Animal experiments were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals with approval from the Ethics Committee of Chongqing Medical University. C57BL/6J male mice (4–6 weeks old, n = 21) were maintained under standard recommended conditions in the Laboratory Animal Center of Chongqing Medical University. Briefly, mice were housed in colony cages under 12-h light-dark cycles at 23 ± 1°C. Animals were on the standard diet for 2 weeks prior to the experiments, and then the mice were fed with a high-fat diet (HFD) (60% fat; Research Diets, United States) for 12 weeks. After that, the body weight curve of these mice were compared to that of the control mice on normal diet (ND), and glucose tolerance test (GTT) was performed to evaluate the success of obesity model. Then the mice were randomly divided into three groups (n = 7 per group). The groups were further fed on high fat diet while receiving vehicle (5% Gum Arabic solution) (Sangon, China, dissolved in ddH₂O), OA 25 mg/kg or OA 50 mg/kg (selleck, United States, suspended in 5% Gum Arabic solution) per day individually by intragastric administration for 4 weeks, followed by a measurement of mice’s body weight and GTT. Finally, the mice were overnight fasted, assessed for the level of fasting blood glucose (FBG), and then sacrificed. The blood and WAT samples were collected for the follow-up experiments.

The mice were fasted for 14 h for GTT, followed by measurements of body weight and FBG. Mice were then injected intraperitoneally with glucose solution. The blood glucose levels at 30, 60, 90 and 120 min were monitored. Notably, the obese mice fed with HFD for 12 weeks were administered with 50% glucose solution (2 mg/g body weight) while the obese mice intragastric administrated for 4 weeks received 25% glucose solution (1 mg/g body weight), due to the fact that the mice at this stage were more obese and their blood glucose levels were prone to exceed the detection limit).

RAW264.7 (ATCC, United States) macrophages were grown in DMEM (Gibco, United States) supplemented with 10% fetal bovine serum (FBS) (Gibco), 1% penicillin/streptomycin (P/S) (Beyotime Biotechnology, China) at 37°C with 5% CO₂.

Femurs and tibias were isolated from male C57BL/6J mice of 6–8 weeks old and briefly sterilized by 70% ethanol. The bone marrow cells were resuspended in DMEM medium with 10% FBS, 1% P/S, and Macrophage Colony-Stimulating Factor (M-CSF) (10 ng/ml) (PeproTech, United States). The bone marrow-derived macrophages (BMDMs) were ready for further experiments after 7–10 days.

The inflammatory macrophages were established by additional 20 ng/ml IFN-γ (PeproTech) and 100 ng/ml LPS (Sigma-Aldrich, United States) treatment for 16 h. OA was prepared with DMSO (Sigma-Aldrich) as a stock solution of 50 mmol/L.

The cytotoxic effects of OA to RAW 264.7 cells were evaluated by the Methyl Thiazolyl Tetrazolium (MTT) assay, as previously described (Huang et al., 2020). In general, the seeded RAW 264.7 cells were incubated with OA and IFN-γ/LPS for 16 h, mixing with MTT, and then assayed for cell viability. The absorbance was monitored by a microplate reader at the wavelength of 490 nm (Supplementary Figure 1).

Epidydimal white adipose tissue (eWAT) and inguinal white adipose tissue (iWAT) were fixed in 4% paraformaldehyde, embedded in paraffin after dehydration with a series of ethanol solution, and cut into slides with the thickness of 5 μm, then stained with hematoxylin and eosin (H&E). AdipoCount 1.1 was used to calculate the adipocyte area.

Immunofluorescence was performed to evaluate the macrophages recruitment to adipose tissues by immune-staining. The sections were heated in citric acid repair solution for antigen repair and then blocked with 5% normal donkey serum for 2 h. Shook off the serum and added F4/80 antibody (PBST dilution: 0.1% Tween-20 and 0.5% BSA, 1:100 dilution) dropwise, and incubated the sections in a wet box at 4°C overnight (>8 h). The next day, took out the wet box and rewarming for more than 30 min, rinsed the sections in PBS, added the corresponding fluorescent secondary antibody (1:500 dilution) and incubate it in a wet box at room temperature for 1 h. Rinsed the sections in PBS again and stained the nuclei with DAPI (4′,6-Diamidino-2-28 phenylindole dihydrochloride) for 10 min at room temperature. Fully rinsed the sections in PBS, dried the remaining PBS buffer solution, covered with 50% glycerol (diluted with PBS), and applied nail polish around the cover slides to block the air. Immediately observed the sections under a fluorescence microscope (Olympus, Japan).

Serum insulin levels were determined by ELISA kit (Millipore, United States) and the standard operation steps were according to the manufacturer’s protocol. Serum TG, FFAs, and T-CHO levels (Nanjing Jiancheng Company, China), and SOD, Gpx activities (Beyotime Biotechnology, China) were measured using commercial kits according to the manufacturer’s instructions. Optical density (OD) was determined on a microplate reader. HOMA-IR index = [fasted insulin (μIU/ml) × fasted glucose (mmol/L)]/22.5 (Neuschwander-Tetri, 2010). Adipo-IR index = fasted insulin (mmol/L) × fasted NEFA (pmol/L) (Musso et al., 2012).

Total RNA was isolated from eWAT or cells using TRIzol Reagent (Thermo Scientific, 15596026, United States). For qRT-PCR, 1 μg total RNA from each sample was reverse-transcribed by using a Revert Aid first-strand cDNA synthesis kit (Thermo Scientific, 00698284, United States). The cDNA products were amplified using Quantstudio3/5 (Thermo Scientific, United States) real-time PCR instrument with the Power SYBR Green PCR Master Mix (Thermo Scientific, 00736756, United States). The expression levels of target genes were calculated using the 2^−ΔΔCT method with normalization to the standard housekeeping gene 18s, and expressed as relative mRNA levels compared with internal control. Primers used for qRT-PCR are shown in Table 1.

Cells were lyzed in cell lysis buffer on ice for 30 min. The tissue samples were sonicated (70 Hz, 90 s) in cell lysis buffer, followed by an additional incubation on ice for 20 min. Then the lysates were centrifuged at 4°C for 15 min at the speed of 12,000 rpm. The protein in the collected supernatant was degenerated under 100°C and then quantitated. Equal amounts of protein samples were loaded on SDS-PAGE gels, separated by electrophoresis, and transferred onto PVDF membranes. After being blocked with 5% skim milk, the membranes were incubated over-night at 4°C with the primary antibodies and 1 h at room temperature with appropriate secondary HRP-conjugated antibodies. Antibodies are shown in Table 2.

Adipose tissues were minced in PBS containing 0.075% collagenase (Sigma-Aldrich, C2139, United States). After incubated at 37°C for 30 min and filtrated with 100 mesh filter, cell suspensions were centrifuged at 1,500 rpm for 5 min to remove adipocyte. Isolated stromal vascular fraction (SVF) pellet was collected from the bottom. The SVF pellet was resuspended in PBS containing 3% BSA, then red blood cell lysis buffer was added and incubated for 3 min. After washing in 3% BSA, bottom cells were incubated with Fc-Block (CD16/32, 12-0161-85, ebioscience) for 20 min at 4°C. Antibodies against CD45-FITC (11-0451-82, ebioscience), F4/80-PE (123110, Biolegend), CD11b-PerCP/Cy5.5 (101227, Biolegend), CD206-APC (141707, Biolegend) and CD11c-APC (117310, Biolegend) were added, and incubated for 20 min followed by washing in PBS containing 3% BSA. Data analysis and compensation were performed using BD Accuri C6 Plus.

RAW264.7 macrophages were digested with trypsin and terminated with PBS containing 3% BSA, then centrifuged, resuscitated and incubated with Fc-Block. The following steps were the same as above.

For the migration assay, 100,000 BMDMs were placed in the upper chamber of an 8 μm polycarbonate filter (24-transwell format; Corning, United States), whereas the corresponding conditioned medium was placed in the lower chamber. After 16 h, cells were fixed in formalin, stained with crystal violet and observed under the microscope.

Mitochondrial ROS level was determined using MitoSOX™ Red mitochondrial superoxide indicator (Invitrogen, United States) and FCM analysis. The specific operation steps were according to the manufacturer’s protocol.

All data are presented as means ± SEM. Mean value differences between two groups were assessed by two-tailed Student’s t-test. p values less than 0.05 were considered to be statistically significant. Statistical analyses were performed with Graph Pad Prism 8.

The obese murine model was applied here to study the effect of OA on obesity-related metabolic dysfunction. C57BL/6J mice were fed with HFD for 12 weeks, and the mice body weight increased significantly, the impaired glucose tolerance suggested establishment of the obesity model, compared with the mice fed with ND (Supplementary Figure 2). Then the obese mice were intragastric administrated with vehicle, 25 mg/kg or 50 mg/kg OA for 4 weeks. The tested OA concentration was adapted from previous studies that used doses of 20, 40, 250 mg/kg/day, or 50 mg/kg/3 days in mice (Wang et al., 2015; Su et al., 2018; Nakajima et al., 2019), and 5–100 mg/kg/day in rats (Ying et al., 2014; Lee et al., 2016; Matumba et al., 2019).

OA administration in obese mice lowered the bodyweight (Figure 1A) while improved glucose tolerance (Figure 1B), fasting insulin level (Figure 1C) and HOMA-IR index (Figure 1D). Consistently, OA administration also induced activation of the AKT pathway, which is considered as a marker event of insulin sensitivity improvement (Figure 1E). Moreover, OA decreased basal plasma concentrations of total cholesterol (TC), triglyceride (TG) and FFA (Figure 1F). In addition, the Adipo-IR index (Figure 1F) was decreased by OA treatment. Obesity-related IR can up-regulate lipolysis (Degerman et al., 2003) and increase FFA levels. Figure 1G showed the ratio of phosphorylated hormone sensitive lipase (p-HSL) to HSL, which usually used as the indicator of adipose lipolysis, was down-regulated by OA treatment. The 25 mg/kg dose of OA treatment had maximal effect on these improvements. These data showed that OA could improve glucose and lipid metabolism in HFD-induced obese mice. Notably, the improvement in inflammation, glucose tolerance, and insulin sensitivity in the OA-treated mice also sustained in the body-weight matched groups (Supplementary Figure 3).

Figure 1 has shown that OA treatment improves IR, thus we next tried to underline how OA treatment achieves its effects on IR. It is known that changes in adipocyte morphology and infiltration of macrophages in adipose tissue contribute to IR development (Xu et al., 2020), we then examined the effect of OA on adipocyte morphology and macrophage infiltration. As shown in Figure 2A, OA treatment reduced the ratio of adipose tissue weight to body weight, and significantly decreased the adipocyte size in HFD-treated mice (Figures 2B,C). HE staining and immunofluorescence showed fewer CLSs in the eWAT after OA treatment, which suggested a decreased macrophage accumulation (Figures 2B,D). The mRNA expression of F4/80 (a macrophage marker) in eWAT was evaluated by qPCR, and the proportion of F4/80⁺ cells in eWAT was measured by FCM. The results showed a decreasing trend of macrophage infiltration by OA treatment (Figures 2D,E).

Macrophages infiltrating adipose tissue are generally chemotactic from peripheral blood. In order to further confirm whether OA could reduce the chemotaxis of macrophages, the effect of OA on macrophage migration under inflammatory condition was assayed using trans-well chemotaxis assay in vitro. As shown in Figure 2F, OA significantly inhibited IFN-γ/LPS-induced macrophage migration. These data suggested that OA decrease macrophage infiltration by inhibiting chemotaxis of macrophages.

An significant increase in ATM infiltration is often observed in eWAT rather than iWAT in the process of obesity (Gómez-Ambrosi et al., 2004; Cancello et al., 2005; Amano et al., 2014). Consequently, infiltrated ATM induced expression of inflammatory markers that play key regulatory roles in the development of obesity-related IR (Dong et al., 2014). Here we next investigated whether OA treatment can impact inflammation in eWAT. The qPCR results showed that pro-inflammatory markers derived from M1 macrophages (iNOS, MCP1, TNFα) were significantly down-regulated in eWAT of OA administrated mice compared with the vehicle treated mice (Figure 3A). The decreased expression of M2 macrophage related genes in eWAT (Figure 3A) may be due to the decrease in the total number of infiltrated macrophages in adipose tissue after OA treatment. These findings were associated with attenuated phosphorylation of c-Jun N-terminal kinase (JNK) and extracellular signal-regulated kinase (ERK) (the key proteins in MAPK signaling pathway) in eWAT of diet-induced obesity (DIO) mice (Figure 3B).

To explain how inflammation is down-regulated in eWAT of OA-administrated mice, we tested whether this decrease is due to regulation of macrophages polarization in eWAT. FCM analysis showed that the proportion of M1 ATMs decreased in WAT of obese mice with OA treatment, whereas the proportion of M2 ATMs increased, the ratio of M1/M2 decreased significantly (Figures 3C–F). The polarity transition of ATMs could lead toward an anti-inflammatory phenotype.

Taken together, these results suggested that OA treated mice showed attenuated inflammation and decreased M1/M2 ratio in WAT of DIO mice.

Given the association between OA treatment and ATM polarization in obese mice, RAW264.7 macrophages and bone marrow-derived macrophages (BMDMs) were engaged to determine whether OA directly regulates macrophage activation and/or polarization. Consistently with results in vivo, the expression levels of M1 marker genes (iNOS, MCP1, IL-6 and TNFα) were decreased in these macrophages stimulated with the combination of IFN-γ/LPS and OA (Figure 4A; Supplementary Figure 4B), and the phosphorylation of JNK, ERK and p38 were also inhibited by the treatment as expected (Figure 4B; Supplementary Figure 4A). Furthermore, flow cytometry analysis showed that OA significantly inhibited the IFN-γ/LPS induced M1 polarization (Figure 4C).

To investigate how OA reduce adipose tissue inflammation and inhibit macrophage M1 polarization, we further tested whether it depends on regulating activation of the NLRP3 inflammasome. The NLRP3 inflammasome (NLRP3/ASC/caspase-1 complex) is a key player of inflammation and M1 macrophage polarization (Ślusarczyk et al., 2018), and plays a central role in the induction of obesity and IR (Stienstra et al., 2011; Vandanmagsar et al., 2011).

We firstly examined the expression of caspase-1, which was the effector of NLRP3 inflammasome, in RAW264.7, BMDMs and mice eWAT. The data showed that the up-regulation of caspase-1 induced by IFN-γ/LPS or DIO was inhibited significantly by OA treatment (Figures 5A–D; Supplementary Figures 5A,B). Consistently, the levels of IL-1β and IL-18, the inflammatory cytokines processed by inflammasome, were also decreased significantly with OA treatment (Figures 5B,D; Supplementary Figure 5B). Our results suggested that OA could resist the activation of NLRP3 inflammasome.

Due to ROS production in macrophages triggers the activation of NLRP3 inflammasome (Dostert et al., 2008; Eisenbarth et al., 2008), the levels of the mitochondrial ROS was detected. Figures 5E,F revealed that OA attenuated the up-regulation of ROS production induced by IFN-γ/LPS stimulation in RAW 264.7 cells.

Voltage dependent anion channels, the most abundant proteins of the outer mitochondrial membrane (Colombini, 2004), is known to regulate mitochondrial ROS production (Da-Silva et al., 2004) and associated with the NLRP3 inflammasome (Wolf et al., 2016). Previous investigators had noted that the knockdown of the VDAC somehow blocks NLRP3 inflammasome activation (Zhou et al., 2011). With IFN-γ/LPS stimulation in vitro or DIO in vivo, VDAC protein levels were up-regulated, which is consistent with ROS production. However, OA attenuated the increase of VDAC protein levels (Figures 5A,C; Supplementary Figure 5A).

To further determine whether OA improves inflammation and regulates macrophage polarization through VDAC and ROS, rotenone (selleck, United States, 10 μmol/L), a mitochondrial complex Ι inhibitor that can lead to the up-regulation of VDAC and robust ROS production (Zhou et al., 2011; Jiang et al., 2017), was added 6 h before cell harvest to counteract the effects of OA on VDAC expression and ROS production in RAW264.7 macrophages. We observed that the addition of rotenone up-regulated VDAC, increased the production of ROS, and significantly weakened the anti-inflammatory and the inhibition effect of the MAPK pathway and inflammasome activation of OA. In addition, inhibition of VDAC by the inhibitor VBIT-12 (selleck, United States, 20 μmol/L) in RAW264.7 macrophages treated with IFN-γ/LPS mimicked the OA’s effects (Figure 6). Endogenous antioxidant enzymes such as superoxide dismutase (SOD) and glutathione peroxidase (GPx) have been known as the main regulator in the mitochondrial ROS production, and their activation can lead to the remission of inflammatory response (Gasparrini et al., 2017; Zhao et al., 2020). We found that OA increased the activities of SOD and GPx in RAW264.7 macrophages stimulated by IFN-γ/LPS, while rotenone attenuated this effect. As expected, VBIT-12 also increased the activities of SOD and GPx (Figure 6H). Taken together, these data supported that OA could inhibit the activation of NLRP3 inflammasome by reducing ROS production and VDAC expression.