Introduction: Prostate cancer (PCa) in the transition zone (TZ) accounts for approximately 30% of disease and tends to present with higher PSAs with a lower risk of seminal vesicle invasion, extra-capsular extension, and risk of biochemical recurrence compared with peripheral zone (PZ) tumors. The underlying biological mechanism for these differences is poorly understood. Here, we performed spatial transcriptomic profiling to elucidate the molecular differences between TZ and PZ PCa.
Methods: We identified three patients who underwent radical prostatectomy for PCa (one each with PZ only, TZ only, and both PZ and TZ tumors) and used the NanoString’s Digital Spatial Profiling (DSP) platform to quantify whole transcriptomic gene expression data (17,128 genes) in multiple regions of interest (ROI) per patient (42 cancer and 8 normal samples). Four morphology markers to facilitate ROI selection in both cancer and normal areas (SYTO13 for nucleus, PanCK for epithelium, SMA for stroma and CD45 for immune cells) were selected. Raw counts for 17,128 genes were imported to R v.4.1.0 for downstream data analyses. Counts were Q3 normalized and scaled (Z-score) to enable plotting of all genes on the same axes. Differential gene expression analysis using a linear model fit by empirical Bayes moderation, gene set enrichment analysis (GSEA) by cancer hallmarks, XCell gene sets for pathway enrichment and immune cell deconvolution using CIBERSORT was performed.
Results: There were grade group (GG) 4 (n=10) and 5 (n=10) tumors in PZ and GG 3 (n=10), 4 (n=11) and 5 (n=1) cancers in TZ regions. We observed distinct gene expression profiles between PZ (n = 20) and TZ (n=22) tumors. Interestingly, androgen receptor (AR) signaling was significantly higher in TZ PCa ROIs compared to PZ ROIs in both GSEA (false discovery rate < 5%) and the androgen subcomponent of the genomic prostate score (p < 0.001), regardless of grade, epithelial, stromal or immune component of the region. To standardize the comparison, CIBERSORT’s absolute immune signature scores were only computed for GG4 tumors and found to be significantly higher in PZ GG4 tumors compared to TZ GG4 tumors. Notably, CD4+ memory T cells were significantly higher in PZ GG4 regions compared to TZ GG4 tumor regions (p < 0.05).
Conclusions: Here, we demonstrate distinct gene expression profiles of PZ and TZ PCa. Specifically, we observed higher AR signaling in TZ cancers and higher levels of immune infiltration on PZ cancers. This is in concordance with prior knowledge that TZ tumors may be associated with higher serum PSA and PZ tumors may be associated with inflammation. Further studies are needed to discern the biological and clinical significance of the different molecular features of PZ and TZ PCa.