Research Article

Two distinct interstitial macrophage populations coexist across tissues in specific subtissular niches

See allHide authors and affiliations

Science  15 Mar 2019:
Vol. 363, Issue 6432, eaau0964
DOI: 10.1126/science.aau0964

Tissue macrophages have a split personality

Resident tissue macrophages (RTMs) reside in various tissue-specific niches during development. They evince microenvironment-directed phenotypes that support host defense and tissue homeostasis. Chakarov et al. used single-cell RNA sequencing and fate-mapping of murine lung RTMs to interrogate RTM-subset heterogeneity, interrelationships, and ontogeny (see the Perspective by Mildner and Yona). In addition to alveolar macrophages, they identified two different interstitial macrophage populations. One population mostly abutted nerve fibers; the other population preferentially localized near blood vessels and appeared to support vessel integrity and inhibit inflammatory cell infiltration into tissues.

Science, this issue p. eaau0964; see also p. 1154

Structured Abstract


Resident tissue macrophages (RTMs) are a heterogeneous population of immune cells occupying multiple tissue niches and exhibiting microenvironment-specific phenotypes and functions. In certain tissues such as the brain, lung, and liver, embryonically derived RTMs maintain themselves by self-renewal, whereas others, including those in the gut, dermis, and pancreas, are replaced by monocytes, at levels that are tissue specific. Once they arrive in their tissue of residence, monocytes undergo extensive differentiation according to molecular cues provided by their distinct tissue-specific niches, enabling their development into specialized RTMs that support local tissue function.


As a result of this ontogenetic and tissue niche heterogeneity, each tissue contains multiple populations of macrophages. For example, in the murine lung, alveolar macrophages are the major embryonically derived population in the alveolar spaces, whereas a minor population named interstitial macrophages (IMs) resides within the lung parenchyma. Previous results reported several phenotypically distinct IM subpopulations, whose relationship remained unknown. Do they represent independent populations or, rather, different points on the spectrum of maturation and activation states? How do these differences relate to their localization in tissue or roles in tissue function in health and disease? Does such macrophage heterogeneity also exist in other tissues?


Here, using single-cell mRNA sequencing, we unbiasedly identified two independent populations exhibiting distinct gene expression profiles and phenotypes: Lyve1loMHCIIhiCX3CR1hi (Lyve1loMHCIIhi) and Lyve1hiMHCIIloCX3CR1lo (Lyve1hiMHCIIlo) RTMs. We uncovered evidence of parallel populations in multiple others tissues, including the heart, fat, and dermis, as well as in human lung and omental and subcutaneous fat tissues, suggesting that a similar dichotomy is observed in human tissues.

We further demonstrated that both populations are slowly replaced by Ly6Chi monocytes. Importantly, using complementary fate-mapping models, we showed that monocyte-derived RTMs (MRTMs) are two separate lineages, rather than representing points along a developmental or maturation continuum. Notably, these distinct MRTM populations preferentially reside within different, but conserved, subtissular niches, located either adjacent to nerve bundles and fibers (Lyve1loMHCIIhi) or blood vessels (Lyve1hiMHCIIlo) across tissues.

Finally, by acutely depleting Lyve1hiMHCIIlo MRTMs using a mouse model of inducible macrophage depletion during the induction of fibrosis, we found that the absence of Lyve1hiMHCIIlo IMs exacerbated experimental lung and heart fibrosis, demonstrating their critical role in tissue inflammation.


Two independent MRTMs populations exist across tissues with specific niche-dependent phenotype and functional programming. Their different roles in homeostasis, immune regulation, and fibrosis renders them attractive and separate cellular targets for the therapeutic exploitation of RTM subsets.

Two independent populations of MRTMs exist across most tissues in steady state with conserved specific subtissular niche-dependent phenotype.

(Top) Lyve1loMHCIIhi are mostly located adjacent to nerves, wherease Lyve1hiMHCIIlo preferentially reside alongside blood vessels. CX3CR1, fractalkine receptor; MHCII, major histocompatibility complex class II; IL-10, interleukin 10; CD206, mannose receptor. (Bottom) Depletion of Lyve1hiMHCIIlo MRTMs in a model of lung fibrosis exacerbated vessel permeability, immune cell infiltration, and collagen deposition, demonstrating their critical role in restraining inflammation and fibrosis. DC, dendritic cell.


Macrophages are a heterogeneous cell population involved in tissue homeostasis, inflammation, and various pathologies. Although the major tissue-resident macrophage populations have been extensively studied, interstitial macrophages (IMs) residing within the tissue parenchyma remain poorly defined. Here we studied IMs from murine lung, fat, heart, and dermis. We identified two independent IM subpopulations that are conserved across tissues: Lyve1loMHCIIhiCX3CR1hi (Lyve1loMHCIIhi) and Lyve1hiMHCIIloCX3CR1lo (Lyve1hiMHCIIlo) monocyte-derived IMs, with distinct gene expression profiles, phenotypes, functions, and localizations. Using a new mouse model of inducible macrophage depletion (Slco2b1flox/DTR), we found that the absence of Lyve1hiMHCIIlo IMs exacerbated experimental lung fibrosis. Thus, we demonstrate that two independent populations of IMs coexist across tissues and exhibit conserved niche-dependent functional programming.

Macrophages are a diverse population of immune cells occupying multiple tissue niches and exhibiting microenvironment-specific phenotypes and functions. Together, these immune cells respond to invading pathogens and support tissue development and homeostasis. By contrast, their dysregulation is linked with numerous pathological processes, including inflammatory diseases, cancer, and fibrosis, making them potential therapeutic targets.

All resident tissue macrophage (RTM) populations are established during development by embryonic precursors that are either yolk sac macrophages or fetal liver monocytes (16). In certain tissues, such as the brain, lung, and liver, these macrophages maintain themselves in adulthood by self-renewal (16). Other macrophage populations, including those in the gut, dermis, and pancreas, are maintained by replacement from the blood-circulating monocyte pool, at tissue-specific levels (7). Once progenitors arrive at their tissue of residence, they undergo extensive differentiation according to molecular cues provided by their tissue-specific niche. This process enables progenitors to develop into specialized RTMs that support tissue function. As a result of this ontogenetic and tissue niche heterogeneity, each tissue contains multiple populations of macrophages. For example, in the murine lung, alveolar macrophages (AMs) are the major embryonically derived population in the alveolar spaces (3). A minor RTM population of interstitial macrophages (IMs), thought to derive from blood monocytes (8), also resides within the lung parenchyma (9) and comprises three phenotypically distinct subpopulations (10). However, the relationship between these IM subsets is unknown.

Do they represent independent populations or, rather, different points on the spectrum of macrophage maturation and activation states? How do these differences relate to their localization in tissue or roles in tissue function in health and disease? Finally, does such macrophage heterogeneity also exist in other tissues? To address these questions, we isolated IMs from murine lungs and used single-cell mRNA sequencing (scRNA-seq) to identify two independent monocyte-derived populations exhibiting distinct gene expression profiles, phenotypes, and functions, including key roles in protection from induced fibrosis. We uncovered evidence of comparable populations in other tissues, including the heart, fat, and dermis, where these two populations consistently reside in discrete microenvironmental niches. A Lyve1loMHCIIhiCX3CR1hi (Lyve1loMHCIIhi) population was mostly found surrounding the nerves, whereas Lyve1hiMHCIIloCX3CR1lo (Lyve1hiMHCIIlo) macrophages were often closely associated with blood vessels across tissues.


Unbiased identification of lung IM subpopulations by nonsupervised scRNA-seq

We isolated IMs from the lungs of wild-type (WT) C57BL/6 mice by fluorescence-activated cell sorting (FACS) (see fig. S1A for gating strategy) and used scRNA-seq to generate transcriptional profiles for each individual cell using the Smart-seq2 method (11) (see fig. S1B for quality control and table S1 for number of expressed genes detected in individual cells). From these transcriptomes, 1309 genes exhibited significant variability in expression level between cells. Using principal components analysis (PCA) and dimensionality reduction by the t-distributed stochastic neighbor embedding (t-SNE) method, the expression patterns of these 1309 genes delineated two distinct cell clusters within the lung IM population (Fig. 1, A and B). Cluster 2 showed higher expression levels of genes associated with wound healing, such as Tfgb2 and Plaur, whereas cluster 1 had increased expression of immune response activation and antigen presentation genes [such as the major histocompatibility complex class II (MHCII)–encoding genes H2-Aa, H2-Ab1, H2-DMb1, and H2-Eb1] and Cx3cr1 (fractalkine receptor) (Fig. 1C and fig. S2A). The differential expression of MHCII protein at the cell surface of IMs was confirmed by flow cytometry and similarly separated IMs into two subpopulations (Fig. 1D).

Fig. 1 Unbiased identification of lung IM subpopulations by nonsupervised scRNA-seq.

(A to C) Seurat analysis of single FACS-purified lung IMs defines two populations: cluster 1 (cyan) and cluster 2 (red), with distinct patterns of gene expression. Seurat PC analysis and heat-map representation of the 30 highly DEGs that define PC1 are shown in (A). Each row defines the gene expression in the corresponding cell in the PCA above. Genes highlighted in cyan are those MHCII-related genes expressed in cluster 1. Seurat t-SNE analysis and heat map–style representation of the six most highly DEGs that explain the t-SNE1 and t-SNE2 are shown in (B). Heat map of the 100 most highly DEGs in clusters 1 and 2 defined by Seurat are shown in (C). (D) Flow cytometric analysis of lung AMs and IMs and identification of two IM populations: population 1 in cyan and population 2 in red. (E) Volcano plots of the relative difference in expression level of all DEGs (n =1837) between MHCII and MHCII+ IMs (blue). −log10 (P value) ≥ 4 and −2 ≥ log2 (fold change) ≤ 2 DEGs (n = 126) are shown in red; genes of interest are highlighted in yellow. (F) The 50 genes from (E) with the highest –log10 (P value) and log2 (fold change) are plotted as a heat map. Genes highlighted in red are specific for cluster 2, and genes highlighted in cyan are specific for cluster 1. DEG P values were calculated using the bimodal test from the Seurat package.

Lung Lyve1loMHCIIhi and Lyve1hiMHCIIlo IM subpopulations arise from monocytes

To understand whether the IM subpopulations delineated by differential MHCII expression were the same as the two cell clusters identified by scRNA-seq, we sorted lung MHCII and MHCII+ IM subpopulations for bulk RNA-seq analysis (Fig. 1D). We detected 1837 significantly differentially expressed genes (DEGs) between MHCII and MHCII+ subpopulations (Fig. 1E). Among the 50 top DEGs were genes encoding surface markers, including Lyve1 (lymphatic endothelium hyaluronan receptor-1) and Cd38 in the MHCII cluster. Cx3cr1 and Ccr2 expression was higher in MHCII+ IMs (Fig. 1, E and F). We first compared our lung IM scRNA-seq data to the bulk RNA-seq data using connectivity MAP (cMAP) (12), which generates scores (as scaled dimensionless quantities) indicative of the degree of “closeness” of one cell subset to a defined signature gene set. cMAP confirmed that scRNA-seq clusters 1 and 2 corresponded to MHCII+ and MHCII IM subpopulations, respectively (fig. S2B). Moreover, transcriptional comparison of both IM populations to sorted AMs and lung monocytes by RNA-seq revealed their similarities with lung monocytes in PCA analysis (fig. S2C), sharing common genes such as Cd63, Maf, Fcgrt, Cd36, and Marco with AMs as well as monocyte core genes, including Ccr2, Ceacam1, Ly6e, and Cx3cr1, while exhibiting their own signature comprising MHCII-related genes (H2-Aa, H2-Ab1, and H2-Eb1), Cd38, Cd163, and S1pr1 (fig. S2D).

We validated these results by measuring the expression of selected surface proteins that were differentially expressed between MHCII versus MHCII+ IM subpopulations at the mRNA level. We found that CX3CR1 (Fig. 1F and Fig. 2A), CD44, CLEC7A, and CD72 were more highly expressed on MHCII+ IMs, whereas CD36, CD38, C5AR1 (CD88), CD169, Lyve1, Mrc1 (CD206), and MafB were more highly expressed on MHCII IMs (fig. S2E). On the basis of their specfic and clear Lyve1 expression, and the previous descriptions of tissue macrophages expressing Lyve1 (1316), we named these cells MHCII Lyve1hiMHCIIlo and MHCII+ Lyve1loMHCIIhi. Interestingly, we also observed that Lyve1hiMHCIIlo IMs seemed larger compared with their Lyve1loMHCIIhi counterparts, yet exhibited a similar morphology (Fig. 2A).

Fig. 2 Identification of two populations of monocyte-derived IMs present in murine lungs.

(A) Representative flow cytometric plot showing identification of Lyve1loMHCIIhi and Lyve1hiMHCIIlo macrophages in the lung are shown on the left, and electron microscopy images of FACS-purified Lyve1loMHCIIhi and Lyve1hiMHCIIlo macrophages and flow cytometric measurements of their respective levels of CX3CR1 expression in Cx3cr1GFP/+ mice (geometric means of GFP signal are shown) are on the right. (B) Flow cytometric measurement of the frequency of EYFP+ IM subpopulations (top) or AMs and lung monocytes (bottom) in 8-week-old S100a4Cre-RosaEYFP mice are shown on the left, and the percentages of EYFP+ cells over time in Lyve1hiMHCIIlo and Lyve1loMHCIIhi IMs, lung monocytes (Lg Mono), and AMs from birth (week 0) to adulthood (week 20) are shown on the right. (C) Representative flow cytometric plots showing the relative frequencies of CD45.1+ and CD45.2+ cells within the Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs lung monocytes, and AM populations from the Cd45.2+ Ccr2−/− parabiont after 6 months of surgical attachment to WT Cd45.1+ counter partner are shown on the left; the average percentage chimerism of each population is displayed on the right. (D) Flow cytometric analysis of the frequency of lung macrophages and monocytes from LyzMCre-SLco2b1flox/DTR mice 0 to 120 days after DTx treatment (top); mean ± SEM absolute cell numbers of AM, lung monocytes (Lg Mono), and Lyve1hiMHCIIlo and Lyve1loMHCIIhi IMs at the indicated time points post-DTx treatment (bottom). A representative plot of two independent experiments is shown (n = 4). Bars represent means ± SEM. **P ≤ 0.01, ****P ≤ 0.0001, and ns is not significant using Student’s t test. Scale bars in (A) represent 1 μm.

To further ascertain the macrophage identity of Lyve1hiMHCIIlo and Lyve1loMHCIIhi cells, we first measured their respective expression of the dendritic cell (DC) master regulator transcription factor ZBTB46 (17, 18) by flow cytometry. As expected, both Lyve1hiMHCIIlo and Lyve1loMHCIIhi macrophages, like lung monocytes and AMs, did not express ZBTB46 compared with classical DCs (fig. S2F). We also evaluated their recombination labeling in the Csf1rMerCreMer-RosatdTomato macrophage fate-mapping model. Both populations were labeled at comparable levels to AMs and monocytes (fig. S2G), supporting their macrophage identity.

Murine lung IM populations may be replenished by circulating blood monocytes in steady state in adults (8, 19). We therefore assessed the contribution of monocytes to the identified lung IM subsets, using the S100a4Cre-RosaEYFP murine fate-mapping model (4, 19). In adult S100a4Cre-RosaEYFP mice, all bone marrow (BM) hematopoietic cells, including monocytes, are 100% enhanced yellow fluorescent protein (EYFP)–labeled (4). However, we found that the S100a4 gene was expressed either not at all or at a very-low relative level in adult AMs and IMs compared with its expression in monocytes (fig. S2H). Thus, if murine lung IMs are replaced by monocytes, the percentage of labeled IMs should increase with time. About 50% of AMs and IMs were EYFP+ at birth (Fig. 2B) (4), and AM labeling frequency did not increase with age (Fig. 2B). This is consistent with a lack of blood monocyte replacement and rather local self-renewal (3). However, EYFP+ IM frequency increased between birth and 20 weeks of age, up to ~80% in both subpopulations (Fig. 2B), suggesting replacement by blood monocytes.

We confirmed these observations using parabiotic mouse models. We surgically joined the circulatory systems of CD45.1+ WT mice and Cd45.2+ Ccr2−/− mice, which are defective in monocyte egress from the BM. Cd45.1+ WT cell chimerism within the Cd45.2+ Ccr2−/− host was then monitored for 6 months (Fig. 2C). Efficient parabiosis was confirmed by examining the chimerism of blood B cell and T cell populations, which is independent of CCR2 status (fig. S2I, left panel). By contrast, and as expected (19), WT Ly6Clo and Ly6Chi monocytes represented >70 and >90% of blood circulating monocytes in Ccr2−/− mice, respectively (fig. S2I, right panel). No exchange was observed in the AM population of either parabiont (Fig. 2C) (19). However, WT partner cells reconstituted 20% of Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs (Fig. 2C), indicating that both IM populations are equally replenished by CCR2-dependent cells in adulthood.

We also adoptively transferred monocytes into a flox–diptheria toxin receptor (DTR)–based macrophage-depletion mouse model and assessed their contribution to lung IMs. A previous study identified the high gene expression of the solute carrier organic anion transporter family member 2B1 (Slco2b1) in lung macrophages (20). We also confirmed Slco2b1 expression in tissue macrophages (spleen, brain, lung, and gut), but not in Ly6Chi or Ly6Clo monocytes or DC subsets (fig. S3A). We verified Slco2b1 expression in sorted lung IM subsets (fig. S3B), which was further confirmed by flow cytometry (fig. S3C). Thus, we generated a mouse model in which a loxP-flanked stop cassette upstream of the DTR-coding region was placed before the 3′ untranslated region of Slco2b1 (Slco2b1flox/DTR mouse) (fig. S3D). We crossed Slco2b1flox/DTR to LyzMCre mice (LyzMCre-Slco2b1flox/DTR, called Lyz-SLCO hereafter) to specifically target macrophages (fig. S3D). As expected, 16 hours after intranasal diptheria toxin (DTx) administration, the MerTK+CD64+ population, including both AMs and IMs, was totally depleted (Fig. 2D). Other myeloid and lymphoid populations were unaffected by DTx treatment (fig. S3E). Different repopulation kinetics were observed for AMs and IMs (Fig. 2D), with rapid repopulation of both IM subsets after monocyte infiltration (Fig. 2D). AMs repopulated only 3 weeks after depletion and returned to normal levels 120 days after DTx treatment (Fig. 2D). Using a lower dose of DTx, we observed the partial depletion of both AMs and IMs (fig. S3F). Consequently, we used this latter dose to assess whether monocytes contribute to both populations. Cd45.2+ Lyz-SLCO mice were treated with DTx and transferred with CD45.1+ Ly6Chi monocytes. We observed monocyte-derived Cd45.1+ macrophages in both AM and IM populations 9 days after transfer, confirming that Ly6Chi monocytes can give rise to both IM populations (fig. S3G).

Macrophage dichotomy is conserved across tissues and species

There is emerging evidence that macrophage heterogeneity has important roles across different tissues. Similar to lung Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs (Fig. 2A), Wolf et al. described two populations of CX3CR1 and CX3CR1+ brown adipose tissue macrophages (ATMs) (21), whereas Hulsmans et al. observed CX3CR1+ distal atrioventricular macrophages (AVMs) in the heart that have important roles in cardiac conduction (22). We hypothesized that such populations correspond to our lung IM subsets, and consequently, this IM dichotomy is conserved across tissues.

We first compared lung IM transcriptomes generated by scRNA-seq with published transcriptomes of ATMs (21) and AVMs (22). cMAP analysis revealed that lung Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs aligned closely with the gene expression patterns of CX3CR1 and CX3CR1+ brown ATMs, respectively (Fig. 3A, upper panel). When comparing lung IMs with AVM populations, we found that some cells did not exhibit transcriptional similarities with our IM subsets (13/76), although most (34/76) of the AVMs exhibited transcriptional similarities with lung Lyve1loMHCIIhi IMs (table S2). By contrast, a minority (29/76) exhibited transcriptional similarities to lung Lyve1hiMHCIIlo IMs (table S3) (Fig. 3A, lower panel). Comparing the DEGs of the Lyve1loMHCIIhi-like with those of the Lyve1hiMHCIIlo-like AVMs, we detected a higher expression level of MHCII-related genes (H2-Eb1, and Cd74) in the former and transcripts encoding Lyve1 in the latter (fig. S4A). Finally, we took advantage of the recently published mouse cell atlas single-cell database (23), projecting our lung IM subset gene signatures onto bladder, lung, mammary gland, muscle, pancreas, and uterus atlas single-cell macrophages (fig. S4B). Importantly, we found that some of these macrophages exhibit transcriptional similarity with our lung IM subsets in all tissues (fig. S4B). Here again, among the DEGs, Lyve1loMHCIIhi-like macrophages expressed MHCII-related genes, whereas Lyve1hiMHCIIlo-like macrophages expressed Lyve1 and Mrc1 (mannose receptor or CD206) (fig. S4C).

Fig. 3 Identification of Lyve1loMHCIIhi and Lyve1hiMHCIIlo IM subpopulations across tissues.

(A) cMAP analysis of single lung IMs showing their enrichment for CX3CR1+ and CX3CR1 macrophage gene sets (top) and cMAP analysis of single heart macrophages showing their enrichment for lung Lyve1loMHCIIhi and Lyve1hiMHCIIlo IM gene sets (bottom). (B) Representative flow cytometric plots showing the identification of Lyve1loMHCIIhi and Lyve1hiMHCIIlo macrophages in heart, fat, and skin (left); electron microscopy images of FACS-purified Lyve1loMHCIIhi and Lyve1hiMHCIIlo macrophages from the indicated tissues (middle); and their respective levels of CX3CR1 expression in Cx3cr1GFP/+ mice (geometric means of GFP signal are shown) (right). A representative plot of three independent experiments is shown. (C and D) PCA analysis of gene expression patterns in Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs from the lung, heart, fat, and skin before (C) and after [(D), left] the removal of tissue-specific probes. The heat map representation in [(D), right] shows the 50 most highly DEGs that define PC1. Genes expressed in Lyve1loMHCIIhi IMs are highlighted in cyan; genes expressed in Lyve1hiMHCIIlo IMs are highlighted in red. (E) cMAP analysis of human macrophages showing their enrichment for lung Lyve1loMHCIIhi and Lyve1hiMHCIIlo IM gene sets. (F) Volcano plot of the relative difference in expression levels of DEGs between 1003 Lyve1loMHCIIhi and 754 Lyve1hiMHCIIlo hIM cells identified by cMAP in (E). −log (P value) ≥ 2 DEGs are shown in blue; −log10 (P value) ≥ 3 and −0.25 ≥ log2 (fold change) ≤ 0.25 DEGs are shown in red; and genes of interest are highlighted in yellow. (G) Representative flow cytometric plots showing identification of Lyve1loMHCIIhi and Lyve1hiMHCIIlo-like macrophages in human lung. t-SNE analyses were performed using Cytofkit in gated DAPICD45+CD11b+CD16CD14+ cells. A representative plot of three independent patient samples is shown.

We validated that Lyve1loMHCIIhi and Lyve1hiMHCIIlo macrophage subsets were clearly detectable by flow cytometry in the heart, visceral fat, and skin, as well as in brown adipose tissue, the diaphragm, ganglia, trachea, tongue, testis, meninges, and bladder, with tissue-specific ratios of each subset (fig. S4D). Similar to lung IMs (Fig. 2A), heart, fat, and skin Lyve1hiMHCIIlo macrophages seemed larger compared with Lyve1loMHCIIhi macrophages but otherwise exhibited a similar morphology (Fig. 3B and fig. S4E). We next performed a comparative gene expression profile analysis to assess how closely IM subsets related to each other across tissues. We isolated total RNA from sorted Lyve1loMHCIIhi or Lyve1hiMHCIIlo IM subsets purified from lung, heart, fat, and skin (Fig. 3B) and identified 3652 DEGs between the two populations. PCA analysis of these DEGs indicated a strong effect of tissue imprinting, as PC1, PC2, and PC3 were clearly driven by the tissue of residence (Fig. 3C). We identified and removed tissue-specific probes from the analysis, as previously described (24) to reveal the “core” macrophage transcriptional profile of subsets from different tissues. This approach allowed us to cluster cells based on subtype, evidenced by PC1 and PC2 being mostly devoted to cell type, which separated tissue Lyve1loMHCIIhi and Lyve1hiMHCIIlo subpopulations (Fig. 3D). Among the top 50 genes defining the IM dichotomy across tissues were Lyve1, Timd4, Cd5l, Fcna, and Vsig4 for Lyve1hiMHCIIlo IMs, and Axl, Ccr2, Cx3cr1, and MHCII-related genes (H2-DMa, H2-Aa, H2-Eb1, H2-Ab1, Cd74, and H2-K1) for Lyve1loMHCIIhi IMs (Fig. 3D and fig. S4F). Furthermore, gene ontology analysis of the DEGs between Lyve1loMHCIIhi and Lyve1hiMHCIIlo macrophages revealed marked differences in the expression levels of functional gene sets between subsets. For example, genes involved in the complement system and blood vessel morphology were more highly expressed in Lyve1hiMHCIIlo IMs, whereas inflammation-related and chemotaxis signaling–related genes were more highly expressed in Lyve1loMHCIIhi IMs (fig. S4G).

We then asked whether Lyve1loMHCIIhi and Lyve1hiMHCIIlo populations were detectable in human tissues. We took advantage of a recently published single-cell dataset of human lung (from a lung tumor patient) and identified macrophages from the nonmalignant tissue using the same method described by Lambrechts et al. (25) (fig. S4H). We then projected our mouse lung bulk RNA-seq IM transcriptomes onto these human macrophage single-cell data using cMAP (analysis was performed with mouse orthologs of human transcripts; see table S4 and S5 for Lyve1loMHCIIhi and Lyve1hiMHCIIlo IM gene signatures, respectively). We found that 754/3873 and 1003/3873 of human lung macrophages exhibited transcriptional similarities with mouse lung Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs (called hIMs hereafter), respectively (Fig. 3E). This suggested that macrophage subsets transcriptionally similar to murine Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs likely coexist in the human lung. DEG analysis between these 754 Lyve1loMHCIIhi-related and 1003 Lyve1hiMHCIIlo-related cells revealed distinct gene signatures. Lyve1loMHCIIh hIMs expressed genes such as CXCR4, CLEC10A, and CCL17, whereas Lyve1hiMHCIIlo hIMs were enriched in macrophage-related genes such as MARCO, APOE, and CD163 (Fig. 3F). Moreover, projection of the cMAP score on the human lung t-SNE analysis revealed two distinct clusters mostly composed of Lyve1loMHCIIhi (fig. S4I, blue cells) and Lyve1hiMHCIIlo (fig. S4I, red cells) hIMs. Finally, we extended our analysis of human tissues (lung, omental, and subcutaneous fat tissues) using flow cytometry. t-SNE analysis of gated CD45+CD11b+CD14+CD16 human lung macrophages revealed two CD64+ populations: one characterized by high CD11b expression and the other expressing CD169, CD206, and Lyve1 (Fig. 3G and fig. S4J), markers also found in murine Lyve1hiMHCIIlo IMs. However, in contrast to murine IM subsets, MHCII (HLA-DR) and CX3CR1 were both expressed at higher levels in Lyve1+ cells, highlighting the fact that these markers cannot be used as in murine tissues to identify IM subsets (Fig. 3G). Nevertheless, we report the identification of two IM populations that share similar gene expression signatures across several mouse and human tissues.

Multitissue monocyte-derived Lyve1loMHCIIhi and Lyve1hiMHCIIlo macrophages are independent lineages

We then asked whether other tissue IM subsets were similarly replaced by BM-derived Ly6Chi monocytes, as shown for lung IM subpopulations (Fig. 2B). Using the S100a4Cre-RosaEYFP fate-mapping model (4, 19), we observed increased percentages of EYFP labeling with age in the fat and skin, mirroring the accumulation of monocyte-derived cells in the Lyve1loMHCIIhi and Lyve1hiMHCIIlo lung IM populations. By contrast, Lyve1loMHCIIhi heart macrophages did not show increased labeling (Fig. 4A).

Fig. 4 Lyve1loMHCIIhi and Lyve1hiMHCIIlo macrophage populations, across tissue, arise from monocytes and are independent lineages.

(A) The percentage of EYFP+ Lyve1loMHCIIhi and Lyve1hiMHCIIlo macrophages in indicated tissues from adult (7 to 20 weeks of age) S100Cre-RosaEYFP mice. (B) Flow cytometric analysis of the percentage of Ly6Chi and Ly6Clo monocytes in the blood of Cx3cr1CreER-RosaEYFP mice expressing EYFP after 7 days of dietary tamoxifen (TAM) treatment to induce EYFP expression in CX3CR1+ cells (left), and mean percentages of blood monocytes expressing EYFP at the indicated time points after the end of TAM treatment, measured by flow cytometry (right). (C) Flow cytometric plots showing the frequency of EYFP expression in lung monocytes, AMs, and Lyve1loMHCIIhi and Lyve1hiMHCIIlo macrophages from Cx3cr1CreER-RosaEYFP mice after 7 days of dietary treatment with TAM to induce EYFP expression in CX3CR1+ cells (left), and mean expression frequency of EYFP in Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs from lung, fat, heart, and skin measured at 0 to 15 days after dietary TAM treatment (right). (D) Flow cytometric analysis of Ly6Chi and Ly6Clo monocytes from the blood of Lyve1Cre/GFP-RosatdTomato mice (left), and the percentage of each monocyte population expressing tdTomato (right). (E) Flow cytometric measurement of tdTomato expression in Lyve1loMHCIIhi and Lyve1hiMHCIIlo macrophages from the lungs of 8-week-old Lyve1Cre/GFP-RosatdTomato mice (top), and the mean frequency of tdTomato+ blood Ly6C+ monocytes and Lyve1hiMHCIIlo and Lyve1loMHCIIhi IMs in the indicated tissues in mice at 5, 8, and 10 weeks of age (bottom). Bars represent means ± SEM. *P ≤ 0.01 and **P ≤ 0.001 by Student’s t test.

We next tested whether IM subsets are two separate lineages or instead represent two points on a developmental or maturation continuum, whereby Lyve1loMHCIIhi cells give rise to Lyve1hiMHCIIlo cells or vice versa. To answer this question, we used the tamoxifen-inducible Cx3cr1CreER-RosaEYFP reporter mouse model (26). Because Lyve1loMHCIIhi IMs express CX3CR1 (Fig. 2A), they should be labeled upon a tamoxifen pulse, as previously demonstrated for Ly6Chi and Ly6Clo monocytes (26). One week of continuous tamoxifen exposure to adult animals resulted in the efficient labeling of CX3CR1+Ly6Clo blood monocytes (Fig. 4B) and Lyve1loMHCIIhi IMs in the lung, heart, fat, and skin (Fig. 4C). Moreover, the frequency of EYFP labeling decreased with time after the end of the tamoxifen pulse, indicating their expected replacement by unlabeled monocytes. However, little or no labeling was detected in the Lyve1hiMHCIIlo populations of the tamoxifen-treated Cx3cr1CreER-RosaEYFP mice, and no further increase was observed over time (Fig. 4C). Thus, Lyve1loMHCIIhi cells do not give rise to Lyve1hiMHCIIlo cells. To test whether, instead, Lyve1hiMHCIIlo cells could give rise to Lyve1loMHCIIhi cells, we used Lyve1Cre/GFP-RosatdTomato mice (27). As previously described, hematopoietic stem cells are also labeled to a limited degree owing to the transient expression of Lyve1 during fetal development (28). Consequently, up to 55% of blood Ly6Chi and Ly6Clo monocytes were labeled in 5-week-old mice (Fig. 4D), whereas 100% of Lyve1hiMHCIIlo IMs were labeled in all organs. By contrast, Lyve1loMHCIIhi IMs exhibited lower tdTomato labeling (Fig. 4E), which decreased during adulthood, approaching the levels seen in blood Ly6C+ monocytes. Thus, Lyve1loMHCIIhi IMs appear to arise from blood monocytes. Furthermore, Lyve1hiMHCIIlo IMs do not appear to give rise to Lyve1loMHCIIhi IMs (Fig. 4E). The high level of Lyve1loMHCIIhi IM labeling at 5 weeks of age reflects their origin from highly labeled embryonic progenitors (28). Thus, Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs are two separate lineages, which arise from tissue-recruited monocytes, rather than representing points on a developmental or maturation continuum of the same population.

Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs occupy preferentially distinct niches that are conserved across tissues

Tissue macrophage populations sense microenvironmental cues and respond by modifying their transcriptional signatures and epigenetic marks (2931), leading to the expression of tissue-specific phenotypes and functions. Thus, we hypothesized that the two conserved populations of IMs identified here could only exist across tissues if common tissue cues or minimally shared niches also existed within those tissues. Recently, two groups identified CX3CR1+ sympathetic neuron–associated macrophages in brown adipose tissue, which are implicated in steady-state tissue innervation and norepinephrine-mediated regulation of thermogenesis (21, 32). We showed that these macrophages were likely equivalent to the Lyve1loMHCIIhi IM lineage (Fig. 3). In addition, Lyve1 is expressed on a subpopulation of macrophages in human and mouse tissues (1316). Lastly, our recent work identified a subpopulation of arterial macrophages that express Lyve1 and are associated with blood vessels invested with smooth muscle cells (33). In accordance with the concept that common niches across tissues support parallel IM diversification, these data suggest that Lyve1loMHCIIhi macrophages may be associated with tissue nerves, whereas Lyve1hiMHCIIlo macrophages might be associated with blood vessels.

Fig. 5 Pan-tissue–specific niches for Lyve1loMHCIIhi and Lyve1hiMHCIIlo IM populations.

(A to C) Immunofluorescence micrographs of 20-μm-thick lung sections (A) or whole-mount ear skin (B) and heart (C) from WT C57BL/6 mice. Lyve1hiMHCIIlo cells are visualized with anti-Lyve1 monoclonal antibody (mAb) (green), macrophages with anti-CD68 mAb [red in (A) and (B)], and blood vessels with anti-CD31 mAb [blue in (A) and (B) and red in (C)]. Scale bars represent 50 μm; arrows indicate Lyve1hiMHCIIlo IMs. DAPI, 4′,6-diamidino-2-phenylindole. (D to F) Immunofluorescence microscopy images of 20-μm-thick lung sections (D) or whole-mount ear skin (E) and heart (F) from WT Cx3cr1GFP/+ mice. CX3CR1+ cells are visualized with anti-GFP Ab (green), macrophages with CD68 mAb [red in (D) and (E)], and nerve bundles with anti–class III β-tubulin (TUBB3) Ab [blue in (D) and (E) and red in (F)]. Scale bars represent 50 μm; arrows indicate Lyve1loMHCIIhi IMs. A representative image of three independent experiments (n = 2 per group) is shown.

To test this hypothesis, we first studied the subtissular distribution of Lyve1hiMHCIIlo macrophages in the lung (Fig. 5A), skin (Fig. 5B), and heart (Fig. 5C) of WT C57BL/6 mice by immunostaining for Lyve1 and pan–blood vessel (CD31) and pan-macrophage (CD68) markers. In addition to nonperivascular Lyve1hiMHCIIlo IMs, which most likely associate with capillaries (34), there was a major population of Lyve1+ macrophages closely associated with CD31+ blood vessels in these organs (Fig. 5, A to C). The distribution of CX3CR1+ macrophages (corresponding to Lyve1loMHCIIhi IMs) was also addressed using a Cx3cr1GFP/+ mouse model (35). Further staining with an anti–class III β-tubulin (TUBB3) neuron-specific antibody revealed Lyve1loMHCIIhi IMs associated with nerve bundles in the lung (Fig. 5D), skin (Fig. 5E), and heart (Fig. 5F). In the lung, we found abundant innervation surrounding the bronchioles. In accordance with this observation, CX3CR1+ IMs were more frequently observed around bronchiole nerve bundles. In the heart, sympathetic nerves that are spread over the epicardial surface were closely associated with CX3CR1+ cells (36) (Fig. 5F). Additionally, CX3CR1+ cells were abundantly distributed throughout the epicardial surface of the heart. Finally, most Lyve1loMHCIIhi IMs were associated with nerves in these tissues (Fig. 5, D to F).

To confirm these observations, we used two lineage-tracing models, Cdh5CreER-RosatdTomato and Wnt1Cre-RosatdTomato, to directly image endothelial cells and nerves, respectively (3739). Tamoxifen administration to adult Cdh5CreER-RosatdTomato mice drives tdTomato fluorescent protein expression in all endothelial cells (fig. S5A). Together with Lyve1 and pan-macrophage Iba1 (40, 41) staining, we detected that Lyve1hiMHCIIlo IMs, but not Lyve1loMHCIIhi IMs, were preferentially located around blood vessels in lung, heart, fat, and skin (fig. S5A).

To visualize the subtissular localization of Lyve1loMHCIIhi IMs, we used Wnt1Cre-RosatdTomato crossed with Cx3cr1GFP/+ transgenic mice. As previously observed (Fig. 5, D to F), Lyve1loMHCIIhi IMs, but not Lyve1hiMHCIIlo IMs, were closely associated with tdTomato+ sympathetic neurons in all tissues (fig. S5B). Lyve1loMHCIIhi IMs were also present surrounding nerve bundles or fine nerve fibers (fig. S5B). Thus, Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs are preferentially localized across tissues within common subtissular niches, which likely provide specific cues supporting their differentiation and function.

Lyve1loMHCIIhi IMs possess superior antigen-presentation capacities

Several studies have described IMs as regulatory cells that spontaneously secrete interleukin-10 (IL-10) (8, 9) at levels that increase after CpG treatment (8). Accordingly, RNA-seq revealed significantly higher expression of the Il10 gene in both lung IM subsets, compared with AMs or lung monocytes, with Lyve1hiMHCIIlo IMs expressing significantly higher levels of Il10 than Lyve1loMHCIIhi IMs (fig. S6A). To validate the differential IL-10 expression between IM subsets in vivo, we used IL-10–β-lactamase reporter mice (8) and found that at steady state, both IM subsets spontaneously produced IL-10, but with a significantly higher proportion of Lyve1hiMHCIIlo IMs expressing the cytokine in lung (fig. S6B), heart, fat, and skin (fig. S6C). Moreover, because both IM subsets exhibited higher MHCII expression compared with other RTM populations, we tested whether they were able to process and present foreign antigens to T cells. We cultured sorted AMs, DCs, and IM subsets from WT mouse lungs with ovalbumin protein and purified carboxyfluorescein succinimidyl ester–labeled ovalbumin-specific CD4+ OTII T cells. Both IMs drove T cell proliferation, though to a lesser extent than DCs, whereas AMs, which do not express MHCII, were unable to activate T cells (fig. S6D). In agreement with their higher level of MHCII expression, Lyve1loMHCIIhi IMs more potently activated OTII cells than Lyve1hiMHCIIlo IMs (fig. S6D). Moreover, Lyve1loMHCIIhi IMs also induced a higher proportion of FoxP3-expressing regulatory CD4+ T cells (Tregs) (fig. S6E). Thus, IM subsets are both able to present antigen and stimulate CD4+ T cell proliferation and Treg differentiation. However, in both instances, Lyve1loMHCIIhi IMs are markedly more adept at doing so.

Lyve1hiMHCIIlo IMs restrain induced tissue fibrosis

Given their high bystander IL-10 secretion, we aimed to understand the immunoregulatory role of Lyve1hiMHCIIlo lung IMs in vivo. We first examined the scRNA-seq and bulk RNA-seq data for the expression of genes linked with immunoregulatory functions. Lung Lyve1hiMHCIIlo IMs expressed relatively high levels of genes linked with wound healing, repair, and fibrosis, including Cxcl2 (MIP-2), platelet factor 4 (Pf4 or Cxcl4), Tgfb2 (TGFβ2), and Plaur (μPAR), compared with Lyve1loMHCIIhi IMs (fig. S7A). Thus, to address the potential functional role of Lyve1hiMHCIIlo IMs, we used two models of induced fibrosis: bleomycin-induced lung fibrosis (42) and isoprenaline-induced heart fibrosis (43).

To enable us to specifically target Lyve1hiMHCIIlo macrophages for depletion, we further refined the SLCO model (Fig. 2D and fig. S3) by crossing Lyve1Cre/GFP mice (27) with our Slco2b1flox/DTR (SLCO) mice resulting in the Lyve1Cre/GFP-Slco2b1flox/DTR mouse model (called Lyve1-SLCO hereafter). Because Slco2b1 and Lyve1 genes are both expressed in several nonhematopoietic cells (4446), we generated BM chimeras where Cd45.1+ recipients were lethally irradiated before reconstitution with Lyve1-SLCO Cd45.2+ BM (Lyve1-SLCO→Cd45.1+) or Lyve1Cre/GFP CD45.2+ littermate control (Lyve1?Cd45.1+) BM. The resulting chimeras were injected intraperitoneally with DTx, and Lyve1hiMHCIIlo IM depletion was then measured 24 hours after injection. We found a significant (>80%) reduction in the Lyve1hiMHCIIlo, but not the Lyve1loMHCIIhi, IM populations in the lung (Fig. 6A), heart, and fat (fig. S7B). No significant depletion of other cell types was observed in the lungs of Lyve1-SLCO→Cd45.1+ chimeras after DTx treatment (Fig. 6B). We then induced lung fibrosis by bleomycin and DTx treatment (fig. S7C) and monitored fibrosis by weight loss (Fig. 6C). Fourteen days after bleomycin treatment, we observed greater weight loss (Fig. 6C), collagen deposition (Fig. 6D and fig. S7D), and inflammatory cell infiltration (Fig. 6E and fig. S7D) in the lungs of mice lacking Lyve1hiMHCIIlo IMs compared with that observed in Lyve1hiMHCIIlo IM–sufficient mice. On day 3 after bleomycin treatment, we observed increased MIP-1α and CCL22 chemokine levels (Fig. 6F), in line with the observed neutrophil and monocyte recruitment in the lungs of mice lacking Lyve1hiMHCIIlo IMs (Fig. 6B). Similar results were obtained using an isoprenaline-induced model of heart fibrosis (43) (fig. S7E) in Lyve1→Cd45.1+ and Lyve1-SLCO→Cd45.1+ chimeric mice, with greater weight loss (fig. S7F), heart weight (fig. S7G), and neutrophil and inflammatory cell infiltration (fig. S7H) in mice lacking Lyve1hiMHCIIlo IMs compared with that observed in Lyve1hiMHCIIlo IM–sufficient mice. Thus, Lyve1hiMHCIIlo macrophages appear to have an early antifibrotic role likely by repressing immune cell infiltration. Interestingly, Lyve1hiMHCIIlo IM macrophages were dramatically reduced in both absolute and relative numbers within the first few days after bleomycin treatment (Fig. 6G). Lyve1hiMHCIIlo IM depletion correlated with enhanced granulocyte and monocyte infiltration into the lungs of bleomycin-treated mice.

Fig. 6 Role of Lyve1hiMHCIIlo macrophages in induced fibrosis.

(A) Flow cytometric measurement of Lyve1hiMHCIIlo IM frequency in lungs from WT Cd45.1+ mice lethally irradiated and reconstructed with Cd45.2+ Lyve1Cre/GFP (Lyve1→Cd45.1+) (left) or Lyve1Cre/GFP-Slco2b1flox/DTR (Lyve1-SLCO→Cd45.1+) (middle) BM and treated intraperitoneally with 1 μg/200 μl DTx 2 months after reconstitution, and mean percentage of Lyve1hiMHCIIlo IMs within all live hematopoietic cells (CD45+) in lung (right). The chimeras were analyzed 16 hours after DTx treatment. (B) Number of cells in lung immune populations in BM chimeras of WT CD45.2 mice reconstituted with Lyve1→Cd45.1+ or Lyve1-SLCO→Cd45.1+ then injected intraperitoneally with 1 μg/200 μl DTx 2 months after reconstitution. Mice were treated with 0.5 μg bleomycin per gram of body weight 16 hours after DTx treatment. Lungs were analyzed by flow cytometry 3 days after treatment. (C to E) BM chimeras Lyve1→Cd45.1+ or Lyve1-SLCO→Cd45.1+ treated with DTx, followed by intranasal bleomycin to induce lung fibrosis. Weight loss was measured for 14 days (C). ELISA measurement of collagen levels in mouse lungs 14 days post-bleomycin treatment is shown in (D). The mean frequency (left) and absolute number (right) of inflammatory SiglecFMerTK+CD64+ cells in the lungs of mice, measured by flow cytometry at 14 days post-bleomycin treatment are shown in (E). (F) Concentration of TARC, MIP-1α, and CCL22 chemokines in lungs of mice in (B), measured by Luminex. (G) WT mice were treated intranasally with bleomycin, and absolute numbers of Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs in lungs were measured by flow cytometry at the indicated time points. (H) Lyve1→Cd45.1+ or Lyve1-SLCO→Cd45.1+ chimeras were treated intraperitoneally with DTx. Vessel permeability was measured 48 hours after treatment, by intravenous injection of 0.9% Evans blue dye solution. (I and J) Lyve1→Cd45.1+ or Lyve1-SLCO→Cd45.1+ BM chimeras treated with DTx followed by intravenous transfer of 5 × 106 Ly6Chi monocytes purified from ubi-GFP mice. Snapshots from time-lapse videos 24 hours after GFP+ monocyte (green) transfer in Lyve1→Cd45.1+ (left; movie S1) or Lyve1-SLCO→Cd45.1+ (right; movie S2) BM chimeras (blood vessels labeled in red) are shown in (I). Scale bars represent 40 μm. Quantifications of track duration (top) and speed (bottom) of GFP+ monocytes 24 hours after transfer into Lyve1→Cd45.1+ (red) or Lyve1-SLCO→Cd45.1+ (blue) BM chimeras are shown in (J). Bars represent means ± SEM. *P ≤ 0.1; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001; and n.s. is not significant, Student’s t test was used for (A), (B), (D) to (F), (H), and (J), and two-way analysis of variance was used for (C). Representative plots of two independent experiments with three to six animals per group are shown.

To investigate how Lyve1hiMHCIIlo IMs influence blood vessel permeability, we treated control and Lyve1-SLCO→Cd45.1+ with intraperitoneal DTx. Vessel permeability was monitored by intravenous Evans blue dye injection 48 hours after injection. We observed more dye in Lyve1hiMHCIIlo IM–depleted mice, suggesting their role in vessel integrity (Fig. 6H). We also used intravital imaging to visualize how Lyve1hiMHCIIlo IM depletion influences cell recruitment. Lyve1→Cd45.1+ and Lyve1-SLCO→Cd45.1+ were treated intranasally with DTx and intravenously transfered with Ly6C+ BM and blood monocytes purified from ubiquitous–green fluorescent protein (ubi-GFP) mice. Lungs were imaged 24 hours after transfer. We observed longer track durations and lower speed (Fig. 6, I and J) in Lyve1hiMHCIIlo IM–depleted mice (movie S2) compared with that in mice sufficient for Lyve1hiMHCIIlo IMs (movie S1). This finding indicates that monocytes are better at infiltrating the lung in the absence of Lyve1hiMHCIIlo IMs. Thus, Lyve1hiMHCIIlo IMs appear to support blood vessel integrity at steady state, and their absence promotes the infiltration of inflammatory cells such as monocytes (tested in Fig. 6I) and likely neutrophils and DCs especially during inflammation.


Here we show that blood monocytes recruited to peripheral tissues differentiate into two distinct RTM populations that come to preferentially reside within different, but conserved, subtissular niches located adjacent to either nerve fibers (Lyve1loMHCIIhi) or blood vessels (Lyve1hiMHCIIlo). Our results extend the heterogeneity of RTM populations and demonstrate that two independent populations of monocyte-derived RTMs (MRTMs) exist across tissues and exhibit conserved niche-dependent functional programming.

In addition to their developmental heterogeneity, a considerable amount of RTM heterogeneity arises from the diversity of tissue environments in which they reside. Several studies have shown that macrophage populations exhibit distinct transcriptional signatures (29, 30) and epigenetic marks (30, 31) that are specific to their tissue of residence. Such findings highlight the key role of tissue factors in imprinting the macrophage transcriptional program and influencing macrophage development, activation, and functional diversity, with consequences for disease and homeostasis. However, heterogeneity within other populations of macrophages, such as IMs has not yet been fully understood. Subpopulations of MHCII versus MHCII+ dermal macrophages (47), as well as dermal perivascular macrophages (48, 49), have already been described. In addition, Gibbings et al. were the first to suggest a level of heterogeneity in lung IMs by identifying three phenotypically distinct IM subpopulations (10). Here we conducted a detailed investigation of the murine lung IM population. Using scRNA-seq, we showed that two, not three, IM lineages coexist within the lung. We demonstrated that the same dichotomy exists in other tissue IM populations and that most Lyve1hiMHCIIlo cells are consistently located in a specific niche surrounding blood vessels. By contrast, most Lyve1loMHCIIhi IMs instead surround nerve bundles or fine nerve fibers. Thus, alongside the major embryonically derived RTM populations already described, other MRTM subsets coexist within specialized subtissular niches that are conserved across tissues.

Importantly, using complementary fate-mapping models, we provide evidence that Lyve1loMHCIIhi and Lyve1hiMHCIIlo IMs are two separate lineages arising from tissue-recruited monocytes, rather than representing points along a developmental or maturation continuum. The molecular cues driving the development of the two IM lineages remain to be identified: It will be interesting to understand whether an intrinsic imprinting programming exists as for DC lineages before tissue entry (50) or if tissue cues encountered after recruitment are the main determinants of MRTM fate.

Our current data highlight the complex and multilayered development of macrophage identity. All macrophages share a common core of expressed markers (F4/80, CD64, and MerTK) and the ability to phagocytose, which defines them as macrophages. Additionally, ontogeny (embryonic versus adult) and tissue of residence add another level of identity (7). Of these, tissue of residence is the major contributor, with the ability to almost completely reprogram ontogenically defined identity (29, 30, 51, 52). Here we add another dimension to macrophage identity: the subtissular niche, which is shared across diverse tissue types, likely as a result of overlapping features common to blood vessels and nerves irrespective of tissue type. In agreement with this hypothesis, our in silico investigation of published datasets uncovered gene signatures consistent with the presence of Lyve1hiMHCIIlo and Lyve1loMHCIIhi MRTMs in fat, heart, bladder, mammary gland, muscle, pancreas, and uterus (2123). Additionally, we detected these populations directly in the skin, brown adipose tissue, diaphragm, ganglia, trachea, tongue, testis, meninges, and bladder. Nevertheless, tissue imprinting remained the primary influence on gene expression patterns in MRTM subpopulations in the lung, fat, heart, and skin. Furthermore, the evidence that the extent of MRTM turnover was tissue specific or even that some populations exhibited very low exchange reinforces this idea. Specially in the heart, Lyve1loMHCIIhi MRTMs appear to have almost no exchange with monocytes in our S100a4Cre-RosaEYFP model, which is in agreement with a recent study showing that such populations mostly self-renew with no contribution of monocytes (53). However, underneath this effect, we uncovered core programming for each subset, likely imprinted by common cues from their conserved cross-tissue niches.

Each IM subpopulation also exhibited distinct functional specializations. Both MRTM subsets can present antigen and stimulate proliferation of CD4+ T cells in vitro. However, despite their higher IL-10 secretion, Lyve1hiMHCIIlo MRTMs were less efficient at inducing Treg differentiation. This effect was likely due to their low MHCII expression, whereas Lyve1loMHCIIhi MRTMs that express higher MHCII combined with relatively potent IL-10 secretion (although lower than Lyve1hiMHCIIlo MRTMs) were more efficient at inducing Tregs (54). Thus, Lyve1loMHCIIhi MRTMs exhibit potent immune-regulatory potential, similar to macrophages from the gut (55) or fat (56), which is complementary to their relatively high IL-10 secretion. In comparison to Lyve1loMHCIIhi MRTMs, Lyve1hiMHCIIlo MRTMs expressed higher levels of genes involved in wound healing, repair, and fibrosis, as well as blood vessel morphology and leukocyte migration. In agreement with these observations, we previously found that arterial macrophages expressing Lyve1 maintain homeostatic vascular tone via the regulation of collagen production (33). Here we extended this functional characterization in murine models of induced lung and heart fibrosis, mimicking life-threatening human diseases with poorly understood pathogenesis. Although monocytes and tissue-resident macrophages are involved in the processes of inflammation and tissue repair during fibrosis (57, 58), a role for MRTMs or MRTM subpopulations has not been investigated in such models. By acutely depleting Lyve1hiMHCIIlo MRTMs during the induction of fibrosis, we have demonstrated their critical role in restraining collagen deposition, immune cell infiltration, and tissue inflammation in both heart and lung fibrosis models. Lyve1hiMHCIIlo MRTMs were implicated in the early stages of fibrosis as early as 3 days after their depletion. In addition, enhanced cell recruitment and chemokine secretion were observed, setting the tone for a dysregulated healing process. Our data support a scenario whereby Lyve1hiMHCIIlo MRTM depletion leads to direct or indirect increases in expression of the CCR4 ligands TARC and CCL22 (59), as well as CCR1, CCR4, and the CCR5 ligand MIP-1α (60), which together induce neutrophil and monocyte infiltration.

Thus, two independent MRTM populations exist that are conserved across tissues and exhibit distinct functional profiles. Their roles in immune regulation and fibrosis render them attractive and separate cellular targets for the therapeutic exploitation of tissue macrophage subsets.

Materials and Methods

Generation of Slco2b1-IRES-LoxP-STOP-LoxP-hDTR gene-targeted mice

A synthetic targeting construct was designed to introduce an IRES-LoxP-STOP-LoxP-hDTR cassette in the 3′ untranslated region (ENSMUSE00000672087) of the Slco2b1 gene, 22 bp downstream of the stop codon. IRES corresponds to an internal ribosomal entry site, the LoxPSTOPLoxP cassette consists of two consecutive polyadenylation signals originating from the bovine growth hormone and SV40 genes that are flanked by loxP sequences, and of a cDNA coding for human diphtheria toxin receptor [hDTR; (61)]. The 5′ and 3′ homology arms of the targeting vector were 3 and 3.3 kb, respectively. The loxP-flanked transcriptional stopper cassette prevents hDTR expression in the absence of Cre activity. The IRES-LoxP-STOP-LoxP-hDTR cassette was abutted to a frt-flanked cassette containing a neomycin-kanamycin resistance (neor) gene that can be expressed under the control of a prokaryotic (gb2) or eukaryotic (Pgk1) promoter. The final targeting construct also contained a cassette coding for the diphtheria toxin fragment A expression cassette. JM8.F6 C57BL/6N ES cells (62) were electroporated with the targeting vector that was linearized with Xho1. After selection in G418, ES cell clones were screened for proper homologous recombination by PCR and Southern blotting. A neomycin-specific probe was used to ensure that adventitious nonhomologous recombination events had not occurred in the selected ES clones. Properly recombined ES cells were injected into BalbC/N blastocysts. Following germline transmission, excision of the frt-neor-frt cassette was achieved via genetic cross with transgenic mice expressing the FLP recombinase under the control of the actin promoter (63). The pair of primers: sense 5′-TCAGCAGACAGTCCTTACCC-3′ and antisense 5′-ACCGTCAACCTCCTGGAATC-3′ amplified a 682-bp band in case of the wild-type Slco2b1 allele, whereas the pairs of primers: sense 5′-CACCCGGGTTACCATGGAGA-3′ and antisense 5′-ACCGTCAACCTCCTGGAATC-3′ amplified a 321-bp band in the case of the Slco2b1-IRES-LoxP-STOP-LoxP-hDTR allele. Here we denote the Slco2b1-IRES-LoxP-STOP-LoxP-hDTR allele as B6-Slco2b1tm1Ciphe and the mice expressing as Slco2b1flox/DTR mice.

Flow cytometry and sorting

Multiparameter analyses of labeled cell suspensions were performed on an LSR II (Becton Dickinson). For intracellular staining, cell surfaces were stained prior to cell fixation and permeabilization using an eBioscence FoxP3 kit. Data were analyzed with FlowJo software (TreeStar). A FACSAria II (Becton Dickinson) was used for flow cytometry. Fluorochrome-conjugated or biotin-conjugated monoclonal antibodies described in the Key Resource Table were used.

For adoptive transfer, single-cell suspensions of BM and blood cells from Cd45.1+ mice were stained with biotin-labeled anti-CD115 antibodies (eBioscience) and positively selected using anti-biotin microbeads and were separated on an AutoMacs (Miltenyi). Monocytes were sorted from the remaining cells by collecting single, live, lineage (CD3, CD19, CD49b, Ly6G)CD11b+CD115+Ly6Chi cells. Each LyzMCre-Slco2b1flox/DTR Cd45.2+ recipient mouse received the indicated number of Ly6Chi monocytes intravenously 24 hours after intranasal treatment with the indicated amount of DTx. Recipient mice were analyzed 9 days after transfer or at the indicated time point.

Scanning electron microscopy

For imaging by scanning electron microscopy, sorted cells were fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer for 1 hour (pH 7.4) at room temperature, treated postfixation with 1% osmium tetroxide (Ted Pella Inc.) at room temperature for 1 hour, and then dehydrated through a graded ethanol series from 25 to 100% and critical point dried using a CPD 030 critical point dryer (Bal-Tec AG, Liechtenstein). The cell surface was coated with 15 nm of gold by sputter coating using a SCD005 high-vacuum sputter coater (Bal-Tec AG). The coated samples were examined under a field emission JSM-6701F Scanning Electron Microscope (JEOL Ltd., United States) at an acceleration voltage of 8 kV using the in-lens secondary electron detector.

Single-cell RNA-sequencing

For SMARTseq v2, single lung IMs (LinCD45+MerTK+CD64+SiglecFCD11b+) were collected by FACS into 96-well plates. Single-cell cDNA libraries were generated using the SMARTSeq v2 protocol (11) with the following modifications: (i) use of 10 μM TSO and (ii) use of 250 pg of cDNA with 1/5 reaction of an Illumina Nextera XT kit (Illumina, San Diego, CA, USA). The length distribution of the cDNA libraries was monitored using a DNA High Sensitivity Reagent Kit on the Perkin Elmer Labchip (Perkin Elmer, Waltham, MA, USA). All samples were subjected to an indexed paired-end sequencing run of 2 × 51 cycles on an Illumina HiSeq 2000 system (Illumina, San Diego, CA, USA) (188 samples/lane).

Total bulk RNA-sequencing

Cells collected by FACS from several animals were pooled. Total RNA was extracted using Arcturus PicoPure RNA Isolation kit (Applied Biosystems Thermo Fisher Scientific) according to manufacturer’s protocol. All mouse RNAs were analyzed on an Agilent Bioanalyser for quality assessment with a median RNA Integrity Number (RIN) of 8.3. cDNA libraries were prepared using 2 ng total of RNA and 1 μl of a 1:50,000 dilution of ERCC RNA Spike in Controls (Ambion Thermo Fisher Scientific) using the SMARTSeq v2 protocol (64) with the following modifications: (i) addition of 20 μM TSO and (ii) use of 250 pg of cDNA with 1/5 reaction of an Illumina Nextera XT kit (Illumina, San Diego, CA, USA). The length distribution of the cDNA libraries was monitored using a DNA High Sensitivity Reagent Kit on the Perkin Elmer Labchip (Perkin Elmer, Waltham, MA, USA). Six samples were subjected to an indexed PE sequencing run of 2 × 51 cycles on an Illumina HiSeq 2500 Rapid mode (14 samples/lane), 16 samples were subjected to an indexed PE sequencing run of 2 × 51 cycles on an Illumina HiSeq 2000 (Illumina, San Diego, CA, USA) (16-20 samples/lane), and nine samples were subjected to an indexed paired-end sequencing run of 2 × 151 cycles on an Illumina HiSeq 4000 system (Illumina) (24 samples/lane).


Cryosections (20 μm) from C57BL/6 and Cx3cr1GFP/+ mouse lung were prepared as previously described (65). Briefly, tissue was fixed in 2% paraformaldehyde (PFA) with 30% sucrose overnight at 4°C and frozen in Tissue-Tek Optimum Cutting Temperature compound for 20-μm cryosectioning. The primary antibodies used were anti-LYVE-1 (rabbit polyclonal; Abcam), anti-CD68 (rat clone FA-11; Serotec), anti-CD31 (armenian hamster clone 2H8; Millipore), and anti-GFP (chicken polyclonal; Aves Lab). Fluorescence conjugated antibodies from Jackson Immunoresearch were used to reveal the staining. Cell nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) and sections were mounted for analysis.

Whole-mount staining on ear skin and heart was performed as described previously (66). Briefly C57BL/6 and Cx3cr1GFP/+ mice were perfused with 2% PFA, and the tissue was removed and placed into 2% paraformaldehyde (PFA) fixative overnight at 4°C. Tissues were rinsed in PBS and ears were split into dorsal and ventral halves. The samples were blocked with PBS containing 0.5% bovine serum albumin and 0.3% Triton X-100 overnight at 4°C, followed by incubation at 4°C with primary antibodies against LYVE-1, CD31, CD68, and/or anti-GFP for 3 days. Finally, tissues were washed and stained for 2 days with fluorescent (AF488-, AF647-, and Cy3-conjugated) secondary antibodies (Jackson Immunoresearch). Specimens were visualized under a fluorescence widefield microscope (Axio Imager.Z1, AxioCam HRM camera 426511-9901-000; Zeiss EC Plan-NEOFLUAR 20x/0,5 420350-9900; Carl Zeiss Micro Imaging, Inc., Jena, Germany) or confocal microscope with photomultiplier tubes to detect fluorescence signals from samples (Leica TCS SP5; Leica 506251 HCX PL APO 40x/1.25-0.75 Oil CS; Leica 506192 HCX PL APO 63x/1.40-0.60 Oil CS; Leica Microsystems, Inc., Deerfield, IL) using LAS AF confocal software (version 1.8.2; Leica Microsystems, Inc).

Adult Cdh5CreER-RosatdTomato mice were treated with tamoxifen (5 mg in corn oil, 200-μl gavage) to induce tdTomato expression in all endothelial cells. One week later, mice were euthanized and perfused with PBS. Wnt1Cre-RosatdTomato-Cx3cr1GFP/+ were euthanized without tamoxifen treatment. In both cases, the indicated organs were fixed for 2 hours in Antigen Fix (Diapath), washed in Tris 0.1 M phosphate buffer, embedded in 4% agarose, and sectioned using a vibratome. Slices (250 μm) were stained for 24 hours for the indicated markers, cleared, and imaged in Rapiclear 1.47 (Sunjin Lab). Images were acquired with a Zeiss LSM880 confocal microscope (Zeiss Plan-Apochromat 20X/0.8 DIC II lens) and processed with Imaris software (Bitplane).

Spinning-disk confocal intravital imaging of lung

Lung imaging was performed as previously described (67). In brief, mice were anesthetized with a cocktail of 150 mg/kg ketamine and 10 mg/kg xylazine (i.p.) and placed on a electric heat pad to maintain body temperature at 37°C. Tracheostomy and catheterization of the jugular vein were performed surgically before the mice were connected to a mechanical ventilator (Mini Vent; HSE) and the lung was exposed. To immobilize the lung, a customized vacuum window was positioned over the lung surface to apply a negative pressure (40 mm Hg). The lung microvasculature was visualized using a LaVision TriM Scope II microscope (LaVision BioTec), with a water dipping objective (20× magnification, 1.0 NA, 2 mm WD; XLUMPLFLN20×W, Olympus) and a Chameleon-pulsed infrared laser (titanium sapphire; Coherent). The acquisitions of cellular activity in the lung microvasculature were performed using a 990-nm laser for excitation. A back-thinned electron-multiplying charge-coupled device camera (512 × 512 pixels; C9100-13; Hamamatsu) was used for fluorescence detection. Evan’s blue (50 μg in 50 μl sterile PBS) was injected into the mice (i.v.) to label the lung vasculature. Five different fields were recorded for 15 min and then analyzed to determine the total number of monocytes per field. Monocytes were defined as adherent if they remained stationary in the field for the entire imaging period. Data analysis was performed using Imaris software.

Statistical Analyses

All statistical analyses were performed with Prism 5.0 (GraphPad Software). All P values are two tailed.

Supplementary Materials

Material and Methods

Figs. S1 to S7

Tables S1 to S5

Movies S1 and S2

References and Notes

Acknowledgments: We would like to thank L. Robinson of Insight Editing London for editing the manuscript. We thank E. Koh from the Advanced Imaging core facility (Life Science Institute, Immunology Programme, NUS) for sharing expertise. Funding: The CyTOF, bioinformatics, and immunogenomics platforms are part of the SIgN Immunomonitoring platform (supported by a BMRC IAF 311006 grant and BMRC transition funds #H16/99/b0/011). This work was supported by EMBO YIP, Singapore Immunology Network core funding, Agency for Science, Technology and Research (A*STAR), and Singapore NRF Senior Investigatorship (NRFI2017-02) to F.G.; AXA Research Fund (F.G. and B.Mali.); and NMRC (CBRGnov094) and NRF grants to V.A. A.S. is funded by an Emmy Noether fellowship (SCHL 2116/1-1) of the German Research Foundation and a Young Investigator Award of the Biomedical Research Council Singapore. Author contributions: Conception: F.G. and S.C.; discussion: F.G., S.C., H.Y.L., A.S., T.M., B.Mali., J.C., M.P., K.K., M.B., L.G.N., and V.A.; research design and experimentation: S.C., H.Y.L., S.Y.L., L.T., P.S., J.L., S.F., S.N., W.T.K., R.G., A.B., C.B., and B.Mali.; data analysis: S.C., H.Y.L., S.Y.L., X.-M.Z., D.K., J.C., M.P., and F.G.; resource assistance and clinical samples: J.K.C.T., S.B., M.S., S.-A.E.S.T., A.L., T.M., B.Mali., and V.A.; writing of draft and editing: S.C. and F.G.; and project administration: F.G. Competing interests: The authors declare no competing interests. Data and materials availability: All sequencing data sets are deposited in the Genome Expression Omnibus under accession number GSE125691. All other data supporting the findings of this study are available within the paper or in the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article