For single-cell RNA-seq, we collected three datasets of ccRCC patients and normal kidney tissues were downloaded from GEO database, including GSE131685, GSE152938 and GSE156632 (26, 27). We integrated all these scRNA through “Harmony” algorithm and gathered total 9 ccRCC and 9 normal kidney samples (28, 29). The standard workflow of cell clustering in Seurat was utilized to identify distinct groups of cells based on the integrated data. In brief, PCA was performed on the scaled data, and the top 20 PCs were used for graph-based clustering to identify cell clusters. Cluster marker genes were identified using “FindAllMarkers” function in Seurat (https://satijalab.org/seurat/) based on the “RNA” assay (30, 31). Next, respective reduction of cell clustering, including UMAP and PCA were performed, and cell cluster was obtained through the UMAP method. Finally, we used the “singleR” package to get the cell type for cell population annotation (32).

Then, for bulk RNA-seq, we integrated the normalized RNA-seq profiles (TPM), matched clinical characteristics and survival information of ccRCC samples and normal kidney samples from The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov) and GTEx database were downloaded (33). Meanwhile, GEO dataset GSE40435 was also applied to analyze (34). Differentially expressed cuproptosis regulators between tumor and normal tissue samples were screened out with the Wilcoxon test and “Limma” R package. Additionally, proteogenomic expression profiles of ccRCC patients was downloaded and pre-processed from CPTAC database and the supplemental materials of Ding’s research (35, 36).

Genes or proteins with false discovery rate (FDR) adjusted P < 0.05 and | log₂FC (fold-change) | > 0.5 were considered as DEGs.

We conducted Gene Ontology (GO) enrichment analysis, which including biological process (BP), cellular components (CC), molecular function (MF), together with Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways to explore the biological functions and underlying signaling pathways. Subsequently, gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA) were performed to evaluate the pathways enriched among “h.all.v7.5.1.symbols.gmt” and “c2.cp.kegg.v7.4.symbols.gmt” gene sets from the molecular signature database (37, 38). We applied “AddModuleScore” algorithm to calculate the copper-induced cell death score in our scRNA datasets.

To exhibit the comprehensive landscape of immune cell infiltration in different subgroups, we conducted several deconvolution algorithm algorithms, including XCELL (39, 40), TIMER (41, 42), QUANTISEQ (43, 44), MCPCOUNT (45), EPIC (46), CIBERSORT (42, 47) and CIBERSORT-ABS (48) to estimate the subpopulations of immunity infiltration scores. Differences between two risk groups were analyzed by the Wilcoxon signed-rank test and the results were obtained according to p-value< 0.05. Subsequently, we used correlation analysis when exploring the relationship between the risk score and immune infiltrated cells.

Renal clear cell carcinoma and adjacent noncancerous renal samples were obtained by radical nephrectomy from the First Affiliated Hospital of Nanjing Medical University (Jiangsu Province Hospital) between 2005 and 2018. All diagnoses were confirmed by senior pathologists independently. Informed consent from all patients was acquired in the study. The study design and protocol were approved by the ethics committee of the First Affiliated Hospital of Nanjing Medical University (Jiangsu Province Hospital). IHC assays were performed as previously described (49, 50). Briefly, the primary antibody was diluted as follows: anti-PDHB (1:100, Abcam, USA).

RCC cell lines (786-O, 769-P, A498, Caki-1) and human renal tubular epithelial cell line (HK-2) were purchased from ATCC and cultured in RPMI 1640 (786-O, 769-P), McCoy’s 5A (Caki-1), DMEM (A498) and DMEM/F12 (HK-2) (Gibco, Thermo Fisher Scientific, USA) containing 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (Gibco, Thermo Fisher Scientific, USA). Small interfering RNA targeting PDHB (si-PDHB), and negative control (shNC) were constructed and transfected. Elesclomol (Selleck, China), a specific copper-induced cell death activator, was also applied. Cells were transfected with si-PDHB and si-NC using Lipofectamine 3000 (Invitrogen, Thermo Fisher Scientific, USA).

Pretreated RCC cells were counted and seeded into a 96-well plate at a density of 1.0x10³ cells/well. Cell proliferation was detected after 24h, 48h, 72h, and 96h using the CCK-8 Cell Counting Kit (Vazyme, China). The absorbance was measured at 450 nm with a microplate reader following incubation at 37°C for 1h according to the manufacturer’s protocols. For the colony formation assay, pretreated cells were seeded into 6-well plates (1000 cells/well). The cells were incubated for 10 days. Colonies were fixed in 4% paraformaldehyde for 20 min, washed with PBS twice, and stained with 0.1% crystal violet for further analysis.

For transwell cell migration and invasion assay, 1.5×10⁵ cells RCC cells were seeded into the 8μm PET membranes 24-well Transwell (Corning, USA) upper chambers with serum-free medium for the migration assays. Medium containing 15% FBS was added to the bottom chamber. After incubation at 37°C for 24 h, the cells were fixed in 4% paraformaldehyde for 20 min and stained with 0.1% crystal violet for 20 min. Cells were captured on a microscope in five randomly selected fields and repeated three times.

Total RNA was isolated using Trizol (Invitrogen, Thermo Fisher Scientific, USA). HiScript III All-in-one RT SuperMix (Vazyme, China) was used for cDNA synthesis. qRT-PCR was performed with SYBR qPCR Master Mix (Vazyme, China) using StepOne Plus (Applied Biosystems, USA) and LightCycler 480 PCR instrument (Roche Diagnostics, Switzerland) according to the manufacturer’s instructions. The primers and siRNA Oligo used were listed in Table S1 .

All mice involved in this research were approved by the Institutional Animal Care and Use Committee (IACUC) of Nanjing Medical University. Briefly, total 2.5 × 10⁷ 786-O cells with knockdown-PDHB (shPDHB) and negative control cells were collected and suspended with PBS and Matrigel (1:1, Corning, USA), then subcutaneously injected into 4-week-old female BALB/c nude mice. The formula of tumor volume was calculated as follows: Tumor volume= (length*width)²/2.

All analyses were performed using GraphPad Prism software and R 4.2.2. All statistical tests were two-sided, and P-value <0.05 was considered statistically significant unless otherwise noted. Continuous variables in normal distribution were between-group compared through the independent Student’s two-tailed t-test, while continuous variables in skewed distribution through the Mann-Whitney U test. Spearman order correlation analysis was used to determine the relationship between different subgroups. The differences in clinical outcomes were calculated with the Log-rank test through the Kaplan-Meier method. The univariate regression model was constructed to analyze the effect of each variable on the survival. All experiments were repeated independently three times. Data are shown as the mean ± standard deviation (SD). The P-value < 0.05 was considered to be statistically significant.

Firstly, after summary copper-induced cell death regulators by genomic-wide CRISPR-Cas9 loss of function screening results, total 10 copper-induced cell death regulators were categorized into two groups: cuproptosis resistances (FDX1, LIAS, LIPT1, DLD, DLAT, PDHA1, and PDHB) and cuproptosis sensitizers (MTF1, GLS, and CDKN2A). For investigating the activity of copper-induced cell death across human cancers, single sample gene set enrichment (ssGSEA) algorithm was applied to calculate the cuproptosis score (CPS) based on the gene expression from the TCGA database. We found that CPS is significantly downregulated in the majority cancers ( Figure 1A ). Consistently, CPS was dramatically decreased in paired samples among human cancers ( Figure 1B ). These results confirmed that CPS based on different approaches or cancer subtypes is robust. We observed the aberrant expression patterns of these cuproptosis regulators ( Figure 1C ). Meanwhile, we visualized somatic copy number alterations (SCNA) frequency and the expression of these cuproptosis regulators in TCGA pan-cancer cohort ( Figure 1D ). Additionally, we analyzed the association between these cuproptosis regulators and overall survival outcomes by log-rank test and Cox regression. Interestingly, high expression of these cuproptosis resistances revealed favorable overall survival ( Figure 1E ). Functional enrichment further demonstrated that these intersecting genes were mainly enriched in glyoxylate metabolism and glycine degradation, metabolic reprogramming, and biosynthesis of cofactors ( Figure 1F ). These suggested that transcriptional alternations and genetic mutation of cuproptosis regulators are probably the underlying mechanisms leading to perturbations in copper-induced cell death.

Based on above results, we discovered that ccRCC tissues had significantly lower copper-induced cell death index compared to adjacent normal tissues. Moreover, these cuproptosis regulators are abnormally expressed and associate with clinical outcome in ccRCC. Accordingly, we chose ccRCC for further research. First, we applied renal cancer cell lines to elesclomol exposure, a specific copper-dependent cell death activator. The cell viability of 786-O, Caki-1, A498 and 769-P cells decreased significantly after elesclomol treatment, which exhibited a concentration-dependent effect ( Figure 2A ). To further investigate the expression pattern of these cuproptosis-related genes in ccRCC, we quired TCGA-KIRC cohort to compare the transcriptional alternations, which was illustrated in a heatmap ( Figure 2B ). In addition, we analyzed the correlation between the expression of different genes and survival outcomes in ccRCC patients, which revealed strong associations. Among them, FDX1, DLAT, DLD, LIAS and PDHB showed positive correlation in cuproptosis resistances subgroup ( Figure 2C ). We also observed PDHA1, PDHB, FDX1, GLS, DLD, DLAT, LIAS and LIPT1 were down-regulated in ccRCC tumor tissues, while CDKN2A exhibited higher protein expression among tumor samples in CPTAC-ccRCC database ( Figure 2D ).

In consideration of the heterogeneity of ccRCC, we applied single-cell RNA sequence (GSE) for further validation. Firstly, we after using “Harmony” to remove batch effects, we gathered a total of 33 clusters by UMAP algorithm ( Figure S1A ). Then, we explored the distribution of copper-induced cell death scores by single-cell signature scorer and found that overwhelming majority copper-induced cell death signature was enriched in normal samples compared with ccRCC tumors by “AddModuleScore” algorithm (P<0.001; Figures 2E, F ). Marker genes between each cluster were calculated and illustrated in Figure S1C . ccRCC and normal kidney samples could mainly be divided into epithelial cells (Malignant tumor cells), endothelial cells, Myeloid cells, Mast cells, T cells, Fibroblast cells and Fibroblast_Endothelial_like cells ( Figures S1B , S2E ). Finally, we tried to explore the exact distribution of cuproptosis regulator genes in ccRCC tissues. Specifically, we conducted single-cell analysis to demonstrate and validate the detail changes of immune composition alternations and found that higher proportion of T and Myeloid cells among patients with high expression of PDHB, especially in patient sample T1, T6 and T9. In contrast, among low PDHB expression patients, we observed lower immune cell infiltration, such as T cells and myeloid cells ( Figure 2G ). The results showed that cuproptosis regulator genes were mainly concentrated in hepatocytes and epithelial cells, indicating their vital role in immune cell infiltration among tumor microenvironments.

To further identify the essential cuproptosis regulators in ccRCC, we combined expression and prognostic analysis to detect these candidates ( Figure 3A ). We gathered four differently expressed cuproptosis regulators (FDX1, PDHB, PDHA1 and CDKN2A) among TCGA-KIRC cohort and two regulators among GSE40435 cohort ( Figures 3C, D ). Meanwhile, the prognostic value of these cuproptosis-related genes was calculated by the ROC curve, which illustrated the AUC value of CDKN2A was 0.991 (95%CI: 0.982-1.000); FDX1 was 0.965 (95%CI: 0.946-0.983), and the AUC value of PDHB was 0.956 (95%CI: 0.933-0.979) ( Figure 3B ). Next, we conducted univariate Cox regression analysis and identified total 9 prognostic genes ( Figure 3E ). Taken together, the multi-omics analyses confirmed that PDHB might be the key gene involved in copper-induced cell death. ( Figure 3F ).

We first analyzed the transcriptional profiles of PDHB in ccRCC through TCGA and GTEx databases. PDHB expression was significantly lower in ccRCC tumor tissues compared with normal kidney tissues both in TCGA-ccRCC and TCGA+GTEx ccRCC cohorts (P<0.001; Figures 4A, B ). This result was also validated in both TCGA paired samples and (P<0.001; Figure 4E ). ccRCC samples of our NJMU cohorts, ( Figure 4F ). The ROC curve was also applied to assess the prognostic of PDHB. The AUC value of PDHB in TCGA-ccRCC cohort was 0.956 (95% CI: 0.933-0.979) and 0.844 (95% CI: 0.784-0.903) in TGCA+GTEx database ( Figures 4C, D ). Moreover, PDHB protein expression level was significantly downregulated in ccRCC tissues from CPTAC and Chinese FUSCC cohort ( Figures 4G,H ). Ultimately, its tissue abundance was measured using IHC both in HPA database and our clinical samples, which achieved consistent results from above achieved ( Figures 4I, J ). We also evaluated the association between PDHB expression and clinicopathological features. As shown in Figure 4K , the expression of PDHB in patients with lower stage (Stage I-II) was found significantly higher compared to patients who were highe stage (Stage III-IV) level (P<0.05; Figure S2A ). The distribution of PDHB showed a significant difference among the T classification. PDHB was highly expressed in T1-2 patient compared with T3-4 patient (P<0.05; Figure S2B ). Similarly, PDHB was decreased with advanced M classification ( Figure S2C ). We also performed qRT-PCR experiments to detect the expression level of PDHB in ccRCC or normal kidney cell lines and found that PDHB was down-regulated in ccRCC cell lines (P<0.05; 786-O, 769-P, Caki-1 and A498) compared with normal kidney cell line HK2 ( Figure 4L ). Additionally, about the subsequent analyses, median cut was used to dichotomize 539 individuals into high-PDHB (n=270) and low-PDHB (n=269) subgroup based on mRNA expression level. As shown in Table 1 , PDHB was significantly correlated with the pathologic stage and T classification (P<0.001). Furthermore, logistic regression analysis was adopted to describe the exact correlativity between PDHB expression and clinicopathological characteristics ( Table 2 ). Taken together, above results suggested that PDHB played a vital role in ccRCC.

From TCGA-ccRCC database, we found that the patients with low level of PDHB displayed poor prognosis in overall survival (OS), disease-specific survival (DSS) and progression free interval (PFI) ( Figures 5A–C ). In addition, conducting univariate and multivariate cox regression, PDHB could serve as an independent predictive marker for ccRCC patients’ overall survival (Univariate: HR=0.553, 95% CI=0.407−0.751, P<0.001; Multivariate: HR=0.696, 95% CI=0.503-0.963, P=0.029), revealing that low levels of PDHB expression were correlated with shorter OS ( Figure 5D ). Moreover, a nomogram based on age, gender, pathologic stage, and PDHB was developed to predict the 1-, 3-, and 5-year OS for individual ccRCC patients. ( Figure 5E ). Additionally, by analyzing PDHB expression from IHC tissue microarray staining from our NJMU ccRCC cohort (N=90), we divided patients into PDHB-high and PDHB-low subgroups ( Figure 5F ). Kaplan-Meier survival curves demonstrated that low expression of PDHB were correlated with shorter overall survival (P=0.022; Figure 5G ). Therefore, these findings suggested that PDHB might serve as an indicator for the clinical prognosis of ccRCC patients.

PPI network was constructed and illustrated in ComPPI database ( Figure 6A ). We next investigated the difference among biological function, hallmarks and pathways involved. GSVA analysis demonstrated that oxidative phosphorylation, adipogenesis, mTORC1 signaling and fatty acid metabolism pathway was significantly enriched in PDHB low subgroup ( Figure 6B ). GSEA analysis also acquired similar enrichment of Hallmark bile acid metabolism and apical surface signature in PDHB-low subgroup ( Figures 6C, S3A, B ). By calculating PDHB co-expressed genes ( Figures S3A, B ), we found that co-expressed genes were involved in tRNA processing and GDP binding pathway. These findings confirmed that down-regulated PDHB was mainly participated in metabolism-related pathway.

We applied ssGSEA algorithm to evaluate 24 types of immune cell infiltration level among tumor microenvironment (TME). Detailed characteristics of immune cell proportion and differences between PDHB expression subgroup was further identified. The results revealed that regulatory T cells (Treg) cells were dramatically increased in low-PDHB patients, which demonstrated a suppressive tumor immune microenvironment ( Figure 6D ). Then, we investigated the component of immune cell and stromal cell using ESTIMATE algorithm. As shown in Figure 6E , high-PDHB expression subgroup tend to illustrate more immune and stromal cell infiltration phenomenon. Additionally, correlation analysis indicated that Treg, cytotoxic cells, NK CD56bright cells and T cells was negatively correlated with PDHB expression level, while mast cells, Tgd, DC, eosinophils and Tfh cells has a significant positive correlation with PDHB ( Figure 6F ). Additionally, we found negative association between the expression level of PDHB and Treg cell marker FOXP3 in TCGA-ccRCC cohort. Ultimately, we performed IHC assays to detect the expression level of PDHB and Treg cell marker FOXP3, which found a negative correlation between PDHB and Treg cells in our NJMU ccRCC cohort ( Figure 6I ). Above results suggested that low expression of PDHB plays a vital role in regulating suppressive tumor immune microenvironment mainly via up-regulating Treg cells by ssGSEA and CIBERSOFT algorithms ( Figure 6G ). We assessed the IPS score among different PDHB expression level patients, which could predict the response to immunotherapy. Among them, IPS score of CTLA-4 block therapy and CTLA4+PD-1 combined block therapy was significantly increased in high-PDHB group ( Figure 6H ). These results illustrated low-PDHB expression patients are more likely to benefit from CTLA-4 block therapy and CTLA4+PD-1 combined block immunotherapy.

To further determine the biological oncogenic role of PDHB in ccRCC, PDHB-was knocked down in 786-O and Caki-1 cell models and validated by qRT-PCR and western blotting ( Figure 7A , P< 0.05). Cell counting kit-8 (CCK-8) assay indicated that PDHB knockdown significantly increased cell proliferation ability ( Figure 7B ). Colony formation assay was also employed to determine the long-term impact of PDHB on cells proliferation. We observed higher colony-formation efficiency in PDHB knockdown group than control group both in 786-O and Caki-1 cell lines accordingly ( Figures 7C, D ). Ultimately, in vivo experiment confirmed that knocking down PDHB dramatically accelerated tumor growth ( Figure 7E ). In addition, Transwell migration assay and wound healing assay demonstrated that knockdown PDHB increased the migration ability of RCC cells ( Figures 7F–I ). These findings corroborated that PDHB was essential for ccRCC proliferation and metastasis in vitro and in vivo.

Currently, sunitinib is a first line recommended clinical treatment drug that targets multiple RTKs, such as VEGFR2 (Flk-1) and PDGFRβ (51, 52). Considering that sunitinib resistance is still a common challenge for targeted therapies among renal cell carcinoma (53), we sought to investigate the combined therapeutic strategies to overcome sunitinib resistance in RCC cells. Importantly, the combination therapy of elesclomol and sunitinib profoundly suppressed the proliferation ability of ccRCC cells in a synergistic manner, as demonstrated by the HSA and Bliss synergy scores (786-O: ZIP-score: 17.89, Bliss-score: 17.79; Caki-1: ZIP-score: 11.93, Bliss-score: 11.88; Figures 7J, K ). Ultimately, among the most synergistic area score also demonstrated that elesclomol and sunitinib could suppress ccRCC cells proliferation synergistically (Most synergistic area score: 786-O: ZIP-score: 26.75, Bliss-score: 27.31; Caki-1: ZIP-score: 21.05, Bliss-score: 20.27; Figures 7J, K ).