3.1. Construction of a prognostic risk model for ccRCC based on CDRLRs
Using the STRING database, we constructed a network relationship map of CDRGs (Fig. 2A) and demonstrated the association of cuproptosis and disulfidptosis death-related genes. In accordance with published literature and TCGA-KIRC data, 21 CDRGs were identified via co-expression analysis (Fig. 2B). Following Pearson analysis, 247 eligible lncRNAs were established with parameters |Pearson R|> 0.5 and p < 0.001. Univariate Cox regression analysis was applied to identify 108 lncRNAs significantly correlated with OS (p < 0.05, Fig. 2C). To prevent overfitting, LASSO regression was utilized, mitigating lncRNAs with a high correlation to prognosis. Subsequent multifactorial Cox regression resulted in the selection of four CDRLRs (Fig. 2D,E). The risk models were constructed using ACVR2B-AS1, AC095055.1, AL161782.1, and MANEA-DT. The corresponding risk score equations for ccRCC patients are provided below:
Risk score = (-0.406602113479572 ×ExpACVR2B-AS1) + (-0.988256841487476 ×ExpAC095055.1) + (-0.526107034426687×ExpAL161782.1) + (0.988504048700137 ×ExpMANEA-DT)
Subsequently, we generated a correlation heatmap to visualize the associations between four CDRLRs and CDRGs. This heatmap revealed that ten CDRGs—specifically, OXSM, NUBPL, NDUFS1, NCKAP1, LRPPRC, LIAS, GCSH, DBT, ATP7B, and ATP7A—showed a strong correlation with CDRLRs (Fig. 2F).
3.2. Intergroup validation of prognostic risk models
Median risk scores were computed based on CDRLRs. Subsequently, the training set, test set and the entire set were divided into high- and low-risk groups for survival analysis. This revealed an increasing mortality rate among ccRCC patients in correlation with escalating risk scores (Fig. 3A-F). In the low-risk group, ACVR2B-AS1, AC095055.1, and AL161782.1 exhibited significant expression, whereas MANEA-DT was highly expressed in the high-risk group (Fig. 3G-I). This reinforced that ACVR2B-AS1, AC095055.1, and AL161782.1 are beneficial prognostic factors and MANEA-DT is a poor prognostic factor. The high-risk group demonstrated significantly lower OS compared to the low-risk group (Fig. 3J-L). These findings were confirmed in the three data sets.
3.3. Independent prognosis of risk scores
A comparison of survival probability among ccRCC patients in high- and low-risk groups, based on patient age, gender, histological grade and tumor stage, revealed that the risk score effectively assessed prognosis across all these clinical characteristics (p < 0.01, Fig. 4A-H). Risk scores were indeed adept at predicting OS in ccRCC patients, independent of clinical characteristics. After univariate and multivariate Cox regression analyses, risk scores and clinical characteristics such as age, histological grade and tumor stage emerged as independent prognostic factors in ccRCC patients (p < 0.01, Fig. 5A-B). Factors correlating negatively with prognosis were excluded, and nomogram plots were drawn based on independent prognostic factors (Fig. 5C). Calibration curves were subsequently used to verify the reliability of these findings, which revealed a C-index value of 0.783 (95% CI: 0.7750–0.816) (Fig. 5D). The 1-year, 3-year, and 5-year subject operating characteristic curves (ROC) were then plotted, with an area under the curve (AUC) of 0.725, 0.718, and 0.762, respectively (Fig. 5E). When the AUC values of clinical characteristics within each group were compared, the risk score's AUC was 0.718, second only to tumor stage (Fig. 5F). The 10-year Concordance index further validated these findings (Fig. 5G). These results suggest that the risk score surpasses other clinical characteristics, except tumor stage, as a factor in assessing prognosis. Thus, the risk score can effectively serve as a biomarker for predicting ccRCC patient prognosis.
3.4. GO and GSEA of high- and low-risk groups
From the high- and low-risk groups, we identified 683 DEGs that met the selection criteria (Padjust < 0.05, |log2 (fold change)| ≥ 1). These DEGs were used for functional and pathway enrichment analyses to explore potential biological differences between the groups (Fig. 6A). The Gene Ontology (GO) enrichment analysis showed an enrichment of biological processes (BP), such as antigen binding and immunoglobulin receptor binding. Cellular components (CC) involving the immunoglobulin complex and the external side of the plasma membrane were also enriched, along with molecular functions (MF) like humoral immune response and immunoglobulin production (Fig. 6B). Utilizing Gene Set Enrichment Analysis (GSEA), we identified differences in pathways between the risk groups, with 89 signaling pathways significantly enriched (p < 0.05). The high-risk group displayed significant enrichment in the top five pathways: complement and coagulation cascades, drug metabolism by cytochrome P450, metabolism of xenobiotics by cytochrome P450, retinol metabolism and steroid hormone biosynthesis (Fig. 6C). Conversely, the low-risk group exhibited enrichment in the following top five pathways: endocytosis, insulin signaling pathway, neurotrophin signaling pathway, pathways in cancer, and valine leucine and isoleucine degradation (Fig. 6D). These enrichment patterns may offer valuable insights into the prognostic differences observed between the high- and low-risk groups.
3.5. Immune cell infiltration and immunotherapy sensitivity
Significant differences were observed in TME scores, notably ESTIMATE scores (p < 0.001) and immune scores (p < 0.001), among ccRCC patients, with the high-risk group presenting notably higher scores than the low-risk group (Fig. 7A). To explore potential relationships in immune cell infiltration between the risk groups, we compared 22 immune cell enrichment scores and 29 immune-related function enrichment scores. Using the CIBERSORT algorithm, we created an immune infiltration landscape for the high- and low-risk groups. The correlation box line plot illustrated significant associations between multiple immune cells and risk scores (Fig. 7B). High-risk groups showed an enrichment of T cells CD8, T cells follicular helper and T cells regulatory (Tregs) (p < 0.001), while the low-risk group exhibited a significant upregulation of T cells CD4 memory resting, Macrophages M1, Macrophages M2, and Mast cells resting (p < 0.001, Fig. 7C). Our immune function analysis indicated that the risk models demonstrated significant discrepancies across multiple immune function scores, including the checkpoint (p < 0.001, Fig. 7D). Guided by these immune function analyses, we compared the differential expression of five key immune checkpoints (PD1, PD-L1, CTLA-4, IL-6, LAG3) using the IPS. Results showed that except for PD-L1, which was highly expressed in the low-risk group, all other checkpoints were overexpressed in the high-risk group (p < 0.001, Fig. 7E-I). This suggests the IPS 's potential in predicting the immune response to checkpoint inhibitors in ccRCC patients based on risk score grouping. The immune efficacy, predicted by PD-1 and CTLA-4 expression in the TCIA database, yielded significantly different risk scores in the ctla4(-) pd1(+), ctla4(+) pd1(-), and ctla4(+) pd1(+) groups (p < 0.05). Higher scores were found in the high-risk group, indicating that high-risk ccRCC patients demonstrated a heightened sensitivity to PD-1 and CTLA-4 single-agent and dual-agent combination immunotherapies (Fig. 7J-M).
3.6. TMB prognostic analysis and potential drug sensitivity
To explore somatic mutations within the high- and low-risk groups, TCGA-KIRC mutation data were downloaded and categorized. The results demonstrated identical 15 driver genes with the highest mutation frequencies in both groups, with STED2 and BAP1 hypermutations being more prevalent in the high-risk group (Fig. 8A,B). Although no statistically significant association was found between the risk groups and TMB (p = 0.11, Fig. 8C), both TMB grouping and TMB + risk grouping significantly differentiated survival statuses of ccRCC patients (p < 0.001, Fig. 8D,E). Here, the High-TMB + high risk group exhibited the lowest overall survival rate, while the Low-TMB + low risk group showed the highest. Thus, a combination of the risk score and TMB presents a promising prognostic marker for patients. Several common drugs were selected to analyze their sensitivity in the risk groups. The results indicated that Alpelisib, Ipatasertib, Lapatinib, Selumetinib and Pictilisib demonstrated higher sensitivity in the low-risk group, whereas AZD4547 showed high sensitivity in the high-risk group (p < 0.0001, Fig. 8F-K).
3.7. Differential expression and prognostic validation of CDRLRs in ccRCC
To further examine CDRLRs' expression in ccRCC, we utilized two ccRCC cell lines (769-P, Caki-1), with normal renal tubular epithelial cells (HK-2) as a control. RT-qPCR evaluated the mRNA expression levels of the four key CDRLRs in these cell lines. The findings demonstrated a significantly higher expression of AC095055.1 in both 769-P (p < 0.05) and Caki-1 (p < 0.01) cell lines compared to the HK-2 cell line (Fig. 9B). However, AL161782.1 showed significant expression only in the 769-P cell line (p < 0.0001, Fig. 9C). Conversely, ACVR2B-AS1 (p < 0.01) and MANEA-DT (p < 0.0001) expressions were significantly lower in 769-P and Caki-1 cell lines compared to HK-2 (Fig. 9A,D). To corroborate the independent prognostic role of CDRLRs in ccRCC patients, we performed a prognostic analysis of ACR2B-AS1 and MANEA-DT using the KM Plotter database (AC095055.1 and AL161782.1 were not found in the database). The results identified ACR2B-AS1 as a protective prognostic factor (HR = 0.48 (0.35–0.65), p < 0.0001), while MANEA-DT (HR = 2.05 (1.51–2.79), p < 0.0001) indicated a poor prognosis (Fig. 9E,F). This prognosis based on CDRLRs aligns with the survival analysis results from external databases.