Novel Cell-to-Cell Communications Between Macrophages and Fibroblasts Regulate Obesity-Induced Adipose Tissue Fibrosis Free

Macrophage infiltration into adipose tissue, in obesity, was first reported in 2003 (1,2). Accumulating evidence indicates that in addition to parenchymal mature adipocytes, a variety of stromal cells, such as immune cells, endothelial cells, and fibroblasts, play critical roles in adipose tissue function, thereby regulating systemic metabolic homeostasis through adipokine production and lipid storage (3). Chronic overnutrition induces adipocyte hypertrophy, which is followed by immune cell infiltration and angiogenesis, eventually resulting in overproduction of the extracellular matrix (ECM). Such dynamic histological changes are referred to as “adipose tissue remodeling.” Recent evidence indicates significant individual differences in the phenotype of adipose tissue remodeling associated with obesity (4). Most individuals with obesity belong to the category of “metabolically unhealthy obesity,” characterized by insulin resistance, type 2 diabetes, and cardiovascular complications, while a relatively small number of individuals, known as “metabolically healthy obesity,” do not exhibit these complications. The major difference between these categories is adipose tissue remodeling, with persistent inflammatory changes leading to interstitial fibrosis, which is thought to limit adipocyte expansion and thus induce ectopic lipid accumulation. Indeed, there is an inverse correlation between adipose tissue fibrosis and hepatic steatosis of individuals with obesity (5,6). Moreover, collagen VI deficiency in mice inhibits obesity-induced adipose tissue fibrosis and ameliorates systemic metabolic derangements through uninhibited adipocyte expansion (7). However, the molecular mechanism of adipose tissue fibrosis remains unclear, because myriad cell types communicate with each other during this process.

Stromal cell diversity was first noted regarding proinflammatory M1 and anti-inflammatory M2 macrophages, in which phenotypic changes occur in adipose tissue during the development of obesity (8). Recently, single-cell RNA sequencing (scRNA-seq) has dramatically increased our knowledge of stromal cell diversity in adipose tissue. For instance, a macrophage subcluster characterized by triggering receptor expressed on myeloid cells 2 (Trem2) and CD9, termed “lipid-associated macrophages (LAMs),” increases its population in parallel with adiposity, exerting a proinflammatory effect (9,10). Fibroblasts also include subclusters, such as mesenchymal stem cells, adipocyte progenitor cells, and ECM-producing profibrotic cells. Marcelin et al. (11) and Hepler et al. (12) reported that a subpopulation of progenitor cells expressing platelet-derived growth factor receptor-α (PDGFRα) and CD9 have profibrotic effects in the adipose tissue during obesity. Although knowledge of the diversity of specific cell types has accumulated, little is known about cell-to-cell communication among distinct cell types. Notably, it is crucial to distinguish the fibrotic phase from the inflammatory phase, because unlike inflammation, fibrosis is difficult to reverse once it develops.

There are unique histological structures in adipose tissue termed “crown-like structures” (CLS), where infiltrated macrophages aggregate around and engulf dead/dying adipocytes (13). We previously demonstrated that factors secreted from adipocytes and macrophages form a vicious cycle that accelerates chronic inflammation in adipose tissue (14,15). In particular, macrophage-inducible C-type lectin (Mincle), an innate immune sensor for Mycobacterium, is selectively expressed in macrophages constituting CLS and promotes obesity-induced adipose tissue fibrosis upon sensing adipocyte cell death, without affecting overall proinflammatory cytokine levels in adipose tissue (16,17). This indicates that CLS are the sites of cellular interactions in obesity-induced adipose tissue fibrosis, although its molecular mechanisms remain largely unknown.

In this study, we used scRNA-seq and spatial transcriptomics to investigate the interactions among stromal cells in adipose tissue from diet-induced obese mice, with a particular focus on CLS during the fibrotic phase. We identified several macrophage and fibroblast subclusters with a novel cross talk potentially responsible for adipose tissue fibrosis. Moreover, we found that oncostatin M (Osm), a member of the interleukin-6 cytokine superfamily secreted from Mincle-expressing macrophages, acts on fibroblasts to inhibit collagen expression, thereby regulating ECM remodeling. The strategies used in this study will help us understand the complex cell-to-cell communication involved in chronic inflammation-regulated tissue remodeling.

All reagents were purchased from Sigma-Aldrich (St. Louis, MO) or Nacalai Tesque (Kyoto, Japan), unless otherwise stated. Antibodies used for immunohistochemistry, flow cytometry, and magnetic cell sorting are listed in Supplementary Table 1.

C57BL/6J mice were purchased from CLEA Japan (Tokyo, Japan). Mincle-deficient mice and Col1a2/EGFP-transgenic mice with a C57BL/6J background were kindly provided by Dr. Shizuo Akira (Osaka University, Osaka, Japan) and Dr. Yutaka Inagaki (Tokai University, Tokyo, Japan) (18), respectively. Osm-deficient mice with 129 genetic backgrounds were backcrossed with C57BL/6 mice more than seven times (19). All animals were housed in a temperature-, humidity-, and light-controlled animal room (12 h light/dark cycle) and allowed free access to water and standard diet (SD: 343.1 kcal/100 g, 12.6% energy as fat; CE-2, CLEA Japan). Eight-week-old animals were fed the SD or high-fat diet (HFD: 556 kcal/100 g, 60% energy as fat; D12492, Research Diets, New Brunswick, NJ) for 8–36 weeks. Blood analyses were performed as described (20). All animal experiments were conducted in accordance with the Guidelines for the Care and Use of Laboratory Animals at Nagoya University. All experimental protocols were approved by the Animal Care and Use Committee of the Nagoya University Research Institute of Environmental Medicine (approval no. R240021).

Subcutaneous adipose tissue was provided by Toho University Sakura Medical Center (Sakura, Japan). The specimens were obtained from patients with obesity (mean BMI 43.5 kg/m² [31.5–61.7], 16 men aged 19–59 years and 19 women aged 23–64 years) receiving bariatric surgery. The control specimens were obtained during dermatological surgery (mean BMI 24.5 kg/m² [18.1–29.5 kg/m²] without metabolic diseases, two men aged 58 and 66 years, and six women aged 65–82 years). The Toho University Sakura Medical Center Ethics Committee (batch number: 2013-093, 2013-094) approved this study, and all patients who provided specimens signed an informed consent form before the operation. The specimens were promptly soaked in 5% saline (obesity) or RNAlater (QIAGEN Bioinformatics, Redwood City, CA) (control), returned to the laboratory, and stored at –80°C until further use. Relevant clinical information from patients was duly recorded. Analysis of visceral adipose tissue was performed using the Gene Expression Omnibus DataSet (GSE235696).

The epididymal adipose tissue was cut into small pieces and incubated for 30 or 60 min in collagenase solution (2 mg/mL collagenase type 2 [Worthington, Lakewood, NJ] or 0.5 mg/mL Brightase-C [Nippi, Tokyo, Japan], respectively) with gentle shaking at 37°C. For scRNA-seq analysis, the adipose tissue was digested for 45 min with 0.25 mg/mL Liberase TM solution at 37°C. After being filtered through a 180-μm mesh, it was centrifuged and resuspended as stromal vascular fraction (SVF).

Flow cytometric analysis of SVF was performed as described (17). Cells were analyzed using SA3800 (SONY, Tokyo, Japan) and FlowJo software version 10 (BD Biosciences, Franklin Lakes, NJ) or sorted using SH800 (SONY). Sorted cells were used for mRNA expression analysis, Giemsa staining (FUJIFILM Wako Pure Chemical, Osaka, Japan), BODIPY staining (5 µmol/L; Thermo Fisher Scientific, Waltham, MA), and electron microscopy.

Macrophages from epididymal adipose tissue were fixed by a conventional fixation method (1.6% paraformaldehyde and 3% glutaraldehyde in 0.1 mol/L phosphate buffer [pH 7.4], followed by an aqueous solution of 1% osmium tetroxide). Fixed samples were embedded in Epon 812, and thin sections (70–80 nm) were then cut and stained with uranyl acetate and lead citrate for observation under a Jeol-1010 electron microscope (Jeol, Tokyo, Japan) at 80 kV (21).

Thioglycolate-elicited peritoneal macrophages and bone marrow–derived macrophages were obtained as described (16,17). RAW264 macrophages were cultured in DMEM supplemented with 10% FBS and antibiotics. For Mincle activation, macrophages were stimulated with trehalose-6,6’-dimycolate (TDM), as described (17). Adipose tissue–derived fibroblasts were prepared from the SVF of HFD-fed mice using a magnetic cell sorting system (AutoMACS; Miltenyi Biotec) as CD45-, CD31-, and Ter119-negative cells. Mouse embryonic fibroblasts (MEFs) were obtained as described (22). Adipose tissue–derived fibroblasts or MEFs were treated with recombinant human transforming growth factor-β1 (TGF-β1) (3 ng/mL; BioLegend) or recombinant mouse Osm (0.3 and 3 ng/mL; BioLegend) for 24 h. Adipose tissue–derived fibroblasts were treated with conditioned media obtained from RAW264 macrophages stimulated with TDM for 24 h.

The Osm concentrations in the culture media were measured using a commercially available ELISA kit (R&D Systems, Minneapolis, MN) according to the manufacturer’s protocols. Serum insulin concentrations were measured using a commercially available ELISA kit (Morinaga Institute of Biological Science, Kanagawa, Japan) according to the manufacturer’s protocols.

Eight-week-old male wild-type mice were fed the SD for 16 weeks or the HFD for 8 or 16 weeks. Epididymal adipose tissue was cut into small pieces, followed by permeation with CELLBANKER 1 (Nippon Zenyaku Kogyo, Fukushima, Japan) for 2 h at 4°C with gentle rotation. Samples were stored at −80°C, slowly frozen by using CoolCell (Corning, Corning, NY), and stored at −80°C before single-cell preparation step. Single-cell suspensions from the SVF were trapped and reverse transcribed using BD Rhapsody (BD Biosciences), according to the manufacturer’s instructions. scRNA-seq was performed using the terminator-assisted solid-phase complementary DNA amplification and sequencing (TAS-Seq) platform (23). Sequencing was performed using a NovaSeq 6000 (Illumina, San Diego, CA) to a depth of ∼50,000 reads per cell. Details of data processing are provided in another section. The processed data set was analyzed and visualized using Seurat v4.1.0 (24) in R 4.2.0 software. Pathway analysis of fibroblast marker genes in subcluster 3 was performed using Ingenuity Pathway Analysis (IPA) (QIAGEN Bioinformatics) to reveal the biological relevance of the marker genes. Intercellular communications between two cell types was inferred by using the CellChat v1.5.0 (25) on normalized data generated separately for each condition. The LAM and perivascular macrophage scores were calculated by using the AddModuleScore function in Seurat (26). Pseudo-bulk analysis was performed using DESeq2 v1.44.0 (27,28).

After adapter removal and quality filtering by Cutadapt-2.1.0 (29) in R 3.6.3 software, gene expression libraries were aligned to mouse Ensemble RNA by Bowtie2-2.4.2, and count matrices were generated using the modified python script of BD Rhapsody workflow. Valid cell barcodes were identified as cell barcodes above the inflection threshold of knee plot of total read counts of each cell barcode identified by the DropletUtils package (30). Sample origins and doublets were identified based on fold change (FC) of the normalized read counts of the hashtags. The resultant data set was mainly analyzed using Seurat v4.1.0 (24) in R 4.2.0. As quality control, doublets were filtered out. The log-normalized gene counts were calculated using NormalizeData function (scale.factor = 1,000,000), and highly variable genes were defined by FindVariableFeatures function (selection.method = mvp, mean.cutoff = c[0.1, Inf], dispersion.cutoff = c[0.5, Inf]). Principal component analysis was performed on the variable genes, and principal components with their P value <0.05 calculated by the jackstraw method, were subjected to cell clustering (resolution = 2.0 for Fig. 2C: 0.5 for Fig. 7B) and fast-Fourier transform–accelerated interpolation-based t-stochastic neighborhood embedding (FItSNE) or uniform manifold approximation and projection (UMAP) dimensional reduction. For after Fig. 2G, macrophage or fibroblast clusters were subjected to reclustering analysis (resolution = 1.0 for Fig. 2G and 0.6 for Fig. 2I). Macrophage and fibroblast subclusters in Fig. 8C and 8D were labeled with each subcluster in Fig. 2G and 2I using the R package SingleR 2.2.0 (31).

Eight-week-old male wild-type mice were fed SD or HFD for 36 weeks. Five-micrometer-thick paraffin-embedded sections of epididymal adipose tissue were placed on a 10x Visium Spatial Gene expression slide (10x Genomics, Pleasanton, CA). cDNA synthesis and library preparation were conducted according to the 10x Genomics instructions. Sequencing was performed using NovaSeq X Plus (Illumina) with a 150 base pair paired-end configuration. Raw sequencing data were processed using Space Ranger v2.0.1 software (10x Genomics). Alignment to the mouse reference genome mm10 and initial quality control steps were performed according to software guidelines. The resulting feature barcode matrices were loaded into Seurat, and data were normalized (using the “SCTransform” function in Seurat for independent tissue sections), reduced, and visualized.

Adipose tissue–derived fibroblasts were treated with recombinant human TGF-β1) (3 ng/mL; BioLegend) or recombinant mouse Osm (3 ng/mL; BioLegend) for 24 h. Total RNA was purified using RNeasy Plus Universal Mini Kit (QIAGEN). RNA integrity was assessed using the Bioanalyzer 2100 system (Agilent Technologies, Santa Clara, CA). mRNA was purified from total RNA using poly-T oligo–attached magnetic beads. After fragmentation, first-strand cDNA was synthesized using random hexamer primers, followed by second-strand cDNA synthesis. The library was ready after end repair, A-tailing, adapter ligation, size selection, amplification, and purification. After library quality control, libraries were sequenced with 150 base pair paired-end reads on a NovaSeq X Plus sequencer (Illumina). Low-quality reads, reads containing adapter, and reads containing ploy-N were trimmed using fastp software from raw data. The reads were mapped to the mm39 reference genome using HISAT2 v2.0.5. featureCounts v1.5.0 was used to count the reads numbers mapped to each gene. Read counts were adjusted using edgeR 3.22.5, followed by differential expression analysis. An interactive network was created based on a select list of genes associated with collagen assembly and organization using IPA (QIAGEN). Differentially expressed genes were selected using a cutoff value of |log₂FC| >1.

Epididymal adipose tissue was fixed in 10% neutral-buffered formalin and embedded in paraffin. Four-micrometer-thick sections were stained with Sirius red, and immunohistochemical staining was performed using antibodies against collagen III, F4/80, Osm, and Osm-specific receptor subunit β (Osmr). All samples were analyzed using a fluorescence microscope (BZ-X710; KEYENCE, Osaka, Japan) or AX confocal laser-scanning microscope (Nikon Solutions, Tokyo Japan). Adipose tissue fibrosis was defined as an area positive for Sirius red or collagen III. F4/80 immunostaining was used to detect CLS, and the number of CLS was counted in the entire area of each section and expressed as the mean number/mm². For the measurement of adipocyte diameter, >500 cells were counted per each sample using an image analysis software (Hybrid cell count; KEYENCE). Quantitative histological analysis was performed by two investigators who were blinded to the origin of the slides.

In situ hybridization analyses were performed using formalin-fixed and paraffin-embedded mouse epididymal adipose tissue samples with RNAscope technology (RNAscope Multiplex Fluorescent Reagent Kit v2; Advanced Cell Diagnostics, Newark, CA) and custom-designed probes according to the manufacturer’s instructions. Briefly, tissue sections were deparaffinized and incubated with an H₂O₂ solution for 10 min at room temperature. Slides were boiled in target-retrieval solution for 15 min, incubated with protease solution for 30 min at 40°C, incubated with the relevant probes for 2 h at 40°C, and then successively incubated with Amp1 to Amp3 reagents. Staining was visualized with Opal570 (AKOYA, Marlborough, MA) for C2 and Opal690 (AKOYA) for C1, followed by incubation with DAPI for nuclear staining and mounting in PermaFluor Aqueous Mounting Medium (TA-030-FM; Thermo Fisher Scientific). The RNAscope probes used were as follows: mouse Mincle (Clec4e) (cat no. 401281-C2) and mouse Acta2 (cat no. 319531). Images were acquired using the sectioning mode of a fluorescence microscope (BZ-X710; KEYENCE) to obtain high-resolution optical sections.

Quantitative real-time PCR was performed as described (32). The primer sequences are shown in Supplementary Table 2. Data were normalized to 36B4 levels for mice or ACTB (β-Actin) levels for humans and analyzed using the comparative cycle threshold method.

Bone marrow transplantation was performed as described (17). Briefly, bone marrow cells obtained from donor mice were washed three times with cold PBS and injected intravenously (1.6 × 10⁶ cells) into 10-Gy irradiated 8-week-old male recipient mice. After 4 weeks, the mice were fed the HFD for 24 weeks.

Data are presented as the mean ± SEM, and P < 0.05 and P < 0.01 were considered statistically significant. Statistical analysis was performed using ANOVA, followed by the Tukey-Kramer test. An unpaired t test was used to compare the two groups. Relationships between the expression levels of MINCLE and OSM were evaluated by calculating the Spearman correlation coefficient. Data were analyzed using GraphPad Prism 9 software.

Data sets in this publication have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus and are accessible through GSE275590, GSE276112, GSE276113, and GSE285937. The source code used to analyze the data is available on GitHub at https://github.com/hikois/DB24-0762. Further information and requests for resources and reagents should be directed to T.S. or H.K. ([email protected]). All of the study data are provided in this article and the Supplementary Material.

We determined the nature of interaction between macrophages and fibroblasts, with focus on the morphological features of Mincle-expressing macrophages isolated from the epididymal adipose tissue of diet-induced obese mice (Fig. 1A). Flow cytometry detected Mincle-expressing macrophages as a minor subpopulation, accounting for 2.7% of the SVF in obese mice, whereas nonexpressing macrophages contributed ∽37.1% (Fig. 1B). Giemsa staining revealed numerous vacuoles in Mincle-expressing macrophages compared with nonexpressing macrophages (Fig. 1C), and cell-to-cell adhesion rates were higher in Mincle-expressing macrophages (Fig. 1D and Supplementary Fig. 1A). A lipid-rich phenotype similar to LAMs was anticipated, but Mincle-expressing macrophages showed less lipid accumulation than nonexpressing macrophages (Supplementary Fig. 2). Electron microscopy confirmed the adhesive traits of Mincle-expressing macrophages (Fig. 1E). Mincle-expressing macrophages were characterized by round and elongated membrane protrusions (Supplementary Fig. 1B), active endocytosis, enhanced exosome budding (Fig. 1F), and abundant lysosomes and autolysosomes (Supplementary Fig. 1C). Endosome and autolysosome contents were released outside the cells (Supplementary Fig. 1D), whereas these features were not observed in adipose tissue macrophages from lean mice (Fig. 1G) or in those from obese mice without Mincle expression (Fig. 1H). We previously reported that Mincle is selectively expressed in macrophages constituting CLS, which fuse with each other to surround and engulf dead adipocytes, thereby accelerating fibrosis (17). Our histological observations support the notion that Mincle plays a critical role in the cellular interaction within CLS during obesity-induced adipose tissue fibrosis. Thus, we further investigated communications between Mincle-expressing macrophages and fibroblasts using scRNA-seq.

In this study, we used C57BL/6J mice fed a HFD for 8 and 16 weeks to represent the inflammatory and fibrotic phases, respectively, as previously reported (17,33). We validated the mRNA expression of inflammation- and fibrosis-related genes in epididymal adipose tissue (Fig. 2A). Notably, Mincle (Clec4e) expression was significantly elevated only in the fibrotic phase, similar to Tgfb1 and Col1a1, suggesting a role of Mincle in obesity-induced adipose tissue fibrosis. We then performed scRNA-seq on the SVF prepared from the epididymal adipose tissue during the development of obesity (Fig. 2B). Seurat clustering revealed 33 transcriptionally distinct cell clusters (Fig. 2C). Macrophages and fibroblasts were annotated with the predicted cell type information using the SingleR package (31). Macrophages and fibroblasts were defined as clusters expressing Ptprc (Cd45), Adgre1 (F4/80), and Itgam (Cd11b) and as those expressing Pdgfra, Col1a1, and Acta2 (α-Sma), respectively (Fig. 2D and E). The number of macrophages was markedly increased by the HFD load (Fig. 2F). We then conducted subclustering of macrophages and fibroblasts. Macrophages were divided into 11 subclusters, which differed substantially depending on the mice nutritional status (Fig. 2G and H). On the other hand, fibroblasts were divided into seven subclusters; subcluster 3 was predominant in mice with adipose tissue fibrosis (Fig. 2I and J). Fibroblasts in subcluster 3 showed high Midkine (Mdk) and C4b expression (Fig. 2K). Mdk, a proinflammatory heparin-binding growth factor, promotes fibrosis in the lung and kidney (34,35). Pathway analysis of fibroblast characterizing gene sets in subcluster 3 highlighted pathways such as “collagen synthesis and degradation” and “fibrosis” (Fig. 2L), indicating fibroblasts in subcluster 3 as “fibrosis-associated fibroblasts.” Collectively, macrophage and fibroblast subclusters in the adipose tissue differed distinctly between the inflammatory and fibrotic phases.

To investigate the cellular interaction between Mincle-expressing macrophages and fibrosis-associated fibroblasts with potent profibrotic properties, we next performed cellular network analysis using CellChat and found that the intensity and number of interactions between macrophages and fibroblasts markedly increased during the development of obesity (Fig. 3A). In this study, LAMs and Lyve1-expressing perivascular macrophages (36) corresponded to macrophage subclusters 0, 1, 3, 5, 6, 7, and 9 and 8 (Fig. 3A and Supplementary Fig. 3A–C). The intensity of signals involving fibrosis-associated fibroblasts was particularly strong in the fibrotic phase relative to the inflammatory phase. For instance, LAM-corresponding subclusters 1 and 3 showed active interactions with fibrosis-associated fibroblasts via PDGF, VCAM, and TWEAK signaling pathways (Fig. 3A and Supplementary Fig. 3D), which are implicated in the pathogenesis of fibrosis (37).

Mincle-expressing macrophages were abundant in macrophage subclusters 4 and 10, with genetic profiles distinct from LAMs and perivascular macrophages (Supplementary Fig. 3B and C). Mincle expression levels were higher in these subclusters, particularly in the fibrotic phase (Fig. 3B). Top marker genes for subclusters 4 and 10 were characterized by Ccr2, Fn1, and Cd226, and Ceacam1, Cx3cr1, and Ly6c2, respectively (Fig. 3C). These observations suggest that these subclusters are derived from circulating monocytes with proinflammatory and profibrotic properties. Next, we examined the cross talk between Mincle-expressing macrophages and fibrosis-associated fibroblasts during the fibrotic phase. Three pathways, Wnt4, Osm, and Itgb2-Icam1, were identified as potential signals from Mincle-expressing macrophages to fibrosis-associated fibroblasts (Fig. 3D and E). Among these genes, only Osm and its receptor Osmr were predominantly expressed in macrophages and fibroblasts, respectively (Fig. 3F), whereas the expression of other candidates was not restricted to specific cell type, making it technically difficult to investigate their roles as downstream effectors of Mincle-expressing macrophages.

Spatial transcriptomic analysis of the epididymal adipose tissue was performed to investigate the localization of genes identified by CellChat. Of eight clusters identified, clusters 6 and 7 were histologically rich in large vessels, with high expression of endothelial markers (Pecam1, Tek, Vwf, and Cd34) (Fig. 4A–C). Among the remaining six clusters, the population of cluster 2 was markedly increased in obese mice compared with lean mice. Cluster 2 localized to the area containing CLS with increased expression of genes associated with macrophages (Ptprc, Cd68, and Itgax) and fibroblasts (Col1a1 and Tgfb1) (Fig. 4C–E). Notably, Mincle and LAM-specific genes were highly expressed in CLS-rich cluster 2, as described previously (17,38), whereas perivascular macrophages are present in the area containing blood vessels (Fig. 4F, left, and Supplementary Fig. 3E). Among the genes suggested by CellChat (Fig. 3E), Itgb2 and Osm, together with their receptor Osmr, were highly expressed in cluster 2 (Fig. 4F). Since our transcriptomic data include multiple cell types in a single spot, we performed in situ hybridization and confirmed the presence of Mincle-expressing cells in the proximity of Acta2-expressing myofibroblasts, presumably within the regions corresponding to CLS (Fig. 4G). These results suggest that CLS are the sites of cellular interactions between Mincle-expressing macrophages and fibroblasts.

We next examined the mRNA expression of the identified genes by CellChat in peritoneal macrophages stimulated with TDM, a mycobacterial cell wall glycolipid known to be an exogenous Mincle ligand (39). We confirmed that TDM stimulation promotes Mincle expression as reported previously (17), and only Osm expression was significantly increased by TDM treatment (Fig. 5A). Similar results were obtained for the other cultured macrophages (Fig. 5B). In the obese mice, the expression of Mincle and Osm was selectively upregulated in the fibrotic phase, suggesting a role for Osm downstream of Mincle in adipose tissue fibrosis (Fig. 2A). Moreover, Osm and Osmr were mainly expressed in macrophages and CD45-negative cells, including fibroblasts in the SVF, respectively (Fig. 5C and D). Mincle-expressing macrophages sorted from the SVF of obese mice showed higher Mincle and Osm expression than nonexpressing macrophages (Fig. 5E). Consistent with these results, immunostaining revealed that Osm was localized to macrophages forming CLS (Fig. 5F) and that Osmr-positive cells were also localized in the proximity of CLS (Fig. 5G). In addition, MINCLE and OSM expression levels were positively correlated in human adipose tissue (Fig. 5H and I). Taken together, these observations strongly suggest that Osm expression increases downstream of Mincle signaling in adipose tissue macrophages during the development of obesity in mice and humans. Given that Osm secreted from Mincle-expressing macrophages in CLS acts on Osmr-positive fibroblasts, we next examined the functional role of Osm in the pathogenesis of adipose tissue fibrosis.

Because Osm has been reported to play differential roles in the pathogenesis of tissue fibrosis in various organs (40–44), we examined the effect of recombinant Osm on the expression of fibrosis-related genes using adipose tissue–derived fibroblasts (Fig. 6A and B). Different from Tgfβ1, a master regulator of fibrogenesis, stimulation with recombinant Osm did not induce expression of fibrosis-related genes such as Acta2 and Tgfb1 (Fig. 6C and E and Supplementary Fig. 4). On the other hand, Osm stimulation significantly increased mRNA levels of tissue inhibitor of metalloproteinase 1 (Timp1) and decreased those of Col1a1 and Loxl2, a major regulator of collagen cross-linking, in a dose-dependent manner (Fig. 6E). Similar results were obtained in MEFs (Fig. 6D and F). Moreover, Osm potently antagonized the effects of Tgfb1 (Fig. 6G). Next, we examined the effects of conditioned media prepared from Mincle-stimulated macrophages on adipose tissue–derived fibroblasts (Fig. 6H). TDM stimulation markedly increased Osm expression and secretion in RAW264 macrophages (Fig. 6I and J). Treatment with culture media showed expression patterns similar to those of recombinant Osm in adipose tissue–derived fibroblasts (Fig. 6K). Collectively, these results suggest that the Mincle-Osm axis regulates ECM remodeling in adipose tissue fibroblasts. However, whether Osm promotes or suppresses adipose tissue fibrosis in vivo remains unclear.

To investigate the effects of macrophage-derived Osm on adipose tissue fibrosis in vivo, Osm-deficient or wild-type bone marrow cells were introduced into lethally irradiated wild-type mice, which were then rendered obese by HFD feeding (Fig. 7A). There was a slight but significant reduction in epididymal adipose tissue weight in bone marrow cell–specific Osm-deficient mice relative to control mice, whereas the body, subcutaneous adipose tissue, and liver weights did not differ between the groups (Fig. 7B). Serum ALT concentrations were significantly increased, and serum insulin concentrations tended to increase in Osm-deficient mice (Fig. 7C and D). Most gene expression profiles related to inflammation, fibrogenesis, and adipogenesis showed no significant changes between the genotypes; however, mRNA levels of Il1b, Col6a1, and Loxl2 were significantly increased in epididymal adipose tissue from Osm-deficient mice (Fig. 7E). Histological analysis revealed that Osm deficiency slightly increased the fibrotic area without affecting the number of CLS and significantly reduced adipocyte diameters compared with control mice (Fig. 7F–I). Given that Osm reduces Loxl2 expression in adipose tissue fibroblasts, the degree of collagen cross-linking and stiffness may be augmented in Osm-deficient mice. If so, Osm deficiency limits adipose tissue expandability and induces ectopic lipid accumulation in the liver. Taken together, Osm exerts its antifibrotic properties during the development of obesity.

Finally, to examine the role of Mincle-expressing macrophages in obesity-induced adipose tissue fibrosis in vivo, we performed scRNA-seq analysis on the SVF of epididymal adipose tissue from wild-type or Mincle-deficient mice in the fibrotic phase (Fig. 8A). Mincle deficiency did not affect the number of clusters within each cluster (Fig. 8B). We performed functional clustering using the clusters identified in Fig. 2 (Fig. 8C and D). Pseudo-bulk analysis revealed that Mincle deficiency significantly increased mRNA expression of genes related to collagens, regulators of collagen cross-linking, and the Tgfb1 pathway in fibrosis-associated fibroblasts (in subcluster 3) (Fig. 8E). These data suggest that that Mincle-expressing macrophages regulate fibroblasts during obesity-induced adipose tissue remodeling, possibly through the Osm-Osmr pathway.

Recently, scRNA-seq data have provided insights into cell diversity in tissues and organs that exhibit dynamic changes in response to internal and external stimuli. In this study, we used scRNA-seq and spatial transcriptomics to reveal novel cell-to-cell communications between macrophages and fibroblasts in the adipose tissue of diet-induced obese mice. Notably, the interactions between these cells become more intense and complex as obesity develops. We identified the fibrosis-associated fibroblast cluster, which actively interacts with several macrophage subclusters, including LAMs and Mincle-expressing macrophages, in fibrotic adipose tissue. LAMs were further divided into subclusters, some of which predominated in the inflammatory phase and others in the fibrotic phase. In contrast, Mincle-expressing macrophages, localizing to CLS, were a relatively small subpopulation that may be overlooked by conventional bulk RNA-seq. On the basis of our data, CLS are excellent targets for spatial transcriptomics in which adipocyte death–induced complex cellular interactions involving macrophages and fibroblasts drive interstitial fibrosis. Novel spatial transcriptomic methodology at the one-cell level would further elucidate the details of the cellular component of CLS. In this regard, we identified CLS in the liver and kidney in metabolic dysfunction–associated steatohepatitis and acute kidney injury, respectively (32,45,46). Thus, this study serves as a prototype for cellular network analysis of CLS-driven fibrotic pathologies.

Osm belongs to the interleukin-6 cytokine superfamily and plays differential roles in tissue fibrosis, depending on the organ and pathology (47), which may be attributed to the local cytokine milieu. Osm expression in adipose tissue increases with the degree of obesity in mice and humans (48) and adipocyte-specific Osmr deficiency results in accelerated adipose tissue inflammation in obese mice (49). In this study, we demonstrated for the first time that bone marrow cell–specific Osm deficiency in mice enhances obesity-induced adipose tissue fibrosis and suppresses adipocyte expandability, without affecting expression levels of most proinflammatory and profibrotic cytokines in adipose tissue at the organ level. In addition to its suppressive effects on collagen expression, Osm induces mRNA expression of collagen cross-linking genes, such as Loxl2, in both in vitro and in vivo, suggesting that Osm plays a role in regulating both quantity and quality of ECM. In other words, Osm regulates the degree of collagen cross-linking and stiffness; thereby, Osm deficiency could suppress adipocyte hypertrophy. This study is unique in highlighting the qualitative changes of fibrosis, beyond the qualitative changes that have been the focus to date. Development of novel methodologies to evaluate the qualitative changes of tissue fibrosis will provide insight into the dynamic process of obesity-induced adipose tissue remodeling.

We now discuss the similarities and differences between Mincle-expressing macrophages and LAMs. Both cell types are localized to CLS in adipose tissue. While the role of LAMs has been extensively studied (9,38), Mincle-expressing macrophages remain less thoroughly investigated. These macrophages share morphological features, such as increased number of lysosomes and protrusions, whereas lipid droplets are minimal in Mincle-expressing macrophages (26). In contrast, both cell types were distinct with respect to their gene expression profiles, and Mincle-expressing macrophages must have different functional roles in adipose tissue fibrosis. Given that LAMs are numerous and actively interact with fibrosis-associated fibroblasts, LAMs may play a major role as effectors in the pathogenesis of adipose tissue fibrosis. Besides fibrosis-associated fibroblasts, we noted that Mincle-expressing macrophages communicate with LAMs, which is consistent with their close localization in CLS. In this study, we identified Osm as a novel downstream target of Mincle signaling. Since we found that Osm is a negative regulator of obesity-induced adipose tissue fibrosis, the mechanism underlying Mincle-mediated profibrotic effects was not fully elucidated in this study; thus, further investigation of the interaction of Mincle-expressing macrophages with LAMs and other cell types is required to fully understand the mechanism of Mincle-mediated adipose tissue fibrosis.

We need to discuss how CLS are involved in obesity-induced adipose tissue fibrosis. Tissue fibrosis is a dynamic and complex process in which degradation of ECM occurs in parallel with its synthesis. Indeed, Tgfb, the master regulator of fibrogenesis, induces gene expression of matrix metalloproteases, which degrade ECM. Thus, active tissue remodeling is assumed to occur around CLS. We previously reported that Mincle plays a key role in promoting obesity-induced adipose tissue fibrosis, which may be attributed to the induction of Tgfb and Pdgfb (16,17). In this study, we shed light on another aspect of the Mincle-mediated effects on tissue fibrosis: macrophage-derived Osm may play a role in fine-tuning obesity-induced adipose tissue remodeling around CLS, likely through suppression of collagen expression and cross-linking.

This study has some limitations. A causal role of Mincle-regulated Osm in adipose tissue fibrosis was not clarified, because other cell types may also be responsible for Osm production. Further studies involving Mincle-expressing cell-specific Osm-deficient mice are required. This study demonstrates a positive correlation between the expression of MINCLE and OSM in human adipose tissue. Exploring the association between adipose tissue fibrosis and the expression of these genes to understand full clinical implications would be valuable.

In summary, through scRNA-seq and spatial transcriptomics, we identified novel cell-to-cell communications among macrophages and fibroblasts in the adipose tissue of diet-induced obese mice (Supplementary Fig. 5). Spatial transcriptomic analysis is useful for understanding the complex cellular cross talk in CLS, a distinctive histological feature of adipose tissue in obesity, driving fibrosis. Osm secreted from Mincle-expressing macrophages is involved in dynamic ECM remodeling in the proximity of CLS. Considering that CLS are also observed in other organs, such as liver and kidney, the analytical approaches used in this study may be applicable to other chronic inflammatory conditions.

Acknowledgments. The authors thank Dr. Yutaka Inagaki (Tokai University) for a generous gift of the Col1a2/EGFP-transgenic mice, Akinori Osada and Ryosuke Ishikawa (Graduate School of Medicine, Nagoya University) for technical assistance, Chika Matsui (Research Core, Institute of Research, Tokyo Medical and Dental University) for assistance with electron microscopic analysis, Tatsuro Ogawa (Tokyo University of Science) and Bin Wu (ImmunoGeneTeqs) for assistance with handling the scRNA-seq data, CyberomiX (Kyoto, Japan) for assistance with the spatial transcriptomic analysis, Editage for English language editing, and Center for Animal Research and Education (CARE), Nagoya University for support on animal experiments. The authors also thank the members of the Suganami laboratory for their helpful discussions.

Funding. This work was supported in part by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan (24K10076, 23K27377, 23H04776, 23K16815, 22K19524, and 22K19723), and Japan Agency for Medical Research and Development (CREST [JP24gm1210009s0106], Research Program on Hepatitis [JP24fk0210154s0501], and Research Program on Rare and Intractable Diseases). This study was also supported by research grants from SEI Group CSR Foundation, SECOM Science and Technology Foundation, The Hori Science & Art Foundation, Uehara Memorial Foundation for Life Sciences, Takeda Science Foundation, Suzuken Memorial Foundation, Harmonic Ito Foundation, and Foundation of Public Interest of Tatematsu. This work was also supported by “Quantum-Based Frontier Research Hub for Industry Development” and “Innovative Research Center for Preventive Medical Engineering,” Nagoya University, Nagoya, Japan. H.K. was supported by Nagoya University CIBoG (Convolution of Informatics and Biomedical Sciences on Global Alliances) WISE program (Doctoral Program for World-leading Innovative & Smart Education) from Ministry of Education, Culture, Sports, Science and Technology.

Duality of Interest. S.Shic., S.U., and K.M. have stocks of ImmunoGeneTeqs, Inc. S.Shic. has an advisory role with ImmunoGeneTeqs, Inc. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. H.K. performed most of the experiments, analyzed data, and generated the figures. H.K., M.T., and T.S. designed research and wrote the manuscript. M.T., E.W., and K.O. performed bone marrow transplantation experiments. S.Shic., S.U., and K.M. performed scRNA-seq analysis. S.A. and S.Shim. performed electron microscopic analysis. T.K. and Y.Mo. did animal experiments involving Osm-deficient mice. A.I. and Y.O. contributed to critical discussion. X.Y., T.T., Y.Mi., and A.E. performed histological analysis. A.S., I.T., and K.I. provided and analyzed human samples. T.S. supervised the whole study. All authors approved the final version of the manuscript and agreed to be accountable for all aspects of the work, including any questions related to the accuracy or integrity of any part of the work. T.S. and M.T. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.