Recent research has identified a number of genes playing critical roles in the central regulation of energy homeostasis. Subsequently, models of the neurocircuitry regulating energy balance have been suggested, although their physiological relevance remains mostly untested. Using the Cre/loxP system, we can now genetically dissect these neurocircuits and establish the specific roles of these genes in small neuronal subpopulations. Here we focus on two receptors shown to be critical in the central regulation of energy homeostasis: leptin (LepR) and melanocortin-4 receptors (MC4R). Mice and humans deficient in either leptin or melanocortin signaling are severely obese. A prominent model of leptin action places the arcuate nucleus of the hypothalamus, and in particular arcuate proopiomelanocortin (POMC) neurons, at the center stage of energy balance regulation. By deleting LepR specifically from POMC neurons in mice, we showed that LepR on POMC neurons are required but not solely responsible for leptin's regulation of body weight homeostasis. Thus, LepR on other neurons must also be critically important in leptin-mediated regulation of body weight homeostasis. Data from MC4R-deficient mice have shown that MC4Rs regulate both sides of the energy intake/energy expenditure balance. Our recent experiments used MC4R-deficient mice with restored MC4R expression only in the paraventricular hypothalamus and a subpopulation of amygdala neurons. We showed that MC4Rs in the paraventricular hypothalamus and/or amygdala are sufficient to control food intake but that MC4Rs elsewhere control energy expenditure, thereby discovering the novel concept of functional and anatomical divergence of MC4Rs.
Diabetes, obesity, and related metabolic disorders are among the most pressing of today's health care concerns. The World Health Organization predicts a doubling in the prevalence of diabetes over the next two decades, fueled predominantly by the increasing incidence of obesity. To identify targets for new therapeutic agents, detailed understanding of the neuronal pathways controlling energy homeostasis is urgently needed. Recent research has started to unravel some of the major players regulating body weight balance, but the exact mechanisms by which peripheral signals are sensed by the central nervous system (CNS)1 and translated into coordinated responses remain largely unclear.
Neuroanatomy–Physiological Function Disconnect
In the past few years, we have gained tremendous insight into several aspects of the neuroanatomy and physiology behind the central regulation of energy homeostasis. Key genes encoding neuropeptides, neurotransmitters, and receptors have been discovered, and their localization in the CNS has been detailed. Furthermore, neuronal colocalization of the expression of these peptides, transmitters, and receptors has partially been established, giving us insight into the potential interplay of these genes. In addition, sophisticated techniques of monitoring neuronal activity of single cells or “microcircuits” in response to external stimuli have been established. Elegant viral tracing studies have shown connections between neurons in different structures of the CNS and their connection to peripheral organs. This invaluable wealth of neuroanatomical information has led to speculations on the physiological function of genes and neurocircuits. Speculations on function have also been based on data from lesion studies and stereotaxic injections of pharmacological agents. However, lesion studies usually destroy CNS areas of complex neuronal composition and connection, and whereas pharmacological injection studies can identify particular areas of the brain as functional sites of agonist action, it remains unclear to what degree these accurately reflect endogenous signaling events. Thus, although the neuroanatomical evidence may give us clues about the particular function of a gene or neurocircuit, the true link between neuroanatomy and physiological function remains largely unresolved.
Creating That Connection
Recent advances in the technologies of genetic manipulation in rodents now allow us to manipulate genes in small, specified subsets of neurons or neuroanatomical locations. Using the available neuroanatomical and pharmacological clues, we can decide where and how to perturb specific neurocircuits of a suggested physiological function and then use functional genomics approaches to directly test their physiological relevance. Using the Cre/loxP system, researchers have begun to establish the function of genes expressed in defined neuronal subsets of the CNS by studying the physiological outcome of the genetic perturbation in the whole, awake unrestrained animal (1, 2). In a complex organ such as the CNS, cells directly adjacent to each other can express very different sets of genes. The Cre/loxP system has the distinct advantage that manipulation of the gene of interest can be directed to very specific subsets of neurons by the use of specific promoters to drive Cre-expression, rather than being limited to an anatomically defined area, as is the case for any kind of stereotaxic injection. The only limitation to how defined a manipulation can be achieved is the availability of neuronal-specific Cre-recombinase–expressing mice, and that list is continuously growing.
Admittedly, these physiological genomics studies are not without caveat, because the body has a tremendous ability to defend itself against perturbation of particular genes or anatomical areas. Compensatory events can thus mask some of the true physiological functions of genes. However, effects that are observed despite the body's ability to compensate for perturbations must thus be extraordinarily important. Furthermore, the generation of better Cre-inducible systems will allow us to avoid developmental compensatory mechanisms by deleting or reactivating genes in later stages of development and will add precious information about the function of genes in adult animals.
Lesion studies have shown that the hypothalamus is one of the CNS areas essential to the control of energy homeostasis. Specific lesioning of the arcuate nucleus of the hypothalamus (ARC), paraventricular hypothalamus (PVH), or ventromedial nucleus (VMH) of the hypothalamus leads to obesity with hyperphagia (3). The hypothalamus receives input from a number of peripheral signals, including the adipocyte-derived hormone leptin, pancreatic insulin, and gut hormones, such as ghrelin and peptide YY. It integrates this information and relays it to several other CNS sites, including the brain stem and spinal cord. A well-studied example is the adipocyte-derived hormone leptin. Mutation of leptin (ob) or the leptin receptor (LepR; db) leads to obesity in humans and rodents (4, 5, 6, 7, 8). While LepR is expressed in many peripheral tissues as well as the CNS, it was physiological genomics (neuronal-specific deletion of LepR) (2) and pharmacological (intracerebroventricular injection of leptin) (9) approaches that showed leptin's anorexigenic effects being mediated entirely by LepR in the CNS. The signaling form of the LepR is expressed in several areas of the brain including the hypothalamic ARC, VMH, and the dorsomedial hypothalamus (10), but most studies on leptin-mediated regulation of energy homeostasis have focused on the ARC. Indeed, the ARC itself is required for leptin's anorexic effects, because ARC-lesioned Lepob/ob mice are resistant to leptin's effects on body weight (11).
In the ARC, leptin activates the anorexigenic cocaine- and amphetamine-regulated transcript/proopiomelanocortin (POMC)-expressing neurons (12), whereas it inhibits the orexigenic neuropeptide Y/agouti-related peptide (NPY/AgRP)-expressing ARC neurons (13). In addition to the direct effect of leptin on these first-order leptin-responsive neurons, POMC neurons are also indirectly inhibited through the actions of γ-aminobutyric acid and NPY released from leptin-responsive NPY/AgRP neurons synapsing onto POMC neurons (14). AgRP and α-melanocyte stimulating hormone (α-MSH; a proteolytic product of the POMC gene) bind to melanocortin-4-receptors (MC4Rs) on so-called second-order neurons; whereas α-MSH activates the MC4R and decreases food intake, AgRP inhibits it, and overexpression of AgRP leads to hyperphagia (15).
Apart from the presence of leptin-responsive LepR-expressing neurons in other CNS sites and hindbrain injections of leptin causing a reduction in food intake in rats (16), there has been little evidence for a physiological role of leptin in energy homeostasis in other sites. This has led to the idea that leptin signaling in ARC POMC and AgRP neurons mediates most of leptin's effects on energy homeostasis. Indeed, adenoviral expression of LepR in the ARC of LepR-deficient rats ameliorated their obesity (17). However, the physiological importance of endogenous LepR signaling in the ARC and in specific ARC cell types remained to be established. Considering the important role of α-MSH in the regulation of body weight homeostasis (18) and that ARC lesions cause obesity, not leanness (3), the POMC neuron was favored as the main regulator of energy homeostasis in the ARC.
Testing (and Abandoning) the Leptin–ARC Model
We set out to definitively test the hypothesis that leptin's anorexigenic effects are mediated by ARC LepR signaling. Using recombineering techniques (19), we generated bacterial artificial chromosome transgenic mice expressing Cre-recombinase specifically in POMC neurons and crossed these with mice bearing a loxP-flanked exon 17 of the LepR, resulting in deletion of functional LepR only from POMC neurons (20). (As a side note, recombineering techniques allow fast manipulation of large genomic sequences without the use of restriction enzymes and, thus, generation of bacterial artificial chromosome transgenic and loxP-manipulated mice in a very flexible and time-efficient process). Mice lacking LepR on POMC neurons had a mild obesity phenotype that was only ∼20% of that observed in mice lacking all LepR signaling (20). The increased body weight was a result of an increase in fat mass in conjunction with hyperleptinemia. We hypothesized that this hyperleptinemia was being sensed by other LepR-bearing neurons, which would signal to curb body weight increases. Indeed, mice lacking LepR on POMC neurons have decreased hypothalamic AgRP mRNA expression. Our data showed that LepR on POMC neurons are required for normal body weight homeostasis. However, even if the degree of body weight increase in POMC-specific LepR knockout mice is an underestimate because of compensatory mechanisms, it is in blatant disparity to the massive obesity observed in mice lacking all LepR. These data led us to conclude that LepR on other first-order, leptin-response neurons must also contribute importantly to leptin's anorexigenic effects. As LepR is expressed widely throughout the CNS, these could be within the ARC, other areas of the hypothalamus, or anywhere outside the hypothalamus.
We recently found that mice lacking LepR from AgRP neurons, POMC neurons, or both AgRP and POMC neurons all have similar body weight phenotypes of mild obesity (unpublished data). These data lead us to abandon the ARC-focused view of the importance of leptin's anorexigenic action. Indeed, Dhillon et al. (21) recently showed that steroidogenic factor 1 neurons in the VMH are first-order, leptin-responsive neurons. Removal of LepR signaling from these VMH neurons resulted in a mild increase in body weight, and interestingly, also increased sensitivity to diet-induced obesity. Thus, leptin action at this widely ignored CNS site plays an important role in body weight regulation and diet-induced thermogenesis. Collectively, our data showed that the previous view of the ARC as the main site for leptin-mediated effects on body weight homeostasis is incorrect and needs, at the minimum, to be amended to include the VMH and to acknowledge that leptin's anorexigenic effects are also mediated by extrahypothalamic sites (Figure 1). One among many candidate sites is the nucleus of the solitary tract (NTS) in the hindbrain, where leptin injections have been shown to reduce food intake and body weight (16).
Collectively, these neuronal-specific manipulations of leptin signaling give us a powerful insight into the mechanism of leptin-mediated control of body weight. Clearly, leptin's action to control body weight homeostasis is distributed over a number of different neuronal cell types and CNS areas. Importantly, we have been able to assign functions to LepR on specific cell types; whereas LepR on VMH neurons plays an important role in protecting animals from diet-induced obesity, LepR on POMC neurons has no such function (unpublished data). Furthermore, at least in high-fat diet feeding, the two arms of the energy balance, i.e., food intake and energy expenditure, are controlled by LepR on the same neuronal cell type. CNS regulation of energy homeostasis must comprise a combination of sensory, afferent components, integrative components, and efferent components, splitting at some point into the control of food intake and energy expenditure; LepR seems to be part of the integrative component, because its signaling in specific CNS areas affects food intake and energy expenditure.
PVH MC4Rs: An Example of Inferring Function from Neuroanatomy
Directly downstream of POMC and AgRP neurons lie the MC4R-expressing neurons; we set out to identify key MC4R-expressing neuronal subpopulations in the regulation of energy homeostasis. A major role for α-MSH and the MC4R in the control of body weight homeostasis is supported by the fact that loss of either POMC or MC4R leads to severe obesity with hyperphagia, hyperinsulinemia, and increased nasal-anal length in humans and mice (18, 22, 23, 24, 25, 26). MC4R mRNA expression is distributed in several specific areas and nuclei throughout the CNS (27, 28). These include two areas most neuroanatomical and pharmacological studies have focused on, the PVH and nuclei in the hindbrain, such as the NTS, dorsal motor nucleus of the vagus (DMV), and intermediolateral column of the spinal cord (IML) (27, 28). Because of these neuroanatomical data, we chose to initially focus on PVH MC4Rs.
The PVH regulates a variety of neuroendocrine and autonomic functions and can be divided into magnocellular and parvocellular subdivisions. The magnocellular neurons project to the posterior pituitary and show little MC4R mRNA (27, 28). The dorsal, ventral, and lateral parvocellular subdivisions of the PVH, however, show intense labeling for MC4R mRNA. These particular PVH neurons project to structures involved in central autonomic control, including parasympathetic preganglionic neurons in the DMV and sympathetic preganglionic cell columns in the spinal cord.
The PVH has long been implicated in the control of feeding behavior, because lesions in the PVH cause profound obesity (3). Melanocortin receptor agonist melanotan II (MTII) injections into the PVH inhibit nocturnal feeding and increase energy expenditure (29), whereas MC4R antagonist PVH injections exert opposite effects (30, 31). Experiments in MC4R knockout mice had also shown that the MC4R is required for appropriate responses to a high-fat diet in terms of energy expenditure and activity (32). Brown adipose tissue is thought to regulate energy expenditure, and it receives relayed inputs not only from the IML, but also from many other CNS areas including the PVH (33). Furthermore, it has been shown that MC4R knockout mice develop hyperinsulinemia and insulin resistance before the onset of obesity (34). In addition to MC4Rs playing a role in insulin secretion, they are also required for normal insulin sensitivity in peripheral tissues, independent of their effects on energy homeostasis (35). Viral tracing studies have revealed that the pancreas receives relayed input from a number of CNS sites, including the PVH and the IML (36). The anatomical location and connections of the PVH, expression of MC4Rs in specific PVH subdivisions, and the pharmacological MC4R agonist and antagonist injection data led to the hypothesis that MC4Rs in the PVH control food intake and also provide descending inputs to preganglionic neurons, contributing to melanocortin's autonomic effects, such as regulation of energy expenditure and insulin secretion. We aimed to test the accuracy of ascribing function by inference from neuroanatomical and pharmacological evidence using a loxP-modified MC4R null mouse model.
Testing the PVH MC4R Hypothesis
To relate signaling of MC4R in specific brain regions to specific physiological consequences within the obesity phenotype, we generated a mouse model with a loxP-modified, MC4R null allele, which is reactivatable by Cre-expression through the excision of a loxP-flanked transcriptional blocker (loxTB MC4R) (37). As expected, mice homozygous for the loxP-modified, MC4R null allele do not express MC4Rs and are obese (37). We chose this model of Cre-mediated reactivation over Cre-mediated ablation of the MC4R, because if some MC4R-mediated pathways were redundant, ablation of such MC4R pathways would lead to little or no effect on body weight homeostasis. However, reactivation of functionally important MC4Rs, in an animal that is otherwise devoid of MC4Rs, is expected to be a more sensitive assay of MC4R function and have a significant effect on body weight.
To reactivate MC4Rs specifically in the PVH, we generated mice expressing Cre-recombinase driven by the Sim1 (single-minded 1) promoter, a hypothalamic developmental transcription factor strongly expressed throughout the PVH (37, 38). Crossing these Sim1-Cre transgenic mice with the loxTB MC4R mice reactivated MC4Rs in the PVH, nucleus of the lateral olfactory tract (a structure of the olfactory amygdala subdivision), scattered cells of the medial and basomedial amygdala, and the posterior hypothalamus; we were able to exclude the posterior hypothalamus as a key MC4R site in the regulation of energy homeostasis, because posterior hypothalamus-specific reactivation of MC4Rs had no effect on body weight (37). These mice with MC4R expression in the PVH and amygdala only had a 60% amelioration of their obesity phenotype compared with loxTB MC4R littermates. Interestingly, this body weight reduction was entirely caused by rescued hyperphagia, with no effect of PVH/amygdala MC4Rs on energy expenditure. This was confirmed by intraperitoneal injection of the MC4R agonist MTII and, thus, acute activation of MC4Rs in the PVH and amygdala only: whereas the reduction in food intake was comparable to wild-type, there was no increase in energy expenditure in response to MTII injection. We concluded that, contrary to our initial predictions, MC4Rs in the PVH and amygdala are not involved in mediating melancortin's autonomic effects on energy expenditure. However, PVH/amygdala MC4Rs alone are sufficient in mediating all of melanocortin's effects on food intake, independent of MC4Rs in other CNS areas. This divergence of melanocortin pathways in the control of food intake vs. energy expenditure was a surprising and novel finding (Figure 1). These findings also showed that agonist injection and neuroanatomical data alone did not allow accurate predictions of MC4R function in a specific CNS area. Whether the PVH/amygdala is the only MC4R-expressing area responsible for the regulation of food intake or whether other areas of MC4R expression are also able to control food intake remains to be determined.
If PVH/amygdala MC4Rs are not involved in the regulation of energy expenditure, which areas of MC4R expression might be? Candidate sites include sympathetic preganglionic neurons in the IML (the final common pathway for sympathetic activation) and various CNS sites projecting to the IML, such as the NTS or the raphe pallidus. Indeed, recent data in decerebrate rats have shown that the caudal brainstem is sufficient to mediate the reduction in energy expenditure in response to starvation (39), as well as fourth ventricle administration of MTII stimulating uncoupling protein 1 expression in brown adipose tissue in decerebrate rats (40).
In future experiments, we will be able to study the physiological roles of other CNS areas of MC4R expression, using various transgenic mouse models of Cre-expression. In addition, we will be able to define which particular PVH/amygdala neuronal MC4R subpopulation may be responsible for MC4R effects on food intake by creating more and more refined Cre-expressing mice.
In contrast to LepR, which was found to regulate both food intake and energy expenditure in one CNS area, MC4Rs in the PVH/amygdala are involved in only one arm of the energy balance. Thus, the LepR seems to be part of the integrative component of energy homeostasis, while the MC4R is part of the efferent component that splits into the control of food intake and energy expenditure.
In summary, our recent data have shown that, by the use of functional genomics approaches in neuroanatomy (neuroanatomics), i.e., using genetic manipulation in mice, we have been able to draw a more refined and accurate picture of the central control of energy homeostasis (Figure 1). We have shown that we need to move away from the assumption that leptin's anorexigenic effects are mediated by ARC neurons or even just hypothalamic neurons alone, although these other functionally important leptin-responsive areas still require identification. We have further shown that MC4Rs in the PVH/amygdala only control the food intake arm of the energy balance, revealing a previously unknown divergence of melanocortin pathways in the control of energy homeostasis. Other areas of MC4R expression regulating the energy expenditure arm of the energy balance are currently being studied.
These studies are just the tip of the iceberg in defining the true neuronal pathways regulating energy homeostasis. Using the Cre/loxP system, we will be able to ascribe physiological function to specific CNS areas, neuropeptides, transmitters, and receptors in specific neuronal subpopulations, not only in the field of energy homeostasis, but also in other complex behaviors and physiologies.
The author thanks Bradford B. Lowell (Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA), Joel K. Elmquist (Southwestern Medical School, Dallas, TX), and Roberto Coppari (University of Ancona, Ancona, Italy), as well as members of the Lowell and Elmquist laboratories for support and helpful discussions. N.B. was funded by The Wellcome Trust, American Diabetes Association/European Association for the Study of Diabetes, Boston Obesity and Nutrition Research Center, and the American Heart Foundation.
Nonstandard abbreviations: CNS, central nervous system; ARC, arcuate nucleus of the hypothalamus; PVH, paraventricular hypothalamus; VMH, ventromedial nucleus; LepR, leptin receptor; POMC, proopiomelanocortin; NPY, neuropeptide Y; AgRP, agouti-related peptide; αMSH, α-melanocyte stimulating hormone; MC4R, melanocortin-4-receptor; NTS, nucleus of the solitary tract; DMV, dorsal motor nucleus of the vagus; IML, intermediolateral column of the spinal cord; MTII, melanocortin receptor agonist melanotan II; Sim1, single-minded 1.