1 Introduction
Ovarian cancer (OC) is one of the most common malignant gynecological tumors, and its mortality rate remains high [
1]. Because the disease is often asymptomatic and there is a lack of related biomarkers, a large proportion of patients are diagnosed at an advanced stage [
2]. Despite the application of standard therapy such as debulking surgery and adjuvant chemotherapy, more than 70% of patients exhibit tumor recurrence due to chemotherapy resistance [
3]. There has been tremendous research effort, and numerous treatment options, such as poly ADP-ribose polymerase (PARP) inhibitors and immunotherapy for ovarian cancer patients, have been tested [
4,
5]. However, the molecular heterogeneity at the genomic level, prominent toxicity, and possible acquisition of drug resistance substantially decrease efficacy in the clinical applications of adjuvant chemotherapy and targeted therapies [
6]. Therefore, in addition to the basic strategy of complete tumor resection, efficient therapeutic targets for sensitization and signatures of ovarian cancer related to the response to targeted therapy and immunotherapy should be explored to improve prognosis and survival.
Copper is an essential trace metal for living organisms (bacteria, animals, and humans) and plays an important role as a cofactor of copper-binding enzymes in many biological processes, including mitochondrial respiration, iron uptake, antioxidant activity, and detoxification [
7]. Copper maintains dynamic equilibrium at low concentrations via evolutionarily conserved homeostatic mechanisms and participates in cell proliferation (cuproplasia), while abnormal accumulation of copper can cause cytotoxicity and induce cell death (cuproptosis) [
8]. Copper ions induce abnormal oligomerization of lipoylated proteins by binding lipoylated components of the tricarboxylic acid (TCA) cycle, subsequently reducing iron-sulfur cluster protein levels and inducing a proteotoxic stress response, ultimately leading to cell death [
9]. Significantly elevated copper levels in serum and tumor tissue have been found in patients with a variety of cancers, including breast cancer, thyroid cancer, pancreatic cancer, and bladder cancer [
10‐
13]. An increasing number of studies have demonstrated that an abnormal increase in copper stress is involved in the proliferation, invasion, and metastasis of cancers [
14]. Léo Aubert found that the copper-exporter ATP7A is essential for neoplastic growth by regulating intracellular copper icon levels and further indicated that copper bioavailability is a KRAS-specific vulnerability in colorectal cancer[
15]. Additionally, copper ions contributed to tumor angiogenesis by activating many angiogenic factors, such as vascular endothelial growth factor (VEGF), VEGF2, and fibroblast growth factor 1 (FGF1) [
16‐
18]. However, some studies showed that the prognosis of OC patients was correlated to the procuproptosis gene-lipoic acid synthase (LIAS), and patients with OC who have high expression of LIAS showed a better overall survival and post-progressive survival [
19]. This demonstrated that promoting cuproptosis was beneficial to the prognosis of OC, but whether more CRGs had a positive effect on the prognosis of OC needed to be further explored.
In the present study, a scoring model based on cuproptosis-related differentially expressed genes was constructed, and this model may provide novel potential strategies to predict prognosis, immune infiltration, immunotherapy, and chemotherapy response. Furthermore, ligand–receptor pairs and RNA velocity in single-cell RNA-seq analysis were assessed based on this system, and these strategies supplement the current methods for evaluating novel connections between cancer cells and immunocytes. In general, the cuproptosis-related scoring model might assess aggressiveness and provide more efficient immunotherapeutic strategies for OC patients.
2 Materials and methods
2.1 Data preparation
The copy number variation (CNV), clinical information, and mRNA sequence data of ovarian cancer are downloaded from The Cancer Genome Atlas (TCGA) database (
https://portal.gdc.cancer.gov/). This samples from the Gene Expression Omnibus (GEO) (
https://www.ncbi.nlm.nih.gov/geo/) database are set as validation cohort, including clinical information and mRNA sequence. There are 583 OC patients in the training cohort from the TCGA dataset and 260 OC patients from the GEO dataset (GSE32062).
Data for single-cell RNA-seq (scRNA-seq) analysis is downloaded from the GEO database (GSE173682 and GSE158937) (
https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi). Expression data are normalized with R packages ‘Seurat’ and ‘NormalizeData’. The distribution of cells components is mapped with R package ‘UMAP’ [
20].
2.2 Unsupervised clustering for cuproptosis-related genes
Initially, the ten cuproptosis-related genes (CRGs) were identified from previous studies (PMID:35298263). Basis on the expression profiles of these 10 CRGs, 583 OC patients from the TCGA cohort were identified two clusters using the unsupervised clustering analysis, which used consensus clustering algorithm, and a follow-up analysis was conducted. TCGA-OV counts data are converted into transcripts per kilobase million (TPM) by using the R package “GeoTcgaData”.
2.3 Functional enrichment analysis and immune cell infiltration analysis
The Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway, and Gene set variation analysis (GSVA) were utilized for enrichment analysis for bulk RNA-seq and scRNA-seq, to investigate the differences in biological processes and pathways among the differences cuproptosis-related scoring (CuRS) clusters [
21]. The hallmark gene (c2.cp. Kegg.v7.2) was derived from the MSigDB database to run GSVA enrichment analysis. The single sample gene set enrichment analysis (ssGSEA) algorithm of the R package “GSEAbase” was performed to elucidate the enrichment of two different risk subgroups in 29 gene sets related to immune function [
22]. CIBERSORT, a versatile computational method for quantifying cell fractions from bulk tissue gene expression profiles (GEPs), and can accurately estimate the immune composition of a tumor biopsy [
23]. Subsequently, the Estimate algorithm is used to describe the infiltration rate of immune cells and stromal cells [
24].
p < 0.05 were considered to indicate significant differences.
2.4 Construction of a prognosis signature associated with cuproptosis
The deferentially expressed genes (DEGs) between the cuproptosis clusters are identified using the DESeq2 R package (
p < 0.05 and |log2FoldChange|> 0.58). Primarily, DEGs are subjected to univariate Cox regression analysis to identify these 33 prognostic-related DEGs. Afterward, utilizing the “glmnet” R package, the LASSO Cox regression algorithm is recommended to minimize the risk of over-fitting depended on CRG prognostic genes [
25]. In the TCGA training group, the multivariate Cox analysis identify 21 candidate genes and their correlative coefficients acquired for constructing the prognostic risk score. The following formula was used to determine the risk score:
$${\text{GuAscore = }}\sum\limits_{i = 1}^{n} {\exp ression\,of\,gene\,i\,*\,lasoo\,coefficient\,of\,gene\,i}$$
The patients in TCGA and GSE32062 OC cohort are further stratified as high-risk or low-risk subgroups, and then Kaplan–Meier survival analysis is performed, and the development of receiver operating characteristic (ROC) curves using the survival R package “timeROC” used to assess the predictive power of the signature [
26]. Furthermore, applying the univariate and multivariate analyses on the clinical variables in training and GSE32062 groups to assess whether CRRS could be served as an independent prognostic factor.
2.5 Evaluation of the sensitivity of chemotherapy and targeted therapy drugs
We downloaded from Genomics of Drug Sensitivity in Cancer (GDSC) the drug sensitivity of about 1000 cancer cells is the drug sensitivity of data (
http://www.cancerrxgene.org). The Spearman correlation analysis is used to calculate the correlation between drug sensitivity and CRRS, taking half-maximal inhibitory concentration (IC50) as the drug response index in cancer cell lines, and considering p < 0.05 and coef > 0.2 was a significant correlation. The same method is also done in Cancer Cell Line Encyclopedia (CCLE) database (
https://depmap.org/portal/download/).
2.6 Transcription factor regulatory network, RNA velocity, and cells communication
RcisTarget database of humans (
https://resources.aertslab.org/cistarget/) and R package “SCENIC” are used for transcription factors regulating network building [
27]. AUCell algorithm is utilized to assess transcription factor activation and regulatory modules based on the connection-specific index. Besides, RNA velocity analysis of tumor cells is calculated by package “monocle3” [
28]. Different states of OC cells are figured to reveal their internal transformation. Communication between immunocytes and OC cells is analyzed using the R package “Celltalker” and “Cellchat”, and differential ligand-receptor pairs are identified [
29].
2.7 Human OC tissue specimens and immunohistochemical staining
In our work, the clinical patient tissue microarray contained 8 40 epithelial ovarian cancer from department of Gynecologic Oncology, Hunan Cancer Hospital. All tissue specimens are confirmed by pathologist diagnosis and bedded in paraffin for immunohistochemistry (IHC). IHC staining and score criteria are described as previous research [
30]. The primary antibody used was anti-KIF26B (dilution 1:200, 17422-1-AP, Proteintech). The informed consent is given to patients before this research.
2.8 Quantitative real-time PCR and small interfering RNA
Total mRNA was extracted from OC cells using Trizol reagent (Takara) following the operating protocol. Reference gene GAPDH is used to normalization. Primer sets used for KIF26B and GAPDH RNA examination are as follows: KIF26B forward 5ʹ- TTTGCGCCACTCACCTGAA -3ʹ, KIF26B reverse 5ʹ- GGCGTCGTAGTGCTCACTG -3ʹ; 18 s forward 5ʹ-TGCGAGTACTCAACACCAACA-3ʹ,18 s reverse 5ʹ- GCATATCTTCGGCCCACA-3ʹ. The formula RQ = 2 − ΔCT is used to calculate gene expression levels.
The siRNAs against KIF26B were purchased from Gene Pharma (Shanghai, China). Transfection according to the manufacturer’s protocols uses Lipofectamine 3000. For KIF26B siRNA: siKIF26B-1: 5- GGACAACCGCUGUGACAUUTT-3, siKIF26B-2: 5-CAUCGAGAGUCUUGAGGAUTT-3.
2.9 Cell viability, colony formation, and Edu stain assay
The CCK-8 cell proliferation, colony formation, and apoptosis assay were performed as previously described [
30].
2.10 Statistical analysis
Statistical analyses are executed utilizing R software (version 4.1.0) and RStudio (version 2022.07.2). The SPSS 19.0 and GraphPad Prism 8.0 software is used for statistical analysis. Student’s t-test is used to analyze two groups of data. All statistical p values are two-sided and less than 0.05 is considered statistically significant. All methods were carried out in accordance with relevant guidelines and regulations.
4 Discussion
Copper is a valuable coenzyme in many physiological activities, such as the oxidative stress response, lipid metabolism, and mitochondrial respiration [
33]. Cuproptosis, an unconventional mechanism of cell death, is playing an increasingly important role in promoting tumor progression, recurrence, TME, and chemotherapy response [
34,
35]. The mechanisms of copper-induced tumor cell death include induction of reactive oxygen species production, decreasing the expression of signaling pathways, and overcoming chemoresistance, but its effects on OC remain poorly understood. In ovarian cancer, previous studies showed that CDKN2A, the mutations of which led to loss of growth control in ovarian cancer cells, is related to the cell cuproptosis sensitivity [
9,
36], indicating that cuproptosis plays essential role in ovarian cancer.
Infiltration of immune cells was validated to be associated with cancer cell proliferation, invasion, and metastasis in multiple cancers [
37‐
40]. Some immune cells were differentially infiltrated in the CuRS subgroups. While the high-risk group had more M0 macrophages, resting memory CD4 T cells, activated NK cells, and activated mast cells, the low-risk group had more CD8 T cells and M1 macrophages. The ssGSEA algorithm results also revealed that patients in the low-risk group had higher immune activity. Tumor-associated macrophages (TAMs) are the key regulatory agent of treatment responses in the TME [
41]. M2 macrophages, as a major subtype of macrophages, are conducive to tumor cell growth, invasion, and angiogenesis, and these cells are associated with a poor prognosis in many cancers [
42]. In contrast, a high level of M1 macrophages is related to antigen presentation and indicates antitumor activity in patients with ovarian cancer [
43]. Many studies have shown that a high level of T-cell infiltration, particularly cytotoxic CD8 T cell infiltration, indicates a favorable prognosis, and our findings support these conclusions [
44]. According to the ESTIMATE algorithm, a low-risk group had higher immune scores than a high-risk group. As previously reported, the higher TMB is associated with better prognosis in OC patients [
45], which is consistent with the findings in this study. These results suggested that our model can accurately predict the prognosis and immune status of OC patients and develop personalized immunotherapeutic strategies.
As previously shown, no studies have revealed the cellular relationship between cuproptosis and the progression of OC or the immune landscape. By integrating bulk and single-cell RNA sequencing, we first systematically analyzed the heterogeneity of OC. Based on the cuproptosis genes, we classified 15 clusters and 8 subtypes of OC cells using the Seurat package and copy number variation (CNV) profile, and these clusters were visualized by using the UMAP algorithm. Pseudotime trajectory analysis via the Monocle 3 algorithm also illustrated the time-dependent evolution of OC cells among these clusters. In our study, we show that the differentiation and progression of OC cells is classified along with increasing CuRS, which demonstrates that the CuRS scoring model can be used to evaluate tumorigenesis.
The development of chemotherapy drug resistance is an important cause of recurrence and death in patients, but there is no effective treatment [
46]. Copper transporters mediate resistance of OC cells to platinum-based drugs, and may predict platinum sensitivity and prognosis of OV patients. Prevent study showed that cisplatin chemotherapy resistance by glutathione-resistant copper-based nanomedicine via cuproptosis [
47]. Based on the chemotherapy sensitivity analysis, it is possible that patients with high-CuRS are more responsive to platinum chemotherapy drugs. Consistently, increased drug responsiveness to olaparib and niraparib was identified in patients with high cuproptosis scores, suggesting that PARP inhibitors have potential value and providing a useful reference for combining PARP inhibitors and with cuproptosis-targeting agents to alleviate platinum resistance. Furthermore, our data show that OC patients with high cuproptosis scores had higher sensitivity to other targeted chemotherapy drugs, including VEGFR inhibitors, mTOR inhibitors, and PI3K signal inhibitors, than those with low cuproptosis scores. These results suggest that cuproptosis-related genes may predict or affect the response to chemotherapy in patients with OC.
In this study, we built a prognostic scoring model relying on cuproptosis-related DEGs in ovarian cancer samples. High CuAS samples show a more aggressive growth pattern and worse clinical outcomes than low CuRS samples. Tumor cells with different CuRS values communicate with immunocytes in a distinct manner, implying that cuproptosis activation may modulate immunocyte function. We assumed that targeting high CuRS samples may improve the patient’s prognosis, and novel potential compounds were predicted via a machine learning algorithm. The specific mechanisms underlying the function of these model genes and the correlation of prognostic and therapeutic gene signatures in ovarian cancer need further clarification. Nevertheless, our model can accurately predict the prognosis and immune status of OC patients and guide precision treatment. In summary, comprehensive bioinformatics analysis demonstrated that the CuRS model can be used to evaluate ovarian cancer aggressiveness and modulate crosstalk with immunocytes, offering a new option for chemotherapy combination treatments and response evaluation.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.