Leptin signaling in GFAP-expressing adult glia cells regulates hypothalamic neuronal circuits and feeding

We have shown that synaptic re-organization of hypothalamic feeding circuits in response to metabolic shifts involves astrocytes, cells that can directly respond to the metabolic hormone, leptin, in vitro. It is not known whether the role of glia cells in hypothalamic synaptic adaptions is active or passive. Here we show that leptin receptors are expressed in hypothalamic astrocytes and that conditional, adult deletion of leptin receptors in astrocytes leads to altered glial morphology, decreased glial coverage and elevated synaptic inputs onto pro-opiomelanocortin (POMC)and Agouti-related protein (AgRP)-producing neurons. Leptin-induced suppression of feeding was diminished, while rebound feeding after fasting or ghrelin administration was elevated in mice with astrocyte-specific leptin receptor deficiency. These data unmask an active role of glial cells in the initiation of hypothalamic synaptic plasticity and neuroendocrine control of feeding by leptin. Contributions: J.G.K., M.O.D. and T.L.H. designed the study. J.G.K., M.H.T. and T.L.H. interpreted the results. J.G.K. and S.J. performed experiments and analyzed the data. M.K. and K.S.B. contributed to Figure 1h-j and 2a-c. J.K.J. and S.D. contributed to Figure 1c. S.S. and Z.L. contributed to Figure 2 d-g and Supplementary Figure 6. M.R.Z., N.S. and F.M.V. contributed to Supplementary Figure 1 and 4a. P.A., J.C. and J.A. contributed to Supplementary Figure 3b. Y.G., C.G. and C.Y. contributed to the generation of animal model. J.G.K and T.L.H. wrote the paper with input from the other authors. NIH Public Access Author Manuscript Nat Neurosci. Author manuscript; available in PMC 2014 July 28. Published in final edited form as: Nat Neurosci. 2014 July ; 17(7): 908–910. doi:10.1038/nn.3725. N IH -P A A uhor M anscript N IH -P A A uhor M anscript N IH -P A A uhor M anscript Astrocytes are the most abundant cells in the central nervous system (CNS), yet at times, they have been relegated a less than prominent role in the control of complex brain functions supported by neuronal circuits1,2. The regulation of food intake and energy expenditure is tightly linked to synaptic plasticity of hypothalamic neural circuits3,4, processes in which glial cells have also been implicated5,6. It has not yet been explored whether this involvement of glia is secondary or plays an active role in the promotion of these processes initiated by leptin7. As previously reported, the long form of leptin receptors (LepR) was found to be located in astrocytes through the use of immunocytochemistry8,9. However, due to questions regarding antibody specificity, it still remains controversial whether astrocytes express functional LepR. We found immunolabeling of glial fibrillary acidic protein (GFAP) in a subset of leptin receptor (LepR)-driven EGFP-expressing cells (Fig. 1a). Second, LepR mRNA was detected from translating ribosomes of astrocytes (Supplementary Fig. 1). Third, mRNA of LepR was expressed in purified mouse hypothalamic astrocytes using astrocyte primary culture (Supplementary Fig. 3b). To test the role of the long form of leptin receptors in glial cells, we generated a genetic mouse model in which leptin receptors are time-specifically ablated in astrocytes. Because glial cells are the progenitor cells for neurogenesis during brain development10, we used a tamoxifen-inducible Cre-ERT2 system to allow cell and time-specific knockout of leptin receptor in adult astrocytes (Supplementary Fig. 3a). To assess whether functional Cre protein was restricted to astrocytes and induced by tamoxifen injection, we crossed GFAPCreERT2 mice with tdTomato-loxP reporter mice, which express red fluorescent protein. We confirmed successful Cre-mediated recombination in GFAP-positive cells by detecting tdTomato-positive cells after injection of tamoxifen (Supplementary Fig. 2c). This recombination was found to be specific to astrocytes as the tdTomato-positive cells did not express Iba-1 (a marker for microglia) or NeuN (a marker for neurons) (Supplementary Fig. 2c). In addition, we combined in situ hybridization (ISH) with immunohistochemistry (IHC) to validate the selective loss of functional leptin receptors from GFAP-positive cells in mice that are GFAP-Cre transgenic and homozygous for the loxP-modified leptin receptor allele (Fig. 1b,c). We further confirmed the deletion of leptin receptor exon 17 in astrocyte primary cells of GFAP-LepR−/− mice by reverse transcriptase (RT)-polymerase chain reaction (PCR) (Supplementary Fig. 3b). Because of our previous findings that leptin affects glial morphology,6,11 we first analyzed astrocytes in the arcuate nucleus of mice following leptin receptor knockout. Astrocytespecific loss of leptin receptors did not alter the total number of GFAP-positive cells in the hypothalamus (Fig. 1e). However, GFAP-LepR−/− mice showed fewer numbers (Fig. 1f) and shorter lengths (Fig. 1g) of primary astrocytic projections. We also analyzed astrocytes in the hippocampus. Interestingly, we could detect LepR mRNA in the hippocampus (Supplementary Fig. 4a), but there were no significant changes regarding number and morphology of GFAP-positive cells (Supplementary Fig. 4b–e). Previously, we reported that astrocytic processes are involved in synaptic plasticity of feeding circuits, including those comprising the proopiomelanocortin (POMC) neurons that Kim1 et al. Page 2 Nat Neurosci. Author manuscript; available in PMC 2014 July 28. N IH -P A A uhor M anscript N IH -P A A uhor M anscript N IH -P A A uhor M anscript secrete α-melanocyte stimulating hormone (α-MSH) and AgRP (agouti-related protein) neurons that coproduce neuropeptide Y (NPY) and γ-amino-butyric acid (GABA)5, 6. This led us to evaluate the patterns of glial ensheathment onto the perikaryal membranes of POMC and unlabeled neurons in the arcuate nucleus by electron microscopy (EM). GFAPLepR−/− mice had lower glial coverage on the perikaryal membranes of POMC (Fig. 1i) and unlabeled neurons (Fig. 1j) compared to that of control mice. We then analyzed glial coverage of POMC and AgRP cells of GFAP-LepR−/− mice through the use of double immunofluorescence: GFAP immonolabeled with red fluorescence in association with green fluorescent protein (GFP)-labeled POMC or AgRP neurons (Npy-hrGFP mice were used for the latter; these mice allow visualization of AgRP neurons due to the co-expression of NPY and AgRP in these cells). We found that direct contacts were lower between astrocytes and either POMC (Supplementary Fig. 5a,b) or AgRP (Supplementary Fig. 5c,d) neurons in GFAP-LepR−/− mice relative to control values. Next, we assessed whether reduced astrocyte coverage affects synapse number on arcuate nucleus neurons. First, we analyzed synapse number and type by EM. We found that there were elevated numbers of both symmetric and asymmetric synapses on both POMC (Fig. 2b) and unlabeled neuronal perikarya (Fig. 2c) in GFAP-LepR−/− mice relative to controls. To corroborate these anatomical findings, miniature postsynaptic currents (mPSCs) onto POMC and AgRP neurons were analyzed. We found an elevated frequency of miniature inhibitory postsynaptic currents (mIPSCs) (Fig. 2d) but no change in frequency of miniature excitatory postsynaptic currents (mEPSCs) onto POMC cells (Fig. 2e). AgRP neurons had an increase in the frequency of both mIPSCs and mEPSCs (Fig. 2f,g). Taken together, these data show that leptin receptor signaling in astrocytes regulates the synaptic input organization of AgRP and POMC cells. We also revealed an increased amplitude of both mIPSCs and mEPSCs onto the POMC neurons of GFAP-LepR−/− mice (Supplementary Fig. 6c,d). On the other hand, there was no alteration in the amplitude of mPSCs onto the AgRP neurons (Supplementary Fig. 6a,b). These findings suggest that the reduced astrocyte coverage may affect the signaling pathways linked to the postsynaptic receptors of POMC neurons presumably by buffering trophic factors in the respective synaptic cleft area. We have shown that the synaptic input organization of the melanocortin system predicts behavioral output of the melanocortin system in the face of a changing metabolic milieu3, 5, 12. Thus, we next assessed the metabolic phenotype, including body weight, body composition, food intake and energy expenditure, of GFAP-LepR−/− mice and their littermate controls. Three-month-old GFAP-LepR−/− mice showed no differences in metabolic phenotypes under standard feeding conditions (Supplementary Fig. 7a–h). However, the effect of both single and multiple injections of leptin to suppress feeding were diminished in GFAP-LepR−/− mice relative to controls (Fig. 3a,b). Consistent with these results, leptin-stimulated Fos activity was attenuated in the GFAP-LepR−/− mice (Fig. 3c,d). These findings are in line with the observed increase in the number of inhibitory inputs onto the POMC neurons in these mice, as previously it was shown that leptin exerts its effect on POMC neurons, at least in part, by the suppression of their inhibitory inputs13,14. The effect of the selective knockout of LepR in astrocytes on mIPSCs (but not on mEPSCs) recapitulated the effects of leptin we observed in Lepob/ob mice regarding miniature events Kim1 et al. Page 3 Nat Neurosci. Author manuscript; available in PMC 2014 July 28. N IH -P A A uhor M anscript N IH -P A A uhor M anscript N IH -P A A uhor M anscript on the POMC neurons11. On the other hand, while the morphological observations of the current study mimicked the electrophysiological findings regarding IPSCs seen here and in our earlier work11, the lack of a measurable effect of leptin on mEPSCs of POMC cells was not reflected in morphological alterations regarding putative excitatory inputs. These discrepancies may be due to the fact that leptin signaling is more broadly impacted in Lepob/ob mice, but it also highlights the idea that the interrogation of circuit integrity and function cannot be reliably asserted by a single approach. Next, we determined the responses of GFAP-LepR−/− mice to fasting or ghrelin, a gut hormone that is elevated during negative energy balance and promotes feeding behavior15,16. Fasting-induced hyperphagia was significantly enhanced in these mice compared to controls (Fig 3e,h). They also showed elevated ghrelin-induced food intake (Fig. 3f). Aligned with these findings, AgRP neurons of GFAP-LepR−/− mice exhibited an increased number of Fos expressing nuclei in response to fasting compared to controls (Fig. 3g,h). These observations are consistent with the findings that food deprivation or ghrelin administration elevates AgRP neuronal activity, at least in part, by mediation of presynaptic excitatory inputs, which are now revealed to be controlled by astrocytes. Collectively, our data show that leptin receptor signaling in astrocytes plays a previously underappreciated active role at the arcuate nucleus interface between afferent hormones, hypothalamic synaptic adaptations and strength and CNS control of feeding. To what extent these processes may be involved in the development of obesity in response to overnutrition and the identity of the intercellular signaling modalities that enables glial cells to alter synaptology need to be determined.