Abstract

Background

   In the ongoing battle against BCR-ABL+ leukemia, despite significant
   advances with tyrosine kinase inhibitors (TKIs), the persistent
   challenges of drug resistance and the enduring presence of leukemic
   stem cells (LSCs) remain formidable barriers to achieving a cure.

Methods

   In this study, we demonstrated that Disulfiram (DSF) induces
   ferroptosis to synergize with TKIs in inhibiting BCR-ABL+ cells,
   particularly targeting resistant cells and LSCs, using cell models,
   mouse models, and primary cells from patients. We elucidated the
   mechanism by which DSF promotes GPX4 degradation to induce ferroptosis
   through immunofluorescence, co-immunoprecipitation (CO-IP), RNA
   sequencing, lipid peroxidation assays, and rescue experiments.

Results

   Here, we present compelling evidence elucidating the sensitivity of
   DSF, an USA FDA-approved drug for alcohol dependence, towards
   BCR-ABL+ cells. Our findings underscore DSF’s ability to selectively
   induce a potent cytotoxic effect on BCR-ABL+ cell lines and effectively
   inhibit primary BCR-ABL+ leukemia cells. Crucially, the combined
   treatment of DSF with TKIs selectively eradicates TKI-insensitive stem
   cells and resistant cells. Of particular note is DSF’s capacity to
   disrupt GPX4 stability, elevate the labile iron pool, and intensify
   lipid peroxidation, ultimately leading to ferroptotic cell death. Our
   investigation shows that BCR-ABL expression induces alterations in
   cellular iron metabolism and increases GPX4 expression. Additionally,
   we demonstrate the indispensability of GPX4 for LSC development and the
   initiation/maintenance of BCR-ABL+ leukemia. Mechanical analysis
   further elucidates DSF’s capacity to overcome resistance by reducing
   GPX4 levels through the disruption of its binding with HSPA8, thereby
   promoting STUB1-mediated GPX4 ubiquitination and subsequent proteasomal
   degradation. Furthermore, the combined treatment of DSF with TKIs
   effectively targets both BCR-ABL+ blast cells and drug-insensitive
   LSCs, conferring a significant survival advantage in mouse models.

Conclusion

   In summary, the dual inhibition of GPX4 and BCR-ABL presents a
   promising therapeutic strategy to synergistically target blast cells
   and drug-insensitive LSCs in patients, offering potential avenues for
   advancing leukemia treatment.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12943-024-02162-0.

   Keywords: Ferroptosis, Disulfiram, BCR-ABL+ leukemia, GPX4, Tyrosine
   kinase inhibitor

Introduction

   Chronic myeloid leukemia (CML) and a subset of acute lymphoblastic
   leukemia (ALL), collectively known as Philadelphia chromosome-positive
   (Ph+) leukemias, are marked by a common pathogenic feature—the
   Philadelphia chromosome, which arises from the t(9;22)(q34;q11)
   translocation [[52]1]. This translocation leads to the formation of the
   fusion gene BCR::ABL (BCR-ABL), producing the dysregulated BCR-ABL
   tyrosine kinase that serves as the primary driver of leukemogenesis and
   remains the central focus of targeted therapies.

   The advent of tyrosine kinase inhibitors (TKIs), including imatinib
   mesylate (IM), dasatinib (DAS), and nilotinib (NL), has ushered in a
   transformative era in CML management. By specifically targeting BCR-ABL
   kinase activity, these TKIs have shifted CML from a once-fatal disease
   to a more manageable condition [[53]2]. Despite substantial clinical
   efficacy observed in chronic phase (CP) CML, achieving complete cure
   remains elusive, with only approximately 10% of CP CML patients able to
   discontinue TKI treatment and maintain therapy-free remission.
   Challenges persist in addressing reduced TKI efficacy during the
   accelerated phase and blast crisis (BC) of CML, along with prevalent
   primary and acquired resistance to these compounds. Similarly,
   BCR-ABL+ acute lymphoblastic leukemia (ALL), sharing resemblances with
   the aggressive lymphoid BC of CML, confronts comparable challenges of
   relapse under current TKI monotherapies [[54]1].

   While TKIs have significantly advanced therapy management in
   BCR-ABL+ leukemia, providing controlled CML and long-term remission in
   BCR-ABL+ B-ALL when tailored to individual conditions, their
   limitations are evident. TKIs fall short of achieving complete cure for
   BCR-ABL+ leukemia due to their inability to eradicate cancer/leukemia
   stem cells (CSCs/LSCs), which harbor kinase-independent survival
   pathways [[55]3]. Additionally, genetic instability observed in
   BCR-ABL+ leukemias results in TKI-resistant point mutations in the
   BCR-ABL kinase domain and other genetic anomalies, constraining TKI
   efficacy. LSCs in BCR-ABL+ leukemias carrying the BCR-ABL fusion gene
   are deemed the root cause of cancer initiation, resistance, and relapse
   [[56]4–[57]6]. These LSCs operate independently of BCR–ABL for
   survival, underscoring the need to identify and target
   kinase-independent pathways [[58]3]. Thus, selective elimination of
   LSCs while sparing their normal counterparts may offer a potentially
   curative therapy against certain cancers, warranting further
   investigation.

   Ferroptosis, a form of regulated cell death (RCD), manifests as cell
   demise through intracellular deposition of a lethal amount of lipid
   reactive oxygen species (lip-ROS) mediated by iron ions [[59]7–[60]9].
   The primary mechanism of ferroptosis involves perturbation of
   intracellular lipid redox reactions. Glutathione peroxidase 4 (GPX4)
   plays a pivotal role in thwarting lipid peroxidation and ferroptosis by
   clearing lipid peroxides [[61]10, [62]11]. Recent studies have
   highlighted the importance of iron metabolism in maintaining CSCs,
   noting that the high concentration of iron in CSCs compared to non-CSCs
   underlines their iron addiction and potential vulnerability to
   ferroptosis. Given their increased resistance to standard therapies and
   tendency for relapse, CSCs’ abnormal regulation of iron
   metabolism-related proteins and dependence on iron make them a
   promising target for innovative cancer treatments that induce
   ferroptosis cell death by disrupting their iron metabolism [[63]12,
   [64]13]. Ferroptosis has thus emerged as a promising therapeutic
   strategy for curbing tumor progression and overcoming drug resistance
   [[65]14]. Targeting GPX4 offers substantial potential to enhance the
   efficacy of chemotherapy, particularly against drug-resistant cancer
   cells, including CSCs [[66]15]. Several ferroptosis-inducing agents
   have been identified to date, including System Xc⁻ inhibitors (e.g.,
   erastin and sorafenib) and GPX4 inhibitors (like RSL3). These agents,
   especially GPX4 inhibitors, are extensively utilized in preclinical
   cancer research and have demonstrated significant efficacy across
   various cancer models [[67]16, [68]17]. However, it is important to
   note that these compounds may also induce ferroptotic cell death in
   immune cells, potentially compromising the antitumor immune response
   [[69]18–[70]20]. Therefore, the development of cancer-specific
   ferroptosis inducers is crucial to mitigate adverse effects and enhance
   the therapeutic efficacy of immunotherapies.

   Disulfiram (DSF), a long-standing treatment for alcohol dependence,
   boasts well-established pharmacokinetics, safety, and tolerability at
   USA FDA-recommended doses [[71]21]. Previous studies have demonstrated
   the efficacy of DSF in preventing and treating diet-induced obesity
   [[72]22], as well as its ability to counteract cell apoptosis and
   septic cell death induced by lipopolysaccharide (LPS) in murine models
   [[73]21]. Notably, emerging research has highlighted DSF’s potent
   anti-inflammatory and anticancer properties [[74]23]. Our previous
   study has also revealed its efficacy against T-cell acute lymphoblastic
   leukemia (T-ALL), while indicating that it does not adversely affect
   normal T cells [[75]24]. These findings suggest DSF’s promise as an
   effective and safer therapeutic option for cancer treatment.

   In this study, we demonstrate that DSF, alone or in combination with
   TKIs, effectively eradicates malignant cells and LSCs of
   BCR-ABL+ leukemias, while exerting minimal effects on normal blood
   cells derived from healthy individuals (HIs). Mechanistically, DSF
   induces ferroptosis by disrupting the interaction between HSPA8 and
   GPX4, which facilitates STUB1-mediated ubiquitination and subsequent
   proteasomal degradation of GPX4. Our findings indicate that GPX4 plays
   a critical role in the anti-ferroptotic mechanisms that sustain blast
   cells and LSCs in BCR-ABL+ leukemia, positioning it as a promising
   therapeutic target for DSF in the eradication of LSCs.

Materials and methods

Clinical samples

   Peripheral blood (PB) or bone marrow (BM) samples were collected from
   patients diagnosed with CML, Philadelphia chromosome-positive B-cell
   acute lymphoblastic leukemia (BCR-ABL+ B-ALL) (see Supplemental Table
   S1), as well as from HIs. Samples were collected from the First
   Affiliated Hospital of Jinan University. Written informed consent was
   obtained from all participants before blood collection, and the study
   was approved by the ethical committee of the First Affiliated Hospital
   of Jinan University (No.2014006).

Cell culture

   Mouse 32D cells were purchased from the China Center for Type Culture
   Collection (CCTCC). The human CML cell line KBM5, K562-R, and
   32D-BCR-ABL were generously provided by Prof. Jingxuan Pan, Sun Yat-Sen
   University. Other hematological malignancy cell lines including U266,
   RPMI8226, H929, CCRF-CEM, Molt4, Jurkat, NB4, MOLM-13, KG1-α, HL-60,
   MV-411, K562, K562-R, Ku-812, TOM-1, Kasumi-1, SUP-B15 and 293T cells
   were provided by the Institute of Hematology, Jinan University.
   SUP-B15, U266, RPMI8226, H929, CCRF-CEM, Molt4, Jurkat, NB4, MOLM-13,
   KG1-α, Kasumi-1, HL-60, MV-411, K562, Ku-812, TOM-1 were cultured in
   90% RPMI 1640 with 10% Gibco fetal bovine serum (FBS). K562-R cells
   were cultured with 5µM IM under the same conditions. KBM5, and
   32D-BCR-ABL were cultured in 90% Iscove’s Modified Dulbecco’s Medium
   (IMDM) with 10% FBS. 293T cells were cultured in 90% DMEM with FBS. All
   cell lines were maintained in a humidified atmosphere containing 5%
   CO[2] at 37 °C. Plasmocin prophylactic was added to all media to
   prevent potential mycoplasma contamination.

   CD34 + cells were isolated using the CD34 Microbeads Kit (#130-046-702,
   Miltenyi Biotec, Germany) according to the manufacturer’s instructions
   and cultured in IMDM supplemented with 10% FBS along with SCF
   (#300-07-500, Proteintech; 100 ng/mL), IL-3 (#213-13-100, Proteintech;
   20 ng/mL), IL-6 (#216-16-100, Proteintech, 20 ng/mL), and GM-CSF
   (#300-03-100, Proteintech; 100 ng/mL) at 37℃ in a humidified incubator
   with 5% CO[2].

Reagents

   Disulfiram (DSF; S1680), Imatinib (IM; S1026), Dasatinib (DAS; S1021),
   Z-VAD-FMK (S7023), N-acetylcysteine (NAC; S1623), Ferrostatin-1 (Fer-1;
   S7243), Deferoxamine (DFO; S5742), MG132 (S2619), Chloroquine (CQ;
   S6999), Cycloheximide (CHX; C7698), Vitamin E (Vit-E; S4686), VX-765
   (S2228), MG132 (S2619) and Necrosulfonamide (NSA; S8251) were purchased
   from Selleck Chemicals (Houston, TX, USA). 4-Diethylaminobenzaldehyde
   (DEAB; HY-W016645) was purchased from MedChemExpress (Monmouth
   Junction, NJ, USA). Z-VAD-FMK, Fer-1, DFO, Vit-E, NSA were dissolved in
   DMSO and added to culture media at a final concentration of 10 µM. NAC
   was dissolved in DMSO and added to culture media at a final
   concentration of 1 µM.

RNA interference

   KBM5 cells were transfected using the Neon^® Transfection System
   (Invitrogen) with 100 pmol of oligonucleotides in 10 µl reactions
   [[76]25]. Briefly, 2 × 10^5 cells were suspended in 100 pmol of
   siHSPA8, siALDHA1, or siGPX4 in 10 µl reactions. After electroporation,
   cells were cultured in IMDM medium containing 10% FBS at 37 °C with 5%
   CO[2]. The transfected cells were used for subsequent experiments after
   24–48 h. siRNA sequences targeting HSPA8, STUB1, BCR-ABL and GPX4/Gpx4
   are listed in Supplemental Table S2.

In vivo siRNA delivery

   The preparation of amine-terminated, generation 5 polyamidoamine
   (G5-PAMAM, hereafter called G5) dendrimer-siRNA nanoparticles for use
   in mouse models was conducted according to the protocol established by
   Prof. Daoguang Yan from Jinan University. Briefly, in vivo siRNA
   delivery was conducted using G5 (CYD-150 A), obtained from Weihai
   Chenyuan Molecular New Materials Co., Ltd. The N: P ratio of G5 to
   siRNA was set at 30:1, and the complexes were formed in PBS at room
   temperature with gentle vortexing for 10 min. A dosage of 1.0 mg/kg of
   the G5-siRNA complex was administered via tail vein injection every 3
   days.

Establishment of stable cell lines

   A total of 3 µg of the PLKO-puro-NC or PLKO-puro-sh-GPX4 plasmid along
   with 1.5 µg of psPAX2 and 1.5 µg of pMD2.G were transfected into 293T
   cells using lipofectamine 3000 (Life Technologies, USA). Virus
   supernatant was harvested 24 and 48 h after transfection. Subsequently,
   KBM5 cells and mouse lineage- cells were co-cultured with virus
   supernatant. Then 5 µg/mL puromycin was applied to select positively
   infected cells for at least 10 days. shRNA sequences targeting
   GPX4/Gpx4 are listed in Supplemental Table S2. Lentiviruses
   overexpressing HSPA8 (LV-HSPA8, GV705), GPX4 (LV-GPX4, GOSL0438191),
   control lentivirus (LV-NC, CON525), Flag-tagged GPX4-WT (wild type),
   and Flag-tagged seven cysteines-mutated GPX4 (GPX4-7CA mutant) were
   purchased from Genechem (Shanghai, China). KBM5 cells were transduced
   with LV-HSPA8 or LV-GPX4 lentivirus, and positively transduced cells
   were selected by treatment with 5 µg/mL puromycin for a duration of 10
   days.

Cell viability assay

   Cell viability was assessed using the Cell Counting Kit-8 (CCK-8)
   (#96992, Sigma) following the manufacturer’s protocol. In brief, cancer
   cell lines were seeded at 10,000 cells per well, and primary cells at
   50,000 cells per well, in 96-well plates. After treatment, 10 µl of
   CCK-8 reagent was added to each well and incubated for 4 h. The
   absorbance was then measured using a microplate reader (Biotek
   Synergy4, USA).

Colony forming cells (CFC) assays

   GFP + lineage- c-Kit + cells sorted from BCR-ABL mice were plated in
   M3434 methylcellulose medium (Stem Cell Technologies). Cultures were
   incubated at 37 °C in a humidified atmosphere of 5% CO[2] for 10–14
   days. For CFC assays using BCR-ABL+ leukemia cell lines, 1,000 cells
   were seeded into 24-well plates after a 24-hour drug treatment.
   Subsequently, the cells were cultured in IMDM supplemented with 2% FBS
   and Methylcellulose Stock Solution (HSC001, Bio-Techne). Cultures were
   incubated at 37 °C in a humidified atmosphere of 5% CO[2] for 10 to 14
   days before counting.

Lipid peroxidation and cell death analysis

   BCR-ABL+ leukemia cell lines, primary cells from patients or mouse
   primary cells were harvested and washed three times in PBS following
   various treatments. Lipid peroxidation and cell death analysis were
   using BODIPY-C11(D3861, Invitrogen) and Annexin-V-APC/PI kit
   (AP107-100, MultiSciences, China), and analyzed by flow cytometry.

The labile iron pool assay

   The FeRhoNox-1 probe (MX4558, MKBio, China) was used to assess the
   labile iron pool. Briefly, a 500 µl cell suspension from each treatment
   group was seeded into triplicate wells of a 24-well plate.
   Subsequently, 2.5 µM FeRhoNox-1 probe was added to each well and
   incubated for 20 min at 37 °C. Flow cytometry was then conducted for
   analysis.

Malondialdehyde (MDA) assay

   The relative concentration of malondialdehyde (MDA), an end product of
   lipid peroxidation, was measured using a Lipid Peroxidation Assay Kit
   (S0131M, Beyotime, China) following the manufacturer’s instructions.

GSH and GSSG detection

   The GSH and GSSG levels of BCR-ABL+ leukemia cell lines or tissue from
   mice were measured using a GSH and GSSG Assay Kit (S0053, Beyotime,
   China) according to the manufacturer’s instructions. The total GSH and
   GSSG concentrations were calculated using a standard curve and
   normalized to the total protein level in each sample.

Immunofluorescence analysis

   Cells were washed with cold phosphate-buffered saline (PBS), fixed in
   4% paraformaldehyde for 15 min at room temperature, and permeabilized
   with 0.1% triton X100 for 10 min at room temperature. Cells were
   stained with GPX4 antibody (Santa Cruz, sc-166570; 1:100) or HSPA8
   antibody (ab51052, Abcam; 1:200) at room temperature for 1 h. After
   washing 3 times with PBS, cells were incubated with the secondary
   antibody of Alexa Fluor 555 labeled goat anti-rabbit IgG or Alexa Fluor
   488 labeled goat anti-mouse IgG at room temperature for 30 min.
   4′,6-Diamidine-2′-phenylindole dihydrochloride (DAPI) (Thermo Fisher
   Scientific) was used to stain the nucleus. MitoTracker Red (M7512,
   Invirogen) was used to stain mitochondria. Images were captured using a
   Leica SP8 confocal microscope (Leica Corp, Germany).

Western blot analysis

   Cells were harvested and lysed in RIPA buffer supplemented with
   protease inhibitors. Proteins were then separated using a 12% SDS-PAGE
   Criterion X-gel (Bio-Rad) and transferred to a PVDF membrane (Bio-Rad).
   The membranes were blocked with QuickBlock™ Blocking Buffer (Beyotime,
   China) and incubated overnight at 4 °C with primary antibodies against
   GPX4 (#59735S, Cell Signaling Technology; ab125066, Abcam; 1:1000),
   HSPA8 (#8444, Cell Signaling Technology; 1:1000), Ubiquitin (ab134953,
   Abcam; 1:2000), and Flag (AF0036, Beyotime; 1:1000). The anti-β-actin
   (ACTB) antibody (A3853, SIGMA; 1:5000) was used as a control. Bands
   were visualized using enhanced chemiluminescence (ECL; Beyotime, China)
   and the UVITEC photo documenter.

Coimmunoprecipitation (CO-IP)

   Cell protein extracts were isolated using Pierce IP Lysis Buffer
   (Thermo Fisher Scientific), then, cell lysates were incubated specific
   antibody (Anti-GPX4, sc-166570, Santa Cruz, USA) at 4 °C overnight.
   Dynabeads Protein A/G beads (CXB-PA/G0004-1, EeasunBio, Guangzhou) were
   incubated with the obtained protein complex for 6 h at 4 °C. The
   magnetic bead-antibody-antigen complex was washed three times with IP
   buffer, then eluted with loading buffer and heated at 100 °C for
   10 min. The samples were resolved by sodium dodecyl
   sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and subsequently
   analyzed by immunoblotting.

RNA extraction and quantitative real-time RT-PCR

   Total RNA from BCR-ABL+ cell lines or primary cells was extracted with
   Trizol reagent (Invitrogen) according to the manufacturer’s
   instructions. RNA was reverse transcribed into cDNA using the
   High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, CA,
   USA). Real-time PCR was conducted using SYBR Green (Tiangen, China),
   following a cycling program of 45 cycles at 95 °C for 30 s, 60 °C for
   30 s, and 72 °C for 30 s. Relative expression levels were determined
   using the comparative 2-ΔΔCt method, with data normalized to ACTB mRNA
   levels. The primers used for real-time PCR analysis are listed in
   Supplemental Table S3.

RNA-seq analysis

   KBM5 cells were treated with DMSO (Ctrl) or 0.25 µM DSF, both in the
   absence and presence of 0.25 µM IM for 48 h. Total RNA was extracted
   and assessed for RNA integrity using the RNA Integrity Number (RIN).
   cDNA library construction and sequencing were performed by Shanghai
   Biotechnology Corporation using the VAHTS Stranded mRNA-seq Library
   Prep Kit for Illumina^®. High-quality reads were aligned and mapped to
   the human reference genome (GRCh38) using Hisat2 (version 2.0.4). Gene
   expression levels were calculated in terms of fragments per kilobase of
   exon model per million mapped reads (FPKM) and are provided in
   Supplemental Table S4.

Flow cytometric analysis

   BM and spleen cells harvested from recipient mice were analyzed by flow
   cytometry (BD LSRFortessa) following staining with the specified
   antibodies. For the detection of leukemia myeloid cells, cells were
   stained with anti-mouse Gr-1-APC (E-AB-F1120UE, Elabscience),
   anti-mouse Mac-1-PE (E-AB-F1081UD, Elabscience), or anti-human
   CD45-FITC (#368508, BioLegend). To identify lineage-negative
   Sca-1 + c-Kit+ (LSK) cells, cells were stained with lineage-PerCP-Cy5.5
   (#561317, BD), Sca-1-APC (160904, BioLegend), and c-Kit-BV421 (#567818,
   BD). For the identification of long-term hematopoietic stem cells
   (LT-HSCs) and short-term hematopoietic stem cells (ST-HSCs), cells were
   stained with lineage-PerCP-Cy5.5 (#558074, BD), Sca-1-APC (#17-5981-82,
   eBioscience), c-Kit-BV421 (#105828, BioLegend), CD135-PE (#12-1351-83,
   eBioscience), CD150-PE-Cyanine7 (#25-1502-82, eBioscience), and
   CD48-APC-eFluor780 (#47-0481-82, Invitrogen). The gating strategies
   used for flow cytometry analysis are detailed in Supplemental Fig. S1A.

Molecular docking

   The molecular structures necessary for docking studies were obtained
   from the Protein Data Bank ([77]https://www.rcsb.org/) and AlphaFold
   ([78]https://alphafold.ebi.ac.uk/). For molecular docking predictions
   on the Windows platform, AutoDock 4.2.6 and PyMOL 2.5.4 were employed,
   while AutoDock Vina and GRAMM (Global RAnge Molecular Matching,
   [79]https://gramm.compbio.ku.edu/) were utilized on the Ubuntu system.

KBM5 mice

   sh-NC or sh-GPX4 KBM5 cells (2 × 10^6 cells/mouse) were transplanted
   into B-NDG mice via tail vein injection to establish CML-CDX mouse
   models (n = 8). B-NDG mice were procured from Biocytogen (Beijing,
   China). Seven days post-administration of KBM5 cells, flow cytometry
   using a BD LSRFortessa was employed to assess the expression of human
   CD45 + cells. The mice were subsequently divided into three groups:
   sh-NC, sh-GPX4-1, and sh-GPX4-2. At day 28, three mice were randomly
   selected from each group, and the weight and size of the spleen (n = 3)
   were recorded. Tumor burden in the spleen and bone marrow of each group
   (n = 3) was assessed, and the survival times of the remaining mice in
   each group (n = 5) were documented. The animal experimental protocols
   were approved by the Laboratory Animal Ethics Committee of Jinan
   University (Approval No. 20230812-12).

K562 mice

   K562 cells stably expressing both GFP and luciferase (K562-GL, 5 × 10^6
   cells/mouse, IV) were transplanted into B-NDG mice via tail vein
   injection to establish CML-CDX mouse models (n = 5). B-NDG mice were
   obtained from Biocytogen (Beijing, China). Bioluminescence imaging was
   performed using the IVIS imaging system one week post-injection of K562
   cells, following luciferin administration. The mice were then randomly
   assigned to one of four treatment groups: a control group receiving
   vehicle, an IM monotherapy group (100 mg/kg/day), a DSF monotherapy
   group (100 mg/kg/day), and a combined treatment group (100 mg/kg/day
   IM + 100 mg/kg/day DSF). Tumor burden progression was monitored weekly
   via bioluminescence imaging, and the mice were observed throughout the
   study period to evaluate survival outcomes. The animal experimental
   protocols were approved by the Laboratory Animal Ethics Committee of
   Jinan University (Approval No. 20211109-19).

BCR-ABL mice

   The Scl-tTa-BCR-ABL1/GFP (hereafter referred to as BCR-ABL) mice were
   provided by Prof. Weizhang Wang, Guangdong Pharmaceutical University.
   FVB-N mice were purchased from HFK Bio-Technology (Beijing, China). To
   evaluate the therapeutic potential of DSF and TKIs, BM cells were
   harvested from BCR-ABL mice after a 4-week induction of BCR-ABL
   expression via tetracycline withdrawal. BM GFP + cells (1 × 10^6
   cells/mouse, IV) were transplanted into lethally irradiated wild-type
   FVB/N recipient mice, which received 900 cGy of radiation. Blood
   samples were collected 4 weeks post-transplantation to confirm the
   onset of leukemia. The mice were then subjected to various treatment
   regimens: IM (100 mg/kg/day, oral gavage), DSF (100 mg/kg/day, oral
   gavage), a combination of DSF and IM, or a control group receiving only
   a vehicle (0.5% CMC-NA). After 4 weeks of treatment, a subset of
   animals from each cohort was euthanized for post-treatment analysis,
   while the remaining mice were monitored for survival. To assess the
   impact of DSF and TKIs treatment on LSC burden, 1 × 10^6 BM GFP + cells
   from treated mice were transplanted into secondary (2nd) recipient
   mice, and engraftment rates were monitored.

   To evaluate the antileukemic efficacy of GPX4 deletion in maintaining
   BCR-ABL+ leukemia, 1 × 10^6 BM GFP + cells from 4-week Tet-off BCR-ABL
   mice were transplanted into lethally irradiated primary recipients.
   These mice were divided into two groups and treated with G5-siNC or
   G5-siGpx4 for 2 weeks. Circulating blasts and LSKs were assessed, and
   survival outcomes were compared between the two groups. To further
   investigate the antileukemic efficacy of GPX4 deletion during
   leukemogenesis, 5 × 10^5 preleukemic lineage-c-Kit + cells were sorted
   and transplanted into lethally irradiated recipient mice. BCR-ABL mice
   received G5-siGpx4 (1 mg/kg/day, IV) or G5-siNC (1 mg/kg/day, IV) for 4
   weeks, starting the day after transplantation and Tet-off BCR-ABL
   induction. Circulating blasts, LSKs, and survival were compared between
   the two groups. The animal experimental protocols were approved by the
   Laboratory Animal Ethics Committee of Jinan University (Approval Nos.
   20230313-06, 20230602-13, 20230920-01, and 20240709-01).

Statistical analysis

   All statistical analyses were conducted in R language software (version
   3.8.2, [80]https://www.r-project.org/) and GraphPad Prism (version 8.0,
   CA, USA), as appropriate. Differences in two independent subgroups in
   clinical samples were tested by Mann-Whitney-Wilcoxon. For comparisons
   between two groups in vivo and in vitro experiments, two-tailed
   unpaired or paired Student’s t test was utilized. For comparisons among
   three or more groups, a one-way ANOVA followed by Bonferroni’s post hoc
   test was used. To compare a control group with multiple experimental
   groups, a one-way ANOVA with Dunnett’s post hoc test was applied,
   unless otherwise specified. The “surv_cutpoint” function in the
   “survminer” package was used to determine the optimal cut-off values of
   GPX4. Kaplan-Meier curves were compared by log-rank test using the R
   package “survival”. The differentially expressed genes in RNA-seq data
   were identified utilizing the package “edgeR”. Gene Ontology (GO) and
   Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment were
   analyzed by the packages “clusterProfiler” and “org.HS.eg.db”. A
   two-tailed P value < 0.05 was considered statistically significant.

Result

The combination of DSF and TKIs synergistically promotes cell death in
BCR-ABL+ leukemia cells

   DSF, a well-established treatment for alcohol dependence for over seven
   decades, exhibits established pharmacokinetics, safety, and
   tolerability at doses recommended by the USA FDA. Recent studies unveil
   DSF’s remarkable anti-cancer properties [[81]26]. To evaluate its
   impact on blood cancer, we subjected various cancer cell lines to
   varying DSF concentrations, revealing a dose-dependent inhibition of
   cancer cell growth. Notably, BCR-ABL+ cells exhibited heightened
   sensitivity to DSF (Fig. [82]1A). Subsequently, two BCR-ABL+ cell
   lines, KBM5 (CML) and TOM-1 (BCR-ABL+ B-ALL), demonstrated significant
   induction of cell death by DSF in a dose-dependent manner
   (Fig. [83]1B). DSF also markedly impeded colony number and size in KBM5
   and TOM-1 cells (Fig. [84]1C). Further investigations explored DSF’s
   potential to enhance cytotoxic effects in cells from newly diagnosed
   BCR-ABL+ leukemia patients, consistently promoting cell death in
   primary BCR-ABL+ leukemia cells (Fig. [85]1D). To discern DSF’s
   anti-leukemia effects in cells harboring BCR-ABL relative to BCR-ABL-
   cells, we included BCR-ABL1-transformed 32D-BCR-ABL and parental mouse
   hematopoietic 32D cells. The results indicated more pronounced cell
   death in 32D-BCR-ABL cells compared to 32D cells under equivalent
   concentrations (Fig. S1B), emphasizing DSF’s robust cytotoxic effect on
   BCR-ABL+ leukemia cells.

Fig. 1.

   [86]Fig. 1
   [87]Open in a new tab

   DSF cooperates with TKIs to induce the cell death of BCR-ABL+ leukemia
   cells. A A panel of BCR-ABL+ leukemia cells and cancer cell lines were
   treated with increasing concentrations of DSF for 48 h and the cell
   viability was determined by CCK-8 assay (left panel), and the IC50 was
   calculated by CompuSyn 2.0 (right panel). B BCR-ABL+ leukemia cell
   lines were treated with the indicated concentrations of DSF (left
   panel), and the percentage of Annexin V + cells was detected by the
   flow cytometry (right panel). C Effect of DSF on the colony-forming
   ability of BCR-ABL+ leukemia cells. Representative results are shown in
   the left panel, with quantification from three experiments displayed in
   the right panel. D BM MNCs from newly diagnosed patients with CML
   (n = 6) or BCR-ABL+ B-ALL (n = 6) were exposed to indicated
   concentrations of DSF for 48 h, and the percentage of Annexin V + cells
   was detected by the flow cytometry. E Heatmaps of drug combination
   responses were generated using the online SynergyFinder software
   ([88]https://www.synergyfinder.fimm.fi/), assessing the synergistic
   effects of DSF + TKIs in KBM5 and TOM-1 cells. Synergistic effects were
   indicated by scores over 0, and those surpassing 10 suggested strong
   synergistic effects. In these heatmaps, red signals synergism, while
   green represents antagonism. The drug combinations exhibiting the most
   potent synergistic effects are highlighted with white squares (left
   panel). KBM5 and TOM-1 cells were treated with DSF alone, TKIs alone,
   or the combination of DSF and TKIs at the indicated concentrations for
   48 h and the cell viability was measured by CCK8 assay (right panel). F
   KBM5 and TOM-1 cells were treated with DSF (0.25 µM) alone, IM (0.25
   µM) for KBM5 cells, DAS (0.25 µM) for TOM-1 cells, or the combination
   of DSF and TKIs at the indicated time points, and cell viability was
   measured by CCK8 assay(two-way ANOVA test, data are presented as
   mean ± SD). G KBM5 (upper panel) and TOM-1 cells (lower panel) were
   treated with 0.25 µM DSF, 0.25 µM TKIs or the combination of DSF and
   TKIs, and the percentage of Annexin V + cells was determined by flow
   cytometry. H Representative image and quantification of the
   colony-forming ability of BCR-ABL+ leukemia cells treated with DSF
   (0.25 µM), with or without IM (0.25 µM). I Cell death in Ku-812 and
   SUP-B15 cells with indicated treatment (0.25 µM DSF, 0.25 µM TKIs or
   the combination of DSF and TKIs for 48 h) was evaluated by flow
   cytometry labeling with Annexin-V and PI. J-L Cell viability of BM MNCs
   from newly diagnosed patients with CML (n = 10) or BCR-ABL+ B-ALL
   (n = 5) or HI PB (n = 3) with indicated treatment (0.25 µM DSF, 0.25 µM
   TKIs or the combination of DSF and TKIs) for 48 h were measured by
   CCK-8 assay (J) and flow cytometry(K-L). Data are expressed as the
   mean ± SD. n = 3 or more independent biological replicates, * P
    < 0.05; ** P  < 0.01**** P  ≤ 0.0001

   Given that TKIs represent the current first-line therapy for
   BCR-ABL+ leukemia, we investigated whether DSF synergized with TKIs in
   BCR-ABL+ leukemia cells. We chose DAS for TOM-1 cells (BCR-ABL+ B-ALL)
   due to its ability to overcome various mechanisms of imatinib
   resistance and its enhanced activity against BCR-ABL. For KBM5 cells
   (CML), we selected IM, as it remains the standard treatment and is
   effective in achieving cytogenetic responses. Treatment of KBM5 and
   TOM-1 cells with TKIs alone or in combination with low dose DSF
   (0.25µM) induced a synergistic reduction in cell activity
   (Fig. [89]1E). Further experiments involving time gradients of the
   drugs consistently demonstrated a synergistic effect between DSF and
   TKIs (Fig. [90]1F-G). Importantly, combination treatment led to
   increased cell death and notable reduction in colony number and size
   (Fig. [91]1H). Similar findings were observed in two additional
   BCR-ABL+ cell lines, KU-812 and SUP-B15 (Fig. [92]1I and Fig. S1C).
   This synergistic effect was minimally observed in normal human PB
   mononuclear cells (MNCs) but significantly affected BM MNCs from
   BCR-ABL+ leukemia patients (Fig. [93]1J-L). Collectively, these results
   underscore the synergistic effect between DSF and TKIs.

DSF induces BCR-ABL+ leukemia cell ferroptosis

   To delve deeper into the potential pathways underlying the effects of
   DSF in BCR-ABL+ leukemia cells, we analyzed the enriched pathways of
   differentially expressed genes in KBM5 cells treated with Ctrl (DMSO),
   IM, DSF, or IM + DSF for 48 h (Fig. [94]2A). Comparing the DSF + IM
   treated cells with the Ctrl, we observed significant enrichment changes
   in the oxidative stress pathway and glutathione pathway (Fig. [95]2B,
   left panel). Furthermore, comparing the DSF + IM treated cells with the
   IM treated cells, we found significant changes in pathway enrichment
   related to ferroptosis, suggesting that DSF may induce cell death in
   the KBM5 cell line through the ferroptosis pathway (Fig. [96]2B, right
   panel). To confirm these results, we utilized ferroptosis inhibitors
   Fer-1, Vit-E, and DFO, oxidative stress inhibitors NAC, and cell death
   inhibitors NSA, VX-765, and z-VAD-FKM on DSF-induced cell death.
   Notably, DSF-induced cell viability reduction could be rescued by
   Fer-1, Vit-E, DFO, and NAC (Fig. [97]2C), but not by Z-VAD-FMK, NSA,
   VX-765, indicating that DSF indeed induces leukemic cell death through
   the ferroptosis and oxidative stress pathways.

Fig. 2.

   [98]Fig. 2
   [99]Open in a new tab

   DSF induces BCR-ABL+ leukemia cell ferroptosis. A Schema of the
   experimental design to define the pathway in DSF-treated
   BCR-ABL+ leukemia cells. B Bubble diagram of KEGG pathway enrichment
   analysis was performed in KBM5 cells. The top 16 pathways most
   significantly changed in DSF + IM group compared with Ctrl group (left
   panel). The top 12 pathways most significantly changed in DSF + IM
   group compared with IM group (right panel). C The inhibitory effects of
   DSF (0.25 µM), IM (0.25 µM), or the combination of DSF and IM on KBM5
   cells were assessed in the presence of the indicated inhibitors using
   the CCK-8 assay. D Fluorescence intensities of BODIPY-C11 in DSF
   treated KBM5 and TOM1 cells by flow cytometry. E The MDA in cell
   lysates of KBM5(up panel) and TOM-1(lower panel) were treated with the
   indicated concentrations of DSF for 36 h. F-G KBM5 and TOM-1 were
   treated with 0.25 µM DSF, 0.25 µM TKIs, or a combination of DSF and
   TKIs for 48 h. Fluorescence intensities of BODIPY-C11 (F) and ROS level
   (G) were determined by flow cytometry. H The MDA concentration of BM
   MNCs from newly diagnosed patients with CML treated with DSF (0.25 µM)
   or TKIs (0.25 µM) alone, or a combination of DSF and TKIs for 48 h.
   I The primary cells from HIs, CML patients and BCR-ABL+ B-ALL patients
   were treated with DSF (0.25 µM) alone, TKIs (0.25 µM) alone, or a
   combination of DSF and TKIs for 48 h. The levels of lipid peroxidation
   were determined by flow cytometry using the BODIPY-C11 probe.
   Representative results are shown in (left panel), and quantification
   are shown in (right panel). Data are expressed as the mean ± SD. n = 3
   or more independent biological replicates, * P  < 0.05; ** P  < 0.01,
   *** P  ≤ 0.001, **** P  ≤ 0.0001

   Moreover, we assessed key biochemical events of ferroptosis, including
   lipid peroxidation measured by fluorescence probes BODIPY-C11 and
   malondialdehyde (MDA), a product of lipid peroxidation, in the KBM5 and
   TOM-1 cell lines with graded doses of DSF. We observed a dose-dependent
   increase in lipid peroxidation and MDA (Fig. [100]2D-E). Flow cytometry
   analysis of lipid peroxidation also showed an increase in DSF-induced
   ferroptosis after combined treatment with TKIs (Fig. [101]2F),
   accompanied by an increase in reactive oxygen species (ROS)
   (Fig. [102]2G), MDA (Fig. S1D) and a decrease in GPX4 enzymatic
   activity (Fig. S1E). These results further underscore the crucial role
   of ferroptosis in DSF-induced cell death in BCR-ABL+ cell lines.

   Furthermore, we measured MDA levels in primary BM cells from CML
   patients treated with DSF or in combination with IM. Consistent with
   the cell line results, DSF combined with IM treatment led to increased
   MDA in CML cells (Fig. [103]2H). Additionally, we analyzed lipid
   peroxidation levels and found that the combination of DSF and IM
   induced lipid peroxidation in BCR-ABL+ leukemia cells, while having no
   significant impact on HIs (Fig. [104]2I). Interestingly, compared with
   BCR-ABL+ cell lines, DSF had no impact on lipid peroxidation of acute
   myeloid leukemia (AML) cells, consistent with previous studies
   demonstrating that DSF induces apoptosis in AML cells (Fig. S1F). These
   data collectively indicate that DSF induces ferroptosis in
   BCR-ABL+ leukemia cells.

DSF induces ferroptosis by repressing GPX4

   Subsequently, we investigated the expression of GPX4, closely
   associated with ferroptosis, in BCR-ABL+ cells following DSF treatment.
   While mRNA expression showed no significant changes (Fig. [105]3A),
   protein levels of GPX4 decreased in response to DSF alone or in
   combination with TKIs in KBM5 and TOM-1 cells (Fig. [106]3B). Moreover,
   DSF in combination with IM led to a significant reduction in GPX4
   protein levels in a time-dependent manner (Fig. S2A). Confocal
   microscopy analysis shows that cytosolic GPx4 (cGPx4) is the key
   isoform dominantly expressed relative to mitochondrial GPx4 (mGPx4) and
   nucleolar GPx4 (nGPx4), and further confirmed decreased GPX4 protein
   levels in DSF and DSF + IM-treated KBM5 cells (Fig. [107]3C and Fig.
   S2B). Compared to BCR-ABL+ cells, we found that DSF treatment does not
   induce GPX4 repression in multiple myeloma cell lines with low
   sensitivity to DSF (Fig. S2C). These findings suggest that DSF may
   induce BCR-ABL+ cell ferroptosis through GPX4 suppression.

Fig. 3.

   [108]Fig. 3
   [109]Open in a new tab

   DSF induce ferroptosis through GPX4 targeting. A The mRNA level of GPX4
   in RNA-seq data and the FPKM values were used to compare differences in
   gene expression among samples. B The protein level of GPX4 in
   BCR-ABL+ leukemia cell lines treated with the indicated concentrations
   of DSF, or DSF (0.25 µM) alone, TKIs (0.25 µM) alone, or a combination
   of DSF and TKIs for 48 h. C Representative immunofluorescence images of
   GPX4 protein in KBM5 cells treated with DSF (0.25 µM) alone, IM (0.25
   µM) alone, or a combination of DSF and IM for 48 h. Nuclei are stained
   with DAPI (×10). Scale bars, 5 μm. D Western blot assay of GPX4 and
   ACTB proteins in KBM5 cells transfected with sh-GPX4 or its control
   sh-NC. E The colony formation assay was utilized to detect cell
   proliferation of KBM5 cells transfected with sh-GPX4 or its control
   sh-NC. F-H KBM5 cells transfected with sh-GPX4 were treated with 0.25
   µM IM for 48 h, Cell viability (F), cell death (G), and fluorescence
   intensities of BODIPY-C11(H) were analyzed using the CCK8 assay or flow
   cytometry. I Western blot analysis of the GPX4 expression in
   GPX4-overexpressing KBM5 cells. Cell viability (J), cell death analysis
   (K), and fluorescence intensities of BODIPY-C11 (L) of
   GPX4-overexpressing KBM5 cells treated with 0.25 µM DSF or 0.25 µM IM
   for 48 h. Data are expressed as the mean ± SD. n  = 3 or more
   independent biological replicates, * P  < 0.05; ** P  < 0.01, *** P
    ≤ 0.001, **** P  ≤ 0.0001

   To explore the relationship between GPX4 and DSF, we generated KBM5
   cell lines with GPX4 knockdown (KD) (Fig. [110]3D). Depletion of GPX4
   markedly reduced the clonogenic potential of KBM5 cells in
   colony-forming assays (Fig. [111]3E), significantly impaired their
   proliferative capacity (Fig. [112]3F), and heightened ferroptosis in
   response to IM treatment (Fig. [113]3G-H). Consistent with the effect
   of GPX4 depletion, the combination of the ferroptosis inducer Fin56 and
   TKIs more effectively inhibited BCR-ABL+ leukemia cells than each
   treatment alone (Fig. S3A-C). Finally, we overexpressed GPX4 in KBM5
   cell lines (Fig. [114]3I), and GPX4-overexpressing KBM5 cells
   demonstrated resistance to growth inhibition and ferroptosis by IM and
   DSF (Fig. [115]3J-L). The effect of DSF has previously been shown to
   ALDH1A1. Next, we also explored whether ALDH1A1 mediates DSF induced
   ferroptosis. Our results indicate that pharmacological and genetic
   ALDH1A1 inhibition does not significantly enhance the effects of GPX4
   KD or DSF treatment on KBM5 or leukemic stem cells (Fig. S4A-C).
   Collectively, our data reveal that DSF induces ferroptosis by
   suppressing GPX4 expression.

DSF sensitized TKI resistant cells and leukemia stem cells to TKIs

   Despite the continuous improvement in efficacy with the use of
   second-generation and third-generation TKIs, resistance to TKIs,
   disease progression, and recurrence due to resistant cells or leukemic
   stem cells (LSCs) remain formidable challenges in CML treatment. To
   address this, we investigated whether the combination of DSF and TKIs
   could overcome TKI resistance in cell lines. Consequently, we employed
   CCK8 and flow cytometry to assess changes in cell viability and cell
   death in IM resistant cells K562-R following the combination of DSF and
   IM. The results unveiled a clear synergistic effect of the two drugs on
   IM-resistant K562-R cells (Fig. [116]4A-B). Additionally, we measured
   changes in GPX4 protein levels and lipid peroxidation after combination
   treatment, indicating that the combination of DSF and IM could induce
   ferroptosis in these cells (Fig. [117]4C-D). Furthermore, sorted
   CD34 + cells from newly diagnosed CML patients were treated with IM and
   DSF. The combination of the two drugs increased cell death, inhibited
   cell viability, and augmented lipid peroxidation in CML CD34 + cells
   (Fig. [118]4E-G and Fig. S4D).

Fig. 4.

   [119]Fig. 4
   [120]Open in a new tab

   Dual BCR-ABL and GPX4 inhibition more effectively targets IM-resistant
   cells and LSCs than either treatment alone. A-D Prior to the
   experiment, K562-R cells were cultured in an IM-free medium for 48 h.
   IM-resistant cells were treated with IM (5 µM) alone, DSF (0.25 µM)
   alone, or a combination of DSF and IM for 48 h (A-C) or 56 h (D). Their
   responses were assessed using CCK8, western blot and flow cytometry to
   measure cell viability (A), cell death (B), the protein levels of GPX4
   (C), and fluorescence intensities of BODIPY-C11 (D). E-G CD34 + cells
   from CML patients were treated with IM (0.25 µM) alone, DSF (0.25 µM)
   alone, or a combination of DSF and IM for 48 h. The effects were
   evaluated using CCK8 and flow cytometry to assess cell viability (E),
   cell death (F), and fluorescence intensities of BODIPY-C11 (G). H-I
   Mouse LSK cells from BCR-ABL mice 4 weeks after induction of BCR-ABL
   expression were treated with IM (0.25 µM) alone, DSF (0.25 µM) alone,
   or a combination of DSF and IM. Their responses were evaluated using
   CCK8 (H) and colony formation assays (I). Data are expressed as the
   mean ± SD. n = 3 or more independent biological replicates, * P
    < 0.05; ** P  < 0.01, *** P  ≤ 0.001, **** P  ≤ 0.0001

   To further validate this discovery, we utilized the stem cell promoter
   (Scl)-driven transgenic Scl-tTa-BCR-ABL1 (hereafter referred to as
   BCR-ABL) mouse model. Withdrawal of tetracycline (Tet) induces BCR-ABL
   expression and the development of BC CML-like disease in these mice.
   Mouse LSCs (LSK cells) were sorted from Tet-off BCR-ABL mice, revealing
   a dramatic decrease in cell viability and clonogenic potential when LSK
   cells were treated with IM in combination with DSF (Fig. [121]4H-I).
   Overall, these results indicate that the combination of DSF and TKIs
   can effectively promote the death of drug-resistant and DSF-insensitive
   cells, including drug-insensitive LSCs, in BCR-ABL+ leukemia.

GPX4 is highly expressed in BCR-ABL+ leukemia, and its elevated expression is
associated with a poor prognosis in BCR-ABL+ ALL

   Given GPX4’s essential role in inhibiting ferroptosis, our data
   demonstrate that DSF induces ferroptosis and decreases GPX4 expression.
   Based on this, we hypothesized that BCR-ABL+ leukemia cells may rely on
   high GPX4 levels. To investigate, we compared GPX4 protein levels in PB
   MNCs from CML patients, BCR-ABL+ B-ALL patients, and HIs. We found
   elevated GPX4 protein levels in primary BCR-ABL+ leukemia samples
   (Fig. [122]5A-B), consistent with observations in BCR-ABL+ leukemia
   cell lines compared to HIs (Fig. [123]5C). Additionally, CD34 + cells
   from CML patients exhibited higher GPX4 levels compared to those from
   HIs (Fig. [124]5D). Furthermore, in the BCR-ABL mouse model, we
   observed increased Gpx4 expression in BM MNCs (Fig. [125]5E, left and
   middle panel) and LSK cells (Fig. [126]5E, right panel) from Tet-off
   BCR-ABL mice. Consistently, Gpx4 was increased in 32D-BCR-ABL(p210)
   cells compared to their parental 32D cells (Fig. [127]5F).
   Overexpression of BCR-ABL1(p190) also led to increased Gpx4 protein
   levels (Fig. S5A), while BCR-ABL KD attenuated GPX4 expression at both
   mRNA and protein levels (Fig. S5B). These results suggest that BCR-ABL
   promotes GPX4 expression and contributes to BCR-ABL+ leukemia
   development.

Fig. 5.

   [128]Fig. 5
   [129]Open in a new tab

   GPX4 is highly expressed in BCR-ABL+ leukemia cells and is associated
   with a poor prognosis in BCR-ABL+ B-ALL. A-B Protein levels of GPX4
   were assessed in primary cells from CML (A) and BCR-ABL+ B-ALL leukemia
   (B) patients and HIs. C GPX4 protein levels were examined in
   BCR-ABL+ leukemia cell lines, and primary cells from HIs. D RNA levels
   of GPX4 were evaluated in CD34 + cells obtained from HIs and CML
   patients. E Expression of GPX4 in cells from BCR-ABL mice before or
   after tetracycline withdrawal. Representative western blot (left panel)
   and immunofluorescence images (middle panel) of GPX4 protein in BM
   cells. RNA levels of GPX4 in LSK cells from BCR-ABL mice were also
   examined (right panel). F Protein levels of GPX4 were measured in 32D
   and 32D-BCR-ABL cells (left panel). Representative immunofluorescence
   images of GPX4 in these cells were analyzed (right panel). G Levels of
   lipid reactive oxygen species were determined in PB MNCs from HIs and
   CML patients. H Fluorescence intensities of BODIPY-C11 in 32D and
   32D-BCR-ABL cells following treatment with IM (0.25 µM) alone, DSF
   (0.25 µM) alone, or a combination of DSF and IM for 48 h were measured.
   I Kaplan‒Meier analysis was performed to evaluate the associations of
   GPX4 expression with overall survival (OS) in patients with
   BCR-ABL+ B-ALL in the [130]GSE34941-GP11088. The cutoff value is 13.9.
   J GPX4 expression levels in 477 newly diagnosed BCR-ABL+ CML patients
   and 74 patients receiving TKI treatment more than 1 year. Data are
   expressed as the mean ± SD. n = 3 or more independent biological
   replicates, * P < 0.05; ** P < 0.01, *** P ≤ 0.001, **** P ≤ 0.0001

   Concurrent with high GPX4 expression, CML patient BM MNCs exhibited
   lower lipid peroxidation levels compared to HIs, indicating ferroptosis
   inhibition (Fig. [131]5G). This observation was consistent in
   32D-BCR-ABL(p210) and 32D cells (Fig. S5C-D). Notably, CML leukemic
   cells displayed significantly increased Fe2 + levels (Fig. S5E). Next,
   we examined the effect of DSF on BCR-ABL+ cell death. DSF alone or in
   combination with IM markedly induced cell death (Fig. S5F) and lipid
   peroxidation (Fig. [132]5H) in 32D-BCR-ABL cells with high Gpx4
   expression, but not in their parental 32D cells. Next, we further
   investigated the relationship between GPX4 and the prognosis of
   BCR-ABL+ B-ALL patients. Analysis of RNA-seq data from bone marrow
   samples in the [133]GSE34941-GP11088 dataset revealed a notable trend:
   high GPX4 expression was associated with poor overall survival (OS) in
   BCR-ABL+ B-ALL patients (Fig. [134]5I). Although this association did
   not achieve statistical significance, it may be influenced by the
   limited sample size. Furthermore, we examined the correlation between
   GPX4 expression levels and clinical outcomes in BCR-ABL+ CML patients
   using the [135]GSE76312 dataset. The results indicated a significant
   decrease in GPX4 expression in CML patients after ≥ 1 year of TKI
   treatment (Fig. [136]5J). Collectively, these findings suggest that
   elevated GPX4 expression is significantly linked to adverse clinical
   outcomes in both BCR-ABL+ B-ALL and CML patients.

Targeting GPX4 compromises BCR-ABL+ leukemia initiation/progression

   The results presented above demonstrate the elevated levels of GPX4 in
   BCR-ABL+ leukemia and the inhibitory effect of GPX4 KD on ferroptosis
   induction and leukemia cell growth in vitro. Subsequently, we
   investigated the role of GPX4 in a xenograft mouse model
   (Fig. [137]6A). Consistent with our previous findings, GPX4 KD via
   shRNAs led to a substantial reduction in the number of CML leukemia
   cells in the BM (Fig. [138]6B). Notably, compared to the sh-NC group,
   GPX4 KD resulted in significantly decreased spleen size and weight
   (Fig. [139]6C), ultimately leading to prolonged survival in recipient
   mice (Fig. [140]6D).

Fig. 6.

   [141]Fig. 6
   [142]Open in a new tab

   GPX4 is essential for LSC development and the maintenance of BCR-ABL+
   leukemia. A Schema of the experimental strategy to define the
   dependence of CML cells on GPX4 in the maintenance of leukemia. (n = 8
   mice per group). B Flow cytometry analysis of the CD45 + cell in the BM
   of CML mice. C Representative images of spleen. Statistical analysis of
   spleen weight is presented. D Kaplan–Meier survival curves showing
   effect of GPX4 KD on leukemia maintenance in immunodeficient mice.
   subsequent to sh-GPX4 KBM5 cell transplantation (n = 5). E Schematic
   overview of G5-siGpx4 nanoparticle therapy of BCR-ABL+ leukemia mice.
   (n = 8 mice per group for this experiment). F The frequency of LSKs,
   LT-HSCs, and ST-HSCs was analyzed by flow cytometry.
   G-H Mac-1 + Gr1 + myeloid cells in spleen (G) and BM (H). I-L Survival
   (I), spleen weight (J), the ratio of GSH and GSSG (K), MDA (L), and CFC
   (M) of the G5-siGpx4-treated vs. G5-siNC-treated mice are shown.
   N-R Experimental design (N). To assess the impact of G5-siGpx4
   treatment on the LSC development, lineage-c-Kit + cells from BCR-ABL
   mice (BCR-ABL were induced for two days by Tet-off) were transplanted
   into recipient mice, starting on the day after transplantation, mice
   were treated G5-siGpx4 or G5-siNC. Spleen weight (O),
   Mac-1 + Gr1 + myeloid cells (P), LSK cells (Q), Survival (R), the ratio
   of GSH/GSSG (S) and MDA(T) of the G5-siGpx4-treated vs. G5-siNC-treated
   mice are shown. Data are expressed as mean ± SD from three or more
   independent biological replicates. Statistical significance is denoted
   as * P < 0.05; ** P < 0.01; *** P ≤ 0.001; **** P ≤ 0.0001

   We then examined whether GPX4 deletion impacts BCR-ABL+ leukemia
   maintenance in the BCR-ABL mouse model (Fig. [143]6E). For gene
   delivery, we utilized amine-terminated, G5 (G5-PAMAM) dendrimers. After
   transplantation of cells from 4-week Tet-off BCR-ABL mice into lethally
   irradiated primary recipients, the mice were divided into two groups
   and treated with G5-siNC or G5-siGpx4, followed by an assessment of
   leukemia cell engraftment in the BM. As anticipated, a profound
   reduction in LSCs and quiescent LSCs (Fig. [144]6F) was observed in the
   BM of mice treated with G5-siGpx4 nanoparticles compared with G5-siNC.
   Notably, the siGpx4 nanoparticles significantly decreased myeloid cells
   (Gr-1 + Mac-1+) in the spleen (Fig. [145]6G) and BM (Fig. [146]6H),
   ultimately prolonging the survival of recipient mice (Fig. [147]6I).
   Additionally, compared with the G5-siNC group, the spleen size and
   weight were significantly decreased in the G5-siGpx4 group
   (Fig. [148]6J), accompanied by reduced leukemia cell infiltration in
   spleen and BM (Fig. S6A). Moreover, reduced Gpx4 protein levels were
   observed in the spleens of G5-siGpx4 group mice compared with the
   control group (Fig. S6B). Consistent with the in vitro findings,
   G5-siGpx4 nanoparticles-mediated Gpx4 KD resulted in high levels of
   malondialdehyde (MDA) and low levels of reduced glutathione (GSH) in
   the spleen and BM (Fig. [149]6K-L). Moreover, Gpx4 KD significantly
   inhibited colony-forming capacity in BM cells from Tet-off BCR-ABL mice
   (Fig. [150]6M).

   To explore the requirement for Gpx4 in leukemogenesis, we employed the
   BCR-ABL mouse model (Fig. [151]6N). In this model, BM preleukemic
   lineage-c-kit + cells were sorted and transplanted into lethally
   irradiated recipient mice. Withdrawal of Tet led to the induction of
   BCR-ABL expression, generating self-renewing LSCs, and causing BCR-ABL+
   leukemia. Compared with the G5-siNC group, the spleen size and weight
   were significantly decreased in the G5-siGpx4 group (Fig. [152]6O),
   accompanied by reduced myeloid cells (Gr-1 + Mac-1+, Fig. [153]6P) and
   LSCs (LSK cells, Fig. [154]6Q). Additionally, preleukemic cells treated
   with G5-siGpx4 exhibited a significantly longer latency period for the
   onset of leukemia compared to controls (Fig. [155]6R). Consistently,
   the anti-leukemia effects of G5-siGpx4 nanoparticles were associated
   with a substantial decrease in GSH levels (Fig. [156]6S) and an
   increase in MDA levels (Fig. [157]6T) in the spleen and BM. Moreover,
   in the presence of DSF, GPX4 depletion led to the lowest survival rate
   of LSCs, as indicated in Fig. S6C-D. Collectively, our data suggest
   that GPX4 is required for the maintenance and initiation/development of
   BCR-ABL+ leukemia.

DSF promotes GPX4 degradation by inhibiting HSPA8 binding to GPX4

   We then delved into the molecular mechanism by which DSF regulates the
   protein level of GPX4. To explore the impact of DSF on the expressions
   of GPX4 upon CHX treatment, we observed a significant decrease in GPX4
   protein half-life in KBM5 cells treated with DSF (Fig. [158]7A).
   Furthermore, we utilized MG132 (a proteasome pathway inhibitor) and CQ
   (an autophagy inhibitor) to treat CML cells with or without DSF
   treatment. Our findings revealed that MG132 could reverse the effect of
   DSF on GPX4 protein degradation, while CQ did not (Fig. [159]7B).
   Co-immunoprecipitation (Co-IP) assays revealed substantial
   polyubiquitination of GPX4 in DSF-treated cells (Fig. [160]7C),
   indicating that DSF modulates GPX4 protein stability via the
   ubiquitin-proteasome pathway. Utilizing UbiBrowser
   ([161]http://ubibrowser.bio-it.cn/ubibrowser/), we identified several
   potential interactors (HSPA8 and STUB1) of GPX4 (Fig. S7A). Given that
   HSPA8 is regulated by BCR-ABL and plays a vital role in leukemia cell
   survival [[162]27], we explored whether HSPA8 affects GPX4 expression.
   Molecular docking analysis suggested a potential interaction between
   HSPA8 and GPX4 (Fig. [163]7D). Co-IP analysis further indicated the
   GPX4-HSPA8 interaction, which was disrupted upon DSF treatment
   (Fig. [164]7E). Previous research has demonstrated that DSF binds to
   multiple cysteine residues on GPX4, disrupting its interaction with
   HSC70 (HSPA8) and thereby preventing erastin-induced autophagy-mediated
   GPX4 degradation [[165]28]. In line with these findings, we observed
   that GPX4 with seven cysteine site mutations reduced DSF’s effects.
   However, while prior studies indicated that this mutation impaired
   DSF’s ability to block erastin-induced autophagy-mediated degradation
   of GPX4, our results suggest that it instead inhibits DSF-induced
   ubiquitin-mediated degradation of GPX4 (Fig. [166]7F). To assess
   whether HSPA8 regulates the post-translational expression of GPX4, we
   downregulated HSPA8 in KBM5 cells and observed a significant decrease
   in GPX4 protein levels (Fig. [167]7G). HSPA8 KD also led to GPX4
   suppression, which was reversible by MG132 (Fig. [168]7H), suggesting
   that HSPA8 influences GPX4 expression via the ubiquitin-proteasome
   pathway. Conversely, HSPA8 overexpression increased GPX4 protein levels
   (Fig. [169]7I) and significantly reduced GPX4 polyubiquitination,
   enhancing GPX4 protein stability even after DSF treatment
   (Fig. [170]7J-K). As seen with GPX4 KD, HSPA8 KD significantly induced
   cell death and lipid peroxidation in DSF-treated KBM5 cells (Fig. S7B).
   To explore if DSF might inhibit GPX4 through HSPA8, we knocked down
   GPX4 in KBM5 cells overexpressing HSPA8 (Fig. [171]7L). In these cells,
   HSPA8 overexpression led to reduced cell death and lipid peroxidation
   upon GPX4 KD (Fig. [172]7M). Notably, HSPA8 expression did not decrease
   with DSF treatment, indicating that DSF’s inhibitory effect on GPX4 may
   not involve HSPA8 downregulation (Fig. S7C-D). Recent studies have
   identified STUB1 as a novel ubiquitin E3 ligase targeting GPX4
   [[173]29]. In line with this, Co-IP analysis confirmed the interaction
   between GPX4 and STUB1, which was enhanced by DSF treatment
   (Fig. [174]7N). Silencing STUB1 increased GPX4 levels (Fig. [175]7O)
   and suppressed DSF-induced cell death and lipid peroxidation (Fig.
   S7E-F). Furthermore, STUB1 KD significantly inhibited GPX4
   ubiquitination (Fig. [176]7P). Immunoblot analysis following GPX4 IP
   revealed that elevated HSPA8 levels reduced the interaction between
   GPX4 and STUB1 (Fig. [177]7Q). These findings suggest that DSF
   modulates GPX4 stability via HSPA8, promoting STUB1-mediated GPX4
   ubiquitination and proteasomal degradation.

Fig. 7.

   [178]Fig. 7
   [179]Open in a new tab

   GPX4 degradation and Ferroptosis induced by DSF was dependent on
   HSPA8. A KBM5 cells were treated with CHX alone (20 µM) or in
   combination with DSF (0.25 µM) or MG132 (0.25 µM). The protein level of
   GPX4 was assayed by western blots. B Western blot assay of GPX4 in KBM5
   cells treated with DMSO, DSF (0.25 µM), or MG132 (0.25 µM) and CQ (25
   µM) for 48 h. C Anti-Ubiquitin immunoblotting assay of GPX4
   polyubiquitination in KBM5 cells treated with DSF (0.25 µM, 36 h) or
   control. D Molecular docking predicted a direct interaction between
   HSPA8 and GPX4. E Co-immunoprecipitation (Co-IP) of HSPA8 and GPX4
   proteins in the whole cell lysates of KBM5 cells with or without DSF
   treatment (0.25 µM, 36 h). F GPX4-reconstituted cells were treated with
   0.25 µM DSF for 46 h, and cells were collected for Western blot
   analysis. G qPCR analysis of HSPA8 (left panel) expression in KBM5
   cells transfected with siHSPA8 or siNC. Western blot analysis of
   GPX4(right panel). H Western blot analysis of GPX4 expression in KBM5
   cells transfected with siNC, siHSPA8-1, or siHSPA8-2 and treated with
   DMSO or MG132 (0.25 µM, 48 h). I qPCR analysis of HSPA8 (left panel)
   expression in HSPA8-overexpressing KBM5 cells. Western blot analysis of
   GPX4 (right panel) expression in these cells is also shown. J Western
   blot analysis of GPX4 expression in KBM5 cells transduced with LV-NC or
   LV-HSPA8 and treated with DSF or control. K The effect of HSPA8 on the
   ubiquitination modification of GPX4. L Western blot analysis of GPX4
   expression in LV-NC or LV-HSPA8 transfected KBM5 cells following GPX4
   KD. M Cell death (left panel) and fluorescence intensities of
   BODIPY-C11 (right panel) in LV-NC or LV-HSPA8 transfected KBM5 cells
   following GPX4 KD are displayed. N Co-IP of STUB1 and GPX4 proteins in
   the whole cell lysates of KBM5 cells with or without DSF treatment. O
   qPCR analysis of STUB1 (upper panel) expression in KBM5 cells
   transfected with siSTUB1 or siNC. Western blot analysis of GPX4(lower
   panel). P The effect of STUB1 KD on the ubiquitination modification of
   GPX4. Q The effect of HSPA8 on the STUB1-GPX4 binding. Data are
   expressed as mean ± SD from three or more independent biological
   replicates

The combination of DSF and TKI exerts synergistic anti-leukemic effects in
vivo

   The aforementioned results underscore the anti-leukemia efficacy of DSF
   alone or in combination with TKIs in BCR-ABL+ leukemia, indicating
   their potential to induce ferroptosis. To further evaluate their
   therapeutic potential, we investigated their effects in
   BCR-ABL+ leukemia mouse models. Firstly, K562 Xenograft models were
   established (Fig. [180]8A), revealing significantly smaller leukemia
   burden and prolonged survival in mice treated with the combination of
   DSF and TKIs compared to those treated with DSF or TKIs alone
   (Fig. [181]8B-C). Subsequently, BCR-ABL mice were utilized to assess
   the therapeutic potential of DSF and TKIs (Fig. [182]8D). Treatment
   with DSF and DSF + IM led to a notable reduction in spleen size and
   weight compared to control mice (Fig. [183]8E), with evident
   morphological improvements in BM and spleen (Fig. S8A-B). Additionally,
   reduced Gpx4 protein levels were observed in the spleens of mice
   treated with DSF plus IM compared to control and IM treatment alone
   (Fig. S8C). Notably, populations of myeloid leukemic cells in BM and
   spleen were significantly diminished in CML mice treated with DSF,
   further reduced in those treated with the combination of DSF and IM
   (Fig. [184]8F). Moreover, a lower proportion of LSK and LT-HSCs cells
   were observed upon treatment with DSF alone or in combination with IM
   (Fig. [185]8G-H). DSF treatment alone and in combination with TKIs
   significantly prolonged survival in murine models (Fig. [186]8I). In
   line with our findings, the IM and DSF combination group showed
   elevated levels of MDA, along with reduced levels of GSH and GPX4
   (Fig. [187]8J-L). To assess the impact of combination treatment on the
   LSC burden, we transplanted 1 × 10^6 BM cells from mice treated with
   either DSF alone or DSF plus IM into 2nd recipients (Fig. S8D). The
   recipients of BM from DSF plus IM treated donors exhibited
   significantly smaller leukemia burden (Fig. S8E), decreased size and
   weight of the spleens (Fig. S8F). These findings collectively
   demonstrate that the combination of DSF and TKIs markedly exerts
   cytotoxicity against BCR-ABL+ leukemia cells, including stem/progenitor
   cells, compared to individual treatments alone.

Fig. 8.

   [188]Fig. 8
   [189]Open in a new tab

   Enhanced anti-leukemia activity in BCR-ABL+ leukemia by the combination
   of DSF and TKIs in vivo. A Schematic of xenotransplantation assay with
   K562 cells and B-NDG immune-deficient recipient mice. (n = 5 mice per
   group for this experiment). B Representative in vivo bioluminescent
   images of NRGS recipient mice xenotransplanted with
   GFP+/luciferase + K562 (K562-GL) cells treated with IM and DSF, either
   alone or in combination. C Kaplan–Meier survival curve following
   treatments with vehicle, TKIs alone, DSF alone, or a combination of DSF
   and TKIs (n = 5 mice per group). D Experimental design using the
   BCR-ABL transgenic mouse model. Treatment with DSF, IM, or DSF in
   combination with IM or vehicle alone (Ctrl) was initiated after
   transplantation and continued for 4 weeks. E-L A group of mice was
   sacrificed and analyzed after Treatment, and the rest for survival
   analysis. Images and weights of spleen (E), Mac-1 + Gr1 + myeloid cells
   (F), frequency of LSKs(G), LT-HSCs/ST-HSCs cells (H), survival (I), the
   ratio of GSH and GSSG (J), MDA (K), and representative
   immunofluorescence images of GPX4 protein (L) from each group are
   shown. M A summary of regulatory mechanisms that are significantly
   affected by the interaction of DSF and TKIs is shown. Data are
   presented as mean ± SD from three or more independent biological
   replicates

Discussion

   BCR-ABL+ leukemia is characterized by the presence of small populations
   of LSCs, which pose a significant challenge to achieving long-term
   remission and cure. While BCR-ABL TKIs like IM, NL, and DAS induce
   remission by targeting CML progenitors, they fail to eliminate LSCs,
   which persist despite prolonged therapy, potentially leading to disease
   relapse. Hence, there is a pressing need for improved therapeutic
   approaches targeting CML LSCs [[190]30]. Our study offers compelling
   evidence for the therapeutic potential of DSF in BCR-ABL+ human
   leukemia. We demonstrate that DSF selectively eliminates BCR-ABL+ cell
   lines and effectively inhibits primary BCR-ABL+ human leukemia cells.
   In both in vivo and in vitro settings, DSF combined with TKIs
   eradicates malignant cells and LSCs in BCR-ABL+ leukemias. Our findings
   address the persistent challenges of drug resistance and LSC
   persistence, which hamper efforts towards achieving a cure.

   Ferroptosis, a regulated form of cell death, can be induced by factors
   such as increased reactive iron ions and disrupted iron pools. Various
   cellular pathways, including the Xc- system-GSH-GPX4 axis,
   NAD(P)H/FSP1/CoQ10 axis, and GCH1/BH4/DHFR axis, can suppress
   ferroptosis by inhibiting lipid peroxidation [[191]7]. Ferroptosis has
   been associated with cancer therapy resistance, and inducing
   ferroptosis has been shown to reverse drug resistance [[192]31]. DSF
   has been reported to exert potent anti-leukemic effects through various
   cell death mechanisms, including NETosis, pyroptosis, apoptosis,
   ferroptosis, and cuproptosis [[193]21, [194]23, [195]24, [196]26]. In
   our study, we demonstrate that DSF alone or combined with TKIs induces
   alterations in the ferroptosis pathway, suggesting that induction of
   ferroptosis contributes to its anti-leukemic effects. Indeed, the
   effects of DSF alone or combined with TKIs on BCR-ABL+ leukemia cells
   can be reversed by various ferroptosis inhibitors. Our data reveal that
   DSF induces ferroptosis by suppressing GPX4 expression, with this
   effect being most pronounced when combined with TKI treatment.
   Importantly, TKI treatment alone is insufficient to induce ferroptosis,
   likely due to only a modest reduction in GPX4 levels. In contrast, DSF
   exerts a more potent inhibitory effect on GPX4, and the combination of
   both agents leads to a synergistic depletion of GPX4, resulting in
   higher levels of ferroptosis (Fig. [197]8M). Recent studies have also
   shown that GPX4 plays a crucial role in counteracting apoptosis
   [[198]11]. It is therefore possible that the synergy between IM and DSF
   stems from IM-induced apoptosis—potentially facilitated in part by GPX4
   suppression—alongside DSF-induced ferroptosis. Furthermore, our
   preclinical studies indicate that the combination of DSF and TKI is
   more effective in suppressing BCR-ABL+ leukemia cells than either agent
   used alone.

   GPX4 serves as a central regulator of ferroptosis [[199]32]. We find
   that GPX4 is overexpressed in BCR-ABL+ leukemia compared to normal
   controls. Furthermore, DSF exerts its cytotoxic effects on
   BCR-ABL+ cells by disrupting GPX4 stability, ultimately leading to
   ferroptotic cell death. This novel finding uncovers a previously
   unknown vulnerability in BCR-ABL+ cells that can be targeted for
   therapeutic purposes. Mechanistically, our findings indicate that DSF
   primarily promotes the ubiquitin-mediated degradation of GPX4 while
   simultaneously inhibiting its interaction with HSPA8. Previous study
   has documented an interaction between these proteins, suggesting that
   DSF disrupts the HSPA8-GPX4 binding, which in turn reduces
   erastin-induced GPX4 degradation [[200]28]. This discrepancy in results
   may stem from differences in the cellular reliance on distinct
   degradation pathways; prior research has shown that erastin induces
   autophagic degradation in HT1080 cells, wherein HSC70 (HSPA8) serves as
   an adaptor protein for chaperone-mediated autophagy [[201]33, [202]34].
   In this context, the application of DSF disrupts the GPX4-HSPA8
   interaction, thus inhibiting erastin-induced GPX4 degradation. In
   contrast, our study reveals that in BCR-ABL+ leukemia, GPX4 degradation
   is predominantly mediated via the ubiquitination pathway. DSF not only
   inhibits the GPX4-HSPA8 interaction but also facilitates GPX4’s
   ubiquitin-mediated degradation. Notably, DSF treatment enhances GPX4
   ubiquitination, while overexpression of HSPA8 effectively inhibits this
   process. Importantly, a recent study supports the role of HSPA8 in
   stabilizing GPX4 expression, thereby enhancing its protein levels in
   liver cancer cells [[203]35]. Although this study did not explore the
   underlying mechanisms, we propose that GPX4 in liver cancer cells may
   similarly depend on ubiquitin-mediated degradation. Furthermore, recent
   investigations have identified STUB1 as a novel ubiquitin E3 ligase
   targeting GPX4 [[204]29]. Indeed, our results show that while DSF
   inhibits the binding of HSPA8 to GPX4, it concurrently enhances the
   interaction between GPX4 and STUB1. This shift suggests that upon DSF
   treatment, GPX4 preferentially associates with STUB1, thereby promoting
   its degradation (Fig. [205]8M). Additionally, we confirmed that STUB1
   plays a crucial role in regulating GPX4 degradation and cell death in
   BCR-ABL+ leukemia cells. Collectively, these findings illuminate the
   complex interplay between GPX4, HSPA8, and STUB1, underscoring the
   potential of targeting these interactions in therapeutic strategies for
   BCR-ABL+ leukemias.

   Inhibition of GPX4 has been shown to sensitize therapy-resistant cancer
   cells to treatment [[206]15, [207]36]. Our study provides evidence that
   GPX4 is the target of DSF in inducing ferroptosis in BCR-ABL+ leukemia
   cells. We demonstrate that GPX4 is crucial for LSC development,
   transformation, and maintenance in BCR-ABL+ leukemia. As a pivotal
   regulator of ferroptosis, GPX4 plays a crucial role in maintaining
   cellular integrity [[208]37–[209]39], as well as in hematopoietic
   system [[210]40, [211]41]. Nevertheless, dietary supplementation with
   vitamin E has been shown to protect hematopoietic stem and progenitor
   cells (HSPCs) from ferroptosis-induced damage in Gpx4-depleted mice
   [[212]42], thereby safeguarding the functional integrity of the
   hematopoietic system. Conversely, inhibition of GPX4 specifically
   impairs the development and propagation of LSCs in mice fed natural
   ingredient diets, highlighting its potential as a therapeutic target in
   BCR-ABL+ leukemia mouse models, as demonstrated in this study. DSF has
   garnered attention for its potent anti-tumor activity across various
   cancer types, coupled with low toxicity [[213]43, [214]44]. In contrast
   to normal tissues, cancer cells typically exhibit elevated levels of
   GPX4, rendering them more tolerant to lipid reactive oxygen species
   (ROS) accumulation. Hence, there arises a possibility that DSF-mediated
   inhibition of GPX4 could enhance ferroptosis specifically in leukemic
   cells while sparing normal tissues. This represents a promising avenue
   for cancer treatment. Our findings lend support to the hypothesis that
   DSF treatment alone or in combination with TKIs significantly
   diminishes the burden of LSCs, suppresses disease progression, and
   enhances survival rates in murine models, underscoring its therapeutic
   potential in BCR-ABL+ leukemia.

   In conclusion, our study presents a promising therapeutic strategy for
   targeting drug-insensitive LSCs in BCR-ABL+ leukemia. Dual inhibition
   of GPX4 and BCR-ABL with DSF and TKIs, respectively, holds potential
   for addressing minimal residual disease and improving outcomes in
   preclinical mouse models. Further clinical trials are warranted to
   validate these findings and optimize the dosing and administration of
   DSF in combination with TKIs.

Supplementary Information

   [215]Supplementary Material 1.^ (903KB, pdf)
   [216]Supplementary Material 2.^ (4.5MB, xlsx)
   [217]Supplementary Material 3.^ (2.5MB, docx)

Acknowledgements