Opposite Effects of the Triple Target (DNA- PK/PI3K/mTOR) Inhibitor PI-103 on the Radiation Sensitivity of Glioblastoma Cell Lines MO59K and MO59J Differing in DNA-PK and ATM Status

Cholpon S. Djuzenova (  djuzenova_t@ukw.de ) Universitatsklinikum Wurzburg Thomas Fischer Department of Radiation Oncology University Hospital of Würzburg Astrid Katzer Department of Radiation Oncology, University Hospital of Würzburg Dmitri Sisario Department of Biotechnology and Biophysics, University of Würzburg Tessa Korsa Department of Biotechnology and Biophysics, University of Würzburg Gudrun Steussloff University Hospital Wurzburg: Universitatsklinikum Wurzburg Vladimir L. Sukhorukov Department of Biotechnology and Biophysics, University of Würzburg Michael Flentje Department of Radiation Oncology, University Hospital of Würzburg

In order to enhance the effects of RT, two major approaches have been proposed. The rst strategy is based on the reduction of the treatment volume, i.e. by irradiating a smaller volume of healthy tissue while irradiating the de ned tumor volume. The second approach utilizes the increase of the differential response of the normal tissue and the tumor using chemotherapeutics, biological agents, proteomic or genetic methods etc. [4][5][6][7]. This strategy is based on the rationale that additional arti cially induced DNA damage may lower the threshold of the IR mediated cancer cell death. DNA double-strand breaks (DSBs) are the most signi cant lesions produced by IR and other exogenous cytotoxic agents. DSBs affect the genomic integrity of cells and, if insu ciently repaired or misrepaired, they may lead to chromosome breaks, gene deletions and translocations [8].
IR targets not only DNA itself but rather DNA in the context of chromatin, i.e. within a highly regulated and complex DNA-protein structure [9,10]. One of the earliest events in the response to IR-induced DNA damage is phosphorylation of histone H2AX at Serin139 to γH2AX, which in turn is believed to recruit DNA repair factors to sites of DSBs [11]. Since γH2AX is associated with DSB repair, the kinetics of induction and decay of γH2AX expression might be related to the repair e ciency of higher-order chromatin structure [12].
Despite technical advances in radiation delivery protocols, poor radiation response or even radioresistance of some tumor entities justi es the urgent need to investigate the underlying network of signaling pathways that de ne tumor response to IR. Among others signaling pathways, the frequent activation of the PI3K pathway in GBM cells and its crucial role in cell growth, survival and migration make it a promising target for pharmacological intervention (for review, see [13]). Besides this, genetic alterations leading to activation of the PI3K/Akt/mTOR pathway are associated with treatment resistance in a variety of solid tumors [14].
We here examine two GBM cell lines MO59K and MO59J derived from the same glioblastoma biopsy specimen regarding their response to irradiation and simultaneous DNA-PK/PI3K/mTOR inhibition by PI-103. The two isogenic cell lines were chosen because they differ considerably in intrinsic radiosensitivity and mutational status of DNA-PK and ATM. The inhibitor PI-103 was added 3 hours before IR and kept 24 h thereafter. Irradiated samples of both cell lines treated with inhibitor were analyzed for colonyforming ability, DNA damage induction and repair, apoptosis, autophagy, marker protein expression, and cell cycle alterations.

Cell lines and preliminary characterization
Glioblastoma cell lines MO59K and MO59J were obtained from the American Type Culture Collection (Manassas, VA) and authenticated by the supplier. Both cell lines were cultivated in CGM containing Dulbecco's modi ed Eagle's medium nutrient mixture F-12 HAM (Sigma D-6421), non-essential amino acids, and sodium pyruvate according to the provider's prescription. The modal chromosome numbers were reported to be 64 and 74 for MO59K and MO59J lines, respectively [15]. Cells were used at low (< 15) passages after thawing and were regularly examined during the study for Mycoplasma (MycoAlert; Lonza, Rockland, ME) contamination.
In our preliminary tests (see Supplementary Figs. S1-S4) we found intrinsic differences in chromatin compaction between the two tested cell lines. These tests were performed because chromatin compaction is well known to play an essential role in DNA repair, and thus can in uence the sensitivity of cells to DNA damaging agents [16]. To this end, we performed a Comet assay and also analyzed microscopically the expression of histone deacetylase 5 (HDAC5). Additionally, we tested the stainability of the nuclear chromatin with 4',6'-diamidino-2-phenylindole (DAPI).
The alkaline Comet assay reveals not only single-strand DNA breaks and alkali-labile sites, but it also provides a measure for chromatin integrity after exposure to a clastogenic agent such as IR [17]. By assessing the DNA damage/fragmentation from the Tail Moment (TM) values of irradiated cells ( Supplementary Fig. S1), we found that the initial radiation-induced DNA fragmentation in MO59J cells was ~ 1.3 times higher than in MO59K cells. Yet, the residual DNA damage as well as the DNA repair kinetics were very similar in both cell lines. The difference between MO59K and MO59J cell lines was seen more evidently if the TM distribution of individual cells was taken into account, i.e. the representative TM values histograms of the radiosensitive MO59J line were strongly shifted towards higher values compared with the TM values histograms of MO59K cells ( Supplementary Fig. S2).
A further preliminary test was immuno uorescence staining of HDAC5 foci. It is known that histone deacetylation promotes a "closed" chromatin conformation and it usually leads to repression of gene activity [18]. As seen in Supplementary Fig. S3, the number of HDAC5 foci in MO59K cells was ~ 10-time higher than in MO59J cells. An additional interesting nding was that MO59K cells show a much higher stainability with DAPI than MO59J cells ( Supplementary Fig. S4). Since a linear relationship exists between chromatin condensation and its stainability with DAPI [19], this result points to a strong intrinsic difference in chromatin compactness between the two cell lines studied here. The nding on DAPI staining ( Supplementary Fig. S4) also corroborates with the observed difference in HDAC5 expression ( Supplementary Fig. S3). To sum up, the preliminary tests, including the Comet assay, HDAC5 and DAPI staining, suggest a strong difference in the intrinsic degree of chromatin condensation between the isogenic MO59K and MO59J cell lines.
The substance was freshly diluted from frozen (-80 °C) aliquots dissolved in DMSO. PI-103 (2 µM, [20]) was added 3 h (short-term treatment) prior to exposure to IR and remained in CGM up to 24 h (long-term treatment) after IR. Control cells were treated in parallel with respective amounts of DMSO.

X-ray irradiation
Irradiation was performed at room temperature using a 6 MV Siemens linear accelerator (Siemens, Concord, CA) at a dose rate of 2 Gy/min. After irradiation, cells were kept in CGM for the indicated time until harvest.

Colony survival assay
Cell survival was assessed by colony formation as previously described [20]. Subcon uent monolayers of control and inhibitor-treated cells were irradiated in culture asks lled with CGM at room temperature by graded single doses (0 8 Gy), seeded 24 h post-IR in Petri dishes, i.e. delayed plating, and then cultured for 10-12 days in CGM. Four replicates were performed for each radiation dose, and the experiments were repeated at least four times. After 10-12 days the cells were xed and stained with crystal violet (0.6%).
Macroscopic colonies containing at least 50 cells were scored as survivors. The mean clonogenic survival data for each cell line were tted to the linear-quadratic (LQ) model: where, SF is the survival fraction, X is the irradiation dose, α and β are the tted parameters.
In some experiments, radiation survival was assessed without replating, i.e. tumor cells were cultured at xed densities, and treated with the inhibitor 3 h before IR with 3 and 5 Gy. Twenty four hours after IR culture medium was replaced with fresh medium, and the dishes were incubated for the next 10-12 days, xed and stained as it was done in case of delayed plating (see above).

Western blotting
For immunoblot assays, whole-cell lysates were prepared 30 min and 24 h post-IR, according to standard procedures. Samples equivalent to 40 µg of protein were separated using 412% SDS-polyacrylamide precast gels (Invitrogen, Karlsruhe, Germany) and transferred to nitrocellulose membranes. For protein detection, membranes were incubated with respective primary and species-speci c peroxidase-labeled secondary antibodies according to standard protocols. The levels of protein expression were quanti ed using the software ImageJ (NIH, Bethesda, MD) and normalized to β-actin levels.

Antibodies
The primary and secondary antibodies are speci ed in the Supplementary Information. 4% formaldehyde (Thermo Scienti c, Rockford, IL). Thereafter, the cells were stained for γH2AX essentially as previously decribed [21]. Confocal uorescence images were acquired with a Zeiss LSM 700 microscope. In each sample, computer-assisted γH2AX foci counting from 3D-CLSM image stacks [22] was performed in about 100 cells.

Cell cycle measurements by ow cytometry
Non-treated and drug-treated cell cultures were irradiated as sub-con uent monolayers in CGM at room temperature. The cells were then kept in CGM under standard conditions, and analyzed by ow cytometry 30 min and 24 h after IR exposure. For analysis, the cells were trypsinized, washed twice in PBS, xed and stained with propidium iodide (PI, Sigma P-4170, 10 µg ml -1 ) in the presence of ribonuclease A (Sigma R-5250, 25 µg ml -1 ) as described elsewhere [20].
At least 20,000 cells were assayed for DNA distribution using a ow cytometer FACSCantoII (Becton Dickinson, San Jose, CA). The output data presented as one-dimensional histograms acquired in linear mode, i.e., the distributions of PI-DNA signals within the cell samples, were analyzed using the Flowing Software program obtained from P. Terho (Turku Centre for Biotechnology, Turku, Finland) and the ModFit LT program (Verity Software House, Topsham, ME).

Statistics
Data are presented as means (± SD or ± SE). Mean values were compared via Student's t-test. The threshold of statistical signi cance was set at P < 0.05. Statistics and tting of experimental data were performed with Origin 8.5 (Microcal, Northampton, MA).

Results
The following experiments were performed to evaluate the effects of simultaneous targeting of DNA-PK, PI3K and mTOR with PI-103 on the radiation sensitivity, marker protein expression, DNA damage/repair, the degree of apoptosis, autophagy, and cell cycle alterations in two GBM tumor cell lines. The two commercially available isogenic (MO59K and MO59J) cell lines originate from different portions of the same human glioblastoma specimen and show an ~ 10-fold difference in their radiation sensitivities [23]. In addition, the radiosensitive MO59J cells carry mutated DNA-PK and ATM genes [24]. Given that MO59J cells are fully devoid of DNA-PK, the pyridinylfuranopyrimidine PI-103, a triple-target (DNA-PK, PI3K and mTOR) inhibitor, has one less target in MO59J cells, as compared to MO59K cells, which express all three target proteins.
Effects of PI-103 and NVP-AUY922 on colony survival after IR Figure 1 shows the cell survival curves of control (DMSO) and drug-treated cells plotted versus the radiation dose, along with the best ts of the LQ model (Eq. 1) to the data. The plating e ciencies (PE) of non-irradiated cell samples, as well as the tted parameters derived with the LQ model, including the surviving fraction at 2 Gy (SF2), the radiation dose required to reduce colony forming ability by 90% (D 10 ) and the growth inhibition factor (IF 10 ) from at least 4 independent experiments are summarized in the Supplementary Table S1. As seen in Fig. 1A, PI-103 had a strong radiosensitizing effect in MO59K cells ( Fig. 1, curve 2 vs. curve 1), as evidenced by the decrease of the SF2 value from 0.63 in non-treated irradiated cells to 0.41 in PI-103-treated irradiated MO59K cells (Supplementary Table S1). In contrast, PI-103 induced radioresistance in MO59J cells ( Fig. 1, curve 4 vs. curve 3), indicated by the increase of the SF2 value from 0.16 in control irradiated MO59J cells to 0.23 (Supplementary Table S1). Likewise, the D 10 value was increased from 2.3 Gy in control to 3.2 Gy in drug-treated MO59J cells.
It is conceivable that radiation resistance induced by PI-103 in MO59J cells may be the result of selective cell loss prior to cell re-plating due to excessive apoptosis and increased detachment of highly damaged cells treated with PI-103 and IR. These cells would escape analysis due to the washing and trypsinization steps prior to subsequent re-plating for colony-survival test. Indeed, the colony-survival test in our study was carried out with delayed plating, i.e. the cell samples were incubated 24 h post-IR with the inhibitors before being re-plated in Petri dishes for further cultivation. In order to analyze the possibility of exclusion of strongly damaged cells during delayed plating, we tested additionally the survival of pre-plated cells.
To this end, the tested cells were grown at xed densities in Petri dishes, treated with the inhibitor for 3 h and irradiated with a single dose of 3 and 5 Gy. Twenty four hours later the inhibitor was washed out and the samples were cultivated just like in the delayed plating protocol. The results of experiments without re-plating are shown in Fig. 1D, 1E. As seen in Fig. 1D, 1E, treatment with PI-103 without re-plating caused the same opposing effects on the radiation sensitivity of the two cell lines as in the experiments with cells recovered to about 50-80% of the background level ( Fig. 2, LHS). However, in PI-103-treated MO59J cells much less p-Akt recovered independent of IR.
In addition to PI3K, a target of PI-103, we analyzed the expression of a further target of PI-103, p-mTOR and its downstream effectors, ribosomal S6 and translational repressor 4EBP1 proteins, which are known to in uence cell-cycle progression and cell growth [25,26]. The expression of p-mTOR was decreased after a 3-h incubation with PI-103 in both cell lines (Fig. 2), independent of IR. However, after a 24-h incubation it was moderately increased in MO59K cells whereas it remained slightly reduced in MO59J cells. As a result of depletion of p-Akt and reduction of p-mTOR, the pS6 protein was also depleted 30 minutes post-IR in both cell lines (Fig. 2) treated with PI-103. In contrast, expression of p-4E-BP1 was reduced to a much lesser extent ( Fig. 2) but in the same pattern as p-S6.
The lack of PTEN in PTEN-mutated cells, such as both tested cell lines, usually leads to a compensatory activation of the PI3K pathway. Thus, activation of Akt, a major hub protein of the pathway, typically results in an inhibition of Raf-1 and its downstream effectors MEK and ERK through a cross-talk between the PI3K/Akt/mTOR and Ras/Raf/MEK/Erk (MAPK signaling) pathways [27]. Normally the MAPK pathway transmits signals from cell surface receptors to promote proliferation and survival programs, and it is frequently mutated in cancer cells [28,29]. In addition, we detected the expression of non-phosphorylated forms of the above mentioned proteins (data not shown). Contrary to the phosphorylated forms, the expression of non-phosphorylated forms of Akt, mTOR, 4E-BP1, S6, MEK1/2, and Erk1/2 remained virtually unchanged after addition of PI-103.
Impact of PI-103 on the IR-induced DNA damage assessed by γH2AX foci counting and Western blot To further elucidate the opposing effects of PI-103 on the colony-forming ability of MO59K and MO59J cells (Fig. 1), we compared the IR-induced DNA damage in PI-103-treated and control drug-free cells by counting immunostained γH2AX foci [11,21] as markers of the DNA DSBs in irradiated cell samples (Figs. 3A, 3B), either untreated or pretreated with PI-103 alone. γH2AX foci counting in cell nuclei (Figs. 3C, 3D) was performed on 3D-CLSM image stacks using an automated foci analysis software reported recently [22]. The γH2AX foci were counted in samples prepared at different times (0-6 h) after IR.
Although the drug-free samples of MO59K and MO59J cell lines were similar in their initial and residual γH2AX foci counts (blue symbols in Figs. 3C, 3D), PI-103 exerted different effects on the time-course of foci induction and decay in two cell lines (red vs blue symbols in Figs. 3C and 3D). However, for a deeper quantitative analysis of γH2AX foci in these isogenic cell lines, the ~ 10% difference in their modal chromosome numbers [15] should also be considered.
Due to its limited spatial resolution, conventional uorescence microscopy only enables recognition and counting of discrete γH2AX foci, but it does not provide any information on the foci size and their intensity. Therefore, the data in Figs. 3C and 3D show only the amounts of γH2AX foci in the different cell samples, without any information on the total γH2AX protein expression over the whole nuclear volume. Therefore, in addition to foci counting, we also analyzed γH2AX expression by Western blot (Fig. 3E). We found that PI-103 strongly increased the expression of γH2AX in MO59K cells, especially 30 min after IR (8 Gy), but not in MO59J cells. However, given that cell samples have to be lysed for Western blotting, apoptotic, highly damaged and/or dead cells with high γH2AX content cannot be excluded from the analysis. Interestingly, background γH2AX expression in MO59J cells was apparently too low, even in PI-103 treated cells, to be detected. This result con rms the ~ 10-fold lower sensitivity of Western blots as compared to γH2AX foci counting by uorescence microscopy, reported elsewhere [31].
Effects of PI-103 and irradiation on the expression of DNA repair proteins Driven by the nding that PI-103 inversely affects the radiation survival of tested cell lines (Fig. 1) we analyzed the expression of several DNA repair proteins. Figures 4 and 5 show representative Western blot detections of several proteins belonging either to non-homologous end-joining (NHEJ, Fig. 4) or homologous recombination (HR, Fig. 5) DNA repair pathways in both cell lines treated with drugs and IR.
As seen in Fig. 4, the components of the heterodimer Ku70/Ku80, which bind to and protect the broken DNA ends, are equally expressed in both cell lines and the expression levels were unaffected by drug and/or IR treatment. The de ciency of MO59J cells in DNA-PK gene is clearly con rmed by our Western blot data (Fig. 4, RHS), i.e. DNA-PK was completely absent in MO59J cells but clearly present in MO59K cells (Fig. 4, LHS). Thirty minutes after IR, the expression levels of DNA-PK in MO59K cells were almost unchanged, independent of drug and/or IR treatment (Fig. 4). We also found no differences in the background expression of the DNA repair protein Rad50 between cell lines (Fig. 4). Moreover, neither chemical inhibition nor IR exposure affected Rad50 expression in both studied cell lines.
In addition to the NHEJ DNA repair pathway, we tested several proteins of the HR pathway (Fig. 5). Our Western blot data (Fig. 5, RHS) clearly con rmed the mutation in ATM gene in MO59J cells, i.e. the expression of ATM protein was strongly reduced in this line compared with MO59K line. Interestingly, despite the ATM de ciency in MO59J cells, addition of PI-103 clearly increased ATM expression thereby enhancing DNA repair. We further tested the CtBP (C-terminal binding protein) interacting protein (CtIP), an interacting partner of the Mre11/Rad50/Nbs1 (MRN) DNA damage sensor protein complex, which recognizes DNA DSBs. As seen in Fig. 5, CtIP as well as p95/NBS1 are strongly expressed in both cell lines, especially 24 h post-IR in MO59J cells. In both cell lines, the activated form of p95/NBS1 protein was strongly induced by IR independent of PI-103 treatment. Additionally we detected p-BRCA1 protein, which, together with BRCA2, is required for localization of Rad51 to DNA DSBs sites. As seen in Fig. 5, MO59J cells exhibited lower expression of p-BRCA1 as compared to MO59K cells. PI-103 treatment slightly diminished p-BRCA1 expression in non-irradiated MO59J cells, but not in irradiated samples. The expression pattern of Rad51 was similar to that of p-BRCA1. The Rad54 helicase, which interacts with Rad51 to regulate its DNA binding and strand exchange activities during HR, was strongly expressed in both cell lines, especially 24 h post-IR in MO59J cells independent of drug treatment.
To sum up, we found that several proteins of the NHEJ pathway are highly expressed not only in the repair-pro cient MO59K cells but also in the repair-de cient MO59J cells. However, due to the absence of DNA-PK, NHEJ likely does not contribute to DSB repair upon drug and/or IR treatment in MO59J cells. As for HR, despite reduced level of ATM in MO59J cells, other marker proteins of the HR pathway were markedly expressed in this cell line, except for p-BRCA1. Interestingly, addition of PI-103 did not reduce the expression of HR-related proteins, but instead it induced the expression of ATM especially in MO59J cells, indicating a certain degree of HR pathway functionality in the radiation-sensitive MO59J cells.
Effects of PI-103 and irradiation on the expression of p53, p-p53 and p53-related Bax protein Both DNA-PK and ATM redundantly phosphorylate similar substrates, e.g. both are required for normal levels of p53 phosphorylation and p53-dependent apoptosis [33]. Therefore, we analyzed the expression of p53 in both cell lines (Supplementary Fig. S6). We found that the diminished expression of p53 in MO59J cells was further reduced by addition of PI-103 (24 h post-IR) but to a much lesser extent in MO59K cells (Supplementary Fig. S6). Another difference between these cell lines was that irradiation activated p53 in MO59K cells by phosphorylation on Serin15 ( Supplementary Fig. S6), but p-p53 was undetectable in irradiated MO59J cells. Next, we found that at the time of IR the background expression of the pro-apoptotic protein Bax was much higher in MO59K cells than in MO59J cells, especially after addition of PI-103 ( Supplementary Fig. S6). Given that Bax is related to p53, this nding is in line with the pro-apoptotic function of p53 and its different expression levels in both cell lines.

Effects of PI-103 and radiation on late-stage apoptosis and autophagy
To further explore the mechanisms underlying the opposing effects of PI-103 on the radiation sensitivity of the GBM cell lines illustrated in Fig. 1, we also analyzed cleaved PARP, a well-known marker of apoptosis. As seen in Supplementary Fig. S7, the background expression of PARP, which plays an important role not only in base excision repair, but also in homologous and non-homologous DNA repair [34], was much higher in MO59K than in MO59J cells. Interestingly, short incubation with PI-103 strongly increased the expression of cleaved PARP, especially in MO59K cells (Supplementary Fig. S7) and this effect was independent of IR. This nding is consistent with the assumption that at the moment of IR, PI-103-treated MO59K cells underwent extensive apoptosis. The net expression of cleaved PARP in drugtreated MO59J cells was much lower than in MO59K cells, which may be due to the de ciency of MO59 cells in DNA-PK, which is normally involved in signaling DNA damage to the apoptosis machinery [35]. In contrast to short-term incubation with the inhibitor, after long-time incubation with the drug, almost no expression of cleaved PARP was seen, independent of IR. Another tested marker of apoptosis, cleaved caspase 3, did not show any changes in response to the inhibitor and/or IR (Supplementary Fig. S7).
Because the PI3K pathway is a major pathway regulating autophagy [36], we also studied the possible role of cytoprotective autophagy in the development of radiation resistance in PI-103-treated MO59J cells (Fig. 1). To this end, we detected the autophagosomal membrane-bound LC3B protein along with the expression of the p62/sequestome protein, a pleiotropic protein that is consumed during autophagy [37]. Interestingly, we found that PI-103 added for 3 h strongly induced autophagy, as evident from the increased levels of LC3B-II protein in both cell lines (Fig. 6). However, prolonged incubation with PI-103 increased autophagy only in MO59J cells. These ndings agree well with the results of Fan et al. (2010) who found that dual inhibition of PI3K and mTOR promotes survival of glioma cells by inducing cytoprotective autophagy [38]. Furthermore, the enhanced autophagy in PI-103-treated (24 h) MO59J samples, indicated by LC3B-II expression, was also corroborated by the strong reduction of p62, another marker of autophagy (Fig. 6) in MO59J cells treated with PI-103. In contrast, the reduction of p62 in MO59K cells treated with PI-103 (24 h post-IR) was not correlated with the LC3B-II marker. The highest extent of autophagy assessed by p62 expression was observed in samples treated with PI-103 alone. To sum up, a 24-h treatment with PI-103 induced autophagy in MO59J cells as evidenced by both markers. In contrast, the respective samples of MO59K cells showed increased autophagy only in case of p62 marker. Both inhibitors in combination induced autophagy especially in MO59J cells if measured by p62 protein, but not by LC3B-II. It is worth to be mentioned, however, that the measurement of p62 expression strictly as a marker of autophagic ux is still controversial and can be misinterpreted mainly because this protein is subject to complex regulation at both the transcriptional and post-translational levels [39].

Effects of PI-103 and irradiation on the cell-cycle progression
Besides apoptosis, the most prominent consequence of p53 activation is cell-cycle arrest [40]. We found that the long-term treatment with PI-103 caused a reduction of S phase and an accumulation of cells in G1 phase (70%) in non-irradiated MO59K cells, whereas irradiated and PI-103-treated MO59K cells showed a strong G2-arrest (45-55%) and a reduction of G1-fraction (Fig. 7A). Unlike MO59K cells, drugfree MO59J cells showed strong G2-arrest (~ 80-90%) 24 h after irradiation (Fig. 7B). Interestingly, addition of PI-103 signi cantly reduced G2-arrest in irradiated MO59J cells and increased the fractions of G1-and S-phase cells, which is indicative of the partial abolishment of cell-cycle arrest.

Discussion
In this study, we tested the effects of targeting the PI3K pathway using the triple DNA-PK/PI3K/mTOR inhibitor PI-103 in combination with IR to induce radiosensitization in two isogenic GBM cell lines differing in DNA-PK and ATM status [23]. The SF2 values obtained here ( Fig. 1 and Supplementary Table S1) revealed that MO59J cells were ~ 4 times more sensitive to IR as compared to MO59K cells, which is in qualitative agreement with the radiosensitivity data for these cell lines reported elsewhere [23].
Unexpectedly we found that the DNA-PK/PI3K/mTOR-inhibitor PI-103 affected radiosensitivity in a cell line-dependent manner, i.e. it acted as a radiosensitizer in MO59K cells, but it induced radiation resistance in MO59J cells (Fig. 1). Figure 8 outlines our key ndings. At the time of IR, both cell lines treated with PI-103 were strongly depleted of Akt. Surprisingly, long-term (24 h) incubation with PI-103 led to the re-activation of p-Akt in MO59K cells (Fig. 2, LHS). Activated Akt is widely recognized as the major mediator of cell survival, which inhibits apoptosis through several mechanisms [41], e.g. preserving mitochondrial integrity, phosphorylation and inactivation of pro-apoptotic BAD (Bcl-2-antagonist of cell death) and caspase 9 etc. [42]. BAD maintains Bcl-2 (B-cell lymphoma 2) and Bcl-xL function thereby inhibiting apoptosis mainly at the mitochondrial level by suppressing cytochrome c release [43]. However, despite re-activation of Akt and p-mTOR, the clonogenic survival data (Fig. 1) showed that PI-103 decreased radiation survival of MO59K cells. In contrast, radiation survival of PI-103-treated MO59J cells was unexpectedly increased, although in these samples the p-Akt protein was reactivated to a much lesser extent (Fig. 2, RHS) than that of MO59K cells. The same samples of MO59J cells showed strong up-regulation of p-Erk1/2 ( Supplementary Fig. S5, LHS) which could explain, at least partly, the radioresistance of PI-103-treated MO59J cells.
(* best place for Fig. 8, Schema *) A further factor of chemically or radiation-induced cell resistance or death is the induction and repair of DNA DSBs, probed in this study by counting of γH2AX foci (Figs. 3A-3D) and by Western blot detection (Fig. 3E). γH2AX foci counting revealed some differences in the foci induction rate between the two cell lines. Yet, the initial radiation-induced foci amounts were very similar, which is in line with the nding of Stiff et al. (2004) [24]. However, if the γH2AX foci counts (Fig. 3) were corrected for the cell line-speci c chromosome content [15], MO59J cells would display a lower γH2AX foci-to-chromatin ratio than MO59K cells.
We found a strong difference in the kinetics of γH2AX induction between irradiated controls of both cell lines (Figs. 3C and 3D). Particularly, MO59J cells displayed a slower induction of γH2AX foci (Fig. 3D) as compared to MO59K cells (Fig. 3C), which is in accordance with the lack of NHEJ in MO59J cells and their high radiation sensitivity [15]. However, after treating MO59K cells with PI-103 the γH2AX induction was delayed (Fig. 3C, red curve vs. blue curve), which could explain the increased radiation toxicity as compared to control MO59K samples (Fig. 1C). In contrast, the induction of γH2AX in irradiated MO59J cells was almost unaffected by PI-103 treatment. Despite these differences in the initial induction phase, the kinetics of foci decay were very similar in both GBM lines (Fig. 3). Accordingly, Wang et al. (1997) found no marked difference in the unwinding/rewinding of DNA supercoils, radiation-induced changes in nucleoid halo size or the kinetics of nucleoid halo lysis between these two cell lines [44]. Likewise, an e cient rejoining of DNA DSBs has been observed in vitro in extracts of MO59J cells, despite the lack of DNA-PK activity in this strain [45]. However, assessment of DNA damage by γH2AX foci counting did not take into account their size and intensity. As a result, foci numbers per nucleus do not necessarily re ect the total amount of phosphorylated H2AX.
We therefore performed additional detection of histone γH2AX in cellular extracts by Western blotting (Fig. 3E). Our Western blot data on γH2AX agree well with the results of Paull et al. (2000), who showed that MO59J cells exhibit a substantially reduced ability to increase the phosphorylation of histone H2AX in response to ionizing radiation [46].
The discrepancy between the γH2AX foci counts and the virtual absence of γH2AX bands in our Western blot experiments of MO59J samples probably resides in the different sensitivity of the two methods, as reported elsewhere [31]. Yet it should be mentioned that in Western blot experiments all cells are lysed, and therefore, apoptotic or highly damaged cells could not be excluded from the analysis.
Nevertheless, we found that a short (3 h) incubation period with PI-103 strongly increased the expression of γH2AX in MO59K cells, especially after IR (Fig. 3E), whereas 24 h after IR it was measurable only in combination with IR. Given that H2AX is phosphorylated to γH2AX, among other factors, by DNA-PK and ATM [24], and exactly these two proteins are either completely absent (DNA-PK) or reduced (ATM) in MO59J cells, it is conceivable that DNA damage assessment based on γH2AX detection underestimates the actual degree of DNA damage, at least in MO59J cells. This line of reasoning is supported by our preliminary tests, which revealed strong difference in chromatin compactness between the two tested cell lines ( Supplementary Figs. S1-S4). Consequently, the radioresistance of PI-103-treated MO59J cells can be related to the inability of PI-103 to induce DNA damage detectable by γH2AX in irradiated MO59J cells, as contrasted to MO59K cells (Fig. 3E). Another reason may be a partial enhancement of the HR pathway in PI-103 treated MO59J cells as compared to control (Fig. 5). In addition, as suggested by Virsik-Köpp and co-workers (2003), the elimination of DNA-PK-dependent NHEJ can recruit a slow, error-prone repair process, which is DNA-PK independent [15].
Another possible reason for the opposite effects of PI-103 on the radiation survival of the two tested cell lines might be a difference in their proneness to PI-103-induced apoptosis. The DNA repair protein DNA-PKcs and the signal transducer ATM are both known to be activated by DNA DSBs [33]. Both DNA-PK and ATM redundantly phosphorylate similar substrates, e.g. both are required for normal levels of p53 phosphorylation and p53-dependent apoptosis [33]. Indeed, our experiments revealed a diminished expression level of p53 and lack of p-p53 in MO59J cells (Supplementary Fig. S6). Accordingly, we found higher expression of the apoptosis marker cleaved PARP in MO59K cells treated with PI-103 (30 min post-IR) compared to respective samples of MO59J cells (Supplementary Fig. S7).
Although DNA-PK is an important DNA repair protein, it can also in uence gene expression by phosphorylation of various transcription factors including Fos, Jun, myc, and p53, which are known to regulate either cell death or cell growth [47]. There are con icting data on the role of DNA-PK in regulation of cell death. Thus, DNA-PK has been reported to protect cells from death via a caspase-independent [48] or p53-independent [49] pathway. On the other hand, DNA-PK can promote cell death by interacting with telomeres, forming a complex with p53, and phosphorylating MDM2 to render it unable to inhibit p53 transactivation [50].
Our ndings clearly indicate a pro-cell death role of DNA-PK. Thus, MO59K cells containing intact DNA-PK were more sensitive to combined PI-103 and IR treatment (Fig. 1) than to IR alone, whereas MO59J cells lacking DNA-PK were more resistant to the combined PI-103-IR treatment than to IR alone. The reasons for these divergent ndings might involve a variety of factors.
As known, the most prominent outcomes of p53 activation are apoptosis and cell-cycle arrest [40]. We found that the expression of pro-apoptotic Bax was much higher in MO59K cells than in MO59J cells, especially after addition of PI-103 ( Supplementary Fig. S6). Given that Bax is related to p53, this nding is in line with the pro-apoptotic function of p53 and its different expression levels in both cell lines ( Supplementary Fig. S6). Moreover, our results are consistent with the p53-regulating function of DNA-PK [47]. It is well known that p53 is able to up-regulate Bax in variety of cell types including glioma [51]. Therefore, it is reasonable to assume that the decreased expression of the pro-apoptotic protein Bax in MO59J cells may be due to reduced level of p53, which is known to result from de ciency of DNA-PK [51].
Another consequence of the diminished expression of p53 and the absence of the activated form of p53 (p-p53) in MO59J cells is a strong alteration of the cell cycle, i.e. the majority (~ 80%) of irradiated drugfree MO59J cells were arrested in G2-phase. However, IR treatment in the presence of PI-103 resulted in a partial cell-cycle progression of MO59J cells, as evidenced by reduction of the G2 fraction and doubling of the S-phase fraction. This might be another reason for the increased radiation survival of MO59J cells treated with PI-103 as compared to the drug-free irradiated controls.
In accordance with our ndings, the pro-cell death role of DNA-PK in human glioblastoma was demonstrated by Chakravarty et al. (1999) who showed that activation of DNA-PK by staurosporine, ceramide or UV radiation increases apoptosis in human neuroblastoma cells [53]. Likewise, Chen et al. (2005) found that MO59K cells were much more sensitive to staurosporine treatment than MO59J cells [47], which corroborates our results on the sensitivity of MO59J cells to combined PI-103-IR treatment.
An additional reason for the opposing effect of PI-103 on the radiation sensitivity of the tested GBM cell lines could be protective autophagy. As known, PI3K signaling is a major pathway regulating autophagy [36]. Indeed, we found an increased autophagy indicated by LC3B-II and p62 markers in irradiated PI-103treated MO59J cells (Fig. 6).

Conclusions
To sum up, our data demonstrate an enhanced radiosensitivity in MO59K tumor cells pretreated with the triple DNA-PK/PI3K/mTOR inhibitor PI-103 added shortly before IR. In contrast, the same treatment caused radioresistance in MO59J cells, which lack one of the target proteins for PI-103, i.e. DNA-PK. However, the lack of the NHEJ pathway in DNA-PK de cient MO59J cells might have been partially compensated by a slow, DNA-PK independent and yet unexplored DNA repair pathway. Another explanation for the radioresistance of PI-103-treated MO59J cells might be their DNA-PK de ciency, leading to diminished expression and activation of the DNA-PK substrate p53. PI-103 treatment further reduced p53 expression in MO59J cells, resulting in a resistance to apoptosis. In contrast, intact DNA-PK in MO59K cells treated with PI-103 activated p53 and related pro-apoptotic proteins. Moreover, the massive G2 arrest in irradiated MO59J cells was partially abolished by PI-103, thus allowing cell-cycle progression of a signi cant portion of cells. Unlike in MO59J cells, the G2 fraction in irradiated and PI-103-treated MO59K cells was increased. Furthermore, DNA damage assessed by γH2AX expression in Western blot showed a strong difference between the two cell lines. Thus, irradiation in the presence of PI-103 caused a much stronger induction of γH2AX in MO59K cells than in MO59J cells. In conclusion, to further elucidate the mechanisms of radiosensitization in DNA-PK-and ATM-de cient tumor cells, studies on an extended panel of cell lines regarding the cell type-speci c activation of p53, cell-cycle arrest, apoptosis and DNA repair will be the subject of future research in our laboratory.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availability of data and materials
All data generated and/or analyzed during this study are included in this published article and its Additional les.

Competing interests
The authors declare that they have no competing interests.

Funding
This work was supported by the grant (No. 70112891 to CSD and VLS) of the Deutsche Krebshilfe e.V.
The funding organization has no role in the design of the study, analysis, and interpretation the data and in writing the manuscript.
Authors' contributions CSD, MF and VLS designed the study, interpreted the data and prepared the manuscript. Most of the experiments were performed by AK with contributions of TK, DS, and GS. TF assisted with the irradiation. All authors read and approved the nal manuscript. the survival data at 3 and 5 Gy extracted from part a. Parts d and e showed the survival data at 3 and 5 Gy obtained without re-plating. "*" means P < 0.05.  100 nuclei were evaluated using the ImageJ plugin FocAn [22]. Part e shows γH2AX expression in cellular lysates detected by Western blot 30 min and 24 h post-IR in control and drug-treated cells irradiated with 8 Gy. The uncropped blots are shown in Supplementary Fig. S8.

Figure 4
Representative Western blot analysis demonstrating the expression of several DNA repair proteins of the NHEJ pathway in MO59K and MO59J cells treated with DMSO (control) or PI-103 for 3 h prior to IR with 8 Gy and detected 30 min and 24 h thereafter. β-actin was used as loading control. The uncropped blots are shown in Supplementary Fig. S8. The experiment was repeated at least three times.

Figure 5
Representative Western blot analysis of the expression levels of several DNA repair proteins of the HR pathway in MO59K and MO59J cells treated with DMSO (control) or PI-103 for 3 h prior to IR with 8 Gy and detected 30 min and 24 h thereafter. β-actin was used as a loading control. The uncropped blots are shown in Supplementary Fig. S8. The experiment was repeated at least three times.