Integrative pan-cancer analysis of cuproplasia-associated genes for the genomic and clinical characterization of 33 tumors

Abstract

Background:

The molecular mechanisms driving tumorigenesis have continually been the focus of researchers. Cuproplasia is defined as copper-dependent cell growth and proliferation, including its primary and secondary roles in tumor formation and proliferation through signaling pathways. In this study, we analyzed the differences in the expression of cuproplasia-associated genes (CAGs) in pan-cancerous tissues and investigated their role in immune-regulation and tumor prognostication.

Methods:

Raw data from 11,057 cancer samples were acquired from multiple databases. Pan-cancer analysis was conducted to analyze the CAG expression, single-nucleotide variants, copy number variants, methylation signatures, and genomic signatures of micro RNA (miRNA)–messenger RNA (mRNA) interactions. The Genomics of Drug Sensitivity in Cancer and the Cancer Therapeutics Response Portal databases were used to evaluate drug sensitivity and resistance against CAGs. Using single-sample Gene Set Enrichment Analysis (ssGSEA) and Immune Cell Abundance Identifier database, immune cell infiltration was analyzed with the ssGSEA score as the standard.

Results:

Aberrantly expressed CAGs were found in multiple cancers. The frequency of single-nucleotide variations in CAGs ranged from 1% to 54% among different cancers. Furthermore, the correlation between CAG expression in the tumor microenvironment and immune cell infiltration varied among different cancers. ATP7A and ATP7B were negatively correlated with macrophages in 16 tumors including breast invasive carcinoma and esophageal carcinoma, while the converse was true for MT1A and MT2A. In addition, we established cuproplasia scores and demonstrated their strong correlation with patient prognosis, immunotherapy responsiveness, and disease progression (P <0.05). Finally, we identified potential candidate drugs by matching gene targets with existing drugs.

Conclusions:

This study reports the genomic characterization and clinical features of CAGs in pan-cancers. It helps clarify the relationship between CAGs and tumorigenesis, and may be helpful in the development of biomarkers and new therapeutic agents.

Introduction

Tumor growth and metastasis are the most significant contributors to mortality in patients with cancer.^[1] The severity of tumor growth and metastasis is promoted by angiogenesis.^[2] The element copper is essential for both cellular proliferation and tumor angiogenesis.^[3] Cuproplasia is defined as the regulation of copper-dependent cellular proliferation.^[4] Consequently, cuproplasia is particularly important in cancer because of the increasing demand for copper in tumor growth and metastasis.^[4] Cuproplasia is associated with various cellular processes, including mitochondrial respiration, antioxidant defenses, redox signaling, kinase signaling, autophagy, and protein quality control.^[4,5] The range and number of cuproplasia-associated genes (CAGs) have been identified in previous studies.^[4] Currently, the expression, functional epigenetic characteristics, and microRNA (miRNA)–messenger RNA (mRNA) interactions of CAGs remain elusive.

In this study, we analyzed the differences in the expression of CAGs in pan-cancerous tissues using several databases, including The Cancer Genome Atlas (TCGA). We preliminarily investigated the value of CAGs in predicting tumor prognosis, and their involvement in regulating multiple tumor-related immune responses. We also analyzed the epigenetic regulation of CAGs and its impact on prognostic outcomes. The targeting of CAGs may emerge as one of the most promising cancer treatment strategies.

Methods

Data download and processing

First, the University of California Santa Cruz Xena, Genomic Data Commons (GDC) and TCGA databases were used to download 11,057 samples comprising 33 tumor types.^[6] Data on the expression profiles, survival, copy number variations (CNVs) (n = 11,495), single-nucleotide variation (SNV) (n = 8663), and methylation (n = 10,129) of CAGs were obtained. In addition, details of patient demographics, follow-up, tumor pathology, staging, and outcomes were retrieved. A total of 36 CAGs were identified based on a previous study.^[4] CAG expression data for normal human tissues were downloaded from the Genotype-Tissue Expression (GTEx) database, and the transcripts per million method was used to normalize gene expression levels. The R package “limma” was used to analyze the data.^[7] Tumor-infiltrating immune cells were calculated by using the gene expression deconvolution algorithm.^[5,6]Figure 1 shows the study flowchart.

Figure 1.

Workflow of this study. CAG: Cuproplasia-associated genes; CNV: Copy number variation; GTEx: Genotype-Tissue Expression; miRNA: microRNA; SNV: single nucleotide variation; TCGA: The Cancer Genome Atlas.

Survival analysis

The association of clinical data with CAG expression data was analyzed using the R packages “survminer” and “survival.”^[8] According to the median expression level on RNA-Seq by Expectation-Maximization (RSEM), patients were divided into two groups for survival analysis, high and low expression. Survival-related hazard ratios (HRs) were computed using the Cox proportional hazards model. A log-rank univariate test was carried out for each gene, and a Kaplan–Meier curve was constructed. Genes with a P-value <0.05 for the log-rank test were retained.

SNV and CNV analyses

The SNV data were collected from the TCGA database (n = 8663). The downloaded data included the following variant types: Missense_Mutation, Silent, 5′ Flank, 3′ untranslated region (UTR), RNA, In_Frame_Del, Nonsense_Mutation, Splice_Site, Intron, 5′ UTR, In_Frame_Ins, Frame_Shift_Del, Nonstop_Mutation, 3′ Flank, Frame_Shift_Ins, and Translation_Start_Site. The number of mutated samples/tumor samples was used to calculate the frequency (percentage) of SNV mutations in the coding region of each gene. The SNV oncoplot was processed using the R package “maftools.”^[9] The Genomic Identification of Significant Targets in Cancer (GISTIC) algorithm version 2.0 was used to retrieve and synthesize raw CNV data, and combine it with the reference human genomic data to obtain genetic CNVs.^[10] To determine the percentage of amplification and deletion of CAGs in pan-cancers, we generated percentage statistics based on CNV isoforms with GISTIC-treated CNV data. Next, we checked for correlations between the original CNV data and the mRNA-RSEM data. Only genes with CNVs >5% were considered significant variants.^[11,12] Pearson's product-moment correlation coefficient was used to analyze the association between the mRNA levels and the CNV.^[13]

Methylation analysis

Methylation errors and mutations in regulatory regions have been considered crucial regulators of tumor gene expression.^[14] As a part of this investigation, we collected DNA methylation data from TCGA. A Student's t-test was performed to identify the methylation differences between tumor and normal tissue samples, and a false discovery rate (FDR) adjusted P <0.05 was considered statistically significant. Utilizing a previously reported method,^[11] we combined the overall survival (OS) and methylation data. Further, hypo- and hypermethylated samples were classified using the mean methylation levels as cut-off points. Survival analysis was performed with the Cox proportional-hazards model, using the R package “survival,” and the Cox coefficient for the methylation levels in each cancer was determined.

Pathway activity and drug sensitivity analyses

The Reverse Phase Protein Array data on TCGA patients were obtained from The Cancer Proteome Atlas Portal (https://tcpaportal.org/tcpa). The pathways of interest included the tuberous sclerosis complex (TSC)/mammalian target of rapamycin (mTOR), receptor tyrosine kinases (RTK), RAS/mitogen-activated protein kinase (MAPK), phosphatidylinositol 3 kinase (PI3K)/protein kinase B (AKT), estrogen receptor (ER), androgen receptor (AR), epithelial–mesenchymal transition (EMT), DNA damage response, cell cycle, and apoptosis pathways, all of which are known to be associated with oncogenesis, cancer progression, and metastasis. By dividing gene expression values by the median, two equal groups were derived. The significant difference in the pathway activity score between the groups was determined by a t-test, with an FDR-corrected P <0.05 considered statistically significant.

Gene–drug interaction databases, such as the Cancer Therapeutics Response Portal (CTRP) and the Genomics of Drug Sensitivity in Cancer (GDSC), facilitate the identification of potential biomarkers by correlating drug responses with well-characterized genomic features.^[15] In this study, these databases were used to determine the half-maximal inhibitory concentration (IC₅₀) of a drug, a value that reflects its effectiveness. The immunotherapeutic response was predicted by tumor immune dysfunction and exclusion algorithms.^[16] Pearson's correlation analysis was used to determine the correlation between mRNA levels and drug IC₅₀ values. P-values were adjusted using FDR <0.05.

miRNA regulatory network analysis

A total of 11,057 tumor samples representing 33 cancer types were obtained from GDC and TCGA database. Only recorded miRNA pairs of genes were used for expression correlations analysis. Next, the miRNA data were linked with the gene expression data. Pearson's correlation coefficient was used to compute correlations between paired mRNA and miRNA levels, and a t-test was used to evaluate their significance.^[17] Paired correlations were calculated after selecting a cutoff value of 0.05 for FDR-corrected P-values. Transcription factors were identified as positive regulators, while negatively correlated miRNA-gene pairs were considered potential negative regulators. Meanwhile, we used nine miRNA–mRNA interaction databases (ENCORI, miRDB, miRWalk, RNA22, RNAInter, TargetMiner, TargetScan, NPInter, and miRTarBase) to assess the interaction between genes and miRNAs.

Results

Gene expression and typing analysis of CAGs

The mRNA expression of CAGs was assessed in normal human tissues from the GTEx database. The CAG expression was found to be widespread, with VEGFA, SOD1, SLC25A3, and MT2A being highly expressed in normal tissues, and MT2A being most highly expressed in the liver [Supplementary Figure 1]. Subsequently, using TCGA expression data, we tested the differential expression of CAGs in a pan-cancer analysis: CAGs were aberrantly expressed in 20 different types of tumors [Figure 2A]. For example, cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC) and head and neck squamous cell carcinoma (HNSC) had a low expression of CP, while kidney renal clear cell carcinoma (KIRC) and lung adenocarcinoma (LUAD) had a high expression. There was also remarkable heterogeneity in the expression of CAGs in the same tumor, with MT1A and DBH demonstrating low levels of expression in cholangiocarcinoma (CHOL), and LOXL2, high levels in CHOL. This implies, therefore, that the different expression of CAGs may be involved in the distinct mechanisms of tumor development. We combined groups of patients with high and low CAG expression and assessed the association of CAG expression with patient survival outcomes. Survival rates were classified into OS, progression-free survival (PFS), disease-specific survival (DSS), and disease-free survival (DFS). A number of these genes, in particular, TYR and SLC31A1, were associated with poor prognoses in the majority of tumors, thereby suggesting the potential link between CAG dysregulation and tumorigenesis [Figure 2B].

Figure 2.

(A) The heatmap shows the expression profiles of the CAGs in the TCGA dataset. (B) Survival analysis of CAGs. The size of the dots represents the significance of the gene's influence on survival for each cancer type, and the statistical significance of differences was determined by Cox regression analysis. ACC: Adenoid cystic carcinoma; BLCA: Bladder urothelial carcinoma; BRCA: Breast invasive carcinoma; CAGs: Cuproplasia-associated genes; CESC: Cervical squamous cell carcinoma and endocervical adenocarcinoma; CHOL: Cholangiocarcinoma; COAD: Colon adenocarcinoma; DFI: Disease free interval; DLBC: Diffuse large B-cell lymphoma; DSS: Disease specific survival; ESCA: Esophageal carcinoma; FC: Fold change; FDR: False discovery rate; GBM: Glioblastoma; HNSC: Head and neck squamous cell carcinoma; KICH: Kidney chromophobe; KIRC: Kidney renal clear cell carcinoma; KIRP: Kidney renal papillary cell carcinoma; LAML: Acute myeloid leukemia; LGGs: Low-grade gliomas; LIHC: Liver hepatocellular carcinoma; LUAD: Lung adenocarcinoma; LUSC: Lung squamous cell carcinoma; MESO: Mesothelioma; OS: Overall survival; OV: Ovarian serous cystadenocarcinoma; PAAD: Pancreatic adenocarcinoma; PCPG: Pheochromocytoma/paraganglioma; PFS: Progression free survival; PRAD: Prostate adenocarcinoma; READ: Rectum adenocarcinoma; RPPA: Reverse-phase protein array; SARC: Sarcoma; SKCM: Skin cutaneous melanoma; STAD: Stomach adenocarcinoma; TCGA: The Cancer Genome Atlas; TGCT: Testicular germ cell tumors; THCA: Thyroid cancer; THYM: Thymoma; UCEC: Uterine corpus endometrial carcinoma; UCS: Uterine carcinosarcomas; UVM: Uveal melanoma.

Next, we evaluated the relationship between CAG expression and immune cell infiltration. For example, ATP7A and ATP7B were negatively correlated with the macrophages of 16 tumors including breast invasive carcinoma (BRCA) and esophageal carcinoma (ESCA). Furthermore, CP, DBH, and MAP2K1 were positively correlated with natural regulatory T cell (nTreg) and T follicular helper cell (Tfh) in 28 tumors including CHOL [Supplementary Table 1]. We conducted a Cox proportional hazards regression analysis for each gene with the outcome as the dependent variable (Figure 3A; P ≥0.05 in gray; P <0.05, HR ≥1 in red; HR <1 in blue). The number of significant genes on Cox analysis of all tumors was counted, and the final risk score was determined by adding the risk factors (+1) and subtracting the protective factors (−1) [Figure 3B]. We found that the majority of CAGs had a score >0, indicating their significant role in cancer development. Cancers can be classified into subtypes based on the expression of specific genes that manifest differently in terms of tumor characteristics and clinical outcomes.^[18] Supplementary Figure 2A illustrates how the gene affects the subtype in KIRC, BRCA, stomach adenocarcinoma (STAD), LUAD, glioblastoma multiforme (GBM), lung squamous cell carcinoma (LUSC), HNSC, bladder urothelial carcinoma (BLCA), and colon adenocarcinoma (COAD).

Figure 3.

(A) Univariate Cox regression analysis of CAGs in pan-cancer. Gray denotes P ≥0.05, red represents P <0.05, HR >1 while blue denotes HR <1. (B) The respective gene scores with statistical significance according to Cox analysis were calculated in all tumors. We summarized the prognostic role (risk factors or protective factors) of genes in each tumour, with a risk factor as +1, and a protective factor as −1. ACC: Adenoid cystic carcinoma; BLCA: Bladder urothelial carcinoma; BRCA: Breast invasive carcinoma; CAGs: Cuproplasia-associated genes; CESC: Cervical squamous cell carcinoma and endocervical adenocarcinoma; CHOL: Cholangiocarcinoma; COAD: Colon adenocarcinoma; DLBC: Diffuse large B-cell lymphoma; ESCA: Esophageal carcinoma; GBM: Glioblastoma; HNSC: Head and neck squamous cell carcinoma; HR: Hazard ratio; KICH: Kidney chromophobe; KIRC: Kidney renal clear cell carcinoma; KIRP: Kidney renal papillary cell carcinoma; LGGs: Low-grade gliomas; LIHC: Liver hepatocellular carcinoma; LUAD: Lung adenocarcinoma; LUSC: Lung squamous cell carcinoma; MESO: Mesothelioma; OV: Ovarian serous cystadenocarcinoma; PAAD: Pancreatic adenocarcinoma; PCPG: Pheochromocytoma/paraganglioma; PRAD: Prostate adenocarcinoma; READ: Rectum adenocarcinoma; SARC: Sarcoma; SKCM: Skin cutaneous melanoma; STAD: Stomach adenocarcinoma; TGCT: Testicular germ cell tumors; THCA: Thyroid cancer; THYM: Thymoma; UCEC: Uterine corpus endometrial carcinoma; UCS: Uterine carcinosarcomas; UVM: Uveal melanoma.

Genetic correlation analysis and genetic variation

Supplementary Figure 2B demonstrates significant positive correlations among CAGs. Thus, each CAG can act either individually or cooperatively. Further, we analyzed the mutation frequency of different CAGs using pan-cancer analysis. The frequency of SNVs for genes ranged from 1% to 54% among different cancer types, with ATP7A having the highest percentage [Supplementary Figure 3A]. Missense_mutation was the most common variant classification, followed by nonsense_mutation. The most common variant type was single nucleotide polymorphism (SNPs), followed by deletion [Supplementary Figure 3B]. Furthermore, SNVs were dominated by C >T. We then created mutation landscape plots using the waterfall function in GenVisR51.^[19] The total SNV frequency of the CAGs was 77.7% (1146/1475 samples). The top 10 genes with the highest mutations were ATP7A, ATP7B, CP, TYR, ULK1, PDE3B, ULK2, LOXL2, DBH, and AOC3 [Supplementary Figure 3C]. On SNV survival analysis, a significant difference was noted between genes with and without mutations [Supplementary Figure 4A]. The association between immune cell infiltration and genomic SNV was assessed using a Wilcoxon test. The infiltrates of 24 immune cells were evaluated using ImmuCellAI. Supplementary Figure 4B shows that the immune infiltration fraction was higher in the GBM wild-type group and lower in the COAD, CESC, and BLCA wild-type groups. The complete data are presented in Supplementary Table 2.

Copy number variation

We examined the CNV of CAGs among all 33 cancer types obtained from TCGA pan-cancer samples, using GISTIC v.2.0,^[20] and found that heterozygous amplification or deletion was the main CNV in CAGs [Figure 4A]. CNV percentage analysis showed that the heterozygous amplification of LOXL2 in testicular germ cell tumors (TGCT), and that of ATP7B, UBE2D4, and AOC3 in KIRP, was >85% [Figure 4B]. CP in LUSC and ATOX1 in KIRC subtypes had >26% homozygous amplification levels, while LOXL2 in prostate adenocarcinoma (PRAD) had >26% homozygous deletion levels [Figure 4C]. Pearson's correlation analysis revealed a positive correlation between CNV and CAGs’ mRNA expression levels [Supplementary Figure 5]. These results suggest that the CNVs of genes mediate their aberrant expression, and are possibly a key factor in cancer progression.

Figure 4.

CNV underlies the dysregulation of CAGs. (A) CNV distribution in 33 cancers. CNV pie chart showing the combined heterozygous/homozygous CNV of each gene in each cancer. A pie chart representing the proportion of different types of CNV of one gene in one cancer, and different colors represent different types of CNV. Light red represents heterozygous amplification; deep red represents homozygous amplification; light green represents heterozygous deletion; deep green represents homozygous deletion; gray represents no CNV. (B-C) Heterozygous CNV profile showing the percentage of heterozygous CNVs, including the percentage of amplification and deletion of heterozygous CNVs for each gene in each cancer (B). Homozygous CNV profile showing the percentage of homozygous CNVs (C). Only genes with >5% CNV in a given cancer are shown as a point in the figure. ACC: Adenoid cystic carcinoma; BLCA: Bladder urothelial carcinoma; BRCA: Breast invasive carcinoma; CAGs: Cuproplasia-associated genes; CHOL: Cholangiocarcinoma; CNV: Copy number variation; COAD: Colon adenocarcinoma; DLBC: Diffuse large B-cell lymphoma; ESCA: Esophageal carcinoma; GBM: Glioblastoma; Hete Amp: Heterozygous amplification; Hete Del: Heterozygous deletion; Homo Amp: Homozygous amplification; Homo Del: Homozygous deletion; KICH: Kidney chromophobe; KIRC: Kidney renal clear cell carcinoma; KIRP: Kidney renal papillary cell carcinoma; LAML: Acute myeloid leukemia; LGGs: Low-grade gliomas; LICH: Liver hepatocellular carcinoma; LUAD: Lung adenocarcinoma; LUSC: Lung squamous cell carcinoma; MESO: Mesothelioma; PAAD: Pancreatic adenocarcinoma; PCPG: Pheochromocytoma/paraganglioma; PRAD: Prostate adenocarcinoma; READ: Rectum adenocarcinoma; RPPA: Reverse-phase protein array; SCNA: Somatic copy number alterations; SKCM: Skin cutaneous melanoma; THCA: Thyroid cancer; THYM: Thymoma; UCEC: Uterine corpus endometrial carcinoma; UCS: Uterine carcinosarcomas; UVM: Uveal melanoma.

Methylation analysis

Cancer development is thought to depend greatly on epigenetic alterations.^[21] Herein, we explored the methylation of CAGs to identify epigenetic regulation abnormalities. Most CAGs were hypomethylated in all 33 cancers, while AOC3 and SLC31A2 were hypermethylated in most cancer types [Figure 5A]. Moreover, the methylation of CAGs and mRNA expression levels exhibited an inverse correlation [Figure 5B]. Only SCO1 in liver hepatocellular carcinoma (LIHC), BRCA, BLCA, sarcoma (SARC); SLC25A3 in TGCT and thyroid carcinoma (THYM); and CCS in KIRP and adrenocortical carcinoma (ACC) showed a positive correlation between methylation and gene expression. Survival analysis indicated that the hypermethylation of COX17 and CCS in uveal melanoma (UVM) was associated with a poor prognosis [Figure 5C].

Figure 5.

(A) Differential methylation in CAGs between tumor (T) and normal (N) samples in each cancer. Red dots represent increased methylation in tumors and blue dots represent decreased methylation in tumors. The darker the dot color, the larger the difference in methylation level. (B) Correlation between methylation and mRNA gene expression. Red dots represent a positive correlation, and blue dots represent a negative correlation. (C) Survival differences between CAGs with high and low methylation levels in various cancers. Red dots represent worse survival of the hypermethylation group; blue dots represent the opposite. The dot size represents the statistical significance, and the larger the dot size means, the higher the statistical significance. ACC: Adenoid cystic carcinoma; BLCA: Bladder urothelial carcinoma; BRCA: Breast invasive carcinoma; CHOL: Cholangiocarcinoma; CAGs: Cuproplasia-associated genes; COAD: Colon adenocarcinoma; DLBC: Diffuse large B-cell lymphoma; ESCA: Esophageal carcinoma; GBM: Glioblastoma; KICH: Kidney chromophobe; KIRC: Kidney renal clear cell carcinoma; KIRP: Kidney renal papillary cell carcinoma; LAML: Acute myeloid leukemia; LGGs: Low-grade gliomas; LICH: Liver hepatocellular carcinoma; LUAD: Lung adenocarcinoma; LUSC: Lung squamous cell carcinoma; mRNA: Messenger RNA; MESO: Mesothelioma; PAAD: Pancreatic adenocarcinoma; PCPG: Pheochromocytoma/paraganglioma; PRAD: Prostate adenocarcinoma; READ: Rectum adenocarcinoma; RPPA: Reverse-phase protein array; SKCM: Skin cutaneous melanoma; THCA: Thyroid cancer; THYM: Thymoma; UCEC: Uterine corpus endometrial carcinoma; UCS: Uterine carcinosarcomas; UVM: Uveal melanoma.

Relevant miRNA regulation analysis

miRNAs play an important role in regulating gene expression.^[22] To determine whether miRNAs can regulate CAG expression, VisNetwork was used to create regulatory miRNA networks.^[23] As shown in Supplementary Figure 6A, miRNAs may regulate gene mRNA levels by targeting PDE3B, UBE2D3, COX11, PTGS1, and VEGFA. Further, VEGFA could be downregulated by more miRNAs, including hsa-miR-939-5p, hsa-miR-378a-5p, hsa-miR-299-3p, hsa-miR-372-3p, and hsa-miR-3143, among others. Supplementary Figure 6B shows the predicted results of UBE2D3 (using nine databases, we obtained eight common miRNAs). The full details of each dataset are represented in Supplementary Table 3. These results indicate that cuproptosis-related gene expression may be regulated by miRNA and may affect cancer progression.

Pathway activity analysis

The pathway activity shown in Supplementary Figure 7A suggests that CAG enrichment is involved in the TSC/mTOR, RTK, RAS/MAPK, PI3K/AKT, hormone ER, hormone AR, EMT, DNA damage response, cell cycle, and apoptosis pathways. For example, most CAGs were mainly involved in the activation of EMT and inhibition of cell cycle and DNA damage [Supplementary Figure 7B]. Gene set enrichment analysis (GSEA) analysis showed that CAG expression was closely correlated with the interferon_gamma_response, interferon_alpha_response, inflammatory_response, and tumor-necrosis factor-alpha (TNFA)_signaling_via_NFκB [Supplementary Figure 7C]. These results suggest that CAGs play a key role in regulating cancer-related metabolic pathways (such as fatty acid metabolism and bile acid metabolism).

Drug sensitivity analysis

On Pearson's correlation analysis for drug sensitivity, positive correlations indicate that high gene expression is associated with resistance to the drug, and vice versa.^[24] Our results showed that in the GDSC database, the expression of UBE2D2, UBE2D3, PDE3B, AOC3, CD274, and PDK1 was negatively correlated with the drug sensitivity including PHA-793887, BMS34554, methotrexate, TAK-715, AT-7519, and FK866, according to their IC₅₀ values [Figure 6A]. Meanwhile, in the CTRP database, UBE2D3, PDK1, PDE3B, COX11, and AOC3 were negatively correlated with drug sensitivity according to their IC50 values [Figure 6B]. Integration of the results from the two databases suggested that UBE2D3, PDK1, PDE3B, and AOC3 are potential therapeutic targets.

Figure 6.

Drug sensitivity analysis of CAGs. (A) The gene set drug sensitivity analysis from Genomics of Drug Sensitivity in Cancer (GDSC) IC₅₀ drug data. (B) The gene set drug sensitivity analysis from CTRP IC₅₀ drug data. Pearson's correlation indicates the correlation between gene expression and drug sensitivity. Blue bubbles represented negative correlations, and red bubbles represented positive correlations; the deeper the color, the higher the correlation. The bubble size was positively correlated with the FDR significance. The black outline indicates an FDR <0.05. CAGs: Cuproplasia-associated genes; CTRP: Cancer Therapeutics Response Portal; FDR: False discovery rate; IC₅₀: Half-maximal inhibitory concentration; mRNA: Messenger RNA.

Cuproplasia score

A cuproplasia score for each tumor included in the TCGA pan-cancer data was determined using the single-sample GSEA (ssGSEA) method, employing the R package “gene set variation analysis (GSVA).”^[25] We found that the cuproplasia score was the highest in both tumor and normal samples from PRAD [Supplementary Figure 8A]. On comparing the cuproplasia score of tumors with different clinical stages, we found a significant correlation in BRCA, KIRC, LUAD, skin cutaneous melanoma (SKCM), stomach adenocarcinoma (STAD), and THYM tumors [Supplementary Figure 8B]. On analyzing the correlation of the cuproplasia score with the prognostic data of patients, we found that high scores in KICH, brain lower-grade glioma (LGG), and KIRC were associated with poor prognoses, while high scores in SKCM were associated with a protective effect [Supplementary Figure 8C]. The association between immune cells’ infiltrates and cuproplasia score was represented by a correlation coefficient, which was evaluated through Spearman correlation analysis. A heatmap was drawn to illustrate the association between the relative proportion of tumor-infiltrating immune cells and the cuproplasia score [Figure 7]. The cuproplasia score was positively correlated with cytotoxic, exhausted, and macrophage types, and negatively correlated with CD8+ T cells, neutrophil, and CD4+ T cells.

Figure 7.

Association between the relative proportion of tumor-infiltrating immune cells and the cuproplasia score. ACC: Adenoid cystic carcinoma; BLCA: Bladder urothelial carcinoma; BRCA: Breast cancer; CD: Cluster of differentiation; CHOL: Cholangiocarcinoma; COAD: Colon adenocarcinoma; DC: Dendritic cell; DLBC: Diffuse large B-cell lymphoma; ESCA: Esophageal carcinoma; GBM: Glioblastoma; iTreg; Inducible regulatory T cells; KICH: Kidney chromophobe; KIRC: Kidney renal clear cell carcinoma; KIRP: Kidney renal papillary cell carcinoma; LAML: Acute myeloid leukemia; LGGs: Low-grade gliomas; LICH: Liver hepatocellular carcinoma; LUAD: Lung adenocarcinoma; LUSC: Lung squamous cell carcinoma; MAIT: Mucosal-associated invariant T cells; MESO: Mesothelioma; NK: Natural killer; NKT: Natural killer T cells; nTreg: Natural regulatory T cell; PAAD: Pancreatic adenocarcinoma; PCPG: Pheochromocytoma/paraganglioma; PRAD: Prostate adenocarcinoma; READ: Rectum adenocarcinoma; RPPA: Reverse-phase protein array; SKCM: Skin cutaneous melanoma; Tfh: T follicular helper cell; Th: T helper; THCA: Thyroid cancer; THYM: Thymoma; Tr1: Type 1 regulatory T cell; UCEC: Uterine corpus endometrial carcinoma; UCS: Uterine carcinosarcomas; UVM: Uveal melanoma.

Association of cuproplasia score with the therapeutic response in multiple cancer types

We first assessed the relevance of biomarkers related to the cuproplasia score, comparing them to standardized biomarkers for the prediction of OS in immune-checkpoint blockade (ICB) subgroups. Interestingly, we found 13 separate cuproplasia scores in the 23 ICB subgroups with an area under the receiver operating characteristic curve (AUC) >0.5 [Supplementary Figure 9A]. In turn, the cuproplasia score exhibited a higher predictive value. Furthermore, the high-risk group (those with cuproplasia scores above the median were assigned to the high-risk group) responded better to immunotherapy [Supplementary Figure 9B]. Patients with higher scores were also associated with a more progressive phase of cancer [Supplementary Figure 9C].

Discussion

Copper, as a vital constituent for enzymes and proteins,^[26] is the third most abundant essential trace metal in living organisms.^[27] Many biochemical processes require copper; these include mitochondrial respiration, skin pigmentation, iron transport, connective tissue formation, blood vessel formation, antioxidant defense, catecholamine synthesis, and peptide amidation.^[28] Copper can also indirectly reduce tumorigenesis by acting on tumor microenvironments (TMEs) to promote changes in myeloid precursor cell recruitment, or through indirect regulation of tumor-associated macrophages (TAM). Carcinogenesis is linked to disruptions of tumor suppressor genes, also known as antioncogenes.^[29] Cells undergo a neoplastic transformation when they acquire the ability to multiply uncontrollably, resist apoptosis, enable angiogenesis, and evade immune surveillance.^[30] Accordingly, cancer has been defined as an uncontrollable proliferation of cells.^[31] Increased copper levels have been detected in cancer cell lines, tissues, and patients’ sera.^[32,33] Therefore, there exists a close association between copper and tumorigenesis. Cuproplasia as a concept was recently proposed by Ge et al.^[4] It includes both the primary and secondary roles of copper through signaling pathways, as well as copper-dependent cell growth and proliferation.^[4,5] Cuproplasia is thus a promising target for cancer treatments, though its role in tumor pathogenesis needs to be studied further.

In the present study, using 11,057 cancer samples from 33 types and subtypes of cancer, we characterized the genomic, clinical, and functional characteristics associated with CAGs. CAG expression was found to be abnormal in various tumor types, which possibly reflects its unique functional aspects. We further investigated the relationship between CAGs and patient prognosis, and identified some CAGs as risk factors for patient prognosis in most tumors.

CNVs cause altered gene expression during tumorigenesis and tumor growth.^[34] In addition to the partial gene deletion, CNVs may also affect the functions of the affected genes, which in turn alters downstream gene expression.^[35] In general, CNVs can be divided into two subtypes—heterozygous and homozygous—which include amplifications and deletions.^[11] Our genetic analysis revealed a high frequency of CNVs in CAGs, and a strong positive correlation between them, indicating that CNVs influence CAG expression, probably contributing to tumor development. This is also consistent with previous findings.^[36]

Numerous studies have reported that hypermethylation is associated with poor prognosis in various types of cancers.^[36] The analysis of epigenetic modifications of individual CAGs revealed that abnormal hypermethylation of CAGs mediates their downregulation and is associated with a poor prognosis in many cancers. Hypermethylation and survival analysis indicated that hypermethylated COX17 and CCS may play a decisive role in UVM.

Considering the important role exerted by CAGs in tumorigenesis, it is crucial to identify their regulatory molecules. We analyzed the miRNA–mRNA interaction network and found that PDE3B, UBE2D3, COX11, PTGS1, and VEGFA could be regulated by miRNAs, which allowed us to identify therapeutic targets. We also found that cuproplasia is involved in tumor-related signaling pathways (RAS/MAPK, PI3K/AKT, hormone ER, hormone AR, EMT, and DNA damage response) in various tumors.

In the analysis of tumor pathogenesis, the TME should also be taken into consideration along with the tumor cells. TMEs alter angiogenesis, release cytokines,^[37] and contain fibroblasts and immune cells.^[38] Previous studies have found that copper can indirectly affect the TME with the help of TAM,^[31] which is in line with our conclusions.

Platinum-based drugs constitute one of the most widely used and effective chemotherapy drugs.^[39] Nevertheless, the exact mechanism underlying chemoresistance to these drugs remains unknown. Resistance to cisplatin was also influenced by copper transporter 1.^[31] Previous studies have suggested that ATOX1, CTR1, ATP7A, and ATP7B were involved in the transport of cisplatin.^[40] In this study, we found that cuproplasia scores were correlated with patient prognosis, clinical stage, and the therapeutic response in multiple cancer types, further emphasizing the important role of cuproplasia in malignancies. We also explored existing drugs that can target these CAGs: PHA–793887, BMS34554, methotrexate, TAK–715, AT–7519, and FK866 might be developed as potential therapeutic agents, and UBE2D3, PDK1, PDE3B, and AOC3 may serve as potential therapeutic targets in the future.

In this study, CAG expression was compared between tumor and normal tissues in the TCGA and GTEx databases. The genomic variation in SNVs and CNVs was found to affect mRNA levels and patient survival. Furthermore, we used several databases to predict the effect of CAGs on the efficacy of immunotherapy. We also found that CAGs are regulated by miRNAs, and identified CAGs for targeted drug therapy. Our study thus outlines potential targets for cuproplasia-related tumor immunotherapy.

Acknowledgments

We thank Figdraw (www.figdraw.com) for its help and support (ID:TITWS6cb57). We thank Bullet Edits Limited for the linguistic editing of the manuscript. We thank the Guo lab (Anyuan Guo, Department of Bioinformatics and Systems Biology, Huazhong University of Science and Technology) for offering their help.

Conflicts of interest

None.