Metabolic state of glioma stem cells and nontumorigenic cells

Recent experimental and clinical data support the hypothesis that many solid tumors, including brain tumors (9), are organized hierarchically and contain a small number of CSCs. We previously reported that glioma stem cells (GSCs) have lower 26S proteasome activity than nontumorigenic cells, and we have used this feature to monitor GSCs in real time via the fluorescent protein ZsGreen fused to the C-terminal degron of murine ornithine decarboxylase (cODC). This fusion protein is rapidly degraded by the 26S proteasome in a ubiquitin-independent fashion but accumulates in cells with low proteasome activity (6). Neurosphere cultures of gliomas are enriched in GSCs with low proteasome activity, whereas differentiated progeny, with high proteasome activity, die under these conditions by anoikis. Combined with the two different culture conditions, our imaging system allows for comparison of differentiated progeny in monolayer cultures (high proteasome activity, ZsGreen-cODC-negative), where GSCs are very rare, with differentiating progenitor cells (high proteasome activity, ZsGreen-cODC-negative) and GSCs (low proteasome activity, ZsGreen-cODC-positive) in neurospheres (Table S1).

It has long been known that most cancer cells perform aerobic glycolysis. This phenomenon was discovered by Otto Warburg in 1924, when he described that cancer cells metabolize glucose to lactate even under normoxic conditions (7). Although the Warburg effect is thought to be a near-universal phenomenon in cancer cells, it has not been specifically described for CSCs. Therefore, we decided to investigate whether the metabolic state of GSCs differs from the bulk of the tumor cells. To study the metabolic state of GSCs in detail we made use of our recently developed imaging system for CSCs and tested monolayer and 3D neurosphere cultures derived from a panel of glioma cell lines.

To take a closer look at the metabolic state of GSCs and to compare it with the nontumorigenic cell population, we allowed differentiated cells from monolayers or progenitor and stem cells from neurospheres to attach on laminin-coated wells overnight. This method has been shown to preserve the nondifferentiated state of GSCs (10). When basal oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) of differentiated cells and stem/progenitor cells were compared in U87MG and GBM-146 cells, stem/progenitor cells were consistently in a less glycolytic state than differentiated cells [Fig. 1 A and B; OCR: monolayer vs. neurospheres, GBM-146: P < 0.05; U87MG: nonsignificant (n.s.), Student t test]. Basal ECARs were lower in stem/progenitor cells in both cell lines (ECAR: GBM-146: n.s.; U87MG: P < 0.05, Student t test). Stem/progenitor cells in U87MG and GBM-176 showed higher maximal respiratory mitochondrial capacity (n.s., Student t test) and higher mitochondrial reserve capacity (U87MG: P < 0.05; GBM-176: n.s., Student t test) than differentiated cells, indicating that stem/progenitor cells derived from neurospheres exhibit higher resistance against oxidative stress (Fig. 1 A and C). However, GBM-146 stem/progenitor cells showed lower maximal respiratory mitochondrial capacity (P < 0.05, Student t test) and lower mitochondrial reserve capacity (n.s., Student t test) than cells from corresponding differentiated cells (Fig. 1B).

The mitochondrial reserve capacity is thought to correlate to the cells’ ability to cope with oxidative stress (11). Because ionizing radiation deploys its cytotoxic effects mainly through the generation of free radicals, we decided to explore the radiation sensitivity of these cells using standard clonogenic assays. This assay measures the ability of single cells to form colonies of more than 50 cells, equivalent to six cell divisions. Thus, it measures the survival of GSCs as well as progenitor cells with limited proliferative potential. Consistent with decreased mitochondrial reserve capacity in GBM-146 GSCs and progenitor cells, GBM-146 cells were found to be far less radioresistant than GBM-176 and U87MG cells (Fig. 2A). This correlation was confirmed when radiation sensitivity of GSCs was assayed, even though GSCs in general were confirmed to be more radiation resistant, consistent with a previous report (3) (Fig. 2B).

Exposure of differentiated cells and stem/progenitor cells from GBM-176 and GBM-146 to H₂O₂ showed that basal and H₂O₂-induced free radical levels were lower in GBM-176 compared with GBM-146 and that basal and H₂O₂-induced free radical levels in differentiated cells exceeded those in stem cell/progenitor populations (Fig. 2C).

To investigate the metabolic state of the different subpopulations in more detail, we sorted GBM-146 and GBM-176 cells into differentiated cells (ZsGreen-cODC-negative from monolayer cultures), progenitor cells (ZsGreen-cODC-negative cells from neurospheres), and GSCs (ZsGreen-cODC-high cells from neurospheres) and repeated the assessment of OCR and ECAR. These experiments confirmed the observations gained when unsorted differentiated cells from monolayers and stem/progenitor cells from sphere cultures were used, with insignificant differences between progenitor cells and GSCs (Fig. 1 B and C, Student t test).

Next we studied whether the energetic state of the different cell populations was reflected in differences in ATP content. Differentiated cells derived from monolayers had lower ATP content than stem/progenitor cells from neurospheres, supporting the idea that GSCs and progenitor cells depend on oxidative phosphorylation and that this dependence is lost at some point during differentiation and accompanied with a switch to aerobic glycolysis in differentiated cells (Fig. 3A) (12).

To determine at which point during differentiation this metabolic switch occurs, we decided to compare the intracellular levels of ATP in sorted progenitor cells and GSCs, both derived from neurospheres. GSCs from GBM-146, U87MG, and from two additional GBM lines, GBM-176 and GBM-189, had significantly higher ATP content than their corresponding progenitor cells (Fig. 3B, Student t test, P < 0.05). Reexpression of Phosphatase and tensin homolog (PTEN) in U87MG did not change this phenotype (Fig. 3A), suggesting that the PTEN status does not play a role in the differences observed between the ATP levels of GSCs and progenitor cells in U87MG. To test whether this difference depended on differential responses to EGF and basic FGF (bFGF) in the media used to grow the neurospheres, measurements were repeated in the presence or absence of both growth factors for GSCs and progenitor cells sorted from GBM-146 and U87MG. The relative ATP content in the two populations was not affected by the presence or absence of growth factors (Fig. 3C).

The standard media used to grow the neurospheres contained glucose at a concentration of 17.5 mM. To exclude nonphysiological glucose concentrations as a cause for the observed differences in ATP content, we decided to repeat the experiment at physiological glucose levels. For this purpose, the two populations were sorted into media containing 5 mM glucose and allowed to adjust overnight to the new glucose levels. The next day ATP levels were measured. Again, GSCs had higher ATP levels than progenitor cells (Fig. 3D), excluding nonphysiologically high concentrations of glucose as the cause for the observed differences in ATP content.

When the glucose uptake of differentiated cells and stem/progenitor cells were compared using the fluorescent 2-deoxyglucose (2-DG) analog 2-[N-(7-nitrobenz-2-oxa-1,3-diaxol-4-yl)amino]-2-deoxyglucose (2-NBDG), we further confirmed the hypothesis that progenitor cells are less glycolytic than differentiated cells. Differentiated cells had 2-NBDG uptake rates that exceeded those of stem/progenitor cells by twofold (Fig. 4A, P < 0.05, Student t test).

Our ECAR measurements showed differences in the ECAR between GCSs and progenitor cells (Fig. 1 A–C, Left). Consistent with these findings, GSCs produced significantly more lactate than progenitor cells (Fig. 4B). This again indicated that the metabolic state of GSCs may differ from the progenitor and differentiated cells and that the metabolic switch to glycolysis is performed late in the differentiation process (P < 0.05, Student t test).

To determine whether the differences in ATP levels between GCSs and progenitor cells were due to the cells relying on one metabolic pathway vs. another, we decided to treat both populations with different metabolic inhibitors and reanalyze the ATP levels after 3 h of drug treatment. 2-Deoxyglucose was used as a specific glycolysis inhibitor, and oligomycin was used to inhibit the production of ATP via the mitochondrial ATP synthase. After application of the drugs, ATP levels in progenitor cells and GSC populations were affected in a similar manner by both inhibitors. Inhibition of either glycolysis or oxidative phosphorylation had only marginal effects on the overall ATP content, whereas the combined effect of both inhibitors was at least additive (Fig. 5 A–C). Even though the combined inhibition of glycolysis and oxidative phosphorylation for 3 h did not completely deplete cells from ATP, it caused a substantial reduction of total ATP (40–70%). These data suggested that ATP in both subpopulations was mainly produced by glycolysis and oxidative phosphorylation and that inhibition of glycolysis could be compensated through increased oxidative phosphorylation, and vice versa. The lactate produced by cancer cells is not always derived from the conversion of glucose during glycolysis. A portion can also be produced by the degradation of the amino acid glutamine to lactate during glutaminolysis (13). To assess the contribution of glutaminolysis to the ATP content of these cell populations, we measured the ATP levels in the two populations plated in 4 mM glutamine or glutamine-free media. However, absence of glutamine in the media had only little effect on the total ATP content, indicating that glutaminolysis contributes only to a lesser extent to the energy production in GSCs and progenitor cells, or at least that glutaminolysis is not an essential pathway for energy production in these cell populations (Fig. 5 A and B).

Proliferating cells and cancer cells preferentially express the M2 isoform of pyruvate kinase (PK-M2) over the PK-M1 isoform at the protein level (14, 15). The Warburg effect has been described for many different tumor types, and Christofk et al. (15) recently attributed the dependence of cancer cells on glycolysis to the exclusive expression of the M2 isoform of pyruvate kinase. The GBM-146, GBM-176, and U87MG glioblastoma cells expressed both PK-M1 and PK-M2 isoforms, indicating that GBM cells could perform glycolysis and oxidative phosphorylation in parallel. However, PK-M1 or PK-M2 protein expression levels did not differ between progenitor cells and GSCs (Fig. S1).

When we assessed the expression of PK-M2 in a tissue microarray of 65 glioblastoma patients by immunohistochemistry, the expression levels of PK-M2 were not correlated with overall survival (Fig. 6 A and C, log–rank test). However, in patients with nonrelapsed primary and secondary glioblastomas, low expression levels of the 26S proteasome subunit PSMD1, indicating the presence of high numbers of GSCs, were correlated with significantly reduced overall survival (Fig. 6B, P = 0.0487, log–rank test). This trend was also seen when recurrent tumors were included (Fig. 6D, P = 0.0603, log–rank test).

Glioma cell lines, U87, and U87-PTEN were a kind gift from Dr. Paul Mischel [Department of Pathology, University of California, Los Angeles (UCLA)]. The glioma samples GBM-146, GBM-176, and GBM-189 were a kind gift from Dr. Harley Kornblum (Department of Molecular and Medical Pharmacology, UCLA). All cells were cultured as adherent monolayers or as neurospheres in suspension. Monolayers were grown in DMEM-F12 (Invitrogen) supplemented with 10% heat-inactivated FBS (Sigma), penicillin (100 U/mL), and streptomycin (100 μg/mL mixture) and were grown at 37 °C in a humidified atmosphere (5% CO₂). Neurospheres were grown in serum-free DMEM-F12 supplemented with 0.4% BSA (Sigma), 10 mL/500 mL B27 (Invitrogen), 5 μg/mL bovine insulin (Sigma), 4 μg/mL heparin (Sigma), 20 ng/mL FGF 2 (bFGF, Sigma), and 20 ng/mL EGF (Sigma). Stable lines expressing the ZsGreen-cODC reporter protein, which reports for proteasome activity in a live cell, were also generated from each cell line via infection with a ZsGreen-cODC retroviral vector as described previously (6).

For clonogenic survival assays, cells were irradiated as single-cell suspensions using an experimental 200-kV X-ray irradiator (Gulmay Medical). After irradiation an appropriate number of cells was plated into 10-cm Petri dishes into DMEM media, supplemented with 10% FBS. After 14 d, cells were fixed with methanol, stained with crystal violet, and colonies consisting of more than 50 cells were counted. To assess sphere formation, cells were irradiated as single-cell suspensions and plated into ultralow-adhesion 96-well plates at clonogenic densities from 1 to 256 cells per well in DMEM/F12 (as described above). After 14 d, the number of neurospheres per well was counted. Curves were fitted using a linear-quadratic model (GraphPad Prism, version 5).

Cells were incubated in suspension with 20 μM 2′7′-dichlorofluorescein diacetate (DCF-DA) in PBS for 30 min while in the dark. DCF-DA was aspirated, and PBS was replaced with PBS containing 500 μM H₂O₂. Cells were incubated for an additional 5 min, resuspended in PBS, and analyzed by flow cytometry (MACSquant Analyzer, Miltenyi Biotec).

For comparison of ATP levels in monolayers vs. spheres and ZsGreen-cODC-negative vs. ZsGreen-cODC-high cells, the one-step ATPlite Assay (PerkinElmer) was used. Cells deriving from monolayers or neurospheres were plated in 96-well nontreated plates at a density of 5,000 cells per well, in 100 μL of media per well. For comparing ZsGreen-cODC-negative vs. ZsGreen-cODC-high cells, neurosphere cultures were digested with TrypLE Express (Invitrogen) and triturated into a single-cell suspension. The cells were then sorted onto 96-well nontreated plates, at 5,000 cells per well, in 100 μL of serum-free media using fluorescence-activated cell sorting. After 3 h, the ATP assay was conducted according to the manufacturer's protocol. The data from ATPlite assays are presented as an average of more than three independent experiments.

The fluorescent 2-DG analog 2-NBDG (Sigma) was used to measure glucose uptake. Cells were incubated with 2-NBDG (10 μM) for 1 h, washed twice, and analyzed by flow cytometry. For assessment of lactate production, the cells were sorted in a similar manner. After 24 h, media was collected and diluted 1:100 in lactate assay buffer. The amount of lactate present in the media was then estimated using the Lactate Assay Kit (BioVision Research Products) according to the manufacturer's instructions. The amount of lactate produced by the cells in each sample was calculated by subtracting the amount of lactate in the media (without cells) from the amount of lactate in the media from each sample.

The following reagents were used at the doses indicated and as described in the text and figure legends: oligomycin (Sigma), carbonylcyanide-p-trifluoromethoxyphenylhydrazone (FCCP, Sigma), rotenone (Sigma), myxothiazol (Sigma), 2-DG (Sigma), and 3-nitropropionic acid (3-NP, Sigma).

The oxygen consumption and ECARs of monolayers vs. neurospheres was determined using the Seahorse XF Extracellular Flux Analyzer (Seahorse Bioscience). The Extracellular Flux Analyzer allows for analyzing oxygen consumption and ECARs of a defined number of cells in a defined small volume of culture media in real time and for monitoring their response to drug treatment. An overview of the technique can be found in ref. 11. Twenty-four-well plates (Seahorse Bioscience) were coated with laminin (Sigma) as described previously to allow the single cells derived from neurosphere cultures to attach for this assay without differentiating (10). Briefly, each well of the 24-well plate was coated with 50 μL laminin diluted in PBS (10 μg/mL) overnight. The next day the wells were washed three times with PBS, and cells from monolayer or sphere cultures were plated at a density of 40,000 cells per well and allowed to attach overnight in either monolayer or neurosphere media. The following day the adherent cells were washed and fresh media was added. The cartridge was loaded to dispense three metabolic inhibitors sequentially as specific time points: oligomycin (inhibititor of ATP synthase, 1 μM), followed by FCCP (a protonophore and uncoupler of mitochondrial oxidative phosphorylation, 0.5 μM), followed by the addition of a combination of rotenone (mitochondrial complex I inhibitor, 100 nM) and myxothiazol (inhibitor of cytochrome C reductase, 100 nM). Basal OCR and ECAR were measured, as well as the changes in oxygen consumption caused by the addition of the metabolic inhibitors described above. Several parameters were deducted from the changes in oxygen consumption, such as basal OCR, maximum mitochondrial capacity, and mitochondrial reserve capacity (= [maximum mitochondrial capacity] − [basal OCR]) as described previously (19).

Tissue microarrays and immunohistochemical staining were used to analyze PKM2 and PSMD1 expression in tissue, as previously described (20, 21). Tumor specimens were obtained according to a protocol approved by the institutional review board of UCLA. Briefly, tissue microarray enable tumor tissue samples from different patients to be analyzed on the same histologic slide. We constructed the array with a 0.6-mm needle to extract three representative tumor tissue cores from each formalin-fixed, paraffin-embedded tissue block of glioblastoma. These tissue microarrays have been used for other studies (22, 23). Tissue microarray slides were counterstained with hematoxylin to visualize nuclei. Expression analysis was performed by two individual pathologists who were unaware of the findings of the molecular analyses. Scores of 0 and 1 were considered “low expression,” and a score of 2 was considered “high expression.” A log–rank test was used for statistical analysis.