All experiments were approved by the Behörde für Gesundheit und Verbraucherschutz Hamburg (N082/2020). Mice were housed at the UKE animal facility at room temperature and a 12h light/12h dark cycle with permanent access to food and water. C57BL/6 wild type mice were bred in-house or obtained from Charles River. Prof. Juergen Schrader provided Cd73^fl/fl mice (University Hospital of Duesseldorf, Germany), which were crossed with Tg(Cdh5-cre/ERT2)1Rha mice (45). By this breeding, we generated Cd73^fl/fl-Cdh5^Cre+ and control litters (Cd73^fl/fl-Cdh5^Cre-). To induce the endothelial-specific Cd73-knockout, the mice were gavaged on 3 consecutive days with 100 µl corn oil containing 2 mg tamoxifen. After two weeks, the mice received another tamoxifen gavage to deplete CD73 also in newly-formed endothelial cells. Age-matched male and female mice (10-16 weeks) were used for the experiments. For cold exposure and for thermoneutral studies, mice were exposed to cold (6°C), at mild cold (22°C) or housed at thermoneutrality (30°C) for 1 week, respectively, using a temperature- and humidity-controlled cabinet (Memmert). For organ harvest and blood collection, mice were anaesthetised by ketamine (180 mg/kg) and xylazine (24 mg/kg) injection after 4 h fasting.

The Ethics Committee of the Hamburg Chamber of Physicians approved the study (PV4889). For the isolation of stromal vascular fractions, WAT samples from three male patients with morbid obesity (age between 36 and 45 years) during bariatric surgery were harvested at the Department of General, Visceral and Thoracic Surgery, University Medical Center Hamburg-Eppendorf. All those participating signed and returned an informed consent.

Cell types were isolated from WAT and BAT as described previously (46). In brief, inguinal depots of WAT or interscapular depots of BAT of 4 mice were pooled and digested for 45 min using a collagenase D (1.5 U/ml) and dispase II (2.4 U/ml) containing buffer. After filtering through a 100 µm cell strainer, the flow through was pelleted by centrifugation for 5 min, at 600 x g. The remaining sample was resuspended (PBS containing 2 mM EDTA, 0.5% BSA, 2 mM glucose), passed through a 40 µm cell strainer and centrifuged (5 min, 600 x g). The cell pellet was resuspended and incubated for 15 min with CD11b MicroBeads (Miltenyi, 10 µl beads/10⁷ cells) for isolation of the macrophage fraction. The CD11b+-fraction was separated using magnetic LS-columns (Miltenyi). The remaining sample was incubated with CD31 MicroBeads (Miltenyi) for endothelial cell isolation. The remaining flow through fraction was considered as adipocyte pool. Cell fractions were centrifuged and the obtained pellets dissolved for RNA isolation in TRIzol^® reagent.

RNA was isolated from whole organ samples, BAT and WAT cell fractions or primary adipocyte cell cultures using TRIzol^® reagent and a NucleoSpin RNAII kit (Macherey & Nagel) as described in protocol of the company. RNA concentration was determined using NanoDrop^® spectrophotometer. Then, 400 ng of RNA was transcribed into cDNA (High-Capacity cDNA Reverse Transcription Kit, Applied Biosystems) following the manufacturer’s instructions. Quantitative real-time RT-PCR was performed on a QuantStudio 5 Real- Time-PCR System (ThermoFischer Scientific) using TaqMan^® Assay-on-Demand primer sets. Taqman^® assays used in this study: h36B4 (Hs99999902_m1), hACACA (Hs01046047_m1), mAcaca (Mm01304285_m1), hCD36 (Hs00169627_m1), mCd36 (Mm00432403_m1), hCHREBP (Hs00975710_g1), mChrebp (Mm00498811_m1), hCHREBP_BETA (AIT9559), mChrebp_beta (AIVI4CH), mDio2 (Mm00515664_m1), hFASN (Hs00188012_m1), mFasn (Mm00662319_m1), hGLUT4 (Hs00168966_m1), mGlut4 (Mm01245502_m1), hLPL (Hs00173425_m1), mLpl (Mm00434764_m1), mNt5e (Mm00501910_m1), hSREBF1 (Hs01088691_m1), mSrebf1 (Mm00550338_m1), hSREBP1C (AI70MDO), mSrebp1c (AI89KJW), hSREBP2 (Hs01081778_m1), mSrebf2 (Mm01306292_m1), mTbp (Mm00446973_m1), mUcp1 (Mm00494069_m1). Relative mRNA-expression was normalized to mTbp (TATA-box binding protein) or h36B4 (hRPLP0= ribosomal protein lateral stalk subunit P0) as housekeeper and calculated via ΔΔCT method.

For oral glucose fat tolerance test (OGFT), mice received orally a lipid-glucose-emulsion containing glucose (2 mg/g body weight) and triglycerides (3.7 mg/g body weight), containing ¹⁴C-deoxyglucose (1.7 kBq/g body weight) and ³H-triolein (1.7 kBq/g body weight) as radiolabelled tracers after 2 hours of fasting. At time points indicated in the figures, glucose concentrations were determined in blood taken from tail vein using commercially available AccuCheck Aviva glucose sticks (Roche). Then, mice were anaesthetized and tissues were collected after transcardial perfusion with phosphate buffer saline containing 10 U/ml heparin. After dissolving organs using Solvable™, radioactivities were determined using liquid scintillation counting (Tricarb, Perkin Elmer).

Indirect calorimetry was performed in metabolic cages (PhenoMaster, TSE systems). For data acquisition, mice were exposed to various temperatures to monitor oxygen consumption and carbon dioxide production. Based on respiratory data, energy expenditure was calculated as kcal per hour.

Plasma lipid levels (cholesterol, triacylglycerides) were measured by enzymatic-colorimetric assays (Roche) following the instructions of the manufacturer.

For primary brown and white adipocyte cell culture, stromal-vascular fractions (SVF) from BAT and WAT of wild type mice (C57BL/6J, male, age 4-6 weeks) were isolated as described previously (47). To obtain mature adipocyte, cells from SVF were cultured in DMEM/high glucose GlutaMAX (Gibco) supplemented with 10% neonatal calf serum, 1% penicillin and streptomycin, 1% antibiotic-antimycotic, 2.4 nM of insulin, and either 1 µM (brown adipocytes) or 100 nM (white adipocytes) of rosiglitazone (Cayman Chemicals).

For primary human white adipocyte cell culture, stromal-vascular fractions of subcutaneous WAT were isolated (as described for murine cells above) and expanded in DMEM/high glucose GlutaMAX (Gibco) containing 20% FCS (Gibco). Cells were differentiated to mature adipocytes in DMEM/high glucose GlutaMAX supplemented with 5% FCS, 1% penicillin and streptomycin, 0,1 µM dexamethasone (Sigma), 450 µM IBMX (Sigma), 2 µM insulin (Sigma), 1 µM rosiglitazone (Cayman Chemicals) and 1 µM U0126 (Sigma).

During adipogenic differentiation, cells were incubated with/without adenosine (10 µM) and/or the adenosine deaminase (ADA) inhibitor erythro-9-(2-hydroxy-3-nonyl)adenine (EHNA, 10µM) and/or adenosine kinase (ADK) inhibitor 5-(3-Bromophenyl)-7-[6-(4-morpholinyl)-3-pyrido[2,3-d]byrimidin-4-amine hydrochloride (ABT-702, 100 nM). Media change was performed every day. Adipocytes were harvested after PBS washing in TRIzol^® reagent for RNA isolation or in RIPA buffer (20 mM Tris-HCl, pH 7.4; 5 mM EDTA; 50 mM NaCl, 10 mM Na-Pyrophosphate; 50 mM NaF, 1% Nonidet P40) for protein analysis.

Snap-frozen adipose tissues were lysed in RIPA buffer supplemented with protease inhibitors (Complete Mini protease inhibitor cocktail; Roche), 0.1% SDS and phosphatase inhibitors (1% phosphatase inhibitor A + B; bimake) using the TissueLyser (Qiagen). Protein concentrations were determined using the BCA method. Protein lysates were separated by SDS-PAGE (10% acrylamide gels) and for Western blotting, transferred to nitrocellulose membranes. After blocking in 5%-milk in TBS-T (20 mM Tris, 150 mM NaCl, 0.1% (v/v) Tween 20) or ROTI^®Block (ROTH; phosphorylated proteins) for 1 h, the membranes were incubated with the indicated primary antibodies over night at 4°C (diluted in 5% BSA in TBS-T). Secondary HRP-conjugated antibodies were diluted 1:5000 and incubated for 1 h. Following primary antibodies were used: rabbit-anti-ACC (1:1000; Cell Signaling #3662; RRID : AB_2219400), rabbit-anti-pACC (1:500; Cell Signaling #3661; RRID : AB_330337), rabbit-anti-AMPKα (1:1000; Cell Signaling #2532; RRID : AB_330331), rabbit-anti-pAMPKα (1:1000; Cell Signaling #2531; RRID : AB_330330), rabbit-anti-ChREBP (1:1000; Novus #NB400-135; RRID : AB_10002435), rabbit-anti-gTubulin (1:2000; Abcam #ab179503; RRID: N/A), mouse-anti-FASN (1:1000; BD Bioscience #610962; RRID : AB_398275). Following secondary antibodies were used: Peroxidase-conjugated AffiniPure goat-anti-rabbit IgG (H+L) (1:5000; Jackson ImmunoResearch #111-035-144; RRID : AB_2307391), Peroxidase-conjugated AffiniPure goat-anti-mouse IgG (H+L) (1:5000; Jackson ImmunoResearch #115-035-146; RRID : AB_10015289). The blots were developed with enhanced chemiluminescence using Amersham Imager 600 (GE Healthcare). Signal quantification was performed using ImageStudio Lite (Licor; RRID : SCR_013715).

Data are expressed as mean ± S.E.M. Two groups were compared by unpaired, two-tailed Student’s t test, more than two groups were compared by ANOVA. The statistical parameters such as p values and numbers of replicates are presented in the figure legends and Supplementary Table 1 . Statistical analysis and presentation of data were conducted using Microsoft Excel 2016 (RRID : SCR_016137) and GraphPadPrism 9.0 (RRID : SCR_002798).

As CD73 (encoded by Nt5e) generates extracellular adenosine, we determined the expression of Nt5e in murine BAT ( Figure 1A ) and WAT ( Figure 1B ) in a cell-type specific manner. For this approach, we performed magnetic-activated cell-sorting (MACS^®) to separate adipocytes, tissue-resident macrophages and endothelial cells. The purity of the isolated cell fractions was confirmed by the expression of specific genes characteristic for brown (Ucp1) and white (Adipoq) adipocytes, macrophages (Emr1) and endothelial cells (Gpihbp1). Notably, Nt5e expression was predominantly detected in the endothelial cell fractions of BAT and WAT, respectively. These data suggest that primarily vascular endothelial cells of adipose tissues are equipped to generate extracellular adenosine.

To study the role of vascular-derived adenosine in adipose tissue function under different environmental conditions, we generated endothelial-specific CD73-knockout mice (CD73 ^fl/fl-Cdh5^Cre+) and compared them to control littermates (CD73 ^fl/fl-Cdh5^Cre-). In BAT and WAT of these tamoxifen-inducible CD73 ^fl/fl-Cdh5^Cre+ mice, Nt5e gene expression was reduced compared to controls on whole tissue level ( Figure 2A ). Successful cell type-specific deletion was confirmed in MACS^®-isolated CD31-positive endothelial cells of BAT ( Figure 2B ) and inguinal WAT ( Figure 2C ) from CD73 ^fl/fl-Cdh5^Cre+ mice. In the flow-through fractions representing adipocytes, CD73 expression was very low and unaltered between genotypes.

Previously, stimulation of the adenosine receptors, A2A and A2B, has been shown to induce thermogenesis in BAT and to stimulate browning in WAT suggesting this pathway as novel target for anti-obesity therapies (41–43). To study the potential relevance of CD73-generated adenosine in the context of thermogenic activation, we performed metabolic studies in CD73 ^fl/fl-Cdh5^Cre+ and control mice housed at room temperature (22°C; mild cold for mice) as well as cold (6°C) conditions. First, to assess a potential effect on systemic glucose metabolism, we performed an oral glucose fat tolerance test (OGFT). In CD73 ^fl/fl-Cdh5^Cre+ and CD73 ^fl/fl-Cdh5^Cre- mice housed at room temperature, we detected a comparable blood glucose clearance after oral gavage ( Figure 2D ). Despite similar glucose tolerance, uptake of glucose and lipid tracers into highly oxidative tissues such as BAT, heart and muscle was higher in CD73 ^fl/fl-Cdh5^Cre+ mice compared to the controls ( Figures 2E, F ). However, higher fuel disposal did not affect body and organ weights as well as plasma triglyceride and cholesterol levels compared to the control mice ( Figures S1A–D ). In line, gene expression of thermogenic (Ucp1: uncoupling protein 1, Dio2: diodinase 2), glucose uptake (Glut4: glucose transporter 4) and lipid handling marker genes (Acaca: acetyl CoA carboxylase, Fasn: fatty acid synthase, Cd36, Lpl: lipoprotein lipase) was unaffected in BAT ( Figure 2G ) and WAT ( Figure 2H ).

To maximally challenge the thermogenic response, CD73 ^fl/fl-Cdh5^Cre+ and CD73 ^fl/fl-Cdh5^Cre- mice were acclimatized to 6°C for one week. Similar to mice housed at room temperature, body and organ weights, plasma lipid levels, glucose tolerance, as well as expression of thermogenic and lipid processing genes in BAT and WAT were largely unaffected between control and endothelial cell-specific CD73-deficient mice ( Figures S1E–K ). Moreover, under sustained cold exposure conditions glucose and lipid tracer uptake into metabolically active tissues was unaltered ( Figures S1L, M ). To directly assess the relevance of endothelial CD73 in the regulation of energy expenditure during different temperature conditions, CD73 ^fl/fl-Cdh5^Cre+ and control mice were housed in a temperature-controlled indirect calorimetry system. In this setup, the mice were housed at thermoneutrality (30°C) and then exposed to decreasing ambient temperatures (5°C per day) until reaching 5°C. With this approach to study systemic energy metabolism under conditions of thermogenic activation, lack in endothelial CD73 did not affect whole body energy expenditure compared to controls ( Figure 2I ). Taken together, these data indicate that endothelial CD73 is dispensable for functional thermogenic adaptation to cold conditions.

To investigate the role of endothelial CD73 at human standard environmental temperature conditions, transgenic male mice were acclimated to thermoneutrality (30°C). Mice housed at this ambient temperature display a lower metabolic rate that more closely resembles the normal metabolic status of humans living at room temperature (48, 49). Body and organ weights as well as plasma lipid levels were similar ( Figures S2A–D ). Compared to control mice, CD73^fl/fl -Cdh5^Cre+ mice showed a slight, but not significant, improvement in blood glucose clearance after oral gavage ( Figure 3A ). Consistently, lack in endothelial CD73 resulted in higher glucose but not lipid tracer uptake into gonadal and inguinal WAT depots but not any other organs ( Figures 3B, C ). Under anabolic conditions of excess carbohydrate availability, glucose internalised by brown and white adipocytes can be used as a precursor for DNL (3). Lack in endothelial CD73 did not significantly alter the expression of DNL-regulating transcription factors and enzymes in BAT ( Figure 3D ). However and consistent with the higher glucose uptake exclusively in WAT ( Figure 3B ), gene expression of DNL-marker genes namely total Chrebp, Chrebp-beta, Srebf1, Acaca and Fasn was significantly higher in WAT of CD73 ^fl/fl-Cdh5^Cre+ mice compared to controls ( Figure 3E ). Similar to the males, thermoneutrally housed female mice lacking endothelial CD73 are characterized by an improved glucose tolerance, displayed increased uptake of glucose tracers in adipose tissue depots and higher expression of DNL genes compared to controls ( Figures S2E–K ).

Thus, higher glucose uptake and upregulated DNL genes observed in WAT of endothelial cell-specific CD73 deficient mice could be a consequence of lower local adenosine levels. To study the direct effect of adenosine on DNL gene expression, primary brown and white murine adipocytes were differentiated in the absence or presence of 10 µM adenosine. In fact, supplementation of adenosine results in lower gene expression of Chrebp as well as its targets Acaca and Fasn in brown and white adipocytes compared to control (mock) treatment ( Figures 3F, G ). Moreover, glucose (Glut4) and lipid uptake mediators (Cd36, Lpl) displayed lower expression upon adenosine treatment. In contrast to the in vivo situation, the adenosine effect on gene expression was also present in primary brown adipocytes. To evaluate a potential impact of rapid adenosine degradation in this setup, the cell culture media was supplemented with adenosine with or without the addition of the adenosine deaminase (ADA) inhibitor EHNA (10 µM). Both conditions caused a comparable lower expression of DNL genes in brown and white adipocytes ( Figures S3A, B ). Altogether, these data suggest a paracrine signaling pathway, in which adenosine generated by CD73 of microvascular endothelial cells regulates energy handling and lipid metabolism of white adipocytes.

Previous studies have shown that extracellular adenosine triggers AMP-dependent protein kinase (AMPK) activity in hepatocytes and an intestinal epithelial cell line (50, 51). Mechanistically, the authors propose that adenosine can enter the cell and is subsequently phosphorylated to AMP by the adenosine kinase (ADK). Independently of these studies, AMPK has been proposed to phosphorylate and thereby inactivate both ChREBP and ACC (52, 53). To investigate a possible link between the intracellular conversion of adenosine to AMP and ChREBP-dependent DNL, we incubated murine primary white adipocytes with adenosine in the absence or presence of 100 nM of the ADK-inhibitor ABT-702 (54). As shown in Figures 4A–C , adenosine supplementation resulted in lower gene and protein expression of ChREBP and DNL-mediating enzymes (ACC and FASN). Moreover, higher ratios of both pAMPK/AMPK and pACC/ACC in response to adenosine supplementation were detected ( Figures 4B, C ). Notably, the effects on gene expression, protein expression and phosphorylation were blunted by the ADK-inhibitor. These findings were validated in a human setting, where a similar modulation of CHREBP and its targets ACC and FASN by extracellular adenosine was found in primary adipocytes derived from human WAT ( Figures S4A–C ). On the other hand and consistent with lower adenosine levels, higher protein levels of ACC and FASN and a decrease in the pAMPK/AMPK ratio were detected in WAT of mice lacking endothelial CD73 ( Figures 4D, E ). Altogether, these data indicate that adenosine generated by CD73 in vascular endothelial cells shapes glucose uptake and DNL by driving an ADK-AMPK-ChREBP pathway in adipocytes and WAT depots.