Background
Hepatocellular carcinoma (HCC), a common abdominal tumor typically originating from the cirrhotic liver, is the fourth most common cause of cancer-related death worldwide, leading to over 400,000 deaths throughout China in 2021 [
1,
2]. Although unremitting efforts have been devoted to the diagnosis and treatment of HCC, patients still suffer from poor prognoses. In China, more than half of the patients are diagnosed with advanced disease at their first visit, and their 5-year overall survival rate (OSR) is < 12.5% [
3]. Surgical excision offers the only possibility for a cure [
4]. However, only 34 to 70% of patients may be suitable for hepatic resection; this limitations cause the overall postoperative mortality to reach up to 3% [
5]. Moreover, molecular target therapy (MTT) and immune checkpoint inhibitors (ICIs) exhibit limiting improvements in overall survival (OS). For instance, the median OS of sorafenib an approved first-line agent for advanced HCC, is merely 14.7 months [
6]. Additionally, only a minority of patients receiving ICIs achieve the treatment response. Nivolumab and pembrolizumab commonly produce a 15–20% rate of objective remissions [
7]. These observations underscore the urgency and importance of widening therapeutic strategies and refining clinical assessments. Recently, cuproptosis, a novel form of programmed cell death (PCD), has been a topic of interest in HCC treatment.
Programmed cell death exceedingly expands the anti-cancer arsenal. With the discovery of ferroptosis, necroptosis and pyroptosis, we have obtained a deeper understanding of carcinogenic mechanism and clinical assessment of multiple cancers [
8‐
10]. For example, SLC7A11 the catalytic subunit of Xc- system in ferroptosis could promote malignant biological properties of renal carcinoma cells [
9]. Ferroptosis regulator SLC1A5 exhibits its cancer-promoting abilities by activating the mTORC1 signaling pathway [
10]. In 2022, Tsvetkov, P et al. has reported noteworthy research on copper-mediated cell death, namely ‘Cuproptosis’ [
11]. Mechanistically, metal reductase FDX1 is activated owing to the accumulation of intracellular copper ions. Subsequently, FDX1 mediates the lipoylation of the tricarboxylic acid cycle (TCA) proteins, thereby inducing the oligomerization of lipoylated proteins with the aid of copper ions. Considering that the immense potentials of cuproptosis in cancer treatment, the use of Cu ionophores has been proposed to be an emerging technological approach for targeting cancer cells [
12].
The discovery of cuproptosis has attracted considerable interest across the oncology community. Several scholars have commented on this remarkble finding and regarded it as a new bellwether for cancer treatment [
13‐
15]. Nevertheless, limited studies have probed into the roles of cuproptosis regulatory genes in cancers, which is the original aim of this research. In the present study, we sought to construct a novel risk signature based on 17 core CR genes using Lasso regression analysis. Moreover, we intended to investigate its great prognostic value and the abilities for indicating the immune microenvironment, metabolic reprogramming and therapeutic outcomes. Our findings provided novel and valuable evidence of the therapeutic potential of utilizing cuproptosis for treating HCC.
Materials and methods
Data source
We obtained the gene expression data and clinical information from TCGA, ICGC and GEO public databases. Owing to the inadequate number of normal samples in TCGA-LIHC cohort (
n = 50), we added 110 normal liver tissue samples from GTEx database (
https://xenabrowser.net/datapages/) to equilibrize the sample sizes of tumor and normal tissues. All transcriptome data was standardized by log2 (FPKM + 1) transformation. The clinical characteristics of TCGA, ICGC and GEO cohorts were presented in Supplementary Table
1–
2.
Reportedly, protein lipoylation in TCA cycle triggers the onset of cuproptosis via FDX1 mediation. Accordingly, based on the findings of the study by Tsvetkov, P et al. [
11], we selected 17 critical cuproptosis regulatory genes for further analysis. The CR genes and their functions in the cuproptosis process were shown in Table
1. We constructed the protein–protein interaction (PPI) network of CR genes using the STRING database (
https://string-db.org/) [
16] and Cytoscape (version 3.71) software [
17]. The biological function analyses of 17 cuproptosis regulators were performed via the DAVID database (
https://david.ncifcrf.gov/) [
18].
Table 1
Seventeen critical genes involved in cuproptosis process
FDX1 | Ferredoxin 1 |
LIPT1 | Lipoyltransferase 1 |
LIAS | Lipoic acid synthetase |
DLD | Dihydrolipoamide dehydrogenase |
PDP1 | Pyruvate dehydrogenase phosphatase catalytic subunit 1 |
DLAT | Dihydrolipoamide S-acetyltransferase |
PDHA1 | Pyruvate dehydrogenase E1 subunit alpha 1 |
PDHB | Pyruvate dehydrogenase E1 subunit beta |
DBT | Dihydrolipoamide branched chain transacylase E2 |
GCSH | Glycine cleavage system protein H |
DLST | Dihydrolipoamide S-succinyltransferase |
SLC31A1 | Solute carrier family 31 member 1 |
ATP7A | ATPase copper transporting alpha |
ATP7B | ATPase copper transporting beta |
MTF1 | Metal regulatory transcription factor 1 |
CDKN2A | cyclin dependent kinase inhibitor 2A |
GLS | Glutaminase |
Consensus clustering analysis
We applied consensus clustering for identifying the distinct prognostic patterns based on the features of CR expressions. This procedure was performed using the ‘ConsensusClusterPlus’ package in R software (version 4.1.2) and was based on the algorithm of cumulative distribution function (CDF).
Establishment of CR risk signature
WE constructed a novel CR risk signature through two steps. First, CR differentially expressed genes (DEGs) were screened out using the ‘Limma’ package in R software (version 4.2.0). The screening criteria were as follows: adjusted p-value < 0.05 and absolute value of log2FC ≥ 0.58 (1.5 fold difference in gene expression). Second, we used CR DEGs to accomplish the modeling process through the Lasso regression analysis using the ‘glmnet’ R package. This process was performed using the sevenfold cross-validation scheme.
Prognostic analysis
The optimal cutoff value of the CR risk score was determined using the Cutoff Finder online tool (
http://molpath.charite.de/cutoff) [
19]. The prognostic differences between high- and low-risk groups were compared based on the Kaplan–Meier method. Independent prognostic factors of HCC were identified using the Cox univariate and multivariate analyses. The accuracy of predicting OSR was assessed using the receiver operating characteristic curve (ROC). Decision curve analysis (DCA) was utilized for determining whether CR risk score could elevate clinical-decision benefit of traditional prognostic models. Furthermore, we conducted clinical subgroup analyses to evaluate the prognostic stratification ability of the CR model in HCC patients with different disease stages. Owing to the low number of samples in M1 (
n = 3) and N1 stages (
n = 3), clinical subgroup analyses were not applied to these subgroups. Using multiple logistic regression, a nomogram comprising of clinical stage and CR risk level were constructed to predict the OSR of individual at 1,3, and 5 years. The calibration curve was used to test its predictive accuracy.
We selected the GSE14520, GSE116174 and ICGC-LIRI cohorts to validate the prognostic value of CR risk signature. Survival difference analysis and ROC were both conducted in each validation cohorts.
Immune and mutational analysis
The CIBERSORT algorithm was performed to quantize the infiltration levels of 21 immune cells in each HCC sample [
20]. As described in previous studies [
21,
22], ssGSEA (single-sample gene set enrichment analysis) method was employed to calculate the activities of ten immune-related signaling pathways. The ESTIMATE method is an effective approach for inferring the fraction of stromal and immune cells in tumour samples using gene expression [
23]. By this method, immune score, stromal score and tumor purity of each HCC sample can be calculated. The cBioPortal database (
http://cbioportal.org) [
24] provided the somatic mutational frequency and patterns of CR signature genes across four HCC projects (
n = 973 samples).
GSEA
GSEA (Gene Set Enrichment Analysis) was used to investigate the influence of CR risk score on multiple metabolic processes, including glycolysis, nucleotide, cholesterol, glutamine, and fatty acid metabolisms. The MSigDB database (
https://www.gsea-msigdb.org/) provided the used gene sets. The detailed descriptions of metabolic gene sets were presented in Supplementary Table
3. The phenotype labels were set as high-CR risk versus low-CR risk samples. The number of permutations was set at 1000, and there was no collapse in gene symbols.
Therapeutic correlation analysis
WE explored the potential associations of CR risk score with the efficacy of sorafenib and ICIs. The GSE109211 dataset, namely the phase 3 STORM trial, contained the transcriptome data and therapeutic outcomes of 140 HCC patients receiving sorafenib treatment [
25]. Thus it was applied to the sorafenib-related analysis. Regarding ICIs, we addressed the issue from four perspectives, namely tumor mutation burden (TMB), TIDE (Tumor immune dysfunction and exclusion) algorithm, the expressions of immune checkpoints (ICs), and the IMvigor 210 cohort. Among these, TIDE algorithm is pivotal for predicting the response to anti-PD-1/L1 and anti-CTLA4 treatments based on the estimation of T cell dysfunction and tumor immune evasion, which was achieved by its online tool (
http://tide.dfci.harvard.edu/login/) [
26]. The IMvigor210 dataset was derived from a real clinical cohort and offered a therapeutic response to atezolizumab (a PD-L1 inhibitor) of 348 patients [
27].
Clinical samples and qPCR
After obtaining informed consent from the patients, 20 pairs of HCC and adjacent normal liver tissues were utilized for confirming the differential expression of DLAT. The study protocol was approved by the Ethics Committees of second affiliated hospital of Xi'an Jiaotong University.
Total RNA was extracted using TRIzol Reagent (TakaRa, Japan). RNA concentration was calculated by the A260/A280 ratio with the aid of Nanodrop 2000 spectrophotometer. Reverse transcription reactions were performed via the PrimeScript RT reagent Kit (Takara, Japan). RT-qPCR reaction was marked by SYBR-Green PCR Reagent (Takara, Japan) and tracked on the ABI Prism 7900 sequence detection system. GAPDH was employed as the reference gene. The relative gene expression was calculated according to the 2-ΔΔCT method. The list of primer sequences was shown in Supplementary table
4.
Immunohistochemistry assay
The Formalin-fixed HCC and the paired adjacent normal tissues were embedded in paraffin and cut into 3 mm sections. The clinical specimens were incubated with rabbit polyclonal antibodies of DLAT (1ug/ml, Abcam, USA) at 4° overnight. Secondary antibodies labeled with horseradish peroxidase (1:400, Abcam, USA) were incubated with the sections at room temperature for 1.5 h. Then, each section was stained with DAB reagent, and finally counterstained with hematoxylin.
Cell culture and transfection
Two hepatocellular cancer cell lines (HepG2 and Huh-7) that were obtained from Procell Life Science&Technology Company (Wuhan, China) were used for further in vitro experiments. HepG2 and Huh-7 cells were cultured in MEM (Minimum Essential Medium) and DMEM (Dulbecco's Modified Eagle Medium) respectively. Each medium was added by 10% FBS (Fetal bovine serum) and 1% P/S (Penicillin/ Streptomycin) (Procell, Wuhan, China). HanHeng Biotechnology (Shanghai, China) designed and synthesized the short hairpin RNA targeting DLAT (sh-DLAT) and the overexpression plasmids (OE-DLAT). Their specific sequences were shown in Supplementary table
5. The lentiviral system created stable transfected cells (HanHeng Biotechnology, Shanghai, China).
Transfected cells at logarithmic growth phase were seeded into 6-well plates with a density of 1 × 103/ per well. After the incubation of 2 weeks, cell colonies were visible and were fixed by methanol. Giemsa was applied to stain the cell colonies. Finally, colonies were counted under the microscope from five random fields.
Transwell assays
Transwell assays followed the similar procedures described previously [
28]. In migrative assays, the mediums with different concentrations of FBS were added into upper (0.1%) and lower (10%) chambers respectively. Cells were cultured for 24 h, and we used PBS and swab to remove non-migrative cells. Then, the migrated cells were fixed by paraformaldehyde for 20 min and stained by 0.1% crystal violet for 20 min. Cell counting was conducted using a high magnification microscope (100-fold) from five random visual fields. For the invasive assays, the upper chambers were precoated with Matrigel (Corning, NY, USA).
Statistical analysis
All statistical analyses were performed using the R software (version 4.2.0) and GraphPad Prism (version 8.0). Differences between groups were compared by unpaired T test or Wilcoxon rank sum test. Correlations between CR risk score and the clinicopathological features of HCC were determined via the Kolmogorov–Smirnov test. The in vitro experiments were performed in triplicated. Statistical significance was set at P < 0.05.
Discussion
Owing to high malignancy and easy metastases, HCC results in a poor prognosis, with a median survival time of 23 months [
45]. Hepatectomy, MTT or ICIs do not fulfill the eager needs of patients for treating liver carcinoma. Recently, the discovery of cuproptosis paints a promising anti-cancer landscape, which may bring a paradigm shift in cancer treatment. Limited available research has reported the roles of cuproptosis regulators in prognosis, immune response and development of cancers, this lack of information prompted us to conduct this investigation.
It is worthy to notice the fact that detective approach of cuproptosis remains obscure, meanwhile no available studies confirmed the existence of cuproptosis in human cancers so far. The most critical issue, whether cuproptosis occurs in HCC needs to be addressed first. We speculated the answer was negative and the following possible reasons resulted in this. First, the accumulation of copper ions couldn’t always trigger cell cuproptosis, but where the copper ions concentrate is the decisive factor [
37]. Copper (Cu) is an essential nutrient for a huge number of biological processes including energy metabolism, iron uptake and antioxidant/detoxification processes [
46]. Therefore, Cu accumulation has been commonly associated with enhanced proliferation and growth, angiogenesis, and metastasis [
46]. Mounting research has determined the upregulation of Cu levels in both serum and tumor tissues in various human cancers such as prostate cancer [
47], lung cancer [
48] and colorectal cancer [
49]. Recently, Tamai Y et al
. have confirmed that Cu levels was positively correlated with higher BCLC (Barcelona clinic liver cancer) stage in HCC [
50]. In light of these findings, Cu levels should elevate in HCC. However, the surge of Cu ions does not directly drive cuproptosis occurrence in HCC. The core reason is the aggregated location of Cu ions. The most critical evidence is the anticancer mechanism of Elesclomol (ES), the only cuproptosis inducer available. Unlike other copper ionophores, ES could selectively promote cellular copper levels in mitochondria, not just in the cytosol [
51]. In conclusion, cuproptosis did not occur spontaneously in liver cancer, but active Cu metabolism and high Cu levels in HCC created its precondition. Indeed, two current copper-related anticancer strategies support the above discussion [
12]. On one hand, researchers have applied Cu chelators to inhibit Cu-dependent cellular proliferation, termed ‘cuproplasia’, through decreasing the intracellular Cu concentration [
52]. On the other hand, Cu ionophores being developed exhibit a promising anticancer direction through stimulating Cu concentration in mitochondria, namely inducing cuproptosis [
12].
Accurate prognostic assessment is the most critical component of individualized cancer treatment. Although some mainstream prognostic systems strongly contribute to predicting survival outcomes of HCC patients, these systems are not without their limitations. For example, the Barcelona Clinic Liver Cancer (BCLC) system is insufficient in providing precise distinguishability for prognostically stratifying HCC patients with the intermediate stage [
53]. Moreover, AJCC 8th edition staging system fails to discriminate survival differences between patients with IVA and IIIA stages [
54]. Giannis D et al. have reported that AJCC 8th edition presented a mediocre predictive ability in a SEER cohort, particularly the advanced TNM stage was not associated with increased risk of death [
55]. Thus, improving the existing models is necessary and meaningful. In the present study, the novel CR risk signature could markedly elevate the decision-benefit and predictive accuracy of the AJCC system (Fig.
3F-H), demonstrating that it acted as an essential supplement to the AJCC system. Moreover, CR risk signature was capable of distinguishing the survival difference of III/IV stage cases, which were the inadequacies of AJCC system [
54]. Thus, our findings validated the remarkable prognostic value of the CR model.
The alterations in tumor immune microenvironment (TIM) profoundly determine the trend of anti-tumor response. Immune analyses revealed that CR risk score was closely associated with the infiltration level of CD8 + T cells and macrophages. The potent anti-cancer potency of CD8 + T cells has long been known, this immune guarder eradicates tumor cells through perforin and Fas/Fasl pathways [
56]. Under different chemokine stimulation, macrophages can differentiate into M1 and M2 subtypes. Macrophages polarization is strongly involved in the cancer immune regulation [
57]. Specifically, M1 subtypes can directly target cancer cells, whereas M2 subtypes can drive immune evasion and tolerance by suppressing the functions of CD8 + T cells [
58]. Therefore, decreased immune abundance of CD8 + T cells and M1 macrophages (Fig.
6A), and increased that of M2 macrophages resulting from high CR risk all pointed toward unfavorable changes to the anti-cancer immune process. CR risk score may be indicative of anti-tumor response.
Immune checkpoint inhibitors (ICIs) well represented by pembrolizumab (PD-1 inhibitor) have changed the paradigm of cancer treatment. Currently, the NCCN (version 2021) guidelines have listed nivolumab, pembrolizumab and atezolizumab as the first-line option for HCC treatment [
59]. Nevertheless, it is inconclusive of identifying a reliable and effective biomarker for predicting the efficacy of ICIs. Here, we found that a high CR risk score may be an indicator of the response to immunotherapy (Fig.
7H-L). Despite the negative result observing in TMB, given that inadequate stimulation for neoantigens formation, high cost of determination and false-negative response population [
60,
61], whether TMB is a valid predictor is controversial [
62]. In contrast, high expression levels of ICs [
63,
64], low TIDE score [
26], and analytical results of the IMvigor 210 cohort supported the associations between CR risk score and ICIs efficacy.
Evidence suggests that metabolic reprogramming is a critical hallmark of cancer biology. Particularly, aerobic glycolysis termed the ‘Warburg effect’ widely participates in malignant progression, therapy resistance and immune tolerance of various cancers [
65,
66]. Owing to meeting the metabolic requirements of cell proliferation [
67,
68], active glycolysis commonly implies cancer development. In this study, we observed that glycolysis was enriched in HCC samples with high CR risk scores (Fig.
7A-D), thereby indicating that glycolysis may be the metabolic driving force of high-risk progression.
Some studies have investigated the functions of CR signature genes in multiple cancers. For instance, FDX1 can promote ATP production and is a risk indicator for LUAD prognosis, but cannot affect the proliferation and apoptosis of LUAD cells [
69]. CDKN2A promoter methylation was associated with an elevated HCC risk and indicated HCC progression [
70]. GLS as a crucial substrate of MET kinase can promote the metabolism and biogenesis of HCC cells [
71]. Nonetheless, only a few studies have reported the roles of DLAT in cancers, which prompted us to conduct further analysis. Through in vitro experimentations, DLAT was established to have pro-oncogenic capacities in HCC, thereby indicating its potentials as an anti-cancer agent. This gene encodes component E2 of the multi-enzyme pyruvate dehydrogenase complex (PDC) and its overexpression leads to cirrhosis and liver failure [
72]. Hence, targeting DLAT can also aid in treating other liver diseases.
Nevertheless, there are some limitations that cannot be neglected in this study. First, the CR risk signature requires further validation in a clinical cohort. Second, the specific cancer-promoting mechanism of DLAT in HCC remains elusive. Third, we did not detect the intensity of intracellular cuproptosis at different expression levels of DLAT. Fourth, since cuproptosis research is still in its infancy, lacking the detective means of cuproptosis is a currently unavoidable drawback, which extremely limits the clinical implications of our study. Thus, utilizing cuproptosis for the clinical assessment and treatment of HCC is a long but promising way.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.