Therapeutic targeting of differentiation-state dependent metabolic vulnerabilities in diffuse midline glioma

Abstract

H3K27M diffuse midline gliomas (DMG), including diffuse intrinsic pontine gliomas (DIPG), exhibit cellular heterogeneity comprising less-differentiated oligodendrocyte precursors (OPC)-like stem cells and more differentiated astrocyte (AC)-like cells. Here, we establish in vitro models that recapitulate DMG-OPC-like and AC-like phenotypes and perform transcriptomics, metabolomics, and bioenergetic profiling to identify metabolic programs in the different cellular states. We then define strategies to target metabolic vulnerabilities within specific tumor populations. We show that AC-like cells exhibit a mesenchymal phenotype and are sensitized to ferroptotic cell death. In contrast, OPC-like cells upregulate cholesterol biosynthesis, have diminished mitochondrial oxidative phosphorylation (OXPHOS), and are accordingly more sensitive to statins and OXPHOS inhibitors. Additionally, statins and OXPHOS inhibitors show efficacy and extend survival in preclinical orthotopic models established with stem-like H3K27M DMG cells. Together, this study demonstrates that cellular subtypes within DMGs harbor distinct metabolic vulnerabilities that can be uniquely and selectively targeted for therapeutic gain.

Fig. 1. In vitro models of H3K27M DIPG molecularly mimic oligodendrocyte precursor (OPC)-like and differentiated astrocyte (AC)-like DIPG and exhibit distinct gene expression programs.

A Schematic depicting the generation of DIPG gliomaspheres (GS) and differentiated glioma cells (DGC): GS (3-D floating spheres) were cultured in serum-free tumor stem cell media containing growth factors and supplements. Monolayer (adherent) DGC were generated by dissociating GS to single cells and culturing for up to 14 days in media containing 10% fetal bovine serum (FBS). B Growth kinetics of DIPG-007, SF7761 and DIPG-XIII GS vs. DGC. Error bars are mean ± SD from three biological replicates; ns (not significant) p = 0.7589, ****p < 0.0001 by area under proliferation curve test followed by two-tailed Student’s t-test. C, D Venn diagrams derived using DESeq2 package in R with adjustment for multiple comparisons illustrating the overlap of upregulated (C) and downregulated (D) genes in GS vs. DGC across three cell lines. E Western blot for OLIG2 in GS and DGC pairs of DIPG cell lines. HSP90 was used as a loading control. The figure shows representative blots from two independent experiments. F Analysis of the gene expression signature of DIPG GS vs. DGC cross-referenced with patient gene signature capturing OPC-like, AC-like gene signatures. G Differential expression of DIPG genes in OPC-like and AC-like tumors from patient single-cell RNAseq. F, G) Data was analyzed using unpaired, two-sided, two-tailed, Students t-test with 95% confidence intervals *p < 0.05; **p < 0.01; ***p < 0.001. H Survival analysis of DIPG/DMG patients (n = 78; female = 41, male = 28, unknown = 9. Age range was 1.2–17.9 years (average 9.8 ± 12 years) based on “GS high” versus “GS low” gene signature. Median survival is 11.9 months versus 8 months; Log Rank p = 0.0176. I, J “Hallmark” gene set enrichment analysis (GSEA) indicating pathways that are (I) upregulated and (J) downregulated in DIPG GS vs. DGC. GSEA plots show enrichment scores and include values for normalized enrichment score (NES), nominal p-value (P), and false discovery rate (FDR) q-value. For all panels, red indicates GS; blue, DGC.

Fig. 2. Metabolomic profiling of gliomaspheres (GS) and differentiated glioma cells (DGC) reveals differentiation state-dependent metabolic features.

A Volcano plot indicating differential metabolite profiles in GS vs. DGC, presented as the average metabolite abundance value from the three cell lines DIPG-007, SF7761, and DIPG-XIII. B Metabolic pathway enrichment as determined using MetaboAnalyst based on differential metabolite abundance in GS vs. DGC. A, B Statistically significant differential metabolites evaluated using t-test with two-tailed distribution and two-sample unequal variance based on p-values from 3 independently prepared samples. Dot size reflects pathway impact. C Venn diagram indicating the number of metabolites significantly altered in GS vs. DGC across all three cell lines. Example metabolites and pathways are indicated below. D–F Differential abundance of select metabolites for (D) glycolysis, (E) tricarboxylic acid (TCA) cycle, and (F) purine nucleotides in DIPG-007, SF7761, and DIPG-XIII GS versus DGC. G Highly enriched metabolites in DIPG DGC vs. GS. D–G Metabolite levels presented as median-centered fold change of GS relative to DGC across the three cell lines. Error bars represent mean ± SD. All metabolomic profiling data (panels A–G) were generated from the average of three independently prepared samples run on the same day. Source data are provided as Source data file 2. F6P, fructose-6-phosphate; F1,6BP, fructose-1-6-bisphosphate; G3P, glyceraldehyde-3-phosphate; DHAP, dihydroxyacetone phosphate; PEP, phosphoenolpyruvate, PYR, pyruvate; LAC, lactate; Gln, glutamine; Glu, glutamate; Cit, citrate; Iso-Cit, isocitrate; Aconi, cis-aconitate; α-KG, alpha-ketoglutarate; Suc, succinate; Mal, malate; ADP, adenosine diphosphate; GDP, guanosine diphosphate; AMP, adenosine monophosphate; IMP, inosine monophosphate; IDP, inosine diphosphate; Hydro-Pro, hydroxyproline. For panels A, C–G, red indicates GS; blue, DGC.

Fig. 3. Ferroptosis is a metabolic vulnerability of differentiated glioma cells (DGC).

A Differential expression of epithelial-mesenchymal transition (EMT) genes by bulk transcriptomics in DIPG-007, SF7761, and DIPG-XIII gliomaspheres (GS) vs. DGC. Data presented from three biological replicates. B Simplified scheme depicting the role of glutathione peroxidase 4 (GPX4) in ferroptosis. C, D RSL3 dose-response in (C) DGC and (D) GS DIPG-007, SF7761, and DIPG-XIII with or without 1 h ferrostatin-1 (Fer-1) pre-treatment. E Flow cytometry assessment of intracellular lipid reactive oxygen species (ROS) using C-11 BODIPY in DIPG-007 DGC treated with RSL3 for 6 h with or without 1 h Fer-1 pre-treatment. Data expressed as mean ± SD mean fluorescent intensity (MFI) (****p < 0.0001; multiple unpaired t test). F Cell viability of RSL3-treated DIPG-007 DGC in the presence of bafilomycin A1 (Baf-1), necrosulfonamide (NSA), ZVAD-FMK (Z-VAD), Fer-1, or Trolox. G RSL3 dose response in DIPG-007 GS cultured in antioxidant-free B-27-supplemented GS, DGC, or reduced-serum (2.5%) DGC media, with or without 1 h Fer-1 pre-treatment. Cell viability (C, D, G) assessed with Cell Titer-Glo 2.0 (DGC) or Cell Titer-Glo 3D (GS) at 48 h and data expressed as percent vehicle control (0.1% DMSO). Error bars represent mean ± SD from three biological replicates with ns = not significant, *p < 0.05, **p < 0.01, ****p < 0.0001 by area under proliferation curve test followed by either two-tailed Student’s t-test) (C, D) or one-way ANOVA for multiple Uncorrected Fisher’s LSD test (G). For panels E and F, error bars represent mean ± SD from three biological replicates with nd = no discovery, *p < 0.00001 by multiple unpaired t-test and Two-stage step-up. Panels C–G are representative of data from three biological replicates. TGFB2, transforming growth factor beta-2; VCAM1, vascular cell adhesion molecule 1; TAGLN, transgelin; ACTA2, actin alpha 2; COL11A1, collagen type XI alpha 1 chain; SNAI1, snail family transcriptional repressor 1; MMP2, matrix metallopeptidase 2; MMP11, matrix metallopeptidase 11; GSH, glutathione. For panels A and C-F, red indicates GS; blue, DGC. Figure 3B created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license.

Fig. 4. Oxidative phosphorylation (OXPHOS) and cholesterol biosynthesis are targetable vulnerabilities in oligodendrocyte precursor (OPC)-like gliomaspheres (GS).

A–F Dose-response curves for DIPG-007 GS vs. differentiated glioma cells (DGC) treated with indicated concentrations of mitochondrial OXPHOS inhibitors (i.e. complex I inhibitors) (A) Metformin, (B) Phenformin, and (C) IACS-010759 (IACS), or statins (D) Atorvastatin, (E) Fluvastatin, and (F) Pitavastatin for 3 days or 7 days (Metformin). Cell viability was assayed using Cell Titer-Glo 2.0 (DGC) or 3D (GS) and results expressed as percent of vehicle control (0.1% DMSO). Error bars for panels A–F represent mean ± SD from three biological replicates (****p < 0.0001 by area under proliferation curve test followed by two-tailed Student’s t-test). G, H) Western blot analysis of PARP cleavage (apoptosis indicator) with an α-TUBULIN loading control in (G) Phenformin (Phen)-treated and (H) Pitavastatin (Pita)-treated DIPG007 DGC and GS. The figure shows representative blots from two independent experiments. Cells were treated for 48 h at the indicated concentrations. For panels A–F, red indicates GS; blue, DGC.

Fig. 5. Oligodendrocyte precursor (OPC)-like gliomaspheres (GS) exhibit enhanced sterol biosynthetic activity to make cholesterol.

A–F Metabolomics-based assessment of glucose-derived carbon entry into downstream metabolism in differentiated glioma cells (DGC) or GS at the indicated timepoints. Representative metabolites are presented for (A) glycolysis, (B–D) the TCA cycle, (E) sterol biosynthesis, and (F) de novo fatty acid biosynthesis. Enrichment is presented as mass + the number of carbons labeled in the metabolite, m + n. Data are presented as the fractional enrichment of the pool. Colored, stacked bars represent the isotopologues, and the isotopologue enrichment within each group is determined by the mean calculated from three biological replicates with ± SD error bars indicated. Cer (XX:X;OX), ceramide (carbons:unsaturation;oxygens); m/z, mass/charge. (G) Schematic of sterol biosynthesis pathway indicating key intermediates in the biosynthesis of cholesterol and coenzyme Q_10. H–L Cell viability of DIPG-007 GS following treatment with vehicle (0.4% DMSO) or Pitavastatin (Pita) with or without co-treatment with (H) mevalonate (Meva), (I) farnesyl pyrophosphate (FPP), (J) geranylgeranyl pyrophosphate (GGPP), (K) cholesterol, and (L) coenzyme Q₁₀ (CoQ₁₀) at the indicated concentrations. Cell viability was assayed at 72 h post-treatment using Cell Titer-Glo 2.0. Results expressed as percent of vehicle control and error bars represent mean ± SD from three biological replicates (ns = not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 by Unpaired t test, two-tailed with 95% confidence interval). HMG-CoA, β-Hydroxy β-methylglutaryl-coenzyme A; HMGCR, HMG-CoA reductase.

Fig. 6. Bioenergetic properties of oligodendrocyte precursor (OPC)-like gliomaspheres (GS) and astrocyte (AC)-like differentiated glioma cells (DGC).

A–C Bioenergetics analysis (Seahorse assay, mitochondrial stress test) of DIPG-007, SF7761, and DIPG-XIII GS and DGC. (A) Oxygen consumption rate (OCR) during mitochondrial stress test; (B) Basal OCR, (C) Spare respiratory capacity (SRC). Determination of (D) energy charge calculated as [(ATP) + 0.5(ADP)]/ [(ATP) + (ADP) + (AMP)], and (E) ATP/ADP ratios in GS and DGC across all three DIPG pairs. F Extracellular acidification rate (ECAR) of DIPG-007, SF7761, and DIPG-XIII GS and DGC. A–C, F n = 2 independent experiments. D Experiments were performed in technical triplicates and bar graphs are calculated values of energy charge expressed as mean; n = 1. E Experiments were performed in triplicates on the same day and expressed as mean of ATP/ADP ratio in GS vs. DGC, n = 1. Data in D and E were extracted from the metabolomics analysis presented in Fig. 2A.

Fig. 7. Statin and Complex I inhibitors prolong survival in an orthotopic model of DIPG.

A Dose-response curves of radiation-treated DIPG-007, SF7761, and DIPG-XIII gliomaspheres (GS) and differentiated glioma cells (DGC). Cell viability assessed at 7 days post-treatment using Cell Titer-Glo 2.0 (DGC) or 3D (GS) with results expressed as percent of control and error bars representing mean ± SD from 3 biological replicates; ns = not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001**** as determined by area under proliferation curve test followed by two-tailed Student’s t-test. B Schematic of in vivo experiment using a luciferase-expressing DIPG-007 orthotopic xenograft model and intraperitoneal administration of Pitavastatin (Pita) (10 mg/kg; n = 10 mice), Phenformin (Phen) (50 mg/kg; n = 10 mice), or the combination (Comb) (n = 9 mice) of both drugs at indicated doses or vehicle control (Veh) (PBS; n = 10). C Quantitation of change in tumor volume based on bioluminescence imaging (BLI). To calculate BLI, animals were injected with luciferin and luminescence was captured and quantitated. Data are presented as change in BLI between weeks 3 and 7 post-tumor implantation in arbitrary units (arb.units) with error bars representing mean ± SD and multiple comparisons made using uncorrected Fisher’s LSD (Pita vs. control, **p = 0.0073). D Percent change in body weight of DIPG-007 tumor-bearing NSG mice in “C” following treatment, with error bars representing mean ± SD. E End-point survival analyzes of treatment and control tumor-bearing mice using Log-rank (Mantel-Cox) test (Pita vs. control, **p = 0.0016; Phen vs. control, **p = 0.0097; Pita + Phen vs. control, ***p < 0.0001). Vehicle, n = 9; Phenformin, n = 10; Pitavastatin, n = 10; Combination, n = 9. F Model of H3K27M DIPG intratumoral heterogeneity, indicating specific vulnerabilities within oligodendrocyte precursor (OPC)-like and differentiated astrocyte (AC)-like tumor populations and their respective targeting strategies. Gy, Gray; OXPHOS, oxidative phosphorylation. For panels A and F, red indicates GS; blue, DGC. For panels C–E, black indicates vehicle control; red, Pita; purple, Phen; olive, combination Pita + Phen. Figure 7B, F were created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license.