HL60, MOLM14, SET2, MV4-11, OCI-AML2, OCI-AML3, K562, THP1, UT7-EPO, SKM1, NB4 and KASUMI AML cell lines were used. Patients provided written informed consent in accordance with the Declaration of Helsinki. Bone marrow (BM) samples were obtained from five patients with newly diagnosed AML (characteristics provided in the Online Supplementary Table S1). Cells were cultured in RPMI with glutamine (Gibco61870, Life Technologies® Saint Aubin, France) supplemented with 10% fetal bovine serum (FBS) and 4 mM glutamine. All AML cell lines were certified using their microsatellite identity (characteristics provided in Online Supplementary Table S2). Ferrostatin-1, necrostatin-1, necrostatin-1S, necrosulfonamide, QVD-OPH, APR-246 for the in vitro study, erastin and RSL3 were sourced from Selleckchem (Houston, TX). Chloroquine and doxycycline were obtained from Sigma–Aldrich (Saint-Louis, MO). FINO2 was purchased from Cayman Chemicals (Ann Arbor, MI). The APR-246 reagent used in the in vivo study was provided by APREA therapeutics (Solna, Sweden).

Inducible short hairpin RNA (shRNA) targeting SLC7A11 or GPX4 were constructed as previously described¹⁷ using the following sequences: SLC7A11#2, GCTGAATTGGGAACAACTATA; SLC7A11#3, GCAGTTGCTGGGCTGATTTAT; GPX4#1, GTGAGGCAAGACCGAAGTAAA; GPX4#2, CTACAACGTCAAATTCGATAT.

293-T packaging cells were used to produce lentiviral constructs through co-transfection with plasmids encoding lentiviral proteins. Supernatants were collected and ultracentrifuged for 48 hours (h) after transfection over two consecutive days and subsequently stored at -80°C. AML cell lines were plated at 2x10⁶/mL and 2-10 mL aliquots of lentiviral supernatants were added for 3 h. Cells were then grown in 10% fetal calf serum medium and further selected with puromycin. For shRNA induction, 200 mg/mL doxycycline was added to the culture medium.

Data acquisition and data analysis were conducted at the Cochin Cytometry and Immunobiology Facility. For glutathione measurements using monochlorobimane (MCB; Thermofischer; Waltham, MA), 2x10⁵ cells were labeled with 40 mM MCB in 1 mL of warm complete medium for 20 minutes (min) in a tissue culture incubator (37°C, 5% CO²) in the dark. The reaction was terminated using 1 mL of cold complete medium, followed by centrifugation (200xg, 1 min). The pelleted cells were then resuspended in 0.5 mL of cold complete medium and placed on ice in the dark until analysis by flow cytometry (FCM). The MCBGSH signal was detected using a 355 nm laser through a 450/50 nm filter. FCM data were collected using a BD Fortessa flow cytometer with DIVA software. 10,000 events were recorded for analysis. Data analysis was then carried out with KALUZA software. For lipid peroxide production measurements using C11- BODIPY (581/591) (2 mM) (Thermofischer, Waltham, MA), 2x10⁵ cells were labeled with C11-BODIPY in 1 mL of warm complete medium for 10 min in a tissue culture incubator (37°C, 5% CO²) in the dark. Cells were then washed twice and resuspended in 200 mL of fresh PBS. For cystine uptake measurements using BioTracker Cystine-FITC Live Cell Dye (5 mM) (Thermofischer, Waltham, MA), 2x10⁵ cells were labeled with Biotracker cystine in 1 mL of warm complete medium for 120 minutes in a tissue culture incubator (37°C, 5% CO²) in the dark. FCM data were collected using a C6 Accuri flow cytometer (Becton Dickinson, Le Pont de Claix, France) with CFlow Plus software. 10,000 events were captured for subsequent analysis with CFlow Plus software (Becton Dickinson, Le Pont de Claix, France).

Whole-cell extraction and western blotting were performed as previously described.¹⁷ Anti-GPX4 antibody was purchased from Proteintech (Manchester, UK). Anti-PARP, caspase 8, caspase 3, cleaved caspase 3, MLKL, pMLKL, p53 and SLC7A11 antibodies were sourced from Cell Signaling Technology (Danvers, MA, USA).

AML cells were plated at 20x10⁴/mL in 100 μl of 10% FBSsupplemented RPMI prior to the addition of compounds. Cells were cultured in the presence of the test compounds for 24 to 48 h at 37°C. Viability was quantified using the fluorescence based Uptiblue assay (Interchim, Montluçon, France). Uptiblue was added to each well in 10 mL aliquots. Fluorescence was then measured with a Typhoon FLA9500 scanner (GE Healthcare; IL). Fluorescence values were normalized to dimethyl sulfoxide (DMSO)-treated controls for each AML cell line. Half maximal inhibitory concentration (IC⁵⁰) values were calculated using a four parameter non-linear regression curve with Graph Pad Prism v8 (GraphPad, La Jolla, CA, USA). For primary AML cells, viability was assayed by FCM analysis using forward scatter (FSC) versus side scatter (SSC).

Cell viability was calculated for every dose combination of APR-246 and ferroptosis inducer using the Synergy Finder webtool (https://synergyfinder.fimm.fi/) and compared to each agent alone. Calculations were based on the ZIP model.¹⁸

Xenograft tumors were generated by randomly injecting 1×10⁶ MOLM14 shCTRL or shSLC7A11 cells into the tail veins of NOD/SCID IL-2 receptor g-chain-null mice (NSG) aged 6–9 weeks. Fourteen days after injection, doxycycline (200 mg/mL) and sucrose (1% weight:volume) were added to the drinking water of these animals. After 3 days, the mice were randomly treated with a daily intraperitoneal injection of APR-246 (100 mg/kg) or vehicle (phosphate-buffered saline [PBS]) for 4 days. All experiments were conducted in accordance with the guidelines of the Association for Assessment and Accreditation of Laboratory Animal Care International. Animals were used in accordance with a protocol reviewed and approved by the Institutional Animal Care and Use Committee of Région Midi- Pyrénées (France). BM (mixed from tibias and femurs) were dissected and flushed in Hanks balanced salt solution with 1% FBS. MNC from BM were labeled with PE-conjugated anti-hCD33, PerCP-Cy5.5-conjugated anti-mCD45.1 and APC-conjugated anti-hCD45 (all antibodies from Becton Dickinson, BD) to determine the fraction of human blasts (hCD45⁺mCD45.1−hCD33⁺ cells) using FCM. Acquisition of data was performed on a CytoFLEX (Beckman Coulter) flow cytometer with CytExpert software. The number of AML cells in the BM was determined using CountBright beads (Invitrogen, CA, USA) in accordance with the manufacturer’s protocol. Data analysis was performed with flowJo software.

In order to determine the activity of APR-246 in AML, we assayed a set of 12 AML cell lines carrying diverse and representative molecular abnormalities, and five primary AML samples (Online Supplementary Table S1). Most of the AML cell lines and primary AML cells were sensitive to cell death induction by APR-246 (Figure 1A and B). The IC⁵⁰ of APR-246 at 24 h for each cell line ranged from 11 mM to more than 200 mM, independently of their TP53 mutational status (Figure 1C). Previous in vivo human data have shown that the plasma concentrations of APR-246 in the 12 h following its intravenous administration range from 50 to 500 mM, suggesting that concentrations above 50 mM are suitable for in vitro studies of the early effects of APR-246.¹⁹ In our subsequent experiments, we selected five AML cell lines sensitive to APR-246 in these concentration ranges, with or without TP53 mutations.

In order to investigate the mechanisms underlying APR-246 activity against AML cells, we exposed the cells treated with APR-246 to inhibitors of various cell death pathways (Figure 2A). The decrease in cell viability was almost completely rescued by either iron chelation via deferoxamine (DFO) or by the lipophilic antioxidant ferrostatin- 1 (Fer1), demonstrating that cell death is both iron and reactive- oxygen species (ROS) dependent. Necrostatin-1 (Nec1) also consistently prevented cell death induced by APR-246 after short-term incubation in all AML cells lines. Nec1 has been used to define necroptotic cell death, but can also protect against ferroptosis through a target which is as yet unknown.²⁰ Induction of necroptosis after treatment with APR-246 in our cells was excluded due to the absence of protection by more specific necroptosis inhibitors (Necrostatin 1s and Necrosulfonamide),^21,22 and due to the absence of MLKL phosphorylation, a key marker in necroptosis²¹ (Online Supplementary Figure S1A and B). Inhibitors of autophagy (chloroquine) or apoptosis (QVD-OPH) failed to block APR-246-induced cell death. We confirmed that the mechanism of APR-246 induced cell death is distinct from apoptosis, as evidenced by the absence of caspases 3/8 or PARP1 cleavage, including in TP53 mutated AML cell lines (Figure 2B; Online Supplementary Figure S2A to C). Notably, the protection against cell death observed with Fer1, DFO or Nec1 was partially lost at higher doses of APR-246 (Figure 2C) and at later time points (Figure 2D). However, apoptosis did not appear to be the mechanism of this late death, since caspases 3/8 and PARP1 were not cleaved and QVD-OPH was still unable to prevent the cell death (Online Supplementary Figure S2A to E). Examination of the ultrastructural changes induced by APR-246 treatment did not reveal any characteristic features of apoptosis (i.e., no plasma membrane blebbing, chromatin condensation or nuclear fragmentation) or autophagy (absence of autophagolysosomes). Necrotic cells were rare, and some mitochondria showed membrane rupture and reduced cristae (Figure 3A; Online Supplementary Figure S3). We analyzed the levels of endogenous lipid peroxidation – a hallmark of ferroptosis induction – following APR-246 treatment by flow cytometry analysis with C11-BODIPY staining. We observed a high induction of lipid ROS in AML cell lines and primary AML cells from two patients treated with APR-246 ex vivo. This increase in lipid ROS was fully reversed by Fer1, indicating that lipid peroxides had been newly generated (Figure 3B and C). All these results allowed to conclude that APR-246 induces early cell death by ferroptosis, a recently described non-apoptotic form of regulated cell death that links together membrane lipid peroxidation, cysteine and iron metabolism, glutathione peroxidase activity and oxidative stress (as summarized in Figure 4A^23–25). As previously reported,^14,16 we observed that APR-246 treatment induced a dramatic decrease in GSH levels in AML cell lines (Figure 4B). Cysteine is the main biosynthetic precursor of GSH. Cysteine can be transported into cells via membrane transporters for neutral amino acids. However, in the extracellular space, cysteine is rapidly reduced to cystine. Thus the main source of intracellular cysteine comes from the entry of cystine into the cell via system xc-.²⁶ b-mercaptoethanol (b-ME) can promote cystine uptake through an alternative pathway.²⁷ b-ME was able to completely rescue the cell death and GSH depletion induced by APR-246 (Figure 4C and D). Cysteine treatment showed similar results as Fer1 treatment (Figure 4C and D). Altogether, these results demonstrate that APR-246 induces GSH depletion which induces ferroptosis in AML cells irrespective of their TP53 mutational status, and that ferroptosis induction is the main mechanism of cell death after early exposure to APR-246.

Since cysteine is a biosynthetic precursor of GSH, we asked whether the ability of cells to provide cysteine for GSH synthesis underlies the sensitivity to APR-246. Using FITC-labeled cystine, we showed that after exposure to APR-246, AML cells increased their uptake of cystine from the extracellular space (Figure 5A). Western blot analysis of the protein levels of SLC7A11 showed an increased amount of SLC7A11 (Figure 5B). This suggests that the cells enhanced cystine uptake by increasing SLC7A11 protein levels to maintain intracellular GSH levels after exposure to APR-246. We modulated the cystine uptake through SLC7A11 overexpression or inhibition. SLC7A11 overexpression in MOLM-14 and OCI-AML2 (Online Supplementary Figure S4) decreased the cell death and prevented the depletion of GSH induced by APR-246 (Figure 5C and D). Then we showed that targeting the SLC7A11 cystine transporter by RNA interference reduced the basal uptake of cystine (Online Supplementary Figure S4B and C) and had a very little effect on cell death (Online Supplementary Figure S4D), but strongly reduced cell proliferation in AML cells in vitro (Online Supplementary Figure S4E). Inhibition of SLC7A11 with RNA interference increased cell death and viability impairment, GSH depletion, and the accumulation of lipid peroxides induced by APR-246 (Figure 6A to C; Online Supplementary Figure S4F). Interestingly, basal GSH levels were not consistently affected by b−ME addition, cysteine addition, or the overexpression or inhibition of SLC7A11, which suggests that the amount of GSH in basal conditions is not a reliable marker of cell cystine uptake ability (Figures 4D, 5D and 6C). Finally, targeting SLC7A11 with erastin, a potent inhibitor of system xc-²⁸ that showed variable sensitivity in our AML cell lines (Online Supplementary Figure S4G), had synergistic activity with APR-246 both on cell death and on cell viability impairment (Figure 6D and E; Online Supplementary Figure S5). The association of erastin and APR-246 also had a synergistic effect on cell viability in five primary AML samples (Figure 6F). There was no correlation between the basal levels of GPX4 and SLC7A11 proteins and the sensitivity to APR-246 (Online Supplementary Figure S6A and B). Altogether, these data suggest that the ability of AML cells to prevent lipid peroxides accumulation by increasing their cystine uptake to support GSH after exposure to APR-246 is a key determinant of the sensitivity to this compound.

We next determined whether targeting ferroptosis pathways in combination with APR-246 can increase the antileukemic activity of this compound, mimicking SLC7A11 inhibition. Downregulation of GPX4 by RNA interference resulted in cell death (Online Supplementary Figure S7A and B). Our AML cell lines panel showed variable sensitivity to two ferroptosis-inducing drugs: RSL3, a direct GPX4 inhibitor;²⁹ and FINO2, an indirect GPX4 inhibitor and direct iron oxidant³⁰ (Online Supplementary Figure S8). We observed that knockdown of GPX4 increased the impairment of cell viability induced by APR-246 (Figure 7A). RSL3 and FINO2 in association with APR-246 synergistically decreased cell viability in AML cell lines (Figure 7B and C; Online Supplementary Figures S9 and 10). Collectively, these results show that pharmacological or genetic activation of ferroptosis enhances the antileukemic activity of APR-246 in AML.

We then examined whether inhibition of GSH synthesis through SLC7A11 inhibition could interfere with AML persistence and could enhance APR-246 activity in vivo. We engrafted MOLM14 cells transduced with either control (shSCR) or anti-SLC7A11 (shSLC7A11) doxycyclineinducible shRNA (Figure 8A). After induction of shRNA expression in vivo, we treated the mice with a 4-day APR- 246 regimen in order to mimic the therapeutic schedule used in clinical trials of APR-246.¹⁶ This treatment scheme varied from those previously published in mice in terms of treatment duration. Indeed, in these studies, APR-246 was administrated for 7 to 28 days, and reduction of tumor volume after 4 days of treatment was minimal at best.^7,8,14,16 However, our aim was not to assess the efficacy of APR-246 alone but to demonstrate that its association with SLC7A11 inhibition enhanced its anti-leukemic activity. As expected, APR-246 alone did not reduce BM tumor cell burden. SLC7A11 knockdown significantly reduced tumor cell burden in the BM (Figure 8B and C). Moreover, the decrease in BM tumor cell burden was enhanced when APR-246 treatment was combined to SLC7A11 knockdown (Figure 8B and C). Overall, these results showed that inhibition of anti-ferroptosis mechanisms enhanced the anti-leukemic activity of APR-246 in vivo.