Targeting distinct amino acid metabolic vulnerabilities in IDH-mutant and IDH-wildtype gliomas

Abstract

Lower grade gliomas frequently harbor mutations in isocitrate dehydrogenase (IDH), which define biologically distinct tumor subtypes. Although IDH-mutant and IDH-wildtype gliomas share similar histological morphology, they display markedly different metabolic profiles that may be exploited for targeted therapy. In this study, we investigated therapeutic approaches tailored to these metabolic differences. Using capillary electrophoresis–mass spectrometry, we compared the metabolomes of engineered IDH-wildtype and IDH-mutant glioma cell models. IDH-mutant cells exhibited elevated asparagine levels and reduced glutamine and glutamate levels compared with IDH-wildtype cells. These differences were corroborated in vivo by proton magnetic resonance spectroscopy of 130 patients with diffuse gliomas, showing lower glutamine and glutamate in IDH-mutant tumors. Pharmacological depletion of asparagine with L-asparaginase, which converts asparagine to aspartate, preferentially inhibited the growth of IDH-wildtype glioma cells, and this effect was potentiated by inhibition of asparagine synthetase. In contrast, inhibition of glutamate dehydrogenase 1 (GLUD1), the enzyme catalyzing the conversion of glutamate to α-ketoglutarate, selectively suppressed proliferation of IDH-mutant glioma cells by inducing reactive oxygen species accumulation and apoptosis. In vivo, L-asparaginase suppressed tumor growth in xenografted IDH-wildtype gliomas, whereas GLUD1 inhibition significantly reduced tumor growth in IDH-mutant glioma xenografts. These findings reveal distinct amino acid metabolic vulnerabilities defined by IDH mutation status and identify L-asparaginase and GLUD1 inhibition (via R162) as promising, mutation-specific therapeutic strategies. L-asparaginase demonstrated potent antitumor activity against IDH-wildtype gliomas, while GLUD1 inhibition selectively suppressed IDH-mutant gliomas both in vitro and in vivo. These results highlight the clinical potential of targeting amino acid metabolism in gliomas and provide a strong rationale for translating these mutation-specific approaches into future clinical trials.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40478-025-02193-8.

Introduction

Isocitrate dehydrogenase (IDH) mutations have been frequently observed in young patients and lower-grade gliomas [1, 2]. Despite similar histological findings, the molecular biological backgrounds and prognoses differ between the IDH-wildtype and IDH-mutant gliomas [3]. In the 2021 WHO Classification of Tumors of the Central Nervous System, adult-type diffuse gliomas are categorized as (i) astrocytoma, IDH-mutant; (ii) oligodendroglioma, IDH-mutant and 1p/19q-codeleted; and (iii) glioblastoma, IDH-wildtype [4].

Wildtype IDH converts isocitrate to α-ketoglutarate; alternatively, mutant IDH converts α-ketoglutarate to 2-hydroxyglutarate. Due to its structural similarity to α-ketoglutarate, 2-hydroxyglutarate inhibits α-ketoglutarate-dependent enzymes, including DNA demethylation enzymes and histone demethylases, such as ten-eleven translocation enzymes. Consequently, cells harboring IDH mutations have higher levels of DNA and histone methylation, leading to the global alterations in gene expression that contribute to tumorigenesis [5, 6].

Isocitrate and α-ketoglutarate are involved in the tricarboxylic acid (TCA) cycle, highlighting the role of IDH in metabolism. Therefore, the metabolism of IDH-wildtype and IDH-mutant gliomas is reported to differ significantly, even though their morphologies are similar [7].

Asparagine is a nonessential amino acid. Intracellular asparagine is synthesized by asparagine synthetase (ASNS) [8], which transfers the amino group from glutamine to aspartic acid. The expression of ASNS is regulated by DNA methylation; the presence of methylation in the ASNS promoter region leads to reduced ASNS expression [9]. Conversely, L-asparaginase converts asparagine into aspartate, thereby reducing the amount of asparagine. Thus, L-asparaginase has been clinically used to treat several childhood leukemias and lymphomas [8, 10].

Glutamine, the most prevalent amino acid in the blood, is essential for rapidly growing cancer cells that require additional external sources. Glutamine provides carbon and nitrogen to the cells, serving as a precursor for glucose, amino acids, aspartate, and glutamate. Through glutaminolysis, glutamine carbon contributes to the TCA cycle and supports rapid proliferation by supplying precursors for biosynthetic pathways [8, 11]. Extracellular glutamine is transported into the cytoplasm through alanine, serine, and cysteine transporter 2 (ASCT2). Once inside, glutamine is converted into glutamate by glutaminase, and glutamate is converted into α-ketoglutarate by glutamate dehydrogenase 1 (GLUD1) [12].

Previously, a glioma model was created in which normal human astrocytes (NHA) expressing E6, E7, and hTERT [13] were made tumorigenic by the additional expression of an oncogenic form of HRas or a mutant IDH [14]. In the current study, we used these two models to identify metabolic differences between cells that became tumorigenic in IDH-wildtype or -mutant settings and explored novel therapeutic approaches based on these differences.

Methods

Cell culture and creation of cell lines

The generation and culture of NHA expressing E6E7hTERT, E6E7hTERT plus mutant G12V HRas (HRasV12), and E6E7hTERT plus IDH1^mut have been described previously [13–15]. U251 and U87 cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FBS and 1% penicillin/streptomycin at 37 °C in a 5% CO₂ atmosphere. U251 cells expressing IDH1^mut were generated as described previously [16]. MGG4 cells (derived from a glioblastoma) were maintained in a serum-free medium containing 20 ng/mL of recombinant human epidermal growth factor (EGF) and 20 ng/mL of recombinant human fibroblast growth factor (FGF). BT142 cells (IDH-mutant, oligodendroglioma-derived) were maintained in a serum-free medium containing 20 ng/mL of recombinant human EGF, 100 ng/mL of recombinant human platelet-derived growth factor, 20 ng/mL recombinant human FGF, and 2 µg/mL of heparan sulfate.

Reagents

L-asparaginase, 6-diazo-5-oxo-L-norleucine, R162, asparagine, and dimethyl 2-oxoglutarate (α-ketoglutarate) were purchased from Sigma. 3-methyladenine, V-9302, CB-839, and N-acetyl-L-cysteine were purchased from Selleck Chemicals.

Genetic suppression of ASNS

The cells were plated at 10⁵/mL in six-well plates containing DMEM cell growth media without antibiotics. Twenty-four hours later, the cells were transfected with an optimized amount (5 nmol/L) of siRNA targeting human ASNS (SMARTpool, Dharmacon) or nontargeting siRNAs as a negative control, using DharmaFECT reagent (Dharmacon) according to the manufacturer’s protocol.

Colony formation assay

The colony formation efficiency was determined using the colony formation assay as described previously [14].

Protein extraction and immunoblot analyses

Cells were lysed in RIPA lysis buffer (Life Technologies) supplemented with 1 × PhosStop and protease inhibitor cocktail (Roche). Protein (30 μg) was used for Western blot analysis using primary antibodies against IDH1R132H (Dianova), HRas (Santa Cruz Biotechnology), LC3 (Cell Signaling Technology), ASNS (Santa Cruz Biotechnology), β-actin (Cell Signaling Technology), and the appropriate horseradish peroxidase-conjugated secondary antibodies (Cell Signaling Technology). Antibody binding was detected using enhanced chemiluminescent (ECL) reagents (Fujifilm).

RNA extraction and sequencing

Following the manufacturer’s recommendations, intact Poly(A)+ RNA was isolated from 1 µg of total RNA using the NEBNext Poly(A) mRNA Magnetic Isolation Module (New England Biolabs). The enriched RNA was used to construct the RNA-seq library with the NEBNext Ultra RNA Library Prep Kit for Illumina (New England Biolabs). Three biological replicates were generated per sample, and the prepared libraries were sequenced using the Illumina HiSeq 1500 platform at Fujita Health University (126 bp single-end reads). Base calling was performed using bcl2fastq v1.8.4, and sequencing was performed for a subset of samples via GENEWIZ, Inc. Raw sequences underwent quality control and trimming with FastQC v0.11.5 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) to ensure data quality, followed by mapping to the human reference genome (hg19) with Hisat2 v2.0.5 [17] using default parameters. Aligned reads were converted to and sorted in BAM format using SAMtools v1.3.1 [18]. Gene expression levels were estimated using HTSeq v0.6.0 [19] against the Homo_sapiens_UCSC_hg19.gtf annotation file. DESeq2 v1.14.1 [20] was used to identify significantly differentially expressed genes (cut-offs: adjusted P-value < 0.05, log2 fold change > 1 or <  − 1), employing the Wald test. Finally, Metascape [21] was used for GO analysis on these genes. Genes with significant expression changes were subjected to a KEGG analysis [22] to identify their association with metabolic pathways. Among the annotated genes, those classified under Metabolic pathways were further grouped into KEGG metabolic subcategories, and the number of genes in each subcategory was counted to summarize the distribution of metabolic genes.

Proteomics

The nanoscale liquid chromatography coupled to tandem mass spectrometry (nano-LC–MS/MS) analysis was performed as described previously [23]. Briefly, the samples were analyzed using an LC system (EASY-nLC 1000; Thermo Fisher Scientific) coupled to an MS system (Orbitrap Fusion; Thermo Fisher Scientific). Peptide ions were detected using MS combined with the Xcalibur software (version 4.1; Thermo Fisher Scientific). MS/MS searches were conducted with the MASCOT (version 2.6.2; Matrix Science) and SEQUEST HT search algorithms against the SwissProt Homo sapiens database (2017-10-25) using Proteome Discoverer (PD) (version 2.2.0.388; Thermo Fisher Scientific). Label-free quantification was performed using PD, and the differential analysis was conducted using the Perseus software (version 1.6.15.0) [24]. The significance was assessed using the Student’s t-test (cutoff: p < 0.05). Proteins with significant expression changes were subjected to KEGG analysis [22] to identify their association with metabolic pathways.

Quantitative PCR

Total RNA was isolated from the cells using the PureLink RNA Mini Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Total RNAs (2500 ng) were reverse-transcribed to produce first-strand cDNA using the SuperScript VILOMaster Mix (Thermo Fisher Scientific) kit, according to the manufacturer’s protocol. The Thermal Cycler Dice System (TaKaRa, Shiga, Japan) was used to perform a two-step reverse transcription polymerase chain reaction (PCR). The mRNA transcripts were quantified using SYBR Premix ExTaq (TaKaRa) using the following amplification conditions: 30 s at 95 °C; 40 cycles at 95 °C for 5 s and 60 °C for 30 s each; dissociation for 15 s at 95 °C and 15 s at 60 °C; and then 15 s at 95 °C on a Thermal Real-Time PCR System 7900HT (Thermo Fisher Scientific). The results for each sample were normalized to a housekeeping gene encoding β-actin.

Metabolome analysis

After collection, 1 × 10⁶ cells were centrifuged at 2,000 g for 5 min at 4 °C and washed with phosphate-buffered saline (PBS) twice. Then, 1 ml of methanol containing the internal standards (20 μmol/l each of L-methionine sulfone and D-camphor-10-sulfonic acid [Wako]) was added. A 400 μl aliquot of extracted solutions was mixed with 160 μl of Milli-Q water and 400 μl of chloroform. After centrifugation, 300 μl of the separated upper (methanol–water) layer was ultrafiltered using a 5 kDa-cutoff filter (Human Metabolome Technologies) to remove proteins. The filtrate was dried using an evacuated centrifuge and dissolved in 25 μl of Milli-Q water containing 200 μmol/l of reference compounds (3-aminopyrrolidine and trimesic acid) prior to CE-MS analysis. CE-MS-based metabolomic profiling and data analysis were performed as described previously [25–29].

Apoptosis assay

The apoptosis assay was performed according to the manufacturer’s instructions. In brief, cells were plated at 105/well in a 6-well plate 24 h before treatment with R162, NAC, or a vehicle for 3 days. The cells were collected and measured by flow cytometry using the Annexin V-633 Apoptosis Detection Kit (Nacalai Yesque Inc.).

Reactive oxygen species assay

The ROS assay was performed according to the manufacturer’s instructions. In brief, cells were plated at 105/mL in a six-well plate 24 h before treatment with R162, NAC, or vehicle for 3 days, followed by staining with CellROXTM Deep Red reagent (ThermoFisher) for 1 h. The cells were collected, and intracellular ROS was measured using flow cytometry.

Proton magnetic resonance spectroscopy

Single-voxel proton (1H) magnetic resonance spectroscopy (MRS) was acquired with a point-resolved spectroscopy sequence for both short and long TE (35 and 144 ms, respectively) using the following standard parameters: TR, 2000 ms; flip angle, 90°; voxel size, 15 mm × 15 mm × 15 mm (this was the standard protocol, but the dimensions could be varied according to the size and shape of the lesion); bandwidth, 1.27 Hz/point; and NEX, 128. The voxel of interest was selected to include the largest part of the tumor based on the axial T2WI. The spectra data were automatically analyzed with the LCModel quantification program for the proton MR spectra (Stephen Provencher). Metabolites with Cramer-Rao lower bound values of < 30% were used for the analysis; this cut-off was selected based on previous studies [30, 31]. Patients who underwent the first operation and were diagnosed with diffuse gliomas (WHO grade 2–4) at our institute from 2014 to 2022 were retrospectively evaluated (n = 130). Histopathologic diagnoses and grading were assigned according to the 2021 WHO Classification of Tumors of the Central Nervous System. This retrospective study was approved by the ethics review boards of our institution.

Xenograft studies

Animal experiments were performed according to protocols approved by the Institutional Animal Care and Use Committee of Fujita Health University. Female, 6-weeks-old, nude mice (athymic nu/nu) were subcutaneously injected with 2 × 10⁶ U251 cells harboring green fluorescent protein (GFP) or U251 cells expressing IDH1R132H and GFP on the right flank. The efficacy of R162 was evaluated by administering the drug daily via intraperitoneal injection (20 mg/kg); 50% of dimethyl sulphoxide (DMSO) in PBS was used as a diluent control. The efficacy of L-asparaginase was assessed by administering the drug every other day by intraperitoneal injection (5U/g); PBS was used as a diluent control. Tumor growth was recorded by measuring the two perpendicular diameters of the tumors, while the tumor size was calculated using the following formula: length × (width)² × 1/2 [32]. Animals were randomly assigned to each treatment group to minimize bias. Blinding was not performed during tumor measurements.

Statistical analyses

Data are reported as mean ± standard error of the mean (SE) from at least three experiments. The unpaired Student’s t-test was applied to compare the two groups. One-way analysis of variance with a post hoc Tukey–Kramer multiple comparisons test was used to compare more than two groups. A P-value of < 0.05 was considered statistically significant.

Results

Metabolic characterization of NHAE6E7hTERT, NHAE6E7hTERTRas, and NHAE6E7hTERTIDHmut cells

NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut were used to reveal the genetic or metabolic difference between the IDH-wildtype and -mutant gliomas (Fig. 1A, B). Pathway enrichment analyses were performed by Metascape on NHAE6E7hTERT, NHAE6E7hTERTRas, and NHAE6E7hTERTIDHmut cells to explore the metabolic differences between cells transformed in the IDH-wildtype or -mutant contexts. Heatmaps of the top 20 most enriched terms are listed in supplementary Fig. 1A (NHAE6E7hTERTIDHmut vs. NHAE6E7hTERTRas), 1 B (NHAE6E7hTERT vs. NHAE6E7hTERTIDHmut), 1C (NHAE6E7hTERT vs. NHAE6E7hTERTRas). Genetic and proteomics analyses were performed to evaluate the percentages of differentially expressed genes or proteins associated with metabolism. The analyses revealed that 7.6% and 14.6% of differentially expressed genes and proteins between NHAE6E7hTERTIDHmut and NHAE6E7hTERTRas, 7.1% and 15.4% of differentially expressed genes and proteins between NHAE6E7hTERT and NHAE6E7hTERTRas, and 10.0% and 14.6% of differentially expressed genes and proteins between NHAE6E7hTERT and NHAE6E7hTERTIDHmut were involved in metabolism (Supplementary Table 1). Among the annotated genes, those classified under Metabolic pathways were further grouped into KEGG metabolic subcategories, and the number of genes in each subcategory was counted to summarize the distribution of metabolic genes (Fig. 1C (NHAE6E7hTERTIDHmut vs. NHAE6E7hTERTRas), 1 D (NHAE6E7hTERT vs. NHAE6E7hTERTIDHmut), 1E (NHAE6E7hTERT vs. NHAE6E7hTERTRas). Across contrasts, amino-acid metabolism was prominently represented among metabolism-related differentially expressed genes.

Fig. 1.

A Schematic of the creation of Isocitrate dehydrogenase (IDH)-wildtype (NHAE6E7hTERTRas) and IDH-mutant (NHAE6E7hTERTIDHmut) glioma models. B Validation of the identity of the cell lines used in this panel via Western blotting. 1, NHAE6E7hTERT; 2, NHAE6E7hTERTRas; 3, NHAE6E7hTERTIDHmut. Differentially expressed genes identified by RNA-seq were annotated using the KEGG pathway. Genes categorized under Metabolic pathways were grouped according to KEGG metabolic subcategories, and the number of genes in each subcategory was counted. C NHAE6E7hTERTRas versus NHAE6E7hTERTIDHmut. D NHAE6E7hTERT versus NHAE6E7hTERTIDHmut. E NHAE6E7hTERT versus NHAE6E7hTERTRas

The IDH-wildtype and -mutant glioma cells were analyzed by CE-MS to define their metabolic differences. Principal component analysis (PCA) defined three groups (NHAE6E7hTERT, NHAE6E7hTERTIDHmut, and NHAE6E7hTERTRas) that could be separated from each other (Fig. 2A). The PCA loading plot (Fig. 2B) highlighted that 2-hydroxyglutarate, asparagine, glutamine, and glutamate were among the top contributors to the variance between IDH-wildtype and -mutant glioma cells. The two latter cells were then compared. Consistent with a previous report [33], the amount of 2-hydroxyglutarate was increased in NHAE6E7hTERTIDHmut compared to that in NHAE6E7hTERTRas cells (Fig. 2C). To highlight the most biologically relevant pathways in our study, we extracted three key metabolites—asparagine, glutamine, and glutamate—and presented them in a dedicated panel (Fig. 2D). A complete metabolomic profile comparing the two cell types is provided in Supplementary Fig. 2. Asparagine levels were lower in NHAE6E7hTERTRas cells than in NHAE6E7hTERTIDHmut cells; alternatively, glutamine and glutamate levels were lower in NHAE6E7hTERTIDHmut than in NHAE6E7hTERTRas cells (Fig. 2D).

Fig. 2.

The metabolic differences between IDH-wildtype and -mutant glioma cells. A Principal component analysis (PCA) of the metabolomics profiles of NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells. B PCA loading plot highlighting key metabolites differentiating the IDH-wild-type and -mutant glioma cells. C The amount of 2-hydroxyglutarate in NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells. *P < 0.05. D The amount of asparagine (Asn), glutamine (Gln), and glutamate (Glu) in NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells. *P < 0.05, ***P < 0.001

Confirmation of lower levels of glutamine and glutamate in samples of patients with diffuse glioma harboring mutant IDH by proton MRS

Metabolite values measured by MRS were compared between diffuse gliomas harboring mutant IDH and those without mutant IDH. Across the full MRS cohort (n = 130; mean age 60.8 ± 16.9 years; 79 male/51 female), IDH status was wildtype in 96 cases and mutant in 34. Final WHO 2021 diagnoses comprised IDH-mutant astrocytoma (grade 2, 9; grade 3, 7; grade 4, 2), IDH-mutant and 1p/19q-codeleted oligodendroglioma (grade 2, 12; grade 3, 4), glioblastoma, IDH-wildtype (n = 86), and other IDH-wildtype diffuse gliomas (n = 10) (Supplementary Table 2).Consistent with the CE-MS data, the glutamine, glutamate, and glutathione (GSH) values in diffuse glioma harboring mutant IDH were lower than those in diffuse glioma without mutant IDH (Fig. 3A–C). Within the IDH-mutant group, glutamine, glutamate, and GSH did not differ significantly between astrocytoma and 1p/19q-codeleted oligodendroglioma (Supplementary Table 3).

Fig. 3.

Metabolite values measured by proton magnetic resonance spectroscopy between diffuse gliomas harboring mutant IDH and those without mutant IDH (IDH-WT). Metabolites whose levels were statistically different between the two groups in the 35 ms echo time spectrum. Gln, glutamine; Glu, glutamate; GSH, glutathione

Asparagine as a therapeutic target in IDH-wildtype gliomas

The suppression of asparagine levels might be more deleterious to IDH-wildtype cells than IDH-mutant cells because asparagine and aspartate were less abundant in the IDH-wild-type NHAE6E7hTERTRas cells compared to NHAE6E7hTERTIDHmut cells. Consistent with this idea, exogenous L-asparaginase, which converts asparagine into aspartate, reduced colony formation efficiency in the NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells in a dose-dependent manner but did so at lower concentrations in IDH-wild-type NHAE6E7hTERTRas cells (Fig. 4A). The deletion of amino acids sometimes leads to autophagy, which was also evaluated in cell lines treated with L-asparaginase. Immunoblot analysis revealed an L-asparaginase-induced increase in the ratio of LC3-II to LC3-I that was selective for IDH-wild-type cells and could be blocked by the autophagy inhibitor 3-MA, indicative of asparaginase-induced, IDH-wildtype-selective autophagy (Fig. 4B). The addition of 3-MA did not affect the colony formation efficiency of IDH-mutant cells (alone or exposed to asparaginase); however, it significantly reduced asparaginase-induced cytotoxicity in IDH-wild-type NHAE6E7hTERTRas cells, indicating that autophagy contributes to L-asparaginase-induced cell death rather than cell survival (Fig. 4C). Inhibition of autophagy with 3-MA reduced LC3-II and partially rescued L-asparaginase–induced cytotoxicity in IDH-WT cells (Fig. 4B, C), supporting an autophagy-associated contribution to cell death under asparagine depletion. Exogeneous asparagine was added to IDH-wildtype or -mutant cells treated with L-asparaginase to further verify that the depletion of asparagine contributed to reduced colony formation efficiency. As shown in Fig. 4D, exogenous asparagine rescued the suppressed colony formation efficiency of cells treated with L-asparaginase, suggesting that L-asparaginase-induced cytotoxicity was at least partially due to the deletion of asparagine.

Fig. 4.

Effect of L-asparaginase in IDH-wildtype and -mutant glioma cells. A Colony formation efficiency of engineered glioma models: IDH-wildtype (NHAE6E7hTERTRas) and IDH-mutant (NHAE6E7hTERTIDHmut) cell studied following L-asparaginase exposure (0–1 unit). B NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells were treated with L-asparaginase with/without 3-MA (autophagy inhibitor). Whole-cell extracts were examined by Western blotting. C Colony formation efficiency of cells (NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut) following L-asparaginase exposure (0.05 and 0.1 units) with/without 3-MA. D Colony formation efficiency of cells (NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut) following L-asparaginase exposure with/without asparagine. E NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells were treated with L-asparaginase. The mRNA expression of ASNS was examined by qPCR. F Western blot analysis of ASNS and β-actin levels in NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells three days after exposure to scramble or ASNS-targeting siRNA. G Colony formation efficiency of control and ASNS-suppressed cells (NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut) after treatment with L-asparaginase. H Western blot analysis of LC3 and β-actin levels in NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells 3 days after exposure to scramble or ASNS-targeting siRNA treated with/without L-asparaginase. I Colony formation efficiency of cells (NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut) after treatment with L-asparaginase combined with/without DON. J Colony formation efficiency of IDH-wild-type glioma cells (NHAE6E7hTERTRas, U87, and U251) and IDH-mutant glioma cells (NHAE6E7hTERTIDHmut, U251IDHmut) treated with L-asparaginase. K Relative cell viability of MGG4 cells (IDH-wildtype) treated with L-asparaginase. L Relative cell viability of BT142 cells (IDH-mutant) treated with L-asparaginase. *P < 0.05

However, exposure to L-asparaginase resulted in increased expression of mRNA of ASNS in the IDH-mutant and -wildtype cells (Fig. 4E), likely as a pro-survival compensatory response to the L-asparaginase-induced deletion of asparagine. Further studies were performed in which the ASNS activity was genetically or pharmacologically inhibited, following which the effects on L-asparaginase-induced toxicity were monitored. The expression of an siRNA targeting ASNS suppressed the ASNS levels in both IDH-wildtype and -mutant cells relative to that of a scrambled control siRNA (Fig. 4F); additionally, L-asparaginase-induced cytotoxicity was enhanced in the IDH-wild-type NHAE6E7hTERTRas cells, but not the NHAE6E7hTERTIDHmut cells (Fig. 4G). The combined treatment of L-asparaginase and siRNA targeting ASNS enhanced autophagy in IDH-wildtype cells but not in IDH-mutant cells (Fig. 4H). Similar results were noted when DON, a glutamine analog that inhibits ASNS activity [34], was combined with L-asparaginase. Specifically, DON enhanced L-asparaginase-induced cytotoxicity in NHAE6E7hTERTRas cells, whereas the effect was less prominent in NHAE6E7hTERTIDHmut cells (Fig. 4I).

U251 and U87 cells were treated with L-asparaginase to confirm the effect of L-asparaginase on other IDH-wildtype glioma cells. In addition, colony-formation assays in U251-IDH1R132H cells were performed to examine the L-asparaginase response (Fig. 4J). Cell viability was measured to reveal the antitumor effect of asparaginase inhibition on patient-derived MMG4 (IDH-wildtype glioma) and BT142 (IDH-mutant glioma) cells. Consistent with the cell line data, MMG4 was more sensitive to L-asparaginase than BT142 (Figs. 4K, L). Taken together, these findings indicate that asparagine depletion preferentially affects IDH-wildtype backgrounds and support L-asparaginase as a potential therapy for IDH-wildtype gliomas.

Inhibition of glutaminolysis could be therapeutic in IDH-mutant gliomas

The amount of glutamine and glutamate was lower in NHAE6E7hTERTIDHmut cells than in NHAE6E7hTERTRas cells (Fig. 2C). Therefore, the glutaminolysis pathway was considered as a target for treating IDH-mutant gliomas. V-9302, CB-839, and R162 selectively inhibit ASCT2, glutaminase, and GLUD1, respectively [35]. The antitumor effects of these three inhibitors (V-9302, CB-839, or R162) were initially evaluated in the NHAE6E7hTERTRas cells and NHAE6E7hTERTIDHmut cells using the colony formation assay (Fig. 5A, supplementary Fig. 3A, B). Each reagent showed antitumor effects in a dose-dependent manner. The antitumor effect of R162 was most significantly different between the NHAE6E7hTERTRas cells and NHAE6E7hTERTIDHmut cells, with the latter more sensitive to the reagent. Therefore, R162 was selected for further experiments. R162 decreased GLUD1 activity at a concentration of 25 μM that reduced colony formation efficiency in NHAE6E7hTERTIDHmut cells (Fig. 5B).

Fig. 5.

Effect of R162 in IDH-wildtype and -mutant glioma cells. A Colony formation efficiency of NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells treated with R162. B GLUD1 activity of NHAE6E7hTERTIDHmut cells treated with/without R162. C Intracellular ROS measured in NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells treated with/without R162. D Percentage of apoptosis cells in NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells treated with/without R162. E Colony formation efficiency of cells (NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut) following R162 exposure with/with dimethyl 2-ketoglutarate. F Intracellular ROS measured in NHAE6E7hTERTIDHmut cells treated with R162 or/and NAC. G Percentage of apoptosis cells among NHAE6E7hTERTIDHmut cells treated with R162 and/or NAC. H Colony formation efficiency of NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells treated with R162 and/or NAC. I Colony formation efficiency of IDH-wild-type (NHAE6E7hTERTRas, and U251) and IDH-mutant (NHAE6E7hTERTIDHmut, and U251IDHmut) glioma cells treated with R162. J Relative cell viability of MGG4 cells (IDH-wildtype) treated with R162. K Relative cell viability of BT142 cells (IDH-mutant) treated with R162. *P < 0.05

Glutaminolysis is associated with ROS generation, and high concentrations of ROS have been reported to induce apoptosis [36]. ROS and apoptosis were evaluated in NHAE6E7hTERTRas and NHAE6E7hTERTIDHmut cells following R162 exposure to determine their association with the antitumor effect induced by the inhibitor. ROS levels significantly increased in NHAE6E7hTERTIDHmut cells treated with R162 compared to those treated with the vehicle (DMSO); conversely, no significant changes in ROS levels were observed in NHAE6E7hTERTRas cells treated with R162 (Fig. 5C). Similarly, the percentage of annexin-positive and PI-negative cells increased in NHAE6E7hTERTIDHmut cells but not IDH-wildtype cells exposed to R162 (Fig. 5D). Cell-permeable dimethyl α-ketoglutarate [37] was added to the cultured medium of cells treated with R162 to confirm that the suppression of α-ketoglutarate contributed to reduced colony formation efficiency, which partially rescued the suppressed colony formation efficiency in IDH-mutant cells treated with R162 (Fig. 5E).

NAC, a ROS scavenger, was added to the IDH-mutant cells treated with either a vehicle or R162 to confirm that the increased ROS caused by R162 led to increased apoptosis and antitumor effects. NAC exposure reduced the levels of ROS in the control and R162-treated cells and the percentage of apoptotic cells in the R162-treated group (Fig. 5F, G). Furthermore, the addition of NAC partially rescued the IDH-mutant cells, but not the IDH-wildtype cells, from R162-induced loss of clonogenicity, indicating that R162-induced ROS contributed to increased apoptosis and antitumor effects (Fig. 5H).

NHAE6E7hTERTIDHmut cells were transfected with shRNA targeting GLUD1 (Supplementary Fig. 3C), and the ROS levels, number of apoptotic cells, and colony formation efficacy were measured (Supplementary Fig. 3D−F) to verify that the effect of R162 was due to GLUD1 inhibition. Similar to the R162-mediated pharmacological inhibition of GLUD1, genetic inhibition of GLUD1 resulted in a comparable increase in ROS and the percentage of apoptotic cells, along with a loss of clonogenicity, confirming that on-target inhibition of GLUD1 mediated the antitumor effect of R162.

Finally, U251 cells were infected with a lentivirus encoding GFP and IDH1R132H to verify that the effects of GLUD1 inhibition were not specific to NHAE6E7hTERTIDHmut cells (Supplementary Fig. 3G). As expected, the ROS levels significantly increased in U251IDHmutGFP cells treated with R162 compared to those treated with the vehicle; no difference in ROS levels was observed between the R162-treated and DMSO-treated U251GFP cells (Supplementary Fig. 3H). The percentage of annexin-positive and PI-negative cells also increased in U251IDHmutGFP cells exposed to R162, while no apoptosis induction was seen in the U251GFP cells (Supplementary Fig. 3I). The colony formation efficacy of IDH-wildtype glioma cells (NHAE6E7hTERTRas, U251GFP) was significantly less impacted by R162 exposure compared to that of IDH-mutant glioma cells (NHAE6E7hTERTIDHmut, U251IDHmutGFP). Thus, GLUD1 inhibition caused preferential cytotoxicity in IDH-mutant cells, including in the isogenic U251 ± IDH1-R132H pair, and was partially rescued by NAC/α-KG, indicating an on-target, metabolism-linked effect (Fig. 5I). Cell viability assays were conducted to evaluate the antitumor effects on patient-derived glioma cells. Consistent with the cell line data, BT142 was more sensitive to R162 than MGG4 (Figs. 5J, K). These findings support the efficacy of GLUD1 inhibition in IDH-mutant gliomas.

L-asparaginase and GLUD1 inhibition suppresses tumor growth in IDH-wildtype glioma and IDH-mutated glioma xenografts, respectively

The effect of L-asparaginase or R162 was evaluated using subcutaneous xenograft models. U251GFP and U251IDHmutGFP cells were transplanted into the flanks of mice. After the tumors reached a size of 100 mm³, L-asparaginase, R162, or a vehicle was administered intraperitoneally. As shown in Fig. 6A, B, L-asparaginase inhibited tumor growth in U251GFP xenografts but not in the U251IDHmutGFP xenografts. Conversely, R162 suppressed the growth of U251IDHmutGFP xenografts but not the U251GFP xenografts (Figs. 6C, D).

Fig. 6.

Effect of L-asparaginase and R162 on tumor growth in a subcutaneous xenograft model of IDH-wild-type and -mutant glioma cells. Effect of L-asparaginase on the in vivo growth of A U251 cells harboring GFP or B U251 cells expressing IDH1R132H and GFP in a subcutaneous xenograft model. Values represent the mean ± standard error of the mean (SE). Effect of R162 on the in vivo growth of C U251 cells harboring GFP or D U251 cells expressing IDH1R132H and GFP in a subcutaneous xenograft model. Values represent the mean ± SE. *P < 0.05

Discussion

Gliomas are divided into IDH-mutant and IDH-wildtype categories, with significant differences in their genetic background and prognosis, despite having similar histopathological appearances. The tumorigenic pathways also differ between these two types. Mutant IDH1/2 is recognized as a driver mutation in IDH-mutant glioma [2], whereas RTK/RAS/PI3K, p53, or RB signaling pathways are associated with the development of glioblastomas [38]. Previously, we created models of IDH-wildtype and -mutant gliomas from NHA cells [13, 14]. In this study, we compared the genetic, proteomic, and metabolic differences between these two types of gliomas. Previous studies have shown metabolic differences between IDH-wildtype and -mutant gliomas, including variations in lactate levels [39], glucose consumption, glutamine metabolism, and TCA cycle intermediates [40]. In cancer metabolism, glycolysis has been the most studied pathway, followed by moderate research on glutaminolysis, while studies targeting asparagine have been less frequent. This study focused on amino acids such as asparagine, glutamine, and glutamate. The in vitro data on these amino acids were consistent with clinical data obtained using MRS, implying that the results could apply to clinical cases. Category composition placed amino-acid metabolism consistently high across contrasts, prompting hypothesis-driven analyses of asparagine/glutamine pathways (Figs. 4 and 5).

Our study revealed that IDH-wildtype gliomas are more vulnerable to asparagine depletion compared to IDH-mutant gliomas, which was also shown in a patient-derived glioma line. Consistent with previous research using glioblastoma cell lines [34], L-asparaginase reduces the proliferation of IDH-wildtype glioma cells. Extracellular asparagine is taken up into the cell. In addition, intracellular asparagine is synthesized from aspartate by ASNS. L-asparaginase converts asparagine into aspartate, reducing the extracellular asparagine level and depleting intracellular asparagine, which is necessary for cell proliferation. L-asparaginase induces autophagy and increases the expression of ASNS to compensate for the reduced extracellular asparagine intake. Interestingly, the effect of L-asparaginase on cell death differed between NHAE6E7hTERTRas cells and NHAE6E7hTERTIDHmut cells. NHAE6E7hTERTRas cells had lower intracellular asparagine levels despite higher ASNS protein levels, indicating an insufficient supply of asparagine and a compensatory increase in ASNS. Thus, NHAE6E7hTERTRas cells are more susceptible to asparagine depletion. Consequently, combining L-asparaginase with ASNS inhibitors is expected to enhance its antitumor effect, as shown by the genetic or pharmacological inhibition of ASNS in glioblastoma cell lines [34]. While DON is not approved due to severe side effects [41], L-asparaginase is approved for diseases such as childhood acute lymphoblastic leukemia [42], greatly facilitating its potential for use alone or in combination with ASNS inhibitors in clinical trials for IDH-wild-type gliomas. Furthermore, although L-asparaginase itself does not cross the blood–brain barrier (BBB), systemic administration significantly reduces asparagine levels in the plasma and cerebrospinal fluid (CSF) [43]. Thus, L-asparaginase may exert antitumor effects within the brain by depleting CSF asparagine, despite its inability to directly penetrate the BBB.

Previous studies have highlighted the critical role of glutamine metabolism in tumor progression and maintenance. For instance, ASCT2-mediated glutamine transport and glutaminase activity have been identified as essential for sustaining proliferative and biosynthetic demands in tumor cells, and their inhibition leads to significant tumor suppression in preclinical models [35]. Furthermore, GLUD1, the terminal enzyme in the glutaminolysis pathway, contributes to TCA cycle replenishment and modulates redox homeostasis via the GPx1 axis, thereby promoting tumor cell survival under oxidative stress [44]. Our findings that GLUD1 inhibition increases ROS and apoptosis in IDH-mutant gliomas are consistent with these mechanisms. Collectively, these insights support our conclusion that GLUD1 is a promising and multifaceted therapeutic target, particularly in IDH-mutant glioma subtypes.

Several enzymes and transporters are associated with pathways that lead from extracellular glutamine to α-ketoglutarate and the TCA cycle, any of which may serve as therapeutic targets. Based on the analysis of several inhibitors, data from the present study suggest that inhibition of GLUD1 produces the greatest differential antitumor effect between IDH-wildtype and -mutant gliomas. GLUD1 is the final enzyme in the pathway that converts glutamine into α-ketoglutarate, which enters the TCA cycle. Genetic or pharmacological (R162) inhibition of GLUD1 showed consistent results in multiple cell models with IDH1 mutation. Moreover, among the patient-derived glioma lines, IDH-mutant gliomas were more sensitive to GLUD1 inhibitors than IDH-wildtype gliomas.

According to one study, siRNA targeting GLUD1 or R162 increased ROS and decreased cell proliferation in lung and breast cancer; the inhibition of GLUD1 decreased α-ketoglutarate and fumarate levels, which inactivated GPx, resulting in increased ROS [44]. The mechanism that makes IDH-mutant gliomas more sensitive to GLUD1 inhibition is multi-factorial. First, inhibition of GLUD1 induces ROS. Second, IDH-mutant gliomas show low levels of GSH. The conversion of isocitrate to α-ketoglutarate involves the conversion of NADP+ to NADPH. Wildtype IDH cells metabolize isocitrate to α-ketoglutarate, converting NADP+ to NADPH. In contrast, mutant IDH metabolizes α-ketoglutarate into 2- hydroxyglutarate, with NADPH being converted into NADP+. Therefore, the NADPH level is expected to be lower in IDH-mutant gliomas than in IDH-wildtype gliomas. In the present study, NHAE6E7hTERTIDHmut cells showed lower NADPH levels than NHAE6E7hTERTRas cells (Supplementary Fig. 4). Finally, GSH is a key antioxidant that mitigates oxidative stress. Oxidized glutathione (GSSG) is converted to GSH using NADPH as an electron donor; consequently, ROS activation increases in IDH-mutant gliomas, which is believed to contribute to the differing antitumor effects observed between IDH-mutant and -wildtype gliomas.

NAC did not completely rescue the colony formation efficiency suppressed by R162, despite the ROS levels in cells treated with NAC and R162 being almost similar to those in cells treated with the vehicle only; hence, other factors induced by R162 might also contribute to cytotoxic effects. For example, R162 probably suppresses α- ketoglutarate-dependent enzymes by reducing α-KG. Another possibility is that GLUD1 has been reported to activate the NF-kB pathway, contributing to tumorigenesis [45]. While off-target metabolic effects of R162 cannot be excluded, GLUD1 knockdown phenocopied R162 and NAC/α-KG partially rescued its effects, supporting an on-target GLUD1 mechanism rather than general cytotoxicity.

Currently, there are no clinically available GLUD1 inhibitors, and the lack of data on R162 BBB permeability limits its testing in preclinical and clinical settings. However, new methods to open the BBB, such as ultrasound and laser treatment [46, 47], have been reported, and these, along with convection-enhanced delivery, may speed the development of the novel approaches identified here to target differences in metabolism between IDH-wildtype and -mutant glioma [48].

One limitation of the present study is the use of subcutaneous xenograft models rather than orthotopic brain tumor models. While subcutaneous models allow for the straightforward measurement of tumor growth and response to treatment, they do not fully recapitulate the unique microenvironment of the brain, including the presence of the BBB and regional hypoxia. Accordingly, our in vivo data should be viewed as tumor-volume–based proof-of-concept without survival or histopathological endpoints. Future studies employing orthotopic implantation will be necessary to evaluate BBB penetration, intratumoral drug exposure, and efficacy of metabolism-targeted therapies under more physiologically relevant conditions.

Another limitation of the present study is the reliance on an HRAS-mutant glioblastoma model to represent IDH-wild-type glioma. To minimize genetic differences between the IDH-mutant and IDH-wild-type models, we first immortalized normal human astrocytes (NHA) by introducing HPV16 E6E7 and TERT, then added mutant IDH1 or HRAS to generate the respective lines. The immortalized intermediate (NHAE6E7hTERT) was not used as a control because it does not undergo tumorigenesis [14]. HRAS- or mutant IDH-transduced immortalized models have been widely used in various studies [13–15] and are considered appropriate models for IDH-wildtype and -mutant gliomas, respectively. However, we acknowledge that HRAS mutations are rare in glioblastomas and that the metabolic effects of RAS signaling cannot be entirely ruled out. Therefore, the findings of this study should be regarded as fundamental insights into the metabolic impact of the IDH mutation status.

Conclusions

In this study, we elucidated the metabolic differences between IDH-wildtype and -mutant glioma cells and explored novel therapeutic approaches based on these differences. Both L-asparaginase and GLUD1 inhibitors showed promise as treatments for IDH-wild-type and IDH-mutant gliomas, respectively. With advancements in techniques to improve BBB permeability, the clinical application of these treatments is highly anticipated. Our results underscore the importance of distinct therapeutic strategies for IDH-wild-type and -mutant gliomas, paving the way for more effective, personalized treatments for individual patients.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

α-KG

α-Ketoglutarate

ASNS

Asparagine synthetase

CE–MS

Capillary electrophoresis–mass spectrometry

GLUD1

Glutamate dehydrogenase 1

IDH

Isocitrate dehydrogenase

MRS

Magnetic resonance spectroscopy

ROS

Reactive oxygen species

Author contributions

Conceptualization: S.O., H.S., R.P., and Y.H. Methodology: S.O., H.A., K.H., H.Y., K.M., and Kunihiro Tsuchida. Investigation: S.O., H.A., T.T., K.H., H.Y., K.M., M.N., Kensuke Tateishi, and Kunihiro Tsuchida. Formal analysis: S.O., H.A., T.T., K.H., H.Y., K.M., H.S., R.P., and Y.H. Writing—original draft: S.O., H.A., K.H., H.Y., K.M., Writing—review and editing: M.N., Kensuke Tateishi, H.W., H.S., R.P. and Y.H. All authors reviewed and approved the manuscript.

Funding

This work was supported by JSPS KAKENHI (22K09243 and 19K09468), TOYOAKI SCHOLARSHIP FOUNDATION, Aichi Cancer Research Foundation, DAIKO FOUNDATION, The Nitto Foundation, Ichihara International Scholarship Foundation, Takeda Science Foundation, and Niigata University Brain Research Institute Collaborative Project.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committees of Fujita Health University (HM21-424).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.