Tumor-associated macrophages (TAM) play a detrimental role in triple-negative breast cancer (TNBC). In-depth analysis of TAM characteristics and interactions with stromal cells, such as cancer-associated fibroblast (CAF), could provide important biological and therapeutic insights. Here we identify at the single-cell level a monocyte-derived STAB1+TREM2high lipid-associated macrophage (LAM) subpopulation with immune suppressive capacities that is expanded in patients resistant to immune checkpoint blockade (ICB). Genetic depletion of this LAM subset in mice suppressed TNBC tumor growth. Flow cytometry and bulk RNA sequencing data demonstrated that coculture with TNBC-derived CAFs led to reprogramming of blood monocytes towards immune suppressive STAB1+TREM2high LAMs, which inhibit T-cell activation and proliferation. Cell-to-cell interaction modeling and assays in vitro demonstrated the role of the inflammatory CXCL12–CXCR4 axis in CAF–myeloid cell cross-talk and recruitment of monocytes in tumor sites. Altogether, these data suggest an inflammation model whereby monocytes recruited to the tumor via the CAF-driven CXCL12–CXCR4 axis acquire protumorigenic LAM capacities to support an immunosuppressive microenvironment.

Significance:

This work identifies a novel lipid–associated macrophage subpopulation with immune suppressive functions, offering new leads for therapeutic interventions in triple-negative breast cancer.

Tumor-associated macrophages (TAM) play an important role in cancer development (1). In breast cancer, high TAM density is a poor prognostic factor (2, 3); however, given the substantial biological and therapeutic diversity of breast cancer subtypes, transcriptomic studies of the human myeloid compartment in heterogeneous patient cohorts have some biological limitations (4, 5). Triple-negative breast cancer (TNBC) lacks specific therapeutic targets and is heterogeneous in its immune composition and natural history (6–9). Transcriptomic signatures highlight the presence of immune-activated and -suppressive subtypes, with distinct prognosis and clinical outcome (8, 10, 11). Likewise, bulk signatures of inflammation in solid tumors poorly segregate responders from nonresponders to immune checkpoint blockade (12, 13), calling for in-depth interrogation of specific immune subsets, including human myeloid cells. Two broad functional states were described in TAM: proinflammatory, associated with tumor protection and tissue reparative, associated with tumor growth (14). However, recent technological advances revealed the complexity of TAM, in terms of origin, cell plasticity, and tissue specificity (4, 15–17). Novel subset-specific markers were recently identified, including TREM2, an immune suppressive receptor and potential target for modulating of tumor-infiltrating myeloid cells (18). In addition to the in-depth characterization of myeloid subsets in TNBC, the understanding of their interactions with stromal cells, such as cancer-associated fibroblasts (CAF), is critical to the definition of the tumor microenvironment (TME) ecosystem. CAFs, which express fibroblast activation protein (FAP), and integrin β1 (CD29) are essential for tumor growth, metastasis (19), and support of immune suppressive regulatory T cells (Treg; ref. 7). Like TAMs, human CAFs are heterogeneous; single-cell RNA sequencing (scRNA-seq) of FAP+ CAF isolated from human breast cancer shows inflammatory (iCAF) and myofibroblastic (myCAF) clusters (20). Operational networks, however, between myeloid cells and CAFs have yet to be elucidated.

In this study, we deconvoluted at the single-cell level the heterogeneity of TNBC TAM from juxta-tumor and tumor samples from patients with early-stage, treatment-naïve TNBC. We identified—at RNA and protein levels—the monocyte (mo)-derived STAB1+TREM2high lipid-associated macrophage (LAM), an immune suppressive subpopulation expanded in patients with TNBC resistant to anti–PD-1. The depletion of these cells in Trem2–/– mice reduced TNBC growth, demonstrating their key function in tumorigenesis. We discovered that human FAP+ CAFs, derived from TNBC tissues, induced the reprogramming of monocytes towards the STAB1+TREM2high suppressive LAM phenotype. Finally, we pinpointed the involvement of the CXCL12–CXCR4 axis in the FAP+ iCAF–myeloid cell–cell interaction, which promotes recruitment of the monocytes at the tumor site, ultimately shaping the myeloid landscape of TNBC and suggesting future therapeutic leads.

Breast cancer patient cohorts and tissue specimens

Human sample collection and use for the studies were approved by Institutional Review Board and Ethics committee of the Institut Curie (Paris, France; Feb 12, 2014) and Commission Nationale de l'informatique et des Libertés (CNIL; N. approval: 1674356; SCANDARE, NCT03017573), and the national ethics committee. The studies were conducted in accordance with the recognized ethical guidelines of Helsinki Declaration. Tumor (TUM) and adjacent juxta-tumor (JT) tissue samples were from early-stage, treatment-naïve TNBC (N = 32), and treatment-naïve luminal breast cancer (LBC; N = 21) patients. Normal mammary tissues (N = 2) from prophylactic mastectomies were included in the study. For all the patients included in the study we obtained written informed consent. Invasive ductal carcinoma of the TNBC histology was defined by estrogen receptor (ER), progesterone receptor (PR), and, ERBB2 (HER2/Neu) negativity; LBC was defined by positive immunostaining for ER and/or PR. The cut-off used to define hormone receptor positivity was 10% of stained cells according to Cardoso and colleagues (21). For scRNA-seq studies, fresh TUM specimens from treatment-naïve TNBC (N = 4) patients were obtained, of which, 3 were paired with JT samples.

Normal mammary gland, TUM, and JT specimens both from patients with treatment-naïve TNBC (N = 29) and LBC (N = 21), were included in the flow cytometry (FC) analysis. Buffy coats from healthy donors (HD) were obtained from Etablissement Français du Sang (Paris, France). Blood from patients with treatment-naïve TNBC was collected (N = 20).

Myeloid cell sorting, scRNA-seq, and library preparation

HLA-DR+CD11c+ myeloid cells from TUM- and JT-infiltrating cells were FACS sorted (ARIA-BD Biosciences). DAPI solution (BD Biosciences) used for dead cells exclusion. CD11c (BioLegend, catalog no. 301605, RRID:AB_314175), HLA-DR (BioLegend, catalog no. 307616, RRID:AB_493588), CD45 (BD Biosciences, catalog no. 557833, RRID:AB_396891), CD3-CD56-CD19 (BioLegend, catalog no. 300424, RRID:AB_493741; BD Biosciences, catalog no. 557919, RRID:AB_396940; BD Biosciences, catalog no. 557921, RRID:AB_396942) were included for the staining. Single-cell suspension was loaded into a Chromium Single Cell Chip (10× Genomics) according to the manufacturer's instructions. Target capture rate correspond to 5,000 to 10,000 individual cells/sample. scRNA-seq libraries were prepared using Chromium Single Cell 3′ v3 Reagent Kit (10X Genomics, catalog no. CG000183) according to manufacturer's protocol. Indexed libraries were equimolarly pooled and sequenced on an Illumina NovaSeq 6000 system using paired-end 28 × 91 bp as sequencing mode. A coverage corresponding to 50,000 reads/cell was obtained. For further details see the Supplemental material.

CAF primary cell lines, culture with monocytes, and migration assay

FAP+CD29+ and FAPCD29+ CAF primary cell lines were established based on previously described protocols (7, 19, 22). For further details about migration assay see the Supplemental material.

Establishment of CAF primary cell lines and culture with monocytes

Both FAP+CD29+ and FAPCD29+ CAF primary cell lines (from maximum 6–9 culture passages) were seeded in DMEM (10% FBS and 1% Pen-Strep) in 6-multiwell at the concentration of 2 × 104 per well; complete DMEM (Hyclone, catalog no. SH30243.01) was replaced at 24 hours and 5 × 105 CD14+ monocytes enriched by Miltenyi kit (Miltenyi, catalog no. 130–050–201; from HD and TNBC) were added. Both FAP+CD29+ and FAPCD29+ CAFs plus CD14+ monocytes were incubated 3 days at 37°C. Each experiment was performed three times, including four HD and four TNBC patients, in triplicates, for a total of 12 HD and 12 TNBC pts. Freshly isolated (day 0) and day 3 monocytes were evaluated by FC for both cocultures. FAPCD29+ CAF cocultures were performed twice.

FAP+CD29+ were seeded as described above, after ON CAF-supernatants were collected and freeze at −20°C.

Miltenyi freshly isolated CD14+ monocytes from HD (N = 3) and TNBC (N = 3) were cultured for 3 days with FAP+CD29+ supernatants. CAF supernatants were obtained from three different CAF cell lines. To assess the differentiation, FC analysis was performed at day 3.

Mono-derived LAM-T cell cocultures in vitro

Circulating CD14+ monocytes from HD and TNBC patients were coculture with FAP+CD29+ and FAPCD29+ CAF as described above. Upon 3 days of monocyte-CAF cocultures, mono-derived LAMs were recovered and then cocultured at 1:1 ratio with autologous CD3+ T cells (Pan T-cell isolation kit, Miltenyi, catalog no. 130–095–130) from either peripheral blood mononuclear cell (PBMC) of HD or patients with TNBC. CD3+ T cells were prelabeled with carboxyfluorescein diacetate succinimidyl ester (CFSE; 5 μmol/L; Invitrogen) for 15 minutes at 37°C. 96 plates were precoated with soluble anti-CD3 (1μg/mL; Clone OKT3, Thermo Fischer Scientific, catalog no. 16–0037–81) in PBS1x, 2 hours at 37°C. Mono-derived LAM and CD3+ T cells were cocultured in complete RPMI medium (Gibco, catalog no. 61870036; 10% FBS and 1% Pen-Strep) for 3 days, plus IL2 (100 U/mL). After 3 days, T cells were assessed for CFSE dilution, PD1, and CD25 expression. Three independent experiments were performed.

Animals

Mice used in this study include wild-type (WT) C57BL/6J and Trem2–/– animals bred at Washington University School of Medicine (St. Louis, MO) animal facility. All animals were backcrossed until at least more than 98% C57BL/6J confirmed by genotype-wide microsatellite typing. Mice were housed under specific pathogen-free conditions, were all females and age matched. Mice from different genotypes (Trem2+/+ and Trem2–/–) were cohoused from birth and separated during the experiment (the day of tumor injection). Mice did not undergo any procedures prior to their stated use and were injected at 8 to 10 weeks of age. All studies performed on mice were done in accordance with the Institutional Animal Care and Use Committee (IACUC) at Washington University in St. Louis. IACUC at Washington University in St. Louis have approved these studies.

Tumor models

PY8119 cells (purchased from ATCC, CRL-3278) were washed and resuspended in PBS. 5 × 105 cells were injected into the mammary fat pad (intramammary) that was previously shaved. Mice were monitored every day and tumors were measured by caliper every other day. Mice were sacrificed at day 15 and or day 30.

For FC in human and mouse, see Supplemental material.

Statistics

Wilcoxon matched-pairs test, two-tailed; paired matched t test, two tailed; unpaired t test, two-tailed, were applied to compare groups of FC analysis, levels of cytokines/chemokines by Luminex and in vitro CAF-monocytes cocultures. Two-way ANOVA test (Tukey multiple comparison test) performed for the migration assays in vitro. Mann–Whitney test was used for mice experiments. P < 0.05 was considered statistically significant in all tests. Correlations were calculated using the nonparametric Spearman and/or Pearson correlation test, two-tailed.

Mouse data were shown as mean ± SEM. Two-way ANOVA for repeated measures was used to model longitudinal tumor growth between groups followed by posthoc comparisons on group difference at time points. Mann–Whitney U test was used to compare two groups. Statistics were calculated with GraphPad Prism 6 and 8 (GraphPad Software).

Data availability

Data generated from this study are publicly available in Gene Expression Omnibus (GEO) repository: GSE206637 (bulk data) and GSE206638 (single-cell data).

Heterogeneity of the myeloid cell landscape of TNBC by scRNA-seq

We performed scRNA-seq analysis to capture the heterogeneity of the myeloid cell compartment in TNBC. First, we isolated all mononuclear phagocytes by FACS-sorting the myeloid HLA-DR+CD11c+ cell fraction from both tumor and adjacent JT tissues obtained from 7 surgical breast specimens from 4 patients with early-stage, treatment-naïve TNBC (Fig. 1A; Supplementary 1A gating strategy). The transcriptomes of individual myeloid cells were generated using the 10× Genomics platform. After quality control and filtering (Supplementary Fig. S1B–S1C), we obtained a total of 14,134 HLA-DR+CD11c+ myeloid cells.

Figure 1.

Unsupervised analysis identifies 16 clusters of HLA-DR+CD11c+ myeloid cells. A, Schematic representation of our scRNA-seq experimental workflow of HLA-DR+CD11c+ myeloid cells. B, UMAP representation of 14134 myeloid cells (JT and TUM merged samples) identifying 16 clusters by Seurat pipeline. C, Heatmap representation showing the genes (both by DEGs and classical markers) for each cluster, averaged by cluster. D, Feature plots of selected genes for each myeloid cluster. E, Cluster-averaged heatmap of inhibitory and stimulatory checkpoints in merged (JT and TUM) monocytes-MAC clusters. ICP, immune checkpoint.

Figure 1.

Unsupervised analysis identifies 16 clusters of HLA-DR+CD11c+ myeloid cells. A, Schematic representation of our scRNA-seq experimental workflow of HLA-DR+CD11c+ myeloid cells. B, UMAP representation of 14134 myeloid cells (JT and TUM merged samples) identifying 16 clusters by Seurat pipeline. C, Heatmap representation showing the genes (both by DEGs and classical markers) for each cluster, averaged by cluster. D, Feature plots of selected genes for each myeloid cluster. E, Cluster-averaged heatmap of inhibitory and stimulatory checkpoints in merged (JT and TUM) monocytes-MAC clusters. ICP, immune checkpoint.

Close modal

Unsupervised clustering analysis identified 16 clusters and projected using Uniform Manifold Approximation and Projection (UMAP) algorithm (Fig. 1BC). UMAP representation of tumor and JT samples showed similar distribution of single cells considering each patient and each sample analyzed (Supplementary Fig. S1D–S1E). Differential expression analyses revealed cluster-specific transcriptomic profiles (Fig. 1C). 10 clusters of monocytes (based on the expression of CD14, VCAN, S100A8, S100A9) and macrophages (MAC; based on the expression of MRC1, CD163, MSR1). Three monocyte clusters were identified as Mono-S100A8, Mono-FCGR3A, and Mono-TREM1. Mono-S100A8 displayed differential expression of S100A8, S100A9, VCAN, THSB1 monocytes genes; Mono-FCGR3A exhibited high levels of FCGR3A, LST1, and CD52 expression (Fig. 1C). Both clusters showed a trending expansion in JT (Supplementary Fig. S1F). In addition, we identified a Mono-TREM1 cluster, differential expressing S100A8–9, VCAN, CSTB, and high levels of TREM1 (Fig. 1BD). These monocyte clusters showed low levels of immune-stimulatory or inhibitory markers (Fig. 1E).

We also detected two macrophage clusters at early state of differentiation. Early–MAC-ISG15 differentially expressed ISG and IFN genes (i.e., CXCL11, CXCL10), IDO1 and CD274, upregulated upon IFNγ response (Fig. 1CE). The early–MAC-CXCR4 cluster was similarly distributed in tumor and JT (Supplementary Fig. S1F).

Five clusters of MACs were identified (Fig. 1BD). MAC-CXCL2 cluster characterized by differential expression of C5AR1 was similarly distributed in tumor and JT samples. MAC-FBP1 cluster, partially expanded in tumor (Supplementary Fig. S1F), differentially expressed MT1X-G-H (metallothionines) genes, and FN1, a profibrotic gene. MAC-FCGBP cluster was also detected (Fig. 1BC). We identified two subsets of LAMs (based on the LAM signature from Jaitin and colleagues (23), sharing the expression of APOE, TREM2, GPNMB, and APOC1 genes. We proposed to name them, LAM-APOC1 and mo-derived LAM-STAB1. Mo-derived LAM-STAB1 were defined as such based on a combination of monocytic marker expression and trajectory models (Supplementary Fig. S2A–S2C); in addition, they displayed differential expression of LYVE-1, STAB1, FOLR2, and CD209 genes, similarly reported in hepatocellular carcinoma (HCC), where a FOLR2+ mo-derived MAC subset was enriched (Fig. 1BD; ref. 15). Mo-derived LAM-STAB1 showed high expression of CD276 and PDCD1 immune checkpoints (Fig. 1E). The LAM-APOC1 cluster displayed differential levels of APOC1, complement component genes such as C1QA, C1QB, C1QC (Fig. 1BD), and high expression of LAG3, PDCD1LG2, and CD200 (Fig. 1E). LAM-APOC1 cells were similarly distributed in JT and tumor; while, mo-derived LAM-STAB1 were rather expanded in tumor (Supplementary Fig. S1F). In addition, two clusters of cycling cells: cycling 1-PTTG1 and cycling 2-MCM7 that exhibited genes involved in cell proliferation (i.e., PTTG1, MKI67, HMG) were identified, suggestive of a potential turnover of myeloid cells in the TME (Fig. 1BE).

Our gating strategy also captured dendritic cells (DC), which included four DC clusters with distinct transcripts (Fig. 1,BD). A small cDC1-CLEC9 cluster expresses CLEC9A, CPNE3, C1orf54, CPVL, SNX3, and IRF8, transcription factors (TF) of cDC1 lineage differentiation (24) particularly enriched, compared with the other DC clusters, in both costimulatory TNFSF9, ICOSLG, CD226, and inhibitory LAG3- and PDCD1-associated genes and slightly enriched in JT (Supplementary Fig. S1G). Two clusters of type 2 DCs, cDC2-CD1c and DC-AXL, mainly characterized by the coexpression of CD1C, CLEC10A, IL2RA, the former expressing AREG (Fig. 1C). cDC2-CD1c cells showed differential expression of TLR4, TLR5, and STAT4 expression, associated with DC activation and antitumor properties (25, 26) and low levels or absence of inhibitory immune checkpoints; the DC2-AXL were characterized by high expression of MMPs genes (MMP1, 2, 9, 12, 14), AXL, CD276, and HAVCR2—potent inhibitors of T-cell functions; CD200R1 and ENTPD1—regulators of immune suppression in the TME, potentially suggestive of an immune suppressive role for the tumor-expanded DC2-AXL (Supplementary Fig. S1F). We then identified a cluster of activated cDC2-LAMP3 cells, characterized by CCR7, CCL19, CCL22, and MARCKSL1 genes (Fig. 1BD). cDC2-LAMP3 were similarly distributed in tumor and JT (Supplementary Fig. S1F) and exhibited higher expression of stimulatory genes, compared with the other DC subpopulations, in addition to immune checkpoints i.e., TNFSF4, TNFRSF4, TNFRSF9, TNFRSF14, ICOS, CD80, and IDO1 and costimulatory receptors like CD28, CD40, and CD27 (Supplementary Fig. S1G).

APOE, a common tumor-associated marker expressed in LAM

To investigate cluster specificities in tumor and JT tissues, we analyzed differentially expressed genes (DEG) in the two series of samples. In line with the distributions in Supplementary Fig. S1F, density plot analysis showed an accumulation of monocyte clusters and cDC2-CD1c in JT; whereas early-MAC and LAM clusters were enriched in tumors (Fig. 2A). DEGs analysis revealed that, among tumor-associated genes (i.e., SPP1, APOC1, GPNMB, and OLR1), APOE was expressed in all MAC-clusters (including LAM-APOC1 and mo-derived LAM-STAB1); while, among JT-associated genes (i.e., CFP, IFITM2, CD52, CFD) S100A9 was high in monocytes subclusters (Fig. 2B and C). To verify APOE and S100A9 protein expression levels, we tested by FC the frequency of APOE+/– and/or S100A8–9+/– cells in tumor-infiltrating HLA-DR+CD14+ myeloid cells from patients with TNBC.

Figure 2.

APOE, a common tumor-associated marker of LAM. A, Density plot represented by UMAP visualization displaying myeloid cellular density in JT and TUM. B, Volcano plot showing DEGs comparing total JT and TUM across all merged clusters. Wilcoxon test. C, Violin plots displaying S100A9 and APOE gene expression in JT and TUM (white line median of expression). ***, P < 0.001 by unpaired t test. D, Representative APOE/S100A8–9 FC staining plots gated on CD14+HLA-DR+ myeloid cells in JT and TUM. Percentage of APOE+S100A8–9, APOE+S100A8–9+, APOES100A8–9+ cells in CD14+HLA-DR+ gated myeloid cells in JT and TUM (N = 13). *, P < 0.05 by paired t test, two-tailed. E, MFI expression of CD206 (N = 9), CD204 (N = 8), CD163 (N = 8), and CD9 (N = 5) evaluated by FC in APOE+S100A8–9, APOE+S100A8–9+, APOES100A8–9+ gated on myeloid cells in TUM. *, P < 0.05; **, P < 0.01 by unpaired t test, two-tailed. F, Chemokine concentration in JT- and TUM-CM from untreated TNBC samples (N = 78). **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 by paired t test, two-tailed.

Figure 2.

APOE, a common tumor-associated marker of LAM. A, Density plot represented by UMAP visualization displaying myeloid cellular density in JT and TUM. B, Volcano plot showing DEGs comparing total JT and TUM across all merged clusters. Wilcoxon test. C, Violin plots displaying S100A9 and APOE gene expression in JT and TUM (white line median of expression). ***, P < 0.001 by unpaired t test. D, Representative APOE/S100A8–9 FC staining plots gated on CD14+HLA-DR+ myeloid cells in JT and TUM. Percentage of APOE+S100A8–9, APOE+S100A8–9+, APOES100A8–9+ cells in CD14+HLA-DR+ gated myeloid cells in JT and TUM (N = 13). *, P < 0.05 by paired t test, two-tailed. E, MFI expression of CD206 (N = 9), CD204 (N = 8), CD163 (N = 8), and CD9 (N = 5) evaluated by FC in APOE+S100A8–9, APOE+S100A8–9+, APOES100A8–9+ gated on myeloid cells in TUM. *, P < 0.05; **, P < 0.01 by unpaired t test, two-tailed. F, Chemokine concentration in JT- and TUM-CM from untreated TNBC samples (N = 78). **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 by paired t test, two-tailed.

Close modal

APOE+S100A8–9 MACs and APOE+S100A8–9+ double-positive early-MACs were significantly enriched in tumor, as opposed to JT (Fig. 2D). Conversely, APOES100A8–9+ monocytes were accumulated in JT (Fig. 2D). The phenotype of APOE+S100A8–9 MACs by mean fluorescence intensity (MFI) was consistent with protein expression of CD206, CD204, and CD163 as common markers of MACs and of CD9, as a common marker of LAM (23). CD206, CD204, and CD9 were upregulated in APOE+S100A8–9MACs compared with the APOE+S100A8–9+ double-positive early-MACs, and decreased in the APOES100A8–9+ monocyte counterpart (Fig. 2E). CD163 was mainly enriched in the APOE+S100A8–9 MAC subpopulation, although at lower levels than CD206 and CD204 receptors (Fig. 2E).

To explore the contribution of soluble factors in shaping the myeloid cell compartment in the TME, we assessed by Luminex assay the conditioned media (CM) generated from tumor and JT TNBC specimens. While CCL2, CCL3, CCL4, and CCL8, conventional monocytes-recruiting chemokines in the tissue (5) were significantly enriched in JT; CSF-2 and CSF-1 were enriched in tumor (Fig. 2F). Although we do not provide mechanistic evidence, we hypothesize a potential role for CCL2, CCL3, CCL4, and CCL8 chemokines in monocyte recruitment to the surrounding JT, and a potential role for the CSF-1 and CFS-2 in inducing the MAC differentiation in the tumor.

LAMs comprise resident and monocyte-derived MACs

To investigate the origin of LAMs in TNBC, we evaluated—by trajectory inference analysis—the differentiation fate of monocytes. Using the two complementary algorithms Monocle-3 and PHATE, we identified a directional flow originating from monocytes (the origin based on the monocyte marker S100A8), respectively mono-FCGR3A, mono-S100A8, and mono-TREM1, to early–MAC-CXCR4, then next on the time scale to early–MAC-ISG15 and MAC-FBP1, suggesting that the early-MAC-CXCR4 cluster was highly plastic. Afterwards, early-MAC-ISG15 was linked to MAC-FBP1 and mo-derived LAM-STAB1 clusters, which suggested to give rise to the MAC-CXCL2 (Supplementary Fig. S2A and S2B). Accordingly, while the LAM-APOC1 cluster was found outside the trajectory, suggesting no direct origin from mono–early-MACs, the mo-derived LAM-STAB1 cells appeared linked from mono–early-MAC subpopulations, suggesting that LAM-APOC1 and the mo-derived LAM-STAB1 may have different origins. Both LAM subpopulations shared the expression of FOLR2 (Fig. 1C), consistent with recent data in HCC (15). Indeed, comparison of HCC-resident versus monocyte-derived FOLR2+ TAM signatures revealed close resemblance between LAM-APOC1 and resident FOLR2+ HCC TAM, and between mo-derived LAM-STAB1 and monocyte-derived FOLR2+ counterparts (Supplementary Fig. S3A and S3B), indicative of the potential coexistence of both resident and monocyte-derived LAMs in TNBC.

Profiling the single-gene expression, we confirmed a strong reduction of monocyte genes (i.e., S100A8–9, TREM1) and the progressive upregulation of MACs genes (i.e., MAFB, TREM2, APOC1, and APOE) along the trajectory, resulting in an increased expression of APOE, APOC1, TREM2, and MAFB detected in early-MACs and upregulated in LAM clusters (Supplementary Fig. S2C). Hypothesizing a key role for inflammation in coordinating the myeloid cell fate in TNBC, high levels of NLRP3 and—as consequence—of the proinflammatory cytokine IL1B were observed in all monocytes, early-MACs and mo-derived LAM-STAB1, as opposed to LAM-APOC1 (Supplementary Fig. S3C and S3D), which suggests a common state of IL1B-associated inflammation in all monocyte-derived subsets. To explore the hypothesis that mo-derived LAM-STAB1 exhibited a more plastic phenotype than LAM-APOC1, we assessed their potential differentiation rate by using CytoTRACE algorithm, which leverages transcript diversity to infer differentiation potential. CytoTrace analysis suggested higher plasticity for the mo-derived LAM-STAB1 subpopulation (Supplementary Fig. S2D, left panel) and lower values of predicted CytoTRACE index (Supplementary Fig. S2D, right panel) for LAM-APOC1, suggesting a terminal differentiation state for these cells. Similarly, by the ROGUE algorithm, a method to quantify cluster purity whereby high ROGUE index indicates high cluster homogeneity, the LAM-APOC1 cells displayed higher homogeneity in both JT and tumor than mo-derived LAM-STAB1(Supplementary Fig. S2E).

Two transcriptionally distinct LAMs identified in the tumor

Recently identified in the adipose tissue of obese subjects (23), CD9+CD63+ LAMs were similarly detected in our SC-dataset. The LAM signature was shared by mo-derived LAM-STAB1 and LAM-APOC1 and mainly enriched in tumor for both subpopulations (Fig. 3A). They shared the expression of APOE and APOC1 apolipoprotein genes; TREM2 triggering lipid receptor; CD9, CD63, ACP5, LIPA, LAMP1, LGMN genes involved in lysosome pathways, and FABP5 fatty acid-binding protein involved in lipid metabolism (Fig. 3B), FOLR2, MSR1, GPNMB were also shared.

Figure 3.

Two transcriptionally distinct LAMs identified in tumor. A, Violin plots of LAM signature (top 100 genes) in JT and TUM of mo-derived LAM-STAB1 and LAM-APOC1 clusters (medians in white; ***, P < 0.001 by unpaired t test). B, Venn diagram representing DEGs in mo-derived LAM-STAB1 and LAM-APOC1 clusters. C and D, Pathway enrichment analysis by Reactome, in total mo-derived LAM-STAB1 and LAM-APOC1 clusters. E, Spearman correlations between JT or TUM mo-derived LAM-STAB1 and LAM-APOC1 myeloid signatures and cytotoxic/experienced CD8 T-cell and Treg signatures by bulk RNA analysis of the METABRIC TNBC cohort (N = 332). F, Volcano plots showing DEGs of mo-derived LAM-STAB1 and LAM-APOC1 clusters in JT versus TUM. Wilcoxon test. G, Bulk RNA analysis of single genes (STAB1, LYVE-1, FOLR2, APOE, TREM2, CD9, APOC1, SPP1, MMP9) by The Cancer Genome Atlas basal-like cohort (N = 212) evaluating JT and TUM specimens. H, Violin plots of CD9, APOE, TREM2, and STAB1 in total mo-derived LAM-STAB1 and LAM-APOC1 clusters (means in white; ***, P < 0.001 by unpaired t test). I, Representative APOE/CD14 and CD9/CD63 FC staining gated on CD14+HLA-DR+ myeloid cells of one TNBC patient. Percentage of CD9+CD63+ cells in the APOE+CD14+ gate from healthy (N = 2), JT (N = 5), and TUM (N = 8; left). Bodipy MFI by FC in CD9+CD63+ cells in healthy (N = 2), JT (N = 6), TUM (N = 7; right). *, P < 0.05; **, P < 0.01; ***, P < 0.001 by paired t test, two-tailed between JT and TUM and unpaired Student t test, two-tailed, between healthy and JT/TUM. J, Cytokeratin, DAPI, APOE, CD11c, and STAB1 expression by IHC in three representative treatment-naïve TNBC samples (left) and the HALO analysis of the triple-positive APOE+CD11c+STAB1+ (mo-derived LAM STAB1+) cells is shown in green (right). Red arrows, presence of triple-positive APOE+CD11c+STAB1+ cells. Quantification of APOE+CD11c+STAB1+ cells is shown in tumor and stroma areas (N = 6). *, P < 0.05 by paired t test. K, Top, UMAP of myeloid cells representing annotated clusters from ref. 29 in anti–PD-1–treated patients with TNBC. Bottom, frequency of the C10-LYVE1 cluster in patients with T-cell–expanded versus nonexpanded TNBC receiving anti–PD-1. Freq, frequency.

Figure 3.

Two transcriptionally distinct LAMs identified in tumor. A, Violin plots of LAM signature (top 100 genes) in JT and TUM of mo-derived LAM-STAB1 and LAM-APOC1 clusters (medians in white; ***, P < 0.001 by unpaired t test). B, Venn diagram representing DEGs in mo-derived LAM-STAB1 and LAM-APOC1 clusters. C and D, Pathway enrichment analysis by Reactome, in total mo-derived LAM-STAB1 and LAM-APOC1 clusters. E, Spearman correlations between JT or TUM mo-derived LAM-STAB1 and LAM-APOC1 myeloid signatures and cytotoxic/experienced CD8 T-cell and Treg signatures by bulk RNA analysis of the METABRIC TNBC cohort (N = 332). F, Volcano plots showing DEGs of mo-derived LAM-STAB1 and LAM-APOC1 clusters in JT versus TUM. Wilcoxon test. G, Bulk RNA analysis of single genes (STAB1, LYVE-1, FOLR2, APOE, TREM2, CD9, APOC1, SPP1, MMP9) by The Cancer Genome Atlas basal-like cohort (N = 212) evaluating JT and TUM specimens. H, Violin plots of CD9, APOE, TREM2, and STAB1 in total mo-derived LAM-STAB1 and LAM-APOC1 clusters (means in white; ***, P < 0.001 by unpaired t test). I, Representative APOE/CD14 and CD9/CD63 FC staining gated on CD14+HLA-DR+ myeloid cells of one TNBC patient. Percentage of CD9+CD63+ cells in the APOE+CD14+ gate from healthy (N = 2), JT (N = 5), and TUM (N = 8; left). Bodipy MFI by FC in CD9+CD63+ cells in healthy (N = 2), JT (N = 6), TUM (N = 7; right). *, P < 0.05; **, P < 0.01; ***, P < 0.001 by paired t test, two-tailed between JT and TUM and unpaired Student t test, two-tailed, between healthy and JT/TUM. J, Cytokeratin, DAPI, APOE, CD11c, and STAB1 expression by IHC in three representative treatment-naïve TNBC samples (left) and the HALO analysis of the triple-positive APOE+CD11c+STAB1+ (mo-derived LAM STAB1+) cells is shown in green (right). Red arrows, presence of triple-positive APOE+CD11c+STAB1+ cells. Quantification of APOE+CD11c+STAB1+ cells is shown in tumor and stroma areas (N = 6). *, P < 0.05 by paired t test. K, Top, UMAP of myeloid cells representing annotated clusters from ref. 29 in anti–PD-1–treated patients with TNBC. Bottom, frequency of the C10-LYVE1 cluster in patients with T-cell–expanded versus nonexpanded TNBC receiving anti–PD-1. Freq, frequency.

Close modal

Interestingly, mo-derived LAM-STAB1 selectively expressed STAB1, MRC1, MAFB, and CD163 genes. Conversely, together with SPP1, recently reported in many tumor-MAC sc-datasets (20, 22) an accumulation of IFN-genes (i.e., CXCL9, CXCL10, CXCL11, ISG15, IFI27) was observed in LAM-APOC1 (Fig. 3B; Supplementary Fig. S3E). A pathway enrichment analysis revealed endosomal/vacuolar pathways, as well, lipid metabolism genes in both LAM subpopulations, however, mo-derived LAM-STAB1 exhibited MMP and collagen degradation pathways, compared with the LAM-APOC1 that displayed antigen cross-presentation, IFNα and -γ signaling pathways (Fig. 3C and D). In order to test the relationship with T-cell subsets, we analyzed by bulk RNA-sequencing from the METABRIC TNBC cohort (27) the correlations between mo-derived LAM-STAB1 or LAM-APOC1 versus cytotoxic and experienced CD8 (28, 29) T cells and Tregs (29). In the JT, LAM-APOC1 were strongly correlated with cytotoxic/experienced CD8 and Tregs as opposed to mo-derived LAM-STAB1 (Fig. 3E). These observations suggest distinct functional properties: immune suppressive for mo-derived LAM-STAB1 and immune activatory for the LAM-APOC1 subset (30).

By gene expression profiling, LAM-APOC1 showed similar expression patterns in JT versus tumor; whereas mo-derived LAM-STAB1 displayed an upregulation of the LAM genes APOE, CD9, APOC1, TREM2, and GPNMB in tumors, in addition to high expression of LYVE-1, STAB-1, and FOLR2 in JT (Fig. 3F). Interestingly, bulk analyses in a TNBC cohort (METABRIC; ref. 27), confirmed high expression of LYVE-1, STAB-1, and FOLR2 in JT, and the enrichment of APOE, CD9, APOC1, and TREM2 in tumor (Fig. 3G). In line with these findings, scRNA violin plot analysis revealed that both LAMs shared CD9 and APOE expression, while TREM2 and STAB-1 were highly expressed in mo-derived LAM-STAB1 (Fig. 3H). FC analysis of APOE+ LAMs showed a striking enrichment of CD9+CD63+ LAMs in tumor, as opposed to JT and normal mammary tissues (Fig. 3I, left panel). In accordance with the high lipid metabolism, the bodipy content (quantifying intracellular neutral lipids) was highly detected in tumor-infiltrating APOE+CD9+CD63+ LAMs, compared with its JT counterpart (Fig. 3I, right panel). To evaluate the spatial distribution of mo-derived LAM-STAB1+, we analyzed samples from untreated, primary TNBC by multicolor IHC. In line with transcriptomic analyses, APOE+CD11c+ STAB1+ were enriched in the stroma localized at the invasive front of the tumor (Fig. 3J; Supplementary Fig. S3F). We further tested whether the mo-derived LAM-STAB1 could be associated with outcome to ICB. Thanks to a recently published dataset in patients with TNBC treated with anti–PD-1 (29) we discovered that bona-fide mo-derived LAM-STAB1 cells, which corresponded to the cluster C10-LYVE1 and expressed high levels of MAF, LYVE1, and STAB1 (Fig. 3K, top panel; Supplementary Fig. S3G), were significantly enriched in patients with TNBC with impaired T-cell expansion (nonexpanded) compared with those, whose T cells expanded under PD-1 blockade (Fig. 3K, bottom panel). On the contrary, the other clusters were equally distributed in the two cohorts (Supplementary Fig. S3H). These data suggest an association of mo-derived LAM-STAB1 cells with an impaired T-cell expansion in nonresponding patients.

Depletion of mo-derived LAM-STAB1 in trem2 KO mice is associated with delay in tumor growth

To test the protumorigenic activity of mo-derived LAM-STAB1, we developed a TNBC mouse model by implanting the TNBC cell line PY8119 in WT and Trem2–/– immunocompetent mice. Highly expressed by mo-derived LAM-STAB1 (Fig. 3H), TREM2 is a key modulator of the myeloid compartment in a sarcoma mouse model, in which TREM2+ MACs displayed an immunosuppressive activity (18). According to cross-reference and label transfer analyses of our human TNBC scRNA-dataset projected into the mouse sarcoma one (Supplementary Fig. S4A–S4C), the mo-derived LAM-STAB1 cluster was significantly enriched in the CX3CR1+-MAC cluster (Maf, Stab1, Folr2, and Mrc1mouse-human shared genes; Supplementary Fig. S4C), was strongly depleted in the sarcoma Trem2–/– mouse model where it showed a protumor activity.

Bulk RNA analysis (METABRIC; ref. 27) revealed that human TREM2 was positively correlated with common LAM genes (MSR1, LIPA, APOE, APOC1, CD63, GPMNB, ACP5), and particularly with STAB1, CD163, MAFB, and MAF mo-derived LAM-STAB1 genes; as opposed to the LAM-APOC1-specific genes CXCL9, ISG15, IFI27, and IFIT1, for which no correlation was seen (Fig. 4A). In PY8119 implanted-TNBC model, we tested tumor growth in WT and Trem2–/– mice. PY8119-TNBC tumor growth was significantly delayed in Trem2–/– compared with Trem2wt mice (Fig. 4B). 15 days after tumor injection, Trem2–/– showed significant reduction in the percentage of macrophages (MHCIIhighLyC6lowLy6GCD11b+); particularly, the CD9+CD63+ LAM-like and the FOLR2+STAB1+CD206high mo-derived LAM-STAB1 macrophages (Fig. 4C and D). A trend in the accumulation of resident FOLR2+(bona fide LAM-APOC1 MAC) was noted. CD11b+ were accordingly depleted and Ly6Chigh monocytes accumulated (Supplementary Fig. S4D and S4E); equal distribution of neutrophils (CD11b+Ly6G+), B and CD4 T cells was observed (Supplementary Fig. S4E). In addition, Trem2–/– mice showed a trend in CD8 accumulation and Treg reduction, resulting in a significantly increased CD8/Treg ratio (Fig. 4E). PY8119 Trem2–/– tumor infiltrates also contained significantly more natural killer T cells (NKT) and natural killer (NK) cells, (Fig. 4E). Collectively, TREM2 deficiency may promote, through the selective depletion of CD9+CD63+ LAM-like and, particularly, FOLR2+STAB1+ mo-derived LAM-STAB1 macrophages, the partial control of PY8119 tumor growth.

Figure 4.

Trem2−/− deficiency controls the tumor growth by the depletion of CD9+CD63+FOLR2+STAB1+ mo-derived LAM-STAB1 macrophages. A, Heatmap showing Spearman correlation between TREM2 and common-LAM, LAM-APOC1, and mo-derived STAB1 genes by bulk RNA analysis in the METABRIC TNBC cohort (N = 332). B, Tumor growth in Trem2+/+ and Trem2–/– mice injected intramammary with PY8119. Data represent mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001 by two-way ANOVA with multiple comparison test. C, t-SNE representation of one representative Trem2+/+ and Trem2–/– tumor gated on CD45+ showing MHCII, FOLR2, CD9, CD63, CD206, and STAB1 expression. Circle, population of interest. D, Graphs showing tumor FC analysis of the percentage of MHCIIhighLyC6low, CD9+CD63+ gated on CD11b+Ly6GLy6Clow and CD206+, FOLR2+STAB1+ and FOLR2+ MAC gated on CD9+CD63+. The frequencies were calculated in CD45+ cells. Trem2+/+ (N = 4); Trem2–/– (N = 6). E, Graphs showing tumor FC analysis of the percentage of NKT, NK gated on CD45+ cells, CD8 T cells and Tregs gated on total T cells, and the ratio of CD8/Tregs cells. Trem2+/+ (N = 4); Trem2–/– (N = 5). *, P < 0.05; **, P < 0.01 by Mann–Whitney t test.

Figure 4.

Trem2−/− deficiency controls the tumor growth by the depletion of CD9+CD63+FOLR2+STAB1+ mo-derived LAM-STAB1 macrophages. A, Heatmap showing Spearman correlation between TREM2 and common-LAM, LAM-APOC1, and mo-derived STAB1 genes by bulk RNA analysis in the METABRIC TNBC cohort (N = 332). B, Tumor growth in Trem2+/+ and Trem2–/– mice injected intramammary with PY8119. Data represent mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001 by two-way ANOVA with multiple comparison test. C, t-SNE representation of one representative Trem2+/+ and Trem2–/– tumor gated on CD45+ showing MHCII, FOLR2, CD9, CD63, CD206, and STAB1 expression. Circle, population of interest. D, Graphs showing tumor FC analysis of the percentage of MHCIIhighLyC6low, CD9+CD63+ gated on CD11b+Ly6GLy6Clow and CD206+, FOLR2+STAB1+ and FOLR2+ MAC gated on CD9+CD63+. The frequencies were calculated in CD45+ cells. Trem2+/+ (N = 4); Trem2–/– (N = 6). E, Graphs showing tumor FC analysis of the percentage of NKT, NK gated on CD45+ cells, CD8 T cells and Tregs gated on total T cells, and the ratio of CD8/Tregs cells. Trem2+/+ (N = 4); Trem2–/– (N = 5). *, P < 0.05; **, P < 0.01 by Mann–Whitney t test.

Close modal

FAP+ CAFs induce the differentiation of blood circulating monocytes towards mo-derived LAMs

Given our observations of an accumulation of mo-derived LAM-STAB1 in the TNBC stroma, we hypothesized a role for CAFs in shaping phenotypical and functional features of myeloid cells in the tumor. First, the proportion of FAP+CD29+ CAFs in breast cancer tissues was quantified ex vivo by FC. FAP+CD29+ CAFs were significantly accumulated in the tumor versus JT, as shown by higher FAP MFI (Supplementary Fig. S5A and S5B). To interrogate the interaction between CAFs and immune cells in the TME, we analyzed the percentage of total tumor-infiltrating CD45+ immune cells. CD45+ cells were highly enriched in the tumor versus JT of TNBC and LBC (Supplementary Fig. S5C). There was a numerically greater accumulation of CD45+ cells in TNBC than in LBC, suggestive of an inflammatory state in TNBC compared with LBCs, thereby confirming previous data from patients with breast cancer (7). In addition, a positive correlation between the percentage of CD45+ cells and FAP+CD29+ CAFs in tumor of patients with breast cancer existed (Supplementary Fig. S5D), suggestive of a brisk immune infiltrate in FAP+CD29+ CAFs-rich tumors. FAP+CD29+ CAFs and APOE+S100A8–9 LAMs were significantly correlated in tumors by FC ex vivo (Supplementary Fig. S5E), which was confirmed by bulk RNA analysis (METABRIC; ref. 27) of mo-derived LAM-STAB1 and FAP+CD29+ CAF signatures in patients with TNBC. No correlation was observed with LAM-APOC1 cluster (Supplementary Fig. S5F).

To evaluate whether FAP+CD29+ CAFs could influence the myeloid cell phenotype, blood monocytes (derived from both HD and patients with TNBC) were cocultured with primary FAP+CD29+ and as control with FAPCD29+ CAF cell lines, established from fresh breast cancer tissues (7, 19). FAP+CD29+ CAFs induced a robust phenotypical differentiation ex vivo of blood monocytes towards mo-derived LAM-STAB1 cells, characterized by a significant upregulation of APOE (Fig. 5A), not observed with monocytes cocultured with FAPCD29+ CAFs (Supplementary Fig. S5G). In line, although the TREM2 receptor was upregulated in circulating monocytes from HDs, its expression was significantly higher in circulating TNBC-monocytes (Fig. 5B; Supplementary Fig. S5G). In addition, the bodipy content was significantly increased at day 3 particularly in TNBC-derived monocytes and exclusively in coculture with FAP+CD29+ CAFs, indeed, no increase in bodipy content was observed in the monocyte coculture with FAPCD29+ CAFs (Fig. 5C; Supplementary Fig. S5G).

Figure 5.

FAP+CD29+ CAFs induce the differentiation of blood circulating monocytes towards mo-derived LAMs. FC analysis of HD- (N = 4) and TNBC-derived (N = 4), magnetically enriched blood monocytes, before (day 0) and after (day 3) CAF-circulating monocyte cocultures in vitro. One representative experiment, performed in triplicates of four is shown. Both FAP+CD29+ and FAPCD29+ primary CAF cell lines were derived from primary TNBC tissues. A and B, Representative APOE/CD14 FC staining and percentages of APOE+ cells in CD14+ gated cells at day 0 and 3 from FAP+CD29+ (left) CAF-monocyte coculture. B and C, MFI of TREM2 and Bodipy by FC in CD14+ gated cells at day 0 and 3 from FAP+CD29+ CAF-monocyte coculture. *, P < 0.05; **, P < 0.01; ***, P < 0.001 by paired t test, two-tailed between day 0 and D3 in HD and TNBC; unpaired Student t test, two-tailed between HD and TNBC. D, Bulk data from CD14+ Miltenyi-enriched monocytes from HD (N = 3) and TNBC (N = 3) cocultured with FAP+CD29+ for 3 hours. Heatmap with selected LAM, monocyte, and suppressive transcripts per million (TPM) gene expression values. TPM values are row z-score normalized. HD (N = 3) and TNBC (N = 3) ex vivo and FAP+CD29+-cocultured monocytes are shown. D0, day 0; D3, day 3.

Figure 5.

FAP+CD29+ CAFs induce the differentiation of blood circulating monocytes towards mo-derived LAMs. FC analysis of HD- (N = 4) and TNBC-derived (N = 4), magnetically enriched blood monocytes, before (day 0) and after (day 3) CAF-circulating monocyte cocultures in vitro. One representative experiment, performed in triplicates of four is shown. Both FAP+CD29+ and FAPCD29+ primary CAF cell lines were derived from primary TNBC tissues. A and B, Representative APOE/CD14 FC staining and percentages of APOE+ cells in CD14+ gated cells at day 0 and 3 from FAP+CD29+ (left) CAF-monocyte coculture. B and C, MFI of TREM2 and Bodipy by FC in CD14+ gated cells at day 0 and 3 from FAP+CD29+ CAF-monocyte coculture. *, P < 0.05; **, P < 0.01; ***, P < 0.001 by paired t test, two-tailed between day 0 and D3 in HD and TNBC; unpaired Student t test, two-tailed between HD and TNBC. D, Bulk data from CD14+ Miltenyi-enriched monocytes from HD (N = 3) and TNBC (N = 3) cocultured with FAP+CD29+ for 3 hours. Heatmap with selected LAM, monocyte, and suppressive transcripts per million (TPM) gene expression values. TPM values are row z-score normalized. HD (N = 3) and TNBC (N = 3) ex vivo and FAP+CD29+-cocultured monocytes are shown. D0, day 0; D3, day 3.

Close modal

Mo-derived LAMs generated in vitro expressed LAM-associated genes and a suppressive/tumor-promoting transcriptional profile

We next investigated the transcriptomic profile by bulk RNA-seq of mo-derived LAMs generated in vitro after a 3-hour coculture with primary FAP+CD29+ CAFs. Although monocytes from HD and TNBC ex vivo segregated slightly in the principal component analysis (Supplementary Fig. S5H), they commonly showed high expression of classical monocyte genes like VCAN, LYZ, S100A8, S100A9, and S100A12 (Fig. 5D); after FAP+CD29+ coculture, both acquired a LAM transcriptional program comparable with that one reported in sc-data of LAM ex vivo (Fig. 3B). These cells displayed increased CD163, APOE, TREM2, CD9, CD63 APOC1, GPNMB, DAB2, MSR1, FABP5, ACP5, and LGMN RNA expression (Fig. 5D). Baseline levels of STAB1, MRC1, and ARG1 were observed higher in TNBC compared with HD, suggesting a potential systemic monocyte preconditioning by the tumor, as already reported in breast cancer (31). In addition, a suppressive/tumor promoting transcriptional profile was increased in HD and TNBC samples by the expression of IL10, VEGFA, IL6, PTGS2, CXCL1, CXCL2, IL1A, IL1B, and CCL2 genes (32), accordingly, we observed suppressive functions for both HD- an TNBC-derived LAM cells (Fig. 6). To explore which signals may contribute to monocyte differentiation and, based on supportive data about the involvement of cytokines in the CAF-monocyte reprogramming (33–36), we set up an assay in vitro in which circulating CD14+ monocytes, from HD and TNBC, were cultured for 3 days with FAP+CD29+-generated supernatants. APOE and TREM2 protein expression at day 3 was higher in FAP+CAF-derived supernatant compared with the control (medium) condition (Supplementary Fig. S6A and S6B). Regarding the soluble profile of the FAP+CD29+-supernatants, we detected IL6, CCL2, and VEGFα by the Human Procarta-Plex Immunoassay (Supplementary Fig. S6C). Interestingly, IL6 was associated with a monocyte reprogramming towards M2-phenotype, therefore, we speculate a key role for IL6 in inducing the monocyte differentiation, as recently reported by many groups (33–36).

Figure 6.

Mo-derived LAM-STAB1 cells show a suppressive activity in vitro. CD14+-Miltenyi enriched monocytes from HD (N = 4) and TNBC (N = 4) were cocultured with FAP+CD29+ and FAPCD29+ CAF for 3 days. Enriched mono-derived LAM were then cocultured with autologous CFSE-labeled CD3 T cells at 1:1 ratio. At day 3, mono-derived LAM–T-cell cultures were assessed for CFSE dilution in CD3 T cells, PD1, and CD25 expression in CD4 and CD8 T cells. A, Left, graphs show the percentage of CFSE dilution in CD3 T cells gated on live cells from both HD and TNBC conditions derived from monocytes-FAP+CD29+ or FAPCD29+ CAF cocultures. Right, representative CFSE dilution plot in CD3 T cells from one representative patient with TNBC. B and C, Graphs show CD25 and PD1 frequencies evaluated in CD4 and CD8 T cells derived from monocytes–FAP+CD29+ or –FAPCD29+ CAF cocultures. Two pooled experiments in duplicate are shown. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 by two-way ANOVA Tukey multiple comparisons test. D, Grzm-B and IL10 (pg/mL) were evaluated by CBA from supernatants generated in the mono-derived LAM–T-cell cocultures (at day 3). *, P < 0.05; **, P < 0.01 by unpaired Student t test, two-tailed. One representative experiment run in duplicate is shown, pooled HD (N = 2) and TNBC (N = 2). Three independent experiments were analyzed.

Figure 6.

Mo-derived LAM-STAB1 cells show a suppressive activity in vitro. CD14+-Miltenyi enriched monocytes from HD (N = 4) and TNBC (N = 4) were cocultured with FAP+CD29+ and FAPCD29+ CAF for 3 days. Enriched mono-derived LAM were then cocultured with autologous CFSE-labeled CD3 T cells at 1:1 ratio. At day 3, mono-derived LAM–T-cell cultures were assessed for CFSE dilution in CD3 T cells, PD1, and CD25 expression in CD4 and CD8 T cells. A, Left, graphs show the percentage of CFSE dilution in CD3 T cells gated on live cells from both HD and TNBC conditions derived from monocytes-FAP+CD29+ or FAPCD29+ CAF cocultures. Right, representative CFSE dilution plot in CD3 T cells from one representative patient with TNBC. B and C, Graphs show CD25 and PD1 frequencies evaluated in CD4 and CD8 T cells derived from monocytes–FAP+CD29+ or –FAPCD29+ CAF cocultures. Two pooled experiments in duplicate are shown. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 by two-way ANOVA Tukey multiple comparisons test. D, Grzm-B and IL10 (pg/mL) were evaluated by CBA from supernatants generated in the mono-derived LAM–T-cell cocultures (at day 3). *, P < 0.05; **, P < 0.01 by unpaired Student t test, two-tailed. One representative experiment run in duplicate is shown, pooled HD (N = 2) and TNBC (N = 2). Three independent experiments were analyzed.

Close modal

Mo-derived LAM-STAB1 cells showed a suppressive activity in vitro

To inquire about the suppressive activity of mo-derived LAM-STAB1, these cells were differentiated in vitro from circulating monocytes (from HD and TNBC) upon a 3-day c-culture with FAP+CD29+ and FAPCD29+ primary cell lines (Fig. 5AC; Supplementary Fig. S5G; and Fig. 6A). They were then enriched and cocultured with autologous CD3+ T cells from either HD or patients with TNBC (Fig. 6A). T-cell proliferation by CFSE dilution was inhibited in FAP+CD29+-derived LAMs conditions, as opposed to FAPCD29+ ones, demonstrating the specific FAP+CD29+ effect in inducing a LAM suppressive phenotype (Fig. 6A). In addition, the frequency of activated CD4 and CD8 T cells expressing PD1 and CD25 was significantly higher in FAPCD29+-derived LAM, than in FAP+CD29+-derived ones (Fig. 6,BC; Supplementary Fig. S6D and S6E). Although the activatory markers PD1 and CD25 were highly expressed on CD4 from FAPCD29+-derived LAM compared with FAP+CD29+-derived LAM cocultures, we could not exclude that a portion of those CD4 T cells (PD1high, CD25high) may belong to the Treg subpopulation.

Notably, the amount of soluble granzyme B and IL10, assessed by Cytometric Bead Array (CBA), was significantly higher in FAPCD29+–derived LAM T-cell cocultures and in the FAP+CD29+-derived LAM cocultures, respectively (Fig. 6D). These data suggested a suppressive role of mo-derived LAM by inhibiting T-cell proliferation and effector functions.

iCAFs attract circulating monocytes to the tumor site by the CXCR4—CXCL12 axis

Based on recent findings on the coexistence—within FAP+CD29+ CAFs—of iCAF and myCAF subsets in breast cancer (20), we asked whether these CAF subsets engaged into specific interactions with myeloid cells and regulated their functions.

FC analysis of ex vivo surgical specimens revealed higher iCAF (ANTRX1) frequency in TNBC, while myCAF (ANTRX1+) frequency was enriched in patients with luminal BC (Fig. 7A). Likewise, the iCAF transcriptomic signature was statistically enriched in TNBC compared with LBC and HER2+ cohorts; while, no difference in myCAF signatures was observed (Fig. 7B), confirming previous observations on an independent cohort of breast cancer (20). To assess the ability of iCAF in balancing immune responses in TNBC, patients were segregated according to iCAFhigh and iCAFlow signature by bulk RNA. The iCAFhigh group displayed higher Tregs, cytotoxic and exhausted CD8 T signatures (37). Notably, strong correlations between iCAF (global signature, IL-iCAF or IFNγ-iCAF signatures), mono–early-MAC clusters (mono-S100A8, mono-FCGR3A, early–MAC-CXCR4), and mo-derived LAM-STAB1 were observed; whereas, only a weak correlation was found with LAM-APOC1 (Fig. 7C). To further explore a CAF monocyte–derived myeloid network, we assessed cell–cell communications via receptor–ligand interactions between iCAF/myCAF and myeloid clusters. We performed cell–cell interaction analysis using CellPhoneDB tool by matching scRNA-seq data from FAP+CD29+ CAFs (20) and our TNBC myeloid HLA-DR+CD11c+ dataset. We observed a total of 138 interactions for both iCAF/myCAF and myeloid cells, of which, the majority were shared (N = 98), 10 observed for iCAF, and 20 for myCAF (Fig. 7D). By pathway enrichment analysis of those selective interactions, we found that iCAF interacted preferentially via interleukins and cytokines; whereas myCAF via collagen and extracellular matrix factors (Fig. 7D), which revealed the existence of a cross-talk between iCAF and myeloid cells mediated by soluble factors.

Figure 7.

FAP+ iCAF accumulate in TNBC and attract circulating monocytes at tumor site by CXCR4–CXCL12 axis. A, Representative FAP/ANTRX1 FC staining (left) and percentages (right) of FAP+ANTRX1 (iCAF) and FAP+ANTRX1+ (myCAF) in FAP+CD29+ CAF gated cells in TUM from TNBC (N = 7) and LBC (N = 8). *, P < 0.05 by unpaired t test. B, Bulk RNA analysis of iCAF (top) and myCAF tumor signatures (bottom) from METABRIC TNBC (N = 332), LBC (N = 1,409), and HER2+ breast cancer (N = 248) cohorts. *, P < 0.05; **, P < 0.01; ***, P < 0.001 by unpaired t test. C, Heatmap showing Spearman correlation between the signatures of myeloid and CAF clusters by bulk RNA analysis in the METABRIC TNBC cohort (N = 332). D, Cell–cell interactions by CellPhoneDB of i/myCAF versus all myeloid cell clusters showing all significant interactions (P < 0.05). Venn diagram representing shared (N = 98) and specific iCAF (N = 10) or myCAF (N = 20)–myeloid interactions. Pathway enrichment analysis by Reactome of specific i/myCAF top nine enriched pathways. E, Dotplot showing cell—cell interactions by CellPhoneDB of iCAF or myCAF with all myeloid cell clusters. Dot size corresponds to ligand-receptor pairs according to cluster specificity (P values). The means of the average expression level of interacting are indicated by colors. F, CXCL12 concentrations in JT- and TUM-CM from untreated TNBC (N = 78) and LBC (N = 103) samples. ***, P < 0.001; ****, P < 0.0001 by paired t test, two-tailed between JT and TUM in TNBC, and by unpaired Student t test, two-tailed between TNBC and LBC cohorts. G, Percentage of CXCL12+ cells by FC in FAP+CD29+ gated CAF in TUM of patients with TNBC (N = 4) and LBC (N = 4), **, P < 0.01, by unpaired Student t test, two-tailed. H, Representative APOE/S100A8–9 FC expression in CD14+HLA-DR+ gate from one TNBC TUM; representative CXCR4 MFI expression and quantification in APOE+S100A8–9, APOE+S100A8–9+, APOES100A8–9+ myeloid cells. *, P < 0.05; **, P < 0.01 by paired t test. I, Representative APOE/CXCR4 FC expression plots in one TNBC and one LBC sample (left); percentage of CXCR4+ cells in CD14+HLA-DR+ gated myeloid cells. ****, P < 0.0001 by unpaired Student t test, two-tailed; bulk RNA analysis of CXCR4 expression in LBC (N = 1409), HER2 (N = 248), and TNBC (N = 332) METABRIC cohorts (right); ***, P < 0.001 by unpaired t test. J, Migration assay in vitro of blood monocytes from HD pretreated or not with AMD3100 (CXCR4 antagonist) in response to either recombinant CXCL12, CAF supernatants, or medium. One representative experiment of four is shown. Mean cell index for each condition (in duplicates) was evaluated. ****, P < 0.0001 by two-way ANOVA multiple comparisons. NS, not significant.

Figure 7.

FAP+ iCAF accumulate in TNBC and attract circulating monocytes at tumor site by CXCR4–CXCL12 axis. A, Representative FAP/ANTRX1 FC staining (left) and percentages (right) of FAP+ANTRX1 (iCAF) and FAP+ANTRX1+ (myCAF) in FAP+CD29+ CAF gated cells in TUM from TNBC (N = 7) and LBC (N = 8). *, P < 0.05 by unpaired t test. B, Bulk RNA analysis of iCAF (top) and myCAF tumor signatures (bottom) from METABRIC TNBC (N = 332), LBC (N = 1,409), and HER2+ breast cancer (N = 248) cohorts. *, P < 0.05; **, P < 0.01; ***, P < 0.001 by unpaired t test. C, Heatmap showing Spearman correlation between the signatures of myeloid and CAF clusters by bulk RNA analysis in the METABRIC TNBC cohort (N = 332). D, Cell–cell interactions by CellPhoneDB of i/myCAF versus all myeloid cell clusters showing all significant interactions (P < 0.05). Venn diagram representing shared (N = 98) and specific iCAF (N = 10) or myCAF (N = 20)–myeloid interactions. Pathway enrichment analysis by Reactome of specific i/myCAF top nine enriched pathways. E, Dotplot showing cell—cell interactions by CellPhoneDB of iCAF or myCAF with all myeloid cell clusters. Dot size corresponds to ligand-receptor pairs according to cluster specificity (P values). The means of the average expression level of interacting are indicated by colors. F, CXCL12 concentrations in JT- and TUM-CM from untreated TNBC (N = 78) and LBC (N = 103) samples. ***, P < 0.001; ****, P < 0.0001 by paired t test, two-tailed between JT and TUM in TNBC, and by unpaired Student t test, two-tailed between TNBC and LBC cohorts. G, Percentage of CXCL12+ cells by FC in FAP+CD29+ gated CAF in TUM of patients with TNBC (N = 4) and LBC (N = 4), **, P < 0.01, by unpaired Student t test, two-tailed. H, Representative APOE/S100A8–9 FC expression in CD14+HLA-DR+ gate from one TNBC TUM; representative CXCR4 MFI expression and quantification in APOE+S100A8–9, APOE+S100A8–9+, APOES100A8–9+ myeloid cells. *, P < 0.05; **, P < 0.01 by paired t test. I, Representative APOE/CXCR4 FC expression plots in one TNBC and one LBC sample (left); percentage of CXCR4+ cells in CD14+HLA-DR+ gated myeloid cells. ****, P < 0.0001 by unpaired Student t test, two-tailed; bulk RNA analysis of CXCR4 expression in LBC (N = 1409), HER2 (N = 248), and TNBC (N = 332) METABRIC cohorts (right); ***, P < 0.001 by unpaired t test. J, Migration assay in vitro of blood monocytes from HD pretreated or not with AMD3100 (CXCR4 antagonist) in response to either recombinant CXCL12, CAF supernatants, or medium. One representative experiment of four is shown. Mean cell index for each condition (in duplicates) was evaluated. ****, P < 0.0001 by two-way ANOVA multiple comparisons. NS, not significant.

Close modal

Among the common iCAF/myCAF–myeloid interactions, we found the CXCL12–CXCR4 axis (Fig. 7E) highlighted in all monocyte–early-MAC subclusters and in mo-derived LAM-STAB1, but not in LAM-APOC1. By analyzing scRNA-seq, iCAF, and myCAF showed CXCL12 expression, although iCAF expressed the highest levels (Supplementary Fig. S6F), suggesting their preferential, but not exclusive, involvement in a CXCL12–CXCR4 cross-talk with mono-derived myeloid cells in TNBC.

To assess the levels of CXCL12 in breast TME, we quantified by Luminex assay its concentration in the CM of JT and tumor from breast cancer specimens. CXCL12 was more accumulated in tumor than JT counterpart in patients with TNBC, potentially reflecting the high iCAF percentage detected by FC (Fig. 7A). CXCL12 levels were statistically higher in TNBC compared with LBCs (Fig. 7F) and, accordingly, the percentage of CXCL12+ CAFs was higher in TNBC than LBCs (Fig. 7G). Then, to confirm the presence of the receptor CXCR4, we analyzed its expression by scRNA-seq and FC. By scRNA-seq all monocyte clusters (Mono-S100A8, Mono-FCGR3A, and Mono-TREM1) and monocyte-derived clusters (early-MAC-CXCR4 and -ISG15, mo-derived LAM-STAB1) showed high CXCR4 expression, unlike the LAM-APOC1 cluster (Supplementary Fig. S6G). Consistently, the MFI of CXCR4 was higher in mono–early-MACs (APOE–/+/S100A8–9+) compared with APOE+S100A8–9 LAMs (Fig. 7H). In addition, CXCR4+ myeloid cells, evaluated by FC, were significantly enriched in TNBC than in LBC (Fig. 7I); similarly, CXCR4-bulk RNA expression was higher in TNBC (Fig. 7I, right panel). To assess the ability of iCAFs to promote the migration of monocytes at the tumor site via CXCL12–CXCR4, we set up the xCELLigence assay in vitro. We first generated CAF-derived and CXCL12-containing supernatants (24 hours); then we tested their ability to induce monocytes migration. The monocyte migration, evaluated as mean of cell index (at 2 hours) was higher in the presence of CAF-derived supernatant and drastically reduced when monocytes were pretreated with the CXCR4-antagonist (AMD3100). CXCL12 positive control of monocyte migration had the strongest effect (Fig. 7J).

These data demonstrate the implication of the CXCL12–CXCR4 pathway in the orchestration of the monocyte entrance at the tumor site. Moreover, the CXCL12–CXCR4 axis appeared to be selective for TNBC rather than LBC TME, as suggested by the negligible levels of CXCL12–CXCR4 expression in the latter (Fig. 7FI).

Here, we generated a comprehensive, single-cell atlas of the myeloid ecosystem of primary TNBC and identified, at the transcriptomic and protein levels, monocyte-derived STAB1+TREM2high LAMs as immune suppressive and expanded in patients resistant to PD-1 blockade.

Recent scRNA-seq studies proposed TREM2 as an immunosuppressive receptor in mice (18). Indeed, the monocyte-derived STAB1+TREM2high cells in patients with TNBC exhibited selective expression of proangiogenic factors (i.e, STAB1, LYVE-1; refs. 38, 39). Our data in vitro demonstrated a suppressive LAM phenotype and the capacity of monocyte-derived LAMs to inhibit T-cell proliferation and effector functions. To corroborate the role of the TREM2 receptor in modulating the tumor growth, we took advantage of the Trem2 KO mouse model implanted with Py8119 TNBC cell line. Despite the intrinsic limitations of using the full Trem2–/– KO mouse model to target the Trem2 expression on immune cells, the depletion of bona fide murine monocyte-derived STAB1+TREM2high LAMs resulted in the partial control of tumor growth, supporting our observations in patients about the suppressive role of monocyte-derived LAMs. When investigating the potential mechanisms behind the tumor inhibition, the Py8119 TNBC model did not recapitulate the human TNBC, being poorly controlled by CD8 T cells, hence unresponsive to anti-PD1. In immunogenic tumor models instead, Trem2 depletion was able to unleash CD8 T cell antitumor responses (18, 40).

In line with our findings, STAB1+ MACs were found in human colorectal cancer and breast cancer (41, 42) and in several murine tumor models (42, 43), indicating that targeting Stab1 could reactivate CD8 T cells and unleash antitumor immune responses (43). Consistent with the detrimental role of the mo-derived STAB1+ MACs in the mouse, these cells were enriched in patients with TNBC, whose T cells did not expand and were unresponsive to anti–PD-1 treatment. These preliminary findings require further validation in a larger patient cohort.

As we demonstrated that iCAFs accumulated preferentially in TNBC, and that iCAFhigh TNBC patients displayed a brisker T- and myeloid-cell infiltration, we speculated—based on a recent study (37)that iCAF may contribute to immune suppression in TNBC by interacting with myeloid cells. Indeed, we demonstrated among several interactions, a direct cross-talk between myeloid cells and CAFs mediated by CXCR4–CXCL12. This pathway regulated the formation of breast cancer metastasis (44, 45), cancer cell migration, epithelial—mesenchymal transition (19), and contributed to intratumoral Treg recruitment (7). CXCR4 was upregulated in monocytes, early-MACs, and mo-derived STAB1+ clusters suggesting their overall susceptibility to CXCL12. Hence, we suggested that the CXCL12 production in situ may preferentially derive from iCAF. Indeed, CAF-derived supernatants induced blood monocyte migration, which was drastically reduced after monocyte pretreatment with the CXCR4-antagonist. Collectively our results resolved the heterogeneity of the macrophage ecosystem in TNBC, demonstrated the protumorigenic and immune suppressive role of mo-derived STAB1+TREM2high MACs in human TNBC and the TME involvement in the generation of these cells.

M. Molgora reports grants from NIH and grants from Cancer Research Institute during the conduct of the study. A. Vincent-Salomon reports grants and personal fees from IBEX, Daiichi Sankyo; personal fees from Roche; grants, personal fees, and nonfinancial support from AstraZeneca; and personal fees from BMS outside the submitted work. V. Soumelis is a full-time employee at Owkin since March 2022; in addition, V. Soumelis reports grant support from Sanofi and personal fees from Leo Pharma and Aummune. M. Colonna reports grants and personal fees from NGM Biopharmaceutical during the conduct of the study; grants from Oncorus; and grants and personal fees from Vigil Neuroscience outside the submitted work; in addition, M. Colonna has a patent for TREM2 pending. S. Amigorena reports other support from INSERM, Institut Curie, CNRS during the conduct of the study; personal fees from Biomunex; grants and personal fees from Mnemo Therapeutics; and personal fees from Innate Pharma outside the submitted work. E. Romano reports grants from Fonds Amgen France pour la Science et l'Humain during the conduct of the study; Fondation BMS, Replimune, Astra Zeneca; and other support from MSD France outside the submitted work. No disclosures were reported by the other authors.

E. Timperi: Conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. P. Gueguen: Data curation, software, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. M. Molgora: Data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. I. Magagna: Resources, validation, methodology. Y. Kieffer: Formal analysis, validation, methodology. S. Lopez-Lastra: Validation, investigation, methodology. P. Sirven: Resources, formal analysis, validation, methodology. L.G. Baudrin: Data curation, methodology. S. Baulande: Data curation, methodology. A. Nicolas: Validation, methodology. G. Champenois: Validation, methodology. D. Meseure: Supervision, validation, methodology. A. Vincent-Salomon: Resources. A. Tardivon: Resources. E. Laas: Resources. V. Soumelis: Resources, writing–review and editing. M. Colonna: Resources, supervision writing–review and editing. F. Mechta-Grigoriou: Resources, supervision, writing–review and editing. S. Amigorena: Supervision, writing–review and editing. E. Romano: Conceptualization, resources, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft, writing–review and editing.

The authors thank all the platforms and services at Curie involved in the study: 10X, Cytometry, and Pathology. High-throughput sequencing was performed by the ICGex NGS platform of the Institut Curie supported by the grants ANR-10-EQPX-03 (Equipex) and ANR-10-INBS-09–08 (France Génomique Consortium) from the Agence Nationale de la Recherche ("Investissements d'Avenir" program), by the ITMO-Cancer Aviesan (Plan Cancer III), and by the SiRIC-Curie program (SiRIC Grant INCa-DGOS- 4654). The authors thank Maija Hollmén for the collaboration and for providing us with the antibody (STAB1); Immunologie Clinique and Olivier Lantz for their collaboration; Rodrigo Nalio Ramos, Jimena Tosello, Mercedes Tkach, Giulia Vanoni, and Arnaud Manon for technical advice; Julie Helft and Elodie Segura for the constructive discussions; Jules Gilet for bioinformatic analysis; Leclerc Renaud from the experimental pathology platform for technical support on Multiplex IHC analysis; Maud Kamal, Christophe le Tourneau, Charlotte Martinat, and all the colleagues involved in the SCANDARE and STROMA circuits; and all the patients and families. This work was supported by the following grants to E. Romano: Foundation ARC (grant no. AAP SIGN'IT 2019), Fonds Amgen France pour la Science et l'Humain; CIC IGR-Curie 1428; ANR-10-IDEX-0001-02 PSL; and ANR-11-LABX-0043; S. Amigorena: INSERM, Institut Curie, CNRS; M. Colonna: NIH (grant no. R01 CA262684).

E. Timperi was supported by a postdoctoral fellowship abroad from the AIRC (2018/2020-number: 20934). M. Molgora is a recipient of the Cancer Research Institute—Lloyd J. Old Memorial Fellowship in Tumor Immunology.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

1.
Ugel
S
,
De Sanctis
F
,
Mandruzzato
S
,
Bronte
V
.
Tumor-induced myeloid deviation: when myeloid-derived suppressor cells meet tumor-associated macrophages
.
J Clin Invest
2015
;
125
:
3365
76
.
2.
DeNardo
DG
,
Brennan
DJ
,
Rexhepaj
E
,
Ruffell
B
,
Shiao
SL
,
Madden
SF
, et al
.
Leukocyte complexity predicts breast cancer survival and functionally regulates response to chemotherapy
.
Cancer Discov
2011
;
1
:
54
67
.
3.
Miyasato
Y
,
Shiota
T
,
Ohnishi
K
,
Pan
C
,
Yano
H
,
Horlad
H
, et al
.
High density of CD204-positive macrophages predicts worse clinical prognosis in patients with breast cancer
.
Cancer Sci
2017
;
108
:
1693
700
.
4.
Azizi
E
,
Carr
AJ
,
Plitas
G
,
Cornish
AE
,
Konopacki
C
,
Prabhakaran
S
, et al
.
Single-cell map of diverse immune phenotypes in the breast tumor microenvironment
.
Cell
2018
;
174
:
1293
308
.
5.
Cassetta
L
,
Fragkogianni
S
,
Sims
AH
,
Swierczak
A
,
Forrester
LM
,
Zhang
H
, et al
.
Human tumor-associated macrophage and monocyte transcriptional landscapes reveal cancer-specific reprogramming, biomarkers, and therapeutic targets
.
Cancer Cell
2019
;
35
:
588
602
.
6.
Annaratone
L
,
Cascardi
E
,
Vissio
E
,
Sarotto
I
,
Chmielik
E
,
Sapino
A
, et al
.
The multifaceted nature of tumor microenvironment in breast carcinomas
.
Pathobiology
2020
;
87
:
125
42
.
7.
Costa
A
,
Kieffer
Y
,
Scholer-Dahirel
A
,
Pelon
F
,
Bourachot
B
,
Cardon
M
, et al
.
Fibroblast heterogeneity and immunosuppressive environment in human breast cancer
.
Cancer Cell
2018
;
33
:
463
79
.
8.
Pareja
F
,
Geyer
FC
,
Marchiò
C
,
Burke
KA
,
Weigelt
B
,
Reis-Filho
JS
.
Triple-negative breast cancer: the importance of molecular and histologic subtyping, and recognition of low-grade variants
.
NPJ Breast Cancer
2016
;
2
:
16036
.
9.
Timperi
E
,
Vissio
E
,
Marchiò
C
,
Romano
E
.
The immune landscape in women cancers
.
Cancer Treat Res
2020
;
180
:
215
49
.
10.
Burstein
MD
,
Tsimelzon
A
,
Poage
GM
,
Covington
KR
,
Contreras
A
,
Fuqua
SAW
, et al
.
Comprehensive genomic analysis identifies novel subtypes and targets of triple-negative breast cancer
.
Clin Cancer Res
2015
;
21
:
1688
98
.
11.
Masuda
H
,
Baggerly
KA
,
Wang
Y
,
Zhang
Y
,
Gonzalez-Angulo
AM
,
Meric-Bernstam
F
, et al
.
Differential response to neoadjuvant chemotherapy among 7 triple-negative breast cancer molecular subtypes
.
Clin Cancer Res
2013
;
19
:
5533
40
.
12.
Ayers
M
,
Lunceford
J
,
Nebozhyn
M
,
Murphy
E
,
Loboda
A
,
Kaufman
DR
, et al
.
IFN-γ-related mRNA profile predicts clinical response to PD-1 blockade
.
J Clin Invest
2017
;
127
:
2930
40
.
13.
Wang
B
,
Liu
M
,
Ran
Z
,
Li
X
,
Li
J
,
Ou
Y
.
Analysis of gene signatures of tumor microenvironment yields insight into mechanisms of resistance to immunotherapy
.
Front Bioeng Biotechnol
2020
;
8
:
348
.
14.
Mantovani
A
,
Biswas
SK
,
Galdiero
MR
,
Sica
A
,
Locati
M
.
Macrophage plasticity and polarization in tissue repair and remodelling
.
J Pathol
2013
;
229
:
176
85
.
15.
Sharma
A
,
Seow
JJW
,
Dutertre
C-A
,
Pai
R
,
Blériot
C
,
Mishra
A
, et al
.
Onco-fetal reprogramming of endothelial cells drives immunosuppressive macrophages in hepatocellular carcinoma
.
Cell
2020
;
183
:
377
94
.
16.
Wu
K
,
Lin
K
,
Li
X
,
Yuan
X
,
Xu
P
,
Ni
P
, et al
.
Redefining tumor-associated macrophage subpopulations and functions in the tumor microenvironment
.
Front Immunol
2020
;
11
:
1731
.
17.
Zhang
L
,
Li
Z
,
Skrzypczynska
KM
,
Fang
Q
,
Zhang
W
,
O'Brien
SA
, et al
.
Single-cell analyses inform mechanisms of myeloid-targeted therapies in colon cancer
.
Cell
2020
;
181
:
442
59
.
18.
Molgora
M
,
Esaulova
E
,
Vermi
W
,
Hou
J
,
Chen
Y
,
Luo
J
, et al
.
TREM2 modulation remodels the tumor myeloid landscape enhancing anti-PD-1 immunotherapy
.
Cell
2020
;
182
:
886
900
.
19.
Pelon
F
,
Bourachot
B
,
Kieffer
Y
,
Magagna
I
,
Mermet-Meillon
F
,
Bonnet
I
, et al
.
Cancer-associated fibroblast heterogeneity in axillary lymph nodes drives metastases in breast cancer through complementary mechanisms
.
Nat Commun
2020
;
11
:
404
.
20.
Kieffer
Y
,
Hocine
HR
,
Gentric
G
,
Pelon
F
,
Bernard
C
,
Bourachot
B
, et al
.
Single-cell analysis reveals fibroblast clusters linked to immunotherapy resistance in cancer
.
Cancer Discov
2020
;
10
:
1330
51
.
21.
Cardoso
F
,
Kyriakides
S
,
Ohno
S
,
Penault-Llorca
F
,
Poortmans
P
,
Rubio
IT
, et al
.
Early breast cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up†
.
Ann Oncol
2019
;
30
:
1194
220
.
22.
Givel
A-M
,
Kieffer
Y
,
Scholer-Dahirel
A
,
Sirven
P
,
Cardon
M
,
Pelon
F
, et al
.
miR200-regulated CXCL12β promotes fibroblast heterogeneity and immunosuppression in ovarian cancers
.
Nat Commun
2018
;
9
:
1056
.
23.
Jaitin
DA
,
Adlung
L
,
Thaiss
CA
,
Weiner
A
,
Li
B
,
Descamps
H
, et al
.
Lipid-associated macrophages control metabolic homeostasis in a Trem2-dependent manner
.
Cell
2019
;
178
:
686
98
.
24.
Maier
B
,
Leader
AM
,
Chen
ST
,
Tung
N
,
Chang
C
,
LeBerichel
J
, et al
.
A conserved dendritic-cell regulatory program limits antitumour immunity
.
Nature
2020
;
580
:
257
62
.
25.
Fang
H
,
Ang
B
,
Xu
X
,
Huang
X
,
Wu
Y
,
Sun
Y
, et al
.
TLR4 is essential for dendritic cell activation and anti-tumor T-cell response enhancement by DAMPs released from chemically stressed cancer cells
.
Cell Mol Immunol
2014
;
11
:
150
9
.
26.
Rhee
SH
,
Im
E
,
Pothoulakis
C
.
Toll-like receptor 5 engagement modulates tumor development and growth in a mouse xenograft model of human colon cancer
.
Gastroenterology
2008
;
135
:
518
28
.
27.
Curtis
C
,
Shah
SP
,
Chin
S-F
,
Turashvili
G
,
Rueda
OM
,
Dunning
MJ
, et al
.
The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups
.
Nature
2012
;
486
:
346
52
.
28.
Guo
X
,
Zhang
Y
,
Zheng
L
,
Zheng
C
,
Song
J
,
Zhang
Q
, et al
.
Global characterization of T cells in non-small-cell lung cancer by single-cell sequencing
.
Nat Med
2018
;
24
:
978
85
.
29.
Bassez
A
,
Vos
H
,
Van Dyck
L
,
Floris
G
,
Arijs
I
,
Desmedt
C
, et al
.
A single-cell map of intratumoral changes during anti-PD1 treatment of patients with breast cancer
.
Nat Med
2021
;
27
:
820
32
.
30.
Qu
Y
,
Wen
J
,
Thomas
G
,
Yang
W
,
Prior
W
,
He
W
, et al
.
Baseline frequency of inflammatory Cxcl9-expressing tumor-associated macrophages predicts response to avelumab treatment
.
Cell Rep
2020
;
32
:
107873
.
31.
Ramos
RN
,
Rodriguez
C
,
Hubert
M
,
Ardin
M
,
Treilleux
I
,
Ries
CH
, et al
.
CD163+ tumor-associated macrophage accumulation in breast cancer patients reflects both local differentiation signals and systemic skewing of monocytes
.
Clin Transl Immunol
2020
;
9
:
e1108
.
32.
Bonavita
E
,
Bromley
CP
,
Jonsson
G
,
Pelly
VS
,
Sahoo
S
,
Walwyn-Brown
K
, et al
.
Antagonistic inflammatory phenotypes dictate tumor fate and response to immune checkpoint blockade
.
Immunity
2020
;
53
:
1215
29
.
33.
Chomarat
P
,
Banchereau
J
,
Davoust
J
,
Palucka
AK
.
IL-6 switches the differentiation of monocytes from dendritic cells to macrophages
.
Nat Immunol
2000
;
1
:
510
4
.
34.
Braune
J
,
Weyer
U
,
Hobusch
C
,
Mauer
J
,
Brüning
JC
,
Bechmann
I
, et al
.
IL-6 regulates M2 polarization and local proliferation of adipose tissue macrophages in obesity
.
J Immunol
2017
;
198
:
2927
34
.
35.
Sanmarco
LM
,
Ponce
NE
,
Visconti
LM
,
Eberhardt
N
,
Theumer
MG
,
Minguez
ÁR
, et al
.
IL-6 promotes M2 macrophage polarization by modulating purinergic signaling and regulates the lethal release of nitric oxide during Trypanosoma cruzi infection
.
Biochim Biophys Acta Mol Basis Dis
2017
;
1863
:
857
69
.
36.
Fernando
MR
,
Reyes
JL
,
Iannuzzi
J
,
Leung
G
,
McKay
DM
.
The pro-inflammatory cytokine, interleukin-6, enhances the polarization of alternatively activated macrophages
.
PLoS One
2014
;
9
:
e94188
.
37.
Wu
SZ
,
Roden
DL
,
Wang
C
,
Holliday
H
,
Harvey
K
,
Cazet
AS
, et al
.
Stromal cell diversity associated with immune evasion in human triple-negative breast cancer
.
EMBO J
2020
;
39
:
e104063
.
38.
Chakarov
S
,
Lim
HY
,
Tan
L
,
Lim
SY
,
See
P
,
Lum
J
, et al
.
Two distinct interstitial macrophage populations coexist across tissues in specific subtissular niches
.
Science
2019
;
363
:
eaau0964
.
39.
Salmi
M
,
Koskinen
K
,
Henttinen
T
,
Elima
K
,
Jalkanen
S
.
CLEVER-1 mediates lymphocyte transmigration through vascular and lymphatic endothelium
.
Blood
2004
;
104
:
3849
57
.
40.
Binnewies
M
,
Pollack
JL
,
Rudolph
J
,
Dash
S
,
Abushawish
M
,
Lee
T
, et al
.
Targeting TREM2 on tumor-associated macrophages enhances immunotherapy
.
Cell Rep
2021
;
37
:
109844
.
41.
Algars
A
,
Irjala
H
,
Vaittinen
S
,
Huhtinen
H
,
Sundström
J
,
Salmi
M
, et al
.
Type and location of tumor-infiltrating macrophages and lymphatic vessels predict survival of colorectal cancer patients
.
Int J Cancer
2012
;
131
:
864
73
.
42.
Riabov
V
,
Yin
S
,
Song
B
,
Avdic
A
,
Schledzewski
K
,
Ovsiy
I
, et al
.
Stabilin-1 is expressed in human breast cancer and supports tumor growth in mammary adenocarcinoma mouse model
.
Oncotarget
2016
;
7
:
31097
110
.
43.
Viitala
M
,
Virtakoivu
R
,
Tadayon
S
,
Rannikko
J
,
Jalkanen
S
,
Hollmén
M
.
Immunotherapeutic blockade of macrophage clever-1 reactivates the CD8+ T-cell response against immunosuppressive tumors
.
Clin Cancer Res Off J Am Assoc Cancer Res
2019
;
25
:
3289
303
.
44.
Liang
Z
,
Yoon
Y
,
Votaw
J
,
Goodman
MM
,
Williams
L
,
Shim
H
.
Silencing of CXCR4 blocks breast cancer metastasis
.
Cancer Res
2005
;
65
:
967
71
.
45.
Müller
A
,
Homey
B
,
Soto
H
,
Ge
N
,
Catron
D
,
Buchanan
ME
, et al
.
Involvement of chemokine receptors in breast cancer metastasis
.
Nature
2001
;
410
:
50
6
.

Supplementary data