Abstract
Background. Gastric cancer is the most common malignant tumor of the digestive system. It has a poor prognosis and is clinically challenging to treat. Ferroptosis is a newly defined mode of programmed cell death. The roles and prognostic value of ferroptosis-related long noncoding RNAs (lncRNAs) in gastric cancer remain unknown. Results. In the current study, 20 ferroptosis-related lncRNAs were identified via univariate Cox analysis, least absolute shrinkage, and selection operator Cox regression analysis and used to construct a prognostic signature and classify gastric cancer patients into high-risk and low-risk groups. The signature was validated using TCGA training and testing cohorts. The risk signature was an independent prognostic indicator of survival and accurately predicted the prognoses of patients with gastric cancer. It was also associated with immune cell infiltration. Gene set enrichment analysis was used to investigate underlying mechanisms that the 20 ferroptosis-related lncRNAs were involved in. Chemosensitivity and immune checkpoint inhibitor analyses indicated that high-risk patients were more sensitive to the immune checkpoint inhibitor programmed cell death protein 1. Conclusions. The important role of ferroptosis-related lncRNAs in immune infiltration identified in the current study may assist the determination of personalized prognoses and treatments in patients with gastric cancer. These 20 lncRNAs can be used as the diagnostic and prognostic markers for gastric cancer.
1. Introduction
Gastric cancer is the most common malignant tumor of the digestive system, ranking fifth in incidence and fourth in mortality globally. In 2020, more than one million new cases of gastric cancer and 769,000 deaths were reported worldwide [1]. Because of the increasing incidence of autoimmune gastritis and dysbiosis of the gastric flora, the incidence of gastric cancer is also gradually increasing in young people [2, 3]. To date, the precise mechanisms underlying the initiation and progression of gastric cancer remain unknown.
Ferroptosis is a newly defined mode of cell death that is iron dependent and is triggered by lipid peroxidation and lethal reactive oxygen species (ROS). It is distinct from necrosis, apoptosis, and autophagy [4]. Long noncoding RNAs (lncRNAs) are noncoding RNA molecules of approximately 200 nucleotides in length that induce the occurrence, development, and metastasis of tumors by mediating chromosome modification, transcriptional activation, and interference [5, 6]. Recent reports suggest that ferroptosis-related lncRNAs play important roles in tumorigenesis, progression, and metastasis via multiple mechanisms. Wu et al. [7] reported that lncRNA NEAT1 promoted ferroptosis and ferroptosis sensitivity in non-small-cell lung cancer by regulating the expression of ACSL4. Qi et al. [8] reported that lncRNA GABPB1-AS1 could induce ferroptosis in hepatocellular carcinoma cells by inhibiting the translation of GABPB1. Wang et al. [9] reported that LINC00336 regulated the expression of cystathionine-β-synthase to promote lung cancer cell ferroptosis by serving as an endogenous microRNA 6852 “sponge.” Gai et al. [10] proposed that MT1DP regulates the miR-365a-3p/NRF2 signaling pathway, resulting in the sensitization of non-small-cell lung cancer cells to erastin-induced ferroptosis.
The roles and prognostic value of ferroptosis-related lncRNAs in gastric cancer remain unclear. In the current study, ferroptosis-related lncRNAs potentially involved in gastric cancer were screened to construct a prognostic signature, and the potential mechanisms involved were investigated. To the best of our knowledge, the study is the first to construct and validate a ferroptosis-related lncRNA prognostic signature for used in gastric cancer patients. Functional enrichment, immune cell infiltration, chemosensitivity, and immune checkpoint inhibitors were also analyzed. The study resulted in the development of an effective, practical, and quantitative approach for clinicians to use to predict survival and formulate individualized treatments in gastric cancer patients.
2. Materials and Methods
2.1. Data Collection and Differential Ferroptosis-Related lncRNAs
Raw counts from RNA-seq transcriptome data and corresponding clinical data derived from gastric cancer tissues were extracted from the TCGA database. Ferroptosis-related genes were downloaded from the FerrDb database [11]. Differentially expressed ferroptosis-related lncRNAs were screened via Pearson’s correlational analyses (, ).
2.2. Construction and Validation of the Prognostic Ferroptosis-Related lncRNA Signature
Ferroptosis-related lncRNAs associated with survival were evaluated via univariate Cox regression analysis. Gastric cancer samples were randomly divided into a training set and a testing set at a 1 : 1 ratio. A prognostic signature was then constructed in patients with gastric cancer based on LASSO Cox regression. The formula used to calculate prognostic signature risk scores was where Exp [lncRNA] is the corresponding expression of the included lncRNA, and coef [lncRNA] represents the regression coefficient. The risk score of each patient was calculated, and each patient was assigned to a low-risk group or a high-risk group based on the median risk score in the training and testing cohorts. Kaplan-Meier analysis and areas under ROC curves were used to evaluate the performance of the prognostic signature.
2.3. Correlations with Clinicopathological Characteristics and the Establishment of a Nomogram
Univariate and multivariate Cox regression analyses were conducted to investigate whether risk scores and relevant clinicopathological characteristics were associated with overall survival in gastric cancer patients. Nomograms that included age, sex, grade, stage, TNM classifications, and risk score were used to calculate the total score and predict 1, 3, and 5-year survival probabilities. Resulting 1, 3, and 5-year-dependent ROC curves were used to evaluate nomogram performance.
2.4. Tumor-Infiltrating Immune Cells and Functional Enrichment Analysis
Infiltration levels of distinct immune cells were quantified and evaluated in the low-risk group and the high-risk group using the “CIBERSORT” package [12]. KEGG by gene set enrichment analyses was performed to explore different molecular mechanisms in high-risk and low-risk patients. A false discovery rate value of <0.05 was considered statistically significant.
2.5. Prediction of Responses to Chemotherapy and Immunotherapy
IC50s of common chemotherapeutics were calculated to evaluate clinical responses to gastric cancer treatment using pRRophetic [13] and ggplot2 packages in . The Wilcoxon signed rank test was conducted. Relationships between risk score and expression levels of genes related to immune checkpoints were investigated, including PD1, PDL1, and CTLA4.
2.6. Statistical Analysis
All statistical analyses were conducted using statistical software version 4.0.4 and strawberry-perl-5.32.0.1. Differentially expressed lncRNAs were identified using the Benjamini-Hochberg method. Hazard ratios and a 95% confidence intervals were evaluated via univariate and multivariate Cox regression models. was considered to indicate statistical significance.
3. Results
3.1. Identification of Prognostic Ferroptosis-Related lncRNAs in Gastric Cancer
RNA-seq transcriptome data and matched clinical data for 32 normal gastric tissues and 371 gastric cancer tissues were downloaded from the TCGA database. A total of 259 ferroptosis-related genes were obtained from the FerrDb database (Table S1). A total of 1,378 differentially expressed ferroptosis-related lncRNAs were identified via Pearson’s correlational analyses (, ). Preliminary screening via univariate Cox analysis identified 95 ferroptosis-related lncRNAs that were significantly correlated with survival (Table S2).
3.2. Construction and Validation of a Ferroptosis-Related lncRNA Prognostic Signature
A total of 371 gastric cancer samples were randomly allocated to a training set () or a testing set (). Twenty lncRNAs were identified and used to construct a prognostic signature via least absolute contraction and selection operator (LASSO) Cox regression analysis: AC114271.1, AC147067.2, AL353796.1, AC104958.1, AC087521.1, AL590705.3, AC068790.7, AC090772.1, LINC01094, AC007405.3, AC083902.1, LINC00460, AC005165.1, AC048382.2, AC106782.5, STX18-AS1, AL355574.1, CYMP-AS1, AC006547.1, and LINC02696 (Figures 1(a) and 1(b)). Compared with normal gastric tissues, seven of these lncRNAs (AC114271.1, AL353796.1, AC104958.1, AC007405.3, STX18-AS1, AL355574.1, AC006547.1) were downregulated in cancer tissues, whereas the other thirteen were upregulated. The following risk score was calculated:
(a)
(b)
-.
Using the median risk score as the cut-off value, patients in the training set and the testing set were divided into high-risk and low-risk groups. Distribution patterns of risk scores and survival status are shown in Figures 2(a)–2(d). The changing trends in expression levels of the 20 ferroptosis-related lncRNAs as determined via heatmapping were concordant with their risk scores in the prognostic signature (Figures 2(e) and 2(f)). In Kaplan–Meier survival analysis, the high-risk group exhibited worse overall survival than the low-risk group in both the training set and the testing set (Figures 2(g) and 2(h)). Receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) values were 0.830 in the training set and 0.716 in the testing set (Figures 2(i) and 2(j)).
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
3.3. Associations between Prognostic Risk Score and Clinicopathological Characteristics
Univariate and multivariate Cox regression analyses were conducted in the training set and the testing set to investigate whether risk score independently predicted the prognoses of patients with gastric cancer. Univariate Cox regression stage and risk score were significantly associated with overall survival in both cohorts (Figures 3(a) and 3(b)). Multivariate Cox regression age, stage, and risk score were significant prognostic indicators in both cohorts (Figures 3(c) and 3(d)).
(a)
(b)
(c)
(d)
Associations between risk score and clinicopathological features were assessed. Overall survival was significantly longer in the low-risk group than in the high-risk group in patients aged both >65 years and ≤65 years, in both sexes, in grades 1-2 and grade 3, in stages I-II and III-IV, and in T1-2, T3-4, M0, M1, N0, and N1–3 classifications (Figures 4(a)–4(n)). Prognostic nomograms derived from both cohorts composed of clinicopathological characteristics and risk scores were established as a quantitative and visual method for predicting 1, 3, and 5-year overall survival probability in gastric cancer patients (Figures 5(a) and 5(b)). In the training set, the respective AUC values for 1, 3, and 5-year overall survival were 0.830, 0.852, and 0.947, and in the testing set, they were 0.716, 0.724, and 0.684 (Figures 5(c) and 5(d)).
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
(k)
(l)
(m)
(n)
(a)
(b)
(c)
(d)
3.4. Risk Scores and Immune Cell Infiltration
To further investigate the reasons for the different prognoses in the high-risk and low-risk groups, differences in immune cell infiltration and correlations between immune cell infiltration and risk score were analyzed. Resting dendritic cells, eosinophils, monocytes, and M2 macrophages were significantly more prevalent in the high-risk group. Plasma cells and follicular helper T cells were significantly more prevalent in the low-risk group (Figure 6(a)). Numbers of resting dendritic cells, eosinophils, M2 macrophages, monocytes, and resting CD4 memory T cells were positively correlated with risk score, and numbers of follicular helper T cells were negatively correlated with risk score (Figure 6(b)). Numbers of resting dendritic cells, eosinophils, monocytes, M2 macrophages, and follicular helper T cells were strongly correlated with risk score (Figure 6(c)).
(a)
(b)
(c)
3.5. Functional Enrichment Analysis
Kyoto Encyclopedia of Genes and Genomes (KEGG) functional enrichment analyses of 20 ferroptosis-related lncRNAs were conducted to investigate differences in biological functions between the high-risk and low-risk groups (Table S3). The top five pathways enriched in the high-risk group were the hypertrophic cardiomyopathy signaling pathway, dilated cardiomyopathy signaling pathway, focal adhesion signaling pathway, extracellular matrix receptor interaction signaling pathway, and calcium signaling pathway (Figures 7(a)–7(e)). The top five pathways enriched in the low-risk group were the spliceosome signaling pathway, homologous recombination signaling pathway, oxidative phosphorylation signaling pathway, Huntington’s disease signaling pathway, and RNA polymerase signaling pathway (Figures 7(f)–7(j)).
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
3.6. Responses to Chemotherapy and Immunotherapy in High-Risk and Low-Risk Patients
The pRRophetic algorithm was used to predict the IC50s of cisplatin, docetaxel, and paclitaxel, which are common chemotherapeutic agents used in gastric cancer patients. There were no differences in sensitivity to the three chemotherapeutics based on risk score (Figures 8(a)–8(c)). Potential susceptibility to immune checkpoint inhibitors targeting the immune checkpoint proteins programmed cell death protein 1 (PD1), programmed death ligand 1 (PDL1), and cytotoxic T-lymphocyte-associated protein 4 (CTLA4) was also investigated in both groups. Samples from the high-risk group had higher expression of PD1 (Figure 8(d)), suggesting that high-risk patients may respond better to immune checkpoint inhibitors targeting PD1.
(a)
(b)
(c)
(d)
(e)
(f)
4. Discussion
Studies indicate that ferroptosis plays important roles in biological processes associated with gastric cancer [14–19]. The circ-0008035/miR-599/EIF4A1 axis can reportedly promote gastric cancer cell proliferation and suppress apoptosis and ferroptosis [14]. Exosomal miR-522 secreted by cancer-associated fibroblasts targets ALOX15 and blocks lipid-ROS accumulation, inhibiting ferroptosis in gastric cancer cells [15]. C-Myb regulates CDO1, inhibiting erastin-induced ferroptosis in gastric cancer cells by upregulating the GPX4 expression [16]. The polyunsaturated fatty acid biosynthesis pathway reportedly plays an essential role in ferroptosis and determines ferroptosis sensitivity in gastric cancer [17]. Guan et al. [18] reported that tanshinone IIA could induce p53 to upregulate gastric cancer cell ferroptosis. Niu et al. [19] confirmed that physcion 8-O-β-glucopyranoside stimulated gastric cancer cell ferroptosis by regulating the miR-103a-3p/GLS2 axis. Notably however, studies investigating ferroptosis-related lncRNA and the development of a prognostic tool for gastric cancer based on it are lacking.
In the current study, prognostic ferroptosis-related lncRNAs were screened via univariate Cox regression analyses, and 20 ferroptosis-related lncRNAs were identified via LASSO regression. A ferroptosis-related lncRNA prognostic signature for application in gastric cancer patients was then constructed and validated in two independent cohorts. In univariate and multivariate Cox regression analyses, the risk score was an independent prognostic indicator in gastric cancer patients. In survival and clinicopathological analyses, the signature accurately predicted prognoses and was an independent prognostic indicator in gastric cancer patients. Nomograms provided a quantitative and visual method for predicting 1, 3, and 5-year overall survival probabilities in gastric cancer patients. ROC curves indicated that the ferroptosis-related lncRNA prognostic signature was highly accurate and reliable. The roles of immune cells infiltrating the tumor microenvironment and responses to common chemotherapeutic agents and immune checkpoint inhibitors in gastric cancer patients were also investigated. The results of the study highlighted a novel biomarker and potential therapeutic target in gastric cancer.
The prognostic signature proposed in the present study was derived from 20 lncRNAs. Some of them reportedly participate in the development and occurrence of various tumors by regulating drug resistance. Meng et al. [20] demonstrated that the LINC00460-miR-149-5p/miR-150-5p-mutant p53 feedback loop could induce oxaliplatin resistance in colorectal cancer. LINC00460 promotes gefitinib resistance in non-small-cell lung cancer by targeting epidermal growth factor receptor by sponging miR-769-5p [21]. In clear cell renal cell carcinoma, LINC01094 can reportedly target the miR 577/CHEK2/FOXM1 axis, promoting radio resistance [22]. KEGG functional enrichment analyses elucidated the probable mechanisms of the high-risk group and low-risk group, but the specific mechanisms of ferroptosis-related lncRNAs in gastric cancer are fiendishly complex and still unclear, and this will be one focus of our future work. In the current study, correlations between the prognostic signature and chemotherapeutics and immune checkpoint inhibitors were analyzed. The model suggested that high risk scores were associated with sensitivity to immunotherapies such as PD1, but not associated with sensitivity to the common chemotherapeutic agent cisplatin, docetaxel, and paclitaxel. We surmised that immunotherapy is of greater benefit than chemotherapy in patients with high risk scores.
Immune cells that infiltrate tumors have diverse effects on tumor progression. In the present study, numbers of resting dendritic cells, eosinophils, monocytes, M2 macrophages, and follicular helper T cells were strongly correlated with risk score, indicating that these immune cells may play important roles in the occurrence and progression of gastric cancer. The roles of immune cell infiltration in gastric cancer are gradually being determined. Tumor-associated macrophages reportedly participate in the progression of gastric cancer via the TGFβ2/NF-κB/Kindlin-2 axis [23]. Macrophage-derived exosomal miR-21 mediates cisplatin resistance in gastric cancer cells by downregulating PTEN, resulting in activation of the PI3K/AKT signaling pathway [24]. Eosinophils may represent a T helper 2-biased response preventing cancer development, or they may promote a T helper 1-type response leading to the progression of precancerous lesions [25]. Melanoma antigen gene-1 may regulate CCL3 and CCL20, causing recruited dendritic cells to stimulate antitumor immunity specific to gastric cancer in vivo or in vitro [26].
To the best of our knowledge, the current study is the first to construct a ferroptosis-related lncRNA-based gastric cancer prognostic signature and validate it in gastric cancer patients. Functional enrichment, immune cell infiltration, immune checkpoint inhibitors, and chemosensitivity were also analyzed. Investigating the effects of ferroptosis-related lncRNAs on tumor immune cell infiltration will contribute to a better understanding of how the tumor microenvironment is modulated and facilitate better predictions of prognoses and treatment outcomes in patients with gastric cancer. Despite its strengths, the present study had some limitations. The main datasets in the study were obtained from the TCGA database, and other datasets should be investigated using the prognostic ferroptosis-related lncRNA signature, to reduce selection bias. Additionally, the function of the signature must be validated in clinical research with larger samples.
5. Conclusions
Twenty ferroptosis-related lncRNAs associated with prognoses in gastric cancer patients were identified, the role of immune cell infiltration was systematically explored, and correlations between chemosensitivity and immune checkpoint inhibitors were assessed. The signature developed has many potential prognostic applications and may contribute to determining individual therapeutic strategies and expanding insights into therapeutic approaches in gastric cancer patients. These 20 lncRNAs can be used as the diagnostic and prognostic markers for gastric cancer.
Data Availability
All data used in the study can be downloaded from the TCGA data repository (https://gdac.broadinstitute.org/; accessed 26 February 2021) and the FerrDb database (http://www.zhounan.org/ferrdb; accessed 26 February 2021).
Conflicts of Interest
The authors declare no conflicts of interest.
Authors’ Contributions
W.Z.C., Z.G.J., Z.R.L., and Y.C. performed the study conceptualization. Y.T. performed the data curation. J.B.X. performed the data analysis. Z.G.J. and Y.C. performed the funding acquisition. T.X. contributed to the investigation. W.Z.C. performed the study design. Z.G.J. and Y.L. performed the project administration. W.Z.C. contributed to the data acquisition. Z.R.L. performed the supervision. Z.F.F., J.F.H., and P.C.F performed the data validation. W.Z.C. performed writing the original manuscript draft. Z.R.L. contributed to the manuscript review and editing. WenZheng Chen, ZongFeng Feng, JianFeng Huang, and PengCheng Fu contributed equally to this work.
Acknowledgments
We thank the peer reviewers for their input, which improved this manuscript. This research was funded by the National Natural Science Foundation (grant number 81960503) and the Natural Science Foundation of Jiangxi Province (grant number 20202BABL216051).
Supplementary Materials
Table S1: ferroptosis-related genes downloaded from the FerrDb database. Table S2: univariate Cox regression analysis of ferroptosis-related genes. Table S3: detailed results of gene set enrichment analyses in the high-risk group and the low-risk group. (Supplementary Materials)