N-Myc-induced metabolic rewiring creates novel therapeutic vulnerabilities in neuroblastoma

N-Myc is a transcription factor that is aberrantly expressed in many tumor types and is often correlated with poor patient prognosis. Recently, several lines of evidence pointed to the fact that oncogenic activation of Myc family proteins is concomitant with reprogramming of tumor cells to cope with an enhanced need for metabolites during cell growth. These adaptions are driven by the ability of Myc proteins to act as transcriptional amplifiers in a tissue-of-origin specific manner. Here, we describe the effects of N-Myc overexpression on metabolic reprogramming in neuroblastoma cells. Ectopic expression of N-Myc induced a glycolytic switch that was concomitant with enhanced sensitivity towards 2-deoxyglucose, an inhibitor of glycolysis. Moreover, global metabolic profiling revealed extensive alterations in the cellular metabolome resulting from overexpression of N-Myc. Limited supply with either of the two main carbon sources, glucose or glutamine, resulted in distinct shifts in steady-state metabolite levels and significant changes in glutathione metabolism. Interestingly, interference with glutamine-glutamate conversion preferentially blocked proliferation of N-Myc overexpressing cells, when glutamine levels were reduced. Thus, our study uncovered N-Myc induction and nutrient levels as important metabolic master switches in neuroblastoma cells and identified critical nodes that restrict tumor cell proliferation.

a better understanding of cell autonomous and cell non-autonomous regulatory circuits governed by oncogenic N-Myc functions, insights into mechanistic effects on the level of metabolic circuits is still largely lacking.
Deregulated Myc activity comes along with enhanced metabolic stress and increased sensitivity towards apoptosis due to a dependency on continuous supply with nutrients. Glutamine has been identified as a limiting factor for Myc-dependent cell growth and glutamine deprivation was preferentially inducing apoptosis in Myc-high cells 6 . In neuroblastoma, the most common solid tumor of childhood, elevated N-Myc levels are often found due to amplification of the coding gene, MYCN, which is correlated with poor overall survival of affected patients. N-Myc overexpression in the absence of MYCN amplification is not prognostic, pointing to the importance of additional genetic factors such as telomerase maintenance for determining disease outcome 7 . However, forced expression of N-Myc is sufficient to induce neuroblastoma in different model organisms including mice 8,9 and zebrafish 10,11 indicating a causative role for N-Myc expression in disease onset and maintenance. Ectopic N-Myc expression in neuroblastoma cells is accompanied with increased aggressiveness, but also a higher sensitivity towards drug-induced apoptosis in vivo and in vitro 12,13 . Interestingly, glutamine deprivation was reported to preferentially sensitize MYCN-amplified cells towards programmed cell death 14 . N-Myc also promoted expression of genes involved in glycolysis including lactate dehydrogenase (LDH) and hexokinase 2 (HK2) in neuroblastoma in vitro 15 and in vivo 16 . The notion that depletion of LDHA inhibited MYCN-dependent tumorigenesis 17 and that N-Myc cooperated with HIF1α 15 suggested that N-Myc expression could also contribute to the Warburg effect. Thus, tumor-specific Myc deregulation directly affects tumor metabolism in multiple pathways.
At present, it is not clear how metabolic dependencies of tumors can be best exploited for therapeutic purposes. It is evident that metabolic features are remarkably flexible, especially those driven by Myc proteins. Moreover, tumor-type specific metabolic adaptations have been identified in Myc-driven tumors, exemplified by the Myc-mediated de novo synthesis of glutamine 18 . By contrast, Myc-driven liver tumors rather consume glutamine by a process termed glutaminolysis, which allows for fueling into the tricarboxylic acid cycle (TCA cycle) at the level of α-ketoglutarate by activation of glutaminase, another Myc-target 19 . MYCN-driven glutaminolysis has been recently suggested as a strategy to treat Myc-driven cancers 20 . Still, tumor cell-intrinsic response patterns as a consequence of MYCN activation under varying nutrient conditions largely remain to be identified. We thus set out to profile metabolic shifts in neuroblastoma cell lines with inducible N-Myc expression and correlate their phenotypic responses upon variations in the two most common carbon sources, glucose and glutamine.

Materials and methods
Cell culture and reagents. Neuroblastoma cell lines SHEP, SH-SY5Y, SK-N-AS and SK-N-SH were cultivated in RPMI1640 medium containing 10% fetal bovine serum (FBS) and antibiotics as described [21][22][23] . Protocols for generating inducible expression of a gene of interest have been described before 24 . In brief, cell lines were sequentially transfected with pcDNA6/TR, harboring the tetracycline repressor gene, and pT-REx-DEST30 (ThermoFisher/ Invitrogen) containing MYCN cDNA. Single cell clones were selected by limiting dilution in medium containing blasticidine and G418 (ThermoFisher/ Invitrogen). For all cell lines transfected to express N-Myc upon addition of tetracycline, the suffix "TR-MYCN" was added to distinguish them from the parental cells. N-Myc induction was realized by adding 1 μg tetracycline per ml medium. Cell lines were authenticated by STR genotyping prior and post transfections. All reagents used for cell culture were obtained from Gibco/ ThermoFisher. Absence of Mycoplasma sp. in cultivated cells was confirmed by performing PCR with Mycoplasma-specific primers (IDT).
Real-time reverse transcriptase-PCR. Total RNA was isolated from cells using the High Pure RNA Isolation Kit (Roche, #11828665001), and cDNA synthesis was performed using the Transcriptor First Strand cDNA Synthesis Kit (Roche, #04379012001). Expression of target genes was monitored using a StepOnePlus ™ Real-Time PCR System (Applied Biosystems) and Fast SYBR ® Green Master Mix (Roche, #04913914001).
Expression values were normalized to β-actin expression and the fold change was calculated using the ΔΔ − Ct method.
Glycolysis flux assays. Glycolysis flux assays were performed in a Seahorse 96-well XF Cell Culture Microplate using a XF96 sensor cartridge according to the manufacturer's instructions (Seahorse Bioscience, #102416-100). Briefly, 20,000 cells per well were allowed to grow overnight in a 37 °C, 5% CO2 incubator in pyruvate-free RPMI1640 (Gibco/ThermoFisher Scientific, #21875-034, #31870-025, #11879-020) with L-glutamine (Gibco/ThermoFisher Scientific, #25030-024) and varying concentrations of D-glucose (Gibco/ ThermoFisher Scientific, #A2494001), 10% fetal bovine serum (Gibco/ThermoFisher Scientific #10270-106) and antibiotics. On the next day, medium was replaced with pyruvate-and glucose-free Seahorse XF Base Medium (#102353-100) supplemented with 2 mM glutamine according to the manufacturer's protocol. The microplates were incubated at 37 °C for 45 min and then transferred to the microplate stage of a Seahorse XFe96 Extracellular Flux Analyzer to quantify the extracellular acidification rate (ECAR, in [mpH/min]) and the oxygen consumption rate (OCR, in [pmol/min]) using glucose concentrations as indicated. For each step, three separate ECAR and OCR readings using a mix and measure cycling protocol (3 min each) were recorded. Data for each experimental condition were calculated from a minimum of 10 technical replicates.
Metabolic profiling. SHEP-TR-MYCN cells with and without induction of MYCN were incubated under varying glucose or glutamine concentrations. Upon harvesting, samples were prepared using the automated MicroLab STAR ® system (Hamilton). To recover chemically diverse metabolites, proteins were precipitated with methanol under vigorous shaking for 2 min (Glen Mills GenoGrinder 2000) followed by centrifugation. The resulting extract was analyzed either by separate reverse phase (RP)/UPLC-MS/MS with positive ion mode electrospray ionization (ESI), RP/UPLC-MS/MS with negative ion mode ESI or HILIC/UPLC-MS/MS with negative ion mode ESI. The sample extracts were stored overnight under nitrogen before preparation for analysis. All methods utilized a Waters ACQUITY ultra-performance liquid chromatography (UPLC) and a Thermo Scientific Q-Exactive high resolution/accurate mass spectrometer interfaced with a heated electrospray ionization (HESI-II) source and Orbitrap mass analyzer operated at 35,000 mass resolution. Raw data was extracted, peak-identified and QC processed using hardware and software developed by Metabolon Inc..

Statistical analysis.
Statistical analyses of metabolic profiling data were performed using R, version 3.3.2 (available at cran.r-project.org). Prior to analyses, samples were normalized according to protein content determined by Coomassie G-250 based assays. For detailed procedures see Supplemental Methods. If not indicated otherwise, data are presented as mean and standard error of the mean (SEM).

Metabolic reprogramming of SHEP cells depends on carbon source supply and N-Myc expression.
We had previously described a correlation of MYCN with HK2 mRNA expression in SHEP human neuroblastoma cells engineered to stably express N-Myc or to activate N-Myc only in the presence of tamoxifen 27 . To better understand the cellular and the metabolic response towards N-Myc induction and varying glucose / glutamine ratios, we here used SHEP cells with tightly controlled conditional upregulation of N-Myc in the presence of tetracycline (SHEP-TR-MYCN). Expression analyses confirmed both, N-Myc induction and concomitant hexokinase II protein (HKII) upregulation in these cells (Fig. 1A). SHEP-TR-MYCN cells were then cultivated with or without N-Myc induction and varying glucose and glutamine levels. Of note, N-Myc and HKII induction appeared to be independent of glucose and glutamine concentrations in the medium (Fig. 1A). We defined ten experimental groups that differed in N-Myc expression levels and nutrient supply (Table 1) 1B). We performed two condition-specific analyses of variance (ANOVAs) for each metabolite, one for constant glutamine conditions to examine the effect of glucose variation, and one for constant glucose conditions to examine the effect of glutamine variation. For each condition, we only analyzed metabolites which were present in at least one third of samples considered ( Table 1). Induction of N-Myc was associated with consistent metabolic shifts, as 144/496 (29%, for constant glutamine) or 365/486 (75%, for constant glucose) of the metabolites were significantly altered (FDR < 0.05) between N-Myc high and N-Myc low expressing cells (Fig. 1C,D; Supplemental Figs. 1 and 2). A comparable number of metabolites were significantly affected when glucose (199/496, 40%) or glutamine levels (352/486, 73%) were varied. We then examined a possible interaction effect between N-Myc expression levels and glucose or glutamine supply. ANOVAs revealed a significant interaction between varying N-Myc expression and glucose concentrations affecting 24% of all metabolites (120/496 metabolites, FDR < 0.05) when only samples with constant glutamine concentrations were considered (Fig. 1C). An even higher number of metabolites (240/486, FDR < 0.05) had a significant interaction effect between N-Myc and glutamine when analyzing samples that had constant glucose concentrations (Fig. 1D). Classification of these metabolites according to "super pathways" 28 was performed to uncover patterns according to metabolic categories. Here, metabolites classified in the category "energy metabolism" had the highest relative enrichment (Supplemental Figs. 1 and 2). Thus, global profiling suggested tight interaction of N-Myc and carbon source levels in metabolic reprogramming of SHEP cells, while HKII was confirmed as a N-Myc target independent of carbon source availability.

N-Myc shifts the balance from oxidative phosphorylation to glycolysis. To examine if
up-regulation of pacemaker enzymes of glycolysis including HKII were concomitant with a switch in energy harvest, we tested different scenarios in SHEP cells with varying N-Myc levels and glucose supply. In all conditions analyzed, we observed a shift towards preferential use of glycolysis upon MYCN induction as seen in higher ECAR levels ( Fig. 2A). This effect was independent of glucose concentration ( Fig. 2A) and cell number (Fig. 2B). Interestingly, N-Myc-induced upregulation of HKII did not result in altered steady state HK enzymatic activity. N-Myc low and N-Myc high cells also had comparable activity of other glycolytic proteins, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), lactate dehydrogenase (LDH), triosephosphate isomerase (2020) 10:7157 | https://doi.org/10.1038/s41598-020-64040-1 www.nature.com/scientificreports www.nature.com/scientificreports/ (TIM) and pyruvate kinase (PK, Supplemental Fig. 3). Of note, we could confirm that c-Myc levels decrease upon N-Myc induction (Supplemental Fig. 3). Moreover, subcellular fractionation revealed no difference in HK activity for N-Myc low and N-Myc high cells in different cellular compartments (Supplemental Fig. 4). While siRNA-mediated knockdown of either HKI or HKII was tolerated regardless of N-Myc expression levels (Supplemental Fig. 5), blocking of glycolysis downstream of HK by using 2-deoxyglucose (2-DG) preferentially affected SHEP cells with high N-Myc levels (Fig. 2C). To check whether this effect could be recapitulated in other neuroblastoma cells, we ectopically expressed inducible N-Myc in three additional human neuroblastoma cell lines without MYCN gene amplification, SK-N-AS, SK-N-SH and SY5Y. Again, induction of N-Myc was accompanied by upregulation of HKII (Fig. 2D). N-Myc-mediated sensitization towards 2-DG was also observed for SK-N-AS cells, while SY5Y cells were intrinsically more sensitive to 2-DG regardless of MYCN expression levels (Supplemental Fig. 6). Thus, N-Myc overexpression induced a shift to preferential usage of glycolysis and also increased sensitivity towards inhibition of the glycolytic pathway in a cell line specific manner.
N-Myc induces cell-type specific responses upon interference with different steps of glutamine biosynthesis. Cellular reprogramming by Myc proteins has been shown to affect glutamine metabolism, consistent with an increased demand of tumor cells for glutamine and metabolic adaptions to fluctuations in nutrient supply (reviewed in 29 ). As for glucose, we observed a high number of significantly altered metabolites depending on MYCN expression levels when glutamine supply was varied. As indicated above, bifactorial ANOVAs revealed that 49% of all metabolites displayed a significant interaction effect between N-Myc induction and glutamine levels suggesting that these factors are intricately linked to metabolic reprogramming. As Myc proteins upregulate both, the import of glutamine and mitochondrial glutaminolysis, we aimed to investigate the cellular response to inhibition of glutamine metabolism upon MYCN induction. Blocking glutamine uptake by L-γ-Glutamyl-p-nitroanilide (GPNA) did not affect cell viability regardless of N-Myc expression levels (Fig. 3A). Paradoxically, inhibition of glutaminase by a small molecule inhibitor, CB-839, increased relative cell viability in SK-N-SH TR-MYCN and SHEP-TR-MYCN cells upon MYCN induction when glutamine supply was limited. Low and high levels of glucose and glutamine, respectively, were used as defined in Table 1. (C) Numbers of metabolites significantly affected by N-Myc, glucose abundance or with significant interaction effect according to bifactorial ANOVAs using sample groups with constant glutamine conditions and varying glucose levels (cf. Table 1). (D) Numbers of metabolites significantly affected by N-Myc, glutamine abundance or with significant interaction effect according to bifactorial ANOVAs using the sample groups with constant glucose conditions and varying glutamine levels (cf. Table 1 N-Myc affects the entire pathway of glutathione biosynthesis, but exogenous glutathione does not restore viability of N-Myc high cells. In our analysis of metabolome data, classification of known metabolites according to "super pathways" 28 was also performed to uncover MYCN-induced patterns according to metabolic categories under varying glutamine concentrations. Significant overrepresentation of metabolites upon MYCN induction under these conditions was observed only for the group of "peptides" (Supplemental Fig. 9). Here, 26/26 di-peptides were significantly altered when cells with different N-Myc levels were compared. Detailed analyses of this group revealed that 12 of the di-peptides contained gamma-glutamyl residues and were linked to glutathione metabolism. Additional metabolites annotated as members of "glutathione metabolism" 30 were also significantly altered by N-Myc levels independent of glutamine supply (32/34 metabolites, Fig. 4A); glutathione-associated metabolites were significantly over-represented among the MYCN-affected metabolites (p-value 0.003, Fisher's exact test). Strikingly, almost all metabolites involved in glutathione metabolism, including glutamine, were found at significantly lower levels in N-Myc high cells compared to N-Myc low cells regardless of glutamine supply (Fig. 4B). While this suggested a limiting role of glutathione availability for N-Myc high cells, DON-mediated reduced viability of SHEP-TR-MYCN cells could not be rescued by exogenous glutathione supply (Fig. 4C). Additionally, increasing reactive oxygen species (ROS) by mono-or dimethyl fumarate (MMF, DMF) did also not affect cell viability upon MYCN induction and varying glutamine supply (Fig. 4D, Supplemental Fig. 10). These results indicate that MYCN induction causes a significant depletion of glutamine-related metabolites including glutathione, and that the resulting glutamine addiction cannot be attributed to modulation of ROS levels.

Discussion
The wealth of data on plasticity of tumor cells suggests that blocking of single metabolic pathways may not be sufficient for long-term control of cancer growth. However, it is crucial to analyze oncogene-induced adaptations of metabolic pathways on a cellular level to understand mechanisms of metabolic plasticity contributing to development of therapy resistance in cancer. Still, tumor cell-centric assays are limited as they will miss interactions with the tumor microenvironment. A prime example is the notion that KRAS driven lung tumor cells are   www.nature.com/scientificreports www.nature.com/scientificreports/ glutamine dependent in vitro, as they need glutamine-derived α-ketoglutarate to enter the TCA cycle, but are glutamine-independent in vivo due to fueling of the TCA cycle by glucose 31 . Expression of glutamine synthetase (GS) by cells of the tumor microenvironment might explain compensatory effects in cancer cells with reduced GS activity 32 . It has also been argued that routinely applied cell culture conditions will not appropriately mimic metabolic conditions found in human plasma 33 . Thus, experiments to model tumor cell responses to oncogenic activation in vitro have to be carefully designed. We here investigated the effects of ectopic MYCN expression in neuroblastoma cells lines in the presence of high or low supply with glucose or glutamine, respectively, as an approach to mimic both, oncogene-induced changes in metabolic patterns and response to altered carbon source supply. For metabolic profiling of these conditions, we chose SHEP neuroblastoma cells, in which we had previously noticed MYCN-induced HK2 upregulation indicating metabolic reprogramming 16,27 . Using a newly adapted system for inducible MYCN expression, designated SHEP-TR-MYCN, we were able to demonstrate that HKII induction by N-Myc was independent of carbon source supply (Fig. 1A). Moreover, N-Myc dependent upregulation of HKII could be confirmed in other neuroblastoma cell lines (Fig. 2D). Metabolome analyses revealed that both, carbon source and N-Myc expression levels, cause significant and specific changes allowing for separation of these experimental groups by principal component analyses (Fig. 1B). Of note, we also observed significant interaction effects between N-Myc and carbon source levels (Fig. 1C,D) suggesting that N-Myc induced reprogramming also affects downstream metabolic processes under varying nutrient conditions. While our analyses uncovered that N-Myc mediated regulation of metabolites affects all pathways investigated, classification of N-Myc regulated metabolites confirmed enrichment for metabolites within the categories "energy metabolism" and a highly significant enrichment for metabolites involved in the glutathione pathway. We thus focused on analyses of two main pathways for energy harvest, glycolysis and glutamine metabolism, as the latter feeds also into glutathione biosynthesis. Indeed, we found a shift towards preferential use of glycolysis in MYCN-induced cells and an increased sensitivity towards glycolysis inhibition (Fig. 2). While this paper was in review, this was confirmed by results from Oliynyk et al. who showed that N-Myc increases the glycolytic and oxidative phosphorylation capacity of MYCN-amplified neuroblastoma cells. Moreover, they could demonstrate glutamine independence of mitochondrial respiration in MYCN-amplified neuroblastoma cells compared with non-MYCN amplified cells and that N-Myc high cells are especially vulnerable to inhibition of fatty acid oxidation 34 . We can conclude from these results that N-Myc reprograms neuroblastoma cells towards a highly energy producing and consuming phenotype. Glycolysis addiction in cancer cells with high levels of Myc has been described and linked to mTOR pathway activation 35 . Interestingly, we and others previously noticed an increased sensitivity of MYCN-dependent neuroblastoma towards mTOR inhibition 27,36 , suggesting that simultaneous blockade of glycolysis and mTOR would be an option to interfere with both, metabolism and survival pathways.
The addiction of MYCN-amplified neuroblastoma cells to glutamine availability has been described 14 . A recent report indicated higher glutamate/glutamine ratios in neuroblastoma compared to other cell types 37 . These findings suggested that neuroblastoma can be categorized as a "glutamine-consuming" rather than a "glutamine-producing" tumor type. Of note, c-Myc regulates glutamine metabolism, at least in part, by repressing miR-23a and miR-23b resulting in enhanced expression of their target gene, glutaminase 19 . Recent reports pointed to a possible feedback mechanism, as MYC translation itself is regulated by intracellular levels of glutamine-derived adenosine nucleotides via a sequence element in the 3′-UTR of the MYC mRNA 6 . In our model cell lines, this effect cannot be analyzed in the absence of endogenous regulator elements in the 3′UTR of the MYCN encoding vectors. The emerging picture, however, is far from clear: blocking glutamine uptake and its metabolic conversion to α-ketoglutarate was well tolerated in all conditions analyzed, indicating compensatory mechanisms. Paradoxical effects were observed for CB-839, an inhibitor of mitochondrial glutaminase (GLS1) that is currently tested in early clinical trials for treatment of several solid tumors multiple myeloma and acute myeloid leukemia 38 : while two of four neuroblastoma cell lines did not respond to CB-839, N-Myc expression combined with glutamine starvation enhanced metabolic activity of SK-N-SH and SHEP cells, however, this was not translating in increased cell proliferation (Supplemental Fig. 7). Viability of the www.nature.com/scientificreports www.nature.com/scientificreports/ latter two cell lines was not affected by DON, an irreversible glutamine-competitive inhibitor, at concentrations of up to 10 µM. However, limiting glutamine concentrations increased sensitivity towards DON in SY5Y and SKNAS cells, while N-Myc expression enhanced DON sensitivity 10-fold in SK-N-SH cells. Dissection of cellular responses on the level of the entire metabolome in SHEP cells revealed significant changes in glutathione-related metabolites upon variation of glutamine availability, which affected the vast majority of intermediates on the route from glutamine to GSH (Fig. 4A,B). We thus hypothesized that glutamine addiction of N-Myc high cells could be attributed to an increased sensitivity towards ROS production. Increased demand for ROS detoxification has been recognized as a vulnerability of cancer cells in the past years. The concept has also helped to understand the mechanisms of action for cytotoxic drugs including As 2 O 3 and anthracyclines, which are effective at least in part by glutathione oxidation and increased ROS production, respectively 39 . However, sensitivity of N-Myc high cells to inhibition of glutamine metabolism could not be rescued by external glutathione supply. Moreover, N-Myc high cells were not significantly more sensitive to the ROS-inducing agent, dimethyl fumarate, which was described to suppress neuroblastoma cell proliferation 20 . Thus, preferential killing of N-Myc high cells in the presence of glutamine metabolism inhibitors and limited glutamine supply cannot be explained by increased consumption of glutathione and increased ROS sensitivity.
As mentioned above, strategies aiming to limit carbon source availability of inhibition of primary metabolic pathways have to take into account not only plasticity of tumor cells themselves but also that other cell types in the tumor microenvironment could serve as a carbon source in vivo 32 . Our data consistently show that responses to interference with primary metabolism are cell-type specific and that these cell-type specific differences largely contribute to metabolic plasticity reflecting inherent heterogeneity of our cell models. However, our results confirm interference with glucose and glutamine availability are possible avenues for exploiting metabolic dependencies of N-Myc high cells, which have to be further explored in in vivo models.