Making Sense of the Sensory Regulation of Hunger Neurons

Insights & Perspectives

Making sense of the sensory regulation

Think again of hunger neurons

Yiming Chen and Zachary A. Knight*

AgRP and POMC neurons are two key cell types that regulate feeding in such as leptin, and increases the level of response to hormones and nutrients. Recently, it was discovered that these hormones that promote feeding, such as neurons are also rapidly modulated by the mere sight and smell of food. This ghrelin. This hormonal switch activates rapid sensory regulation ‘‘resets’’ the activity of AgRP and POMC neurons AgRP neurons and inhibits POMC neu- before a single bite of food has been consumed. This surprising and rons, creating a “hunger drive” that counterintuitive discovery challenges longstanding assumptions about the motivates animals to find and consume food, and that persists until food intake function and regulation of these cells. Here we review these recent findings and replenishes the body of nutrients. In this discuss their implications for our understanding of feeding behavior. We way, a simple negative feedback loop is propose several alternative hypotheses for how these new observations might thought to explain the remarkable be integrated into a revised model of the feeding circuit, and also highlight some coordination of feeding behavior with of the key questions that remain to be answered. physiologic need. While this homeostatic model is Keywords: widely accepted, one key piece of data has been missing: information about anticipatory; arcuate nucleus; feeding; homeostasis; hunger; hypothalamus; . the activity dynamics of these neurons neural circuit in vivo. In the past year, this gap has Introduction consumption [3–6]. POMC neurons, in been closed, as three groups have contrast, are activated by energy surfeit, reported measurements of AgRP and The body weight of most animals is and their activity promotes fasting and POMC neuron dynamics in awake, remarkably stable over time, suggesting weight loss [3, 7]. These two cell types are behaving mice [12–14]. Contrary to that a homeostatic system balances food often described as the “gas pedal” and expectations, these experiments intake and energy expenditure over the the “brake” for the neural control of revealed that AgRP and POMC neurons long-term [1]. Two neural cell types in the feeding. Together, they have been stud- are rapidly modulated by the mere sight hypothalamus, termed Agouti-related ied in far greater depth than any other and smell of food, in a way that “resets” protein (AgRP) and Proopiomelanocortin neurons in the brain that have a special- their activity before any food is con- (POMC) neurons, are thought to be ized role in energy balance. sumed. This paradoxical discovery has critical components of this homeostatic AgRP and POMC neurons are regu- prompted reassessment of the regula- system (Fig. 1A). AgRP neurons are lated by circulating signals of nutrition and function of these long-studied activated by energy deficit [2] and their tional state, which modulate these cells cells [12–15]. In this essay, we summa- activity promotes food seeking and in opposite directions consistent with rize these recent findings and discuss their function. For example, the hor- their implications for our understanding mone leptin, which signals energy of the neural regulation of feeding. DOI 10.1002/bies.201500167 availability, activates POMC neurons and inhibits AgRP neurons [8, 9]. The Department of Physiology, University of California, hormone ghrelin, which signals energy The sensory detection of San Francisco, CA, USA scarcity, has the opposite effect [10, 11] (Fig. 1B). Based on these observations, food rapidly resets AgRP *Corresponding author: Zachary A. Knight an intuitive and widely accepted model and POMC neurons E-mail: [email protected] has emerged for how these two cell types control feeding. According to this AgRP and POMC neurons were first Abbreviations: AgRP, agouti-related protein; GABA, gamma- model, food deprivation decreases the implicated in the control of feeding aminobutyric acid; POMC, proopiomelanocortin. level of hormones that inhibit feeding, almost 20 years ago [2, 16–20], yet their

316 www.bioessays-journal.com Bioessays 38: 316–324, ß 2016 WILEY Periodicals, Inc...... Insights & Perspectives Y. Chen and Z. A. Knight

types. To overcome these obstacles, it was necessary to develop new methods that enable optical [21, 22] or hn again Think electrophysiological [23] recordings from genetically defined cell types at deep brain sites and in freely behaving animals. Three such methods have now been developed and applied to investigate the dynamics of AgRP and POMC neurons (Fig. 2). The first study used an approach called fiber photometry, which utilizes an optical fiber to record fluorescence signals from a genetically encoded calcium reporter (e.g. GCaMP6s) targeted to a specific cell type [22]. This approach measures population-level calcium dynamics but does not resolve single cells (Fig. 2). In the key experiment in this study, mice were fasted overnight and then presented with a Figure 1. AgRP and POMC neurons regulate feeding in response to nutritional signals. piece of food, and during this time the A: AgRP and POMC neurons are intermingled in the arcuate nucleus of the hypothalamus. B: The anorexigenic hormone leptin and the orexigenic hormone ghrelin regulate AgRP/ calcium dynamics of AgRP and POMC POMC neurons in opposite directions. neurons were optically recorded [12]. Contrary to the predictions of homeostatic models, it was found that food in vivo dynamics were only described in the arcuate nucleus (ARC) of the presentation alone rapidly inhibited the past year [12–14]. This long delay hypothalamus, which is located at the AgRP neurons and activated POMC was due to technical challenges associ- base of the forebrain and contains a neurons. This response began the ated with recording neural activity from diversity of intermingled neural cell moment that food was presented and was complete within seconds, often before a single bite of food could be consumed. Thus sensory cues associated with food, rather than the post-ingestive effects of feeding, are primarily responsible for resetting the activity of these cells [12]. This finding was counterintuitive in part because it revealed that AgRP neurons, which promote food intake, are actually less active when a mouse is eating, and conversely that POMC neurons, which inhibit food intake, are more active during feeding. This paradoxical activity pattern creates a puzzle for explain- ing how these neurons are able to perform their presumed functions. Further characterization of the rapid sensory modulation of these neurons revealed five important properties [12]. (i) The response depends on Figure 2. Techniques for cell-type-specific recording of deep brain neural activity. (Left) nutritional state, since neurons from Microendoscopic calcium imaging utilizes a head-mounted, miniaturized microscope to record fasted mice respond more strongly to fluorescence signals from a genetically encoded calcium indicator targeted to a specific cell food cues than neurons from fed mice. type. This method can be used to probe deep brain structures by coupling the microscope to a (ii) It depends on food palatability, gradient refractive index (GRIN) lens of appropriate length. (Middle) Optrode recordings utilize an since energy dense foods such as electrode array paired with an optical fiber to perform extracellular recordings. The optical fiber peanut butter induce a stronger enables cells expressing channelrhodopsin to be identified by their short latency responses to light stimulation. (Right) Fiber photometry uses a single optical fiber to both excite and record response than energy poor foods such fluorescence from a population of neurons expressing a calcium reporter. The publications that as chow. (iii) It depends on food used each of these approaches to investigate AgRP/POMC neurons are listed below. accessibility, since food that is hidden

Bioessays 38: 316–324, ß 2016 WILEY Periodicals, Inc. 317 Y. Chen and Z. A. Knight Insights & Perspectives.....

or inaccessible induces a weaker and electrode recordings based on their light that predicts food availability [28]. The less durable response. (iv) It is contin- sensitivity [23] and, unlike calcium anticipatory response in this case is gent on eventual food consumption, imaging, provide reliable measurement the secretion of saliva containing diges- since removal of food before it can be of single action potentials (Fig. 2). In tive enzymes, which functions to pre- consumed causes reversion of neural this study ChR2 was targeted to AgRP pare the oral cavity for food ingestion activity back to its prior state. (vi) It is neurons, so that AgRP neurons could be just moments before the food arrives. By integrative, because these factors identified as light-activated units and contrast, the function of the anticipa- interact with each other to determine putative POMC neurons could be iden- tory regulation of AgRP and POMC the magnitude of the response. As an tified as light-inhibited units. Using this neurons is less clear. We consider below example of this last property, presen- approach AgRP neuron firing rates were five alternatives.

Think again tation of a palatable food such as found to gradually increase throughout peanut butter was shown to modulate the course of the light phase, consistent these neurons even in fed mice, which with the slow development of energy Hypothesis #1: Sensory were otherwise insensitive to presenta- deficit that occurs during the day [14]. feedback gates cephalic tion of chow [12]. Taken together, these Subsequent food presentation rapidly phase responses findings reveal that AgRP and POMC inhibited AgRP neurons and activated neurons are regulated in a way that POMC neurons, similar to the findings Food consumption is necessary for anticipates the nutritional or incentive from calcium imaging. One important survival but also represents an acute value of a forthcoming meal [12, 15]. difference between these studies; how- threat, since it floods the body with This anticipatory regulation resembles ever, was that AgRP neuron activity was nutrients that can disrupt physiologic an expected value calculation, in which shown to remain somewhat elevated balance [29]. To deal with this chal- the brain weighs factors such as the following food presentation in the lenge, animals have developed a large need for food, the energy density or optrode experiment [14]. This persistent class of peripheral adaptations that are palatability of the food, and the likeli- elevated activity was interpreted to triggered by the sight, smell, and taste hood that it is obtainable [12]. represent a residual “hunger drive” that of food and function to prepare the body These findings were extended by a is responsible for promoting subsequent to metabolize and absorb nutrients second study that used fluorescence food consumption. We discuss this [30, 31]. These anticipatory responses, microendoscopy to analyze calcium observation and its possible implica- which Pavlov called “psychic secre- dynamics of AgRP neurons in freely tions later in the text. tions,” are now known as cephalic behaving mice [13]. An important phase responses because they are advantage of this approach is that it controlled by the brain. In addition to provides single cell resolution and Feeding is regulated by salivation, cephalic phase responses therefore can reveal heterogeneity in include the secretion of gastric acid, responses within a genetically defined both homeostatic and bile, and digestive enzymes into the population of neurons (Fig. 2). This anticipatory mechanisms stomach and intestines; the release of study added two additional important hormones such as insulin, cholecysto- pieces of information. First, it showed The rapid sensory modulation of AgRP kinin, and pancreatic polypeptide into that essentially every AgRP neuron is and POMC neurons was unexpected in the bloodstream; and an increase in rapidly inhibited by food presentation, part because thinking about these cells body temperature [30, 31]. Most of these at least at the level of calcium dynamics has been dominated by the concept of responses can be triggered by both food (106/110 cells) [13]. Thus AgRP neurons, homeostasis. A homeostatic mechanism cues and food absorption. For example, despite having diverse projection pat- is one in which a deviation from a the cephalic phase of insulin release is terns, appear to be subject to remark- physiologic set-point triggers a counter- triggered by sensory cues that occur at ably homogeneous regulation by regulatory response [24, 25]. For exam- the onset of feeding and precedes sensory cues. Second, it showed that ple, a decline in plasma leptin activates changes in blood glucose [32, 33]. Later, AgRP neurons can be modulated by an AgRP neurons to induce feeding. The a second phase of insulin release occurs arbitrary cue, such as a light or sound, sensory modulation of AgRP and POMC following food absorption in response to that animals have been trained to neurons by contrast is not homeostatic, hyperglycemia. associate with food [13]. This indicates because it occurs before any physiologic Similarities between the sensory that the response to food cues is, at least change has taken place. This response is regulation of AgRP and POMC neurons in part, learned. This finding is consis- instead “anticipatory,” because it pre- and the activation of cephalic phase tent with the fact that response of these dicts the nutritional changes that will responses suggest these neurons could neurons to novel foods accelerates upon occur in the future after the food has be part of the upstream pathway. For repeated exposure, which also indicates been consumed (Fig. 3). example, cephalic phases responses are learning [12]. Anticipatory mechanisms are wide- triggered more strongly by palatable A third study reported optrode spread in neurobiology [24, 26, 27], but foods [34–37], by multisensory com- recordings of AgRP and POMC neuron are often overlooked in discussions of pared to unisensory food cues [38], and activity in head fixed mice [14]. Optr- feeding behavior. The classic example of following food deprivation [37]. The odes enable the use of channelrhodop- anticipatory regulation is Pavlov’s dogs, sensory regulation of AgRP/POMC neu- sin (ChR2) to identify cells from which salivate follow ringing of a bell rons shares all of these properties [12].

318 Bioessays 38: 316–324, ß 2016 WILEY Periodicals, Inc...... Insights & Perspectives Y. Chen and Z. A. Knight

thereby calibrate their food intake more precisely. An implication of this finding is that hn again Think food delivered directly to the stomach, thereby bypassing sensory cues, should be experienced as less satiating than food consumed orally. This prediction has been confirmed by experiments showing that enteral feeding in humans fails to fully suppress appetite [48–51]. This failure does not appear to reflect decreased production of gastrointestinal satiation signals [49, 52], suggesting it involves the absence of a cognitive signal triggered by sensory cues. Con- Figure 3. Fast and slow regulation of AgRP and POMC neurons by anticipatory and sistent with this, some enteral fed homeostatic signals. (Right) Circulating hormonal signals such as leptin, ghrelin, and human subjects report that simply insulin have traditionally been thought to play a primary role in the regulation of AgRP chewing food decreases their residual and POMC neurons. The levels of these signals fluctuate slowly over minutes to hours in hunger, even though the food cannot be accordance with changes in nutritional state. (Left) In vivo recordings; however, revealed swallowed [50, 53]. a dominant role for anticipatory signals in the regulation of AgRP and POMC neurons. The rapid inhibition of AgRP neu- These anticipatory signals are triggered by sensory cues from the outside world and rons by food cues could contribute to communiated by neural input, and therefore develop much faster than hormonal changes. In addition, these anticipatory signals precede rather than respond to changes this phenomenon of anticipatory sati- in the nutritional state of the body. ety. The fact that these neurons are more strongly inhibited by energy dense foods [12] and respond to Pavlovian Cephalic phase responses can also be While this seems paradoxical, there is conditioning [13] are both consistent learned by Pavlovian conditioning evidence that learned associations with with this possibility. One way to test this [28, 34, 39, 40], similar to the sensory sensory cues contribute to the termina- model would be to determine whether modulation of AgRP/POMC neurons [13]. tion of feeding [43]. The most compel- AgRP neurons can be conditioned by While the forebrain circuitry that con- ling data come from sham feeding gastrointestinal negative feedback sig- trols cephalic phase responses is largely experiments, performed primarily in nals. For example, would an arbitrary unknown, manipulations of the para- rats, which used a gastric fistula to cue that predicts the infusion of ventricular, ventromedial, and lateral drain the ingested food from the stom- nutrients into the stomach attain the hypothalamus can trigger gastric acid ach [44, 45]. This preparation enables ability to inhibit AgRP neurons follow- release and other gastrointestinal disconnection of the effects of gastroin- ing training? Conversely, would the responses, indicating a role for the testinal signals from external sensory response of AgRP neurons to a food hypothalamus [34, 41, 42]. There is also cues on feeding behavior. These studies cue undergo extinction if the ingested some evidence that chemogenetic mod- showed, first, that rats consume much food was drained from the stomach each ulation of AgRP neurons can rapidly more food during sham feeding com- time the cue was presented? These types alter peripheral metabolism [4], pared to real feeding [44, 45], confirm- of experiments would clarify the nature although this has not been explored ing the importance of post-ingestive of the unconditioned stimulus that in detail. In future studies, it will be negative feedback signals such as trains AgRP neurons to respond to food important to measure the effects of cell- gastric distension in meal termination. cues, and in doing so provide insight type-specific manipulations of AgRP However, it was found that during into the function of this anticipatory and POMC neurons on cephalic phase repeated sham feeding trials the amount modulation. responses in peripheral tissues, in order of food these animals consumed in- to understand the role of these neurons creased even further [44, 45]. This in gating this response. progressive increase in sham food Hypothesis #3: Sensory consumption was shown to reflect the feedback suppresses appetitive extinction of a learned association behaviors Hypothesis #2: Sensory between the sensory properties of feedback induces anticipatory specific foods (the conditioned stimu- AgRP neurons regulate not only food satiety lus) and their post-ingestive consequen- intake but also appetitive behaviors that ces (the unconditioned stimulus). This promote food obtainment. For example, AgRP neurons are thought to promote learned association functions to reduce optogenetic or chemogenetic activation hunger, and thus their rapid inhibition the rate of food intake at the beginning of AgRP neurons motivates animals to by food cues would be predicted to of a meal [46, 47], perhaps so that engage in vigorous lever pressing or result in “anticipatory satiety” – satiety animals can anticipate at the outset of a nose poking for a food reward [4, 54]. that occurs before the food is consumed. meal some of its physiologic effects and AgRP neuron activation also stimulates

Bioessays 38: 316–324, ß 2016 WILEY Periodicals, Inc. 319 Y. Chen and Z. A. Knight Insights & Perspectives.....

locomotor activity speciﬁcally in the continuous optogenetic stimulation. function as a teaching signal that trains absence of food, which has been Consistent with this second possibility, animals to search for and consume interpreted as representing foraging it has been shown that AgRP activation food [13]. This mechanism for encour- [4, 55]. Pharmacologic experiments induced by high-dose ghrelin treatment aging a behavior by linking it to the have shown that delivery of AgRP or can be largely reversed by presentation removal of an aversive stimulus is NPY into the brain can preferentially of a single piece of chow [12], highlight- known as negative reinforcement [63]. promote appetitive rather than consum- ing the potency of this inhibitory A prediction of this negative rein- matory behaviors under some condi- sensory input. forcement model is that mice should tions [56–59]. Together, these An important unresolved question is learn to perform instrumental responses observations highlight the importance whether AgRP neurons are required for that lead to a reduction in AgRP neuron

Think again of AgRP neurons for promoting behav- appetitive behaviors in food deprived activity. However this is not the case [13], iors that lead to food discovery, and mice. One study found that neonatal as it has been shown that mice fail to raise the question of how the transition ablation of AgRP neurons reduced food execute simple operant tasks that have from appetitive to consummatory anticipatory activity (FAA), which is an been experimentally paired with AgRP behaviors is controlled. increase in locomotor activity that neuron inactivation. For example, fed The discovery that AgRP neurons occurs before food availability and mice will neither lever press nor nose are rapidly inhibited by food cues resembles foraging in mice subjected poke in order to pause optogenetic suggested that this inhibition may gate to scheduled feeding protocols [60]. By stimulation of AgRP neurons [13]. Simi- the transition from foraging to feed- contrast, adult ablation of AgRP neu- larly, fasted mice fail to nose poke in ing [12, 14]. Indeed, the response rons reduced food intake primarily due order to induce optogenetic silencing of properties of AgRP neurons to sensory to “visceral malaise,” a defect that is naturally elevated AgRP neuron activ- cues appear almost perfectly designed consummatory in nature since AgRP ity [13]. This indifference to AgRP neuron to serve this function. As described neuron ablated mice will reject food silencing is in stark contrast to the above, food cues modulate AgRP neu- delivered directly into their mouths [61]. dramatic lever pressing and nose poking rons in a way that resembles an However, this nausea may be specific to that AgRP neuron stimulated animals expected value calculation, in which the extreme case of acute AgRP neuron will perform for an actual food reward the animal weighs factors such as the ablation in adults, since otherwise the [4, 54]. Thus these data do not support accessibility of the food, the energy mere sight and smell of food would a primary role for negative reinforce- density of the food, and its own need for trigger sickness and eating would be ment as the motivational mechanism nutrition. This integration would enable impossible. that AgRP neurons utilize to achieve animals to make foraging decisions that Experiments that test the require- their remarkable behavioral effects. are adaptive in the face of changing ment for AgRP neurons in specific An alternative possibility is that internal and external circumstances. appetitive behaviors will be important AgRP neurons function by increasing For example, discovery of a suboptimal in order to clarify how the sensory the positively rewarding properties of source of nutrition by a starving animal modulation of these cells influences the food, such as its sight, smell, and would nonetheless inhibit AgRP neu- transition from foraging to feeding. taste [15]. It is well known that food rons, thereby ensuring that foraging is While AgRP neuron dynamics appear deprivation makes food more appeal- blocked when food is of greatest well-suited to serve this function, it is ing, a concept known as alliesthe- value [12–14]. By contrast AgRP neurons possible that these anticipatory dynam- sia [64], and that this enhanced from a well-fed animal would be ics are merely a consequence of this palatability and incentive salience can insensitive to food cues unless the food behavioral transition rather than its promote food seeking and consump- was particularly palatable [12]. cause, as has been proposed for other tion [15]. Consistent with this, palatable While this model is appealing, it is cell types [62]. foods have enhanced ability to modu- inconsistent in its simplest form with all late AgRP neurons [12, 15]. At present, of the available evidence. One problem however, there is little data that directly is that if AgRP neuron inhibition is Hypothesis #4: Sensory addresses this hypothesis. required for the transition from foraging feedback acts as a teaching to feeding, then continuous optogenetic signal stimulation of AgRP neurons should Hypothesis #5: Sensory result in animals that perpetually forage Hunger is an unpleasant state, and one feedback connects the present and as a result do not eat. However this reason animals eat is to eliminate this to the future is not the case [3]. One way to explain negative feeling. Recently, it was shown this discrepancy might be that AgRP that AgRP neuron activity has negative It is possible that the rapid sensory neuron inhibition promotes, but is not valence, meaning that mice find it modulation of AgRP and POMC neurons required, for the transition from forag- aversive and therefore learn to avoid may not trigger any immediate behaving to feeding. Alternatively, it is places and flavors that are associated ioral transition, motivational change, or possible that food presentation sends with elevated AgRP neuron activity [13]. physiologic response. Instead, this sen- a sufficiently strong inhibitory signal to It has been proposed that, by alleviating sory modulation could function primar- AgRP neurons that they become this negative state, the rapid sensory ily to synchronize rapid feeding silenced even in the presence of inhibition of AgRP neurons may behavior with slower nutritional

320 Bioessays 38: 316–324, ß 2016 WILEY Periodicals, Inc...... Insights & Perspectives Y. Chen and Z. A. Knight

changes. To illustrate why this may be that they actually regulate AgRP and Whatever the reason, it is clear that necessary, consider three types of POMC neurons, with the possible ex- AgRP and POMC neurons are rapidly signals that could communicate ongo- ception of peptide YY ([65, 66] but see modulated by food associated sensory hn again Think ing meal status to AgRP and POMC also [67]). cues. This raises another question: if neurons (Fig. 4). The first is post- In the absence of post-absorptive these neurons are “reset” by the sight absorptive signals, such as leptin, that and post-ingestive signals, external and smell of food, then how do they report on the nutritional state of the sensory cues are the primary remaining control feeding at all? body. These signals have been the focus type of information that AgRP and of most studies of the regulation of POMC neurons could use to learn about AgRP and POMC neurons, yet they act the status of an ongoing meal (Fig. 4C). too slowly to control behavior directly Because these sensory cues are How do AgRP and POMC (Fig. 4A). For example, changes in detected before food has been con- neurons control feeding? plasma leptin require metabolic and sumed, they can most easily be used to transcriptional responses in adipocytes make predictions about impending A vast body of evidence shows that AgRP which develop over hours, and there- food consumption, which is how they and POMC neurons control food intake. fore leptin could not directly terminate a are utilized in practice [12]. Thus the This includes the results of optogenetic, “hunger drive” following feeding. A anticipatory regulation of AgRP and chemogenetic, and cell ablation studies second candidate is post-ingestive sig- POMC neurons may have evolved in mice demonstrating that bidirectional nals that arise from the stomach and primarily as a facile mechanism for manipulation of these cells alters intestine following feeding, including the brain to coordinate rapid food feeding [3–7], as well as pharmacologic gastric distension and gut derived ingestion with much slower homeo- and genetic studies showing that the satiation peptides such as cholecystoki- static changes, rather than as a way to neurotransmitters produced by these nin. These signals have the right kinet- trigger a specific behavioral or meta- cells (gamma-aminobutyric acid (GABA), ics for controlling feeding behavior bolic transition upon the discovery of neuropeptide Y (NPY), AgRP, and POMC) (Fig. 4B), but there is little evidence food. modulate feeding inwaysconsistentwith their putative functions [16–18, 68–71]. These findings have been extended to humans by the discovery that loss-of-function mutations in the melanocortin 4-receptor (MC4R) are a com- mon cause of extreme obesity [72–75]. Thus a mechanism must exist by which the activity of these neurons is translated into changes in food consumption. Nonetheless, the natural activity patterns of these neurons are difficult to reconcile with their presumed functions. It is generally assumed that, if a neuron drives a behavior, then the neuron should be more active during or immediately preceding the behavior’s execution. Yet this is not thecaseforAgRPandPOMCneurons. AgRP neurons, which are thought to drive food intake, are much less active during feeding compared to minutes before. POMC neurons, which are thought to inhibit food intake, are more active during the act of feeding Figure 4. Three models for the inactivation of AgRP neurons by feeding. A: Post-absorptive itself. These counterintuitive trends feedback reports on the nutritional state of the body and includes circulating signals such as apply to even the small fluctuations in leptin. These signals evolve over hours and therefore are too slow to explain the rapid AgRP and POMC neuron activity that suppression of hunger by food intake. B: Post-ingestive signals arise from the stomach and surround individual bouts of eating, intestine immediately following food intake, and including gastric distension as well as gut- whichoftenhavetheoppositesign derived satiation peptides. These signals directly control meal termination, and therefore relative to what would be predicted would be well-suited to inactivate AgRP neurons after feeding. However, current data does based on the known functions of these not support a prominent role for these gastrointestinal signals in the actual regulation of AgRP neuron activity. C: Cephalic feedback involves sight, smell, taste, and possibly other cells [12, 14]. external sensory cues. In vivo recordings reveal that this is the major mechanism for the How can we explain these paradox- regulation of AgRP and POMC neurons during meals. ical findings? We consider below two

Bioessays 38: 316–324, ß 2016 WILEY Periodicals, Inc. 321 Y. Chen and Z. A. Knight Insights & Perspectives.....

hypotheses, which are not mutually activated? In response to other stimuli, presentation to “spill over” and inﬂu- exclusive, for how these cells may AgRP activation was robustly detected by ence the food intake that occurs later. regulate feeding. calcium imaging [13]. Second, photome- There have been hints that such a try recordings showed that presentation sustained mechanism might operate. of palatable food to fed mice can reduce One clue is that feeding following Mechanism #1: Residual AgRP calcium signals considerably below the optogenetic stimulation of AgRP neu- neuron activation persists after ad libitum fed baseline [12]. Therefore rons has an unusually long latency food presentation and drives calcium imaging is not inherently limited (6 minutes) [3]. By contrast, optoge- feeding by linear range or sensitivity in its ability netic stimulation of inputs to the to detect activity reductions below the lateral hypothalamus can induce feed-

Think again An important question is whether food baseline level of fed animals. ing within 10 seconds [76]. This presentation completely resets the activity An alternative possibility is that these suggests that AgRP neurons do not of AgRP and POMC neurons to baseline, or discrepancies reflect differences in control the food intake machinery whether a residual orexigenic activity experimental paradigm. One important directly, but rather produce some pattern persists in these cells. Micro- difference is that the optrode study used factor that must “build up” in the endoscopic calcium imaging found that head-fixed mice presented with a liquid downstream circuit before feeding is essentially all AgRP neurons were rapidly diet [14], whereas the calcium imaging triggered. inhibited by food presentation (106/110 studies used freely behaving mice pre- How could such a mechanism cellsinhibitedvs.1/110cellsactivated)[13]. sented with solid food [12, 13]. It has been operate? It has been shown that a By contrast, optrode recordings found that shown that the magnitude and durability single injection of the AgRP neuropep- only 64% of AgRP neurons were inhibited of the response of AgRP and POMC tide into the brain can increase feeding by food presentation (14/22 cells),whereas neurons to food cues is very sensitive for up to 1 week, indicating that the 23% of AgRP neurons were activated (5/22 to factors such as the accessibility and effects of AgRP peptide release can cells) [14]. Consistent with this mixed palatability of the food [12]. Thus it may extend far beyond the duration of AgRP response, optrode recordings showed that be that the use of a head-fixed prepara- neuron activation [77]. Similarly, acti- food presentation to food restricted mice tion or liquid diet reduces the anticipa- vation of POMC neurons inhibits feed- in the dark phase did not reduce the firing tory response of these neurons, either ing on a time-scale of hours to days rate of AgRP neurons all the way to the due to stress or some change in the [3, 7]. Thus modulation of the melano- level of fed mice at the start of the light animal’s expectation of food availability cortin system can have delayed and phase, a time that mice generally do not or value. This would explain why the chronic effects on feeding. However, eat. This residual activity of AgRP neurons calcium imaging and electrophysiologic this mechanism cannot explain feeding was proposed to represent a hunger drive measurements showed different percen- that results from release of NPY and that persists after food discovery [14]. tages of AgRP neurons that were inhib- GABA, which in many contexts are Why would optroderecordings reveal ited by food cues. more important than AgRP itself [3, 54, residual AgRP activity that was not To clarify these issues, it will be 61, 68, 70, 78]. In this regard, activation detected by calcium imaging? One possi- important to perform optrode recordings of AgRP neurons lacking both GABA bility is that the difference is technical. from AgRP neurons in freely behaving and NPY does not promote feeding in Calcium dynamics are only a surrogate mice and measure whether any residual the first 2 hours after stimulation [70]. for neural firing and provide relative, not activation persists after food presenta- This implies that any mechanism that absolute, measurements of neural activ- tion. In addition, it will be important to regulates acute feeding would neces- ity. Calciumsensorscan alsodisplaynon- test the functional importance of any sarily be mediated by a melanocortin linearity in their response properties residual AgRP activity by using methods independent signal (either GABA or outside a certain range. By contrast, with high temporal control, such as NPY). electrophysiologic recordings provide optogenetic silencers, to selectively A prediction of this model is that data on absolute firing rates by measur- inhibit AgRP neurons after food presen- there should exist a population of ing individual spikes. Thus it is possible tation and measure the kinetics of the feeding-regulatory neurons that serve that a residual activation of AgRP cessation of feeding. These experiments as the substrate for this sustained neurons following food presentation will enable dissection of the contribution response. These neurons would be was simply not detected in the calcium of residual AgRP neuron activity to located downstream of AgRP neurons imaging experiments due to a technical subsequent food consumption. in the feeding circuit and integrate AgRP limitation of the approach. neuron activity over time, so that their However there are reasons to suspect response to sudden changes in AgRP this is not the whole story. One reason is Mechanism #2: AgRP neurons neuron firing rate was delayed. As a that technical differences would not transmit a sustained hunger result, these downstream cells would obviously explain why the two studies signal appear to “remember” the history of found a different direction of modulation AgRP neuron activity, enabling them to for a significant subset of AgRP neurons; An alternative hypothesis is that AgRP promote feeding even after AgRP neu- that is, why would calcium imaging show neurons act through a sustained or rons have been silenced by sensory that 96% of AgRP neurons were inhibited persistent mechanism that enables the cues. If such neurons exist, their by food cues if 23% were actually activity of these neurons before food dynamics would correlate more closely

322 Bioessays 38: 316–324, ß 2016 WILEY Periodicals, Inc...... Insights & Perspectives Y. Chen and Z. A. Knight

with the subjective notion of hunger and NPY in fasting-activated hypothalamic 22. Gunaydin LA, Grosenick L, Finkelstein JC, than AgRP neurons themselves. neurons. Nat Neurosci 1: 271–2. Kauvar IV,etal. 2014. Natural neural projec- 3. Aponte Y, Atasoy D, Sternson SM. 2011. tion dynamics underlying social behavior. Cell

AGRP neurons are sufficient to orchestrate 157: 1535–51. again Think feeding behavior rapidly and without training. 23. Lima SQ, Hromadka T, Znamenskiy P, Conclusions and outlook Nat Neurosci 14: 351–5. Zador AM. 2009. PINP: a new method of 4. Krashes MJ, Koda S, Ye C, Rogan SC,etal. tagging neuronal populations for identifica- 2011. Rapid, reversible activation of AgRP tion during in vivo electrophysiological Over the past year we have witnessed neurons drives feeding behavior in mice. recording. PLoS ONE 4: e6099. some of the first glimpses into the J Clin Invest 121: 1424–8. 24. Berridge KC. 2004. Motivation concepts in dynamics of the key neural cell types 5. Luquet S, Perez FA, Hnasko TS, Palmiter behavioral neuroscience. Physiol Behav 81: RD. 2005. NPY/AgRP neurons are essential 179–209. that control feeding, including AgRP for feeding in adult mice but can be ablated in 25. Cannon WB. 1929. Organization for physio- and POMC neurons. This has revealed neonates. Science 310: 683–5. logical homeostasis. Physiol Rev 9: 399–431. that these cell types, long regarded as 6. Gropp E, Shanabrough M, Borok E, Xu AW, 26. Woods SC, Ramsay DS. 2007. Homeostasis: merely sensors of the internal state of et al. 2005. Agouti-related peptide- beyond Curt Richter. Appetite 49: 388–98. expressing neurons are mandatory for feed- 27. Stricker EM, Hoffmann ML. 2007. Presys- the body, in fact respond rapidly to ing. Nat Neurosci 8: 1289–91. temic signals in the control of thirst, salt sensory cues from the outside world. 7. Zhan C, Zhou J, Feng Q, Zhang JE,etal. appetite, and vasopressin secretion. Physiol They use this sensory information to 2013. Acute and long-term suppression of Behav 91: 404–12. feeding behavior by POMC neurons in the 28. Pavlov IP. 1902. London: Charles Griffin Co. predict nutritional changes that will brainstem and hypothalamus, respectively. Ltd. occur in the future, revealing anticipa- J Neurosci 33: 3624–32. 29. Woods SC. 1991. The eating paradox: how tory dynamics that are both remarkable 8. Cowley MA, Smart JL, Rubinstein M, we tolerate food. Psychol Rev 98: 488–505. Cerdan MG,etal. 2001. Leptin activates 30. Power ML, Schulkin J. 2008. Anticipatory and puzzling, since they confound anorexigenic POMC neurons through a neural physiological regulation in feeding biology: traditional explanations for how these network in the arcuate nucleus. Nature 411: cephalic phase responses. Appetite 50: neurons control behavior. An important 480–4. 194–206. challenge for the field will be to 9. Pinto S, Roseberry AG, Liu H, Diano S,etal. 31. Zafra MA, Molina F, Puerto A. 2006. The 2004. Rapid rewiring of arcuate nucleus neural/cephalic phase reflexes in the physiol- integrate these new observations into feeding circuits by leptin. Science 304: 110–5. ogy of nutrition. Neurosci Biobehav Rev 30: a revised model of the feeding circuit, 10. Cowley MA, Smith RG, Diano S, Tschop M, 1032–44. and we have highlighted here some of et al. 2003. The distribution and mechanism of 32. Steffens AB. 1976. Influence of the oral cavity action of ghrelin in the CNS demonstrates a on insulin release in the rat. Am J Physiol 230: the key questions that remain to be novel hypothalamic circuit regulating energy 1411–5. answered. As AgRP and POMC neurons homeostasis. Neuron 37: 649–61. 33. Strubbe JH, Steffens AB. 1975. Rapid insulin represent only a small piece of the 11. Nakazato M, Murakami N, Date Y, Kojima release after ingestion of a meal in the neural network that controls feeding, it M,etal. 2001. A role for ghrelin in the central unanesthetized rat. Am J Physiol 229: regulation of feeding. Nature 409: 194–8. 1019–22. is likely that additional surprises lie 12. Chen Y, Lin YC, Kuo TW, Knight ZA. 2015. 34. Powley TL. 1977. The ventromedial hypotha- ahead as we uncover the dynamics of Sensory detection of food rapidly modulates lamic syndrome, satiety, and a cephalic this circuitry. arcuate feeding circuits. Cell 160: 829–41. phase hypothesis. Psychol Rev 84: 89–126. 13. Betley JN, Xu S, Cao ZF, Gong R,etal. 2015. 35. Brand JG, Cagan RH, Naim M. 1982. Neurons for hunger and thirst transmit a Chemical senses in the release of gastric negative-valence teaching signal. Nature 521: and pancreatic secretions. Annu Rev Nutr 2: Acknowledgments 180–5. 249–76. 14. Mandelblat-Cerf Y, Ramesh RN, Burgess 36. Janowitz HD, Hollander F, Orringer D, Levy Y.C. is an HHMI International Student CR, Patella P,etal. 2015. Arcuate hypotha- MH,etal. 1950. A quantitative study of the Research Fellow. Z.A.K. is a New York lamic AgRP and putative POMC neurons gastric secretory response to sham feeding in Stem Cell Foundation – Robertson show opposite changes in spiking across a human subject. Gastroenterology 16: Investigator and acknowledges support multiple timescales. eLife 4: e07122. 104–16. 15. Seeley RJ, Berridge KC. 2015. The hunger 37. Klajner F, Herman CP, Polivy J, Chhabra R. from the New York Stem Cell Founda- games. Cell 160: 805–6. 1981. Human obesity, dieting, and anticipa- tion, the Rita Allen Foundation, the 16. Fan W, Boston BA, Kesterson RA, Hruby tory salivation to food. Physiol Behav 27: McKnight Foundation, the Alfred P. VJ,etal. 1997. Role of melanocortinergic 195–8. neurons in feeding and the agouti obesity 38. Feldman M, Richardson CT. 1986. Role of Sloan Foundation, the Brain and syndrome. Nature 385: 165–8. thought, sight, smell, and taste of food in Behavior Research Foundation, the 17. Ollmann MM, Wilson BD, Yang YK, Kerns the cephalic phase of gastric acid secre- Esther A. & Joseph Klingenstein Foun- JA,etal. 1997. Antagonism of central tion in humans. Gastroenterology 90: dation, the Program for Breakthrough melanocortin receptors in vitro and in vivo 428–33. by agouti-related protein. Science 278: 39. Woods SC, Kuskosky PJ. 1976. Classically Biological Research, and the UCSF 135–8. conditioned changes of blood glucose level. Diabetes Center (P30 DK063720). This 18. Seeley RJ, Yagaloff KA, Fisher SL, Burn P, Psychosom Med 38: 201–19. work was supported by an NIH New et al. 1997. Melanocortin receptors in leptin 40. Woods SC, Vasselli JR, Kaestner E, Szak- effects. Nature 390: 349. mary GA,etal. 1977. Conditioned insulin Innovator Award (DP2-DK109533) and 19. Huszar D, Lynch CA, Fairchild-Huntress V, secretion and meal feeding in rats. J Comp R01-DK106399. Dunmore JH,etal. 1997. Targeted disruption Physiol Psychol 91: 128–33. of the melanocortin-4 receptor results in 41. Kermani M, Eliassi A. 2012. Gastric acid obesity in mice. Cell 88: 131–41. secretion induced by paraventricular nucleus 20. Lu DS, Willard D, Patel IR, Kadwell S,etal. microinjection of orexin A is mediated through 1994. Agouti protein is an antagonist of the activation of neuropeptide Yergic system. References melanocyte-stimulating-hormone receptor. Neuroscience 226: 81–8. Nature 371: 799–802. 42. Takahashi N, Okumura T, Yamada H, 1. Weigle DS. 1994. Appetite and the regulation 21. Ziv Y, Burns LD, Cocker ED, Hamel EO, Kohgo Y. 1999. Stimulation of gastric acid of body composition. FASEB J 8: 302–10. et al. 2013. Long-term dynamics of CA1 secretion by centrally administered orexin-A 2. Hahn TM, Breininger JF, Baskin DG, hippocampal place codes. Nat Neurosci 16: in conscious rats. Biochem Biophys Res Schwartz MW. 1998. Coexpression of Agrp 264–6. Commun 254: 623–7.

Bioessays 38: 316–324, ß 2016 WILEY Periodicals, Inc. 323 Y. Chen and Z. A. Knight Insights & Perspectives.....

43. Smith GP. 2000. The controls of eating: a shift locomotor activity and accumbal dopamine- 68. Tong Q, Ye CP, Jones JE, Elmquist JK, from nutritional homeostasis to behavioral overﬂow via central cholinergic systems in et al. 2008. Synaptic release of GABA by neuroscience. Nutrition 16: 814–20. mice: implications for its involvement in brain AgRP neurons is required for normal regula- 44. Young RC, Gibbs J, Antin J, Holt J,etal. reward. Addict Biol 11: 45–54. tion of energy balance. Nat Neurosci 11: 1974. Absence of satiety during sham feeding 56. Sederholm F, Ammar AA, Sodersten P. 998–1000. in the rat. J Comp Physiol Psychol 87: 2002. Intake inhibition by NPY: role of 69. Clark JT, Kalra PS, Kalra SP. 1985. Neuro- 795–800. appetitive ingestive behavior and aversion. peptide Y stimulates feeding but inhibits 45. Davis JD, Campbell CS. 1973. Peripheral Physiol Behav 75: 567–75. sexual behavior in rats. Endocrinology 117: control of meal size in the rat. Effect of sham 57. Day DE, Keen-Rhinehart E, Bartness TJ. 2435–42. feeding on meal size and drinking rate. 2005. Role of NPY and its receptor subtypes 70. Krashes MJ, Shah BP, Koda S, Lowell BB. J Comp Physiol Psychol 83: 379–87. in foraging, food hoarding, and food intake by 2013. Rapid versus delayed stimulation of 46. Weingarten HP, Kulikovsky OT. 1989. Siberian hamsters. Am J Physiol Regul Integr feeding by the endogenously released AgRP Taste-to-postingestive consequence condi- Comp Physiol 289: R29–36. neuron mediators GABA, NPY, and AgRP.

Think again tioning: is the rise in sham feeding with 58. Ammar AA, Nergardh R, Fredholm BB, Cell Metab 18: 588–95. repeated experience a learning phenome- Brodin U,etal. 2005. Intake inhibition by NPY 71. Poggioli R, Vergoni AV, Bertolini A. 1986. non? Physiol Behav 45: 471–6. and CCK-8: a challenge of the notion of NPY ACTH-(1–24) and alpha-MSH antagonize 47. Davis JD, Smith GP. 1990. Learning to sham as an “Orexigen.” Behav Brain Res 161: 82–7. feeding behavior stimulated by kappa opiate feed: behavioral adjustments to loss of 59. Woods SC, Figlewicz DP, Madden L, Porte agonists. Peptides 7: 843–8. physiological postingestional stimuli. Am J D, Jr.,etal. 1998. NPY and food intake: 72. Farooqi IS, Keogh JM, Yeo GS, Lank EJ, Physiol 259: R1228–35. discrepancies in the model. Regul Pept et al. 2003. Clinical spectrum of obesity and 48. Stratton RJ, Stubbs RJ, Elia M. 2008. Bolus 75–76: 403–8. mutations in the melanocortin 4 receptor tube feeding suppresses food intake and 60. Tan K, Knight ZA, Friedman JM. 2014. gene. New Engl J Med 348: 1085–95. circulating ghrelin concentrations in healthy Ablation of AgRP neurons impairs adaption to 73. Vaisse C, Clement K, Durand E, Hercberg subjects in a short-term placebo-controlled restricted feeding. Mol Metab 3: 694–704. S,etal. 2000. Melanocortin-4 receptor trial. Am J Clin Nutr 88: 77–83. 61. Wu Q, Howell MP, Cowley MA, Palmiter mutations are a frequent and heterogeneous 49. Stratton RJ, Stubbs RJ, Elia M. 2003. Short- RD. 2008. Starvation after AgRP neuron cause of morbid obesity. J Clin Invest 106: term continuous enteral tube feeding sched- ablation is independent of melanocortin 253–62. ules did not suppress appetite and food signaling. Proc Natl Acad Sci USA 105: 74. Farooqi IS, Yeo GS, Keogh JM, Aminian S, intake in healthy men in a placebo-controlled 2687–92. et al. 2000. Dominant and recessive inheri- trial. J Nutr 133: 2570–6. 62. Berridge KC. 2007. The debate over dop- tance of morbid obesity associated with 50. Stratton RJ, Elia M. 1999. The effects of amine’s role in reward: the case for incentive melanocortin 4 receptor deﬁciency. J Clin enteral tube feeding and parenteral nutrition salience. Psychopharmacology 191: Invest 106: 271–9. on appetite sensations and food intake in 391–431. 75. Hinney A, Schmidt A, Nottebom K, Heibult health and disease. Clin Nutr 18: 63–70. 63. Skinner BF. 1938. The Behavior of Organ- O,etal. 1999. Several mutations in the 51. Cecil JE, Castiglione K, French S, Francis isms. New York: Appleton-Century-Crofts. melanocortin-4 receptor gene including a J,etal. 1998. Effects of intragastric infusions 64. Cabanac M. 1971. Physiological role of nonsense and a frameshift mutation associ- of fat and carbohydrate on appetite ratings pleasure. Science 173: 1103–7. ated with dominantly inherited obesity in and food intake from a test meal. Appetite 30: 65. Parkinson JR, Dhillo WS, Small CJ, humans. J Clin Endocrinol Metab 84: 1483–6. 65–77. Chaudhri OB,etal. 2008. PYY3–36 injection 76. Jennings JH, Rizzi G, Stamatakis AM, Ung 52. LeGall-Salmon E, Stevens WD, Levy JR. in mice produces an acute anorexigenic RL,etal. 2013. The inhibitory circuit architec- 1999. Total parenteral nutrition increases effect followed by a delayed orexigenic effect ture of the lateral hypothalamus orchestrates serum leptin concentration in hospitalized, not observed with other anorexigenic gut feeding. Science 341: 1517–21. undernourished patients. JPEN J Parenter hormones. Am J Physiol Endocrinol Metab 77. Hagan MM, Rushing PA, Pritchard LM, Enteral Nutr 23: 38–42. 294: E698–708. Schwartz MW,etal. 2000. Long-term 53. Wolf SG, Wolff HG. 1943. Human Gastric 66. Batterham RL, Cowley MA, Small CJ, orexigenic effects of AgRP-(83–132) involve Function, an Experimental of a Man and His Herzog H,etal. 2002. Gut hormone mechanisms other than melanocortin recep- Stomach. London, New York etc.: Oxford PYY(3–36) physiologically inhibits food intake. tor blockade. Am J Physiol Regul Integr Comp university press. Nature 418: 650–4. Physiol 279: R47–52. 54. Atasoy D, Betley JN, Su HH, Sternson SM. 67. Halatchev IG, Ellacott KL, Fan W, Cone RD. 78. Qian S, Chen H, Weingarth D, Trumbauer 2012. Deconstruction of a neural circuit for 2004. Peptide YY3–36 inhibits food intake in ME,etal. 2002. Neither agouti-related protein hunger. Nature 488: 172–7. mice through a melanocortin-4 receptor- nor neuropeptide Y is critically required for the 55. Jerlhag E, Egecioglu E, Dickson SL, independent mechanism. Endocrinology regulation of energy homeostasis in mice. Mol Andersson M,etal. 2006. Ghrelin stimulates 145: 2585–90. Cell Biol 22: 5027–35.

324 Bioessays 38: 316–324, ß 2016 WILEY Periodicals, Inc.