Immune landscape of distinct subtypes in urothelial carcinoma based on immune gene profile

Abstract

Immune checkpoint blockade (ICB) has become a promising therapy for multiple cancers. However, only a small proportion of patients display a limited antitumor response. The present study aimed to classify distinct immune subtypes and investigate the tumor microenvironment (TME) of urothelial carcinoma, which may help to understand treatment failure and improve the immunotherapy response. RNA-seq data and clinical parameters were obtained from TCGA-BLCA, E-MTAB-4321, and IMVigor210 datasets. A consensus cluster method was used to distinguish different immune subtypes of patients. Infiltrating immune cells, TME signatures, immune checkpoints, and immunogenic cell death modulators were evaluated in distinct immune subtypes. Dimension reduction analysis was performed to visualize the immune status of urothelial carcinoma based on graph learning. Weighted gene co-expression network analysis (WGCNA) was performed to obtain hub genes to predict responses after immunotherapy. Patients with urothelial carcinoma were classified into four distinct immune subtypes (C1, C2, C3 and C4) with various types of molecular expression, immune cell infiltration, and clinical characteristics. Patients with the C3 immune subtype displayed abundant immune cell infiltrations in the tumor microenvironment and were typically identified as "hot" tumor phenotypes, whereas those with the C4 immune subtype with few immune cell infiltrations were identified as "cold" tumor phenotypes. The immune-related and metastasis-related signaling pathways were enriched in the C3 subtype compared to the C4 subtype. In addition, tumor mutation burden, inhibitory immune checkpoints, and immunogenic cell death modulators were highly expressed in the C3 subtype. Furthermore, patients with the C4 subtype had a better probability of overall survival than patients with the C3 subtype in TCGA-BLCA and E-MTAB-4321 cohorts. Patients with the C1 subtype had the best prognosis when undergoing anti-PD-L1 antibody treatment. Finally, the immune landscape of urothelial carcinoma showed the immune status in each patient, and TGFB3 was identified as a potential biomarker for the prediction of immunotherapy resistance after anti-PD-L1 monoclonal antibody treatment. The present study provided a bioinformatics basis for understanding the immune landscape of the tumor microenvironment of urothelial carcinoma.

Keywords: biomarker; immune subtypes; immunotherapy; tumor microenvironment; urothelial carcinoma.

Figure 1

Robust classification of distinct immune subtypes of bladder urothelial carcinoma. (A) Consensus clustering analysis of bladder urothelial carcinoma patients. (B) Cumulative distribution function (CDF) curve of the training cohort. (C) CDF delta area curve of the training cohort. (D) Distribution of immune subtypes according to sex, race, and tumor stage. (E) Kaplan plot of overall survival in TCGA-BLCA cohort. (F) Kaplan plot of progression-free survival in the E-MTAB-4321 cohort. (G) Kaplan plot of overall survival in the IMVigor210 cohort. (H) The overlap between our four immune subtypes and Thorsson’s six subtypes.

Figure 2

Landscape of tumor mutational burden (TMB) in four different immune subtypes. (A) TMB in four different immune subtypes of bladder urothelial carcinoma. (B) Mutation status of the top 10 immune genes with genomic alterations in the four different immune subtypes of bladder urothelial carcinoma.

Figure 3

Expression pattern of immune molecules in the TME of bladder urothelial carcinoma according to immune subtypes. (A-C) Differential expression of immune checkpoints in the four immune subtypes in TCGA-BLCA cohort (A), E-MTAB-4321 cohort (B), and IMVigor210 cohort (C). (D-F) ICD expression in the four immune subtypes in TCGA-BLCA cohort (D), E-MTAB-4321 cohort (E), and IMVigor210 cohort (F). ns, p ≥ 0.05; *p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001.

Figure 4

Distributions of tumor-infiltrating immune cells and molecular signatures of immune subtypes. (A-F) Heatmap and correlation analysis between immune subtypes and tumor-infiltrating cells in TCGA-BLCA (A, B), E-MTAB-4321 (C, D), and IMvigor210 (E, F). (G) Distribution of 29 immune cells in distinct immune subtypes. (H) Scores of 29 molecular signatures in distinct immune subtypes. ns, **p < 0.01; ***p < 0.001; and ****p < 0.0001.

Figure 5

Enriched KEGG signaling pathways in the C3 and C4 immune subtypes using the GSEA method. NES, normalized enrichment score; NOM P-val, nominal p value; KEGG, Kyoto Encyclopedia of Genes and Genomes; and GSEA, Gene Set Enrichment Analysis.

Figure 6

Immune gene-related landscape of bladder urothelial carcinoma. (A) Distribution of patients after reduced dimension. (B) Correlations between principal component analysis and immune cells. (C) Distribution of patients representing the status of each immune subtype. (D–G) Stratification analysis of immune subtypes. (D) Prognostic value and immune cell components in different subgroups of the C1 subtype. (E) Prognostic value and immune cell components in different subgroups of the C2 subtype. (F) Prognostic value and immune cell components in different subgroups of the C3 subtype. (G) Prognostic value and immune cell components in different subgroups of the C4 subtype. , ns, p ≥ 0.05; *p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001.

Figure 7

Exploration of coexpression modules of immune-related genes. (A) Clustering of patients with bladder urothelial carcinoma based on the expression profile of immune-related genes. (B) Scale-free topology model for the identification of multiple soft thresholds. (C) Mean connectivity for multiple soft thresholds. (D) Cluster dendrogram and module colors. (E) Module eigengenes of seven modules in the immune subtypes of bladder urothelial carcinoma ns, p ≥ 0.05; and ****p < 0.0001.

Figure 8

Identification of hub genes of bladder urothelial carcinoma based on immune genes. (A) Forest plot of univariate Cox analysis of seven modules of bladder urothelial carcinoma. (B) The degree of correlation between different modules and survival information is shown. (C) Prognostic value of the black modules with the median as a cutoff. (D) Prognostic value of the pink modules with the median as a cutoff. (E) Prognostic value of the red modules with the median as a cutoff. (F) Dot plot of the top 10 biological processes in terms of the black module. Correlation between the black module and principal component 1. (G) Dot plot of the top 10 biological processes in terms of the pink module. Correlation between the pink module and principal component 1. (H) Dot plot of the top 10 biological processes in terms of the red module. Correlation between the red module and principal component 1.