A novel Signature Constructed using Ferroptosis-Related lncRNA Pairs May Predict the Prognosis of Bladder Cancer Patients

Background:Bladder cancer is one of the most common malignant tumors of the urinary system, and its incidence has been increasing in recent years. Ferroptosis is a recently discovered type of cell death, and some studies have suggested that it is closely associated with tumors. It can promote tumor apoptosis and also promote tumor development. Moreover, it has been reported that a correlation exists between long non-coding RNAs (lncRNAs) pairs and tumors. Herein, we developed an lncRNA pair signature associated with ferroptosis to predict the prognosis of bladder cancer. Methods: We combined the bladder cancer transcriptome data from the Cancer Genome Atlas (TCGA) database to identify ferroptosis-related lncRNA (FRlncRNA) pairs. Using univariate and multivariate Cox analyses and LASSO regression analysis, we identied a FRlncRNA pair signature. We subsequently assessed the predictive prognostic value of this signature and validated the results. Results: The signature included 18 lncRNA pairs and was highly accurate for clinical prediction in patients with bladder cancer. Univariate and multivariate Cox analyses and stratied analysis indicated that the model was an independent prognostic factor. Additionally, we detected a positive correlation between this signature and the tumor immune microenvironment. Conclusion: The FRlncRNA pair signature has good prognostic and clinical predictive value in patients with bladder cancer. validation the DEFRlncRNA pair was performed using univariate and multivariate Cox analyses. The association between this signature and clinicopathological features was assessed using the chi-square test and Wilcoxon signed-rank test. The risk scores were analyzed for immune inltration using the Spearman correlation test by combining several accepted immune inltration methods (XCELL, TIMER, QUANTISEQ, MCPCOUNTER, EPIC, CIBERSORT − ABS, and CIBERSORT). The relationship between risk scores and genes involved in immune checkpoint inhibitors (ICIs) was assessed using the limma package of R. Risk groups were compared for their sensitivity to common chemotherapeutic drugs for bladder cancer by analyzing drug IC using the limma package and the pRRophetic package in R language. Statistical signicance was determined at p < 0.05; * indicates p < 0.01 and indicates p < 0.001.


Background
Bladder cancer is the most common malignancy of the urinary system and its mortality rate ranks rst among urinary system malignancies (1). Approximately 25% of bladder cancer cases are muscle-invasive bladder cancer (MIBC), and patients with MIBC usually have poor prognosis, with a ve-year survival rate of less than 10% (2,3). Although the prognosis of non-muscle-invasive bladder cancer (NMIBC) is better than that of MIBC, the recurrence rate of NMIBC is 50-70% after ve years (4). At present, the main treatments for bladder cancer are surgery, chemotherapy, and immunotherapy (5,6). Because of the poor survival rate of MIBC and the high recurrence rate of NMIBC, effective biomarkers are needed to predict bladder cancer prognosis.
Ferroptosis is a recently discovered mechanism of cell death and is characterized by excessive lipid peroxidation (7). The molecular mechanism between ferroptosis and tumors has not yet been clearly elucidated; however, tumor growth has been shown to be highly susceptible to ferroptosis(8). Long noncoding RNAs (lncRNAs) are > 200 nucleotides long(9, 10) and cannot encode proteins; however, they are involved in various biological functions (11). Recently, studies have shown lncRNAs may regulate ferroptosis and reduce tumor progression (12,13). Wang et al. (14) showed that the lncRNA, LINC00336, plays a role in inhibiting iron ptosis in lung cancer. It has also been suggested that a nuclear lncRNA called LINC00618 plays a role in promoting iron ptosis in leukemia (15). However, the role of ferroptosisrelated lncRNAs (FRlncRNAs) in bladder cancer remains unexplored to date.
Herein, we constructed a FRlncRNA pair signature characterized by the lack of speci c expression of lncRNAs in the samples using the iterative and odd and even methods proposed by Sun et al.(16) Subsequently, we performed a series of assessments and validations of the prognostic value of the FRlncRNA pair signature in patients with bladder cancer.

Identi cation of DEFRlncRNAs
The study work ow is illustrated in Fig. 1. The co-expression analysis of ferroptosis-related genes and lncRNAs is shown in Fig. 2A. Following differential analysis, we obtained 126 DEFRlncRNAs, which included 105 upregulated genes and 21 downregulated genes ( Fig. 2B and C).

Identi cation Of Prognostic Signature
Subsequently, univariate Cox, LASSO regression, and multivariate Cox regression analyses resulted in 18 DEFRlncRNA pairs being included in the prognostic signature ( Fig. 3A-C). The risk value was calculated for each sample (Fig. 4A) and cut-off points were selected for high and low risks. The ROC curves at 1, 3, and 5 years were plotted, and all three curves had high area under the curve (AUC; Fig. 4B). The AUC values of the 5-year ROC curve were compared with other clinical characteristics, and the 5-year ROC AUC values were far higher than those of the other clinical features (Fig. 4C). Subsequently, we performed survival difference analysis for the high-and low-risk groups. We found survival in the low-risk group was signi cantly greater than that of the high-risk group ( Fig. 5A-C).

Independent Prognostic Validation Of Defrlncrna Pairs Signature
We then performed univariate and multivariate Cox regression analyses of risk scores, age, sex, tumor grade, and tumor stage. Age, tumor stage, and risk score had independent prognostic values ( Fig. 6A and   B). Therefore, we conducted a strati ed analysis where samples of patients with bladder cancer into two age groups: young (age ≤ 65 years) and elderly (age > 65 years) and performed survival analysis of highand low-risk patients within each age group. The survival rate of low-risk patients was higher than that of high-risk patients in both the young and elderly groups (p < 0.001, Fig. 6C and D). When patients were divided by tumor stage into early (stages I-II) and late (stages III-IV) and survival analysis was performed, low-risk patients had improved survival compared to high-risk patients (p < 0.001, Fig. 6E and F).

Correlation Between Defrlncrna Pairs Signature And Clinicopathological Features
The correlation between DEFRlncRNA pair signature and clinicopathological features was compared using the chi-square test (Fig. 7A) and the Wilcoxon signed-rank test ( Fig. 7B-F). Both tests showed that the risk score was signi cantly correlated with age, grade, clinical stage, T stage, and N stage.
Tumor immune in ltration in DEFRlncRNA pair signatures and analysis of ICI-related gene expression in different risk scores Risk scores were then assessed by seven immune in ltration methods. Categorization in high-risk group was positively correlated with cancer-associated broblasts, macrophages, monocytes, neutrophils, NK cells, and CD8 + T cells, but negatively correlated with CD4 + naive T cells and B cell plasma (Fig. 8A).
We analyzed the association between several signi cant ICI-related genes and risk scores and found high risk correlated positively with the expression of PDL1 (

Correlation analysis between DEFRlncRNA pair signatures and sensitivity to chemotherapeutic drugs
We analyzed the association between commonly used chemotherapeutic agents and the risk scores calculated using DEFRlncRNA pair signature. High-risk score was negatively correlated sensitivity to cisplatin (

Discussion
Ferroptosis is an iron-dependent mechanism by which various pathways decrease cellular antioxidant capacity, leading to the accumulation of reactive oxygen species and subsequent cell death (7,17). Many studies have focused on inducing ferroptosis to treat cancer(18-20); however, it has also been suggested that ferroptosis may have both tumor-promoting and tumor-inhibiting actions (8, 21). Additionally, studies have suggested that lncRNAs play crucial roles in tumor development. Schmitt et al. (22) suggested that lncRNAs can promote the development of many tumor phenotypes through interactions with DNA and RNA. Elena et al. (23) showed that lncRNAs can affect the occurrence and development of urinary tumors and can be used as biomarkers for urinary tumors. Although many lncRNA signatures have been used as prognostic markers for bladder cancer patients (24)(25)(26), there is no signature of ferroptosis-related lncRNAs used to predict the prognosis of patients with bladder cancer. Therefore, we established a ferroptosis-related lncRNA signature and validated its ability to predict prognosis in patients with bladder cancer.
After performing numerous analyses on transcriptional and clinical data from TCGA we selected 18 lncRNA pairs to construct a signature for bladder cancer prognosis. Both KM survival analysis and ROC curve demonstrated that the signature successfully predicted the prognosis of patients with bladder cancer. Moreover, our calculated risk score based on this signature was positively associated with hightumor grade and stage. Our analysis found that age and tumor stage also had independent prognostic values. To further validate the risk score, we performed a strati ed analysis validation using age and tumor stage. The results also showed that the risk score has an independent prognostic value and can be used as an independent prognostic factor.
The 18 lncRNA pairs we obtained included some lncRNAs that have been associated with tumors, such as AC010186.3 (27), LINC01605(28), LINC00641 (29), and MAP3K14 − AS (30). Notably, LINC01605 and LINC00641 have been reported in bladder cancer. Qin et al. (31) suggested that LINC01605 promotes the proliferation and invasion of bladder cancer. In contrast, Li et al. (32) suggested that LINC00641 inhibits the progression of bladder cancer. In addition to well-known lncRNAs, we discovered novel lncRNAs, which we speculate could be new tumor biomarkers.
When we compared our risk score to immune in ltration, we found our risk score positively correlated with several cell populations, including CD8 + T cells. In addition, our calculated risk score correlated with the patient response to ICI, suggesting our model may predict responses to ICI. ICIs exert an anti-tumor effect mainly by reactivating immune cells and are most effective in in ammatory tumors with high CD8 + T cell in ltration and high PD-L1 expression (33). Our signature suggests that ICIs are effective for treating bladder cancer, in line with the National Comprehensive Cancer Network (NCCN) guidelines (34). When we analyzed ability of our signature to predict tumor sensitivity to commonly used chemotherapeutic agents, the results showed that chemosensitivity was lower in the high-risk group than in the low-risk group. Therefore, we believe that immunotherapy may achieve better results than chemotherapy in high-risk patients.
Our study also has some limitations. We only utilized TCGA database and did not validate our model with other external data sets. Further validation of our results would bene t from collecting additional samples and clinical trial data to verify our results. Our method offers some bene ts compared to previous studies. The advantage of our lncRNA pair signature is that it does not need to adopt the speci c expression level of each lncRNA in the sample. Additionally, it does not suffer from a batch effect, so the prognostic model constructed by this method is also more accurate than other signatures(16, 35).

Methods
Collection and processing of raw data Transcriptome and clinical data from bladder cancer patients were acquired from the Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/). ene transfer format les from Ensembl (http://asia.ensembl.org) were analyzed using Strawberry Perl (version 5.32.01) to distinguish mRNAs from lncRNAs. Then, 60 ferroptosis-related genes were obtained from the study by Liang et al.(36) Correlations between ferroptosis-related genes and lncRNAs were tested using the limma package in R (version 4.0.2) and signi cance was determined with correlation coe cient > 0.4 and p < 0.001. Resulting signi cant lncRNAs were de ned as FRlncRNAs. Differential analysis of the FRlncRNAs was performed using the limma package of R, and signi cant differences were determined by a false discovery rate (FDR) < 0.05 and log fold change (FC) > 1.5. Differentially-expressed FRlncRNAs (DEFRlncRNAs) were used for subsequent analysis.

Data analysis
Expression of DEFRlncRNA in each sample was compared in pairs and the value was de ned by A. In each DEFRlncRNA pair, if the rst lncRNA expression level was greater than that of the second lncRNA, A = 1; otherwise, A = 0. Univariate Cox analysis was subsequently used to determine prognostically-relevant DEFRlncRNA pairs. Then, prognostically-relevant DEFRlncRNA pairs were screened using LASSO regression analysis (iteration = 1000) and the DEFRlncRNA pair signature was constructed using multivariate Cox analysis. A risk score for each patient was calculated using the following equation: where C i represents the coe cient obtained after multivariate Cox analysis for the i th lncRNA pair and A i represents the expression value of the i th lncRNA pair.
A time-dependent receiver-operating characteristic (ROC) curve was drawn for 5 years and the maximum in ection point was considered as the cut-off value. If the risk value was greater than the cut-off value, the DEFRlncRNA pair was divided into the high-risk group; otherwise, it was divided into the low-risk group.
ROC curves were constructed using the survivalROC package in R to assess the DEFRlncRNA pair signature precision. Survival differences between high-and low-risk groups were assessed using Kaplan-Meier (KM) survival analysis. Independent prognostic validation of the DEFRlncRNA pair signature was performed using univariate and multivariate Cox analyses. The association between this signature and clinicopathological features was assessed using the chi-square test and Wilcoxon signed-rank test. The risk scores were analyzed for immune in ltration using the Spearman correlation test by combining several accepted immune in ltration methods (XCELL, TIMER, QUANTISEQ, MCPCOUNTER, EPIC, CIBERSORT − ABS, and CIBERSORT). The relationship between risk scores and genes involved in immune checkpoint inhibitors (ICIs) was assessed using the limma package of R. Risk groups were compared for their sensitivity to common chemotherapeutic drugs for bladder cancer by analyzing drug IC 50 using the limma package and the pRRophetic package in R language. Statistical signi cance was determined at p < 0.05; * indicates p < 0.01 and * indicates p < 0.001.

Conclusion
In conclusion, we constructed a novel ferroptosis-related lncRNA pair signature that can predict the prognosis of bladder cancer and guide immunotherapy and chemotherapy for bladder cancer patients.