Metabolic plasticity underlies ferroptosis resistance driven by mutant p53 in pancreatic ductal adenocarcinoma

Abstract

Background

Pancreatic ductal adenocarcinoma (PDAC) is an aggressive malignancy characterized by late diagnosis, rapid progression, and resistance to conventional therapies. Mutations in the tumor suppressor gene TP53 are prevalent in this cancer and have been associated with gain-of-function activities that promote tumor survival. Ferroptosis, an iron-dependent form of regulated cell death driven by lipid peroxidation, represents a potential therapeutic vulnerability. However, the influence of mutant p53 on ferroptosis susceptibility remains poorly understood.

Methods

We employed isogenic pancreatic cancer cell models, including CRISPR-Cas9-mediated TP53 Knockout (TP53-KO) and transient overexpression of common mutant TP53 variants, to examine ferroptosis sensitivity. Cell viability, reactive oxygen species accumulation, lipid peroxidation, mitochondrial integrity, and metabolic profiling were assessed following treatment with ferroptosis inducers. Transcriptional changes were analysed by RNA sequencing, and functional contributions of glycolysis and mitochondrial respiration were evaluated using metabolic flux assays. Pharmacological reactivation of wild-type p53 was tested in combination with ferroptosis inducers in vitro and in orthotopic murine models.

Results

Deletion of mutant TP53 sensitized pancreatic cancer cells to ferroptosis, accompanied by increased oxidative stress, lipid peroxidation and mitochondrial dysfunction. Mutant TP53 expression preserved mitochondrial integrity and sustained bioenergetic flexibility under ferroptotic stress. Transcriptomic analyses revealed a multi-layered adaptive program, including upregulation of antioxidant and metabolic genes, and activation of PI3K–AKT signaling, which is associated to a selective glycolytic shift to maintain ATP levels. Functional modulation confirmed that glucose supplementation enhanced, whereas glycolytic inhibition impaired, survival under ferroptotic conditions in mutant TP53-expressing cells, but not in TP53-KO cells. Pharmacological reactivation of wild-type p53 disrupted this adaptive network, abrogating glycolytic reprogramming and significantly increasing ferroptotic cell death both in vitro and in vivo.

Conclusions

Mutant p53 orchestrates transcriptional and metabolic plasticity that confers ferroptosis resistance in pancreatic cancer. Reactivating wild-type p53 in combination with ferroptosis induction overcomes this resistance, highlighting a therapeutically actionable vulnerability and supporting the development of combinatorial strategies for this aggressive malignancy.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12943-026-02746-y.

Background

Pancreatic ductal adenocarcinoma (PDAC) remains one of the deadliest malignancies worldwide, with a dismal five-year survival rate of just 13% [1]. Its poor prognosis is largely driven by late diagnosis, rapid metastatic progression, profound metabolic plasticity, and intrinsic resistance to conventional therapies [2]. Despite advances in targeted and immunotherapeutic strategies, effective treatment options for advanced PDAC remain extremely limited, underscoring the urgent need to identify novel therapeutic vulnerabilities.

Among the key genetic alterations driving PDAC progression, mutations in the tumor suppressor gene TP53 are particularly prevalent, occurring in about 50–70% of cases [3, 4]. Mutant p53 promotes tumorigenesis through multiple mechanisms, including loss of its tumor-suppressive function [5], dominant-negative inhibition of wild-type p53 [6], and the acquisition of gain-of-function (GOF) activities that actively drive cancer progression, metastasis, and therapy resistance [7–9]. Importantly, different TP53 missense mutations exhibit substantial functional heterogeneity. Common hotspot mutants such as TP53^R175H, TP53^R248W, and TP53^R273H display distinct structural and transcriptional properties, leading to divergent biological outcomes depending on cellular and tissue context [10, 11]. However, the contribution of specific PDAC-associated mutant p53 variants to therapy resistance mechanisms remains incompletely understood. Recent advances have highlighted ferroptosis, a regulated iron-dependent form of cell death characterized by lipid peroxidation, as a promising therapeutic vulnerability in PDAC [12, 13]. PDAC cells exhibit profound alterations in iron metabolism, redox homeostasis, and lipid remodeling, rendering them particularly susceptible to ferroptotic stress under specific metabolic conditions [14, 15]. Importantly, modulation of ferroptosis-related pathways, such as GPX4, iron metabolism, and lipid peroxidation, has been shown to overcome chemotherapy resistance by enhancing ferroptotic cell death in tumor cells [16]. Nevertheless, the molecular determinants governing ferroptosis sensitivity in PDAC remain poorly defined. Wild-type p53 has been implicated in ferroptosis regulation through context-dependent mechanisms involving SLC7A11 repression, reactive oxygen species accumulation, and modulation of lipid peroxidation pathways, generally promoting ferroptotic sensitivity [17, 18]. In contrast, the role of mutant p53 in ferroptosis remains far less clear and appears highly context-specific. While some studies suggest that certain p53 mutants may enhance oxidative stress tolerance and metabolic adaptation, thereby suppressing ferroptotic cell death [19], the mechanistic relationship between PDAC-associated mutant p53 variants and ferroptosis susceptibility has not been systematically investigated. In particular, whether ferroptosis resistance represents a shared feature across different GOF mutants or a mutation-specific phenotype remains an unresolved question. In this study, we investigate how mutant p53 modulates ferroptosis sensitivity in PDAC cells, focusing on both functional outcomes and underlying molecular mechanisms. Using CRISPR-Cas9 knockout models, overexpression systems, and multi-level metabolic and transcriptional analyses, we reveal that mutant p53 confers resistance to ferroptotic cell death by preserving mitochondrial integrity and orchestrating a metabolic adaptation towards glycolysis that supports survival under oxidative stress. Importantly, our work provides mechanistic insight into how mutant p53 rewires ferroptosis-related metabolic pathways in PDAC and identifies a combinatorial therapeutic strategy capable of increasing ferroptosis sensitivity in mutant p53-expressing cells. These findings not only clarify a previously underexplored link between mutant p53 and ferroptosis regulation in PDAC but also highlight mutant p53-driven metabolic plasticity as a clinically actionable barrier to ferroptosis-based therapies.

Results

Mutant p53 deletion sensitizes PDAC cells to ferroptosis

To investigate the role of mutant p53 in regulating ferroptosis in PDAC, we used CRISPR-Cas9 knockout (KO) clones generated from the human PDAC cell line PANC-1 [20, 21], harbouring the hotspot TP53^R273H mutation commonly found in PDAC patients. Three independent TP53-KO clones were selected following confirmation of p53 loss by western blotting (Fig. 1A). To assess their sensitivity to ferroptosis, we treated both parental and TP53-KO cells with three distinct ferroptosis inducers: Erastin (a system Xc⁻ inhibitor [22]), RAS-Selective Lethal 3 (RSL3; a GPX4 inhibitor [23]), and Ferroptocide (a ferroptosis inducer targeting the thioredoxin antioxidant system [24]), across a range of concentrations. Cell viability was evaluated using three complementary approaches: MTT assay (for metabolic activity evaluation) (Fig. 1B), CellTox Green cytotoxicity assay (for membrane integrity assessment) (Fig. 1C), and Annexin V/PI staining followed by flow cytometry (for cell death profiling) (Fig. 1D and S1A). All three assays consistently demonstrated that TP53-KO clones are significantly more sensitive to ferroptosis-inducing agents compared to parental PANC-1 cells. The effect was robust and reproducible across all three TP53-KO clones and all three ferroptosis inducers. To confirm that the observed cell death was specifically attributable to ferroptosis, TP53-KO cells were co-treated with established ferroptosis inhibitors Ferrostatin-1 [25] and Liproxstatin-1 [26]. Both inhibitors fully rescued cell viability in TP53-KO clones treated with the aforementioned ferroptosis inducers (Fig. 1E). In contrast, pharmacological blockade of alternative cell death pathways, including apoptosis (z-VAD-fmk), necroptosis (Necrostatin-1), and autophagy (Bafilomycin A1), failed to prevent cell death, confirming the specificity of the ferroptotic response (Fig. S1B-C). Mechanistically, the increased susceptibility to ferroptosis of TP53-KO cells was associated with significantly elevated levels of reactive oxygen species (ROS) (Fig. 1F), and enhanced lipid peroxidation, measured by 4-Hydroxynonenal (4-HNE) western blot (Fig. 1G and Fig. S1D). Ultrastructural analysis by transmission electron microscopy (TEM) further revealed early and profound organelle damage in TP53-KO PANC-1 cells exposed to RSL3, including dilated endoplasmic reticulum, loosened Golgi apparatus, and a pronounced dilatation of the perinuclear space, all evident within 6 h of treatment. On the contrary, parental cells remained morphologically intact at this time point (Fig. 1H). Altogether, these findings suggest that TP53^R273H mutation in PANC-1 cells confers resistance to ferroptosis and that its genetic ablation re-sensitizes cells to ferroptotic cell death by amplifying oxidative stress and lipid peroxidation. Importantly, the differential sensitivity to ferroptosis was not associated with altered proliferative capacity, as Ki67 staining performed following treatment with ferroptosis-inducing agents revealed comparable proliferation rates (Fig. S1E). To further validate these observations under a more physiological ferroptosis-inducing condition, cells were subjected to cyst(e)ine (CYS) deprivation. Consistent with the results obtained using pharmacological ferroptosis inducers, CYS starvation [27, 28] triggered ferroptotic cell death with similar kinetics, with parental PANC-1 cells displaying significantly greater resistance than TP53-KO clones (Fig. S1F). Furthermore, in line with previous findings demonstrating that ferroptosis evasion could be related to chemotherapy resistance [29], we found that prolonged exposure to low-dose gemcitabine for 5 days resulted in a significantly greater loss of viability in TP53-KO cells compared with parental PANC-1 cells (Fig. S1G). Finally, the increased ferroptosis sensitivity associated with mutant p53 loss was confirmed in previously established TP53-KO clones derived from the independent PDAC cell line MIA PaCa-2 (Fig. S1H) [20], harboring the hotspot TP53^R248W mutation, suggesting that this phenotype is not restricted to a single cellular model (Fig. S1I).

Fig. 1.

Mutant p53 protects PDAC cells from ferroptosis. A Western Blot of p53 in TP53^R273HPANC-1 and TP53-KO PANC-1 clones. Vinculin was used as loading control; B Cell viability curves (MTT assay) of TP53^R273H PANC-1 and TP53-KO PANC-1 cells after 24 h treatment with Erastin (IC₅₀: PANC-1 = 38 µM; TP53-KO = 6 µM), RSL3 (IC₅₀: PANC-1 = 2.5 µM; TP53-KO = 0.3 µM), and Ferroptocide (IC₅₀: PANC-1 = 8 µM; TP53-KO = 5 µM); (C) Images of TP53^R273H PANC-1 and TP53-KO PANC-1 cells stained with CellTOX Green dye and treated with Erastin, Ferroptocide (5 µM – 6 h) and RSL3 (0.5 µM – 4.5 h). Scale bar: 100 μm. D Flow cytometry evaluation of cell viability of TP53^R273H PANC-1 and TP53-KO PANC-1 cells after 24 h treatment with Erastin, RSL3 and Ferroptocide. The bar graphs represent the percentage of Annexin V-/PI- (viable) cells (n = 3). p-values were determined by 2wayANOVA using Tukey’s multiple comparisons test (**p = < 0.01, ***p < 0.001, ****p = < 0.0001); E Flow cytometry evaluation of cell viability of TP53-KO PANC-1 cells pre-treated with Ferrostatin-1 and Liproxstatin-1 (5 µM − 1 h) before adding Erastin (5 µM), RSL3 (0.5 µM) and Ferroptocide (5 µM) for further 24 h. The bar graphs represent the percentage of Annexin V-/PI- (viable) cells (n = 3). p-values were determined by ordinary one-way ANOVA using Tukey’s multiple comparisons test (****p = < 0.0001); F Flow cytometry evaluation of ROS accumulation in TP53^R273H PANC-1 and TP53-KO PANC-1 cells after 24 h treatment with Erastin (5 µM), RSL3 (0.5 µM) and Ferroptocide (5 µM). The bar graphs and overlay histograms represent the Mean Fluorescence Intensity (MFI) of the cells positive for the CellROX dye (n = 3). p-values were determined by ordinary one-way ANOVA using Tukey’s multiple comparisons test (**p = < 0.01, ****p = < 0.0001); G Western blot of 4-HNE in TP53^R273H PANC-1 and TP53-KO PANC-1 cells after 24 h treatment with RSL3 (0.5 µM). Vinculin was used as loading control; H Ultrastructural images from Transmission Electron Microscopy (TEM) of TP53^R273H PANC-1 and TP53-KO PANC-1 cells after 6 h treatment with RSL3 (0.5 µM). N, nucleus; golgi apparatus (in blue); SeR, smooth endoplasmic reticulum (in red); arrowhead, dilatation of space between inner and outer nuclear membranes. Bars: 200 nm

Over-expression of PDAC-associated mutant p53 variants limits ferroptotic cell death in p53-null pancreatic cancer cells

To further dissect the contribution of GOF mutant p53 isoforms to ferroptosis resistance, we employed a complementary approach based on transient expression of mutant p53 in p53-null PDAC cells AsPC-1. As a preliminary validation, we compared the response of PANC-1 and p53-null AsPC-1 cells to ferroptosis induction and we found that AsPC-1 cells were significantly more sensitive to RSL3 and Ferroptocide treatment than PANC-1 cells (Fig. S2A), consistent with a ferroptosis-protective role of mutant p53. Then, six of the most prevalent TP53 hotspot mutations identified in PDAC patients (TP53^R273H, TP53^R175H, TP53^R282W, TP53^R248H, TP53^R248Q, and TP53^Y220C) were individually expressed and protein expression was confirmed by western blotting (Fig. 2A). Following transfection, cells were exposed to Erastin, RSL3, or Ferroptocide for 24 h and ferroptotic cell death was quantified by Annexin V/PI staining and flow cytometry. Notably, expression of each mutant p53 isoform significantly attenuated ferroptosis-induced cell death compared with mock-transfected controls, resulting in increased cell viability across all three ferroptosis-inducing conditions (Fig. 2B and Fig. S2B). These findings are consistent with the increased ferroptosis sensitivity observed following TP53 deletion in PANC-1 and MIA PaCa-2 cells and indicate that ferroptosis resistance represents a shared functional property of multiple PDAC-associated mutant p53 proteins. As a biological comparison, we next examined the effect of wild-type p53 expression in the same experimental setting. Given the rapid turnover and tight physiological regulation of wild-type p53, lower amounts of wild type TP53 expression plasmid were used compared with those employed for mutant p53 isoforms. In contrast to the protective effect exerted by mutant p53 protein, wild-type p53 expression increased the susceptibility of AsPC-1 cells to RSL3-induced ferroptosis in a dose-dependent manner, resulting in enhanced cell death relative to mock-transfected cells (Fig. S2C-D). These findings are consistent with the established tumor suppressor function of wild-type p53 and with previous studies demonstrating that wild type p53 promotes ferroptosis [17]. Mechanistically, mutant p53 expression was associated with reduced ROS accumulation following RSL3 treatment compared with mock-transfected controls (Fig. 2C). Furthermore, all mutant p53 variants significantly attenuated lipid peroxidation, as evidenced by decreased accumulation of 4-HNE protein adducts detected by western blot analysis (Fig. 2D). Consistent results were obtained using C11-BODIPY staining, which revealed reduced lipid ROS levels in cells expressing TP53^R273H following ferroptosis induction (Fig. S2E). Together, these data indicate that distinct GOF mutant p53 isoforms converge on a common ferroptosis-resistant phenotype characterized by suppression of oxidative stress and lipid peroxidation.

Fig. 2.

Mutant-p53 transfection confers ferroptosis resistance in p53-null AsPC-1 cells. A Western blot evaluation of p53 in AsPC-1 cells after 48 h transfection with mutant p53 isoforms or mock control. Vinculin was used as loading control; B Flow cytometry evaluation of AsPC-1 cell viability after transfection with mutant-p53 isoforms or mock control and treatment with Erastin (5 µM), RSL3 (0.5 µM) and Ferroptocide (5 µM) for further 24 h. The results are shown in fold change (related to each untreated control) and the bar graphs represent the percentage of Annexin V-/PI- (viable) cells (n = 3). p-values were determined by ordinary one-way ANOVA using Tukey’s multiple comparisons test (*p = < 0.05, **p = < 0.01, ***p < 0.001, ****p = < 0.0001); (C) Flow cytometry evaluation of ROS accumulation in AsPC-1 cells after transfection with mutant-p53 isoforms or mock control and treatment with RSL3 (0.5 µM) for 6 h. The results are shown in fold change (related to each control) and bar graphs represent the Mean Fluorescence Intensity (MFI) of the cells positive for the CellROX dye (n = 3). p-value was determined by 2wayANOVA using Tukey’s multiple comparisons test (*p = < 0.05); D Western blot evaluation of 4-HNE in AsPC-1 cells after transfection with mutant-p53 isoforms or mock control and treatment with RSL3 (0.5 µM) for 24 h. Vinculin was used as loading control, and the bar graph shows the protein level compared to the mock control; E–F Violin plots showing the anti-ferroptotic score (E) and the redox-ferroptotic score (F) derived from transcriptomic and mutational data from the TCGA pancreatic adenocarcinoma (TCGA-PAAD) cohort, stratified by wild-type or mutant TP53 status. p values were determined using two-sided Wilcoxon rank-sum tests; G–J Dot plots showing RNA-seq expression of ferroptosis-related markers in PDAC cell lines following gemcitabine exposure or acquisition of gemcitabine resistance; G-H) PANC-1 (G) and BxPC-3 (H) cells treated with gemcitabine versus control cells (GSE249302); I-J CFPAC-1 (I) and BxPC-3 (J) gemcitabine-resistant (Gem-R) derivatives versus parental cells (GSE140077). p values were calculated using two-sided unpaired t-tests on log2-transformed expression values

To investigate whether this phenotype is reflected in human disease, we analysed transcriptomic and mutational data from the TCGA pancreatic adenocarcinoma (TCGA-PAAD) cohort. Patients were stratified according to TP53 mutational status and expression of genes involved in ferroptosis regulation and redox homeostasis was examined. TP53-mutated tumors displayed significantly higher anti-ferroptotic and redox-related transcriptional signatures than TP53 wild-type tumors. Specifically, a composite GPX4/SLC7A11 anti-ferroptotic score was significantly increased in TP53-mutated samples (p = 0.043), and an extended GPX4/SLC7A11/TXN redox-ferroptosis score showed an even stronger association with TP53 mutational status (p = 6.5 × 10⁻⁶) (Fig. 2E-F). These findings support the existence of an enhanced anti-ferroptotic transcriptional program in TP53-mutated PDAC. Given our observation that TP53-deficient cells display increased sensitivity to prolonged gemcitabine treatment, we next investigated whether anti-ferroptotic programs are associated with adaptation to chemotherapy. To this end, we interrogated publicly available RNA-seq datasets from mutant p53-PDAC cell lines (PANC-1, BxPC-3, CFPAC-1) exposed to gemcitabine or exhibiting acquired gemcitabine resistance (GEM-R). Analysis of GSE249302 and GSE140077 revealed a consistent increase in a composite anti-ferroptotic signature comprising GPX4, SLC7A11, and TXN following gemcitabine exposure and in gemcitabine-resistant derivatives. This trend was observed across all PDAC models analysed (Fig. 2G-J). These findings suggest that activation of ferroptosis-protective pathways may represent an adaptive response associated with chemotherapy tolerance and resistance. Notably, we previously demonstrated that chemotherapeutic treatments, including gemcitabine, induce Ser15 phosphorylation of mutant p53, a modification associated with its activation [21, 30]. Therefore, the induction of anti-ferroptotic programs upon gemcitabine exposure may reflect the engagement of mutant p53-dependent adaptive responses, raising the possibility that the acquisition of gemcitabine resistance is mediated, at least in part, through the ferroptosis-protective mechanisms described here.

Mutant p53 status is associated with preserved mitochondrial integrity during ferroptosis in pancreatic cancer cells

Mitochondria play a crucial role in ferroptosis, as their morphology and functionality can influence cellular redox status and susceptibility to lipid peroxidation [31]. Building on our recent publication demonstrating distinct mitochondrial morphology and oxidative phosphorylation (OXPHOS) profiles in PDAC cells harboring mutant p53 compared to p53 wild-type or null cells [20], we conducted a detailed ultrastructural and functional analysis of mitochondria in our PANC-1 models upon ferroptosis induction. TEM analysis performed after only 6 h of RSL3 treatment revealed that TP53-KO cells exhibited severely compromised mitochondrial integrity, characterized by disrupted cristae architecture, damaged mitochondrial membranes, and marked alterations in overall shape (Fig. 3A). In these cells, mitochondria displayed a reduced aspect ratio, reflecting a shift toward a more rounded morphology (Fig. 3B), together with an overall increase in mitochondrial area (Fig. 3C). This enlargement is consistent with mitochondrial swelling, a feature typically found in cells affected by ferroptosis and not observed in RSL3-treated parental cells. Indeed, PANC-1 mitochondria largely preserved their structural integrity under RSL3 treatment, showing only minor morphological changes and no evidence of swelling (Fig. 3A-C). Interestingly, flow cytometry assessment of mitochondrial membrane potential (Δψm) revealed opposite trends between mutant p53-expressing and TP53-KO cells in response to RSL3-induced ferroptosis. Indeed, while parental PANC-1 cells exhibited rapid mitochondrial hyperpolarization upon RSL3 treatment, TP53-KO cells displayed significant depolarization (Fig. 3D). This divergence is particularly relevant given that mutant p53 PANC-1 cells are generally more sensitive to the mitochondrial uncoupler agent CCCP, a feature typically associated with cells exhibiting a high reliance on OXPHOS [32]. In addition, immunoblot analysis of the mitochondrial respiratory complexes (CI–CV) revealed that, at 8 h post-ferroptosis induction with RSL3, mutant p53 PANC-1 cells upregulate all five complexes (Fig. 3E), further suggesting an active compensatory response to maintain electron transport chain function and mitochondrial bioenergetics under ferroptotic stress. We therefore performed the Seahorse Mitostress assay to evaluate OXPHOS capacity. Confirming our previous observation [20], PANC-1 cells harboring mutant p53 display significantly higher baseline of OXPHOS levels compared to TP53-KO PANC-1 (Fig. 3F). Taking this intrinsic difference into account, we found that under ferroptotic conditions, both cell types exhibited a reduction in key respiratory parameters including basal respiration, maximal respiration, and ATP production but the decline was significantly more pronounced in TP53-KO cells (Fig. 3F). Collectively, these results indicate that mutant p53 is associated with mitochondrial integrity during early ferroptotic stress, contributing to ferroptosis resistance in PDAC cells.

Fig. 3.

Ferroptosis affects mitochondria of PDAC cells in the absence of TP53. A Ultrastructural representative images of mitochondria in TP53^R273H PANC-1 and TP53-KO PANC-1 cells after 6 h treatment with RSL3 (0.5 µM). Bars: 200 nm; B Bar graph indicating the quantification of the mitochondrial Aspect Ratio (n = 25). p-value was determined by 2wayANOVA using Sidak’s multiple comparisons test (****p = < 0.0001); C Bar graph representing the quantification of the mitochondrial Area (n = 25). p-value was determined by 2wayANOVA using Sidak’s multiple comparisons test (****p = < 0.0001); D Flow cytometry evaluation of mitochondrial membrane potential (ΔΨm) in TP53^R273H PANC-1 and TP53-KO PANC-1 cells after treatment with RSL3 (0.5 µM). The bar graph represents the Mean Fluorescence Intensity (MFI) of the cells positive for the TMRM staining (n = 3). p-values were determined by 2wayANOVA using Tukey’s multiple comparisons test (**p = < 0.01, ****p = < 0.0001); E Western blot evaluation of the mitochondrial complexes in TP53^R273H PANC-1 and TP53-KO PANC-1 cells after 8 h treatment with RSL3 (0.5 µM). Vinculin was used as housekeeping control; F Oxygen Consumption Rate (OCR) profiles obtained from TP53^R273H PANC-1 and TP53-KO PANC-1 cells following 4 h treatment with RSL3 (0.5 µM). OCR traces are shown as values normalized to the DNA content (µg) measured in each well. The accompanying bar graphs summarize basal respiration, maximal respiration, and ATP production parameters derived from the Mito Stress Test; these values were normalized to DNA content and expressed as fold change relative to the corresponding control condition (n = 4). Statistical significance was assessed by 2wayANOVA followed by Tukey’s multiple-comparisons test (***p < 0.001, **p < 0.0001)

Mutant p53 orchestrates a complex gene expression pattern preventing ferroptosis in PDAC

To elucidate the molecular mechanisms underlying mutant p53-mediated resistance to ferroptosis, we performed comprehensive RNA sequencing (RNA-seq) profiling on TP53^R273H PANC-1 cells and three independent TP53-KO PANC-1 clones following 6 h of treatment with the three ferroptosis inducers (Erastin, RSL3, and Ferroptocide; Sup. Tables 1–2). This early time point was selected to capture transcriptional changes in TP53-KO viable cells, as later time points would compromise cell integrity and alter the following analysis. Principal Component Analysis (PCA) revealed a clear distinct clustering of parental and TP53-KO PANC-1 cells, as well as between the different treatment groups (Fig. 4A and S3A). Notably, parental PANC-1 cells harbouring mutant p53 displayed a roughly five-fold greater number of differentially expressed transcripts (both upregulated and downregulated) in response to ferroptosis inducers compared to TP53-KO clones, indicating a more robust transcriptional program (Fig. 4B). Among the analysed transcripts, we identified several genes that were consistently upregulated in TP53^R273H-cells (log₂FC > 0.8, p < 0.05) by at least two different ferroptosis inducers and are known to functionally contribute to cellular protection against ferroptosis through complementary mechanisms (Fig. 4C). These genes act through diverse strategies: Ten-Eleven Translocation methylcytosine dioxygenase 1 (TET1) regulates gene expression and control inflammation, helping cells to manage oxidative stress and survive harmful conditions [33, 34]; Nitric Oxide Synthase 3 (NOS3) and Superoxide Dismutase 3 (SOD3) directly modulate oxidative stress by balancing reactive oxygen species and nitric oxide levels, thereby preventing lipid damage that drives ferroptosis [35–37]. Increased levels of Aldehyde Oxidase 1 (AOX1) counteract oxidative stress by detoxifying lipid-derived aldehydes [38, 39], while metabolic regulators such as Pyruvate Dehydrogenase Kinase 4 (PDK4) fine-tune energy pathways and produce antioxidant metabolites that further shield cells from oxidative injury [28, 40, 41]. Alpha-Tocopherol Transfer Protein (TTPA) provides targeted antioxidant defence and modulate inflammatory signaling to reduce susceptibility to ferroptosis [42, 43]. Transcriptional and mitochondrial regulators such as SRY-Box Transcription Factor 13 (SOX13) and Forkhead Box Q1 (FOXQ1) enhance resistance to ferroptosis by sustaining bioenergetics and activating antioxidant pathways that limit lipid peroxidation and intracellular iron accumulation [44, 45]. We confirmed the upregulation of these transcripts in PANC-1 parental cells by real-time quantitative PCR (RT-qPCR) on independent samples, validating the RNA-seq findings and highlighting their coordinated role in anti-ferroptotic defence (Fig. 4D). To investigate the biological pathways associated with the mutant p53-dependent transcriptional response to ferroptotic stress, we performed Over-Representation Analysis (ORA) on up-regulated genes using KEGG pathway annotations (Fig. 4E and S3B-D). Notably, among all enriched pathways, the PI3K-AKT signaling pathway is consistently maintained as the most significantly represented across all three ferroptosis-inducing treatments. Consistent with these transcriptomic findings, protein-level analyses demonstrated that RSL3 treatment activates a compensatory PI3K-AKT signaling response in mutant p53-expressing PANC-1 cells, characterized by increased AKT Ser473 phosphorylation and preservation of downstream mTOR signaling. This adaptive response was lost in p53-deficient cells, which displayed reduced AKT Thr308 and phospho-S6 levels following ferroptotic stress (Fig. 4F and Supplementary Fig. S4A). In addition, flow cytometry analyses demonstrated that pharmacological inhibition of the PI3K-AKT signaling pathway significantly enhanced ferroptosis sensitivity in PANC-1 cells. This effect was consistently observed using four independent inhibitors targeting different nodes of the pathway, including the PI3K inhibitors LY294002, PI-103, and Copanlisib, as well as the AKT inhibitor MK-2206 (Fig. 4G-H and Fig. S4B-C). Notably, MK-2206 potentiated cell death induced not only by RSL3 but also by Erastin and Ferroptocide (Fig. S4D-E), indicating that PI3K–AKT signaling broadly contributes to ferroptosis resistance irrespective of the ferroptosis-inducing stimulus. Together, these findings indicate that mutant p53 promotes a broad, multi-component protective program, encompassing coordinated transcriptional adaptations and activation of PI3K-AKT signaling, thereby contributing to ferroptosis resistance.

Fig. 4.

Mutant p53 modulates a complex gene expression pattern in PANC-1 cells under ferroptosis: A Principal Component Analysis (PCA) plots of TP53^R273H PANC-1 and TP53-KO PANC-1 cells after 6 h treatment with Erastin (5 µM), RSL3 (0.5 µM) and Ferroptocide (5 µM) (n = 3). B Bar charts representing the number of genes up- and down- regulated in TP53^R273H- and TP53 KO PANC-1 cells. C Volcano plots representing the differential gene expression after treatment with Erastin, RSL3 and Ferroptocide (as indicated above) in TP53^R273H PANC-1 compared to TP53 KO PANC-1 cells. Upregulated genes of interest are indicated on the side (arrows). D Relative expression of genes upregulated after ferroptosis induction with Erastin, RSL3 and Ferroptocide (6 h treatment) (n = 3). The results are indicated as Fold Change (F.C.) of the relative expression in TP53^R273H PANC-1 cells compared to TP53-KO cells. E Dotplot showing the most represented pathways upregulated in TP53^R273HPANC-1 cells (compared to TP53-KO PANC-1 cells) after 6 h treatment with RSL3. F Western Blot of p-AKT (Ser473) and total AKT in TP53^R273H PANC-1 and TP53-KO PANC-1 cells after 8 h treatment with RSL3 (0.5 µM). Actin was used as loading control. G-H Flow cytometry analysis of PANC-1 cell viability after 24 h treatment with RSL3 and MK-2206 or Copanlisib. The bar graphs represent the percentage of Annexin V-/PI- (viable) cells (n = 3). p-values were determined by ordinary one-way ANOVA using Tukey’s multiple comparisons test (****p = < 0.0001)

Mutant p53 promotes glycolysis to sustain ATP production under ferroptotic stress

PI3K-AKT signaling is a central regulator of glycolytic metabolism in cancer, promoting glucose uptake and ATP production to support cell survival [46]. The persistent activation of this pathway observed in mutant p53-expressing PANC-1 cells under ferroptotic stress suggests its potential contribution in maintaining ATP levels. To directly examine how cellular metabolism responds to ferroptotic stress, we focused on the metabolic profile of TP53^R273H and TP53-KO PANC-1 cells by measuring glycolytic and mitochondrial ATP production rates under treatment with the ferroptosis inducers RSL3, Erastin, and Ferroptocide. These analyses confirmed the expected metabolic differences between the two genotypes, with TP53^R273H PANC-1 cells displaying a more metabolically active phenotype at baseline, characterized by higher ATP production rates from both glycolytic and mitochondrial pathways, compared to their TP53-KO counterpart. Upon ferroptosis induction, compared to parental cells, TP53-KO cells exhibited a marked reduction in ATP production rate, consistent with an impaired metabolic adaptation to ferroptotic stress. In contrast, mutant p53-expressing cells largely maintained ATP production and showed a trend toward increased glycolytic ATP generation following treatment (Fig. 5A and Fig. S5A-B). To further investigate this phenotype, we performed a glycolysis stress test. In line with the ATP rate assay, RSL3-treated TP53^R273H-PANC-1 cells exhibited higher glycolytic activity, glycolytic capacity, and glycolytic reserve compared to TP53-KO cells (Fig. 5B), corroborating the notion that mutant p53 supports and promotes a glycolytic response under ferroptotic stress. Importantly, this phenotype was recapitulated in MIA PaCa-2 cells (TP53^R248W) (Fig. 5C), but not in PDAC cells with TP53^wt (Hs766T) (Fig. 5D) or in non-tumorigenic human pancreatic ductal epithelial cells (TP53^wt-HPDE) (Fig. 5E), which in contrast displayed a reduction in glycolysis after ferroptosis induction. These data indicate that glycolytic adaptation is a mutant p53-specific feature. To assess the functional impact of metabolic plasticity, parental and TP53-KO PANC-1 cells were treated with ferroptosis inducers in presence of either increasing concentrations of glucose or the glycolytic inhibitor 2-deoxy-D-glucose (2-DG) (0.5-5 mM). In TP53^R273HPANC-1 cells, supplementation with glucose significantly increased survival under ferroptotic stress, whereas inhibition of glycolysis with 2-DG markedly increased cell death (Fig. 5F). In contrast, TP53-KO cells showed no appreciable change in survival under both condition (Fig. 5G). These results demonstrate that glucose metabolism specifically supports ferroptosis resistance in mutant p53-expressing cells, highlighting glycolytic reprogramming as a key adaptive mechanism.

Fig. 5.

Ferroptosis onset leads to a metabolic shift towards glycolysis in PDAC cells with mutant p53. A Energetic map and bar graphs representing the levels of glycoATP, mitoATP and total ATP production rate in TP53^R273H PANC-1 and TP53-KO PANC-1 cells after 4 h treatment with RSL3 (0.5 µM). Data were normalized for the quantity of DNA (µg) quantified for each well. Results are indicated in fold change relative to each control (n = 3). p-values were determined by 2wayANOVA using Tukey’s multiple comparisons test (*p = < 0.05); B ECAR profile and bar graphs representing the levels of glycolysis (the rate of glycolysis measured after glucose addition following glucose starvation), glycolytic reserve (the difference between glycolytic capacity and basal glycolysis, representing the additional capacity available to increase glycolytic flux under stress) and capacity (the maximum glycolytic rate achieved in response to mitochondrial inhibition and increased energetic demand) in TP53^R273H PANC-1 and TP53-KO PANC-1 cells after 4 h treatment with RSL3 (0.5 µM). Data were normalized for the quantity of DNA (µg) quantified for each well. Results are indicated as fold change relative to each control (n = 4). p-values were determined by 2wayANOVA using Tukey’s multiple comparisons test (*p = < 0.05, ***p < 0.001) (C) Energetic maps and bar graphs representing the level of glycoATP, mitoATP and total ATP production rate in MIA PaCa-2 (n = 3), Hs766T (D) and HPDE (E) cells (n = 4) after 6 h treatment with RSL3 (1 µM). Data were normalized for the quantity of DNA (µg) quantified for each well. Results are indicated in fold change relative to each control. p-values were determined by unpaired T test (*p = < 0.05, **p = < 0.01). F-G Bar graphs representing cell viability (MTT assay) in TP53^R273H- and TP53 KO PANC-1 cells after 24 h treatment with RSL3, alone or in combination with glucose or 2-DG. Results are indicated as Fold Change relative to RSL3-treated cells (n = 3). p-values were determined by ordinary one-way ANOVA using Tukey’s multiple comparisons test (*p = < 0.05, **p = < 0.01, ****p = < 0.0001)

Restoring wild type p53 functions to enhance ferroptosis-based therapies in PDAC

Our findings provide evidence that mutant p53 acts as a central coordinator of a robust, multi-layered transcriptional and metabolic network that enables PDAC cells to withstand ferroptotic stress. By simultaneously preserving mitochondrial integrity, sustaining bioenergetic flexibility, and activating complementary antioxidant and metabolic defence programs, mutant p53 establishes a highly redundant ferroptosis resistance architecture. This functional redundancy suggests that targeting individual downstream effectors is unlikely to effectively dismantle ferroptosis resistance. To translate these mechanistic insights into a therapeutic strategy for advanced PDAC, we therefore shifted our focus upstream to mutant p53 itself. Specifically, we evaluated the effect of the pharmacological reactivation of mutant-p53 using APR-246 (Eprenetapopt), a first-in-class compound capable of restoring wild-type p53 conformation and neutralizing mutant p53 oncogenic functions [47]. We found that co-treatment with ferroptosis inducers, Erastin, RSL3 or IKE, and APR-246 strongly sensitized PANC-1 and MIA PaCa-2 cells to ferroptosis, suggesting that targeting mutant p53 can mitigate its anti-ferroptotic effects (Fig. 6A and S6A-D). Notably, this combinatorial approach significantly reduced the effective doses of both APR-246 and the ferroptosis inducers, enhancing the translational potential of this strategy and making it more clinically relevant. Mechanistically, this sensitization was associated with a failure to activate the glycolytic response previously observed, highlighting the critical role of this compensatory metabolic adaptation in supporting cell survival under ferroptotic stress (Fig. 6B and S6E-F). To validate these findings and translate them in vivo, we first introduced a new murine PDAC model, the 7940b KPC cell line harbouring TP53^R172H, derived from a highly aggressive pancreatic tumor and characterized by rapid growth, typically forming big masses in less than one month [48, 49]. Functionally, 7940b cells reproduced the strong synergistic response to Erastin in combination with APR-246 previously observed in the human system, further supporting the conservation of the mutant p53-driven anti-ferroptotic program across species (Fig. S6G). For in vivo experiments, we selected IKE as ferroptosis inducer, given its superior metabolic stability relative to Erastin and its widespread use in the literature, for eliciting ferroptosis in animal models [50–52]. In vitro, 7940b cells exhibited ferroptotic cell death in response to IKE, which was effectively blocked by the ferroptosis inhibitors Ferrostatin-1 and Liproxstatin-1 (Fig. 6C), but was markedly potentiated when combined with APR-246 (Fig. 6D). The potentiating effect of APR-246 was also observed with Erastin in 7940b cells (S6G). Synergism between APR-246 and IKE was quantitatively assessed by CompuSyn software analysis, which identified strong synergistic interactions at effect levels below 50% cell viability as compared to untreated control (Fig. 6E-F). Consistent with these observations, a second murine PDAC line derived from KPC tumors, MT3, exhibited the same pattern of sensitivity and combinatorial potentiation with both Erastin (Fig. S6H) and IKE (Fig. S6I-K), thus reinforcing the reproducibility of our mechanistic model. On the basis of these results, we proceeded with in vivo validation using an orthotopic model in immunocompetent C57BL/6 mice (Fig. 6G). 7940b cells were orthotopically implanted and, 14 days post-implantation, the presence of localized tumors smaller than 100 mm³ was confirmed via ultrasounds, indicating early yet established disease. At this stage, animals were randomized into 4 treatment groups to evaluate the effect of single agents and the APR-246 + IKE combination within a physiologically relevant tumor microenvironment. Mice received intraperitoneal (ip) administration of APR-246 (40 mg/kg), IKE (40 mg/kg), or their combination over four treatment cycles. On day 20 post-implantation, ultrasound imaging was performed to monitor tumor growth. Notably, only tumors in mice treated with the combination therapy showed a significant reduction in growth compared to those receiving either single agent or vehicle, highlighting the enhanced therapeutic efficacy of co-targeting mutant p53 and inducing ferroptosis in vivo (Fig. 6H).

Fig. 6.

Wild type p53 restoration counteracts the ferroptosis-induced glycolytic shift both in vitro and in vivo. A Flow cytometry evaluation of PANC-1 cell viability after 24 h treatment with Erastin and APR-246. The bar graphs represent the percentage of Annexin V-/PI- (viable) cells (n = 3). p-values were determined by ordinary one-way ANOVA using Tukey’s multiple comparisons test (****p = < 0.0001); B Energetic map and bar graphs representing the levels of glycoATP, mitoATP and total ATP production rate in PANC-1 cells after 6 h treatment with Erastin (5 µM) and APR-246 (25 µM). The results are indicated in Fold Change relative to the control (n = 3). p-values were determined by ordinary one-way ANOVA using Tukey’s multiple comparisons test (*p = < 0.05, **p = < 0.01); C Bar graphs representing the percentage of 7940b viable cells (MTT assay) after 24 h treatment with IKE with or without the ferroptosis inhibitors Ferrostatin-1 and Liproxstatin-1 (5 µM - n = 3). p-values were determined by ordinary one-way ANOVA using Tukey’s multiple comparisons test (*p = < 0.05, ***p = < 0.001, ****p = < 0.0001); D Flow cytometry evaluation of cell viability of 7940b cells after 24 h treatment with IKE and APR-246. The bar graphs represent the percentage of Annexin V-/PI- (viable) cells (n = 3). p-values were determined by ordinary one-way ANOVA using Tukey’s multiple comparisons test (****p = < 0.0001); E Cell viability curve (MTT assay) of 7940b cells after 24 h treatment with IKE in combination with APR-246 (ratio 1:5) (n = 3). F Dotplot representing the Combination index (CI)–fraction affected (Fa) values of 7940b cells after treatment with IKE and APR-246 as indicated above (CI < 0.5 = strong synergism). G Schematic representation of the experimental setting used for the in vivo experiment; H Line graph +/-SEM representing the tumor size (mm³) detected in mice treated with IKE (40 mg/kg) alone or in combination with APR-246 (40 mg/kg) (n = 14/group) detected on day 20, after 4 rounds of treatment. p-values were determined by ordinary one-way ANOVA using Tukey’s multiple comparisons test (**p = < 0.01); I Energetic map and bar graphs representing the level of glycoATP, mitoATP and total ATP production rate of cells collected from 5 different tumors (n= 5/group); J ECAR profile and bar graphs showing the level of glycolysis (rate of glycolytic process after glucose starvation), glycolytic reserve (capacity available to utilize glycolysis beyond the basal rate) and capacity (glycolytic response to energetic demand from stress) of cells collected from 5 different tumors (n = 5/group). Data were normalized for the quantity of DNA (µg) quantified for each well. Results are indicated in fold change relative to the DMSO-treated mice (n = 5)

Importantly, the antitumor activity of the combination treatment was not associated with overt signs of systemic toxicity. Body weight measurements collected at the experimental endpoint revealed no significant differences among treatment groups, and treated animals did not exhibit clinical signs of cachexia (Fig. S7A). Moreover, serum analyses performed at sacrifice showed no significant alterations in alanine aminotransferase (ALT) activity or creatinine levels, indicating the absence of detectable hepatic or renal toxicity following treatment with APR-246 and IKE (Fig. S7B-C). Together, these findings suggest that the therapeutic benefit of the combination regimen is achieved within a favorable preliminary tolerability window. Following two additional treatment cycles, control mice began to exhibit signs of distress. We therefore sacrificed the animals, harvested tumors, and prepared tumor lysates for metabolic profiling to investigate how combination therapy affects tumor bioenergetics. Both ATP production rate assays and glycolysis stress tests on these tumor lysates revealed a ferroptosis-induced glycolytic shift in tumors treated with IKE alone, which was completely reversed by co-treatment with APR-246, effectively abolishing the mutant-p53 metabolic reprogramming underlying ferroptosis resistance (Fig. 6I-J). To further validate ferroptosis induction in vivo, we analysed 4-HNE accumulation in tumor lysates collected at sacrifice. Tumors treated with the combination of IKE and APR-246 exhibited significantly higher levels of 4-HNE compared with vehicle- or single-agent-treated tumors (Fig. S7D), indicating enhanced lipid peroxidation in vivo. These findings further support the ability of APR-246 to potentiate ferroptosis induction in mutant p53-driven PDAC tumors.

Together, our data identify mutant p53-driven metabolic plasticity as a key determinant of ferroptosis resistance in PDAC. Reactivating wild-type p53 functions collapses this adaptive state and restores ferroptotic vulnerability in vitro and in vivo, uncovering a therapeutically actionable dependency in pancreatic cancer (Fig. 7).

Fig. 7.

Schematic representation of the metabolic adaptation model: Graphical abstract illustrating the proposed model of mutant p53-driven metabolic adaptation and its contribution to ferroptosis resistance in PDAC. Under ferroptotic stimulation by ferroptosis inducers (FINs), mutant p53-expressing cells (top left) exhibit a limited oxidative stress response together with activation of the PI3K-AKT pathway, leading to enhanced glycolysis and increased resistance to ferroptotic cell death. In contrast, TP53-KO cells (top right) display greater sensitivity to ferroptotic stimuli, characterized by increased lipid peroxidation, elevated oxidative stress, and marked mitochondrial dysfunction. Restoration of wild-type p53 activity in mutant p53-expressing cells through treatment with the mutant p53 reactivator APR-246 (bottom) reduces glycolytic activity, thereby sensitizing cells to ferroptosis. These findings support a model in which mutant p53 promotes metabolic rewiring that favors ferroptosis resistance in PDAC

Discussion

PDAC exemplifies the challenges of modern oncology, with mortality exceeding that of many other solid tumors due to late diagnosis, aggressive local invasion, and multifactorial therapy resistance. Despite advances in immunotherapy and targeted therapies, median survival in patients with metastatic disease remains below 12 months, underscoring the urgent need to identify novel therapeutic vulnerabilities [53]. In this context, ferroptosis emerges as a promising paradigm to overcome these barriers bypassing apoptosis, which is frequently impaired in PDAC. Recent studies have shown that ferroptosis inducers, such as Erastin and RSL3, exert selective cytotoxic effects on PDAC cells, both in vitro and in xenograft models, suggesting immediate clinical potential [14]. However, the translational success of these strategies is hindered by robust intrinsic resistance mediated by prevalent genetic alterations, including TP53 mutations. TP53 mutations not only abrogate wild-type tumor-suppressive functions, but also confer GOF pro-tumorigenic activities that remodel cellular metabolism, promote survival, and drive chemoresistance. In non-PDAC contexts, mutant p53 has been implicated in regulating ferroptosis via mechanisms such as NRF2 and SLC7A11 upregulation, which detoxify lipid peroxides [54, 55]. Yet, the specific role of PDAC-associated hotspot TP53 variants in modulating ferroptosis sensitivity remains poorly defined. Indeed, to date the only prior evidence for an anti-ferroptotic function of mutant p53 comes from triple-negative breast cancer, where genetic ablation of hotspot mutation induced tumor ferroptosis through loss of NRF2-mediated transcriptional control of lipid peroxide detoxifying enzymes [19]. In addition to TP53 alterations, PDAC progression is driven by recurrent mutations in KRAS, CDKN2A, and SMAD4, which may also influence cellular stress responses and therapeutic adaptation [56]. Among these, KRAS has been implicated in ferroptosis regulation through metabolic rewiring and modulation of redox-associated pathways [57, 58], whereas the contribution of CDKN2A and SMAD4 to ferroptosis remains less clearly defined. Although additional PDAC drivers may participate in shaping ferroptotic susceptibility, our findings support a direct role for mutant p53 in promoting ferroptosis resistance in PDAC cells.

Our study fills this gap, demonstrating for the first time that mutant p53 establishes a multifactorial barrier to ferroptosis in PDAC cells, integrating mitochondrial preservation, glycolytic metabolic adaptation, and antioxidant/stress-response pathway activation. Using isogenic CRISPR-Cas9 knockout models, mutant p53 overexpression systems, and integrated transcriptomic and metabolic analyses, we demonstrate that mutant p53 coordinates a multilayered protective network that maintains mitochondrial integrity, preserves bioenergetic flexibility, and engages complementary antioxidant and stress-response pathways. These mechanisms collectively buffer oxidative stress and lipid peroxidation, enabling survival under ferroptotic insults. A key component of this adaptive network is the PI3K-AKT signaling pathway. Consistent with its established role in regulating glycolysis and survival in cancer, PI3K-AKT is selectively hyperactivated in mutant p53-expressing PDAC cells under ferroptotic stress. This activation preferentially increases glycolytic ATP production, thereby maintaining energy homeostasis when mitochondrial respiration is compromised. In contrast, TP53-KO cells experience a collapse of both glycolytic and mitochondrial ATP production, underscoring the essential role of metabolic plasticity in ferroptosis resistance. Functional manipulation of glucose availability further highlights this dependency: glucose supplementation enhances survival in mutant p53 cells, whereas glycolysis inhibition with 2-deoxy-D-glucose accelerates ferroptotic cell death. These results define the glycolytic shift as a critical adaptive mechanism downstream of mutant p53, directly linking metabolic reprogramming to ferroptosis resistance.

Our data extend prior observations focused on wild-type p53, demonstrating its role in promoting ferroptotic sensitivity via SLC7A11 downregulation and context-dependent enhancement of lipid peroxidation, positioning mutant p53 as pro-survival driver [17]. From a translational perspective, our findings highlight the challenges of targeting ferroptosis resistance in PDAC: the redundancy and robustness of mutant p53-mediated protective programs suggest that single-target interventions may prove insufficient. In contrast, pharmacological reactivation of mutant p53 with APR-246, which restores a wild-type-like conformation and transcriptional activity, not only blocks mutant p53’s anti-ferroptotic barrier, but also exploits restored wild-type p53 pro-ferroptotic functions, abrogating glycolytic adaptation, and potently sensitizing cells to ferroptosis both in vitro and in vivo. This dual-action combinatorial strategy (suppressing anti-ferroptotic GOF while unleashing wild-type pro-ferroptotic activity) demonstrates the feasibility of co-targeting mutant p53 and ferroptosis to overcome intrinsic resistance, potentially enabling lower doses of ferroptosis inducers and broadening clinical applicability.

Conclusions

In summary, our study identifies mutant p53-driven metabolic and transcriptional plasticity as a central barrier to ferroptosis in PDAC. By linking mitochondrial preservation, transcriptional control, and glycolytic reprogramming through PI3K-AKT signaling, we provide a mechanistic framework for understanding ferroptosis resistance in this aggressive cancer. Importantly, pharmacological reactivation of mutant p53 with APR-246, which restores wild-type-like p53 activity, disrupts this adaptive network, thereby restoring ferroptotic vulnerability and offering a therapeutically actionable strategy. These findings underscore the potential of co-targeting mutant p53 and ferroptosis as a precision-based approach to overcome intrinsic resistance and improve treatment outcomes in PDAC.

Materials and methods

Cell cultures

The PDAC human cell lines PANC-1 (TP53^R273H), AsPC-1 (p53-null), MIAPaca-2 (TP53^R248W), and Hs766T (TP53^wt) were purchased from the American Type Culture Collection (ATCC). The Human Pancreatic Duct Epithelial Cell Line (HPDE) was kindly provided by Prof. Aldo Scarpa (University of Verona, Italy). The mouse KPC-derived cell lines 7940b and MT3 were kindly provided by D. Tuveson (Cold Spring Harbor Laboratory) and G.L. Beatty (University of Pennsylvania) via Material Transfer Agreement (MTA). For routine propagation, unless otherwise indicated, all cell lines were cultured in RPMI 1640 (Gibco, 52400025) supplemented with 10% FBS (Gibco, A5256701) and 50 µg/mL Gentamicin (Gibco, 15750037). Cells were kept in a humified incubator at 37 °C with 5% CO₂. Phosphate-Buffered Saline (PBS) (Gibco, 14190094) and trypsin-EDTA (Gibco, 25200056) were used for cell washing and cell detachment during maintenance.

TP53 CRISPR/Cas9 knock-out (KO) cell lines

TP53-KO PANC-1 and TP53-KO MIA PaCa-2 cells were generated in our laboratory using the CRISPR/Cas-9 technology (CRISPR KN 2.0 Gene Knockout Kit, Origene Technologies, Rockville, MD, USA) as previously described [20]. Unless otherwise indicated, three independent TP53 KO clones were used in all experiments. The absence of TP53 expression in these KO cells was periodically verified by Western blot analysis.

Drugs and chemicals

The drugs used in this study are listed in Table 1:

Table 1.

List of compounds

Name	Company	Formula	Cat. Number
Erastin	Med Chem Express (New Jersey, USA)	C30H31ClN4O4	HY-15,763
RSL3	Med Chem Express (New Jersey, USA)	C23H21ClN2O5	HY-100,218 A
Ferroptocide	Tokyo Chemical Industry (Tokyo, Japan)	C30H36ClN3O7	F1239
Ferrostatin-1	Med Chem Express (New Jersey, USA)	C15H22N2O2	HY-100,579
Liproxstatin-1	Cayman chemical company (Michigan, USA)	C19H21ClN4	17,730
Z-VAD fmk	Selleckchem (Houston, USA)	C22H30FN3O7	S7023
Bafilomycin A1	Selleckchem (Houston, USA)	C35H58O9	S1413
Necrostatin-1	Selleckchem (Houston, USA)	C13H13N3OS	S8037
APR-246	Selleckchem (Houston, USA)	C10H17NO3	S7724
IKE	Selleckchem (Houston, USA)	C35H35ClN6O5	S8877
LY294002	Cell Signaling Technologies, Massachusetts, USA	C19H17NO3	9901
PI 103	Cayman chemical company (Michigan, USA)	C19H16N4O3	10,009,209
2-DG	Sigma-Aldrich	C6H12O5	D8375
Glucose	Agilent Technologies	C6H12O6	103577-100
MK-2206 Dihydrocloride	Selleckchem (Houston, USA)	C25H21N5O.2HCl	S1078
Copanlisib	Selleckchem (Houston, USA)	C23H28N8O4	S2802
Gemcitabine (hydrocloride)	Cayman chemical company (Michigan, USA)	C9H11F2N3O4⋅HCl	CAY-9,003,096

For cell treatments, all drugs were dissolved in DMSO (stock solutions: 10 mM, except for IKE, LY294002 and APR-246, which were prepared at 50 mM). Copanlisib was dissolved in 5% TFA (stock: 10 mM), while Gemcitabine in PBS (stock: 33.3 M).

Overexpression of mutant TP53 via transient plasmid transfection

For the transient expression of different mutant-TP53 isoforms, p53-null human PDAC AsPC-1 cells were transfected with the following plasmids: pcDNA3-mutp53R273H, pcDNA3-mutp53R175H, pcDNA3-mutp53R282W, pcDNA3-mutp53R248H, pcDNA3-p53R248Q, pcDNA3-p53Y220C (GenScript - SC1200M), pcDNA-TP53 (GenScript – SC1200) and the mock plasmid (GenScript - SC2092). In detail, 4 × 10⁵ AsPC-1 cells/well were seeded in 6-wells plates. The following day, cells were transiently transfected using Lipofectamine 3000 Reagent (Invitrogen - L30000015) according to the manufacturer’s instructions. For cell transfection, Opti-MEM Reduced Serum Medium (Gibco − 31985-062) was used and the different plasmid DNAs were added at a concentration of 5 µg/mL (for mutant p53) and 2.5–1 µg/mL (for wild-type p53). In these experiments, the mock empty vector was used as control. After 48 h transfection, TP53 overexpression was confirmed by Western Blot analysis. Vinculin was used as loading control.

CellTOX green cytotoxicity assay

TP53^R273H PANC-1 and TP53-KO PANC-1 cells were seeded in 96-well plates (1 × 10⁴ cells/well). After overnight adhesion, cells were treated with the ferroptosis inducers in presence of the CellTox™ Viability Dye (diluted in cell culture media 1:4.000). Cell images were captured using an APX100 microscope (Evident–Olympus), considering a time-point of 4.5 h for RSL3 (0.5 µM) and 6 h for Erastin and Ferroptocide (5 µM).

MTT cell viability assay

MTT Cell staining by Invitrogen (M6494) was used for cell viability evaluation. PANC-1 and TP53-KO PANC-1 cells were seeded at 1 × 10⁴ cells/well, AsPC-1 cells were seeded at 1.3 × 10⁴ cells/well, while 7940b and MT3 cells were seeded at 7 × 10⁴ cells/well in 96-well plates. The following day, cells were treated with the compounds of interest. After treatment, cells were incubated with MTT solution (5.5 mg/ml in pre-warmed PBS) for 2.5 h at 37° with 5% CO₂. Then, medium was removed and cells were lysed by adding 25 µl RPMI + 50 µl DMSO in each well, according to the manufacturer’s instructions. Once the cells were lysed, absorbance was measured at 540 nm by using a microplate reader, Tecan Infinite^® M Nano Plus microplate reader (Tecan Group Ltd., Zürich, Switzerland). The results are expressed as the percentage of viable cells relative to the control and represent the mean ± standard deviation (SD) of at least three independent replicates.

Annexin V/PI double staining for cell viability

Cell viability was assessed by flow cytometry with the eBioscience Annexin V Apoptosis Detection Kit APC (Invitrogen, 88800774). Cells were seeded in triplicate in 24-well plates at the following densities: 6 × 10⁴ cells/well for TP53^R273HPANC-1, TP53-KO PANC-1, and AsPC-1 cells, and 4 × 10⁴ cells/well for MIA PaCa-2, 7940b and MT3 cells. After overnight cell adhesion, cells were treated as indicated for 24 h. Then, cells were collected after trypsinization and stained according to the manufacturer’s protocol. Flow cytometry analysis was performed using a BD LSR-Fortessa (BD Bioscience) flow cytometer. The population of Annexin V-/PI- (double negative) cells was considered viable. The results are presented as mean ± SD from at least three independent replicates.

CellROX deep red staining for ROS detection

PANC-1, TP53-KO PANC-1 cells and AsPC-1 cells were seeded in triplicate in 24-well plate (6 × 10⁴ cells/well). The following day, cells were treated as indicated. At the end of the treatment, cells were stained with the CellROX Deep Red (Invitrogen, C10422) according to the manufacturer’s instructions. Then, cells were collected and incubated for 3 min with the SYTOX AADvanced probe (Invitrogen, S10349) (0.5 µl of SYTOX in 50 µl of flow buffer – 2% FBS, 3 mM EDTA, PBS). Finally, 100 µl of flow buffer were added and the samples were analysed by flow cytometry using a BD LSRFortessa flow cytometer (BD Biosciences). The population of SYTOX- cells was considered viable and used for evaluating ROS levels. The results are presented as mean ± SD from at least three independent replicates.

Ki-67 staining for cell proliferation

PANC-1 and TP53-KO PANC-1 cells (3.5 × 10⁵ cells/well) were seeded in triplicate in 6 wells plates. The following day, cells were treated as indicated and, after 24 h, cell pellets were collected. On the day of the assay, samples were processed using the eBioscience™ Foxp3 / Transcription Factor Staining Buffer Set (eBioscience™ - cat. # 00-5523-00) according to the manufacturer’s instructions. Cell viability was evaluated using the eBioscience™ Fixable Viability Dye eFluor™ 780 (0.2 µl in 50 µl FACS buffer/sample. Invitrogen, cat. # 65-0865-15), and samples were incubated for 30 min at 4 °C. Then, Ki-67 positivity was assessed using the Brilliant Violet 421™ anti-human Ki-67 Antibody (1 µl/sample. Biolegend – cat. # 350506), and cells were kept at room temperature for 30 min. Samples were analysed by flow cytometry using a BD LSRFortessa flow cytometer (BD Biosciences), and population of viable cells was considered to assess Ki-67 positivity. Results are presented as mean of mean of fluorescence intensity (MFI) ± SD from three independent replicates.

Cysteine starvation

PANC-1 and TP53-KO PANC-1 cells (4 × 10⁴ cells/well) were seeded in 24-wells plates. The following day, cells were washed twice with PBS. Then, cysteine-free media (RPMI-1640 – MP Biomedicals, cat. # 165464) supplemented with 10% dialyzed fetal bovine serum (Gibco, Cat# A33820-01), 2 mM L-glutamine (Agilent #103579-100) and L-methionine (15 mg/L. MP Biomedicals, cat #102291). L-cysteine (10 mg/L Sigma Aldrich – cat. #C7352) was added only in the cysteine-supplemented conditions. After 24 h, cells were collected for Annexin V/PI staining. Flow cytometry analysis was performed using a BD LSR-Fortessa (BD Bioscience) flow cytometer. The population of Annexin V-/PI- (double negative) cells was considered viable. The results are presented as mean ± SD from three independent replicates.

C11-bodipy staining for lipid peroxidation

24 h after plasmid transfection, AsPC-1 cells were seeded in triplicate in 24-well plates (6 × 10⁴ cells/well). The following day, cells were treated for 24 h with the ferroptosis inducers. At the end of the treatment, cells were stained with the Lipid Peroxidation Assay Kit (Abcam, ab243377) according to the manufacturer’s protocol. Samples were analysed by flow cytometry using a BD LSR Fortessa flow cytometer (BD Biosciences). Lipid peroxidation was expressed as C11-Bodipy, a probe that shifts its fluorescence emission from red (PE channel) to green (FITC channel), that was quantified by calculating the ratio between the fluorescence intensity of the FITC channel (oxidized probe) and the PE channel (non-oxidized probe). For each condition (mock and R273H), results are presented as Fold Change (F.C.) related to the control ± SD from three independent replicates.

TMRM staining for mitochondrial membrane potential evaluation

PANC-1 and TP53-KO PANC-1 cells were seeded in triplicate in 24-well plate (6 × 10⁴ cells/well). The following day, cells were treated with 0.5 µM of RSL3 for 4 and 6 h or 5 min of Carbonyl cyanide m-chlorophenyl hydrazone (50 µM CCCP), which serves as a mitochondrial uncoupler of oxidative phosphorylation. At the end of the treatment, cells were stained with the MitoProbe TMRM (Invitrogen, M20036) according to the manufacturer’s protocol. Cells were then collected after trypsinization and resuspended in 100 µl of flow buffer. Samples were analysed by flow cytometry using a BD LSR Fortessa flow cytometer (BD Biosciences). The results are presented as mean ± SD from three independent replicates.

Transmission Electron Microscopy (TEM)

PANC-1 and TP53 KO-PANC-1 were seeded in 6-wells plates (3.5 × 10⁵ cells/well). The following day, cells were treated with 0.5 µM of RSL3 for 6 hours. Then, 1 × 10⁶ cells were counted and collected for further processing. Cell pellets were fixed in a solution of 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, at 4 °C for 2 h. Cell pellets were then rinsed in PBS, postfixed in 1% OsO4 and 1,5% potassium ferrocyanide at 4 °C for 1 h, dehydrated with acetone, and embedded in Epon 812 resin. Seventy-thick ultrathin sections were collected on a Formvar-carbon-coated cooper grids, stained with lead citrate for 1 min and observed with a Philips Morgagni transmission electron microscope operating at 80 kV and equipped with a Megaview III camera for digital image acquisition. Micrographs (×18,000) of 25 randomly selected mitochondria from 20 different cells were taken for each sample. The mitochondrial sectional area and the lengths of the major and the minor mitochondrial axes were measured. The aspect ratio parameter (as an index of mitochondrial elongation) was calculated as the ratio of the two lengths. The measurements were performed by using ImageJ software (NIH Image, Bethesda, MD, USA).

Protein sample preparation and western blot analysis

Cells were lysed in RIPA buffer (150 mM NaCl, 50 mM Tris-HCl at pH 8.1% IGEPAL, 0.50% Na Doc, 0.10% SDS, at pH 8) supplemented with protease and phosphatase inhibitor cocktails (Thermo Scientific 78420). Lysates were clarified by centrifugation at 10,000 x g for 10 min at 4 °C, and protein concentrations were determined using the Coomassie reagent (Thermo Scientific 1856209) according to manufacturer’s instructions. Equal amounts of proteins were separated by SDS-polyacrylamide gels and transferred onto PVDF membrane (Thermo Scientific – AC40447). Red ponceau staining (Sigma-Aldrich, P7170) was performed to verify the protein transfer on the membrane. Membranes were blocked with 5% non-fat dry milk or 3% BSA (bovine serum albumin) in TBS-T (20 mM Tris-HCl, pH 7.6, 150 mM NaCl, 0.1% Tween-20) for 1 h at room temperature, and then incubated with primary antibodies in 5% non-fat dry milk or 3% BSA in TBS-T overnight at 4 °C with gentle agitation. After three 10-min TBS-T washing, membranes were probed with appropriate HRP-conjugated secondary antibodies (1:2000) for 1 h at room temperature. Protein bands were detected using Chemiluminescence Western Blot reagents, Pierce™ ECL Western Blotting Substrate (1:1 of luminol and peroxide solution) (Thermo Fisher Scientific Inc., Waltham, MA USA, 32106) and imaged on a Chemi Doc Image System (Bio-Rad^®). Band intensities were quantified by densitometry after background subtraction, normalized to loading controls, and analysed using Image J software.

A list of the primary and secondary antibody (primary diluted 1:1000, secondary diluted 1:2000) used in this study is indicated in Table 2:

Table 2.

List of primary and secondary antibodies used for western blot

Antibody	Company	Cat. Number	Clone number
β- ACTIN rabbit monoclonal antibody	ABclonal, Germany	AC026	ARC5115-01
AKT (pan) mouse monoclonal antibody	Cell Signaling Technologies, Massachusetts, USA	2920	40D4
phospho AKT (Ser473) rabbit monoclonal antibody	Cell Signaling Technologies, Massachusetts, USA	4060	D9E
phospho AKT (Thr308) rabbit monoclonal antibody	Cell Signaling Technologies, Massachusetts, USA	4056	244F9
Phospho-p70 S6 Kinase (Thr389) rabbit monoclonal Antibody	Cell Signaling Technologies, Massachusetts, USA	9234	108D2
4-HNE rabbit polyclonal antibody	AbClonal	A24456
P53 mouse monoclonal antibody	Santa Cruz Biotechnology	sc-126	DO1
Total OXPHOS antibody cocktail	Abcam	110,413
Vinculin Rabbit monoclonal antibody	Cell Signaling Technologies, Massachusetts, USA	13,901 S	E1E9V
Anti-rabbit IgG-HRP secondary antibody	Cell Signaling Technologies, Massachusetts, USA.	7074
Anti-mouse IgG-HRP secondary antibody	Cell Signaling Technologies, Massachusetts, USA	7076

RNA samples preparation

PANC-1 and TP53-KO PANC-1 cells were seeded in 6-well plates (3.5 × 10⁵ cells/well). After overnight adhesion, cells were treated for 6 h with the ferroptosis inducers. Then, total RNA was extracted using the RNeasy Mini Kit (Qiagen, Venlo, Netherland, Cat. No. 74104) according to the manufactures’ instructions. The RNA was eluted in RNAse-free water and stored at – 80° C until quantification. The concentration and purity of the RNA were measured using a Nano Drop spectrophotometer (Thermo Fisher Scientific) by evaluating the absorbance at 260 nm and the 260/280 ratio. Only high-quality RNA samples (260/280 ratio between 1.8 and 2.1) were used for downstream applications, including RNA-seq analysis and further RT-qPCR validation analyses.

RNA-sequencing analysis

Total RNA was extracted from 24 samples as described above, and RNA integrity was evaluated using the Fragment Analyzer (Agilent Technologies). RNA-seq library preparation was performed starting from 1,000 ng high-quality total RNA using the “TruSeq Stranded mRNA library prep kit” (Illumina, San Diego, USA). Briefly, the mRNA fraction was purified from total RNA by polyA capture, fragmented and subjected to cDNA synthesis. Barcoded DNA adapters were ligated to both ends of the double-stranded cDNA and subjected to PCR amplification. The library products were evaluated using Fragment Analyzer (Agilent Technologies), then sequenced on an Illumina NextSeq2000 sequencer using 75 bp single-end reads, generating 17 million reads per sample. After quality filtering according to DRAGEN RNA pipeline v3.10.12, reads were aligned to human reference genome GRCh38 (hg38) using an alt-aware reference bundle, and a raw, unnormalized gene-level read count matrix was generated and aggregated across all samples. For each treatment (DMSO/CTRL, Erastin, RSL3, Ferroptocide), differential gene expression analysis was performed by comparing TP53-KO PANC-1 cells with the corresponding PANC-1 control condition, using three biological replicates per group. Differential expression analysis was carried out in R using DESeq2 [59]. Genes with low expression were filtered by retaining those with at least 10 reads in 50% of the samples. Count data were normalized using the median-of-ratios method, dispersion estimates were obtained according to the DESeq2 pipeline, and p values were adjusted for multiple testing using the Benjamini–Hochberg method. For data exploration and visualization, normalized counts were transformed using Variance Stabilizing Transformation (VST). Principal Component Analysis (PCA) was performed on the 1,000 most variable genes.

Over-representation analysis (ORA)

Over-Representation Analysis (ORA) was performed using clusterProfiler (KEGG) [60]. Differentially expressed genes were selected using adjusted p value ≤ 0.05 and |log2 fold change| ≥ 1, and were separated into upregulated and downregulated gene sets. ORA was performed for KEGG pathways analysis, using Benjamini–Hochberg correction, with a significance threshold of adjusted p value ≤ 0.05.

TCGA-PAAD analysis

To evaluate the clinical translatability of the ferroptosis/redox-related phenotype, we analysed publicly available pancreatic ductal adenocarcinoma data from The Cancer Genome Atlas pancreatic adenocarcinoma cohort (TCGA-PAAD). RNA-seq gene expression data, somatic mutation data, and clinical annotations were retrieved from the Genomic Data Commons using the TCGAbiolinks R package. Patients were classified according to TP53 mutational status using somatic mutation data. Patients carrying non-synonymous TP53 alterations, including missense, nonsense, frameshift, in-frame, splice-site, nonstop, or translation-start-site mutations, were classified as TP53-mutated. All remaining patients were classified as TP53 wild-type. Expression values for GPX4, SLC7A11, TXN, and TP53 were extracted from the RNA-seq dataset and analysed as log2(TPM + 1) values. To summarize the anti-ferroptotic/redox-related signal, two exploratory composite scores were generated. An anti-ferroptotic score was calculated as the average of the z-scored log2(TPM + 1) expression values of GPX4 and SLC7A11. An extended redox/ferroptosis score was calculated as the average of the z-scored log2(TPM + 1) expression values of GPX4, SLC7A11, and TXN. Associations between TP53 mutation status and marker expression or score values were evaluated using Wilcoxon rank-sum tests. All analyses were performed with R software.

Gemcitabine treatment in PANC-1 cells

PANC-1 and TP53-KO PANC-1 cells were seeded in 24-wells plates (5 × 10⁴ cells/well). After overnight adhesion, cells were treated with Gemcitabine at 0.25, 0.5, 0.75 and 1 µM. After 5 days of treatment, cells were collected and analysed with Annexin V/PI staining. Flow cytometry analysis was performed using a BD LSR-Fortessa (BD Bioscience) flow cytometer. The population of Annexin V-/PI- (double negative) cells was considered viable. The results are presented as mean ± SD from three independent replicates.

Gemcitabine treatment and ferroptosis in PDAC

Publicly available processed RNA-seq data from GSE249302 and GSE140077 were downloaded from GEO. GSE249302 was used to evaluate the transcriptional effect of gemcitabine treatment in PDAC cell lines, comparing gemcitabine-treated versus control BxPC3 and PANC-1 cells. GSE140077 was used to evaluate transcriptional features associated with acquired gemcitabine resistance, comparing gemcitabine-resistant derivatives with their corresponding parental PDAC cell lines, namely BXPC-3-GR versus BXPC-3 and CFPAC-1-GR versus CFPAC-1. Gene-level FPKM values were transformed as log2(FPKM + 1). Expression of GPX4, SLC7A11 and TXN was analysed separately within each cell line/model. To summarize ferroptosis/redox-related features, an exploratory composite redox/ferroptosis score was calculated at the sample level as the mean of z-scored GPX4, SLC7A11 and TXN expression. Z-score normalization was performed within each cell-line comparison. Differences between treated/resistant and control/parental cells were assessed using two-sided t-tests on log2-transformed expression values or score values. Nominal p-values were reported for these predefined marker-based analyses, together with Benjamini–Hochberg adjusted p-values as a more conservative reference.

Seahorse metabolic assays and DNA normalization

PANC-1, TP53-KO PANC-1 (2 × 10⁴ cells/well), Hs766T (1.5 × 10⁴ cells/well), HPDE (2.5 × 10⁴ cells/well), and MIAPaca-2 cells (3 × 10⁴ cells/well) were seeded in Seahorse XF 24-well cell culture microplates previously coated with poly-D Lysine (100 µg/mL; Thermo Fisher Scientific, A3890401), depending on the cell line. The following day, cells were treated with the ferroptosis inducers, and metabolic analyses were performed using the Seahorse XFe24 Analyzer (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer’s protocols.

For the Mito Stress Test, PANC-1 and TP53-KO PANC-1, cells were treated with RSL3 for 4 h prior to the assay performed using the Seahorse XF Cell Mito Stress Test kit (Agilent Technologies, 103015-100). For the ATP Rate Assay (PANC-1, TP53-KO PANC-1, Hs766T, HPDE, and MIA PaCa-2), cells were analysed with the Seahorse XF ATP Rate Assay kit (Agilent Technologies, 103582-100) after treatment. For the Glycolysis Stress Test (PANC-1 and TP53-KO PANC-1), cells were treated with RSL3 for 4 h prior to analysis with the Seahorse XF Glycolysis Stress Test kit (Agilent Technologies, 103020-100). For all assays, Seahorse XF RPMI media (Agilent Technologies, 103576-100) supplemented as appropriate was used: for Mito Stress and ATP Rate assays, 10 mM glucose (103577-100), 2 mM glutamine (103579-100), and 1 mM sodium pyruvate (103578-100); for the Glycolysis Stress Test, 2 mM glutamine only. At the end of the treatment, culture media was replaced with fresh XF RPMI, and plates were incubated for 1 h at 37 °C in a non-CO₂ incubator before running the assay. All cartridges were incubated overnight at 37 °C in a non-CO₂ incubator with calibrant solution (Agilent Technologies, 100840-100) before the day of the assay. For the Mito Stress Test, oligomycin (1 µM), FCCP (2 µM), and rotenone/antimycin A (0.5 µM) were loaded; for the ATP Rate Assay, oligomycin (1.5 µM) and rotenone/antimycin A (0.5 µM) were loaded; for the Glycolysis Stress Test, glucose (10 mM), oligomycin (1 µM), and 2-deoxyglucose (50 mM) were loaded. All conditions were performed in at least technical triplicate.

At the end of each assay, the 24-well plates were collected, the medium discarded, and the dried cell plates stored at − 80 °C overnight. Total DNA was quantified the following day using the CyQUANT™ Cell Proliferation Assay (Thermo Fisher Scientific, C7026) according to the manufacturer’s protocol. The CyQUANT™ working solution was prepared by diluting the dye 1:400 and the lysis buffer 1:20 in sterile, autoclaved H₂O. A total of 200 µL of this solution was added to each well and, after cell lysis, the contents were transferred to a 96-wells half-area Greiner plate (675096) for quantification. Fluorescence intensity was measured on a Tecan Infinite^® M Nano Plus microplate reader (Tecan Group Ltd., Zürich, Switzerland) with excitation at 480 nm and emission at 520 nm. DNA quantity (µg) was calculated using the standard curve provided in the kit and uploaded to Wave software for Seahorse data normalization. Considering the intrinsic metabolic differences observed at baseline between the cell lines, data obtained after normalization were referred to the correspondent control for each condition. Results are shown as fold change to control (CTRL).

Synergism evaluation and Compusyn analysis

For these evaluations, 7940b and MT3 cells were seeded in 96-wells plate as previously described. The following day, cells were treated with IKE and APR-246 as indicated in the figure, using a fixed ratio of 1:5. After 24 h treatment, the percentage of viable cells were analysed by MTT assay. Then, the percentage of dead cells were calculated for each condition, and used to evaluate synergism. In detail, data were analysed with CompuSyn (1.0) software, and the Combination Index (CI) was used to calculate the degree of the synergism in the combination (where CI < 1 = synergism), while the Fa (Fraction affected) indicates the percentage of the cells affected and represents the response level of the combinatory treatment.

Mouse models and orthotopic tumor implantation

Seven-week-old female C57BL/6 N (H-2b) mice were obtained from Charles River Laboratories (Calco, Italy). Mice were orthotopically implanted with mouse-derived KPC pancreatic ductal adenocarcinoma (PDAC) 7940b cells (carrying the mutant TP53 allele R172H, equivalent of the human TP53 R175H hotspot mutation). Briefly, mice were anesthetized, and a small incision was made in the left abdominal flank to expose the pancreas. A total of 5 × 10⁴ cells, resuspended in cold phosphate-buffered saline (PBS), were injected into the pancreatic tail using a 25-gauge insulin syringe (BD Micro-Fine). Successful engraftment was confirmed by the appearance of a localized bleb without leakage. The peritoneum was closed with absorbable sutures (Vetsuture), and the skin was secured with wound clips. Following surgery, mice were monitored daily for recovery, signs of pain or distress, and proper wound healing.

Tumor growth and progression were monitored by high-resolution ultrasound imaging (Vevo 2100, FUJIFILM VisualSonics). Tumors were identified using established anatomical landmarks, including the spleen, kidneys, liver, portal vein, and inferior vena cava. Imaging was performed in a blinded fashion, and data analysis was independently supervised by two investigators (Y.H. and T.W.). Tumor volume was determined by measuring the maximal longitudinal (length) and transverse (width) diameters and calculated using a modified ellipsoid formula: volume = ½ × (length × width²).

Tumor-bearing mice with comparable mean tumor volumes were randomized into treatment groups (n = 14/group). The experimental groups included vehicle control (DMSO), IKE (40 mg/kg), APR-246 (40 mg/kg), and the combination of IKE and APR-246. All compounds were administered via intraperitoneal injection (i.p.) once every other day. Tumor volumes were reassessed by ultrasound following four treatments. After two additional treatments, mice were euthanized, and tumors were collected for downstream analyses. Tumor burden did not exceed the maximum size permitted by institutional ethical approval (OPBA) and European regulations.

Mice were maintained under specific pathogen-free conditions at the University of Verona animal facility, with ad libitum access to food and water. All animal experiments were conducted in accordance with national and European regulations (protocols C46F4.26 and C46F4.30; Principal Investigator: Stefano Ugel) and approved by the University of Verona Ethical Committee. All procedures complied with FELASA guidelines and the Amsterdam Protocol for animal welfare. Animals were monitored daily and euthanized upon signs of excessive distress.

Isolation of tumor cells

Collected tumors from the in vivo experiment were then cut into small pieces with scissors under sterile conditions and incubated in freshly prepared dissociation buffer (collagenase 1 mg/mL, hyaluronidase 1 mg/mL, DNase I 0.025 mg/mL in RPMI) for 45 min at 37 °C. Tumor cell culture medium was added to terminate the dissociation and then pass through the 70 μm strainer to obtain single-cell suspensions. Cells from five different mice were kept in RPMI + FBS 10% and seeded in 24-wells plates at 3 × 10⁴ cells/well. The following day, Seahorse ATP Rate Assay and Glycolysis Stress Test were performed as previously described.

Alanine transaminase (ALT) activity evaluation

Alanine Transaminase (ALT) activity was measured in sera from mice (n = 8) from the in vivo experiment as index of liver damage. In detail, on the day of the sacrifice, blood samples were collected and sera were extracted and stored at -80 °C. ALT activity was evaluated by using the Alanine Transaminase Activity assay kit (ab105134) according to the manufacturer’s instructions. For this assay, sera were diluted 1:2. After 10 min incubation, outputs were measured at OD₅₇₀ nm using a kinetic mode and considering an incubation time of 20 min. Results were obtained using a Tecan Infinite^® M Nano Plus microplate reader (Tecan Group Ltd., Zürich, Switzerland) and concentrations were obtained using the standard curve provided by the kit.

Creatinine assay

Creatinine was measured in sera from mice (n = 8) from the in vivo experiment as index of renal damage. In detail, sera were collected as previously described, and centrifuged for 10 min at 4 °C (10,000 x g) using the 10 kDa spin columns (ab93349) and the filtrate was used for the following analysis. For this assay, sera were diluted 1:2. Then, creatinine levels were measured using the Creatinine Assay Kit (ab65340) according to the manufacturer’s instructions. After 1 h incubation at 37 °C, outputs were read at OD₅₇₀ nm using a Tecan Infinite^® M Nano Plus microplate reader (Tecan Group Ltd., Zürich, Switzerland). Concentrations were quantified using the standard curve provided by the kit.

Statistical analysis and software

Results are presented as mean ± SD, except for the in vivo experiments, which are presented as mean + SEM. For experiments with n ≤ 15, individual data points are shown. Statistical significance between two groups was determined using a two-tailed unpaired t-test, while comparisons among three or more groups were assessed using 2 wayANOVA. All the data were plotted using GraphPad Prism 10.3.1, which was also used for statistical analyses. Statistical significance is described in the figure legends as: ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001.

For western blot analysis, Image Lab software 6.1 was used, while flow cytometry data were analysed using the FlowJo software v10.8.1. For Seahorse data analysis, the Wave v2.6 software was used. For synergism evaluation, the CompuSyn (1.0) software was used. For RNA-seq data processing, R 4.5.2 software was used. In vivo scheme represented in Fig. 6G and graphical abstract were created using Biorender.com.

Abbreviations

PDAC

Pancreatic ductal adenocarcinoma

TP53

Tumor protein p53

KO

Knockout

GOF

Gain-of-function

ROS

Reactive oxygen species

4

HNE − 4-Hydroxynonenal

TEM

Transmission electron microscopy

Δψm

Mitochondrial membrane potential

OXPHOS

Oxidative phosphorylation

OCR

Oxygen Consumption Rate

MTT − 3

(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide

2

DG − 2-deoxy-D-glucose

ORA

Over representation analysis

KEGG

Kyoto Encyclopedia of Genes and Genomes

PI3K

Phosphoinositide 3-kinase

AKT

Protein kinase B

APR

246-Eprenetapopt

IKE

Imidazole ketone erastin

Authors' contributions

A.C. contributed to conceptualization, investigation, data curation, formal analysis, and writing. F.P. performed experiments and data analysis. F.D., R.P. M.P., and C.L. performed laboratory experiments. T.W., Y.H., and S.U. conducted the in vivo experiments. A.A. performed the bioinformatic analyses. B.C. carried out the TEM analyses. M.T.S., S.U., and M.D. supervised the study. S.U. and M.D. acquired the funding. A.F. contributed to conceptualization, investigation, data curation, formal analysis, writing, and funding acquisition. The manuscript was written with input from all authors.

Funding

This work was supported by Associazione Italiana per la Ricerca sul Cancro (AIRC, project number: 27080, PI A.F.); Fondo Italiano per la Scienza FIS3 (grant number, FIS-2024-04982) from Italian Ministry of University and Research (to A.F.); PNRR project MNESYS PE00000006 (CUP: B33C22001060002), NextGenerationEU (PNRR “HEAL ITALIA - Health Extended Alliance for Innovative Therapies, Advanced Lab-research, and Integrated Approaches of Precision Medicine” (project number: PE00000019, CUP: B33C22001030006); Excellence Project 2023–2027 of the University of Verona, Department of Neurosciences, Biomedicine and Movement Sciences. Associazione Italiana per la Ricerca sul Cancro (AIRC Projects: 21509 and 28730, PI: S.U.).

Data availability

The data generated in this study are publicly available in the Gene Expression Omnibus (GEO) under accession number GSE319384.

Declarations

Competing interests

The authors declare no competing interests.