Abstract
Pleural mesothelioma (PM) is an aggressive cancer with a dismal prognosis. Co-targeting of immune checkpoint inhibitors (ICI) CTLA-4 and PD-1, the new standard of care for PM, is still clinically unsatisfying, regardless of PM histotype. Moreover, no predictive biomarkers of ICI efficacy are available in PM. We performed multi-omics analysis of pre-ICI therapy lesions from 91 PM patients enrolled in the multicenter NIBIT-EPI-MESO study and identified four clinically meaningful PM subsets with progressively increasing global DNA methylation profiles from demethylated (DEM), LOW, intermediate (INT), and CpG island methylator phenotype (CIMP). This tumor methylome classification predicted both response and survival to ICI therapy. Indeed, the LOW subset was enriched in responder patients, who also had the longest median Overall Survival (mOS) and the highest 3-year OS rate, and showed a T- and B cell-rich immune microenvironment. The LOW subtype was also the most predictive variable of ICI therapy outcome, irrespective of the PM subtype, in a multivariate model, including gender, age, tumor mutational burden, and an interferon-gamma gene expression signature. Conversely, the CIMP subtype was enriched in non-responder patients who had the shortest mOS and OS rate, along with a depleted tumor immune microenvironment. We also developed a publicly available, methylation-based, decision-making probabilistic classification tool to predict the outcome to ICI treatment of PM patients. The NIBIT-EPI-MESO study included PM patients enrolled in NCT01655888, NCT02588131, and NIBIT MESO-2 Study Protocol Number 22973, as well as patients treated in routine clinical practice.