Combined inhibition of de novo glutathione and nucleotide biosynthesis is synthetically lethal in glioblastoma

SUMMARY

Understanding the mechanisms by which oncogenic events alter metabolism will help identify metabolic weaknesses that can be targeted for therapy. Telomerase reverse transcriptase (TERT) is essential for telomere maintenance in most cancers. Here, we show that TERT acts via the transcription factor forkhead box O1 (FOXO1) to upregulate glutamate-cysteine ligase (GCLC), the rate-limiting enzyme for de novo biosynthesis of glutathione (GSH, reduced) in multiple cancer models, including glioblastoma (GBM). Genetic ablation of GCLC or pharmacological inhibition using buthionine sulfoximine (BSO) reduces GSH synthesis from [U-¹³C]-glutamine in GBMs. However, GCLC inhibition drives de novo pyrimidine nucleotide biosynthesis by upregulating the glutamine-utilizing enzymes glutaminase (GLS) and carbamoyl-phosphate synthetase 2, aspartate transcarbamoylase, and dihydroorotatase (CAD) in an MYC-driven manner. Combining BSO with the glutamine antagonist JHU-083 is synthetically lethal in vitro and in vivo and significantly extends the survival of mice bearing intracranial GBM xenografts. Collectively, our studies advance our understanding of oncogene-induced metabolic vulnerabilities in GBMs.

In brief

Metabolic reprogramming and telomere maintenance are hallmarks of cancer. Udutha et al. demonstrate that TERT, which drives telomere maintenance, also upregulates glutathione synthesis in glioblastoma. Targeting glutathione synthesis, however, rewires metabolism to augment nucleotide biosynthesis. Importantly, combined inhibition of glutathione and nucleotide synthesis is synthetically lethal in preclinical glioblastoma models.

Graphical abstract

INTRODUCTION

Telomerase reverse transcriptase (TERT) expression is a defining characteristic of ~85% of human tumors, including glioblastoma (GBM), oligodendroglioma, melanoma, and hepatocellular carcinoma.^1,2 TERT is the catalytic, rate-limiting component of the enzyme telomerase that synthesizes telomeric DNA for incorporation into telomeres.^1,2 Telomeres are cap-like nucleoprotein complexes that protect the ends of linear chromosomes from damage during replication.^1,2 They shorten with every cell division until the cell undergoes either senescence or apoptosis, thereby imposing a natural limit on cell proliferation.^1,2 TERT expression is silenced early in development in normal somatic cells, except for stem cells, and is reactivated in cancer cells via hotspot mutations in the TERT promoter.² In contrast, ~15% of tumors, including astrocytoma, pediatric high-grade hemispheric glioma, and osteosarcoma, utilize a homologous-recombination-based method of telomere maintenance known as the alternative lengthening of telomeres (ALT) pathway.^3–7 Although the mechanistic basis for the ALT pathway is less well understood, inactivating mutations in the histone chaperone protein α-thalassemia retardation X-linked (ATRX) are necessary for induction of the ALT pathway.^3–9

Due to its role in driving tumor immortality, TERT has long been an attractive therapeutic target.^10,11 However, drugs that directly inhibit TERT, such as imetelstat, have not shown efficacy in clinical trials, possibly due to the observed lag period before growth inhibition due to telomere shortening.¹² This delay is likely to be particularly detrimental for aggressive tumors like GBMs. Targeting TERT directly also has the potential to lead to toxicity toward stem cells and germline cells that also require telomere synthesis for proliferation.^10–12

Metabolic reprogramming is a hallmark of cancer.¹³ Oncogenic events in cancer cells alter metabolism to generate biosynthetic intermediates and maintain redox homeostasis.¹³ However, metabolic networks are flexible, and exploiting cancer metabolism for therapy requires a mechanistic understanding of how targeting a specific oncogene-induced metabolic alteration reshapes metabolism.^14,15 We and others have previously linked TERT expression to metabolic reprogramming and, particularly, to altered redox in cancer.^16–22 Specifically, we showed that TERT negatively regulates the transcription factor forkhead box O1 (FOXO1),¹⁷ which is a master regulator of metabolism,^23,24 in TERT-driven cancers, including GBM, oligodendroglioma, melanoma, neuroblastoma, and hepatocellular carcinoma. FOXO1 suppresses the expression of nicotinamide phosphoribosyl transferase, the rate-limiting enzyme for NAD+ biosynthesis, and glyceraldehyde-3-phosphate dehydrogenase, the glycolytic enzyme that converts NAD+ to NADH.¹⁷ As a result, TERT elevates steady-state NAD+, NADH, and the NADH/NAD+ ratio.¹⁷ In other studies, we showed that TERT upregulates the expression of glucose-6-phosphate dehydrogenase and drives elevated glucose flux through the pentose phosphate pathway in oligodendrogliomas.^18,19 The pentose phosphate pathway is the major source of NADPH, which maintains the antioxidant glutathione (GSH, reduced) in the reduced state, and TERT expression leads to elevated NADPH.^18,19 In all these studies, we also demonstrated that GSH abundance is elevated in TERT-expressing cells and intracranial tumor xenografts.^16–19 However, the precise mechanism by which TERT upregulates GSH and whether this alteration is a druggable vulnerability remains unknown.

GSH is the most abundant antioxidant within mammalian cells and exists predominantly in a reduced state in the cytosol.^25,26 GSH plays an essential role in cell signaling and defense by functioning as an enzyme cofactor, modulating protein function via glutathionylation, and detoxifying reactive metabolites and xenobiotics.^25,26 Importantly, GSH mitigates oxidative stress by scavenging reactive oxygen species (ROS) and is converted to oxidized GSH (GSSG) in this process. Conversion back to GSH is mediated by the NADPH-dependent enzyme GSH reductase.^25,26 GSH is a tripeptide of glutamate, cysteine, and glycine (γ-glutamylcysteinylglycine). Although cells can import GSH, extracellular GSH concentrations are ~3 orders of magnitude lower than intracellular concentrations, with the result that intracellular GSH concentrations are largely dependent on de novo synthesis.²⁷ The first and rate-limiting step in GSH synthesis is the formation of γ-glutamyl cysteine by γ-glutamyl cysteine ligase (GCL), which is a heterodimer of a catalytic subunit (glutamate-cysteine ligase catalytic [GCLC]) and a regulatory subunit (glutamate-cysteine ligase modifier [GCLM]). GSH synthetase (GS) subsequently conjugates γ-glutamyl cysteine with glycine to produce GSH.²⁸

The goal of our study was to delineate the molecular mechanism by which TERT maintains GSH homeostasis and determine whether this process can be exploited for therapy. Our studies indicate that GCLC is a specific, targetable vulnerability in TERT-dependent cancers. However, GCLC inhibition causes a compensatory increase in de novo pyrimidine nucleotide biosynthesis from [U-¹³C]-glutamine via MYC-driven upregulation of the glutamine-utilizing enzymes glutaminase (GLS) and carbamoyl-phosphate synthetase 2, aspartate transcarbamoylase, and dihydroorotatase (CAD). Importantly, combined inhibition of GCLC, GLS, and CAD is synthetically lethal and induces tumor regression in preclinical GBM models in vivo. Collectively, our studies identify TERT-dependent metabolic vulnerabilities that are synthetically lethal and can be exploited for therapy in GBMs and potentially other cancers.

RESULTS

TERT upregulates de novo GSH synthesis and GCLC expression in a FOXO1-dependent manner in an isogenic astrocyte model

To begin with, we studied immortalized normal human astrocytes (NHAs) with (NHA_TERT) or without (NHA_CONTROL) TERT expression. We also examined immortalized NHAs in which ATRX was silenced via short hairpin RNA (shRNA) (ATRX-knockout NHA [NHA_ATRX-KO]). We confirmed TERT expression in NHA_TERT cells relative to NHA_CONTROL and NHA_ATRX-KO cells (Figure 1A). We also confirmed the loss of ATRX in NHA_ATRX-KO cells (Figure 1B). As shown in Figure 1C, GSH abundance was significantly higher in NHA_TERT cells relative to NHA_CONTROL and NHA_ATRX-KO cells. Elevated GSH abundance can result from higher de novo GSH synthesis or higher recycling of GSSG to GSH. Neither GSSG abundance nor GSH reductase activity was altered in NHA_TERT cells relative to NHA_CONTROL and NHA_ATRX-KO cells (Figures S1A and S1B).

Figure 1. TERT acts via FOXO1 to upregulate GCLC in isogenic models.

(A–E) TERT mRNA (A), ATRX mRNA (B), GSH pool size (C), percentage of ¹³C labeling of GSH from [U-¹³C]-glutamine (D), and GCLC mRNA (E) in NHA_CONTROL, NHA_TERT, and NHA_ATRX-KO cells.

(F) Western blots for GCLC in NHA_CONTROL, NHA_TERT, and NHA_ATRX-KO cells. β-actin was used as the loading control.

(G) GCL activity in NHA_CONTROL, NHA_TERT, and NHA_ATRX-KO cells.

(H) Western blots for phosphorylated and total FOXO1 in NHA_CONTROL, NHA_TERT, and NHA_ATRX-KO cells.

(I) FOXO1 binding to the GCLC promoter as measured by ChIP-qPCR in NHA_CONTROL, NHA_TERT, and NHA_ATRX-KO cells.

(J) Western blots for the FLAG tag in NHA_CONTROL, NHA_TERT, and NHA_TERT cells expressing a FLAG-tagged constitutively active form of FOXO1 (CA-FOXO1). β-actin was used as the loading control.

(K) FOXO1 binding to the GCLC promoter measured by ChIP-qPCR in NHA_CONTROL, NHA_TERT, and NHA_TERT cells expressing CA-FOXO1.

(L–N) GCLC protein expression (L), GSH pool size (M), and percentage of ¹³C labeling of GSH from [U-¹³C]-glutamine (N) in NHA_CONTROL, NHA_TERT, and NHA_TERT cells expressing a FLAG-tagged CA-FOXO1.

Data are presented as mean ± standard deviation. **p < 0.01, ***p < 0.001, and ****p < 0.0001; ns indicates lack of statistical significance. See also Figure S1.

Glutamine is the key carbon donor for de novo GSH synthesis.^27,28 [U-¹³C]-glutamine (m+5) is converted by GLS to glutamate (m+5), which is then conjugated with cysteine by GCL to produce γ-glutamyl cysteine. The addition of glycine to γ-glutamyl cysteine produces GSH (m+5). GSH (m+5) was significantly higher in NHA_TERT cells relative to NHA_CONTROL and NHA_ATRX-KO cells, an effect that was associated with elevated GCLC mRNA, GCLC protein, and GCL activity (Figures 1D–1G). There was no difference in the expression of GCLM or GS between NHA_CONTROL, NHA_TERT, or NHA_ATRX-KO cells (Figures S1C and S1D).

Next, we examined the mechanism by which TERT upregulates GCLC expression. The FOXO family of transcription factors coordinates metabolism and redox homeostasis.^23,24 We previously showed that TERT negatively regulates FOXO1 via inhibitory phosphorylation in cancer, including GBMs, oligodendroglioma, melanoma, and hepatocellular carcinoma.¹⁷ Phosphorylation sequesters FOXO1 in the cytoplasm, thereby preventing FOXO1 activity in the nucleus.^23,24 As shown in Figure 1H, FOXO1 phosphorylation was significantly higher in NHA_TERT cells relative to NHA_CONTROL and NHA_ATRX-KO cells. Concomitantly, chromatin immunoprecipitation quantitative PCR (ChIP-qPCR) showed significantly reduced binding of FOXO1 to the GCLC promoter in NHA_TERT cells relative to NHA_CONTROL and NHA_ATRX-KO cells (Figure 1I).

To confirm the role of FOXO1 in regulating GCLC expression, we examined the effect of expressing a constitutively active form of FOXO1 (hereafter named CA-FOXO1) in NHA_TERT cells (Figure 1J).²⁹ CA-FOXO1 is a FLAG-tagged form of FOXO1 that cannot be phosphorylated due to the mutation of all three phosphorylation sites to alanine and is, therefore, constitutively active.²⁹ As shown in Figures 1K–1N, the expression of CA-FOXO1 in NHA_TERT cells restored FOXO1 binding to the GCLC promoter and reduced GCLC expression, GSH abundance, and de novo synthesis of GSH (m+5) from [U-¹³C]-glutamine to levels observed in NHA_CONTROL cells. Collectively, these results indicate that TERT acts via FOXO1 to upregulate GCLC expression and drive de novo GSH synthesis from glutamine in an isogenic model system.

TERT acts via FOXO1 to upregulate GCLC expression in patient-derived cancer models and patient biopsies

To broadly confirm the relevance of these results, we examined the effect of silencing TERT in multiple cancer models (Figure S1E). As shown in Figures 2A–2C, silencing TERT significantly upregulated FOXO1 binding to the GCLC promoter, downregulated GCLC expression, and reduced GSH abundance in patient-derived and syngeneic GBM (GBM6, U251, and SB28), oligodendroglioma (SF10417), melanoma (A375), and hepatocellular carcinoma (HepG2) models. In contrast, re-expressing ATRX did not alter FOXO1 binding to the GCLC promoter, GCLC expression, or GSH abundance in patient-derived astrocytoma (BT142), pediatric diffuse hemispheric glioma (KNS42), or osteosarcoma (U2OS) models (Figures 2D–2F). To validate the clinical relevance of our results, we examined GBM patient tissue and compared it to non-neoplastic gliosis (since obtaining normal brain tissue is challenging).^16–18 As added controls, we examined tissue from patients with astrocytoma.^16,30 We confirmed that TERT expression was significantly higher while FOXO1 binding to the GCLC promoter was significantly lower in GBM biopsies relative to gliosis and astrocytoma (Figures 2G and 2H), consistent with our prior studies.¹⁷ Importantly, as shown in Figure 2I, GCLC expression was significantly higher in GBM patient tissue relative to gliosis or astrocytoma. Collectively, our results indicate that TERT relieves FOXO1-mediated repression of GCLC expression, thereby upregulating de novo GSH synthesis in cancer.

Figure 2. TERT upregulates GCLC in patient-derived models and patient biopsies.

(A–C) FOXO1 binding to the GCLC promoter (A), GCLC protein expression (B), and GSH pool size (C) in GBM6, U251, SB28, SF10417, A375, and HepG2 cells transfected with non-targeting control small interfering RNA (siRNA; siNT) or two non-overlapping siRNA sequences against TERT (siTERT-1 and siTERT-2).

(D–F) FOXO1 binding to the GCLC promoter (D), GCLC mRNA (E), and GSH pool size (F) in BT142, KNS42, and U2OS cells transfected with an empty vector or with a plasmid expressing ATRX.

(G–I) TERT mRNA (G), FOXO1 binding to the GCLC promoter (H), and GCLC mRNA (I) in GBM, astrocytoma, or gliosis biopsies.

Data are presented as mean ± standard deviation. **p < 0.01, ***p < 0.001, and ****p < 0.0001; ns indicates lack of statistical significance.

Genetic ablation or pharmacological inhibition of GCLC depletes GSH and inhibits the viability of TERT-dependent cancer cells

Directly targeting TERT using small-molecule inhibitors is an approach that is complicated by issues of toxicity to normal stem cells and the lag period before tumor growth inhibition can be observed.^10–12 Metabolic reprogramming associated with TERT expression provides targets that are potentially druggable. To determine whether GCLC is a targetable vulnerability in the context of TERT expression, we examined the effect of genetic ablation of GCLC (Figure S2A) or pharmacological inhibition of GCL activity using buthionine sulfoximine (BSO; Figure S2B)³¹ in TERT-dependent cancer cells. As shown in Figures 3A–3F, targeting GCLC reduced GSH, increased ROS, and abrogated the viability of GBM, oligodendroglioma, melanoma, and hepatocellular carcinoma cells with IC50 values of ~10–25 μM for BSO. BSO did not inhibit the viability of ReN human neural stem cells or NHAs (Figure 3F). In contrast to TERT-driven cancer cells, targeting GCLC did not alter the abundance of GSH or ROS or the viability of ALT-dependent cancer cells, including astrocytoma, pediatric diffuse hemispheric glioma, or osteosarcoma cells (Figures S2C–S2J). Taken together, these results identify GCLC as a specific metabolic vulnerability in TERT-dependent cancers.

Figure 3. Targeting GCLC depletes GSH, induces oxidative DNA damage, and inhibits the viability of TERT-dependent cancer cells.

(A and B) Effect of GCLC silencing (A) or BSO (B) on GSH pool size in tumor cells.

(C and D) Effect of GCLC silencing (C) or BSO (D) on levels of ROS in tumor cells.

(E) Effect of GCLC silencing on live cell viability in the GBM6, U251, SB28, SF10417, A375, and HepG2 models.

(F) Dose-response curves for BSO in the GBM6, U251, SB28, SF10417, A375, and HepG2 models. ReN human neural stem cells or primary human astrocytes were used as controls. Cells were treated with the indicated concentrations of BSO and IC50 measured via the percentage of inhibition of viability using the RealTime-Glo assay.

(G) Left: representative flow cytometric histograms of live GBM6 cells stained with Vybrant DyeCycle green showing DNA content distribution. Cells were transfected with non-targeted control siRNA (siNT) or siRNA against GCLC (siGCLC-1). siGCLC-1-transfected cells were incubated with GSH ethyl ester to rescue the effect of GCLC silencing (rescue). Right: quantification of the percentage of cells in the G1, S, and G2/M phases in siNT, siGCLC-1, and rescue GBM6 cells.

(H) Left: representative histograms of live GBM6 cells stained with Vybrant DyeCycle green showing DNA content distribution. Cells were treated with vehicle (control), 10 μM BSO (BSO), or 10 μM BSO + 100 μM GSH ethyl ester (rescue). Right: effect of BSO on the percentage of cells in the G1, S, and G2/M phases in GBM6 cells.

(I and J) Effect of GCLC silencing (I) or BSO (J) on levels of 8-OHdG in GBM6 cells. 100 μM GSH ethyl ester was used to rescue the effect of GCLC silencing or BSO.

(K and L) Effect of GCLC silencing on caspase activity (K) or levels of malondialdehyde (L) in GBM6, U251, or SB28 cells.

Data are presented as mean ± standard deviation. **p < 0.01, ***p < 0.001, and ****p < 0.0001; ns indicates lack of statistical significance. See also Figures S2 and S3.

Targeting GCLC induces oxidative DNA damage and blocks cell cycle progression in GBM cells

Next, we characterized the effect of GCLC silencing or BSO on cell cycle progression, doubling time, and cell death in our GBM models. To add rigor to our studies, we also examined the murine syngeneic SB28 model,^32,33 in addition to the human GBM6 and U251 models. As shown in Figures 3G, 3H, and S3A–S3D, targeting GCLC resulted in a significant increase in the proportion of cells arrested in the G1 phase, with a concomitant decrease in the proportion of cells in the S and G2/M phases. GCLC silencing and BSO also significantly increased the doubling time of GBM6, U251, and SB28 cells (Figures S3E and S3J). Mechanistically, ROS can damage DNA via base modifications, such as the oxidation of guanine to 8-hydroxy-2-deoxyguanosine (8-OHdG).³⁴ We confirmed that targeting GCLC significantly increased the abundance of 8-OHdG in all GBM models (Figures 3I, 3J, and S3K–S3N). Importantly, the effects of BSO on cell cycle progression, doubling time, and oxidative DNA damage could be rescued by supplementation with cell-permeable GSH (Figures 3G–3J and S3A–S3N), confirming that the effect of BSO on viability was mediated by the on-target loss of GSH. Neither GCLC silencing nor treatment with BSO induced apoptosis in GBM cells (Figures 3K and S3O). GSH depletion and disruption of redox can damage lipids via peroxidation, leading to the formation of reactive products such as malondialdehyde and the induction of ferroptosis.³⁵ However, we did not observe ferroptosis following GCLC silencing or BSO treatment in GBM cells (Figures 3L and S3P). Taken together, our findings indicate that targeting GCLC depletes GSH and induces oxidative stress that inhibits the viability of TERT-driven cancer cells in a cytostatic manner.

Targeting GCLC downregulates de novo GSH synthesis but causes compensatory upregulation of pyrimidine nucleotide biosynthesis in GBM cells

Metabolic plasticity allows cancer cells to adapt to therapy.^14,15 To determine whether the lack of cell death resulted from compensatory metabolic alterations, we examined the effect of GCLC silencing or BSO on [U-¹³C]-glutamine metabolism in GBM cells. Figure 4A shows a schematic illustration of glutamine metabolism in mammalian cells. In addition to GSH synthesis, glutamine-derived glutamate can be transaminated to α-ketoglutarate (α-KG), which is then metabolized via the tricarboxylic acid (TCA) cycle to produce oxaloacetate, and aspartate.^36–38 Glutamine, aspartate, and bicarbonate are converted to dihydroorotate by a single trifunctional, rate-limiting enzyme called CAD.³⁹ Subsequent metabolism generates the pyrimidine nucleotides uridine triphosphate (UTP) and cytidine triphosphate (CTP). Alternately, glutamine can be shunted toward purine biosynthesis and incorporated into guanosine monophosphate (GMP) and adenosine monophosphate (AMP).³⁹ As shown in Figures 4B–4F, both GCLC silencing and treatment with BSO reduced GSH synthesis from [U-¹³C]-glutamine, further confirming on-target activity in GBM cells. However, GCLC silencing and BSO also upregulated ¹³C labeling of glutamate and oxidative metabolism of [U-¹³C]-glutamine via the TCA cycle to m+4 α-KG, succinate, and malate (Figures 4B–4F). We did not observe any change in m+5 citrate, indicating that reductive glutamine metabolism was not altered by GCLC inhibition.^37,40 Importantly, GCLC inhibition and BSO increased m+4 aspartate, m+4 dihydroorotate, and m+3 UTP and CTP (Figures 4B–4F). We confirmed that these differences in dynamic glutamine metabolism resulted in corresponding differences in metabolite pool sizes, i.e., reduced GSH abundance and elevated glutamate, TCA cycle metabolites, aspartate, and pyrimidine nucleotides (Figures S4A–S4D). GCLC inhibition did not impact purine nucleotide abundance or synthesis from [U-¹³C]-glutamine in either model (see Figures 4B–4F and S4A–S4D). Taken together, these results suggest that targeting GCLC causes an on-target reduction in GSH synthesis but results in compensatory upregulation of oxidative glutamine metabolism and pyrimidine nucleotide biosynthesis in GBM cells.

Figure 4. GCLC silencing or inhibition rewires [U-¹³C]-glutamine metabolism in GBM cells.

(A) Schematic illustration of [U-¹³C]-glutamine metabolism in cancer cells. [U-¹³C]-glutamine (m+5 glutamine) is catabolized by GLS to m+5 glutamate, which can then be converted by GCLC to m+5 γ-glutamyl cysteine and then to m+5 GSH. Glutamate m+5 can also be transaminated to m+5 α-KG by transaminases such as GLUD1, GLUD2, BCAT1, and GPT1. α-KG m+5 is oxidized via the TCA cycle to m+4 succinate, m+4 malate, and m+4 oxaloacetate. Oxaloacetate m+4 is conjugated with acetyl-coenzyme A (acetyl-CoA) derived from glucose to form m+4 citrate. In addition to oxidative metabolism, m+5 α-KG can be reductively carboxylated by IDH1 to m+5 isocitrate and then m+5 citrate. Alternately, m+4 oxaloacetate is converted by GOT1 to m+4 aspartate. CAD combines aspartate, glutamine, and bicarbonate to produce m+4 dihydroorotate (DHO), which is subsequently metabolized by DHODH and UMPS to the pyrimidine nucleotides m+3 UTP and m+3 CTP. Aspartate m+4, along with glutamine and glycine, also produces the purine nucleotides m+3 AMP and m+3 GMP.

(B and C) Effect of silencing GCLC (B) or pharmacological inhibition using BSO (C) on [U-¹³C]-glutamine metabolism in GBM6 cells.

(D and E) Effect of silencing GCLC (D) or BSO (E) on [U-¹³C]-glutamine metabolism in U251 cells.

(F) Effect of silencing GCLC on [U-¹³C]-glutamine metabolism in SB28 cells.

Data are presented as mean ± standard deviation. **p < 0.01, ***p < 0.001, and ****p < 0.0001; ns indicates lack of statistical significance. See also Figure S4.

Targeting GCLC upregulates GLS and CAD in an MYC-dependent manner in GBMs

Next, we examined the mechanism by which GCLC inhibition upregulates pyrimidine nucleotide biosynthesis in GBMs. Expression profiling of enzymes involved in glutamine metabolism showed that GCLC silencing or BSO treatment significantly upregulated the expression of GLS and CAD (Figures 5A–5F), which are the rate-limiting enzymes for glutamate and dihydroorotate synthesis (see schematic illustration in Figure 4A).^36,39 We also observed the upregulation of the glutamate transaminases GOT1 and GLUD1 and the downregulation of GS. Since MYC is a known transcriptional regulator of both GLS and CAD,^41,42 we examined whether MYC plays a role in the upregulation of GLS and CAD following GCLC inhibition. As shown in Figures 5G and 5H, both GCLC silencing and treatment with BSO upregulated MYC expression in the GBM6, U251, and SB28 models. Importantly, silencing MYC (see Figures S4E and S4F for confirmation of MYC silencing) abrogated the increase in GLS and CAD expression induced by targeting GCLC in GBM6, U251, and SB28 cells (Figures 5I–5L). Collectively, these results suggest that GCLC inhibition upregulates MYC, which then drives the expression of GLS and CAD in GBM cells.

Figure 5. Targeting GCLC upregulates GLS and CAD in an MYC-dependent manner in GBM cells.

(A and B) Effect of silencing GCLC (A) or BSO (B) on mRNA levels of enzymes involved in glutamine metabolism in GBM6 cells.

(C and D) Effect of GCLC silencing (C) or BSO (D) on mRNA levels of enzymes involved in glutamine metabolism in U251 cells.

(E and F) Effect of GCLC silencing (E) or BSO (F) on mRNA levels of enzymes involved in glutamine metabolism in SB28 cells.

(G and H) Effect of GCLC silencing (G) or BSO (H) on mRNA levels of MYC in GBM6, U251, and SB28 cells.

(I) GLS mRNA in GBM6, U251, or SB28 cells transfected with non-targeted siRNA (siNT), siRNA against GCLC (siGCLC), siRNA against MYC (siMYC), or siRNA against both MYC and GCLC (siGCLC siMYC).

(J) GLS mRNA in GBM6, U251, or SB28 cells treated with vehicle (DMSO) or 10 μM BSO, either alone or transfected with siRNA against MYC.

(K) CAD mRNA in GBM6, U251, or SB28 cells transfected with siNT, siGCLC, siMYC, or siGCLC siMYC.

(L) CAD mRNA in GBM6, U251, or SB28 cells treated with vehicle (DMSO) or 10 μM BSO, either alone or transfected with siRNA against MYC.

Data are presented as mean ± standard deviation. **p < 0.01, ***p < 0.001, and ****p < 0.0001; ns indicates lack of statistical significance.

6-diazo-5-oxy-L-norleucin inhibits de novo pyrimidine nucleotide biosynthesis in GBM cells

Since our data indicated that GCLC inhibition upregulated the expression of GLS and CAD, we examined the therapeutic potential of targeting GLS and CAD in combination with GCLC in GBM cells. 6-diazo-5-oxy-L-norleucin (DON) is a potent inhibitor of glutamine-utilizing enzymes, including GLS and the carbamoyl phosphate synthetase II domain of CAD.⁴³ First, we confirmed that DON as monotherapy significantly downregulated GLS and CAD activity and inhibited the viability of GBM6, U251, and SB28 cells (Figures 6A–6C). DON did not inhibit the viability of ReN human neural stem cells or NHAs (Figure 6C). Interrogation of [U-¹³C]-glutamine metabolism showed that DON significantly downregulated oxidative metabolism to m+5 glutamate, m+5 α-KG, m+4 succinate, m+4 malate, and m+4 aspartate, consistent with inhibition of GLS activity (Figures 6D and 6E). DON also downregulated ¹³C labeling of m+4 dihydroorotate, m+3 UTP, and m+3 CTP, as expected with the inhibition of CAD activity, in both GBM6 and U251 models (Figures 6D and 6E). DON did not alter the reductive metabolism of [U-¹³C]-glutamine to m+5 citrate or m+3 AMP or GMP (Figures 6D and 6E). DON reduced steady-state pool sizes of TCA cycle metabolites and pyrimidine nucleotides in our GBM models (Figures S5A and S5B).

Figure 6. DON inhibits GLS and CAD and downregulates de novo pyrimidine nucleotide biosynthesis from [U-¹³C]-glutamine in GBM cells.

(A and B) Effect of DON on GLS activity (A) and CAD activity (B) in GBM6, U251, and SB28 cells.

(C) Dose-response curves for DON in the GBM6, U251, and SB28 models. ReN human neural stem cells and primary human astrocytes were used as controls. (D and E) Effect of DON on percentage of ¹³C metabolite enrichment from [U-¹³C]-glutamine in GBM6 (D) or U251 (E) cells.

(F) Representative histograms of live GBM6 (left), U251 (middle), and SB28 (right) cells stained with Vybrant DyeCycle green showing DNA content distribution. Cells were treated with vehicle (control), 5 μM DON (DON), or 5 μM DON + 100 μM uridine (rescue).

(G) Effect of DON on the percentage of GBM6 cells in the G1, S, and G2/M phases.

(H) Effect of DON on caspase activity in GBM6, U251, and SB28 cells.

(I and J) Bliss synergy score maps for the combination of DON and BSO in GBM6 (I) and U251 (J) cells.

(K) Effect of treatment with vehicle (control) or the combination of 1 μM each of DON and BSO (combo) on caspase activity in GBM6, U251, and SB28 cells.

(L and M) Effect of treatment with vehicle (control) or the combination of 1 μM each of DON and BSO (combo) on the percentage of ¹³C metabolite enrichment from [U-¹³C]-glutamine in GBM6 (L) or U251 (M) cells.

Data are presented as mean ± standard deviation. **p < 0.01, ***p < 0.001, and ****p < 0.0001; ns indicates lack of statistical significance. See also Figures S5 and S6.

Next, we examined the effect of DON on cell cycle progression, doubling time, and cell death in our GBM models. As shown in Figures 6F, 6G, S5C, and S5D, DON arrested GBM cells in the G1 phase of the cell cycle, which is consistent with the observed inhibition of de novo pyrimidine nucleotide biosynthesis. Concomitantly, we observed a significant increase in the doubling time of DON-treated cells (Figure S5E). Importantly, the effect of DON on cell cycle progression and doubling time was rescued by supplementation with uridine (Figures 6F, 6G, and S5C–S5E). Like BSO, treatment with DON did not induce apoptosis in GBM cells (Figure 6H). These findings suggest that DON inhibits the viability of GBM cells in a cytostatic manner by inhibiting de novo pyrimidine nucleotide biosynthesis.

The combination of BSO and DON is synthetically lethal in GBM cells

Next, we examined the effect of treating GBM cells with a combination of BSO and DON. As shown in Figures 6I, 6J, and S6A, the combination of BSO and DON was synergistically lethal, with Bliss synergy scores of 31.37, 29.19, and 31.91 for the GBM6, U251, and SB28 models, respectively (>10 indicates synergy). We then tested the effect of treatment with a combination of 1 μM each of DON and BSO (based on the dose-response matrix of the combination; see Figure S6B–S6D) on GBM cells. As shown in Figure 6K, the combination of DON and BSO induced apoptosis in GBM6, U251, and SB28 cells. Importantly, in line with our results indicating that GCLC is a specific vulnerability in TERT-dependent cancers, we confirmed that the combination of BSO and DON did not impact the viability of astrocytoma, pediatric high-grade diffuse hemispheric glioma, or osteosarcoma cells (Figure S6E).

To confirm the mechanism of action, we examined [U-¹³C]-glutamine metabolism in GBM6 and U251 cells treated with vehicle or the combination of DON and BSO. As shown in Figures 6L and 6M, the combination of DON and BSO abrogated the synthesis of m+5 GSH, m+5 glutamate, m+5 α-KG, m+4 succinate, m+4 malate, m+4 aspartate, m+4 dihydroorotate, m+3 UTP, and m+3 CTP. Taken together, these results highlight the synthetic lethality of combining BSO and DON in GBM cells.

The combination of BSO and JHU-083 induces tumor regression and extends the survival of mice bearing intracranial GBM xenografts in vivo

The use of DON has been associated with disease stabilization or remission in clinical trials in patients with cancer.⁴³ However, clinical use was halted due to issues of gastrointestinal toxicity.⁴³ JHU-083 is a novel prodrug of DON that has shown efficacy in multiple cancers, including a cytostatic effect in preclinical GBM models.^37,44 JHU-083 is also brain penetrant, which is an important consideration for GBM therapy.⁴⁴ We first confirmed that JHU-083 inhibits the viability of GBM6, U251, and SB28 cells with IC50 values of ~6–11 μM, consistent with prior studies⁴⁴ (Figure S7A). Like DON, JHU-083 did not inhibit the viability of ReN human neural stem cells or NHAs (Figure S7A). We also confirmed that JHU-083 inhibited GLS and CAD activity in the GBM6, U251, and SB28 models (Figures S7B and S7C). Importantly, the combination of JHU-083 and BSO was synergistically lethal, with Bliss synergy scores of 22.29 (GBM6), 14.34 (U251), and 23.5 (SB28) (Figures 7A–7C).

Figure 7. The combination of BSO and JHU-083 is synthetically lethal in GBMs.

(A–C) Bliss synergy score maps for the combination of JHU-083 and BSO in GBM6 (A), U251 (B), and SB28 (C) cells.

(D) Schematic illustration of study design with mice bearing intracranial GBM12 xenografts. Tumor-bearing mice were treated with vehicle, BSO (20 mg/kg), JHU-083 (20 mg/kg), or the combination of 20 mg/kg each of JHU-083 and BSO (combo) for 15 ± 2 days, and tumor tissue was resected for analysis.

(E and F) Flow cytometric quantification of the percentage of Ki67+ cells (E) and Annexin V+ cells (F) in tumor tissue from mice bearing intracranial GBM12 tumors.

(G) Schematic illustration of study design with mice bearing intracranial GBM6 xenografts. Tumor-bearing mice were treated with vehicle or the combination of 20 mg/kg each of JHU-083 and BSO (combo) for 7 days and then infused with [U-¹³C]-glutamine, and tumor tissue was resected for liquid chromatography-mass spectrometry (LC-MS).

(H) Percentage of ¹³C metabolite enrichment from [U-¹³C]-glutamine in tumor tissue from mice bearing intracranial GBM6 tumors treated with vehicle or the combination of 20 mg/kg each of JHU-083 and BSO.

(I) GCL activity, GLS activity, and CAD activity in tumor tissue from mice bearing intracranial GBM6 tumors treated with vehicle or the combination of 20 mg/kg each of JHU-083 and BSO as described in (G).

(J) Schematic illustration of in vivo efficacy study design with mice bearing intracranial GBM6 or SB28 xenografts. Tumor-bearing mice were treated with vehicle, BSO (20 mg/kg), JHU-083 (20 mg/kg), or the combination of 20 mg/kg each of JHU-083 and BSO (combo) daily for 5 days every week. Mice were treated until they needed to be euthanized or the tumor was no longer visible on MRI.

(K) Representative serial T2-weighted MRI from mice bearing intracranial GBM6 tumors treated as described in (J).

(L) Kaplan-Meier survival curves for mice bearing intracranial GBM6 tumors treated as described in (J).

(M) Representative serial T2-weighted MRI from mice bearing intracranial SB28 tumors treated as described in (J).

(N) Kaplan-Meier survival curves for mice bearing intracranial SB28 tumors treated as described in (J).

Data are presented as mean ± standard deviation. **p < 0.01, ***p < 0.001, and ****p < 0.0001; ns indicates lack of statistical significance. See also Figure S7.

To assess the potential of combination therapy in vivo, we first examined the effect of short-term treatment with BSO, DON, JHU-083, or the combination on mice bearing intracranial patient-derived GBM12 xenografts. Mice were intracranially implanted with GBM12 cells, and once tumors were visible by bioluminescence (see Figure S7D), tumor-bearing mice were treated with vehicle, BSO, JHU-083, or the combination of BSO and JHU-083 for 15 ± 2 days, and tumor tissue was resected for flow cytometric analysis of proliferation via Ki67 staining and apoptosis via Annexin V staining (see schematic in Figure 7D). As shown in Figures 7E and 7F, monotherapy with BSO and JHU-083 modestly reduced the proportion of Ki67+ cells but did not increase Annexin V+ cells. In contrast, the combination of BSO and JHU-083 induced a significant and massive increase in Annexin V+ cells with a concomitant decrease in Ki67+ cells (Figures 7E and 7F). These results are consistent with our in vitro studies and highlight the in vivo synthetic lethality of combined BSO and JHU-083 in mice bearing intracranial GBM xenografts.

Next, to confirm the mechanism of action of combination therapy in vivo, we examined the effect of treatment with vehicle or the combination of JHU-083 and BSO on the metabolism of mice bearing intracranial patient-derived GBM6 tumors. To this end, we infused mice bearing intracranial GBM6 tumors with [U-¹³C]-glutamine at day 7 after the start of treatment and quantified metabolite labeling in tumor tissue (see schematic in Figure 7G). The combination of JHU-083 and BSO caused a significant reduction in the synthesis (Figure 7H) and steady-state levels (Figure S7E) of GSH, glutamate, α-KG, succinate, malate, aspartate, dihydroorotate, UTP, and CTP in vivo. Of note, these metabolic effects were associated with an on-target reduction in GCL, GLS, and CAD activity in treated tumors relative to controls (Figure 7I).

Finally, to rigorously establish the synthetic lethality of combination therapy vs. monotherapy against well-established tumors, we examined the effect of treatment with BSO, JHU-083, or their combination on tumor growth and animal survival in mice bearing intracranial GBM6 and SB28 tumors in vivo. In both cases, once intracranial tumors were clearly detectable by magnetic resonance imaging (MRI) (~35 mm³ in volume), mice were randomized and treated with vehicle, BSO as monotherapy, JHU-083 as monotherapy, or the combination of BSO and JHU-083. Mice were treated until they needed to be euthanized per IACUC guidelines or the tumor was no longer visible on MRI (see schematic in Figure 7J). As shown in the representative serial T2-weighted MR images in Figure 7K, tumor growth curves in Figure S7F, and the survival curves in Figure 7L, while BSO and JHU-083 as monotherapy delayed tumor growth and significantly extended survival relative to vehicle-treated tumors, the combination induced tumor regression and resulted in tumor eradication in mice bearing intracranial GBM6 xenografts. With the highly aggressive SB28 model, monotherapy with BSO or JHU-083 did not attenuate tumor growth or extend survival relative to vehicle-treated tumors (Figures 7M, 7N, and S7G). Importantly, although we did not observe tumor eradication, the combination of BSO and JHU-083 induced tumor shrinkage and significantly increased animal survival in mice bearing intracranial SB28 tumor xenografts (Figures 7M, 7N, and S7G). Collectively, our studies highlight the in vivo synthetic lethality of inhibiting GCLC, GLS, and CAD using the combination of BSO and JHU-083 in preclinical GBM models in vivo.

DISCUSSION

Metabolic reprogramming and telomere maintenance are both hallmarks of cancer.⁴⁵ TERT expression is essential for telomere maintenance and replicative immortality in the majority of human cancers.^1,2 Tumors also rewire metabolism to maintain redox balance and mitigate oxidative stress.⁴⁶ By combining information from isogenic models that differ only in TERT expression or ATRX loss with loss-of-function experiments in patient-derived models, we show that TERT rewires glutamine metabolism to upregulate GSH synthesis in multiple cancers, including GBMs, oligodendrogliomas, melanomas, and hepatocellular carcinoma. While blocking GSH synthesis inhibits tumor viability in a cytostatic manner, it results in compensatory upregulation of de novo pyrimidine nucleotide biosynthesis in GBM cells. Importantly, the combined inhibition of de novo GSH and pyrimidine nucleotide biosynthesis is synthetically lethal in mice bearing intracranial GBM xenografts in vivo.

Our metabolomics, stable isotope tracing, and expression profiling studies in cells and in vivo in mice bearing intracranial xenografts show that TERT drives GSH synthesis from [U-¹³C]-glutamine by upregulating GCLC in TERT-dependent cancers. We show that TERT drives de novo GSH synthesis by relieving FOXO1-mediated repression of GCLC expression. FOXO1 binding to gene promoters can lead to the activation or repression of gene expression in a context-dependent manner.^23,24 Silencing TERT results in the inhibitory phosphorylation of FOXO1, which is known to sequester FOXO1 in the cytoplasm.⁴⁷ Indeed, ChIP-qPCR studies indicate that TERT expression is associated with the reduced binding of FOXO1 to the GCLC promoter in isogenic models and that, conversely, silencing TERT is associated with increased binding of FOXO1 to the GCLC promoter in patient-derived models. Although the precise mechanism by which TERT causes inhibitory phosphorylation of FOXO1 needs to be delineated, our studies provide a mechanistic framework that links TERT via FOXO1 to GCLC upregulation in cancer.

Our studies show that genetic ablation of GCLC or pharmacological inhibition using BSO downregulates GSH synthesis and selectively inhibits the viability of TERT-dependent cancer cells in a cytostatic manner. In contrast, we did not find any relationship between ATRX, GCLC, or GSH synthesis in ALT-dependent cancers. Mechanistic studies show that targeting GCLC induces oxidative DNA damage, arrests cells in the G1 phase of the cell cycle, and increases the doubling time of GBM cells. However, GCLC inhibition upregulates GLS and CAD and drives de novo pyrimidine nucleotide biosynthesis from [U-¹³C]-glutamine. This rewiring of glutamine metabolism is mediated by MYC, as silencing MYC abrogates the increase in GLS and CAD that is induced by GCLC inhibition in GBM cells. Collectively, our studies identify GCLC as a druggable vulnerability in the context of TERT expression in cancer and point to a hitherto unknown role for MYC in compensating for GCLC loss in TERT-driven cancer cells.

We demonstrate that targeting GLS and CAD using the glutamine antagonist DON abrogates de novo pyrimidine nucleotide biosynthesis, an effect that also inhibits the G1-S phase transition and increases the doubling time of GBM cells. Like BSO, DON as monotherapy inhibits viability in a cytostatic manner. Importantly, the combination of BSO and DON (or its prodrug JHU-083) is synergistically lethal in vitro and induces tumor regression in mice bearing intracranial GBM xenografts in vivo. Of note, prior studies have demonstrated an inhibitory effect of BSO or DON on tumor metabolism and/or growth, independent of the tumor genotype.^44,48 Nevertheless, our studies are the first to show that GCLC is a druggable vulnerability in the context of TERT expression and provide a rationale for utilizing DON/JHU-083 in the context of GCLC inhibition in TERT-driven cancer cells.

Our short-term experiments with mice bearing intracranial GBM12 xenografts indicate that monotherapy with BSO or JHU-083 inhibits proliferation but does not induce cell death. In contrast, the combination of BSO and JHU-083 induced a significant and massive increase in apoptotic cells with a concomitant decrease in tumor proliferation. These results are consistent with our in vitro studies and serve to confirm the in vivo synthetic lethality of combined BSO and JHU-083 in mice bearing intracranial GBM xenografts. Importantly, the combination of BSO and JHU-083 induces tumor regression and significantly extends the survival of mice bearing intracranial tumors for both the patient-derived GBM6 and the murine SB28 models. It should be noted that the SB28 model is highly aggressive and resistant to most therapies, including standard-of-care radiation and temozolomide.^32,33 Therefore, our findings are significant and serve to identify the combination of BSO and JHU-083 as a novel therapeutic strategy for GBMs.

Regarding clinical translation, BSO is limited by poor brain penetrance and pharmacokinetic properties.⁴⁹ Our studies provide a rational impetus for the development of novel GCLC inhibitors with improved pharmacokinetic properties. JHU-083 is a prodrug of DON that was designed to cross the blood-brain barrier and get selectively activated within the tumor, leading to the inhibition of glutamine-utilizing enzymes, including GLS and CAD.^37,43,44 Although JHU-083 is rapidly converted to DON in mouse plasma, due to the high esterase activity in rodents, large animals, including humans, are known to have lower esterase activity,⁵⁰ which makes JHU-083 a more clinically relevant candidate than DON. Of note, a related DON prodrug (DRP-104) that was specifically designed for solid body tumors is in clinical testing (ClinicalTrials.gov: NCT06027086).⁵¹

In summary, our studies leverage a mechanistic understanding of TERT-associated metabolic rewiring for the identification of synthetically lethal metabolic vulnerabilities in GBMs. Clinical translation of our results has the potential to enable the rational development of novel combination therapies for patients with GBM.

Limitations of the study

Our results indicate that both DON and JHU-083 inhibit GLS and CAD activity in GBM cells. Consistent with these results, treatment with DON as monotherapy reduced ¹³C labeling of TCA cycle metabolites and the pyrimidine nucleotides UTP and CTP. Interestingly, despite reports suggesting that DON inhibits the purine biosynthesis enzymes formylglycinamide ribonucleotide amidotransferase and GMP synthetase,⁴³ we did not observe inhibition of purine nucleotide (AMP and GMP) biosynthesis following treatment with DON or JHU-083 in our models. It is possible that differences in the affinity of DON/JHU-083 to different glutamine-utilizing enzymes result in context-dependent differences in the metabolic consequences of DON/JHU-083 treatment. Regarding our in vivo efficacy studies, we find that the combination of BSO and JHU-083 significantly extends survival in both the patient-derived GBM6 and murine SB28 models. Combination therapy resulted in complete tumor eradication in all treated mice in the patient-derived GBM6 model. In contrast, while it induced tumor shrinkage, the combination did not result in tumor eradication in the SB28 model. It is possible that the tumor microenvironment limits the efficacy of combination therapy in the SB28 model. Additional studies in a larger cohort of syngeneic models would overcome this limitation.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Pavithra Viswanath (Pavithra.Viswanath@ucsf.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• The raw data, including metabolomics data, generated in this manuscript have been deposited in Dryad (https://doi.org/10.5061/dryad.bcc2fqzq).

• This manuscript does not report any original code.

• Any additional information required to reanalyze the data reported in this manuscript is available from the lead contact upon reasonable request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cell culture

Immortalized p53/pRb-deficient normal human astrocytes (NHA_CONTROL) have been described previously.^52,53 To generate NHA_TERT+ cells, immortalized normal human astrocytes were seeded into 10 cm dishes and transfected with pCDNA-3xHA-hTERT (Addgene, plasmid #51631) using Lipofectamine 3000. TERT-expressing cells were selected using G418. Expression was confirmed using QPCR as described below. To generate NHA_ATRX-KO cells, immortalized normal human astrocytes were seeded in 10 cm dishes and transfected with ATRX shRNA against ATRX using polybrene and selected using puromycin. GBM6, GBM12, and U251 cells were isolated from isocitrate dehydrogenase wild-type glioblastoma male patients as previously described.^54,55 GBM12 cells were transduced with lentiviral particles expressing firefly luciferase and an eGFP cassette (Luc-P2A-eGFP) under the control of a CMV promoter. Luciferase-expressing cells were selected using puromycin. SF10417 cells were isolated from male patients harboring isocitrate dehydrogenase mutant oligodendrogliomas.^56,57 A375 (female) and HepG2 (male) cells were a kind gift from Dr. Joseph Costello and have been described earlier.⁵⁸ BT142 (male) cells were isolated from a patient with an isocitrate dehydrogenase mutant astrocytoma.⁵⁹ KNS42 (male) cells were isolated from a patient with diffuse hemispheric glioma.⁶⁰ U2OS (female) cells were isolated from a patient with osteosarcoma.⁶¹ ReN human neural progenitor cells were purchased from Sigma.

NHA_TERT, NHA_CONTROL, NHA_ATRX-KO, U251, SB28, A375, HepG2, KNS42, and U2OS cells were grown as monolayers in DMEM supplemented with 10% fetal bovine serum, 2 mM L-glutamine, and 100 U/ml each of penicillin and streptomycin. SF10417 cells were grown as monolayers in laminin-coated flasks in DMEM/F-12 supplemented with 1X B-27,1X N-2, and 20 ng/mL each of human EGF and FGF-basic, 2 mM L-glutamine, and 100 U/ml each of penicillin and streptomycin. BT142 cells were grown as neurospheres in serum-free Neurobasal medium supplemented with 20 ng/mL EGF, 20 ng/mL FGF-basic, 2 mM L-glutamine, and 100 U/ml each of penicillin and streptomycin. ReN human neural progenitor cells were purchased from Sigma and maintained in ReNcell NSC Maintenance Media according to the manufacturer’s instructions. Cell lines were routinely tested for mycoplasma contamination, authenticated by short tandem repeat fingerprinting, and assayed within 6 months of authentication.

Biopsies

Patient biopsies were obtained from the UCSF Brain Tumor Center Biorepository in compliance with the written informed consent policy.^16–18 Biopsy use was approved by the Committee on Human Research at UCSF and research was approved by the Institutional Review Board at UCSF according to ethical guidelines established by the U.S. Common Rule.

Animal studies

Efficacy studies:

Animal studies were conducted in accordance with UCSF Institutional Animal Care and Use Committee guidelines. Animals were housed and cared for according to standard guidelines with free access to food and water. 3×10⁵ GBM6 and GBM12 cells were intracranially injected into female SCID mice. 5×10⁴ SB28 cells into female C57BL/6 mice. Intracranial implantation was performed using a stereotactic frame at 2 mm to the right of the medial suture, 2 mm behind the bregma, and a depth of 2 mm. For the GBM12 model, mice were intracranially implanted with GBM12 cells and monitored by bioluminescence until tumors reached a radiance of ~1×10⁶ p/s/cm²/sr. This timepoint was considered day 0 (D0) and mice were randomized and treated with vehicle (saline), BSO (20 mg/kg), JHU-083 (20 mg/kg), or their combination via intraperitoneal injection. Mice were treated daily for 5 days/week. On days 15 ± 2, mice were euthanized, and tumor tissue was resected for flow cytometric analysis. For the GBM6 and SB28 models, mice were intracranially implanted with tumor cells, and tumor growth was monitored by T2-weighted MRI. Once tumors reached a volume of 30 ± 5 mm³, this timepoint was considered D0, and mice were randomized and treated with vehicle (saline), BSO (20 mg/kg), JHU-083 (20 mg/kg), or their combination via intraperitoneal injection. These doses were selected based on prior studies in preclinical cancer models.^31,44 Mice were treated until they needed to be euthanized, or the tumor was no longer visible on MRI. Animal survival was assessed by Kaplan-Meier analysis.

Stable isotope tracing:

Stable isotope tracing experiments in vivo were performed using a previously established protocol³⁸ on mice bearing intracranial GBM6 tumors treated with vehicle (saline) or 20 mg/kg each of JHU-083 and BSO as described above. At day 7 post-treatment, mice were intravenously infused with [U-¹³C]-glutamine under anesthesia on a heating pad: a bolus of 428 mg/kg of [U-¹³C]-glutamine diluted in 0.18 mL in saline was injected within 1 min, and then, 14 mg/kg/min was infused for 2 h. At the end of the infusion, brain (tumor and contralateral) tissue was collected and snap frozen. ~15–25 mg of each tissue sample was homogenized in 1 mL of pre-cooled methanol/water (80:20 v/v) and centrifuged at 14,000 rpm for 10 min. The supernatant was then lyophilized and used for LC-MS as described below.

METHOD DETAILS

Silencing and overexpression studies:

Two non-overlapping siGENOME or Accell siRNA sequences (Dharmacon) against TERT, GCLC, MYC, or non-targeting siRNA were used to silence gene expression. 3 million cells were seeded in a T25 flask and transfected with 25 nM siRNA using Lipofectamine 3000 at 24 h. Cells were incubated at 37°C for an additional 48 h prior to harvesting for further studies. For exogenous expression of CA-FOXO1, TERT-expressing cells were transiently transfected with human CA-FOXO1 (pcDNA3 Flag-FKHR-AAA mutant; Addgene).¹⁷ CA-FOXO1 expression was verified at 72 h by western blotting for the FLAG tag.

QPCR:

Gene expression was measured by QPCR according to standard protocols. Briefly, RNA was isolated using the RNeasy Kit (Qiagen) according to the manufacturer’s protocol. The SYBR Green quantitative PCR kit (Sigma) was used to perform quantitative real-time PCR using SYBR green fluorescence on 10 ng of cDNA template in a final volume of 20 μL. Data analysis was performed using the DDCt method with β-actin as a reference/control gene.

ChIP-QPCR:

Chromatin was immunoprecipitated using an antibody specific to FOXO1 (Cell Signaling, #2880) or rabbit IgG (Cell Signaling, #2729) and the High-Sensitivity ChIP Kit (ab185913) according to the manufacturer’s instructions. QPCR was then performed for GCLC using the primers described above in the key resources table. Data was expressed as fold enrichment relative to IgG control.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
GCLC	Cell Signaling	48005; RRID: AB_3101780
FOXO1	Cell Signaling	2880; RRID: AB_2106495
Phospho-FOXO1	Cell Signaling	84192; RRID: AB_2800035
FLAG	Cell Signaling	14793; RRID: AB_2572291
GLS	Abcam	ab156876; RRID: AB_2721038
CAD	Cell Signaling	93925; RRID: AB_2750933
Rabbit IgG	Cell Signaling	2729; RRID: AB_1031062
β-actin	Cell Signaling	4970; RRID: AB_2223172
Ki67, PE conjugate	Invitrogen	12-5698-82; RRID: AB_11150954
Annexin V, PE conjugate	BD Biosciences	563795; RRID: AB_2869071
Biological samples
GBM, astrocytoma, and gliosis biopsies	UCSF Brain Tumor Center Repository	N/A
Chemicals, peptides, and recombinant proteins
DMEM	Gibco	11965–092
DMEM/F-12	Gibco	11330032
Neurobasal™-A Medium	Gibco	10888022
B27	Gibco	12587010
N2	Gibco	17502001
Fetal Bovine Serum	Gibco	10438–026
L-Glutamine	Gibco	25030–081
Antibiotics	Gibco	15240–062
ReNcell NSC Maintenance Media	Sigma-Aldrich	SCM005
Human EGF	Gibco	AF-100–15
Human FGF-basic	Gibco	AF-100–18B
Laminin	Corning	354239
G418	Sigma-Aldrich	G1279
Puromycin	Sigma-Aldrich	P8833
Lipofectamine 3000	Gibco	L3000001
Polybrene	Sigma-Aldrich	TR-1003
RIPA buffer	Thermo Scientific	89901
Halt™ Protease Inhibitor Cocktail	Thermo Scientific	87785
2X Laemmli SDS sample buffer	Biorad	1610737
4–20% polyacrylamide gel	Biorad	4561096
Precision Plus Protein Ladder	Biorad	1610374
10x Tris/Glycine/SDS running buffer	Biorad	1610732
Immobilon® -FL PVDF Membrane	Sigma-Aldrich	IPFL00010
Blot Absorbent Filter Paper	Biorad	1703965
25X Tris-glycine transfer buffer	Thermo Scientific	LC3675
Nonfat dried milk	Sigma-Aldrich	M7409
Immobilon ECL Ultra Western HRP Substrate	Sigma-Aldrich	WBULS0100
BSO	Sigma-Aldrich	B2515
DON	Cayman Chemicals	17580
JHU-083	Selleckchem	S8891
DMSO	Sigma-Aldrich	D2650
Vybrant™ DyeCycle™ Green Stain	Thermo Scientific	V35004
[U-13C]-glutamine	Cambridge Isotope Laboratories	CLM-1822-H-PK
2,3-naphthalenedicarboxyaldehyde	Sigma-Aldrich	70215
BD Pharmingen™ Annexin V Binding Buffer, 10X concentrate	BD Biosciences	556454
BD Cytofix/Cytoperm™ Fixation and Permeabilization Solution	BD Biosciences	554722
D-luciferin	Gold Biotechnology	LUCK-100
Uridine	Sigma-Aldrich	U3750
Glutathione ethyl ester	Cayman Chemicals	14953
Critical commercial assays
SYBR® Green Quantitative RT-qPCR Kit	Sigma-Aldrich	QR0100
High-Sensitivity ChIP Kit	Abcam	ab185913
2′,7′-dichlorofluorescin diacetate (DCFDA) assay kit	Abcam	ab113851
Glutathione reductase activity assay kit	Abcam	ab83461
Glutaminase activity assay kit	Abcam	ab284547
Caspase activity assay kit	Abcam	ab39401
Malondialdehyde assay kit	Abcam	ab118970
FOXO1 transcription factor activity assay kit	Abcam	ab207204
RealTime-Glo™ MT Cell Viability Assay	Promega	G9711
Tumor dissociation kit	Miltenyi Biotec	130-095-929
Mouse cell depletion kit	Miltenyi Biotec	130-104-694
Luna 3 μm column	Phenomenex	PRD-756484
8-OHdG DNA Damage ELISA	Cell Biolabs	STA-320-T
Deposited data
Raw data	This paper	https://doi.org/10.5061/dryad.bcc2fqzq
Metabolomics data	This paper	https://doi.org/10.5061/dryad.bcc2fqzq
Experimental models: Cell lines
NHACONTROL	Dr. Russell Pieper	N/A
NHATERT	Dr. Russell Pieper	N/A
NHAATRX-KO	This paper	N/A
GBM6	Dr. Jann Sarkaria	N/A
GBM12	Dr. Jann Sarkaria	N/A
U251	Dr. Joseph Costello	N/A
SB28	Dr. Hideho Okada	N/A
SF10417	Dr. Joseph Costello	N/A
A375	Dr. Joseph Costello	N/A
HepG2	Dr. Joseph Costello	N/A
BT142	ATCC	ACS-1018
KNS42	JCRB Cell Bank	IFO50356
U2OS	ATCC	HTB-96
ReN human neural progenitor cells	Sigma-Aldrich	SCC007
Primary human astrocytes	Celprogen	36058
Experimental models: Organisms/strains
Fox Chase SCID mice	Charles River	236
C57BL6/J mice	Charles River	027
Oligonucleotides
Human TERT siRNA	Dharmacon	D-003547-02, D-003547-03
Mouse TERT siRNA	Dharmacon	E-48320-00, M-048320-00
Human GCLC siRNA	Dharmacon	A-009212-15, A-009212-16
Human MYC siRNA	Dharmacon	D-003282-14, D-003282-15
Mouse MYC siRNA	Dharmacon	E-40813-00, M-040813-02
Non-targeting siRNA (mouse and human)	Dharmacon	D-001206-14
Human TERT primers	Origene	HP230441
Mouse TERT primers	Origene	MP216756
Human FOXO1 primers	Origene	HP205770
Mouse FOXO1 primers	Origene	MP204948
Human GCLC primers	Origene	HP205366
Mouse GCLC primers	Origene	MP205545
Human GCLM primers	Origene	HP205810
Mouse GCLM primers	Origene	MP205544
Human GS primers	Origene	HP200163
Human GLS primers	Origene	HP211243
Mouse GLS primers	GeneCopoeia	MQP054544-GC
Human CAD primers	Origene	HP207669
Mouse CAD primers	Origene	MP201690
Human β-actin primers	Origene	HP204660
Mouse β-actin primers	Origene	MP200232
ATRX shRNA (GATCCCCGAGGAAACCTTCAATTGTATTCAAGAGATACAATTGAAGGTTTCCTCTTTTTA)	IDT	N/A
Recombinant DNA
pcDNA3 Flag-FKHR-AAA mutant	Addgene	13508
pCDNA-3xHA-hTERT	Addgene	51631
Software and algorithms
SynergyFinder web application (version 3.0)	https://synergyfinder.fimm.fi/synergy/20241116012111964438/	N/A
GraphPad Prism 10	GraphPad Prism	N/A
XCalibur	Thermo Scientific	N/A
Freestyle	Thermo Scientific	N/A
TraceFinder	Thermo Scientific	N/A
Other
Chemisolo western blot imaging system	Azure Biosystems	CS1000

Western blotting:

Cells (~10⁷) were lysed by sonication in RIPA buffer containing protease inhibitors. Lysates were cleared by centrifugation at 14000 rpm for 15 min at 4°C and boiled in SDS-PAGE sample buffer (95°C for 10 min). Total cellular protein (~20 μg) was separated on a 4–20% polyacrylamide gel by sodium dodecyl sulfate polyacrylamide gel electrophoresis and transferred onto Immobilon-FL PVDF membrane. Membranes were blocked overnight in 5% milk in Tris-buffered saline containing Tween 20 (TBST: 20 mM Tris-HCl, pH 7.5, 500 mM NaCl, 0.1% Tween 20) at 4°C. Membranes were then washed 3 times for 5 min each in TBST and incubated with primary antibodies diluted in TBST for 1 h at room temperature. Following 3 washes of 10 min each with TBST, HRP-conjugated secondary antibodies were added for 1 h in TBST at room temperature. Membranes were washed thrice in TBST for 10 min each and developed using an enhanced chemiluminescence substrate. Western blot images were acquired using an automated Chemsolo imager.

Activity assays:

ROS levels were assessed using a 2′,7′-dichlorofluorescin diacetate (DCFDA)-based assay kit (Abcam, #ab113851). 8-OHdG levels were measured by ELISA using a kit (Cell Biolabs, #STA-320-T) according to the manufacturer’s instructions. Glutathione reductase activity (Abcam, #ab83461), glutaminase activity (Abcam, #284547), and caspase activity (Abcam, #ab39401) were measured using kits according to the manufacturer’s instructions. Lipid peroxidation was assayed by measuring malondialdehyde levels using a kit (Abcam, Ab118970). FOXO1 transcription factor activity was measured using a kit (Abcam, #ab207204). Briefly, the assay measures the binding of active FOXO1 to a DNA sequence containing the FOXO1 binding site. FOXO1 is detected using a primary antibody that recognizes an epitope of FOXO1 accessible only when the protein is active and bound to its target DNA. GCL activity was measured as described previously.⁶² Cells or tissue were lysed in buffer containing 20 mM Tris-HCl, 1 mM EDTA, 250 mM sucrose, 20 mM sodium borate, and 2 mM serine buffer and added to a reaction mixture (400 mM Tris-HCl, 40 mM adenosine triphosphate, 40 mM L-glutamic acid, 30 mM cysteine, 2 mM EDTA, 20 mM sodium borate, 2 mM serine, 40 mM MgCl₂). After incubation at 37°C for 15 min, the enzymatic reaction was stopped by precipitation of proteins with 200 mM 5-sulfosalicylic acid. Following incubation with 2,3-naphthalenedicarboxyaldehyde (NDA) in the dark at room temperature for 30 min, the product (γ-glutamylcysteine-NDA) was detected by fluorimetry (472 nm excitation/528 nm emission). Data was compared to a standard curve and expressed as mmoles per mg protein. The carbamoyl phosphate synthetase II activity of the multifunctional CAD enzyme was measured as described previously.^63,64 A two-step assay was used to measure carbamoyl phosphate synthesis by coupling the carbamoyl phosphate synthetase reaction to that of ornithine transcarbamoylase and quantifying the product i.e., citrulline. Specifically, cell lysate was added to a reaction mixture (50 mm HEPES, 100 mm KCl, 10 mm ATP, 20 mm MgCl₂, 20 mm NaHCO₃, 1 mm dithiothreitol, 5 mm ornithine, 0.2 units of ornithine transcarbamoylase, and 10 mm glutamine, pH 7.6) and incubated at 37°C for 20 min. After incubation, citrulline was quantified by colorimetry with diacetylmonoxime at 490 nm as previously described.⁶⁴

Flow cytometry:

Cells were seeded in 6-well plates at a density of 3×10⁵ cells/well and treated with vehicle (DMSO), BSO (10 μM), or DON (5 μM). Alternately, cells were treated with non-targeting or GCLC-targeted siRNA to silence GCLC expression. To assess whether the effects of BSO are mediated via loss of GSH, we examined the ability of glutathione ethyl ester (100 μM) to rescue the phenotype. To assess whether the effects of DON are mediated via loss of pyrimidine nucleotide biosynthesis, we examined the ability of exogenous uridine (100 μM) to rescue the phenotype. Cell cycle progression was measured by staining live cells for DNA content using the Vybrant DyeCycle Green stain (Thermo Scientific). 1×10⁶ cells were stained with 10 μM Vybrant DyeCycle Green and incubated in the dark at 37°C for 30 min. Cells were analyzed by flow cytometry on a MACSQuant 10 analyzer with excitation at 488 nm and emission at 525 nm. To measure doubling time, cells were seeded in a 96-well plate at 10,000 cells per well and treated as described above. After 4 days, cell number was measured using the MACSQuant 10 analyzer. The doubling time was calculated using the following formula: doubling time = [4 days × (ln2)]/[ln (day 4 cell count/day 0 cell count)]. To quantify proliferation in GBM12 tumors, tumor tissue was dissociated into single cells using the tumor dissociation kit (Miltenyi Biotec) according to the manufacturer’s instructions. Mouse cells were then depleted using the mouse cell depletion kit (Miltenyi Biotec) according to the manufacturer’s instructions. Cells were fixed and permeabilized by resuspension in the BD Cytofix/Cytoperm Fixation and Permeabilization Solution followed by incubation in the dark at 4°C for 20 min. Cells were washed and intracellularly stained with 0.03 μg/mL anti-Ki67-PE antibody for 20 min at room temperature. The % of Ki67+ cells was measured using a MACSQuant 10 analyzer with excitation at 488 nm and emission at 585 nm. To quantify apoptosis in GBM12 tumors, tumor tissue was dissociated into single cells and depleted of mouse cells as described above. Single cells were resuspended in annexin binding buffer and stained with a 1:50 dilution of the anti-annexin V-PE antibody for 30 min in the dark at room temperature. Cells were washed once in the annexin binding buffer and the % of annexin V+ cells measured using a MACSQuant 10 analyzer with excitation at 488 nm and emission at 585 nm.

IC50:

Live cell viability was measured using the RealTime-Glo Assay (Promega). Briefly, 10,000 cells/well were seeded in a 96-well plate and treated with vehicle (dimethyl sulfoxide, DMSO) or a 10-point serial dilution of BSO, DON, or JHU-083 for 48 h. Alternately, cells were treated with non-targeting or GCLC siRNA (25 nM) for 48 h. Following the addition of the MT Cell Viability Substrate and NanoLuc Enzyme, the luminescence signal was measured. IC50 was calculated by non-linear regression. To assess synergy, 10,000 cells/well were seeded in a 96-well plate and treated with a 5-point serial dilution of DON or JHU-083 and BSO. Live cell viability was measured as described above using the RealTime-Glo Assay. Bliss synergy scores were calculated using the SynergyFinder web application (version 3.0).⁶⁵

Metabolomics and stable isotope tracing in cells:

~5×10⁵ cells were seeded in 6-well plates and treated with vehicle (DMSO), DON (5 μM), BSO (10 μM), or siRNA (25 nM) in regular cell culture media for 48 h (drug treatment) or 72 h (siRNA). For combination therapy, cells were treated with 1 μM each of DON (or JHU-083) and BSO for 72 h. For stable isotope tracing, cells were seeded and treated as above and incubated in media in which glutamine was replaced with [U-¹³C]-glutamine (99% enrichment; Cambridge Isotope Laboratories, final concentration 5.7 mM). Following incubation for 48 or 72 h, cells were washed with ice-cold ammonium acetate (150 mM, pH 7.3). 1 mL of pre-cooled methanol/water (80:20 v/v) was added to each well and plates incubated at −80°C for 30 min. Cells were collected by scraping and centrifuged at 14,000 rpm for 15 min at 4°C to remove debris. The supernatant was lyophilized, samples were reconstituted with 60 μL of pre-chilled acetonitrile/water (50:50, v/v), transferred into glass vials and utilized for liquid chromatography mass spectrometry (LC-MS) as described below.

LC-MS:

LC-MS was performed using a Vanquish Ultra High-performance LC system coupled to an Orbitrap ID-X Tribrid mass spectrometer (Thermo Scientific), equipped with a heated electrospray ionization (H-ESI) source capable of both positive and negative modes simultaneously.⁶⁶ Before analysis, the MS instrument was calibrated using a calibration solution (FlexMix, Thermo Fisher). Cell or tissue samples, along with blank controls, were placed in the autosampler. Individual samples were run alongside a pooled sample made from an equal mixture of all individual samples to ensure chromatographic consistency. Chromatographic separation of metabolites was achieved by hydrophilic interaction liquid chromatography (HILIC) using a Luna 3 NH2 column (150 mm × 2.1 mm, 3 μm, Phenomenex) in conjunction with a HILIC guard column (Phenomenex, 2.1 mm). The column temperature and flow rate were maintained at 27°C and 0.2 mL/min respectively. Mobile phases consisted of A (5 mM ammonium acetate, 48.5 mM ammonium hydroxide pH 9.9) and B (100% Acetonitrile). The following linear gradient was applied: 0.0–0.1 min: 85–80% B, 0.1–17.0 min: 80–5% B, 17.0–24.0 min: 5% B, 24.0–25.0 min: 5–85% B, 25.0–36.0 min: 85% B. The injection volume and auto sampler temperature were kept at 5 μL and 4°C respectively. High-resolution MS was acquired using a full scan method alternating between positive and negative polarities (spray voltages: +3800kV/−3100kV; sheath gas flow: 45 arbitrary units: auxiliary gas flow: 15 arbitrary units; sweep gas flow: 1 arbitrary unit; ion transfer tube temperature: 275°C; vaporizer temperature: 300°C). Three mass scan events were set for the duration of the 36-min run time. The first was the negative polarity mass scan settings at full-scan-range; 70–975 m/z. Positive polarity mass scan settings were split to two scan events, full-scan-range; 70–360 m/z and 360–1500m/z, in that order. MS1 data were acquired at resolution of 60,000 with a standard automatic gain control and a maximum injection time of 100 ms. Data was acquired using Xcalibur software. Chromatograms were reviewed using FreeStyle (Thermo Fisher) and a 5 ppm mass tolerance. Peak areas were quantified using TraceFinder from either positive or negative modes depending on previously run standards. Peak areas were corrected to blank samples and normalized to cell number or wet weight of tissue and the total ion count of that sample. For ¹³C isotope tracing, % enrichment for each isotopomer was calculated after correcting for natural abundance using Escher Trace.⁶⁷

MRI:

Tumor volume was determined by T2-weighted MRI using a small animal horizontal Bruker 3T (GBM6 studies) or 9.4T (SB28 studies) scanner. For the 3T scanner, we used a ¹H quadrature volume coil and a T2 rapid acquisition with relaxation enhancement (RARE) sequence (TE/TR = 64/3700 ms, FOV = 30 × 30 mm², matrix = 256 × 256, slice thickness = 1.5 mm, NA = 5). Studies at 9.4T were performed using a ¹H/¹³C linear/linear volume coil on a preclinical 9.4T MR scanner (Biospec, Bruker). Axial T2-weighted images were acquired using a spin-echo TurboRARE sequence (TE/TR = 8.25/3200ms, FOV = 30 × 30mm², 256 × 256, slice thickness = 1.2mm, NA = 3).

Bioluminescence:

Tumor growth was monitored by bioluminescence on a Xenogen IVIS Spectrum 10 min after intraperitoneal injection of 150 mg/kg D-luciferin prepared according to the manufacturer’s directions. The total flux signal was used for quantification in all studies.

QUANTIFICATION AND STATISTICAL ANALYSIS

All experiments were performed on a minimum of 3 biological replicates (n ≥ 3) and results presented as mean ± standard deviation. Statistical significance was assessed in GraphPad Prism 10 using a two-way ANOVA or two-tailed Welch’s t test with p < 0.05 considered significant. Analyses were corrected for multiple comparisons using Tukey’s method, wherever applicable. * = p < 0.05, ** = p < 0.01, *** = p < 0.001 and **** = p < 0.0001. Details on the statistical methods are included in the figure legends.

Supplementary Material

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2025.115596.

Highlights.

• TERT upregulates GCLC and drives GSH synthesis in glioblastoma cells

• Targeting GCLC inhibits viability but causes compensatory upregulation of GLS and CAD

• Combined inhibition of GCLC, GLS, and CAD is synthetically lethal in glioblastomas in vivo