To test whether our cancer cell lines exhibited the characteristics of the Warburg effect, we analyzed both, normal human fibroblasts AG1522D and NFF primary (NFFp) and human cancer cell lines HT1080, HeLa and HCT116 under normoxic conditions, using a Seahorse XFp metabolic analyzer and a Real-Time ATP-Rate assay to measure the total ATP production rate through oxidative phosphorylation (OXPHOS) and aerobic glycolysis ( Figure 1A ). Our results show that the tested normal fibroblast cell lines exhibited a slight preference for oxidative phosphorylation, with app. 60-65% of ATP being generated through this pathway ( Figure 1B ). In contrast, the tested cancer cell lines did not show a strict preference for aerobic glycolysis to supply ATP under standard cell culture conditions. The cervix carcinoma HeLa cells and the colon carcinoma HCT1116 cell lines gained their ATP by employing app. 20-35% of aerobic glycolysis. Only the sarcoma HT1080 cells showed a preference for this metabolic pathway, generating 55% of their ATP via aerobic glycolysis ( Figure 1B ). Significant metabolic differences between the tested cell lines were analyzed using one-way ANOVA followed by Tukey post-hoc multiple comparisons. While the tests showed that AG1522D and NFFp fibroblasts were not significantly different from each other, HT1080 and HeLa cell lines were significantly different to both fibroblasts cell lines ( Figure 1B ). Only the cancer cell line HCT116 did not exhibit a significant difference to the fibroblast cell lines. In addition, the cancer cell lines showed significantly different metabolisms when compared to each other, reflecting their heterogeneity. In summary, we found that both normal fibroblast cell lines and HCT116 cells exhibited a similar metabolism but HT1080 and HeLa cells were significantly metabolically different from each other and the fibroblast and HCT116 cells, respectively. Nevertheless, neither cell line exhibited a strong reliance on a single metabolic pathway for ATP production, but rather were able to employ both pathways to a certain extend.

After we noticed that both energy-producing pathways are operational in all of our cell lines, we tested them for their metabolic spare capacities. The spare capacities of a cell line reflect its maximum ability to respond to an increased energy demand or stress. We again employed the Seahorse XFp metabolic analyzer and Cell Mito Stress- and Glycolysis Stress Tests to characterize both the cellular respiratory- and glycolytic spare capacities, respectively. In Figure 2A , the oxygen consumption rates (OCR) of the normal fibroblast cell line AG1522D and the cancer cell lines HT1080 and HeLa during the Mito-Stress Test are shown. While normal AG1522D cells exhibited a strong FCCP-induced increase of the OCR and consequently a large respiratory spare capacity, HT1080 and HeLa cells showed only a minor increase, indicating a small respiratory spare capacity. As visible in Figure 2B , the extracellular acidification rate (ECAR) of HeLa cells increased significantly after addition of Oligomycin, indicating a large glycolytic spare capacity for this cell line. In contrast, AG1522D and HT1080 cells did not significantly react to the addition of Oligomycin and consequently exhibit only small glycolytic spare capacities. In summary, our tested normal fibroblast cell lines AG1522D and NFFp possess large respiratory spare capacities of app. 200-218%, while the tested cancer cells only have small respiratory spare capacities of app. 11-46% ( Figure 2C ). In contrast, the normal fibroblast AG1522D and NFFp cells exhibited only small glycolytic spare capacities of app. 25-30%, while the cancer cell lines HeLa and HCT116 had large glycolytic spare capacities of app. 106% and 130%, respectively ( Figure 2C ). The third tested cancer cell line HT1080 showed both a low respiratory- and glycolytic spare capacity with app. 11% and 20%, respectively ( Figure 2C ). Significant differences of metabolic spare capacities between the tested cell lines were analyzed using a one-way ANOVA followed by Tukey post-hoc multiple comparisons. Because the AG1522D and NFFp fibroblasts were not significantly different from each other, we compared their pooled data with the single cancer cell lines. Our analysis showed that all tested cancer cell lines differed significantly in their OXPHOS spare capacity from the fibroblast cells ( Figures 2A, C ). When we compared the glycolytic spare capacities of fibroblasts with the cancer cell lines, HeLa and HCT116 cancer cells differed again significantly in their spare capacity from the fibroblast cell lines ( Figures 2B, C ). Only the cancer cells HT1080 did not exhibit a significant difference of a glycolytic spare capacity when compared to fibroblasts ( Figures 2B, C ). In summary, our tested fibroblast cells exhibited large OXPHOS spare capacities and small glycolytic spare capacities while the cancer cells HeLa and HCT116 possessed small OXPHOS spare capacities and large glycolytic spare capacities. Only the cancer cell line HT1080 neither possessed an OXPHOS- nor a glycolytic spare capacity, probably because it already reached its metabolic limit to achieve its fast proliferation rate. As a result of the different metabolic spare capacities, our tested normal fibroblast- and cancer cell lines were expected to react differently to treatment with metabolic inhibitors.

Previous in vitro studies treated cancer cell lines with 4-20mM 2-DG (33–35), while clinical trials used 1.2-1.8mM 2-DG in combination with irradiation (37, 38). To employ inhibitor concentrations similar to those applied in clinical trials, we used 2.5mM 2-DG in culture medium containing 5mM glucose. This glucose concentration resembles the human blood glucose levels (41) and leads to an app. 1:2 ratio of 2-DG to glucose in our experiments. When our cell lines were cultured under these conditions for 24h and compared with untreated control cells, a proliferation reduction of app. 20-50% became visible ( Supplementary Figure 1A ). Using a Welch´s two-tailed t-test, the observed growth defect induced by 2.5mM 2-DG was significant for HT1080 and HCT116 cancer cells, respectively. To characterize the effect of 2.5mM 2-DG on the cellular energy metabolism, we incubated cells in either control medium or medium containing 2-DG and determined the ATP production rate. As shown in Figure 3A , 2-DG reduced the glycolytic ATP production rate by app. 34-53% (lower panel). HT1080 cells exhibited the strongest 2-DG dependent decrease of the glycolytic- and even total ATP production rate with 53% and 24%, respectively ( Figure 3A , upper and lower panel). The OXPHOS-dependent ATP production rate in all cell lines tested did not significantly change after treatment with 2.5mM 2-DG (Welch´s two-tailed t-test; Supplementary Figure 1B ). The test further revealed that the fibroblast cell line AG1522D showed no significant decrease in the glycolytic ATP production rate after treatment with 2.5mM 2-DG while NFFp cells and the tested cancer cell lines HT1080 and Hela exhibited a significant decrease in glycolytic ATP production rate. While the total ATP production rate did not significantly change for the AG1522D and NFFp fibroblast cells, it was significantly reduced for the tested HT1080 and HeLa cancer cells ( Figure 3A , upper panel, Supplementary Figure 1B ). In general, the OXPHOS-dependent ATP production rate did not significantly increase to compensate the loss of glycolytic ATP production, implying that cells can cope with less ATP under stress-free conditions, although all cell lines exhibited a decrease in proliferation ( Supplementary Figure 1A ). To analyze this growth delay in more detail, we determined cell cycle distribution and debris/dead cells with flow cytometry. Our with 2.5mM 2-DG treated cell lines exhibited a decreased S-phase cell population of about 34-50% and an increased G1-phase cell population of about 15-38%, suggesting that the cell lines arrest at the G1/S checkpoint ( Supplementary Figure 1C ). A Welch´s two-tailed t-test was used to analyze the significance of the observed cell cycle alterations after treatment with 2-DG. In all tested cancer cell lines, the addition of 2.5mM 2-DG induced significant changes in the cell cycle distribution, except for the G2/M phase of HCT116 cells ( Supplementary Figure 1C ). Although the amount of cell debris/dead cells increased in some of the cell lines, it represented less than 8% of the total cell population and most likely did not significantly contribute to the observed proliferation defect after 24h ( Supplementary Figure 1D ). The increase of debris/dead cells was significant for HT1080 and HCT116 cells treatment with 2.5mM 2-DG but not for HeLa cancer cells (Welch´s two-tailed t-test; Supplementary Figure 1D ). 2-DG does not only affect glycolysis but also the linked PPP that provides ribose 5-phosphate, which serves as source for nucleotide synthesis (9, 42). Consequently, treatment of cells with 2-DG could disturb nucleotide synthesis and induce the observed proliferation defect and accumulation of cells in G1, respectively ( Supplementary Figures A, C ). In addition, the PPP also provides NADPH for biosynthetic reactions and the cellular redox defense (9, 42). To characterize the effect of 2-DG on the NADP+/NADPH level and ROS, we used the HT1080 cancer cell line that reacted most sensitive to treatment with this inhibitor. We cultured the cells in control medium or medium containing 2.5mM 2-DG for 4h and 24h, respectively, and determined their total NADP+/NADPH level using a NADP+/NADPH assay. While we could not detect any significant changes after 4h between the control and 2.5mM 2-DG treated cells using a Welch´s two-tailed t-test, we saw a significant 46% reduction of the total NADP+/NADPH level after 24h treatment with 2.5mM 2-DG ( Figures 3B, C ). To complement this experiment, we analyzed ROS in 2-DG treated HT1080 cells because the observed decrease of NADP+/NADPH levels ( Figure 3C ) can negatively affect the cellular redox defense, leading to an increase of oxidative stress. To analyze ROS within the cells, the fluorescent probe CellRox Green was added after a 4h treatment with 2-DG. CellRox Green is a weak fluorescent dye that emits a stable fluorescent signal upon interaction with ROS. As visible in Figure 3D , 2-DG treated cells showed a significant increase of 30% of fluorescent signal, confirmed through a Welch´s two-tailed t-test, indicating that 2-DG can induce oxidative stress. Although NADPH depletion can induce ROS (10), oxidative stress was detectable at an earlier time point than the change in total NADP+/NADPH levels. This may be due to the fact that our NADP+/NADPH assay detects the overall decrease in NADP+/NADPH levels, but not the change in the ratio of NADP+ to NADPH, which might occur already at an earlier time point and triggers oxidative stress.

In the next step, we determined the efficiency of the NQO1 inhibitor 5-Methoxy-1,2-dimethyl-3-[(4-nitrophenoxy)methyl]indole-4,7-dione (ES-936). While no data of clinical studies with ES-936 are available, pre-clinical in vitro and in vivo analyses show that ES-936 induces a growth defect in human cancer cell lines and lead to a significant decrease in growth rates in xenograft tumors (43, 44). As the authors in these studies showed that 50nM or higher concentrations of ES-936 led already to a strong reduction of plating efficiency and anchorage-dependent growth (43), we decided to test lower ES-936 concentrations. For our experiment, HT1080 cells were treated with different concentrations of ES-936 for 1h and 24h, respectively. Subsequently, cell lysates were prepared and analyzed in an NQO1-activity test ( Supplementary Figure 2A , Figure 4 ). As visible in Figure 4 , 12.5nM ES-936 lead to an app. 19% inhibition of NQO1, while 25nM and 50nM ES-936 induced app. 47% and 60% inhibition of NQO1, respectively. The inhibitor-induced decrease of NQO1-activity was significant for all ES-936 concentrations tested at time point 4min (Welch´s two-tailed t-test; Figure 4 ). We decided to use 25nM ES-936 in our subsequent experiments as it exhibited a substantial and enduring inhibition of NQO1 of app. 41% after 24h of treatment ( Figure 4 , lower panel). In addition, this concentration also permitted the analysis of the inhibitor in combination with 2.5mM 2-DG and irradiation without lethal toxicity, as a 24h treatment of HT1080 cells with 25nM ES-936 alone did not induce a defect in cellular proliferation or a decrease in S-phase cell population ( Supplementary Figures 2B, C ). Using 25nM ES-936, we could not detect a change in the total NADP+/NADPH level in HT1080 cells after 4h and only a small decrease of app. 12% after 24h treatment ( Figure 3B, C ). Nevertheless, it was shown before that NQO1 affects the NAD(P)+/NAD(P)H ratio rather than the total NAD(P)+/NAD(P)H level (21). Additionally, we could not observe an increase in oxidative stress after 4h treatment with the rather low concentration of 25nM ES-936 alone ( Figure 3D ). Oxidative stress induced by ES-936 might be either persistent but below the detection limit of CellRox Green or requires a longer incubation time than 4h to be observable. In summary, 25nM ES-936 is an efficient inhibitor of the NQO1 enzyme and mildly affects NADP+/NADPH levels after 24h treatment at this concentration.

Following the characterization of the single inhibitors 2-DG and ES-936, we incubated our cells lines for 24h in control medium or medium containing a combination of both 2.5mM 2-DG and 25nM ES-936 and subsequently cultivated the cells for app. 10 days in inhibitor-free medium to assess proliferation. When untreated control cells reached app. 80% confluence, we fixed and stained all samples with Sulforhodamin B to determine their protein content, which correlates to cell number and thus, proliferation. As visible in Figure 5 , the treatment of the indicated cell lines with a combination of 2-DG and ES-936 led to a pronounced decrease of cell proliferation of about 38-78%, thus showing a mainly additive effect when compared to the effects of the single inhibitors. The normal fibroblast cell line AG1522D reacted least sensitive to 2-DG and the combination of 2-DG and ES-936, confirming our hypothesis that normal fibroblast cells can compensate 2-DG treatment better than cancer cells. In contrast, the normal fibroblast cells reacted similar to the inhibitor ES-936 than the cancer cell lines ( Figures 5A, B ), in line with the assumption that this inhibitor targets both normal fibroblast- and cancer cells. The decrease in proliferation was significant for the 2.5mM 2-DG and 2.5mM 2-DG+25nM ES-936 treated cancer cells lines, respectively (Welch´s two-tailed t-test). Although a slight proliferation defect of 25nM ES-936 treatment was visible in all cell lines, it was not significant except for the HCT116 cancer cells. The AG1522D fibroblasts showed only a significant decrease when treated with both, 2-DG and ES-936, although the observed proliferation defect was weaker than in the tested cancer cell lines ( Figures 5A, B ). Additionally, the effect of the inhibitor combination was tested in regard to NADP+/NADPH levels and ROS. For this, HT1080 cells were treated with the inhibitors and assessed after 4h and 24h, respectively. While a 4h incubation with both inhibitors did not show any change in NADP+/NADPH level, it was significantly decreased after 24h incubation, being only slightly lower than the level of 2-DG treated cells ( Figure 3B, C ). This is in line with the observation that a combination of 2-DG and ES-936 led to a pronounced reduction of the mitochondrial amount of the reduced forms of NAD(P)H in HT1080 cells after 6h incubation time ( Supplementary Figure 1E ). For this experiment, 2-Photon Fluorescence Lifetime Imaging Microscopy (2P-FLIM) was used to image and quantify mitochondrial NAD(P)H lifetime decay (45). An increase of oxidative stress in HT1080 cells was again detectable 4h after treatment with both inhibitors similar to that of 2-DG treated cells ( Figure 3D ). In summary, the combination of 2-DG and ES-936 showed a combined, robust inhibitory effect on cell proliferation under stress-free conditions in our tested cancer cell lines, while normal fibroblast AG1522D cells were also affected but to a lesser extend when compared to the cancer cell lines ( Figure 5 ). The decrease of the NADP+/NADPH level and increase in ROS in HT1080 cells was similar to 2-DG treated cells after 24h treatment.

To test whether inhibitor-treated normal fibroblast- and cancer cells react differently when challenged with radiation, we incubated normal AG1522D fibroblasts and HT1080, HeLa and HCT116 cancer cells with a combination of 2.5mM 2-DG and 25nM ES-396 for 4h. Subsequently, cells were irradiated with low- or high LET radiation and seeded for colony formation in inhibitor-free medium 24h post irradiation. In this experimental setting, DNA-repair takes place in the presence of the inhibitors, while colony formation proceeded in inhibitor-free medium. As shown in Figure 6A , the cancer cell line HT1080 was sensitized to both, low- and high LET irradiation (X-rays and alpha-particles, respectively) when treated with both inhibitors, visible through decreased colony formation. The cancer cell lines HeLa and HCT116 also showed – to a different extend – radiosensitization after treatment with both inhibitors and low- or high LET irradiation ( Supplementary Figures 3A, B ). In contrast, the normal fibroblast cell line AG1522D was not sensitized to low- and high LET irradiation (X-rays and Carbon-ions, respectively) when treated with a combination of both inhibitors ( Figure 6B ). Significance was assessed as indicated with either a Welch´s two-tailed t-test or the Chi-squared method (46). In summary, the combination of 2-DG and ES-936 seems to negatively affect the colony formation ability of different cancer cell lines after irradiation with low- or high LET but not of the normal fibroblast cell line AG1522D.

In the next step, we examined the DNA repair efficiency of inhibitor-treated and irradiated cells. For this, cells were treated with 2.5mM 2-DG and 25nM ES-936 for 4h and subsequently irradiated with either low- or high-LET radiation. Cells remained in inhibitor-containing medium up to 24h post irradiation and were fixed and immunostained at the indicated time points. Double-strand break (DSB) repair was analyzed using yH2AX foci as a read-out at 0.5h post irradiation to assess induction of DNA DSBs and at 24h post irradiation to detect residual damage ( Supplementary Figure 4 ). While the inhibitor-treated cancer cell line HT1080 did not exhibit a significant difference in residual yH2AX foci numbers to untreated cells 24h after irradiation with 1Gy or 2Gy of X-rays, a significant DNA repair defect could be observed after irradiation with 1Gy Lithium-ions or 2Gy alpha-particles ( Figures 7A–C , Supplementary Figure 5A ). As visible in Figure 7C , both single inhibitors contributed to the observed DNA repair defect in an additive manner in the HT1080 cancer cells. A DNA-repair defect was also observable in inhibitor-treated HeLa cells after irradiation with Lithium-ions and alpha-particles, respectively ( Supplementary Figures 5C, D ). Because S-phase cells are more susceptible to the formation of yH2AX foci due to ongoing replication that can convert base damages or DNA single-strand breaks into DSBs, we used EdU pulse labeling to identify S-phase cells and excluded them from our evaluation. As shown in Supplementary Figure 5B , S-phase cells did not contribute significantly to the observed DNA repair defect in HT1080 cells, confirming that the detected DSB-repair defect was based on irradiation induced DSBs. To exclude the possibility that the residual numbers of yH2AX foci in inhibitor-treated HT1080 cancer cells 24h post high-LET irradiation were based on apoptotic cells, we performed a TUNEL-assay. In Supplementary Figure 6 , it becomes visible that a high-LET irradiated HT1080 population contains no detectable apoptotic cells 24h after irradiation compared to DNAseI- or Staurosporine-treated or low-LET (X-ray) irradiated cells. Consequently, the elevated number of detected yH2AX foci seem not to be based on apoptotic cells. As a comparison, we analyzed whether normal AG1522D fibroblasts also showed a DNA-repair defect after inhibitor treatment and high-LET irradiation. In contrast to the cancer cell lines tested, no repair defect could be detected in the AG1522D cells 24h after alpha-particle irradiation ( Figure 7D ). We conclude that the inhibitor combination of 2-DG and ES-936 induces a DNA-repair defect after high- but not low-LET irradiation in our tested cancer cell lines but not in normal fibroblast AG1522D cells. Significance was assessed using a Kolmogorov-Smirnov test. Finally, we tested the hypothesis of Zhou et al. (47) that connects metabolism and DNA repair with each other. The authors establish in their work that specifically GTP activates a signaling cascade that promotes non-homologous end joining (NHEJ). To examine whether 2-DG and ES-936 can affect irradiation induced DSB repair through interference with GTP availability, HT1080 cells were supplemented 24h prior irradiation with either 50µM Adenosine or 50µM and 100µM Guanosine, respectively. Adenosine and Guanosine can enter the cells where the molecules are converted to ATP and GTP, respectively (47). 4h prior irradiation with 2Gy alpha-particles, cells were treated with 2.5mM 2-DG and 25nM ES-936. DNA repair efficiency was assessed again 24h post irradiation. As visible in Figure 7E , the addition of Guanosine but not Adenosine decreased the number of DNA damage foci after 24h significantly (Kolmogorov-Smirnov test), implying that specifically GTP can mediate DSB repair and that the inhibitor combination of 2-DG and ES-936 could indeed interfere with GTP availability.

The cancer cell lines HT1080 (CLS), HeLa (ATCC) and HCT116 (kindly provided by B. Vogelstein) were cultured in DMEM with either 4.5 g/L (high) (Gibco) or 1.0 g/L (low) (Sigma-Aldrich) glucose, 10% FCS, 1% Glutamine and 1% Penicillin/Streptomycin at 37°C and 95% humidity. The normal fibroblast cell lines AG1522D (Coriell) and NFFp (51) were cultured in DMEM with either high- or low glucose, 20% FCS, 1% Glutamine, 1% non-essential amino acids and 1% Penicillin/Streptomycin at 37°C and 95% humidity. Cell lines were maintained in high glucose medium, while experiments were performed in low glucose medium to allow a 1:2 ratio of 2-DG and glucose.

A Seahorse XFp analyzer (Agilent) was used to characterize the metabolic properties of the indicated cell lines. Cells were analyzed 24h post seeding, following the manufacturer´s protocol for the Real-Time ATP Rate Assay, Cell Mito Stress Test and Glycolysis Stress Test.

Cells were seeded to obtain 30% confluence and 24h later treated with 2.5mM 2-DG in low glucose medium to allow a 1:2 ratio of 2-DG and glucose. After a 24h incubation time, untreated and treated cells were counted using a coulter counter (Beckman). Untreated samples were set as 100% proliferation and treated samples were normalized to this value.

Cells were seeded to obtain 30% confluence and 24h later treated with 2.5mM 2-DG in low glucose medium to allow a 1:2 ratio of 2-DG and glucose. After a 24h incubation time, cells were harvested, fixed with ice cold 70% EtOH and stained with DAPI. Cell cycle distribution and amount of dead cells/debris were analyzed with a CytoflexS (Beckman Coulter) and FlowJo (BD) software.

Cells were seeded into a 96-well plate to obtain 30% confluency and 24h later treated with the indicated inhibitors in low glucose medium. After 4h and 24h, cells were processed following the manufacturer´s protocol of the NADP/NADPH-Glo assay (Promega). For the 24h time point, cell numbers were normalized to account for the altered cell proliferation due to treatment with 2-DG. For determination of the mitochondrial amount of NAD(P)H, cells were seeded in 35mm glass bottom petri-dishes (Greiner) and were either untreated or treated with the inhibitors for 6h prior analyses. 2-Photon Fluorescence Lifetime Imaging Microscopy (2P-FLIM) was used to image and quantify mitochondrial NAD(P)H lifetime decay (45). The 2P-FLIM setup is based on an Olympus IX71 microscope equipped with a Tokai Hit climate chamber (37°C, 5%CO2), a DCS-120 confocal FLIM System (Becker&Hickl) and a mode locked titanium sapphire laser (Coherent Chameleon II) operated at 720nm and 80MHz. Images were collected by time correlated single photon counting through a water immersion 60X 1.2 NA lens. Emission was recorded by a non-descanned HPM-100–07 hybrid detector (Becker&Hickl) after reflection by a long pass dichroic mirror and passing a FSP-680 blocking filter, a T500SPXR beam splitter and a 458/64 band-pass filter (all Semrock). Decay curves of the 2P-FLIM Images were generated by a bi-exponential fit fixing the fast decay time for free NAD(P)H to 350ps (SPCimage software Becker and Hickl). The quality of the fits were checked by Chi-squared and the uniform distribution of residuals. Pixel intensities and relevant fitting parameters were exported and pixels corresponding to the mitochondria analyzed using in-house written software (Image D; https://davideilenstein.github.io/ImageD/index.html).

Cells were seeded to obtain 30% confluence and 24h later treated with the indicated inhibitors in low glucose medium. After 4h, CellRox green (Invitrogen) was added for 30 minutes to the cells. Subsequently, cells were PFA fixed and analyzed with a fluorescent microscope.

Cells were seeded to obtain 30% confluence and 24h later treated with the indicated concentrations of ES-936 in low glucose medium. After 1h and 24h cells were processed following the manufacturer´s protocol of the NQO1 Activity Assay (Abcam). Protein concentrations were determined using a Pierce BCA Protein Assay (Thermo Fisher Scientific) and Spectramax i3X (Molecular Devices) and 25µg of all lysates were loaded onto a 10% SDS-gel to confirm the use of equal amounts of lysates in the assay.

Cells were seeded to obtain 30% confluence and 24h later treated with 25nM ES-936 in low glucose medium. After 24h, 10µM EdU was added for 30 minutes to the cells. Subsequently, cells were PFA fixed and cells were processed following the manufacturer´s protocol of the EdU Cell Proliferation Kit for Imaging (Baseclick), using a 594nm antibody (BCL-EdU 594-1, Base-Click; AF594-Picolyl-Azide, Jena Bioscience) and imaged with a fluorescent microscope.

Cells were seeded in 96-well plates and 24h later treated with the indicated inhibitors in low glucose medium. After 24h, the medium was replaced by inhibitor-free, low glucose medium and cells were cultured for several days until the untreated control sample of each cell line reached 80% confluency. Subsequently, cells were fixed with 10% TCA and stained with 0.057% Sulforhodamin B in 1% acetic acid. The dye was removed from the fixed cells by adding 10mM Tris Base pH 10.5 and measured with a Spectramax i3x (Molecular Devices) to correlate the intensity of the dye solution with cellular proliferation (52). Control samples were set as 100% proliferation and treated samples were normalized to this value.

Cells were seeded to obtain 30% confluence and 24h later treated with the indicated inhibitors in low glucose medium. After 4h, cells were irradiated with the indicated doses and radiation qualities and 24h later seeded in inhibitor-free, low glucose medium for colony formation. Depending on the cell lines, cells were cultured for up to 10 days until colonies with more than 50 cells were formed. Subsequently, colonies were fixed and stained with 3x Methylenblue for 30 minutes and counted for analysis.

Cells were seeded to obtain 30% confluence and 24h later treated with the indicated inhibitors in low glucose medium. After 4h, cells were irradiated with the indicated qualities of irradiation and PFA fixed after either 0.5h or 24h. Prior fixation, 10µM EdU was added to label S-phase cells. Subsequently, cells were processed following the manufacturer´s protocol of the EdU Cell Proliferation Kit for Imaging (Baseclick), using a 594nm antibody (BCL-EdU 594-1, Base-Click; AF594-Picolyl-Azide, Jena Bioscience), followed by antibody staining for yH2AX (05-636, Millipore; Alexa Fluor 488 A-11017, Invitrogen) and DAPI. Samples were imaged as z-stacks using a Leica SPE confocal microscope equipped with a 63x and 100x lens, respectively. Maximum intensity projections of the fluorescence image stacks were analyzed using manual foci counting in Fiji.

Cells were seeded to obtain 30% confluence and 24h later treated with the indicated substances in low glucose medium. After 4h, inhibitor-treated cells were irradiated with the indicated qualities of irradiation and PFA fixed after 24h. Subsequently, cells were processed following the manufacturer´s protocol of the In Situ Cell Death Detection Kit (Roche) and analyzed with a fluorescent microscope.

For X-ray irradiation, an X-ray machine (MXR 320/26, GE Sensing & Inspection Technologies. Ahrensburg, Germany) with a tube voltage of 250 kV and app. dose rate of 2.0 Gy/min was used. Alpha-particle irradiation was performed employing an 241Am-source. Lithium-ion irradiation (1 Gy) was performed at the linear accelerator UNILAC of the GSI Helmholzzentrum für Schwerionenforschung GmbH (10.9 MeV/u, 39.5 keV/μm). Carbon-ion irradiation took place at the Heavy-ion Synchrotron (SIS) of the GSI Helmholzzentrum für Schwerionenforschung GmbH or at the Marburger Ionstrahl-Therapiezentrum with samples positioned in the midst of a 40mm SOPB (mean dose LET of app. 65 keV/µm; dose 0.5-1Gy).

2-Deoxy-D-Glucose (Sigma); ES-936 (Santa Cruz).

´N´ indicates number of independent experiments. Technical replicates are given by ´n´. Significance was assessed using the following tests as indicated: one-way ANOVA and Tukey post-hoc multiple comparisons, Welch´s two-tailed t-test, Chi-squared method (46), Kolmogorov-Smirnov test; (n.s.= not significant, *p<0.05, **p<0.01; ***p<0.001). Error bars indicate standard error of the mean (SEM).