Multi-Omics Integration Identifies a CDH3-Associated Malignant Epithelial State and Immunosuppressive Niche to Predict Prognosis in Thymic Epithelial Tumors.

Thymic epithelial tumors (TETs) are rare and heterogeneous malignancies whose aggressive epithelial states and microenvironmental organization remain poorly defined. Here, we integrated single-cell RNA sequencing, spatial transcriptomics, multiplex immunofluorescence, bulk transcriptomics, functional assays, xenograft validation, and computational pathology to characterize malignant epithelial heterogeneity in TETs. We identified a CDH3-associated malignant epithelial state located at the origin of malignant-state trajectories and enriched for stem-like and EMT-related features. Spatial transcriptomics and multiplex immunofluorescence showed that CDH3+ tumor cells preferentially localized within an M2 macrophage-rich immunosuppressive niche, while cell-cell communication analyses nominated CCN2-LRP1 as a candidate epithelial-myeloid crosstalk axis. A 68-gene CDH3-associated signature stratified TCGA-THYM into biologically distinct subgroups with differences in survival, histology, genomic instability, and immune contexture. Patient-derived thymic carcinoma organoids showed elevated CDH3 expression, and CDH3 silencing suppressed thymic carcinoma cell proliferation, migration, invasion, EMT/PI3K-Akt-related signaling, and macrophage-associated crosstalk. Candidate inhibitors showed antitumor activity in xenograft models. We also established a deep learning pathology model that captured CDH3-associated morphology from routine H&E slides and predicted patient outcome. Together, these findings define CDH3 as a biomarker and therapeutic target linking malignant epithelial plasticity to immunosuppressive niche formation and adverse clinical behavior in TETs.