Peritoneal resident macrophages constitute an immunosuppressive environment in peritoneal metastasized colorectal cancer

Abstract

Patients with peritoneal metastasized colorectal cancer (PM-CRC) have a dismal prognosis. We hypothesized that an immunosuppressive environment in the peritoneal cavity underlies poor prognosis. We define the composition of the human peritoneal immune system (PerIS) using single-cell technologies in 18 patients with- and without PM-CRC, as well as in matched peritoneal metastases (n = 8). Here we show that the PerIS contains abundant immunosuppressive C1Q⁺VSIG4⁺ and SPP1⁺VSIG4⁺ peritoneal-resident macrophages (PRMs), as well as monocyte-like cavity macrophages (mono-CMs), which share features with tumor-associated macrophages, even in homeostasis. In PM-CRC, expression of immunosuppressive cytokines IL10 and VEGF increases, while simultaneously expression of antigen-presenting molecules decreases in PRMs. These intratumoral suppressive PRMs originate from the PerIS, and intraperitoneal depletion of PRMs in vivo using anti-CSF1R combined with anti-PD1 significantly reduces tumor burden and improves survival. Thus, PRMs define a metastatic site-specific immunosuppressive niche, and targeting PRMs is a promising treatment strategy for PM-CRC.

Fig. 1. The homeostatic peritoneal immune system is dominated by peritoneal-resident cavity macrophages.

A Gene expression Uniform Manifold Approximation and Projection (UMAP) overlay of color coded PBMC (red) derived immune cells (n = 13,242 cells) and PF (blue) derived immune cells (n = 14,441 cells) using Harmony for batch correction. B Gene and protein weighted nearest neighbor (GP-WNN) UMAP color coded PF derived immune cells (n = 14,441 cells) identifying 27 subsets. C GP-WNN UMAP feature plots of reclustered MNPs in PF of HCs showing log transformed expression of scRNA markers used for myeloid subset identification DCs: CLEC4C, XCR1, and CD1C and CITE-seq CD1C; mononocytes + macrophages: ITGAM, MARCO, CD14, CCR2, and CD163 and CITE-Seq (CD14, CD192 and CD163). D GP-WNN UMAP color coded PF derived monocytes + macrophages (n = 3,937 cells) identifying 4 clusters. E GP-WNN UMAP feature plot of reclustered monocytes + macrophages PF of HCs showing log transformed expression of VSIG4. F GP-WNN UMAP feature plots of reclustered monocytes + macrophages in PF of HCs showing log transformed expression of scRNA markers CD163, VCAN, CCR2, C1QA, SPP1, and CD163 (CITE-Seq). G Boxplot showing the annotation and subsequent proportion of monocytes + macrophages subsets relative to total MNPs in PF of HCs (n = 5 donors) (does not add to 100% as DC subsets are not included in the graph). Box: 25th to 75th percentiles; line: median; whiskers: minimum and maximum values. UMAP Uniform Manifold Approximation and Projection, scRNA-seq single cell RNA sequencing, CITE-seq cellular indexing of transcriptomes and epitopes sequencing, HC healthy controls, PBMC peripheral blood mononuclear cell, PF peritoneal fluid, Mono-macs monocytes + macrophages, MNPs mononuclear phagocytes, T T cells, NK/ILC natural killer cells/innate lymphoid cells, B B cells, TCM T central memory, Treg regulatory T, CTL cytotoxic T lymphocyte, TEM T effector memory, MAIT mucosal-associated invariant T, GDT gamma delta T cells, NKT natural killer T, CDC1s conventional dendritic cells type 1, CDC2s conventional dendritic cells type 2, PDCs plasmacytoid dendritic cells, DCs dendritic cells, Class classical, mono-CM monocyte-like cavity macrophage, CM cavity macrophage. Source data are provided as a Source Data file.

Fig. 2. Peritoneal-resident cavity macrophages are immunosuppressive by nature.

A Radarplot showing TAM median Ucell score signature per macrophage subset in PF. B Dot plot of selected genes encoding pro- (red) or anti-inflammatory (green) genes in all four monocytes + macrophages subsets. C Heatmap showing the row-scaled mean log₂transformed expression of top 5 unique marker proteins per monocytes + macrophages subset. D GP-WNN UMAP feature plots of reclustered mononocytes + macrophages showing log transformed expression of C1QA and SPP1 in healthy PF, colon, and liver tissue. E Boxplot analysis of monocytes + macrophages, C1Q⁺ CMs, and SPP1⁺ CMs comparing PF (n = 5 donors), colon (n = 5 donors), and liver (n = 5 donors). Box: 25th to 75th percentiles; line: median; whiskers: minimum and maximum values. Statistics: Mann–Whitney test, two-tailed. *p = 0.032, **p = 0.0079. UMAP Uniform Manifold Approximation and Projection, scRNA-seq single cell RNA sequencing, CITE-seq cellular indexing of transcriptomes and epitopes sequencing, HC healthy controls, PF peritoneal fluid, Mono-macs monocytes + macrophages, mono-CM monocyte-like cavity macrophage, CM cavity macrophage, RTM resident-tissue macrophages, Angio pro-angiogenic, IFN interferon-primed, INFLAM inflammatory-cytokine enriched, LA lipid-associated, Prolif proliferating, Reg immune regulatory, pro-inflam pro-inflammatory, anti-inflam anti-inflammatory. Only statistically significant differences are shown (*p ≤ 0.05; **p ≤ 0.01). Source data are provided as a Source Data file.

Fig. 3. Peritoneal resident macrophages shape an immunosuppressive niche for PM-CRC.

A UMAP color-coded PF-derived macrophage subsets (n = 8661 cells) comparing HC (left) and PM-CRC (right) identifying 4 macrophage subsets. B Stacked bar chart analysis of macrophage subsets as a proportion of macrophages comparing PF of HC to PM-CRC. C Heatmap showing effect size comparing PM-CRC with HC (Wald statistic) as obtained using DESeq2 of significant altered genes in either HC (turquoise) or PM-CRC (brown) of macrophage subsets from selected KEGG pathways. DESeq2 Wald test derived p values. D Bar chart showing mean with SEM of absolute gene expression levels of HLA-DRB1, HLA-C, IL10, and VEGFA in all macrophage subsets in PF comparing HC (n = 5 donors) with PM-CRC (n = 13 patients). Statistics: Wald test using DESeq2. Multiple comparison adjustments and adjusted p values are used for testing. HLA-DRB1: **p = 0.007; *p = 0.014; HLA-C: *p = 0.022; ***p = 0.0003; IL10: *p = 0.035; 0.037; 0.039; VEGFA: **p = 0.007; 0.008. E Normalized Protein Expression (NPX) levels of IL-10 and VEGFA, using Olink technology in PF comparing HC (n = 5 donors) to PM-CRC patients (n = 12 patients). Line at median. Statistics: Mann–Whitney test, two-tailed. ***p = 0.0006; 0.0003. F Correlation plot of NPX levels of IL-10 and VEGFA in the PF of PM-CRC patients (n = 21 patients) and PCI score. Statistics: Spearman correlation. *p = 0.043; 0.030 (two-tailed). G UMAP color-coded PF and PM-CRC (non-)immune subsets (n = 35,329 cells). H Feature plots of (non-)immune subsets identified in g, showing log-transformed expression of scRNA markers IL10 and VEGFA. UMAP Uniform Manifold Approximation and Projection, HC healthy controls, PM-CRC peritoneal metastasized colorectal cancer, CMs cavity macrophages, mono-CMs monocyte-like cavity macrophages, rec receptor, KEGG Kyoto Encyclopedia of Genes and Genomes, PCI Peritoneal Cancer Index, NPX Normalized Protein Expression, NK/ILC natural killer cells/innate lymphoid cells, Mono-macs monocytes + macrophages. Only statistically significant differences are shown (*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001). All UMAPs in this figure were generated using only gene expression data. Source data are provided as a Source Data file.

Fig. 4. Peritoneal-resident immunosuppressive macrophages infiltrate PMs.

A UMAP color coded for PMs from PM-CRC patients derived mono-mac subsets (n = 8 patients; 1,747 cells). B Boxplot of proportion of macrophage subsets out of total macrophages in PMs from PM-CRC patients (n = 8 patients). Box: 25th to 75th percentiles; line: median; whiskers: minimum and maximum values. Statistics: Mann–Whitney test, two-tailed. *p = 0.038, **p = 0.0019; 0.003, ***p = 0.0002. C Feature plot showing VSIG4 expression in PM derived monocytes + macrophages subsets. D UMAP color coded for monocytes or macrophages for HC (n = 5 patients) as well as the PM-CRC patients (n = 8 patients) of PBMC-, PF-, and, from PM-CRC only, PM-derived monocytes and macrophages (n = 17,930 cells) using Harmony for batch correction. E UMAP color coded for tissue and disease state of monocytes and macrophages for HC (n = 5 patients) as well as the PM-CRC patients (n = 8 patients) of PBMC-, PF- and, from PM-CRC only, PM-derived monocytes and macrophages (n = 17,930 cells) using Harmony for batch correction. F Dendrogram based on the Euclidian pairwise distances calculated between monocytes and macrophage samples obtained from HC (n = 5 patients) and PM-CRC patients (n = 8 patients). G Schematic overview of experimental set-up of PKH26 in vivo macrophage labeling (Created in BioRender. Vermeulen, J. (2025) https://BioRender.com/oi0rimk). H Gating strategy used to identify peritoneal PKH26⁺ macrophages: Plots are representative for their respective group. I Boxplot showing the percentage of PKH26⁺ F4/80⁺ macrophages in Spleen, PF, and PM, 2 weeks after tumor cell injection without (−, n = 1 mouse) or with (+, n = 6 mice) PKH26 linker injection. Box: 25th to 75th percentiles; line: median; whiskers: minimum and maximum values. UMAP Uniform Manifold Approximation and Projection, CMs cavity macrophages, mono-CMs monocyte-like cavity macrophages, mono-macs monocyte-macrophage, Spl spleen, PF peritoneal fluid, PM peritoneal metastasis, HC healthy controls, PM-CRC peritoneal metastasized colorectal cancer, i.p. intraperitoneal. (ns = non-significant; *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001). All UMAPs in this figure were generated using only gene expression data. Source data are provided as a Source Data file.

Fig. 5. Peritoneal-resident immunosuppressive macrophages contribute to the peritoneal metastatic TIME.

A Dot plots of selected genes encoding receptors of key macrophage chemoattractants. B Boxplot showing the normalized protein expression (NPX) levels using Olink technology in PF comparing HC (n = 5 donors) to PM-CRC patients (n = 12 patients). Line at median. Statistics: Mann–Whitney test, two-tailed. ***p = 0.0003; 0.0006; 0.0003. C Radarplot showing TAM median Ucell score signature per macrophage subset in PM. D Dot plots representing the 5 most differentially expressed TAM-associated genes per macrophage subset. Size represents percentage expressing cells and color represents the median gene expression. E Heatmap of bulk RNA-sequencing data (GSE190609) of paired patients’ primary tumor (PT), liver metastasis (LM), and peritoneal metastasis (PM) of colorectal cancer showing Z-score of immunosuppressive macrophage markers. F Paired patient analysis of gene set score in (E) (n = 12 patients with PT and PM, 6 patients also had LM). Statistics: paired t-test, two-tailed. *p = 0.026; 0.013. CMs cavity macrophages, mono-CMs monocyte-like cavity macrophages, mono-mac monocyte-macrophage, PM peritoneal metastasis, PT primary tumor, LM liver metastasis, HC healthy controls, PM-CRC peritoneal metastasized colorectal cancer, RTM resident-tissue macrophages, Angio pro-angiogenic, IFN interferon-primed, INFLAM inflammatory-cytokine enriched, LA lipid-associated, Prolif proliferating, Reg immune regulatory. (ns = non-significant; *p ≤ 0.05; ***p ≤ 0.001). Source data are provided as a Source Data file.

Fig. 6. Improved overall survival in mice after macrophage-depleting therapy combined with immune checkpoint blockade.

A Gating strategy used to identify peritoneal macrophages: CD11b⁺F4/80⁺. Plots are representative for their respective group. B Quantification of flow cytometry data of % peritoneal macrophages out of myeloid cells in PMs after 21 days. Mice received control (n = 7 mice), anti-CSF1R (n = 5 mice), anti-PD1 (n = 4 mice) or anti-CSF1R/anti-PD1 combination therapy (n = 3 mice). Bar chart showing mean with SEM. Statistics: Kruskal–Wallis. *p = 0.027, 0.046 C IF of CD206⁺ macrophages in PM. D Quantification of CD206⁺ IF staining in PMs. Data shown as mean number of CD206⁺ cells per HPF with SEM. (control n = 5 mice; anti-CSF1R n = 5 mice; anti-PD1 n = 4 mice; anti-CSF1R/anti-PD1 n = 4 mice; minimally 3 high-power field (HPF) per mouse). Statistics: Kruskal–Wallis. *p = 0.038, **p = 0.005 E IF of CD8⁺ T cells in PM split per different treated groups. F Quantification of CD8⁺ IF staining in PM. Data shown as mean number of CD8⁺ cells per HPF with SEM. (control n = 5 mice; anti-CSF1R n = 5 mice; anti-PD1 n = 4 mice; anti-CSF1R/anti-PD1 n = 3 mice; minimally 3 HPF per mouse). Statistics: Kruskal–Wallis. G, H Subgroup analysis of mice that either reached their humane endpoint (control: n = 10 mice; anti-CSF1R: n = 6 mice; anti-PD1: n = 4 mice and anti-CSF1R/anti-PD1: n = 1 mouse) or were sacrificed after 21 days (anti-CSF1R/anti-PD1: n = 2 mice). Data shown as mean with SEM. Bar chart of ascites (G) Statistics (G): Fisher’s exact test. *p = 0.0016; 0.0035. Bar chart of mPCI (H). Statistics (H): Kruskal–Wallis. *p = 0.013. I Representative pictures of vehicle control (upper panel) and combination therapy (lower panel). J Schematic overview of in vivo experiments. Experiment 1: anti-CSF1R at t = 0 days. Experiment 2: anti-CSF1R at t = 7 days (Created in BioRender. Vermeulen, J. (2025) https://BioRender.com/hb9202t). K Survival analysis after 4 weeks (see J; exp 1). Statistics: Log-rank test. l Survival analysis after 6 weeks (see J; exp 2). Statistics: Log-rank test. M Stacked bar chart analysis showing proportion of mice achieving complete remission after 42 days. N Representative picture of combination therapy showing complete remission. Mφ macrophage, α anti, CFS1R colony stimulating factor 1 receptor, PD1 programmed cell death protein 1, HPF high-power field, PM peritoneal metastasis, mPCI modified peritoneal cancer index, CR clinical remission, i.p. intraperitoneal. Pre-clinical model: intraperitoneal injection of CT26 colorectal cancer cell line in BALB/c mice, followed by administration of anti-CSF1R and anti-PD1. (*p ≤ 0.05; **p ≤ 0.01). Source data are provided as a Source Data file.