Modulation of Feeding and Associated Behaviors by Lateral Hypothalamic Circuits

Abstract

Our ability to modulate and observe neuronal activity in defined neurons in freely moving animals has revolutionized neuroscience research in recent years. Findings in the lateral hypothalamus (LHA) highlighted the existence of many neuronal circuits that regulate distinct phenotypes of feeding behavior, emotional valence, and locomotor activity. Several of these neuronal circuits do not fit into a common model of neuronal integration and highlight the need to improve working models for complex behaviors. This review will specifically focus on recent literature that distinguishes LHA circuits based on their molecular and anatomical characteristics and studies their role in feeding, associated behaviors (e.g., arousal and locomotion), and emotional states (e.g., emotional valences).

The central nervous system (CNS) is responsible for generating feelings of hunger and controlling food intake behaviors in response to internal and external stimuli (1). To understand the central processes involved in the pathogenesis of obesity and to identify potential drug targets, researchers have focused their studies on understanding hypothalamic functions. In the past decades, our understanding of hypothalamic modulations of many behaviors and physiological adaptations (including reproductive, sexual, aggression, social, thermoregulation, and food intake) has evolved tremendously (2–10).

Early research distinguished the hypothalamus into feeding and satiety centers, located in the lateral hypothalamic area (LHA) and the ventromedial hypothalamus, respectively, because electric stimulation or ablation showed robust and opposing changes in feeding behavior (11–14). The LHA also elicited robust electric self-stimulation (15), and the important interaction of reward, learning, and food intake was suggested (16). However, the neuronal origin that contributed to the observed feeding behaviors remained a matter of debate, because electric lesions and stimulations could modulate local cell bodies as well as fibers of passage (14). A major move forward came with methodological advances in immunohistochemistry and pathway tracing that allowed an unprecedented resolution of hypothalamic circuits and confirmed a rich LHA interconnection (17). With the discovery of individual neurons within the arcuate nucleus (ARC) that harbor their own sets of neurons to drive opposing feeding behaviors (18, 19), the idea of feeding and satiety centers has been replaced with opposing acting feeding and satiety neurons.

Over the past decade, a revolution took place in neuroscience research, with many exciting new technologies that allow the live observation or remote control of neuronal activity from identified neurons and their specific projection sites in freely behaving animals. This development resulted in a plethora of newly discovered neuronal populations and circuits that contribute to energy homeostasis and connect many central sites to feeding behavior or aspects of feeding behavior (20–24).

In the LHA, γ-aminobutyric acid (GABA)ergic and glutamatergic neuronal populations were recently identified according to their opposing effects on feeding and satiety. However, several LHA-induced behaviors are inconsistent with ARC feeding circuits. This inconsistency raises the question of whether and how LHA feeding circuits integrate into well-studied ARC feeding circuits. This review specifically focuses on recent literature of distinct GABAergic and glutamatergic LHA neuronal circuits, their role in feeding behavior, feeding-associated behaviors (emotional valences, arousal, and locomotion) and their relation to ARC feeding circuits. Finally, we highlight remaining gaps in our understanding of LHA feeding circuits.

Distinguishing Feeding Behaviors

Energy intake and energy expenditure are both highly variable and modulated parameters that ensure body weight maintenance. Energy intake and food intake are loosely defined terms that can refer to a variety of ingestive behaviors that have been well characterized in the past (1, 25). New molecular genetic methods that modulate or monitor acute changes in neuronal activation (e.g., optogenetics or fiber photometry) have more extensively focused on real-time behavioral changes (e.g., short-term food intake, food seeking, food consumption, positive and negative valences associated with feeding and energy status), in contrast to measuring daily food intake, and a clear distinction of these terminologies is warranted.

In this review, we use the term metabolic food intake to represent the daily caloric food intake necessary to fulfill metabolic needs and to maintain a stable body weight. This is in contrast to acute feeding behaviors, which contribute to metabolic food intake (Fig. 1) (1, 25) and can be distinguished into appetitive feeding (food seeking, foraging, food approach) and consumatory feeding (actual food consumption, meal, feeding bout). The decision to eat and how much to eat or not to eat is influenced by circadian time (sleep-wake cycle, arousal), external and internal factors that modulate the internal need states and determine whether we feel hungry or satiated. External factors include possible danger (e.g., predators), availability of food, and recognition of food sources. Internal factors are available energy stores, associated changes in circulating nutrients (glucose, fatty acids, and amino acids), and hormones (e.g., leptin, ghrelin, asprosin) and learned emotional valences that are associated with a food source (23, 26–28).

Figure 1.

Model of feeding behavior regulation. The internal state informs the body about energy need states, such as hunger and satiety, and is induced by changes in neuronal activity of ARC^AgRP neurons. Hunger and satiety are associated with negative or positive valences and play an integral role in modulating reward value or the incentive value of food and change the motivation to seek out food (e.g., lever pressing in an operant conditioning task). Activation of select neuronal populations in the ARC and LHA modulates both emotional valences and feeding behaviors, depending on activation timing and activation context.

Emotional valence is a measure of pleasure associated with an internal state or related context and can modulate the motivation to pursue a food reward [for further review, see Berridge and Kringelbach (29)]. These changes in motivation modulate learned Pavlovian and operant behavior and are used to measure emotional valences in conditioned place and food preference tasks (Pavlovian conditioning) or lever pressing for rewards (operant conditioning) (30–33). Importantly, emotional valences are modulated by internal state changes and are powerful enough to reverse the rewarding effect and incentive value of food, as measured in food reactivity tests (34) or in Pavlovian and operant conditioning tasks (32, 33). Food intake measured in conditioned tasks and with palatable foods (e.g., rich in sugar or fat) is called hedonic food intake (35), to emphasize the associated positive valence, and typically measures short-term feeding episodes. Although metabolic food intake is composed of these short-term feeding events, the distinction is important because the regulation of acute feeding behavior seems distinct from long-term regulation of metabolic food intake. For example, increased hedonic food intake episodes may be compensated throughout the day by reduced chow intake. Furthermore, only neuronal circuits that ultimately can modulate metabolic food intake and body weight are attractive targets to treat obesity.

ARC Neurons Mediate Hunger Valences to the LHA

The ARC mediates metabolic need states via two opposing sets of ARC neurons: anorexigenic pro-opiomelanocortin (POMC) expressing neurons and orexigenic agouti-related peptide (AgRP) expressing neurons (18, 19, 36). During metabolic need states, such as fasting, humoral signals change dramatically (e.g., the adipose tissue–derived hormone leptin decreases to signal depletion of body fat stores, whereas the gut hormone ghrelin increases before meals) (37–40). ARC^POMC and ARC^AgRP neurons are differentially regulated by leptin and ghrelin, resulting in neuronal silencing of ARC^POMC neurons and neuronal activation of ARC^AgRP neurons in fasted, hungry mice (41–45).

The behavioral feeding responses to fasting can be mimicked by activation of ARC^AgRP neurons alone (notably in the absence of a metabolic need state), strongly supporting ARC^AgRP neurons as an integral way to communicate the feeling of hunger that is associated with metabolic need states (18, 19, 36, 46). The LHA receives strong input from ARC^POMC and ARC^AgRP neurons, and both neurons and their peptides are able to modulate neuronal activity in LHA neurons (47–49). Importantly, ARC^AgRP neurons mediate hunger-induced feeding behavior via several brain sites that include the paraventricular nucleus (PVN), bed nucleus of stria terminalis (BNST), and the LHA itself (33, 50).

Activation of ARC^AgRP neurons results in long-term metabolic food intake changes that are able to cause body weight changes according to the activation level of ARC^AgRP neurons and also increases the incentive value of food in a progressive ratio task (19, 46, 51). Thus, activation of ARC^AgRP neurons and hunger information is identified by changes in metabolic food intake (24 hours food intake) and hedonic food intake (operant responding, palatable food intake). Importantly, hunger or activation of ARC^AgRP neurons mediates a negative valence that will cause conditioned avoidance (32), which is in line with the idea of informing the body about a need state and motivating the animal to escape the need state, in this case by seeking and eating food. Conversely, ARC^AgRP neuronal activation is quickly reversed by meal initiation (consumatory feeding) and even by the mere sight or approach of food (appetitive feeding) (42), indicating that this fast deactivation signal must be independent of the internal need state.

Additional clues that the feeling of hunger underlies a more complex regulatory system than initially thought came from the observation that ARC^AgRP activation seemed to dynamically change emotional valences from positive (initial increased operant responding) to negative (decreased operant responding) with ongoing activation during operant testing (32). Indeed, additional studies highlighted that ARC^AgRP neurons induced a positive valence with neuronal prestimulation where food remains absent, whereas operant testing (food accessibility) occurred without ongoing neuronal activation. This paradigm more authentically reproduced the physiological ARC^AgRP activation pattern observed in the transition from fasting to refeeding. Prestimulation of ARC^AgRP neurons in the absence of food caused strong and long-lasting (days) increases in appetitive (lever pressing) and consumatory (caloric intake) feeding responses and induced positive valence (food preference and place preference) (33). Taken together, these data indicate that activation of ARC^AgRP neurons underlies modulation by fast sensory signals (seconds) and long-term internal signals (e.g., humoral factors). Furthermore, the timing and context of ARC^AgRP neuronal firing dynamically change emotional valences from negative to positive. Finally, these prestimulation effects modulate appetitive and consumatory feeding via LHA circuits (as well as ARC^AgRP > BNST and ARC^AgRP > PVN circuits), even though metabolic feeding and emotional valences were not directly evaluated (33). Thus, it remains to be determined which LHA neurons could integrate ARC^AgRP feeding circuits.

The LHA as a Central Interconnection Hub Within the CNS

The rodent hypothalamus is organized into well-defined nuclei with densely packed cell bodies (e.g., ARC, ventral and dorsomedial hypothalamus, PVN). The larger LHA is an exception with its organization into less dense neuronal cell bodies that allows the passage of neuronal processes via massive bundles known as the median forebrain bundle (mfb). The mfb reaches through the brain like a highway to connect neuronal cell bodies from caudally arising neurons (e.g., midbrain and brainstem) to forebrain structures and vice versa. Also, cell bodies within the LHA enter the mfb and receive innervation from mfb fibers (52). The fornix (fx) is another prominent fiber bundle within the LHA that carries neuronal fibers from the hippocampus to posterior sites, leading through the LHA, and extends anterior into the septal nuclei and nucleus accumbens (NAc). The fx is also widely used as an LHA landmark.

The mfb plays also an integral part in connecting key components of the limbic system. Dopaminergic (DA) neurons in the ventral tegmental area (VTA) propagate their signals from the VTA to the forebrain (striatum and NAc) and vice versa via the mfb (53). The NAc has been implicated as a key structure in translating motivation into motor function (54), and thus it is important for reward behavior. The LHA shows prominent interconnection with VTA circuits and receives innervation from the NAc, which is in line with many studies that connect LHA function to reward behavior and hedonic feeding (55–57).

Another unique feature of the LHA is its enormous interconnection with other central sites and number of distinct nuclei that have been distinguished (58, 59). Therefore, the LHA interconnects with all other major output systems of the brain, including the autonomic nervous system via the brainstem, spinal cord, and PVN, the neuroendocrine axis via the PVN and other hypothalamic sites, and the cognitive processing, memory, and emotional valences via the cortex, thalamus, and limbic system (Fig. 2). Based on projection patterns or functional studies, the LHA is further implicated in a variety of behaviors such as feeding (60), foraging and exploration (58, 61), sleep/wake states and arousal (62, 63), and anxiety (64, 65).

Figure 2.

LHA central connectivity. The LHA interconnects to all major output systems of the brain (endocrine and autonomic control, behavior, learning, and memory). The relative importance of LHA targets and inputs for feeding and associated behavior is not fully understood and requires systematic and unbiased comparison of LHA connections. Acb, nucleus accumbens; DMH, dorsomedial nucleus of the hypothalamus; DMV, dorsal motor nucleus of vagus; LC, locus coeruleus; NTS, nucleus of the solitary tract; PAG/MLR, periaqueductal gray/mesencephalic locomotor region; PBN, parabrachial nucleus; SNC, substantia nigra pars compacta; Thal, thalamus; VMH, ventromedial hypothalamus.

Thus, the interconnected organization of the LHA is well suited to orchestrate environmental (sight, small, taste, temperature, cognition) and metabolic (blood glucose, circulating hormones, inflammation) stimuli into appropriate contextual responses from autonomic function, the neuroendocrine axis, and behavior [for detailed review, see Berthoud and Münzberg (53)]. A more detailed study of small LHA subdivisions has further highlighted that each of these sites shows a high degree of interconnectivity within and outside the LHA (59, 66, 67). Despite the massive interconnectivity, anatomical projection specificity exists for limbic inputs to the LHA, with possible functional relevance to modulate motivated behaviors (68, 69). Yet a consistent model of information flow and integration across various LHA inputs and outputs that can predict the modulation of food intake is not in place.

GABA and Glutamate LHA Neurons Mediate Different Aspects of Feeding Behavior

Recent work in the LHA highlighted the role of opposing acting neurons in feeding behavior based on their neurotransmitter characteristic of excitatory glutamate or inhibitory GABA. The transporters for glutamate or GABA, vesicular glutamate transporter-2, and vesicular GABA transporter, respectively, have been most useful to visualize and manipulate these two neuronal populations in many central sites (70, 71). The LHA shows wide distribution of both stimulatory glutamate (LHA^glutamate) and inhibitory GABA (LHA^GABA) neurons that have opposing roles in metabolic and hedonic feeding and promote emotional valences in animals that modulate conditioned preferences (57).

Emotional valences and feeding behavior

As outlined earlier, activation of ARC^AgRP neurons can induce positive and negative valences depending on the timing of stimulation and availability of food during stimulation. In a continuous stimulation paradigm that induced a negative valence in ARC^AgRP neurons, LHA^GABA neurons promote food intake in conjunction with a positive valence (72), whereas LHA^glutamate neurons inhibit food intake in conjunction with a negative valence (73). These data indicate that ARC^AgRP-induced feeding behavior overall may differ from LHA^GABA-induced feeding behavior, at least with regard to the associated emotional valence. On the other hand, the opposing effects of LHA^GABA and LHA^glutamate neurons to modulate food intake and emotional valance fit well into the overarching model for hunger-driven food intake (Fig. 1). Furthermore, prestimulation paradigms in ARC^AgRP neurons will induce hedonic and metabolic feeding that are associated with a positive valence and involve interaction with LHA neurons (33).

The dynamic nature of emotional valences is also demonstrated when they compete with other motivational states, such as anxiety or social interactions. Fasted mice (physiological hunger) and activation of ARC^AgRP neurons would overcome competing motivational states such as anxiety and sexual or social interaction to gain access to food. Yet this effect was not observed when LHA^GABA neurons were activated (74), arguing again that ARC^AgRP- and LHA^GABA-induced food intake may represent divergent feeding circuits.

Neuronal connectivity is another variable that contributes to the complexity of behavioral and emotional modulations. Positive and negative valences of both LHA populations are reproduced by selective stimulation of their projections to the VTA, where LHA^GABA > VTA afferents stimulate VTA^dopamine neurons (by disinhibition of VTA^GABA interneurons), whereas LHA^glutamate > VTA afferents inhibit VTA^dopamine neurons (stimulation of VTA^GABA interneurons) (75). However, these LHA^GABA > VTA projections mediate only metabolic feeding, not hedonic feeding, and LHA^glutamate > VTA afferents have no effect on food intake. Instead, LHA^glutamate neurons mediate negative valence (and inhibition of food intake) via the lateral habenula (76), thus supporting the existence of several redundant and possibly independent-acting circuits that modulate feeding and emotional valences.

Taken together, the opposing effects of LHA^GABA and LHA^glutamate neurons fit well into an internal state–driven model of food intake and emotional valences that integrates feeding signals from the ARC. However, it remains unclear how the dynamic change of emotional valences promoted by ARC^AgRP neurons fits into these circuits. Even though ARC^AgRP > LHA afferents were confirmed to induce acute and operant feeding (32, 33), the associated negative valence was not evaluated for specific ARC^AgRP projections. ARC^AgRP > LHA afferents may also target subsets of LHA populations that would explain diverse behavioral phenotypes (e.g., the suprafornical LHA region, which receives many inputs relevant for feeding behavior) (66).

Overall, the molecular and anatomical diversity within the population of LHA^GABA and LHA^glutamate neurons must be considered, which probably combines molecularly heterogeneous neurons, and divergent projection patterns, specifically considering the strong interconnectivity of the LHA as a whole as well as small LHA subdivisions (59, 66, 67). Indeed, LHA^GABA and LHA^glutamate neurons are distinguishable into several neuronal subpopulations, according to their expression of hormone receptors, neuropeptide expression, projection sites, or their feeding-related activation pattern (Fig. 3).

Figure 3.

GABA and glutamate subpopulations in the LHA. The LHA distinguishes stimulatory glutamatergic (GLUT) neurons and inhibitory GABAergic neurons. Additional subpopulations are found within each group with distinct projection sites. Dyn, dynorphin; GABA_A, GABA-receptor; Gal, galanin; GalR, galanin receptor; Hcrt/Ox, hypocretin/orexin; LHb, lateral habenual; MCH, melanin concentrating hormone; NMDA, N-methyl-d-aspartate; NT1, neurotensin receptor-1; Nts, neurotensin; OxR, hypocretin/orexin/receptor.

Consumatory and appetitive LHA neurons

LHA^GABA neurons are composed of discrete neuronal populations that are activated with appetitive or consumatory feeding behaviors (72). The distinction was made possible by the use of an implantable gradient-index lens to visualize individual neurons and simultaneously detect calcium-sensitive changes in neuronal activity. Such studies are missing in LHA^glutamate neurons. However, LHA^glutamate subpopulations such as hypocretin/orexin (Hcrt/ox) (LHA^Hcrt/ox)-expressing neurons are activated with arousal and food rewards (77–79), whereas lateral habenula–projecting LHA^glutamate neurons suppress consumatory feeding and promote aversion (76). Thus, LHA^glutamate neurons may similarly dissociate into neuronal subgroups based on their feeding-related activation patterns (e.g., consumatory vs appetitive feeding). The LHA is typically viewed as an integrator of hypothalamic signals, yet the dissociation of neurons that are associated with appetitive vs consumatory feeding indicates further distinction of feeding circuits at the level of the LHA. Future studies in ARC^AgRP neurons and other feeding-related populations should elucidate how common the distinction into appetitive and consumatory responsive neurons is throughout the hypothalamus.

Metabolic and hedonic feeding mediated by LHA neurons

Ablation of LHA^GABA and LHA^glutamate neurons have opposing effects on metabolic and hedonic feeding, sufficient to cause body weight changes (72, 76), and both densely innervate the VTA. LHA projections to the VTA and subsequent regulation of DA neurons are proposed as a key component of hedonic feeding (72, 75, 80, 81). Thus, the failure of LHA > VTA projections to modulate hedonic feeding after optogenetic stimulation was surprising (75). However, it was noted earlier that chemical lesions or blockade of dopamine signaling in the NAc had strong effects on locomotion but did not affect food intake (82, 83), so that dopamine neurons may promote the motivational aspects of feeding rather than influencing food intake per se (84).

The comparison of LHA neurons with and without VTA projections gave further insights into alternative circuits that promote hedonic feeding (Fig. 4). LHA^GABA neurons are composed of long-form leptin receptor (Lepr)– (85), neurotensin (Nts)- (86), and galanin- (87) expressing neurons, and all of these innervate LHA^Hcrt/ox neurons and are well associated with food reward (78, 86, 88). Nts and galanin are coexpressed in many LHA neurons (53); however, LHA^Nts and LHA^galanin neurons are further separated by their distinct projection profiles (86, 88, 89). Only LHA^Nts, but not LHA^galanin neurons, innervate the VTA, whereas LHA^galanin but not LHA^Nts neurons innervate the locus coeruleus (LC) (86, 88, 89) (Fig. 3). In line with these differential projection profiles, neuronal activation induced clearly distinguishable feeding responses. Activation of LHA^galanin neurons (no VTA projections) is sufficient to enhance food reward, similar to activation of overall LHA^GABA neurons, but had no effect on metabolic food intake (90). In contrast, activation of LHA^Nts neurons (including substantial VTA-projecting neurons) suppressed metabolic food intake and had no effect on palatable food intake (91). This demonstrates that LHA^GABA neurons are composed of several neuronal subsets that drive different aspects of food intake. Of particular interest is that LHA^galanin neurons increase hedonic feeding without affecting metabolic feeding, further demonstrating that several possibly redundant and independently regulated circuits can be distinguished based on their feeding behavior, at least in experimental settings.

Figure 4.

Diverse LHA^GABA and LHA^glutamate circuits and their associated feeding behavior and emotional valence. Behavioral changes follow neuronal activation or projection-specific activation of select LHA circuits.

Furthermore, LHA^Nts > VTA projecting neurons drive rewarding valences and self-stimulation via Nts receptor-1 and N-methyl-d-aspartate glutamate receptors (80), indicating that VTA-projecting LHA^Nts neurons also include glutamatergic neurons. Similarly, LHA^Hcrt/ox neurons (distinct from LHA^Nts neurons) contribute to glutamatergic VTA inputs that promote hedonic feeding (55, 92, 93), whereas overall LHA^glutamate > VTA projections have no effect on food intake (75). Again, these inconsistent data suggest a more complex organization of LHA > VTA inputs, with distinct circuits that promote discrete sets of behavioral phenotypes and may not require direct interaction with the VTA. Such indirect interaction could involve inhibitory inputs to LHA^Hcrt/Ox neurons (94), even though enhanced reward behavior has been associated with activation of LHA^Hcrt/Ox neurons (78, 95).

Furthermore, it remains unclear how neuropeptides and humoral inputs contribute to the observed behaviors at a circuit level. The anorexigenic (leptin, ghrelin, POMC-derived peptides) and orexigenic (Nts, galanin, Hcr/Ox, neuropeptide Y, AgRP) properties of many neuropeptides and hormones have been well described [for review, see Berthoud and Münzberg (53)]. Furthermore, many gut hormones that were first described to act in the brainstem were found to also interact with hypothalamic and limbic sites [for review, see Williams (96) and Grill and Hayes (97)], suggesting further complexity to modulate feeding circuits. However, there has been little progress at the circuit levels to integrate neuropeptide and humoral impact. Some data exist on melanocortin-4 receptors [(MC4Rs) which are stimulated by POMC-derived neuropeptides and inhibited by AgRP] that highlight MC4R-dependent (ARC^AgRP > PVN) and MC4R-independent (ARC^AgRP > LHA) feeding circuits (98). Also, the different temporal effects of neuropeptide Y, AgRP, and GABA, or G-protein coupled receptor signaling (G_svs G_q), have been addressed for ARC^AgRP-induced feeding behavior but with no context for how it may affect emotional valences or LHA feeding circuits (51, 99). Furthermore, the role of LHA^Hcrt/ox neurons and Hcrt/Ox peptide itself has not been addressed on a circuit level. These studies are further complicated because new studies using single-cell quantitative PCR suggest that at least some LHA^Hcrt/ox neurons may have GABA and glutamatergic characteristics and can express galanin mRNA (100), despite histologic evidence that these populations do not overlap (87).

Arousal, Locomotion, and Anxiety

Beyond feeding, the LHA plays an important role in arousal, which accompanied the discovery of LHA^Hcrt/Ox neurons (101, 102). LHA^Hcrt/Ox neurons are critical for arousal, and their activation level modifies sleep/wake states via their projections to the LC (103). LHA^Hcrt/Ox neurons are implicated in hedonic food intake (77–79, 93), but reported effects on metabolic feeding are minimal (104). Therefore, the feeding effects via LHA^Hcrt/Ox neurons may be an indirect result of the heightened arousal, because low arousal or even sleep is incompatible with feeding behavior. Indeed, recent studies further emphasized the role of LHA^Hcrt/Ox neurons in arousal, indicated by enhanced locomotor activity. Activation of glutamatergic LHA^Hcrt/ox neurons induces robust locomotor activity (104, 105), which was sufficient to blunt high-fat diet–induced weight gain in mice but had no effect on metabolic feeding (105). LHA^Hcrt/Ox neuronal stimulation also influences LHA circuitries (e.g., activation of LHA^GABA neurons), which similarly induces robust locomotion (75, 90, 106).

Currently, our understanding of locomotor activity and its relevance for diverse behavior is not well defined. Locomotion is easily evaluated by measuring infrared beam breaks, but it fails to reflect the underlying motivation that causes locomotor events. For instance, activation of LHA^GABA neurons (strong VTA projections) or LHA^galanin neurons (no VTA projections) both induced very similar home cage locomotor activity and energy expenditure (90), but locomotion was unchanged in stress-related open field settings after LHA^GABA activation (107). More thorough testing revealed distinct locomotor modalities. Activation of LHA^galanin neurons increased the overall occurrence of normal mouse behavior with frequent transitions across diverse behaviors (grooming, exploration, digging, feeding), and this behavior was associated with an anxiolytic valence, possibly reflecting increased arousal and willingness to explore. In striking contrast, activation of LHA^GABA neurons induced repetitive behavior with compulsive gnawing or digging and few behavioral transitions, which was associated with anxiogenic valence (90). LHA^GABA > VTA projections also mediated repetitive gnawing and licking behavior (75, 81), which can lead to noncaloric consumatory intake (107), suggesting that LHA^GABA > VTA circuits are responsible for compulsive locomotion. Furthermore, activation of LHA^Nts neurons increases locomotor activity but is accompanied by decreased feeding behavior (91), in striking contrast to activation of LHA^GABA and LHA^galanin neurons that increased metabolic or hedonic feeding (90). An increase in locomotor activity is also in line with food-seeking and foraging behavior, which may be difficult to interpret properly in the absence of food-directed consumatory behavior. However, it is conceivable that foraging behavior may be more clearly defined by increased locomotor activity, coupled with reduced anxiety, consistent with findings that physiological hunger and ARC^AgRP stimulation are able to overcome fear of predator odor for the sake of finding food (74).

Locomotion inherently intertwines with almost all behaviors (e.g., feeding, learning tasks, arousal, drinking, anxiety, social interaction), and it is difficult to dissociate locomotor function from ingestive behavior. This effect has been also noted for the mesolimbic dopamine system, where VTA^DA neurons are tightly connected with food reward and the motivation to work for food. However, deletion of DA neurons or inputs to the NAc alone only diminishes locomotor activity but is not sufficient to modulate food intake (82, 83). Such depletion may result in reduced effort to obtain a food reward but would not change intake of freely available food (84). Similarly, compulsive locomotion as observed after activation of LHA^GABA neurons or its projections to the VTA may interfere or even mask hedonic feeding. For example, chemogenetic activation of LHA^GABA neurons required individual dose-response analysis to unmask increased operant responding, due to overwhelming compulsive locomotor behavior with higher stimulation doses (90); this response was also noted as a confounding effect with LHA^GABA > VTA stimulation and the inability to increase compulsive operant responding (75, 81).

Thus, locomotor activity can be associated with different feeding and non–food-related behavior, with diverse influences on emotional valences. An important task for future research is a better classification of contextual behavior that defines locomotor activity as it relates to feeding behavior, behavioral setting, and emotional valences [e.g., by improving the visualization of complex behavior analysis, as elegantly shown by Sterley et al. (108)]. Furthermore, projection-specific studies of feeding and associated behaviors (e.g., emotional valence), as performed comprehensively for feeding responses via ARC^AgRP neurons (50), will allow evaluation if select feeding circuits integrate into each other or build independent feeding circuits. Particularly for the LHA, the VTA-centric view to modulating feeding behavior requires an unbiased evaluation of the targets and their relative contribution to feeding and associated behaviors.

Conclusion and Perspective

CNS circuits that modulate feeding behavior ultimately need to modulate the final motor neurons that allow feeding behavior coordination for food approach and consumption (holding, biting, chewing, and swallowing of food). The need to coordinate many behavioral components and sensory inputs (e.g., metabolic state, emotional state, sight and smell of food) has led to the overall view that these components must be integrated at various levels from central circuits to final output systems (neuroendocrine release, autonomic outflow, and motor function and behavior).

From an anatomical perspective, the impressive interconnectivity of the LHA with other central sites and its ability to sense and respond to physiological state changes make the LHA an ideal integration center for all components of feeding behavior [for further review, see Berthoud and Münzberg (53) and Petrovich (56)]. In contrast, some LHA populations (e.g., LHA^Nts and LHA^galanin neurons) show limited extra-LHA projection sites and modulation of distinct feeding behaviors.

We have highlighted that LHA^GABA and LHA^glutamate neurons consist of diverse neuronal populations based on neurotransmitter and neuropeptide expression patterns, extrahypothalamic projection patterns, and the behaviors elicited by neuronal activation. These data suggest that the LHA not only plays an integrating role for internal and external stimuli but also dissociates behavioral components (e.g., appetitive, consumatory, hedonic, metabolic food intake). Thus, one future task is to identify additional genetic markers that identify more homogeneous LHA subpopulations [e.g., as recently performed with single–LHA neuron analysis (100)] that allow identification of more specific neuronal markers.

The finding that complex behavior such as food intake involves distinct subsets of neurons that respond to select behavioral components (appetitive vs consumatory food intake) also shows that we may be just beginning to properly define and categorize complex behavior. Specifically, as other central sites have found similar behavioral dissociations of complex behavior, prey killing can be dissected into prey pursuit (appetitive behavior) via amygdala > brainstem projecting neurons and a killing bite (this fits the category of aggressive behavior and consumatory behavior) via amygdala midbrain projecting neurons (109). This task is particularly difficult, because most behaviors include some form of locomotion, and interpretation is particularly prone to investigator bias, depending on the system investigated (e.g., food intake, aggression, anxiety, arousal). Therefore, the compulsive locomotor behavior that has been noted by some investigators when stimulating LHA^GABA neurons has been interpreted as consumatory (gnawing or food intake), increased arousal, repetitive behavior, or anxiety associated, depending on the experimental approaches or context of study (72, 81, 90, 107, 110). This finding highlights an important need for future studies to find a more systematic way to classify locomotor behavior and identify the main behavioral drive (e.g., food intake vs anxiety) that underlies the change in locomotion.

Additionally, it is important to account for the internal (physiological and emotional status) and external (environment) context of the animal when interpreting behavioral outcomes. The discussed studies strengthen the view that positive and negative valences are an important motivating aspect for feeding behavior. We have only begun to unravel the importance of timing for neuronal firing (prestimulation vs continuous) and the environmental context when neuronal firing occurs (e.g., food availability, competing need states such as anxiety, social interaction, or mating). In the future it will be important to be more conscientious about the dynamic changes that are possible when it comes to emotional valences and associated changes in motivational drives for feeding.

Glossary

Abbreviations:

AgRP

agouti-related-peptide

ARC

arcuate nucleus

BNST

bed nucleus of stria terminalis

CNS

central nervous system

DA

dopaminergic

fx

fornix

GABA

γ-aminobutyric acid

Hcrt/ox

hypocretin/orexin

LC

locus coeruleus

Lepr

leptin receptor

LHA

lateral hypothalamus

MC4R

melanocortin-4 receptor

mfb

median forebrain bundle

NAc

nucleus accumbens

Nts

neurotensin

POMC

pro-opiomelanocortin

PVN

paraventricular nucleus

VTA

ventral tegmental area