Loss of cAMP Signaling in CD11c Immune Cells Protects Against Diet-Induced Obesity Free

Obesity is associated with a state of chronic, low-grade inflammation that contributes to insulin resistance and susceptibility to type 2 diabetes in rodents and humans (1). Recent studies have revealed that inflammation and metabolism are highly integrated. Diet-induced obesity (DIO) is characterized by the recruitment of immune cells, including CD11c⁺ macrophages and dendritic cells (DCs), into adipose tissue (2). Tissue-resident immune cells in a lean animal generally have an anti-inflammatory phenotype, and DIO provokes trafficking of CD11c⁺ monocyte-derived macrophages and DCs into adipose tissue (3). These recruited CD11c⁺ immune cells are important factors in the pathogenesis of metabolic disease because their acute ablation attenuates adipose tissue inflammation and improves glucose tolerance (4).

The G-protein Gαs (encoded by Gnas) regulates production of cAMP and mediates the sympathetic nervous system (SNS) effects on brown adipose tissue (BAT) to stimulate thermogenesis and promotes lipolysis on white adipose tissue (WAT) (5). Denervation of adipose tissue reduces both lipolysis and thermogenesis (6,7). Norepinephrine (NE) is the major effector of the SNS and stimulates adaptive thermogenesis and lipolysis through β-adrenergic receptors (β-ARs) (8). Mice lacking all β-ARs develop massive obesity due to a defect in stimulated lipolysis and a deficiency in diet-induced thermogenesis (9). Therefore, activation of BAT to increase energy expenditure via uncoupled thermogenesis is an attractive antiobesity strategy (10). Recently, brown adipocyte-like cells, also referred to as beige cells, have been described within WAT under thermogenic conditions and display similar physiological functions of brown adipocytes (11).

Here, we demonstrate that disruption of Gαs signaling in CD11c⁺ innate immune cells improved glucose metabolism in mice by promoting energy expenditure and inducing adipose beiging, consequently preventing obesity and insulin resistance. Notably, we identified an important role for Gαs in CD11c⁺ innate immune cells to regulate adipose tissue catecholamine levels that likely contribute to the lean phenotype that constitutes a potential target for the treatment of obesity.

Full, detailed methods are given in the online Supplementary Methods.

Gnas^ΔCD11c knockout [Gnas^{tm5.1Lsw/tm5.1Lsw}/Tg(Itgax-cre)1–1Reiz] mice (hereafter, KO mice) were generated from Gnas^fl/fl (Gnas^{tm5.1Lsw/tm5.1Lsw}) mice (hereafter, Flox mice), as previously described (12). Rag.ΔCD11c mice were generated by crossing Rag1^−/− mice with Gnas^ΔCD11c mice. Mice were housed inside a clean specific pathogen–free barrier facility. Under these conditions, mice up to 6 months of age are atopic with elevated IgE levels, but in the absence of an immunological challenge, their lungs and airways do not show any histological changes that would indicate an allergic response, lung CD4 and CD8 T-cell numbers are not altered, and cytokines are unchanged (12). Metabolic studies were performed in a Comprehensive Laboratory Animal Monitoring System. Glucose, insulin, and lipid tolerance tests were performed as published previously (13). For bone marrow transplantation, wild-type (WT) mice were lethally irradiated and then reconstituted intravenously with 6 × 10⁶ bone marrow cells harvested from age- and sex-matched KO or Flox donor mice on the day after the irradiation. All the recipients received sulfamethoxazole and trimethoprim in drinking water for 2 weeks after irradiation. All procedures were approved by the Institutional Animal Care and Use Committee of the University of California San Diego, School of Medicine.

Epididymal fat pads were digested with 1 mg/mL collagenase type II supplemented with 1% fetal BSA for 30 min at 37°C with shaking. Cell suspensions were filtered through a 100-µm filter and centrifuged at 2,000g for 10 min. The pellet containing the stromal vascular cells (SVCs) was then incubated with red blood cell lysis buffer and recentrifuged. The SVCs were resuspended in PBS with 2% FBS, incubated with fluorescently labeled primary antibodies for 30 min on ice and then analyzed on a Cytoflex flow cytometer.

NE, epinephrine (Epi), and dopamine (DA) were measured on a Waters ultra performance liquid chromatography (UPLC) system after alumina extraction, followed by separation and analysis with high-performance liquid chromatography. Dihydroxybenzylamine was used as an internal standard in each tissue sample. The data were analyzed with Empower software.

Western blots and quantitative PCR assays were perfomed using standard procedures. Primers and antibodies are listed in supplementary tables. Hormones and cytokines were measured using commercial assays, as listed in supplementary tables.

Data were analyzed using Prism software and are presented as mean ± SEM or mean ± SD. Statistical significance was determined using the unpaired two-tailed Student t test or one- or two-way ANOVA followed by Tukey post-tests for multiple variables. A P value of <0.05 was considered significant. Calorimetric data were analyzed by general linear regression models using body mass or activity as covariates using the CalR (14).

All data are available upon request from the corresponding author. The mice generated in this study are available from the University of California San Diego under a universal material transfer agreement.

Mice with deletion of Gnas and cAMP signaling in CD11c⁺ cells (KO) and control mice (Flox) have been described previously (12). In addition to the published immune phenotype, these mice have an unexpected metabolic phenotype. The body weight (BW) of both male and female KO mice on a normal chow diet (NCD) was significantly lower than that of control littermate Flox mice (Fig. 1A). At 5 months, male KO mice showed improved glucose tolerance (Fig. 1B). An insulin tolerance test (ITT) using a standard insulin dose (0.6 units/kg) caused severe hypoglycemia in the KO mice and the mice had to be rescued with a bolus of glucose (Supplementary Fig. 1A). Repeating the ITT with a lower dose of insulin (0.4 units/kg) showed that KO mice have improved insulin sensitivity (Fig. 1C), had lower fasting insulin and glucose levels, and, consequently, lower HOMA of insulin resistance (HOMA-IR) compared with control mice (Fig. 1D). Adipose tissue weights were lower in the KO mice. (Fig. 1E). Liver, BAT, epididymal WAT (eWAT), or inguinal WAT (iWAT) histologies were indistinguishable between genotypes (Fig. 1F), and adipocyte size distribution was the same (Supplementary Fig. 1B).

In metabolic cages, BW-matched KO mice consumed the same amount of food during dark and light phases but they ate a slightly lower cumulative amount of food (Fig. 1G). Respiratory exchange ratio (RER), water intake, Vo₂, and energy expenditure (EE) showed the expected circadian changes between dark and light phases, but there were no between genotypes (Fig. 1H–K, raw data shown in Supplementary Fig. 1C–E). Cumulative water consumption was increased in the KO, but EE was unchanged (Supplementary Fig. 1F and G). Covariate analysis of EE indicated significant body mass and genotype effects with lower EE for a given body mass in KO mice (Fig. 1K). Similar results were found for Vo₂ and Vco₂ (Supplementary Fig. 1H). Despite lower EE, the cage temperatures were elevated in cages containing the KO mice during both the light and dark phases (Fig. 1L). Ambulatory locomotor activity was slightly elevated in the KO mice in the dark phase (Fig. 1M) and showed the expected circadian changes (Supplementary Fig. 1I). These results indicated that KO mice had a lean phenotype, which was associated with increased glucose and insulin tolerance, decreased food consumption, increased activity, and increased heat production in the face of reduced oxidative energy expenditure, culminating in decreased energy balance compared with Flox mice (Fig. 1N).

We then challenged the mice with a high-fat diet (HFD). Both male and female KO mice showed resistance to HFD-induced BW gain (Fig. 2A). Flox mice fed the HFD showed the expected whitening of the BAT (Fig. 2B), with adipocyte enlargement in eWAT (Fig. 2C) and iWAT (Supplementary Fig. 1K) and hepatic steatosis, but the KO mice did not have any of these changes. The KO mice had reduced adipose tissue by DEXA (Fig. 2D and Supplementary Fig. 1L), but bone density was unchanged (Supplementary Fig. 1M). Tissue weights of iWAT, eWAT, BAT, and liver were significantly lower in the KO mice after euthanasia (Fig. 2E). Consistent with their lean phenotype, KO mice had improved glucose tolerance in both males and females (Fig. 2F), as well as improved insulin sensitivity (Fig. 2G). A lipid tolerance test showed no alteration in lipid absorption (Fig. 2H). The temperature was elevated in metabolic cages containing KO mice, particularly during the light phase, as we observed before (Supplementary Fig. 1N). Cumulative food and water intakes were unchanged (Supplementary Fig. 1O and P). Diurnal changes in RER were suppressed in both genotypes due to the high fat content of the diet (Supplementary Fig. 1Q). Vo_2, EE, and Vco₂ values were similar between genotypes (Supplementary Fig. 1R–T), and covariate analysis did not indicate a significant body mass or genotype effect for mice on the HFD. Unlike mice fed the NCD, there was no difference in physical activity of energy balance in mice fed the HFD (Supplementary Fig. 1U and V).

Flow cytometric analysis of the SVCs from eWAT (Supplementary Fig. 2A) indicated reduced cellular content per gram of fat in the KO eWAT and reduced CD45⁺ cells as a percentage of SVCs (Fig. 2I), reduced adipose tissue macrophages (ATMs) with decreased M1 and increased M2 (Fig. 2J), and decreased adipose tissue DCs (ATDCs) with decreased conventional type 1 DCs (cDC1) and increased conventional type 2 DCs (cDC2) (Fig. 2K) but increased eosinophils (Fig. 2L). Analysis of macrophage markers by quantitative (QPCR) indicated an increase in Adgre1 (EMR1) and Itgax (CD11c) in eWAT from WT mice fed an HFD that was not seen in KO mice, but an increase in Itgam (CD11b) in KO mice consistent with the reduced M1 to M2 ratio (Fig. 2M). Comparing the overall populations, the KO mice had a reduced proportion of CD45⁺ cells with decreased proportions of macrophages and DCs (Fig. 2N). A similar analysis of iWAT showed decreased numbers of ATMs, ATDCs, and eosinophils but increased numbers of neutrophils in the KO mice (Supplementary Fig. 2B) and the same M1→M2 and cDC1→cDC2 phenotype switch (Supplementary Fig. 2C). Unlike the eWAT, nonmyeloid immune cells composed the greatest proportion of cells in iWAT (Supplementary Fig. 2D), which likely represents B and T cells from the inguinal lymph node within the inguinal fat pad.

The difference in metabolic phenotype between NCD and HFD was not caused by the composition of the diet. Metabolic profiling during a diet-switch protocol showed that switching lean mice to the HFD alters circadian rhythms in oxidative metabolism, RER, and EE, and increased Vo₂ and EE in both genotypes; however, differences in cage temperatures were still observed (Supplementary Fig. 3).

CD11c is expressed on multiple hematopoietic cells but is also expressed on microglia in the brain (3). To assess the role of hematopoietic cells in the lean phenotype, we generated chimeric mice by transplanting bone marrow from KO or Flox donor mice (CD45.2) into lethally-irradiated, WT C57BL/6J recipient mice (CD45.1), which were then fed the HFD (Fig. 3A). Mice receiving KO donor cells started losing weight 3 weeks after bone marrow transplant (BMT) (Fig. 3B), despite greater weekly food intake (Fig. 3C). iWAT and eWAT were depleted when mice were sacrificed 8 weeks after BMT, but the liver and spleen were normal (Fig. 3D).

We assessed engraftment by measuring the percentage of donor CD45.2 cells in peripheral blood mononuclear cells (Supplementary Fig. 4A). Engraftment was 85–90% for both genotypes (Fig. 3E). We also assessed tissue engraftment of donor cells in eWAT (Supplementary Fig. 4B). Most (70–80%) of SVCs were derived from the donor mice (Fig. 3F). The mice had significantly more CD4 T cells and fewer ATMs when the bone marrow was transplanted from the KO mice (Fig. 3G). Engraftment into iWAT was similar (Supplementary Fig. 4C).

We then repeated the BMT but used recipient mice that had been made obese by 8 weeks of HFD feeding prior to the BMT to test whether the KO bone marrow could reverse the obese phenotype. Mice receiving the KO bone marrow started to lose weight 7 weeks after BMT (Fig. 3K) despite no change in food intake (Fig. 3L). The Flox and KO chimeric mice did not show a difference in glucose tolerance at 7 weeks (Fig. 3M), but the KO chimeric mice had improved insulin sensitivity and glucose tolerance at weeks 8 and 9 after BMT compared with the Flox chimeric mice (Fig. 3L and M), and this was still evident at week 11 (Supplementary Fig. 4D). These results suggested that lack of Gαs in bone marrow–derived immune cells is sufficient for metabolic improvement and resistance to DIO.

To determine whether T or B cells played a role in the metabolic improvement, we generated Rag1^−/−/Gnas^ΔCD11c double-KO mice (Rag^ΔCD11c). The resulting Rag^ΔCD11c mice were resistant to DIO when fed an HFD (Fig. 3N) and consequently had reduced iWAT and eWAT (Fig. 3O). The Rag^ΔCD11c mice had decreased lipid accumulation in BAT and smaller adipocytes in eWAT and iWAT (Supplementary Fig. 5A). The mice also had improved glucose tolerance and insulin sensitivity (Fig. 3P and Q) and decreased fasting blood glucose levels (Fig. 3R), insulin levels (Fig. 3S), and HOMA-IR (Fig. 3T) relative to Rag^fl/fl controls. Metabolic studies of the Rag^ΔCD11c mice fed an HFD showed reduced food intake, EE, and Vo₂ and Vco₂, relative to the Rag^fl/fl control mice (Supplementary Fig. 5B–D). Interestingly covariate analysis of the data against locomotor activity, but not body mass, showed that the Rag^ΔCD11c mice expended less energy and consumed fewer calories for the equivalent activity (Supplementary Fig. 5E). These results suggested that adaptive immunity was not required for the lean phenotype. Deletion of Gnas in macrophages using LysM-Cre did not confer a lean phenotype (Supplementary Fig. 6A–C) nor did altering the gut microbiota by cohousing or fecal transfer (Supplementary Fig. 6D–F).

As the cage temperature was increased in metabolic cages containing KO mice, we investigated body temperature of Flox and KO mice by using infrared imaging. We observed increased temperature in the KO mice as compared with Flox mice, especially those fed an HFD (Fig. 4A). Because KO mice have reduced adipose tissue, we tested whether the KO mice could adapt to a cold challenge. The Flox mice were able to maintain their body temperature, but the body temperature of KO mice dropped by almost 5°C after 4.5 h (Fig. 4B). Body temperature quickly recovered when mice were removed from the cold room. This suggested either a defective adrenergic response to cold or defective thermogenesis. To assess the thermogenic capacity of various tissues, we performed an oxygen flux analysis using an Oroboros system. Isolated soleus muscle, an oxidative slow-twitch muscle, showed increased basal oxidative phosphorylation, increased maximal mitochondrial complex I activity, increased complex I+II activity, and increased maximal uncoupled activity (Fig. 4C). The glycolytic fast-twitch gastrocnemius muscle did not show these alterations (Supplementary Fig. 7A) nor did BAT or liver (Supplementary Fig. 7B and C), although liver showed greater uncoupled complex II activity. Oxygen flux measurements in eWAT and iWAT, however, were indistinguishable from baseline noise. To determine if the immune cells could contribute to the change in oxidative metabolism, we analyzed splenic CD11c⁺ cells on a Seahorse instrument. Cells from KO mice had elevated basal respiration that was primarily driven by an increase in ATP-linked respiration (Fig. 4D) rather than an increase in proton leak.

To assess the thermogenic potential of WAT, we measured UCP1 expression by Western blot. UCP1 protein was increased in eWAT and iWAT but was reduced in BAT (Fig. 4E), so we then measured the expression of brown- and beige-associated genes. Ucp1 mRNA was reduced in BAT, consistent with the protein data (Fig. 4F), as was expression of Oplah, Pdk4, and Serca2b, but it was highly elevated in eWAT, as was Pgc1a, Cidea, and Pparg expression (Fig. 4G). Changes in iWAT were less pronounced (Fig. 4H). Expression of adipose triglyceride lipase (ATGL/PNPLA2), the major lipase involved in basal lipolysis (15), and phosphorylation of hormone sensitive lipase (HSL/LIPE) on Ser563, the major lipase involved in cAMP-stimulated lipolysis, were increased in both iWAT and eWAT from mice fed an HFD (Fig. 4I). Phosphorylation of cAMP-dependent protein kinase (PKA) substrates was also increased in KO iWAT and eWAT (Supplementary Fig. 8A). In contrast, however, phosphorylation of p70S6Kinase on Thr389, which is repressed by PKA, was reduced in the KO WAT (Fig. 4I). We did not see phosphorylation of HSL or induction of UCP1 or ATGL in mice of either genotype that were fed the NCD (Supplementary Fig. 8B). These results indicate that cAMP signaling is selectively increased in WAT of KO mice fed the HFD.

To test for a circulating factor that could increase cellular cAMP in adipocytes, we treated 3T3-L1 adipocytes with serum from either KO or Flox mice and found that both could increase intracellular cAMP levels (Fig. 5A). Treatment of 3T3-L1 adipocytes with serum from either Flox or KO mice fed an NCD activated HSL(Ser563), perilipin(Ser522), and CREB(Ser133) phosphorylation (Fig. 5B). Similar results were observed with serum from HFD-fed mice (Fig. 5C). The increase of cAMP and PKA signaling was blunted when the cells were pretreated with propranolol, suggesting activation of β-ARs. We then tested whether the presence of immune cells could influence the lipolysis response of 3T3-L1 adipocytes to the selective β3-adrenergic receptor agonist CL-316,243 (CL). CD11c bone marrow–derived macrophages (BMDMs) from the KO stimulated greater glycerol release from 3T3-L1 adipocytes in response to CL (1 μmol/L for 3 h) when cocultured in a Transwell or in contact culture compared with cells from Flox mice (Fig. 5D). Similar results were obtained with splenocytes from KO mice (Supplementary Fig. 9A). The communication in the coculture experiments was bidirectional as 3T3-L1 adipocytes altered cytokine expression in BMDM in the immune cells. BMDM from Flox mice had elevated levels of Il1b and Il6 cytokine expression, which was further stimulated by the 3T3-L1 adipocytes, whereas BMDMs from KO mice had much lower cytokine expression and were refractory to 3T3-L1 stimulation (Fig. 5E).

The inhibition of serum-stimulated cAMP signaling by propranolol suggested β-adrenergic signaling, so we measured NE, Epi, and DA levels in plasma from Flox and KO mice. Plasma levels of NE and DA were not altered between KO and Flox mice (Fig. 5G), nor was plasma Epi level (Supplementary Fig. 9B). In contrast, tissue NE and DA levels were significantly higher in eWAT of KO mice under both NCD and HFD feeding (Fig. 5H). Similar NE increases were found in iWAT (Supplementary Fig. 9C), but DA was not detected in the majority of iWAT samples. Epi was not detected in either eWAT or iWAT, consistent with the sympathetic origin of the NE. Adrenal catecholamine levels did not change between genotypes or diets (Supplementary Fig. 9D).

To test WAT sensitivity to adrenergic stimulation, we challenged mice fed either the NCD or HFD with CL-316,243 and measured glycerol levels as a surrogate for lipolysis. Because the KO mice have much less WAT (Fig. 2D), we normalized the glycerol levels to the adipose tissue weight. Basal glycerol levels were increased in KO mice compared with Flox mice fed the NCD and were not suppressed by the HFD (Fig. 5I). Stimulated glycerol release was not different between genotypes but was reduced by the HFD. Raw glycerol levels were lower in the KO mice, reflecting the depleted adipose tissue lipids (Supplementary Fig. 9E). These results would indicate that a given amount of adipose tissue in KO mice is more lipolytic than the equivalent in Flox mice, and that the major genotype difference is seen in basal glycerol levels, consistent with the elevated ATGL (Fig. 4I). Because the stimulated glycerol levels were lower in mice fed the HFD, we measured β-AR expression. The adipocyte-specific β-3AR (Adrb3) was decreased dramatically in both eWAT and iWAT of Flox mice fed the HFD (Fig. 5J,). Adrb3 was partially reduced in the KO mice fed the NCD but did not change upon HFD feeding in either depot (Fig. 5J). In contrast, there were no changes in the expression levels of β-1AR (Adrb1) and β-2AR (Adrb2) between the groups or diets (Supplementary Fig. 9F).

To explore the cause of elevated catecholamine levels in the KO WAT, we measured expression of catecholamine transporters and enzymes of catecholamine synthesis and degradation. Expression of the organic cation transporter 3 (Slc22a3; OCT3) was decreased in the eWAT of KO mice fed the NCD, and the NE transporter (Slc6a2; NET) was decreased in the eWAT of KO mice fed the HFD (Fig. 5K). Expression of Slc22a3 did not change significantly in iWAT (Supplementary Fig. 9G); however, Slc6a2 was not detected. The expression of monoamine oxidase A (Maoa) was significantly increased in both eWAT and iWAT from Flox mice fed an HFD, and this increase was blunted in the KO mice. Expression of catecholamine-O-methyl transferase (Comt) was significantly lower in eWAT from KO mice but did not change in iWAT (Fig. 5L and Supplementary Fig. 9H). Expression of tyrosine hydroxylase (Th), the rate-limiting enzyme for catecholamine synthesis, was not detected in the WAT of Flox and KO mice and did not differ in the adrenal glands of KO and Flox mice (Supplementary Fig. 9I). Collectively, these data suggest that Gαs deficiency in CD11c⁺ innate immune cells elevated tissue NE but prevented Adrb3 downregulation, allowing continued adrenergic stimulation of the adipose tissue.

Obesity is associated with immune cell infiltration and inflammation of adipose tissue. We observed fewer macrophage crown-like structures in the eWAT of KO mice compared with Flox mice fed the HFD, consistent with reduced inflammation (Fig. 6A). This was not due to a migratory defect, because BMDM migration in vitro was slightly increased in KO mice in response to 3T3-L1 adipocyte conditioned medium, but the response to serum was unchanged (Fig. 6B and Supplementary Fig. 10A). Migration of fluorescently labeled (PKH26) CD11c⁺ BMDM into adipose tissue in vivo by flow cytometry (Supplementary Fig. 10B) showed that BMDMs from KO mice migrated to WAT in obese WT recipient mice to the same extent as cells from Flox mice (Fig. 6C). We then assessed chemokine and cytokine expression. In Flox mice, the HFD increased the expression of Ccl2 and Il6 that was markedly blunted in the KO mice (Fig. 6D). Plasma CCL2 was unchanged, however, and IL-6 was slightly elevated in KO mice fed the HFD (Fig. 6E). In contrast, Il10 was increased in KO mice in either NCD or HFD groups, and IL-10 level was significantly higher (200 pg/mL; ∼50-fold) in the KO mice, especially those fed the HFD (Fig. 6E). No significant differences in the mRNA and circulating levels of TNFα or IL-1β were observed between the groups (Supplementary Fig. 10C).

Cytokine induction is mediated partly by increased expression and activation of the inflammasome, especially NLRP3. HFD induced NLRP3 protein expression and processing of pro-IL1β in eWAT in Flox mice, but this was absent in the KO mice (Fig. 6F and Supplementary Fig. 11A). Similarly, the inflammasome components Caspase-1 and adaptor protein apoptosis-associated speck-like protein containing a caspase-recruitment domain (Pycard) were induced by the HFD in Flox mice but not KO mice (Fig. 6G). Activation of the NLRP3 inflammasome by cAMP and lipopolysaccharide (LPS) was then tested in BMDMs as an in vitro model system. The induction of Nlrp3, Il1b, and Il6 mRNAs by LPS was suppressed in the BMDMs from KO mice (Fig. 6H). We also detected decreased protein expression and secretion in the supernatants of KO BMDMs (Fig. 6I and J). As expected, Gnas expression was greatly reduced in KO BMDMs, but TLR, MyD88, and Irf1–4 expression was unchanged, and Trif, Irf7, and Klf4 expression was slightly elevated (Supplementary Fig. 11B–D). LPS induction of p38MAPK, JNK, and ERK was preserved (Supplementary Fig. 11E), but degradation of Iκb and phosphorylation of p65 RelA was impaired in KO BMDMs (Supplementary Fig. 11F), consistent with suppressed inflammatory signaling. Mechanistically, expression of PPARγ, a known repressor of NLRP3 (16,17), was much higher in KO BMDMs and decreased with LPS (Fig. 6I).

Taking a pharmacological approach, BMDMs from Flox and KO mice were treated with the cell-permeable cAMP analog 8-(4-chlorophenylthio) adenosine 3′,5′-cyclic monophosphate (CPT) or formoterol, a long-acting β2 agonist (LABA). CPT induced NLRP3 and pro-IL1β expression in both Flox and KO BMDMs, but LABA only worked in Flox BMDMs due to the lack of Gαs in the KO BMDMs (Fig. 6K). PPARγ expression was suppressed by CPT in BMDMs from both genotypes, but LABA, again, had no effect in KO BMDMs (Fig. 6K). At the mRNA level, CPT induced Crem, Il1b, and Nlrp3 and repressed Pparg in BMDMs from both genotypes, but LABA had no effect in the KO BMDMs (Fig. 6L and M). Similar effects were observed with dibutyryl-cAMP, isoproterenol, and forskolin (Supplementary Fig. 11G). These data indicate that inflammatory signaling requires cAMP signaling in addition to TLR activation.

We report in this article the reciprocal regulation of cAMP signaling in CD11c⁺ innate immune cells and adipocytes. Suppression of cAMP generation in CD11c⁺ cells caused an anti-inflammatory phenotype in adipose tissue with, paradoxically, increased cAMP signaling in adipocytes, increased adipose tissue NE, and increased lipolysis and beiging. The increased NE was accompanied by decreased expression of catecholamine transporters and metabolizing enzymes, which we interpret to indicate that reduced tissue catecholamine clearance in immune cells was the primary cause of the elevated NE. The cumulative effects of these changes were to create a lean, obesity-resistant phenotype (Fig. 7). The effect was mediated by bone marrow–derived cells and did not require adaptive immunity. Although our results support the importance of cAMP signaling in both the innate immune cell and adipocyte compartments and the interaction between these cell types to coordinate sympathetic regulation of adipose tissue function, we cannot exclude effects outside of the adipose tissue, because CD11c immune cells are found in many other tissues. Indeed, we see increased oxidative metabolism in selected slow-twitch muscles that may contribute to the phenotype.

That selective elevation of Gαs signaling and lipolysis in adipocytes could explain the observed phenotype is supported by evidence that selective chemogenetic activation of Gαs in adipocytes increases energy expenditure, improves glucose metabolism, and decreases fat mass (18), and deletion of Gαi2 in adipocytes similarly protects against DIO (19). Conversely, genetic deletion of Gαs in adipocytes reduces lipolysis and impairs thermogenesis (20). If Gαs is deleted in both adipocytes and immune cells, the phenotype is very different because the KO mice are lean, have severely reduced adipose tissue, impaired glucose tolerance, and reduced thermogenic response to catecholamines or diet (21), suggesting that the loss of Gαs in immune cells is dominant over the effect of Gαs loss in adipocytes.

The connection between sympathetic activity and adipose tissue lipolysis has been studied in both animal models and humans (22–25). Peripheral sympatho-facilitators or direct optogenetic stimulation of adipose tissue sympathetic nerves activate lipolysis and cause fat mass loss (8,26–28). Conversely, adipose tissue denervation eliminates the thermogenic response to a wide variety of stimuli, including fasting and cold exposure (6). Interestingly, complete loss of catecholamine production, however, has no effect on DIO and the mice have normal glucose tolerance (29) suggesting that the ratio of β- to α-adrenergic signaling is important. Many of the changes that we observe can be explained by elevated adrenergic signaling because β-3AR–specific agonists, which activate Gαs signaling in adipocytes, rescue metabolic dysfunction in mice (30–32), cause beiging of WAT (33), and enhance lipolysis and insulin secretion (34), and β-3AR KO mice are more susceptible to DIO (35,36). Similarly, β-2AR–specific antagonists that would inhibit Gαs signaling in immune cells improve glucose and insulin tolerance without affecting BW (37).

The human data are less clear but generally agree with the animal data (38,39). Humans have a different distribution of BAT than rodents (40), but BAT activation has been proposed as an antiobesity therapy (38,41). Catecholamine reuptake inhibitors cause weight loss by increasing thermogenesis (42), and high urinary catecholamine levels are inversely associated with obesity (43). Conversely, patients taking β-blockers have decreased energy expenditure and increased BW (44). In humans, the β3-agonist mirabegron improves glucose homeostasis, but side effects of β-agonist therapy include an increase in macrophage infiltration into adipose tissue and potential increases in blood pressure (45,46).

Last, the identity of the CD11c⁺ innate immune cells that are responsible for our observed phenotype remains to be determined. Whether a recently identified population of sympathetic nerve–associated CD11c⁺ macrophages (47) are responsible for our lean phenotype is unclear, because deletion of Gαs using LysM-Cre did not have the same phenotype. A limitation, however, is that LysM-Cre is not expressed in all macrophages (48). There is also a potential role for inflammation in the lean phenotype. The most striking change in cytokines in our KO mice was the strong elevation of IL-10 levels, which could potentially reverse obesity and improve glucose and insulin resistance (49,50). The role for IL-10 in obesity is controversial, however, because mice lacking IL-10 have increased thermogenesis (51,52). Results may be different in humans, however, because IL-10 has no effect on human adipocytes, although it has anti-inflammatory effects on immune cells (53).

Acknowledgments. We acknowledge the assistance of Dr. Valeria Estrada and the Biorepository and Tissue Technology Shared Resource at the Moores Cancer Center.

Funding. This work was funded primarily by National Institutes of Health (NIH) grants R01HL141999, U01AI125860, and R01CA196853 to E.R. and N.J.G.W., and, in part, by a VA Merit Review Award I01BX004848 and Senior Research Career Scientist Award IBX005224 to N.J.G.W. The research was also supported by the NIH Cancer Center Support grant P30CA023100, the Diabetes Research Center grant P30DK063491, and the San Diego Digestive Disease Center grant P30DK120515. The ANCOVA analysis of calorimetric data done for this work was provided by the National Institute of Diabetes and Digestive and Kidney Diseases Mouse Metabolic Phenotyping Centers (www.mmpc.org) using their Energy Expenditure Analysis page (https://www.mmpc.org/shared/regression.aspx) and supported by grants DK076169 and DK115255.

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

Author Contributions. E.R. and N.J.G.W. designed and conceived the study. L.Z. performed the experiments with the assistance of S.M.L. L.Z. was supported and mentored by A.T. D.S.H. assisted with animal husbandry. M.D. assisted with the glucose and insulin tolerance tests. S.B. assisted with cytokine measurements. L.E. performed the microbiota experiments. S.M. measured catecholamine levels and substrate uptake assays. M.H., J-W.B., and P.W. assisted with tissue harvest. A.J.A.M. performed the Seahorse measurements, S.D. and H.H.P. performed the Oroboros measurements. O.O. performed the cold-room thermogenesis assay. M.C. assisted with genotyping. L.Z. and N.J.G.W. wrote the manuscript and generated the figures. N.J.G.W. is the guarantor of this work and, as such, had full access to all the data in this study and takes responsibility for the integrity of the data and accuracy of the data analysis.