Spatial ecostructural modelling of endometrial cancer identifies the key role of CD90 + CD105 + endothelial cells in tumour heterogeneity and predicts disease recurrence.

BACKGROUND: Current therapeutic strategies for endometrial cancer are mainly based on aggressive histological types and molecular subtypes. However, ignoring the spatial distribution of immune/stromal cells fails to account for the heterogeneity of the local tumour microenvironment, leading to biased prediction of treatment response. The goal of precision medicine is to delineate the biological characteristics of local functional units based on molecular labelling, which adequately reflects spatially adaptive changes during treatment or metastasis. METHODS: Single-cell resolution analysis of 40 endometrial cancer cases across four molecular subtypes was performed using imaging mass cytometry (IMC) to quantify the frequency, spatial distribution, and intercellular crosstalk of distinct immune and stromal cell populations. These ecosystem-level features were systematically correlated with clinical features and outcomes, including treatment response and survival. We further identified CD90 + clusters as key regulators of macrophage polarization and T-cell infiltration dynamics, with flow cytometry used to validate their functional role in tumour subtype specification and microenvironmental remodelling. Finally, machine learning-based spatial phenotyping was employed to construct molecular subtype-specific signatures and a highly accurate recurrence prediction model for high-risk endometrial cancer. RESULTS: Single-cell profiling revealed that CD90 + clusters constitute a critical immunomodulatory component within the tumour microenvironment, demonstrating significant enrichment in macrophage differentiation pathways and serving as key mediators of intercellular signalling networks. Furthermore, computational models integrating functional molecular signatures with cell-cell interaction profiles demonstrated high predictive accuracy for both molecular subtyping and recurrence risk stratification in patients with endometrial carcinoma. CONCLUSIONS: Our study establishes a spatial eco-context framework for molecular subtypes of endometrial cancer by integrating single-cell spatial multiomics data. This approach enables high-resolution mapping of tumour-immune-stromal interaction networks and reveals novel targets for personalized therapeutic strategies.