The stimulatory G protein Gsα is required in melanocortin 4 receptor–expressing cells for normal energy balance, thermogenesis, and glucose metabolism

Central melanocortin 4 receptors (MC4Rs) stimulate energy expenditure and inhibit food intake. MC4Rs activate the G protein Gsα, but whether Gsα mediates all MC4R actions has not been established. Individuals with Albright hereditary osteodystrophy (AHO), who have heterozygous Gsα-inactivating mutations, only develop obesity when the Gsα mutation is present on the maternal allele because of tissue-specific genomic imprinting. Furthermore, evidence in mice implicates Gsα imprinting within the central nervous system (CNS) in this disorder. In this study, we examined the effects of Gsα in MC4R-expressing cells on metabolic regulation. Mice with homozygous Gsα deficiency in MC4R-expressing cells (MC4RGsKO) developed significant obesity with increased food intake and decreased energy expenditure, along with impaired insulin sensitivity and cold-induced thermogenesis. Moreover, the ability of the MC4R agonist melanotan-II (MTII) to stimulate energy expenditure and to inhibit food intake was impaired in MC4RGsKO mice. MTII failed to stimulate the secretion of the anorexigenic hormone peptide YY (PYY) from enteroendocrine L cells, a physiological response mediated by MC4R–Gsα signaling, even though baseline PYY levels were elevated in these mice. In Gsα heterozygotes, mild obesity and reduced energy expenditure were present only in mice with a Gsα deletion on the maternal allele in MC4R-expressing cells, whereas food intake was unaffected. These results demonstrate that Gsα signaling in MC4R-expressing cells is required for controlling energy balance, thermogenesis, and peripheral glucose metabolism. They further indicate that Gsα imprinting in MC4R-expressing cells contributes to obesity in Gsα knockout mice and probably in individuals with Albright hereditary osteodystrophy as well.

G s ␣ is a ubiquitously expressed G protein ␣-subunit that couples various hormone and neurotransmitter receptors, including melanocortin receptors, to adenylyl cyclase, leading to increased generation of intracellular cAMP. Central melanocortins act via melanocortin receptors, such as melanocortin 4 (MC4R) 5 receptors, which are expressed in distinct regions of the central nervous system (CNS) (1) and have divergent effects on energy and glucose homeostasis, thermogenesis, and cardiovascular function (2)(3)(4)(5). Both patients with MC4R mutations (6,7) and MC4RKO mice (8,9) develop obesity associated with hyperphagia and reduced energy expenditure, as well as increased linear growth. G s ␣ is encoded by the gene GNAS (Gnas in mice) that undergoes genomic imprinting, an epigenetic phenomenon leading to differential expression from the two parental alleles. Heterozygous loss-of-function mutations of G s ␣ lead to Albright hereditary osteodystrophy (AHO), which presents with skeletal and neurobehavioral abnormalities. AHO patients who inherit the mutation maternally or who have a de novo mutation on the maternal allele also develop multihormone resistance, including insulin resistance, as well as early-onset obesity, whereas these features are absent in patients with mutations on the paternal allele (10 -12). Similarly, mice with germ line G s ␣ deletion on the maternal allele develop obesity and insulin-resistant diabetes, whereas these features are absent in mice with paternal G s ␣ deletion (13)(14)(15). This parentof-origin-specific metabolic phenotype is due to G s ␣ being imprinted in a tissue-specific manner (16). G s ␣ imprinting in one or more regions of the CNS is involved in the parent-oforigin-specific metabolic phenotype, as mice with CNS-specific disruption of the maternal, but not paternal, G s ␣ allele develop obesity and insulin-resistant diabetes (17). More recently, the dorsomedial hypothalamus (DMH) has been identified as at least one site where G s ␣ imprinting impacts the effects of G s ␣ mutations on energy homeostasis (18). However, the extent to which G s ␣ deficiency in MC4R-expressing neurons is responsible for the parent-of-origin metabolic abnormalities in AHO patients and G s ␣ knockout mice is undefined.
There are several differences in the obesity phenotype observed in mice with maternal G s ␣ mutation versus MC4R mutation. Unlike MC4RKO mice, mice with a CNS-specific maternal G s ␣ mutation (mBrGsKO) do not have a primary effect on food intake or on acute regulation of food intake by a melanocortin agonist; nor does the mutation affect linear growth (17). Recently, we have shown that G q ␣/G 11 ␣ rather than G s ␣ mediates the actions of MC4R on food intake and linear growth in the paraventricular nucleus of the hypothalamus (PVN) (19). This and the fact that some MC4R mutations identified in obese patients do not affect activation of G s ␣/ cAMP signaling (20) raise the question of whether G s ␣ is required for all of the actions of MC4Rs.
The present study investigated the consequences of G s ␣ deficiency in MC4R-expressing cells and provides evidence that G s ␣ signaling in MC4R-expressing cells plays a critical role in the control of food intake and energy expenditure, thermogenesis, and peripheral glucose metabolism. In addition, we show that that G s ␣ imprinting within these cells contributes to the parent-of-origin-specific metabolic phenotype observed with G s ␣ mutations.

G s ␣ deficiency in MC4R-expressing cells leads to severe obesity
Mice with homozygous deletion of G s ␣ in MC4R-expressing cells (MC4RGsKO mice) were generated by breeding G s ␣floxed mice (E1 fl/fl ) (21) with MC4R-cre mice (22). To confirm proper targeting of the MC4R-cre line and specific loss of G s ␣ in targeted cells, MC4R-cre:E1 fl mice were serially mated with Ai14 reporter mice with a Cre-dependent tdTomato transgene (23). Examination of two brain regions known to harbor a large number of MC4R-expressing neurons (the PVN and the dorsal motor nucleus of the vagus (DMV)) showed a large number of tdTomato-expressing neurons with a lower number of tdTomato-expressing cells in surrounding areas (Fig. 1A). Costaining with a G s ␣-specific antibody by immunofluorescence demonstrated specific loss of G s ␣ expression in tdTomato-positive cells in both PVN and DMV (Fig. 1, B and C). No tdTomato staining was observed in cardiac or skeletal muscle (quadriceps) or in BAT, indicating that MC4R-cre is not expressed in these peripheral tissues (data not shown).
MC4RGsKO offspring were born and weaned at expected Mendelian ratios. MC4RGsKO mice steadily gained more weight than their littermate controls starting after 5 weeks of age, and by 12 weeks of age, male MC4RGsKO mice were 103% and female MC4RGsKO mice were 143% heavier than their littermate controls ( Fig. 2A). This increased weight gain was due to a marked increase in fat mass, with a lesser increase in lean mass (Fig. 2B). MC4RGsKO mice also had a small, but significant, increase in body length (Fig. 2C). Consistent with their increased fat mass, MC4RGsKO mice had enlarged adipocytes in both brown and white adipose tissue (BAT and WAT, respectively) with greater intracellular lipid accumulation as compared with controls ( Fig. 2J) and had markedly increased serum leptin levels (Table 1). MC4RGsKO mice also developed obesity with increased fat mass and a lesser increase in lean mass when raised at thermoneutrality (30°C), where sympathetic nervous system (SNS) activity is at a minimum (Fig. S1, A and B).
Mice with a heterozygous deletion within the G s ␣ maternal allele in MC4R-expressing cells (mMC4RGsKO) also gained significantly more weight than controls (Fig. 2, D and E), although the differences were much less than those observed with homozygous MC4RGsKO mice ( Fig. 2A). Both male and female mMC4RGsKO mice had significantly greater fat mass and body length, whereas lean mass was only slightly greater in male mMC4RGsKO mice (Fig. 2, E and F). WAT adipocytes tended to be slightly enlarged, although the differences in cell size did not reach statistical significance (Fig. 2K). Similar small increases in body weight and fat mass were also observed in mMC4RGsKO mice maintained at thermoneutrality (Fig. S1, C green, staining using a G s ␣-specific antibody; blue, 4Ј,6-diamidino-2-phenylindole. Scale bar, 20 m. C, similar images taken from the DMV. In both PVN and DMV, merged images indicate loss of G s ␣ expression from MC4R-expressing neurons. No evidence of tdTom staining was observed in the heart muscle, quadriceps muscle, or BAT of MC4RGsKO mice (data not shown).

G s ␣ deficiency in MC4R-expressing cells leads to obesity
and D). In contrast, G s ␣ deletion within the paternal allele in MC4R-expressing cells in pMC4RGsKO mice showed no changes in body weight, composition, or length or histological appearance of WAT or BAT (Fig. 2, G-I and K). Consistent with a parent-of-origin effect of G s ␣ mutation on fat mass, serum leptin levels were increased in mMC4RGsKO mice but were unaffected in pMC4RGsKO mice (Table 1). These results indicate that G s ␣ in MC4R-expressing cells is essential for the control of energy homeostasis and that there is a parent-oforigin effect of G s ␣ deletion in heterozygotes consistent with G s ␣ imprinting within a population of MC4R-expressing cells that are important for metabolic regulation.

MC4RGsKO mice are hyperphagic
To avoid the impact of the large differences in body weight and adiposity that developed between the mutants and con-

G s ␣ deficiency in MC4R-expressing cells leads to obesity
trols on energy balance, we examined food intake and energy expenditure in mice before the onset of significant weight gain. We examined food intake weekly over a 3-week period (ages 7-10 weeks) at 22°C. At 7 weeks, the body weights of MC4RGsKO mice were similar to controls, and subsequently, MC4RGsKO mice gained more weight than controls over the next 3 weeks (Fig. 3A). During this period, the mutants were clearly hyperphagic as compared with controls ( Fig. 3B). Total energy expenditure (TEE) over this period was calculated using an energy balance technique by measuring body composition and subtracting the gain in total body energy stores from the total food intake (24). xy plots of TEE versus average body weight during this experiment showed that TEE relative to body weight was reduced and that this probably also contributes to the rapid weight gain observed in MC4RGsKO mice (Fig. 3C). Total activity levels measured by beam break interruption were significantly reduced in MC4RGsKO mice at 22°C, whereas ambulatory activity also tended to be reduced (Fig. 3H). There were no differences in the respiratory exchange ratio (RER; vCO 2 /vO 2 ) between controls and mutants at 22°C (Fig. 3G).
Similar studies performed over 2 weeks (5-7 weeks of age) in mice housed at thermoneutrality (30°C) also showed increased weight gain and food intake in MC4RGsKO mice (Fig. 3, D and E). Although interpretation is difficult due to the large disparity in body weights between the two groups, linear regression analysis of TEE versus body weight indi-cated that, in contrast to what is observed at 22°C, there was no reduction in TEE relative to body weight at 30°C (Fig. 3F). This discrepancy between what was observed at 22°C versus 30°C was confirmed by indirect calorimetry (data not shown) and is consistent with what we observed previously in a whole brain-specific G s ␣ knockout model (17). Both total and ambulatory activity levels were decreased when measured at 30°C. Overall, our findings indicate that the primary driver of obesity in MC4RGsKO mice is hyperphagia, although reduced TEE at temperatures below thermoneutrality probably contributes as well.

mMC4RGsKO mice have reduced energy expenditure
In contrast to MC4RGsKO mice, maternal heterozygous mice (mMC4RGsKO) showed no evidence of hyperphagia (Fig. 4A). TEE measured by indirect calorimetry (normalized to lean mass) was significantly reduced, whereas resting energy expenditure (REE) tended to be reduced in mMC4RGsKO mice at 22°C (p ϭ 0.095; Fig. 4B). No differences in REE or TEE at 30°C or activity levels at either temperature were observed in these mice (Fig. 4, B and C). No differences in food intake, energy expenditure, or activity levels were observed in pMC4RGsKO mice (Fig. 4, D-F). Therefore, the mild obesity observed in mMC4RGsKO mice is primarily the result of a small decrease in energy expenditure, consistent with what has been observed in mice with heterozygous disruption of the maternal Gnas allele in other mouse models (17,18). and daily food intake averaged by week (B) measured in male MC4RGsKO mice and control littermates housed at 22°C from 7 to 10 weeks of age (n ϭ 6/group; week 0 in A is at 7 weeks of age). C, plots of mean daily total energy expenditure versus mean body weight for individual mice from A and B with linear regression lines. D and E, body weight curves (D) and daily food intake averaged by week (E) measured in male MC4RGsKO mice and control littermates housed at thermoneutrality (30°C) from 5 to 7 weeks of age (n ϭ 7-9/group; week 0 in D is at 5 weeks of age). F, plots of mean daily total energy expenditure versus mean body weight for individual mice from D and E with linear regression lines. G, respiratory exchange ratios (RER; vCO 2 /vO 2 ) measured at 22°C in 3-4-month-old male MC4RGsKO mice and control littermates (n ϭ 5/group). H, total and ambulatory (Amb) activity levels of 3-4-month-old male MC4RGsKO mice and control littermates (n ϭ 5/group). The linear regression line for MC4RGsKO mice at 22°C is significantly shifted to the right (p Ͻ 0.01), whereas the linear regression line at 30°C is not shifted. Data are expressed as mean Ϯ S.E. (error bars). *, p Ͻ 0.05; **, p Ͻ 0.01 versus controls.

MC4RGsKO and mMC4RGsKO mice have impaired responses to an MC3/4R agonist
We previously showed that whole brain-specific loss of G s ␣ expression from the maternal allele (mBrGsKO) leads to an impaired energy expenditure response to the MC3/4R agonist melanotan-II (MTII), whereas the food intake response to MTII is unaffected (17). To determine the extent to which G s ␣ deficiency in MC4R-expressing cells alters responses to a melanocortin agonist, we measured the energy expenditure and food intake responses to MTII in homozygous and heterozygous MC4RGsKO mice. MTII-stimulated oxygen consumption (a measure of energy expenditure) was markedly reduced in MC4RGsKO and mMC4RGsKO, but not in pMC4RGsKO, mice (Fig. 5, A and D). This is consistent with impaired melanocortin stimulation of energy expenditure as the underlying cause of the obesity associated with maternal G s ␣ mutation and confirms that G s ␣ imprinting in a population of MC4R-expressing cells plays a role in the parent-of-origin-specific effects of G s ␣ deletion on energy expenditure observed in mBrGsKO and pBrGsKO mice, respectively (17).
Similar to what was observed previously in mBrGsKO and pBrGsKO mice (17), the food intake response to MTII was unaffected in both mMC4RGsKO and pMC4RGsKO mice (Fig.  5E), which is consistent with the obesity in mMC4RGsKO mice being primarily due to lower energy expenditure. Interestingly, the ability of MTII to inhibit food intake in homozygous MC4RGsKO mice was impaired (Fig. 5B), suggesting that resistance to the effects of melanocortins on food intake may contribute to the hyperphagia observed in these mice.
MC4Rs are expressed in cells outside of the CNS, including the enteroendocrine L cells in the gastrointestinal tract (25), where their activation stimulates secretion of peptide YY (PYY), a circulating peptide that acts to reduce appetite, and this response is mediated by MC4R activation of G s ␣ (26). Loss of PYY secretion from L cells in response to MTII due to G s ␣ deficiency in these cells could be one potential explanation for the impaired food intake response to MTII in MC4RGsKO mice. Whereas circulating PYY levels significantly increased in response to MTII in control mice, there was no significant response to MTII in MC4RGsKO mice, although this lack of response may reflect the fact that baseline PYY levels were significantly increased in MC4RGsKO mice (Fig. 5C). Circulating levels of glucagon-like peptide 1 (GLP-1), another peptide secreted from enteroendocrine L cells that inhibits food intake, were also found to be elevated in MC4RGsKO mice in the basal fed state ( Table 1). The elevated baseline levels of PYY and GLP-1 may be a secondary response to counteract the hyperphagia observed in these mice and, in the case of GLP-1, may reflect direct resistance to its actions on food intake, as GLP-1 receptors mediate their responses via G s ␣ (27). Lack of hormonal responsiveness of L cells to MTII may contribute to the A-C, food intake (kcal/day; n ϭ 6 -7/group) (A), resting and total energy expenditure normalized to lean mass (REE and TEE, respectively; n ϭ 7-10/group) (B), and total and ambulatory (Amb) activity levels (n ϭ 7-10/group) (C) of male mMC4RGsKO and control littermates at 3-4 months of age. D-F, food intake (kcal/day; n ϭ 7/group) (D), resting and total energy expenditure normalized to lean mass (n ϭ 5-8/group) (E), and total and ambulatory activity levels (n ϭ 5-8/group) (F) of male pMC4RGsKO and control littermates at 3-4 months of age. Data are expressed as mean Ϯ S.E. (error bars). *, p Ͻ 0.05 versus controls.

G s ␣ deficiency in MC4R-expressing cells leads to obesity
impaired anorectic response to this agonist that we observed in MC4RGsKO mice.

Loss of G s ␣ in MC4R-expressing cells alters glucose metabolism
Homozygous MC4RGsKO mice had significantly increased baseline glucose, insulin, and leptin levels ( Table 1) and severe glucose intolerance and insulin resistance (Fig. 6 (A and B) and Fig. S2 (A and B)) when measured at 12-20 weeks of age, well after the establishment of obesity. In contrast, these mice showed no significant differences in serum-free fatty acid, triglyceride, or cholesterol levels (Table 1). To determine whether loss of G s ␣ expression in MC4R-expressing cells directly alters glucose metabolism, we studied MC4RGsKO mice at 5-6 weeks of age, before the onset of obesity. At this age, when body weight was still normal (Fig. 6G), MC4RGsKO mice had impaired glucose tolerance (Fig. 6H), whereas insulin levels during the glucose tolerance test tended to be higher in the mutants (Fig. 6I). These results suggest that G s ␣ signaling in MC4R-expressing cells directly affects glucose tolerance and insulin sensitivity independently of its effects on energy balance.
Both male and female mMC4RGsKO mice had mild glucose intolerance ( Fig. 6C and Fig. S2C). Insulin sensitivity tended to be reduced in male mMC4RGsKO mice, although the areas below baseline calculated based on the insulin tolerance tests were not significantly different (Fig. 6D). Male mMC4RGsKO mice also tended to have higher insulin and leptin levels compared with control mice (Table 1). Female mMC4RGsKO mice did not show evidence of insulin resistance based upon insulin tolerance tests (Fig. S2D). Male pMC4RGsKO mice displayed slightly but significantly impaired glucose tolerance (Fig. 6E), despite having normal body weight, serum glucose, and insulin levels ( Fig. 2G and Table 1). Insulin sensitivity tended to be lower in these mice, although the areas below baseline were not significantly different (Fig. 6F). Female pMC4RGsKO mice had normal glucose tolerance and insulin sensitivity (Fig. S2, E and F).

MC4RGsKO mice, but not mMC4RGsKO or pMC4RGsKO mice, have impaired cold tolerance
Reduced energy expenditure in MC4RGsKO and mMC4RGsKO mice at ambient temperature but not at thermoneutral temperature (Figs. 3 (C and F) and 4B), at which the SNS activity is minimal, suggests an impairment of SNS activity in these mice. We measured tolerance to an acute cold environment, which requires SNS-mediated BAT thermogenesis. Body temperature of MC4RGsKO mice rapidly declined within 1 h of cold exposure, and 3 of 10 mice had to be removed from the cold environment prematurely, as their body temperature had dropped below 25°C (Fig. 7A) before the end of the experiment. BAT temperature measured with an implanted probe was also significantly decreased in MC4RGsKO mice that were placed in the cold, indicating a defect in BAT thermogenesis (Fig. 7B). These mice had significantly lower basal BAT Ucp1 (uncoupling protein 1) mRNA levels, a marker of thermogenic capacity that is stimulated by SNS activity, at room temperature compared with controls (Fig. 7C), consistent with inactive brown adipocyte morphology (Fig. 2J). Cold exposure significantly stimulated Ucp1 gene expression in control but not in MC4RGsKO mice (Fig. 7C), indicating that these mice have impaired BAT activation in response to acute cold challenge. mMC4RGsKO mice were able to maintain their body temperature for up to 6 h at 6°C (Fig. 7D) and had normal baseline BAT Ucp1 mRNA levels, although the response of BAT Ucp1 mRNA to cold was impaired (Fig. 7F). BAT temperature also remained similar to controls except for a slight decrease at time points 1 and 3 h after being placed in the cold (Fig. 7E).
pMC4RGsKO mice maintained their body temperature during cold exposure, and baseline and cold-induced levels of BAT Ucp1 mRNA were similar to those of controls (Fig. 7, G  and H).
We next examined the response of the mutants to chronic cold adaptation. In these experiments, environmental temperature was lowered by 2°C each day down to 6°C, and then the mice were maintained at 6°C for 5 days. MC4RGsKO mice maintained their body temperature ϳ2°C lower than that of controls throughout most of the experiment (Fig. 8A). Immunostaining of inguinal WAT (iWAT) for UCP1 in MC4RGsKO mice showed areas of increased UCP1 staining but little formation of obvious beige-appearing cells (Fig. 8C). Whether this represents a failure of beiging or is a reflection of the increased lipid accumulation in iWAT cells is unclear. It is also unclear whether the small decrease in body temperature throughout this experiment is due to insufficient cold-induced thermogenesis or rather is due to a change in the central temperature set point at lower environmental temperatures. Interestingly, after longer cold exposure, there was now significant induction of Ucp1 mRNA expression in interscapular BAT (Fig. 8E) in these mutants.
Similar chronic cold adaptation experiments in mMC4RGsKO mice showed that these mice were able to maintain normal body temperature in the cold (Fig. 8B). Similar to MC4RGsKO mice, these mice showed increased UCP1 staining in iWAT but no clear-cut beige-appearing cells (Fig. 8D). Induction of Ucp1 mRNA in interscapular BAT was normal (Fig. 8F).

G s ␣ deficiency in MC4R-expressing cells leads to obesity Reduced expression of adipogenesis and lipogenesis-related genes in adipose tissue from MC4RGsKO mice
Examination of gene expression in BAT and epididymal WAT (eWAT) samples from MC4RGsKO mice that were maintained at 22°C showed that expression of virtually all of the genes examined that are involved in beiging/browning, adipogenesis, or lipogenesis was either significantly reduced or tended to be reduced in MC4RGsKO mice (Fig. 9, A and B). These changes are consistent with the reduced levels of Ucp1 mRNA that we observed in BAT from MC4RGsKO mice (Figs. 7C and 8E). We previously showed that loss of G s ␣ expression in adipose tissue, which results in reduced sensitivity of adipose tissue to SNS activity, also led to significantly reduced expression of adipogenesis, browning, and lipogenesis genes (28). Together with the adipose tissue histology and cold intolerance, the present data are consistent with reduced efferent SNS activity to adipose tissue in MC4RGsKO mice. Similar analysis in adipose tissue from mMC4RGsKO mice showed no significant differences in gene expression except for reduced Pgc1a expression in BAT, although the expression of many of the genes tended to be lower in eWAT from mMC4RGsKO mice (Fig. 9, C and D).

MC4RGsKO mice have reduced heart rate
MC4RGsKO mice had significant decreases in heart rate, whereas blood pressure was unaffected (Fig. 10, A and B), indicating that G s ␣ deficiency in MC4R-expressing neurons leads to a reduction of cardiac SNS activity. Heart rate and blood pressure were unaffected in both mMC4RGsKO (Fig. 10, C and D) and pMC4RGsKO mice (Fig. 10, E and F).

Discussion
MC4R mutations are the most common cause of severe monogenic obesity and are associated with hyperphagia, reduced energy expenditure, increased body length, impaired BAT thermogenesis, insulin resistance, and decreased heart rate and blood pressure (4,7,8). The present study investigated to what extent G s ␣ signaling might play a role in these diverse physiologic effects of MC4R mutations. Our results show that mice with homozygous G s ␣ mutation in MC4R-expressing cells were severely obese primarily due to increased food intake, although reduced energy expenditure most likely contributes to this phenotype as well.
Consistent with increased food intake observed in MC4RGsKO mice, the ability of the MC4R agonist MTII to reduce food intake was impaired in these mice, suggesting resistance to the effects of MC4R on food intake. Although we have shown that MC4R mediates its effects on food intake within the PVN via G q/11 ␣ signaling, loss of G q/11 ␣ signaling in PVN did not completely abolish the effect of MTII on food intake (19), suggesting that MC4R may also mediate effects on food intake via G s ␣ in other brain regions, most likely outside of the PVN. The fact that we saw no effect on food intake in mMC4RGsKO mice or in more global deletion of the maternal Gnas allele

G s ␣ deficiency in MC4R-expressing cells leads to obesity
within the CNS (17) indicates that the loss of MC4R action on food intake requires complete loss of G s ␣ expression and its signaling and is probably occurring in one or more areas where G s ␣ does not undergo imprinting. However, it should be stressed that other anorexigenic hormones and neurotransmitters, such as GLP-1, mediate their actions to lower food intake via G s ␣ (29) and that melanocortins may alter the sensitivity to other satiety factors (30), so that defects in signaling pathways besides MC4R also probably contribute to the hyperphagia observed in MC4RGsKO mice.
Intestinal endocrine L cells are one potential site of action outside of the CNS where loss of MC4R-G s ␣ signaling might mitigate the acute anorectic response to MTII, as MC4R-G s ␣ signaling in these cells stimulates the secretion of various anorexigenic peptides, including PYY and GLP-1 (26). In contrast to controls, MC4RGsKO mice showed no increase in serum PYY levels in response to MTII, which may partly account for the impaired ability of MTII to reduce food intake in these mice. However, it should be noted that PYY levels (as well as GLP-1 levels) were elevated in MC4RGsKO mice at baseline. Circulating PYY levels are known to be low in the fasting state and increase postprandially. Obesity in humans is associated with elevated fasting PYY levels (31) and attenuated PYY release in response to feeding (32,33). It is unclear whether the elevated fasting PYY levels observed in MC4RGsKO mice is secondary to obesity or to a direct perturbation in enteroendocrine L cells due to loss of G s ␣. It is possible that the elevation in PYY is a response to hyperphagia, but this effect would be secondary, as PYY does not mediate its actions primarily by G s ␣. In contrast, GLP-1 does mediate its actions via G s ␣ so that G s ␣ deficiency in MC4R-expressing cells may directly contribute to elevated GLP-1 levels in MC4RGsKO mice.
MC4RGsKO mice also had a very small but significant increase in body length, which was not observed in mouse models with CNS-specific maternal G s ␣ mutations (17,34). In fact, a greater increase in body length was observed in mice with PVN-specific loss of G q/11 ␣ signaling, and similar increases in linear growth are observed in humans and mice with loss of SIM1 (35,36), a relatively PVN-specific transcription factor that is regulated by PVN MC4R-G q/11 ␣ signaling (19). The extent to which MC4R-G s ␣ signaling contributes to the effect of MC4R mutations on linear growth appears to be small, and further studies will be required to determine whether the small changes in linear growth observed in MC4RGsKO mice represent a primary effect due to loss of MC4R-G s ␣ signaling as opposed to a secondary effect due to other metabolic perturbations or disturbance of other G s ␣-mediated signaling pathways in MC4RGsKO cells.

G s ␣ deficiency in MC4R-expressing cells leads to obesity
Existing evidence shows that activation of central MC4R signaling inhibits basal plasma insulin secretion and improves glucose tolerance and insulin sensitivity (4) and that MC4Rs expressed in cholinergic neurons are required for their inhibitory effect on plasma insulin levels (37,38). Our data revealed that young MC4RGsKO mice with normal body weight exhibited impaired glucose tolerance, which is probably due to a primary defect in insulin sensitivity, as insulin levels during the glucose tolerance test showed no evidence of insulin deficiency. The more severe effects observed in older MC4RGsKO mice probably reflect the additional effects of severe obesity on glucose metabolism. Both male mMC4RGsKO and pMC4RGsKO mice showed similar and more subtle effects on glucose metabolism, suggesting a small effect on glucose metabolism due to haploinsufficiency that is not affected by genomic imprinting. However, female mMC4RGsKO mice had mild glucose intolerance, whereas pMC4RGsKO mice did not. Loss of response to other ligands whose response is mediated via G s ␣ in MC4Rexpressing cells may also contribute to the effects that we observed on glucose metabolism.
MC4R-expressing neurons can promote SNS-mediated BAT thermogenic function (39 -42), and this effect has been recently shown to primarily involve MC4Rs in cholinergic autonomic sympathetic preganglionic neurons (38). MC4RGsKO mice had inactive BAT with larger lipid droplets and significantly reduced Ucp1 mRNA levels at room temperature and failed to increase BAT Ucp1 expression in response to cold, which resulted in a rapid drop in body temperature during acute cold exposure. In contrast, BAT Ucp1 expression was induced normally during a more chronic cold adaptation paradigm, which is consistent with a prior study showing lack of Ucp1 induction during acute cold but normal Ucp1 induction during chronic cold conditions in MC4R knockout mice (39). MC4RGsKO mice also maintain a somewhat lower body temperature during chronic cold, but whether this represents a defect in cold-induced thermogenesis or a change in the centrally regulated temperature set point is not known. These results clearly demonstrate that G s ␣ signaling in MC4R-expressing cells is required for normal BAT activation and cold-induced thermogenesis. MC4RGsKO mice also exhibited reduced heart rate. Thus, G s ␣ in MC4R-expressing cells is critical for maintenance of BAT and cardiac SNS activity. It is possible that the defects in cold-induced thermogenesis observed in MC4RGsKO mice are not directly due to loss of MC4R-G s ␣ signaling, as MC4R knockout mice were shown not to develop cold intolerance (9,43).
mMC4RGsKO mice with G s ␣ deletion on the maternal allele had mild obesity with decreased energy expenditure and mildly increased body length. In contrast, pMC4RGsKO mice with G s ␣ deletion on the paternal allele showed normal body weight and length and normal energy expenditure. Moreover, mMC4RGsKO mice have mildly impaired induction of BAT Ucp1 expression in the cold, whereas pMC4RGsKO mice have normal cold-induced BAT Ucp1 expression. These results indicate that G s ␣ imprinting in MC4R-expressing cells contributes to the parent-of-origin-specific effects on adiposity, energy expenditure, and thermogenesis observed with heterozygous G s ␣ mutations. This is because G s ␣ is preferentially expressed from the maternal allele in a subset of MC4R-expressing cells that are involved in energy homeostasis, and therefore deletion of G s ␣ from the maternal allele in these cells leads to a greater loss of G s ␣ expression (Ͼ50%) than the same deletion on the paternal allele. The fact that the mMC4RGsKO phenotype is less severe than that of homozygotes is consistent with G s ␣ imprinting being both tissue-specific and incomplete in terms of loss of expression of G s ␣ from the paternal allele, leading to only a partial loss of G s ␣ expression. Complete loss of G s ␣ in one or more brain regions leads to hyperphagia in the homozygotes, which is absent in heterozygotes. We have recently shown that G s ␣ imprinting in the DMH, a site at which MC4R is also expressed (44), plays a major role in the parent-oforigin-specific metabolic phenotype observed with G s ␣ mutations and that loss of MC4R in the DMH leads to a similar phenotype as mice with DMH-specific deletion of the maternal G s ␣ allele (18). Thus, disruption of G s ␣ signaling in MC4Rexpressing neurons significantly contributes to obesity associated with G s ␣ mutations in mice and probably in AHO patients as well.

G s ␣ deficiency in MC4R-expressing cells leads to obesity Body weight, composition, food intake, and metabolism
Body composition was measured using an Echo 3-in-1 NMR analyzer (Echo Medical Systems, Houston, TX). Food intake was performed on individually housed male mice that were acclimated for 5 days, followed by food and body weight measurements every other day for 7 days. REEs and TEEs were measured over a 24-h period by indirect calorimetry using a 12-chamber CLAMS/Oxymax system (Columbus Instruments, Columbus, OH) as described previously (17). Total and ambulatory activity levels were measured using IR beam interruption (OptoVarimex mini, Columbus Instruments). Day (light cycle) was 0600 -1800 h, and night (dark cycle) was 1800 -0600 h. For energy balance studies, mice were acclimated to a home cage environment for a minimum of 72 h. Body weight and food intake were measured weekly for the duration of the experiment. Mice housed at thermoneutral temperatures (30°C) were pair-caged for energy balance studies, whereas mice were singly housed for studies at room temperature (22°C). Body composition was measured weekly, and TEE was calculated by subtracting the gain in total body energy stores from the total energy intake over the measurement interval (24).

Glucose and insulin tolerance tests
Blood glucose or insulin tolerance tests were performed on overnight fasted mice with intraperitoneal injection of glucose (2 mg/g) or insulin (Humulin, 0.75 mIU/g) as described (21). Blood glucose was obtained from tail blood and measured using a Contour glucometer (Bayer) at the indicated times.

Responses to MTII
For measurement of food intake response to MTII, singlecaged mice were fasted for 24 h before receiving vehicle (saline, 100 l i.p.) at 30 min before lights out, and food intake was measured during the first 3.5 h after injection. Mice were allowed to recover for 2 days before a second 24-h fast, followed by administration of MTII (200 g i.p.) and measurement of food intake. For measurement of the energy expenditure response to MTII, mice were given MTII (10 g/g i.p.) or saline on separate days, and total oxygen consumption was measured at 30°C before and between 1 and 3 h after injection of MTII or saline.

Responses to acute and chronic cold exposure
Rectal temperature was measured with a TH-5 rectal probe (Thermalet, TX) inserted 1 cm deep, and interscapular BAT (iBAT) temperature was directly assessed using a telemetry temperature probe (IPII-300, Bio Medic Data Systems Inc., Seaford, DE) that was surgically implanted into the iBAT of mice under isoflurane anesthesia and detected by a receiver (DAS-7007S, Bio Medic Data Systems). Before acute cold tolerance testing, mice were acclimated to experimental conditions at room temperature for 3 days with daily measurement of rectal and iBAT temperatures. During the acute cold tolerance test, mice were individually housed without bedding and provided ad libitum access to food and water. Rectal and iBAT temperatures were measured before (time 0) and at the indicated time points after exposure to 6°C. For chronic cold adaptation, sin-gle-caged mice were housed in a temperature-controlled chamber (Memmert 750 LIFE Chamber) with bedding and were provided with food and water ad libitum. The ambient temperature of the chamber decreased in 2°C/day increments from 22 to 6°C, and then mice were housed at 6°C for the remainder of the experiment. Rectal temperatures were measured at the indicated time points.

PYY measurements
Male mice were injected with saline (100 l i.p.) and MTII (200 g i.p.) on separate days following a 24-h fast to reduce postprandial hormones to basal levels. Mice were allowed to recover for 1 week between administration of saline or MTII. Blood samples were collected 10 min after injection of saline or MTII by submandibular bleeding. Serum PYY levels were measured using the Milliplex MAP mouse metabolic hormone magnetic bead panel (EMD-Millipore, MMHMAG-44K), following the manufacturer's instructions. The PYY values were read on a Bio-Plex Magpix multiplex reader (Bio-Rad) and analyzed with xPONENT software (Luminex).

Biochemical assays
Serum glucose levels were measured with an Elite glucometer (Bayer). Insulin and leptin levels were measured using an ELISA kit (obtained from Crystal Chem and R&D Systems, respectively). Free fatty acids were measured using reagents from Roche Applied Sciences, and triglycerides and cholesterol levels were measured using reagents from Thermo (Middletown, VA). Serum GLP-1 levels were measured by ELISA following the manufacturer's protocol (EMD Millipore, Darmstadt, Germany).

Blood pressure and heart rate measurements
Blood pressure and heart rate were measured using a BP-2000 specimen platform (Visitech).

Gene expression
Gene expression in adipose tissue was assessed by quantitative RT-PCR as described previously (28).

Adipocyte size quantification
BAT, eWAT, and iWAT fat pads were fixed in 10% formalin. Tissue sections (5 m thick for BAT; 8 m thick for eWAT and iWAT) were subsequently H&E-stained and imaged at ϫ20 (BAT) or ϫ10 (eWAT and iWAT) magnifications with a Hamamatsu NanoZoomer 2.0RS slide scanner (Meyer Instruments). Adipocyte size was measured for each tissue section using the FIJI plug-in Adiposoft, an open source software for the quantification of adipose tissue cellularity in histological sections (45).

UCP1 immunohistochemistry
iWAT fat pads were fixed in 10% formalin and then cut into 8-m-thick sections. Tissue sections were treated in 10 mM sodium citrate with 0.05% Tween 20 at 85°C for 20 min, followed by 3% hydrogen peroxide treatment for 10 min. After blocking for 20 min in 5% BSA, tissue sections were incubated with anti-UCP1 antibody (Abcam; catalogue no. Ab10983; G s ␣ deficiency in MC4R-expressing cells leads to obesity 1:1000 dilution) at 4°C overnight. The sections were then washed with PBS and subsequently incubated with biotinylated goat anti-rabbit IgG secondary antibody (Agilent DAKO; catalogue no. E043201-6; 1:500 dilution) at room temperature for 1 h. UCP1 signal was detected with streptavidin horseradish peroxidase (Vector Laboratories) and visualized with diaminobenzidine tetrahydrochloride (Sigma). Tissue sections were counterstained with hematoxylin.

G s ␣ immunohistochemistry
Mice were anesthetized with avertin and transcardially perfused with cold PBS followed by cold 4% paraformaldehyde. After removal, brains were post-fixed in 4% paraformaldehyde overnight at 4°C and subsequently cryoprotected in 30% sucrose in PBS at 4°C for 3 days. Free-floating brain sections (40 m) were prepared using a microtome, and sections were stored at Ϫ80°C. For immunohistochemistry, brain sections were blocked in 2.5% normal horse serum (Vector Laboratories) plus 0.3% Triton X-100 at room temperature for 2 h and then incubated with anti-G s ␣ antibody (46) in blocking solution overnight at 4°C. After washing with PBS, brain sections were incubated with an Alexa Fluor-conjugated secondary antibody (Alexa Fluor 488, Life Technologies) for 1 h at room temperature. The signals in MC4R ϩ neurons in the PVN and DMV were captured and visualized using confocal microscopy (Carl Zeiss).

Statistical analysis
Data are presented as mean Ϯ S.E. Statistical significance was determined using unpaired t tests or linear regressions by analysis of covariance. Differences were considered significant if p Ͻ 0.05.