Tumor tissues from a human GBM surgical specimen, which were collected from a 23-year-old man in The First Affiliated Hospital of Soochow University (Suzhou, China), were washed, deprived of vessels, acutely dissociated in phosphate buffered saline (PBS) and subjected to enzymatic dissociation. The patient provided written informed consent to participate in the study, which was approved by the Ethics Review Board of the First Affiliated Hospital of Soochow University (no. 2012070). Cells were cultured in high glucose Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum (FBS) for ~2 months, then the glioma cells were subsequently placed in serum-free DMEM/F12 medium supplemented with 1% N2 (Gibco; Thermo Fisher Scientific Waltham, MA, USA), 20 ng/ml epidermal growth factor [EGF (Invitrogen; Thermo Fisher Scientific)], and 20 ng/ml basic fibroblast growth factor [bFGF (Invitrogen; Thermo Fisher Scientific)] for ~7 days and formed non-adhesive neurospheres. Neurospheres were maintained by changing half of the medium every 3 days and collected by centrifugation at 1,000 × g for 10 min. Subspheres were formed for 3~4 days after primary spheres were dissociated mechanically to single cell suspension. Neurospheres of ~12 passages were used for sorting. Magnetic isolation of CD133⁻/A2B5⁺ GICs population was performed using the Miltenyi Biotec A2B5 and CD133 Cell Isolation kit (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany). Cells were cultured at 37°C in a humidified 5% CO₂/95% air atmosphere.

Tumor spheres were plated onto poly-L-lysine-coated glass coverslips. Cells were fixed with 4% paraformaldehyde for 20 min at room temperature, washed three times with PBS, blocked with 2% goat serum (Boster Bio-Engineering, Wuhan, China) for 30 min and permeabilized with 0.1% Triton X-100 (Beyotime Institute of Biotechnology, Shanghai, China). Cells were incubated with primary monoclonal anti-human mouse CD133 (1:200; cat. no. 130-090-422; Miltenyi Biotec GmbH), monoclonal anti-human rabbit nestin (1:300; cat. no. ab105389; Abcam, Cambridge, MA, USA) and monoclonal anti-human mouse A2B5 (1:200; cat. no. ab53521; Abcam) antibodies overnight at 4°C. Subsequently, cells were washed three times with PBS. Secondary anti-mouse IgGs conjugated with Alexa fluor 555 (1:1,000; cat. no. A-21422; Molecular Probes; Thermo Fisher Scientific) or anti-rabbit IgGs conjugated with Alexa fluor 488 (1:1,000; cat. no. A-11008; Molecular Probes; Thermo Fisher Scientific) were used in darkness for 30 min at room temperature. The nuclei were counterstained with anti-fade sealant containing 4′6-diamidino-2-phenylindole (DAPI) (5 µg/ml; SouthernBiotech, Birmingham, AL, USA). Fluorescence images were captured using a fluorescence microscope (Olympus BX40; Olympus Corporation, Tokyo, Japan).

GICs were dissociated into single cell suspension, then incubated with PE-conjugated anti-human mouse IgG A2B5 antibody (1:100; Miltenyi Biotech GmbH) and APC-conjugated anti-human mouse IgG CD133 antibody (1:100; Miltenyi Biotech GmbH). Labeled cells were analyzed by a flow cytometer Beckton Dickinson FACScan (BD Biosciences, San Jose, CA, USA) (13). Unsorted GICs were also analyzed as a control.

GICs were differentiated into non-initiating tumor cells in a high glucose DMEM of 10% FBS without N2/EGF/bFGF for 28 days. The cells were blocked with normal goat serum for 30 min. Next, the cells were incubated with monoclonal anti-human rabbit SOX2 (GSC marker; 1:200; cat. no. ab97959; Abcam), monoclonal anti-human mouse Tuj1 (neuron marker; 1:200; cat. no. ab14545; Abcam), monoclonal anti-human mouse GFAP (astrocyte marker; 1:300; cat. no. ab10062; Abcam) and monoclonal anti-human mouse Galc (oligodentrocyte marker; 1:100; cat. no. ab125086; Abcam) primary antibodies overnight at 4°C. Secondary anti-mouse IgGs conjugated with Alexa fluor 555 (1:1,000; cat. no. A-21422; Molecular Probes; Thermo Fisher Scientific) or anti-rabbit IgGs conjugated with Alexa fluor 488 (1:1,000; cat. no. A-11008; Molecular Probes; Thermo Fisher Scientific) were used in darkness for 30 min at room temperature. The nuclei were counterstained with DAPI (5 µg/ml). Fluorescence images were captured using a fluorescence microscope (Olympus BX40; Olympus Corporation) to confirm that differentiation had occurred.

The in vitro cell invasion assay was performed using a Matrigel-coated (3 mg/ml) invasion chamber (BD Biosciences), while the migration assay was performed by control inserts. Cell culture medium supplemented with 10% FBS was added to the lower chamber to act as a chemoattractant. Cells were seeded at a density of 1×10⁵ cells/well onto the upper inserts. After incubation for 24 h at 37°C, the non-migratory or non-invasive cells were removed from the upper side of the filter by gentle scrubbing with a moist cotton swab. The migrating or invading cells in the reverse side of the filter were fixed in 100% methanol and stained with crystal violet (Beyotime Institute of Biotechnology), then counted under an inverted microscope (Olympus CKX41; Olympus Corporation) in 6 fields at random for analysis.

Cells were transfected with red fluorescence protein (RFP) gene using a lentivirus mediated gene transfection kit (Shanghai GenePharma Co.,Ltd., Shanghai, China) according to the manufacturer's instructions. More than 90% of GICs and non-initiating tumor cells expressed RFP under fluorescence microscope (Olympus IX51; Olympus Corporation), the RFP gene integrated stably in target cell genome and maintained a stable RFP expression in the transfected tumor cells. Invasive or migratory potential of GICs-RFP and non-initiating tumor cells-RFP cells in vitro was detected by matrigel assay or polycarbonate filter, respectively, and was observed directly under fluorescence microscope (Olympus IX51; Olympus Corporation).

A total of 3×10⁵ cells/well were seeded into 6-well plates in triplicate and incubated at 37°C for 24 h. The cells were lysed, and total RNA was isolated using TRIzol reagent (Invitrogen; Thermo Fisher Scientific) according to the manufacturer's instructions. First strand cDNA synthesis was performed using reverse transcriptase and random hexanucleotide primers (Thermo Fisher Scientific, Inc., Waltham, MA, USA). The complementary DNA was subsequently used to perform qPCR on a LightCycler480 Thermal Cycler (Roche Diagnostics, Basel, Switzerland) using SYBR Green (Molecular Probes; Thermo Fisher Scientific) with gene-specific primers (Gene Pharma, Shanghai, China) and JumpStart™ Taq DNA polymerase (Invitrogen; Thermo Fisher Scientific). The crossing threshold value was normalized to GAPDH, and quantitative changes in mRNA were expressed as fold-change relative to the control ± standard error of the mean (SEM) value.

Cell extracts were prepared by lysing cells in RIPA buffer containing a mixture of protease and phosphatase inhibitor cocktails (Thermo Fisher Scientific, Inc.) followed by sonication and centrifugation at 10,000 × g for 10 min at 4°C. Protein concentration was determined using a BCA protein assay (Pierce Biotechnology, Inc., Rockford, IL, USA). A total of 50 µg of protein was loaded onto a 10% SDS-PAGE gel and transferred to nitrocellulose membranes (EMD Millipore, Billerica, MA, USA). The membranes were blocked in 5% non-fat dry milk for 1 h and probed overnight with primary antibodies at 4°C. Subsequently the membranes were incubated with secondary antibodies conjugated with horseradish peroxidase for 1 h at room temperature. The blots were visualized using enhanced chemiluminescence (ECL) reagents (Invitrogen; Thermo Fisher Scientific) and densitometric quantification was analyzed using Launch Sensi Ansys software (Shanghai Peiqing Science & Technology Co., Ltd., Shanghai, China). The primary antibodies including anti-epithelial cadherin (E-cadherin, Santa Cruz Biotechnology, Inc., Dallas, TX, USA), anti-intercellular adhesion molecule 1 (ICAM-1, Santa Cruz Biotechnology, Inc.), anti-matrix metalloproteinase 2 (MMP-2, Abcam), anti-MMP-9 (Abcam) and anti-tissue inhibitor of metalloproteinase 3 (TIMP3, Abcam) were used. β-actin (Santa Cruz Biotechnology, Inc.) expression was evaluated as a control for protein loading.

BALB/c mice (Shanghai Laboratory Animal Center, CAS, Shanghai, China) were divided into 2 groups, and each consisted of 8 mice. Animal care and experimental procedures described in the study were performed in accordance with the Guidelines for Animal Experiments at Soochow University from Institutional Animal Care and Use Committee with the approval of Ethics Committee of the university [Certificate no. SYXK (Su) 2007-0035]. Athymic/nude immunocompromised mice at 6–8 weeks of age were maintained under pathogen-free conditions within the institutional animal facility. Food and water were provided ad libitum. Surgical procedures were performed in a sterile fashion. Mice were anesthetized by intraperitoneal injection of chloral hyfrate (400 mg/kg), then positioned in a rodent stereotaxic frame. A total of 2×10⁵ CD133⁻/A2B5⁺ GICs or matched non-initiating tumor cells were stereotactically injected into the right putamen (1 mm forward, 2 mm right lateral from the bregma, and 2.5 mm down from the dura) by use of a Hamilton syringe to establish GBM xenografts. The mice were sacrificed when symptoms of brain tumor, including sustained weight loss, ataxia and periorbital bleeding, were observed. Prior to the collection of mouse brains bearing GBM tumors, cardiac perfusion with PBS followed by perfusion with 4% paraformaldehyde was performed. The whole brain was harvested and continuously sectioned at a thickness of 5 µm, then either stained with hematoxylin and eosin (H&E; Beyotime Institute of Biotechnology) using routine histopathological procedures, or directly observed under fluorescent microscopy. Red fluorescence indicated the presence of tumor cells, whereas normal brain cells exhibited no fluorescence.

All determinations were performed from ≥3 independent experiments as indicated. The data are presented as the mean ± SE. Statistically significant differences in mean values between the 2 groups were tested by Student's t test. A value of P<0.05 was considered to indicate a statistically significant difference. All statistical calculations were performed using SPSS software, version 11.0 (SPSS Inc., Chicago, IL, USA).

Following magnetic sorting, a majority of GICs were shown to exhibit negative expression of glioma stem cell marker CD133 and positive expression of marker A2B5 (CD133⁻/A2B5⁺ cells), which were involved in the self-renewal and proliferation of stem cells (Fig. 1A). Co-expression of A2B5 and nestin was observed in the majority of GICs (Fig. 1A). To determine the percentage of cells that expressed the CD133⁻/A2B5⁺ phenotype and the sorting efficiency of the magnetic sorting method, quantitative analysis of CD133 and A2B5 positive cells was performed by flow cytometry. The results demonstrated that the GICs sorted by magnetic beads consisted of 8.5% CD133⁺ cells and 91.6% A2B5⁺ cells; the unsorted GICs consisted of 13.9% CD133⁺ cells and 80.6% A2B5⁺ cells (Fig. 1B).

A high level of human SOX2, a stemness marker of GSCs, and a low expression level of human Tuj1, GFAP and Galc, differentiation markers of tumor cells, were observed in the GICs. In contrast, a reduced number of SOX2⁺ cells and an increased number of Tuj1⁺, GFAP⁺ and Galc⁺ cells were observed in the non-initiating tumor cell population compared with GICs (Fig. 2A). There were significant differences in the expression levels of the stemness marker SOX2 (P<0.001) and differentiation markers GFAP (P<0.001), Galc (P<0.001) and Tuj1 (P=0.002) between GICs and matched non-initiating tumor cells (Fig. 2B).

CD133⁻/A2B5⁺ GICs and matched non-initiating tumor cells were assessed for their invasive potential by a matrigel invasion assay and migratory capability by polycarbonate filters. The results revealed that an increase in invasive cells through the matrigel and more migratory cells through filters were observed in GICs comparing with matched non-initiating tumor cell populations. A total number of 72.7±6.7 GICs in a field passed the matrigel, compared with 30.3±4.6 non-initiating tumor cells (Fig. 3A). The number of migratory cells was 118.3±8.5 in the GIC group when compared with 56.7±5.7 in the non-initiating tumor cells group. The same events occurred in migratory assay (Fig. 3C). Quantified data demonstrated 2.42 folds more invasive (Fig. 3B) and 2.11 folds more migratory (Fig. 3D) cells in GICs comparing to matched non-initiating tumor cells populations. These findings indicated that the migratory and invasive potential of CD133⁻/A2B5⁺ GICs was greater than that of the matched non-initiating tumor cells in vitro (P<0.001). No significant differences in migration or invasion were observed in GICs-RFP compared with GICs or non-initiating tumor cells-RFP (P>0.05). These results indicated that RFP transfection did not change the migration and invasion of GICs or matched non-initiating tumor cells, and cells with RFP may be used for in vivo invasive detection.

In addition to possessing the abilities of sphere-forming and growing, CD133⁻/A2B5⁺ GICs possessed certain genetic features during differentiation. Given that E-cadherin, ICAM-1, MMP-2, MMP-9 and TIMP3 are involved in the acquisition of highly migratory or invasive characteristics by different subsets of glioma cells, the mRNA and protein expression levels of these genes were examined using RT-qPCR and western blot analysis, respectively, to determine whether they were functionally associated with aggressive migration and invasion of GICs (Fig. 4A). The mRNA levels determined by RT-qPCR demonstrated that the expression levels of ICAM-1 (P=0.016), MMP-2 (P<0.001) and MMP-9 (P<0.001) were increased in CD133⁻/A2B5⁺ GICs compared with matched non-initiating tumor cells. The protein expression levels of ICAM-1, MMP-2 and MMP-9 were assessed using western blotting and matched the pattern of mRNA expression, demonstrating a significant increase in these proteins in GICs compared with matched non-initiating tumor cells (Fig. 4B). The mRNA and protein expression levels of E-cadherin and TIMP3 were also detected in GICs. A significant reduction in the mRNA (P=0.003) and protein expression levels of TIMP3 was observed in GICs compared with matched non-initiating tumor cells. On the other hand, no significant difference in E-cadherin mRNA (P=0.137) or protein levels was observed between GICs and matched non-initiating tumor cells.

Aggressive invasion of tumor cells into brain tissue is one of the most malignant characteristics of GBM. To test the infiltrative potential of GICs in vivo, CD133⁻/A2B5⁺ GICs or matched non-initiating tumor cells were transplanted into brains of athymic/nude mice through intracranial injection. Macroscopically, yellow-gray tumors were observed in mice with intracranial GICs or non-initiating tumor cells inoculation. Microscopically, the majority of tumor cells were primitive round or oval, distributed evenly and densely, after transplantation with GICs or non-initiating tumor cells. As presented in Fig. 5, in the brains implanted with GICs, aggressive invasion of tumor cells infiltrating into surrounding normal tissue was observed in 7 mice, and these tumor cells spread particularly far away from the original injection site. Obvious invasion occurred in 87.5% mice when implanted with CD133⁻/A2B5⁺ GICs. However, in the brains implanted with non-initiating tumor cells, the majority of tumor cells remained close to the injection site and no dispersal of tumor cells was detected following cell transplantation in all 8 mice. These data demonstrated that CD133⁻/A2B5⁺ GICs demonstrated infiltrating growth patterns and displayed greater invasive potential compared with matched non-initiating tumor cells.