Dysregulation and mutation of copper induced cell death related genes in cancers
Copper induced cell death has gradually become the focus of attention for cancer research. We first explored the expression pattern of copper induced cell death related genes across cancers. As shown in Fig. 1A, most copper induced cell death genes were downregulated in various cancers, such as FDX1 in cholangiocarcinoma (CHOL); LIAS and LIPT1 in bladder cancer (BLCA); PDHB in glioblastoma (GBM) and kidney chromophobe cancer (KICH). The results demonstrated that copper induced death was suppressed across cancers. Nonetheless, CDKN2A expression was significantly up regulated KICH and breast cancer (BRCA). The classic regulator of copper induced death FDX1 was investigated and was significantly reduced in renal cancer, liver cancer, gastric adenocarcinoma, and other cancers compared with the normal control (Additional file 1: Fig. S2A). To comprehensively understand the dysregulated expression of copper induced cell death genes, we investigated the characteristics of copy number variation (CNV) and single-nucleotide variation (SNV) across cancers (Fig. 1B). In most cancers, CNV and gene expression were significantly correlated, especially in BRCA, BLCA, HNSC, LUSC and OV. As depicted in Fig. 1C, heterozygous amplifications frequently appeared in DLD, PHDA1, and LIPT1, while heterozygous deletions often occurred in PHDB, FDX1, LIAS, and DLAT. Figure 1D shows the genomic location of copper induced cell death genes, which were distributed on several chromosomes. The SNV frequency of copper induced cell death genes was analyzed, and all 729 samples tested had at least one mutation site (Fig. 1E). CDKN2A, MTF1, DLD, GLS, PDHA1 and DLAT exhibited higher mutation frequencies, and the SNV rate of CDKN2A even exceeded 50%. Cancers with high rates of SNVs included HNSC, LUSC, PAAD, BLCA, LUAD, SKCM and UCEC (Additional file 1: Fig. S2B). Methylation is usually negatively correlated with gene expression. Likewise, the expression of most copper induced cell death genes was negatively correlated with their methylation status, such as PHDB, DLAT, LIPT1 and LIAS (Fig. 2F). Thus, most of the copper induced cell death genes were hypermethylated states in cancers. CDKN2A expression was positively related to methylation levels in HNSC, ESCA, PCPG and LUSC. MTF1 expression in PRAD, COAD, UVM and FDX1 expression in LIHC exhibited positive correlation with methylation status. Hypermethylated PDHB, LIPT1, LIAS and GLS were mainly associated with poor prognosis in cancer, whereas hypomethylated FDX1 was dominantly related to poor prognosis of LGG and ccRCC (Additional file 1: Fig. S2C). Therefore, copy number variation, single nucleotide variation and methylation status together contributed to the dysregulation of copper induced cell death in pan-cancer.
Pathway enrichment analysis and drug sensitivity assessment of copper induced cell death genes
After exploring the fundamental reasons for copper induced cell death gene dysregulation, we analyzed the effect of these genes on cancer signaling pathways. GSEA analyses indicated that copper induced cell death genes were mainly positively related to the MYC, MTOR, G2M checkpoint and oxidative phosphorylation signaling pathways, but negatively correlated with epithelial mesenchymal transition (EMT), myogenesis and TNFA—NFκB (Fig. 2A). Specifically, PDHA1 was associated with cell cycle, hormone AR pathway activation, EMT inhibition and RTK inhibition. LIPT1 was associated with DNA damage response, hormone AR activation, and EMT inhibition. LIAS was enriched in hormone AR activation and apoptosis inhibition. FDX1 was mainly enriched in hormone AR, ER activation and EMT inhibition. MTF1 was related to RTK pathway, RAS-MAPK pathway activation and cell cycle inhibition. CDKN2A was associated with cell cycle activation and RTK, RAS-MAPK inhibition (Fig. 2B). The positive factors related to copper induced death were mainly enriched in cell cycle, metabolism, RAS-MAPK activation and EMT inhibition pathways.
Considering the importance of copper induced death, we determined copper induced death potential index (CPI), similar to ferroptosis in cancer and normal tissues [32]. CPI index was significantly lower in most cancers than in normal tissues, especially in ccRCC, breast cancer and liver cancer (Fig. 2C). In addition, we found that high expression of most of these genes predicted poor prognosis in cancer patients (Fig. 2D), such as FDX1 in LUAD and LGG; LITP1 in SKCM and UVM; DLAT in BLCA and LIHC; and CDKN2A in KICH, ACC, THCA, LIHC, LIHC, KIRC UCEC and COAD. Notably, most genes except CDKN2A indicated a protective function for ccRCC patients. The GDSC and CTRP databases were used to perform drug sensitivity analysis of copper induced cell death-related genes. Spearman correlation analysis showed that high CDKN2A expression exhibited good sensitivity to PD-0332991 and Nutlin-3a. Similar results were found for PDHB to Gefitinib and Afatinib (Fig. 2E). GLS was responsible for GSK-J4 and tivantinib sensitivity (Additional file 1: Fig. S3). LIAS was associated with multiple drug resistance, such as Methotrexate and TPCA-1. GLS was closely related to CHIR-99021, dasatinib and bortezomib resistance. These drug sensitivity results may be developed to find effective targets for cancer treatment. LIA and GLS expression might be responsible for therapy resistance, which needs further research.
Establishment of two clusters by clustering analysis of copper induced cell death-related genes in ccRCC
As we found above, copper-induced cell death-related genes were protective factors for ccRCC, which was unique and different from other cancers. Thus, we investigated the characteristics of copper-induced cell death-related genes in ccRCC. The TCGA ccRCC samples were classified into several subtypes using an unsupervised clustering method based on the expression levels of copper-induced cell death-related genes. The optimal classification method was validated, and the PAC method was used to assess the robustness of the analysis. Consequently, the TCGA ccRCC dataset was divided into two subtypes, namely, copper-pattern cancer type 1 (CPCS1) and type 2 (CPCS2) (Fig. 3A–D). After excluding patients lacking tumor stage and grade information, the clinicopathological characteristics of the 506 ccRCC patients with the two subtypes were compared, as shown in Additional file 2: Table S2. Compared with CPCS1 subtype patients, CPCS2 subtype patients had higher T stage and shorter overall survival (OS) and progression-free survival (PFS) (Fig. 3E, F). We analysed the expression of copper-induced cell death-related genes in ccRCC subtypes and normal kidney tissues. The CSP2 subtype, similar to the desert of copper, induced cell death and expressed lower levels of copper-induced cell death-related genes than the CPCS1 subtype and normal tissues. Conversely, CDKN2A showed higher expression levels in the CPCS2 subtype than in CPCS1. The desert of copper-induced cell death-related genes in CPCS2 contributed to the suppression of copper-induced death, which trained CPCS2 to aggressive clinical subtypes.
Functional enrichment analysis of ccRCC subtypes
Since there were differences in the copper-induced cell death profiles and clinical characteristics between the subgroups, we next investigated the biological function and hallmarks of CPCS1 and CPCS2. GO analysis demonstrated that differentially expressed genes were mainly enriched in cornified envelope, blood microparticle and lipoprotein particle in cellular component; cornification, keratinization, and epidermal cell differentiation in biological process; and serine hydrolase activity, active ion transmembrane transporter activity and anion transmembrane activity in molecular function (Fig. 4A–C). GSEA pathway analysis indicated that these genes were mainly enriched in cellular responses to external stimuli, metabolism of lipids and vesicle-mediated transport pathways (Additional file 1: Fig. S4A). Compared with the CPCS1 subtype, the CPCS2 subtype was more correlated with the PI3K-AKT-mTOR, fatty acid metabolism, oxidative phosphorylation and KRAS and MYC pathways, while the CPCS1 subtype was more related to the myogenesis, EMT, hypoxia and P53 pathways (Fig. 4D). The CSP2 subtype, but not the CPCS1 subtype, was activated in tumorigenicity and cancer progression.
The tumor microenvironment (TME) and metabolic pathways were compared between the CPCS1 and CPCS2 subtypes. The CSP2 subtype was significantly activated in exosome secretion, ferroptosis and extracellular vesicle biogenesis and was inhibited in m6A methylation modification (Fig. 4E). The TME fraction pathways included EMT, CAF, and cytokine signaling pathways. The CSP2 subtype was stimulated for immune checkpoint, CD8 T cells, chemokines, and interleukin receptors. The CSCP1 subtype was associated with mast cells, EMT, WNT and TGF-β receptors (Figure S4B). Metabolic processes play crucial roles in ccRCC [33]. We observed activation of multiple metabolic pathways in CPCS2 subtypes, including amino sugar and nucleotide sugar metabolism, sulfur metabolism, oxidative phosphorylation, glutathione metabolism, cardiolipin metabolism, and valine, leucine and riboflavin metabolism. The CPCS1 subtype was stimulated for inositol phosphate metabolism, lysine degradation, remethylation and terpenoid backbone biosynthesis. The TME and metabolic environment shape the different ccRCC subtypes to a certain degree.
Comparison of immune infiltration characteristics between subtypes
Immunotherapy has gradually become the main characteristic of ccRCC treatment in recent years. To define the effect of copper-induced death on immune profiling, we analysed the immune infiltration environment of the two subgroups using GSVA. Immune-related genes of the CPCS2 subtype showed an overall upwards-regulated trend compared with the CPCS1 subtype. The CSP2 subtype expressed higher levels of CCL5, CCL21, CCL26, CXCR4, CXCR5, CXCR10, IL10RB, LAG3, CD276, CD48, CD70 and TCFRSF8 (Fig. 5A). Similar results were seen in several immune infiltration scoring models, including TIMER and CIBERSORT, QUANTISEQ, XCELL and EPIC. The CSP2 subtype was correlated with CD8 + T cells, Tregs, cancer-associated fibroblasts and NK cells (Fig. 5B). The above investigation confirmed the greater immune infiltration in CPCS2 than in CPCS1. Consistently, ESTIMATE algorithm analysis showed that the CPCS2 subtype contained a higher stromal score, immune score, and ESTIMATE score (Fig. 6A). The specific immune components were compared between CPCS1 and CPCS2 subtypes. The CSP2 subtype lacked activated dendritic cells, which prompted the reduction of identifying tumor cells (Fig. 6B). The CSP2 subtype was less capable of repairing DNA damage (Fig. 6C). The CSPC2 subtype expressed more CD276, IL-6, PDCD1 and TGFB1, while CD274 (PD-L1) expression was significantly downregulated (Fig. 6D). In the antitumour processes, the CPCS2 subtype expressed a higher degree of immune infiltration, but tumor clearance abilities were severely impaired, possibly due to the immunosuppressive states (Fig. 6E). The immune function scores were also evaluated in the two subtypes. The CSP1 subtype gained a higher microsatellite instability (MSI) score, whereas the CPCS2 subtype obtained a higher cancer-associated fibroblast (CAF) score, stemness-associated score, dysfunction score and tumor immune dysfunction and rejection score (TIDE) (Fig. 6F). Nearly half of the cases in CPCS1, a copper-induced death-activated subtype, developed an effective immune response. Conceivably, the C2 subtype may generate more effective immune responses by activating copper-induced death to form abundant presented antigens.
Comparison of tumor somatic mutations and CNVs in two subtypes
In addition to the influence of the microenvironment on drug therapy, genome mutations are also core factors in drug effectiveness. We examined the differential distribution of tumor somatic mutations between the two subtypes. The gene mutations of the CPCS2 subtype were more frequent than those of the CPCS1 subtype (Fig. 7A). The picture depicts the mutation frequency of the top 20 mutant genes. VHL, PBRM1, TTN, SETD2 and BAP1 were the most frequently mutated genes for both subtypes. Compared to CPCS1, the CPCS2 subtype had several genes with a higher mutation frequency, including KDM5C (9% vs 4%), PTEN (6% vs ≤ 4%), XIRP (6% vs ≤ 4%), LPR2 (6% vs ≤ 4%), RYR2/3 (6% vs ≤ 4%) and ANKS1B (5% vs ≤ 4%). The forest analysis also confirmed the above findings. As shown in Fig. 7B, the CPCS2 subtype was more mutated in DNAH11, ZGRF1, ESPL1, LAMC1, RYP2, PTEN and ANKS1B. The fraction of pathways affected in CPCS1 was more frequent than that in CPCS2, while CPCS2 shared a greater fraction of affected samples (Fig. 7C). For instance, the CPCS1 subtype was more frequent than the CPCS2 subtype for the fraction of pathways affected in NOTCH (35/71 vs 17/71), NOTCH (28/68 vs 18/68), Hippo (22/38 vs 12/38) and MYC (6/13 vs 3/13). The CPCS2 subtype was more frequent for the fraction of samples affected in WNT (21/109 vs 51/211). The DGldb database was used to investigate potential therapeutic targets of mutated genes. The potential therapeutic targets of the CPCS1 subtype included ARID1A, ATM, BAP1, KDM5C and KMT2C, while CPCS2 targets were mainly BAP1, FAT3, KDM5C, LAMPCCS1 and mTOR (Fig. 7D). The somatic interactions analysis suggested that comutation of PBRM1 and PKHD1 caused cell death in CPCS1, and comutation of VHL and RYR2/LRP2 also led to death (Fig. 7E). These synthetic lethal mutations could potentially be used to develop treatments for different subtypes. For mutations of copper-induced cell death-related genes, the CPCS2 subtype reserved a higher mutation frequency (4.59% vs 2.37%) and more abundant mutation types than the CPCS1 subtype (Fig. 7F). The DLD gene contained frame deletion, frame insertion and missense mutations in the CPCS2 subtype.
Copy number variations and not just gene somatic mutations were also compared between the two subtypes. For amplification frequencies, the CPCS2 subtype presented higher CNV frequencies than the CPCS1 subtype on chromosomes 3q, 8q, 12p, 12q, 20p, and 20q. Regarding deletion frequencies, the CPCS2 subtype contained significant CNV frequencies on chromosomes 6p, 8p, 9p, 9q, 10q, 11q, 13q, 17p, 18p, 18q, and 22q (Fig. 8A). The amplification and deletion regions on chromosomes were decoded and analysed using GISTIC 2.0 software (Fig. 8B–D, Additional file 3: Table S3). The recurrent CNVs of CPCS1 included amplification of 5q35.2 (CPEB4), 5q31.3 (KCTD16), 5q33.2 (SGCD), 5q15 (NR2F1), 14q13.1 (CFL2, NFKBIA, PSMA6, SRP54, PPP2R3C), and 7q34 (EPHB6, TRPV6) and deletion of 9p21.3 (CDKN2A), 1p36.13 (UBR4), 9p21.3 (CDKN2B), 9p23 (PTPRD), 2q37.1 (ALPI, COL4A3, GPR35, PTPRN), and 3p21.2 (ABHD14A, PARP3, RBM15B). The specific CNVs of CPCS2 were the amplification of 5q35.3 (FGFR4, HNRNPH1, MAPK9, RNF44), 5q23.3 (HINT1), 11q22.2 (MMP7), 3q26.33 (PIK3CA, ZNF639), and 8q24.22 (ADCY8, ADCY8, GPR20) and the deletion of 2q37.3 (AGXT, KIF1A), 9p23 (PTPRD), 1p31.1 (NEGR1), 1p36.11 (C1QA, CD52, GPR3, RUNX3), and 9p21.3 (CDKN2A, CDKN2B) (Fig. 8B–D). Differences in the amplification and deletion of genomic regions may lead to the formation of the two subtypes.
Drug sensitivity analysis of two subtypes
After comprehensively inspecting the clinical characteristics, prognosis, immune profile and mutation information of the two subtypes, we further searched for sensitive targets and potential drugs of the two subtypes based on the above investigations. The GDSC database was used to perform drug sensitivity analysis of the two ccRCC subtypes. Significantly different responses were observed between the CPCS1 and CPCS2 subtypes. The CSP1 subtype was more sensitive to axitinib, crizotinib, pazopanib and temsirolimus, while the CPCS2 subtype displayed sensitivity to dasatinib, erlotinib, lisitinib, saracatinib and gefitinib (Fig. 9A). Although most of them were TKI inhibitors, there were certain differences in the effective targets of CPCS1 and CPCS2 subtypes, which were VEGFR/PDGFR/MET and EGFR/SRC, respectively. We further analysed two subtypes of potential molecular inhibitor drugs. CPCS1 subtype indicated sensitivity to PAC.1, GW.441756, AKT inhibitor VIII, FH535 and Epothione. B, while the CPCS2 subtype was more responsive to CGP.6047, sunitinib, GSK269962A, and LFM. A13 and vinblastine (Fig. 9B). We then investigated potential drugs targeting the oncogenic process. The CellMiner database was utilized to determine the relationships between copper-induced death-related genes and drug sensitivities. Negative correlations were observed between GLS expression and the IC50 of TYROTHRICIN, EMD-534085, Batasertib and PDHA1 expression and the IC50 of LY-3023414 (Additional file 1: Fig. S5A). The results suggested that these were appropriate for ccRCC patients with high expression of GLS and PDHA1. In addition, MI-503, LY-3154567 or KPT-9274 may be suitable for patents with low DLD, CDKN2A or DLAT expression, respectively.
Verification of classification model in external dataset
To further confirm the robustness of the classification model, we performed identification using the GDSC ccRCC cell line database and JAPAN-KIRC cohort. A significant difference was observed in cell lines between CPCS1 and CPCS2 subtypes. Most copper-induced cell death-related genes except CDKN2A were downregulated in CPCS2, similar to the TCGA cohort (Fig. 10A). The nearest template prediction (NTP) algorithm suggested that the dysregulated hallmarks identified from ccRCC subtyping can divide the JAPAN-KIRC cohort into two different groups (Fig. 10B). The CSP2 subtype predicted a poorer prognosis of renal cancer patients than the CPCS1 subtype, which was consistent with previous data (Fig. 10C). Compared to CPCS2, the CPCS1 subtype displayed more sensitivity to most tested drugs (ACY-1215, BX-912, CP466722, ETOPOSIDE, TEMSIROLIMUS, LESTAURTINIB, QS11, KIN001-206, PALBOCICLIB, and SN-38) (Fig. 10D, Additional file 1: Fig. S5B, Additional file 4: Table S4). The results suggested the effectiveness of targeting the CPCS1 subtype.
Construction of a four-copper-induced-cell death-related genes risk model
To further evaluate the reliability of the subtyping model, univariate Cox regression analysis was used to explore dysregulated biomarkers between the two subtype samples (Fig. 11A). The random forest supervised classification algorithm identified the 10 most relevant genes (Fig. 11B). To establish the best risk assessment model, we performed Kaplan‒Meier (KM) analysis, and based on the p value of every model, we finally screened out a risk assessment model composed of four genes (MGAM, PTPRB, PAGE2B, RTL1), named RCC-CUPT4 (Fig. 11C, D). The risk score of each patient was calculated as follows: RCC-CUPT4 = − 7.553304*MGAM-6.184020*PTPRB + 3.895654*PAGE2B + 4.645926*RTL1. To further identify the accuracy of RCC-CUPT4, both TCGA-ccRCC and JAPAN-KIRC cohorts were divided into high-risk and low-risk groups according to median scores (Additional file 1: Fig. S9A). The high-risk group predicted worse OS and PFS than the low-risk group in both cohorts (Fig. 11E, F, Additional file 1: Fig. S6B). The area under the ROC curves also confirmed the high sensitivity and specificity of the RCC-CUPT4 model for predicting prognosis. The AUC scores for the TCGA ccRCC cohort were 0.6956, 0.7175, 0.7041, 0.7154 and 0.705 at 0.5 years, 1 year, 2 years, 3 years and 5 years, respectively (Fig. 11G). Better AUCs were obtained in the JAPAN-KIRC cohort, and the AUC scores were 0.9485, 0.733, 0.8478, 0.7804 and 0.7485 at 0.5 years, 1 year, 2 years, 3 years and 5 years, respectively (Additional file 1: Fig. S6C). The above results confirmed the reliability and practicality of our classification.
The core role of DLAT and functional verification
In view of the importance of copper-induced cell death in ccRCC, we evaluated which gene shared the most importance proportion in clinical outcome. Unlike previous studies, DLAT rather than FDX1 may play the core role among copper death signatures when we performed random forest analysis (Fig. 12A); a similar result was shown in the JAPAN-KIRC cohort (Additional file 1: Fig. S7A). As previously observed, DLAT was downregulated in renal cancer, and the magnitude of regulation increased with the stage (Additional file 1: Fig. S7B, C). The immunohistochemical score of DLAT was downregulated in high-stage ccRCC compared with low-stage ccRCC (p < 0.05) (Fig. 12B, Additional file 1: Fig. S7D). Accordingly, the high expression of DLAT predicted good prognosis of ccRCC and was a protective factor of ccRCC (Fig. 12C). The TIMER database found that DLAT was strongly correlated with neutrophils, macrophages, and DCs (Fig. 12D). In addition, we evaluated the AUC efficacy of DLAT with several classic cancer risk prediction models and the effect of DLAT mutation on immune cell infiltration in ccRCC (Figs. 12E, Additional file 1: Fig. S7F). The AUC curve of DLAT was 0.54 in Miao Kidney (ICB) data, while the scores achieved 1.00 in Zhao glioblastoma (PD-1) and 0.80 in Nathanson melanoma (CTLA4). In addition, DLAT expression may significantly influence multiple immune signatures across cancers, especially in ccRCC (Figure S7G). To verify the function of DLAT in renal cancer cell lines, we constructed a DLAT-overexpressing lentivirus (DLAT). Using the CCK-8 assay, overexpression of DLAT significantly inhibited the proliferation of ccRCC cell lines (p < 0.001) (Fig. 12F). Transwell assays showed that overexpression of DLAT significantly reduced the migration ability of ccRCC cell lines (p < 0.001) (Fig. 12G). In vivo experiments found that upregulation of DLAT inhibited the increase in the volume and weight of xenografts in mice (p < 0.01) (Fig. 12H). These results suggested that upregulation of DLAT expression could effectively inhibit the growth and metastasis of renal cancer. Therefore, it makes sense to prompt copper-induced death by activating DLAT expression to achieve the goal of tumor eradication.
Cuprotosis in ccRCC could enhance tumor immunity though cGAS-STING signaling
The results mentioned above reminded us that CPCS1, an activated cuproptosis phenotype of ccRCC, led to a better prognosis, enhanced antitumor immunity and a higher response rate to ICI therapy. Cuproptosis functioned as a novel programmed cell type. Figure 6C also indicates that the DNA damage repair signal was activated in CPCS1; we thus speculated that cuproptosis could activate tumor immunity through DDR-related signatures, which then stimulated tumor immunity through cGAS-STING signaling. Cell cytotoxicity was performed to choose the effective and safe concentration of cuproptosis inducer agent, and we applied 2 μmol/L as the optimal concentration for the activation of cuproptosis for RENCA, since the cytotoxicity rate of RENCA was higher than CP-M062 at 2 μmol/L (Fig. 13A). Activated cuproptosis significantly inhibited tumor growth in vitro and vivo (Fig. 13B, C, Additional file 1: Fig. S8A, B). Through three independent ccRCC datasets, (TCGA-KIRC, JAPAN-KIRC and Cacner cell-cohort), containing 1439 ccRCC samples, we found that the cuproptosis score calculated by ssGSEA was significantly correlated with the cGAS-STING-related signature, including TMEM173/STING, TBK1, MB21D1/cGAS and IRF3, except for TBK1 in Motzer’s cohort (Fig. 13D). Similar to the results in Fig. 2C, the cuproptosis score was lower in malignant cells in ccRCC and higher in tumor and stromal cells at single cell level (Figure S8C, D). Previous studies have indicated that cGAS-STING signaling was correlated with DC in anti-tumor immunity [34,35,36,37,38]. We thus adopted the three ccRCC cohorts to investigate correlation between cuproptosis score and DC signatures, which also showed that cuproptosis score was significantly positive correlated with DC infiltration score (Figure S9A).
We next applied a co-culture system of DC and renal cancer cells to verify our findings (Fig. 13E). Expression of the cGAS-STING pathway at both the protein and RNA levels was increased in a dose-dependent pattern in DCs cocultured with cuproptosis-activated tumor cells (Fig. 13F, G). In addition, the intracellular activity level of cGAMP in DCs was higher in the cuprotosis-treated group (Fig. 13H). The secretion levels of IL2, TNF-α, IFN-γ, CXCL10 and CXCL11 were increased in medium supernatant from co-culture system (Fig. 13I). 547-553As Fig. 13J indicated that, at 14 days after RENCA cells inoculation in C57BL/6 mice, all mice carrying similar size of tumors were randomized into four groups receiving DMSO, anti-PD-1mAB, cuprotosis inducer reagents, and combined therapy (anti-PD-1 and cuprotosis inducer reagents) by intraperitoneal (i.p., given at days 14, 17, 20), respectively. After one week, all mice treated with those agents were euthanized and peripheral blood were harvested to quantity the percentage of CD8 + T cells among different groups. Finally, injection of cuproptosis induce pre-treated RENCA could increase the percentage of CD45+CD8+ T cells from peripheral blood, and combined therapy with cuproptosis and ICI further increased this percentage (Fig. 13K). All these results suggested that cuprotosis could enhance tumor immunity although cGAS-STING signaling in ccRCC (Figure S9B).