Development of a novel copper metabolism-related risk model to predict prognosis and tumor microenvironment of patients with stomach adenocarcinoma

Abstract

Background: Stomach adenocarcinoma (STAD) is the fourth highest cause of cancer mortality worldwide. Alterations in copper metabolism are closely linked to cancer genesis and progression. We aim to identify the prognostic value of copper metabolism-related genes (CMRGs) in STAD and the characteristic of the tumor immune microenvironment (TIME) of the CMRG risk model. Methods: CMRGs were investigated in the STAD cohort from The Cancer Genome Atlas (TCGA) database. Then, the hub CMRGs were screened out with LASSO Cox regression, followed by the establishment of a risk model and validated by GSE84437 from the Expression Omnibus (GEO) database. The hub CMRGs were then utilized to create a nomogram. TMB (tumor mutation burden) and immune cell infiltration were investigated. To validate CMRGs in immunotherapy response prediction, immunophenoscore (IPS) and IMvigor210 cohort were employed. Finally, data from single-cell RNA sequencing (scRNA-seq) was utilized to depict the properties of the hub CMRGs. Results: There were 75 differentially expressed CMRGs identified, 6 of which were linked with OS. 5 hub CMRGs were selected by LASSO regression, followed by construction of the CMRG risk model. High-risk patients had a shorter life expectancy than those low-risk. The risk score independently predicted STAD survival through univariate and multivariate Cox regression analyses, with ROC calculation generating the highest results. This risk model was linked to immunocyte infiltration and showed a good prediction performance for STAD patients' survival. Furthermore, the high-risk group had lower TMB and somatic mutation counters and higher TIDE scores, but the low-risk group had greater IPS-PD-1 and IPS-CTLA4 immunotherapy prediction, indicating a higher immune checkpoint inhibitors (ICIs) response, which was corroborated by the IMvigor210 cohort. Furthermore, those with low and high risk showed differential susceptibility to anticancer drugs. Based on CMRGs, two subclusters were identified. Cluster 2 patients had superior clinical results. Finally, the copper metabolism-related TIME of STAD was concentrated in endothelium, fibroblasts, and macrophages. Conclusion: CMRG is a promising biomarker of prognosis for patients with STAD and can be used as a guide for immunotherapy.

Keywords: IMvigor210; biomarkers; copper; copper metabolism-related genes; immunotherapy; prognosis; single-cell RNA sequencing.

FIGURE 1

Workflow of the study.

FIGURE 2

Identification of CMRGs in STAD. (A) Volcano map of CMRGs with different expressions. (B) Predictive value of CMRG. (C) Venn diagram of selected prognostic CMR-DEGs. (D, E) Lasso Cox regression analysis of 6 prognostic CMR-DEGs. (F) mRNA expression of CP, F5, LOX, S100A12 and SNCG in STAD and normal tissues. *p < 0.05; **p < 0.01; ***p < 0.001.

FIGURE 3

Predictive value assessment of the risk model. Distribution of overall survival status (A), and risk scores (B). (C) Heatmap of CMRG expression. (D) Kaplan-Meier curves for OS in the TCGA-STAD cohort. (E) ROC curves for OS prediction. The distribution of overall survival status (F), and risk score (G). (H) CMRG expression in the validation cohort. (I) Kaplan-Meier survival curves for OS in the validation cohort. (J) Validation of ROC curves for OS prediction in the validation cohort.

FIGURE 4

Predictive value of the risk model. Independent prediction analysis by univariate (A) and multivariate Cox regression (B). (C) Predictive accuracy of the risk model in terms of age, gender, grade, and stage. (D) Nomogram for OS predictions. (E) Calibration curves for OS predictions. (F) DCA curves for 5-year clinical predictions.

FIGURE 5

Clinicopathological characteristics in the risk model. (A) Heatmap of clinical features and risk score distribution. (B) Incidence of high-risk and low-risk tumor stages. The proportion of patients by age (C), gender (D), grade (E), and stage (F). (G–J) OS analysis of subgroup.

FIGURE 6

CMRG enrichment analysis. GSEA of KEGG pathways of the high-risk group (A) and low-risk group (B). (C) Heatmap of pathways enrichment related to CMRGs.

FIGURE 7

TMB analysis. Oncoprint of mutations in the high-risk group (A) and low-risk group (B). (C) Gene mutations of CMRGs. (D) Differences in TMB between the two risk groups. (E) Scatter plot of correlation between risk score and TMB. (F) Kaplan-Meier curves for the high- and low-TMB groups. (G) Kaplan-Meier curves for patients with different TMB and risk scores.

FIGURE 8

Immune feature in the risk model. (A) TICs distribution. (B) Immune function scores comparison. (C) Correlation between TICs and risk scores. (D) Correlation between CMRGs and immune scores. (E) Correlation between CRMG and ESTIMATE scores and stromal scores. Immune scores were correlated with the expression of CP (F), F5 (G), LOX (H), S100A12 (I), and SNCG (J). *p < 0.05; **p < 0.01; ***p < 0.001.

FIGURE 9

Predicted response to immunotherapy. Analysis of (A) dysfunction, (B) exclusion, (C) MSI, and (D) TIDE scores between the high-risk and low-risk groups. Comparison of IPS in the two groups with CTLA4_negative/PD-1_negative (E), CTLA4_negative/PD-1_positive (F) CTLA4 _positive/PD-1_negative (G) CTLA4_positive/PD-1_positive (H). Sensitive drugs in the low-risk group (I–M) and high-risk group (N–P).

FIGURE 10

Response to chemotherapy in the risk model. (A) Kaplan-Meier analysis of OS in the IMvigor210 cohort. (B) The proportion of immunotherapy responses and stages in risk groups. cr, complete response; pr, partial response; sd, stable disease, pd, disease progression.

FIGURE 11

Characteristics of CMRG clusters. (A) Cluster plots for subtype analysis of the STAD sample. k = 2 for higher intra-group correlations but lower inter-group correlations. CDF (B) and delta plots (C) for consensus analysis. PCA analysis of two subgroups (D), tSNE (E) and UMAP (F). (G) Risk scores for the probability of survival for clusters A and (B) (H) Kaplan-Meier survival curves showing the probability of survival for clusters A and (B) (I) Heat map showing expression of clusters. (I) Heatmap of the 5 CMRGs expression in clinical features and clusters. (J) Sankey plots of CMRG clusters, risk scores, and survival.

FIGURE 12

Validation of CMRG. (A) CMRG mRNA expression levels. (B) IHC analysis of CP, S100A12, and SNCG from the HPA database. (C) IHC analysis of LOX and F5 from STAD samples. *p < 0.05; **p < 0.01; ***p < 0.001.

FIGURE 13

scRNA-seq analysis of CMRG. (A) Cells were divided into 17 clusters. (B) Annotation of the cell clusters. (C, D) The proportion of TICs in GSE167297. (E) Expression of CMRG mRNA in different TICs. (F) Distribution of CMRG in different cell types. Comparison of the expression levels of F5(G) and SNCG (H) in tumor and normal cells.