Unique metabolic activation of adipose tissue macrophages in obesity promotes inflammatory responses
Recent studies have identified intracellular metabolism as a fundamental determinant of macrophage function. In obesity, proinflammatory macrophages accumulate in adipose tissue and trigger chronic low-grade inflammation, that promotes the development of systemic insulin resistance, yet changes in their intracellular energy metabolism are currently unknown. We therefore set out to study metabolic signatures of adipose tissue macrophages (ATMs) in lean and obese conditions.
F4/80-positive ATMs were isolated from obese vs lean mice. High-fat feeding of wild-type mice and myeloid-specific Hif1α−/− mice was used to examine the role of hypoxia-inducible factor-1α (HIF-1α) in ATMs part of obese adipose tissue. In vitro, bone marrow-derived macrophages were co-cultured with adipose tissue explants to examine adipose tissue-induced changes in macrophage phenotypes. Transcriptome analysis, real-time flux measurements, ELISA and several other approaches were used to determine the metabolic signatures and inflammatory status of macrophages. In addition, various metabolic routes were inhibited to determine their relevance for cytokine production.
Transcriptome analysis and extracellular flux measurements of mouse ATMs revealed unique metabolic rewiring in obesity characterised by both increased glycolysis and oxidative phosphorylation. Similar metabolic activation of CD14+ cells in obese individuals was associated with diabetes outcome. These changes were not observed in peritoneal macrophages from obese vs lean mice and did not resemble metabolic rewiring in M1-primed macrophages. Instead, metabolic activation of macrophages was dose-dependently induced by a set of adipose tissue-derived factors that could not be reduced to leptin or lactate. Using metabolic inhibitors, we identified various metabolic routes, including fatty acid oxidation, glycolysis and glutaminolysis, that contributed to cytokine release by ATMs in lean adipose tissue. Glycolysis appeared to be the main contributor to the proinflammatory trait of macrophages in obese adipose tissue. HIF-1α, a key regulator of glycolysis, nonetheless appeared to play no critical role in proinflammatory activation of ATMs during early stages of obesity.
Our results reveal unique metabolic activation of ATMs in obesity that promotes inflammatory cytokine release. Further understanding of metabolic programming in ATMs will most likely lead to novel therapeutic targets to curtail inflammatory responses in obesity.
Microarray data of ATMs isolated from obese or lean mice have been submitted to the Gene Expression Omnibus (accession no. GSE84000).
KeywordsAdipose tissue Glycolysis obesity Immunometabolism Inflammation Macrophages Oxidative phosphorylation
AMP-activated protein kinase
Adipose tissue macrophage
Bone marrow-derived macrophage
Extracellular acidification rate
Gene set enrichment analysis
Chemokine (C-X-C motif) ligand-1
Oxygen consumption rate
Principal component analysis
Quantitative reverse-transcription PCR
Spare respiratory capacity
In obesity, macrophages accumulate in adipose tissue and trigger chronic low-grade inflammation which promotes the development of systemic insulin resistance [1, 2, 3]. Based on transcriptional profiles and expression markers derived from in vitro experiments, macrophages are generally classified as classically/inflammatory (M1) or alternatively/anti-inflammatory (M2) activated [4, 5]. Applying this phenotypical classification to adipose tissue macrophages (ATMs) has led to the identification of M2 macrophages in lean adipose tissue vs M1 macrophages in obese adipose tissue . In recent years, however, the two-dimensional M1/M2 spectrum has been challenged and macrophages in different tissue environments have been shown to adopt a variety of inflammatory phenotypes that fall outside this classification [7, 8]. Indeed, macrophages in obese adipose tissue display surface-proteins that resemble neither classical nor alternative activation, but rather represent a state of metabolic activation . Intracellularly, ATMs in obese adipose tissue are characterised by lysosomal activity , suggestive of robust changes in intracellular energy metabolism of ATM in obesity.
Recent developments in the field of immunology have identified macrophage intracellular energy metabolism as a fundamental determinant of its functional response. M1 macrophages are characterised by a high glycolytic rate whereas M2 macrophages rely mainly on oxidative phosphorylation (OXPHOS) [11, 12]. A central role in driving macrophage polarisation has been appointed to hypoxia-inducible factor-1α (HIF-1α), a master regulator of glycolysis that is critically involved in the development of the M1 phenotype [13, 14]. Profiling of intracellular metabolism in ATMs, as well as identifying key regulators involved, is expected to further the understanding of their metabolic functions and may ultimately bring forward targets for modulating their inflammatory traits. Using various approaches, we identified unique metabolic activation of ATMs in obesity that does not resemble M1 or M2 macrophages. Metabolic activation of macrophages, characterised by increased OXPHOS and glycolysis, was dose-dependently induced during a co-culture with adipose tissue and translated into increased cytokine secretion. Although various metabolic pathways contributed to cytokine release by ATMs, glycolysis accounted mostly for the higher cytokine production by ATMs from obese mice. Inflammatory activation of ATMs during early stages of obesity, however, appeared to be independent of HIF-1α. Further understanding of the functional consequences of metabolic programming in macrophages in lean adipose tissue and metabolic activation in ATMs residing in obese adipose tissue is expected to lead to novel therapeutic targets to curtail inflammatory responses that will ultimately reduce obesity-induced metabolic complications.
Male C57Bl/6 mice (Harlan, Horst, Germany) were on a high-fat diet (HFD) containing either 45% (D12451) or 60% energy derived from fat (D12492), or on a low-fat diet (LFD) containing 10% energy derived from fat and matching most other components present in either the 45% HFD (D12450B) or 60% HFD (D12450J), for 16 weeks. Mice were stratified based upon body weight at the start of the LFD or HFD intervention. All procedures were approved by the ethics committee for animal experiments at Wageningen University.
For studying the role of HIF-1α in ATMs during the development of obesity, 9- to 12-week-old male C57/Bl6 mice with floxed Hif-1α (also known as Hif1a) (exon 2) crossed into a background of lysozyme M-driven cre recombinase (LysM Hif-1α−/−) or C57/Bl6 controls not carrying lysozyme M-driven cre recombinase (LysM Hif-1α+/+)  were exposed to an HFD (D12492) for 8 weeks. After 7 weeks, an insulin tolerance test was performed in mice fasted for 5 h, by injecting insulin (1 U/kg body weight) intraperitoneally. Blood was taken from the tail at specific time points and glucose was measured using Accu-check glucose meters (Roche Diagnostics, Almere, the Netherlands). The study was carried out in accordance with recommendations in the Guide for the Care and Use of Laboratory Animals of the National Research Council. The protocol was approved by the Dartmouth IACUC.
All mice were individually housed and had ad libitum access to food and water. All diets were obtained from Research Diets (New Brunswick, NJ, USA). Experimenters were not blinded to group assignment and outcome assessment.
ATMs and peritoneal macrophages were isolated from male, wildtype C57Bl/6 mice (Harlan). For details of tissue and cell collection, see ESM Methods.
Freshly isolated ATMs were cultured in RPMI 1640 (Lonza, Verviers, Belgium) supplemented with 10% (vol./vol.) FCS and 1% (vol./vol.) penicillin/streptomycin (PS) (RPMI/FCS/PS) for 24 h (200,000 cells/well). The contribution of various metabolic pathways to cytokine release was examined by providing 5.5 mmol/l 2-deoxy glucose (2-DG) (Merck, Darmstadt, Germany), 50 μmol/l Etomoxir (Merck), 10 μmol/l UK5099 (Merck) or 10 μmol/l BPTES (Merck) 2 h after plating until the end of the 24 h culture period.
Bone marrow-derived macrophages
Bone marrow cells were cultured in DMEM (Lonza) supplemented with 10% (vol./vol.) FCS and 1% (vol./vol.) PS (DMEM/FCS/PS) and 5% (vol./vol.) L929-conditioned medium (L929). After 3–4 days, adherent bone marrow-derived macrophages (BMDMs) were re-plated and exposed for 3 days to an insert containing 25 mg or 100 mg of minced epididymal adipose tissue collected from LFD-fed or HFD-fed mice. Control BMDMs were held in DMEM/FCS/PS containing 5% (vol./vol.) L929 with an empty insert for the same length of time. M1 and M2 macrophages were generated by 24 h incubation with 10 ng/ml lipopolysaccharide (LPS) (M1) or 25 ng/ml IL-4 (M2). For measuring cytokine and lactate production of BMDMs after adipose tissue priming, inserts were removed and BMDMs were held in fresh DMEM/FCS/PS for an additional 24 h. For extracellular flux analysis, BMDMs were scraped, counted and re-plated in a 96-well Seahorse microplate (Seahorse Bioscience, Santa Clara, CA, USA).
Epididymal adipose tissue was brought into culture and exposed to 17.5 nmol/ml insulin in DMEM/PS for 20 min to measure insulin sensitivity. The tissue was kept in DMEM/FCS/PS with or without LPS (10 ng/ml) for 24 h to measure IL-6 release or was cultured in DMEM/FCS/PS for 3 days for leptin and lactate measurements.
Extracellular flux analysis
The real-time oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) of ATMs and BMDMs were analysed using an XF-96 Extracellular Flux Analyzer (Seahorse). See ESM Methods for further details.
Cytokine and lactate measurements
Levels of IL-6, chemokine (C-X-C motif) ligand-1 (KC), TNF-α, IL-1β, IL-10 and leptin in cell culture supernatant fractions were measured with murine DuoSet ELISA Development kits (R&D Systems, Abingdon, UK). An enzymatic assay adapted from the Lactate Assay kit (Merck) was used to determine lactate levels.
Paraffin-embedded sections of epididymal adipose tissue were stained with an F4/80 antibody (Bio-Rad, Veenendaal, the Netherlands) and counterstained with haematoxylin. Macrophages were visualised with 3,3-diaminobenzidene (Merck).
Primary antibodies for actin (Merck), AMP-activated protein kinase (AMPK) (no. 2532L), phospho-AMPK Thr172 (no. 2531) and p-Akt Ser473 (no. 4060) (Cell Signalling, Leiden, the Netherlands) were all used at a ratio of 1:1000 according to manufacturers’ instructions, and incubated overnight at 4°C. See ESM Methods for further details.
RNA isolation and qRT-PCR
RNA from ATMs and peritoneal macrophages isolated from HFD-fed or LFD-fed mice, or from BMDMs, was used for quantitative reverse-transcription PCR analysis (qRT-PCR). The following genes were measured: Cd11c (also known as Itgax), Cd206 (also known as Mrc1), Cd36, Cd68, Glut1, Hk2, Hif-1α, Ldhα (also known as Ldha), Lipa, Pdk4, Plin2 and Vegfα (also known as Vegfa), normalised against 36b4 (also known as Rplp0) (ESM Table 1). For further details, see ESM Methods.
Microarray analysis and interpretation
Four pools of ATMs isolated from epididymal adipose tissue of male C57Bl/6 mice fed an LFD or HFD in four separate experiments were subjected to expression profiling by microarray. In addition, raw transcriptome data from various tissue macrophages, including ATMs (GEO accession no. GSE56682), from LPS-stimulated BMDMs (GSE53986) and from obese diabetic and obese non-diabetic humans (GSE54350) were obtained from the Gene Expression Omnibus. Details of the microarray analysis and interpretation are in the ESM Methods. Microarray data have been submitted to the Gene Expression Omnibus (accession number GSE84000).
Results are shown as mean ± SEM. Statistically significant differences between two groups were calculated using Student’s t test. For comparisons between more than two groups, a one-way ANOVA and post hoc Bonferroni’s multiple comparison test was done. When comparing diet and treatment effects within one experiment, data were analysed with a two-way ANOVA with post hoc Bonferroni test (treatment vs control). A p value ≤0.05 was considered significant.
Unique metabolic and inflammatory activation of ATMs in obesity
Metabolic and inflammatory activation of macrophages in obese adipose tissue is distinct from classical activation by LPS and associates with the presence of type 2 diabetes in obese humans
Functional consequences of metabolic activation of ATMs in obesity
Exposing macrophages to adipose tissue explants ex vivo dose-dependently induces metabolic activation and cytokine release
Glycolysis largely controls cytokine release by ATMs in obesity
Myeloid-specific Hif-1α does not affect adipose tissue inflammation in HFD-fed mice
We observed increased spare respiratory capacity (SRC) of macrophages in an adipose tissue environment compared with control macrophages (Fig. 6f), indicative of enhanced cellular flexibility and capacity to manage stressful situations [28, 29]. Interestingly, macrophages lacking Hif-1α had lower SRC than wild-type BMDMs (Fig. 6g) and appeared to be less glycolytic (Fig. 6h, i). The absence of compensation for lower glycolytic rates by Hif-1α−/− BMDMs (i.e. by increasing oxygen consumption) (Fig. 6j), is suggestive of reduced metabolic capacity and flexibility in the absence of Hif-1α and this may outweigh the potential anti-inflammatory effects of lower glycolysis in an adipose tissue environment.
In obesity, macrophages fuel adipose tissue inflammation, promoting the development of insulin resistance and type 2 diabetes . The inflammatory state of ATMs has been studied extensively. To our knowledge, however, we are the first to measure real-time metabolic fluxes in freshly isolated ATMs. Interestingly, our data strongly point to unique metabolic activation that drives cytokine release by ATMs in obesity, resembling neither the metabolic nor inflammatory signatures seen in M1- or M2-primed macrophages or peritoneal macrophages.
Cumulative evidence from the field of immunology shows that robust metabolic rewiring fuels differential inflammatory activation of macrophages. On the one hand, M2 macrophages require OXPHOS for responses, whereas M1 macrophages rely on aerobic glycolysis [11, 12]. In contrast to these two extremes, we found that macrophages in an adipose tissue environment adopt a unique metabolic profile in obesity, characterised by activation of various metabolic routes including both OXPHOS and glycolysis. Metabolic and inflammatory adaptations in obesity were specific for ATMs, as no metabolic rewiring was found in peritoneal macrophages. In line with our finding of unique metabolic rewiring in ATMs and supportive of various studies reporting diverse inflammatory activation of ATMs in obese adipose tissue [9, 16, 31, 32], we found inflammatory activation of ATMs to be different from that of classically activated macrophages.
Interestingly, macrophages co-cultured with obese adipose tissue developed similar phenotypical adaptations in a dose-dependent manner, suggestive of obesity-induced changes in the adipose tissue microenvironment shaping the ATM phenotype. Indeed, the composition of adipose tissue is importantly affected in obesity, with resultant adipocyte hypertrophy and both accumulation and phenotypical changes of immune cells including macrophages. In our co-culture system we have used lean and obese adipose tissue explants of equal weight. This may not have accounted for all the shifts in relative cell numbers occurring in obese adipose tissue, yet strongly points toward the existence of divergent factors secreted by obese vs lean adipose tissue that may critically influence the macrophage phenotype in a dose-dependent manner. Potential factors may include adipokines, cytokines, fatty acids or other metabolites . Interestingly, leptin [24, 25] and lactate  have been shown capable of remodelling intracellular metabolism and changing the inflammatory state of macrophages. We found both to be secreted more by obese adipose tissue than by lean adipose tissue, yet neither leptin nor lactate induced metabolic rewiring similar to that seen in macrophages in an adipose tissue environment. Additionally, differences in cell death between lean and obese adipose tissue in vivo may have an effect on metabolic rewiring in macrophages. Most likely, a mixture of signals is responsible for shaping ATM metabolic phenotypes in the lean and obese state, although this needs further investigation.
Importantly, metabolic activation of ATMs contributes to their inflammatory cytokine release. First, metabolically active ATMs from obese mice secreted far more IL-6 and KC than the less metabolically active ATMs isolated from lean mice. Second, we found that interference with metabolic routes directly affected cytokine release by ATMs. Especially in ATMs from lean mice, several metabolic routes including fatty acid oxidation, glycolysis and glutaminolysis contribute to cytokine release. Glycolysis appears to play a dominant role in fuelling the inflammatory trait of ATMs from obese adipose tissue, since inhibiting glycolysis with 2-DG almost completely abolished the greater basal cytokine secretion by ATMs from obese vs lean mice.
Our finding of lower TNF-α secretion by ATMs from obese vs lean mice was unexpected, as was the lower level of TNF-α in supernatant fractions of macrophages exposed to obese vs lean adipose tissue. Despite lower cytokine levels, we found Tnf upregulated at the mRNA level. A similar discrepancy between mRNA and protein level has been reported in ATMs before . In obese adipose tissue, however, macrophage influx and proliferation as well as an increase in other immune cell populations likely overrules lower TNF-α secretion per macrophage and might be responsible for higher TNF-α levels found in the adipose tissue and circulation of obese individuals [6, 33, 34, 35]. Alternatively, enhanced autocrine TNF signalling in ATMs may explain the lower levels of TNF-α measured in ATM supernatant fractions.
Next to cytokine release, intracellular metabolism most likely controls several other macrophage functions. For example, OXPHOS has been found to contribute to phagocytosis by human monocytes , and lysosomal biogenesis and function in T cells . Both phagocytic and lysosomal genes were found to be strongly upregulated in ATMs of obese mice and in obese individuals with type 2 diabetes. Phagocytosis of dead adipocytes by macrophages [38, 39, 40] and lysosomal function of ATMs [10, 39] are considered to be important for maintaining adipose tissue homeostasis. Hence, the obesity-induced increase in OXPHOS might fuel ATM functions in expanding adipose tissue, not directly related to inflammatory cytokine release yet may greatly affect adipose tissue function.
Our data show that interfering with metabolic routes alters the inflammatory phenotype of ATMs and that glycolysis importantly contributes to inflammatory cytokine release by ATMs. Unexpectedly, however, myeloid-specific deletion of a previously identified key regulator of glycolysis, Hif-1α, did not alleviate inflammatory activation of ATMs during the early stages of obesity. Because mice lacking Hif-1α in myeloid cells were significantly heavier yet did not display increased adipose tissue inflammation or insulin resistance, one could speculate that HIF-1α may be partly protective for the development of obesity-induced adipose tissue inflammation, as has been reported before in mice fed an HFD for 18 weeks . One might also hypothesise that during the earlier stages of HFD-induced obesity HIF-1α is important for controlling other metabolic properties of macrophages not related to cytokine production. For example, our data revealed a decreased SRC in BMDMs from Hif-1α−/− mice, suggestive of a role for HIF-1α in maintaining metabolic flexibility of macrophages. Indeed, despite decreased capability of using glycolysis as an energy source, Hif-1α−/− macrophages lack the flexibility to increase their oxidative capacity . Metabolic flexibility is probably needed for ATMs, as we observed enhanced SRC in macrophages in an adipose tissue environment which was even further increased upon exposure to obese adipose tissue. Reduced metabolic flexibility in macrophages lacking Hif-1α in expanding adipose tissue might have overruled effects on cytokine release during the development of obesity.
In conclusion, we identified unique metabolic activation of ATMs in obesity, characterised by increased OXPHOS and glycolysis. Blocking metabolic routes in isolated ATMs led to the identification of glycolysis as a main contributor to their proinflammatory trait, especially in obesity. Interestingly, metabolic signatures, similar to those found in ATMs upon obesity, including the induction of OXPHOS and lysosomal genes, were observed in human macrophages isolated from adipose tissue of obese individuals with type 2 diabetes. Further understanding of metabolic programming in ATMs will most likely lead to novel therapeutic targets to modulate macrophage metabolism and curtail inflammatory responses that drive insulin resistance and type 2 diabetes in obese individuals.
We would like to thank R. Leen (Laboratory Genetic Metabolic Diseases, Academic Medical Center, University of Amsterdam, the Netherlands) for his excellent support during Seahorse measurements.
Microarray data of F4/80+ cells isolated from obese or lean adipose tissue have been submitted to the Gene Expression Omnibus (accession number GSE84000). All other data generated and/or analysed during the current study are available from the corresponding author on reasonable request.
LB, RAC, MGN and RS designed the experiments. LB, RS and SD conducted the experiments and data acquisition. LB, RS, GJH and MGN analysed or interpreted the data. LB and RS wrote the paper. GJH, SD, RAC and MGN critically revised the paper for important intellectual content. All authors approved the final manuscript to be published. RS is the guarantor of this work.
Duality of interest
The authors declare that there is no duality of interest associated with this manuscript.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.