Leveraging spatial proteomics to identify epithelial cell subtypes driving poor prognosis in prostate cancer.

Abstract

5126 Background: Epithelial cell features, such as morphology and cellular organization within the tumor architecture (e.g. Gleason Score), have long been known to be prognostic in prostate cancer. However, the specific epithelial cell subtypes driving clinical outcomes in patients remains unclear. Traditional tumor profiling techniques lack the resolution needed to interrogate complex single-cell biomarkers. Imaging Mass Cytometry (IMC) now enables high-dimensional identification and detailing of epithelial, stromal, and immune cell subtypes within patient tumors. Using a custom IMC panel, we profiled tumor biopsies from a large prospective biopsy cohort with long-term (median >12-year) clinical follow-up to identify potentially targetable epithelial cell populations associated with adverse outcomes in localized prostate cancer. Methods: Spatially resolved protein expression profiling was performed on primary prostate cancer tumor biopsies using a custom IMC assay enriched for targetable tumor markers. Cell segmentation and single-cell expression measurement were performed using the published ‘steinbock’ toolkit. Biochemical progression-free survival (bPFS) and cancer-specific survival (CSS) were pre-specified clinical endpoints. Univariate and multivariable survival analyses stratified by cell abundance tertiles were performed using a Cox proportional hazards model. The Mann-Whitney U test was used to assess for pairwise differences in cell abundance between patient groups. All significance testing was performed using a two-tailed significance level of 0.05. Results: Protein co-expression patterns in >3.4 million cells comprising 573 biopsy samples obtained from 385 patients were measured. Single-cell analysis revealed 15 cell clusters representing luminal prostate cancer cells (including PSMA-high, PSMA-intermediate, and PSMA-low), basal epithelial cells, and lymphocytes. Patients with tumors enriched for PSMA-high epithelial cells demonstrated impaired bPFS ( P <0.001) and CSS ( P =0.016). PSMA-high epithelial cells were enriched in high-grade (Gleason Score 8+) tumors (P<0.05). Multivariable analysis revealed PSMA-high epithelial cell enrichment to be prognostic of bPFS independently of Gleason Score and serum PSA at time of diagnosis ( P <0.05). Co-expression analysis demonstrated that these prognostic PSMA-high cells also expressed high levels of the cell-surface targets KLK2, B7-H3, and protein. Conclusions: We identified an epithelial single-cell biomarker associated with adverse clinical outcomes in localized prostate cancer. If validated through ongoing experiments in an independent cohort, our findings support new potential strategies for treatment intensification using targeted therapies in select high-risk patients.