Differences in clinical features and prognosis of two ccRCC subtypes based on the expression of DNA damage repair (DDR) genes.
According to previous studies[21, 22], we obtained 231 DDR genes (Supplementary Table 2). Consensus clustering analysis was adopted to identify subtypes based on the expression of these selected 231 DDR genes in TCGA ccRCC cohort. Consequently, compared with consensus clustering matrix for k=3, 4, 5 and 6 (Supplementary Figure 1), the k=2 was determined as the most appropriate clustering from k = 2 to 6 (Figure 1a-1c). Thus, a total of 530 ccRCC patients were split into two subtypes, i.e., Subtype1 (S1, n=165) and Subtype2 (S2, n=365). To explore the heterogeneity between two subtypes, we further analyzed the clinical characteristics and prognosis of patients of these two subtypes. As displayed in Figure 1d, patients of subtype1 exhibited a higher proportion of males and a higher cancer stage and tumor grade compared to patients of subtype2. (P<0.05). Additionally, subtype1 presented unfavorable overall survival (OS) (P=0.0026), progression free survival (PFS) (P=0.0039), and disease specific survival (DSS) (P=0.00058) in comparison with subtype2 (Figure 1e-1g). These findings suggested a marked heterogeneity between two subtypes.
Discrepant tumor microenvironment (TME) of two ccRCC subtypes
Given the marked heterogeneity between two subtypes, immune-related analyses were also performed to investigate the TME of two subtypes. Firstly, in comparation with the subtype1 and normal group, a remarkably elevated PD-L1 expression levels were noted in subtype2 and ccRCC group, respectively (P<0.001, Figure 2a-b). Subsequently, by using CIBERSORT algorithm, discrepant enrichment of tumor‐infiltrating lymphocytes (TIL) was observed in two subtypes (Figure 2c), notably, subtype1 exhibited higher enrichment of Tregs, CD8+ T cell and T follicular helper cell (Figure 2d), whereas subtype2 tightly linked with the infiltration of CD4+ memory resting cell, activated Mast cell and Monocyte (Figure 2d). GSEA was subsequently conducted to investigate the regulatory mechanisms that lead to this difference in TME of two subtypes. Consequently, there were 3 KEGG pathways significantly enriched in subtype1 (Figure 3e-3g), including the TGF-β signaling pathway (NES=1.806; Figure 3e), Focal adhesion (NES=1.735; Figure 3f) and JAK-STAT signaling pathway (NES=1.588; Figure 3g). These signaling pathways might implicated in the discrepancy in TME of two subtypes.
Identification of RAD54L as the potential immune-related DDR gene in ccRCC
As previously described, distinct PD-L1 expression and TME were noted between two subtypes defined according to the expression of DDR genes. These findings implied the potential roles of DDR genes in regulating TME of ccRCC. Further, with the help of R software “stat” package, most DDR genes (191/231, Supplementary Table 3) were found to be positively associated with the expression of PD-L1 (Spearman Cor>0, P<0.05, Figure 3a). then by analyzing the RNA-seq and clinical information from the TCGA database, 1437 genes (including 6 DDR genes) were found to be remarkably upregulated in ccRCC compared to normal samples (adj. P<0.05, log2FC>1, Figure 3b), and 34 DDR genes were shown to be detrimental to overall survival (OS) in ccRCC patients (P<0.05, Figure 3c). Subsequently, an intersection of DDR genes positively linked to PD-L1 expression, highly expressed in ccRCC and correlated with worse OS were used to determine potential immune-related DDR genes among the 231 DDR genes. As displayed in Figure 3d, RAD54L was a key immune-related DDR gene in accordance with the three conditions. Further analysis of expression patterns of RAD54L in ccRCC demonstrated that RAD54L expression not only elevated in paired ccRCC samples (Figure 3e), but also correlated with the T stage (Figure 3f), N stage (Figure 3g) and M stage (Figure 3h) in ccRCC. Moreover, the increased expression level of RAD54L was also verified in 21 pairs of ccRCC and adjacent normal tissues (Figure 3i).
Associations of RAD54L expression with prognosis in ccRCC
Further, the prognostic values of elevated RAD54L expression were analyzed, where, highly expressed RAD54L group showed unfavorable OS (P=0.0027) and DSS (P=0.00017) in comparation with the lowly group (Figure 4a-b). Cox analyses were undertaken to analysis the prognostic values of RAD54L expression and other clinical variates, univariate analysis revealed that RAD54L expression and 6 clinical variates including age, TNM stage, pathological stage and histologic grade were tightly linked to the OS of ccRCC patients (P<0.05, Figure 4c), and multivariate analysis, displayed as forest plot in Figure 4d, demonstrated that RAD54L expression, age, and M stage can independently predict OS for ccRCC patients (P<0.05). Moreover, with the help of R “rms” and “survival” packages, a nomogram model was developed via combining RAD54L expression with other variates to help predict 1-y, 3-y and 5-y OS of ccRCC patients (Figure 4e).
Associations of RAD54L and its co-expressed genes with TME in ccRCC
GSEA was then undertaken to explore the potential KEGG pathways of upregulated RAD54L in ccRCC, and the results revealed that 5 signaling pathways were remarkably enriched in highly expressed RAD54L samples, including T cell receptor signaling pathway, cell cycle, JAK-STAT signaling pathway, antigen processing presentation, and cell adhesion molecules pathway (adj. p and FDR <0.05, Supplementary Table 4, Figure 5a). Further, through performing spearman correlation analysis, EZH2, KIF18B and EME1 were noted to be the 3 strongest correlated genes (Figure 5b), besides, these 3 genes also remarkably overexpressed in ccRCC (Supplementary Figure 2a) and their elevated expression tightly linked to worse OS and DSS of ccRCC patients (Supplementary Figure 2b-c). Subsequently, ssGESA algorithm was adopted to initially assess the relationships between these 4 genes and TIL in ccRCC, consequently, RAD54L and its 3 co-expressed genes tightly linked to the enrichment of T helper cells, Th2 cells and Tregs, while negatively associated with the infiltrating levels of NK cells, Th17 cells and Mast cells (Figure 5c-5f). These findings revealed that RAD54L and its co-expressed genes might be implicated in regulating TME of ccRCC.
Correlations of RAD54L with TME in pan-cancer
Furthermore, the xCell algorithm was adopted to evaluate the links between RAD54L expression and the TME in pan-cancer. As depicted as a heat map in Figure 6a, a correlation analysis involving 33 cancer types revealed significant associations of RAD54L expression with the infiltrating levels of Th2 cells in 32 cancer types, in addition, RAD54L also showed positive correlations with the excessive production of immune-escape markers (Figure 6b). Moreover, the ssGSEA algorithm was also carried out to further evaluate the links between the RAD54L expression and Th2 cells in pan-cancer, as shown in a series of scatter plots, striking correlations were noted between the enrichment of Th2 cells and RAD54L expression in 30 cancer types (Supplementary Figure 3). These findings revealed that RAD54L might be implicated in regulating the enrichment of Th2 cells and modulating the TME of cancers.
The expression and prognostic values of RAD54L in pan-cancer
To comprehensively evaluate the transcriptional and translational expression levels of RAD54L in pan-cancer, the RNA-seq from TCGA and GTEx database were combined with IHC staining results of HPA database to illustrate the expression levels of RAD54L in multiple cancer types. As displayed in Figure 7a, compared to corresponding normal tissues, the expression of RAD54L were increased in 24 cancer types. Specially, Breast invasive carcinoma (BRCA) (Figure 7b), Cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC) (Figure 7c), Colon adenocarcinoma (COAD) (Figure 7d), Kidney renal clear cell carcinoma (KIRC) (Figure 7e), Liver hepatocellular carcinoma (LIHC) (Figure 7f), Lung adenocarcinoma (LUAD) (Figure 7g), Lung squamous cell carcinoma (LUSC) (Figure 7h), Pancreatic adenocarcinoma (PAAD) (Figure 7i) and Stomach adenocarcinoma (STAD) (Figure 7j) exhibited stronger IHC staining intensity in comparation with their corresponding normal tissues (Figure 7b-7j). Subsequently, the prognostic values of RAD54L in pan-cancer were analyzed, as depicted as a forest plot in Figure 8a, univariate analyses presented that the elevated RAD54L expression tightly linked with unfavorable OS in 13 cancer types, including Adrenocortical carcinoma (ACC), Kidney Chromophobe (KICH), KIRC, Kidney renal papillary cell carcinoma (KIRP), Brain Lower Grade Glioma (LGG), LIHC, LUAD, Mesothelioma (MESO), PAAD, Prostate adenocarcinoma (PRAD), Sarcoma (SARC), Uterine Corpus Endometrial Carcinoma (UCEC) and Uveal Melanoma (UVM), while showed good OS in STAD and Thymoma (THYM). Moreover, the associations of RAD54L expression with PFS in pan-cancer were also explored, as plotted in the forest plots, elevated RAD54L expression showed poor PFS (Figure 8b) in 13 cancer types. These results implied that the expression of RAD54L is remarkably upregulated in multiple prevailing cancers and its elevated expression predicted unfavorable clinical outcomes in various cancers.