Development and validation of manganese metabolism-related genes prognostic model in colorectal cancer and its immunological characteristics

Introduction

The worldwide incidence and mortality of colorectal cancer (CRC) have demonstrated a continuous surge, cementing its status as one of the most significant oncological threats within the digestive tract (1,2). CRC stands as the third highest-ranking cancer diagnosis worldwide, while simultaneously accounting for the second largest proportion of cancer deaths, underscoring its profound impact on public health (3,4). Early symptoms of this disease are often subtle, with approximately 40–50% of patients presenting at an advanced stage with distant metastasis at diagnosis. The 5-year survival rate for such patients is below 12% (5,6). Current treatment modalities for CRC include surgery, chemotherapy, and immunotherapy (7). Chemotherapy is widely limited by strong drug resistance, significant toxicity, and poor safety profiles (8). While the introduction of targeted and immune-based therapies has refined individualized care for CRC patients, the ongoing debate surrounding their long-term robustness and safety profiles has effectively prevented the establishment of a unified therapeutic consensus (9,10). Therefore, in-depth analysis of CRC pathogenesis is crucial for achieving early precision diagnosis, developing personalized treatment strategies, and implementing effective early prevention.

The integrity of various physiological processes relies heavily on the availability of manganese, an essential trace mineral with multifaceted regulatory properties. It not only participates in antioxidant defense and the activation of various enzymes but also serves as a cofactor for manganese superoxide dismutase (MnSOD), thereby playing a crucial role in maintaining cellular homeostasis (11-13). Manganese exerts significant effects on tumorigenesis and progression (14). For instance, manganese deficiency promotes mitochondrial-mediated cell death by activating the tumor suppressor protein p53 (14). In oral squamous cell carcinoma, manganese activates the YAP/TAZ signaling pathway, enhancing ferroptosis (15). Recent studies indicate manganese’s critical role in intestinal homeostasis. Dietary manganese exacerbates colonitis-associated inflammation and tissue damage in mice, whereas appropriate supplementation enhances intestinal barrier function, mitigates increased intestinal permeability, and reduces colonic injury (16,17). Furthermore, the gut microbiota can mitigate neuroinflammation during manganese exposure by inhibiting brain NLRP3 inflammasomes (18). Despite the recognized involvement of manganese in modulating gut-associated inflammatory milieus and neoplastic growth, the precise signaling cascades it coordinates during CRC pathogenesis remain inadequately defined.

To investigate the potential role of manganese metabolism-related genes (MMRGs) in the development of CRC, this study integrated data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases to systematically screen MMRGs closely associated with CRC prognosis. Based on this, a prognostic risk model was constructed to predict patient survival. To further investigate disease heterogeneity, consensus clustering was implemented to stratify patients into molecular subtypes based on the messenger RNA (mRNA) expression patterns of MMRGs. An in-depth assessment of the immune cell landscapes across these stratified subtypes was performed, uncovering a distinct relationship between the proposed classification and the heterogeneous nature of the tumor microenvironment. To confront the clinical challenge posed by disparate survival outcomes in CRC patients under standard care, this study provides a comprehensive evaluation of the molecular factors driving this heterogeneity. We present this article in accordance with the TRIPOD reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-1-1005/rc).

Methods

Data source

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study integrated transcriptomic data from the University of California, Santa Cruz (UCSC) Xena database (https://xena.ucsc.edu/) for colon adenocarcinoma (COAD; 41 normal samples and 473 tumor samples) and rectal adenocarcinoma (READ; 10 normal samples and tumor samples: 167) as the training set for the CRC, concurrently acquiring corresponding clinical and copy number variation (CNV) data for each sample. The GEO (https://www.ncbi.nlm.nih.gov/geo/) database provided the GSE12945 (62 tumor samples) and GSE29621 (65 tumor samples) datasets as independent validation sets. Based on research by Mariathasan et al. (19), we sourced transcriptomic profiles from the IMvigor210 study, which includes bladder urothelial carcinoma patients characterized by diverse clinical outcomes following PD-1 pathway inhibition. In addition, we performed a search in the GeneCards database (https://www.genecards.org/) for “manganese metabolism” and identified 1,930 MMRGs related to this process after applying a protein coding gene filter.

Preliminary screening of CRC prognostic genes

To identify MMRGs as prognostic markers for CRC, we first performed differential expression analysis on normal and tumor tissue samples from the training set using the limma package (v3.62.2). The differentially expressed genes (DEGs) were defined as those candidates fulfilling the predefined selection criteria of an absolute log2 fold change exceeding 1 and an adjusted P value below 0.05. The WGCNA package (v1.73) was employed to perform weighted gene co-expression network analysis (WGCNA) on the training set data. To ensure the construction of a robust scale-free network, a power of 18 and a fit index of 0.85 were established as thresholds, enabling the aggregation of genes with highly correlated expression into unified functional modules. To isolate highly relevant co-expression clusters, we assessed the associations with CRC by computing the Pearson’s r values between module eigengenes and clinical phenotypes. Subsequently, integrating MMRGs, DEGs and genes from WGCNA key modules, univariate Cox analysis (P value <0.05) was performed using the survival package (v3.5-8) to identify candidate prognostic genes for CRC.

Construction and validation of a CRC prognostic model

To further identify CRC prognostic genes, screening was performed using least absolute shrinkage and selection operator (LASSO) regression analysis (glmnet, v4.1-9) and multivariate Cox regression analysis (survival, v3.5-8). The final prognostic framework was established through a stepwise regression strategy, which prioritized the most informative genes into a parsimonious model to ensure high survival-predictive accuracy. The risk score calculation method is as follows:

Riskscore=∑(Geneexpressioni×Coefficienti)[1]

To verify the reliability of the prognostic model, Kaplan-Meier (K-M) curves and receiver operating characteristic (ROC) metrics were generated for both cohorts. Furthermore, the model’s robustness was further assessed by mapping survival status in relation to the risk score distribution.

Constructing a nomogram based on clinical characteristics

A prognostic scoring system for CRC was established by integrating MMRGs-based signatures with relevant clinical parameters through univariate and multivariate Cox proportional hazards modeling. We assessed the net benefit and consistency of the prognostic score by generating calibration curves and decision curve analysis (DCA), ensuring its robust performance in predicting patient outcomes at multiple time points (1, 3, and 5 years). ROC curves were employed to compare the predictive accuracy of the risk score with individual clinical variables across 1-, 3-, and 5-year intervals. Furthermore, the generalizability and robustness of the prognostic nomogram were rigorously assessed through external validation in an independent cohort (GSE12945).

Characterizing the immunological landscape and predicting sensitivity to immunotherapy

The immunological landscape and functional infiltration profiles within distinct cohorts were dissected by integrating various computational approaches, including MCP-counter, single-sample gene set enrichment analysis (ssGSEA) and CIBERSORT, to ensure a robust evaluation of the cellular tumor microenvironment architecture. Clinical responsiveness to anti-PD-1 therapy was assessed using the IMvigor210 dataset by grouping individuals with complete response (CR) or partial response (PR) into the responder (R) cohort. Conversely, patients exhibiting stable disease (SD) or progressive disease (PD) were designated as non-responders (NRs). This allowed for a rigorous analysis of how different risk groups respond to immune checkpoint blockade. Immunophenoscore (IPS) data for CRC were retrieved from the TCIA database to predict response potential to cytotoxic T lymphocyte-associated antigen-4 (CTLA-4) and anti-PD-1 therapies across different risk cohorts.

Tumor mutational burden (TMB) and drug sensitivity analysis

Genomic variations in the discovery set were quantified to pinpoint the leading 20 most frequently mutated genes across the high- and low-risk subgroups. Utilizing the maftools (v2.22.0) framework, we visualized the distribution and spectrum of these mutations to illustrate the mutational heterogeneity between the two cohorts. The pRRophetic package (v0.5) was used to quantify the half-maximal inhibitory concentration (IC50) for CRC-associated chemotherapeutic agents. To explore the pharmacogenomic landscape, we retrieved drug-gene interaction data from the CellMiner platform. The relationship between the transcriptional levels of our prognostic markers and the potency of various antineoplastic agents was quantified using Spearman’s rank correlation.

Molecular typing of CRC

Distinct molecular subtypes were established within the CRC cohort by implementing a consensus clustering framework (ConsensusClusterPlus, v1.70.0) drawing on the identified gene signatures. We assessed the stability of these stratifications through rigorous analysis of the delta area plots and cumulative distribution function (CDF) to ensure clustering consistency.

Reverse transcription-quantitative polymerase chain reaction (RT-qPCR)

For the in vitro experiments, human CRC cells (HCT116) and normal colonic epithelial cells (CACO2) were utilized. Both cell lines were maintained in DMEM containing 10% fetal bovine serum under sterile, aseptic conditions. Following total RNA isolation using TRIzol reagent, reverse transcription was performed with PrimeScript^TM RT Master Mix to synthesize cDNA. After verifying nucleic acid concentrations via a NanoDrop 2000 spectrophotometer, RT-qPCR was executed on the QuantStudio 5 platform. Relative gene expression was determined using the 2^−ΔΔCT method, with GAPDH acting as the internal control and primer sequences listed in Table 1.

Data analysis

R software (v4.4.0) was applied to data analysis and visualization. Wilcoxon test was utilized for inter-group comparison. The strength of correlations was quantified via Pearson correlation metrics, with the threshold for significance rigorously maintained at P<0.05 throughout the study.

Results

Preliminary screening of CRC prognostic genes

This study first integrated TCGA-COAD and TCGA-READ samples as the CRC training dataset, and employed principal component analysis (PCA) to visualise the overall structure of the dataset (Figure 1A). Subsequently, differential expression analysis identified 3,692 DEGs (Figure 1B). Further, WGCNA was applied with thresholds β =18 and R²=0.85 to construct a scale-free co-expression network (Figure 1C) and generate a hierarchical clustering tree (Figure 1D). To refine the co-expression network, highly correlated modules were integrated, leading to the identification of MEgreen and MEyellow as the primary modules linked to CRC progression. Quantitative analysis revealed that MEgreen (cor =0.8) and MEyellow (cor =0.52) possessed the highest association metrics with the clinical features (Figure 1E). To verify the reliability of the identified modules, we calculated the correlation between MM and GS. The results revealed that genes central to MEgreen and MEyellow exhibited a high degree of consistency with their clinical importance, with correlation coefficients of 0.77 and 0.39, respectively (Figure 1F). Intersecting the 2,328 genes from these modules with DEGs and MMRGs yielded 112 common genes (Figure 1G). Gene enrichment analysis (Figure 1H) revealed enrichment in pathways including TNF signalling, cellular senescence, and cytokine-cytokine receptor interaction. Univariate Cox regression analysis of these common genes identified 17 candidate prognostic genes for CRC (Figure 1I), among which human neutrophil elastase (ELANE), SELENOP, sulfotransferase family 1B member 1 (SULT1B1), XDH, and RIMKLA exhibited significantly higher expression levels in the normal group compared to the tumor group (Figure 1J). Correlation analysis (Figure 1K) revealed positive correlations between matrix metallopeptidase 10 (MMP10) and CXCL1 with MMP3. Furthermore, the chromosomal CNV landscape of the candidate prognostic genes is illustrated (Figure 1L).

Figure 1 Preliminary screening of CRC prognostic genes. (A) PCA plot of CRC training set samples. (B) Volcano plot of DEGs. (C) CRC WGCNA analysis (β =18, R²=0.85). (D) Hierarchical clustering-derived co-expression module dendrogram. (E) Heatmap of module-trait relationships illustrating the associations between co-expression gene clusters and CRC clinical characteristics. (F) Scatterplots of significant gene correlations within MEgreen and MEyellow modules. (G) Identification of candidate genes through the overlap of WGCNA-derived modules, DEGs, and MMRGs. (H) Gene enrichment analysis of shared genes. (I) Prognostic gene prioritization using univariate Cox analysis in CRC. (J) Expression profiles of the prognostic signature genes in various sample subsets. (K) Correlation analysis among CRC candidate prognostic genes. (L) Chromosomal CNV distribution map of 17 candidate prognostic genes. ***, P<0.001. BP, CC, CNV, copy number variation; COAD, colon adenocarcinoma; CRC, colorectal cancer; DEGs, differentially expressed genes; KEGG, Kyoto Encyclopedia of Genes and Genomes; MF, MMRGs, manganese metabolism-related genes; PCA, principal component analysis; READ, rectal adenocarcinoma; TNF, WGCNA, weighted gene co-expression network analysis.

Construction and validation of a CRC prognostic model

Based on 17 candidate prognostic genes for CRC, 13 genes were selected via LASSO regression analysis (Figure 2A,2B). Consequently, six genes were determined as independent prognostic predictors for CRC using multivariable Cox modeling (Figure 2C). The key CRC risk prognostic model derived from this gene set employs the following risk scoring formula:

Riskscore=0.331×SNAI1+0.213×TERT−0.147×ELANE−0.730×CCNF−0.084×SULT1B1−0.076×MMP10[2]

Figure 2 Construction and validation of a CRC prognostic model. (A) Shrinkage trajectories of LASSO regression coefficients. (B) Identification of the optimal tuning parameter λ based on the partial likelihood deviance from cross-validation analysis. (D) Predictive efficacy of the model across the discovery cohort (left) and independent validation sets (GSE12945, center; GSE29621, right) evaluated via ROC curves. (E) Survival differentiation illustrated by K-M curves and (F) risk score stratification patterns. (G) Expression level comparison of the six prognostic genes between tumor and normal groups. (H) Using RT-qPCR technology, the relative mRNA expression levels of six MMRGs were detected in human CRC cells HCT116 and human normal colonic epithelial cells CACO2. *, P<0.05; **, ***, P<0.001. AUC, area under the curve; CI, confidence interval; CRC, colorectal cancer; K-M, Kaplan-Meier; LASSO, least absolute shrinkage and selection operator; MMRGs, manganese metabolism-related genes; mRNA, messenger RNA; ROC, receiver operating characteristic; RT-qPCR, reverse transcription-quantitative polymerase chain reaction.

The CRC cohort was stratified into discrete high- and low-risk subgroups by utilizing the median risk score as the definitive classification threshold. Both the differential mRNA abundance (Figure S1A) and the independent survival-forecasting capability (Figure S1B) were confirmed for all six prognostic genes. The discriminative power of our prognostic signature was confirmed via time-dependent ROC analysis, which demonstrated consistently high accuracy across both the training and validation populations (Figure 2D). K-M estimates demonstrated a notably superior prognosis for the low-risk cohort, whose survival duration was significantly extended relative to the high-risk subgroup (Figure 2E). A survival status versus risk score distribution diagram visually illustrates the outcome disparities between risk groups (Figure 2F). Furthermore, differential expression analysis showed that ELANE and SULT1B1 exhibited significantly higher expression levels in the normal group than in the tumor group, while the remaining four genes demonstrated the opposite expression trend (Figure 2G). The RT-qPCR data demonstrated a high degree of consistency with our computational predictions, thereby reinforcing the validity of the identified prognostic genes (Figure 2H).

Construction of a CRC prognostic nomogram

The distribution of risk scores across various clinical strata indicated a significant escalation alongside tumor advancement. As illustrated in Figure 3A, the risk score exhibited a significant upward trend as the tumor (T), node (N), metastasis (M), and clinical stages progressed, with high-risk scores predominantly clustered in the cohorts representing more aggressive phenotype. Univariate Cox modeling (Figure 3B) demonstrated that the calculated risk score and several standard clinical markers (including age and TNM categories) possessed robust prognostic value, showing significant correlations with patient outcomes. The predictive autonomy of the risk score, patient age, pathological stage, and distant metastasis (M staging) was further substantiated through a multivariable Cox proportional hazards model, identifying them as self-standing prognostic indicators (Figure 3C). For the purpose of providing an individualized estimation of clinical outcomes, an integrative nomogram was constructed by amalgamating the signature-derived risk scores with independent clinicopathological variables in the CRC cohort (Figure 3D). The probabilistic alignment between predicted and observed outcomes was substantiated by calibration plots, while DCA evidenced the significant clinical net benefit of the model in forecasting long-term survival at 1-, 3-, and 5-year intervals (Figure 3E,3F). As illustrated in Figure 3G, the risk score consistently outperformed individual clinical variables across 1-, 3-, and 5-year intervals, demonstrating better and more stable predictive accuracy compared to age and M-stage. Finally, external validation of the nomogram was performed in the GSE12945 set (Figure S2).

Figure 3 Construction and validation of a CRC prognostic nomogram. (A) Comparison of risk scores for various clinical characteristics. (B) Univariate Cox analysis. (C) Multivariate Cox analysis. (D) Nomogram drawing on clinical characteristics and risk score. (E) DCA curve. (F) Calibration curve of nomogram. (G) ROC curves evaluating the predictive accuracy of the risk score and clinical parameters. CI, confidence interval; CRC, colorectal cancer; DCA, decision curve analysis; M, metastasis; N, node; ROC, receiver operating characteristic; T, tumor.

Immune infiltration analysis

ssGSEA showed higher abundances of B cells, DCs and Th1 cells in the low-risk cohort than in the high-risk cohort (Figure 4A). Based on the CIBERSORT deconvolution results, the low-risk cohort was characterized by a notably higher abundance of plasma cells, resting memory CD4⁺ T cells, and neutrophils compared to the high-risk group (Figure 4B). A panel of immune checkpoint genes was found to be more highly expressed in patients classified as low risk (Figure 4C). Patients stratified into the low-risk group were characterized by a more favorable immunophenotype, indicating an enhanced likelihood of responding to dual-checkpoint inhibition (anti-PD-1/CTLA-4) compared to the high-risk group (Figure 4D). Subsequent validation in the IMvigor210 cohort demonstrated a significantly higher response rate to anti-PD-1 therapy in the low-risk group (Figure 4E), with the NR group exhibiting significantly higher risk scores than the R group (Figure 4F). Log-rank test results (Figure 4G) revealed that low-risk patients experienced significantly enhanced survival outcomes relative to their high-risk counterparts. Furthermore, correlation analysis indicated that plasma cell immune infiltration levels positively correlated with SULT1B1 gene expression and negatively correlated with snail family transcriptional repressor 1 (SNAI1) expression (Figure 4H).

Figure 4 Immune infiltration analysis. (A) ssGSEA assessment of immune-cell enrichment. (B) Profiling the relative abundance of distinct immune cell subsets based on the CIBERSORT algorithm. (C) Inter-group comparison of immune checkpoint gene expression (high vs. low risk). (D) IPS violin plots. (E) Frequency of clinical response to anti-PD-1 therapy in different risk cohorts. (F) Risk score distribution patterns across different immunotherapeutic response categories (R vs. NR). (G) K-M survival curves assessing patient survival rates across risk groups. (H) Association of immune cell infiltration subsets with the mRNA expression levels of prognostic markers. ns, not significant; *, P<0.05; **, P<0.01; ***, P<0.001. APC, adenomatous polyposis coli; DCs; IPS, immunophenoscore; K-M, Kaplan-Meier; mRNA, messenger RNA; NR, non-responder; PD-1, programmed death-1; R, responder; ssGSEA, single-sample gene set enrichment analysis.

TMB and drug sensitivity analysis

The genomic landscape of somatic alterations was delineated across the stratified risk cohorts through TMB evaluation (Figure 5A). This analysis uncovered a distinct enrichment of tumor protein p53 (TP53) mutations in high-risk individuals (35%) relative to the low-risk subgroup (27%). As shown in Figure 5B, high-risk patients exhibited higher predicted IC50 for 5-fluorouracil, trametinib and erlotinib—suggesting decreased sensitivity—whereas pazopanib displayed an inverse association with risk group. Additionally, correlation analysis (Table S1) revealed positive correlations between telomerase reverse transcriptase (TERT) expression and sensitivity to oxaliplatin and irinotecan, between cyclin F (CCNF) expression and sensitivity to raltitrexed, and between SNAI1 expression and sensitivity to lenvatinib (Figure 5C).

Figure 5 TMB and drug sensitivity analysis. (A) Distribution of the top 20 somatic mutations across stratified risk groups (left panel: high-risk; right panel: low-risk). (B) Discrepancy in IC₅₀ metrics for four standard-of-care agents among CRC risk groups. (C) Evaluating the link between the prognostic gene signature and chemosensitivity via pharmacogenomic analysis. CRC, colorectal cancer; IC₅₀, half-maximal inhibitory concentration; TMB, tumor mutational burden.

Molecular subtypes of CRC

This study classified CRC into two molecular subtypes (Group 1 and Group 2) based on prognostic genes using consensus clustering (Figure 6A). The reliability of the clustering results was confirmed by evaluating the relative change in the area under the CDF curve (Figure 6B,6C), ensuring an optimal and robust partition of the CRC samples. Gene expression analysis revealed significant overexpression of SULT1B1 and MMP10 in Group 2 (Figure 6D). K-M survival analysis demonstrated significantly higher survival rates in Group 2 patients compared to Group 1 (Figure 6E). Subsequent enrichment analysis of DEGs between groups revealed significant pathway differences, including the JAK-STAT signaling pathway and IL-17 signaling pathway (Figure 6F). Evaluation of the immunological landscape via MCP-counter (Figure 6G) underscored notable variations in the recruitment of neutrophils, fibroblasts, and CD8⁺ T cells, all of which showed statistical divergence between the two groups. Immune cell infiltration analysis using the CIBERSORT algorithm showed higher infiltration levels of T cells CD8 and Tregs in Group 1 compared to Group 2 (Figure 6H). Immune cell infiltration analysis based on the ssGSEA algorithm indicated higher infiltration levels of DCs and CCR in Group 2 (Figure 6I).

Figure 6 Molecular subtypes of CRC. (A) Heatmap of consensus clustering (k =2). (B) CDF for cluster stability assessment. (C) Assessment of clustering stability via delta area plot to identify the ideal number of molecular subcategories. (D) Box plot illustrating the heterogeneous mRNA abundance of prognostic candidates within stratified risk tiers. (E) Discrepancies in clinical outcomes and survival probability among different clusters, as determined by log-rank analysis. (F) KEGG and GO enrichment analysis. (G) Immune landscape by MCP-counter. (H) Immune composition by CIBERSORT. (I) Immune enrichment scores from ssGSEA. ns, not significant; *, P<0.05; **, P<0.01; ***, P<0.001. APC, adenomatous polyposis coli; BP, CC, CCR, CDF, cumulative distribution function; CRC, colorectal cancer; DCs, GO, Gene Ontology; IFN, KEGG, Kyoto Encyclopedia of Genes and Genomes; MCP, MF, mRNA, messenger RNA; pDCs, ssGSEA, single-sample gene set enrichment analysis.

Discussion

This study leveraged transcriptomic data from patients in the TCGA and GEO databases to investigate the potential role of MMRGs in CRC. We successfully identified six prognosis-associated genes and constructed an effective prognostic risk model. This research not only provides new insights into the molecular mechanisms of CRC but also demonstrates the clinical utility of MMRGs-based prognostic models in guiding patient treatment decisions.

This study identified six MMRGs as prognostic genes for CRC (SNAI1, ELANE, CCNF, SULT1B1, MMP10, and TERT) that serve as independent predictors of prognosis in CRC patients. The six MMRGs effectively stratified CRC patients into high-risk and low-risk groups, with the high-risk group exhibiting shorter overall survival and poorer survival rates. A nomogram was developed by integrating various clinical variables with the risk score, providing a comprehensive perspective on the prognostic capability of these genes. These prognostic genes are closely associated with tumor initiation and progression. SNAI1, reported as a key driver of epithelial-mesenchymal transition (EMT), plays a crucial role in tumorigenesis, invasion, and metastasis (20,21). Research indicates that abnormal expression of SNAI1 in intestinal epithelium promotes CRC development by intensifying EMT processes and synergistically activating the Wnt/β-catenin signaling pathway (22). Furthermore, tigecycline, as a therapeutic strategy, has been found to effectively inhibit EMT by targeting SNAI1 and β-catenin. This manifests as upregulation of the epithelial marker E-cadherin alongside decreased expression of pluripotency and mesenchymal markers, suggesting its potential in reversing tumor malignant phenotypes (23). On the other hand, inhibiting SNAI1 weakens the migration and invasive capabilities of cancer cells (24). MMP10 is primarily expressed by myeloid cells in human colitis, with its expression levels positively correlated with disease severity (25). In CRC, MMP10 overexpression significantly enhances tumor cell invasiveness (26,27). Conversely, SULT1B1 shows reduced expression in CRC tissues, with its low expression associated with poor patient prognosis (28). TERT shows elevated expression in CRC tumor tissues (29). It forms a complex with the zinc finger E-box binding domain 1 (ZEB1) transcription factor, binds to the E-cadherin promoter, inhibits its expression, and induces EMT (30). ELANE, released by neutrophils in the tumor microenvironment, can proteolytically release the CD95 death domain. This domain interacts with histone H1 subtypes, thereby selectively inducing cancer cell death (31). CCNF, a key regulator of cell death, angiogenesis, and EMT, exhibits abnormal expression in multiple cancers (32). For instance, CCNF expression is upregulated in ovarian cancer, and its high expression correlates with poor prognosis in patients (33). In breast cancer, CCNF overexpression not only correlates with increased tumor malignancy but also elevates the risk of metastatic recurrence (34). Notably, while most prognostic genes are associated with the EMT process, whether they drive CRC progression by regulating manganese metabolism and EMT actions warrants further investigation.

As an integral part of the tumor niche, infiltrating immune subsets play a decisive role in modulating cancer development and the efficacy of clinical interventions (35). This study’s immune infiltration analysis revealed that B-cell infiltration levels were significantly higher in low-risk CRC patients compared to high-risk patients. Previous research indicates that B cells are one of the primary infiltrating immune cells in CRC, and their enrichment is closely associated with improved treatment response and survival outcomes in patients (36). Specifically, increased CD20⁺ B cell density has been confirmed as an independent predictor of favorable prognosis (37). Further IPS analysis revealed that low-risk patients demonstrated markedly superior responsiveness to CTLA-4 and PD-1 antibody therapies compared to high-risk patients. PD-1, as a T-cell surface receptor, transmits inhibitory signals upon binding its ligand PD-L1, thereby weakening T-cell tumor killing capabilities (38,39). CTLA-4 is another immunosuppressive receptor expressed on T-cell surfaces. Its inhibitors can reverse T-cell immunosuppression, promote cell activation and proliferation, thereby enhancing antitumor immune responses (40,41).

CRC exhibits high mutation rates, with significant heterogeneity in tumor immune responses and metabolic characteristics among patients (42). The most common molecular pathway in CRC is driven by inactivation of the adenomatous polyposis coli (APC) gene, whose mutations promote adenoma formation and progression (43). Consistent with the findings of this study, the APC mutation rate in high-risk patients (41%) was significantly higher than in low-risk patients (37%), suggesting that targeted therapy against this gene may yield greater benefits for high-risk patients. Additionally, this study observed that the IC50 value of 5-fluorouracil was significantly higher in the high-risk patient group compared to the low-risk group, suggesting stronger drug resistance to this agent in high-risk patients. This phenomenon may be related to differences in TP53 mutation frequency between the two groups: the TP53 mutation rate was 35% in the high-risk group, significantly higher than the 27% rate in the low-risk group (44,45). TP53, a key tumor suppressor gene, is activated under cellular stress conditions to regulate multiple biological processes including cell cycle arrest, apoptosis initiation, and senescence induction. This mechanism exerts tumor suppression effects and enhances cellular sensitivity to chemotherapeutic agents (46,47). However, TP53 mutations not only result in the loss of tumor suppressor function in wild-type p53 (wtp53), but certain mutant p53 (mutp53) forms may also gain functional advantages, thereby promoting malignant tumor progression (48). TP53 mutations mediate chemotherapy resistance through multiple pathways, including enhanced drug efflux and metabolic inactivation, promotion of DNA damage repair, and suppression of apoptosis (49,50). Additionally, studies have reported that ubiquitin C-terminal hydrolase L3 (UCHL3) may serve as a potential therapeutic target by reversing elevated glycolysis in TP53-mutant CRC and enhancing cellular sensitivity to 5-fluorouracil chemotherapy (51). Therefore, TP53 mutations likely represent a key mechanism contributing to reduced sensitivity to 5-fluorouracil in high-risk patients. Whether the prognostic genes identified in this study can influence CRC progression by regulating TP53 mutations and manganese metabolism pathways warrants further investigation.

While this study offers a holistic characterization of MMRGs in CRC, certain caveats must be acknowledged. Firstly, although molecularly-defined CRC subgroups displayed distinct immunological landscapes, their predictive utility as clinical biomarkers for treatment response remains to be empirically substantiated through prospective trials. Secondly, the exclusive reliance on open-access repositories may introduce retrospective bias and limit population diversity. Consequently, large-scale, multi-institutional cohort studies are essential to ensure the broad applicability of these findings. Furthermore, while the CellMiner database was utilized to identify potential therapeutic candidates, several limitations must be acknowledged. Drug sensitivity estimations derived from in vitro cell line data are often subject to cell-specific biases and may not accurately reflect in vivo drug responses. Additionally, systemic factors such as pharmacokinetics, liver-mediated drug metabolism, and the heterogeneity of the intestinal absorption barrier cannot be fully simulated through cell-based assays. Lastly, the precise regulatory networks underpinning CRC malignancy are still partially obscure. In particular, the intricate crosstalk between manganese metabolic flux, the EMT process, and the identified prognostic signatures warrants further mechanistic dissection to clarify their synergistic contributions to tumor evolution. Moving forward, it is essential to empirically substantiate the biological roles of these prognostic signatures and lead compounds using both in vitro and in vivo CRC models. Furthermore, leveraging spatially resolved transcriptomics will provide deeper mechanistic insights into their intramodular interactions, ultimately facilitating the engineering of diagnostic platforms for enhanced early-stage malignancy screening.

Conclusions

MMRGs are not only related to CRC’s tumor microenvironment, but also may be involved in regulating EMT processes. Modulating the activity of these manganese-related genes may pave a new avenue for precision oncology, offering fresh perspectives for CRC clinical treatment strategies.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-1-1005/rc

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-1-1005/prf

Funding: This work was supported by grant from Natural Science Foundation of Fujian Province (No. 2022J05300), and Huzhou Science and Technology Plan Project (No. 2025GYB59).

Conflicts of Interest: Both authors have completed the ICMJE uniform disclosure form (https://jgo.amegroups.com/article/view/10.21037/jgo-2025-1-1005/coif). P.W. received funding from Natural Science Foundation of Fujian Province (grant No. 2022J05300) and Huzhou Science and Technology Plan Project (No. 2025GYB59). The other author has 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. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

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. Siegel RL, Wagle NS, Cercek A, et al. Colorectal cancer statistics, 2023. CA Cancer J Clin 2023;73:233-54. [Crossref] [PubMed]
2. Duan B, Zhao Y, Bai J, et al. Colorectal Cancer: An Overview. In: Morgado-Diaz JA, editor. Gastrointestinal Cancers. Brisbane (AU): Exon Publications; 2022.
3. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
4. Jacobsson M, Wagner V, Kanneganti S. Screening for Colorectal Cancer. Surg Clin North Am 2024;104:595-607. [Crossref] [PubMed]
5. Forster S, Radpour R. Molecular Immunotherapy: Promising Approach to Treat Metastatic Colorectal Cancer by Targeting Resistant Cancer Cells or Cancer Stem Cells. Front Oncol 2020;10:569017. [Crossref] [PubMed]
6. Ullah F, Pillai AB, Omar N, et al. Early-Onset Colorectal Cancer: Current Insights. Cancers (Basel) 2023;15:3202. [Crossref] [PubMed]
7. Eng C, Yoshino T, Ruíz-García E, et al. Colorectal cancer. Lancet 2024;404:294-310. [Crossref] [PubMed]
8. Ayala-de Miguel C, Jiménez-Castro J, Sánchez-Vegas A, et al. Third-line treatment and beyond in metastatic colorectal cancer: What do we have and what can we expect? Crit Rev Oncol Hematol 2024;202:104454. [Crossref] [PubMed]
9. Abedizadeh R, Majidi F, Khorasani HR, et al. Colorectal cancer: a comprehensive review of carcinogenesis, diagnosis, and novel strategies for classified treatments. Cancer Metastasis Rev 2024;43:729-53. [Crossref] [PubMed]
10. Singh M, Morris VK, Bandey IN, et al. Advancements in combining targeted therapy and immunotherapy for colorectal cancer. Trends Cancer 2024;10:598-609. [Crossref] [PubMed]
11. Zhang K, Qi C, Cai K. Manganese-Based Tumor Immunotherapy. Adv Mater 2023;35:e2205409. [Crossref] [PubMed]
12. Xiong Z, Zhang Z, Cheng S, et al. A novel defined manganese metabolism-related gene signature for predicting the prognosis of pancreatic ductal adenocarcinoma. Oncol Lett 2025;30:436. [Crossref] [PubMed]
13. Liu Y, Ye H, Zhang R, et al. A manganese metabolism-related gene signature stratifies prognosis and immunotherapy efficacy in kidney cancer. Discov Oncol 2025;16:1242. [Crossref] [PubMed]
14. Tinkov AA, Paoliello MMB, Mazilina AN, et al. Molecular Targets of Manganese-Induced Neurotoxicity: A Five-Year Update. Int J Mol Sci 2021;22:4646. [Crossref] [PubMed]
15. Wu J, Minikes AM, Gao M, et al. Intercellular interaction dictates cancer cell ferroptosis via NF2-YAP signalling. Nature 2019;572:402-6. [Crossref] [PubMed]
16. Choi EK, Aring L, Das NK, et al. Impact of dietary manganese on experimental colitis in mice. FASEB J 2020;34:2929-43. [Crossref] [PubMed]
17. Choi EK, Rajendiran TM, Soni T, et al. The manganese transporter SLC39A8 links alkaline ceramidase 1 to inflammatory bowel disease. Nat Commun 2024;15:4775. [Crossref] [PubMed]
18. Wang H, Yang F, Xin R, et al. The gut microbiota attenuate neuroinflammation in manganese exposure by inhibiting cerebral NLRP3 inflammasome. Biomed Pharmacother 2020;129:110449. [Crossref] [PubMed]
19. Mariathasan S, Turley SJ, Nickles D, et al. TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells. Nature 2018;554:544-8. [Crossref] [PubMed]
20. Lv XL, Peng QL, Wang XP, et al. Snail family transcriptional repressor 1 radiosensitizes esophageal cancer via epithelial-mesenchymal transition signaling: From bioinformatics to integrated study. World J Gastrointest Oncol 2025;17:97644. [Crossref] [PubMed]
21. E H. SNAI1 promotes epithelial-mesenchymal transition and maintains cancer stem cell-like properties in thymic epithelial tumors through the PIK3R2/p-EphA2 Axis. J Exp Clin Cancer Res 2024;43:324. [Crossref] [PubMed]
22. Qing F, Xue J, Sui L, et al. Intestinal epithelial SNAI1 promotes the occurrence of colorectal cancer by enhancing EMT and Wnt/β-catenin signaling. Med Oncol 2023;41:34. [Crossref] [PubMed]
23. Ruiz-Malagón AJ, Rodríguez-Sojo MJ, García-García J, et al. Tigecycline suppresses colon cancer stem cells and impairs tumor engraftment by targeting SNAI1-regulated epithelial-mesenchymal transition. Acta Pharmacol Sin 2026;47:222-41. [Crossref] [PubMed]
24. Wang H, Chirshev E, Hojo N, et al. The Epithelial-Mesenchymal Transcription Factor SNAI1 Represses Transcription of the Tumor Suppressor miRNA let-7 in Cancer. Cancers (Basel) 2021;13:1469. [Crossref] [PubMed]
25. He L, Kang Q, Chan KI, et al. The immunomodulatory role of matrix metalloproteinases in colitis-associated cancer. Front Immunol 2022;13:1093990. [Crossref] [PubMed]
26. Klupp F, Neumann L, Kahlert C, et al. Serum MMP7, MMP10 and MMP12 level as negative prognostic markers in colon cancer patients. BMC Cancer 2016;16:494. [Crossref] [PubMed]
27. Derakhshani A, Javadrashid D, Hemmat N, et al. Identification of Common and Distinct Pathways in Inflammatory Bowel Disease and Colorectal Cancer: A Hypothesis Based on Weighted Gene Co-Expression Network Analysis. Front Genet 2022;13:848646. [Crossref] [PubMed]
28. Lian W, Jin H, Cao J, et al. Identification of novel biomarkers affecting the metastasis of colorectal cancer through bioinformatics analysis and validation through qRT-PCR. Cancer Cell Int 2020;20:105. [Crossref] [PubMed]
29. Kienzl P, Deloria AJ, Hunjadi M, et al. Telomere transcripts act as tumor suppressor and are associated with favorable prognosis in colorectal cancer with low proliferating cell nuclear antigen expression. Cell Oncol (Dordr) 2025;48:239-47. [Crossref] [PubMed]
30. Zou Y, Cong YS, Zhou J. Implications of telomerase reverse transcriptase in tumor metastasis. BMB Rep 2020;53:458-65. [Crossref] [PubMed]
31. Cui C, Chakraborty K, Tang XA, et al. Neutrophil elastase selectively kills cancer cells and attenuates tumorigenesis. Cell 2021;184:3163-3177.e21. [Crossref] [PubMed]
32. Gao X, Bu H, Ge J, et al. A Comprehensive Analysis of the Prognostic, Immunological and Diagnostic Role of CCNF in Pan-cancer. J Cancer 2023;14:2431-42. [Crossref] [PubMed]
33. Li Y, Guo H, Wang Z, et al. Cyclin F and KIF20A, FOXM1 target genes, increase proliferation and invasion of ovarian cancer cells. Exp Cell Res 2020;395:112212. [Crossref] [PubMed]
34. Liu Y, Pan B, Qu W, et al. Systematic analysis of the expression and prognosis relevance of FBXO family reveals the significance of FBXO1 in human breast cancer. Cancer Cell Int 2021;21:130. [Crossref] [PubMed]
35. Zhao F, Chen M, Wu T, et al. Integration of single-cell and bulk RNA sequencing to identify a distinct tumor stem cells and construct a novel prognostic signature for evaluating prognosis and immunotherapy in LUAD. J Transl Med 2025;23:222. [Crossref] [PubMed]
36. Xia J, Xie Z, Niu G, et al. Single-cell landscape and clinical outcomes of infiltrating B cells in colorectal cancer. Immunology 2023;168:135-51. [Crossref] [PubMed]
37. Berntsson J, Nodin B, Eberhard J, et al. Prognostic impact of tumour-infiltrating B cells and plasma cells in colorectal cancer. Int J Cancer 2016;139:1129-39. [Crossref] [PubMed]
38. Qiu J, Cheng Z, Jiang Z, et al. Immunomodulatory Precision: A Narrative Review Exploring the Critical Role of Immune Checkpoint Inhibitors in Cancer Treatment. Int J Mol Sci 2024;25:5490. [Crossref] [PubMed]
39. 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]
40. Pavelescu LA, Enache RM, Roşu OA, et al. Predictive Biomarkers and Resistance Mechanisms of Checkpoint Inhibitors in Malignant Solid Tumors. Int J Mol Sci 2024;25:9659. [Crossref] [PubMed]
41. Ahmed H, Mahmud AR, Siddiquee MF, et al. Role of T cells in cancer immunotherapy: Opportunities and challenges. Cancer Pathog Ther 2023;1:116-26. [Crossref] [PubMed]
42. van den Driest L, Johnson CH, Rattray NJW, et al. Development of an Accessible Gene Expression Bioinformatics Pipeline to Study Driver Mutations of Colorectal Cancer. Altern Lab Anim 2022;50:282-92. [Crossref] [PubMed]
43. Guinney J, Dienstmann R, Wang X, et al. The consensus molecular subtypes of colorectal cancer. Nat Med 2015;21:1350-6. [Crossref] [PubMed]
44. Hu J, Cao J, Topatana W, et al. Targeting mutant p53 for cancer therapy: direct and indirect strategies. J Hematol Oncol 2021;14:157. [Crossref] [PubMed]
45. Nzitakera A, Uwamariya D, Kato H, et al. TP53 mutation status and consensus molecular subtypes of colorectal cancer in patients from Rwanda. BMC Cancer 2024;24:1266. [Crossref] [PubMed]
46. Kastenhuber ER, Lowe SW. Putting p53 in Context. Cell 2017;170:1062-78. [Crossref] [PubMed]
47. Zhou Y, Fan X, Jiao T, et al. SIRT6 as a key event linking P53 and NRF2 counteracts APAP-induced hepatotoxicity through inhibiting oxidative stress and promoting hepatocyte proliferation. Acta Pharm Sin B 2021;11:89-99. [Crossref] [PubMed]
48. Mantovani F, Collavin L, Del Sal G. Mutant p53 as a guardian of the cancer cell. Cell Death Differ 2019;26:199-212. [Crossref] [PubMed]
49. Xu J, Wang J, Hu Y, et al. Unequal prognostic potentials of p53 gain-of-function mutations in human cancers associate with drug-metabolizing activity. Cell Death Dis 2014;5:e1108. [Crossref] [PubMed]
50. Cao X, Hou J, An Q, et al. Towards the overcoming of anticancer drug resistance mediated by p53 mutations. Drug Resist Updat 2020;49:100671. [Crossref] [PubMed]
51. Xin H, Zhao Z, Guo S, et al. Targeting the JAK2-STAT3-UCHL3-ENO1 axis suppresses glycolysis and enhances the sensitivity to 5-FU chemotherapy in TP53-mutant colorectal cancer. Acta Pharm Sin B 2025;15:2529-44. [Crossref] [PubMed]