Autophagy inhibition sensitizes radiotherapy responses in high-grade mutant IDH1 glioma

Abstract

Mutant isocitrate dehydrogenase 1 (mIDH1) catalyzes 2-hydroxyglutarate (2HG) production which leads to epigenetic reprogramming in astrocytomas with tumor protein p53 (TP53)/α-thalassemia/mental retardation, X-linked (ATRX) loss. RNA-sequencing, single-cell RNA-sequencing, and Chromatin Immunoprecipitation sequencing (ChIP-seq) followed by bioinformatics analysis shows that human and mouse mIDH1 gliomas exhibit downregulated gene ontologies (GOs) related to mitochondrial metabolism and upregulated autophagy-related GOs. Decreased mitochondrial metabolism is accompanied by decreased glycolysis, rendering autophagy as a source of energy in mIDH1 gliomas. Mutant IDH1 glioma cells exhibit increased expression of autophagy-related proteins and enhanced microtubule-associated protein 1 light chain 3 (LC3) I/II conversion, indicating augmented autophagy. Inhibiting autophagy in vivo by administration of synthetic protein nanoparticles (SPNPs) encapsulating autophagy related gene 7 (ATG7) silencing RNA sensitizes mIDH1 glioma cells to radiation, resulting in tumor regression, long-term survival, and immunological memory. This work uncovers autophagy as a critical pathway for survival in mIDH1 gliomas and its inhibition elicits radiosensitivity in vitro in human and mouse mIDH1 glioma cells, and in vivo in mIDH1 models.

Mutations in IDH1 are linked to altered metabolism and increased autophagy in glioma. Here, authors show that inhibiting autophagy sensitizes mutant IDH1 glioma cells to radiation and improves outcomes in preclinical models.

Introduction

Gliomas are primary brain neoplasms characterized by high morbidity, recurrence, and mortality¹. This pathology includes a genetically and phenotypically heterogeneous group of CNS tumors, which are molecularly classified into different subtypes according to the presence of specific genetic lesions and epigenetic profiles^2,3. A gain-of-function mutation in isocitrate dehydrogenase 1 (IDH1^R132H; mIDH1) is a common genetic alteration in adult gliomas⁴. IDH1^R132H catalyzes the conversion of α-ketoglutarate (α-KG) to 2-hydroxyglutarate (2HG)⁵. 2HG inhibits α-ketoglutarate-dependent enzymes such as histone demethylases (KDMs) and cytosine dioxygenases [ten-eleven translocation (TET) methylcytosine dioxygenases], inducing epigenetic reprogramming of the tumor transcriptome⁶. IDH1^R132H is associated with a better prognosis in gliomas^4,7,8 and impacts multiple cellular functions, including cell differentiation⁹, invasion¹⁰, immune response^11–13, and responses to cellular stress and DNA damage^14–16. Previous studies demonstrated that the epigenetic reprogramming in mIDH1 results in enhanced DNA-damage responses and impaired response to radiotherapy^15–18. Furthermore, mIDH1 was shown to elicit cellular metabolic reprogramming that could modify the energetic state of mIDH1 glioma cells and the tumor microenvironment (TME)^16,19–22. The metabolic alterations in mIDH1 gliomas have been associated with a less aggressive tumor phenotype^16,23–25. Mutant IDH1 glioma cells lose the ability to use NADPH-reducing equivalents and gain NADPH-coupled α-ketoglutarate-reducing activity²⁶, resulting in a change in the cellular NADPH to NADP+ ratio⁵. This leads to enhanced susceptibility to changes in the cellular redox status and oxidative stimuli in mIDH1 glioma cells^27–30.

Autophagy is a conserved process that allows the recycling of intracellular components, including damaged organelles, protein aggregates, and cellular waste, maintaining energetic homeostasis and protecting cells against stress³¹. The autophagic pathway involves the generation and degradation of autophagic vesicles, where the cellular material is collected, contained, processed, and recycled³². This phenomenon occurs in several steps: initiation, expansion, fusion, and degradation, and involves numerous autophagy-related genes (ATG) that participate in an orchestrated assembly, driving the autophagic flux^33,34. This flux can be activated via cell signaling in response to stimuli such as starvation, cellular stress, or DNA damage^32,35,36. This physiological process plays crucial roles during development, and its deregulation is involved in disease progression, including Parkinson’s disease, diabetes, and cancer^37–41. While previous studies have examined autophagy in gliomas, including IDH1 wild-type glioblastomas and IDH-mutant gliomas^14,42–44, the role of autophagy and the mechanisms regulating it in mIDH1 astrocytomas harboring concomitant loss of TP53 and ATRX remain largely underexplored. For instance, Gilbert et al., 2014, showed that mutant IDH1/2HG can induce oxidative stress, autophagy, and apoptosis¹⁴. In addition, Viswanath et al. showed that 2HG induces endoplasmic reticulum (ER) stress which mediates autophagic degradation of the ER via a mechanism termed ER-phagy⁴². They also demonstrated that autophagic flux was higher in both WT-IDH1 cells exposed to 2HG and mIDH1 cells⁴².

In this study, we demonstrate that autophagy is activated and plays a critical role in limiting DNA-damaged induced cell death and supports energetic homeostasis in mouse mIDH1 glioma models and human mIDH1 gliomas co-expressing TP53 and ATRX inactivating mutations^2,15. Through epigenetic, transcriptomic, metabolomic, and signaling pathway analysis, we show that the epigenetic changes in mIDH1 gliomas, with TP53 and ATRX loss of function mutations, modulate the expression of several genes that functionally impact mitochondrial metabolism and upregulate autophagy. We demonstrate that targeting the autophagy pathway by synthetic protein nanoparticles (SPNPs) incapsulating siRNA against ATG7 enhance radiosensitivity in mIDH1 mouse glioma model, prolong survival, and result in a sustained immunological memory^45,46. Our findings highlight that combining autophagy inhibition with ionizing radiation leads to increased survival of mIDH1-glioma-bearing mice.

Results

Transcriptomic reprogramming in mIDH1 glioma impacts autophagy and mitochondrial metabolic pathways

For this study, we used the Sleeping Beauty Transposon system to create three genetically engineered mIDH1 mouse glioma models which include the following genetic lesions: (i) NPA/NPAI: NRAS-G12V, shTP53, shATRX, and IDH1-R132H; and two RAS independent mouse glioma models: (ii) CPA/CPAI: CDKN2A -/-, shTP53, shATRX, and IDH1-R132H; and (iii) RPA/RPAI: PDGFRα-D842V, shTP53, shATRX, and IDH1-R132H. We also employed human mIDH1 glioma cell cultures obtained from surgical biopsies and tumor tissue for our “omics” and functional studies. To understand the biological significance of the transcriptomic alteration in mIDH1 glioma, we first analyzed the RNA-seq data generated from our previously described mIDH1 mouse glioma model (NPA/NPAI; Fig. 1a)¹⁵. We specifically focused on the differential expression of genes related to cellular metabolism and energetic homeostasis. We observed that mIDH1 neurospheres (NS) had upregulated genes related to autophagy gene ontologies (Fig. 1b; Supplementary Table 1). In addition, these cells have a profound downregulation of electron transport chain (ETC) complex genes (Fig. 1c; Supplementary Table 2), suggesting impaired mitochondrial function and bioenergetics. This revealed a downregulation of mitochondrial metabolism pathways and an enrichment in autophagy (Fig. 1d). To confirm the differences in the gene ontologies (GOs) related to autophagy observed in our RNA-seq data, we also analyzed scRNA-seq data obtained from the primary tumors of mIDH1 glioma patients¹². We found alterations in GOs related to several autophagy genes including ULK1, UVRAG, ATG4b, and ATG7 (Fig. 1e–h). Additionally, we have analyzed RNA-seq data from The Cancer Genome Atlas Program (TCGA) and found significant differences in mRNA expression in UVRAG, ATG9b, ATG7, and ULK1 in mIDH patients when compared to WT-IDH patients (Supplementary Fig. 1). To determine the effect of mIDH1 on cell metabolism, we performed mass spectrometry-based metabolomics analysis (Supplementary Fig. 2). We observed that the disrupted mitochondrial respiration gene program (Fig. 1i–k; and Supplementary Figs. 3 and 4) is accompanied by the accumulation of TCA cycle metabolites (Fig. 1k), suggesting suppressed pathway activity. Consistently, TCA cycle anaplerotic substrates such as aspartate and glutamate (Fig. 1j and Supplementary Fig. 5), also accumulated in the mIDH1 cells, supporting that mIDH1 disrupts metabolic activities in the glioma NS. In addition, we measured several TCA cycle metabolites including citric acid, succinic acid, α-ketoglutarate, malic acid, and glutamine synthase in two different mouse cell lines: RPA/RPAI and NPA/NPAI (Supplementary Fig. 6). Our findings show that in both genetically engineered mouse glioma models, there is a significant increase in TCA cycle metabolites in the mIDH1 cells when compared to the WT-IDH1 cells (Supplementary Fig. 6).

Fig. 1. mIDH1 glioma exhibits upregulated autophagy and disrupts mitochondrial metabolism.

a Illustration of genetically engineered mIDH1 glioma models harboring ATRX and TP53 knockdown and molecular analysis performed. Scale bar = 20 µm. Heatmaps showing RNA-seq¹⁵ gene expression patterns for b autophagy and c electron transport chain (ETC) genes in WT-IDH1 and mIDH1 neurospheres (n = 3 biological replicates; upregulated genes are red; downregulated genes are blue; FDR ≤ 0.05; ≥1.5-fold). d GSEA comparing transcriptional changes in mIDH1 NS versus WT-IDH1 NS. Negative/positive normalized enrichment scores (blue/red scale; FDR < 0.05) show GO terms linked to metabolism (mitochondrial, OXPHOS) (downregulated) and autophagy (upregulated) in mIDH1 NS. e–h Autophagy-related gene expression of human scRNA-seq from glioma patients¹². Human glioma scRNA-seq was analyzed for the expression of ULK1, UVRAG, ATG4B, and ATG7 in WT-IDH1 (n = 8 specimens) and mIDH1 glioma tumor cells (n = 10 specimens). Violin plots represent the expression of e ULK1, f UVRAG, g ATG4B, and h ATG7. Two-tailed unpaired t-test. i Illustration showing TCA cycle intermediates and amino acids disrupted in mIDH1 NS versus WT-IDH1 NS. Significantly altered amino acids depicted in red. Metabolomics profiling data showing relative levels of j amino acids and k TCA cycle intermediates in WT and mIDH1 NS (arb. units: arbitrary units). Results expressed as mean ± SD. (n = 4 biological replicates; two-tailed unpaired t-test; ns = not significant). Aspartate: p < 0.0001; Methionine: p = 0.0003; Proline: p = 0.0548; Glutamate: p = 0.0002; Tyrosine: p = 0.0005; Leusine: p = 0.0039; Serine: p = 0.0330. Citrate: p < 0.0001; Aconitate: p < 0.0001; α-KG: p = 0.0323; Succinate: p < 0.0001; Malate: p < 0.0001. Source data are provided as a Source Data file.

Histone hypermethylation in mIDH1 glioma NS is associated with metabolism and autophagy pathways

To evaluate the impact of epigenetic reprogramming mediated by mIDH1 on tumor metabolism and autophagy, we used previously published ChIP-seq data generated through native ChIP followed by sequencing (Fig. 2a)¹⁵. To determine the impact of the mutation in IDH1 on histone marks’ hypermethylation levels, we measured H3K4me3, H3K27me3, and H3K36me3 via western blot in four different glioma cell model systems, including human mIDH1 glioma cells obtained from surgical biopsies (SF10602) (Supplementary Fig. 7)⁴⁷. Mutant IDH1 glioma models tested were as follows: (i) NPA/NPAI mouse NS, (ii) RPA/RPAI mouse NS, (iii) WT-IDH1 SJGBM2/mIDH1 SJGBM2 human glioma cells, and (iv) SF10602 human glioma cells treated with and without mIDH1 inhibitor AG-881. Our results show a significant increase in the histone marks H3K4me3, H3K27me3, and H3K36me3 in mIDH1 cells when compared to WT-IDH1 cells (Supplementary Fig. 7). We previously demonstrated the efficacy of mIDH1 inhibitors at decreasing the production of 2HG in both mouse and human mIDH1 glioma cells^15,18. We then proceeded to evaluate the differential enrichment at promoter regions of H3K4me3, a transcriptional activation mark, and of H3K27me3, a transcriptional repressive mark, in mIDH1 compared to WT-IDH1 NS (Fig. 2b, c). Differential H3K4me3-enriched GO terms included regulation of several cell metabolic processes (Fig. 2b), resulting in epigenetic regulation of these metabolic pathways in mIDH1 cells. GO terms enriched in H3K27me3 histone indicate epigenetic repression in mIDH1 versus WT-IDH1 glioma NS (Fig. 2c). To better understand the relationship between epigenetic regulation and autophagy in mIDH1 glioma, we conducted comprehensive CUT&Run analyses that demonstrated autophagy enrichment, including differential deposition of H3K4me3 and H3K27ac marks at promoter and regulatory regions of genes which mediate autophagy (Fig. 2d). Of note, we observed differential peaks’ deposition for H3K4me3, a histone mark classically associated with transcriptional activation, around the promoter and genomic regions of the autophagy-related gene ATG9b in mIDH1 glioma NS (Fig. 2e)^48,49. We also observed differential enrichment of H3K27ac, a marker of active enhancers, at regulatory regions in mIDH1 NS compared to WT-IDH1 NS (Fig. 2e). Additionally, these results were validated by CUT & RUN -qPCR using RPA and RPAI cells. We found that H3K4me3 and H3K27ac were significantly enriched in the promoter and enhancer regions (boxes 2 and 3) in mIDH1 NS versus NS for Atg9b (Supplementary Fig. 8). Consistent with our previous results, our Bru-seq analysis performed on mIDH1 NS showed a higher transcription rate for ATG9b (>2-fold) versus WT-IDH1 NS (Fig. 2f)¹⁵. Further analysis revealed that mIDH1 cells express significantly higher levels of ATG9b protein in both mouse and human mIDH1 glioma cells (Fig. 2g, h; Supplementary Fig. 9). These results indicate that mIDH1, through epigenetic reprogramming, alters metabolic and autophagic pathways in glioma cells (Fig. 2i).

Fig. 2. Epigenetic reprogramming in mIDH1 glioma alters metabolic pathways and induces an autophagy regulatory network.

a Native H3K4me3 ChIP-seq diagram performed on mIDH1 NS and WT-IDH1 NS (n = 3 biological replicates)¹⁵. b, c Genes enriched in H3K4me3 or H3K27me3 linked to distinct functional GO terms by ChIP-seq analysis. Bar graphs representing GO terms containing genes having enrichment of H3K4me3 (red) or H3K27me3 (green) at their promoter regions in mIDH1 NS. GO terms’ significance determined by FDR (<0.05); enrichment expressed as OR. d Genes enriched in H3K4me3 or H3K27ac linked to distinct functional GO terms by CUT&Run (n = 1 biological replicate). e H3K4me3 and H3K27ac occupancy in specific genomic regions of Atg9b. The y-axis represents the estimated number of immunoprecipitated fragments normalized to the total number of reads in each dataset. Reference sequence (RefSeq) gene annotations are shown. f Bru-Seq traces showing differential transcriptional rates (<1.5-fold; P < 0.05) of Atg9b in mIDH1 NS (blue; n = 12 biological replicates) versus WT-IDH1 (red; n = 3 biological replicates). Arrow indicates sequence strand reading direction; positive y-axis represents positive strand signal of transcription moving left to right; negative y-axis represents negative strand signal of transcription moving right to left. Vertical red mark indicates Atg9b position within chromosome. Genes and chromosome locations indicated on maps. Data expressed in reads per kilobase per million mapped reads (RPKM). Gene maps generated from RefSeq¹⁵. g WB showing ATG9b protein expression in WT-IDH1 NS, mIDH1 NS human glioma SJGBM2 WT-IDH1 and mIDH1, and mIDH1 human glioma cells, SF10602, ±IDH1^R132H inhibitor AG-881 (250 nM). h ImageJ densitometric quantification of the western blots for ATG9b and β-actin shown in (g) (n = 3 technical replicates). All western blots were independently repeated 3 times to ensure experimental rigor. i Diagram of initial steps of autophagy pathways, including phagophore formation with participation of Atg9b protein in the nucleation process. Source data are provided as a Source Data file.

Epigenetic reprogramming in human mIDH1 glioma cells impacts autophagy-related genes

In addition to ATG9b, the formation of phagophores and autophagosomes involves several other genes, including UVRAG, STK26 (MST4), ATG7, and MAP1LC3B (Fig. 3a). To further validate the role of mIDH1 in the epigenetic regulation of these genes, we performed ChIP-seq analysis on human mIDH1 SF10602 glioma cells, in the presence or absence of the mIDH1 inhibitor (Fig. 3b)¹⁸. Our results showed that the specific inhibition of mIDH1 alters peak enrichment for H3K36me3, H3K4me3, or H3K27ac marks (Fig. 3c–i) around promoter and enhancer regions of UVRAG (Fig. 3c), ATG7 (Fig. 3d), MAP1LC3B (Fig. 3e, f), STK26 (MST4) (Fig. 3g), and ATG9b (Fig. 3h, i). H3K36me3 was included in the analysis as it is a histone mark associated with transcriptional elongation and gene body regulation and is sensitive to α-KG–dependent dioxygenases, which are inhibited by 2HG in IDH-mutant cells⁵⁰.

Fig. 3. Epigenetic regulation of autophagy-related genes involves histone methylation in mIDH1 human glioma cells.

a Diagram of autophagosome formation in the autophagic pathway involving autophagy-related genes: UVRAG, Atg7, LC3 (MAP1LC3B), and MST4 (STK26). b Diagram of experimental design of ChIP-seq analysis performed in primary mIDH1 human glioma cells, SF10602, ±IDH1^R132H inhibitor AGI-5198 (5 µM), prior to ChIP¹⁸. c–f H3K36me3 and g–i H3K4me3 and H3K27ac occupancy in specific genomic regions of autophagy-related genes UVRAG, Atg7, MAP1LC3B (LC3), STK26 (MST4), and Atg9b. The y-axis of each profile represents the estimated number of immunoprecipitated fragments at each position normalized to the total number of reads in each dataset. RefSeq gene annotations are shown. Differential peaks (FDR < 0.05) in untreated SF10602 mIDH1 glioma cells are represented in blue compared to AGI-5198-treated cells in red (n = 3 biological replicates per group).

Single-cell RNA-seq shows an enhancement of autophagy in human mIDH1 glioma cells

To assess the impact of epigenetic reprogramming mediated by mIDH1 on autophagy, scRNA-seq analysis was conducted on SF10602 cells, under both the presence and absence of the mIDH1 inhibitor AGI-5198 (Fig. 4a)¹². The examination included an evaluation of enriched gene ontology (GO) terms associated with autophagy. The results demonstrated that the inhibition of mIDH1 led to decreased positive regulation of autophagy (Fig. 4b, c), increased negative regulation of autophagy (Fig. 4d; Supplementary Fig. 10a), decreased cellular components of the autophagosome (Fig. 4e; Supplementary Fig. 10b), and decreased processes utilizing autophagic mechanisms (Fig. 4f, g). The observed changes in gene ontology terms associated with autophagy highlight a significant influence of mIDH1 on the regulation of autophagy-related processes. The specific alterations identified, such as the modulation of positive and negative regulation of autophagy, as well as changes in cellular components and processes related to autophagosome function, provide valuable insights into the intricate relationship between IDH1 mutation and the dysregulation of autophagy pathways in glioma. Taken together, these findings suggest a link between IDH1 mutation and the upregulation of autophagy-related pathways in glioma.

Fig. 4. Single-cell RNA-seq shows an enhancement of autophagy pathways in human mIDH1 glioma cells.

a Diagram of experimental design of scRNA-seq analysis performed in primary mIDH1 human glioma cells, SF10602, ±IDH1^R132H inhibitor AGI-5198, prior to sequencing (n = 2 biological replicates). b GO term enrichment score for GO biological process of positive regulation of autophagy. c Violin plot of scRNA-seq in human mIDH1 glioma cells related to GO enrichment score for GO biological process of positive regulation of autophagy. NPC neural progenitor cells, OPC oligo progenitor cells, AC astrocytic cells, MES mesenchymal. The center line of violin plots denotes the median, and the plot width represents the kernel density estimate of the distribution. d GO term enrichment score for GO biological process of utilizing the autophagy mechanism. e Violin plot of scRNA-seq in human mIDH1 glioma cells related to GO term enrichment score for GO biological process of utilizing the autophagy mechanism. NPC neural progenitor cells, OPC oligo progenitor cells, AC astrocytic cells, MES mesenchymal. The center line of violin plots denotes the median, and the plot width represents the kernel density estimate of the distribution. GO term enrichment score for GO cellular component of f negative regulation of autophagy and g autophagosome. Significance of the violin plots was measured via the Wilcoxon Rank Sum Test from the Seurat/Presto package. All UMAPs show n = 21,656 total cells (n = 9968 from Control and n = 11,688 from IDH inhibitor treatment).

Mutant IDH1 NS exhibit altered mitochondrial activity and slower proliferative rate

To gain further insight into the metabolic activity of mIDH1 glioma cells, we used the Seahorse extracellular flux analyzer to compare the bioenergetic profiles of primary NS isolated from WT-IDH1 and mIDH1 tumors. Given our transcriptomics analysis indicating dysregulated mitochondrial activity and the accumulation of TCA cycle metabolites, we performed the Mito Stress Test to assess differences in mitochondrial function and metabolic profile between WT-IDH1 and three different mIDH1 NS clones. When compared to WT-IDH1, mIDH1 NS exhibited lower oxygen consumption rate (OCR) at baseline, suggesting mIDH1 glioma cells utilize mitochondrial respiration to a lesser extent compared to that of WT-IDH1 cells (Fig. 5a). In line with this finding, the maximal rate of respiration in mIDH1 NS was lower compared to that of WT-IDH1 cells, as seen by the reduction in OCR following treatment with the mitochondrial uncoupler 4-(trifluoromethoxy) phenylhydrazone (FCCP) (Fig. 5a and Supplementary Fig. 11a). Additionally, examination of glycolytic activity, measured by the extracellular acidification rate (ECAR), revealed that mIDH1 NS exhibit a significant reduction (~50%) in ECAR compared to WT-IDH1 counterparts at baseline (Fig. 5b and Supplementary Fig. 11b). Together, these findings demonstrate that mIDH1 NS exhibit less mitochondrial respiration and glycolytic activity indicative of slower proliferative rate when compared to the WT-IDH1 NS (Fig. 5c).

Fig. 5. mIDH1 glioma cells exhibit a slower proliferative state.

a Mitochondrial respiration of glioma NS derived from WT-IDH1 or mIDH1 tumors (Clones 1–3) examined through Mito Stress Test. Error bars depict SD of 3 biological replicates. Mitochondrial respiration determined by oxygen consumption rates (OCR). b Glycolytic activity determined by extracellular acidification rates (ECAR) in mIDH1 NS and WT-IDH1 NS. c Metabolic profile of WT-IDH1 NS and mIDH1 NS comparing ECAR versus OCR. d Mito Stress Test performed on WT-IDH1 NS, mIDH1 NS, and WT-IDH1 NS cultured with cell-permeable (R)-2HG (-NS+O-2HG) for 4 h at indicated concentrations. Error bars depict SD of 3 biological replicates from a representative plot of 3 independent experiments. e ECAR of WT-IDH1 NS, mIDH1 NS, and WT-IDH1 NS+O-2HG. f Metabolic profile of WT-IDH1 NS, mIDH1 NS, and WT-IDH1 NS+O-2HG, comparing ECAR versus OCR. g Diagram representing mitochondrial complex 1 (C1) inhibition by metformin and rotenone. mGPDH mitochondrial glycerol-3-phosphate dehydrogenase. Cell viability assay of WT-IDH1 NS and mIDH1 NS (n = 3 biological replicates) treated with h metformin and i rotenone for 72 h at increasing concentrations. Cell viability expressed in relative luminescence units (RLU) as a percentage of control (100%). Linear regression model and generalized additive models (GAM) for flexible regression modeling were used to model the dose-response relationship. Two-tailed unpaired t-test. Cell viability assay showing the effect of j, k Metformin at 30 μM doses or l, m Rotenone at 2.5 µM in combination with radiation (3 Gy) on cell proliferation in j, l WT-IDH1 and k, m mIDH1 mouse NS. Results expressed in relative luminescence units (RLU). Two-tailed unpaired t-test; error bars depict SEM of 3 biological replicates; ns = not significant. Cell viability assay showing the effect of n Metformin and o Rotenone at increasing concentrations in combination with radiation (20 Gy) on the proliferation of AG-881-treated SF10602 and untreated SF10602 cells. Results expressed as % of Cell survival relative to control populations (two-tailed unpaired t-test; error bars depict SEM of 3 biological replicates; ns not significant). Source data are provided as a Source Data file.

Mutant IDH1 glioma cells promote the conversion of α-ketoglutarate to (R)-2-hydroxyglutarate (2HG), an inhibitor of α-ketoglutarate-dependent enzymes. This can modify the transcriptome of tumor cells, and therefore indirectly influence metabolism, in a process that typically requires multiple days and cell divisions⁵¹. In contrast, 2HG can also directly impact metabolism by inhibiting metabolic enzymes that use α-ketoglutarate as a substrate, including but not limited to transaminases. A recent study demonstrated that 2HG can inhibit BCATs, a mechanistic understanding of which identified additional metabolic dependencies in IDH-mutant brain tumors⁵². To determine whether 2HG influences the metabolic profile in our models of mIDH1 tumor cells, WT-IDH1 NS were cultured with cell-permeable 2 R Octyl-alpha-hydroxyglutarate (O-2HG) for 4 h, after which we performed Seahorse extracellular flux analysis. WT-IDH1 NS supplemented with O-2HG appeared more similar in basal and maximal respiration to mIDH1 (Fig. 5d), whereas glycolytic activity was impacted to a lesser degree (Fig. 5e). This data illustrates that 2HG supplementation can directly alter the energetic state of glioma NS (Fig. 5f), which likely contributes in part to the observed metabolic alterations in mIDH1 cells.

Based on the metabolic defects observed with exogenous 2HG, we then evaluated if mIDH1 glioma cells are more susceptible to mitochondrial inhibitors than those expressing WT-IDH1. To evaluate the sensitivity of mIDH1 to mitochondrial inhibition, we treated mouse WT-IDH1 and mIDH1 cells with different concentrations of the mitochondrial complex 1 (C1) inhibitors, metformin and rotenone (Fig. 5g), after which we evaluated cell viability (Fig. 5h–m). Results showed that mIDH1 glioma cells were more sensitive to rotenone and metformin treatment than WT-IDH1. Also, inhibition of mitochondrial complex I renders the mIDH1 mouse glioma cells more susceptible to radiation (Fig. 5j–m). We found a significant difference in cell viability when the mIDH1 glioma cells are treated with either rotenone or metformin in combination with IR compared to radiation alone (Fig. 5j–m). In addition, we evaluated the sensitivity of two human mIDH1 glioma cell cultures to treatment with mitochondrial complex 1 inhibitors, metformin and rotenone, measuring cell viability (Fig. 5n, o and Supplementary Fig. 12). Our results showed that mIDH1 SJGBM2 cells and SF10602 cells are more sensitive to metformin alone and metformin in combination with IR (Fig. 5n and Supplementary Fig. 12), and to rotenone alone and rotenone in combination with IR (Fig. 5o and Supplementary Fig. 12).

IDH1^R132H is associated with increased levels of autophagy-related proteins in glioma cells

Our results described above indicate that mIDH1 glioma cells have disrupted mitochondrial activity with a consequently lower energetic charge when compared with WT-IDH1 glioma cells (Fig. 5). This is in line with previous studies indicating that metabolite changes associated with IDH1 mutation alter the energetic state of glioma cells^23,25,53. However, mIDH1 glioma cells can grow and develop tumors in the brains of animals, indicating that they could use alternative mechanisms to generate energy and maintain cell survival. According to our RNA-seq, CHIP-seq, and CUT&RUN data, autophagy is upregulated in mIDH1 glioma cells in correlation with a functional disruption of mitochondrial activity. We then evaluated autophagy activity by protein expression and phosphorylation state of key regulators in the autophagic pathway, including pULK1 (S555) and pULK1 (S757), which are associated with autophagy activation and inhibition, respectively. We also evaluated expression of UVRAG, MST4, and pATG4 that participate in the nucleation and expansion process, involved in autophagosome formation in both mIDH1 and WT-IDH1 glioma NS (Fig. 6a). Western blot (WB) analysis revealed that in two different mouse cells (NPA/NPAI and RPA/RPAI) and human glioma cells, IDH1 mutation results in increased expression of pULK1 (S555) but decreased expression of pULK1 (S757), when compared with WT-IDH1 glioma cells (Supplementary Fig. 13). Moreover, mIDH1 glioma cells showed upregulation of p62, UVRAG, MST4, pATG4b⁵⁴, and ATG7 (Fig. 6b–q; see quantification in Supplementary Fig. 14). Interestingly, inhibition of mIDH1 using AG-881 in mIDH1 SF10602 glioma cells, mIDH1 MGG119 glioma cells, and mIDH1 LC1035 glioma cells reverted the expression pattern of all the key autophagy regulators evaluated (Fig. 6n–q; Supplementary Fig. 14g, h; Supplementary Fig. 15). Also, when we treated three different mIDH1 mouse glioma cell models and human mIDH1 SF10602 glioma cells with α-ketoglutarate (α-KG) (Fig. 7a), cell viability was significantly reduced in the α-KG-treated mIDH1 glioma cells when compared to untreated (Supplementary Fig. 16). In addition to reduced cell viability, α-KG rendered both mouse and human mIDH1 glioma cells sensitive to radiation (Supplementary Fig. 16). We then measured histone trimethylation status via western blot in two mouse mIDH1 glioma NS supplemented with and without α-KG and found that untreated mIDH1 cells had a significant increase in the histone marks H3K4me3, H3K27me3, and H3K36me3 when compared to α-KG supplemented cells (Supplementary Fig. 17). Furthermore, we found that the mIDH1 glioma cells reverted the expression patterns showing significant reduction in several key autophagy regulators including pATG4b, UVRAG, ATG7, ATG9b, and p62 (Fig. 7b–k; Supplementary Fig. 18).

Fig. 6. Enhanced autophagy activity in mIDH1 glioma cells.

a The autophagy pathway is responsible for the recycling of cellular residues and damaged organelles, thus maintaining homeostasis. This process includes initiation, expansion, and degradation steps, where several proteins participate in the formation and transformation of the different cellular structures involved (phagophores, autophagosomes, lysosomes). CQ chloroquine. WB assay showing the expression of proteins involved in the autophagy pathways and their activated form of them, evaluated in b–e mIDH1 and WT-IDH1 NPA mouse NS, f–i mIDH1 and WT-IDH1 RPA mouse NS, j–m human glioma SJGBM2, and in n–q mIDH1 SF10602 human glioma cells ±IDH1^R132H inhibitor AG-881. The red arrows indicate the proteins and phosphorylated (p) status related to autophagy activation (p62, UVRAG, MST4, pATG4B (S383), ATG7); β-actin, loading control. Bar graph represents densitometric analysis of the ratio of c, g, k, o p62, d, h, l, p pATG4b/ATG4b, and e, i, m, q ATG7/β-actin WB assay (n = 3 technical replicates). All western blots were independently repeated 3 times to ensure experimental rigor. Source data are provided as a Source Data file.

Fig. 7. Impact of mIDH1 on modulating the autophagy pathway.

(a) Diagram showing the method and sample processing for α-ketoglutarate-treated mIDH1 glioma cells prior to western blot analysis. Cell culture flask and Pipette from NIAID Visual & Medical Arts. (10/7/2024). (Low Friction Flask: bioart.niaid.nih.gov/bioart/303; Pipette: bioart.niaid.nih.gov/bioart/408). b–k Western blot measurement of proteins involved in the autophagy pathway of either NPAI or RPAI mIDH1 mouse NS treated with α-ketoglutarate. WB assay showing the expression of proteins involved in the autophagy pathways and their activated form of them, evaluated in b mIDH1 NPAI mouse NS ± treatment with α-ketoglutarate (α-KG) or g mIDH1 RPAI mouse NS ± treatment with α-KG. β-actin, loading control. ImageJ densitometric quantification of western blot assays of c pATG4b, d UVRAG, e ATG7, and f ATG9b in the mIDH1 NPAI mouse NS treated with α-KG. ImageJ densitometric quantification of western blot assays of h pATG4b, i UVRAG, j ATG7, and k ATG9b in the mIDH1 RPAI mouse NS treated with α-KG (n = 3 technical replicates). l Diagram indicating the role of LC3 in autophagy expansion. m–p WB assay showing the conversion of LC3I to LC3II as an indicator of autophagy activation. The increased ratio of LC3II/LC3I (red arrows) indicates autophagy activation. β-actin is a loading control. The bar graph represents densitometric analysis of the ratio of LC3II/LC3I WB assay (n = 3 technical replicates). All western blots were independently repeated 3 times to ensure experimental rigor. Source data are provided as a Source Data file.

We then explored the impact of 2HG on both histone marks and autophagy-related proteins in mouse WT-IDH1 glioma NS (Supplementary Figs. 19 and 20). We found that WT-IDH1 glioma NS supplemented with O-2HG had significantly increased trimethylated histone marks (H3K4me3, H3K27me3, and H3K36me3) when compared to untreated WT-IDH1 glioma NS (Supplementary Fig. 19), indicating a direct correlation between 2HG and histone modification. Furthermore, we found that WT-IDH1 glioma NS supplemented with O-2HG showed significantly increased levels of key autophagy regulators (pULK1 (S555), UVRAG, p62, and ATG9b) when compared to untreated WT-IDH1 glioma NS (Supplementary Fig. 20), indicating a direct correlation between 2HG and increased autophagy.

Another important marker of autophagy activation is the conversion of LC3I to LC3II (Fig. 7l). A cytosolic form of LC3 (LC3I) is cleaved by ATG4b then conjugated to phosphatidylethanolamine to form LC3-phosphatidylethanolamine conjugate (LC3II), which is recruited to autophagosomal membranes⁵⁵ during the expansion step of autophagy (Fig. 7l). Thus, the presence and increased intracellular ratio of LC3II over LC3I is a signal of autophagy activation. Results showed that in mIDH1 glioma cells, there was more than a 4-fold increase in LC3II/LC3I in both human and mouse glioma cells (Fig. 7m–o). As expected, blocking mIDH1 results in a 2-fold decrease in the LC3II/LC3I ratio (Fig. 7p). These results indicate that IDH1 mutation is associated with autophagy activation in gliomas. When we supplemented mIDH1 glioma NS with α-KG, we found a significant decrease in the LC3II/I ratio when compared to untreated mIDH1 glioma NS (Supplementary Fig. 21), indicating decreased autophagy. Furthermore, when we supplemented WT-IDH1 glioma NS with O-2HG, we found a significant increase in the LC3II/I ratio when compared to untreated WT-IDH1 glioma NS (Supplementary Fig. 22), indicating increased autophagy.

Autophagy flux is enhanced in mIDH1 glioma cells

Autophagy is a continuous process that ends in a recycling of cellular materials (Fig. 6a). During autophagic flux, autophagosomes fuse with lysosomes, forming autolysosomes, which degrade autophagosome contents⁵⁶. Using a reporter system where LC3 is fused to both GFP and mCherry (Supplementary Fig. 23)⁵⁷, we found that mIDH1 glioma cells had an increased number of both autophagosomes and autolysosomes when compared with WT-IDH1 glioma cells (autophagosomes >10-fold increase, P < 0.001; autolysosomes >1.4-fold increase in mIDH1 glioma cells, P < 0.001). The increase in autophagosomes and autolysosomes could reflect increased autophagy or stalled autophagy. To discern between these two fates, we blocked autophagy with chloroquine (CQ), and we observed an increase in the accumulation of LC3-GFP in mIDH1 glioma cells (autophagosomes >1.6-fold increase in CQ-treated mIDH1 glioma cells, P < 0.01; no significance observed in autolysosomes in CQ-treated mIDH1 cells vs. CQ-treated WT-IDH1 glioma cells) (Fig. 8a–j). In this assay, CQ inhibits autolysosome formation, allowing autophagosomes to form and accumulate within the cell. We then measured cell viability of two mouse mIDH1 glioma NS and one mIDH1 human glioma cell (SF10602) treated with and without CQ (Supplementary Fig. 24) and hydroxychloroquine (HCQ) (Supplementary Fig. 25) and with and without radiation. Our data showed that both CQ and HCQ significantly reduced cell survival in both mouse and human mIDH1 glioma cells and rendered the cells sensitive to radiation (Supplementary Figs. 24 and 25). Together, these results indicate that glioma cells harboring IDH1^R132H have an augmented autophagy flux when compared with WT-IDH1 glioma cells.

Fig. 8. mIDH1 increases autophagy flux and in vitro inhibition of autophagy restores radiosensitivity in mIDH1 glioma cells.

a–h Representative images of GFP and mCherry fluorescence in WT and mIDH1 human glioma cells, SJGBM2, in basal conditions or treated with 10 µM chloroquine (CQ). Scale Bar: 10 µm. Quantification of i mCherry-GFP foci (indicating autophagosome formation) or j mCherry-only foci (indicating autolysosome formation) in WT and mIDH1 SJGBM2 cells, in basal conditions or treated with 10 µM chloroquine. Bar graphs represent the average total foci per cell means ± SEM; two-tailed unpaired t-test (untreated WT-IDH1 cells = 38 biological replicates; untreated mIDH1 cells = 26 biological replicates; WT-IDH1 cells + CQ = 10 biological replicates; mIDH1 cells +CQ = 42 biological replicates). k NSC185058 compound inhibits Atg4b. Cell viability assay of l mIDH1 mouse NS (n = 8 biological replicates), m human glioma cells, SJGBM2 (n = 4 biological replicates), and n SF10602 treated with NSC185058 and IR (n = 4 biological replicates). Results expressed as percent of RLU in relation to the untreated control (100%); two-tailed unpaired t-test. o ULK101 compound inhibits ULK1 during autophagy initiation. Cell viability assay of p mIDH1 mouse NS, q SJGBM2, and r SF10602 treated with ULK101 and IR. Results expressed as percent of RLU in relation to the untreated control (100%). Error bars represent SEM from 4 biological replicates; two-tailed unpaired t-test. s ATG7 inhibition by siRNA (siATG7). Cell viability assay performed in t mIDH1 mouse and u SJGBM2 and v SF10602 treated with siATG7 and IR. Results expressed as percent of RLU in relation to the untreated control (100%). Error bars represent SEM from 4 biological replicates; two-tailed unpaired t-test. w ATG4 inhibition by siRNA (siATG4). Cell viability assay performed in x mIDH1 mouse NS, y SJGBM2, and z SF10602 treated with siATG4b and IR. Results expressed as percent of RLU in relation to the untreated control (100%). Error bars represent SEM from 4 biological replicates; two-tailed unpaired t-test. Source data are provided as a Source Data file.

Inhibition of autophagy sensitizes mIDH1 glioma cells to ionizing radiation

Autophagy has been associated with DNA-damage response activation and resistance to ionizing radiation (IR) in tumor cells⁵⁸. We and others have previously reported that mIDH1 gliomas are radioresistant^15,16. We hypothesized that the inhibition of autophagy will impair a proper response against a DNA-damage insult in mIDH1 glioma cells, making them more sensitive to IR treatment. To test this hypothesis, we performed a cell viability assay in glioma cells harboring mIDH1, treated with multiple autophagy inhibitors in the presence or absence of irradiation (Fig. 8k–r). We first tested the effect of NSC185058, which targets ATG4b (Fig. 8k)⁵⁴. Treatment with NSC185058 or irradiation alone decreased IDH1 glioma cell viability compared to vehicle (Fig. 8l–n). However, combination treatment of NSC185058 with IR dramatically decreased mIDH1 glioma cell viability when compared to control or monotherapy (mIDH1 mouse NS: P < 0.0001, Fig. 8l; SJGBM2: P < 0.001, Fig. 8m; SF10602: P < 0.0001, Fig. 8n). We then tested whether upstream inhibition of the autophagy pathway would be more potent in mIDH1 as a monotherapy. We used ULK101, which inhibits the initial steps of autophagy by targeting ULK1 (Fig. 8o). Monotherapy with ULK101 significantly decreased cell viability in mIDH1 glioma cells (Fig. 8p–r) when compared with control and IR alone (P < 0.0001 in all cells). However, combination therapy of IR and ULK101 resulted in a more dramatic decrease in mIDH1 glioma cells viability when compared to control and monotherapy (mIDH1 mouse NS: P < 0.05; SJGBM2: P < 0.05; SF10602: P < 0.001) or IR alone (mIDH1 mouse NS: P < 0.0001; SJGBM2: P < 0.001; SF10602: P < 0.0001) (Fig. 8p–r). In addition, we evaluated the impact of the molecular inhibition of autophagy on glioma cells’ radio-responses through siRNAs (Supplementary Table 3) against autophagy-related genes ATG7 and ATG4b (Fig. 8s–z) and through the knockdown of ATG7 in mouse glioma cells. Using siRNAs (Supplementary Table 3) against autophagy-related genes ATG7 (Fig. 8s), ATG4b (Fig. 8w), we assessed the viability of both mouse and human glioma cells treated with radiation. We examined the impact of siRNA-ATG7 (Fig. 8t–v) and siRNA-ATG4b (Fig. 8x–z) in the presence or absence of IR. Our observations revealed a significant reduction in viability across all tested cells when treated with a combination of siRNAs targeting ATG7 or ATG4b (mIDH1 mouse NS: P < 0.0001; SJGBM2: P < 0.0001; SF10602: P < 0.0001).

To validate effective silencing of ATG4b, ATG7, and ATG9b following siRNA treatment, mIDH1 mouse NS cells were transfected with the respective siRNAs at a concentration of 100 nM (Supplementary Fig. 26). Western blot analysis revealed significant reduction in protein expression-2.65-fold for ATG4b (P < 0.0001), 2.59-fold for ATG7 (P < 0.0001), and 2.06-fold for ATG9b (P < 0.0001) following 48hrs of siRNA treatment (Supplementary Fig. 26).

We validated these results in two NRAS-independent genetically engineered mouse models (RPAI and CPAI). To this end, we used siRNAs targeting ATG4b (Supplementary Figs. 27a, b) and ATG7 (Supplementary Figs. 27c, d) and assessed the viability of mIDH1 glioma NS treated in combination with radiation. Our results showed a significant reduction in viability across all cells (RPAI mIDH1 mouse NS: P < 0.0001; CPAI mIDH1 NS: P < 0.0001). We also evaluated the impact of inhibiting the autophagy-related gene, ATG9b, by using siRNAs (Supplementary Fig. 28). Cell viability was assessed in both human (SJGBM2 mIDH1) and murine glioma (mIDH1 NPAI, mIDH1 CPAI, and mIDH1 RPAI) cells treated with ATG9b siRNAs, both in the presence and absence of IR. Silencing of ATG9b in mIDH1 human glioma cells (Supplementary Fig. 28a) and mIDH1 mouse NS models (Supplementary Fig. 28b–d) significantly reduced cell viability when combined with IR, compared to siRNA treatment alone (SJGBM2 mIDH1: P < 0.0001; CPAI mIDH1 mouse NS: P < 0.0001; RPAI mIDH1 NS: P < 0.0001; NPAI mIDH1 NS: P < 0.0001). Collectively, these findings indicate that molecular inhibition of autophagy, achieved by targeting distinct proteins involved at different stages of the autophagy signaling pathway, enhances the radiosensitivity of mIDH1 murine and human glioma cells.

In addition, we used a shRNA against ATG7 to generate an autophagy-deficient mouse glioma cell (Supplementary Fig. 29). Like transient autophagy inhibition using siRNA (against ATG7, ATG4b, or ATG9b), mIDH1 autophagy-deficient cells (mIDH1-ATG7KD) showed significant radiosensitization when compared with mIDH1 cells in which the autophagy pathway was not inhibited (P < 0.0001) (Supplementary Fig. 30). The impact of concurrently inhibiting autophagy and applying radiation was assessed in both mouse (Supplementary Fig. 31) and human glioma cells (Supplementary Fig. 32) harboring mIDH1. Our observations revealed a notable shift in cytokine profiling, with a significant increase in cytokine concentration observed in mIDH1 glioma cells (Supplementary Figs. 31 and 32; P < 0.01) upon combining autophagy inhibition with radiation. These findings suggest the potential of the combined treatment to elicit immune responses in mIDH1 glioma cells. In summary, our results collectively propose that autophagy inhibition not only heightens the sensitivity of mIDH1 glioma cells to radiotherapy but also modulates immune responses.

Molecular inhibition of autophagy prolongs the survival of mIDH1-glioma-bearing mice

To study the impact of autophagy inhibition on mIDH1 glioma in vivo, we targeted the autophagy pathway via two approaches: (i) a genetically engineered Sleeping Beauty model that expresses shRNA against ATG7 (shATG7; Fig. 9a and Supplementary Figs. 29, 33); and (ii) pharmacological inhibition of the autophagy pathway using Synthetic Protein Nanoparticles (SPNPs) loaded with an siRNA targeting ATG7 (ATG7i-SPNP; Fig. 9a and Supplementary Figs. 34 and 35). To ensure ATG7i-SPNPs impact glioma cell survival when loaded into SPNPs, we measured cell viability in both mouse and human mIDH1 glioma cells (Supplementary Figs. 36 and 37). We also evaluated the impact of autophagy inhibition on the radiosensitivity of glioma cells using ATG7i-SPNPs (Supplementary Fig. 36). Cell viability was assessed in three mIDH1 glioma murine cell models, NPAI, CPAI, and RPAI, treated with ATG7i-SPNPs, in the presence or absence of IR. Silencing of ATG7 using SPNPs significantly reduced cell viability when combined with IR in all cells tested versus ATG7i-SPNPs treatment alone (RPAI mIDH1: P < 0.0001; NPAI mIDH1: P < 0.001; CPAI mIDH1: P < 0.0001; Supplementary Fig. 36). These findings suggest that using ATG7i-SPNPs to inhibit autophagy enhances the radiosensitivity of mIDH1 glioma cells. We also evaluated the effect of autophagy inhibition on the radiosensitivity of human glioma and patient-derived endogenous mIDH1-expressing SF10602 cells using ATG7i-SPNPs (Supplementary Fig. 37). Silencing of ATG7 significantly reduced cell viability when combined with IR versus ATG7i-SPNP treatment alone (P < 0.0001). These results indicate that inhibiting autophagy through ATG7i-SPNPs enhances the sensitivity of both mouse and human mIDH1 glioma cells to IR.

Fig. 9. In vivo inhibition of autophagy promotes tumor progression in mIDH1 gliomas.

a SB plasmid encoding shRNA targeting Atg7 (shAtg7) and iRGD SPNPs delivering siRNA (ATG7i) developed for Atg7 inhibition in vivo. b Autophagy-deficient mIDH1 mouse gliomas generated using the SB transposon system with plasmids encoding SB-transposase-LUC, NRAS/shP53/GFP, shATRX-GFP, IDH1^R132H/Katushka, and shAtg7/CFP. Transfection efficiency was monitored by luminescence. Symptomatic mouse brain tumors expressing fluorescent proteins were identified macroscopically. LUC: luciferin. c Kaplan–Meier survival analysis for mice bearing mIDH1 (n = 30) or mIDH1 + shATG7 (n = 19) gliomas (p = 0.0053, Mantel–Cox test). MS median survival. d In mIDH1 glioma cells treated with autophagy inhibition alone, the mIDH1 glioma cells express enhanced proliferation markers, increased invasion markers, and increased tumor aggressiveness signaling, which all lead to increased tumor growth. Source data are provided as a Source Data file.

The genetically engineered autophagy-deficient mouse mIDH1 glioma model was developed using the Sleeping Beauty (SB) Transposon system through a combination of SB-transposase/LUC; NRAS/shP53/GFP; shATRX/GFP; IDH1^R132H/Kat and shATG7/CFP plasmids (Fig. 9b). We found that mice with ATG7-KD had a significantly lower median survival (MS) when compared with the control (154 vs. 120 days post-plasmid injection (DPI); P < 0.001) (Fig. 9c), indicating that inhibition of autophagy promotes tumor progression (Fig. 9d).

We then proceeded to assess the impact of autophagy inhibition in combination with radiation treatment using our cell-implantable mouse mIDH1 glioma model¹⁵. We intracranially implanted mIDH1 glioma NS in mice, and 7 days later the animals were split into five groups: (i) saline control; (ii) treated with empty SPNPs; (iii) treated with IR; (iv) ATG7i-SPNP; and (v) treated with IR + ATG7i-SPNP at indicated doses and schedule (Fig. 10a and Supplementary Table 4). In the genetically engineered mouse mIDH1 glioma model, autophagy inhibition alone significantly decreased MS (P < 0.05) when compared with the control group (Fig. 10a and Supplementary Table 4). However, IR + ATG7i-SPNP led to increased MS (MS = 60 DPI; P < 0.001) when compared with all other experimental groups (Fig. 10a and Supplementary Table 4). Taken together, while autophagy inhibition alone promotes tumor progression, it also sensitizes the tumors to radiation (Fig. 10b). We observed that the combined treatment prolonged the survival of mice long-term (>90 DPI in 40% of the treated animals). To ensure that ATG7 was suppressed in vivo in response to ATG7i-SPNP treatment, we performed immunohistochemistry on mIDH1-glioma-bearing mice treated with ATG7i-SPNP (Fig. 10c, Supplementary Fig. 38). We found that mIDH1-glioma-bearing mice treated with ATG7i-SPNPs (Fig. 10c, lower panel) exhibited no ATG7 expression when compared to saline-treated mIDH1-glioma-bearing mice (Fig. 10c, upper panel).

Fig. 10. In vivo inhibition of autophagy radiosensitizes mIDH1 gliomas and enhances survival.

a Autophagy inhibition using ATG7i-SPNP on radio-response in mIDH1 glioma model. Adult mice implanted with 50,000 mIDH1 NS (Day 0). At 7 DPI, animals split into groups: (i) untreated saline control (n = 10); (ii) 2 Gy/day (n = 5); (iii) 9 doses of 2 × 10¹¹ empty SPNPs every other day (n = 5); (iv) 9 doses of 2 × 10¹¹ ATG7i-SPNP every other day (n = 13); (v) 9 doses of 2 × 10¹¹ ATG7i-SPNP every other day and 2 Gy/day (days 7-16) (n = 10). Kaplan–Meier survival analysis performed for each experimental group (Mantel–Cox test). b In mIDH1 glioma cells treated with autophagy inhibition combined with IR, the mIDH1 glioma cells block cellular stress adaptation, increase IR-induced tumor cell death, and increase radiosensitization, leading to tumor shrinking. c mIDH1 glioma NS-implanted mice treated with ATG7i-SPNP were subsequently perfused at 14 DPI and analyzed for IHC. IHC staining of 5 μm paraffin-embedded brain sections from saline and Atg7i-SPNP–treated groups, stained for ATG7. Low-magnification (10×) panels (black scale bar = 100 μm), High-magnification (40×) panels (black scale bar = 20 μm) indicate positive staining for the areas in the low-magnification panels. d Experimental design showing mice implanted with mIDH1-OVA NS and either received (i) no treatment, (ii) 2 Gy radiation for 5 days, (iii) ATG7i-SPNP for 3 doses, or (iv) both radiation and ATG7i-SPNP combination. Splenocytes were harvested on day 21, CFSE labeled, and cultured with SIINFEKL for 24 h before proliferation analysis. e CD8 T cell proliferation, as measured by CFSE dilution, of splenocytes harvested from the treated mIDH1-OVA NS-implanted mice. Bars (mean ± SEM) represent proliferating CD8⁺ T cell percentage within treatment groups: untreated (n = 5), ATG7i-SPNP alone (n = 5), IR alone (n = 5), ATG7i-SPNP + IR (n = 6), SIINFEKL alone (n = 4), ATG7i-SPNP + SIINFEKL (n = 4), IR + SIINFEKL (n = 5), and ATG7i-SPNP + IR + SIINFEKL (n = 6). All replicates are biological; one-way ANOVA followed by Tukey’s test; ns non-statistically significant. Source data are provided as a Source Data file.

Long-term survivors from the IR + ATG7i-SPNP treatment groups were rechallenged with mIDH1 mouse NS in the contralateral hemisphere (Supplementary Fig. 39). These animals remained tumor free without further treatment, whereas control mice implanted with glioma cells succumbed due to tumor burden (MS = 33 DPI; P ≤ 0.0001) (Supplementary Fig. 39). The animals were sacrificed at 60 DPI after tumor rechallenge, mice showed no evidence of microscopic intracranial tumor. Together, these results suggest the development of anti-glioma immunological memory in mIDH1 glioma rechallenged animals previously treated with IR + ATG7i-SPNP. None of these treatments produced hepatic alterations compared with the control group (Supplementary Figs. 40 and 41).

We then aimed to study the anti–mIDH1 glioma-specific immune response elicited by IR + ATG7i-SPNP therapy, using a T cell proliferation assay. Mice bearing mIDH1 tumors harboring a surrogate tumor antigen, OVA, were treated with saline, IR, ATG7i-SPNP, or IR + ATG7i-SPNP (Fig. 10d). Mice were euthanized 14 days after the last treatment dose (30 DPI), and the spleen was removed and processed for flow cytometry analysis. Splenocytes were fluorescently labeled with 5- and 6-CFSE and stimulated for 4 days with 100 nM SIINFEKL (Fig. 10d), where T cell proliferation was measured as the reduction of CFSE staining in the CD45+/CD3+/CD8+ population. After SIINFEKL treatment, we observed that T cell proliferation was significantly increased (P < 0.05) in mIDH1-glioma-bearing mice treated with IR + ATG7i-SPNP (Fig. 10e and Supplementary Fig. 42) when compared with the control group. This indicates that the combined treatment was able to induce a systemic anti–mIDH1-glioma-immune response associated with an enhancement in T cell proliferation.

In our investigation, we conducted a comprehensive assessment of tumor growth using histological analysis via H&E staining, along with immunohistochemical markers CD3 (reflecting infiltrating T cells) and CD68 (indicating an inflammatory response or the presence of tumor-associated macrophages) in long-term survivor mice bearing mIDH1 tumors subjected to treatment with ATG7i-SPNP+IR. The results revealed no tumor was present in the IR + ATG7i-SPNP-treated mice, concomitant with a significant (P < 0.0001) decrease in the presence of CD3- and CD68-positive cells following the combined treatment, as illustrated in Supplementary Figs. 43 and 44. To validate our findings in the long-term survivor mice, we next harvested tumors at 15 days post-tumor implantation, 5 days post-initial treatment (saline, IR, or ATG7i-SPNP+IR), and performed detailed immunohistochemical analysis to determine infiltration of immune cells into the TME. We found that mIDH1 tumor-bearing mice treated with ATG7i-SPNP + IR exhibited significantly increased infiltration of CD4+ cells and CD8+ cells (Supplementary Fig. 45); DCs and natural killer cells (NK1.1) (Supplementary Fig. 46); macrophages (CD68) and microglia (IBA1) (Supplementary Fig. 47), when compared to both the saline-treated group and the IR-treated group. Furthermore, mIDH1 tumor-bearing mice treated with IR alone exhibited significantly increased infiltration of CD4+ cells, CD8+ cells, and natural killer cells (Supplementary Figs. 45 and 46) when compared to saline-treated animals. However, the IR-treated animals did not have increased infiltration of DCs, macrophages, or microglia (Supplementary Figs. 46 and 47) when compared to saline-treated animals. Collectively, these results demonstrate that autophagy inhibition, when combined with IR, exerts a robust anti-tumor effect in the mIDH1 glioma models. This combined therapeutic approach not only enhances T cell infiltration but also significantly alters the inflammatory tumor microenvironment and the development of anti-glioma immunological memory. These findings highlight the multifaceted mechanisms by which autophagy inhibition augments the efficacy of radiotherapy, leading to durable, systemic anti-tumor immune responses in mIDH1-glioma-bearing mice.

Discussion

Metabolic reprogramming is a hallmark alteration observed in glioma tumors harboring the IDH1^R132H mutation⁵⁹; however, the biological consequences and its therapeutic implications remain to be fully elucidated. In this study, we show that in mIDH1 astrocytomas with inactivating mutations in ATRX and TP53, autophagy is upregulated in conjunction with functional disruption of mitochondrial activity. Analysis of RNA-seq data shows an upregulation of several autophagy-related genes and a downregulation of genes involved in mitochondrial metabolism. The downregulation of mitochondrial activity and oxidative phosphorylation is evident in mIDH1 glioma cells, as indicated by the signaling pathway enrichment analysis data. Consistently, mass spectrometry-based metabolomics analysis uncovered a marked alteration in the mIDH1 glioma metabolic profile and an alteration of the TCA cycle, with an accumulation of TCA cycle metabolites, suggesting impaired utilization of this pathway. In line with our findings, previous studies suggested that IDH1 mutation modulates the TCA cycle, mitochondrial metabolism, and oxidative stress pathways^{16,20,23,60,61}. The presence of 2HG reduces both oxidative phosphorylation and ATP levels in mIDH1 glioma cells⁶². These metabolic alterations were also observed in the genetically engineered mIDH1 glioma model, which displays an upregulation of DNA-damage responses (DDR)¹⁵. In addition, the enhancement of DDR has been related to metabolic alterations present in mIDH1 gliomas¹⁶. This supports our findings showing that the metabolic alterations in mIDH1 glioma cells are involved in the TCA cycle and mitochondrial dysfunctions.

Autophagy is activated in response to mitochondrial dysfunction, serving as a mechanism for cell protection and survival^63–65. Our ChIP-seq data confirms that epigenetic reprogramming of mIDH1 gliomas regulates the expression of several ATG genes as observed in both mouse and human mIDH1 glioma cells. In both the patient-derived glioma cells and patient-tumor samples, TFEB/TFE3 and other MiT/TFE family members were examined but did not display great transcriptional or epigenetic changes. This indicates that the upregulation of autophagy-related genes in mIDH1 gliomas is driven by direct epigenetic regulation through the mIDH1/2HG axis rather than by canonical TFEB/TFE3 pathway activation⁶⁶. Our OCR and ECAR analysis demonstrate reduced mitochondrial and glycolytic activity in mIDH1 glioma cells. These results functionally support our molecular data, indicating mitochondrial dysfunction along with an energetic state as encountered in slow proliferative cells. It has been suggested that autophagy may provide nutrients necessary to meet bioenergetic demands during transition from a slower proliferative state to an energy-dependent cell state⁶⁷. Our data corroborates with this hypothesis. We show that autophagy activity is functionally increased in mouse and human mIDH1 glioma cells and that the inhibition of IDH1^R132H reverts this phenotype in patient-derived mIDH1 glioma cells, suggesting that the enhancement of autophagy is 2HG-dependent. Our study describes epigenetic mechanisms involved in this phenotype, where histone methylation plays a key role in metabolic reprogramming and activation of autophagy in mIDH1 glioma cells.

Our results indicate that autophagy could serve as a cell protection and survival mechanism in mIDH1 gliomas. Autophagy has been shown to influence cancer therapeutic responses^68–73. We evaluated the impact of autophagy inhibition in response to radiation in vitro and in vivo and show that autophagy inhibition sensitizes mIDH1 glioma cells to IR treatment, decreasing cell viability of mouse glioma NS and human glioma cells. This notion is further supported by the fact that autophagy can induce DNA-damage response activation, thus inducing resistance to IR in tumor cells⁵⁸. We generated a genetically engineered autophagy-deficient mIDH1 mouse glioma model, which displayed decreased median survival when compared with the control mIDH1 glioma mouse model. These results align with the established role of autophagy in protecting cells from DNA damage³⁶ and maintaining chromosomal stability^74–76. Notably, autophagy blockade in our Sleeping-Beauty mIDH1 models, as well as treatment of mIDH1-glioma-bearing mice with autophagy inhibition (ATG7i-SPNP) alone, resulted in increased tumor aggressiveness (Figs. 9c and 10a). The plausible reason why autophagy inhibition alone reduces survival in mIDH1-glioma-bearing animals is that autophagy contributes to maintaining more stable genomes, which are associated with less mutagenic states and less aggressive tumors in vivo. Therefore, loss of a normal functioning autophagic pathway could lead to decreased DNA-damage repair, enhanced proliferation, increased invasion, and increased tumor aggressiveness (Fig. 9d).

Our results also show an upregulation of UVRAG, an ATG gene involved in chromosomal stability, which is in line with an increased DDR reported in mIDH1 gliomas^15,16. Conversely, this phenomenon could increase cell protection against DNA-damage assault and contribute to radio-resistance. Using our preclinical implantable mIDH1 glioma model¹⁵, we evaluated the therapeutic efficacy of autophagy inhibition using SPNPs loaded with siRNA targeting ATG7 (ATG7i-SPNP) in combination with IR. This combination increased median survival and elicited 40% of long-term tumor-free survivors in mIDH1-glioma-bearing mice, demonstrating that autophagy inhibition sensitizes mIDH1 tumors to radiotherapy, aligning with our in vitro findings. Additionally, we did not observe any overt toxicities as the treated animals were examined twice daily by expert veterinary staff at the University of Michigan’s Institutional Animal Care and Use Committee. We also evaluated the effects of the ATG7i-SPNP in combination with IR treatment in long-term survivors via histopathological analysis of brains and livers. We found that this treatment did not cause any alterations in the brain or liver architecture. Normal liver and kidney function in the treated animals was evidenced by the mouse serum biochemistry analysis (Supplementary Figs. 40 and 41).

In addition, our rechallenge experiment displayed a 100% survival rate of mIDH1 animals that were previously treated with IR in combination with ATG7i-SPNP and remained tumor-free for 90 days post-initial tumor implantation (Supplementary Fig. 39). These results suggest that the combined therapeutic approach induces anti-tumor immunological memory²². This was further confirmed as we observed increased surrogate tumor antigen (ovalbumin) specific T cell proliferation in mIDH1 glioma animals treated with IR and ATG7i-SPNP. Moreover, ovalbumin does not naturally occur in humans or other mammals; when it is expressed in tumor cells, it creates a “surrogate” tumor antigen. This enabled the detection of tumor-antigen-specific T cells in the spleen from the treated animals. Using SIINFEKL (cognate ovalbumin peptide), we demonstrate that the ATG7i-SPNP+IR treatment led to effector T cells that can recognize and proliferate when stimulated by the surrogate tumor antigen, “SIINFEKL”.

The rationale for targeting ATG7 in this study is that it functions as a key enzyme required for autophagosome expansion and maturation^77–79; thus, the inhibition of ATG7 results in a robust blockage of autophagy initiation and flux, allowing for a more complete blockade of autophagy from its early initiation steps. Our findings suggest that autophagy inhibition could be further developed as an upfront therapeutic strategy for biologically aggressive IDH-mutant tumors where intensification of local control is warranted, as well as in recurrent disease in which radiation and/or IDH inhibitor (AG-881) therapy has already been used, and autophagy dependence persists. Our data, along with independent findings from other laboratories using our genetically engineered animal models, which recapitulate critical features observed in glioma patients and molecular databases^{12,15,22,80–84}, provide evidence that our findings are relevant to a better understanding of the biological significance of metabolic reprogramming in mIDH1 gliomas and the role autophagy plays in mediating therapeutic responses. Our results apply to a particular mIDH1 molecular subtype, which harbors concomitant inactivating mutations in TP53 and ATRX, recently classified by the WHO as mIDH1 astrocytomas⁸⁵. This genotype was included in all in vitro, ex vivo, and in vivo experiments presented.

Lastly, patient-derived glioma models do have limitations, including poor engraftment of patient-derived glioma cells harboring the mIDH1 lesion, such as glioma cell culture, SF10602, and the inability of immunodeficient xenograft systems to recapitulate interactions encountered in the presence of a functional immune system⁴⁷. Despite this, autophagy inhibition significantly sensitized human mIDH1 glioma cells to radiotherapy in vitro, supporting the translational relevance of the findings.

In summary, our work demonstrates that autophagy inhibition radiosensitizes mIDH1 glioma in vivo, thereby leading to improved survival, long-term survival, and anti-tumor immunity (Fig. 10). Thus, our data support a combinatorial therapeutic approach for mIDH1 gliomas that warrants further investigation for both biologically aggressive tumors requiring intensified local control and recurrent disease resistant to radiation and/or IDH inhibition but still dependent on autophagy.

Methods

Study approval

All animal studies were conducted according to guidelines approved by the IACUC of the University of Michigan (protocols PRO00012850 and PRO00012853). All animals were housed in an AAALAC-accredited animal facility with a standard temperature of 22 °C ± 2 °C, 45-50% humidity, and were kept on a strict 12:12-h light:dark cycle. Mice were provided with PicoLab Laboratory Rodent Diet (Cat# 5L0D) and monitored daily for food and water consumption as well as bedding and living conditions. Studies did not discriminate by sex; both male and female mice were used. The strains of mice used in the study were C57BL/6 (the Jackson Laboratory, strain no. 000664) and CDKN2A-KO mice (Frederick National Library for Cancer Research strain no. 01XB1).

Study design

To study the impact of autophagy in the context of IDH1^R132H mutant gliomas, with TP53 and ATRX inactivating mutations, we previously generated a genetically engineered animal model by injecting SB plasmids encoding NRAS-G12V, shp53, and shATRX, and with or without IDH1^R132H into the lateral ventricle of one-day-old mouse pups¹⁵. Sample size and any data inclusion/exclusion were defined individually for each experiment. We also used an animal model previously generated by intracranial implantation of glioma NS (WT-IDH1 and mIDH1) derived from our genetically engineered animal model to test therapeutic responses¹⁵. Furthermore, we used human glioma cells derived from patients harboring IDH1^R132H, in the context of TP53 and ATRX inactivating mutations, to confirm the results obtained from our animal models. The number of replicates is reported in the figure legends. Our studies were not randomized. We performed blinding for quantitative IHC scoring. All RNA-seq and ChIP-seq data were deposited in public databases as indicated in the respective sections. Materials and Methods are detailed in the Supplementary Materials.

Genetically engineered mutant IDH1 glioma model

All animal studies were conducted according to guidelines approved by the IACUC at the University of Michigan. All animals were housed in an AAALAC-accredited animal facility, and they were monitored daily. Studies did not discriminate by sex, and both males and females were used. The strains of mice used in the study were C57BL/6 (The Jackson Laboratories, 000664).

The wild-type (WT)-IDH1 and mIDH1 glioma models used in this study were generated previously for our group using the SB Transposon System¹⁵. The plasmid used to generate those models are: (i) SB-transposase and LUC (pT2C-LucPGK-SB100X, henceforth referred to as SB/Luc), (ii) a constitutively active mutant of NRAS, NRAS-G12V (pT2CAG-NRASV12, henceforth referred to as NRAS), (iii) a short hairpin against p53 (pT2-shp53-GFP4, henceforth referred to as shp53), (iv) a short hairpin against ATRX (pT2-shATRX53-GFP4, henceforth referred to as shATRX), and (v) mutant IDH1^R132H (pKT-IDH1^R132H-IRES-Katushka, henceforth referred to as mIDH1). One-day-old P01 C57BL/6 mice were used in all experiments. The genotype of SB- generated mice involved these combinations: (i) NRAS, shP53, and shATRX (WT-IDH1) and (ii) NRAS, shp53, shATRX, and IDH1^R132H (mIDH1). Mice were injected according to a previously described protocol⁸⁶. Plasmids were mixed in mass ratios of 1:2:2:2 or 1:2:2:2:2 (20 μg plasmid in a total of 40 μL plasmid mixture) with in vivo-jetPEI (Polyplus Transfection, 201-50G) (2.8 μL per 40 μL plasmid mixture) and dextrose (5% total) and maintained at room temperature for at least 15 min prior to injection. The lateral ventricle (1.5 mm AP, 0.7 mm lateral, and 1.5 mm deep from lambda) of one-day-old mouse pups (P01) was injected with 0.75 μL plasmid mixture (0.5 μL/min) that included: (1) SB/Luc, (2) NRAS, (3) shp53, (4) shATRX, and (5) with or without IDH1^R132H. To monitor plasmid uptake in one-day-old pups, 30 μL of luciferin (30 mg/mL) was injected s.c. into each pup 24–48 h after plasmid injection. In vivo bioluminescence was measured on an IVIS Spectrum (PerkinElmer, 124262) imaging system. For the IVIS Spectrum, the following settings were used: automatic exposure, large binning, and aperture f = 1. For in vivo imaging of tumor formation and progression in 6–8-week-old adult mice, 100 μL of luciferin solution was injected i.p., and mice were then anesthetized with oxygen/isoflurane (1.5–2.5% isoflurane). To score luminescence, Living Image Software Version 4.3.1 (Caliper Life Sciences) was used. A region of interest (ROI) was defined as a circle over the head, and luminescence intensity was measured using the calibrated unit’s photons/s/cm²/sr. Multiple images were taken over a 25 min period following injection, and maximal intensity was reported.

For survival studies, animals were monitored daily for signs of morbidity, including ataxia, impaired mobility, hunched posture, seizures, and scruffy fur. Animals displaying symptoms of morbidity were intracardially perfused using Tyrode’s solution, followed by fixation with 4% paraformaldehyde (PFA) in PBS.

Alternative genetically engineered mouse glioma models

In CPA/CPAI models, the genetic lesions were incorporated into one-day-old CDKN2A homozygous knocked-down (CDKN2A^-/-) mice¹². CDKN2A^-/- mice were obtained from the Frederick National Library for Cancer Research as frozen embryos. Embryos were implanted into receptive adult B6 females by the Transgenic Animal Core of the University of Michigan, and the obtained transgenic mice, which were CDKN2A heterozygous, were mated with B6 mice. The CDKN2A-KO heterozygous mice of the second generation were mated, and the progeny obtained were genotyped to confirm the homozygous CDKN2A deletion. From this point, the CDKN2A^-/- colony was maintained by mating CDKN2A^-/- mice and monitored regularly by CDKN2A genotyping. The genotype of the genetically engineered mice involved these combinations: (i) CDKN2A deletion, TP53 and ATRX short hairpin knockdown, with or without IDH1-R132H. The CPA/CPAI models were confirmed via western blotting analysis.

In the RPA/RPAI models, glioma was driven by platelet-derived growth factor receptor alpha mutation (PDGFRA^D842V) to promote constitutive activation of the MAPK pathway¹⁸. These mouse glioma cells include ATRX and TP53 short hairpin knockdown, with or without endogenous expression of IDH1-R132H, to simulate the molecular genetic lesions that define the molecular features of astrocytoma.

Primary glioma neurosphere cultures (NS)

Mouse glioma neurosphere (NS) cultures were generated as described above^15,87. Briefly, brain tumors in mice were harvested at the time of euthanasia by intracardial perfusion with Tyrode’s solution only. The tumor mass was dissociated using non-enzymatic cell dissociation buffer, filtered through a 70 µm strainer and maintained in neural stem cell medium DMEM/F12 with L-Glutamine (Gibco, 11320-033), B-27 supplement without Vitamin A (1×) (Gibco, 12587-010), N-2 supplement (1×) (Gibco, 17502-048), Penicillin-Streptomycin (100X) (10,000 IU Penicillin) (10,000 µg/mL Streptomycin) (Corning, Cellgro, 30-001-CI), and Normocin (1×) (Invivogen, ant-nr-1) at 37 °C, 5% CO₂. Human-EGF (Peprotech, 100-15) and human-FGF (Peprotech, 100-18) were added twice weekly at 1 µL (20 ng/µL each stock) per 1 mL medium for a final concentration of 20 ng/mL. For reduced growth conditions, human-EGF and human-FGF were added twice weekly at 1 µL (20 ng/µL) per 3 mL medium for a final concentration of 6.6 ng/mL. These genetically engineered mouse cell models were confirmed via western blotting analysis.

Human glioma cell culture

Patient-derived mIDH1 glioma cells, SF10602, were characterized and generously provided by Dr. Joseph Costello from the University of California, San Francisco⁴⁷. Patient-derived mIDH1 glioma cells, MGG119, were characterized and generously provided by Dr. Daniel Cahill from Harvard University Medical School⁸⁸. Patient-derived mIDH1 glioma cells, LC1035, were characterized by the University of Michigan Neurosurgery Department (IRB approval code HUM00175135 and HUM0024610)⁸⁹. The work reported in this manuscript does not constitute Human Subjects Research.

Human SF10602⁴⁷ cells were grown in NeuroCult NS-A Basal Medium supplemented with 100 units/mL antibiotic-antimycotic, 10 mL B-27 without vitamin A, 5 mL N-2, 100 μg/mL normocin, 20 ng/mL FGF, 20 ng/mL EGF, and 20 ng/mL PDGF-AA. Human MGG119⁸⁸ and LC1035 GCCs were grown in Neurobasal medium supplemented with 100 units/mL antibiotic-antimycotic, 10 mL B-27, N-2, 100 μg/mL normocin, 20 ng/mL FGF, 20 ng/mL EGF, and 20 ng/mL of PDGF-AA. Glioma cells were dissociated using StemPro Accutase solution and passaged weekly. Human glioma cells were shared through the following collaborations: Dr. Daniel Cahill's laboratory, Harvard Medical School (MGG119), and Dr. Joseph Costello's laboratory, UCSF (SF10602). LC1035 was generated in our laboratory through our collaboration with the University of Michigan’s Department of Neurosurgery.

RNA-seq and bioinformatics analysis

RNA-seq was performed in collaboration with the University of Michigan sequencing core. All the RNA-seq datasets used to identify differential enrichments related to metabolism and autophagy were previously generated in our lab¹⁵ with public access. Briefly, RNA was isolated from tumor NS, and 100 ng samples of purified RNA were sent for RNA-seq analysis. After passing all quality controls, Illumina HiSeq 2500/4000 fastq files were processed using the Tuxedo Suite for alignment and differential expression analysis. Genes and transcripts were identified as differentially expressed based on three criteria: test status = “OK”, FDR ≤ 0.05, and fold-change ≥ ±1.5. Gene Ontology (GO) enrichment analysis, which was performed using iPathwayGuide (http://www.advaitabio.com/ipathwayguide) separately for upregulated and downregulated genes. GO Biological Processes, selected for relevance to phenotype, were plotted in a horizontal bar graph using R. Gene set analysis (GSA) was performed using the R package GSA (http://statweb.stanford.edu/~tibs/GSA/), which implements a version of Gene Set Enrichment Analysis (GSEA)⁹⁰ with improved power statistics⁹¹. GSA takes an expression matrix and a gene set collection as input. The Gene Set Knowledgebase (GSKB) data package (version 1.3.0) was downloaded from Bioconductor. GSKB is specific to mice and includes annotation from over 40 sources. GSA outputs significant gene sets (FDR ≤ 0.05), both negatively and positively correlated to phenotype, as well as normalized enrichment scores, and the lists of genes contributing to the significance of each gene set (FDR ≤ 0.05). For each list of genes found in a GSKB functional term we tested for enrichment in a small set of Gene Ontology terms, selected for relevance to the phenotype, using the hypergeometric test in R. After RNA-seq analysis, enrichment maps were generated using the Cytoscape platform with enrichment map; the designation of node color was amended to complement our RNA-seq expression color scheme of positively (red) and negatively (green) regulated pathways. The stringency of FDR was reduced to truly encapsulate the underlying biological ramifications of ATRX and IDH1 mutations in our transposon-mediated tumor formation model. Heatmaps were plotted with the Heatmap2 package in R. RNA-seq datasets have been deposited in NCBI’s Gene Expression Omnibus with identifiers GSE94902, GSE94974, and GSE94975.

Bru-seq and bioinformatics analysis

The Bru-seq dataset used in this work to compare transcriptional levels of genes related to metabolism and autophagy was previously generated¹⁵. The protocol used was described in Paulsen et al.⁹². Nascent RNA labeling was performed for 30 min at 37 °C with 2 mM 5-bromouridine in conditioned medium. The cells were then incubated for 6 h at 37 °C. At the completion of labeling, the cells were lysed in TRIzol, and the Bru-containing RNA was isolated using anti-BrdU antibodies conjugated to magnetic beads. The isolated RNA was converted into cDNA libraries and prepared for sequencing using the Illumina TruSeq RNA Library Preparation Kit v2 followed by deep sequencing to around 50 million single-end 50-nucleotide reads. A bioinformatics and data analysis pipeline was implemented using the q pipeline manager (http://sourceforge.net/projects/qppln-mngr/). The bioinformatics programs used, read mapping, genome annotation, and expression scoring, were previously described in section 2.4 of ref. ⁹². RPKM (reads per kilobase per million mapped reads) values were calculated for individual genes that were at least 300 bp long. For genes with lengths of 30 kb and less, RPKM values were calculated using read counts from the entire gene. For genes longer than 30 kb, an RPKM value was calculated using read counts from the first 30 kb downstream of the TSS. The R package DESeq35 was used to test differential expression of genes where the mean RPKM between samples was greater than 0.5. Significant changes in transcription initiation were defined as follows: adjusted p-values < 0.05; fold-change <1.5.

ChIP-sequencing and bioinformatics analysis

Both mouse and human ChIP-seq datasets were previously published^15,18. For the ChIP-seq data analysis for this study, the ChIP-seq raw data quality was assessed using FastQC 0.11.3, and the adapters were trimmed using TrimGalore 0.4.4. ChIP-seq and input reads were aligned to the mouse reference genome (mm10) using bowtie2⁹³ with default options, and the uniquely mapped reads were extracted for peak calling. The post-alignment quality control steps were performed by Deeptools 2 3.5.1⁹⁴. PePr⁹⁵ was used to identify the differential peaks between WT-IDH1 and mIDH1 groups using FDR < 0.05 and fold-change >3 as the cutoffs. The identified peaks were annotated to genomic features 1–5 kb upstream of the promoter, promoter, intron, exon, UTR, CDS, and intergenic regions (using the Bioconductor annotatr package⁹⁶). As a comparison, the same number of peaks were randomly generated across the same chromosome and annotated the same way by annotatr. Gene set enrichment testing was performed on the identified differential peaks by the Bioconductor package ChIPenrich⁹⁷ to detect enriched Gene Ontology Biological Process Terms.

Cleavage Under Targets & Release Using Nuclease (CUT&RUN)

CUT&RUN was performed using the CUTANA ChIC/CUT&RUN kit (Epicypher #14-1048) following the manufacturer’s protocol. Briefly, 1 × 10⁶ NS cells were coupled with Concanavalin A beads, permeabilized with 0.01% Digitonin, and incubated overnight at 4 °C with 0.5 μg target antibody in antibody buffer (H3K4me3 Antibody - SNAP-Certified™ for CUT&RUN, Epicypher #13-0041, H3K27ac Antibody, SNAP-Certified™ for CUT&RUN and CUT&Tag, Epicypher #13-0059). The following day, cells were first incubated for 10 min with micrococcal nuclease fused to proteins A and G (pAG-MNase), which was then activated by CaCl₂ addition. After 2 h incubation at 4 °C, the reaction was stopped with stop buffer, and E. coli DNA was added as spike-in DNA. DNA was then isolated using SPRI beads and quantified with Qubit. Libraries were prepared using CUTANA™ CUT&RUN Library Prep Kit, following the manufacturer’s recommendations. Fragment size was detected using a TapeStation system, and multiplexed libraries were sequenced on an AVITI24 sequencer (Element Biosciences) at 150 bp paired-end reads. Data analysis was performed the same way as described in the “ChIP-sequencing and bioinformatics analysis” section.

CUT & RUN qPCR

CUT & RUN was performed as previously described. qPCR was performed on a QuantStudio 3 Real-Time PCR System (Applied Biosystems) using SYBR Green qPCR Master Mix Fast SYBR Green Master Mix (Applied Biosystems). Primers were designed based on the CUT & RUN sequencing. The primers sequences we used are: Atg9b_1_F: CTTTGTTGTTGCCTCACCCG, Atg9b_1_R: ACTCATGGCCTTGGGAATGG, Atg9b_2_F: GAGATCCTCCTTCCCCCAGT, Atg9b_2_R: TCAGTGCCTCTGCTCCCTAT, Atg9b_3_F: CGGGTACATGAAGCAAGGGT, Atg9b_3_R: GGGCTTCTCAGACTCCACAC.

Human single-cell RNA-seq

Human glioma scRNA-seq was performed and analyzed as described above¹². The data has been deposited in NCBI’s Gene Expression Omnibus with identifier GSE152273. Briefly, primary mIDH1 human glioma cells, SF10602, were cultured for 7 days in the presence or absence of specific IDH1^R132H inhibitor AGI-5198. 3′ single-cell libraries were generated using the 10X Genomics Chromium Controller following the manufacturer’s protocol for 3′ V3.1 chemistry with NextGEM Chip G reagents (10X Genomics). Final library quality was assessed using the TapeStation 4200 (Agilent), and libraries were quantified by Kapa qPCR (Roche). Pooled libraries were subjected to 150 bp paired-end sequencing according to the manufacturer’s protocol (Illumina NovaSeq 6000). Raw sequencing data files were converted to fastq files and aligned to the human reference genome hg38 using the Cell Ranger Pipeline 7.0.0 (10X Genomics). The data were clustered, and gene expression was analyzed using the Seurat R package. Pathway enrichments were performed by the Gene Set Variation Analysis (GSVA) package⁹⁸. Two biological replicates for each condition (wt-IDH1 and mIDH1) were analyzed. An average sequencing depth of 20,000 reads/cell was recovered for each sample we processed.

Human glioma cells stably transfected to express IDH1^R132H

SJGBM2 pediatric glioma cells are ATRX mutant, and they were cultured in IMDM medium with L-glutamine (0.3 mg/mL) (Gibco, 12440061), 20% FBS (Peak Serum, PS-FB3), and antibiotic-antimycotic (1×) (Gibco, 15240-062) at 37 °C, 5% CO₂. SJGBM2 cells were seeded in a 6-well plate (1.5 × 10⁵ each per well), and after 24 h they were transfected with the p-CMV-IDH1^R132H-Entry plasmid using the jetPRIME transfection system (VWR, 89129-922). One day after transfection, the medium was replaced with selection medium containing G418 (Gibco, 10131027) at a concentration of 800 µg/mL. On day 15 of selection, individual cell colonies were taken from the well using autoclaved filter paper squares (2 × 2 mm approx.) previously embedded in HyQTase Cell Detachment Solution (TroyBiologicals, SV3003001) and placed within wells of 24-well plates with the appropriate cell culture medium. After 24 h, the medium was replaced with selection medium. Each well corresponded to an isolated colony that was expanded for in vitro experiments. IDH1^R132H protein expression was confirmed by WB assay¹⁵. For reduced growth conditions, FBS was reduced to a final concentration of 5%. SJGBM2 cells were confirmed by Dr. Patrick Reynolds⁹⁹, and SJGBM2-mIDH1 cells were confirmed in our laboratory via western blotting analysis.

Metabolomics profiling

Mutant IDH1 and WT NS (3 × 10⁶ cells) were seeded in four T75 flasks (n = 4 per cell type) in complete media as described above (Supplementary Fig. 2). A parallel plate for protein estimation and sample normalization was also set up with the same number of cells. After overnight incubation, the culture media was aspirated off and replaced with fresh media. The cells were then cultured for a further 24 h. For intracellular metabolites, the media and cells were centrifuged to separate the cells and media. The media was aspirated off and the samples washed once with 1 mL cold PBS before isolating the cells again. The cells were then incubated with 1 mL ice-cold 80% methanol on dry ice for 10 min. Thereafter, cell lysates were collected from each well and transferred into separate 1.5 mL Eppendorf tubes. At this point, the samples could be placed at −80 °C until ready for use. Once removed from −80 °C, they were thawed on ice for 5 min before being centrifuged at 12,000 × g. The volume of supernatant to collect for each experimental condition was then determined based on the protein concentration of the parallel plate. The collected supernatants were dried using a SpeedVac Concentrator, reconstituted with 50% v/v methanol in water, and analyzed by mass spectrometry¹⁰⁰.

Targeted metabolomics was performed on an Agilent 1290 Infinity II Binary Bio LC coupled with an Agilent 6495 d QqQ mass spectrometer. Two methods were utilized in this study on the same instrumentation. For method one, the column used was an Agilent InfinityLab Poroshell 120 HILIC-Z, 2.1 × 150 mm, 2.7 μM (p/n 683775-924). Method parameters are as follows: Solvent A is water + 20 mM ammonium acetate (pH 9.3) + 5 μM medronic acid. Solvent B is acetonitrile. Wash solvent is 1:1:1 water, acetonitrile, and methanol. The solvent gradient is 10% A – 90% B until 1 min, 22% A – 78% B until 8 min, 40% A – 60% B until 12 min, 90% A – 10% B until 15 min, hold until 18 min, and 10% A – 90% B until 19 min, hold until 23 min. Flow rate is 0.4 mL/min. Column temperature is set to 15 °C. Source parameters are optimized for the Agilent dMRM (dynamic multiple reaction monitoring) library. The dMRM library was acquired from Agilent for their HILIC-Z platform and further optimized to screen for 435 targets in both positive and negative ion modes.

Method 2 utilized an Agilent ZORBAX RRHD Eclipse Plus C18, 100 × 2.1 mm, 1.8 μm (p/n 959758-902) column. Solvent A was water + 0.1% formic acid. Solvent B is methanol + 0.1% formic acid. Wash solvent is 1:1 water and methanol. The solvent gradient is100% A until 3 min, 5% A – 95% B until 5 min, hold until 7 min, 100% A until 8 min, hold until 10 min. Flow rate is 0.2 mL/min. Column temperature is set to 45 °C. Source parameters reflect method 1. The dMRM acquisition list includes 12 targets optimized by analytical standards in both positive and negative ion modes.

Data from both methods was preprocessed using Agilent MassHunter Workstation Quantitative Analysis for QQQ Version 12.1. This process includes manual inspection of chromatograms to verify accurate area-under-the-curve (AUC) readings, which are then reported as raw abundance of metabolites. Quality control pooled samples are inspected to evaluate instrument stability and signal-to-noise ratios for metabolites within the experiment dataset. While transitions for 435 metabolites are included in the screen, only those with verified peak shapes are kept during preprocessing.

Finalized raw data was normalized by median-centering, and then further fold-change analysis was performed relative to control groups to achieve relative abundance. Student’s t-tests were performed with an alpha level of 0.1. Heatmaps were generated in Morpheus (Morpheus, https://software.broadinstitute.org/morpheus).

Measurement of metabolites in genetically engineered mouse mIDH1 glioma models

Mouse mIDH1 glioma NS (RPA/RPAI and NPA/NPAI) were prepared according to the protocols outlined by the following metabolomics kits: Citric Acid Colorimetric Assay Kit (Elabscience, E-BC-K351-M), Succinic Acid Colorimetric Assay Kit (E-BC-K902-M), Alpha-Ketoglutarate Assay Kit (Cell Biolabs Inc., MET-5131), Malic Acid Colorimetric Assay Kit (Elabscience, E-BS-K905-M), and Glutamine Synthetase Activity Colorimetric Assay Kit (Elabscience, E-BC-K664-M).

Western blot

Mouse NS and human glioma cells (1.0 × 10⁶ cells) were harvested for each genotype, and total protein extracts were prepared in a RIPA lysis and extraction buffer (Thermo Fisher Scientific, Pierce, 89900) with 1× Halt protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific, 78442). 20 μg of protein extract (determined by bicinchoninic acid assay (BCA), Pierce, 23227) was separated by 4–12% SDS-PAGE (Thermo Fisher Scientific, NuPAGE, NP0322BOX) and transferred to nitrocellulose membranes (Bio-Rad, 1620112). The membrane was probed with 1:1000 of a rabbit anti-p62 Ab (Cell Signaling Technology, 5114, Lot: 6), 1:2000 of a rabbit anti-ATG9b Ab (Novus Biologicals, NBP1-77169, Lot #5797-1702), 1:1000 of a rabbit anti-ATG7 Ab (Cell Signaling Technology, 8558, Lot: 5), 1:1000 of a rabbit anti-pULK1 (S555) Ab (Cell Signaling Technology, 5869, Lot: 3); 1:1000 of a rabbit anti-pULK1 (S757) Ab (Cell Signaling Technology, 6888, Lot: 3); 1:1000 of a rabbit anti-ULK1 (Cell Signaling Technology, 8054, Lot: 9); 1:2000 of a rabbit anti-MST4 Ab (Abcam, ab52491, Lot #1007309-3); 1:1000 of a rabbit anti-ATG4b Ab (Cell Signaling Technology, 5299, Lot: 2); 1:1000 of a rabbit anti-phosphoATG4b Ab (Cell Signaling Technology, 19386, Lot: 1); 1:1000 of an anti-UVRAG Ab (Abcepta, AP1850d); 1:1500 of a rabbit anti-LC3I/II Ab (Novus Biologicals NB100-2220, Lot #D155067-7); 1:2000 of an anti-β-actin Ab (Cell Signaling Technology, 3700, Lot: 23 and 16), then followed by secondary [Dako, Agilent Technologies, goat anti-rabbit 1:4000 (P0448), Lot #41372205; goat anti-mouse 1:4000 (P0447), Lot: 41705282]. Enhanced chemiluminescence reagents were used to detect the signals following the manufacturer’s instructions (SuperSignal West Femto, Thermo Fisher Scientific, 34095). Blots were imaged using a ChemiDoc (Bio-Rad ChemiDoc™ MP System). WB quantification was performed using ImageJ, and the reported data are from three replicates. For the cells treated with α-ketoglutarate (α-KG), mIDH1 mouse NS (5.0 × 10⁵ cells) were treated with 2.5 mM α-KG (Cayman Chemicals, 876150-14-0) for 5 h prior to protein extraction. For o-2HG treatment, cells were incubated in 1.5 mM (2 R)-Octyl-α-hydroxyglutarate (Cayman Chemicals, 1391194-67-4) for 5 h prior to protein extraction.

Western blot assay of histones

To assess specific histone markers' post-translational modifications, histone extracts were obtained using the Histone Purification Mini Kit (Active Motif, 40026), and WB were performed on 1.5 µg of histone extract. Histone markers’ specific antibodies were as follows: 1:1000 of H3K4me3 antibody (Hologic Diagenode, C15410003, Lot #A8034D); 1:1000 of H3K36me3 antibody (Hologic Diagenode, C15410192, Lot #A1845P); 1:2000 of H3K27me3 antibody (Millipore, B07- 449, Lot #4258307) for human glioma cell protein extracts; 1:1000 of H3K27me3 antibody (Hologic Diagenode, C15410195, Lot #A3236P); and 1:2000 of total histone H3 antibody (Cell Signaling Technologies, 4499, Lot: 20).

Metabolic flux assay

To assess metabolic activity, oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) were performed using the XF-96 Extracellular Flux Analyzer (Agilent). Cells were washed and resuspended in nonbuffered DMEM, adjusted to pH~7.4. WT-IDH1 and mIDH1 cells (1 × 10⁵ cells/well) were seeded on laminin-coated plates and allowed to equilibrate for 30 min in a non-CO₂, 37 °C incubator. OCR and ECAR were measured under basal conditions and in response to the following mitochondrial inhibitors: oligomycin (1 μM), FCCP (1 μM), rotenone (100 nM), and antimycin A (1 μM). After the assay, OCR and ECAR measurements were normalized to cell number using CyQuant NF analysis (Invitrogen). For o-2HG treatment, cells were incubated in 2.5 mM (2 R)-Octyl-α-hydroxyglutarate (Cayman Chemicals, 1391194-67-4) for 5 h before assessing metabolic activity.

Autophagy flux assay using mCherry-GFP-LC3 construct

Lentiviral particles expressing mCherry-GFP-LC3 were generated by the University of Michigan Vector Core using the FUW mCherry-GFP-LC3 plasmid⁵⁷, which was a gift from Anne Brunet. WT-IDH1 and mIDH1 SJGBM cells were cultured in a T-12.5 flask and infected with Lenti-mCherry-GFP-LC3 particles at a concentration of 1X of lentivirus. After 36 h, 5 × 10⁴ cells were plated in a confocal microscopy chamber cover glass system, in the presence or absence of 10 μM of chloroquine, and cultured for 12 h before analysis of GFP and mCherry expression by confocal microscopy.

Confocal microscopy and image analysis

Cells were placed into the incubator chamber of the microscope at 37 °C with a 5% CO₂ atmosphere. Confocal microscopy images were acquired using a single-photon laser scanning inverted confocal microscope LSM 880 AxioObserver (Carl Zeiss, Jena, Germany). Two laser lines were used for simultaneous excitation with a Plan-apochromat 63×/1.4 numerical aperture (NA) oil DIC M27 objective. Cells transfected with mCherry-GFP-LC3 were excited at 488 nm and 561 nm, and cells stained with MitoTracker™ green dye at 405 nm and 488 nm wavelength, respectively. Line-sequential scan mode and 1.4 scan zoom were used. The system was driven by ZEN Black software. Multichannel immunofluorescence images were processed and analyzed using Fiji (NIH). Autophagosome (green) and autolysosome (red) were manually identified and quantified using the plug-in Cell Counter.

In vitro experiments with radiation and autophagy inhibitors

WT-IDH1 and mIDH1 mouse NS and human glioma cells (i.e., NPA/NPAI, SJGBM2/SJGBM2 mIDH1, and SF10602/SF10602 with AGI-5198) were plated at a density of 1 × 10³ cells/well in a 96-well plate 24 h before treatment. WT-IDH1 and mIDH1 mouse NS (NPA/NPAI) were then incubated with either ATG7 siRNAs (100 nM; Origene, SR427399), ATG4b siRNAs (100 nM; Origene, SR413163), ULK101 (1 µM; SelleckChem, S8793), or NSC185058 (5 µM; Cayman Chemical, 23957) in combination with 3 Gy of IR, 2 h post-inhibitor treatment. Human glioma cells (SJGBM2/SJGBM2 mIDH1, and SF10602/SF10602 with AGI-5198) were then incubated with either ATG7 siRNAs (150 nM; Origene, SR323157), ATG4b siRNAs (150 nM; Origene, SR323518), ULK101 (3 µM; SelleckChem, S8793), or NSC185058 (15 µM; Cayman Chemical, 23957) in combination with 5 Gy (SJGBM2/SJGBM2 mIDH1) or 20 Gy (SF10602/SF10602 with AGI-5198) of IR, 2 h post-inhibitor treatment. Cell viability was evaluated 72 h post-IR treatment using CellTiter-Glo 2.0 (Promega, G9242) luminescence cell viability assay, following the manufacturer’s protocol. The resulting luminescence was read with the Enspire Multimodal Plate Reader (PerkinElmer, 2300-0000). Data were represented graphically using GraphPad Prism software (version 8), and statistical significances were determined using Student’s t-test.

In vitro experiments with radiation and α-ketoglutarate

Mutant IDH1 mouse NS and human mIDH1 glioma cells (i.e., CPAI, RPAI, NPAI, and human mIDH1 SF10602) were plated at a density of 1 × 10³ cells/well in a 96-well plate 24 h before incubation with Octyl-α-ketoglutarate (α-KG; Cayman Chemicals, 876150-10-0). Mouse mIDH1 NS were incubated with 0.3 mM α-KG in combination with 3 Gy of IR, 2 h post-inhibitor treatment. Human glioma cells (human mIDH1 SF10602 cells) were then incubated with 1.5 mM α-KG in combination with 30 Gy of IR, 2 h post-inhibitor treatment. Cell viability was evaluated 72 h post-IR treatment using CellTiter-Glo 2.0 (Promega, G9242) luminescence cell viability assay, following the manufacturer’s protocol. The resulting luminescence was read with the Enspire Multimodal Plate Reader (PerkinElmer, 2300-0000). Data were represented graphically using GraphPad Prism software (version 8), and statistical significances were determined using Student’s t-test.

In vitro experiments with radiation and chloroquine and hydroxychloroquine

Mutant IDH1 mouse NS and human mIDH1 glioma cells (i.e., NPAI, RPAI, and human mIDH1 SF10602) were plated at a density of 1 × 10³ cells/well in a 96-well plate 24 h before incubation with either Chloroquine (CQ; Invivogen, tlrl-chq-4) or Hydroxychloroquine (HCQ; Millipore Sigma, H0915). Mouse mIDH1 NS were treated with 10 µM of CQ in combination with 3 Gy of IR, 2 h post-inhibitor treatment. Human glioma cells (human mIDH1 SF10602 cells) were treated with 10 µM of CQ in combination with 30 Gy of IR, 2 h post-inhibitor treatment. Mouse mIDH1 NS cells were treated with 50 µM of HCQ in combination with 3 Gy of IR, 2 h post-inhibitor treatment. Human glioma cells (human mIDH1 SF10602 cells) were treated with 50 µM of HCQ in combination with 30 Gy of IR, 2 h post-inhibitor treatment. Cell viability was evaluated 72 h post-IR treatment using CellTiter-Glo 2.0 (Promega, G9242) luminescence cell viability assay, following the manufacturer’s protocol. The resulting luminescence was read with the Enspire Multimodal Plate Reader (PerkinElmer, 2300-0000). Data were represented graphically using GraphPad Prism software (version 8), and statistical significances were determined using two-tailed Student’s t-test.

In vitro dose-response and evaluation of radiosensitivity using mitochondrial electron transport chain inhibitors

To assess the susceptibility of both mutant and WT-IDH1 human (SJGBM2 WT-IDH1/ SJGBM2 mIDH1 and SF10602/SF10602 with AG-881) and mouse NS (NPA/NPAI, CPA/CPAI, and RPA/RPAI) to mitochondrial electron transport chain inhibitors (mETCi), two Complex I inhibitors were used: Rotenone (SelleckChem, S2348) and Metformin (SelleckChem, S1950) under normal and growth factor-starved conditions. Both cell lines were plated at a density of 1000 cells per well in a 96-well plate (Fisher, 12-566-00) 24 h prior to treatment, where wells per inhibitor dose (1 μM, 2.5 μM, 5 μM, 10 μM, 20 μM, and 30 μM) were evaluated for each cell type. Cells were then incubated for 3 days under normal (EGF: 1:1000, FGF: 1:1000, and PDGFα: 1:1000 dilutions) as well as growth factor-starved conditions (EGF: 1:3000, FGF: 1:3000, and PDGFα: 1:3000 dilutions) and viability was assessed using the CellTiter-Glo 2.0 assay (Promega, G9242) following manufacturer’s protocol. To assess radiosensitivity of each cell line, cells were incubated with either free Rotenone or Metformin, or in combination with radiation at their respective IC50 doses for 72 h in triplicate wells per condition. Cells were pre-treated with both inhibitors 2 h prior to IR with 3 Gy for mouse, 5 Gy for SJGBM2, and 20 Gy for SF10602 of radiation, respectively. Resulting luminescence was read with the Enspire Multimodal Plate Reader (PerkinElmer, 2300-13 0000).

Release of cytokines in response to autophagy inhibition and radiation

To assess the release of cytokines (DAMPs and type-I IFNs) into the conditioned media, mIDH1 mouse NS and human glioma cells were seeded at a density of 1 × 10⁶ cells/well in 6-well plates and allowed to settle overnight before treatment. The following day, mIDH1 mouse NS were treated with either ATG7 (100 nM; Origene, SR427399) or ATG4b (100 nM; Origene, SR413163) for 2 h prior to exposure to 3 Gy of IR, and the human glioma cells were treated with either ATG7 (150 nM; Origene, SR323157) or ATG4b (150 nM; Origene, SR323518) for 2 h prior to exposure to 5 Gy of ionizing radiation. After 72 h, the levels of various cytokines in the cultures’ supernatants were measured using ELISA, either following the manufacturer’s protocol (Novus Biologicals) or at the Cancer Center Immunology Core, University of Michigan.

Implantable syngeneic murine glioma models

Both male and female C57BL/6 mice, aged 6–8 weeks old, were used for implantation models. Intracranial tumors were established by stereotactic injection of 5 × 10⁴ WT-IDH1 or mIDH1 mouse tumor NS into the right striatum using a 22-gauge Hamilton syringe (1 μL/min) with the following coordinates: +1.00 mm anterior, 1.8 mm lateral, and 3.5 mm deep. The presence of tumors was verified 5 days post-implantation (DPI) by bioluminescence imaging.

For survival studies, animals were monitored daily for signs of morbidity, including ataxia, impaired mobility, hunched posture, seizures, and scruffy fur. Animals displaying symptoms of morbidity were intracardially perfused using Tyrode’s solution, followed by fixation with 4% paraformaldehyde (PFA) in PBS.

Generation of autophagy-deficient mouse model using SB transposon system

Using the SB transposon system⁸⁶, we generated a glioma model harboring IDH1^R132H, shATRX, and shP53 as described above¹⁵, plus a shRNA for ATG7 (shATG7) to disrupt the autophagy pathway in the tumor. The shATG7 candidates were obtained from the Codex database (http://cancan.cshl.edu/cgi-bin/Codex/Codex.cgi) and cloned into the pT2 plasmid, replacing shATRX sequences from our previous PT2-shATRX construction⁸⁷, generating the pT2-ATG7-BFP. A WB assay was performed to select the best candidate to use for the generation of the animal model. The selected shATG7-A (HP_603911) (CTCGAGTGCTGTTGACAGTGAGCGACCAGAAGAAGTTGAACGAGTATAGTGAAGCCACAGATGTATACTCGTTCAACTTCTTCTGGGTGCCTACTGCCTCGGAGAATTC) was finally used in the combination cocktail of plasmids. Sleeping-beauty-derived glioma cells NPA and NPAI were transfected using jetPEI (VWR, catalog 89129-960) with the plasmid pT2-shATG7-EBFP and sleeping-beauty transposase plasmid (Supplementary Fig. 30a, b). After transfection, cells were allowed to grow for 72 h then they were subjected to FACS for isolation of BFP cells. These cells were sorted 3 times each before confirming ATG7 knockdown via Western Blot assay.

Generation of iRGD synthetic protein nanoparticle (SPNP) with siRNA against ATG7

SPNP formulation

The preparation of the SPNP formulations followed previously reported methods but with modifications⁴⁶. In the basic formulation, two solutions were prepared separately, then combined. HSA (7.5% w/v) was solubilized in a solvent system comprised of ultrapure deionized water and ethylene glycol (80:20 v/v). The peptide, iRGD, was then added along with BSA Alexa Fluor 647 conjugate (0.25% w/w relative to the albumin) to produce fluorescently labeled SPNPs. Then, the siRNA, which was resuspended according to the manufacturer’s protocol, was complexed for 30 min at room temperature under rotation with 60 kDa branched polyethyleneimine (5% w/v) before also being added to the serum albumin solution. Next, a bi-functional macromer, O,O′-Bis[2-(N-Succinimidyl-succinylamino)ethyl]polyethylene glycol, was added at 10% w/w relative to the HSA solution. Once mixed, the two solutions were mixed to form the final formulation. For empty SPNP groups, all the components that were added in the iRGD ATG7i-SPNPs were included except for the ATG7 siRNA.

SPNP fabrication

SPNPs were fabricated via electrohydrodynamic jetting⁴⁶. The parameters used were the same as above. Briefly, the final formulation was loaded into 1 mL syringes equipped with a 1.5” 25-gauge stainless steel blunt needle. It was pumped at a rate of 0.2 mL/h to form droplets at the base of the needle. A voltage source was connected to the needle and grounded at the collection plate, located 6 inches from the base of the pump. The voltage was typically adjusted to a range between 8 kV and 15 kV to achieve a stable Taylor cone whereby rapid evaporation of the solvent occurred, creating solid nanoparticles on the collection plate. The collection plate was replaced with a clean plate every 30 min until the solution within the syringe emptied. The collection plates were enclosed and incubated for 7 days at 37 °C to form stable crosslinks.

SPNP collection and processing

After the 1wk period of incubation, the SPNPs on collection plates were removed. About 3–4 mL of 0.01% Tween 20 in DPBS was added to each pan and physically agitated with plastic razor blades to release the SPNPs from the plate. The SPNP suspension was collected into a Falcon tube. Each pan was agitated with fresh 0.01% Tween 20 in DPBS three times. The collected SPNPs were tip sonicated at an amplitude of 7 for 30 s (1 s on and 3 s off) in an ice bath to break up aggregates, strained through a 40 µm filter into a new Falcon tube, then centrifuged at 3220 RCF for 5 min. The supernatant was removed and distributed into 2 mL Eppendorf tubes and centrifuged for 1 h at 4 °C at 21,500 RCF. The supernatant was discarded, and the resulting pellets were combined into a single 2 mL tube. The particles were washed two times with fresh DPBS without Tween 20.

SPNP characterization

Scanning electron Microscopy: Scanning electron microscopy samples were prepared by placing a silicon wafer on top of the collection plate during the jetting process. The samples were then placed on a copper tape-covered scanning electron microscopy stub, then gold-coated for 40 s and visualized through the FEI NOVA 200 SEM/FIB instrument. The dry-state SPNP quantification of morphology parameters was conducted through ImageJ analysis¹⁰¹.

Dynamic Light Scattering: The hydrodynamic diameter and zeta potential were determined through dynamic light scattering on the Malvern Zetasizer. Samples were prepared by diluting the stock sample in DPBS and measured in folded capillary zeta cells. An average of at least three measurements was used to characterize each sample.

Bicinchoninic acid assay (BCA assay): BCA assay was used to quantify SPNP concentration. A standard curve was prepared for every SPNP concentration quantification measurement.

In vitro mouse T cell proliferation assay

To evaluate the effect of the combination of radiation (IR) treatment and autophagy inhibition, we tested the effect of IR (2 Gy) and the administration of iRGD nanoparticles loaded with a siRNA against ATG7 (ATG7i-SPNP) in vivo. Twenty mice were implanted with mIDH1-OVA mouse tumor NS as stated above, and 7 days post-implantation, mice were divided into the following 4 groups (n = 5 mice): Control (no treatment); ATG7i-SPNP only; IR only; and IR + ATG7i-SPNP. Mice were treated with 2 Gy IR for five consecutive days. ATG7i-SPNP (2.0 × 10¹¹ particles) was administered 3 times total, once every other day starting on day 5. The mice were euthanized, and the spleens were collected on day 21 post-implantation.

The spleens were collected to analyze the development of anti-tumor immune response. Splenocyte processing was performed as detailed previously¹⁰². The splenocytes were cultured with 100 nM of SIINFEKL for 24 h in 10% FBS media with 55 µM 2-ME. Cells were then stained with anti-CD45, anti-CD3, and anti-CD8 antibodies, and proliferation was assessed by CFSE dye dilution. All flow data were acquired on a FACSAria flow cytometer (BD Biosciences) and analyzed using FlowJo version 10 (Treestar).

IHC of paraffin-embedded brains

Following perfusion, mouse brains were fixed in 4% paraformaldehyde (PFA) for an additional 48 h at 4 °C, then transferred to 70% ethanol, and processed and embedded in paraffin at the University of Michigan Microscopy & Image Analysis Core Facility using a Leica ASP 300 paraffin tissue processor/Tissue-Tek paraffin tissue embedding station (Leica). Tissue was sectioned using a rotary microtome (Leica) set to 5 μm in the z-direction. Antigen retrieval and IHC of paraffin-embedded sections were performed using antibodies and dilutions as follows: Tissue sections were blocked with blocking solution (1× PBS with 0.2% Tween 20 with 5% Goat Serum) for 2 h. Sections were incubated with the following primary antibodies: anti-CD3ε (Cell Signaling, 99940, Lot: 4); anti-CD68 (Abcam, ab125212, Lot #1091902); anti-MBP (Millipore, MAB386, Lot: 4231803); anti-GFAP (Millipore, AB5541, Lot #4031343) at 4 °C overnight. Tissue sections were then labeled with Secondary Biotinylated Ab (1:1000) in PBS with 0.2% Tween 20 for 15 min at RT, then 4 °C overnight. ABC Avidin-Biotin-COMPLEX Binding was done using VECTASTAIN ABC Reagent (Vectastain Elite ABC-HRP kit; Vector Laboratories, PK-6100). Sections were incubated for 1 h with VECTASTAIN ABC Reagent in the dark and then washed with gentle agitation. Colorimetric Detection of Peroxidase was performed using the Betazoid DAB Chromogen kit (BioCare BDB2004) according to the manufacturer’s instructions. Afterward, Hematoxylin Counterstain was done, and coverslips were mounted with a xylene-based mounting medium and allowed to dry on a flat surface at room temperature until imaging. Images were obtained using brightfield/epifluorescence (Zeiss Axioplan2, Carl Zeiss MicroImaging) or laser scanning confocal microscopy (Leica DMIRE2, Leica Microsystems) and analyzed using LSM5 software (Carl Zeiss MicroImaging). Immunostaining was performed using the Discovery XT processor (Ventana Medical Systems).

To assess ATG7 expression levels post treatment of ATG7i-SPNPs, we implanted mice intracranially with 2 × 10⁴ NPAI cells on Day 0. At day 10, mice were randomly divided into 2 groups of 3 mice each. One group was treated with saline and the other group with ATG7i-SPNPs in combination with IR. Mice received a total of 10 Gy of radiation over 5 days. Mice were administered three doses of ATG7i-SPNPs (2 × 10¹¹) at days 10, 12, and 14. At the end of day 14, mice were perfused, and brain and liver tissues were collected for further analysis. Brain sections from these mice were stained for MBP (myelin basic protein) (Millipore, MAB386, Lot: 4231803), GFAP (glial fibrillary acidic protein) (Sigma, AB5541, Lot #4205396), ATG7 (Invitrogen, PA5-35203, Lot #ZH4433739), and cleaved Caspase-3 (Cell Signaling, 9661, Lot: 47). Quantification of markers was done using absolute counts of 3 representative images and the statistical analyses were conducted using two-tailed Student’s t-test.

To assess the immune profile in mIDH1-glioma-bearing mice immediately post treatment with ATG7i-SPNPs, we implanted mice intracranially with 2 × 10⁴ NPAI cells on Day 0. At day 10, mice were randomly divided into 3 groups of 3 mice each. One group was treated with saline, one group was treated with IR, and the final group was treated with ATG7i-SPNPs in combination with IR. Mice received a total of 10 Gy of radiation over 5 days. Mice were administered three doses of ATG7i-SPNPs (2 × 10¹¹) at days 10, 12, and 14. At the end of day 14, mice were perfused, and brain and liver tissues were collected for further analysis. Brain sections from these mice were stained for anti-CD4 (Cell Signaling Technology, 48274, Lot: 3), anti-CD8 (Abcam, ab217344, Lot: 1112658-20), anti-CD11c (Novus biologicals, MAB6950-SP, Clone #435421), anti-NK1.1 (BioLegend, 108702, Lot: B255763), anti-CD68 (Abcam, ab125212, Lot #1091902), and anti-IBA1 (Abcam, ab178846, Lot #1090399-30). Quantification of markers was done using absolute counts of 3 representative images, and the statistical analyses were conducted using two-tailed Student’s t-test.

Statistical analysis

All quantitative data are presented as the mean ± SEM from at least three independent samples. ANOVA and two-sample t-tests were used to compare continuous outcomes between groups. Survival curves were analyzed using the Kaplan–Meier method and compared using Mantel–Cox tests; the effect size is expressed as median survival (MS). Differences were considered significant if P < 0.05. All analyses were conducted using GraphPad Prism software (version 6.01), SAS (version 9.4, SAS Institute), or R (version 3.1.3). The statistical tests used are indicated in each figure legend.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: M.G.C., F.J.N. Methodology: All authors. Investigation: All authors. Visualization: All authors. Project administration: M.G.C. Supervision: M.G.C., P.R.L., C.A.L., J.L. Writing – original draft: All authors. Writing – review & editing: All authors. Co-first authors: F.J.N. is listed first because he was part of the conceptualization, developed the genetically engineered mouse model, and he also performed the initial experiments shown in this manuscript. All other co-first authors (K.B., A.A.M., A.M., C.E.T., Z.Z.) were placed in alphabetical order and contributed equally to the experiments presented, writing and editing, data analysis, and preparation of figures for publication.

Peer review

Peer review information

Nature Communications thanks Kelong Fan and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the NIH/National Institute of Neurological Disorder & Stroke (NINDS) Grants R37-NS094804, R01-NS122165, and R21-NS123879-01, the National Cancer Institute (NCI) Cancer Center Support Grant 2P30CA46592, and the Rogel Cancer Center Faculty Scholar Award (to M.G.C.); NIH/NINDS Grant R01-NS124167 (to M.G.C. and J.L.); NIH/NINDS Grants R01NS122234 and R01-NS122378 (to P.R.L.); The Pediatric Brain Tumor Foundation, Pediatric Cancer Foundation and Ian’s Friends Foundation (to M.G.C. and P.R.L.); NCI Grants R01CA248160 and R01CA244931 (to C.A.L.); and the National Science Foundation Graduate Research Fellowship Grant DGE 1256260 (to A.M.).

Data availability

There are no limitations on data availability. The mouse RNA-seq data used in this study are available in the Gene Expression Omnibus (GEO) database under accession code GSE94902, and the mouse ChIP-seq data under accession code GSE99806. The Bru-seq data generated from mouse glioma neurosphere cultures have been deposited in the GEO database under accession code GSE344389. These datasets were previously published by Núñez et al.¹⁵ and are publicly available. The human ChIP-seq data used in this study are available in the GEO database under accession code GSE220715. This dataset was previously published by Carney et al.¹⁸ and is publicly available. The single-cell RNA-seq data from human glioma specimens used in this study are available in the GEO database under accession code GSE152273. This dataset was previously published by Alghamri et al.¹² and is publicly available. The mouse CUT&RUN data generated in this study have been deposited in the GEO database under accession code GSE344027. The single-cell RNA-seq data generated from human glioma cultures in this study have been deposited in the GEO database under accession code GSE261042. All other data generated within this study are available via the supplementary data files; cells and plasmids will be freely distributed upon request from the corresponding author, Maria G Castro. Source data are provided with this paper.

Competing interests

In the past three years, C.A.L. has consulted for Astellas Pharmaceuticals, Odyssey Therapeutics, Third Rock Ventures, and T-Knife Therapeutics, and is an inventor on patents pertaining to Kras regulated metabolic pathways, redox control pathways in pancreatic cancer, and targeting the GOT1-ME1 pathway as a therapeutic approach (US Patent No: 2015126580-A1, 05/07/2015; US Patent No: 20190136238, 05/09/2019; International Patent No: WO2013177426-A2, 04/23/2015). All other authors declare that there is no conflict of interest regarding the publication of this article.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77320-7.