A novel 7-chemokine-genes predictive signature for prognosis and therapeutic response in renal clear cell carcinoma

Abstract

Background: Renal clear cell carcinoma (ccRCC) is one of the most prevailing type of malignancies, which is affected by chemokines. Chemokines can form a local network to regulate the movement of immune cells and are essential for tumor proliferation and metastasis as well as for the interaction between tumor cells and mesenchymal cells. Establishing a chemokine genes signature to assess prognosis and therapy responsiveness in ccRCC is the goal of this effort. Methods: mRNA sequencing data and clinicopathological data on 526 individuals with ccRCC were gathered from the The Cancer Genome Atlas database for this investigation (263 training group samples and 263 validation group samples). Utilizing the LASSO algorithm in conjunction with univariate Cox analysis, the gene signature was constructed. The Gene Expression Omnibus (GEO) database provided the single cell RNA sequencing (scRNA-seq) data, and the R package "Seurat" was applied to analyze the scRNA-seq data. In addition, the enrichment scores of 28 immune cells in the tumor microenvironment (TME) were calculated using the "ssGSEA" algorithm. In order to develop possible medications for patients with high-risk ccRCC, the "pRRophetic" package is employed. Results: High-risk patients had lower overall survival in this model for predicting prognosis, which was supported by the validation cohort. In both cohorts, it served as an independent prognostic factor. Annotation of the predicted signature's biological function revealed that it was correlated with immune-related pathways, and the riskscore was positively correlated with immune cell infiltration and several immune checkpoints (ICs), including CD47, PDCD1, TIGIT, and LAG-3, while it was negatively correlated with TNFRSF14. The CXCL2, CXCL12, and CX3CL1 genes of this signature were shown to be significantly expressed in monocytes and cancer cells, according to scRNA-seq analysis. Furthermore, the high expression of CD47 in cancer cells suggested us that this could be a promising immune checkpoint. For patients who had high riskscore, we predicted 12 potential medications. Conclusion: Overall, our findings show that a putative 7-chemokine-gene signature might predict a patient's prognosis for ccRCC and reflect the disease's complicated immunological environment. Additionally, it offers suggestions on how to treat ccRCC using precision treatment and focused risk assessment.

Keywords: chemokine; gene signature; immunotherapy; renal clear cell carcinoma; tumor microenvironment.

FIGURE 1

Construction of 7-chemokine-genes prognostic signature. (A) Flow chart of model construction. (B) Cross-validation was performed to optimize the parameter selection of the LASSO regression model. (C) Distribution of lasso coefficients of 11 prognosis-related chemokine genes. (D, E) Correlation between chemokine genes signature and profiles of seven chemokine genes’ expression in the training cohort and validation cohort.

FIGURE 2

Association between chemokine-based gene signature and clinical features in ccRCC. (A) Correlation of riskscore and clinicopathological characteristics of patients in the training cohort. (B) Correlation of riskscore and clinicopathological characteristics of patients in the validation cohort. (C, E) In the training and validation cohorts, riskscore considerably rise at higher-grade ccRCC. (D, F) In the training and validation cohorts, riskscore considerably rise at higher-stage ccRCC. The significance of the difference was tested with Wilcoxon test.

FIGURE 3

K-M survival analysis and nomogram survival prediction. (A, B) Distribution of riskscore, survival status and survival time in ccRCC patients, and heat map of 7 chemokine genes. (C, D) K-M survival curves for OS and ROC curves for 1-, 3-, and 5-year survival rates. (E) Nomogram prediction combining clinicopathological features and riskscore. (F) Predicted and observed 1-year, 3-year and 5-year survival in calibration plots for training and validation cohorts. (G) The C-index is used to visualize the predictive effect of predictive model, riskscore, predictive model without riskscore, and clinicopathological factors.

FIGURE 4

Biological functions associated with the 7-chemokine-genes signature. (A–D) The biological processes (BP) and cell components (CC) that are enriched by genes that are positively correlated with riskscore. (E, F) Correlation of riskscore with the HALLMARK gene set. The heat map shows the enrichment scores of the HALLMARK for each patient. Bar and line plots show R- and p-values for correlation analysis.

FIGURE 5

Differences in mutations between high- and low-risk groups. (A, B) Somatic mutation waterfall plots in the training cohort. (C) TMB difference between low-risk and high-risk group. (Wilcoxon test). (D) K-M survival curves comparing the groups with high and low TMB levels. (E) Intergroup K-M survival curves for the four groups.

FIGURE 6

scRNA-seq data analysis in GSE152938. (A) Cell-type annotation of clusters. (B) Signature genes expression levels in several cell subtypes. (C) UMAP shows expression of signature genes in all cell subtypes. (D) Subtypes of Cancer cells. (E) Signature genes expression levels in different subtypes of cancer cells. (F) UMAP shows expression of signature genes in subtypes of cancer cells. (CCL11 is unavailable).

FIGURE 7

Immune infiltration reflected by gene signature. (A) Correlation between riskscore and inhibitory ICs. The R-value is shown by the band’s width. The p-value is indicated by the band’s color. The correlation was examined using Pearson correlation analysis. (B) Expression levels of inhibitory ICs in various subtypes of cancer cells. (C) Comparison of 28 immune cell enrichment scores. (D) Comparison of the difference in the abundance of immune infiltrating cells by the CIBERSORT algorithm. The significance of the difference was tested with Wilcoxon test. *p < 0.05, **p < 0.01, and ***p < 0.001.

FIGURE 8

Twelve drug agents were identified as well as four VEGFR inhibitors for analysis. (A, C) Correlation between riskscore and the IC50 estimates for the 12 agents. The correlation was examined using Spearman correlation analysis. (B, D) Differences in the estimated IC50 for 12 agents between the high- and low-risk groups. (E, F) Comparison of IC50 estimates of four VEGFR targeted drugs (sunitinib, sorafenib, pazopanib and axitinib). The significance of the difference was tested with Wilcoxon test. *p < 0.05, **p < 0.01, and ***p < 0.001.