Identification and validation of prognostic biomarkers in ccRCC: immune-stromal score and survival prediction

Abstract

Background: The tumor microenvironment (TME) is integral to tumor progression. However, its prognostic implications and underlying mechanisms in clear cell renal cell carcinoma (ccRCC) are not yet fully elucidated. This study aims to examine the prognostic significance of genes associated with immune-stromal scores and to explore their underlying mechanisms in ccRCC.

Methods: Data from the Cancer Genome Atlas (TCGA) and International Cancer Genome Consortium (ICGC) were subjected to analysis to compute immune and stromal scores utilizing the ESTIMATE algorithm. The weighted gene co-expression network analysis (WGCNA) was employed to identify gene modules associated with these scores. Differentially expressed genes were assessed using the limma package. Prognostic biomarkers were subsequently identified through univariate, LASSO, and multivariate Cox regression analyses, culminating in the development of a risk score model. Gene expression was confirmed in ccRCC cell lines (786-O, Caki-1) and tumor tissues. Functional assays, such as wound healing and Transwell assays, were employed to evaluate tumor invasion and migration. The prognostic accuracy was assessed through ROC curve analysis, and a nomogram integrating risk scores with clinical variables was constructed. Analyses of immune infiltration, human leukocyte antigen (HLA) expression, immune checkpoint expression, immunophenoscore (IPS), tumor immune dysfunction and exclusion (TIDE) scores, and responses to six targeted therapies were conducted across different risk groups.

Results: Twelve critical prognostic markers, including CAPRIN1, CXCR3, FERMT3, HAPLN3, HBP1, MACF1, MPEG1, OSCAR, STAT1, UBA7, VAMP1, and VSIG4, were identified. The risk score model exhibited a high degree of predictive accuracy for survival outcomes in ccRCC. Immune profiling revealed significant differences in the TME between risk groups, with high-risk patients displaying elevated expression of HLA and immune checkpoints. Drug sensitivity analyses suggested that high-risk patients had a better response to erlotinib, temsirolimus, axitinib, and sunitinib, whereas low-risk patients demonstrated greater sensitivity to pazopanib. Variability in immunotherapy responsiveness between groups was observed based on IPS and TIDE analyses.

Conclusion: This study highlights the prognostic value and TME-related mechanisms of immune-stromal score signatures in ccRCC, developing a risk score model and nomogram for predicting patient prognosis.

Keywords: Clear cell renal cell carcinoma; Immune score; Prognosis; Risk score; Stromal score.

Fig. 1

The immune (A-F) and stromal (G-L) scores of different ccRCC groups stratified by clinic characteristics (T stage, M stage, N stage, gender, tumor stage, tumor grade)

Fig. 2

WGCNA identifies immune score-related genes and stromal score-related genes. A Six outlier samples were eliminated based on the sample clustering results obtained from the WGCNA analysis. B Sample dendrogram and trait heatmap. C The optimal soft threshold power was determined to be 14, in which R² was 0.85 approximately. D Merging similar modules and identifying 23 modules from the co-expression network. E Through an examination of the relationship between modules and sample traits depicted in the module-trait heatmap, three modules (MEgreenyellow, MEgreen, MEdarkred) exhibiting the most significant correlation coefficients with the Immune Score trait were identified, collectively encompassing 415 genes. Similarly, for the Stromal Score trait, three modules (MEgreenyellow, MEgreen, MEblue) displaying the highest correlation coefficients were chosen, encompassing a total of 1100 genes

Fig. 3

Immune score-related (A, B) and stromal score-related genes (C, D) were identified by the LASSO algorithm. E After removing duplicated genes, remaining immune-stromal score-related genes were input into multivariate Cox regression. 12 genes were significantly associated with the prognosis of ccRCC. *P < 0.05, **P < 0.01, ***P < 0.001

Fig. 4

VAMP1 is significantly overexpressed in ccRCC and associated with the prognosis. A-B The expression levels of VAMP1 are significantly higher in renal cell carcinoma cell lines compared to HK-2 cell lines. C-D VAMP1 is highly expressed in the cancer tissues of ccRCC patients. E–F The efficiency of VAMP1 knockdown was validated in 786-O and Caki-1 cell lines using qRT-PCR and WB. G-H VAMP1 knockdown in 786-O and Caki-1 cells resulted in a significant reduction in migration ability (Scale bar = 10 μm). I-J VAMP1 knockdown in 786-O and Caki-1 cells led to decreased invasion (Scale bar = 10 μm). **P < 0.01, ***P < 0.001

Fig. 5

Construction of immune-stromal score-related risk score model. A High- and low-risk groups from the TCGA training set. B Kaplan–Meier survival curves for high- and low-risk score groups. C The AUC for 1-, 3- and 5-year, respectively, indicating high accuracy of risk score model in predicting the survival of ccRCC patients. D-F The consensus results from the ICGC validation set

Fig. 6

A Multivariate analysis showing the risk score, tumor stage and T stage were significantly associated with prognosis of ccRCC patients. B The nomogram for predicting the 1-, 3- and 5-year survival of ccRCC patients based on risk score, tumor stage and T stage. C The calibration curves indicated a high level of agreement between the predicted and observed overall survival rates. D The AUCs of the nomogram for 1-, 3- and 5-year respectively. *P < 0.05, ***P < 0.001

Fig. 7

Low- and high-risk groups of ccRCC had different therapeutic sensitivity. A-D High-risk patients were more sensitive to erlotinib, temsirolimus, axitinib and sunitinib. E Patients in low-risk group were more sensitive to pazopanib. F There was no statistically significant disparity in therapeutic response to sorafenib between the low- and high-risk groups. ***P < 0.001