GLIA 54:850–860 (2006)

Glioblastoma-Derived Tumorospheres Identify a Population of Tumor Stem-Like Cells with Angiogenic Potential and Enhanced Multidrug Resistance Phenotype ANDREA SALMAGGI,1* AMERIGO BOIARDI,1 MAURIZIO GELATI,1 ANNAMARIA RUSSO,2 CHIARA CALATOZZOLO,1 EMILIO CIUSANI,1 FRANCESCA LUISA SCIACCA,1 ARIANNA OTTOLINA,1 EUGENIO AGOSTINO PARATI,1 CATERINA LA PORTA,3 GIULIO ALESSANDRI,1 CARLO MARRAS,1 DANILO CROCI,1 AND MARCO DE ROSSI1 1 Istituto Nazionale Neurologico ‘‘Carlo Besta,’’ Via Celoria 11, 20133 Milan, Italy 2 Central Hospital Sant’Anna, Via Napoleona 60, 22100 Como, Italy 3 Department of Biomolecular Science and Biotechnology, Via Celoria 26, 20133 Milan, Italy

KEY WORDS brain tumor stem-like cells; CXCL12; VEGF; MDR proteins and adhesion molecules

ABSTRACT We investigated in vitro the properties of selected populations of cancer stem-like cells defined as tumorospheres that were obtained from human glioblastoma. We also assessed their potential and capability of differentiating into mature cells of the central nervous system. In vivo, their tumorigenicity was confirmed after transplantation into the brain of non-obese diabetic/severe combined immunodeficient (NOD-SCID) mice. The angiogenic potential of tumorospheres and glioblastoma-derived cells grown as adherent cells was revealed by evaluating the release of angiogenic factors such as vascular endothelial growth factor and CXCL12 by ELISA, as well as by rat aortic ring assay. The proliferative response of tumorospheres in the presence of CXCL12 was observed for the first time. Multidrug resistance-associated proteins 1 and 3 as well as other molecules conferring multidrug resistance were higher when compared with primary adherent cells derived from the same tumor. Finally, we obtained cells from the tumor developing after grafting that clearly expressed the putative neural stem cell marker CD133 as shown by FACS analysis and also nestin and CXCR4. The cells’ positivity for glial fibrillary acidic protein was very low. Moreover these cells preserved their angiogenic potential. We conclude that human glioblastoma could contain tumor cell subsets with angiogenic and chemoresistance properties and that this chemoresistance potential is highly preserved by immature cells whereas the angiogenic potential is, to a higher extent, a property of mature cells. A better understanding of the features of these cell subsets may favor the development of more specifically targeted therapies. V 2006 Wiley-Liss, Inc. C

INTRODUCTION Human brain tumors are a significant cause of morbidity and mortality in the pediatric age as well as in adulthood and remain difficult to cure despite advances in surgery and adjuvant therapy. As a matter of fact, median survival in patients with glioblastoma (the most aggressive C 2006 V

Wiley-Liss, Inc.

malignant glioma, WHO grade IV) is
12 months after surgery and adjuvant radiotherapy. This survival time has shown only negligible improvement in the last 20 years with a slight prolongation obtained by adjuvant chemotherapy (Chang et al., 2005; Stupp et al., 2005). Until now, most current brain tumor research has been focused on the molecular and cellular analysis of the bulk tumoral mass that typically comprises morphologically diverse cells. However, the critical role of a restricted population of stem cells in tumor progression has been emphasized recently in other malignancies, such as acute myeloid leukemia (Bonnet and Dick, 1997; Reya et al., 2001) and breast cancer (Al-Hajj et al., 2003). Along this line of research, following the same technique already in use for isolation of neural stem cells (NSCs) (Bez et al., 2003), we generated clusters of glioblastoma (GBM) stem-like cells growing as floating tumorospheres (TS) (Galli et al., 2004; Hemmati et al., 2003; Singh et al., 2003). TS may represent selected GBM-derived cell subsets since, under NSCs culture conditions, only more ‘‘undifferentiated’’ neoplastic cells can survive whereas more ‘‘differentiated’’ tumor cells die and are negatively selected (Galli et al., 2004). The pivotal role in tumor growth played by TS is highlighted by their ability to reconstitute (with a truly similar phenotype) the human tumor from which they take origin; on the contrary, NSCs never give rise to neoplastic lesions (Galli et al., 2004). Only recently CD133, a neural stem cell marker, has been shown to identify a population of GBM cells able to recapitulate the original GBM when injected in NODSCID mice (Singh et al., 2004b). CD1331 cells are therefore obviously considered as a potentially relevant target tumor cell subpopulation for therapy in the context of the whole tumoral mass (Singh et al., 2004b).

Marco De Rossi is currently at Dipartimento di Oncologia Sperimentale, Istituto Europeo di Oncologia, Via Ripamonti 435, 20141 Milano, Italy. *Correspondence to: Dr. Andrea Salmaggi, Istituto Nazionale Neurologico ‘‘Carlo Besta,’’ Via Celoria 11, 20133 Milan, Italy. E-mail: [email protected] Received 1 June 2006; Accepted 1 August 2006 DOI 10.1002/glia.20414 Published online 15 September 2006 in Wiley InterScience (www.interscience. wiley.com).

851

BRAIN TUMOR STEM-LIKE CELLS CHARACTERIZATION

Apart from these important results, very little has been investigated regarding the angiogenic, adhesive, and chemoresistance profile of this GBM subpopulation grown as TS but we consider these features fundamental for a cell population assuming a pivotal role in tumor growth, progression, and relapse. GBMs are extremely vascularized brain tumors (Fischer et al., 2005); therefore we investigated the ability of TS to induce angiogenesis by evaluating microvessel sprouting in a rat aortic ring assay as well as the amount of vascular endothelial growth factor (VEGF) and CXCL12 released by TS culture supernatants. We also investigated the expression of cell adhesion molecules (CAMs), such as the ligands of lymphocyte function-associated antigen (LFA)-1, vascular CAM (VCAM)-1, intercellular adhesion molecule (ICAM)-1, and ICAM-2, as well as the expression of integrins (a5b1, aVb3, and aVb5) described as critical for the invasion of GBM in the central nervous system (CNS) (Bellail et al., 2004; Felding-Habermann and Cheresh, 1993; Goldbrunner et al., 1998). Furthermore, in order to clarify the possible role of the chemoresistance potential of TS in the failure of GBM response to chemotherapy treatment, we analyzed by FACS and real time quantitative (RTQ)-PCR, the multidrug resistance (MDR) associated proteins (MRPs) (Tews et al., 2000) and glutathione-S-transferase (GST)p expression in TS cells (Calatozzolo et al., 2005). These proteins are certainly very well studied in brain tumors (Habgood et al., 2000) but only partial data are available on their expression by TS.

MATERIALS AND METHODS Tissue Samples The use of human CNS tissue was authorized by the Ethics Committee of ‘‘C. Besta’’ National Neurological Institute and Dipartimento di Ostetricia, Ginecologia e Neonatologia, IRCCS Ospedale Maggiore Policlinico and Regina Elena, Milano, Italy. Human NSCs were derived from the brains of 12-week old human embryos obtained following the ethical guidelines of the European Network for Transplantation. Human GBM samples were obtained from three adult patients undergoing craniotomy at ‘‘C. Besta’’ Neurological Institute for right-sided temporal lesions with radiological features typical of GBM multiforme. After surgical resection, the tumor specimens were sent for routine neuropathological evaluation and in part used for extraction of total RNA or preserved in culture medium and processed immediately for cell culture.

In all the cases, histology was consistent with GBMs (WHO classification grade IV) and tumor location was the right temporal lobe.

Histopathologic Analysis Histology was performed on specimens fixed in a 4% paraformaldheyde (PFA) solution in phosphate-buffered saline (PBS) (Gibco) and subsequently embedded in paraffin, cut at 3 lm, and stained with haematoxylin-eosin. Cell Culture and Establishment of NSCs, TS, and GBM-Derived Primary Adherent Cells The experimental conditions driving the formation of NS from single cells and generating NSCs are already described in detail (Bez et al., 2003; Galli et al., 2004) and were followed also for the culture of GBMs-derived TS. Briefly, TS cells were cultured in serum-free medium containing epidermal growth factor (EGF) and basic fibroblast growth factor (bFGF). These culture conditions negatively select mature cells in primary CNS cultures, leaving neural stem-like cells free to proliferate and expand exponentially (Galli et al., 2004). TS-derived cells were obtained by enzymatic disruption with collagenase type IV (37°C; 15 min) and mechanical dissociation and appeared as floating cells capable of reaggregating into spherical units and of differentiating into oligodendrocytic, astrocytic or neuronal phenotypes. We considered dissociation into single cells and the following reaggregation in TS as a culture passage. In this way we satisfied some of the conditions known to isolate CNS precursors in GBMs and their multipotency by immunocytochemistry (ICC). Primary cell cultures were grown to adherence from the same surgical specimens giving rise to TS as previously described (Calatozzolo et al., 2005) and identified as GBM adherent cells (GBM-ADCs). On the contrary, TS obtained in NOD-SCID mice after implantation of human TS cells were utilized at the 3rd passage and called ‘‘Secondary TS.’’ The U87MG human GBM cell line was obtained from the American Type Culture Collection. Human brain microvascular endothelial cells (HBMECs) were obtained from the tumor of a 51-year old male undergoing surgical treatment for a right frontal GBM (Salmaggi et al., 2004). All the cells were maintained at 37°C in an atmosphere with 5% CO2 and the medium was replaced every 3 days. Immunocytochemistry

Patients Clinicopathological Data TS are indicated as TS1, TS2, and TS3. TS1 was obtained from a 77-year old female, TS2 was obtained from a 57-year old male, and TS3 was obtained from an 18-year old female patient.

ICC was performed according to the manufacturer’s instructions and as already described in detail (Parati et al., 2002). Briefly, GBM-derived stem-like cells were obtained after TS disruption by enzymatic digestion (15 min at 37°C) with collagenase type IV (Gibco, InviGLIA DOI 10.1002/glia

852

SALMAGGI ET AL.

trogen S.R.L, Italy) and pipetting mechanically up and down many times in order to obtain a single cell suspension that was then grown for 10 days in the previously mentioned medium (Bez et al., 2003; Galli et al., 2004) in the absence of EGF and bFGF and in the presence of 2% fetal bovine serum (FBS). During this period they differentiated and became mature due to the presence of the serum. Primary antibodies were rabbit antiglial fibrillary acidic protein (GFAP, dilution 1:200, Chemicon), mouse anti-b-III tubulin (dilution 1:100, Chemicon), anti-Galactocerebroside (Gal C, dilution 1:200, R&D), mouse antiCD133 (dilution 1:20, R&D), mouse anti-CXCR4 (dilution 1:50, R&D). Secondary antibodies were goat antimouse (Cy3) or goat antirabbit (Cy2) IgG conjugate (dilution 1:600; Jackson ImmunoResearch). Samples were viewed and photographed with an Eclipse 80I (Nikon) fluorescence microscope with a 203 magnification. Real Time Quantitative Polymerase Chain Reaction Total RNA was isolated from surgical specimens, TS, GBM-ADCs, U87MG, and HBMECs and reversely transcribed as already described (Calatozzolo et al., 2005). Fluorescent labeled probes were specific for MDR-1/ P-glycoprotein (Pgp), MRP-1, MRP-3, MRP-5, GST-p, CXCL12, VEGF), KDR, vascular endothelial (VE)- cadherin, CD34, CD133, and nestin (Assay on demand, Applied Biosystem). All primers and probe sets were controlled for sequence accuracy and, as a positive control, functionally tested with a pooled human cDNA (Applied Biosystem). All reactions were performed in duplicate with a negative control (no template) and the mean value of the threshold cycle (the start of exponential amplification) of each sample was normalized with the threshold cycle of Glyceraldehyde-3-phosphate dehydrogenase (GAPD), obtaining the DCt value: the higher the DCt, the lower the expression level. Data were analyzed with the Gene Amp 5700 software (Applied Biosystem). To make the results of RTQ-PCR clearer and more evident, we showed an mRNA expression index defined as (25-DCt) value. This was because the lowest GAPD threshold value was 15 in our analysis and the total cycles number was 40; thus, the highest possible DCt value of 25 indicated no mRNA expression. Fluorescence-Activated Cell Sorter Analysis Flow cytometry analysis was performed in cells from NSCs, TS, GBM-ADCs, ‘‘Secondary’’ TS, and the U87MG cell line. In order to obtain a single cell suspension, cells from NSCs, TS, GBM-ADCs or ‘‘Secondary’’ TS were disrupted (see ‘‘Immunocytochemistry’’ under ‘‘Materials and Methods’’); GBM-ADCs and U87MG grew adherent and were detached by trypsin. GLIA DOI 10.1002/glia

The cell suspensions were then incubated with albumin 4% for 15 min at 4°C to reduce nonspecific staining and cells were permeabilized and fixed 20 min at 4°C with BD Cytofix/Cytoperm solution (BD Biosciences). To investigate the expression of MRPs, cells were double stained with monoclonal antibodies antihuman MRP-1 (MRPr1, Alexis, San Diego, CA), MRP-3 (M3II-9, Alexis), MRP-5 (M5I-1, Alexis), GST-p (GST-p 353-10, Dako), MDR-1/Pgp (Pgp ,C1, Ylem, Roma, Italy) together with markers of NSCs or mature cells of the CNS such as antihuman GFAP (polyclonal rabbit, Dako), CD133/1 PE (Miltenyi), and nestin (R&D systems Minneapolis, MN 55413). Adhesion molecules were also investigated with anti human VCAM-1 (R&D systems Minneapolis, MN 55413), ICAM-1 (Caltag Laboratories Burlingame, CA 94010), ICAM-2 FITC (Bender Med Systems, Burlingame, CA 94010), a5b1 (Santa Cruz Biotechnology, CA 95060) PE-CAM (platelet endothelial cell adhesion molecule)/CD31-PE (Becton Dickinson, NJ 07417), aVb3 and aVb5 (Chemicon Temecula, CA 92590) and VE-cadherin (R&D systems Minneapolis, MN 55413). Finally, receptors of angiogenic factors were investigated with mouse antihuman CXCR4 and KDR (R&D systems Minneapolis, MN 55413). In several experiments mouse anti human b-III tubulin (Chemicon) and rabbit antihuman GFAP were used together to detect double positive cells. Secondary antibodies were goat antimouse IgG fluorescein-conjugated (BD Biosciences) for nestin, CXCR4, VCAM-1, VE-cadherin, a5b1, aVb3, MRP-3, GST-p, and Pgp, swine antirabbit biotin-conjugated (Code n°E0467, Dako, Denmark) and streptavidin PE (Becton Dickinson NJ 07417) for MRP-1 and MRP-5, and swine antirabbit TRITC-conjugated (Dako) for GFAP. Cells were analyzed by flow cytometry (FACStar Plus, BD Biosciences). Appropriate isotype controls were performed for each marker. Data were always expressed as percentage of positive cells and also as percent ratio between fluorescence intensity of a specific antibody (MRP-3, GST-p, and Pgp) and fluorescence intensity of the relative isotype control.

Proliferation Assay The assay was performed as already described (Salmaggi et al., 2004). Briefly, TS and NS were disrupted (see ‘‘Immunocytochemistry’’ under ‘‘Materials and Methods’’) and collected with PBS plus 2% bovine serum albumin (BSA), washed and plated on a substrate of matrigel. Cells were seeded at the concentration of 5,000 cells/ well in a 96-well multiplate. The growth medium was changed after 1 day and replaced with their medium plus 2% FBS without EGF and bFGF. Again the next day cells were stimulated for 3 days with the CXCL12, EGF, and bFGF (10 ng/mL). After 72 hr the culture medium was removed and cells were processed as already published (Salmaggi et al., 2004).

853

BRAIN TUMOR STEM-LIKE CELLS CHARACTERIZATION

Enzyme-Linked Immunosorbent Assays CXCL12 and VEGF proteins (pg/ml) were evaluated by ELISA performed following manufacturer’s instructions (Quantikine immunoassay for human CXCL12, R&D systems, code n° DSAOO and VEGF, code n° DVE00). Supernatants of TS at different passages (from 3rd to 15th) were used immediately after collection or frozen at 220°C until they were processed.

Rat Aorta Model of Angiogenesis Angiogenic growth can be evaluated in rat aortic rings producing spontaneously microvessels (Nicosia and Ottinetti, 1990). It is self-limited and can be modulated by a number of angiogenic factors such as VEGF which is the most powerful (Nicosia et al., 1994). Rings of rat (4 weeks old Fisher 344 male rats) aorta were cultured in type I collagen gel under serum-free conditions and newly formed microvessels were counted every day. Each ring was embedded in type I collagen (35 lL) containing TS1 and TS3 (about 4 spheres) and the same amount of GBM-ADCs 1 and GBM-ADCs 3 derived from the same human GBM of TS1 and TS3, respectively. After gel consolidation, the ring was cultured (at 37°C with 5% CO2) in the presence of Endothelial Basal Medium (EBM, Clonetics, Cambrex, Walkersville, MD USA) plus gentamicin (10 U/mL). Vessels outgrowth was checked for 1 week starting the count after 3 days and compared, as a negative control, with vessel outgrowth of the aortic ring embedded in collagen in the absence of TS. Medium was replaced every 3 days and our data express the mean value of 4 experiments. Anesthesia and surgery were performed following the NIH Guide for the Care and Use of Laboratory Animals, US National Research Council, 1996). Intracranial Cell Transplantation into NOD-SCID Mice An aliquot of 200,000 cells from TS (TS1 and TS2) were resuspended in 10 mls of PBS and injected stereotactically into the frontal lobe of 6-week old NOD-SCID mice following administration of general anesthesia; the injection stereotaxic coordinates of the inoculation point were 1.2 mm (x plane), 4 mm rostral (y plane), and 1.2 mm ventral (z plane) from the bregma. The bregma, recognized in mice by visual examination of the exposed skull, is the intersection of the coronal and sagittal sutures. After 5 weeks, mice were again subjected to general anesthesia and human tumor growing inside the brain was removed for histological analysis, cell culture (TS generation) or human nuclei staining. We refer to TS generated in NOD-SCID mice as ‘‘Secondary’’ TS.

Fig. 1. Panel A: Histopathologic features of the GBM grade IV generating TS. Peculiar histomorphologic areas showing highly packed tumor cells and pseudopalisades (magnification 320). Panel B: Spheres from GBM or TS. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Histological Analysis for Identifying Human Cells in the Grafted Neoplasm In order to identify the human ‘‘Secondary’’ TS derived cells, NOD-SCID mice were anesthetized and treated by transcardial perfusion, after open-chest surgery, with cold PBS followed by 4% PFA. The already fixed brain tissue was removed and again treated for 48 hr in 4% PFA. Fresh brain sections (30 lm thickness) were cut on vibratome (Zeiss) and permeabilized in PBS with 5% triton X-100 in the presence of 10% normal goat serum for 1 hr at 37°C. Then slides were free floating labeled overnight at 4°C with a specific mouse antihuman nuclei (1:50) (Chemicon MAB 1281) in order to identify human cells; next they were washed and then incubated with secondary antibody Cy-3-conjugated antimouse IgG (1:1000) (Jackson) for 1 hr at room temperature and then washed again and counter stained with 40-6-diamidino-2-phenylindole (DAPI) to identify cell nuclei (data not shown). Immunofluorescence measurements were made using a confocal microscopy (Leika system). GLIA DOI 10.1002/glia

854

SALMAGGI ET AL. TABLE 1. GBM Derived TS and GBM-ADCs FACS Analysis

Panel A CD133 Nestin CXCR4 Panel B MRP-1 MRP-3 MRP-5 Pgp GST Panel C ICAM-1 ICAM-2 VCAM-1 PE-CAM aVb5 aVb3 a5b1 VE-cadherin

NSCs

TS

GBM-ADCs

1 6 0.5 85 6 5 95 6 1

5 6 2.6 95 6 1 98 6 2.3

1.3 6 0.7 55 6 7 87 6 2

18 16 8 20 52

6 6 6 6 6

1 3.2 2.7 3.2 5

74 94 21 85 90

6 6 6 6 6

6 2 3 4.1 2

60 6 5 83 6 2 50 6 2 0 13 6 1 37 6 7 90 6 1 0

8 15 60 44 51

6 6 6 6 6

2 3 5 3 4

52 6 5 83 6 4 0 0 ND 35 6 5 2 6 0.5 ND

U87MG 0 90 6 2 100 6 5 63 53 75 64 78

6 6 6 6 6

2 1 5 4 3

15 6 2 53 6 3 11 6 1 0 ND 11 6 0.5 13 6 1 ND

Cytofluorimetric analysis of cell suspension, permeabilized and fixed, obtained from TS, NSCs, GBM-ADCs, and U87MG was performed with specific antibodies (‘‘Materials and Methods’’). Data were expressed as percentage of positive cells (mean value 6 SD). Results represent the average obtained from 3 independent experiments; in TS and GBM-ADCs, these values represent the average values obtained in three different patients.

Phenotype, Differentiation Capacity, and NSC Markers Expression of GBM-Derived TS

Fig. 2. TS immunocytochemistry (ICC). We show cells from NSCs and TS obtained following the technique described in the ‘‘Materials and Methods.’’ Cells b-III tubulin positive are shown in panels A (NSCs) and B (TS1); GFAP positive in panel C (NSCs) and D (TS1); nestin positive in panels E (NSCs) and F (TS1); Gal C positive in panels G (NSCs) and H (TS1) and finally CXCR4 positive in panels I (NSCs) and L (TS1). Nuclei appear colored in blue (DAPI nuclear-staining). Magnification was 320.

RESULTS Histological Analysis of the Human GBM Generating TS Human GBMs originating TS were routinely stained for haematoxylin-eosin. The staining was consistent with a grade IV malignant glioma showing palisades (Fig. 1; panel A) and extensive neovascularization of the tumor (data not shown). GLIA DOI 10.1002/glia

We used long term proliferating (10–15 passages) TS (Fig. 1; panel B) to check their ability to acquire markers of mature neural cells. As shown in Fig. 2, TS differentiation produces cells expressing mature CNS markers such as b-III tubulin (neurons, Fig. 2; panel B), GFAP (astrocytes, panel D) and GalC (oligodendrocytes, panel H). Normal NSCs, used as positive control, are also shown in Fig. 2 (panels A, C, and G). It is important to note that, as already reported by others (Hemmati et al., 2003; Tunici et al., 2004), GFAP and b-III tubulin positive cells were also present in undifferentiated TS culture. The expression of NSCs markers such as CD133, nestin and CXCR4 (the CXCL12 receptor) were analyzed by FACS in TS, NSCs and also in GBM (GBM-ADCs) and U87MG growing under adherent conditions; U87MG cells were used as a reference ‘‘standard’’ GBM cell line (see also ‘‘Materials and Methods’’). As shown in Table 1 (panel A) the expression of nestin and CXCR4 was significantly high in TS as well as in NSCs as also confirmed by ICC analysis (Fig. 2; panels F and E for nestin and L and I for CXCR4); U87MG was also strongly positive for both markers (Table 1; panel A) whereas GBM-ADCs were less positive for nestin. The percentages of positive cells for CD133 was higher in TS (5% 6 2.6%) than in NSCs (1% 6 0.5%) and in GBM-ADCs (1.3% 6 0.7%); whereas the U87MG cell line was completely negative (Table 1; panel A). Nestin, an NSC marker always expressed in CD1331 cells (Singh et al., 2004a), has the highest percentage of positive cells in TS compared with NSCs, GBM-ADCs, or U87MG and its expression was generally high in any case. When comparing gene expression of NSCs’ markers in TS, NSCs, GBM-ADCs, and U87MG line by RTQPCR, we found that CD133 was equally expressed in TS and NSCs, whereas GBM-ADCs and U87MG cells had a sig-

855

BRAIN TUMOR STEM-LIKE CELLS CHARACTERIZATION TABLE 2. GBM Derived TS and GBM-ADCs Gene Expression Evaluated by RTQPCR 25-DCtval NSCs Panel A CD133 Nestin CD34 Panel B MDR-1 MRP-1 MRP-3 MRP-5 GST-p Panel C CXCL12 VEGF KDR VE-cad

TS

16.8 6 0.7 12.4 6 1 21.3 6 1 23.5 6 1.5 15.0 6 0.4 17.4 6 0.5 15.6 20.1 15.4 24.3

0 6 6 6 6

0.5 1 0.6 0.8

18.0 19.2 11.5 8.9

6 6 6 6

0.5 3.5 6 0.4 0.7 19.4 6 2.3 0.3 10.5 6 2.5 0.1 15.7 6 1.1

16.5 19.6 15.6 23.2

0 6 6 6 6

2 1.1 1.1 1.1

GBM-ADCs

U87MG

2.6 6 0.8 21.8 6 1.2 9.8 6 1

1.5 6 0.3 12.1 6 0.7 6.5 6 0.5

8.4 6 0.4 18.2 6 1.3 19.8 6 0.7 15.1 6 0.4 ND

17.8 15.0 13.9 24

9.4 20.5 16.6 15.0

6 6 6 6

0.8 1 0.9 0.4

0 6 6 6 6

HBMECs

0.5 0.6 0.3 1.4

9.7 6 0.3 20.5 20.2 6 0.6 18.8 14.6 6 0.8 21.8 0 23.1

6 6 6 6

0.8 0.6 0.4 1

25-DCt value: expressing the amount of mRNA gene-specific (mean value 6 SD) as described in ‘‘Materials and Methods,’’ RTQPCR paragraph; NSCs, TS, GBMADCs, U87MG, and HBMECs see ‘‘Materials and Methods,’’ cell culture paragraph. Results represent the average obtained from 3 independent experiments; in TS and GBM-ADCs, these values represent the average values obtained in three different patients. The Applera specific probes are already described elsewhere and ND is not done.

nificantly lower expression. Nestin was expressed at high levels in all different samples. We also investigated the gene expression of CD34, another putative stem marker expressed in brain tumors (Chaubal et al., 1994). We observed that TS expressed the highest level of this marker, followed by NSCs, GBM-ADCs, and U87MG (see Table 2; panel A). This result seems to suggest that TS can be more heterogeneous in the expression of stem cell markers compared with GBM-ADCs and NSC, where even hemathopietic stem cell marker such as CD34 can be detected. Summarizing these results, all the markers of staminality checked seem to be upregulated in the different GBM derived TS preparations. In Vivo Tumorigenic Potential of TS In order to verify the tumorigenic potential of TS, we next implanted in the frontal lobe of NOD-SCID mice around 200,000 cells obtained from two different GBMs, TS1 (15th passage) and TS2 (10th passage). Animals were followed clinically with daily examination for neurological signs. The expected time point for the appearance of neurological signs was actually 7 weeks as suggested by the literature (Galli et al., 2004). The NOD-SCID mice injected with TS, but not those injected with NSCs used as control cells (Galli et al., 2004), showed neurological disturbance starting at 5 weeks after transplantation. Animals showed signs of tetraparesis and marked cyphosis, hardly being able to move at that time. At 6– 7 weeks mice were sacrificed and brains dissected. After fixation, brains were stained with haematoxylineosin to investigate the presence of newly generated brain tumor. Formation and histopathology of the human brain tumor is shown in Fig. 3 (panel A); the human origin was demonstrated by the positive staining with anti human

Fig. 3. IHC of the human GBM grafted into NOD-SCID mouse. IHC of the tumor obtained in 5 weeks by transplant of GBM derived TS cells (panel A). Presence of human cells in the neoplasm and infiltrating the mouse brain identified with antihuman nuclei staining (panel B).

nuclei, which identified cells positive to human marker in the slides (Fig. 3; panel B). Moreover human cells were present not only in the area of development of the engrafted tumor but also in the surrounding tissue, which suggested an infiltrative pathway. Angiogenic Profile of GBM-Derived TS We first checked if TS were able to release angiogenic factors in the culture supernatants and, indeed, TS were able to release VEGF and CXCL12 as assessed by ELISA test. In TS, release of VEGF increased and release of CXCL12 started to be detected after repeated in vitro passages (see Fig. 4). Our positive control, i.e. the U87MG cell line, a well known high producer of VEGF (Xu et al., 2002), released nearly as much VEGF as TS did at its maximal level. At low concentration, CXCL12 stimulated proliferation of NSCs while GLIA DOI 10.1002/glia

856

SALMAGGI ET AL.

highly positive for ICAM-1 and ICAM-2 while U87MG showed a lower expression. VCAM-1 was significantly expressed in TS and to a lesser extent in U87MG but absent in GBM-ADCs, whereas the cell–cell contact adhesion molecules PECAM and VEcadherin were not found in the different samples (Table 1; panel C). The analysis of avb5, avb3, and a5b1 showed a significant expression of these integrins only in TS. In particular a5b1was present at high level on TS (90% 6 1%) followed by avb3 (37% 6 7%) and avb5 (13% 6 1%); the GBM-ADCs expressed only avb3 at a significant level, while U87MG expressed very low levels of these integrins. Fig. 4. CXCL12 and VEGF release by TS. VEGF (empty bars) and CXCL12 (filled bars) was measured by ELISA (pg/mL) in supernatants of TS1 and U87. TS1 were analyzed at different passages in culture. Results are the average 6 SD of three experiments.

higher concentration was required to stimulate TS (data not shown). The proliferative activity of CXCL12 is not surprising given the high expression of its receptor CXCR4, in TS and NSCs (Table 1; panel A and Fig. 2 panels I and L, respectively). On the contrary, by FACS analysis, KDR, the VEGF receptor type 2, was positive in a small percentage of TS (data not shown). Moreover, we performed a molecular phenotype of TS analyzing gene-specific mRNA of selected molecules of angiogenic relevance (Table 2; panel C). The expression of gene encoding for VEGF was high in all samples. CXCL12 was expressed more in NSCs than in TS, with intermediate values in GBM-ADCs. HBMECs were included as positive control. Aortic Ring Assay in GBM-Derived TS and GBM-ADCs Angiogenesis can be studied ex vivo in a rat aortic ring assay. In order to test if TS and their adherent counterpart GBM-ADCs were able to functionally interact with healthy vessels, we co-cultured both in collagen gel with rat aortic rings. A massive cell spreading from the TS core occurred after 3 days of culture in the presence of aortic rings (Fig. 5; panel B) when compared with control (i.e. TS cultured without aortic rings; Fig. 5; panel A). The number of vessels of the aortic rings increased in the presence of TS (Fig. 6; panels A and B and Fig. 5; panel D). Interestingly, GBM-ADCs was able to induce a huge increase of the number of vessels outgrowing from the aortic rings (Fig. 6; panels A and B). Apparently, neo-vessels growing in the presence of TS were longer and thinner compared with our control (Fig. 5; panel C). Adhesion Molecules and Integrins Expression of TS TS, GBM-ADCs and U87MG were checked for the expression of adhesion molecules by FACS analysis (Table 1; panel C). We found that TS and GBM-ADCs were GLIA DOI 10.1002/glia

MDR Phenotype of GBM-Derived TS and GBM-ADCs In order to better understand whether TS could allow recovery and proliferation after chemotherapeutic treatment and to evaluate the possible role played by GBM stem-like cells in chemoresistance, we analyzed by FACS the expression of the MRPs proteins (MRP-1-3-5) and GST-p in single cells obtained from TS and, as a control, from NSCs, GBM-ADCs and U87MG. Overall, MRP-1, MRP-3, Pgp, and GST were all highly expressed (MRP-3 the highest) by TS, followed by U87MG, while their expression was variable (but consistently lower) in NSCs and GBM-ADCs; on the contrary, MRP-5 was more expressed in U87MG than in TS and NSCs . GBM-ADCs always showed massive downregulation of MDR proteins with the only exception of MRP-5, which starting from a very low percentage of positive TS cells, was three times higher in GBM-ADCs (see Table 1; panel B). When we compared the analysis of mean fluorescence intensity (MFI) of Pgp, MRP-3, and GST-p performed in cells from TS with cells from whole brain tumors (WBT) previously reported (Calatozzolo et al., 2005), we found the following differences: Pgp (241% WBT versus 524% TS1), MRP-3 (170% WBT versus 493% TS1), and GST-p (214% WBT versus 941% TS1). We then evaluated by RTQPCR the expression of the MDR-1 gene encoding for Pgp and we found that this gene was weakly expressed in GBM-ADCs and negative in spheres (both NSCs and TS). The GBM cell line U87MG was also negative. These data were in contrast to RT-PCR published data (Galli et al., 2004) and our FACS analysis and we believe our molecular and protein data do not recognize the same target. Concerning MRP-1-3-5 and GST-p, RTQPCR analysis showed good concordance with cytofluorimetric data concerning TS in which MRP-3 is absolutely the most expressed. On the other hand, in GBM-ADCs these molecules were poorly expressed as detected by FACS analysis, but mRNA levels were not significantly affected by adherence (see Table 2; panel B). ‘‘Secondary’’ TS are CD1331 and Maintain Strong Angiogenic Potential We then grew the brain tumor obtained after sacrifice of the NOD-SCID mouse injected with TS1 in culture.

BRAIN TUMOR STEM-LIKE CELLS CHARACTERIZATION

857

Fig. 5. Angiogenesis induced in aortic ring by TS (Magnification 340). TS spreading in absence (panel A) or presence (panel B) of aorta vessels after 3 days of culture. Density and morphology of aortic ring vessels in the absence (panel C) or presence (panel D) of TS after 7 days of culture. [Color figure can be viewed in the online issue, which is available at www. interscience.wiley.com.]

CD1331 (92% see Fig. 7; panel A) and, like TS1, positive for CXCR4 and nestin in a percentage close to 100%. GFAP, expressed on mature astrocytes, was negative in ‘‘Secondary’’ TS cells suggesting altogether a truly immature phenotype of the newly generated tumor. Moreover, ‘‘Secondary’’ TS cells were all ICAM-1 positive (not shown) and preserved the angiogenic potential of TS1 as evaluated by the rat aortic ring assay and shown in Fig. 7 panel B; the mRNA specific for VEGF was 22 in ‘‘Secondary’’ TS versus 18 in TS1 (value expressed as 25-DCt; see ‘‘Real Time Quantitative Polymerase Chain Reaction’’ under ‘‘Materials and Methods’’) and MRP-3 was still the most expressed of the MDR genes. DISCUSSION

Fig. 6. Angiogenesis induced by TS and GBM-ADCs in rat aortic ring assay (vessel density). Number of vessels in the absence (negative control) or presence of TS and GBM-ADCs cells. TS 1 and GBM-ADCs 1 in panel A and TS 3 and GBM-ADCs 3 in panel B. [Color figure can be viewed in the online issue, which is available at www.interscience. wiley.com.]

Cultured cells originated TS in 1 day and will be defined as ‘‘Secondary’’ TS. We analyzed by FACS analysis ‘‘Secondary’’ TS cells and to our great surprise, almost all of the cells were

GBM is a very aggressive primary brain tumor resulting in a short life expectancy for patients after diagnosis and with a grim prognosis despite maximal treatment (Stupp et al., 2005). It has been recently suggested that tumorigenicity and resistance to therapy might not be a feature of all the cells of the bulk tumoral mass but rather of a selected population of cells with stem-like phenotype (Galli et al., 2004; Singh et al., 2004b). We obtained GBM-derived TS and further characterized them for the expression of NSCs markers and their capability of expressing markers of mature CNS cells; we found an altered differentiation pathway with promiscuous expression of astrocytic and neuronal markers as already described (Galli et al., 2004). Concerning NSCs markers, we found that TS express nestin and CXCR4 in the majority of the cells and CD133 in only 5% of the same cells. GLIA DOI 10.1002/glia

858

SALMAGGI ET AL.

Fig. 7. CD133 expression and angiogenic properties of ‘‘Secondary’’ TS. FACS analysis showing CD1331 cells from ‘‘Secondary’’ TS (panel A). Rat aortic ring assay performed with ‘‘Secondary’’ TS exactly as previously reported (see ‘‘Materials and Methods’’) (panel B). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Our finding is in partial agreement with a previous publication (Singh et al., 2004b) in which, however, a higher level of CD133 expression in GBM was reported. This partial discrepancy could be related to the different expression of CD133 in the fresh bulk tumor mass and TS undergoing serial subcultivation. Moreover, TS downregulated CD133 with passages and GBM-ADCs completely lost CD133 by adherence and presence of serum. Actually all the NSCs markers checked (CD34, nestin and CXCR4) were down regulated in GBM-ADCs, thus confirming that stem cell-like GBM-derived cells are more represented in TS as compared with GBM-ADCs. We further checked whether this selected population of stem-like cells from human GBM displayed potential pro-angiogenic activity, adhesion molecule expression, and chemoresistance: all features likely responsible for GBM relapse and progression. Concerning the pro-angiogenic phenotype, when we compared TS cells with GBM-ADCs in the aortic ring assay we found that a higher number of vessels (see Fig. 6) was induced by GBM-ADCs, suggesting that serum and adhesion conditions trigger the expression of pro-angiogenic factors to some extent. TS were also able to release VEGF in the culture medium and we found an increased production and release of angiogenic cytokines after time and passages in culture which eventually exceeded the high release by GLIA DOI 10.1002/glia

U87MG (Xu et al., 2002). Nevertheless, TS cells are described as stable at least when regarding karyotype and cellular or molecular phenotype (Galli et al., 2004). TS cells are therefore high producers of VEGF which they most probably utilize for the surrounding cell population rather than for themselves since indeed, their KDR is quite low or absent. When investigating the CXCL12/CXCR4 ligand/receptor system, we found that CXCL12 is released in small amounts by TS in the supernatants and like VEGF, increases after passages in culture; whilst the expression of its receptor CXCR4 is extremely high and close to the totality of TS cells. TS cells could therefore be a likely target of CXCL12 released in close proximity by other cells rather than the main producers of this chemokine. Actually, in other neoplastic lesions such as breast cancer, fibroblasts within invasive breast carcinomas contribute to tumor progression in a large part through the secretion of CXCL12 (Orimo et al., 2005). The presence of CXCL12 in the endocavitary fluids of brain tumors and in the supernatants of tumor endothelial cell cultures has already been reported (Salmaggi et al., 2004), suggesting that other cells may support growth of malignant cells in brain tumors exactly like stroma cells do in breast cancers. The relevance of CXCL12/CXCR4 interaction in malignant glioma growth/invasion is further supported by the evidence that a small-molecule antagonist of CXCR4 is able to inhibit tumor growth in an in vivo model (Rubin et al., 2003) and by the proliferative response to CXCL12 of TS cells shown by our data for the first time; also, GBM cells expressing high levels of CXCR4 show increased invasive ability and response to a chemiotactic gradient of CXCL12 (Ehtesham et al., 2006). Although we did not perform inhibition experiments, it is likely that VEGF/CXCL12 produced by TS/GBM-ADCs play a major role in the angiogenic response shown by the sprouting of neo-vessels. It has to be remembered that CXCL12 can not only be considered a survival and proliferation factor but also a chemotactic one, able to mobilize NSCs (Imitola et al., 2004) from their site of location to the neoplasm as well as able to promote angiogenesis by recruiting endothelial progenitor cells (EPCs) (Mohle et al., 1998). The totality of TS and NS cells was roughly found positive for CXCR4 by FACS analysis but we emphasize that TS cells were all positive. These data explain how these cells not only grow and recruit other cells in the tumor but also provide support for understanding the mechanism(s) of development of the angiogenic network needed for the expansion and infiltration of the neoplasm. This pro-angiogenic potential is fully expressed by more differentiated GBM-derived cells grown in adherent conditions. We can summarize the data concerning the angiogenic phenotype of TS suggesting that vessel growth is certainly a requisite of mature cells more than of stem cells. The extensive infiltrative nature of GBM led us to analyze the adhesion pathway in TS; adhesion molecules

859

BRAIN TUMOR STEM-LIKE CELLS CHARACTERIZATION

were always more expressed in TS than in U87MG. Some of them, like ICAM-2 and a5b1, were positive in almost the totality of TS cells checked. The interaction between ICAM-2 and LFA-1, the latter one expressed on immune cells, is a priority of HBMECs but certainly not limited to these cells; indeed, TS cells in our data and also microglia cells have been shown to express LFA-1 (Mizuno et al., 1999). Concerning a5b1, the molecule that is strongly expressed on TS, is responsible for contacts with ECM proteins of the brain. Other CAMs affect microvascular proliferation, a histopathological hallmark of GBM, like platelet endothelial cell adhesion molecule 1 (PECAM-1/CD31) and vascular endothelial (VE)-cadherin (Aroca et al., 1999). These molecules were also examined in TS cells but were found negative. In order to verify if GBM derived stem-like cells could represent a sort of primary therapeutic target, we investigated their expression of chemoresistance-related molecules such as MRP-1, 3, 5, Pgp and GST-p. MDR proteins were already shown, in our previously reported manuscript, to be highly expressed in gliomas but their expression was lower in the whole group of glioma samples analyzed at surgery (>40) (Calatozzolo et al., 2005) once compared with that of TS cells investigated in the present work. A very high expression of MDR resistance molecules by TS cells was found by assessing the percentage of positive cells in comparison with our controls (NSCs, GBM-ADCs, and U87MG). MRP-3 can be considered a hallmark of GBM derived TS since all the data we produced clearly indicated its overexpression once compared with our controls. MRP-1, GST, and Pgp complete the pattern of MDR molecules strongly expressed by TS whereas MRP-5 was failing. Our data partially support the idea that few cells in the bulk tumoral mass could indeed be highly resistant to chemotherapy. However, it has recently been reported, in vitro, that increased resistance to apoptosis (rather than higher activity of chemoresistance proteins) may be a feature of the resistance of GBM-derived stem cells to chemotherapy (Eramo et al., 2006). Further data are therefore needed to fully clarify this issue. Another important point concerns TS cells following their grafting. Several phenotypic changes occurred in TS cells following their grafting in the forebrain of NODSCID mice; the tumor derived after transplantation in NOD-SCID mice showed, by FACS analysis, elevated CD133 expression suggesting further selection of cancer stem cells, in contrast to the low percentage of CD1331 cells present in the original TS cell population. CD1331 cells further decreased over passages in culture before grafting. We cannot explain the previous downregulation in vitro of CD133 and its absolute upregulation in vivo after grafting, but we find the data intriguing. We speculate that, as recently described (Singh et al., 2004b), the initial ‘‘in vivo’’ tumoral growth could be warranted by cells expressing this marker; the marker might eventually be lost after passages ‘‘in vitro’’. The low expression

of GFAP also suggests the truly immature phenotype of human TS cells grown in NOD-SCID mice. Moreover, we expected ‘‘in vitro’’ TS formation from the grafted neoplasm but the time required (only 1 day) was indeed very short when considering that the formation of TS needs about 30–40 days (Galli et al., 2004) when starting from the original human tumor. The mice-derived tumor was characterized by the presence of human cells, its exceptional capability of forming spheres and its angiogenic potential. In conclusion, our data underline the presence of a restricted population of cells isolated from human GBM, with characteristics of potential pro-angiogenic activity, of adhesion molecule expression and chemoresistance, that are able to rebuild the tumor. Due to their chemoresistance, we consider TS cells as the major candidate involved in the persistence of minimal residual disease after treatments, whereas adherent cells originating from TS cells might be more involved in enhancing tumor progression because of their elevated angiogenic properties. Selectively targeted therapeutic approaches should provide further information on the clinical relevance of this cell population in brain tumor behavior. ACKNOWLEDGMENTS We thank Andrea Smith for helping us with the English draft of this article and Prof. Giovanni Broggi and colleagues of the Neurosurgery Department of the National Neurological Institute C. Besta, Milan, Italy for their collaboration. REFERENCES Al-Hajj M, Wicha MS, Benito-Hernandez A, Morrison SJ, Clarke MF. 2003. Prospective identification of tumorigenic breast cancer cells. Proc Natl Acad Sci USA 100:3983–3988. Aroca F, Renaud W, Bartoli C, Bouvier-Labit C, Figarella-Branger D. 1999. Expression of PECAM-1/CD31 isoforms in human brain gliomas. J Neurooncol 43:19–25. Bellail AC, Hunter SB, Brat DJ, Tan C, Van Meir EG. 2004. Microregional extracellular matrix heterogeneity in brain modulates glioma cell invasion. Int J Biochem Cell Biol 36:1046–1069. Bez A, Corsini E, Curti D, Biggiogera M, Colombo A, Nicosia RF, Pagano SF, Parati EA. 2003. Neurosphere and neurosphere-forming cells: Morphological and ultrastructural characterization. Brain Res 993:18–29. Bonnet D, Dick JE. 1997. Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nat Med 3:730–737. Calatozzolo C, Gelati M, Ciusani E, Sciacca FL, Pollo B, Cajola L, Marras C, Silvani A, Vitellaro-Zuccarello L, Croci D, Boiardi A, Salmaggi A. 2005. Expression of drug resistance proteins Pgp, MRP1, MRP3, MRP5 and GST-p in human glioma. J Neuroncol 74:113–121. Chang SM, Parney IF, Huang WAnderson FA Jr, Asher AL, Bernstein M, Lillehei KO, Brem H, Berger MS, Laws ER. 2005.For the glioma outcomes project investigators. Patterns of care for adults with newly diagnosed malignant glioma. JAMA 293:557–564. Chaubal A, Paetau A, Zoltick P, Miettinen M. 1994. CD34 immunoreactivity in nervous system tumors Acta Neuropathol (Berl) 88:454–458. Ehtesham M, Winston JA, Kabos P, Thompson RC. 2006. CXCR4 expression mediates glioma cell invasiveness. Oncogene 25:2801–2806. Eramo A, Ricci-Vitiani L, Zeuner A, Pallini R, Lotti F, Sette G, Pilozzi E, Larocca LM, Peschle C, De Maria R. 2006. Chemotherapy resistance of glioblastoma stem cells. Cell Death Differ 13:1238–1241. Felding-Habermann B, Cheresh DA. 1993. Vitronectin and its receptors. Curr Opin Cell Biol 5:864–868.

GLIA DOI 10.1002/glia

860

SALMAGGI ET AL.

Fischer I, Gagner JP, Law M, Newcomb EW, Zagzag D. 2005. Angiogenesis in gliomas: Biology and molecular pathophysiology. Brain Pathol 15:297– 310. Galli R, Binda E, Orfanelli U, Cipelletti B, Angela Gritti A, De Vitis S, Fiocco R, Foroni C, Dimeco F, Vescovi A. 2004. Isolation and characterization of tumorigenic, stem-like neural precursors from human glioblastoma. Cancer Res 64:7011–7021. Goldbrunner RH, Bernstein JJ, Tonn JC. 1998. ECM-mediated glioma cell invasion. Microsc Res Tech 43:250–257. Habgood MD, Begley DJ, Abbott NJ. 2000. Determinants of passive drug entry into the central nervous system. Cell Mol Neurobiol 20: 231–253. Hemmati HD, Nakano I, Lazareff JA, Masterman-Smith M, Geschwind DH, Bronner-Fraser M, Kornblum HI. 2003. Cancerous stem cells can arise from pediatric brain tumors. Proc Natl Acad Sci USA 100: 15178–15183. Imitola J, Raddassi K, Park KI, Mueller FJ, Nieto M, Teng YD, Frenkel D, Li J, Sidman RL, Walsh CA, Snyder EY, Khoury SJ. 2004. Directed migration of neural stem cells to sites of CNS injury by the stromal cell-derived factor 1a/CXC chemokine receptor 4 pathway. Proc Natl Acad Sci USA 101:18117–18122. Mizuno T, Yoshihara Y, Kagamiyama H, Ohsawa K, Imai Y, Kohsaka S, Mori K. 1999. Neuronal adhesion molecule telencephalin induces rapid cell spreading of microglia. Brain Res 849:58–66. Mohle R, Bautz F, Rafii S, Moore MA, Brugger W, Kanz L. 1998. The chemokine receptor CXCR-4 is expressed on CD341 hematopoietic progenitors and leukemic cells and mediates transendothelial migration induced by stromal cell-derived factor-1. Blood 91:4523–4530. Nicosia RF, Nicosia SV, Smith M. 1994. Vascular endothelial growth factor, platelet-derived growth factor, and insulin-like growth factor-1 promote rat aortic angiogenesis in vitro. Am J Pathol 145:1023–1029. Nicosia RF, Ottinetti A. 1990. Growth of microvessels in serum-free matrix culture of rat aorta: A quantitative assay of angiogenesis in vitro. Lab Invest 63:115–122. Orimo A, Gupta PB, Sgroi DC, Arenzana-Seisdedos F, Delaunay T, Naeem R, Carey VJ, Richardson AL, Weinberg RA. 2005. Stromal fibroblasts present in invasive human breast carcinomas promote tumor growth and angiogenesis through elevated SDF-1/CXCL12 secretion. Cell 121:335–348.

GLIA DOI 10.1002/glia

Parati EA, Bez A, Ponti D, de Grazia U, Corsini E, Cova L, Sala S, Colombo A, Alessandri G, Pagano SF. 2002. Human neural stem cells express extra-neural markers. Brain Res 925:213–221. Reya T, Morrison SJ, Clarke MF, Weissman IL. 2001. Stem cells, cancer, and cancer stem cells. Nature 415:105–111. Rubin JB, Kung AL, Klein RS, Chan JA, Sun Y, Schmidt K, Kieran MW, Luster AD, Segal RA. 2003. A small-molecule antagonist of CXCR4 inhibits intracranial growth of primary brain tumors. Proc Natl Acad Sci USA 100:13513–13518. Salmaggi A, Gelati M, Pollo B, Frigerio S, Eoli M, Silvani A, Broggi G, Ciusani E, Croci D, Boiardi A, De Rossi M. 2004. CXCL12 in malignant glial tumors: A possible role in angiogenesis and cross-talk between endothelial and tumoral cells. J Neurooncol 67:305–317. Singh SK, Clarke ID, Hide T, Dirks PB. 2004a. Cancer stem cells in nervous system tumors. Oncogene 23:7267–7273. Singh SK, Clarke ID, Terasaki M, Bonn VE, Hawkins C, Squire J, Dirks PB. 2003. Identification of a cancer stem cell in human brain tumors. Cancer Res 63:5821–5828. Singh SK, Hawkins C, Clarke ID, Squire JA, Bayani J, Hide T, Henkelman RM, Cusimano MD, Dirks PB. 2004b. Identification of human brain tumour initiating cells. Nature 432:396–401. Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJ, Belanger K, Brandes AA, Marosi C, Bogdahn U, Curschmann J, Janzer RC, Ludwin SK, Gorlia T, Allgeier A, Lacombe D, Cairncross JG, Eisenhauer E, Mirimanoff RO for the European Organization for Research and Treatment of Cancer Brain Tumor and Radiotherapy Groups;National Cancer Institute of Canada Clinical Trials Group. 2005. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med 352:987–996. Tews DS, Nissen A, Kulgen C, Gaumann AK. 2000. Drug resistanceassociated factors in primary and secondary glioblastomas and their precursor tumors. J Neurooncol 50:227–237. Tunici P, Bissola L, Lualdi E, Pollo B, Cajola L, Broggi G, Sozzi G, Finocchiaro G. 2004. Genetic alterations and in vivo tumorigenicity of neurospheres derived from an adult glioblastoma. Mol Cancer 6:3–25. Xu L, Fukumura D, Jain RK. 2002. Acidic extracellular pH induces vascular endothelial growth factor (VEGF) in human glioblastoma cells via ERK1/2 MAPK signaling pathway: Mechanism of low pHinduced VEGF. J Biol Chem 277:11368–11374.