Abstract RNA 5-methylcytosine (m^5C), a prevalent epitranscriptomic modification that critically regulates gene expression and cellular homeostasis. While its roles in solid tumors have been increasingly recognized, the functional landscape of m^5C in acute myeloid leukemia (AML) remains unexplored. Here, we identified NSUN2, the principal RNA m^5C methyltransferase, as a key regulator of AML progression. NSUN2 was aberrantly upregulated in AML patient samples and correlated with poor prognosis. Functional studies demonstrated that NSUN2 promoted leukemic cell proliferation, enhanced tumor growth in xenograft models, and conferred resistance to ferroptosis—a regulated cell death process driven by lipid peroxidation. Mechanistically, NSUN2 catalyzed m⁵C deposition on the 3’UTR of FSP1 (ferroptosis suppressor protein 1) mRNA, facilitating its recognition and stabilization by the m^5C reader protein YBX1. This NSUN2-YBX1-FSP1 axis protected AML cells from ferroptotic stress by suppressing lipid peroxidation and oxidative damage. Depletion of NSUN2 or FSP1 induced mitochondrial remodeling, which primed cells for ferroptosis. Reconstitution of wild-type NSUN2 or FSP1 rescued ferroptosis resistance, whereas catalytically inactive NSUN2 (C271A/C321A) or non-functional FSP1 mutants (G2A/E156A) failed to reverse this phenotype. Pharmacological inhibition of NSUN2 with MY-1B or targeting FSP1 with iFSP1 exhibited potent anti-leukemic effects, synergizing robustly with ferroptosis inducers, standard chemotherapy, and the BCL-2 inhibitor venetoclax. Our study unveils NSUN2 and FSP1 as prognostic biomarkers and therapeutic targets in AML. We highlight a novel epitranscriptomic mechanism linking RNA methylation to ferroptosis evasion, providing a dual-strategy approach to overcome AML treatment resistance. Supplementary Information The online version contains supplementary material available at 10.1186/s12943-025-02394-8. Keywords: Acute myeloid leukemia, m^5C modification, NOP2/Sun RNA methyltransferase 2, Ferroptosis, Ferroptosis suppressor protein 1 Introduction Acute myeloid leukemia (AML) is an aggressive hematologic malignancy driven by the clonal expansion of immature myeloid blasts, which disrupt normal hematopoiesis and lead to life-threatening cytopenias [[50]1]. Despite advancements in targeted therapies and chemotherapeutic regimens, the 5-year survival rate for AML patients remains below 30%, primarily due to relapse and therapy resistance [[51]2, [52]3]. While genomic alterations (e.g., FLT3, NPM1 mutations) and epigenetic dysregulation (e.g., DNMT3A, TET2 mutations) are well-established drivers of AML pathogenesis [[53]4, [54]5], emerging evidence underscores the critical role of post-transcriptional RNA modifications in rewriting oncogenic circuits and therapeutic responses [[55]6, [56]7]. Among these modifications, 5-methylcytosine (m^5C) has recently emerged as a key regulator of RNA stability, translation, and protein interactions, with profound implications in cancer biology [[57]8, [58]9]. NOP2/Sun RNA methyltransferase 2 (NSUN2), the principal enzyme responsible for m^5C deposition in RNA, catalyzes cytosine methylation in diverse RNA species, including mRNAs and non-coding RNAs [[59]10]. In solid tumors such as bladder and lung cancers, NSUN2 overexpression stabilizes oncogenic transcripts such as HDGF and QSOX1 via m^5C-dependent recruitment of reader proteins like YBX1, thereby promoting proliferation and metastasis [[60]11–[61]13]. However, the functional significance of NSUN2 in AML, particularly its role in modulating RNA epitranscriptomics and therapeutic resistance, remains unexplored. Addressing this gap is critical, as RNA modification-driven mechanisms may unveil novel targets to overcome the limitations of conventional AML therapies. A parallel therapeutic opportunity lies in ferroptosis, an iron-dependent form of regulated cell death driven by lipid peroxidation [[62]14–[63]16]. Ferroptosis susceptibility is governed by the balance between pro-oxidant factors (e.g., iron overload, PUFA-rich membranes) and antioxidant systems, including glutathione peroxidase 4 (GPX4) and ferroptosis suppressor protein 1 (FSP1, also known as AIFM2) [[64]17–[65]25]. FSP1, a NAD(P)H-dependent CoQ oxidoreductase, protects cells from ferroptosis by reducing coenzyme Q10 (CoQ10) to its antioxidant form—ubiquinol, independently of the GPX4 pathway [[66]19]. While FSP1 is overexpressed in solid tumors and linked to therapy resistance, its prognostic relevance and regulatory mechanisms in AML are unclear. Intriguingly, recent studies suggest that RNA modifications may intersect with ferroptosis pathways [[67]20, [68]21], yet whether m^5C methylation directly regulates FSP1 or ferroptotic vulnerability in AML remains to be elucidated. In this study, we identify NSUN2 as a critical regulator of AML progression and ferroptosis resistance through m^5C-dependent stabilization of FSP1 mRNA. We demonstrate that NSUN2 is overexpressed in AML patients and correlates with poor survival across multiple cohorts. Mechanistically, NSUN2 mediates m⁵C methylation in the 3’UTR of FSP1 mRNA, enabling YBX1 binding to stabilize the transcript and maintain FSP1 protein levels. Depletion of NSUN2 or FSP1 triggers mitochondrial priming for ferroptosis, sensitizing AML cells to ferroptosis inducers (RSL3 and erastin). We further validate the efficacy of pharmacological inhibitors targeting this axis. The NSUN2 inhibitor MY-1B and the FSP1 inhibitor iFSP1 exhibited potent anti-leukemic activity, both as monotherapies and in combination with ferroptosis inducers and standard chemotherapies (daunorubicin, cytarabine and venetoclax). Strikingly, MY-1B and iFSP1 synergized robustly with the BCL-2 inhibitor venetoclax. This dual death pathway inhibition—targeting both ferroptosis evasion and apoptotic resistance—achieved near-complete leukemia clearance. Our work establishes NSUN2 and FSP1 as prognostic biomarkers and therapeutic targets, while uncovering a novel epitranscriptomic mechanism that bridges RNA methylation to cell death regulation, offering a dual-strategy approach to improve AML treatment outcomes. Materials and methods Patient samples Patient samples were obtained from 303 individuals diagnosed with acute myeloid leukemia (AML) at the First Affiliated Hospital of Zhejiang University School of Medicine between July 2010 and April 2016. Written informed consents were secured from all participants, and the study protocol was approved by the Institutional Review Board of Zhejiang University. Bone marrow and peripheral blood samples were collected at the time of diagnosis, prior to any treatment administration. Mononuclear cells were isolated through Ficoll density gradient centrifugation and subsequently cryopreserved in liquid nitrogen for further analysis. Comprehensive clinical data, including patients age, gender, complete blood count, cytogenetic profile, and treatment response, were documented. All samples were anonymized and handled in accordance with ethical standards to maintain patient confidentiality. CD34⁺ hematopoietic progenitor cell isolation and culture Peripheral blood mononuclear cells from healthy donors were isolated by Ficoll gradient centrifugation. CD34⁺ cells were purified using CD34 MicroBeads (Miltenyi Biotec, Cat# 130-046-702) and MS columns according to the manufacturer’s protocol. Purity (> 90%) was confirmed by flow cytometry using anti-human CD34 antibody (BioLegend, Cat# 378604). Cells were cultured in StemSpan™ SFEM II medium (Stemcell Technologies) supplemented with SCF (100 ng/mL), Flt3-L (100 ng/mL), and TPO (50 ng/mL) at 37 °C with 5% CO₂. Cell lines and culture The human AML cell lines MOLM-13, THP-1, MV4-11, HL-60, OCI-AML2, OCI-AML3, Kasumi-1, U937, MONO-MAC-6 and KG-1 were used for detailed functional studies. All cell lines were authenticated by short tandem repeat profiling (Biowing Applied Biotechnology, Shanghai). Cells were cultured in either Iscove’s Modified Dulbecco’s Medium (IMDM) or RPMI-1640 medium (Gibco) respectively, supplemented with 10% fetal bovine serum (Gibco), 100 U/mL penicillin, and 100 µg/mL streptomycin (Gibco) at 37 °C in a humidified atmosphere containing 5% CO2. Kasumi-1 cells were cultured in RPMI-1640 supplemented with 20% FBS to maintain optimal growth. ShRNA knockdown and lentivirus production Lentiviral vectors targeting NSUN2, YBX1, and FSP1 were constructed using the pLKO.1 vector, with shRNA sequences detailed in Supplementary Table [69]1. Lentiviruses were produced by co-transfecting HEK293T cells with the shRNA plasmid, psPAX2, and pMD2.G using Lipofectamine 3000 (Invitrogen). Viral supernatants were collected 48 h post-transfection, filtered, and concentrated by ultracentrifugation. MOLM-13 and THP-1 cells were transduced with lentivirus in the presence of 6 µg/mL polybrene (Sigma-Aldrich). Stable knockdown lines were selected with 2 µg/mL puromycin (Thermo Fisher) for 48 h. Plasmid construction for NSUN2 and FSP1 wild-type and mutants The full-length coding sequences of human NSUN2 (NCBI RefSeq: [70]NM_017755.6) and FSP1 (NCBI RefSeq: [71]NM_032797.6) were cloned into the pLVX-EF1α -IRES-Neo lentiviral vector (Takara Bio). Site-directed mutagenesis introduced cysteine-to-alanine substitutions at residues 271 (C271A) and 321 (C321A) in NSUN2, targeting conserved catalytic residues critical for m^5C methyltransferase activity. For FSP1, mutations were introduced to disrupt the N-terminal myristoylation site (G2A) and the enzymatic active site (E156A), abolishing membrane localization and CoQ10 reductase activity. All constructs were validated by Sanger sequencing. Lentiviral particles were generated as above, and stable AML cell lines (MOLM-13, THP-1) were selected with 1 mg/mL G418 (Thermo Fisher) for 7 days. RNA extraction and RT-qPCR Total RNA was extracted from cultured cells using the TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. The quantity and quality of the extracted RNA were examined by NanoDrop-3000 spectrophotometer (Thermo Fisher Scientific). Reverse transcription was performed with 1 µg of total RNA using PrimeScript RT Reagent Kit (Takara). Then qRT-PCR procedure was conducted with SYBR Green PCR Master Mix (Takara, Japan) with GAPDH as internal reference. The relative mRNA expression levels were quantified as fold changes using the comparative Ct (ΔΔCt) method. The detailed primer sequences were shown in Supplementary Table [72]1. Western blotting Whole-cell lysates were prepared in RIPA lysis buffer with protease inhibitor cocktail (Cat #78442, Thermo Fisher Scientific). After centrifugation at 4 °C, the supernatants were quantified by bicinchoninic acid assays (Pierce™ BCA Protein Assays Kit, Thermo Fisher Scientific). Proteins were separated using SDS-PAGE and transferred onto PVDF membranes (Millipore). Membranes were incubated with primary antibodies at 4 °C overnight after blocking with milk. Membranes were incubated with chemiluminescent signal was detected using a chemiluminescent Image System (Bio-rad). Primary antibodies used are as follows: anti-NSUN2 (Cat #20854-1-AP, Proteintech), anti-FSP1 (Cat #20886-1-AP, Proteintech), anti-YBX1 (Cat #ab76149, Abcam), anti-α-tubulin (Cat #2128, CST). Cell viability assay Cell viability was assessed using the CellTiter-Glo (Promega, CTG assay). Cells were seeded into 96-well plates at a density of 1,000 cells per well. At indicated time points, an equal volume (100 µl) of CellTiter-Glo reagent was added into each well and measurements were done using Varioskan Flash multimode reader (Thermo Fisher Scientific). DAPI staining, edu incorporation assays and colony formation assay For DAPI nucleus stain, cells were incubated with DAPI diluted 1:5000 in PBS for 15 min, followed by washing with PBS. Results were acquired by flow cytometry (CytoFlex, Beckman Coulter) and processed using FlowJo software. For EdU incorporation assays, cells were seeded into 6-well plates and treated with 10 µM EdU for 2 h. After, incubation, cells were fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton X-100, and stained using Click-iT™ EdU Imaging Kit (Invitrogen) according to the manufacturer’s instructions. The proportion of EdU-positive cells was determined by flow cytometry. Colony formation assays were performed to evaluate clonogenic potential. Approximately 500 cells were plated in growth medium containing 1.2% methylcellulose in a single well of a 35 mm dish. After two weeks, colonies were stained with 5 mg/ml MTT solution to assess viability. Cell cycle analysis The cell cycle distribution was analyzed by flow cytometry. Cells were harvested, washed with PBS, and fixed in 70% ethanol at 4 °C overnight. After fixation, cells were stained with propidium iodide (PI) solution containing RNase A (Sigma-Aldrich) and incubated at 37 °C for 30 min. The percentages of cells in G0/G1, S, and G2/M phases were determined by flow cytometry. In vivo xenograft models NOD.CB17-Prkdc^scid Il2rg^tm1/Bcgen (B-NDG) mice (6 weeks old) were purchased from Jiangsu Biocytogen Co., Ltd. Mice were maintained under specific pathogen-free conditions. All animal procedures were approved by the Ethics Committee for Zhejiang Centre of Laboratory Animals and were carried out in accordance with established guidelines for animal welfare. MOLM-13 luciferase cells were transduced with either control lentivirus or lentivirus carrying NSUN2-targeting or FSP1-targeting shRNA, as described above. One million NSUN2 knockdown, FSP1 knockdown or control MOLM-13 luciferase cells were resuspended in PBS and injected intravenously into the tail vein of each mouse. Tumor growth was monitored using bioluminescence imaging (IVIS Spectrum, PerkinElmer) after injecting D-luciferin (150 mg/kg) (VivoGlo™ Luciferin, Promega). Mice were sacrificed at the study endpoint or when they displayed signs of total hind limb paralysis. The liver and spleen were harvested and weighed. The femurs were collected for HE staining. The human CD45 positive (hCD45^+) cells were measured and analyzed from bone marrow and spleen by flow cytometry. Overall survival was recorded and analyzed using Kaplan-Meier survival analysis. BODIPY 581/591 C11 analysis and detection of ferrous ions (Fe^2+) Cells were seeded into 6-well plates at a density of 2 × 10^5 per well, and treated with DMSO (vehicle) or 250 nM RSL3 (MCE, HY-100218 A) for 75 min. After treatment, cells were washed with HBSS and incubated with 5 µM BODIPY 581/591 C11 (Thermo Fisher Scientific, #D3861) for 30 min in a 37 °C incubator equilibrated with 5% CO[2]. Cells were then washed and analyzed by flow cytometry. The fluorescence intensity ratio of oxidized (green) to reduced (red) C11-BODIPY was used as an indicator of lipid peroxidation. To detect cellular ferrous ions (Fe^2+), cells were harvested after treatment with vehicle or RSL3 and washed with HBSS three times. The cells were then labeled with 1 µmol/L Fe^2+ indicator working solution (Ferrorange, Donjindo) and incubated for 30 min in a 37 °C incubator, followed by analysis using flow cytometry. Transmission electron microscopy (TEM) Transmission electron microscopy (TEM) was utilized to examine the ultrastructural characteristics of AML cells. Cells were first fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.4) at 4 °C overnight. Following primary fixation, cells were washed, post-fixed in 1% osmium tetroxide for 1 h at room temperature, dehydrated through a graded ethanol series (30%, 50%, 70%, 90%, and 100%) and then infiltrated with a mixture of ethanol and epoxy resin (Epon 812). Ultrathin sections were prepared and stained before observation using an HT7800 TEM (PHILIPS TECNAI 10). Images were collected for analysis. RNA sequencing The RNA sequencing (RNA-seq) was performed by Oebiotech (Shanghai, China). Briefly, 1 µg of total RNA was used to generate cDNA libraries. The libraries were prepared using the TruSeq Stranded mRNA Library Prep Kit (Illumina), following the manufacturer’s instructions. This process involves mRNA enrichment, fragmentation, reverse transcription, and ligation of sequencing adapters. The prepared cDNA libraries were sequenced on an Illumina NovaSeq 6000 platform to produce paired-end reads. The quality of the raw sequencing data was assessed using FastQC. Differentially expressed genes between control and NSUN2- or FSP1-knockdown cells were identified using DESeq2, a statistical package for analyzing count data from RNA-Seq experiments. Gene set enrichment analysis (GSEA) was performed using the Broad Institute’s GSEA software to identify significantly enriched pathways and biological processes. The results were visualized and interpreted to understand the impact of NSUN2 knockdown on gene expression. RNA-bisulfite sequencing The RNA-bisulfite sequencing (Bis-seq) experiment was performed by E-GENE (Shenzhen, China). Total RNA from control and NSUN2-knockdown cells was extracted using the TRIzol reagent (Invitrogen), following the manufacturer’s protocol. The quality and quantity of the RNA were assessed using a NanoDrop spectrophotometer and agarose gel electrophoresis. To identify m^5C modification sites, 1 µg of mRNA purified from total RNA, was bisulfite-converted using the EZ RNA Methylation Kit (Zymo Research, Irvine, CA, USA). The converted RNA was then purified according to the kit instructions. The purified, bisulfite-converted RNA was sequenced on an Illumina NovaSeq 6000 platform (Illumina, San Diego, CA, USA) to generate high-throughput sequencing data. Raw sequencing data were processed using Bismark, a bioinformatics tool specifically designed for bisulfite sequencing data. The processed data were further analyzed for m^5C sites using meRanTK software, which provides a comprehensive analysis of RNA methylation. The site-specific methylation level was normalized to local transcript coverage and calculated as the proportion of unconverted cytosine reads over the total reads covering that site. Specifically, m⁵C level = (Number of unconverted C reads) / (Total reads covering the site). Luciferase reporter construction and luciferase activity assays The luciferase reporter plasmids were constructed by inserting annealing oligonucleotides including target sequences into the pmirGlo vector (Promega). Constructs were verified by sequencing. For luciferase activity assays, HEK293T were co-transfected with the luciferase reporter plasmids using Lipofectamine 3000 (Invitrogen). After 24 h, cells were lysed, and luciferase activity was measured using the Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer’s instructions. Firefly luciferase activity was normalized to Renilla luciferase activity (included in the pmirGlo vector as an internal control) to control for transfection efficiency. RNA stability assays MOLM-13 cells with or without NSUN2 knockdown were treated with actinomycin D at a final concentration of 5 µg/mL and collected at the indicated time points. The mRNAs were analyzed via RT-qPCR. RNA immunoprecipitation (RIP) The RIP experiment was performed with the Magna RIP Kit (Millipore) according to the manufacturer’s instructions. Cell lysates were incubated with magnetic beads conjugated to antibodies against NSUN2 or YBX1. The rabbit IgG was used as a control. After overnight incubation at 4 °C, the beads were washed extensively with RIP wash buffer. Co-precipitated RNAs were then eluted and purified using the RNeasy Mini Kit (Qiagen). Purified RNAs were reverse-transcribed, and the resulting cDNA was analyzed via qPCR. m^5C-RNA immunoprecipitation m^5C RNA immunoprecipitation was performed as previously described [[73]20]. Briefly, 100 µg total cellular RNA was extracted using TRIzol and fragmented to an average size of 200 nucleotides using RNA fragmentation reagents (AM8740, Ambion). The fragmented RNA was then incubated with 5 mg an anti-m^5C antibody (ab10805, Abcam) in IP buffer (150 mM NaCl, 0.1% NP-40 and 10 mM Tris-HCl, pH 7.4) at 4 °C overnight with gentle rotation. The antibody-RNA complex was isolated by incubation with Protein A/G magnetic beads (Thermo Fisher Scientific) for 2 h at 4 °C. The beads were washed three times and eluted competitively with an m^5C monophosphate solution. RNA in the eluate was isolated using RNeasy Mini Kit (Qiagen) and used for downstream qPCR analysis. Statistical analysis All data were presented as mean ± SD. Statistical analysis was performed using GraphPad Prism 8.0 or R software. Survival analysis was performed using Kaplan-Meier method, and the log-rank test was used to evaluate the differences. The Student’s t-test was employed to assess the statistical significance between two groups. One-way ANOVA was used for multiple group comparisons. Drug synergy was analyzed using the SynergyFinder package (Bliss independence model), with synergy scores > 10 indicating significant synergy. P < 0.05 was considered statistically significant. Asterisks indicate significant differences (^*P < 0.05; ^**P < 0.01; ^***P < 0.001, ^****P < 0.0001). Result NSUN2 drives acute myeloid leukemia progression and predicts poor prognosis To define the clinical significance of NSUN2 in AML, we first analyzed CD34^+ bone marrow (BM) cells from the [74]GSE30029 dataset, which revealed significantly elevated NSUN2 mRNA levels in AML patients compared to healthy donors (n = 46 vs. 31; P < 0.0001, Fig. [75]1a). This differential expression was validated in whole bone marrow specimens from our cohort, where newly diagnosed AML patients (n = 303) exhibited markedly higher NSUN2 levels than healthy controls (n = 24; P < 0.01, Fig. [76]1b). Western blotting of primary BM samples confirmed NSUN2 protein overexpression in AML patients compared to healthy counterparts (Fig. [77]1e). TNMplot meta-analysis (407 normal vs. 151 AML) and BloodSpot cytogenetic stratification universally validated this dysregulation (Supplementary Fig. [78]1a-b). Critically, survival analysis across three independent cohorts established NSUN2 as a potent prognostic determinant. High NSUN2 expression predicted significantly shorter overall survival in the TCGA-AML cohort (P = 0.047, Supplementary Fig. [79]1c), [80]GSE12417 dataset (P = 0.0325, Fig. [81]1c), and our institutional cohort (P < 0.0001, Fig. [82]1d). Baseline characteristics of our cohort are provided in Supplementary Table [83]2. Fig. 1. [84]Fig. 1 [85]Open in a new tab NSUN2 drives acute myeloid leukemia progression and predicts poor prognosis. (a) NSUN2 mRNA levels in CD34^+ bone marrow (BM) cells from AML patients (n = 46) versus healthy donors (n = 31) from [86]GSE30029 dataset. ****P < 0.0001. (b) NSUN2 mRNA expression in BM cells from newly diagnosed AML patients (n = 303) compared to healthy controls (n = 24) from our cohort. **P < 0.01. (c) Kaplan-Meier survival analysis of AML patients in [87]GSE12417 dataset, stratified by NSUN2 median expression. P = 0.0325, log-rank test. (d) Kaplan-Meier plot for overall survival of AML patients from our center, stratified by NSUN2 expression (median cutoff: NSUN2^low, n = 151, NSUN2^high, n = 152; P < 0.0001, log-rank test). (e) Western blot analysis of NSUN2 protein levels in primary BM samples from healthy donors (N1-N3) and AML patients (P1-P6). α-Tubulin served as a loading control. (f) RT-qPCR validation of NSUN2 knockdown efficiency in MOLM-13 and THP-1 cells transduced with control shRNA (shCtrl) or two independent NSUN2-targeting shRNAs (shNSUN2#1 / #2). n = 3, ****P < 0.0001. (g) Western blot confirmation of NSUN2 knockdown in MOLM-13 and THP-1 cells. (h-i) Cell viability of NSUN2-knockdown MOLM-13 (h) and THP-1 cells (i) was measured by CellTiter-Glo assay over 5 days (n = 3, ***P < 0.001, ****P < 0.0001). (j-k) EdU incorporation assay quantifying proliferative cells in NSUN2-knockdown MOLM-13 (j) and THP-1 (k), analyzed by flow cytometry. (l) In vivo leukemic burden monitored via bioluminescence imaging in B-NDG mice transplanted with shCtrl- or shNSUN2-transduced MOLM-13-luc cells. Bioluminescence imaging was performed weekly from Day 7 to 42. (m) Quantification of bioluminescent signal intensities (photons/s) in mice from (l). (n) Kaplan-Meier survival curves comparing mice transplanted with shCtrl- vs. shNSUN2-transduced MOLM-13 cells. n = 6, P = 0.001, log-rank test. (o) Representative image of HE staining of femur from shCtrl or shNSUN2 mice group. Scale bar represents 100 μm, n = 3. (p) Flow cytometry analysis of hCD45^+ leukemic blasts in bone marrow (BM) and spleen (n = 3, **P < 0.01). (q) The weight of spleen from mice sacrificed on Day 21 (n = 3, *P < 0.05) To further delineate NSUN2’s biological impact, we engineered stable NSUN2-knockdown MOLM-13 and THP-1 cells through lentiviral delivery. The knockdown efficiency validated at both transcriptional (Fig. [88]1f) and protein levels (Fig. [89]1g). NSUN2-depleted cells exhibited significant viability loss over time (Fig. [90]1h-i). This was accompanied by decreases in proportion of proliferating cells (Fig. [91]1j-k, Supplementary Fig. [92]1d-e) and near-complete abrogation of clonogenic capacity (Supplementary Fig. [93]1f), indicating critical roles in sustaining AML cell proliferation. To determine whether NSUN2’s methyltransferase activity is essential for AML cell proliferation, we performed rescue experiments in NSUN2-knockdown MOLM-13 and THP-1 cells by reintroducing either wild-type NSUN2 (NSUN2-WT), a catalytically inactive double mutant (NSUN2-DM, C271A/C321A), or an empty vector control. Western blot analysis confirmed comparable or even higher expression levels of NSUN2-DM compared to NSUN2-WT (Supplementary Fig. 7a). However, only NSUN2-WT rescued proliferation capacity (Supplementary Fig. [94]7b-e) and recovered global m⁵C methylation (Supplementary Fig. [95]7f), whereas NSUN2-DM did not exhibit such effects. We next generated xenograft models using MOLM-13-Luciferase cells either with control or with NSUN2 knockdown. Bioluminescence imaging demonstrated a markedly reduced leukemic burden in the NSUN2-deficient group compared to controls (Fig. [96]1l-m). Notably, mice transplanted with NSUN2-knockdown cells exhibited extended median survival (43 vs. 27 days) and overall survival (48 vs. 28 days) compared to controls (P = 0.0001, Fig. [97]1n). Additionally, three more mice from each group were sacrificed on day 21. Terminal analysis demonstrated attenuated leukemic infiltration in NSUN2-deficient group, as evidenced by decreased hCD45 positive blasts in bone marrow and spleen (Fig. [98]1p), along with normalized spleen weights (Fig. [99]1q). Histopathological assessment confirmed diminished leukemic involvement (Fig. [100]1o). NSUN2 deficiency sensitizes AML cells to ferroptosis To delineate the mechanistic basis of NSUN2 in AML progression, RNA-seq profiling of NSUN2-knockdown MOLM-13 cells (4 days post-transduction) revealed profound transcriptomic changes, with 1,769 genes downregulated and 400 upregulated (Supplementary Table [101]3). Gene Set Enrichment Analysis (GSEA) uncovered coordinated suppression of cell cycle progression pathways (G0/G1 checkpoint: NES = − 1.97, FDR = 0.011; Fig. [102]2a), consistent with flow cytometry showing G0/G1 phase arrest in NSUN2-knockdown cells (Fig. [103]2b, Supplementary Fig. [104]2a). Western blot confirmed the downregulation of CDK2/CDK6 level alongside p21/p27 upregulation (Fig. [105]2c). Fig. 2. [106]Fig. 2 [107]Open in a new tab NSUN2 deficiency sensitizes AML cells to ferroptosis. (a) GSEA analysis showed cell cycle pathway was upregulated in NSUN2-knockdown MOLM-13 cells. (b) Quantification of cell cycle distribution in MOLM-13 and THP-1 cells, analyzed by Flow cytometry. (c) Western blot analysis of cell cycle-related proteins in MOLM-13 and THP-1 following NSUN2 knockdown. (d-e) Ferroptosis pathway was significantly enriched in NSUN2 deficient cells, as shown in gene enrichment plot (d) and heatmap (e). (f) Transmission electron microscopy (TEM) of MOLM-13 cells (left panel: ×2500, right panel: ×8,300). Yellow arrows in control cells indicate normal mitochondria with typical morphology, while yellow arrows in NSUN2-knockdown cells highlight characteristic ferroptotic features including mitochondrial shrinkage and cristae disassembly. (g–j) MOLM-13 (g, i) and THP-1 (h, j) cells with control or NSUN2 knockdown were treated with RSL3 or Erastin at gradient concentrations for 48 h. Cell viability was measured using CTG assay. Data shown represent mean ± SD from three independent experiments. (k-l) Control and NSUN2 knockdown cells treated with 250 nM RSL3 for 75 min, then incubated with Fe^2+ probe and assessed by flow cytometry. (m-n) Control and NSUN2 knockdown cells were treated with 250 nM RSL3 for 75 min. Then lipid peroxidation level was assessed using BODIPY 581/591 C11 staining by flow cytometry (m). Quantitative result of lipid peroxidation levels was shown (n). Ox.: oxidized. NS2-WT: wildtype-NSUN2; NS2-DM: double-mutant NSUN2 (C271A and C321A) Intriguingly, parallel transcriptomic analysis exposed a significant activation of ferroptosis pathways (NES = 1.90, FDR = 0.023; Fig. [108]2d-e) in NSUN2-knockdown cells, suggesting a potential link between NSUN2 loss and ferroptosis regulation. However, under basal conditions, lipid peroxidation levels remained unchanged in NSUN2-knockdown cells (Supplementary Fig. [109]2b-e), indicating that transcriptional activation of ferroptosis pathways alone is insufficient to trigger spontaneous ferroptosis. Notably, ultrastructural analysis uncovered mitochondrial hallmarks of ferroptotic priming, including marked shrinkage, condensed membrane density and partial dissolvement of cristae (Fig. [110]2f). This morphological alteration implied that NSUN2 loss establishes a metabolically primed state, sensitizing cells to exogenous ferroptotic triggers. Functional validation demonstrated that NSUN2 deficiency sensitized AML cells to ferroptosis inducers. In MOLM-13 cells, NSUN2 knockdown reduced RSL3 IC[50] by 1.98-fold (shNSUN2#1: 190.64 nM vs. shCtrl: 378.39 nM, Fig. [111]2g) and Erastin IC[50] by 1.27-fold (shNSUN2#1: 239.92 nM vs. shCtrl: 304.37 nM, Fig. [112]2i). Similar sensitization was observed in THP-1 cells (Fig. [113]2h and j). DAPI staining of cells treated with RSL3 revealed significantly higher cell death in NSUN2- deficient cells (Supplementary Fig. [114]2f-g). Mechanistically, NSUN2 loss (shNSUN2#1) amplified RSL3-induced iron overload (6.2-fold Fe^2+ increase in MOLM-13; 6.3-fold in THP-1 (Fig. [115]2k-l) and potentiated lipid peroxidation, with oxidized BODIPY C11 signals surging from 7.1 to 42.6% in MOLM-13 and 5.8–32.7% in THP-1 (Fig. [116]2m-n). Re-expression of wild-type NSUN2 (NSUN2-WT), but not catalytic mutant (NSUN2-DM, C271A/C321A), restored lipid ROS homeostasis, confirming methyltransferase activity dependence (Fig. [117]2m-n). FSP1 expression depends on NSUN2-mediated RNA m^5C modification As an RNA m⁵C methyltransferase, NSUN2 depletion in MOLM-13 and THP-1 leukemia cells markedly reduced global RNA m⁵C levels (Fig. [118]3a, Supplementary Fig. [119]3a-b). To map site-specific methylation dynamics, we performed bisulfite sequencing (Bis-seq) in control and NSUN2-deficient MOLM-13 cells. While the overall transcriptomic distribution of m⁵C sites (5’UTR/CDS/3’UTR) remained largely unchanged (Fig. [120]3b), we identified 5,954 hypomethylated sites (Δm⁵C < − 0.1, FDR < 0.05) mapping to 948 genes, far exceeding hypermethylation events (1,897 sites, 203 genes) (Fig. [121]3c, Supplementary table [122]4). Chromosomal profiling revealed broad m⁵C reduction across multiple genomic regions (Fig. [123]3d, Supplementary Fig. [124]3c-d), suggesting systemic rather than locus-restricted effects. Fig. 3. [125]Fig. 3 [126]Open in a new tab FSP1 expression depends on NSUN2-mediated RNA m^5C modification. (a) Dot blot analysis of total RNA m^5C levels in MOLM-13 and THP-1 cells transfected with shCtrl, shNSUN2#1, or shNSUN2#2. Methylene blue (MB) staining served as a loading control. (b) Density distribution of m⁵C sites across transcriptomic regions (5’UTR, CDS, 3’UTR) in shNSUN2 versus shCtrl MOLM-13 cells. (c) Volcano plot of differentially methylated sites identified by Bis-seq in NSUN2-knockdown MOLM-13 cells. Thresholds:|Δm⁵C|>0.1 and FDR < 0.05. Hypomethylated (blue) and hypermethylated (red) sites are highlighted (n = 3 biological replicates). (d) Average m^5C level across chromosomes in control and NSUN2 knockdown MOLM-13 cells. (e) Venn diagram intersecting genes with hypomethylation (Δm⁵C < − 0.1 and FDR < 0.05, Bis-seq) and downregulated expression (|log[2]foldchange|>1 and FDR < 0.05, RNA-seq). FSP1 was prioritized as a candidate gene. (f) RNA-seq result of FSP1 downregulation upon NSUN2 knockdown (n = 3; ***P < 0.001, Student’s t-test). (g) Site-specific m⁵C levels at eight cytosine residues (C1236–C1300) on FSP1 mRNA in NSUN2-knockdown MOLM-13 cells, quantified by Bis-seq. Data shown as mean ± SD (n = 3; *P < 0.05, **P < 0.01, Student’s t-test). (h-i) NSUN2 knockdown reduced FSP1 mRNA (h, qPCR) and protein (i, Western blot) levels in MOLM-13 and THP-1 cells. (j) RIP-qPCR using anti-NSUN2 antibody demonstrated significant enrichment of FSP1 mRNA compared to IgG control (n = 3; **P < 0.01). (k) m⁵C-RIP-qPCR demonstrated diminished FSP1 enrichment using m⁵C antibody in NSUN2-knockdown cells (n = 3, *P < 0.05, **P < 0.01). (l) RIP-qPCR detecting FSP1 enriched by anti-YBX1 antibody in MOLM-13 cells (n = 3, *P < 0.05). (m) YBX1 knockdown impaired FSP1 protein expression in MOLM-13 and THP-1 cells. (n) FSP1 mRNA stability assay following transcriptional arrest with 5 µg/ml actinomycin D. NSUN2 knockdown reduced FSP1 half-life in MOLM-13 cells (shCtrl: 19.75 h, shNSUN2: 5.47 h, shYBX1 4.27 h, *P < 0.05). (o) Schematic of FSP1-3’UTR mutagenesis to disrupt m⁵C sites. 293T cells were co-transfected with shNSUN2/shYBX1 and reporter plasmids (FSP1-3’UTR-WT or FSP1-3’UTR-Mut). (p) Luciferase assay showed NSUN2/YBX1 knockdown suppressed wild-type (WT) FSP1-3’UTR activity (n = 3, **P < 0.001, ***P < 0.001 vs. shCtrl, Student’s t-test) but not mutant constructs Integrating methylation and transcriptomic profiles, we identified 30 genes exhibiting concurrent m⁵C loss and transcriptional downregulation (Fig. [127]3e). Among these candidates, FSP1 emerged as a priority target both mechanistically and clinically, due to its dual association with ferroptosis pathway and the highest prognostic hazard ratio in TCGA-AML cohort (HR = 2.4, P = 0.0029, Supplementary Table [128]5). RNA-seq showed NSUN2 knockdown-dependent FSP1 mRNA suppression (log[2]FC = − 1.53, P < 0.0001; Fig. [129]3f), while Bis-seq pinpointed specific m⁵C hypomethylation at eight cytosine residues (C1236–C1300) clustered within the 3’UTR of FSP1 transcripts (Fig. [130]3g, Supplementary Fig. [131]3e). Consistent with these findings, NSUN2 depletion reduced FSP1 mRNA and protein levels in both MOLM-13 and THP-1 cells (Fig. [132]3h-i), establishing FSP1 as a key downstream target of NSUN2-mediated methylation. Supporting this conclusion, rescue experiments showed that re-expression of wild-type NSUN2, but not the catalytically inactive mutant, restored FSP1 expression to baseline levels, although NSUN2-DM was more abundantly expressed (Supplementary Fig. [133]7a). Mechanistic dissection revealed direct NSUN2-FSP1 mRNA interaction (Fig. [134]3j) and m⁵C-dependent recognition, as NSUN2 depletion abolished m⁵C-RIP signals (Fig. [135]3k). We further identified YBX1 as a co-factor binding FSP1 mRNA (P < 0.05; Fig. [136]3l), whose knockdown mirrored NSUN2 deficiency in reducing FSP1 level (Fig. [137]3m). Functional validation demonstrated that both NSUN2 and YBX1 stabilize FSP1 transcripts, with mRNA half-life decreasing from 19.75 h (control) to 5.47 h (NSUN2-KD) and 4.27 h (YBX1-KD) (Fig. [138]3n). Crucially, luciferase reporters harboring wild-type FSP1-3’UTR, but not mutants disrupting m⁵C sites, lost responsiveness to NSUN2/YBX1 knockdown (Fig. [139]3o-p), pinpointing these residues as essential regulatory elements. To further validate the methylation-dependent interaction between FSP1 mRNA and YBX1, we conducted RNA immunoprecipitation (RIP) assays in HEK293T cells transfected with reporter constructs containing either wild-type or m⁵C-deficient mutant FSP1 3’UTRs. Quantitative analysis showed significantly reduced enrichment of the mutant FSP1 3'UTR reporter RNA in the anti-YBX1 immunoprecipitate, confirming that m⁵C modification within the 3’UTR is critical for YBX1 binding (Supplementary Fig. [140]9a-c). FSP1 deficiency significantly represses AML progression in vitro and in vivo Comparative qPCR profiling of bone marrow specimens revealed pronounced FSP1 overexpression in AML patients relative to healthy donors (P = 0.0005, Fig. [141]4a). Analysis of AML patient cohorts demonstrated that high FSP1 expression associated with significantly worse overall survival in TCGA-AML cohort (P = 0.0029, Fig. [142]4b), [143]GSE6891 (P = 0.016, Fig. [144]4c) and our institutional cohort (P = 0.0059, Fig. [145]4d). To delineate the biological role of FSP1 in AML, we initiated our investigation by knocking down FSP1 expression in MOLM-13 and THP-1 cells (Fig. [146]4e-f). FSP1 silencing markedly impaired leukemia cell viability over days (Fig. [147]4g-h), while EdU incorporation assays revealed a marked reduction in proliferating cells (Fig. [148]4i-j). Fig. 4. [149]Fig. 4 [150]Open in a new tab FSP1 deficiency significantly represses AML progression in vitro and in vivo. (a) qPCR analysis of FSP1 expression in bone marrow cells from healthy donors (n = 24) and AML patients (n = 303). P = 0.0005, unpaired t-test. (b) Kaplan-Meier survival analysis of AML patients from the TCGA-AML cohort stratified by FSP1 median expression levels. P = 0.0029, log-rank test. (c) Kaplan-Meier curve of AML patients from the [151]GSE6891 cohort stratified by FSP1 expression levels (FSP1^low, n = 160; FSP1^high, n = 160; P = 0.016). (d) Kaplan-Meier plot of AML patients from our institutional cohort stratified by FSP1 expression levels (FSP1^low, n = 151; FSP1^high, n = 152; P = 0.0059). (e-f) The knockdown efficiency was validated by qPCR (e) and Western blotting (f) in MOLM-13 and THP-1 cells transduced with shRNA targeting FSP1 (shFSP1#1 or shFSP1#2) or a scrambled control shRNA (shCtrl). (g-h) CTG assays was performed to observe the cell growth ability of MOLM-13 (g) and THP-1(h) over time following FSP1 knockdown. (i-j) EdU assays were carried out to detect the proliferation ability of MOLM-13 (i) and THP-1 (j) cells with FSP1 knockdown. (k) B-NDG mice was injected via tail vein with MOLM-13-luc cell transduced with shRNA targeting FSP1 (shFSP1) or scramble control (shCtrl). Tumor burden was monitored by fluorescence imaging on Day 7, 14, 21, 28, 35. (l) Quantification of average fluorescence intensity of mice on Day 7, 14, 21. Data were presented as mean ± SD. (m) Kaplan-Meier survival curves for B-NDG mice injected intravenously with MOLM-13-luc cells (control or FSP1-knockdown). Statistical significance was determined by log-rank test. n = 6 mice per group, ***P = 0.0001. (n) Flow cytometry showing the proportion of hCD45 positive blasts in the bone marrow from femur. (o) HE staining of femur from two groups. Scale bar: 100 μm To assess whether the enzymatic function of FSP1 is essential for supporting AML proliferation, we conducted rescue assays in FSP1-knockdown MOLM-13 cells by reintroducing either wild-type FSP1 (FSP1-WT) or an enzymatically inactive double mutant (FSP1-DM, G2A/E156A). The G2A mutation disrupts N-terminal myristoylation required for membrane localization, and E156A impairs catalytic activity. Western blot analysis confirmed that both constructs were comparably expressed (Supplementary Fig. [152]8a). However, FSP1-WT, but not FSP1-DM, was able to restore the proliferation capacity of AML cells (Supplementary Fig. [153]8b–e). We next validated these findings in vivo by transplanting B-NDG mice with luciferase-expressing MOLM-13 cells transduced with shFSP1 or control constructs. Tumor burden was monitored through bioluminescence imaging at multiple time points, and the results demonstrated that FSP1 deficiency significantly suppressed tumor growth in the mice, as evidenced by lower photon intensities in the FSP1 knockdown group compared to controls (Fig. [154]4k-l). This reduction in tumor burden translated into a significant survival advantage for the FSP1-deficient mice, as shown by Kaplan-Meier survival analysis (P = 0.0001) (Fig. [155]4m). Furthermore, flow cytometry analysis of the bone marrow from these mice revealed a substantial decrease in the proportion of human CD45-positive blasts in the FSP1-knockdown group, indicating reduced leukemic infiltration (Fig. [156]4n). Histological examination confirmed these findings, with less leukemic cell infiltration observed in the bone marrow of FSP1-deficient mice (Fig. [157]4o). These comprehensive in vitro and in vivo studies highlight the pivotal role of FSP1 in AML progression. FSP1 is necessary for NSUN2-mediated ferroptosis in AML cells To investigate whether FSP1 is involved in NSUN2-mediated ferroptosis in AML cells, RNA-seq combined with GSEA analysis was performed. Results indicated activation of ferroptosis-related signaling pathways in FSP1-knockdown MOLM-13 cells (Fig. [158]5b). Subsequently, cell viability assays demonstrated that knockdown of FSP1 markedly decreased the survival of MOLM-13 and THP-1 cells treated with increasing concentrations of ferroptosis inducers RSL3 or Erastin for 48 h (Fig. [159]5a). TEM imaging revealed characteristic ferroptotic mitochondrial morphology, including reduced mitochondrial size and increased membrane density, in MOLM-13 cells with FSP1 knockdown (Fig. [160]5c). Flow cytometric analysis using BODIPY 581/591 C11 indicated significantly elevated lipid peroxidation levels in MOLM-13 and THP-1 cells following NSUN2 or FSP1 knockdown under RSL3 treatment (Fig. [161]5d-e). Restoring wild-type FSP1 into either NSUN2 and FSP1-knockdown cells effectively reversed their heightened sensitivity to ferroptosis inducers, leading to a reduction in lipid peroxidation. In contrast, the reintroduction of FSP1-DM, failed to replicate this protective effect (Fig. [162]5d-e). In vivo, TEM images of femoral bone marrow cells from mice infused with MOLM-13-luc cells displayed mitochondria exhibiting ferroptotic morphological changes following FSP1 or NSUN2 knockdown compared to control cells (Fig. [163]5f). Furthermore, flow cytometry analysis showed significantly increased lipid peroxidation levels in the bone marrow cells of mice receiving cells with FSP1 or NSUN2 knockdown compared with controls (Fig. [164]5g). Fig. 5. [165]Fig. 5 [166]Open in a new tab FSP1 is necessary for NSUN2-mediated ferroptosis in AML cells. (a) Control or FSP1-knockdown cells of MOLM-13 and THP-1, were treated with RSL3 or Erastin at gradient concentrations for 48 h, respectively. Cell viability was measured using CTG assay. (b) Pathway enrichment analysis heatmap illustrating the expression levels of key ferroptosis-related genes in FSP1-knockdown (shFSP1) versus control (shCtrl) MOLM-13 cells. (c) Representative mitochondrial morphology of MOLM-13 cells, captured by TEM. Scale bar: 2 μm (left), 1 μm (right). Orange arrowheads indicate mitochondria with ferroptotic features like smaller size and increased membrane density. (d-e) Flow cytometry analysis (d) and statistical results (e) showing lipid peroxidation levels detected by BODIPY 581/591 C11. MOLM-13 Cells with shCtrl, shNSUN2, shFSP1, and shNSUN2 co-transfected with either wild-type (FSP1-WT) or mutant FSP1 (FSP1-Mut), were treated with vehicle and RSL3 (200 nM) for 75 min. Data are presented as mean ± SD, n = 3. (f) TEM images of femoral bone marrow cells from mice infused with MOLM-13-luc cells transduced with control (shCtrl), FSP1 knockdown (shFSP1), or NSUN2 knockdown (shNSUN2). Top panels show 2500× magnification, and bottom panels show 8300× magnification. (g) Flow cytometry analysis of femoral bone marrow cells from mice infused with MOLM-13-luc cells transduced with control (shCtrl), FSP1 knockdown (shFSP1), or NSUN2 knockdown (shNSUN2). The left panels show representative flow cytometry plots of lipid peroxidation (Ox.C11) and hCD45 expression. The right bar graph shows the quantification of lipid peroxidation levels (%). Data are presented as mean ± SD, *P < 0.05, **P < 0.01 NSUN2 inhibitor MY-1B exhibit anti-leukemic activity and synergize with RSL3 and standard AML therapies To systematically characterize the anti-leukemic potential of the NSUN2 inhibitor MY-1B and the FSP1 inhibitor iFSP1, we first evaluated their cytotoxicity across ten AML cell lines. MY-1B demonstrated potent activity, with 48-hour IC[50] values ranging from 3.816 µM (MOLM-13) to 11.930 µM (OCI-AML2) (Fig. [167]6a, Supplementary Fig. 6f), while iFSP1 exhibited higher IC[50] values, ranging from 8.489 µM (MOLM-13) to 44.022 µM (Kasumi-1) (Supplementary Fig. [168]4a, Supplementary Fig. [169]6g). Mechanistic profiling in MOLM-13 and THP-1 cells revealed that MY-1B treatment dose-dependently decreased global RNA m^5C modifications levels (Fig. [170]6b) and suppressed FSP1 protein expression (Fig. [171]6c). Notably, pretreatment with MY-1B or iFSP1 for 24 h followed by exposure to RSL3 synergistically amplified lipid peroxidation, with ROS levels exceeding those of single-agent treatments (Fig. [172]6d–g). Synergy analysis using the BLISS model further confirmed the combinatorial effects of MY-1B or iFSP1 with RSL3. In MOLM-13 cells, MY-1B combined with RSL3 exhibited robust synergy (scores: -5.23 to 24.86), especially at higher drug concentrations (Fig. [173]6h), while iFSP1 required elevated doses to achieve strong synergy (scores: -0.65 to 16.59, Supplementary Fig. [174]4b). In THP-1 cells, MY-1B + RSL3 combination demonstrated stable synergy (scores: 7.40–23.90, Fig. [175]6i), whereas iFSP1 combined with RSL3 exhibited a dose-dependent synergy profile with peak scores at 14.73 (Supplementary Fig. [176]4c). Fig. 6. [177]Fig. 6 [178]Open in a new tab NSUN2 inhibitor MY-1B exhibit anti-leukemic activity and synergize with RSL3 and standard AML therapies. (a) Dose-response curves of MY-1B in AML cell lines (MV4-11, THP-1, MOLM-13, HL-60, OCI-AML2, OCI-AML3) after 48 h treatment. Calculated IC[50] values for each cell line are indicated. (b) Dot-blot analysis showing global RNA m^5C modification levels in MOLM-13 and THP-1 cells treated with indicated concentrations of MY-1B (0, 2, 4 µM) for 48 h. (c) Western blot analysis of FSP1 protein expression in MOLM-13 and THP-1 cells upon MY-1B treatment (0, 2, 4 µM) for 48 h. α-tubulin served as a loading control. (d-g) Representative flow cytometry plots and quantification of lipid ROS levels in MOLM-13 (d-e) and THP-1 (f-g) cells treated with vehicle, MY-1B, iFSP1, RSL3, or indicated combinations (MY-1B + RSL3, iFSP1 + RSL3) for 75 min. Lipid peroxidation (%) levels were shown. (h-i) Synergistic anti-leukemia effects of MY-1B and RSL3 combinations. Left panels: Heatmaps illustrating viability reduction in MOLM-13 (h) and THP-1 (i) cells treated with MY-1B and RSL3 at indicated concentrations for 48 h. Color gradients represent viability percentages normalized to untreated controls. Right panels: Corresponding synergy landscapes calculated using BLISS model. Positive values (orange) indicate synergy and negative values (blue) denote antagonism. Data presented as mean ± SD (n = 3). (j-l) Synergistic efficacy of MY-1B with standard chemotherapies or Venetoclax. Left panels: Viability matrices of MOLM-13 cells treated for 48 h with MY-1B (0–4 µM) combined with (j) Daunorubicin (DNR, 0–10 nM), (k) Cytarabine (Ara-C, 0-250 nM), or (l) Venetoclax (0-100 nM). Right panels: BLISS synergy score landscapes for each combination. (m-n) Annexin V/PI assay for drug combinations: RSL3 (200 nM), daunorubicin (DNR, 7.5 nM), cytarabine (Ara-C, 250 nM), Venetoclax (100 nM) with MY-1B (4 µM) or iFSP1 (16 µM) in MOLM-13 cells. Representative flow cytometry plots (m) and quantification (n) are shown. Data are presented as mean ± SD (n = 3) To further evaluate the specificity and tolerability of NSUN2/FSP1 inhibition, we expanded our analysis to include normal hematopoietic progenitors and primary AML samples. CD34⁺ hematopoietic progenitors from healthy donors (n = 4) were treated with MY-1B or iFSP1 across a range of concentrations (0.125–32 µM). Cell viability remained largely unaffected at lower doses, with only a modest reduction observed at the highest concentrations, indicating good tolerability in normal hematopoietic cells (Supplementary Fig. [179]6a–c). In contrast, primary AML mononuclear cells (n = 8) exhibited a dose-dependent decrease in viability following treatment with MY-1B or iFSP1 inhibitors (Supplementary Fig. [180]6d, e), confirming preferential cytotoxicity in AML cells. Extending these analyses to standard AML chemotherapies, combination of MY-1B with daunorubicin or cytarabine showed concentration-dependent, moderate-to-strong synergy in both MOLM-13 (daunorubicin scores: -1.06 to 15.95, cytarabine scores: 0.73 to 13.26) and THP-1 cells (daunorubicin scores: -3.62 to 23.30, cytarabine scores: -5.90 to 16.79) (Fig. [181]6j-k, Supplementary Fig. [182]4d-e). In contrast, iFSP1 displayed cell type-specific effects, showing synergistic with daunorubicin in both MOLM-13 (scores: -10.92 to 12.43) and THP-1 (scores: -8.37 to 17.10) (Supplementary Fig. [183]5a, d), but antagonistic with cytarabine in MOLM-13 (scores: -16.77 to 0.69) and weak synergy in THP-1 (scores: -0.73 to 15.87; Supplementary Fig. 5b, e). Remarkably, MY-1B combined with venetoclax consistently exhibited strong synergistic effects in both cell lines (MOLM-13: 2.07 to 28.03; THP-1: 8.57 to 25.18), resulting in near-complete cytotoxicity at higher concentrations (Fig. [184]6l, Supplementary Fig. [185]4f). Similarly, iFSP1 in combination with venetoclax displayed pronounced synergy in THP-1 cells (scores: -3.02 to 24.77) and substantial synergy in MOLM-13 cells (scores: -2.34 to 18.73) (Supplementary Fig. [186]5c, f). Consistent with these synergy analyses, Annexin V/PI apoptosis assays validated that combination treatments markedly enhanced cell death in MOLM-13 and THP-1 cells compared to monotherapies. (Fig. [187]6m-n, Supplementary Fig. [188]5g-i). Discussion Despite recent advances in targeted and immune-chemotherapeutic strategies, acute myeloid leukemia (AML) remains a devastating malignancy with poor survival outcomes. There is an urgent need to identify novel molecular vulnerabilities and therapeutic targets. This study reveals a pivotal role for NSUN2-mediated RNA m⁵C methylation in conferring ferroptosis resistance in acute myeloid leukemia (AML), establishing a direct link between epitranscriptomic regulation and cell death evasion. We demonstrate that NSUN2 is significantly overexpressed in AML patients and serves as a robust prognostic marker, correlating with aggressive disease progression and inferior survival outcomes across multiple patient cohorts. Mechanistically, NSUN2 catalyzes m⁵C deposition within the 3’UTR of FSP1 mRNA, facilitating its recognition by the RNA-binding protein YBX1, which stabilizes the transcript and sustains FSP1 protein expression. This axis protects AML cells from ferroptotic stress by suppressing lipid peroxidation. Mutational inactivation of NSUN2 (C271A/C321A) or FSP1 (G2A/E156A) abolished this protective effect, confirming the functional specificity of the pathway. The discovery of NSUN2 as a critical regulator of AML progression expands the functional repertoire of RNA modifications in hematologic malignancies. While NSUN2 has been previously implicated in stabilizing oncogenic transcripts in solid tumors, its role in AML remained unexplored. Our findings reveal that NSUN2-driven m⁵C methylation is not merely a passive modification but actively rewires AML cell survival pathways by maintaining FSP1 expression. This observation aligns with emerging evidence that RNA modifications dynamically regulate cancer cell adaptability, yet diverges from prior studies by identifying ferroptosis evasion as a novel output of epitranscriptomic reprogramming. Notably, FSP1, a well-characterized ferroptosis suppressor, emerges here as a critical dependency in AML, bridging RNA methylation to redox homeostasis. The prognostic significance of FSP1 overexpression in AML patients further underscores its clinical relevance, suggesting that FSP1 could serve as a biomarker for risk stratification and therapeutic targeting. The mechanistic interplay between NSUN2 and FSP1 unveils a previously unrecognized epitranscriptomic circuit that operates independently of the canonical GPX4-glutathione axis. By stabilizing FSP1 mRNA, NSUN2 ensures continuous regeneration of reduced CoQ10, thereby neutralizing lipid peroxidation and preserving membrane integrity. This pathway’s reliance on YBX1, a multifunctional RNA-binding protein, emphasizes the importance of RNA-protein interactions in AML biology. The rapid degradation of FSP1 mRNA upon NSUN2 or YBX1 depletion, as evidenced by reduced transcript half-life, further substantiates the dynamic regulation of this axis. These findings expand the predominantly DNA- and m^6A-centric view of epigenetic dysregulation in AML, advocating a broader perspective that incorporates m^5C-based epitranscriptomic mechanisms into the molecular framework of leukemia progression. Although NSUN2 or FSP1 depletion did not elicit spontaneous lipid peroxidation under basal conditions, transmission electron microscopy revealed marked mitochondrial alterations, including reduced size, increased membrane density, and cristae collapse. These ultrastructural features are consistent with a primed ferroptotic state, suggesting that NSUN2-FSP1 loss lowers the ferroptosis threshold without directly triggering cell death. This “ferroptosis priming” sensitizes AML cells to ferroptosis inducers, revealing a two-step regulatory mechanism—whereby epitranscriptomic disruption preconditions the cells, and external stimuli execute ferroptosis. In vitro, NSUN2 or FSP1 inhibitors (MY-1B, iFSP1) synergized with ferroptosis inducers (RSL3, Erastin) to enhance AML killing. This uncoupling of mitochondrial priming from immediate lipid peroxidation underscores the importance of NSUN2-FSP1 in maintaining cellular redox resilience and points to a previously underappreciated layer of ferroptosis regulation. This ferroptosis-sensitized state also creates a therapeutic window wherein co-targeting the NSUN2-FSP1 axis and existing therapies. The synergistic effects of MY-1B and iFSP1 with venetoclax, a BCL-2 inhibitor widely used in AML regimens, are particularly compelling. Venetoclax resistance, often driven by compensatory anti-apoptotic mechanisms, remains a major barrier to durable remission. Our data suggest that co-inhibition of NSUN2-FSP1 and BCL-2 disrupts parallel survival pathways—ferroptosis protection and apoptotic blockade, thereby circumventing resistance. The differential synergy patterns observed between MY-1B/iFSP1 and chemotherapeutic agents (daunorubicin, cytarabine) further highlight the context-dependent nature of drug interactions, advocating for tailored combinatorial approaches based on molecular profiles. Critically, our data demonstrate that NSUN2/FSP1 inhibition exerts selective cytotoxicity toward AML cells while sparing normal hematopoietic progenitors. In viability assays, CD34^+ progenitor cells from healthy donors maintained robust survival across a broad concentration range of MY-1B and iFSP1. Conversely, primary AML cells and ten distinct AML cell lines exhibited pronounced, dose-dependent sensitivity. These findings underscore the translational potential of NSUN2/FSP1 inhibitors, offering a favorable therapeutic window for clinical application. In addition, several questions remain to be addressed. Preliminary experiments included HEL and K562 cell lines (derived from erythroleukemia and chronic myeloid leukemia in blast crisis, respectively), which exhibited initial sensitivity to NSUN2 or FSP1 inhibitors. However, we ultimately excluded these non-AML models from our final analysis as there is currently insufficient evidence to support extending our findings to other types of hematologic malignancies. Therefore, all mechanistic and therapeutic interpretations in this study are confined to AML, while the implications of NSUN2/FSP1 inhibition in other hematologic malignancies remain to be investigated. Furthermore, preclinical validation of MY-1B and iFSP1 is required, including pharmacokinetic profiling, safety evaluation, and efficacy assessments. Additionally, potential crosstalk between m⁵C and other RNA modifications, such as m⁶A or pseudouridylation, may form complex regulatory networks that shape ferroptosis sensitivity. Moreover, longitudinal tracking of NSUN2/FSP1 expression during AML therapy, particularly under venetoclax-based regimens, could shed light on resistance dynamics and offer predictive biomarkers. In conclusion, we identify NSUN2 and FSP1 as key regulators of ferroptosis resistance and therapeutic vulnerability in AML. The NSUN2-YBX1-FSP1 axis provides a novel epitranscriptomic mechanism that links RNA methylation to redox control and cell survival. Targeting this axis in combination with ferroptosis or apoptosis inducers offers a promising strategy to overcome therapy resistance and improve clinical outcomes in AML. Conclusion Our study uncovers an epitranscriptomic mechanism in which NSUN2-mediated m⁵C modification stabilizes FSP1 mRNA via YBX1, promoting ferroptosis resistance and AML progression. NSUN2 and FSP1 are both upregulated in AML and predict poor prognosis. Disruption of this axis sensitizes leukemic cells to ferroptosis inducers and enhances responses to chemotherapy and venetoclax. These findings identify the NSUN2–YBX1–FSP1 circuit as a key regulator of ferroptotic vulnerability and highlight RNA methylation as a therapeutic target. This work supports a dual-pathway strategy to overcome resistance in AML and encourages further exploration of RNA modification–based interventions in hematologic malignancies. Electronic supplementary material Below is the link to the electronic supplementary material. [189]Supplementary Material 1^ (2.2MB, pdf) [190]Supplementary Material 2^ (553.4KB, pdf) [191]Supplementary Material 3^ (92.2KB, xlsx) [192]Supplementary Material 4^ (773.4KB, xlsx) [193]Supplementary Material 5^ (10.8KB, xlsx) [194]Supplementary Material 6^ (1.1MB, pdf) Acknowledgements