Myc-Driven Glycolysis Is a Therapeutic Target in Glioblastoma

The MYC gene family (MYC, MYCN, and MYCL) consists of potent oncogenes that play critical roles in the pathogenesis of diverse types of human cancers (1, 2). The Myc oncoprotein contributes to malignancy by various mechanisms, including driving cellular proliferation, blocking differentiation, facilitating chromosomal instability, increasing cell migration, and inducing angiogenesis (1, 2). In glioblastoma, Myc enhances the self-renewal capacity of glioma stem-like/tumor neurosphere (tumorsphere) cells and maintains their tumorigenic potential (3–6). High-level amplification of the MYC and MYCN genes are observed in a subset of glioblastoma (7–11). Myc is therefore a compelling therapeutic target in glioblastoma.

Despite extensive efforts, direct inhibition of the Myc transcription factor has remained a challenge. Several indirect strategies that selectively target the pleiotropic Myc-driven downstream effects have recently shown promise, including small-molecule inhibitors of BET chromatin adapters (12, 13), Chk1 (14), and CDK7 (15). Another potential Myc-specific target is the Myc-reprogrammed metabolic state, which is evident in glioblastoma (16–18). Deregulated Myc has been shown to increase glycolysis and glutaminolysis to support the increased biosynthetic demand of rapidly proliferating cancer cells, and the altered cell metabolism may render Myc-driven cancers vulnerable to strategic nutrient deprivation (16, 19).

Here, we tested whether inhibition of the Myc-induced glycolytic drive would be a selective strategy for Myc-driven glioblastoma. We confirmed that Myc increases glycolytic flux in glioblastoma cells and found that Myc generates a dependency on glycolysis for survival. Using a panel of patient-derived glioblastoma tumorsphere lines (20, 21), we found glycolytic inhibition to be strikingly selective for lines with highly amplified MYC/MYCN. We identified small-molecule inhibitors of the NAD⁺ salvage enzyme nicotinamide phosphoribosyl-transferase (NAMPT), which mediates cytotoxicity by inhibiting glycolysis (22), as a therapeutic strategy with potential for rapid translation in a genetically defined subset of glioblastoma patients.

To create glioblastoma tumorsphere lines, fresh patient glioblastoma surgical specimens were dissociated and cultured in serum-free neural stem cell medium, which enriches for tumor-initiating cells (20, 21). All tumorsphere lines were cryopreserved at passage number 3 or less prior to use for in vitro experiments. For genetic manipulation, lines passaged <10 times were used. All tumor samples were collected with patient consent under protocols approved by the Massachusetts General Hospital (MGH, Boston, MA) Institutional Review Board. All tumors were confirmed to be glioblastoma by formal pathologic review. U87, H1975, D283 Med, IMR-32, Daoy, and Raji cell lines were obtained from ATCC and were cultured using ATCC-formulated Eagle minimum essential medium (for IMR-32, D283 Med, and Daoy), ATCC-formulated RPMI-1640 (for H1975 and Raji), or DMEM (for U87)_. Kelly was obtained from Sigma-Aldrich and cultured with EMEM. UACC257 was a gift from David E. Fisher (MGH, Boston, MA) and cultured with DMEM. Normal human astrocytes (NHA) were obtained from ScienCell and cultured in DMEM. All standard cell line media were supplemented with 10% FBS and penicillin–streptomycin–amphotericin B. FK866, nicotinamide mononucleotide (NMN), nicotinic acid (NA), NAD⁺, and 2-deoxyglucose (2-DG) were purchased from Sigma-Aldrich. GMX1778 was purchased from Cayman Chemical.

293T cells were cotransfected with lentivirus vectors containing MYC (pCDH-puro-cMYC, Plasmid #46970; Addgene) or GFP (pCDH-CMV-MCS-EF1-copGFP, CD511B-1; System Bioscience), pCMV-dR8.2 dvpr, and pCMV-VSVG with ScreenFect (Wako). U87 cells were infected with lentivirus with polybrene (8 μg/mL) for 6 hours, and then selected with puromycin (0.7 μg/mL) for 7 days, then maintained in puromycin (0.2 μg/mL).

To knockdown MYC, MGG4 and H1975 cells were infected with GIPZ lentiviral human MYC shRNA (V2LHS_152053; GE Dharmacon) in polybrene (8 μg/mL) for 8 hours. Two weeks later, cells were selected with puromycin (0.3 μg/mL) for 5 days. To knockdown MYCN, IMR-32 cells were infected with GIPZ lentiviral human MYCN shRNA (V2LHS_36751; GE Dharmacon) in polybrene (8 μg/mL) for 8 hours. Three weeks later, cells were selected with puromycin (0.3 μg/mL) for 7 days. GIPZ nonsilencing lentiviral shRNA control (RHS4348; GE Dharmacon) was used as control.

Gene amplification status of MYC and MYCN was evaluated by FISH. BAC clones CTD-3066D1 and RP11-480N14 were used to make the MYC and MYCN probes, respectively, and BAC clones RP11-301H15 (Chr8) and RP11-984I21 (Chr2) were used for centromere controls. The probes were labeled in Cy3-dCTP or FITC-dUTP. Gene-amplified cells were counted in at least three different high-power fields, and the proportion of amplification-positive per total cells was calculated. Gene/control probe copy number ratios of >2.0 were considered amplified.

Cells were lysed in radioimmunoprecipitation buffer (Boston Bioproducts) with a cocktail of protease and phosphatase inhibitors (Roche). 12.5 μg of protein was separated by 4% to 20% SDS-PAGE and transferred to polyvinylidene difluoride membranes by electroblotting. After blocking with 5% nonfat dry milk in TBST [20 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, 0.1% Tween20], membranes were incubated at 4°C overnight with primary antibodies. After washing and incubation with horseradish peroxidase–conjugated secondary antibodies (Cell Signaling Technology), blots were washed, and signals were visualized with an ECL Kit (Amersham Biosciences). Primary antibodies that were used are as follows: c-Myc, N-Myc, hexokinase II (HKII), PKM2, LDHA, cleaved PARP, phospho-AMPK, phospho-Raptor, phospho-mTOR, phospho-S6, phospho-4E-BP1, actin (Cell Signaling Technology); NAMPT (Bethyl Laboratories); Naprt1 (Sigma-Aldrich); and Vinculin (Thermo Scientific).

To assess cell viability, glioblastoma tumorspheres were dissociated into single cells and seeded into 96-well plates at 7,000 to 8,000 cells/well. After 12 to 24 hours, chemical inhibitors and/or NMN, NAD⁺, or NA were serially diluted and added to wells. Cell viability was measured by CellTiter-Glo (Promega) assays and the IC₅₀ values (drug concentrations causing 50% viability of cells) were determined. To determine cytotoxicity, 7,000 to 8,000 cells were treated with DMSO, 12.5 nmol/L of FK866 or GMX1778, and the number of viable cells that excluded Trypan blue was counted using TE2000-U inverted microscope (Nikon) at 48, 72, and 96 hours after treatment.

Glioblastoma tumorspheres were dissociated into single cells and 1 × 10⁵ cells were seeded per well (12-well plates) and 24 hours later treated with NAMPT inhibitors. After 96-hour incubation, cells were collected and stained with propidium iodide (PI) and allophycocyanin (APC)-conjugated Annexin V (Annexin V Apoptosis Detection Kit APC; eBioscience), and analyzed by an Accuri flow cytometer and the BD CSampler software (BD Biosciences). To evaluate caspase-3/7 activities, cells were treated with DMSO or inhibitors (12.5 nmol/L) for 24 hours and were tested by Caspase-Glo 3/7 Assay (Promega) according to the manufacturer's recommendations.

Total NAD (NADt: NADH and NAD⁺) and NADH were measured using NAD⁺/NADH Quatification Colorimetric Kit (BioVision Incorporated) following the manufacturer's recommendations. NAD⁺ was calculated by subtracting NADH from NADt. Data were expressed as pmol/1 × 10⁶ cells or normalized to DMSO control and presented as percent change. To evaluate NAD⁺ change in tumor tissue, H1975 cells were implanted into the flank of SCID mice. Three weeks after implantation, mice were randomized to GMX1778 (250 mg/kg, n = 3) or vehicle (n = 3). Forty-eight hours after treatment, mice were sacrificed, tumor tissues were harvested, and NAD⁺ concentration was measured. Protein concentrations were measured and used for normalization. To evaluate qualitative values of NAD⁺, the NAD/NADH-Glo Assay (Promega) was used as described in the Supplementary Methods.

To measure dihydroxyacetone phosphate (DHAP), Pico probe Dihydroxyacetone Phosphate Fluorometric Assay Kit (BioVision Incorporated) was used according to manufacturer's recommendations. Cells were treated with DMSO, 12.5 nmol/L of FK866 or GMX1778 for 48 hours, and washed with cold PBS. A total of 3 × 10⁵ cells were suspended with DHAP Assay Buffer and mixed with PicoProbe, DHAP Enzyme Mix, and DHAP Developer. DHAP signal was measured by fluorescence (BIOTEK). Data were compared with DMSO-treated cells and expressed as percent change.

Steady-state labeling of glycolytic intermediates was accomplished by culturing cells (U87-Myc; 2.5 × 10⁶, U87-GFP 4.5 × 10⁶ cells/sample) in DMEM-no glucose medium (Gibco) containing 25 mmol/L of d-glucose (U-¹³C₆, 99%; Cambridge Isotope Laboratories) for 6 hours. For evaluation of MGG4 with/without NAMPT inhibitor, 4.5 × 10⁶ cells/sample were treated with DMSO or FK866 (50 nmol/L) for 42 hours, washed, and incubated in stem cell medium based on glucose-free Neurobasal medium (Gibco) supplemented with 25 mmol/L of d-glucose (U-¹³C₆, 99%) and DMSO or FK866 (50 nmol/L) for 6 hours. All treatments were conducted in triplicate. Intracellular metabolites were then extracted and measured by capillary electrophoresis time-of-flight mass spectrometry (CE-TOFMS) by Human Metabolome Technology as described in the Supplementary Methods (23–26). For the U87 assay, the amounts of each metabolite labeled at specific carbons were measured and C₃- or C₆-labeled carbons were used for analysis. For the MGG4 assay, the total amounts of each metabolite, both labeled and unlabeled, were measured and used for analysis. Data were expressed as pmol/1 × 10⁶ cells.

For glioblastoma orthotopic xenograft survival experiments, 1 × 10⁵ MGG8 or 2.5 × 10⁵ U87 cells were implanted into the right striatum of SCID mice as described previously (21). After 5 days, animals were randomized to treatments with weekly oral gavage with GMX1778 (250 mg/kg; n = 7) or vehicle [20% Captisol (CyDex) and 5% dextrose in water; n = 7]. Mice were monitored daily and body weight was measured two to three times a week. Mice were sacrificed when neurologic deficits or general conditions reached the criteria for euthanasia. For H1975 xenograft experiments, 2.5 × 10⁶ cells were subcutaneously implanted into the right flank of 7- to 10-week-old female SCID mice. When maximum tumor diameter reached 5 mm, cohorts were randomized to weekly GMX1778 (250 mg/kg; n = 7) or vehicle (n = 7) by oral gavage. Tumor diameters were measured three times a week using a digital caliper. Calculated volume (mm³) = length (mm) × width (mm)² × 0.5. All mouse experiments were approved by the Institutional Animal Care and Use Committee at MGH (Boston, MA).

Tumor tissue specimens (n = 3 for each group, drug, or vehicle) were fixed in 10% neutral-buffered formalin and embedded in paraffin. Hematoxylin and eosin staining was performed using the standard procedures. For IHC, 5-μm thick sections were deparaffinized, treated with 0.5% H₂O₂ in methanol, rehydrated, and heated for 25 minutes for antigen retrieval. Sections were blocked with serum, and incubated with anti-Ki-67 (1:125; Wako) primary antibody at 4°C overnight. The next day, sections were washed in PBS, incubated with biotinylated secondary antibody for 60 minutes at room temperature, and then incubated with ABC solution (PK6200; Vector Laboratories) for 30 minutes at room temperature. Finally, sections were stained with DAB (Dako) and counterstained with hematoxylin. To detect apoptosis in situ, TUNEL assay was performed using TACS XL (Trevigen) followed by hematoxylin counterstain. For staining quantification, microscopic images of three randomly chosen fields/tumor were acquired with SPOT 5.0 software, and more than 1,000 cells/tumor were counted with ImageJ software, 1.48v. Two investigators (K. Tateishi and H. Wakimoto) blinded to treatment group quantified signals, and only strongly stained cells were considered positive.

MGG4, MGG8, and MGG85 cells were treated with DMSO, 12.5 nmol/L of FK866, or GMX1778 for 72 hours, washed with cold PBS, and mounted on slides by cytospin. Cells were fixed with methanol/acetic acid (3:1) for 20 minutes, and permeabilized with 0.1% Triton X in PBS for 5 minutes. After blocking with 10% goat serum in PBS, cells were incubated with Ki-67 (1:125) at 4°C overnight, and then incubated with fluorophore-conjugated secondary antibody (1:200, Alexa Fluor 546 goat anti-mouse IgG; Life Technologies Corporation) at 4°C for 4 hours. After washing with PBS, slides were mounted with DAPI (H-1200; Vector Laboratories), and analyzed using a Fluorescent TE2000-U inverted microscope (Nikon). Microscopic images of three randomly chosen fields/cell line were acquired with SPOT 5.0 software, and more than 200 cells/cell line were counted with ImageJ software, 1.48v. Only strongly stained cells were considered positive.

For parametric analyses, two-tailed Student t tests were used, and for analysis of frequencies of nominal data, two-tailed Fisher exact test was used. Data were expressed as mean ± SE. Survival analysis was performed using the Kaplan–Meier method using the log-rank test to compare treatment arms. P values less than 0.05 were considered statistically significant. All in vitro and in vivo experiments, including cell viability assays, metabolite quantification, Western blot analysis, and flow cytometry were performed at least twice to confirm reproducibility. ¹³C metabolic flux analysis in Myc-transduced U87 glioblastoma cells and MGG4 were also performed twice with at least three replicates per condition.

Inappropriate Myc activation drives aerobic glycolysis (“Warburg effect”) in part by inducing expression of key glycolytic pathway regulators (1, 17, 18, 27). To evaluate the effect of overactive Myc on glycolysis in glioblastoma cells, we applied ¹³C metabolic flux analysis to Myc-transduced U87 glioblastoma cells, which do not have significant Myc or N-Myc expression nor MYC/MYCN gene amplification at baseline (Fig. 1A and B). Lentivirus constructs carrying either MYC or GFP driven by a constitutively active cytomegalovirus promoter were used to infect U87, generating U87-Myc and U87-GFP lines. Of note, Myc overexpression induced only a slight reduction in proliferation rate (Fig. 1C). To gauge flux through glycolysis, both U87-Myc and U87-GFP were cultured in normoxia in the presence of [U-¹³C₆] glucose for 6 hours and isotope enrichment in glycolytic metabolites was determined via CE-TOFMS. The quantity of ¹³C-labeled glycolytic intermediates glucose 6-phosphate (G6P), fructose 6-phosphate (F6P, C₆ labeled), DHAP, pyruvate, and lactic acid (C₆ labeled) were significantly increased in U87-Myc compared to U87-GFP, indicating Myc overexpression increased glycolytic flux (Fig. 1D; Supplementary Table S1). In addition, Myc overexpression resulted in significantly increased expression of the glycolytic enzymes HKII, PKM2, and LDHA in U87 cells (Fig. 1E).

We then tested whether inappropriate Myc activation would induce a dependence on glycolytic metabolism for growth. We assessed the effect of glucose deprivation on growth of cells with or without Myc overexpression. We found that U87-GFP cells survived significantly longer in glucose-free culture conditions than U87-Myc cells, indicating a switch to dependence on glucose for growth was generated when Myc was overexpressed (Fig. 1F). In addition, we inhibited glycolysis using 2-DG, a small-molecule inhibitor of hexokinase (28), in U87-Myc and U87-GFP. Similar to our findings with glucose deprivation, growth of U87-Myc cells was more sensitive to glycolytic inhibition with 2-DG than U87-GFP (Fig. 1G).

We then tested the effect of glycolytic inhibition in a panel of patient-derived glioblastoma tumorsphere lines, where Myc has been reported to play an important role in self-renewal and maintenance of tumorigenicity (3–6). We tested for MYC and MYCN gene copy number status by FISH and expression of the respective proteins by Western blot analysis. We found that three glioblastoma tumorsphere lines in our panel had high-level amplification of MYC (MGG4) or MYCN (MGG 6, MGG8; Fig. 2A), and that all amplified lines had high-level expression of the corresponding proteins (Fig. 2B). There were two lines that expressed Myc without MYC amplification (MGG18 and MGG91; Fig. 2A and B).

We next tested the effect of 2-DG in our panel of glioblastoma tumorspheres, and found that the MYC/MYCN–amplified glioblastoma tumorspheres were significantly more sensitive to the deleterious effects of 2-DG (Fig. 2C). We then knocked down MYC expression in MGG4 (MYC amplified) using shRNA and tested the effects on glycolysis (Fig. 2D). In contrast to Myc overexpression, MYC knockdown inhibited expression of the glycolytic enzymes HKII, PKM2, and LDHA (Fig. 2E). Stable Myc knockdown significantly decreased cell proliferation of MGG4 (Fig. 2F). However, after Myc knockdown, we observed significantly enhanced survival of MGG4 in glucose-free culture and with exposure to 2-DG (Fig. 2G and H), confirming that Myc activity was a major contributor to the dependence of MGG4 on glycolysis for growth. We attempted to overexpress Myc in two non-MYC/MYCN amplified glioblastoma tumorsphere lines; however, stable lines could not be generated after Myc transduction. Although this may have been due to technical limitations, it is known that ectopic Myc expression can invoke apoptosis in certain cell contexts (reviewed in ref. 2).

To evaluate the contribution of MYCN to glycolytic dependency in a highly MYCN-amplified cell line, we knocked down MYCN in IMR-32 cells (Fig. 2I and J). Similar to results observed when knocking down MYC in a highly MYC-amplified line, MYCN knockdown mildly decreased IMR-32 cell proliferation (Fig. 2K). MYCN knockdown also inhibited expression of the glycolytic enzymes HKII, PKM2, and LDHA (Fig. 2J). In addition, MYCN knockdown significantly decreased the sensitivity to the glycolysis inhibitor 2-DG (Fig. 2L). These results indicate that MYCN is a driver of cell proliferation in IMR-32 and induces a dependence on glycolysis for cell growth.

NAMPT is the rate-limiting enzyme of the NAD⁺ salvage pathway in eukaryotes, and its inhibition mediates cytotoxicity in cancer cells by attenuating glycolysis at the NAD⁺-requiring GAPDH step, among other metabolic perturbations (22, 29). We therefore tested the effect of small-molecule inhibitors of NAMPT in our panel of glioblastoma tumorsphere lines. We tested two chemically distinct, highly specific NAMPT inhibitors, FK866 (30–32) and GMX1778 (33, 34), on the growth of glioblastoma tumorspheres. Both FK866 and GMX1778 significantly decreased NAD⁺ levels in both MYC/MYCN–amplified and MYC/MYCN–nonamplified glioblastoma tumorspheres (Supplementary Fig. S1A). However, reduction of NAD⁺ with NAMPT inhibitors resulted in strikingly potent cytotoxicity in a MYC/MYCN–dependent manner. In glioblastoma tumorsphere lines tested for MYC/MYCN amplification (n = 9), the IC₅₀ values ranged between 1.3 and 6.4 nmol/L for MYC/MYCN–amplified lines (n = 3), whereas nonamplified glioblastoma tumorspheres (n = 6) and NHA remained resistant at concentrations up to 10 μmol/L, even up to 5 days (Fig. 3A; Supplementary S1B and Supplementary Table S2; P = 0.01, two-tailed Fisher exact test).

To test whether sensitivity to NAMPT inhibitors was mediated by Myc or N-Myc overactivity, we tested the effect of FK866 and GMX1778 in our MYC-transduced, MYC knockdown, and MYCN knockdown systems. We found that MYC knockdown using shRNA significantly protected MYC-amplified MGG4 from the cytotoxic effect of NAMPT inhibition (Fig. 3B), and similarly MYCN knockdown protected highly MYCN-amplified IMR-32 from NAMPT inhibitor cytotoxicity (Fig. 3C). Correspondingly, Myc overexpression induced sensitivity to NAMPT inhibitors in U87 cells, which are completely resistant to NAMPT inhibitors at baseline (Fig. 3D).

To confirm that NAMPT inhibitors inhibit glycolysis in Myc-driven glioblastoma cells, we measured total levels of glycolytic intermediates in MGG4 (MYC-amplified glioblastoma tumorsphere line) with or without NAMPT inhibitor using CE-TOFMS. Incubation with NAMPT inhibitor resulted in marked accumulation of glycolytic intermediates upstream of the GAPDH step [i.e., G6P, F6P, FBP, glyceraldehyde 3-phosphate (G3P), and DHAP], whereas glycolytic intermediates downstream of GAPDH were not significantly changed (3-phosphoglycerate and 2-phosphoglycerate) or decreased (pyruvate and lactic acid; Fig. 4A and Supplementary Table S3). These results indicated NAMPT inhibitor blocked glycolysis at the GAPDH step in MGG4. To determine whether the NAMPT inhibitor–induced glycolysis blockade observed in the MGG4 metabolic flux data were selective for MYC/MYCN–amplified glioblastoma cells, we used a second assay to quantify levels of DHAP, which rapidly and reversibly isomerizes with the GAPDH substrate G3P, in our panel of glioblastoma tumorspheres with and without NAMPT inhibition. We found that NAMPT inhibitor exposure significantly increased the levels of DHAP in the MYC/MYCN–amplified (and NAMPT inhibitor sensitive) glioblastoma tumorsphere lines, whereas those in nonamplified tumorspheres were unchanged (Fig. 4B). These data confirm that NAMPT inhibitors disrupt glycolysis at the NAD⁺-dependent GAPDH step in MYC/MYCN–amplified glioblastoma tumorspheres.

Because NAMPT inhibitors deplete intracellular NAD⁺ to mediate their effects on tumor growth, we further studied the relationship between Myc and NAD⁺. Intriguingly, we discovered that MYC/MYCN–amplified glioblastoma tumorspheres had significantly lower basal intracellular NAD⁺ levels compared with nonamplified glioblastoma (Fig. 5A and B). NAMPT inhibitor exposure significantly reduced the absolute NAD⁺ levels in both MYC/MYCN–amplified and nonamplified lines; however, NAD⁺ levels in amplified lines were nearly completely eliminated, whereas the NAD⁺ levels in nonamplified lines after NAMPT inhibitor exposure remained higher than the basal levels observed in untreated amplified lines (Fig. 5A and B). This threshold difference in residual NAD⁺ levels may have contributed to the profound cytotoxic effect of NAMPT inhibitors observed only in MYC/MYCN–amplified lines.

We then tested the apparent association between Myc activity and NAD⁺ metabolism using our MYC/MYCN knockdown and overexpression systems. Knockdown of MYC in MGG4 (MYC amplified) and MYCN in IMR-32 (MYCN amplified) resulted in increased intracellular NAD⁺ levels (Fig. 5C and D), whereas overexpression of Myc in U87 cells resulted in significant reduction of NAD⁺ (Fig. 5E). Similar to our observations in glioblastoma tumorspheres (Fig. 5A and B), NAMPT inhibitor treatment significantly reduced NAD⁺ levels in both U87-GFP and U87-Myc cells (Fig. 5E and Supplementary Fig. S2A), but the posttreatment levels in U87-GFP remained comparable with the pretreatment (basal) levels in U87-Myc (Fig. 5E), again suggesting a threshold NAD⁺ level for survival was maintained without Myc overactivity.

We then confirmed that the NAMPT inhibitor–mediated effects observed in Myc-driven glioblastoma were dependent on both NAMPT and NAD⁺. Exogenous supplementation with NAD⁺ increased intracellular NAD+ levels and rescued U87-Myc and MYC/MYCN–amplified cells from NAMPT inhibitor cytotoxicity (Fig. 5F–I). Furthermore, MYC/MYCN–amplified cells were rescued from NAMPT inhibitor cytotoxicity with NMN, a product of the NAMPT reaction and a direct precursor of NAD⁺, indicating direct NAMPT inhibition (Supplementary Fig. S2B). In contrast, supplementation with NA, a precursor for an alternate NAD⁺ salvage pathway rate-limited by nicotinic acid phosphoribosyltransferase (Naprt1), could not rescue cells from the effects of NAMPT inhibition (Supplementary Fig. S2C).

Because Myc has been reported to induce NAMPT expression (35) and Naprt1 expression is known to modulate sensitivity to NAMPT inhibitors (36, 37), we evaluated the expression of Naprt1 and NAMPT in our glioblastoma tumorsphere lines. We found that NAMPT expression was similar in all glioblastoma tumorsphere lines tested (Supplementary Fig. S3A). However, Naprt1 expression varied and was notably absent in all MYC/MYCN–amplified glioblastoma tumorspheres, whereas in nonamplified lines Naprt1 expression was highly variable and not clearly associated with Myc expression (Supplementary Fig. S3A). In addition, we found that neither forced overexpression nor knockdown of Myc altered the expression of NAMPT or Naprt1 (Supplementary Figs. S3B and S3C). These results suggest NAMPT and Naprt1 expression per se do not absolutely determine sensitivity to NAMPT inhibitors in glioblastoma tumorspheres. Together with the differences observed in glycolytic intermediates between MYC/MYCN–amplified versus Myc-expressing/non-MYC/MYCN–amplified lines after GAPDH inhibition (Fig. 4B), our findings point to MYC/MYCN genetic amplification, rather than expression of NAMPT, Naprt1, or Myc (without gene amplification), as critical for determining sensitivity to glycolytic inhibition.

Prior studies have revealed that NAMPT inhibitor–mediated glycolytic blockade causes intracellular depletion of NAD⁺ and ATP, resulting in metabolic perturbation and cytotoxicity (22). We similarly found that NAMPT inhibitors caused ATP depletion and cell death in MYC/MYCN–amplified glioblastoma tumorspheres (Fig. 6A–C). Cytotoxicity was mediated by apoptosis, as indicated by significant increases in Annexin V staining and double staining for Annexin V and PI by flow cytometry, and caspase-3/7 activation after treatment with NAMPT inhibitor (Fig. 6C and Supplementary Fig. S4A). We further observed activation of the intracellular energy regulator AMPK, inhibition of mTOR signaling, and induction of PARP cleavage, indicating intracellular metabolic perturbation after NAMPT inhibition in MYC/MYCN–amplified cells (Supplementary Fig. S4B), but not in MGG85 (MYC/MYCN nonamplified) cells (Supplementary Fig. S4C). To assess whether NAMPT inhibitors affected cell proliferation, we assessed Ki-67 expression in glioblastoma tumorspheres. Although the fraction of Ki-67–expressing cells in MGG85 (MYC/MYCN nonamplified) cells was unchanged, NAMPT inhibitors significantly decreased the Ki-67–positive fraction in MYC/MYCN–amplified cells (Fig. 6D and Supplementary Fig. S4D).

To test the effect of glycolytic inhibition on Myc-driven glioblastoma cells in vivo, we investigated NAMPT inhibitors due to their striking selectivity for MYC/MYCN–amplified glioblastoma tumorspheres and their feasibility for use in vivo. We tested the effect of GMX1778 on MGG8 (MYCN amplified), which rapidly and diffusely infiltrates mouse brains (21), and U87 (MYC/MYCN nonamplified) orthotopic xenografts. Oral GMX1778 significantly extended the survival of mice bearing MGG8 orthotopic xenografts compared with vehicle dosing (P < 0.0005, log-rank test; Fig. 6E). On necropsy, xenografts tumors from animals treated with GMX1778 compared with tumors from vehicle-treated mice had significantly decreased cells positive for Ki-67 staining (80.1% ± 1.4% for vehicle vs. 53.0% ± 8.1% for GMX1778, P = 0.03) and increased apoptotic cells as measured by TUNEL staining (1.4% ± 0.2% for vehicle vs. 15.1 ± 3.3% for GMX1778, P = 0.03; Fig. 6F). In contrast, oral GMX1778 had no effect on the survival of mice bearing U87 orthotopic xenografts (Fig. 6G).

To explore whether our finding that Myc-driven glioblastoma cells are selectively sensitive to glycolytic inhibition is generalizable to other cancer types, we tested the effect of NAMPT inhibitors in several standard cell lines from multiple cancer types known to harbor genetically activated MYC, including H1975 lung (MYC amplified), D283 Med medulloblastoma (MYC amplified), IMR-32, and Kelly neuroblastoma (MYCN amplified), and Raji Burkitt's lymphoma (IGH-MYC translocation; Fig. 2I and Supplementary Fig. S5). Similar to our findings in MYC/MYCN–amplified glioblastoma tumorspheres, NAMPT inhibitors displayed low nanomolar activity against genetically Myc-driven cell lines (Supplementary Fig. S6A). Furthermore, consistent with MGG4, shRNA knockdown of Myc in H1975 cells (highly MYC amplified; Fig S5) resulted in increased intracellular NAD⁺ levels and rescued H1975 from the cytotoxic effects of NAMPT inhibition (Supplementary Figs. S6B–S6D).

We also tested the effect of GMX1778 on established H1975 (MYC amplified) subcutaneous flank xenograft tumors. Mice bearing H1975 xenografts treated with NAMPT inhibitors had significantly inhibited tumor growth (mean tumor weight ± SE, 101.7 ± 28.1 mg for GMX1778 treatment vs. 612.0 ± 129.8 mg for vehicle treatment, P = 0.002, t test; Supplementary Figs. S7A–S7C). Tumors from mice treated with GMX1778 had increased TUNEL staining (Supplementary Fig. S7D) and tumor regressions were observed the day after the weekly dose (Supplementary Fig. S7A). Notably, detectable NAD⁺ was nearly eliminated in tumors from GMX1778-treated mice (Supplementary Fig. S7E). All animals with flank or intracerebral xenografts tolerated oral GMX1778 well, with no significant body weight change or apparent toxicities during treatment (Supplementary Figs. S8A–S8B).

Here, we demonstrate that overactive Myc drives an abnormal metabolic program in glioblastoma cells, increasing glycolytic flux and altering NAD⁺ metabolism. We found that Myc-driven glioblastoma cells are sensitive to inhibition of the Myc-induced increased glycolytic drive, and identified NAMPT inhibitors, which attenuate glycolysis by depleting NAD⁺ required for the GAPDH step, as a selective and strikingly potent strategy for MYC/MYCN–amplified glioblastoma, Ultimately, our findings support and extend on the work of others that have demonstrated that overactive Myc generates an altered metabolic state in cancer cells that can be exploited by identifying their addiction to specific metabolites (19, 38).

It is now clear that glioblastoma represents a group of molecularly heterogeneous subtypes with distinct clinical phenotypes (9–11) and that therapeutic development strategies must focus on specific biologic subtypes to derive maximal benefit. Our strategy would target a small proportion of primary glioblastoma, as approximately 4% of primary glioblastomas harbor high-level MYC or MYCN amplification (7–11). However, most large-scale genetic studies have focused on primary (initial) glioblastoma tumors and the few comprehensive studies that analyzed the genetic changes at develop at recurrence included only small numbers of paired tumors (39, 40). The prevalence of MYC/MYCN copy number alterations developing at recurrence in primary glioblastoma is unclear. Interestingly, recent large-scale molecular studies of isocitrate dehydrogenase 1 (IDH1) and IDH2-mutant gliomas have identified a high proportion of MYC amplification (up to 31%) in this subtype of glioma (41–44). Although most MYC amplification events were low-level in these studies, the majority were observed at progression (44) and in the non-1p/19q codeleted subtype (43), suggesting an association with more aggressive clinical behavior in IDH1/2–mutant gliomas (45). Thus, our strategy could apply to a significant proportion of glioma subtypes.

We also identified an intriguing association between overactive Myc and reduced NAD⁺ levels, suggesting Myc reprograms NAD⁺ metabolism. Our findings contrast with a prior study that showed Myc induced expression of NAMPT and resulted in increased NAD⁺ levels (35). In our study, Myc overexpression and knockdown had no effect on NAMPT expression, and increased Myc led to lower NAD⁺ levels. The reasons for this discrepancy are unclear and may relate to cell context, as all our NAD⁺ measurements were performed in malignant cells. NAD⁺ is a fundamental metabolite for numerous critical intracellular pathways and further studies will be needed to elucidate the complex mechanisms of NAD⁺ regulation, the effects of NAMPT inhibition and the relationship between Myc and NAD⁺. Intriguingly, we found evidence that a critical NAD⁺ threshold may exist for cancer cell survival. This preliminary finding warrants further detailed investigation because together with our observation that NAMPT inhibitors nearly completely eliminate detectable NAD⁺ levels in xenograft tumors, our data suggests intratumoral NAD⁺ level represents a potential predictive and/or pharmacodynamic biomarker in Myc-driven cancers.

It is likely that other genetic or epigenetic alterations that similarly impact cancer cell metabolism may also sensitize to NAMPT inhibitors, as NAMPT inhibitor activity has been observed in diverse cancer types (34, 46). Whether increased Myc/N-Myc expression without MYC/MYCN genetic alteration sensitizes to glycolytic inhibition remains to be completely determined. We found that the association with glycolytic inhibitor sensitivity in glioblastoma cells was most tightly correlated with gene amplification, particularly with NAMPT inhibitors. It is possible that genetic events that inappropriately activate Myc, such as amplification, translocation, or lentiviral transduction, are more stable or potent means of driving overactive Myc and altering cancer metabolism. In addition, there may be differences in downstream effects between MYC/MYCN genetic alterations and Myc/N-Myc protein expression without genetic alteration. We find evidence of this in our studies as the effect on glycolytic intermediates differed between MYC/MYCN–amplified versus Myc overexpressed (nonamplified) glioblastoma tumorspheres after inhibition of the key glycolytic enzyme GAPDH.

Nevertheless, our findings support the paradigm of glycolytic inhibition as a novel therapeutic strategy for a genetically defined subgroup within a wide array of cancers. Amplification of the MYC and MYCN genes are among the most frequent amplification events in human cancer (47), and we provide evidence that our findings in glioblastoma may translate to other cancer types, as we observed potent cytotoxicity with NAMPT inhibitors in numerous genetically Myc-driven cancer cells in vitro and in vivo. In addition, because NAMPT inhibitors have entered clinical development (48), our findings have the potential for rapid translation in a genetically enriched subset of glioblastoma patients.

A.J. Iafrate holds ownership interest (including patents) in ArcherDx, and is a consultant/advisory board member for Chugai, Constellation, DebioPharm, and Pfizer. T. Batchelor reports receiving commercial research grants from Pfizer; speakers bureau honoraria from Merck; and is a consultant/advisory board member for Oxigene, Roche, and Upsher. D.P. Cahill is a consultant/advisory board member for Merck and Vertex. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Tateishi, A.J. Iafrate, D.P. Cahill, A.S. Chi, H. Wakimoto

Development of methodology: K. Tateishi, A.J. Iafrate, A.S. Chi, H. Wakimoto

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Tateishi, A.J. Iafrate, W.T. Curry, T.T. Batchelor, M.L. Onozato, N. Lelic, A.S. Chi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Tateishi, A.J. Iafrate, K.T. Flaherty, A.S. Chi

Writing, review, and/or revision of the manuscript: K. Tateishi, A.J. Iafrate, W.T. Curry, T.T. Batchelor, K.T. Flaherty, D.P. Cahill, A.S. Chi, H. Wakimoto

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.J. Iafrate, Q. Ho, T.T. Batchelor, N. Lelic, S. Sundaram, A.S. Chi

Study supervision: D.P. Cahill, A.S. Chi, H. Wakimoto

The authors thank Hideyuki Takeuchi, Fares Nigim, and Mara Koerner for technical assistance, Shinichi Esaki, and Jianfang Ning for growing glioblastoma tumorsphere lines, and Laura Shelton and Takushi Oga at Human Metabolome Technologies, Inc., for assistance with and analysis of ¹³C metabolite labeling assays.

This work is supported by U.S. NIH P50CA165962-01A1 [to T.T. Batchelor (Principal Investigator), D.P. Cahill, H. Wakimoto, and A.S. Chi] and K24 CA125440-06 (to T.T. Batchelor), a Burroughs Wellcome Fund Career Award (to D.P. Cahill). K. Tateishi is supported by a Society of Nuclear Medicine and Molecular Imaging Wagner-Torizuka Fellowship, the Japan Brain Foundation, and the KANAE foundation for the Promotion of Medical Science. A.S. Chi is supported by a Richard B. Simches Scholars Award and the MGH Glioblastoma Research Fund in memory of David M. King, Kirit N. Patel, Sarah E. Rippetoe, and Eric W. Schwartz.

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.