The normalized RNA sequencing (RNA-seq) expression data as transcripts per million (TPM) and corresponding clinical information of 321 KIRP samples were acquired from TCGA database (https://xenabrowser.net/datapages/, until December 01, 2021). After screening, the 72 samples were rejected based on the merged sample quality annotations (https://gdc.cancer.gov/about-data/publications/pancanatlas). Additionally, 28 normal kidney samples were collected from the Genotype-Tissue Expression (GTEx) database (https://xenabrowser.net/datapages/, until December 01, 2021). All RNA-seq data were log₂-transformed for further analysis.

As shown in Supplementary Table S1, the 52 PRGs were retrieved from GSEA (http://www.gsea-msigdb.org/gsea/index.jsp) and previous research (Supplementary Table S1) (Qi et al., 2021). The “limma” R package was utilized to determine differentially expressed genes (DEGs) between 28 normal kidney samples and 249 KIRP samples, with |Log₂FC|>1 and p < 0.05. The differentially expressed PRGs were selected through the “VennDiagram” package, and their protein–protein interaction (PPI) network was acquired from the STRING database (https://www.string-db.org/, version 11.5).

To investigate the biological characteristics of differentially expressed PRGs in KIRP patients, we classified the patients into different subtypes using the “ConsensusClusterPlus” R package with a resampling rate of 80% and 500 iterations. The differential clinical information and expression of different subtypes were shown in the heat-map. The survival differences among clusters were visualized with the Kaplan–Meier curve using the “survival” R package.

Univariate Cox proportional hazards regression analysis determined the prognostic value of PRGs in KIRP patients, and genes with p < 0.2 were selected for subsequent analysis. The candidate PRGs were selected using 10-fold cross-validation of the least absolute shrinkage and selection operator (LASSO)-penalized Cox regression analysis in the “glmnet” R package. Then, the prognostic model was built based on the six genes (CASP8, CASP9, CHMP2A, GPX4, IL6, and IRF1) and their coefficients, and the penalty parameter (λ) was decided by the minimum criteria. The risk score of each patient was calculated according to regression coefficients derived from the LASSO-Cox regression model multiplied with its gene expression level, as follows: Risk score = ∑ i 6 Xi*Yi (X: coefficients, Y: gene expression level). Next, 249 patients were separated into the low-risk group and the high-risk group based on the median risk score, and Kaplan–Meier survival analysis and a log-rank test were performed to compare the survival outcomes between two risk groups. The area under the curve (AUC) of the receiver-operating characteristic (ROC) curve based on the “survival,” “survminer,” and “time-ROC” R packages was used to evaluate the predictive performance of the prognostic model.

To identify independent prognostic factors and validate the independent prognostic value of risk score, the risk score and clinical characteristics including age, gender, and T-stage, N-stage, M-stage, and tumor stage in TCGA dataset were analyzed via univariate and multivariate Cox regression models, respectively. These multivariate prognostic analysis results were calculated and then visualized by the “forestplot” R package.

The DEGs between the low-risk group and the high-risk group were identified via the “limma” R package. |Log₂FC|>1 and p < 0.05 were considered to be statistically significant. The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of those DEGs were performed via the “clusterProfiler” R package. The “CIBERSORT” package was used to explore the landscape of 22 tumor-infiltrating immune cells and their connection with the signature.

A total of three pairs of tissues from KIRP patients and their paired normal tissues were collected from the Department of Pathology of The Second Affiliated Hospital of Chongqing Medical University, which was approved by the Human Research Ethics Committee.

Paraffin sections were placed in a 60°C oven to melt the paraffin and soaked in xylene and ethanol at different concentrations to elute the paraffin. Then, the sections were incubated with 3% H₂O₂ at room temperature for 10 min to eliminate endogenous peroxidase activity. The sections were immersed in boiling EDTA repair solution for 10 min and allowed to cool naturally. Then, the sections were incubated with 5% BSA blocking solution at 37°C for 30 min. The sections were incubated with appropriately IL-6 primary antibody (1:50, Proteintech, 21865-1-AP) and CASP9 primary antibody (1:200, Abcam, ab202068) at 4°C overnight. The next day, the sections were washed three times with PBS for 10 min. The sections were incubated with secondary antibody at room temperature for 60 min. After washing three times with PBS for 10 min, the tissues were stained with DAB and hematoxylin. Then, the sections were sequentially immersed in 60, 75, 80, 95, and 100% ethanol for dehydration. Finally, the sections were sealed with neutral gum and observed with a light microscope.

The DEGs between the normal and KIRP tissues were analyzed with one-way analysis of variance. The Kaplan–Meier curve with a two-sided log-rank test was utilized to assess the survival difference. Cox regression models were applied to identify prognostic factors, with hazard ratios (HRs) and 95% confidence intervals (CIs). All statistical analyses were completed by R software (v4.1.0), and p < 0.05 was considered statistically significant.

The mRNA expression of 52 PRGs from 249 tumor and 28 normal tissues was examined on the basis of TCGA data. 20 genes were considered differentially expressed PRGs with |Log2FC|>1 and p < 0.05. As shown in Figure 1A, 13 genes (GRX4, BAX, CHMP2A, PYCARD, CHMP48, IL18, CASP4, PLCG1, TP53, CASP1, CHMP6, CASP8, and CASP3) of the above genes were upregulated, and 7 genes (TIRAP, CHMP7, IL6, IRF1, CASP9, PRKACA, and CHMP3) were downregulated in the tumor group. To further explore the interactions of these 20 differentially expressed PRGs, PPI network analysis was conducted, and the results are shown in Figure 1B. And the correlation network of these genes is shown in Figure 1C.

To investigate whether differentially expressed PRGs had an impact on survival outcomes, we carried out the consensus clustering analysis of 249 KIRP patients. Based on the above PRGs, the results showed that the clustering variable (k) = 2 was considered to have the optimal stability from k = 2 to 9, implying that KIRP patients could be divided into two clusters (cluster 1 and cluster 2) with the highest intragroup correlations and the lowest intergroup correlations (Figures 2A–D). Notably, compared with those in cluster 1, KIRP patients in cluster 2 had a significantly longer survival (Figure 2E, p = 0.0041), indicating a significant prognostic value of these PRGs. Moreover, clinical characteristics including gender, age, and tumor TNM stage were presented in two clusters without significant differences (Figure 2F).

The clinical implication of PRGs was further assessed in KIRP patients. As shown in univariate Cox regression analysis (Figure 3A), 11 (IL6, CHMP2A, GPX4, CASP3, CASP4, CASP8, CASP9, CHMP7, PRKACA, TP53, and IRF1) of PRGs were survival-related with p < 0.2. And then, LASSO-Cox regression analysis was performed using 11 prognostic genes, and a signature consisting of CASP9, CHMP2A, GPX4, IL6, IRF1, and CASP8 was constructed based on the optimal λ score (Figures 3B,C). The risk score was calculated by the following formula: Risk score = (0.067*IL6 exp.) + (0.01011*CASP8 exp.) + (0.5066*IRF1 exp.) + (−0.4791*CASP9 exp.) + (−0.0988*CHMP2A exp.) + (−0.119*GPX4 exp.). 249 KIRP patients were approximately divided into the low-risk group and the high-risk group according to the median risk score (Figure 3D). As shown in Figure 3E, the result of principal component analysis (PCA) indicated patients of two risk groups could be distributed into two directions. Figure 3F shows that patients of the low-risk group tended to have a low probability of mortality compared to those of the high-risk group. Consistently, Kaplan–Meier analysis showed that the high-risk group had a significantly shorter survival time (Figure 3G, p < 0.001). A time-dependent ROC curve was performed to evaluate the predictive performance. And the AUC was 0.85 at 1 year, 0.785 at 2 years, and 0.707 at 3 years (Figure 3H), showing that this risk model exhibited high accuracy and sensitivity in predicting the prognosis of KIRP patients.

Univariate Cox regression analysis and multivariable Cox regression analysis were carried out to determine whether the risk score could serve as the independent prognostic predictor for survival in KIRP. The univariate Cox regression analysis showed that the risk score was significantly associated with poor survival (HR: 3.276, 95% CI 1.528–7.025, p = 0.002, Figure 4A). Moreover, other clinical characteristics including tumor stage and N-stage were found as the risk factors as well. After the adjustment of confounding factors, the result of multivariable Cox regression analysis suggested that the risk score was still a risk prognostic factor (HR: 2.655, 95% CI 1.192–5.911, p = 0.016, Figure 4B). As shown in Figure 4C, the patients suffering from advanced tumor stage had a higher probability of high risk score.

To further investigate the differences in biological function and pathway between the low-risk group and the high-risk group, DEGs were generated using the “limma” R package. Then, these DEGs were further analyzed with the GO term and KEGG pathway enrichment analysis. As presented in Figure 5A, the top-rank biological processes were lymphocyte mediated immunity, complement activation pathway, and humoral immune response mediated by circulating immunoglobulin. Moreover, the most highly enriched cellular components associated with DEGs were immunoglobulin complex, external side of plasma membrane, and T cell receptor complex (Figure 5B). As for the molecular functions, antigen binding, immunoglobulin receptor binding, and immune receptor activity were on the top list (Figure 5C). Furthermore, the KEGG pathway analysis is shown in Figure 5D, and the cytokine-cytokine receptor interaction pathway and the viral protein interaction with cytokine and cytokine receptor signaling pathway were mostly associated with these DEGs. These results showed that these DEGs were significantly enriched in immune-related functions or pathways.

Based on functional enrichment, we speculated that tumor immune status of KIRP played an important role in the development of cancer. Therefore, the tumor immune microenvironment (TIM) of KIRP was further explored. Firstly, the abundance of immune cell infiltration was investigated. The overview of tumor microenvironment immune cell compositions is shown in Figure 6B, in which 22 type cells had differential distributions in KIRP. To be specific, M2 macrophages were found with an especially high infiltration level. The infiltration levels of naïve B cells, CD8⁺ T cells, regulatory T cells, and M1 macrophages in the high-risk group were significantly upregulated, while the infiltration levels of memory B cells, activated mast cells, and resting mast cells decreased (Figures 6A,C, p < 0.05).

To further explore genetic interrelationships in the PRG signature, the PPI network of these genes was obtained using the STRING database (Figure 7A). IL6 was the hub node in the obtained interactive network. Next, the CytoHubba plugin in Cytoscape was used, and the genes with the top three MCC values (IL6, CASP8, and CASP9) were identified as candidate hub genes (Figure 7B). Meanwhile, we assessed the prognostic role of these PRGs in the signature, and three genes (CASP9, IL6, and IRF1) were survival-related (p < 0.05). High expression of CASP9 was significantly correlated with longer overall survival in KIRP patients (p = 0.031), while high IL6 (p = 0.004) and IRF1 (p < 0.001) expressions had a shorter survival time (Figures 7C–E). The intersections of the above genes were selected as hub genes. Finally, two hub genes (IL6 and CASP9) were identified and further validated through protein expression levels in vitro. As depicted in immuno-histochemical staining, the protein expression levels of IL6 and CASP9 were significantly downregulated in KIRP tissues compared with normal renal tissues (Figures 7F,G).