Hypothalamic PKA regulates leptin sensitivity and adiposity

Mice lacking the RIIβ regulatory subunit of cyclic AMP-dependent protein kinase A (PKA) display reduced adiposity and resistance to diet-induced obesity. Here we show that RIIβ knockout (KO) mice have enhanced sensitivity to leptin's effects on both feeding and energy metabolism. After administration of a low dose of leptin, the duration of hypothalamic JAK/STAT3 signalling is increased, resulting in enhanced POMC mRNA induction. Consistent with the extended JAK/STAT3 activation, we find that the negative feedback regulator of leptin receptor signalling, Socs3, is inhibited in the hypothalamus of RIIβ KO mice. During fasting, RIIβ–PKA is activated and this correlates with an increase in CREB phosphorylation. The increase in CREB phosphorylation is absent in the fasted RIIβ KO hypothalamus. Selective inhibition of PKA activity in AgRP neurons partially recapitulates the leanness and resistance to diet-induced obesity of RIIβ KO mice. Our findings suggest that RIIβ–PKA modulates the duration of leptin receptor signalling and therefore the magnitude of the catabolic response to leptin.

T he adipose-derived hormone, leptin, regulates energy balance by binding to receptors in the hypothalamus and regulating neural circuits that suppress feeding and increase energy expenditure [1][2][3][4] . Plasma leptin concentration is directly correlated with the level of stored triglyceride in white adipocytes providing a negative feedback signal to the brain to help maintain body weight homeostasis. However, the chronic consumption of calories in excess of those required for daily energy expenditure leads to elevated adiposity, a proportionate increase in circulating leptin, and a decrease in the response of neurons to leptin 5 . This hypothalamic leptin resistance has limited the usefulness of leptin as a treatment for obesity. The leptin receptor (LepRb) signals primarily through stimulation of the associated JAK2 kinase which then leads to phosphorylation of STAT3 transcription factors and changes in gene transcription 6 . In addition, JAK2 triggers the PI3K pathway leading to multiple effects including activation of Akt and subsequent phosphorylation and inactivation of the transcription factor, FoxO1 (ref. 7). The intracellular mechanisms which regulate the overall sensitivity of neurons to circulating leptin are poorly understood. The protein kinase A (PKA) regulatory (R) subunit, RIIb, is highly expressed in mouse brain, brown adipose tissue and white adipose tissue with limited expression elsewhere [8][9][10][11] . RIIb knockout (KO) mice exhibit a 50% reduction in white adipose tissue, a four to fivefold reduction in serum leptin, and are resistant to diet-induced obesity and diabetes 10,[12][13][14] . Previous studies have shown that deficiency of RIIb-PKA in GABAergic hypothalamic neurons leads to the lean phenotype 15 but it remains unclear what the intracellular signalling events are that account for this phenotype. Although reduced serum leptin is a strong anabolic signal to promote feeding and suppress energy expenditure 16 , RIIb KO mice exhibit only a slight increase in food intake 12,14 and a normal basal metabolic rate 15 in the context of greatly reduced serum leptin. Based on these observations, we hypothesize that RIIb-PKA may regulate energy balance by modulating leptin signalling in the hypothalamus and its deficiency may lead to sensitized responses to leptin.
PKA holoenzyme is a heterotetramer containing a dimer of two R subunits with each binding a catalytic (C) subunit. Binding of cyclic AMP (cAMP) to the R subunits leads to its conformational change and release of active C subunits. Four R subunits genes (encoding RIa, RIb, RIIa and RIIb) and two C subunits genes (encoding Ca and Cb) have been identified in mouse 17 . The R subunits act as intrinsic inhibitors of the C subunits and also protect the C subunits from degradation 18 . The RIIb-PKA holoenzyme contains a homodimer of RIIb and two C subunits (Ca or Cb). We have shown previously that RIIb deficiency leads to decreased C subunits and PKA activity in adipose tissues 10,11 and striatum 8,19 .
In this report, we show that RIIb-PKA regulates the sensitivity of mice to the catabolic effects of leptin on feeding and energy expenditure. We demonstrate that RIIb-PKA is abundant in LepRb-expressing neurons in the hypothalamus and becomes activated during a fast. Inhibition of PKA by expression of a dominant negative Prkar1a allele in AgRP neurons inhibited hypothalamic CREB phosphorylation in fasted mice and resulted in a lean phenotype and resistance to high-fat diet-induced obesity, partially mimicking the phenotype of the RIIb KO mouse. We suggest that the cAMP/PKA pathway plays an important physiological role in modulating the gain of LepR signalling in the hypothalamus and that this results in significant effects on overall adiposity.

Results
Disruption of RIIb results in increased sensitivity to leptin. The mRNA levels of leptin-regulated orexigenic peptides, NPY and AgRP, and the anorexigenic aMSH precursor POMC in the hypothalamus are similar between RIIb KO and wild-type (WT) control mice in both fed and fasted states despite the very low leptin levels (Fig. 1a,b). This indicates that RIIb KO mice have an intact leptin-responsive system in the hypothalamus but suggests that they might be hypersensitive to leptin.
To test leptin sensitivity, we injected low or high doses of leptin directly into the third ventricle of 6-8-week-old male RIIb KO mice and WT controls. By measuring 24-h food intake, we determined that RIIb KO and WT littermates respond equivalently to a single injection of a high dose of leptin (500 ng). However, when the leptin dose was reduced to 100 ng, there was a greater reduction of 24-h food intake in KO mice compared with WT mice (Fig. 1c). To further examine the effect of central leptin administration on body weight loss, leptin at a dose of 100 ng was injected through a cannula into the third ventricle of KO and WT mice once a day for 7 days. Food intake was suppressed to a much greater extent in KO mice compared with control WT mice (Fig. 1d). As expected from the food intake data, this low dose of leptin caused only a slight decrease in body weight in WT mice but a very significant loss of body weight in the KO mice which was rapidly reversed within 48 h after leptin was discontinued (Fig. 1e). Energy expenditure, as measured by oxygen consumption, was increased by 410% in KO mice but not in control WT mice after 100 ng leptin infusion into the third ventricle (Fig. 1f). We considered the possibility that the increased sensitivity to leptin in the adult KO mice might simply reflect their leanness at a time when the age-matched WT mice are beginning to gain fat mass and develop leptin resistance. To test this possibility, we examined weight and adiposity-matched WT and KO mice at a young age (Fig. 1g, j). Leptin induced greater suppression of food intake (Fig. 1h) and loss of body weight (Fig. 1i) in adipositymatched RIIb KO mice compared with WT mice, indicating that the elevated leptin sensitivity of RIIb KO mice is not a secondary effect of their lean phenotype.
PKA regulates the duration of JAK/STAT signalling by LepRb. Leptin activates both the JAK2-STAT3 pathway and the phosphatidylinositol 3-kinase (PI3K)-Akt pathway in the hypothalamus to regulate energy balance 5 . STAT3 tyrosine phosphorylation (pSTAT3-Y705) is a downstream effect of LepRb activation of JAK2, phosphorylation of the LepRb on Y1138, and recruitment of STAT3 to the LepRb/JAK2 complex. pSTAT3-Y705 was induced to similar extents in KO and WT mice 1 h after low leptin (100 ng) administration to overnight fasted animals. However, the pSTAT3 signal was sustained at least 6 h longer in KO mice compared with WT mice (Fig. 2a-c). At 6 h after 100 ng leptin administration, pSTAT3-positive neurons were much more abundant in KO hypothalamus (arcuate nucleus (ARC), ventromedial hypothalamic nucleus (VMH) and dorsomedial hypothalamic nucleus (DMH)) compared with WT (Fig. 2a). Interestingly, we also observed a small but significant increase in basal hypothalamic pSTAT3 signal in overnight fasted RIIb KO mice compared with WT control (Fig. 2a,c inset). Consistent with the prolonged leptin-induced pSTAT3 signal, the level of POMC mRNA induced by low dose leptin administration was significantly greater in KO mice than in WT mice 12 h after leptin injection (Fig. 2d).
Socs3 acts as a feedback inhibitor of the JAK/STAT pathway by inhibiting JAK2 activity 5 . Hypothalamic Socs3 mRNA levels were comparable between RIIb KO and WT mice after an overnight fast, but were significantly lower in fed RIIb KO mice compared with WT control (Fig. 2e). The induction of Socs3 mRNA by low dose leptin was also attenuated in RIIb KO mice (Fig. 2e). This indicates that RIIb-PKA deficiency may impair hypothalamic Socs3 expression and prevent the feedback inhibition that normally limits the duration of leptin signalling.
Leptin activation of PI3K/Akt pathway is regulated by PKA. Leptin also activates the PI3K/Akt pathway in several regions of the hypothalamus including the ARC triggering Akt-dependent phosphorylation and inhibition of the transcription factor, FoxO1. FoxO1 binds to the Pomc promoter and antagonizes the action of STAT3 leading us to ask whether this leptin receptor signalling pathway was also enhanced in RIIb KO mice. Phosphorylation of FoxO1 by Akt triggers nuclear exclusion and proteosomal degradation of phospho-FoxO1 (refs 20,21). To explore the effect of low leptin on ARC FoxO1 localization we administered leptin to fasted animals at 100 ng per mouse, i.c.v. and examined FoxO1 by immunohistochemistry. In fasted animals, FoxO1 was predominantly nuclear in the ARC in both WT and KO. Three hours after leptin treatment, FoxO1 remained nuclear in WT ARC neurons but became diffusely associated with the cytoplasm in many of the ARC neurons of RIIb KO mice (Fig. 2f). We measured hypothalamic FoxO1 mRNA and protein levels and found that both were significantly decreased in the hypothalamus of RIIb KO mice compared with WT controls (Fig. 3a-c). A higher basal level of activated Akt (phospho-Akt (T 308 )) was observed in fasted RIIb KO hypothalamus compared with WT control and after leptin stimulation hypothalamic pAkt was markedly increased in both WT and RIIb KO mice as expected (Fig. 3b,c). Taken together, these data indicate that RIIb KO mice have an increased response to leptin through both the JAK2/ STAT3 and Akt/FoxO1 signalling pathways in the hypothalamus.
Regulation of LepRb signalling by PKA is cell autonomous. The increased duration of STAT3 and FoxO1 signalling in RIIb KO hypothalamus after leptin administration is occurring in multiple regions of the hypothalamus including the ARC, DMH, lateral hypothalamus (LH) and VMH. We used two approaches to ask whether the RIIb-PKA regulation of leptin sensitivity is cell autonomous. We injected AAV-Cre into one side of the ventral hypothalamus of RIIb lox/lox mice to activate RIIb expression ( Fig. 4a and Supplementary Fig. 1a,b). STAT3 phosphorylation in the ARC was similar between the AAV-Creinjected side and the saline-injected control side at 1 h after leptin injection (Fig. 4b) and many cells showed co-expression of Cre and pSTAT3 ( Supplementary Fig. 1a), indicating that Cre expression and Cre-induced RIIb re-expression did not affect the acute response to leptin. However, at 4 h after leptin injection, STAT3 phosphorylation was decreased in Cre-expressing cells but still robustly present in cells without Cre expression ( Fig. 4a,b). As a second approach, we activated RIIb re-expression in all GABAergic neurons by crossing the RIIb lox/ À mice to a Vgat-Cre transgenic line 22 , which activated RIIb expression in multiple hypothalamic nuclei except the primarily glutamatergic paraventricular nucleus (PVN), VMH 15 and non-GABAergic neuronal populations in the ARC ( Supplementary Fig. 1c). As shown previously, the Vgat-Cre/RIIb lox/ À (RIIb Vgat ) mice are rescued to WT adiposity and leptin levels 15 . In the ARC, DMH and LH of RIIb Vgat mice, RIIb was re-expressed in most leptinresponsive neurons as indicated by co-localization with leptininduced pSTAT3 expression ( Supplementary Fig. 1d 4 h after leptin injection, most of the pSTAT3 signals disappeared in the DMH and LH but remained high in the VMH where RIIb was not re-expressed (Fig. 4c,d). In the ARC, the number of pSTAT3-positive cells was significantly decreased at 4 h after leptin but there were still a significant number of pSTAT3positive cells (Fig. 4d) but many of them were negative for RIIb staining and therefore not expressing Vgat-Cre ( Supplementary  Fig. 1e). Our data demonstrates that RIIb À PKA negatively regulates the duration of leptin-induced pSTAT3 signalling in hypothalamic neurons and that the enhanced leptin sensitivity in RIIb KO neurons is cell autonomous and likely the cause rather than the effect of the lean phenotype.
RIIb-PKA is activated during a fast. To determine if RIIb-PKA is physiologically responding to nutritional signals in leptin receptor-expressing neurons as might be expected if it is regulating leptin sensitivity, we examined hypothalamic RIIb localization and activation in fed and fasted WT mice. After a 1-h leptin treatment, pSTAT3-positive neurons were shown by  immunohistochemistry to be co-expressing RIIb (Fig. 5a). It has been suggested that RIIb autophosphorylation at Ser114 occurs in the inactive PKA holoenzyme but does not lead to dissociation of the R/C complex 23 . Consistent with this observation, we found that activation of PKA in hypothalamic extracts by the cAMP analogue, 8-Br-cAMP, resulted in dephosphorylation of P-Ser 114 on RIIb and this was blocked by phosphatase inhibitors (Supplementary Fig. 2a). This allows us to directly monitor the nutritional activity of RIIb-PKA by western blot and immunohistochemistry with pRIIb S114 indicating inactive PKA and loss of pRIIb S114 as an indication of cAMP activation. As shown in Fig. 5b, a 24-h fast leads to the depletion of pRIIb S114 while 2-h re-feeding greatly increases pRIIb S114 , suggesting that hypothalamic RIIb-PKA is activated during fasting and inactivated after re-feeding. The fast caused an increase in pCREB S133 compared with fed or re-fed states (Fig. 5b)   consistent with previous findings 24 . Immunostaining of pRIIb S114 showed that it was increased in multiple hypothalamic nuclei including the ARC, VMH and DMH in re-fed mice compared with fasted mice (Fig. 5c). As an internal control, the level of pRIIb S114 in the cortex is not changed during fasting and re-feeding (Fig. 5c). We also tested whether leptin could substitute for re-feeding and reverse the activation state of RIIb-PKA; we were surprised to find that a 2-h treatment with leptin mimicked re-feeding and caused an increase in pRIIb S114 and returned pCREB to fed levels (Fig. 5b). The specificity of anti-RIIb and anti-pRIIb antibodies for immunohistochemical staining was verified by staining of brain sections from RIIb KO mice (Supplementary Fig. 2b,c). We conclude that the RIIb-PKA holoenzyme is actively expressed in leptin-responsive neurons and is activated by fasting and inhibited by either re-feeding or leptin treatment.
RIIb-PKA regulates CREB phosphorylation during fasting. The level of pCREB was greatly reduced in the RIIb KO compared with age-matched WT under either fasting or re-fed states as determined by western blot (Fig. 5d) and immuohistochemistry (Fig. 5e). To show that this effect is independent of body fat content, we examined hypothalamic pCREB levels using WT (6-week old, body weight: 20.9 ± 0.51 g) and RIIb KO (12-week old, body weight: 23±0.5 g) mice that had similar fat content as indicated by gonadal fat pads weight. As shown in Supplementary Fig. 2d, pCREB levels were dramatically reduced in RIIb KO mice in either fed, fasted or re-fed state compared with fat-matched WT control. In RIIb Vgat mice, which had similar fat content and blood leptin level as age-matched WT mice 15 , pCREB expression in response to fasting was decreased in the VMH where the Vgat-Cre is not expressed and therefore RIIb is not re-expressed. In contrast, ARC neurons exhibited both RIIb and pCREB staining in fasted RIIb Vgat mice ( Supplementary  Fig. 2e). The co-localization of RIIb and pCREB in hypothalamic neurons was shown in the ARC and VMH of fasted WT mice ( Supplementary Fig. 2f). These results indicate that the phosphorylation of CREB in response to fasting is dependent on RIIb-PKA in a cell-autonomous manner and that pCREB signalling in VMH and other non-GABAergic neurons does not play a major role in the lean phenotype of RIIb KO mice. Consistent with the greatly attenuated pCREB expression, the protein levels of PKA C subunits (including Ca, Cb1 and Cb2) ( Supplementary Fig. 2g) and total PKA activity ( Supplementary  Fig. 2h) were significantly decreased in the hypothalamus of RIIb KO mice, indicating that RIIb is one of the major PKA isoforms in mouse hypothalamus.
Inhibition of PKA in AgRP neurons causes leanness. AgRP neurons in the arcuate nucleus represent only a fraction of the GABAergic neurons in the hypothalamus but they play an essential role in the regulation of energy balance in adult mice 22,25 . We next asked whether impaired PKA signalling in AgRP neurons would have an effect on body composition. We generated mice with selective expression of a dominant negative PKA subunit allele (RIaB) in AgRP neurons. The dominant negative RIaB allele was generated as a knock-in mutation in Prkar1a that was silenced by a lox-flanked intragenic neo-stop sequence 26,27 . Expression of RIaB protein in AgRP neurons was initiated by Cre recombinase-dependent excision of the neo-stop sequence in Agrp-CreEGFP-expressing mice 28 . These Agrp-CreEGFP/RIaB mice are referred to as RIaB-On. The localization of Cre expression in the ARC was visualized by the fluorescence of EGFP fused to the Cre (Fig. 6a, left). Agrp-CreEGFP-mediated recombination has been shown to be specific to AgRP neurons in the hypothalamus 28 . By crossing the Agrp-Cre mouse to a tdTomato reporter mouse line 29 , we could observe Cre-activated tdTomato expression in AgRP but not in adjacent POMC neurons (Fig. 6a, right). RIaB expression significantly suppressed 24-h fasting-induced CREB phosphorylation in AgRP neurons (Fig. 6b,c). Activation of GABAergic AgRP neurons promotes feeding and inhibits energy expenditure at least partially by inhibiting the activity of PVN neurons that express melanocortin receptors 25,30 . We observed decreased c-Fos-positive cells in the ARC in RIaB-On mice compared with RIaB-Off mice following a 24-h fast (Fig. 6d,e). In contrast, we observed a 1.6-fold increase in c-Fos-positive cells in the PVN of fasted RIaB-On mice (Fig. 6d,e). These results indicated that PKA inhibition in AgRP neurons partially suppressed their activation by fasting and led to increased activity in a subset of PVN neurons. We examined whether RIaB-On mice recapitulated the obesity-resistant phenotype of RIIb KO mice 12 by placing them on a high-fat diet for 12 weeks. RIaB-On mice gained significantly less weight on a high-fat diet compared with RIaB-Off and Agrp-Cre/WT mice (Fig. 6f). Furthermore, RIaB-On mice had a significant decrease in fat pad weight compared with RIaB-Off and Agrp-Cre/WT mice on either chow or high-fat diet (Fig. 6g). These results demonstrate that PKA inhibition in AgRP neurons partially mimics the leanness of RIIb KO mice. However, our previous studies demonstrate that re-expression of RIIb only in AgRP neurons did not reverse the lean phenotype of RIIb KO mice 15 , suggesting that normal PKA signalling in AgRP neurons is required but not sufficient to keep adiposity at WT level.

Discussion
In this report, we show that RIIb-PKA is a major PKA isoform in the LepR-expressing neurons of the hypothalamus and is being regulated by fasting and re-feeding. RIIb deficiency leads to impaired PKA signalling, inhibition of pCREB induction during a fast and an increased duration of LepRb signalling through both the pSTAT3 and FoxO1 pathways in the hypothalamus. The extended duration of LepRb signalling in the RIIb KO hypothalamus leads to an increase in the catabolic effects of low doses of leptin on feeding, energy expenditure and body weight. The ability of leptin to induce a major negative feedback regulator of LepRb signalling, Socs3, is inhibited in the RIIb KO. An attractive hypothesis is that cAMP activation of RIIb-PKA plays a synergistic role in the induction of Socs3 by pSTAT3, perhaps by phosphorylation of CREB (Fig. 7). The Socs3 promoter contains CRE elements that bind CREB 31 and Socs3 can be induced in hypothalamic cell lines by cAMP 32,33 . Furthermore, the leanness and elevated leptin sensitivity of RIIb KO mice resembles the phenotype of mice with either a neuronspecific KO of Socs3 (ref. 34) or global haploinsufficiency of Socs3 (ref. 35). This defective CREB signalling might also contribute to the decreased transcription of FoxO1 in RIIb KO mice (Fig. 3a) since a recent study showed that PKA/CREB/p300 signalling promoted the expression of FoxO1 (ref. 36). The increase in leptin sensitivity we see in the RIIb KO hypothalamus occurs in all regions where the LepRb is expressed including the ARC, VMH, DMH and LH. We have shown previously that re-expressing RIIb in just the AgRP or POMC neurons of the ARC is not sufficient to reverse the lean phenotype. However re-expression of RIIb selectively in GABAergic neurons with Vgat-Cre does reverse the lean phenotype 15 and yet the glutamatergic neurons in the VMH remain deficient in RIIb and continue to be hypersensitive to leptin (Fig. 4). In the ARC of RIIb Vgat mice, a significant number of neurons show prolonged leptin-induced pSTAT3 activation and are negative for RIIb staining ( Fig. 4d and Supplementary Fig. 1d). These RIIb-negative neurons are likely to be the leptin-responsive glutamatergic POMC neurons 22 that would not be expected to re-express RIIb in the RIIb Vgat animals. These observations demonstrate that PKA signalling interacts with LepRb signalling in a cellautonomous manner. We have not yet been able to identify the specific subset of GABAergic neurons in the hypothalamus that are responsible for the effects of RIIb disruption on body weight regulation. It seems likely that these neurons are the same as those primary leptin-responsive neurons, still unidentified, which account for the majority of the effects of leptin on feeding and energy metabolism.
Although GABAergic neurons regulate the majority of leptin's effects on body weight, leptin signalling in other neuronal types such as glutamatergic neurons in the VMH also contributes to the regulation of body weight 37 and other metabolic processes including glucose homeostasis 37,38 and bone metabolism 39 . We showed previously that RIIb-PKA re-expression in SF1-expressing VMH neurons does not significantly restores the adiposity of RIIb KO mice 15 . However, it is possible that RIIb-PKA in the VMH might affect the susceptibility to diet-induced obesity or glucose homeostasis 12 . In addition to its role in leptin signalling, hypothalamic PKA is also a downstream effector of other metabolic signals such as glucagon like peptide (GLP1) and glucagon, which are implicated in the neuronal regulation of hepatic glucose production 40 . It is likely that RIIb-PKA is playing other roles in overall metabolic regulation in addition to its effects on leptin signalling and adiposity focused in this report.
Recently it was reported that cAMP promotes leptin resistance by activating the Epac/Rap1 signalling pathway and increasing Socs3 expression. These authors provided evidence that the pharmacological inhibition of PKA with H89 did not prevent the forskolin-induced increase in Socs3 mRNA and protein 41 . Our studies used mouse genetic approaches to inhibit the PKA pathway and this led to an increase in leptin sensitivity and resistance to diet-induced obesity. It is likely that both PKA and Epac/Rap1 signalling pathways are being co-regulated during fasting and re-feeding as cAMP levels change in response to metabolic signals and perhaps the relative contribution of the two cAMP effector pathways depends on both the level of cAMP and the intracellular localization of the cAMP effector. Elevated leptin sensitivity would be expected to lead to a decrease in adiposity and circulating leptin just as we see in the RIIb KO mice. The elevated leptin sensitivity of RIIb KO mice appears to be maintained for the animal's lifetime and their lean phenotype may underlie the increased healthy lifespan of RIIb KO mice 13 . The nutritional regulators that interact with hypothalamic neurons and regulate the levels of cAMP are unknown. It has been suggested that leptin itself might be capable of affecting cAMP by acting through the Akt pathway to activate PDE3 and cause a decrease in cAMP 42 . It is probable that there are also Gs-coupled GPCR pathways that stimulate adenylyl cyclase activity in LepRb-expressing neurons. A more detailed understanding of the crosstalk between the cAMP and LepRb signalling pathways may suggest a strategy for increasing leptin sensitivity therapeutically.

Methods
Mice. RIIb lox/lox and RIIb Vgat mice were generated and characterized as described previously 15 . We have deposited the RIIb lox/lox mice in the Mutant Mouse Regional Resource Center (MMRRC) Stock No. 036960-UCD and the RIIb Vgat mice were generated by crossing the RIIb lox/lox mice with Vgat-Ires-Cre mice sent to us by Brad Lowell (Harvard University). Agrp-CreEGFP mice were provided by Richard Palmiter (University of Washington). Conditional RIaB mice were described previously 26,27 and are available as Stock No. 032879-UCD from the MMRRC. All strains were on a C57Bl/6 background. The mice were fed standard chow (Picolab mouse diet 20) or high-fat diet (Research Diets #D12492) and had free access to water. Mice were housed at 22-24°C with a 12-h light/dark cycle and were individually housed for studies of food intake, energy expenditure and, after implantation of cannulas, for stereotactic injections. Otherwise, mice were group housed (two to five animals per cage). The number of mice used in each experiment was chosen based on the expected variance of the measurements to be used and a power analysis that would allow detection of at least a 30% change in values with a Po0.05. All procedures were approved by the Institutional Animal Care and Use Committee of the School of Medicine of the University of Washington.
Leptin administration. Recombinant mouse leptin was obtained from Dr Parlow (Harbor-UCLA Medical Center, CA). Age-matched male WT and RIIb KO mice (8-10-week old) were anesthetized and implanted with a cannula (Plastics One) into the third ventricle at the midline coordinates of 0.5 mm posterior to the bregma and 3.0 mm below the surface of the skull. After cannulation, mice were individually housed with free access to food and water, and body weight and ad libitum food intake were measured daily. After at least 1 week recovery from the surgery, 1 ml vehicle (artificial cerebrospinal fluid) with or without leptin was injected through the cannula into the third ventricle once a day between 1700 to 1900 hours (0-2 h before the start of dark phase). Daily food intake and body weight were measured at the time of injections. For oxygen consumption, mice were injected i.c.v. with vehicle or leptin at 1200 to 1300 hours and, 20 min later, put in the Oxymax metabolic chamber (Columbus Instruments) at room temperature for VO2 measurements for 2 h. The chamber was 4 Â 8 inches, allowing limited locomotion, and the entire apparatus was housed in an isolated room away from other animals and stimuli. Air flow to the cage was 500 ml min À 1 as previously described 43 . For each mouse, vehicle and leptin were injected on different days and VO2 was compared. Body weight-matched male WT and RIIb KO mice were used for leptin sensitivity study with i.p. injections between 1700 to 1900 hours. The mice were killed after the experiments and the major fat pads including gonadal, inguinal and retroperitoneal fat pads were isolated and weighed; the blood was collected for serum leptin assay.
AAV-Cre-mediated RIIb expression. Recombinant adeno-associated virus (AAV1-CreGFP) was kindly provided by Richard Palmiter (the University of Washington). Male RIIb lox/lox mice at 8 weeks of age were anesthetized and used for Cre virus injection (1 Â 10 9 genomic particles per ml per site) into the ventral hypothalamus (1.4 mm posterior to the bregma, 0.5 mm lateral to the midline, 5.8 mm below the bregma) with a 10 ml Hamilton syringe attached to a Micro4 Micro Syringe Pump Controller (World Precision Instruments). After injection, the mouse was housed individually with free access to food and water. At least 10 days after AAV injection, mice were fasted overnight and leptin was injected (1 mg kg À 1 , i.p.). Mice were killed and perfused for pSTAT3 immunofluorescent staining at 1 or 4 h after the injection. AAV-Cre-induced RIIb expression was determined by immunohistochemistry.
Quantitative reverse transcription-PCR assay. We isolated total RNA from the hypothalamus using Trizol (Invitrogen). RNA concentration was determined by RiboGreen assay. SYBR Green PCR Master Mix was used for quantitative reverse transcription-PCR. LepRb mRNA level was determined using the primers shown in Supplementary Table 1 and characterized previously 44 . The data were normalized to either b-actin or Gapdh. No difference was observed between LepRb mRNA levels in WT and RIIb KO mice. mRNA levels of Agrp, Npy, Pomc, Socs3 and FoxO1 were all normalized to LepRb mRNA content and expressed as a percentage change from corresponding WT control. Primers for Agrp, Npy, Pomc, b-actin, Socs3 and FoxO1 are shown in Supplementary Table 1.
Leptin measurements. Mice were killed with CO 2 , and whole blood was collected by cardiac puncture. The blood was allowed to clot at room temperature and centrifuged at 12,000 r.p.m. for 2 min. The supernatant serum was collected and saved at À 80°C for leptin assay by ELISA following the manufacturer's protocol (Millipore, Cat# EZML-82 K).