BAF60a Deficiency in Macrophage Promotes Diet-Induced Obesity and Metabolic Inflammation

Insulin resistance (IR) and pancreatic β-cell dysfunction are the major causes of type 2 diabetes. Obesity-associated metabolic dysfunction of adipose tissue plays an important role in the pathogenesis of IR and type 2 diabetes (1). Accumulating evidence suggests that cross talk between metabolic and immune cells is critical for the maintenance of systemic energy homeostasis and metabolic balance in adipose tissues (2). While adipose tissue–resident immune cells are essential to maintain normal adipocyte expansion and function, overnutrition-induced aberrant activation of innate immune system and chronic low-grade inflammation are closely linked to the metabolic dysfunction of adipose tissue (2–4). In fact, persistent inflammation recurrence drives adipose tissue fibrosis and necrosis and impairs angiogenesis, thereby leading to adipose tissue dysfunction and IR (5).

Obesity-induced chronic inflammation in adipose tissue is characterized by increased expression of proinflammatory cytokines and chemokines, such as TNF-α, IL-1β, chemokine (C-C motif) ligand 5 (Ccl5), and Ccl2, in which the interaction between adipose tissue macrophage (ATM) and microenvironment in adipose tissue plays a critical role (6–8). ATMs are actively involved in regulating metabolic homeostasis and the immune microenvironment in adipose tissue under both physiological and pathophysiological conditions. Upon diverse stimuli, ATMs can undergo different polarization, resulting in two distinct phenotypes: proinflammation (M1-like, classical activation) and anti-inflammation (M2-like, alternative activation) (9). M1-like macrophages are normally activated by T helper (Th)1-derived cytokines such as interferon-γ (IFN-γ) and are characterized by the overproduction of proinflammatory cytokines like TNF-α and IL-6, which could block insulin action in adipocytes. In contrast, M2-like macrophages are induced by Th2-derived cytokines such as IL-4/IL-13 and protect adipocytes from inflammation and damage (10). It is known that the M1- and M2-like polarizations of ATM keep equilibrium in normal physiological conditions, whereas monocytes are recruited into adipose tissue and inducing a proinflammatory phenotype of ATM during obesity (11,12). In addition, previous study has demonstrated that ATM may possess a metabolically activated (MMe) phenotype that is distinct from classical M1/M2 activation (13). Despite the growing understanding of the role of ATM in controlling the immune microenvironment in adipose tissue, how ATM integrates the nutritional and environmental cues to switch on proinflammatory response, to lead to adipose tissue metabolic inflammation and IR, remains elusive.

Previous study has demonstrated that epigenetics mechanisms are essential for metabolic homeostasis, since they serve important roles in the pathogenesis of obesity and diabetes through integrating the environmental factors to the transcriptional regulation of gene programs (14). ATP-dependent chromatin remodeling complexes regulate the biological function of target cells through selectively orchestrating the chromatin accessibility and transcriptional activity of target genes in response to diverse environmental and metabolic stimuli (15,16). Among the four major classes of chromatin remodeling complexes in mammals, switch/sucrose-nonfermentable (SWI/SNF) complex has been shown to play critical roles in diverse physiological and pathophysiological processes. SWI/SNF complex is composed of catalytic ATPase subunits Brg1 or Brm and multiple noncatalytic homologous subunits termed as Brg/Brm-associated factors (BAFs) (15,16). The BAF60 family members, including BAF60a, BAF60b, and BAF60c, can recruit the SWI/SNF complex to target genes by interacting with context-specific transcription factors (17). Our recent studies have demonstrated that BAF60a is highly expressed in liver and adipose tissue and is mainly involved in regulating hepatic fatty acid oxidation, cholesterol homeostasis, and adipose tissue thermogenesis (18–21). BAF60c is preferentially expressed in skeletal muscle, acting as an important driver of glycolytic myofiber specification and glucose metabolism (22–25). However, the role of BAF60 proteins in obesity-induced macrophage activation and chronic inflammation in adipose tissue remains unclear.

In this study, we found that accumulation and activation of macrophages in adipose tissue from obese and type 2 diabetic mice are associated with markedly lower expression of BAF60a in stromal vascular fractions (SVFs). Using myeloid-specific BAF60a knockout (BaMKO) mice, we demonstrated that BAF60a deficiency promotes metabolic inflammation in both epididymal white adipose tissue (eWAT) and inguinal white adipose tissue (iWAT). Such inflammation in these adipose tissues exacerbates high-fat diet (HFD)-induced obesity and leads to impairment of systemic insulin sensitivity and glucose homeostasis in mice. Peritoneal macrophages (PMs) derived from BaMKO mice exhibit M1-like polarization and overexpression of proinflammatory cytokines compared with controls. In contrast, PMs derived from myeloid cell–specific BAF60a knock-in (BaMKI) mice display inhibition of M1-like polarization and anti-inflammatory phenotype. In accordance, BAF60a deficiency in mice triggers the infiltration of M1-like macrophages in the SVFs of both eWAT and iWAT, leading to transcriptional induction of proinflammatory gene programs. In macrophages, further integrative analyses of BAF60a-dependent chromatin accessibility and BAF60a genome-wide DNA binding profile with Assay for Transposable-Accessible Chromatin with high-throughput sequencing (ATAC-Seq) and CUT&Tag sequencing (CUT&Tag-Seq), respectively, reveal that BAF60a can directly regulate a cluster of genes involved in macrophage activation through orchestrating the chromatin landscape surrounding target genes. Mechanistically, BAF60a physically interacts with transcription factor Aft3 to synergistically repress the transcription of proinflammatory gene in macrophages. These findings identify BAF60a as a pivotal checkpoint factor that can integrate the local nutritional and environmental cues to transcriptional activation of M1-like macrophage through chromatin remodeling. This work also unveils Atf3 as a critical regulator of proinflammatory macrophage through its interaction with BAF60a.

BAF60a and BAF60c flox/flox mice on the C57BL/6J background were generated as previously described (18,24). For the generation of BAF60b flox/flox mice, a targeting vector that carries two loxP sites flanking exons 3–5 of mouse BAF60b was constructed using bacterial artificial chromosome recombineering, followed by standard homologous recombination into C57BL/6 embryonic stem cells by Cyagen Biosciences (Suzhou, China). Atf3 flox/flox mice on the C57BL/6J background were generated by GemPharmatech (Nanjing, China). Two loxP sites were inserted flanking exons 2–3 of mouse Atf3 using bacterial artificial chromosome recombineering. To obtain myeloid-specific BAF60a overexpression (BaMKI) mice, we first generated BAF60a knock-in (BaKI) mice using an approach similar to that used to generate Cas9 knock-in mice by GemPharmatech (Nanjing, China) as previously described (26). Briefly, Flag/HA-mouse BAF60a-P2A-GFP expression cassette was synthesized to replace the 3× FLAG-NLS-SpCas9-NLS-P2A-EGFP expression cassette in the LSL-Cas9-Rosa26TV targeting vector, a gift from Dr. Feng Zhang of Broad Institute of MIT and Harvard (Addgene plasmid no. 61408 [https://n2t.net/addgene:61408; RRID: Addgene_61408]). The targeting vector was further verified by sequencing before application for the generation of BaMKI mice through standard homologous recombination. Correctly targeted embryonic stem cell colonies were injected into blastocysts on C57BL/6N background for generating chimeric mice. The germline transmitted founders were further backcrossed with mice on C57BL/6J background for at least six generations for metabolic studies. Lyz2-Cre (no. 004781), leptin receptor–deficient db/db mice were purchased from The Jackson Laboratory.

Mice were housed in 12/12 h light/dark cycles at an ambient temperature of 23°C and fed with standard chow diet or HFD (D12492; Research Diets). All experiments were performed on male mice unless otherwise indicated. After experiments, mice were sacrificed and tissues were immediately frozen in liquid nitrogen before RNA and protein analyses. All animal studies were performed according to procedures approved by the University Committee on Use and Care of Animals at Zhejiang University.

Mouse bone marrow–derived macrophages (BMDMs) were generated as previously described (27). Mouse PMs were collected 3 days after thioglycolate (Millipore) injection. Human embryonic kidney (HEK)293T cells, PMs, and RAW264.7 cells were cultured in DMEM (11995065; Gibco) supplemented with 10% FBS (SE100-011; Visual Technologies Inc.), 50 μg/mL streptomycin, and 50 units/mL penicillin (15140122; Gibco). BMDMs were grown in DMEM containing 10% FBS, 10 ng/mL macrophage colony-stimulating factor (M-CSF) (315-02-50; PeproTech), 50 μg/mL streptomycin, and 50 units/mL penicillin. All cells were cultured and maintained at 37°C in a 5% CO₂ incubator. For stimulation, BMDM or PMs were cultured in DMEM medium and then treated with lipopolysaccharides (LPS) (100 ng/mL) for 4 h or IL-4 (20 ng/mL) for 24 h before subsequent analysis. Cell lysates and supernatants were analyzed with Western blot, quantitative PCR (qPCR), and ELISA.

The adipose tissue was excised from the mouse and minced with scissors. The tissue was then digested for 40 min in a solution containing collagenase D (2 mg/mL) and Dispase II (1.5 mg/mL) at 37°C with constant agitation. After digestion, the tissue was filtered through a nylon mesh (70 μm) and centrifuged at 1,200 rpm for 10 min. The filtered cells were allowed to separate with the adipocytes floating on top of the buffer. The SVFs were resuspended in red blood cell lysis buffer for 5 min at room temperature before further analysis. Finally, the cells were washed, centrifuged, and resuspended for the subsequent analysis.

Supernatants from cell cultures and plasma were collected, and the concentrations of TNF-α and IL-1β were determined with ELISA assays according to the manufacturer’s instructions (eBioscience).

Freshly dissected murine adipose tissues were fixed in 4% paraformaldehyde (P6148; Sigma-Aldrich) for 24 h at 4°C and embedded with paraffin, and hematoxylin-eosin (H-E) staining was performed. The crown-like structures (CLS) number and adipocytes area were quantified with ImageJ software (National Institutes of Health). At least four biological replicates were used in each group. For each biological replicate four randomly selected sections were included for analysis.

The adipose tissues were embedded in Tissue-Tek optimum cutting temperature (OCT) compound and sliced into sections with 20-μm thickness. After fixed with precold 4% paraformaldehyde in PBS for 10 min at 4°C, slides were washed 5 min in PBS with 0.1% Tween-20. Samples were permeabilized with 0.3% Triton X-100 for 15 min. After washing twice with PBS, sections were blocked with 10% BSA [36106ES25; Yeasen Biotechnology (Shanghai) Co., Ltd.] in PBS for 1 h at room temperature and then stained with primary antibodies F4/80 (1:200, no. 123116; BioLegend) overnight at 4°C. Next day, slides were washed with PBS three times and then incubated with Alexa Fluor–conjugated secondary antibodies (1:200; Invitrogen) for 1.5 h at room temperature. The sections then were washed with PBS followed by DAPI staining (1:2,000, P3693l; Invitrogen). Sections were then mounted with Fluoromount-G mounting medium [36307ES08; Yeasen Biotechnology (Shanghai) Co., Ltd.] and sealed with neutral resin. Images were acquired with use of the ZEISS LSM 800. At least four biological replicates were used in each group.

Glucose tolerance tests (GTT) and insulin tolerance tests (ITT) were performed as previously described (22). For GTT, mice fasted overnight (∼16 h) and subsequently injected with glucose solution in saline (2 g/kg body wt i.p.). Blood glucose was determined before and 15, 30, 60, and 120 min after glucose injection. For ITT, mice fasted for 4 h and injected with insulin solution in saline (1.8 units/kg body wt i.p.). Blood glucose was measured before and 15, 30, 60, and 120 min after insulin injection.

Adipose tissues, SVFs, or whole cells were lysed and quantified. Protein samples were separated with SDS-PAGE and transferred onto a polyvinylidene difluoride membrane (Millipore), followed by immunoblotting with the following primary antibodies: BAF60a (1:1,000, no. 611728; BD Biosciences), BAF60b (1:1,000, customized; Abcam), Atf3 (1:1,000; ab254268; Abcam), β-actin (1:1,000, A4700; Sigma-Aldrich), and Myc (1:1,000, sc-40; Santa Cruz Biotechnology). Horseradish peroxidase (HRP)–conjugated goat anti-mouse (1:5,000, A4416; Sigma-Aldrich) and goat anti-rabbit (1:5,000, A6154; Sigma-Aldrich) were used as secondary antibodies. For immunoprecipitation (IP), a small portion of the protein lysate was taken as input and the remaining part incubated with anti-HA agarose (26182; Thermo Fisher Scientific) or anti-Myc agarose (A7470; Sigma-Aldrich) on a rotator at 4°C overnight.

For endogenous co-immunoprecipitation (co-IP) assay of BAF60a and Atf3, the cell lysate was incubated with anti-Atf3 (1:200, no. ab254268; Abcam) overnight and protein A/G agarose [36403ES08; Yeasen Biotechnology (Shanghai) Co., Ltd.] for 3 h on a rotator at 4°C. The IP samples were washed with IP wash buffer five times and boiled with 1× loading buffer for immunoblotting analysis.

Total RNA was extracted from adipose tissue and SVF with a commercial kit (DP430; TIANGEN Biotech). RNA from other tissues and cultured cells was extracted with use of TRIzol (Invitrogen) reagent. Normalized RNA was reverse transcribed with the High-Capacity cDNA Reverse Transcription Kit (R222-01; Vazyme) and cDNA was analyzed with real-time qPCR through the Roche LightCycler 480 System. Relative gene expression levels were expressed as ratios relative to internal 36B4 mRNA levels, and the qPCR primers used are listed in Supplementary Table 1.

Cell preparation for flow cytometry was performed as previously described (27). Cells were suspended in FACS buffer (0.5% BSA, 2 mmol/L EDTA). Recovered SVFs were washed and incubated with the desired combination of fluorochrome-conjugated antibodies, including Fixable Viability Dye eFluor 450 (eBioscience), APC/Cyanine7 anti-mouse CD45 Antibody (clone 30-F11; BioLegend), APC anti-mouse F4/80 Antibody (clone BM8; BioLegend), FITC anti-mouse/human CD11b Antibody (clone M1/70; BioLegend), PE anti-mouse CD11c Antibody (clone N418; BioLegend), and Purified anti-mouse CD206 (MMR) Antibody (clone C086C2; BioLegend). Validation information is available on the manufacturers’ websites. Cells were then subjected to flow cytometry analysis with BD LSRFortessa (BD Biosciences). Data were analyzed with FlowJo software, version 10.3.1.

For RNA Sequencing (RNA-Seq) analysis, total RNA samples were sent for library preparation and sequencing by BGI group (Wuhan, China). Data were processed following the standard BGI mRNA analysis pipeline. Briefly, raw reads were filtered and trimmed with SOAPnuke (version 1.4.0) and Trimmomatic (version 0.36) (parameters: -l 5 -q 0.5 -n 0.1 for SOAPnuke and ILLUMINACLIP:2:30:10 LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:50 for Trimmomatic). Filtered clean data were then aligned to mm10 reference genome (Mus_musculus [GRCm38.90]) with HISAT2 (version 2.1.0) under recommended parameters (--dta --phred64 unstranded --new-summary -x index −1 read_r1–2 read_r2(PE)). Expression level of mRNA was transferred as fragments per kilobase of transcript per million mapped reads for statistical analysis performed with DESeq2 (version 1.20.0) (28). Gene Ontology (GO) and pathway grouping and enrichment studies were performed with clusterProfiler (version 3.12.0), and pathway visualization was conducted with pathview (version 1.26.0) (29,30).

Nuclei from F4/80⁺ SVF cells were isolated as previously described (31). In each transposase reaction (Ilumina), 5 × 10⁴ nuclei were used, followed by barcoding and library preparation. HiSeq 4000 instrument with 150-bp paired-end reads was used for DNA sequencing. The reads were trimmed to 38-bp paired-end reads with fastx_trimmer (http://hannonlab.cshl.edu/fastx_toolkit) for further processing. Downstream analysis pipeline was adapted from previous study (32). Briefly, MACS2 (2.1.1.20160309) package was used for broad peaks calling with the parameter of (--nomodel --shift -100 --extsize 200 -B --broad) and differential accessed peaks were called with DESeq2 (version 1.20.0) (28,33). Motif enrichment analysis and peak-associated-gene annotation were performed with HOMER (version 4.10) with use of peaks filtered with corresponding criteria mentioned previously (34). Browser tracks were visualized with Integrative Genomics Viewer (IGV) (version 2.4.14) after normalizing of the reads from each individual sample to its own library size (35). ataqv (version 1.0.0) package developed by the Parker laboratory from the University of Michigan was used to perform the ATAC-Seq data quality control analysis.

The library preparation for CUT&Tag-Seq was performed as previously reported (36). In brief, ∼1 × 10⁵ single cells were washed twice in 1 mL PBS. Cells were incubated with 10 µL activated concanavalin A–coated magnetic beads (BP531; Bangs Laboratories) in 700 µL Wash Buffer (20 mmol/L HEPES, pH 7.5, 150 mmol/L NaCl, 0.5 mmol/L spermidine, and protease inhibitor cocktail; Roche) for 10 min at room temperature. Cell-bound beads were collected and resuspended with 50 µL Dig-wash buffer (20 mmol/L HEPES, pH 7.5, 150 mmol/L NaCl, 0.5 mmol/L spermidine, protease inhibitor cocktail, 0.05% digitonin) containing 2 mmol/L EDTA, 0.1% BSA, and a 1:50 dilution of the primary antibody (rabbit anti-SMARCD1, no. HPA004101, Sigma-Aldrich, and normal IgG, 2729, Cell Signaling Technology) and incubated at 4°C overnight. Secondary antibody (goat anti-rabbit IgG, SAB3700883; Sigma-Aldrich) diluted at 1:50 in 100 µL Dig-wash buffer was then administered into the beads and incubated for 60 min at room temperature following primary antibody removal with magnet stand (CM101; Vazyme). The preparation of pG-Tn5 adapter complex was performed according to the manufacturer’s instruction with Hyperactive pG-Tn5 Transposase for CUT&Tag (S602; Vazyme). Standard tagmentation and amplification were performed as previously described (36). Amplified DNA libraries were purified with VAHTS DNA Clean Beads (N411; Vazyme) and shipped for next-generation sequencing by Annoroad Gene Technology (Beijing, China). CUT&Tag-Seq data were analyzed with a similar protocol adapted from the pipeline developed by Ye Zheng of Vaccine and Infectious Disease Division, Fred Hutchinson Cancer Research Center, Seattle, WA (https://yezhengstat.github.io/CUTTag_tutorial/).

Statistical analyses were carried out with GraphPad Prism 8. Statistical differences were evaluated with two-tailed unpaired Student t test for comparisons between two groups or ANOVA and appropriate post hoc analyses for multigroup comparisons. For GTT and ITT, two-way ANOVA with multiple comparisons was used for statistical analysis. A P value <0.05 (*P < 0.05; **P < 0.01; and ***P < 0.001) was considered statistically significant. No statistical methods were used to predetermine sample size. The experiments were not randomized, and the investigators were not blinded to allocation during experiments or outcome assessment.

The data sets generated during or analyzed during the current study are available from the corresponding authors on reasonable request.

Energy is stored in the form of triglycerides in white adipose tissues and can affect the health status of individuals according to its differential distribution in vivo, thereby causing various metabolic diseases such as obesity and type 2 diabetes (37). Previous study has suggested that subcutaneous white adipose is capable of browning under cold exposure or exercise through the activation of macrophages to further prevent the occurrence of obesity and diabetes (38). To understand the role of BAF60s in the progress of macrophage polarization during the development of obesity and type 2 diabetes, we first assessed the expression of BAF60s in adipose tissues and BMDM obtained from WT mice. The results showed that BAF60a and BAF60b were highly expressed in white adipose tissues (eWAT and iWAT), adipose tissue SVFs, and BMDMs, while BAF60c expression was much lower (Fig. 1A–C). As such, we focused on the function of BAF60a and BAF60b in this study. Gene expression analyses of BAF60s expression in adipose tissues from db/db and control mice showed that mRNA levels of BAF0a and BAF60b were similar between the two groups (Supplementary Fig. 1A and B).

To specifically examine the expression of BAF60s in mature adipocyte and SVF in adipose tissue, we separated SVF and adipocyte in adipose tissue. As revealed with qPCR analysis, the purity of mature adipocyte and SVF was confirmed by the successful expression of mature adipocyte (Supplementary Fig. 1C) and SVF (Supplementary Fig. 1D) marker genes, respectively. Interestingly, BAF60a protein expression in SVFs from both iWAT and eWAT was robustly decreased in db/db mice compared with their littermate controls, whereas protein levels of BAF60b remained similar between the two groups (Fig. 1D and E). As expected, the mRNA expression of F4/80, a marker gene of macrophage, in SVFs from both iWAT and eWAT was markedly elevated in db/db mice compared with controls (Fig. 1F and G). Consistent with the protein expression levels, BAF60b mRNA expression was not altered in SVFs from iWAT or eWAT in db/db mice compared with controls (Supplementary Fig. 1E and F). While the mRNA levels of BAF60a in SVF from eWAT were significantly reduced in db/db mice compared with controls, its mRNA expression in SVFs from iWAT was trending higher (Supplementary Fig. 1E and F), suggesting that posttranscriptional mechanisms might be involved in regulating BAF60a expression in SVFs. Previous studies have demonstrated that elevations of LPS, lipids, and inflammatory cytokines such as TNF-α in the circulation play important roles in the pathogenesis of obesity-associated IR and type 2 diabetes (39–41). To dissect the upstream signals that trigger the downregulation of BAF60a in macrophages, we treated the RAW264.7 cells with LPS, palmitic acid (PA), and TNF-α and found that LPS (100 ng/mL) robustly decreased both mRNA and protein expression of BAF60a (Fig. 1H and I), whereas treatment with PA (0.5 mmol/L) and TNF-α (50 ng/mL) elicited a modest downregulation of BAF60a protein expression in RAW264.7 cells (Supplementary Fig. 1G). These results suggest that the decreased BAF60a protein expression in macrophages, downstream of metabolic endotoxemia, inflammatory signaling, and lipid toxicity, may be involved in the recruitment and proinflammatory activation of ATM in adipose tissue during the pathogenesis of type 2 diabetes.

To test the role of BAF60a in macrophage in vivo, we generated BaMKO mice by crossing the BAF60a flox/flox mice with Lyz2-Cre mice (Fig. 2A and Supplementary Fig. 2A). As shown in Fig. 2B and C, mRNA and protein levels of BAF60a, but not BAF60b and BAF60c, were markedly decreased in BMDM from BaMKO mice compared with those from BAF60a flox/flox mice. Under chow diet feeding condition, body weight and fasting blood glucose were similar between the two groups (Supplementary Fig. 2B and C). We next subjected the mice to HFD feeding and, interestingly, observed that BaMKO mice gained significantly more body weight (Fig. 2D). Consistently, body composition analysis revealed that BaMKO mice have higher fat mass compared with controls, while lean mass was similar between two groups (Fig. 2E).

We next went on to measure the metabolic parameters in HFD-fed BaMKO mice. Fasting blood glucose (Fig. 2F) and plasma insulin (Fig. 2G) levels were significantly elevated in BaMKO mice compared with controls, whereas plasma concentrations of total cholesterol, glycerol, and triglyceride were indistinguishable between the two groups (Fig. 2H and J). These results suggest that inactivation of BAF60a in myeloid cells may selectively affect systemic insulin sensitivity and glucose homeostasis in mice. Consistent with this hypothesis, further GTT and ITT assays revealed that BaMKO mice exhibited impaired glucose tolerance and insulin sensitivity compared with control mice (Fig. 2K and L). Together, these data demonstrate that conditional inactivation of BAF60a in myeloid cells promotes HFD-induced obesity, IR, and glucose intolerance in mice.

It is well-known that accumulation and activation of ATM in adipose tissue play an important role in pathogenesis of obesity-associated chronic metabolic inflammation, IR, and impaired glucose homeostasis. In fact, ATM account for <10% of the cell population in adipose tissue under normal healthy conditions but can increase up to 40% of cells in obese adipose tissue (42,43). Previous study has shown that proinflammatory macrophages are clustered around the damaged or necrotic adipocytes to form CLS, which is the classical histological feature of local inflammation in adipose tissue in the late stage of obesity (11).

Notably, H-E staining of adipose tissue section revealed that the numbers of CLS were robustly increased in iWAT (Fig. 3A and B) and eWAT (Fig. 3C and D) from HFD-fed BaMKO mice compared with those from control mice. Such histological alteration was accompanied by upregulation of proinflammatory genes in adipose tissues of BaMKO mice as revealed in qPCR analysis (Supplementary Fig. 3A). The adipocyte size and tissue weight were similar in iWAT between the two mice groups (Fig. 3E and F), while a marked decrease in these two indices was observed in eWAT from HFD-fed BaMKO mice (Fig. 3G and H). Further immunofluorescent staining also demonstrated enrichment of F4/80⁺ macrophages in iWAT from BaMKO mice compared with controls (Fig. 3I and J). Moreover, liver weight was trending higher in HFD-fed BaMKO mice compared with controls, while tissue weights of quadriceps muscle and brown adipose tissue (BAT) were similar between the two groups (Supplementary Fig. 3B). Measurement of liver triglyceride content and H-E staining of liver section revealed a significant elevation of triglyceride and more lipid accumulation, respectively, in livers from HFD fed-BaMKO mice compared with controls, suggesting more severe liver steatosis in HFD fed-BaMKO mice (Supplementary Fig. 3C and D). However, expression levels of proinflammatory genes remained largely unchanged in liver and BAT between two groups (Supplementary Fig. 3E and F). These results indicate a tissue-specific effect of BAF60a on inflammatory response in metabolic tissues and that BaMKO may promote HFD-induced obesity and IR through recruitment and activation of proinflammatory macrophages selectively in white adipose tissue.

Accumulating evidence suggests that polarization of ATM from anti-inflammatory phenotype (M2-like) to proinflammatory phenotype (M1-like) plays a central role in the pathogenesis of obesity-induced adipose tissue inflammation, IR, and type 2 diabetes (10,44–46). To examine the effects of BaMKO on macrophage recruitment and polarization in white adipose tissues, we performed flow cytometry analysis on SVFs from iWAT and eWAT of control and BaMKO mice. Interestingly, the proportion of CD11b⁺F4/80⁺CD11c⁺ cells (M1-like macrophages) in SVFs from both iWAT and eWAT depots was significantly higher in BaMKO mice compared with controls, while there was no significant difference in the proportion of CD11b⁺F4/80⁺CD206⁺ cells (M2-like macrophages) between the two groups (Fig. 4A and B). These results suggest that BAF60a deficiency in myeloid cells aggravates obesity-induced adipose tissue inflammation through recruitment and polarization of proinflammatory M1-like macrophages in response to HFD.

For exploration of the effects of BAF60a inactivation on the transcriptional profiles of macrophages in white adipose tissue, ATM were enriched with anti-F4/80 magnetic beads in iWAT-SVF and eWAT-SVF from HFD-fed control and BaMKO mice for subjection to total RNA isolation and RNA-Seq analysis. Results of heat map analysis of differentially expressed genes showed a remarkable upregulation of genes encoding proinflammatory cytokines and chemokines in BaMKO F4/80⁺ cells from both iWAT-SVF (Fig. 4C) and eWAT-SVF (Fig. 4D). GO analysis of these differentially expressed genes suggested that the most significantly upregulated biological processes were related to myeloid leukocyte activation and immune and inflammation response (Fig. 4C and D). Further qPCR assays also confirmed a robust transcriptional induction of M1-like proinflammatory cytokines such as IFN-γ, IL12p40, IL-1β, iNOS, Ccl5, Ccl2, and IL-6 in iWAT-SVF and eWAT-SVF from HFD-fed BaMKO mice, while such induction in cytokines was absent in controls (Fig. 4E and F).

To further examine whether BAF60a regulates macrophage polarization and activation in a cell-autonomous manner, we isolated primary PMs from control and BaMKO mice followed by treatment with LPS or IL-4 for qPCR analysis of the proinflammatory cytokines mRNA expression. Consistent with the gene expression profiles from iWAT-SVF and eWAT-SVF, mRNA expression of several proinflammatory cytokines including IL-1β, Ccl5, Ccl2, IL-6, and TNF-α was significantly elevated in PMs from BaMKO mice compared with those from control mice. Conversely, mRNA expression of anti-inflammatory cytokines such like Mrc1, Mgl1, IL-10, Ym1, and Arg1 was decreased in BaMKO mice (Fig. 4G and H). Moreover, ELISA assays revealed that PMs from BaMKO mice exhibited a marked increase in the production and secretion of TNF-α and IL-1β in the culture media following LPS and ATP stimulation (Fig. 4I and J). Additionally, the BMDM from BaMKO mice showed a similar robust increase in M1-related genes and a significant decrease in M2-related genes on LPS or IL-4 stimulation. (Supplementary Fig. 4A and B).

To further assess the effects of BAF60a gain of function on macrophage polarization and activation, we generated BaMKI mice by crossing the BaKI mice with Lyz2-Cre mice (Supplementary Fig. 5A). Successful overexpression of BAF60a mRNA and protein levels in myeloid cells was validated with qPCR and Western blot analyses, respectively, of BMDMs derived from control and BaMKI mice (Supplementary Fig. 5B and C). In contrast to BAF60a inactivation, BAF60a overexpression in PMs from BaMKI mice displayed a markedly decreased expression of M1-like proinflammatory genes such as IFN-γ, IL-1β, iNOS, and Ccl5 and an increased expression of M2-like anti-inflammatory genes such as Arg1 and IL-10 (Supplementary Fig. 5D and E). Moreover, we assessed the mRNA expression levels of CD11c and CD206 in SVFs from eWAT and iWAT. As shown in the Supplementary Fig. 5F, the gene expression of CD206 was trending higher in iWAT-SVF from BaMKI mice, while the expression of CD11c was significantly decreased in eWAT-SVF from BaMKI mice compared with controls. In accordance, histological analysis of iWAT and eWAT showed that the numbers of CLS were significantly reduced in both fat depots from BaMKI mice (Supplementary Fig. 5G). Together, these data demonstrate that BAF60a cell-autonomously regulates the transcriptional programs associated with macrophage polarization and activation. BAF60a deficiency promotes the polarization of ATM toward M1-like proinflammatory phenotype through transcriptional induction of proinflammatory genes. Conversely, overexpression of BAF60a reverses the polarization shift to M2-like anti-inflammatory phenotype via downregulation of proinflammatory genes and upregulation of anti-inflammatory genes.

Previous study has demonstrated that BAF60 family proteins function as important subunits of SWI/SNF complex to sense the local nutritional and environmental cues to alter the chromatin accessibility and transcriptional activity through interacting with context-dependent transcription factors (17). To further investigate whether and how BAF60a can directly regulate the transcription of proinflammatory cytokines through chromatin remodeling, we performed ATAC-Seq analysis on nuclei isolated from iWAT-SVF F4/80⁺ cells and eWAT-SVF F4/80⁺ cells from controls and BaMKO mice, both under HFD condition (Fig. 5A). We obtained 1,009 peaks and 1,267 peaks that are significantly altered in iWAT-SVF F4/80⁺ and eWAT-SVF F4/80⁺ cells, respectively, from BaMKO mice compared with those from control mice. We annotated these peaks to their closest genes and identified two sets of genes that are associated with significant upregulation and downregulation of chromatin accessibility in BaMKO mice compared with controls (Fig. 5B and C). GO analysis of genes annotated by the significantly upregulated peaks revealed a remarkable enrichment of proinflammatory responses such as cell chemotaxis, leukocyte migration, and myeloid cell differentiation in both iWAT-SVF F4/80⁺ cells and eWAT-SVF F4/80⁺ cells from BaMKO mice (Fig. 5B and C).

To further identify which specific genes are likely to be directly regulated by BAF60a among all the genes that are significantly altered in SVF F4/80⁺ cells due to BAF60a inactivation, we performed overlapping analysis of RNA-Seq data (Fig. 4C and D) and ATAC-Seq data (Fig. 5B and C). We obtained 80 genes that are mainly enriched in chemotaxis and inflammatory response from this combined analysis (Fig. 5D). Notably, several proinflammatory cytokines including Ccl5, Cd24a, Gpx1, and Tarm1 are involved in the list of which genes are upregulated, while M2-like macrophage–associated anti-inflammatory genes such as IL-10 and Mgl1 are present in the list with downregulated genes. Interestingly, previous study demonstrated that Ccl5 but not Ccl2 enhances macrophage recruitment and survival in human adipose tissue (7). Collectively, the results suggest that BAF60a may regulate macrophage polarization and activation in white adipose tissue through directly orchestrating the chromatin accessibility of genomic loci surrounding genes that are involved in inflammatory response.

In the process of SWI/SNF complex–mediated chromatin remodeling, BAF60s subunits serve as the linkers between SWI/SNF complex and tissue-specific transcription factors (17). To identify the transcription factors that directly interact with BAF60a for SWI/SNF-mediated chromatin remodeling and transcriptional reprogramming of genes involved in macrophage polarization and activation, we performed known motif enrichment analysis on the overlapping peaks from the differential accessible peaks (DAPs) in ATAC-Seq (Fig. 5D) and BAF60a CUT&Tag-Seq in F4/80⁺ cells in WAT-SVFs from WT mice. Among all the 722 overlapping peaks, Fos, Fra1, Atf3, JunB, and Fra2 were five significantly enriched transcription factors (Fig. 6A). Interestingly, these five transcription factors all belong to the AP1 family, which has been shown to be closely related to immune response pathways (47). Activated AP1 serves as the key mediator to influence inflammatory response through transcriptional regulation of cytokines, such as IL-1β, TNF-α, IL-6, and IL-8. Notably, RNA-Seq analysis revealed that the expression of Fra1 in ATMs from both iWAT and eWAT depots is barely detectable with a much lower expression level than the other four transcription factors (Supplementary Fig. 6A).

Furthermore, integrative analysis of ATAC-Seq DAPs with BAF60a CUT&Tag-Seq data and chromatin immunoprecipitation sequencing (ChIP-Seq) data for the four transcription factors (Fra1 excluded) (obtained from GSE111854) was performed to demonstrate whether the ATAC-Seq DAPs were bound by these transcription factors and cofactor. As expected, a large portion of the ATAC-Seq DAPs were bound by BAF60a (Fig. 6B). In addition, among the four transcription factors, Atf3 binding sites were highly enriched in BAF60a-dependent ATAC-Seq DAPs (Fig. 6B), suggesting that BAF60a may directly interact with Atf3 to regulate the downstream target gene expression. To test this possibility, we first performed genome browser track analysis of the proinflammatory genes regulated by BAF60a ablation in macrophage including Ccl5 (Fig. 6C), Ccl2, IL-6, TNF-α, IL-1β, and iNOS (Supplementary Fig. 6B–D). In this case, ChIP-Seq data for H3K4Me3 and H3K27Ac were downloaded from the ENCODE database, and the ENCODE-annotated cis-regulatory elements (48) were included to indicate the promoter and enhancer regions in the genome. Compared with controls, eWAT F4/80⁺ cells from BaMKO mice exhibited a significantly elevated Ccl5 gene expression by RNA-Seq signals followed by higher ATAC-Seq signals on the promoter and enhancer regions flanking Ccl5 gene (Fig. 6C). These higher ATAC-Seq peaks are also occupied by BAF60a as revealed by BAF60a CUT&Tag-Seq analysis. Intriguingly, we also observed that among all four transcription factors only Atf3 displayed the same binding profile as BAF60a on the promoter and enhancer regions flanking Ccl5 gene locus (Fig. 6C), indicating that BAF60a can directly interact with Atf3 to regulate proinflammatory gene expression in macrophages.

Moreover, we also found that BAF60a and Atf3 are capable of colocating to both promoter and enhancer regions of Ccl2, IL-6, and TNF-α (Supplementary Fig. 6B), while they only bind to the enhancer region of IL-1β gene locus (Supplementary Fig. 6C). Together, these findings suggest that BAF60a can physically interact with Atf3 to regulate the transcription of these proinflammatory cytokines through chromatin remodeling. Notably, we also observed that only the binding signals of Atf3, and not BAF60a, are present on the enhancer regions around the iNOS gene locus, indicating that BAF60a may regulate iNOS expression indirectly (Supplementary Fig. 6D). Co-IP assays were performed in HEK293T cells transiently transfected with FH-BAF60a and Myc-Atf3, and the direct physical interaction between BAF60a and Atf3 was in turn validated (Fig. 6D). To further confirm the interaction of endogenous proteins, we performed IP assays using anti-Atf3 antibody in RAW264.7 macrophages and showed a strong interaction between endogenous BAF60a and Atf3 proteins (Fig. 6E). In addition, we also observed a pronounced attenuation of such endogenous interaction between BAF60a and Atf3 on LPS (100 ng/mL) stimulation, which mimics the metabolic endotoxemia in type 2 diabetes pathogenesis (Fig. 6F).

Next, we obtained PM from myeloid-specific Atf3 deficiency mice and qPCR analysis of proinflammatory gene expression. Remarkably, Atf3 ablation led to robust transcriptional induction of proinflammatory cytokines (Fig. 6G), closely resembling the effects of BAF60a inactivation in macrophages. Conversely, similar to the effects of BAF60a overexpression on proinflammatory gene expression in macrophages, retroviral vector–mediated overexpression of Atf3 and/or BAF60a in RAW264.7 cells markedly decreased LPS-induced expression of proinflammatory cytokines (Fig. 6H). Together, these results demonstrate that BAF60a physically interacts with Atf3 to regulate proinflammatory gene expression program in macrophage, thereby playing a critical role in the pathogenesis of obesity-induced adipose tissue inflammation, IR, and type 2 diabetes (Fig. 6I).

Besides BAF60a, BAF60 family members of the SWI/SNF chromatin remodeling complex also contain BAF60b and BAF60c. In this study, we examined the roles of BAF60b and BAF60c in the regulation of macrophage activation and inflammatory response in adipose tissues in obesity and diabetes. Myeloid cell–specific knockout of BAF60b (BbMKO) and BAF60c (BcMKO) mice were generated by crossing of Lyz2-Cre mice with the BAF60b flox/flox mice and BAF60c flox/flox mice, respectively. Consistent with the aforementioned data on absence of observed difference in BAF60b protein level in eWAT-SVF and iWAT-SVF from db/db mice (Fig. 1D and E), BAF60b deficiency exhibited no obvious effects on HFD-induced obesity and blood glucose levels in mice (Supplementary Fig. 7A and B). Echoing the existing evidence on a much lower expression of BAF60c in adipose tissue, SVFs, and ATMs (high enrichment in skeletal muscle), conditional inactivation of BAF60c in myeloid cells displayed no differences in HFD-induced obesity and glucose homeostasis (Supplementary Fig. 7C and D). Therefore, our data suggest that BAF60a is the only subunit of the BAF60s family that can be affected by HFD-induced obesity in macrophage, thereby impairing glucose homeostasis and causing aberrant activation of proinflammatory ATMs.

The cross talk between metabolic cells and the immune system is critically linked to metabolic homeostasis under physiological and disease conditions. It has been widely accepted that type 2 diabetes is an immune-related inflammatory disease, in which the obesity-induced macrophage activation and chronic inflammation are closely associated with its pathogenesis and disease progression (4). While several metabolic stress signals and transcription factors have been identified to play important roles in this process (11–13,49,50), the epigenetic checkpoint factors that can integrate the metabolic and stress cues to the transcriptional reprograming of macrophage activation and inflammatory response have yet to be defined. In this study, we unveil BAF60a as an environmental sensing component of the SWI/SNF chromatin remodeling complex to serve as an epigenetic rheostat of macrophage activation, adipose tissue inflammation, and systemic energy homeostasis. Mechanistically, we identify the transcription factor Aft3 to be the key downstream factor bound by BAF60a in mediating the selective chromatin remodeling and transcriptional reprogramming of proinflammatory cytokines in ATMs.

The ATP-dependent SWI/SNF chromatin-remodeling complexes include 11 subunits, among which the BAF60 subunits have been identified as a vital regulator in energy and glucose metabolism (15–17). BAF60 subunits are comprised of BAF60a, -b and -c, each of which is characterized by distinct tissue distribution patterns and regulatory machineries. They have emerged as critical linkers between the SWI/SNF complex and transcription factors to regulate target gene expression in different organs. BAF60a is encoded by Smarcd1, and its expression is enriched in adipose tissue and liver to accelerate hepatic fatty acid oxidation through interaction with PPAR-α and PGC-1α (18–20). BAF60b, encoded by Smarcd2, is highly expressed in immune cells, serving as an important regulator of granulocyte development (51,52). BAF60c, encoded by Smarcd3, is mainly expressed in skeletal muscle and heart and has been reported to work as a glucose sensor in skeletal muscle (22,24,25). In our previous studies we found that BAF60c in myocytes could be downregulated by inflammatory cytokines such as TNF-α, suggesting that muscle BAF60c serves as a connector between metabolic inflammation and systemic glucose homeostasis (23). In this study using myeloid-specific knockout mice, we have also explored the roles of BAF60b and BAF60c in macrophage activation, metabolic inflammation in adipose tissues, and systemic energy homeostasis. Consistent with their gene expression patterns and intactness in response to obesity-associated metabolic stress, myeloid-specific inactivation of BAF60b or BAF60c exhibited mild effects on diet-induced obesity and hyperglycemia (Supplementary Fig. 7), suggesting a cell type–specific and context-dependent role of BAF60 proteins.

The transcription factor Atf3 is a member of AP1 family, and it was previously identified as a player in regulating proinflammatory cytokine expression in macrophages (53). In this study, we demonstrated that myeloid-specific Aft3 ablation in mice enhances proinflammatory gene expression in primary PMs. Moreover, Atf3 displayed binding profiles similar to those of BAF60a on the promoter or enhancer regions flanking proinflammatory gene loci such as Ccl5, Ccl2, IL-1β, IL-6, and TNF-α, suggesting that BAF60a could suppress expression of those proinflammatory cytokines through directly interacting with Atf3. This is further confirmed by co-IP assays in transiently transfected 293 cells and direct IP experiments of endogenous BAF60a and Aft3 proteins in macrophages. Notably, we also found that only Atf3 occupies the enhancer region of the iNOS gene locus, while there was no obvious BAF60a binding peak, suggesting that BAF60a may indirectly regulate iNOS expression in macrophages. Besides Atf3, our motif analysis also revealed four other transcription factors of AP1 family that may interact with BAF60a. In fact, results of motif analysis worked in concert with our co-IP assay suggested that Fra2 might also be involved in BAF60a-mediated transcriptional regulation in macrophages via a relatively weaker binding between BAF60a and Fra2 (data not shown). Further studies are needed to test this possibility.

Macrophages have been reported to interact with multiple metabolic cells, and they are involved in the regulation of tissue homeostasis and metabolic balance in major metabolic organs (4). However, here we found that, except for white adipose tissues (eWAT and iWAT), BAF60a deficiency in myeloid cells elicits modest effects on metabolic inflammation in metabolic tissues such as BAT and liver, suggesting a white fat–specific effect of BAF60a in regulating macrophage polarization and proinflammatory response. Two factors may contribute to this tissue specificity. First, accumulating evidence indicates that the dynamic and extensive cross talk between the resident tissue macrophages and the local tissue microenvironment plays a critical role in shaping the identities and functions of tissue-specific resident tissue macrophages (54). As such, the unique local environment in white adipose tissue, the main body site for lipid storage and fatty acid release, may confer the distinct regulatory mechanisms and functions of ATMs at steady state and in obesity (4,55); Second, our recent studies have revealed that BAF60a functions as a key component of the lipid sensing and metabolism pathway in hepatocyte, adipocyte, and myocyte (18–20,25). In accordance, we showed that treatment with PA and TNF-α significantly decreased BAF60a expression in macrophages (Supplementary Fig. 1G), highlighting the specific role of BAF60a, rather than BAF60b and BAF60c, in regulating the macrophage activation and metabolic inflammation in white adipose tissue. While elucidating the underlying mechanisms of the processes mentioned above has been challenging, recent technological advances have revolutionized the study of macrophage biology. Notably, recent studies using single-cell RNA sequencing (scRNA-seq) approach have identified a unique population of triggering receptor expressed on myeloid cells 2 (Trem2)-expressing macrophages in adipose tissue homeostasis and nonalcoholic steatohepatitis pathogenesis (56,57). We also measured the mRNA expression of Trem2 and its associated marker genes Gpnmb and CD9 in iWAT, eWAT, and liver from HFD-fed control and BaMKO mice and observed a trending higher expression of Trem2 and Gpnmb in both white fat depots and liver (Supplementary Fig. 8). The link between BAF60a and Trem2 expression in macrophages and the exact role of Trem2⁺ macrophages in BaMKO-induced adipose tissue inflammation remain to be assessed in future studies using the emerging scRNA-seq and multiomics technologies.

In summary, our work identifies BAF60a as a key chromatin remodeling checkpoint factor that links obesity-associated stress signals to metabolic inflammation and systemic energy homeostasis. In addition, we also uncover Atf3 as an important downstream effector in BAF60a-mediated chromatin remodeling and transcriptional reprogramming of macrophage polarization and activation in adipose tissue. Our findings provide a novel framework for the development of new treatment targeting obesity-induced metabolic inflammation and its associated metabolic disorders.

Acknowledgments. The authors thank the members of the laboratory of Z.-X.M. for helpful discussion and technical support for this study. The authors also thank the Core Facilities of Zhejiang University School of Medicine for technical support.

Funding. This work was supported by grants from the National Key Research and Development Program of China (2018YFA0800403 and 2021YFC2701903 to Z.-X.M.), the Training Program of the Major Research Plan of the National Natural Science Foundation of China (91857110 to Z.-X.M.), the National Natural Science Fund for Excellent Young Scholars of China (81722012 to Z.-X.M.), the National Natural Science Foundation of China (81670740 to Z.-X.M. and 82100904 to S.H.), the Zhejiang Provincial Natural Science Foundation of China (LZ21H070001 to Z.-X.M. and LQ21C110001 to S.H.), the Innovative Institute of Basic Medical Sciences of Zhejiang University to Z.-X.M., the Construction Fund of Medical Key Disciplines of Hangzhou (OO20200055 to Y.G.), the Hangzhou Science and Technology Bureau (20150733Q13 and ZD20200129 to Y.G.), and the Fundamental Research Funds for the Central Universities to Z.-X.M. and by K.C. Wong Education Foundation (to Z.-X.M.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.G. and Z.-X.M. conceived the project and designed research. Q.K., J.Z., R.P., and Y.X. performed the experiments. Z.Z. carried out the bioinformatics analysis. S.H. and Y.X. contributed to discussion and data interpretation. Q.K., J.Z., Z.Y.Z., and Z.-X.M. analyzed data and wrote the manuscript. The authorship order of the co-first authors was determined by rolling dice. Z.-X.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.