A novel manganese metabolism- and immune-related prognostic risk model for acute myeloid leukemia

Introduction

Acute myeloid leukemia (AML) is a malignant clonal disease that originates from hematopoietic stem cells and is the most common and highly fatal acute leukemia in adults (1). AML is characterized by the clonal proliferation of leukemic blast cells and the impaired differentiation of myeloid lineages, leading to the suppression of the normal hematopoietic function of the bone marrow (2). AML patients present with clinical symptoms such as anemia, infection, and bleeding (2). In recent years, an increasing number of studies have shown that AML patients often exhibit extensive cytogenetic and molecular biological alterations (3-5). AML is classified into various biological subtypes based on cytogenetic abnormalities and gene mutations. The overall survival (OS) of AML patients with different biological subtypes varies (6,7). To guide personalized treatment strategies, patients are stratified into three risk categories: favorable, intermediate, and adverse. Age and physical conditions also affect the prognosis of AML patients (6). However, these prognostic factors often lack accuracy; thus, improved models are needed to guide AML treatment decisions.

Most enzymes require metals for catalytic activity (8). Evidence shows that potassium, calcium, manganese (Mn), and other metals play critical roles in regulating immune function (9). Mn is essential for life (10,11), and plays a key role in critical processes such as gene expression (12,13) and signal transduction (14,15). Recent studies have highlighted the immunoregulatory role of Mn, particularly in activating the cyclic GMP-AMP (cGAMP) synthase (cGAS)-stimulator of interferon genes (STING) pathway, a key mediator of antiviral and anti-tumor immunity (16-18). Mn enhances cGAS sensitivity to double-stranded DNA and its enzymatic activity, enabling cGAMP production, even at low double-stranded DNA concentrations, while strengthening cGAMP-STING binding to boost STING activity and antiviral innate immunity (19-22). Recent studies have shown that Mn also plays a critical role in regulating antibacterial innate immunity (19,23,24).

The type VI secretion system (T6SS), first identified in Vibrio cholerae in 2006, is a key microbial immunity mechanism (25,26). T6SS is widely present in Gram-negative bacteria and is closely related to the pathogenicity of these bacteria (27). Si et al. discovered that the T6SS of Burkholderia thailandensis secretes a T6SS-4-dependent Mn²⁺-binding effector (TseM) to mediate Mn²⁺ transport, enhancing bacterial oxidative stress resistance and virulence (23). Wang et al. showed that viral infection triggers host Mn release from organelles, activating cGAS-STING antiviral defense (19). Later, research demonstrated that bacterial lipopolysaccharide and cyclic dimeric guanosine monophosphate (c-di-GMP) induce host Mn release to activate STING, while the bacterial effector TssS (a T6SS-secreted micropeptide) sequesters intracellular Mn to impair c-di-GMP-mediated STING activation and suppress antibacterial immunity (28). The findings revealed novel bacterial immune evasion mechanisms and provide potential therapeutic strategies for autoimmune diseases. Moreover, Lv et al. discovered that tumor growth and metastasis in Mn-deficient mice were significantly enhanced, and the number of infiltrating cluster of differentiation 8 positive (CD8⁺) T cells in tumors was reduced, indicating that Mn may play a crucial role in inhibiting tumor occurrence and development (24).

Tumor vaccines (29) and immune checkpoint blockade therapy (30) represent promising approaches in the treatment of cancer. However, the efficacy of tumor immunotherapy remains limited due to poor immune cell infiltration, weak antigen presentation, and the immunosuppressive nature of the tumor microenvironment (31). Recent studies have shown that combining Mn²⁺ with programmed cell death protein-1 (PD-1) antibodies significantly enhances anti-tumor efficacy (24,32). Further, Mn-based nanomaterials can precisely deliver immunotherapeutic drugs to tumor cells and reshape the immunosuppressive tumor microenvironment by alleviating hypoxia, macrophage polarization, metabolite clearance, among other mechanisms. Ultimately, they enhance anti-tumor immunity through the generation of reactive oxygen species and the activation of the cGAS-STING pathway (33). Thus, Mn plays a crucial role in regulating immune responses and exerting anti-tumor effects, and thus holds significant potential in immune adjuvant therapy and tumor immunotherapy.

In this study, we developed a manganese metabolism- and immune-related gene (MIRG)-based prognostic risk model using publicly available transcriptomic data and clinical information to better predict the prognosis of AML patients. External independent datasets (GSE12417 and GSE71014) demonstrated the practicality of the model. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-933/rc).

Methods

Data collection and organization

The gene expression profiles of 151 AML patients were downloaded from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/). The data of 70 healthy bone marrow samples were acquired from the Genotype-Tissue Expression (GTEx) database (https://www.gtexportal.org/). After excluding AML patients with missing survival time and status data in TCGA data set, the data of 140 AML patients with complete survival data were retained for the subsequent survival analysis. Two independent datasets (GSE71014 and GSE12417) from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) were used for validation. A single-cell dataset was also downloaded from the GEO database. Based on previously published literature (34), we retrieved 1,399 manganese metabolism-related genes (MRGs) and 2,483 immune-related genes (IRGs) from the Gene Cards database (https://www.genecards.org/) and ImmPort database (https://www.immport.org/), respectively. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. As all the data used in this study were sourced from public databases, no ethical approval was required.

Construction and validation of the risk model

In total, 32 differentially expressed MIRGs underwent univariate Cox regression analysis, followed by multivariate Cox regression analysis to build the prognostic risk model. Five MIRGs were ultimately included in the AML model. The AML patients were then stratified into high- and low-risk groups based on the median risk score. A Kaplan-Meier analysis was conducted to compare the OS of patients between the risk subgroups. To evaluate the predictive accuracy of our risk model, the same coefficients were applied to GSE71014 and GSE12417 for validation. To determine whether the MIRG-based signatures could serve as independent prognostic factors, univariate and multivariate Cox regression analyses were performed, incorporating the risk score, as well as other clinical data, such as age, gender, blood cell count, French-American-British (FAB) classification, and cytogenetic risk.

Establishment of a nomogram

A nomogram, which integrated the risk score and clinical factors, was used to predict the OS of AML patients at 1, 2 and 3 years. A calibration plot was used to evaluate the predictive efficacy of the nomogram.

Functional enrichment analysis

“Limma” package was used to screen the differentially expressed genes (DEGs) between the high- and low-risk subgroups, with the thresholds of |log₂ fold change| >1 and adjusted P<0.05. Using the “clusterProfiler” package, we performed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses on these DEGs. Terms and pathways with P<0.05 were considered statistically significant.

Immune infiltration and immune checkpoint analysis

Our study investigated the differences in immune cell infiltration between the high- and low-risk groups, and evaluated the correlations among the MIRG-based risk score, the infiltration levels of immune cells, and the expression levels of the five MIRGs. We also analyzed the expression levels of 10 immune checkpoints to investigate their association with the risk score and their potential effect on treatment response. Additionally, a survival analysis was conducted by combining the risk score with the expression levels of immune checkpoints.

Single-cell RNA-sequencing data

To explore the distribution of the five MIRGs in the AML tumor microenvironment, we analyzed a single-cell RNA-sequencing dataset GSE116256 using the “Seurat” package (version 5.3.0) following standard quality control and normalization (35,36).

Statistical analysis

All the statistical analyses in this study were conducted using R software (version 4.4.2) (https://www.R-project.org/). Significant values were set at P<0.05.

Results

Data collection and processing

We collected the gene expression profiles of 151 AML patients from TCGA-LAML dataset, and 70 healthy bone marrow samples from the GTEx dataset. The clinical and survival information were primarily obtained from TCGA dataset. The detailed clinical information of the patients in the TCGA data set is shown in Table 1. After excluding AML samples with incomplete survival information from the TCGA data set, 140 patients with complete survival data were included in the subsequent survival analysis. We then combined the transcriptome information of 151 AML patient samples and 70 healthy bone marrow samples, and performed a principal component analysis (PCA) to eliminate any batch effects caused by the merging of the two datasets (Figure 1A,1B). Subsequently, we performed a differential expression analysis using the thresholds of |log₂ fold change| >1 and adjusted P<0.05, identifying 2,006 DEGs (Figure 1C), of which 1,604 were upregulated and 402 were downregulated. To identify the MIRGs, we intersected the DEGs with the MRGs and IRGs, and ultimately identified 32 differentially expressed MIRGs (Figure 1D).

Figure 1 Differentially expressed MIRGs. (A) Plot before the PCA analysis. (B) Plot after the PCA analysis. (C) Heatmap of DEGs. (D) Venn diagram of overlapping genes among the DEGs, MRGs, and IRGs. DEGs, differentially expressed genes; IRGs, immune-related genes; MIRGs, manganese metabolism- and immune-related genes; MRGs, manganese metabolism-related genes; PCA, principal component analysis.

Construction and validation of prognostic risk model in AML

Based on the univariate Cox regression analysis of 32 differentially expressed MIRGs, eight potential prognosis-related genes were identified (Table 2). Subsequently, through a multivariate Cox regression analysis, we established a prognostic risk model comprising CGA, TGFA, GKN1, S100G, and CCL23 (Table 3). The risk score was calculated using the following formula: risk score = 1.51 × CGA − 0.61 × TGFA − 2.60 × GKN1 − 0.78 × S100G + 0.33 × CCL23. The AML patients were divided into high- and low-risk groups based on the median risk score. The Kaplan-Meier analysis showed that the high-risk group had inferior OS in TCGA cohort (P<0.001) (Figure 2A). We also verified the predictive efficacy of this risk model by drawing receiver operating characteristic (ROC) curves. The area under the curve (AUC) values at 1, 2 and 3 years were 0.720, 0.769, and 0.764, respectively, indicating that the model had good predictive ability (Figure 2B). In addition, the survival outcomes and risk scores in TCGA cohort, as well as the differential expression levels of the five MIRGs among the subgroups, were examined (Figure 2C). To test the reliability of the model, the same coefficients were applied to the two validation datasets (GSE71014 and GSE12417). Consistent with the results of TCGA cohort, the high-risk group had a shorter survival time in the GSE71014 (P=0.006) and GSE12714 (P=0.047) datasets (Figure 3A,3B). The AUC values at 1, 2 and 3 years were 0.567, 0.634, and 0.696, and 0.719, 0.661, and 0.645 in the GSE71014 and GSE12417 datasets, respectively, showing the accuracy and reliability of the model (Figure 3C,3D). Additionally, the differences in the expression levels of the five MIRGs, risk scores, and survival statuses of the two GEO cohorts are shown in Figure 3E,3F.

Figure 2 Construction of the prognostic risk model. (A) Kaplan-Meier survival curve for the prognostic risk model in TCGA cohort. (B) Time-dependent ROC curves evaluating the predictive accuracy of the MIRGs in TCGA cohort. (C) The risk scores, survival status, and a heatmap for TCGA cohort. AUC, area under the curve; MIRGs, manganese metabolism- and immune-related genes; ROC, receiver operating characteristic; TCGA, The Cancer Genome Atlas.

Figure 3 Validation of the prognostic risk model. (A,B) Kaplan-Meier survival curves for the prognostic risk model in GSE71014 and GSE12417, respectively. (C,D) Time-dependent ROC curves evaluating the predictive accuracy of the MIRGs in GSE71014 and GSE12417, respectively. (E,F) The risk scores, survival status, and gene expression heatmap for GSE71014 and GSE12417, respectively. AUC, area under the curve; MIRGs, manganese metabolism- and immune-related genes; ROC, receiver operating characteristic.

Independent prognostic analysis and construction of a nomogram

To verify the independent nature of the prognostic risk model with other clinical factors in the AML patients, we performed univariate and multivariate Cox regression analyses. The univariate Cox regression analysis demonstrated that age [hazard ratio (HR) =1.044; 95% confidence interval (CI): 1.026–1.061; P<0.001], cytogenetic risk (HR =0.560; 95% CI: 0.393–0.797; P=0.001), and risk score (HR =2.045; 95% CI: 1.465–2.855; P<0.001) were independent predictors affecting the prognosis of the AML patients (Figure 4A). Subsequently, the clinical factors with P values <0.05 in the above analysis were included in the multivariate Cox regression analysis. We found that age (HR =1.039; 95% CI: 1.021–1.057; P<0.001) and risk score (HR =1.866; 95% CI: 1.296–2.685; P<0.001) remained the key prognostic factors for AML (Figure 4B). To further assess the predictive value of these factors for individual OS, we built a nomogram integrating age and risk score (Figure 4C). The nomogram can be used to predict 1-, 2- and 3-year OS in AML. Our calibration plot indicated that the nomogram reliably predicted the prognosis of the patients with AML (Figure 4D). These findings suggest that the five MIRGs could serve as novel prognostic biomarkers.

Figure 4 Independent prognostic analysis. (A) Forest plot of the univariate Cox regression analysis in TCGA cohort. (B) Forest plot of the multivariate Cox regression analysis in TCGA cohort. (C) Nomogram predicting the 1-, 2- and 3-year survival rate of the AML patients in TCGA cohort. (D) Calibration curves for the nomogram in TCGA cohort. AML, acute myeloid leukemia; CI, confidence interval; OS, overall survival; TCGA, The Cancer Genome Atlas.

DEG, GO, and KEGG analyses

We performed a differential analysis between the high- and low-risk groups to characterize their molecular profiles (|log₂ fold change| >1; adjusted P<0.05). A total of 766 DEGs were identified, of which 554 were upregulated and 212 were downregulated. A volcano plot (Figure 5A) and a heatmap (Figure 5B) were used to visualize the DEGs. In the GO-biological process (BP) analysis, the pathways mainly focused on the positive regulation of leukocyte migration, chemotaxis, and taxis. In the GO-cellular component (CC) analysis, the main pathways included the external side of plasma membrane, collagen-containing extracellular matrix, and endocytic vesicles. While in the GO-molecular function (MF) analysis, the main pathways were immune receptor activity, cytokine binding, and pattern recognition receptor activity (Figure 5C). Further, the KEGG enrichment analysis indicated that the DEGs were mainly associated with pathways such as cytokine-cytokine receptor interaction, and phagosome and neutrophil extracellular trap formation (Figure 5D).

Figure 5 Differential analysis between the risk subgroups, and the functional enrichment analysis. (A) Volcano plot of the DEGs between the risk subgroups. (B) Heatmap of the DEGs between the risk subgroups. (C) The top six GO-BP, -CC and -MF terms. (D) The top 10 KEGG pathways. DEGs, differentially expressed genes; GO, Gene Ontology; BP, biological process; CC, cellular component; MF, molecular function; KEGG, Kyoto Encyclopedia of Genes and Genomes.

Immune infiltration analysis

In the immune infiltration analysis, we found that six immune cell subsets exhibited differential infiltration levels between the risk subgroups. Specifically, monocytes, M2 macrophages, activated dendritic cells (DCs), and neutrophils were enriched in high-risk AML, while resting CD4 memory T cells and resting mast cells were enriched in low-risk AML (Figure 6A). Further, the correlation analysis showed that the MIRG-based risk score was positively correlated with M2 macrophages (R=0.36; P<0.001) and monocytes (R=0.32; P<0.001), but negatively correlated with resting natural killer (NK) cells (R=−0.19; P=0.03), and resting mast cells (R=−0.39; P<0.001) (Figure 6B-6E). These findings suggest that the tumor immune microenvironment is closely associated with the prognosis of patients with AML. Further, the MIRG-based risk score was found to be positively correlated with messenger RNA (mRNA) expressions of CCL23 (R=0.54; P<0.001) and CGA (R=0.1; P=0.23) (Figure 6F,6G), but negatively correlated with mRNA expressions of GKN1 (R=−0.39; P<0.001), TGFA (R=−0.54; P<0.001) and S100G (R=−0.52; P<0.001) (Figure 6H-6J). Notably, the MIRG-based risk score was positively correlated with CCL23 mRNA expression, which itself was positively correlated with M2 macrophages, activated DCs, monocytes, and neutrophils (Figure 6K). Thus, CCL23 may be a potential mediator of poor AML outcomes in the tumor microenvironment.

Figure 6 Immune infiltration analysis. (A) Infiltration differences in immune cells between the risk subgroups. (B-E) Correlations among the MIRG-based risk score and immune cells. (F-J) Correlations among the MIRG-based risk score and the five MIRGs. (K) Correlations among the immune cells and the five MIRGs. *, P<0.05; **, P<0.01; ***, P<0.001. MIRGs, manganese metabolism- and immune-related genes.

Immune checkpoint analysis

Recent studies have shown that immune checkpoint inhibitors, which are a cornerstone of tumor immunotherapy, can significantly improve the OS of cancer patients (37,38). We selected 10 key immune checkpoints (LAG3 and CD274, which encode the PD-L1 protein; IDO1, HAVCR2, CTLA4, ICOS, TIGIT, BTLA, and PDCD1, which encode the PD-1 protein; and PDCD1LG2, which encodes the PD-L2 protein) from published literature to investigate their associations with the MIRG-based risk score and identify the mechanisms underlying immunotherapy response. High-risk AML exhibited elevated CTLA4, PDCD1, and PDCD1LG2 expression compared to low-risk AML (P<0.05) (Figure 7A). Further, the survival analysis revealed that the low-risk AML patients with low immune checkpoint (PDCD1, PDCD1LG2, and CTLA4) expression had a superior prognosis (P<0.001) (Figure 7B-7D). These results suggest that the MIRG-based risk score may serve as a predictive biomarker for immunotherapy response.

Figure 7 Immune checkpoint analysis. (A) Comparison of immune checkpoint expression. (B-D) Kaplan-Meier survival curves combining the risk scores with the expression of immune checkpoints.

Single-cell RNA-sequencing analysis

To analyze the expression levels of the five MIRGs at the single-cell level, Uniform Manifold Approximation and Projection (UMAP) was used to cluster and annotate the cells (Figure 8A). The results showed that TGFA was expressed in progenitor-like cells, monocytes, promonocyte-like cells, granulocyte-monocyte progenitor (GMP)-like cells, T cells, and promonocytes, but the expression levels were low (Figure 8B). CCL23 was mainly highly expressed in promonocytes, promonocyte-like cells, monocytes, progenitors, and progenitor-like cells, and was also lowly expressed in GMP, T cells, plasma cells, and conventional DC-like cells (Figure 8C). In addition, we performed a violin plot analysis to show the expression levels of TGFA and CCL23 in the cells (Figure 8D). Unfortunately, we failed to observe the expression of CGA, GKN1, and S100G at the single-cell level, which might be related to quality control issues in data processing.

Figure 8 Single-cell RNA-sequencing analysis. (A) UMAP plot of 17 cell clusters. (B,C) UMAP plots of AML samples showing the expression levels of TGFA and CCL23, respectively. (D) Volin plots comparing the expression levels of TGFA and CCL23 in the different types of cells. AML, acute myeloid leukemia; UMAP, Uniform Manifold Approximation and Projection.

Discussion

AML is the most common and highly heterogeneous leukemia in adults (39). It is characterized by the abnormal proliferation of myeloid blasts and has high incidence and mortality rates (39). Prognostic stratification in AML management is based on cytogenetic and molecular genetic profiles, immunophenotypic characteristics, as well as dynamic assessments of treatment response and patient-specific factors (40). However, the lack of reliable prognostic biomarkers in AML remains a major clinical challenge.

Mn plays a crucial role in regulating immune responses and exerting anti-tumor effects, and thus has significant potential in immune adjuvant therapy and cancer immunotherapy (24). Building on previous findings, our study integrated MRGs and IRGs to construct a prognostic risk model, aiming to identify new prognostic biomarkers and therapeutic targets for AML. This study developed a prognostic risk model integrating five MIRGs (CGA, TGFA, GKN1, S100G, and CCL23) using public databases. Previous studies have shown that these genes play critical roles in the pathogenesis of multiple diseases.

The α-subunit encoded by the CGA gene is a common component of four glycoprotein hormones (i.e., thyroid-stimulating hormone, luteinizing hormone, follicle-stimulating hormone, and human chorionic gonadotropin) (41). It has been reported that mutations or the abnormal expression of CGA may lead to disorders in the synthesis of glycoprotein hormones, causing endocrine diseases such as hypothyroidism and hypogonadism. Further, CGA is overexpressed in pituitary adenomas, thyroid carcinomas, and germ cell tumors, and thus can be used as a biomarker for the diagnosis and monitoring of these diseases (41). The aforementioned studies indicate that CGA has the potential to serve as a disease biomarker, and its high expression may be associated with a poor prognosis in certain diseases. Our results showed that the elevated expression level of CGA in AML patients was correlated with worse OS. However, further investigations are required to elucidate the underlying mechanism.

TGFA, a member of the epidermal growth factor family of polypeptide growth factors, binds with high affinity to epidermal growth factor receptor (EGFR) (42). TGFA-EGFR binding triggers receptor dimerization and kinase activation, initiating RAS/MAPK and PI3K/AKT signaling, driving tumor cell proliferation, invasion, angiogenesis, and apoptosis resistance (43-48). In breast and colon cancer, the high expression of TGFA has been shown to promote the proliferation, invasion and metastasis of cancer cells (49,50). In addition, Ma et al. showed that TGFA was highly expressed in cervical cancer tissues and cells. TGFA knockdown has been shown to inhibit the proliferation, migration, and invasion of cervical cancer cells (51). However, we found that TGFA was lowly expressed in high-risk AML, and thus may exert protective effects in the pathogenesis of AML. This phenomenon may be attributed to specific cell-type dependencies or microenvironmental contexts. Future research should further explore the dual roles of TGFA in cellular proliferation and investigate its underlying molecular mechanisms.

GKN1, a stomach-specific tumor suppressor (52), exerts anti-tumor effects by inducing cellular senescence via p16/Rb pathway activation (53), promoting apoptosis through Fas/FasL-Caspase cascades (54), and inhibiting NF-κB signaling to suppress metastasis (55). Its downregulation is significantly associated with gastric cancer progression and a poor prognosis, highlighting its potential as a diagnostic and prognostic biomarker (56). Our study found that the expression level of GKN1 was increased in the low-risk AML patients, which suggests that it exerts a protective effect. This finding supports previous reports and shows that GKN1 has significant potential as a diagnostic and prognostic biomarker for AML patients.

S100G is a member of the S100 protein family. Numerous studies have confirmed that the abnormal expression of S100 family proteins is closely related to the occurrence and development of various tumors such as liver cancer, kidney cancer, ovarian cancer, and colorectal cancer (57-61). In this study, elevated S100G expression was significantly associated with a favorable prognosis in AML patients, which suggests that it exerts protective effects in the pathogenesis of AML. However, given the current limited understanding of the mechanistic functions of S100G, further investigations are required to explore its molecular mechanisms and biological roles.

CCL23, a relatively novel member of the CC chemokine family, has garnered significant attention due to its biological roles. CCL23 has been reported to be frequently downregulated at both the mRNA and protein levels in hepatocellular carcinoma, and its reduced expression has been reported to be significantly correlated with a poor prognosis (62). In biliary tract cancer, serum CCL23 levels are markedly elevated, and are correlated with tumor aggressiveness and poor postoperative prognosis (63). Further, a previous study found that the increased expression of CCL23 in AML cells is associated with a poor prognosis (64). In the bone marrow of adult and pediatric AML patients, the average level of CCL23 is higher than that of healthy individuals (64). In this study, the high expression of CCL23 was associated with a poor prognosis in the AML patients, suggesting its potential as an adverse prognostic factor for AML. Although these genes have been implicated in the prognoses of certain diseases, their precise molecular mechanisms and biological functions require further investigation for comprehensive elucidation.

Our model demonstrated strong predictive accuracy between the risk subgroups. Age and risk score were identified as independent prognostic factors and integrated into a practical nomogram. The high-risk patients showed significant enrichment in cytokine signaling and immune-related pathways, along with distinct immune microenvironment characteristics, including elevated monocytes, M2 macrophages, activated DCs, and neutrophils, while the low-risk patients exhibited more resting CD4 memory T cells and resting mast cells. These findings provide valuable insights into AML prognosis assessment and reveal potential immune-related therapeutic targets. Unlike their anti-tumor M1 counterparts, M2 macrophages are known to promote tumor progression (65). Notably, we found that M2 macrophages were significantly enriched in the high-risk AML patients. Some studies have shown that M2 macrophages play a crucial role in the pathogenesis of AML. They create an immunosuppressive microenvironment by secreting cytokines, thereby inhibiting the activity of T cells and NK cells. They also promote disease progression through metabolic reprogramming and resistance to phagocytosis (66,67), ultimately leading to poor clinical outcomes, including increased relapse risk and reduced survival (68). These findings suggested that the poor prognosis of AML patients might be related to the infiltration of M2 macrophages. We also found that the expression of immune checkpoints (PDCD1, PDCD1LG2, and CTLA4) was upregulated in high-risk AML patients, and correlated with a worse outcome. These findings suggest that elevated levels of these immune checkpoints contribute to a poor prognosis in high-risk AML patients, and thus could serve as therapeutic targets. Notably, while our findings revealed their involvement in shaping the tumor microenvironment and immune response, emerging evidence indicates that these molecules may also exert direct regulatory effects on leukemic stem cells through mechanisms independent of their classic immune-modulatory roles (69). This dual function requires further investigation.

It should be acknowledged that this study has some limitations, and further research is required to fully elucidate the mechanisms of Mn metabolism in AML.

Conclusions

We developed a novel prognostic risk model for AML that demonstrated robust predictive performance, offering clinically actionable insights for treatment decision-making.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-933/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-933/prf

Funding: This study was supported by the Youth Science Fund Project of the National Natural Science Foundation of China (No. 81800100).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-933/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Since all the data used in this study were sourced from public databases, no ethical approval was required.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. De Kouchkovsky I, Abdul-Hay M. 'Acute myeloid leukemia: a comprehensive review and 2016 update'. Blood Cancer J 2016;6:e441. [Crossref] [PubMed]
2. Döhner H, Weisdorf DJ, Bloomfield CD. Acute Myeloid Leukemia. N Engl J Med 2015;373:1136-52. [Crossref] [PubMed]
3. Döhner H, Estey E, Grimwade D, et al. Diagnosis and management of AML in adults: 2017 ELN recommendations from an international expert panel. Blood 2017;129:424-47. [Crossref] [PubMed]
4. Bullinger L, Döhner K, Döhner H. Genomics of Acute Myeloid Leukemia Diagnosis and Pathways. J Clin Oncol 2017;35:934-46. [Crossref] [PubMed]
5. Al-Harbi S, Aljurf M, Mohty M, et al. An update on the molecular pathogenesis and potential therapeutic targeting of AML with t(8;21)(q22;q22.1);RUNX1-RUNX1T1. Blood Adv 2020;4:229-38. [Crossref] [PubMed]
6. Pogosova-Agadjanyan EL, Moseley A, Othus M, et al. AML risk stratification models utilizing ELN-2017 guidelines and additional prognostic factors: a SWOG report. Biomark Res 2020;8:29. [Crossref] [PubMed]
7. Koenig KL, Sahasrabudhe KD, Sigmund AM, et al. AML with Myelodysplasia-Related Changes: Development, Challenges, and Treatment Advances. Genes (Basel) 2020;11:845. [Crossref] [PubMed]
8. Andreini C, Bertini I, Cavallaro G, et al. Metal ions in biological catalysis: from enzyme databases to general principles. J Biol Inorg Chem 2008;13:1205-18. [Crossref] [PubMed]
9. Wang C, Zhang R, Wei X, et al. Metalloimmunology: The metal ion-controlled immunity. Adv Immunol 2020;145:187-241. [Crossref] [PubMed]
10. Schroeder HA, Balassa JJ, Tipton IH. Essential trace metals in man: manganese. A study in homeostasis. J Chronic Dis 1966;19:545-71. [Crossref] [PubMed]
11. Wedler FC. Biological significance of manganese in mammalian systems. Prog Med Chem 1993;30:89-133. [Crossref] [PubMed]
12. Uchida A, Oh-hashi K, Kiuchi K, et al. Manganese regulates caspase-3 gene promoter activity by inducing Sp1 phosphorylation in PC12 cells. Toxicology 2012;302:292-8. [Crossref] [PubMed]
13. Guo Z, Zhang Z, Wang Q, et al. Manganese chloride induces histone acetylation changes in neuronal cells: Its role in manganese-induced damage. Neurotoxicology 2018;65:255-63. [Crossref] [PubMed]
14. El Mchichi B, Hadji A, Vazquez A, et al. p38 MAPK and MSK1 mediate caspase-8 activation in manganese-induced mitochondria-dependent cell death. Cell Death Differ 2007;14:1826-36. [Crossref] [PubMed]
15. Kim DS, Jin H, Anantharam V, et al. p73 gene in dopaminergic neurons is highly susceptible to manganese neurotoxicity. Neurotoxicology 2017;59:231-9. [Crossref] [PubMed]
16. Liu N, Pang X, Zhang H, et al. The cGAS-STING Pathway in Bacterial Infection and Bacterial Immunity. Front Immunol 2021;12:814709. [Crossref] [PubMed]
17. Samson N, Ablasser A. The cGAS-STING pathway and cancer. Nat Cancer 2022;3:1452-63. [Crossref] [PubMed]
18. Zhang Z, Yang J, Zhou Q, et al. The role and mechanism of the cGAS-STING pathway-mediated ROS in apoptosis and ferroptosis induced by manganese exposure. Redox Biol 2025;85:103761. [Crossref] [PubMed]
19. Wang C, Guan Y, Lv M, et al. Manganese Increases the Sensitivity of the cGAS-STING Pathway for Double-Stranded DNA and Is Required for the Host Defense against DNA Viruses. Immunity 2018;48:675-687.e7. [Crossref] [PubMed]
20. Li J, Ren H, Qiu Q, et al. Manganese Coordination Micelles That Activate Stimulator of Interferon Genes and Capture In Situ Tumor Antigens for Cancer Metalloimmunotherapy. ACS Nano 2022;16:16909-23. [Crossref] [PubMed]
21. Li Z, Chu Z, Yang J, et al. Immunogenic Cell Death Augmented by Manganese Zinc Sulfide Nanoparticles for Metastatic Melanoma Immunotherapy. ACS Nano 2022;16:15471-83. [Crossref] [PubMed]
22. Lu Q, Chen R, Du S, et al. Activation of the cGAS-STING pathway combined with CRISPR-Cas9 gene editing triggering long-term immunotherapy. Biomaterials 2022;291:121871. [Crossref] [PubMed]
23. Si M, Zhao C, Burkinshaw B, et al. Manganese scavenging and oxidative stress response mediated by type VI secretion system in Burkholderia thailandensis. Proc Natl Acad Sci U S A 2017;114:E2233-42. [Crossref] [PubMed]
24. Lv M, Chen M, Zhang R, et al. Manganese is critical for antitumor immune responses via cGAS-STING and improves the efficacy of clinical immunotherapy. Cell Res 2020;30:966-79. [Crossref] [PubMed]
25. Pukatzki S, Ma AT, Sturtevant D, et al. Identification of a conserved bacterial protein secretion system in Vibrio cholerae using the Dictyostelium host model system. Proc Natl Acad Sci U S A 2006;103:1528-33. [Crossref] [PubMed]
26. Chen Z, Mao Y, Song Y, et al. Refined egoist: The toxin-antitoxin immune system of T6SS. Microb Pathog 2024;196:106991. [Crossref] [PubMed]
27. Cianfanelli FR, Monlezun L, Coulthurst SJ. Aim, Load, Fire: The Type VI Secretion System, a Bacterial Nanoweapon. Trends Microbiol 2016;24:51-62. [Crossref] [PubMed]
28. Zhu L, Xu L, Wang C, et al. T6SS translocates a micropeptide to suppress STING-mediated innate immunity by sequestering manganese. Proc Natl Acad Sci U S A 2021;118:e2103526118. [Crossref] [PubMed]
29. Sahin U, Türeci Ö. Personalized vaccines for cancer immunotherapy. Science 2018;359:1355-60. [Crossref] [PubMed]
30. Gong N, Sheppard NC, Billingsley MM, et al. Nanomaterials for T-cell cancer immunotherapy. Nat Nanotechnol 2021;16:25-36. [Crossref] [PubMed]
31. Topalian SL, Drake CG, Pardoll DM. Immune checkpoint blockade: a common denominator approach to cancer therapy. Cancer Cell 2015;27:450-61. [Crossref] [PubMed]
32. Zhao Z, Dong S, Liu Y, et al. Tumor Microenvironment-Activable Manganese-Boosted Catalytic Immunotherapy Combined with PD-1 Checkpoint Blockade. ACS Nano 2022;16:20400-18. [Crossref] [PubMed]
33. Zhang K, Qi C, Cai K. Manganese-Based Tumor Immunotherapy. Adv Mater 2023;35:e2205409. [Crossref] [PubMed]
34. Han X, Leng C, Zhao S, et al. Development and verification of a manganese metabolism- and immune-related genes signature for prediction of prognosis and immune landscape in gastric cancer. Front Immunol 2024;15:1377472. [Crossref] [PubMed]
35. Butler A, Hoffman P, Smibert P, et al. Integrating single-cell transcriptomic data across different conditions, technologies, and species. Nat Biotechnol 2018;36:411-20. [Crossref] [PubMed]
36. Stuart T, Butler A, Hoffman P, et al. Comprehensive Integration of Single-Cell Data. Cell 2019;177:1888-1902.e21. [Crossref] [PubMed]
37. Rui R, Zhou L, He S. Cancer immunotherapies: advances and bottlenecks. Front Immunol 2023;14:1212476. [Crossref] [PubMed]
38. Naimi A, Mohammed RN, Raji A, et al. Tumor immunotherapies by immune checkpoint inhibitors (ICIs); the pros and cons. Cell Commun Signal 2022;20:44. [Crossref] [PubMed]
39. Vago L, Gojo I. Immune escape and immunotherapy of acute myeloid leukemia. J Clin Invest 2020;130:1552-64. [Crossref] [PubMed]
40. Döhner H, Wei AH, Appelbaum FR, et al. Diagnosis and management of AML in adults: 2022 recommendations from an international expert panel on behalf of the ELN. Blood 2022;140:1345-77. [Crossref] [PubMed]
41. Fiddes JC, Goodman HM. The gene encoding the common alpha subunit of the four human glycoprotein hormones. J Mol Appl Genet 1981;1:3-18.
42. Kudlow JE, Bjorge JD. TGF alpha in normal physiology. Semin Cancer Biol 1990;1:293-302.
43. Di Marco E, Pierce JH, Aaronson SA, et al. Mechanisms by which EGF receptor and TGF alpha contribute to malignant transformation. Nat Immun Cell Growth Regul 1990;9:209-21.
44. Grandis JR, Tweardy DJ. TGF-alpha and EGFR in head and neck cancer. J Cell Biochem Suppl 1993;17F:188-91. [Crossref] [PubMed]
45. Greten FR, Wagner M, Weber CK, et al. TGF alpha transgenic mice. A model of pancreatic cancer development. Pancreatology 2001;1:363-8. [Crossref] [PubMed]
46. Schultz G, Rotatori DS, Clark W. EGF and TGF-alpha in wound healing and repair. J Cell Biochem 1991;45:346-52. [Crossref] [PubMed]
47. Wong DT. TGF-alpha and oral carcinogenesis. Eur J Cancer B Oral Oncol 1993;29B:3-7. [Crossref] [PubMed]
48. Kavanagh S, Mirzai B, Fuller K, et al. TGFα expression in myeloid malignancies. J Clin Pathol 2016;69:543-6. [Crossref] [PubMed]
49. Yu CY, Chang WC, Zheng JH, et al. Transforming growth factor alpha promotes tumorigenesis and regulates epithelial-mesenchymal transition modulation in colon cancer. Biochem Biophys Res Commun 2018;506:901-6. [Crossref] [PubMed]
50. Liu Y, Zhang A, Bao PP, et al. MicroRNA-374b inhibits breast cancer progression through regulating CCND1 and TGFA genes. Carcinogenesis 2021;42:528-36. [Crossref] [PubMed]
51. Ma X, Zheng J, He K, et al. TGFA expression is associated with poor prognosis and promotes the development of cervical cancer. J Cell Mol Med 2024;28:e18086. [Crossref] [PubMed]
52. Yoon JH, Ashktorab H, Smoot DT, et al. Uptake and tumor-suppressive pathways of exosome-associated GKN1 protein in gastric epithelial cells. Gastric Cancer 2020;23:848-62. [Crossref] [PubMed]
53. Xing R, Li W, Cui J, et al. Gastrokine 1 induces senescence through p16/Rb pathway activation in gastric cancer cells. Gut 2012;61:43-52. [Crossref] [PubMed]
54. Rippa E, La Monica G, Allocca R, et al. Overexpression of gastrokine 1 in gastric cancer cells induces Fas-mediated apoptosis. J Cell Physiol 2011;226:2571-8. [Crossref] [PubMed]
55. Xing R, Cui JT, Xia N, et al. GKN1 inhibits cell invasion in gastric cancer by inactivating the NF-kappaB pathway. Discov Med 2015;19:65-71.
56. Yoon JH, Park YG, Nam SW, et al. The diagnostic value of serum gastrokine 1 (GKN1) protein in gastric cancer. Cancer Med 2019;8:5507-14. [Crossref] [PubMed]
57. Zhang C, Yao R, Chen J, et al. S100 family members: potential therapeutic target in patients with hepatocellular carcinoma: A STROBE study. Medicine (Baltimore) 2021;100:e24135. [Crossref] [PubMed]
58. Chen H, Xu C, Jin Q, et al. S100 protein family in human cancer. Am J Cancer Res 2014;4:89-115.
59. Li F, Men X, Zhang W. S100 protein in breast tumor. Indian J Cancer 2014;51:e67-71. [Crossref] [PubMed]
60. Bresnick AR, Weber DJ, Zimmer DB. S100 proteins in cancer. Nat Rev Cancer 2015;15:96-109. [Crossref] [PubMed]
61. Yang M, Cui W, Lv X, et al. S100P is a ferroptosis suppressor to facilitate hepatocellular carcinoma development by rewiring lipid metabolism. Nat Commun 2025;16:509. [Crossref] [PubMed]
62. Meng J, Wang L, Hou J, et al. CCL23 suppresses liver cancer progression through the CCR1/AKT/ESR1 feedback loop. Cancer Sci 2021;112:3099-110. [Crossref] [PubMed]
63. Roderburg C, Labuhn S, Bednarsch J, et al. Elevated Serum Levels of CCL23 Are Associated with Poor Outcome after Resection of Biliary Tract Cancer. Mediators Inflamm 2022;2022:6195004. [Crossref] [PubMed]
64. Korbecki J, Bosiacki M, Stasiak P, et al. Clinical Aspects and Significance of β-Chemokines, γ-Chemokines, and δ-Chemokines in Molecular Cancer Processes in Acute Myeloid Leukemia (AML) and Myelodysplastic Neoplasms (MDS). Cancers (Basel) 2024;16:3246. [Crossref] [PubMed]
65. Liu J, Wei Y, Jia W, et al. Chenodeoxycholic acid suppresses AML progression through promoting lipid peroxidation via ROS/p38 MAPK/DGAT1 pathway and inhibiting M2 macrophage polarization. Redox Biol 2022;56:102452. [Crossref] [PubMed]
66. Al-Matary YS, Botezatu L, Opalka B, et al. Acute myeloid leukemia cells polarize macrophages towards a leukemia supporting state in a Growth factor independence 1 dependent manner. Haematologica 2016;101:1216-27. [Crossref] [PubMed]
67. Weinhäuser I, Pereira-Martins DA, Almeida LY, et al. M2 macrophages drive leukemic transformation by imposing resistance to phagocytosis and improving mitochondrial metabolism. Sci Adv 2023;9:eadf8522. [Crossref] [PubMed]
68. Xie Y, Yang H, Yang C, et al. Role and Mechanisms of Tumor-Associated Macrophages in Hematological Malignancies. Front Oncol 2022;12:933666. [Crossref] [PubMed]
69. Reinhardt C, Ochsenbein AF. Immune checkpoints regulate acute myeloid leukemia stem cells. Leukemia 2025;39:1277-93. [Crossref] [PubMed]

(English Language Editor: L. Huleatt)