AgRP neurons: Regulators of feeding, energy expenditure, and behavior

Abstract

Neurons in the hypothalamic arcuate nucleus (ARC) that express agouti-related peptide (AgRP) govern a critical aspect of survival: the drive to eat. Equally important to survival is the timing at which food is consumed—seeking or eating food to alleviate hunger in the face of a more pressing threat, like the risk of predation, is clearly maladaptive. To ensure optimal prioritization of behaviors within a given environment, therefore, AgRP neurons must integrate signals of internal need states with contextual environmental cues. In this state-of-the-art review, we highlight recent advances that extend our understanding of AgRP neurons, including the neural circuits they engage to regulate feeding, energy expenditure, and behavior. We also discuss key findings that illustrate how both classical feedback and anticipatory feedforward signals regulate this neuronal population and how the integration of these signals may be disrupted in states of energy excess. Finally, we examine both technical and conceptual challenges facing the field moving forward.

Introduction

To coordinate the behavioral and physiological responses necessary to maintain energy homeostasis, the central nervous system (CNS) must rapidly and precisely calculate the needs of the body throughout varying environmental conditions. A central representation of the internal state is achieved through bidirectional communication between the CNS and the body, or interoception [1]. To optimally direct behavior, representations of the internal state must be counterbalanced by signals of the outside world provided by exteroception that indicate impending changes in energy demand. The relevance of this bidirectional relationship between the periphery and the CNS is essential for even a seemingly simple decision by an animal—whether to expend effort to forage for food.

Interoceptive AgRP neurons are critical to producing a central representation of hunger [2]. The decision to forage for food, driven by AgRP neurocircuits, is based upon this formed representation and informed by myriad environmental factors such as ambient temperature, risk of predation, and competing demand for social interactions that are balanced against a host of internal factors (e.g., energetically costly processes such as growth and reproduction) [3–7]. AgRP neurons function by integrating internal signals of energy state with external signals that impact current or future energy state, including sensory input relaying food attributes and availability, competitive motivational drives, and environmental cues of energy demand, such as temperature. In turn, activated AgRP neurons promote food-seeking, in part, by suppressing distracting motivators, including innate fears and pain that might keep an animal from searching for food. Activation of these neurons further suppresses the expression of energetically expensive social behaviors, such as mating and fighting, until animals locate food. Although AgRP neurons are commonly referred to as ‘hunger-driving’, and their activation unmistakably drives food intake, this reference oversimplifies the broad control they have over many primal behaviors and essential physiological processes. When and to what magnitude AgRP neurons engage downstream neurocircuits in response to an ever-changing environment is only beginning to be understood. This review describes recent advances that have provoked a new understanding of the central role AgRP neurons play at the system, network, and behavioral levels in the regulation of energy balance.

AgRP Neurons: neural substrates of hunger

The melanocortin system

In addition to AgRP, AgRP neurons coexpress both neuropeptide Y (NPY) and the neurotransmitter gamma-aminobutyric acid (GABA) and are exclusively located in the ARC [21]. Due to their proximity to the median eminence (a circumventricular organ), AgRP neurons have greater access levels of circulating hormones and nutrients than regions of the brain protected by a more fully formed blood–brain barrier. This privileged anatomical position enables AgRP neurons to sense circulating signals conveying levels of both readily available energy (e.g., free fatty acids and glucose) and energy stores (e.g., fat depots). Broadly, AgRP neurons are activated by hormonal and neural signals of heightened energetic demand, and they stimulate feeding, in part, by antagonizing postsynaptic MC3R/MC4Rs through the synaptic release of AgRP, a biased agonist of melanocortin receptors (Fig. 1) [9,22]. Within the ARC and juxtaposed to AgRP neurons are neurons that express the precursor polypeptide POMC. These neurons release multiple cleavage products of POMC, including α-MSH, that bind to and activate downstream melanocortin receptors (MC3R/MC4R) to suppress feeding and enhance energy expenditure [15]. Alongside their disparate effects on energy balance, POMC and AgRP neurons are reciprocally regulated by circulating cues of nutritional state, including the adiposity hormones leptin and insulin. Leptin is secreted from adipocytes in proportion to levels of body fat stores [23] and acts on its receptor (LepR) within the brain [24] to suppress food intake and increase energy expenditure [25]. As genetic deficiency of leptin (ob/ob) or its receptor (db/db) produces profound hyperphagia and obesity, enormous efforts have been invested in elucidating the neural mechanisms by which central leptin signaling regulates food intake and energy expenditure. Among central targets of leptin signaling, AgRP and POMC neurons are the best studied. In contrast to POMC neurons, which are directly depolarized by leptin [26], AgRP neurons are inhibited by both leptin and insulin [27,28], whereas they are activated by the gastric hormone ghrelin, which increases prior to a meal [29]. Therefore, in conditions of reduced leptin signaling, such as fasting or diet-induced weight loss, AgRP neurons are activated, while POMC neurons are inhibited. POMC neurons are also directly inhibited by AgRP neurons through the release of GABA [26]. This combination of responses promotes positive energy balance and the recovery of lost weight (Fig. 1).

Fig. 1.

Central Melanocortin System. The central melanocortin system includes two distinct neuronal populations located in the arcuate nucleus of the hypothalamus (ARC) that express either agouti-related peptide (AgRP) or pro-opiomelanocortin (POMC). Neurons expressing POMC are inhibited by insulin and leptin and release alpha-melanin-stimulating hormone (α-MSH), an agonist that binds to downstream melanocortin-4 receptor (MC4R)-expressing neurons in the paraventricular nucleus of the hypothalamus (PVH^MC4R) to inhibit food intake and induce weight loss. In contrast, AgRP neurons are inhibited by insulin and leptin and are activated by the stomach-derived hormone, ghrelin, and stimulate food intake and reduce energy expenditure when activated. These neurons also express neuropeptide Y (Npy), and the inhibitory neurotransmitter, gamma aminobutyric acid (GABA). AgRP neurons mediate their effects through the release of AgRP, which antagonizes the binding of α-MSH to PVH^MC4R neurons, Npy acting on downstream Y1 and Y5 receptors, and by inhibiting POMC neurons via synaptic release of GABA.

AgRP neurons are necessary and sufficient for food intake

The contribution of AgRP neurons to energy homeostasis was initially challenged by evidence that embryonic deletion of Npy [30], Agrp, or both did not alter body weight or food intake [31]. However, as neurocircuit connectivity and function are inherently shaped by genetics and experience, compensation across development can often mask a particular gene’s importance [32]. To test the necessity of AgRP neurons for feeding without the possibility of developmental compensation, the human receptor for diphtheria toxin was expressed specifically in AgRP neurons so that the timing of diphtheria dosing could provide temporal control over ablation. The loss of AgRP neurons using this model was first examined in the neonatal mouse, where their ablation produced little effect on food intake and body weight [33]. In stark contrast, ablation of AgRP neurons in the adult mouse produced eventual death as, without them, the animal will not eat and cannot be motivated to do so [33,34]. Additional models of AgRP neuron ablation leading to either progressive postnatal cell death [35] or incomplete ablation (< 50%) [36] corroborated these findings. Given this critical function in the adult, it is striking how the neonatal brain can adapt to AgRP neuronal loss, and research into the plasticity of feeding circuits during development is a growing field [37,38]. Although the ablation models demonstrate the necessity of AgRP neurons for survival, the application of tools aimed at controlling AgRP neuron activity is required to study the dynamics of hunger.

Our understanding of AgRP neurons’ role in feeding accelerated with the application of both optogenetics [39] and chemogenetics approaches [40]. By applying these technologies, multiple groups found that activation of AgRP neurons is sufficient to stimulate voracious feeding [41–43]. Notably, this effect occurs even in sated mice, without training, and at any time of day. In optogenetic models, feeding responses scale with stimulation frequency, laser power, and the number of AgRP neurons expressing the channelrhodopsin transgene [41]. Moreover, chronic activation using either chemogenetics [43] or expression of a bacterial sodium channel [44] causes marked hyperphagia and weight gain, although a recent study suggests that these effects are moderated as excess weight is gained [45]. Conversely, chemogenetic inhibition reduces food intake [43], an effect that underscores the importance of this population of neurons in regulating food intake.

Neuromediators: AgRP, NPY, and GABA

The capacity of AgRP neurons to stimulate food intake is made possible by the synaptic release of three neuromediators: AgRP, an endogenous biased agonist of MC3R/MC4R; NPY, an agonist for the Gi-coupled NPY family of receptors (NPY1–5R); and the inhibitory neurotransmitter, GABA, an agonist for the ionotropic GABA_B and metabotropic GABA_A receptors [46]. NPY is the most prevalent neuropeptide in the brain [47] and is one of the most potent orexigenic agents known. Delivering NPY directly into the ventricular system of the brain (intracerebroventricular (icv)) [48] rapidly and robustly increases acute (~ 1 h) food intake, in part, via NPY Y1 and Y5R signaling in the hypothalamic paraventricular nucleus (PVH), a major target of AgRP neuron innervation [49,50]. When administered chronically, intra-PVH NPY induces obesity [50]. By comparison, a single icv injection of AgRP augments food intake for up to a week [51], illustrating the complementary but temporally distinct effects of NPY and AgRP on food intake. This redundancy may help explain the lack of a body weight phenotype in mice lacking either Npy or Agrp [30,31,51], which highlights the role of GABA release in the regulation of food intake by AgRP neurons. In support of this hypothesis, direct application of the GABA_A receptor agonist, muscimol, to the PVH stimulates feeding in rats [52]. In contrast, genetic deletion of the vesicular GABA transporter, Vgat, from AgRP neurons (thereby precluding GABAergic neurotransmission from this population) produces a lean phenotype and attenuates diet-induced obesity (DIO) in mice [53].

Recent studies have sought to resolve how the bioactive molecules released from AgRP neurons contribute to feeding. During chemogenetic stimulation of AgRP neurons, targeted antagonism of either NPY1R or GABA_A signaling at the PVH blocked acute (1-h) feeding responses [42]. Thus, NPY and GABA signaling in the PVH contribute to feeding evoked by chemogenetic AgRP neuron activation. Similarly, combinatorial genetic mouse models lacking Mc4r, Npy, and Vgat reveal that loss of both Vgat and Npy delayed acute but not chronic increases in food intake following activation of AgRP neurons [54]. However, this deficit was recovered over a 24-h period, suggesting that AgRP signaling via MC4R mediates long-term increases in food intake [54], consistent with extensive literature implicating the melanocortin system in the regulation of energy homeostasis [15]. Optogenetic paradigms show that, food intake is sustained after the stimulating laser is turned off [41,55], and this prolonged food intake drive depends upon intact NPY signaling [56]. Indeed, loss of NPY specifically from AgRP neurons both delays and blunts food intake following either opto- or chemogenetic activation [56,57]. Thus, whereas NPY and GABA rapidly modulate food intake, AgRP signaling via MC4R drives sustained increases in food intake induced by AgRP neuron activation.

While these studies have greatly improved our understanding of downstream NPY, GABA, and AgRP signaling and food intake, the contribution of these three neuromediators to other features of hunger regulated by AgRP neurons (see section 1.4) is just beginning to be explored. Furthermore, as we discuss in the next section, AgRP activation also influences other aspects of energy and glucose homeostasis independent of feeding. Clarifying the roles of NPY, GABA, and AgRP in these diverging effects is an important goal for the field.

AgRP neurons in energy homeostasis

AgRP neurons regulate energy expenditure and substrate utilization

A common frustration for dieters trying to lose weight is the marked, compensatory reduction in energy expenditure [58] associated with fasting or caloric restriction [59]. Early studies implicated AgRP and NPY in food intake and in this adaptive response aimed at conserving energy. NPY, given either icv or intra-PVH, is associated with reduced energy expenditure, an effect due, in part, to reduced sympathetic outflow to thermogenic tissues such as brown adipose tissue (BAT) [60]. Comparably, icv administration of AgRP reduces energy expenditure and increases body adiposity, independent of changes in food intake [61]. Chemogenetic studies corroborated these early pharmacological studies and emphasized the role of AgRP neurons in the suppression of energy expenditure: AgRP activation not only rapidly increased the respiratory exchange ratio (RER), indicating elevated carbohydrate utilization and reduced lipolysis [43,54,62–64], but also suppressed the thermogenic program of white fat [65]. These findings support the hypothesis that the fasting-evoked activation of AgRP neurons promotes the conservation of energy by suppressing sympathetic outflow to thermogenic tissues. The importance of this mechanism to energy homeostasis is captured by models of constitutive AgRP neuron activation, in which heightened lipogenesis exacerbates the obesogenic effect of high-fat diet (HFD) feeding [64]. These studies show that activated AgRP neurons shift metabolism toward energy conservation and lipid storage, in part, by reducing energy expenditure in thermogenic tissues. Weight loss therapeutics aimed at preventing this shift to lipogenesis and reduced metabolic rate, despite restricted caloric intake, offer a promising avenue to mitigate the homeostatic counterregulatory responses to weight loss.

AgRP neurons in glucoregulation

The role of AgRP neurons in glucose homeostasis is consistent with their regulation by nutrient- and hormonal-related signals, including insulin, leptin, and glucose. Early pharmacological studies found that central administration of NPY causes peripheral insulin resistance independent of changes in food intake. This effect was partly attributed to decreased insulin-induced suppression of hepatic glucose production (HGP) [66,67], a critical determinant of fasting blood glucose levels. While the loss of insulin receptors from AgRP neurons also impairs hepatic insulin sensitivity [68], chemo- and optogenetic activation of AgRP neurons induces insulin resistance by impairing insulin-stimulated glucose uptake into BAT, without effects on HGP [69]. Interestingly, insulin resistance induced by artificial AgRP neuron activation does not occur in NPY genetic knockout mice. Since this observation is reversed by ARC-specific NPY re-expression, we infer NPY signaling plays a key role in driving glucose metabolic impairment induced by activation of these neurons [57]. Consistent with this interpretation, activation of POMC neurons does not appear to alter insulin sensitivity or BAT glucose uptake [69], suggesting that these metabolic responses are driven by changes in NPY rather than melanocortin signaling.

One caveat to the interpretation of these types of experimental findings is that under natural conditions, AgRP neuron activation is typically associated with POMC neuron inhibition and vice versa, as noted earlier, so the physiological insights that can be drawn from activation of one neuronal subset in isolation are limited. For example, in many rodent models of diabetes, hyperglycemia is associated with both AgRP neuron activation and POMC neuron inhibition. This combination reduces net melanocortin signaling, an effect that favors elevation of the defended blood glucose level.

Beyond feeding: AgRP neurons and the features of hunger

Under normal circumstances, animals prefer to stay in safe, familiar environments and avoid potential dangers and threats. However, in conditions of energy deprivation, the perceived value of food increases, and animals are more willing to enter potentially dangerous situations as the risk/benefit analysis favors food-seeking behavior [4]. AgRP neurons are activated in the fasted state, and their activation, using either chemogenetic or optogenetic means, recapitulates many of the behaviors and physiological effects associated with starvation, in addition to stimulating food intake [3,70]. These include foraging behaviors such as digging and marble burying [70], the willingness and motivation to work for food (as assessed by lever pressing or progressive ratio test), and enhanced rewarding properties of food (assessed by conditioned place preference test) [43,55,71]. Moreover, AgRP neuron activation promotes anxiolytic responses that increase willingness to explore novel and potentially threatening environments to seek food [70,72], including those associated with an intruder or the presence of a predator [3,72]. These findings suggest that the effect of energy deprivation to activate AgRP neurons results in increases of both appetitive and consummatory drives that are mediated in part by suppressing or overriding the activity of brain areas involved in competing behavior states such as fear and anxiety that might otherwise prevent an animal from searching for food [4–6].

Postsynaptic targets of AgRP neurons

AgRP neurons project to multiple hypothalamic subnuclei, including the lateral, ventromedial, and dorsal hypothalamic areas (VMH, DMH, and LHA, respectively), preoptic area (POA), suprachiasmatic area (SCN), and the PVH. AgRP projections also extend beyond the borders of the hypothalamus, innervating the anterior bed nucleus of the stria terminalis (aBNST), central and medial nuclei of the amygdala (CeA and MeA, respectively), paraventricular thalamic nucleus (PVT), lateral septum (LS), periaqueductal gray (PAG), and parabrachial nucleus (PBN) [73–75]. An interesting characteristic of AgRP projections is that they are noncollateralizing [74], implying that distinct subsets of AgRP neurons, based on their postsynaptic target, may mediate separable features of hunger [3,70]. Ongoing efforts to identify discrete subpopulations of AgRP neurons through neurocircuit mapping and functional interrogation of projection sites are crucial to an improved understanding of the broad behavioral and metabolic effects mediated by these neurons (Fig. 2).

Fig. 2.

Mapped AgRP neuron projections. (A) AgRP neurons project to a broad array of brain areas implicated in a wide range of behaviors. (B) Circuits related to feeding include AgRP projections to the LHA, PVH, BNST, PVT, MeA, and MPOA, which stimulate food intake, while projections to the PBN suppress discomfort and malaise. (C) Downstream projection sites in the PVT have also been found to be involved in smell, the amygdala in fear and aggression, the LHA in taste, Kisspeptin-expressing neurons of the ARC in reproduction, MPOA in maternal behaviors, and the PBN in pain and satiety. AgRP, agouti-related peptide; ARC, arcuate nucleus; BNST, bed nucleus of the stria terminalis; CeA, central amygdala; DMH, dorsomedial hypothalamus; KISS, Kisspeptin neurons; LHA, lateral hypothalamic area; LS, lateral septum; MPOA, medial preoptic area; MeA, medial amygdala; PAG, periaqueductal gray; PBN, parabrachial nucleus; PVH, paraventricular hypothalamus; PVT, paraventricular thalamic nucleus; SCN, suprachiasmatic nucleus; VMH, ventromedial hypothalamus.

AgRP circuits in feeding

Paraventricular nucleus (PVH)

Of the identified AgRP neuron projection sites, activation of AgRP projections within the PVH, aBNST, LHA, PVT, medial POA (mPOA), and MeA is sufficient to stimulate food intake [3,42,74,76]. Among these, the PVH is a brain region of particular interest as it receives the densest AgRP neuron innervation [74,75]. While neurons in this hypothalamic area participate in a wide range of neuroendocrine and autonomic responses, early PVH lesion studies led to hyperphagia and obesity [77], suggesting a key role in energy homeostasis. Consistent with this conclusion, administration of MC4R agonists directly into the PVH reduces feeding [78], and PVH neurons that express MC4R (PVH^MC4R) are implicated in the control of food intake [79], but not energy expenditure [80], by AgRP neurons. The robust feeding response evoked by stimulation of AgRP→PVH terminals, nearly comparable to activation at AgRP neuron cell bodies, underscores the role of the melanocortin pathway in the control of food intake [42,74]. However, central NPY administration stimulates food intake in MC4R-deficient mice [81], and optogenetic activation of AgRP neurons promotes food intake even in sated agouti (A^y) mice [41], suggesting AgRP neuron activation can drive feeding independently of MC4R. Consistent with this interpretation, AgRP neurons also synapse onto many neurons implicated in feeding that have incomplete overlap with PVH^MC4R neurons, such as oxytocin, prodynorphin, and glucagon-like-peptide1 receptor-expressing subpopulations [42,82,83] that may explain this finding.

Lateral Hypothalamic Area (LHA) and Paraventricular Thalamus (PVT)

The LHA plays a role in arousal, feeding, motivation, and reward [84]. LHA lesions reduce food intake and cause weight loss [85], while electrical stimulation of the LHA increases feeding [86]. Both glutamatergic and GABAergic LHA (LHA^GABA) subpopulations exist, with activation of the latter tending to increase food intake and reward while activation of LHA glutamatergic neurons (LHA^Glut) suppresses feeding and is aversive [87,88]. Conversely, selective inhibition of these neurons elicits the opposite set of responses, implying a physiological role for both subsets of neurons in food intake control [84]. While the LHA is richly supplied with AgRP neuronal projections, the specific contribution made by AgRP neurons in the control of discrete LHA neuronal subsets is unresolved. However, recent work implicates projections onto glutamatergic neurons in this brain area in taste modification. Thus, the effect of starvation and hunger to enhance sweet taste responsiveness and decrease aversive taste sensitivity [89] can be recapitulated by the activation of AgRP→LHA^Glut neurons [90].

In addition to taste, hunger also promotes attraction to food odors over other olfactory cues [91]. AgRP neurons projecting to the PVT mediate this effect as optogenetic stimulation of AgRP→PVT projections promotes food odor attraction, while inhibition of these neurons decreases food odor attraction in fasted mice, an effect via the NPY Y5 receptor [92]. Together, these findings suggest that the effect of hunger to modulate enhanced olfactory and orosensory perceptions is mediated in part by projections from AgRP neurons to the PVT and LHA, respectively.

Parabrachial nucleus (PBN)

The parabrachial nucleus (PBN) is comprised of a small group of nuclei located at the junction of the midbrain and pons and functions as a relay station for visceral and gustatory signals, as well as somatosensory information, including pain, temperature, and itch. AgRP neurons also regulate this region in ways that have profound physiological consequences [93]. The following observations highlight the importance of this region to AgRP-associated food intake. First, AgRP neuron ablation (inducing a loss of inhibitory input onto postsynaptic targets) induces marked neuronal hyperactivity and gliosis in the PBN, associated with profound anorexia and life-threatening weight loss [33,34]. Furthermore, inhibition of PBN neurons postsynaptic to AgRP rescued the anorexia phenotype induced by AgRP neuron ablation [94], implicating PBN neuron activation as the driver of this response and suggesting that this activation plays a physiological role to suppress feeding when inhibitory tone from AgRP neurons is reduced [95].

The PBN weighs internal state discomfort against metabolic need. Information related to malaise / discomfort is provided by excitatory input from the nucleus of the solitary tract (NTS) [96] and the spinal cord, while hunger state modulates the inhibitory input from AgRP neurons [94]. Within the external lateral subnucleus of the PBN, AgRP neurons directly synapse onto a population that expresses calcitonin gene-related peptide (PBN^CGRP neurons), activation of which potently suppresses food intake. When discomfort outweighs appetite, PBN^CGRP projections suppress appetite by exciting the central nucleus of the amygdala (CeA), a brain region associated with fear, anxiety, and related emotional responses [95]. Visceral malaise induced by lithium chloride, systemic inflammation induced by the endotoxin lipopolysaccharide (LPS), or cancer activate this PBN^CGRP→CeA circuit [95,97].

In contrast, while inhibition of these neurons has no effect on daily food intake in normal mice, it both increases meal size (implying a physiological role in satiation; see below) and blocks anorexia induced by interventions such as lithium chloride or lipopolysaccharide [95], implying a key role in the perception of sickness or malaise. Consistent with this notion, synaptic silencing of PBN^CGRP neurons, using cell-specific expression of a Tetanus toxin light chain [98], mitigates the anorexia, lethargy, anxiety, and malaise associated with cancer [97]. Thus, silencing these neurons may have therapeutic potential in conditions or diseases associated with wasting.

As noted above, the AgRP→PBNCGRP circuit also plays a physiological role in regulating meal size. Meal-related satiety signals such as amylin and cholecystokinin (CCK) activate PBN^CGRP neurons [99], and stimulation of AgRP→PBN projections is sufficient to increase food intake following administration of these gastrointestinal (GI)-related satiety peptides [99]. Synaptic silencing of PBN^CGRP neurons also blunts the satiety response to meal-related peptides, resulting in increased meal size without affecting total food intake due to a compensatory decrease in meal frequency [100]. These findings further our understanding of how hunger and satiety circuits interact to control food intake and energy homeostasis, whereby activation of AgRP neurons can inhibit the satiety effects of meal-related peptides to promote the consumption of larger meals.

AgRP circuits that regulate competing motivated drives

Physiological needs also essential for survival, that is, sleep, thirst, and sex, can compete with the motivational drive to eat. Eating and drinking are tightly coordinated. During meal consumption, prandial thirst ensures adequate water is consumed to aid ingestion and digestion [101], and while activation of AgRP neurons stimulates water intake in the presence of food, their activation is not sufficient to promote water intake independent of food consumption [72]. Conversely, dehydration prioritizes water over food intake. While the neurocircuits linking drinking and eating remain to be mapped, brain areas involved in the control of fluid balance and thirst, including the lamina terminalis, comprised of the subfornical organ (SFO), organum vasculosum of the lamina terminalis (OVLT), and the median preoptic nucleus (MnPO) [102], are implicated in thirst driven by dehydration.

Since sleeping, by definition, precludes engaging in motivated behavior, sleeping and feeding are not competing behaviors as such. But if you are both hungry and tired, do you eat, or do you sleep? Fasting increases wakefulness by reducing sleep depth and quality, and this effect is recapitulated by optogenetic or chemogenetic AgRP neuron activation [103], although Agrp neuron activation cannot prevent sleep altogether. On the other hand, sleep deprivation robustly tempers the effects of either AgRP activation or a prolonged fast to drive food intake [103], thus prioritizing sleep over feeding. During fasting, sleep depth is restored not only by AgRP neuron inhibition but also by Pomc neuron activation, suggesting a connection between appetite-regulating circuitry and sleep induction [103]; that is, sleep–wake circuitry and ARC neurons may be connected to one another via as yet unidentified projections. As the LHA is involved in both arousal and food intake and receives both AgRP and Pomc neuron innervation [84], connectivity between the ARC and the LHA seems likely.

Activated AgRP neurons are also implicated in territorial behaviors. During caloric restriction, where food is present but limited, mice will aggressively defend their resources from a conspecific intruder, but this territorial behavior is lost in cases of extreme caloric need if food resources remain absent [3]. This fasting-associated aggressive behavior is recapitulated by both chemogenetic activation of AgRP neurons and optogenetic activation of their terminals in the medial amygdala (MeA) [3]. These downstream MeA neurons express NPY1R and project to the pBNST, suggesting that an AgRP→MeA^Npy1R→pBNST circuit can drive territorial behavior [3], at least when food resources are limited. However, additional studies have found that competition between social interactions and food-seeking behaviors when no food source can be located does not always result in reduced territoriality. Driven either by fasting or artificial AgRP neuron activation, male mice balance food-seeking with the innate drive to dominate a submissive male if they cannot sate their hunger [7].

Multiple aspects of successful reproduction compete with hunger, consistent with the teleological concept that reproduction is maladaptive in the face of deficient fuel stores or food availability. For example, interest in receptive females is inhibited after both fasting and artificial activation of AgRP neurons if food is present [7,72]. Although the AgRP neuron projections that directly inhibit copulation in favor of food-seeking are unknown, one promising postsynaptic candidate is the ventromedial hypothalamus (VMH) [104,105], which is involved in social behaviors ranging from aggression and dominance to mating [105]. Given the steep metabolic demands of reproduction, the effect of fasting / chronic negative energy balance to deplete energy reserves and lower plasma leptin levels can cause infertility [106]. AgRP neuron activation in this setting is tied to inhibition of a hypothalamic reproductive circuit (involving kisspeptin-expressing neurons), reducing fertility [107]. Starvation also inhibits reproductive behaviors [7], and if pups are born, the motivation to perform parental behaviors necessary for their survival is blunted [76]. Mapped to AgRP neuron projections to the mPOA [76] this inattentive parental response during periods of fasting underscores the crucial, multidimensional role played by AgRP neuron activation during periods of energy deficiency or privation. By inhibiting energy-demanding physiological processes such as mating and reproduction, while also promoting food-seeking and consumption, AgRP neurons play an essential role in survival.

Hunger can also suppress the behavioral response to pain [108], allowing animals to prioritize finding food. This critical survival strategy involves an AgRP→PBN circuit, as optogenetic activation of AgRP neurons that project to the PBN selectively inhibits the behavioral response to long-term inflammatory pain [109], evidently mediated by NPY signaling on NPY Y1 receptors on PBN neurons [109]. These findings suggest that AgRP neurons, when activated during conditions of energy deficiency, suppress or override the perception of (or responsiveness to) pain or other somatosensory stimuli that compete with the animal’s need to focus their attention on searching, finding, and consuming food. These findings illustrate how physiological, safety, and social factors create motivational drives often in competition with hunger.

Regulation of AgRP neuron activity

Previous work provided a conceptual framework wherein hormonal and nutrient-related input regulate AgRP neuron activity (Fig. 1). This model is aligned with a straightforward hormonal feedback model of hunger and satiety, whereby ‘adiposity negative feedback signals’ leptin and insulin provide tonic inhibition of AgRP neurons, and the gastric hormone ghrelin has the opposite effect [110]. While these inputs are no doubt relevant to both control of AgRP neuron activity and energy homeostasis more broadly, our understanding of how AgRP neuron activity is regulated has evolved to incorporate both feedback and feedforward mechanisms. This sea change came about following three seminal and elegant studies, which monitored in vivo AgRP neuron activity. These studies found that the expected increase in AgRP neuron activity associated with caloric restriction or fasting is rapidly suppressed upon food presentation or learned feeding-related cues prior to consumption, even in the absence of food consumption [71,111,112]. The rapid onset and magnitude of this reduction in activity have forced a reconsideration of where AgRP neurons fit into the overall control of food intake.

Two of the aforementioned studies relied on the genetically encoded calcium sensor, GCaMP6, to monitor neuronal activity [113]. The more straightforward of the two techniques, fiber photometry [114,115], has enabled an explosion of studies documenting AgRP neuron calcium activity changes in real-time in response to sensory and hormonal input and how HFD feeding impacts this activity [116,117]. These studies suggest that, in addition to being regulated by feedback mechanisms (e.g., changing circulating levels of signals relevant to nutritional state of body fuel stores), AgRP neurons are also rapidly regulated by feedforward cues that promote homeostasis in an anticipatory manner. That is, by predicting future need, rather than responding in a compensatory manner to events that have already occurred (e.g., depletion of body fuel stores). Control by both feedback and feedforward mechanisms adds precision and efficiency to the defense of body fuel stores by predicting metabolic needs in response to changing external cues before the consequences of the changing environment become manifest. In comparison, homeostatic responses typically rely on the perception of a change in the defended parameter (e.g., body fuel stores) rather than in anticipation of a future change [118].

Feedforward regulation of AgRP activity

AgRP neuron activity varies dynamically throughout the sensory and metabolic processes that accompany the food presentation and meal consumption. This dynamic response can generally be segregated into discrete and continuous waves [119]. The first wave is rapid; food is presented, and sensory cues anticipating consumption inhibit AgRP neuron activity [71,111,112]. Sight, smell, and taste cues lead to robust inhibition in direct relation to the predicted value of the meal to be consumed [92]; by definition, these responses cannot be explained by changing circulating levels of leptin, insulin, or ghrelin. As the meal progresses, longer-lasting changes mediated by gut signals that predict the consequences of ingestion provide a second wave of feedforward signals [120,121]. Among these are GI tract-secreted hormones associated with satiety, for example, CCK and peptide YY (PYY), which increase across a meal and provide feedforward information related to nutrient content [120,121]. Vagal afferent signaling upon detection of macronutrients within the gut—specifically infusion of fat in the duodenum—also inhibits AgRP neurons once feeding commences [122]. Mechanosensation elicited by gastric distension also plays a feedforward role over food intake mediated in part by inhibition of AgRP neurons [123,124]. Together, these findings point to multiple sensory and gut–brain pathways that are engaged by preabsorptive signals to maintain energy homeostasis in part by inhibiting AgRP neurons, highlighting the complex nature of interactions between the periphery and brain that control food intake (Fig. 3).

Fig. 3.

Feedback and Feedforward Regulation of AgRP neurons. Feedback control of AgRP neuron activity is primarily regulated by hormonal (i.e., insulin, leptin, and ghrelin)- and nutrient-related input (i.e., glucose and free fatty acids). Visual and olfactory input related to food presentation and orosensory information related to taste inhibit AgRP neuron activity early in a meal, while mechanosensation and GIderived hormone input inhibit AgRP neuron activity as nutrients enter the gut but before nutrient absorption. Ambient temperature can modulate AgRP neuron activity in either direction, with cold temperatures increasing AgRP neuron activity and warmer temperatures reducing activity.

Afferent input to AgRP neurons

In contrast to our rapidly growing understanding of AgRP neuron efferents and their contributions to food intake, food-seeking, metabolism, motivation, and anxiety, information regarding afferent input to these neurons is comparatively diminutive. Retrograde tracing studies [75,125] and ex vivo electrophysiology [125,126] approaches show numerous inputs to AgRP neurons, with the largest innervations from the PVH and DMH. Among these are glutamatergic afferents from a subset of neurons in the PVH that coexpress thyrotropin-releasing hormone (TRH) and pituitary adenylate cyclase-activating polypeptide (PACAP). Chemogenetic activation of this population activates AgRP neurons and is sufficient to drive food intake in sated mice [125]. The functional role of this population in feeding, that is, what physiological state(s) activate the PVH^Pacap→AgRP circuit, is still unknown (Fig. 4).

Fig. 4.

Functionally mapped AgRP neuron afferents. AgRP neurons receive afferent input from several different brain areas. Of these, the PVH sends strong, excitatory, glutamatergic input to AgRP neurons in the ARC that stimulate feeding. In contrast, AgRP neurons receive GABAergic input from leptin-receptor-expressing neurons in the DMH that, when activated, inhibit feeding. AgRP, agouti-related peptide; ARC, arcuate nucleus; BNST, bed nucleus of the stria terminalis; DMH, dorsomedial hypothalamus; GABA, gamma aminobutyric acid; LepR, leptin receptor; LS, lateral septum; LHA, lateral hypothalamic area; MPOA, medial preoptic area; NI, nucleus incertus; PVH, paraventricular hypothalamus; PACAP, pituitary adenylate cyclase-activating peptide; PAG, periaqueductal gray; PBN, parabrachial nucleus; SON, supraoptic nucleus; VMH, ventromedial hypothalamus; VTG, ventral tegmental area.

AgRP neurons also receive input from a population of GABAergic neurons in the DMH that express the leptin receptor (DMH^GABA/LepR) and dynorphin [126]. Inhibition of AgRP neuron activity upon food presentation in the fasted mouse is provided, at least in part, by activation of this population [125,126]. Like AgRP neuron inhibition upon food presentation, the relative activity of this population scales according to both food availability and food palatability. While optogenetic activation of DMH^GABA/LepR→ARC afferents rapidly decreases food intake in fasted mice (or at the dark cycle onset), chemogenetic silencing of these neurons does not increase light cycle food intake, suggesting that this population of neurons is not required for maintaining physiological satiety [126]. Thus, in addition to hormonal cues, AgRP neurons are also controlled by neuronal afferents, which provide feedforward control rather than feedback information to prevent homeostatic changes. Indeed, leptin inhibition of AgRP neurons is mediated at least in part by activating upstream leptin-responsive DMH neurons, in addition to the direct action of leptin on AgRP neurons themselves.

Regulation of AgRP neurons by thermal sensory input

The use of fiber photometry has profoundly impacted our understanding of how sensory, nutrient, and hormonal input regulate AgRP neuron activity over varying time frames to control feeding in conscious, freely behaving mice. An important consideration when interpreting the growing body of fiber photometry recordings is that most such studies measure the change of activity from the value maintained in the setting of either fasting or caloric restriction when the majority of AgRP neurons are activated. This approach, however, may obscure other important, more subtle input that regulates AgRP neuron activity on a meal-to-meal basis that does not rely on, or may even be obscured by, states of energy restriction or deprivation. While fiber photometry is an extremely valuable tool, it has limitations. For one, it lacks resolution, measuring only the total AgRP neuron population activity and thus cannot detect differences in how distinct subsets of AgRP neurons respond to different types of feedback or participate in feedforward mechanisms via projection to distinct downstream targets. For example, besides the sight and smell of food, recent evidence suggests that AgRP neurons can respond to other sensory inputs unrelated to the status of body fuel stores, including environmental temperature [127] and circadian rhythms [111].

A recent example includes the response to cold exposure. In this setting, the increased energy expenditure required to maintain core temperature and avert hypothermia must be offset by an associated increase of food intake if energy homeostasis and body fat mass are to be preserved [128,129]. While the adaptive feeding response to cold exposure has been attributed to signals mounted in response to negative energy balance [130–132], this now seems unlikely since a subset of AgRP neurons is rapidly activated in response to cold exposure, and this activation is required for the associated increase in food intake (but not for the increase in energy expenditure) [127]. Moreover, this activation occurs so rapidly as to be incompatible with a compensatory response to negative energy balance; instead, cold-induced AgRP neuron activation appears to be part of a predictive, feedforward control system that recognizes the future need for increased calories to meet changing environmental circumstances.

Because these experiments were conducted under ad libitum feeding conditions, these observations point to the existence of discrete subsets of AgRP neurons, projecting to divergent downstream regions, with the potential to drive distinct physiological and behavioral effects to meet an environmental challenge. This model offers a viable explanation for how in one condition (e.g., during fasting or starvation), the effect of AgRP neuron activation to increase food intake is accompanied by reductions in energy expenditure and fertility, while in another (e.g., in the setting of cold exposure), AgRP neuron activation elicits increased energy intake that is accompanied by a proportionate increase in energy expenditure, with no loss of fertility [133]. Identifying upstream neurocircuits linking thermoregulation to control the activity of distinct subsets of AgRP neurons and associated feeding responses will improve our understanding of both thermoregulation and energy homeostasis. How AgRP neuron activity changes in other physiological conditions associated with hyperphagia besides caloric restriction, such as pregnancy, lactation [134], or growth, remain to be determined.

Circadian regulation of AgRP neuron activity

The regulation of circadian rhythm and energy homeostasis are inherently coupled. Circulating levels of nutritionally relevant hormones that regulate AgRP neuron activity (e.g., leptin, insulin, ghrelin, and corticosterone) are not static; instead, they follow cyclical patterns associated either with the light–dark cycle or with meals [135]. It is, therefore, not surprising that AgRP neuron activity exhibits a rhythm that increases across the light cycle in association with differences in food intake in ad-lib fed mice [71,111]. Although the neonatal loss of AgRP neurons has little to no effect on body weight or overall food intake, the loss of these associated rhythms of AgRP neuron activity renders mice unable to adapt to a restricted feeding schedule [136]. Moreover, recent work suggests that in AgRP neurons, cell autonomous ‘clock’ activity is required for feeding and transcriptional rhythms [137]. However, a better understanding of the neurocircuitry linking circadian rhythms and AgRP neurons is necessary to know how feeding is coordinated across the light:dark cycle. Recent work from our group demonstrates that silencing of DMH^LepR neurons, which project to and inhibit AgRP neurons, not only increases body weight and adiposity, but also phaseadvances diurnal rhythms of feeding and metabolism into the light cycle and abolishes the expected increase in dark cycle locomotor activity characteristic of nocturnal rodents [138]. The role of AgRP neurons in the circadian control of feeding is thus an active area of interest. These observations place AgRP neurons within a larger picture of feeding where, in addition to hormonal- and nutrient-related input, AgRP neurons receive and integrate sensory input that influences the decision of when and how much to eat.

AgRP neurons in health and disease

An interoceptive neurocircuit governs a function central to survival: thirst, salt appetite, sleep, by creating an internal representation of where the basic need they regulate falls on a spectrum of surfeit vs. insufficiency. Not only do such circuits create this representation, but they also put into action the behavioral and physiological responses necessary to either prevent or reverse a deviation from the established norm. Should the established norm be incorrect, however, or if the ability of the circuit to sense or respond to such a deviation is faulty, metabolic disease can result.

Obesity

AgRP neurons undeniably play a central role in energy homeostasis. To perform this function, they must be part of a circuit that generates a valid representation of current and impending energy needs related to the internal state. Given that diet and obesity cause dysfunction in the CNS mechanisms that regulate energy balance [59], impairments in the intrinsic sensitivity of AgRP neurons to hormonal- or nutrient-related input or the connection between the internal state and the CNS/AgRP neurons may mechanistically explain this dysfunction. Such dysfunction may weaken the ability of AgRP neurons to enable a correct representation of the energy state, and their ability to drive hunger, or any of the features of hunger, will miss the mark. Early studies examining the effect of an HFD on Agrp mRNA levels were inconsistent and found that the response varied on the composition and duration of the diet [139–141]. Subsequent electrophysiology studies showed that AgRP neurons are hyperactive in long-term, HFD-fed mice, and their responsiveness to leptin is also blunted [142–144]. However, leptin sensitivity could be restored if diet exposure was kept short and the mice were switched back to a standard chow diet [142] or by reducing the fat content of their diet [144]. Although exclusively ex vivo, these results suggest both that obesogenic diets disrupt basal activity patterns of AgRP neurons provided by long-term hormonal signals and that diet exposure on its own can be associated with AgRP neuron dysfunction.

With the introduction of fiber photometry, recent work has examined the activity of AgRP neurons in response to sensory, hormonal, and nutrient-related stimuli at multiple time points across HFD exposure [116,117]. Contrary to the ex vivo work detailed above, the application of fiber photometry found that, although the neurons continue to increase their activity in response to a fast, HFD exposure reduced AgRP neuron basal activity [117]. While the effect of diet exposure on basal AgRP neuron activity is unsettled, HFD feeding blunts in vivo AgRP neuron responsiveness to both sensory and hormonal cues that typically cause distinct and robust inhibitory effects [116,117]. These effects are observed before weight gain and may, or may not, be reversible. Thus, following the return to a chow diet, the AgRP neuron response to food cues is restored, while the response to GI hormones and nutrients is not [116]. Of note, six weeks of HFD exposure causes a specific, irreversible defect in the response of these neurons to dietary fat and the associated release of the satiety hormone CCK [116]. However, the time course when over which resistance occurs is unclear. HFD exposure studies noted decreased satiation, based on increased meal size and length within the first seven days of diet exposure [145], an effect that might be due to impaired CCKmediated satiety, the ability of CCK to inhibit AgRP neuron activity remained intact following four weeks of HFD exposure. These findings suggest that HFD feeding can impair AgRP neuronal responses to nutrients and hormones in ways that contribute to the pathogenesis of obesity and the difficulty inherent in maintaining weight loss [116,117].

Diabetic hyperglycemia

AgRP neurons may also contribute to diabetic hyperglycemia, as the activity of AgRP neurons is increased activity across several rodent models of diabetes. These include leptin-deficient ob/ob mice, mice with streptozotocin-induced diabetes mellitus (STZ-DM), and Zucker diabetic fatty rats [146–149]. In the STZ-DM model, inhibition of AgRP neurons reduces diabetic hyperglycemia [148], and the effects of either central leptin or FGF1 administration to ameliorate diabetic hyperglycemia are associated with inhibition of hypothalamic Agrp mRNA expression [150,151]. Combined with evidence that CRISPR-mediated deletion of the leptin receptor (LepR) from AgRP neurons is sufficient to induce severe obesity and diabetes and is required for leptin’s antidiabetic effects in STZ-DM [148], these findings collectively suggest that inappropriate activation of AgRP neurons contributes to diabetic hyperglycemia.

Challenges for the field

In the natural environment, animals may risk foraging for food in one set of circumstances, but not another, as they weigh intersecting and potentially competing physiological needs (hunger, thirst, sleep, illness, exhaustion, ambient temperature) and safety (threat, pain, aggression). However, activation of AgRP neurons, by either opto- or chemogenetic means, recapitulates the fasted state and, although tremendously informative, limits our ability to determine the hierarchy of only the most dominant drives. The role of AgRP neurons under other physiological circumstances is less understood. Conditions such as pregnancy and cold exposure increase food intake and AgRP transcript or activity, yet both fertility and heat production are inhibited in starvation and recapitulated in chemo and optogenetic activation models. The interpretation of this dichotomy is that AgRP activation models signal an insufficiency of available food, whether food is present or not, and when energy demands cannot be met, energy conservation mechanisms aimed at survival dominate. Of note, many of the stereotypical behaviors following stimulation of AgRP neurons are context-dependent and significantly influenced by the availability of food and energy state [7,72], where, depending on food availability, the activation of AgRP neurons is either aversive [71] or rewarding [112]. As such, a possible workaround for this concern is to measure the physiological or behavioral readout either in the presence or in the absence of food [3,7,72]. These same concerns apply to the growing body of fiber photometry recordings as the majority measure a change in activity from levels produced by fasting or food restriction. While the rationale for this approach is understandable, AgRP neurons are far less active during the light cycle [111] or in the fed state [71,112], and discerning subtle inhibition from an already suppressed signal is not easy. However, an approach that relies on fasted animals may obscure the role played by these neurons under more ordinary physiological conditions.

A plausible conceptual model to explain the day-today regulation of AgRP neurons posits that they do not simply cycle between binary ‘off’ and ‘on’ states through periods of fasting and feeding. Rather, we suspect that under usual physiological conditions, the overall balance of multiple inputs onto these neurons is inhibitory and that activation requires that this collective inhibitory input decrease or be overcome sufficiently to cross a threshold value. Stated differently, because of the highly consequential nature of behavioral and physiological responses triggered by AgRP neuron activation, multiple inhibitory inputs arising from negative humoral feedback, sensory, and interoceptive cues hold AgRP neuron activity in check, and only with withdrawal, perhaps in combination with stimulatory input from ghrelin or lower ambient temperature, is this ‘need threshold’ surpassed. This system of control over feeding behavior by these neurons makes sense from a teleological perspective; natural selection has favored a response system that can very powerfully promote a positive energy balance state (e.g., AgRP neuron activation). It follows then that once the threshold is exceeded, this inhibition can be lifted rapidly, motivating the animal to prioritize feeding.

A next challenge for our field is to determine the inputs capable of moving AgRP neuron activity across this threshold and relate their influence to scaled food intake changes. This goal is ambitious and not without technical challenges. Its lack of a biophysical mean limits our ability to view subtle activity changes by photometry, but more advanced technologies, for example, neuropixels [152,153] and in vivo electrophysiology [154], may work around these limitations.

Conclusions

Our understanding of the role played by AgRP neurons in feeding, metabolism, and behavior is growing, particularly with the implementation of increasingly sophisticated neuroscience tools. The functional dissection of AgRP neurons and their projections has broadened how we view hunger and its associated metabolic and behavior response. The field will continue to progress as our understanding of which neuromodulators are involved and mediate their distinct effects becomes apparent. Whether anatomically and functionally discrete subsets of AgRP neurons exist, whether they are regulated differentially by distinct afferent input, and whether they mediate distinct physiological responses are important unanswered questions. In pursuing this question, however, it is essential to recognize that the behavioral effects following activation of AgRP neurons are context-dependent, influenced by the availability of food and energy state (i.e., fasted vs. fed). As the AgRP neuron and food intake fields continue to grow, there will be a better understanding of how these neurons are regulated, the specific cell types and neurocircuits they engage under different physiological conditions, and whether dysfunction of these neurons contributes to the pathogenesis of either obesity, disordered eating or pathological anorexia.

BOX 1. A Brief History of AgRP.

In 1997, the cloning of the Agouti-related peptide (Agrp) gene identified a 132 amino acid peptide with homology to agouti [8,9]. Normally, agouti serves as a paracrine signaling molecule that regulates skin pigmentation by antagonizing the melanocortin receptor-1 (MC1R) in the hair follicle, switching the synthesis of the black pigment eumelanin to the yellow/red pigment, pheomelanin [10,11]. While the expression and action of agouti are normally limited to the skin, a mutation in the agouti gene, A^y, leading to the ectopic expression [12,13], results in mice with characteristic yellow coat color and a pronounced metabolic phenotype including obesity, hyperinsulinemia, and increased body length [14]. This agouti obesity syndrome is attributed to the antagonism of melanocortin receptors 3 and 4 (MC3R/MC4R) expressed in the central nervous system [12,13]. The melanocortin system consists of a family of peptide agonists, including alpha-melanocyte-stimulating hormone (α-MSH) and corticotropin hormone (ACTH) derived from the polypeptide precursor pro-opiomelanocortin (POMC), that mediate their biological effects by binding to one of five melanocortin receptors [15]. As a biased agonist of the MC3R/MC4R [9], AgRP is a critical regulator of melanocortin signaling. A role for the central melanocortin system in regulating energy homeostasis was first inferred following evidence that pharmacological agents that agonize or antagonize MC4R reduce or increase food intake [16]. Moreover, similar to both A^y mice [14] and mice overexpressing AgRP [9,17], mice with a targeted disruption of MC4R present with severe obesity and increased linear growth [18], a finding that has been extended to humans [19,20].

Abbreviations

ACTH

corticotropin hormone

AgRP

agouti-related peptide

Alpha-MSH

alpha-melanocortin-stimulating hormone

ARC

arcuate nucleus

A^y

Agouti, ectopic expression mutant

BAT

brown adipose tissue

BNST

bed nucleus of the stria terminalis

CCK

cholecystokinin

CeA

central nuclei of the amygdala

CGRP

calcitonin gene-related peptide

CNS

central nervous system

DIO

diet-induced obesity

DMH

dorsomedial hypothalamus

DREADD

Designer Receptors Exclusively Activated by Designer Drugs

GABA

gamma aminobutyric acid

GI

gastrointestinal

HFD

high-fat diet

HGP

hepatic glucose production

ICV

intracerebroventricular

LepR

leptin receptor

LHA

lateral hypothalamus

LPS

lipopolysaccharide

LS

lateral septum

MC1–4R

Melanocortin receptors 1–4

MeA

medial nuclei of the amygdala

MnPO

median preoptic nucleus

MPOA

medial preoptic area

NPY

Neuropeptide Y

NPY1–5R

Neuropeptide Y receptors 1–5

NTS

nucleus of the solitary tract

OVLT

organum vasculosum of the lamina terminalis

PACAP

pituitary adenylate cyclase-activating polypeptide

PAG

periaqueductal gray

PBN

parabrachial nucleus

POA

preoptic area

POMC

pro-opiomelanocortin

PVH

paraventricular nucleus

PVT

paraventricular thalamic nucleus

PYY

peptide YY

RER

respiratory exchange ratio

SCN

suprachiasmatic nucleus

SFO

subfornical organ

STZ-DM

streptozotocin-induced diabetes mellitus

TRH

thyrotropin-releasing hormone

VGAT

vesicular GABA transporter

VMH

ventromedial nucleus