Background
Head and neck cancer (HNC) ranked sixth among all cancer types worldwide, which includes cancers originating in the lip, oral cavity, nasopharynx, larynx, hypopharynx, etc. [
1]. According to the data based on the GLOBOCAN produced by the International Agency for Research on Cancer in 2020, the global incidence counts of HNC were approximately 870,000 and the global mortality counts of HNC were approximately 440,000 [
2], which indicates that more attention should be paid on the significant burden caused by HNC.
Head and neck squamous cell cancer (HNSCC), as the most common pathological type of HNC, accounts for approximately 90% of HNC cases. During the past decades, although significant improvement has been observed in the treatment strategies including surgery, radiotherapy, chemotherapy, and immunotherapy for HNSCC, the overall survival and quality of life of HNSCC has not increased accordingly [
3]. Therefore, it is urgently needed to further explore the development and progression mechanisms, screening effective biomarkers for predicting early diagnosis, and long-term prognosis of HNSCC.
As an important cofactor, copper plays a vital role in various physiological processes [
4]. Dysregulated intracellular bioavailability of copper can lead to oxidative stress and cytotoxicity [
5]. However, many associations between disease status and Cu have been observed. In particular, copper levels in the serum and tumour tissues of patients with various cancers were significantly changed [
6‐
11]. In a recent study published in the journal Science, a novel form of cell death, copper-dependent cell death (referred to as cuproptosis), showed that copper binds directly to lipoylated components of the tricarboxylic acid (TCA) cycle, followed by lipoylated mitochondrial protein aggregation and subsequent loss of the Fe–S cluster, triggering proteotoxic stress and a unique form of cell death [
12]. Therefore, the potential role and mechanism of cuproptosis in the development and progression of HNSCC needs to be further explored.
To the best of our knowledge, there were only 19 coding genes verified to be related to cuproptosis, which was not enough to establish a prognostic model for HNSCC. Previous studies suggested that lncRNAs have a potential role in the diagnosis and therapy of HNSCC [
13]. At present, there are few studies focusing on CRLs in HNSCC, which are used to predict the prognosis and chemotherapy sensitivity of HNSCC. Therefore, we aimed to explore cuproptosis-associated lncRNAs, which might provide new ideas and insights for HNSCC treatment and prediction.
Based on seven CRLs, our study revealed a predictive signature model and evaluated the prognosis, drug sensitivity, and tumour immune functions of patients with HNSCC. Ultimately, we also tentatively verified the expression levels of seven CRLs among HNSCC cell lines and human normal nasopharyngeal cell lines.
Materials and methods
Data download and screening of cuproptosis-related lncRNAs
The transcriptome profiles and relevant clinical information of patients with HNSCC were downloaded from the TCGA database. Gene expression was normalized by the “limma” R package. A total of 19 CRGs were acquired from published studies (Additional file
1: Table S1).
The association between the CRLs and HNSCC was estimated by Pearson correlation. The correlation coefficient |R2|> 0.4 at P < 0.001 was deemed significant. A Sankey diagram was drawn to present the extent of association between CRLs and CRGs via the ‘ggalluvial’ R package.
Establishment of the cuproptosis-related lncRNA prognostic model
Cox regression was performed by the “survival” and “glmnet” packages. Patients with HNSCC were randomly separated into training or test groups. We used the 783 CRLs identified to establish the HNSCC prognostic model. Then, prognostic candidates were identified by LASSO Cox regression analysis. Ultimately, we generated a prognostic model for seven lncRNAs associated with cuproptosis, selecting the best penalty parameter λ correlated with at least tenfold cross-validation. According to the median risk score, the samples were classified into a high-risk group and low-risk group. The K–M curve was generated by the ‘survminer’ R package. A receiver operating characteristic curve (ROC) was generated by the ‘timeROC’ R package.
Construction and validation of a predictive nomogram
The nomogram of HNSCC patients was plotted by the “rms” package. The calibration curve was performed to evaluate the consistency between the OS and PFS rates in this study.
Principal component analysis (PCA)
PCA is a helpful tool that is widely applied in reducing dimensionality and extracting features in the computer vision field [
14]. We also estimated differences between the above two risk groups by the “scatterplot3d” package.
Functional enrichment analysis
The genes differentially expressed between the high-risk and low-risk groups were identified (|log
2(fold change)|> 1 and FDR < 0.05) by the ‘edgeR’ R package [
15]. GO and KEGG analyses were conducted by the ‘clusterProfiler’ R package [
16].
Tumour mutation burden (TMB) analysis
TMB refers to the number of mutations per million bases in the tumour genome [
17]. Tumours with a higher TMB were usually more sensitive to immunotherapy. The TMB was analysed by the “maftools” R package.
Immune functions and tumour immune dysfunction and exclusion (TIDE) score
Based on the CRL signature, single-sample gene set enrichment analysis (ssGSEA) and the ‘‘gsva’’ package were compared to determine the immune functions between the high-risk and low-risk groups. The differences in immune functions were uncovered using a heatmap. The TIDE score was used for the prediction of outcome and response to immunotherapy for cancer patients. The TIDE score, CAF, IFN-g (IFNG), CD8 score, CD274 score, MDSC, dysfunction score, merck18 (T-cell-inflamed signature) score, TAM M2, and exclusion score were obtained from the TIDE web (
http://tide.dfci.harvard.edu).
Drug susceptibility analysis
The half-maximal inhibitory concentration (IC50) of common chemotherapeutic drugs was generated by the “pRRophetic” R package in different risk groups. Cisplatin, paclitaxel, docetaxel, doxorubicin, etoposide, gemcitabine, methotrexate, and cytarabine were included in this research.
Cell culture and qRT-PCR
The human normal nasopharyngeal cell line (NP69), the human tongue squamous cell carcinoma cells (SCC25) and the human NPC cell line (HNE-2) were donated by the Cancer Research Institute of Central South University (China). Human hypopharyngeal squamous cell carcinoma cells (FaDu) were purchased from Suzhou Bei Na Chuanglian Biotechnology Co., Ltd. (China). The cell culture conditions and qRT-PCR were as described earlier [
18]. The list of primer sequences is described in Additional file
2: Table S2.
Statistical analysis
Statistical analysis was visualized using R software (version 4.2.0,
https://www.r-project.org/). The significance of the differences in the expression of lncRNAs between tumour and normal cell lines was assessed by the Wilcoxon test. Generally,
P < 0.05 was considered significant unless otherwise specified.
Discussion
As a novel form of programmed cell death, cuproptosis is a type of copper-dependent cell death that is distinct from apoptosis, necroptosis, and ferroptosis [
19]. Ferroptosis is triggered by the iron-dependent peroxidation of oxidizable membrane phospholipids [
20]. Similar to ferroptosis, cuproptosis derives from excessive intracellular copper-induced accumulation of lipoylated dihydrolipoamide S-acetyltransferase (DLAT), which is related to the mitochondrial tricarboxylic acid (TCA) cycle; finally, it results in proteotoxic stress and the occurrence of cell death [
12]. Moreover, Tsvetkov et al. further identified a series of key genes or proteins involved in the process of cuproptosis, which was closely associated with copper imbalance [
12]. Several studies have indicated that copper homeostasis plays an important role in the development of HNSCC and that regulating copper metabolism affects the cell growth and proliferation of different cancers [
21‐
23]. To the best of our knowledge, this was the first study to investigate the possible roles and mechanisms of cuproptosis in the development, prognosis, and chemotherapy efficacy of HNSCC.
In the present study, we collected and obtained crucial genes related to cuproptosis from previous studies to identify the cuproptosis-related lncRNAs that were candidates for the prognostic signature for HNSCC. Then, a novel prognostic 7-lncRNA model was constructed that was significantly correlated with the prognosis of HNSCC. In previous investigations, several ferroptosis-related signatures were developed to predict the prognosis of HNSCC patients [
24]. Some studies have also shown that prognostic cuproptosis-related signatures have been established for various tumours, including osteosarcoma, lung adenocarcinoma, and liver cancer [
25‐
27]. However, the prognostic signature of cuproptosis-related lncRNAs in HNSCC has rarely been explored. Here, we reported that the seven CRLs were differentially expressed between tumour and normal cell lines and were associated with the OS and PFS of HNSCC, suggesting a potential role in the prediction of HNSCC survival. Additionally, we found that the expression of AP001372.2, MIR9-3HG, AL160314.2, POLH-AS1, and AL109936.2 was upregulated and AC090587.1 and WDFY3-AS2 were downregulated in tumour cell lines compared with normal cell lines. The detailed mechanisms of seven cuproptosis-related lncRNAs in cancer progression and prognosis are unknown. Hence, despite the important prognostic value of the cuproptosis-related lncRNA signature identified in this study, future research is urgently needed to elucidate their mechanisms in HNSCC.
In this study, the low-risk groups presented higher TIDE scores. However, a higher TIDE score has been reported to be associated with lower responsiveness to both anti-PD-1 and anti-CTLA-4 treatment [
28]. Due to our results, no significant differences in CD274 (PD-L1) or TMB score were observed between the two risk groups. However, several studies have suggested that high TMB may serve as a preferable biomarker, which optimizes the efficacy of PD-1/PD-L1 inhibition in improving survival, and the Food and Drug Administration (FDA) has approved pembrolizumab for the treatment of solid tumours with high TMB (TMB-H or TMB ≥ 10) [
17,
29,
30]. Interestingly, patients with lower TMB scores and lower risk scores had a better prognosis in the long term. Therefore, in our opinion, in contrast to other solid cancers, the immunotherapy efficacy and prognosis of HNSCC might be determined by a combination of multiple factors, including the TIDE score, PD-L1 expression, TMB scores, and risk scores.
Of the seven CRLs in our study, MIR9-3HG, POLH-AS1, and WDFY3-AS2 are reported to play important roles in various cancers by regulating ferroptosis [
31‐
33]. Additionally, POLH-AS1 was identified to regulate the process of necroptosis in hepatocellular carcinoma [
34]. It was also reported that MIR9-3HG knockdown inhibited cell proliferation and promoted apoptosis in cervical cancer [
35]. Silencing WDFY3-AS2 significantly inhibited proliferation, migration and invasion but accelerated cell apoptosis in cisplatin-resistant ovarian cancer. Obviously, these three lncRNAs were involved in a series of programmed cell death pathways, including cuproptosis, ferroptosis, apoptosis, and necroptosis, indicating that the different types of programmed cell death could be seen as a single, coordinated cell death system in which the individual pathways are highly interconnected and can flexibly compensate for one another. Hence, despite the important prognostic value of the CRL signature identified in our study, future studies are needed to elucidate their mechanisms in HNSCC.
There were some limitations in our study. First, we only performed internal verification based on the TCGA database, and we still need other databases for external validation of our signature. Second, the mechanism of CRLs in HNSCC remains to be further identified in HNSCC tissues compared with control tissues.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.