Efferocytosis fuels malignant pleural effusion through TIMP1

The clearance of apoptotic cells by Mφs favors malignant pleural effusion progression in a murine cancer model.

Abstract

Malignant pleural effusion (MPE) results from the capacity of several human cancers to metastasize to the pleural cavity. No effective treatments are currently available, reflecting our insufficient understanding of the basic mechanisms leading to MPE progression. Here, we found that efferocytosis through the receptor tyrosine kinases AXL and MERTK led to the production of interleukin-10 (IL-10) by four distinct pleural cavity macrophage (Mφ) subpopulations characterized by different metabolic states and cell chemotaxis properties. In turn, IL-10 acts on dendritic cells (DCs) inducing the production of tissue inhibitor of metalloproteinases 1 (TIMP1). Genetic ablation of Axl and Mertk in Mφs or IL-10 receptor in DCs or Timp1 substantially reduced MPE progression. Our results delineate an inflammatory cascade—from the clearance of apoptotic cells by Mφs, to production of IL-10, to induction of TIMP1 in DCs—that facilitates MPE progression. This inflammatory cascade offers a series of therapeutic targets for MPE.

INTRODUCTION

Malignant pleural effusion (MPE), often the result of several tumors metastasizing in the pleural cavity (1), is a devastating complication in patients with cancer. There is currently no effective treatment for this lethal complication (1).

MPE is characterized by an accumulation of exudative effusion into the pleural space (2), leading to the compression of the lung, which can ultimately end in suffocation. Angiogenesis and vascular remodeling are essential pathophysiological steps in MPE formation (3). Moreover, it is now recognized that the immune response within MPE affects both the angiogenesis and vascular remodeling process. However, the cascade of cellular and molecular interactions promoting MPE progression still remains to be elucidated.

Several human and mouse studies report an association between Mφ infiltration and MPE development (4, 5). Mφs found in MPE secrete transforming growth factor–β (TGF-β) that favors Foxp3⁺ (forkhead box P3) regulatory T cells (Foxp3⁺ T_regs) and reduces T cell cytotoxicity (6), possibly leading to tumor progression in MPE (7). Moreover, Mφs characterized by high mRNA levels of Il10 have been proposed as a diagnostic biomarker for MPE (8). Interleukin-10 (IL-10) is released by Mφs upon phagocytosis of apoptotic cells (9, 10).

Phagocytosis of apoptotic cells by Mφs is mediated by several phagocytic receptors. Among them, we and others have described the critical role of receptor tyrosine kinases AXL and MERTK (Mer proto-oncogene tyrosine kinase) in the regulation of Mφ anti-inflammatory functions (11, 12). Immunosuppression as a result of phagocytosis renders the clearance of dying cells an active mechanism by which Mφs negatively regulate the inflammatory response (9, 10). However, little has been done so far to define whether the generation of an anti-inflammatory environment as a result of the clearance of dying cells by Mφs can lead to MPE progression.

Moreover, because of technical limitations, it has been so far assumed that pleural Mφs are homogeneous population of cells. However, other studies investigating Mφs at the resolution of a single cell have shown their large functional diversity in other tissues, including the lung (13, 14). Therefore, investigating their potential functional heterogeneity might allow a deeper understanding of pleural Mφ function and help to better examine their contribution to the progression of MPE.

Similar to Mφs, dendritic cells (DCs) are also present in exudative pleural effusions (15). A meta-analysis evaluating the significance of immune cells in non–small cell lung cancer (NSCLC) showed that increased tumor and stromal conventional DCs (cDCs) were associated with improved survival. These results show that DCs are present in exudative pleural effusions and may be involved in the development of cell-mediated immune reactions in the pleural space. However, whether DCs play a key role in MPE progression establishing a functional network with Mφs has not been tested so far. Consequently, potential pathways mediating this connection and thus affecting MPE are unknown.

In this study, we used a mouse model to investigate the role of phagocytosis, Mφs, and DCs in MPE progression. Our results delineate an inflammatory cascade consisting of four distinct pleural Mφ populations that favor fluid accumulation by initially phagocytosing the apoptotic cells and subsequently secreting IL-10; IL-10 promotes DCs to express tissue inhibitor of metalloproteinase 1 (Timp1), which lastly mediates the development of MPE.

RESULTS

Phagocytosis of apoptotic cells by Mφs promotes MPE

The injection of the MPE-competent Lewis lung carcinoma (LLC) cell line into the pleural cavity of C57BL/6 mice (Fig. 1A) promoted MPE formation. In parallel to the increased infiltration of immune and stromal cells during MPE development (fig. S1A), we observed accumulation of early- and late-stage apoptotic cells (Fig. 1B), which were enriched in the CD45⁺ cell fraction (Fig. 1C). When Mφs, which are known to accumulate in MPE (16), ingest apoptotic cells, they are educated to acquire an anti- inflammatory phenotype, often associated with a protumorigenic function (17). However, whether clearance of apoptotic cells by Mφs, rather than a direct mechanism of cancer cell immune evasion, controls MPE progression has not yet been tested.

Fig. 1. Phagocytosis of apoptotic cells by Mφs promotes MPE.

(A) Schematic overview of the MPE model. MPE was induced via intrapleural injection of LLC cells in wild-type (WT) mice. (B) Percentage of apoptotic cells in peripheral blood (PB) (n = 5) and MPE (n = 5) was assessed by flow cytometry via annexin V/propidium iodide (PI) staining. Pooled data (left) and representative dot plot (right) showing the frequency of early (bottom right quadrant) or late (top right quadrant) apoptotic cells. (C) Frequency of CD45⁺ and CD45⁻ cells within annexin V⁺ apoptotic cells. (D) Expression of MERTK receptor on CD11b⁺F4/80⁺ (Mφs) and CD3⁺ (T cells) immune cells in MPE (n = 5). Left: Dot plots showing the delta mean fluorescence intensities (ΔMFI) of MERTK receptor on Mφs and T cells relative to that of FMO (fluorescence minus one) controls from murine MPE. Right: Flow cytometric histogram graphs showing a representative example from CD11b⁺F4/80⁺ (Mφs) expressing MERTK receptor (black open histogram) compared to FMO control (filled gray histogram). (E) Representative images (left) of MPEs (dashed lines) from Csf1rCre⁻Axl^f/fMertk^f/f (n = 21) and Csf1rCre⁺Axl^f/fMertk^f/f (n = 19) mice through the diaphragm and respective MPE quantification (right). (F) IL-10 concentration in MPEs isolated from Csf1rCre⁻Axl^f/fMertk^f/f (n = 19) and Csf1rCre⁺Axl^f/fMertk^f/f (n = 18) mice was analyzed by enzyme-linked immunosorbent assay (ELISA). (B and D) Data of one experiment. (C) Data of two experiments. (E and F) Data of three experiments. Each symbol represents a measurement from a single mouse. Data are means ± SEM. *P < 0.05 and **P < 0.01. Statistical analysis was performed using Wilcoxon signed-rank test (B) or Mann-Whitney U test (E and F). Photo credit: L. Zhao (University Medical Center Hamburg-Eppendorf).

Here, we observed that MERTK is expressed by Mφs in the MPE through fluorescence-activated cell sorting (FACS) (Fig. 1D) and quantitative polymerase chain reaction (qPCR) analysis (fig. S1B). Since we found trace levels of AXL protein and mRNA (fig. S1, B and C), we decided to use Axl and Mertk double-deficient mice to test the role of phagocytosis in MPE development. In particular, we injected LLC cells in previously characterized Csf1rCre⁺Axl^f/fMertk^f/f mice (12), a transgenic mouse model in which genetic ablation of Axl and Mertk in Mφs impairs their capacity to phagocytose apoptotic cells. Although pleural tumor growth was not affected (fig. S1D), we observed a significant reduction of MPE progression in Csf1rCre⁺Axl^f/fMertk^f/f mice compared to Csf1rCre⁻Axl^f/fMertk^f/f mice (Fig. 1E). Notably, we confirmed these data by comparing Csf1rCre⁺Axl^f/fMertk^f/f with littermate controls Csf1rCre⁺Axl^wt/wtMertk^wt/wt (fig. S1E).

Last, phagocytosis of apoptotic cells has been linked to IL-10 release (18). Similarly, IL-10 has been suggested as a critical factor accounting for MPE formation (19, 20). We therefore analyzed the concentration of IL-10 in the pleural effusions of the two groups of mice described above. A lower IL-10 concentration was observed in the pleural effusion of Csf1rCre⁺Axl^f/fMertk^f/f mice that also had a reduced MPE volume compared to littermate controls (Fig. 1F).

The concentration of IL-10 is reported to be higher in human pleural effusion than in peripheral blood (19, 20). However, little is known regarding whether IL-10 levels are higher in human MPE compared to benign pleural effusion (BPE). Therefore, we measured IL-10 concentrations in parapneumonic pleural effusion (PPE) of 15 patients, in transudative pleural effusion (TrPE) of 17 patients, in lung cancer–induced MPE of 19 patients, and in other cancer–induced MPE of 10 patients. The concentration of IL-10 in MPE was higher than in all the other test samples (fig. S2A and table S1), confirming increased levels of IL-10 in patients undergoing MPE development.

Thus, we found that mice with an impairment in Mφ-mediated phagocytosis have a reduced MPE formation, and this is associated with less IL-10 accumulation in the pleura. We also extended the observation regarding IL-10 to MPE patients; however, whether IL-10 is causally linked to MPE remains unclear.

Mφs producing IL-10 promote MPE

To clarify whether IL-10 is also causally linked to MPE, we first confirmed that IL-10^−/− mice show reduced MPE progression (fig. S3, A to F) (21). However, the major cellular source of IL-10 remained to be revealed. Many immune cells have the ability to produce IL-10, including T cells and Mφs (22). Using the IL-10^eGFP reporter mice (tiger mice) (Fig. 2A) (23), we first found that IL-10^eGFP expression was restricted to the CD45⁺ cells in MPE (Fig. 2B). Then, we performed a multiparameter FACS analysis on CD45⁺IL-10^eGFP+ cells and subsequently simplified its complexity using viSNE (visual interactive Stochastic Neighbor Embedding) . This unsupervised analysis showed that the largest fraction of IL-10^eGFP–producing cells were CD11b⁺LY6C⁻LY6G⁻F4/80⁺MERTK⁺ cells [AXL was excluded from the viSNE analysis due to low detection (fig. S1, B and C)]. A smaller source of IL-10^eGFP was the FOXP3⁺CD4⁺ T cells (Fig. 2C). We also quantified this analysis and confirmed that the large majority of IL-10^eGFP+ was CD11b⁺LY6C⁻LY6G⁻ cells (Fig. 2D). We further validated these results using a different reporter mouse—10BiT mouse—and intracellular staining for IL-10, indicating that the major source of IL-10 is CD11b⁺ cells (fig. S4A). Next, we analyzed IL-10^eGFP expression on MERTK^high and MERTK^low Mφ populations isolated from MPE and observed that MERTK^high Mφs express higher IL-10^eGFP than MERTK^low Mφs (Fig. 2E).

Fig. 2. Mφs producing IL-10 promote MPE.

(A) MPE was induced in Il10^eGFP reporter mice. (B) Left: Gating strategy for the representative FACS plot of IL-10^eGFP expression. Right: Percentages of CD45⁺ IL-10^eGFP+ and CD45⁻ IL-10^eGFP+ within alive cells from MPE (n = 9). (C) viSNE of IL-10^eGFP–producing CD45⁺ immune cells. Clustering is based on MFI of IL-10^eGFP, CD3, FOXP3, CD11b, LY6G, LY6C, F4/80, and MERTK. Black circles indicate the IL-10^eGFP–rich region. (D) Source of IL-10^eGFP in CD45⁺ immune cells (left) and CD11b⁺ myeloid cells from MPE (right). (E) IL-10^eGFP expression of CD45⁺ CD11b⁺ F4/80⁺ MERTK^high (n = 7) and CD45⁺ CD11b⁺ F4/80⁺ MERTK^low (n = 7) cells in MPE. IL-10^eGFP histogram (left) and quantitative MFI graphs (right). (F) MPE was induced in LysMCre⁺-Il10^f/f mice and Foxp3Cre⁺-Il10^f/f mice. (G) Representative images of MPEs (left) and respective MPE quantification (right) in LysMCre⁺-Il10^wt/wt (n = 31) and LysMCre⁺-Il10^f/f (n = 20) mice. (H) Correlation between IL-10 concentration and MPE volume in LysMCre⁺-Il10^wt/wt mice (n = 12). Top-right text (r and r²) represents Pearson’s correlation and its coefficient of determination. (I) Representative images of MPEs (left) and respective MPE quantification (right) in Foxp3Cre⁺-Il10^wt/wt (n = 8) and Foxp3Cre⁺-Il10^f/f (n = 14) mice. (J) Intracellular IL-10 staining of CD45⁺ CD11b⁺ F4/80⁺ cells in MPEs isolated from Csf1rCre⁻Axl^f/fMertk^f/f (n = 5) and Csf1rCre⁺Axl^f/fMertk^f/f (n = 5) mice. IL-10 histogram (left) and graph reporting MFI (right). Data of one (B, D, E, H, and J), two (I), or three (G) experiments, respectively, are shown. Each symbol represents a measurement from a single mouse. Data are means ± SEM. ns, nonsignificant. *P < 0.05, **P < 0.01, and ***P < 0.001. Statistical analysis was performed using Mann-Whitney U test (B, G, I, and J) or paired t test (E). Photo credit: L. Zhao (University Medical Center Hamburg-Eppendorf).

Next, we generated LysMCre⁺Il10^f/f mice in which IL-10 expression is specifically depleted in Mφs. These mice exhibited reduced MPE accumulation, but no differences in tumor burden, compared with littermate controls upon LLC injection (Fig. 2, F and G, and fig. S4B). Furthermore, we detected less IL-10 in the MPE of LysMCre⁺Il10^f/f mice compared with littermate controls by enzyme-linked immunosorbent assay (ELISA) (fig. S4C). When we performed a Pearson’s correlation, we observed that a statistically significant linear relationship exists between IL-10 and MPE volume (Fig. 2H). In addition, according to our characterization, Foxp3⁺ T_regs are another source of IL-10; therefore, we tested their relevance by inducing MPE in Foxp3Cre⁺Il10^f/f mice. No significant difference in MPE volume or in tumor weight was observed (Fig. 2, F and I, and fig. S4D). Together, our results indicate that myeloid cells, and particularly a population of CD11b⁺LY6C⁻LY6G⁻ Mφs present in the pleural cavity during MPE progression, are the major sources of IL-10, which is a critical cytokine for pleural effusion accumulation.

Next, we wondered what promotes the expression of IL-10 in Mφs during MPE development. Considering the known role of IL-10 upon sensing of apoptotic cells (9, 24), we evaluated the expression of IL-10 in Mφs in which phagocytosis is impaired (Csf1rCre⁺-Axl^f/fMertk^f/f mice) during MPE development. We observed that IL-10 expression was significantly reduced in CD11b⁺F4/80⁺ cells isolated from Csf1rCre⁺-Axl^f/fMertk^f/f mice compared to cells isolated from littermate controls (Fig. 2J and fig. S4E). Last, to further confirm the increased IL-10 secretion upon apoptotic cell clearance (9), we cocultured Mφs with apoptotic neutrophils (aNs) before lipopolysaccharide (LPS) stimulation. As previously described (24) and in our experimental setting, phagocytosis of apoptotic cells before Toll-like receptor (TLR) engagement led to increased IL-10 secretion (fig. S4F).

Together, these findings show that Mφs are the major relevant source of IL-10. Moreover, during MPE development, Mφs produce IL-10 upon phagocytosis of apoptotic cells in an AXL- and MERTK-dependent manner.

The transcriptional landscape of Mφs in MPE

Our data show that CD11b⁺ Mφs are the key regulators of MPE development. To investigate CD11b⁺ Mφ heterogeneity and achieve an unsupervised characterization of these cells, we performed single-cell RNA sequencing (scRNA-seq). We isolated myeloid cells from MPE using a broad myeloid marker, such as CD11b, to include not only Mφs but also other myeloid cell types and potentially other cell populations to be used as reference in our analysis. The t-distributed stochastic neighbor embedding (t-SNE) analysis clustered 1989 cells into nine discrete cell populations (Fig. 3A). Gene expression patterns of established canonical markers of immune cells allowed us to assign putative biological identities to each cluster (Fig. 3, A and B). Enrichment of typical Mφ-associated genes such as Adgre1, Csf1r, Cd68, Fcgr1, F13a1, and Lyz2 was observed (fig. S5A). Despite the fact that all Mφs express Il10 (Fig. 3C and fig. S5B) and Mertk (Fig. 3, D and E), they still represent a transcriptionally heterogeneous population that can be subdivided into four distinct clusters (Mφ_1, Mφ_2, Mφ_3, and Mφ_4 populations).

Fig. 3. Identification of MPE-associated myeloid cell populations.

(A) t-SNE representation of aligned gene expression data in single cells (n = 1991) extracted from murine MPE. Cell identities were assigned on the basis of the expression of canonical markers. (B) Heatmap showing the statistically up-regulated genes (ordered by decreasing P value) in each cluster defined in (A) and selected enriched genes used for biological identification of each cluster (scale, log₂ fold change). (C) Il10 mRNA expression level and percentage in each cluster. (D) Mertk mRNA expression level and percentage in each cluster. (E) Top: t-SNE representation of four Mφ clusters isolated from murine MPE. Bottom: t-SNE plot of Mertk gene expression (red dot, detected Mertk gene expression; yellow dot, undetected Mertk gene expression).

Gene ontology (GO) term analysis was therefore performed among these four different Mφ populations, to thoroughly investigate the heterogeneity of MPE-associated Il10 expressing Mφs (Fig. 4, A to D). The Mφ_1 population contained multiple transcripts related to aerobic glycolysis, such as Tpi1, Pgk1 (25), Pkm, and Ldha (26). These genes code for glycolytic enzymes involved in the modulation of tumor angiogenesis and confer chemotherapy resistant features (27). For example, phosphoglycerate kinase 1 is an enzyme responsible for the first adenosine 5′-triphosphate (ATP)–generating step in the glycolysis pathway as it catalyzes the reversible conversion of 1,3-bisphosphoglycerate and adenosine 5′-diphosphate (ADP) to 3-phosphoglycerate and ATP, respectively (28). This underlines the high metabolic activity of the Mφ_1, a feature usually associated with cancer metabolism. The Mφ_2 population was enriched for genes such as Ppbp (29) (also known as Cxcl7) and Ccl2, both molecules are critical for neutrophil activation, chemotaxis, and migration (30).

Fig. 4. Functional enrichment analysis of Mφs in MPE.

(A to D) Top 10 terms identified by GO enrichment analyses for Mφ_1, Mφ_2, Mφ_3, and Mφ_4 populations. Adjusted P value for each annotation is represented by color scale. Gene ratio is represented by dot size. Enriched terms were identified as significant at an adjusted P ≤ 0.01 and a false discovery rate of ≤0.05. (E) UpSet plot showing the overlap of differentially expressed genes identified in different monocyte/Mφ clusters (Mφ_1, Mφ_2, Mφ_3, and Mφ_4 populations). (F) Top 24 distinct differentially expressed genes identified in (E). Red labeling indicates characteristic cluster genes. (G) Phagocytic capacity of Mφs isolated from pooled MPE of 17 mice. Left: FACS plots showing the four subclusters of sorted MPE Mφs (CD11b⁺) that bind (4°C) or bind and uptake (37°C) CellTracker-labeled apoptotic thymocytes (aTs) upon 1 hour of coculture. Right: Phagocytic index reporting the difference between percentage of Mφs binding and taking up apoptotic cells (37°C) and percentage of Mφs binding apoptotic cells (4°C).

The Mφ_3 population, similar to the Mφ_2 population, was also enriched for genes associated with migration, such as Pf4 (31) (a heparin-binding protein with chemotactic activity for neutrophils and monocytes) and Trem2 (32) (reported to regulate Mφ chemotaxis and phagocytosis). These genes suggest a promigratory gene signature of the Mφ_3 population in unresolved inflammation such as MPE. In addition to leukocyte migration, the Mφ_3 population also expressed genes such as Hmox1, Arg1, and Gpx1, which are related to epithelial proliferation (33), suggesting that these cells might be involved in tissue remodeling, especially of the vascular system. Last, the Mφ_4 population was characterized by genes involved in cell migration and cholesterol biosynthesis, including Fdps, Hmgcs1, Cyp51, Gch1, and Sqle (34), which were reported to be essential for cancer cell propagation.

Furthermore, to identify unique features that distinguish these different Mφ clusters from each other, we further analyzed the genes that are specifically expressed by each cluster via UpSetR package. We identified 24 genes selectively expressed by the Mφ_1, 61 genes by the Mφ_2, 98 genes by the Mφ_3, and 143 genes by the Mφ_4 (Fig. 4, E and F).

This analysis led us to identify the highly metabolically active cells of the Mφ_1 population as selectively enriched for Ccr2 [which encodes the cognate receptor of chemokine ligand 2 (CCL2)] (35). Ccr2-deficient mice have been described to be protected from MPE, suggesting that tumor cells drive MPE development via systemic CCL2 signaling to CCR2⁺ host cells (36).

The cells from the Mφ_2 population, which showed chemotactic ability, selectively expressed the proangiogenic cytokine Cxcl3, which was reported in BPE. In parallel, the Mφ_2 population expressed the highest levels of Il4Ra, thus suggesting a high capability to respond to IL-4. In line with this, the cells of the Mφ_2 population also expressed high levels of Msr1, scavenger receptor A, known not only to be induced upon IL-4 stimulation but also to be associated with tumor promotion and metastasis in ovarian and pancreatic cancer in mice (37).

Notably, the “promigratory” cells of the Mφ_3 population were enriched in molecules critical for phagocytosis such as Cd36, a phagocytic receptor with a key role in the uptake of apoptotic cells and oxidized lipids, as well as Trem2 and Lpl, genes associated with phagocytosis and lipid metabolism. In addition to this, the cells of the Mφ_3 population also selectively produced Spp1-encoding osteopontin, a key mediator of vascular permeability that leads to MPE accumulation (38).

The chemotactic cells of the Mφ_4 population were selectively characterized by the expression of Ly6c2, a feature strongly supporting the monocyte identity of these cells. Moreover, this cell population selectively expressed Cd44, a receptor on monocyte-derived Mφs, which augments its phagocytosis of aNs and contributes to the migration of activated leukocytes to sites of inflammation (39), and Tgm2, a member of the transglutaminase family of enzymes, which, in conjunction with CD14, plays a key role in phagocytosis of apoptotic cells (40).

To substantiate the scRNA seq characterization, we tested the phagocytic capacity of the four subclusters of Mφs identified via scRNA seq. First, we selected surface markers that define the four different Mφ subclusters enriched in MPE (Ccr2 for Mφ_1, Il4ra for Mφ_2, Cd36 for Mφ_3, and Ly6c2 for Mφ_4; Fig. 4F). Second, on the basis of these markers, we sorted the different populations (fig. S5C), cultured them in vitro, and analyzed their phagocytic capacity upon exposure to labeled apoptotic thymocytes (aTs). As expected, most likely because of the expression of MERTK in all the four Mφ subclusters analyzed (Fig. 3, D and E), Mφs all showed binding and uptake capability in vitro with Mφ_1 and Mφ_3 being the most phagocytic cell subpopulations. Notably, Mφ_3, which was also enriched for the receptors CD36, and Trem2, both regulating Mφ clearance activity, showed the highest phagocytic capacity. Both Timp2 (characteristic of Mφ_3) and Cd36 have been shown to be up-regulated in Mφs upon coculture with apoptotic tumor cells (Fig. 4G) (41).

Together, these data illustrate that although IL-10 expression resulting from Mφ phagocytic activity can be considered a key characteristic of Mφs present in MPE, the analysis of a combination of different functions is a more reliable approach to fully comprehending the Mφ polarization status. In particular, aerobic glycolysis, neutrophil activation, cell migration, and cholesterol biosynthesis are the distinct features typical of the IL-10–producing pleural Mφ.

The target cells of IL-10 in MPE are the DCs

To identify the cellular target of IL-10 during MPE development, we examined the expression of IL-10 receptors in several types of cells that are known to play a key role in MPE progression. First, we evaluated Il10ra (IL10R1) and Il10rb (IL10R2) gene expression in LLC and MC38 cells. No Il10ra expression was detected, while Il10rb was ubiquitously expressed in both cancer cell lines (fig. S6A). We also checked the expression of IL-10RA on immune cells by flow cytometry. First, as a proof-of-concept experiment and to avoid the potential bias of an inflammatory environment, we analyzed the expression of IL-10RA on the surface of CD45⁺ and CD31⁺ cells isolated from the lungs of mice under steady-state conditions by flow cytometry. The mean fluorescence intensity (MFI) of IL-10RA was significantly higher in CD45⁺ cells than in CD31⁺ cells, indicating that immune and not endothelial cells are more likely the target cells of IL-10 (fig. S6B). To test their responsiveness to IL-10 in vitro, we isolated CD45⁺ immune cells and CD31⁺ endothelial cells from mice without MPE and measured signal transducers and activators of transcription 3 phosphorylation (pSTAT3) in these cells upon exposure to IL-10. IL-6 was used as a positive control. The data showed that among CD45⁺ cells, CD3⁻ cells exhibited the highest STAT3 phosphorylation, while CD31⁺ endothelial cells did not respond to either IL-6 or IL-10 (Fig. 5A and fig. S6C). Briefly, these data showed that in our experimental setting, CD45⁺CD3⁻ cells were the cells with the highest sensitivity to IL-10.

Fig. 5. The target cells of IL-10 in MPE are the DCs.

(A) pSTAT3 expression in CD31⁺, CD45⁺, CD45⁺CD3⁺, CD45⁺CD4⁺, CD45⁺CD8⁺, and CD45⁺CD3⁻ cells cultured in the presence of either IL-6 (100 ng/ml) or IL-10 (100 ng/ml), as shown by the histogram of relative expression of pSTAT3 MFI (IL-10 versus IL-6). (B) Graphical representation of cell-specific mouse models of Il10ra deletion or impairment. (C to G) CD11cCre⁺Il10ra^wt/wt (n = 19) and CD11cCre⁺Il10ra^f/f (n = 17) mice (C), LysMCre⁺Il10ra^wt/wt (n = 7) and LysMCre⁺Il10ra^f/f (n = 10) (D), Cdh5Cre⁺Il10ra^wt/wt (n = 17) and Cdh5Cre⁺Il10ra^f/f (n = 11) (E), WT (n = 7) and DN Il10Rα (n = 7) (F), and Foxp3Cre⁺Il10ra^wt/wt (n = 11) and Foxp3Cre⁺Il10ra^f/f (n = 11) (G) mice were intrapleurally injected with LLC cancer cells. Representative images of MPEs (dashed lines) (left) through the diaphragm and respective MPE quantification (right). Each symbol represents a measurement from a single mouse. Data are means ± SEM. Results were representative of three independent experiments. *P < 0.05 and **P < 0.01. Statistical analysis was performed using Mann-Whitney U test (C to G). Photo credit: L. Zhao (University Medical Center Hamburg-Eppendorf).

Next, we used a variety of conditional knockout (KO) mouse lines to identify the cell population responding to IL-10, which is critical for MPE development. To dissect the contribution of various CD45⁺CD3⁻ cell populations as targets of Mφ-derived IL-10 to MPE progression, we crossed LysMCre (granulocyte- and Mφ-specific) and CD11cCre (DC-specific) mice with Il10ra^f/f mice, respectively. As controls, Il10ra^f/f mice were crossed with Cdh5Cre (endothelial cell–specific) mice or with Foxp3Cre mice (Foxp3⁺ T_reg–specific), and we also used the CD4 dominant-negative IL-10R receptor (DN Il10Rα) mice to specifically down-regulate IL-10R signaling in T cells (Fig. 5B). Following MPE induction, CD11cCre⁺Il10ra^f/f mice showed significantly less fluid than littermate controls (Fig. 5C). Therefore, the data suggested that CD11c⁺ DCs might be the possible target of IL-10 in MPE. On the contrary, LysMCre⁺Il10ra^f/f (Fig. 5D) and Cdh5Cre⁺Il10ra^f/f mice (Fig. 5E) did not exhibit reduced MPE progression compared with littermate controls. Last, both DN Il10Rα mice (Fig. 5F) and Foxp3Cre⁺Il10ra^f/f mice (Fig. 5G) showed reduced MPE compared to littermate controls, indicating that Foxp3⁺ T_regs are also potential target cells of IL-10 in MPE. None of these mouse lines showed significantly altered tumor burden (fig. S7).

Therefore, DCs and Foxp3⁺ T_regs are the relevant target cells of IL-10 during MPE. While the role of Foxp3⁺ T_regs (42) and of T helper (T_H) cells in Il10^−/− mice during MPE progression has been already studied (21), how DCs, in response to IL-10, promote MPE remains unexplored. We therefore decided to further address the mechanisms behind the role of DCs in MPE.

DC-derived Timp1 promotes MPE

Further analysis of the scRNA-seq data led to the identification of a cluster of DCs (Itgax and Cd209a) characterized by the expression of MHC (major histocompatibility complex) class II–related genes (H2-Aa, H2-Ab1, and H2-Eb1). This DC population was also enriched for genes involved in biological processes such as T cell activation, antigen processing, and presentation through MHC class II molecules (fig. S8A). We next tested whether this population was enriched for plasmacytoid (Clec4c, Tcl1a, Irf8, and Tlr7) or cDCs (Cd11b, Itgax, Cd24a, Cd115, and Flt3 and MHC class II–related genes). The expression pattern of these cells indicated that they are most likely cDCs (fig. S8B). We observed that Timp1 was significantly enriched in this DC cluster compared to all the other populations (Fig. 6A). TIMP1 is a glycoprotein that regulates the structure of the extracellular matrix. Previous studies have reported a link between TIMP1 and angiogenesis, a key aspect in MPE progression (43).

Fig. 6. DC-derived Timp1 promotes MPE.

(A) Violin plots of log-transformed Timp1 gene expression in the indicated cell populations. (B) TIMP1 expression in human monocyte-derived DC (moDC) upon IL-10 exposure for 0 min, 45 min, 2 hours, 4 hours, 8 hours, and 12 hours from GSE45466 dataset. P value or ns compared with 0 hours. (C) The expression levels (counts per million) of TIMP1 in human immature moDCs and immature IL-10 antigen-presenting cells (APCs) from GSE92852 dataset (P < 0.05). (D) qPCR of Timp1 expression in pleural DCs from CD11cCre⁺Il10ra^f/f (n = 8) and CD11cCre⁺Il10ra^wt/wt (n = 12) mice under MPE condition. (E) qPCR of Timp1 expression in pleural Mφs from LysMCre⁺Il10ra^wt/wt (n = 5) and LysMCre⁺Il10ra^f/f (n = 5) mice under MPE condition. (F and G) Timp1^−/− (n = 19) and WT (n = 13) littermate control mice were intrapleurally injected with LLC cancer cells. (F) Representative images of MPEs (dashed lines) captured through the diaphragm and respective MPE quantification. (G) Representative images of pleural tumors (t), lungs (l), and hearts (h) and respective tumor weight. (H) IL-10 concentration in MPE of WT (n = 11) and Timp1^−/− (n = 7) mice quantified via ELISA. (I) qPCR of Timp1 expression in pleural DCs from Csf1r-Cre⁻ Axl^f/fMertk^f/f (n = 7) and Csf1r-Cre⁺ Axl^f/fMertk^f/f (n = 7) mice under MPE condition. (J) Graphical abstract: Mφs produce IL-10 upon phagocytosis via MERTK and promote MPE by inducing the secretion of TIMP1 from DC cells. (D) Data of two experiments. (E, H, and I) Data of one experiment. (F and G) Data of three experiments. Each symbol represents a measurement from a single mouse. Data are means ± SEM. *P < 0.05 and **P < 0.01. Statistical analysis was performed using Mann-Whitney U test (C to I). Photo credit: L. Zhao (University Medical Center Hamburg-Eppendorf).

We next sought to determine whether IL-10 could promote TIMP1 expression in DCs. Therefore, we examined two different GSE datasets and found that stimulation with IL-10 induces TIMP1 expression in in vitro cultured DCs. TIMP1 expression was increased in human monocyte-derived DCs (moDCs) cultures upon IL-10 stimulation (GSE45466; Fig. 6B). In addition, another dataset showed that TIMP1 expression is higher in human immature IL-10⁺ antigen-presenting cells than in immature moDCs (GSE92852; Fig. 6C). Furthermore, we sorted DCs from CD11cCre⁺Il10ra^f/f mice and littermate controls and checked the Timp1 expression in DCs. We observed a trend in the reduction of Timp1 expression in DCs due to Il10 signaling ablation (Fig. 6D). In contrast, Mφ did not show any reduction in Timp1 expression when they do not respond to IL-10 (Fig. 6E), suggesting that despite Mφs outnumbering DCs, the latter might be more reactive to IL-10 with regard to the expression of TIMP1.

Then, to investigate whether TIMP1 is able to promote MPE, we injected LLC cells into Timp1^−/− mice and wild-type (WT) littermate controls. Timp1^−/− mice showed less MPE (Fig. 6F), while tumor growth was not affected (Fig. 6G). We also checked the concentration of IL-10 in MPE supernatant from Timp1^−/− mice and their WT littermate controls, and we did not find any significant difference between the two groups (Fig. 6H), indicating that TIMP1 is downstream of IL-10 in MPE progression. Last, we checked Timp1 expression in DCs isolated from Csf1r-Cre⁻Axl^f/fMertk^f/f and Csf1r-Cre⁺Axl^f/fMertk^f/f mice during MPE development. We observed a significant reduction of Timp1 expression in Csf1r-Cre⁺Axl^f/fMertk^f/f mice (Fig. 6I). All in all, these results argue in favor of a cellular cascade characterized by an initial phagocytosis of apoptotic cells by Mφs, followed by IL-10 production that signals through DCs, promoting the expression of Timp1 (Fig. 6J).

DISCUSSION

MPE is a common complication of advanced oncological conditions, including lung and breast cancer (44). Studies in humans and mice indicate that complex inflammatory interactions among tumor, immune cells, and endothelial cells lead to pleural blood vessel leakiness culminating in MPE formation (45). For example, tumor foci have been described to secrete several vasoactive mediators, which contribute to the development of MPE by increasing blood vessel permeability, including vascular endothelial growth factor (VEGF), angiopoietins, and endostatin. Newer evidence shows that proinflammatory mediators such as IL-6, tumor necrosis factor (TNF), monocyte chemotactic protein 1, CCL2, and osteopontin also play an important role in the process (30, 38, 46–48). Here, we report an additional mechanism behind MPE progression, characterized by Mφ efferocytosis of apoptotic cells followed by IL-10 production, which in turn promotes TIMP1 secretion by DCs.

It is known that human Mφs can acquire an anti-inflammatory phenotype in MPE and this associates with a poor prognosis in lung cancer (8). In addition, in a mouse model of MPE, the depletion of Mφs leads to a reduced MPE progression (49). However, the molecular mechanisms mediating this were unclear to a large extent. We focused here on the role of phagocytosis performed by Mφs during MPE progression. Our data are in line with previous studies reporting accumulation of apoptotic cells in the pleural effusion during malignancy and in various inflammatory settings such as tuberculosis (50). Our results show that phagocytosis is a critical mechanism for the progression of MPE, defining this process as one contributor of the etiology of this disease and thus a previously unidentified therapeutic target.

An in-depth and unsupervised characterization of Mφs has never been performed so far. scRNA-seq enabled us to uniquely address this gap in knowledge and to precisely quantify differences among identified subpopulations of Mφ.

In line with the literature (35–37), the gene expression profiles of Mφ_1 and Mφ_2 were reminiscent of anti-inflammatory and tissue regenerative Mφs, suggesting a pro-MPE phenotype. On the contrary, the other two populations of Mφs (Mφ_3 and Mφ_4) did not display any gene expression patterns linking them to previously proposed anti-inflammatory Mφ subtypes. However, the cells from Mφ_3 are armed with a group of molecules associated with phagocytosis and lipid metabolism, such as the phagocytic receptor Cd36, which, through binding of oxidized lipids and lipoproteins, participates in the internalization of apoptotic cells, as well as the lipid receptor Trem2 and the gene Lpl coding for lipoprotein lipase, which regulates Fc receptor–mediated phagocytosis (51). Moreover, in Mφ_4, we found markers characterizing discrete populations of monocytes/Mφs, such as Ly6c2 and Tgfbi, previously described as cells with an intermediate monocyte/Mφ phenotype in a model of glioma (52). Together, our and other findings (53) indicate that the rigid and simplified nomenclature of Mφs cannot mirror the functional heterogeneity that Mφs acquire in vivo, including during MPE progression. These data set the basis for future functional experiments aiming to elucidate the role of these distinct Mφ populations in MPE development.

Although at different levels, all four Mφ populations within MPE express Il10 and the receptor Mertk. This is in line with previous findings, which identified MERTK as not homogeneously distributed among human Mφ populations cultured in vitro but mostly restricted to a subset of IL-10–producing cells (54). IL-10 secretion by Mφs has been extensively linked to the response to TLR ligands and type I interferon (22). However, increasing evidence demonstrates that phagocytosis, also via engagement of the receptor tyrosine kinases AXL and MERTK, leads to IL-10 expression. For example, this is the case in an in vivo murine model of dextran sulfate sodium–induced colitis (18), as well as in human in vitro cultured Mφ, where engagement of MERTK in inflammatory settings via the ligand growth arrest specific 6 amplifies IL-10 secretion and positively regulates homeostasis of Mφs (54). Briefly, this unsupervised method allowed us to appreciate the heterogeneity of these populations, which goes far beyond the classical M1/M2 classification.

The role of IL-10 in MPE was apparently controversial. Some studies report a higher concentration of IL-10 in MPE than in peripheral blood (19, 20), while others demonstrate no differences in IL-10 levels between MPE and peripheral blood (6). To clarify these apparent controversies, we performed a loss-of-function experiment and found that although IL-10 does not affect tumor growth, it promotes MPE progression. Consistently, Wu et al. (21) also showed that IL-10 promotes MPE formation in mice by regulating T_H1 and T_H17 cell differentiation and migration. Therefore, our findings further support IL-10 as a pro- MPE cytokine.

Our data define the cellular source of IL-10 in MPE. It has been extensively documented that various leukocyte populations (22), as well as normal and malignant epithelial cells (55), can express IL-10. In this study, by taking advantage of the LysMCre⁺Il10^f/f mice and of IL-10^eGFP tiger mice and 10BiT reporter mice, we identified Mφs as a critical source of IL-10 in MPE. This finding is in line not only with the idea that Mφs can support tumor development in many ways, such as by regulating cancer initiation and promotion, immune regulation, metastasis, and angiogenesis (56), but also with current evidence suggesting that depletion of circulating monocytes/Mφs protects from MPE formation (49). One limitation of the results obtained using the LysMCre⁺Il10^f/f mice is the presence of one group of mice with a low accumulation of MPE and one group with an intermediate accumulation. This effect was reproducible across the three experiments performed. We therefore speculate that either the tumor cells are quickly controlled by the immune response, contributing to the formation of a low amount of MPE, or if they escape the initial check point inhibition, they over proliferate partially evading the immune response. Further experiments are needed to address this point.

Which are the cells targeted by Mφ-secreted IL-10? How does IL-10 regulate MPE progression? We found that both Foxp3⁺ T_regs and DCs are the key target cells of IL-10.

It has been previously shown that Foxp3⁺ T_regs need to sense IL-10 (57) to maintain a strong regulatory activity. In addition, we showed that when Foxp3⁺ T_regs lose their regulatory activity, T_H17 and T_H1/T_H17 cells overexpand (58). It has been recently shown that the expansion of T_H17 and T_H1 cells associates with a reduced MPE formation (21). Therefore, our results using the Foxp3Cre⁺IL-10^f/f mice could be explained by a loss of regulatory activity of Foxp3⁺ T_regs leading to an anti-MPE function mediated by the expansion of T_H17 and T_H1 cells.

Regarding the role of DCs in MPE, few studies had characterized their immune functions until now. A meta-analysis evaluating the significance of immune cells in patients with NSCLC showed that increased tumor and stromal DCs with a mature phenotype were associated with improved survival (15). This improved survival was also associated with an accumulation of T cells with a T_H1 phenotype (59). In addition, Gu et al. (60) characterized DCs from pleural effusion of patients with NSCLC as immature and observed that when stimulated by a TLR-4/7/8 agonist in vitro, DCs were able to induce T_H1 differentiation. A different report associates the expansion of T_H1 cells with reduced MPE development in a murine model (21). In line with this, our results show that when DCs are not able to respond to IL-10, the MPE progression is reduced. By reanalyzing an already published dataset, we found that DCs express Timp1 when stimulated with IL-10. Last, our data suggest that when murine DCs do not respond to IL-10, the expression of Timp1 appears to be down-regulated. Timp1 KO mice also have a reduced MPE development. Briefly, we could speculate that, when DCs maintain an immature phenotype by responding to IL-10, they promote TIMP1-mediated MPE progression, while when they do acquire a proinflammatory phenotype, they can mediate the anti- MPE process by promoting T_H1 cells. However, one limitation of our study is the use of the whole-body Timp1 KO mice. Therefore, the development of conditional CD11cCre⁺Timp1^f/f mice would be ideal to further support the role of DC-derived TIMP1in MPE formation.

Until now, divergent functions of TIMP1 during tumor development and metastasis formation have been described, most likely because of the metalloprotease-independent versus metalloprotease-dependent functions that TIMP1 can exert. Elevated levels of TIMP1 have been associated with poor prognosis in a variety of cancers (61, 62). The prometastatic function of TIMP1 has not only been attributed to its direct effect on tumor cell intrinsic metastatic potential, via its ability to promote tumor cell proliferation (63) and inhibit apoptosis (64), but also due to its capacity to promote angiogenesis, as shown by the increased vascularization upon TIMP1 overexpression in tumor cell lines (43). Furthermore, Psallidas et al. (65) described a proangiogenic function of TIMP1 with TIMP1 concentration in MPE being associated with the survival of patients. On the contrary, the antiangiogenic effect of TIMP1 has been associated with the inhibition of metalloproteases. This inhibition in turn blocks degradation of the extracellular matrix, thereby inhibiting tumor invasion and metastasis (66). Furthermore, TIMP1 is known to affect vascular permeability (67), which might also affect MPE progression.

Our results indicate that TIMP-1 promotes MPE progression. It is tempting to speculate that, in our experimental setting, TIMP-1 promotes MPE progression by affecting angiogenesis and/or vascular permeability. Further experiments will be essential to test this hypothesis.

Clinical MPE can be caused by virtually any tumor type, as a consequence of pleural metastasis. Therefore, we chose the LLC cell line as a representative one, as previously shown (36). One limitation of this model is that there is no spontaneous formation of the primary tumor and the secondary pleura metastasis. However, the model used here is, to the best of our knowledge, the only one currently existing. In addition, several clinical targets have been identified using this model, such as VEGF, TNF, IL-6, and CCL2, as well as specific mutations including epidermal growth factor receptor mutations and KRAS mutations (1, 2, 8–11), which can strengthen the link between our model and the clinical scenario. These preclinical data in conjunction with our new mechanistic study suggest that TIMP-1 might be a novel potential therapeutic target in MPE.

Briefly, our data reveal an inflammatory cascade underlying MPE progression and pinpoint some key targets—e.g., TIMP1, IL-10, and MERTK—for potential future MPE immune treatments. Understanding of tumor immunology in conjunction with technological innovations will yield novel immune-based treatments and will result in unprecedented development of therapies for MPE treatment.

MATERIALS AND METHODS

Cancer cell lines

LLC (NCI Tumor Repository, Frederick, MD) and MC38 cells were cultured at 37°C in 5% CO₂ to 95% air using Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum (FBS), 1 mM pyruvate, 2 mM l-glutamine, streptomycin (100 mg/ml), and penicillin (100 U/ml). For in vivo injections, cells were harvested, counted, and injected through a left intercostal space, as described elsewhere (46).

Animals

C57BL/6 mice were obtained from the Jackson laboratory (Bar Harbor, MN), Axl^f/fMertk^f/f mice, crossed with Csf1rCre⁺ mice, were provided by Carla V. Rothlin and colleagues (12), IL-10^Thy1.1 (CD90.1) (10BiT) mice were provided by C. Weaver (68), Il10^f/f were provided by G. Tiegs (69), Il10ra^f/f were generated by Y. Kobayashi (details are available on request). Il10^−/− mice (70), Foxp3^mRFP (71) and IL-10^eGFP (23) reporter mice, and LysMCre⁺ (72), Foxp3 YFP Cre⁺ (73), CD11cCre⁺ (74), Cdh5Cre⁺ (75), CD4dnIL-10Rα (76), and Timp1^−/− (77) mice are described elsewhere. Age- and sex-matched littermates between 8 and 16 weeks of age were used. All mice were in a C57BL/6 background. Animals were maintained under specific pathogen–free conditions and handled according to protocols approved by the responsible federal health authorities of the state of Hamburg (Behörde für Gesundheit und Verbraucherschutz).

MPE model induction

Intrapleural injections were performed under inhalation anesthesia using isoflurane (4%) and direct stereoscopic vision through an incision in the left anterolateral thoracic skin and fascia. For this purpose, a needle was inserted into the pleural cavity at an angle of 45° under direct contact with the upper rib, and tumor cell suspensions (0.15 × 10⁶) were injected under direct visual inspection to develop MPE. After 2 weeks, the mice were appropriately euthanized. Mice with a pleural fluid volume of ≥100 μl (equal to the initial injection volume) were subjected to pleural fluid aspiration, whereas animals with a pleural fluid volume of <100 μl were subjected to pleural lavage. For this, 1 ml of normal saline was injected intrapleurally and was withdrawn after 30 s. Following pleural fluid or lavage retrieval, the chest was opened, and pleural tumors were stripped and weighed. For analgesic treatment, the mice received metamizole via drinking water from 1 day before surgery until 2 days after. The metamizole was used in a dose of 1200 mg/kg of body weight. For sufficient analgesia during surgery, carprofen (5 mg/kg of body weight) was also administered subcutaneously to the mice 1 hour before surgery.

Flow cytometry

The cell suspension was filtered through 100- and 40-μm cell strainers and centrifuged at 400g for 5 min at 4°C. Cells were resuspended in cell blood lysis buffer (0.15 M NH₄Cl, 10 mM KHCO₃, and 0.1 mM EDTA) for 5 min, washed in phosphate-buffered saline (PBS), resuspended in FACS buffer (PBS, 2 mM EDTA, and 25 mM Hepes) at a concentration of 1 million cells/100 μl, and incubated for 20 min on ice with a mixture of appropriate antibodies (table S2). Labeled cells were washed with PBS and resuspended in FACS buffer. Cells were labeled with eBioscience Fixable Viability Dye eFluor 506 (catalog no. 65-0866-14) for cell viability.

For intracellular cytokine staining, cells were restimulated for 3 hours at 37°C with phorbol 12-myristate 13-acetate (50 ng/ml; Sigma-Aldrich) and ionomycin (1 mg/ml; Sigma-Aldrich) in the presence of GolgiStop (BD Bioscience). The cells were then fixed in paraformaldehyde for 20 min at room temperature. After washing, the cells were permeabilized (NP-40) and stained at 4°C with intracellular antibodies for 30 min. Flow cytometric analysis was performed with BD LSRFortessa cytometer (BD Biosciences, Alameda, CA), and data were examined using FlowJo software (FlowJo, Ashland, OR). In addition, data were analyzed using the Cytobank platform (viSNE analysis).

Bioluminescence imaging

Cells and mice were serially imaged on a Xenogen Lumina II after addition of d-luciferin (300 μg/ml) to culture media or intravenous delivery of 1 mg of d-luciferin. Data were analyzed using Living Image v.4.2 (PerkinElmer, Waltham, MA).

Human samples

Pleural fluid samples were received during diagnostic thoracenteses in patients suffering from MPE caused by lung cancer (n = 19), other cancers (n = 10), as well as patients with PPE (n = 15) and TrPE (n = 17) treated at the Department of Respiratory Medicine, University of Thessaly, Biopolis, Larissa, Greece. All protocols abided by the Helsinki Declaration were approved a priori by the local hospital ethics committees and by all patients via written informed consent.

Enzyme-linked immunosorbent assays

IL-10 levels in cell culture supernatants, cell-free MPE, and sera were determined using dedicated murine or human ELISA kits according to the manufacturer’s instructions (PeproTech, London, UK and R&D Systems, Minneapolis, MN).

Apoptosis assay

Apoptotic cells were examined using the Annexin V–FITC Apoptosis Detection Kit (ab14085, Abcam, Cambridge, UK) following the manufacturer’s instructions. Briefly, cells were harvested by careful centrifugation, washed twice with 1× annexin V binding buffer, resuspended in binding buffer, and stained with annexin V and propidium iodide (PI). Cell apoptosis was detected using the FACS (BD LSRFortessa, BD Biosciences).

Neutrophil and thymocyte isolation and induction of apoptosis

Neutrophils were isolated from bone marrow cell suspensions by negative selection using the Neutrophil Isolation Kit (Miltenyi) following the manufacturer’s instructions. aNs were generated by aging for 24 hours in complete media containing 90% RPMI 1640, 5% FBS, 0.5% gentamicin, and 1% l-glutamine. Thymocytes were isolated via disruption of thymic tissue. aTs were generated by aging overnight in complete media containing 90% RPMI 1640, 10% FBS, 0.5% gentamicin, and 1% l-glutamine. Apoptosis was verified by annexin V and PI staining. aTs were then labeled with eF450 dye following the manufacturer’s instructions and used for coculture experiments.

LPS stimulation of phagocytic Mφs

Bone marrow–derived Mφs from WT mice were differentiated from bone marrow precursors. Briefly, bone marrow cells were isolated and propagated for 7 days in RPMI 1640 containing 20% FBS, 30% L929-conditioned media, 0.5% gentamicin, and 1% l-glutamine. Mφs (0.33 × 10⁶) cultured in 24-well plates were then exposed to aNs at a 5:1 ratio. After 1 hour, cells were washed five times with ice-cold PBS. Mφs were then resuspended in RPMI 1640 containing 20% FBS, 30% L929-conditioned media, 0.5% gentamicin, and 1% l-glutamine, and stimulated with LPS (100 ng/ml) (Sigma-Aldrich) for 24 and 48 hours. The amount of secreted IL-10 in the supernatant was analyzed by ELISA.

Phagocytosis assay

Mφ clusters (Mφ_1, Mφ_2, Mφ_3, and Mφ_4) FACS-sorted from MPE were seeded into a 48-well plate (0.09 × 10⁶ to 0.15 × 10⁶ cells per well) and cocultured with aTs at a 5:1 ratio. Coculture of Mφs and aTs was performed under two conditions simultaneously, at 4° and 37°C to respectively discriminate Mφs that only bind to aTs (4°C condition) and Mφs that bind and uptake aTs (37°C condition). After 1 hour, cells were washed five times with ice-cold PBS. Mφs were then collected and restained for F4/80 and CD11b. Analysis was performed via flow cytometry (BD LSRFortessa). Phagocytic Mφs have been defined as the difference between the percentage of CD11b⁺F4/80⁺eF450⁺ cells under the 37°C condition and the percentage of CD11b⁺F4/80⁺eF450⁺ cells under the 4°C condition.

Vascular permeability assays

Mice with MPE received 0.8 mg of Evans’ blue intravenously, were euthanized after 1 hour, and the albumin-binding dye levels in the MPE were determined (46).

Reverse transcriptase PCR and real-time qPCR

RNA isolation was performed using TRIzol LS Reagent (Life Technologies) according to the manual. The isolated RNA was subjected to reverse transcription with the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). cDNA was semiquantified using commercially available primer/probe sets from Applied Biosystems or using SYBR Green Master Mix in a StepOnePlus cycler (Applied Biosystems, Carlsbad, CA). Samples were analyzed with the change in cycle threshold method. Results were normalized to hypoxanthine phosphoribosyltransferase (Hprt) quantified in parallel amplification reactions. Reverse transcriptase PCR primers are given in table S3.

Single-cell RNA sequencing

Single cells were encapsulated in droplets using 10X Genomics GemCode Technology and processed following the manufacturer’s specifications. Briefly, every cell and every transcript were uniquely barcoded using a unique molecular identifier (UMI), and cDNA ready for sequencing on Illumina platforms was generated using the Single Cell 3′ Reagent Kits v2 (10X Genomics). Libraries were sequenced across a HiSeq 4000 (Illumina) in paired end to reach approximately 50,000 reads per single cell.

scRNA-seq data analysis

Data were demultiplexed using Cell Ranger software (version 2.0.2) based on 8–base pair 10× sample indexes, and paired-end FASTQ files were generated. The cell barcodes and transcript UMIs were processed as previously described (78). The reads were aligned to mouse UCSC mm10 reference genome using STAR aligner. The alignment results were used to quantify the expression level of mouse genes and generation of gene-barcode matrix. Low-quality cells, doublets, and potentially dead cells were removed according the percentage of mitochondrial genes and number of genes and UMIs expressed in each cell. The description of scRNA-seq quality controls is reported in table S4. The remaining data were normalized and log transformed, and the log-transformed matrix was used for all downstream analysis. We used “Harmony” (version 0.99.9) and “Seurat” (version 2.3.4) packages to combine the scRNA-seq data of two experiments using five mice in each experiment. Applying the same filtering criteria, we captured 1989 cells (after filtering out low-viability cells). The data of the two scRNA-seq sets were superimposable, suggesting that we can perform downstream analysis with the integrated data (fig. S5D). Data clustering was performed using Seurat R package. Highly variable genes—genes with relatively high average expression and variability—were detected with Seurat, and these genes were used for clustering analysis. Principle components analysis (PCA) was used for dimensionality reduction, and the number of significant principal components was calculated using built in “JackStraw” function. t-SNE was used for data visualization in two dimensions. Two complementary approaches were used to identify differentially expressed genes. In the results of Seurat package, genes with P < 0.01 were considered differentially expressed genes.

Preprocessing analysis with Seurat Package

The Seurat pipeline was applied to the dataset. Genes that were expressed in less than three cells and cells that expressed less than 200 and more than 4000 genes were excluded. Data were normalized with a scale factor of 10⁴. Latent variable—number of UMIs—was regressed out using a negative binomial model (function ScaleData). Most variable genes were detected by the “FindVariableGenes” function and used for subsequent analysis. PCA was performed on about 2000 genes with PCA function. A t-SNE dimensional reduction was performed on the scaled matrix (with most variable genes only) using first 15 PCA components to obtain a two-dimensional representation of the cell states. For clustering, we used the function “FindClusters” that implements SNN (shared nearest neighbor) modularity optimization–based clustering algorithm on 20 PCA components with resolution 0.6, leading to nine clusters for the analysis.

Identification of cluster-specific genes and marker-based classification

To identify marker genes, the “FindAllMarkers” function was used with likelihood ratio test for single-cell gene expression. For each cluster, only genes that were expressed in more than 25% of cells with at least 0.25-fold difference were considered. To perform GO analysis, we used “clusterProfiler” (3.12.0). For heatmap representation, mean expression of markers inside each cluster was used.

Public gene expression data analysis

The gene expression profiles GSE45466 and GSE92852 were extracted from GEO (Gene Expression Omnibus) database, including two experiments of human moDCs differentiated in the presence or absence of IL-10. The processed transcript abundance data were downloaded and used directly.

Statistics

Sample size was calculated using power analysis on G*power academic freeware (79), assuming α = 0.05, β = 0.8, and ρ = 0.3 (www.gpower.hhu.de/). Animals were distributed to different treatment groups by alternation and transgenic animals were enrolled case control–wise. Data were collected by at least two blinded investigators from samples coded by a nonblinded investigator. Sample size (n) always refers to biological and not technical replicates. Differences in means between two or multiple groups were examined, respectively, by t test or one-way analysis of variance (ANOVA) with Bonferroni post hoc tests and in medians between two or multiple groups by Mann-Whitney U test or Kruskal-Wallis test with Dunn’s post hoc tests, as appropriate. All P values are two-tailed and were considered significant when <0.05. All statistical analyses were performed, and plots were created using Prism v8.0 (GraphPad, La Jolla, CA).

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/7/33/eabd6734/DC1

View/request a protocol for this paper from Bio-protocol.