Fig. 5: Immune infiltration in mUC.

Different bioinformatics methods were applied to deconvolute immune cell populations from bulk transcriptome data using immune cell–specific signatures (CIBERSORT, XCELL, MCPcounter, Epic) and microenvironment composition signatures (ESTIMATE). N = 98 samples. Source data are provided as a Source Data file. Global immune infiltration (estimated from transcriptomic profiles using the ESTIMATE algorithm) was associated to TMB (a), prior FGFR inhibitor therapy (N = 88 naive and N = 10 treated patients) (b), prior immunotherapy (N = 80 naive and N = 18 treated patients) (c), and biopsy sampling site (N = 20 lung; N = 34 lymph nodes; N = 25 liver and N = 19 other) (d). Whiskers box plot represent interquartiles with 1.5x IQR, and outliers dots. e Subtype-specific immune enrichment in eMIBC vs mUC. The level of immune and stromal infiltration (estimated by MCPcounter in upper panel and ESTIMATE in lower panel) was compared in one subtype versus the others, or in FGFR3-mutated tumors vs wild-type, in either mUC after removing lymph-node biopsies (left panel) or in eMIBC from the TCGA (right panel). f Forest plot of univariate survival models in mUC of stromal and immune estimates, stratified by presence of visceral metastasis and lymph node biopsy site. The squares represent HR with the confidence intervals in gray lines. Two-tailed Mann–Whitney test was applied g Forest plot of univariate survival models in eMIBC of stromal and immune estimates. The squares represent HR with the confidence intervals in gray lines. Two-tailed Mann–Whitney test was applied. TMB Tumor Mutational Burden, LN lymph node, MCP Microenvironment Cell Population, HR Hazard Ratio. N = 98 mUC versus N = 408 eMIBC from TCGA.