Deficient Caveolin-1 Synthesis in Adipocytes Stimulates Systemic Insulin-Independent Glucose Uptake via Extracellular Vesicles

Adipose tissue (AT) is one of the largest metabolic organs by mass in most humans. As such, adequate lipid storage and endocrine functions of this organ are essential for the overall metabolic health of the organism (1). In obesity, adipocytes and AT become dysfunctional, resulting in severe tissue inflammation and fibrosis (1). The defining features of an adipocyte are the presence of large single or multilocular lipid droplets and a plasma membrane replete with 50–100-nm flask-shaped invaginations called caveolae. These lipid rafts are enriched with cholesterol and sphingolipids and specialized for facilitating signal transduction through multiple plasma membrane receptors. Caveolins are the structural proteins that are indispensable for caveolae formation. In adipocytes and endothelial cells, the dominant caveolin isoform is caveolin-1 (cav1). Although cav1 is essential for caveolae formation at the plasma membrane, caveolin protein has also been found in most subcellular compartments (2). This observation reflects the fact that cav1 plays diverse roles throughout the cell, including signal transduction, secretory pathway vesicle trafficking, lipid homeostasis, and mitochondrial function (2). In adipocytes, cav1 is particularly important in lipid droplet biogenesis and metabolism (3). Cav1 is significantly upregulated in visceral and subcutaneous AT in human obese subjects with or without type 2 diabetes (4,5). Another study reported a decrease of cav1 expression in visceral AT in obese compared with control subjects (6). Similarly, in mouse studies, cav1 is downregulated in isolated adipocytes from obese mice compared with their lean counterparts (7). The inconsistent results in humans can likely be explained by anthropometric differences between the studies; however, it is clear that the dysregulation of cav1 in AT and other tissues is detrimental during the development of diabetes in obesity in both human cells and mouse models (8). Furthermore, mutations have been identified in the cav1 gene that causes lipoatrophic diabetes in humans, similar to what has been described in cav1 knockout (KO) mice (9).

Cav1KO mice have significant metabolic and cardiovascular phenotypes. KO mice are partially lipodystrophic and resistant to diet-induced weight gain, characteristics that are consistent with human lipodystrophy caused by cav1 loss of function (10–12). They display impaired whole-body triglyceride clearance, insulin resistance in major metabolic organs, vascular abnormalities, dilated cardiomyopathy, and pulmonary hypertension (11–16). Reexpression of cav1 in endothelial cells of null mice corrected the pulmonary and cardiovascular phenotypes (17). We wanted to determine whether cav1 loss of function in adipocytes could account for the metabolic phenotypes in cav1KO mice. We generated an adipocyte-specific cav1KO mouse (ad-cav1KO) by crossing cav1-floxed mice with adiponectin-cre mice. We have previously reported that the ad-cav1KO mouse is not a true KO; while the cav1 mRNA is depleted, the cav1 protein is only reduced by 40–50% (18). This is due to extensive trafficking of cav1 in small extracellular vesicles (sEVs) from neighboring endothelial cells to adipocytes (18). The sEVs are 30–150-nm vesicles released from many cell types and are enriched with various macromolecules capable of inducing significant signaling events in the receiving cells. These include miRNAs, mRNAs, signaling proteins, enzymes, and metabolites (19). In the case of ad-cav1KO mice, the sEV-associated cav1 protein that reconstitutes the adipocyte cav1 pool is not able to form caveolar structures at the plasma membrane (18). Thus, this model is in many ways a functional KO despite the presence of the cav1 protein.

In this study, we determined the effect of adipocyte cav1 loss of function on whole-body metabolism in further detail. We found that ad-cav1KO mice on a short-term high-fat diet (HFD) exhibited the opposite glucose metabolism phenotype of whole-body cav1-null mice. Ad-cav1KO mice displayed improved glucose tolerance, despite a loss of glucose-stimulated insulin secretion (GSIS) and partial lipodystrophy. This was due to increased insulin-independent glucose uptake and anaerobic glycolysis in white AT (WAT) depots and reduced hepatic gluconeogenesis. Reliance on glycolysis and fermentation for ATP production is likely the result of significantly impaired mitochondrial function in ad-cav1KO mouse WAT. The loss of adipocyte lipid storage capacity and impaired mitochondrial function contribute to the severe inflammation and fibrosis that are characteristic of ad-cav1KO mouse fat tissue. We found that AT-sEVs from ad-cav1KO mice were sufficient to induce this same glucose clearance phenotype when injected into control mice, suggesting that AT-sEVs can relay this phenotype to otherwise healthy tissues. In fact, RNA sequencing analysis demonstrated that AT-sEVs from ad-cav1KO mice can propagate the negative AT phenotype to the liver, triggering increased collagen production and inflammation. We therefore conclude that adipocyte cav1 promotes AT health during the expansion associated with overnutrition and exerts important systemic functions by preventing the spread of adverse metabolic phenotypes to other organs by sEVs.

All procedures performed on animals have been approved by the institutional animal care and use committee of the University of Texas (UT) Southwestern Medical Center. All experiments were conducted using littermate-controlled male mice. Males were chosen for this study to be able to compare the data with the body of literature describing whole-body cav1KO mice, which was mostly conducted in males. Experiments were conducted in chow-fed mice at 10–12 weeks of age. For HFD studies, mice were started on the diet at 10 weeks of age (60% calories from fat, S5867; Bio-Serv). Cav1-floxed mice in a Friend leukemia virus B (FVB) background were bred to adiponectin-cre mice with a mixed C57BL/6 and FVB background [B6;FVB-Tg(Adipoq-cre)1Evdr/J, stock no. 028020; The Jackson Laboratory). Whole-body cav1KO mice were used in the FVB background. Mice were maintained on a 12-h dark/light cycle with unrestricted access to chow and water.

GW4869 was dissolved in DMSO at 5 mg/mL then diluted to 0.3 mg/mL in PBS. Each mouse received a 200 μL injection solution (60 μg/mouse) once a week. Control mice were injected with a 6% DMSO solution in PBS. Where indicated, mice were injected intravenously with 1 × 10⁹ AT EVs once a week. Control mice were injected with an equal volume of PBS.

For the glucose tolerance test (GTT), mice were fasted for 3 h before the administration of 1.25 g/kg body weight i.p. glucose. At the indicated time points, venous blood samples were collected in capillary tubes from the tail vein. Glucose levels were measured using an oxidase/peroxidase assay (MilliporeSigma), and insulin concentration was determined using a commercial insulin ELISA kit (90095; Crystal Chem). Mice were also fasted for 3 h before the insulin tolerance test (ITT) where mice received 1 unit/kg body weight insulin i.p. Blood glucose was measured at the indicated time by glucometer. For the triglyceride clearance procedure, mice were fasted for 15 h, then administered 20% Intralipid (15 μL/g body weight, I141; MilliporeSigma) by gastric gavage. Blood was taken at the indicated time points for serum measurements of triglycerides (Infinity Triglyceride Reagent, TR22421; Thermo Fisher Scientific, Waltham, MA) and nonesterified free fatty acids (NEFAs) [HR Series NEFA-HR (2); Fujifilm Wako Diagnostics] measurements. For in vivo lipolysis assay, mice were administered with β3-adrenergic receptor agonist CL-316,243 (C5976; MilliporeSigma) at a dose of 1 mg/kg body weight. Blood was collected from the tail vein at the indicated time points for serum. NEFA and glycerol (F6428; MilliporeSigma) were assayed in serum samples. Before an arginine tolerance test (ArgTT), a pyruvate tolerance test (PTT), or a glutamine tolerance test (GlnTT), mice were fasted for 15 h and then administered 1 g/kg arginine, 2 g/kg pyruvate, or 2 g/kg glutamine i.p., respectively. Blood glucose was measured at the indicated time points by glucometer. For the ArgTT, plasma samples were collected at each time point for insulin determination by ELISA.

Mice were sacrificed, tissues were collected, and ∼100 mg of each tissue was sent to the UT Southwestern Metabolic Phenotyping Core for tricarboxylic acid cycle (TCA) intermediate profiling using liquid chromatography tandem mass spectrometry technology (Nexera X2 UHPLC coupled to an LCMS-8060 triple-quadrupole mass spectrometer; Shimadzu Scientific Instruments, Columbia, MD).

Frozen liver tissues were used for lipid extraction and measurement as previously described. In brief, 25–35 mg of liver tissue was homogenized in 0.5–0.7 mL PBS (20× liver weight in μL). Homogenates (0.4 mL) were then mixed with 1.6 mL of CHCl₃-CH₃OH (2:1 v/v), and the suspension was centrifuged at 3,000 rpm for 10 min at room temperature. The lower organic phase was transferred and air dried overnight in a chemical hood. The residual liquid was resuspended in 800 mL of ethanol containing 1% Triton X-100, and the concentrations of triglycerides and cholesterol were determined using a serum triglyceride and cholesterol determination kit (Thermo Fisher Scientific).

The metabolic cage studies were performed by the UT Southwestern Medical Center Metabolic Phenotyping Core. Briefly, the mice were maintained on a 12 h dark/light cycle at room temperature. Metabolic parameters, including oxygen consumption, carbon dioxide generation, food intake, and water consumption, were monitored and recorded continuously using the TSE PhenoMaster system. All transgenic mice and their littermate controls were single housed in the metabolic chambers.

The measurement of mouse whole-body composition, including total body fat and lean mass, were performed with the minispec mq10 system (Bruker Scientific, Billerica, MA). Serum adiponectin and leptin levels were determined by commercial ELISA kits (90080 and 90030; Crystal Chem). Serum ketones were measured within 2 h of collection using the Wako Autokit Total Ketone Bodies (415-73301; Wako Chemicals).

¹⁴C-labeled 2-deoxyglucose (2-DG) (13 μCi/mouse) was administered after a 3 h fast by retro-orbital injection. After 25 min, tissues were rapidly dissected and frozen. A similar-sized piece of each tissue was dissected from each mouse. The uptake of 2-DG was calculated as previously described (20).

After 5 weeks of HFD feeding, mice were euthanized and perfused with 6 mL PBS through the left ventricle of the heart. Subcutaneous WAT (sWAT) depots were harvested and minced in Hanks’ balanced salt solution (HBSS). The remaining steps in the protocol were conducted under sterile conditions in a laminar flow hood. The minced tissue was transferred to a 100 μm cell strainer and rinsed with 40 mL of sterile HBSS to remove debris from damaged cells. The washed tissue pieces were then transferred to a 10 cm cell culture dish containing cell culture media (FluoroBrite DMEM supplemented with 10% sEV-depleted FBS, EXOFBSHI50A1; System Biosciences) and 1× penicillin/streptomycin). Tissues were incubated for ∼16 h in a cell culture incubator (37°C, 5% CO₂). Media were recovered and centrifuged at 600g for 15 min at 4°C to remove any cells. The supernatant was then centrifuged at 1,200g at 4°C for 20 min to remove cell debris and apoptotic bodies. The resulting supernatant was centrifuged at 10,000g for 30 min at 4°C to remove large EVs. The final supernatant was concentrated to 1 mL using a centrifugal filter with a 100-KD cutoff (Amicon Ultra-15, UFC910024; MilliporeSigma). The full 1 mL was loaded onto a 10 mL gravity flow size exclusion column (89898; Pierce) packed with Sepharose CL-2B particles (CL2B300; MilliporeSigma) as previously described (21). Sterile PBS (10 mL) was used to elute sEVs from the column, and 1 mL fractions were collected. Fractions 2–5 were found to have the highest and purest sEV yield, so they were combined and concentrated again to ∼500 μL. sEV concentration was determined using the ZetaView nanoparticle tracking analyzer (Particle Metrix).

Inguinal fat pads were dissected from 4–6-week-old mice. Fat tissues were minced and digested for 1 h at 37°C in buffer containing 100 mmol/L HEPES, 120 mmol/L NaCl, 50 mmol/L KCl, 1.5% BSA, 5 mmol/L glucose, 1 mmol/L calcium, and 1 mg/mL collagenase D. The dispersed tissue was then filtered through a 100-μm cell strainer and centrifuged at 600g for 5 min at 4°C. The pelleted stromal vascular fraction (SVF) cells were resuspended in culture media (DMEM/F12 containing 10% FBS, GlutaMax, 1× penicillin/streptomycin, and gentamicin) and filtered through a 45 μm cell strainer. Cells were centrifuged again as described above. The pelleted cells were resuspended in culture media and grown at 37°C in 5% CO₂. For in vitro differentiation experiments, SVF cells were allowed to grow to ∼95% confluency. Adipogenesis was induced by culture media supplemented with 500 μmol/L 3-isobutyl-1-methylxanthine, 1 μmol/L dexamethasone, 5 μg/mL insulin, and 1 μmol/L rosiglitazone for 2 days. Following the 2 days of induction, cells were maintained in media containing only 5 μg/mL insulin. Cells were used for sEV secretion assays at 8 days of differentiation.

sEVs were purified from adipocyte culture media by ultracentrifugation. Freshly harvested media containing 10% EV-depleted FBS (EXOFBSHI50A1; System Biosciences) was centrifuged at 600g for 15 min at 4°C to remove whole cells. The supernatant was then centrifuged at 1,200g for 20 min at 4°C to remove cell debris and apoptotic bodies. The resulting supernatant was centrifuged at 10,000g for 30 min at 4°C to remove large EVs. The final supernatant was centrifuged in a swinging bucket rotor at 100,000g for 1 h to pellet sEVs. The pellet was resuspended in PBS and centrifuged again at 100,000g for 1 h. Purified sEVs were stained with PKH26 per the manufacturer’s instructions, and excess dye was removed with a sucrose cushion. Mice received 2 × 10¹⁰ PKH26-labeled sEVs via retro-orbital injection every 12 h for a 24 h period. Mice were euthanized by cervical dislocation under isoflurane. The heart was perfused with ice cold PBS to remove blood. Tissues were sliced into pieces ∼10 mm thick. Pieces were mounted onto a microscope slide with Dako Fluorescence Mounting Medium, and PHK26 fluorescence was detected using a Zeiss LSM8800 Airyscan confocal microscope. For 1,1'-dioctadecyl-3,3,3',3'-tetramethylindotricarbocyanine iodide (DiR) AT-sEV membrane labeling and biodistribution, AT-sEVs were isolated as described in the Ex Vivo AT Incubation for sEV Production section. Concentrated AT-conditioned media were incubated with DiR (1 μmol/L, D12731; Thermo Fisher Scientific) for 15 min. A size exclusion column was used to isolate sEVs and remove excess dye. sEVS (3 × 10⁸) were injected into mice retro-orbitally, and mice were euthanized 16 h later. A dye-only condition was used as control where 1 μmol/L DiR was incubated with PBS instead of conditioned media and processed with the rest of the samples. Tissues were harvested and visualized with the Chemidoc MP Gel imaging system (Bio-Rad).

AT was disrupted in ice cold isolation buffer (10 mmol/L MOPS, 1 mmol/L EDTA, 210 mmol/L mannitol, and 70 mmol/L sucrose [pH 7.4]) using a Potter-Elvehjem homogenizer. The homogenate was centrifuged at 600g for 10 min at 4°C. The resulting supernatant was filtered through cheese cloth and centrifuged at 10,000g for 15 min to pellet mitochondria. Mitochondria were resuspended in isolation buffer to a final concentration of 15 mg/mL. Five micrograms of mitochondria was transferred to each well of Xfe24 cell culture microplates and centrifuged at 3,500g for 20 min at 4°C to pellet the mitochondria at the bottom of the plate. The plate was then subjected to a 5 min equilibration period in a 37°C incubator. MAS1 buffer (70 mmol/L sucrose, 220 mmol/L mannitol, 10 mmol/L KH₂PO₄, 5 mmol/L MgCl₂, 2 mmol/L HEPES, 1 mmol/L EGTA, and 0.2% fatty acid–free BSA [pH 7.2]) (450 μL) containing 10 mmol/L pyruvate, 2 mmol/L malate, and 4 μmol/L carbonyl cyanide-p-trifluoromethoxyphenylhydrazone was carefully added to each well without disturbing the pellet. Where noted, 25 μmol/L palmitoylcarnitine was used instead of pyruvate. Oxygen consumption rates were determined at room temperature using a Seahorse Flux Analyzer following sequential additions of rotenone (2 μmol/L final), succinate (10 mmol/L final), antimycin A (4 μmol/L final), and ascorbate (10 mmol/L final)/tetramethyl-p-phenylenediamine (100 μmol/L final).

Following treatment, 500 μL of media was cleared through sequential centrifugation: 600g for 15 min, then 1,200g for 20 min. The cleared supernatant was diluted 1:100 in PBS and analyzed using the ZetaView nanoparticle tracking analyzer. For sEV quantification in serum samples, 10 μL of serum was diluted to 1 mL in PBS. Samples were filtered through a 0.2 μm syringe-driven filter, followed by a 1:100 dilution for nanoparticle tracking analysis with ZetaView. sEVs were taken to have the size distribution of 45–200 nm.

Islets were isolated as previously described (22). Briefly, the mouse was euthanized by cervical dislocation under full anesthesia. The ampulla was clamped with surgical clamps on the duodenum wall to block the bile entry to the duodenum. Four milliliters of collagenase solution (1,000 units/mL, C7657 Collagenase IX; Sigma) was slowly injected into the common bile duct to distend the pancreas. The pancreas was removed and placed into a 50 mL conical tube containing collagenase solution and incubated in a water bath at 37.5°C for 15 min. After digestion, the dispersed cells were pelleted by centrifugation at 390g for 30 s at 4°C. The pellet was resuspended in 20 mL of ice cold HBSS, 1 mmol/L CaCl₂ and centrifuged at 390g for 30 s at 4°C. Cells were resuspended in HBSS, 1 mmol/L CaCl₂. Isolated islets were handpicked using a pipette with a wide-open tip, counted, and placed in a 5% CO₂ incubator at 37°C. Isolated islets of three mice with the same genotype were pooled for the GSIS assay (four replicates/condition and 20 islets/well). After isolation, islets were cultured overnight in 11 mmol/L RPMI medium. The next morning, the media were switched to starvation media (2 mmol/L RPMI medium) for 2 h before stimulation. The starvation media were removed from the islets and replaced with 1 mL of Krebs-Ringer bicarbonate HEPES buffer (140 mmol/L NaCl, 3.6 mmol/L KCl, 0.5 mmol/L KH₂PO₄, 0.5 mmol/L MgSO₄, 2.5 mmol/L CaCl₂, 10 mmol/L NaHCO₃, 10 mmol/L HEPES [pH 7.4], and 0.7% w/v fatty acid–free BSA) supplemented with the following compounds: 5.5 mmol/L glucose, 16.7 mmol/L glucose, or 35 mmol/L KCl. Following 1 h of static incubation, supernatants were collected for insulin determination using a commercial insulin ELISA kit (90095; Crystal Chem). The result was normalized to total islet insulin content. Total insulin extraction of islets was prepared by adding 1 mL of 75% ethanol and 1.5% v/v HCl to each well and incubating at 4°C overnight. The next day, samples were vortexed and centrifuged at 5,000g. A 200-fold dilution of the supernatant was made for insulin measurement by ELISA.

sWAT or liver was excised and fixed in 10% PBS-buffered formalin for 24 h. Tissues were paraffin embedded and sectioned (5 μm). Following hydration, tissues were stained with hematoxylin-eosin as well as Masson’s trichrome C staining or primary antibodies as previously described (18) (see Supplementary Table 1). Collagen hybridizing peptide (Advanced BioMatrix) was used to stain collagen in liver sections according to the manufacturer’s instructions.

Digital microscopy images were analyzed using ImageJ software. Each image was subjected to the following process: tissue was thresholded apart from the background, the analyze particles command was used to identify adipocytes as large particles of background-level pixel intensity in the thresholded image, and the extracellular matrix thickness surrounding each adipocyte was calculated using the thresholded image. The analyze particles command was operated with the following detection settings: minimum circularity of 0.4, maximum circularity of 1.0, minimum particle size of 150 μm², and maximum particle size of 30,000 μm², with the “exclude on edges” option checked.

Protein was extracted from AT by homogenization in PBS supplemented with 1 mmol/L EDTA, 20 mmol/L NaF, 2 mmol/L Na₃VO₄, and protease inhibitor cocktail. Five times the final concentration of radioimmunoprecipitation assay (RIPA) buffer was added to the homogenate for a final concentration of 10 mmol/L Tris-HCl, 2 mmol/L EDTA, 0.3% NP40, 0.3% deoxycholate, 0.1% sodium dodecyl sulfate, and 140 mmol/L NaCl (pH 7.4). The sample was cleared by centrifugation at 10,000g for 5 min. Twenty to 50 μg/lane of supernatant protein was separated by SDS-PAGE (NP0335BOX; Thermo Fisher Scientific) and transferred to nitrocellulose membrane. The blots were then incubated overnight at 4°C with primary antibodies in a 1% BSA Tris-buffered saline with Tween blocking solution. An Odyssey Infrared Imager was used to visualized Western blots with LI-COR IRDye secondary antibodies. See Supplementary Table 1 for a list and the source of all primary antibodies.

Tissues were homogenized in TRIzol (12034977; Thermo Fisher Scientific) using the QIAGEN TissueLyser II. RNA was isolated per the manufacturer’s protocol. RNA yield and quality were determined using a NanoDrop spectrophotometer. cDNA was prepared by reverse transcribing 1 μg of RNA with the iScript cDNA Synthesis Kit (cat. no.1708890; Bio-Rad). Gene expression was calculated by the standard threshold cycle method using β2-microglobulin for normalization. For each experiment, it was confirmed that β2-microglobulin did not change between groups. Primer sequences are listed in Supplementary Table 2.

RNA was extracted from whole AT or liver and submitted to Novogene for RNA sequencing and analysis.

Electron microscopy samples were prepared by the UT Southwestern Electron Microscopy Core. Briefly, sWAT pieces were fixed by perfusion with 4% paraformaldehyde and 1% glutaraldehyde in 0.1 mmol/L sodium cacodylate buffer. The fixed tissue was then transferred to 2.5% glutaraldehyde in 0.1 mmol/L sodium cacodylate buffer, postfixed in buffered 1% osmium tetroxide, and en bloc stained in 4% uranyl acetate in 50% ethanol. Thin sections were cut on a Leica Ultracut UCT ultramicrotome and poststained with 2% uranyl acetate and lead citrate. Images were acquired on the JEOL 1200 EX transmission electron microscope.

All data are presented as mean ± SEM. Significance was determined by two-tailed Student t test or in the case of systemic assays, two-way ANOVA. P < 0.05 was considered statistically significant. All quantitative PCR data points were the average of technical duplicates. For all mouse studies, the n value corresponds to individual mice of a given treatment. Data were analyzed using GraphPad Prism software.

The data sets generated in this study are available upon reasonable request from the corresponding author. No applicable resources were generated during is study.

Confirming our previous results (18), cav1 mRNA was significantly depleted in AT from mice expressing adiponectin promotor-driven cre (Fig. 1A). This effect is specific to AT, as no changes in cav1 mRNA were detected in the liver, kidney, or gastrocnemius muscle (Fig. 1B). At the protein level, cav1 was reduced by ∼50% in the sWAT, epididymal WAT (eWAT), and brown AT (BAT) (Fig. 1C). In contrast, no changes in cav1 protein were detected in liver or kidney (Fig. 1D). Although 50% of cav1 protein remained in ad-cav1KO AT, electron microscopy images showed a complete lack of caveolar structures and disrupted mitochondrial morphology, similar to whole-body cav1KO mice compared with their controls (Fig. 1E and F). Interestingly, the ad-cav1KO adipocytes displayed structures that are reminiscent of lamellar bodies, which were not seen in the wild-type controls or the whole-body null mice (Fig. 1E). Lamellar body formation in adipocytes may be a reflection of aberrant lipid storage, as these structures are generally enriched in polar lipids that include glycerophospholipids, sphingolipids, and cholesterol, all lipid species that are found in caveolar structures (23). As cav1 is an important regulator of cholesterol metabolism and mitochondrial function (12,24), this provides further evidence that the ad-cav1KO mouse is a functional KO. After a 4-week HFD, the sWAT of ad-cav1KO mice developed large vacuolar structures that appeared empty but potentially lipid filled (Fig. 1G). Other vacuolar structures were doubled membraned and reminiscent of autophagosomes. No such structures were observed in control mouse sWAT.

We subsequently subjected the ad-cav1KO mice to comprehensive metabolic phenotyping. The body weights of ad-cav1KO mice fed a chow diet did not differ significantly from control mice (Fig. 2A). On an HFD, body weights between the genotypes began to diverge at 13 weeks of feeding (Supplementary Fig. 1A). By 16 weeks on an HFD, the ad-cav1KO mice had significantly lower body weights compared with control mice on the same diet (Fig. 2A and Supplementary Fig. 1A). Ad-cav1KO mice presented with a selective loss of AT mass on either a chow diet or an HFD compared with controls (Fig. 2B). Furthermore, the weights of the sWAT and eWAT fat pads were significantly reduced under both chow and HFD feeding in ad-cav1KO mice compared with controls (Fig. 2C). The loss of AT mass did not reflect a healthy phenotype, as sWAT and eWAT depots of the ad-cav1KO mice contained more infiltrating macrophages and fibrosis on either a chow diet or an HFD compared with their control littermates (Fig. 2D and F and Supplementary Fig. 1). This phenotype resembles the adipose morphology in the whole-body KO mice (Fig. 2D). Although adipocyte size was not different between genotypes, the ad-cav1KO AT displayed a wider size distribution, and cav1KO AT distribution was skewed toward smaller adipocytes (Fig. 2E). Immunofluorescent staining of sWAT depots of ad-cav1KO mice confirmed increased macrophage infiltration into the tissue and reduced adipose viability, as indicated by the presence of perilipin-negative lipid droplet structures compared with controls (Fig. 2F). No histological changes were observed in the BAT depot of ad-cav1KO mice, and changes in the eWAT mirrored that of the sWAT (Fig. 2D and Supplementary Fig. 1). Both the sWAT and eWAT of ad-cav1KO mice exhibited a reduction in the mRNA expression of key mature adipocyte markers, such as peroxisome proliferator–activated receptor-γ2 and leptin, particularly in the eWAT, compared with control mice (Fig. 2G). However, at the protein level, circulating leptin was trending toward an increase in the ad-cav1KO mice compared with controls (Fig. 2H). Furthermore, we did not detect a reduction in the major insulin-sensitizing adipokine adiponectin in the serum of ad-cav1KO mice on an HFD (Fig. 2I), which was previously reported to be reduced in the whole-body cav1KO mice (12).

Mitochondrial integrity is an important determinant of adipocyte health and function. We therefore tested whether the mitochondrial activity in the AT of ad-cav1KO mice is altered, as suggested by the electron micrographs (Fig. 1F and G). Interestingly, the sWAT depot of ad-cav1KO mice on a chow diet displayed severely reduced mitochondrial protein content (Fig. 2J), but this phenotype was absent in ad-cav1KO mice after 6 weeks of HFD feeding (Fig. 2J). Although there was an apparent normalization of mitochondrial content upon HFD feeding in the ad-cav1KO mice, mitochondria isolated from the sWAT of these mice demonstrated a reduced oxygen consumption rate when respiring on either pyruvate or palmitoylcarnitine (Fig. 2K and L). This suggests that the AT mitochondria in ad-cav1KO mice are dysfunctional.

RNA sequencing analysis of sWAT after 5 weeks of HFD feeding revealed a robust effect of cav1 loss of function on AT immune system activation. The top 10 most significantly upregulated pathways spanned a broad range of immune responses, including both innate and adaptive immunity (Supplementary Fig. 2A). Collagen synthesis and fibrosis-related pathways were also highly upregulated in ad-cav1KO sWAT, as further evidenced by trichrome and MAC2 stains of adipose depots (Fig. 2D and F). As suggested by the reduction in mitochondrial function (Fig. 2K and L), the pathways that were the most significantly downregulated at the mRNA level were related to mitochondria (Supplementary Fig. 2B and C). All aspects of cellular respiration were downregulated in ad-cav1KO mice, including mitochondrial electron transport chain constituents, TCA enzymes, and respiration-stimulating pathways controlled by key transcription factors, such as peroxisome proliferator–activated receptor γ coactivator-1α and sirtuin 3 (Supplementary Fig. 2C). Furthermore, TCA intermediates accumulated considerably in the eWAT of ad-cav1KO mice (Supplementary Fig. 3). This likely was not due to increased anaplerosis, as the mRNA levels of major anaplerotic enzymes were either unchanged (pyruvate carboxylase) or reduced (glutamine dehydrogenase 1 and malic enzyme 3) (data not shown). Thus, accumulation of TCA intermediates is likely the result of a slowed oxidation of energetic substrates by dysfunctional mitochondria.

Dysfunctional AT is a major contributor to the whole-body derangements in glucose and lipid metabolism observed in obesity and type 2 diabetes. The ad-cav1KO mice are partially lipodystrophic, which, by definition, reflects nonfunctional AT. Therefore, we wanted to determine the effects of adipocyte cav1 loss of function on whole-body glucose and lipid metabolism. On a chow diet, ad-cav1KO mice exhibit a phenotype similar to what has been reported for the whole-body cav1-null mice: impaired glucose tolerance and enhanced GSIS (Supplementary Fig. 4A and B). No changes in insulin sensitivity were observed, as measured by an ITT (Supplementary Fig. 4C). Furthermore, no difference in pyruvate-supported gluconeogenesis in the liver was detected between genotypes, as measured by a PTT (Supplementary Fig. 4D). Interestingly, in ad-cav1KO mice, insulin secretion was abrogated following an arginine injection to stimulate the exocytosis of secretory granules, as measured by an ArgTT (Supplementary Fig. 4E). Unexpectedly, fasted ad-cav1KO mice had better systemic clearance of triglycerides after an oral triglyceride gavage (Supplementary Fig. 4F), a phenotype that is opposite of the whole-body KO mouse that displays severely impaired triglyceride clearance (11). In response to an injection of the β3-adrenergic receptor agonist CL-316,243, lower levels of fatty acids and glycerol were liberated into the circulation, reflecting impaired lipolysis in adipocytes (Supplementary Fig. 4G).

The reduced glucose clearance and enhanced triglyceride clearance may indicate that the lack of cav1 in adipocytes triggers a whole-body switch in substrate preference to favor triglyceride metabolism. Indeed, metabolic cage data demonstrated that ad-cav1KO mice had a slightly lower respiratory exchange ratio during the dark cycle (Supplementary Fig. 4H). This can also reflect metabolic inflexibility, as the higher insulin levels in the fed state (dark cycle) should trigger fatty acid storage rather than oxidation in a healthy animal. No changes were detected in activity and food intake between control and ad-cav1KO mice (Supplementary Fig. 4I and J).

To further characterize the lipid metabolism phenotype, we challenged ad-cav1KO mice with an HFD. We chose to use a 4–5-week HFD challenge in subsequent studies because this is the time frame where body weight does not change between genotypes. Ad-cav1KO mice displayed a lower respiratory exchange ratio in both the light and dark phases compared with the controls (Fig. 3A and B), with no change in food intake (Fig. 3C). Consistent with this result, the rate of triglyceride clearance in fasted ad-cav1KO mice was significantly improved compared with controls, reflecting the increased lipid consumption (Fig. 3D). However, serum triglycerides and NEFAs were elevated in ad-cav1KO mice on an HFD compared with controls in the refed state (Fig. 3E), suggesting impaired AT storage capacity in line with the partial lipodystrophy of ad-cav1KO mice. Furthermore, this is consistent with the observed reduction of postprandial insulin levels in ad-cav1KO mice, which would be expected to result in less esterification of free fatty acids compared with controls. Despite the partial lipodystrophy of ad-cav1KO mice, HFD did not result in exacerbated liver steatosis as determined by histological assessment of the presence of lipid droplets (Fig. 3F), liver weight (Fig. 3G), and liver triglyceride levels (Fig. 3H). The livers from ad-cav1KO mice did, however, display higher collagen and macrophage levels compared with the controls (Fig. 3I). Although lipids are the major systemwide energy substrate in the ad-cav1KO mice on HFD, these mice did not display ketosis during fasting and had lower serum ketones in the refed state (Fig. 3J).

To our surprise, glucose tolerance was improved in the ad-cav1KO mice on a 4-week HFD challenge compared with control mice, despite the complete absence of GSIS in the KOs (Fig. 4A and B). This contrasts with the phenotype of whole-body cav1KO mice on an HFD, which displayed impaired glucose tolerance and no change in GSIS (Supplementary Fig. 5A and B). No change in glucose clearance after an insulin injection (ITT) was detected in the ad-cav1KO mice on an HFD compared with controls (Fig. 4C), while whole-body cav1-null mice on the same diet displayed an impaired ITT, suggesting an insulin resistant state (Supplementary Fig. 5C). Like the chow diet condition, insulin secretion following an arginine injection was severely blunted in the HFD-fed ad-cav1KO mice compared with controls (Fig. 4D). No change in the ArgTT was observed in whole-body cav1KO fed an HFD (Supplementary Fig. 5D). Furthermore, no histological differences were observed for islets between the genotypes (data not shown). Notably, the higher glucose tolerance exhibited by the ad-cav1KO mice was transient, present at 4–6 weeks of HFD feeding but absent during chronic HFD feeding (16 weeks) (Supplementary Fig. 5E).

Because the AT of the ad-cav1KO mice contains ∼50% of the wild-type levels of cav1 protein, we determined whether the divergent phenotypes between the adipocyte-specific and whole-body KO mice can be explained by the residual cav1 protein remaining in the conditional model. For this reason, we included the heterozygous cav1KO littermates in our metabolic analysis as a control comparison. The heterozygous cav1KO mice did not phenocopy the ad-cav1KO mice; no changes were detected in any of the metabolic tests conducted on the heterozygous cav1KO mice compared with the wild-type controls (Supplementary Fig. 5A–D).

We confirmed that the rate of glucose clearance from the blood in the ad-cav1KO mouse on HFD was higher than controls by radioactive glucose tracing (Fig. 4E). Furthermore, islets isolated from the ad-cav1KO mice on an HFD displayed moderately blunted insulin secretion under the basal (5 mmol/L glucose) and fully stimulated (35 mmol/L KCl) conditions, with significant differences seen compared with control mice (Fig. 4F). Therefore, in vitro, the β-cells responded to glucose and KCl in contrast to the in vivo conditions where insulin levels were nonresponsive during the intraperitoneal GTT (IPGTT) and ArgTT (Fig. 4B and D). An additional surprising finding is that WAT and liver displayed insulin resistance in the ad-cav1KO mice, as measured by phosphorylated AKT (pAKT) levels following an insulin injection (Fig. 4G and H). Thus, the glucose handling phenotype of the ad-cav1KO mice on an HFD suggests that an insulin-independent glucose uptake mechanism is activated systemically. To better understand this phenotype, we assessed other pathways that could result in glucose uptake without insulin stimulation. We found that unstimulated baseline levels of pAKT in the liver were increased in the ad-cav1KO mice compared with controls after HFD feeding (Fig. 4I). This could be evidence that IGF-I, rather than insulin, is mediating the AKT activation. However, we did not detect any changes in circulating IGF-I levels (Fig. 4J). Rather than explain the insulin-independent glucose uptake phenotype, these data may instead be further indication of insulin resistance in the liver, as attenuated insulin-stimulated pAKT levels in the liver are associated with higher basal levels of pAKT in obesity (25). No changes were detected in the expression of the insulin-independent GLUT1 in metabolic organs of ad-cav1KO mice, except for the gastrocnemius muscle, which exhibited reduced GLUT1 expression (Fig. 4K). Therefore, we investigated further which tissues were responsible for the insulin-independent glucose uptake. A tissue survey for glucose uptake revealed that WAT depots of ad-cav1KO mice took up significantly more glucose than the corresponding depots of control mice (Fig. 4L). The classical BAT depot and the other tissues surveyed did not display any differences in glucose uptake rates between genotypes (Fig. 4M and Supplementary Fig. 5F). Because WAT mitochondria are dysfunctional in ad-cav1KO mice, the adipocytes are likely utilizing glucose for energy production through fermentation and not oxidative phosphorylation. In keeping with this, eWAT from ad-cav1KO mice exhibited a significant two- to threefold increase in basal levels of both pyruvate and lactate compared with controls (Fig. 4N). Therefore, the observed insulin-independent glucose uptake is likely a compensatory response of WAT to both the reduction in islet insulin secretion (Fig. 4B) and the requirement of the cell to produce ATP outside the mitochondria. We also assessed whole-body gluconeogenic activity to determine whether glucose output affects the clearance rate. We used a PTT to assess primarily hepatic glucose production and a GlnTT to evaluate extrahepatic glucose production. The PTT demonstrated that despite being more insulin resistant and having lower basal circulating insulin, the livers of ad-cav1KO mice produced less glucose compared with controls on an HFD (Fig. 4O). No changes were found with the GlnTT (Fig. 4P). As a control, we did not detect any changes in glucose concentration in the urine between ad-cav1KO mice and controls, suggesting that no glucose wasting mechanisms were activated (Fig. 4Q). Therefore, the combination of enhanced WAT glucose uptake and reduced hepatic glucose output likely accounts for improved glucose clearance in ad-cav1KO mice.

We have previously demonstrated that the cav1 protein is transferred from AT endothelial cells to adipocytes in ad-cav1KO mice through sEVs (18). We also previously reported that mitochondrial dysfunction in adipocytes leads to increased sEV release (26). Indeed, we found that there was an overall increase in AT-sEVs in the sWAT but not the eWAT of ad-cav1KO mice compared with control mice on an HFD for 5 weeks (Fig. 5A). Furthermore, we detected an increase in serum sEVs in ad-cav1KO mice compared with controls under the same conditions (Fig. 5B). We determined whether these EVs play a role in the whole-body metabolic phenotype of these mice. Mice were treated with GW4869, a compound that suppresses exosome production through inhibition of neutral sphingomyelinase-2. After 5 weeks of HFD feeding, GW4869 normalized glucose clearance and prevented the drop in in vivo GSIS in ad-cav1KO mice (Fig. 5C and D). However, GW4869 increased the insulin levels in both the control and ad-cav1KO mice in the basal and glucose-stimulated state (Fig. 5D and data not shown). Therefore, we used a gain-of-function model to confirm that sEVs participated in the glucose clearance phenotype of ad-cav1KO mice. We isolated sEVs from the sWAT of control (AT-sEV^contr) and ad-cav1KO mice (AT-sEV^adKO) on an HFD for 5 weeks. Isolated sEVs were between 30 and 200 nm in size as measured by nanoparticle tracking analysis and electron microscopy (Supplementary Fig. 6A and B). sEVs also carried flotillin-1, a common EV marker, and were devoid of histone H3 (Supplementary Fig. 6C). Control mice (homozygous for cav1flox) were fed an HFD and injected with 1 × 10⁹ AT-sEVs once a week for 4 weeks. Surprisingly, injection of control mice with AT-sEV^adKO phenocopied the endogenous glucose phenotype of the ad-cav1KO mice with an improved IPGTT (Fig. 5E) and reduced GSIS (Fig. 5F) compared with mice injected with PBS. AT-sEV^contr had the opposite effect: blunted glucose clearance (Fig. 5G) and a similar GSIS (Fig. 5H) compared with mice that received PBS. The improved IPGTT in ad-cav1KO AT-sEV–treated mice was not due to compensatory IGF-I signaling, as circulating levels were lower compared with mice treated with PBS (Fig. 5I). We wanted to determine the primary site of uptake for AT-sEVs. The majority of AT-sEVs are derived from adipocytes (27). Therefore, we differentiated the SVF of sWAT from wild-type mice into mature adipocytes. sEVs were purified from the adipocyte conditioned media, and membranes were labeled with PKH26 red fluorescent dye. Wild-type mice were injected intravenously with labeled sEVs to determine the natural homing of adipocyte-derived sEVs to various metabolic organs. The liver displayed the most robust uptake of adipocyte sEVs, followed by the sWAT (Fig. 5J). We detected a small amount of PKH26-sEVs in the pancreas as well, but none in the kidney (Fig. 5J). We confirmed these findings using AT-sEVs labeled with DiR, a membrane-incorporating dye that is commonly used for sEV biodistribution studies. We found that AT-sEVs, like adipocyte sEVs, targeted mostly to the liver (Supplementary Fig. 7). AT-sEVs also robustly targeted the heart and to a lesser extent the sWAT, eWAT, and pancreas. No targeting to the gastrocnemius muscle was detected (Supplementary Fig. 7).

Because adipocyte sEVs target to the pancreas, we determined whether the loss of GSIS observed in ad-cav1KO mice was due to direct action of AT-sEVs on β-cells. MIN6 cells were cultured and treated with PBS, AT-sEV^contr, or AT-sEV^adKO for 24 h before GSIS measurements. No changes were detected between treatment groups under low-glucose conditions (2.8 mmol/L) (Fig. 5K). When MIN6 cells were stimulated with 16.7 mmol/L glucose, those pretreated with AT-sEV^contr displayed slightly augmented insulin secretion compared with those pretreated with PBS (Fig. 5K). In contrast, high GSIS was abolished in MIN6 cells pretreated with AT-sEV^adKO (Fig. 5K). These data suggest that AT-sEVs from ad-cav1KO mice directly blunt β-cell insulin secretion (Fig. 5F, J, and K) but simultaneously enhance systemic glucose uptake (Fig. 5E). This would classically be explained by increased insulin sensitivity in metabolic tissues. We assessed insulin-stimulated pAKT levels in the liver, a large glucose sink and a primary site of adipocyte sEV uptake. In keeping with the endogenous phenotype of ad-cav1KO mice, HFD-fed control mice injected with AT-sEV^adKO displayed significantly reduced liver insulin sensitivity (pAKT) compared with mice injected with vehicle (Fig. 5L). Under the same conditions, injection with AT-sEVs from control mice resulted in the same insulin desensitizing effect in the liver (Fig. 5L). These data suggest that AT-sEVs from mice fed an HFD induced insulin resistance in major metabolic organs, like the liver, regardless of cav1 levels in the sourced AT; however, only sEVs from ad-cav1KO mice improve systemic glucose clearance. The glucose uptake phenotype imparted on the system by AT-sEVs is likely the result of the action of the sEVs on the AT itself, as the sWAT and eWAT are targets of intravenously injected AT-sEVs, and WAT depots were the only organs that contributed to enhanced glucose disposal in the ad-cav1KO mice (Fig. 4L, Supplementary Figs. 5F and 7). We wanted to determine whether the enhanced glycolysis phenotype could be transferred between adipose depots by sEVs. We found that lactate levels were higher in eWAT from mice injected with sWAT-derived AT-sEV^adKO compared with PBS- or sWAT AT-sEV^contr–injected mice (Fig. 5M). This suggests that the phenotype can be transferred by sEVs between AT depots. We confirmed that AT-sEV^adKO were not phenocopying the ad-cav1KO mice by affecting cav1 levels through transfer of cre protein. Cav1 protein was the same in the eWAT of mice injected with PBS or AT-sEV^adKO (Fig. 5N).

To determine how AT-sEVs from ad-cav1KO mice are signaling in the liver, we did an RNA sequencing experiment comparing transcriptional changes in the liver after weekly injections of AT-sEV^contr or AT-sEV^adKO during 4 weeks of HFD feeding. These data were compared with the RNA sequencing results from the sWAT of control and ad-cav1KO mice. Interestingly, the most strongly upregulated pathways in the AT with cav1KO were also significantly upregulated in the liver in mice treated with AT-sEV^adKO compared with those treated with AT-sEV^contr (Fig. 6A). This included multiple inflammatory processes and collagen-synthesizing pathways (Fig. 6A), suggesting that ATs-EVs from the ad-cav1KO mice relay the proinflammatory and fibrogenic environment from the AT to the liver. We confirmed that AT-sEV^adKO provokes inflammation in the liver by demonstrating that the livers of mice injected with AT-sEV^adKO contained more macrophages than those from mice injected with control AT-sEVs (Fig. 6C). As shown in Supplementary Fig. 1, the most highly downregulated pathways in the AT of ad-cav1KO mice are mitochondria-associated pathways. We did not see this same effect in the livers of control mice injected with AT-sEV^adKO. However, we did find pathways specific to the catabolism of lipids and lipid homeostasis to be significantly downregulated in the sWAT of ad-cav1KO mice and the liver of control mice injected with AT-sEV^adKO compared with their respective controls. Together, these data suggest that the systemic metabolic phenotype of ad-cav1KO mice may, at least in part, be generated by communication of the dysfunctional AT with other organs like the liver and pancreas through EVs.

The major goal of this study was to determine the effect of adipocyte cav1 on AT health and whole-body metabolism. Previous studies have demonstrated that whole-body cav1 KO mice on an HFD have significantly impaired triglyceride clearance and a blunted glucose-lowering response to insulin (11,14) (Supplementary Fig. 5). In addition, a more detailed analysis of the metabolic dysfunction of whole-body cav1KO mice revealed a complex phenotype in which altered mitochondrial function in AT is compensated for by the liver (12). Adipocyte-specific cav1 loss of function also resulted in impaired glucose clearance on a chow diet, but an HFD stimulated improved glucose clearance compared with control mice (Supplementary Fig. 4A and Fig. 4A). At the level of the AT, whole-body and adipocyte-specific cav1KO mice were comparable, presenting with reduced mass, impaired mitochondrial function, reduced insulin-stimulated pAKT, and increased fibrosis and inflammation (11–14) (Figs. 2 and 4G and H). Cav1 is not required for adipogenesis (28); therefore, in both mouse models of cav1 depletion, adipocytes form but display dysfunctional characteristics like reduced insulin signaling (3,14). These dysfunctional adipocytes likely lead to the profibrotic and proinflammatory environment in AT. So, rather than adipocyte-autonomous differences between the whole-body and adipocyte-specific cav1KO mice, our data suggest that it is the communication between adipocytes and cells in other tissues that account for the divergent whole-body metabolic phenotypes between the two mouse models. This is further evidenced by the fact that the whole-body cav1-null mice display a significant loss of insulin receptor protein expression and reduced insulin-stimulated pAKT in AT only, not in muscle or liver (14). In contrast, the ad-cav1KO mice showed a striking loss of AKT activation by insulin in both the liver and AT (Fig. 4G and H). Furthermore, it is interesting to note that muscle is not a strong target for AT-sEVs (Supplementary Fig. 7) and was the only tissue tested in ad-cav1KO mice that retained the ability to activate AKT during insulin stimulation (Fig. 4G and H). Thus, whole-body cav1 ablation leads to metabolic alterations and adaptations that partly overrule the effect of adipocyte-specific cav1 deficiency, such as the EV-mediated communication between adipocytes and the liver.

Adipokines and EVs are major means of intercellular signaling from the adipocyte. Our previous work demonstrated that the AT endothelial cells release cav1-containing sEVs that are taken up by adipocytes to the extent that the adipocyte cav1 pool can be reconstituted by >50% in the ad-cav1KO mouse (18). Others have reported that cav1 regulates EV cargo selection and EV production and uptake (29). Because of this and the diversity in sEV cargo, we pursued the potential involvement of adipocyte sEVs in the regulation of systemic metabolism in ad-cav1KO mice. We found that the total level of sEVs was increased in the sWAT and serum of ad-cav1KO mice compared with controls (Fig. 5A and B). This could be due to dysregulation of sEV production by the loss of cav1. For example, cav1 ablation in fibroblasts results in increased EV secretion due to cholesterol accumulation in the sEV production compartment, the multivesicular body (30). Alternatively, sEV production in ad-cav1KO AT could be stimulated by adipocyte mitochondrial dysfunction (26). Injection of isolated sEVs from the sWAT of ad-cav1KO mice into control mice resulted in a similar metabolic phenotype than that of the original ad-cav1KO mice: augmented glucose clearance and reduced insulin secretion (Fig. 5E and F). We found that fluorescently labeled adipocyte sEVs target the pancreas as well as the AT and liver when injected intravenously into control mice (Fig. 5J and Supplementary Fig. 7). This suggests that these metabolic tissues are major sites of adipocyte sEV action, and the resulting alterations in signaling could account for the phenotype produced by injections of AT-sEVs from ad-cav1KO mice. Consistent with this model, we found that sEVs from ad-cav1KO mice directly inhibit GSIS in MIN6 cells, whereas sEVs from control AT enhanced insulin secretion (Fig. 5K). Similarly, a recent study reported that EVs from healthy adipocytes enhanced β-cell insulin secretion by increasing the survival and proliferation of β-cells (31). In contrast, EVs from inflamed adipocytes induced β-cell death and dysfunction, leading to reduced insulin secretion. Therefore, AT-sEVs from ad-cav1KO mice potentially propagate the proinflammatory signal from the AT to β-cells and other cells that take up these EVs. We tested this by doing an RNA sequencing experiment on the liver from control mice on an HFD and injected with AT-sEVs derived from ad-cav1KO mice (AT-sEV^adKO) and compared these results to AT of ad-cav1KO mice. Interestingly, the top upregulated pathways in ad-cav1KO AT were also the most significantly upregulated pathways in control livers after AT-sEV^adKO injection (Fig. 6A). These included pathways involved in both innate and adaptive immunity and collagen synthesis. The AT-sEVs can also transmit the enhanced glycolysis phenotype to other AT depots. sWAT-derived AT-sEV^adKO, when injected into control mice, can target the eWAT and stimulate higher lactate production compared with those injected with PBS or control AT-sEV (Fig. 5M). Overall, these data suggest that AT-sEV^adKO can phenocopy the ad-cav1KO mice by relaying signals from the AT to other organs. In conclusion, ad-cav1 is essential for AT functionality and is involved in whole-body metabolic regulation through EV-mediated actions on other organs.

Acknowledgments. The authors thank the UT Southwestern Metabolic Phenotyping Core for help and Charlotte E. Lee of UT Southwestern for assistance in embedding and processing of histological samples and the UT Southwestern Electron Microscopy Core for help in sample processing for electron microscopy. The authors also thank Shimadzu Scientific Instruments for the collaborative efforts in mass spectrometry technology resources.

Funding. This study was supported by National Institutes of Health (NIH) grants R01-DK55758, R01-DK127274, R01-DK099110, R01-DK131537, and RC2-DK118620 (to P.E.S.). C.C. is supported by K99-DK122019 and R00-DK122019. C.M.G. is supported by NIH grant F32-DK122623. D.Y.O. is supported by NIH grant R01-DK108773. I.W.A. is supported by Swedish Research Council grants 2013-07107, 2017-00792, and 2020-01463; Swedish Diabetes Foundation grant DIA2019-419; Novo Nordisk Foundation grant NNF19OC0056601, and Diabetes Research & Wellness Foundation grant 2334. O.V. and J.J.R. are supported by NIH grants P51-OD01192 for operation of the Oregon National Primate Research Center and 1S10-OD025002-01.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. C.C. conducted all experiments except for those listed below. C.C. and P.E.S designed the studies, analyzed and interpreted data, and wrote the manuscript. S.C. conducted islet isolations and the in vitro GSIS assay with islets. S.C. and X.X.Y. performed the radioactive glucose uptake experiments. D.B. conducted histological staining. C.M.G. and N.J. assisted in mouse experiments, tissue harvesting, and processing. I.W.A. and D.Y.O. contributed valuable resources and key insight. C.O.d.S. performed the experiments with MIN6 cells. J.-B.F. helped with electron microscopy experiments. O.V. and J.J.R. conducted the histological image analysis. R.G. performed the TCA intermediate measurements and analysis and, along with the metabolic core, conducted tissue triglyceride measurements, metabolic cage measurements, and the serum ketone assay. P.E.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.