Arginine Deprivation Therapy: Putative Strategy to Eradicate Glioblastoma Cells by Radiosensitization

Arginine is a semi-essential amino acid required for numerous cellular processes such as protein synthesis and arginylation, polyamine production, metabolism of other amino acids and nucleotides, as well as creatine, urea, and nitric oxide (NO) synthesis (1, 2). The intracellular arginine pool is usually fueled by arginine uptake from the blood and interstitial fluid through cationic amino acid transporters (CAT), via protein degradation, or by conversion from citrulline catalyzed by the enzymes argininosuccinate synthase (ASS) and argininosuccinate lyase (ASL; refs. 1, 3). The latter two processes are sufficient to maintain cell survival in normal tissues in the event of exogenous arginine shortage. Aggressive cancer cells with their high proliferative activity requiring enhanced energy disposal, DNA synthesis, and protein metabolism have an extraordinary demand for arginine that cannot be covered by intracellular sources. This is particularly striking for malignancies deficient in one key enzyme of arginine synthesis due to genetic or epigenetic modifications. In most cases, except for hepatocellular carcinoma, the enzyme ASS is lacking in these arginine auxotrophic tumors, and enzymotherapeutic arginine deprivation therapy (ADT) has already proven to effectively inhibit growth of such cancers in preclinical studies and clinical phase I–III trials (1, 4–8). ADT combined with radiotherapy is not yet in clinical trial. However, our group has shown recently that ADT can augment the radioresponse even in ASS-positive colorectal cancer cells (9, 10), raising hope for improving the treatment outcome also for other arginine nonauxotrophic tumor entities.

ADT is a particularly attractive strategy for brain tumors and metastases as it is a systemic approach based on the catabolic depletion of arginine from the blood supply. It is thus independent of the blood–brain barrier, which, in spite of its partially enhanced permeability in brain cancer patients (11, 12), often impairs adequate drug distribution. The translation of the initial finding that arginine withdrawal induces growth arrest in various tumor entities in vitro via ADT-based animal studies into clinical trials has just commenced and progresses rapidly (1, 6, 8, 9, 13–15).

Recent observations in glioblastoma (GBM) models (16–18) support the idea that ADT might be a treatment option for brain cancer patients. The aim of our project was to systematically extend these previous studies by addressing the impact of ADT not only on growth behavior but also on all three major characteristics of GBMs—the high proliferation rate and survival capacity, the strong invasive potential, and the severe intrinsic radiation resistance (19).

The GBM cell lines U251-MG and U138-MG (both p53mut) as well as isogenic U87-MG-shLuc control cells (p53wt) and their p53-knockdown counterpart U87-MG-shp53 were used. The original cell lines were purchased in 2010 (CLS Cell Lines Service). The generation of isogenic and knockdown cell lines is described in Supplemental Materials and Methods and (20–22). The correct genetic profiles of the frozen stocks derived from originally and genetically modified cells were confirmed prior to use via the multiplex PCR kits Mentype NonaplexQS Twin (Biotype AG) and PowerPlex 16 (Promega Corp.; Institute of Legal Medicine, TU Dresden, Germany). Cultures were free of mycoplasms and routinely regrown from the validated frozen stocks for <100 cumulative population doublings using Eagle's Minimum Essential Medium (MEM) containing 1.0 g/L D-glucose, 292 mg/L l-glutamine, 25 mmol/L HEPES, and 2.2 g/L NaHCO3, and supplemented with 10% heat-inactivated FCS as well as 1% penicillin/streptomycin (10,000 U/mL/10 mg/mL; all from PAN-Biotech). Exponentially growing cultures were kept at 37°C in a humidified 5% CO₂ in air atmosphere. Single-cell suspensions for passaging and for setting up experiments were obtained by enzymatic and mechanic means using a 0.05%/0.02% trypsin/EDTA solution in PBS (PAN-Biotech). A CASY TTC device (Roche Innovatis AG) was applied for cell counting, cell volume analysis, and culture quality assessment.

Spheroid cultures were initiated by seeding defined concentrations of single-cell suspensions from exponentially growing monolayer cultures onto 1.5% agarose-coated 96-well plates. The designated standard spheroid size at day 4 in culture was 360 to 400 μm. Spheroids were further cultured in liquid overlay and monitored for integrity and volume growth as described earlier (refs. 9, 23, 24; for details, see Supplemental Materials and Methods).

ADT was mimicked by the use of arginine-free culture medium (Pan Biotech) or achieved by exposure to 2 U/mL yeast-derived recombinant human arginase I (rhA, provided by the Institute of Cell Biology, NASU, Lviv, Ukraine). Successful deprivation of arginine by the enzyme at this concentration was scrutinized before use by a reverse-phase HPLC technique as described in ref. 25. Arginine in culture media is degraded within <10 minutes, and rhA activity remains at maximum level for at least 72 hours. It was documented earlier that the treatment induces an essential drop in intracellular free arginine within 15 to 30 minutes; arginine decreases below detection level within 1 hour (25, 26). In case of amino acid dietary medium, both control and arginine-deficient media were supplemented with 10% heat-inactivated dialyzed FCS deprived of molecules <10 kDa (Pan Biotech). Spheroids exposed to arginine-free media required the transfer onto 96-well plates coated with 1.5% agarose prepared with the corresponding serum-free amino acid-deprived medium. In some settings, media contained citrulline (Sigma Aldrich) at an arginine-equimolar, hyperphysiological concentration of 0.6 mmol/L or at a concentration of 0.05 mmol/L reflecting the blood/plasma level. Irradiation was performed at room temperature using a single dose X-ray approach (200 kV, 0.5 mm Cu filter, YxlonY.TU 320; Yxlon.international).

For two-dimensional (2-D) growth assessment, 1 × 10⁴ single cells/well in 2 mL culture medium were seeded into 6-well plates and allowed to adhere for 24 hours; rhA was then added, and cells were counted after 1, 3, and 5 days of treatment as well as 3 days after treatment, i.e., after retransfer to normal conditions. At least two independent experiments (N = 2) were performed each with n = 3 to 4 wells per condition.

Spheroid volume was determined before, during, and after exposure to arginine-comprised and arginine-free conditions, respectively, with or without citrulline. For this purpose, spheroids were washed with PBS and carefully transferred in the specific dietary media onto the respective agarose-coated 96-well-plates. After defined time intervals, spheroids were conveyed back into standard liquid-overlay culture and monitored for another 24 days. A minimum of 12 individual spheroids were analyzed per condition, and three independent aliquots of at least 8 spheroids per treatment condition were collected and dissociated to determine the numbers of membrane-intact cells per spheroid. The operating schedule of the 2-D and three-dimensional (3-D) growth, regrowth, and survival assays is outlined in Supplemental Fig. 1A.

For dynamic cell-cycle assessment, exponential monolayer cells treated with or without rhA for a total of 24 hours were pulse-chased with the thymidine analogue 5-ethynyl-2′-deoxyuridine (EdU; Invitrogen). The EdU pulse (10 μmol/L, 1 hour) and chase periods (0 and 8 hours) were scheduled in a way that all samples could be harvested at the same time. Single-cell suspensions were prepared, and EdU incorporation was detected using the Click iT EdU Flow Cytometry Assay Kit (Invitrogen). RNase A (0.1 mg/mL, 37°C, 30 minutes; Invitrogen) was applied for RNA elimination and propidium iodide (PI, 50 μg/mL; Sigma-Aldrich) for stoichiometric DNA counterstaining. For more details, see ref. 27. PI was also applied for static DNA analyses using the handling protocol from (28). All samples were measured with a BD FACS Canto II flow cytometer (BD Biosciences). A minimum of 2 × 10⁴ single cells were recorded using the PI area and width signals for doublet exclusion. G_1/0, S-, and G₂–M-phase fractions were determined in DNA histograms using the univariate Dean–Jett Fox model implemented in the FlowJo software version 7.6.4 (Tree Star). Gating of particular EdU-positive cell fractions 0 and 8 hours after pulse-labeling allowed to detect and follow actively cycling S-phase cells.

Clonogenic survival was assessed in colony formation assays. Here, low cell numbers (150–6,000 cells/well), adapted to the individual GBM cell line and to the increasing dose of irradiation (0–10 Gy), were seeded into 6-well plates and allowed to adhere for 1 day. Control cells were then irradiated, whereas in the other samples, supernatant was exchanged to rhA-containing media with or without citrulline. Accordingly, arginine-deprived cells which were growth-arrested during treatment were irradiated 24 hours later. In a separate set of experiments, rhA exposure was combined with the endoplasmic reticulum (ER) stress response modifiers salubrinal (20 μmol/L; Sigma Aldrich) and dimethyl sulfoxide (DMSO, 2%; Sigma Aldrich). A 4 mmol/L stock solution of salubrinal in DMSO was prepared and freshly diluted in medium prior to use; accordingly, control cells were exposed to medium containing 0.5 mmol/L DMSO. Twenty-four hours later, treated cells were irradiated (0–10 Gy and 0, 6, and 8 Gy for salubrinal/DMSO assays, respectively). Standard culture conditions in all samples were restored 1 hour after irradiation, and cells were incubated for another 9 to 13 days depending on the cell-line–specific doubling times to allow >5–6 doublings. Colonies were then fixed, stained, and counted manually to determine the plating efficiencies (PE). Surviving fractions (SF) were normalized for cell-line– and treatment-dependent differences in PE to quantify radioresponse. Data were reproduced (N = 3 with n ≥ 3 wells per experiment and treatment condition), and cell survival curves were fitted employing the linear-quadratic model (D: dose; A and B: variables defining the irradiation dose):

Cell survival curves after combined treatment were considered distinct from controls (irradiation only) if values A or B significantly differed.

Standard sized spheroids were washed with PBS, transferred into arginine-free medium with or without citrulline, and irradiated with single doses of 0 to 30 Gy at day 5 of arginine depletion. Twenty-four hours after irradiation, arginine was resupplemented to the standard concentration of 0.6 mmol/L in MEM. Control spheroids were irradiated in the presence of arginine at standard diameter. Cultures were monitored for a period of 60 days after treatment. Spheroids which did not regrow to >200 μm and/or did not enlarge at least 3 times in sequence were declared as "controlled." A minimum of 29 individual spheroids were analyzed per treatment condition.

Spheroid growth delay t_SGD represents the different time intervals of treated versus control spheroids to reach 5 times the starting volume (5 x V₀), with the starting volume (V₀) being defined as the spheroid volume directly before irradiation and treatment, respectively. To determine spheroid growth delay, growth and regrowth kinetics were mathematically modeled via the Gompertz function as described earlier (23, 24).

Spheroid control probability (SCP) describes the proportion of controlled spheroids as a function of irradiation dose. SCP curves were fitted with a sigmoid dose-response model according to the tumor control probability in vivo assays (D: dose; a and b: variables):

The dose at which SCP amounts to 50% (SCD₅₀) was deduced from the SCP curves; the quotient of SCD₅₀ values (combined treatment/irradiation) yields the dose reduction factor (DRF).

Spheroids at standard size were washed in PBS and either studied under acute arginine-deprivation or pre-exposed to arginine-free medium for a period of 5 days to mimic chronic arginine deficiency. Spheroids were transferred onto flat-bottom wells coated with fibronectin (FN) or collagen-I (Col-I, migration), or were embedded into Col-I gel (invasion) and monitored under acute (no pre-exposure) and chronic (5-day pre-exposure) dietary conditions. The invasion protocol included the utilization of fluorescence-labeled U251-MG-eGFP, U87-MG-shLuc-eGFP, and U87-MG-shp53-eGFP spheroids and the combination of z-stacks of about 40 images with a distance of 5 μm. Migration and invasion of spheroid single cells were recorded over a period of 12 hours via time-lapse microscopy using an AxioObserver.Z1 system. Experimental and analytical details are given in Supplemental Fig. 1B. A minimum of 8 individual spheroids were monitored per condition.

Whole-cell protein was extracted from exponentially growing 2-D and 3-D cultures untreated or exposed to ADT for 48 hours. The basic extraction procedure and protein quantification have been described elsewhere (9). In brief, the lysis buffer for protein extraction contained 2.0% phosphatase-inhibitor-cocktail 1 + 2 (Sigma Aldrich), 10% complete mini protease inhibitor cocktail (Roche Applied Science) dissolved in PBS, 1% Na₃VO₄, 1% PMSF (100 mmol/L), and 0.2% NaF in RIPA buffer. Protein isolated from HepG2 cells (hepatocellular carcinoma) served as positive standard. Protein lysates were completed by dithiothreitol-supplemented (1:19) protein loading buffer (1:5) and denatured at 95°C for 5 minutes. Equal protein weights were disassembled in 10% SDS-PAGE. Blotting, antibody incubation, and detection were performed as detailed earlier (9) using the antibodies listed in Supplemental Materials and Methods. ASS and ASL signal intensities were semi-quantitatively assessed relative to an α-tubulin loading control and normalized to a positive protein standard (HepG2 protein extract). Three independent protein lysates were analyzed for each condition.

Differences in SF, cell numbers per spheroid, and migration and invasion distances were evaluated by independent Mann—Whitney U tests, whereas differences in cell-cycle distribution were analyzed by the paired Wilcoxon rank-sum test. Both tests included the Bonferroni–Holm correction for multiple testing. ANOVA comprising Bonferroni correction was applied for 3-D cell growth assays. Linear regression was used for the analysis of cell survival curves and for the comparison of the parameters of the linear-quadratic model. All tests were performed using SPSS Statistics 21 (IBM Corporation). Logistic regression was employed to compare SCD₅₀ values as well as DRF and to fit the corresponding curves. To determine 95% confidence intervals, subsequent bootstrapping with 1,000 samples was performed using STATA/SE 11.2 (StataCorp LP). Values are documented according to their significance level as P < 0.05 (*), P < 0.01 (**), and P < 0.001 (***).

Arginine deprivation over a period of 48 hours by exposure to a modified rhA was recently found to be cytotoxic for GBM cells in 2-D culture (17). We examined the impact of ADT in more detail by using different GBM cell line models, variable treatment intervals, and by analyzing not only acute growth effects but also the cells' potential to recover.

We observed a complete growth arrest under acute arginine-deprived conditions for all GBM cell lines tested herein. Furthermore, cells were not capable to recover and show regrowth after completion of ADT, if the treatment interval was ≥5 days (Fig. 1A). Basically, all three GBM lines included in the initial experimental series express the enzymes involved in citrulline to arginine conversion, i.e., ASS and ASL, but intrinsic protein levels differ (Fig. 1C). Accordingly, the cells cannot be considered intrinsically auxotrophic for arginine when the arginine precursor citrulline is bioavailable. We therefore also monitored changes in ASS and ASL protein upon ADT and found a positive, yet not significant, trend of ASS in U87-MG-shLuc and U251-MG, whereas ASL protein seemed rather reduced, a phenomenon that remains ambiguous. Overall, the ASS/ASL^high U87-MG-shLuc cells appeared to be less sensitive to ADT than the other GBM cell line models as they more efficiently regenerate after an arginine deprivation of >3 days and upon bioavailability of citrulline. However, citrulline—even at concentrations far above blood level—neither abolished the ADT-induced growth arrest nor was it sufficient to fully compensate the ADT-related loss of regeneration capacity (Fig. 1B).

GBMs are known for their high intrinsic radioresistance. Our previous study in colorectal cancer models indicated that ADT can enhance radiosensitivity of human epithelial cancer cells (9). This motivated us to examine whether combination of systemically applicable ADT plus irradiation could be a treatment option for GBM patients.

Twenty-four–hour ADT reduced the PE of U251-MG from 44% ± 7% to 26% ± 10% and of U87-MG-shLuc cells from 10% ± 4% to 6% ± 3%. Normalization of the SF after irradiation was thus essential to identify a radiosensitizing potential. The respective data shown in Fig. 2 reveal that both U251-MG and U87-MG-shLuc cells are sensitized to irradiation by arginine deprivation in the 2-D assay. However, the effect strongly differed in magnitude. The ASS^high/ASL^low U251-MG revealed much higher susceptibility with a significant difference in radioresponse for doses ≥2 Gy (p_{2Gy,4Gy,6Gy,8Gy,10Gy} < 0.01) in the absence of citrulline and for doses ≥4 Gy (p_{4Gy,6Gy,8Gy,10Gy} < 0.01) in the presence of citrulline. The ASS^high/ASL^high U87-MG-shLuc cells were sensitized at high doses only (p_8Gy,10Gy < 0.05 without citrulline, P_10Gy < 0.05 with citrulline).

Beside the ASS expression rate, the GBM models used herein differ in their intrinsic p53 status. U138-MG and U251-MG are p53-mutant with loss of functional protein (29, 30), whereas U87-MG cells are p53-wildtype (p53-wt). Because of the high frequency of TP53 mutations in GBM, we implemented a p53-knockdown derivative of the U87-MG cell line (U87-MG-shp53) for comparison with its counterpart U87-MG-shLuc to study the role of p53 in ADT response in more detail. In this model, loss of functional p53 (Supplemental Materials and Methods) did not per se alter radioresponse in the clonogenic survival assay (Fig. 2). Also, U87-MG-shp53 cells displayed intrinsic ASS levels at least as high as the U87-MG-shLuc cells, and the ASL contents did not differ (Fig. 1D). Putative differences in the response of U87-MG-shLuc and U87-MG-shp53 to ADT can thus not be attributed to p53-related changes in intrinsic radiosensitivity or ASS/ASL expression.

Indeed, the cell growth and regrowth behavior of U87-MG-shp53 cells upon rhA treatment with and without citrulline supplementation resembled that of its p53-wt ancestor (Fig. 1A and B). Accordingly, p53 knockdown in U87-MG cells did not overcome ADT-induced growth and cell-cycle arrest in 2-D culture. However, the radiosensitizing effect of ADT and arginase treatment, respectively, was enhanced upon p53 knockdown (Fig. 2) with a significant loss in clonogenic cell survival at ≥6 Gy single-dose irradiation (p_6Gy,8Gy,10Gy < 0.01). Furthermore, this effect was not at all compensated by citrulline, assuming that p53-knockdown/mutant (i.e., loss of functional p53 protein) accounts for a greater citrulline-independent radiosensitization.

In summary, our data reveal that the radiosensitizing effect of ADT in 2-D culture is more prominent in the p53-deficient U87-MG-shp53 GBM cells and appears autonomous of ASS and ASL protein expression.

The relative radioresponse is critically affected by the proliferative activity and cell-cycle distribution of cells. Proliferating cells are in general more susceptible than arrested cells. On the other hand, cells in G₂–M phase are most sensitive to DNA-damaging factors, whereas G₁-phase cells have a much lower sensitivity (31). We performed flow cytometric cell-cycle analyses to evaluate if growth arrest by ADT could relate to radiosensitization due to accumulation of cells in G₂–M phase. DNA histograms revealed that all GBM cell cultures show a higher proportion of cells in G_1/0 phase, whereas the S-phase and/or G₂–M-phase fractions are rather reduced upon ADT (Fig. 3A and Supplemental Fig. 2A).

Dynamic cell-cycle analysis using a 1-hour EdU pulse was applied because (i) a phase-independent cell-cycle arrest by ADT was described recently in colorectal cancer cells (27) and (ii) the changes seen in the DNA histograms of GBM cultures did not fully explain their growth arrest. Figure 3B representatively documents the flow cytometric DNA dot blot diagrams for U87-MG-shp53 cells, showing that roughly all cells arrest under ADT, i.e., S-phase cells do no longer actively take up EdU (0-hour measurement: active S-phase fraction, 24.8% for control, 0.7% for ADT, and 2.8% for ADT + citrulline). The effect is not sufficiently compensated by hyperphysiologic citrulline, because 8 hours later, <1% of the cells has entered the next G₁-phase in both rhA-treated arms in contrast to 18.5% in the arginine-rich control. This behavior was confirmed for all other cell lines (Supplemental Fig. 2B). Growth inhibition by ADT is thus clearly due to cell-cycle arrest but not causally related to radiosensitization.

Arginine deprivation induces ER stress in human solid cancer cells including U251-MG (10). Observations in cell lines from other tumor entities indicate a causal link between ER stress response and radiosensitization by ADT. In order to prove this hypothesis for GBM cells, we assessed clonogenic survival at 6 and 8 Gy for U87-MG-shp53 treated with and without rhA in the absence and presence of two different ER stress inhibitors. The results documented in Fig. 3C indeed show a reduced radioresponse when ADT is combined with the ER stress response modulators. However, application of both 2% DMSO and 20 μmol/L salubrinal resulted in a pronounced loss of radiosensitivity in U87-MG-shp53 cells even in the presence of arginine. The same was true for U251-MG cells. The data are thus not sufficient to unequivocally prove our hypothesis but rather imply that ER stress response pathways are highly relevant for GBM radioresponse in 2-D culture in general.

We described earlier that a 3-D culture environment renders epithelial cancer cells profoundly less susceptible to single amino-acid starvation (32). It was therefore of utmost relevance to validate the effect of combined treatment in a 3-D cell context. We applied a sophisticated long-term spheroid assay that was designed to assess the putative curative potential of new combination treatment strategies in the 3-D environment before turning into whole animal studies.

The two p53-mutated GBM cell lines U251-MG and U138-MG form regular, round-shaped, and membrane-intact multicellular spheroids, but turned out to show no intrinsic 3-D volume growth under standard culture conditions even in the presence of arginine (Supplemental Fig. 3A). In contrast, U87-MG-shLuc and U87-MG-shp53 grow well under these conditions, indicating that the proliferation arrest is not causally related to the p53 status. As a further prerequisite, we compared ASS and ASL protein levels in 2-D and 3-D cultures. ASS but not ASL level in GBM cells tends to change from 2-D to 3-D conditions; however, the differences were neither systematic nor significant. Nonetheless, ADT-related alterations in protein expression were qualitatively comparable in the 2-D and 3-D environment (Supplemental Fig. 3B and 3C).

We next analyzed the impact of 1, 5, and 10 days of ADT on 3-D volume growth and regrowth capacity for U87-MG-shLuc and U87-MG-shp53 spheroids. ADT led to a clear spheroid growth arrest in both models (Fig. 4A and C). This observation was confirmed by counting the viable cell number in spheroids grown either in the absence or in the presence of arginine for 5 and 10 days (Fig. 4B and D). In contrast to the situation in 2-D culture, all spheroids completely restored growth upon arginine resupplementation. The regrowth kinetics was exiguously slowed only after long-term (10 days) ADT, i.e., the time to reach the analytical end volume (5 x V₀) after arginine supplementation increased from 4 (control) to 5 days.

Note that the application of rhA to therapeutically eliminate arginine in the supernatant of U87-MG-shp53 spheroids resulted in an acute growth arrest analogous to arginine-free media (Supplemental Fig. 4A). Spheroid assays were therefore performed using dietary ADT.

For the evaluation of radioresponse in 3-D, we adapted our well-established growth delay and SCP assays to the U87-MG models. The approach allows the estimation of SCD₅₀ as analytical endpoint reflecting the clinically relevant readout of tumor control dose 50% (TCD₅₀) determined in xenograft models after irradiation.

We first applied the p53-knockdown U87-MG-shp53 cells due to their higher susceptibility to ADT in the 2-D radioresponse assays. In the SCD₅₀ approach, ADT combined with single-dose irradiation up to 12.5 Gy could not prevent U87-MG-shp53 spheroid regrowth but led to a massive growth delay (Fig. 5A and Supplemental Fig. 4B). This growth delay effect was less than additive at lower doses but increased exponentially to become supra-additive at intermediate doses of 12.5–15 Gy (Fig. 5B and Supplemental Table 1). Citrulline at physiologic concentrations during ADT did not abolish this effect. First signs of controlled (non-regrown) spheroids over a period of 60 days after irradiation were seen at a dose of 15 Gy only under ADT. Controlled spheroids in all groups, yet different proportions, were documented at 17.5 Gy shown in Fig. 5C as function of time after treatment. Here, an advantage but not full compensation of the ADT effect in the presence of citrulline became obvious.

The complete SCP curves finally revealed that U87-MG-shp53 spheroids are significantly and strongly radiosensitized in the absence of arginine (Fig. 5D). The SCD₅₀ for U87-MG-shp53 control spheroids was 20.8 Gy (95% CI, 20.2–21.4) and differed significantly (P < 0.001) from the values determined under arginine-free conditions in the absence (SCD₅₀: 16.5 Gy; 95% CI, 16.0–17.0) and presence of citrulline (SCD₅₀:17.3 Gy; 95% CI, 16.7–17.9). The therapeutic benefit of ADT is reflected by a DRF of 1.27 (95% CI, 1.21–1.32, P < 0.001). Note that bioavailability of the arginine precursor citrulline did not abrogate the ADT-induced radiosensitization in this model resulting in a DRF of 1.21 (95% CI, 1.15–1.27, P < 0.001).

ADT-induced radiosensitization of p53-wt U87-MG-shLuc cells was quite weak in the 2-D assay. It was unclear if the effect would be entirely lost in the 3-D environment or more pronounced due to the fact that higher doses are applied. The assessment of spheroid volume growth delay and monitoring of SCP revealed that (i) U87-MG-shLuc cells in spheroids in contrast to the 2-D assay are more sensitive to radiotherapy than their p53-knockdown counterparts, (ii) ADT-induced radiosensitization is clearly visible in the 3-D environment, and (iii) the cells' response to treatment is qualitatively but not quantitatively similar to the U87-MG-shp53.

Again, the growth delay effect became supra-additive at intermediate doses, i.e., 10–12.5 Gy for this spheroid type (Fig. 5E and F; Supplemental Fig. 4B and Supplemental Table 1). Controlled U87-MG-shLuc spheroids in all treatment arms were seen at doses ≥15 Gy (Fig. 5G), and the SCD₅₀ values calculated from the SCP curves (Fig. 5H) were 16.4 Gy (95% CI, 15.8–16.9) under arginine-rich, 13.8 Gy (95% CI, 13.3–14.4) under arginine-free, and 15.0 Gy (95% CI, 14.5–15.4) under arginine-free, citrulline-supplemented conditions. As for the U87-MG-shp53, SCD₅₀ values of U87-MG-shLuc spheroids in the treatment arms significantly differed (P < 0.001) from the controls. However, the DRF calculated for arginine-free conditions was 1.18 (95% CI, 1.12–1.24, P < 0.001) and only 1.09 (95% CI, 1.05–1.14, P < 0.001) in the presence of citrulline. This reflects a lower therapeutic effect of the ADT treatment and a higher compensating potential of citrulline bioavailability in the p53-wt U87-MG-shLuc compared with the U87-MG-shp53 model.

Pavlyk and colleagues (16) demonstrated recently that arginine deprivation reduces the migration and invasion capacity of U87-MG and U251-MG in various 2-D assays. We therefore hypothesized that ADT—best case—can have a triple benefit for GBM treatment outcome by (i) inhibiting growth, (ii) inducing radiosensitivity, and (iii) diminishing invasion. Accordingly, we analyzed the migration of U87-MG-shp53 spheroid cells on FN and Col surfaces and their invasion when embedded into a 3-D Col-I gel under acute and chronic ADT, i.e., without and with a 5-day pre-exposure to arginine-deprived conditions.

A significant reduction in migration distance on FN was observed for both acute and chronic ADT conditions (P < 0.02; Fig. 6A and Supplemental Fig. 5). Yet, the effect mainly manifests as parallel shift but not as reduced slope in the curves describing migration distance as function of time. This suggests that not migration rate per se is affected. In culture dishes, we observed reduced adherence of single cells in arginine-free medium. The reduced migration radius on FN may thus also be the consequence of a delayed adherence of spheroids and spheroid cells, respectively, to the extracellular matrix (ECM)-coated surface under arginine dietary conditions. A similar effect is not seen on Col; here, ADT caused no difference in adherence or migration rate (Fig. 6A and Supplemental Fig. 5). We further applied a spheroid 3-D invasion assay, which requires degradation of the ECM for cells to migrate out of the spheroid. Supplemental Videos 1 to 3 provide representative time-lapse microscopic videos, which indicate that the invasion capacity of U87-MG-shp53 spheroid cells is not modified by ADT. Figure 6B documents the corresponding average invasion distances as function of time. Although we could not prove the hypothesized triple effect in our 3-D model, the data insistently demonstrates that ADT-induced growth arrest and radiosensitization are unlikely to be accompanied by an adversely stimulated invasion.

Over the past decades, various therapy strategies to target GBM have been invented, but none of them evidently advanced patients' outcome. Since the establishment of the current standard-of-care radiochemotherapy with temozolomide about 10 years ago, the prognosis of newly diagnosed GBM patients has not further improved with an overall survival of only 14.6 months (19, 33, 34). Large epigenetic and genetic heterogeneities contribute to the challenges of this disease (19), with the tumor-suppressor gene TP53 as the most common gene altered in approximately 28% of primary and 65% of secondary GBMs (35–37). The increasingly recognized potential of ADT as new metabolic targeting strategy for numerous solid cancers (1, 6, 9), and its systemic, blood–brain barrier–independent mode of action, motivated us to systematically examine the impact of ADT in various GBM models. By using conventional 2-D and advanced 3-D survival, growth, migration, and invasion assays combined with specially formulated media and an enzymotherapeutic approach with rhA, we could show for the first time that p53-/knockdown GBM cells are particularly sensitive to ADT and ADT-induced radiosensitization. Our data further reveal that this effect is unlikely to be impaired by an enhanced citrulline-to-arginine conversion, which is a prerequisite for selective in vivo effectiveness.

We found all GBM cell models to express the enzymes relevant for intrinsic arginine synthesis. However, the protein levels were quite variable and those that were most sensitive to ADT-induced growth arrest and reduced recovery in the 2-D assays were extremely poor in either ASS1 (U138-MG) or ASL (U251-MG) and did not upregulate these proteins in the absence of arginine. This effect might be due to epigenetic silencing via methylated CpG islands in the ASS1 or ASL gene, respectively, which was recently shown by Syed and colleagues (38) (i) to occur in 8/22 and 5/22 primary GBM cell cultures, (ii) to correlate with reduced median overall survival time, and (iii) to functionally alter the response to arginine deiminase (ADI) treatment. As described for other ASS/ASL-positive cancer cells (16–18), all GBM models studied herein showed reduced regrowth capacity after long-term ADT reflecting some loss of reproductive survival. The effect clearly attenuates in the presence of citrulline, which is in line with findings in other previously studied tumor entities (4, 7, 13, 32, 39). The loss of efficacy was most evident in the ASS/ASL^high-expressing models, indicating that ASS/ASL^low GBM cells might indeed be more sensitive to ADT due to a limited capability of intrinsic de novo arginine synthesis.

The observations are also relevant in the context of the two therapeutic enzymes in clinical trial, i.e., ADI and arginase. ADI degrades arginine via production of ammonia and citrulline, resulting in enhanced serum citrulline levels. It is unclear, however, if the increased serum citrulline detected after ADI treatment (40) leads to a critical augmentation of intratumoral citrulline levels which could reduce the efficacy upon long-term treatment. The primary product of arginase-dependent arginine degradation is ornithine. Ornithine is converted via intermediates to arginine only in cells with a highly active urea cycle, i.e., liver cells, and the respective negative loop can thus be ignored in GBM treatment. However, ornithine also serves as precursor for various aliphatic polyamines involved in multiple regulatory processes such as ion channel function, DNA stabilization, regulation of gene expression as well as cell proliferation, and might therefore adversely promote tumor growth and survival (41). In the present study, both arginine-free media and rhA were applied to achieve ADT in vitro with no evidence for a detrimental impact of catabolites on treatment outcome. Furthermore, as for citrulline, the increase in mean ornithine blood level may not be accompanied by intratumoral accumulation of ornithine, e.g., due to the locoregionally and spatiotemporally poor perfusion and blood supply. Yet, it might be speculated that patients with cancers showing high levels and activity of ornithine decarboxylase (ODC), the enzyme catalyzing the conversion of ornithine to putrescine as the first polyamine in the putrescine–spermidine–spermine axis might profit from combining ADT with an ODC inhibitor. In a previous clinical phase III trials, combination of the ODC inhibitor α-difluoromethylornithine (DFMO) with radiotherapy and/or various chemotherapeutics did not result in a survival benefit for newly diagnosed GBM patients even if their tumors overexpressed ODC (42, 43). The proposed multimetabolic targeting approach may thus only be considered in the case of critically enhanced intratumoral ornithine levels during ADT and if DFMO does not, as also indicated, critically interfere with the arginine-depriving enzymes.

In contrast to GBM cell growth and regrowth, radiosensitization upon ADT appears to be rather independent of ASS/ASL protein levels, and citrulline did not impair the synergistic effect sine qua non for in vivo effectiveness. Vice versa, it is a matter of debate whether high citrulline concentrations—if indeed not interfering with ADT-induced tumor cell radiosensitization—might at the same time protect normal cells. Citrulline and reduced citrulline plasma levels, respectively, are mainly considered as biomarker for intestinal failure and mucosal injury after radio-/chemotherapy (44–46). A recent study using a chemically induced murine mucositis model revealed that pretreatment with l-citrulline positively affected mucosal architecture and function of the small intestine (47). Also, orally administered l-citrulline was found to improve memory deficits by inhibiting neuronal death and cerebrovascular injury in a bilateral common carotid artery occlusion mouse model (48). The proposed mechanism of cerebrovascular protection was associated with increased eNOS expression and required the recycling of l-Arg via the ASS/ASL enzyme system followed by production of NO. These data indicate a protective potential of citrulline, an interesting phenomenon to be further assessed in the context of combinatorial treatment strategies with ADT.

Independent of any treatment, radioresistance of U87-MG-shp53 spheroids was found to be higher than of p53-wt U87-MG-shLuc spheroids at standard size (360–400 μm) and comparable cell numbers per spheroid. It is noted though that the U87-MG-shLuc spheroids were on average slightly smaller at the time of irradiation, i.e., the difference at absolutely identical size might be less pronounced. The discrepancy in radioresponse of U87-MG-shLuc and U87-MG-shp53 cells was not seen in our monolayer assays but is indeed in line with earlier studies using several GBM cell lines with distinct TP53 mutational status as well as U87-MG models with and without inactivation of the TP53 gene (49, 50). Differences in cell numbers during irradiation are of course an important aspect for data interpretation in the 3-D assay. This has been addressed in our experimental design by attempts to irradiate treated and untreated spheroids at similar size, i.e., by exposing the controls at different time points to radiation than ADT-deprived cultures. Although the mean number of viable cells/spheroid was not absolutely identical, differences were always <15% and cannot explain the highly significant effects seen in the SCP assays. From independent experiments using various head and neck squamous cell carcinoma, colorectal, and pancreatic cancer cell spheroids of different size ranges between 250 and 650 μm, we know that the cell number/spheroid has to change considerably to result in a significant change in SCD₅₀. An increase in cell number by factors of 5 and 7, respectively, resulted in an increase in SCD₅₀ of 6 and 7 Gy in two HNSCC models independent of their intrinsic radiosensitivity (unpublished data). A reduction in cell number by about 14% as seen in the U87-MG-shp53 spheroids after 5 days of ADT is therefore unlikely to translate into an SCD₅₀ change of >0.2 Gy.

An important outcome of our work is that TP53-mutated and -knockdown cells appear exceedingly responsive to the combined therapy design of ADT and irradiation. ADT-induced radiosensitization was proven in our sophisticated 3-D radioresponse assay which was particularly useful to also substantiate the clear, yet less pronounced, radiosensitizing effect of ADT in the p53-wt model. Conceivably, the particularly high efficiency of ADT in the p53-lacking cells might be linked to decreased p53-associated DNA repair or lack of p53-induced cell-cycle arrest as previously shown in colorectal cell models (26), leading to enhanced mitotic catastrophe. However, p53 also plays a regulatory role in various fundamental metabolic pathways. Among others and in contrast to its proapoptotic function during DNA-damage stress, p53 is known to support cell survival under metabolic stress conditions such as lack of glucose, glutamine, or serine (51–54). We also have first indications that functional p53 is relevant for metabolic stress response and survival during ADT in epithelial cancer cells (26).

Cumulative evidence suggests that caspase-dependent apoptotic pathways play a major role in cell death after ADT in different tumor cell types (55). However, tumor cells may evade apoptosis through the onset of autophagy, which is frequently described as an early consequence and a possible prosurvival mechanism after amino acid starvation. On the other hand, induction of extensive autophagy may lead to cell death through activation of stress-related pathways, mainly due to the accumulation of reactive oxygen species (55, 56). In colorectal cancer cells, arginine starvation was found to induce prolonged ER stress resulting in an upregulation of the AKT/MAPK pathway but not massive apoptosis (10). In the respective study, 2% DMSO significantly reduced the arginine deprivation–induced ER stress. DMSO is not a specific ER stress inhibitor, but clearly alleviates UPR pathways and correspondent ER stress gene activation. It is known to reduce protein aggregation and accumulation in the ER lumen and to trigger all UPR pathways (57). In contrast to DMSO, salubrinal is a specific inhibitor of eIF2α phosphatase enzymes and indirectly inhibits eIF2 as a result of reduced dephosphorylation. It thus enhances eIF2α phosphorylation and ATF4 expression, shuts down protein translation when the ER lumen is overloaded with unfolded protein, and reduces UPR activation (58, 59). Today, salubrinal is widely used experimentally to study ER stress response and has also shown earlier to inhibit drug-related ER stress–induced apoptosis in neuroblastoma cells (60, 61) and T98G GBM cells (62) at concentrations of 10 to 40 μmol/L. Nonetheless, our initial attempt to causally link ER stress response to ADT-induced radiosensitization by simultaneous application of either DMSO or salubrinal remained inconclusive.

About one-third of GBMs show mutations or deletions of TP53, and these cancers appear to be more resistant to the current standard-of-care chemotherapeutic agent temozolomide (35, 63, 64). Although the causal relation and underlying mechanistic link for GBM still needs to be unraveled, the observation of ADT-induced radiosensitization is of great clinical interest, in particular in GBM cells showing p53 loss of function. Mechanistic studies and transfer into appropriate GBM in vivo models are thus envisioned to further explore the potential and challenges of ADT in combination with radiotherapy and standard-of-care therapeutics to allow rapid implementation of this promising approach into a clinical trial.

No potential conflicts of interest were disclosed.

Conception and design: C.N. Hinrichs, M. Ingargiola, O. Stasyk, L.A. Kunz-Schughart

Development of methodology: C.N. Hinrichs, M. Ingargiola, A. Temme, O. Stasyk, L.A. Kunz-Schughart

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.N. Hinrichs, M. Ingargiola, T. Käubler, O. Vovk, L.A. Kunz-Schughart

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.N. Hinrichs, M. Ingargiola, T. Käubler, S. Löck, A. Köhn-Luque, L.A. Kunz-Schughart

Writing, review, and/or revision of the manuscript: C.N. Hinrichs, M. Ingargiola, S. Löck, A. Temme, A. Deutsch, L.A. Kunz-Schughart

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.N. Hinrichs, M. Ingargiola, A. Temme, L.A. Kunz-Schughart

Study supervision: M. Ingargiola, O. Stasyk, L.A. Kunz-Schughart

This work was supported by the European Social Fund and the Free State of Saxony (GlioMath DD 100098214 to L.A. Kunz-Schughart, A. Deutsch, and A. Temme), and the Else Kröner-Fresenius Foundation (to C.N. Hinrichs).

We thank Melanie Hüther, Marit Wondrak, and Alexander Krüger for excellent technical assistance and gratefully acknowledge Drs. Yaroslav Bobak, Yuliya Kurlishchuk, Oleg Chen, Barbara Klink, and Ralf Wiedemuth for constructive discussions and methodological support.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.