Redox Biology 55 (2022) 102408 Contents lists available at ScienceDirect Redox Biology journal homepage: www.elsevier.com/locate/redox Acute lymphoblastic leukemia necessitates GSH-dependent ferroptosis defenses to overcome FSP1-epigenetic silencing Lucas B. Pontel a, b, **, Alberto Bueno-Costa a, Agustín E. Morellato b, Juliana Carvalho Santos c, Gaël Roué c, Manel Esteller a, d, e, f, * a Cancer Epigenetics Group, Josep Carreras Leukaemia Research Institute (IJC), Badalona, Barcelona, Catalonia, Spain b Instituto de Investigación en Biomedicina de Buenos Aires (IBioBA), CONICET-Partner Institute of the Max Planck Society, Buenos Aires, Argentina c Lymphoma Translational Group, Josep Carreras Leukaemia Research Institute (IJC), Badalona, Barcelona, Catalonia, Spain d Centro de Investigacion Biomédica en Red de Cáncer (CIBERONC), Madrid, Spain e Institucio Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Catalonia, Spain f Physiological Sciences Department, School of Medicine and Health Sciences, University of Barcelona (UB), Barcelona, Catalonia, Spain A R T I C L E I N F O A B S T R A C T Keywords: Ferroptosis is a form of cell death triggered by phospholipid hydroperoxides (PLOOH) generated from the iron- Glutathione dependent oxidation of polyunsaturated fatty acids (PUFAs). To prevent ferroptosis, cells rely on the antioxidant Ferroptosis glutathione (GSH), which serves as cofactor of the glutathione peroxidase 4 (GPX4) for the neutralization of Acute lymphoblastic leukemia PLOOHs. Some cancer cells can also limit ferroptosis through a GSH-independent axis, centered mainly on the GPX4 FSP1 ferroptosis suppressor protein 1 (FSP1). The significance of these two anti-ferroptosis pathways is still poorly DNA Methylation understood in cancers from hematopoietic origin. Here, we report that blood-derived cancer cells are selectively sensitive to compounds that block the GSH-dependent anti-ferroptosis axis. In T- and B- acute lymphoblastic leukemia (ALL) cell lines and patient biopsies, the promoter of the gene coding for FSP1 is hypermethylated, silencing the expression of FSP1 and creating a selective dependency on GSH-centered anti-ferroptosis defenses. In-trans expression of FSP1 increases the resistance of leukemic cells to compounds targeting the GSH-dependent anti-ferroptosis pathway. FSP1 over-expression also favors ALL-tumor growth in an in vivo chick chorioallantoic membrane (CAM) model. Hence, our results reveal a metabolic vulnerability of ALL that might be of therapeutic interest. 1. Introduction through ferroptosis [7]. Blocking GSH synthesis by the compound L-buthionine sulfoximine (L-BSO) also triggers ferroptosis, though this Cells can undergo death through several mechanisms such as effect largely depends on the cell line [8]. GSH is synthesized from apoptosis and necroptosis [1]. In the last decade, an iron-dependent cysteine and glutamate by the enzyme glutamate cysteine ligase (GCL), form of cell death was described termed ferroptosis. This mechanism which is the target of L-BSO and it is formed by a catalytic (GCLC) and a is independent of the classical apoptosis pathway, and it is characterized regulatory (GCLM) unit [9]. Cysteine, for GSH synthesis, can be ob­ by the accumulation of phospholipid hydroperoxides (PLOOH), inflict­ tained from the metabolization of cystine, which is imported through ing lethal damage to cell membranes [2]. In vitro, ferroptosis can be xCT (SLC7A11/SLC3A2). Accordingly, inhibition of xCT by erastin or reverted by iron chelators or by PLOOH-specific antioxidants such as sulfasalazine also triggers ferroptosis [2]. The last step of GSH synthesis ferrostatin-1 (Ferr-1) and vitamin E (α-tocopherol) [3–5]. To prevent is the addition of glycine carried out by the enzyme glutathione synthase this form of cell death, cells evolved the protein glutathione peroxidase 4 (GSS). (GPX4), which metabolizes PLOOH using glutathione (GSH) as cofactor Several reports have described mechanisms that prevent ferroptosis [6] (Fig. 1A). The genetic inactivation or the chemical inhibition of independently of GSH. The main factor associated to this GSH- GPX4 by compounds such as (1S,3R)-RSL3 (RSL3) trigger cell death independent ferroptosis defense is known as ferroptosis suppressor * Corresponding author. Cancer Epigenetics Group, Josep Carreras Leukaemia Research Institute (IJC), Badalona, Barcelona, Catalonia, Spain. ** Corresponding author. Cancer Epigenetics Group, Josep Carreras Leukaemia Research Institute (IJC), Badalona, Barcelona, Catalonia, Spain. E-mail addresses: lpontel@carrerasresearch.org (L.B. Pontel), mesteller@carrerasresearch.org (M. Esteller). https://doi.org/10.1016/j.redox.2022.102408 Received 14 June 2022; Accepted 12 July 2022 Available online 31 July 2022 2213-2317/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). L.B. Pontel et al. Redox Biology 55 (2022) 102408 protein 1 (FSP1) [10–12]. Initially, FSP1 was named as AIFM2 2.2. Patient data (Apoptosis inducing factor mitochondrial 2) based on sequence ho­ mology to apoptosis factor mitochondrial-associated 1 (AIFM1), and All the patient-derived data shown in this manuscript were obtained later renamed as FSP1 [13]. This factor presents NADH:ubiquinone from anonymized open databases. Specifically, the CpG promoter oxidoreductase activity and it has been shown to limit ferroptosis by methylation data were obtained mining primary tumors from hemato­ regenerating the coenzyme Q10 [11]. This ubiquinone appears as a poietic and lymphoid origin available at The Cancer Genome Atlas lipid-soluble antioxidant and might catalyze the detoxification of (TCGA) (https://www.cancer.gov/about-nci/organization/ccg/resea PLOOHs [14]. The interplay between FSP1 and GSH-dependent fer­ rch/structural-genomics/tcga) and from published datasets generated roptosis defenses has been addressed in several systems, including by Esteller’s lab [15–17]. CpG methylation from normal donors was xenograft models of lung cancer [10]. However, it is less understood the obtained from the previous works [18–21]. A total of 1259 tumor bi­ role of FSP1 in hematopoietic cancers. Here, we show that FSP1 is barely opsies and normal samples from hematopoietic and lymphoid expressed in ALL cell lines. This lack of expression correlates with the compartment were analyzed (data shown in Fig. 5a). Survival data were DNA hypermethylation of FSP1 promoter in cells and patient biopsies obtained from TCGA and TARGET cohorts through the platform cBio­ and creates a selective dependency of ALL cells on GSH-dependent fer­ Portal. Gepia2 was used to analyze survival data across the 32 TCGA roptosis defenses. cohorts [22]. 2. Material and methods 2.3. Viability assays 2.1. Cell lines Viability was determined by resazurin dye (Sigma-Aldrich, #R7017). Briefly, 3000 cells per well were seeded into 96-well plate at a final Jurkat, CTV-1, MOLT-4, MOLT-13, MOLT-16, SUPT-1, Nalm6, REH, volume of 200 μL. Depending on the cell line, at day 4 or 6 resazurin was Kasumi, Karpas-45, K562, ATN-1 and HL-60 were grown in Roswell Park added to a final concentration of 30 μmol L− 1, and plates were left in the Memorial Institute 1640 medium (RPMI-1640) supplemented with 10% incubator for 3–5 h. Fluorescence (λex = 525 nm; λem = 590 nm) was fetal bovine serum (FBS) and 1% penicillin/streptomycin. HCT-116 and measured 3 h later and viability calculated as a percentage relative to HEK293T cells were maintained in Dulbecco’s Modifies Eagle’s Medium the untreated well. Every experiment was performed by technical trip­ (DMEM), high glucose (4.5 g/L), supplemented with 1% penicillin/ licate and repeated at least 3 times. Reagents used were: 1S,3S-(RSL3) streptomycin and 10% FBS. Unless otherwise stated, cell lines were (MedChem Express, #HY-100218A), L-buthionine sulfoximine L-BSO obtained from American Type Culture Collection (ATCC). All cell lines (Sigma, #B2515), Ferrostatin-1 (MedChem Express, # HY-100579), were authenticated by short tandem repeat profiling (LGS Standards) Necrostatin-1 (MedChem Express, # HY-15760), quinoline-Val-Asp- and tested negative for mycoplasma infection. Difluorophenoxymethylketone (QVD) (MedChem Express, # HY- 12305), tert-butyl hydroquinone tBHQ (MedChem Express, #HY- Fig. 1. GSH metabolism is a vulnerability in acute lymphoblastic leukemia (ALL) cell lines. A. Scheme depicting glutathione (GSH) metabolism and the main cellular ferroptosis defenses. B. Cancer dependency maps for SLC7A11, GCLC, GCLM, GSS, GPX4 and FSP1. The statistical analysis was performed applying a one-way ANOVA and a Tukey test for multiple comparison. C. IC50 (Inhibitory concentration at 50%) of 1S,3S-RSL3 (RSL3) for the cell lines enumerated in the x-axis. The IC50 was calculated from viability curves shown in Sup. Fig. 1A (mean ± SD, n = 3). D. IC50 of L-buthionine-sulfoximine (L-BSO) from viability curves shown in Sup. Fig. 1B (mean ± SD, n = 3). # indicates that for HCT-116 the IC50 was >100 μmol L− 1 E. Log10(P value) of the Person correlation between area under the curve (AUC) toxicity data for RSL3 and gene expression in all the cell lines with available data at depmap.org. FSP1 ranks 6th among the genes with positive correlation, suggesting high FSP1 expression is usually accompanied by high resistance to RSL3. xCT (SLC7A11/SLC3A2), GCLC/GLCM (Gamma glutamyl-cysteine ligase), GSS (GSH synthase), GPX4 (GSH peroxidase 4), FSP1 (Ferroptosis suppressor protein 1), α-Toc (α-tocopherol), μM (μmol L− 1). 2 L.B. Pontel et al. Redox Biology 55 (2022) 102408 100489), 5′ -aza-2′ -deoxycytidine (DEC) (Sigma, # A3656). Fluores­ methylation analyses across cell lines from different tissue origin (data cence was determined in a Synergy H1 microplate reader (Biotek). shown in Fig. 4A), we used the Infinium Human Methylation BeadChip 450K (Illumina) [23]. The in-silico methylation analysis focused on the 2.4. Flow cytometry hematopoietic compartment was performed using the MethylationEPIC BeadChip 850K (Illumina) [24], and filtering the Sanger cell lines from Independent cultures were set in 24-well plates, grown in presence of hematopoietic and lymphoid origin [23]. For hematopoietic-derived the indicated condition for 24 h, and then stained with C11-BODIPY patient and normal samples, the DNA methylation microarray used 581/591 (Thermofisher #D3861) at a final concentration of 5 μmol was the Infinium Human Methylation BeadChip 450K (Illumina). DNA L− 1. Stained cells were transferred to a 96-well plate and fluorescence quality checks, bisulfite modification, hybridization, data normaliza­ intensity (FI) was determined using phycoerythrin (PE, λex = 488 λem ¼ tion, statistical filtering, β methylation were calculated after ssNoob 585 nm) and fluorescein isothiocyanate (FITC, excitation/emission of normalization using Minfi R package (Bioconductor) as described pre­ 488/530 nm) channels in a FACS-BD Canto II flow cytometer with a high viously [25]. Heatmaps were built using Next-Generation Clustered throughput (HTS) adaptor. FlowJo v9 software (BD Biosciences) was Heat Map Builder [26]. For bisulfite genomic sequencing, genomic DNA used for analysis. The ratio between the geometric mean FI of PE and was converted using the EZ DNA Methylation-Gold Kit (ZYMO Research, FITC was calculated and plotted relative to the untreated samples. C11- #D5005). Bisulfite PCR products were transformed into competent BODIPY shifts the fluorescence emission maximum from 590 nm to 510 bacteria and a minimum of 8 clones were sequenced to calculate nm when it is oxidized. methylation frequency. Bisulfite PCR primers for FSP1 amplification used were Fwd 5′ GYGTGAGTTAGGTTTTTAATT 3′ and Rv 5′ AAAAA­ 2.5. FSP1 in trans expression CACTTTAAACCAAATCTAAA 3’. Results were analyzed with BioEdit software and methylated cytosines were mapped using a custom perl For complementation of FSP1 expression, the lentivirus construct script (BSMapR). pLOC-AIFM2 (FSP1, #PLOHS_100010824) was obtained from Horizon discovery. HEK293T cells were transfected with the vector expressing 2.8. Glutathione (GSH) determination FSP1, psPAX2 (Addgene #12260) and pMD2. G (Addgene #12259) using JetPrime® Transfection Reagent (Polyplus #101000046). After GSH was determined using GSH-Glo Glutathione Assay (#V6911, 72 h, virus-containing medium was collected, filtered through a 0.45 μm Promega). Briefly, cells were grown for 24 h in presence of RSL3 or L- membrane, and delivered into CTV-1 and Jurkat cells. Blasticidin at 4 BSO. Afterwards, 10000 live cells were transferred into a white 96-well μg/mL was used for selection, and GFP-positive cells were purified by plate, and the reagents applied following manufacturer instructions. two rounds of cell sorting, obtaining >90% of GFP + cells. Then, FSP1 Luminescence was recorded in a Synergy H1 microplate reader (Biotek). expression was confirmed by western blot in the pool of sorted cells and A GSH-standard curve was performed in parallel, and the content of GSH using an anti-FLAG antibody. For generating the empty vector control, per cell in each sample calculated considering the 10000 cells seeded FSP1 was removed by restriction enzyme digestion from the pLOC- initially. AIFM2 plasmid. Then, the empty vector was re-generated by using Gibson assembly, yielding the vector pLOC-EV. 2.9. Chick embryo chorioallantoic membrane (CAM) model 2.6. Expression analysis Fertilized white Leghorn chicken eggs were purchased from Granja Santa Isabel, S. L. (Córdoba, Spain) and incubated for 9 days at 37 ◦ C Quantitative reverse transcription PCR (qRT-PCR) was used to assess with 55% humidity. At day 9 of their embryonic development, eggs were FSP1 mRNA expression. Briefly, total RNA was extracted from cell pel­ cleaned with alcohol 70◦ and a window of an approximate 2 cm-diam­ lets using the SimplyRNA kit (Promega #AS1340) in the automated eter was drilled on top of the air chamber of the eggshell. Then, half Maxwell RSC device (Promega, #AS4500). 2 μg of total RNA were million pEV or pFSP1 Jurkat cells were resuspended in 25 μL RPMI converted to cDNA using the RevertAid First Strand cDNA synthesis kit medium containing 10% FBS and 100 U/mL penicillin and streptomycin (Thermofisher Scientific, #K1622). For quantitative PCR, SYBR Green (Thermo Fisher Scientific, Inc.) and 25 μL Matrigel (Cultek, # 354234). PCR Master Mix (Life technologies, #4312704) was used and GAPDH The mix was incubated for 15 min at 37 ◦ C and subsequently implanted expression as housekeeping control. The primers used were: hGADPH into the CAM of each egg. The window was then covered with a sterile Fwd 5′ TCAAGGCTGAGAACGGGAAG 3’; hGADPH Rv 5′ tape and the eggs were placed back in the incubator. At days 12 and 14 CGCCCCACTTGATTTTGGAG 3’; hFSP1 Fwd 5′ AGACGGACAAAGGCA­ of their embryonic development, 0.5 μmol L− 1 of RSL3 or vehicle CAGAG 3’; hFSP1 Rv 5′ CAATGGCGTAGACGTTGCTG 3’. (DMSO) diluted in RPMI medium were administered topically on the Western blot was used to determine protein expression. Briefly, cell tumor-bearing CAMs. On the 16th day of development (7 days post- pellets were lysed in RIPA buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, implantation), chick embryos were sacrificed by decapitation. Tumors 1 mM EDTA, 1 mM EGTA, 1% NP-40), water-bath sonicated and lysed in were excised and carefully weighed to determine their mass. Laemmli 1x sample buffer (62.5 mM Tris-HCl pH 6.8, 25% glycerol, 2% SDS, 0.01% bromophenol blue, 5% β-mercaptoethanol). Antibodies used 2.10. Statistics were: anti-LaminB1 (Abcam, #ab16048, 1:2000); anti-GCLC (Santa Cruz Biotechnology, #sc-390811, 1:500); anti-GCLM (Santa Cruz Spearman’s correlation was used to address the association between Biotechnology, #sc-55585, 1:500); anti-FSP1 (Santa Cruz Biotech­ expression and methylation data. Kaplan-Meier plots and Logrank nology, #sc-377120, 1:500); anti-GPX4 (R&D Systems, #MAB5457, (Mantel-Cox) tests were used to estimate overall survival (OS) through 1:1000); anti-β-tubulin-HRP (Abcam, #21058, 1:2000). Secondary an­ univariate and multivariate Cox proportional hazards regression tibodies were anti-mouse HRP conjugated-secondary antibody (Abcam, models. Patients were stratified in those with expression higher or lower #ab9044, 1:5000) and anti-rabbit HRP conjugated-secondary antibody than the mean for the whole set. For TARGET T-ALL, we selected the (Sigma, #A0545, 1:5000). samples containing FSP1 expression determined by microarray. Statis­ tical analyses (One-way ANOVA adjusted with a Tukey’s test for mul­ 2.7. DNA methylation analyses tiple comparison) were carried out with GraphPad Prism 8. Comparison between two groups was carried out by unpaired t-test. Values of p < DNA methylation status at gene promoter was determined by DNA 0.05 were considered statistically significant. Specific details for statis­ methylation microarrays and bisulfite genomic sequencing. For the tical analysis can be found in the legend of each figure. 3 L.B. Pontel et al. Redox Biology 55 (2022) 102408 2.11. Datasets showing that induction of mitochondrial caspase activation by the Second Mitochondria-Derived Activator of Caspase (SMAC) mimetic All the data used were downloaded from depmap. org or cBioPortal compound BV6 is also necessary to induce cell death by GSH-depletion and analyzed using R-environment or GraphPad Prism v8. [31,32]. To further assess the cell death mechanism in ALL exposed to GPX4 or GSH-synthesis inhibitors, we evaluated RSL3 and L-BSO 3. Results and discussion toxicity in presence of the caspase inhibitor quinoline-Val-Asp-Difluorophenoxymethylketone (QVD) and of 3.1. GSH metabolism is required for growth of hematopoietic cancer cell Receptor-interacting serine/threonine-protein kinase 1 (RIPK1) inhibi­ lines tor Necrostatin-1 (Nec-1). RSL3 tolerance was prevented by Ferr-1 in both Jurkat and CTV-1 cells (Fig. 2D and E). Interestingly, Nec-1 also To understand the relevance of both GSH-dependent and GSH- prevented RSL3 toxicity, in agreement with a recent report showing that independent anti-ferroptosis mechanisms in hematopoietic cancers, we Nec-1 prevents ferroptosis independently of RIPK1 [33]. On the other explored the dependency of blood cancer cell lines on GPX4, GCLC, hand, QVD mildly reverted L-BSO toxicity in Jurkat cells, suggesting at GCLM, GSS, FSP1 and SLC7A11 genes. We selected the dependency score least in part cell death is executed by apoptosis in Jurkat cells upon GSH CHRONOS, which is available from the depmap portal data [27]. Genes depletion. A plausible mechanism to explain the increased sensitivity of that are required for cell growth will present negative CHRONOS scores, ALL cell lines towards GPX4 inhibition could be that ALL cells sensitive indicating that the inactivation of those genes causes a growth defect. A to RSL3 contain less GSH than those cells resistant to this drug. To gene completely essential will generate a CHRONOS score around − 1, challenge this hypothesis, we evaluated the level of reduced GSH in whereas a non-essential gene will show scores ≥ 0. We detected that Jurkat, CTV-1 and MOLT-16 (all sensitive to RSL3) and in K562 (resis­ cancer cell lines from hematopoietic origin (including leukemia, lym­ tant to RSL3). K562 presented as much reduced GSH as the cell lines phoma and myeloma cells) present negative CHRONOS scores for genes sensitive to GPX4 inhibition (Sup. Fig. 2B). The only exception was that code for factors involved in GSH synthesis (GCLC, GCLM and GSS) MOLT-16, which showed a significant less amount of reduced GSH than and for GPX4 (Fig. 1B). This observation suggests a selective dependency Jurkat, CTV-1 or K562. Moreover, the exposure of these cell lines to on GSH synthesis and GPX4 for hematopoietic-derived cancer cells. In RSL3 did not significantly altered GSH cellular content (Sup. Fig. 2B). In contrast, SLC7A11 and FSP1 presented a selective requirement for contrast, blocking GCL significantly depleted GSH in both RSL3 sensitive lymphoma cell lines, though were not required for growth of and resistant cells, overall supporting that the selective sensitivity of ALL leukemia-derived cells (Fig. 1B). To validate these data, we measured cells to RSL3 is not caused by differences in the GSH content. the tolerance of a panel of hematopoietic cancer cell lines to the GPX4 inhibitor RSL3 and to the GCL inhibitor L-BSO (Fig. 1C and D and Supp. 3.3. ALL cells lack the expression of FSP1 Fig. 1A and B). All the cell lines tested derived from B- and T-ALL were strikingly sensitivity to RSL3 and L-BSO. On the other hand, the chronic The analysis of the correlation between the tolerance to RSL3 myeloid leukemia (CML) cell line K562 and the colorectal carcinoma cell measured as area under de curve (AUC) and gene expression (Fig. 1E) line HCT-116 were among the most resistant cells to GPX4 and GCL suggests that cells presenting lower expression of FSP1 are more sensi­ inhibition. Both GPX4 and GSH are key players on preventing ferrop­ tive to the GPX4 inhibitor RSL3. This observation might indicate that tosis, thus suggesting that hematopoietic-derived cancer cells might be ALL cells present low expression of FSP1. Hence, to evaluate the un­ more vulnerable to this form of cell death. Indeed, the evaluation of the derlying cause of the increased sensitivity of hematopoietic cancer cells correlation between gene expression and the tolerance to RSL3 to GSH synthesis and GPX4 inhibition, we analyzed mRNA expression (measured as area under de curve (AUC) in the 823 cancer cell lines from data from all the available cell lines found at depmap. FSP1 presented a cancer therapeutic response portal) [28] suggests that expression of the significant low expression in cell lines classified as leukemia (Supp. Fig. anti-ferroptosis factor FSP1 positively correlated with RSL3 AUC 3A). To corroborate these data in hematopoietic cells, we performed (Pearson correlation score of 0.376, Fig. 1E). Thus, cell lines presenting western blots in an extended set of leukemia-derived cell lines. GCLC, reduced expression of FSP1-dependent anti-ferroptosis defenses might GCLM and GPX4 were detected in most of the hematopoietic cell lines be more prone to undergo ferroptosis upon GPX4 inhibition. tested (Fig. 3A), whereas FSP1 was not detected in any of these cell lines but K562 (Fig. 3A and Supp. Fig. 3B). Interestingly, K562 was the only 3.2. Ferroptosis inducers selectively eliminate ALL cells hematopoietic cell line resistant to RSL3 and L-BSO (Fig. 1C and D and Supp. Fig. 1A and 1B). The non-hematopoietic cell line HCT-116 also GSH is one of the main cellular antioxidants and participates in the expressed FSP1 (Fig. 3A), in accordance with previously published data detoxification of xenobiotics, conferring resistance to chemotherapeutic [11]. The expression of FSP1 might be controlled at transcriptional or drugs and toxins [29]. GSH is also a cofactor for GPX4, which has a more post-transcriptional level. To gain insights into the regulation of FSP1 in specific role on the detoxification of PLOOHs, suppressing ferroptosis blood cancer cells, we determined the level of FSP1 mRNA in a subset of [30]. To address whether the selective sensitivity of ALL cell lines to cell lines by RT-qPCR. FSP1 mRNA levels were significantly lower in ALL GPX4 and GSH-synthesis inhibition was because they are more prone to cells compared to cell lines that show high FSP1 expression such as K562 undergo ferroptosis, we determined the tolerance of CTV-1 (T-ALL), and HCT-116 (Fig. 3B). In addition to FSP1-lack of expression, the Jurkat (T-ALL), MOLT-16 (T-ALL), MOLT-13 (T-ALL), REH (B-ALL), and increased sensitivity of ALL cell lines towards ferroptosis inducers might K562 (CML) cell lines to RSL3 and L-BSO in presence of the ferroptosis be caused by a differential regulation of GSH biosynthesis genes antioxidant Ferr-1 (Fig. 2A and B and Supp. Fig. 2A). Remarkably, the (GCLC/GCLM) or GPX4. We addressed the expression of FSP1, GCLC, sensitivity of ALL cell lines against RSL3 was significantly limited by GCLM and GPX4 in response to RSL3 or L-BSO in Jurkat, CTV-1, Ferr-1. In support of this finding, we evaluated the generation of phos­ MOLT-16 and K562 by western blot (Fig. 3C). No significant differ­ pholipid hydroperoxides (PLOOH) by the specific dye C11-BODIPY ences in FSP1, GCLC and GCLM protein levels were detected upon 24-h (Fig. 2C). PLOOHs were detected upon exposure to RSL3 in CTV-1, exposure to RSL3 or L-BSO (Fig. 3D and E). We noted a change in the Jurkat and MOLT-16, and only mildly detected in the CML cell line electrophoretic mobility of GPX4 upon exposure to RSL3 likely caused K562. L-BSO also increased PLOOH but to a lesser extent than RSL3. by any of the post-translational modifications described in GPX4 [34], Accordingly, Ferr-1 was able to significantly limit the production of lipid and a net reduction in GPX4 protein level in K562 cells (Fig. 3D and E). hydroperoxides induced by RSL3 and to less extent L-BSO (Fig. 2C). Remarkably, K562 -resistant to RSL3- presented the lowest expression of Interestingly, the toxicity of L-BSO was prevented by Ferr-1 only in some GCLC among the four cell lines, which however did not correlate with ALL cell lines, consistently with the previous observation in Jurkat cells the GSH level in this cell line (Supp. Fig. 2B). Overall, the results 4 L.B. Pontel et al. Redox Biology 55 (2022) 102408 Fig. 2. Ferroptosis underlies GSH metabolism vulnerability in acute lymphoblastic leukemia (ALL). A. Viability assays performed with the indicated cell lines exposed to increasing concentrations of 1S,3S-RSL3 (RSL3). Ferrostatin-1 (Ferr-1) at 1 μmol L− 1 was used as ferroptosis-specific antioxidant. Data are represented as mean ± SD; n = 3 for CTV-1 (6 days); n = 4 for Jurkat (4 days); n = 3 for MOLT-16 (6 days); n = 6 for K562 (4 days). Two-way ANOVA corrected for multiple comparison using a Sidak test. B. Viability assays performed with the indicated cell lines exposed to increasing concentrations of L-buthionine-sulfoximine (L-BSO). Ferr-1 was used at 1 μmol L− 1 in the indicated samples. Data are represented as mean ± SD; n = 3 for CTV-1 (6 days); n = 5 for Jurkat (4 days); n = 3 for MOLT-16 (6 days); n = 6 for K562 (4 days). Two-way ANOVA corrected for multiple comparison using a Sidak test. C. Determination of lipid peroxides by C11-BODIPY in cells exposed to RSL3 0.25 μmol L− 1 or L-BSO 100 μmol L− 1 for 24 h. The ratio of the geometric mean corresponding to the C11-BODIPY fluorescence intensity calculated for the R-phycoerythrin (PE) channel (Exc. 488 nm/Emm. 585 nm) and for fluorescein isothiocyanate (FITC) channel (Exc. 488 nm/Emm. 530 nm) was calculated and plotted as % of the ratio obtained in the untreated samples (mean ± SD, one-way ANOVA corrected for multiple comparison using a Tukey test, ***P = 0.0001, ****P < 0.0001). D. Viability assays performed in Jurkat cells exposed to increasing concentrations of RSL3 or L-BSO and Ferr-1 at 1 μmol L− 1, necrostatin 1 (Nec-1) 30 μmol L− 1, or quinoline-Val-Asp-Difluorophenoxymethylketone (QVD) 10 μmol L− 1 (n = 3, mean ± SD, unpaired t-test comparing viability at the highest RSL3 or L-BSO concentration, ****P < 0.0001, *P = 0.0159). E. Viability assays performed in CTV-1 cells exposed during 6 days to increasing concentrations of RSL3 or L- BSO and Ferr-1 at 1 μmol L− 1, Nec-1 30 μmol L− 1, or QVD 10 μmol L− 1 (n = 3, mean ± SD, unpaired t-test comparing viability at the highest RSL3 or L-BSO concentration, ***P = 0.0001 (RSL3), ***P = 0.0009 (L-BSO)). 5 L.B. Pontel et al. Redox Biology 55 (2022) 102408 Fig. 3. FSP1 is silenced in acute lymphoblastic leukemia (ALL) cell lines. A. Immune blot detection of FSP1, GCLC, GCLM, and GPX4 in total protein extracts from the cell lines depicted in the figure. LaminB1 was used as loading control. B. RT-qPCR analysis of the expression of FSP1 in MOLT-16, CTV-1, Jurkat and HCT- 116. Data are plotted as expression relative to the level of RNA detected in MOLT-16 (mean ± SD, one-way ANOVA corrected for multiple comparison using a Tukey test, ****P < 0.0001, ***P = 0.0004). C. Immune blot detection of FSP1, GCLC, GCLM, and GPX4 in total protein extracts from the cell lines exposed to 1S,3S-RSL3 (RSL3) or L-buthionine sulfoximine (L-BSO) for 24 h. The concentrations used were 0.25 and 1 μmol L− 1 RSL3 for CTV-1, Jurkat, K562 and HCT-116. For MOLT-16 RSL3 was used at 0.05 and 0.25 μmol L− 1. L-BSO was used at 100 μmol L− 1 in all the cell lines. D. Quantification of FSP1 immuneblot shown in C. The data plotted correspond to 3 independent biological replicates. b-tubulin was used as loading control. E. Quantification of GCLC, GCLM and GPX4 immuneblots shown in Fig. 3C. The data plotted correspond to 3 independent biological replicates. b-tubulin was used as loading control (mean ± SD; n = 3). presented in this section suggest that ALL vulnerability to GPX4 or GSH − 0.5210, P < 0.0001, Fig. 4B) in the whole panel of cell lines included in synthesis inhibition is manly caused by the lack of FSP1 expression. the depmap portal [35]. Moreover, this correlation was detected in the subset of ALL cells (Spearman = − 0.4247, P = 0.0306, Fig. 4B). Overall, 3.4. FSP1 is hypermethylated in ALL cell lines these results suggest that the expression of FSP1 might be at least in part under DNA epigenetic control in ALL-derived cell lines. Indeed, FSP1 The striking downregulation of FSP1 in blood-derived cancer cell promoter harbors a defined CpG island upstream of its transcriptional lines correlated with the increase sensitivity to GPX4 and GCL inhibitors. start site (TSS) (Fig. 4C). To determine the methylation status of this CpG To address the underlying mechanism of FSP1 downregulation, we island, we performed a DNA methylation analysis using the Methyl­ interrogated whether the FSP1 promoter presents DNA hyper­ ationEPIC BeadChip 850K microarray on multiple hematopoietic cell methylation, which might indicate its selective epigenetic silencing. To lines (Fig. 4C, Supp. Fig. 4B). We detected an enrichment of methylated this end, we mined the Sanger set of human cancer cell lines from CpGs in the promoter of FSP1 mainly in ALL, suggesting that DNA different origins comparing the methylation of CpGs found in the pro­ methylation controls GSH-independent defenses against ferroptosis in moter region of genes involved in ferroptosis defenses: SLC7A11, cancer cells from hematopoietic origin. To corroborate this observation, SLC3A2, NFE2L2, NFE2L1, KEAP1, GSS, GCLM, GCLC, GPX4 and FSP1 we determined the CpG methylation status of the CpG island detected in [23]. We observed that FSP1 promoter presents some degree of DNA FSP1 promoter by bisulfite sequencing in CTV-1, Jurkat and K562. methylation in most of the cell lines tested (Fig. 4A), suggesting that Interestingly, Jurkat and CTV-1 presented DNA hypermethylation of DNA hypermethylation might impact on FSP1 expression. Accordingly, FSP1 promoter (>90%), whereas K562 showed partial methylation a significant inverse correlation could be detected between FSP1 (73.1%), concordantly with augmented FSP1 expression detected in this expression and FSP1 promoter DNA hypermethylation (Spearman = cell line (Fig. 4D). Concordantly, in the hypermethylated T-ALL cell lines 6 L.B. Pontel et al. Redox Biology 55 (2022) 102408 Fig. 4. DNA hypermethylation of FSP1 is a feature of acute lymphoblastic leukemia (ALL). A. Heatmap showing the DNA methylation of FSP1 promoter in cancer cell lines originated from the tissues described. CNS: Central nervous system. The number in parenthesis represents the number of cell lines included in each group. B. Top: Correlation between FSP1 RNA expression and FSP1 promoter methylation (1 kb upstream of the transcriptional start site (TSS). Data were obtained from depmap portal. The Spearman correlation and the corresponding P value are shown. Bottom: Correlation between FSP1 RNA expression and FSP1 promoter methylation (1 kb upstream of the TSS) in acute lymphoblastic leukemia (ALL) cell lines. C. DNA methylation analysis by using a Illumina EPIC array in Sanger cell lines from hematopoietic origin. FSP1 gene structure is shown on the top part of the heatmap highlighting the CpG island found in the promoter region. D. FSP1 promoter methylation determined by bisulfite sequencing in CTV-1, Jurkat and K562 cells (sequenced chromosomal region: chr10:71892987 to chr10:71892584, hg19). E. FSP1 promoter sequence including the CpG island and the overlapping NRF2-binding site. F. FSP1 mRNA expression in CTV-1, Jurkat and K562 cells cultured 8 days in presence 0.5 μmol L− 1 of 5′ -aza-2′ -deoxycytidine (decitabine, DEC). Data is plotted relative to the FSP1 mRNA level detected in HL-60 untreated. For CTV-1 100 nmol L− 1 of DEC was used. On day 7, tert-butylhydroquinone (tBHQ) was added at 30 mmol L− 1 (n = 5, mean ± SD, two-tailed unpaired t-test, *P = 0.029, **P = 0.0097 (CTV-1), ***P = 0.007 (Jurkat), ****P < 0.0001 (K562), *P = 0.0107 (HL-60). Jurkat and CTV-1, a 72-h exposure to the DNA methyltransferase in­ sufficient to upregulated FSP1 expression. The K562 cell line (CML, hibitor decitabine (5′ -aza-2′ -deoxycytidine, DEC) reduced FSP1 pro­ partially methylated) presented strong basal expression of FSP1. In this motor DNA hypermethylation to 52.5 and 72.5%, respectively (Supp. model, the combined exposure to tBHQ and DEC increases FSP1 Fig. 4A). DEC interferes with the activity of DNA methyltransferases expression from 52.3 ± 10.5 (untreated) to 88.2 ± 5.2 (tBHQ + DEC) reducing the CpG methylation of the DNA [36]. It is still plausible that folds compared to HL-60 basal expression (Fig. 4F). additional mechanisms might contribute to regulate FSP1 expression in ALL. Indeed, a recent report indicated that FSP1 is under the control of 3.5. FSP1 expression favors cancer progression nuclear factor erythroid 2-related factor 2 (NFE2L2 or NRF2) in lung cancer [37]. FSP1 promoter presents a conserved binding site for NRF2 To address whether the DNA epigenetic silencing of FSP1 is a feature that overlaps the CpG island (Fig. 4E). To evaluate the contribution of of leukemic cells from patients, we investigated the methylation status NRF2 to FSP1 regulation in ALL, we determined the FSP1 mRNA level of the FSP1 CpG island in cancer biopsies using the Infinium Human upon exposure to the NRF2 activator tert-butylhydroquinone (tBHQ). In Methylation 450 BeadChip (Fig. 5A and Supp. Fig. 5A). A defined subset addition, we used the DNA methyltransferase inhibitor DEC, which in­ of primary leukemic cells presented increased FSP1 DNA methylation, terferes with the activity of DNA methyltransferases [36]. The combi­ which was not detected in hematopoietic samples from normal donors nation of tBHQ and DEC was able to significantly upregulate FSP1 in (Fig. 5A). Accordingly, expression data from cBioPortal show that FSP1 Jurkat and CTV-1 cells, which present FSP1 DNA hypermethylation was barely expressed in leukemia, contrasting with the expression (Fig. 4F). Interestingly, in HL-60 -an AML cell line that does not present observed in non-hematopoietic cancers (Supp. Fig. 5B). To gain insights FSP1 DNA hypermethylation (Supp. Fig. 4B), the exposure to tBHQ was into the role of FSP1 in hematopoietic malignancies in comparison with 7 L.B. Pontel et al. Redox Biology 55 (2022) 102408 non-hematopoietic cancers, we performed a prognosis analysis in the 32 is amplified. Out of 10967 tumor samples, only 39 show FSP1 amplifi­ TCGA datasets, which include AML (LAML) but not ALL (Fig. 5B) [22]. cation and include patients from adrenocortical carcinoma (ACC), FSP1 high expression appeared as a significant worst prognosis marker bladder urothelial carcinoma (BLCA), breast invasive carcinoma for overall survival of LAML (Logrank Test P value: 0.0059) (Fig. 5C). In (BRCA), cholangio carcinoma (CHOL), colon adenocarcinoma (COAD), the TARGET dataset, the ALL patients that present FSP1 expression > glioblastoma multiforme (GBM), head and neck squamous cell carci­ median also showed a worst overall survival (Logrank Test P value: noma (HNSC), liver hepatocellular carcinoma (LIHC), lung adenocar­ 0.032) (Fig. 5C). Thus, those cancer cells that express FSP1 might pre­ cinoma (LUAD), ovarian serous cystadenocarcinoma (OV), pancreatic sent an advantage leading to a more aggressive cancer evolution. To carcinoma (PAAD), prostate adenocarcinoma (PRAD), stomach adeno­ further support this hypothesis, we look for TCGA tumors in which FSP1 carcinoma (STAD), and uterine corpus endometrial carcinoma (UCEC). (caption on next page) 8 L.B. Pontel et al. Redox Biology 55 (2022) 102408 Fig. 5. FSP1 expression confers an advantage to acute lymphoblastic leukemia (ALL) by limiting ferroptosis. A. Heatmap showing the methylation status of the CpGs found in the promoter of FSP1 in samples from hematopoietic origin (ALL: T- and B- acute lymphoblastic leukemia; AML: acute myeloid leukemia; CLL: chronic lymphocytic leukemia; MDS: myelodisplastic syndrome). B. Survival prognosis analysis according to FSP1 expression performed in TCGA cohorts. Red and blue tones indicate whether the expression of FSP1 is a poor or favourable prognostic marker, respectively. Squares indicate the difference between high and low expression groups is significant. Data were calculated using a Mantel–Cox test with p-values corrected for False discovery rate, and the cohort grouped with cut off higher or lower than 50% of the median for the high- and low-FSP1 expression populations, respectively. C. Prognosis analysis in a subset of samples from TARGET project (pediatric ALL). A z-score for the expression of FSP1 in all the samples was calculated, and samples with z-scores for FSP1 > or < than the mean are considered high or low FSP1-expressing samples, respectively (TCGA/cBioPortal). D. Survival for the pool of TCGA cohorts stratifying samples according to the genomic amplification of FSP1 (TCGA/cBioPortal). E. Viability curves for CTV-1 and Jurkat cell lines stably expressing the protein FSP1. Data are represented as mean ± SD, n = 3. Two-way ANOVA corrected for multiple comparison using a Sidak test. F. Determination of lipid peroxides by C11-BODIPY in cells exposed to RSL3 0.25 and 1 μmol L− 1, or L-BSO 100 μmol L− 1 for 24 h. The ratio of the geometric mean corresponding to the fluorescence intensity calculated for the R-phycoerythrin (PE) channel (Exc. 488 nm/Emm. 585 nm) and for fluorescein isothiocyanate (FITC) channel (Exc. 488 nm/Emm. 530 nm) was calculated, and plotted as % of the ratio obtained for the untreated sample (mean ± SD, one-way ANOVA corrected for multiple comparison using a Tukey test, **P = 0.0026, *P = 0.0357, ****P < 0.0001). G. Scheme depicting the chick embryo chorioallantoic membrane (CAM) tumor assay H. Left, tumor weights at day 16 plotted as percentage of the untreated group for the same genotype. Right, net tumor weights at day 16 for Jurkat/pEV and Jurkat/pFSP1 exposed to DMSO or 0.5 mmol L− 1 (mean ± SD, n = 5, two-tailed unpaired t-test). I. Representative pictures of engrafted tumors in CAM at day 16. The dotted line delimitates the tumor. Scale: 1 cm. These patients with FSP1 amplification present a worst prognosis sug­ induced in cancer but not normal cells [46]. Few advances have been gesting FSP1 might function as an oncogene (Fig. 5D) [38]. To address achieved by targeting GSH-dependent axis [47,48], and more recently whether FSP1 confers an advantage to ALL cells in conditions of fer­ by using an inhibitor of FSP1 [11,49]. In this work, we identified that roptosis induction, we generated Jurkat and CTV-1 cells stably ALL tumors lack the expression of the anti-ferroptosis factor FSP1. FSP1 expressing FSP1. FSP1-expressing cells were more resistant to ferrop­ is under the control of NRF2 in lung cancer [37]. Indeed, FSP1 promoter tosis induced by GPX4 inhibition (RSL3) and by GSH-synthesis inhibi­ contains an NRF2-binding site that overlaps with the CpG island. The tion (L-BSO), overall indicating that the lack of FSP1 expression DNA methylation of this CpG island at the FSP1 promoter prevents the sensitizes ALL to ferroptosis (Fig. 5E). To further support these findings, expression of FSP1 in response to NRF2 inducers. In contrast, the we measured PLOOHs in CTV-1 and Jurkat cells overexpressing FSP1 by non-methylated cell line HL-60 shows upregulation of FSP1 in response C11-BODIPY staining. PLOOHs were significantly reduced in both to an NRF2 inducer. CTV-1 and Jurkat cells exposed to RSL3 (Fig. 5F). In the 24-h exposure The silencing of FSP1 in ALL creates a dependency on the anti- protocol, L-BSO was not able to induce significant C11-BODIPY oxida­ ferroptosis axis centered on GPX4, which requires GSH metabolism. tion and, accordingly, no effect on PLOOHs was detected upon over­ We addressed that the ferroptosis inducers RSL3 and L-BSO are signifi­ expression of FSP1 (Fig. 5F). It is likely that longer L-BSO treatment cantly more toxic to ALL than to non-hematopoietic cancer cell lines. might trigger an accumulation of PLOOHs in CTV-1 cells, though for Moreover, RSL3 impairs tumor growth of Jurkat cells (ALL) but not of Jurkat, the cell death mechanism triggered by GSH depletion likely in­ the isogenic cells overexpressing FSP1 in a CAM model. Overall, our volves caspases and canonical apoptosis (Fig. 2D and E). Finally, to findings suggest that inducing ferroptosis might be a selective vulnera­ address whether FSP1 also protects ALL cells from ferroptosis during bility with therapeutic impact for ALL. tumor growth, we established a chick embryo chorioallantoic membrane (CAM) model using Jurkat/pEV and Jurkat/pFSP1 (Fig. 5G). This in vivo Funding model has long been used in cancer research to evaluate angiogenesis, tumor growth, metastasis, and treatment responses [39,40]. In this This work was supported by MCIN/AEI/10.13039/501100011033 assay, cells were deposited in matrigel medium on top of chick embryo and European Union “NextGenerationEU”/PRTR” Project PCI2021- membrane (day 9 after fertilization). Three and five days later, RSL3 (or 122045-2B. We thank CERCA Programme/Generalitat de Catalunya for DMSO) at 0.5 μmol L− 1 was applied in the same place where cells were institutional support. Work at M.E. laboratory is supported by the Health deposited. At day 16, eggs were opened and tumors weighed. We Department PERIS—project no. SLT/002/16/00374 and AGAUR—pro­ detected that RSL3 significantly impaired the growth of tumors consti­ ject no. 2017SGR1080 of the Catalan Government (Generalitat de Cat­ tuted by Jurkat/pEV. On the other hand, the tumors formed by alunya); Ministerio de Ciencia e Innovación (MCI), Agencia Estatal de Jurkat/pFSP1 were not impaired by RSL3 0.5 μmol L− 1. Moreover, Investigación (AEI) and European Regional Development Fund (ERDF) FSP1-overexpressing tumors show a growth advantage in presence of project no. RTI2018-094049-B-I00; the Cellex Foundation; and “la RSL3 (Fig. 5H and I). GPX4-inhibition in conditions of FSP1 over­ Caixa” Banking Foundation (LCF/PR/GN18/51140001). Studies at expression might lead to an increase in reactive oxygen species (ROS), Roué lab were partially funded by the ERDF through the Interreg V-A which are not able to kill the cells (because of the presence of FSP1) but Spain-France-Andorra (POCTEFA) program (EFA360/19, PROTEOblood can promote growth as previously shown for canonical ROS [41]. project). J.C.S was recipient of a Sara Borrell research contract (CD19/ 00228) from Instituto de Salud Carlos III. L.B.P. laboratory receives 4. Conclusions support from IBioBA-MPSP-CONICET (FOCEM COF 03/11, PICT-PRH 2017–4668), MPI for Metabolism Research (Cologne, Germany) and ALL accounts for 26% of all the pediatric cancers including young MPI for Biophysical Chemistry (Gottingen, Germany). A.E.M. is a adults [42]. The most common childhood leukemia is B-ALL, whereas CONICET fellow. A.B.C is a fellow of the Spanish Ministry of Education the incidence of T-ALL increases in adults reaching up to 25% of di­ and Vocational Training, under FPU contract no. FPU17/02423. M.E. is agnoses leukemia [43]. The overall survival is above 90% in the pedi­ an ICREA Research Professor. atric population but drops to less than 50% in adults [43]. Identifying genetic features in leukemic cells can guide personalized treatments Declaration of competing interest with optimal outcome. However, conventional chemotherapy is still widely used and usually produces long-term side effects [44], high­ M.E. is a consultant of Ferrer International and Quimatryx. The au­ lighting an urgent need for more selective targets to treat ALL thors declare they have no conflict of interest. malignancies. Since the identification of ferroptosis in 2012 [45], intensive work has been carried out to address whether this form of cell death could be 9 L.B. Pontel et al. Redox Biology 55 (2022) 102408 Acknowledgements RNA decapping enzyme NUDT16 mediates C-MYC activation in T-cell acute lymphoblastic leukemia, Leukemia 31 (2017) 1622–1625, https://doi.org/ 10.1038/leu.2017.99. We thank Josep Carreras Foundation for institutional support. [18] H. Heyn, N. Li, H.J. Ferreira, S. Moran, D.G. Pisano, A. Gomez, J. Diez, J. v. Sanchez-Mut, F. Setien, F.J. Carmona, A.A. Puca, S. Sayols, M.A. Pujana, J. Serra-Musach, I. Iglesias-Platas, F. Formiga, A.F. Fernandez, M.F. Fraga, S. Appendix A. Supplementary data C. Heath, A. Valencia, I.G. Gut, J. Wang, M. Esteller, Distinct DNA methylomes of newborns and centenarians, Proc. Natl. Acad. Sci. U. S. A. 109 (2012) Supplementary data to this article can be found online at https://doi. 10522–10527, https://doi.org/10.1073/pnas.1120658109. [19] H. Heyn, S. Moran, I. Hernando-Herraez, S. Sayols, A. Gomez, J. Sandoval, org/10.1016/j.redox.2022.102408. D. Monk, K. Hata, T. Marques-Bonet, L. Wang, M. Esteller, DNA methylation contributes to natural human variation, Genome Res. 23 (2013) 1363–1372, References https://doi.org/10.1101/gr.154187.112. [20] H. Heyn, F. Carmona Javier, A. Gomez, H.J. Ferreira, J.T. Bell, S. Sayols, K. Ward, O.A. Stefansson, S. Moran, J. Sandoval, J.E. Eyfjord, T.D. Spector, M. Esteller, DNA [1] J.G. Nirmala, M. Lopus, Cell death mechanisms in eukaryotes, Cell Biol. Toxicol. 36 methylation profiling in breast cancer discordant identical twins identifies DOK7 as (2020) 145–164, https://doi.org/10.1007/S10565-019-09496-2. novel epigenetic biomarker, Carcinogenesis 34 (2013) 102–108, https://doi.org/ [2] S.J. Dixon, K.M. Lemberg, M.R. Lamprecht, R. Skouta, E.M. Zaitsev, C.E. Gleason, 10.1093/carcin/bgs321. D.N. Patel, A.J. Bauer, A.M. Cantley, W.S. Yang, B. Morrison, B.R. Stockwell, [21] S. Moron-Lopez, V. Urrea, J. Dalmau, M. Lopez, M.C. Puertas, D. Ouchi, A. Gómez, Ferroptosis: an iron-dependent form of nonapoptotic cell death, Cell 149 (2012) C. Passaes, B. Mothe, C. Brander, A. Saez-Cirion, B. Clotet, M. Esteller, M. Berdasco, 1060–1072, https://doi.org/10.1016/j.cell.2012.03.042. J. Martinez-Picado, The genome-wide methylation profile of CD4+T cells from [3] G. Miotto, M. Rossetto, M.L. Di Paolo, L. Orian, R. Venerando, A. Roveri, A. individuals with human immunodeficiency virus (HIV) identifies distinct patterns M. Vučković, V. Bosello Travain, M. Zaccarin, L. Zennaro, M. Maiorino, S. Toppo, associated with disease progression, Clin. Infect. Dis. 72 (2021) E256, https://doi. F. Ursini, G. Cozza, Insight into the mechanism of ferroptosis inhibition by org/10.1093/cid/ciaa1047. –E264. ferrostatin-1, Redox Biol. 28 (2020), 101328, https://doi.org/10.1016/J. [22] Z. Tang, B. Kang, C. Li, T. Chen, Z. Zhang, GEPIA2: an enhanced web server for REDOX.2019.101328. large-scale expression profiling and interactive analysis, Nucleic Acids Res. 47 [4] A. Hinman, C.R. Holst, J.C. Latham, J.J. Bruegger, G. Ulas, K.P. McCusker, (2019) W556–W560, https://doi.org/10.1093/nar/gkz430. A. Amagata, D. Davis, K.G. Hoff, A.H. Kahn-Kirby, V. Kim, Y. Kosaka, E. Lee, S. [23] F. Iorio, T.A. Knijnenburg, D.J. Vis, G.R. Bignell, M.P. Menden, M. Schubert, A. Malone, J.J. Mei, S.J. Richards, V. Rivera, G. Miller, J.K. Trimmer, W.D. Shrader, N. Aben, E. Gonçalves, S. Barthorpe, H. Lightfoot, T. Cokelaer, P. Greninger, E. van Vitamin E hydroquinone is an endogenous regulator of ferroptosis via redox Dyk, H. Chang, H. de Silva, H. Heyn, X. Deng, R.K. Egan, Q. Liu, T. Mironenko, control of 15-lipoxygenase, PLoS One 13 (2018), https://doi.org/10.1371/ X. Mitropoulos, L. Richardson, J. Wang, T. Zhang, S. Moran, S. Sayols, JOURNAL.PONE.0201369. M. Soleimani, D. Tamborero, N. Lopez-Bigas, P. Ross-Macdonald, M. Esteller, N. [5] L. Probst, J. Dächert, B. Schenk, S. Fulda, Lipoxygenase inhibitors protect acute S. Gray, D.A. Haber, M.R. Stratton, C.H. Benes, L.F.A. Wessels, J. Saez-Rodriguez, lymphoblastic leukemia cells from ferroptotic cell death, Biochem. Pharmacol. 140 U. McDermott, M.J. Garnett, A landscape of pharmacogenomic interactions in (2017) 41–52, https://doi.org/10.1016/J.BCP.2017.06.112. cancer, Cell 166 (2016) 740–754, https://doi.org/10.1016/j.cell.2016.06.017. [6] M. Maiorino, M. Conrad, F. Ursini, GPx4, lipid peroxidation, and cell death: [24] S. Moran, C. Arribas, M. Esteller, Validation of a DNA methylation microarray for discoveries, rediscoveries, and open issues, Antioxidants Redox Signal. 29 (2018) 850,000 CpG sites of the human genome enriched in enhancer sequences, 61–74, https://doi.org/10.1089/ARS.2017.7115. Epigenomics 8 (2016) 389–399, https://doi.org/10.2217/epi.15.114. [7] W.S. Yang, R. Sriramaratnam, M.E. Welsch, K. Shimada, R. Skouta, V. [25] J. Sandoval, H.A. Heyn, S. Moran, J. Serra-Musach, M.A. Pujana, M. Bibikova, S. Viswanathan, J.H. Cheah, P.A. Clemons, A.F. Shamji, C.B. Clish, L.M. Brown, A. M. Esteller, Validation of a DNA methylation microarray for 450,000 CpG sites in W. Girotti, V.W. Cornish, S.L. Schreiber, B.R. Stockwell, Regulation of ferroptotic the human genome, Epigenetics 6 (2011) 692–702, https://doi.org/10.4161/ cancer cell death by GPX4, Cell 156 (2014) 317–331, https://doi.org/10.1016/j. epi.6.6.16196. cell.2013.12.010. [26] B.M. Broom, M.C. Ryan, M. Stucky, C. Wakefield, J.M. Melott, R. Akbani, J. [8] Y. Sun, Y. Zheng, C. Wang, Y. Liu, Glutathione depletion induces ferroptosis, N. Weinstein, Interactive Clustered Heat Map Builder: an Easy Web-Based Tool for autophagy, and premature cell senescence in retinal pigment epithelial cells article, Creating Sophisticated Clustered Heat Maps, vol. 8, 2020, p. F1000Res, https:// Cell Death Dis. 9 (2018), https://doi.org/10.1038/s41419-018-0794-4. doi.org/10.12688/f1000research.20590.2. [9] K. Aquilano, S. Baldelli, M.R. Ciriolo, Glutathione: new roles in redox signalling for [27] R.M. Meyers, J.G. Bryan, J.M. McFarland, B.A. Weir, A.E. Sizemore, H. Xu, N. an old antioxidant, Front. Pharmacol. 5 (AUG) (2014) 1–12, https://doi.org/ V. Dharia, P.G. Montgomery, G.S. Cowley, S. Pantel, A. Goodale, Y. Lee, L.D. Ali, 10.3389/fphar.2014.00196. G. Jiang, R. Lubonja, W.F. Harrington, M. Strickland, T. Wu, D.C. Hawes, V. [10] K. Bersuker, J.M. Hendricks, Z. Li, L. Magtanong, B. Ford, P.H. Tang, M.A. Roberts, A. Zhivich, M.R. Wyatt, Z. Kalani, J.J. Chang, M. Okamoto, K. Stegmaier, T. B. Tong, T.J. Maimone, R. Zoncu, M.C. Bassik, D.K. Nomura, S.J. Dixon, J. R. Golub, J.S. Boehm, F. Vazquez, D.E. Root, W.C. Hahn, A. Tsherniak, A. Olzmann, The CoQ oxidoreductase FSP1 acts parallel to GPX4 to inhibit Computational correction of copy number effect improves specificity of CRISPR- ferroptosis, Nature 575 (2019) 688–692, https://doi.org/10.1038/s41586-019- Cas9 essentiality screens in cancer cells, Nat. Genet. 49 (2017) 1779–1784, https:// 1705-2. doi.org/10.1038/ng.3984. [11] S. Doll, F.P. Freitas, R. Shah, M. Aldrovandi, M.C. da Silva, I. Ingold, A. Goya [28] M.G. Rees, B. Seashore-Ludlow, J.H. Cheah, D.J. Adams, E.v. Price, S. Gill, Grocin, T.N. Xavier da Silva, E. Panzilius, C.H. Scheel, A. Mourão, K. Buday, S. Javaid, M.E. Coletti, V.L. Jones, N.E. Bodycombe, C.K. Soule, B. Alexander, A. Li, M. Sato, J. Wanninger, T. Vignane, V. Mohana, M. Rehberg, A. Flatley, A. Schepers, P. Montgomery, J.D. Kotz, C.S.Y. Hon, B. Munoz, T. Liefeld, V. Dancík, D.A. Haber, A. Kurz, D. White, M. Sauer, M. Sattler, E.W. Tate, W. Schmitz, A. Schulze, C.B. Clish, J.A. Bittker, M. Palmer, B.K. Wagner, P.A. Clemons, A.F. Shamji, S. V. O’Donnell, B. Proneth, G.M. Popowicz, D.A. Pratt, J.P.F. Angeli, M. Conrad, L. Schreiber, Correlating chemical sensitivity and basal gene expression reveals FSP1 is a glutathione-independent ferroptosis suppressor, Nature 575 (2019) mechanism of action, Nat. Chem. Biol. 12 (2016) 109–116, https://doi.org/ 693–698, https://doi.org/10.1038/s41586-019-1707-0. 10.1038/nchembio.1986. [12] C. Mao, X. Liu, Y. Zhang, G. Lei, Y. Yan, H. Lee, P. Koppula, S. Wu, L. Zhuang, [29] C. Umansky, A.E. Morellato, M. Rieckher, M.A. Scheidegger, M.R. Martinefski, G. B. Fang, M.V. Poyurovsky, K. Olszewski, B. Gan, DHODH-mediated ferroptosis A. Fernández, O. Pak, K. Kolesnikova, H. Reingruber, M. Bollini, G.P. Crossan, defence is a targetable vulnerability in cancer, Nature 593 (2021) 586–590, N. Sommer, M.E. Monge, B. Schumacher, L.B. Pontel, Endogenous formaldehyde https://doi.org/10.1038/s41586-021-03539-7. scavenges cellular glutathione resulting in redox disruption and cytotoxicity, Nat. [13] M. Wu, L.G. Xu, X. Li, Z. Zhai, H.B. Shu, AMID, an apoptosis-inducing factor- Commun. 13 (2022) 1–15, https://doi.org/10.1038/s41467-022-28242-7, 2022 homologous mitochondrion-associated protein, induces caspase-independent 13:1. apoptosis, J. Biol. Chem. 277 (2002) 25617–25623, https://doi.org/10.1074/JBC. [30] W.S. Yang, R. Sriramaratnam, M.E. Welsch, K. Shimada, R. Skouta, V. M202285200. S. Viswanathan, J.H. Cheah, P.A. Clemons, A.F. Shamji, C.B. Clish, L.M. Brown, A. [14] B. Frei, M.C. Kim, B.N. Ames, Ubiquinol-10 is an effective lipid-soluble antioxidant W. Girotti, V.W. Cornish, S.L. Schreiber, B.R. Stockwell, Regulation of ferroptotic at physiological concentrations, Proc. Natl. Acad. Sci. USA 87 (1990) 4879–4883, cancer cell death by GPX4, Cell 156 (2014) 317–331, https://doi.org/10.1016/j. https://doi.org/10.1073/PNAS.87.12.4879. cell.2013.12.010. [15] J. Sandoval, H. Heyn, J. Méndez-González, A. Gomez, S. Moran, M. Baiget, [31] H. Schoeneberger, K. Belz, B. Schenk, S. Fulda, Impairment of antioxidant defense M. Melo, I. Badell, J.F. Nomdedéu, M. Esteller, Genome-wide DNA methylation via glutathione depletion sensitizes acute lymphoblastic leukemia cells for Smac profiling predicts relapse in childhood B-cell acute lymphoblastic leukaemia, Br. J. mimetic-induced cell death, Oncogene 34 (2015) 4032–4043, https://doi.org/ Haematol. 160 (2013) 406–409, https://doi.org/10.1111/bjh.12113. 10.1038/ONC.2014.338. [16] A.F. Fernandez, Y. Assenov, J.I. Martin-Subero, B. Balint, R. Siebert, H. Taniguchi, [32] Y. Sun, Y. Zheng, C. Wang, Y. Liu, Glutathione depletion induces ferroptosis, H. Yamamoto, M. Hidalgo, A.C. Tan, O. Galm, I. Ferrer, M. Sanchez-Cespedes, autophagy, and premature cell senescence in retinal pigment epithelial cells, Cell A. Villanueva, J. Carmona, J.v. Sanchez-Mut, M. Berdasco, V. Moreno, G. Capella, Death Dis. 9 (2018), https://doi.org/10.1038/S41419-018-0794-4. D. Monk, E. Ballestar, S. Ropero, R. Martinez, M. Sanchez-Carbayo, F. Prosper, [33] H. Yuk, M. Abdullah, D.-H. Kim, H. Lee, S.-J. Lee, Necrostatin-1 prevents X. Agirre, M.F. Fraga, O. Graña, L. Perez-Jurado, J. Mora, S. Puig, J. Prat, ferroptosis in a RIPK1- and Ido-independent manner in hepatocellular carcinoma, L. Badimon, A.A. Puca, S.J. Meltzer, T. Lengauer, J. Bridgewater, C. Bock, Antioxidants 10 (2021) 1347, https://doi.org/10.3390/antiox10091347. M. Esteller, A DNA methylation fingerprint of 1628 human samples, Genome Res. [34] C. Cui, F. Yang, Q. Li, Post-translational modification of GPX4 is a promising target 22 (2012) 407–419, https://doi.org/10.1101/gr.119867.110. for treating ferroptosis-related diseases, Front. Mol. Biosci. 9 (2022) 501, https:// [17] C. Anadón, G. van Tetering, H.J. Ferreira, C. Moutinho, A. Martínez-Cardús, doi.org/10.3389/fmolb.2022.901565. A. Villanueva, M. Soler, H. Heyn, S. Moran, M. Castro De Moura, F. Setien, A. Vidal, E. Genescà, J.M. Ribera, J.F. Nomdedeu, S. Guil, M. Esteller, Epigenetic loss of the 10 L.B. Pontel et al. Redox Biology 55 (2022) 102408 [35] A. Tsherniak, F. Vazquez, P.G. Montgomery, B.A. Weir, G. Kryukov, G.S. Cowley, [42] E. Ward, C. DeSantis, A. Robbins, B. Kohler, A. Jemal, Childhood and adolescent S. Gill, W.F. Harrington, S. Pantel, J.M. Krill-Burger, R.M. Meyers, L. Ali, cancer statistics, 2014, CA A Cancer J. Clin. 64 (2014) 83–103, https://doi.org/ A. Goodale, Y. Lee, G. Jiang, J. Hsiao, W.F.J. Gerath, S. Howell, E. Merkel, 10.3322/CAAC.21219. M. Ghandi, L.A. Garraway, D.E. Root, T.R. Golub, J.S. Boehm, W.C. Hahn, Defining [43] V. Cordo’, J.C.G. van der Zwet, K. Canté-Barrett, R. Pieters, J.P.P. Meijerink, T-Cell a cancer dependency map, Cell 170 (2017) 564–576, https://doi.org/10.1016/j. acute lymphoblastic leukemia: a roadmap to targeted therapies, Blood Cancer cell.2017.06.010, e16. Discov. 2 (2021) 19–31, https://doi.org/10.1158/2643-3230.BCD-20-0093. [36] M.F. Paz, M.F. Fraga, S. Avila, M. Guo, M. Pollan, J.G. Herman, M. Esteller, [44] H. Inaba, C.G. Mullighan, Pediatric acute lymphoblastic leukemia, Haematologica A systematic profile of DNA methylation in human cancer cell lines, Cancer Res. 63 105 (2020) 2524–2539, https://doi.org/10.3324/HAEMATOL.2020.247031. (2003) 1114–1121. [45] S.J. Dixon, K.M. Lemberg, M.R. Lamprecht, R. Skouta, E.M. Zaitsev, C.E. Gleason, [37] P. Koppula, G. Lei, Y. Zhang, Y. Yan, C. Mao, L. Kondiparthi, J. Shi, X. Liu, D.N. Patel, A.J. Bauer, A.M. Cantley, W.S. Yang, B. Morrison, B.R. Stockwell, A. Horbath, M. Das, W. Li, M.v. Poyurovsky, K. Olszewski, B. Gan, A targetable Ferroptosis: an iron-dependent form of nonapoptotic cell death, Cell 149 (2012) CoQ-FSP1 axis drives ferroptosis- and radiation-resistance in KEAP1 inactive lung 1060–1072, https://doi.org/10.1016/j.cell.2012.03.042. cancers, Nat. Commun. 13 (2022) 2206, https://doi.org/10.1038/s41467-022- [46] X. Jiang, B.R. Stockwell, M. Conrad, Ferroptosis: mechanisms, biology and role in 29905-1. disease, Nat. Rev. Mol. Cell Biol. 22 (2021) 266–282, https://doi.org/10.1038/ [38] M. Yoshimoto, S.R.C. de Toledo, E.M. Monteiro Caran, M.T. de Seixas, M.L. de s41580-020-00324-8, 2021 22:4. Martino Lee, S. de Campos Vieira Abib, S.M.R. Vianna, S.T. Schettini, J.A. [47] N. Zaffaroni, G.L. Beretta, Ferroptosis inducers for prostate cancer therapy, Curr. D. Andrade, MYCN gene amplification: identification of cell populations containing Med. Chem. 29 (2022), https://doi.org/10.2174/0929867329666220111120924. double minutes and homogeneously staining regions in neuroblastoma tumors, [48] B. Hassannia, B. Wiernicki, I. Ingold, F. Qu, S. Van Herck, Y.Y. Tyurina, H. Bayir, B. Am. J. Pathol. 155 (1999) 1439–1443, https://doi.org/10.1016/S0002-9440(10) A. Abhari, J.P.F. Angeli, S.M. Choi, E. Meul, K. Heyninck, K. Declerck, C. 65457-0. S. Chirumamilla, M. Lahtela-Kakkonen, G. Van Camp, D.V. Krysko, P.G. Ekert, [39] H. el Hasasna, A. Saleh, H. al Samri, K. Athamneh, S. Attoub, K. Arafat, S. Fulda, B.G. De Geest, M. Conrad, V.E. Kagan, W. Vanden Berghe, N. Benhalilou, S. Alyan, J. Viallet, Y. al Dhaheri, A. Eid, R. Iratni, Rhus coriaria P. Vandenabeele, T. Vanden Berghe, Nano-targeted induction of dual ferroptotic suppresses angiogenesis, metastasis and tumor growth of breast cancer through mechanisms eradicates high-risk neuroblastoma, J. Clin. Invest. 128 (2018) inhibition of STAT3, NFÎ◦ B and nitric oxide pathways, Sci. Rep. 6 (2016) 1–15, 3341–3355, https://doi.org/10.1172/JCI99032. https://doi.org/10.1038/srep21144. [49] H. Yoshioka, T. Kawamura, M. Muroi, Y. Kondoh, K. Honda, M. Kawatani, H. Aono, [40] I.W. Achkar, S. Kader, S.S. Dib, K. Junejo, S.B. Al-Bader, S. Hayat, A.M. Bhagwat, H. Waldmann, N. Watanabe, H. Osada, Identification of a small molecule that X. Rousset, Y. Wang, J. Viallet, K. Suhre, A. Halama, Metabolic signatures of tumor enhances ferroptosis via inhibition of ferroptosis suppressor protein 1 (FSP1), ACS responses to doxorubicin elucidated by metabolic profiling in ovo, Metabolites 10 Chem. Biol. (2022), https://doi.org/10.1021/ACSCHEMBIO.2C00028 (2020) 268, https://doi.org/10.3390/metabo10070268. acschembio.2c00028. [41] C.R. Reczek, N.S. Chandel, The two faces of reactive oxygen species in cancer, Annu. Rev. Cell Biol. 1 (2017) 79–98, https://doi.org/10.1146/annurev-cancerbio- 041916-065808. 11