Molecularly defined tumour microenvironment ecosystems enhance prognostic stratification in low- and intermediate-risk prostate cancer

Background & Objectives: Tumour progression in prostate cancer (PCa) is shaped by complex interactions between epithelial cancer cells and the surrounding tumour microenvironment (TME). While some relations are evident on haematoxylin and eosin (H&E) slides, transcriptomic profiling offers deeper functional insights. This study aimed to integrate RNA-seq and histopathological data to identify TME-defined tumour ecosystems (Ecotypes) and functional cell states, assessing their association with pathological features and clinical outcome. Methods: On a clinicopathological well-annotated multi-institutional cohort of 400 prostate tissue samples (179 benign and 221 malignant) with matched bulk RNA-seq, we applied EcoTyper, a computational framework that deconvolutes transcriptomic data into distinct cell types with unique transcriptional states (“cell states”) and assigns to each PCa an overall TME molecular profile called “Ecotype” (E). Associations between Ecotypes, cell states, Grade Groups (GG), and morphology were assessed. A graph neural network was trained to predict Ecotypes from H&E images. Results: A predominant Ecotype was identified in 181/221 (82%) PCa cases. Interestingly, GG1 PCa (n=62) were enriched in E6 (normal-like TME), however GG1 patients lacking E6 showed a markedly higher recurrence rate with nearly fivefold increased odds (OR= 4.76, 95% CI: 0.83–40.0). Whereas in GG2-3 PCa (n=70), E10 (longer overall survival) showed a negative correlation with the percentage of Gleason pattern 4 (p=0.02). Moreover, specific cell states (e.g., myofibroblasts, chymase-positive mast cells) were enriched in low-grade tumours. Finally, distinct H&E patterns predictive of Ecotypes were identified. Conclusion: Characterizing the TME using transcriptome-based, ecosystem- level tools provides relevant biological and clinical information beyond conventional clinicopathological features. Ecotype E6 may help identify truly GG1 indolent PCa suitable for active surveillance, while Ecotype E10 could refine prognostic stratification in intermediate-risk cases. Integrating these molecular ecosystem signatures with standard histopathological evaluation has the potential to enhance diagnostic accuracy and shape tailored treatment strategies.