Regenerative potential of prostate luminal cells revealed by single-cell analysis

Abstract

Androgen deprivation is the cornerstone of prostate cancer treatment. It results in involution of the normal gland to ~90% of its original size because of the loss of luminal cells. The prostate regenerates when androgen is restored, a process postulated to involve stem cells. Using single-cell RNA sequencing, we identified a rare luminal population in the mouse prostate that expresses stemlike genes (Sca1 ⁺ and Psca ⁺) and a large population of differentiated cells (Nkx3.1 ⁺, Pbsn ⁺). In organoids and in mice, both populations contribute equally to prostate regeneration, partly through androgen-driven expression of growth factors (Nrg2, Rspo3) by mesenchymal cells acting in a paracrine fashion on luminal cells. Analysis of human prostate tissue revealed similar differentiated and stemlike luminal subpopulations that likewise acquire enhanced regenerative potential after androgen ablation. We propose that prostate regeneration is driven by nearly all persisting luminal cells, not just by rare stem cells.

Fig. 1.. Three subsets of luminal cells identified by scRNA-seq of the intact mouse prostate.

(A) Single cell census of the intact prostate. tSNE of scRNA-seq profiles colored by unsupervised clustering of 15 subsets and labeled post hoc. (B) Prostatic luminal subtypes. tSNE of scRNA-seq profiles only from the luminal clusters in (A). (C) Validation of luminal subset markers in situ. IF staining of L1 (CD26/Dpp4, Cyan, top) and L2 (Tacstd2/Trop2, Red, middle) markers in the proximal and distal anterior lobe, along with Epcam (for epithelial cells, white), Ck5 (basal cells, green) and DAPI (nuclei). Bottom: IHC staining of Foxi1 in the proximal and distal anterior lobe. (D) Sharp transition from L2 to L1 cells. IF staining of L1 (CD133/Prom1 or CD26/Dpp4) and L2 (Tacst2/Trop2) markers, along with Epcam (for epithelial cells), Ck5 (basal cells) and DAPI (nuclei). A distinct border is observed between proximal and distal prostatic regions. Scale bars: 100 or 50 μm as labeled.

Fig. 2.. Transcriptomic changes in murine luminal subpopulations during castration and organ regeneration.

(A) Schematic overview of the castration/regeneration cycle with experimental time-points. (B) Scatterplots of the L1 (x axis) and L2 (y axis) intact signature score (z score) for each cell (dot) assigned to L1 (red) or L2 (blue) at each time point (panel). Dot color intensity is scaled by the strength of their classifier assignment probability for their assigned class (colorbar). (C) Similar transcriptional states of L1 and L2 during castration. PHATE graph of scRNA-seq profiles from luminal cells, colored by time point (left panel) or L1, L2, L3 based on expression profiles in T0 (right panel). L1 cells undergo the most substantial transcriptional changes. On castration Day 28 (dark green, left panel) and regeneration day 1 (light green, left panel) L1 are co-embedded with L2 cells (orange, right panel). (D) Rapid entry of L1 and L2 cells into the cell cycle during regeneration. Each plot shows the distribution of Mki67 mRNA expression (y axis) throughout the C/R cycle (x axis) for L1, L2 and L3 cells. Fraction of cells with Mki67 expression detected is noted on top. *expression is significantly different from intact (T0) (Bonferroni corrected P < 0.05, one sided Wilcoxon rank-sum test) **designates fold change of 1.5 or greater, and AUC of 0.65. (E) IF staining of Ki67 in the anterior lobe at regeneration day 2. Left: Low magnification showing proximal and distal regions. Right: representative higher magnification (20x) of proximal and distal regions. Ki67 (red), Ck8 (Green), Ck5 (white) and DAPI (purple). Scale bars: 200 or 50μm as labeled.

Fig. 3.. Enhanced regenerative potential of murine luminal cells after castration in organoid culture and in vivo.

(A-C) Enhanced organoid formation of L1 and L2 cells isolated post castration or during regeneration. (A) Relative organoid formation (%, y axis; mean ± standard deviation) in the presence of 1 nM DHT in cultures initiated by L1 or L2 cells isolated by CD26/Dpp4 (L1), CD133/Prom1 (L1) or Sca1/Ly6a (L2) expression from hormonally intact prostate (blue) or prostate 28 days post castration (red). The number of organoids was quantified 7 days after seeding 200 cells. (N=3 * designates p< 0.05, t-test). (B) Relative organoid formation (%, y axis; mean ± standard deviation) from L1 or L2 cells isolated by CD26/Dpp4 or Sca1/Ly6a expression, respectively, from a prostate 28 days post castration (red) or a prostate 2 days into regeneration (blue) in the presence or absence of 1 nM DHT. The number of organoids was quantified 7 days after seeding 200 cells. (N=3 * designates p< 0.05, t-test). (C) Representative images of organoids derived from CD26/Dpp4⁺ L1 cells (top) or Sca1 L2 cells (bottom). Left: brightfield; Right: Confocal images (single Z and maximum projection) stained with Ck8 (red), Ck5 (green), Epcam (white) and DAPI (purple) 7 days post establishment. Scale bars: 100μm. (D) Lineage tracing strategy. (E, F) Contribution of multiple clones to prostate gland regeneration. (E) Top: Maximum projection of a castrated prostate 7 days post tamoxifen injection (left) and 4 weeks post regeneration (right). Only red fluorescence protein (RFP) and yellow fluorescence protein (YFP) are shown. Scale bars: 500μm. Bottom: Higher magnification of lineage-traced prostates showing contribution of multiple clones to gland regeneration. Scale bars: 100μm. (F) Distribution of size of different color clones. Log₂ clone size (y axis) is plotted from three independent mice compared to control (castrate 7 days post tamoxifen) (x axis). Raw data is in table S2. As observed previously, GFP+ clones are infrequent in the prostate (34).

Fig. 4.. Androgen receptor-mediated induction of neuregulin in mesenchymal cells is a potential driver of luminal regeneration.

(A) Changes in expression of key stromal ligands over the C/R time course. Smoothed mean expression relative to intact prostate (T0) (y axis) of ligands in different subsets of stromal and epithelial cells (per color code). (B) In situ validation of growth factor expression by RNA-FISH of prostate tissue isolated on regeneration day 2. Representative growth factors (Egf, Nrg2, Rspo3, Fgf10; green), luminal cells (CD24a; white) and proliferating cells (Mki67; red) are shown. Scale bar: 25μm. (C-E) Nrg promotes luminal regeneration in mouse and human organoids. (C) Relative proliferation of murine L1 cells (CD26/Dpp4⁺; top) and L2 cells (Sca1/Ly6a⁺; bottom) in the presence of Egf, Nrg, Fgf10 Igf or no growth factor, in the presence of DHT (1nM) or enzalutamide (10μM). The data are displayed as average growth ± standard deviation (y axis) of 5,000 cells measured by Cell titer glo at 7 days. Base organoid medium contains noggin, R-spondin, A83-001 and Y-27632. N=3. *designates p< 0.05, **designates p< 0.01, t test. (D) Relative proliferation of murine L1 and L2 cells measured as in (C), in the presence of EGF alone or EGR in combination with either Nrg, Ffg10 or Igf, all in the presence of DHT (1nM) (x axis). N=3. *designates p< 0.05, **designates p< 0.01, t test. (E) Relative proliferation of human prostate luminal cells (CD26/DPP4⁺) measured as in (C), in the presence of EGF, NRG or ERG plus NRG in base human organoid medium (NOGGIN, R-SPONDIN1, FGF2, FGF10, PGE2, A83-001, NICOTINAMID, SB202190, DHT and Y-27632). N=3. *designates p< 0.05, **designates p< 0.01, t test. Human organoids for this panel were derived from normal prostate tissue isolated during cystectomy surgery.

Fig. 5.. Androgen deprivation enhances the regenerative potential of human prostate luminal cells.

(A) Enhanced organoid formation by human luminal cells obtained post castration. Left: Relative organoid formation (mean ± standard deviation) of CD26/DPP4 luminal cells isolated from prostates obtained by radical prostatectomy from hormonally intact patients (N=5, blue) or patients treated with androgen deprivation therapy (N=5, red). Organoids were quantified 14 days after seeding of 200 cells. N=4. **designates p< 0.01, Welch’s t test. (B) Representative brightfield image (right), H&E stain (middle) and confocal image (right) of a human organoid derived a patient treated with ADT as in (A). For the confocal image: CK8 (red), CK5 (green), EPCAM (white) and DAPI (purple). Scale bar: 100μm. (C) Schematic of human prostate processing for scRNA-seq. (D) Top: PHATE map of luminal cells from all samples stratified by treatment (left) and by sample (right). Bottom: PHATE maps colored by correlation to RNA signatures derived from Henry et. al. (30). L1 (left); L2 “Club” (right). (E) Pairwise correlation of signature scores for L1 and L2 “Club” cells (30) per patient after CNA filtering. Signatures were generated using previously published human prostate luminal cell data (30). * designates significant change of the median correlation (p<0.05, Welch’s t-test, one-sided test). (F) Model of prostate regeneration. The prostate gland shrinks ~90% following androgen deprivation (castration) due to loss of luminal epithelial cells. During this process, the transcriptome of L1 cells closely resembles that of more stem-like L2 cells. Androgen addback stimulates production of growth factors by distinct populations of mesenchymal cells, which rapidly recruit nearly all persisting luminal cells into cell cycle. Each of these proliferating luminal cells collectively contributes to regeneration of the prostate gland, rather than a rare stem cell population.