Enhanced inhibition of clonogenic survival of human medulloblastoma cells by multimodal treatment with ionizing irradiation, epigenetic modifiers, and differentiation-inducing drugs

Medulloblastoma (MB) is the most common pediatric brain tumor. Current treatment regimes consisting of primary surgery followed by radio- and chemotherapy, achieve 5-year overall survival rates of only about 60 %. Therapy-induced endocrine and neurocognitive deficits are common late adverse effects. Thus, improved antitumor strategies are urgently needed. In this study, we combined irradiation (IR) together with epigenetic modifiers and differentiation inducers in a multimodal approach to enhance the efficiency of tumor therapy in MB and also assessed possible late adverse effects on neurogenesis. In three human MB cell lines (DAOY, MEB-Med8a, D283-Med) short-time survival (trypan blue exclusion assay), apoptosis, autophagy, cell cycle distribution, formation of gH2AX foci, and long-term reproductive survival (clonogenic assay) were analyzed after treatment with 5-aza-2′-deoxycytidine (5-azadC), valproic acid (VPA), suberanilohydroxamic acid (SAHA), abacavir (ABC), all-trans retinoic acid (ATRA) and resveratrol (RES) alone or combined with 5-aza-dC and/or IR. Effects of combinatorial treatments on neurogenesis were evaluated in cultured murine hippocampal slices from transgenic nestin-CFPnuc C57BL/J6 mice. Life imaging of nestin-positive neural stem cells was conducted at distinct time points for up to 28 days after treatment start. All tested drugs showed a radiosynergistic action on overall clonogenic survival at least in two-outof-three MB cell lines. This effect was pronounced in multimodal treatments combining IR, 5-aza-dC and a second drug. Hereby, ABC and RES induced the strongest reduction of clongenic survival in all three MB cell lines and led to the induction of apoptosis (RES, ABC) and/or autophagy (ABC). Additionally, 5-aza-dC, RES, and ABC increased the S phase cell fraction and induced the formation of gH2AX foci at least in oneout-of-three cell lines. Thereby, the multimodal treatment with 5-aza-dC, IR, and RES or ABC did not change the number of normal neural progenitor cells in murine slice cultures. In conclusion, the radiosensitizing capacities of epigenetic and differentiation-inducing drugs presented here suggest that their adjuvant administration might improve MB therapy. Thereby, the combination of 5-aza-dC/IR with ABC and RES seemed to be the most promising to enhance tumor control without affecting the normal neural precursor cells.


Background
Medulloblastoma (MB) is the most common malignant brain tumor (WHO°IV) in children aged < 15 years [1]. The MB standard therapy consists of primary tumor surgery followed by radiation therapy and/or chemotherapy. Current therapy regimes are mainly guided by tumor histology, metastasic disease, extend of resection, and patient age. New trials additionally include molecular factors like ß-catenin protein and MYC/MYCN gene amplification status. Patients with metastases receive a more intensive treatment compared to metastases-free patients. No radiation therapy but a more aggressive chemotherapy is given to children below 4 years to avoid radiation-related adverse late effects, like neuroendocrine and neurocognitive deficits. The 5-year overall survival of approximately 60 % implies the urgent need of improved antitumor therapies to enhance the outcome especially of high-risk patients (infants, metastatic disease:~55 % of all MB patients).
During embryonal development, MB arises from neural precursor cells in different zones of the rhombic lip, wherefrom they are growing into the cerebellum or the brain stem [2]. It is widely believed that tumor formation is initiated by genetic, gene-regulatory, or epigenetic abnormalities, which inhibit the normal neuronal or glial differentiation [3]. In 70-90 % of primary MBs, hypermethylation of gene promotors of tumor suppressor genes (TSG) is observed, which leads to their inactivation and, finally, to unrestricted proliferation and blockage of apopotosis [3]. Hence, the application of approved epigenetic modifiers, like 5-aza-2′-deoxycytidine (5-aza-dC, decitabine), valproic acid (VPA), or suberanilohydroxamic acid (SAHA, Vorinostat®), which have been shown by us [4] and others [5][6][7][8] to demethylate TSG, seems to be a suitable approach to inhibit tumor cell growth. These substances induce a cell cycle arrest at G2/M [9][10][11], where cells are most radiosensitive (reviewed in [12]). Therefore, synergistic effects with IR might be anticipated. Besides, MBs are mostly poorly differentiated tumors [13,14] containing 6-21 % potential tumor stem cells (TSC) [15], which are often chemo-and radiotherapy-resistant (reviewed in [16]) and held responsible for tumor relapse (reviewed in [17]). Differentiation-inducing drugs like all-trans retinoic acid (ATRA), abacavir (ABC), or resveratrol (RES) are applied in this study for their potential to induce the maturation of MB tumor stem cells and, thereby, to suppress their cancer-forming capacities (previously described in [18]). Besides, ATRA is able to inhibit MB cell growth by suppression of the OTX2 (orthodenticle homeobox 2) gene, which is overexpressed in the majority of medulloblastomas [19]. In addition, tumor cell-specific activity might be exerted by ABC through telomerase inhibition, which is enhanced in 70 % of MBs [20] and by RES through its Cu(II)dependent oxidative effects, which may induce cell death especially in low pH environments. As tumor cells are often acidic due to their glycolysis dependence and have elevated copper levels, they might be more sensitive to RES than normal cells [21].
Recently, we tested single and combinatorial effects of the de novo methyltransferase (DNMT) inhibitor 5-aza-dC with IR [4] or with other epigenetic/differentiationinducing drugs on the metabolic activity and reproductive survival of human MB cells [18]. Here, we combined for the first time IR, an integral part of MB standard therapy in children > 4 years, with 5-aza-dC and previously evaluated [18] drugs (VPA, SAHA, RES, ABC, ATRA) in a multimodal approach to further enhance the efficiency of MB tumor therapy. As brain tumor treatment is known to induce cognitive deficits by inhibition of hippocampal neurogenesis [22], also potential adverse effects have been evaluated using a murine entorhino-hippocampal slice culture model.

Methods
All methods were performed at room temperature unless otherwise noted.

Cell lines
The human MB cell line MEB-Med8a was kindly provided by Prof. T. Pietsch (Department of Neuropathology, University of Bonn Medical Centre, Bonn, Germany). The MB cell lines D283-Med and DAOY were purchased from ATCC cell biology collection (Manassas VA, USA). DAOY cells are TP53-mutated, whereas MEB-Med8a and D283-Med cell lines are wild-type TP53 [29,30].

Animals
Mouse breeding was performed in the animal facility of the Faculty of Medicine, University of Leipzig according to European (Council Directive 86/609/EEC) and German (Tierschutzgesetz) guidelines for the welfare of experimental animals. Nestin-CFPnuc C57BL/J6 mice, generated by Encinas et al. [31] were housed in a 12 h/12 h light/ dark cycle with access to food and water ad libitum.

Single drug and multimodal treatment of cell lines
To examine concentration effects of the single drugs, cells were seeded in cell culture flasks and five different concentrations of the modulator were added after 24 h and again at medium renewal 4 days after seeding. For combination experiments, cells were seeded and, 24 h later, treated simultaneously with IC 90 of 5-aza-dC and a second drug (concentrations see Table 1). After three treatment days, cells were irradiated with 2 or 8 Gy, medium was renewed, and 5-aza-dC and the second drug were supplemented. After overall six treatment days, cells were seeded for clonogenic assay.

Single drug and multimodal treatment of tissue slices
To assess the neurotoxic potential of the most promising multimodal settings, murine enthorino-hippocampal tissue slices were treated (day 0) with 0.5 μM 5-aza-dC, 40 μM RES or 250 μM ABC, corresponding to the maximal IC 90 in the three cell lines. After 72 h, single dose irradiation with 8 Gy was performed, medium was renewed, and the drugs were supplemented for additional 72 h as done for the clonogenic assay. Ethanol control contains 0.023 % ethanol which corresponds to the ethanol concentration in the resveratrol treated samples.

Irradiation
Cell/slice cultures were irradiated with a single dose using a 150 kV X-ray machine (DARPAC 150-MC) at a dose rate of 0.86 Gy/min.

Relative number of vital cells
To determine the number of vital cells, cells were counted using the trypan blue exclusion assay. The relative number of vital cells is calculated from vital cell counts after treatment normalized to the initially seeded vital cell number.

Cell cycle distribution
To examine changes on cell cycle distribution 24 h and 72 h after treatment, cells were harvested, washed twice with PBS, and fixed with 70 % ethanol at 4°C overnight. Afterwards, cells were washed twice in PBS, incubated with RNAseA solution (0.1 mg/ml) 20 min at 37°C, and stained with propidium iodide (50 μg/ml) 5 min at 4°C. The DNA staining by propidium iodide and the cell cycle distribution were then assessed by flow cytometry (Beckman coulter, EPICS XL) using EXPO32 software.

DNA double-strand break detection
To detect DNA double-strand breaks, nuclear staining of gH2AX repair protein was executed after 1 h/24 h treatment or 1 h/24 h after IR. Cells were seeded on chamber slides (DAOY, MEB-Med8a; adherent cells) or in 6-well cell culture plates (D283-Med, suspension cells) and irradiated (8 Gy) or treated with IC 90 of 5-

Apoptosis
Apoptotic cells were determined using the FAM-FLICA™ in vitro Caspase 3/7 Detection Kit (Immunochemistry Technologies) according to the manufacturers' instructions. In brief, 18 h after IR with 2 Gy or treatment with IC 90 of 5-aza-dC, RES, or ABC (concentrations see Table 1), cells were harvested and stained with FLICA™ solution for 60 min at 37°C. After washing three times, cells were counterstained with propidium iodide (2 μg/ ml) and analyzed by flow cytometry (Beckman coulter, EPICS XL).
In addition, the PI staining of sub-G1 cells (see cell cycle distribution in method section) was quantified as a second detection method for apoptotic cells.

Autophagy
Autophagy was analyzed with the Cyto-ID® Autophagy Detection Kit (ENZO), according to the manufacturers' instructions. The CytoID® detection reagent, a cationic amphiphilic tracer (CAT) selectively labels pre-autophagosomes, autophagosomes and auto(phago)lysosomes with negligible staining of lysosomes. Rapamycin and chloroquine were applied as positive and negative controls. Cells were harvested 18 h after irradiation (2 Gy) or treatment with IC 90 of 5-aza-dC, RES, or ABC (concentrations see Table 1) and stained with Cyto-ID® solution for 30 min at 37°C before flow cytometric analysis.
The autophagy activity factor (AAF) was calculated as follows: AAF autophagy activity factor MFI mean fluorescence intensity

Clonogenic survival
Subsequently to cell treatment, vital cells were counted using trypan blue exclusion test and seeded at two different cell densities in triplicates in 6 well cell culture plates. Normal medium without modulators was added. Ten to 14 days later, colonies were washed with PBS, fixed with ice-cold ethanol/acetone (1:2) for 10 min, stained with Giemsa solution (1:2 with distilled water) for 5 min, and washed with distilled water. Colonies with > 50 cells were counted; the quotient of counted colonies to seeded cells determined the plating efficiency (PE). The ratio between PE of treated cells and PE of untreated cells represented the clonogenic surviving fraction (SF) of seeded cells. The overall clonogenic survival was calculated from the relative number of vital cells multiplied with the SF.

Statistics
Statistical analysis of two treatment groups within a single cell line or in tissue slice experiments was conducted with parametric two-tailed, paired Student's t-test using Microsoft Excel 2003 software. The number of experiments/number of animals executed is given as "n". Joint analysis of the three cell lines was conducted with the non-parametric Wilcoxon rank sum test using SigmaPlot 11.0. In the joined analysis "n" represents the number of cell lines. Values are presented as mean ± SEM, if not otherwise noticed. P-values ≤ 0.05 (*; # ) and ≤ 0.01 (**; ## ) were considered as statistically significant; p-values ≤ 0.001 (***; ### ) as highly statistically significant. Drug interaction analyses regarding synergistic or additive effects were conducted as described previously [18] using the Bliss independence (BI) model which is based on the non-interaction theory [35,36].

Short-and long-term survival of tumor cells
First, we recorded the short-time survival of human MB cells after treatment with single modulators for 6 days using the trypan blue exclusion assay. The rise of vital untreated cells after 6 days compared to initially seeded cells (D283-Med, 16-fold; MEB-Med8a, 41-fold; DAOY, 83-fold) reflects the different growth rates of the three MB cell lines (Fig. 1). Although, short-time cytotoxic effects of the tested drugs may reduce the tumor bulk, some (tumor stem) cells might not be targeted and lead to quick tumor regrowth. Otherwise, cells might grow well in short-term assays but lose their colony-forming ability after a few cell divisions due to the accumulation of DNA damages. Therefore, surviving cells were seeded for clonogenic assays to examine the long-term reproductive survival 10-14 days later (Fig. 2). The plating efficiencies (PE) of untreated cell lines were very similar, between 0.49 ± 0.05 (MEB-Med8a) and 0.56 ± 0.1 (DAOY) (data not shown). From single drug clonogenic assays an inhibitory concentration by which 90 % of clonogenic cells were killed (IC 90) was calculated for each drug (listed in Table 1). If not otherwise noted, these concentrations were used for combination experiments presented individually in Fig. 3 and for joined analysis of all three MB cell lines in Fig. 4. Detailed statistical analysis of overall clonogenic survival is summarized in Table 2.

Valproic acid (VPA)
Relatively high doses of VPA (>1 mM) were required to reduce the short-and long-time survival in all MB cell lines (Figs. 1b and 2b). Combination of 5-aza-dC and VPA showed additive effects of VPA in D283-Med, synergism in MEB-Med8a, but no further reduction in DAOY on overall clonogenic survival compared to 5aza-dC alone. VPA acted radiosynergistically in D283-Med and MEB-Med8a and radioadditively in DAOY cells. Addition of VPA to 5-aza-dC and IR (2 Gy) led to synergistic effects in MEB-Med8a only, reducing the overall clonogenic survival further by 7-fold ( Fig. 3a; significances in Table 2). Joint analysis of all MB cell lines (n = 3) revealed that VPA alone induced a 3.7-fold decrease (SF = 0.27 ± 0.13; p ≤ 0.001) and combined with 2 Gy and 5-aza-dC, a 104fold decrease (SF = 0.01 ± 0.007; p ≤ 0.001) of overall clonogenic survival compared to untreated control ( Fig. 4a; significances Table 2).

Suberanilohydroxamic acid (SAHA)
The maximal tolerated SAHA plasma concentration in humans of 2 μM [37] reduced the relative number of vital cells only by 46-77 % compared to untreated cells (Fig. 1c). In clonogenic assays, the IC 90 was reached only in D283-Med cells. In MEB-Med8a and DAOY cells SAHA (2 μM) resulted in a SF of 0.29 and 0.18 (Fig. 2c). The combination of SAHA and 5-aza-dC revealed synergistic effects in D238-Med and additive effects in MEB-Med8a and DAOY on overall clonogenic survival. Combined with IR, SAHA showed radiosynergistic effects in D283-Med and DAOY and radioadditive effects in MEB-Med8a. In 5-aza-dC/IR-treated cells, SAHA induced additive effects in D283-Med cells and synergistic effects in MEB-Med8a and DAOY cells. Thereby, it further reduced the overall clonogenic survival of  Table 1) for 3 days pre-and 3 days post-IR (2 Gy). Clonogenic survival was examined 10-14 days later. Results after 8 Gy irradiation were similar (Table 2). Data were normalized to the untreated, nonirradiated control (=1) and relative values presented as mean ± SEM from sextuplicates. Statistical significances of treatment groups vs. control (0 or 2 Gy) are indicated by asterisks (*, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001) and for combination of all treatments (white bar) vs. 5-aza-dC + control by hash (#, respectively). Additional information about significant synergism/additivity and intergroup significances is given in Table 2 Fig  Table 2).

Abacavir (ABC)
The ABC treatment diminished the relative number of vital cells (Fig. 1d) and the overall clonogenic survival (Fig. 2d) Table 2).

Mechanistic studies of IR, 5-aza-dC, RES, and ABC in tumor cells
The combination of IR and 5-aza-dC with RES or ABC reduced the clonogenic survival of human MB cells most efficiently compared to any other single or combined drug treatment tested. To reveal the mechanism of action we analyzed apoptosis, autophagy, cell cycle distribution, and DNA double-strand break repair after (See figure on previous page.) Fig. 4 Joint analysis of clonogenic survival in all three cell lines after combined treatment with 5-aza-dC, IR, and a) VPA, b) SAHA, c) ATRA, d) ABC, e) RES . Relative data (untreated, nonirradiated control = 1) are presented as mean ± SEM, n = 3. Statistical significance is presented vs. control (0, 2 or 8 Gy, black bars) by asterisks (*, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001) and for multimodal combination (white bar) vs. 5-aza-dC + control (dark grey bar) by hash (#, respectively) to clarify (1) the effects compared to irradiation alone and (2) the enhancement of the previously described effect of 5-aza-dC/IR [4]. Additional information about significant synergism/additivity is given in Table 2. Drug concentrations are shown in Table 1  Table 2 Statistics -Clonogenic survival after combinatorial treatment P-values were calculated by two-tailed, paired Student's t-test in the single cell lines or by non-parametric Wilcoxon rank-sum test in the joint analysis of all cell lines. The x-fold reductions of overall clonogenic survival compared to the indicated reference group are listed. Combinatorial effects conducted with the Bliss independent model are indicated as A = additive or S = synergistic these most promising treatments using the cell line specific IC 90 of of 5-aza-dC, RES, or ABC (see Table 1) and IR at 2 Gy.

Vital cell count, Apoptosis, and Autophagy
The joint analysis of all three MB cell lines revealed a significant reduction of vital cell count 18 h after IR (2 Gy) or treatment with 5-aza-dC, RES, or ABC by 25, 19, 37, or 23 %, respectively (n = 3, p ≤ 0.05. Fig. 5a). Responses were similar in all three cell lines, except for DAOY cells being the least sensitive for ABC and IR treatments.
Treatments led also to a significant increase of apoptotic cells in caspase 3/7 induction assays (Fig. 5b) and sub-G1 phase quantification (Fig. 6b) in joint analyses (n = 3, p ≤ 0.05 to p ≤ 0.01). The strongest effects on caspase 3/7 induction (18 h) and on sub-G1 apoptotic cell fraction (24/72 h) were observed after RES treatment and IR resulting in a significant 2.5-fold and 1.8-fold  Table 1) or irradiation with 2 Gy. Data were normalized to the untreated control group and presented as mean ± SEM, n = 3. Statistical significances are presented vs. control by asterisks (*, p ≤ 0.05; **, p ≤ 0.01). The joint analysis included the relative data of the three MB cell lines Autolysosome staining to assess autophagy induction showed significant enhancements only after ABC or IR treatment compared to control cells (joint analysis, n = 3, p ≤ 0.05, Fig. 5c). Most pronounced effects were detected in MEB-Med 8a cells (29-fold after IR, p ≤ 0.05).

Cell cycle distribution
Joint analyses of the three MB cell lines revealed a significant increase of sub-G1 apoptotic cell population in all treatment groups after 72 h (n = 3, p ≤ 0.05 to ≤ 0.01). This effect was most pronounced after IR (2 Gy) and RES treatment (Fig. 6b). 5-Aza-dC (72 h) and RES (24 h) additionally induced a significant enriched S phase fraction with a concomitant drop in G2/M cell fraction. Similarly, 24 h after treatment with ABC, a significant enrichment of S phase cells could be observed, though without an impact on the G1 or G2/M cell fraction. Only IR caused a significant G2/M arrest.
Within the cell lines, D283-Med cells showed the strongest response, all single drugs induced a G1 arrest accompanied by a decrease of G2/M cell fraction 72 h after treatment.

DNA double-strand breaks
The IR-induced G2/M arrest goes along with significant enhancement of gH2AX DNA repair protein expression 1 h after IR (2 Gy) in all three MB cells (Fig. 7). The number of gH2AX foci felt below control level after 24 h indicating complete repair of IR-induced DSBs in surviving cells. Also RES and ABC are able to induce gH2AX foci at least in two of the three cell lines, whereas 5-aza-dC induced the formation of gH2AX foci only in D283-Med cells.

Neurotoxic studies in neural progenitor cells
Nestin-positive neural progenitors were analyzed by lifeimaging microscopy of cultured murine hippocampal  Table 1) or 1 h and 24 h after irradiation with 2 Gy. a Representative images of gH2AX (red) and DAPI (blue) staining in DAOY cells; scale bar = 12 μm. b Treatment-induced gH2AX-foci relative to control cells in the single MB cell lines. At least 50 nuclei were counted with a maximum of 30 foci per cell. Experiment was performed once and data are presented as mean ± SEM. Statistical significances are presented vs. control by asterisks (*, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001) slices from transgenic nestin-CFPnuc C57BL/J6 mice to examine potential neurotoxic effects of the most promising multimodal approach (5-aza-dC + IR and RES or ABC) (Fig. 8). To detect short-and long-term effects, analysis were conducted at day 7, 14, 21 and 28 after treatment start. Administration of 5-aza-dC and IR combined with RES or ABC did not reduce the nestin-positive neural progenitor cell pool compared to untreated control.

Discussion
Despite of surgery, irradiation, and chemotherapy, some remaining therapy-resistant tumor cells may lead to tumor relapse. Here, we follow the hypothesis that tumor control might be enhanced by a multimodal treatment approach, exploiting some of the hallmarks of cancer (detailed in [38]). Based on the positive results of our previous studies investigating the effects of IR and 5-aza-dC [4], and of 5-aza-dC with additional modulators [18], we investigated here for the first time the effect of a triple combination: IR, 5-aza-dC and one additional epigenetic modifier (SAHA, VPA) or differentiation-inducer (RES, ABC and ATRA).
To assess the impact of each treatment component, we examined single and combinatorial effects on short-term survival and long-term reproductive survival in three human MB cell lines. We emphasize that all drug concentrations used here, except of VPA and ABC, are within the range achievable in human or animal plasma [23][24][25][26][27][28].

Irradiation
In accordance to our previous results [18], the MB cell line with the slowest growth rate (D283-Med) was found to be most sensitive to IR, whereas the one with the fastest growth rate (DAOY) was the least sensitive concerning the reduction of short-term vitality, enrichment of apoptotic sub-G1 cells, and also of long-term reproductive survival. Although a higher amount of untreated D283-Med cells is located in the radiosensitive G2/M cell cycle phase compared to DAOY cells which exhibited a higher amount of radioresistant G1/G0 cells, more DSBs were induced in the DAOY then in D283-Med cells. We hypothesize that the p53-wild-type D283-Med cells may prefer non-homologous end-joining (NHEJ) as the main DNA repair pathway which is usually executed within minutes after IR (reviewed in [39]) and, therefore, it is probably not seen 1 h after IR. As NHEJ is more error-prone than the DSB repair by homologue recombination (HR), this might explain the enhanced sub-G1 apoptotic cell fraction observed in D238-Med. Nevertheless, DSB repair was completed 24 h after IR in the surviving cells of both cell lines. Our results show, that the induction of gH2AX foci by IR is accompanied with an enhanced G2/M arrest in p53-mutant DAOY cells. This is in line with findings in other tumor cells [40]. Additionally, we observed that IR induces autophagy in human MB cells. It is already known that IR may induce autophagy as a rescue mechanism resulting in radioresistance but also as cytotoxic process resulting in cell death (reviewed in [41,42]). However, we revealed that autophagy did not account for the long-term radiosensitivity of clonogenic MB cells indicating that other mechanisms such as DSB repair efficiency, suggested by the results above, may play a role here.
The multimodal treatments with 8 Gy IR generally support the findings after IR with 2 Gy (Table 2). Fewer significances found for combinatorial treatments are  (8 Gy) or combinatorial treatment with 5-aza-dC, IR and RES or ABC. Ethanol control was included as RES was dissolved in ethanol. Data are normalized to the untreated control group (=1). Results are presented as mean ± SEM; *, p ≤ 0.05; n = 6 compared to untreated control possibly due to the much lower overall surviving fraction after 8 Gy IR compared with 2 Gy IR.

5-Aza-2′-deoxycytidine (5-aza-dC)
The 5-aza-dC-induced enhancement of mortality, inhibition of proliferation, and reduction of clonogenic survival in all three MB cell lines confirm our previous data [4]. Both, its ability to form DNA enzyme adducts (5-aza-dC/ DNMT) inhibiting cell cycle progression and also leading to DSB as well as its function as de novo methyltransferase inhibitor (DNMTi) [43] are thought to contribute to these anti-cancer effects. The lack of 5-aza-dC-induced apoptosis and autophagy demonstrated here goes along with the only weak induction of gH2AX foci, mainly seen in D283-Med cells. Therefore, we assume that the 5-aza-dCinduced reduction of the short-term and long-term survival in MB cells might be rather caused by its inhibition of proliferation and/or epigenetic properties [4,44,45].
We demonstrated a radiosynergistic reduction of overall clonogenic survival after combining 5-aza-dC with IR. By modulation of the former treatment schedule (5aza-dC only prior to IR) to 5-aza-dC treatment prior and after IR, we could improve the previously found radioadditive effect of 5-aza-dC [4]. This 5-aza-dC-induced enhancement of radiosensitivity has also been described in other tumor entities [46][47][48][49]. It has been suggested by others that the induction of a G2/M arrest by 5-aza-dC treatment prior to IR may account at least partly for its radiosensitizing effect [48]. This could not be confirmed by our findings. Joint analysis of our MB cell lines demonstrated a slightly prolonged DNA synthesis cell cycle phase possibly resulting in the reduced G2/M fraction, after 24 h 5-aza-dC treatment. Additional mechanisms like the p53-dependent induction of senescence demonstrated in hepatocellular carcinoma [50] might play a role here and also could explain the increased sensitivity of p53-wild-type D283-Med and MEB-Med8a cells compared to p53-mutated DAOY cells. Similarly, in experiments with colon cancer cells, Karpf et al. revealed that 5-aza-dC is more effective to mediate growth arrest and cytotoxicity in p53-wildtype versus p53-mutated cells [51]. These observations indicate that the treatment of MB patients with p53 wild-type tumors (the majority of patients) might require lower 5-aza-dC concentrations than the one with p53-mutated tumors.

Valproic acid (VPA)
VPA, a histone deacetylase inhibitor (HDACi) [52] and a replication-independent DNMTi [53] induced a moderate decline in relative vital MB cell number and clonogenic survival, which goes along with our previous findings [18], and supports its anti-tumor effects reported in other tumor models [54,55].
VPA induces apoptosis and growth arrest in MB cells by a p53-independent mechanism through alteration of histone acetylation and thereby gene and protein expression [52]. Indeed, all three tested MB cell lines showed largely similar sensitivities to VPA alone. The combination of VPA and 5-aza-dC showed differential effects on MB cell overall clonogenic survival and the previously measured metabolic activity [18]. After treatment, an enhanced metabolic activity but a reduced clonogenic potential was found in MEB-Med8a cells. In DAOY cells however, despite of reduced metabolic activity, the clonogenicity was slightly enhanced. These results emphasize the importance of long-term reproductive survival assays to evaluate the clonogenic potential of tumor stem cells after different treatment regimes. The radiosynergistic effect of VPA on clonogenic survival, demonstrated in two cell lines here, goes along with findings in glioma cell lines and in a glioma xenograft mouse model, where an inhibition of radiation-induced DSB repair by VPA has been suggested [56]. Our results underline the potential of VPA in combination with radio-and chemotherapy, recently demonstrated in retrospective studies of glioblastoma patients [57,58].

Suberanilohydroxamic acid (SAHA)
The HDACi SAHA alone only slightly reduced the clonogenic survival despite of its strong effect on the metabolic activity published previously [18]. Interestingly, addition of SAHA significantly diminished the 5-aza-dCinduced reduction of overall clonogenic survival in all MB cell lines. Strong additive effects of 5-aza-dC and SAHA are described also by others in leukemic and ovarian cancer cell lines and are thought to be transmitted through the induction of cell cycle arrest (G1/S by SAHA; G2/M by 5-aza-dC) and apoptosis [59,60].
The combination of SAHA with IR showed radiosynergistic effects on overall clonogenic survival in all tested MB cell lines, even slightly stronger than found with VPA. Similarly, other groups observed a radiosensitizing action of SAHA on osteosarcoma and atypical teratoid rhabdoid tumor cells in vitro and in a rhabdoid tumor mouse model in vivo [61,62]. Regarding 5-aza-dC-treated irradiated MB cells, SAHA induced synergistic effects on overall clonogenic survival in all three tested MB cell lines and in this combination seems to be more effective than the HDACi VPA.

Abacavir (ABC)
ABC has been shown to down-regulate the telomerase activity, which is overexpressed in the vast majority of cancers including MB [63]. Sparing normal brain cells due to their low or even not existing telomerase activity, the inhibition of telomerase in brain tumor cells seems to be a promising therapeutic strategy for MB. ABC alone decreased the overall clonogenic survival in all MB cell lines. The ABC-induced reduction of vital cell count goes along with its induction of apoptosis and autophagy. Importantly, we found that ABC acts strongly additive with 5-aza-dC on the reduction of the overall clonogenic survival in all three MB cell lines. It seems worthwhile to investigate this combination in animal models to clarify if it could be an alternative therapy option for infants (<4 years), who cannot obtain radiotherapy due to unwanted adverse effects.
Our experiments revealed also radiosynergistic (DAOY, D283-Med) and radioadditive (MEB-Med8a) effects of ABC on the overall clonogenic survival. In accordance, it has been shown that the inhibition of telomerase activity can sensitize tumor cells to irradiation [64,65], partly mediated by prolongation of DSB repair. In two-out-of-three cell lines we found indeed a higher amount of gH2AX foci after 24 h ABC treatment compared to control. MEB-Med8a behaved differentially possibly due to a high amount of initial (control) DSB. It has been suggested that the upregulation of telomerase activity in tumor cells may represent a protective mechanism against DNA damage [65]. This hypothesis is supported by Akiyama et al., who showed that the overexpression of telomerase protects leukemia cells from DSB-triggered apoptosis [66].
The ABC-induced significant reduction of overall clonogenic survival in all 5-aza-dC/IR-treated MB cell lines indicated a therapeutic potential of the triple combination.

All-trans retinoic acid (ATRA)
ATRA, the prototype of differentiation therapy, induced differential responses in the three MB cell lines. Similar to our previous studies on metabolic activity [18], the strongest effect on overall clonogenic survival was found in the OTX2-expressing D283-Med cell line, supporting recent findings of OTX2 as a medulloblastoma oncogene and possible target for ATRA [67]. MEB-Med8a cells seemed to be ATRA-resistant, whereas in DAOY cells a moderate effect was seen. In accordance, an ATRAmediated induction of caspase 3 followed by inhibition of proliferation and apoptosis was reported in D283-Med cells and to a lesser extend in DAOY cells [68]. Also, enhancement of OTX2 (orthodenticle homeobox 2) protein levels, involved in ATRA-mediated tumor cell death, has been shown in D283-Med but not in DAOY cells [67].
By combination of ATRA and 5-aza-dC as well as ATRA and IR, we showed synergism in D283-Med (1.5 μM), a moderate reduction of clonogenicity in DAOY and also in the ATRA-resistant MEB-Med8a cell line, although at relatively high doses (10 μM) only. We hypothesize that this synergistic effect could be caused by 5-aza-dCinduced demethylation of the CRABP-II gene as suggested by Fu et al. [69]. Therefore, ATRA, if combined with demethylating agents, might be an interesting substance to treat MB tumor stem cells to bypass their radio-and chemoresistance.

Resveratrol
The plant polyphenol RES exhibits tumor-preventive as well as anticancer effects depending on concentration, cell type, and microenvironment [21,[70][71][72]. Studies in MB cells have shown the induction of cell cycle arrest, apoptosis and neuronal differentiation by RES [73][74][75]. In accordance, we could show a RES-mediated prolongation of DNA synthesis cell cycle phase with an increase of gH2AX foci formation. This is followed by the reduction of vital cell count partially due to the induction of apoptosis. Our previous experiments revealed synergistic effects with 5-aza-dC on the metabolic activity of human MB cell lines [18]. Our recent data from clonogenic survival experiments confirm an anti-cancer effect of RES combined with 5-aza-dC and/or IR in all three MB cell lines. Although RES has been shown to prevent normal cells from treatment-induced damage [18,72,76,77], used on its own, it seems to induce adverse effects [78].
It is of special interest that combinatorial treatments with RES reduced the overall clonogenic survival most strongly in the p53-mutated DAOY cell line. Numerous clinical studies confirmed the unfavorable prognosis of patients with p53-mutated MBs (about 10 % of all MB patients); they all died in the first 5 years after diagnosis [79][80][81][82][83]. Animal experiments are warranted to reveal if RES might be an interesting adjuvant therapy option for high-risk patients with p53-mutated tumors.

Neurotoxicity of 5-aza-dC, RES, and ABC
This is the first report showing data concerning the neurotoxicity of a multimodal anti-tumor treatment approach with 5-aza-dC, IR, and RES or ABC using a murine hippocampal tissue slice model. Thereby, no toxic effects on neural progenitor cells were observed. Park et al. revealed adverse effects of RES alone on the murine hippocampal neurogenesis and cognitive function in vivo and in vitro [78]. However, in accordance to our observations, Denissova et al. [84] showed that RES enhances the survival of irradiated mouse embryonic stem cells by accelerating and refining radiation-induced DNA repair.
The nucleoside analogue ABC is known to have a high potential risk of neuronal damage in primary cell cultures of rat forbrain [85]. Nevertheless, in our murine ex vivo hippocamal tissue slice model we could not detect any significant neurotoxicity of ABC in combination with 5-aza-dC and IR regarding nestin-positive neural precursor cells.

Conclusion
The differential results of short-and long-term survival in these and also metabolic tests in previous studies [18] underline the importance of clonogenic assays. Successful antitumor therapy may benefit from personalized medicine reflected here in the case of ATRA, working best in selected OTX2 expressing samples. Nevertheless, the triple combination of 5-aza-dC/IR and ABC or RES turned out to work effectively in all three cell lines warranting further in vivo investigations.
Long-term analysis of neural stem cell numbers in a murine hippocampal slice culture model revealed no neurotoxic potential for these drug combinations. Also VPA and SAHA reduced the reproductive survival of 5-aza-dC/IR treatments further but to a lower extend.
The most powerful combinations of our in vitro experiments warrant in vivo experiments in an orthotopic MB mouse model. If successfully translated in vivo, especially the combination experiments without IR might provide adjuvant alternative therapy strategies in infants (<4 years).
Authors' contributions IP is participated in the design of the study, carried out the experimental assays and draft the manuscript. FM was involved in the performance of tissue slice experiments. RK take part in research instruction and development of the manuscript. AG is participated in conceiving the study, provided technical help and partially draft the manuscript. All authors read and approved the final manuscript.