Single-Cell Transcriptome Comparison of Bladder Cancer Reveals Its Ecosystem

Abstract

Bladder carcinoma (BLCA) is a highly heterogeneous disease, and the underlying biological behavior is still poorly understood. Here, single-cell RNA sequencing was performed on four clinical samples of different grades from three patients, and 26,792 cell transcriptomes were obtained revealing different tumor ecosystems. We found that N-glycan biosynthesis pathway was activated in high-grade tumor, but TNF-related pathway was activated in cystitis glandularis. The tumor microenvironment (TME) of different samples showed great heterogeneity. Notably, cystitis glandularis was dominated by T cells, low-grade and high-grade tumors by macrophages, while TME in patient with high-grade relapse by stromal cells. Our research provides single-cell transcriptome profiles of cystitis glandularis and BLCA in different clinical states, and the biological program revealed by single-cell data can be used as biomarkers related to clinical prognosis in independent cohorts.

Keywords: N-glycan biosynthesis; bladder cancer; different grades; microenvironment; single-cell sequencing.

Figure 1

Overview of all cells in bladder samples. (A) The experimental process of single-cell RNA sequencing workflow. (B) UMAP plot of the general cell cluster. One point represents one cell, and different colors represent different cell types. (C) Dot plot of marker gene expression in different cell types. The size of the dot indicates the percentage of marker gene expression, and the color of the dot indicates the average expression of marker genes. (D) CNV score inferred by infercnv in different epithelial cell clusters and endothelium and fibroblasts. Clusters 1 and 2 with the highest scores were identified as malignant epithelium. ^*** p < 0.001 using two-sided unpaired Wilcoxon rank sum test.

Figure 2

Analysis of differential gene expression between malignant epitheliums in different clinical states. (A) UMAP plot of 10,339 malignant epithelial cells. Color indicates different cell clusters. (B) Scatter plot showed that a small number of genes were upregulated in CG. (C) Violin plots showing the expression level of the GC signature (average expression of CG upregulated genes) in different clinical states. ^*** p < 0.001 using two-sided unpaired Wilcoxon rank sum test. (D) Bar plots showing the pathway differentially activated between LG and HG. The color indicates significance p < 0.05, and the x-axis is the enrichment fraction. (E) Violin plots showing the expression level of N-glycan signature (the average expression of genes enriched into this pathway) in different clinical states. ^*** p < 0.001 using two-sided unpaired Wilcoxon rank sum test. (F) Violin plots showing the expression level of CD63 in epithelial cells of different clinical states. ^*** p < 0.001 using two-sided unpaired Wilcoxon rank sum test. (G) Boxplots showing the distribution of N-glycan signature (the average expression of genes enriched into this pathway) in TCGA BLCA cohort pathological stages. ^*** p < 0.001 using two-sided unpaired Wilcoxon rank sum test. (H) Kaplan-Meier analysis of N-glycan signature with overall patient survival in TCGA dataset. p = 0.011 using log-rank test. (I, J) Therapeutic response to inhibition of N-glycan pathway activity in T24 (I) and MBT2 (J) cells. ^* p < 0.05, ^** p < 0.01, ^*** p < 0.001 using two-sided unpaired Student’s t-test. T24 and MBT2 cells were treated with NGI-1 or cisplatin (DDP) or combined DDP and NGI-1. Error bars indicate addition and subtraction of standard deviation.

Figure 3

Changes of cell composition in microenvironment of different bladder samples. (A) UMAP plot of immune cells and stromal cells, showing different cell types. (B) Dot plot of marker gene expression in different cell types. (C) Sample specificity of each immune or stromal cell type. (D) Changes in the proportion of each cell type in samples with different clinical states. The error bar represents the 95% confidence interval of the cell proportion. (E) Immunofluorescence images showed changes in cellular components of CG, LG, and HG. CD8 and CD3 are T-cell markers; CD68 and CD14 are macrophage markers.

Figure 4

Enrichment of immune memory T-cell population in high-grade recurrent bladder cancer. (A) UMAP plot of all T cells. Color indicates T-cell subtype. (B) Violin plots showing the expression of marker genes in T-cell subtypes. (C) Changes in the proportion of each T-cell subgroup in samples with different clinical states. The error bar represents the 95% confidence interval of the cell proportion. (D) T-cell pseudotime development trajectory. Color indicates T-cell cluster.

Figure 5

Immunosuppressive macrophages infiltrate in tumor microenvironment. (A) UMAP plot of macrophages; color indicates cell population subtypes. (B) Correlation between proinflammatory (average expression of M1 signatures) and reparative (average expression of M2 signatures) macrophages; each point represents one cell. Two-sided p-value was calculated for Pearson’s correlation coefficient. (C) Changes in the proportion of each macrophage subgroup in samples with different clinical states. The error bar represents the 95% confidence interval of the cell proportion. (D) Heatmap of differential activation pathways in macrophage clusters. (E) Violin plots showing the expression of marker genes in macrophage subtypes.

Figure 6

Endothelium and fibroblasts enriched in high-grade relapse cancer. (A) UMAP plot of endothelial cells; color indicates cell population subtypes. (B) Violin plots showing the expression level of marker genes among endothelial cells. (C) Bar plot showing the most enrichment GO terms for E0 cluster. (D) UMAP plot of fibroblasts; color indicates cell population subtypes. (E) Bar plot showing the most enrichment GO terms for F1 cluster. (F) Violin plots showing the expression level of marker genes among fibroblasts.