Resolving the immune landscape of human prostate at a single-cell level in health and cancer

Abstract

The prostate gland produces prostatic fluid, high in zinc and citrate and essential for the maintenance of spermatozoa. Prostate cancer is a common condition with limited treatment efficacy in castration-resistant metastatic disease, including with immune checkpoint inhibitors. Using single-cell RNA-sequencing to perform an unbiased assessment of the cellular landscape of human prostate, we identify a subset of tumor-enriched androgen receptor-negative luminal epithelial cells with increased expression of cancer-associated genes. We also find a variety of innate and adaptive immune cells in normal prostate that were transcriptionally perturbed in prostate cancer. An exception is a prostate-specific, zinc transporter-expressing macrophage population (MAC-MT) that contributes to tissue zinc accumulation in homeostasis but shows enhanced inflammatory gene expression in tumors, including T cell-recruiting chemokines. Remarkably, enrichment of the MAC-MT signature in cancer biopsies is associated with improved disease-free survival, suggesting beneficial antitumor functions.

Keywords: human prostate; immune landscape; macrophage; prostate cancer; single-cell RNA sequencing; zinc.

Graphical abstract

Figure 1

Single-cell RNA sequencing reveals immune and epithelial cell heterogeneity in paired normal-cancer samples (A) Schematic describing experimental set-up for 10x genomic single-cell RNaseq of matched tumor-normal prostate samples from n = 10 patients. (B) UMAP of 15,492 cells post-QC from all prostate samples. (C) UMAP of embedding density of source of samples (normal—green to blue gradient, top; tumor—orange to red gradient, bottom). (D) Violin plot of canonical marker genes for each cell types found in prostate. Gene expression values per cell are standardized to a range from 0 to 10. (E) Patient demographics displayed as a color-coded heatmap and stacked bar charts of single-cell cell type proportions. (F) Quantification of absolute cells counts by flow cytometry per 0.5mg of normal and malignant human prostatic tissue for the indicated immune subsets. Data are a combination of n = 6 donors with each dot representing an individual donor. (G) Quantification of absolute cells counts by flow cytometry per murine prostate lobe for the indicated immune subsets. Each dot represents an individual mouse (n = 3 biological replicates). Abbreviations: MNP—mononuclear phagocyte; DC—dendritic cell; ILC—innate-like lymphoid cell; PMN—polymorphonuclear; NK natural killer (H) Confocal imaging of CD3, CD4 and CD8 in human normal and tumor prostate. Scale bars, 50 μm. (I) Confocal imaging of CD3, CD8, F4/80 and CD31 in normal murine prostate. Scale bars, 100 μm. See also Figure S1 and Table S1.

Figure 2

Androgen-receptor-negative prostate luminal epithelial cell type (A) UMAP expression plot of keratin genes in prostate cells. Increasing color gradient from purple, blue, green to yellow corresponds to increasing (standardized) expression value. (B) UMAP of normal prostate sample cells. Expression of kallikrein genes marking luminal epithelial cell types, including luminal cell type, is presented as a heatmap where cells with no expression (0 expression) are colored gray and increasing expression is colored according to increasing gradient from purple, blue, green to yellow. (C) Pre-ranked GSEA of hallmark gene sets between normal KLK3+ versus KLK4+ LE clusters. Pathways with FDR < 0.25 are colored from purple, blue, green to yellow according to decreasing FDR value. Grey circles indicate pathways that attained p < 0.05 and FDR > 0.25. Size of circles indicate the significance (signed -log₁₀(p value)). (D) Mean expression dot plot of gene encoding androgen receptor (AR). Expression values are scaled from 0 to 1. Size of circles indicate percentage of cells expressing the gene and increasing color gradient from purple, blue, green to yellow corresponds to increasing (standardized) expression value. (E) Immunohistochemistry images of KLK3, KLK4 and KRT5 in prostate tissue. Images are sourced from the Human protein atlas (https://www.proteinatlas.org). See also Figures S2–S3.

Figure 3

Immune landscape of the prostate includes a prostate-specific macrophage subset enriched in metallothionein transcripts (A) UMAP of 793 cells in myeloid compartment after integration of myeloid/MNP cells from n = 10 patients with Henry et al. myeloid/MNP cells. (B) Mean expression dot plot of top five significant marker genes for each myeloid cluster. Marker genes were identified using Wilcoxon rank sum test and p adj < 0.05 was considered statistically significant. Size of circles indicate percentage of cells expressing the gene and increasing color gradient from white to red corresponds to increasing expression value. (C) UMAP plot of predicted MNP clusters in prostate cancer single cell data from (Karthaus et al., 2020, Chen et al., 2021, Crowley et al., 2020). (D) (top) Representative RNAscope images of probes targeting MT1 family genes (magenta) and CD68 (yellow). ‘L’ indicates lumen. Arrows point to single cells that are marked by both probes in sub-panels i and ii. Scale bar, 20 μm. (bottom) Representative immunofluorescence microscopy images of a human prostate section labeled for metallothionein (α-MT)/isotype control (yellow), HLA-DR (cyan), CD206 (purple) and DAPI (blue). White arrows point to structure displaying colocalization of α-MT with HLA-DR and/or CD206 labeling. Scale bars, 50 μm. (E) Mean expression dot plot of Zinc transporter genes for each myeloid cluster. Size of circles indicate percentage of cells expressing the gene and increasing color gradient from white to red corresponds to increasing expression value. (F) Heatmap of mean AUCell enrichment of F4/80^hi/lo gene sets, corresponding to yolk sac (YS) versus hematopoetic stem-cell (HSC) lineage. Row enrichment value is scaled from 0 to 1 and presented as an increasing gradient from black, gray, yellow to orange which corresponds to increasing enrichment score. (G) Representative immunofluorescence microscopy images of cross sections of mouse prostate labeled for F4/80 (green), MHCII (red), CD11b (blue), CD31 (yellow) and phalloidin. Scale bars, 120 μm. (H) Cell counts per gram of prostate for rat IgG2a isotype or anti-Csf1r antibody (Ab) treated male mice. N = 5 per group. ^∗∗∗∗p < 0.0001; n.s denotes not significant (p > 0.05) (Two-way ANOVA with Tukey’s multiple correction). (I) Zinc concentration of anterior prostate lobe, liver lobe, and kidney from male mice treated with either rat IgG2a isotype control or anti-Csf1r Ab. N = 6 per group. (shown is representative quantification from one of two independent experiments). ^∗p < 0.05; n.s. not significant (Mann-Whitney test). See also Figures S4–S5.

Figure 4

Lymphoid single-cell landscape of normal prostate and prostate cancer (A) UMAP of 1694 lymphoid cells from n = 7 patients. Expression of marker genes for NK cells (FCGR3A, GNLY), CD8 T cells (CD8B), tissue residency and activation (CD69) and cytolytic molecule (GZMA) are shown as a heatmap where gray indicates no expression and increasing expression is colored from purple, orange to yellow. (B) Dot plot of top five significant marker genes for each lymphoid clusters. Marker genes were identified using Wilcoxon rank sum test and p adj < 0.05 was considered statistically significant. Size of circles indicate percentage of cells expressing the gene and increasing color gradient from white to blue corresponds to increasing expression value. (C) Pie chart showing proportion of cells expressing markers for (left) memory (CD27+IGHD-), naive (IGHD+CD27-), non-naive (remainder) and (right) heavy gene constant gene expression. (D) Confocal imaging of CD19, IgG and CD31 in normal murine prostate section. Scale bars, 50 μm. (E) (Top) Volcano plot showing top 15 significant DEGs between NK CD16pos and NK CD16neg (normal only). (Bottom) Violin plots of gene set testing (AUCell) for NK cell gene sets (KEGG and GO) and lymphocyte tissue residency gene sets from (Mackay et al., 2016). Significance is denoted by ^∗∗p < 0.01; ^∗∗∗p < 0.001 (Mann-Whitney test). Position of asterisks indicate the group with higher expression. (F) Confocal imaging of NKp46, MHCII, F4/80 and CD31 in normal murine prostate section. Scale bars, 35 μm. See also Figure S6.

Figure 5

Perturbed immune cell transcriptomes and cellular interactions in prostate tumor (A) Violin plot of gene module score of GO term corresponding to antigen processing and presentation in tumor versus normal in myeloid cells. Kruskal-Wallis test was performed between normal and tumor for each cluster and p < 0.05 was considered statistically significant. (B) Pie chart of co-activating and co-inhibitory DEGs between normal (light colored sections) and tumor (dark colored sections) in myeloid clusters. Darker sections indicate genes that were that were upregulated in tumor versus normal and lighter section indicate upregulation is in normal versus tumor (> = 1.25 log₂ fold change). Grey sections indicate < 1.25 log₂ fold change. (C) Mean expression dot plot of costimulatory/coinhibitory molecules. Increasing expression corresponds with increasing gradient from white to red (MNP/B cell/epithelial/stromal) or white to blue (T/NK/B cell) corresponding to increasing expression value. Size of circles indicate the percentage of cells expression the gene. (D) Violin plot of gene set test (AUCell) results in CD8 cytotoxic T cell cluster for murine CD8 T cell exhaustion gene set curated from (Doering et al., 2012). Significance is denoted by ^∗p < 0.05 (Mann-Whitney test). (E) GSEA of KEGG pathways for NK CD16^neg tumor versus normal. Statistically significant pathways are colored and labeled. (F) Mean expression dot plot of leading edge genes in cytokine-cytokine receptor interaction as in (E). Only genes that are expressed by at least 20% of cells are plotted. (G) (Top) Expression of CCL5 and prediction/label transfer score of NK CD16^neg cells in visium data of normal and tumor prostate sections. (Bottom) Spatial correlation of CCL5 with NK CD16^neg cells in prostate cancer visium data. Only positive correlations are plotted; increasing value of correlation is shown as a gradient from white to red. (H) CellPhoneDB receptor-ligand interaction analysis between B cell and myeloid clusters. (I) Representative immunofluorescence confocal microscopy of BAFF and HLA-DR in human prostate tumor section. Scale bars, 20 μm. (J–L) (J) Expression of TNFSF13B and prediction/label transfer score of MAC-MT cells, and correlation of TNFSF13B with MAC-MT cells in prostate cancer visium data. Only positive correlations are plotted; increasing value of correlation is shown as a gradient from white to red. CellPhoneDB receptor-ligand interaction analysis between (K) fibroblasts and T cell clusters, and fibroblast and myeloid clusters and (L) myeloid clusters with LE clusters split by group (N = normal; T = tumor). The order of the receptor-ligand interactions corresponds to the order of the cell-types i.e., cell type A expressing molecule A interacts with cell type B expressing molecule B. Size of circles and color gradient corresponds to the receptor-ligand interaction score, which purple, blue, green to yellow for increasing values. Significant interactions (p < 0.05) are highlighted in red. See also Figure S7.

Figure 6

MT1-expressing macrophages in tumor have increased metallothionein and pro-inflammatory gene expression and are associated with improved tumor event-free survival (A) Violin plots show genes than achieved a p adj < 0.05 after statistical analyses with Wilcoxon Rank Sum Tests. Color of adjusted p value indicates the group where expression is higher (orange = tumor). (B) Mean expression dot plot of metal ion transport genes in MAC-MT cluster separated by normal (N) or tumor (T) in rows. Size of circle indicates the percentage of cells expressing the genes and color indicates which group (normal or tumor) expresses higher (dark red) levels of the genes. (C) Heatmap of mean AUCell enrichment of 27 macrophage-stimulation signatures split by normal or tumor. Row expression value is scaled from 0 to 1 and presented as a gradient from purple, blue, green to yellow. (D) GSEA of KEGG pathways in tumor versus normal for LE-KLK4. Pathways were considered statistically significant if p value < 0.05 (marked by vertical dashed red line). Size of circles indicate normalized enrichment score (NES) and colors indicate if pathways achieved FDR < 0.25 starting from purple, blue, green to yellow as significance values decreases. (E) String-DB analysis of leading edge genes from selected pathways enriched in tumor MAC-MT. (F) Mean expression dot plot of CCL5, CXCL9, CXCL10 in MNP clusters and CXCR3 in lymphoid clusters. Size of circle indicates the percentage of cells expressing the genes and increasing expression (scaled from 0 to 1) corresponds to increasing color gradient from purple, blue, green to yellow. (G) Representative immunofluorescence confocal microscopy of CD8 and HLA-DR in human prostate tumor section. Scale bars, 10 μm. (H) Expression of CXCL9 and CXCL10 and prediction/label transfer scores of MAC-MT and CD8 cytotoxic cells in visium data of tumor prostate sections. (Bottom) Spatial correlation of CXCL9 and CXCL10 with MAC-MT or MAC-MT with CD8 cytotoxic cells in prostate cancer visium data. Only positive correlations are plotted; increasing value of correlation is shown as a gradient from white to red. (I) Kaplan-Meier survival curve for TCGA-PRAD disease free index with deconvolved MAC-MT score. Samples were categorised into high (black, top 25%) and low (red, bottom 25%) of deconvolved score. Statistical analysis was performed with log rank test and p < 0.05 was considered statistically significant. See also Figures S8–S9 and Table S2.