Tissue-Specific Immunosuppressive and Proliferating Macrophages Fuel Early Metastatic Progression of Human Colorectal Cancer to the Liver

Abstract

Early synchronous colorectal liver metastasis (CRLM) represents a clinical condition characterized by the simultaneous presence of primary colorectal cancer and metastatic liver lesions. In this study, we characterized the tissue-specific transcriptomes, phenotypes, and functional relevance of tumor-associated macrophages (TAM) within the tumor microenvironment (TME) of colorectal cancer and CRLM specimens from patients who underwent simultaneous surgical removal of these malignancies. The high-throughput single-cell transcriptional analysis revealed an inverse ratio of inflammatory and immunoregulatory TAMs in the colorectal cancer and CRLM TMEs, along with heterogeneity in both tumoral tissues. Furthermore, we found that inflammatory TAMs in colorectal cancer expressed inhibitory ligands that might support immune escape, thus favoring liver metastatic progression. In contrast, CRLM lesions possessed a highly immunosuppressive milieu characterized by large proliferative CTLA4+ immunoregulatory TAMs and the presence of IL7R+ cytotoxic TAMs. Higher frequencies of these specific TAM subsets in CRLM were associated with shorter disease-free survival and worse patient prognosis. The identification and characterization of immunoregulatory TAMs preferentially enriched in CRLM is key for the development of novel immunotherapeutic strategies aimed at boosting anticancer immune responses within the TME.

Figure 1.

Myeloid cell characterization in synchronous metastatic colorectal cancer by scRNA-seq. A, Schematic overview of the experimental workflow depicting sample processing and scRNA-seq of matched peritumoral samples of primary colorectal cancer (n= 3), metastatic liver tumor (n= 3), and peripheral blood cells (n= 2). B, Tissue-integrated UMAP of the myeloid cell reclustering, a total of 11,283 single cells from all the patients (CRLM-1–3, n= 3), color-coded by tissue assignment. C, UMAP plot (left) color-coded by the main myeloid cell families. The dot plot (right) shows the expression of canonical markers used for the annotation of the cell families, in which the dot size indicates the proportion of expressing cells and the color indicates the normalized mean expression levels. D, Fractions (%) of cell families detected in each tissue, colored as in (C), normalized on the total cells per tissue. The alluvial plot shows the correspondence between cell families and tissues, in which the stratum indicates the contribution of each cell family in each tissue as frequency and the flow indicates the shared cell families among tissues. E, The heatmap shows the scaled gene expression of the top five DEGs identified among the colon and liver cell families reported on the right.

Figure 2.

Heterogeneity of TAM transcriptional profiles in the colon peritumoral area. A, UMAP clustering projection of the integrated colon myeloid immune cells from all the patients (ncells = 1,127). B, The dot plot shows the expression of canonical markers used for the annotation of myeloid cell subtypes, in which the dot size indicates the proportion of expressing cells, colored by normalized mean expression levels, and the clusters are reordered according to cell family. C, The nested pie chart shows the percentage frequency distribution of each cluster (outer circle) and the cell family correspondences (inner circle), colored as in Fig. 1C. DC clusters are reported in gray. D, The heatmap shows the scaled gene expression of the top 10 DEGs (rows) per cluster (columns) identified among the colon macrophage clusters (C0, C2–6, and C8). DC clusters (C1, C7, and C9) were excluded from the analysis. E, The radar plots show the mean expression levels of the antigen-presentation (ngenes = 24), lipid-associated (ngenes = 32), and angiogenic (ngenes = 31) scores per cluster. Clusters with the highest scores for each signature are highlighted in red. F, The dot plot shows the expression of immune checkpoint and ligand genes, in which the dot size indicates the proportion of expressing cells, colored by normalized mean expression levels, and the clusters are reordered according to cell family. DC clusters (C1, C7, and C9) were excluded from the analysis.

Figure 3.

Heterogeneity of TAM transcriptional profiles in the liver peritumoral area. A, UMAP clustering projection of the integrated liver myeloid immune cells from all the patients (n cells = 6,127). B, The dot plot shows the expression of canonical markers used for the annotation of myeloid cell subtypes, in which the dot size indicates the proportion of expressing cells, colored by normalized mean expression levels, and the clusters are reordered according to cell family. C, The nested pie chart shows the percentage frequency distribution of each cluster (outer circle) and the cell family correspondences (inner circle), colored as in Fig. 1C. DC clusters are reported in gray. D, The heatmap shows the scaled gene expression of the top 10 DEGs (rows) per cluster (columns) identified among the liver macrophage clusters (L0–8). DC clusters (L9–11) were excluded from the analysis. E, The radar plots show the mean expression levels of the antigen-presentation (ngenes = 24), lipid-associated (n genes = 32), and angiogenic (n genes = 31) scores per cluster. Clusters with the highest scores for each signature are highlighted in red. F, The dot plot shows the expression of immune checkpoint and ligand genes, in which the dot size indicates the proportion of expressing cells, colored by normalized mean expression levels, and the clusters are reordered according to cell family. DC clusters (L9–11) were excluded from the analysis. G, The feature plot shows the expression of the cytotoxic inflammatory score (n genes = 21). DC clusters (L9–11) were excluded from the UMAP plot.

Figure 4.

Transcriptomic correlation and trajectory analysis reveal both similarities and tissue-specific traits in colon and liver TAMs. A, The correlation matrix shows the unsupervised hierarchical clustering computed using the Pearson correlation coefficient taking into consideration all the monocytes (blood) and macrophages (colon and liver) clusters. DC clusters were excluded from the analysis. B, The ordering of different cell clusters along pseudotime in a two-dimensional state-space defined by Monocle2. The pseudotime trajectory was computed on all the monocytes (blood) and macrophages (colon and liver) clusters, without including DC clusters and setting the starting root. Each point corresponds to a single cell, color-coded according to the tissue identity with the proportions of cells per branch. C, Left, the pseudotime trajectory is color-coded according to the pseudotime value, in which each dot corresponds to a single cell. Right, the Monocle2 pseudo-temporal ordering of cells is superimposed on the UMAP plot, in which cells are colored based on their progression along the pseudo-temporal space. D, Cell density distribution plot representing cluster frequency (y-axis) along the pseudotime (x-axis). The median pseudotime value of each distribution is indicated by the solid line, whereas the total mean pseudotime value is indicated by the dotted line.

Figure 5.

Transcriptional profile showing proliferative features of liver immunoregulatory TAMs with large macrophages. A, The circular bar plots show a selection of significantly enriched pathways with FDR values < 0.05, identified among significant DEGs (percent >10% and adjusted P value <0.05 and log₂ fold change >0.25) of inflammatory (left) and immunoregulatory TAMs (right) as the comparison between the colon (orange) and liver (green) using the Reactome pathway browser. The bars are color-coded according to the tissue identity and sized by the −log₁₀(P value) value of the pathway enrichment. The dotted lines separate the different pathway groups. B, The radar plot shows the mean expression levels of proinflammatory (TNF, IL1A, IL1B, IL6, IL23A, CXCL9, CXCL10, and CXCL11) and immune-suppressive (IL10, TGFB1, CCL2, CCL5, CCL17, and CCL22) module scores for the colon and liver inflammatory and immunoregulatory TAMs. C, The bar plot shows the proportion (%) of each TAM cluster per cell-cycle phase (G1, S, and G2M) in the colon and liver. For the statistical analysis, the SEM was calculated according to the absolute cell count, and the unpaired nonparametric Mann–Whitney test was used to determine P values. D, The bar plot shows the module score of cell-cycle genes (ngenes = 57) in the liver. For the statistical analysis, the SEM was calculated according to the absolute cell count, and the unpaired nonparametric Mann–Whitney test was used to determine P values. E, Double IHC for CD163/PCNA on liver metastases. In the peritumoral area of the liver, some cells are double positive for CD163 (red) and PCNA (brown). The nuclei are counterstained with hematoxylin. F, The Kaplan–Meier survival curve shows the proportion of disease-free survival in relation to the CD163⁺PCNA⁺ macrophage frequency in 18 CRLM specimens. Log-rank Mantel–Cox test used to determine P values. G, The PCA plots show the distribution of the liver macrophage clusters (L0–8) across the first three dimensions (PC1–3) based on their similarity according to the large TAM signature (top) or to the small TAM signature (bottom). The red arrows delineate the principal component 1 (PC1) dimension, which captures the highest variance and has been selected for downstream analysis. Dotted circles denote clusters characterized by high PC1 values. H, The box plot shows the distribution of the S and G2M phase gene scores (ngenes = 20) computed in the most representative clusters enriched in the small (L3 and L8) and large (L2, L5, and L7) TAM scores, in which the inner line represents the median value. For the statistical analysis, the unpaired nonparametric Mann–Whitney test was used to determine P values. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. CI, confidence interval; TCA, tricarboxylic acid.