Hexokinase 2-mediated histone H3K18la promotes PAI-1-dependent thrombosis in acute myeloid leukemia via tumor-endothelial crosstalk

Abstract

Acute myeloid leukemia is an aggressive hematological malignancy frequently complicated by coagulation disorders, including thrombosis and hemorrhage, which contribute to poor outcomes. Here, we identify lactate-driven histone lactylation as a mechanism promoting thrombosis in acute myeloid leukemia. We demonstrate that hexokinase 2-mediated glycolysis in leukemic cells leads to lactate accumulation, which enhances histone H3 lysine 18 lactylation and upregulates plasminogen activator inhibitor-1 expression, impairing fibrinolysis. Lactate released by acute myeloid leukemia cells is internalized by vascular endothelial cells via monocarboxylate transporter 1, amplifying plasminogen activator inhibitor-1 expression and thrombotic risk. Inhibition of hexokinase 2-mediated lactate production or monocarboxylate transporter 1-mediated lactate uptake attenuates thrombosis. Our findings reveal a critical link between tumor metabolism, epigenetic modifications, and coagulation dysfunction in acute myeloid leukemia.

The molecular mechanisms underlying thrombosis in acute myeloid leukemia (AML), which contribute to poor clinical outcomes, remain to be explored. Here, the authors show that hexokinase 2 (HK2)-mediated glycolysis leads to lactate-driven histone lactylation, inducing thrombosis in AML.

Introduction

Thromboembolism (TE) is a prevalent complication in cancer patients and represents the second leading cause of mortality among individuals with malignant diseases¹. Compared to the general population, cancer patients exhibit a sevenfold increased risk of thrombosis². In hematological malignancies, thrombocytopenia and neutropenia are common at diagnosis and throughout the disease course, leading to a predominant focus on infection- and bleeding-related morbidity and mortality. Consequently, TE has historically been underrecognized in this population. However, emerging evidence indicates that the risk of thrombosis in hematological malignancies is substantial, often comparable to or even exceeding that of solid cancers^3–5. Acute myeloid leukemia (AML) is particularly associated with a high incidence of thrombotic complications. Venous thromboembolism (VTE) frequently involves cerebral venous sinuses, portal veins, and hepatic veins, while arterial thrombosis can manifest as peripheral artery occlusion, myocardial infarction, or ischemic stroke⁶. Clinical studies have demonstrated that coagulation disorders are independently associated with failure to achieve remission in AML patients, and acute promyelocytic leukemia (APL) patients with thrombosis exhibit higher mortality rates compared to those without thrombosis⁷. Risk factors for thrombotic events in AML include mutations in WT-1 and FLT3/ITD, the SERPINE1 gene 4 G/4 G polymorphism, CD15 expression, and high white blood cell counts^7,8. Despite these associations, the precise mechanisms driving thrombosis in AML remain poorly understood, necessitating further investigation.

A hallmark of cancer metabolism is the shift from oxidative phosphorylation to aerobic glycolysis, known as the Warburg effect, which is particularly pronounced in leukemic cells. This metabolic reprogramming results in increased glucose consumption and lactate production^9–11. Elevated lactate levels are associated with poor 5-year overall survival and event-free survival in AML patients¹². Beyond its role as a metabolic byproduct, lactate has emerged as a key substrate for histone lactylation, a recently discovered post-translational modification that regulates gene transcription and influences diverse pathological processes¹³. However, the role of histone lactylation in AML-associated coagulation disorders remains unclear.

In this study, we investigate the interplay between lactate metabolism, histone lactylation, and coagulation dysfunction in AML. Our findings demonstrate that the crosstalk between leukemia cells and endothelial cells within the tumor microenvironment amplifies thrombotic propensity. This research provides mechanistic insights into the pathogenesis of thrombosis in AML.

Results

Elevated lactate associates with histone lactylation and thrombotic risk in AML

Enhanced glycolysis in tumor cells has been widely recognized^14–16. To investigate whether AML blasts display a similar metabolic alteration, we analyzed glycolytic activity in peripheral blood mononuclear cells (PBMCs) from AML patients and healthy controls. Extracellular acidification rate (ECAR) demonstrated significantly elevated glycolytic activity in AML-derived PBMCs (Fig. 1a). Considering the established relationship between glycolytic activation and lactate accumulation^11,17,18, coupled with clinical observations linking elevated lactate to inferior 5-year overall survival in AML¹², we quantified circulating lactate levels. AML patients exhibited substantially increased plasma lactate concentrations relative to healthy individuals (Fig. 1b). To investigate the clinical implications of elevated lactate in AML, we stratified patients into two groups based on lactate levels and analyzed the incidence of thrombotic events. Notably, AML patients with higher lactate exhibited increased incidence of thrombotic events (Fig. 1c), suggesting a potential link between lactate accumulation and thrombosis in AML. To further explore this association, we performed correlation analysis between plasma lactate and D-dimer, a well-established marker of thrombosis^19,20. The results showed that lactate was positively correlated with D-dimer in newly diagnosed AML patients (Fig. 1d), supporting the connection between lactate and thrombotic risk.

Fig. 1. Elevated lactate associates with histone lactylation and thrombotic risk in AML.

a ECAR of PBMCs isolated from AML patients and healthy donors, as determined by Seahorse metabolic analyzer (n = 6 biological replicates), mean ± SEM. b Plasma lactate in AML patients (n = 48) and healthy donors (n = 32). French-American-British (FAB) subtypes of 48 AML patients: M1 (n = 7), M2 (n = 11), M3 (n = 14), M4 (n = 3), and M5 (n = 13). c Cumulative incidence of thrombosis in AML patients stratified by plasma lactate (n = 23 for normal lactate, 2 thrombosis events [M3 and M5; Patients 47 and 24]; n = 25 for elevated lactate, 5 events [2 M1, M3, M4, M5; Patients 26, 28, 19, 27, 20]). d Pearson correlation between plasma lactate and D-dimer in AML patients (n = 42). e Experimental timeline for sodium lactate (NaLa) intervention in AML mice (by figdraw.com). f Bioluminescence imaging of leukemic burden in murine models, prior to IVC stenosis. g Quantitative assessment of plasma lactate in murine models using a commercial Lactate Assay Kit. h Representative images and weights of thrombi from IVC 48 h post-stenosis. i Representative immunohistochemical staining images of Pan-Kla in bone marrow specimens from healthy donors (n = 10) and AML patients (n = 11 per group). Boxed areas are magnified; arrows indicate positive cells (scale bar=25 μm). j, k Western blot of Pan-Kla and site-specific histone lactylation in PBMCs. PBMCs from healthy donors were pooled (n = 5 donors per pool). D-dimer<4 mg/L (n = 8, included 1 APL (M3), FAB subtypes shown in the figure are M3, M5, and M5). D-dimer>4 mg/L (n = 9, included 3 APL, FAB subtypes shown in the figure are M1, M3, M2). f–h n = 6 mice per group. Data were analyzed by two-tailed Student’s t test (g), two-tailed Mann-Whitney test (a, b, h), two-way ANOVA (c), Pearson correlation analysis (d), one-way ANOVA with Tukey’s multiple comparisons test (i). Data are presented as mean ± SD. Source data are provided as a Source Data file.

To validate lactate’s thrombogenic potential in AML, we generated xenograft models through intravenous injection of luciferase-expressing AML cell lines (NB4 or MOLM13) into NCG mice. Leukemia engraftment was confirmed by bioluminescence imaging, after which mice were randomized into two groups and treated with sodium lactate or saline for 3 days. Thrombosis was then induced by inferior vena cava (IVC) stenosis surgery (Fig. 1e). Leukemia cells infiltration in NCG mice was shown in Fig. 1f. Sodium lactate administration significantly increased plasma lactate levels compared to the saline group (Fig. 1g) without affecting blood cell counts (Supplementary Fig. 1). Importantly, sodium lactate-treated mice developed significantly larger thrombi following IVC partial obstruction (Fig. 1h). These findings indicate that elevated lactate might promote thrombosis in AML, both in clinical settings and experimental models.

Given the established role of lactate as a precursor for protein lactylation, we hypothesized that lactate-mediated histone lactylation might contribute to AML-associated coagulopathy. Since elevated D-dimer levels (>4 mg/L) are a strong predictor of thrombosis in newly diagnosed AML patients^19,20, we next investigated whether lactate-induced histone lactylation might contribute to the prothrombotic state in high D-dimer AML patients. Indeed, we observed elevated levels of pan-lysine lactylation (Pan-Kla), particularly histone lactylation, in bone marrow mononuclear cells (BMMCs) and PBMCs of AML patients with D-dimer levels > 4 mg/L (Fig. 1I, j). Further analysis revealed significant increases in lactylation at histone residues H3K9, H3K18, H4K5, and H4K12 in these patients (Fig. 1k), suggesting that histone lactylation may play a functional role in AML-associated thrombosis.

Lactate links to PAI-1 upregulation and impair fibrinolysis through histone lactylation in AML

To explore potential mechanisms underlying the lactate-thrombosis association, we treated AML cells with exogenous lactate and observed a significant induction of Pan-Kla and specific histone lactylation residues (Fig. 2a, Supplementary Fig. 2a, b). RNA sequencing (RNA-seq) analysis of lactate-treated NB4 cells (Lac) compared to control cells (Control) revealed 582 upregulated and 189 downregulated genes (Fig. 2b). Gene ontology (GO) analysis highlighted significant enrichment of pathways related to hemostasis and blood coagulation regulation in lactate-treated cells (Fig. 2c). The heatmap of the RNA-seq data displayed genes enriched in the coagulation process (Fig. 2d), with a significant upregulation of SERPINE1 mRNA, which encodes plasminogen activator inhibitor 1 (PAI-1). PAI-1 inhibits plasmin generation by forming a complex with plasminogen activators, thereby suppressing fibrin degradation. Dysregulation of PAI-1 is implicated in thrombotic events and abnormal bleeding^21–23. Recent studies have also linked elevated PAI-1 levels to an increased risk of deep venous thrombosis (DVT) in AML patients, identifying PAI-1 as a potential biomarker for DVT in acute leukemia²⁴. The RNA-seq findings were validated by qPCR and Western blot, confirming increased PAI-1 expression in lactate-treated AML cells (Fig. 2e, f, Supplementary Fig. 2c, d).

Fig. 2. Lactate links to PAI-1 upregulation and impair fibrinolysis through histone lactylation in AML.

a NB4 and MOLM13 cells were exposed to 20 mM lactate for 12 or 24 hours, the levels of Pan-Kla were detected by Western blot (n = 3 biological replicates). b Volcano plot depicting the differential gene expression between lactate-treated NB4 cells (Lac) and controls (n = 3 biological replicates). Upregulated genes are shown in purple, and downregulated genes are shown in green (adjusted p-values < 0.05 and | log₂(fold change) |>2). c Bubble chart showing gene ontology terms enriched in upregulated genes (n = 3 biological replicates). d Heatmap showing the expression of differential genes in the coagulation pathway from RNA-seq data (n = 3 biological replicates). e, f NB4 and MOLM13 cells were exposed to 20 mM lactate, followed by qPCR and Western blot to assess PAI-1 expression (n = 3 biological replicates). g Cells were tested for plasmin generation with plasminogen and the substrate S2251, the absorbance (Abs) was continuously measured at 405 nm (n = 3 biological replicates). h Fibrinolysis assays were conducted in lactate-treated NB4 and MOLM13 cells, fibrinolysis rate was recorded based on turbidity at 405 nm, dashed line represents 50% lysis of the clot (n = 3 biological replicates). i TAT complex in plasma was measured using an ELISA kit (n = 6 mice per group). j Representative immunofluorescence images showing Pan-Kla and PAI-1 expression in mouse bone marrow cells (scale bar=25 μm) (n = 6 mice per group). k PAI-1 levels in mouse plasma measured using a commercial ELISA kit (n = 6 mice per group). l Representative immunofluorescence images of PAI-1 content (yellow) in thrombi from thrombotic IVC segments of mice (scale bar = 200 μm) (n = 6 mice per group). m ELISA quantification of t-PA/PAI-1 complex levels in mouse plasma (n = 6 mice per group). Data were analyzed by two-tailed Wald test with Benjamini-Hochberg multiple comparisons (b), two-tailed hypergeometric test with Benjamini-Hochberg multiple comparisons (c), two-tailed Student’s t test (e, i, l, m), two-way ANOVA with Sidak’s multiple comparisons test (g, h), two-tailed Mann-Whitney test (k). Data are presented as mean ± SD. Source data are provided as a Source Data file.

In addition, we assessed the impact of endogenously increased lactate production in AML cells on the expression of PAI-1. AML cells were treated with rotenone which could switch oxidative phosphorylation to glycolysis, leading to increased lactate production. Consistent with exogenous lactate treatment, rotenone-induced lactate accumulation significantly increased Pan-Kla levels and PAI-1 expression (Supplementary Fig. 2e–h).

To assess the functional impact of lactate-mediated PAI-1 upregulation on fibrinolysis, we performed plasmin generation and fibrinolysis assays. Lactate-treated AML cells exhibited reduced plasmin production and prolonged fibrinolysis (Fig. 2g, h, Supplementary Fig. 2i, j), while fibrin production remained unchanged. (Supplementary Fig. 2k). Consistent with these findings, plasma levels of thrombin-antithrombin (TAT) complex in sodium lactate-treated AML mice were comparable to controls (Fig. 2i), suggesting that lactate-induced thrombosis is not driven by enhanced coagulation activation. Next, we focused on the potential role of lactate-mediated histone lactylation in enhancing PAI-1 expression and its implications in thrombosis, multiplex immunofluorescence was employed to evaluate Pan-Kla and PAI-1 levels in mouse bone marrow. Sodium lactate treatment significantly increased global lactylation and PAI-1 expression in BMMCs (Fig. 2j). Furthermore, sodium lactate elevated plasma PAI-1 levels (Fig. 2k) and promoted PAI-1 accumulation in venous thrombi following IVC stenosis (Fig. 2l). Since PAI-1 inhibits fibrinolysis by forming complex with t-PA, we further found elevated concentrations of plasma t-PA/PAI-1 complex in the sodium lactate -treated mouse (Fig. 2m). Taken together, these results indicated that lactate enhances thrombosis in AML by upregulating PAI-1 expression through histone lactylation, thereby impairing thrombus resolution.

HK2-mediated lactate production enhances PAI-1 expression through H3K18la

Tumor cells exhibit enhanced glycolytic flux through upregulation of rate-limiting glycolytic enzymes, culminating in lactate accumulation¹⁴. Building upon recent discoveries that histone lactylation can be mediated by both p300/CREB-binding protein C (CBP) acetyltransferases and aminoacyl-tRNA synthetases (AARS)^25,26, we sought to elucidate the key glycolytic regulator responsible for lactate overproduction and PAI-1 regulation in AML through comprehensive expression profiling. Our analysis revealed that among glycolytic enzymes and lactyltransferases examined, hexokinase 2 (HK2) displayed the most pronounced upregulation, with significantly elevated expression at both mRNA and protein levels in AML patient-derived PBMCs compared to healthy controls (Fig. 3a, b). This prominent overexpression of HK2 suggested its potential role in promoting glycolytic flux in AML. To further investigate this relationship, we performed correlation analysis between HK2 expression in patient PBMCs and corresponding plasma lactate, we observed a strong positive correlation (Supplementary Fig. 3a), supporting HK2 as a major regulator of lactate production in AML^27,28. To identify whether HK2-dependent lactate production is sufficient to drive histone lactylation in AML, we generated HK2-knockdown AML cells (Fig. 3c, Supplementary Fig. 3b). HK2 inhibition significantly reduced the glycolysis rate (Fig. 3d, Supplementary Fig. 3c), decreased lactate production, and histone lactylation (Fig. 3e, f), particularly at H3K9, H3K18, H4K5, and H4K12 residues (Supplementary Fig. 3d). Additionally, lactate supplementation rescued the reduced histone lactylation levels in HK2-knockdown AML cells (Fig. 3g), with H3K9, H3K18, and H4K5 lactylation showing a concentration-dependent recovery (Fig. 3h). Consistent with these findings, HK2 knockdown downregulated PAI-1 expression at both mRNA and protein levels (Fig. 3i, j), while lactate supplementation reversed this downregulation (Fig. 3k, l). Functional assays demonstrated that HK2 knockdown enhanced plasmin generation and accelerated fibrinolysis (Supplementary Fig. 3e, 3m).

Fig. 3. HK2-mediated lactate production enhances PAI-1 expression through H3K18la.

a mRNA levels of glycolytic enzymes and lactyltransferases were assessed by qPCR in PBMCs from healthy donors and AML patients (HK2, AARS1, P300, CBP: n = 24 vs. 24; LDHA: n = 17 vs. 18; PFKP: n = 22 vs. 18; PKM2: n = 23 vs. 24). b Western blot analysis of HK2, PFKP, PKM2 and AARS1 in PBMCs from healthy donors and AML patients (n = 5), PBMCs from healthy donors were pooled (n = 5 donors per pool). c The knockdown efficiency of HK2 was evaluated by Western blot (n = 3 biological replicates). d Seahorse Metabolic Analysis of Extracellular Acidification Rate in HK2 knockdown-NB4 Cells (n = 3 biological replicates), mean ± SEM. e Lactate concentration in culture medium after HK2 knockdown (n = 3 biological replicates). f Pan-Kla levels in HK2-knockdown cells (n = 3 biological replicates). g, h HK2-knockdown cells were treated with increasing concentrations of lactate for 12 hours, and Pan-Kla as well as site-specific histone lactylation were analyzed by Western blot (n = 3 biological replicates). i, j The mRNA and protein levels of PAI-1 in HK2-knockdown cells were detected by qPCR and Western blot, respectively (n = 3 biological replicates). k, l The expression of PAI-1 in HK2-knockdown cells exposed to different concentrations of lactate for 12 hours was assessed by qPCR and Western blot, respectively (n = 3 biological replicates). m Fibrinolysis assays in HK2-knockdown cells, with lysis rate continuously monitored at 405 nm, dashed line represents 50% lysis of the clot (n = 3 biological replicates). n ChIP-qPCR analysis of H3K18la enrichment at the PAI-1 promoter in lactate-treated NB4 cells (n = 3 biological replicates). o Relative occupancy of H3K18la at the PAI-1 promoter in shHK2 NB4 cells was determined by ChIP-qPCR (n = 3 biological replicates). p CUT&Tag tracks showing H3K18la enrichment at the SERPINE1 (PAI-1) promoter (n = 1 biological replicates). shNC, negative control; shHK2, shRNA targeting HK2. Data were analyzed by two-tailed Mann-Whitney test (a), two-tailed Student’s t test (d, e, i, n, o), one-way ANOVA with Dunnett’s multiple comparisons test (k), two-way ANOVA with Sidak’s multiple comparisons test (m). Data are presented as mean ± SD. Source data are provided as a Source Data file.

To identify the specific histone lactylation residue responsible for PAI-1 regulation in AML cells, we conducted ChIP-qPCR using antibodies specific to lactylated histone residues. The results revealed that lactate exposure significantly increased H3K18la enrichment at the PAI-1 promoter (Fig. 3n), whereas HK2 knockdown reduced this enrichment (Fig. 3o). CUT&Tag analysis in HK2-knockdown cells further revealed that HK2 knockdown decreased H3K18la enrichment at the PAI-1 promoter (Fig. 3p). These findings indicate that H3K18 lactylation mediates PAI-1 upregulation. Conversely, HK2 overexpression in leukemic cells increased lactate production, histone lactylation (Supplementary Fig. 3f–i), and PAI-1 expression (Supplementary Fig. 3j, k), while suppressing plasmin activation and prolonging fibrinolysis (Supplementary Fig. 3l, m). Taken together, these data suggest that HK2-mediated lactate production enhances PAI-1 expression via H3K18 lactylation at its promoter.

HK2 knockdown in leukemic cells attenuates thrombosis in AML mice

To validate the role of HK2-mediated lactate production in AML-associated thrombosis in vivo, we established xenograft models by injecting HK2-knockdown NB4 or MOLM13 cells into NCG mice (Fig. 4a). After confirming leukemia engraftment by bioluminescent imaging (Fig. 4b), thrombosis was induced by IVC stenosis. lower lactate levels, and decreased thrombus weight were observed in mice injected with HK2-knockdown cells (Fig. 4c, d), without significant alterations in blood cell counts or TAT complex levels (Supplementary Fig. 4a, b). Knockdown of HK2 induced a reduction in lactate, leading to decreased H3K18 lactylation, and PAI-1 levels in mouse BMMCs (Fig. 4e), Additionally, mice transplanted with HK2-knockdown cells showed decreased PAI-1 levels in both plasma and thrombi (Fig. 4f, g), as well as reduced t-PA/PAI-1 complex content (Fig. 4h). To further explore the clinical relevance of the HK2/H3K18la/PAI-1 signaling axis, we performed multicolor immunofluorescence analysis on BMMCs from AML patients. Notably, HK2, H3K18la, and PAI-1 were upregulated in AML patients with D-dimer levels>4 mg/L (Supplementary Fig. 4c), supporting the clinical relevance of the HK2/H3K18la/PAI-1 axis in AML-associated thrombosis. These data provide evidence that HK2-mediated lactylation at the H3K18 residue promotes PAI-1 expression in AML cells, potentially contributing to the pathogenesis of thrombosis in AML.

Fig. 4. HK2 knockdown in leukemic cells attenuates thrombosis in AML mice.

a The outline depicts the steps of the animal experiments (by figdraw.com). b NGG mice were injected with shHK2 or shNC AML cells via tail vein, and leukemia engraftment was confirmed by bioluminescence imaging (n = 6 mice per group). c Plasma lactate was measured using a Lactate Assay Kit (n = 6 mice per group). d Representative images and weights of thrombi isolated from mice 48 hours after IVC stenosis (n = 6 mice per group). e Immunofluorescence staining of H3K18la and PAI-1 in mouse bone marrow cells (scale bar=25 μm) (n = 6 mice per group). f Levels of PAI-1 in mice plasma were measured by ELISA (n = 6 mice per group). g Representative immunofluorescence images of PAI-1 (yellow) expression in thrombi from mouse IVC (scale bar = 200 μm) (n = 6 mice per group). h Plasma t-PA/PAI-1 complex levels were determined by ELISA (n = 6 mice per group). shNC, negative control; shHK2, shRNA targeting HK2. Data were analyzed by two-tailed Student’s t test (b, c, g, h), two-tailed Mann-Whitney test (d, f). Data are presented as mean ± SD. Source data are provided as a Source Data file.

AML cells-derived lactate promotes endothelial PAI-1 expression through histone lactylation

We observed a significant elevation in plasma PAI-1 levels in AML mice treated with sodium lactate (Fig. 2k). Furthermore, a positive correlation was detected between plasma lactate and PAI-1 concentrations in AML patients (Fig. 5a). Given that vascular endothelial cells are an important source of PAI-1 in vivo^29,30, we next investigated the expression of key molecules in the inferior vena cava of sodium lactate or saline treated-AML mice following IVC stenosis. Notably, the sodium lactate-treated group exhibited significantly higher levels of H3K18la, and PAI-1 in the IVC tissues (Fig. 5b), indicating that lactate supplementation enhances histone lactylation and PAI-1 expression in vascular endothelial cells. Based on these findings, we hypothesized that lactate released from AML cells could be taken up by vascular endothelial cells, subsequently promoting PAI-1 expression through histone lactylation. To validate this hypothesis, in vitro experiments were conducted by co-culturing HUVECs with supernatants collected from AML cell cultures at different time points. Intriguingly, supernatants from metabolically active AML cells significantly increased histone lactylation and H3K18la levels in endothelial cells (Fig. 5c, Supplementary Fig. 5a). Concurrently, PAI-1 expression in HUVECs was markedly elevated in the presence of metabolically active AML cell supernatants (Fig. 5d, e, Supplementary Fig. 5b, c). Additionally, plasminogen activation and fibrinolysis were inhibited under these conditions (Fig. 5f, g, Supplementary Fig. 5d, e). These data indicate that lactate or other metabolites in AML cell supernatants might absorb by endothelial cells, leading to the regulation of PAI-1 expression.

Fig. 5. AML cells-derived lactate promotes endothelial PAI-1 expression through histone lactylation.

a PAI-1 in plasma from AML patients (n = 39) was measured using a commercial assay kit, followed by Pearson correlation analysis with corresponding plasma lactate. b Immunofluorescence staining was performed to evaluate the expression of H3K18la, and PAI-1 in the IVC tissues of thrombus-forming segment. (CD31, an endothelial cell marker; scale bar=50 μm; n = 6 mice per group). c Conditioned media (CM) from NB4 and MOLM13 cells cultured for 12 or 24 hours were co-cultured with HUVECs. Pan-Kla and H3K18la levels in HUVECs were analyzed by Western blot (n = 3 biological replicates). d, e mRNA and protein expression of PAI-1 in HUVECs co-cultured with conditioned media from AML cell lines were assessed by qPCR and Western blot, respectively (n = 3 biological replicates). f Plasmin generation in HUVECs co-cultured with conditioned media from AML cell lines was continuously monitored at 405 nm (n = 3 biological replicates). g Fibrinolysis rate was monitored at 405 nm after co-culture of HUVECs with conditioned medium from AML cell lines, dashed line represents 50% lysis of the clot (n = 3 biological replicates). h, i HUVECs were transfected with siRNA targeting MCT1 (siMCT1), and MCT1 knockdown efficiency was assessed by qPCR and Western blot (n = 3 biological replicates). j siMCT1 HUVECs were co-cultured with conditioned media from THP-1 cells, and Pan-Kla and H3K18la levels were analyzed by Western blot (n = 3 biological replicates). k, l siMCT1 HUVECs were co-cultured with conditioned media from THP-1 cells, and PAI-1 mRNA and protein expression were assessed by qPCR and Western blot, respectively (n = 3 biological replicates). m Plasmin generation was continuously monitored at 405 nm (n = 3 biological replicates). n Fibrinolysis rate was continuously monitored at 405 nm after co-culture of THP-1-conditioned medium with HUVECs, dashed line represents 50% lysis of the clot (n = 3 biological replicates). siNC, negative control. Data were analyzed by Pearson correlation analysis (a), two-tailed Student’s t test (d, h, k), and two-way ANOVA with Sidak’s multiple comparisons test (f, g, m, n). Data are presented as mean ± SD. Source data are provided as a Source Data file.

Monocarboxylate transporters (MCTs), particularly MCT1, play a critical role in transporting extracellular lactate into cells^10,31,32. To determine whether lactate mediates the observed histone lactylation and PAI-1 upregulation in endothelial cells, we knocked down MCT1 in HUVECs prior to co-culture with AML cell supernatants (Fig. 5h, i). Silencing MCT1 significantly attenuated the AML cell supernatant-induced upregulation of histone lactylation and PAI-1 in HUVECs (Fig. 5j–l). Moreover, MCT1 deficiency similarly impaired plasminogen activation and fibrinolysis in HUVECs (Fig. 5m, n). Together, these findings indicate that AML-derived lactate enters endothelial cells via MCT1, promoting PAI-1 expression through histone lactylation.

HK2-mediated lactate production in AML cells promotes endothelial PAI-1 expression via H3K18la

To elucidate the role of HK2 in mediating lactate production by AML cells and its subsequent impact on endothelial lactylation, we conducted carbon-13 (¹³C) tracer experiment. Specifically, we labeled HK2-knockdown and control NB4 cells with [1,2-¹³C] glucose, which undergoes aerobic glycolysis to produce (M + 2) lactate. The labeled supernatant containing the radiolabeled metabolites was then co-cultured with HUVECs, and the amount of labeled lactate in the HUVECs was quantified (Fig. 6a). Metabolic flux analysis demonstrated that HUVECs co-cultured with the supernatant of HK2-knockdown NB4 cells exhibited significantly reduced total lactate levels (Fig. 6b) and a decreased proportion of (M + 2) lactate (Fig. 6c). Furthermore, HUVECs co-cultured with supernatants from HK2-knockdown AML cells showed decreased histone lactylation and H3K18la (Fig. 6d), along with a downregulation in both mRNA and protein expression of PAI-1 (Fig. 6e, f). As a result, plasminogen activation was enhanced, resulting in accelerated fibrinolysis (Fig. 6g, h). Conversely, co-culture of HUVECs with the supernatants from AML cells overexpressing HK2 reversed these effects (Supplementary Fig. 6). Consistent with these in vitro findings, AML mice transplanted with HK2-knockdown cells exhibited lower levels of H3K18la and PAI-1 in the IVC tissues following IVC stenosis (Fig. 6i). To further investigate the mechanistic link between H3K18 lactylation and PAI-1 expression, we performed ChIP assay to assess the enrichment of H3K18la on the PAI-1 promoter region. The results revealed a significant reduction in H3K18la enrichment on the PAI-1 promoter in HUVECs co-cultured with supernatants from HK2-knockdown AML cells (Fig. 6j). These data indicate that lactate produced through HK2-mediated metabolism in AML cells can be absorbed by endothelial cells, promoting H3K18 lactylation and subsequently enhancing PAI-1 expression.

Fig. 6. HK2-mediated lactate production in AML cells promotes endothelial PAI-1 expression via H3K18la.

a Schematic diagram: NB4 cells were cultured in [1,2-¹³C] glucose medium, and the labeled supernatant was co-cultured with HUVECs (by figdraw.com). Isotopically labeled metabolites in HUVECs were analyzed by metabolic flux analysis. b Heatmap showing the total levels of metabolites in HUVECs. c Metabolic flux analysis of (M + 2) lactate intensity in HUVECs co-cultured with conditioned media (CM) under different conditions. d Western blot analysis of Pan-Kla and H3K18la in HUVECs co-cultured with conditioned media from HK2-knockdown NB4 or MOLM13 cells. e, f mRNA and protein levels of PAI-1 in HUVECs co-cultured with conditioned media from AML cells were assessed by qPCR and Western blot, respectively. g Plasmin generation in HUVECs co-cultured with conditioned media from AML cells was monitored at 405 nm. h Fibrinolysis rate in HUVECs co-cultured with conditioned media from HK2-knockdown AML cells was monitored at 405 nm, dashed line represents 50% lysis of the clot. i Immunofluorescence staining was performed to evaluate the expression of H3K18la, and PAI-1 in the IVC tissues of thrombus-forming segment. (CD31, an endothelial cell marker; n = 6 mice per group; scale bar=50 μm). j ChIP-qPCR analysis of H3K18la enrichment at the PAI-1 promoter in HUVECs co-cultured with conditioned media from HK2-knockdown NB4 cells. b–h, j n = 3 biological replicates. Data were analyzed by two-tailed Student’s t test (c, e, j), two-way ANOVA with Sidak’s multiple comparisons test (g, h). Data are presented as mean ± SD. Source data are provided as a Source Data file.

Pharmacological inhibition of HK2 or MCT1 attenuated thrombosis in AML mice

To investigate the therapeutic potential of targeting HK2-mediated lactate production or MCT1-mediated lactate uptake, we utilized a MOLM13 xenograft model with four treatment groups: vehicle control, the HK2 inhibitor 2-deoxy-D-glucose (2-DG)^29,33, the MCT1 inhibitor AZD3965^26,34–36, and combination therapy (Fig. 7a). Following successful engraftment confirmed by bioluminescence imaging (Fig. 7b), thrombosis was induced via IVC stenosis. 2-DG monotherapy and combination treatment significantly reduced circulating lactate levels versus controls, whereas AZD3965 alone (AZD) showed no such effect (Fig. 7c). All treatments - monotherapies and combination - significantly reduced thrombus weight (Fig. 7d). This antithrombotic effect occurred independently of changes in peripheral blood cell counts or systemic coagulation activation as measured by TAT complexes (Supplementary Fig. 7a, b). Molecular analysis revealed decreased H3K18 lactylation and PAI-1 expression in BMMCs with 2-DG or combination treatment, but not with AZD3965 alone (Fig. 7e). IVC tissues exhibited reduced H3K18 lactylation and PAI-1 expression across all treatment groups (Fig. 7f). Systemic analyses further demonstrated consistent suppression of PAI-1 across all treatment groups, with significantly decreased plasma PAI-1 levels (Fig. 7g) and t-PA/PAI-1 complexes (Fig. 7h), along with decreased PAI-1 accumulation within thrombi (Fig. 7i). These results indicate that inhibiting leukemic lactate production (via HK2) or endothelial lactate uptake (via MCT1) disrupts the lactylation-PAI-1 axis, resulting in attenuated thrombosis in AML.

Fig. 7. Pharmacological inhibition of HK2 or MCT1 attenuated thrombosis in AML mice.

a The outline depicts the steps of the animal experiments (by figdraw.com). b Bioluminescence images were acquired after leukemia implantation in mice before IVC stenosis. c Plasma lactate was measured using a Lactate Assay Kit. d Representative images and thrombus weights from mice 48 hours after IVC stenosis surgery. e Immunofluorescence staining of H3K18la and PAI-1 in mouse bone marrow cells (scale bar = 25 μm). f Immunofluorescence staining was performed to evaluate the expression of H3K18la and PAI-1 in the IVC tissues of thrombus-forming segment. (CD31, an endothelial cell marker; scale bar = 50 μm). g, h The levels of PAI-1 and t-PA/PAI-1 complex in mouse plasma were detected using ELISA kits. i Representative immunofluorescence images of PAI-1 (yellow) expression in thrombi from mouse IVC (scale bar = 200 μm). b-i n = 6 mice per group. Data were analyzed by one-way ANOVA with Dunnett’s multiple comparisons test. Data are presented as mean ± SD. Source data are provided as a Source Data file.

Discussion

AML is frequently associated with coagulation disorders, including both thrombotic and hemorrhagic complications, which significantly contribute to morbidity and mortality. Emerging evidence highlights the interplay between genetic alterations and thrombotic risk in AML. Specific mutations (e.g., FLT3-ITD, NPM1, IDH1/2) and clonal hematopoiesis of indeterminate potential (CHIP)-related mutations (DNMT3A, TET2, ASXL1) have been epidemiologically linked to increased thrombosis incidence in clinical studies^3,20,37,38, though the underlying mechanisms remain incompletely characterized. In our small cohort, mutations (FLT3-ITD/TKD, NPM1, WT1) were detected in 85.7% of thrombotic cases, aligning with previous observations that certain molecular alterations may disproportionately contribute to thrombotic propensity. Intriguingly, Hu et al. demonstrated that FLT3-ITD drives a marked increase in aerobic glycolysis through AKT signaling-mediated upregulation of HK2, rendering leukemia cells highly glycolytic-dependent and vulnerable to pharmacological inhibition of glycolytic activity³⁹. Our current findings regarding HK2-mediated lactate production and its role in thrombosis could be one potential mechanism underlying FLT3-ITD-associated thrombosis, though clinical validation is needed. The potential association of FLT3-ITD with HK2 and thrombosis is hampered by the limited size of our sample and the lack of exhaustive molecular characterization of some samples. Further studies with a larger sample size, complete genetic characterization, and a prospective design are necessary to verify our conjecture.

Given the close interaction between tumor cells and vascular endothelial cells within the tumor microenvironment, as well as the high expression of MCT1 on endothelial cells⁴⁰, which mediates the transport of extracellular lactate into the intracellular milieu^10,29,30 we reveal that leukemic cell-endothelial cell interactions, mediated by MCT1-dependent lactate transport⁴⁰, amplify thrombotic effects. Inhibition of HK2-mediated lactate production or MCT1-mediated lactate uptake attenuates thrombosis, suggesting that lactate may promote thrombosis through epigenetic modifications. This study directly connects lactate metabolism to tumor-associated thrombosis, offering insights into AML-associated coagulation disorders. Consistent with the therapeutic potential of targeting glycolysis in AML, prior work showed that 2-DG synergizes with sorafenib to suppress FLT3/ITD-mutated leukemia by counteracting Warburg-effect dependency³⁹. Long et al. reported that histone lactylation promotes PD-L1 expression to drive immunosuppression in AML³³. Here, we identify HK2-mediated lactate as a pivotal mediator of thrombotic complications, extending the pathophysiological impact of metabolic dysregulation beyond tumor progression. The limitations of our thrombus model could not further continue to observe the effect of targeting HK2 on leukemic burden. Future studies could explore tumor-targeted HK2 inhibitors (e.g., nanosphere-encapsulated drugs) or combine HK2-knockout models with chemotherapy to investigate whether targeting lactate metabolism concurrently suppresses leukemia and thrombosis.

Compared to the general population, cancer patients exhibit a heightened risk of venous thromboembolism. Our research highlights the significance of lactate in thrombosis through histone lactylation in AML, offering insights that may extend to coagulation disorders in other diseases. Clinical studies have reported a positive correlation between plasma lactate levels and clot lysis time in patients with acute pulmonary embolism⁴¹, although systematic evidence remains limited. Gangaraju et al. conducted a cohort study on 358 adult acute leukemia patients, linking high extracellular vesicle tissue factor activity to increased hemorrhage risk and elevated PAI-1 levels to a higher risk of DVT. Tissue factor and PAI-1 have emerged as biomarkers for hemorrhage and DVT in acute leukemia, respectively²⁴. However, the molecular mechanisms driving PAI-1 elevation in leukemic cells remain poorly understood. Here, our data demonstrate that lactate activates PAI-1 transcription through H3K18 lactylation, thereby inhibiting clot lysis and increasing thrombosis risk. One limitation of our study is the lack of multivariate analysis to identify blood lactate levels as an independent risk factor for thrombosis in AML patients, which warrants further investigation in large-scale prospective cohort studies. Although our findings indicate that AML-derived lactate can be internalized by endothelial cells to enhance H3K18 lactylation and upregulate PAI-1 expression, we acknowledge that intrinsic endothelial cell glycolysis may also be altered within the leukemic microenvironment. Unfortunately, validating this possibility remains challenging due to the inherent difficulty of from patients with blood cancers. Thus, further investigation is needed to clarify endothelial cell metabolic reprogramming in this context. Our study focuses on the role of the HK2-lactate-H3K18la-PAI-1 axis in AML-associated thrombosis. Future studies could further identify additional potential targets regulated by histone lactylation, which will be a crucial direction for expanding the epigenetic regulatory network of lactate metabolism.

In summary, our study highlights lactate-driven histone lactylation as a mechanism promoting thrombosis in AML, revealing a link between metabolic reprogramming and coagulation dysfunction (Fig. 8). Targeting glycolysis in leukemic cells or reducing blood lactate levels may represent a promising strategy to mitigate thrombosis risk, potentially facilitating precise stratification and management of thrombosis in AML.

Fig. 8. Schematic model proposed in this study (by figdraw.com).

HK2-mediated lactate production regulates PAI-1 expression and thrombosis via histone lactylation.

Methods

Cell culture

The human acute myeloid leukemia (AML) cell lines NB4, THP-1, and MOLM13 were cultured in RPMI 1640 medium (Gibco, USA) supplemented with 10% fetal bovine serum (FBS; Procell, China). HL-60 cells were cultured in IMDM medium (Procell) containing 20% FBS. Fresh peripheral blood mononuclear cells (PBMCs) were isolated from AML patients and healthy donors using Ficoll-Hypaque density gradient centrifugation, and cultured in RPMI 1640 medium containing 20% FBS. Human umbilical vein endothelial cells (HUVEC) were cultured in HUVEC cell-specific medium (Procell) and used within 5 passages. For co-culture experiments, AML cells were incubated for 12–24 hours to generate conditioned medium. The supernatant was collected by centrifugation and subsequently applied to HUVEC cultures. HUVEC was obtained from Procell Life Science & Technology Co., Ltd. THP-1, HL-60 were from American Type Culture Collection; NB4 and MOLM13 were from Deutsche Sammlung von Mikroorganismen und Zellkulturen. These cells were authenticated by short tandem repeat (STR) profiling and confirmed to be mycoplasma-free prior to use in experiments.

Plasminogen activation assay

Plasminogen activation assay was performed as previously described^42–45. Cells were incubated with 200 nM plasminogen (Diapharma, USA) followed by the addition of 200 μM S-2251 (Boatman, China). The formed plasmin was cleaved to produce p-nitroaniline, and its absorbance was continuously recorded at 405 nm.

Fibrin formation and fibrinolysis assay

Fibrin formation and Fibrinolysis assay were conducted as described^43,45,46. For fibrin formation, cells were incubated with 5 mM CaCl₂ (Sangon, China), 0.8 mg/ml fibrinogen (Boatman), and 0.5 U/ml thrombin (Boatman). For fibrinolysis, cells were incubated with 5 mM CaCl₂, 0.8 mg/ml fibrinogen, 0.5 U/ml thrombin, 0.5 μM plasminogen (Diapharma), and 0.5 μg/ml t-PA (ProSpec, Israel). Absorbance at 405 nm was monitored immediately in a microplate reader.

Quantitative PCR (qPCR)

Total RNA was isolated by TRIzol (Takara, Japan), and cDNA was synthesized with PrimeScript™ RT reagent Kit (Takara). qPCR was performed using TB Green Premix Ex Taq™ II (Takara), with target mRNA levels normalized to ACTB. Primer sequences are listed in Supplementary Table 1.

Western blot

Cells were lysed using RIPA buffer, sonicated, and centrifuged at 12,000 × g for 10 minutes to obtain the supernatant. Protein concentrations were determined using a BCA assay (GlpBio, USA). Proteins were separated by SDS-PAGE, transferred to nitrocellulose membranes (Merck Millipore, Germany), and probed with antibodies anti-Pan-Kla (PTM Bio, PTM-1401RM, 1:1000), anti-H3K9la (PTM Bio, PTM-1419RM, 1:1000), anti-H3K14la (PTM Bio, PTM-1414RM, 1:1000), anti-H3K18la (PTM Bio, PTM-1406RM, 1:1000), anti-H4K5la (PTM Bio, PTM-1407, 1:1000), anti-H4K8la (PTM Bio, PTM-1415RM, 1:1000), anti-H4K12la (PTM Bio, PTM-1411RM, 1:1000), anti-H4K16la (PTM Bio, PTM-1417RM, 1:1000), anti-Histone H3 (PTM Bio, PTM-1001RM, 1:1000), anti-PAI-1 (Abmart, T56901S, 1:500), anti-HK2 (Abmart, TD6176, 1:1000), anti-MCT1 (ABclonal, A3013, 1:1000), anti-AARS1 (Proteintech, 17394-1-AP, 1:1000), anti-PKM2 (Proteintech, 15822-1-AP, 1:1000), anti-PFKP (ABclonal, A20983, 1:1000), and anti-β-actin (Proteintech, 66009-1-Ig, 1:10000). Secondary antibodies (Proteintech, SA00001-2, 1:5000; Proteintech, SA00001-1, 1:5000) were used for detection.

Chromatin immunoprecipitation (ChIP)-qPCR

ChIP was performed using the BeyoChIP™ Enzymatic ChIP Assay Kit (Beyotime, China) according to the manufacturer’s instructions^47,48. Cells were crosslinked with 1% formaldehyde for 10 minutes, and the reaction was quenched with 125 mM glycine. After collection, cells were fragmented using micrococcal nuclease and ultrasonication. The supernatant was incubated overnight at 4 °C with 4 μg anti-H3K18la or IgG antibody. Immune complexes were precipitated using Protein A/G magnetic beads, washed, and cross-links were reversed. DNA was purified using the BeyoMag™ Magnetic Bead PCR/DNA Purification Kit (Beyotime), and PAI-1 promoter fragments were quantified by qPCR using primers listed in Supplementary Table 2.

Establishment of overexpression and knockdown cell Lines

For gene overexpression or knockdown, the full-length cDNA or short hairpin RNA (shRNA) sequences were cloned into pHBLV-ZsGreen or pLVX-ZsGreen vector, respectively. The constructed plasmids were then co-transfected with the packaging plasmids pSPAX2 and pMD2G into HEK293T cells using to produce lentiviruses. The viral supernatant was collected and used to infect the target cells for 72 hours, puromycin (2 μg/ml) was added to the screening cells. For the transient knockdown of MCT1 in HUVECs, cells were transfected with MCT1-targeting small interfering RNA (siRNA) using the Hieff Trans™ siRNA/miRNA Transfection Reagent (YEASEN, China), following the manufacturer’s protocol. Oligonucleotide sequences are listed in the Supplementary Table 3.

Experimental animals

NOD/ShiLtJGpt-Prkdc^em26Cd52Il2rg^em26Cd22/Gpt (NCG) mice (both male and female, 6 weeks) were purchased from Gempharmatech (Nanjing, China). All mice were maintained under specific pathogen-free conditions at the transgenic animal facility, Institute of Translational Medicine in Nanchang University (Nanchang, China). The mice were maintained under a 12 hour light/12 hour dark cycle at a controlled temperature of 23 ± 2 °C and relative humidity of 40–50%, with 4–6 animals per cage. They were provided with a standard rodent diet (#XTI01FZ, Jiangsu Xietong Pharmaceutical Bio-engineering Co., Ltd., China). A minimum of six mice per group were used, balanced for age and body weight. For the in vivo xenograft models, NCG mice were intravenously injected with 8 × 10⁵ luciferase-labeled NB4 or MOLM13 cells to obtain AML mouse models. Humane endpoints were strictly enforced; euthanasia by cervical dislocation under deep isoflurane anesthesia was performed upon observation of systemic tumor infiltration, defined as the presence of severe lethargy, weight loss exceeding 20% of initial body weight, or systemic dissemination confirmed by bioluminescence imaging.

Thrombosis models

To induce thrombosis, mice were subjected to inferior vena cava (IVC) stenosis surgery as previously described^49–52. Mice were anesthetized with Delivector Avertin⁵³, the IVC was exposed and ligated with a 4-0 silk suture over a 30-gauge needle. The needle was then removed to allow partial flow restriction. The incision was closed, and mice were placed on a warming pad for postoperative recovery. Forty-eight hours post-surgery, mice were euthanized, and blood was collected into citrate-anticoagulated tubes. The thrombi were carefully harvested, photographed, and weighed. Subsequently, both the thrombi and surrounding IVC tissues were fixed in 4% formaldehyde for further immunofluorescence staining. Bone marrow aspirates were collected for smear preparation. For treatment groups, sodium lactate (0.5 g/kg, Sigma-Aldrich) by intraperitoneal injection for 3 days. 2-deoxy-D-glucose (2-DG; 0.5 g/kg, MCE) by intraperitoneal injection for 7 days^29,36. 50 mg/kg AZD3965 (AZD, MCE) orally twice daily for 7 days^26,35.

Immunofluorescence and immunohistochemistry

Thrombus and IVC tissues were fixed, paraffin-embedded, and stained with anti-CD31 (AIFang, AFRM0001, 1:100), anti-H3K18la (PTM Bio, PTM-1406RM, 1:100), anti-PAI-1 (AIFang, AF03419, 1:100) antibodies. Bone marrow smears were stained with anti-HK2 (Abmart, TD6176, 1:100), anti-Pan-Kla (PTM Bio, PTM-1401RM, 1:100), anti-H3K18la (PTM Bio, PTM-1406RM, 1:100), anti-PAI-1 antibodies (Abmart, T56901S, 1:50). The stained sections were scanned on a digital slide scanner and analyzed using K-Viewer analysis software.

Enzyme-linked immunosorbent assay (ELISA)

Plasma was prepared from anticoagulated whole blood of mice or patients by centrifuging at 1500 × g for 10 minutes at room temperature. The concentrations of PAI-1 (Bio-swamp, China), t-PA/PAI-1 complex (Camilo Biological, China), TAT complex (Elabscience, China) in mouse plasma were determined using commercial enzyme-linked immunosorbent assays (ELISA) according to the manufacturers’ instructions. Patients’ plasma PAI-1 levels were measured using a commercial ELISA kit (BOSTER, China).

Lactate measurement

The concentration of lactate in the patients’ plasma was measured using the Lactate Assay Kit (DiaSys, Germany) with a biochemistry analyzer (Mindray, China) according to the manufacturer’s instructions. For lactate in mouse plasma, cytoplasm or cell culture supernatants were detected using the L-Lactate Assay Kit (Solarbio, China) according to the manufacturer’s instruction.

[1,2-¹³C]-glucose metabolic flux analysis

NB4 cells with HK2 knockdown or control cells were cultured in glucose-free RPMI-1640 medium containing 2 g/L [1,2-¹³C]-glucose for 24 hours. Supernatants were collected, centrifuged, and co-cultured with HUVECs for 12 hours. Cells were washed with PBS, and metabolites were extracted using pre-cooled methanol: acetonitrile: water (2:2:1, v/v/v). After scraping, cells were ground in liquid nitrogen, mixed with chloroform, and centrifuged. Water-soluble metabolites were analyzed using a Q Exactive PLUS mass spectrometer (Thermo Scientific, USA) with hydrophilic interaction chromatography (HILIC). Data were processed using El-MAVEN and corrected for natural abundance using MATLAB.

Measurement of glycolytic rate

The glycolytic rate was determined using the Seahorse XF Glycolysis Rate Assay Kit (Agilent Technologies, USA) following the manufacturer’s instructions. Briefly, XFe 24-well cell culture microplates were pre-coated with Cell-Tak (Corning, USA). On the day of the assay, cells were resuspended in assay medium supplemented with 1 mM pyruvate, 2 mM glutamine, and 10 mM glucose, and then seeded into the pre-coated plates. Cells were sequentially treated with 0.5 μM rotenone/antimycin A (Rot/AA) and 50 mM 2-DG. The assay was conducted using the Agilent Seahorse XFe 24 Analyzer, and data were acquired and analyzed using Wave Desktop software. The glycolytic rate was calculated based on extracellular acidification rate (ECAR) measurements, and the results were graphically represented for comparative analysis.

RNA sequencing

Total RNA was extracted with TRIzol reagent (Takara), and quality was verified using the RNA Nano 6000 Assay Kit (Agilent Technologies). Following mRNA enrichment with oligo(dT) beads, cDNA libraries were prepared through fragmentation, double-stranded cDNA synthesis, end repair/adapter ligation, and PCR amplification. Sequencing was performed on Illumina platforms. Differential expression (adjusted p-values < 0.05, fold change > 4) analysis was performed using the DESeq2 R package (1.42.0). Gene Ontology (GO) enrichment analysis of differentially expressed genes was implemented by the clusterProfiler (4.8.1).

CUT&Tag

The CUT&Tag was performed using the Hyperactive Universal CUT&Tag Assay Kit for Illumina (Vazyme, China) according to the manufacturer’s instructions. Briefly, cells were bound to ConA beads and incubated sequentially with primary antibodies against H3K18la (PTM Bio, PTM-1406RM, 1:50) and species-matched secondary antibody (Vazyme, Ab207, 1:100), followed by tagmentation using pA/G-Tn5 transposase. After reaction termination with SDS, DNA fragments were captured by DNA Extract Beads Pro, washed with B&W buffer, and PCR-amplified using dual-indexed primers. Final libraries were sequenced on an Illumina NovaSeq platform. Raw reads were quality-checked by FastQC and filtered using cutadapt. Clean reads were aligned to the reference genome using bowtie2, and binding peaks were called for each sample using MACS2. Peaks with | log₂(fold change) | >0.5 and p-values < 0.05 were employed to differentially peaks.

Statistics and reproducibility

Normally distributed data were analyzed by Student’s t-test, one-way ANOVA, or two-way ANOVA, as appropriate. Non-normally distributed data were evaluated using Mann-Whitney test. Pearson correlation coefficients were used to assess the correlation between two continuous variables. The specific statistical tests applied are described in their respective figure legends. The sample sizes and experimental replicates are indicated in the figure legends. Data are presented as mean ± SD or mean ± SEM, as specified. p-values < 0.05 were considered significant Fig. 8.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Ethics statement

Peripheral blood and bone marrow samples from newly diagnosed AML patients and healthy donors were obtained from the Second Affiliated Hospital of Nanchang University with approval from the Ethics Committee for Biomedical Research, and written informed consent from all participants. Patients received no financial compensation for study participation. Since bone marrow samples from healthy donors were difficult to obtain, we selected iron-deficiency anemia patients with normal coagulation function as controls for AML bone marrow samples. Since sex were not a predetermined focus of the research, no sex-based data collection or separate analysis was performed. Clinical characteristics and laboratory parameters are detailed in Supplementary Data 1.

All animal experiments were approved by the Experimental Animal Welfare Ethics Committee of Nanchang University (Approval number: NCULAE-20250318001) and performed in compliance with the Guide for the Care and Use of Laboratory Animals.

Author contributions

X.Z.W. and Q.B. designed the study. Q.B. performed the experiments, analysis of data and wrote the manuscript. X.Z.W. revised the manuscript. Y.X.Q. and H.Z. contributed to design the methodology. BH and JL provided valuable suggestions. J.Z., H.S., X.H., and Z.H.W. helped to collect blood samples. F.M.Z. helped to conduct the bioinformatics analysis. YMX collected Bone marrow samples. L.H.Y. and K.Y.D. helped to conduct the animal experiment.

Peer review

Peer review information

Nature Communications thanks Natalia Baran and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The RNA-seq and CUT&Tag data generated in this study have been deposited in the Genome Sequence Archive under accession code HRA011045 and HRA012625. The uncropped blot figures and original data underlying Figs. 1–7 and Supplementary Figs. 1–7 generated in this study are provided in the Source Data file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-65259-0.