HuPSA and MoPSA Atlases Reveal Novel Cell Populations and Lineage Plasticity in Prostate Cancer at Single-Cell Level

Androgen deprivation therapy has improved patient survival. Nevertheless, treatment resistance inevitably emerges due to the complex interplay of tumor heterogeneity and lineage plasticity. We integrated scRNAseq data from multiple studies, comprising both publicly available cohorts and data generated by our research team, and established the HuPSA (Human Prostate Single cell Atlas) and MoPSA (Mouse Prostate Single cell Atlas) datasets. Through unsupervised clustering and manual annotation, we found that both atlases consisted of previously known cell clusters including prostate adenocarcinoma (AdPCa), neuroendocrine prostate cancer (NEPCa), stromal, and immune cell populations. Our analysis also unearthed the less described populations including MMP7+ prostate club cells and two novel lineage plastic cancerous populations, namely Progenitor-like and KRT7. Immunohistochemical staining analysis confirmed the presence of these populations in both human and mouse PCa tissues, reinforcing their significance in PCa pathobiology. Furthermore, employing HuPSA based deconvolution, we scrutinized 877 high-quality human PCa bulk RNAseq samples and reclassified them into different molecular subtypes, including the newly discovered KRT7 and Progenitor-like categories. Moreover, employing supervised dimensional reduction and label transferring techniques, we projected the scRNAseq data derived from C4-2B xenograft tumors onto HuPSA. Our analysis effectively identified the diverse subpopulations within the tumor, including C4-2B derived AdPCa and NEPCa cells as well as the murine cells in the tumor microenvironment. We launched the “PCaAtlas” app (https://pcatools.shinyapps.io/ProAtlas_dev/) for users to visualize genes expression in re-classified human prostatic tissue samples. In conclusion, our study has successfully integrated multiple scRNAseq datasets and made these data easily accessible to researchers. Our results elucidate a roadmap of PCa progression, showcasing the development of heterogeneous populations and the involvement of lineage plasticity. This understanding holds promise for guiding the development of precision medicine in PCa field. Additionally, the HuPSA and MoPSA provide invaluable blueprints for analyzing and interpreting user-generated PCa single-cell RNAseq data. Overall design: C4-2B xenograft and TRAMP tumor tissues were harvested and dissociated using “Tumor Dissociation Kit, human, #130-095-929” and “Tumor Dissociation Kit, mouse, #130-096-730” from Miltenyi Biotec. The dead cells were removed by “Dead Cell Removal Kit, #130-090-101” from Miltenyi Biotec. The red blood cells were removed by “Red Blood Cell Lysis Solution, # 130-094-183” from Miltenyi Biotec. The cDNA libraries were constructed using Chromium Next GEM Single Cell 3' Reagent Kit v3.1. The resulting cDNA was fragmented, processed, and size selected. Adaptors were ligated and dual indexes were attached via PCR. The libraries were cleaned up and assessed for quality with the Agilent TapeStation D5000 High Sensitivity Kit. The libraries were quantitated with qPCR (NEBNext Library Quant Kit for Illumina, Bio-rad CFX96 Touch Real-Time PCR) and normalized to 1.5 pM and pooled. The library pool was denatured and diluted to approximately 300 pM. A library of 1.5 pM PhiX was spiked in as an internal control. Paired-end sequencing was performed on an Illumina NovaSeq 6000.