Unraveling tumor microenvironment heterogeneity in malignant pleural mesothelioma identifies biologically distinct immune subtypes enabling prognosis determination

Background: Malignant pleural mesothelioma (MPM) is a rare and intractable disease exhibiting a remarkable intratumoral heterogeneity and dismal prognosis. Although immunotherapy has reshaped the therapeutic strategies for MPM, patients react with discrepant responsiveness.

Methods: Herein, we recruited 333 MPM patients from 5 various cohorts and developed an in-silico classification system using unsupervised Non-negative Matrix Factorization and Nearest Template Prediction algorithms. The genomic alterations, immune signatures, and patient outcomes were systemically analyzed across the external TCGA-MESO samples. Machine learning-based integrated methodology was applied to identify a gene classifier for clinical application.

Results: The gene expression profiling-based classification algorithm identified immune-related subtypes for MPMs. In comparison with the non-immune subtype, we validated the existence of abundant immunocytes in the immune subtype. Immune-suppressed MPMs were enriched with stroma fraction, myeloid components, and immunosuppressive tumor-associated macrophages (TAMs) as well exhibited increased TGF-β signature that informs worse clinical outcomes and reduced efficacy of anti-PD-1 treatment. The immune-activated MPMs harbored the highest lymphocyte infiltration, growing TCR and BCR diversity, and presented the pan-cancer immune phenotype of IFN-γ dominant, which confers these tumors with better drug response when undergoing immune checkpoint inhibitor (ICI) treatment. Genetically, BAP1 mutation was most commonly found in patients of immune-activated MPMs and was associated with a favorable outcome in a subtype-specific pattern. Finally, a robust 12-gene classifier was generated to classify MPMs with high accuracy, holding promise value in predicting patient survival.

Conclusions: We demonstrate that the novel classification system can be exploited to guide the identification of diverse immune subtypes, providing critical biological insights into the mechanisms driving tumor heterogeneity and responsible for cancer-related patient prognoses.