In silico Computed Clusters of Prostate Luminal Progenitors Match FACS-Enriched LSC^med cells

Abstract

Several groups recently published single-cell (sc) expression atlases of the adult mouse prostate cells based on RNA sequencing (scRNA-seq) data. All studies identified one computerized cluster of non-secretory luminal progenitor cells enriched in luminal and stemness-related gene transcripts. The actual correspondence between these luminal progenitor cell clusters has not been investigated. In addition, the presence of Krt4 (encoding cytokeratin 4) in these in silico-identified luminal progenitors suggested the overlap with FACS-enriched LSCmed luminal progenitor cells earlier identified as a stem-like, castration-tolerant and tumor-initiating cell population. Here, we used a unified bioinformatics pipeline to re-analyze published prostate scRNA-seq datasets and perform various pan-transcriptomic comparisons including the LSCmed cell signature. Our study demonstrates that i) the mouse prostate luminal progenitor cell clusters identified in the different scRNA-seq studies largely overlap and can be defined by a common 15-gene signature including Krt4, ii) mouse LSCmed cells match both mouse and human luminal progenitors identified by scRNA-seq analysis. Bridging these in silico-identified and ex vivo-characterized prostate luminal progenitor subsets should benefit our understanding of their actual involvement in prostate diseases.

Prostate luminal progenitor cells: from mouse to human, from health to disease

“…Strikingly, several of them, including the LSC med cell marker Krt4, were previously identified in the 111-gene mouse LSC med cell signature 31 (TaBle 3). Accordingly, a significant enrichment of the latter signature was observed in each scRNA-seq dataset for the sole population identified as luminal progenitor cells 50 (Fig. 2a-d).…”

“…In the past 3 years, several single-cell RNA sequencing (scRNA-seq) studies of healthy mouse prostates have provided unprecedented evidence that the molecular diversity of the prostate epithelium far exceeds the classical histological description [32][33][34][35][36] . Despite the largely different number of cells sequenced in each scRNA-seq study 50 , a unique cluster of basal cells was identified in all of them, suggesting a relative molecular homogeneity. By contrast, luminal cell heterogeneity ranged from three to six distinct clusters (TaBle 2), which most probably reflects methodological differences in bioinformatic processes (for example, parameters used for cell clustering).…”

“…Although no secretoglobin gene expression could be detected in mouse LSC med -like cells 36 , the LSC med cell signature was enriched only in Club and Hillock cells of the human prostate (P < 2e-16) 50 (Fig. 2e).…”

“…Each publication named this cell population differently (TaBle 2). To investigate the actual overlap between these cell clusters, we reanalysed the scRNA-seq transcriptomic data [32][33][34][35][36] to identify the top biomarkers of these putative luminal progenitor cells in each individual study and compare their expression level in each cell population across studies 50 . Although one study claimed that the non-secretory luminal cell cluster was actually of urethral origin 36,37 (see Spatial distribution), their scRNA-seq data were included in the comparison given the obvious molecular proximity of these 'urethral' luminal cells with the putative luminal progenitor cells (herein used as the generic term for this cluster) referred to as prostatic cells by all other groups [32][33][34][35] .…”

“…2 | Transcriptomic similarity between FACS-enriched LSC med cells and putative luminal progenitor cell clusters identified by scRNA-seq analyses of wild-type mouse and healthy human prostates. a-c | The transcriptomic correspondence between lineage-negative (Lin − )SCA-1 + CD49f med (LSC med ) cells and prostate luminal progenitor cells identified by single-cell RNA sequencing (scRNA-seq) is illustrated using the dataset of Guo et al 32 as a representative example (see Puig Lombardi et al 50 for technical details, including the number of cells sequenced in all the studies compared). a | UMAP projections based on linear dimensionality reduction by principal component analysis (PCA) for 19,503 single-cell transcriptomes.…”