Abstract The formation of intratumoral spatial niches has been reported for many human malignancies. However, the translational potential of such spatial niches is understudied. Herein, we utilize digital spatial profiling (DSP) to explore the prognostic relevance of spatially defined protein expression in high-risk prostate cancer. A total of 49 patient samples were analyzed for the expression of 46 proteins in 463 regions of interest (ROIs) from the tumor center ( n = 198) and the tumor periphery ( n = 265) resulting in 21,298 primary data points (mean per patient n = 9.4). Expression data from either the tumor center or the tumor periphery were not found to be prognostic. Protein expression of tumor center and periphery was then integrated into single datapoints by calculating the log 2 -transformed relative expression between the two niches for each protein and patient. Unsupervised hierarchical clustering of these data yielded two distinct patient subgroups. These clusters did not show a statistically significant correlation with known prognostic parameters yet significantly correlated with progression-free survival ( p = 0.014, log-rank, HR 0.43; 95% CI, 0.22–0.86). Our results thus reveal that spatial protein expression contains prognostic information, however, only when expression data from both spatial niches are taken into account. In conclusion, our proof-of-concept study shows that DSP can be exploited for the development of novel prognostic biomarkers that rely on spatially resolved protein expression.
Schneider et al. (2026) studied this question.