A novel cuproptosis-related lncRNA prognostic risk score model for gastric cancer based on bioinformatics

Abstract

Objective

To establish an effective cuproptosis-related long non-coding ribonucleic acid (lncRNA) (CRL) prognostic risk score (RS) model (CRLPRSM) for the prediction of the outcomes of patients with gastric cancer (GC).

Introduction

Cuproptosis is an up-to-date mode of cell death, and its action mechanism is different from all other known mechanisms that regulate cell death. LncRNAs are RNA species that are over 200 nt long and do not encode proteins. They have prominent actions in tumor onset and development, and their involvement in a variety of intracellular regulatory processes is vital for cell proliferation and differentiation, so they may serve as prognostic biomarkers in tumor patients.

Methods

We retrieved cuproptosis-associated clinical and lncRNA expression data of GC cases from The Cancer Genome Atlas Stomach Adenocarcinoma (TCGA-STAD). Then a CRLPRSM was built based on univariate and multivariate Cox regression (UCR and MCR) analyses. As per the RS, the patients fell into high- and low-risk group. Later, the predictive efficacy of the CRLPRSM was confirmed with the aid of Kaplan-Meier (KM) analysis and receiver operating characteristic (ROC) curve analysis. Next, combining independent prognostic factors in clinical characteristics, we plotted a prognosis-related nomogram to predict one-year, three-year and five-year overall survival (OS) in GC patients. Finally, we implemented Gene Ontology enrichment analysis (GOEA) and Kyoto Encyclopedia of Genes and Genomes enrichment analysis (KEGGEA) for clarifying the possible biological actions and molecular mechanisms.

Results

The constructed CRLPRSM consisted of 3 CRLs, namely, AC092574.1, MAGI2-AS3, AC090204.1. It was found that the hazard ratio (HR) was 1.911 (1.337–2.731) (p < 0.001) in UCR analysis and 1.852 (1.286–1.668) (p < 0.001) in MCR analysis, and the AUC of the CRLPRSM was 0.649. Moreover, the KM analysis showed a pronounced intergroup difference in survival, and the nomogram illustrated some clinical benefits of CRLPRSM. Furthermore, GO terms and KEGG pathways were unveiled to be significantly enriched.

Conclusion

The constructed CRLPRSM has a significant predicted value for GC patient prognosis, and CRLs may become novel hallmarks for clinical treatment of GC.

Keywords

cuproptosis LncRNA gastric cancer prognostic TCGA

Introduction

Gastric cancer (GC) remains a prominent malignancy endangering human health, despite a significant decrease in its occurrence over the previous several decades.¹ Effective GC therapies include targeted therapy, surgery, chemotherapy, radiotherapy and immunotherapy, and multidisciplinary comprehensive therapy is proven to improve the survival of patients.² Classification based on molecular subtypes provides options for individualized treatment of GC. However, the prognosis is poor in patients with poorly differentiated histological subtypes of GC and subtypes with immune-related markers of GC. Therefore, it is still necessary to find new targets for tumor therapy in improving the outcome of patients with GC.

Copper ion is a double-edged sword: it is recognized as an important cofactor of essential enzymes throughout the animal kingdom, and excessive copper ion can lead to cell death.^3,4 Although the mechanisms of toxicity induced by other important metals, such as iron, have been well defined, the mechanisms of copper-induced cytotoxicity are unclear.⁵ A study of Tsvetkov et al.⁶ has revealed that the mode by which copper exerts its toxicity is different from other known modes of cell death, such as apoptosis, ferroptosis, pyroptosis, and necroptosis. This previously unknown mode of cell death is called cuproptosis, which is regulated via an ancient mechanism known as protein lipoylation. Moreover, this study has also demonstrated that the abundance of Ferredoxin 1 (FDX1) and lipoylated proteins are highly correlated across a large spectrum of human tumors, and that cell lines containing a high concentration of lipoylated proteins are sensitive to cuproptosis, implying that copper ionophore therapies should be targeted at cancers with such a metabolic profile.⁶ The above findings may also stimulate the exploration of the use of copper in cancer therapies. It follows that building a cuproptosis-related lncRNA (CRL) prognostic risk score (RS) model (CRLPRSM) is critical for predicting the prognosis of GC.

Long non-coding RNAs (LncRNAs) are vital modulators of various life activities.⁷ They are implicated in several processes in the development of cancer and utilized as potential biomarkers for the prediction of the prognosis of cancer patients.⁸ Additionally, multiple studies have shown that lncRNAs stimulate tumor onset and progression in a variety of malignancies.⁹ Consequently, constructing a CRL to establish a prediction model is of value for the prediction of the prognosis of GC cases.

Therefore, on the strength of bioinformatics analysis, we built a CRLPRSM and validated its value in estimating the outcome of GC cases.

Methods and materials

Data source

This retrospective study was implemented with the aid of a public database. We retrieved clinical data (age, gender, grade, stage) and lncRNA expression data of GC cases from The Cancer Genome Atlas Stomach Adenocarcinoma (TCGA-STAD, https://cancergenome.nih.gov/). Totaling 407 tissue specimens were enrolled, comprising 375 GC tissues and 32 normal controls, and subjects with incomplete clinical data were excluded.

Identification of CRLs

Thirteen CRLs were retrieved from previous publications^[6]. Pearson correlation analysis was conducted to establish a cuproptosis-related mRNA-lncRNA co-expression network by use of “limma” package in R v3.48.3, with | Coefficient| > 0.3 and p < 0.001 as the threshold.^10,11 Then we visualized this network with the aid of Cytoscape v3.9.0.

CRLPRSM construction

We distinguished CRLs influencing survival via univariate and multivariate Cox regression (UCR and MCR) analyses, and a difference with p < 0.01 indicated statistical significance. We implemented MCR analysis with the aid of “survival” package in R, and the optimal CRLPRSM was developed. The RS was calculated on the strength of the formula: RS = Σ (coef_(lncRNAs) × expr_(lncRNAs)), where coef_(lncRNAs) is the coefficient of a lncRNA related to survival, and expr_(lncRNAn) stands for a lncRNA expression.

Validation of the CRLPRSM

We implemented UCR and MCR analyses to analyze the association of prognosis with clinical characteristics and the RS. Time-dependent ROC curves were plotted with the aid of the “survival ROC” package in R to evaluate the predicted reliability for survival via various clinical characteristics and the RS. According to the RS, GC patients fell into high- and low-risk groups, and Kaplan-Meier (KM) survival analysis was employed for the estimation of the intergroup difference in survival using “survival” package in R.

Studies have evidenced the wide-spectrum application of nomogram in the prediction of cancer prognosis. Herein, independent prognostic factors for GC were identified by MCR analysis, and clinical factors and RS were combined to establish a nomogram to analyze the 1-year, three-year, and five-year overall survival (OS) rates. Lastly, we took a measurement of the validity of the nomogram by way of the concordance index and calibration curves.

Gene Ontology enrichment analysis (GOEA) and Kyoto Encyclopedia of Genes and Genomes enrichment analysis (KEGGEA)

We implemented GOEA and KEGGEA by way of clusterProfiler, org.Hs.eg.db, enrichplot and ggplot2 packages in R, and the enrichment results with p < 0.05 were output. Then the diagrams of the top 10 terms with the most statistically significant biological processes (BP), cellular components (CC) and molecular functions (MF) in GOEA and those of the top 29 terms with the greatest statistical significance in KEGGEA were drawn.

Statistical analysis

R v4.1.2 was utilized for statistical processing of a blanket inclusion of data. p < 0.05 hinted a statistically significant difference.

Results

CRLs showing striking prognostic value in GC

As demonstrated by the UCR analysis, 6 CRLs showed substantial relevance to the survival of TCGA GC patients (p < 0.01), including 2 CRLs with low RS (HR <1) and 4 CRLs with high RS (HR > 1) (Table 1). Further, the MCR analysis identified 3 CRLs with prognostic value from the above 6 CRLs, namely, AC092574.1, MAGI2-AS3, AC090204.1 (Table 2). Later, the optimal CRLPRSM was built by utilizing the 3 CRLs and a visualized cuproptosis-related mRNA-lncRNA co-expression network (Figure 1). The RS was computed as follows: RS = -0.54171903 × expr_(AC092574.1) + 0.311543963698179 × expr_(MAGI2-AS3) + 0.053274377 × expr_(AC090204.1).

Table 1.

Univariate Cox analysis identified 6 cuproptosis-related lncRNAs.

Gene	HR	HR.95 L	HR.95H	p-value
AC092574.1	0.691708	0.494211	0.968128	0.031651
MAPKAPK5-AS1	0.907836	0.825633	0.998224	0.04586
MAGI2-AS3	1.207486	1.023054	1.425167	0.025782
AP001528.2	1.424289	1.106673	1.833062	0.006009
MSC-AS1	1.420108	1.066114	1.891642	0.016503
AC090204.1	1.041297	1.000125	1.084163	0.049293

HR < 1 indicate low risk, HR > 1 indicate high risk.

Table 2.

Multivariate Cox analysis identified 3 cuproptosis-related lncRNAs.

Gene	Coef	HR
AC092574.1	-0.54171903	0.58174735
MAGI2-AS3	0.311543963698179	1.36553182
AC090204.1	0.053274377	1.054719

Figure 1.

Filtering of CRLs with prognostic value in GC. (a) Prognostic CRL‐mRNA co-expression network. and (b) The Sankey diagram of lncRNA-mRNA relationship.

Here, GC patients fell into high- and low-risk groups as per the RS, and into high- and low-expression groups as per the expressions of 3 different lncRNAs. As unveiled by KM survival analysis the OS was shorted in high-risk group relative to in low-risk group, demonstrating that the RS may be a predictor of the prognosis (Figure 2(a)). Likewise, high-expression group with high-RS lncRNAs exhibited a shorter OS relative to low-expression group (Figure 2(b)–(d)). Risk curves and scatterplots were plotted to display the RS-survival relationship in GC patients (Figure 3(a) and (b)), which illustrated that the death rate was dependent on the RS. The heatmap of the 3 CRLs denoted the up-regulation of AC090204.1 and MAGI2-AS3 in high-risk group and the notable up-regulation of AC092574.1 in low-risk group (Figure 3(c)).

Figure 2.

Survival curves of GC patients, based on data obtained from the TCGA. (a) Survival curves in different groups. and (b–d) Survival curves of GC patients with discrepant expression levels of 3 CRLs.

Figure 3.

RS analysis of the 3 prognostic CRLs, obtained from the TCGA. (a) RS-based risk curves of specimens. (b) The scatterplot on the strength of the survival of specimens. and (c) The heatmap denoted the expression levels of 3 CRLs in high and low risk groups.

Evaluation of the CRLPRSM for GC

Whether the aforementioned 3 CRLs were independent prognostic indicators for GC was validated by UCR and MCR analyses. The HR was 1.911 (95% CI 1.337–2.731) (p < 0.001) in the UCR analysis (Figure 4(a)) and 1.852 (95% CI: 1.286–2.668) (p < 0.001) in the MCR analysis (Figure 4(b)). As a result, the 3 CRLs were identified to be independent prognostic indicators for GC. With the aim of assessing the sensitivity and specificity of the RS in predicting the outcome of GC cases, the area under curve (AUC) of the RS was determined. It was uncovered that the AUC of the RS (0.649) was higher relative to that of other clinical indicators (Figure 5(a)), demonstrating that 3 CRLs were rather valuable for a GC predictive RS model.

Figure 4.

UCR and MCR regression analyses of the 3 CRLs in GC. (a) UCR analysis results of RS and clinical hallmarks. and (b) MCR analysis results of RS and clinical hallmarks.

Figure 5.

Assessment of the prognostic RS model. (a) The AUC for RS and clinical hallmarks and the ROC curves. (b) Calibration curve of nomogram. and (c) Nomogram predicting OS for GC patients.

Plotting of a predictive nomogram

Furthermore, combined with clinical factors, we plotted a nomogram to predict one-year, three-year and five-year OS in GC patients, including gender, grade, age, TNM stage and RS (Figure 5(c)). As illustrated by the correction curve, the actual OS of GC patients was identical to that estimated by the nomogram (Figure 5(b))

Gene enrichment analysis

For the purpose of clarifying the molecular mechanism of GC signature, we carried out gene enrichment analysis. BP analysis results unveiled striking enrichment in the extracellular matrix (ECM) structural constituent and ECM binding. CC analysis manifested remarkable enrichment in the collagen-containing ECM. Moreover, MF analysis demonstrated that enrichment primarily appeared in the extracellular structure organization and ECM organization (Figure 6(a)–(d)). As revealed by the KEGGEA, pronounced enrichment appeared in the cGMP-PKG signaling pathway, ECM-receptor interaction, proteoglycans in cancer, cAMP signaling pathway and the PI3K-Akt signaling pathway (Figure 7(a)–(c)).

Figure 6.

Plots for GO ORA results of the TCGA in GC patients. GOEA including BP, CC and MF. (a) Bubble plot of GO enrichment analysis. (b, and c) Column chart of enrichment analysis. and (d) Loop graph of GO enrichment.

Figure 7.

Plots for KEGG ORA results of the TCGA in GC patients. (a) KEGG enrichment column diagram, (b) KEGG enrichment bubble chart, and (c) KEGG enrichment loop graph.

Discussion

GC is emerging as an increasingly common malignancy in the globe, which is seriously harmful to human health, and the biological differences between tumors increase the complexity of determining standard treatments for GC.² More patients with GC benefit from individualized therapy based on molecular subtype classification. However, for poorly differentiated GC and tissue subtypes without immunotherapeutic activity markers, the therapeutic effect is poor, so novel treatment targets are still required for improving survival. With the recent advancement of sequencing technology, it is not uncommon to utilize biomarkers diagnose cancer and predict the prognosis.¹²

Copper ion is a core cofactor of enzymes in animals, and excessive copper ion may induce cell death.⁴ Genetic variations in copper homeostasis can lead to diseases endangering life.¹³ Both copper ion carriers and copper chelating agents are considered as anticancer agents.^14,15 Cuproptosis is a novel mode of cell death, and its action mechanism is different from the existing mechanisms regulating cell death, including necroptosis, apoptosis, pyroptosis and ferroptosis.⁶

LncRNAs, a new type of transcripts, are encoded by the genome and are mostly not translated into proteins,¹⁶ and they have various cellular and physiological functions.¹⁷ Recently, it is reported that functions of lncRNAs as tumor promotors or suppressors to influence the onset and progression of human tumors.¹⁸ Beyond that, it has been evidenced that considerable lncRNAs have relevance to multifold cancers, and the change or mutation of lncRNA expressions can propel tumor incidence and metastasis.¹⁹ Over the past 10 years, researchers have made great efforts in the clinical application of RNA-based therapies, and the lncRNA-based targeted therapy for cancer and many other diseases has come into the spotlight.²⁰ Mounting evidence showed the involvement of lncRNAs in the occurrence, metastasis and recurrence of GC, suggesting that they are valuable in the diagnosis, treatment and prognosis of GC.²¹

Cancer cells have a higher demand for copper than non-mitotic cells in tumors. Studies have shown that copper concentrations are elevated in tumors or serum in animal models and in patients with multiple tumors, including GC, breast cancer, lung cancer, oral cancer, thyroid cancer, gallbladder cancer, gynecological cancer and prostate cancer.¹⁴ Cuproptosis also brings new treatment opportunities for human beings to treat copper metabolism disorders, such as Menke disease, hepatolenticular degeneration (Wilson disease), CTR1 deficiency and other copper metabolism disorders.²² These new results may stimulate the exploration of the use of copper in cancer and metabolic disease therapies. Therefore, it is important to establish a CRLPRSM to predict cancer outcome.

Here, a CRLPRSM was established and verified. It was the first time that the CRLPRSM had been employed to predict GC prognosis. The ROC curve illustrated that the AUC of the CRLPRSM formed by 3 CRLs was 0.649, demonstrating that the CRLPRSM had a high predictive value. The nomogram combined with clinical prognostic factors and 3 CRLs-based RS showed high value in predicting the one-year, three-year and five-year OS of GC patients. It can be concluded that the CRLPRSM formed by these 3 CRLs can be used as useful predictors of the survival of GC patients. In addition, we implemented GOEA and KEGGEA to reveal the potential biological roles and molecular mechanisms of these lnRNAs.

Recently, considerable studies have investigated the prognosis of associated genes and diseases by bioinformatics analysis.²³ Cuproptosis is a new mode of cell death, which may open a new way to kill cancer cells.⁶ At present, there is no CRLPRSM for GC based on TCGA database. Meanwhile, research on the role of CRLs in tumors will also be the trending spot. In this study, 3 CRLs were filtered and employed to establish a CRLPRSM model. This study may assist physicians in making personalized and optimal treatment decisions, and may also stimulate the exploration of the use of copper in the treatment of cancer.

However, limitations of this study included: (1) Typical statistical analysis methods were adopted to evaluate the CRLPRSM formed by 3 CRLs. Although the effectiveness of these analysis methods has been validated in numerous studies, more sophisticated methods and techniques are required in future studies. (2) The sample is drawn from a public database with a small sample size, which will be validated by future experiments with large sample data. (3) To verify bioinformatics predictions, CRLs should be investigated in the aspects of functional annotation and molecular pathways. (4) In this study, experimental validation was not conducted in vitro and in vivo, and thus relevant experimental validation will be needed in the future.

Conclusion

To sum up, we established and identified a CRLPRSM consisting of 3 CLRs (AC092574.1, MAGI2-AS3 and AC090204.(1) for GC, which shows accuracy in predicting the outcome of GC patients. In subsequent research, these 3CRLs may become novel hallmarks for GC therapy and provide an indication on the path towards a more individualized treatment and accurate prediction of prognosis.

Footnotes

Acknowledgements

The authors gratefully acknowledge the Cancer Genome Atlas (TCGA) database, which made the data available.

Author contribution

This research was conducted in collaboration with all authors. Xue Du and Chenbao Chen performed the data curation and analysis. Xue Du, Chenbao Chen and Yu Xiao analyzed and interpreted the results. Xue Du, Lu Yang, Yu Cui and Bangxian Tan drafted and reviewed the manuscript. All authors read and approved the final manuscript.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Xue Du

Chunbao Chen

References

Waldum

Fossmark

. Gastritis, Gastric Polyps and Gastric Cancer. Int J Mol Sci, 22(12). 2021, 6548. DOI: 10.3390/ijms22126548.

Joshi

Badgwell

. Current treatment and recent progress in gastric cancer. CA Cancer J Clin 2021; 71(3): 264–279. DOI: 10.3322/caac.21657.

Kim

Nevitt

Thiele

. Mechanisms for copper acquisition, distribution and regulation. Nat Chem Biol 2008; 4(3): 176–185. DOI: 10.1038/nchembio.72.

Bush

Casini

, et al. Connecting copper and cancer: from transition metal signalling to metalloplasia. Nat Rev Cancer, 22(2). 2022, 102–113. DOI: 10.1038/s41568-021-00417-2.

Hasinoff

Yadav

Patel

, et al. The cytotoxicity of the anticancer drug elesclomol is due to oxidative stress indirectly mediated through its complex with Cu (II). J Inorg Biochem 2014; 137: 22–30. DOI: 10.1016/j.jinorgbio.2014.04.004.

Tsvetkov

Coy

Petrova

, et al. Copper induces cell death by targeting lipoylated TCA cycle proteins. Science 2022; 375(6586): 1254–1261. DOI: 10.1126/science.abf0529.

Mei

, et al. A potential oncogene in human cancers. Cancer Cell Int, 20(1). 2020, 536. DOI: 10.1186/s12935-020-01608-x.

Tang

Gan

, et al. New insights into long non-coding RNAs in non-small cell lung cancerEpub 2020 Sep 26. Erratum in. Biomed Pharmacotherbiomed Pharmacother 2020; 131132. DOI: 10.1016/j.biopha.2020.110775.

Pan

Xiao

Zhao

, et al. The functional roles of long noncoding RNA DANCR in Human Cancers. J Cancer, 11(23). 2020, 6970–6981. DOI: 10.7150/jca.44384.

10.

Schober

Boer

Schwarte

. Correlation Coefficients: Appropriate Use and Interpretation. Anesth Analg 2018; 126(5): 1763–1768. DOI: 10.1213/ANE.0000000000002864.

11.

Subramanian

Tamayo

Mootha

, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A, 102(43). 2005, 15–50. DOI: 10.1073/pnas.0506580102.

12.

Magalhães

Alvarez-Lorenzo

Concheiro

, et al. RNAi-based therapeutics for lung cancer: biomarkers, microRNAs, and nanocarriers. Expert Opin Drug Deliv 2018; 15(10): 965–982. DOI: 10.1080/17425247.2018.1517744.

13.

Bandmann

Weiss

Kaler

. Wilson's disease and other neurological copper disorders. Lancet Neurol, 14(1). 2015, 103–113. DOI: 10.1016/S1474-4422(14)70190-5.

14.

O'Day

Eggermont

Chiarion-Sileni

, et al. Final results of phase III SYMMETRY study: randomized, double-blind trial of elesclomol plus paclitaxel versus paclitaxel alone as treatment for chemotherapy-naive patients with advanced melanoma. J Clin Oncol 2013; 31(9): 1211–1218. DOI: 10.1200/JCO.2012.44.5585.

15.

Tsang

Posimo

Gudiel

, et al. Copper is an essential regulator of the autophagic kinases ULK1/2 to drive lung adenocarcinoma. Nat Cell Biol, 22(4). 2020, 412–424. DOI: 10.1038/s41556-020-0481-4.

16.

Khalil

Guttman

Huarte

, et al. Many human large intergenic noncoding RNAs associate with chromatin-modifying complexes and affect gene expression. Proc Natl Acad Sci U S A 2009; 106(28): 11667–11672. DOI: 10.1073/pnas.0904715106.

17.

Guttman

Amit

Garber

, et al. Chromatin signature reveals over a thousand highly conserved large non-coding RNAs in mammals. Nature, 458(7235). 2009, 2–7. DOI: 10.1038/nature07672.

18.

Wang

, et al. LncRNA HOXA-AS2 and its molecular mechanisms in human cancer. Clin Chim Acta 2018; 485: 229–233. DOI: 10.1016/j.cca.2018.07.004.

19.

Bhan

Soleimani

Mandal

. Long Noncoding RNA and Cancer: A New Paradigm. Cancer Res, 77(15). 2017, 3965–3981. DOI: 10.1158/0008-5472.CAN-16-2634.

20.

Winkle

El-Daly

Fabbri

, et al. Noncoding RNA therapeutics - challenges and potential solutions. Nat Rev Drug Discov, 20(8). 2021, 629–651. DOI: 10.1038/s41573-021-00219-z.

21.

Liu

Ding

, et al. Long non-coding RNAs: Biogenesis, functions, and clinical significance in gastric cancer. Mol Ther Oncolytics, 23. 2021, pp. 458–476. DOI: 10.1016/j.omto.2021.11.005.

22.

Aggarwal

Bhatt

. Advances in Treatment of Wilson Disease. Tremor Other Hyperkinet Mov (N Y), 8. 2018, 525. DOI: 10.7916/D841881D.

23.

Ren

Zhang

Ding

, et al. Development and validation of a novel survival model for head and neck squamous cell carcinoma based on autophagy-related genes. Genomics 2021; 113(1 Pt 2): 1166–1175. DOI: 10.1016/j.ygeno.2020.11.017.