1. ABSTRACT Background: This study set out to elucidate the biological functions and prognostic role of ferroptosis-related lncRNAs based on a synthetic analysis of competing endogenous RNA (ceRNA) networks in clear cell renal cell carcinoma (ccRCC). Methods: Ferroptosis-related genes were obtained from the FerrDb database. The expression data and matched clinical information of lncRNAs, miRNAs, and mRNAs from The Cancer Genome Atlas (TCGA) database were obtained to identify differentially expressed RNAs (DERNAs). The lncRNA-miRNA-mRNA ceRNA network was established utilizing the common miRNAs that were predicted in the RNAHybrid, StarBase, and TargetScan databases. Then, by progressive univariate Cox regression, LASSO,and multivariate Cox regression analysis of gene expression data and clinical information, a ferroptosis-related lncRNA prognosis signature was constructed based on the lncRNAs in ceRNA. Finally, the influence of independent lncRNAs on ccRCC was explored through a series of functional and unsupervised cluster analysis. Results: A total of 35 ferroptosis-related DEmRNAs, 356 DElncRNAs, and 132 DEmiRNAs were sorted out from the KIRC cohort of TCGA database. Overlapping DElncRNA-DEmiRNA and DEmiRNA-DEmRNA interactions among the RNAHybrid, StarBase, and TargetScan databases were constructed and identified, then a ceRNA network with 77 axes related to ferroptosis was established utilizing mutual DEmiRNAs in two interaction networks as nodes. Through synthetic analysis of the expression data and clinical information of 27 lncRNAs in the ceRNA network, a 6-ferroptosis-lncRNA signature including PVT1, CYTOR, MIAT, SNHG17, LINC00265, and LINC00894 was identified in the training set. Kaplan-Meier, PCA, t-SNE analysis, risk score curve, and ROC curve were performed to confirm the validity of the signature in the training set and secondly verified in the validation set. Finally, ssGSEA and ESTIMATE analysis showed that the signature was related with immune cell infiltration and could predict immune-related phenotypes. Conclusions: Our research underlines the role of the 6-ferroptosis-lncRNA signature as a predictor of prognosis and a therapeutic alternative for KIRC.
2. INTRODUCTION Globally, about 4.4 per 100,000 people are diagnosed with renal cell carcinoma (RCC) every year, and the incidence is twice as common in males as in females, which poses a serious health threat to men. There are many subtypes of RCC, of which nearly 75–80% are clear cell renal cell carcinoma (ccRCC). Surgical resection is the main treatment for RCC, whereas, metastasis and recurrence are the main causes of treatment failure and important factors affecting the prognosis [1], therefore, immunotherapy and targeted therapy have become important means for the treatment of metastasis or recurrence RCC [2]. Nevertheless, resistance to immunotherapy and targeted drugs always occurs, ultimately leading to disease progression. Hence, looking for more effective therapies or biomarkers based on tumor gene-expression profiles to treat patients with different subsets of advanced RCC are urgently needed. In the past decade, a new type of programmed cell death has been identified by many researchers. Ferroptosis is an iron-dependent form of regulatory cell death caused by excessive lipid peroxidation and is involved in the development and progression of different tumors, including ccRCC [3]. Currently, emerging shreds of evidence have indicated that Von Hippel Lindau (VHL) gene mutation can cause hereditary and familial ccRCC. VHL gene mutation can lead to the stable production of hypoxia-induced transcription factor (HIF), which is considered to regulate hypoxia adaptation by inducing metabolic reprogramming [4]. At the same time, HIF can induce cells to absorb more lipids, which in turn drives ccRCC to synthesize more GSH to prevent the accumulation of lipid peroxides and maintain cell viability. Miess et al. have proved that inhibiting the synthesis of GSH can make ccRCC sensitive to ferroptosis and finally prevent tumor growth, and what's more surprising is that allowing the defective cells to re-express VHL can make them resistant to ferroptosis again [5]. In addition, another study also confirmed that the Hippo pathway effector TAZ can regulate the ferroptosis sensitivity of RCC [6]. Thus, modulating ferroptosis may have a number of important implications for future therapeutic practice of RCC. Long non-coding RNAs (lncRNAs) are non-coding transcripts containing more than 200 nucleotides [7]. Accumulating studies have confirmed that lncRNAs can regulate the biological processes of various cancers through various forms, and can be used as a reliable prognostic predictor of cancers. There is a growing body of studies revealed that lncRNAs can regulate tumor EMT process, pyroptosis, autophagy, ferroptosis, etc [8–11]. Instead of using a single lncRNA to analyze its prediction of disease, it is more effective to comprehensively analyze the expression profile of certain pathway-related lncRNAs. Therefore, a growing number of prognostic lncRNA signatures have been established as prognostic and therapeutic targets for cancers. For instance, in our previous research, we constructed an EMT-related lncRNA prognostic signature and analyzed its role in identifying different molecular and immune characteristics and prognosis of colorectal cancer [12], in a recent study, Xing et al. developed a prognostic signature of three ferroptosis-related lncRNA in kidney clear cell carcinoma, which could accurately predict the outcome of ccRCC [13]. Those signatures may provide a new perspective for cancer prognostic screening. Nevertheless, to our knowledge, very little information is available on lncRNA signature to explain the relationship between lncRNA and Ferroptosis-related genes (FRG) using the competing endogenous RNA (ceRNA) network. In the current research, we attempted to establish a lncRNA signature based on the ceRNA network and try to explain the relationship between lncRNA and ferroptosis-related genes as well as its specific biological characteristics. Therefore, we analyzed the RNAseq data and matched clinical profiles retrieved from the KIRC cohort of the TCGA (N = 583) database to comprehensively mine the prognostic role of ferroptosis-related lncRNA, and a 6-lncRNA signature was constructed in the training cohort and validated in the test cohorts. Furthermore, through a series of bioinformatics analyses, we identified the signature could be used as a prognostic predictor for ccRCC and was significantly correlated with immune-related functions, which suggests the signature could be a promising therapeutic alternative and prognostic biomarker for ccRCC patients.
3. METHODS
4. 1 Data Acquisition and Differentially Expressed Gene Analysis 112 Ferroptosis-related genes (annex Table S1) were obtaine from the FerrDb Database (http://www.zhounan.org/ferrdb/) concerning the following screening conditions: validated in human and protein-coding. Gene expression data and matched clinical profiles (including lncRNA, miRNA, and mRNA) of the KIRC cohort were downloaded from the TCGA database using the "GDCRNATools" package of R software [14], duplicate samples were removed, and only samples with sample_type of "PrimaryTumor" and "SolidTissueNormal" were retained. Then, the "GDCRNATools" package of R software was applied to analysis the differential expression of lncRNA, miRNA, and ferroptosis-related mRNA based on the conditions: method: DESeq2, Normalization: Voom, P = 0.05 and log| (fold change(FC))| =2.
5. 2 Construction of the ceRNA Network miRNA could regulate gene expression by targeting the 3’UTR of mRNA, while ceRNA network refers to that non-coding RNA (such as lncRNA) can competitively bind to miRNAs and reduce their inhibition on mRNA. The interactions of differentially expressed lncRNA and miRNA, as well as the interactions of miRNA and ferroptosis-related mRNA, were predicted utilizing TargetScan (www.targetscan.org), StarBase (https://starbase.sysu.edu.cn/) and RNAhybrid (bibiserv.cebitec.uni-bielefeld.de/rnahybrid/) databases. The intersection between the three databases was identified using the "vennR" package of R software. Then, the ceRNA network was established using the common miRNAs in the three databases that connect lncRNAs and mRNAs. Following the mechanism of the ceRNA network, we only retain the relationships as up-regulated lncRNA\down-regulated miRNA\up-regulated mRNA and down-regulated lncRNA\up-regulated miRNA\down-regulated mRNA, which were used to construct the ceRNA network, and then the results are imported into Cytoscape 3.7.1 for visualization.