Human Tumor-Associated Macrophage and Monocyte Transcriptional Landscapes Reveal Cancer-Specific Reprogramming, Biomarkers, and Therapeutic Targets

Abstract

The roles of tumor-associated macrophages (TAMs) and circulating monocytes in human cancer are poorly understood. Here, we show that monocyte subpopulation distribution and transcriptomes are significantly altered by the presence of endometrial and breast cancer. Furthermore, TAMs from endometrial and breast cancers are transcriptionally distinct from monocytes and their respective tissue-resident macrophages. We identified a breast TAM signature that is highly enriched in aggressive breast cancer subtypes and associated with shorter disease-specific survival. We also identified an auto-regulatory loop between TAMs and cancer cells driven by tumor necrosis factor alpha involving SIGLEC1 and CCL8, which is self-reinforcing through the production of CSF1. Together these data provide direct evidence that monocyte and macrophage transcriptional landscapes are perturbed by cancer, reflecting patient outcomes.

Keywords: CCL8; SIGLEC1; breast cancer; endometrial cancer; human circulating monocytes; human macrophages; tumor microenvironment.

Graphical abstract

Figure 1

Cancer Alters the Transcriptome of Human Monocytes (A) Principal-component analysis (PCA) plot of n = 12,157 genes expressed in monocytes from healthy individuals (Mo) (n = 45) and TEMo from cancer patients (n = 35; breast cancer [BrCa] = 32; endometrial cancer [EnCa] = 3). (B) Hierarchical clustering of all differentially expressed genes (DEGs) between Mo and TEMo. Expression values are Z score transformed. Samples were clustered using complete linkage and Euclidean distance. (C) Gene ontology (GO) analysis of DEGs between TEMo and Mo (blue, downregulated genes; red, upregulated genes). (D) Bar plot of selected DEGs in TEMo (FDR <= 0.05). (E) Expression of CD200R1, TNFSF10, HGF, and ANGPT1 mRNA in Mo and breast TEMo (n = 3–5; independent from the RNA-seq cohort). (F) Relative distribution of non-classical monocytes from healthy controls and BrCa and EnCa patients determined by flow cytometry shown as percentage in the monocyte gate. Cohort 1: Mo, n = 31, BrCa TEMo, n = 22, EnCa TEMo, n = 12. Cohort 2, BrCa and controls only: Mo, n = 18, TEMo, n = 33. (G) ELISA quantification of CX3CL1 and CCL2 levels in the sera of control (CTR) (n = 15) and BrCa patients (n = 45). (H) Expression of CX3CR1 and CCR2 in Mo (n = 10) and breast TEMo (n = 31). Data are expressed as geometric mean (Geo mean). (I and J) Confusion matrix (I) and summary of results of Recursive Feature Elimination with Random Forest (RFE-RF) classification on the testing set (n = 22) for breast TEMo (J). (K) Receiver operating characteristic curves of RFE-RF classification in the training and test set. (E and H) Data depicted as means ± SEM; (F and G) horizontal bars represent the mean of the individual values ± SD; (E–H) Student's t test; ^∗p < 0.01, ^∗∗p < 0.001, ^∗∗∗p < 0.0001. See also Figure S1 and Table S1.

Figure 2

TAMs from Breast and Endometrial Cancers Exhibit Cancer-Specific Transcriptional Profiles (A) PCA plot of n = 13,668 genes expressed in breast tissue-resident macrophages (Br-RM) (n = 4) and breast cancer TAMs (Br-TAM) (n = 4). (B) Hierarchical clustering of all DEGs between Br-RM and Br-TAM. Expression values are Z score transformed and samples clustered using complete linkage and Euclidean distance. (C) GO analysis of DEGs between Br-TAM and Br-RM (blue, downregulated genes; red, upregulated genes). (D) Bar plot of selected DEGs in Br-TAM (FDR ≤ 0.05). (E) Venn diagram of commonly regulated transcripts in Br-TAM and TEMo (red, upregulated; blue, downregulated). (F) PCA plot of n = 13,739 genes expressed in endometrial tissue-resident macrophages (En-RM) (n = 5) from healthy individuals and endometrial cancer TAMs (En-TAM) (n = 9). (G) Hierarchical clustering of all DEGs between En-RM and En-TAM. Expression values are Z score-transformed and samples clustered using complete linkage and Euclidean distance. (H) GO analysis of DEGs between En-TAM and En-RM (blue, downregulated genes; red, upregulated genes). (I) Bar plot of selected DEGs in En-TAM (FDR ≤ 0.05). (J and K) Venn diagram of commonly regulated transcripts between En-TAM and TEMo (J) and En-TAM and Br-TAM (red, upregulated; blue, downregulated) (K). See also Figure S2, and Table S2.

Figure 3

Breast TAM Signature Is Associated with Clinical Outcomes (A and B) Boxplot showing TAM signature score stratified by the CSF1 signature (A) and across breast cancer subtypes in cohort 3 (n = 47) (B). (C) TAM signature score across PAM50 molecular subtypes in the METABRIC cohort (n = 1,350). (D) Disease-specific survival of the METABRIC cohort according to the TAM signature expression. Boxplots depict the first and third quartiles, with the median shown as a solid line inside the box and whiskers extending to 1.5 interquartile range from first and third quartiles. (A–C) One-way ANOVA with Tukey's post hoc multiple comparisons test (^∗∗∗p < 0.0001). (D) The p value is based on the Wald test. See also Table S3.

Figure 4

Breast TAM Transcriptomes Are Associated with Clinical Outcomes and Reveal TAM-Specific Markers (A) Scatterplot showing Pearson's correlation between CD163 and SIGLEC1 expression in the METABRIC cohort. Red line indicates local regression (LOESS) fit. (B) Disease-specific survival according to the mRNA level of SIGLEC1 in the METABRIC cohort. (C) Expression of SIGLEC1 mRNA in Br-RM (n = 4) and Br-TAM (n = 6). (D and E) SIGLEC1 expression in the Finak et al. (2008) dataset (left) and the Karnoub et al. (2007) dataset (right). Expression calculated from the median centered normalized values. The p values were estimated using a Wilcoxon rank-sum test. Boxplots depict the first and third quartiles, with the median shown as a solid line inside the box and whiskers extending to 1.5 interquartile range from first and third quartiles (D). Data points beyond the limit of lines represent outliers (black dots). CD163 and SIGLEC1 immunofluorescent (IF) staining (n = 5) (E). Stains from cancer (top) and benign sample (bottom) are shown representative of n = 12 independent tumors analyzed. Single channels and merge are shown. Inset representing a double-positive SIGLEC1 and CD163 macrophage (top) and a single-positive CD163 macrophage (bottom). Scale bars, 50 μm, and 5 μm (inset). (F) Quantification of CD163⁺ (left), SIGLEC1⁺ (center), and CD163⁺ and SIGLEC1⁺ (right) cells per mm² of tissue in benign (n = 4) and breast cancer samples (n = 8). Boxplots depict the first and third quartiles, with the median shown as a solid line inside the box and whiskers extending to 1.5 interquartile range from first and third quartiles. (G and H) SIGLEC1 expression in primary MDM- (G) and PMA-treated THP1 cells (H) stimulated for 24 h with culture medium (CTR) normalized as 1, MDA-MB-231 conditioned medium (CM) or MDA-MB-468 CM. Data are depicted as fold change versus CTR (n = 3). (I and J) Flow cytometric analysis of SIGLEC1 expression in iPSDM cells without stimulation (CTR) or stimulated with MDA-MB-231 (I) or MDA-MB-468 (J) CM (n = 3). (K) TNF-α levels in supernatants of iPSDM incubated for 24 h with CTR plus isotype control or CTR plus anti-TNF-α antibody. Same conditions are shown for MDA-MB-231 and MDA-MB-468 CM (n = 3). Results are expressed as pg/mL. (L) Expression of TNFA mRNA in Br-RM (n = 4) and Br-TAM (n = 6). (M) TNF-α protein levels in supernatants of MDM incubated for 24 h with MDA-MB-231 and MDA-MB-468 CM or CTR. Results are expressed as optical density at 450 nm (OD₄₅₀) (n = 3). (N and O) SIGLEC1 mRNA expression in iPSDM stimulated for 24 h with MDA-MB-231 CM normalized as 1 (CTR), MDA-MB-231 CM + TNF-α neutralizing antibody and MDA-MB-231 CM + isotype control antibody (N) or with MDA-MB-468 CM normalized as 1 (CTR), MDA-MB-468 CM + TNF-α neutralizing antibody and MDA-MB-468 CM + isotype control antibody (O) (n = 3 each). (C and L) Horizontal bars represent the mean of the individual values ± SD; (G–K and M–O) data depicted as means ± SEM; (B) The p value is based on the Wald test; (C, D, I, J, and L) Student's t test; (F) two-way ANOVA; (H, K, and M–O) one-way ANOVA; ^∗p < 0.01, ^∗∗p < 0.001, ^∗∗∗p < 0.0001, ^∗∗∗∗p < 0.00001. See also Figures S3 and S4 and Table S4.

Figure 5

TAMs and Cancer Cells Engage in Cytokine Feedback Loops to Support CCL8 and SIGLEC1 Expression in Breast Cancer TAMs (A and B) Volcano plot showing genes whose expression was significantly (Log₂FC ± 1, p < 0.05) deregulated in PMA-THP1 cells after incubation with MDA-MB-231 (A) or MDA-MB-468 (B) CM for 24 h (n = 3 each). (C) Venn diagram of commonly upregulated transcripts between MDA-MB-231-treated (left circle) and MDA-MB-468-treated (right circle) THP1 cells. (D) Selection of pro-inflammatory genes commonly upregulated in Br-TAM (n = 4) (from RNA-seq analysis) and PMA-THP1 (n = 3) (qPCR). (E) Scatterplot showing Pearson's correlation between CD163 and CCL8 expression in the METABRIC cohort. Red line indicates local regression (LOESS) fit. (F) Disease-specific survival according to the mRNA level of CCL8 in the METABRIC cohort. (G) CCL8 mRNA expression in Br-RM (n = 4) and Br-TAM (n = 7). Data are expressed as fold change versus Br-RM. (H) CCL8 levels in CM from MDA-MB-231, MDA-MB-468, MDM, and MDM incubated for 24 h with the two cancer cell CM, respectively (n = 3). (I) IF and fluorescence in situ hybridization for CCL8 mRNA (top) or a DapB-control RNA (bottom) in breast cancer samples. Scale bars, 10 μm (n = 3). Inset representing a SIGLEC1⁺CD163⁺ macrophage-expressing CCL8 mRNA (top) or DapB-control mRNA (bottom). XY, XZ, and YZ projections are shown (right panels). (J and K) CCL8 mRNA expression in iPSDM stimulated for 24 h with MDA-MB-231 CM normalized as 1 (CTR), MDA-MB-231 CM + TNF-α neutralizing antibody and MDA-MB-231 CM + isotype control antibody (J), or with MDA-MB-468 CM normalized as 1 (CTR), MDA-MB-468 CM + TNF-α neutralizing and MDA-MB-468 CM + isotype control antibody (K) (n = 3 each). (L and M) CSF1 levels (L) and TNF-α and IL-1β levels (M) in supernatants from unstimulated MDA-MB-231 or MDA-MB-468 (CTR), and MDA-MB-231 or MDA-MB-468 incubated for 24 h with 10 or 20 ng/mL (or 20 ng/mL for CSF1) of rCCL8 (n = 3 each). (N) In vitro scratch assay of untreated MDA-MB-231 or treated with CCL8 or CCL2 for the indicated period of time, yellow line = cell culture margins (n = 4). Scale bars, 500 μm. (O) Quantification of in vitro scratch assay covered by MDA-MB-231 after 24 h (calculated as area covered at 24–1 h) in untreated (CTR), and CCL8- and CCL2-treated cells. Same symbols represent mean of technical replicates (n = 4). (P) THP1 chemotaxis assay for CCL2 and CCL8. Cells were incubated with medium alone (CTR) or with 20 ng/mL of rCCL2 or rCCL8. Results shown as fold change versus CTR at 72 h (n = 3). (H, J, K, M, and P) Data depicted as mean ± SEM; (G and L) horizontal bars represent the mean of the individual values ± SD; (O) horizontal bars represent the mean of the individual values; (F) the p value is based on the Wald test; (G and L) Student's t test; (H, J, K, M, and P) one-way ANOVA; (O) two-way ANOVA; ^∗p < 0.01, ^∗∗p < 0.001, ^∗∗∗p < 0.0001. See also Figures S5 and S6, Table S5.

Figure 6

High Expression of SIGLEC1/CCL8 Is Associated with Poor Outcome in Breast Cancer Patients (A) Heatmap and recurrence-free survival according to mRNA levels of SIGLEC1 and CCL8 in the breast cancer stroma dataset (Finak et al., 2008). (B and C) Heatmap and disease-specific survival in all (B) and ER⁺Her2⁻ (C) patients from the METABRIC cohort. All significant cutoff points (p < 0.05) are shown in black. Black vertical lines indicate positivity for ER and Her2 expression or grade III tumors. All p values are based on the Wald test.

Figure 7

Schematic Representation of the Crosstalk between Br-TAM and Cancer Cells Tumor cells upregulate SIGLEC1, TNF-α, and CCL8 expression in Br-TAM. In turn, cancer cells respond to CCL8 stimulation by producing CSF1, IL-1β, and TNF-α, which further contribute to the positive feedback loop.