Distributed Neural Control of Energy Balance: Contributions from Hindbrain and Hypothalamus


Graduate Groups of Psychology and Neuroscience, University of Pennsylvania, Philadelphia, PA 19104. E-mail: grill@psych.upenn.edu


Data are reviewed that support the hypothesis that the neural control of energy expenditure is distributed among several brain sites. This view contrasts with that expressed most commonly in literature, that a single site—the arcuate hypothalamic nucleus—receives and integrates signals of relevance to energy status assessment and engages the effector circuits that orchestrate responses that maintain energy balance. The data reviewed support a contribution from medullary neurons, including those of the nucleus of the solitary tract, in the integration of signals of relevance to energy balance and in the issuing of commands to local behavioral and autonomic effectors. Experimental evidence is discussed that supports the following specific conclusions: hindbrain neurons integrate oral and gastrointestinal signals and issue commands to local motor circuits that control meal size; leptin's effect on food intake may be mediated, in part, by a direct action on the hindbrain neurons that respond to gastric distention; deprivation signals, such as the fall in leptin level, affect gene expression outside of the hypothalamus with reductions in proglucagon and proopiomelanocortin message seen in nucleus of the solitary tract-rich tissue; and that hindbrain neurons contribute to the control of energy expenditure seen with food deprivation and increases in expenditure after cold exposure or starvation. Future work is needed to define how the nucleus of the solitary tract and arcuate nodes of the central energy balance control network interact to collectively, or separately, influence specific aspects of energy balance control in the intact brain.


The dramatic increase in the prevalence of obese and overweight individuals has intensified interest in obesity as a major health problem. The overweight and obesity outcomes are attributed to the combined impact of increased high-energy food consumption and decreased physical activity. Atherosclerosis and type 2 diabetes are among the comorbidities of obesity. The costs of treating obesity and its related pathologies are considerable, with estimates in the vicinity of 60 billion dollars; this represents ∼6% of total U.S. health care costs. Feeding-related pathologies are not limited to those associated with excess food consumption. Also relevant to the focus of the “International Symposium on the Neurobiology of Obesity” are a variety of pathologies that result from insufficient food consumption and, like obesity, these pathologies bring humans of all ages into the health care system. Approximately 30% of admissions at the Children's Hospital of Philadelphia present with gastroesophageal reflex. A large percentage of these children require in-patient care for the reduced weight and the limited growth that results from the underconsumption of food that is central to their diagnoses (e.g., failure to thrive, food selectivity, and early satiety). Insufficient food consumption also brings the elderly into the health care system with dysgeusia (swallowing pathology) becoming a more common diagnosis. Adults, too, can be affected by a range of feeding pathologies, such as the weight loss resulting from early satiety, a symptom of dyspepsia and reflux, or the anorexia associated with various chronic disease states. What is remarkable is that despite the increased prevalence of these pathologies of feeding for humans of all ages, there are no effective pharmacological treatments currently available. To address this urgent need, the pharmaceutical industry relies on findings from basic research on the neurobiology of energy balance to guide its targets for drug development.

Hypothalamic Arcuate Model of Energy Balance

A number of dramatic discoveries made within the past 10 years have introduced to basic scientists a set of new peripheral signals and several novel neuropeptides. Intense research activity has highlighted the role of these signals and neuropeptides in energy balance physiology. The 1994 discovery of leptin (1) revolutionized thinking about 1) the neural systems that control energy balance and 2) the physiology of white adipose tissue. Leptin's multimodal contributions to energy balance are attributed to its action on central nervous system (CNS)1 receptors. While the functional leptin receptor (ObRb) is expressed on neurons in various brain regions, attention has focused almost exclusively on the ObRb-bearing neurons of the arcuate nucleus of the hypothalamus. These neurons also express receptors for other peripheral signals, such as those for ghrelin and insulin, whose levels, like those of leptin, vary with energy status. The arcuate nucleus contains two anatomically and functionally distinct populations of neurons—the neuropeptide Y/agouti-related peptide (NPY/AgRP) and the proopiomelanocortin (POMC)/cocaine- and amphetamine-regulated transcript (CART) neurons. A series of important discoveries (2, 3) showed that melanocortin receptor ligands [the agonist, α-melanocyte stimulating hormone α-MSH)], derived from POMC and the antagonist AgRP made in the brain affect feeding and energy expenditure through their actions on CNS melanocortin receptors (MC4-R and MC3-R). Mutations of the MC4-R result in hyperphagia and obesity in mice; similar effects on feeding and body weight are linked to mutations of the MC4-R in humans and account for a significant percentage of pediatric obesity (4). Additional support for a critical role of arcuate nucleus neurons in energy balance comes from studies showing that changes in energy status triggered by food deprivation or excessive caloric loading triggers alterations in the gene expression of AgRP and NPY and of POMC and CART (5, 6, 7). These and a variety of other data support the hypothesis that arcuate nucleus neurons receive blood-borne and neural signals of relevance to the assessment of energy status and that the integration of these signals is reflected in the excitability of these cells. Through direct and multisynaptic projection pathways—to caudal brainstem, spinal cord, and pituitary—the excitability of arcuate nucleus neurons is seen as directly controlling the three effector systems: behavioral, autonomic, neuroendocrine—whose activity maintains energy balance. This perspective, that a single site receives and integrates signals of relevance to energy balance control and issues commands to effectors, is described here as the arcuate model.

Distributed Model of Energy Balance Control

The localized control that is the hallmark of the arcuate model provides a backdrop for an alternate view of the neurology of energy balance. Our perspective on this critical regulatory system—one that governs survival and reproductive capacity—is that function is distributed across different brain regions with integrations made and commands to effectors issued by several sites within the brain. One extra-arcuate site, the nucleus of the solitary tract (NTS) of the dorsal medulla is considered here, but we acknowledge that other sites may also participate in a distributed system. With respect to the broad outlines of a feeding control system, we take the perspective that all signals that ultimately control food intake must engage hindbrain motor output circuits. It is well known that the final common path neurons to jaw, tongue, swallowing, and salivatory effectors, as well as the pattern generator circuits that gate the activity of these motor neurons to produce rhythmic patterns of chewing, licking, and swallowing, are localized to the caudal brainstem (CBS) nuclei. No controversy surrounds the notion that feeding behavior requires the activation of these CBS feeding motor circuits. Brain systems hypothesized to control energy balance must project to and operate on these CBS circuits to start, or end, a bout of feeding. The question this paper addresses is whether the integrations of relevance to energy balance control are performed by caudal brainstem neurons.

Caudal Brainstem Is a Site of Integration for Meal Size Control

There is broad agreement that signals arising in the gastrointestinal tract (mechanosensory, chemosensory, hormonal) and in the mouth are the principal determinants of meal size. These signals relayed to the brain through cranial nerve afferent fibers (e.g., vagus nerve) are processed at several CNS sites. The first central projection for both gastrointestinal and gustatory afferents is the NTS in the dorsal medulla (8). From the NTS, visceral afferent input is relayed rostrally to a variety of locations, including hypothalamic nuclei, for further processing. From a neural systems perspective, a key question is: which of the central sites integrating oral and gastrointestinal sensory information is required to drive the CBS motor output circuitry that controls meal size? Although the caudal brainstem is the entrance portal for visceral afferents responsive to gut and taste inputs and is also the brain region that contains feeding motor output circuitry, the general perspective of the field is that hypothalamic processing of visceral afferent information is required for meal size control. Known descending projections from hypothalamus to CBS are hypothesized to engage the feeding motor output. To examine the necessity of hypothalamic processing and to evaluate an alternate hypothesis—that endemic caudal brainstem processing is sufficient for meal size control, we performed a number of studies on chronically maintained brainstem transected rats (chronic decerebrate or CD rats). Patterns of rhythmic oral motor behavior (licking, chewing, swallowing) driven by oral chemical stimuli were found to be identical in decerebrate and pair-fed intact rats (9, 10). Qualitatively similar taste-driven oral motor responses from anencephalic and intact human neonates support the conclusions drawn from the rat data (11). Other studies showed comparable adjustments in the meal size of decerebrate and intact controls in response to treatments that 1) stimulate gastrointestinal afferents (gastric nutrient preloads, cholecystokinin injection) or 2) reduce gastrointestinal stimulation (sham feeding where ingested food drains through an open gastric cannula) (12, 13, 14). In the sham feeding study, increases in the sugar concentration of the liquid food increased the meal size of decerebrates and pair-fed intact rats comparably whether the ingested food drained from the cannula or emptied normally (13). Collectively, these data support the hypothesis that the neural circuits within the caudal brainstem are sufficient to integrate oral and gastrointestinal sensory signals that in turn engage endemic circuits that govern the size of individual meals.

Neural Mediation of Leptin's Effect on Meal Size

The neurology of meal size control is also referred to as short-term intake control. Some investigators use language of this sort to parse intake control function into phases including short- (meal size), intermediate- (between meals), long- (daily caloric intake), and longer-term (over days) phases. These temporal distinctions would be heuristically useful should evidence arise that indicates that for each of these hypothesized phases of regulatory control distinct signals operate on specific CNS circuits. There is, however, some limitation to these notions because the boundary between one phase and another can be difficult to discern. Consider, for example, the action of leptin and the neuropeptides that are hypothesized to mediate its effect (e.g., melanocortin) on energy balance. Leptin and its mediating neuropeptides are typically described as mediators of long-term intake control. Despite the long-term designation, it is clear that exogenously delivered leptin (or melanocortin receptor agonists) acts rapidly to reduce food intake through an action on meal size (15, 16). We now address where in the neuraxis leptin acts to affect meal size control and whether an action on extra-arcuate, caudal brainstem circuits might contribute to these effects?

The arcuate model localizes leptin's action on food intake to the NPY/AgRP and POMC/CART neurons of this nucleus (6, 7, 17). Intake reduction is viewed as the result of alterations in the excitability of these neurons that, in turn, affects the excitability of neurons in the feeding motor systems of the caudal brainstem. A variety of findings using a range of methods, however, show that leptin receptors are found on extra-arcuate neurons both within other hypothalamic nuclei and outside the hypothalamus. A systematic assessment of the contribution to energy balance that arises from leptin signaling at these extra-arcuate sites has not been undertaken. As mentioned above, gastrointestinal signals acting on CBS neurons are the principal determinants of meal size. One way that leptin's effect on meal size could be mediated is through an action on hindbrain gastrointestinal signal processing. Schwartz and Moran (18) showed that ventricularly delivered leptin alters the electrophysiological response of NTS neurons to the vagal mechanosensory inputs that arise from gastric distention. In the presence of leptin, lesser gastrointestinal loads produced greater electrophysiological response, and the function describing the effect of increasing gastric distention on the magnitude of the electrophysiological response was dramatically left-shifted. These data suggest that the observed reduction in meal size by leptin is achieved by amplifying the gastrointestinal afferent signals that directly control meal size. Where does the ventricularly delivered leptin act to produce these excitability effects? The authors (18) suggest that the observed effect is mediated by hypothalamic leptin signaling that, in turn, affects NTS excitability through descending projections. Leptin receptors, however, are also found on NTS neurons (19, 20), and low-dose leptin injected directly into this region reduced short- and longer-term food intake (19).

A series of collaborative experiments with Bjorbaek's laboratory addressed whether a leptin action on the NTS neurons that process gastric distention signals could contribute to its meal size effect. Systemic leptin injection rapidly stimulates cellular signal transducer and activator of transcription 3 (STAT3) phosphorylation, a marker of neurons that respond directly to leptin (21). Detailed anatomical analysis of the distribution of the pSTAT3 immunohistochemical (IHC) signal in the neurons of the NTS in response to leptin injection revealed that leptin signaling is highly localized to the neurons of the medial subnucleus at the level of the area postrema and is not found elsewhere in the rostro-caudal extent of the nucleus (22). The medial subnucleus contains neurons that process gastric distention signals and was the NTS recording site in the electrophysiological experiments described above (18). To determine whether leptin receptor is expressed on the same neurons that process gastric distention, we measured the expression of pSTAT3 IHC induced by peripheral leptin injection and c-Fos induced by gastric balloon distention. Double fluorescent IHC for pSTAT3 and c-Fos revealed that 39% of the leptin-responsive cells (pSTAT3 IR) responded to gastric distension (c-Fos IR) (23).

We hypothesize that these results are relevant to leptin's effects on meal size and are currently pursuing the neurochemical phenotype of the leptin-responsive NTS neurons. POMC neurons are localized to two sites in the CNS: the arcuate nucleus and the NTS. Arcuate POMC neurons express ObRb, but for a variety of reasons the coexpression of POMC and ObRb has not been evaluated for NTS neurons. NTS POMC neurons are notoriously difficult to detect by conventional methods because of low POMC expression in individual cells. To study potential leptin signaling in POMC neurons in the NTS, we took advantage of POMC-enhanced green fluorescent protein (EGFP) transgenic mice that express high levels of EGFP in POMC neurons in the NTS (24). As stated above, leptin injection stimulates cellular STAT3 phosphorylation in NTS neurons that are located in the medial subnucleus at the rostral-caudal level of the area postrema. In contrast, the majority of POMC neurons, as measured by expression of EGFP, were found caudal to the area postrema in the dorsomedial- and commissural-subnuclei of the NTS. No pSTAT3 signal was observed in POMC/EGFP neurons. We conclude that POMC neurons in the NTS of mice of this genotype do not respond directly to leptin, in contrast to those in the arcuate nucleus of the hypothalamus. This suggests that leptin action in the CNS through the melanocortin peptide system is mediated exclusively through signaling in POMC neurons located in the hypothalamus. Preproglucagon expressing neurons are also found in the NTS. These neurons are reported to express leptin receptor (25). Vrang et al. (26) showed that some of the NTS neurons expressing c-Fos after gastric distension were preproglucagon positive. It is not known whether the same preproglucagon neurons are driven by gastric afferent signals and by leptin. Obviously, neurons with other neurochemical phenotypes are present in this region, and their contribution to a mechanism underlying a hypothesized synergy between leptin and gastrointestinal fill remains to be explored.

Caudal Brainstem Contributions to Longer-Term Energy Balance

As mentioned earlier, the effects of taste and/or gastrointestinal stimulation on meal size are comparable in decerebrate and pair-fed control rats. Daily caloric intake, like meal size, seems to be a regulated parameter, because it is accurately defended against experimental perturbation in experiments in rats and other species (27). We now consider whether the caudal brainstem participates in this aspect of energy balance control as it does for meal size? Simple 24-hour food deprivation is routinely used to study the signals and the neural circuits that mediate the control of daily caloric intake. Several experiments studied whether food deprivation would comparably increase the meal size of decerebrates and control rats. In one, decerebrate and pair-fed intact rats were implanted with gastric cannulas, and their meals were sizes assessed under three conditions. In the reference condition, rats were fed normally, but before meal size assessment, the gastric cannulas were opened and the stomachs emptied of their contents. Meal size under the fed-empty stomach reference condition was compared with results from two different conditions: in one, rats received a gastric nutrient load before testing, and in the another, rats were food deprived for 24 hours before testing (food deprivation results in an empty stomach and gastrointestinal tract). With reference to the fed-empty stomach condition, nutrient preload comparably reduced the meal size of intact and decerebrate rats—consistent with results discussed above. Food deprivation increased the meal size of the intact rat but failed to affect the meal size of the decerebrates beyond that observed in the empty stomach-fed condition (14). These data suggest that, in the absence of connections between forebrain and caudal brainstem, the isolated CBS is insufficient to control motor circuits that adjust meal size in response to the sequelae of food deprivation. These data, however, do not rule out a contribution of CBS receptors or circuits in the isolated caudal brainstem to adjustments in energy expenditure triggered by food deprivation. In addition, these data do not rule out a contribution of CBS receptors and circuits within the intact brain to the processing of deprivation signals or to activating the effector pathways that mediate longer-term intake control.

Food deprivation is known to produce changes in neuropeptide gene expression in the arcuate nucleus. The highly reproducible changes in NYP/AgRP and in POMC/CART mRNA levels with food deprivation are often marshaled in support of the arcuate model of energy balance (5). The expression of these genes, and those for other neuropeptides that are hypothesized to play a role in energy balance control, is not limited to the arcuate nucleus. They are expressed in other regions of the brain both within and outside the hypothalamus. The effect of food deprivation on gene expression in these extra-arcuate regions has not been systematically studied. In collaboration with Bjorbaek's laboratory, we examined the effect of food deprivation on preproglucagon and POMC gene expression in NTS-rich tissue samples from the dorsal medial medulla of the mouse. Similar to results from arcuate-rich tissue samples, 48 hours of food deprivation significantly lowered POMC gene expression in hindbrain samples (22). Unlike the data from arcuate-rich samples, however, leptin administration (to reverse the fall in circulating leptin that results from food deprivation), in otherwise food-deprived mice, did not restore POMC mRNA levels to that seen in the fed state. This result is consistent with the absence of leptin signalling in NTS POMC/EGFP neurons, discussed above. This finding raises a question for future work: what factors other than leptin mediate the effect of deprivation on hindbrain POMC neurons? Food deprivation significant lowered preproglucagon gene expression in NTS-rich tissue. Furthermore, peripheral leptin injection elevated the preproglucagon mRNA levels of food deprived mice to levels observed in fed mice (22). These findings establish that caudal brainstem neurons, like those of arcuate hypothalamus, receive and respond to signals associated with food deprivation. It remains to be determined how the alterations in hindbrain neuron POMC and preproglucagon gene expression that arise with food deprivation contribute to the maintenance of energy balance.

Energy balance involves the coordinated control of two effector systems—that underlying energy intake and another that controls energy expenditure. In the intact brain, food deprivation triggers compensatory responses in both systems. In the absence of environmental sources of food, signals associated with food deprivation (such as the fall in leptin level) act on brain receptors to trigger a reduction in energy expenditure through an action on autonomic effectors. In collaboration with Harris and Bartness, the effect of food deprivation on energy expenditure was examined in chronic decerebrate rats (28). One week after decerebration, decerebrate and pair-fed intact rats were housed in an indirect calorimeter, and daily energy expenditure was measured for 4 days. One half of each of the CD and control groups was food deprived for 48 hours (water was provided). Responses to starvation were similar in CD and control rats; they reduced energy expenditure, decreased respiratory quotient (indicating lipid utilization), defended body temperature, mobilized fat, decreased serum leptin and insulin, and regulated plasma glucose. It is possible that, in intact animals, these responses are refined further by a contribution of more rostral brain areas or by the bidirectional communication between hypothalamic and hindbrain nuclei. Nevertheless, these data clearly show that the isolated caudal brainstem is sufficient to mediate many aspects of the energetic response to starvation that are attributed to hypothalamic signal processing (29). Support for this conclusion comes from other recent experiments with decerebrate rats. MC-R agonist applied to the hindbrain's fourth ventricle was shown to significantly elevate brown adipose tissue and core temperature. This result suggests that, in the absence of connections to and neural involvement of the hypothalamus, stimulation of hindbrain melanocortin receptors triggers endemic CBS sympathetic output circuits to engage heat production. When CD rats were exposed to the cold (8 and 12 °C) for a 6-hour period, they, like controls, maintained their core temperature. The defense of the decerebrate's core temperature involved dramatic increases in their heart rate and in their locomotor activity, indicating that cold afferent signals processed in the isolated brainstem can drive endemic autonomic and behavioral effectors (30).


Data were reviewed that support the hypothesis that the neural control of energy expenditure is distributed among several brain sites. This view contrasts with that expressed most commonly in a literature that focuses on the arcuate nucleus as the single locus of regulatory control for energy balance. The data reviewed support a contribution from medullary neurons including those of the NTS in the integration of signals of relevance to energy balance and in the issuing of commands to local behavioral and autonomic effectors. Experimental evidence is discussed that supports the following specific conclusions: hindbrain neurons integrate oral and gastrointestinal signals and issue commands to local motor circuits that control meal size; leptin's effect on food intake may be mediated, in part, by a direct action on the hindbrain neurons that respond to gastric distention; deprivation signals, such as the fall in leptin level, affect gene expression outside the hypothalamus with reductions in preproglucagon and POMC message seen in NTS-rich hindbrain tissue; and hindbrain neurons contribute to the control of energy expenditure with comparable reductions in expenditure seen with food deprivation and increases in expenditure with cold exposure. In addition to the arcuate and the NTS nodes of this regulatory system, there may be other nuclei that participate in the distributed control system including, for example, contributions made by ventromedial hypothalamus, nucleus accumbens, and orbito-frontal cortical neurons. The task ahead involves 1) studying the involvement of these and other sites in a distributed control network and 2) defining how the various sites implicated act in concert to achieve a energy balance.


The author thanks collaborators Christian Bjordaek, Ruth Harris, Tim Bartness, and Joel Kaplan. This study was supported by NIH-DK21397.


  • 1

    Nonstandard abbreviations: CNS, central nervous system; ObRD, functional leptin receptor; NPY, neuropeptide Y; AgRP, agouti-related peptide; POMC, proopiomelanocortin; CART, cocaine- and amphetamine-regulated transcript; αMSH, α-melanocyte stimulating hormone; MC-R, melanocortin receptor; NTS, nucleus of the solitary tract; CBS, the caudal brainstem; CD, chronic decerebrate; STAT3, signal transducer and activator of transcription 3; IHC, immunohistochemistry; EGFP, enhanced green fluorescent protein.