Oxidative stress-related genes in uveal melanoma: the role of CALM1 in modulating oxidative stress and apoptosis and its prognostic significance

Abstract

Background: Uveal melanoma (UVM) is a rare yet aggressive form of ocular cancer with a poor prognosis. This study aims to investigate the role of oxidative stress-related genes (OSGs) in UVM, focusing on their involvement in key signaling pathways and immune infiltration and their potential as prognostic biomarkers and therapeutic targets.

Method: Differential gene expression analysis was conducted using 175 samples of normal retinal pigmented epithelium-choroid complex samples and 63 samples from UVM. Protein-protein interaction (PPI) networks were constructed to identify hub genes, and machine learning algorithms were utilized to screen for diagnostic genes, employing methods such as least absolute shrinkage and selection operator (LASSO) regression, random forest, support vector machine (SVM), gradient boosting machine (GBM), neural network algorithm (NNET), and eXtreme gradient boosting (XGBoost). A risk signature model was developed using data from The Cancer Genome Atlas (TCGA) cohort and validated using the International Cancer Genome Consortium (ICGC), GSE84976 dataset. Clinical samples were used to validate the diagnostic value. Experimental validation encompassed H₂O₂-induced oxidative stress assays and CALM1 overexpression analysis in UVM cells to evaluate its protective effects.

Results: A total of 2,576 differentially expressed genes (DEGs) were identified, with 185 overlapping OSGs enriched in pathways such as HIF-1, FoxO, PI3K-Akt, and apoptosis. Prognostic hub OSGs, including ACACA, CALM1, and DNM2, were associated with poor survival outcomes in the training set and multiple validation data. Revalidation using clinically collected samples confirmed that CALM1 exhibits superior diagnostic value. The risk signature model demonstrated strong predictive accuracy for a 5-year overall survival (AUC = 0.844). Immune infiltration analysis revealed increased CD4⁺ memory-activated T cells and mast resting cells in the high-risk group. Additionally, CALM1 overexpression attenuated H₂O₂-induced oxidative stress and apoptosis in UVM cells. CALM1 upregulation also mitigated the inhibitory effects of H₂O₂ on key cellular processes, including proliferation, migration, and invasion.

Conclusion: This study underscores the critical role of OSGs in the progression of UVM and their potential as prognostic biomarkers and therapeutic targets. The identified risk signature model and the protective role of CALM1 offer valuable insights for developing targeted therapies and enhancing patient clinical outcomes in UVM.

Keywords: CALM1; machine learning algorithms; oxidative stress; risk signature; uveal melanoma.

Figure 1

The flowchart of this study.

Figure 2

The identification of oxidative stress-related genes (OSGs). (A) Principal component analysis (PCA) before batch effects in merging the RNA sequencing data of GSE22138 and GSE29801. (B) PCA analysis after batch effects. (C) The Venn diagram showing the intersection of differentially expressed genes (DEGs) and OSGs. (D) Heatmap showing the differential expression of the intersection of OSGs between normal and UVM samples. (E) The enrichment function of Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis.

Figure 3

Screening prognostic hub OSGs in UVM. (A) The protein–protein interaction (PPI) analysis. (B) The hub genes network. (C) The Venn diagram showing the overlapping of prognosis and hub genes of the PPI network. Kaplan–Meier survival curve for overall survival according to the expression levels of OSGs, including ACACA (D), AKT2 (E), CALM1 (F), CALR (G), CDK2 (H), CXCR4 (I), DNM2 (J), EDN1 (K), HMOX1 (L), HSP90B1 (M), IL6 (N), POMC (O), SPP1 (P), TEK (Q), and TNFSF10 (R).

Figure 4

Candidate biomarkers identified by the machine learning algorithms. The ROC curves of the LASSO model (A), random forest (B), SVM (C), NNET (D), GBM (E), and XGBoost (F). The feature importance in different machine learning algorithms (G).

Figure 5

The construction of the risk signature model in the TCGA cohort. (A) The multivariate regression analysis. (B) Survival curves to evaluate the risk stratification ability of OSGs. (C) Risk plots to illustrate the risk scores of different risk groups. (D) Risk plots to illustrate the survival status of different risk groups. (E) Heatmap showing the expression levels and risk scores in the risk model. (F) ROC curves to evaluate the sensitivity and specificity of the risk signature to predict the 1-, 3-, and 5-year overall survival.

Figure 6

Validation of the risk signature model in the ICGC database. (A) Survival curves to investigate the risk stratification ability of OSGs. (B) The risk scores of the high- and low-risk groups. (C) The survival status of the high- and low-risk groups. (D) The expression levels and risk scores in the risk model. ROC curves to evaluate the sensitivity and specificity of the risk signature to predict the 1-year (E), 3-year (F), and 5-year (G) overall survival.

Figure 7

The immune function between the high- and low-risk groups. (A) The difference of immune cell infiltration scores between the high- and low-risk groups. (B) The heatmap was generated to show the relationship between OSGs and immune cells. (C) The correlation analysis between risk scores and immune cells. (D) The difference of stromal scores, immune scores, and estimate scores between the high- and low-risk groups. “ns” represents no statistical significance; * represents P < 0.05, ** represents P < 0.01, and *** represents P < 0.001.

Figure 8

CALM1 overexpression attenuated H₂O₂-induced oxidative stress in MP65 cells. (A) Cell viability was inhibited by H₂O₂. (B) The expression level of CALM1 was detected by Western blot upon transfection of H₂O₂-induced MP65 cells with negative control (vector) or CALM1 overexpression. The activity of SOD (C), MDA (D), and LDH (E) in H₂O₂-induced MP65 cells was determined by ELISA. (F) Western blot evaluating the expression levels of SOD2 and CAT. * or ^# represents P < 0.05, ** or ^## represents P < 0.01, and *** or ^### represents P < 0.001. “0 µmol/L” denotes the untreated control.

Figure 9

CALM1 overexpression attenuated H₂O₂-induced apoptosis in MP65 cells. (A) After transfection with negative control (vector) or CALM1 overexpression, MP65 cells were determined by Annexin V-FITC/propidium iodide (PI) staining. (B) Percentage of apoptotic cell death and necrosis. (C) The expression level of BAX and CASP3 (caspase 3) was detected by Western blot. Three independent experiments were carried out. “ns” represents no statistical significance; * or ^# represents P < 0.05, ** or ^## represents P < 0.01, and *** or ^### represents P < 0.001.