Author contributions: P.B.: conception and design, collection and assembly of data, data analysis and interpretation, and manuscript writing; B.O.: collection, assembly, and analysis of confocal data; L.F.:statistical analysis; D.L. and G.B.: GBM provision; G.P.: conception and design, data analysis and interpretation, manuscript writing, and final approval of manuscript.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS January 10, 2013.
The role of the cell surface CD133 as a cancer stem cell marker in glioblastoma (GBM) has been widely investigated, since it identifies cells that are able to initiate neurosphere growth and form heterogeneous tumors when transplanted in immune-compromised mice. However, evidences of CD133-negative cells exhibiting similar properties have also been reported. Moreover, the functional role of CD133 in cancer stem/progenitor cells remains poorly understood. We studied the biological effects of CD133 downregulation in GBM patient-derived neurospheres. Our results indicate that there is not a hierarchical relation between CD133-positive and CD133-negative cells composing the neurospheres. Indeed, CD133 appears in an interconvertible state, changing its subcellular localization between the cytoplasm and the plasmamembrane of neurosphere cells. Silencing of CD133 in human GBM neurospheres using lentivirus-mediated short hairpin RNA impairs the self-renewal and tumorigenic capacity of neurosphere cells. These results imply that CD133 could be used as a therapeutic target in GBMs. STEM CELLS 2013;31:857–869
Glioblastoma (GBM) (World Health Organization grade IV astrocytoma) is the most common primary brain tumor and the most lethal among gliomas in adults. Despite advances in surgical and chemotherapeutic treatments, no therapies have a clear benefit in improving the survival time of patients that remains less than 1 year. GBM infiltrates throughout the brain making complete surgical resection impossible, its cellular composition is heterogeneous and it is largely resistant to radiotherapy and chemotherapy.
There are now compelling evidences that the bulk of malignant cells in GBM is generated by a rare fraction of self-renewing, multipotent tumor-initiating cells, named cancer stem cells (CSCs) [1, 2]. The CSC hypothesis assumes that tumors consist of a cellular hierarchy with a subpopulation of cells able to maintain and propagate the tumor because of their capacity of self-renewal and resistance to chemotherapy and radiotherapy .
The identification of stem cells in brain tumors provides a powerful tool to investigate the tumorigenic process in the central nervous system and to develop therapies targeted to these cells. However, the methods for the isolation/enrichment of CSCs remain still imperfect and require improvements, as a consequence of the lack of universal markers . Up-to-date, there are no specific markers that might be used as a tool to distinguish a normal stem cell from a CSC as well as stem cells from progenitors.
CD133 is largely used as CSC marker in several tumors, including GBM. It was originally identified as a surface antigen expressed on hematopoietic stem cells . CD133 was then used in the isolation of neural stem cells from human fetal brain . Despite the demonstration that cell sorting for CD133 expression can enrich for cells with tumorigenic potential in brain tumors [1, 2], the utility of CD133 in the isolation of brain tumor stem cells has been questioned in several studies [7–13]. The high variability of CD133 expression in gliomas (from 1% to 60%), even if stem cells are supposed to be rare within the tumors, and the existence of gliomas with CD133-negative cells able to self-renew and regenerate tumors in xeno-transplantation assay, has led to the reconsideration of CD133 as a reliable CSC marker in all brain tumor cases. These conflicting findings could be the results of experimental differences or variations between individual brain tumors and may reflect assay sensibility/specificity.
CD133 is the first identified member of prominin family of pentaspan membrane glycoproteins . Its function in normal and cancer tissues is not definitely understood. Recent studies conducted in several human tumor cell lines reported that CD133 regulates the proliferation and the colony-forming ability of the cancer cells [15, 16].
In this study, we examined the in vitro and in vivo biological effects of CD133 knockdown in GBM-derived neurospheres, in order to better understand the function of CD133 in initiation and maintenance of tumors, the regulation of its expression and cellular localization, and the molecular mechanisms that it controls in tumorigenesis. Our findings strongly suggest that CD133 is essential for the self-renewal and the tumorigenic potential of stem/progenitor cells in human GBMs and encourage its use as therapeutic target in brain tumors.
MATERIALS AND METHODS
Preparation of Cell Suspensions from Patient Tumors
This study was approved by the Ethical Committee for human experimentation of European Institute of Oncology and all patients signed an approved consent document prior to surgery. Surgical specimens of tumors were collected at the Neurosurgery Dpt at Istituto Neurologico Carlo Besta (Milan), and tissue fragments without necrotic areas were processed. The tissue was dissociated into single cell suspension with papain (2 mg/ml) (Worthington Biochemical, Lakewood, NJ, http://worthington-biochem.com) at 37°C for 2 hours.
Neurosphere Culture and Clonogenic Assay
Neurosphere cultures were maintained in Dulbecco's modified Eagle's medium—Ham's F-12 Nutrient Mixture (DMEM-F-12) supplemented with B27 Supplement (Life Technologies, Paisley, U.K., http://www.lifetech.com), 20 ng/ml epidermal growth factor, 10 ng/ml basic fibroblast growth factor (PeproTech, Rocky Hill, NJ, http://www.peprotech.com), and 0.0002% heparin (Sigma-Aldrich, St. Louis, MO, http://www.sigmaaldrich.com), at 37°C in a 5% CO2 humidified incubator, as previously described . All cultures were passaged by mechanical dissociation of the spheres and the cells seeded at the density of 104 cells per square centimeter. Cell growth was measured seeding the cells in 96-well plates (2,000 cells per well) in DMEM/F-12 complete medium and using the CellTiter 96 AQueous Non-Radioactive Cell Proliferation Assay (Promega, Madison, WI, http://www.promega.com). To measure the clonogenic capacity, the cells resuspended in DMEM/F-12 with methylcellulose (Stem Cell Technologies, Vancouver, BC, Canada, http://www.stemcell.com) were seeded in a 35-mm culture plate (100 cells per square centimeter). Reported numbers represent a minimum of three plates per condition. Ten to fifteen days after plating, neurospheres containing more than 20 cells were scored.
Fluorescence-Activated Cell Sorting Analysis
For CD133 extracellular staining, CD133/1-PE (AC133, Miltenyi, Biotec GmbH, Bergisch Gladbach, Germany, http://www.miltenyibiotec.com, 1:10) was added to the cells for 10 minutes at +4°C. For CD133 intracellular staining, the cells were pre-incubated with a fivefold excess of biotin-conjugated CD133/1 antibody to saturate extracellular binding sites. Cells were then permeabilized with the Inside Stain Kit (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany, http://www.miltenyibiotec.com) and stained with CD133/1-PE resuspended in Inside Perm buffer. Gating for single cells was established using forward scatter in the isotype control sample. The isotype control sample was used to establish a gate in the PE channel. Cells showing signal for CD133 above the gate established by the isotype control were deemed to be CD133-positive cells. FACS Vantage SE flow cytometer (BD Biosciences, Franklin Lakes, NJ, http://www.bdbiosciences.com) was used for the analysis. Sorted CD133-positive and CD133-negative cell fractions had at least 98% of purity in all cases analyzed.
Lentiviral-Mediated CD133 Silencing
The pLentiLox3.7-Puro-GFP (pll) was used to generate lentiviral plasmids that express short hairpin RNAs (shRNAs): sh1 (GACCCAACATCATCCCTGT) , sh2 (AAGGCGTTCACAGATCTGAAC), or a nontargeting (NT) (CGTACGCGGAATACTTCGA). shRNAs were cloned into the XhoI and HpaI sites of the vector. To generate lentiviral particles, 293T cells were cotransfected with packaging vectors pMDLg/pRRE, pRSV-REV, and pMD2G and pll-sh1, pll-sh2, or pll-NT using calcium phosphate. Viral particle-containing supernatants were collected at 48 hours after transfection and used. Cells (105) were infected with 5 × 107 viral particles and selected in puromycin (Sigma-Aldrich, St. Louis, MO, http://www.sigmaaldrich.com) for 3 days.
Cloning of Human CD133
The Caco2 cell line was used as a source for CD133. Total RNA was purified with RNeasy Mini Kit (Quiagen, Valencia, CA, http://www.qiuagen.com) and immediately converted to cDNA using random hexamers and Improm-II Reverse Transcriptase (Promega). The wild-type full-length human CD133 Open Reading Frame was amplified by polymerase chain reaction (PCR) using Phusion High-Fidelity DNA Polymerase (Finnzymes, Vantaa, FI, www.thermoscientificbio.com/finnzymes/) and directionally cloned into the BamHI and SmaI sites of the pWPXL lentiviral vector. The PCR primers were as follows: 5′-GCACGGATCCTGGAGGATCTTGCTAGCTATG-3′ (forward); 5′-GAGCTCGAGTCAATGTTGTGATGGGCTTGTC-3′ (reverse). The recombinant construct was verified by DNA sequencing. A shRNA-resistant wild-type CD133 was generated using the QuickChange site-directed mutagenesis kit (Stratagene, LA Jolla, CA, http://www.stratagene.com). A point mutation in the sh2 annealing site was introduced without generating alterations in the amino-acid identity. The mutagenic primers designed were as follows: 5′-GACAAGGCGTTCACAGACCTGAACAGTATCAATTC-3′ (forward); 5′-GAATTGATACTGTTCAGGTCTGTGAACGCCTTGTC-3′ (reverse). The mutant construct was verified by DNA sequencing. Lentiviral constructs containing the CD133 gene and the additional packaging plasmids (pMD2G and ΔR8.2) were cotransfected into the packaging cell line 293T using calcium phosphate. The lentiviral supernatants were collected 48 hours after transfection and used to infect CD133 silenced cells with the sh2 oligo in presence of 8 μg/ml polybrene.
Western Blotting and qPCR
Performed according to the established procedures. Primary antibodies and dilutions used were: CD133/1 (W6B3C1, Miltenyi, Biotec GmbH, Bergisch Gladbach, Germany, http://www.miltenyibiotec.com 1:100) and actin (Sigma-Aldrich, St. Louis, MO, http://www.sigmaaldrich.com, 1:1,000) used to normalize the amount of lysate loaded on the gels.
Total RNA was isolated with RNeasy Mini kit and formaldehyde fixed-paraffin-embedded (FFPE) RNeasy kit from neurosphere cells and FFPE brain slides, respectively (Quiagen, Valencia, CA, http://www.qiuagen.com). Quantitative PCR analysis was performed with the 7.500 Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, http://appliedbiosystems.com) using the Syber Green PCR Master Mix (Applied Biosystems, Foster City, CA, http://appliedbiosystems.com). The threshold cycle (CT) values for each gene were normalized to β-actin expression levels. The primers used were as follows: CD133: ACCAGGTAAGAACCCGGATCAA (forward), CAAGAATTCCGCCTCCTAGCACT (reverse); ACTIN: AGAAAATCTGGCACCACACC (forward), AGAGGCGTACAGGGATAGCA (reverse); hypoxanthine phosphoribosyltransferase (HPRT1, human specific primers): TGACCTTGATTTATTTTGCATACC (forward), CGAGCAAGACGTTCAGTCCT (reverse).
In vivo tumor-initiation analyses were done with CD-1 nude mice in accordance with the Italian Laws (D.L.vo 116/92 and following additions), which enforces EU 86/609 Directive (Council Directive 86/609/EEC of 24 November 1986 on the approximation of laws, regulations and administrative provisions of the Member States regarding the protection of animals used for experimental and other scientific purposes). Cells from dissociated neurospheres were resuspended in 2 μl of phosphate buffered saline (PBS) and stereotaxically injected into the nucleus caudatus (0.7–1 mm posterior, 3 mm left lateral, 3.5 mm in depth from the dura) of 5-week-old female nu/nu CD1 mice (Charles River, Wilmington, MA, http://www.criver.com). The mice were maintained until development of neurological signs and the brains of euthanized mice were collected.
Mouse brains and human surgical samples were fixed with 4% formalin and paraffin-embedded. For hematoxylin and eosin staining, sections (5 μm) were deparaffinized in xylene, rehydrated in a graded alcohol series, and stained in Mayer hematoxylin and eosin-phloxine B solution. For immunohistochemistry, sections were blocked with a PBS solution containing 10% goat serum and 0.2% bovine serum albumin and incubated first overnight with the primary antibody CD133/1 (AC133, Miltenyi, Biotec GmbH, Bergisch Gladbach, Germany, http://www.miltenyibiotec.com, 1:100) and with the appropriate secondary antibody for 45 minutes at room temperature. Prior to coverslip application, nuclei were counterstained with Mayer's hematoxylin and images were visualized with a bright-field microscope.
In the case of cell surface versus intracellular labeling of CD133, the Inside Stain Kit was used (Miltenyi, Biotec GmbH, Bergisch Gladbach, Germany, http://www.miltenyibiotec.com) as previously described . Briefly, cells were first cell surface labeled with CD133 (W6B3C1, Miltenyi, Biotec GmbH, Bergisch Gladbach, Germany, http://www.miltenyibiotec.com, 1:100) and Cy3-conjugated α-mouse IgG (Sigma-Aldrich, St. Louis, MO, http://www.sigmaaldrich.com) at +4°C. Then the cells were fixed in Inside Fix Buffer (Miltenyi, Biotec GmbH, Bergisch Gladbach, Germany, http://www.miltenyibiotec.com) and finally labeled with CD133 (W6B3C1, Miltenyi, Biotec GmbH, Bergisch Gladbach, Germany, http://www.miltenyibiotec.com, 1:100) and a FITC-conjugated α-mouse IgG (Sigma-Aldrich, St. Louis, MO, http://www.sigmaaldrich.com). Nuclei were counterstained with 4′-6-diamidino-2-phenylindole (DAPI) (Molecular Probes, Eugene, OR, http://www.invitrogen.com) and images were captured using a Leica TCS SP2 Leica upright confocal microscope (Leica Microsystems, Germany, http://www.leica.com). Individual sections (0.5/1.0 μm interval) or a composite of 10/16 optical sections are shown. The images shown were prepared from the confocal data files using ImageJ software (rsbweb.nih.gov/ij/).
Analysis of Apoptosis
For the analysis of apoptosis, infected cells were first fixed in 1% formaldehyde for 20 minutes on ice, washed once in PBS, and fixed again in ethanol 75% for 30 minutes on ice. Fixed cells were incubated in propidium iodide (PI) (BD Biosciences, Franklin Lakes, NJ, http://www.bdbiosciences.com, 2.5 μg/ml) and RNase (Sigma-Aldrich, St. Louis, MO, http://www.sigmaaldrich.com, 250 μg/ml) for 12–16 hours at +4°C and analyzed by flow cytometry.
For cleaved caspase-3 immunostaining, the cells were fixed in formaldehyde and permeabilized with 0.2% Triton X-100 prior to incubation with cleaved caspase-3 antibody (Cell Signaling, Danvers, MA, http://www.cellsignal.com/, 1:100) 45 minutes at room temperature. Cy3-conjugated AffiniPure donkey-α-mouse IgG (H+L) Fab fragment (Jackson ImmunoResearch Laboratories, West Grove, PA, http://www.jacksonimmuno.com/), incubated for 45 minutes at room temperature, was used as secondary antibody and (DAPI) (Molecular Probes, Eugene, OR, http://www.invitrogen.com) to stain nuclei.
Statistical analyses were performed using Prism 5.0 software (http://www.graphpad.com/). Data graphed with error bars represent mean ± SEM. Two-tailed Student's t test or analysis of variance (ANOVA), corrected for multiple comparisons as appropriate, was used to determine the significance of any differences between experimental groups. In Kaplan Meyer curves, the survivals were compared with log-rank analysis. Differences were considered “statistically significant” when p < .05 (*), p < .01 (**), and p < .001 (***).
CD133 Is Variably Expressed in Human GBMs and Their Derivative Neurospheres
We examined a series of newly diagnosed and recurrent GBM cases, classified as grade IV according to the WHO system, to investigate the expression level of CD133 by fluorescence-activated cell sorting (FACS) analysis in established neurosphere cultures  and in acutely dissociated tumors.
Our screening revealed that the neurospheres analyzed had a wide range of CD133 expression varying from 1.0% to 84.8% (Fig. 1A and supporting information Fig. S1). Four out of eleven neurosphere lines included in the study had an abundance of CD133-positive cells (from 84.8% to 58.1%), (called CD133-high cells), similar to what has been previously reported [2, 20]. The other seven neurosphere lines contained smaller populations of CD133-positive cells (called CD133-low cells), ranging from 8.4% to 1.0%. No GBM-derived neurospheres with undetectable levels of CD133 expression were found, in contrast with other reports [7–10]. We have also studied CD133 expression in acutely dissociated tumors. Using flow cytometry, we found that three out of six tumors had less than 1% of CD133-positive cells, while the other three cases displayed higher CD133 positivity, ranging from 3.6% to 13.9% positive cells (Fig. 1A), confirming the high variability in the expression of CD133 in GBMs. No significant differences in the capacity to grow as neurospheres and in the tumorigenic potential was observed between CD133-high and CD133-low neurospheres (supporting information Fig. S2). The percentage of CD133-positive cells remained constant for individual neurosphere lines over several in vitro passages (Fig. 1A).
In order to determine whether the CD133-positive cells identified in the neurospheres exist in the corresponding human GBMs in situ, we performed immunohistochemical analysis of patient-derived GBM paraffin-embedded sections. A subpopulation of CD133-positive cells was detected in GBM patients, revealing the consistency between in vitro data and qualitative immunohistochemical evaluation (Fig. 1B). In some cases, such as GBM#20, the neurospheres contained few CD133-positive cells while the corresponding tumor sample presented high CD133 expression. Interestingly, we noticed a clear cytoplasmic localization of CD133 rather than exclusive plasmamembrane associated expression, in accordance to previous results [21–24]. This result suggested an exclusive cytoplasmic localization of CD133 in some cases, not detected in not permeabilized cells, as those analyzed by FACS in our first screening.
CD133 Is Localized in the Cytoplasm of Cells Composing the GBM Patient-Derived Neurospheres
In an attempt to further characterize the GBM-derived neurospheres for the expression of CD133, we analyzed CD133 mRNA expression levels by qRT-PCR. We found that in the majority of the samples analyzed, the CD133 mRNA expression levels were consistent with the percentages of CD133-positive cells measured by FACS: CD133-high neurospheres showed the highest CD133 mRNA levels. However, in five out of seven CD133-low cells, we observed considerable levels of CD133 transcript (Fig. 2A) suggesting that the mRNA levels not always reflect the CD133 expression levels on the plasmamembrane. This discrepancy induced us to investigate in depth the expression of CD133 protein independently of its cell surface localization.
We investigated the CD133 localization within the neurosphere cells by confocal analysis. The cell surface labeling revealed that CD133 is uniformly distributed in the plasmamembrane (Fig. 2B). Remarkably, a second labeling performed upon cell permeabilization revealed an additional intracellular staining of CD133 (Fig. 2B). This observation is in accordance with other recent studies that reported a cytoplasmic localization of CD133 in tumor cells [21, 24]. Thus, two distinct pools of CD133 molecules coexist in GBM-derived neurosphere cells. Of note, the majority of cells showed both an extracellular and intracellular CD133 localization in GBM#10 cells, while on the other hand CD133 had an almost exclusive intracellular localization in GBM#18 cells. We were not able to find cells with only an extracellular staining of CD133.
In addition, we performed a flow cytometry analysis of intracellular CD133 and we found a cytoplasmic localization of CD133 in almost all the neurospheres analyzed; in particular the CD133-low neurospheres had a large amount of intracellular CD133 (Fig. 2C). Notably, not all CD133-low cells had cytoplasmic expression of CD133, suggesting that a population of cells not expressing CD133 at all does exist in the neurosphere.
CD133 Is Localized in Cytoplasm of CD133 Plasmamembrane-Negative Cell Fraction
The previous results demonstrated the existence of at least two populations of cells within the neurosphere relating to CD133 expression: (a) cytoplasm and cell surface double positive cells and (b) only cytoplasm-positive cells. In order to further investigate the phenotype of GBM-derived cells with respect to CD133 expression and its subcellular localization, we used the FACS to sort the cells into CD133-positive (CD133 plasmamembrane-positive cells) and CD133-negative fractions (CD133 plasmamembrane-negative cells) and we studied the subcellular distribution of CD133 in the sorted CD133-positive and CD133-negative cell fractions by confocal analysis. As expected, we found that CD133 was not expressed on the plasmamembrane of CD133-negative cells; however, it was exclusively diffused in the cytoplasm (Fig. 3A and supporting information Figs. S3A, S4A). After 15 days in culture, the CD133-positive and CD133-negative cells had all the same subcellular distribution of CD133, with the CD133-negative cells re-expressing the protein on the plasmamembrane (Fig. 3A and supporting information Fig. S3A). Indeed the CD133-negative cells resembled the parental cell population from which they were sorted, already after 3 days of culture post-sorting (supporting information Fig. S4B).
These results were in accordance with the observations that the CD133-positive and CD133-negative cell fractions showed comparable expression levels of CD133 protein and mRNA when total cell lysates were analyzed by Western blotting and qRT-PCR, respectively. The total CD133 expression levels did not change after culturing the cells in vitro for 15 days (supporting information Fig. S5A, S5B).
Flow cytometric analysis of intracellular CD133 in GBM neurospheres showed cytoplasmic expression of CD133 in the unsorted, CD133-positive, and CD133-negative cells (Fig. 3B and supporting information Fig. S3B). Moreover, almost all CD133 plasmamembrane-positive cells have intracellular CD133, while only the 60% of CD133 plasmamembrane-negative cells of GBM neurospheres was positive to the intracellular staining. This result suggests that a discrete amount of cells not expressing any CD133 composes the neurosphere. Of note, after 15 days in culture, the three cell fractions had similar extracellular and intracellular CD133 localization. This is a strong indication of a dynamic plasmamembrane localization of CD133 in the cell, probably subjected to a re-cycling from the plasmamembrane to the cytoplasm and vice versa. Nevertheless the existence of a hierarchical organization between the CD133-positive and CD133-negative (cytoplasmic or not) cells cannot be excluded.
CD133 Plasmamembrane-Negative Cells Have Reduced Self-Renewal and Tumorigenic Capacity
The CD133-positive and CD133-negative cell fractions sorted from different GBM patient-derived neurospheres displayed qualitative similar results for neurosphere growth, in the sense that both fractions generated expandable neurospheres, although CD133-negative cells formed spheres slightly smaller in size than the CD133-positive cells (supporting information Fig. S5C).
The clonogenic efficiency was analyzed seeding the cells at low density in methylcellulose whose viscous composition does not allow cell aggregation. Single clones were counted after 15 days and the clonogenic cells were calculated as percentage on the total number of seeded cells. In all cases examined, a slight difference in the clonogenicity of CD133-positive and CD133-negative cells (Fig. 3C and supporting information Fig. S3C) was observed. Interestingly, when the spheres formed in methylcellulose were dissociated and replated in the same conditions, there was no difference in clonogenic capacity and sphere size between the two cell fractions (Fig. 3C and supporting information Fig. S3C). Intriguingly, the changes in subcellular localization of CD133 could correspond to the acquisition by CD133-negative cells of self-renewal capacity, feature of CD133-positive cells.
We examined the ability of the CD133-positive and CD133-negative cell progeny to generate orthotopic tumors in nude mice. For these studies, we inoculated sorted cells (n = 105 per mouse) into the forebrain of 5-week-old nude mice. Unsorted cells were injected as controls. All the three cell populations (unsorted, CD133+, and CD133−) formed tumors in mice, although the overall survival of tumor-bearing animals was significantly longer in CD133-negative cell-injected mice (Fig. 3D and supporting information Fig. S3D).
Histological analysis of tumors arose in all mice inoculated with the three cell populations showed that the tumors were similar in cytoarchitecture, vascularization, and invasiveness at the tumor center and periphery and were classified as WHO grade IV astrocytomas. Consistent with the in vitro data, immunohistochemical detection of CD133 in the xeno-tumors revealed that CD133 expression was similar within the tumors formed by CD133-positive and CD133-negative cells (Fig. 3E and supporting information Fig. S3E).
CD133 Shuttles Between the Plasmamembrane and the Cell Cytoplasm
In order to understand whether a hierarchy could be established between cell surface CD133-positive and CD133-negative cells, we cloned single cells of each fraction and analyzed the CD133 expression on the cell surface in each secondary population. Secondary clones from CD133-positive cells invariably contained a mixture of positive and negative cells and the percentage of CD133 plasmamembrane-positive cells in these expanded clonal cultures did not differ from the unsorted population. In addition, all the secondary clones derived from CD133-negative cells contained a mixture of positive and negative cells, although there was a greater variability in the content of CD133 plasmamembrane-positive cells among clones (Fig. 4). The possible contamination of CD133-positive cells in the CD133-negative fraction has been excluded by applying the binomial probability law, showing that the probability to obtain more than two wells with CD133-positive cells in the negative cell population was less than 1% in GBM#9, 3.2% in GBM#10, and 1.6% in GBM#18 (supporting information Fig. S6).
We were not able to identify clones not re-expressing CD133 on the plasmamembrane nor clones expressing any CD133. These findings are consistent with a dynamic regulation of CD133 subcellular localization, suggesting a shuttle of the protein between the cytoplasm and the plasmamembrane.
CD133 Silencing Reduces the Self-Renewal Capacity of GBM Stem Cells
A direct role of CD133 in maintaining the tumorigenic potential of GBM stem cells remains to be defined. To investigate the function of CD133, we silenced CD133 gene expression in the entire cell population of neurospheres derived from four different patients, using shRNAs.
To achieve a stable knockdown of CD133 in these cells, we cloned shRNAs directed against different regions of human CD133 mRNA within a lentiviral vector. One of the two shRNAs (sh2) gave a nearly complete knockdown of CD133, while the other (sh1) was less efficient (Fig. 5A). CD133-silenced cells showed reduced cell growth (Fig. 5B). The magnitude of the reduction in cell growth was proportional to the efficacy of the two shRNA constructs. These dose-dependent effects strongly suggest that the decreased growth we observed was a specific effect caused by the reduction of CD133 expression. The reduction in cell proliferation caused by CD133 knockdown is a robust response common to stem/progenitor cells derived from different GBM patients (Fig. 5B).
No difference in the percentage of apoptotic cells between CD133 silenced cells and the control cells was detected, as revealed by PI staining (supporting information Fig. S7A, S7B) and measurement of the cleavage of caspase-3 by FACS and immunofluorescence staining, respectively (supporting information Fig. S7C, S7D). These results indicate that CD133 regulates the cellular proliferation and is not a survival factor for GBM-derived neurospheres. Since a key characteristic of the CSCs is the self-renewal capacity, we studied the clonogenicity of CD133-silenced cells. Single cells were seeded on methylcellulose-coated plates and allowed to form colonies for 2 weeks. CD133 silenced cells formed fewer colonies than the control cells. Notably, the decreased clonogenic capacity of CD133 silenced cells was dose-dependent, consistent with the level of knockdown produced by each oligonucleotide (Fig. 5C). Interestingly, the reduction of the self-renewal of GBM#10 cells was less pronounced than in other cell lines, consistent with the less efficient CD133 silencing obtained in these cells (Fig. 5A).
The neurospheres formed by cells with reduced CD133 expression were markedly smaller than the spheres formed by the control cells (Fig. 5D). When the single spheres were dissociated and single cells were plated in a 96-well plate, the CD133-silenced cells reformed less spheres, additionally suggesting reduced self-renewal capacity (Fig. 5D).
Together, these data indicate that CD133 is required for the self-renewal potential of GBM stem cells, although its expression is not associated with the expression of other putative stem cell markers. Indeed, the expression of Olig2, Nestin, and CD15 was not modified by CD133 knockdown (supporting information Fig. S8).
Silencing of CD133 Impairs the Tumorigenic Potential of GBM Stem Cells
In order to determine the tumorigenic potential of CD133 silenced cells in vivo, 105 infected cells were injected into the brain of 5-week-old nude mice. The animals were sacrificed at the first appearance of neurological signs. We found that CD133 silencing significantly increased tumor-free survival. Notably, only the mice injected with cells interfered with the sh2, which was the more efficient oligo in silencing CD133, had a significant prolonged survival (Fig. 6A and supporting information Fig. S9A). Interestingly, when lower numbers of cells (103 and 102) were injected in mice, CD133-silenced cells did not form tumors (Fig. 6A, 6B and supporting information Fig. S9A, S9B). Interestingly, tumor incidence was already reduced at high doses (105 cells) in the GBM#18 case (supporting information Fig. S9A, S9B) where a marked CD133 silencing was obtained (Fig. 5A). All tumors formed by the CD133-silenced cells expressed the protein at levels comparable to the control (Fig. 6C and supporting information Fig. S9C, S9D, S9E). The tumor formation was dependent from both CD133 expression and the number of inoculated cells (Wilk's G test: respectively, G2 = 17.00, 2 D.F., p = .0002, G2 = 13.92, 3 D.F., p = .003). Considering that CD133 re-expression is essential for tumor formation, the ELDA algorithm (http://bioinf.wehi.edu.au/software/elda/) was used to estimate the rate of CD133 re-expression in the silenced cell population (NT 1:1, sh1 1:260, sh2 1:8,880; difference between sh1 and sh2: χ2 = 16.9, 1 D.F., p = .00004) and resulted in a lower frequency of CD133 re-expression in the CD133-silenced cells with both sh1 and sh2 oligos compared to the control cells.
Re-expression of CD133 Recovers the In Vitro and In Vivo Phenotype of CD133-Silenced GBM Stem Cells
CD133 full-length open reading frame was amplified from a human colon carcinoma cell line and modified through a point mutation in the complementary sequence to the sh2 oligo, in order to generate a shRNA-resistant CD133 without modification in the amino-acidic sequence. The CD133 cDNA was cloned into a lentiviral vector and then over-expressed in the CD133-silenced cells (with the sh2 oligo) derived from three different patients (Fig. 7A). Re-expression of CD133 in the CD133-silenced cells was able to rescue the proliferation and the clonogenic capacity of neurosphere cells at the same extent of the control cells (Fig. 7B, 7C). Moreover, the inhibitory effect of CD133 silencing on tumor-initiating ability was completely rescued by the shRNA-resistant wild-type CD133 over-expression upon xeno-transplantation in the brain of immune-compromised mice (Fig. 7D, 7E). These data further demonstrate that CD133 plays an essential role in the maintenance and in the tumorigenic potential of GBM stem cells.
In GBMs, CD133 identifies a subpopulation of stem-like tumor cells, named CSCs, which drive tumor formation and are highly resistant to conventional chemotherapy and radiotherapy. However, several studies report the existence of glioma stem cells not expressing CD133 able to self-renew and retain tumorigenic potential.
These discrepancies result in the reconsideration of CD133 as a universal marker for CSCs in gliomas. Thus, an in-depth characterization of the complex CD133 phenotype in GBMs is much needed to re-evaluate the significance of CD133 in these tumors. Furthermore, questioning the utility of CD133 in the identification of CSCs implies the study of its possible functional role in gliomagenesis and in stem cell properties, since its function in normal and cancer cells is not clearly defined.
We choose the neurosphere system as a good surrogate for in vitro study of glioma stem and progenitor cells [19, 25]. A recent study indicates that in vitro expansion of GBM stem/progenitor cells as neurospheres does not alter the differentiation ability and the tumorigenic potential of these cells, neither their karyotype and gene expression pattern . Thus, each neurosphere cell line reflects an image of the tumor, which it was derived from, and remains representative after moderate expansion. The major findings of this investigation relate to the localization of CD133, reversibly modulated between plasmamembrane and cytoplasm, and to the essential role of CD133 in the maintenance of human GBM stem/progenitor cells.
We found distinct expression patterns of CD133 both in neurospheres and in freshly dissociated tumors. Substantial amounts of membrane-bound CD133 were detectable only in a fraction of neurosphere lines, that we called CD133-high, whereas CD133 mRNA and intracellular CD133 protein were found expressed at higher levels in almost all neurosphere lines examined. The existence of an intracellular pool of CD133 in addition to the one on the cell surface has been demonstrated in human hematopoietic stem cells and in established GBM cell lines in two very recent studies [18, 21], confirming our results.
Cell sorting, based on cell surface CD133 expression, allowed the isolation of a CD133-negative cell fraction composed of cells with an exclusive cytoplasmic expression of CD133. These cells were less clonogenic and formed tumors with a longer latency than the CD133-positive cells did. Notably, the impairment in self-renewal capacity and tumorigenic potential of CD133-negative cells was lost when these cells re-expressed CD133 on the plasmamembrane. The rapid re-expression of CD133 on the surface of CD133-negative cells has been described in different cancer cell types  suggesting a conserved mechanism of CD133 regulation in tumors of different tissues. These results suggest that the localization of CD133 on the plasmamembrane is functionally related to the CSC biology and characterizes cell populations with enhanced self-renewal and tumorigenic capacity. Our findings confirm the observation that the cell surface CD133 is a marker for self-renewing and tumor-initiating GBM cells , but meanwhile add the notion that the mere isolation of CD133 plasmamembrane-positive cells is not sufficient to enrich for a stem cell population. We indeed observed that also neurospheres with low cell surface CD133 expression are able to grow and propagate the tumor, strengthening the idea of CD133 as a nonessential element for stem cell properties in all GBM cases.
A recent study shows that self-renewing CD133-negative cells are present in neurosphere cultures derived from human GBM . These cells can be divided in two different types: (a) CD133-negative cells that generate aggressive tumors comprising a mixture of CD133-positive and CD133-negative cells, (b) CD133-negative cells that give rise to slow-growing circumscribed tumors formed of CD133-negative cells alone. The authors proposed that these cells are hierarchically related and represent different stages of differentiation. However, any kind of observation about a possible intracellular localization of CD133 in these two types of CD133-negative cells was not reported. They suggested indeed a possible transcriptional regulation of CD133 rather than changes in subcellular localization.
Through clonal analysis, we have been able to describe the intrinsic capabilities of individual tumor cells and their progeny. We found that every clone formed by a single CD133-negative cell contains cells with a considerable expression of cell surface CD133, although in some cases to a lesser extent, demonstrating that there is not a hierarchical relation between CD133-positive and CD133-negative cells. Indeed, CD133 appears in an interconvertible state, changing its expression levels and subcellular localization between the cytoplasm and the plasmamembrane. These changes might be regulated by several kinds of stimuli, such as the progression through the cell cycle phases or the supply of oxygen and nutrients. It has been demonstrated that the CD133 expression is strongly influenced by hypoxic conditions and is associated with alterations in mitochondrial function . It was also found a downregulation of CD133 expression in the G0/G1 phase of the cell cycle in neural stem cells .
It remains obscure whether a population of cells that do not express CD133 at all exists in the neurosphere. Interestingly, the intracellular staining of CD133 in the CD133 plasmamembrane-negative cell fraction reveals the existence of a cell population not expressing any CD133. The analysis of the stem cell properties of these cells might be relevant, although it is impossible to isolate them using the FACS sorting.
The functional relation between CD133 protein expression and putative cancer stemness is still a matter of debate. Previous observations showed that targeting CD133 results in decreased proliferation, anchorage-independent growth, invasion, and xenograft growth in different cancer cell types [15, 30–32]. However, little insights have been made into the function of CD133 in the stem cell compartment of human gliomas. In this study, we show that CD133 is an important mediator of CSC biology since it is involved in the maintenance and in the tumorigenic potential of GBM stem cells. The silencing of CD133 in neurospheres derived from human GBMs reduces the growth, the self-renewal, and the tumor-initiating ability of stem/progenitor cells. This inhibitory phenotype was completely recovered re-expressing CD133 in the silenced cells, strengthening the idea of a relevant role of CD133 in GBM stemness.
Surprisingly, we observed similar biological effects of CD133 silencing in GBM-derived neurospheres expressing different basal levels of CD133 protein. Considering that the exact mechanism of CD133 activation (post-translational modifications as tyrosine phosphorylation, acetylation or glycosylation) is not known, we can speculate that cells expressing different levels CD133 mRNA could have comparable amount of functionally active protein. Moreover, it has to be considered that the peculiar heterogeneity among the GBMs implies that cells derived from different patients may have different genetic lesions.
Taken together, our results indicate a functional role of CD133 in gliomagenesis, giving further significance to its use in the therapy of GBM. It has to be taken in consideration that cytoplasmic CD133 is relocalized on the plasmamembrane, rendering the anti-CD133 therapy, based exclusively on the targeting of cell surface CD133, not effective in every case. Targeting CD133 using shRNAs might be more effective.
CD133 is essential to the maintenance and the tumorigenic potential of GBM stem cells, since its silencing impairs both the self-renewal and tumorigenic capacity of GBM stem cells. The mere isolation of CD133 plasmamembrane-positive cells is not sufficient to enrich for a stem cell population in GBM since there is a cytoplasmic pool of CD133 recycling to the plasmamembrane. The existence of the cytoplasmic CD133 reservoir might render ineffective a therapy exclusively based on the blocking of the cell surface CD133, while the targeting of the total protein might have more efficacy.
We thank D. Riva and E. Bellani for the technical assistance, S. Ronzoni for assistance with flow cytometry and C. Richichi, D. Osti, and M. Setti for critical review of the manuscript. We also thank P.G. Pelicci and J.N. Rich for critical comments and helpful advice. P.B. was supported by Foundation Umberto Veronesi (FUV). G.P. was supported by Ministry of Health grants.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.