Extracellular matrix collagen remodeling predicts resistance to immune checkpoint blockade in mismatch repair-deficient endometrial cancer.

BACKGROUND: Immune checkpoint inhibitors (ICIs) have revolutionized the therapeutic paradigm for mismatch repair-deficient endometrial cancer (MMRd EC). However, a considerable subset of patients fails to achieve durable clinical benefit. This pronounced interpatient variability underscores the urgent unmet need for mechanistic predictive biomarkers beyond MMRd status alone to guide clinical decision-making. METHODS: We performed a multiproteomic mass spectrometry analysis on tumor specimens from an advanced or recurrent MMRd EC cohort stratified into ICI responders (ICI-R) and non-responders (ICI-NR). Proteomic signatures were validated spatially and structurally using whole-section Masson’s Trichrome digital imaging and polarization-resolved second-harmonic generation microscopy. Machine learning algorithms were constructed to integrate extracellular matrix (ECM) structural metrics with immune parameters to build a multiparametric predictive model of ICI resistance. RESULTS: ECM remodeling emerged as the dominant proteomic program characterizing ICI resistance. Quantitative spatial imaging revealed that increased intratumoral collagen density and peripheral fiber disorganization were significantly associated with impaired CD8+ T-cell infiltration and immunotherapy failure. Integrating collagen structural metrics with immune parameters yielded a predictive score that showed promising discrimination between ICI-Rs and ICI-NRs in our cohort, providing a framework for future validation in independent cohorts. CONCLUSIONS: Our findings implicate ECM remodeling as a key determinant of immune exclusion and ICI resistance in MMRd EC. These findings support developing collagen-based spatial biomarkers and stromal-targeted approaches, though the clinical utility of this predictive score requires further validation in larger, independent cohorts.