Metformin reduces macrophage HIF1α-dependent proinflammatory signaling to restore brown adipocyte function in vitro

Abstract

Therapeutic potential of metformin in obese/diabetic patients has been associated to its ability to combat insulin resistance. However, it remains largely unknown the signaling pathways involved and whether some cell types are particularly relevant for its beneficial effects. M1-activation of macrophages by bacterial lipopolysaccharide (LPS) promotes a paracrine activation of hypoxia-inducible factor-1α (HIF1α) in brown adipocytes which reduces insulin signaling and glucose uptake, as well as β-adrenergic sensitivity. Addition of metformin to M1-polarized macrophages blunted these signs of brown adipocyte dysfunction. At the molecular level, metformin inhibits an inflammatory program executed by HIF1α in macrophages by inducing its degradation through the inhibition of mitochondrial complex I activity, thereby reducing oxygen consumption in a reactive oxygen species (ROS)-independent manner. In obese mice, metformin reduced inflammatory features in brown adipose tissue (BAT) such as macrophage infiltration, proinflammatory signaling and gene expression, and restored the response to cold exposure. In conclusion, the impact of metformin on macrophages by suppressing a HIF1α-dependent proinflammatory program is likely responsible for a secondary beneficial effect on insulin-mediated glucose uptake and β-adrenergic responses in brown adipocytes.

Graphical abstract

Highlights

• •Macrophage-mediated inflammation in brown adipocytes stabilizes HIF1α.
• •Stabilization of HIF1α reduces brown adipocyte hormonal responses.
• •Metformin attenuates inflammation and BAT dysfunction in obese mice.
• •Mitochondrial complex I inhibition by metformin reduces HIF1α in macrophages.
• •Targeting macrophages HIF1α by metformin alleviates brown adipocyte inflammation.

1. Introduction

Since 1980 the global prevalence of overweight and obesity has been duplicated and currently almost one third of the world's population is affected by this pathology [1]. Obesity is characterized by a number of pathophysiological changes, including hypertrophy/hyperplasia of adipocytes, as well as chronic low-grade inflammation [2], both contributing to insulin resistance and related metabolic diseases such as type 2 diabetes (T2D) and cardiovascular diseases [3].

During the process of weight gain, the number of macrophages in white adipose tissue (WAT) is increased mainly by the recruitment of circulating monocytes shifting their polarity to a proinflammatory (M1) state. These macrophages secrete proinflammatory cytokines such as tumor necrosis factor (TNF) α, interleukin (IL) 1β and IL6 and increase the levels of the inducible nitric oxide synthase (iNOS) [2,4]. This inflammatory process and its deleterious impact on adipose cells' functionality has been mainly studied in WAT since, in the classical view, brown adipose tissue (BAT) [5,6] was thought to have limited physiological relevance in adult humans. However, in the last two decades imaging studies have identified metabolically active BAT depots in the interscapular and cervical areas of healthy adults [[7], [8], [9]]. Further research associates the amount of BAT with body mass index and other metabolic parameters such as fasting glucose, triglycerides, and cholesterol. Altogether, these seminal studies suggest that a decrease in active BAT may be associated with the accumulation of visceral WAT and unfavorable metabolic parameters [10,11]. Studies conducted in murine models of diet-induced obesity have shown infiltration of immune cells in BAT depots [12,13], one of them reporting reduced uncoupling protein 1 (UCP1) content [13].

Different processes have been proposed as triggers of the inflammatory state during obesity such as endoplasmic reticulum/oxidative stress and hypoxia of the adipose tissue [[14], [15], [16]]. The hypoxia-inducible factor HIF1α is responsible for regulating oxygen homeostasis [17]. During normoxia, HIF1α protein is degraded by prolyl hydroxylases (PHDs) through a mechanism involving the von Hippel-Lindau (VHL) protein in an ubiquitin- and proteasome-dependent manner [18]. HIF1α regulates the transcription of many genes involved in glucose and energy metabolism, iron metabolism, vascular development, cell proliferation, and viability [15,19]. Although PHDs are largely considered as direct oxygen sensors, some other studies have reported that accumulation of mitochondrial reactive oxygen species (ROS) can also mediate HIF1α stabilization in hypoxia [[20], [21], [22], [23], [24]]. In this line, it has been shown that ROS accumulation lead to inactivation of PHDs by promoting the oxidation of FeII into FeIII [25]. Moreover, it seems that this ROS-dependent HIF1α stabilization is particularly operative in hypoxia, but not in conditions close to anoxia [26].

Inflammation is another hypoxia-independent HIF1α stabilization condition [27,28] which has been attributed at least in part to the presence of an NFκB binding site in the proximal promotor of HIF1α [29]. Remarkably, genetic downregulation of HIF1α in adipocytes protects against inflammation and insulin resistance in obese mice [[30], [31], [32]]. On the other hand, Halberg et al. [30] reported that overexpression of a constitutively active form of HIF1α in adipocytes leads to weight gain and glucose intolerance, and indicates that aberrant HIF1α activity produces high local inflammation in WAT and an exacerbated diabetic phenotype. All these studies have shed light on the relevant role of HIF1α in developing obesity-associated inflammation and insulin resistance. However, the specific involvement of HIF1α in brown adipocytes and thermogenic BAT responsiveness during the inflammatory context of obesity has been less explored.

In this study we identified a molecular mechanism involving HIF1α in the negative impact of macrophages-mediated inflammation on brown adipocytes regarding their sensitivity to insulin and β-adrenergic responses. Since several pharmacological studies using metformin have revealed antiinflammatory properties of this antidiabetic drug [[33], [34], [35]], we also studied the effect of metformin in the interactome between macrophages and brown adipocytes, particularly related to HIF1α regulation.

2. Material and methods

Reagents and antibodies. Cell culture reagents were from Invitrogen (Carlsbad, CA, USA). Insulin (l0516), metformin (D150959) and CL316243 (C5976) were from Sigma (Sigma-Aldrich, Saint Louis, MO, USA). The anti-Large T antigen (LTAg) antibody was kindly provided by J. de Caprio (Dana Farber Cancer Institute, Boston, MA). Anti-phospho JNK (Thr 183/Tyr185) (#9251), anti-phospho STAT3 (Tyr705) (#9131), anti-phosho p38 MAPK (Thr180/Tyr182) (#9211), anti-p38 MAPK (#9212), anti-phospho Akt (Ser473) (#4058), anti-phospho Akt (Thr308) (#9275), anti-Akt (#9272), anti-phospho HSL (Ser660) (#4126), anti-HSL (#4107T), anti-phospho AMPK (Thr172) (#2531) and anti-AMPK (#2532) antibodies were from Cell Signaling Technology (MA, USA). Anti-GLUT4 (sc-53566), anti-GLUT1 (sc-1603), anti-JNK (sc-571), anti-iNOS (m19sc650) and anti-Vinculin (sc-73614) antibodies were from Santa Cruz (Palo Alto, CA, USA). Anti-IRS1 (06–248), anti-phospho IRS1 (Ser307) (07–247) and anti-AMPKα1 (07–350) antibodies were from Merck Millipore (Merck KGaA, Darmstadt, Germany). The anti-UCP1 antibody was purchased from Abcam (Ab109803,Cambridge, UK). The anti-HIF1α (10006421) antibody was from Cayman Chemical laboratories and the anti-F4/80 (MF 4F) antibody was from Immunostep (Salamanca, Spain).

Animals. Eight-week-old C75BL/6J male mice were purchased from Charles River. Mice were fed a HFD (60% of calories from fat, TD-06414, Envigo, Spain) for 10 weeks and then divided into two groups: mice that continued on a HFD for 6 additional weeks and mice fed a HFD and treated with metformin hydrochloride (100 mg/kg) daily by oral gavage. Then, half of the mice of each group were exposed to the cold (4 °C) for 16 h. Animal experimentation was approved by the Ethics Committees of Consejo Superior de Investigaciones Científicas and Comunidad de Madrid (Spain) in accordance with the European Union guidelines.

Generation of brown preadipocyte cell lines and differentiation. Brown preadipocytes were isolated from the interscapular BAT of 20-day old suckling mice and immortalized as previously described [36]. Brown preadipocytes were also immortalized from Ampkα1^−/− mice [37] or mice containing loxP-flanked alleles of HIF1α previously described [38], referred as BA^Ampkα1−/- and BA^Hif1afl/fl, respectively.

For differentiation, immortalized brown preadipocytes were grown in DMEM supplemented with 10% fetal bovine serum (FBS), 20 nM insulin and 1 nM triiodothyronine (T3) (differentiation medium, DM) until reaching confluence. Next, the cells were cultured for 2 days in induction medium (IM) consisting of DM supplemented with 0.5 μM dexamethasone, 0.125 μM indomethacin, 1 μM rosiglitazone and 0.5 mM isobutyl-methyl-xanthine (IBMX). Then, cells were cultured in DM until day 7 in which they exhibited a fully differentiated phenotype.

To generate Hif1α null brown adipocytes (BA^Hif1αΔ/Δ), at day 6 of differentiation BA^Hif1αfl/fl were transduced with adenoviruses encoding cre recombinase or β galactosidase at a titer of 4 x 10¹⁰ plaque-forming units. After 48 h (day 8), cells were used for experiments.

To generate a brown adipocyte cell line stably expressing HIF1α, we used the retroviral vector pRv-HIF1a(P2)* encoding a mutated version of HIF1α at positions P402 and P564 which lacks critical proline residues and is therefore not recognized by VHL [39]. This construct is constitutively active under normoxic conditions irrespective of the oxygen tension [40]. As a control, we utilized the empty vector pRv. For retroviral infection, Phoenix cells seeded in p100 culture dishes were transfected with 10 μg of each retroviral vector and 10 μg of the pCLAmpho Retrovirus Packaging Vector (Imgenex) using Lipofectamine 2000 (Invitrogen). Cell culture supernatants were harvested 24 h after transfection, filtered through a 0.45-μm filter, diluted (1:2) with fresh medium containing 8 mg/ml polybrene (Sequabrene, Sigma) and preadipocytes were differentiated until day 7.

Isolation and culture of bone marrow derived macrophages (BMDM). BMDM were obtained by flushing bone marrow from 8 weeks-old C57BL/6J mice femurs and tibias. Briefly, mice were euthanized by CO₂ inhalation and death was ensured by cervical dislocation. The skin of the mouse was disinfected with 70% alcohol. The tibia and femur were isolated and any tissue was removed from the bones. The epiphyses were cut of the bones and flushed the marrow into a 50 ml centrifuge tube using a 5 ml syringe and a 25G needle. The cell suspension was filtered with a 75 μm cell strainer in a new 50 ml tube to remove the excess of tissue, fat or other debris (all this procedure was performed under a laminar flow hood). Cell suspension was centrifuged at 300 × g for 5 min and the pellet was resuspended in 2 ml of ACK lysis buffer for 2 min at RT to lysate the red blood cells. Cells were washed with complete DMEM medium and centrifuged at 300 × g for 5 min. Then, bone marrow cells were cultured with L929 differentiation medium (10% FBS, 1% P/S, 10–30% of L929-CM) in non-treated p100 plates for 6–8 days to differentiate into macrophages.

Generation of conditioned media in macrophages and treatment of brown adipocytes. Raw 246.7 murine macrophages cell line and BMDM were used to obtain conditioned media for the in vitro experiments. Raw 246.7 macrophages were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with 10% heat-inactivated FBS, antibiotics (1% P/S; 100 U/ml penicillin, 100 μg/ml streptomycin) and glutamine (2 mM) at 37 °C in a humidified atmosphere with 5% CO₂. To generate a proinflammatory CM (CM-Lps), confluent (90–95%) macrophages were treated for 4 h with 100 ng/ml LPS (InvivoGen, California, USA) in RPMI medium supplemented with 2% FBS, 1% P/S and 2 mM glutamine. Then, cells were washed twice with phosphate buffer saline (PBS) and kept in RPMI-2% FBS for 16 h. To obtain an antiinflammatory CM (CM-LpsMet), confluent Raw 246.7 macrophages were treated with 100 ng/ml LPS as detailed above and then the cells were washed twice with PBS and treated with metformin (10 mM) in RPMI medium-2% FBS for 16 h. To produce CM from Raw 246.7 macrophages with inhibited complex I activity (CM-LpsRtn), cells were treated with 100 ng/ml LPS as detailed above and then the cells were treated with rotenone (1 μM) for 2 h after which they were washed twice with PBS and maintained in RPMI-2% FBS for a further 16 h. To produce CM from macrophages with inhibited HIF1α transcriptional activity (CM-InhLps), Raw 246.7 cells were pretreated with a transcriptional inhibitor of HIF1α (sc-205346, Santa Cruz) at 30 μM concentration in RPMI-2% FBS, 1% P/S and 2 mM glutamine for 1 h. Then, LPS (100 ng/ml LPS) was added for 4 h in the presence of the HIF1α inhibitor. After this time, the cells were washed twice with PBS and maintained in RPMI-2% FBS for a further 16 h. As control, CM was collected from macrophages maintained for the corresponding time-periods in RPMI-2% FBS (referred as CM-Ctl). Similar treatment to generate CM was conducted in BMDM using 150 ng/ml LPS and 1 mM metformin, dose previously used to treat BMDM [41]. Before addition to brown adipocytes, all CM were harvested and filtered through a 0.22-μm filter to remove cell debris.

Measurement of nitrite production. NO production was evaluated in the CM by measuring nitrites, an oxidative product of NO. Nitrite content was determined by the Griess reaction [42]. Aliquots of CM were collected and mixed with 1 mM sulphanilic acid (151057.1208, Panreac) and 1 mM N-(l-naphthyl) ethylenediamine (NEDA) (N912510G, Sigma-Aldrich). Nitrites production was measured spectrophotometrically at 548 nM and the concentration was determined by a standard curve of NO₂.

Conditioned medium cytokine detection assay. CM were collected and immediately frozen at −80 °C. TNFα, IL6 and IL1β were determined using Luminex 100 IS (Merck Millipore, Darmstadt, Germany).

Treatment of differentiated brown adipocytes with CM. For experiments, differentiated brown adipocytes were treated with CM-Lps, CM-LpsMet, CM-LpsRtn, CM-InhLps or the corresponding CM-Ctl for several time-periods to analyze proinflammatory signaling cascades. For insulin signaling, brown adipocytes were pretreated with either CM for 16 h prior to stimulation with 10 nM insulin (Sigma-Aldrich, Missouri, USA) for 10 min. To study β-adrenergic responses, brown adipocytes were treated with either CM in the presence of 2 μM CL316243 for several time-periods as described in results and figure legends. At the end of the treatment cells were washed twice with PBS and collected for further analysis.

Treatment of brown adipocytes with cytokines. Differentiated brown adipocytes were treated with a cocktail of TNFα (300-01A, PeproTech), IL6 (216–16, PeproTech) and IL1β (211-11B, PeproTech) (30 ng/ml each) for 16 h in serum-free RPMI medium. Responses to insulin and CL316243 were determined as described for the treatment with CM.

Measurement of glucose uptake. Differentiated brown adipocytes were treated with CM-Ctl or CM-Lps for 16 h and after cells were serum-deprived for 1 h. Then, cells were washed with ice-cold Krebs-Ringer-phosphate buffer (KRP) (135 mM NaCl, 5.4 mM KCl, 1.4 mM CaCl₂, 1.4 mM MgSO₄, 10 mM sodium pyrophosphate, pH 7.4) and incubated with KRP buffer with or without insulin for 10 min at 37 °C. Then, 2-Deoxy-D [1–³H] glucose (500 nCi/ml) was added to this solution, and the incubation was continued for 5 min at 37 °C. The cells were washed with ice-cold KRP buffer and solubilized with a syringe in 1% SDS as previously described [43]. After 2 h incubation at 37 °C, radioactivity was measured by liquid scintillation counting. Total protein was determined by BCA method and results were normalized to the protein amount of each culture plate.

Measurement of glycerol release. After treatment of brown adipocytes with 2 μM CL316243 for 3 h, cell medium was removed and glycerol content was measured using a commercial kit (Biosystems, Barcelona, Spain) according to the manufacturer's instructions.

Homogenization and preparation of tissue extracts. Frozen BAT samples were homogenized in 16 vol (w/v) of ice-cold lysis buffer containing 50 mM Tris-HCl, 1% Triton X-100, 2 mM EGTA, 10 mM EDTA, 100 mM NaF, 1 mM Na₄P2O₇, 2 mM Na₃VO₄, 100 μg/ml phenylmethylsulphonyl fluoride (PMSF), 1 μg/ml aprotinin, 1 μg/ml pepstatin A and 1 μg/ml leupeptin. Samples were homogenized in the same lysis buffer using the Brinkman PT 10/35 Polytron. Extracts were kept ice-cold at all times. Tissue extracts were cleared by centrifugation at 40,000×g for 20 min at 4 °C twice. The supernatant was aliquoted and stored at −80 °C.

Preparation of protein cell extracts and Western blot. Cells were scraped off in ice-cold PBS, pelleted by centrifugation at 4000×g for 10 min at 4 °C and resuspended in buffer containing 10 mM Tris-HCl, 5 mM EDTA, 50 mM NaCl, 30 mM disodium pyrophosphate, 50 mM NaF, 100 μM Na₃VO₄, 1% Triton X-100, 1 mM PMSF, 10 μg/ml leupeptin and 10 μg/ml aprotinin pH 7.6 (lysis buffer). Cell lysates were clarified by centrifugation at 12,000×g for 10 min at 4 °C twice. Protein content was determined by the Bradford method, using the Bio-Rad reagent and bovine serum albumin (BSA) as the standard. After SDS-PAGE, proteins were transferred to Immobilon membranes (Merk-Millipore), blocked using 5% non-fat dried milk or 3% BSA in 10 mM Tris-HCl, 150 mM NaCl pH 7.5, and incubated overnight with the antibodies indicated in 0.05% Tween-20, 10 mM Tris-HCl, 150 mM NaCl pH 7.5. Immunoreactive bands were visualized using the ECL Western blotting protocol (Merk-Millipore).

Quantitative real-time PCR (RT-qPCR). Total RNA was extracted with Trizol (Invitrogen). Complementary DNA (cDNA) was synthesized by reverse transcription PCR (RT-PCR) from 1 μg of RNA using SuperScript™ III First-Strand Synthesis System for quantitative PCR (qPCR) following the manufacturer's indications (Invitrogen). RT-qPCR was performed with an ABI 7900 sequence detector (Applied Biosystems).

Lipid staining. To visualize lipid droplets in brown adipocytes, BODIPY 493/503 stock solution (1 mg/ml in methanol) (d-3922; Molecular Probes) was diluted 1:1000 in RPMI medium supplemented with 2% FBS before addition to differentiated brown adipocytes for 30 min. Then, cells were fixed with 4%-PFA. Mounting medium (Vectashield with 4,6-diamidino-2-phenylindole (DAPI), H-1200, Vector Laboratories) was used for fluorescence imaging. Brown adipocytes were observed and images were captured using an LSM 700 confocal microscope (Zeiss).

Immunofluorescence analysis in brown adipocytes. Brown preadipocytes were differentiated on glass coverslips, fixed in 4%-paraformaldehyde (PFA) and processed for immunofluorescence. Briefly, anti-GLUT4 (sc-56566, Santa Cruz) or anti-GLUT1 (sc-377,228, Santa Cruz) primary antibodies were applied for 1 h at 37 °C in PBS-1% BSA at 1/50 dilution. The secondary antibody used was Alexa 488 goat anti-mouse (Invitrogen) at 1/500 dilution. Cell nuclei were counterstained with DAPI (Sigma). Immunofluorescence was examined using an LSM 700 confocal microscope (Zeiss).

Histological analysis and immunostaining. For immunohistochemistry, BAT tissue was fixed in 4%-PFA (16,005, Sigma-Aldrich) for 24 h, washed twice with PBS, dehydrated with ascending ethanol solutions, incubated with xylene, and then embedded in paraffin. Blocks were cut into 5 μm sections. Prior to H&E staining or immunohistochemistry, the sections were deparaffinized in xylene and hydrated in descending ethanol solutions and distilled water. The slides were stained with Mayer's hematoxylin (MHS32-1L, Sigma-Aldrich) for 15–20 min and eosin (1.15935.0025, Merck) for 1 min. After, the slides were dried and mounted and the images were captured with an Axiophot light microscope (Zeiss) using a 40 × objective.

For UCP1 and HIF1α immunohistochemistry, sections were heated for 20 min in 100 mM sodium citrate buffer (pH 6.0) and 0.05% Tween-20. After washing with PBS, endogenous peroxidase was blocked by incubation with a solution of 9% H₂O₂ in 10% methanol for 10 min at RT. The tissue sections were then incubated in 6% BSA and 2% normal horse serum (NHS) in PBS-0.1% Triton X-100 for 1 h at room temperature to block non-specific binding. The sections were incubated with primary antibody against UCP1 (1:500, Ab10983, Abcam) or primary antibody against HIF1α (1:100, anti HIF-1α Rat monoclonal antibody (SIMA343B) kindly donated by Dr. Martín-Puig from Centro Nacional de Investigaciones Cardiovasculares, CNIC & Universidad Francisco de Vitoria, Madrid, Spain) overnight at 4 °C. Sections were then washed with PBS and incubated with biotinylated secondary anti-rabbit (1:250) (BA-1100, Vector Laboratories) in the case of UCP1 or biotinylated secondary anti-rat (1:250) (BA-4000, Vector Laboratories) for HIF1α for 1 h at room temperature. After washing with PBS, the slides were incubated with DAB-immunoperoxidase staining (SK-4100, Vector Laboratories). After washing with water for 5 min, the sections were counterstained with hematoxylin (H3136, Sigma-Aldrich) solution for 10 s and rinsed again with water. After dehydration and xylene incubation, the sections were assembled with DePeX (18243.02, SERVA Electrophoresis GmbH), covered with coverslips, and dried overnight.

For immunofluorescence staining of BAT, paraffin sections were deparaffinized and rehydrated. The sections were boiled in 10 nM citrate buffer for 5 min and treated with 0.3% Triton X-100 in PBS for 10 min. Then, the sections were blocked with 0.3% Triton X-100 in PBS containing 0.5% BSA for 1 h at room temperature. Thereafter, sections were incubated with blocking solution containing primary antibodies (1:250, anti-HIF1α Rat monoclonal antibody (SIMA343B) or 1:500, anti F4/80, Immunostep) overnight at 4 °C. Finally, the sections were washed and probed with Alexa-conjugated secondary antibodies. Fluorescent images were taken with the confocal microscope.

MitoTracker Green and MitoSOX Red labelling. Raw 264.7 cells were loaded with 100 nM green-fluorescent MitoTracker Green (MitoGreen, Invitrogen, M7514) for 30 min at 37 °C to measure mitochondrial content, followed by a wash with warmed complete culture medium. To examine mitochondrial reactive oxygen species (ROS) levels, cells were loaded with 5 μM MitoSOX Red (MitoSOX, Invitrogen, M36008) for 30 min at 37 °C, which is a mitochondrial superoxide indicator. Quantification of fluorescence intensity was carried out by using a BD LSR II flow cytometer (Becton Dickinson, Franklin Lakes, NJ, USA) with at least 10,000 events for each sample and analyzed with Becton Dickinson FACS Diva (version 6.1.2).

Oxygen consumption measurements. Raw 264.7 cells were treated with 100 ng/ml LPS for 4 h and then treated with or without 10 mM metformin for 16 h. The average intracellular oxygen consumption rate (OCR) was measured in a sealed chamber using a Clark-type S1 electrode (Oxygraph System, Hansatech Instruments) according to the manufacturer's instructions, as described [44].

Seahorse analysis. Mitochondrial function was evaluated with a Seahorse XF24 analyzer (Agilent, CA, USA). Raw 246.7 macrophages were seeded at a density of 8 × 10⁴ cells/well in XF-24 plates in RPMI medium supplemented with 10% heat-inactivated FBS, antibiotics (1% P/S; 100 U/ml penicillin, 100 μg/ml streptomycin) and glutamine (2 mM). The following day, cells were treated with LPS without or with metformin as described above. Briefly, culture medium was replaced with 500 μl RPMI supplemented with 1 mM HEPES, 1 mM sodium pyruvate, and 2 mM l-glutamine, and the plates were equilibrated in a non-CO₂ incubator at 37 °C. Metabolic toxins (0.5 μM oligomycin, 1 μM carbonyl cyanide p-trifluoromethoxy-phenylhydrazone [FCCP], 0.5 μM rotenone and antimycin A [Rot/AA]) were loaded into the cartridge drug ports. The cellular oxygen consumption rate (OCR) was measured in real-time by adding the drugs in order according to the manufacturer's protocol.

Determination of mitochondrial complex I activity. The activity of the respiratory chain complex I was measured as previously described [45] using a Beckman Coulter DU 800 spectrophotometer.

Quantification and statistical analysis. Statistical details of the experiments including statistical tests used, exact value of n, dispersion and precision measures (mean ± SEM), and statistical significance are reported in the Figures and Figure Legends. The differences between 2 groups with similar variances were analyzed using a two-tailed Student's t-test. When the variances of the groups were significantly different, a two-tailed t-test with Welch's correction was used. Differences between the variances of the groups were analyzed with the Fisher test (F-test). When more than 2 groups were analyzed, statistical analysis was performed by one-way ANOVA followed by Tukey's post hoc test. All statistical analyses were performed using GraphPad Prism 8 software. A p value lower than 0.05 was considered significant.

3. Results

Metformin reduces macrophage-mediated proinflammatory signaling cascades in brown adipocytes. Metformin is a widely prescribed drug for T2D that besides its well-known effects in insulin-sensitive tissues also reduces the proinflammatory responses of macrophages [33,46,47]. Based on that, we investigated the effect of metformin in the crosstalk between macrophages and brown adipocytes in the context of inflammation. To achieve this, we collected conditioned media from M1 (LPS-stimulated) Raw 264.7 macrophages without or with a further treatment with metformin (referred to as CM-Lps or CM-LpsMet, respectively) as detailed in Materials and Methods. As control, CM was collected from macrophages that did not receive treatment (CM-Ctl). Quantification of proinflammatory cytokines in the CM (Table S1) revealed a significant decrease in TNFα and IL6 content in the CM-LpsMet compared to levels detected in CM-Lps. Next, activation of proinflammatory signaling cascades was tested in brown adipocytes exposed to the different CM. As shown in Fig. 1, STAT3, JNK, and p38 MAPK phosphorylation was rapidly detected in brown adipocytes stimulated with CM-Lps, an effect that concurred with the degradation of IκBα. Notably, activation of proinflammatory signaling was markedly attenuated in brown adipocytes treated with CM-LpsMet.

Fig. 1.

Metformin attenuates macrophage-mediated proinflammatory signaling cascades in brown adipocytes. Differentiated brown adipocytes were treated with CM-Ctl, CM-Lps, or CM-LpsMet for the indicated time-periods. Representative Western blots of phospho-STAT3, phospho-JNK, JNK, IκBα, phospho-p38 MAPK, p38 MAPK, and Vinculin levels and quantification (n = 3 independent experiments). Results are expressed as fold change versus CM-Ctl condition. *p < 0.05, **p < 0.01, ***p < 0.001 versus CM-Ctl; ^#p < 0.05, ^##p < 0.01, ^###p < 0.001 versus CM-Lps with one-way ANOVA followed by Tukey's post hoc.

CM-LpsMet prevented the impairment in insulin signaling in brown adipocytes induced by CM-Lps. Several studies have demonstrated that activation of proinflammatory signaling cascades interferes with the early steps of insulin signaling by phosphorylating IRS1 at serine residues, thereby triggering its degradation via 26S proteasome [48,49]. In this regard, we previously reported that cytokines secreted by M1-polarized macrophages induced insulin resistance in hepatocytes [50]. As depicted in Fig. 2A, stimulation of brown adipocytes with CM-Lps led to a rapid and transient phosphorylation of IRS1 at serine 307 peaking at 30 min and declining after that. This effect concurred with a marked reduction in IRS1 protein levels 24 h after treatment with the proinflammatory CM-Lps (Fig. 2B). Both IRS1 serine 307 phosphorylation at its subsequent degradation were abolished in brown adipocytes receiving CM-LpsMet (Fig. 2A and B). To monitor insulin responses downstream of IRS1, CM-Lps, CM-LpsMet or CM-Ctl were added to brown adipocytes for 24 h and then, cells were stimulated with 10 nM insulin for a further 10 min. Fig. 2C shows a substantial decrease in insulin-induced Akt phosphorylation at both serine 473 and threonine 308 residues in brown adipocytes preincubated with CM-Lps compared to those treated with CM-Ctl or CM-LpsMet. Similar effect of the CM-Lps in decreasing IRS1 and insulin-induced Akt phosphorylation in brown adipocytes and the CM-LpsMet in preventing these responses was found when the CM was generated in BMDM (Fig. S1A).

Fig. 2.

The impairment of insulin signaling in brown adipocytes treated with CM-Lps was prevented in brown adipocytes exposed to CM-LpsMet.A) Differentiated brown adipocytes were treated with CM-Ctl, CM-Lps or CM-LpsMet for the indicated time-periods. Representative Western blot of phospho-IRS1 (Ser307), IRS1 and Vinculin protein levels and quantification (n = 3 independent experiments). Results are expressed as fold change versus CM-Ctl. *p < 0.05, **p < 0.01 versus CM-Ctl; ^##p < 0.01 versus CM-Lps. B) IRS1 and Vinculin levels in brown adipocytes treated with CM-Ctl, CM-Lps or CM-LpsMet for 24 h and quantification (n = 3 independent experiments). ***p < 0.001 versus CM-Ctl; ^###p < 0.001 versus CM-Lps. C) Differentiated brown adipocytes were treated with CM-Ctl, CM-Lps or CM-LpsMet for 24 h followed by stimulation with 10 nM insulin for 10 min. Representative Western blots using the indicated antibodies and respective quantification (n = 4 independent experiments). *p < 0.05, ***p < 0.001 versus without insulin; ⁺p<0.05, ⁺⁺⁺p<0.001 versus CM-Ctl; ^#p < 0.05, ^##p < 0.01 versus CM-Lps. D) Glucose uptake under conditions described in C (n = 3 independent experiments performed in duplicate). *p < 0.05 versus without insulin. E) Representative GLUT4 immunofluorescence images in differentiated brown adipocytes treated with CM-Ctl, CM-Lps or CM-LpsMet for 18 h followed by stimulation with 100 nM insulin for 10 min (nuclei in blue (DAPI), GLUT4 in green). Scale bars, 10 μm. F) mRNA levels of Slc2a1 after 18 h of treatment with CM-Ctl, CM-Lps or CM-LpsMet (n = 4 independent experiments performed in duplicate). **p < 0.01 versus CM-Ctl; ^##p < 0.01 versus CM-Lps. G) Representative GLUT1 immunofluorescence images of differentiated brown adipocytes treated with CM-Ctl, CM-Lps or CM-LpsMet for 18 h (nuclei in blue (DAPI), GLUT1 in green). Scale bars, 10 μm. Data are shown as mean ± SEM. Statistical analysis was performed using two-tailed unpaired t-test when comparing two groups and one-way ANOVA followed by Tukey's post hoc test when comparing three or more groups. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Since insulin induces glucose uptake in brown adipocytes in a PI-3-kinase/Akt-dependent manner [6], we assessed the effect of the inflammatory environment and metformin in this response. Whereas almost a three-fold increase in glucose uptake was found in insulin-stimulated brown adipocytes upon pretreatment with CM-Ctl, this effect was markedly attenuated in cells pretreated with CM-Lps (Fig. 2D) and correlated with a decrease in GLUT4 translocation to the plasma membrane (Fig. 2E). Of note, CM-Lps increased basal (insulin-independent) glucose uptake (Fig. 2D) in parallel with an increase in Glut1 (Slc2a1) mRNA levels and GLUT1 immunostaining (Fig. 2F and G) resulting in only about 1.4-fold increase in insulin-dependent glucose uptake. Brown adipocytes treated with CM-LpsMet did not show impairments in insulin-induced glucose uptake and GLUT4 translocation to the plasma membrane and also GLUT1 up-regulation was not observed. Importantly, in these cells the increase in insulin-dependent glucose uptake was comparable to the controls (Fig. 2D, E, 2F, 2G).

CM-Lps, but not CM-LpsMet, decreased β-adrenergic-mediated lipolysis and UCP1 levels in brown adipocytes. The impact of inflammation was evaluated in β-adrenergic stimulated lipolysis as an early trigger of brown adipocyte thermogenic response. Brown adipocytes were stimulated with the β3-adrenergic agonist CL316243 (2 μM) for 3 h in the presence of CM-Lps, CM-LpsMet or CM-Ctl. Fig. 3A shows that cells treated with CM-Ctl were sensitive to CL316243 manifested by hormone-sensitive lipase (HSL) phosphorylation and, consequently, the release of glycerol to the culture medium (Fig. 3B). Both responses were significantly altered in brown adipocytes treated with CL316243 and CM-Lps, but not CM-LpsMet. We also analyzed the effect of the proinflammatory environment in lipolysis by performing PLIN1 (Perilipin 1) immunofluorescence and Bodipy staining. As shown in Fig. 3C, smaller lipid droplets surrounded by PLIN1 were found in brown adipocytes co-treated with CL316243 and CM-Ctl. In cells exposed to CM-Lps and CL316243 lipid droplet size was similar to those of unstimulated cells. However, in the presence of CM-LpsMet, brown adipocytes maintained their β3-adrenergic lipolytic response manifested by the smaller size of their lipid droplets. The reduced sensitivity of brown adipocytes to β3-adrenergic stimulation by the CM-Lps and the protection by CM-LpsMet was also evident in Ppargc1a (encoding PGC1α) mRNA levels and UCP1 protein content (Fig. 3D and E). As occurred with the insulin response, the effect of the CM-Lps in decreasing CL316243-mediated UCP1 expression and the protective effect of CM-LpsMet was found in brown adipocytes exposed to the CM generated by BMDM (Fig. S1B).

Fig. 3.

CM-Lps, but not CM-LpsMet, decreased β3-adrenergic-mediated lipolysis and UCP1 levels in brown adipocytes.A) Representative Western blots of phospho-HSL, HSL and Vinculin levels and quantification in differentiated brown adipocytes treated with CM-Ctl, CM-Lps or CM-LpsMet for 18 h followed by stimulation with 2 μM CL316243 for the last 3 h (n = 6 independent experiments). Results are expressed as fold change versus CM-Ctl. B) Glycerol released to the culture medium was measured as described in Materials and Methods (n = 4 independent experiments performed in triplicate). ***p < 0.001 versus without CL316243; ⁺p<0.05 versus CM-Ctl with CL316243; ^#p < 0.05 versus CM-Lps with CL316243. C) Representative images of immunofluorescence of PLIN1 (green) in differentiated brown adipocytes upon treatment with CM-Ctl, CM-Lps or CM-LpsMet for 24 h and CL316243 (2 μM) during the last 8 h. BODIPY 493/503 staining before fixation is shown in red. Scale bars, 10 μm. D) mRNA levels of Ppargc1a in differentiated brown adipocytes treated with CM-Ctl, CM-Lps or CM-LpsMet for 18 h (n = 4 independent experiments performed in duplicate). ***p < 0.001 versus CM-Ctl; ^##p < 0.01 versus CM-Lps. E) Representative Western blot images of UCP1 and Vinculin levels and quantification in brown adipocytes stimulated with 2 μM CL316243 in the presence of CM-Ctl, CM-Lps or CM-LpsMet for 18 h (n = 5 independent experiments). ***p < 0.001 versus without CL316243; ⁺⁺⁺p<0.001 versus CM-Ctl + CL316243; ^##p < 0.01 versus CM-Lps + CL316243. Data are shown as mean ± SEM. Statistical was performed using one-way ANOVA followed by Tukey's post hoc test. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Treatment of obese mice with metformin reduced BAT inflammation and increased cold responsiveness. We aimed to provide a physiological relevance to the in vitro findings in obese C57BL/6J mice treated or not with a daily dose of metformin (100 mg/kg) for 6 weeks. Treatment of obese mice with metformin reduced glucose levels and body weight (Fig. S2). As shown in Fig. 4A, metformin administration decreased the phosphorylation of STAT3, JNK and IκB, all readouts of proinflammatory signaling, in BAT. Likewise, treatment with metformin reduced Il6 and Il1b mRNA levels in BAT from obese mice (Fig. 4B), revealing a decrease in tissue inflammation. A step further, in a cohort of mice exposed to 4 °C (CE) for 6 h, Ppargc1a and Ucp1 mRNA levels were elevated in BAT to a higher extend in mice that received metformin prior to cold exposure (Fig. 4C and D). This effect was confirmed by UCP1 immunohistochemistry in BAT sections (Fig. 4E).

Fig. 4.

Treatment of obese mice with metformin reduced BAT inflammation and increased cold responsiveness.A) Representative Western blot analysis of phosho-STAT3, phospho-JNK, JNK, phospho-IκBα and Vinculin in BAT from mice fed a HFD without (HFD) or with metformin treatment (HFD-Met) (n = 4 mice/group). *p < 0.05 versus HFD. B) mRNA levels of Il6 and Il1b in BAT of HFD and HFD-Met groups (n = 4 mice/group). *p < 0.05, **p < 0.01 versus HFD. C, D) mRNA levels of Ppargc1a and Ucp1 in BAT from HFD and HFD-Met mice exposed or not to 4 °C (CE) for 6 h (n = 4 mice/group). **p < 0.01, ***p < 0.001 versus non-cold exposure; ^###p < 0.001 versus HFD cold exposure (CE). E) Representative images of UCP1 immunohistochemistry in BAT from HFD and HFD-Met mice exposed or not to 4 °C for 6 h (n = 4 mice/group). Data are shown as mean ± SEM. Statistical analysis was performed using two-tailed unpaired t-test when comparing two groups and one-way ANOVA followed by Tukey's post hoc test for multiple comparisons.

Stabilization of HIF1α is sufficient and necessary to decrease β-adrenergic responses in brown adipocytes. Due to the existence of a direct relationship between hypoxia and inflammation at the level of HIF1α-mediated transcriptional regulation of proinflammatory mediators [51,52], we tested whether HIF1α was stabilized in brown adipocytes upon treatment with the proinflammatory CM-Lps. Fig. 5A shows HIF1α stabilization in brown adipocytes treated with CM-Lps, an effect markedly attenuated in cells receiving CM-LpsMet. Likewise, expression levels of the proinflammatory markers Tnfa, Il6, Il1b and Nos2 were elevated in brown adipocytes exposed to CM-Lps, pointing to a secondary inflammatory response in brown adipocytes that significantly decreased in those receiving CM-LpsMet. Increased HIF1α immunostaining was detected in the nucleus of both brown adipocytes (rounded) and immune cells (smaller and enlarged) in BAT sections from obese mice (Fig. 5B) and, importantly, these two different nuclear HIF1α signals were not detected in metformin-treated mice. These results prompted us to hypothesize that stabilization of HIF1α in differentiated brown adipocytes could mimic the impairment of insulin and β-adrenergic responses of the proinflammatory CM-Lps. To test this, brown adipocytes were transduced with retroviral particles expressing a HIF1α mutant resistant to degradation by PHDs [39,53]. Of note, HIF1α overexpression reached levels comparable to those of brown adipocytes treated with CM-Lps (Fig. S3). We first tested the response of brown adipocytes to CL316243. As shown in Fig. 5C, CL316243-mediated induction of Ppargc1a mRNA and UCP1 protein expression were reduced in brown adipocytes upon ectopic expression of HIF1α. Since these results indicated that stabilization of HIF1α is sufficient to impair thermogenic-related expression in brown adipocytes, we next evaluated whether endogenous HIF1α stabilization by CM-Lps-mediated inflammation is also responsible for the decline of UCP1. Fig. 5D shows that control brown adipocytes (BA^Hif1afl/fl) did not respond to CL316243 in the presence of CM-Lps while HIF1α-deficient brown adipocytes (BA^Hif1aΔ/Δ) retained CL316243-dependent UCP1 expression. These results suggest that HIF1α expressed in brown adipocytes is critical for the impairment of β-adrenergic-dependent UCP1 expression in an inflammatory context.

Fig. 5.

Stabilization of HIF1α was sufficient to decrease insulin and β3-adrenergic responses in brown adipocytes. A)(left panel) Representative Western blots of HIF1α and Vinculin levels and quantification in differentiated brown adipocytes treated with CM-Ctl, CM-Lps and CM-LpsMet for 18 h (n = 3 independent experiments). (right panel) mRNA levels of Tnfa, Il6, Il1b and Nos2 under similar experimental conditions (n = 3 independent experiments). ***p < 0.001 versus CM-Ctl; ^###p < 0.001 versus CM-Lps. B) Representative images of HIF1α immunohistochemistry in BAT from HFD and HFD-Met mice (n = 4 mice/group). Black square: amplification image showing brown adipocyte nuclei. Red square: amplification image showing immune cell nuclei. C) mRNA levels of Ppargc1a (upper panel) and representative Western blot analysis of HIF1α, UCP1 and Vinculin levels and quantification (lower panel) in differentiated brown adipocytes expressing HIF1α(P2)* or GFP upon treatment or not with 2 μM CL316243 for 18 h (n = 3 independent experiments). *p < 0.05, ***p < 0.001 versus GFP D) BA^Hif1afl/fl and BA^Hif1aΔ/Δ were treated with CM-Ctl and CM-Lps for 18 h in the absence or presence of 2 μM CL316243 and the expression of HIF1α, UCP1, and Vinculin was analyzed by Western blot (n = 5 independent experiments). **p < 0.01, ***p < 0.001 versus without CL316243; ⁺⁺⁺p<0.001 versus CM-Ctl + CL316243. E) (upper panel) Representative Western blot and quantification of HIF1α, IRS1 and Vinculin levels in differentiated brown adipocytes expressing HIF1α(P2)* and their corresponding GFP controls (n = 3 independent experiments). (middle panel) mRNA levels of Il6, Tnfa and Nos2 (n = 3 independent experiments). *p < 0.05, **p < 0.01 versus GFP. (lower panel) Representative Western blot of HIF1α, phospho-Akt (Ser473), Akt and Vinculin levels and quantification in brown adipocytes overexpressing HIF1α(P2)* and GFP control cells upon stimulation with 10 nM insulin for 10 min (n = 4 independent experiments). ***p < 0.001 versus without insulin; ^##p < 0.01 versus GFP + Insulin. F) (upper panel) Representative Western blot analysis of IRS1 and Vinculin protein levels and quantification in BA^Hif1fl/fl and BA^Hif1aΔ/Δ treated with CM-Ctl and CM-Lps for 24 h *p < 0.05, **p < 0.01 versus CM-Ctl. (lower panel) Representative Western blot of phosho-Akt (Ser473), Akt and Vinculin levels and quantification in BA^Hif1afl/fl and BA^Hif1aΔ/Δ cells treated with CM-Ctl and CM-Lps for 18 h followed by stimulation with 10 nM insulin for 10 min (n = 5 independent experiments). **p < 0.01 versus without insulin. Data are shown as mean ± SEM. Statistical analysis was performed using two-tailed unpaired t-test when comparing two groups and one-way ANOVA followed by Tukey's post hoc test for multiple comparisons. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Proinflammatory environment leads to HIF1α-dependent and -independent suppression of insulin signaling in brown adipocytes. We next addressed the relevance of HIF1α in the impairment of insulin signaling under proinflammatory conditions. Ectopic expression of HIF1α in brown adipocytes markedly reduced IRS1 protein levels, as well as insulin-mediated Akt phosphorylation (Fig. 5E). Of note, Il6, Tnfa and Nos2 mRNAs were elevated in HIF1α-transduced brown adipocytes (Fig. 5E), as occurred upon exposure to CM-Lps (Fig. 5A), suggesting that macrophage-dependent inflammatory signals activate a secondary inflammatory response in nearby brown adipocytes characterized by HIF1α-dependent Il6, Tnfa, and Nos2 expression. Since it has been reported IRS1 serine 307 phosphorylation-mediated IRS1 degradation by inflammatory cytokines [48] and, therefore, not involving transcriptional mechanisms, we wanted to explore whether HIF1α stabilization is still relevant to reduce IRS1 levels in the presence of CM-Lps. As shown in Fig. 5F, IRS1 degradation and impaired insulin-induced Akt phosphorylation were also found in HIF1α-deficient brown adipocytes (BA^Hif1aΔ/Δ) exposed to CM-Lps. In fact, these cells lacking full-length HIF1α still showed marked activation of proinflammatory signaling and phosphorylation of IRS1 at serine 307 in the presence of CM-Lps (Fig. S4). These data led us to suggest that although HIF1α stabilization per se is sufficient to impair insulin signaling, the proinflammatory cytokine-enriched CM overrides HIF1α due to the ability to trigger IRS1 serine 307 phosphorylation, IRS1 degradation and an impairment of insulin signaling. In fact, direct addition of a cocktail of TNFα, IL6 and IL1β to brown adipocytes led to IRS1 degradation, resulting in impaired insulin-mediated Akt phosphorylation (Fig. S5).

The protective effect of metformin in the crosstalk between macrophages and brown adipocytes does not require AMPKα1. Interestingly, short-term stimulation of brown adipocytes with CM-LpsMet transiently increased AMPK phosphorylation (Fig. S6A). Notably, CM-LpsMet does not contain metformin since the drug was washed out as detailed in Materials and Methods. Moreover, as shown in Fig. S6B, the effect of CM-LpsMet in inducing AMPK phosphorylation in brown adipocytes was comparable to the response of brown adipocytes treated with CM-Ctl supplemented with metformin. To directly address if AMPK was required for the effect of CM-LpsMet in the protection of brown adipocytes against an inflammatory environment, we immortalized and differentiated brown adipocytes from AMPKα1-deficient mice (referred as BA^Ampka1−/-) (Fig. S6C) that, as shown in Fig. S6D, preserved the capacity to fully differentiate, as well as maintained insulin and CL316243 sensitivity in the presence of CM-LpsMet (Figs. S6E and S6F). These results suggest that in brown adipocytes AMPKα1 is dispensable to preserve the hormonal responses in the presence of proinflammatory CM-LpsMet.

Metformin alleviated BAT inflammation by promoting HIF1α degradation in macrophages. Since on the one hand, the secretome of macrophages receiving LPS and metformin protected brown adipocytes against inflammation and, on the other, treatment of HFD-fed mice with metformin reduced HIF1α levels in BAT, we postulated that metformin could modulate HIF1α in BAT macrophages and therefore, alleviate inflammation. In fact, Fig. 5B showed HIF1α immunostaining in immune cells in BAT. We also performed double HIF1α and F4/80 (as a marker of BAT macrophages) immunofluorescence in BAT sections. As shown in Fig. S7, HIF1α and F4/80 co-localized in areas of macrophages infiltration in BAT from obese mice similar to what was found in WAT [54], and this effect was ameliorated by the treatment with metformin. These results correlated with the absence of HIF1α and iNOS protein levels in macrophages treated with LPS and metformin (Fig. 6A). Moreover, the effects of LPS and metformin in HIF1α and iNOS expression were similar in BMDM from mice lacking AMPKα1 (Fig. S6G). On the contrary, the effect of metformin in preventing LPS-mediated HIF1α stabilization in macrophages was abolished in the presence of the proteasome inhibitor MG132, suggesting that metformin promotes HIF1α degradation via proteasome (Fig. 6A).

Fig. 6.

Metformin alleviated inflammation in macrophages independently of AMPKα1.A). (Left panel) Representative Western blots of HIF1α, iNOS and Vinculin levels in BMDM from AMPKα1-deficient mice treated with LPS (Lps) or LPS and Metformin (LpsMet) as indicated in material and methods (n = 3 independent experiments). ***p < 0.001 versus Ctl; ^###p < 0.001 versus Lps. (Right panel) Representative Western blots of HIF1α and Vinculin levels in Raw 246.7 macrophages treated as indicated in materials and methods. Proteasome activity was inhibited by addition of 1 μM MG-132 for 4 h before cell lysis. ***p < 0.001 versus Ctl; ⁺⁺p<0.01 versus Lps; ^###p < 0.001 versus LpsMet. B) (Left panel) Changes in mitochondrial complex I activity and (Right panel) oxygen consumption in Raw 264.7 macrophages in response to Lps or Lps and Metformin (LpsMet) treatment (n = 3 independent experiments). ***p < 0.001 versus Ctl, **p < 0.01 versus Ctl, ^##p < 0.01 versus Lps, ^#p < 0.05 versus Lps. C) (Left panel) Real-time changes in OCR of Raw 246.7 macrophages treated with Lps, Lps and Metformin or Metformin alone before treatment with oligomycin (Olig) (0.5 μM), FCCP (1 μM), and rotenone and antimycin A (ROT + AA) (0.5 μM). (Right panel) Basal respiratory capacity of Raw 246.7 macrophages treated with Lps, Lps and Metformin or Metformin alone (indicated in Material and Methods) before the starting Seahorse determinations. Average and standard deviation of 8 individual wells from two experiments is shown. Results are expressed as pmol/min/μg total protein. **p < 0.01 and ***p < 0.001 by two-way ANOVA compared with Ctl or ^###p < 0.001 compared with Lps. D) Mitochondrial mass and E) MitoSOX positive cells analysis in Raw 246.7 macrophages treated with Lps or Lps and Metformin (LpsMet) (n = 4 independent experiments). Data are shown as mean ± SEM. Statistical analysis was performed using two-tailed unpaired t-test when comparing two groups and one-way ANOVA followed by Tukey's post hoc test for multiple comparisons.

Then, we considered whether inhibition of mitochondrial complex I and oxygen consumption by metformin might redistribute the spared oxygen to reactivate PHDs leading to HIF1α degradation through the proteasome pathway. To further evaluate this possibility, we measured the activity of mitochondrial complex I and oxygen consumption rate (OCR) in Raw 267.4 macrophages. As shown in Fig. 6B, both parameters did not decrease significantly by LPS compared with the control cells, but were abrogated when metformin was added on LPS-treated cells. Similar modulation of the basal OCR by LPS and metformin in Raw 267.4 macrophages was observed in the Seahorse profile (Fig. 6C). In addition, mitochondrial mass and ROS production were analyzed by flow cytometry using Mitotracker Green and MitoSOX, respectively. The results revealed no significant differences among Raw 264.7 macrophages untreated or treated with LPS in absence or presence of metformin (Fig. 6D and E). Therefore, it might be suggested that exposure to metformin by inhibiting mitochondrial complex I leads to increased oxygen availability (not consumed by mitochondria) for the prolyl-4-hydroxylase domain (PHD) oxygen sensors, which leads to enhanced HIF1α degradation. To reinforce this hypothesis, Raw 264.7 macrophages were treated with LPS in the absence or presence of the well-recognized complex I inhibitor rotenone. As shown in Figs. S8A and S8B, the combination of LPS and rotenone produced a similar effect in oxygen consumption and over HIF1α protein levels than that induced by LPS plus metformin (Fig. 6A and B), supporting the limitation of oxygen consumption by metformin as a trigger for HIF1α degradation in macrophages. A step further and in agreement with the results with metformin, when brown adipocytes were treated with the CM collected from Raw 264.7 macrophages stimulated with LPS plus rotenone (CM-LpsRtn), IRS1 levels were preserved as well as the response to insulin and CL316243 (Figs. S8C and S8D).

Inhibition of HIF1α in macrophages reduced the proinflammatory potency of CM-Lps and protected against insulin resistance in brown adipocytes. Since we found that the negative impact of HIF1α on insulin signaling in brown adipocytes is overridden by proinflammatory cytokines (Fig. S4), we tested whether HIF1α inhibition in macrophages might have a more profound effect. For this goal, Raw 264.7 macrophages were exposed to LPS together with a HIF1α inhibitor which represses its transcriptional activity [55,56]. Fig. 7A shows that co-treatment of macrophages with LPS and the HIF1α inhibitor blunted iNOS expression. We also determined the proinflammatory profile of the macrophages and found decreased Il6 and Il1b mRNA levels in those co-treated with LPS and the HIF1α inhibitor (Fig. 7B) in parallel to a decrease in IL6 and IL1β content of the CM (referred as CM-InhLps) (Table S1).

Fig. 7.

Inhibition of HIF1α in macrophages reduced the proinflammatory potency of CM-Lps and protected against insulin resistance in brown adipocytes. A) Representative Western blots of HIF1α, iNOS and Vinculin levels in Raw 246.7 cells treated with LPS in the absence (Lps) or presence (InhLps) of a HIF1α inhibitor as indicated in material and methods (n = 3 independent experiments). **p < 0.01 versus Ctl; ^#p < 0.05 versus Lps. B) mRNA levels of Il6 and Il1b under similar experimental conditions (n = 3 independent experiments performed in duplicate). ***p < 0.001 versus Ctl; ^##p < 0.01 versus Lps. C) Representative Western blots of phospho-STAT3, phospho-JNK, JNK, phospho-p38 MAPK and p38 MAPK and quantification in brown adipocytes treated with CM-Ctl, CM-Lps or CM-InhLps for several time-periods (n = 3 independent experiments). ***p < 0.001, **p < 0.01, *p < 0.05 versus CM-Ctl; ^#p < 0.05, ^##p < 0.01 versus CM-Lps. D) Representative Western blots of phospho-Akt (Ser473), phospho-Akt (Thr308), Akt and Vinculin levels and quantification in brown adipocytes treated with CM-Ctl, CM-Lps or CM-InhLps for 18 h followed by stimulation with 10 nM insulin for 10 min (n = 5 independent experiments). **p < 0.01, *p < 0.05 versus without insulin; ⁺⁺p<0.01 versus CM-Ctl + Insulin; ^#p < 0.05 versus CM-Lps + Insulin. E) Representative Western blots of UCP1 and vinculin protein levels and quantification in brown adipocytes stimulated with 2 μM CL316243 in the presence of CM-Ctl, CM-Lps or CM-InhLps for 18 h (n = 4 independent experiments). *p < 0.05, ***p < 0.001 versus without CL316243; ⁺p<0.05 versus CM-Ctl + CL316243. Data are shown as mean ± SEM. Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test.

Next, we analyzed the effect of HIF1α inhibition in macrophages in the crosstalk with brown adipocytes. Fig. 7C shows that STAT3, JNK and p38 MAPK phosphorylation in brown adipocytes stimulated with CM-Lps was attenuated when macrophages were treated with the HIF1α inhibitor. Consistently, the drop in insulin-induced Akt phosphorylation found in brown adipocytes exposed to CM-Lps was absent in the case of CM-InhLps (Fig. 7D). Also, brown adipocytes exposed to CM-InhLps were able to respond to CL316243 regarding UCP1 expression (Fig. 7E). Altogether, these results point to the relevance of HIF1α in the proinflammatory fate of macrophages leading to a secondary HIF1α activation in brown adipocytes that negatively modulates insulin and β-adrenergic responses of brown adipocytes.

4. Discussion

The molecular and cellular mechanisms involved in BAT dysfunction in the context of obesity are not completely understood. In particular, the interplay between macrophages and brown adipocytes is largely unknown. Herein, we provide in vitro settings that recapitulate precisely in vivo BAT dysfunction regarding insulin and adrenergic responsiveness, as well as the therapeutic benefit of metformin. Specifically, our results in vitro support the in vivo findings in obese mice regarding i) stabilization of HIF1α in BAT macrophages and reduction by metformin. ii) activation of proinflammatory signaling cascades and elevation of Il1b and Il6 mRNA levels in BAT; both effects being also attenuated by metformin. iii) restoration by metformin of BAT responsiveness to cold exposure. Overall, our study has revealed relevant molecular insights in the role of macrophages in BAT dysfunction in obesity using both in vitro and in vivo settings.

Although HIF1α stabilization is mainly dependent on oxygen concentration, other studies have reported an oxygen-independent regulation of HIF1α by hyperglycemia, oxidative stress and, importantly, inflammatory mediators [[57], [58], [59]]. Our results have confirmed this alternative HIF1α induction under normoxic conditions in brown adipocytes exposed to the secretome of M1-polarized macrophages (CM-Lps) concurrently with intracellular pathways (i.e. STAT3, JNK, p38 MAPK and NFκB) and gene expression (Tnfa, Il6, Il1b and Nos2) associated to proinflammatory responses. This suggests that, under normoxic conditions, but in the context of local inflammation due to macrophage activation, brown adipocytes respond with HIF1α stabilization and induction of proinflammatory genes as occurred in white adipocytes [2,4,30,31]. As an additional sign of HIF1α activation, we also found that brown adipocytes exposed to the proinflammatory CM-Lps increased the expression levels of GLUT1, a direct target of HIF1α, resulting in elevated basal (insulin-independent) glucose uptake. A step further, inflammation-mediated HIF1α induction in brown adipocytes was able to attenuate insulin responses such as Akt phosphorylation and, hence, GLUT4 translocation to the plasma membrane and insulin-dependent glucose uptake. Both GLUT1 and glucose uptake were found elevated in human adipose stromal cells exposed to a mixture of TNFα, IL6 and IL1β [60] or in human and 3T3L1 adipocytes exposed to hypoxia [61,62]. Moreover, previous in vivo studies reported decreased Akt phosphorylation and glucose intolerance by HIF1α overexpression in adipose tissue [30]. Similarly, other studies conducted in adipocyte-specific HIF1α-deficient mice or mice treated with a HIF1α inhibitor showed protection against HFD-induced insulin resistance [30,32,[63], [64], [65], [66], [67], [68]].

Of relevance, in brown adipocytes, the proinflammatory CM-Lps led to an increase in IRS1 phosphorylation at the modulatory serine 307 and the subsequent downregulation of its protein content as occurred with other metabolic stressors [48,69]. Our results show for the first time the downregulation of IRS1 protein by the ectopic expression of HIF1α in brown adipocytes concurrently with an elevation in Tnfa, Il6, and Nos2 expression. Therefore, in brown adipocytes HIF1α-mediated elevation of proinflammatory cytokines could act in an autocrine manner, which might explain the downregulation of IRS1 and the subsequent decrease in insulin signaling. In addition, our data suggest that the proinflammatory cytokines present in the CM-Lps also impair insulin signaling involving other pathways such as a the rapid IRS1 serine 307 phosphorylation that overrides the impact of the HIF1α transcription program on IRS1 stability. Indeed, the CM-Lps was still competent to reduce IRS1 protein levels in HIF1α-deficient brown adipocytes, and the solely deletion of HIF1α was not able to rescue brown adipocytes from insulin resistance. These data were supported by the protection against both HIF1α stabilization and inflammation-mediated IRS1 phosphorylation at serine 307 found in brown adipocytes exposed to the CM from macrophages treated with LPS and metformin (CM-LpsMet) that efficiently protected against proinflammatory signaling-mediated IRS1 degradation and insulin resistance. Interestingly, HIF1α-mediated decrease in IRS1 contrasts with the role of HIF2α in the liver positively modulating IRS2 expression and improving insulin sensitivity [70]. Thus, it seems that the two HIF isoforms might have contrasting actions on insulin signaling in a tissue-specific manner. Of relevance, in contrast to the beneficial effects of HIF1α deficiency, adipocyte-specific deletion of HIF2α in mice on a HFD increased inflammation and exacerbated insulin resistance [32]. In this line, antagonistic physiological or pathological functions of HIF isoforms have been reported in other cell types, including macrophages in which HIF1α modulates iNOS whereas HIF2α regulates arginase-1, readouts of M1 and M2 polarization, respectively [71].

In addition, the proinflammatory conditions surrounding brown adipocytes impaired the early steps of β-adrenergic signaling (i.e. HSL phosphorylation) and the lipolytic response as visualized by the higher size of the lipid droplets in brown adipocytes receiving CM-Lps and CL316243 compared to the controls. As expected from the impaired β-adrenergic signaling, Ppargc1a mRNA and UCP1 protein levels were significantly reduced in agreement with data in C3H10T1/2 (10T1/2) mouse mesenchymal stem cells [72] and our recent study [36]. Of relevance, herein we show that, in brown adipocytes, ectopic expression of HIF1α is sufficient to suppress their responses to CL316243. As such, either BA^Hif1aΔ/Δ exposed to CM-Lps or wild-type brown adipocytes exposed to CM-LpsMet, which markedly reduced HIF1α levels, maintained β-adrenergic responsiveness, suggesting that stabilization of HIF1α is necessary and sufficient for attenuating the thermogenic responses of brown adipocytes in the context of inflammation. In this regard, our in vitro results could explain the reduced thermogenic function of BAT found in mice overexpressing HIF1α in adipose tissue [64].

An unresolved issue is whether inflammation in adipose tissue is an early trigger for HIF1α stabilization and if the subsequent HIF1α-mediated (hypoxia-related) gene response worsens the metabolic dysfunction in this tissue during obesity. In this line, our data show that the proinflammatory mediator LPS has the potential to induce primarily HIF1α in activated macrophages, which release cytokines that subsequently induce HIF1α in brown adipocytes under normoxic conditions. These data point out to the role of inflammation – and not necessarily hypoxia – in the activation of HIF1α in BAT tissue of obese mice. Accordingly, in BAT sections from HFD-fed mice we observed HIF1α immunostaining in macrophages that was nearly absent in mice treated with metformin, concomitantly with attenuated proinflammatory signaling cascades in this fat depot. Thus, our results revealed for the first time that, in addition to the role of HIF1α in brown adipocytes, stabilization of HIF1α in tissue macrophages is a key feature of BAT inflammation in obesity. In this regard, the expression of HIF1α in myeloid cells promotes inflammatory responses in WAT and tissue remodeling towards insulin resistance [73,74]. Also, Li and co-workers [73] associated the antiinflammatory effect of metformin with a reduction of HIF1α levels in WAT, thereby preventing inflammation and local fibrosis. These data prompted us to analyze the impact of metformin in the HIF1α-dependent proinflammatory profile of macrophages. We found that both the HIF1α inhibitor and metformin attenuated the proinflammatory responses in M1-(LPS-polarized) macrophages, as reported in other studies [33,46,47]. Moreover, our results showed that metformin promotes HIF1α degradation in macrophages via proteasome due to its well-known effect in reducing mitochondrial oxygen consumption by inhibiting mitochondrial complex I [5,75] without affecting mitochondrial ROS generation. Our data are in line with the study of Li et al. [73] showing an effect of metformin in 3T3L1 adipocytes by limiting oxygen consumption, thereby preventing hypoxia. This effect was further evaluated using rotenone, a well-known specific inhibitor of mitochondrial complex I and oxygen consumption. Indeed, rotenone, at doses that abrogated oxygen consumption, markedly reduced HIF1α accumulation in LPS-treated Raw 264.7 macrophages. Therefore, these results strongly suggest that metformin and rotenone decrease HIF1α by reducing mitochondrial oxygen consumption. Moreover, addition of rotenone as well as metformin to M1 polarized macrophages generated CM that prevented the attenuation of insulin and β-adrenergic responses in brown adipocytes. It is also important to highlight that the effect of metformin through inhibition of complex 1 is controversial [76] and, in this regard, the inhibition of hepatic gluconeogenesis by this drug independently of complex 1 activity has been reported [77].

Inflammation in BAT leads to the impairment of insulin and β-adrenergic responses in vivo, as we and others reported in db/db mice and mice injected LPS [30,36,55]. Regarding the impact of metformin in the interactome between macrophages and brown adipocytes in BAT functionality in obesity, we found an increase in Ppargc1a and Ucp1 mRNAs, as well as UCP1 protein expression upon cold exposure in mice treated with metformin for 6 weeks, an effect likely related to the inhibition of HIF1α. A recent study associated the effect of metformin in HFD-fed mice with lipid metabolism in BAT by reducing Cpt1b and Cpt2 expression [78]. These authors conclude that this reduction might attenuate diet-induced thermogenesis; although, this contrasts with the well-known relevance of fatty acid oxidation for BAT thermogenesis. Since our results evidence an increase in UCP1 upon cold exposure in BAT from HFD-fed mice receiving metformin and also an effect of this drug in attenuating HIF1α immunostaining in macrophages as discussed above, we suggest an additional effect of metformin in BAT metainflammation by targeting HIF1α in resident macrophages. Indeed, treatment of M1-polarized macrophages with either metformin or a HIF1α inhibitor led to the production of CM that did not increase HIF1α (in the case of CM-LpsMet) or triggered proinflammatory signaling cascades when added to brown adipocytes, thereby favoring insulin and β-adrenergic responsiveness. For metformin, this effect concurred with a reduction of HIF1α in both macrophages and brown adipocytes, independently of AMPKα1. However, we cannot exclude a role for AMPKα2 in the effects of CM-LpsMet in brown adipocytes since this subunit is also expressed in BAT [79]. In addition to the mechanistic insights provided herein in cell-based systems, additional in vivo research will be needed to prove that metformin directly acts on BAT macrophages.

Evidence reported so far that metformin may influence tumor progression by modulating the responses of the tumor microenvironment, including reduction of tumor-associated macrophages [80,81]. Nevertheless, to the best of our knowledge, this is the first study reporting the protective effect of this antidiabetic drug in brown adipocytes specifically by targeting macrophages and reducing HIF1α-dependent brown adipocyte dysfunction. These results might have therapeutic relevance to alleviate BAT metainflammation in obesity.

Authors contribution

Conceptualization, A.M.V., N.P.; Funding Acquisition, A.M.V., M.J.O., M.AF-M., J.A. B.V., Investigation, N.P. V.F, P.V., E.M.E., C.E-G., M.P.V., L.R., C.G-P., I.G.M., A.A.U. Supervision, A.M.V., N.P.; Writing Original Draft, Review, and Editing, A.M.V., N.P., J.A. All Authors revised and approved the final version of the manuscript.

Declaration of competing interest

The authors declare no competing interests.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Multimedia component 1

Body weight and glucose levels in obese mice treated with metformin for 6 weeks. Body weight change (left panel) and blood glucose levels (right panel). ⁺p<0.05, ⁺⁺p<0.01, versus initial body weight. ***p < 0.001 versus HFD. Data are shown as mean ± SEM. Statistical analysis was performed using two-tailed unpaired t-test.

Multimedia component 2

Comparable HIF1α levels in brown adipocytes exposed to CM-Lps or expressing HIF1α(P2)*. Representative Western blots of HIF1α and Vinculin levels and quantification in differentiated brown adipocytes treated with CM-Lps for 18 h or expressing HIF1α(P2)* (n = 3 independent experiments). Data are shown as mean ± SEM. Statistical analysis was performed using two-tailed unpaired t-test.

Multimedia component 3

Deletion of HIF1α in brown adipocyte does not protect against proinflammatory signaling. Representative Western blots of phospho-IRS1 (Serine307), IRS1, phospho-STAT3, phospho-JNK, JNK and Vinculin levels and quantification in BA^Hif1αfl/fl and BA^Hif1aΔ/Δ treated with CM-Ctl and CM-Lps for several time periods (n = 3 independent experiments). **p < 0.01, ***p < 0.001 versus CM-Ctl. Data are shown as mean ± SEM. Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test.

Multimedia component 4

Direct addition of proinflammatory cytokines to brown adipocytes suppressed insulin signaling. (Left panel) IRS1 and Tubulin levels in brown adipocytes treated with a cocktail of TNFα, IL6 and IL1β for 24 h and quantification (n = 3 independent experiments). ***p < 0.001 versus Ctl. (Right panel) Brown adipocytes were treated with a cocktail of TNFα, IL6 and IL1β for 24 h followed by stimulation with 10 nM insulin for 10 min. Representative Western blots using the indicated antibodies and respective quantification (n = 4 independent experiments). ***p < 0.001 versus without insulin; ^###p < 0.001 versus Ctl + Insulin. Data are shown as mean ± SEM. Statistical analysis was performed using two-tailed unpaired t-test when comparing two groups and one-way ANOVA followed by Tukey's post hoc test for multiple comparisons..

Multimedia component 5

The protective effect of metformin in the crosstalk between macrophages and brown adipocytes does not require AMPKα1.A) Representative Western blots of phospho-AMPK, AMPK and Vinculin levels and quantification in brown adipocytes treated with CM-Ctl, CM-Lps or CM-LpsMet for several time-periods (n = 3 independent experiments). **p < 0.01, ***p < 0.001 versus CM-Ctl; ^#p < 0.05, ^##p < 0.01versus CM-Lps. B) Representative Western blots of phospho-AMPK, AMPK and Vinculin levels and quantification in brown adipocytes treated with CM-LpsMet or CM-Ctl plus metformin (10 mM) (CM-Ctl + Met) for several time-periods (n = 4 independent experiments). **p < 0.01, ***p < 0.001 versus CM-Ctl. C) Representative Western blots of AMPKα1, AMPK and Vinculin levels in wild-type (BA-WT) and AMPKα1-deficient (BA^Ampkα1−/-) brown adipocytes. D) Representative phase contrast photomicrographs of differentiated BA-WT and AMPKα1-deficient (BA^Ampkα1−/-) brown adipocytes. E) Representative Western blots of phospho-Akt (Ser473), Akt and Vinculin levels and quantification in AMPKα1-deficient brown adipocytes treated with CM-Ctl, CM-Lps or CM-LpsMet for 18 h followed by stimulation with 10 nM insulin for 10 min (n = 4 independent experiments). ***p < 0.001 versus without insulin; ⁺⁺p<0.01 versus CM-Ctl + Insulin; ^###p < 0.001 versus CM-Lps + Insulin. F) Representative Western blots of UCP1 and Vinculin levels and quantification in AMPKα1-deficient brown adipocytes stimulated with 2 μM CL316243 in the presence of CM-Ctl, CM-Lps or CM-LpsMet for 18 h (n = 3 independent experiments). **p < 0.01 versus without CL316243; ⁺⁺⁺p<0.001 versus CM-Ctl + CL316243; ^##p < 0.01 versus CM-Lps + CL316243. G) Representative Western blots of HIF1α, iNOS and Vinculin levels in BMDM lacking AMPKα1 treated with LPS (150 ng/ml, 4 h) (Lps) or LPS (150 ng/ml, 4 h) followed by metformin (10 mM, 16 h) (LpsMet) (n = 3 independent experiments). **p < 0.01, ***p < 0.001 versus Ctl; ^#p < 0.05, ^###p < 0.001 versus Lps. Data are shown as mean ± SEM. Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test..

Multimedia component 6

Co-location of HIF1α and F4/80 in areas of macrophages infiltration in BAT from obese mice. Immunofluorescence images of HIF1α and F4/80 in BAT from mice fed a HFD without (HFD) or with metformin treatment (HFD-Met). DAPI (blue), F4/80 (green), and HIF1α (red) (n = 4 mice/group). Scale bar 10 μm..

Multimedia component 7

Complex I inhibitor rotenone leads to similar effect of metformin in the crosstalk between macrophages and brown adipocytes. A) Changes in oxygen consumption in Raw 264.7 macrophages in response to Lps or Lps and Rotenone (LpsRtn) treatment (n = 3 independent experiments). **p < 0.01 versus Ctl, ^#p < 0.05 versus Lps. B) Representative Western blots of HIF1α, iNOS and Vinculin levels in Raw 264.7 macrophages treated with LPS (Lps) or LPS and Rotenone (LpsRtn) as indicated in material and methods (n = 3 independent experiments). ***p < 0.001 versus Ctl; ^###p < 0.001 versus Lps. C)(Left panel) Representative Western blots of IRS1, HIF1α and Tubulin levels and quantification in differentiated brown adipocytes treated with CM-Ctl, CM-Lps and CM-LpsRtn for 24 h (n = 3 independent experiments). ***p < 0.001, **p < 0.01 versus CM-Ctl; ^###p < 0.001, ^#p < 0.05 versus CM-Lps. (Right panel) Representative Western blot of phosho-Akt (Ser473), Akt and Vinculin levels and quantification in brown adipocytes treated with CM-Ctl, CM-Lps or CM-LpsRtn for 18 h followed by stimulation with 10 nM insulin for 10 min (n = 4 independent experiments). ***p < 0.001 versus without insulin; ^###p < 0.001 versus CM-Lps + Insulin. D) Representative Western blots of UCP1 and Vinculin levels and quantification in brown adipocytes stimulated with 2 μM CL316243 in the presence of CM-Ctl, CM-Lps or CM-LpsRtn for 18 h (n = 5 independent experiments). **p < 0.01, *p < 0.05 versus without CL316243. Data are shown as mean ± SEM. Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test..

Multimedia component 8

Levels of proinflammatory mediators in the different CM (CM-Ctl, CM-Lps, CM-LpsMet, CM-InhLps) released by macrophages.

FigS1.

CM-Lps generated by BMDM, but not CM-LpsMet, increased HIF1α, decreased IRS1 levels and impaired insulin and CL316243-mediated responses in brown adipocytes. A) (Upper panel) IRS1, HIF1α and Tubulin levels in brown adipocytes treated with CM-Ctl, CM-Lps or CM-LpsMet (generated by BMDM) for 24 h and quantification (n = 4 independent experiments). ***p < 0.001, **p < 0.01 versus CM-Ctl; ^###p < 0.001, ^##p < 0.01 versus CM-Lps. (lower panel) Representative Western blot of phosho-Akt (Ser473), Akt and Vinculin levels and quantification in brown adipocytes treated with CM-Ctl, CM-Lps or CM-LpsMet (generated by BMDM) for 18 h followed by stimulation with 10 nM insulin for 10 min (n = 3 independent experiments). ***p < 0.001, *p < 0.05 versus without insulin; ^#p < 0.05 versus CM-Ctl + Insulin. B) Representative Western blot images of UCP1 and Vinculin levels and quantification of CL316243-induced UCP1 in brown adipocytes stimulated with 2 μM CL316243 in the presence of CM-Ctl, CM-Lps or CM-LpsMet (generated by BMDM) for 18 h (n = 4 independent experiments). ***p < 0.001 versus without CL316243; ^###p < 0.001 versus CM-Ctl + CL316243; ⁺⁺⁺p<0.001 versus CM-Lps + CL316243. Data are shown as mean ± SEM. Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test.