Histone deacetylase inhibitors promote glioma cell death by G2 checkpoint abrogation leading to mitotic catastrophe

Glioblastoma multiforme is resistant to conventional anti-tumoral treatments due to its infiltrative nature and capability of relapse; therefore, research efforts focus on characterizing gliomagenesis and identifying molecular targets useful on therapy. New therapeutic strategies are being tested in patients, such as Histone deacetylase inhibitors (HDACi) either alone or in combination with other therapies. Here two HDACi included in clinical trials have been tested, suberanilohydroxamic acid (SAHA) and valproic acid (VPA), to characterize their effects on glioma cell growth in vitro and to determine the molecular changes that promote cancer cell death. We found that both HDACi reduce glioma cell viability, proliferation and clonogenicity. They have multiple effects, such as inducing the production of reactive oxygen species (ROS) and activating the mitochondrial apoptotic pathway, nevertheless cell death is not prevented by the pan-caspase inhibitor Q-VD-OPh. Importantly, we found that HDACi alter cell cycle progression by decreasing the expression of G2 checkpoint kinases Wee1 and checkpoint kinase 1 (Chk1). In addition, HDACi reduce the expression of proteins involved in DNA repair (Rad51), mitotic spindle formation (TPX2) and chromosome segregation (Survivin) in glioma cells and in human glioblastoma multiforme primary cultures. Therefore, HDACi treatment causes glioma cell entry into mitosis before DNA damage could be repaired and to the formation of an aberrant mitotic spindle that results in glioma cell death through mitotic catastrophe-induced apoptosis.

Glioblastoma multiforme (GBM) is an incurable cancer due to its aggressiveness and its resistance to conventional antitumoral therapies. Multiple genetic alterations are involved in gliomagenesis leading to an aberrant activation of key pathways involved in mitogenic signaling and cell cycle control. 1,2 The intratumoral heterogeneity combined with a putative cancer stem cell subpopulation underlies the difficulty to treat this cancer. The median survival of GBM patients treated with multimodal therapies including surgical resection, radiation and chemotherapy is less than 16 months due to tumor relapse after surgical removal. 3 Histone deacetylases (HDAC) are key regulators of cell development and cancer, by deacetylating histones and other proteins. 4 Recent studies found that class I HDAC expression was high in locally advanced, dedifferentiated and strongly proliferating tumors, sometimes associated with compromised patient prognosis. 5 In contrast, a reduction in class II HDAC expression was described in different types of tumors, including GBM samples. 6 Nevertheless, HDAC inhibitors cause the acetylation of both histone and non-histone proteins and exert multiple anti-tumoral effects by inducing differentiation, apoptosis, cell cycle arrest, susceptibility to chemotherapy and inhibition of migration and angiogenesis. 7 Therefore, HDACi are widely investigated and tested as anticancer drugs. Initial clinical trials indicate that HDAC inhibitors from several structural classes are well tolerated and exhibit therapeutic activity against a variety of human malignancies, and the pleiotropic molecular mechanisms of action of these drugs are being uncovered. [8][9][10] The elucidation of the key molecular targets of HDACi involved in glioma cell death is relevant for the development of more specific therapeutic strategies.
Here, we characterize the response of glioma cell lines and primary GBM cultures to two broad range HDACi being tested in clinical trials against GBM: suberanilohydroxamic acid (SAHA, vorinostat) and valproic acid (VPA). Both drugs are able to kill glioma cells more efficiently than the chemotherapeutic drug temozolomide (TMZ). We also present the analysis of the molecular alterations associated with glioma cell death, showing that HDACi drive cells to mitotic catastrophe and cell death by apoptosis.

Results
SAHA and VPA affect glioma cell viability, proliferation and clonogenicity. On WST-1 assays, SAHA and VPA decreased cell viability in a concentration-dependent manner (Figure 1a). Only at intermediate concentrations, differences between glioma cell lines were observed, being U251-MG cells less sensitive than U87-MG cells. LC50 values (Figure 1a) showed that U251-MG has the lower sensitivity to both HDACi. Similar results were obtained by viable cell counting using trypan blue exclusion at selected HDACi concentrations (Figure 1b), being 10 μM SAHA more effective than 10 mM VPA.
Next, we analyzed the effect of VPA and SAHA on cell proliferation by Ki-67 immunodetection. Treatment with 10 μM SAHA for 24 h reduced the percentage of proliferating cells from a range of 70-90% to less than 20% in all cell lines (Figure 1c), whereas 10 mM VPA had a weaker effect and only significant in U251-MG and LN229 cells. Clonogenic assays showed that HDACi reduced significantly the capability of cells to grow clonally after a 48-h treatment, (Figure 1d).
HDACi promote nuclear condensation and the activation of the mitochondrial apoptotic pathway. Our results showed that SAHA and VPA decrease cell viability dose dependently. Therefore, to assess the participation of the apoptotic cell death pathway on this effect, we analyzed the nuclear morphology of glioma cells after 24 h of treatment. As a reference, apoptosis inducer TRAIL was added to cells at 100 ng/ml. Nuclear staining with Hoechst showed that HDACi cause nuclear condensation and abnormal nuclear morphology in some cells (Figure 2a). However, nuclear fragmentation was absent and it was very infrequent in TRAIL-treated cells. Therefore, glioma cells do not show a typical apoptotic nuclear fragmentation after VPA and SAHA treatment. In addition, HDACi induced an increase in nuclear size in U87-MG and LN229 cells, indicating that chromatin was remodeled due to histone acetylation (Supplementary Figure 1).
To determine whether HDACi induced the intrinsic apoptotic pathway, we checked the expression and cleavage of caspase 9 and caspase 3 by western blot (Figure 2b) and executioner caspase activity by a fluorogenic assay (Figure 2c). We observed that 10 μM SAHA promoted the cleavage of both Caspase 9 and Caspase 3, whereas 10 mM VPA had a weaker effect in all glioma cell lines. As expected, Caspase-3 cleavage correlated with an increase in the DEVDase activity of SAHA-treated cells (Figure 2c). Therefore, HDACi, mainly SAHA at the concentration tested, are able to promote the activation of the mitochondrial apoptotic pathway in glioma cells. HDACi used in this study did not activate the extrinsic apoptosis pathway on glioma cells (unpublished results).
HDACi promote DNA fragmentation in glioma cells, which is dependent on the activation of the caspase/ CAD pathway. Next, we checked whether the activation of the caspase cascade by HDACi promoted DNA fragmentation. We analyzed DNA integrity by Pulsed-field gel electrophoresis (PFGE) and agarose gel electrophoresis ( Figure 3a). SAHA treatment for 48 h caused the formation of 50-kb DNA fragments and the final DNA degradation in a smear pattern (Figure 3a). Interestingly, the addition of a pancaspase inhibitor (Q-VD-OPh) blocked both high-and lowmolecular weight DNA fragmentation, indicating that caspase activation was necessary for DNA degradation. Staurosporine (STS), an apoptosis inducer, was added for 24 h to U251-MG cells as a positive control for DNA fragmentation. The formation of 50-Kbp DNA fragments was observed, but typical oligonucleosomal laddering was not detected (Figure 3a). These results indicated that U251-MG cells exposed to an apoptosis inducer do not perform DNA internucleosomal fragmentation. To further corroborate the implication of the mitochondrial apoptotic pathway in SAHAinduced DNA degradation, we overexpressed the antiapoptotic gene BCL-X in U251-MG glioma cells (Figure 3b). We observed that Bcl-xL-overexpressing cells were protected against the induction of DNA degradation by SAHA, further suggesting the involvement of the intrinsic apoptotic pathway in SAHA effects.
One of the nucleases responsible for DNA processing on caspase activation is Caspase-associated DNAse (CAD). 11 Therefore, we next verified whether CAD endonuclease was implicated in SAHA effects by using two shRNAs against CAD mRNA that efficiently downregulate CAD protein levels in U251-MG cells (Figure 3d). After SAHA treatment, we observed that CAD downregulation blocked the formation of 50-Kbp and low-molecular weight DNA fragments, leading to bigger DNA fragments showing a smear pattern on the upper part of the gel. This indicates that CAD is the nuclease implicated in the formation of 50-Kbp fragments on HDACi treatment and suggests that 50-Kbp DNA processing is required for the final low-molecular weight DNA degradation. As suggested by the smear pattern that appears in PFGE experiments when CAD expression is reduced ( Figure 3c) but not when caspases are inhibited with Q-VD-OPh (Figure 3a) or on Bcl-xL overexpression (Figure 3b), other caspaseactivated nucleases may participate in DNA initial processing; however, the identity of this nuclease/s remains to be investigated.
As caspases were involved in DNA fragmentation, we next checked the relevance of caspase activation in HDACiinduced glioma cell death ( Figure 3e). Trypan blue exclusion cell counts showed that caspase inhibition did not prevent cell death, indicating that cells died even in the absence of caspase activation, thus indicating that the apoptotic pathway is not required for HDACi-induced glioma cell death.
HDAC inhibition in glioma cells promotes DNA damage partly due to ROS production. One of the mechanisms involved in the activation of the mitochondrial pathway is the activation of the DNA damage response. Therefore, we next verified whether SAHA and VPA were able to promote the formation of DNA double-strand breaks (DSBs). Cells treated with temozolomide (TMZ), an alkylating drug in current use for GBM treatment, were used as positive control (Figure 4a). SAHA and VPA promoted an increase in the percentage of γH2AX-positive cells, a marker of DSBs formation, and significant changes in γH2AX staining for both drugs were attained in U251-MG cells. As expected, TMZ induced a strong increase in γH2AX-positive nuclei in glioma cells.
As TMZ was more efficient in promoting DNA damage, we next verified the sensitivity of glioma cell lines to this drug. Surprisingly, all three cell lines showed a higher resistance to TMZ than to HDACi used in this study (Figure 4b compared with Figure 1a). In fact, at the maximal concentration of 500 μM, TMZ did not even reach a 40% decrease in cell viability of the most sensitive cell line U87-MG compared with more than 70% attained by HDACi at the highest doses.
To determine the role of caspase activation in HDACiinduced DNA damage, we next analyzed the effect of caspase inhibition in γH2AX-positive cells after SAHA treatment ( Figure 4c). Results obtained show that caspase inhibition with Q-VD-OPh did not block the formation of DSBs in genomic DNA, indicating that SAHA-induced DNA damage occurs independently of caspase activation.
It has been reported that HDAC inhibition produces an increase in reactive oxygen species (ROS) and that this could contribute to the promotion of DNA damage. We therefore analyzed ROS production in HDACi-treated cells ( Figure 4d). Both HDACi tested caused a significant increase in ROS abundance, whereas TMZ did not have this effect. ROS production correlated with an increased protein carbonylation in SAHA-treated U87-MG cells (Supplementary Figure 2A). SAHA-induced ROS generation was blocked by glutathione (GSH) addition ( Figure 4d). The mechanism involved in SAHA-induced ROS production in tumoral cells has been proposed to be by the upregulation of thioredoxin-bindingprotein-2 (TBP-2) and by the downregulation of thioredoxin activity. 12 We assessed the effect of both HDACi on TBP-2 and thioredoxin protein levels in our cell model, but we did not detect any change neither in TBP-2 nor in thioredoxin expression (Supplementary Figure 2B).
With the aim to determine whether ROS production is involved in HDACi-induced DNA damage, we checked γH2AX staining on SAHA-and TMZ-treated cells in the presence or absence of ROS inhibitors. We observed that ROS scavenging with 15 mM NAC partially prevented DNA DSBs formation in U87-MG cells and in U251-MG cells, but not in LN229 cells (Figure 4e), pointing out to the existence of an additional molecular alteration that results in DSBs formation in addition to oxidative stress.
HDACi affect cell cycle progression by decreasing G2 checkpoint kinase expression: Wee1 and Chk1. To further characterize glioma cell response to SAHA and VPA, we next analyzed cell cycle distribution by flow cytometry. We observed that, after 24 h, HDACi promoted a significant decrease of U87-MG and LN299 cells in S phase and an increase of cells in G2/M (Figure 5a). Changes on U251-MG cells reached significance after 48h of incubation (data not shown). In addition, an increase of cells in G0/G1 phase was observed in LN229 cells. The reduction of cells in S phase can be due to the promotion of cell cycle arrest or a faster progression to the G2/M phase. However, only LN229 cells showed an arrest in G0/G1. The increase of G2/M-cell population at the expense of G0/G1 and S phases suggests  Figure 3). Therefore, we next checked whether SAHA induced apoptosis in glioma cells, using Annexin V-Propidium Iodide (PI) double staining. Data presented on Figure 4e show that SAHA promoted an increase in the percentage of Annexin V-positive cells in the PI-negative population (conserved membrane integrity, left bar-graph), whereas no significant changes were observed in the PI-positive cell population, indicating phosphatidylserine exposure in the cell membrane. These results indicate that SAHA promotes apoptotic cell death in glioma cells.
The accumulation of cells in G2/M prompted us to check whether HDACi affect the expression of G2 checkpoint kinases: Chk1 and Wee1. We observed that 10 μM SAHA caused a strong decrease in Chk1 and Wee1 protein abundance (Figure 6a), correlating with a decrease in cdc2 phosphorylation on Tyr-15, which is the final effector for M-phase progression and Wee1 substrate. VPA treatment had a weaker effect and also reduced the amount of Chk1 and Wee1 in U87-MG and LN229 cells, respectively. In contrast, the treatment with TMZ, which directly causes DNA damage, did not affect the amount of Chk1 and Wee1 kinases. Interestingly, the more resistant cell line U251-MG As the expression of cdc2 was not affected by HDACi treatment, but the inhibitory phosphorylation on Tyr-15 diminished, we next checked the expression of its cyclin partner (Figure 6a). We observed that cyclin B1 decreased after HDACi treatment and the expression of the cell cycle inhibitor p21 was strongly increased concurring with published data. 13 In contrast, the amounts of cdc25A, a phosphatase involved in the regulation of G1/S and G2/M checkpoints, 14 were unaffected, suggesting that cells were not arrested through a decrease in cdc25A. As cdc25A is regulated posttranslationaly and by subcellular localization, a more detailed characterization of other cell cycle regulators in synchronized cells would be necessary to confirm whether cell cycle arrest occurs on a subpopulation. Wee1 inhibition has been reported to increase replication stalling and DSBs induced formation by the endonuclease Mus81. 15,16 Therefore, we checked whether Mus81 is implicated in HDACi-induced DNA damage. As shown in Figure 6c, none of the drugs altered the expression of Mus81. Moreover, Mus81 downregulation by shRNA did not reduce SAHA-induced DSBs formation nor low-and high-molecular weight DNA fragmentation either (Figures 6d and e), discarding the implication of this endonuclease in SAHA DNAdamaging effects. HDACi cause glioma cell death by promoting mitotic catastrophe. The fact that HDACi cause a decrease in the expression levels of Wee1 and Chk1 made us to hypothesize that cell death may occur through mitotic catastrophe due to the accumulation of DNA damage that cannot be repaired before mitosis. 17,18 Therefore, we next analyzed by confocal  (Figure 7a). However, on SAHA treatment dividing cells showed aberrant metaphase morphologies: chromosomes were not aligned at the metaphase plate, Survivin was not recruited to centromeres and an abnormal distribution of mitotic spindle was observed (Figure 7a).
Aberrant mitotic figures were formed, and no cells in telophase or cytokinesis were observed, indicating that cells cannot complete mitosis leading to cell death. In addition we observed that cells enter mitosis with damaged DNA, as chromosomes from SAHA-treated cells were labeled with the γH2AX antibody (Figure 7a, right image). In addition, HDACi promoted a decrease in the amount of Survivin and TPX2 (Figure 7b), indicating a deficit of proteins relevant for correct  mitosis completion. In addition, we determined the expression of Rad51, a protein involved in homology-directed repair of DNA DSBs. Interestingly, HDACi also reduced the expression levels of Rad51 in glioma cell lines in comparison with untreated cells and TMZ-treated cells (Figure 7a). Therefore, this indicates that HDACi impair DNA repair mechanisms in glioma cells, as it has been described in other types of tumor cells. 19 As HDACi induced a strong decrease of key proteins involved in cell cycle control and mitosis, we verified whether these effects were extensive to primary GBM cultures. Cells isolated from human GBM tumors were treated with 10 μM SAHA for 24 h (Figure 7c). Three of the four primary GBM cultures reduced the expression of Wee1, Chk1, Survivin and TPX2 after SAHA treatment. Only cells of GBM C48, which expressed very low levels of these proteins in basal conditions, did not show the same response. These results suggest that HDACi have the same molecular effects on glioma cells independently of their origin and genetic alterations, and validate the results obtained in glioma cell lines. Therefore, HDACi alter the expression of proteins necessary for tumoral cell proliferation, which have been previously proven to be involved in gliomagenesis.
Together, these results indicate that glioma cells treated with SAHA die during mitosis by a mechanism involving mitotic catastrophe. One of the initial events that have been reported during this type of cell death is the activation of Caspase-2, upstream of cytochrome c release. 20 Therefore, we checked Caspase-2 activation after HDACi treatment (Figure 7d). We observed that HDACi caused a decrease of procaspase-2 (51 kDa) protein, but no cleaved forms were detected (p32 or p18). However, Caspase 2 mRNA levels were unchanged in U251-MG and LN229 cells after HDACi treatment (table in Figure 7d). Measurement of caspase-2 activity in U251-MG cell lysates confirmed that VPA and SAHA promoted the activation of caspase 2, and that Q-VD-OPh addition blocked its activation (Figure 7e). Therefore, our results demonstrate the implication of Caspase 2 in glioma cell death induced by HDACi.

Discussion
Our aim was to characterize the molecular changes induced by broad-range HDAC inhibitors that are used for killing chemotherapy-resistant glioma cells. Here we have demonstrated that two structurally different HDACi, SAHA and VPA, have multiple effects in glioma cancer cells that result in oxidative stress, DNA damage and the activation of the apoptotic mitochondrial pathway. However, the HDACiinduced apoptotic pathway is not the main cause of glioma cell death. Rather, HDACi treatment impairs G2 checkpoint, pushes glioma cells to enter mitosis before DNA damage is repaired and decreases the expression of proteins that have crucial roles in mitosis, thus triggering cell death by mitotic catastrophe.
HDAC inhibition in glioma cell lines causes cell death in a dose-dependent manner, whereas those cells show resistance to the chemotherapeutic agent TMZ. The resistance of GBM cells to chemo-and radiotherapy has recently been attributed to the overexpression of the G2 checkpoint kinase Wee1. 18 We have shown that HDAC inhibition alters the expression of two G2 checkpoint kinases, Wee1 and Chk1, and results in cells entering mitosis before DNA-damage repair. In accordance with our results, pharmacologic inhibition of G2 checkpoint kinases Wee1 and Chk1 has been shown to have anti-tumoral effects in neuroblastoma cells both in vitro and in vivo 21 and cause DNA double-strand breaks in cancer cells in the absence of DNA-damaging chemotherapeutic drugs. 22 In addition, Wee1 inhibition sensitizes GBM cells to ionizing radiation and pushes cancer cells onto mitotic catastrophe. 18 In contrast, Chk1 inhibition in combination with a DNA-damaging drug, gemcitabine, drives rapid chromosome fragmentation followed by caspase-independent cell death. 17 In fact, Chk1 has been demonstrated to control mitotic DNA damage checkpoint by delaying mitotic exit of DNA-damaged cells by regulating mitotic catastrophe. 23 The role of Chk1 downregulation in HDACi-induced cell death has been previously demonstrated by Brazelle et al. 24 in non-small cell lung cancer cells. Nevertheless here we describe for the first time the implication of Wee1 kinase in HDACi effects. Our results demonstrate that G2 checkpoint kinases are molecular mediators of HDACi-induced cell death on GBM cells and highlight the relevance of the administration of multitarget therapies for the treatment of resistant tumors.
Thioredoxin inactivation by the induction of thioredoxinbinding protein (TBP-2) has been previously reported to be important for HDACi-induced ROS production and DNA damage in different cancer cell types. 10,25,26 However, we have not detected changes in TBP-2 expression, thus discarding this mechanism in our model. In fact, ROS production was implicated in DNA damage observed in U87-MG and U251-MG cells, but not in LN229, indicating that other molecular alterations may participate, such as alteration of the DNA-repair machinery, 19 the acetylation of histones involved in DNA repair such as Histone 4 27 or histone 3 28 and the decrease in G2 checkpoint kinases. 22 Our results show that HDACi display at least two of these effects on glioma cells, as they decrease the expression of G2 checkpoint kinases Wee1 and Chk1 and DNA damage repair protein Rad51. As the endonuclease Mus81 has been implicated in the DNA damage response induced by Wee1 depletion 16 and in Chk1-deficient cells, 29 we also checked whether Mus81 was a mediator of DSB formation and DNA fragmentation in SAHAtreated cells. However, our results indicate that Mus81 does not participate in the DNA-damaging effects of SAHA.
HDACi are able to activate the mitochondrial caspasedependent apoptosis pathway in glioma cells. However, our results indicate that this event does not determine glioma cell death, as caspase inhibition does not prevent it. The detailed analysis of cell cycle distribution and nuclear morphology rather indicates that HDACi promote cell death by mitotic catastrophe due to an aberrant mitotic spindle formation, and the detection of caspase-2 activation supports this conclusion. In accordance with our results, a previous work described that HDACi treatment temporarily delays mitotic progression affecting prometaphase due to activation of the spindle assembly checkpoint, involving aberrant chromosome segregation and failed cytokinesis. 30 Similarly, Trichostatin A has been shown to induce a delay at the G2/M transition, chromosome missegregation and multi-nucleation, and thereby it leads to cell death by promoting exit from aberrant mitosis without spindle checkpoint. 31 Our results provide an explanation to these observations, as they show that HDACi significantly reduce the expression of G2 gatekeepers Wee1 and Chk1, in addition to other proteins involved in chromosome segregation, DNA repair and mitotic spindle formation. Therefore, HDACi-treated glioma cells die by a mechanism of mitotic catastrophe involving the activation of caspase-2 and the mitochondrial pathway due to DNA-damage accumulation and G2 checkpoint impairment, according to the current concept of mitotic catastrophe. 32 Our results show that the expression of Rad51, a protein involved in DNA repair by homologous recombination, is reduced in HDACi-treated glioma cells. The same effect was described for the HDACi PCI-24781. 33 In fact, it has been reported that SAHA affects the expression of proteins involved in several mechanisms of DNA repair: homologous recombination, non-homologous end joining or base excision repair 19 in different kinds of tumor cells, such as RAD50 and MRE11 in prostate and lung cancer cells, 10 Ku70, Ku80 and RAD50 in melanoma cells 34 and Rad51 and BRCA1 in prostate cancer cells. 35 HDACi-induced impairment of the DNA repairing potential could explain why antioxidant treatment does not prevent DSB formation in LN229 glioma cells. This implies that HDACi effects on proteins involved in DNA-repair are common and can contribute importantly to DNA damage, in conjunction with ROS production.
Our results also show that Survivin, an antiapoptotic protein that has been shown to be important for the regulation of mitosis, 36 is downregulated by HDACi treatment in glioma cells. Survivin is overexpressed in most cancers and it can be used as a prognostic factor for several kinds of tumors. 37 Therefore, therapies designed to downregulate or block Survivin have been experimentally tested. A dominantnegative mutant of Survivin significantly induced mitotic catastrophe and apoptosis, and inhibited tumor growth in a colon cancer xenograft model. 38 It has also been described that cell division in the absence of Survivin results in defects in chromosome alignment, failure of cytokinesis and eventually cell death. 39 Recently, it has been shown that the disruption of the Survivin-Ran complex by a pharmacological inhibitor abolishes survival and growth of glioma stem cells. 40 Therefore, our findings together with previous data point out to the relevance of Survivin in cancer cell proliferation and support the role of this protein as a mediator of HDACi-induced mitotic catastrophe in glioma cells.
In conclusion, our results show that SAHA and VPA, which are more efficient than the chemotherapeutic drug TMZ in killing glioma cells, reduce the expression of proteins involved in processes relevant for tumor growth and survival: G2/M checkpoint gatekeepers, DNA repair mechanisms and proteins involved in mitotic spindle formation and chromosome segregation during mitosis, thus leading to cell death by mitotic catastrophe. These findings support the view that drugs targeting multiple processes are more efficient in killing cancer cells than DNA-damaging agents.
Glioma cell lines: U87-MG and U251-MG cells were obtained from CLS Cell Lines Service (Eppelheim, Germany), and LN229 cell line was from ATCC (Manassas, VA, USA). Cells were maintained in culture as indicated by the supplier at 37°C and 5% CO 2 atmosphere and were subcultured by trypsinization twice a week.
Primary cultures: Tumor biopsies were obtained from Hospital Universitari Arnau de Vilanova (HUAV; Lleida) and processed following the procedure described by Dr Seoane group. 41 Lentiviral Construction and Transduction. Human BCL-X cDNA was cloned into the expression lentiviral vector pEIGW. 42 Primers for small hairpin RNA interference (shRNA) and control plasmid DNA pLKO.1-puro-SHC002 were bought to Sigma (hMus81-1: TRCN00000049727; hMus81-2: TRCN00000290878). Human CAD shRNA were kindly provided by Dr. Victor Yuste (UAB, Barcelona). 43 Viruses were prepared and titered as described in Bahi et al. 44 Cells were usually treated after 2-3 days of viral transduction when expression vectors were used or after 4-5 days when shRNA experiments were performed.
WST-1 viability assay. Cell viability was measured using a colorimetric assay with WST-1 reagent (Roche Diagnostics GmbH, Mannheim, Germany). Glioma cells were seeded in 96-well plates (2-4 × 10 3 cells/well), and, after 24 h, cells were treated for 48 h as indicated in the figures. WST-1 reagent was added and absorbance at 450 nm and 600 nm was measured in a BioTek plate reader (BioTek Instruments, Winooski, VT, USA) every hour up to 3 h. Cell viability was calculated as percentage of absorbance readings of treated versus untreated cultures.
Clonogenic assay. Cells were plated in six-well plates at a density of 3.5-7 × 10 5 cells/well. Once cells reached 70% confluency, HDACi were added and incubated for 48 h. Afterward, cells were trypsinized, counted by trypan blue exclusion and re-plated (500 cells/well) in duplicate in six-well plates. Cells were grown at 37°C and 5% CO 2 in normal medium and left to form colonies for 10 days, with a mid-term medium replacement. At the end, cells were fixed with coldmethanol for 10 min and stained with 0.5% crystal violet solution (Sigma) for 10 min at room temperature. Plates were rinsed carefully in water and left to dry overnight at room temperature. Finally, plates were scanned and clones were counted.

Flow cytometry
Cell cycle analysis: Cells were seeded in 60-mm petri dishes at 1.25-2.5 × 10 6 cells/dish, left to adhere and treated as indicated for 24 h. Afterward, cells were trypsinized and washed in ice-cold PBS, fixed in 70% ethanol and stored at -20°C until analysis. Fixed cells were suspended in 500 μl propidium iodide (PI)/ RNase staining buffer (Becton Dickinson, Franklin Lakes, NJ, USA), incubated for 30 min at 37°C and analyzed in a BD FACSCanto II cytometer (Becton Dickinson). Data analysis was performed using ModFit LT software (Verity software house, Topsham, ME, USA).
Measure of ROS: Cells were plated in 35-mm dishes at 3.5 or 7 × 10 5 cells/dish, left to adhere for 1 day and treated with the indicated drugs for 24 h at 37 ºC and 5% CO 2 . At the end Cell Rox Green Reagent (Life Technologies) was added to a final concentration of 2.5μM and incubated for 30 min at 37 ºC. Cells were washed twice in cold PBS, trypsinized and suspended in 1ml PBS. Cells were analyzed in a BD FACSCanto II cytometer (Becton Dickinson).
Annexin V and propidium iodide staining: After 24 h of treatment with 10 μM SAHA, cells were collected and incubated with 1 μg/ml Annexin V-FITC (ENZO Life Sciences, Farmingdale, NY, USA) and 2.5 μg/ml PI (BD Transduction Laboratories, Becton Dickinson, Franklin Lakes, NJ, USA) in 200 μl buffer containing 10 mM Hepes/KOH (pH 7.4), 140 mM NaCl and 2.5 mM CaCl 2 for 30 min at room temperature, and the cells were analyzed using a BD FACSCanto II cytometer (Becton Dickinson).