Abstract B045: Single cell spatial transcriptomics identifies unique cell populations driving muscle invasive prostate adenocarcinoma

Abstract Prostate cancer cell invasion into the surrounding muscle capsule, known as extracapsular extension (ECE), is associated with a significantly worse prognosis (5-year survival reduction: 98% to 31%). Using a live ECE mouse xenograft model with DU145 tumor cells, our group previously demonstrated that functional heterogeneity among adhesion subtypes within invasive clusters is essential for muscle invasion. Bulk RNA sequencing from three spatial compartments (pre-invasive, invasive, and post-invasive) revealed distinct gene expression profiles, with hypoxia-related and cell–cell/cell–ECM interaction genes prominently dysregulated in muscle-invasive disease. In this study, we sought to define the single-cell transcriptional heterogeneity across the three ECE compartments and compare these patterns to a human patient-derived xenograft (PDX) model. Spatial transcriptomic analysis (STA) was performed using the Bruker CosMx platform with a 6K gene panel. Data was preprocessed (quality control and dimensional reduction) and analyzed in R 4.4.0 using Seurat v5. Notable downstream analyses included unsupervised Louvain clustering, trajectory inference with Monocle 3, and cluster-specific differential gene expression. In the ECE model, STA resolved at least 11 distinct tumor cell identity clusters previously undetectable by bulk methods and revealed a clear spatial demarcation between pre-invasive and invasive populations crossing the muscle barrier. Trajectory analysis identified a pre-invasive population branching into two invasive fates with unique expression profiles. Dispersion scores, calculated using principal component embeddings, exhibited progressively increasing intracluster heterogeneity along one branch. Strong expression enrichment of the Prolaris 31-gene prognostic panel, a clinically established marker of aggressiveness, was observed at the bifurcation point of tumor evolution. STA of the human PDX model similarly revealed a high-dispersion invasive cluster branching into two fates, one exhibiting decreased heterogeneity consistent with specialization of invasive phenotypes. Together, these findings reveal, for the first time, that tumor invasion of the extracapsular muscle layer induces substantial transcriptional heterogeneity. This dynamic process highlights potential therapeutic vulnerabilities during the transition to extracapsular extension. Targeting tumor cells at this critical juncture may represent a key strategy for preventing ECE and improving prostate cancer outcomes. Supported by P30CA023074, University of Arizona Comprehensive Cancer Center Support Grant. Citation Format: Saptarshi Mallick, Rafael Sainz, Jayati Chakrabarti, Kendra D. Marr, Noel A. Warfel, Gregory C. Rogers, Beatrice S. Knudsen, Rashid Sayyid, Kelvin W. Pond, Anne E. Cress. Single cell spatial transcriptomics identifies unique cell populations driving muscle invasive prostate adenocarcinoma [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Innovations in Prostate Cancer Research and Treatment; 2026 Jan 20-22; Philadelphia PA. Philadelphia (PA): AACR; Cancer Res 2026;86(2_Suppl):Abstract nr B045.