Fasting-Induced Lipolysis and Hypothalamic Insulin Signaling Are Regulated by Neuronal Glucosylceramide Synthase

The central nervous system (CNS) balances energy intake to energy expenditure, termed body energy homeostasis. Among several other brain regions, the hypothalamic arcuate nucleus (Arc) harbors first-order neurons sensing peripheral energy signals, such as leptin and insulin (1,2). Subsequently, these neurons adapt energy intake to the energy needs of the body by altering their firing pattern and neurotransmitter expression (1–3).

Most hypothalamic neuronal subpopulations possess receptors for the energy signals leptin and insulin. In the CNS, both hormones exert anorexic actions by suppressing food intake (4,5). Many reports (6,7) support the hypothesis that hypothalamic insulin and leptin systems act in a concerted manner in order to regulate feeding and glucose homeostasis, as well as body weight. In line with this, a novel concept of hypothalamic insulin/leptin cross talk demonstrates that leptin receptor signaling regulates mitochondrial function in hypothalamic neurons, which directly influences their insulin sensitivity (8). In the hypothalamic Arc, leptin signaling depolarizes anorexigenic proopiomelanocortin (POMC) neurons (3) and hyperpolarizes Agouti-related peptide (AgRP) neurons (9), thereby regulating energy homeostasis and body weight. In addition, central insulin signaling is an important regulator of peripheral glucose homeostasis and fat tissues (10,11). Signaling pathways for leptin and insulin converge on insulin receptor (IR) substrate/phosphatidylinositol 3 kinase (PI3K) in hypothalamic neurons (6,7). It has, however, also been shown that the effects of leptin and insulin as energy balance signals do not completely overlap on all hypothalamic neuronal subpopulations (12). In POMC neurons, PI3K integrates the synergistic effects of insulin and leptin (12). Insulin also directly induces PI3K in AgRP neurons, whereas leptin-stimulated PI3K activation in AgRP neurons requires additional inhibitory synaptic input (12).

Increased mediobasal hypothalamus (MBH) insulin signaling restrains white adipose tissue (WAT) lipolysis (11). Exaggerated PI3K signaling has further been shown to silence POMC neurons (13). Besides the Arc, the ventromedial hypothalamus has recently emerged as a critical MBH region, where increased insulin signaling is involved in the promotion of diet-induced obesity (14,15). Glucosylceramide synthase (GCS [gene Ugcg])–derived gangliosides are acidic glycosphingolipids that are prominently expressed by neurons (16). They contribute to the formation of membrane microdomains, regulating intracellular signal transduction (17). We have recently demonstrated (18) that adequate function of the hypothalamic leptin receptor (ObR) requires GCS expression.

In our previous study, we used a Ugcg f/f mouse model harboring tamoxifen-inducible Cre activity in forebrain neurons, including regions of the MBH. GCS deletion by tamoxifen injection at the age of 6 weeks resulted in the development of marked obesity because of increased fat mass, hypometabolism, decreased sympathetic nervous activity, hypothermia, and glucose intolerance, as described earlier by us (18).

The sympathetic nervous system (SNS) regulates WAT function (19). Appropriate WAT function is seminal for body weight maintenance and glucose homeostasis. A decade ago, ganglioside GM3, which is dominantly expressed in WAT, was proposed to influence IR signaling (20). GM3 inhibits IR signaling (21) by sequestering IRs from caveolae (22). We hypothesized that GCS deficiency, resulting in the loss of neuronal gangliosides (GM1, GD1a, GD1b, and GT1b), would potentially perturb neuronal insulin signaling in a similar fashion. In the current study, we indeed observed increased phosphorylation of the IR and Akt in the MBH of fasted insulin-stimulated Ugcgf/f//CamKCreERT2 mice. In line with data demonstrating that hypothalamic insulin signaling regulates WAT function (11), fasting-induced lipolysis is impaired in Ugcgf/f//CamKCreERT2 mice. Thus, the current results propose hypothalamic gangliosides as a novel class of CNS regulators for adipose tissue homeostasis.

Ugcgf/f//CamKCreERT2 mice were generated as previously described (16,18,23). Mice were induced with tamoxifen 6 weeks after birth (18). Experiments were performed in overnight (o/n)-fasted females at the age of 3–9 weeks after tamoxifen induction (postinduction [pi]), as stated in the figure legends.

Adult hypothalamic cells were purchased from Cedarlane and were cultured according to the manufacturer's guidelines. The generation of these cell lines has been described earlier (24), and their characteristics and application have been reviewed (25).

Mice were killed and transcardially perfused with 4% paraformaldehyde. Parametrial WAT was collected and stained with hematoxylin-eosin. Cell size was determined with ImageJ (National Institutes of Health).

Blood was withdrawn from the tail vein of mice (fed, fasted, 4 h after refeeding). Serum free fatty acids (FFAs) were determined by enzymatic assays (BIO-CAT).

For MBH analysis, MBH of fasted insulin-injected (5 units/kg i.p., 15 min; Eli Lilly) mice was collected and processed as previously described (18). For WAT analysis, WAT was collected, shock frozen in liquid N₂, and subsequently homogenized in RIPA buffer (12 mmol/L Na-deoxycholate, 1% NP-40 in PBS) containing proteinase inhibitor and phosphatase inhibitor cocktails (Roche). After centrifugation (15,000g, 10 min, 4°C), the intermediate phase containing proteins was carefully collected, excluding the top fat layer and the cell debris pellet. Protein concentrations were determined by the Lowry method.

Adult mHypoA2–12 cells were treated as indicated, lysed, and subjected to immunoprecipitation (rabbit-α-IRβ) (1:50; Santa Cruz Biotechnology [SCBT]) or, for Western blots, lysed as previously described (18). SDS gel electrophoresis and Western blots were performed according to standard procedures. The primary antibodies used were rabbit α-p-HSL563, α-p-HSL660, α-p-HSL565, α-HSL, α-p-Akt Ser473, α-Akt, α-FAS, and α-ACC (1:1,000; Cell Signaling Technology); rabbit-α-IRβ (1:200; SCBT); rabbit-α-p-Tyr (1:500, 4G10; Millipore); and rabbit-α-actin (1:500; SCBT). The secondary antibody used was horseradish peroxidase–conjugated α-rabbit-IgG (1:1,000; Dako). Bands were visualized by chemiluminescence (Amersham) and quantified with ImageJ (National Institutes of Health).

WAT was homogenized in buffer (4 mmol/L EDTA, 1 mmol/L Na₂S₂O₅, 0.01N HCl, pH 7.0; concentration 200 mg WAT/mL buffer) and centrifuged, and the infranatant omitting the top fat layer and debris pellet was used for norepinephrine (NE) extraction. NE was measured by NE alumina extraction with subsequent detection by ELISA (BioVision). The analytical specificity for this ELISA is 100% for NE, 0.14% for adrenaline, and 0.2% for dopamine. The detection limit of the ELISA is specified as 0.1 ng/mL × correction factor C. C is calculated as follows: C = 10 μL (volume of standards extracted)/sample volume (microliters) extracted. The analytical sensitivity (750 µL undiluted sample) for this ELISA is 1.3 pg NE/mL, and the functional sensitivity is 2 pg NE/mL. The coefficient of variation for NE concentrations measured in this study is stated to range from 8.4% to 11.7%, according to the manufacturer’s manual.

The in vitro lipolysis assay was performed as described previously (26). Twenty-five milligrams of parametrial WAT were preincubated for 15 min in DMEM/0.5% fatty acid–free BSA at 37°C. Samples were shaken at 400 rpm. After adding the β3-adrenergic agonist BRL37344 (BRL) (50 ng/mL), samples were collected from the DMEM at regular time points (0, 30, 60, 120 min). DMEM FFA levels were determined by enzymatic assays (BIO-CAT).

mHypoA2–12 cells were treated with Genz123346 (Genz) (27) (5 µmol/L, 7 days). Ganglioside depletion was verified by thin-layer chromatography, as described previously (18).

The 4-h serum-starved mHypoA2–12 cells were stimulated with insulin (10 nmol/L; Eli Lilly) during the indicated time. Proximity ligation assay (PLA; Duolink; Sigma-Aldrich) was performed according to the manufacturer's guidelines, using primary α-IRα antibodies (1:50; Santa Cruz Biotechnology) and α-GD1a antibodies (1:100; Millipore). In case of close proximity between the target epitopes (∼20–30 nm), the PLA probe oligonucleotides can be ligated to form a template for subsequent rolling circle amplification. These amplified sequences are then recognized by fluorescently labeled probes, resulting in fluorescent spots corresponding to the presence of proximity events. PLA spots were visualized by fluorescence microscopy (Keyence), and the number of spots per cell was quantified.

Mice received injections of either saline or insulin (5 units/kg, 15 min). Subsequently, mice were anesthetized and transcardially perfused with 4% paraformaldehyde (PFA). Brains were postfixed in PFA o/n, cryoprotected in 30% sucrose, and subsequently frozen. Cryosections were prepared for both PLA (on glass slides) and immunofluorescence (IF) (free-floating sections), including the hypothalamic Arc, approximately between bregma −1.7 and −2.2. The PLA was performed according to the manufacturer's guidelines and was evaluated as described above. Nuclei were stained with DAPI. PLA spots in Arc neurons were visualized by fluorescence microscopy (Keyence). The number of spots in the vicinity of DAPI-stained neuronal nuclei was quantified. Only spot-positive neurons were included in the quantification, in order to account for the heterogeneity of Arc neurons. For IR quantification, the following antibodies were used: primary α-IRα and α-IRβ (extracellular epitope) (1:50; 4°C o/n in PBS/1% BSA/0.01% Triton-X [PBT]; Santa Cruz Biotechnology). For the detection of IR phosphorylation, the following antibodies were used: primary α-IRβ (intracellular epitope; 1:30; Santa Cruz Biotechnology) and α-p-Tyr (1:50; 4°C o/n in PBT; 4G10; Millipore).

Free-floating sections (40 µm) were prepared and stained with α-IRα (1:50; 4°C o/n in PBT; Santa Cruz Biotechnology). The secondary antibody was donkey-α-rabbit Alexa Fluor 488 (1:200; Life Technologies). Nuclei were stained with DRAQ5. IF was monitored by confocal microscopy (TCS-SL; Leica), and IR fluorescence intensity in reference to background staining was estimated by ImageJ (NIH).

Cells were treated with either saline or insulin (10 nmol/L, for 5 or 10 min) and were immediately washed with ice-cold PBS and fixed in 4% PFA (15 min). IRs and GD1as were stained with α-IRα (1:50; Santa Cruz Biotechnology) and α-GD1a (1:100; Millipore) at 4°C o/n. Secondary antibodies used were donkey-α-rabbit Alexa Fluor 546 and goat-α-mouse Alexa Fluor 488 (1:200; Life Technologies). IF was evaluated by confocal microscopy, and fluorescence overlap (Pearson coefficient) was estimated in double-positive regions of interest in the vicinity of the nuclei by ImageJ (NIH).

Results were analyzed by a two-tailed, unpaired Student t test (Prism; GraphPad Software). P values were marked as follows: *P ≤ 0.05 (considered to be statistically significant), **P ≤ 0.01, and ***P ≤ 0.001.

Forebrain neuron-specific GCS deletion results in ganglioside loss in targeted neurons (Fig. 1A). Gangliosides are located in cellular membranes where they are proposed to modulate neuronal signal transduction (Fig. 1B). In the current study, we used a Ugcg f/f mouse model harboring tamoxifen-inducible Cre activity in forebrain neurons, including neurons in the MBH, as described by us earlier (18). Upon tamoxifen injection, target neurons are devoid of gangliosides within 3 weeks. These mice were shown to develop prominent obesity due to progressively increasing fat mass. Mutant mice have deficient hypothalamic leptin receptor signaling, as well as brown adipose tissue dysfunction (18). Despite normal feeding behavior, Ugcgf/f//CamKCreERT2 mice displayed a decreased metabolic rate and an increased respiratory exchange ratio, indicating preferential use of carbohydrates as fuel. Furthermore, Ugcgf/f//CamKCreERT2 mice displayed hypothermia because of decreased sympathetic nervous outflow to brown adipose tissue (18).

In order to analyze WAT function, we monitored several parameters as well as hypothalamic IR signaling in fasted Ugcgf/f//CamKCreERT2 and control mice.

When metabolically challenged by fasting, Ugcgf/f//CamKCreERT2 mice defend their body weight more efficiently than control mice and lose less weight (Fig. 1C). Furthermore, total parametrial WAT weight is increased in fasted Ugcgf/f//CamKCreERT2 mice (Fig. 1D). While WAT cell size is slightly increased in fed Ugcgf/f//CamKCreERT2 mice (Fig. 1E and F), WAT cells of fasted Ugcgf/f//CamKCreERT2 mice are significantly larger than those of fasted control mice (Fig. 1E and G). These results propose that impaired WAT function may contribute to body weight defense of Ugcgf/f//CamKCreERT2 mice.

Fasting stimulates lipolysis in WAT primarily through catecholaminergic effects, leading to increased FFA release into the blood (reviewed in the study by Duncan et al. [28]). Although baseline serum FFA levels are unchanged in fed Ugcgf/f//CamKCreERT2 mice, fasting-induced FFA release into the blood is significantly decreased (Fig. 2A). FFA release restriction upon refeeding, observed to an equal extent in both groups (Fig. 2B), is mainly mediated by IRs present on WAT cells (reviewed in the study by Desvergne et al. [29]). In line with this, IR expression is normal in peripheral WAT (Fig. 2B and C)

Hormone-sensitive lipase (HSL) catalyzes a key step in lipolysis, namely hydrolysis of diacylglycerides to monoacylglycerides, and is activated upon β3-adrenergic receptor (β3-AR) stimulation. In order to further investigate whether lipolysis is impaired in Ugcgf/f//CamKCreERT2 mice, HSL protein levels and its phosphorylation state were analyzed. The relative occurrences of phospho-Ser563 and phospho-Ser660, which are indicative of active HSL, are significantly decreased in Ugcgf/f//CamKCreERT2 mice, whereas HSL levels themselves remain unchanged (Fig. 2D and E). Relative phosphorylation at Ser565, which is indicative of inactive HSL, is not significantly altered (Fig. 2D and E). Levels for enzymes involved in lipogenesis (acetyl-CoA synthase and fatty acid synthase) are also slightly, but not significantly, decreased (Fig. 2D–F). These results indicate that WAT dysfunction in Ugcgf/f//CamKCreERT2 mice primarily manifests itself as impaired lipolysis. Furthermore, the data support the suggestion that inadequate fasting-induced lipolysis may contribute to obesity in Ugcgf/f//CamKCreERT2 mice.

Since obesity, which had been observed in Ugcgf/f//CamKCreERT2 mice 6 weeks pi (18), itself contributes to impaired WAT function, we analyzed a group of total body weight–matched Ugcgf/f//CamKCreERT2 and control mice 3–4 weeks pi, a time point at which ganglioside depletion is already established in target neurons (18). Just like in mice 6 weeks pi, total parametrial WAT weight is increased in fasted Ugcgf/f//CamKCreERT2 mice 3–4 weeks pi (Fig. 3A). A morphometric analysis of WAT cell size reveals enlarged cells in fasted Ugcgf/f//CamKCreERT2 mice (Fig. 3B and C). Comparable to Ugcgf/f//CamKCreERT2 mice 6 weeks pi, weight-matched mutant mice also defend their body weight during a fasting challenge (Fig. 3D). Furthermore, fasting does not stimulate FFA release in these mice, compared with control mice (Fig. 3E), indicating that the catecholamine-induced stimulation of FFA release upon fasting is already impaired in mutant mice before the onset of severe obesity.

In order to evaluate WAT function in more detail, as has been done in mice 6 weeks pi, we analyzed lipolysis and lipogenesis enzymes in weight-matched mice 3–4 weeks pi by Western blot analysis. In concordance with the observed decreased in vivo release of FFA and the body weight defense despite fasting, HSL phosphorylations at Ser563 and Ser660 are significantly decreased in Ugcgf/f//CamKCreERT2 mice (Fig. 3F and H), whereas levels for HSL, IRs, and phospho-Ser565 remain unchanged (Fig. 3F–H). The expression of the enzymes ACC and FAS, which are involved in lipogenesis, are not significantly altered (Fig. 3F and I).

These data indicate that GCS deficiency and the resulting ganglioside depletion in forebrain neurons result in decreased lipolysis. At that stage, we can exclude that lipolysis impairment is solely due to obesity. Rather, these results support the hypothesis that neuronal GCS plays a role in regulating WAT function.

WAT is densely innervated by sympathetic fibers stimulating fasting-induced lipolysis. We found that WAT of fasted Ugcgf/f//CamKCreERT2 mice contains less NE than WAT of control littermates (Fig. 4A). NE content was still significantly lower when normalized to WAT protein content in order to account for larger WAT cells in fasted Ugcgf/f//CamKCreERT2 mice (Fig. 4B). WAT sensitivity to the synthetic β3-AR–specific agonist BRL has been investigated in an in vitro lipolysis assay (Fig. 4C). Time-dependent and BRL-induced augmentation of FFA release from WAT explants into the medium, indicative of sensitivity to β3-AR agonists, is normal in both groups (Fig. 4D). β3-AR expression in WAT does not differ significantly between groups (Fig. 4E). These findings indicate that fasting-induced SNS activation is impaired in WAT of fasted Ugcgf/f//CamKCreERT2 mice. Furthermore, generalized dysfunction of WAT can be ruled out, since WAT of mutant mice is responsive to a synthetic β3-AR agonist.

MBH insulin signaling inhibits WAT lipolysis, supposedly by dampening SNS outflow to fat pads (11,30). We followed the hypothesis that impaired fasting-induced lipolysis might be linked to altered MBH insulin signaling. Whereas baseline Akt phosphorylation is normal (Fig. 5A and B), Akt phosphorylation is more pronounced in MBH of fasted and insulin-injected Ugcgf/f//CamKCreERT2 mice (Fig. 5C and D). Subsequently, we intended to study IR phosphorylation, for example in the Arc as one MBH region harboring a large number of Cre-targeted GCS- and ganglioside-deficient neurons. We have applied the PLA Duolink method in order to analyze IR tyrosine phosphorylation (Fig. 5E). PLA enables the detection and quantification of close proximity events between two antibody targets (18,31) and has recently been successfully applied to the detection of protein tyrosine phosphorylations (32). Indeed, the p-Tyr/IR PLA shows significantly more PLA signals in insulin-stimulated Ugcgf/f mice, compared with saline-injected mice (Fig. 5F and G). At the same time, the negative controls where we have omitted either anti-p-Tyr or anti-IR only show a background occurrence of a few solitary detectable spots as well as some nonspecific cloudy nuclear signal, which is clearly distinguishable from spots. This indicates that the method enables an estimation of IR phosphorylation. Compared with Ugcgf/f mice, baseline IR phosphorylation is unchanged in Ugcgf/f//CamKCreERT2 mice, while insulin injection leads to significantly more PLA spots in Arc neurons of mutant mice (Fig. 5F and G).

Next, we analyzed IR levels in Arc neurons by applying both PLA (Fig. 6A) and conventional IF. PLA analysis shows that IR levels are slightly but significantly increased in Arc neurons of Ugcgf/f//CamKCreERT2 mice, regardless of insulin stimulation (Fig. 6B and C). These results have further been confirmed by IF analysis of IR staining in Arc neurons (Fig. 6D and E).

These results indicate that IR levels as well as IR signaling are increased in Cre-targeted GCS-deficient neurons of Ugcgf/f//CamKCreERT2 mice, which may contribute to the increased insulin sensitivity observed in the MBH.

We studied the effects of pharmacological GCS inhibition in vitro. We used an immortalized hypothalamic cell line as a homogenous model system (24,25). These cells express markers of mature neurons, like neuron-specific enolase and neurofilament, as well as syntaxin and neurosecretory granules (25). Furthermore, these neurons exhibit a depolarization-induced calcium response (24). In these cells, gangliosides can be depleted by pharmacological GCS inhibition, using the ceramide analog Genz (Fig. 7A), resulting in a prominent reduction of GCS-derived gangliosides (Fig. 7B). Similar to Arc neurons of Ugcgf/f//CamKCreERT2 mice, Genz-treated cells display increased IR levels (Fig. 7C and D). When stimulated with 10 nmol/L insulin, Genz-treated cells show significantly more IR phosphorylation, specifically after short-term stimulation (3 and 5 min) (Fig. 7E and F). When normalized to the slightly increased number of immunoprecipitated IRs in Genz-treated neurons, tyrosine phosphorylation is still significantly increased 3 min after stimulation (Fig. 7E and G).

These results indicate that genetic GCS deletion and pharmacological GCS inhibition exert similar effects on hypothalamic neurons, and that they appear to restrain the quantity of IR signaling and IR protein levels.

In order to acquire further hints of the regulation of IRs by GCS-derived gangliosides, we analyzed proximity dynamics between the IRs and, for example, the abundant complex neuronal ganglioside GD1a. A PLA (Fig. 8A) reveals that proximity events between the neuronal ganglioside GD1a and IR increase dynamically upon insulin stimulation, peaking between 3 and 5 min after insulin exposure (Fig. 8B and C). These proximity events decrease subsequently during longer times of insulin stimulation (Fig. 8B and C). An IF staining for IRα and GD1as and subsequent analysis of fluorescence overlap by confocal microscopy confirmed the findings of the PLA that short-term insulin stimulation results in significantly increased overlap between IR and GD1a staining, which decreases slightly during longer insulin exposure (Fig. 8D and E).

These results indicate that neuronal gangliosides interact dynamically with the IR upon insulin stimulation.

The current study proposes neuronal GCS as a novel CNS regulator of WAT homeostasis. Besides energy storage, a major homeostatic function of WAT is the liberation of FFA and glycerol upon metabolic challenges, like fasting. In Ugcgf/f//CamKCreERT2 mice, fasting-induced lipolysis is impaired. Consequently, these mice defend their body weight more efficiently than control mice. Insulin-evoked Akt phosphorylation as well as IR phosphorylation and IR levels are increased in the MBH of fasted Ugcgf/f//CamKCreERT2 mice. Increased insulin signaling and increased IR levels could furthermore be confirmed in an adult hypothalamic cell line treated with the GCS inhibitor Genz. These results are consistent with recent findings proposing that IR signaling in the MBH restrains lipolysis in peripheral WAT (11). As seen in Ugcgf/f//CamKCreERT2 mice 6 weeks pi (18), obesity in itself can cause impairment of WAT function. For that reason, we analyzed key parameters of WAT function, such as fasting-induced FFA release and phosphorylation status of lipolysis enzymes, in body weight-matched Ugcgf/f and Ugcgf/f//CamKCreERT2 mice 3–4 weeks pi. At that stage, Cre-targeted neurons are devoid of gangliosides (18). We show that lipolysis is already impaired in these mice 3–4 weeks pi, surmising that GCS deletion itself is an important factor in regulating WAT function. On the basis of this work, elucidating whether GCS might regulate IR signaling and fat tissue homeostasis differentially in distinct hypothalamic neuronal subpopulations constitutes a promising target for future studies.

GM3, the major ganglioside in adipose tissue, inhibits IR signaling by directly binding and subsequently sequestering IRs from caveolae (20–22). Furthermore, systemic pharmacological GCS inhibition improves insulin sensitivity in ob/ob mice (33,34). In line with the concept that membrane lipid microdomain integrity is crucial for receptor signaling, our in vivo data indicate that even more complex neuronal gangliosides than GM3 also regulate IR signaling in neurons. Insulin-induced Akt phosphorylation is enhanced in Ugcgf/f//CamKCreERT2 mice. Furthermore, we could demonstrate that IR phosphorylation is also increased in hypothalamic neurons of Ugcgf/f//CamKCreERT2 mice. For example, we chose to study IR levels and IR phosphorylation in Arc neurons by PLA. This method has been widely used for studying protein-protein interactions (31), protein-ganglioside interactions (18), as well as receptor tyrosine phosphorylations (32). Our negative controls furthermore confirm that PLA displays a very low occurrence of nonspecific spots, in addition to a certain degree of nonspecific cloudy nuclear signal. This is, however, clearly distinguishable from PLA spots. PLA has confirmed increased IR levels as well as increased IR tyrosine phosphorylation in Arc neurons of Ugcgf/f//CamKCreERT2 mice. These results have furthermore been verified in vitro in Genz-treated hypothalamic cells. We are, however, not yet able to pinpoint whether increased neuronal IR signaling results from increased IR sensitivity, from increased IR levels, or from a combination of both effects.

Our hypothesis that complex ganglioside GD1a has a regulatory role in IR signaling is, however, supported by PLA data from our in vitro model, showing that GD1a/IR proximity changes in a time-dependent manner upon insulin stimulation. This finding has further been confirmed by an IF analysis of GD1a/IR staining in hypothalamic cells. In light of these findings, an important target for future studies would be to elucidate the regulatory mechanism of the observed proximity dynamics.

Fasting-induced lipolysis is mainly initiated through catecholamine stimulation of β3-AR (28), as SNS activity increases in WAT (35). SNS activity in adipose tissue correlates with NE content, NE turnover rate (36), and HSL phosphorylation (11,28,37). Reduced FFA release, NE content, and less HSL phosphorylation in Ugcgf/f//CamKCreERT2 mice therefore indicate impaired fasting-induced SNS activation in WAT. Contrary to IR signaling (11,30), MBH leptin signaling increases SNS-induced lipolysis in WAT (37). Thus, our findings that Ugcgf/f//CamKCreERT2 mice have decreased MBH leptin signaling (18), as well as increased MBH insulin signaling, indicate that potential synergistic effects may lead to decreased SNS outflow to WAT.

In conclusion, our study indicates that increased IR signaling in MBH of fasted Ugcgf/f//CamKCreERT2 mice contributes to the observed impaired SNS-dependent lipolysis in peripheral WAT. We propose that disturbed WAT homeostasis contributes to the development of obesity in Ugcgf/f//CamKCreERT2 mice. Appropriate GCS expression in neurons therefore constitutes a hitherto unknown essential mechanism for managing WAT homeostasis and body weight.

Acknowledgments. The authors thank Richard Jennemann from the Department of Cellular and Molecular Pathology, German Cancer Research Center in Heidelberg, Germany, for the generation of Ugcgf/f mice.

Funding. This work was funded by grants from the Deutsche Forschungsgemeinschaft (grant SFB 1118, C04 to H.-J.G. and grant NO 1107/1-1 to V.N.) and the European Foundation for the Study of Diabetes (EFSD) (EFSD/Amylin Grant to V.N.).

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

Author Contributions. S.H. conceived the research, designed and performed the experiments, and reviewed the article. S.M. performed the experiments. H.-J.G. conceived the research and reviewed the article. V.N. conceived the research, designed and performed the experiments, analyzed the data, and wrote the article. V.N. 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.