A Ferroptosis-Related Long Non-Coding RNA Signature Predicts Prognosis and Immune Microenvironment for Lung Squamous Cell Carcinoma

Background: Ferroptosis-related lncRNAs (FerLncRNAs) were developed to play a signicant role in cancer treatment and prognosis. However, the relationship between FerLncRNAs and Lung squamous cell carcinoma (LUSC) remains unclear. Method: Based on ferroptosis-related differentially expressed lncRNAs in LUSC, we established a prognostic 8-lncRNA signature. Results: 8 Ferroptosis-related lncRNAs (LUCAT1, AL161431.1, AL122125.1, AC104248.1, AC016924.1, MIR3945HG, C10orf55 and AP006545.2) had prognostic value for LUSC by multivariate COX analysis (P<0.05), and possessed signicant association with patient outcomes. Kaplan–Meier curves showed an obvious difference in OS that the high-risk group patients exhibited poorer survival than the low-risk group patients. The clinical receiver operating characteristic (ROC) curve and decision curve analysis (DCA) revealed that the ferroptosis-related lncRNAs prognostic signature can (FerRLSig) emerged more outstanding performance than clinical features in predicting the prognosis of LUSC. GSEA revealed that the majority of the novel Ferroptosis-related lncRNAs signature-regulated immune responses and the immune system processes were enriched in the high-risk group. 34 immune checkpoints (ICs) were detected signicantly different expression between high-risk and low-risk groups. Conclusion: A novel FerRLSig based on 8 Ferroptosis-related lncRNAs provided important prognostic value for LUSC patients and developed new insights about ferroptosis-related immunotherapy targets in clinic.


Introduction
Despite great efforts made in recent years, the incidence of Lung squamous cell carcinoma (LUSC), a histological subtype of non-small cell lung carcinoma (NSCLCs) (Travis, 2002;Travis et al., 2015;Youlden, Cramb, & Baade, 2008), remains increasing and is a leading cause of cancer death worldwide (Stinchcombe, 2014). A poor 5-year overall survival (OS) rate of LUSC by current treatments further prompts that novel and e cient clinical managements are urgent to be established (Garon et al., 2019;Miller et al., 2012). Therefore, it is strongly required that de nite diagnostic biomarkers, new therapeutic targets, and favorable prognostic signatures for patients with LUSC be found.
Iron not only is a vital microelement in normal cell physiological growth and development, but also plays a vital role in tumor cell of LUSCs ( Long non-coding RNAs (lncRNAs) are transcripts composed by nucleotides ranging in length from 200bp to 10kbp (D. Wang et al., 2020). Although limited in protein-encode function, lncRNAs have been found to perform various functions in a wide variety of important biological and pathological processes, such as chronic in ammatory response, cell migration and invasion (Bai et al., 2020;Qian et al., 2018). Notably, previous studies have shown lncRNAs play as the essential regulators in ferroptosis and iron metabolism in lung cancers. For instance, LINC0036 exerts anti-ferroptosis effect in lung carcinoma by acting as an endogenous sponge of microRNA 6852 to regulate the expression of cystathionine-β-synthase (CBS), a surrogate marker of ferroptosis (M. Wang et al., 2019).
Metallothionein 1D pseudogene (MT1DP) sensitizes NSCLCs toward ferroptosis via elevating lipid reactive oxygen species (ROS) (Gai et al., 2020). Silencing nuclear enriched transcript 1 (NEAT1) aggravates erastin-induced ferroptosis through decreasing levels of ACSL4, SLC7A11, and GPX4 in NSCLCs (H. . Those have indicated ferroptosis-related lncRNAs (FerLncRNAs) are closely related to pathological outcome of NSCLCs, but whether to regulate LUSC and the speci c mechanism maintains unclear. Therefore, identifying FerLncRNAs in LUSC is a key link in creating a prognostic signature based on FerLncRNAs, which may be contribute to a theoretical basis for novel strategies to treat patients with LUSC.
Immune checkpoint molecules, the receptors expressed on immune cell, inhibit immune response by triggering immune cells into a state of "exhaustion" In this study, a prognosis signature of FerLncRNAs was constructed to evaluate actual scienti c relevance and applicability to LUSC and identi ed the roles in tumor immunity, which contributes to clarifying the relationship between FerLncRNAs and LUSC, providing potential diagnostic biomarker and therapeutic targets to LUSC in clinic.

Data Source
The RNA sequencing (RNA-seq) data (49 normal and 502 tumors) and corresponding clinical information of patients with LUSC were from TCGA-LUSC. Table.1 showed the clinical features of the patients. What's more, some patients were excluded with incomplete clinical information. Eventually, a total of 102 patients were excluded from the study ( Table 1). The ferroptosis-relates genes were extracted from FerrDb. Finally, 246 ferroptosis-related genes were identi ed (Supplementary Table S1). Pearson correlation was used to assess the relationship between the ferroptosis-related lncRNAs and LUSC (|R2|>0.4 at P<0.001). The criteria was set for the ferroptosis-related lncRNAs to FDR<0.05 and |log 2 FC|≥1. First, we made expression and enrichment analysis of both upregulated and downregulated ferroptosis-related differentially expressed genes (DEGs). Then Gene ontology (GO) was used to evaluate the biological pathways associated with the DEGs. Through the use of ggplot2 package of R software, further function regulated by ferroptosis-related DEGs was analyzed based on Kyoto Encyclopedia of Genes and Genomes (KEGG) data.

Development and Validation of a Ferroptosis-Related lncRNAs Prognostic Signature
In order to establish the FerLncRNAs prognostic signature (FerRLSig), Lasso-penalized Cox regression and Univariate Cox regression analysis were performed.
The computational formula was as follows: Risk Score=sum (each lncRNA's expression×corresponding regression coe cient). Next, we calculated the scores of each patient in the data collected from TCGA. Patients were classi ed into low-risk and high-risk groups on the basis of the median risk score. By Kaplan-Meier analysis, the OS of high-risk and low-risk groups were compared. Time-dependent receiver operating characteristic curves (ROC) and risk survival status together proved the sensitivity of our model in predicting the prognosis of LUSC. And decision curve analysis (DCA) veri ed the speci city of FerRLSig.

Gene set enrichment analyses and the predictive nomogram
Gene set enrichment analysis (GSEA) was used to study the high-risk and low-risk groups, and visualized their pathways closely related to immunity and tumorigenesis. In order to explore the clinical signi cance of the FerRLSig, a nomogram integrating prognostic signals was constructed to predict the 1-, 3-and 5-year OS of HNSCC patients.

Immune correlation analysis
The difference of immune response, between high-risk and low-risk groups, under distinct algorithms (TIMER, CIBERSORT, CIBERSORT−ABS, QUANTISEQ, MCPCOUNTER, XCELL, EPIC) was revealed by heat map. Furthermore, we used ssGSEA to quantify the tumor-in ltrating immune cell subgroups and immune function between the two groups. Potential immune checkpoints between high-risk and low-risk groups were listed.

Statistical analysis
All statistical analyses were completed by using Bioconductor packages in R software, version 4.1.0., Cytoscape and GSEA 4.1.0. Benjamin Hochberg method was used to identify different expressed FerLncRNAs. "Gsva" (R-package) was utilized in ssGSEA. Logistic regression analysis was used to evaluate the relationship between FerLncRNAs and clinicopathological manifestations. Statistical tests were bilateral with P value≤0.05 indicated statistically signi cant differences.

The expression and enrichment analysis of ferroptosis-related DEGs in LUSC
The 102 ferroptosis-related differentially expressed genes (FerDEGs) were rst discovered in total from The Cancer Genome Atlas (TCGA) database which contains clinical information of patients with lung squamous cell carcinoma (LUSC), including 35 downregulated genes and 67 upregulated genes (Supplementary Table S2). The GO enrichment signi cantly revealed that FerDEGs were obviously enriched in many immune-related biological processes (BP) which generally appeared in the apical part of cell plasma membrane, such as response to oxidative stress, cellular response to chemical stress, cellular response to oxidative stress and reactive oxygen species metabolic process (Fig. 1A). The overexpressed FerDEGs was mainly involved in uid shear stress and atherosclerosis, HIF−1 signaling pathway, Bladder cancer and AGE−RAGE signaling pathway in diabetic complications through the KEGG pathway analysis (Fig. 1B). . The visualized coexpression network between 8 FerLncRNAs and ferroptosis-related genes constructed by Cytoscape was shown in (Fig. 1.D). Pro ting from this model, patients with LUSC were classi ed into low-risk and high-risk groups on the basis of the corresponding median risk score which was calculated by the favorable prognostic signature (0.1052×ExpressionLUCAT1+0.004×ExpressionAL161431.1-0.319513084×ExpressionAL122125.1+0.146203641×ExpressioAC104248.1+0.592710446×ExpressionAC016924.1+0.434557078×ExpressionMIR3945HG+0.2 0.170496616×ExpressionAP006545.2).

Multivariate examination of FerLncRNAs prognostic Signature
Kaplan-Meier curves showed an obvious difference in OS that the high-risk group patients exhibited poorer survival than the low-risk group patients (P<0.001, Fig. 2.A), suggesting that the building-related signature of the 8-FerLncRNAs effectively predicts survival. Time-dependent receiver operating characteristic curves exhibited the AUC (Area Under the Curve) at 1, 3, 5-years were 0.658, 0.693, 0.687 respectively, showing predictive value of the novel FerRLSig (Fig. 2. B). Risk survival status was shown in (Fig. 2. C). The death risk augmented and survival duration diminished following the increased risk score, indicating risk score was inversely proportional to the survival rate of patients with LUSC. The heatmap also demonstrated that the 8-FerLncRNAs was positively correlated with the prognostic signature (Fig. 2. D). Moreover,the clinical receiver operating characteristic (ROC) curve and decision curve analysis (DCA) revealed that the FerRLSig emerged more outstanding performance than clinical features in predicting the prognosis of LUSC (Fig. 2.E,F). In the univariate and multivariate Cox independent prognostic analysis, risk scores (P<0.001) revealed the risk signature was prognostic factor for predicting the OS of LUSC patients (Fig. 3.A). To investigate whether the FerRLSig participated in the development of LUSC, the heatmap analyzed the association between 8 FerLncRNAs and clinicopathological manifestations, which showed there were signi cant differences between high-risk and low-risk groups (Fig. 3.B). The clinical nomogram based on prognostic signature combined with the clinical factors also con rmed a FerRLSig might be applied in clinic as a novel signature for prognostic diagnosis and therapeutic management of LUSC patients (Fig. 3.C). Those results veri ed FerRLSig was an independent prognostic factor characterized by stability and accuracy for OS in patients with LUSC.

Immunity-related status and functions responsible for different OS in patients with LUSC
To assess the correlations between the LUSC prognosis and immune system function, gene set enrichment analyses (GSEA) were applied to compare between the high-risk and the low-risk, revealing that the majority of the novel FerRLSig-regulated immune responses and the immune system processes were enriched in the high-risk group, such as hematopoietic cell lineage, cell adhesion molecules cams, leukocyte transendothelial migration, antigen processing and presentation, natural killer cell mediated cytotoxicity, intestinal immune network for IgA production, chemokine signaling pathway, JAK stat signaling pathway, and Toll like receptor signaling pathway (P < 0.05, FDR < 0.05; Fig. 4).
Based on TIMER,CIBERSORT CIBERSORT−ABS QUANTISEQ MCPCOUNTER XCELL EPIC the immune-related status heatmap con rmed the differences of T cell CD8+_TIMER, Myeloid dendritic cell_TIMER, Monocyte_XCELL, Monocyte_MCPCOUNTER and Mast cell resting_CIBERSORT−ABS between high-risk and low-risk group (Fig. 5.A). The correlation analysis of single-sample gene set enrichment analysis (ssGSEA) based on TCGA-HNSCC data in immune cell subsets and functions showed that there were signi cant differences between two risk groups except for APC_co_inhibition (Fig. 5.B). In addition, the difference was further found in the expression of immune checkpoints, such as CD86, CD48, HAVCR2 and LAIR1 (Fig. 5.C), suggesting that FerRLSig may involve in immune regulation of tumor microenvironment via affecting immune cells. m6A related mRNA expression was signi cant including FTO, YTHDC2, METTL3, YTHDC1, HNRNPC and YTHDF1 by comparing high and low risk groups (Fig. 5.D), which means genes that regulate methylation may be involved in the regulation of the immune system.

Discussion
The incidence and mortality rate of LUSC show an increasing trend internationally, causing serious harm to human health (Torre et al., 2015). A good high-risk and low-risk groups were compared. And the analysis indicated most of ICIs were high-expressed in high-risk groups. In addition, the GSEA analyses of the high-risk and low-risk groups exhibited that the immune response and immune system processes were signi cantly enriched in the high-risk group, and thus, the roles of FerLncRNAs had been more con rmed in the regulation of tumor immune in ltration.
Currently, few studies have explored the relevance between N6-methyladenosine (m6A) and ferroptosis. From the m6A analysis, YTHDC2 and ALKBH5 were rich in low-risk group. YTHDC2, an m6A methylase, is regarded as a powerful endogenous ferroptosis inducer via suppressing SLC3A2, which is essential for anti-oxidation as a subunit of system XC − in cell (Ma et al., 2021). Suppression of ALKBH5 is a vital process to promote sorafenib(SF)-induced ferroptosis in HCC (Z. Liu et al., 2020). It may speculate that FerlncRNAs mediate methylation in LUSC to facilitate ferroptosis.
The signature of 8 FerlncRNAs has an independent prognostic value for patient with LUSC and exhibits a promising prospect of ferroptosis and immunity in tumor therapy. Therefore, this novel FerRLSig is signi cant for improving the clinical prediction of prognosis and management among LUSC patients.