Anorexia and Impaired Glucose Metabolism in Mice With Hypothalamic Ablation of Glut4 Neurons

The central nervous system (CNS) plays an integral role in maintaining glucose homeostasis (1). It exerts systemic effects on organismal turnover and storage of glucose by promoting hormone release and through sympathetic and parasympathetic innervation of peripheral organs (2). In addition, the CNS is a target organ of diabetes complications, such as stroke, hypoglycemia unawareness, as well as sympathetic dysfunction, not to mention the longstanding and poorly understood relationship between cognitive dysfunction and diabetes (3). Despite the established role of the CNS in metabolism, the identities of effector neurons remain elusive.

We have previously shown that the insulin-responsive glucose transporter, Glut4, is a marker of a distinct subset of neurons (4). We are agnostic as to the function of Glut4 itself in these neurons, but we have pursued the hypothesis that it represents a vestigial marker of their regulatory metabolic functions. This concept is borne out by the finding that ablating insulin receptors (InsR) in a constellation of Glut4-expressing tissues (muscle, fat, and neurons) causes overt diabetes (4), in contrast with the mild metabolic phenotypes observed following InsR knockout in either muscle, fat, or both (5). These findings highlight the potential metabolic function of InsR signaling in Glut4 neurons, as well as the latter’s distinct neuroanatomical identity.

Previous studies identified Glut4 expression in the brain (6,7). We have found Glut4-expressing neurons in the CNS among cortical pyramidal neurons, cerebellar Purkinje-like, olfactory bulb, hippocampal, and hypothalamic neurons (4). Given their anatomic and morphological differences, Glut4 neurons are likely to mediate several processes. To begin to clarify their metabolic contribution, we combined cell ablation and metabolic profiling studies to interrogate their hypothalamic functions. Ablation of hypothalamic Glut4-expressing neurons resulted in anorexia and fasting hyperglycemia, altering the response to insulin and leptin. We propose that hypothalamic Glut4 neurons constitute a distinct neuronal entity and mediate CNS effects on energy balance and peripheral metabolism.

Gt(Rosa)26Sor^{tm9(CAG-tdTomato)Hze} and C57BL/6-Gt(ROSA)26Sor^{tm1(HBEGF)Awai/J} mice were from The Jackson Laboratory. The Columbia University Animal Care and Utilization Committee approved all procedures. Normal chow diet (NCD) had 62.1% calories from carbohydrates, 24.6% from protein, and 13.2% from fat (PicoLab rodent diet 20, 5053; Purina Mills); high-fat diet (HFD) had 20% calories from carbohydrates, 20% from protein, and 60% from fat (D12492; Research Diets). We measured weight and length to calculate BMI and estimated body composition by nuclear magnetic resonance (Bruker Optics). We generated Glut4-Cre;Rosa-iDTR and Pomc-Cre;Rosa-iDTR mice by mating Glut4-Cre or Pomc-Cre male transgenic mice with female Rosa-iDTR mice and genotyped them as previously described (4).

We measured food intake and refeeding response as described (8) using a TSE Labmaster Platform (TSE Systems) for indirect calorimetry and activity measurements (9). We used ELISA for leptin, insulin, and glucagon measurements (Millipore) and colorimetric assays for metabolite measurement (4).

We performed stereotactic manipulations in 3- to 4-month-old mice. We placed cannulae and allowed 1 week for recovery (bilateral targeting −1.5 mm anterior and 0.4 mm lateral to bregma and 4.7 mm below the skull surface). We verified correct positioning of the cannulae by injecting methylene blue (1%) after killing. To ablate cells, we injected diphtheria toxin A (DTA) at a dose of 4 ng/mouse (injector with 1-mm projection). Control mice (carrying Rosa-DTR allele) are littermates of Glut4 neuron–ablated mice (carrying Glut4-Cre;Rosa-DTR alleles). Both control and Glut4 neuron–ablated mice received DTA injection. Only mice carrying Glut4-Cre;Rosa-DTR alleles are susceptible to DTA-mediated cell death. To trace neuronal projection, we injected Ad-iZ-EGfpf 1 μL to the bilateral cannula. For GABA infusion, Alzet 14-day micro-osmotic pumps (model 1002; Durect, Cupertino, CA) loaded with bretazenil (60 ng/mL in saline plus 0.2% DMSO; Sigma-Aldrich, St. Louis, MO) were implanted subcutaneously on the back of cannulated mice immediately after DTA injection. The minipumps dispensed bretazenil at a rate of 15 ng/h. For immunohistochemistry studies, we perfused mice 4 to 5 h after refeeding.

We processed mouse brains and cut 10-μm-thick coronal (unless otherwise noted) sections for immunohistochemistry as described (8), using phospho-Akt (pAkt; #4060; Cell Signaling Technology), phospho-Stat3 (pStat3; #9131; Cell Signaling Technology), phospho-ribosomal protein S6 (pS6; #4858; Cell Signaling Technology), and TUNEL kit (Roche Applied Sciences). Sagittal sections and confocal microscopy (LSM710; Zeiss) were used for tracing studies in the hindbrain (scale bar, 100 μm).

We extracted RNA with TRIzol (Invitrogen) and reverse transcribed with SuperScript II reverse transcriptase. We performed quantitative PCR using primers spanning introns. Primer sequences are available upon request.

We dissociated mediobasal hypothalami from neonates with the papain dissociation kit (Worthington Biochemical). We gated live neurons to collect Rfp-positive and/or Gfp-positive neurons. We analyzed the FACS data using FlowJo software. We quantified Glut4, neuropeptide Y (NPY), and proopiomelanocortin (POMC) neurons in adults by counting fluorescent cells in arcuate nucleus (ARH) sections using ImageJ software.

We analyzed data with Student t test or one-way ANOVA using GraphPad Prism and Partek Genomic Suite software. We used the customary threshold of P < 0.05 to declare statistical significance.

Mice are not susceptible to DTA, due to lack of diphtheria toxin receptor (DTR) (10). We took advantage of this property and introduced a Rosa-iDTR transgene expressing DTR into Glut4-Cre mice to generate Glut4-Cre; Rosa-iDTR mice. DTR is inserted into the Rosa26 locus and maintained inactive by a lox-stop-lox sequence preceding the DTR cDNA. Upon Cre-mediated excision, the lox-stop-lox sequence is removed, leading to DTR expression, and rendering Glut4-expressing cells susceptible to DTA in Glut4-Cre; Rosa-iDTR mice.

To ablate basal hypothalamic Glut4 neurons, we delivered DTA to that location by stereotactic injection. Forty-eight hours after DTA injection, we killed mice and performed terminal deoxynucleotidyl TUNEL to document cell loss. We detected TUNEL-positive cells in the basal hypothalamus, including arcuate, ventromedial, and lateral hypothalamic nuclei, with sporadic distribution to the dorsomedial nucleus (Fig. 1A). We did not see any TUNEL-positive cells in other brain regions, including regions with high levels of Glut4 expression, such as hippocampal field CA1 and dentate gyrus (Fig. 1A) (4). In addition, we did not detect peripheral toxicity induced by DTA in Glut4-expressing tissues outside the CNS, such as muscle and fat (data not shown). To confirm the loss of Glut4 neurons, we performed immunostaining with Glut4 antibody (Fig. 1B). Glut4 neurons were detectable in the basal hypothalamus of control mice, but were absent in the ablated mice (Fig. 1B). As a control, hippocampal Glut4 neurons were well preserved (Fig. 1B).

Three days after DTA injection, we began to see changes in energy homeostasis concomitant with the onset of neuronal apoptosis. Glut4 neuron–ablated mice showed significant reductions of oxygen consumption VO₂ (Fig. 2A) and respiratory quotient (RQ) during the dark phase of the light cycle (Fig. 2C), indicating impaired energy use. In contrast, locomotion remained normal throughout the light cycle (Fig. 2E). Food intake was severely curtailed during the dark phase (Fig. 2G). The pronounced anorexia after Glut4 neuron ablation is expected to yield a negative energy balance.

It could be argued that any extensive ablation of basal hypothalamic neurons would impair eating as a result of tissue reaction, inflammation, and associated vascular changes. Therefore, as a control, we performed neuronal ablations in mice carrying a Cre-inducible DTR in POMC neurons (Pomc-Cre; Rosa-iDTR), a known effector population of anorexigenic signals. As expected, POMC neuron ablation caused opposite effects to Glut4 neuron ablation. POMC neuron–ablated mice exhibited normal VO₂ (Fig. 2B), increased RQ (Fig. 2D), reduced locomotor activity during the dark phase (Fig. 2F), but increased food intake throughout the light cycle (Fig. 2H). The change in food intake is consistent with previous POMC ablation experiments (11), as well as with impaired melanocortin signaling (12).

Compared with POMC neurons, Glut4 neurons constitute a much larger population, and their ablation may cause significant inflammation. To examine the contribution of the inflammatory response to the observed phenotype, we measured hypothalamic expression of mRNA encoding the astrocyte marker, Gfap, as a marker of gliosis. Gfap was significantly elevated after Glut4 neuron ablation (Supplementary Fig. 1A). Next, we measured other markers of inflammation, including interleukins (ILs) and Toll-like receptors (Tlrs). Levels of IL-1, IL-6, IL-12, and tumor necrosis factor-α showed a trend toward increase in samples from Glut4 neuron–ablated mice, but were exceedingly low (i.e., threshold cycle values between 33 and 39), raising the question of whether they are pathophysiologically significant (Supplementary Fig. 1B–E). In contrast, Tlr2 and Tlr4, for which threshold cycle values were in the usual range (∼25), were either unaltered or significantly reduced (Supplementary Fig. 1F and G). Therefore, the inflammatory response is mainly associated with gliosis.

To rule out that the observed changes of food intake were due to the acute gliosis, we followed mice for up to 2 months after ablation and characterized the metabolic consequences of removing Glut4 neurons. The anorectic effect of Glut4 neuron ablation leveled off, but persisted. Throughout the light cycle, Glut4 neuron–ablated mice continued to show significantly lower cumulative food intake (Fig. 3A). Quantitative analyses of the data showed a 23% reduction of food intake during the dark phase (Fig. 3B). Interestingly, Glut4 neuron–ablated mice showed increased O₂ consumption (Fig. 3C and D) and energy expenditure (Fig. 3I and J), which were not observed immediately after ablation. RQ (Fig. 3E and F) and locomotor activity (Fig. 3G and H) were comparable with control mice. Overall, the negative energy balance phenotype persisted throughout the duration of the study. Therefore, we conclude that Glut4 neurons promote positive energy homeostasis by regulating food intake and energy expenditure.

We further characterized the response to fasting and refeeding in Glut4 neuron– or POMC neuron–ablated mice. Glut4 neuron–ablated mice had 50% lower locomotion during fast, indicating reduced foraging (Fig. 4A). They showed lower rebound food intake (Fig. 4B). To test whether Glut4 neuron–ablated mice had defective responses to orexigenic neuropeptides, we administered intracerebroventricular NPY before and after Glut4 neuron ablation and found identical responses (Fig. 4C), indicating that animals can still respond to orexigenic cues. In contrast to the anorexigenic effect observed in Glut4 neuron–ablated mice, mice with POMC neuron ablation displayed hyperphagia (Fig. 4D), and increased body weight (Fig. 4E).

Glut4 neuron ablation causes persistent negative energy balance, while POMC neuron ablation causes positive energy balance. As a result, 2 months after neuronal ablation, Glut4 neuron–ablated mice showed a 26% decrease in body weight (Fig. 4F), while POMC neuron–ablated mice showed a 60% increase (Fig. 4H). The changes in body weight reflected changes in adiposity, as Glut4 neuron–ablated mice showed a 3.6-fold decrease of epididymal fat weight (Fig. 4G), while POMC neuron–ablated mice showed a 3.5-fold increase (Fig. 4I). Moreover, we assayed energy balance in a diet-induced obesity model, in which mice were fed HFD to induce weight gain. NCD-fed mice were injected with DTA and switched to HFD on day 0 (Fig. 4J). Although feeding on a more calorie dense diet blunted the weight loss in Glut4 neuron–ablated mice (Fig. 4J, ablated HFD vs. ablated NCD), these mice still failed to gain weight compared with controls on HFD (Fig. 4J, control HFD vs. ablated HFD). Therefore, we conclude that Glut4 neurons are orexigenic neurons that promote feeding.

We considered different possibilities to explain the remarkable anorexia associated with Glut4 neuron ablation. First, we evaluated ARH neuropeptides. We assessed the extent to which Glut4 neurons overlap with agouti-related protein (AgRP)/NPY or POMC neurons. We generated Glut4-Cre; Rosa26-tomato; Npy-Gfp mice and Glut4-Cre; Rosa26-tomato; Pomc-Gfp mice (Supplementary Fig. 2A). In both strains, Glut4 neurons are labeled red, whereas NPY or POMC neurons are labeled green. Quantitative analyses showed limited overlap of Glut4 neurons with either NPY (∼17%) or POMC neurons (∼13%) (Supplementary Fig. 2B). We also quantified the percentage of NPY- or POMC-positive hypothalamic Glut4 neurons by flow cytometry (Supplementary Fig. 2C–F, upper right quadrants). In the absence of Glut4-Cre, we detected no Tomato signal in NPY (Supplementary Figure 2C) or POMC neurons (Supplementary Fig. 2E). When Tomato expression was activated by Glut4-Cre, ∼18% of NPY neurons (Supplementary Figure 2D) and ∼23% of POMC neurons in the mediobasal hypothalamus (Supplementary Fig. 2F) were Glut4 positive.

We measured Npy/Agrp and Pomc mRNA in Glut4 neuron–ablated mice, but found only a nonsignificant trend toward increased Agrp/Npy (Supplementary Fig. 2G), consistent with the limited colocalization. Therefore, the anorexigenic phenotype observed in Glut4 neuron–ablated mice was not due to defects in NPY/AgRP and POMC neurons. Curiously, this neuropeptide expression pattern mirrors that seen after ablating orexigenic Rip_HER neurons (13), thereby providing further evidence that Glut4 neurons are orexigenic in nature.

Next, we examined other hypothalamic regions that control food intake and energy balance. The lateral hypothalamus (LH) regulates food intake and glucose metabolism (14). Melanin-concentrating hormone (MCH) is a potent orexigen that is produced from Pmch synthesized by LH neurons and acts through Mchr1/r2 receptors (15). Following toxin-mediated ablation of MCH neurons, mice become hypophagic, lean (16), and hyperactive (17). Using MCH immunohistochemistry in Glut4-Cre; Rosa-Gfp mice, we found that Glut4 and MCH colocalize in the LH (Fig. 5A). Moreover, immunohistochemistry demonstrated a sizable decrease in number and intensity of Glut4 staining following toxin-mediated ablation (Fig. 5B), indicating potential defects in the MCH system. In fact, Pmch expression was significantly reduced in hypothalamic samples after ablation (Fig. 5C). Thus, hypothalamic Glut4 neuron ablation affects MCH neuron number and decreases Pmch levels, providing a potential explanation for the hypophagia.

Recent work has implicated GABAergic transmission in orexigenic signaling (18). Transcriptome analyses showed that GABAergic signaling elements are enriched in Glut4 neurons (19). Following Glut4 neuron ablation, key genes required for GABAergic signaling, such as glutamate decarboxylase 1 (Gad1), GABA-A receptor γ1 (Gabrg1), and GABA vesicular transporter member 1 (vGat) were decreased (Fig. 5D), consistent with the possibility that GABAergic activity is reduced. To test the hypothesis that decreased GABAergic signaling underlies hypophagia, we infused the GABA receptor agonist bretazenil into the mediobasal hypothalamus after Glut4 neuron ablation. Before ablation, all treatment groups showed comparable food intake (data not shown). After ablation, the control group, regardless of whether they received saline or bretazenil, ate comparably to preablation. As expected, Glut4 neuron–ablated mice treated with saline had significantly reduced food intake (Fig. 5E). In contrast, the Glut4 neuron–ablated group that received bretazenil partly recovered food intake, even though it was not normalized to preablation levels. Saline or bretazenil infusion was continued for 14 days. Bretazenil infusion attenuated the weight loss caused by Glut4 neuron ablation (Fig. 5F).

Insulin signaling in the CNS affects systemic insulin sensitivity and glucose metabolism (20). We tested whether ablating Glut4 neurons alters peripheral metabolism. Glut4 neuron–ablated mice had significantly elevated fasting glucose and normal ad libitum glucose (Fig. 6A, Supplementary Fig. 3A and C). This is in contrast with elevated ad libitum and fasting glucose in POMC neuron–ablated mice (Supplementary Fig. 3B and D). Glut4 neuron–ablated mice had comparable glucagon levels (Fig. 6B), but significantly reduced insulin during fasting and refeeding (Fig. 6C). These findings suggest that the hyperglycemia results from insufficient insulin levels. Immunohistochemical and morphometric studies indicated that β- and α-cell distribution and mass were normal in Glut4 neuron–ablated mice (Supplementary Fig. 3E), suggesting that the reduced plasma insulin is likely to reflect insufficient insulin secretion rather than decreased cell number.

Ablating Glut4 neurons impaired glucose tolerance (Fig. 6D and E) and increased plasma glucose levels during pyruvate tolerance tests (Fig. 6F and G). As the complexity of the manipulations required for neuronal ablation prevented us from subjecting mice to clamps, we examined liver gene expression for surrogate markers of hepatic insulin resistance. After an overnight fast, glucose-6-phosphatase (G6pc) and phosphoenolpyruvate carboxykinase 1 (Pck1) increased by 2.4- and 1.8-fold, respectively (Fig. 6H). After refeeding, Glut4 neuron–ablated mice showed a 4.3-fold increase of hepatic G6pc (Fig. 6I), but comparable Pck1 expression (data not shown). Pyruvate dehydrogenase kinase 4 (Pdk4) controls the activity of pyruvate dehydrogenase in an insulin-inhibitable manner, limiting conversion of pyruvate to acetyl-CoA. Glut4 neuron–ablated mice had significantly increased hepatic Pdk4 expression after refeeding, consistent with reduced plasma insulin levels (Fig. 6I). Curiously, expression of sterol regulatory element-binding protein (Srebp1c) was nearly fourfold higher under refed, but not fasted conditions (Fig. 6H and I) and was associated with increased hepatic triglyceride content in Glut4 neuron–ablated mice (Fig. 6J). The increased hepatic lipid content in Glut4 neuron–ablated mice may be caused by increased de novo lipogenesis through sterol regulatory element-binding protein and/or increased lipolysis from adipose tissue, due to lower insulin levels.

To understand the mechanism of defective pancreatic endocrine function in Glut4-ablated mice, we sought to determine their efferent pathways using neuron-tracing studies. We injected adenoviral Ad-iZ-EGfpf in the basal hypothalamus of Glut4-Cre mice. This virus allows for Cre-dependent expression of membrane-tethered, farnesylated Gfp (21). We traced Gfp-labeled projections to two main sites, the vagal dorsal motor nucleus in the hindbrain and the hypothalamic paraventricular nucleus (PVN) (Fig. 7A). Vagal dorsal motor nucleus is critical for the regulation of pancreatic secretion, as well as sympathetic and parasympathetic afferents (22,23). Thus, Glut4 neuron ablation causes reduced insulin secretion via impaired vagal efferent to the pancreas. Moreover, we examined c-fos expression as readout of neuronal activation after Glut4 neuron ablation. The c-fos immunoreactivity localized to the PVN, dorsomedial hypothalamus (DMH), and, most prominently, anterior paraventricular thalamic nucleus (Fig. 7B). PVN and DMH relay feeding and metabolic cues from the ARH (13,24,25). The findings are consistent with the possibility that Glut4 neurons are integrated with critical hypothalamic nuclei that regulate food intake and hindbrain nuclei that regulate insulin secretion.

The changes in energy homeostasis and peripheral metabolism following Glut4 neuron ablation suggest that Glut4 neurons mediate CNS actions of metabolic hormones. To test this hypothesis, we explored hormone signaling following Glut4 neuron ablation. First, we measured pAkt after acute ablation to correlate metabolic phenotypes (Fig. 2) with insulin pathway. Western blots showed that pAkt increased in the hypothalamus of Glut4 neuron–ablated mice after fasting (Supplementary Fig. 4A) and refeeding (Supplementary Fig. 4B). The increased pAkt in mice with Glut4 neuron ablation could represent an acute response to inflammation. We did immunohistochemistry to colocalize pAkt and Gfap (Supplementary Fig. 4C). We found that increased pAkt could only be partially attributed to glial cells (Gfap-positive cells). The acute increase of pAkt in other cells of Glut4 neuron–ablated mice may be a result of compensation or loss of inhibitory tone from Glut4 neurons.

Next, we measured hypothalamic pAkt 2 months after ablation. We found that pAkt immunoreactivity decreased in the basal hypothalamus of Glut4 neuron–ablated mice compared with WT (Fig. 8A, left panel), indicating impaired insulin signaling. Leptin signaling, assessed from the number of pStat3-immunoreactive cells (Fig. 8A and B) as well as their individual intensity (Fig. 8C), increased in the basal hypothalamus of Glut4 neuron–ablated mice, despite a significant decrease of fasting and fed plasma leptin (Fig. 8D). Glut4 neuron–ablated mice have lower body weight and less adiposity, which can be associated with reduced circulating leptin. However, the change of body weight exceeds the reduction of leptin. Therefore, reduced leptin cannot only be attributed to the reduced body weight, but rather indicates that Glut4 neuron–ablated mice are more leptin sensitive.

We also used pS6 to assess mTOR signaling. We found that generation of pS6 in response to feeding in the hypothalamus of Glut4 neuron–ablated mice was comparable to controls (Fig. 8A, right panel). Based on these results, we conclude that disrupting basal hypothalamic Glut4 neurons impairs insulin signaling, but increases leptin signaling and sensitivity.

In this study, we sought to define the metabolic functions of basal hypothalamic Glut4 neurons by targeted cell ablation and metabolic phenotyping in adult mice. Combined evidence from these investigations is consistent with a role of this cell population in food intake and glucose metabolism. Key findings of this work are that: 1) basal hypothalamic Glut4 neurons regulate satiety and glucose homeostasis; 2) Glut4 neurons regulate energy homeostasis by promoting food intake and reducing energy expenditure; 3) Glut4 neurons regulate hepatic glucose and lipid metabolism; and 4) Glut4 neurons regulate insulin secretion and β-cell function, possibly via direct projection to hindbrain. When viewed in the context of previous studies with Glut4-promoter driven insulin receptor knockout (GIRKO) mice (4), the impairment of pAkt immunoreactivity in the current study supports the conclusion that ablating Glut4 neurons causes a defect in the insulin signaling pathway. Qualitative and quantitative changes in pStat3 immunohistochemistry and decreases in plasma leptin levels after ablation indicate that Glut4 neurons normally dampen leptin signaling, emphasizing their integrative role in energy homeostasis. Increased leptin signaling in Glut4 neuron–ablated mice may represent a compensatory response to reduced insulin signaling, consistent with the notion that leptin administration can rescue insulin-deficient diabetes (26). The findings expand our understanding of the neuroanatomy and cellular physiology of insulin action in the CNS and integrate the latter into the broader picture of diabetes pathogenesis.

We detected increased gliosis shortly after Glut4 neuron ablation. Yet, we think that this is unlikely a cause of anorexia. First, the role of inflammation on food intake is complex. Lipopolysaccharide-induced inflammation stimulates the release of cytokines and promotes anorexia (27). In contrast, increased caloric intake associated with an HFD is also associated with hypothalamic inflammation (28). Second, the negative energy balance following Glut4 neuron ablation persisted for 2 months after DTA injection, as did all of the energy balance phenotypes examined in our study. Third, the acute changes in hypothalamic pAkt content are consistent with an anti-inflammatory response. Fourth, Glut4 neuron–ablated mice show reduced weight gain on HFD, which is consistent with negative energy balance.

The anorectic phenotype linked to Glut4 neuron ablation in the hypothalamus is striking. It bears more than a passing resemblance with that of mice lacking Rip_HER neurons, suggesting that the latter partly overlap with the former (13). With regard to the mechanism of this observation, we provide three clues: 1) Glut4 neurons project to PVN and DMH, two relay sites for feeding cues from the ARH (13,24,25); 2) Glut4 neurons partly overlap with MCH neurons in the LH, and MCH neuron number and Pmch levels fall after Glut4 neuron ablation, potentially contributing to anorexia (29); and 3) finally, the anorexia can be rescued by GABA supplementation, consistent with an important role of glutamatergic input onto hypothalamic neurons to regulate eating (30–32), as well as with the reversal of starvation caused by NPY neuron ablation (33) by delivery of GABA-A receptor agonists to the hindbrain (18).

Notably, the effects of glutamate and GABA extend to the LH (34). GABA can promote weight gain when released from leptin-inhibited neurons (35). Deleting synaptic release of GABA from Rip_HER neurons increases body weight by decreasing energy expenditure and brown adipose tissue thermogenesis, though not food intake (25). These findings provide more circumstantial evidence that Glut4 neurons feed into these neural circuits.

Another key finding of this study is that ablating Glut4 neurons increases hepatic glucose production and decreases insulin levels, suggesting that this neuron population integrates various aspects of metabolic physiology. These observations are consistent with the fact that GIRKO mice (lacking InsR in muscle, fat, and Glut4 neurons) develop diabetes with profound hepatic insulin resistance and β-cell dysfunction (4).

There is a functional link between CNS and pancreatic islet secretion (36). We found that Glut4 neurons are: 1) enriched in components of the K_ATP channel (19), a central regulator of pancreatic hormone release (37); and 2) project to the vagal DMN, a critical relay station for the CNS–endocrine pancreas axis (38). The alterations of hepatic gene expression and lipid content in Glut4 neuron–ablated mice add to a growing body of evidence linking hypothalamic hormone and nutrient sensing to insulin sensitivity in the liver (39). Whether this process applies to larger mammals remains to be determined (40).

As the medical scientific community scrambles to find new approaches to treat insulin resistance, the fundamental abnormality of type 2 diabetes, the present findings add to a growing catalog of evidence suggesting that the CNS be considered a target organ of insulin sensitizers. In this regard, the identification of a discrete, if heterogeneous, “insulin-sensitive” neuronal population represents a necessary step in developing treatments that meet appropriate standards of safety and efficacy.

Acknowledgments. The authors thank C. Rhodes (University of Chicago) and M.G. Myers (University of Michigan) for providing Ad-iZ-EGfpf adenovirus, X. Xu and A.W. Ferrante (Columbia University) for critical discussion and technical help with assaying inflammation response, and members of the Accili laboratory for critical discussion of the data.

Funding. H.R. is the recipient of a mentor-based postdoctoral fellowship from the American Diabetes Association and the Naomi Berrie Diabetes Research Fellowship. This work is supported by National Institutes of Healthhttp://dx.doi.org/10.13039/100000002 grants K99-DK-098294 (to H.R.), 5R37-DK-058282 (to D.A.), and DK-63608 (to Columbia University Diabetes Research Center).

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

Author Contributions. H.R. designed and conducted experiments, analyzed data, and wrote the manuscript. T.Y.L. conducted experiments. T.E.M. provided research tools. D.A. designed experiments, analyzed data, and wrote the manuscript. D.A. 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.

Prior Presentation. Parts of the study were presented in abstract form at the 71st Scientific Sessions of the American Diabetes Association, San Diego, CA, 24–28 June 2011.