Constructing an immune- and ferroptosis-related lncRNA signature to predict the immune landscape of human bladder cancer

Abstract

Background: LncRNAs play a variety of roles in the tumor microenvironment and cancer immune responses. Determining the significance of bladder cancer (BLCA)-related genes to predict the prognostic and therapeutic response of BLCA is important.

Methods: IrlncRNA/ frlncRNA pairs were determined using univariate analysis. The signature was constructed based on this pairs. Finally, analysis and internal validation were performed from several aspects.

Results: We identified 60 immune- and ferroptosis-related lncRNA pairs, among which 12 were included in the Cox proportional hazards model. Patients in low-risk group survived for significantly longer. Survival and riskScore analyses showed that the low-risk group had a significantly better clinical outcome. ROC curve analysis showed that AUC of OS values were more than 0.75 in the training set and the whole cohort. As assessed using Cox analysis, the riskScore was an independent prognostic predictor in the training, testing set and the whole cohort. The areas under the multi-index ROC in the training set, the testing set, and the whole cohort were 0.777, 0.692, and 0.748, respectively. High-risk group was positively associated with most of tumor-infiltrating immune cells. High-risk Scores correlated positively with high expression of CD274, but not with PD-1. Low riskScores correlated positively with high expression levels of the genes ERBB2 and nectin-4. High-risk Score was associated with a lower IC50 value for Docetaxel, cisplatin, and Pazopanib, while there was an opposite result for metformin.

Conclusions: The signature constructed by pairing irlncRNAs and frlncRNAs showed a notable clinical predictive value.

Keywords: bladder cancer; ferroptosis; immune infiltration; immunity; lncRNA; signature.

FIGURE 1

Workflow of the present study

FIGURE 2

Identification of differentially expressed immune‐related lncRNAs (DEirlncRNAs) and ferroptosisrelated lncRNAs (DEfrlncRNAs) from TCGA data and Ensembl‐based annotation. (A) DEirlncRNAs shown on a volcano plot (B) DEfrlncRNAs shown on a volcano plot. (C) Intersection of the two sets of DElncRNAs

FIGURE 3

Establishment of the Risk Assessment Model using DE‐IFRL Pairs (A, B) The 60 pairs of prognostic lncRNAs were used to construct a LASSO Cox regression model, and the partial likelihood deviance incorporating 10‐fold cross‐validation was used to derive the tuning parameter (λ). In the plot, the vertical black line indicates an optimal log λ. (C, D) Results of (C) univariate and (D) multivariate Cox regression analyses of lncRNA pairs involved in the model

FIGURE 4

Validation of the Risk Assessment Model (A, B, C) Kaplan–Meier tests in training set (A), the testing set (B), and the whole cohort (C). (D, E, F) time‐dependent ROC analysis of risk scores based on 1‐, 3‐, and 5‐year OS in the raining set (D), the testing set (E), and the whole cohort (F). (G‐L) Risk scores of each case and Survival outcome of each case in the training set (G, J), testing set (H, K), and the whole cohort (I, L)

FIGURE 5

Use of the risk assessment model for clinical evaluation. (A‐F) Results of Univariate Cox and Multivariate Cox analysis showing the relationship of the risk score and clinical variables including age, sex, and TNM stage to overall survival (OS) in the training set (A, D), the testing set (B, E), and the whole cohort (C, F); (G‐I) multi‐index ROC curve analysis of the signature demonstrated that the areas under the curves in the training set, the testing set, and the whole cohort were 0.777, 0.692, and 0.748, respectively

FIGURE 6

Use of the risk assessment model to estimate tumor‐infiltrating cells, immunosuppressed molecules, and ADC targets. (A) Spearman correlation analysis showing that tumor‐infiltrating immune cells such as neutrophils, monocytes, fibroblasts, and macrophages, were associated positively with patients in the high‐risk group, whereas these patients were associated negatively associated with CD4+ T cells and fibroblasts. (B, C) The upregulated level of CD274 correlated positively with high‐risk scores (C), whereas the expression level of PDCD1 was not different among the groups (B). (D, E) Upregulated level of ERBB2 (D) and nectin‐4 (E) correlated positively with low‐risk scores. *p < 0.05; **p < 0.01; ***p < 0.001

FIGURE 7

The model could function to predict chemosensitivity. High‐risk scores were associated with lower IC50 values for targeted drugs (e.g., Pazopanib) (C) and chemotherapeutics (e.g., cisplatin and doxorubicin) (A, B), but were associated with higher IC50 scores for Metformin (D)