Relationship of spatial transcriptomic analyses with distance-associated transcriptomic spectrum and morphology-specific signatures of localized prostate cancer.

Abstract

422 Background: Current clinical management and classification of prostate cancer is dependent upon specimen morphology assessment. Limited studies have focused on identifying and mapping transcriptomic signatures within morphological variants. Additionally, the impact of transcriptomic distance-relationship between benign and tumor glands has yet to be identified. In this analysis, we profiled and derived signature gene sets for morphological variants and assessed the spatial impact of distance from the adjacent tumor areas on surrounding benign tissues. Methods: Four (three African American and one European ancestry) formalin fixed paraffin embedded prostate cancer samples were obtained via radical prostatectomy and underwent the GeoMx DSP workflow with gene expression level counts. The four patients each had different dominant morphologic features on the slide used, with one patient with extensive foamy gland morphology, one with extensive atrophic morphology, one with extensive mucinous morphology, and one with extensive Gleason pattern 5 carcinoma. Other regions of interest (ROIs) were benign (including benign glands, benign stromal) or mixed tumor/benign. Transcriptomic analysis was performed on these ROIs to derive the morphology-specific gene signatures. Additionally, distance-based spectrum of proximity of benign ROIs to tumor ROIs was constructed to identify genes that are significantly associated with distance to the nearest tumor ROI. Results: 96 ROIs from four patients (denoted by the dominant morphological tumor features) were annotated. In total, 95 ROIs and 8,102 genes passed quality control for analyses. A distance-based spectrum analysis revealed that benign ROIs that are closer to tumor ROIs were highly enriched in transcriptomic androgen signature and MYC signature and endothelial or club-like features were enriched in benign ROIs further from tumor ROIs. Moreover, distinct gene set signatures were established for each morphological feature. Conclusions: Benign ROIs closer to the tumor are more tumor-like and enriched in androgen receptor pathways and MYC signatures. We identified transcriptomic differences among the morphological tumor features that demonstrate inter-patient and intra-patient tumor heterogeneity.