Low oxygen tension reverses antineoplastic effect of iron chelator deferasirox in human glioblastoma cells

Under experimental normoxic conditions, treatment with 10 μM of DFX for 3 days significantly inhibited proliferation of both U251 (Fig. 1a) and U87 (Fig. 1d) cells in comparison to control cells. The repression of proliferation was linked to high LDH release (Fig. 1b and e) and high caspase 3 activity (Fig. 1c and f), suggesting that the antiproliferative effect of DFX is probably mediated by cell death, including necrosis and apoptosis. In addition, phase-contrast microscopy photography clearly showed damage to the cells (Additional file 1: Figure S1). However, despite the ability of DFX to repress the proliferation at the same rate in glioblastoma cells, it led to more LDH release and more caspase 3 activity in U251 cells than in U87 cells. Importantly, when ferric ammonium sulfate (FAC) as Fe3+ donor was simultaneously added to the cell culture medium with DFX in stoichiometric proportion (i.e. 5 μM since two molecules of the tridentate DFX chelate one molecule of Fe3+), the antineoplastic effect of DFX was completely abolished, and, used alone, FAC failed to modulate the markers. This result confirms that the antineoplastic effect of DFX can be ascribed to its ability to chelate iron, thus emphasizing that iron is a key molecule for biological and biochemical processes.

Since radiotherapy is the main therapy for GBM to date, the same treatment was performed in combination with external irradiation. Sixteen hours after the beginning of DFX treatment, external X-ray beam radiation was applied at a dose of 16 Gy. Three days after irradiation, a significant decrease in proliferation was observed in U251 cells (Fig. 1a). Moreover, the impact of irradiation on U251 cells was correlated with the induction of LDH release (Fig. 1b) and caspase 3 activity (Fig. 1c). However, importantly, irradiation had no significant effect on U87 cells (Fig. 1d and f). These results suggest that U251 cells are more radiosensitive than U87 cells. These observations are consistent with previous work and may be explained by the fact that U251 cells have less DNA damage repair activity of Ape1 than in U87 cells [13].

Upon irradiation, DFX conserved its intrinsic characteristic of an antineoplastic agent in both cell lines, i.e. a repressor of proliferation, through increased LDH release and caspase 3 activity, compared to control irradiated cells (Fig. 1). This result showed that irradiation did not impair the activity of DFX in vitro. Importantly, irradiation did not overload the intrinsic antineoplastic activity of DFX nor generate a synergistic effect with DFX. It is important to note that DFX led to more caspase 3 activity at 0 Gy than at 16 Gy, demonstrating the potent apoptotic activity of this molecule (Fig. 1c and f). In addition, iron supplementation upon irradiation did not modulate the response in terms of proliferation, toxicity and apoptosis in the two cell lines, showing that excess of exogenous iron did not affect cell sensitivity to X-rays (Fig. 1).

As a hypoxic environment is frequently encountered in GBM due to the presence of areas of necrosis [14] and since the presence of hypoxic areas in GBM is correlated with the aggressive phenotype [15], we decided to perform the same treatment in a mild-hypoxic brain tumor environment. Surprisingly, the antiproliverative, cytotoxic and apoptotic effect of DFX was completely lost when cells from the two cell lines were cultivated in 3 % of oxygen (Fig. 2). Only the effect of the external irradiation at a dose of 16 Gy was conserved in U251 cells (Fig. 2a-c). To illustrate these results, no major cells damages were observed at 3 % of oxygen as shown in phase-contrast microscopy photography (Additional file 1: Figure S1). However, DFX led a slight significant induction of LDH for U251 cells (Fig. 2b) and caspase 3 activity for U87 cells in non-irradiated and irradiated conditions (Fig. 2f).

Importantly, same result was observed in human colon carcinoma, HCT116 cells in terms of proliferation where DFX treatment (10 μM for 3 days) led to inhibition of HCT116 proliferation under experimental normoxic conditions but was lost in in vitro hypoxic conditions (Additional file 2: Figure S2).

A mechanism by which iron chelators exert their antineoplastic effects is by targeting iron dependant proteins that are key component in the progression of the cell cycle, such as ribonucleotide reductase [2], causing cell cycle arrest. Therefore, analysis of U251 and U87 cell cycles has been performed in absence or presence of DFX in combination or not with an irradiation scheme (Fig. 3 and Additional file 3: Figure S3). Under experimental normoxic conditions, treatment with 10 μM of DFX for 3 days induced an important accumulation of U251 cells in S phase and a slight accumulation in G2-M phase (Fig. 3a). Concerning U87 cells, 10 μM of DFX induced an important accumulation of cells in G2-M phase and a slight accumulation in S phases (Fig. 3c). This increase in the GBM cell number in S and G2-M phases was correlated to a diminution in the number of cells in G0-G1 phase (Fig. 3a and c). Importantly, the capacity of DFX to profoundly affect cell cycle distribution under experimental normoxic condition is lost at 3 % of oxygen (Fig. 3b and d) except for U251 cells where a slight accumulation in phase S is still observed at 3 % of oxygen (Fig. 3b).

Three days after irradiation, elevated accumulation of cells in the G2-M phase of the cell cycle is observed for both GBM cells and both conditions of oxygenation and accumulation in S phase is also observed but only for U251 cells (Fig. 3). Interestingly, although irradiation did not affect significantly U87 proliferation (Fig. 1), an accumulation in G2-M phase is observed (Fig. 3c and d).

Finally, when irradiation was combined to DFX, at 21 % of oxygen, cell cycle profiles are very similar to the one obtained with DFX alone (Fig. 3a and c and Additional file 3: Figure S3). In contrast, at 3 % of oxygen, the combination of irradiation with DFX resulted in a cell cycle profile rather similar to the one obtained with radiation alone, while DFX was even capable to slightly reduce the impact of radiations on cell accumulation in G2-M phase.

In this last condition, the increase in S phase, particularly for U251 cells, revealed an effect of the iron chelator that was initially hidden in the overall analysis of cell proliferation (Fig. 3b and d and Additional file 3: Figure S3). As such, DFX seems to mainly interfere with S-phase, which has already been related to the specific expression of the ribonucleotide reductase [2], requiring iron for its activity which catalyzes the rate limiting step in the production of deoxyribonucleotides needed for DNA synthesis. Meanwhile, irradiation leads more a G2-M blockade, important phase for DNA damage checkpoint upstream DNA repair or cell death. However, whatever their intrinsic impact, nearly abolish for DFX used alone at 3 %, DFX and irradiation did not presented any synergism nor on proliferation or on the cell cycle.

To better understand why the effect of DFX is lost at 3 % of oxygen in vitro, iron dosage was performed. The iron contents measured in both GBM cells are in line with values found in primary rat astrocytes (9.3 ± 1.2 nmol per mg protein) [16] (Fig. 4). DFX treatment at 10 μM for 3 days in normoxic condition led to significant intracellular iron depletion in both cell lines (Fig. 4a and c). Moreover, irradiation significantly repressed intracellular iron contents but only in U251 cells in normoxic condition (Fig. 4a). Importantly, no repression of iron content by DFX was measured with 3 % O₂ although there was no difference between the total concentration of iron in the two oxygenation conditions (Fig. 4b and d). This result suggests that the effect of DFX on intracellular iron depletion observed at 21 % of oxygen is absent at 3 % of oxygen. Since the antineoplastic effect of DFX is due to its ability to chelate iron, as FAC addition abolishes the toxicity of DFX, this could explain the absence of antineoplastic effect of DFX at 3 % of oxygen. As oxidation is less frequent in hypoxic condition than in normoxic condition [17], it is possible that at 3 % of oxygen, more Fe2+ is present than Fe3+. As DFX is selective for Fe3+ chelation [18], this hypothesis could explain the loss of DFX effect in mild-hypoxic condition. In order to confirm this hypothesis, an iron dosage to distinguish the ferrous (Fe2+) and ferric (Fe3+) forms should be used to get round the limit of detection of these two forms of iron.

The loss of antineoplastic activity of DFX in an in-vitro mild-hypoxia environment could raise the question of the benefits of using iron chelators in anticancer therapy where the tumor is often hypoxic [12]. The synthesis of iron chelators that efficient only in hypoxic cells, such as bioreductive prodrugs [17], might overcome this limitation. Another alternative would be the synthesis of iron chelators that are only active in cancer cells. This has already been achieved in neurodegenerative diseases where iron accumulation has been clearly linked to these diseases [4]. In Alzheimer’s disease, the finding that acetylcholinesterase (AChE) colocalizes with amyloid-β and accelerates its aggregation has led to the development of a new class of selective AChE inhibitors with site-activated chelating activity. The prochelator HLA20A exhibits low affinity for metal ions, but can be activated following AChE cleavage to liberate an active chelator and an AChE inhibitor [19]. The active molecule HLA20 possesses neuroprotective properties both in vitro and in vivo with the ability to inhibit β-amyloid aggregation induced by metal ions [19]. An additional strategy could be to take advantage of cancer cell metabolism to build a site-directed iron chelator. In this context, the synthesis of a new generation of iron chelators such as quilamines has produced encouraging results [20]. Quilamines are linked to linear polyamine vectors that use the polyamine transport system, which is overexpressed in most cancer cells [20].

Another dilemma linked with iron chelators is that iron depletion may inhibit prolyl hydroxylase domain (PHD) enzyme activity and therefore activate the hypoxia-inducible factor-1 (HIF-1) pathway, which is implicated in tumor aggressiveness and invasion [21]. Indeed, in both GBM cells, HIF-1α protein stabilization has been observed following DFX treatment in normoxic condition (Fig. 5). Importantly, the mild-hypoxic condition led to stabilization of HIF-1α in U87 cells (Fig. 5b) but not in U251 cells (Fig. 5a), suggesting that in this GBM cell line, an oxygen concentration of 3 % is not low enough to observe intracellular accumulation of HIF-1α. However, at this low level of oxygen, DFX was still able to stabilize the HIF-1α protein in U251 cells and the same trend was observed in U87 despite the fact that the stabilization was largely masked by the high level of HIF-1α in these low oxygen conditions. Moreover, irradiation did not significantly impact HIF-1α expression or DFX-mediated HIF-1α stabilization in either type of cell. Finally, HIF-1α protein stabilization mediated by DFX observed in all the conditions tested (Fig. 5), suggests that the difference in the antineoplastic effect by DFX observed at 21 and 3 % of oxygen should not be attributed to HIF-1 signaling. The consequence of this HIF-1α stabilization could be dramatic for anticancer therapy. Indeed, it has been shown that iron depletion by desferrioxamine (DFO), which leads to HIF-1 activation and expression of urokinase plasminogen activator (uPAR) and matrix metallopeptidase 2 (MMP2), enhanced invasion of GBM cells by degrading the extracellular matrix (ECM) [22]. In addition, in clinical trials with DFO, toxicities including edema were reported [23], probably due to its ability to increase the potent angiogenic factor vascular endothelial growth factor (VEGF) [24], a well characterized HIF-1 target gene. Taken together, these data underline the need for caution when using chelators in cancer therapy, particularly for tumors with high invasive potential.

Moreover, it is still difficult to predict if iron chelation is able to affect all the GBM tumor margins and how homogeneously tumor mass rapid dividing cells and glioblastoma stem cells (GSC) would respond to iron chelation. Concerning GSC, chelation therapy may have an impact on cancer stem cells since it has been recently shown that iron dependency is enhanced in GSC [25]. However, it is important to notice that GBM tumors are composed of GSC plastic cells (defined as proliferative, symmetrically dividing and less invasive cells) and GSC rigid cells (defined as quiescent, asymmetrically dividing and more invasive cells) [26]. Since iron chelators preferentially target cells with high proliferative capacity, DFX might not be selective for GSC rigid cells responsible for tumor recurrence. Taken together, impact of DFX on GBM cells and GSC warrants further investigation.

Another limit to using iron chelators in brain tumors is the difficulty in crossing the blood–brain barrier (BBB). In neurodegenerative diseases, Novartis claims improved penetration of DFX into the brain through the co-administration of an efflux protein inhibitor in the patent US20090306160A1. Among other alternatives, drug delivery nanosystems derived from nanotechnologies are perhaps the most appropriate and potentially the most useful in this biological context [27]. Drugs encapsulated in nanoparticles may be more soluble, and have improved biological barrier crossing properties and better controlled release kinetics, with substantial clinical advantages including dose reduction, prevention of side effects and improvement of bioavailability within the targeted tumor cells [27]. Such nano-objects can either be implanted inside the tumor or within the resection cavity or, alternatively, delivered via the blood to the CNS tumor site. Brain locoregional active targeting by direct infusion by convection-enhanced delivery (CED) into the brain could lead to a major breakthrough in efficacy while allowing optimum specificity and safety [28].

Concerning iron chelation therapies, some studies of nano-carriers are currently being conducted for use in neurodegenerative diseases [29‐31]. Conjugating a derivative of deferiprone with nanoparticles did not alter its ability to chelate iron. This nano-deferiprone analog conjugate was shown to be able to inhibit amyloid-β aggregation in vitro and to protect neuronal cells from amyloid-β-associated neurotoxicity [30, 31]. DFX has been conjugated to lactoferrin, which was able to cross the BBB via its receptors. The neuroprotective effects of this nano-object have been assessed in vitro and in vivo. The results revealed a significant reduction in learning deficits induced by amyloid-β injection in a rat model of Alzheimer’s disease [29]. However, none of these objects have yet been tested in the context of brain tumors, including in GBM, and locoregional application should probably be improved.