance, and Mr. Philip Benjamin for photographic work. This work was supported by the “Verein zur Foerderung der Krebsforschung in Deutschland e.V. (CHM, HHS) and by the “Tumorzentrum Heidelberg-Mannheim” (CHM, HHS).