Curcumin decreases malignant characteristics of glioblastoma stem cells via induction of reactive oxygen species

Human Glioblastoma Multiforme (GBM) tissue was obtained from five adult patients from the University of Miami Department of Neurosurgery diagnosed with WHO-IV gliomas based on the World Health Organization (WHO) classification of tumors of the Central Nervous System. Patients or guardians provided written informed consent prior to tumor sample retrieval. Samples were named Glio3, Glio4, Glio9, Glio11 and Glio14. GBM stem-like cell lines were generated as previously described [38]. Briefly, tumors were mechanically and enzymatically dissociated, red blood cells were removed using Red Cell Lysis buffer (SigmaAldrich, St. Louis, MO), Cells were filtered and plated in a 3:1 ratio of Dulbecco’s Modified Eagle’s medium (DMEM): F12 (Gibco, Carlsbad, Ca) media supplemented with 1% penicillin and streptomcycin (penn/strep), 20 ng/ml each of human epidermal growth factor and human fibroblast growth factor, and 2% Gem21 NeuroPlex Serum-Free Supplement (Gemini Bioscience, Sacramento, CA); a formulation consistent for the generation of neurospheres. The GBM cell lines U87, U251 and U235 were purchased from ATCC (Manassas, VA) and were maintained in RPMI media supplemented with 10% FBS and 1% penn/strep. These established GBM cell lines grew in an adherent fashion. All cell lines were routinely tested for mycoplasma using LookOut mycoplasma PCR detection kit (SigmaAldrich, St. Louis, MO) according to the manufacturer’s instructions and were maintained at 37 °C in a humidified 5% CO₂ incubator.

To evaluate stem cell marker expression, neurospheres were dissociated mechanically or enzymatically with Accutase (Gemini Bioscience, Sacramento, CA). To facilitate adherence, cells were plated on poly-L-lysine/laminin coated four-well plates in neurosphere media. Cells were fixed in 4% paraformaldehyde, blocked and permeabilized with a 5% bovine serum albumin (BSA) with 0.6% Triton-× 100 and then treated with the primary antibodies Nestin (Abcam, Cambridge, MA), Sox2, Musashi 1, CD44, Bmi-1 (Cell Signaling Technology, Danvers, MA), CD133 (Biorbyt, Cambridge, UK) and A2B5 (A2B5 clone 105, ATCC, Manassas, VA). A “no primary control” was included for all antibodies tested for all cell lines. For these, the cells were incubated with only the antibody diluent (2.5% BSA, 0.3% triton, balance PBS). Cells were then treated with a fluorochrome-conjugated secondary antibody followed by Prolong Gold Antifade Reagent with DAPI (Thermo Fisher Scientific, Waltham, MA). Samples were examined under an EVOS FLoid Cell Imaging Station fluorescent microscope (Thermo Fisher Scientific, Waltham, MA).

Viability was determined using the CellTiter 96® AQueous One Solution Cell Proliferation Assay (MTS) assay (Promega Madison, WI). Cells were seeded into 96-well plates using a modified neurosphere media containing 5% FBS at a density of 10,000 cells per well in 100 μl of cell culture media. Following treatment, media was aspirated and 100 μl of a 1:5 solution of MTS to cell culture media was added to each well and incubated for 1–4 h. Optical density was measured at 490 nm using BoiTek Synergy HT plate reader. To examine the effect of temozolomide (Sigma-Aldrich, St. Louis, MO), GBM stem cells were treated with 100 μM for 72 h or U87 cells were treated with 10–100 μM. Data is represented as the average of 3 separate experiments in which the viability was calculated as the percent of non-treated cells. To determine the effect of curcumin, cells were treated with increasing concentrations of curcumin (Sigma-Aldrich, St. Louis, MO) for 72 h. The IC_50, the concentration of curcumin at which 50% of cells were non-viable, was determined for a minimum of 3 separate experiments. Data is presented as the average IC₅₀ for each cell line examined.

To determine the effect on cell proliferation 100,000 cells were plated in 10 ml of neurosphere media (100 mm dish for Glio9, and T25 flask for Glio3). Curcumin was added at a concentration of 2.5 μM on day 0. Cells were counted on days 4, 7 and 10 using Orflo Technologies Cell Counter Moxi z (Ketchum, ID). Experiments were done in triplicate.

The effect of curcumin on clonogenic growth potential was determined using sphere-forming assays. Single cells were seeded at 50–100 cells per well in a 96-well plate and treated with 2.5 μM of curcumin on day 0. Spheres were manually counted under microscopy on day 14. All experiments were done in triplicate.

Colony counting was performed to determine colony forming potential of the adherent GSC line. Cells were plated at 200 cells per well in 6-wells plates and treated with 2.5 μM of curcumin at day 0. Colonies were stained with 0.01% crystal violet (Sigma-Aldrich, St. Louis, MO) and counted under microscopy on day 14. Cell clusters of less than 50 cells were not considered colonies and therefore were not counted. Experiments were done in triplicate.

Curcumin-induced ROS was visualized and quantitated using the general oxidative stress indicator CM-H2DCFDA (Thermo Fisher Scientific, Waltham, MA). CM-H2DCFDA passively diffuses into cells and reacts with ROS to yield a fluorescent adduct. For quantification, cells were split into 96-well plates in cell culture media with the addition of 5% FBS to cause adherence to the well bottoms. Samples were treated with 25 μM of curcumin in phenol red free media for 30 min, 4 h, and 24 h. Cells were incubated with 0.5 μM CM-H2DCFDA in PBS for 5 min subsequently washed in PBS and read at an excitation of 495 nm and an emission of 525 nm using BoiTek Synergy HT plate reader. Data is presented as fold change from non-treated cells. Curcumin-induced ROS activity was also examined using fluorescent microscopy. Dissociated GSCs were plated in neurosphere media on poly-L-lysine/laminin coated four-well plates. CM-H2DCFDA fluorescence was evaluated at 1, 6 and 24 h post curcumin (25 μM) treatment. Images were obtained using the EVOS FLoid Cell Imaging Station fluorescent microscope (Thermo Fisher Scientific, Waltham, MA).

Neurospheres cultures, Glios 3, 4, 11 and 14 were plated and treated as neurospheres ranging in size from 100–300 μm as determined by light microscopy. At 8 or 24 h of treatment, the effect of curcumin, N-acetylcysteine (NAC, Sigma-Aldrich, St. Louis, MO) or the combination of curcumin and NAC on protein levels was determined by western blot analysis.

Our method for western blot analysis has previously described [39]. Briefly, GSCs were lysed in RIPA buffer, protein concentrations determined by using BCA protein assay and 20 μg of protein was loaded onto 8, 12 or 15% polyacrylamide gel (BioRad Hercules, CA) gels for electrophoresis and subsequently transferred onto nitrocellulose membranes. The membranes were then blocked for 1 h in 5% non-fat milk (Biorad, Hercules, CA) at room temperature (RT) and incubated with the primary antibody diluted in 2.5% BSA overnight. All primary antibodies were purchased from Cell Signaling (Danvers, MA) except for alpha-tubulin, which was purchased from Abcam (Cambridge, UK) and STAT3, which was purchased from Santa Cruz Biotechnology (Dallas, TX). Membranes were then incubated at room temperature with anti-mouse or anti-rabbit secondary antibodies for 1 h. Blots were developed using SuperSignal™ West Pico Chemiluminescent Substrate (Thermo Scientific Waltham, MA).

In neurosphere media four out of five cell lines formed spheres, where as the Glio9 grew in an adherent fashion (Fig. 1a). Since there is no definitive marker for GBM stem cells, we examined the expression of multiple putative cancer stem cell markers by immunocytochemistry [40‐45]. Except for Glio9 the cell lines demonstrated expression of all markers examined (Fig. 1a). Negative controls for each antibody are shown in Additional file 1: Figure S1A. No SOX2 expression was observed in Glio9. Recently it has been shown that GBM stem cells can be further classified into subgroups, proneural and mesenchymal. These differ both morphologically (neurosphere verse a more adherent phenotype) and in stem cell marker expression [46]. The adherent fashion and the lack of SOX2 expression suggests that glio9 falls into the mesenchymal subgroup. In order to determine if our patient derived cell lines exhibited the cancer stem cell property of chemoresistance [47], we treated five cell lines with 100 μM temozolomide, the chemotherapeutic agent of choice for GBM. We chose a concentration of 100 μM since this is well above the reported (approximately 10 μM) peak levels in cerebral spinal fluid and brain tissue of treated GBM patients [48, 49]. Our results demonstrate that temozolomide had no significant effect on the viability of these GBM cell lines compared to non-treated controls (Fig. 1b). In contrast, the non-GBM stem cell line U87 was sensitive to temozolomide treatment at doses as low as 10 μM, the lowest dose examined (Fig. 1c). These data suggest that our patient-derived GBM cell lines demonstrate progenitor cell properties consistent with glioblastoma stem cells (GSCs).

Several reports have demonstrated that curcumin has anti-neoplastic effects on glioblastoma cells [9, 50‐52]. To determine the effect of curcumin on GSC viability we treated five GSC cell lines with increasing concentrations of curcumin for 72 h. In all cell lines analyzed, curcumin demonstrated a does-dependent decrease in viability (Fig. 2a). All cell lines reached levels less than 20% viability at 70 μM curcumin—the highest concentration tested. The concentration of curcumin at which 50% of cells were non-viable is known as the IC₅₀. The IC₅₀s were as follows: Glio3 25.5 μM (SEM: 2.7 μM), Glio4 39.5 μM (SEM: 5.4 μM), Glio9 22.5 μM (SEM: 1.7 μM), Glio11 20.3 μM (SEM: 3.7 μM), and Glio14 13.9 μM (SEM: 5.0 μM) (Fig. 2b). We also verified that curcumin decreases the viability of GBM non-stem cells using the established GBM cell lines U87, U251 and CH235. The IC₅₀s of these common GBM cell lines were 30.0 μM (SEM: 2.2 μM) for U87, 26.8 μM (SEM: 11.5 μM) for U251, and 23.4 μM (SEM: 1.6 μM) for CH235 (Fig. 2c). Taken together, these results show that curcumin has a does-dependent effect on the viability of both GBM stem cells and non-stem cells.

Cancer stem cells are marked by their ability to proliferate indefinitely and by their sphere- and colony-forming potential at the single cell level in vitro [53, 54]. We chose to carry out the remainder of the experiments in this study using Glio3, a non-adherent GSC cell line, and Glio9, an adherent GSC cell line, due to their similar IC₅₀s and differing adherence patterns. In order to determine if curcumin affects the proliferative ability of GSCs, we plated Glio3 and Glio9 at 1×10⁵ cells and treated with 2.5 μM curcumin on day 0. Curcumin treated Glio3 showed a statistically significant decrease in cell number on days 7 and 10 (p < 0.05) compared to non-treated controls, whereas Glio9 showed a non-significant decrease in cell number on days 7 and 10 (Fig. 3a). To investigate whether curcumin has an effect on the sphere-forming capacity of GSCs, we seeded the non-adherent cell line Glio3 at 50–100 cells per well and treated it with 2.5 μM curcumin on day 0. Spheres were counted on day 14. Glio3 demonstrated a 60% decrease in sphere formation when treated with curcumin compared to non-treated controls (p <0.05) (Fig. 3b). The adherent cell line Glio9 was used to determine if curcumin affects the colony-forming ability of GSCs. Glio9 was plated at 200 cells per well and 2.5 μM curcumin was treated at day 0. On day 14, the curcumin treated cells showed a dramatic 95% reduction in colony number compared to non-treated controls (p < 0.05) (Fig. 3c). These data show that low doses of curcumin inhibit proliferation, sphere-forming and colony-forming potentials of GSCs.

Curcumin has been demonstrated to induce reactive oxygen species (ROS) in various cancer cell lines [55‐57]. To determine if curcumin has the same effect on GSCs we used the molecular probe CM-H2DCFDA, a general oxidative stress indicator, to measure ROS via fluorescence in two cell lines. Under fluorescence microscopy, Glio9 showed an induction of ROS at the 1 and 6 h time points after treatment with 25 μM curcumin with a return to control levels at 24 h (Fig. 4a). After quantification, a one time treatment of 25 μM curcumin was shown to significantly induce ROS in Glio3 and Glio9 with a peak increase of approximately 6–8 fold relative fluorescence at 4 h post-treatment relative to non-treated controls (p < 0.05). ROS were shown to decrease 24 h post-treatment (Fig. 4b). These data suggest that curcumin may cause its effects in GSCs via induction of ROS.

Studies have demonstrated that ROS can induce the activation of multiple signaling pathways including the MAPK pathways in several cell types [58, 59]. We used western blot analysis to determine curcumin’s, and potentially ROS activation’s, modulation on different signaling pathways. Following 8 h of 25 μM curcumin treatment, the phosphorylated (activated) form of ERK, p38 and c-jun (as an indicator of JNK activation) was increased in the GSCs Glio3 and Glio9 (Fig. 5a). This was also demonstrated in all other GSC cell lines (Additional file 2: Figure S2), ERK has been shown to cause the repression of STAT3 activity via dephosphorylation at the Tyr705 position and phosphorylation at the Ser727 location [60]. Here we show that treatment with curcumin decreases the Tyr705 phosphorylated form of STAT3 and increases the Ser727 form in Glio3 and Glio9 (Fig. 5b). When STAT3 is dephosphorylated at the Tyr705 position and phosphorylated at the Ser727 position it is rendered inactive and is incapable of translocating to the nucleus to carry out its downstream effects. We also demonstrate the decreased expression of STAT3’s downstream target Survivin as well as the other anti-apoptosis proteins IAP1 and IAP2 (Glio9 only) in these GSCs (Fig. 5c). These results suggest that curcumin induces the activation of MAPKs and the inhibition of STAT3 activity in GSCs.

N-acetylcysteine (NAC) is an antioxidant shown to decrease ROS [61, 62]. To test whether ROS induction was truly the mechanism for curcumin’s anti-malignant effects on GSCs, we conducted a cell viability assay and western blot analysis to determine if NAC could rescue curcumin’s effects on GSCs. We treated cells with 5 mM NAC, 25 μM curcumin, and a combination of both treatments and viability was determined at 72 h. Treatment with NAC alone had no significant effect on viability on all cell lines except for Glio4, which showed an 18.7% increase in viability (p < 0.05). Treatment with curcumin alone showed significant decreases in viability in all cell lines compared to non-treated controls (p < 0.001). When cells were pretreated with NAC to prevent ROS induction, cell viability was significantly rescued in all cell lines compared to curcumin only treated cells (p < 0.001) (Fig. 6a). To determine if NAC treatment reverses curcumin’s effects on signaling pathways in Glio3 and Glio9, cells were treated with 5 mM NAC, 25 μM curcumin, and a combination of both treatments for 8 h. Western blot analysis indicates that NAC reversed the curcumin-induced MAPK activation (Fig. 6b) and STAT3 deactivation—signified by an increase in p-STAT3 (Tyr705) and a decrease in p-STAT3 (Ser727) (Fig. 6c). This was also demonstrated in all GSC cell lines at the Tyr705 position (Additional file 3: Figure S3). These data demonstrate that ROS induction may be the mechanism behind curcumin’s anti-cancer effects.