HIF1α activation in dendritic cells under sterile conditions promotes an anti-inflammatory phenotype through accumulation of intracellular lipids

Obesity is among the leading causes of elevated cardiovascular disease mortality and morbidity. Adipose tissue dysfunction, insulin resistance and inflammation are recognized as important risk factors for the development of cardiovascular disorders in obesity. Hypoxia appears to be a key factor in adipose tissue dysfunction affecting not only adipocytes but also immune cell function. Here we examined the effect of hypoxia-induced transcription factor HIF1α activation on classical dendritic cell (cDCs) function during obesity. We found that deletion of Hif1α on cDCs results in enhanced adipose-tissue inflammation and atherosclerotic plaque formation in a mouse model of obesity. This effect is mediated by HIF1α-mediated increased lipid synthesis, accumulation of lipid droplets and alter synthesis of lipid mediators. Our findings demonstrate that HIF1α activation in cDCs is necessary to control vessel wall inflammation.

Obesity is among the leading causes of elevated cardiovascular disease mortality and morbidity. Adipose tissue dysfunction, insulin resistance and inflammation are recognized as important risk factors for the development of cardiovascular disorders in obesity. Hypoxia appears to be a key factor in adipose tissue dysfunction affecting not only adipocytes but also immune cell function. Here we examined the effect of hypoxia-induced transcription factor HIF1α activation on classical dendritic cell (cDCs) function during obesity. We found that deletion of Hif1α on cDCs results in enhanced adiposetissue inflammation and atherosclerotic plaque formation in a mouse model of obesity. This effect is mediated by HIF1α-mediated increased lipid synthesis, accumulation of lipid droplets and alter synthesis of lipid mediators. Our findings demonstrate that HIF1α activation in cDCs is necessary to control vessel wall inflammation.
Classical dendritic cells (cDCs) are the main antigen presenting cells and key regulators of immune homeostasis and immunity 1 . Constantly probing their surroundings, cDCs are poised to detect invading pathogens, initiate and direct adaptive immune responses. cDCs are found in peripheral tissue in an immature state, where they sense and sample the environment for self-and non-self-antigens. Under pathogen driven danger signals, e.g. toll-like receptor (TLR), cDCs undergo an intricate maturation process characterized by increased expression of co-stimulatory molecules, cytokine/chemokine production and enhanced migration to draining lymph nodes (dLN) to present antigen to naïve T cells 1 . In tissue, cDCs must adapt to specific metabolic cues such as nutrient restriction or low O 2 tensions, which may modulate their biological functions. Indeed, in recent years it became apparent that disruption of key metabolic pathways can alter immune cell activation and differentiation.
Reduced oxygen or hypoxia, is a hallmark of multiple acute and chronic diseases and arises especially in inflamed tissues. Cellular adaptation to hypoxia rely on stabilization of the hypoxia-inducible factor-1 (HIF1) transcription factor, which drives the metabolic changes required for adaptation to hostile microenvironments. HIF1 is a heterodimer transcription factor consisting of a constitutively expressed β-subunit (HIF1β) and an oxygen-regulated α subunit (HIF1α) 2 . Under normoxic conditions, HIF1α is regulated through hydroxylation of proline residues, ubiquitination and rapid degradation by the proteasome. Under low oxygen concentrations, however, prolyl hydroxylase is inactivated and HIF1α can translocate into the nucleus for transcriptional regulation. HIF1α accumulation does not always occur under low O 2 tension but can be triggered by oxygenindependent signalling events or cellular stress, such as toll-like receptor signalling and reactive oxidase species (ROS)-a phenomenon known as 'pseudohypoxia' 3 .
Activation of the hypoxia pathway is believed to fuel inflammation, largely due to HIF1α activation in macrophages. Lipopolysaccharide (LPS) activation induces HIF1α protein accumulation in macrophages under normoxic conditions 4 leading to metabolic reprogramming, succinate release, and IL-1β production 5 . HIF1αdeficient macrophages exhibit impaired inflammatory responses, with reduced glycolytic rate, motility and migration 6 . Similarly, toll-like receptor (TLR) activation in dendritic cells, leads to HIF1α stabilization, glucose Scientific Reports | (2020) 10:20825 | https://doi.org/10.1038/s41598-020-77793-6 www.nature.com/scientificreports/ uptake and glycolytic metabolic shift 7,8 . In this context, deletion of HIF1α results in decreased ATP production, cell survival and T cell priming [8][9][10] . However, the effect of HIF1α on dendritic cells function remains controversial. In other studies, particularly under hypoxic conditions, HIF1α activation has been shown to promote anti-inflammatory response 11,12 . HIF1α has pleiotropic effects regulating hundreds of genes encoding glycolytic enzymes, glucose transporters, matrix metalloproteinases, angiogenic and survival factors, that may depend on time and context. In addition, these divergent reports may in fact be explained by differences in experimental conditions or cell type targeted. The present study was designed to investigate the role of HIF1α activation in cDCs in an experimental model of obesity by using mice with HIF1α deletion specifically in cDCs. Contrary to what was expected, deletion of HIF1α in cDCs promoted adipose tissue inflammation and atherosclerotic plaque formation. This effect was mediated by intracellular lipid accumulation and altered production of lipid mediators, which translated into reduced T cell stimulation.

Results
Deletion of Hif1α in cDCs leads to increased adipose tissue inflammation under high lipid and cholesterol diet. Adipose tissue hypoxia has been causally implicated in obesity-induced inflammation and insulin resistance. Under lipid rich diet, adipocyte expansion decreases oxygen tension and ROS production, which leads to rapid activation of HIF1α 13,14 . cDCs are present in visceral adipose tissue (VAT) where they control tissue homeostasis 15 . To investigate the effect of HIF1α in cDCs under chronic, sterile inflammation, we generated cDC-specific Hif1α knockdown by crossing Zbtb46-CRE with Hif1α fl/fl mice. As Zbtb46 is expressed in some non-haematopoietic cells, we performed mice chimeras where irradiated C57BL/6 mice were reconstituted with bone marrow cells from Hif1α fl/fl × Zbtb46-CRE (referred to as Hif1α −/− ) or HIF1α fl/+ × Zbtb46-CRE (referred to as WT) mice. HIF1α deletion is shown in Suppl Fig. S1A. Hif1α −/− or WT mice were fed a lipidrich diet (western diet) for 12 weeks, which resulted in HIF1α increase in cDCs (Fig. 1a). Surprisingly, deletion of HIF1α in cDCs results in increased immune cell infiltration into visceral adipose tissue (VAT) from Hif1α −/− mice, with increased numbers of cDCs, T cells and decreased regulatory T cells (Tregs) (Fig. 1b, S1c). Production of IL-17 in VAT-infiltrated T cells was increased in Hif1α −/− compared to WT mice together with increased IFNγ production in lymph nodes (LN) ( Fig. 1d; Suppl Fig. S1d). No differences were observed in splenic T cells (Fig. 1d). This contributes to the enhanced VAT inflammation in Hif1α −/− mice as detected by RT-PCR (Fig. 1e).
Chronic inflammation in VAT plays a central role in the development of obesity-related insulin resistance and systemic metabolic abnormalities 16 . Thus, we evaluated if the increased inflammation in Hif1α −/− mice could translate into systemic metabolic abnormalities. Body and VAT weight were similar between mice (Suppl Fig. S1e,f). Hif1α −/− mice displayed a small increase in fasting glucose levels and serum insulin levels compared to WT mice, albeit this was not significant (Fig. 1g,h). Circulating concentrations of adiponectin were reduced in Hif1α −/− mice indicative of adipocyte dysfunction ( Fig. 1f-h). Despite signs of metabolic misbalance, Hif1α −/− mice showed no sign of increased glucose and insulin tolerance during a glucose tolerance (GTT) and insulin tolerance (ITT) test (Suppl Fig. S1g,h). Thus, the results presented here indicate that inhibition of the Hif1 pathways under chronic, low oxygen conditions, promotes cDC activation.
Deletion of Hif1α in cDCs accelerates the development of atherosclerosis. Obesity-induced chronic inflammation is tightly linked to the development of atherosclerotic plaque 17 . The atherosclerotic plaque is characterized by arterial wall thickening, ROS production and reduced oxygen supply in certain areas of the vascular intima creating a hypoxic milieu 18 . HIF1α expression was observed in macrophages and smooth muscle cells bordering the necrotic core, and was associated with inflammation and angiogenesis 19 . In addition, arteries are surrounded by perivascular adipose tissue which, under metabolic abnormalities, appears to promote vascular dysfunction and atherosclerosis 20 . Thus, we hypothesized that increased VAT inflammation in Hif1α −/− mice might have a negative impact on the development of atherosclerosis. To test this, lethally-irradiated low density lipoprotein receptor knockout (Ldlr −/− ) mice were reconstituted with bone marrow from Hif1α fl/fl × Zbtb46-CRE (referred as Hif1α −/− → Ldlr −/− ) or HIF1α fl/+ × Zbtb46-CRE mice (referred as WT → Ldlr −/− ). Ldlr −/− mice, under fat rich diet, are known to show elevated plasma cholesterol (25 mmol/L) and rapid atherosclerotic plaque formation 21 .
To confirm that this phenotype was a consequence of increased cDC activation, we tested cDC T cell stimulatory capacity in vivo. Enhanced proliferation of ovalbumin (OVA)-specific CD4 + T cells (OT-II) could be observed in the aorta, VAT and dLN but not spleen from Hif1α −/− → Ldlr −/− compared to control mice (Fig. 2i). www.nature.com/scientificreports/ Similarly, HIF1 −/− cDCs isolated from VAT an dLN of obese mice showed enhanced OT-II cell proliferation ex vivo compared to WT cDCs (Suppl Fig. S2g). OVA uptake was not affected by Hif1α deletion (Suppl Fig. S2h). Thus, depletion of HIF1α in cDCs under sterile conditions appears to unleash their activation.

HIF1α induces intracellular lipid accumulation.
Stabilization of HIF1α is known to promote aerobic glycolysis shifting the cells away from glucose oxidative phosphorylation. Upon TLR activation, HIF1αdependent glycolytic switch enables macrophages and cDCs to produce metabolic intermediates and to fulfil cellular energetic demands required for full activation 22 (Suppl Fig. S3a). However, HIF1α has recently been shown to regulate mitochondrial glutamine metabolism (Suppl Fig. S3a). Glutamine once converted to α-ketoglutarate (αKG), can be either oxidized to succinate or reduced to isocitrate and then citrate by a mechanism known as reversed TCA cycle. The latter is favoured in cells where HIF1α is stabilized 23,24 .
In an attempt to understand the metabolic changes induced by HIF1α in cDCs, we moved to an in vitro system where cDCs were generated from bone-marrow with Flt3 ligand (Flt3L) and Hif1α was activated with desferrioxamine (DFO) ( Fig. 3a; Suppl Fig. S3b),which facilitates HIF1 activation during seahorse assays. As expected, DFO induced a glycolytic switch characterized by reduced mitochondria O 2 consumption (OCR) and increased extracellular acidification (ECAR) detected with the XF Mito Stress and the Glycolysis stress test respectively (Fig. 3b). This metabolic shift was HIF1α dependent (Fig. 3b) and equally affecting both main cDC www.nature.com/scientificreports/ subsets, cDC1 and cDC2 (Suppl Fig. S3c,d). DFO treatment alone did not affect expression of co-stimulatory molecules (Fig. 3c). Despite a shift towards aerobic glycolysis, a functional electron transport chain and glutamine-derived carbon may still be required for cellular function. Indeed, in tumour cells, metabolic adaptation under hypoxia involves a switch from glucose to glutamine as a source of carbon for the synthesis of fatty acids 25 . We therefore tested glutamine consumption, which was indeed increased in DFO-treated compared to control cells (Fig. 3d).
During the reverse TCA cycle, glutamine-derived citrate can be transported to the cytoplasm to generate acetyl-CoA, necessary for anabolic reactions such as fatty acid synthesis. In tumour cells, HIF1α-induced uptake and reduction of glutamine induces lipid synthesis and storage in lipid droplets (LD). Interestingly, intracellular accumulation of lipids has been associated with poor cDCs activation and function in tumours 26 . Consistent with an increased glutamine uptake, DFO-treated cells exhibit increased intracellular lipid accumulation in vitro and in vivo as detected by bodipy staining (Fig. 3e,f). Similar results were observed with DMOG and hypoxia treatment (Suppl. Fig. S3e). DFO-induced intracellular lipid accumulation was HIF1α dependent (Fig. 3g). Uptake of C16-Bodipy (palmitate) was unchanged after DFO stimulation in vitro and in vivo suggesting that lipid droplets accumulation is not mediated by lipid uptake but rather lipid synthesis (Fig. 3h,i). Transcriptional levels of the hypoxia-inducible lipid droplet-associated protein (Hilpda), which is involved in intracellular neutral lipid deposition, were also elevated after DFO treatment (Fig. 3j).  www.nature.com/scientificreports/ To test if glutamine metabolism was responsible for lipid accumulation in cDCs, cells were treated with the fatty-acid synthase inhibitor TOFA and the glutamine antagonist DON. Inhibition of fatty-acid synthesis and glutamine usage, significantly reduced lipid droplets accumulation in DFO-treated cells (Fig. 3k). Collectively, these data suggest that HIF1α activation in cDCs stimulates fatty-acid synthesis and accumulation in lipid droplets.

HIF1α-mediated lipid droplets accumulation suppresses cDC activation. Accumulation of lipids
in DCs has been shown to impair their antigen-presenting function in tumours. Mice with deleted expression of HIF1α on cDCs exhibit increased aortic and VAT inflammation. Thus, we tested whether DFO treatment could alter cDCs function in vitro. Indeed, DFO treatment significantly reduces cDCs stimulatory capacity in a mixedleukocyte reaction (MLR). This effect was reverted after fatty-acid synthesis inhibition with TOFA (Fig. 4a). Similarly, pre-treating cells with DFO reduces cDC cytokine production after LPS treatment (Fig. 4b,c). Thus, similar to tumour-infiltrating DCs, lipid-laden cDC exhibit reduced immunogenicity. www.nature.com/scientificreports/

Lipid accumulation in cDCs alters the expression of lipid mediators. Neutral lipid stored in LD
are mobilized by lipases to provide metabolic energy (through fatty acid oxidation) and lipids for membrane synthesis. However, recently, it has become evident that LD function as a major intracellular pool of arachidonic acid, which is the precursor for eicosanoid biosynthesis 27 . In addition, LD can often interact with peroxisomes apparently facilitating peroxisomal fatty acid oxidation 28 . To investigate a possible consequence of lipid accumulation and the synthesis of lipid mediators, we performed a lipid mediator profiling of DFO-treated and control cells (Fig. 5a,b). In these incubations, we identified mediators from all four major fatty acid metabolomes (Table S1). Production of prostaglandins and leukotrienes remained unchanged after DFO treatment, probably due to the absence of TLR stimulation (Table S1). However, incubation of these cells with DFO significantly increased resolvin D2 (RvD2) and thromboxane B 2 (TXB 2 ) a stable metabolite of TXA 2 (Fig. 5c,d) concentrations. Accordingly, key enzymes contributing to the generation of resolvins were elevated in WT compared to HIF1 −/− VAT-cDCs from obese mice (Fig. 5e). RvD2 is a pro-resolving lipid-mediator that was recently shown to suppress T cell responses and to decrease TLR4 expression in monocytes 29 . Similarly, TXA 2 modulates DC-T cell interaction and adaptive immune responses 30 . Interestingly, treatment with RvD2 or deletion of the TXA 2 receptor, TP, significantly delays atherosclerosis progression 31 . Consistent with previous observations, treatment of cDCs-T cell cultures with RvD2 and TP agonist (I-Bop) reduces T cell proliferation in vitro (Fig. 5f). Similarly, the inhibitory effect of HIF1α activation in cDCs, was partially reverted with the TP antagonist SQ29548 (Fig. 5g). Thus, this data suggest that intracellular accumulation of lipids also impacts the production of bioactive lipid mediators by cDC.

Discussion
The transcriptional factor HIF1α plays a central role in cellular adaptation to hypoxia. Physiologic O 2 tension found in tissues varies between 2 and 9%. Hypoxic niches can be observed in a range of tissues including bone marrow, medulla of the kidney, intestinal epithelial and thymus [32][33][34] . Tissue oxygenation can be severely disrupted during pathological conditions such as obesity, solid tumours, stroke and inflammation leading to hypoxia and HIF1 pathway activation [35][36][37] . In obesity, HIF1 stabilization is propagated by mitochondrial ROS production 14 . In either of these conditions, HIF1α initiates a transcriptome response to help the cells adapt to a hostile environment. For example, HIF1α induces production of angiogenic factors to promote blood vessel formation; switches cellular metabolism to anaerobic glycolysis; reduces reactive oxygen species (ROS) and upregulates BCL-XL expression thus promoting cell survival 38,39 . The link between HIF1 and inflammation has been increasingly appreciated over recent years. Inflammatory conditions are frequently characterized by tissue hypoxia, ROS production and HIF1α stabilization due to altered metabolic supply/demand ratios 40 . HIF1α activation in adipose tissue hypoxia is a triggering factor for obesityinduced inflammation 41 . Similarly, arterial wall thickening and inflammatory deposition in the atherosclerotic plaque leads to HIF1α stabilization, which appears to promote disease progression 18 . This pro-inflammatory effect appears to be partially mediated by HIF1α activation in myeloid cells. Indeed, mice with myeloid-specific deletion of HIF1α (via lysozyme M-Cre) exhibit improved glucose metabolism, reduced adipose tissue inflammation and decreased atherosclerotic plaque formation 42,43 . With this in mind, we investigated the implications of HIF1α activation in cDCs and its physio-pathological consequences in obesity-induced chronic inflammation.
In this study, we found that inhibition of HIF1α in cDCs promotes adipose tissue inflammation and accelerates the development of atherosclerosis. Deletion of HIF1α appears to unleash cDC activation with an increased in IL12 production and induction of IL17 + T cells. Not apparent difference on lymph node migration and phagocytosis could be detected but this is not an exhaustive study and other mechanisms may be involved. These findings oppose the anti-inflammatory effect observed in myeloid cells. Of note, HIF1 transcription has also been shown to activate multiple anti-inflammatory pathways such as the expression of arginase 1 and VEGF in M2-like tumor-infiltrating macrophages and the upregulation of A2B adenosine receptor promoting Th2 and www.nature.com/scientificreports/ Treg differentiation 44 . Therefore, it appears that HIF1 regulates cell effector function in a cell-dependent manner. Contrary to pro-inflammatory macrophages, cDCs in steady-state are poised to suppress immune responses in tissue, which may account for the phenotype observed. Indeed, cDC are important for the control of adipose tissue homeostasis while Flt3L-mediated expansion of cDCs protects mice from atherosclerosis 15,45 . Along those lines, expression of HIF1α in intestinal dendritic cells, which are crucial for the maintenance of mucosal immune homeostasis, suppresses colitis-induced inflammation 12 . These results suggest that activation of HIF1α facilitates cDCs cellular adaptation and survival, allowing them to pursue their anti-inflammatory function.
A key feature of the HIF1 transcriptional response is the upregulation of genes encoding glycolytic enzymes, glucose transporters and the pyruvate dehydrogenase kinase Pdk1 thereby shunting pyruvate away from mitochondria 6 . Under these conditions, mitochondria are not supplied with glucose-derived acetyl-CoA, a central biosynthetic precursor for lipid synthesis. To compensate for this effect, recent studies in tumor cells have shown a complementary switch towards glutamine metabolism to support anabolism and growth under hypoxic conditions 46 . In cDCs, glutamine-derived lipid synthesis results in the accumulation of lipids in LDs. Interestingly, cDCs in solid tumors, which are in general hypoxic, are also characterized by LD accumulation and an anti-inflammatory phenotype 26 . Indeed, similar findings were described as a result of the ER stress response factor XBP1 activation in ovarian tumor-infiltrated cDCs 47 . XBP1 regulates HIF1α pathway activation and is essential for cell survival under hypoxic conditions 48 .
The mechanism by which lipid-laden DCs acquire a tolerogenic function is unclear. Recently it has been suggested that LD in tumor-DCs contain oxidatively truncated electrophilic lipids that interfere with cross-presentation 49 . However, the role of lipid mediators is not known. Lipid mediators are considered key signalling molecules in inflammation as they are locally produced through specific biosynthetic pathways in response to extracellular stimuli 50 . Once produced, they bind to their cognate G protein receptor to exert their function after which they are rapidly inactivated through specific processes 50 . Lipid mediators are involved in many physiological processes and their dysregulation has often been linked to various diseases including atherosclerosis 51   www.nature.com/scientificreports/ can be subcategorized into different groups, such as AA-derived eicosanoids (e.g. prostaglandins); and the newly identified anti-inflammatory and pro-resolving mediators derived from ω-3 polyunsaturated fatty acids such as resolvins 52 . They can have pro-or anti-inflammatory function: for example, pro-inflammatory lipid mediators, in general terms, include prostaglandins and leukotrienes, while pro-resolving mediators include lipoxins, resolvins, maresins and protectins 52 . The effect of several lipid mediators on DC function is well documented 34,53 , however, production of these mediators by cDC and its regulation remains uncharacterized. We here show that activation of HIF1 pathways promotes the expression of RvD2 and TXB 2 , two mediators involved in the regulation of T cell responses. The profile of lipid mediators is not only regulated at the enzymatic levels but also through substrate availability that may differ between tissues and physiological and pathological conditions 54 . Our data indicated that HIF1α activation can alter the production of lipid mediators through LD accumulation. How the lipid mediator profile is altered in tissue-cDCs in vivo will require further investigations. There is currently an increasing interest in targeting immunometabolic pathways in inflammatory diseases and cancer. However, questions arise about the specificity and off-target effect of many metabolic drugs. Although deletion of HIF1α in cDCs results in increased obesity-tissue inflammation, therapeutic treatment with the selective HIF1α inhibitor PX-478 slows down the progression of obesity-induced adipose tissue inflammation 55 . This opposite effect can be mediated through inhibition of HIF1α activation in recruited inflammatory macrophages, inhibition of HIF1-dependent VEGF production or directly on adipocytes [56][57][58][59] . Thus, the use of HIF1α inhibitors should be evaluated with caution as it may increase the risk of obesity-induced inflammatory complications such as psoriasis and rheumatoid arthritis or risk of infections.

Methods
Mice. B6.129s7-Ldlr tm1Her /J (Ldlr-KO), B6.129-Hif1a tm3Rsjo /J (HIFa loxp ) and B6.Cg-Tg(TcraTcrb)425Cbn/J (OT-II) mice were purchased from Jackson Laboratory (US); C57BL/6 were purchased from Charles River laboratories (UK); Zbtb46-Cre + mice were kindly provided by Nussenzweig (The Rockefeller University, NY). As previously described 15 , mice were housed in temperature-and humidity-controlled rooms at 22 °C and 55% humidity, with a 12 h light/12 h dark cycle. Mice were fed Chow or Test Diet AIN-76A (Test Diet IPS Ltd) and given water ad libitum; animals were rehoused in clean cages weekly. 15 g of Z-NEST (IPS Ltd) was used as nesting material to help regulate temperature and light levels. For chimeras, Ldlr-KO male mice were γ-irradiated twice with 500 rad 3 h apart. Three hours later, mice were reconstituted with marrow cells (3 × 10 6 ) that had been harvested from the femurs and tibias of Hif1α −/− and control WT littermates. Mice were maintained on acidified water during the critical 4 weeks reconstitution period. Mice fed Western diet that failed to gain more than 20% of body weight were excluded from the study. All animal experimental protocols were carried out in accordance with UK government Home Office licensing procedures and regulations, and approved by QMUL's Animal Welfare and Ethical Review Board (AWERB-7007443). www.nature.com/scientificreports/ For in vivo cDC T cell stimulatory capacity, mice were injected intravenously with ovalbumin-specific OTII cells purified from spleen and lymph nodes by CD4 + bead positive selection (MACS Miltenyi Biotec) and labelled with CFSE 3 μM (Invitrogen). The next day, mice were immunized with 200 μg of Ovalbumin (Sigma) i.p. Three days later, immune cells were isolated from spleen, lymph nodes, aorta and visceral adipose tissue and OT-II cell division was evaluated by CFSE dilution by flow cytometry. Alternatively, mice were immunized with 200 μg of Ovalbumin (Sigma) i.p. and cDCs from VAT and dLN were isolated four hours later, purified by cell sorting and cultured with CFSE-labelled OT-II cells at a ratio of 1:5 in vitro. CFSE dilution was evaluated by flow cytometry. To evaluate ovalbumin uptake, mice were injected i.p. with 10 μg of OVA-Alexa Fluor-555. Two hours later, visceral adipose tissue was harvested and digested. Ova uptake by cDCs was analysed by flow cytometry.
BM-cDCs were treated with DFO overnight. Cells were then plated at 10 6 cells/well and stimulated with 20 ng/ml LPS (sigma) in the presence of DFO. After culturing overnight, the supernatant was collected and levels of IL-10, IL-6 and IL-12p70 were determined by murine ELISA (eBioscience). Alternatively, cells were stimulated with LPS for 6 h and in combination with 10 µg/ml Brefeldin A (Sigma) for the last 5 h. Intracellular expression of IL-12 was assessed by flow cytometry as previously described. For in vivo IL-12 production, mice were immunized with 1 μg of LPS (Sigma), 2 h later, tissues were harvested, digested and incubated with 10 µg/ ml Brefeldin A (Sigma) for an additional 4 h.
Western blotting. Western blotting was performed as describe in 15 . Protein lysates were prepared from purified cDCs using RIPA buffer with phosphatase and protein inhibitors (Pierce). Proteins were separated with SDS-PAGE and transferred to Immobilon™ PVDF membrane (Millipore). Membranes were blocked for 1 h at room temperature in PBST containing 5% (w/v) milk, incubated overnight at 4 °C with anti-Hif1α antibody (Cell signalling) and subsequently with HRP-conjugated secondary antibody (Amersham Bioscience). Blotted proteins were detected using an EZ-ECL™ HRP chemiluminescence detection kit (Biological Industries) and exposed on to Hyperfilm™ photo film (Amersham).
Quantitative Real-time PCR. RT-PCR was performed as previously described 15 . Briefly, RNA from 100 mg VAT was isolated using RNeasy Lipid Tissue Mini Kits (Qaigen) following the manufacturer's instructions. Reverse transcription to cDNA was performed using High-Capacity RNA-to-cDNA Kits (Applied Biosystems) and stored at − 80 °C. Gene expression was performed using SYBR Green Supermix (Bio-Rad), according to the manufacturer's instructions, and analysed using a CFX connect light cycler (Bio-Rad). Gene-relative expression was calculated using the ΔΔCT method and normalised to a reference control (GAPDH) with control sample set as 1. Average ratio from replicates was calculated using geometric mean. The following primers were used for the analysis (Invitrogen, UK); www.nature.com/scientificreports/ opened longitudinally for en face Oil Red O staining. To prepare the stain 3 parts 5 mg/ml Oil Red O solution in isopropanol was diluted with 2 parts distilled water and allowed to solubilise for 1 h prior to 0.45 μm filtration. The aortas were washed in distilled water then transferred into freshly prepared Oil Red O solution for 15 min at room temperature (RT). They were then washed in 60% isopropanol for 5 min. This was repeated until no more red colour came out into the wash. Then the aortas were transferred into distilled water. Glycerol gelatin (Sigma) was used to mount the aortas onto slides and they were visualised using a M205 FA microscope (Leica). Images of the aortas were analysed blindly using ImageJ to quantify the percentage lesion density. The top half of the hearts were paraffin embedded, sectioned and mounted onto slides ready for Sirius Red staining by the BCI Pathology Core Service. The sectioned were visualised using a Pannoramic 250 High Throughput Scanning Microscope.
Lipid mediator profiling. Lipid profiling was performed by QMUL lipidomic facility as described in 60 .
Cell incubations were placed in 2 volumes of ice-cold methanol (Thermo Fisher Scientific) containing internal standards (d 8 -5S-hydroxy eicosatetraenoic acid, d 5 -RvD2, d 5 -LXA 4 , d 4 -PGE 2 , and d 4 -leukotriene LTB 4 (all from Cayman Chemical); 500 pg each were added to the sample. LMs were extracted and identified as described previously 60 . Briefly, the samples in methanol were incubated for 45 min at − 20 °C for protein precipitation, and centrifuged at 1900g at 4 °C for 10 min. The methanol content of the supernatant was evaporated to less than 1 ml using nitrogen gas stream, and the LMs were extracted with automated Extra-Hera system (Biotage, Uppsala, Sweden) employing solid-phase extraction. The methyl formate eluates were concentrated and injected to liquid chromatography-tandem mass spectrometry (LC-MS/MS) system (LC-20AD HPLC (Shimadzu) and SIL-20AC autoinjector (Shimadzu) paired with QTrap 6500 + (ABSciex, Framingham, MA, USA). LMs were identified and quantified using multiple reaction monitoring the parent (Q1) and daughter (Q3) ions in negative ionization mode. Identification was conducted in accordance with published criteria, matching the retention time with authentic and synthetic standards (from Cayman Chemical, prepared in house, or provided by Charles N. Serhan, Harvard Medical School, Boston, MA, USA) and identifying at least 6 diagnostic ions from the MS/ MS spectra PMID: 29363065.
Metabolic assays. Adiponectin, Insulin (Merk Millipore), and total cholesterol (Abcam) were measured in the plasma from fasted mice by ELISA. Glucose metabolism was assessed by Glucose (GTT) and Insulin (ITT) tolerance tests after 6 h fasting as previously described in 15 . For GTT, mice were administered with 1.5 mg d-glucose/g of body weight (Sigma) or 0.5 U insulin/kg of body weight (Actrapid) via intraperitoneal injection for ITT. In both cases, blood glucose measurements were taken from tail bleeds at 15, 30, 60, 90, 120, 180 min after injection using blood glucose meter and test strips (FreeStyle Optium Neo, Abbott). Body weight and food intake was recorded weekly. Weight of total visceral adipose tissue harvested was recorded for cell number calculations and to determine the percentage of body weight. To test mitochondrial respiration and glycolysis, BM-cDCs were treated with DFO overnight and plated at 2 × 10 5 /well onto Seahorse XF96 cell plates. Mito stress and Glycolysis stress test were performed with the Seahorse XF96 Extracellular Flux Analyzer (Agilent) flowing manufacturer's instructions. Glutamine concentration was evaluated in supernatant from overnight DFO-treated BM-cDCs by ELISA (Abcam). For intracellular lipid detection, BM-cDCs were incubated with Bodipy FL or Bodipy-C16 (Invitrogen) at RT for 15 min in PBS. For ex vivo staining, mice were injected 50 μg/g bw DFO i.p. for two consecutive days Mice were euthanised the following day and spleens were digested with 400 U/ml Collagenase D (Roche). Splenocytes were stained with Bodipy FL or Bodipy-C16 as explained above. Statistical analysis. Statistical analysis was carried out with Graph Prism 8 (GraphPad Software LLC) and significance was evaluated by Student' two tailed t test. For GTT and ITT, statistical significance was evaluated with 2-way ANOVA followed by post-hoc Bonferroni test. Differences within means were significant when P values of 0.05 or less. Normality was assessed using the Kolmogorov-Smirnov and Shapiro-Wilk test. Data are represented as mean and standard deviation.