Neuronal Expression of Glucosylceramide Synthase in Central Nervous System Regulates Body Weight and Energy Homeostasis

Body weight and energy homeostasis are regulated by leptin receptor interactions with gangliosides, a class of plasma membrane lipids, in forebrain neurons of mice.


Introduction
The investigation of pathogenetic mechanisms underlying obesity has attained significant interest, as obesity has become an endemic metabolic disturbance worldwide. Elevated peripheral energy storage can develop as a consequence of alterations in the neuronal feedback circuits regulating energy homeostasis. The hypothalamus is the main CNS integrator of peripheral energy signals, matching energy intake to energy expenditure for body weight maintenance [1].
Among the most extensively studied peripheral molecules involved in regulating energy homeostasis and feeding behavior in the CNS are the adipocyte-derived hormone leptin as well as insulin [2,3]. Among numerous leptin-and insulin-sensitive brain areas, the hypothalamic Arc is one of the main regions integrating peripheral energy signals and initiating adaptive metabolic and behavioral responses [4].
Recently, several CNS regions targeted by leptin have emerged that are involved in the regulation of energy metabolism, such as the brain stem nucleus of the solitary tract (NTS) and reward circuits involving the ventral tegmental area [5,6]. Still, leptin is suggested to exert anti-obesity effects by signaling through ''long form'' leptin receptors (ObR) abundantly present on both orexigenic neuropeptide Y (NPY)/agouti-related peptide (AgRP) neurons and anorexigenic pro-opiomelanocortin (POMC) neurons in the Arc. Excess NPY signaling abates sympathetically mediated thermogenesis, thereby reducing energy expenditure [7]. NPY and AgRP expression is attenuated upon ObR-induced phosphatidylinositol-3-OH-kinase (PI3k) signaling [8]. Conversely, leptin stimulates the expression of the POMC-derived neurotransmitter a-melanocyte-stimulating hormone (a-MSH) through the Janus kinase/signal transducer and activator of transcription (Jak-Stat) pathway [9]. Alpha-MSH, a potent agonist of melanocortin receptors, inhibits food intake and stimulates the expenditure of excess energy in the body, thus preventing obesity development [10].
Insulin exerts its anorexigenic effects in hypothalamic neurons by directly stimulating insulin receptor autophosphorylation and activation of PI3k. Even though both insulin and leptin receptor stimulation leads to activation of PI3k and subsequent formation of phosphatidylinositol (3,4,5)-triphosphate (PIP3) [11], it has been shown that both hormones exert converging direct actions on POMC neurons, while having opposite effects on AgRP/NPY neurons [12].
GCS is the key enzyme for the biosynthesis of glycosphingolipids (GSLs) and gangliosides, a class of acidic GSLs abundantly expressed by neurons and glial cells [13,14]. Ganglioside-depleted neurons are viable and show apoptosis rates comparable to wildtype neurons [15]. GSLs including gangliosides contribute to the formation of membrane microdomains, which are important mediators of intracellular signal transduction [16]. GCS expression is crucial for initial postnatal brain maturation and Ugcg f/f//NesCre mice with constitutive Ugcg deletion in brain tissue under the control of the nestin promoter die within 3 wk after birth [15]. In 2003, it was shown that GM3 synthase-deficient mice are more sensitive to insulin, thereby protecting these mice from high-fat-diet-induced insulin resistance [17]. A different ganglioside species, GD1a, has been shown to exert activating effects on tyrosine kinase receptors [18]. To address the functional role of GCS in neuronal regulation of energy homeostasis, we have generated and characterized mice with inducible neuron-specific Ugcg deletion in adult mouse CNS (Ugcg f/f//CamKCreERT2 mice). Cre activity in this mouse model was restricted to distinct populations of forebrain neurons. Hypothalamic nuclei involved in the regulation of energy homeostasis were targeted by this approach. Explicitly, Cre activity was absent in the brain stem NTS, which also contributes to regulation of energy homeostasis.
The present study highlights GCS-derived gangliosides as mediators for ObR-dependent signal transduction at the hypothalamic neuronal membrane. GCS-depleted neurons failed to show ObR activation upon leptin stimulation. Major neuronal gangliosides GM1 and GD1a were recruited to ObR upon ligand stimulation and subsequent signal transduction depended on ganglioside expression in hypothalamic neurons.
Ugcg f/f//CamKCreERT2 mice deficient in GSLs in hypothalamus developed progressive obesity and decreased sympathetically mediated thermogenesis. rAAV-mediated Ugcg delivery to the hypothalamic Arc with ensuing nucleus-specific GSL synthesis significantly ameliorated obesity.

Ganglioside Depletion in Cre-Targeted Neurons in Vivo and in Vitro
Ugcg flox/flox (Ugcg f/f ) mice were bred with mice expressing the inducible CreERT2 recombinase under the control of the Calcium/Calmodulin-dependent Kinase II-alpha (CamK) promoter, resulting in forebrain neuron-specific Ugcg deletion (Ugcg f/f//CamKCreERT2 ) followed by ganglioside depletion after tamoxifen injection ( Figure 1A). Generation of Ugcg f/f mice and CamKCreERT2 mice has been described earlier [15,19].
Beta-galactosidase (X-Gal) staining of brains from R26R/Ugcg f/+//CamKCreERT2 reporter mice indicated strong Cre activity in distinct hypothalamic nuclei, namely in the Arc ( Figures 1B and S1B), in the paraventricular nucleus, and in median preoptic area (MnPO) ( Figure S1A,B). Additional Cre activity was detected in the lateral hypothalamic area (LHA), in hippocampus, and in the cerebral cortex ( Figure S1A,B). Notably, Cre activity was absent in the ventromedial hypothalamus and the NTS in the brain stem ( Figure S1A,B). Ganglioside depletion was confirmed in Cre-targeted areas by GD1a immunofluorescence, whereas non-targeted areas retained GD1a expression ( Figure 1B and Figure S1A).
Consistent with the expected Cre-activity pattern, in situ hybridization showed Ugcg mRNA depletion in hippocampus, cerebral cortex, amygdala, as well as hypothalamic nuclei ( Figure  S1C). Recombination events were absent in peripheral organs and peripheral nervous tissue ( Figure S1D).
Neuron-dense total hippocampi showed significant and stable ganglioside reduction 3 wk postinduction (p.i.), as assessed by thin layer chromatography (TLC) ( Figure S1E). Residual gangliosides in the dissected tissue were assumed to result from glial cells as well as from innervating nerve fibers emerging from nontargeted neurons [14]. Ceramide levels in Cre-targeted neuronal

Author Summary
Obesity is a growing health threat that affects nearly half a billion people worldwide, and its incidence rates in lower income countries are rising dramatically. As obesity is a major risk factor for type II diabetes and cardiovascular disease, significant effort has been put into the exploration of causes, prevention, and potential treatment. Recent research has demonstrated that a region of the brain called the hypothalamus is a major integrator of metabolic and nutrient signals, adapting food intake and energy expenditure to current metabolic needs. Leptin or insulin receptors located in the plasma cell membrane of neurons sense energy signals from the body. They transmit this information inside the cell, which then regulates neuronal function. In this study, we show that leptin receptors interact with gangliosides, a class of plasma membrane lipids. This interaction is a prerequisite for proper receptor activation. Consequently, ganglioside loss in hypothalamic neurons inhibits leptin receptor signal transduction in response to energy metabolites. Furthermore, mice lacking gangliosides in distinct forebrain areas, amongst them the hypothalamus, develop progressive obesity and hypothermia. Our results suggest a previously unknown regulatory mechanism of plasma membrane lipids for hypothalamic control of body weight.
populations were unchanged ( Figure 1C), and a slight increase in sphingomyelin could be detected ( Figure S1F).
In order to investigate if ganglioside depletion abated general neuronal function and integrity in Ugcg f/f//CamKCreERT2 mice, both electron microscopy and electrophysiological slice recordings were done at late time points p.i. Electron microscopy from Arc neurons displayed normal ultrastructure of the neuronal nucleus, organelles, and an intact, regular plasma membrane of Ugcg f/f//CamKCreERT2 mice both 6 and 12 wk p.i. ( Figure 1D). Basic biophysical parameters [spontaneous firing rate, action potential (AP) width, and AP rate of rise] from slice recordings of Arc neurons 12 wk p.i. were unaltered ( Figure S2A). The resting membrane potential and the AP threshold were marginally increased in Ugcg f/f//CamKCreERT2 mice, however not to an extent that impairs neuronal function ( Figure S2B).
In order to confirm these findings in vitro, immortalized mouse hypothalamic cells (N-41 cells) expressing GCS-derived gangliosides ( Figure S3A,B) were treated with n-butyldeoxynojirimycin (NB-DNJ) specifically inhibiting GCS [20]. NB-DNJ treatment resulted in approximately 80%-90% ganglioside depletion ( Figure  S3C). Consistent with the findings in Ugcg f/f//CamKCreERT2 mice, membrane integrity and normal cellular ultrastructure of ganglioside-depleted N-41 cells was confirmed by electron microscopy ( Figure S3D). Additionally, passive and active membrane properties of cultured primary GCS-deficient hypothalamic Ugcg f/f//NesCre neurons [15] were examined by whole-cell recordings. There were no differences toward control cells in membrane resistance, capacitance, and resting potential ( Figure S3E). Spikes evoked by somatic current injection had unaltered threshold, amplitude, and duration ( Figure S3F). These results indicate that basic neuronal integrity and general function are not affected by Ugcg deletion and subsequent lack of plasma membrane gangliosides.
Hematoxylin and eosin (HE) staining revealed enlarged adipocytes in epigonadal white adipose tissue (WAT) ( Figure 2D). In line with this, epigonadal WAT pad weight was significantly elevated ( Figure 2E). Whole body nuclear magnetic resonance (NMR) analysis revealed that body weight increase was due to progressive accumulation of body fat ( Figure 2F); lean mass was only marginally elevated 4 wk p.i. ( Figure S4B). Adjusted for body weight the initial increase of fat and lean mass was proportional, whereas at later stages fat mass overrode lean mass gain ( Figure  S4C). Liver steatosis and morphological changes in major peripheral organs of obese Ugcg f/f//CamKCreERT2 mice were not detected 9 wk p.i. ( Figure S4D). Serum enzyme activities indicative for liver function (glutamate dehydrogenase, glutamic oxaloacetic transaminase, and glutamic pyruvic transaminase) were unaltered ( Figure S4E). Likewise, serum cholesterol, urea, glucose, and creatinine did not show any biologically relevant abnormalities ( Figure S4F). Coincident with obesity, Ugcg f/f//CamKCreERT2 mice were less glucose tolerant than Ugcg f/f mice 12 wk p.i. ( Figure S4G) and insulin sensitivity was marginally impaired 10 wk p.i. ( Figure S4H). These results demonstrate that Ugcg f/f//CamKCreERT2 mice develop progressive obesity that is evident in all adipose compartments with constant lean mass and a shift in body composition toward fat accumulation.
As tight regulation of energy homeostasis is crucial for body weight maintenance [1], a metabolic characterization was carried out in order to study the relation of energy intake to energy expenditure. Food intake and metabolizable energy (E MET ) adjusted to body weight were slightly elevated in Ugcg f/f//CamKCreERT2 mice before the onset of obesity 3 wk p.i. ( Figure 3A,B) when gangliosides were already depleted in Cretargeted brain regions. Hyperphagia was no longer evident 6 and 11 wk p.i., as food intake and E MET were simply elevated due to higher body weight ( Figure 3A,B). Fecal excretion of free fatty acids (FFAs) as well as energy content of feces and extraction efficiency from the food ( Figure S5A) were unaltered. Thus, abnormalities in food intake do initially contribute to obesity development, but not for obesity maintenance.
Energy expenditure was monitored by indirect calorimetry for 21 h. Before onset of body weight gain, the metabolic rate was indistinguishable from Ugcg f/f mice 2 wk p.i. ( Figure S5B). When adjusted for body weight, the average metabolic rate tended to be lower in Ugcg f/f//CamKCreERT2 mice at 5 and 9 wk p.i. ( Figure 3C). Spontaneous locomotor activity is one contributor to daily energy expenditure and has been reported to be decreased in obese rodents [21]. However, both before the onset of weight gain and during progressive adiposity, spontaneous open field activity of Ugcg f/f//CamKCreERT2 mice was indistinguishable from control littermates ( Figure S5C).
The respiratory exchange ratio (RER) provides information on metabolic fuel preferences [22]. Ugcg f/f//CamKCreERT2 mice displayed significantly elevated average daily RER values ( Figure 3D). This finding suggests a shift from lipid oxidation toward lipid storage [22]. In line with this, fat mobilization in response to fasting as assessed by measuring plasma nonesterified free fatty acids (NEFAs) was impaired. Significantly decreased plasma NEFAs were detected in Ugcg f/f//CamKCreERT2 mice 11 wk p.i. (Figure S5D), suggesting a reduced capability to mobilize lipid stores when challenged by food withdrawal.
After the onset of weight gain, Ugcg f/f//CamKCreERT2 mice displayed a prominent drop in core body temperature, as exemplarily depicted 10 wk p.i. ( Figure 3E). Adipocytes in intrascapular brown adipose tissue (iBAT) were enlarged ( Figure  S6A), suggesting reduced triglyceride turnover. Ultrastructural analysis of iBAT furthermore revealed mitochondrial disorganization as well as a lower average mitochondrial size ( Figure  S6B,C). Thermogenesis in iBAT is regulated by synergistic actions of thyroid hormones and sympathoadrenergic signaling [23]. Free triiodothyronine (fT3) and free thyroxine (fT4) levels were normal in Ugcg f/f//CamKCreERT2 mice ( Figure S6D,E). Thus, thyroid dysfunction was unlikely to account for inappropriate thermoregulation. Decreased sympathetic outflow to adipose tissue is assumed to be associated with impaired lipid mobilization [24]. In fact, both iBAT sympathetic activity, as assessed by norepinephrine (NE) turnover rate ( Figure 3F, Figure S6F), and NE content ( Figure S6G) were decreased in Ugcg f/f//CamKCreERT2 mice.
These results demonstrate that Ugcg f/f//CamKCreERT2 mice develop progressive obesity and a shift in body composition toward fat accumulation initially supported by hyperphagia, but maintained due to hypometabolism and hypothermia.

Reconstitution of Ugcg Gene Expression in the Hypothalamic Arc Ameliorates Obesity
Several distinct hypothalamic and nonhypothalamic brain regions were targeted by Cre activity in Ugcg f/f//CamKCreERT2 mice. Arc neurons in Ugcg f/f//CamKCreERT2 mice expressing the long form of the ObR were targeted by Cre activity, as demonstrated by co-immunofluorescence of PStat3 and betagalactosidase (b-Gal) in R26R/Ugcg f/+//CamKCreERT2 reporter mice ( Figure 4A). Other leptin-responsive neurons outside the Arc also targeted by Cre activity, such as the MnPO are likely in part contributing to the observed phenotype. However, ObR-expressing neurons in the LHA seem to be recessed by Cre activity ( Figure S7A,B).
In order to furthermore clarify the role of the Arc in obesity development, we injected recombinant adeno-associated viruses encoding either Ugcg and lacZ (rAAV-Ugcg/LacZ) or only lacZ (rAAV-Empty/LacZ) bilaterally into the Arc of Ugcg f/f//CamKCreERT2 mice after ganglioside depletion before 4 wk p.i. Injection of rAAV-Ugcg/LacZ significantly ameliorated obesity, underlining the importance of Ugcg expression in the Arc for body weight maintenance ( Figure 4B). Consistently, serum leptin levels tended to be lower in rAAV-Ugcg/LacZ-treated mice ( Figure 4C). We verified correct targeting of the Arc by X-Gal staining of the brains injected with rAAV-Ugcg/LacZ and displayed targeted regions in a schematic drawing as well as a typical staining (Figures 4D-F and S7C). Animals that were not targeted by rAAV-Ugcg/LacZ in the Arc (rAAV-Ugcg/LacZ missed) did not improve their weight gain ( Figure S7D). Restored ganglioside biosynthesis in the Arc of rAAV-Ugcg-treated animals compared to mice injected with viruses encoding empty plasmid was demonstrated by GD1a immunofluorescence ( Figure 4G and Figure S7E).
Taken together, these results indicate that loss of GCS expression in the Arc is significantly involved in part of the metabolic deregulation seen in Ugcg f/f//CamKCreERT2 mice.

GCS-Derived Gangliosides Regulate Leptin Receptor Signaling in Hypothalamic Neurons at the Plasma Membrane
Since the number of neurons in the Arc did not differ between Ugcg f/f//CamKCreERT2 mice and controls ( Figure S8A), a functional analysis of the Arc was performed. Leptin signaling in the hypothalamus is crucial for the maintenance of body weight and energy homeostasis. As adipocyte-secreted leptin is a major regulator of body weight in the CNS, we hypothesized that leptin signaling might be disturbed in GCS-deficient neurons of Ugcg f/f//CamKCreERT2 mice. In order to test this hypothesis, we investigated hypothalamic Stat3 phosphorylation (PStat3) in the Arc after peripheral leptin stimulation. Decreased PStat3 was detected by immunofluorescence in the Arc ( Figure 5A) and by Western blot in mediobasal hypothalamus ( Figure S8B). Interestingly, baseline Stat3 levels were elevated in Ugcg f/f//CamKCreERT2 mice ( Figure S8C). The PStat3/Stat3 ratio was decreased both at baseline and upon leptin challenge ( Figure S8D).
It has been shown that deficient ObR signaling due to leptin resistance of the Arc in mice with diet-induced obesity (DIO) is a consequence of long-term elevated leptin levels [25][26][27]. The suppressor of cytokine signaling 3 (SOCS-3) is a major negative regulator of the ObR that is elevated in rodent models of leptin resistance [25,28]. In line with progressive obesity, Ugcg f/f//CamKCreERT2 mice show indeed elevated leptin levels 7 wk p.i. ( Figure 5B). However, expression of hypothalamic Socs-3 did not rise with increasing obesity and leptin levels, as measured 2, 6, and 9 wk p.i. ( Figure 5C). Moreover, hypothalamic ObR expression, usually elevated in leptin-resistant rodents [29,30], was normal in Ugcg f/f//CamKCreERT2 mice 6 wk p.i. ( Figure 5D).
To further investigate if GCS-derived gangliosides regulate proper leptin receptor signaling at the level of the plasma membrane in hypothalamic neurons, we first assured that loss of gangliosides would not interfere with ObR transport to the membrane, which would have impaired ObR signaling per se. ObR was labeled by an in situ proximity ligation assay (PLA) on non-detergent-perturbed cells by two ObR antibodies. The number of detected surface ObR PLA spots on cells treated with NB-DNJ was similar to control cells ( Figure 5E), indicating that ObR at the plasma membrane of ganglioside-depleted hypothalamic cells is not significantly changed compared to control cells.
As GCS-derived gangliosides have previously been shown to modulate the activity of plasma-membrane-located receptors through close interactions in both adipocytes [17] and neurons [31], we investigated ObR interactions with major neuronal gangliosides. The PLA indicating close proximity events [32] indeed revealed proximity between ObR and gangliosides GM1 and GD1a. In demonstration of activity-dependent interaction between GSL and ObR, the number of GD1a/ObR and GM1/ ObR PLA spots increased upon stimulation with leptin (Figures 5F,G and S8E). Complex formation between GD1a/ GM1 with ObR was further corroborated by co-immunoprecipitation (Co-IP) of ObR and GD1a/GM1 in saline-and leptinstimulated N-41 cells ( Figures 5H and S8F). As N-41 cells do not express the complex neuronal gangliosides GD1b and GT1b, potential interactions with ObR had to be analyzed in hypothalamic tissue of Ugcg f/f mice. GD1b and GT1b could not be coprecipitated with ObR ( Figure 5H).
Ganglioside-depleted cells were then assessed for leptin-dependent signal transduction. Ganglioside-depleted cells did not show the leptin-stimulated increased complex formation between ObR and Jak ( Figures 5I and S8G). Time-and dose-dependent Jak phosphorylation could be induced by leptin treatment in N-41 cells and was decreased in NB-DNJ-treated GSL-depleted cells ( Figures 5J and S8H). It has to be noted that NB-DNJ evokes ganglioside depletion by only approximately 80%-90% ( Figure  S3C). Thus, residual gangliosides in the plasma cell membrane may explain the appearance of a P-Jak signal at a late time point after stimulation of NB-DNJ-treated cells. Ganglioside-depleted N-41 cells showed decreased Jak phosphorylation 30 min after stimulation with 0.5 mg/ml leptin ( Figure 5J,K).
These results have now shown that two major neuronal GCSderived gangliosides, GD1a and GM1, form dynamically leptinstimulated complexes with ObR on the plasma membrane and that loss of gangliosides decreases signal transduction in hypothalamic neurons.

Distinct Hypothalamic Neurons of Ugcg f/f//CamKCreERT2 Mice Are Less Responsive to Peripheral Leptin
It is known that mice with deficient leptin receptor (db/db mice) function develop obesity and lack hypothalamic responsiveness to leptin stimulation [33]. Regarding the finding that neuronal gangliosides enhance ObR signaling, we hypothesized that hypothalamic neuronal function may be altered in Ugcg f/f//CamKCreERT2 mice. In order to investigate this question, neuronal activity after intraperitoneal (i.p.) leptin injection was evaluated by c-Fos staining [34]. Leptin-induced c-Fos formation was normal in non-obese Ugcg f/f//CamKCreERT2 mice 1-2 wk p.i. ( Figure 6A). Since ganglioside depletion coincides with the start of the obesity development, Ugcg f/f//CamKCreERT2 mice that were weight-matched to control littermates were analyzed 3-4 wk p.i. Decreased leptin responsiveness could already be observed in the Arc of these mice ( Figure 6B) as well as in the Arc of obese mice 6 wk p.i. ( Figure 6C).
Neurons in the nontargeted and non-ganglioside-depleted VMH retained responsiveness to leptin at all time points ( Figure 6D-F). As expected, the nontargeted brain stem NTS of Ugcg f/f//CamKCreERT2 mice showed regular leptin-induced c-Fos staining 6 wk p.i. ( Figure S9).
Altogether, these results indicate a primary deficiency of ganglioside-depleted hypothalamic neurons to respond adequately to peripheral leptin signals.

Ganglioside-Depleted NPY/AgRP and POMC Neurons in the Arc Are Less Responsive to Leptin
Antagonistic orexigenic NPY and anorexigenic POMC neurons in the hypothalamic Arc are first-order responsive neurons initiating metabolic adaptations to altered peripheral leptin levels [4]. In order to determine leptin-dependent NPY and POMC neuronal function, neuronal activity and ObR activation were assessed by semiquantitative analysis of c-Fos, PStat3, and PIP3 formation in response to peripheral leptin injections. Leptin engaged POMC neurons (a-MSH positive) in control mice, as indicated by increased c-Fos ( Figure 7A). Significantly elevated PStat3 ( Figure 7B) and PIP3 formation ( Figure S10A) confirmed activation of their ObR. Before ganglioside depletion (1-2 wk p.i.), POMC neurons of Ugcg f/f//CamKCreERT2 mice responded normally to leptin. However, c-Fos, PStat3, and PIP3 formation were not elevated in response to leptin in obese GSL-deficient mice 6 wk While a slight decrease in c-Fos-positive NPY neurons was found in leptin-injected control mice, leptin did not show any such effect in Ugcg f/f//CamKCreERT2 mice 6 wk p.i. ( Figure 7C). Similarly, leptin did not raise PStat3 levels in NPY neurons of Ugcg f/f//CamKCreERT2 mice 6 wk p.i. ( Figure 7D) and did not have any direct effect on PIP3 formation ( Figure S10C). Remarkably, basal mRNA expression of Agrp and Npy was markedly elevated in the MBH of Ugcg f/f//CamKCreERT2 mice 6 and 9 wk p.i., with Agrp already increasing 2 wk p.i. ( Figure S10D).
In summary, this study has indicated that GCS expression and sufficient gangliosides in neurons of the adult CNS play a seminal role in the regulation of body weight and energy homeostasis. Analysis of the leptin receptor signaling pathway, being one of the most prominent regulators of CNS metabolic control [35,36], revealed that GCS-derived gangliosides interact with ObR on the plasma cell membrane, thereby facilitating ObR-dependent signal transduction ( Figure 8A). In Ugcg f/f//CamKCreERT2 mice, leptin responsiveness and neuronal function are impaired in hypothalamic neurons involved in the regulation of energy metabolism ( Figure 8B). Consequently, defective ObR signaling contributes to the observed metabolic imbalance and obesity development of mice with ganglioside deficiency in the CNS.

Discussion
Although the seminal role of CNS feedback responses to peripheral energy signals for the regulation of energy homeostasis has been extensively studied, the role of the lipid microenvironment for energy signal receptor function has not yet been addressed. The present study demonstrates that GCS-derived GSLs are critically involved in a to-date unknown mechanism of hypothalamic control of body weight. In line with the finding that neurons of the constitutive Ugcg f/f//NesCre mice do not show increased apoptosis [15], ganglioside-deficient hypothalamic neurons are viable and they show normal membrane and organelle appearance both in vivo and in vitro. Electrophysiological recordings from Arc neurons in slices of Ugcg f/f//CamKCreERT2 mice at 12 wk p.i. did not show a major disruption of membrane functions. However, resting membrane potential and action potential threshold were both shifted to slightly more depolarized values. The molecular mechanism underlying the altered membrane potential remains presently elusive. However, it is well feasible that the shift of threshold is secondary to the slight depolarization, which might inactivate a fraction of Na + channels. In line with the largely normal properties of neurons from brain slices, biophysical parameters of primary hypothalamic neurons devoid of gangliosides were unaltered. Thus, failure of basic electrophysiological membrane functions is unlikely to cause the observed phenotype of mice with ganglioside deficiency.
Therefore, the present work focuses on interactions of leptin receptors with the ganglioside-containing lipid microevironment in which receptors are embedded. We show with independent methods that two major neuronal GCS-derived gangliosides, GD1a and GM1, closely interact with leptin receptors on the neuronal membrane. This interaction is dynamically enhanced by stimulation with leptin. Both Ugcg f/f//CamKCreERT2 mice and ganglioside-depleted hypothalamic cells display deficient ObR signal transduction upon leptin stimulation, as assessed by decreased leptin-induced Jak phosphorylation, Stat3 phosphorylation, and PIP3 formation. Corroborated in situ by deficient leptin responsiveness in Arc neurons of Ugcg f/f//CamKCreERT2 mice, these results indicate that GCS-derived GSLs, primarily gangliosides, are seminal regulators for neuronal leptin signal transduction. Consequently, Ugcg f/f//CamKCreERT2 mice with deficient leptin-induced hypothalamic neuronal responsiveness develop progressive obesity.
Numerous hypothalamic feedback systems involved in body weight maintenance are known [1,4]. Admittedly, the robust phenotype of Ugcg f/f//CamKCreERT2 mice may be caused by several peripheral hormones and defective ensuing signaling events occurring in various Cre-targeted CNS regions of this mouse model. The brain stem NTS, though an important mediator of metabolic control [37], is not targeted by activity under the CamK II alpha-dependent Cre recombinase used in this study. Consequently, the NTS shows normal responsiveness to leptin in obese Ugcg f/f//CamKCreERT2 mice and can be excluded to contribute to the observed phenotype. Recent reviews also highlight the LHA as an important regulator of energy balance [38,39]. In fact, compensating neurocircuits involving nontargeted CNS regions may be considered for the return of food intake from initial hyperphagia to normal levels in obese mice despite the striking increase in orexigenic neuropeptides. Even though X-Gal staining could be seen in parts of the LHA, we could not verify Cre targeting of a major part of ObR-expressing LHA neurons in Ugcg f/f//CamKCreERT2 mice ( Figure S7B). In strong support to this line of reasoning, we demonstrate that partial Ugcg replenishment in the Arc mediated by stereotactic injection of rAAV significantly  ameliorates obesity and hyperleptinemia in Ugcg f/f//CamKCreERT2 mice. Even though limited infection of closely attached tissue by rAAV injection could not be definitely excluded, mainly Arc neurons were targeted by this approach, as assessed by X-Gal stainings of brains co-injected with LacZ-expressing viruses. The present investigation has thus been restricted to GCS effects focused on the MBH harboring Arc neurons.
Deficient leptin signaling as a consequence of leptin resistance occurs predominantly in the Arc of DIO mice with severe longterm hyperleptimenia [25][26][27]40]. Socs-3 is a major negative feedback pathway of ObR signaling [41]. Thus, elevated Socs-3 expression levels are found in the hypothalamus of leptin-resistant rodent models [25,42]. In line with observations in obese db/db mice with nonfunctioning ObR [25,42], Socs-3 expression in the Arc remains indistinguishable from control littermates in nonobese and obese Ugcg f/f//CamKCreERT2 mice 2, 6, and 12 wk p.i. Elevated hypothalamic ObR expression, as it occurs in DIO mice [29,30], has also been proposed as a potential mechanism playing a role in the development of leptin resistance [28]. However, normal ObR expression in Ugcg f/f//CamKCreERT2 mice supports the hypothesis  that the ObR signaling in their neurons must be deficient due to ganglioside loss and not merely due to secondary leptin resistance. Furthermore, the nontargeted hypothalamic VMH and brain stem NTS retain leptin responsiveness even in obese mice 6 wk p.i. These results in combination with the decreased ObR signal transduction in ganglioside-depleted and non-leptin-resistant N-41 cells strongly suggest that loss of GCS-derived GSLs including gangliosides GD1a and GM1 is the reason for failing ObR activation and subsequently inhibited intracellular signaling.
GCS-depleted Arc neurons display normal leptin sensitivity 1-2 wk p.i., a time point when gangliosides are still present. Furthermore, onset of body weight gain, deficient neuronal activity in the Arc, and abolished ObR signaling coincide with ganglioside depletion 3 wk p.i. This strongly suggests that the mentioned defects are due to ganglioside depletion in these cells rather than due to lack of the enzyme GCS itself.
Further evidence for the postulate that ganglioside deficiencydependent inhibition of ObR signaling in hypothalamic neurons leads to impaired neuronal function is based on our in situ results in ganglioside-depleted Arc of both obese and nonobese Ugcg f/f//CamKCreERT2 mice. Whereas leptin injection increases c-Fos immunoreactivity and thus neuronal activity in the Arc neurons of fasted lean mice, this response did not occur in GCSdeficient neurons. Leptin specifically engages POMC neurons. Even though the effects of PI3k-and Stat3-dependent signaling in POMC neurons do not overlap [12,43], both pathways are activated by leptin [11,12,44,45] and contribute to maintenance of energy homeostasis [46]. In ganglioside-depleted POMC neurons, neither PStat3 nor PIP3 formation is increased by peripheral leptin injections, strongly suggesting that defects in both pathways may contribute to partial failure of obesity prevention. As peripheral leptin stimulates both pathways through ObR activation [11,47], defective ObR function is very likely to be assumed.
In NPY neurons, it has been demonstrated that Jak-Stat3 signaling plays an important role in maintaining NPY/AgRPmediated energy homeostasis [48]. Additional ObR-mediated PI3k activation seems to be required for inhibiting Npy and Agrp gene expression [8]. Npy and Agrp expression is markedly increased in the MBH of Ugcg f/f//CamKCreERT2 mice, which may be a consequence of absent leptin-induced PStat3 formation in NPY neurons. On the other hand, leptin-induced PIP3 formation does not differ in neither of the groups, which goes in line with the hypothesis that leptin-dependent PIP3 formation in AgRP/NPY neurons is stimulated by an indirect mechanism involving synaptic transmission [12]. Overactive NPY neurons in obese ObRdeficient Lepr fa/fa rats were shown to inhibit sympathetic nervous outflow to BAT and cause hypothermia [7] as observed in Ugcg f/f//CamKCreERT2 mice. With regard to the fact that Npy and Agrp but not Pomc expression differ in Ugcg f/f//CamKCreERT2 mice, the role of GCS expression in regulating neuropeptide expression and secretion has to be elucidated. Especially the role of hypothalamic insulin receptor signaling, which also regulates the expression of Pomc and Npy/Agrp in part similar to ObR signaling [11] and is antagonized by GM3 in the periphery [17], constitutes a promising target for further clarifying the differential neuropeptide expression. Moreover, a potential contribution of Cretargeted ObR-expressing neurons in the median preoptic area of Ugcg f/f//CamKCreERT2 mice to hypothermia may also be considered.
Dynamic membrane microdomains are widely accepted as critical components involved in membrane receptor functions [16,49]. Since GCS-derived gangliosides are important constituents of these microdomains, they potentially interact with and regulate a variety of membrane components including receptors such as Trk receptors [31] and insulin receptors [17]. In contrast to mice with neuron-specific insulin receptor deletion, which only display a gender-and diet-dependent subtle increase in body weight [11,50], the obesity and glucose intolerance observed in db/db mice can be rescued by neuron-specific re-expression of ObR [51]. Furthermore, deficient ObR signaling in POMC neurons of the Arc itself leads to the development of mild obesity [52]. In consideration of these findings-despite the existence of potential alternative pathways that might be impaired in neurons of Ugcg f/f//CamKCreERT2 mice-we ascribe ObR and its regulation of activity to a major function in our model pointing to a novel mechanism for CNS metabolic regulation.
We demonstrate that GCS-derived gangliosides GD1a and GM1 closely interact with ObR. The leptin-induced increase in GD1a/ObR and GM1/ObR interaction assumes recruitment of these gangliosides to the ObR upon leptin stimulation. These results in combination with the demonstrated deficient ObR signaling in ganglioside-depleted hypothalamic neurons both in vivo and in vitro leads us to surmise that the lipid microenvironment surrounding the ObR can significantly modulate leptindependent intracellular signal transduction in hypothalamic neurons. Altogether, these results provide evidence that GM1 and GD1a are actively involved in enhancing the effects of leptin in hypothalamic neurons.
As insulin receptors contain a lysine residue predicted for interaction with GM3 [53], loss of GM3 synthase showed already a prominent effect on peripheral insulin receptor signaling [17]. It is a widely accepted concept that in the state of insulin resistance in peripheral adipocytes, the IR segregates from caveolae into GM3enriched microdomains [53], an endogenous inhibitory mechanism [17]. Indeed, elevated GM3 synthase expression could be detected in adipose tissue of obese Zucker fa/fa rats and ob/ob mice [54]. Pharmacologic GCS inhibition in the periphery has been shown to exert beneficial effects on peripheral insulin sensitivity and liver steatosis [55,56]. With regard to the fact that different ganglioside species can exert either stimulatory [18,57] or inhibitory [17] effects on membrane receptors, the mentioned studies including the present work support the concept that any perturbation, either loss or excess, of membrane GSLs can alter receptor function. Contributions of GCS-derived lipid raft components apart from gangliosides, namely neutral GSLs in the CNS, to leptin receptor function have yet to be elucidated and constitute a challenging target for future investigations. Besides gangliosides, lactosylceramide has been shown to contribute to formation of lipid microdomains [58]. We, however, propose in the present study that in line with the findings for the insulin receptor, hypothalamic leptin receptor signaling is to a significant extent regulated through interactions with the dominant gangliosides GD1a and GM1.
Recent studies have highlighted the central role of systemic ceramide biosynthesis and GCS in the regulation of energy homeostasis [59,60]. In accordance with earlier findings [15,61], we show that neuronal ceramide levels in Ugcg f/f//CamKCreERT2 mice are indistinguishable from control mice, virtually excluding any effects of ceramides.
In conclusion, our study highlights the expression of neuronal GCS-derived GSLs, foremost gangliosides, as a novel class of hypothalamic metabolic regulators. Gangliosides GM1 and GD1a enhance the action of leptin on intracellular signaling and neuronal activity, most likely through dynamic changes of the lipid microenvironment of the ObR. We demonstrate by independent methods that gangliosides GD1a and GM1 strongly interact with the ObR upon leptin stimulation. Loss of these gangliosides leads to impaired responsiveness. By this relevant influence on hormone signaling, Ugcg deletion in adult mouse CNS leads to development of progressive obesity, hyperleptinemia, and glucose intolerance. The obesity can be partially ameliorated by restoration of GCS activity and ganglioside expression in the hypothalamic Arc of Ugcg f/f//CamKCreERT2 mice. Neuronal GCS expression therefore constitutes a novel mechanism for hypothalamic regulation of body weight maintenance.

Materials and Methods
Ugcg f/f//CamKCreERT2 Mice Animals were kept in specific-pathogen-free barrier facilities.
Ugcg f/f mice [15] and inducible CamKCreERT2 mice were bred to generate Ugcg f/f//CamKCreERT2 mice and control littermates. Mice were induced with tamoxifen 6 wk after birth for 1 wk as described [19]. We performed experiments in female mice, unless stated otherwise.

Glucose Tolerance and Insulin Sensitivity
Mice were fasted overnight (o/n). Blood glucose levels were analyzed prior to i.p. injection of glucose (2 g/kg body weight). Glucose levels were determined from tail vein blood (Glucometer Accu Check, Aviva, Roche). Food was withdrawn 4 h prior to the insulin sensitivity assay. Mice were injected i.p. with 0.75 U/kg human insulin (Eli Lilly), and glucose levels were determined as described above (see also Text S1).

Generation of rAAV
The AAV Helper-free System (Agilent Technologies Inc.) was used for preparation of rAAV. Full-length mouse Ugcg cDNA was cloned into the pAAV-MCS vector from the Helper-free System (pAAV-Ugcg). Viruses were generated according to the manufacturer's guidelines and purified as described earlier [64].
rAAV Injection Into the Arc Bilateral stereotaxic injections were performed as described [65]. We injected 400 nl virus solution containing equal volumes of viruses carrying Ugcg and lacZ (,1.8*10 11 genome copies/ml) into the Arc of each hemisphere (caudal to bregma: 1.4 mm, 1.44 mm; lateral: 0.25 mm; ventral: 5.7 mm). After surgery, mice were maintained with ad libitum access to lab chow and body weight was monitored weekly.

Measurement of Serum Leptin and Nonesterified Free Fatty Acids (NEFAs)
Serum leptin and NEFAs were determined by commercially available kits according to the manufacturer's guideline [Leptin-ELISA (Linco); NEFA-HR2 kit (WAKO Chemicals)]. NEFAs were measured in male mice.

Determination of NETO and NE Content in iBAT
NETO rate in iBAT was determined as described earlier [67]. Tissue NE was measured by reversed-phase HPLC with electrochemical detection (Chrome Systems, Germany) (see also Text S1).
Metabolic Characterization, Core Body Temperature, and Locomotor Activity Body weight was measured once a week. Metabolic measurements were carried out in an open circuit respiratory system (SM-MARS, Sable Systems, USA). VO 2 and VCO 2 per mouse were analyzed for 21 h to determine the RER = V CO2 /V O2 and HP (mW). Whole body composition was determined by noninvasive NMR analysis (Mini-Spec, Bruker Optics). Core body temperature was measured with a rectal probe (ALMENO 2390-1, Ahlborn) (see also Text S1).

Quantitative mRNA Analysis
Total RNA of the MBH was extracted from nonfasted mice as described earlier [68]. RNA was reversely transcribed by Superscript II Reverse Transcriptase (Invitrogen) and cDNA was quantified using the LC FastStart DNA Master SYBR Green I kit (Roche) according to the manufacturer's guidelines and the Light Cycler (Roche) (see also Text S1).

Extraction and Analysis of GSLs and Ceramide
GSLs were extracted and separated into neutral and acidic fractions containing gangliosides as described earlier [15]. The amount of GSLs spotted onto a plate by a TLC applicator (Camag, USA) was normalized to tissue protein content determined by the Lowry method [70]. TLC running solvent for acidic GSL was chloroform/methanol/0.2% CaCl 2 (45:45:10 by vol). GSLs were visualized with 0.2% orcinol in 10% sulphuric acid at 120uC for 10 min. Ceramide was extracted as described earlier [15] and spotted onto a TLC plate. Running solvent for ceramide was chloroform/methanol/acetic acid (190:9:1 by vol), and ceramide was visualized with 10% CuSO 4 in 8% H 3 PO 4 at 180uC for 10 min. Lipid content was quantified by densitometry (Shimadzu, Japan).

Statistical Analysis
Unless stated elsewhere, results were analyzed by a two-tailed, unpaired Student's t test (Graph Pad Prism, Graph Pad Software, Inc.). To analyze main effects of genotype on metabolizable energy or energy expenditure, body weight was employed as a co-factor in a linear regression model to account for the confounding effect of body size on energy metabolism parameters [71]. p#0.05 was considered statistically significant and marked *. p#0.01 was marked **, and p#0.001 was marked ***.  Figure 5E) for GD1a/ObR similar to the experiment depicted in Figure 5F was performed using a different ObR-specific antibody. GD1a/ObR interactions were detected and leptin treatment dynamically increased the GD1a/ObR PLA spots per cell similar to the result depicted in Figure 5F. Pre-adsorption of the antibody by a blocking peptide abolished PLA signals (n = 74-150 cells). (F) Similar to the experiments depicted in Figure 5H, extracts from saline-and leptintreated N-41 cells were immunoprecipitated with a second ObR antibody, lipids were extracted, and GD1a and GM1 were visualized by immune overlay TLC. Similar to the results obtained in Figure 5H, GD1a and GM1 co-immunoprecipitated (Co-IP) with ObR. (G) Similar to the result in Figure 5I, Jak was co-precipitated with the ObR, which increased upon leptin stimulation of N-41 cells, using a second ObR antibody. Addition of the blocking peptide abolished the signal. (H) Weak Jak phosphorylation was induced in N-41 cells after 30 min of leptin treatment (0.1 mg/ml). *p#0.05; **p#0.01; ***p#0.001. Means 6 SEM. (TIF) Figure S9 The brain stem NTS of Ugcg f/f//CamKCreERT2 mice retained leptin responsiveness 6 wk p.i. c-Fos expression was unaltered in the nontargeted NTS (n = 2-4 sections). Datasets for each time point were acquired individually. Immunofluorescence and image acquisition for each dataset (treated and untreated controls and knockouts) were performed simultaneously. *p#0.05; **p#0.01. Means 6 SEM. (TIF) Immunofluorescence and image acquisition for each dataset (treated and untreated controls and knockouts) were performed simultaneously. n = 3-7 sections; *p#0.05; **p#0.01; ***p#0.001.