Characterization of the immune landscape in healthy mouse prostate and during prostate cancer progression

ABSTRACT

The immune landscape of healthy prostate and its alterations during prostate cancer (PCa) progression remain poorly characterized. Using scRNA-sequencing and multiparametric flow-cytometry analysis, we comprehensively characterized immune cells in wild-type and PTEN^(i)pe−/− mouse prostates, a model that closely recapitulates human PCa. PCa in PTEN^(i)pe−/− is marked by the recruitment of tumor-associated neutrophils (TANs), which represent the dominant immune cell population and resolved into eight distinct states, Trem2⁺ tumor-associated-macrophages (TAMs), and exhausted CD8⁺ T cells. Trem2⁺ TAMs differ from the three main resident macrophage populations in the healthy prostate, exhibiting a strong metabolic and immunosuppressive signature, likely driven by the MIF/HIF1A-signaling axis. This study provides the first detailed characterization of immune cells in the healthy mouse prostate and reveals changes in the immune landscape associated with prostate cancer progression.

SUMMARY

Using scRNA-sequencing and flow cytometry, this study uncovers the immune cell landscape in healthy prostate and its dynamic changes during prostate cancer progression, identifying three resident macrophage populations and early immune alterations linked to tumor development (TAMs, TANs, Exhausted T cells).

Introduction

The prostate is an exocrine gland of the male reproductive system, composed of epithelial and few neuroendocrine cells supported by a fibroelastic stroma.^1–3 Prostate cancer (PCa) arises from the malignant transformation of prostatic luminal epithelial cells at adulthood. It is the most common visceral malignancy in men over 65 years old,⁴ and progresses over decades from prostatic intraepithelial neoplasia (PIN) to adenocarcinoma, eventually leading to castration-resistant metastatic prostate cancer. The human prostate has a lobular shape and is divided in three zones, whereas mouse prostate consists of four distinct lobes. Human prostate adenocarcinoma predominantly originates in the peripheral zone, which corresponds anatomically to the dorsolateral (DL) lobes in mice.

Epithelial and stromal cells have been studied in both human and mouse prostates, at steady-state and during PCa progression.^5–8 The immune microenvironment plays a pivotal role in tumorigenesis and recent studies have begun to delineate immune alterations associated with human prostate cancer, including the involvement of myeloid-derived suppressor cells (MDSCs), immunosuppressive macrophages and exhausted CD8⁺ T cells.^9–11 Notably, high blood neutrophil-to-lymphocyte ratio predicts shorter overall survival in patients with metastatic castration-resistant prostate cancer (mCRPC).¹²

Among genetic alterations implicated in PCa, the loss of PTEN (Phosphatase and TENsin homolog deleted on chromosome 10), occuring in 20% of primary and more than 40% of metastatic PCa, is the most extensively studied.^13,14 However, many PCa mouse models using PTEN deletion exhibit rapid PCa progression, failing to replicate the slow progression of human PCa.¹⁵ In these models, polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs) and pro-tumorigenic macrophages are commonly observed, particularly in aggressive PCa induced by probasin (Pb)-driven Cre recombinase.^16–20 Conversely, PTEN^(i)pe−/− mice more closely reproduce key features of human PCa.^21–24 Indeed, in this model, PTEN is selectively inactivated in prostatic luminal epithelial cells (PECs) at adulthood via the tamoxifen (Tam)-dependent Cre-ER^T2 system. This leads to PEC proliferation and PIN formation in DL prostate lobes within one month, followed by a latency phase characterized by cell senescence and a progression to adenocarcinoma occurring over approximately nine months.²¹ Extensive stromal reaction evidenced by extracellular matrix (ECM) remodeling and infiltration of Gr1⁺ MDSCs were observed during tumor progression,^21–24 but the immune cell composition in this model remained poorly characterized.

Tumor-associated macrophages (TAMs) are increasingly recognized for their critical role in tumor progression and dissemination, contributing to poor prognosis in various solid tumors.²⁵ Derived from monocytes recruited to tumor lesions, TAMs differ functionally and phenotypically from tissue-resident macrophages (RTMs) found in healthy tissues.²⁶ Some RTMs, such as perivascular macrophages (PVMs) and nerve-associated macrophages, are conserved across multiple tissues.^27,28 Additionally, Cx3cr1^hi ductal macrophages reside within epithelial layers of exocrine glands, including the mammary gland, pancreas, salivary gland, and testis.^29,30 However, RTM populations in the adult mouse prostate, both in healthy conditions or following tumor-induction in the PTEN^(i)pe−/− model, remain uncharacterized.

In this study, we conducted an in-depth analysis of immune infiltrates in healthy and cancerous mouse prostates. In PTEN^(i)pe−/− prostates, we observed a marked infiltration of tumor-associated Ly6G⁺ neutrophils (TANs), which represent the dominant immune cell population and resolved into eight distinct states, as well as Trem2⁺ tumor-associated-macrophages (TAMs), and PD-1^hiTim3⁺ exhausted CD8⁺ T cells. Trem2⁺ TAMs were distinct from the CD11b⁻, Tim4⁻ and Tim4⁺ RTM populations in the healthy prostate and exhibited immunosuppressive features possibly mediated by the MIF/HIF1A signaling axis.

Materials and methods

Mice

CD64^dtr, CCR2^−/−, CX3CR1^GFP/WT, Zbtb46^GFP and CSF1R^ΔFIRE mice were previously described. All mice are bred in CIML/CIPHE mouse facility platforms.^31–36 C57BL/6J (CD45.2) and C57BL/6J (CD45.1) were purchased from Janvier. Generation of Slco2b1-IRES-iCre-P2A-hCD2t mice (official name B6-Slco2b1^tm2Ciphe) is described below. PTEN^(i)pe−/− mice were generated as described.²¹ Briefly, mice carrying one copy of the PSA-Cre-ER^T2 transgene, expressing the Tam-inducible Cre-ER^T2 recombinase in prostatic luminal epithelial cells under the control of the human PSA promoter, were intercrossed with mice carrying LoxP-flanked (floxed) alleles of PTEN (L2 allele) to generate PSA-Cre-ER^T2(tg/0)/PTEN^L2/L2 (tg, transgenic) and PSA-Cre-ER^T2(0/0)/PTEN^L2/L2 (called PTEN^L2/L2 in the manuscript) mice. Gene ablation was induced by intraperitoneal injection of Tam for 5 days (1 mg/mouse/day) to 8- to 10-week-old mice, to generate the mutant PTEN^(i)pe−/− (pe, prostate epithelium; (i), induced) and the respective control PTEN^L2/L2 mice.

Mice were housed under specific pathogen-free conditions in temperature- and humidity-controlled animal facility with a 12 h light/dark cycle. Animals were euthanized with carbon dioxide or cervical dislocation, and tissues were immediately collected, weighed and processed for single-cell or histological analysis.

Generation of Slco2b1-IRES-iCre-P2A-hCd2t (Slco2b1^hCD2) mice

A targeting construct was designed to introduce an IRES-iCre-P2A-hCD2t cassette in the 3’ untranslated region of the Slco2b1 gene, 30 bp downstream of the stop codon. IRES corresponds to an internal ribosomal entry site, iCre to a sequence coding for a codon-improved Cre recombinase, P2A to the sequence coding for a self-cleaving 2A peptide, and hCD2t, to a sequence coding for the human CD2 (Uniprot-P06729). JM8.F6 C57BL/6N ES cells were electroporated with the targeting vector.³⁷ After selection in G418, ES cell clones were screened for homologous recombination by PCR and Southern blot. Properly recombined embryonic stem cells were injected into Balb/CNRj blastocysts. Following germline transmission, excision of the frt-neo^r-frt cassette was achieved through crossing with mice expressing the FLP recombinase (Tg(CAG-flpo)1Afst, EMMA-05149). Two pairs of primers were used to distinguish the WT and edited Slco2b1 alleles. The first pair (sense 5’-TCAGCAGACAGTCCTTACCC-3’ and antisense 5’- ACCGTCAACCTCCTGGAATC-3’) amplified a 682 bp band in case of the wild-type Slco2b1 allele, whereas the second pair (sense 5’-CTCATCATTGGCATATGTGG-3’ and antisense 5’- ACCGTCAACCTCCTGGAATC-3’) amplified a 295 bp band in the case of the Slco2b1- IRES-iCre-P2A-hCD2t allele.

Intravascular in vivo labeling

To distinguish circulating cells from non-circulating cells, mice were injected intravenously with 3 μg of anti CD45.2-FITC for 3 minutes before euthanasia. This procedure ensured the study of immune cells specifically located within the prostate parenchyma (CD45.2 negative), by exclusion of cells present in the vasculature (CD45.2 positive).

Cell extraction from prostates and lymphoid organs

To isolate cells from spleen and lymph nodes, these organs were processed as previously described.³⁸ Prostates were collected and weighted either as a whole or after the DLV (dorsal-lateral-ventral) lobe isolation. They were minced and incubated under regular manual mixing for 20 min at RT, in 2% FCS RPMI medium containing 140 µg/ml DNAse I (Sigma) and 1 mg/ml collagenase II (Worthington). The enzymatic reaction was stopped by adding 60 μL of 0.5 M EDTA. The cellular suspension was passed through a 5 ml syringe with 20 G needles for complete tissue dissociation. The single cell solution was filtered through a 70 μm cell strainer and centrifuged. The pellet was resuspended in FACS buffer (PBS-5 mM EDTA- 2% FCS) and was re-filtered through a 30 μm cell strainer (Miltenyi) solution prior to antibody staining.

Spleens were collected and dissociated by mechanical disruption in RPMI medium containing 2% FCS. Red blood cells were lysed using RBC lysis buffer (eBioscience) for 5 min at RT. Single cell suspensions were filtered through a 70 μm cell strainer, centrifuged and resuspended in FACS buffer prior to antibody staining.

Analysis of blood cells

Blood was collected in EDTA-containing tubes and stained with the appropriate antibody mix in PBS-5 mM EDTA solution for 30 min at RT at dark. Red blood cells were lysed with BD FACS Lysing Solution (Becton Dickinson) and the absolute numbers were estimated using CountBright^TM absolute count microbeads for flow cytometry (Invitrogen) and normalized by 1 μl of blood volume.

In vivo depletion of CD64⁺ cells using the CD64^dtr mice

CD64^dtr mice were injected intraperitoneally twice and 24 h apart with 1 μg DT (Calbiochem, EMD Millipore) in PBS as previously described.³¹

In vivo uptake of dextran

200 μM of Dextran-FITC with average mol wt 40,000 (fluorescent Isothiocyanate-dextran, 250 MG FD40-250 MG Merck/Sigma-Aldrich) were administered i.v. (100 μl). The recipient mice were sacrificed after 1 h and the blood and prostates were collected for flow cytometry (blood and prostate) or confocal microscopy (prostate).

In vivo treatments

Three months post tamoxifen, PTEN^(i)pe−/−mice were treated i.p. with a combination of 100 μg murinized anti-Ly-6 G (clone 1A8, Vivopurex, Cat# AB00295–2.0-VXL), and 500 μg of murinized rat anti-CSF1R (clone AFS98, Biolegend, Ultra-LEAF™ Purified anti-mouse CD115, Cat # 135542) and 100 μg of murinized rat anti-PD1 (clone RMP1–14, BioX Cell, Cat # BE0146) at days 0, 3, 5, 7 of treatment. Alternatively, mice were treated with Rat IgG2a, κ Isotype control (Biolegend, Ultra-LEAF™ Purified, Cat # 400566). Blood was collected prior to the treatment at d0, and at d8 (24 h after the last treatment injection), prostate and blood were collected for further FACS analysis of the immune compartment.

Competitive BM chimeras

Anesthetized 8-week-old male C57BL/6J (CD45.2) recipient mice were placed into a 6 mm thick lead cylinder to selectively expose their upper body to irradiation and protect their lower body including the prostate. Mice were then irradiated (8 Gy) and transplanted i.v. with 3 × 10⁷ BM cells. BM cells were obtained from femurs and tibias of male C57BL/6J (CD45.1).

Eight weeks after reconstitution, the level of reconstitution was determined by FACS analysis in blood and prostate. Chimeras were kept on antibiotic-containing water (0.2% Bactrim, Roche).

Ex vivo suppressive assay

CD8⁺ T cells were isolated from spleen of control PTEN^L2/L2 mice using the Dynabeads® Untouched™ Mouse CD8 Cells kit (Invitrogen, Cat 11417D) and following the manufacturer’s protocol. Isolated CD8⁺ T cells were stained with Cell Trace Violet (CTV) according to manufacturer’s instructions and were cultured (1 ×10⁴ CD8⁺ T cells/well) with activating a-CD3/CD28 beads at a stable ratio 1:1 (Dynabeads™ CD3/CD28 Mouse T Activator for T Cell Expansion and Activation, Invitrogen, Cat 11452D) in 96 well V-bottom plates in culture medium (RPMI-1640 supplemented with 10% fetal calf serum (FCS), 1% GM-CSF, 10% M-CSF, 1% penicillin/streptomycin, 10 mM HEPES, 1 mM glutamine, 1 mM non-essential amino-acids, and 50 μM 2-ME). Decreasing numbers of CD45⁺Ly-6 G⁺CD11b⁺ neutrophils or CD45⁺CD11b⁺CD64⁺CD88⁺ macrophages sorted from tumor prostates of PTEN^(i)pe−/−mice at 3 months post tamoxifen (PIN stage), were added to the 1 × 10⁴ CD8⁺ T cells/well to have the following T cell: neutrophil or macrophage ratios, 1:1, 2:1, 4:1, 8:1 and 16:1. Negative controls included non-activated CD8⁺ T cells, whereas positive controls were CD8⁺ T cells cultured only with activating beads. Cells were collected after 48 h and CD8⁺ T cell division was measured by assessing the relative CTV dilution by FACS.

Flow cytometry

For surface staining, single cell suspensions were preincubated with anti-Fc receptor antibody (clone 2.4G2) at 4°C for 15 min. Cells were centrifuged and stained with appropriate fluorescent labeled monoclonal antibodies (Supplementary Table S2) in Brilliant FACS Buffer (BD) at 4°C for 30 min.

Cell viability was determined using DAPI (4,’6-Diamidino-2-Phenylindole, Dihydrochloride) (Life Technologies) or Zombie UV^TM Fixable Viability (Biolegend). For intracytoplasmic staining, cells were fixed and permeabilized using a Cytofix/Cytoperm™ fixation-permeabilization kit (BD Biosciences). For cytokines staining, 1 × 10⁶ cells were activated for 4 h at 37°C in 0.5 ml of 1x Cell Stimulation Cocktail plus inhibitors (eBioscience).

Stained cell samples were analyzed on a LSR Fortessa or a FACSymphony Flow Cytometer equipped with FACSDiva software (BD Biosciences), and both instruments were validated prior to data acquisition using Flow Cytometry Calibration Particles (Spherotech, RQC-30-5A). Photomultiplier tube voltages were also adjusted to minimize fluorescence spillover. Single-stain controls were prepared with UltraComp eBeads (Thermo Fisher Scientific) following the manufacturer’s instructions and were used to calculate a compensation matrix. To determine the positive staining for a given marker, Fluorescent Minus One (FMO) controls were performed, consisting of the full staining antibody mix except one antibody. Debris were removed based on forward and side scatter, dead cells were excluded using live/dead staining and doublets were excluded by plotting FSC-A versus FSC-H. Leukocytes were subsequently selected by plotting CD45 versus FSC-A. Immune cell populations were identified using a multiplexed antibody panel targeting 25 cell surface markers specific to either myeloid or lymphoid lineages. The resulting data set was submitted to supervised analysis using either BD FACSDiva™ V9 Software or FlowJo™ V10.7 Software (BD Biosciences) or unsupervised analysis using OMIQ Software. Absolute cell counts were determined using CountBright^TM absolute count microbeads for flow cytometry (Invitrogen) and normalized to prostate tissue weight.

For cell sorting, prostate single-cell suspension was pre-enriched using CD45 MicroBeads following manufacturer’s protocol (Miltenyi, Cat 130–052–301) in the AutoMacs separator (Miltenyi). Subsequently, the enriched sample of CD45⁺ prostate cells was stained as described above and the populations of interest were sorted on a FACSAria SORP (BD Biosciences).

Cell morphology

Sorted cells were spun onto Shandon Cytoslides with a Thermo Shandon Cytospin 4 cytofuge (4 min at 400 g), fixed with methanol and stained with hematoxylin and eosin.

Confocal microscopy

The whole urogenital system was fixed in Antigen Fix (Microm Microtech) for 4 h, washed in 0.1 M phosphate buffer, then the prostate was isolated under binocular microscope and dehydrated overnight in 30% sucrose in 0.1 M phosphate buffer. Prostates were frozen in Tissue-Tek OCT compound (Electron Microscopy Sciences, Hartfield, PA) and 25-μm sections were permeabilized with 2% BSA Tris buffer. Unspecific binding was blocked with homemade buffer (Tris +2% BSA + triton 1% + 1% FCS + 1% serum) for 30 min and sections were stained with the indicated antibodies for cell identification (Supplementary Table 2) and hoechst for nuclei staining. Immunofluorescence confocal microscopy was performed using a Zeiss LSM 880 confocal microscope. Separate images were collected for each fluorochrome and merged to obtain a multicolor image. Final image processing was performed with Image J.

Data analysis

FACS data were analyzed using OMIQ (app.OMIQ.ai). First, the data were cleaned by manual gating to remove doublets, debris, and dead cells. Second, UMAP (Uniform Manifold Approximation and Projection) was calculated for all samples together, with subsampling to 100,000 CD45⁺ cells/sample. Third, PARC (Phenotyping by Accelerated Refined Community) was subsequently used to cluster the events based on UMAP parameters (30KNN).

Data visualization and statistical analysis were performed using the GraphPad Software Prism 10.

scRNA-seq

DLV lobes of mouse prostates were isolated and dissociated in single-cell suspension as described above. Prostate samples from three PTEN^(i)pe−/− mice at 3 months post tamoxifen were pooled. Five prostate samples from sex-matched PTEN^L2/L2 control littermates at 3 months post-tamoxifen were also pooled. The same approach was applied for the 9 months samples of PTEN^(i)pe−/− and control mice respectively. This experimental procedure was repeated twice in order to generated duplicates for each timepoint and sample type to eliminate possible batch effects during the analysis. Next, samples were FACS-sorted using a BD FACSAriaTM Fusion flow cytometer. Cell viability was determined using DAPI (4,’6-Diamidino-2-Phenylindole, Dihydrochloride). The control samples were sorted as bulk non-circulating immune cells (CD45.2⁻CD45⁺). However, for the tumorous PTEN^(i)pe−/− samples, the non-circulating neutrophils (CD45.2⁻CD45⁺CD11b⁺Ly-6 G⁺) were sorted independently from the rest of the immune cells (CD45.2⁻CD45⁺CD11b⁺Ly-6 G⁻). Post sorting, they were re-pooled in a ratio of 30% neutrophils to 70% of the remaining immune cells.

Trypan blue exclusion assay was used to determine cell number viability of each FACS-sorted sample using a Neubauer Chamber. Samples with a cell viability of > 95% were processed with a Chromium Controller (10X Genomics, Leiden, The Netherlands). Sixteen thousand cells were loaded per well in nanoliter-scale Gel Beads-in Emulsion (GEMs). Single-cell 3’ mRNA-seq libraries were generated according to “User guide Chromium Next GEM Single Cell 3 prime Reagent Kits v3.3 (Dual index) (10x Genomics, PN CG000315- Rev.D). Briefly, cells were partitioned into droplets with barcoded gel beads and reverse transcription master mix using Chromium Chip G. After complementary DNA (cDNA) synthesis and barcoding from poly-adenylated mRNA, GEMs were disrupted and pooled before amplification of cDNA by 11 polymerase chain reaction (PCR) cycles. Following enzymatic fragmentation and size selection of cDNA amplicons, sequencing libraries were constructed by adding Illumina P5 and P7 adapters (San Diego, USA) and i5 and i7 sample indexes by end repair, A-tailing, adaprot ligation, and 10 cycles of PCR amplification. Quality control and quantification of libraries was performed with a Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA). The generated libraries were sequenced on NextSeq2000 sequencer using a P3-type flow cell and 100-type cartridge.

Analysis of scRNA-seq datasets

The complete scRNA-seq dataset is composed of 16 datasets: 2 conditions ×2 time points (3 months/9 months) ×2 replicates (R1/R2) ×2 resequencing (Illumina RUN ID 144 and 188). To ensure a rigorous and traceable analysis, these datasets were analyzed through an iterative process.

i. Analysis of individual datasets : Each individual dataset was first independently analyzed to check for the cell quality and the cell heterogeneity using the following workflow. mRNA FASTQ raw files were processed using Cell Ranger v6.0.1 (10X genomics Inc.) software with default parameters to perform alignment, filtering, barcode counting, and unique molecular identifier (UMI) counting. Reads were aligned to the mouse GRCm38/mm10 genome. Datasets were analyzed using the R package Seurat 4.9.9. Quality control was performed to remove poor quality cells. Observing distribution of co-variables we fixed different thresholds for the number of UMIs (max 40,000), number of features (max 6000), percentage of mitochondrial genes (max 10%). We did not fix any threshold for the percentage of ribosomal genes since we observed that good quality neutrophils cells have a low content in ribosomal genes. Cells outside the QC threshold ranges were removed. Identification of putative multiplets was performed using scDblFinder (v 1.12).³⁹ However, since previous tests using multiple samples with HTO identification showed us that the result of automatic multiplet identification is not fully reliable, we decided to flag the identified multiplets but not to remove them at first sight. Instead, we traced them during the whole analysis and removed them only in the final datasets, when all evidence clearly identified them as multiplets.

Expression data were normalized using the NormalizeData function of the Seurat R package (logNormalize method and scale factor of 10,000). We centered the expression data from these factors using the Seurat R package ScaleData function (centering true and scaling false). Feature selection was performed using the FindVariableFeatures Seurat R package function with the vst algorithm and 2000 genes selected. Principal component analysis (PCA) was run using Seurat RunPCA function on those 2000 most variable genes. UMAP dimensionality reduction was run using Seurat RunUMAP function on the first 30 principal components from PCA. Clustering was done with Seurat FindClusters based on the first 30 principal components from PCA dimensionality reduction. Best resolution for clustering was determined using the treeMap package (v2.4–4) and the analysis of the cluster’s marker genes. Marker genes for each cluster were identified using Wilcoxon rank-sum test from Seurat FindAllMarkers with the log fold change threshold set to 0.1 and an adjusted FDR threshold at 0.001. Cell-cycle scores and cell phases were calculated using CellCycleScoring function with s.genes and g2m.genes from the Seurat package.

ii. Analysis of merged datasets by time points : The pairs of re-sequenced samples were then combined together using 10x Genomics Cell Ranger in a single analysis to obtain a complete dataset for each combination of condition, time points and replicate. We then used the obtained count matrices to create merged datasets. We first combined separately the two conditions and two replicates datasets for each time point to obtain one dataset per time point. A complete analysis similar to the one performed for individual dataset was applied (for QC, the main filter was on the percentage of mitochondrial genes set to 10% for both datasets, cluster resolution was set to 0.1 and 0.4 for both datasets). After QC, we obtained 27,535 and 8,674 cells, respectively, for the 3-months and 9-month datasets. We observed that, in both datasets, the cells from the two conditions overlapped very well, and we decided not to apply any integration algorithm. Heatmap of expression of genes across clusters was performed using the Heatmap function of the ComplexHeatmap (v2.10.0) R package.⁴⁰ We then made an automatic cell type annotation using Azimuth (v 0.4.6). Since no reference specific to prostate was available, we used the human PBMC reference as proxy and convert the human genes to mouse genes using Babelgene (v 22.9). Cell were assigned to cell types using 3 levels of annotation. Cell type annotation was then refined through a manual curation of marker genes of clusters in order to identify specific subtypes and cell states.

Using this annotation, we defined four main cell types: B cells, T/NK cells, myeloids and neutrophils. Each of these groups was then extracted to create a separate dataset that was analyzed using a similar workflow as for the previous dataset. The new clusters obtained from the separate datasets were assigned to cell types using the newly identified marker genes. The differentially expressed genes (DEG) inside each cluster between cells of the two conditions were identified using the FindMarkers function of the Seurat package. Functional enrichment analysis of the DEG was performed using ClusterProfiler (v 4.6.2).

iii. Analysis of global merged dataset : The analysis of the time points datasets allowed us to have a good understanding of the data at each time point. We decided to merge the dataset of all conditions, time points and replicates to generate a single complete dataset. A complete analysis similar to the one performed for individual dataset was applied (for QC, thresholds were number of UMIs max 70,000, number of features max 8000, percentage of mitochondrial genes max 10%, cluster resolution was set to 0.1 and 0.4). After QC, we obtained 36,175 cells. We observed that, in both datasets, the cells from the two conditions and two time points overlapped very well, and we decided not to apply any integration algorithm. As before, Azimuth was used to roughly identify cell types and cell type annotation was refined through manual curation of identified marker genes of clusters.

Here again, using this annotation, we defined four main cell types: B cells, T/NK cells, myeloids, and neutrophils. Each of these groups was then extracted to create a separate dataset that was analyzed using a similar workflow as for the previous dataset. Moreover, thanks to the increase in cell number, we were able to identify a few cell clusters that were identified as contaminants (Stroma, epithelial, blood vessel, adipocytes and basal cells). Finally, automatic multiplets identification allowed to flag cells that we identified as true multiplets considering gene expression and spatial position in UMAP. Contaminant and multiplets were removed from analysis.

The new clusters obtained from the separate datasets were assigned to cell type using the newly identified marker genes. The differentially expressed genes (DEG) inside each cluster between cells of the two conditions were identified using the FindMarkers function of the Seurat package. Functional enrichment analysis of the DEG was performed using ClusterProfiler (v 4.6.2). Statistical analysis of comparison of gene expression among time points or conditions through clusters was performed by a Kruskal-Wallis test followed by Dunn post-hoc tests. In each dataset, differential composition and variability analysis of clusters, respect to change in condition, and time points was performed using sccomp (v 1.5.6).⁴¹

Cell trajectories analysis was first performed using Monocle 3.⁴² Cells were clustered using the Monocle 3 cluster_cells function with a resolution set to 0.003. Trajectories were computed using the Monocle 3 learn_graph and order_cells function. Scores of gene sets were computed using the AddModuleScore method of the Seurat package and plotted along the pseudotime. Since several datasets showed unconnected clusters, we split the dataset in several cluster groups for which valid trajectory hypotheses were possible. Analysis of gene expression or gene module score along pseudotime was performed using a LOESS regression with the ggplot package.

To confirm trajectories identified by Monocle, we also applied cell dynamic inference using RNA velocity method with velocyto.R (v 0.6). Maps of cell velocity vectors were compared to the trajectories from Monocle and divergent conclusions were not considered for analysis.

Human datasets analyses

The Trem2 TAM and exhausted-like CD8⁺ T cell signature were generated by selecting the top 10 globally distinguishing genes of the Trem2⁺ TAM cluster (PMN-Cluster 4) and CD8⁺_T2 (T/NK-Cluster3) in our single-cell RNA sequencing dataset, respectively. The expression of the signature was detected in the single-cell RNA sequencing data of human intraductal cribriform PTEN-deficient tumors and adjacent benign tissue generated by Wong et al.¹⁰

The correlation of the signatures with the Gleason score and the Disease-Free Survival were analyzed in the TCGA (The Cancer Genome Atlas) PRAD (Prostate Adenocarcinoma) cohort (n = 496 patients with prostate cancer). The survival curve was generated using GEPIA2 tool.⁴³

Results

CD11b⁻, Tim4⁻ and Tim4⁺ macrophage subsets predominate among the mononuclear phagocytes residing in healthy mouse prostate

To characterize the immune landscape of the mouse prostate, we first examined the complexity of the mononuclear phagocytes (MNPs), encompassing monocytes, conventional dendritic cells (cDCs) and macrophages, using flow cytometry with a comprehensive panel of cell surface markers. Non-circulating myeloid cells residing in healthy prostate were separated from blood contaminant using intravascular in vivo labeling, then pre-gated as CD45⁺ non-B-T-NK-Neutrophils-Eosinophils (Supplementary Figure S1a-b) and gated into CD11b⁻ and CD11b⁺ populations (Figure 1(a)). The CD11b⁻ population was further analyzed using XCR1 versus CD11c to identify XCR1⁺CD11c⁺ cells corresponding to cDC1 (Figure 1(a)). Within the XCR1⁻CD11b⁻ cells, we identified CD11c⁺MHCII⁺ cells that also expressed the macrophage marker F4/80, and were denoted as CD11b⁻ macrophages (CD11b⁻ Mf) (Figure 1(a)). Conversely, the CD11b⁺ population was further analyzed using CD64 (FcγRI) versus CD88 (C5aR1).^44,45 The CD64⁺CD88⁺ macrophages were further subdivided into Tim4⁻ and Tim4⁺ macrophage subsets (Figure 1(a)). The CD64⁻CD88⁻ cells were further analyzed using MHCII versus Ly-6C, allowing us to identify Ly-6C⁺MHCII⁻ cells (P1) and Ly-6C⁺MHCII⁺ cells (P2), corresponding to classical and activated monocytes, respectively, as well as Ly-6C⁻MHCII⁺ cells (P3) (Figure 1(a)).⁴⁵

Figure 1.

Unraveling the myeloid cell heterogeneity in steady-state mouse prostate.

(a) Myeloid gating strategy of flow cytometry analysis in adult (10 weeks old) C57BL/6J prostates. Following exclusion of doublets, dead cells, blood contaminants (CD45.2⁺), T, B, NK cells, neutrophils, and eosinophils (see Supplementary Figure S1a), non-circulating CD45⁺ immune cells were divided into CD11b⁻ and CD11b⁺ cells. CD11b⁻ cells contained XCR1⁺CD11c⁺ cDC1 cells and within the XCR1⁻ non-cDC1 gate, the CD11c⁺MHC-II⁺ population was further gated as CD64⁺F4/80⁺ cells, identified as CD11b⁻ macrophages. CD11b⁺ cells were plotted to identify the CD11b⁺CD64⁺CD88⁺ macrophages, that were further separated into Tim4⁺ and Tim4⁻ macrophages using the Tim4 and MHC-II markers. From the CD64⁻CD88⁻ non-macrophage gate, the expression of Ly-6C and MHC-II allowed to discriminate three populations, including P1 classical monocytes (Ly-6C^HighMHC-II⁻), P2 non-classical/activated monocytes (Ly-6C^{int to low}MHC-II⁺), and P3 population (Ly-6C⁻MHC-II⁺).

(b) Distribution of MNP subpopulations, described in Figure 1(a) and Supplementary Figure S1a, in the adult steady-state prostate (10-week-old C57BL/6J) based on their frequency.

(c) Counts of prostate macrophage subsets in adult steady-state mice (10 weeks old C57BL/6J) treated with rat IgG2a isotype control or anti-CSF1R antibody. Data are presented as mean ±SEM, are representative of 2 independent experiments with 4 animals/group. Statistically significant difference was calculated using Ordinary one-way ANOVA followed by multiple comparisons test; *, p < 0.05; **, p < 0.01; ****, p < 0.0001.

(d-f) Qualitative confocal microscopy analysis of prostate tissue sections from adult 3 months old CX3CR1^GFP/WT mice, with representative staining from the dorso-lateral (DL) prostate lobes (scale bar 50 μm). (d) Staining with Phalloidin (blue), CD326/EpCAM (magenta), Iba1 (white) and CX3CR1-GFP (green) at 40x magnification. White arrows indicate the CX3CR1^−/lowIba1⁺ macrophages within the stroma and yellow arrows indicate the CX3CR1^highIba1⁺ intra-epithelial macrophages. (e) Staining with Phalloidin (blue), CD326 [EpCAM] (magenta) and CX3CR1-GFP (green) at 40x magnification (top), CX3CR1-GFP (green) at 40x magnification (bottom) (f) Staining with Phalloidin (blue), Iba1 (white), CX3CR1-GFP (green) and Tim4 (red) at 40x magnification. Red arrows indicate the CX3CR1⁻Iba1⁺Tim4⁺ macrophages and white arrows indicate the CX3CR1^−/lowIba1⁺Tim4⁻ macrophages located in the stroma.

(g) Qualitative confocal microscopy analysis showing the staining of Hoechst-Phalloidin (blue), CD31⁺ blood vessels (red), Iba1⁺ macrophages (white), and Tim4⁺ macrophages (green) in prostate tissue sections from adult Csf1r^ΔFIRE/+ and Csf1r^{ΔFIRE/ΔFIRE} mice (10 weeks old). Representative staining from DL prostate lobes (scale bar 50 μm).

(h) Counts of prostate macrophage subsets in adult (10 weeks old) Csf1r^ΔFIRE/+ and Csf1r^{ΔFIRE/ΔFIRE} mice. Data are presented as mean ±SEM and are a cumulative representation of 2 independent FACS experiments with 2-4 animals per group. Statistically significant difference was calculated using Ordinary one-way ANOVA followed by multiple comparisons test; ***, p < 0.001; ****, p < 0.0001.

(i) In vivo dextran-FITC uptake by the different macrophage subsets in the prostate of adult steady-state mice (9 weeks old C57BL/6J) from treated (green) and non-treated (gr-ey) mice. Data are representative of 3 independent experiments with 3–5 animals per group.

(j) Qualitative confocal microscopy analysis of prostate tissue sections from adult steady-state mice (10 weeks old C57BL/6J). Top: staining with Hoechst-Phalloidin (blue), CD31⁺ blood vessels (red), Iba1⁺ macrophages (white) and dextran-FITC (green), at 20x and 40x magnification. Bottom: staining with Hoechst-Phalloidin (blue), Tim4⁺ macrophages (red), Iba1⁺ macrophages (white) and dextran-FITC (green) at 20x and 40x magnification. Representative staining from the DL prostate lobes (scale bar 50 μm).

(k) Graphical abstract of protected BM chimera: Adult CD45.2 mice were placed in shielded conditions to protect the prostate and hindlimbs during irradiation and were transferred intravenously with BM-derived CD45.1⁺ donor cells.

(l) BM chimerism is presented as the frequency of host (CD45.2⁺) and donor (CD45.1⁺) cells for blood and prostate monocytes, Tim4⁻, Tim4+ and CD11b⁻ prostate macrophages in protected BM chimeras, 8 weeks post reconstitution. Data are presented as mean ±SEM of n = 3-5 and are representative of 3 independent experiments.

(m) Absolute cell numbers of Tim4⁻, Tim4+ and CD11b⁻ macrophages in prostates of 10 weeks old CCR2^−/− and WT mice. Data are presented as mean ± SEM of n = 3-8, and statistical difference was calculated using Kruskal-Wallis test with Dunn’s multiple comparison. Data are representative of 2 independent experiments.

(n) Quantification of the replenishment kinetics of Tim4⁻ (orange), Tim4⁺ (pink), and CD11b⁻ (green) macrophage subsets in CD64^dtr prostates (based on gating strategy on FACS plots in Supplementary Figure S1j) showing untreated (-DT) and 1, 3, 7, 10, 14, 30, 45 and 60 days (d) after the last DT injection. Data are presented as mean ±SEM of n = 3-14 and are a cumulative representation of 4 independent experiments with 3-5 animals per group. Statistically significant difference was calculated using Kruskal-Wallis one-way ANOVA followed by Dunn’s multiple comparisons test; *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001

The cDC1 identity of XCR1⁺CD11c⁺ cells was confirmed by their expression of the transcription factor Zbtb46 (Supplementary Figure S1c).³⁴ Moreover, a fraction of the Ly-6C⁻MHCII⁺ cells (P3) subset expressed CD11c, Zbtb46 and the Dendritic Cell-Specific Intercellular adhesion molecule 3 (ICAM-3)-Grabbing Nonintegrin (DC-SIGN) (Supplementary Figure S1d), suggesting they correspond to cDC2. The DC-SIGN⁻CD11c⁺ cells within P3 were heterogeneous and divided in Zbtb46⁺ cDC2 and Lyz⁺ cells (Supplementary Figure S1d). The macrophage identity of CD11b⁻, Tim4⁻ and Tim4⁺ subsets was confirmed by their expression of the solute carrier organic anion transporter family member 2B1 (Slco2b1) and F4/80 (Supplementary Figure S1c) and their dependency on the colony-stimulating factor 1 (CSF1) (Figure 1(c)).^27,45–48 Additionally, Tim4⁺ Mf express high levels of Lyve-1, FolR2, and CD206, similar to the perivascular macrophages (PVMs) described in the literature (Supplementary Figure S1c).²⁷ A fraction of Tim4⁻ Mf express CX3CR1,³³ Lyve-1, and FolR2, and CD11b⁻ Mf express high levels of CX3CR1 (Supplementary Figure S1c). The CD11b⁻, Tim4⁻, and Tim4⁺ macrophage subsets are the dominant myeloid cells (76%) found in steady-state prostate tissue of adult mice (Figure 1(b)). Altogether, our flow cytometry analysis showed that in the prostate, MNPs consist of cDC1, cDC2, monocytes, Lyz⁺ cells, and three macrophage subsets, CD11b⁻, Tim4⁻, and Tim4⁺.

Prostate-resident CD11b^– and Tim4⁺ macrophage subsets are associated to the epithelium and the vasculature respectively, while Tim4^– macrophages are located in the stroma

Imaging of DLP sections from CX3CR1^GFP/+ mice revealed the presence of CX3CR1^highIba1⁺ macrophages positioned at the basal side of CD326⁺ epithelial cells, beneath the basement membrane (Figure 1(d) and Supplementary Figure S1e). CX3CR1^highIba1⁺ intra-epithelial macrophages are CD11b⁻ (Supplementary Figure S1f) and harbor extended dendrites forming a network within the duct epithelium (Figure 1(e)). CX3CR1^−/lowIba1⁺ macrophages were observed in the DLP stroma (Figure 1(d) and Supplementary Figure S1e) and could be further divided into CX3CR1⁻Iba1⁺Tim4⁺ and CX3CR1^−/lowIba1⁺Tim4⁻ macrophages (Figure 1(f) and Supplementary Figure S1g). Iba1⁺Tim4⁺ Mfs are localized near CD31⁺ blood vessels of the prostate (Figure 1(g)). Perivascular macrophage (PVM) survival is independent of the super-enhancer fms-intronic regulatory element (FIRE) within the Csf1r locus.³⁶ Consistent with this, Iba1⁺Tim4⁺ Mfs are not depleted in Csf1R^{ΔFIRE/ΔFIRE} prostate (Figure 1(g–h)). To confirm the identity of Tim4⁺ macrophages as PVMs, we performed an in vivo scavenging assay. Both FACS and confocal microscopy analyses demonstrated that perivascular Tim4⁺ Mfs efficiently internalized intravenously injected FITC-labeled dextran (Figure 1(i,j)). This activity correlated with their high expression of the mannose receptor CD206 (Supplementary Figure S1c). Therefore, the healthy mouse prostate harbors three major resident macrophage populations. Tim4^– macrophages are located in the stroma, while CD11b^– and Tim4⁺ macrophage subsets form distinct network within the epithelium and around the vasculature, respectively.

Prostate-resident macrophages subsets exhibit a low turn-over and CD11b^– macrophages harbor a protracted reconstitution kinetic after depletion

To assess the turnover the three distinct populations of prostate-resident macrophages and investigate the contribution of circulating monocytes to their renewal, we generated protected bone marrow (BM) chimeric mice (Figure 1(k)). The frequency of CD45.1 donor cells of monocytes infiltrated in prostate (50%) and Tim4^– Mf (30%) was in the same range than monocytes from circulation (37%), while Tim4⁺ and CD11b^– Mf remain predominantly of host origin (5% and 1.8% of donor cells respectively) (Figure 1(l) and Supplementary Figure S1h). To address the monocyte-origin of prostate macrophages, we used Ccr2^–/– mice,⁴⁹ as CCR2 is essential for monocytes to exit the bone marrow and migrate to peripheral tissues. Monocytes infiltrating the prostate were significantly reduced akin to blood monocytes (Supplementary Figure S1i), whereas the number of Tim4^– Mfs was slightly reduced (not significant) and the number of Tim4⁺ and CD11b^– Mfs remain unaffected (Figure 1(m)). Altogether, these results suggested that prostate-resident macrophage populations exhibit partial (Tim4^– Mf) to low (Tim4⁺ and CD11b^– Mf) input from adult monopoiesis in their long-term maintenance in healthy conditions.

Additionally, we used CD64^dtr mice to induce diphtheria toxin (DT)-mediated ablation of tissue resident macrophages, monocytes, and monocyte-derived cells.³¹ The three macrophage populations of the prostate were ablated after DT injection and their reconstitution patterns differed significantly (Supplementary Figure S1j). Notably, higher numbers of Tim4^– Mfs were detected at day 3 post-DT (Figure 1(n)), suggesting a rebound of cells and increased myelopoiesis following DT treatment, as previously observed in circulating monocytes.³¹ Interestingly, while the complete replenishment of Tim4^– and Tim4⁺ Mfs was observed at day 7 post-DT treatment, CD11b^– Mfs exhibited no rebound and demonstrated a protracted reconstitution kinetics, achieving full restoration by day 45 post-DT treatment (Figure 1(n)). Therefore, these findings indicate that, monocyte-derived differentiation supports the reconstitution of Tim4^– and Tim4⁺ Mfs following DT-mediated macrophage depletion, while in situ proliferation likely contributes to the generation of CD11b^– Mfs.

This first part focusing on the MNPs in the healthy prostate of adult mice revealed a great heterogeneity of cDCs (cDC1, cDC2), monocytes (P1, P2, P3) and RTM subsets (Tim4^–, Tim4⁺, and CD11b^– Mfs). Consistent with RTMs in other tissues, the prostate macrophage subsets exhibited low turn-over rates and distinct spatial distribution within the tissue.

Macrophages and neutrophils are recruited in PIN-containing prostates

Following the detailed characterization of the myeloid immune cell composition in the healthy prostate, we next aimed to investigate the alterations in the immune cell compartment in tumors of PTEN^(i)pe−/− mice. Following PTEN inactivation, prostatic glands exhibited an increase of non-circulating CD45⁺ immune cell infiltration (Supplementary Figure S2a). At PIN and adenocarcinoma stages, respectively 3 and 9 months after gene inactivation, the predominant infiltrating immune cells were Ly6G⁺ neutrophils, previously characterized as Gr1⁺ MDSCs,²² and F4/80⁺ macrophages (Figure 2(a–c)), which accumulate in the lumen and tumorous epithelium, and in the stroma respectively (Figure 2(d)). Consistent with a selective tumor development in the dorsolateral and ventral (DLV) lobes,²¹ the infiltration of CD45⁺ immune cells, and more specifically neutrophils and macrophages, was observed in the DLV lobes, but not in the anterior prostate lobes (Supplementary Figure S2b). To further dissect the changes of the immune cells in prostate, we performed droplet-based single-cell RNA-sequencing (scRNA-seq) on FACS-sorted infiltrating CD45⁺, extracted from DLV prostate of PTEN^(i)pe−/− mice at PIN and adenocarcinoma stages and of age-matched control mice (Figure 2(e) and Supplementary Figure S2c). Unsupervised analysis showed that the major immune cell lineages consisted of four MNP clusters (Clusters 0, 3, 8, 11), three T cells and NK cell clusters (Clusters 1, 4, 7), three B lymphocyte clusters (Clusters 5, 6, 12), and one neutrophil cluster (Cluster 2) (Figure 2(f), Supplementary Figure S2d-e and Supplementary Table S1).¹¹ We also identified three non-immune clusters (Clusters 9, 10, 13) corresponding, respectively, to epithelial, endothelial, and stromal cells, as few contaminants of the sorted immune cells (Figure 2(f), Supplementary Figure S2d-e and Supplementary Table S1).¹¹ DEG analysis comparing PTEN^(i)pe−/− mice over the controls revealed that most of the tumor-associated-immune cells harbor an enrichment for leukocyte chemotaxis and migration (Figure 2(g)). This suggested that following PTEN inactivation in prostate epithelial cells, both immune and non-immune cells contribute to promote the recruitment of immune cells in the prostate tumor microenvironment, notably neutrophils and MNPs. In summary, this comprehensive analysis revealed that PTEN^(i)pe−/− prostates exhibit significant alterations in the immune infiltrate beginning at the PIN stage.

Figure 2.

Deciphering the CD45⁺ prostate immune cells in PTEN^(i)pe–/– and control mice at 3- and 9-months post tamoxifen administration.

(a) Unsupervised flow cytometry analysis of control (PTEN^L2/L2) (top) and tumor bearing PTEN^(i)pe–/– (bottom) prostates at 3 months post-tamoxifen administration (PIN stage), showing the expression of surface markers used for the unsupervised FACS analysis and immune cell population annotation.

(b) Frequencies of neutrophils, T-NK cells, B cells, F4/80⁺ cells and F4/80^– cells in the prostates of control and PTEN^(i)pe–/– mice at 3 months post-tamoxifen administration.

(c) Counts of neutrophils, T-NK cells, B cells, F4/80⁺ cells and F4/80^– cells in the prostates of control and PTEN^(i)pe−/− mice, at 3- and 9-months post-tamoxifen and after normalization per mg of prostate weight. Dot size represents the cell count. Data are presented as mean ±SEM of n = 3-6 and are representative of 5 independent experiments per time point with 3-6 animals/group. Statistically significant differences were calculated using two-way ANOVA followed by Sidak’s multiple comparisons test; *, p < 0.05; ****, p < 0.0001.

(d) Confocal microscopy analysis showing Hoechst (blue), CD3⁺ T cells (red), Iba1⁺ macrophages (white), and Ly-6G⁺ neutrophils (green) in prostate tissue sections from control and tumor bearing PTEN^{(i)pe-–/–} mice, at 3- and 9-months post-tamoxifen. Representative staining from DL prostate lobes (scale bar 50 μm).

(e) Strategy for scRNA-seq analysis as described in the materials and methods.

(f) Integrative analysis of scRNA-seq of prostate immune cells for control (left) and PTEN^(i)pe–/– (right) samples comprising 2 replicates at 3- and 9-months post-tamoxifen for each genotype. The color-coded cluster annotation mentioned is based on the top marker genes as shown in Supplementary Figure S2d.

(g) Functional enrichment of the different immune clusters as identified in (f) based on the top differential gene expression (DEG) of PTEN^(i)pe−/− versus control samples at 3 months timepoint for selected Gene Ontology (GO) Biological Process (BP) terms. Dot color represents the adjusted p-value of the test and dot size represents the ratio of DEG present in the GO BP term.

Trem2⁺ tumor-associated-macrophages accumulate in PINs of PTEN^(i)pe−/− mice

To dissect further the changes within the MNP compartment between healthy prostate and prostate tumors, the four MNPs clusters (Clusters 0, 3, 8, 11) were pooled and subjected independently to unsupervised computation that revealed 18 distinct clusters (Figure 3(a) and Supplementary Table S1). The analysis of the top differentially expressed genes by each cluster and module scoring from published data (Supplementary Figure S3a and Supplementary Figure S3b),^{27,29,50–52} confirmed the presence of cDC1 (Xcr1, Clec9a, Wdfy4) divided in three subclusters (Clusters 7, 14, 16), cDC2 (Cd209a, Cluster 2), monocyte (Plac8, Ly-6C2, Ccr2, Cluster 6), Lyz1⁺ cells (Lyz1, Ccr2, Ear2, Fn1, Cluster 5), Tim4⁺ Mf (Ccl8, Lyve1, FolR2, Timd4, Mrc1, Cluster 3) and CD11b^– Mf (Cx3cr1, Mmp12, Ccl4, Cluster 0) (Figure 3(a)). The identity of cluster 0 as CD11b^– Mf was confirmed by its low expression of Itgam (CD11b) and high level of Itgax (CD11c) (Supplementary Figure S3c). The scRNA-seq dataset also permitted the detection of rare cells, including mast cells (Cma1, Tpsb2, Fcer1a, Cluster 17),⁵³ plasmacytoid dendritic cells (pDC, Grm8, Ly6d, Siglech, Plac8, Ly-6C2, Cluster 18), mature DC cluster (Ccr7⁺ DC, Ccr7, Fscn1, Mreg, Cluster 13),^54,55 and minor macrophage clusters (Clusters 8, 10, 11, 12, 15).

Figure 3.

Trem2⁺ tumor associated macrophages infiltrated the cancerous prostates.

(a) Independent integrative analysis of scRNA-seq of mononuclear phagocyte clusters (0, 3, 8, 11) of prostates from the general UMAP shown in Figure 2(f), for control (left) and PTEN^(i)pe−/− (right) samples encompassing 2 replicates of 3 months timepoint for each genotype. Cell type annotation of the color-coded clusters was performed based on the marker genes of the new clusters.

(b) Fractional composition for control and PTEN^(i)pe−/− samples of 3-month timepoint of the MNP color-coded clusters described in (a). Significance between the conditions for the different clusters is indicated by an asterisk, based on the statistical analysis presented in Supplementary Figure S3d.

(c) Average expression level of selected marker genes among the top 10 for the myeloid clusters of concatenated control and PTEN^(i)pe−/− samples at 3 months timepoint, as annotated in (a). The color of the dots represents average of normalized and scaled expression of marker genes in each cell type, and their size indicates the percent of cells expressing the marker genes in the cluster.

(d) Quantification of the different macrophage populations from control and PTEN^(i)pe−/− DLV prostates at PIN stage (3 months post Tam), based on the gating strategy shown in Figure 1(a). Counts are normalized per mg of prostate. Data are represented as mean ±SEM of n = 3 and are representative of 5 independent experiments involving 3-6 animals per group. Statistical significance was calculated using mixed effects model followed by Sidak’s multiple comparisons test; ****, p < 0.0001.

(e) Gating strategy identifying Tim4⁻ macrophage subsets based on the expression of Trem2 and FolR2 from control and PTEN^(i)pe−/− prostates at PIN stage (3 months post Tam).

(f) Quantification of the different Tim4⁻ macrophage subsets (Trem2⁻FolR2⁺, Trem2⁺FolR2⁺, Trem2⁺FolR2⁻ and Trem2⁻FolR2⁻) from control and PTEN^(i)pe−/− DLV prostates at 3 months post tamoxifen, based on the gating strategy shown in (e). Counts are normalized per mg of prostate. Data are represented as mean ±SEM of n = 3 animals and they are representative of 2 independent experiments. Statistical significance was calculated using ordinary two-way Anova followed by Sidak’s multiple comparisons test; ***, p < 0.001; ****, p < 0.0001.

(g) Pseudotime differentiation trajectories of the indicated monocyte and macrophage clusters of PTEN^(i)pe−/− samples at PIN stage (3 months post Tam), predicted using Monocle3. The indicated trajectories are presented in (Supplementary Figure S3i).

(h) Average expression level of selected marker genes for the myeloid clusters of PTEN^(i)pe−/− samples at 3-month timepoint, as annotated in (a). The color represents the average of normalized and scaled expression of marker genes in each cell type, and the size indicates the percent of cells expressing the marker genes.

(i) Functional enrichment of cluster 1 (Tim4⁻FolR2⁺ Mac) and cluster 4 (Trem2⁺ TAM) based on the expression of top marker genes of PTEN^(i)pe−/− prostates at 3 months timepoint, for pathways from the Kyoto Encyclopedia of Genes and Genomes (KEGG) (left) and top 10 Gene Ontology (GO) Biological Process (BP) terms (right). Dot color represents the adjusted p-value of the test and dot size represents the count of marker genes present in the enriched functions.

Interestingly, in PTEN^(i)pe−/− mice, we observed the emergence of a tumor-associated-macrophage (TAM) subset (cluster 4) expressing the triggering receptor expressed on myeloid cells 2 (Trem2) and elevated Secreted Phosphoprotein 1 (SPP1) transcript levels (Trem2⁺ TAM), together with a significant increase of Tim4^–FolR2⁺ Mf (cluster 1) as well as monocytes (cluster 6) and cycling mac (cluster 8) (Figure 3(a,b) and Supplementary Figure S3d). Analysis of the top genes differentially expressed by Trem2⁺ TAM (Spp1, Gpnmb, Cxcl10, Rsad2, Bglap) and module scoring from published data,^56–58 suggested that Trem2⁺ TAMs correspond to the immunosuppressive TAMs described in other tumor microenvironments (Figure 3c and Supplementary Figure S3e).^56,59,60 Based on these findings, we investigated whether the Trem2⁺ TAM signature could predict disease progression in human prostate cancer. The dataset published by Wong et al. included paired benign and cancerous prostate tissues obtained from patients who underwent radical prostatectomy.¹⁰ Analysis of this dataset revealed that the Trem2⁺ TAM signature score was higher in tumor tissues compared to benign prostate tissues and was predominantly detected in myeloid cells (Supplementary Figure S3f). Additionally, analysis of the TCGA PRAD cohort demonstrated that prostate tumors with a Gleason score of 9–10 exhibited a significantly higher Trem2⁺ TAM signature score than tumors with lower Gleason scores (Supplementary Figure S3g). Moreover, patients with high Trem2⁺ TAM signature expression showed reduced disease-free survival (Supplementary Figure S3g). We next showed by FACS that Tim4^– Mfs were increased in PTEN^(i)pe−/− prostate at PIN stage by 8-fold (Figure 3(d)), while Tim4⁺ and CD11b^– Mfs remained unchanged. FACS analysis of Tim4^– Mf using FolR2 and Trem2 markers further confirmed the massive infiltration of PTEN^(i)pe−/− prostate at PIN stage by Trem2⁻FolR2⁺ Mf, Trem2⁻FolR2⁻ Mf and Trem2⁺ TAM (Figure 3(e,f)). This infiltration pattern persisted and intensified at the adenocarcimoma stage (Supplementary Figure S3h) . Cell trajectory analysis indicated that Tim4^–FolR2⁺ Mf (cluster 1) originated from monocytes (cluster 6), while Trem2⁺ TAM (cluster 4) and Trem2⁺ Mf (cluster 9) derived from cycling mac (cluster 8) (Figure 3(g) and Supplementary Figure S3i). Collectively, these findings indicate that PTEN^(i)pe−/− prostates are characterized by a pronounced infiltration of Tim4^– Mfs, predominantly composed of Trem2⁺ TAMs that are found in advanced human prostate cancer patients with poor prognosis.

Trem2⁺ TAMs in PTEN^(i)pe−/− prostate harbor a strong protumoral signature

Differential gene expression analysis (DEG) revealed an enrichment for extracellular matrix (ECM) remodeling with high expression levels of various cathepsins, including Ctsd and Ctss, matrix metalloproteinase Mmp9 and hypoxia-induced factor Hif1a and Hif1a-target genes (Eno1, Gapdh, Ldha) in MNPs present in tumors of PTEN^(i)pe−/− mice compared to those in healthy prostate (Figure 3(h)). Moreover, Trem2⁺ TAMs express high level of the Mif gene, encoding for the macrophage migration inhibitory factor, involved in various protumoral pathways,^61,62 and harbor a strong metabolic signature (Apoe, Eno1, Gapdh, Ldha) (Figure 3(h)). KEGG pathway database interrogation and Gene Ontology (GO) analysis indicated an enrichment for ROS-mediated carcinogenesis, HIF-1 signaling pathway and metabolic processes, suggesting protumoral function in Trem2⁺ TAM and an enrichment of endocytosis and autophagy for Tim4⁻ FolR2⁺ Mf in PTEN^(i)pe−/− prostate at PIN stage (Figure 3(i)). Therefore, these data suggest that PIN-containing PTEN^(i)pe−/− prostates are infiltrated by protumoral Trem2⁺ TAMs bearing an immunosuppressive signature.

Mature PD-L1⁺ neutrophils massively invade PIN lesions and exhibit immunosuppressive function

Next, we extended our analysis to tumor-associated neutrophils (TANs), which were frequently identified as Gr1⁺ PMN-MDSCs.^22,63,64 In PTEN^(i)pe−/− prostate, neutrophil prevalence started to be observed from the PIN stage, and further showed a sustained recruitment into late adenocarcinoma (Figure 2(c) and Supplementary Figure S4a). Hematoxylin-eosin stained cytospin revealed that the prostate-tumor-infiltrating neutrophils exhibited multi-lobed and hypersegmented nuclei, suggesting a mature phenotype (Figure 4(a)).⁶⁵ Moreover, neutrophils from both prostate and blood of PTEN^(i)pe−/− mice expressed similar levels of CD101, confirming a mature phenotype, while CXCR2 levels were decreased in prostate TANs, implying restricted recirculation to the BM (Figure 4(b)). In contrast to other tumor models, TANs in PTEN^(i)pe−/− mice did not express CD84 or Trem2 (Supplementary Figure S4b).^18,66 scRNA-seq of PTEN^(i)pe−/− prostate identified cluster 2, as neutrophil cluster (Figure 2(f)), expressing S100a9, Ptgs2, Cd274, Arg2, Il1b, Il1rn (Figure 4(c)), suggestive of immunosuppressive functions. S100A9 (Mrp14) and PD-L1 expression by TANs was confirmed by confocal microscopy and FACS respectively (Supplementary Figure S4c-d). To assess whether TANs possess immunosuppressive properties we sorted them from PTEN^(i)pe−/− PINs and cocultured them ex vivo with CD8⁺ T cells. We observed that they efficiently suppressed the CD8⁺ T cell proliferation, in contrast to bulk CD11b⁺ sorted macrophages, which lacked this suppressive capacity (Figure 4(d,e) and Supplementary Figure S4e-f). In summary, PTEN^(i)pe−/− prostates were massively invaded by TANs, exhibiting a mature phenotype, PD-L1 expression, and immunosuppressive function.

Figure 4.

Distinct neutrophil states infiltrated the PIN and adenocarcinoma stages of PTEN^(i)pe−/− prostate.

(a) Morphological characteristics of Ly-6G⁺CD11b⁺ neutrophils sorted from PTEN^(i)pe−/− prostates at PIN stage (3 months post tamoxifen). Two representative images are shown, and scale bars indicate the size of the sorted neutrophils.

(b) FACS plots showing CD101 and CXCR2 expression by Ly-6G⁺CD11b⁺ neutrophils from prostate and blood of PTEN^(i)pe−/− mice at PIN stage (3 months post tamoxifen).

(c) Normalized expression of selected marker genes (S100a9, Ptgs2, Cd274, Arg2, Il1b, Il1rn) in the total CD45⁺ immune cells presented in the UMAP in Figure 2(f) from PTEN^(i)pe−/− samples including both time points.

(d) Cell Trace Violet (CTV) histograms for the ex vivo suppressive assay showing the proliferation of WT splenic CD8⁺ T cells stimulated with anti-CD3 and anti-CD28 and co-cultured with Ly-6G⁺CD11b⁺ neutrophils sorted from prostate of PTEN^(i)pe−/− mice at PIN stage (3 months post tamoxifen) at different ratios (CD8⁺ T : Neutrophils). Negative control: non-stimulated CD8⁺ T cells cultured alone. Positive control: CD8⁺ T cells stimulated with anti-CD3 and anti-CD28 and cultured alone. (e) Proliferation and expansion indexes of CD8⁺ T cells based on the CTV histograms shown in (d). Data are presented as mean +/- SEM of n=4 culture replicates for each condition and are representative of 2 independent experiments. Statistical significance was calculated using Kruskal Wallis test followed by Dunn's ultiple comparisons test; ***, P < 0.001; ****, P < 0.0001

(f) Integrative re-analysis of scRNA-seq of the neutrophil cluster (2) from the general UMAP shown in Figure 2(f), for control (left) and PTEN^(i)pe−/− (right) samples comprising 2 replicates of 3 months and 9 months for each genotype. Colors indicate the recalculated clusters. Cell type annotation was performed using the marker genes of the new clusters.

(g) Dot plot of the average expression level of the 5 top marker genes of each cluster identified in (f). The color of the dots represents the average of normalized and scaled expression of marker genes in each cell type, and their size indicates the percent of cells expressing the marker genes in the cluster.

(h) Integrative analysis of scRNA-seq of the neutrophil clusters shown in f, visualized using a UMAP embedding for PTEN^(i)pe−/− samples at 3 months PIN stage (left) and 9 months adenocarcinoma stage (right) comprising 2 replicates of each timepoint. Colors represent the recomputed clusters, similar to f-g. (i) Stacked bar plot showing the fractional composition, for PTEN^(i)pe−/− samples at PIN (3-months) and adenocarcinoma (9-months) stages, of the neutrophil clusters described in (f). The color code corresponds to the clusters depicted in (f). Significance between the conditions for the different clusters is indicated by an asterisk, based on the statistical analysis presented in (Supplementary Figure S4i). (j) Pseudotime differentiation trajectories within neutrophil clusters, predicted using Monocle3 from integration of PTEN^(i)pe−/− samples of both time points. The indicated trajectories are presented in (Supplementary Figure S4j).

(k) Functional enrichment of the neutrophil clusters based on the expression of top marker genes of each cluster for several selected Gene Ontology (GO) Biological Process (BP) terms. Dot color represents the adjusted p-value of the test and dot size represents the ratio of marker genes present in the GO BP term.

(l) Functional enrichment of the neutrophil clusters based on the expression of top marker genes of each cluster for several selected Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Dot color represents the adjusted p-value of the test and dot size represents the ratio of DEG present in the GO BP term.

Tumor-associated neutrophils harbor distinct activation states

Although neutrophils have been formerly considered as a homogeneous and transient population, recent studies have highlighted a high TAN heterogeneity.^63,67–72 To explore the diversity of neutrophil infiltrating the PTEN^(i)pe−/− prostates, cluster 2 cells from Figure 2 were re-analyzed, revealing eight distinct clusters, N0-N7 (Figure 4(f–g) and Supplementary Table S1). Cluster N3 was predominant among the few neutrophils found in control prostates and exhibited a signature consistent with steady-state mature blood neutrophils and early tumor stages described in other tumor types (Figure 4(f–g) and Supplementary Figure S4g).^63,68,69 Additionally, the cluster N5 is associated with late tumor stages as described in other tumor models (Figure 4(f–g) and Supplementary Figure S4g).^63,69 Clusters N0 and N2 were predominantly observed at the PIN stage, whereas clusters N5 and N7 were more prominent at the adenocarcinoma stage (Figure 4(h,i) and Supplementary Figure S4i). The distinct distribution patterns of these clusters between the PIN and adenocarcinoma stages suggested that tumor-infiltrating neutrophils display significant plasticity, adopting distinct states and likely following diverse developmental trajectories. Cell trajectory analysis on neutrophil clusters suggested that cluster N2, which is predominant at 3 months PIN stage of PTEN^(i)pe−/− prostate, originated from cluster N3. Similarly, cluster N5 which is predominant at 9 months adenocarcinoma stage, originated also from cluster N3 (Figure 4(j) and Supplementary Figure S4j). Moreover, cluster N5 further differentiate into cluster N6 which also increases in numbers at 9 months adenocarcinoma stage (Figure 4(j) and Supplementary Figure S4j). In summary, our findings revealed significant heterogeneity among TANs in PTEN^(i)pe−/− prostates, likely corresponding to distinct activation states.

Tumor-associated neutrophil subsets display transcriptomic profiles indicative of protumoral activities

Transcriptomic profile analysis of the neutrophil clusters revealed that N0 exhibited an interferon-stimulated gene (ISG) signature, expressing high levels of Cxcl10, Gbp5, Gbp2, Ifit1, Ifit2 and Ifit3 (Figure 4(g)), largely matching the recently described ISG-neutrophils found in other tumor types (Supplementary Figure S4g).^68,71,72 N0 also expressed high levels of cathepsin S (Ctss), which could promote pericellular proteolysis and ECM alterations, allowing tumor cell and leukocyte migration to the TME and high levels of PD-L1 (CD274) suggestive of immunosuppressive functions (Supplementary Figure S4k). BP GO and KEGG pathway analysis confirmed that cluster 0 was associated with active immune response involved in antiviral immunity, TNF signaling pathway, phagosome, apoptosis and necroptosis (Figure 4(k–l)). N2, which prevailed at the PIN stage, exhibited an immunosuppressive signature with Il1b and alarmins S100a8 and S100a9 gene expression. BP GO and KEGG analysis for N2 highlighted enrichment in the regulation of actin cytoskeleton, leukocytes transendothelial migration, VEGF and chemokine signaling pathways, and neutrophil extracellular trap (NET) formation (Figure 4(g,k–l) and Supplementary Figure S4k). Notably, N5, which was significantly increased at the adenocarcinoma stage, expresses Ccl6, Hif1a genes, and Hif1a-target genes (Eno1, Gapdh, Ldha), as well as the gene Tnfrsf23, coding for the protein dcTRAIL-R1 (Figure 4(g) and Supplementary Figure S4k), recently described in TANs.⁶⁹ Moreover, BP GO terms and KEGG analysis indicated that N5 is associated to leukocyte migration, cytokine response, HIF-1 signaling, efferocytosis and ferroptosis (Figure 4(k–l)). The N7 cluster showed high expression of mitochondrial genes mt-Atp6, mt-Co1, mt-Co2, mt-Co3, coding, respectively, for the ATP synthase F0 subunit 6, cytochrome c oxidase subunits 1, 2, and 3, indicatives of significant metabolic activity. Additionally, BP GO and KEGG analysis for N7 highlighted enrichment in mRNA metabolic processes, efferocytosis, and necroptosis (Figure 4(g,k–l)). In summary, the transcriptomic profiles of TAN subsets suggested diverse biological functions and potential protumoral activity.

Prostate tumors at PIN stage are infiltrated by exhausted-like CD8⁺ T cells

We last focused our analysis on NK and T cells identified in the scRNA-seq dataset. Clusters 1, 4, and 7 described in Figure 2 were isolated and reclustered, yielding to 19 sub-clusters (Figure 5(a) and Supplementary Table S1). Based on their differentially expressed gene signature and module scoring from published data,^73–75 we identified five NK cell clusters (Clusters 0, 4, 8, 12, 17), five CD8⁺ T cell clusters (clusters 1, 3, 5, 16, 18), five CD4⁺ T cell clusters (clusters 6, 7, 10, 11, 13) including Tregs (Cluster 13, expressing Foxp3, Ctla4 and Tnfrsf4), and three clusters comprising γδ T cells (cluster 2), ILCs (clusters 9, 15) (Figure 5(b) and Supplementary Figure S5a). Among these clusters, four clusters were significantly increased in prostates of PTEN^(i)pe−/− mice as compared to control, including three CD8⁺ T cell clusters (clusters 3, 16, and 18) and one NK cluster (cluster 12) (Figure 5(c) and Supplementary Figure S5b). DEG analysis comparing PTEN^(i)pe−/− versus control revealed enrichment of ribosomal activity for cluster 3 (Figure 5(d)), suggestive of strong translation. Interestingly, cluster 3 exhibited an increase expression of Pdcd1 and Lag3 genes in PTEN^(i)pe−/− mice (Figure 5(e)), suggesting an exhausted phenotype. Subsequently, we performed flow cytometry analysis to detect T lymphocyte subsets and observed a significant increase of CD4⁺ and CD8⁺ T cells in the prostates of PTEN^(i)pe−/− mice from 3 months post gene invalidation (Supplementary Figure S5c-d). Moreover, PD-1^highTim3⁻ and PD-1^highTim3⁺ within CD8⁺ T cells were strongly increased in PTEN^(i)pe−/− mice at PIN and adenocarcinoma stages with the frequency of PD-1^highTim3⁻CD8⁺ T cells being significantly higher at adenocarcinoma stage (Figure 5(f)). Notably, analysis of the TCGA PRAD cohort demonstrated that human prostate tumors with a Gleason score of 9–10 had significantly higher exhausted-like CD8⁺ T cell signature score than tumors with lower Gleason scores (Supplementary Figure S5e). Moreover, the exhausted-like CD8⁺ T cell signature correlated with Trem2⁺ TAM signature expression in patients with prostate cancer (Supplementary Figure S5f). To address the concurrent infiltration of exhausted CD8⁺ T cells, along with immunosuppressive TAMs and TANs, in PTEN^(i)pe−/− prostates at the PIN stage, we implemented a therapeutic strategy designed to modulate the intratumoral CD8⁺ T cell phenotype. This regimen consisted of a one-week treatment combining anti-Ly-6G to deplete neutrophils,⁷⁶ anti-Csf1R to target macrophages, and anti-PD-1 to inhibit T cell checkpoint signaling (Figure 5(g) and Supplementary Figure S5g). Following treatment, we assessed the intratumoral CD8⁺ T cell phenotype by analyzing their production of IFN-γ and TNF-α (Figure 5(h)). The proportions of IFN-γ⁺TNF-α⁻ and IFN-γ⁺TNF-α⁺ CD8⁺ T cells in PTEN^(i)pe−/− prostates were comparable across all treatment groups, indicating that the combined therapy did not favor the infiltration of cytotoxic intratumoral CD8⁺ T cells, likely due to the short one-week window of treatment (Figure 5(i)). In summary, PTEN^(i)pe−/− prostates exhibit multiple immune alterations including significant infiltration of CD8⁺ T cells with features of exhaustion that are well correlated with human prostate tumors.

Figure 5.

Exhausted CD8⁺ T cells infiltrated the cancerous prostates.

(a) Integrative re-analysis of scRNA-seq of T and NK clusters (1, 4, 7) from the general UMAP shown in Figure 2(f), of control (left) and PTEN^(i)pe−/− (right) samples encompassing 2 replicates of 3 months tumor timepoint for each genotype. Color coded cluster annotation is mentioned.

(b) Dot plot of the average expression level of the top 5 marker genes used to annotate the cell clusters in (a). The color represents the average of normalized and scaled expression of marker genes in each cell type, and the size indicates the percent of cells expressing the marker genes.

(c) Fractional composition for control and PTEN^(i)pe−/− samples of 3 months timepoint of the T and NK color-coded clusters described in (a). Significance between the conditions for the different clusters is indicated by an asterisk, based on the statistical analysis presented in Supplementary Figure S5b.

(d) Enrichment of Top 10 Gene Ontology Biological Process terms for cluster 3 (CD8⁺ T_2) based on the top differential gene expression PTEN^(i)pe−/− versus Control samples at 3 months timepoint. The color represents the adjusted p-value for each term and the dot size indicates the cell count bearing each term. The gene ratio is also indicated.

(e) Heatmap of the normalized mean expression of the Pdcd1 and Lag3 genes for each cluster of Control and PTEN^(i)pe−/− samples at 3 months tumor timepoint and as annotated in (a). Significance is calculated based on the Dunn adjusted p-value; **, p < 0.01; ***, p < 0.001

(f) Gating strategy and quantification of PD1^highTim3⁻ and PD1^highTim3⁺ CD8⁺ T cells in control and PTEN^(i)pe−/− prostates of DLV lobes at 3 and 9 months post tamoxifen. Quantification is presented as frequency of the CD8⁺ T cells. Data are represented as mean ±SEM of n = 3-9 animals and they are a concatenation of 2 independent experiments. Statistical significance was calculated using Two-way ANOVA with mixed effects model; ***, p < 0.001; ****, p < 0.0001

(g) Treatment scheme for combined therapy. PTEN^(i)pe−/− mice at 3 months post tamoxifen were treated with 100 μg of murinized anti-Ly-6 G (clone 1A8), 500 μg of rat anti-CSF1R (clone AFS98) and 100 μg of rat anti-PD1 (clone RMP1-14) at day 0, 3, 5, 7. On day 8, the mice were euthanized and prostates were analyzed by flow cytometry.

(h) IFN-γ and TNF-α production by CD8⁺ T cells from DLV prostate lobes of mice treated with the isotype control or with the combined triple therapy. Plots are presented as overlays of FMO (blue) and IFN-γ and TNF-α staining (red).

(i) Frequency of IFN-γ⁺TNF-α⁻ and IFN-γ⁺TNF-α⁺ CD8⁺ T cells in PTEN^(i)pe−/− prostate (DLV lobes) of the different treated groups: Isotype control, anti-PD1 alone, anti-CSF1R + anti-Ly-6 G, or anti-CSF1R + anti-Ly-6 G + anti-PD1.

Discussion

Our study provided a comprehensive characterization of the immune cells in the healthy mouse prostate, as well as of PIN- and adenocarcinoma-containing prostates of PTEN^(i)pe−/− mice, which faithfully recapitulates human PCa features.²¹ In healthy prostate, we observed the presence of the major immune cell lineages, including MNPs, NK, B and T cells and few neutrophils, and we further dissected the heterogeneity of the tissue-resident macrophages. In PTEN^(i)pe−/− mice, prostates exhibited a significant infiltration of heterogeneous neutrophils, Trem2⁺ TAMs and exhausted CD8⁺ T cells from the PIN stage onward. Notably, the presence of Trem2⁺ macrophages and dysfunctional T cells are also characteristic features of aggressive human PCa, as reported by Wong et al.,¹⁰ indicating that the PTEN^(i)pe−/− mouse model possesses significant translational relevance to human disease.

Building on studies of tissue-resident macrophage heterogeneity in various tissues as well as endocrine and exocrine glands,^26,30 and our prior research on skin macrophages,^31,45,77 we explored macrophage complexity in the murine prostate, using flow cytometry, confocal analysis and scRNA-seq. Three distinct resident macrophage subsets were identified in healthy prostate, Tim4⁺, Tim4^– and CD11b^– Mfs. Prostatic Tim4⁺ Mfs expressed Tim4, FolR2 and Lyve-1, exhibited high scavenging capacity for blood-borne ligand (intravenously injected dextran), localized near CD31⁺ endothelial cells,^27,28 and were unaffected by the absence of the FIRE super-enhancer of the Csf1r locus.³⁶ Based on these features, Tim4⁺ Mfs were identified as perivascular macrophages. Prostatic CD11b^– Mfs, which expressed high level of CX3CR1 and harbored long dendrites, formed a continuous network of the prostate duct epithelium, resembling CX3CR1⁺ ductal macrophages described in murine mammary glands.²⁹ Although, we did not detect metallothionein family genes in CD11b^– Mfs, their anatomical location suggests they may correspond to the zinc transporter-expressing prostate-specific macrophage subset previously described in the human prostate, alongside two other macrophage populations.⁹ Further investigation is required to characterize the Tim4^– Mfs in greater detail, as they likely represent a heterogeneous population within the stromal region of the prostate, consistent with the multiple macrophage clusters (Mac1-Mac4) identified through scRNA-seq analysis.

Tissue-resident macrophages were shown to be long lived cells, originating from embryonic stages, and capable of local self-renewal without input from circulating monocytes as seen in macrophages of the lung, liver, brain, and skin.^26,78,79 Alternatively, some can be replaced by monocytes generated in the bone marrow, as seen in intestine.^26,78,80 In our study, all macrophage subsets were depleted following aCsf1-R treatment, confirming their reliance on Csf-1 for survival. Moreover, prostate-protected BM chimera experiments confirmed that Tim4⁺ and CD11b^– Mfs remain of host origin in prostate tissue, indicating a low turn-over and self-renewal ability. In contrast, Tim4^– Mfs displayed partial replacement by donor BM-derived cells, suggesting a mixed origin of self-renewing and monocyte-derived macrophages. Further studies are needed to elucidate the prenatal origin of these macrophage subsets and identify their corresponding Csf1-producing niches,⁷⁹ and their respective role in prostate function at homeostasis.

Although the number of Tim4⁺ and CD11b^– resident macrophages was comparable between healthy and PTEN^(i)pe−/− prostates throughout tumor progression, a significant alteration in the mononuclear phagocyte compartment was observed, marked by a strong increase in Tim4^– macrophages, corresponding to the recruitment of Trem2⁺ TAMs in PIN-containing glands. The substantial number of cells needed for in vitro assay to evaluate the immunosuppressive function of intratumoral immune cells, combined with the limited yield of Trem2⁺ TAMs due to the small size of the murine prostate, prevented us from conducting in vitro immunosuppressive assays with Trem2⁺ TAMs isolated from PTEN^(i)pe−/− prostates. In other cancer types, Trem2⁺ TAMs are critical mediators of immune suppression.^56,81–83 In human PCa, Trem2⁺ TAMs are likewise present and have been associated with poor clinical outcomes.^10,84 Moreover, metabolically altered Trem2⁺ TAMs have been implicated in the progression of human bladder cancer.⁸⁵ Trem2-mediated immunosuppression is partly dependent on ligand engagement, including interactions with apolipoprotein E (ApoE) and phospholipids. Consequently, Trem2 deficiency has been shown to suppress tumor growth by attenuating immunosuppression, promoting T cell responses and reprogramming TAM phenotypes.^56,59,85 In PTEN^(i)pe−/− prostate, Trem2 expression is restricted to macrophages, while Trem2⁺ immunosuppressive neutrophils have been detected in the probasin-mediated aggressive prostate adenocarcinoma.¹⁸ Transcriptomic analysis indicated that Trem2⁺ TAMs, in PTEN^(i)pe−/− prostates, expressed high levels of macrophage migration inhibitory factor (MIF), a known promoter of intratumoral TAMs, MDSCs and neutrophils recruitment, immunosuppression and tumor progression.^62,86 Their recruitment and immunosuppressive function may be linked to the HIF1A-associated hypoxic microenvironment described in PTEN^(i)pe−/− prostate.²⁴ Interestingly, in hypoxic microenvironment, MIF and HIF1A synergistically increase lactate production and the senescence-associated secretory phenotype (SASP), promoting the Warburg effect and stabilization of P53 observed in senescence.^61,62,87 Thus, in PTEN^(i)pe−/− prostates, Trem2⁺ TAMs could sustain tumor progression to adenocarcinoma through the MIF/HIF1a axis. Targeting Trem2 or MIF has shown promise in enhancing antitumoral immune responses in other cancer models, such as triple-negative breast cancer and colon carcinoma.^{56,81–83,87}

Research on tumor-associated neutrophils has been hindered by the use of the Gr1 marker to describe MDSCs, which include neutrophil and monocyte lineages with potent immunosuppressive activity.⁶⁴ Gr1⁺ myeloid cells or S100a8⁺ MDSCs were observed in the prostate of PTEN^(i)pe−/− mice.^22,24 As Gr1 recognizes both Ly-6C and Ly-6 G, markers commonly used to distinguish monocytes and neutrophils respectively, we did not use the Gr1 antibody and the MDSCs nomenclature, favoring the monocyte, macrophage and neutrophil nomenclature. In PTEN^(i)pe−/− prostates, we confirmed that the most prominent alteration is the extensive infiltration of Ly-6G⁺ neutrophils, which predominantly accumulate within the lumen of PIN-containing glands. In human PCa, PMN-MDSCs were primarily investigated using immunohistochemistry and have been shown to contribute to disease progression and immune evasion.⁸⁸

Recent studies have demonstrated that neutrophils represent a heterogeneous population, with reprogramming potential.^63,68–72 However, none of these investigations have studied neutrophils in the context of human prostate tissue. This heterogeneity is evident even in blood in physiological conditions, where three distinct neutrophil states, including an ISG-expressing subset that expands during bacterial infection, have been identified.⁶⁸ However, a standardized nomenclature for the increasingly complex differentiation and activation states of neutrophils in tumors remains lacking.⁸⁹ In an orthotopic pancreatic ductal adenocarcinoma (PDAC) model, tumor-associated neutrophils were shown to undergo reprogramming to a terminal state characterized by dcTRAIL-R1 and VEGFα expression. These neutrophils exhibited extended lifespan and localized to hypoxic-glycolytic niches, where they exert pro-angiogenic, immunosuppressive and pro-tumoral functions.⁶⁹ Our transcriptomic analysis revealed that neutrophils, within PTEN^(i)pe−/− prostate tumors, displayed marked heterogeneity, resolving into eight distinct states. Among these, dcTRAIL-R1-expressing neutrophils (N5) emerged predominantly at the prostatic adenocarcinoma stage, suggesting that tumor progression drives their reprogramming. This reprogramming may be influenced by the hypoxic microenvironment observed in PTEN^(i)pe−/− prostates,²⁴ potentially regulated through the Trem2⁺ TAMs and the MIF/HIF1a axis described above. Consistent with this, PTEN/HIF1A^(i)pe−/− prostates, where both PTEN and HIF1A were inactivated in PECs, exhibited a reduction in MDSCs and an increase in CD8⁺ T cells and NK cells.²⁴ Furthermore, N5 neutrophils showed high expression of Ccl6, suggesting their implications in remodeling the TME and sustaining immunosuppressive conditions, as recently described during breast cancer metastasis.⁹⁰ However, the lack of specific surface marker to discriminate the distinct neutrophil clusters impeded their precise identification by flow cytometry and limited further functional analysis ex vivo.

A recent neutrophil atlas across various human cancers described the ISG-expressing IFIT1⁺ISG15⁺ neutrophil subset, which exhibited high PD-L1 expression, indicative of an immunosuppressive role.⁷² Another study highlighted that effective anti-CD40 immunotherapy correlated with the infiltration of ISG-expressing neutrophils, which exhibited high cytotoxic potential via CXCL10 secretion and anti-tumor activity.⁷¹ Elevated CXCL10 levels have been associated with effector CD4⁺, CD8⁺, and NK cell tumor infiltration, reduced tumor growth and indicated improved immunotherapy response in metastatic melanoma.^91,92 In PTEN^(i)pe−/− prostates, the predominant N0 neutrophil state at the early PIN stage, is characterized by high CXCL10 and PDL-1 expression. It is hypothesized that these early infiltrating neutrophils (N0) may possess both anti-tumoral properties, akin to those described by Gungabeesoon et al.,⁷¹ and immunosuppressive potential through PDL-1.⁷² Further investigations on neutrophil reprogramming in PTEN^(i)pe−/− prostates, particularly in response to anti-CD40 and anti-PD-L1 therapies, are needed to determine whether sustaining or enhancing the ISG-expressing state could hinder prostate tumor progression.

Last but not least, one of the main features of the PTEN^(i)pe−/− model is the tumor-related CD8⁺ T cell exhaustion. This is clinically relevant as exhaustion signature genes have been extensively described in human prostate cancer.^9–11

Overall, in the PTEN^(i)pe−/− model, our findings revealed the early mobilization of distinct immune cells, including Trem2⁺ TAMs, dcTRAIL-R1⁺ and ISG-TANs, and exhausted CD8⁺ T cells, during prostate tumorigenesis, identifying these immune populations as potential therapeutic targets. Alterations of the immune compartment, particularly involving the HIF1a/MIF and Ccl6 axes, present opportunities for therapeutic intervention. Combined therapies such as anti-Trem2 and anti-PD1, or strategies targeting MIF in Trem2⁺ TAMs or ISG pathways in TANs could potentially mitigate immune suppression and enhance antitumoral immunity in PTEN^(i)pe−/− PCa model.^93–95

Limitations of our study

Using scRNA-seq signature scores, our study provided compelling evidence and a strong concordance between cell phenotypes and their putative biological functions, as validated in studies with sufficient cell numbers. However, the small size of the mouse prostate constrained the yield of specific immune cell subset of interest. Consequently, performing ex vivo functional assays, for instance, on macrophage and CD8⁺ T cell subsets, would have required a substantial number of animals from both genotypes (PSA-Cre-ER^T2(tg/0)/PTEN^L2/L2 and PSA-Cre-ER^T2(0/0)/PTEN^L2/L2), which was not ethically justifiable. In addition, we encountered challenges with long-term neutrophil depletion using monoclonal antibodies when attempting to assess neutrophil function in vivo (data not shown). To overcome this limitation, genetic approaches may offer more effective and sustainable strategies to interrogate immune cell function in vivo. For instance, future studies could consider crossing PSA-Cre-ER^T2(tg/0)/PTEN^L2/L2 mice with genetically modified strains such as Genista mice, which carry a point mutation in the transcriptional repressor Growth Factor Independence 1 (Gfi1) and display neutropenia,⁹⁶ Ccr2^−/− mice, which exhibit markedly reduced numbers of circulating monocytes and monocyte-derived macrophages,⁴⁹ or Trem2^−/− mice, as Trem2 deficiency has been shown to suppress tumor growth in several studies.^56,59,85 Finally, our primary objective was to provide a comprehensive characterization of the immune cell landscape in the mouse prostate under both healthy and tumor-bearing conditions. As a result, our study offers insights into the potential translational relevance of our findings to human prostate cancer, encouraging a reevaluation of the role and diversity of neutrophils and macrophages in the prostate cancer microenvironment diagnosis and treatment.

Funding Statement

This work was supported by the French National Cancer Institute (INCA) to G.L. and S.H. ([INCA2019-191]; PLBIO23-177) and to D.M (INCA2019-191); the Fondation ARC pour la recherche sur le cancer [PJA20191209674], the INCA Région Sud, the Institute for Cancer Immunology (ICI) and Canceropôle to S.H [2023-01Kpole]; PHENOMIN (French National Infrastructure for mouse Phenogenomics; [ANR-10-INBS-07]) and the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (grant agreement n° [787300](BASILIC) to B.M.; The Ministère de l’Enseignement supérieur et de la Recherche to D.P., K.L.T. and D.Y., the Fondation pour la Recherche Médicale (FRM) to D.P. and D.Y. and the Fondation ARC pour la recherche sur le cancer to K.L.T.; ANR-10-INBS-04-01 France Bio Imaging to CIML Imagim platform; and the Centre d’Immunologie de Marseille Luminy, which receives its core funding from Aix Marseille University, CNRS, and INSERM. Institut National Du Cancer INCA2019-191, Institut for Cancer Immunology

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

All data and material are available within the article and its supplementary information file. The scRNA-sequencing data have been deposited in the Gene Expression (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) under accession number GSE271451. Other data from the project have been deposited on the french national research data repository. Four datasets are available (Dataset 1: https://doi.org/10.57745/CRGTDZ ; Dataset 2 https://doi.org/10.57745/NVZQ25 ; Dataset 3 https://doi.org/10.57745/R1VNP3 ; Dataset 4: https://doi.org/10.57745/4EFY2Z). Source code of bioinformatics analysis is available on GitHub (https://github.com/CIML-bioinformatic/BMMlab_Prostate-CD45Pos-PTENKO) and on the Zenodo open archive (https://doi.org/10.5281/zenodo.14670913).

Ethics statement

In vivo procedures were handled in accordance with national and European laws for laboratory animal welfare and experimentation, and protocols approved by the Marseille Ethical Committee for Animal Experimentation and French Animal Ethics Committee (approval APAFIS #25240–2020042716503331 v5 and #46078–2023112412266817 v4).

The use of publicly available human datasets for research purposes is exempt from formal review by a research ethics committee according to the French law.

Supplementary Information

Supplemental data for this article can be accessed online at https://doi.org/10.1080/2162402X.2025.2562220