EXOSC4 as a novel regulator of neddylation in acute myeloid leukemia

Abstract

Acute myeloid leukaemia (AML) is a common haematological malignancy with an unsatisfactory prognosis despite recent therapeutic advances. Neddylation and ferroptosis have been implicated in the regulation of malignant cell proliferation, survival and therapy response. However, the relationship between neddylation-related dysregulation and ferroptosis-associated pathways in AML remains poorly understood. Transcriptomic data from TCGA-AML and GTEx normal controls were integrated for differential expression and functional enrichment analyses. Prognostic neddylation-related genes were identified using univariate Cox regression, LASSO regression and multivariate Cox regression analyses. The expression and regulatory association of EXOSC4 with NEDD8 and Cullin1 neddylation were validated in clinical samples and AML cell lines using qRT-PCR and Western blotting. CCK-8 assays were performed to evaluate the effects of EXOSC4 overexpression, MLN4924, erastin and Ferrostatin-1 on AML cell viability. A ceRNA regulatory network was constructed based on the key neddylation-related genes. GO and KEGG enrichment analyses showed that differentially expressed neddylation-related genes were mainly enriched in post-translational modification-related biological processes and pathways. Cox regression analyses identified EXOSC4 as a prognostic neddylation-related gene significantly associated with AML prognosis. Validation in clinical samples showed that EXOSC4 and NEDD8 were differentially expressed at both the mRNA and protein levels. In vitro experiments further showed that EXOSC4 expression was positively associated with NEDD8 protein levels and Cullin1 neddylation, suggesting that EXOSC4 may be involved in neddylation-related regulation in AML. Correlation analysis indicated a potential association between neddylation-related dysregulation and ferroptosis-associated pathways. Moreover, CCK-8 assays showed that EXOSC4 overexpression enhanced AML cell viability, whereas MLN4924 suppressed this effect and further enhanced erastin-induced reduction in cell viability, suggesting a potential role of EXOSC4 in ferroptosis-related cellular responses. In conclusion, this study reveals a potential association between neddylation-related dysregulation and ferroptosis-associated pathways in AML. EXOSC4 was identified as a candidate regulatory molecule that may connect neddylation activity with ferroptosis-related cellular responses. These findings provide new insights into the molecular mechanisms of AML and support further investigation of EXOSC4 as a potential therapeutic target.

Supplementary Information

The online version contains supplementary material available at10.1038/s41598-026-52987-6.

Introduction

Acute myeloid leukaemia is a common haematological malignancy characterised by the clonal proliferation of abnormal myeloid cells¹. Despite advances in molecular targeted therapies, immunotherapies, and chemotherapy regimens in recent years, which have improved overall survival rates for AML patients², long-term prognosis remains poor due to the disease’s complex pathogenesis, high recurrence rate, and frequent development of drug resistance³. The abnormal proliferation of AML cells remains a major challenge impeding patient survival⁴. Against this backdrop, in-depth investigation into the mechanisms of AML cell proliferation is crucial for developing more precise and effective therapeutic strategies.

Neddylation represents a significant post-translational modification of proteins, primarily achieved by activating and covalently attaching NEDD8 (Neural precursor cell expressed developmentally downregulated protein 8) to substrate proteins⁵. As a pivotal form of protein post-translational modification, neddylation participates in numerous cancer-related biological processes, including cell cycle regulation, DNA repair, and protein degradation^6,7. NEDD8 is a ubiquitin-like small protein exhibiting high structural homology with ubiquitin yet possessing unique functional properties. Its primary targets are Cullin family proteins, which constitute the core components of Cullin-RING ligase complexes (CRLs)⁸. In hepatocellular carcinoma, NEDD8 protein significantly enhances proliferation by activating the Wnt/β-catenin pathway. NEDD8, a key protein in neddylation modification, is closely associated with abnormal proliferation and poor prognosis in AML⁹. In AML, the NEDD8 protein promotes AML cell apoptosis by regulating the expression of genes associated with the NF-κB signalling pathway¹⁰. Consequently, neddylation has become a significant focus in AML research. Notably, the interaction of the NEDD8 protein with multiple tumour-associated proteins promotes AML cell proliferation and survival¹¹. Although several NEDD8-targeting drugs have entered clinical trials, their efficacy and mechanisms remain incompletely understood¹². As a potential therapeutic target in AML treatment, the precise role of neddylation in the abnormal proliferation of AML cells warrants further investigation to provide novel insights and strategies for future personalised therapeutic approaches.

Ferroptosis represents a novel form of programmed cell death distinct from traditional mechanisms such as apoptosis and necrosis. Its hallmark features include the accumulation of intracellular lipid peroxides and elevated reactive oxygen species (ROS) levels¹³. Research indicates that histone deacetylase inhibitors promote ferroptosis in AML by regulating iron metabolism¹⁴. The prolactin-like inhibitor MLN4924 promotes ferroptosis in AML cells by suppressing expression of the SLC7A11/GPX4 axis¹⁵. Consequently, the regulatory mechanism linking neddylation modification and ferroptosis in AML may represent a potential therapeutic target for AML.

The EXOSC4 gene constitutes a crucial component of the RNA exosome complex, participating in multiple biological processes including RNA degradation, gene expression regulation, and maintenance of cellular homeostasis¹⁶. In colorectal cancer, EXOSC4 promotes tumour cell proliferation via the Wnt/β-catenin signalling pathway¹⁷. Within breast cancer, EXOSC4 expression is closely associated with tumour cell proliferation¹⁸. However, the functional characteristics of EXOSC4 in AML and its potential molecular mechanisms remain unclear.

Therefore, this study aimed to investigate the potential relationship among EXOSC4, NEDD8-mediated neddylation and ferroptosis-related pathways in AML by integrating public database analyses, clinical sample validation and in vitro experiments. We further explored the prognostic significance of EXOSC4 and its potential association with NEDD8 expression, Cullin1 neddylation and ferroptosis-related cellular responses, with the goal of providing new insights into the molecular mechanisms underlying AML progression and identifying candidate targets for future mechanistic and therapeutic investigation.

Materials and methods

Data collection and differential expression analysis

RNA sequencing data and clinical information from AML patients and healthy controls were sourced from the Genome-Tissue Expression (GTEx) database. The training cohort comprised a combined cohort of samples from the Cancer Genome Atlas (TCGA), the Therapeutic Application Research for Generating Effective Therapies (TARGET) project, and GTEx. As the datasets were standardised, no additional preprocessing was required. Differential expression analysis between AML and normal samples was performed using the DESeq2 package (version 1.42.1) in R. Genes with low expression (total counts ≤ 1) were removed. Significant DEGs were defined as those with padj < 0.05 and |log₂FC| > 1. Neddylation-related genes were obtained from the GeneCards database (filtered for relevance scores > 1) and intersected with differentially expressed genes (DEGs), defined as “Neddylation-related DEGs” for subsequent analysis.

Functional enrichment analysis

Employing the ClusterProfiler package in R for GO and KEGG pathway enrichment analysis to explore the biological functions of DEGs. GO entries are categorised into biological processes (BP), cellular components (CC), and molecular functions (MF).

Key gene screening

The key genes were obtained by using a multi-step approach. First, genes significantly associated with overall survival were identified by ​univariate Cox proportional-hazards regression (HR = 95%, p < 0.05). Subsequently, to prevent overfitting and select the most informative features, ​the least absolute shrinkage and selection operator (LASSO) Cox regression​ was applied to these candidate genes (HR = 95%, p < 0.05). Finally, the variables selected by LASSO were incorporated into a ​multivariate Cox regression​ analysis to filter key gengs in AML neddylation modifications.

Survival analysis

Survival analysis was performed using the survival package (version 3.8.3) in R to evaluate overall survival (OS) among patients with different expression levels.

Correlation analysis of pro-apoptotic and ferroptosis-related genes

Ferroptosis-associated genes were obtained from the FerrDb database (http://www.zhounan.org/ferrdb/). Pearson correlation coefficients were calculated between AML key neddylation genes and ferroptosis-associated genes. Genes meeting Pearson correlation coefficient > 0.7 and P < 0.05 were selected as significantly correlated genes, with results visualised using the pheatmap package (version 3.5.1) .

Patients and clinical samples

This study included 72 adult patients diagnosed with acute myeloid leukaemia and 30 healthy donors who attended Guizhou Medical University Affiliated Hospital between September 2023 and September 2025. Among these, 36 were newly diagnosed and 36 were relapsed cases. All cases were diagnosed by Guizhou Medical University Affiliated hospital’s Haematology Department through comprehensive evaluation of morphology, immunophenotyping, cytogenetics, and molecular biology (MICM). In this study, newly diagnosed acute myeloid leukemia patients were required to meet the following inclusion criteria: a confirmed diagnosis of AML without prior treatment, and no history of other types of leukemia or related hematological disorders. Patients were aged between 18 and 65 years, with adequate organ function (liver, kidney, heart, etc.) and hematological parameters meeting the study requirements. Relapsed patients were required to meet the following inclusion criteria: a confirmed diagnosis of AML with relapse after achieving complete remission (CR), defined by the reappearance of blast cells in the bone marrow or peripheral blood. Patients were aged between 18 and 65 years, with adequate organ function (liver, kidney, heart, etc.) and hematological parameters meeting the study requirements. Relevant clinical data are presented in Table 1. This study adheres to the principles of the Declaration of Helsinki and was approved by the Basic Research Ethics Committee of Guizhou Medical University Affiliated Hospital (Approval No.: 2023-253). Written informed consent was obtained from all subjects for any potentially identifiable images or data included herein.

Table 1.

Correlation between EXOSC4 expression and clinical parameters of 72 AML patients.

Varibies	case	EXOSC4 expession		P value
		High	Low
Sex				0.3411
Female	31	18	13
Male	41	18	23
Age				0.0739
≥ 60	30	14	5
< 60	36	9	2
WBC(*109/L)				0.3109
≥ 10	52	25	27
< 10	20	11	9
Blasts in bone marrow (%)				0.2270
≥ 50	53	26	27
< 50	19	10	9
Cytogenetics				0.9469
Favorable	13	6	7
Intermediate	28	14	14
Unfavorable	31	16	15
PLT(*109/L)				0.1203
≥50	28	15	13
< 50	44	21	23
CR				0.0328
Yes	32	11	21
No	40	25	15

Isolation of bone marrow mononuclear cells

Under strict aseptic conditions, approximately 5 mL of bone marrow was collected via bone marrow puncture from AML patients and healthy donors. This was placed into EDTA-containing anticoagulant tubes (transported at room temperature and processed within 2 h). The sample was diluted 1:1 with an equal volume of PBS (or 0.9% NaCl) and mixed thoroughly. Add an equal volume of human peripheral blood/bone marrow mononuclear cell isolation solution (Solarbio Technologies, Beijing) to the bottom of a 15 mL conical centrifuge tube. Slowly aspirate the diluted bone marrow suspension along the tube wall, then centrifuge at 2000 rpm at room temperature for 15 min. After centrifugation, aspirate the middle cell layer and transfer to a fresh tube. Centrifuge at 1500 rpm for 5 min, discard the supernatant, and wash the remaining pellet three times with physiological saline. Obtain the pellet and wash it three times with physiological saline.

Quantitative real-time fluorescent PCR

Total RNA extraction was performed using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s protocol. cDNA synthesis employed a reverse transcription kit (MedChemExpressMCE, USA). Real-time PCR was conducted using the SYBR Green PCR Master Mix (MedChemExpressMCE, USA) and a PRISM 7500 real-time PCR detection system (Thermo Fisher Scientific, USA). Relative expression levels of target genes were calculated using β-actin as a reference gene by comparing cycle threshold (CT) values (2 − ΔΔCT). qRT-PCR primers (Generay Biotech Co. Ltd, Shanghai, China) are listed in Table 2.

Table 2.

Primer sequences used for RT-PCR.

Primer name	Sequence (5ʹ-3ʹ)
SIRT6-F	CCCAAGTTCGACACCACCTTTG
SIRT6-R	CTGACCAGGAAGCGGAGGAG
SLC25A28-F	CAGATCCCCTCCACAGCCATC
SLC25A28-R	TTGCCAGCCCTCCACTCTTC
OTUB1-F	TGAGCAGGTGGAGAAGCAGAC
OTUB1-R	GGTGAGCAGCCGCAGGTAG
STK11-F	CGACTCCACCGAGGTCATCTAC
STK11-R	ACCTTGCCGTAAGAGCCTTCC
CISD3-F	GCGGCGGGACATCTCCTC
CISD3-R	TGCCACCAGCTCCACCTTG
GPX4-F	CTGCTCTTCCAGAGGTCCTG
GPX4-R	GAGGTGTCCACCAGAGAAGC
EXOSC4-F	GGCACTGGCTGTGGTCTACG
EXOSCA-R	CTGTGCTGAAGGTCGCTGAAC
NEDD8-F	CGCTGACCGGAAAGGAGATT
NEDD8-R	CAGAGCCAACACCAGGTGAA
β-Actin-F	CTCGCCTTTGCCGATCC
β-Actin-R	ATCCTTCTGACCCATGCCC

Cell culture

Human myeloid cell lines U937 and THP-1 originate from this research centre—the Guizhou Provincial Haematopoietic Stem Cell Transplantation Centre Laboratory. All cell lines have undergone mycoplasma contamination testing and were genotyped using short tandem repeat (STR) analysis to verify their origin and purity. Following thawing, cells were cultured at 37 °C, 5% CO₂, and saturated humidity in RPMI-1640 medium supplemented with 10% foetal bovine serum, 100 U/mL penicillin, and 100 µg/mL streptomycin. Both cell lines were maintained in suspension at a routine seeding density of 2–5 × 10⁵ cells/mL. Passaging was performed at a density approaching 1 × 10⁶ cells/mL using fresh complete medium diluted 1:3–1:5, maintaining viability > 90%. Cells were expanded at low passage numbers, with all experiments conducted on cells with < 20 passages.

Western blot

For Western blot analysis, primary antibodies against EXOSC4 and NEDD8 (Kanghexin Biotechnology Co., Ltd.) were diluted at 1:500, whereas the primary antibody against Cullin1 (Kanghexin Biotechnology Co., Ltd.) was diluted at 1:2000. The β-actin primary antibody (Wuhan Sanjiang Company, China) was diluted at 1:3,000. The HRP-labeled secondary antibody (Proteintech, China) was diluted to 1:10,000. Protein lysates were extracted from cells using RIPA lysis buffer (Solarbio Science and Technology Co.) supplemented with 1 mM PMSF. The mixture was vigorously vortexed and incubated on ice for 30 min. Subsequently, it was centrifuged at 12,000 rpm for 15 min at 4 °C to collect the supernatant. Protein concentration was determined using a BCA protein assay kit (Pierce, USA). Proteins were mixed with loading buffer at a 1:4 ratio and boiled at 100 °C for 10 min. Then, 40 µg of protein per sample was loaded onto a 10% SDS-PAGE gel. Electrophoresis was initially performed at a constant voltage of 80 V, and then switched to 120 V after the samples entered the separating gel. Following electrophoresis, the separated proteins were transferred onto a PVDF membrane at 250 mA for 1 h. The membrane was blocked with 5% skim milk in PBS on a shaker at room temperature for 2 h, followed by washing. The primary antibody was then added and incubated with the membrane at 4 °C for over 8 h. After washing, the secondary antibody was added and incubated at room temperature for 45 min. All protein bands were visualized using an enhanced chemiluminescence kit (Qihai Biotech, Shanghai, China). β-actin was used as an internal control. For the MLN4924 treatment experiments, THP-1 and U937 cells with stable EXOSC4 overexpression were treated with 300 nM MLN4924 for 24 h, while the corresponding vehicle-treated cells served as controls. After treatment, total cellular proteins were extracted and subjected to Western blot analysis to detect the expression levels of EXOSC4, Cullin1-NEDD8, and Cullin1.

Cell transfection and lentiviral infection

Human EXOSC4 overexpression lentiviral particles (LV-EXOSC4) and knockdown lentiviral particles (sh-EXOSC4) were procured from GeneChem (Shanghai, China) and used to transduce THP-1 and U937 cells according to the manufacturer’s instructions; cells treated with empty vector (EV) served as controls. Cells were subsequently expanded and maintained for 5 days in RPMI-1640 containing 10% FBS. THP-1 and U937 cells were then selected with 1.5 µg/mL puromycin to establish stable cell lines expressing LV-EXOSC4 or sh-EXOSC4.

CCK-8 cell viability assay

Cell viability was assessed using the Cell Counting Kit-8 assay. THP-1 and U937 cells in the logarithmic growth phase were collected by centrifugation, resuspended in complete culture medium, and adjusted to an appropriate cell density. Cells were then seeded into 96-well plates at 5 × 10³ cells/well in 100 µL medium. Each group was prepared in triplicate, and wells containing medium without cells were used as blank controls. After seeding, cells were treated according to the experimental design. The groups included the EV control group, LV-EXOSC4 group, EV + MLN4924 group, and LV-EXOSC4 + MLN4924 group. To further evaluate ferroptosis-related changes in cell viability, LV-EXOSC4 cells were treated with erastin, erastin plus Ferrostatin-1 (Ferr-1), or erastin plus MLN4924. The final concentrations of erastin, Ferr-1, and MLN4924 were 1 µM, 1 µM, and 100 nM, respectively, and the treatment duration was 48 h. All drugs were dissolved in the corresponding solvents, and the final solvent concentration was kept consistent among all treatment groups. After treatment, 10 µL of CCK-8 reagent was added to each well, followed by incubation at 37 °C in a 5% CO₂ incubator for 1–2 h protected from light. The absorbance at 450 nm was then measured using a microplate reader. After subtracting the background absorbance of blank wells, relative cell viability was calculated by normalization to the corresponding control group. All experiments were independently repeated at least three times, and data are presented as the mean ± standard deviation.

Image processing and statistical analysis

Statistical analysis was performed using GraphPad Prism 10 (GraphPad Software, San Diego, CA, USA). The distribution characteristics of clinical data were assessed for normality using the Shapiro–Wilk test. Data from three independent replicates are presented as mean ± standard deviation (mean ± SD). Comparisons between two groups were performed using a two-tailed, unpaired Student’s t-test; comparisons involving three or more groups were analysed using one-way analysis of variance (ANOVA). Survival data were plotted using the Kaplan–Meier method, with differences assessed by the log-rank test. P values < 0.05 were considered statistically significant.

Results

Screening of genes differentially expressed in AML versus normal controls associated with neddylation

Through RNA sequencing analysis of AML and normal control samples from the combined cohort of GTEx, TCGA, and TARGET databases, we identified 8870 differentially expressed genes. Among these, 3760 genes were highly expressed in the AML group, while 5110 genes were downregulated (Fig. 1A). Subsequently, we further screened 417 genes associated with neddylation using the GeneCards database. Intersecting these genes with the AML DEGs yielded 135 AML neddyaltion-related differentially expressed genes (Fig. 1B). Illustrates the expression differences of these neddyaltion-related genes between AML patients and normal samples( Fig. 1C).

Fig. 1.

(A) Differentially expressed genes in AML from the combined TCGA and GTEx cohorts. (B) Venn diagram illustrating the intersection of neddylation-related genes and AML differentially expressed genes, identifying AML neddylation-related genes. (C) Heatmap displaying neddylation-related differentially expressed genes in AML.

Enrichment analysis revealed the significant biological roles of these neddylation modification-associated differentially expressed genes in AML

To further explore the potential biological functions of the differentially expressed genes in AML, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed. The GO enrichment results of the differentially expressed genes indicated that the significantly enriched biological process (BP) terms were mainly associated with pattern specification process, regulation of cell-cell adhesion, leukocyte migration, regionalization, and positive regulation of cell adhesion. In the cellular component (CC) category, these genes were mainly enriched in collagen-containing extracellular matrix, external side of plasma membrane, apical plasma membrane, specific granule, and synaptic membrane. In the molecular function (MF) category, the enriched terms were mainly related to immune receptor activity, DNA-binding transcription activator activity, cytokine activity, glycosaminoglycan binding, and ion channel activity (Fig. 2A). KEGG pathway enrichment analysis further demonstrated that these differentially expressed genes were mainly involved in cadherin signaling, cytokine-cytokine receptor interaction, cornified envelope formation, neuroactive ligand-receptor interaction, chemokine signaling pathway, and calcium signaling pathway, suggesting that these genes may participate in cell adhesion, immune regulation, and signal transduction (Fig. 2B). Subsequently, GO and KEGG enrichment analyses were further conducted for the neddylation-related candidate genes. The GO results showed that these genes were significantly enriched in BP terms associated with proteasome-mediated ubiquitin-dependent protein catabolic process, SCF-dependent proteasomal ubiquitin-dependent protein catabolic process, protein neddylation, and regulation of post-translational protein modification. In the CC category, the enriched terms were mainly concentrated in cullin-RING ubiquitin ligase complex, ubiquitin ligase complex, SCF ubiquitin ligase complex, and proteasome complex. In the MF category, these genes were primarily enriched in ubiquitin-like protein transferase activity, ubiquitin-protein transferase activity, ubiquitin protein ligase binding, and ubiquitin-like protein binding(Fig. 2C). Consistently, KEGG analysis revealed that the neddylation-related candidate genes were predominantly enriched in ubiquitin mediated proteolysis, as well as several tumor- and signaling-related pathways, including FoxO signaling pathway, chronic myeloid leukemia, prostate cancer, and ErbB signaling pathway(Fig. 2D). These findings suggest that neddylation-related genes may play important roles in AML progression through regulation of the ubiquitin-proteasome system and multiple cancer-associated signaling pathways.

Fig. 2.

(A) GO analysis of differentially expressed genes in the TCGA-GTEx combined dataset. (B) KEGG^37–39 analysis of differentially expressed genes in the TCGA-GTEx combined dataset. (C) GO analysis of neddylation-associated differentially expressed genes. (D) KEGG analysis of neddylation-associated differentially expressed genes. In the bar plots, the x-axis represents the gene count, and the color gradient indicates the p-value. In the bubble plots, the x-axis represents -log10(p-value), bubble size indicates the GeneRatio, and bubble color represents the p-value.

Identification of key genes for AML ubiquitin-like modification via univariate Cox-Lasso and multivariate Cox analyses

Based on the 135 neddylation-related DEGs identified through the above screening, we first identified 67 candidate genes significantly associated with AML prognosis(HR = 95%, p < 0.05) via univariate Cox regression analysis (Fig. 3A). To avoid overfitting, LASSO Cox regression was further applied to select 10 genes with stronger prognostic associations(HR = 95%, p < 0.05) (Fig. 3B, C). Ultimately, through multivariate Cox regression analysis, we identified four key genes—IGF2BP3, EXOSC4, SOCS2, and KLHL13—as key genes for AML neddylation modification (HR = 95%, p < 0.05) (Fig. 3D).

Fig. 3.

(A) Forest plot displaying hazard ratios (HR), 95% confidence intervals (CI), and P-values for differentially expressed genes identified via univariate Cox regression analysis. (B) Ten gene expression signatures selected using the LASSO Cox model based on the aforementioned univariate Cox analysis results. (C) Cross-validation results for parameter selection in the LASSO model. (D) Forest plot displaying hazard ratios (HR), 95% confidence intervals (CI), and P-values for differentially expressed genes identified via multivariate Cox regression analysis.

The efficacy of these key genes in AML neddylation modification was validated via AUC-ROC analysis, alongside survival analysis of the key genes

To evaluate the prognostic and diagnostic value of the key neddylation-related genes, Kaplan–Meier survival and ROC curve analyses were performed. Survival analysis revealed that high expression of IGF2BP3 (HR = 1.74, p = 0.012; Fig. 4A), EXOSC4 (HR = 2.05, p = 0.001; Fig. 4B), and SOCS2 (HR = 1.68, p = 0.016; Fig. 4C) was significantly associated with worse overall survival in AML patients. In contrast, KLHL13 expression was not significantly correlated with overall survival (HR = 0.95, p = 0.819; Fig. 4D). ROC analysis showed that IGF2BP3, EXOSC4, SOCS2, and KLHL13 had AUC values of 0.645, 0.816, 0.744, and 0.950, respectively (Fig. 4E). Among them, KLHL13 exhibited the best diagnostic performance, whereas EXOSC4 and SOCS2 showed moderate diagnostic value. These results indicate that EXOSC4, IGF2BP3, and SOCS2 may have prognostic significance, while KLHL13 may serve as a robust diagnostic biomarker in AML.

Fig. 4.

Kaplan-Meier survival curves for the aforementioned key genes IGF2BP3 (A), EXOSC4 (B), SOCS2 (C), and KLHL13 (D) in the TCGA-GTEx cohort. (E) Area under the ROC curve (AUC) values for IGF2BP3, EXOSC4, SOCS2, and KLHL3.

Identification of ferroptosis-related genes associated with key neddylation-related genes in AML through correlation analysis

To explore the relationship between neddylation modification and ferroptosis in AML, we retrieved ferroptosis-related genes from the FerrDb database and calculated their Pearson correlation coefficients with genes associated with neddylation modification (Figs. 5 A–B). Results revealed significant positive correlations between EXOSC4 and multiple core ferroptosis regulators (r > 0.7, p < 0.05) (Fig. 5C). Notably, the strong correlations between EXOSC4 and SLC25A28 (mitochondrial iron transporter), STK11 (energy stress kinase), and GPX4 (key ferroptosis antioxidant enzyme) suggest that EXOSC4 may participate in ferroptosis regulation in AML by neddylation modification.

Fig. 5.

(A) Correlation analysis between AML neddylation key genes and ferroptosis-associated genes. (B) Lollipop plot illustrating the correlation analysis between EXOSC4 and ferroptosis-associated genes. (C) Scatter plot illustrating the correlation between EXOSC4 and the top six ferroptosis-associated genes: SIRT6, SLC25A28, OTUB1, CISD3, STK11, and GPX4. (ns p ≥ 0.05; *: p < 0.05; **: p < 0.01; ***: p < 0.001)

Differential expression of ferroptosis-related genes between AML patients and healthy donors

We compared mRNA levels of the aforementioned ferroptosis genes associated with EXOSC4 via qRT-PCR analysis in bone marrow blood samples from newly diagnosed AML patients (AML n = 30) and healthy donors (Health Donor n = 30), with significance denoted by p-values. Notably, SIRT6 expression levels showed no significant difference between the AML patient group and healthy donors (Fig. 6A). Conversely, SLC25A28 expression was 16.21-fold higher in the AML cohort than in healthy donors (Fig. 6B). OTUB1 expression in the AML cohort was 3.26-fold higher than in healthy donors (Fig. 6C). CISD3 expression in healthy donors was 4.88-fold higher than in the AML cohort (Fig. 6D). STK11 expression in the healthy donor group was 14.13-fold higher than in the AML group (Fig. 6E). GPX4 expression in the AML group was 10.06-fold higher than in the healthy donor group. It can be observed that, apart from SIRT6 showing no significant difference between the healthy donor and AML groups, other ferroptosis genes closely associated with EXOSC4 exhibited significant differences between the healthy donor and AML groups.

Fig. 6.

qRT-PCR detection of mRNA expression levels for SIRT6 (A), SLC25A28 (B), OTUB1 (C), CISD3 (D), STK11 (E), and GPX4 (F) in bone marrow blood samples from AML patients and healthy donors. (ns p ≥ 0.05; *: p < 0.05; **: p < 0.01; ***: p < 0.001)

EXOSC4 participates in neddylation modification in AML by regulating NEDD8 expression

To further investigate the clinical relevance of EXOSC4 and its association with NEDD8, a key protein involved in neddylation, the mRNA and protein expression levels of EXOSC4 and NEDD8 were examined in bone marrow samples from newly diagnosed AML patients (ND, n = 36) and relapsed AML patients (R, n = 36). qRT-PCR analysis showed that EXOSC4 mRNA expression was significantly increased in relapsed AML patients and was approximately 3.22-fold higher than that in newly diagnosed patients (Fig. 7A). Similarly, NEDD8 mRNA expression was also significantly elevated in the relapsed group, showing an approximately 1.67-fold increase compared with the newly diagnosed group (Fig. 7B). To further validate these findings at the protein level, Western blot analysis was performed. Representative blot images showed that both EXOSC4 and NEDD8 protein expression levels were markedly higher in relapsed AML samples than in newly diagnosed AML samples (Fig. 7C). Quantitative analysis further confirmed that the protein levels of EXOSC4 and NEDD8 were significantly increased in the relapsed group (Fig. 7D). Collectively, these results indicate that EXOSC4 is positively associated with NEDD8 expression in AML clinical samples, suggesting that EXOSC4 may be involved in the regulation of neddylation during AML progression. To further explore the mechanism by which EXOSC4 regulates neddylation in AML, lentiviral vectors were used to establish stable EXOSC4 overexpression and knockdown models in THP-1 and U937 cells. qRT-PCR analysis demonstrated that EXOSC4 was efficiently overexpressed in the LV-EXOSC4 group and markedly reduced in the sh-EXOSC4 groups in both cell lines (Fig. 8A, B). In THP-1 cells, EXOSC4 expression in the LV-EXOSC4 group was 1.74-fold higher than that in the control group, whereas the three sh-EXOSC4 constructs all significantly reduced EXOSC4 expression. In U937 cells, EXOSC4 expression in the LV-EXOSC4 group was 1.98-fold higher than that in the control group, while the sh-EXOSC4 constructs also markedly decreased EXOSC4 expression. Based on the knockdown efficiency, sh-EXOSC4-1 was selected for subsequent experiments. Western blot analysis further confirmed the regulatory efficiency of EXOSC4 manipulation at the protein level in both THP-1 and U937 cells (Fig. 8C, D). In parallel, NEDD8 protein expression increased significantly following EXOSC4 overexpression, but decreased markedly after EXOSC4 knockdown (Fig. 8C, E). Specifically, in THP-1 cells, NEDD8 protein expression in the LV-EXOSC4 group was 1.60-fold higher than that in the control group, whereas EXOSC4 knockdown reduced NEDD8 expression by 2.23-fold. In U937 cells, NEDD8 protein expression in the LV-EXOSC4 group was 2.94-fold higher than that in the control group, whereas EXOSC4 knockdown caused an 11.76-fold reduction in NEDD8 expression. These results suggest that EXOSC4 positively regulates NEDD8 expression in AML cells. To further determine whether EXOSC4 affects the neddylation pathway, the neddylation inhibitor MLN4924 was applied to EXOSC4-overexpressing THP-1 and U937 cells. Western blot analysis showed that EXOSC4 overexpression increased the levels of both EXOSC4 and Cullin1-NEDD8, whereas treatment with MLN4924 significantly reduced Cullin1-NEDD8 expression in both THP-1 and U937 cells (Fig. 8F, H). In addition, MLN4924 treatment had no significant effect on EXOSC4 protein expression. (Fig. 8F, G). These results showed that EXOSC4 overexpression increased Cullin1-NEDD8 levels, whereas MLN4924 treatment significantly reduced Cullin1-NEDD8 expression without significantly affecting EXOSC4 protein expression.

Fig. 7.

qRT-PCR detection of EXOSC4 mRNA and NEDD8 mRNA expression levels in bone marrow samples from newly diagnosed (ND) and relapsed (R) acute myeloid leukaemia patients. mRNA expression levels of EXOSC4 and NEDD8 in newly diagnosed (ND) and relapsed (R) patients (A, B). Western blot analysis of EXOSC4 and NEDD8 protein expression in AML newly diagnosed and relapsed patients (C, D). (ns p ≥ 0.05; *: p < 0.05; **: p < 0.01; ***: p < 0.001)

Fig. 8.

(A, B) qRT-PCR analysis of EXOSC4 expression in THP-1 and U937 cells after EXOSC4 overexpression or knockdown. LV indicates EXOSC4-overexpressing cells, EV1 and EV2 indicate corresponding empty vector control cells, and sh-1, sh-2 and sh-3 indicate EXOSC4 knockdown cells generated using different shRNAs. (C) Representative Western blot images showing the protein levels of NEDD8 and EXOSC4 in THP-1 and U937 cells after EXOSC4 overexpression or knockdown. β-Actin was used as the loading control. (D, E) Quantitative analysis of EXOSC4 and NEDD8 protein expression levels based on Western blot results. Protein levels were normalized to β-Actin. (F) Representative Western blot images showing the expression levels of EXOSC4, Cullin1-NEDD8 and total Cullin1 in THP-1 and U937 cells with or without MLN4924 treatment. GAPDH was used as the loading control. (G, H) Quantitative analysis of EXOSC4 and Cullin1-NEDD8 protein levels following MLN4924 treatment. Protein levels were normalized to GAPDH. Data are presented as the mean ± SD from three independent experiments. (ns p ≥ 0.05; *: p < 0.05; **: p < 0.01; ***: p < 0.001)

MLN4924 restored ferroptosis sensitivity in EXOSC4-overexpressing AML cells

To investigate whether EXOSC4 affects ferroptosis sensitivity in AML cells, CCK-8 assays were performed to evaluate cell viability in THP-1 and U937 cells following treatment with erastin, Ferrostatin-1(Ferr-1)and MLN4924. First, we examined the effect of EXOSC4 overexpression on AML cell viability. Compared with the EV control group, LV-EXOSC4 cells showed significantly increased relative cell viability in both THP-1 and U937 cells. This result indicated that EXOSC4 overexpression enhanced the viability of AML cells. However, treatment with MLN4924 markedly reduced cell viability in both EV and LV-EXOSC4 cells. Notably, the viability of LV-EXOSC4 cells was significantly decreased after MLN4924 treatment, suggesting that MLN4924 could suppress the viability-promoting effect induced by EXOSC4 overexpression (Fig. 9A). Next, to further determine whether EXOSC4 was involved in ferroptosis-related regulation, EXOSC4-overexpressing AML cells were treated with the ferroptosis inducer erastin. The results showed that erastin significantly reduced the viability of LV-EXOSC4 THP-1 and U937 cells compared with the untreated LV-EXOSC4 group. In contrast, combined treatment with the ferroptosis inhibitor Ferr-1 partially restored erastin-induced reduction in cell viability, indicating that the decrease in cell viability was associated with ferroptosis. These findings suggested that EXOSC4-overexpressing AML cells still responded to ferroptosis induction, whereas inhibition of ferroptosis could rescue cell viability to a certain extent (Fig. 9B, C). Importantly, compared with erastin treatment alone, combined treatment with erastin and MLN4924 further decreased cell viability in both THP-1 and U937 cells. This result indicated that MLN4924 enhanced erastin-induced ferroptosis-related cytotoxicity in EXOSC4-overexpressing AML cells. Taken together, these data suggest that EXOSC4 overexpression promotes AML cell viability and may attenuate ferroptosis sensitivity, whereas MLN4924 counteracts the protective effect of EXOSC4 and restores the sensitivity of AML cells to erastin-induced ferroptosis (Fig. 9B, C).

Fig. 9.

(A) CCK-8 assay showing the relative cell viability of THP-1 and U937 cells after EXOSC4 overexpression and/or MLN4924 treatment. EXOSC4 overexpression increased cell viability, whereas MLN4924 markedly reduced cell viability in both EV and LV-EXOSC4 cells. (B, C) CCK-8 assay showing ferroptosis-related changes in cell viability in EXOSC4-overexpressing THP-1 (B) and U937 (C) cells. Erastin treatment reduced cell viability, whereas Ferrostatin-1 (Ferr-1) partially rescued erastin-induced cell viability inhibition. In contrast, combined treatment with erastin and MLN4924 further decreased cell viability, suggesting that MLN4924 may enhance erastin-induced ferroptosis-related cytotoxicity. Data are presented as the mean ± SD from three independent experiments. (ns p ≥ 0.05; *: p < 0.05; **: p < 0.01; ***: p < 0.001)

Construction and functional analysis of ceRNA networks

Competitive endogenous RNAs (ceRNAs) influence gene expression by regulating microRNA (miRNA) expression, thereby playing a role in the pathogenesis and progression of acute myeloid leukaemia¹⁹. We employed the ENCORI platform (https://rnasysu.com/encori/index.php) to construct ceRNA networks for key genes. Through the miRNA-mRNA plugin, we identified miRNAs interacting with key genes. The mRNA–miRNA interaction network revealed that key mRNAs such as EXOSC4, IGF2BP3, and SOCS2 may be regulated by multiple miRNAs (e.g., hsa-miR-216b-5p and hsa-miR-2467-3p) (Fig. 10A). Further, via the miRNA-lncRNA plugin, we identified multiple lncRNAs simultaneously interacting with both hsa-miR-216b-5p and hsa-miR-2467-3p. These lncRNAs (e.g., NEAT1, KCNQ1OT1, LINC00961) may participate in ceRNA regulation by binding to the aforementioned miRNAs (Fig. 10B). Finally, we visualised the ceRNA network using a Sankey diagram (Fig. 10C).

Fig. 10.

(A) AML neddylation-modified miRNA-miRNA network diagram. (B) LncRNA-miRNA network. (C) LncRNA-miRNA-miRNA network displayed using a Sankey diagram.

Discussion

Although recent years have seen some progress in studying neddylation modification in AML tumour proliferation²⁰, studies employing bioinformatics analysis using public databases remain to be further supplemented. In this study, the GTEx-TCGA-Target data cohort in the GTEx database was analyzed using a univariate Cox-Lasso Cox regression analysis-multivariate Cox bioinformatics analysis method to identify key genes involved in AML neddylation modification: IGF2BP3, EXOSC4, SOCS2, and KLHL13. By validating the efficacy of these genes and examining their relationship with AML patient survival, it was observed that EXOSC4 exhibits higher efficacy,, and patients in the high-expression group of EXOSC4 demonstrated significantly reduced survival compared to those in the low-expression group.

Ferroptosis is an iron-dependent form of programmed cell death characterised by the accumulation of lipid peroxides and iron ion-mediated oxidative stress²¹. In acute myeloid leukaemia, ferroptosis is closely associated with the abnormal proliferation of AML cells²². To investigate the relationship between neddylation modification and ferroptosis in AML, we analysed correlations between the aforementioned key neddylation genes and AML ferroptosis genes. Genes exhibiting correlations > 0.7 and P values ≤ 0.05 were selected as ferroptosis genes associated with AML neddylation modification. In the present study, EXOSC4 was identified as a key neddylation-related gene in AML and was found to be strongly correlated with six ferroptosis-associated genes, namely SIRT6, SLC25A28, OTUB1, CISD3, STK11, and GPX4. Among these genes, several have established mechanistic links to ferroptosis. SLC25A28 has been reported to promote ferroptosis by enhancing lipid peroxidation²³, whereas CISD3 participates in mitochondrial iron-sulfur cluster transfer and the maintenance of mitochondrial iron/ROS homeostasis, and its inhibition markedly sensitizes cells to cystine deprivation-induced ferroptosis²⁴. In addition, STK11 status has been linked to ferroptosis sensitivity through regulation of monounsaturated fatty acid metabolism²⁵, while the role of SIRT6 appears to be context-dependent²⁶, suggesting that the ferroptosis-related network associated with EXOSC4 may be biologically complex in AML. Notably, GPX4 and OTUB1 are particularly relevant in the context of post-translational regulation. GPX4 is a core suppressor of ferroptosis, and growing evidence indicates that its abundance and activity are tightly controlled by post-translational modifications, including ubiquitination-related regulation and palmitoylation, which directly affect ferroptosis sensitivity²⁷. OTUB1, on the other hand, has been shown to inhibit ferroptosis by stabilizing SLC7A11 through deubiquitination, and its anti-ferroptotic activity can itself be weakened by SET7-mediated methylation²⁸. Therefore, the significant correlation of EXOSC4 with GPX4 and OTUB1 is of particular interest, because it links EXOSC4 not only to ferroptosis-associated genes, but specifically to genes whose ferroptosis-regulating functions are already known to depend on post-translational control. Importantly, neddylation and ubiquitination are mechanistically connected through cullin-RING ligases, whose activity depends on cullin neddylation²⁹. Recent studies showed that the CRL3^KCTD10–USP18 axis coordinately regulates SLC7A11 stability and ferroptosis³⁰. Moreover, previous AML studies have reported that the neddylation inhibitor MLN4924 can induce ferroptosis by suppressing the SLC7A11/GPX4 pathway³¹. Consistent with these findings, our CCK-8 results showed that MLN4924 reduced the viability of EXOSC4-overexpressing AML cells and further enhanced erastin-induced inhibition of cell viability. In the context of our study, the significant association of EXOSC4 with ferroptosis-related genes, together with the positive association between EXOSC4 and NEDD8 expression as well as Cullin1-NEDD8 levels, provides a biologically plausible framework suggesting that EXOSC4 may participate in a neddylation–ubiquitination–ferroptosis regulatory network in AML. This hypothesis was further supported by our functional assays, in which EXOSC4 overexpression increased AML cell viability, whereas MLN4924 treatment suppressed this effect and enhanced erastin-induced ferroptosis-related cytotoxicity. However, although the CCK-8 assay combined with erastin and Ferr-1 treatment provides preliminary functional evidence supporting a ferroptosis-related effect, it does not directly measure ferroptotic events. Therefore, further validation using lipid ROS detection, intracellular iron quantification, GPX4/SLC7A11 protein analysis, and additional rescue experiments is required to confirm the precise role of EXOSC4 in regulating ferroptosis in AML.

EXOSC4 is a core component of the RNA exosome complex involved in RNA degradation and post-transcriptional regulation³², and its dysregulation may contribute to tumour-related biological processes. In pancreatic cancer, EXOSC4 has been reported to influence cell survival and apoptosis by regulating key proteins³³. NEDD8, a ubiquitin-like modifier essential for neddylation, plays an important role in tumourigenesis and tumour progression³⁴. In the present study, both EXOSC4 and NEDD8 were significantly upregulated in relapsed AML samples compared with newly diagnosed AML samples, suggesting a potential association between EXOSC4 and NEDD8 in AML progression. In AML cell lines, EXOSC4 overexpression increased NEDD8 protein levels, whereas EXOSC4 knockdown reduced NEDD8 protein expression. Moreover, EXOSC4 overexpression enhanced Cullin1-NEDD8 levels, while the neddylation inhibitor MLN4924 markedly decreased Cullin1-NEDD8 expression without significantly altering EXOSC4 protein levels. These findings suggest that EXOSC4 may be associated with enhanced neddylation activity in AML, particularly through changes in NEDD8 expression and Cullin1 neddylation. However, the molecular mechanism underlying this association remains to be elucidated. In particular, the current data do not allow us to determine whether EXOSC4 affects NEDD8 at the transcriptional, post-transcriptional or protein level. Given that EXOSC4 functions as a component of the RNA exosome complex, its effect on NEDD8 may involve post-transcriptional mechanisms, including altered RNA degradation, RNA processing or transcript stability. It is also possible that EXOSC4 may indirectly affect NEDD8 by modulating transcripts encoding upstream regulators of the neddylation pathway, rather than directly controlling NEDD8 itself. In addition, protein-level mechanisms, such as altered NEDD8 protein stability or changes in Cullin1 neddylation efficiency, cannot be excluded. Therefore, further experiments, including NEDD8 mRNA quantification, actinomycin D-based RNA stability assays, RNA immunoprecipitation, cycloheximide chase assays and proteasome inhibition assays, are needed to clarify the precise mechanism by which EXOSC4 influences NEDD8 expression and neddylation activity in AML.

Furthermore, we constructed a ceRNA network to explore potential regulatory mechanisms involving key neddylation genes in AML neddylation. We identified a potential regulatory pathway between EXOSC4 and NEAT1 mediated by hsa-miR-216b-5p and hsa-miR-2467-3p. NEAT1 inhibits the development and progression of AML by suppressing the Wnt/β-catenin pathway in AML³⁵. Meanwhile, previous studies have demonstrated that EXOSC4 promotes the progression of various tumors, such as ovarian cancer and colon cancer, through the Wnt/β-catenin pathway³⁶. Therefore, we hypothesize that EXOSC4 may regulate abnormal cell proliferation in AML through the Wnt/β-catenin pathway. Further research is needed to confirm this.

Although this study combined bioinformatics analyses with validation in clinical samples and in vitro experiments, several limitations should be acknowledged. First, the limited number of clinical samples may have affected the generalizability and robustness of the findings. Therefore, future studies with larger and independent AML cohorts are required to further validate the clinical relevance of EXOSC4. Second, although our data showed that EXOSC4 expression was positively associated with NEDD8 expression and Cullin1 neddylation, the precise molecular mechanism underlying this association remains unclear. In particular, the current data do not determine whether EXOSC4 affects NEDD8 at the transcriptional, post-transcriptional or protein level. Given that EXOSC4 is a component of the RNA exosome complex, future studies should investigate whether its effect on NEDD8 is mediated through RNA degradation, RNA processing or altered transcript stability. In addition, our CCK-8 assays using erastin, Ferrostatin-1 and MLN4924 provided preliminary evidence that EXOSC4 may be involved in ferroptosis-related regulation of AML cell viability. However, CCK-8 assays do not directly measure ferroptotic events. Further validation using lipid ROS detection, intracellular iron quantification, GPX4/SLC7A11 protein analysis and additional ferroptosis rescue experiments is needed to confirm the role of EXOSC4 in ferroptosis regulation. Finally, although EXOSC4 may represent a potential therapeutic target in AML, the efficacy, specificity and safety of EXOSC4-targeted strategies require further investigation before clinical translation. In summary, our study supports a potential link between EXOSC4-associated neddylation changes and ferroptosis-related pathways in AML. These findings identify EXOSC4 as a candidate molecule for further mechanistic and translational investigation, but additional studies are required to clarify its regulatory mechanism and therapeutic value.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

AML

Acute myeloid leukaemia

ANOVA

Analysis of variance

AUC

Area under the curve

BCA

Bicinchoninic acid

BP

Biological process

CC

Cellular component

CCK-8

Cell Counting Kit-8

cDNA

Complementary DNA

ceRNA

Competitive endogenous RNA

CI

Confidence interval

CISD3

CDGSH iron-sulfur domain 3

CR

Complete remission

CRLs

Cullin-RING ligases

Ct

Cycle threshold

DEGs

Differentially expressed genes

EDTA

Ethylenediaminetetraacetic acid

EV

Empty vector

EXOSC4

Exosome component 4

FBS

Foetal bovine serum

FC

Fold change

Ferr-1

Ferrostatin-1

FerrDb

Ferroptosis database

GAPDH

Glyceraldehyde-3-phosphate dehydrogenase

GO

Gene Ontology

GPX4

Glutathione peroxidase 4

GTEx

Genotype-tissue expression

HR

Hazard ratio

HRP

Horseradish peroxidase

IGF2BP3

Insulin-like growth factor 2 mRNA-binding protein 3

KEGG

Kyoto Encyclopedia of Genes and Genomes

KLHL13

Kelch-like family member 13

LASSO

Least absolute shrinkage and selection operator

lncRNA

Long non-coding RNA

LV

Lentiviral vector

MF

Molecular function

MICM

Morphology, immunophenotyping, cytogenetics and molecular biology

miRNA

MicroRNA

MLN4924

Pevonedistat

mRNA

Messenger RNA

NEDD8

Neural precursor cell expressed developmentally downregulated protein 8

NEAT1

Nuclear paraspeckle assembly transcript 1

NF-κB

Nuclear factor kappa B

OS

Overall survival

OTUB1

OTU deubiquitinase 1

PBS

Phosphate-buffered saline

PMSF

Phenylmethylsulfonyl fluoride

PVDF

Polyvinylidene fluoride

qRT-PCR

Quantitative real-time polymerase chain reaction

R

Relapsed

RIPA

Radioimmunoprecipitation assay

ROC

Receiver operating characteristic

ROS

Reactive oxygen species

RPMI

Roswell Park Memorial Institute

SCF

Skp1-Cullin-F-box

SD

Standard deviation

SDS-PAGE

Sodium dodecyl sulfate–polyacrylamide gel electrophoresis

shRNA

Short hairpin RNA

SIRT6

Sirtuin 6

SLC7A11

Solute carrier family 7 member 11

SLC25A28

Solute carrier family 25 member 28

SOCS2

Suppressor of cytokine signalling 2

STK11

Serine/threonine kinase 11

STR

Short tandem repeat

TARGET

Therapeutically Applicable Research to Generate Effective Treatments

TCGA

The Cancer Genome Atlas

Author contributions

H.G. and As.H. conceived the study, designed and performed the experiments, analyzed the data, and wrote the manuscript. They contributed equally to this work. Sf.X. and W.L. provided technical support. Zh.Z. is in charge of visualization support. Y.C. and S.L. collected patient samples. Js.W. supervised the project. All authors reviewed the manuscript.

Funding

Support was received from the National Natural Science Foundation of China [82370168, 8246010443]; the Basic Research Program of Guizhou Province [QianKeHe JiChu MS (2026) 604]; the Clinical Key Specialty Project of Guizhou Province [GZWJWPF2025007]; and the Clinical Medicine Research Center of Guizhou Province [QianKeHe LCZX(2025)006].

Data availability

The datasets analyzed during the current study are publicly available in the Genotype-Tissue Expression (GTEx) Portal (https://gtexportal.org/home/) and The Cancer Genome Atlas (TCGA) via the Genomic Data Commons Data Portal (https://portal.gdc.cancer.gov/). The combined TCGA-GTEx dataset used in this analysis can be accessed and downloaded from the UCSC Xena browser (https://xenabrowser.net/).

Declarations

Ethics approval and consent to participate

This study involving human participants was conducted in accordance with the Declaration of Helsinki. The protocol was approved by the Institutional Ethics Committee of Basic Medical Research, Affiliated Hospital of Guizhou Medical University (Approval No. 2023-253). Written informed consent was obtained from all adult participants for the research use and publication of de-identified data. For the use of residual, de-identified by-products of routine care, the requirement for informed consent was waived by the ethics committee in accordance with national regulations.

Consent for publication

All participants who provided bone marrow samples for this study signed written informed consent, agreeing to the use of their de-identified data/images for research and publication. For the use of residual, de-identified clinical samples obtained after routine diagnosis or treatment, their utilization was conducted in accordance with national regulations, following the acquisition of participant informed consent or a waiver of informed consent. This consent procedure and the complete study protocol were reviewed and approved by the Institutional Ethics Committee of Basic Medical Research, Affiliated Hospital of Guizhou Medical University (Approval No. 2023-253).

Competing interests

The authors declare no competing interests.