Identification of a Prognostic Signature Associated With the Homeobox Gene Family for Bladder Cancer

Abstract

Background: Bladder cancer (BLCA) is a common malignant tumor of the genitourinary system, and there is a lack of specific, reliable, and non-invasive tumor biomarker tests for diagnosis and prognosis evaluation. Homeobox genes play a vital role in BLCA tumorigenesis and development, but few studies have focused on the prognostic value of homeobox genes in BLCA. In this study, we aim to develop a prognostic signature associated with the homeobox gene family for BLCA. Methods: The RNA sequencing data, clinical data, and probe annotation files of BLCA patients were downloaded from the Gene Expression Omnibus database and the University of California, Santa Cruz (UCSC), Xena Browser. First, differentially expressed homeobox gene screening between tumor and normal samples was performed using the "limma" and robust rank aggregation (RRA) methods. The mutation data were obtained with the "TCGAmutation" package and visualized with the "maftools" package. Kaplan-Meier curves were plotted with the "survminer" package. Then, a signature was constructed by logistic regression analysis. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were performed using "clusterProfiler." Furthermore, the infiltration level of each immune cell type was estimated using the single-sample gene set enrichment analysis (ssGSEA) algorithm. Finally, the performance of the signature was evaluated by receiver-operating characteristic (ROC) curve and calibration curve analyses. Results: Six genes were selected to construct this prognostic model: TSHZ3, ZFHX4, ZEB2, MEIS1, ISL1, and HOXC4. We divided the BLCA cohort into high- and low-risk groups based on the median risk score calculated with the novel signature. The overall survival (OS) rate of the high-risk group was significantly lower than that of the low-risk group. The infiltration levels of almost all immune cells were significantly higher in the high-risk group than in the low-risk group. The average risk score for the group that responded to immunotherapy was significantly lower than that of the group that did not. Conclusion: We constructed a risk prediction signature with six homeobox genes, which showed good accuracy and consistency in predicting the patient's prognosis and response to immunotherapy. Therefore, this signature can be a potential biomarker and treatment target for BLCA patients.

Keywords: biomarkers; bladder cancer; homeobox gene family; immunotherapy; prognostic signature.

FIGURE 1

Flow chart showing the design of the study, with GSE7476 (N = 3; T = 9), GSE13507 (N = 68; T = 188), GSE37815 (N = 6; T = 18), GSE65635 (N = 4; T = 8), and TCGA (N = 18; T = 406) datasets.

FIGURE 2

Identification of DEHGs for BLCA, analysis of the mutation landscape, and correlation analysis of the six DEHGs. (A) DEHGs for BLCA in the GSE7476 dataset. (B) DEHGs for BLCA in the GSE13507 dataset. (C) DEHGs for BLCA in the GSE37815 dataset. (D) DEHGs for BLCA in the GSE65635 dataset. (E) DEHGs for BLCA in the TCGA dataset. (F) LogFC values of each gene in different datasets (GSE7476, GSE13507, GSE37815, GSE65635, and TCGA). (G) Analysis of the correlations among the six DEHGs. (H) Mutation landscape of the six DEHGs in TCGA BLCA patients.

FIGURE 3

Gene expression profile of these six genes in TCGA cohort. (A) Differences in the expression of the six genes between BLCA tissues and normal tissues. (B) Differences in the expression of the six genes in different T stages. (C) Differences in the expression of the six genes in different N stages. (D) Differences in the expression of the six genes in different M stages. (E) Differences in the expression of the six genes in different clinical stages.

FIGURE 4

A high risk score was associated with a poor clinical outcome. The BLCA cohort was divided into two groups based on the median estimated score, and the two groups were then compared. The ranked dot plot indicates the risk score distribution in the training dataset (A) and testing dataset (B). Scatter plot presenting the patients’ overall survival status in the training dataset (C) and testing dataset (D). Heat map with the gene expression profiles of these six genes in the training dataset (E) and testing dataset (F). Kaplan–Meier curve analysis of the signature in the training set (G), testing set (H), and entire dataset (I).

FIGURE 5

Functional enrichment between the high-risk and low-risk groups. (A) GO function annotation. (B) KEGG pathway analysis.

FIGURE 6

Correlation analysis of the signature and immune characteristics. (A) Correlations between the signature and each immune cell infiltration score. (B) Correlations between each signature gene and each immune cell infiltration score. (C) Correlations between the expression level of immune checkpoints and the six signature genes. (D) Prediction of the difference in risk scores between immunotherapy responders and non-responders.

FIGURE 7

Evaluation of the signature model. ROC curves for predicting one-, three-, and five-year survival in the training set (A), testing set (B), and entire dataset (C). External validation of the signature model using the GEO BLCA cohorts GSE13507 (D), GSE19423 (E), and GSE37815 (F).