Caffeine Confers Radiosensitization of PTEN-Deficient Malignant Glioma Cells by Enhancing Ionizing Radiation–Induced G₁ Arrest and Negatively Regulating Akt Phosphorylation

Glioblastoma multiforme (GBM) is the most frequent primary malignancy of the adult central nervous system and the most malignant form of glioma (1). Current treatment strategies involve neurosurgery, polychemotherapy, and ionizing radiation (IR). Long-term survival for patients with GBM is still limited, especially due to the emergence of resistance to DNA-damaging procedures such as IR.

The mechanisms of glioma cell resistance to IR are not completely understood. However, despite the expression of wild-type p53 and p21 in primary GBM cells, they often fail to arrest in cell cycle following IR-induced DNA damage (2). In general, cells arrest in cell cycle to enable DNA repair and, in case of irreparable damage, undergo apoptosis; both mechanisms are governed by p53 (3). However, cell fate is not only dependent on the amount and type of damage but also on the functional interaction of various damage-inducible signal transduction cascades. The ability of p53 to induce cell cycle arrest or apoptosis on IR can be antagonized by survival signals, such as by the phosphatidylinositol-3-kinase (PI3K)–dependent activation of Akt/protein kinase B, which indirectly leads to the inhibition of p53 functions by activating its negative regulator MDM2 (4). Because Akt is constitutively activated in cells with mutations of the phosphatase and tension homologue gene (PTEN; refs. 5, 6) and because PTEN mutations are frequently found in GBM (7–9), a pivotal role of Akt in functional p53 deficiency (4, 10) and radioresistance of GBM can be expected.

Caffeine, a methylxanthine, is well known to radiosensitize cancer cells including malignant gliomas through the inhibition of ATM and ATR kinase activity and the abrogation of the G₂-M checkpoint (11). However, it has also been reported that caffeine has inhibitory effects on the phosphoinositide metabolism, including inhibition of the enzymatic activity of PI3K (12). Furthermore, caffeine and PTEN have been found to act synergistically regarding growth inhibition and the induction of apoptosis in cancer cells by the downregulation of the PI3K/Akt pathway (13), suggesting the antiproliferative effects of caffeine that are comparable with those of PTEN. Based on these reports and because only limited data exist on the effects of caffeine on PI3K/Akt signaling and the cell cycle checkpoints in malignant glioma, we asked if caffeine differentially affects the radiosensitivity of malignant glioma cells displaying different PTEN status. We further aimed at determining by which underlying signaling pathways a potential radiosensitization might be mediated.

The U87MG GBM cell line, which is deficient for PTEN but contains wild-type p53 (14), was obtained from the American Type Culture Collection. The EA14 malignant glioma cell line containing wild-type PTEN and p53 has originally been established by Stuschke et al. (15) from biopsy samples of an anaplastic astrocytoma WHO grade III. The wild-type p53 status of both cell lines (15, 16) was confirmed by our group by temperature gradient gel electrophoresis followed by sequence analysis of the exons 5, 6, 7, and 8 of the p53 gene (data not shown). The T98G and U373MG cell lines, which harbor mutations both in PTEN and p53 (16), were obtained from American Type Culture Collection.

Cells were cultured as previously described (15). In brief, cells were grown in monolayers in MEM (Life Technologies Ltd.) supplemented with 15% heat-inactivated FCS, 1% nonessential amino acids, penicillin (1 × 10⁵ U/L), and streptomycin (100 mg/L). Cultures were incubated at 37°C and 5% CO₂ in a humidified atmosphere.

Before treatment, cells were arrested in the G₀-G₁ phase of the cell cycle by contact inhibition. Synchronization of cultures was confirmed by flow cytometric analysis. Cells were then harvested by trypsinization and were plated in cell culture flasks (25 cm²) at a defined cell density. Because caffeine has been shown to interfere both with G₁ and G₂-M checkpoints of cell cycle, we irradiated cells either 5 h after plating when >90% of cells were still in G₀-G₁ (stationary phase; Fig. 1A and B) or 24 h (U87MG) or 28 h (EA14) after plating when at least 20% of cells were in the G₂-M phase of the cell cycle (exponential growth phase; Fig. 1C and D), respectively. IR with graded doses up to 8 Gy was performed by using a ⁶⁰Co source at 1 to 2 Gy/min. Caffeine (Sigma) was used at a final concentration of 2 mmol/L because this dosage has been previously reported to effectively inhibit ATR and ATM (11). Because caffeine plasma concentrations of 2 mmol/L cannot be clinically achieved because of the dose-limiting toxicity of this drug, we also tested pentoxifylline (Sigma), a methylxanthine derivative with reduced toxicity, for its radiosensitizing effect in PTEN-deficient and PTEN-proficient glioma cells. Furthermore, the selective Akt inhibitor triciribine (Calbiochem) was used to determine whether Akt or ATM/ATR are central targets of caffeine-modulated radiosensitivity of PTEN-deficient glioma cells. Caffeine (2 mmol/L), pentoxifylline (2 mmol/L), or triciribine (5 μmol/L) were added to the medium 30 min before IR and were removed by the exchange of culture medium 72 h later.

Flow cytometric analysis was used to quantify the relative number of cells in different phases of the first (G₀-G₁, S, and G₂-M) as well as the consecutive cell cycle (G₀'/G₁', S', and G₂'/M'; ref. 17). Briefly, cells were labeled continuously during the whole culture period with 5-bromodeoxyuridine (BrdUrd; Serva) and, at the end of the experiment, they were stained with the DNA dyes Hoechst 33258 and ethidium bromide (EB; Sigma-Aldrich). Because incorporated BrdUrd quenches the fluorescence of DNA-bound Hoechst dye and because the amount of incorporated BrdUrd positively correlates with the number of replications, cells can be distinguished according to how many cell cycles they have passed. In contrast, labeling with EB is unaffected by BrdUrd and allows one to differentiate between cells in the cell cycle phases G₀-G₁, S, and G₂-M.

In detail, cells were continuously labeled with 60 μmol/L BrdUrd immediately after IR. Cells were then harvested at different time points between 0 and 60 h after IR and stored at −80°C in cell culture medium containing 10% DMSO until analyzed. After thawing, samples were pelleted by centrifugation and resuspended in 1 mL ice-cold Hoechst staining buffer (1.2 μg/mL Hoechst 33258, 10 mmol/L Tris-HCl, 154 mmol/L NaCl, 0.5 mmol/L MgCl₂, 1 mmol/L CaCl₂, 0.1% Nonidet P40, and 0.2% bovine serum albumin). Cells were incubated at 4°C for 15 min. EB was added to a final concentration of 0.2 μg/mL, and samples were incubated again at 4°C for 15 min. Flow cytometric analysis was done by means of a PAS III three-channel flow cytometer (PARTEC GmbH) equipped with an HBO 100 mercury arc lamp source. Excitation at 365 nm and detection of red fluorescence for EB at 570 nm versus blue fluorescence for Hoechst 33258 at 435 nm were used. A minimum of 2 × 10⁴ cells was analyzed per sample. Analysis of the DNA content distribution in each cell cycle phase was done using WINMDI (supplied by Dr. Joe Trotter, Salk Institute, La Jolla, CA).

Cell survival after IR alone or in the presence of caffeine or pentoxifylline was determined using a clonogenic survival assay. Cells were arrested in G₀-G₁ by contact inhibition and subsequently seeded into 25-cm² culture flasks at defined densities. In their stationary or exponential growth phase, cells were treated with caffeine or pentoxifylline, IR, or a combination of both. Colonies were grown for 14 d, then fixed with 70% ethanol and stained with Giemsa (Sigma). Colonies with >50 cells were counted. Plating efficiency and survival fractions for given treatments were calculated on the basis of survival of nonirradiated cells treated with the solvent control, or caffeine or pentoxifylline alone. Each sample was done in triplicate and at least three independent experiments were carried out.

The quantitation of expression levels of p53, p21^WAF1, pan-Akt, or Akt phosphorylated at serine 473 (p-Akt^Ser473) in cell lysates from EA14 and U87MG cells was done by standard Western blot analysis. Briefly, standard SDS-PAGE was done using 90 μg of total protein per cell lysate, followed by transfer to nitrocellulose membranes (Amersham GE Healthcare). The following antibodies were used for detection: mouse anti-p53 (clone DO-1, Santa Cruz Biotechnology), mouse anti-p21^WAF1 (clone EA10, Calbiochem), mouse pan-Akt (clone 40D4) and mouse anti–p-Akt^Ser473 (clone 587F11, both from Cell Signaling Technology), mouse anti–β-actin (clone AC-74, Sigma), and peroxidase-conjugated goat anti-mouse IgG (Jackson Immuno Research Laboratories).

All statistical analyses were done using StatView 5.1 (Abacus Concepts, Inc.). The significance of differences in surviving fractions and cell cycle distribution of cell lines treated with increasing doses in the absence or presence of caffeine was determined by paired t test. The significance of differences in the effect of caffeine, pentoxifylline, or triciribine on surviving fractions at 2 Gy (SF2) in U87MG, EA14, T98G, and U373MG cells (Table 1A and B; Fig. 5) was determined using paired t test for normally distributed and Wilcoxon-Signed rank test for non-normally distributed data, respectively.

We first assessed the basal radiosensitivity of synchronized cell cultures from the PTEN-deficient U87MG and the PTEN-proficient EA14 cells by irradiating them with increasing doses (0–8 Gy) either in their stationary growth phase (>90% of cells were in G₀-G₁; Fig. 1A and B) or exponential growth phase (≥20% of cells were in G₂-M; Fig. 1C and D). Following a culture period of 14 days, survival fractions were determined. As reported earlier (15), U87MG cells were relatively resistant to IR (Fig. 1A and C). In contrast, the EA14 cell line was significantly more sensitive (Table 1; Fig. 1B and D). Because caffeine has been shown to inhibit PI3K activity and should therefore counteract constitutive PI3K activity in the PTEN-deficient U87MG cells, we treated EA14 and U87MG cells with 2 mmol/L caffeine 30 minutes before irradiation and again assessed their clonogenic survival. As shown in Fig. 1A and C and Table 1A, pretreatment with caffeine significantly sensitized U87MG to IR. In contrast, caffeine pretreatment had no radiosensitizing effects on the PTEN-proficient E414 cells (Table 1A; Fig. 1B and D).

Because the concentration of caffeine that we used for our in vitro studies cannot be achieved in a clinical setting because of the dose-limiting toxicity of caffeine, we also tested pentoxifylline, a less toxic methylxanthine derivative for its radiosensitizing effects. Of clinical importance, we also observed a selective radiosensitizing activity of pentoxifylline in PTEN-deficient glioma cells (Table 1B).

We next addressed whether the selective interaction of caffeine with radiosensitivity of PTEN-deficient cells occurred at distinct cell cycle checkpoints. We first focused on the G₁ DNA damage checkpoint. We therefore irradiated U87MG and EA14 cells in their stationary growth phase with 6 Gy in the absence or presence of caffeine and analyzed whether the IR-induced G₁ checkpoint of the ongoing cell cycle was modulated by caffeine in a PTEN-specific manner. As shown in Fig. 2A (top), in the PTEN-deficient and radioresistant U87MG cells, only a minor percentage of cells remained arrested in G₁ overtime following IR of cells with 6 Gy. Pretreatment with 2 mmol/L caffeine significantly increased the number of U87MG cells that remained arrested in G₁. This temporal extension of the G₁ checkpoint by caffeine in U87MG cells was sustained because 30% ± 1.5% of cells remained in G₁ for up to 60 h after IR (Fig. 2A, top). Caffeine-mediated prolongation of G₁ arrest of irradiated U87MG cells was also reflected by a decreased number of cells progressing to G₂-M (Fig. 2A, bottom). In contrast, IR with 6 Gy alone induced an arrest in G₁ in the radiosensitive PTEN-proficient EA14 cells, and caffeine did neither affect this IR-induced G₁ arrest (Fig. 2B, top) nor the IR-induced inhibition of cell cycle progression to G₂-M (Fig. 2B, bottom).

Besides the modulation of the G₁ checkpoint, caffeine has been shown to cause an override of the IR-induced G₂-M checkpoint (11); however, the role of PTEN at this checkpoint and the effect of caffeine therein are rather unclear. We therefore irradiated U87MG and EA14 cells with 6 Gy in their exponential growth phase in the absence or presence of caffeine and analyzed the relative numbers of U87MG and EA14 cells at G₂-M of the ongoing cell cycle at 0 to 60 h postirradiation. In contrast to the observed significant differences in the extent and duration of the G₁ checkpoint, both U87MG and EA14 cells became initially arrested in G₂ for up to 20 h on IR (Fig. 3A and B, top). Pretreatment of cells with 2 mmol/L caffeine completely abrogated this IR-induced G₂ arrest both in the PTEN-deficient and PTEN-proficient cells: twenty hours after IR in the presence of caffeine, only 15% + 2% of U87MG cells and 9% ± 0.6% of EA14 were still in G₂ (Fig. 3A and B, top). This abrogation of the IR-induced G₂ arrest by caffeine was again persistent because even 60 hours after IR, both cell lines still revealed a significantly lower proportion of cells in G₂ after caffeine pretreatment when compared with IR alone (Fig. 3A and B, top).

We next investigated whether caffeine would also modulate the G₁ checkpoint of cells in which DNA damage had occurred in a cell cycle phase other than G₁. We irradiated U87MG and EA14 cells with 6 Gy in their exponential growth phase in the absence or presence of caffeine and again used the BrdUrd/Hoechst quenching technique for the discrimination between the ongoing and consecutive cell cycles. We found that both U87MG and EA14 cells irradiated during the exponential growth phase reached the G₁ phase of the consecutive cell cycle (G₁') to a degree that, 60 hours after IR, was similar to that of untreated controls or to cells treated with caffeine alone (Fig. 3A and B, bottom). Pretreatment of both cell lines with 2 mmol/L caffeine resulted in a significant IR-induced arrest in G₁' (Fig. 3A and B, bottom). In EA14 cells, the inhibitory effect of caffeine on the G₂ block was equalized by its augmenting effect at the G₁' checkpoint activation, which might explain the overall absent effect of caffeine on clonogenic survival after IR. In contrast, in U87MG cells, caffeine more efficiently increased the G₁' checkpoint activation than it decreased the IR-induced G₂ arrest, resulting in an overall inhibition of cell cycle progression.

In our model, caffeine did not affect the radiosensitivity of PTEN-proficient cells (Fig. 1B) but significantly increased the radiosensitivity of PTEN-deficient cells (Fig. 1A), which was preceded by an increase of IR-induced G₁ arrest by caffeine in cells with impaired PTEN function. DNA damage–induced G₁ checkpoint is particularly regulated through the induction of p21^Waf1 expression by p53 (18), the function of which is downregulated in PTEN-deficient glioma cells (10). We therefore asked whether the interaction of caffeine with the G₁ checkpoint in PTEN-deficient cells would involve p53 reactivation. First, we analyzed p53 and p21^Waf1 protein expression in U87MG and EA14 cells after IR in the absence or presence of caffeine. In both U87MG (Fig. 4A) and EA14 cells (Fig. 4B), p53 was induced 2 hours after IR with 6 Gy. Accordingly, p21^Waf1 expression levels were also significantly increased in both cell lines 6 hours after IR (Fig. 4A and B), suggesting that the observed impaired IR-induced G₁ checkpoint in U87MG cells (Fig. 2A, top) was not a result from low to absent p21^Waf1 induction. Pretreatment of U87MG (Fig. 4A) or EA14 cells (Fig. 4B) with caffeine resulted only in a slight inhibition of IR-induced p53 or p21^Waf1 expression. This moderate effect of caffeine on IR-induced p53 and p21 expression is in line with the absent interaction of caffeine with the extent or duration of the IR-induced G₁ checkpoint in EA14 cells (Fig. 2B, top). Our data further suggest that the significant effect of caffeine on the G₁ checkpoint activation (Fig. 2A) in PTEN-deficient cells is independent of an enhancement of IR-induced p53 or p21^Waf1 upregulation. To confirm the p53 independency of the caffeine-mediated radiosensitization of PTEN-deficient glioma cells, we used T98G and U373MG cells, which carry mutations in PTEN and p53. Indeed, in both cell lines, caffeine pretreatment significantly reduced surviving fractions after IR with 2 Gy (Table 1A).

Caffeine has been shown to inhibit PI3K (12), the upstream activator kinase of Akt. Because increased Akt activation in PTEN-deficient cells has been directly linked with their resistance to IR (19), caffeine might exert its radiosensitizing effects by reducing Akt activation in our model. Therefore, we investigated the levels of p-Akt^Ser473 in PTEN-proficient and PTEN-deficient glioma cells at the time points 30, 60, or 120 minutes after the cells had been irradiated in the absence or presence of caffeine. As shown in Fig. 4A, untreated U87MG cells contained significantly higher levels of p-Akt^Ser473 compared with untreated EA14 cells (Fig. 4B). No significant changes of p-Akt^Ser473 expression levels were induced by IR in either cell line. Independently of the PTEN status, caffeine rapidly, i.e., already 30 minutes after IR, and significantly decreased the expression of p-Akt^Ser473 in U87MG (Fig. 4A) and EA14 cells (Fig. 4B). The extent of inhibition of p-Akt levels was comparable for each time point analyzed (data not shown for 60 and 120 minutes following IR). Of note, caffeine treatment before IR decreased the p-Akt^Ser473 levels of U87MG cells to a level comparable with untreated E414 cells (Fig. 4A and B), which well correlated with its radiosensitizing efficacy in U87MG cells (Table 1A).

Because caffeine also negatively interferes with ATM/ATR (11) aside from inhibiting the PI3K/Akt pathway, we next asked whether the radiosensitizing effect of caffeine in PTEN-deficient glioma cells is mainly due to the inhibition of the PI3K/AKT pathway or whether inhibition of ATM/ATR is also involved. We therefore compared the radiosensitizing effect of caffeine with that of the specific Akt inhibitor triciribine. As shown in Fig. 5, the pretreatment of U87MG cells with triciribine (5 μmol/L) before IR with 2 Gy significantly reduced the surviving fractions of U87MG cells with an efficacy comparable with caffeine (Fig. 5).

It has been shown by numerous studies that the activation of the PI3K/Akt signaling pathway in tumor cells, either by increased growth factor receptor signaling (20, 21) or decreased activity of specific pathway inhibitors such as PTEN (14), is significantly associated with resistance to radiotherapy. We show in this study that constitutive activation of the PI3K/Akt signaling pathway in PTEN-deficient glioma cells can be targeted by caffeine, resulting in radiosensitization. Interestingly, this caffeine-mediated radiosensitization was absent in PTEN-proficient glioma cells (Fig. 1). Our results corroborate findings from a recent study that revealed the higher efficacy of several small-molecule inhibitors of PI3K in radiosensitizing glioma cells lacking wild-type PTEN (22). Of importance, not only caffeine but also the less toxic pentoxifylline selectively radiosensitized PTEN-deficient glioma cells (Table 1B), supporting the clinical relevance of our finding.

The identification of agents that preferentially radiosensitize tumor cells deficient in genes required for treatment efficacy is clinically highly relevant. This would not only increase the desired effects of therapy in tumor tissue but would also decrease the unwanted side effects because normal tissue lacking such genetic abnormalities would not be affected. It has already been shown in the past that caffeine might belong to such an interesting class of agents because it proved to effectively radiosensitize p53-deficient cells but was significantly less effective in p53-proficient cells (23, 24). Our results confirmed the p53 independency of caffeine-mediated radiosensitization (23, 24): caffeine radiosensitized U87MG cells (Fig. 1) despite its concurrent slight inhibition of p53 expression and transcriptional activity (Fig. 4) and it was able to radiosensitize the p53-deficient cell lines T98G and U373MG (Table 1A). However, our study also provides new evidence for a preferential activity of caffeine in PTEN-deficient tumors (Table 1A; Fig. 1). Our study further supports the important role of Akt rather than ATM/ATR in caffeine-mediated radiosensitization of PTEN-deficient glioma cells, and thereby corroborate conclusions from previous reports (19, 22) and a very recent study (25) that Akt represents an important therapeutic target in combination with IR in the treatment of PTEN-deficient brain tumors.

Given the ability of PTEN to indirectly stabilize p53 protein through antagonizing the Akt-MDM2 pathway (4, 10, 26) or by directly increasing p53 acetylation (27), decreased p53 activity in PTEN-deficient tumor cells should be expected and would explain the comparably preferential activity of caffeine in radiosensitizing p53-deficient (Table 1A; refs. 23, 24) and PTEN-deficient tumor cells (Table 1A; Fig. 1). However, the results from the mechanistical studies performed by us (Figs. 2 and 3) by and Powell et al. (1) suggest a distinct mode of interaction of caffeine in cells with p53- or PTEN-deficiencies: the study of Powell et al. (1) revealed a quantitative difference at the IR-induced G₂-M checkpoint and a different threshold for the caffeine-induced override of G₂-M arrest in p53-null compared with p53 to wild-type cells. In contrast, we observed a significant difference in IR-induced G₁ checkpoint activation and in the potential of caffeine to interfere at this specific DNA damage checkpoint particularly in PTEN-deficient compared with PTEN-proficient cells (Fig. 2), whereas the degree of G₂-M checkpoint activation and its well-known modulation by caffeine was the same for both cell lines in our model (Fig. 3). The results from our cell cycle studies are in accordance with the recent study of Sturgeon and Roberge (28) who showed in a very comprehensive analysis that caffeine was only able to radiosensitize breast cancer cell lines if given at the time when most cells progress from G₁ into S phase rather than at the time of maximal G₂ arrest. Furthermore, our results are in line with the study of Deplanque et al. (29) who showed in the model of normal human skin fibroblasts that the radiosensitizing effect of caffeine was associated with modulation of G₁ rather than G₂ checkpoint.

In our study, the caffeine-mediated increase of IR-induced G₁ checkpoint activation in U87MG cells did not depend on an increase in IR-induced p21^Waf1 expression. This result was somewhat surprising because constitutive activation of Akt in PTEN-deficient cells should downregulate p53 transcriptional activity and block p53-induced p21^Waf1 induction (4, 10), both of which should be rescued by PI3K/Akt inhibition by caffeine. This suggests that cell cycle regulators other than p21^Waf1 might mediate the positive effect of caffeine on G₁ DNA damage checkpoint activation in our model. Alternatively, the interaction of Akt with posttranslational modification rather than transcriptional regulation of p21^Waf1 might be targeted by caffeine: besides the indirect suppression of p21^Waf1 expression by triggering MDM2-mediated p53 degradation, Akt has been shown to directly phosphorylate p21^Waf1 within its nuclear localization sequence, thereby causing its cytoplasmic accumulation and preventing its access to nuclear cyclin-Cdk targets (30). The role of p21 mislocalization rather than reduced transcription in G₁ checkpoint defects and radioresistance of PTEN-deficient cells remains to be determined in future studies.

The negative cross-talk between Akt and p53 signaling pathways is not one sided but mutual in its nature. It is well know that the activation of p53 and its target genes such as PTEN by IR can negatively interfere with Akt activation (31). Thus, the increased reduction of p-Akt⁴³⁷ levels observed in U87MG cells irradiated in the presence of caffeine (Fig. 4A) might result from IR-induced p53 activation. However, a role of PTEN as mediator of such p53-dependent inhibition of Akt phosphorylation can be excluded in our model of PTEN-deficient glioma cells. Whether or not PHLDA3, a further p53-regulated repressor of Akt identified very recently (32), contributes to the suppression of Akt phosphorylation in PTEN-deficient glioma cells irradiated in the presence of caffeine remains to be determined in future studies.

In conclusion, we have confirmed previous studies that the PI3K/AKT signaling pathway is an attractive target for radiosensitization of glioma cells (19). We show for the first time that the methylxanthine caffeine or the less toxic derivative pentoxifylline are candidate drugs for targeting PI3K/Akt signaling particularly in PTEN-deficient glioma cells. Future direct comparison of the therapeutic ratio of caffeine or the related methylxanthine derivative pentoxifyllline with improved pharmacologic features (24) in tumor compared with normal tissue will reveal the clinical potential of this substance class in the treatment of PTEN-deficient glioma.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.