INTRODUCTION AND OBJECTIVE:
Analysis of the transcriptomic landscape of prostate adenocarcinoma has shown multidimensional gene expression variabilities. Understanding the complexity of cancer transcriptome can provide biological insight and therapeutic guidance. To avoid potential confounding factors such as stromal contamination and stress-related material degradation, we utilized a set of prostate epithelial cell expressed genes from single cell transcriptome data of the human prostate gland.
By analyzing bulk and single cell RNA sequencing data publicly available, we defined 1,629 genes expressed by prostate epithelial cells. Consensus clustering and the CIBERSORT deconvolution were used for class discovery and proportion estimate analysis. The Cancer Genome Atlas Prostate Adenocarcinoma (TCGA-PRAD) dataset was used as a training set. The resulting clusters were analyzed in association with clinical, pathologic, genomic characteristics and impact on survival.
The TCGA-PRAD tumors could be separated into four subtypes—subtype A (35.6%), subtype B (30.3%), subtype C (9.0%), subtype D (16.1%) and mixed (9.0%). Subtype A was characterized by low frequency of ETS family fusions and high levels of KLK3 gene expression which encodes PSA. Subtype B showed highest expression of ACPP gene encoding PAP (Prostatic Acid Phosphatase). Subtype C and D were commonly associated with advanced T/N stages, high Gleason grades, p53 or PI3KCA mutations. In silico drug sensitivity screening suggested that Subtype B is likely to be sensitive to docetaxel and cabazitaxel. Serum PSA/PAP ratio was predictive of radiographic response to docetaxel in metastatic castration resistant prostate cancer patients.
We propose four subtypes of prostate adenocarcinoma with distinct transcriptomic, genomic and pathologic characteristics. PSA/PAP ratio in advanced cancer may be useful to determine who benefit most from maximized AR inhibition or early use of anti-microtubule agents. The molecular subtyping and biomarkers need validation in prospective cohort study.
Source of Funding:
This study was supported by the Korea Health Industry Development Institute (KHIDI)’s grant for research under the Biomedical Global Talent Nurturing Program of KHIDI (HI19C0723).