Limits of CD133 as a marker of glioma self-renewing cells



In human gliomas, self-renewing and tumor-initiating cells are characterized by the expression marker CD133. Although, widely used, the validity of CD133 is debated as recent data show that CD133+ and CD133 cells share similar stemness and tumorigenic properties. To clarify this “CD133 controversy”, we reexamined the methods of purification and the stem behavior of both CD133 compartments in fresh gliomas and gliomasphere cultures. Using human anti-CD133-coupled microbeads and magnetic activated cell sorting, we observed a nonspecific sorting of glioma cells irrespective of their CD133 expression. In contrast, when purified by fluorescence activating cell sorting, a specific expression and enrichment of CD133 was successfully observed in fresh human gliomas and gliomasphere cultures. However, neither the expression of stemness genes nor the long-term self-renewal capacities of CD133+ and CD133 cells were significantly different, even after fresh isolation. Altogether, our data show that purification of CD133+ glioma cells using hCD133-microbeads presents a lack of specificity and demonstrate that the use of CD133 as a unique glioma stem cell marker is likely not sufficient to tag the whole self-renewing tumor cell reservoir. © 2009 UICC

The human prominin-1/CD133 protein has recently gained attention as a cell surface marker to isolate human embryonic neural stem cells and tumor-initiating cells (TICs).1–4 When isolated from human brain tumors, CD133+ cells display stem cell properties in vitro and initiate tumor growth in vivo.1, 2 However, compelling studies report that some CD133 glioma cells share similar properties with CD133+ cells, including self-renewal and tumorigenicity, suggesting that CD133 is insufficient to fully define TICs.5–8 Nonetheless, it remains accepted that CD133 is the only available marker and nearly the whole body of published work attempting to characterize TICs in gliomas is based on the use of CD133. Here, we reexamined the use and the relevance of CD133 as a marker of self-renewing cells in human gliomas by comparing the magnetic activated cell sorting (MACS) vs. fluorescence activating cell sorting (FACS) sorting methodologies, technically and functionally.1, 2, 5–8

Material and methods

Human specimen processing, gliomasphere cultures, limiting dilution assays and statistic tool

Peripheral blood mononuclear cells (PBMCs) and primary brain tumors were obtained following approved protocol by the ethical committee. Tumor processing, cell cultures and secondary spheres assays were done as described in Refs. 3 and 9. Briefly, the tumour was chopped and digested with papain. The spherogenic capacity of purified CD133+ and CD133 single cells after fresh isolation or from enriched stem cell cultures was tested by limiting dilution (1 cell/μl) and determined by quantifying the number of primary spheres and secondary spheres. The long-term self-renewal ability of individual clones was followed by dissociating individual sphere clones mechanistically along successive passages. Paired t-test was used for statistics.

RNA extraction and reverse transcription-polymerase chain reaction

Total RNAs were extracted using the RNAqueous-Micro kit (Ambion). Reverse transcription and semi-quantitative polymerase chain reaction (PCR) reactions were performed using Superscript II (Invitrogen) and red master mic (Peqlab). Referred to Ref. 9 for primer sequences.

CD133 staining, MACS and FACS sorting

For FACS analyses and purification, fresh glioma and spheres were dissociated, washed, and incubated with unconjugated mouse anti-AC133 or AC133-PE or 293C3-PE or AC133-APC (Miltenybiotech) at a dilution of [1/2] in phosphate buffered saline-bovine serum albumin 0.5%-ethylene-diamine-tetra-acetic acid 2 mM for 30 min at 4°C. As a control, cells were incubated with the secondary antibody (anti-mouse-Alexa 568 1:100, Invitrogen) or the isotype control. Dead cells were analysed and excluded using trypan blue at 1:1,000 (FL3 channel). Expression level analysis and sorting were done on FACScan and/or FACSAria, respectively (BD Bioscience). Purified cells were collected in microfuge tubes.

For MACS purification, fresh glioma and spheres were dissociated, washed, and incubated either with AC133- or with mCD4- coupled microbeads (MiltenyiBiotech) at a concentration of 1 μl/106 cells according to manufacturers' instructions. Magnetic cell separation was performed using manual MS-MACS columns or an AUTOMACS machine according to manufacturers' instructions. MACS-purified cell fractions were then tested by FACS using the FACS analysis procedure and antibodies described above. Postacquisition analysis was performed using CellQuest and Diva Softwares.


The percentage of CD133+ cells and the stability of its expression along in vitro passages were evaluated in freshly dissociated gliomas and gliomasphere cultures using flow cytometry (n = 13, Fig. 1, Table I). In freshly dissociated primary gliomas, CD133+ cells represent less than 1% of the viable cell population irrespective of the tumor grade (n = 7, Fig. 1a, Table II). In gliomasphere cultures,4, 10 only a very small subset of viable glioma cells (average < 2.5%, n = 5, Fig. 1b, Table III) expressed CD133 along several in vitro passages. Only 1 glioma stem cell culture out of 5 (GSM IV-1) contained higher amounts of viable CD133+ cells reaching 6.33 ± 2.6% (Fig. 1b, Table III). Importantly, the expression of CD133 was overall stable along in vitro passages (Figs. 1d and 1e). As a control, freshly purified PBMCs were co-stained with human anti-CD34-FITC and anti-CD133-PE or -APC antibodies, showing that 0.1% of the viable population are CD133+/CD34+ progenitor cells (Fig. 1c), which is consistent with previously described results.11, 12

Figure 1.

CD133 expression levels in human fresh gliomas and gliomasphere cultures. (a) hCD133 expression levels in freshly purified human glioma cells derived from various grade (II, top panel; III, middle panel; IV, bottom panels). (b) hCD133 expression levels in gliomasphere cultures. (c) hCD133 and hCD34 expression levels in control PBMCs from a healthy individual. P1 gates viable cells and is depicted in red. (d), (e) stability of CD133 expression along in vitro culture passages after short-term (d, fresh to passage 5), and long-term cultures (e, passage 6 to 23). Lines represent the linear trendline resulting from individual percentage along time.

Table I. Tumor List
Tumor typeGradeLocationGenderAge
O.A II-2Oligo-astrocytoma grade IIParietal leftM41
O.G III-2Oligodendroglioma grade IIIFrontal rightM64
O.A III-1Oligo-astrocytoma grade IIITemporo-amygdala leftM52
Primary GBM-1GlioBlastoma multiforme grade IVParietal leftM80
Primary GBM-2#Temporal leftF67
Primary GBM-3#Fronto-temporal leftM50
Primary GBM-4#Occipital rightM75
Primary GBM-13#Tempo-parietal rightM79
Primary GBM-15#Parietal rightF80
Primary GBM-16#Temporal leftF52
Primary GBM-17#Fronto-parietal rightM45
Secondary GBM-1#Frontal rightM64
GSM IV-1Gliosarcoma grade IVTemporal rightM70
Table II. Fresh Specimens
Samples% CD133+ cells determined by FACS% CD133+ cells counted after MACS sorting
O.A II-20.06nd
O.G III-20.145.60
Primary GBM-40.013.75
Primary GBM-130.1nd
Primary GBM-150.8nd
Primary GBM-161.0nd
Primary GBM-170.2nd
Table III. Gliomasphere Cultures
In vitro cell cultureMean of % CD133+ cells analysed by FACSMean of % CD133+ cells counted after MACS sorting
O.A III-11.41 ± 0.904 ± 1.97
Primary GBM-21.60 ± 0.6414.35 ± 0.63
Primary GBM-30.70 ± 0.269.08 ± 0.96
Secondary GBM-10.97 ± 0.327.14 ± 0.41
GSM IV-16.33 ± 2.615.72 ± 1.25
Mean2.20 ± 0.9510.05 ± 1.04

Although CD133 expression levels were overall low but detectable, we tried to enrich for CD133+ cells using hCD133 microbeads and MACS separation.1, 2, 5–7 After selection, the number of viable cells in both MACS-fractions was counted using trypan blue exclusion, showing that the percentage of viable column-bound cells was 2.5 to13 times higher than the number of CD133 + cells determined by FACS (n = 2/7 fresh tumor, n = 5/5 gliomasphere cultures, Tables II and III). Surprisingly, the FACS analysis of MACS-purified fractions revealed no enrichment for hCD133 expression levels in any of the glioma-purified cells tested (n = 5/5) contrasting with the control PBMCs sorting where there was a significant enrichment of CD133+ cells (Fig. 2b). The specificity of cell binding using hCD133 microbeads was therefore tested using microbeads coupled to an irrelevant antibody, mouse CD4, which do not recognize the human CD4 epitope. In the majority of the gliomasphere cultures, we found that the proportion of cells binding to the mCD4 microbeads was comparable to that we observed with hCD133 microbeads, whereas in PBMCs and GSM IV-1 a significantly lower number of cells bound to mCD4 microbeads (n = 3/4, Fig. 2a, Table III). Expectedly, no enrichment for CD133 was ever detected in any of the mCD4 MACS-fractions derived from gliomasphere cultures including GSM IV-1. This is likely due to the saturation of CD133 by the antibodies coupled-microbeads (Figs. 2b and 2c). We nevertheless assessed the presence of self-renewing cells in vitro in MACS-purified fractions using clonal assays. Neither the ability to form secondary spheres nor the long-term self-renewal revealed a functional difference between both hCD133-MACS purified fractions (Figs. 2d and 2e).

Figure 2.

Self-renewing properties of CD133+ cells isolated by hCD133-microbeads and MACS.(a) Percentage of bound viable cells after mCD4- or hCD133-microbeads and MACS separation. Only glioma cells from GSM IV-1 denote significance (asterisk) regarding the total number of bound cells when purified using hCD133. Experiments were done in duplicate. (b) Representative dot plot FACS of the hCD133 expression levels after hCD133-microbeads and MACS cell purification in GSM IV-1 (left panel) and control PBMCs (right panel), showing 176-fold enrichment for the latter. (c) Representative dot plot FACS of the mCD4 expression levels after mCD4-microbeads and MACS cell purification in GSM IV-1. (d), (e) Percentage of secondary sphere forming (d) and long-term passages of individual clone (e) after hCD133-MACS selection, and limiting clonal dilution assay. Experiments were done in triplicate.

In an effort to overcome the lack of specificity hCD133 microbeads, we performed standard CD133 staining and FACS.8 Selection of CD133+ cells in gliomasphere cultures (n = 5/5) or control PBMCs leads to a specific enrichment up to 77% and 12.7%, respectively (Figs. 3a and 3b). At the RNA level, a moderate enrichment for CD133 and several putative stemness genes9 expression including NESTIN was found in the CD133+ fraction (Fig. 3b) and the clonal assays revealed that the purification of CD133+ cells from GSM IV-1 culture (n = 1/5) or freshly dissociated gliomas (n = 3/3) gives rise to a significant enrichment for cells with a higher capacity to generate primary clones compared to that of CD133 cells (Figs. 3c and 3d). However, the testing of long-term self-renewal ability of individual CD133+ and CD133 clones revealed that the majority of the CD133+ cells were not able to sustain long-term sphere formation along in vitro passages (Fig. 3e).

Figure 3.

Self-renewing properties of CD133+ cells isolated by hCD133 antibody and FACS. (a) Representative dot plot FACS of CD133 expression after FACS cell purification in GSM IV-1 (left panel) and control PBMCs (right panel), showing a respective 31 and 423-fold enrichment. (b) Representative PCR gel pictures of the gene expression levels in CD133+ and CD133 FACS purified fractions in GSM IV-1 and primary GBM-3. Samples were analysed by semi-quantitative PCR. (c) Percentage of secondary spheres after purification of CD133+ cells by FACS from gliomasphere cultures. The asterisk denotes significance. Six independent experiments were done. (d), (e) Percentage of primary spheres (d) and long-term passages of individual clones (e) after purification of CD133-expressing cells by FACS from fresh gliomas and clonal assays. The asterisk denotes significance.


Here we show that the use of CD133 as a cell surface marker to identify and isolate progenitor/stem like cells depends on the biological system studied and the techniques used. Using FACS analysis, we confirmed that fresh human gliomas and gliomasphere cultures express CD133 at low and sometimes barely detectable levels, and we showed that the purification of CD133+ cells by FACS, although accurate, robust and specific, failed to enrich for long-term self-renewing glioma cells even after fresh dissociation. Furthermore, our study revealed that the use of hCD133-coupled microbeads present a lack of specificity due to an unspecific binding of glioma cells to microbeads irrespectively of the antibody used. The fact that glioma cells unlike PBMCs, bound equally well to hCD133- or mCD4- microbeads suggests that they may express cell surface proteins including adhesion molecules and/or integrins that would confer them a particular affinity to microbeads.

The difficulties to detect and purify reliable and comfortable amount of viable CD133+ cells and its moderate benefit for isolating self-renewing cells raise a general caution about the use of CD133 as an exclusive stem cell marker. Beyond practical and methodological aspects,13 which might have influenced the variable results seen in the literature, we wonder whether a single surface marker prone to environmental, epigenetic and genetic stresses, would be stable and reliable enough to tag the whole self-renewing cell population. We therefore plead for the necessity of developing alternative or complementary selection strategies to identify, isolate and characterize the whole TIC reservoir for a better comprehension of the brain tumor biology.