Abstract
While macrophages are among the most abundant immune cell type found within primary and metastatic mammary tumors, how their complexity and heterogeneity change with metastatic progression remains unknown. Here, macrophages were isolated from the lungs of mice bearing orthotopic mammary tumors for single-cell RNA sequencing (scRNA-seq). Seven distinct macrophage clusters were identified, including populations exhibiting enhanced differential expression of genes related to antigen presentation (H2-Aa, Cd74), cell cycle (Stmn1, Cdk1), and interferon signaling (Isg15, Ifitm3). Interestingly, one cluster demonstrated a profile concordant with lipid-associated macrophages (Lgals3, Trem2). Compared with nontumor-bearing controls, the number of these cells per gram of tissue was significantly increased in lungs from tumor-bearing mice, with the vast majority costaining positively with the alveolar macrophage marker Siglec-F. Enrichment of genes implicated in pathways related to lipid metabolism as well extracellular matrix remodeling and immunosuppression was observed. In addition, these cells displayed reduced capacity for phagocytosis. Collectively, these findings highlight the diversity of macrophages present within metastatic lesions and characterize a lipid-associated macrophage subset previously unidentified in lung metastases.
scRNA-seq of macrophages isolated from lung metastases reveals extensive macrophage heterogeneity and identifies a novel subpopulation enriched for genes involved in lipid metabolism, extracellular matrix remodeling, and immunosuppression.
Introduction
The majority of cancer-related deaths are the result of metastatic spread from the primary-tumor site to vital secondary organs (1). In breast cancer, the most common sites of metastasis are the bone, lung, liver, and brain (2). It is well appreciated that the tumor microenvironment (TME), including immune cells, plays a critical role in facilitating metastatic dissemination. Among these, macrophages are among the most abundant immune cells found in primary breast tumors and metastatic lesions, with increased macrophage density correlating to reduced overall survival in patients with breast cancer (3). Tumor associated macrophages (TAM) exert their effects through a variety of mechanisms, promoting tumor-cell survival, angiogenesis, invasion, and protumorigenic inflammation and immune suppression (4). Furthermore, macrophages have been linked to metastatic spread in mouse mammary tumor models, with early depletion resulting in reductions in metastatic burden (5, 6).
Tissue-resident and recruited macrophages at the metastatic site also play critical roles in the seeding of malignant cells. In the lung, there are 2 primary resident macrophage populations that have been characterized under homeostatic conditions. Alveolar macrophages (AM) reside in the airway lumen and are critical for proper lung development and the generation of balanced inflammatory responses (7). Under normal conditions, they serve to suppress pathologic inflammation, while they act as a first line of defense in mounting a response against various pulmonary infections. Furthermore, these cells participate in the maintenance of metabolic homeostasis in the lungs, catabolizing and recycling lipids, primarily cholesterol, from surfactant (8). These cells are seeded from fetal precursors early in development, and are generally self-renewing, although can be replaced by circulating monocytes following injury or insult (9, 10). Interstitial macrophages (IM) are found in the parenchymal space of the lung in close proximity to blood vessels, nerves, and airways and are continuously replenished from the circulating monocyte pool (11, 12). Recently, the extent of lung-macrophage diversity in both steady state and disease has become more appreciated, with distinct subsets of these populations characterized based on localization and function (9, 12).
In breast cancer, AM contribute to the establishment of the premetastatic niche, in part through suppression of antitumor T-cell responses (13). Similarly, classical inflammatory monocytes, which differentiate into TAM following recruitment, also promote breast cancer metastasis, by enhancing tumor-cell survival and dissemination and suppressing cytotoxic immune functions (14, 15). Interestingly, patrolling monocytes have been shown to have the opposite effect, controlling tumor metastasis to the lung (16).
In the present study, single-cell RNA sequencing (scRNA-seq) was utilized to examine the diversity of macrophages within the lungs of mice bearing metastases from primary mammary tumors. We find extensive heterogeneity, identifying at least 7 distinct macrophage subsets. Importantly, we demonstrate the presence of a population of lipid-associated macrophages (LAM), which bear a marked similarity to foamy macrophages found within atherosclerotic and neurodegenerative disease models (17, 18). There is a dramatic expansion in the number of these cell-per-gram of lung tissue in metastasis-bearing lungs compared with controls. Furthermore, these cells positively costain with markers of AM, and both populations display similar enrichment for genes defining pathways related to lipid metabolism. LAM are unique in their enrichment for genes involved in extracellular matrix (ECM) remodeling and immunosuppression pathways, and display a reduced phagocytic capacity compared with non–lipid-laden controls. Taken together, these data define a population of macrophages with protumorigenic features that has previously remained uncharacterized in metastasis-bearing lungs.
Materials and Methods
Mice
Female BALB/c mice were purchased from Envigo Laboratories. Mice were housed in a specific pathogen-free facility and all procedures were performed in accordance with policies outlined by the University of Minnesota Institutional Animal Care and Use Committee and detailed in the Guide for Care and Use of Laboratory Animals.
Tumor transplantation
4T1-Luc cells were obtained from Thomas Griffith (University of Minnesota, Minneapolis, MN) and parental HC11/R1 cells were provided by Jeffrey Rosen (Baylor College of Medicine, Houston, TX). All cells are tested for mycoplasma (MycoAlert Mycoplasma Detection Kit, Lonza) and mycoplasma-negative cells are frozen in aliquots. For tumor-cell inoculation, cells are thawed from these aliquots and used within one to two passages. For tumor induction, 1 × 104 4T1-Luc or 5 × 104 HC11/R1-LM (see Supplementary Materials and Methods) cells suspended in 50% Matrigel (BD Biosciences) in PBS were injected into the right inguinal mammary fat pads of 5- to 8-week-old female mice. For mice receiving HC11/R1-LM cells, i.p. injection of 1 mg/kg B/B Homodimerizer (Takara Bio) was administered twice weekly. Tumor volume was measured by caliper 3 times per week and calculated as V = (L*W2)/2. Mice were euthanized when the primary tumor volume reached an endpoint of 1,250 to 1,500 mm3.
scRNA-seq and pathway analysis of DE genes
See Supplementary Materials and Methods. For scRNA-seq, each biological replicate consisted of cells isolated from the lungs of a single metastasis-bearing mouse. For replicate 1, clustering at a resolution of 0.6 was determined to produce biologically relevant populations and was used for all further analyses. For replicate 2, clustering at a resolution of 0.5 was selected. Data have been deposited in the Gene Expression Omnibus (GEO) GSE164302. Ranked lists were created with genes from the DE analysis by taking the negative log10 of the p value and adding the direction of change from the fold change (FC). This results in a ranked list of genes arranged by the most significant p value. Gene set enrichment analysis (GSEA; ref. 19) was performed using the R-based clusterProfiler package (version 3.14.3; ref. 20) with the Molecular Signature Databases (MsigDB v7) Hallmark, C2 (specifically KEGG), C5, and gene sets created from a previous study (17). Ranked lists were created for the bulk RNA sequencing (RNA-seq) and scRNA-seq data from Kim and colleagues using the same protocol. These lists were filtered (adjusted p value < 0.001 and log2 FC > 0.25) to create gene sets for GSEA (Bulk_RNAseq_foamy_DOWN_nonfoamy_UP, Bulk_RNAseq_foamy_UP_nonfoamy_DOWN, scRNAseq_foamy_DOWN_nonfoamy_UP, scRNAseq_foamy_UP_nonfoamy_DOWN; Supplementary Dataset 3). Comparisons between the top 25 signature genes from cluster 6 (Supplementary Table S1) and unstimulated and oxLDL-treated wild-type (WT) macrophages from the microarray dataset GSE32358 were performed as described in the Supplementary Materials and Methods.
Immunostaining and histology
At endpoint, lungs were harvested, taking care to preserve the trachea, and fixed in 2% paraformaldehyde (PFA) for 2 hours on ice. Tissues were cryopreserved in 30% sucrose overnight at 4°C. Lungs were inflated through the trachea with a mixture of 50% optimal cutting temperature (OCT) compound (Fisher Scientific) in PBS and flash frozen in OCT compound in cryomolds. For each mouse, a minimum of 3 6 μm sections taken at least 100 μm apart were stained with hematoxylin and eosin (H&E) for histologic analysis. Sections of 20 μm were used for immunofluorescence (IF) staining. See Supplementary Materials and Methods.
Macrophage depletion by clodronate administration
Twenty-five days post injection with 4T1-Luc cells, primary tumors were resected and mice rested for 3 days. Macrophages were then depleted via liposomal clodronate (Fisher Scientific) by intranasal (i.n., 0.25 mg/mouse) administration every other day. Control mice were given an equal volume of empty liposomes. Fourteen days after the initial dose, mice were euthanized and lungs harvested for flow cytometry as described.
Phagocytosis assays
See Supplementary Materials and Methods for generation of bone marrow–derived macrophages (BMDM). Following generation and overnight attachment, cells were starved for 1 hour in media containing 1% FBS (Atlanta Biologicals) and loaded with 40 μg/mL soluble cholesterol (1:6 with methyl-β-cyclodextran in 0.2% BSA in RPMI) or 100 μg/mL oxLDL (Thermo Fisher), or left unloaded as controls. After 24 hours, cells were incubated with prelabeled Zymosan particles (Abcam) for 3 hours according to manufacturer protocols. Cells were then harvested by cell scraping and Zymosan uptake determined by flow cytometry. In vitro assays were performed in triplicate using BMDMs harvested from 3 independent mice.
Statistical analysis
Analysis of differential gene expression and pathway analysis were performed as described above. All experiments in vivo were performed a minimum of 2 times, and all in vitro experiments were performed in technical triplicate with cells harvested from 3 independent biological samples. Comparisons between datasets were made using ANOVA followed by Sidak or Dunnett multiple comparisons test as appropriate. Data are represented as mean ± SE with the following notations of statistical significance: *P < 0.05, **P < 0.01, ***P < 0.001.
Results
Metastasis-bearing lungs exhibit extensive macrophage heterogeneity
To characterize the heterogeneity among macrophages found within metastasis-bearing lungs, scRNA-seq was performed on macrophages isolated from the lungs of a single female BALB/c mouse 5 weeks after orthotopic transplant of a 4T1-Luc mammary gland tumor, when the primary-tumor volume reached 1,250 mm3. Lung macrophages (CD64+MerTK+; ref. 21) were FACS sorted and sequenced using the 10x Chromium platform. Clustering analysis revealed 11 unique cell clusters defined by distinct gene-expression profiles (Fig. 1A and B; Supplementary Table S1; Supplementary Dataset 1). The majority of these clusters (clusters 0–3, 5, 6, 8) demonstrate expression of known macrophage-associated genes including Cd68 and Fcgr1 (Cd64) (Fig. 1C). Small clusters of nonmacrophage leukocytes were also identified (Fig. 1B). Cells in cluster 7 show enrichment for neutrophil markers such as S100a9/8, Retnlg, and Lcn2. Genes related to the T-cell receptor, including Cd3g/d/e and Trbc2/1 define cluster 9, indicating that this cluster represents T cells. Differentially expressed (DE) genes defining cluster 10 include MHC-II molecule-encoding genes (H2-Ab1, H2-DMb2) and Cd209a, indicative of dendritic cells.
scRNA-seq reveals extensive macrophage heterogeneity in metastasis-bearing lungs. Macrophages (CD64+MerTK+) were isolated from the lungs of a mouse 5 weeks after transplant of a 4T1-Luc mammary-gland tumor and sequenced using the 10x Genomics platform (11,013 cells sequenced; 10,979 cells analyzed after filtering for quality). Using the Seurat R package, clusters were identified and differential gene expression analysis performed. A, UMAP projection showing the 11 unique cell clusters identified. B, Heatmap of the top 20 DE genes for each cluster, with representative enriched genes used for cluster identification listed. C, Feature and violin plots demonstrating that the majority of the clusters sequenced express known macrophage-related genes Cd68 and Fcgr1. D and E, Feature and violin plots demonstrating differential expression of known AM (D) or IM (E) markers.
scRNA-seq reveals extensive macrophage heterogeneity in metastasis-bearing lungs. Macrophages (CD64+MerTK+) were isolated from the lungs of a mouse 5 weeks after transplant of a 4T1-Luc mammary-gland tumor and sequenced using the 10x Genomics platform (11,013 cells sequenced; 10,979 cells analyzed after filtering for quality). Using the Seurat R package, clusters were identified and differential gene expression analysis performed. A, UMAP projection showing the 11 unique cell clusters identified. B, Heatmap of the top 20 DE genes for each cluster, with representative enriched genes used for cluster identification listed. C, Feature and violin plots demonstrating that the majority of the clusters sequenced express known macrophage-related genes Cd68 and Fcgr1. D and E, Feature and violin plots demonstrating differential expression of known AM (D) or IM (E) markers.
Historically, macrophages in the lung have been broadly characterized as either AM or IM, based on their localization and function, which can be identified based on both gene and surface-marker expression. Our analysis demonstrates robust expression of AM-associated genes, including Chil3 and Itgax, in clusters 0 and 8 (Fig. 1D), while clusters 1, 2, 3, and 5 express varying levels of genes attributed to IM, such as Cx3cr1 and Itgam (Fig. 1E). Thus, our results confirm the maintenance of macrophages fitting these two classifications in metastasis-bearing lungs, while further identifying additional macrophage subsets, highlighting their complexity and suggesting myriad functions for macrophages in the context of metastatic disease.
Beyond the 2 broadly defined populations, scRNA-seq revealed a number of more distinct macrophage subsets based on differential gene expression (Fig. 2; Supplementary Dataset 1). As noted, cluster 0 is defined by expression of canonical AM markers, with cluster 8 demonstrating similar enrichment (Fig. 1D). The most highly DE genes defining cluster 1 are associated with mature macrophages involved in antigen presentation, including MHC-II molecule-encoding genes (H2-Eb2, H2-Aa) and Cd74, as well as several genes related to the complement pathway (C1qa/c/b; Fig. 2A). Expression of antigen presentation–related genes is also high in macrophage clusters 2, 3, 5, and 6, suggesting that the classical role as an antigen-presenting cell may be conserved in macrophages with otherwise varying genetic or phenotypic profiles. Clusters 2 and 8 both demonstrate increased expression of cell cycle–related genes such as Stmn1 and Cdk1, indicating that these clusters represent proliferating or self-renewing populations of IM and AM, respectively (Fig. 2B). Further analysis of cell cycle–related genes verifies that the majority of cells in clusters 2 and 8 are in G2–M or S phase, whereas most cells in the remaining clusters are in G1 phase (Supplementary Fig. S1A). Cells in cluster 3 are primarily characterized by expression of IFN-stimulated genes (Isg15, Ifitm3, Irf7; Fig. 2C). In cluster 4, significant enrichment was not observed for any of the top 2,000 overall DE genes, and the most highly expressed genes are predominantly mitochondrial in origin, a feature that has been shown to correlate with cell death (22). As such, this cluster was excluded from further analysis. Interestingly, cells in cluster 5 show increased expression of genes associated with an immunosuppressive, matrix-remodeling phenotype (Arg1, Fn1), as well as those associated with protumorigenic inflammation and myeloid-cell recruitment, including the chemokines Cxcl1 and Cxcl2 (Fig. 2D). Cluster 6 is comprised of cells with a gene-expression profile associated with cholesterol and lipid metabolism, including Trem2, Lgals3, and Apoe (Fig. 2E). Notably, the gene signature for cluster 6 is highly concordant to that of lipid-associated, foamy macrophages found in atherosclerotic plaques, neurodegenerative diseases, and other metabolic disorders (17, 23, 24). Recently, similar lipid-laden macrophages have also been reported within some tumor types, including primary and metastatic tumors originating from the colon and in renal cell carcinoma (25–27). Importantly, the presence of this unique population of cells was confirmed by scRNA-seq analysis of an independent biological replicate (cluster 5, Supplementary Fig. S1B–S1D; Supplementary Dataset 2). Clustering analysis of the integrated data from both samples demonstrates a high degree of concordance between the replicates (Supplementary Fig. S1E).
Distinct metastasis-associated macrophage subsets are identified by expression of key gene signatures. A–E, Feature and violin plots of key genes demonstrate that macrophages isolated from metastasis-bearing lungs separate into unique clusters defined by gene signatures related to antigen presentation (A); cell cycle and proliferation (B), IFN responsiveness (C); immunosuppressive and protumorigenic immune responses and matrix remodeling (D); and cholesterol and lipid metabolism (E).
Distinct metastasis-associated macrophage subsets are identified by expression of key gene signatures. A–E, Feature and violin plots of key genes demonstrate that macrophages isolated from metastasis-bearing lungs separate into unique clusters defined by gene signatures related to antigen presentation (A); cell cycle and proliferation (B), IFN responsiveness (C); immunosuppressive and protumorigenic immune responses and matrix remodeling (D); and cholesterol and lipid metabolism (E).
A population of LAM is found in metastasis-bearing lungs
We next sought to verify whether the cells in cluster 6 do in fact represent the same population previously defined as foamy macrophages in other models. To do this, we compared our data with those published by Kim and colleagues, in which foamy macrophages were isolated from the aortas of mice with atherosclerotic plaques and subjected to either scRNA-seq or bulk RNA-seq (17). Using these datasets, we developed lists comprised of genes that are either upregulated in foamy macrophages or downregulated in nonfoamy macrophages, or vice versa. GSEA was then performed comparing the differentially upregulated genes found in cluster 6 of our scRNA-seq analysis to these reference datasets (Fig. 3A, Supplementary Dataset 3). Importantly, compared with all other macrophages identified in our analysis (clusters 0–3, 5, 8), cluster 6 demonstrated significant positive enrichment for the genes found to be upregulated in foamy macrophages in both the scRNA-seq (scRNAseq_foamy_UP_nonfoamy_DOWN, NES = 1.85, Padj = 0.0023) and bulk RNA-seq (Bulk_RNAseq_foamy_UP_nonfoamy_DOWN, NES = 1.91, Padj = 0.0023) datasets. Similarly, cluster 6 showed negative enrichment for those genes found to be upregulated in the nonfoamy macrophage populations (scRNAseq_foamy_DOWN_nonfoamy_UP, NES = −1.81, Padj = 0.014 and Bulk_RNAseq_foamy_DOWN_nonfoamy_UP, NES = −2.06, Padj = 0.014). Furthermore, in assessing relative expression of the core enrichment genes defining the scRNAseq_foamy_UP_nonfoamy_DOWN dataset from Kim and colleagues, cluster 6 shows the greatest enrichment compared with all other clusters (Fig. 3B). The high level of concordance observed between the DE genes defining cluster 6 and known foamy-macrophage datasets thus validates the characterization of these cells as LAM, a population previously unseen in metastasis-bearing lungs.
GSEA verifies that LAMs within metastasis-bearing lungs exhibit the gene expression profile of foamy macrophages. A, Waterfall plots demonstrate that, compared with other macrophage clusters, LAMs in metastasis-bearing lungs (cluster 6) demonstrate highly similar gene signatures to defined bulk and scRNA-seq datasets of foamy macrophages isolated from atherosclerotic plaques. scRNAseq_foamy_UP_nonfoamy_DOWN (NES = 1.85, Padj = 0.002), Bulk_RNAseq_foamy_UP_nonfoamy_DOWN (NES = 1.91, Padj = 0.002), Bulk_RNAseq_foamy_DOWN_nonfoamy_UP (NES = –2.06, Padj = 0.014), scRNAseq_foamy_DOWN_nonfoamy_UP (NES = –1.80, Padj = 0.014). B, Heatmap of relative expression among metastasis-associated macrophage subsets of core enrichment genes defining foamy macrophages based on a published scRNA-seq dataset (17). C–E, GSEA of cluster 6 versus other macrophage clusters (C0–3, 5, 8) identified by scRNA-seq demonstrates positive enrichment for genes implicated in pathways related to lipid metabolism. Waterfall plots showing relative enrichment of genes involved in lysosome function (NES = 1.79, Padj = 0.12, FDR = 0.11; C); cholesterol efflux (NES = 1.62, Padj = 0.17, FDR = 0.15; D); and lipid catabolism (NES = 1.50, Padj = 0.15, FDR = 0.13; E) in cluster 6 compared with other macrophage clusters. NES, normalized enrichment score.
GSEA verifies that LAMs within metastasis-bearing lungs exhibit the gene expression profile of foamy macrophages. A, Waterfall plots demonstrate that, compared with other macrophage clusters, LAMs in metastasis-bearing lungs (cluster 6) demonstrate highly similar gene signatures to defined bulk and scRNA-seq datasets of foamy macrophages isolated from atherosclerotic plaques. scRNAseq_foamy_UP_nonfoamy_DOWN (NES = 1.85, Padj = 0.002), Bulk_RNAseq_foamy_UP_nonfoamy_DOWN (NES = 1.91, Padj = 0.002), Bulk_RNAseq_foamy_DOWN_nonfoamy_UP (NES = –2.06, Padj = 0.014), scRNAseq_foamy_DOWN_nonfoamy_UP (NES = –1.80, Padj = 0.014). B, Heatmap of relative expression among metastasis-associated macrophage subsets of core enrichment genes defining foamy macrophages based on a published scRNA-seq dataset (17). C–E, GSEA of cluster 6 versus other macrophage clusters (C0–3, 5, 8) identified by scRNA-seq demonstrates positive enrichment for genes implicated in pathways related to lipid metabolism. Waterfall plots showing relative enrichment of genes involved in lysosome function (NES = 1.79, Padj = 0.12, FDR = 0.11; C); cholesterol efflux (NES = 1.62, Padj = 0.17, FDR = 0.15; D); and lipid catabolism (NES = 1.50, Padj = 0.15, FDR = 0.13; E) in cluster 6 compared with other macrophage clusters. NES, normalized enrichment score.
To further characterize this population of cells in relation to the other macrophage subsets, we again performed GSEA, this time using predefined gene sets in the MSigDB. Specifically, we focused on the Hallmark, KEGG (C2), and Gene Ontology (C5) gene sets. Compared to the other macrophage clusters, cluster 6 demonstrated positive enrichment for genes involved in pathways related to lipid metabolism, including lysosome, cholesterol efflux, and lipid catabolism (Fig. 3C–E; Supplementary Dataset 3). This is again in agreement with published data in canonical foamy macrophages, which demonstrate enrichment of related pathways.
LAM costain with AM markers and undergo robust expansion in metastasis-bearing lungs
Foamy macrophages can be identified histologically by the presence of cytosolic lipid droplets, responsible for their characteristic “foamy” appearance. We next investigated whether lipid-containing macrophages could be detected in metastasis-bearing lung tissue by IF staining with neutral lipid dyes. Mice were transplanted with 4T1-Luc mammary gland tumors and allowed to progress to a primary-tumor volume of 1,250 to 1,500 mm3 (Supplementary Fig. S2A). Lungs were then harvested and sectioned for IF staining. Metastatic lesions were confirmed by H&E staining of the tissues (Supplementary Fig. S2B). F4/80 was used as a general macrophage marker, while Siglec-F was used to identify AM. The neutral lipid dye LipidTox was used to determine the presence of intracellular lipids. Importantly, we were able to identify rare F4/80+ macrophages containing punctate, cytoplasmic lipid droplets (Fig. 4A). The vast majority of these cells also stained positively for Siglec-F. Additionally, they were found to stain positively for the LAM marker Trem2 (Supplementary Fig. S2C and S2D). In order to determine whether these findings could be replicated in an independent model, we repeated our analysis following transplant with HC11/R1-LM mammary tumors (Supplementary Fig. S2A). In this model, HC11 mammary epithelial cells express an inducible FGFR1 construct; upon administration of a homodimerizer, these cells develop mammary-gland tumors with a propensity to metastasize to the lung (Supplementary Fig. S2B). Once again, we were able to observe the presence of cytoplasmic lipids localized within F4/80+Siglec-F+ macrophages (Fig. 4B), which also stained positively for Trem2 (Supplementary Fig. S2C and S2D). Thus, in 2 distinct lung-metastasizing mammary-tumor models, we have identified macrophages containing punctate, cytoplasmic lipid droplets characteristic of foamy macrophages.
Metastasis-bearing lungs show increased numbers of LAMs within the AM compartment. A and B, IF microscopy of the lungs from mice bearing 4T1-Luc (A) or HC11/R1-LM mammary tumors (B) revealed punctate lipid droplets in the cytoplasm of F4/80+Siglec-F+ macrophages. C and D, Quantification of the total number of SSChiBODIPYhi macrophages per gram of lung tissue demonstrates an increase in LAMs in mice bearing metastases from 4T1-Luc (n = 4 mice/group; C) or HC11/R1-LM (n = 4 mice/group; D) tumors compared with nontumor control mice (n = 4–5 mice/group). Data are shown as mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001 as determined by ANOVA followed by Sidak multiple comparisons test.
Metastasis-bearing lungs show increased numbers of LAMs within the AM compartment. A and B, IF microscopy of the lungs from mice bearing 4T1-Luc (A) or HC11/R1-LM mammary tumors (B) revealed punctate lipid droplets in the cytoplasm of F4/80+Siglec-F+ macrophages. C and D, Quantification of the total number of SSChiBODIPYhi macrophages per gram of lung tissue demonstrates an increase in LAMs in mice bearing metastases from 4T1-Luc (n = 4 mice/group; C) or HC11/R1-LM (n = 4 mice/group; D) tumors compared with nontumor control mice (n = 4–5 mice/group). Data are shown as mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001 as determined by ANOVA followed by Sidak multiple comparisons test.
In addition to IF staining, foamy macrophages can be identified by flow cytometry by gating on the population of SSChiBODIPYhi macrophages (17). Five weeks after transplant of either 4T1-Luc or HC11/R1-LM mammary tumors reaching a volume of 1,250 to 1,500 mm3, lungs were harvested and immune cells isolated for analysis by flow cytometry. Lungs from nontumor-bearing, age-matched mice served as controls to establish the baseline levels of lipid staining in normal lung macrophages. After gating on live, CD45+ singlets, total macrophages were identified as Ly6G−CD64+MerTK+ cells (Supplementary Fig. S3A). From the total macrophage population, AM were then characterized as Siglec-F+ and IM as CD11b+. Quantification of SSChiBODIPYhi macrophages per gram of lung tissue demonstrated a significant increase in these cells among total macrophages compared with nontumor controls in both the 4T1-Luc (Fig. 4C) and HC11/R1-LM (Fig. 4D) tumor models. The number of SSChiBODIPYhi cells per gram of tissue within the Siglec-F+ gate similarly increased, accounting for the vast majority of all SSChiBODIPYhi macrophages (Fig. 4C and D). In comparison with metastasis-bearing lungs, the number of SSChiBODIPYhi macrophages per gram of primary 4T1-Luc tumor tissue was significantly lower (Supplementary Fig. S3B). Interestingly, the accumulation of SSChiBODIPYhi macrophages, particularly among the Siglec-F+ subset, occurred as early as 3 weeks post 4T1-Luc transplant, a time point that precedes metastatic colonization (Supplementary Fig. S3C and S3D). Along with the IF staining, these data suggest that the lipid-containing macrophages represent a subset of AM within metastasis-bearing lungs.
Metabolic and ECM-remodeling pathways are enriched in LAM
The observed relationship between LAM and AM prompted further interrogation of our scRNA-seq data to determine similarities and differences between these cell types at the gene expression level. Macrophage clusters identified by scRNA-seq were broadly grouped into lipid-associated (cluster 6), alveolar (clusters 0 and 8), or interstitial macrophages (clusters 1, 2, 3, and 5) for GSEA (Supplementary Dataset 4). Compared with IM, both LAM and AM demonstrated positive enrichment for genes involved in lipid metabolism, including peroxisome proliferator-activated receptor (PPAR) signaling, cholesterol homeostasis, and lipid catabolism (Fig. 5A and B). Similarly, both showed negative enrichment compared with IM for pathways such as interferon signaling and binding to MHC-II (Fig. 5A). Interestingly, while a number of metabolic pathways demonstrated similar enrichment between LAM and AM, significant differences were observed in pathways related to ECM remodeling, with LAM demonstrating enrichment of ECM (Fig. 5C) and collagen-associated pathways relative to both IM and AM. Genes related to lysosomal function were also selectively enriched in LAM compared with either of the other subsets. Additionally, LAM demonstrated enrichment for genes related to the regulation of T-cell responses, as well as epithelial–mesenchymal transition (EMT; Fig. 5A). Compared with AM, LAM also demonstrate enrichment of genes involved in endothelial-cell proliferation, which may be suggestive of a role in promoting angiogenesis (Fig. 5A). These data suggest that, while lipid-associated and alveolar macrophages are related, distinct differences in gene expression do exist that may point to unique functions within the TME. Overall, LAM demonstrate selective enrichment for genes involved in a wide array of protumorigenic pathways.
LAMs demonstrate enrichment of genes involved in pathways related to lipid metabolism, ECM remodeling, and immunosuppression. A, Dot plots displaying similarities and differences between LAM, IM, and AM in enrichment of genes related to defined MSigDB pathways involved in lipid metabolism, extracellular matrix, and immune responses. B, Heatmap of relative expression among metastasis-associated macrophage subsets of core enrichment genes defining the GO_LIPID_CATABOLIC_PROCESS pathway. C, Heatmap of relative expression among metastasis-associated macrophage subsets of core enrichment genes defining the GO_EXTRACELLULAR_MATRIX pathway.
LAMs demonstrate enrichment of genes involved in pathways related to lipid metabolism, ECM remodeling, and immunosuppression. A, Dot plots displaying similarities and differences between LAM, IM, and AM in enrichment of genes related to defined MSigDB pathways involved in lipid metabolism, extracellular matrix, and immune responses. B, Heatmap of relative expression among metastasis-associated macrophage subsets of core enrichment genes defining the GO_LIPID_CATABOLIC_PROCESS pathway. C, Heatmap of relative expression among metastasis-associated macrophage subsets of core enrichment genes defining the GO_EXTRACELLULAR_MATRIX pathway.
LAM exhibit impaired phagocytosis
Based on our findings suggesting that LAM represent a subset of AM, we next sought to determine whether AM depletion would result in a loss of the LAM population. To deplete AM, liposomal clodronate was delivered via i.n. administration every other day for 2 weeks, following resection of primary 4T1-Luc tumors. Control mice received an equal volume of empty liposomes. Administration of i.n. clodronate leads to a significant reduction in lung-metastatic burden (Supplementary Fig. S4A). After 2 weeks of treatment, lungs were harvested and processed for flow cytometry. Quantification of the relative frequencies of alveolar and interstitial macrophages confirmed robust and selective depletion of AM (Fig. 6A). Additionally, the number of SSChiBODIPYhi macrophages per gram of tissue significantly decreased in both the total and AM populations (Fig. 6B). No such change was observed in the IM population. These data lend support to the hypothesis that LAM are related to AM and may represent a subset of this tissue-resident cell-type that is expanded in the context of metastatic disease. Interestingly, while the total number of SSChiBODIPYhi AM per gram of tissue decreased, indicating efficacy of depletion by liposomal clodronate, the frequency of SSChiBODIPYhi LAM within the total AM population increased (Fig. 6C), suggesting that this population may be more resistant to clodronate depletion than other alveolar macrophages.
LAMs demonstrate reduced phagocytic capacity compared with non–lipid-laden macrophages. A, Intranasal administration of clodronate liposomes results in a decrease in the relative frequency of AMs compared with interstitial macrophages (n = 5 mice/group). B, AM depletion results in a reduction in the number of SSChiBODIPYhi cells per gram of tissue among total macrophages and AM (n = 5 mice/group). C, The frequency of AM exhibiting positive BODIPY 493/503 staining increases following liposomal clodronate depletion compared with controls (n = 5 mice/group). D, BMDMs loaded with soluble cholesterol or oxLDL in culture exhibit decreased phagocytosis of Zymosan particles compared with non–lipid-loaded controls. Data are shown as mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001 as determined by ANOVA followed by Sidak or Dunnett multiple comparisons test.
LAMs demonstrate reduced phagocytic capacity compared with non–lipid-laden macrophages. A, Intranasal administration of clodronate liposomes results in a decrease in the relative frequency of AMs compared with interstitial macrophages (n = 5 mice/group). B, AM depletion results in a reduction in the number of SSChiBODIPYhi cells per gram of tissue among total macrophages and AM (n = 5 mice/group). C, The frequency of AM exhibiting positive BODIPY 493/503 staining increases following liposomal clodronate depletion compared with controls (n = 5 mice/group). D, BMDMs loaded with soluble cholesterol or oxLDL in culture exhibit decreased phagocytosis of Zymosan particles compared with non–lipid-loaded controls. Data are shown as mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001 as determined by ANOVA followed by Sidak or Dunnett multiple comparisons test.
One potential explanation for the resistance to liposomal-clodronate depletion among LAM is impaired uptake of the particles, prompting us to investigate the impacts of excess lipid content on phagocytosis. BMDMs were harvested from WT BALB/c mice and differentiated in culture in the presence of M-CSF. Foamy macrophages were produced by overnight culture in low-serum media containing either oxidized low-density lipoprotein (oxLDL) or soluble cholesterol. Lipid-loaded BMDMs closely resemble LAM, as evidenced by the high concordance in their expression of the top LAM-defining genes (NES = 1.51, FDR = 0.15; Supplementary Fig. S4B and C; ref. 28). Control or lipid-loaded macrophages were then assessed for their capacity to phagocytose Zymosan particles. Compared with control macrophages, both cholesterol-loaded and oxLDL-loaded macrophages demonstrated a diminished phagocytic capacity (Fig. 6D). Together, these data suggest that reduced phagocytosis represents one mechanism by which lipid accumulation contributes to the development of protumorigenic macrophages within metastasis-bearing lungs.
Discussion
In the present study, we employed scRNA-seq to determine the extent of macrophage heterogeneity in the lungs of mice bearing metastases from primary mammary tumors. Our data suggest that at least 7 unique macrophages clusters are present in metastasis-bearing lungs. Macrophages have been shown to be critical regulators of tumor progression, both at the primary and metastatic sites (4, 29) and an understanding of their complex interplay with tumor cells and other components of the TME could provide novel avenues for improved treatment of patients with breast cancer.
At homeostasis, AM represent the most abundant lung-tissue resident population, with IM making up a smaller fraction of resident-mononuclear phagocytes. However, in response to insult or injury, increases in both populations can be observed as inflammatory monocytes are recruited and local proliferation is initiated to fill the open niche (9, 14, 30). Importantly, our work herein demonstrates that the gene-expression profiles of IM in the context of metastatic disease are highly varied. For instance, we find evidence of macrophages whose gene-expression signature suggests a predominant role in antigen presentation (cluster 1) and, perhaps, stimulation of antitumor immune responses. Alternatively, our data also reveal the presence of macrophages with enriched expression of genes related to matrix remodeling and myeloid-cell recruitment (cluster 5), suggestive of a protumorigenic signature. We also find evidence of an IFN-sensitive macrophage population (cluster 3). IFN signaling in tumors represents a complex interplay between various pro- and antitumor effects. In addition to serving critical functions in the early priming and activation of T cells, IFNs can have direct impacts on tumor-cell survival and angiogenesis; conversely, upon prolonged exposure, the effects become predominantly immunosuppressive, promoting T-cell exhaustion and tumor escape (31, 32). Future work defining the contributions of these populations to metastatic progression is warranted.
Our data also demonstrate the existence of self-renewing populations of both AM (cluster 8) and IM (cluster 2). These findings offer support of previous studies in which scRNA-seq identified subpopulations of proliferative resident AM in the lung airways (33, 34). Our data demonstrate that a proliferative subpopulation of IM also exists; these cells may be similar to a recently identified immunoregulatory subset of IM within naïve lungs (35). Importantly, our data does not exclude the possibility that BMDMs also contribute to either AM or IM subsets, as this progenitor population was not included in our scRNA-seq analysis.
Of particular interest, this work identifies a population of LAM previously uncharacterized within metastasis-bearing lungs. Lipid-associated, or foamy, macrophages have been extensively characterized in atherosclerosis and neurodegenerative diseases (18, 36). Foamy macrophages have also been described in the progression of tuberculosis and other pulmonary disorders (37, 38). Furthermore, a recent investigation of colorectal liver metastases demonstrated that TAM with larger area display higher intracellular complexity and are enriched for genes related to lipid metabolism (25), suggesting that a similar population may exist at this metastatic site. Importantly, our data demonstrate an accumulation of these cells at a premetastatic time point, suggesting a role for LAM in the early development of the metastatic niche. Data demonstrating the importance of lipid accumulation in TAM has begun to delineate the mechanisms through which altered metabolism promotes protumorigenic function (39). Interestingly, there is also recent evidence that neutrophils can accumulate lipids in the premetastatic lung, promoting lesion progression by acting as an energy reserve for colonizing breast-cancer cells (40).
Our data suggest that, in the context of breast cancer metastasis to the lung, lipid-laden macrophages represent a subset of tissue-resident AM. In addition to similarities in gene expression related to metabolic pathways, we find that the vast majority of LAM costain with the AM marker Siglec-F, and that selective AM depletion by intranasal clodronate administration significantly diminishes this population. Consistent with the role of AM in controlling inflammation, these LAM demonstrate enrichment for genes involved in negative regulation of T-cell responses and immunosuppression. Importantly, studies utilizing mice deficient in Trem2 have demonstrated its role in suppression of antitumor T-cell responses in numerous murine-tumor models (41). As such, Trem2 is emerging as a key regulator of macrophage function in cancer, akin to its role in other disorders characterized by metabolic dysfunction (42). Additionally, recent work has shown that the hypoxic conditions within breast tumors promote the expression of Lgals3 in macrophages; galectin-3 in turn promotes tumor growth and progression through multiple mechanisms, including immunosuppression (43, 44). Similarly, evidence from the literature indicates that lipid-laden macrophages are associated with immunosuppressive functions. For example, lipid accumulation in macrophages in response to lung injury drives an antiinflammatory, profibrotic phenotype (38). However, AM also serve to regulate surfactant homeostasis, and surfactant proteins have been shown to inhibit tumor progression, in part through regulating proinflammatory macrophage polarization, leading to enhanced immune-cell recruitment (45). Finally, macrophages are key modulators of arginine metabolism, which has been linked to surfactant regulation, highlighting the complexities of the microenvironment in which LAM reside (46). Thus, further investigation is needed to identify how signaling between surfactant and LAM may contribute to their dysregulated metabolism and immunosuppressive phenotype.
In the present study, we determined that lipid-laden macrophages display reduced phagocytic capacity compared with controls. One mechanism proposed for suppression of phagocytosis is via binding of surfactant proteins to SIRP-α (47). While increased differential expression of SIRP-α was observed in AM (cluster 0), no such enrichment was observed in LAM, suggesting an alternative mechanism for the observed reduction in phagocytosis. In fact, discrepancy in the literature regarding whether surfactant proteins promote or suppress phagocytosis by AM indicates that its function may be highly context dependent (47, 48). In atherosclerosis, it has been shown that foamy macrophages demonstrate an impaired ability to clear apoptotic cells in vivo, through mechanisms including competition for macrophage binding (49). Furthermore, in aging and degenerative brain tissue, a robust buildup of lipids is found within brain-resident microglia, resulting in defective phagocytosis, increased reactive oxygen species (ROS) production, and inflammation (36). In tumors, a failure to recognize and phagocytose malignant cells could result in an impaired capacity for antigen presentation and stimulation of an adaptive antitumor immune response. Recent efforts aimed at targeting phagocytosis checkpoints, such as the CD47-SIRPα signaling axis, have demonstrated the importance of intact phagocytic function in controlling tumor progression (50).
In addition to their contributions to lipid metabolism, we also found that LAM show an enrichment for genes related to matrix remodeling, including cathepsins (Cstl, Ctsb, Ctsd) and matrix metalloproteinases (Mmp12, Mmp14). This is again in concordance with foamy macrophages in atherosclerosis (51, 52). Importantly, such activity is likely to promote metastatic progression, as these proteases are well characterized for their roles in degrading the ECM and promoting invasion and vascularization (53, 54). In both settings these cells also express osteopontin (OPN, Spp1), which has been shown to be a key mediator of immunosuppression in the metastatic niche (55). Enrichment for genes in pathways related to endothelial-cell proliferation, suggestive of angiogenesis, and EMT further support a role for LAM in the creation of a microenvironment permissive to metastatic progression. A subpopulation of profibrotic macrophages enriched for genes associated with LAM, including Spp1, Fn1, Trem2, and Apoe, was identified in human lung samples from patients with pulmonary fibrosis (56), suggesting the presence of similar subpopulations in human disease. More in-depth analysis of protein expression paired with scRNA-seq, in addition to further phenotypic analyses, will help to further elucidate the roles of LAM in metastatic progression and how this population might be targeted therapeutically.
Given the abundance of macrophages in both primary and metastatic breast-cancer lesions, and their potential for promoting or controlling tumor growth, understanding their immense heterogeneity is of critical importance. The identification of novel metastasis-associated macrophage subsets, and methods for distinguishing them, is key to developing a complete understanding of their vast potential. Here, we characterize the transcriptional heterogeneity of macrophages found within the lungs of mice bearing metastases from primary mammary tumors, identifying a LAM subset previously unseen in this context.
Authors' Disclosures
D.N. Huggins reports grants from American Cancer Society during the conduct of the study. J.W. Williams reports grants from NIH NHLBI during the conduct of the study. K.L. Schwertfeger reports grants from NIH and grants from American Cancer Society during the conduct of the study. No disclosures were reported by the other authors..
Authors' Contributions
D.N. Huggins: Conceptualization, formal analysis, funding acquisition, validation, investigation, visualization, methodology, writing–original draft. R.S. LaRue: Data curation, software, formal analysis. Y. Wang: Resources, methodology. T.P. Knutson: Software, formal analysis. Y. Xu: Software, formal analysis. J.W. Williams: Conceptualization, funding acquisition, writing–review and editing. K.L. Schwertfeger: Conceptualization, funding acquisition, methodology, project administration, writing–review and editing..
Acknowledgments
The authors thank Dr. Thomas Griffith (University of Minnesota) for providing the 4T1-Luc cells and Dr. Jeffrey Rosen (Baylor College of Medicine) for providing the initial HC11/R1 cells. They thank the University of Minnesota Genomics Center, Flow Cytometry Resource, and Center for Immunology Imaging Facility for expertise and contributions to this work. This work was supported by an American Cancer Society Postdoctoral Fellowship (PF-18-140-01-CSM, to D.N. Huggins), R00HL138163 (to J.W. Williams), R21CA235285, R01CA215052, and R01HD095858 (to K.L. Schwertfeger).
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.