Protein kinase B (PKB/AKT) protects IDH-mutated glioma from ferroptosis via Nrf2

Abstract

Purpose:

Mutations of the isocitrate dehydrogenase (IDH) gene are common genetic mutations in human malignancies. Increasing evidence indicates that IDH mutations play critical roles in malignant transformation and progression. However, the therapeutic options for IDH-mutated cancers remain limited. In this study, the investigation of patient cohorts revealed that the phosphatidylinositol-3 kinase (PI3Ks)/protein kinase B (AKT) signaling pathways were enhanced in IDH-mutated cancer cells.

Experimental Design:

In this study, we investigated the gene expression profile in IDH-mutated cells using RNA sequencing after the depletion of AKT. Gene set enrichment analysis (GSEA) and pathway enrichment analysis were used to discover altered molecular pathways due to AKT depletion. We further investigated the therapeutic effect of the AKT inhibitor, Ipatasertib, combined with Temozolomide (TMZ) in cell lines and preclinical animal models.

Results:

GSEA and pathway enrichment analysis indicated that the PI3K/AKT pathway significantly correlated with Nrf2-guided gene expression and ferroptosis-related pathways. Mechanistically, AKT suppresses the activity of GSK3β and stabilizes Nrf2. Moreover, inhibition of AKT activity with ipatasertib synergizes with the genotoxic agent temozolomide, leading to overwhelming ferroptotic cell death in IDH-mutated cancer cells. The preclinical animal model confirmed that combining ipatasertib and TMZ treatment prolonged survival.

Conclusions:

Our findings highlighted AKT/Nrf2 pathways as a potential synthetic lethality target for IDH mutated cancers.

Statement of translational relevance

Isocitrate dehydrogenase (IDH) mutations are frequent genetic abnormalities in several human malignancies. Effective molecular targeting approaches are unavailable to improve cancer disease outcomes with IDH-mutated cancers. Our present study demonstrated that the latest inhibitors of Akt, such as ipatasertib and MK-2206, could effectively impact the expansion of IDH-mutated cancer cells in vitro and in vivo. Mechanistically, we demonstrate that Akt/Nrf2-guided antioxidant pathway is indispensable to protecting IDH-mutated cells from metabolic and genotoxic stresses. The efficacy of the Akt inhibitor was confirmed through a preclinical orthotopic xenograft model, highlighting the potent synergy between the Akt inhibitor and alkylating agent temozolomide and prolonged overall survival and slowed tumor onset. The Akt inhibitors tested in the present study are actively pursued in several early-phase clinical studies. Including Akt inhibitor is likely safe and benefits patients with brain tumors.

Introduction

Mutations in the isocitrate dehydrogenase (IDH1/2) are common genetic abnormalities in human malignancies, such as glioma, leukemia, cholangiocarcinoma, chondrosarcoma, and neuroendocrine tumors(1,2). The mutant IDH enzyme exhibits neomorphic activity that results in the production of oncometabolite D-2-hydroxyglutarate (D-2-HG). IDH-mutated cancers exhibit unique biologic patterns, such as genome-wide DNA/histone hypermethylation phenotype, homologous recombination DNA repair deficiencies, and metabolic network fluctuations (3). Several therapeutic approaches have been developed for more effective molecular targeting approaches for IDH-mutated malignancies. Multiple clinical studies described the development and validation of synthetic mutant IDH inhibitors (4,5). Although these inhibitors of IDH mutations exhibit ideal efficacy in hematopoietic malignancies, the same treatment provided only modest response rates in solid tumors(6,7). DNA repair inhibitors, such as inhibitors of poly(ADP-ribose) polymerase (PARP), have been reported to have therapeutic values for IDH-mutated glioma. PARP inhibitors, alone or in combination with standard genotoxic therapy, have been proposed by several studies and are currently under investigation in early-phase clinical studies to treat IDH-mutated cancers(8,9). However, despite the remarkable progress in IDH-mutated hematopoietic malignancies, similar responses for these therapies, when used against IDH-mutated solid tumors, remain extremely limited.

The phosphatidylinositol-3 kinase (PI3Ks)/AKT modulates numerous pathways associated with cell survival, cell cycle, and cell proliferation. The abnormal activation of the PI3K/AKT pathway has been frequently identified in human malignancies, which is essential for tumor development and treatment response(10). Gain-of-function mutations in PI3K/AKT pathway are commonly seen in IDH-mutated malignancies such as oligodendrogliomas and astrocytomas(11). Mutant IDH enzyme results in the hyperactivation of the PI3K/AKT/mTOR pathway through inhibition of Lysine-specific demethylase 4A (KDM4A)(12,13). The activation of the PI3K/AKT pathway is essential for promoting cancer aggressiveness, such as enhanced endocytosis and cellular motility in IDH-mutated cells(14). Although increasing evidence implies the critical role of the PI3K/AKT pathway in IDH-mutated glioma(12,15), it is not clear which downstream pathways are essential for the disease progression. There is also a lack of evidence determining if targeting the PI3K/AKT pathway could effectively control IDH-mutated tumor progression.

In the present study, we confirmed the activation of the PI3K/AKT pathway in IDH-mutated cells and patient specimens. Gene expression profiling suggested that AKT guides cytoprotective pathways in IDH-mutated cells. Mechanistically, AKT activation suppresses the Nrf2 E3 ligase βTrCP, which results in constitutive activation of Nrf2-mediated gene transcription and resistance to ferroptotic cell death. Combining AKT inhibitor with genotoxic therapy showed synergistic cytotoxicity in IDH1-mutated cells, significantly suppressing tumor progression and improving disease outcomes.

Materials and Methods

Patients Samples information

This study was approved by the Institutional Review Board of the National Cancer Institute. All patients gave written informed consent. Glioma patient samples were obtained from patients undergoing surgery at the National Institutes of Health (NIH). Eight glioma samples, including 3 anaplastic oligodendroglioma, 1 oligodendroglioma, and 4 astrocytoma samples, were used in this study. The information on patient samples is listed in Supplementary Table.1. The study were conducted in accordance with the U.S. Common Rule.

Cell culture and reagents

The U87 IDH wild-type and R132H cell lines were described previously(16). Cells were cultured in DMEM/F12 supplemented with 10% FBS, penicillin, and streptomycin (Gibco). The IDH1 wild type and R132H transduced immortalized human astrocyte (NHA) were kindly provided by Dr. Russell Pieper. TS603 cells (RRID:CVCL_A5HW) were kindly provided by Dr. Timothy Chan. NHA cells were maintained in complete astrocyte medium (ScienCell Research Laboratories). The brain tumor-initiating cells (BTIC) were reported previously(17). The GSC827 and GSC923 cells were derived from IDH1 wild-type patient samples. BT142 cells (RRID:CVCL_D718) were purchased from American Type Culture Collection (ATCC). All BTIC cells were cultured in the Neurobasal-A medium (Gibco) supplemented with N2, B27, EGF (20 ng/mL), and bFGF (20 ng/mL). All cell lines are tested negative for mycoplasma within 5 years. The AKT inhibitors ipatasertib and MK-2206 (TargetMol) are dissolved in DMSO.

Luciferase Reporter Assay

Nrf2/ARE transcriptional activity was determined using a Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer’s protocol. Forty-five nanogram of plasmid pGL4.37[luc2P/ARE/Hygro] (Promega, E3641) was mixed with 5 ng pRL-CMV (Promega, E2261) and EGFP-Nrf2 plasmids (Addgene 21549). The plasmid mixtures were transfected into 10⁴ cells using FuGene HD (Promega). Luminescence was measured using a microplate reader and normalized to the Renilla luciferase signal.

Western blotting

Protein was extracted using RIPA lysis buffer supplemented with Halt protease and phosphatase inhibitor cocktail (Thermo Fisher). Protein quantification was performed with a DC Protein Assay Kit (Bio-Rad). Equal amounts of protein were separated on NuPAGE 4-12% Bis-Tris mini gels (Life Technologies) and transferred to polyvinylidene difluoride membrane (Millipore). The membrane was incubated with primary antibodies and visualized with an HRP/chemiluminescence kit on the ChemiDoc Imaging System (Bio-Rad). The primary antibodies used in the present study were phospho-AKT (CST, 4060, 1:1,000), AKT (CST, 4691, 1:1,000), AKT1 (CST, 2938, 1:1,000), AKT2 (CST, 3063, 1:1000), phosphor-GSK3β (Ser9, CST, 5558, 1:1000) , GSK3β (CST, 12456, 1:1,000), xCT (Abcam, ab37185, 1:1,000), NQO1 (Abcam, ab34173, 1:1,000), GCLC (Abcam, ab41463, 1:1,000), HA-tag (CST, 3724, 1:1,000), EGFP (Thermo Fisher, A-11122; Proteintech, 66002-1-lg, 1:1,000), β-actin (CST, 4967, 1:2,000).

RNA extraction and quantitative real-time PCR (RT-PCR)

RNA was extracted using an RNeasy Mini Kit (Qiagen) and reverse transcribed to cDNA using Superscript IV VILO Master Mix (Thermo Fisher). Gene expression was analyzed with the Human Oxidative stress Plus PCR Array (PAHS-065YA, Qiagen) or with the SYBR Green Power Master Mix Kit (Applied Biosystems). Primers purchased from Qiagen include GCLC, GCLM, HMOX1, NQO1, and ACTB.

RNA sequencing and data analysis.

RNA sequencing was performed as previously described(14). Three hundred nanogram of total RNA was used as the input to an mRNA capture approach with oligo-dT coated magnetic beads. The mRNA is fragmented, and then a random-primed cDNA synthesis is performed. The resulting double-stranded cDNA is used as the input to a standard Illumina library prep with end-repair, adapter ligation, and PCR amplification performed to give you a sequencing-ready library. The final purified product is then quantitated by qPCR before cluster generation and sequencing. These TruSeq Stranded mRNA libraries were pooled and sequenced on one NextSeq2000 using a 2 x 76 cycle kit for a pair-end run. The HiSeq Real-Time Analysis software (RTA 1.18.64) was used for processing raw data files. The Illumina bcl2fastq2.17 was used to demultiplex and convert binary base calls and qualities to a fastq format. The sequencing reads were trimmed adapters and low-quality bases using Cutadapt (version 1.18). The trimmed reads were mapped to the human reference genome (hg38) and GENCODE annotation v30 (Ensembl v96 GTF file) using STAR aligner (STAR, RRID:SCR_004463, version 2.7.0f) with a two-pass alignment option. RSEM (version 1.3.1) was used for gene and transcript quantification based on the GENCODE v30 GTF file. The transcript quantification was further analyzed with Qlucore Omics Explorer (v3.8).

Gene set enrichment analysis (GSEA)

RNA sequencing data was proceeded for GSEA analysis using the GSEA software (Gene Set Enrichment Analysis, RRID:SCR_003199, v4.2.1) from the Broad Institute(18). Gene collections were obtained from MSigDB (v7.5.1). The GSEA result is further visualized with Cytoscape (Cytoscape, RRID:SCR_003032, v3.9.1) with Enrichment map (v3.3.2) and clusterMaker2 (v2.0) plug-ins according to the manufacturer’s protocol. The GSEA result is also plotted by OriginPro 2022 (v9.9.0.225).

Cell viability assay

One thousand cells were seeded in a 96-well plate. Cell viability was monitored using a Cell Counting Kit-8 (Dojindo) according to the manufacturer’s protocol. The plate was analyzed at 450 nm on an Epoch plate reader (BioTek).

Live/dead cell staining

The dead cell population was determined using a LIVE/DEAD violet cell stain kit (Thermo Fisher) according to the manufacturer’s protocol. Cells were lifted and stained with 1 μL of the fixable violet dead cell stain for 30 min. Cells were analyzed on a flow cytometer, and the dead cell population was highlighted with a 450 nm channel.

Immunohistochemistry (IHC) and TUNEL assay

Frozen normal brain tissue (n=1) and patient samples, including astrocytoma (n=4) and oligodendroglioma (n=4), were sectioned and proceeded with hematoxylin and eosin (H&E) and IHC staining. Antibodies targeting p-AKT and GCLC were used. The slides were detected using diaminobenzidine and counterstain using hematoxylin. The slides were analyzed using a light microscope. For each sample, three Regions of Interest (ROI) were imaged and quantified using the color deconvolution function of ImageJ (RRID:SCR_003070). Mice brain tissues were fixed in formalin, embedded in paraffin, and sectioned. Formalin-fixed, paraffin-embedded (FFPE) sections were processed for H&E and IHC staining. Antibodies targeting ki67 (Abcam, ab15580), PCNA (Santa Cruz, sc-56), gH2A.X (Millipore, 05-636), GCLC (Abcam, ab41463), and NQO1 (Abcam, ab34173) were used. TUNEL assay was performed using DeadEnd Colorimetric TUNEL System (Promega) following the manufacturer’s protocol. Six ROIs for each group were imaged and quantified.

Immunoprecipitation

Co-immunoprecipitation was performed using a crosslink magnetic IP/Co-IP kit (Thermo Fisher). Cells were washed with ice-cold PBS, and the protein was extracted using IP lysis buffer supplemented with Halt proteinase and phosphatase inhibitor cocktail (Thermo Fisher). Prewashed beads were crosslinked with antibodies using disuccinimidyl suberate and washed with IP lysis buffer. The cell lysate was incubated with antibody-conjugated beads overnight, washed twice with wash buffer, and eluted with elution buffer. The eluent was resolved on bis-tris gels and detected with western blotting protocol.

Cycloheximide pulse-chase assay

Cycloheximide pulse-chase assay was performed as previously described(19). The Nrf2-EGFP plasmid (Addgene 21549) was transfected into NHA cells using Lipofectamine 2000 (Thermo Fisher). Forty-eight hours later, cells were exposed to 200 μg/mL cycloheximide (Sigma). Samples were collected at designated time points and analyzed through immunoblotting.

Annexin V/PI staining

Cellular apoptosis was analyzed using Pacific Blue conjugated Annexin V and propidium iodide (PI, Thermo Fisher). Cells were lifted and washed in cold PBS. Cells were labeled with Annexin V and PI for 15 min and analyzed on a flow cytometer.

Lipid peroxidation assay

Lipid peroxidation was measured using a lipid peroxidation (Malondialdehyde, MDA) assay kit (Abcam). Briefly, cell lysate was mixed with thiobarbituric acid (TBA) solution at 95°C for 60 min to generate an MDA-TBA adduct. Then, the reaction mix was analyzed with a microplate reader colorimetrically at OD = 532 nm. The MDA signal was normalized to protein quantification.

Limiting dilution assay

BTIC GSC827 and TS603 were seeded in a 96-well plate at 200, 100, 50, 25, and 12 cells per well (8 replicate wells per condition). Cells were treated with 200 mM TMZ, 1 mM Ipatasertib and 5 mM Ferrstain-1. After14 days, each well was analyzed for the presence of at least one tumor sphere. The results were analyzed as previously described (16).

Xenograft animal modeling and treatment

CB-17 Scid mice (n=40) were intracranially injected with 2x10⁵ TS603 cells. Fourteen days after the injection, the mice were randomized into four groups (n=10). Mice were treated with 10% DMSO as control, TMZ (10 mg/kg), ipatasertib (40 mg/kg), or TMZ + ipatasertib. The first-round gavage treatment was performed on day 15, followed by the second-round treatment on day 30. The mice’s body weight was monitored until they reached the experimental endpoint. The brain tissues were harvested for histology analysis. Kaplan-Meier analysis was used to analyze the mice’s survival over time. All animal studies were approved and conducted under the National Cancer Institute (NCI) Animal Use and Care Committee.

Statistical analysis

Statistical analysis was performed with GraphPad Prism software (RRID:SCR_002798, v8.4.3). A one-way ANOVA or Student t-test was used for statistical comparisons. All statistical analyses were conducted with two tails. Results are shown as mean ± SD, and p < 0.05 was considered statistically significant.

Data Availability Statement

The data generated in this study are available within the article and its supplementary data files. The gene expression profile data were generated at Center for Cancer Research Sequencing Facility (CCR-SF) but processed data are available from authors.

Results

AKT phosphorylation is essential for glioma cells with pathogenic IDH1 mutation

We and others have reported elevated PI3K/AKT activity in IDH-mutated cells(14,20,21). To better understand the PI3K/AKT pathway in glioma, we analyzed the gene expression profile in the patient cohort through The Cancer Genome Atlas (TGCA) lower-grade glioma (LGG) dataset, which contains 119 IDH1 wild-type and 395 IDH1-mutated samples(22). Several critical genes in PI3K/AKT pathway, e.g., PIK3CA, PIK3CB, MYC, EIF4EBP1, and RPS6KB1, are significantly upregulated in gliomas carrying IDH mutation (Fig.1A). Similarly, the gene set enrichment analysis (GSEA) demonstrated that the mTOR signaling pathways are highly enriched in IDH1-mutated gliomas (Fig.1B, Supplementary Fig. 4A). We confirmed the activity of PI3K/AKT pathway in clinical specimens and patient-derived cell line models. AKT phosphorylation was significantly elevated in tumor specimens from IDH-mutated astrocytoma and oligodendroglioma (Fig.1C–D). We confirmed the up-regulation of p-AKT in brain tumor-initiating cells (BTIC) with IDH mutation (Fig.1E). The up-regulation of p-AKT was also observed in U87 or NHA cells transduced with stable expression of pathogenic IDH R132H mutant (Fig. 1E).

Figure 1. AKT phosphorylation is essential for glioma cells with pathogenic IDH1 mutation.

(A) mRNA expression of PI3K/AKT pathway signature genes (PIK3CA, PIK3CB, MYC, EIF4EBP1, and RPS6KB1) was analyzed in 514 glioma patients in TCGA. (B) GSEA compared the mTOR signaling pathway between IDH1^WT and IDH1^Mut patient data from TCGA. (C) Immunohistochemistry (IHC) staining of H&E, p-AKT, and GCLC in glioma patient samples. Bar = 50 μM. (D) Quantification and statistical analysis of IHC staining. (E) Immunoblotting analysis measured the AKT phosphorylation in brain tumor-initiating cells (BTICs), as well as established glioma cell lines U251 and U87 according to IDH1 mutation status. (F and G) CCK8 assay measured the cell viability in IDH1^WT and IDH1^Mut cells after the depletion of AKT using siRNA (F) or 1 μM ipatasertib (G). (H) Dose-response curve evaluated the cell viability of IDH1^WT and IDH1^Mut NHA cells in response to AKT inhibitor Ipatasertib and MK-2206. *p < 0.05. **p < 0.01. ***p < 0.001.

To better understand the role of the AKT pathway in IDH-mutated cells, we established a loss-of-function cellular model through small interference RNA (siRNA) targeting AKT (Supplementary Fig.1A). Growth suppression was found in both IDH wild-type and mutant cells after depletion of AKT. A greater extent of suppression was seen in IDH mutated cells through a 4-day observation (Fig. 1F). Similarly, treatment with the AKT inhibitor ipatasertib resulted in growth suppression in both IDH wild-type and mutated cells, with greater efficacy in IDH-mutated NHA or U87 cells (Fig. 1G). A dose-response study confirmed that NHA cells contained an IDH-mutation were more vulnerable to the treatment with AKT inhibitors ipatasertib (IDH1^WT, IC50 = 15.12 μM; IDH1^Mut, IC50 = 5.847 μM) or MK-2206 (IDH1^WT, IC50 = 4.045 μM; IDH1^Mut, IC50 = 2.006 μM) (Fig. 1H). Ipatasertib treatment significantly increased cell death in NHA IDH1-mutated cells compared with their wild-type counterparts (Supplementary Fig1B–D).

AKT activity correlates with Nrf2/antioxidant pathway in IDH-mutated cells

AKT is a protein kinase that affects multiple biological pathways by phosphorylating its myriad substrates. For example, AKT promotes cellular survival by phosphorylating BAD on Ser136, which disaggregates BAD from the pro-apoptotic Bcl-2/Bcl-X complex(23). Similarly, AKT affected cellular metabolism by inhibiting glycogen synthase kinase 3 (GSK-3) via phosphorylation(24). Our findings above showed that the AKT signaling pathway correlates with cell survival in IDH-mutated cells, in agreement with several recent reports showing that AKT-dependent metabolic pathways assist cell survival and tumor expansion for IDH-mutated malignancies (15,25–27). To better understand the precise molecular targets of AKT in these cancers, we investigated the gene expression profile in AKT-depleted IDH-mutated cells through RNA sequencing. Firstly, the loss of AKT expression resulted in the down-regulation of metabolism-related genes such as glutamate-cysteine ligase (GCLC and GCLM), aldo-keto reductase (AKR1C1 and AKR1B10), and glutathione S-transferase Mu 3 (GSTM3). Furthermore, some of the interferon pathway-related genes were found to be upregulated, such as the interferon-induced protein with tetratricopeptide repeats 2 (IFIT2), Interferon-induced GTP-binding protein Mx1 (MX1), and 2′-5′-oligoadenylate synthetase 2 (OAS2) (Fig.2A). We investigated the changes in molecular pathways through GSEA analysis. We found that Nrf2-guided gene expression (NES = −1.64, q = 0.029), as well as ferroptosis-related pathways (NES = −1.76, q = 0.031), were under-represented when AKT expression was depleted (Fig.2B). In addition, pathway enrichment analysis confirmed the downregulation of Nrf2 and mTOR/AKT-related pathways (Fig.2C). Cancer aggressiveness pathways, such as metabolic reprogramming, angiogenesis, and lipid metabolism, were also found reduced with AKT depletion. Overall, we confirmed that the AKT depletion resulted in apparent changes in PI3K/mTOR/metabolic pathways. The Nrf2/antioxidant pathway also seems to be extensively depleted, suggesting that PI3K/AKT supports IDH-mutated cells by prompting the Nrf2-related cytoprotective pathway (Fig.2D). Other pathways, such as cellular differentiation, interferon signaling, and immune activation, were prominent in AKT depleted cells, suggesting that AKT suppression may lead to additional charges in the biological pattern of IDH mutant cells, which may be helpful to understand AKT-related molecular pathways in glioma.

Figure 2. AKT activity correlates with Nrf2/antioxidant pathway in IDH-mutated cells.

(A) Volcano plots of downregulated and upregulated genes in AKT–depleted NHA cells with IDH1 mutation. (B) GSEA compared the Nrf2 pathway and ferroptosis pathway between the control and siAKT group in IDH1-mutated NHA cells. (C) Pathway enrichment analysis showed the down-regulated pathways related to AKT depletion. (D) Illustration of pathways that are upregulated (red) or down-regulated (blue) in AKT-depleted, IDH1-mutated NHA cells.

AKT stabilizes Nrf2 through deactivating GSK3β/β-TrCP E3 ligase

Nrf2 is active in many human malignancies with a metabolic deficiency or abnormality, such as solid tumors with mutations in the Krebs cycle (16,28). The unexpected finding of the AKT/Nrf2 pathway suggests that AKT may strengthen Nrf2-related gene transcription in IDH-mutated glioma cells. To better understand the relationship between AKT catalytic function and Nrf2 activation, we examined the expression of Nrf2-guided cytoprotective genes in U87 cells with doxycycline-inducible expression of IDH1 R132H mutant. We found the expression of IDH mutant resulted in up-regulation of GCLC, GCLM, HMOX1, and NQO1. In contrast, the expression of these genes is suppressed with the presence of small interference RNA targeting AKT (Fig.3A). Similarly, immunoblotting revealed that expression of IDH mutant resulted in up-regulation of p-AKT and p-GSK3β, as well as Nrf2-regulated genes such as xCT transporter, NQO1, and GCLC (Fig.3B). Pharmacologic inhibition of AKT with ipatasertib resulted in the accumulation of phosphorylated AKT in both IDH wild type and mutant NHA cells, whereas the downstream AKT pathway was inhibited. This phenomenon is consistent with several prior studies (29,30). We discovered that the ipatasertib suppressed GSK3β phosphorylation. The expression of Nrf2-regulated genes xCT, NQO1, and GCLC was also suppressed by ipatasertib (Fig.3B). Additionally, we demonstrated the suppression of AKT activity suppression results in a reduction of Nrf2-driven transcription activity, evidenced by decreased ARE-luciferase activity (Fig.3C–D).

Figure 3. AKT stabilizes Nrf2 through deactivating GSK3β/β-TrCP E3 ligase.

(A) Real-time PCR measured the mRNA expression of Nrf2-guided downstream genes (GCLC, GCLM, HMOX1, and NQO1) in U87 cells with doxycycline-inducible expression of IDH1 R132H mutant. AKT was depleted using siRNA. (B) Immunoblotting analysis measured the alternations of AKT and Nrf2 signaling pathway in IDH1^WT and IDH1^Mut NHA cells with treatment of ipatasertib (Ipa, 1 mM). (C and D) Luciferase reporter assay measured the antioxidant response element (ARE) activity in IDH1^WT and IDH1^Mut NHA cells after the depletion of AKT using Ipa (C) or siRNA (D). (E) Immunoprecipitation assay evaluated the ubiquitination and interaction with β-TrCP of wild-type Nrf2 (Nrf2^WT), Nrf2 lacking DIDLID element (Nrf2^Δ), and Nrf2 lacking both DIDLID element and SDS motif (Nrf2^ΔΔ). (F) Cycloheximide (CHX) pulse-chase assay measured the stability of different Nrf2 variants (Nrf2^WT, Nrf2^Δ, and Nrf2^ΔΔ) in IDH1^WT and IDH1^Mut NHA cells. (G) Immunoprecipitation showed the ubiquitination of Nrf2 variants Nrf2^Δ and Nrf2^ΔΔ in response to Ipa (1 μM) treatment in IDH1^Mut NHA cells. *p < 0.05. **p < 0.01.

AKT suppresses the biological function of GSK3β by phosphorylating the Ser9 residue (31). GSK3β is a serine/threonine-protein kinase that regulates Nrf2 stability through phosphorylating the SDS motif. Phosphorylated Nrf2 is recognized by its E3 ligase β-TrCP for rapid proteasomal degradation(32). We confirmed that the GSK3β/β-TrCP E3 ligase complex is suppressed in IDH mutated cells, evidenced by reduced physical interaction between Nrf2 and β-TrCP (Fig. 3E). Furthermore, we found wild-type Nrf2 (Nrf2^WT) is extensively ubiquitinated in IDH wild-type NHA cells. The Nrf2 lacking DIDLID element (Nrf2^Δ) or both DIDLID and SDS motif (Nrf2^ΔΔ) exhibited a gradual loss of protein ubiquitination. In contrast, protein ubiquitination was absent in all Nrf2^WT, Nrf2^Δ, and Nrf2^ΔΔ in IDH-mutated NHA cells (Fig.3E). The stabilization of Nrf2 is further confirmed with a cycloheximide pulse-chase assay (Fig.3F). Nrf2 degradation was found in IDH wild-type NHA cells. Nrf2^Δ and Nrf2^ΔΔ exhibit prolonged protein half-lives. In IDH-mutated cells, Nrf2 stabilization was identified in all Nrf2^WT, Nrf2^Δ, and Nrf2^ΔΔ variants. Ipatasertib treatment in IDH1 mutated cells rescued protein ubiquitination in Nrf2^Δ but not Nrf2^ΔΔ (Fig.3G), suggesting that the GSK3β/β-TrCP-associated phosphorylation at the SDS motif in Nrf2 is related to Nrf2 degradation.

AKT/Nrf2 axis protects cancer cells from ferroptotic cell death

The bioinformatics findings described above implied that the AKT activity correlates with the alterations in the ferroptotic pathway in IDH-mutated cells. Ferroptosis is a programmed cell death mechanism initiated by lipid peroxidation (33). Our previous investigation suggested that macromolecules, such as protein, DNA, and lipid, are generally oxidized in IDH-mutated cells due to the intrinsic metabolic stress caused by IDH-mutant neomorphic activity (16). It is likely that cells carrying IDH mutation are exposed to oxidative damage, whereas AKT/Nrf2 pathway serves as a protective mechanism to relieve the stress. We confirmed that the expression of GPX4 and SLC7A11, two critical genes that protect cells from lipid peroxidation, were upregulated in IDH-mutated cells (Supplementary Figure 2A). In line with this finding, we found that IDH-mutated cells are more resistant to the ferroptosis inducers, erastin and RSL3 (Fig.4A. For Erastin, IDH^WT IC50 = 2.152 μM; IDH^Mut IC50 = 6.517 μM. For RSL3, IDH^WT IC50 = 16.60 nM; IDH^Mut IC50 = 101.3 nM). However, the resistance to ferroptosis is strongly reduced by ipatasertib (For Erastin + Ipa, IDH1^WT IC50 = 1.049 μM; IDH1^Mut IC50 = 1.888 μM; For RSL3 + Ipa, IDH1^WT IC50 = 10.18 nM, IDH1^Mut IC50 = 22.91 nM). Moreover, ipatasertib increased reactive oxygen species (ROS) levels and reduced glutathione levels in IDH1-mutated cells, which might indicate elevated ferroptotic changes (Supplementary Fig.2B–C).

Figure 4. AKT/Nrf2 axis protects cancer cells from ferroptotic cell death.

(A) Dose-response curve measured the cell viability of IDH1^WT and IDH1^Mut NHA cells in response to Erastin and RSL3, combined with treatment of Ipatasertib (Ipa, 1 μM). (B) Dose-response curve measured the cell viability in IDH1^Mut NHA cells in response to TMZ/Ipa combination treatment. (C) isobologram analysis showed the synergistic effect of TMZ and Ipa in IDH1^Mut NHA cells. (D) Combination index (C.I.) was calculated to show the synergistic effect of TMZ and Ipa. (E) Representative flow cytometric analysis of Annexin V and PI staining measured cell apoptosis of IDH1^WT and IDH1^Mut NHA cells in response to TMZ (200 μM)/Ipa (1 μM) treatment. (F) Statistical analysis of apoptotic alternations in 4E. (G) Representative flow cytometric analysis of live/dead staining in IDH1^WT and IDH1^Mut NHA cells in response to TMZ/Ipa treatment. Ferroptosis inhibitor, Ferrstain-1 (Ferr-1, 5 μM) was used to rescue the ferroptotic cell death. (H) Statistical analysis of dead/live cell ratio in IDH1^WT and IDH^Mut NHA cells in 4G. (I) CCK8 assay measured the cell viability of IDH1^WT and IDH1^Mut NHA cells after being treated with TMZ/Ipa. (J) lipid peroxidation (Malondialdehyde, MDA) assay measured the lipid peroxidation of IDH1^WT and IDH1^Mut NHA cells after being treated with TMZ/Ipa. (K) CCK8 assay measured the cell viability of IDH1^WT and IDH1^Mut NHA cells after being treated with TMZ/Ipa. The iron chelator deferoxamine (DFO, 100 μM) was used to rescue the ferroptotic cell death. **p < 0.01.

AKT inhibition synergizes with genotoxic therapy for IDH-mutated cells

The discovery of the AKT/Nrf2 cytoprotective mechanism suggests that this may be a novel therapeutic target to better impact IDH-mutated cancers such as glioma. To evaluate the drug-drug interaction between AKT inhibitor and temozolomide (TMZ), the current standard for care for glioma, we performed a dose-response study on NHA cells transduced with a pathogenic IDH mutant (Fig.4B). Ipatasertib treatment resulted in a significant reduction in the IC50 dose in IDH mutant NHA cells (TMZ IC50 = 535 μM, TMZ + Ipa IC50 = 250 μM). In addition, a more in-depth dose-response study and isobologram analysis indicated distinctive drug-drug interactions in IDH mutant cells (Fig.4C). Ipatasertib exhibited strong synergy with TMZ in IDH-mutated cells, evidenced by remarkably reduced combination index (C.I. = 0.498) (Fig.4D). In contrast, the same drug combination resulted in minimal synergy in IDH wild type cells (C.I. = 0.912, Supplementary Fig. 4B–D). Further, we performed an annexin V/PI flow cytometry analysis on ipatasertib and/or TMZ-treated cells (Fig.4E). Consistent with the dose-response study, combination therapy was more effective in IDH mutant cells, with a larger propidium iodide (PI)⁻positive dead cell population identified under combination treatment (Fig.4F). Interestingly, we noticed that the elevated cell death may not come from an apoptotic route, as the Annexin V positive population was not vastly increased (Fig.4F). Similarly, the apoptotic marker cleaved PARP was found unchanged comparing cells treated with TMZ and combination regimen (Supplementary Fig.2D), suggesting the ipatasertib resulted in cell death that is not mediated by caspase-related mechanism. Phosphorylation-insensitive Nrf2 (Nrf2^ΔΔ), but not wild-type Nrf2, abolished Ipa chemosensitization, suggesting the combination regimen acts through Nrf2 phosphorylation at the β-TrCP site (Supplementary Fig. 2E).

To better understand the molecular mechanism of Nrf2-derived cytoprotection, we investigated the phenotypic outcome of combination treatment with the presence of a known ferroptosis inhibitor, ferrostatin-1 (Ferr-1). Ferr-1 is a small molecule that rescues ferroptotic cell death by preventing iron-induced lipid peroxidation (34). Firstly, we performed a flow cytometry assay on NHA cells to distinguish between live and dead cells. Quantitative analysis showed more robust cell death induction in the Ipa and TMZ combination treatment in IDH mutant cells. The portion of dead cells decreased with the addition of Ferr-1 in both IDH wild-type and mutant cells, suggesting the possible ferroptotic cell death in both cell lines. (Figure 4G–H). Further, we performed a cell viability test and showed that Ferr-1 treatment increased cell viability in IDH-mutated cells after receiving TMZ and Ipa treatment (Figure 4I). To further validate the ferroptotic changes in IDH1 mutant cells after TMZ and Ipa treatment, we measured the lipid peroxidation level using a malondialdehyde (MDA) assay. The result showed that ipatasertib increased lipid peroxidation while Ferr-1 restored the lipid peroxidation to the baseline level (Fig.4J). Lastly, as ferroptosis depends on intracellular iron, cell death induced by TMZ and ipatasertib combination treatment was suppressed by co-treatment with the iron chelator deferoxamine (DFO) (Fig.4K).

TMZ/ipatasertib combination treatment led to improved disease outcomes in vivo

The results reported above strongly indicate that AKT inhibition sensitizes IDH-mutated cells for traditional cancer therapeutics such as genotoxic agents. To better evaluate whether this combination regimen could enhance the treatment response, we assessed the therapeutic efficacy of the TMZ/ipatasertib combination regimen in patient-relevant disease models.

We first evaluated the combination regimen’s therapeutic efficacy in patient-derived BTIC models. Phase-contrast microscopy and live and flow cytometry confirmed that the TMZ/ipatasertib treatment induced a greater extent of cell death in IDH-mutated BTIC TS603 and BT142. The co-treatment of Ferr-1 restored the cell viability in TS603 and BT142 that received combination treatment (Supplementary Fig.3A–C). Moreover, the limiting dilution assay showed that the TMZ/ipatasertib treatment significantly suppressed the frequency of generating tumor spheres in IDH-mutated BTIC compared with its wild-type counterpart. Consistently, the Ferr-1 treatment reversed the cytotoxicity of the combination treatment (Supplementary Fig.3D).

Further, we tested the efficacy of the TMZ/ipatasertib combination regimen in a preclinical animal model bearing patient-derived IDH-mutated glioma cells. TS603 cells were intracranially injected into CB-17 Scid mice (Fig.5A). The Kaplan Meier analysis demonstrated that TMZ/ipatasertib combination treatment significantly prolonged the survival of the tumor-bearing mice (Fig.5B). Histological staining showed decreased cell proliferation markers PCNA and Ki67 in the combination treatment group (Fig.5C) with increased cytotoxic markers γH2AX and TUNEL assay (Fig.5D). Nrf2 downstream targets GCLC and NQO1 were suppressed with the combination of ipatasertib (Fig.5E). The alterations in histological staining were confirmed by densiometric analysis (Figs.5F to 5H).

Figure 5. AKT inhibition synergizes with genotoxic therapy for IDH-mutated cells.

(A) The schematic Illustration of xenograft animal modeling. Mice were treated with TMZ and/or ipatasertib (Ipa) for two rounds on day 14 and day 42 post-tumor injection. (B) Kaplan-Meier plots showed the survival curves of mice bearing TS603 xenograft treated with vehicle, TMZ (10 mg/kg), and/or Ipa (40 mg/kg). 10 mice per group. (C to E) Immunohistochemistry (IHC) staining of H&E, cell proliferation markers PCNA and Ki-67 (C), DNA damage markers γH2A.x and TUNEL assay (D), Nrf2 downstream genes NQO1 and GCLC (E), in TS603 xenograft samples. Bar = 50 μM. (F to H) Statistical analysis of IHC staining of PCNA (F), TUNEL assay (G), and GCLC (H). *p < 0.05. **p < 0.01.

Discussion

We studied the role of the AKT/Nrf2 pathway in glioma and discovered that the AKT is constitutively phosphorylated in glioma with IDH mutation, which protects the cancer cells from ferroptotic cell death. Mechanistically, AKT phosphorylation activates cytoprotective pathways via stabilizing Nrf2. Nrf2-induced transcriptional activation of metabolic genes supports the expansion of IDH-mutated cells in vitro and in vivo (Figure 6). Molecular targeting of AKT synergized with standard genotoxic therapy, which resulted in improved disease outcomes in preclinical models bearing IDH-mutated xenografts.

Figure 6. Schematic illustration of AKT/Nrf2 pathway in IDH-mutated cells.

In normal cells, Nrf2 is deactivated through βTrCP recognition and proteasomal degradation. The βTrCP recognition requires GSK3β phosphorylation on Nrf2 at the Neh6 domain. In IDH-mutated cells, D-2-HG activates AKT, which compromises the Nrf2 degradation pathway. Stabilized Nrf2 serves as a transcription factor, activates the expression of ROS-scavenging genes, and protects cells from ferroptosis.

Activation of PI3K/AKT/mTOR pathway in IDH-mutated cells

The activation of the PI3K/AKT/mTOR pathway is frequently identified in human malignancies, which supports the process of oncogenesis by promoting cell growth and survival (35,36). In IDH-mutated glioma, genes that encode the critical elements in the PI3K/AKT/mTOR pathway, such as PIK3CA, PIK3R1, and MTOR, are frequently mutated, which enable this oncogenic pathway and support the disease progression (11). Additionally, the mTOR/AKT pathway could be activated through an epigenetic mechanism, which promotes gliomagenesis by enhancing cancer aggressiveness, such as cell proliferation, endocytosis, and cellular migration (12). Pharmacological suppression of mTOR signaling is accompanied by delayed tumor manifestation with improved disease outcomes (37). Our investigation revealed that the PI3K/AKT/mTOR pathway is frequently expressed in patient specimens and brain tumor cells carrying IDH mutation. The RNA sequencing data in the TCGA-LGG dataset revealed a strong PI3K/mTOR/AKT signature in glioma carrying IDH mutation (Fig.1A and 1B). This bioinformatic finding is confirmed through laboratory investigation in patient-derived tissue and cell lines (Fig.1C–1E). The prompted PI3K/AKT/mTOR signature brings about conceptual advances in the oncogenic mechanism for IDH-mutated glioma, which also implies novel molecular targeting approaches by targeting AKT kinase activity. Indeed, the AKT inhibitor ipatasertib exhibited superior efficacy in cancer cells with an IDH-mutated genetic background (Figure 1F–H). Combining ipatasertib with TMZ resulted in remarkably enhanced chemosensitivity, evidenced by reduced IC50 and prolonged survival benefit in a preclinical animal model (Fig.4B–4D, and C; Fig.5).

PI3K/AKT/mTOR pathway protects IDH-mutated cancer cells from ferroptotic cell death

The neomorphic activity of cancer-associated IDH mutants results in not only the production of oncometabolite D-2-HG but also a profound impact on cellular metabolism (38,39). IDH mutant converts αKG into D-2-HG, which converts NADPH to NADP⁺ (40). The depletion of the NADPH pool affects many essential homeostatic pathways, such as the Krebs cycle, de novo lipogenesis, and redox homeostasis (41,42). The metabolic fluctuations, abnormal lipid metabolism, and overwhelming oxidative damage jeopardize cell physiology, leading to lipid peroxidation and ferroptotic cell death (43). The present study showed that the constitutive activation of PI3K/AKT/mTOR signaling protected IDH-mutated cells from lipid peroxidation and oxidative damage. Unbiased gene profiling on AKT-depleted cells showed absent gene sets related to Nrf2-guided gene transcription and ferroptotic pathways (Fig.2B–D). IDH-mutated cells exhibit more robust expression of the genes associated with de novo glutathione synthesis (e.g., SLC7A11, GCLC, and GCLM) and ferroptotic pathways (e.g., FTH1, FTL, TXN, and TXNRD1, Fig.2B). The activation of Nrf2 downstream prompts the resistance to ferroptotic cell death induced by erastin or RSL3 (Fig.4A). Further, we demonstrated that targeting AKT pathway synergized with TMZ-induced cytotoxicity (Figure 5B–D). The chemosensitivity is related to enhanced ferroptosis, as treatment ferrostatin-1, iron chelator DFO, salvaged cell death caused by TMZ and ipatasertib (Fig.5G, 5H, 5J, and 5I). Besides the vital role in chemosensitization, ferroptotic cell death has been frequently identified in radiation-induced cancer suppression (44). Recent studies have extensively discussed the combination of ferroptosis inducers with radiation therapy (45). Pharmacologic targeting AKT kinase activity diminishes the ROS scavenging pathway and establishes the vulnerability to ferroptosis, which may also improve the disease outcome by synergizing with radiation-induced cell death.

Synthetic lethality of targeting PI3K/AKT/mTOR pathway in IDH1-mutated cancer

IDH-mutated cancers exhibit distinctive biological and metabolic patterns and a molecular signature characterized by compromised DNA repair pathways, genome-wide epigenetic shift, and accumulation of reactive oxygen species (7,46). These signatures not only expanded the scope of oncogenic signaling in human malignancies but also implied actionable therapeutic targets. We and others have demonstrated that the combination of PARP inhibitors substantially increases the efficacy of genotoxic agents in IDH-mutated tumors (9,47,48). Similarly, suppression of the glutamine/glutathione metabolic route led to potent synergistic cytotoxicity in the context of IDH1 mutation (49,50). These findings suggest synthetic lethality may be a valuable strategy for IDH1-mutated cancers. In the present study, we demonstrated that targeting PI3K/AKT pathway is a novel therapeutic vulnerability in IDH1-mutated cancers. Our data indicated that the AKT signaling pathway protected IDH1-mutated cells from ferroptotic cell death by abrogating βTrCP-mediated Nrf2 ubiquitination and degradation (Fig.3C and 3D). Suppression of AKT using ipatasertib elevated TMZ-induced DNA damage and cytotoxicity (Fig.4B–4G). Further, a proof-of-concept preclinical study confirmed that the combination regimen of TMZ and ipatasertib suppress the manifestation of IDH-mutated xenografts with improved disease outcomes (Fig.5). Overall, our investigation demonstrated the essential roles of AKT/Nrf2 pathway in glioma with IDH mutation. We discovered the AKT inhibitor sensitizes IDH-mutant glioma for TMZ-induced cytotoxicity. A combination regimen with an AKT inhibitor and genotoxic agent may improve the therapeutic outcome for patients with IDH-mutant glioma.