Glutathione reductase deficiency potentiates the immunogenicity of ferroptosis and cuproptosis via amplified reactive oxygen species accumulation and cGAS-STING pathway activation

Abstract

Background

Cancer remains a major therapeutic challenge due to drug resistance and metastasis, processes driven by oxidative stress and redox imbalance. Targeting this vulnerability through ferroptosis (iron-dependent lipid peroxidation) and cuproptosis (copper-driven mitochondrial dysfunction), two ROS-mediated cell death pathways, offers a promising therapeutic strategy. However, clinical translation is hindered by incomplete understanding of their redox regulation and limited immunogenicity.

Methods

A genome-wide CRISPR knockout screen was performed to identify key regulators of ferroptosis. Genetic depletion or pharmacological inhibition of candidate genes was evaluated across multiple cancer cell lines for sensitivity to ferroptosis inducer RSL3 and the cuproptosis inducer elesclomol (Es). Antitumor efficacy was assessed in xenograft, orthotopic, metastatic, and syngeneic mouse models, alone or combined with immune checkpoint inhibitors. Mechanistic studies also examined ROS production, mitochondrial stress, mitochondrial DNA release, cGAS-STING activation, and immune responses within the tumor microenvironment.

Results

Glutathione reductase (GSR), a central enzyme maintaining reduced glutathione (GSH) homeostasis, was identified as the top suppressor of ferroptosis. GSR knockout or pharmacological inhibition markedly sensitized diverse cancer cell lines to RSL3-induced ferroptosis, while GSR overexpression conferred resistance. Strikingly, GSR depletion also enhanced sensitivity to cuproptosis triggered by the copper ionophore Es. In multiple in vivo tumor models, GSR inhibition synergizes with RSL3 or Es to suppress tumor growth, inhibit lung metastasis, and prolong survival. Mechanistically, GSR deficiency amplified ROS production, induced mitochondrial stress, and triggered the cytosolic mitochondrial DNA release under ferroptotic or cuproptotic stress, activating the cGAS-STING pathway in vitro and in vivo. This increased inflammatory cytokine production, promoted immunogenic cell death, and enhanced the release of damage-associated molecular patterns (DAMPs), including HMGB1. Together, GSR inhibition combined with a ferroptosis or cuproptosis inducer transformed the tumor microenvironment into a highly immune stimulatory state, thereby enhancing the efficacy of immune checkpoint blockade through increased dendritic cell activation and T-cell infiltration and activation.

Conclusions

GSR represents a key molecular node connecting and modulating ferroptosis and cuproptosis through redox regulation. Targeting GSR amplifies ROS-mediated immunogenic cell death, triggers cGAS-STING activation in cancer cells, and enhances the efficacy of cancer immunotherapy, providing a promising redox-based therapeutic strategy.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13045-026-01840-4.

Background

Oxidative stress in cancer

Cancer remains a major challenge due to therapeutic resistance and metastasis, leading to poor clinical outcomes. A central factor underlying cancer’s resilience is oxidative stress, where cancer cells both generate and neutralize reactive oxygen species (ROS) to sustain growth and survival [1, 2]. The high metabolic demands of rapidly proliferating cancer cells drive excessive ROS production via mitochondrial respiration and metabolic reprogramming [3, 4]. At moderate levels, ROS activate pathways such as AMPK and PI3K/AKT, promoting proliferation, angiogenesis, and adaptation to hypoxia [5, 6]. However, excessive ROS cause macromolecular damage and trigger cell death [7–9]. To cope with oxidative pressure, cancer cells evolve robust antioxidant defenses [1, 10]. Oncogenes such as KRAS, BRAF, and MYC elevate ROS through metabolic acceleration and mitochondrial stress, while activating antioxidant pathways such as the NRF2-KEAP1 axis to upregulate antioxidant enzymes and promote endogenous antioxidant synthesis [1, 11, 12].

Targeting redox homeostasis represents a promising therapeutic strategy [13, 14], given the important role of ROS in tumorigenesis. Although small-molecule antioxidants, such as vitamin C, carotenoid, and resveratrol have shown anticancer potential, clinical evidence remains largely inconclusive and context-dependent [2]. In contrast, targeting antioxidant pathways or combining ROS-inducing agents with standard therapies has demonstrated synergistic potential, particularly in treatment-resistant cancers such as glioblastoma and pancreatic cancer [15–18].

A key player in this strategy is reduced glutathione (GSH), the most abundant intracellular antioxidant, reaching concentrations of 1–10 mM in cancer cells, where it plays a crucial role in ROS scavenging, oxidative repair, and maintaining redox balance [1, 10, 16, 19]. GSH homeostasis is maintained by glutamate-cysteine ligase (GCL) for synthesis and glutathione reductase (GSR) for regeneration. For instance, buthionine sulfoximine (BSO), an irreversible GCL inhibitor, enhances the efficacy of cisplatin or radiotherapy by amplifying oxidative damage [16, 20].

Ferroptosis and cuproptosis in cancer

Ferroptosis is an oxidative stress-driven form of regulated cell death (RCD) characterized by iron-dependent lipid peroxidation [21, 22]. Its regulation centers on the GPX4-xCT-GSH axis, while other regulators such as FSP1, DHODH, NOS2, and GCH1 produce antioxidants with radical-trapping activity [23, 24]. Glutathione peroxidase 4 (GPX4) uses GSH to detoxify lipid peroxides, while xCT (SLC7A11) imports cystine for GSH synthesis. Inhibiting either GPX4 (e.g., RSL3) or xCT (e.g., erastin) depletes GSH, disrupts redox homeostasis, and triggers ferroptotic death [24–26].

Ferroptosis provides a potent mechanism to eliminate ROS-rich, therapy-resistant cancer cells. Its efficacy is particularly heightened in aggressive cancer subtypes, including therapy-resistant, metastatic, and epithelial-mesenchymal transition (EMT)-positive tumors [27, 28]. For instance, triple-negative breast cancers with high EMT activity exhibit elevated lipid peroxidation and increased susceptibility to ferroptosis inducers such as erastin and RSL3 [29]. Similarly, persister cancer cells, a dormant subpopulation that drives relapse, are highly dependent on GPX4, making them vulnerable to ferroptosis-based therapies [30]. Despite its therapeutic potentials, ferroptosis induction by small-molecule inducers does not consistently trigger cell death across cancer types, as susceptibility largely depends on cellular context [31, 32].

Cuproptosis is a recently identified form of RCD that has expanded the arsenal of non-apoptotic cancer therapies [33, 34]. It is triggered by intracellular copper overload, which binds directly to lipoylated enzymes of the tricarboxylic acid (TCA) cycle in mitochondria, such as dihydrolipoamide S-acetyltransferase (DLAT) [33]. This interaction induces protein aggregation, iron-sulfur cluster loss, and ROS accumulation, ultimately leading to mitochondrial dysfunction and cell death. Copper plays a dual role in cancer, while essential for angiogenesis and metastasis, excess copper is cytotoxic. Many tumors upregulate copper transporters (e.g., CTR1, ATP7A) to support their metabolic demands, creating a therapeutic vulnerability that can be exploited by supplying copper ionophores such as elesclomol (Es) to induce cuproptosis [33]. Nonetheless, the precise mechanisms and therapeutic potential of cuproptosis in cancer remain incompletely understood.

Ferroptosis and cuproptosis, though mechanistically distinct, share key features that underscore their relevance in cancer biology and treatment. Both are metal-dependent processes, driven by transition metals that catalyze Fenton-type reactions to generate oxidative stress and disrupt cellular redox homeostasis [35, 36]. Therapeutically, both circumvent apoptosis-related resistance mechanisms common in chemotherapy by inducing non-apoptotic cell death. However, whether a common regulatory pathway exists that can coordinately sensitize ferroptosis and cuproptosis both remains to be elucidated.

Immunogenicity of ferroptosis and cuproptosis in cancer treatment

Traditional cancer therapies primarily target the rapid proliferation of cancer cells to induce apoptosis. However, apoptotic cell death is largely non-immunogenic and therefore fails to effectively activate the host immune system to eliminate cancer cells [37]. In contrast, both ferroptosis and cuproptosis have been reported to display features of immunogenic cell death under certain circumstances, enabling them not only to kill cancer cells but also to enhance antitumor immunity. These metal-dependent death pathways exhibit necrotic morphology and can release damage-associated molecular patterns (DAMPs) [38], such as calreticulin (CRT) [39], high mobility group box 1 (HMGB1) [40, 41] and ATP, potentially rendering them immunogenic. The resulting immune activation promotes dendritic cell (DC) maturation, enhances antigen presentation, stimulates CD8⁺ and CD4⁺ T-cell infiltration, and reprograms the tumor microenvironment from immunosuppressive to inflammatory. For instance, ferroptotic cells have been shown to trigger bone marrow-derived DC maturation in vitro and provide protection against tumor rechallenge in mouse models [42].

Nevertheless, current evidence remains limited, and the immunogenicity of these metal-dependent death pathways appears highly context-dependent. In particular, the immunogenic potential of these processes, especially cuproptosis, has not been systematically characterized. If effectively harnessed, ferroptosis and cuproptosis could offer a powerful therapeutic strategy for hard-to-treat cancers by simultaneously eradicating cancer cells directly and (re)activating the patient’s immune system against their cancer.

In this study, we identified GSR as a key negative regulator of ferroptosis through a genome-wide CRISPR/Cas9 screen. Genetic deletion or pharmacological inhibition of GSR sensitized cancer cells not only to ferroptosis but also to cuproptosis, establishing GSR as a molecular link between these two metal-dependent cell death pathways. Although GSR is functionally linked to redox homeostasis and ferroptosis suppression, its role has not been comprehensively characterized experimentally, and its involvement in cuproptosis was previously unknown. We demonstrate that GSR deficiency enhances ferroptotic and cuproptotic cell death at low inducer doses while increasing their immunogenicity, via activation of the STING signaling pathway. Mechanistically, this treatment regimen imposes a combined oxidative burden on cancer cells by inducing ferroptosis or cuproptosis, both highly ROS-generating processes, while simultaneously depleting antioxidant defenses through GSR inhibition. Beyond promoting cell death, this strategy amplifies the cellular immune response triggered by these metal-dependent RCD pathways. Together, our studies identify GSR as a key regulator of metal-dependent RCD and suggest that its inhibition amplifies the immunogenic potential of ferroptosis and cuproptosis, offering a promising therapeutic avenue for cancer treatment.

Methods

General cell culture

Assays were performed using the following cell lines: 4T1, 4T1-Luc2, HEK293T, CT26, HeLa, T47D, A549, MCF7, and SH-SY5Y. All cell lines were obtained from ATCC unless otherwise indicated and all cell lines except for 4T1 and 4T1-Luc2 were maintained in Dulbecco’s Modified Eagle Medium (DMEM), whereas 4T1 and 4T1-Luc2 cells were cultured in RPMI-1640 medium. All culture media were supplemented with 10% fetal bovine serum (FBS; HyClone, SH30070.03HI), 100 U/mL penicillin, and 100 μg/mL streptomycin. Cells were maintained at 37 °C in a humidified incubator with 5% CO₂ and were passaged every 2–3 days to maintain logarithmic growth. Drugs were dissolved in DMSO to prepare high-concentration stock solutions, then diluted at least 1,000-fold in complete culture medium to the indicated final concentrations. Any drug combinations were prepared immediately before treatment. Cell lines were routinely monitored for morphology and viability and were periodically tested to confirm the absence of mycoplasma contamination. Experiments were performed using cells in the exponential growth phase and within a limited number of passages after thawing.

Generation and co-culture of bone marrow-derived dendritic cells (BMDCs)

Bone marrow-derived dendritic cells (BMDCs) were generated from 6–8-week-old BALB/c mice. Briefly, bone marrow cells were harvested from the femurs and tibiae by flushing the marrow cavities with sterile culture medium using a syringe and needle under aseptic conditions. Following red blood cell lysis, cells were collected by centrifugation and resuspended in complete RPMI-1640 medium supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. To induce dendritic cell differentiation, cultures were supplemented with recombinant mouse GM-CSF (20 ng/mL; MedChemExpress, HY-P7361) and IL-4 (10 ng/mL; MedChemExpress, CK15). Cells were maintained at 37 °C in a humidified incubator with 5% CO₂, and fresh cytokine-containing medium was added every 3 days. BMDCs were harvested and used for experiments after 8 days of differentiation.

For co-culture experiments, 4T1 cells were seeded in tissue culture plates and allowed to adhere overnight before drug treatment. After 12 h of treatment, the culture medium was replaced with fresh medium to remove residual compounds, and differentiated BMDCs were added to the cultures. Following 48 h of co-culture, BMDC maturation was evaluated by flow cytometry.

For immunophenotyping, cells were collected and stained with APC-conjugated anti-mouse CD11c (Proteintech, APC-65130), PE-conjugated anti-mouse CD86 (Proteintech, PE−65068), and FITC-conjugated anti-mouse CD80 (Proteintech, FITC-65076) antibodies for 30 min at 4 °C in the dark. A fixable viability dye (FVS700) was used to exclude dead cells. After staining, cells were washed twice with FACS buffer (PBS containing 2% FBS) and resuspended in the same buffer for acquisition on a CytoFLEX LX flow cytometer (Beckman Coulter). Data were analyzed using FlowJo v10.6.2. Dendritic cells were identified by sequential gating on live (FVS700⁻) singlet cells (FSC-A versus FSC-H) followed by selection of the CD11c⁺ population. The expression of the maturation markers CD80 and CD86 was then quantified within the CD11c⁺ BMDC population.

CRISPR knock out screen and data analysis

SH-SY5Y cells stably expressing Cas9 (SH-SY5Y-Cas9) were transduced with the Brunello human CRISPR knockout pooled library (Addgene #73,178) following the manufacturer’s recommended protocol. In brief, 2 × 10⁸ cells were infected at a multiplicity of infection (MOI) of approximately 0.3, corresponding to ~30% transduction efficiency, to maintain ~1,000-fold representation of each sgRNA throughout the screen. After infection, cells were selected with puromycin (5 μg/mL) for 7 days to eliminate non-transduced cells.

Following selection, cells were split into two conditions: a DMSO control group and an RSL3-treated group (1 μM), each containing 1 × 10⁸ cells. After 3 days of treatment, cells were harvested, counted, and 1 × 10⁸ cells from each condition were re-plated into five 15-cm dishes and cultured for an additional 3 days under the same respective conditions. This treatment and passaging cycle was repeated for a total of five rounds (approximately 15 population doublings under selection pressure) to enrich for sgRNAs conferring resistance to RSL3-induced ferroptosis.

Genomic DNA was extracted from final cell pellets using the Blood & Cell Culture DNA Midi Kit (Qiagen, #13,343). Integrated sgRNA sequences were amplified using a two-step PCR strategy (Takara, RR001B) with staggered primers according to the Brunello screening protocol. PCR products were pooled, purified, and sequenced on an Illumina NovaSeq 6000 platform (paired-end, 150 bp).

Raw sequencing reads were processed by adapter trimming and quality filtering, followed by alignment to the Brunello sgRNA library reference. sgRNA abundance and gene-level enrichment analyses were quantified using CRISPRCloud2 (https://crispr.nrihub.org/). Candidate hits were prioritized based on both effect size (log2 fold change) and ranking. Genes exhibiting a log2 fold change > 0.5 (positively selected) or <−0.5 (negatively selected), a rank within the top 20 enriched genes, and a P < 0.1 were considered candidate hits and selected for downstream validation and functional characterization.

Generation of stable cell lines

Lentiviral particles were generated in HEK293T cells by co-transfecting GSR knockout or doxycycline-inducible GSR expression constructs together with the packaging plasmids psPAX2 and pCMV-VSV-G using PEI 40 K (MKBio, MX2203). At 15 h post-transfection, the culture medium was replaced, and viral supernatants were harvested at 48 h and 72 h post-transfection. Collected supernatants were clarified by centrifugation (1,000 rpm, 3 min) and filtered through a 0.48 μm membrane to remove cellular debris.

Target cells were transduced with viral supernatants supplemented with 8 μg/mL polybrene (Beyotime, C0351). Following infection, cells were selected with puromycin (3 μg/mL; Beyotime, ST551) to establish stable cell lines. For inducible systems, GSR expression was activated by treatment with doxycycline (200 ng/mL; Beyotime, Y266886) prior to downstream experiments.

Generation of CRISPR knockout cell lines

CRISPR/Cas9-mediated gene knockout was achieved by cloning annealed oligonucleotides encoding target-specific sgRNA spacer sequences into the lentiCRISPRv2 vector (Addgene #52,961). sgRNA target sites were designed using the CHOPCHOP online platform (https://chopchop.cbu.uib.no/). The sequences of all sgRNAs used in this study are provided in Supplementary Table S1.

Lentiviral particles carrying sgRNA and Cas9 were generated in HEK293T cells and used to transduce recipient cells as described above. Successfully transduced cells were selected using puromycin to obtain stable knockout populations. Knockout efficiency was verified by Western blotting and/or genomic sequencing prior to functional analyses.

Cell viability assay

Cells were seeded in 96-well plates at a density of 2,000 cells per well and allowed to adhere overnight. The following day, cells were treated with the indicated compounds for 24–48 h. Resazurin was then added to a final concentration of 10 μg/mL, followed by incubation at 37 °C for 7 h to allow metabolic conversion. Fluorescence was measured using a BioTek Synergy HTX multimode microplate reader (BioTek Instruments, USA) with excitation at 570 nm and emission at 585 nm.

Cell viability was calculated after background subtraction and normalized to vehicle-treated control wells, which were defined as 100% viability. Data are presented as percentage of control unless otherwise indicated.

Cell death assay (propidium iodide staining)

Cells were seeded in 96-well plates at a density of 2,000 cells per well and allowed to adhere overnight. Cells were then treated with the indicated compounds for 24 h. Following treatment, cells were incubated with propidium iodide (PI; 1 μg/mL) and Hoechst 33,342 (10 μg/mL) for 30 min at 37 °C in the dark. After staining, cells were directly imaged without fixation using an inverted fluorescence microscope under identical exposure settings across all groups. At least five random fields per well were captured for analysis. Cell death was quantified as the percentage of PI-positive nuclei relative to total Hoechst-positive nuclei using ImageJ.

Colony formation assay

Cells were seeded in 6- or 12-well plates at a density of 1,000 cells per well and cultured overnight to allow attachment. The following day, cells were treated with the indicated compounds and maintained under continuous treatment for 12–16 days, with medium replaced every 3 days. At the end of the experiment, colonies were fixed with 4% paraformaldehyde for 15–20 min at room temperature, washed three times with PBS, and stained with 0.1% crystal violet solution for 10–15 min. Excess stain was removed by extensive washing with running water or PBS until the background was clear, followed by air drying. Colonies containing more than ~50 cells were counted. Quantification was performed using ImageJ after background subtraction and uniform thresholding across all images.

BODIPY-C11 staining (lipid peroxidation assay)

Cells were seeded in 6-well plates at a density of 2 × 10⁵ cells per well and treated with the indicated compounds the following day. After treatment, cells were harvested by trypsinization, collected by centrifugation, and washed three times with cold PBS. Cell pellets were resuspended in 500 μL PBS containing 2 μM BODIPY 581/591 C11 (Thermo Fisher Scientific, D3861) and incubated for 30 min at 37 °C in the dark with gentle mixing. Following incubation, cells were washed once with PBS and immediately analyzed by flow cytometry using a CytoFLEX LX (Beckman Coulter). Lipid peroxidation was quantified by measuring the shift in fluorescence in the FL1 channel (oxidized probe). Data were analyzed using FlowJo v10.6.2, and gating was performed on live singlet populations to exclude debris and aggregates.

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

Total RNA was extracted using RNAiso Plus (Takara, 9037) according to the manufacturer’s instructions. RNA concentration and purity were assessed by nanodrop. One microgram of total RNA was reverse-transcribed into cDNA using HiScript III RT SuperMix (Vazyme, R323-01). Quantitative real-time PCR was performed using 2× Taq Pro Universal SYBR qPCR Master Mix (Vazyme, Q712-02) on a CFX Connect Real-Time PCR System (Bio-Rad). Thermal cycling conditions followed the manufacturer’s recommended protocol. Relative gene expression was calculated using the 2^−ΔΔCt method, with ACTB serving as the internal reference gene. Primer sequences used in this study are listed in Supplementary Table S1.

Intracellular glutathione (GSH) measurement

Cells were harvested and lysed by ultrasonication in an ice-water bath (5 cycles of 5 s on/5 s off at 50% amplitude). Lysates were centrifuged, and supernatants were collected for analysis. Total protein concentration was determined using a BCA assay kit (Beyotime, P0011) for normalization. Intracellular reduced glutathione (GSH) levels were quantified using the Reduced Glutathione Content Assay Kit (Nanjing Jiancheng Bioengineering Institute, A006-2-1) following the manufacturer’s instructions. Absorbance was measured using a microplate reader, and GSH levels were normalized to total protein content and expressed as μmol/g protein.

GSR enzyme activity assay

Glutathione reductase (GSR) activity was measured using a GSR Assay Kit (Beyotime, S0055), based on the enzymatic reduction of oxidized glutathione (GSSG) coupled to NADPH oxidation. The reaction generates reduced glutathione, which reacts with DTNB to form TNB, allowing spectrophotometric detection. Absorbance was recorded at 412 nm, and enzyme activity was calculated according to the manufacturer’s instructions and normalized to protein concentration.

DLAT soluble/insoluble fractionation

Soluble and insoluble fractionation of DLAT was separated as previously described [33]. Briefly, cells were lysed on ice for 30 min in NP-40 lysis buffer (20 mM Tris–HCl pH 7.4, 150 mM NaCl, 1% NP-40, and protease inhibitor cocktail). Lysates were centrifuged at 14,000 rpm for 30 min at 4 °C. The supernatant (soluble fraction) was collected, while the pellet (insoluble fraction) was washed once with lysis buffer to remove residual soluble proteins. Both fractions were resuspended in SDS loading buffer, boiled, and analyzed by SDS-PAGE followed by Western blotting.

Western blotting

Cells were lysed in ice-cold RIPA buffer supplemented with protease inhibitor (Meilunbio, MB2678) and phosphatase inhibitor (Servicebio, G2007) cocktails. Lysates were incubated on ice and clarified by centrifugation to remove debris. Protein concentration was determined using a BCA assay kit (Beyotime, P0011) according to the manufacturer’s instructions. Equal amounts of protein were mixed with 1× SDS loading buffer and boiled prior to separation by SDS-PAGE.

Proteins were transferred onto PVDF membranes, which were subsequently blocked with 5% non-fat milk in TBST (Tris-buffered saline, ~pH 7.4, with 0.1% Tween-20) for 1 h at room temperature. Membranes were incubated overnight at 4 °C with primary antibodies at the indicated dilutions in TBST containing 3% bovine serum albumin (BSA). After washing with TBST, membranes were incubated with HRP-conjugated secondary antibodies for 1 h at room temperature. Immunoreactive bands were detected using enhanced chemiluminescence (ECL) reagents and visualized using a chemiluminescence imaging system.

The antibodies used in this study, including catalog numbers and dilution ratios, are listed in Supplementary Table S1.

Cell immunofluorescence

Cells were seeded onto sterile glass coverslips in 24-well plates and allowed to adhere overnight. Following the indicated treatments, cells were fixed with 4% paraformaldehyde for 10 min at room temperature, permeabilized with 0.25% Triton X-100 for 10 min, and blocked with 3% BSA for 1 h. Cells were then incubated with primary antibodies diluted in blocking buffer overnight at 4 °C. After washing three times with PBS containing 0.1% Tween-20 (PBS-T), samples were incubated with Alexa Fluor-conjugated secondary antibodies (1:1,000 dilution) for 1 h at 37 °C in the dark. Cells were washed again three times with PBS-T and counterstained with DAPI (0.1 μg/mL) for 10 min. Coverslips were mounted using anti-fade mounting medium (Vector Laboratories, H-1200).

Images were acquired using a Nikon Ti2-E inverted microscope equipped with an X-Light V3 spinning disk confocal system (CrestOptics), an LDI-7 laser system (89 North), and a Prime 95B sCMOS camera (Teledyne Photometrics). Imaging was performed using CFI Plan Apochromat Lambda 100× (NA 1.45) and 10× (NA 0.30) objectives. Acquisition parameters were optimized to avoid signal saturation and kept constant across all experimental groups. The antibodies used in this study, including catalog numbers and dilution ratios, are listed in Supplementary Table S1.

Tissue histology and immunostaining

Tumor tissues were fixed in 4% paraformaldehyde, dehydrated through graded ethanol series, embedded in paraffin, and sectioned at 5 μm thickness. For hematoxylin and eosin (H&E) staining, sections were deparaffinized, rehydrated, and processed using standard protocols.

For immunohistochemistry (IHC), antigen retrieval was performed by heating sections in 10 mM sodium citrate buffer (pH 6.0) in a microwave for 10 min. After cooling, sections were blocked with 3% BSA for 1 h at room temperature, incubated overnight at 4 °C with primary antibodies, and then washed and incubated with biotinylated secondary antibodies for 1 h at room temperature. Signal detection was performed using an avidin–biotin complex (ABC) system followed by DAB chromogen development.

For immunofluorescence staining of tissue sections, the same primary and secondary antibody incubation conditions were used as described for cell immunofluorescence. Nuclei were counterstained with DAPI where applicable. Whole-slide images were acquired using an Aperio VERSA 8 digital slide scanner (Leica Microsystems).

Quantitative analysis was performed using QuPath v0.5.0 software [43]. Positive staining was quantified by normalizing total fluorescence or chromogenic signal to the number of DAPI- or Hoechst 33,342-positive nuclei. H-scores were calculated using the formula: (0 × % negative) + (1 × % weak) + (2 × % moderate) + (3 × % strong staining intensity).

The antibodies used in this study, including catalog numbers and dilution ratios, are listed in Supplementary Table S1.

Cytokine release and HMGB1 measurement in cell lines

Cells were seeded in 6- or 12-well plates and cultured until reaching 70–80% confluence. Cells were then treated with the indicated compounds for 12 h, after which culture supernatants were collected and clarified by centrifugation to remove cellular debris. Levels of IFN-γ, TNF-α, IL-6, IFN-β, TNF-β, and extracellular HMGB1 were quantified in the conditioned media using commercial ELISA kits (Biorbyt, orb1806768, orb1806770, and orb1806771; Elabscience, EL-M0676) according to the manufacturers’ instructions. Extracellular ATP levels were measured using an ATP Determination Kit (Promega, GA5010) following the recommended protocol.

Immune profiling by flow cytometry

Fresh mouse tumors were excised and mechanically minced into 1–3 mm fragments under sterile conditions. Tumor tissues were enzymatically digested in Hank’s Balanced Salt Solution (HBSS) containing collagenase IV (1 mg/mL), collagenase I (0.5 mg/mL), and hyaluronidase (0.002 mg/mL) at 37 °C for 30 min with gentle agitation to generate single-cell suspensions. The digested tissue was filtered through a 70 μm cell strainer to remove undigested debris, followed by red blood cell lysis using RBC lysis buffer according to the manufacturer’s instructions. Cells were then washed with FACS buffer and resuspended for antibody staining.

For extracellular immune phenotyping, cells were incubated with fluorophore-conjugated antibodies against surface markers at 4 °C for 30 min in the dark. After staining, cells were washed twice with FACS buffer and prepared for flow cytometric analysis.

For intracellular cytokine staining, isolated cells were stimulated ex vivo with phorbol 12-myristate 13-acetate (PMA; 50 ng/mL), ionomycin (500 ng/mL), and brefeldin A (3.0 μg/mL) in RPMI-1640 medium supplemented with 10% FBS at 37 °C in a humidified 5% CO₂ incubator for 4 h to promote cytokine accumulation. Following stimulation, cells were washed and stained with surface antibodies (e.g., CD4, CD8) for 30 min at 4 °C in the dark. Cells were then fixed and permeabilized using an intracellular fixation and permeabilization buffer set according to the manufacturer’s instructions. Briefly, cells were incubated in fixation/permeabilization solution at 4 °C for 30 min, washed twice with permeabilization buffer, and subsequently stained with anti-IFN-γ and anti-Granzyme B antibodies (or corresponding isotype controls) for 30 min at 4 °C in the dark. After staining, cells were washed and resuspended in FACS buffer for acquisition.

Flow cytometric data were acquired using a CytoFLEX LX flow cytometer (Beckman Coulter) and analyzed with FlowJo v10.6.2 software. The antibodies used in this study, including catalog numbers and dilution ratios, are listed in Supplementary Table S1.

Serum factor analysis

Blood samples were collected by cardiac puncture at the time of sacrifice and allowed to clot at room temperature for 30–60 min. Serum was separated by centrifugation at 2,000 × g for 15 min at 4 °C, aliquoted, and stored at − 80 °C until analysis to avoid repeated freeze–thaw cycles. Serum levels of TNF-α, IL-6, albumin, alanine aminotransferase (ALT), and creatinine were quantified using commercially available ELISA kits (Elabscience, EL-M3063, EL-M0044, EL-M0032, BC-K235-M, BC-K188-M) according to the manufacturers’ instructions.

Mitochondrial and ROS fluorescent staining

Cells (1.5 × 10⁶) were seeded in 35-mm glass-bottom dishes and allowed to adhere overnight. Following the indicated treatments for 12 h, cells were incubated with MitoTracker Red (100 nM) for 30 min at 37 °C to visualize mitochondria, H2DCFDA (2 μM) for 30 min at 37 °C to assess total intracellular reactive oxygen species (ROS), or MitoSOX Red (1 μM) for 30 min at 37 °C to detect mitochondrial superoxide. Nuclei were counterstained with Hoechst 33,342 (10 μg/mL) for 10 min at room temperature. Fluorescence images were acquired using a laser scanning confocal microscope under identical acquisition settings across groups, as described in the “Cell immunofluorescence” section. Imaging parameters were optimized to avoid signal saturation. All experiments were performed in triplicate, and fluorescence intensity was quantified in at least 150 cells per condition using ImageJ software.

PicoGreen staining for mitochondrial DNA (mtDNA) visualization

Cells were first stained with MitoTracker (100 nM) for 30 min at 37 °C to label mitochondria, followed by two washes with PBS. Cells were then incubated with PicoGreen nucleic acid stain (1× working solution) for 30 min at room temperature in the dark. After three washes with PBS, fluorescence imaging was performed using a Nikon Ti2-E microscope equipped with an X-Light V3 spinning disk confocal system (CrestOptics), an LDI-7 laser system (89 North), and a Prime 95B sCMOS camera (Teledyne Photometrics), using a CFI Plan Apochromat Lambda 100× objective (NA 1.45).

Isolation and quantitation of cytosolic mtDNA and nuclear DNA

Cytosolic and nuclear fractions were prepared essentially as described previously [44], with minor modifications. Briefly, 6 × 10⁶ cells were washed twice with ice-cold PBS, pelleted (300 × g, 5 min, 4 °C), and gently resuspended in ice-cold hypotonic buffer (10 mM HEPES, 250 mM sucrose, 10 mM KCl, 1.5 mM MgCl₂, 1 mM EDTA, 1 mM EGTA, 0.02% digitonin). Cells were incubated on ice for 5 min with intermittent gentle mixing to selectively permeabilize the plasma membrane while preserving mitochondrial and nuclear integrity. Nuclei were pelleted by centrifugation at 1,000 × g for 10 min at 4 °C, and the resulting supernatant was collected as the cytosolic fraction. This fraction was further clarified by centrifugation at 10,000 × g for 30 min at 4 °C to remove residual organelles and debris. Nuclear pellets were retained for nuclear DNA extraction.

DNA from cytosolic and nuclear fractions was extracted using the same protocol described in the “RNA extraction and quantitative real-time PCR” section. Cytosolic mtDNA was quantified by qPCR using primers targeting ND1 and normalized to nuclear DNA markers (HBB or Tert). Primer sequences are listed in Supplementary Table S1.

Tumor vaccination and re challenge assay

To assess tumor antigen-specific immune memory, tumor cells were treated in vitro with the indicated compounds to induce immunogenic cell death (ICD). Following treatment, cells were harvested, washed twice with sterile PBS, and resuspended at the desired concentration in PBS. Mice were vaccinated subcutaneously with 2 × 10⁶ drug-treated dying 4T1 or 4T1-NT/STING-KO tumor cells in 100 μL PBS.

Ten days after vaccination, mice were challenged with 5 × 10⁵ viable tumor cells injected into the contralateral flank. Tumor growth was monitored every 2 days using digital calipers, and tumor volume was calculated using the formula: (length × width²)/2. All measurements were performed in a blinded manner where applicable.

Ex vivo CD8⁺ T-cell cytotoxicity assay

To evaluate treatment-induced antitumor T-cell activity, CD8⁺ T cells were isolated from tumors of mice receiving the indicated treatments. Tumors were excised, mechanically minced, and enzymatically digested in RPMI-1640 containing collagenase IV (1 mg/mL), collagenase I (0.5 mg/mL), and hyaluronidase (0.1 mg/mL) at 37 °C for 30 min with gentle agitation. Single-cell suspensions were filtered through a 70 μm cell strainer, washed with PBS, and subjected to magnetic bead-based CD8⁺ T-cell isolation according to the manufacturer’s instructions.

For cytotoxicity assays, firefly luciferase-expressing 4T1 (4T1-Luc2) target cells were seeded in white opaque 96-well plates and allowed to adhere overnight. Purified CD8⁺ T cells were added at the indicated effector-to-target (E) ratios (e.g., 5:1) and co-cultured for 24 h at 37 °C. Following incubation, D-luciferin substrate was added according to the manufacturer’s instructions, and luminescence was measured using a microplate luminometer. Bioluminescence intensity was used as a surrogate for viable tumor cell number. Percent specific lysis was calculated relative to target-only control wells, with decreased luminescence indicating increased CD8⁺ T-cell-mediated cytotoxicity.

Animal models

Female BALB/c or BALB/c nude mice (6–8 weeks old) were purchased from Beijing HFK Bioscience and maintained under specific pathogen-free (SPF) conditions with ad libitum access to food and water. Mice were housed in a controlled environment with a 12 h light/dark cycle and were allowed to acclimate before experiments. All in vivo procedures were approved by the institutional animal ethics committee and conducted in accordance with institutional guidelines (see Declarations). Mice were randomly assigned to experimental groups, with 6–10 mice per group, except for the tumor rechallenge assay, which included 12 mice per group. Treatment regimens, drug administration route, dosing schedules, and treatment intervals are provided in the corresponding Supplementary Figures. Small-molecule drugs were initially dissolved in DMSO and subsequently diluted immediately before administration with an in vivo vehicle composed of DMSO/PEG300/PBS/Tween-80 (5%/40%/50%/5%, v/v/v/v). Antibodies were diluted in the in vivo buffer according to the manufacturer’s instructions. Vehicle control mice received an equal volume of the solvent.

For subcutaneous syngeneic and xenograft models, murine tumor cells (CT26, 1.5 × 10⁵; 4T1, 1 × 10⁵) or human SH-SY5Y cells (1 × 10⁶) were injected subcutaneously into the right flank of BALB/c or BALB/c nude mice, respectively. Tumor growth was monitored every 2–3 days using digital calipers, and tumor volume was calculated using the formula: (length × width²) × 0.5.

For the orthotopic breast cancer model, 5 × 10⁵ 4T1 cells were injected into the fourth mammary fat pad. After 12 days of treatment, mice were euthanized, and tumors were excised, weighed, and processed for histological and immunohistochemical analyses.

For the pulmonary metastasis model, BALB/c mice were injected via the lateral tail vein with 1 × 10⁵ 4T1-Luc2 cells. Tumor colonization was assessed 7 days post-injection and subsequently monitored every 3 days by bioluminescence imaging (BLI) following intraperitoneal injection of D-luciferin (150 mg/kg). Imaging was performed using an IVIS Spectrum system. After 12 days of treatment, mice were euthanized, and lungs were harvested for metastatic nodule quantification and hematoxylin and eosin (H&E) staining.

Animals were monitored daily for body weight, general activity, and respiratory status throughout the study. Humane endpoints were applied in accordance with institutional guidelines, and mice were euthanized when tumor volume reached 1,500 mm³ or when predefined ethical criteria were met.

RNA sequencing and bioinformatics analysis

Total RNA was extracted from cells using RNAiso Plus (Takara, 9037) according to the manufacturer’s instructions. RNA integrity was assessed using an Agilent 2100 Bioanalyzer, and only samples with RNA integrity number (RIN) ≥ 9.5 were used for library preparation. Poly(A)-selected mRNA libraries were constructed following standard protocols and sequenced on an Illumina NovaSeq 6000 platform to generate 150 bp paired-end reads.

Raw sequencing reads were quality-checked and aligned to the reference genome using STAR aligner, and gene-level read counts were quantified using featureCounts. Differential gene expression analysis was performed in R using the DESeq2 package. Genes with an adjusted p-value < 0.05 and absolute log2 fold change > 1 were defined as significantly differentially expressed genes (DEGs).

Functional enrichment analysis of Gene Ontology (GO) Biological Process terms was performed using DAVID Bioinformatics Resources. Pathways with false discovery rate (FDR) ≤ 0.05 and p ≤ 0.01 were considered significantly enriched. Data visualization, including heatmaps and volcano plots, was generated using ggplot2 and pheatmap in R (v4.5.0).

Bioinformatics analysis of cancer cohorts and immunotherapy clinical studies

GSR expression levels and associated clinical and immunotherapy-related features across multiple cancer cohorts were retrieved from the TIDE platform [45], a web-based resource for predicting immunotherapy response. Ferroptosis-related pathways and immune-associated hallmark gene sets were obtained using the msigdbr R package.

Single-sample gene set enrichment analysis (ssGSEA) was performed to calculate pathway activity scores at the individual sample level. Immune cell composition within tumor tissues was estimated from bulk transcriptomic data using CIBERSORT [46]. Correlations between glutathione pathway activity, immune-related signatures, and PD-1/PD-L1 expression were assessed based on ssGSEA enrichment scores and CIBERSORT-derived immune infiltration scores and visualized using a triangular correlation matrix.

Patients with pre-treatment tumor samples collected prior to anti-PD-1 immunotherapy were included in the analysis. Clinical responses were classified according to RECIST (Response Evaluation Criteria in Solid Tumors). Patients who achieved a complete response (CR) or partial response (PR) were categorized as responders, whereas those with stable disease (SD) or progressive disease (PD) were categorized as non-responders.

The tumor immune microenvironment was further analyzed using the tigeR R package (v1.0) [47] to evaluate associations with immunotherapy response and survival outcomes. Hazard ratios (HRs) were estimated using univariate and multivariate Cox proportional hazards regression models.

Quantification and statistical analysis

Data are presented as mean ± standard error of the mean (SEM). Statistical analyses were performed as specified in the corresponding figure legends using GraphPad Prism 10.1.2 (GraphPad Software). Differences between two independent groups were evaluated using a two-tailed unpaired Student’s t-test. For comparisons involving more than two groups, one-way analysis of variance (ANOVA) followed by Tukey’s post hoc multiple comparisons test was applied. A P-value < 0.05 was considered statistically significant. Sample sizes were determined based on previous studies with similar experimental designs, and no data were excluded from the analyses.

Statistical significance is indicated as follows: ns, P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; #, P < 0.05; ##, P < 0.01; ###, P < 0.001. Symbols above bars denote comparisons with the corresponding vehicle-treated or GSR non-targeting (GSR-NT) control groups, as specified in each figure legend.

Results

GSR determines sensitivity to ferroptosis

To identify genetic determinants of ferroptosis sensitivity, we performed a genome-wide CRISPR/Cas9 screen in SH-SY5Y neuroblastoma cells using a proliferation-based readout in the presence or absence of the GPX4 inhibitor RSL3 (Fig. 1A). Among the top-ranked genes, TP53 emerged as the top suppressor of cell proliferation under RSL3 treatment (Fig. 1B), supporting the validity of the screen. Notably, the top scoring gene essential for cell proliferation under RSL3 treatment encodes GSR, which catalyzes the reduction of oxidized glutathione disulfide (GSSG) to its reduced form, glutathione (GSH) (Fig. 1B).

Fig. 1.

GSR deficiency sensitizes cancer cells to ferroptosis. (A) Schematic of a genome-wide CRISPR/Cas9 screen in SH-SY5Y cells treated with ferroptosis inducer RSL3 to identify ferroptosis sensitizers. (B) Volcano plot of gene enrichment under RSL3 treatment, highlighting the top ranked anti-ferroptotic (red) and pro-ferroptotic (blue) genes. (C, D) Treatment with RSL3 for 48 h significantly reduced cell viability and colony formation in GSR-KO cells, compared to non-targeting (NT) controls. (E) Enhanced cell viability in doxycycline-induced GSR-overexpressing cells treated with RSL3 for 48 h. (F) No difference in cell viability was observed between GSR-KO and control cells in the absence of RSL3 treatment. (G) Cellular GSH levels are reduced in GSR-KO cells. (H) Cell viability of GSR-NT, GSR-KO, and GSR-KO cells reconstituted with either wild-type human GSR or a catalytically inactive GSR mutant (mtGSR) after 24 h of RSL3 treatment. (I) Increased lipid peroxidation in GSR-KO cells treated with 2 μM RSL3 for 4 h, measured by C11-BODIPY 581/591 staining via flow cytometry, with quantification on the right. (J) Elevated PTGS2 and CHAC1 mRNA expression in GSR-KO cells treated with 2 μM RSL3 for 4 h, assessed by qPCR. (K, L) Reduced cell viability in GSR-KO cells treated with 2 μM RSL3 for 24 h, compared to NT controls, showing no rescue by apoptosis inhibitor Z-VAD-FMK (10 μM) or necroptosis inhibitor necrostatin-1 (Nec-1, 2 μM), but full rescue by ferroptosis inhibitors ferrostatin-1 (Fer-1, 10 μM) and liproxstatin-1 (Lip-1, 10 μM) or thiol-containing antioxidants GSH (1 mM) and NAC (2 mM). (M) Across 860 CTRP cancer cell lines, higher GSR expression was strongly associated with increased resistance (AUC of cell viability) to ferroptosis inducers (RSL3, ML162, ML210, erastin). (N) Kaplan–Meier survival curves from a neuroblastoma cohort, stratified by high vs. low tumor GSR expression levels. (O) GSR enzymatic activity is decreased upon treatment with GSR inhibitors BCNU (SH-SY5Y: 30 or 60 μM for 12 h, A549: 60 or 100 μM for 12 h). (P) Compared to RSL3 alone, combined RSL3 and BCNU treatment significantly reduced cell viability (treatment for 48 h). (Q, R) increased lipid peroxidation and upregulated PTGS2 and CHAC1 expression (2 μM RSL3 and 20 μM BCNU for 24 h). (S) The cytotoxicity induced by RSL3 plus BCNU is not rescued by 10 μM Z-VAD-FMK or 2 μM Nec-1, but is fully rescued by 1 mM GSH or 3 mM NAC. Statistics were analyzed by unpaired two-tailed Student’s t-test (E–G) or one-way ANOVA with Tukey’s correction (C, D, H–L, O-S). Asterisks indicate comparison between NT and KO#1, and hash symbols denote comparison between NT and KO#2 (C)

To confirm GSR’s role in ferroptosis, we tested a wide range of cancer cell lines, and found that knocking out GSR significantly enhanced sensitivity to RSL3-induced cell death, as measured by cell viability, cell death, and colony formation assays (Fig. 1C, 1D, S1A-D). Conversely, overexpressing GSR conferred resistance to ferroptosis (Fig. 1E). Importantly, GSR knockout (KO) alone did not affect proliferation in the absence of ferroptotic stress (Fig. 1F). This sensitization extended to additional ferroptosis inducers, including GPX4 inhibitors ML162 and ML210, and the system Xc- inhibitor erastin (Fig. S1B). Efficient knockout and overexpression of GSR were confirmed by western blot (Fig. S1E, S1F). Consistent with its established enzymatic function, GSR-KO reduced intracellular GSH levels (Fig. 1G).

To determine whether GSR catalytic activity is required for ferroptosis resistance, GSR-KO 4T1 murine breast cancer cells were reconstituted with either wild-type human GSR or a catalytically inactive GSR mutant harboring C102,107A substitutions (mtGSR). Only wild-type GSR, but not the catalytically inactive mutant, fully restored ferroptosis resistance (Fig. 1H), demonstrating that GSR enzymatic activity is required and functionally conserved across species.

RSL3-treated GSR-KO cells displayed hallmark ferroptotic features, including elevated lipid peroxidation detected by BODIPY-C11 oxidation (Fig. 1I), and upregulated expression of ferroptotic maker genes including PTGS2 and CHAC1 (Fig. 1J). This increased cell death was not inhibited by the apoptosis inhibitor Z-VAD-FMK (Z-VAD), nor the necroptosis inhibitor necrostatin-1 (Nec-1), but fully rescued by ferroptosis inhibitors ferrostatin-1 (Fer-1) and liproxstatin-1 (Lip-1), as well as thiol-containing antioxidants GSH or N-acetylcysteine (NAC) (Fig. 1K, 1L, S1G, S1H).

To determine whether the relationship between GSR expression and ferroptosis sensitivity extends across cancer types, we analyzed drug response data from the Cancer Therapeutics Response Portal (CTRP), encompassing 860 cancer cell lines and 481 compounds. GSR expression positively correlates with resistance to multiple ferroptosis inducers (Fig. 1M), with the strongest association observed for GPX4 inhibitors (RSL3, ML162, and ML210) (Fig. S1I). This relationship was independently validated in a panel of 11 cancer cell lines, where intrinsic resistance to RSL3 positively correlated with GSR expression levels (Fig. S1J).

Clinically, pan-cancer analyses of The Cancer Genome Atlas (TCGA) datasets revealed elevated GSR expression is associated with higher risk scores in several tumor types (Fig. S2A), and worse disease-free survival probability in the neuroblastoma, melanoma, glioma, liver, lung and breast lumA cancer (Fig. 1N, S2B). Furthermore, univariate and multivariate Cox regression analyses across four independent cohorts confirmed GSR as an independent prognostic factor after adjusting for clinical variables (Fig. S2C-F).

Pharmacologically, the FDA-approved chemotherapeutic agent carmustine (brand name BCNU), a potent GSR inhibitor [48] (Fig. 1O), recapitulated the effects of GSR genetic knockout by markedly sensitizing cancer cells to RSL3-induced ferroptotic cell death (Fig. 1P, S3A-C). Similar sensitization was observed with other GSH-disrupting agents, including the dual GSR/ thioredoxin reductase inhibitors LCS3 and 2-AAPA (Fig. S3D-F) [49], as well as BSO, an irreversible inhibitor of GCL (Fig. S3G, S3H). In contrast, inhibition of broader antioxidant pathways, including NRF2 inhibition by ML385 and thioredoxin reductase inhibition by auranofin, had minimal effect (Fig. S3I). Similar to GSR genetic knockout, BCNU treatment also increased RSL3-induced lipid peroxidation (Fig. 1Q, S3J), and expression of ferroptosis marker genes (Fig. 1R, S3K). The synergistic cell death induced by BCNU and RSL3 was not rescued by the apoptosis inhibitor Z-VAD or the necroptosis inhibitor Nec-1, but could be effectively rescued by GSH or NAC (Fig. 1S, S3L).

Collectively, these findings identify GSR as an important determinant of ferroptosis sensitivity across multiple cancer models. GSR deficiency sensitizes cancer cells to ferroptosis, consistent with its role in maintaining glutathione homeostasis and thereby limiting ferroptotic cell death.

GSR deficiency sensitizes cancer cells to cuproptosis

Cuproptosis is a copper-dependent form of RCD triggered by excessive intracellular copper, which directly binds to and induces aggregation of lipoylated proteins in the TCA cycle, thereby disrupting mitochondrial respiration, promoting proteotoxic stress, and ultimately leading to cell death [33]. Because GSR deficiency reduces cellular GSH levels (Fig. 1G), an endogenous copper chelator and antioxidant [50], and strongly sensitizes cells to ferroptosis (Fig. 1, S1, S3), we investigated whether GSR deficiency might similarly sensitize cancer cells to cuproptosis.

To induce cuproptosis, we treated cancer cells with copper ionophore elesclomol (Es) in combination with Cu²⁺ (Es-Cu) (Fig. S4A, S4B). Both GSR genetic knockout and pharmacological inhibition with BCNU, LCS3, or 2-AAPA, synergistically enhanced the cytotoxic effects of Es-Cu, leading to reduced cell viability and colony formation as well as increased cell death (Fig. 2A, 2B, S4C-F), resembling the effects observed during RSL3-induced ferroptosis (Fig. S4G, 4H). Re-expression of human wild-type GSR, but not a catalytically inactive mutant, restored cell viability in Es-Cu-treated GSR-KO cells (Fig. 2C), confirming that GSR catalytic activity is required for cuproptosis resistance. In contrast, inhibition of broader antioxidant pathways, including NRF2 inhibition by ML385 and TrxR inhibition by auranofin, did not sensitize cells to Es-Cu (Fig. S4I).

Fig. 2.

GSR deficiency sensitizes cancer cells to cuproptosis. (A) GSR-KO reduces cell viability compared to NT control cells upon treatment with cuproptosis inducer elesclomol plus copper (Es-Cu) for 24 h. (B) Combined treatment of Es-Cu and BCNU decreased cell viability compared to individual treatment for 24 h. (C) Cell viability of GSR-NT, GSR-KO, and GSR-KO cells reconstituted with either wild-type human GSR or a catalytically inactive GSR mutant (mtGSR) after 24 h of Es-Cu treatment. (D) GSR-KO does not alter cell viability in response to the apoptosis-inducing agent paclitaxel for 24 h. (E) Nuclear and propidium iodide (PI) staining of cells treated with 5 μM RSL3, Es-Cu (100 nM each), or 200 nM paclitaxel for 12 h in NT control or GSR-KO backgrounds, showing cell death without nuclear fragmentation for RSL3 and Es-Cu, unlike paclitaxel. Scale bar: 100 μm. White arrowheads indicate apoptotic nuclear damage. (F, G) Increased aggregation of lipoylated TCA cycle proteins DLAT in GSR-KO cells compared to WT upon Es-Cu (100 nM each) treatment for 12 h, shown by immunofluorescence and western blotting. Scale bar: 10 μm. (H, I) Increased aggregation of DLAT with Es-Cu (100 nM each) + BCNU (100 μM) co-treatment compared to individual treatment alone for 12 h), shown by immunofluorescence and western blotting. Scale bar: 10 μm. (J, K) Genetic or pharmacological inhibition of GSR reduces cell viability upon Es-Cu treatment for 24 h; this effect is rescued by the copper chelator 10 μM TTM, 1 mM GSH, or 3 mM NAC, but not by 10 μM Z-VAD-FMK or 2 μM Nec-1. Statistics were analyzed by unpaired two-tailed Student’s t-test (D) or one-way ANOVA with Tukey’s correction (A-C, E, J, K). In (A), asterisks indicate comparison between NT and KO#1, whereas hash symbols denote comparison between NT and KO#2. In (J) and (K), asterisks and ns (not significant) without connecting lines indicate comparisons versus the DMSO control, whereas hash symbols denote comparison versus Es-Cu alone

Importantly, GSR deficiency did not affect apoptosis. GSR-KO cells showed unchanged viability, cell death, and nuclear morphology in response to the apoptosis inducer paclitaxel (Fig. 2D, 2E, S4J). Consistent with mechanistic differences, both RSL3 and Es-Cu induced membrane damage (assessed by propidium iodide staining) in GSR-KO cells, whereas paclitaxel induced nuclear fragmentation independently of GSR status (Fig. 2E, S4J).

Mechanistically, GSR inhibition markedly promotes aggregation of the lipoylated protein DLAT, a hallmark of cuproptosis, upon Es-Cu treatment (Fig. 2F-I, S5A-D), but not under RSL3 treatment (Fig. 2F, 2G). This DLAT aggregation was fully blocked by supplementation with the antioxidant GSH (Fig. 2F-I, S5A-D). Furthermore, the enhanced cell death induced by GSR deficiency and Es-Cu treatment was fully rescued by the copper chelator tetrathiomolybdate (TTM), GSH, or NAC, but not by inhibitors of apoptosis (Z-VAD), necroptosis (Nec-1), or ferroptosis (Fer-1) (Fig. 2J, 2K, S5E-H). Conversely, inhibition of cuproptosis by TTM failed to protect against RSL3-induced ferroptosis (Fig. S5I), indicating that the two pathways remain mechanistically distinct despite their shared dependence on GSR.

Collectively, these findings demonstrate that GSR deficiency specifically enhances copper-dependent cuproptosis through impaired GSH-mediated copper buffering and redox homeostasis, and identify GSR as a convergent regulator bridging both ferroptosis and cuproptosis.

GSR deficiency enhances ferroptosis- and cuproptosis-mediated tumor suppression in vivo

To test the therapeutic relevance of GSR deficiency in vivo, we subcutaneously implanted nude mice with either GSR-KO human SH-SY5Y cells or isogenic WT controls carrying a non-targeting (NT) plasmid (Fig. S6A). GSR-KO xenografts exhibited significant reductions of tumor growth and prolonged survival following treatment with ferroptosis and cuproptosis inducers (Fig. 3A, 3B, S6B), without affecting overall health or body weight (Fig. S6C). Notably, Es was always administered alone in vivo, leveraging endogenous Cu²⁺ present in the body fluid. Immunohistochemical analysis revealed reduced Ki67 staining, indicative of decreased cell proliferation, alongside elevated levels of extracellular HMGB1, surface-exposed calreticulin (CRT), and the lipid peroxidation product 4-hydroxynonenal (4-HNE) in GSR-deficient tumors following treatment (Fig. 3C).

Fig. 3.

GSR deficiency enhances antitumor efficacy in a GSR-dependent manner. (A, B) Reduced tumor volume (A) and prolonged survival (B) in immunocompromised nude mice bearing GSR-KO versus NT control human SH-SY5Y tumor cells treated with RSL3 or Es. (C) Representative images and quantification of immunohistochemistry for Ki67 and HMGB1, and immunofluorescence for calreticulin (CRT) and 4-HNE in tumor tissues. Scale bars: 200 μm. (D, E) Tumor growth curves and final tumor weights in syngeneic BALB/c mice subcutaneous implanted with 4T1 cells: GSR-NT, GSR-KO, and GSR-KO cells reconstituted with either wild-type human GSR or a catalytically inactive GSR mutant (mtGSR), following Es treatment. Statistics were analyzed by one-way ANOVA with Tukey’s correction (A, C, E), or log-rank test (B). Asterisks indicate comparisons versus the DMSO or PBS control of GSR-NT cells (C, E)

To confirm the requirement for GSR catalytic activity, we used a syngeneic 4T1 tumor model in which GSR-KO cells were reconstituted with wild-type human GSR or a catalytically inactive mutant (Fig. S6D). GSR-deficient tumors exhibited enhanced sensitivity to Es treatment, resulting in significantly reduced tumor growth (Fig. 3D, 3E). Re-expression of wild-type GSR largely restored tumor growth under Es treatment, whereas the inactive mutant failed to do so (Fig. 3D, 3E). No significant differences in body weight were observed among groups (Fig. S6E).

Given that BCNU is a clinically approved chemotherapeutic with potent GSR inhibitory activity, we next exploited its therapeutic potential in combination with ferroptosis- or cuproptosis-inducing agents. RNA-seq analysis on BCNU-treated cells (Fig. S7A) revealed upregulation of ferroptosis-associated pathways, including iron and porphyrin metabolism and lipid transport and metabolism, alongside induction of glutathione biosynthesis genes, likely reflecting a compensatory response to BCNU-induced GSH depletion (Fig. S7C). In contrast, BCNU suppressed gene sets linked to cell proliferation, notably Wnt signaling, cell adhesion, and extracellular matrix (ECM) regulation (Fig. S7D), indicating broad inhibition of tumor growth programs. Expression changes of four representative genes induced by BCNU were confirmed by qRT-PCR (Fig. S7B).

Building on these mechanistic insights, we tested whether BCNU combination therapy in an orthotopic breast cancer model, established by injecting 4T1 murine mammary carcinoma cells into the fourth mammary fat pad of syngeneic immunocompetent female BALB/c mice. BCNU combined with the ferroptosis inducer RSL3 or the cuproptosis inducer Es significantly reduced both tumor volume and final tumor weight compared to either monotherapy (Fig. 4A-C, S8A, S8B).

Fig. 4.

GSR deficiency-induced tumor suppression occurs via ferroptosis and cuproptosis. (A-C) Reduced tumor volume and tumor weight in syngeneic BALB/c mice orthotopically implanted with 4T1 cells. (D-F) Reduced tumor burden in the lung metastasis model, as assayed by in vivo bioluminescence (D), lung weight (E) and number of metastasis nodules (F), in syngeneic mice intravenously inoculated with 4T1-Luc2 cancer cells. (G, H) Representative images of metastatic lungs (G) and corresponding H&E-stained lung sections (H) under the indicated treatment. Blue arrows (top) and black arrows (bottom) indicate cancer nodules. Scale bar: 2 mm. (I-K) Tumor growth curves and final tumor weights of 4T1 tumors treated with BCNU combined with RSL3 or Es, with or without ferroptosis inhibitors (Lip-1, RSG) or cuproptosis inhibitor TTM. All statistics were analyzed by one-way ANOVA with Tukey’s correction. Asterisks indicate comparisons versus the DMSO control (C, E, F, K)

We next investigated whether this effect extends to metastatic cancer using an experimental lung metastasis model generated by tail vein injection of 4T1-Luc2 cells into BALB/c mice (Fig. S8C). BCNU co-treatment with RSL3 or Es significantly outperformed monotherapy, as evidenced by reduced tumor size measured by bioluminescent imaging (Fig. 4D, S8D). This was accompanied by decreased lung weight, indicative of reduced pulmonary edema (Fig. 4E), and a marked reduction in metastatic lung nodules (Fig. 4F, 4G, S8E), and restored normal lung histology (Fig. 4H), with no impact on normal body weight (Fig. S8F).

To verify that the antitumor effects of the combination therapy were mediated by ferroptosis or cuproptosis, we used pathway-specific inhibitors (Fig. S9A). Ferroptosis blockage with Liproxstatin-1 (Lip-1) or rosiglitazone (RSG, an ACSL4 inhibitor) attenuated efficacy of BCNU + RSL3 (Fig. 4I-K), while the copper chelator TTM blocked the efficacy of BCNU + Es (Fig. 4I-K), without affecting body weight (Fig. S9B).

To further validate GSR as the relevant target in vivo, we tested two additional GSR inhibitors, 2-AAPA and LCS3 (Fig. S9C). Both compounds, when combined with either RSL3 or Es, suppressed tumor growth more effectively than monotherapy, although the antitumor effects were generally less pronounced than those observed with BCNU (Fig. S9D-K).

To determine whether the synergistic activity of BCNU was mainly attributable to GSR inhibition, we treated GSR-KO and NT control cells under conditions in which each single treatment induced only minimal cytotoxicity. In NT control cells, BCNU significantly enhanced the cytotoxic effects of both RSL3 and Es-Cu (Fig. S9L). In contrast, this synergistic effect was abolished in GSR-KO cells (Fig. S9L), which were already highly sensitive to RSL3 or Es-Cu alone. Consistently, treatment of GSR-KO cells with RSL3 alone resulted in cell viability comparable to that of NT control cells treated with the BCNU plus RSL3 combination. A similar pattern was observed for Es-Cu treatment. Together, these findings suggest that the synergistic effects of BCNU with RSL3 or Es-Cu are largely attributable to GSR inhibition. Nevertheless, we cannot completely exclude contributions from other cytotoxic mechanisms of BCNU, such as DNA alkylation, despite using treatment conditions designed to minimize these effects.

Collectively, these findings demonstrate that genetic or pharmacological disruption of GSR enhances the antitumor activity of ferroptosis- and cuproptosis-inducing treatments across multiple in vivo tumor models. The ability of multiple GSR-targeting approaches to produce similar phenotypes supports a role for GSR activity in limiting the therapeutic efficacy of these regulated cell death pathways.

GSR deficiency enhances immunogenic cell death in ferroptosis and cuproptosis

Our in vivo experiments demonstrated that GSR deficiency not only significantly enhances ferroptotic and cuproptotic cancer cell death, but also increased key immunogenic cell death markers, including extracellular HMGB1 and surface-exposed CRT (Fig. 3C). To further validate this observation across cellular models, we first quantified three canonical DAMP markers, including extracellular ATP, extracellular HMGB1, and surface-exposed CRT, in multiple cancer cell lines. Either genetic GSR-KO or pharmacological inhibition with BCNU, when combined with RSL3 or Es-Cu, significantly elevated extracellular ATP and CRT exposure, while decreasing intracellular HMGB1 through its release into the extracellular space, thereby confirming enhanced DAMP mobilization (Fig. 5A, 5B, S10A-C).

Fig. 5.

GSR deficiency enhances immune activation in ferroptosis and cuproptosis. (A) Representative images and quantification of HMGB1 and CRT staining in 4T1 cells treated with RSL3 or Es-Cu, alone or combined with BCNU. Scale bars: 200 μm. (B) Genetic (left) or pharmacological (right) GSR inhibition, combined with ferroptosis or cuproptosis inducers, significantly elevated extracellular ATP released from MCF7 cells. (C) Levels of IFN-γ, HMGB1, TNF-α, and IL-6 inco-culture medium of 4T1 cancer cells and bone marrow-derived dendritic cells (BMDCs), assessed by ELISA. (D) Increased activated dendritic cells (CD86⁺) in the co-cultures treated with RSL3 or Es-Cu in combination with BCNU, measured by flow cytometry. (E–G) In the co-culture treated with RSL3 or Es-Cu plus BCNU, HMGB1 neutralization or TLR4 inhibition with TAK-242 (0.2 µM) markedly reduced production of cytokine TNF-α and IL-12 (E), decreased DC activation (F), and attenuated phosphorylation of NF-κB p65 and IRF3 (G). All statistics were analyzed by one-way ANOVA with Tukey’s correction. Asterisks and ns (not significant) without connecting lines indicate comparisons versus the DMSO control

Consistent with this ICD-associated phenotype, media from tumor cells treated with BCNU plus either RSL3 or Es-Cu contained elevated levels of released immunogenic cytokines and mediators, including IFN-γ, HMGB1, TNF-α, and IL-6 (Fig. 5C). In parallel, expression of pro-inflammatory genes, including IFN-α, IFN-β, IL-6, ISG56/IFIT1, and MCP1/CCL2, was robustly induced (Fig. S10D-F), suggesting activation of innate immune signaling.

To assess whether these molecular changes translated into functional immune activation, we established an in vitro DC co-culture system. Mouse bone marrow-derived dendritic cells (BMDCs) were co-cultured with 4T1 cells treated with BCNU in combination with either RSL3 or Es-Cu. Compared with either treatment alone, the combination treatment significantly increased the proportion of CD86⁺ CD80⁺ mature DCs (Fig. 5D), indicating enhanced DC maturation and activation. Consistently, TNF-α and IL-12 levels were also increased in the co-culture media (Fig. 5E).

Because HMGB1 is a major DAMP that signals through the TLR4-MyD88/TRIF axis, we next examined whether this pathway mediates the immune effects of GSR inhibition. In co-cultures treated with RSL3 or Es-Cu plus BCNU, HMGB1 neutralization with an anti-HMGB1 antibody or TLR4 inhibition with TAK-242 markedly reduced the production of immunogenic cytokines, including IFN-β, TNF-α, and IL-6 (Fig. 5E, S11A), decreased DC activation (Fig. 5F, S11B), and attenuated NF-κB p65 and IRF3 phosphorylation (Fig. 5G, S11C), indicating that HMGB1-TLR4 signaling is required for the enhanced innate immune activation induced by GSR inhibition during ferroptosis and cuproptosis. Nevertheless, immune activation was not completely abolished, remaining above basal control levels. These findings suggest that HMGB1 released from cancer cells undergoing GSR inhibition-enhanced ferroptosis or cuproptosis plays a central role in promoting immune activation, while additional DAMPs or signaling pathways are likely to contribute to the overall immune response.

GSR loss cooperates with ferroptosis and cuproptosis to reshapes the tumor immune microenvironment (TME)

Translating these findings to an immunocompetent setting, we employed a subcutaneous CT26 colon carcinoma model in syngeneic BALB/c mice (Fig. S12A, S12B). While RSL3 or Es monotherapy induced modest tumor growth delay, BCNU combination therapy markedly suppressed tumor volume and weight (Fig. 6A-D, S12C, S12D), without affecting overall health or body weight (Fig. S12E, S12F).

Fig. 6.

GSR deficiency enhances immune infiltration in ferroptosis and cuproptosis. (A-D) Reduced tumor volume in immunocompetent mice bearing CT26 cancer cells treated with RSL3 (A, B) or Es (C, D) combined with BCNU. (E, F) Representative immune histochemistry images for CD4, CD8, CD11c, and HMGB1, and immunofluorescence images for CRT in tumors treated with RSL3 (E) or Es (F) in combination with BCNU. Quantification of positive staining areas is shown in the bottom histograms. Scale bars: 50 μm. (G) Flow cytometric profiling of 4T1 tumors subjected to the same treatments as (A, C). (H) In vitro killing assays showed that CD8⁺ T cells isolated from tumors or spleens of mice treated with BCNU together with RSL3 or Es-Cu exhibited enhanced cytotoxic activity against luciferase-expressing 4T1 tumor cells in co-culture. (I-K) Tumor growth curves and final tumor weights of 4T1 tumors treated with BCNU in combination with Es, with or without depletion of CD8⁺, CD11c⁺, or CD4⁺ T cells. All statistics were analyzed by one-way ANOVA with Tukey’s correction. Asterisks and ns (not significant) without connecting lines indicate comparisons versus the DMSO or PBS control (E–H, K)

Histological analyses confirmed pathway-specific cell death. 4-HNE staining (ferroptosis marker) was intensified in tumors treated with RSL3 + BCNU (Fig. S12G), while DLAT aggregation (cuproptosis hallmark) was amplified in tumors treated with Es + BCNU (Fig. S12H). Proliferation marker Ki67 was significantly reduced in both regimens (Fig. S12G, S12H). Most importantly, combination therapy drove robust immune infiltration of CD4⁺ and CD8⁺ T cells, CD11c⁺ DCs, and CD68⁺ macrophages (Fig. 6E, 6F, S12G, S12H), together with elevated CRT (both conditions) and HMGB1 (particularly cuproptosis) (Fig. 6F, S12G).

To further characterize the tumor immune microenvironment beyond histological observations, we performed flow cytometric profiling (Fig. 6G, S13A). Combination treatment significantly increased the frequencies of total T cells (CD3e⁺CD45⁺) and cytotoxic T cells (CD8a⁺CD3e⁺), indicating enhanced antitumor T-cell infiltration. In contrast, immunosuppressive myeloid populations, including MDSCs (Gr-1⁺CD11b⁺) and M2-like macrophages (CD206⁺CD11b⁺), were reduced, accompanied by a decrease in the overall CD11b⁺ myeloid compartment. Regulatory T cells (CD25⁺FoxP3⁺CD3e⁺) were also significantly diminished after treatment. Notably, the proportion of exhausted CD8⁺ T cells expressing PD-1 and TIM-3 was reduced, suggesting partial restoration of effector T-cell functionality. Collectively, these data demonstrate that BCNU treatment, when combined with RSL3 or Es-Cu, reshapes the tumor microenvironment from an immunosuppressive state toward a more immune-active phenotype.

To determine whether increased CD8⁺ T cell infiltration translated into enhanced cytotoxic function, we isolated CD8⁺ T cells from tumors or spleens of treated mice and co-cultured them with luciferase-expressing 4T1 tumor cells. CD8⁺ T cells derived from tumors treated with BCNU in combination with either RSL3 or Es exhibited markedly stronger cytotoxic activity, inducing 4–5 folds greater tumor cell killing than RSL3 or Es monotherapy, and approximately twofold greater killing than BCNU alone (Fig. 6H). This effect was more pronounced in tumor-infiltrating T cells than in splenic T cells, indicating that the local tumor microenvironment contributes to the observed functional response.

Finally, to establish which immune subsets are required for the therapeutic benefit, we performed in vivo immune cell depletion using neutralizing antibodies (Fig. S13B). Depletion efficiency was confirmed by flow cytometry (Fig. S13C). CD8⁺ T cell depletion almost completely abrogated the antitumor efficacy of Es + BCNU treatment, restoring tumor growth to nearly control levels without affecting body weight (Fig. 6I-K, S13D). CD11c⁺ cell depletion also substantially impaired tumor control, whereas CD4⁺ T cell depletion had a more modest effect (Fig. 6I-K). These results demonstrate that the antitumor activity of the combination therapy is primarily dependent on CD8⁺ T cells, with significant contribution from CD11c⁺ antigen-presenting cells and a partial role for CD4⁺ T cells.

Collectively, these findings demonstrate that GSR inhibition enhances the antitumor activity of ferroptosis- and cuproptosis-inducing therapies in immunocompetent hosts while promoting extensive remodeling of the tumor immune microenvironment. This remodeling is characterized by increased infiltration and activation of cytotoxic immune cells, reduced immunosuppressive populations, and a prominent dependence on CD8⁺ T cells and CD11c⁺ antigen-presenting cells for therapeutic efficacy.

GSR deficiency induces immunogenic cell death via mitochondrial stress and STING pathway activation

Our findings reveal that GSR deficiency enhances the immunogenicity of ferroptosis and cuproptosis, two metal-dependent forms of regulated cell death that converge on mitochondrial stress and dysfunction. Although these pathways are initiated by distinct mechanisms, both induce excessive ROS production, in part through Fenton or Fenton-like reactions, leading to oxidative stress and mitochondrial damage. In ferroptosis, mitochondrial morphological changes are well documented, whereas in cuproptosis, excess copper binds to mitochondrial lipoylated TCA cycle proteins such as DLAT, triggering their aggregation, functional inactivation, and disruption of the electron transport chain (ETC), ultimately causing mitochondrial dysfunction.

To determine whether GSR deficiency exacerbates mitochondrial stress, we first measured intracellular ROS levels using H2DCFDA. Both genetic GSR knockout and pharmacological inhibition with BCNU markedly increased ROS accumulation when combined with RSL3 or Es-Cu (Fig. 7A-D, S14A-D). Live-cell imaging with MitoTracker further showed that these combination treatments disrupted the interconnected tubular mitochondrial network, resulting in pronounced mitochondrial fragmentation and aggregation (Fig. 7E, 7F, S14E-J). Consistently, MitoSOX staining demonstrated increased mitochondrial superoxide production (Fig. 7G-J, S14K, S14L). Flow cytometric analysis further confirmed that combined treatment with BCNU and either RSL3 or Es-Cu significantly increased both intracellular ROS and mitochondrial superoxide levels (Fig. 7K, 7L, S14M, S14N). Collectively, these findings demonstrate that GSR deficiency exacerbates oxidative stress and mitochondrial dysfunction during ferroptosis and cuproptosis.

Fig. 7.

GSR deficiency enhances cellular ROS levels and drives mitochondrial dysfunction during ferroptosis or cuproptosis. Genetic GSR knockout or pharmacological GSR inhibition with 100 μM BCNU was combined with ferroptosis inducer RSL3 (3 μM) or cuproptosis inducer ES-Cu (100 nM each) for 6 h. (A-D) Cellular ROS levels detected by H2DCFDA staining (Scale bars: 100 μm). (E, F) Mitochondrial morphological changes assessed by MitoTracker staining (Scale bars: 10 μm). (G-J) Mitochondrial superoxide levels detected by MitoSox staining (Scale bars: 100 μm). (K, L) Flow cytometry analysis of intracellular ROS (K) using H2DCFDA and mitochondrial superoxide (L) using MitoSox following the indicated treatments, with representative histograms shown on the left and quantification presented as bar plots on the right for each experiment. All statistics were analyzed by one-way ANOVA with Tukey’s correction. Asterisks indicate comparisons versus the DMSO control (A-D, G-L)

Because mitochondrial stress can promote the release of mitochondrial DNA (mtDNA) into the cytosol, we next examined cytosolic mtDNA accumulation. PicoGreen staining revealed a marked increase in cytosolic mtDNA puncta, with occasional aggregates, following GSR knock or inhibition under ferroptotic or cuproptotic conditions (Fig. 8A). Consistently, quantitative PCR-based assay confirmed a significant increase in cytosolic mtDNA copy number compared to control treatment (Fig. 8B, S15A). Time-course analysis showed that cytosolic mtDNA accumulation occurred rapidly, beginning at approximately 4 h and peaking between 4–8 h after treatment (Fig. S15B). Pharmacological inhibition of mitochondrial DNA release using cyclosporine A (CsA), an inhibitor of mitochondrial permeability transition pore (mPTP), or VBIT-4, a VDAC1 oligomerization inhibitor, markedly attenuated this increase, supporting a mitochondrial origin of the cytosolic DNA (Fig. S15A), consistent with the previous study [51, 52].

Fig. 8.

GSR deficiency enhances mitochondrial DNA release and STING pathway activation. (A) Enhanced mtDNA release in GSR-KO (left) or BCNU-treated (right) cancer cells during ferroptotic or cuproptotic cell death, visualized by MitoTracker (red, mitochondrial morphology) and PicoGreen (green, DNA) staining. White arrowheads indicate individual mtDNA signals, yellow arrowheads denote mtDNA aggregates in swollen mitochondria, and purple arrowheads highlight released mtDNA aggregates. Scale bars, 10 μm. (B) Quantification of isolated cytosolic mtDNA after indicated treatments. (C) Western blotting showing elevated cGAS, p-STING, p-IRF3, and p-TBK1 levels in GSR-KO (left) or BCNU-treated (right) cancer cells treated with RSL3 or Es-Cu. (D) Measurement of intracellular cGAMP levels in SH-SY5Y or 4T1 cells after the indicated treatments. (E) Representative immunofluorescence images and quantification of p-STING and p-TBK1 in orthotopic 4T1 tumor tissues, showing elevated staining when ferroptosis or cuproptosis inducers are combined with BCNU. Scale bar, 50 μm. (F) Co-localization of p-STING with 4T1 tumor cells, CD8⁺ cells, CD11c⁺ cells, CD4⁺ cells or CD68⁺ cells were assessed in 4T1 tumors. Scale bar, 100 μm. (G, H) Genetic deletion (KO) or pharmacologic inhibition (1 µM H151) of STING abrogated the increased release of immunogenic factors from treated cancer cells, including HMGB1 and IFN-β (G), and markedly impaired dendritic cell activation in the co-culture (H). (I) Tumor rechallenge assay showed robust protection in mice vaccinated with treated wild-type cells (BCNU in combination with RSL3 or Es), whereas this protection was markedly reduced in mice vaccinated with identically treated STING-KO cells. The right “tumor-free mouse” curve shows a subset of lines from the left curve, selected to reduce clutter. Statistics were analyzed by one-way ANOVA with Tukey’s correction (B, D, E, G, H), or log-rank test (I). Asterisks ns (not significant) without connecting lines indicate comparisons versus the DMSO control (B, D, E, G, H)

Given that cytosolic mtDNA is a well-established activator of the cGAS-STING pathway [53, 54], we next examined activation of this signaling axis. Consistent with mtDNA release, we observed robust cGAS-STING activation, as evidenced by elevated levels of cGAS, phosphorylated STING, IRF3, and TBK1 (Fig. 8C, S15C). STING phosphorylation became detectable at around 4 h and remained elevated beyond 24 h (Fig. S15D). In contrast, other nucleic acid sensing pathways, including RIG-I and AIM2, were not significantly altered (Fig. S15E). Importantly, BCNU combined with either RSL3 or Es-Cu significantly increased intracellular cGAMP levels (Fig. 8D), further confirming activation of cGAS-mediated DNA sensing. Together, these findings demonstrate that combination treatments activate cGAS-STING signaling through mtDNA release into the cytosol.

We next examined the downstream inflammatory consequences of STING activation. Activation of this pathway was accompanied by increased release of inflammatory mediators, including HMGB1, IFN-β, IFN-γ, IL-6, TNF-α, and TNF-β (Fig. S15F). These responses were largely abrogated by blockade of mtDNA release with cyclosporine A or VBIT-4, linking mitochondrial stress-induced mtDNA leakage to downstream inflammatory signaling (Fig. S15F). Furthermore, inhibition of BAX/BAK-mediated mitochondrial outer membrane permeabilization (MOMP) using MSN-125, a selective inhibitor of BAX/BAK oligomerization, almost completely abolished mtDNA release (Fig. S15G), consistent with previous studies [52, 55]. Collectively, these findings demonstrate that STING activation is driven by cytosolic mtDNA accumulation resulting from mitochondrial membrane permeabilization involving both mPTP opening and BAX/BAK-dependent MOMP.

Enhanced STING pathway activation was also observed in vivo across tumor models. In GSR-KO SH-SY5Y xenograft tumors treated with either RSL3 or Es (Fig. 3A), immunohistochemical analysis revealed markedly elevated levels of phosphorylated STING (p-STING) and p-IRF3, compared with NT control tumors (Fig. S15H). Similarly, in the syngeneic orthotopic 4T1 breast cancer model (Fig. 3D), BCNU combined with RSL3 or Es induced significantly higher p-STING and p-TBK1 staining than any single-agent treatment (Fig. 8E). Co-localization analysis further suggested that p-STING signals were predominantly detected in tumor cells rather than infiltrating immune cells (Fig. 8F), supporting a tumor cell-intrinsic mechanism of STING activation.

To determine whether STING activation is required for these downstream immunogenic effects, we generated STING-KO 4T1 cell lines (Fig. S16A). Compared to wild-type cells, STING-KO cells failed to upregulate the release of IFN-β, HMGB1 and other inflammatory cytokines upon BCNU combination treatment (Fig. 8G, S16B), and showed markedly reduced ability to stimulate dendritic cell activation (Fig. 8H, S16C). Similar effects were observed with the STING inhibitor H151. These results support that tumor cell-intrinsic STING signaling is required for the induction of key immunogenic signals and antigen-presenting cell activation in this context.

To directly assess immunogenic cell death in vivo, we performed a tumor rechallenge assay (Fig. S16D). 4T1 cells were treated in vitro with BCNU plus RSL3 or Es-Cu for 24 h and then inoculated into the left flank of mice as a “vaccination.” Ten days later, untreated 4T1 cells were implanted into the contralateral flank. Mice vaccinated with treated wild-type cells exhibited significant protection against subsequent tumor challenge, whereas vaccination with identically treated STING-KO cells resulted in markedly impaired protection (Fig. 8I), without affecting body weight (Fig. S16E). These findings demonstrate that STING signaling is required for effective antitumor immune memory and support the conclusion that GSR inhibition converts ferroptosis and cuproptosis into bona fide immunogenic cell death modalities in a STING-dependent manner.

GSR inhibition enhances the efficacy of immune checkpoint inhibitor (ICI) therapy

Given that GSR inhibition enhanced inflammatory signaling, dendritic cell activation, and antitumor immunity, we finally tested whether it could enhance the antitumor efficacy of ICI therapy by combining anti-PD-1 with ferroptosis or cuproptosis induction using a 4T1 syngeneic model (Fig. S17A). Triple combination therapy of anti-PD-1 + BCNU + RSL3 or Es achieved superior tumor suppression, with substantially reduced tumor volume, and prolonged survival compared with any dual treatment regimen (Fig. 9A, 9B, S17B), with only a marginal effect on body weight (Fig. S17C).

Fig. 9.

GSR deficiency synergizes with immune checkpoint inhibitor (ICI) therapy. (A, B) Reduced tumor volume (A) and prolonged survival (B) in a 4T1-derived subcutaneous tumor model treated with anti-PD-1 ICI therapy in combination with BCNU and ferroptosis or cuproptosis inducers. (C) Flow cytometry analysis of intratumoral CD3⁺ leukocytes, CD4⁺, CD8⁺ T cells, activated dendritic cells, and CD8⁺ T cells expressing GZMB and IFNγ in 4T1-derived tumors under indicated treatment. (D) Elevated serum IL-6 and TNF-α levels in treated mice. (E, F) Tumor growth curves and final tumor weights of a CT26-derived subcutaneous tumor model treated with anti-PD-1 ICI therapy in combination with BCNU and ferroptosis or cuproptosis inducers. (G) Correlation matrix of glutathione pathway activity with immune signatures, including immune cell proportions and PD-1/PD-L1 expression, in the cohort. (H) ssGSEA enrichment scores of glutamate and glutathione pathways in the Gide2019_PD1 melanoma cohort. (I) Kaplan–Meier survival curves from the Gide2019_PD1 melanoma cohort receiving anti-PD-1 therapy, stratified by tumor GSR expression levels. (J) Bar graphs showing fractions of memory B cells (adaptive immunity) and immune-suppressive M2 macrophages in responders and non-responders, as determined by CIBERSORT deconvolution. (K) Kaplan–Meier survival curves stratified by tumor fractions of B cells, M2 macrophages, or glutathione pathway activity in anti-PD-1-treated patients. Statistics were analyzed by one-way ANOVA with Tukey’s correction (A, C-F), or log-rank test (B). Asterisks ns (not significant) without connecting lines indicate comparisons versus the DMSO or IgG control (C, D, F)

Concurrently, this triple combination therapy drove a robust influx of CD45⁺ leukocytes, with significant elevations in CD3⁺ T lymphocytes, including both CD4⁺ and CD8⁺ subsets, and activated dendritic cells (Fig. 9C, S17D). Functional activation of T-cell was also enhanced, as evidenced by elevated IFN-γ production by CD4⁺ and CD8⁺ T cells and increased granzyme B (GZMB) secretion, particularly from CD8⁺ T cells (Fig. 9C, S17D). In addition to local immune activation, treatment also improved systemic immune parameters. Anti-PD-1 combination therapy alleviated cancer-associated splenomegaly, as shown by reduced spleen weight (Fig. S17E), and the anti-PD-1 combination in the cuproptosis regimen specifically increased circulating levels of TNF-α and IL-6 (Fig. 9D). No overt short-term toxicity was detected based on body weight and unchanged serum levels of albumin (ALB), alanine aminotransferase (ALT), and creatinine (CR) (Fig. S17F).

To determine whether the benefit of triple therapy extended beyond 4T1 model, we next tested the regimen in the CT26 model and directly compared anti-PD-1 + BCNU with the corresponding triple combination including Es (Fig. S17G). While anti-PD-1 + BCNU produced measurable tumor growth control, the addition of Es resulted in markedly improved therapeutic outcomes, with frequent complete or near-complete tumor regression (Fig. 9E, 9F), without affecting overall body weight (Fig. S17H). These results support a model in which GSR inhibition amplifies therapy-induced immunogenic cell death and tumor antigenicity, thereby improving responsiveness to PD-1 blockade. Consistent with this interpretation, triple therapy also more effectively alleviated splenomegaly (Fig. S17I), suggesting improved systemic immune restoration.

Clinical cohort analysis further supported the translational relevance of these findings. Gene set variation analysis (GSVA) of clinical cohorts showed that immunotherapy responding patients have increased ferroptosis signatures and increased immune response and inflammation, as well as decreased GSH levels (Fig. 9G, 9H, S18A). In parallel, high GSR expression was associated with worse clinical outcomes following ICI treatment (Fig. 9I, S18B), suggesting that elevated GSR-GSH metabolism may be linked to an immunotherapy-resistant phenotype.

To further define this relationship, we examined the association between GSH levels and tumor immune features. GSH level was negatively correlated with the tumor inflammation signature, including the STAT1 and T cells signature, the expression of PD-1 and PD-L1, and the proportion of CD8⁺ T cells, memory B cells, activated DCs and natural killer (NK) cells and M1-type macrophages, but positively correlated with the proportion of M2-type macrophages (Fig. 9G, 9J). Consistently, tumors with a high glutathione signature, increased M2 macrophages, and reduced memory B cells were associated with poorer survival (Fig. 9K).

We next evaluated the predictive value of GSR expression for immunotherapy response. GSR alone showed limited ability to distinguish responders from non-responders. However, combining GSR with PD-L1 expression substantially improved predictive accuracy: tumors with low GSR and high PD-L1 were enriched for responders, whereas tumors with high GSR and low PD-L1 were more likely to be non-responsive (Fig. S18C, 18D). This combinatorial model outperformed PD-L1 alone, suggesting that integrating metabolic and immune biomarkers more effectively captures response heterogeneity. Finally, ROC analysis, together with univariate and multivariate Cox regression analyses, further showed that the GSR-PD-L1 signature was associated with clinical outcomes in the analyzed cohort (Fig. S18E-H), supporting its potential as a hypothesis-generating candidate biomarker for patient stratification and possible immunotherapy responsiveness. Validation in independent cohorts and prospective studies will be important to establish its clinical applicability.

Discussion

The adaptation of cancer cells to oxidative stress underscores a key therapeutic vulnerability: their dependence on enhanced antioxidant defense pathways to maintain redox homeostasis amid elevated ROS stress. In this context, we identify GSR as a pivotal node in this balance, demonstrating that its genetic ablation or pharmacological inhibition potently sensitizes diverse cancer cells to two emerging metal-dependent RCD pathways, namely ferroptosis and cuproptosis, without compromising normal proliferation. This synergy was consistently observed across a broad panel of genetically diverse cancer cell lines and in vivo models. Our findings further show that GSR inhibition disrupts GSH homeostasis, amplifies oxidative stress, and enhances the immunogenicity of cancer cell death through mitochondrial stress and activation of the cGAS-STING pathway. These effects synergize with ICI therapy, supporting the potential of this strategy to enhance antitumor immunity.

Linking GSR to anti-tumor immunity in the context of ferroptosis and cuproptosis

Our genome-wide CRISPR/Cas9 screen identified GSR as the top ferroptosis-sensitizing hit, validated across multiple cancer types through genetic knockout, overexpression, and pharmacological inhibition. GSR catalyzes the NADPH-dependent regeneration of reduced GSH from its oxidized disulfide (GSSG), sustaining the GPX4-xCT-GSH axis that detoxifies lipid peroxides. Remarkably, we also found that GSR inhibition potentiates cuproptosis, a recently described mitochondrial RCD triggered by copper overload in all tested cancer models. GSR deficiency lowers intracellular GSH, a physiological copper chelator, resulting in enhanced DLAT oligomerization and mitochondrial fragmentation. Although ferroptosis and cuproptosis are initiated by distinct triggers, lipid peroxidation in the former and copper binding to lipoylated TCA cycle proteins in the latter, they converge on mitochondrial dysfunction as a shared downstream pathology. GSR deficiency exacerbates this convergence by weakening cellular antioxidant defenses, leading to ROS accumulation, disruption of mitochondrial morphology, and elevated mitochondrial superoxide levels. These observations position GSR as a molecular switch that integrates redox imbalance with metal-dependent cell death across cancer types.

GSR thus emerges as a convergent metabolic regulator, controlling a shared redox vulnerability between ferroptosis and cuproptosis, highlighting its potential as a therapeutic target in redox-dependent cancers (Fig. 10A). Whereas direct GPX4 inhibitors or copper ionophores selectively trigger individual pathways, upstream targeting of GSR provides a strategy to deplete GSH and simultaneously sensitize cancer cells to both metal-dependent RCD modalities.

Fig. 10.

Model illustrating how GSR deficiency links ferroptosis, cuproptosis, and cGAS-STING-mediated immunogenic cell death. (A) Schematic illustrating GSR functions as a central regulatory bridge linking ferroptosis and cuproptosis. GSR (green ovals) deficiency, achieved pharmacologically with BCNU, LCS3, or 2-AAPA, depletes GSH, simultaneously impairing GPX4-dependent lipid peroxide clearance (ferroptosis) and copper buffering (cuproptosis, triggering DLAT aggregation and mitochondrial proteotoxic stress). This convergently amplifies mitochondrial dysfunction across both metal-dependent RCD pathways, sensitizing tumor cells to both ferroptotic and cuproptotic cell death pathways. (B) Model of the immunogenic cascade triggered by GSR deficiency, where mitochondrial collapse under RSL3 or Es-Cu induces mtDNA release, activates the cGAS-STING pathway, promotes inflammatory cytokine secretion, exposes DAMPs, and drives T-cell infiltration and dendritic cell maturation, thereby transforming cell death into potent immunogenic cell death

Beyond direct cytotoxicity, our findings demonstrate that GSR inhibition enhances the immunogenic potential of ferroptosis and cuproptosis. Both processes trigger the release of canonical DAMPs, including HMGB1, CRT, and ATP, which collectively facilitate DC-maturation and T-cell activation. Co-culture experiments confirmed that conditioned media from GSR-inhibited cancer cells undergoing ferroptosis or cuproptosis robustly activated DCs and induced proinflammatory cytokines such as IFN-β, TNF-α, IL-6, IFN-γ, and TNF-β (Fig. 5, S11).

Mechanistically, our findings support a compartmentalized model in which ferroptosis and cuproptosis enhanced by GSR inhibition induce mitochondrial stress and mtDNA release, thereby activating cGAS-STING signaling [44, 54] mainly in tumor cells, whereas tumor-derived HMGB1 may engage TLR4 signaling primarily in BMDCs or other myeloid cells (Fig. 10B). By simultaneously inducing oxidative stress and immunostimulatory signaling, GSR inhibition transforms metal-dependent cell death into a potent immunogenic event. Collectively, these findings extend the role of GSR beyond a classical metabolic enzyme to a regulator of tumor-immune interactions.

GSR as a key molecular node connecting ferroptosis and cuproptosis

Ferroptosis and cuproptosis are non-apoptotic, metal-dependent RCD pathways characterized by intense oxidative stress and mitochondrial dysfunction. While mechanistically distinct, both can elicit immunogenic cell death through overlapping downstream processes, particularly when redox homeostasis is disrupted by GSR inhibition. Nevertheless, their immunogenic programs exhibit important qualitative differences.

Ferroptosis, driven by iron-dependent lipid peroxidation, generates abundant oxidative DAMPs such as 4-HNE [56], thereby amplifying tumor microenvironment inflammation. In EMT-high tumors, ferroptotic cells prime DCs, eliciting potent inflammatory signatures via release of ATP, HMGB1, and CRT exposure. However, whether ferroptosis is immunogenic is controversial. Some research has found that late stage ferroptosis is actually immunosuppressive [57, 58], whereas other studies found otherwise [42, 59, 60]. This apparent discrepancy likely reflects the fact that the immunogenicity of ferroptosis is both highly context-dependent [61] and strongly tumor-type dependent, as different tumor types differ in immune composition and redox state, and intense oxidative stress during ferroptosis may further increase DAMP release and shift its immunogenicity. In contrast, cuproptosis is triggered by copper-induced mitochondrial proteotoxic stress, causing lipoylated TCA cycle protein aggregation and potential neoantigen formation. Discovered more recently, its immunogenicity is underexplored, with studies largely limited to nanoparticle-mediated induction [62–65], where DAMP release and immune infiltration have been observed.

Thus, ferroptosis emphasizes lipid oxidative damage and broad inflammation, whereas cuproptosis leverages proteotoxicity and possible neoantigen generation, offering complementary strategies for immune activation in redox-vulnerable cancers. This is consistent with our findings there is no cross-inhibition (Fig. S5G-I) and that they exhibit distinct hallmarks (Fig. 2F, 2G).

Importantly, GSR emerges as a central regulator that functionally links these cell death pathways. Notably, GSR inhibition synergistically amplifies both ferroptotic and cuproptotic immunogenicity by depleting GSH, disrupting redox homeostasis, impairing mitochondria, and activating cGAS-STING, thereby boosting inflammatory gene expression and immune recognition.

Together, our findings support a model in which ferroptosis and cuproptosis remain mechanistically distinct at the upstream level but converge on shared mitochondrial and innate immune signaling pathways downstream. GSR serves as a key metabolic node that modulates this convergence, shaping the magnitude and nature of immunogenic cell death.

Harnessing GSR for antitumor treatment

Targeting GSH metabolism in therapy-resistant cancers may represent a promising strategy for addressing some of the most aggressive malignancies. Elevated GSR expression and enhanced GSH regeneration are hallmark features of numerous solid tumors, irrespective of tissue origin and driver mutations. For instance, glioblastomas, which are highly resistant to apoptosis due to aberrant PI3K/AKT signaling [66], frequently overexpress xCT and copper transporters [67, 68], making them particularly susceptible to ferroptosis and cuproptosis inducers. Similarly, KRAS-mutant cancers, which upregulate NRF2 and GSH synthesis [11], may be effectively sensitized to ferroptotic and cuproptotic cell death through GSR inhibition. Notably, GSR inhibition spares paclitaxel-induced apoptotic nuclear fragmentation but promotes membrane rupture in metal-dependent RCD, highlighting its distinction from conventional chemotherapy.

A critical aspect of harnessing GSR inhibition also lies in the preferential sensitization of cancer cells to ferroptosis and potentially cuproptosis [27, 28, 69]. This selectivity likely arises from the inherently higher basal reactive oxygen species (ROS) levels, increased mitochondrial susceptibility, and greater dependence on GSH redox buffering in tumor cells compared to normal tissues. These vulnerabilities make cancer cells more prone to oxidative damage when GSR function is compromised.

By targeting redox homeostasis, this study uncovers a role for GSR as a molecular bridge between antioxidant defense, metal-dependent RCD, and tumor immunogenicity. By depleting GSH and disrupting mitochondrial integrity, GSR inhibition shifts the redox balance toward oxidative stress while simultaneously converting non-apoptotic cell death into a more immunogenic form. These findings imply that GSR expression and intracellular GSH levels warrant further investigation as candidate biomarkers for oxidative stress sensitivity and immunotherapy responsiveness [70]. Moreover, therapeutic co-targeting of GSR, using pharmacologic inhibitors such as BCNU or next-generation inhibitors, in combination with metal ionophore-based or lipid peroxidation inducers represents a rational immunotherapy approach. Given BCNU’s FDA-approved status, its clinical repurposing alongside ferroptosis or cuproptosis inducers and ICIs warrants further preclinical investigation as a possible therapeutic strategy.

Limitations, future directions, and conclusions

Our study has several limitations that warrant discussion and highlights interesting directions for future investigation.

While BCNU is not a highly selective GSR inhibitor, it remains a well-characterized pharmacological inhibitor of GSR and has the additional advantage of being an FDA-approved drug with established clinical use. To address its broader pharmacological activities, we complemented BCNU treatment with genetic GSR deletion and additional pharmacological GSR inhibitors. These independent approaches consistently reproduced the key combination-treatment phenotypes, including enhanced sensitivity to ferroptosis and cuproptosis, increased mitochondrial stress, and augmented immunogenic signaling. Together, these findings provide convergent evidence that GSR inhibition makes a substantial contribution to the sensitizing effects of BCNU when combined with RSL3 or Es-Cu.

At the same time, we recognize that BCNU also acts through DNA alkylation and potentially other GSR-independent mechanisms. We therefore do not attribute its monotherapy cytotoxicity or antitumor activity solely to GSR inhibition. Instead, we restrict our GSR-dependent conclusions to the sensitization and combination-treatment phenotypes, for which the genetic and pharmacological evidence consistently supports an important role for GSR inhibition.

Our study also does not directly distinguish between tumor-intrinsic and host-derived contributions of GSR to the therapeutic response. Future studies employing tumor-specific and host-specific GSR knockout models will be necessary to fully delineate their respective roles. Ferroptosis is reported to suppress STING activation by promoting STING carbonylation [71]. Determining whether ferroptosis or cuproptosis modulates STING carbonylation will be valuable for mechanistic understanding.

Despite the therapeutic potential of GSR inhibition, tumor cells may activate compensatory antioxidant pathways, which could limit long-term efficacy. Future studies may need to identify and co-target these adaptive responses.

Another limitation is the lack of formal patient stratification according to copper/iron homeostasis. While we analyzed GSR expression and glutathione metabolism, we did not perform a standardized, cross‑cohort evaluation of copper‑ and iron‑related metabolic gene signatures, in part because consensus gene sets for these pathways are not yet established. Future work should therefore focus on defining and validating robust copper/iron homeostasis signatures and testing their utility for prospective patient stratification and clinical decision-making.

In conclusion, we identify GSR as a critical metabolic regulator that simultaneously governs cellular susceptibility to ferroptosis and cuproptosis by controlling intracellular GSH bioavailability. By depleting GSH and disrupting mitochondrial integrity, GSR inhibition sensitizes cancer cells to ferroptosis and cuproptosis while rendering these death pathways highly immunogenic through cGAS-STING activation. These findings establish GSR as a molecular hub linking redox imbalance to antitumor immunity and provide a strong rationale for targeting GSR to enhance the therapeutic efficacy of metal-dependent regulated cell death-based therapies and cancer immunotherapy.

Supplementary Information

Original western blots ^{(2.7MB, pdf)}

Supplementary figures ^{(63MB, pdf)}

Supplementary Table S1 ^{(46.1KB, docx)}

Gating strategies ^{(554.5KB, pdf)}

Abbreviations

ALB

Albumin

ALT

Alanine aminotransferase

BCNU

1, 3-Bis (2-chloroethyl) − 1-nitrosourea

BLI

Bioluminescence imaging

BMDCs

Bone marrow-derived dendritic cells

BSO

Buthionine sulfoximine

cGAS

Cyclic GMP-AMP synthase

CR

Creatinine

CRT

Calreticulin

CTRP

Cancer therapeutics response portal

DAMPs

Damage-associated molecular patterns

DC

Dendritic cell

DLAT

Dihydrolipoamide S-acetyltransferase

ECM

Extracellular matrix

EMT

Epithelial-mesenchymal transition

Es

Elesclomol

Es-Cu

Elesclomol-copper

ETC

Electron transport chain

Fer-1

Ferrostatin-1

GCL

Glutamate-cysteine ligase

GO

Gene Ontology

GPX4

Glutathione peroxidase 4

GSEA

Gene set enrichment analysis

GSH

Reduced glutathione

GSR

Glutathione reductase

GSSG

Glutathione disulfide

GSVA

Gene set variation analysis

GZMB

Granzyme B

H&E

Hematoxylin and eosin

HMGB1

High mobility group box 1

4-HNE

4-Hydroxynonenal

ICI

Immune checkpoint inhibitor

IRF3

Interferon regulatory factor 3

KO

Knockout

Lip-1

Liproxstatin-1

mtDNA

Mitochondrial DNA

Nec-1

Necrostatin-1

NAC

N-Acetylcysteine

NK

Natural killer

NT

Non-targeting

PD-1

Programmed cell death protein 1

RCD

Regulated cell death

ROS

Reactive oxygen species

SPF

Specific pathogen free

STING

Stimulator of interferon genes

TBK1

TANK-binding kinase 1

TCA

Tricarboxylic acid

TCGA

The cancer genome atlas

TTM

Tetrathiomolybdate

TXNRD1

Thioredoxin reductase 1

Z-VAD

Z-VAD-FMK

Author contributions

Conceptualization: Z.D., X.D., H.Y., Y.L., Y.V.Y., Y.N.J.; Methodology: X.D., Z.D., B.Z., C.L., P.Z., L.L., Y.N.J., Y.V.Y.; Investigation: X.D., Z.D., B.Z., R.H., J.M.; Writing—original draft: X.D., Z.D., Y.V.Y., Y.N.J.; Writing—review and editing: X.D., Y.V.Y., Y.N.J.; Visualization: X.D., Z.D., Y.V.Y., Y.N.J.; Supervision: Y.L., Y.V.Y., Y.N.J.; Funding acquisition: H.Y., Y.L., Y.N.J.

Funding

This work was supported by the National Natural Science Foundation of China (Grant Number. 82172841), Translational Medicine and Interdisciplinary Research Joint Fund of Zhongnan Hospital of Wuhan University (ZNJC202321), and the Fundamental Research Funds for the Central Universities (2042022dx0003).

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information. RNA-seq data were deposited into SRA NCBI (BioProject: PRJNA1206958).

Declarations

Ethics approval and consent to participate

All research in this paper was conducted in full compliance with all applicable ethical standards and regulations. All animal experiments were conducted under protocol (MRI2023-LAC165), which was reviewed and approved by the Laboratory Animal Ethics Committee of Wuhan University. Consent to Participate: not applicable.

Competing interests

The authors declare no competing interests.