**Preface to "Hypothalamic Regulation of Obesity"**

The obesity epidemic is a major socioeconomic problem that urgently necessitates better understanding of the mechanisms that mediate the imbalance between food intake and energy expenditure and of obesity-related metabolic and cardiovascular complications. Gaining insight into the cellular basis of obesity could lay the foundations for the development of new therapeutic strategies.

In the last few decades, it has been demonstrated that the hypothalamus regulates energy homeostasis. The hypothalamus contains hormonal- and nutrient-sensing nuclei that organize central and peripheral responses to maintain normal body weight, food intake, energy expenditure and nutrient partitioning. Within the hypothalamus, numerous specialized neuronal populations are connected to each other and to various extrahypothalamic brain regions to coordinate energy homeostasis. Evidence also suggests non-neuronal populations, such as astrocytes, participate, and interesting interplay may exist between astrocytes and hypothalamic neurons, whose disruption leads to insulin resistance and obesity.

This reprint reports on the most recent insights into the hypothalamic neuronal and non-neuronal pathways involved in obesity development. Novel findings related to key systems, such as endocannabinoid and melanocortin-related pathways, in the hypothalamus are discussed. In addition, the emerging contribution of the crosstalk between the hypothalamus and peripheral tissues, such as gut and adipose tissue, in obesity is a promising topic. Therefore, here, we present an overview of the most recent mechanisms underlying the hypothalamic regulation of obesity and the impact of these investigations in the prevention and treatment of human obesity for the scientific community and doctors. We thank the authors who have contributed their scientific findings to this reprint.

> **Rosal´ıa Rodr´ıguez-Rodr´ıguez and Cristina Miralpeix** *Editors*

## *Editorial* **Hypothalamic Regulation of Obesity**

**Rosalía Rodríguez-Rodríguez 1,\* and Cristina Miralpeix <sup>2</sup>**


Obesity has now reached pandemic proportions and represents a major socioeconomic and health problem in our societies. Up to now, obesity has been considered a medical issue only in high-income countries, but this disease is now dramatically on the rise in lowand middle-income countries, particularly in urban settings, affecting both adults and the pediatric population [1].

The simplest definition of overweight and obesity is an excessive accumulation of fat due to an imbalance between the energy in and the energy out, and this adiposity represents a risk to health [1]. But the whole picture is more complex than this, and obesity is a marker of a serious metabolic dysregulation that involves several diseases such as type 2 diabetes, cancer, cardiovascular pathologies, and most recently for COVID-19 infection [1,2]. Since the current approaches to combatting obesity and its complications have limited clinical effectiveness, gaining insight into the cellular and molecular basis of obesity could lay the foundations for the development of new strategies to prevent metabolic disruption and to treat this somewhat unaddressed medical issue.

In the last few decades, it has been strongly demonstrated that the hypothalamus is the master regulator of energy homeostasis. The hypothalamus contains hormone- and nutrient-sensing nuclei that organize central and peripheral responses for maintaining normal body weight, food intake, energy expenditure, and nutrient partitioning. Within the hypothalamus, specialized sub-populations of neurons are connected to each other and to various extrahypothalamic regions to coordinate energy balance. Evidence has also highlighted the participation of non-neuronal populations (i.e., microglia and astrocytes), and even the interesting cross-talk between these types of brain cells, whose disruption leads to insulin resistance and obesity. In this Special Issue, we report on the most recent insights into the hypothalamic circuitries and pathways involving neurons [3–7], astrocytes [8,9], and microglia [10,11] in obesity development and associated complications. The emerging contribution of astrocyte–neuron [9] and microglia–neuron cross-talks [12] in the hypothalamus and the contribution of microbiota and the gut-brain axis controlling food intake and energy homeostasis [13] are also presented in this Special Issue (Figure 1).

Regarding the importance of hypothalamic targets in obesity development, Quiñones et al. [3] elegantly review the most relevant studies focusing on the deacetylases sirtuins in hypothalamic neurons, SIRT1 and SIRT6, as multifaceted mediators of energy metabolism, controlling processes such as food intake, food preference, puberty, body weight, adiposity, and glucose and insulin homeostasis. Importantly, the complexity of the hypothalamic nuclei in obesity development is supported by the fact that the effect of SIRT1 on energy balance depends on the neuronal type where it is acting. For instance, lacking SIRT1 or SIRT6 in POMC neurons affects energy expenditure and adiposity; however, SIRT1 in AgRP neurons specifically affects eating behavior, and in oxytocin neurons in the PVN controls diet preference. In addition to sirtuins, Fosch et al. [4] provide new insights into other targets in SF1 neurons of the ventromedial hypothalamus that play a critical role not only in body weight control and adiposity, but also in glucose tolerance and insulin and

**Citation:** Rodríguez-Rodríguez, R.; Miralpeix, C. Hypothalamic Regulation of Obesity. *Int. J. Mol. Sci.* **2021**, *22*, 13459. https://doi.org/ 10.3390/ijms222413459

Received: 9 December 2021 Accepted: 13 December 2021 Published: 15 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

leptin sensitivity, with minimal effects on food intake and differential actions between male and female mice. The authors also report how nutrient sensors such as AMPK and SIRT1, glutamatergic transmission, synaptic receptors, and mediators of autophagy in SF1 neurons are promising therapeutic targets against obesity and diabetes, reinforcing the "central role" of the hypothalamus in the control of peripheral metabolism beyond the regulation of feeding behavior [4]. In this sense, Fukumura et al. [5] found new evidence in this special issue about the effects of a novel hypothalamic small protein, named neurosecretory protein GL (NPGL), on glucose and insulin homeostasis. In this study, overexpression of *Npgl* in the mediobasal hypothalamus of mice improved glucose tolerance and attenuated insulin resistance and hyperglycemia under high-fat diet (HFD) exposure, without significant changes in food intake. The concomitant increased mRNA expression levels of galanin (a neuropeptide that co-localizes with NPGL in arcuate neurons and regulates glucose homeostasis) suggests that both peptides could regulate each other for these actions [5]. actions between male and female mice. The authors also report how nutrient sensors such as AMPK and SIRT1, glutamatergic transmission, synaptic receptors, and mediators of autophagy in SF1 neurons are promising therapeutic targets against obesity and diabetes, reinforcing the "central role" of the hypothalamus in the control of peripheral metabolism beyond the regulation of feeding behavior [4]. In this sense, Fukumura et al. [5] found new evidence in this special issue about the effects of a novel hypothalamic small protein, named neurosecretory protein GL (NPGL), on glucose and insulin homeostasis. In this study, overexpression of *Npgl* in the mediobasal hypothalamus of mice improved glucose tolerance and attenuated insulin resistance and hyperglycemia under high-fat diet (HFD) exposure, without significant changes in food intake. The concomitant increased mRNA expression levels of galanin (a neuropeptide that co-localizes with NPGL in arcuate neurons and regulates glucose homeostasis) suggests that both peptides could regulate each other for these actions [5].

insights into other targets in SF1 neurons of the ventromedial hypothalamus that play a critical role not only in body weight control and adiposity, but also in glucose tolerance and insulin and leptin sensitivity, with minimal effects on food intake and differential

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 2 of 4

**Figure 1.** In this Special Issue, we report on the most recent insights into the neuronal targets and neuropeptides of the hypothalamic circuitries involved in obesity development and progression. In addition, the emerging contribution of astrocyte-neuron and microglia-neuron cross-talks in the hypothalamus in response to a high-fat diet, and the implication of the gut-brain axis controlling food intake and energy homeostasis are also presented **Figure 1.** In this Special Issue, we report on the most recent insights into the neuronal targets and neuropeptides of the hypothalamic circuitries involved in obesity development and progression. In addition, the emerging contribution of astrocyte-neuron and microglia-neuron cross-talks in the hypothalamus in response to a high-fat diet, and the implication of the gut-brain axis controlling food intake and energy homeostasis are also presented

When investigating the hypothalamic regulation of obesity, in addition to the identification of potential targets mediating the homeostatic circuits, it is also important to explore the hedonic circuits mediating energy balance, identifying the salient neural, hormonal, and humoral components involved, as remarkably reviewed by Gastelum et al. [7]. They explore the sexual dimorphism on these circuits, paying special attention to the role of two emerging neuropeptides, the nociception/orphanin FQ and the pituitary adenylate cyclase-activating polypeptide (PACAP) in the neuronal activity regulation in positive and negative energy states. Jang et al. [6] also investigate leptin-mediated feeding circuits in hypothalamic neurons, identifying the expression of the angiopoietin-like growth factor (AGF) in POMC neurons of the arcuate nucleus as a downstream factor When investigating the hypothalamic regulation of obesity, in addition to the identification of potential targets mediating the homeostatic circuits, it is also important to explore the hedonic circuits mediating energy balance, identifying the salient neural, hormonal, and humoral components involved, as remarkably reviewed by Gastelum et al. [7]. They explore the sexual dimorphism on these circuits, paying special attention to the role of two emerging neuropeptides, the nociception/orphanin FQ and the pituitary adenylate cyclase-activating polypeptide (PACAP) in the neuronal activity regulation in positive and negative energy states. Jang et al. [6] also investigate leptin-mediated feeding circuits in hypothalamic neurons, identifying the expression of the angiopoietin-like growth factor (AGF) in POMC neurons of the arcuate nucleus as a downstream factor involved in leptin signaling in the hypothalamus.

involved in leptin signaling in the hypothalamus. Hypothalamic circuits controlling energy balance in response to feeding are also mediated by the gut-to-brain nutrient signaling, whose disruption leads to obesity. As extensively reported by Romani-Perez et al. [13], aberrant feeding patterns or unhealthy diets might alter gut microbiota–diet interactions and modify nutrient-sensing information from

the gut to the hypothalamus, impairing energy homeostasis. They identify microbiomebased strategies to improve the gut-brain axis function and hence combat obesity.

An already well-established concept in the neurobiology of obesity is that neurons are not the only characters playing a main role in the response processing and the transmission of information. This is also true in the hypothalamus, where a maladaptive interaction between neurons and their surrounding microglia and astrocytes likely contribute to the development and progression of obesity [9,12]. The perspective elegantly presented by Leon et al. [12] discusses recent evidence that sheds light on the cellular and molecular mechanisms underlying microglia–neuron communication in hypothalamic circuits that are crucial for body weight and food-intake control in response to HFD exposure. Interestingly, microglia, as key players in the immunological response, are sensitive to fuel substrate and can influence neuronal activity in part via cytokines. However, the maladaptive microglia–neuron cross-talk observed in response to overnutrition seems to involve synergy with astrocytes, which are key components of the tripartite synapse, as discussed below. In addition to these cellular interactions, Mendes et al. [10] examine the subtypes of microglia that may be involved in HFD-induced hypothalamic inflammation in the early stages of the disease. They also remarkably discuss which models can be useful for obtaining the most reliable data when exploring distinct subsets of microglia, and clarify the changes in hypothalamic microglia signature in response to HFD-induced obesity based on novel transcriptomic analysis. The importance of microglia activation in the hypothalamus in energy balance has also been reported by Lopez-Gambero et al. [11]. They found that glial activation and metabolic dysfunction in the hypothalamus of a mouse model of Alzheimer's disease (5XFAD mouse) leads a negative energy balance and lower insulin and related hormones in plasma, contributing to the development of the pathology. Interestingly, this phenotype was more marked in female than male mice, indicating useful information for early detection of the disease and sexual dimorphism in the pathophysiology of Alzheimer's disease.

The function of hypothalamic astrocytes and the interplay between astrocytes and neurons are also becoming very important in the study of obesity development. In this special issue, Song et al. [8] show the ability of the adipose tissue-derived hormone adiponectin to control nutrient metabolism in hypothalamic astrocytes by enhancing glucose uptake, glycolytic activity, and lactate and ketone body production. In this study, central administration of adiponectin increases the number of astrocytes in the hypothalamus of mice. The most relevant findings regarding molecular mechanisms by which hypothalamic astrocytes are involved in the pathogenesis of obesity have been well reviewed by Gonzalez-Garcia et al. [9]. They expose the idea of how a disruption in the astrocyte–neuron communication within the hypothalamus is altered by exposure to obesogenic insults and the impact this has on obesity development. In addition, they point out the emerging idea that astrocytes are heterogeneous and that there might exist subpopulations that contribute to specific actions. Therefore, the identification of specific hypothalamic astrocyte subpopulations might be critical for the pathophysiology of obesity [9]. The study of the potential role of other brain cells and how they are interconnected with glia and neuronal cells in the hypothalamic regulation of obesity is also promising (e.g., tanycytes and endothelial cells) [12].

Therefore, the manuscripts published in this Special Issue provide new insights into the hypothalamic neuronal and non-neuronal pathways involved in the pathophysiology and development of obesity and related diseases such as diabetes (Figure 1). Via four original papers, six review manuscripts, and one perspective, this Special Issue will substantially contribute to the identification of novel therapeutic targets in the hypothalamus and novel approaches to combatting the progression of human obesity.

**Author Contributions:** All authors (R.R.-R. and C.M.) have participated in writing and editing. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Spanish Ministerio de Economía, Industria y Competitividad (MINECO), Agencia Estatal de Investigación (AEI), and Fondo Europeo de Desarrollo Regional (FEDER) (Grants PCI2018-092997/AEI to R.R.-R.; PID2020-114953RB-C22 to R.R.-R. by MCIN/AEI/10.13039/501100011033).

**Institutional Review Board Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Perspective* **Microglia–Neuron Crosstalk in Obesity: Melodious Interaction or Kiss of Death?**

**Stéphane Léon , Agnès Nadjar and Carmelo Quarta \***

Neurocentre Magendie, Physiopathologie de la Plasticité Neuronale, INSERM U1215, University of Bordeaux, 33000 Bordeaux, France; stephane.leon@inserm.fr (S.L.); agnes.nadjar@inserm.fr (A.N.)

**\*** Correspondence: carmelo.quarta@inserm.fr

**Abstract:** Diet-induced obesity can originate from the dysregulated activity of hypothalamic neuronal circuits, which are critical for the regulation of body weight and food intake. The exact mechanisms underlying such neuronal defects are not yet fully understood, but a maladaptive cross-talk between neurons and surrounding microglial is likely to be a contributing factor. Functional and anatomical connections between microglia and hypothalamic neuronal cells are at the core of how the brain orchestrates changes in the body's metabolic needs. However, such a melodious interaction may become maladaptive in response to prolonged diet-induced metabolic stress, thereby causing overfeeding, body weight gain, and systemic metabolic perturbations. From this perspective, we critically discuss emerging molecular and cellular underpinnings of microglia–neuron communication in the hypothalamic neuronal circuits implicated in energy balance regulation. We explore whether changes in this intercellular dialogue induced by metabolic stress may serve as a protective neuronal mechanism or contribute to disease establishment and progression. Our analysis provides a framework for future mechanistic studies that will facilitate progress into both the etiology and treatments of metabolic disorders.

**Keywords:** microglia–neuron communication; hypothalamus; obesity; metabolism

## **1. Introduction**

Over the last 25 years, since the discovery of the hormone leptin [1], enormous advancements have been made in understanding how the brain directs changes in feeding behaviors and systemic metabolic pathways to maintain the body's metabolic needs [2]. This advancement has initiated a consensus on a "brain-centric" view, which now considers that obesity does not only result from the dysfunctional activity of peripheral organs, such as the adipose tissue, but can also originate from the brain [3].

Large-scale genomic studies have demonstrated that body weight excess and the associated systemic metabolic defects can derive from spontaneous mutations that involve genes that are expressed in the central nervous system (CNS) [3]. Moreover, multiple lines of evidence have now demonstrated that hypercaloric diet feeding can trigger obesity and multiple metabolic comorbidities via central mechanisms of action [4–7].

Most of these mechanisms involve changes in the activity of neuronal networks located in the hypothalamus that influence fuel (food) ingestion, dissipation, or storage [8–10]. Brain-to-periphery axes controlled by this brain region allow maintenance of constant levels of energy reserves, or the so-called condition of energy balance. The melanocortin circuit, which is formed by different neuronal populations that respectively express the peptidergic precursor pro-opio-melanocortin (POMC) or the neuropeptide agouti-related protein (AgRP), plays a crucial role. When the synaptic plasticity of this circuit is impaired for instance, in response to prolonged feeding with hypercaloric diets—this neurobiological defect can lead to energy balance dysregulation [5,6], and, therefore, to the establishment or the progression of diet-induced obesity (DIO) [7,11].

**Citation:** Léon, S.; Nadjar, A.; Quarta, C. Microglia–Neuron Crosstalk in Obesity: Melodious Interaction or Kiss of Death? *Int. J. Mol. Sci.* **2021**, *22*, 5243. https://doi.org/10.3390/ ijms22105243

Academic Editors: Rosalía Rodríguez-Rodríguez and Cristina Miralpeix

Received: 22 April 2021 Accepted: 13 May 2021 Published: 15 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

The pivotal role played by the hypothalamus in the etiology of obesity and its associated sequelae may not seem surprising given that this brain area has long been known to influence food intake and energy handling. More surprisingly, however, non-neuronal cells may be the key piece of the puzzle, at least based on more recent information.

Astrocytes have classically been known to participate in the so-called "tripartite synapse", e.g., the integrated functional unit whereby these glial cells communicate with two or more neurons to accommodate their synaptic transmission [12]. Within this cell-tocell process of communication, however, microglial cells are emerging as key players given their influence on the formation, function, plasticity, and elimination of synapses [13,14].

During overfeeding, microglial cells accumulate in the hypothalamus [15] ("microgliosis"), undergo morphological activation, and augment the production of proinflammatory cytokines and other neuronally secreted factors that potentially interfere with synaptic transmission [16]. These changes seem to occur specifically in the hypothalamus, but not in other brain areas [16], suggesting that a "healthy" microglial–neuronal crosstalk may be crucial for hypothalamus-based regulation of energy balance, whereas a defective process of intercellular communication may promote neuronal dysfunction, energy imbalance, and ultimately obesity.

Our understanding of the role of microglia–neuron interactions in modulating neuronal function in extrahypothalamic brain areas has significantly improved over the last decade [17]. However, the molecular and cellular underpinnings of such intercellular dialogue at the level of specific hypothalamic circuits involved in body weight control have only begun to emerge.

Here, we will discuss recent evidence that sheds new light on the cellular and molecular mechanisms implicated in microglia–neuron communication in hypothalamic circuits that are relevant for body weight and food intake regulation. We will interrogate whether a defective dialogue between hypothalamic microglia and neurons may be causally linked with the etiology of DIO, while also highlighting the remaining outstanding questions on how this intercellular cross-talk operates to maintain energy homeostasis in the whole body.

#### **2. The Emerging Neuroimmune Theory of Obesity**

Bidirectional interactions between the immune system and the CNS enable immunological, physiological, and behavioral responses. Upon activation by infection or injury, for instance, peripheral cells of the innate immune system synthesize and release cytokines that serve as immune mediators. These peripheral signals travel to the brain and act through complex mechanisms of microglia–neuron communication, ultimately leading to non-specific symptoms of infection, including lethargy or listlessness, amongst others, which are commonly referred to as the "sickness response" [18].

In animal models of DIO, enhanced sickness behavior is observed after a single administration of the bacterial lipopolysaccharides (LPSs) [19]. This phenotype is associated with increased production of proinflammatory mediators in the hypothalamus, including tumor necrosis factor alpha (TNFα), interleukin 1 beta (IL-1β), interleukin 6 (IL-6), the inhibitor of nuclear factor kappa alpha (IκBα), and cyclooxygenase-2 (COX2) [20], thereby suggesting that whole-body neuroimmune axes coordinated by CNS microglia undergo defective regulation in obesity.

In addition to responding to injury or infections, CNS microglia can be also activated by dietary or hormonal factors that influence whole-body energy handling, including, for instance, lipids [21,22], carbohydrates [23], or the hormone leptin [20,24]. This suggests that microglial cell activity may play a critical role in maintaining systemic energy homeostasis. After long-term feeding with hypercaloric diets, an increased number of activated microglial cells has been consistently observed across different studies at the level of the mediobasal hypothalamus, where POMC and AgRP neurons reside [4,25–31]. These microglial changes are accompanied by the upregulation of cytokine-signaling-related pathways in hypothalamic neurons [4,25–31], and are not a rodent-specific phenomenon. Indeed, human obesity is also associated with radiological signs of gliosis [4,32,33]. Moreover, histological analyses of post-mortem brain tissues obtained from obese individuals has revealed the presence of microglial cells with aberrant morphologies at the level of hypothalamic areas [34]. Accordingly, body mass index (BMI) positively correlates with microglial soma size in subjects affected by type-2 diabetes (T2D), which is one of the major glucometabolic complications of obesity [35].

Based on these accumulated findings, one emerging theory in the field is that the exacerbated microglial activity induced by overfeeding may sustain and propagate a condition of low-grade chronic inflammation at both the central (hypothalamus) and systemic level, which contributes to the establishment and the progression of DIO and its associated glucometabolic complication. Several observations support this model: (i) inhibiting expansion of CNS microglia expansion in DIO mice hindered diet-induced weight gain and prevented central and peripheral inflammatory responses induced by overfeeding [15]; (ii) pharmacologically depleting microglia or selectively restraining microglial inflammatory signaling pathways in the hypothalamus sharply reduced hypothalamic microgliosis while also limiting diet-induced hyperphagia and weight gain in a murine model of DIO [36]; (iii) several molecules and pathways have been identified as candidate mediators of hypothalamic inflammatory reactions observed during chronic feeding with a hypercaloric diet (for a detailed review on this topic, see [37]). The targeted disruption of these pathways in a hypothalamus-specific manner, by pharmacologic or genetic means, limits the extent of DIO and its main systemic metabolic defects, e.g., impaired systemic glucose homeostasis [26–30,38].

Of note, emerging microglia-directed neuroimmune responses induced by diet and/or obesity share elements of similarity with other "classic" neuroimmune axes. For instance, a feature of the sickness behavior is the enhanced entry of peripheral immune cells into the CNS [18]. Likewise, during high-fat diet (HFD) consumption, changes in the peripheral and central immune systems are intimately interconnected.

In DIO mice, blood-borne monocyte-derived cells can extravasate the vasculature and enter the brain at the level of the hypothalamic arcuate nucleus (ARC) of the hypothalamus [36], which contains a leaky blood–brain barrier [39]. As they accumulate, these infiltrating cells resume a typical microglia morphology and contribute to propagating the diet-induced inflammatory response [36].

Additionally, diet-induced hypothalamic inflammation, and, similarly, other microglial neuroimmune responses, can have both positive and negative downstream neuronal effects. After brain injury or pathogens invasion, an initial microglial activation is typically observed and is necessary for neuroprotection, but increased or prolonged microglial activity can have neurotoxic effects [40]. Likewise, short-term microglial responses to diet-induced metabolic stress may initially be protective for neurons that are in strict communication with these glial cells, whereas in the long term, the communication process may become maladaptive, thus negatively influencing neurotransmission [41]. Supporting this view, ARC microglial activity and the resulting local increase in proinflammatory cytokine levels observed during DIO progression follow a dynamic pattern, as they occur rapidly during the first hours [42] and days [4] of overfeeding, are normalized for a few weeks during the dietary administration, and rise again thereafter [4]. Moreover, the establishment of a "low-grade" chronic microglial activation state in response to prolonged feeding with hypercaloric diets is often associated with hallmarks of hypothalamic neuronal dysfunction in animal models after long-term feeding of hypercaloric diets, including loss of synapses [5], impaired responsiveness to metabolic hormones [7,43,44], altered intracellular organelles function [4,29,45,46], and possibly cell death [47,48], although the latter was not observed in all of the studies [49].

Thus, an intimate and dynamic relationship links exacerbated hypothalamic microglial activity with hypothalamic neuronal dysfunction and DIO progression. Such a neuroimmune basis of obesity may either originate from or lead to a defective process of communication between neurons and surrounding microglial cells.

#### **3. Role of Microglia in the Remodeling of Neuronal Networks in Obesity**

Microglia can phagocytize neuronal spines, terminals, and cell bodies, and errors in this process can lead to impaired neuronal activity and possibly to cell death [13,50]. In response to pathogen invasion or neuronal damage, phagocytosis of microorganisms, dying cells, or neuronal debris from microglia occurs during the resolution phase of inflammation and involves the release of cytokines and lipid mediators that exert anti-inflammatory and prorepair properties.

Microglia can also remove synaptic elements from neuronal cell bodies, a process that allows the protection of neurons from excitotoxicity, e.g., the neuronal death caused by the overactivation of excitatory amino acid receptors [13]. Early in the course of a disease's process, it might be useful to strip away dying synapses so that healthy synaptic elements can take over. However, as the insult to the CNS persists, abnormal synapse removal can translate into neurodegenerative cellular states [13].

Consumption of dietary fat has been proposed to induce apoptosis of neurons in the arcuate nucleus, paraventricular nucleus, and lateral hypothalamus in mice [47,48], and such a neuronal loss is often paralleled by microgliosis [47,48], which makes one wonder if diet-induced neuronal death and exacerbated microglial activity in the hypothalamus are the two faces of the same phenomenon. However, the observation that DIO is associated with hypothalamic neuronal death has not been reproduced by all the studies [49]. This inconsistency may be due to differences in the diet-composition, caloric content, length of diet exposure, and technical approach used to assess neuronal counts across these studies [7]. Hence, only addressing whether and how these different contributing factors impact microglial phagocytic capacity and, eventually, hypothalamic neuronal death may provide a final answer.

Irrespective of this dispute, changes in microglial activity induced by overfeeding may have more "subtle" consequences on downstream neurons—for instance, on their synaptic transmission. LPS-mediated activation of microglial cells evokes excitatory responses in hypothalamic POMC neurons and inhibitory responses in AgRP neurons [51]. HFDinduced obesity is associated with the synaptic reorganization of hypothalamic POMC and AgRP neurons [5], which present fewer total synapses on their perikaryal [5]. These synaptic changes are intimately associated with the presence of increased hypothalamic microglial activity [5], which suggests that microglial-mediated effects may be directly implicated, although no direct causal evidence supporting this model is yet available.

Changes in microglial-mediated phagocytosis of synaptic elements (or microglial "pruning") seem to occur in response to overnutrition in extrahypothalamic areas—for instance, in the hippocampus. In a DIO animal model, low-grade chronic inflammation induced by a hypercaloric diet was associated with reduced cognitive function and with increased microglial pruning [52,53]. Memory impairment observed in these animals occurred along with reductions in hippocampal dendritic spines [53], and both the memory defects and the loss of neuronal spines could be rescued through pharmacological manipulation of microglial phagocytosis [53].

Whether or not microglia-mediated pruning contributes to the specific hypothalamic neuronal dysfunction observed during DIO has not yet been directly tested, but this hypothesis is supported by a series of indirect observations. For instance, in an animal model of obesity and type 2 diabetes (db/db mice), which was characterized by impaired hypothalamic neuronal synaptic plasticity [54,55], hypothalamic microglial cells presented a reduced expression of the phagocytic marker cluster of differentiation 68 (CD68) [56]. In this model, the deficit in hypothalamic neuronal function and the resulting obesity phenotype derived from a defective action of the metabolic hormone leptin.

Likewise, transgenic mice with a specific leptin receptor deficiency in myeloid cells (which include microglia) presented an obesity-like phenotype, a reduced number of hypothalamic ARC POMC-expressing neurons, and signs of impaired microglial phagocytic capacity in the hypothalamus [24]. Thus, microglia-mediated synaptic remodeling of hypothalamic neuronal cells may be under the influence of hormonal signals, such as leptin. This implies that alterations in these microglial–neuroendocrine axes may contribute to DIO pathophysiology. Of note, leptin can potentiate the microglial response to LPS in an in vitro system, and this is linked to morphological changes that render the microglia more reactive [57], further suggesting that this hormone may have an important role in mediating microglial functions.

Notably, in addition to influencing synapse homeostasis through their phagocytic activity, microglial cells can also promote synapse formation [13], although no studies have thus far investigated whether changes in microglial-mediated synaptogenesis occur during diet-induced metabolic stress, which represents an interesting angle for future investigations.

Thus, hypercaloric diets may trigger microglial-mediated synaptic remodeling, which possibly disrupts the activity of AgRP/NPY and POMC neurons, thereby contributing to disease development. However, the defective synaptic plasticity observed in the ARC of DIO animal models likely involves synergy among not only neurons and microglia but also astrocytes [5], which are key components of the tripartite synapse. Thus, studying celltype-specific mechanisms that link diet-induced inflammation with the impaired synaptic plasticity of ARC neurons is an interesting avenue for future research.

#### **4. Impact of Fuel Substrates on Microglia–Neuron Communication in Obesity**

Microglia are highly dynamic and continuously move around the brain parenchyma in response to chemotactic signals. Under activated states, these glial cells not only display increased phagocytic activity, but also augment the production of proinflammatory and neuromodulatory factors, which are all processes that require large amounts of energy. To meet these continuous and rapid changes in energy demands, microglia express transporters for glucose, fatty acids, and amino acids, the three main fuel substrates. Moreover, these glial cells are also capable of metabolizing alternative nutrients, such as glutamine [58].

The exact identities of the main extracellular influencing factors leading to reactive microglial states during overfeeding are under intense investigation and have not yet been clarified. However, saturated fat (SF) may be a key piece of the puzzle. In animals fed with a saturated fat (SF)-enriched diet, fatty acids, including palmitic acid, are conveyed to the brain after ingestion. They are then taken up by hypothalamic microglia, and ultimately contribute to inducing the hypothalamic inflammatory reaction [22].

Feeding mice with an SF-rich diet augments the expression of the heat-shock protein 72 (Hsp72), which is a chaperone protein that is implicated in cellular stress, in the mediobasal hypothalamus [22]. Such a molecular stress response is independent of the increased caloric intake associated with the SF-rich diet. Indeed, increased neuronal Hsp72 levels were observed by administering milk fat to control mice fed with a standard chow diet, but not following the administration of isocaloric olive oil [22]. Thus, increased SF intake and not fat intake per se—accounts for the presence of signs of hypothalamic neuronal stress. Of note, depleting microglia via a genetic approach nearly abolishes the marked induction of neuronal Hsp72 that is otherwise seen in ARC neurons following SF-rich administration [22]. This suggests that changes in microglial SF sensing causally contribute to the neuronal dysfunction observed in DIO.

Saturated fat is likely not the only fuel substrate that mediates diet-induced microglial activation and the resulting negative downstream neuronal effects. Indeed, long-term feeding with a high-carbohydrate/high-fat diet (HCHF) also leads to increased hypothalamic microglial activity in mice, which is associated with the elevated production of advanced glycation end-products (AGEs) from hypothalamic POMC and AgRP neurons. AGEs are products of glycoxidation and lipoxidation reactions that originate from overexposure to sugars. These bioproducts are implicated in many degenerative conditions, including diabetes and Alzheimer's disease [59]. Certain types of AGEs—for instance, the product N(ε)-(carboxymethyl)-lysine (CML)—are mainly produced by neurons during diet-induced cellular stress, whereas their receptors are specifically localized in non-neuronal cells, including, but not limited to microglia [23]. Genetic deletion of CML receptors in mice fed with an HCHF diet resulted in less microglial reactivity in the hypothalamic ARC and also

led to favorable antiobesity effects [23]. Thus, a putative model can be proposed whereby HCHF diets—and the consequent combined overload of both lipids and glucose in the mediobasal hypothalamus—trigger excessive production of neuronal metabolites (such as AGEs) that activate surrounding microglial cells. This results in a vicious feedback loop that impairs neuronal functions.

However, what are the main intracellular underpinnings of such a maladaptive crosstalk? A final answer is not yet available, but common intracellular nodes may integrate changes in fuel substrate handling between neurons and microglia. In both microglial and neuronal cells, maladaptive changes in mitochondrial function and cellular fuel utilization were observed after hypercaloric diet feeding, and in both cell types, these changes were, in part, mediated by the mitochondrial protein uncoupling protein 2 (UCP2) [60]. Of note, increased UCP2-dependent activity in either microglial or neuronal cells played a key role in mediating DIO susceptibility in murine models [61–63].

Thus, sensing of specific dietary components, such as fat and/or sugar, may lead to exacerbated cellular fuel metabolism in both microglial and neuronal cells located in the hypothalamus via common (UCP2-mediated) molecular pathways that are essential for proper mitochondrial activity. This implies that changes in fuel substrate uptake and/or metabolism induced by diet-induced caloric overload may directly impact the microglia– neuron crosstalk via mitochondria-mediated mechanisms and the subsequent production of yet-to-be identified energy metabolites that contribute to the intercellular communication.

Supporting this view, specific microglial genetic deletion of lipoprotein lipase (LPL), one of the key enzymes involved in cellular fuel uptake, caused mitochondrial dysmorphologies in both microglia and nearby ARC neurons in animals that were fed with a hypercaloric diet [64]. These transgenic mice that specifically lacked LPL activity in microglia also presented higher body weight gain relative to wild-type animals when fed with an obesogenic diet [64].

#### **5. Role of Cytokines in Microglia–Neuron Communication in Obesity**

Activated hypothalamic microglia produce a variety of proinflammatory cytokines that potentially influence metabolic or behavioral outputs by inducing changes in neuronal activity [65–67]. In vitro incubation of hypothalamic POMC neurons with the microglialproduced cytokine TNF-α led to an increased neuronal firing rate, increased intracellular ATP production, and mitochondrial fusion processes [68]. HCHF diet consumption is associated with elevated hypothalamic TNF-α levels [69] and with signs of POMC neuronal mitochondrial stress [68]. Hence, persistently elevated TNF-α signaling during DIO may cause functional impairment of hypothalamic POMC neurons as a result of alterations in neuronal mitochondrial activity [68]. This hypothesis is supported by multiple studies that showed that DIO leads to mitochondrial dysfunction in POMC neurons and that this specific pathogenic mechanism may promote dysregulated neuronal activity and obesity [7]. The molecular underpinnings of TNF-α-mediated neuronal dysfunction in obesity have yet to be fully elucidated, but cytokine-mediated changes in the c-Jun Nterminal kinase (JNK)/activator protein 1(AP-1) and NF-κB signal transduction pathways may have a central role. A detailed analysis of the neuronal molecular pathways implicated in hypothalamic inflammation is beyond the main scope of the present work, but more detailed information can be found elsewhere—for instance, in [37].

Although it is tempting to speculate that proinflammatory cytokines produced by hypothalamic microglial cells in response to overfeeding may globally contribute to the induction of neuronal dysfunction and obesity, this hypothesis appears to be too simplistic given that certain microglia-derived cytokines can have positive effects on metabolism and body weight regulation. For instance, activated hypothalamic microglia can produce the proinflammatory cytokines IL6 and IL-1β [70], which can have favorable effects on energy balance regulation. Administration or overexpression of these cytokines in the brain results in reduced food intake, increased energy expenditure, and decreased body weight gain [71–75]. This is likely due to changes in the activity of neurocircuitries located in the

hypothalamus [76,77]. Accordingly, the central application of IL6 led to potent feeding suppression and to improvements in glucose tolerance in DIO mice via the modulation of neurons located in the paraventricular nucleus of the hypothalamus [78]. Thus, different proinflammatory cytokines can have either positive or negative effects on body weight control, which makes one wonder whether diet-induced hypothalamic inflammation is necessarily a pathophysiological mechanism that contributes to disease progression.

The fact that microglia-derived proinflammatory cytokines may induce dynamic downstream neuronal effects is probably part of the answer. For instance, among families of cytokines that orchestrate hypothalamic inflammatory reactions induced by overfeeding, chemokines act by attracting immune cells to the inflamed site [79]. When proinflammatory cytokines activity is enhanced in the brain of an animal, this generally leads to a condition of initial high-grade inflammation and to a transitory reduction in food intake, which is partly due to changes in the activity of hypothalamic neurons [80,81]. Nevertheless, such acute and robust inflammatory responses and the linked food-intake-reducing effect generally tends to diminish or, in some cases, to fade away a few hours after administration [80,81]. The metabolic or behavioral effects observed after short-term hypothalamic cytokines action during hypercaloric diets feeding may reflect a defense mechanism that limits the introduction of unhealthy diets. Nevertheless, this mechanism of "neuroprotection" may become maladaptive in the long term and have the opposite effects, thereby contributing to disease establishment and progression. Accordingly, certain chemokines, such as the C-X3- C motif chemokine ligand 1 (CX3CL1, or 'fractalkine'), promote obesity when chronically administered in the brain [82], whereas inhibition of CX3CL1 actions in the hypothalamus of DIO mice through an approach of small interfering RNA reduced local inflammation and led to weight loss and to improvements in systemic glucose metabolism [82]. Moreover, the expression of the C-X-C motif chemokine 12 (CXCL12) and its receptors was induced in the hypothalamus of mice fed with an HFD [83]. Such increased CXCL12 activity has been suggested to favor overfeeding by enhancing the synaptic activity of specific neurons located in this hypothalamic area that produce the hunger-promoting neuropeptide enkephalin [83].

From a broader perspective, cytokine-mediated changes in microglia–neuronal communication likely also involve other non-neuronal cells. As previously discussed, DIO leads to hypothalamic astrocyte activation, and following activation, these non-neuronal cells produce a variety of immune mediators, including TNFα, IL-1β, IL-6, or the monocyte chemoattractant protein-1 (MCP-1) [39,84], amongst others.

When cultured in vitro, astrocytes accumulate lipid droplets under a free fatty acid (FFA)-rich environment, while also producing a large amount of proinflammatory cytokines [84]. If cultured microglial cells are treated with a conditioned medium obtained from lipid-laden astrocytes, this leads to a markedly enhanced chemotactic activity of the microglial cells, which is due to MCP-1-mediated actions at the level of the microglial C-C chemokine receptor type 2 (CCR2) [84]. Thus, astrocytes-derived inflammatory signals may have an impact on neighboring CNS microglia, thereby alimenting the hypothalamic inflammatory response. Given that the chemokine MCP-1 disrupts the integrity of the blood–brain barrier (BBB) [85] and that the increased brain infiltration of bonemarrow-derived monocytes/macrophages contributes to the induction of hypothalamic inflammation [36] (see above), this vicious cycle may be also influenced by peripheral immune cells.

In conclusion, a multicellular model can be proposed whereby, in response to hypercaloric diets, astrocytes, local microglial cells, and peripheral monocytes/macrophages co-direct changes in the local production of cytokines and chemokines, which ultimately impact hypothalamic neuronal activity and energy homeostasis in a highly dynamic manner during the progression of DIO.

#### **6. "ON/OFF" Communication between Microglia and Neurons in Obesity**

While long-term changes in the activity of CNS microglia can impair neuronal functions, the opposite relationship can be also relevant, as discussed herein. However, can hypothalamic neurons be initiators of vicious interactions with surrounding non-neuronal cells in response to diet-induced metabolic stress? If one considers that neuronal cells in mammals can release ON signals that promote microglial recruitment and activity [82,86], this hypothesis seems plausible.

For instance, the energy substrates adenosine 50 - triphosphate (ATP) and adenosine 5 0 -diphosphate (ADP) can be released from the neuronal cell body via membrane channels or activity-dependent vesicle exocytosis [87,88]. Microglia contain specific purinergic receptors (such as P2Y12) that are expressed at the level of structural interaction sites located between neuronal cell bodies and microglial processes [88]. P2Y12-containing purinergic junctions allow microglia to constantly monitor the neuronal status. During traumatic neuronal events (e.g., brain stroke) these junctions initiate microglia-driven neuroprotective actions in a targeted manner on the neuronal cell body [88]. Moreover, purinergic receptors are implicated in controlling microglial motility and attraction to synapses [89].

Similar mechanisms of spatially controlled neuronal surveillance and protection may take place in hypothalamic neurons in response to diet-induced neuronal stress. DIO leads to POMC neuronal mitochondrial dysfunctions [7,68] and to a higher number of activated microglial cells that are in close contact with POMC neurons [68], e.g., potential sites of P2Y12-mediated communication. However, no studies have, thus far, directly addressed whether hypothalamic neuronal cells display a plastic ability to release purinergic signals that influence microglia recruitment and functions, which represent an interesting area for future investigation.

Of note, microglial cells in the healthy CNS are immunologically more quiescent than peripheral macrophages [90], which raises the question of whether brain neurons may release OFF signals that keep these glial cells in a non-active state. In hippocampal brain slices, the induction of the major histocompatibility complex class II (MHC-II) in microglia, which is indicative of increased microglial immune response [91], is restricted by neurotrophins released from electrically active neurons, for instance, by the brain-derived neurotrophic factor (BDNF [92]).

BDNF-producing hypothalamic neurons play a major role in energy balance regulation in response to the hormone leptin [93]. Accordingly, mutations in the BDNF gene lead to insatiable appetite and severe obesity [94]. Hence, BDNF might be a neuronal OFF signal that maintains a homeostatic process of neuron–microglia communication in the hypothalamus, while defective BDNF production may ultimately lead to exacerbated microglial activity and obesity.

In addition to neurons, microglia can also release BDNF, for instance, in response to the activation of ATP-sensitive purinergic receptors [95]. Moreover, BDNF-producing hematopoietic cells can migrate from circulation to the hypothalamus; they can make direct contact with neurons in response to the feeding status, thereby influencing whole-body metabolic and behavioral responses [96]. Mice with congenital BDNF deficiency specifically in these hematopoietic cells developed hyperphagia, obesity, and insulin resistance [96], suggesting that defective BDNF production from either neuronal and non-neuronal cells located in the hypothalamus may initiate a vicious cycle leading to defective neuronal functions and obesity.

Neurotransmitters may represent an additional layer of reciprocal ON/OFF regulation between microglia and neurons. Microglia can express receptors for the neurotransmitters glutamate, gamma-aminobutyric acid (GABA), noradrenaline, or dopamine (amongst others) in both in vitro and ex vivo systems [97], although evidence for microglial expression of these markers in vivo is still lacking. In vitro treatment of microglial cells with most of these neurotransmitters inhibits LPS-induced release of various proinflammatory factors from these glial cells [86]. Nevertheless, these neurotransmitter-mediated effects on

microglia activity are unlikely to be only directly mediated. Instead, these influences could be indirectly influenced by extracellular ATP, which is released in response to changes in neurotransmission [98].

While several neurotransmitters may act as an OFF-signals that keep microglia in a quiescent state, one exception may be the excitatory neurotransmitter glutamate. Glutamate can trigger activation and release of the proinflammatory cytokine TNF-α from microglial cells [99,100]. Faced with proinflammatory stimuli, such as TNF-α, microglia can, in turn, release glutamate [101], and thus probably enter a vicious cycle that leads to the increased generation of both glutamate and TNF-α. The reciprocal control of glutamate production by microglial and neuronal cells may be part of a feedback loop that protects neurons from excitotoxicity. Alternatively, neurotransmitter-mediated cross-talks between neurons and microglia may serve as a defense mechanism that limits exacerbated cytokine action or excessive microglial phagocytic activity [86]. Whether or not DIO leads to defective neurotransmitters-mediated microglia–neurons communication has yet to be directly addressed. Given that animal models of DIO display impaired hypothalamic expression of glutamate transporters [102], understanding how and whether changes in the balance between both ON and OFF signals orchestrate the (dys) functional dialogue between microglia and neurons in obesity warrants further investigation.

#### **7. Conclusions and Perspectives**

A better understanding of the neurobiological basis of obesity is urgently needed given that this disease may have an origin in the brain [3] and that currently available antiobesity strategies have not yet achieved a transformative therapeutic impact [103].

New pharmacological tools that selectively and safely target hypothalamic neuronal circuits affected by obesity are on the horizon [38,104]. However, the cellular and molecular mechanisms leading to hypothalamic neurons' dysfunction in obesity have only begun to emerge, and additional research will be needed before functional optimization and clinical translation of these brain-directed agents.

These future efforts toward understanding the neurobiological basis of obesity should consider that hypothalamic neuronal cells are not isolated functional entities, and maladaptive reciprocal interactions between neurons and surrounding microglia cells likely contribute to the establishment and the progression of the disease (Figure 1).

Several outstanding questions related to the extracellular and intracellular mode(s) of action of this process of cell–cell communication remain. For instance, what are the main extrinsic factors that transform a melodious microglia–neuron interaction into a maladaptive process? The caloric content of the diet, diet composition, and duration of diet exposure, may influence hypothalamic neuroinflammatory responses through multiple mechanisms, including changes in gut microbiota composition [32]. Moreover, the overload of certain dietary components, such as fats or carbohydrates, may engage specific cellular metabolic pathways that interfere with mitochondrial function and with the extracellular release of chemical messengers (such as ATP or ADP) influencing the inter-cellular communication.

Of note, both neurons and microglia in the brain may secrete extracellular vesicles (EVs), which are known to contain a broad range of signals possibly involved in the maladaptive microglia–neuron communication, including chemokines, ATP, and inflammationassociated small, noncoding RNA (miRNAs) [105]. Whether or not mitochondrial or EV-mediated mechanisms of communication may contribute to diet-induced hypothalamic neuronal alterations, however, remains to be directly explored.

**Figure 1.** Under physiological conditions (**left**) microglia play a key role in fine-tuning neuronal synaptic function in response to changes in the energy status of the body. This involves different mechanisms including the release of ON-OFF signals, the intercellular production of metabolic messengers in response to the cellular metabolism of fuel substrates, and physical intercellular contacts that are necessary for neurosurveillance. This melodious interaction may become maladaptive during diet-induced obesity (**right**), thereby leading to impaired synaptic plasticity through multiple possible mechanisms, which are highlighted in the figure. The maladaptive cross-talk ultimately results in energy balance dysregulation, obesity, and systemic glucose intolerance. FA: fatty acids; AGEs: advanced glycation end-products; UCP2: uncoupling protein 2 ; CX3CL1: C-X3-C motif chemokine ligand 1; ATP: adenosine 5′- triphosphate; TNFα: tumor necrosis factor-alpha ; IL-1β: interleukin 1 beta, IL6: interleukin 6. **Figure 1.** Under physiological conditions (**left**) microglia play a key role in fine-tuning neuronal synaptic function in response to changes in the energy status of the body. This involves different mechanisms including the release of ON-OFF signals, the intercellular production of metabolic messengers in response to the cellular metabolism of fuel substrates, and physical intercellular contacts that are necessary for neurosurveillance. This melodious interaction may become maladaptive during diet-induced obesity (**right**), thereby leading to impaired synaptic plasticity through multiple possible mechanisms, which are highlighted in the figure. The maladaptive cross-talk ultimately results in energy balance dysregulation, obesity, and systemic glucose intolerance. FA: fatty acids; AGEs: advanced glycation end-products; UCP2: uncoupling protein 2; CX3CL1: C-X3-C motif chemokine ligand 1; ATP: adenosine 50 - triphosphate; TNFα: tumor necrosis factor-alpha; IL-1β: interleukin 1 beta, IL6: interleukin 6.

Several outstanding questions related to the extracellular and intracellular mode(s) of action of this process of cell–cell communication remain. For instance, what are the main extrinsic factors that transform a melodious microglia–neuron interaction into a maladaptive process? The caloric content of the diet, diet composition, and duration of diet exposure, may influence hypothalamic neuroinflammatory responses through multiple mechanisms, including changes in gut microbiota composition [32]. Moreover, the overload of certain dietary components, such as fats or carbohydrates, may engage specific cellular metabolic pathways that interfere with mitochondrial function and with the ex-In addition to being influenced by extrinsic factors, the maladaptive cross-talk between microglia and neurons observed in obesity may then originate from intrinsic variables, including genetic predisposition. Indeed, human subjects carrying common polymorphisms in the Jun N-terminal kinase (JNK) or the melanocortin 4 receptor (MC4R) gene are more susceptible to development of obesity and hypothalamic neuroinflammation [32]. Furthermore, gender may play a key role given that microglia isolated from adult brains present sex-specific molecular features [106–108] and that sex differences in microglial activity determine obesity susceptibility in mice [109,110].

tracellular release of chemical messengers (such as ATP or ADP) influencing the intercellular communication. Of note, both neurons and microglia in the brain may secrete extracellular vesicles (EVs), which are known to contain a broad range of signals possibly involved in the maladaptive microglia–neuron communication, including chemokines, ATP, and inflammation-associated small, noncoding RNA (miRNAs) [105]. Whether or not mitochondrial or EV-mediated mechanisms of communication may contribute to diet-induced hypothalamic neuronal alterations, however, remains to be directly explored. In addition to being influenced by extrinsic factors, the maladaptive cross-talk between microglia and neurons observed in obesity may then originate from intrinsic variables, including genetic predisposition. Indeed, human subjects carrying common polymorphisms in the Jun N-terminal kinase (JNK) or the melanocortin 4 receptor (MC4R) Finally, the maladaptive microglia–neuron crosstalk that is observed in response to overnutrition likely involves synergy with other cell types—for instance, astrocytes [5,111], which are key components of the tripartite synapse. Moreover, peripheral monocytes/ macrophages, which can extravasate from circulation to the brain during DIO progression, are known to contribute to neuroinflammation and microglia–neuron communication [36]. Tanycytes, which are specialized ependymocytes that form a blood–cerebrospinal fluid barrier in the hypothalamus, may have an additional influencing role [39] due to their ability to fine-tune the brain transport of nutrients and hormones implicated in metabolic control [112,113]. Additionally, endothelial cells located in the hypothalamus are emerging as key cellular mediators of obesity and related metabolic disorders [114] and may therefore contribute to the modulation of diet-induced inflammatory responses in hypothalamic neurocircuitries [37].

While trying to fill in this knowledge gap, cellular heterogeneity may represent a challenging obstacle, but also a relevant aspect. Indeed, microglia and neurons in the hypothalamus form highly diverse cellular clusters [115–118], and may therefore operate in response to diverse and cell-specific mechanisms of communication. Accordingly, different subpopulations of microglial cells have been found in the mediobasal hypothalamus of rodents and humans [119], and these subpopulations present a high degree of spatial localization and distinct molecular signatures in response to diet-induced inflammatory conditions [119].

Thus, addressing when, where, and how hypothalamic neurons and surrounding microglial cells communicate in response to diet-induced metabolic stress will be challenging, but this effort may ultimately favor progress into both the etiology and treatment of metabolic disorders.

**Author Contributions:** S.L.: writing original draft, A.N.: writing, C.Q.: conceptualization and writing. All authors have read and agreed to the published version of the manuscript.

**Funding:** C.Q. acknowledges support by INSERM, Agence Nationale de la Recherche (neuroIDobese, ANR-20-CE14-0046), Société Française d'Endocrinologie (Pfizer–SFE Prix de Recherche en Endocrinologie), Société Française de Nutrition and Société Francophone du Diabète.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** No new data were created or analyzed in this study. Data sharing is not applicable to this article.

**Conflicts of Interest:** Authors declare no conflict of interest.

#### **References**


## *Review* **Hypothalamic Astrocytes as a Specialized and Responsive Cell Population in Obesity**

**Ismael González-García 1,2,\* and Cristina García-Cáceres 1,2,3**


**Abstract:** Astrocytes are a type of glial cell anatomically and functionally integrated into the neuronal regulatory circuits for the neuroendocrine control of metabolism. Being functional integral compounds of synapses, astrocytes are actively involved in the physiological regulatory aspects of metabolic control, but also in the pathological processes that link neuronal dysfunction and obesity. Between brain areas, the hypothalamus harbors specialized functional circuits that seem selectively vulnerable to metabolic damage, undergoing early cellular rearrangements which are thought to be at the core of the pathogenesis of diet-induced obesity. Such changes in the hypothalamic brain region consist of a rise in proinflammatory cytokines, the presence of a reactive phenotype in astrocytes and microglia, alterations in the cytoarchitecture and synaptology of hypothalamic circuits, and angiogenesis, a phenomenon that cannot be found elsewhere in the brain. Increasing evidence points to the direct involvement of hypothalamic astrocytes in such early metabolic disturbances, thus moving the study of these glial cells to the forefront of obesity research. Here we provide a comprehensive review of the most relevant findings of molecular and pathophysiological mechanisms by which hypothalamic astrocytes might be involved in the pathogenesis of obesity.

**Keywords:** astrocytes; neurons; hypothalamus; obesity

#### **1. The Hypothalamus: Integrator of Neuroendocrine and Systemic Metabolic Homeostasis**

The organization and cellular components of mammalian neuroendocrine systems are highly evolutionarily conserved and require circuits precisely tuned throughout life. The central nervous system (CNS) senses and processes local and peripheral afferent metabolic signals emanating from circulation (hormones, nutrients), and translates information from sensory nerves into physiological responses regulating feeding behavior and energy expenditure [1–3]. The identification of the hypothalamus as a metabolic center arose from pioneered observations during the mid-20th century. These seminal studies demonstrated first that inflicting lesions or the presence of tumors in specific hypothalamic brain areas elicited either hyperphagic obesity or severe anorexia [4]. Subsequently, the advances in technologies and knowledge allowed for further confirmation that the key role of the hypothalamus is its ability to integrate and process metabolic cues of endocrine and autonomic nervous system origins; thus, positioning this brain area at center stage in the regulation of body weight and energy balance.

The hypothalamus is structured and organized into different nuclei with the arcuate nucleus (ARC) having been more deeply studied for containing specialized neuronal populations. These neurons are vastly interconnected between themselves and with other extra-hypothalamic areas, which allows for a high intercommunication and redundancy in the homeostatic mechanisms [5]. The ARC is located at the base of the hypothalamus

**Citation:** González-García, I.; García-Cáceres, C. Hypothalamic Astrocytes as a Specialized and Responsive Cell Population in Obesity. *Int. J. Mol. Sci.* **2021**, *22*, 6176. https://doi.org/10.3390/ijms22126176

Academic Editors: Rosalía Rodríguez-Rodríguez and Cristina Miralpeix

Received: 14 May 2021 Accepted: 4 June 2021 Published: 8 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

in close proximity to the median eminence (ME) and it is there where two neuronal populations well-known for being fundamental for whole-body energy regulation reside: Neuropeptide Y and agouti-related peptide (NPY/AgRP) expressing neurons whose activation exert orexigenic effects and reduce energy expenditure, and proopiomelanocortin (POMC) expressing neurons with opposing effects on feeding behavior and energy expenditure compared to NPY/AgRP neurons [1]. Both sets of neurons express receptors for peripheral endocrine inputs (e.g., leptin, insulin, GLP1, free fatty acids, and ghrelin), but also for central-derived signals (e.g., NPY, GABA, serotonin, and melanocortin), and the overall result of the integration and processing of these signals is to influence the wiring and flexibility of neuronal circuits for the homeostatic maintenance of energy balance. Thus, the connectivity of these ARC neuronal circuits is flexible, varying depending on the energy storage and nutritional state of the entire organism. Such adaptive plasticity of neuronal connectivity requires the chemical and physical association and cooperation of adjacent glial cells, which are active units in the regulation of synaptic function.

In the last few years, hypothalamic astrocytes have started to draw some attention in the neuroendocrine field for acting as hypothalamic functioning regulators, particularly for their role in the central control of metabolism. Indeed, several findings emphasize the crucial involvement of astrocytes for their functional interactions with specific ARC-resident neuronal populations [6–10] together forming one unique functional circuit in metabolic regulation. The advances in neuroscience and molecular biological technologies designed for targeting non-neuronal cells are allowing us to advance current knowledge by providing novel concepts about the functional contribution of astrocytes in the physiological regulation of food intake and energy homeostasis.

#### **2. Hypothalamic Astrocytes: A Specialized Population in Regulating Metabolism**

The hypothalamus, in particular its ventral border where the ARC is located, is beneficially situated adjacent to the ME, a circumventricular organ with a particular angioarchitecture (i.e., fenestrated microvessels which lack blood brain barrier) positioning it as a more metabolically sensitive brain center over others in the CNS [11–13]. Such high and rapid responsiveness to metabolic disturbances is also due to being highly-enriched in a wide-array of metabolic receptors and transporters acting as primary sensors and responders to nutrients and endocrine cues emanating from the bloodstream [14–17]. Additionally, emerging insights highlight that hypothalamic astrocytes, tightly connected to ARC-resident neurons and blood vessels [18], are strategically situated to survey the organism's metabolic status and to, in turn, remodel local vascular beds thereby controlling the selective access of some circulating factors into the brain [19]. Earlier studies have also indicated that hypothalamic astrocytes are particularly and primarily affected in response to metabolic damage induced by hypercaloric diets [20], positioning these glial cells as the center of attention for obesity research. In a more physiological context though, current literature underlines the role of astrocytes in the control of metabolism by their role of nutrient and hormone sensing regulation, and via the release of gliotransmitters in crosstalk with neurons.

#### *2.1. Hypothalamic Astrocytes: In Bidirectional Tuning with Neurons for Hypothalamic Feeding Control*

In recent years, several studies based on Designer Receptor-Exclusively Activated by Designed Drug (DREADD) have revealed that changes in astrocyte activity can impact the feeding behavior response to leptin and ghrelin. In 2015, a study led by Yang and colleagues showed that activation of astrocytes within the mediobasal hypothalamus (MBH) leads to reduced feeding by inhibiting AgRP neurons via adenosine A1 receptors, both in basal conditions and after ghrelin-evoked feeding. Moreover, pharmacogenetic activation of MBH-astrocytes increases leptin's anorectic effect [6]. This study proposed a mechanism by which astrocytes participate in food intake regulation through the release of adenosine, a gliotransmitter known by its inhibitory action on both pre-and post- synaptic neurons [21], leading to the suppression of AgRP neuronal activity and its orexigenic action.

Contrary to this, a later study showed opposing results, observing an increase in AgRP neuronal stimulation and food intake after chemogenetic activation of astrocytes [7]. These discrepancies between studies could be explained by the heterogeneous nature of the glia– neuronal interconnections in the hypothalamic feeding circuits where their connectivity and plasticity can vary depending on health (also influenced by sexual hormones and age) and metabolic status of the animal. Therefore, small alterations in the experimental paradigm such as time, last meal, environment, the anatomical location, and number of hypothalamic astrocytes chemogenetically manipulated could shift the circuit specific manipulation for controlling feeding and thus explain the diverse result observed. A recent study has pointed out the essentiality of a regulatory role of AgRP neurons over adjacent astrocytes for dynamic fine-tuned adaptations of their activity as part of homeostatic control of feeding [10]. This study has shown that chemogenetic or ghrelin-evoked activation of AgRP neurons promotes the release of GABA to increase astrocytic endfeet coverage on the membrane of AgRP neurons and triggers prostaglandin E2 release to subsequently increase the excitability of AgRP neurons via EP2 receptors. These findings support a dynamic and bidirectional communication between AgRP neurons and astrocytes for a feed-forward autoactivation loop of hunger circuits. Other studies have also reported that a gliopeptide Acyl-CoA–binding protein–derived (ACBP-derived) and its product octadecaneuropeptide (ODN) secreted by astrocytes is important for hypothalamic astrocyte–neuron tuning in the regulation of energy balance. In particular, they showed that pan-brain deletion of the ACBP gene in astrocytes promotes diet-induced hyperphagia and obesity in mice. Mechanistically, they proposed a "gliogenic" endozepine released mechanism based on the ACBP–ODN signaling in ARC astrocytes to influence the activity of POMC neurons to reduce feeding and weight gain [9].

#### *2.2. Hypothalamic Astrocytes: Nutrient and Hormone Sensing-Dependent Regulation*

*Lipid sensing.* Astrocytes uptake lipids and store them in the form of droplets via several lipid transporters, including lipoprotein lipase (LPL), which is essential for the control of cellular lipid storage in the brain. When *lpl* is postnatally deleted in astrocytes, mice develop an aggravated obesogenic phenotype characterized by increased body weight gain and glucose intolerance in response to a high fat diet [22]. Mechanistically, it was found that these mice exhibited increased ceramide accumulation, specifically in hypothalamic neurons, a phenomenon observed in neurodegenerative processes induced by lipotoxicity [13,23]. In vitro studies have reported that hypothalamic astrocytes accumulate lipid droplets under saturated fatty acid-rich conditions, adopting a reactive inflammatory profile and eliciting chemotactic activity of microglia to produce inflammatory mediators [24]. Other in vitro work in human astrocytes has also pointed out a direct link between a disruption of lipid metabolism in astrocytes and insulin resistance [25]; however, the use in this case of cultured astrocytes would require additional confirmation for its relevance at hypothalamic level.

*Ketone body sensing.* Besides lipid homeostasis function, astrocytes play a key role in fatty acid oxidation by synthesizing the majority of ketone bodies (KB) in the CNS. Studies in mice have shown that short-term exposure to a hypercaloric diet (HFD) is associated with higher levels of astrocyte-derived KB in the hypothalamus and an acute decrease in food intake. Accordingly, pharmacological inhibition of ketone synthesis blunts HFDinduced anorexia, supporting a model where astrocytes regulate fatty acids and KB levels in the hypothalamus as a lipid sensing mechanism, which is disrupted in leptin resistant obese rats [26]. Further, in vitro studies with the ketone body β-hydroxybutyrate have shown the direct effect of KBs on lowering the consumption of glucose and higher ability of mitochondria to metabolize pyruvate in mouse astrocytes, and consequently affecting the delivery of glucose and lactate to neurons [27]. Overall, these in vitro and in vivo findings indicate that astrocytes play a key role in maintaining a tight regulation of hypothalamic lipid homeostasis necessary for correct metabolic control. Further studies will need to address in detail how the lipogenic/lipolytic state of hypothalamic astrocytes is able to

determine its cellular bioenergetics and communication with the surrounding neurons for the regulation of metabolism.

*Glucose sensing*. The brain has high energetic requirements, with glucose being its preferred fuel. However, glucose or energy storage in the brain is rather limited and restricted to astrocytes that, unlike neurons, are able to store glucose in the form of glycogen deposits in order to energetically support themselves through its lactate-shuttle transport [28,29]. It must be noted though that other authors have raised some questions about whether lactate oxidation primarily occurs in neurons and therefore the lactate-shuttle transport model probably needs further validation [30]. Independent of this role of astrocytes in energy storing, they also behave as glucose sensors. One of the first pieces of evidence of astrocyte involvement in central glucose sensing was provided by studies based on glucose transporter 2 (GLUT2) interrogation, which demonstrated that its inhibition in the ARC impaired glucose sensing by inhibiting the insulin response in response to intra-carotid glucose injection [31]. Interestingly, it was described that GLUT2 is mainly expressed in astrocytes instead of neurons within the hypothalamus [32–34]. Later studies using *glut2* global knockout (KO) mice confirmed the relevance of astrocytes in central glucose sensing. The mice globally lacking GLUT2 exhibited elevated blood glucagon levels and a decreased glucagon secretion in response to fluctuations in systemic glucose, a phenotype which illustrates that GLUT2 is essential for the counterregulatory response to hypoglycemia. Further, transgenesis-induced GLUT2 rescue specifically in astrocytes was able to restore glucagon secretion [35]. However, these last effects were specifically observed in astrocytes located in the brainstem, a relevant brain region regarding astrocyte glucose sensing and metabolism [36–38]. Glucose sensing is also a mechanism involved in the control of food intake, and thus the effects derived from hypothalamic astrocytes affect the way energy balance is regulated. In this regard, a different study showed that transgenic mice with inhibited GLUT2-mediated glucose detection augmented daily food intake by a mechanism that increased the meal size, but not the number of meals [39]. GLUT2 is located not only in astrocytes [32,35,40], but also in tanycytes and neurons [41,42]; therefore, these findings cannot be attributed solely to the astrocytic glucose transporter. In parallel, astrocytes located within the hypothalamus were also associated with the glucose sensing mechanism via GLUT1 in streptozotocin (STZ)-treated rats. Virogenetic overexpression of GLUT1 in hypothalamic astrocytes of STZ-diabetic rats with reduced GLUT1 normalizes circulating glucose levels and restores hypothalamic glucose sensing in rats, namely it rescues glucose production measured in intrahypothalamic clamp studies [43]. These findings support hypothalamic astrocytes as being essential components of the glucose sensing mechanism.

*Insulin sensing*. Insulin receptor (IR) expressed in hypothalamic astrocytes has been proven critical for the entry of glucose into the brain and for the modulation of glucose metabolism within the brain. Indeed, mice with postnatal *ir* knockout in astrocytes develop a defective anorectic response to glucose and an altered glucose homeostasis [8,44]. Despite this, genetically ablated mice lack IR in the whole brain, the phenotype was reproduced under a virogenetic deletion approach, thus proving that astrocytes within the hypothalamus were sufficient to induce the changes in glucose homeostasis. Moreover, we observed that the glucose-evoked reduction in neuronal activation occurred specifically in the hypothalamus and particularly in POMC neurons, thus reinforcing the special relevance of astrocytes located in this brain region [8]. Subsequent in vitro studies confirmed that loss of IR in astrocytes downregulates the expression of GLUT1 [45], while insulin-like growth factor 1 receptor (IGF1R) cooperates together with insulin in order to stimulate glucose uptake by astrocytes [46]. Importantly, in vivo experiments have also shown that IR ablation in astrocytes induces a compensatory upregulation of IGF1 receptor [45] and therefore, the fact that these two receptors have high homology and convergent signaling pathways underlines the importance of understanding their dual functional role in astrocytes.

*Leptin sensing.* Initial studies in agouti viable yellow (Avy) mice, showed that leptin receptor (LepR) in astrocytes may play a role in leptin transport across the BBB and in the regulation of adult-onset obesity of these animals [47,48]. Molecular analysis of the brain in

the same model, in addition to HFD-fed mice, showed that obesity rapidly increases LepR expression in astrocytes where it then influences the leptin-induced calcium signaling in astrocytes [49]. Later, additional studies showed that central and chronic leptin administration modifies hypothalamic astrocyte morphology and glutamate transporters [50,51]. The subsequent generation of specific astrocyte LepR ablation mouse models confirmed these initial findings. In this regard, constitutive astrocyte-specific *lepr* (KO) mice did not develop an altered phenotype in terms of energy balance, likely due to a compensatory mechanism [52]. However, inducible astrocyte-specific *lepr* KO mice exhibited a lack of proper anorectic response to leptin and showed an exacerbated refeeding after fasting. Interestingly, both physiological alterations were associated with changes in POMC neurons (i.e., lower activation after leptin treatment) and AgRP neurons (i.e., overactivation during the refeeding phase) [53]. Overall, these findings support the existence and functional relevance of LepR in astrocytes. In agreement with previous reports, our work has recently confirmed the specific presence of the truncated LepR<sup>a</sup> in fluorescence-activated cell sorting (FACS)-isolated hypothalamic astrocytes [19].

*GLP1 sensing*. Recent studies have highlighted the role of astrocytic glucagon-like peptide-1 (GLP-1) signaling in the regulation of brain glucose uptake [54]. When GLP-1R was postnatally and genetically ablated in hypothalamic astrocytes, mice displayed an increased brain glucose uptake, which was associated with an improvement in memory formation and systemic glucose response [54]. Further evidence suggested that these effects are mediated through fibroblast growth factor (FGF)-21 action, given that neutralization of central FGF21, as well as astrocyte-specific ablation of FGF21, abrogated the improvements in glucose metabolism observed in the astrocyte-specific *glp-1r* knockout mice [54].

#### **3. Hypothalamic Astrocytes in the Physiopathological Development of Obesity**

Based on human genome-wide association studies (GWAS) from obese patients, it has recently emerged that the brain, in particular the hypothalamus, controls most aspects of systemic metabolism over other classically described peripheral endocrine axes, suggesting that obesity is largely a brain disease [55]. Interestingly, clinical studies have reported that being obese is associated with brain atrophy; describing alterations in food rewardregulating circuits [56], nutrient and hormone sensing-dependent neuronal activity [57,58], hypothalamic gliosis [20,59–61], and angiopathy [62], all being signs of neuronal dysfunction and aging [63]. Most of these changes have been reported to originate from the hypothalamus, which is considered a selective vulnerable area in diet-induced obesity and has been suggested as a hallmark of neuronal dysfunction associated with the initiation and progression of this pathology [64,65]. In fact, it is thought that obesity-induced hypothalamic dysfunction emerges from aberrant glia responses (astrocyte and microglia) inducing impairments in the connectivity and activity (responsiveness) of ARC-resident neuronal circuits responsible for orchestrating metabolic control and contributing to the aggravation of this pathology [64–67]. Despite the relative contribution of each glial cell type [68] and their crosstalk [69] to obesity pathophysiology being a current topic of research in this section, we will focus in the astrocyte-related mechanisms.

#### *3.1. Inflammatory-Related Mechanisms*

One of the most established pathophysiological mechanisms to explain obesity development is the inflammatory and oxidative stress hypothesis, mediated through different pathways, occurring in different endocrine tissues, including in the hypothalamus. Early studies pointed out that rich caloric diets induce a low-grade inflammation within the hypothalamus characterized by a rapid rise in cytokines [70], gliosis, and metabolic stress leading to dysfunction of hypothalamic circuits and consequent obesity [20]. Moreover, hypothalamic inflammation has also been associated with the development of central resistance to endocrine signals before changes in body weight [20,70,71]. Indeed, hypothalamic inflammatory processes are tightly linked with hormone resistance and obesity. This is supported by studies reporting that amelioration of hypothalamic inflammatory processes

i.e., a blunted tumor necrosis factor alpha (TNFα) signaling, an ablated suppressor of cytokine signaling-3 (Socs3), and the disruption of hypothalamic-specific interleukin (IL)-6) led to rescue of both leptin and insulin signaling and protection from the metabolic alterations induced by HFD [72–74]. Several pathways have been proposed as mediators of hypothalamic inflammatory responses in diet-induced obesity. Notably, some studies pointed out that the IkB kinase-β (IKKβ)/nuclear factor κB (NF-κB) pathway mediates the expression of several pro-inflammatory genes within the hypothalamus, including cytokines and chemokines [75], and leads to endoplasmic reticulum (ER) stress, a cellular process characterized by the accumulation of misfolded or unfolded proteins in the ER lumen [76,77]. Both signaling pathways are actually interconnected, and it was shown that HFD leads to increased ER stress and consequently activates IKKβ–NF-κB signaling in the hypothalamus, producing the altered response to hormones and ultimately leading to metabolic homeostasis dysregulation.

#### *3.2. Reactive Astrogliosis in Obesity*

This term, which recently reached a consensus agreement on its definition, together with "astrocyte reactivity" or "reactive astrocytes" [78], entails the morphological and/or functional changes seen in astrocytes responding to healing processes or pathophysiology [79]. Seminal studies led by Horvath and colleagues reported that diet-induced obesity in mice also induces reactive astrogliosis in the hypothalamus [18]. The ultrastructural analyses of these brains also showed that, concomitant with the findings, there were changes in the physical interactions of astrocytes with endothelial cells and neurons causing alterations of the cytoarchitecture and synaptology of hypothalamic circuits [18]. Surprisingly, these astrocytic morphological changes, as well as the increased cytokine production in the hypothalamus, occurred before any sign of systemic inflammation or body weight gain [20]. HFD-induced reactive astrogliosis is characterized by an upregulation of the structural protein glial fibrillary acidic protein (GFAP), promotion of a pro-inflammatory phenotype (producing and releasing cytokine markers), and the acquisition of a hypertrophic morphology, but without proliferation-unlike what was observed in other brain pathologies. Remarkably, a recent study has found that GFAP gene induction occurs very early, and even one hour of HFD exposure is sufficient to find a significant change. That short exposure interferes also in the neuropeptide and an inflammatory-like gene responses exerted by the chemogenetic activation of astrocytes [80].

Blocking inflammatory-induced astrocyte activation via expression of a dominantnegative form of the NFκB inhibitor, IκBα, in GFAP<sup>+</sup> astrocytes in a doxycycline-inducible manner showed an inhibition of HFD-induced reactive astrogliosis in the hypothalamus coupled with a promotion of hyperphagia, even 24 h following the dietary switch [81]. These findings suggest that the NFκB-IκBα pathway in astrocytes controls HFD-induced reactive astrogliosis illustrating acute morphological adaptations in astrocytes for precisely controlling food intake [81]. Of note, astrocyte changes described here were found in the MBH; the genetic mouse model was generated modifying all GFAP<sup>+</sup> astrocytes in the CNS and therefore an indirect effect via non-hypothalamic regions cannot be ruled out. Douglass and colleagues found that postnatal ablation of IKKβ specifically in astrocytes after HFD exposure has a protective effect on weight gain mediated through a reduction in food intake and an increase in energy expenditure, which was also accompanied by an improved glucose tolerance and insulin sensitivity. Importantly, both hypothalamic inflammation and reactive astrogliosis were diminished in HFD-fed IKKβKO mice [82]. These findings were similarly supported by other authors showing that astrocytic IKKβ/NF-κB upregulation is a key signaling pathway during overnutrition, which mediates some of the early-onset effects of obesity. Mechanistically, astrocytic activation via IKKβ/NF-κB seems to modify astrocytic morphology (i.e., decreased astrocytic process density), a downstream regulation of hypothalamic extracellular GABA levels, and BDNF expression within the hypothalamus [83]. That being said, hypothalamic inflammation and reactive astrogliosis have been used as generic terms to describe a variety of morphological and molecular

changes occurring under the effects of high fat diet exposure, sometimes describing phenotypes under the same phenomena, which only partly fulfill the molecular and cellular characteristics of the other phenotypes reported. In this regard, there is a clear demand for neuroscientists to agree on a clear and meaningful nomenclature, which precisely describes the morphological and functional readouts to evaluate.

Some studies indicate that dietary factors such as fat are determinant triggers to the hypothalamic inflammatory response and the changes induced in glial cells. In vitro studies have pointed out that saturated fatty acids (i.e., palmitic, lauric, and stearic acid) induce the release of TNF-α and IL-6 from cultured astrocytes [84]. Mechanistically, saturated fatty acids require toll-like receptor 4 (TLR4) to induce cytokine release, while the pharmacological inhibition of p38 or p42/44 MAPK pathways precludes the pro-inflammatory actions of the saturated fatty acids [84]. In agreement with these findings, TLR2/4-deficient mice were protected from HFD-induced weight gain as well as glucose intolerance and insulin resistance. Moreover, C3H/HeJ/TLR2-deficient mice also showed a decreased expression of IL-6 in both basal and insulin-stimulated states. Following that, treatment with an IL-6-neutralizing antibody led to a rescue of insulin sensitivity in HFD-fed mice despite no changes being found in body weight [85]. Gao and colleagues found that while reactive astrogliosis was mainly a response to the fat content of the diet, it is the combination of sucrose and fat consumption that are responsible for stimulating some other aspects of hypothalamic inflammatory response, such as microgliosis and angiogenesis [86]. Lastly, other studies have highlighted that the astrocytes' response to develop an HFD-induced hypertrophic phenotype can vary depending on the age of the mice. Specifically, Lemus and colleagues observed that aged mice showed a higher GFAP expression in basal conditions than younger ones, which can affect the visualization of differences between mice fed with chow diet and HFD at a certain age [87]. Aside from diet-induced obesity, other studies have also reported the presence of reactive astrogliosis in monogenetic genetic obese mouse models [47,49,88], or by maternal dietary influence during the fetal and early stages of life of the offspring [51]. Therefore, a careful interpretation has to be taken about the context in which reactive astrogliosis is evaluated because the molecular pathways entailed and functional implications in obesity development could vary depending on the nature of the trigger (e.g., dietary factor, aberrant metabolic, or hormone inputs), the brain areas involved, the status of surrounding cellular compounds (microgliosis and neuronal connectivity), as well as other relevant parameters, such as health state and age of the animal that could interfere with the resulting outcomes and conclusions.

#### *3.3. Alterations in the Astroglia-Vascular Interface in Obesity*

HFD-induced reactive astrogliosis has been detected in specific hypothalamic regions, in particular, those ones close to the third ventricle, such as the MBH [18,64,88], which have a higher accessibility of circulating factors and are partly placed out of the BBB [88]. Indeed, HFD-fed mice show a significant increase in blood vessel length and density in the MBH, as well as a lack of BBB integrity [62]. Interestingly, human diabetic patients exhibited an increased number of arteries and arterioles in the hypothalamic infundibular nuclei, suggesting a potential link between hypothalamic vascular remodeling and the development of metabolic diseases in humans [62]. A recent study by our group has further dissected the pathogenic processes leading to such impairments, uncovering the role of hypothalamic astrocytes in promoting hypothalamic microangiopathy and the systemic arterial hypertension associated with obesity. By using virus-mediated gene transfer, it was found that astroglial vascular endothelial growth factor (VEGF) in the hypothalamus exerts significant effects on systemic blood pressure control via tuning sympathetic outflow. These new findings supporting a novel mechanism with hypothalamic angiopathies at its center are important as they could explain how obesity-associated hyperleptinemia rather than adiposity causes hypertension [19] (Figure 1).

**Figure 1.** Hypothalamic astrocytes participate in the regulation of metabolic homeostasis and act as key players in the development of diet-induced obesity. Astrocytes located in the hypothalamus express several endocrine receptors such as, insulin, leptin, and GLP-1. The action of hormones and pharmacological analogues on hypothalamic astrocytes has been described to influence several aspects of normal energy homeostasis. Under energy-dense diet exposure, the hypothalamus experiences a low-grade inflammation characterized by a rapid rise in cytokines, which trigger downstream pathways such as the IKKβ/NF-κB and TLR2/4. This also leads to increased hypothalamic ER stress, which seems to interfere with the adequate response to hormones. Together, there is an HFD-reactive astrogliosis characterized by an overexpression of GFAP, the acquisition of a hypertrophic morphology, and a synaptic input reorganization of the melanocortin system, which may suggest a direct involvement of hypothalamic astrocytes in the pathogenesis of diet-induced obesity. At vascular level, obesogenic diets trigger astroglial VEGF-driving angiogenesis in the hypothalamus to promote the development of systemic hypertension. ER, endoplasmic reticulum; IKKβ, IkB kinase-β; GFAP, glial fibrillary acidic protein; GLP-1, glucagon-like peptide-1; NF-κB, nuclear factor κB; POMC, proopiomelanocortin; TLR2/4, toll-like receptor 2/4; VEGF, vascular endothelial growth factor.

#### **4. Outlook**

The hypothalamus, with its different nuclei, is a good example of how different neuronal types and connections translate to functional specialized roles. Given the complex nature of hypothalamic cytoarchitecture where circuits that play opposing roles in the control of metabolism are tightly intermingled, it will not be surprising that the study of astrocytes in the control of metabolism should be approached to specifically address the precise anatomical location of the circuits and the interaction among them. To date, several studies have contributed to understanding the functional relevance of astrocytes in the development of obesity; however, a big gap is still present on how these alterations affect the neural networks in which astrocytes are embedded. Thus, the idea of a disruption in the astrocyte-neuron communication within the hypothalamus seems a plausible, yet poorly investigated, hypothesis that could contribute to the pathophysiology of obesity. Research on how astrocyte–neuron communication is altered by exposure to obesogenic insults will allow a better understanding of the hypothalamic alterations that occur in this context.

Another important factor linked to the regional specialization is astrocyte heterogeneity. Regardless of the classical morphological taxonomy (i.e., fibrous and protoplasmic astrocytes), the specific molecular markers label a wide range of astrocyte subpopulations, which could also indicate a complex array of functional actions. In the upcoming years, an important topic to address will be the relative contribution of each subpopulation to the development of obesity. Taking advantage of the novel technological advances in the study of cell- and synapse-specific functionality by using combined two-photon imaging, patch-clamp recordings, cellular mapping of transcriptomic profiles in the tissue context using single-cell RNA sequencing, and new in situ hybridization technologies. In combination, researchers will be able to understand the tissue structure and function on a single astrocyte level, helping to identify specific hypothalamic astrocyte subpopulations, which might be critical for the pathophysiology and development of obesity.

**Author Contributions:** Conceptualization, I.G.-G.; writing—original draft and figure preparation, I.G.-G.; writing—review and editing, I.G.-G. and C.G.-C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by European Research Council ERC (StG AstroNeuroCrosstalk no. 757393), the German Research Foundation DFG under Germany's Excellence Strategy within the framework of the Munich Cluster for Systems Neurology (EXC 2145 SyNergy-ID 390857198) and Helmholtz Excellence Network, the German Research Foundation DFG Nutripathos Project ANR-15-CE14-0030, the DZD, and the Initiative and Networking Fund of the Helmholtz Association. IG-G is a recipient of a fellowship from European Union's Horizon 2020 research and innovation program under the Marie Sklodowska-Curie actions (842080-H2020-MSCA-IF-2018).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors thank Cassie Holleman (Helmholtz Zentrum München; Germany) for her comments.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in study design, data collection, or analyses, decision to publish, or preparation of the manuscript.

#### **References**


## *Review* **Hypothalamic Microglial Heterogeneity and Signature under High Fat Diet–Induced Inflammation**

**Natália Ferreira Mendes 1,2 , Carlos Poblete Jara 1,2 , Ariane Maria Zanesco 2,3 and Eliana Pereira de Araújo 1,2,\***


**Abstract:** Under high-fat feeding, the hypothalamus atypically undergoes pro-inflammatory signaling activation. Recent data from transcriptomic analysis of microglia from rodents and humans has allowed the identification of several microglial subpopulations throughout the brain. Numerous studies have clarified the roles of these cells in hypothalamic inflammation, but how each microglial subset plays its functions upon inflammatory stimuli remains unexplored. Fortunately, these data unveiling microglial heterogeneity have triggered the development of novel experimental models for studying the roles and characteristics of each microglial subtype. In this review, we explore microglial heterogeneity in the hypothalamus and their crosstalk with astrocytes under high fat diet–induced inflammation. We present novel currently available ex vivo and in vivo experimental models that can be useful when designing a new research project in this field of study. Last, we examine the transcriptomic data already published to identify how the hypothalamic microglial signature changes upon short-term and prolonged high-fat feeding.

**Keywords:** microglia; gliosis; hypothalamus; high-fat diet; obesity; transcriptomic

### **1. Introduction**

Hypothalamic inflammation is a condition frequently observed in experimental models of diet-induced obesity (DIO) [1–3] and obese humans [4–6]. This inflammatory response is mainly triggered by excessive saturated fatty acids (SFAs) from the diet [7–9], which reach the neural tissue mainly through the median eminence (ME), where fenestrated vascular endothelium lacks a blood–brain barrier (BBB) [10,11]. Brain perivascular macrophages (PVMs) also react to excessive free fatty acids (FFAs) circulating in the blood vessels, with a consequent increase in BBB permeability [12,13]. Glial cells, such as astrocytes and microglia, quickly sense and react to the presence of those SFAs in the hypothalamic parenchyma, releasing pro-inflammatory cytokines, chemokines, and reactive oxygen species (ROS) [14,15]. If the stimulus persists, the hypothalamic neuronal network may be damaged, resulting in neuro-inflammation, which eventually leads to energy balance disruption [16,17], and finally, to neuronal dysfunction/apoptosis [18].

Researchers have shown that hypothalamic inflammation initiates just a few hours/ days upon high-fat feeding [5,15,19]. After the onset of the inflammatory response, bone marrow–derived cells (BMDC) and other peripheral immune cells, such as neutrophils, lymphocytes, and regulatory T (Treg) cells can arise into the hypothalamic parenchyma in a time-dependent manner, directly affecting glial functions [12,20,21]. To avoid further metabolic complications, hypothalamic neuronal and non-neuronal cells, along with peripheral immune cells, should act together in an orchestrated mode beginning with the earliest phase of the inflammatory response.

**Citation:** Mendes, N.F.; Jara, C.P.; Zanesco, A.M.; de Araújo, E.P. Hypothalamic Microglial Heterogeneity and Signature under High Fat Diet–Induced Inflammation. *Int. J. Mol. Sci.* **2021**, *22*, 2256. https://doi.org/10.3390/ijms22052256

Academic Editor: Rosalía Rodríguez-Rodríguez

Received: 15 January 2021 Accepted: 19 February 2021 Published: 24 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

The molecular mechanisms underlying microglial immune and metabolic interactions with other cell types under high-fat diet (HFD)-induced hypothalamic inflammation still require a more detailed exploration [22]. Recent data unveiling microglial diversity and signatures throughout the brain have contributed to the development of novel state-ofthe-art approaches in experimental studies, which can be valuable for this field in the coming years.

In this review, we explore novel findings of hypothalamic microglial diversity from rodents and humans. We examine the subtypes of microglia that may be involved in HFD-induced hypothalamic inflammation and investigate how these cells interact with astrocytes upon high-fat feeding. Beyond that, we discuss which models can be useful to get the most reliable data when studying distinct subsets of microglia, myeloid cells, and border-associated macrophages (BAMs). Last, we investigate novel transcriptomic data already published to clarify how the hypothalamic microglia signature changes under saturated fat consumption.

#### **2. Microglial Heterogeneity in the Hypothalamus**

Microglia were identified by Pío del Río-Hortega in 1919, but only in the last two decades has the interest in these cells grown exponentially, with the discovery of their unique origin in the yolk sac and motile capacity [23,24]. Since their identification, microglia have been studied as a unique macrophage-like cell type in the central nervous system (CNS), able to quickly react to a wide range of stimuli by switching their phenotype activation between M1 and M2 subtypes [25,26]. According to this past view, microglia represent a naïve cell type that could equally react to any stimuli by acquiring a predetermined phenotype.

Most recently, advances in genetic tools have been developed and extensively employed in experimental research, enabling a deeper understanding of microglial diversity. Stratoulias et al. [27] recently proposed a new classification, in which microglia constitute a heterogeneous cell group, and each subtype has distinct properties and physiological functions, reacting differently to stimuli. This new view is based on the regional steadystate heterogeneity of microglia and their broad gene marker diversity. Curiously, from six putative microglial subtypes with unique specializations presented by these authors, the only one found in the hypothalamic area is dark microglia (DM). Despite their classification in subtypes, it is important to highlight the existence of many other types of microglia showing distinct features and functions in the hypothalamic area, which were not considered, or at least did not show up, in their categorization.

DM are also found in the hippocampus, cerebral cortex, and amygdala [28]. These cells can only be visualized through high-spatial-resolution transmission electron microscopy and display markers of oxidative stress, such as a condensed, electron-dense cytoplasm and nucleoplasm, dilatation of the Golgi apparatus and endoplasmic reticulum, mitochondrial alteration, and a partial to complete loss of the heterochromatin pattern. Functionally, DM are very active and show extremely thin processes, which allow them to make contact with synaptic elements [28]. Despite being rarely observed in healthy young adult mice, DM are widespread upon chronic stress, ageing, in CX3C chemokine receptor 1 (CX3CR1)-knockout (KO) mice, and in Alzheimer's disease pathology (APP/PS1 model) [27].

It remains unknown through which mechanisms DM are involved in the central inflammatory response, but it is reasonable to speculate that DM functions, similarly to other microglial subtypes, are influenced by peripheral signals. Savage et al. [29] showed that 24 h after an acute systemic injection of lipopolysaccharide (LPS), there are alterations in the inflammatory profile and in the microglial ultrastructure in the hippocampus, with no direct impact on DM. Curiously, DM are largely found in C-C chemokine receptor type 2 (CCR2-KO) mice, which present impaired recruitment of peripheral monocytes to the brain [30]. This finding is particularly interesting because upon prolonged HFD consumption, these peripheral cells are recruited to the hypothalamic parenchyma, as previously mentioned, similarly to what happens in many types of acute CNS injury [21].

Since their first description in 2016 [28], studies have described several DM features in rodents [31–36] and humans [37]. However, none of these studies have focused on DM roles specifically in the hypothalamus. Advances in this subject remain necessary to clarify how DM crosstalk with other microglia under hypothalamic inflammation, whether DM are involved in this recruitment of immune cells to the brain, and which peripheral stimuli can affect hypothalamic DM functions.

Microglia simultaneously express several hallmarks, such as *Iba1, Cx3cr1, P2ry12, Tmem119, Trem2, Cd11b, Hexb, Csf1r, Itgam,* and *Siglec*, among others [38–43]. However, there is no unique expression pattern of these transcripts in these cells, that is, it varies according to the pathological condition, age, sex, species, and brain area [27]. Particularly in mice, Valdearcos et al. [20] recently showed through immunofluorescence assays that the microglial signature varies according to hypothalamic nuclei and the dietary fat content. Thus, at least in mice, some CX3CR1+ cells in the arcuate nucleus (ARC) overlap with Tmem119+ or P2ry12+ cells, while other CX3CR1+ cells do not. When gliosis occurs upon consuming a HFD, ionized calcium binding adaptor molecule 1 (Iba1+) cells become widespread in the ME, ARC, and ventromedial nucleus (VMH), while Tmem119+ and P2ry12+ cells remain more restricted to the VMH when compared with the brain slices obtained from mice fed on chow diet. Some studies involving DM also show this heterogeneity, because these cells barely express *Iba1*, *Cx3cr1*, and *P2ry12*, but robustly express cluster of differentiation molecule 11B (*Cd11b*), which is involved in their synaptic pruning role [44,45].

These data indicate that some hypothalamic microglial subsets may have antiinflammatory functions, depending on the transcriptomic profile of each subtype and their status (steady-state or reactive). It is plausible to consider that it may also happen with other hallmarks, which explains why using only one surface marker for microglial staining, choosing a single cell line for ex vivo experiments, or even examining a single transgenic mouse model for manipulating microglia implicates a methodology bias.

Disease-associated microglia (DAM) are another subset of CNS resident macrophages that have been recently identified in experimental models of neurodegeneration and demyelination [46]. These cells are characterized by the expression of several genes, such as *Apoe*, *Clec7a*, *Cst7*, and *Spp1*. They are mainly found at sites of neurodegeneration and might play a protective role [47]. Interestingly, DAM hallmarks are also observed in human Alzheimer's disease post-mortem brains [46,48]. Triggering receptor expressed on myeloid cells-2 (TREM2) is also highly expressed by DAM; the activation of its intracellular signaling is essential for the transition of homeostatic microglia to the DAM state [49]. Both Toll-like receptor 4 (TLR4) and TREM2 can recognize different pathogen-associated molecular patterns [50,51] and other ligands, such as gram-negative bacteria [52], lipids [53], apolipoproteins (ApoE, ApoJ, and ApoA) [54–56], and nucleic acids released by apoptotic cells [57]. Because TREM2 and TLR4 share some ligands, it is not easy to identify which signaling pathway is activated through each receptor by these stimuli. However, experiments regarding TREM2 inhibition in combination with TLR4 stimulation by LPS have started to clarify the involvement of each receptor in neuro-inflammation [58].

An important characteristic observed in HFD-induced hypothalamic inflammation is the activation of nuclear factor kappa-light-chain-enhancer of activated B cells (NFκB) [14,59–62]. Conversely, inhibition of IKKβ/NF-κB signaling in microglia expressing CX3CR1 ameliorates DIO and hypothalamic inflammation [20]. In the CNS, TREM2 is widely expressed in microglia [63,64], where it acts by negatively regulating the activation of NF-κB [65]. In BV2 cell culture, TREM2 overexpression inhibits PI3K/AKT and NF-κB signaling pathways [66,67]. Interestingly, LPS reduces the expression of TREM2 in these cells through the activation of JNK and NF-κB, resulting in a vicious cycle [66]. Recently, Zhang et al. [68] treated BV2 cells with curcumin, a bioactive compound isolated from *Curcuma longa* with anti-inflammatory and antioxidant activities, and LPS; they observed reduced microglial activation via TLR4/NF-κB when compared with microglial cells treated

only with LPS, which occurred in parallel to the increased TREM2 expression, reinforcing the potential anti-inflammatory role of this receptor.

Microglial dynamics and density vary between brain areas and LPS doses during infection-induced inflammation [69]. In addition, considering that both LPS and SFAs can activate TLR4 and TREM2 intracellular signaling pathways, it is reasonable to speculate that a HFD could also have some effect on these microglial characteristics, especially in the hypothalamus. Yet, whether continued activation of TREM2+ cells under prolonged HFDinduced inflammation is implicated in the development of neurodegenerative diseases remains unexplored. The broad diversity of hypothalamic microglia at steady state and during the HFD-induced inflammatory response is shown in Figure 1. *Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 5 of 24

**Figure 1.** Microglial heterogeneity in the hypothalamus of rodents fed on chow or a high-fat diet (HFD). At steady state, microglia can be targeted by several hallmarks and are widely distributed in the hypothalamic parenchyma. Dark microglia (DM) are only visualized by transmission elec-**Figure 1.** Microglial heterogeneity in the hypothalamus of rodents fed on chow or a high-fat diet (HFD). At steady state, microglia can be targeted by several hallmarks and are widely distributed in the hypothalamic parenchyma. Dark microglia (DM)

tron microscopy and are close to blood vessels and neuronal synapses. Hallmarks of DM are well known in hippocampus, but remains barely explored in the hypothalamus. In HFD-induced hypo-

kines, reactive oxygen species (ROS), and growth factors. In the ARC, ME, VMH, there is a huge increase in Iba1+ cells (wide arrow) and a decrease in P2ry12+ and Tmem119+ cells, which become more restricted to the VMH (thinner arrows when compared to their arrows in the ARC). How other microglial hallmarks (e.g., *Trem2*, *Cd11b*, *Hexb*, *Csfr1*, among others) behave under HFD have not been studied yet in this inflammatory process. The role of DM and changes in their hallmarks in this specific inflammatory response have still not been explored. A leaky blood–brain barrier

are only visualized by transmission electron microscopy and are close to blood vessels and neuronal synapses. Hallmarks of DM are well known in hippocampus, but remains barely explored in the hypothalamus. In HFD-induced hypothalamic inflammation, reactive gliosis is observed and microglia change their spatial distribution and molecular signature. Microglia cells react by increasing the release of pro-inflammatory cytokines, reactive oxygen species (ROS), and growth factors. In the ARC, ME, VMH, there is a huge increase in Iba1+ cells (wide arrow) and a decrease in P2ry12+ and Tmem119+ cells, which become more restricted to the VMH (thinner arrows when compared to their arrows in the ARC). How other microglial hallmarks (e.g., *Trem2*, *Cd11b*, *Hexb*, *Csfr1*, among others) behave under HFD have not been studied yet in this inflammatory process. The role of DM and changes in their hallmarks in this specific inflammatory response have still not been explored. A leaky blood–brain barrier (BBB) allows free fatty acids (FFAs) accumulation in the hypothalamic parenchyma, boosting inflammation. If HFD persists for some weeks, peripheral myeloid cells, such as CD169+ and CCR2+ cells, are chemoattracted and infiltrate the hypothalamic parenchyma, but their functions in the HFD-induced inflammation need to be further studied. Abbreviations: ARC, arcuate nucleus of the hypothalamus; VMH, ventromedial nucleus of the hypothalamus; ME, median eminence.

> Most transcriptomic data currently available about microglia heterogeneity is from studies with rodents [39,70–72]. By contrast, human microglial diversity has only begun to be comprehended in the last few years. Masuda et al. [39] recently identified several clusters of microglia in healthy human brains and in the brains of patients with multiple sclerosis, but these authors evaluated only the cortex and temporal lobe. Similarly, Sankowski et al. [73] identified several microglial subsets in the human brain during homeostasis and some diseases through the combination of two high-dimensional technologies, single-cell RNA sequencing (scRNA-seq) and multiplexed mass cytometry (CyTOF), although the hypothalamus was not included in the analysis. Likewise, Böttcher et al. [74] applied multiplexed CyTOF to detect microglia regional heterogeneity in human postmortem samples from the subventricular zone, thalamus, cerebellum, temporal lobe, and frontal lobe.

> Unfortunately, there is still a lack of data from hypothalamic human microglia in the hypothalamus because there are several methodological difficulties in collecting human brain samples and keeping them cryopreserved without damage. These issues have limited large-scale studies. In addition, while recent advances in transcriptomics, multiplex protein expression analysis, and methods to detect chromatin structure have revealed many facets and details about microglia in rodents and humans, the correlation between the genomerelated information and their features remains to be explored.

#### **3. Hypothalamic Microglia–Astrocyte Crosstalk**

CNS homeostasis, development, injury, and repair are precisely controlled by appropriate cell–cell communication. Nevertheless, studies investigating microglial functions have been focused on microglia isolation, neglecting that physiological actions of these cells are part of a complex network involving other cell types. Therefore, new tools have been widely applied in experimental studies, allowing the determination of an infinite number of intercellular interactions by connecting ligands to target genes across cell types and tissues [75–77]. Unsurprisingly, the inflammatory process in the hypothalamus, observed in obesity, drives multiple harmful outcomes that can affect the interactions between microglia and other cell types [78,79]. As astrocytes and microglia play active roles from the onset of the hypothalamic inflammatory response, here we focus on their crosstalk under this process.

While microglia comprise only 5% to 10% of the total number of CNS cells in humans and rodents [80], astrocytes are the largest cell component of the brain, comprising at least 50% of all CNS cells [81]. Astrocytes make intimate contacts with synapses, blood vessels, and other glial cells, thus controlling synaptic transmission, BBB structure and function, and sensing nutrients and hormones in the blood [82,83]. They are also involved in the regulation of lipid metabolism and storage [84] and some studies have been indicating that astrocytes can also express some hormone receptors, such as leptin, insulin, ghrelin, and glucocorticoid receptors [85–88].

Coordinated crosstalk between glial cells is a determinant in homeostatic and pathological conditions [89]. Microglia and astrocytes interact via contact-dependent and secreted factors, such as growth factors, neurotransmitters, cytokines, chemokines, innate-immunity mediators, mitogenic factors, ROS, and metabolic mediators such as glutamate, which can be used for cell metabolism and may also mediate tissue changes [90]. These glial cells play a crucial role in synapse development and function in the healthy CNS, forming the 'quad-partite' synapse [91], which is essential for neuro-immune communication [90], critically contributing to brain homeostasis [92].

Both microglia and astrocytes quickly become activated upon an injury or any inflammatory stimulus [93]. Reactive microglia release several pro-inflammatory cytokines that induce astrocytic activation [75,93]. Likewise, HFD-induced microglial activation results in astrocytic proliferation, morphological changes, and increased production of cytokines and growth factors [61,84,94–96]. Classically, these morphologic and functional adaptations by glial cells under inflammation are known as gliosis [5,97] and are usually described in rodents fed on HFD for weeks [20,98], or even after very short periods of high-fat feeding [15].

In the hypothalamus, all these astrocytic characteristics are essential for controlling energy homeostasis. NF-κB signaling inhibition by IKKβ deletion in astrocytes (GFAP-Cre mice) reduces HFD-induced hypothalamic inflammation and reactive astrogliosis and attenuates DIO and glucose intolerance [14].

Similarly, knockdown of TGF-β1—which is predominantly synthesized by astrocytes, specifically in the ARC of HFD-fed mice—reduces TGF-β/SMAD and NF-κB signaling pathways and, consequently, also attenuates the inflammation [99]. Interestingly, this approach of hypothalamic TGF-β1 knockdown presents many metabolic benefits, preventing obesity development even under HFD. Countless microglial functions depend on TGF-β1, including their maturation and activation, both in homeostatic conditions and in response to any inflammatory stimuli [100,101]. In addition, the accurate functioning of the TGF-β1 signaling pathway is crucial to prevent excessive microglia activation [102,103].

In astrocyte cell culture, treatment with palmitic acid, an SFA largely found in HFD, evokes lipid droplet accumulation [95], culminating in an inflammatory response, characterized by an increased production of chemokines, such as chemokine C-C motif ligand-2 (CCL2), also called monocyte chemoattractant protein-1 (MCP-1). Curiously, the main chemotactic mechanism described in the hypothalamic inflammation is mediated by fractalkine (CX3CL1), a chemokine produced by hypothalamic neurons that acts through the fractalkine receptor (CX3CR1) [104]. However, another possible chemokine involved in the peripheral recruitment of immune cells to the hypothalamus is MCP-1 [105]. Peripheral immune CCR2+ and CD169+ cells arise in the hypothalamic parenchyma upon chronic periods of HFD [12,20]. CCR2+ cells of obese mice can enter into their white adipose tissue, and this chemo-attraction is mediated by the MCP-1/CCR2 axis [106].

In the CNS, the MCP-1/CCR2 axis has also been presented as an important chemotactic mechanism involved in the recruitment of peripheral immune cells to the paraventricular nucleus of the hypothalamus (PVH) upon inflammatory stimuli [107]. These authors showed that blockage of this recruitment by the peripheral administration of a CCR2 antagonist results in a reduction in local inflammation, suggesting that the MCP-1/CCR2 axis may also be involved in the chemotaxis of peripheral myeloid cells seen in HFD-induced hypothalamic inflammation. Curiously, according to the scRNA-seq data from ARC and ME published by Campbell et al. [108], MCP-1 expression in these areas is mainly observed in microglia. Hence, upon high-fat feeding, neurons could be the main cells involved in fractalkine production, while astrocytes and microglia are responsible for CCL2 (MCP-1) release. Based on these data, activated astrocytes not only directly modulate microglial activity, through the increased synthesis of pro-inflammatory cytokines and growth factors, but also mediate and stimulate the recruitment of peripheral macrophages to the hypothalamus, thus contributing to sustained microglial activation.

A leaky BBB in the hypothalamic area has been observed in the HFD-induced inflammatory response [109]. The mechanism underlying this BBB disruption may be due to various factors, such as altered function/structure of non-fenestrated brain endothelial cells, tanycytes, pericytes, neurons, or glia [11]. These cells constitute the neurovascular unit and confer integrity to the BBB under physiological conditions [110]. However, in response to some inflammatory stimuli, astrocytes and tanycytes increase the secretion of vascular endothelial growth factor (VEGF), increasing BBB permeability [110,111]. It is not well established whether this mechanism is crucial for the entrance of FFAs and peripheral cells into the hypothalamic parenchyma upon chronic HFD intake. Therefore, Lee et al. [12] showed that PVMs in the hypothalamic area produce inducible nitric oxide synthase (iNOS) abundantly under high-fat feeding, resulting in disruption of BBB integrity and in the spread of monocyte-derived macrophages in the ARC. As expected, specific hypothalamic macrophage iNOS inhibition completely abrogates astrocytic lipid droplets and macrophage accumulation and activation in the ARC of obese mice.

Another possible mechanism by which microglia and astrocytes contribute to the leaky BBB – observed in several inflammatory diseases – is the downregulation of tight-junction proteins, namely claudin-5 (CLDN5), occludin, and zonula occludens-1, in activated glial cells [112,113]. Interestingly, under systemic inflammation, vessel-associated microglia are able to phagocytize astrocytic end-feet, an action that damages BBB function [114]. However, how hypothalamic microglial and astrocytic crosstalk impairs BBB integrity under HFD-induced inflammation has been minimally explored.

Unsurprising, upon hypothalamic inflammation, microglial activity can also modulate astrocytic functions. Studies using co-culture of microglia and astrocytes have shown that microglial production of pro-inflammatory cytokines increases astrocytic glucose uptake, thus reducing intercellular glucose trafficking [115] and inhibiting astrocytic gap junctions [116,117]. Phenotypic alterations observed both in microglia and astrocytes at the onset of the inflammatory response are mediated by metabolic changes, switching from mitochondrial oxidative phosphorylation to glycolysis [118]. An important mechanism involved in HFD-induced hypothalamic inflammation in rodents comprises the increased levels of microglial UCP2 levels after HFD consumption, which drives a disruption in microglial mitochondrial dynamics [119]. These authors found that UCP2 deletion specifically in microglia reduced the number of GFAP+ cells in the ARC of mice fed a HFD, suggesting that mitochondrial dynamics in microglia regulate astrogliosis in HFD-induced hypothalamic inflammation.

The schematic representation of the crosstalk between microglia and astrocytes under homeostasis and HFD-induced hypothalamic inflammation is shown in Figure 2.

Over the past decade, accumulating evidence has demonstrated that astrocytes also display high heterogeneity. These cells present many subpopulations spread throughout the CNS, which can vary depending on age, species, and sex, and astrocytic state (surveillance or reactive) [120–122]. Given their complexity, a consensus statement about reactive astrocytes [123] recently recommended that astrocyte phenotypes should be defined by a combination of molecular markers and functional readouts. There is evidence that heterogeneity of astrocytes and microglial hallmarks expression can be bi-directionally controlled [124]. Although understanding how different glial subpopulations regulate their local niche under HFD-induced hypothalamic inflammation still need to be investigated. Future studies regarding single-cell transcriptomic heterogeneity of hypothalamic astrocytes and microglia will elucidate which of their sub-populations are paired, their features, and how their hallmarks change upon HFD-induced reactivity.

**Figure 2.** Microglia and astrocyte crosstalk in the hypothalamus. In homeostatic conditions: **(1)** microglia and astrocytes sense nutrients (amino acids, glucose, and free fatty acids [FFAs], among others) and hormones (e.g., leptin) that enter the hypothalamic parenchyma and react to those environmental changes. Astrocytes regulate lipid metabolism and storage and trigger metabolic changes due to hormone binding to their surface receptors (e.g., leptin and glucose receptors). **(2)** UCP2 expression in microglia is tightly controlled preserving their mitochondrial function while PVMs release normal levels of iNOS in response to FFAs concentration in blood vessels. **(3)** Microglia and astrocytes interact via contact-dependent and secreted factors and **(4)** control synapse development and function. Under high-fat feeding: **(5)** Glial cells quickly sense environmental changes and become reactive, releasing several pro-inflammatory cytokines and growth factors (e.g., VEGF and TGF-β1), and increasing reactive oxygen species (ROS) production. Astrocytes store the excessive FFAs in lipid droplets while microglial UCP2 increases, promoting astrogliosis and mitochondrial dysfunction in microglia. **(6)** PVMs surrounding blood vessels increase the release of iNOS promoting leaking BBB and facilitating the entrance of FFAs into the hypothalamic parenchyma, **(7)** which can stimulate IKKβ/NF-κB signaling by activating TLR4 and TREM2 in microglial cells. **(8)** Glial cells and neurons increase the synthesis of chemokines (e.g., MCP-1 and fractalkine), leading to the recruitment of peripheral myeloid cells to the central nervous sys-**Figure 2.** Microglia and astrocyte crosstalk in the hypothalamus. In homeostatic conditions: (**1**) microglia and astrocytes sense nutrients (amino acids, glucose, and free fatty acids [FFAs], among others) and hormones (e.g., leptin) that enter the hypothalamic parenchyma and react to those environmental changes. Astrocytes regulate lipid metabolism and storage and trigger metabolic changes due to hormone binding to their surface receptors (e.g., leptin and glucose receptors). (**2**) UCP2 expression in microglia is tightly controlled preserving their mitochondrial function while PVMs release normal levels of iNOS in response to FFAs concentration in blood vessels. (**3**) Microglia and astrocytes interact via contact-dependent and secreted factors and (**4**) control synapse development and function. Under high-fat feeding: (**5**) Glial cells quickly sense environmental changes and become reactive, releasing several pro-inflammatory cytokines and growth factors (e.g., VEGF and TGF-β1), and increasing reactive oxygen species (ROS) production. Astrocytes store the excessive FFAs in lipid droplets while microglial UCP2 increases, promoting astrogliosis and mitochondrial dysfunction in microglia. (**6**) PVMs surrounding blood vessels increase the release of iNOS promoting leaking BBB and facilitating the entrance of FFAs into the hypothalamic parenchyma, (**7**) which can stimulate IKKβ/NF-κB signaling by activating TLR4 and TREM2 in microglial cells. (**8**) Glial cells and neurons increase the synthesis of chemokines (e.g., MCP-1 and fractalkine), leading to the recruitment of peripheral myeloid cells to the central nervous system. If the inflammatory stimulus persists, the astrocytic control of synaptic functions become impaired, resulting in neuronal dysfunction. Abbreviations: 3V, third ventricle; MBH, mediobasal hypothalamus.

#### tem. If the inflammatory stimulus persists, the astrocytic control of synaptic functions become impaired, resulting in neuronal dysfunction. Abbreviations: 3V, third ventricle; MBH, mediobasal **4. Advances in Experimental Manipulation of Microglia**

hypothalamus. **4. Advances in Experimental Manipulation of Microglia** The identification of the microglial transcriptomic signature and heterogeneity has given rise to the development of new experimental models for manipulating or labelling specific subtypes of microglia. The hypothalamus has many microglial subpopulations The identification of the microglial transcriptomic signature and heterogeneity has given rise to the development of new experimental models for manipulating or labelling specific subtypes of microglia. The hypothalamus has many microglial subpopulations that distinctly react to hormones and nutrients from the diet. Thus, the availability of these new models can provide a significant advance to comprehend the involvement of each microglial subtype in the HFD-induced hypothalamic inflammation and other

that distinctly react to hormones and nutrients from the diet. Thus, the availability of these

inflammatory or neurodegenerative diseases. To achieve that, researchers should be able to choose the best model to answer their specific questions. In this topic, we discuss the newest approaches that have been developed.

#### *4.1. Ex Vivo Models*

To understand how microglia react to different stimuli, most studies conducted in the past decades have employed ex vivo strategies. A large variety of mouse microglia cell lines have been generated over the years, but the most used has been the murine microglial BV2 cell line. These cells were originally obtained and immortalized from the cerebral cortex of neonatal mice [125]. Other similar cell lines have been obtained from the whole brain, cerebellum, or cortex of adult/embryo rodents and have been less applied in basic research and barely cited in the literature compared with BV2 [126].

Most studies investigating microglial hypothalamic features using ex vivo methods have been conducted with BV2 cells [127–129]. It is well known that there are many phenotypic and genomic differences between microglia from the cortex and from the hypothalamus or microglia from a neonatal mouse compared to an adult or embryo – not to mention the fact that these cells were immortalized 30 years ago, and have very likely already suffered several transcriptomic modifications since that time. Another concern that must be considered when interpreting data is whether these cells are at the surveillance or reactive state when cultivated with no other neuronal or glial cells. Many questions and worries emerged when microglia stopped being seen as a simple macrophage of the CNS. Hence, to reduce any possible bias, published data and ongoing experiments with BV2 cells should be carefully planned, analyzed, and always accompanied by other in vivo strategies.

The development of new transgenic mouse models that express a fluorescent protein driven by microglial hallmarks promises an appealing alternative for researchers interested in cultured microglia. Magnetic-activated cell sorting (MACS) [130] and fluorescenceactivated cell sorting (FACS) [131,132] using these fluorescent reporter mice or even microglia-specific surface markers allow the physical isolation of various subsets of microglia, which can be obtained from primary cultures or co-cultures with other cells. Using these approaches, the brain area of interest can be precisely harvested after a specific dietary or treatment protocol, and the researcher can freely choose the age and the sex of the rodents used in the experiments. These strategies render the collected data much more reliable when compared to studies using only BV2 cells. Although, when choosing cell sorter techniques researchers should keep in mind that any steps of protocols can trigger some reaction in microglia morphology or gene expression and even impair the cell viability (Figure 3).

Grassivaro et al. [133] recently applied FACS to isolate CNS resident microglia and peripheral myeloid cells from some models of neuroinflammation, such as experimental autoimmune encephalomyelitis (EAE). These authors collected samples during embryonic and postnatal periods and subsequently targeted cells from the brain and liver through labelling microglia and peripheral myeloid cells with fluorescent antibodies. By this approach, they observed extraordinary differences and transcriptomic details between those cells. Unfortunately, there is still a lack of data obtained through FACS or MACS for isolating microglia from the hypothalamus of experimental models of obesity.

**Figure 3.** Comparisons between ex vivo experimental models for studying cultured microglia. Pros (in blue) and Cons (in red) of each approach are depicted in figure. Abbreviations: FACS, fluorescence-associated cell sorting; MACS, magnetic-activated cell sorting. **Figure 3.** Comparisons between ex vivo experimental models for studying cultured microglia. Pros (in blue) and Cons (in red) of each approach are depicted in figure. Abbreviations: FACS, fluorescence-associated cell sorting; MACS, magnetic-activated cell sorting.

#### *4.2. In Vivo Models 4.2. In Vivo Models*

Most studies need to evaluate microglial functions in vivo. Conveniently, at least in rodents, functional contributions of specific cell populations can be explored using Cre recombinase–mediated mutagenesis. The identification and cloning of CX3CR1 triggered the development of new tools for studying microglia in vivo [134–136]. Studies regarding the mechanisms through microglia-mediated HFD-induced hypothalamic inflammation have been mostly conducted with a transgenic mouse known as CX3CR1-Cre, which expresses Cre recombinase under the direction of the *Cx3cr1* promoter in the mononuclear phagocyte system, and CX3CR1EGFP knock-in/knock-out reporter mice, which expresses a green fluorescent protein (EGFP) in monocytes, dendritic cells, natural killer cells, and brain microglia under the control of the same locus [20,137]. Consequently, all mechanisms regarding microglial reactivity and gliosis in those research articles can only be interpreted as adaptations and features of CX3CR1+ microglia. Thus, when researchers activate/inhibit microglia in CX3CR1-Cre mice through DREADD, optogenetic, or other Cre recombinase-targeted manipulation, they can only interpret the obtained data in the context of the manipulation of CX3CR1+ microglial, and not to other cells. Most studies need to evaluate microglial functions in vivo. Conveniently, at least in rodents, functional contributions of specific cell populations can be explored using Cre recombinase–mediated mutagenesis. The identification and cloning of CX3CR1 triggered the development of new tools for studying microglia in vivo [134–136]. Studies regarding the mechanisms through microglia-mediated HFD-induced hypothalamic inflammation have been mostly conducted with a transgenic mouse known as CX3CR1-Cre, which expresses Cre recombinase under the direction of the *Cx3cr1* promoter in the mononuclear phagocyte system, and CX3CR1EGFP knock-in/knock-out reporter mice, which expresses a green fluorescent protein (EGFP) in monocytes, dendritic cells, natural killer cells, and brain microglia under the control of the same locus [20,137]. Consequently, all mechanisms regarding microglial reactivity and gliosis in those research articles can only be interpreted as adaptations and features of CX3CR1+ microglia. Thus, when researchers activate/inhibit microglia in CX3CR1-Cre mice through DREADD, optogenetic, or other Cre recombinasetargeted manipulation, they can only interpret the obtained data in the context of the manipulation of CX3CR1+ microglial, and not to other cells.

CX3XR1-CreERT2 is another *Cx3xr1*-driven model widely used in experimental research since it expresses Cre recombinase in a tamoxifen-inducible manner [138–140]. This tamoxifen-inducible model allows the temporal manipulation of microglia. Therefore, Van Hove et al. [141] have recently warned that the use of this strain is not indicated for all types of microglia-related investigations, such as fate-mapping studies, once it exhibits considerable leakiness in the absence of tamoxifen. CX3XR1-CreERT2 is another *Cx3xr1*-driven model widely used in experimental research since it expresses Cre recombinase in a tamoxifen-inducible manner [138–140]. This tamoxifen-inducible model allows the temporal manipulation of microglia. Therefore, Van Hove et al. [141] have recently warned that the use of this strain is not indicated for all types of microglia-related investigations, such as fate-mapping studies, once it exhibits considerable leakiness in the absence of tamoxifen.

For many years, *Cx3cr1*-driven lines were largely employed in basic research – and they still are. However, they are haploinsufficient for *Cx3cr1*, which means that EGFP or Cre coding regions have been designed to replace the endogenous *Cx3cr1* locus. Some studies suggest that this haploinsufficiency could affect microglial functions, such as synaptic plasticity [142–144]. Given this background, all Cx3cr1-drive lines should be used with this limitation in mind. It is fortunate that new transgenic mouse models for genetic labelling and manipulating other microglial subtypes, myeloid cells and border-associated macrophages (BAMs) found in the hypothalamus and in other brain areas have been For many years, *Cx3cr1*-driven lines were largely employed in basic research – and they still are. However, they are haploinsufficient for *Cx3cr1*, which means that EGFP or Cre coding regions have been designed to replace the endogenous *Cx3cr1* locus. Some studies suggest that this haploinsufficiency could affect microglial functions, such as synaptic plasticity [142–144]. Given this background, all Cx3cr1-drive lines should be used with this limitation in mind. It is fortunate that new transgenic mouse models for genetic labelling and manipulating other microglial subtypes, myeloid cells and border-associated macrophages (BAMs) found in the hypothalamus and in other brain areas have been

generated in recent years, allowing a deep investigation of new microglial phenotypes and functions (Table 1).

**Table 1.** Novel in vivo experimental models for studying various subtypes of microglia, myeloid cells, and BAMS.


Masuda et al. [150] recently developed another of these new microglia gene-targeting models. For this endeavor, the authors initially analyzed microglia and CNS-associated macrophage (CAM) signatures obtained from scRNA-seq, during homeostasis and disease, and identified beta-hexosaminidase subunit beta (*Hexb*) as a stably expressed microglia core gene. Based on that, they generated HexbTdTomato and Hexb-CreERT2 mouse strains, which express a red fluorescent protein (TdTomato) or Cre recombinase protein under the control of the *Hexb* locus upon tamoxifen administration, respectively. Although this research article was published a few months ago, there is no data available in the literature about the participation of microglial cells Hexb+ in hypothalamic inflammation. Curiously, *Hexb* is highly expressed in the ARC and ME [108], but in these areas, it is not tightly restricted to microglial when compared with other markers that can also be manipulated through transgenic mouse strains (*Cx3cr1*, *P2ry12*, or *Tmem119*) (Figure 4). Nevertheless, future studies should evaluate whether these Hexb+ microglia are important in some stage of HFD-induced hypothalamic inflammation and characterize their interactions with other glial and non-glial cells in the hypothalamus.

Transmembrane protein 119 (*Tmem119*) and purinergic receptor P2Y12 (*P2ry12*) loci have also been targets for the development of new mouse models (Table 1). Kaiser and Feng [147] generated Tmem119EGFP mice and Tmem119-CreERT2 mice, which express EGFP or CreERT2 under the control of the Tmem119 coding region, respectively. More recently, Ruan et al. [149] generated another reporter mouse, Tmem119TdTomato, which expresses TdTomato rather than EGFP in microglia. All these Tmem119-driven lines have been validated and are now valuable tools to study specifically the role of Tmem119+ microglia in health and disease. McKinsey et al. [148] recently published the new P2ry12-CreERT2 mouse line. As stated by the authors, they chose P2ry12 because it appeared to be the most restricted to brain myeloid cells compared with other markers. Following the pattern, this model expresses Cre protein recombinase under the control of the P2ry12 locus. They also suggested that this model could be useful for studies about middle cerebral artery occlusion–induced ischemic stroke as well as EAE.

As previously mentioned, Valdearcos et al. [20] identified that both Tmem119+ and P2ry12+ cells react differently to Cx3cr1+ or Iba1+ cells in the ARC and the VMH under prolonged HFD intake. Thus, these novel *Tmem119*- and *P2ry12*-driven strains can also be promising to comprehend the mechanisms through which these specific microglial subtypes react to SFAs from the diet in this specific inflammatory condition.

**Figure 4.** Distribution of genes in the arcuate nucleus (ARC) and median eminence (ME) chosen to generate mouse models that target microglia and perivascular macrophages (PVMs). Scatter plot **Figure 4.** Distribution of genes in the arcuate nucleus (ARC) and median eminence (ME) chosen to generate mouse models that target microglia and perivascular macrophages (PVMs). Scatter plot images were obtained from https://singlecell.broadinstitute.org/ accessed on 20 November 2020, ARC-ME DROPSEQ data is from Campbell et al. [108].

Beyond Cre-driven and fluorescent reporter mouse lines, many researchers have chosen to deplete microglia to evaluate their role in some pathologic condition. There are many microglial depletion strategies, which were recently remarkably reviewed by Miron and Priller [151], including the use of liposomes, global knockout of genes required for microglial development or survival, or even transgenic or pharmacological induction of microglial death. Unfortunately, all these methods have some bias and do not allow the depletion of only microglial cells, because BAMs and monocytes may also be targeted. The use of some knockout mice, such as the knockout for colony stimulating factor 1 receptor (*Csfr1*-KO), is also not a good choice, because these mice lack all CNS macrophages; they show serious brain abnormalities and usually die within weeks after birth [152].

Due this dilemma, a novel model has been developed in which microglia are depleted, but BAMs are widely preserved after the genomic deletion of a super enhancer in the colony stimulating factor 1 receptor (*Csf1r*) coding region (fms-intronic regulatory element; Csf1r-FIRE∆/∆) [145]. A few studies have employed microglial depletion to study microglial immune and metabolic functions in the hypothalamus [137,153], because depletion disrupts immune balance and energy homeostasis [154]. Fortunately, Csf1r is also expressed by microglia from the ARC and ME [108] (Figure 4). CSF1R inhibition through a pharmacologic inhibitor (PLX5622) can improve metabolic outcomes in middle-aged female mice [155]. Thus, investigating the role of Csfr1+ microglial cells in the hypothalamic inflammation, through this Csf1r-FIRE∆/<sup>∆</sup> mouse model, could also provide new, important information regarding this subtype of microglia.

As previously described, PVMs are also involved in HFD-induced hypothalamic inflammation [12]. This study was conducted with transgenic mouse models targeting lysozyme M (LysM)-expressing myeloid cells. Although many experimental models have been developed in the past years, there are still some methodological challenges with regard to manipulating specifically microglia or other CAMs/BAMs, such as PVMs. Even the above-mentioned transgenic lines used to study several microglial subsets (Table 1) have the mutation driven by a locus common to microglia and PVMs, at least in the ARC and ME [108] (Figure 4).

Some studies using scRNA-seq have already analyzed these cells separately [71], through dissection of leptomeninges, the perivascular space and parenchyma, and the choroid plexus; however, experimental manipulation of microglia and other CAMs/BAMs independently remains a challenge. Fortunately, Kim et al. [146] have just developed a binary transgenic system involving complementation-competent NCre and CCre fragments. Their expression is driven by two distinct promoters: Sall1ncre:Cx3cr1ccre, which specifically targets parenchymal microglia, and Lyve1ncre: Cx3cr1ccre, which allows the ability to target various CAMs throughout the brain. These new models are very promising and were developed using CRISPR/Cas9 technology. According to the authors, a CCre cassette was inserted after the *Cx3cr1* gene, and Ncre partner transgenes were inserted into *Sall1* and *Lyve1* loci, respectively. In fact, *Lyve1* is scarcely expressed by microglia, at least in the ARC and ME [108] (Figure 4). Hence, both models may be useful to study separately the functions of microglia and CAMs in HFD-induced inflammation, because the knowledge about the role of each glial cell type in this specific condition still need to be better clarified.

#### **5. Microglial Signature Changes Upon HFD-Induced Hypothalamic Inflammation**

There are sufficient clues indicating that microglia present distinct activated signatures under inflammatory conditions. Sousa et al. [72] performed scRNA-seq to investigate the microglial profile in the brain of LPS-injected mice. They found that microglia isolated from these mice exhibit a downregulation of their homeostatic signature together with an upregulation of inflammatory genes. They obtained this data by excluding other CNSresident immune cells and peripheral cells in the analysis.

To explore deeply how HFD-induced inflammation affects microglial transcriptomics, we have searched for transcriptomic data published from HFD-fed rodents at a single-cell resolution. As already mentioned, Campbell et al. [108] performed Drop-seq profiling on

the ARC and ME from mice across different feeding conditions, including 1-week HFD (60% calories from fat). They found that microglia present high *P2ry12* and low *Mrc1* while PVMs show low *P2ry12* and high *Mcr1*. However, when comparing low fat diet–fed mice with HFD-fed mice, they found the transcripts from these clusters downregulated, but they did not describe which HFD-sensitive genes were modulated. on the ARC and ME from mice across different feeding conditions, including 1-week HFD (60% calories from fat). They found that microglia present high *P2ry12* and low *Mrc1* while PVMs show low *P2ry12* and high *Mcr1*. However, when comparing low fat diet–fed mice with HFD-fed mice, they found the transcripts from these clusters downregulated, but they did not describe which HFD-sensitive genes were modulated.

binary transgenic system involving complementation-competent NCre and CCre fragments. Their expression is driven by two distinct promoters: Sall1ncre:Cx3cr1ccre, which specifically targets parenchymal microglia, and Lyve1ncre: Cx3cr1ccre, which allows the ability to target various CAMs throughout the brain. These new models are very promising and were developed using CRISPR/Cas9 technology. According to the authors, a CCre cassette was inserted after the *Cx3cr1* gene, and Ncre partner transgenes were inserted into *Sall1* and *Lyve1* loci, respectively. In fact, *Lyve1* is scarcely expressed by microglia, at least in the ARC and ME [108] (Figure 4). Hence, both models may be useful to study separately the functions of microglia and CAMs in HFD-induced inflammation, because the knowledge about the role of each glial cell type in this specific condition still need to be better clarified.

**5. Microglial Signature Changes Upon HFD-Induced Hypothalamic Inflammation**

There are sufficient clues indicating that microglia present distinct activated signatures under inflammatory conditions. Sousa et al. [72] performed scRNA-seq to investigate the microglial profile in the brain of LPS-injected mice. They found that microglia isolated from these mice exhibit a downregulation of their homeostatic signature together with an upregulation of inflammatory genes. They obtained this data by excluding other

To explore deeply how HFD-induced inflammation affects microglial transcriptomics, we have searched for transcriptomic data published from HFD-fed rodents at a singlecell resolution. As already mentioned, Campbell et al. [108] performed Drop-seq profiling

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 15 of 24

CNS-resident immune cells and peripheral cells in the analysis.

On the other hand, in an experimental model of prolonged high-fat feeding, C57BL/6J male mice were fed a HFD (45% calories from fat) for 10 weeks, and their ARC was collected for single-nucleus RNA-seq (snRNA-seq) [156]. After consuming a HFD for 10-weeks, some genes were upregulated in the cluster of microglia (*Nudt5*, *Gsk3a*, *Oxt*, *Lars2*, and *Il17ra*) while other genes (*Sun1*, *Tmem173* and *Anxa3*) were downregulated (Figure 5). On the other hand, in an experimental model of prolonged high-fat feeding, C57BL/6J male mice were fed a HFD (45% calories from fat) for 10 weeks, and their ARC was collected for single-nucleus RNA-seq (snRNA-seq) [156]. After consuming a HFD for 10 weeks, some genes were upregulated in the cluster of microglia (*Nudt5*, *Gsk3a*, *Oxt*, *Lars2*, and *Il17ra*) while other genes (*Sun1*, *Tmem173* and *Anxa3*) were downregulated (Figure 5).

Some of them have also been described in obesity-related studies. Methylation in *Lars2*, which is a mitochondrial gene, has been reported in an epigenome-wide association study as a gene associated with increased body mass index and waist circumference [157]. The receptor for interleukin-17 (IL-17), known as *Il17ra*, is found in pro-opiomelanocortin (POMC) and agouti-related peptide (AgRP) neurons in the hypothalamus and modulates food intake after IL-17 binding, without affecting whole-body energy expenditure [158].

Although the ARC is the main nucleus involved in energy metabolism and food intake control, functional impairments in neurons and glial cells from other hypothalamic nuclei are also involved in DIO establishment [159,160]. Recently, Rossi et al. [161] identified in the lateral hypothalamic area (LHA) thousands of genes altered upon prolonged HFD intake. This research article highlighted only dynamic transcriptomic details in neurons; thus, the authors did not deeply explore the transcription differences in microglia and other myeloid cell lineages upon. However, the authors distinctively represent these cells by using *Cx3cr1* hallmark for microglia and *Lys2* for myeloid cells, which was not considered in data from ARC published by Deng et al. [156].

Interestingly, from all transcriptomic data obtained from the ARC–ME or LHA, researchers can apply a bioinformatic analysis to better investigate how microglia and other

myeloid cells changes upon prolonged HFD. In the face of the recent and continuous development of novel transcriptomic tools, such as scRNA-seq, snRNA-seq, and CyTOF, among others, future studies should be conducted to better clarify the main changes in microglial signature throughout the hypothalamus under different stages of high-fat feeding. For accurate data interpretation, microglia and other CNS-associated macrophages should be clustered and analyzed separately. The detailed identification of those cells will be valuable to answer more precisely several outstanding questions.

#### **6. Conclusions**

Microglia were first recognized as macrophage-like cells from the CNS a century ago, but for a long time their complexity was unknown. Luckily, the development of assorted transcriptomic tools has boosted the knowledge about these cells in recent years. Currently, it is well known that the hypothalamus presents several microglial subsets that can be identified by their hallmarks: *Iba1*, *Cx3Cr1*, *Tmem119*, *P2ry12*, *Trem2*, *Hexb*, and *Csfr1*, among many other. DM are also found in the hypothalamic area, but their roles in HFD-induced inflammation has been poorly investigated. In fact, microglia play a pivotal role in different stages of the hypothalamic inflammatory process, but how each microglial subtype reacts to SFAs from the diet, communicates with other cells, or even leads to the recruitment of peripheral myeloid cells remains to be explored. Beyond microglia, astrocytes also display high heterogeneity throughout the CNS, and in the hypothalamus, its diversity and its crosstalk with microglia require to be better elucidated. Novel experimental models for manipulating or labelling microglia have been developed and will be useful to answer that question in forthcoming research. On the other hand, the hypothalamic microglial signature under HFD-induced hypothalamic inflammation should still be further studied using different transcriptomic approaches. Together, these advances will allow researchers to take full advantage of crucial insights we have gained about microglial heterogeneity (provided by transcriptomic data) and to exploit this knowledge to determine the mechanisms by which microglia are involved in the inflammatory process observed in obesity.

**Author Contributions:** Conceptualization, N.F.M. and E.P.d.A.; investigation, N.F.M., C.P.J., and A.M.Z.; writing—original draft preparation, N.F.M., C.P.J., and A.M.Z.; writing—review and editing, N.F.M.; supervision, E.P.d.A.; funding acquisition, N.F.M., C.P.J., A.M.Z., and E.P.d.A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Sao Paulo Research Foundation (FAPESP: 2013/07607-8, 2016/17810-3, 2017/22511-8, and 2019/00330-7) and by the Coordination for the Improvement of Higher Education Personnel (CNPq).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** Figures 1–3 and 5 were created with BioRender.com and Figure 4 was created with Corel Draw Graphics Suite 2020.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**



#### **References**


## *Review* **Adaptive Changes in the Central Control of Energy Homeostasis Occur in Response to Variations in Energy Status**

**Cassandra Gastelum 1,†, Lynnea Perez 1,†, Jennifer Hernandez <sup>1</sup> , Nikki Le <sup>1</sup> , Isabella Vahrson <sup>1</sup> , Sarah Sayers <sup>1</sup> and Edward J. Wagner 1,2,\***


**Abstract:** Energy homeostasis is regulated in coordinate fashion by the brain-gut axis, the homeostatic energy balance circuitry in the hypothalamus and the hedonic energy balance circuitry comprising the mesolimbcortical A<sup>10</sup> dopamine pathway. Collectively, these systems convey and integrate information regarding nutrient status and the rewarding properties of ingested food, and formulate it into a behavioral response that attempts to balance fluctuations in consumption and food-seeking behavior. In this review we start with a functional overview of the homeostatic and hedonic energy balance circuitries; identifying the salient neural, hormonal and humoral components involved. We then delve into how the function of these circuits differs in males and females. Finally, we turn our attention to the ever-emerging roles of nociceptin/orphanin FQ (N/OFQ) and pituitary adenylate cyclase-activating polypeptide (PACAP)—two neuropeptides that have garnered increased recognition for their regulatory impact in energy homeostasis—to further probe how the imposed regulation of energy balance circuitry by these peptides is affected by sex and altered under positive (e.g., obesity) and negative (e.g., fasting) energy balance states. It is hoped that this work will impart a newfound appreciation for the intricate regulatory processes that govern energy homeostasis, as well as how recent insights into the N/OFQ and PACAP systems can be leveraged in the treatment of conditions ranging from obesity to anorexia.

**Keywords:** sex difference; estradiol; nociceptin/orphanin FQ; pituitary adenylate cyclase-activating polypeptide; obesity; fasting

### **1. The Hypothalamic Energy Balance Circuit in Homeostatic Feeding**

Energy homeostasis, the intricate balance between energy intake and expenditure, is regulated in coordinate fashion by homeostatic and hedonic neural circuits [1]. Aberrations in these circuits are implicated in the pathophysiology of conditions such as obesity, type-II diabetes, and food addiction [2–5]. Homeostatic control of energy balance is attributed to the hypothalamic energy balance neural circuitry, which integrates information relayed from the brainstem regarding the energy/nutritional status of the organism, based on chemical and mechanical cues from the gastrointestinal (GI) tract [6,7]. These communications are ultimately encoded by various hypothalamic nuclei and their associated neuronal populations, including those in the arcuate nucleus (ARC), ventromedial nucleus (VMN), lateral hypothalamus (LH), dorsomedial nucleus (DMN), and the paraventricular nucleus (PVN); producing orexigenic and anorexigenic signals to, respectively, stimulate and suppress energy intake as well as altering energy expenditure [1,8].

**Citation:** Gastelum, C.; Perez, L.; Hernandez, J.; Le, N.; Vahrson, I.; Sayers, S.; Wagner, E.J. Adaptive Changes in the Central Control of Energy Homeostasis Occur in Response to Variations in Energy Status. *Int. J. Mol. Sci.* **2021**, *22*, 2728. https://doi.org/10.3390/ijms22052728

Academic Editor: Rosalía Rodríguez-Rodríguez

Received: 5 February 2021 Accepted: 4 March 2021 Published: 8 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Excitatory input from steroidogenic factor (SF-1) neurons, located in the VMN, impinging on proopiomelanocortin (POMC) neurons in the ARC represent a critical anorexigenic synapse in homeostatic energy balance that, when activated, suppresses energy intake and enhances energy expenditure [5,8–12]. SF-1 is a transcription factor encoded by the NR5A1 gene, and activation of these neurons in the VMN leads to glutamatergic stimulation of POMC neurons [2,5,8–11]. Activation of POMC expressing neurons within the hypothalamic melanocortin system ultimately leads to formation of POMC posttranslational cleavage products such as α-melanocortin stimulating hormone (α-MSH), which following release from axon terminals go on to bind downstream effectors like melanocortin 4 receptors (MC4R) expressed on corticotropin-releasing hormone (CRH) neurons in the PVN [13–17]. In line with α-MSH functioning as a satiety mediator, the percentage of α-MSH neurons colocalizing with c-Fos in the ARC is greatest at the end of a meal, compared to the beginning of or hours after consumption [18]. In addition to α-MSH, POMC neurons also release β-endorphin (following posttranslational modification), and co-express cocaineand amphetamine- regulated transcript (CART) [19,20]. Overall, POMC signaling reduces food intake, increases energy expenditure, and regulates glucose metabolism [16,17,19,21]. Prevailing glucose concentrations play a key role in relaying nutrition/energy state cues to the homeostatic energy balance circuity, where anorexigenic and orexigenic ARC neurons have glucose sensing abilities. ARC POMC/CART neurons are categorized as glucoseresponsive neurons and take up glucose via a glucose transporter (GLUT2) where it then metabolizes, producing ATP and thereby promoting the closure of ATP-dependent potassium (KATP) channels to reduce the outflow of K<sup>+</sup> , ultimately leading to the depolarization of the cell [22,23]. This aligns POMC cellular excitability and firing rate in direct proportion to glucose concentrations, with satiety signaling accentuated as glucose levels rise (e.g., during or shortly following a meal) [22,23]. It should therefore not be surprising that perturbations in signaling at this VMN SF-1/ARC POMC synapse can pose detrimental consequences for energy balance, where null mutations in POMC, its cleavage enzymes or downstream receptors, as well as lesioning in the VMN, ultimately leads to hyperphagia and obese phenotypes in rodents and humans [16,17,24].

At the opposing end of hypothalamic energy balance control are neuronal populations that promote orexigenic or appetite-stimulating effects. Alongside POMC neurons, the ARC houses neuropeptide Y (NPY) and agouti-related peptide (AgRP) co-expressing neurons, as well as ghrelin-containing somata [25–27]. Activation of this subset of neurons pleiotropically dampens the aforementioned anorexigenic signaling. For example, following NYP/AgRP neuronal activation, direct inhibition of neighboring POMC neurons is achieved via synaptic release of the inhibitory amino acid neurotransmitter γ-aminobutyric acid (GABA) on POMC soma [28]. Further downstream modulation of POMC signaling is mediated by AgRP antagonism on MC4R or via NPY acting on various receptor subtypes, ultimately impeding the anorexigenic signaling of α-MSH to induce feeding and reduce energy expenditure [15,23]. Additionally, NPY/AgRP neurons in the ARC are glucose-sensitive neurons and their activity/firing rate is inversely proportional to ambient glucose levels [23]. Neurons found within the LH also relay orexigenic signals, which are mediated by neurotransmission of melanin-concentrating hormone (MCH) neurons and orexin neurons [29,30]. Indeed, orexin has been shown to electrically silence POMC neurons by enhancing GABAergic and diminishing glutamatergic inputs onto these cells [31]. In addition, endocannabinoids elicit their orexigenic effects in part through retrograde inhibition of GABAergic inputs onto MCH neurons [32].

In addition to the ascending inputs from the brainstem, the neurons comprising the hypothalamic energy balance circuit are also susceptible to the influence of circulating peripheral hormones—leptin, ghrelin, insulin and sex hormones. Sensitivity to these transient peripheral hormones (and other periphery signaling molecules) is especially true for ARC neuronal populations, as this region lies in close proximity to the third ventricle (3V) and median eminence; a circumventricular organ with an incomplete blood–brain barrier [33,34]. Therefore, the ARC is in direct contact with systemic circulating hormones, permitting critical communication about the energy/nutritional status of the body to ARC neurons that promotes energy homeostasis. Leptin is synthesized by white adipose tissue (WAT), with levels fluctuating in proportion to fat mass. Leptin acts as a potent suppressor of food intake, while also stimulating metabolism and reducing excessive stored energy [35]. POMC and SF-1 neurons are depolarized by leptin receptor (LEPR) activation via JAK/signal-transducer-and-activator-of-transcription (STAT) and phosphatidylinositol-4,5-bisphosphate 3-kinase (PI3K) pathway signaling [36–39], while LEPR activity on AgRP neurons activates ATP-gated K<sup>+</sup> (KATP) channels leading to potassium outflow, hyperpolarization and decreases in firing [40–42]. The anorexigenic effect of leptin also involves a reduction in hypothalamic endocannabinoid levels [43]. Conversely, hypothalamic levels of endocannabinoids are increased in leptin receptor-deficient *fa/fa* rats and leptin-deficient *ob/ob* mice [43]. Similar to leptin, insulin signaling in the hypothalamus promotes an anorexigenic tone and elicits transient receptor potential (TRP)C5 channel-induced excitation of POMC neurons following receptor activation [44]. While for insulin, several prior studies reported inhibitory responses in POMC neurons due to activation of KATP channels, it is now known that the proportion of excitatory vs. inhibitory insulin-induced responses is dependent on ambient levels of protein tyrosine phosphatase 1B (PTP1B) and T-cell protein tyrosine phosphatase (TCPTP) activity [45–47]. Both enzymes are expressed in ARC neurons integral to the regulation of energy balance. PTP1B and TCPTP are key regulators of cell metabolism, as PTP1B decreases leptin activity and TCPTP attenuates insulin signaling via dephosphorylation of JAK2 tyrosine kinase and the insulin receptor, respectively, in POMC neurons, whereas in NPY/AgRP neurons TCPTP attenuates insulin but not leptin signaling [48–50]. Deletion of both enzymes in POMC neurons, or of TCPTP in NPY/AgRP neurons, from obese mice results in promoted weight loss due to decreased food consumption and increased WAT browning along with elevated uncoupling protein (UCP)-1 expression due to enhanced leptin and insulin signaling [48,49].

Conversely, levels of the orexigenic gut-derived peptide ghrelin increase in response to negative energy balance. As a result, the peptide binds to receptors on ARC NPY/AgRP neurons to stimulate these neurons, and thereby promote feeding behavior and energy storage [35]. The orexigenic effect of ghrelin also depends on enhanced hypothalamic production of endocannabinoids and activation of the energy-sensing signaling molecule AMP-dependent protein kinase (AMPK) that, in turn, elicits retrograde inhibition of excitatory input impinging on parvocellular neurons in the PVN [51]. These actions require functionally intact cannabinoid CB1 and ghrelin receptor systems [51,52]. AMPK functions as an important signaling molecule within the hypothalamic energy balance circuitry, and exists as a heterotrimeric complex comprising α-, β-, and γ-subunits [53–56]. The α subunit is the catalytic subunit, while the β and γ subunits are involved in glycogen and AMP/ATP binding, respectively [53–56]. AMPK activity is stimulated by phosphorylation via two upstream kinases, human tumor suppressor LKB1 or Ca2+/calmodulin-dependent protein kinase kinase-β (CaMKKβ) that is triggered by cellular stress, cytokines, hormones like those mentioned above, as well as by increases in AMP/ATP ratio [53–56]. This in turn inhibits anabolic pathways and activates catabolic pathways to generate ATP [53–56]. Leptin, insulin, α-MSH, high plasma glucose concentrations and refeeding all inhibit AMPK whereas AgRP increases it [57]. On the other hand, AMPK activation is necessary for leptin to inhibit fatty acid synthesis and thereby promote fatty acid oxidation in skeletal muscle via phosphorylation of acetyl coenzyme A carboxylase [58]. This is consistent with other examples of opposing cannabinoid- and ghrelin-induced effects on AMPK activity in the central nervous system (CNS) versus the periphery [59]. When the effects of AMPK are constitutively manifest in transgenic mice bearing a gain-of-function mutation in the γ2 subunit, this brings about ghrelin-dependent hyperphagia that leads to obesity, glucose intolerance and hyperinsulinemia [60]. The dynamic interplay between the aforementioned peripheral hormones and the homeostatic energy balance circuitry is graphically depicted in Figure 1.

1.

transgenic mice bearing a gain-of-function mutation in the γ2 subunit, this brings about ghrelin-dependent hyperphagia that leads to obesity, glucose intolerance and

hormones and the homeostatic energy balance circuitry is graphically depicted in Figure

**Figure 1.** Schematic diagram illustrating the interplay between peripheral hormones and the homeostatic energy balance circuitry. Hormones like leptin and insulin released from adipose tissue and the endocrine pancreas, respectively, exert anorexigenic effects, whereas ghrelin released from the gastric mucosa exerts orexigenic effects. Leptin's appetite-suppressing actions are mediated via excitatory effects on anorexigenic POMC and SF-1/PACAP neurons, as well as inhibitory effects on orexigenic NPY/AgRP and N/OFQ neurons. Insulin also inhibits NPY/AgRP neurons and, paradoxically, SF-1/PACAP neurons as well. Insulin's effects on POMC neurons are dependent upon prevailing levels of tyrosine protein phosphatases. On the other hand, ghrelin's appetite-promoting effects are due to its excitatory effects on NPY/AgRP and orexin neurons. **Figure 1.** Schematic diagram illustrating the interplay between peripheral hormones and the homeostatic energy balance circuitry. Hormones like leptin and insulin released from adipose tissue and the endocrine pancreas, respectively, exert anorexigenic effects, whereas ghrelin released from the gastric mucosa exerts orexigenic effects. Leptin's appetitesuppressing actions are mediated via excitatory effects on anorexigenic POMC and SF-1/PACAP neurons, as well as inhibitory effects on orexigenic NPY/AgRP and N/OFQ neurons. Insulin also inhibits NPY/AgRP neurons and, paradoxically, SF-1/PACAP neurons as well. Insulin's effects on POMC neurons are dependent upon prevailing levels of tyrosine protein phosphatases. On the other hand, ghrelin's appetite-promoting effects are due to its excitatory effects on NPY/AgRP and orexin neurons.

#### **2. The Mesolimbic Dopamine Network and Hedonic Feeding Behavior 2. The Mesolimbic Dopamine Network and Hedonic Feeding Behavior**

Alongside the homeostatic-hypothalamic circuitry, energy balance is also modulated by hedonic aspects of feeding behavior pertaining to reward-based food intake, or eating for pleasure. Amounting evidence in rodents and humans now support the theory that both drugs of abuse and the consumption of highly palatable foods converge on a shared pathway within the limbic system to mediate motivated behaviors [61–63]. Therefore, the hedonic consumption of palatable foods involves the mesolimbic dopamine (A10) neurons that emanate from the ventral tegmental area (VTA) and project onto structures such as the nucleus accumbens (NAc), prefrontal cortex (PFC), hippocampus, and amygdala [64– 66]. The VTA is comprised of three main neuronal phenotypes including dopaminergic, GABAergic, and glutamatergic neurons. The rate limiting enzyme for catecholamine synthesis is tyrosine hydroxylase (TH), which catalyzes the hydroxylation of tyrosine to 3,4-dihydroxyphenylalanine that is then rapidly decarboxylated to produce dopamine and, in some neuronal populations, norepinephrine and epinephrine [67]. Optogenetic Alongside the homeostatic-hypothalamic circuitry, energy balance is also modulated by hedonic aspects of feeding behavior pertaining to reward-based food intake, or eating for pleasure. Amounting evidence in rodents and humans now support the theory that both drugs of abuse and the consumption of highly palatable foods converge on a shared pathway within the limbic system to mediate motivated behaviors [61–63]. Therefore, the hedonic consumption of palatable foods involves the mesolimbic dopamine (A10) neurons that emanate from the ventral tegmental area (VTA) and project onto structures such as the nucleus accumbens (NAc), prefrontal cortex (PFC), hippocampus, and amygdala [64–66]. The VTA is comprised of three main neuronal phenotypes including dopaminergic, GABAergic, and glutamatergic neurons. The rate limiting enzyme for catecholamine synthesis is tyrosine hydroxylase (TH), which catalyzes the hydroxylation of tyrosine to 3,4-dihydroxyphenylalanine that is then rapidly decarboxylated to produce dopamine and, in some neuronal populations, norepinephrine and epinephrine [67]. Optogenetic studies have shown that local activation of A<sup>10</sup> dopamine neurons or their terminals within the NAc promotes responses of increased reward-seeking behavior [68–71]. On the other hand, activation of VTA GABAergic neurons promotes aversive responses, while their inhibition promotes reward, by dampening or promoting A<sup>10</sup> dopamine signaling, respectively [72,73]. Photoactivation studies, following an intracranial self-stimulation

operant paradigm protocol [74], further support the role of VTA A<sup>10</sup> dopamine neurons in reward-seeking. This reward-seeking role is exemplified through data showing that photoactivation of VTA glutamatergic neurons expressing cation channel rhodopsin-2 (ChR2) under a vesicular glutamate transporter-2 (VGlut2) promoter caused conditioned place preference to a photostimulation-paired chamber and motivated operant responding to earn optical intracranial self-stimulation in mice [75]. These rewarding effects are attributed to local excitatory input onto the VTA A<sup>10</sup> dopamine neurons [75]. A<sup>10</sup> dopamine neurons thereby encode reward processing for natural and drug-induced rewards and are implicated in increasing incentive salience for palatable foods, food-seeking behavior and impulsivity that could, under the right circumstances, lead to binge-feeding behavior [4,64,66,76–78].

To fully grasp the distinct role A<sup>10</sup> dopamine neurons play in hedonic feeding patterns, it is imperative to delineate the multifaceted aspects of global reward processing. The influential theories proposed by Berridge and colleagues make the case that when examining the role of food reward in feeding behavior, distinctions must be made between what he coined as reward 'liking' and reward 'wanting' [79,80]. 'Liking' is associated with the hedonic impact, or the brain reaction underlying sensory pleasure triggered by a rewarding stimulus, such as a highly palatable food [79,80]. Hedonic pleasure ('liking') has been reliably measured through observation of facial affective reactions prompted by exposure to a natural taste stimulus, where sweet tastes elicit positive 'liking' patterns of distinct orofacial expressions (e.g., rhythmic or lateral tongue protrusions) and bitter tastes alternatively evoke "disliking' expressions (e.g., gapes) [81]. These varying patterns of orofacial reactions are homologous to those observed in human infants, orangutan, chimpanzees, monkeys, rats, and mice, insinuating evolutionary conservation of the underlying brain circuits involved [80–83]. Data collected from taste reactivity studies allowed further insights into the neural underpinning of hedonic impact reactions, illuminating involvement of hedonic hotspots in the rostrodorsal quadrant of the medial shell of the NAc, ventral pallidum, and brainstem regions such as the parabrachial nucleus in the pons [80,84,85].

Conversely, hedonic impact ('liking') is distinguishable from 'wanting', which is related to incentive salience or motivation in reward-seeking [79,80]. Further, incentive salience (in the form of cue-triggered 'wanting') is mediated by separate neural networks, like those originating from A<sup>10</sup> dopamine neurons in the VTA, though integration of signals from both components is mechanistically crucial to produce the full phenomenon we typically think of as reward [79,80]. As previously mentioned, VTA dopamine signaling is indicated in incentive salience, a notion vastly supported by evidence wherein rewarding stimuli leads to enhanced dopamine transmission, while suppression of dopamine signaling lessens the motivation for rewards including food, sex, and drugs [65,76,86–88]. However, taste reactivity studies utilizing mutant mice with genetically abolished neural dopamine, or impaired ventral striatal dopamine caused by neurochemical 6-OHDA lesions, have reiterated this point, showing no detectable effects on ability to register the pleasurability/hedonic impact of tastes in the absence of dopamine [80,89]. Recognition of these separate neuronal networks encompassing overall reward is key to be able to discern the contributing neural pathways, and perturbations of these pathways, that lead to pathological reward-seeking behaviors (e.g., in food and drug addiction). In regard our recent study described below in Section 4, we mainly focus on 'wanting' behaviors in hedonic feeding and modulations of A<sup>10</sup> dopamine signaling in the context of food reward and binge feeding.

It should be noted that A<sup>10</sup> dopamine neurons also play a critical role in aversion processing in ways that are related to their firing pattern as well as the topographical pattern of dopaminergic innervation of the NAc. In vivo photometry studies have revealed that responses to aversive stimuli are mediated by A<sup>10</sup> dopamine neurons terminating in the ventromedial shell of the NAc [90]. Aversive responses are induced by excitation of these neurons that occurs, at least in part, via glutamatergic input from the lateral hypothalamus [91]. Optogenetic studies indicate that prolonged stimulation of these A<sup>10</sup> dopamine neurons activates dopamine D1 and D2 receptor-bearing medium spiny neurons that, in turn, increases the firing rate of GABAergic neurons in the ventral pallidum, decreases the excitability of A<sup>10</sup> dopamine neurons and reduces cocaine reward [92].

As with the hypothalamic energy balance circuit, the neurons comprising the hedonic reward circuitry, namely the A<sup>10</sup> dopamine neurons, are also sensitive to circulating leptin, insulin, and ghrelin levels, relaying signals to regulate food-seeking behavior and ultimately changes in body mass [93]. Leptin binding to and activating LEPRs expressed on VTA A<sup>10</sup> dopamine neurons leads to phosphorylation of STAT3, membrane hyperpolarization and reduced firing in these neurons [94,95]. Moreover, site-specific ablation of LEPRs in the VTA heightens the sensitivity of mice to the rewarding aspect of highly palatable foods (e.g., sucrose), while microinjection of leptin into the VTA reduces food intake [95]. Secondary regulation of VTA dopamine neurons by leptin is mediated via direct leptin action on LH LEPR expressing neurons which in turn relay signals onto the A<sup>10</sup> dopamine population, via synaptic contact, to promote decreases in food intake and concomitant decreases in body weight [96]. Interestingly, LH LEPR neurons have been shown to represent a unique neuronal population distinct from the previously mentioned orexin or MCH neurons also found in the LH [96]. Anterograde tract- and retrograde tracing further confirmed that LH LEPR neurons project caudally to densely innervate the VTA, with few to no projections seen in hypothalamic regions (including the ARC) or in the striatum (including the NAc). Therefore, LH LEPR neurons may play a unique role in A<sup>10</sup> dopamine regulation compared to other LEPR expressing neurons in the hypothalamus [96]. This interaction of LH signaling to control VTA neurons importantly highlights how homeostatic and hedonic neural circuits may dynamically and coordinately interact with one another to promote global energy balance. Insulin receptors are also present on VTA (and substantia nigra) dopamine neurons and can induce expression of the dopamine transporter (DAT) in these neuronal populations, as was demonstrated following intracerebroventricular (i.c.v.) insulin treatment [93]. Increased DAT results in quicker dopamine reuptake from the synaptic cleft back into presynaptic neurons, thereby halting stimulation of postsynaptic neurons and ultimately working to decrease the rewarding effect of food [93]. Furthermore, ghrelin binding to and activating ghrelin receptors expressed on A<sup>10</sup> dopamine neurons has been shown to elevate the frequency of action potentials in these neurons, as well as induce increased dopamine turnover in the NAc, promoting appetite [97]. Interestingly, the ghrelininduced increase in locomotor activity and dopamine release in the NAc is negated by CB1 receptor antagonism with rimonabant [98]. Thus, the abundant presence of communication between the brains reward circuitry and fluctuating hormones and neuromodulators, that typically relay nutritional/energy state cues, underscores the important role of A<sup>10</sup> dopamine signaling in energy balance and potential links to feeding pathologies.

The intrinsic separation of the neural systems encoding hedonic impact from pleasure and the incentive salience or motivation for a reward, can give way to possible explanations for aberrant feeding behaviors that may underlie feeding pathologies [80]. Individuals with substance abuse disorders seem to take drugs compulsively even when they no longer derive pleasure from them ('liking'), and their motivation to take the drug ('wanting') may persist due to long-lasting sensitization of their brain mesolimbic systems, brought on by repeated binges [76,80,88]. Food is a natural reward with reinforcing properties, similar to rewards such as sex and drugs, and activates the dopaminergic mesolimbic system by elevating extracellular dopamine concentrations in the NAc [99–101]. In particular, highly palatable, calorie rich foods can critically effect A<sup>10</sup> dopamine neurotransmission [102,103], analogous to modulations caused by other potent reinforcing stimuli, e.g., cocaine, amphetamines, opiate-like drugs, cannabinoids, alcohol, and nicotine [104–108]. Some evidence suggests that similar sensitization-like changes can be induced by exposure to certain regimens of food and restriction, modelling oscillations between dieting and binging on palatable foods [109–113]. Wherein, rats exposed to brief, intermittent bouts of sucrose access (sucrose binges), express sensitization-like changes, especially when

binges are cycled with food-restriction. Observed changes include: increasing propensity to over-consume when allowed, an enduring enhanced neural response to the presentation of food reward and cues, and an over-response to the psychostimulant effects of drugs such as amphetamine (a typical behavioral marker of drug-induced neural sensitization, which suggests a common underlying mechanism) [110,113–115]. Indeed, dopamine release into the NAc and c-fos expression in A<sup>10</sup> dopamine neurons is increased during binge-feeding episodes [77,116], lending further credence to this notion. Additionally, obesity is associated with dysfunction of dopaminergic systems. Obese patients present with reduced striatal dopamine D2 receptor, as measured by positron emission tomography [117]. This similarity between food reward and drug reward gives rise to the notion that feeding disorders and drug abuse and/or dependence share common mechanisms, as neuropsychological diseases involving negative alterations of the neural networks associated with the reinforcing properties of rewards [118–120].

#### **3. Influences of Sex and Diet on Central Energy Balance Circuits**

Sex differences are abundantly present in the context of energy homeostasis and although the prevalence of obesity is similar between men and women, women seem to have a greater risk of developing eating disorders and extreme obesity [121,122]. Indeed, sex is thought to represent one of the main risk factors for food-related disorders, including binge eating disorder [123]. Other lines of evidence illustrating sex differences in foodbased reward processing include women having a reduced ability to control food desire, higher cortical and limbic activation when presented with visual, gustatory and olfactory cues, as well as increased susceptibility to episodes of food-craving and lack of control for sugary foods, compared to men [124–128]. The role of sex hormones, particularly the fluctuating estrogen levels in females throughout the estrous cycle, is of interest in these disparities seen between males and females for reward-based consumption.

Estrogens elicit inhibitory effects on food intake, attributable to activation of estrogen receptors in key brain regions responsible for food intake control and body weight, such as the hypothalamus and nucleus tractus solitarius [5,129–132]. In regard to sex hormones, studies on the cannabinoid system have elucidated activation effects of gonadal steroid hormones on hypothalamic energy balance neural circuitry. Estradiol (E2) in females has been shown to attenuate cannabinoid-induced hyperphagia and hypothermia, as well the decrease in glutamatergic input onto POMC neurons [133]. These estrogenic actions occur through activation of estrogen receptor (ER) and the Gq-coupled membrane ER (GqmER), which triggers a signaling pathway involving PI3K, protein kinase C (PKC), protein kinase A (PKA) and neuronal nitric oxide synthase (nNOS) [131,132,134]. This, in turn, diminishes endocannabinoid tone at VMN SF-1/ARC POMC synapses, thereby relieving the retrograde inhibition of the glutamatergic input [10,11]. Regarding nNOS, while multiple isoforms with differing efficacies in eliciting downstream nitric oxide signaling have been characterized in brain [135], it is currently unknown which isoform mediates the estrogenic diminution in endocannabinoid tone at these synapses. Therefore, it will be necessary for future studies to elucidate the isoform responsible.

Interestingly, functional glutamatergic synapses at these VMN SF-1/ARC POMC synapses are largely silenced in obese females [11]. Indeed, studies of both POMC and NPY/AgRP neurons from *ob/ob* mice reveal that they undergo extensive synaptic plasticity under conditions of obesity; with the former receiving significantly more inhibitory inputs concomitant with a reduction in excitatory inputs, and the latter having appreciably more excitatory inputs and fewer inhibitory ones impinging upon them [136]. Obesity also correlates with chronic inflammation and resistance to leptin and insulin not only in the CNS but also peripherally [137–140]. However, the inflammation, reactive gliosis and subsequent neuronal injury observed in the mediobasal hypothalamus develops more rapidly than for similar maladaptations occurring in peripheral organs [141].

By contrast, testosterone in males rapidly increases energy intake and is reversed by the CB1 receptor antagonist AM251 and the diacylglycerol lipase (DAGL) inhibitor

Orlistat, and potentiates the cannabinoid-induced decrease in glutamatergic input onto POMC neurons [10,142]. This androgen-induced increase in endocannabinoid tone is due to activation of AMPK, which augments retrograde inhibition of glutamate release at VMN SF-1/ARC POMC synapses [10,142]. These effects are further magnified in obese males due to reduced PI3K signaling in the ARC [11,143]. Lastly, the development of central insulin resistance brought on by diet-induced obesity is sexually differentiated. Males are more susceptible to the attenuated activation of both TRPC5 channels in POMC neurons and KATP channels in NPY/AgRP neurons than are females under the protection of E2, which prevents the respective increase in suppressor of cytokine signaling-3 and decrease in PI3K signaling that drives central insulin resistance [143,144].

Estrogen receptors are also expressed within the VTA. Evidence indicates that estrogens increase self-administration of rewards like psychomotor stimulants and alcohol [145–147]. In animal studies, female rats have been more motivated to work for cocaine during the estrus phase, compared to other phases of the estrous cycle [148,149], and E2-treated ovariectomized (OVX) female rats exhibited increased motivation to self-administer cocaine [150]. The effects of estrogen on reward neural circuitry are also evident in motivation for food rewards. In opposition of the findings from self-administration of drug rewards, intra-VTA injections of E<sup>2</sup> significantly reduced the motivation to work for sucrose rewards in a progressive ratio operant conditioning task within 1 h after injection, while overall food intake was not altered by this treatment [151]. Additionally, a study comparing the self-administration of chocolate-flavored beverage (CFB) and concomitant changes in extracellular dopamine in the dialysate obtained from the NAc, between male as well as intact and OVX female rats, showed that female rats in the proestrus and estrus phases of the cycle had reduced lever responding for, and amount of self-administered CFB, paired with lowered extracellular dopamine in the dialysate from the NAc shell [152]. These variable findings between food versus drug rewards raise questions about the role estrogens play in food reward processing and how they may potentially explain the disparate prevalence rates between males and females in feeding behavior pathologies.

#### **4. Nociceptin/Orphanin FQ Regulation of Homeostatic and Hedonic Energy Balance Circuits**

The neuropeptide nociceptin/Orphanin FQ (N/OFQ) is an endogenous opioid hepadecapeptide that is encoded by the prepronociceptin gene [153–155]. N/OFQ binds with high affinity to its cognate Gi/o-coupled nociceptin opioid peptide (NOP) receptor, and despite high structure homology, has minimal affinity for classic opioid receptors (mu, kappa, or delta opioid receptors), nor do classical opioid receptor ligands (e.g., naloxone, endorphin, dynorphin) have high affinity binding for NOP [155,156]. Initial studies on N/OFQ indicated the peptide attenuates locomotor activity, increases sensitivity to pain, while blocking the antinociceptive activity of mu, delta, and kappa analgesics following i.c.v. injections; thus garnering the name nociceptin [154,155,157]. The NOP receptor is expressed extensively throughout the central nervous system (CNS) [153,156,158,159], subserving a role in an array of central processes including pain, learning and memory, emotional states, neuroendocrine control, food intake, and motor control (see [160] for thorough review).

The actions of N/OFQ in these disparate neuronal systems is accredited to peptide binding and subsequent activation of its NOP receptor. In line with agonist activation in all GPCRs, following NOP activation by N/OFQ the Gα and Gβγ subunits dissociate to then promote the various effector pathways [155,158]. Therein, NOP receptor activation inhibits adenylate cyclase (AC) activity and couples to pertussis-toxin-sensitive G-proteins resulting in decreased cyclic adenosine monophosphate (cAMP) production [154,155,160, 161]. Further, NOP receptors canonically couple to G protein-gated inwardly rectifying potassium (GIRK)3 and both N-type as well as P/Q type voltage-gated Ca2+ channels. Upon NOP receptor activation, Ca2+ currents are reduced and GIRK channels activated, causing K<sup>+</sup> efflux, cellular hyperpolarization and attenuated neural activity [162–164]. Additionally, NOP receptor signaling has been shown to promote activation of PKC as well as phospholipase A2 and C [165–167].

As mentioned, the N/OFQ/NOP receptor system is expressed widely throughout the brain in mice, rats, and humans, and importantly is expressed densely in the ARC and VTA [153,156,159,168,169]. The abundant expression of N/OFQ and NOP in these key regions supports a role in both homeostatic and hedonic energy balance regulation. Indeed, i.c.v. injections into the lateral ventricle (LV), 3V, and intranuclear injections into the ARC of N/OFQ or NOP agonists have been shown to produce a dose-dependent increase in feeding behavior even in satiated animals; intra-ARC injections proving to be the most efficacious to induce hyperphagia [170–174]. Additionally, chronic i.c.v. infusions of the neuropeptide have been associated with increased body weight via modifications in feeding and metabolism in mice [175]. In addition, we have found that the rebound hyperphagia seen upon refeeding in fasted NOP receptor null mice is significantly blunted compared to their wildtype littermate controls [176]. The appetite-stimulating properties of N/OFQ can be attributed, at least in part, to the NOP receptor-mediated inhibition of POMC neurons via activation of GIRK channels, as seen with either exogenous bath application of the neuropeptide or high-frequency optogenetic stimulation of ARC N/OFQ neurons [5,176–179]. The observed effects of N/OFQ are not limited to anorexigenic pathways, however, and have been observed to also influence orexigenic circuitry. NOP receptors are expressed in NPY/AgRP neurons, and N/OFQ increases AgRP release from mediobasal hypothalamic explants [180]. Interestingly, these ARC N/OFQ neurons corelease GABA upon low-frequency optogenetic stimulation, which can provide additional inhibitory input onto POMC neurons [181]. They are also regulated by ambient levels of extracellular glucose, and inhibited by leptin [181]. On the other hand, N/OFQ administered into the perifornical/lateral hypothalamic area exerts a hypophagic effect attributed to a NOP receptor-mediated inhibition of orexin neurons via activation of KATP channels [182]. Nevertheless, the prevailing sentiment based on the totality of the available evidence clearly indicates that N/OFQ is profoundly hyperphagic via its collective actions within the homeostatic energy balance circuitry.

Moreover, several lines of in vivo and in vitro evidence illustrate the effects of N/OFQ on the A<sup>10</sup> dopamine system, providing credence to the notion that N/OFQ may be inherently linked to reward system processing, and thus to hedonic feeding behavior. Data obtained from in vivo studies have shown that N/OFQ decreases the outflow of dopamine in the NAc following intracerebral injections, dampens the morphine-induced dopamine release in the NAc, and blocks the acquisition of morphine-dependent place preference [183–186]. Further, bath application of N/OFQ during intracellular electrophysiology recordings in rat midbrain slices caused membrane hyperpolarization and reduced firing under current-clamp, which was associated with an outward current under voltage clamp [187]. These inhibitory effects of N/OFQ in the VTA were reduced by the NOP receptor antagonist [Phe11jCH2-NH)Gly2]NC(1 ± 13)NH2 (1 µM) but were unaltered by presence of tetrodotoxin or the opioid receptor antagonist naloxone [187]. Additionally, activation of NOP receptors expressed in the NAc and dorsal striatum work to pre-synaptically inhibit dopamine synthesis and tyrosine hydroxylase (TH, the rate-limiting enzyme in dopamine production) phosphorylation and act post-synaptically on dopamine responsive neurons by decreasing dopamine D1 receptor signaling via suppression of cAMP/PKA activity [188]. Lastly, microdialysis studies have revealed that N/OFQ significantly reduces extracellular DA levels in both the VTA and NAc [184].

The preceding section on sex differences in energy balance control is particularly relevant when it comes to NOP receptor-mediated regulation. In homeostatic energy balance, males are significantly more sensitive to the inhibitory effects of exogenous N/OFQ on excitatory neurotransmission at VMN SF-1/ARC POMC synapses than are females during E2-dominated phases of the estrous cycle in two ways: (1) the direct hyperpolarization/cessation of firing of both VMN SF-1 and ARC POMC neurons, as well as the underlying outward current, due to the activation of GIRK channels is greater in males than

in females and (2) the presynaptic inhibition of glutamate released of VMN SF-1 neurons at these synapses is more substantive in males than in females [5,189,190]. Additionally, E<sup>2</sup> exerts powerful activational effects by diminishing the inhibitory effects of exogenous N/OFQ at VMN SF-1/ARC POMC synapses, as well as the decreased excitability of POMC neurons caused by optogenetic stimulation of ARC N/OFQ neurons, and protecting against the aberrant hyperphagia and reduction in energy expenditure caused by exogenous N/OFQ administered directly into the ARC of obese OVX females [5,178,179,191]. E<sup>2</sup> also attenuates the pleiotropic actions of N/OFQ on POMC neurons by binding to either ERα or the Gq-mER, which leads to a signaling cascade that includes phospholipase C (PLC), PI3K, PKC, PKA and nNOS [191]. Furthermore, progesterone administered to OVX, estrogenprimed females restores the sensitivity of POMC neurons to these pre- and postsynaptic actions of N/OFQ [192]. The precise actions of N/OFQ and NOP receptor activation in the aforementioned neural circuitry, its effects on feeding-behavior, and potential role in feeding-related pathologies will be discussed in further detail below.

In addition to sex differences, diet modifications resulting lean or obese phenotypes alter the signaling effects of N/OFQ in homeostatic and hedonic neural circuits, and increases the risk for aberrant feeding-behavioral patterns to form. Diet-induced obesity increases the intrinsic excitability of ARC N/OFQ neurons, which augments the inhibitory GABAergic tone received by POMC neurons [181]. Conversely, ablation of ARC N/OFQ neurons hinders the development of obesity caused by a four-week exposure to a HFD [181]. Moreover, diet-induced obesity renders VMN SF-1/ARC POMC synapses more susceptible to the inhibitory effects of exogenous N/OFQ in males and hypoestrogenic OVX females [5]. This greatly curtails anorexigenic signaling at these synapses; causing exaggerated N/OFQinduced increases in energy intake and decreases in energy expenditure, and is entirely in keeping with our recent demonstration that N/OFQ administered directly into the ARC significantly enhances binge-feeding behavior caused by short-term intermittent exposure to a HFD [179].

Endogenous N/OFQ signaling may also be intrinsically associated with or escalate aberrant feeding patterns associated with diet-induced obesity and/or binge-feeding. In one instance, Sprague Dawley rats that had previously been determined as "fat-preferring", were particularly susceptible to N/OFQ-induced acute hyperphagia following i.c.v. injection [173]. Additionally, NOP receptor knockout mice displayed significantly reduced levels of HFD food consumption, compared to their wildtype littermate controls [193]. Additionally, administration of the novel NOP antagonist LY2940094, effectively increases lipid utilization metabolism and reduces fasting-induced hyperphagia of chow in wildtype 129S6 mice, but not in those with genetic deletion of the NOP receptor [194,195]. Further, LY2940094 reduces HFD consumption measured over a 5-h exposure period and also hindered weight gain over 3 days of HFD exposure [195]. Providing further support for the intrinsic contribution of endogenous N/OFQ/NOP signaling in the pathogenesis of obesity and eating disorders, systemic administration of LY2940094 reduced intake of HFD in diet-induced obese rats and mice, and also improved metabolic parameters by reducing the respiratory quotient in mice with access to HFD in their metabolic feeding chamber [195]. In relation to N/OFQ and binge-feeding, systemic treatment with the selective NOP antagonist SB 612111 produced a dose-dependent decrease in intermittent HFD binge eating, but not a change in the total 24-h food intake of mice who were either on an intermittent-HFD or continuous-HFD feeding regimen [196].

While the last few studies mentioned regarding altered feeding behavior certainly provide evidence that the N/OFQ-NOP system is involved in the neuropathology of obesity and related eating disorders, they were not designed to offer insight into the specific neuronal populations acted upon that leads to such pathogeneses. In addition, the exact mechanisms through which sex and diet interact to modulate NOP receptor-mediated inhibition of reward encoding A<sup>10</sup> dopamine neurons and hedonic feeding remained a mystery until very recently. We have discovered that the endogenous release of N/OFQ caused by high-frequency optogenetic stimulation of VTA neurons in mesencephalic slices from

N/OFQ-cre mice powerfully inhibits neighboring VTA neurons; an effect that is faithfully recapitulated by exogenous bath application of N/OFQ during recordings of A<sup>10</sup> dopamine neurons in slices from TH-cre mice [179]. The membrane hyperpolarizations and underlying outward currents were attenuated by E<sup>2</sup> in females, and accentuated by diet-induced obesity in males [179]. These N/OFQ-induced inhibitory effects on A<sup>10</sup> dopamine neurons functionally translated into sex- and diet-dependent changes in binge-eating behavior, as N/OFQ delivered into the VTA decreased the rampant consumption seen during the binge episodes in obese but not lean males, and in both lean and obese females [179]. Just as we saw with the inhibitory effects of N/OFQ in A<sup>10</sup> dopamine neurons from female mice, E<sup>2</sup> counteracted the inhibitory effect of intra-VTA N/OFQ on binge feeding [179]. Thus, it is clear that N/OFQ exerts multifaceted effects on energy balance via NOP receptor-mediated regulation of homeostatic and hedonic circuits that are site-specific as well as sex hormoneand diet-dependent. The site specificity underscores the fact that despite the inhibitory effect of N/OFQ on A<sup>10</sup> dopamine neurons and the associated dampening of binge-feeding behavior, there is clearly a net hyperphagic response caused by the peptide that is due largely, if not exclusively, to its effects within the homeostatic energy balance circuitry.

#### **5. Pituitary Adenylate Cyclase-Activating Polypeptide (PACAP) Regulation of Homeostatic and Hedonic Energy Balance Circuits**

Pituitary adenylate cyclase-activating polypeptide (PACAP) is a hypophysiotropic neurohormone belonging to the largest family of developmental and regulatory peptides that includes vasoactive intestinal polypeptide (VIP)-secretin-growth hormone-releasing hormone (GHRH)-glucagon superfamily, along with secretin, glucagon, human growth hormone-releasing factor (hGRF), and VIP [197]. Containing the most conserved sequence throughout evolution in its family, PACAP is encoded by the ADCYAP1 gene as a 38-amino acid C-terminally *α*-amidated polypeptide [197]. It is transcribed in hypothalamic neurons and peripheral organs, such as the GI tract, pancreas, and gonads; exerting pleiotropic physiological effects such as the regulation of neurotransmitter release and secretion, vasodilation, energy balance, as well as stimulation and inhibition of cell proliferation and/or differentiation [198–200]. The PACAP precursor yields two different forms of PACAP, PACAP1–38 and PACAP1–27, as well as PACAP-related peptide (PRP) [200]. Given its widespread distribution in the CNS and periphery, PACAP is well-equipped to act as a hormone, a neurotransmitter, and a trophic factor in various tissue types [200].

The effects of PACAP are mediated through two classes of PACAP receptors: the pituitary adenylate cyclase-activating polypeptide-specific 1 receptor (PAC1) and two subtypes of VIP/PACAP-receptors termed VPAC1 and VPAC2 [197,200]. PAC1 receptor exhibits almost a twofold higher affinity for PACAP than for VIP, while the VPAC1 and VPAC2 receptors recognize both PACAP and VIP with equally high affinity [197]. PAC1, a metabotropic receptor, is found in various hypothalamic structures including the supraoptic nucleus (SON), PVN, ARC, LH, VMN, as well as extrahypothalamic regions of the brain like the cerebral cortex, Broca's area, the hippocampus, among others [197,200]. VPAC1 and VPAC2 receptors are appreciably expressed in peripheral organs including the lung, duodenum, and thymus, although with less abundance than PAC1 receptors [201,202]. There is considerable evidence demonstrating that the PAC1 receptor signals through G<sup>q</sup> and Gs. For example, physiological studies have determined that PACAP acts on PAC1 receptor in mouse neural stem cells, and upon receptor activation the signal generated is carried via a Gq-mediated PLC/diacylglycerol/inositol 1,4,5-trisphosphate (IP3)-dependent signaling pathway [203]. In the hypothalamus, PACAP binds to PAC1 receptors that induce Gq-coupled stimulation of PLC, PI3K, and PKC to ultimately activate TRPC5 channels [12]. In the neurohypophysis, activation of PAC1 receptors by PACAP can lead to signaling via the G<sup>s</sup> pathway [204] that increases firing and depolarizes the membrane potential of magnocellular neurons in rat brain slices via the activation of adenylate cyclase/cAMP/PKA signaling [204–206]. In doing so, PACAP stimulates the release of oxytocin and vasopressin from the posterior pituitary [204,207]. Activation of PAC1 receptors has also been shown

to elevate intracellular Ca2+ concentrations in dissociated magnocellular neurons from rat SON, and stimulate a quinine- sensitive K<sup>+</sup> outward current in murine microglia [207,208].

In the immune system, PACAP decreases chemotaxis of thymocytes and lymphocytes both via the activation of the PKA pathway [209,210]. Additionally, PACAP inhibits tumor necrosis factor-α and both interleukin-6 and interleukin-12 release, while enhancing the production of the cytokine interleukin-10 in lipopolysaccharide-activated macrophages; suggesting it acts as a protective agent that regulates the release of proinflammatory and anti-inflammatory cytokines [211–214]. The effects of PACAP on cell proliferation/survival are dependent on the downstream channel targeted as well as transcriptional cues. In vitro experiments have demonstrated that the effect of PACAP on cell survival is regulated via the activation of the G<sup>s</sup> pathway, contributing to the phosphorylation of the extracellular signal-regulated (ERK)-type of mitogen-activated protein (MAP) kinase and enhanced c-fos gene expression [215–217]. On the other hand, PACAP dose-dependently inhibits concanavalin A-induced cell proliferation in murine splenocytes [218]. Moreover, PACAP stimulates Ca2+ mobilization and blocks K<sup>+</sup> currents in a variety of neuronal cell types (e.g., magnocellular neurons, cerebellar granule cells), two processes intimately linked to PACAP-induced enhancements in cell excitability [206,219–221]. Furthermore, PACAP stimulates postprandial glucagon-like peptide, leptin and insulin secretion, and also has been shown to promote insulin secretion from pancreatic β-cells via Ca2+ influx through L-type Ca2+ channels [222–226]. This latter effect may be dependent on ambient glucose concentration and its ability to enhance ATP production and thereby negatively gate ATPsensitive potassium (K<sup>+</sup> ATP) channels [227,228]. This indicates that endogenous PACAP acts as a physiological regulator of pancreatic β-cell activity linked to K<sup>+</sup> ATP channels in a manner similar to that described for the vasodilatory and neuroprotective effects of the peptide. [229,230]. Thus, it is clear that PACAP regulates a wide array of bodily functions including hypophysiotropic neurosecretion, glial function, immunomodulation, cell proliferation/survival, glucose homeostasis, vasodilation, neuroprotection and energy homeostasis via Gs- and Gq-mediated signaling. Moreover, given the inexorable link between obesity and inflammation [137–139,141], it is entirely conceivable that the overall anti-inflammatory effect of PACAP contributes significantly to lean phenotypes promoted by this peptide [226,231].

PACAP exerts myriad effects on energy balance at all levels of the brain-gut axis. For example, intravenous injection of PACAP on rats causes secretion of saliva from the submandibular and parotid glands, whereas in the stomach, PACAP decreases histamine- and pentagastrin-activated gastric acid secretion; the latter suggesting that PACAP acts indirectly to regulate gastric acid release [232–234]. In addition, intravenous injection of PACAP increases bicarbonate secretion and chloride secretion in the duodenum and in the distal colon, respectively [235,236]. Moreover, while PACAP administered centrally increases gastric motility, peripherally it evokes a dose-dependent relaxation of the gastric smooth muscles, decreases gastric motility and therefore delays stomach clearing [237–239]. Likewise, PACAP stimulates intestinal smooth muscles to relax in rats and other species [240–243].

Concerning peripheral glucose and lipid homeostasis, PACAP exerts a more potent action in stimulating glucose output from a perfused rat liver as compared to VIP [244]. While PACAP can clearly act centrally to regulate glucose homeostasis, the hyperglycemic role of PACAP in vivo can also be attributed to both an indirect action via increase in plasma glucagon and/or catecholamines, which increase glycogenolysis and gluconeogenesis [245]. Finally, in regard to energy metabolism, PACAP is known to accelerate lipolysis via the sympathetic nervous system (SNS). This suggests that hypothalamic PACAP signaling may promote the use of catabolized lipids as a viable energy source [197,231].

The hypothalamic regions that play a role in the regulation of thermogenesis, energy expenditure, and energy intake such as the PVN, VMN, and ARC abundantly express PACAP and the PAC1 receptor, suggesting that PACAP plays a vital role in the control of these processes [200,201,246]. There are two major populations of hypothalamic PACAP neurons—one with cell bodies residing in the PVN and the other with somata in the VMN [9,12,247]. The PVN PACAP neurons are thought to promote appetite through synaptic connections with, and excitation of, NPY/AgRP neurons [9]. It is the loss of signaling via this population of PACAP neurons that may explain reports of hypoinsulinemia, decreased adiposity, lower body weight and increased insulin sensitivity seen in transgenic PACAP-null mice [248]. The VMN PACAP neurons exhibit extensive colocalization with SF-1, which along with leptin drives PACAP expression [12,247,249]. These neurons are reported to be glucose inhibited, and their selective activation reduces circulating insulin concentrations and glucose tolerance [250,251]. VMN PACAP neurons make synaptic contact with POMC neurons, and like POMC neurons, are excited by leptin [9,12,36,37,252]. As such, this VMN population of PACAP neurons is poised to suppress appetite and enhance energy expenditure. In accordance, studies show that PACAP injected into the VMN causes an increase in body temperature via adaptive thermogenesis and increased levels of UCP-1 [197,247,253,254]. The adaptive thermogenesis brought on in part by increasing WAT browning in rodents increases energy expenditure and suppresses diet-induced obesity and glucose intolerance [49,50]. PACAP also controls activity, as an injection of PACAP given i.c.v. or into the VMN increases locomotion in rodents concomitant with an increase in O<sup>2</sup> consumption [247,253,254]. A systemic injection of PACAP into wild type (WT) mice dose-dependently lowers cumulative energy intake and decreases various indices of meal pattern like meal-size and rate of consumption, which correlates with reduced ghrelin secretion [226]. PACAP delivered directly into the VMN or PVN also reduces energy intake; however, these effects are coupled with somewhat disparate effects on meal pattern. PACAP injected directly into the PVN decreases meal size, rate of consumption, duration, total time spent eating and increased latency to meal initiation, whereas PACAP injected into the VMN only evokes an increase in the latency to meal initiation and a decrease in the rate of eating [254]. Thus, PACAP signaling throughout the homeostatic energy balance circuitry exerts far-reaching effects on energy intake, meal pattern and energy expenditure.

Additional studies investigating the homeostatic energy balance circuitry have demonstrated that PAC1 receptors are expressed in POMC neurons, and that PACAP administered to ad libitum-fed animals elevates POMC expression, c-Fos expression in POMC neurons, and MC4R receptor mRNA expression when injected i.c.v. or into the VMN, as well as enhancing α-MSH release from hypothalamic explants [250,253,255]. In contrast, PACAP had no effect on AgRP, CART or NPY mRNA levels [250]. In addition, PAC1 receptor blockade with PACAP6–38 or PACAP deficiency seen in Adcyap1-/-mice significantly decreases the leptin-induced hypophagia, hyperthermia, and increase in WAT sympathetic nerve activity in vivo [231,247]. We have recently shown that the profound influences PACAP exerts in the homeostatic control of energy balance in ad libitum-fed mice are diet- and sex-dependent [12]. We observed that PACAP evokes an inward current associated with an increase in firing in ARC POMC neurons that was abolished by PAC1 receptor antagonism and TRPC5 channel blockade, and augmented by E<sup>2</sup> [12]. The inward current was significantly attenuated upon inhibition of PLC, PI3K and PKC, but not PKA; suggesting that the PACAP-induced activation of POMC neurons was mediated via, PI3K and Gq-mediated signaling pathways [12]. The stimulation of ARC POMC neurons caused by PACAP administered directly into the ARC suppresses energy intake and enhances energy metabolism, and these effects were markedly attenuated under conditions of diet-induced obesity [12]. These effects of exogenously applied PACAP were effectively mirrored by chemogenetic and optogenetic stimulation of VMN PACAP neurons [12]. Collectively, these data suggest that under ad libitum-fed conditions PACAP functions through a PI3K/PLC/PKC pathway to activate POMC neurons via TRPC5 channels upon PAC1 receptor stimulation via Gq-mediated signaling [12]. Thus, VMN PACAP/ARC POMC synapses constitute a critical anorexigenic component of the homeostatic energy balance circuitry, one that is accentuated by E<sup>2</sup> in females and attenuated by obesity in males. These findings are consistent with other examples of positive estrogenic modulation of the PACAP/PAC1 receptor system that occur during the response to stress. E<sup>2</sup> stimulates PACAP and PAC1 receptor expression in the bed nucleus of the stria terminalis (BTNS) as well as in the

medial basal hypothalamus (MBH) compared to levels seen in the oil-treated OVX controls [256]. Likewise, PACAP expression in the PVN and anterior pituitary is heightened during the elevated E<sup>2</sup> levels seen during proestrus [256]. Concerning the hedonic energy balance circuitry and the consumption of palatable food, prior studies have shown that when PACAP is injected into the NAc it mimics the inhibitory effect of GABA receptor agonists on binge-like feeding behavior and decreases firing in NAc neurons [257]. This effectively reduces hedonic drive for palatable food as gauged by decreases in appetitive orofacial responses to sucrose, as well as by increases in aversive responses when PACAP is administered into the caudal NAc [258].

Whereas obesity exemplifies a state of positive energy balance, fasting represents a state of negative energy balance. Food-restricted mice show low levels of POMC and PACAP mRNA expression coupled with an increase in NPY mRNA expression, and i.c.v. injections of PACAP decrease energy intake after 30 min of refeeding [247,250]. Surprisingly, in recent experiments we found that during voltage clamp recordings in POMC neurons from eGFP-POMC mice subjected to an 18-h fast for five consecutive days, the PACAP response reversed polarity from a predominantly excitatory inward current seen under ad libitum conditions to an inhibitory outward one (Figure 2A,B,E–H); see Supplementary Materials for full material and methods The inset I/V plots corresponding to the representative current traces in Figure 2A,B revealed that the reversal potential shifted from ~−20 mV under ad libitum conditions (indicative of a mixed cation current) to ~−90 mV under fasted conditions (indicative of a K<sup>+</sup> current). The PACAP-induced inhibitory outward current and the corresponding increase in conductance seen under fasting conditions were virtually abolished by the KATP channel blocker tolbutamide and the PAC1 receptor antagonist PACAP6–38 (Figure 2C,D,G,H; 2G: one-way ANOVA/LSD, F = 25.71, df = 3, *p* < 0.0001; 2H: one-way ANOVA/LSD, F = 9.78, df = 3, *p* < 0.0001). *Int. J. Mol. Sci.* **2021**, *22*, x 15 of 32

**Figure 2.** Fasting reverses the polarity of postsynaptic PACAP response in POMC neurons by switching the coupling of PAC1 receptors from TRPC5 to KATP channels. (**A**–**D**), Representative current traces from voltage clamp recordings of POMC neurons that depict the response elicited by 100 nM PACAP under ad libitum and fasting conditions, the latter of which is abrogated by blockade of KATP channels with tolbutamide (100 μM) and PAC1 receptors with PACAP6–38 (200 nM). The inset I/V plots illustrate that the fasting-induced change in polarity is due to a switch from a mixed cation conductance to a PAC1 receptor-mediated K+ conductance via KATP channels. (**E**,**F**), Pie charts that show the proportion of POMC neurons that are excited by, inhibited by, or unresponsive to PACAP under ad libitum (to the left) and fasting (to the right) conditions. (**G**,**H**) highlight the PACAP-induced changes in membrane current (ΔI) and conductance (ΔG); alone (*n* = 12) and in conjunction with tolbutamide (*n* = 12) and PACAP6–38 (*n* = 8). Bars represent means and lines 1 SEM. #, *p* < 0.05 relative to PACAP under ad libitum conditions. \*, *p* < 0.05 relative to **Figure 2.** Fasting reverses the polarity of postsynaptic PACAP response in POMC neurons by switching the coupling of PAC1 receptors from TRPC5 to KATP channels. (**A**–**D**), Representative current traces from voltage clamp recordings of POMC neurons that depict the response elicited by 100 nM PACAP under ad libitum and fasting conditions, the latter of which is abrogated by blockade of KATP channels with tolbutamide (100 µM) and PAC1 receptors with PACAP6–38 (200 nM). The inset I/V plots illustrate that the fasting-induced change in polarity is due to a switch from a mixed cation conductance to a PAC1 receptor-mediated K<sup>+</sup> conductance via KATP channels. (**E**,**F**), Pie charts that show the proportion of POMC neurons that are excited by, inhibited by, or unresponsive to PACAP under ad libitum (to the left) and fasting (to the right) conditions. (**G**,**H**) highlight the PACAP-induced changes in membrane current (∆I) and conductance (∆G); alone (*n* = 12) and in conjunction with tolbutamide (*n* = 12) and PACAP6–38 (*n* = 8). Bars represent means and lines 1 SEM. #, *p* < 0.05 relative to PACAP under ad libitum conditions. \*, *p* < 0.05 relative to PACAP under fasting conditions, one-way ANOVA/LSD. Figure adapted from [12].

**Figure 3.** E2 attenuates the PACAP-induced outward current in POMC neurons observed under fasting conditions. (**A**,**B**) are representative membrane current traces during voltage clamp recordings in EtOH vehicle- (0.01% (*v:v*); *n* = 11) and E2-treated (100 nM; *n* = 8) slices from OVX

female eGFP-POMC mice (Figure 3A,C–E). Bath application of E2 in slices from fasted OVX female eGFP-POMC mice reduced the magnitude of the PACAP-induced outward current by ~50%, and markedly attenuated the increase in K+ conductance (Figure 3B,D:

A nearly identical switch in polarity of the PACAP response in POMC neurons was

PACAP under fasting conditions, one-way ANOVA/LSD. Figure adapted from [12].

Student's t-test, t = 1.51, *p* < 0.15; 3E: Student's t-test, t = 2.664, *p* < 0.02).

A nearly identical switch in polarity of the PACAP response in POMC neurons was observed during voltage clamp recordings in EtOH vehicle-treated slices from fasted OVX female eGFP-POMC mice (Figure 3A,C–E). Bath application of E<sup>2</sup> in slices from fasted OVX female eGFP-POMC mice reduced the magnitude of the PACAP-induced outward current by ~50%, and markedly attenuated the increase in K<sup>+</sup> conductance (Figure 3B,D: Student's *t*-test, *t* = 1.51, *p* < 0.15; 3E: Student's *t*-test, *t* = 2.664, *p* < 0.02). A nearly identical switch in polarity of the PACAP response in POMC neurons was observed during voltage clamp recordings in EtOH vehicle-treated slices from fasted OVX female eGFP-POMC mice (Figure 3A,C–E). Bath application of E2 in slices from fasted OVX female eGFP-POMC mice reduced the magnitude of the PACAP-induced outward current by ~50%, and markedly attenuated the increase in K+ conductance (Figure 3B,D: Student's t-test, t = 1.51, *p* < 0.15; 3E: Student's t-test, t = 2.664, *p* < 0.02).

**Figure 2.** Fasting reverses the polarity of postsynaptic PACAP response in POMC neurons by switching the coupling of PAC1 receptors from TRPC5 to KATP channels. (**A**–**D**), Representative current traces from voltage clamp recordings of POMC neurons that depict the response elicited by 100 nM PACAP under ad libitum and fasting conditions, the latter of which is abrogated by blockade of KATP channels with tolbutamide (100 μM) and PAC1 receptors with PACAP6–38 (200 nM). The inset I/V plots illustrate that the fasting-induced change in polarity is due to a switch from a mixed cation conductance to a PAC1 receptor-mediated K+ conductance via KATP channels. (**E**,**F**), Pie charts that show the proportion of POMC neurons that are excited by, inhibited by, or unresponsive to PACAP under ad libitum (to the left) and fasting (to the right) conditions. (**G**,**H**) highlight the PACAP-induced changes in membrane current (ΔI) and conductance (ΔG); alone (*n* =

PACAP under fasting conditions, one-way ANOVA/LSD. Figure adapted from [12].

*Int. J. Mol. Sci.* **2021**, *22*, x 15 of 32

**Figure 3.** E2 attenuates the PACAP-induced outward current in POMC neurons observed under fasting conditions. (**A**,**B**) are representative membrane current traces during voltage clamp recordings in EtOH vehicle- (0.01% (*v:v*); *n* = 11) and E2-treated (100 nM; *n* = 8) slices from OVX **Figure 3.** E<sup>2</sup> attenuates the PACAP-induced outward current in POMC neurons observed under fasting conditions. (**A**,**B**) are representative membrane current traces during voltage clamp recordings in EtOH vehicle- (0.01% (*v:v*); *n* = 11) and E2 -treated (100 nM; *n* = 8) slices from OVX females that illustrate the estrogenic diminution of the robust and reversible PACAP-induced outward current and change in K<sup>+</sup> conductance (as seen from the inset I/V plots). (**C**), Pie chart that indicates the percentage of POMC neurons from OVX females that are excited by, inhibited by, or unresponsive to PACAP under fasting conditions. (**D**,**E**) show the composite data underscoring the ability of E<sup>2</sup> to negatively modulate the PACAP response. Bars represent means and lines 1 SEM of the PACAP-induced ∆I and ∆G in POMC neurons from OVX females under fasting conditions. \*, *p* < 0.05 relative to EtOH vehicle, Student's *t*-test.

Current clamp recordings in POMC neurons from fasted male eGFP-POMC mice uncovered a more hyperpolarized resting membrane potential (RMP) than those from their ad libitum-fed counterparts (Figure 4A: Student's *t*-test, *t* = 2.237, *p* < 0.04). Once again, the switch in polarity of the PACAP response was evident, such that the PACAPinduced depolarization of POMC neurons and the associated increase in firing seen in the representative trace and composite data from the ad libitum-fed state transformed into a hyperpolarization and a suppression of firing in the fasted state (Figure 4B–E; 4D: Student's *t*-test, *t* = 9.501, *p* < 0.0001; 4E: Kruskal–Wallis/median-notched box-and-whisker analysis, test statistic = 15.6201, *p* < 0.002).

females that illustrate the estrogenic diminution of the robust and reversible PACAP-induced outward current and change in K+ conductance (as seen from the inset I/V plots). (**C**), Pie chart that indicates the percentage of POMC neurons from OVX females that are excited by, inhibited by, or unresponsive to PACAP under fasting conditions. (**D**,**E**) show the composite data underscoring the ability of E2 to negatively modulate the PACAP response. Bars represent means and lines 1 SEM of the PACAP-induced ΔI and ΔG in POMC neurons from OVX females under fasting

Current clamp recordings in POMC neurons from fasted male eGFP-POMC mice uncovered a more hyperpolarized resting membrane potential (RMP) than those from their ad libitum-fed counterparts (Figure 4A: Student's t-test, t = 2.237, *p* < 0.04). Once again, the switch in polarity of the PACAP response was evident, such that the PACAPinduced depolarization of POMC neurons and the associated increase in firing seen in the representative trace and composite data from the ad libitum-fed state transformed into a

Student's t-test, t = 9.501, *p* < 0.0001; 4E: Kruskal–Wallis/median-notched box-and-whisker

conditions. \*, *p* < 0.05 relative to EtOH vehicle, Student's t-test.

analysis, test statistic = 15.6201, *p* < 0.002).

**Figure 4.** The PACAP-induced outward current observed during fasting conditions is associated with a hyperpolarization and a decrease in firing. (**A**), Composite bar graph that demonstrates the more hyperpolarized RMP of POMC neurons under fasting conditions. (**B**,**C**), Representative current clamp traces from POMC neurons showing the reversible PACAP-induced depolarization and increase in firing under ad libitum-fed conditions (*n* = 10) and the reversible hyperpolarization and suppression of firing seen under fasting conditions (*n* = 10). Comparable **Figure 4.** The PACAP-induced outward current observed during fasting conditions is associated with a hyperpolarization and a decrease in firing. (**A**), Composite bar graph that demonstrates the more hyperpolarized RMP of POMC neurons under fasting conditions. (**B**,**C**), Representative current clamp traces from POMC neurons showing the reversible PACAP-induced depolarization and increase in firing under ad libitum-fed conditions (*n* = 10) and the reversible hyperpolarization and suppression of firing seen under fasting conditions (*n* = 10). Comparable effects are seen during recordings in vehicle pre-treated slices from OVX females. (**D**,**E**), Composite data illustrating the PACAP-induced changes in membrane potential (∆V) and firing rate under ad libitum-fed and fasting conditions. Bars represent means and lines 1 SEM. \* *p* < 0.05, relative to ad libitum-fed conditions, Student's *t*-test (**D**); relative to baseline, Kruskal–Wallis/median-notched box-and-whisker analysis (**E**). Figure adapted from [12].

We corroborated these findings during optogenetic recordings in POMC neurons (Figure 5A,C,D) from PACAP-cre/eGFP POMC mice. Selective activation of VMN PACAP neurons (Figure 5B,E) elicited inward currents that were associated with depolarizations and increases in firing in the ad libitum-fed state (Figure 5F,H,J,L–N), and outward currents that were associated with hyperpolarizations and cessation of firing in the fasting state (Figure 5G,I,K–N; 5L: Student's *t*-test, *t* = 6.444, *p* < 0.0001; 5M: Student's *t*-test, *t* = 6.97, *p* < 0.0001; 5N: Kruskal–Wallis/median-notched box-and-whisker analysis, test statistic = 13.8222, *p* < 0.004).

statistic = 13.8222, *p* < 0.004).

effects are seen during recordings in vehicle pre-treated slices from OVX females. (**D**,**E**),

relative to ad libitum-fed conditions, Student's t-test (**D**); relative to baseline, Kruskal– Wallis/median-notched box-and-whisker analysis (**E**). Figure adapted from [12].

Composite data illustrating the PACAP-induced changes in membrane potential (ΔV) and firing rate under ad libitum-fed and fasting conditions. Bars represent means and lines 1 SEM. \* *p* < 0.05,

We corroborated these findings during optogenetic recordings in POMC neurons (Figure 5A,C,D) from PACAP-cre/eGFP POMC mice. Selective activation of VMN PACAP neurons (Figure 5B,E) elicited inward currents that were associated with depolarizations and increases in firing in the ad libitum-fed state (Figure 5F,H,J,L–N), and outward currents that were associated with hyperpolarizations and cessation of firing in the fasting

6.97, *p* < 0.0001; 5N: Kruskal–Wallis/median-notched box-and-whisker analysis, test

**Figure 5.** Optogenetic stimulation of VMN PACAP neurons depolarizes POMC neurons and increases their firing under ad libitum-fed conditions, effects which are flipped under fasting conditions. (**A**)**,** Low power (4×) image of ARC POMC neurons taken from a PACAP-Cre/eGFP POMC mouse. (**B**), Photomicrograph (4×) showing the channel rhodopsin-2 expression in VMN PACAP neurons two weeks after AAV injection as visualized by eYFP. (C), Differential interference contrast image (40×) of a recorded POMC neuron and the corresponding eGFP fluorescence signal from the same neuron (**D**). (**E**), 40X image showing the eYFP-labeled fibers in the immediate vicinity of the recorded neuron. Photostimulation (10-ms pulses delivered at 20 Hz for 10 s) of male VMN PACAP neurons produces a reversible inward current linked to membrane depolarization and increase in the firing of ARC POMC under ad libitum-fed conditions neurons (**F**,**H**,**J**,**L**–**N)**; *n* = 7–11), and the exact opposite is seen under fasting conditions **(G**,**I**,**K**–**N**); *n* = 11– 13**)**. Bars represent means, and lines 1 SEM of the light-induced change in ΔI (**L**), ΔV (**M**) and firing rate (**N**). \* *p* < 0.05, relative to ad libitum-fed conditions, Student's t-test (**L**,**M**); relative to baseline, Kruskal–Wallis/median-notched box-and-whisker analysis (**N**). Figure adapted from [12]. **Figure 5.** Optogenetic stimulation of VMN PACAP neurons depolarizes POMC neurons and increases their firing under ad libitum-fed conditions, effects which are flipped under fasting conditions. (**A**)**,** Low power (4×) image of ARC POMC neurons taken from a PACAP-Cre/eGFP POMC mouse. (**B**), Photomicrograph (4×) showing the channel rhodopsin-2 expression in VMN PACAP neurons two weeks after AAV injection as visualized by eYFP. (C), Differential interference contrast image (40×) of a recorded POMC neuron and the corresponding eGFP fluorescence signal from the same neuron (**D**). (**E**), 40X image showing the eYFP-labeled fibers in the immediate vicinity of the recorded neuron. Photostimulation (10-ms pulses delivered at 20 Hz for 10 s) of male VMN PACAP neurons produces a reversible inward current linked to membrane depolarization and increase in the firing of ARC POMC under ad libitum-fed conditions neurons (**F**,**H**,**J**,**L**–**N)**; *n* = 7–11), and the exact opposite is seen under fasting conditions **(G**,**I**,**K**–**N**); *n* = 11–13**)**. Bars represent means, and lines 1 SEM of the light-induced change in ∆I (**L**), ∆V (**M**) and firing rate (**N**). \* *p* < 0.05, relative to ad libitum-fed conditions, Student's *t*-test (**L**,**M**); relative to baseline, Kruskal–Wallis/median-notched box-and-whisker analysis (**N**). Figure adapted from [12].

Given that the tyrosine phosphatases PTP1B and TCPTP as well as AMPK figure prominently in orchestrating the cellular response to a state of negative energy balance, we then evaluated the role these signaling molecules play in the reversed polarity of the PACAP response in POMC neurons seen during fasting. In slices from fasted male eGFP-POMC mice we found that pretreatment with the PTP1B/TCPTP inhibitor CX08005 dramatically switched the PACAP response in POMC neurons from robust and reversible outward currents and hyperpolarizations (Figure 6A,C,E–I) to robust and reversible inward currents and depolarizations typically seen in the ad libitum-fed state (Figure 6B,D–I; 6E: Student's t-test, t = 8.194, *p* < 0.0001; 6F: Student's t-test, t = 6.878, *p* < 0.0001; Given that the tyrosine phosphatases PTP1B and TCPTP as well as AMPK figure prominently in orchestrating the cellular response to a state of negative energy balance, we then evaluated the role these signaling molecules play in the reversed polarity of the PACAP response in POMC neurons seen during fasting. In slices from fasted male eGFP-POMC mice we found that pretreatment with the PTP1B/TCPTP inhibitor CX08005 dramatically switched the PACAP response in POMC neurons from robust and reversible outward currents and hyperpolarizations (Figure 6A,C,E–I) to robust and reversible inward currents and depolarizations typically seen in the ad libitum-fed state (Figure 6B,D–I; 6E: Student's *t*-test, *t* = 8.194, *p* < 0.0001; 6F: Student's *t*-test, *t* = 6.878, *p* < 0.0001; 6G: Kruskal– Wallis/median-notched box-and-whisker analysis, test statistic = 13.5957, *p* < 0.004).

0.004).

0.004).

analysis (**G**).

6G: Kruskal–Wallis/median-notched box-and-whisker analysis, test statistic = 13.5957, *p* <

*Int. J. Mol. Sci.* **2021**, *22*, x 18 of 32

**Figure 6.** The fasting-induced switch in the polarity of the PACAP response in POMC neurons involves the activation of protein tyrosine phosphatases. Representative traces show that the PACAP-induced outward current (**A**; *n* = 12) and hyperpolarization (**C**; *n* = 10) seen under fasting conditions reverts back to an inward current (**B**; *n* = 9) and depolarization (**D**; *n* = 7) in the presence of the PTP1B/TCPTP inhibitor CX08005 (20 μM; **B**). This is further substantiated by the composite bar graphs in (**E**–**G**) as well as the pie charts in (**H**,**I**). Bars represent means, and lines 1 SEM of the PACAP-induced change ΔI, ΔV or normalized firing rate under fasted conditions, alone and in combination with CX08005, Compound C or metformin. \* *p* < 0.05, relative to PACAP alone, Student's t-test (**E,F**); relative to baseline, Kruskal–Wallis/median-notched box-and-whisker **Figure 6.** The fasting-induced switch in the polarity of the PACAP response in POMC neurons involves the activation of protein tyrosine phosphatases. Representative traces show that the PACAP-induced outward current (**A**; *n* = 12) and hyperpolarization (**C**; *n* = 10) seen under fasting conditions reverts back to an inward current (**B**; *n* = 9) and depolarization (**D**; *n* = 7) in the presence of the PTP1B/TCPTP inhibitor CX08005 (20 µM; **B**). This is further substantiated by the composite bar graphs in (**E**–**G**) as well as the pie charts in (**H**,**I**). Bars represent means, and lines 1 SEM of the PACAP-induced change ∆I, ∆V or normalized firing rate under fasted conditions, alone and in combination with CX08005, Compound C or metformin. \* *p* < 0.05, relative to PACAP alone, Student's *t*-test (**E,F**); relative to baseline, Kruskal–Wallis/median-notched box-and-whisker analysis (**G**). involves the activation of protein tyrosine phosphatases. Representative traces show that the PACAP-induced outward current (**A**; *n* = 12) and hyperpolarization (**C**; *n* = 10) seen under fasting conditions reverts back to an inward current (**B**; *n* = 9) and depolarization (**D**; *n* = 7) in the presence of the PTP1B/TCPTP inhibitor CX08005 (20 μM; **B**). This is further substantiated by the composite bar graphs in (**E**–**G**) as well as the pie charts in (**H**,**I**). Bars represent means, and lines 1 SEM of the PACAP-induced change ΔI, ΔV or normalized firing rate under fasted conditions, alone and in combination with CX08005, Compound C or metformin. \* *p* < 0.05, relative to PACAP alone, Student's t-test (**E,F**); relative to baseline, Kruskal–Wallis/median-notched box-and-whisker analysis (**G**).

**Figure 6.** The fasting-induced switch in the polarity of the PACAP response in POMC neurons

An equally dramatic switch in the polarity of the PACAP response in POMC neurons was observed during voltage clamp recordings in slices from both fasted eGFP-POMC mice that were pre-treated with AMPK inhibitor Compound C (Figure 7A,B,D–F), and from ad libitum-fed animals that were pre-treated with the AMPK activator metformin An equally dramatic switch in the polarity of the PACAP response in POMC neurons was observed during voltage clamp recordings in slices from both fasted eGFP-POMC mice that were pre-treated with AMPK inhibitor Compound C (Figure 7A,B,D–F), and from ad libitum-fed animals that were pre-treated with the AMPK activator metformin (Figure 7C,D,G; 7D: one-way ANOVA/LSD, F = 41.35, df = 2, *p* < 0.0001). An equally dramatic switch in the polarity of the PACAP response in POMC neurons was observed during voltage clamp recordings in slices from both fasted eGFP-POMC mice that were pre-treated with AMPK inhibitor Compound C (Figure 7A,B,D–F), and from ad libitum-fed animals that were pre-treated with the AMPK activator metformin (Figure 7C,D,G; 7D: one-way ANOVA/LSD, F = 41.35, df = 2, *p* < 0.0001).

(Figure 7C,D,G; 7D: one-way ANOVA/LSD, F = 41.35, df = 2, *p* < 0.0001).

**Figure 7.** The fasting-induced reversal of the PACAP response in POMC neurons is also dependent upon activation of AMPK. The outward current caused by PACAP under fasting conditions in POMC neurons (**A**; *n* = 12) is transformed into an inward current in the presence of the AMPK inhibitor Compound C (30 μM; **B**; *n* = 8) The PACAP-induced outward current in (**A**) was reproduced under ad libitum-fed conditions in the presence of the AMPK activator metformin (500 μM; **C**; *n* = 10). The data from these representative traces is summarized in **Figure 7.** The fasting-induced reversal of the PACAP response in POMC neurons is also dependent upon activation of AMPK. The outward current caused by PACAP under fasting conditions in POMC neurons (**A**; *n* = 12) is transformed into an inward current in the presence of the AMPK inhibitor Compound C (30 μM; **B**; *n* = 8) The PACAP-induced outward current in (**A**) was reproduced under ad libitum-fed conditions in the presence of the AMPK activator metformin (500 μM; **C**; *n* = 10). The data from these representative traces is summarized in composite form by the bar graph in (**D**) and the pie charts in (**E**–**G**). Bars represent means and **Figure 7.** The fasting-induced reversal of the PACAP response in POMC neurons is also dependent upon activation of AMPK. The outward current caused by PACAP under fasting conditions in POMC neurons (**A**; *n* = 12) is transformed into an inward current in the presence of the AMPK inhibitor Compound C (30 µM; **B**; *n* = 8) The PACAP-induced outward current in (**A**) was reproduced under ad libitum-fed conditions in the presence of the AMPK activator metformin (500 µM; **C**; *n* = 10). The data from these representative traces is summarized in composite form by the bar graph in (**D**) and the pie charts in (**E**–**G**). Bars represent means and lines 1 SEM. \*, *p* < 0.05 relative to PACAP alone, one-way ANOVA/LSD.

composite form by the bar graph in (**D**) and the pie charts in (**E**–**G**). Bars represent means and lines 1 SEM. \*, *p* < 0.05 relative to PACAP alone, one-way ANOVA/LSD. lines 1 SEM. \*, *p* < 0.05 relative to PACAP alone, one-way ANOVA/LSD. Consistent with these changes in cellular signaling, we also observed through immunohistofluorescent staining of coronal hypothalamic slices that fasting significantly increased the number of PTP1B- and pAMPK-positive ARC neurons as compared to that seen in the ad libitum-fed state (Figure 8; 8C: Student's *t*-test, *t* = 7.748, *p* < 0.002; 8F: Student's *t*-test, *t* = 5.598, *p* < 0.006).

Consistent with these changes in cellular signaling, we also observed through immunohistofluorescent staining of coronal hypothalamic slices that fasting significantly increased the number of PTP1B- and pAMPK-positive ARC neurons as compared to that seen in the ad libitum-fed state (Figure 8; 8C: Student's t-test, t = 7.748, *p* < 0.002; 8F:

Student's t-test, t = 5.598, *p* < 0.006).

**Figure 8.** Fasting augments the activity and expression of PTP1B and AMPK in the ARC. The four 20× images depict the PTP1B (**A**,**B**; 1:100) and pAMPK (**D**,**E**; 1:100) immunoreactivity under ad libitum-fed (**A**,**D**) and fasting (**B**,**E**) as visualized with AF546 (1:600). The composite data in the bar graphs summarize the fasting-induced increase in the number of PTP1B- (**C**) and pAMPKimmunoreactive (**F**) cells per capita in the ARC. Bars represent means and lines 1 SEM. \* *p* < 0.05, relative to ad libitum-fed conditions, Student's *t*-test. **Figure 8.** Fasting augments the activity and expression of PTP1B and AMPK in the ARC. The four 20× images depict the PTP1B (**A**,**B**; 1:100) and pAMPK (**D**,**E**; 1:100) immunoreactivity under ad libitum-fed (**A**,**D**) and fasting (**B**,**E**) as visualized with AF546 (1:600). The composite data in the bar graphs summarize the fasting-induced increase in the number of PTP1B- (**C**) and pAMPKimmunoreactive (**F**) cells per capita in the ARC. Bars represent means and lines 1 SEM. \* *p* < 0.05, relative to ad libitum-fed conditions, Student's *t*-test.

Moreover, the flipped PACAP-induced change in the excitability of POMC neurons yielded parallel alterations in energy intake. Indeed, the pronounced anorexigenic response generated by PACAP administered directly into the ARC of ad libitum-fed wildtype male mice (Figure 9A; repeated measures multifactorial ANOVA/LSD, Ftime = 45.21, df = 1, *p* < 0.0001; FPACACP = 71.82, df = 1, *p* < 0.0001; Finteraction = 0.72, df = 1, *p* < 0.40) was completely reversed in fasted animals, such that there was no change in cumulative energy intake at three hours post-injection and a significant increase six hours after administration (Figure 9B; repeated measures multifactorial ANOVA/LSD, Ftime = 28.18, df = 1, *p* < 0.0001; FPACAP = 4.32, df = 1, *p* < 0.04; Finteraction = 4.79, df = 1, *p* < 0.04; one-way ANOVA/LSD, F = 12.88, df = 3, *p* < 0.0001). A similar profile was observed in OVX wildtype female mice; with PACAP decreasing cumulative energy intake in ad libitum-fed, sesame oil vehicle-treated OVX females, which was significantly potentiated in estradiol benzoate (EB)-treated OVX females (Figure 9C; repeated measures multi-factorial ANOVA/LSD: Ftime = 417.17, df = 1, *p* < 0.0001), FPACAP = 9.00, df = 1, *p* < 0.004, FEB = 20.59, df = 1, *p* < 0.0001, Finteraction = 0.91, df = 1, *p* < 0.35). Conversely, the decrease in cumulative energy intake caused by PACAP in ad libitum-fed, vehicle-treated OVX females was once again completely reversed under fasting conditions. Surprisingly, EB per se failed to exert its prototypical anorexigenic effect in fasted OVX females, but it did abrogate the PACAPinduced increase in cumulative energy intake seen at six hours post-administration (Figure 9D; repeated measures multi-factorial ANOVA/LSD: Ftime = 74.67, df = 1, *p* < 0.0001), FPACAP = 8.64, df = 1, *p* < 0.004, FEB = 3.18, df = 1, *p* < 0.08, Finteraction = 14.88, df = 1, *p* < 0.0003; one-way ANOVA/LSD, F = 15.53, df = 7, *p* < 0.0001). Moreover, the flipped PACAP-induced change in the excitability of POMC neurons yielded parallel alterations in energy intake. Indeed, the pronounced anorexigenic response generated by PACAP administered directly into the ARC of ad libitum-fed wildtype male mice (Figure 9A; repeated measures multifactorial ANOVA/LSD, Ftime = 45.21, df = 1, *p* < 0.0001; FPACACP = 71.82, df = 1, *p* < 0.0001; Finteraction = 0.72, df = 1, *p* < 0.40) was completely reversed in fasted animals, such that there was no change in cumulative energy intake at three hours post-injection and a significant increase six hours after administration (Figure 9B; repeated measures multifactorial ANOVA/LSD, Ftime = 28.18, df = 1, *p* < 0.0001; FPACAP = 4.32, df = 1, *p* < 0.04; Finteraction = 4.79, df = 1, *p* < 0.04; one-way ANOVA/LSD, F = 12.88, df = 3, *p* < 0.0001). A similar profile was observed in OVX wildtype female mice; with PACAP decreasing cumulative energy intake in ad libitum-fed, sesame oil vehicle-treated OVX females, which was significantly potentiated in estradiol benzoate (EB)-treated OVX females (Figure 9C; repeated measures multi-factorial ANOVA/LSD: Ftime = 417.17, df = 1, *p* < 0.0001), FPACAP = 9.00, df = 1, *p* < 0.004, FEB = 20.59, df = 1, *p* < 0.0001, Finteraction = 0.91, df = 1, *p* < 0.35). Conversely, the decrease in cumulative energy intake caused by PACAP in ad libitum-fed, vehicle-treated OVX females was once again completely reversed under fasting conditions. Surprisingly, EB per se failed to exert its prototypical anorexigenic effect in fasted OVX females, but it did abrogate the PACAP-induced increase in cumulative energy intake seen at six hours post-administration (Figure 9D; repeated measures multi-factorial ANOVA/LSD: Ftime = 74.67, df = 1, *p* < 0.0001), FPACAP = 8.64, df = 1, *p* < 0.004, FEB = 3.18, df = 1, *p* < 0.08, Finteraction = 14.88, df = 1, *p* < 0.0003; one-way ANOVA/LSD, F = 15.53, df = 7, *p* < 0.0001).

**Figure 9.** Fasting completely reverses the effect of a direct injection of PACAP into the ARC on energy intake, which is attenuated by EB in OVX females. Under ad libitum-fed conditions, PACAP (30pmo; *n* = 6l) significantly decreases cumulative energy intake in wildtype males compared to saline-treated controls (0.2 μL; **A**; *n* = 6). This PACAP-induced decrease in energy intake is no longer apparent in fasted males, and PACAP actually causes an increase in cumulative consumption which is evident at six hours post-administration (**B**; *n* = 7–9). PACAP also decreases energy intake in ad libitum-fed OVX wildtype females, and this effect is potentiated by EB (20 μg/kg; s.c.; **C**; *n* = 6). Again, the effect of PACAP on consummatory behavior in fasted OVX females is exactly the opposite of that seen under ad libitum conditions, as is the modulatory effect of EB (**D**; *n* = 6). Bars represent means and lines 1 S.E.M. of the cumulative energy intake seen in ad libitum-fed or fasted mice injected with either PACAP or its saline vehicle. #, *p* < 0.05 relative to cumulative energy intake seen at three hours after PACAP injection, repeated-measures, multifactorial ANOVA/LSD; \*, *p* < 0.05 relative to saline vehicle, repeated measures, multi-factorial ANOVA/LSD; ^, *p* < 0.05 relative to sesame oil vehicle, repeated measures, multi-factorial ANOVA/LSD. Figure adapted from [12]. **Figure 9.** Fasting completely reverses the effect of a direct injection of PACAP into the ARC on energy intake, which is attenuated by EB in OVX females. Under ad libitum-fed conditions, PACAP (30pmo; *n* = 6l) significantly decreases cumulative energy intake in wildtype males compared to saline-treated controls (0.2 µL; **A**; *n* = 6). This PACAP-induced decrease in energy intake is no longer apparent in fasted males, and PACAP actually causes an increase in cumulative consumption which is evident at six hours post-administration (**B**; *n* = 7–9). PACAP also decreases energy intake in ad libitum-fed OVX wildtype females, and this effect is potentiated by EB (20 µg/kg; s.c.; **C**; *n* = 6). Again, the effect of PACAP on consummatory behavior in fasted OVX females is exactly the opposite of that seen under ad libitum conditions, as is the modulatory effect of EB (**D**; *n* = 6). Bars represent means and lines 1 S.E.M. of the cumulative energy intake seen in ad libitum-fed or fasted mice injected with either PACAP or its saline vehicle. #, *p* < 0.05 relative to cumulative energy intake seen at three hours after PACAP injection, repeated-measures, multi-factorial ANOVA/LSD; \*, *p* < 0.05 relative to saline vehicle, repeated measures, multi-factorial ANOVA/LSD; ˆ, *p* < 0.05 relative to sesame oil vehicle, repeated measures, multi-factorial ANOVA/LSD. Figure adapted from [12].

Taken together, this demonstrates that under conditions of negative energy balance, PAC1 receptor/effector coupling reverts from TRPC5 channel-induced excitation to KATP channel-induced inhibition, which completely reverses the effect of PACAP on energy intake. Thus, it is clear that postsynaptic PAC1 receptors at VMN PACAP/ARC POMC synapses effectively serve as metabolic switches that provide flexibility in the face of dynamic changes in energy status. Taken together, this demonstrates that under conditions of negative energy balance, PAC1 receptor/effector coupling reverts from TRPC5 channel-induced excitation to KATP channel-induced inhibition, which completely reverses the effect of PACAP on energy intake. Thus, it is clear that postsynaptic PAC1 receptors at VMN PACAP/ARC POMC synapses effectively serve as metabolic switches that provide flexibility in the face of dynamic changes in energy status.

#### **6. Concluding Remarks 6. Concluding Remarks**

To summarize, the N/OFQ/NOP and PACAP/PAC1 systems exert pleiotropic actions in the homeostatic and hedonic regulation of energy balance. In short, the N/OFQ/NOP To summarize, the N/OFQ/NOP and PACAP/PAC1 systems exert pleiotropic actions in the homeostatic and hedonic regulation of energy balance. In short, the N/OFQ/NOP

system elicits a net orexigenic effect via the homeostatic energy balance circuitry, and dampens the consumption of palatable food via inhibitory actions within the hedonic energy balance circuitry. These effects are sexually differentiated, accentuated by dietinduced obesity in males and hypoestrogenic females, and attenuated by E2 in OVX females. On the other hand, the PACAP/PAC1 system contributes a net anorexigenic effect via the homeostatic and hedonic energy balance circuitries. A thorough investigation of the anorexigenic VMN PACAP/ARC POMC synapse reveals that the neurophysiological and accompanying behavioral effects described above are diminished by diet-induced obesity in males, potentiated by E2 in OVX females, and completely reversed under fasting conditions. The manner in which alterations in energy balance status can influence NOP and PAC1 receptor-mediated signaling in POMC neurons is summarized schematically in Figure 10. system elicits a net orexigenic effect via the homeostatic energy balance circuitry, and dampens the consumption of palatable food via inhibitory actions within the hedonic energy balance circuitry. These effects are sexually differentiated, accentuated by dietinduced obesity in males and hypoestrogenic females, and attenuated by E2 in OVX females. On the other hand, the PACAP/PAC1 system contributes a net anorexigenic effect via the homeostatic and hedonic energy balance circuitries. A thorough investigation of the anorexigenic VMN PACAP/ARC POMC synapse reveals that the neurophysiological and accompanying behavioral effects described above are diminished by diet-induced obesity in males, potentiated by E2 in OVX females, and completely reversed under fasting conditions. The manner in which alterations in energy balance status can influence NOP and PAC1 receptor-mediated signaling in POMC neurons is summarized schematically in Figure 10.

**Figure 10.** Schematic diagrams illustrating how PAC1 and NOP receptor signaling in POMC neurons is altered under various energy balance states. (**A**), In POMC neurons from ad libitum-fed animals, N/OFQ activates the NOP receptor that initiates Gi/o-mediated signaling and subsequent activation of GIRK channels through positive allosteric modification by the βγ complex. This in turn promotes K+ efflux and inhibition of POMC neurons, effects which are dampened by E2 acting through ERα and Gq-mER to stimulate PI3K and nNOS as well as PLC, PKC and PKA signaling pathways, respectively. (**B**), In POMC neurons from obese animals, NOP receptor/effector coupling is enhanced; leading to a greater inhibitory effect of N/OFQ on POMC neurons. This N/OFQ-induced inhibition of POMC neurons is once again abrogated by E2 in POMC neurons from obese females. (**C**), Under ad libitum conditions, PACAP activates its cognate PAC1 receptor to elicit Gq-mediated signaling; working through PI3K as well as PLC, IP3, DAG and PKC to promote Ca2+ mobilization from intracellular stores and the coupling of PAC1 receptors to TRCP5 channels. This leads to cation flux through the channel pore that depolarizes and thereby excites POMC neurons. In females, E2 can act via ERα and Gq-mER to potentiate PAC1 receptor/TRPC5 channel coupling and PACAP-induced excitation of POMC neurons. (**D**), Under conditions of diet-induced obesity, the PAC1 receptor-mediated activation of TRPC5 channels in male POMC neurons is attenuated. However, in obese females the PACAP-induced excitation of POMC neurons is maintained due to the potentiating effect of E2. (**E**), Under conditions of fasting, the expression and activity of AMPK and protein tyrosine phosphatases like PTP1B is elevated in POMC neurons. This triggers a switch in the coupling of PAC1 receptors, such that they now are **Figure 10.** Schematic diagrams illustrating how PAC1 and NOP receptor signaling in POMC neurons is altered under various energy balance states. (**A**), In POMC neurons from ad libitum-fed animals, N/OFQ activates the NOP receptor that initiates Gi/o-mediated signaling and subsequent activation of GIRK channels through positive allosteric modification by the βγ complex. This in turn promotes K<sup>+</sup> efflux and inhibition of POMC neurons, effects which are dampened by E<sup>2</sup> acting through ERα and Gq-mER to stimulate PI3K and nNOS as well as PLC, PKC and PKA signaling pathways, respectively. (**B**), In POMC neurons from obese animals, NOP receptor/effector coupling is enhanced; leading to a greater inhibitory effect of N/OFQ on POMC neurons. This N/OFQ-induced inhibition of POMC neurons is once again abrogated by E<sup>2</sup> in POMC neurons from obese females. (**C**), Under ad libitum conditions, PACAP activates its cognate PAC1 receptor to elicit Gq-mediated signaling; working through PI3K as well as PLC, IP3, DAG and PKC to promote Ca2+ mobilization from intracellular stores and the coupling of PAC1 receptors to TRCP5 channels. This leads to cation flux through the channel pore that depolarizes and thereby excites POMC neurons. In females, E<sup>2</sup> can act via ERα and Gq-mER to potentiate PAC1 receptor/TRPC5 channel coupling and PACAP-induced excitation of POMC neurons. (**D**), Under conditions of diet-induced obesity, the PAC1 receptor-mediated activation of TRPC5 channels in male POMC neurons is attenuated. However, in obese females the PACAP-induced excitation of POMC neurons is maintained due to the potentiating effect of E<sup>2</sup> . (**E**), Under conditions of fasting, the expression and activity of AMPK and protein tyrosine phosphatases like PTP1B is elevated in POMC neurons. This triggers a switch in the coupling of PAC1 receptors, such that they now are no longer linked with TRPC5 channels and instead inhibit rather than excite POMC neurons via activation of KATP channels. This inhibitory effect of PACAP in POMC neurons from fasted animals is diminished by E<sup>2</sup> in females.

no longer linked with TRPC5 channels and instead inhibit rather than excite POMC neurons via

Over the past few decades, the neuroscience and neuroendocrine communities have made great strides in advancing our understanding of how the brain coordinates energy balance regulation via the homeostatic and hedonic energy balance circuitries. With the plethora of methodological tools currently at our disposal (e.g., optogenetics, chemogenetics, proteomics and transcriptomics, to name but a few), it is more than reasonable to expect that further advances will be readily and rapidly achieved. Just like the case made for PACAP in this very piece, it will be imperative to systematically evaluate all of the major players implicated in regulating energy homeostasis (including, but certainly not limited to, endocannabinoids, N/OFQ, NPY/AgRP neurons, POMC neurons, A10 dopamine neurons) not only under normophysiologic conditions but also under negative (e.g., fasting) and positive (e.g., obesity) energy balance states. Only in this way will we develop a comprehensive picture of how all of these functioning components are altered under these adaptive (e.g., fasting) and maladaptive (e.g., obesity) scenarios. In addition, it will be equally important to thoroughly dissect the impact of sex on the functioning of the central energy balance circuitries, and the cross-talk between gonadal hormones, the neurotransmitter/modulator systems and the signaling molecules involved. One final critical crowning achievement will be realized once we gain a firmer grasp on how the output from the homeostatic and hedonic energy balance circuitries functionally translates into clearly defined changes in the brain-gut axis and autonomic tone, as well as changes in gustatory and motivated behavior. We currently stand on the shoulders of pioneering giants who have paved the way by elucidating the mechanisms that provide the basis of our current understanding involved in the central control of energy homeostasis. In looking to the future, we welcome the next generation of innovative scientists to carry the torch and further advance our understanding.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/1422-006 7/22/5/2728/s13.S1: Methods and Materials.

**Author Contributions:** C.G. and J.H. performed all stereotaxic and survival surgeries. C.G. and J.H. performed all electrophysiological recordings. C.G., L.P., J.H., N.L., I.V. and S.S. performed all metabolic studies. C.G., J.H. and E.J.W. performed data analysis for all electrophysiology and metabolic studies. C.G., J.H., and E.J.W. created all figures and performed all statistical analyses. E.J.W. generated the manuscript, while all authors edited the final manuscript. E.J.W. and C.G. designed the experiments. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported by PHS Grant DA024314 and intramural funding from Western University of Health Sciences.

**Institutional Review Board Statement:** All care of animals and procedures were in accordance with the NIH Guide for the Care and Use of Laboratory Animals, and approved by the Western University of Health Sciences' IACUC (protocol code R18/IACUC/015,08/06/2020; R19/IACUC/027, 06/19/2020 )."

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available in Gastelum, C.; Perez, L.; Hernandez, J.; Le, N.; Vahrson, I.; Sayers, S.; Wagner, E.J. Adaptive Changes in the Central Control of Energy Homeostasis Occur in Response to Variations in Energy Status. Int. J. Mol. Sci. 2021, 22, 2728. doi.org/10.3390/ijms22052728.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Review* **The Microbiota and the Gut–Brain Axis in Controlling Food Intake and Energy Homeostasis**

**Marina Romaní-Pérez † , Clara Bullich-Vilarrubias † , Inmaculada López-Almela, Rebeca Liébana-García , Marta Olivares and Yolanda Sanz \***

> Microbial Ecology, Nutrition & Health Research Unit, Institute of Agrochemistry and Food Technology, National Research Council (IATA-CSIC), 46980 Valencia, Spain; marina.romani@iata.csic.es (M.R.-P.); clarabullich@iata.csic.es (C.B.-V.); inma.lopez@iata.csic.es (I.L.-A.); reliegar@iata.csic.es (R.L.-G.); m.olivares@iata.csic.es (M.O.)

**\*** Correspondence: yolsanz@iata.csic.es

† These authors contribute equally to this work.

**Abstract:** Obesity currently represents a major societal and health challenge worldwide. Its prevalence has reached epidemic proportions and trends continue to rise, reflecting the need for more effective preventive measures. Hypothalamic circuits that control energy homeostasis in response to food intake are interesting targets for body-weight management, for example, through interventions that reinforce the gut-to-brain nutrient signalling, whose malfunction contributes to obesity. Gut microbiota–diet interactions might interfere in nutrient sensing and signalling from the gut to the brain, where the information is processed to control energy homeostasis. This gut microbiota– brain crosstalk is mediated by metabolites, mainly short chain fatty acids, secondary bile acids or amino acids-derived metabolites and subcellular bacterial components. These activate gut–endocrine and/or neural-mediated pathways or pass to systemic circulation and then reach the brain. Feeding time and dietary composition are the main drivers of the gut microbiota structure and function. Therefore, aberrant feeding patterns or unhealthy diets might alter gut microbiota–diet interactions and modify nutrient availability and/or microbial ligands transmitting information from the gut to the brain in response to food intake, thus impairing energy homeostasis. Herein, we update the scientific evidence supporting that gut microbiota is a source of novel dietary and non-dietary biological products that may beneficially regulate gut-to-brain communication and, thus, improve metabolic health. Additionally, we evaluate how the feeding time and dietary composition modulate the gut microbiota and, thereby, the intraluminal availability of these biological products with potential effects on energy homeostasis. The review also identifies knowledge gaps and the advances required to clinically apply microbiome-based strategies to improve the gut–brain axis function and, thus, combat obesity.

**Keywords:** microbiota; gut–brain axis; nutrient sensing; food intake and obesity

## **1. Gut and Brain Control of Energy Homeostasis**

The gastrointestinal tract is a huge sensory organ able to transmit nutrient-related information to the brain where diverse endocrine and neural inputs converge to ultimately control feeding behaviour and whole-body energy homeostasis through efferent outputs. Nutrient sensors are receptors that bind molecules derived from the macronutrient digestion. They are located in enteroendocrine cells (EECs), which are the primary chemosensory cells in the gut, and are in direct contact with the luminal environment [1]. The activation of nutrient sensors of EECs initiates the secretion of gut hormones that in turn trigger the downstream processes required to maintain energy homeostasis postprandially. The most extensively studied gut hormones are cholecystokinin (CCK), gastric inhibitory polypeptide (GIP), mainly secreted in the upper part of the intestine, and glucagon-like peptide-1 (GLP-1) and peptide tyrosine tyrosine (PYY), mainly secreted in the distal part [2]. EECs

**Citation:** Romaní-Pérez, M.; Bullich-Vilarrubias, C.; López-Almela, I.; Liébana-García, R.; Olivares, M.; Sanz, Y. The Microbiota and the Gut–Brain Axis in Controlling Food Intake and Energy Homeostasis. *Int. J. Mol. Sci.* **2021**, *22*, 5830. https:// doi.org/10.3390/ijms22115830

Academic Editors: Rosalía Rodríguez-Rodríguez and Cristina Miralpeix

Received: 6 April 2021 Accepted: 26 May 2021 Published: 29 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

specifically sense carbohydrates, proteins and lipids through a diverse repertoire of nutrient sensors. Among others, sodium/glucose cotransporter 1 (SGTL1) mediates carbohydrate sensing, mainly sensed in the form of glucose, and, then, induces GIP and GLP-1 secretion; calcium-sensing receptor (CaSR) senses dietary amino acids and, then, secretes CCK and GLP-1; and G-protein coupled receptor (GPR) 120 (FFAR4), GPR40 (FFAR1) and GPR119 sense products from lipid digestion and induce the secretion of CCK, GIP, GLP-1 and PYY [1,3]. EECs are in close proximity to vagal afferent nerve terminals in the lamina propria, which express intestinal hormone receptors. Enteroendocrine hormones secreted after a meal activate nutrient-sensing signalling via endocrine routes, when gut hormones reach the brain or other organs through systemic circulation, or via paracrine routes, when hormones stimulate vagal afferents nearby in the intestinal mucosa and, then, the signal reaches the brain.

The arcuate nucleus (ARC) of the hypothalamus is accessible to humoral signals since is not fully protected by the blood–brain barrier [4]. The ARC contains two subpopulations of neurons; those expressing anorexigenic propiomelanocortin (POMC), the precursor of α-melanocyte-stimulating hormone (αMSH), and the cocaine and amphetamine regulated transcript (CART); and those neurons expressing the agouti gene-related peptide (AgRP) and neuropeptide Y (NPY) [5]. The activation of the POMC/CART neurons after feeding induces the release of α-melanocyte stimulating hormone (α-MSH) that binds to melanocortin 4 receptor (MC4R) in the paraventricular nucleus (PVN) where the information is integrated to suppress food intake and regulate body weight [6]. Thus, the increased levels of gut hormones such as CCK, GLP1 or PYY after a meal reach, via endocrine routes, the ARC to suppress food intake [7]. Via paracrine routes, gut hormones can stimulate vagal afferents [8], which are bipolar neurons whose somas converge in the nodose ganglion and their proximal extensions terminate in the nucleus of the solitary tract (NTS) in the brainstem [9]. This, together with the hypothalamus, represents an important integrative gut–brain hub. The NTS also contains POMC neurons and, through monoamine neurotransmission, transmits sensory information to upstream brain areas including the PVN and the dorsal vagal complex (DVC) that send efferent outputs involved in the vago– vagal reflex [5]. In summary, there are different paths for transmitting nutrient signals to the brain. The hypothalamus senses nutrients through the action of enteroendocrine hormones released in the intestine, which reach the brain by humoral pathways [10–12], that is through the circulatory system, or by paracrine pathways, that is by activating the nerve terminals of intestinal vagal afferents, which are the main focus of the present review.

The increased intake of energy-dense and palatable foods impairs the brain circuits controlling energy homeostasis, whose deficient response to nutrient signals alters feeding behaviour, which contributes to obesity. Accordingly, the restoration of nutrient signalling via the gut–brain axis represents a promising strategy to improve the central control of energy homeostasis in response to meals and, thus, help combat obesity [13–15]. The gut microbiota is a biological factor that might directly or indirectly influence nutrient-sensing and, theoretically, its modulation could aid in the restoration of gut-to-brain communication and maintaining energy homeostasis. The role of gut microbiota in obesity has been proven through faecal transplantation experiments, which produce the metabolic phenotype of the donor in the recipient organism [16]; however, the mechanisms by which microbiota influence energy homeostasis and body-weight regulation are not yet fully understood. Western diets, rich in saturated fat and simple sugars, alter the gut ecosystem reducing bacterial diversity and increasing the abundance of potential pathogens [17]. This, in turn, could alter the metabolism of macronutrients and, thereby, the ligands available for nutrient sensors in the intestinal lumen as well as the presence of structural microbial components that might also act as ligands of sensors that mediate the gut-to-brain communication. Sensors of microbially produced metabolites and bacterial components are located in EECs, vagal afferents and, occasionally, in the hypothalamus and can be activated by ligands reaching the brain through the systemic circulation. These receptors sense microbialderived metabolites such as short chain fatty acids (SCFAs), secondary bile acids (BAs) and

amino acid-derived metabolites and subcellular bacterial components such as caseinolytic peptidase B (ClpB), lipopolysaccharide (LPS) or muramyldipeptide (MDP).

Here, we review the role of dietary factors, including feeding patterns and diet composition, in modulating gut microbiota structure and function and potentially affecting gut-to-brain nutrient sensing, mainly via endocrine and neural routes. We also compile evidence of microbial molecules able to modulate specifically hypothalamic-mediated control of the energy homeostasis, placing special emphasis on their role in regulating food intake and, thereby, obesity.

#### **2. Circadian Rhythms, Eating Patterns and Gut Microbiota in Energy Homeostasis Control**

Most of the physiological functions display circadian rhythms. At cellular level, these are governed by clock proteins that generate self-sustained daily oscillations of the biological processes that allow cells to anticipate and thus to optimally respond and adapt to environmental variations [18,19]. In mammals, most cells throughout the body express clock genes that elicit autonomous circadian oscillations [20]. The hypothalamic clock located in the suprachiasmatic nucleus acts as a master pacemaker that synchronizes the secondary clocks [21,22]. Eating behaviour and energy metabolism show a well-defined circadian pattern through the day. In humans and rodents, the timed feeding patterns are governed by metabolic hormones and nutrients [23] and shift certain secondary clocks without affecting the suprachiasmatic nucleus, whose circadian oscillations are mainly governed by light–dark cycles [24,25]. Circadian rhythmicity, specially that affecting eating behaviour, is required to maintain energy homeostasis. Indeed, humans with disrupted circadian rhythms have higher risk of developing obesity and type 2 diabetes (T2D) [26,27] and mistimed eating increases adiposity in rodents [28] and also favours obesity and increases postprandial glycaemia in humans [29,30].

Circadian rhythmicity is involved in gut microbiota–host interactions. Gut microbiota play a role in maintaining the circadian rhythms of the host, including those related to eating patterns and energy homeostasis, and vice versa. The abundance of bacterial species, as well as of their derived metabolites, vary throughout the day, suggesting associations between gut microbes and eating behaviours in humans [31]. A recent study by Reitmeier et al. [32] supports the link between diurnal fluctuations of the gut microbiota and metabolic health in humans. The authors identified these diurnal oscillations in faecal samples of three independent large-scale population studies and detected aberrant oscillations in subjects with metabolic disorders such as obesity or T2D [32].

#### *2.1. Gut Microbiota Influences Circadian Rhythms Affecting Metabolism*

The influence of gut microbiota on the host circadian rhythms and on energy metabolism has been demonstrated in peripheral tissues such as the gut, liver or white adipose tissue. In mice, gut microbiota is required to maintain the circadian rhythmicity of the expression of the Toll-like receptors (TLRs), through which the microbiota-associated molecular patterns (MAMPs) communicate with the intestinal epithelial cells (IECs) of the host [33]. The absence of gut microbiota impairs the circadian clock of the IEC leading to an increased synthesis of ileal corticosterone, which impairs systemic metabolic homeostasis. This is evidence for the role of the microbiota–IEC dialogue oscillation during the day in metabolic health maintenance [33]. Similarly, germ free mice show altered expression of circadian-related genes in liver and white adipose tissue [34]. Microbiota-derived molecules, especially those that activate aryl hydrocarbon receptor (AHR) and pregnane X receptor (PXR), are crucial to maintain the rhythmicity and sexual dimorphism of the growth hormone secretion, which maintains rhythmic gene expression and metabolome in liver in a sex-dependent manner [34].

#### *2.2. Eating Patterns, Gut Microbial Diurnal Oscillations and Energy Homeostasis*

The eating rhythms of the host cause microbiota oscillations, which seem flexible and able to adapt to the nutritional environment. Mice deficient in the clock gene *Per1/2* or mice submitted to an experimental "jet-lag" show altered eating rhythms throughout the day linked to aberrant microbiota fluctuations [35]. Time-restricted feeding of *Per1/2* knock-out mice partially restores the microbiota oscillations, proving that eating behaviour is the main driver of microbiota diurnal oscillations [35]. A fat-rich diet also blunts the cyclical changes of the gut microbiota [36,37] and, consequently, their derived-metabolites, thus affecting host metabolism. For instance, the fluctuations of species belonging to the family Lachnospiraceae and their derived metabolite butyrate disappear under a high fat diet (HFD), which affects hepatic clock genes and metabolism [36]. Additionally, the spatial distribution of microbes in the gut seems to follow daily fluctuations. In rodents, bacterial adherence to the intestinal epithelium and their proximity to the mucosal surface are higher in the dark than in the light phase and, over the course of the day, some bacterial species show oscillations of epithelial adherence, findings that are dependent on feeding time [38]. Importantly, the nutrient-induced microbiota oscillations throughout the day lead to fluctuations of luminal and serum metabolites, such as amino acids and polyamines, responsible for the circadian hepatic transcriptome required for physiological processes such as detoxification functions [38]. Eating rhythmicity persists in antibiotic-induced microbiota depletion [38] but obese-associated hyperphagia might be transmissible through gut microbiota transplants to germ-free mice; this occurs in hyperphagic TLR5 knock-out mice [39] although not in *ob*/*ob* mice. This suggests that eating rhythmicity is driven mainly by the host circadian clock machinery driving endocrine hormone secretion, but nutrient signalling could also be modulated by gut microbiota activated pathways.

Interestingly, the diurnal bacterial growth dynamics, which are controlled by bacterial quorum sensing, availability of nutrients, gut motility and immunity, apparently overlap with host-feeding cycles [40]. In this regard, a bacterial growth-based model of appetite control has been proposed, in which the exponential and stationary growth phases of the bacteria are associated with satiety-induced signals of the prandial and postprandial phase, respectively, while the decline growth phase is coupled with hunger-related signals, defining intermeal intervals [40]. Therefore, to understand how the gut microbiota shapes eating rhythms it is necessary to investigate the bacterial growth fluctuations parallel to the generation of dietary and bacterially derived ligands of nutrient sensors able to centrally control energy homeostasis (such as SCFAs, secondary BAs or amino acid-derived metabolites—see Section 4 for details). For instance, ClpB, an antigen mimetic of αMSH produced by *Escherichia coli* [41], activates host satiety pathways following nutrient-induced bacterial growth [42]. Similarly, nutrient-induced bacterial growth might enhance the generation and release of bacterial ligands, such as cell wall and membrane components (e.g., lipopolysaccharide, LPS, or muramyldipeptide, MDP) or quorum sensing molecules, which in turn activate nutrient sensing pathways in the gut. Although the causal link between the diurnal oscillations of these molecules and feeding rhythms has yet to be proven, LPS and MDP are known to suppress food intake through immune sensing pathways linked to satiety and sickness behaviour [43–45]. Through an immune-related cascade in which LPS and MDP bind to TLR4 and NOD2, respectively, both molecules seem to modulate GLP-1-mediated signalling [46–48]. Nevertheless, the extent to which nutrient-induced bacterial growth and the consequent release of LPS or MDP synchronize meal-related host rhythms under physiological conditions is still unknown. Indeed, the effects of LPS and MDP on feeding behaviour have mainly been described in the context of bacterial infection, where these bacterial molecules reach the systemic circulation. However, their role as modulators of nutrient signalling in the gut, for instance as GLP-1 secretagogues, remains mostly unexplored. In this regard, LPS could act as an inducer of GLP-1 secretion when administered orally only if the gut barrier integrity is impaired [46]

#### **3. Diet Composition Influences Gut Microbiota and Gut-to-Brain Nutrient-Sensing**

Adherence to a particular dietary pattern, such as a Western, vegetarian or Mediterranean diet, or dietary interventions characterized by large variations in the macronutrients proportions, such as high protein or high fat (ketogenic) diets, impacts gut microbiota

composition and function [49–51]. The composition in macronutrients of these diets exerts an important effect on the availability of microbially derived ligands of dietary and non-dietary nature in the luminal content able trigger gut-to-brain sensing routes and controlling food intake and energy metabolism (summarized in Table 1). The availability of these ligands depends on multiple biological processes, including the microbiota-mediated catabolism of ingested nutrients and their absorption by enterocytes.

Most of the simple carbohydrates, proteins and fats are absorbed in the proximal regions of the small intestine while indigestible complex carbohydrates, the preferred carbon source of gut microbiota, reach the colon, favouring the growth of anaerobic bacteria and species diversity. The gut microbiota structure and composition are flexible, showing rapid adaptations to macronutrient shifts (24–48 h) that remain for short periods while more permanent changes might require longer adherence to a dietary pattern [52,53]. When the dietary macronutrient composition is high in non- or low-fermentable nutrients (e.g., lipids or proteins) these macronutrients overwhelm their intestinal absorption and reach the colon, where they are used by the best adapted microbes with the subsequent shift in their ability to activate gut nutrient sensing routes.

Although the diet per se modulates the energy metabolism, here we focus on effects possibly mediated by the interactions between the gut microbiota and the main macronutrients of different diets for which there is substantial scientific evidence. A special emphasis is placed on those interactions that potentially affect gut-to-brain nutrient sensing and, thereby, energy homeostasis.

#### *3.1. Western Diets*

Over the last two decades, the dietary patterns of modern societies have exhibited a shift towards a diet low in fibre and high in saturated fats, simple sugars and refined foods, termed the "Western diet" [54]. The Western diet is frequently associated with altered eating patterns, leading to hyperphagia and obesity onset. Specifically, the overconsumption of saturated fats and simple sugars in Western diets contributes to impair eating behaviour as a consequence of an overstimulation of the nutrient sensing routes that disturbs how brain senses these nutrient-related signals to control food intake patterns. Particularly, the intake of dietary fats amplifies meal sizes in humans; this phenomenon is exacerbated in obese subjects [55] that also showed reduced PYY secretion following a lipid load [56,57]. Studies in obese individuals reported increased postprandial levels of GLP-1 and CCK and occasional reductions of GLP-1 secretion [58,59] while investigations in rodents demonstrated that HFD diminished the sensitivity of GLP-1 and CCK contributing to alter eating behaviour and energy homeostasis [60,61]. In addition to the effects on eating behaviour, mouse studies also demonstrate that diets rich in saturated fats impair the enteric detection of glucose required to induce glycogen depots in muscle through a GLP-1 receptor mechanism in the arcuate NPY-expressing neurons [62].

Sugar-enriched foods also contribute to increasing weight gain as demonstrated by the meta-analyses of randomised controlled intervention trials and observational studies [63]. Excessive intake of sugar is associated with overeating due to impaired hedonic and homeostatic brain circuits [64–66] related to defective gut-to brain glucose sensing. In lean humans, calories from dietary sugars negatively correlate with glucose-induced GLP-1 secretion and positively with dorsal striatal food cue reactivity to palatable food [67]. Obesity in combination with increased dietary sugars has an additive effect reducing circulating levels of GLP-1 after a glucose load [68]. Functional magnetic resonance imaging also reveals that, compared with lean individuals, obese subjects show altered sucroserelated functional connectivity of lateral hypothalamus and NTS with reward-related brain areas resulting in reduced sucrose-associated hedonic responses [69] while studies in mice indicate that the Western diet impairs glucose sensing in POMC-expression neurons [70].

Together with the altered feeding behaviour and energy homeostasis, the Western diet also decreases the bacterial diversity in the human gut, partly due to its reduced content

of complex carbohydrates [71] and especially if diets are rich in high saturated fatty acids (SFAs) [72].

Compared with children from rural areas of Africa, European children eating a typical Western diet showed a higher abundance of Firmicutes with an overrepresentation of *Faecalibacterium*, while genera from Bacteroidetes, such as *Alistipes* and *Prevotella*, are increased and decreased, respectively [50]. These associations can reflect a metabolic adaptation of the intestinal bacteria to a new nutritional environment. How gut microbiota, through their interactions with Western diet-derived nutrients (i.e., saturated fats and simple sugars), influences the pool of bacterial-derived metabolites potentially activating gut nutrient sensing routes to control eating patterns and energy homeostasis is in an early stage of investigation. Herein, we compile associative studies in humans and mechanistic studies in rodents which provide evidence of these interactions affecting the central control of energy homeostasis.

In humans, short-term adaptations of the gut microbiota to an animal-based diet consist of a predominance of bile-tolerant microorganisms such as *Alistipes* or *Bilophila* and a decrease in bacteria adapted to metabolize dietary plant polysaccharide such *as Roseburia*, *Eubacterium rectale* and *Ruminococcus bromii* [53]. In vitro human gut simulator assays have demonstrated the predominance of lipid degradation-related genes and the downregulation of carbohydrate degradation genes, along with an increase in Gammaproteobacteria and the genera *Alistipes* and *Bilophila* in bacterial communities exposed to a fat-based medium [73]. In mice, a SFAs-enriched diet, especially high in palmitic acid, boosts overgrowth of *Bilophila wadsworthia*, which aggravates the HFD-induced metabolic disturbances [74]. The increased abundance of *B. wadsworthia* seems to be a secondary consequence of the SFAs on BAs metabolism [75]. Specifically, the SFAs-enriched diet increases the host production of BAs conjugated with taurine, which increase the fitness and growth of *B. wadsworthia*, while inhibiting other bacteria. *B. wadsworthia* also exacerbates the HFD-induced decrease in deoxycholic acid (DCA) and hyodeoxycholic acid (HDCA), secondary BAs that mediate gut nutrient sensing routes [75] (see Section 4 for details).

Studies in mice fed saturated fat-enriched diets combined with microbiota transplants or antibiotic treatments have provided insights into the causal relationship between specific bacterial genera or species and nutrient sensing routes involved in the central control of glucose homeostasis. For instance, the abundance of *Lactobacillus gasseri* (which is reduced by Western diets) helps to sense oleic and linoleic acid in the small intestine to ultimately reduce hepatic glucose production through downregulation of the bile acids receptor farnesoid acid receptor (FXR) and upregulation of the long-chain acyl-CoA synthetase-3 (ACSL3) [76]. In addition, the Western diet-induced lactobacilli decrease seems to be involved in the defective GLP-1 receptor/nitric oxide synthase (nNOS)-mediated signalling in response to oral glucose load to centrally control insulin secretion and gastric emptying, a process initiated by LPS and MDP [47].

Western diet also provides an excess of non-absorbable glucose and fructose to the small intestine and subsequently reach the colon. Gavage of [13C]fructose in mice reveals that high doses of fructose saturate the fructose-to-glucose conversion in the small intestine, enhancing fructose utilization by the microbiota in the colon via the hexokinase pathway to further generate tricarboxylic acid cycle (TCA) intermediates, such as essential amino acids (valine and leucine) and SCFAs (succinate, butyrate and acetate) [77,78]. In addition, sucrose-derived glucose and fructose in the colon downregulate a gene involved in gut bacteria colonization, reducing the fitness of bacterial species such as *Bacteroides thetaiotaomicron* associated with a lean phenotype by some studies [79]. The Western diet-induced decrease in the abundance of *Lactobacillus* spp. in the small intestine also impairs the glucose transporter SGLT1 and the GLP-1 receptor-mediated glucose sensing in the gut required to centrally control energy homeostasis in rats [80].

#### *3.2. Vegetarian Diets*

Vegetarian diets largely vary in composition according to the interindividual choice of foods. Commonly, they are devoid of meat and can include eggs and dairy products (lacto-ovo-vegetarian, lacto-vegetarian or ovo-vegetarian); exclude eggs and dairy products (vegan diets) or only include vegetables, fruit, nuts, seeds, legumes and sprouted grains (raw vegan diet) [81].

Compared with omnivorous diets, individuals adhering to vegan or vegetarian diets show less uncontrolled eating and emotional eating [82]. In the short-term, compared with a processed meat meal, a plant-based tofu meal enhances the secretion of GLP-1 in T2D individuals with a concomitant increase in satiety and a reduction of triglycerides in plasma [83,84].

Deciphering which are the vegetarian/vegan diet-associated nutrients and molecules that facilitate the gut-to brain nutrient sensing would provide valuable knowledge for designing effective anti-obesity dietary interventions for subjects who chose this dietary pattern. Overall, well-implemented vegetarian diets meet or exceed the recommended fibre intake [81,85]. A comparative study demonstrated that while the daily intake of sugars does not differ between vegan, vegetarian and omnivorous diets and the intake of proteins shows slight differences, subjects who adhered to a vegetarian/vegan diet consumed fewer calories and saturated and monounsaturated fats and higher amounts of dietary fibre [86], pointing out a main role of fibres in the ability of plant-based diets to modulate the gut–brain axis.

Dietary fibres, complex carbohydrates unable to be digested and absorbed in the upper part of the human intestine and which pass to the distal part, serve as substrate for intestinal microbes [87]. These complex polysaccharides, highly abundant in plant-based diets, per se display protective effects against the progression of obesity; moreover, new investigations suggest that the gut microbiota might be an intermediate player of the fibre benefits. Fibre fermentation generates diverse molecules including lactate, pyruvate and succinate as well as the SCFAs including acetate, propionate and butyrate in a molar ratio of 60:20:20, approximately [88]. Specifically, acetate is produced by phosphate acetyltransferase and acetate kinase; propionate is catalysed via the succinate, acrylate and propanediol pathways; and butyrate is produced by phosphate butyryltransferase and butyrate kinase [89].

Overall, an overrepresentation of the genera *Prevotella* and *Ruminococcus* is present in faeces of humans who adhere to plant-based diets [90]. Interventional studies in humans also show that different types of fibres favour the growth of bacteria with the enzymatic machinery directly involved in their fermentation. For instance, the abundance of *Bifidobacterium* spp., whose genome encodes transporters and enzymes involved in complex carbohydrate catabolism [91], are increased in the human colon as a result of dietary supplementation with inulin-type fructans [92], resistant starch [93], galactoologosacharides [94] or arabyno-oligosaccharides, which also increase *Prevotella* [95,96]. Some studies also associate the bifidogenic effect of the inulin-type fructans supplementation with improvements in oral glucose tolerance in obese women and reductions in body weight z-score in obese children [92]. Nonetheless, most of the human studies only show associations between specific bacterial taxa and metabolic benefits, and further studies are needed to demonstrate causality. Thus, it is also possible that the microbiota-mediated metabolic benefits of prebiotic fibres could depend on the initial microbiota configuration of the human subject and its capability to enhance the production of enough fermented products (e.g., SCFAs or others) and that this variation explains the inconsistency of the results across intervention trials [96,97].

Nevertheless, some studies in germ-free mice intentionally colonized with specific microbiotas have already demonstrated that, for example, plant-based diets induce microbiotadependent benefits. In particular, *Prevotella copri* seems to mediate the fibre-induced glucose tolerance improvement as revealed by a study conducted in germ-free mice colonized with the gut microbiota from individuals that favourably respond to barley kernel-based bread consumption [98]. An important route of communication between the gut and the

brain is the intestinal gluconeogenesis that, via portal glucose sensing, is essential for maintaining the postprandial regulation of the hypothalamic energy homeostasis control, including food intake and endogenous production of glucose [99]. It seems that *P. copri* could modulate the gut–brain axis to centrally control glucose homeostasis antagonizing the effects of other intestinal bacteria. Particularly, *P. copri* seems to limit the effects of *Bacteroides thetaiotaomicron* on energy metabolism impairment, possibly due to the ability of the latter bacterium to reduce colonic gluconeogenesis [98]. This idea was later demonstrated by de Vadder et al. 2016 [100] who evidenced the capacity of the *P. copri* colonization in mice to increase the intraluminal content of succinate, which could be used as substrate for the intestinal gluconeogenesis to then reduce hepatic glucose production via portal glucose sensing.

#### *3.3. Mediterranean Diet*

The Mediterranean diet is characterized by a high-level consumption of fruits, vegetables, grains, and fish and seafood as the main animal protein [101,102]. Scientific evidence supports that most of the benefits, mainly cardiovascular, of the Mediterranean diet are attributed to the high content of monounsaturated fatty acids (MUFAs), mainly oleic acid; ω-3 polyunsaturated fatty acids (PUFAs), mainly α-linolenic acid, eicosapentaenoic acid and docosahexaenoic acid; polyphenols, flavonoids and non-flavonoids, and fibre [103,104].

Contrary to SFAs, unsaturated fats are associated with metabolic benefits [105,106] and, accordingly, the Mediterranean diet is considered adequate to promote metabolic health without restricting total fat intake. Oleic and α-linolenic acid are GLP-1 secretagogues [107–109] and initiate lipid sensing routes to centrally control the endogenous production of glucose [110].

Dietary polyphenols present in vegetables and fruits might also display protective effects against obesity by the modulation of the hypothalamic function. Polyphenols can reach the brain and also initiate gut nutrient sensing routes. In the gut, polyphenols are able to induce the secretion of GLP-1, and PYY [111] in postprandial periods favouring the insulin-mediated glucose-lowering effects [112,113]. Whether or not these effects are mediated by the central nervous system should be further ascertain.

The Mediterranean diet also shapes the gut microbiota and the associated metabolome. Obese individuals adhere to a Mediterranean diet for 2 years show increases in the genera *Bacteroides*, *Prevotella*, and *Faecalibacterium*, and most importantly of *Roseburia* and *Ruminococcus* and the species *Parabacteroides distasonis* and *Faecalibacterium prausnitzii* [114]. Compared with Western diet non-human primate's consumers, Mediterranean diet consumers also show higher levels of genera *Lactobacillus*, *Clostridium* and *Oscillospira* [115]. The overrepresentation of these gut bacteria species could represent an adaptation of the microbial ecosystem to the higher preabsorptive abundance of PUFAs, polyphenols and complex carbohydrates, not absorbed in the upper gut, and thus able to interact with intestinal bacteria to modulate the pool of nutrient and microbe-derived metabolites activating the gut–brain axis.

Studies in mice reveal that, compared with saturated fats, intake of polyunsaturated fats limit the progression of obesity and induce different changes to the gut microbiota composition [116,117]. Overall lipids are not primarily digested by intestinal microbes but lipid-induced gut microbiota changes can equally influence how the gut senses nutrients. *Lactobacillus* spp. might be overrepresented under a diet rich in linoleic acid since they specifically develop resistance to its toxicity [118] and also use this ω6 fatty acid to produce PUFAs-derived metabolites such as 10-hydroxy-cis-12-octadecenoic acid (HYA) [119]. HYA improves metabolic health, suppressing food intake through lipid sensing-mediated signalling in L cells, in which GPR40 and/or GPR120 activation induces GLP-1 secretion in mice [119]. In addition, gut bacteria might influence lipid sensing routes and the exportation of fatty acids to extra-intestinal tissues by influencing the fatty acid storage in the enterocyte. For instance, *Lactobacillus paracasei* and *Escherichia coli* decrease the intestinal secretion of the absorbed oleic acid but through different metabolites derived from the

fermentation of complex carbohydrates [120]. Specifically, intestinal storage of oleic acid is promoted by *L. paracasei*-produced L-lactate by inhibiting malonyl-CoA-induced fatty-acid beta oxidation, a route activated by *E.coli*-produced acetate, which increases oleic acid degradation [120].

Additionally, gut microbiota favours the bioavailability of polyphenols from food in the intestinal lumen through complex multienzymatic reactions [121] modulating the polyphenol-associated gut nutrient sensing. To date, the identification of bacteria species specifically modulating polyphenol metabolic routes is under an early stage of investigation. Nevertheless, some evidence is available indicating, for instance, that *Bacillus subtilis*, which protects against obesity in mice [122], may produce protocatechuic acid from dietary quercetin in the human gut, which can virtually bind to the GLP-1 receptor as predicted by molecular docking simulations [123].

#### *3.4. Diets Based on Macronutrients Ratio Variations*

Herein we review the effects of diets used for weight loss, which are based on variations of the macronutrients proportions, such as high protein diets and low carbohydratehigh fat diets. We also focus on the potential role of these diets in the postprandial modulation of the gut–brain axis by changes in the pool of microbial and nutritional ligands in the intestinal lumen derived from the dietary macronutrients.

High protein diets are characterized by an increased intake of food rich in proteins (25–35% of energy compared with 12–18% of the standard protein diets) [124] and frequently associated with reduced carbohydrate consumption. These type of diets seem to be appropriate to rapidly induce weight loss [125]. Indeed, high protein diets positively regulate energy metabolism since, compared with other macronutrients, proteins strongly induce satiety and stimulate intestinal gluconeogenesis and thermogenesis [99,126]. Nevertheless, in the long-term, diets with different macronutrients ratios but the same energyrestricted content have similar effects on body weight maintenance, which could also be due to a poor long-term adherence to all diets [127–129].

In normal weight or obese human subjects, high protein meals induce the greatest satiety compared with isocaloric diets with high content of carbohydrates or fats, an effect that is dependent on PYY, the secretion of which is preferentially enhanced by proteins [130]. In addition to gut hormones, high protein diets modulate the gut–brain axis to control food intake and energy metabolism by stimulating the intestinal gluconeogenesis from postprandial to postabsorptive periods [99]. Digested peptides in the upper gut antagonize the µ-opioid receptors in the spinal and vagal afferents of the portal vein, a signalling that centrally activates the intestinal gluconeogenesis involved in the glucose sensing-induced food intake suppression [131,132]. Additionally, enriched-protein meals are a source of gluconeogenesis substrates for the gut such as glutamine and glutamate [99]. Compared with carbohydrates and fats, proteins also have the highest effect inducing thermogenesis [126]. Although the underlying mechanisms needs to be clarified specifically for proteins, some studies have identified thermogenesis-dependent gut–brain axis mechanisms mediated by gut hormones. For instance, GLP-1 centrally enhances thermogenesis through sympathetic efferents [133] and the duodenal hormone secretin postprandially activates the thermogenesis to induce satiety [134].

High protein diets also increase the amount of amino acids that can be fermented by gut microbiota in the colon to obtain energy and to produce nutrient sensing ligands. These include amino acid-derived SCFAs, branched chain fatty acids (BCFAs: isobutyrate, 2-methylbutyrate and isovalerate) and other molecules derived from tryptophan or glutamate, among others.

Compared with carbohydrates, the fermentation of proteins produces fewer SCFAs, although it still contributes substantially to microbial organic acid production. Amino acid-derived acetate is produced from glycine, alanine, threonine, glutamate, lysine and aspartate; butyrate is produced from glutamate and lysine and propionate from threonine [135].

Tryptophan is also biochemically transformed by gut microbiota, leading to the production of either serotonin (5-HT), indole, kynurenine or other derivative compounds [136]. Intestinal bacteria modulate the production of 5-HT in the gut directly or indirectly through microbe–host interactions. Members of the gut microbiota also possess tryptophanase activity, mediating the conversion of tryptophan into indole, which serve as interspecies signalling molecule that control bacterial physiology [137]. Indole-producing bacteria include, among others, species belonging to *Bacillus*, *Clostridium*, *Enterococcus*, *Bacteroides*, *Enterobacter*, *E. coli*, *Prevotella*, *Shigella* and *Vibrio* [138]. In addition, the gut microbiota modulates the expression of the rate-limiting host enzyme involved in the conversion of tryptophan to kynurenine, indoleamine 2,3-dioxygenase (IDO1), thus influencing the levels of kynurenine and its derivatives such as kynurenic acid [139]. Although some of these compounds might be toxic, such as indoxyl sulphate and quinolinic acid [140,141], others can potentially activate 5-HT receptors, aryl hydrocarbon receptor (AhR), and GPR35, which affect energy homeostasis (see Section 4 for details).

Intestinal bacteria can also decarboxylase glutamate, producing γ-aminobutyric acid (GABA) via the enzymatic activity of the glutamate decarboxylase, which helps to maintain the intracellular pH of the bacteria [142]. Strains belonging to the genera *Lactobacillus*, *Bifidobacterium*, *Lactococcus*, *Streptococcus*, *Escherichia*, *Listeria*, and *Aspergillus* have been reported to produce GABA [143–145], which also modulate nutrient sensing in the gut (see Section 4 for details).

Nevertheless, few studies address associative or causative links between the gut microbiota and high-protein diets on the control of food intake and energy homeostasis via the gut–brain axis.

Overall, diets high in proteins and low in carbohydrates reduce faecal abundance of SCFAs; i.e. butyrate [146], acetate and propionate [147,148], while increasing branch chain fatty acids; i.e., 2-methylbutyrate [146], isobutyrate and isovalerate [53,148,149]. Additionally, high intake of dietary proteins often increases the levels of kynurenic acid and indoxyl sulphate in plasma [150].

Concerning the effects on gut microbiota composition, interventional studies in humans with high protein diets initially suggest that variations (such as reductions of *Bifidobacterium*, *Roseburia* and *Eubacterium rectale*), were a consequence of the reduced intake of dietary fibres or the caloric restriction associated with these interventions [151]. More recent studies comparing the effects of different protein sources suggest that proteins also play a direct role in driving at least microbial-mediated metabolite changes with potential health impacts [146]. This study shows associations between bacterial abundance and faecal metabolites in overweight subjects conducting a 3-week high-protein diet intervention (either soy- or casein-based diet). For example, the study identified a positive correlation between *Oscillospira* and *Odoribacter* and amino acid-derived bacterial metabolites measured by targeted metabolomics and 1H nuclear magnetic resonance [146].

In humans, protein intake is positively associated with ClpB-like gene function [152] and in vitro studies demonstrate that, compared with other macronutrients (D-fructose and oleic acid), *E.coli* requires protein supplementation (bovine serum albumin) to increase the mRNA and protein levels of ClpB, which induces a dose-dependent stimulating effect on PYY secretion [153]. However, future investigations are needed to assess whether postprandial increases in ClpB coupled with bacterial growth account for the PYY-mediated food intake suppression in high protein diets in humans [130].

Larger-scale human interventions are needed to elucidate to what extent changes in microbiota-derived metabolites from dietary proteins affect nutrient sensing and, ultimately, control energy homeostasis, with an especial emphasis on intestinal gluconeogenesis and gut-mediated enhancement of thermogenesis. This will shed light on the link between protein-derived microbial metabolites and high protein-associated weight loss. Furthermore, additional studies are needed to assess the risk and benefits of high-protein dietary interventions in improving metabolic health, since these reduce butyrate production (the

main energy source for colonic enterocytes) and increase levels of mucosal and renal toxic compounds (e.g., tyrosine-derivate p-cresol and indoxyl sulphate) [151,154].

Low carbohydrate diets are frequently defined as diets with less than 20% of calories from carbohydrates with high content of fats (55–65%) and occasionally high in proteins (25–30%) [155]. These types of diets include ketogenic diets, frequently characterized by high content of fats, which were originally used to treat epilepsy due to their associated anticonvulsant effect driven by the increased production of ketone bodies and the modulation of GABA neurotransmission and mitochondrial metabolism [156]. Currently, these diets are also conceived for weight loss purposes. The metabolic benefits of ketogenic diets are based on their reduced capacity to postprandially increase glycemia and insulinemia and the concomitant enhancement of cellular catabolic routes using fats as the main source of energy. Accordingly, fat depots are reduced and gluconeogenesis, followed by ketogenesis, are also enhanced to supply glucose to cells [157]. These diets are effective to rapidly lose weight although long-term adherence poses challenges due to their associated contraindications [158]. Some studies reveal that ketogenic diets also induce less hunger and reduce the desire to eat in humans when comparing the appetite assessments before and during adherence to the diet [159]. Low carbohydrate diets also induce satiety in T2D patients [160] and, compared with diets with high content in carbohydrates, have lower capacity to stimulate food intake-related brain areas [161]. Nevertheless, the exact underlying mechanisms of the satiety effects caused by these diets remained elusive, especially those possibly dependent on gut microbiota.

Some studies indicate that diet-induced ketogenesis seems to mediate the reduced circulating levels of ghrelin associated with lower appetite in overweigh/obese individuals following a low energy diet [162] although ketone bodies inhibit the GLP-1 release by the EECs [163] and directly activate orexigenic hypothalamic routes in the brain [164]. Similar to SCFAs, ketone bodies also initiate GPR41 and GPR43 signalling to control energy metabolism. In particular, the acetate binds to GPR43 to induce lipid utilization in plasma [119] and β-hydroxybutyrate antagonizes GPR41 in sympathetic neurons to attenuate the sympathetic mediated metabolism [165]. To date, how these ketone bodies impact on the hypothalamus via the GPR41/43 has not been explored.

Ketogenic diets also change the gut microbiota structure and function in obese individuals with an overall decrease in butyrate and butyrate-producing bacteria, such as *Roseburia* spp. and *Eubacterium rectale* and *Bifidobacterium* spp. These changes are mainly attributed to the reduced content of the diet in complex carbohydrates and the concomitant reduction of SCFAs production [148].

Since ketogenic diets are also diets rich in fats, the identification of the main mechanisms through which these diets impact on the gut–brain axis through gut microbiotadependent mechanisms is complex. These mechanisms could be related to the reduction of SCFAs generated in the gut by the intestinal microbiota. Nevertheless, the role of other specific gut microbiota adaptations to these diets in shaping the specific pool of bioactive ligands and the existence of potential interactions with the levels of ketone bodies and their effects remain to be investigated.

In this regard, it is unknown whether the ketogenic diet-associated gut microbiota play a causative role in the postprandial control of the gut–brain axis. A recent study conducted in humans and mice identified that, compared with a conventional HFD, a ketogenic diet, high in fats and low in proteins, specifically reduces species from the genus *Bifidobacterium* by a direct effect of intestinal ketone bodies inhibiting their growth [166]. Additionally, the ketogenic diet displays metabolic improvements in mice, although these were attributed to a protective effect on the intestinal immunity rather than to the modulation of the gut–brain axis [166]. A direct role of ketogenic diets modulating the gut–brain axis was evidenced in another study in mice demonstrating that a carbohydrate-restricted diet with high content of fats favours cross-feeding between *Akkermansia muciniphila* and *Parabacteroides*, resulting in a reduced gamma-glutamyltranspeptidase activity in faeces driving an increased GABA/glutamate ratio in the hippocampus that in turn confers protection

against refractory epilepsy [167]. Although how ketogenic diets centrally control energy homeostasis through gut microbiota-mediated mechanisms remains uncertain, given the role of GABAergic neurotransmission on the central control of energy homeostasis and food intake (see Section 4 for details) it seems plausible that ketogenic diet modulates the gut–brain axis in postprandial periods by a similar gut microbiota-dependent mechanism operating in the hypothalamus. On the other hand, the gut microbiota, by modulating the lipid metabolism, seems to influence the circulating levels of β-hydroxybutyrate, which exerts neuroprotective effects on the brain, although this effect has not been explored for its relationship with metabolic health [168]. tection against refractory epilepsy [167]. Although how ketogenic diets centrally control energy homeostasis through gut microbiota-mediated mechanisms remains uncertain, given the role of GABAergic neurotransmission on the central control of energy homeostasis and food intake (see Section 4 for details) it seems plausible that ketogenic diet modulates the gut–brain axis in postprandial periods by a similar gut microbiota-dependent mechanism operating in the hypothalamus. On the other hand, the gut microbiota, by modulating the lipid metabolism, seems to influence the circulating levels of β-hydroxybutyrate, which exerts neuroprotective effects on the brain, although this effect has not been explored for its relationship with metabolic health [168].

axis was evidenced in another study in mice demonstrating that a carbohydrate-restricted diet with high content of fats favours cross-feeding between *Akkermansia muciniphila* and *Parabacteroides*, resulting in a reduced gamma-glutamyltranspeptidase activity in faeces driving an increased GABA/glutamate ratio in the hippocampus that in turn confers pro-

*Int. J. Mol. Sci.* **2021**, *22*, 5830 12 of 36

#### **4. Microbial Ligands Mediating Gut–Brain Communication and Energy Homeostasis 4. Microbial Ligands Mediating Gut–Brain Communication and Energy Homeostasis**

Here, we review the microbial products, including bacterial metabolites and bacterial cell components, that might impact on brain functions by modulating nutrient sensing signalling through enteroendocrine humoral and neural pathways, and could contribute to controlling energy homeostasis (Figure 1, Table 1). Beyond the control of food intake, gut microbiota might influence the whole-body energy metabolism by modulating the parasympathetic and sympathetic efferent tone [47,169], although this mechanism has been the subject of far fewer studies. Thus, this section focuses on the role of gut microbiota in controlling food intake and energy homeostasis, mainly through effects on the hypothalamusmediated food-intake suppression and, specially, in the postprandial periods. Here, we review the microbial products, including bacterial metabolites and bacterial cell components, that might impact on brain functions by modulating nutrient sensing signalling through enteroendocrine humoral and neural pathways, and could contribute to controlling energy homeostasis (Figure 1, Table 1). Beyond the control of food intake, gut microbiota might influence the whole-body energy metabolism by modulating the parasympathetic and sympathetic efferent tone [47,169], although this mechanism has been the subject of far fewer studies. Thus, this section focuses on the role of gut microbiota in controlling food intake and energy homeostasis, mainly through effects on the hypothalamus-mediated food-intake suppression and, specially, in the postprandial periods.

**Figure 1.** Bacterially produced metabolites from dietary nutrients and structural components of non-dietary origin modulate food intake in the brain (hypothalamus) through humoral and/or enterodocrine and neural signalling pathways. Here, we represent the pathways by which bacterial metabolites and non-dietary bacterial components (LPS, MDP and ClpB) induce an anorexigenic response in postprandial periods and a long-term food intake control. 5-HT3R, 5-hydroxytryptamine type 3 receptor; aa, amino acid; AgRP, agouti gene-related peptide; AHR, aryl hydrocarbon receptor; BA, bile acids; ClpB, caseinolytic peptidase B; GLP-1, glucagonlike peptide-1; GPR41/FFAR3, free fatty acid receptor 3; GPR43/FFAR2, free fatty acid receptor 2; IEC, intestinal epithelial cells; LPS, lipopolysaccharide; MDP, muramyldipeptide; NOD2, Nucleotide-binding oligomerization domain 2; NTS, nucleus tractus solitarius; PYY, peptide YY; POMC, proopiomelanocortin; SCFA, short-chain fatty acids; TGR5, takeda G protein-coupled receptor 5; TLR, Toll-like receptor. **Figure 1.** Bacterially produced metabolites from dietary nutrients and structural components of non-dietary origin modulate food intake in the brain (hypothalamus) through humoral and/or enterodocrine and neural signalling pathways. Here, we represent the pathways by which bacterial metabolites and non-dietary bacterial components (LPS, MDP and ClpB) induce an anorexigenic response in postprandial periods and a long-term food intake control. 5-HT3R, 5-hydroxytryptamine type 3 receptor; aa, amino acid; AgRP, agouti gene-related peptide; AHR, aryl hydrocarbon receptor; BA, bile acids; ClpB, caseinolytic peptidase B; GLP-1, glucagonlike peptide-1; GPR41/FFAR3, free fatty acid receptor 3; GPR43/FFAR2, free fatty acid receptor 2; IEC, intestinal epithelial cells; LPS, lipopolysaccharide; MDP, muramyldipeptide; NOD2, Nucleotidebinding oligomerization domain 2; NTS, nucleus tractus solitarius; PYY, peptide YY; POMC, proopiomelanocortin; SCFA, short-chain fatty acids; TGR5, takeda G protein-coupled receptor 5; TLR, Toll-like receptor.



**Table 1.** Main microbially derived ligands of dietary and non-dietary nature involved in gut-to brain nutrient sensing and control energy homeostasis.




#### *4.1. Short Chain Fatty Acids*

The function of SCFAs in protecting against metabolic alterations of diet-induced obesity is relatively well-established. Supplementing diet with fermentable carbohydrates, such as inulin, oligofructose or pectin, reduces food intake and improves HFD-induced glucose intolerance and weight gain in rodents and humans [170,221–225]. These effects are associated with SCFA production.

Although the three main SCFAs activate both GPR43/FFAR2 and GPR41/FFAR3, acetate is a potent agonist of FFAR2 [226], whereas propionate and butyrate have higher affinity for FFAR3 [226]. Both receptors are expressed in L-cells [227,228], and FFAR3 has been detected in enteric neurons [229] and in nodose ganglia cells [230]. Besides being agonists of FFAR2 and FFAR3, all SCFAs and lactate can potentially affect food intake by inhibiting the signalling of the orexigenic hormone ghrelin. In Hek293a cells stably expressing human growth hormone secretagogue receptor type 1a (GHSR1a), the three SCFAS (acetate, propionate, butyrate) and lactate reduced the Ca2+ influx in presence of ghrelin and showed antagonistic GHSR-1a properties as they attenuated the ghrelinmediated receptor internalization [231]. As this SCFAs-mediated signalling is still under an early stage of investigation (revised by [232]), here we focus on the role of SCFAs in food intake suppression, acting as FFAR2/FFAR3 agonists.

SCFAs are known to regulate food intake by modulating hypothalamic function, either reaching systemic circulation to the brain or via nutrient signalling mediated directly by GLP-1 and PYY generated in EECs or via vagal afferents. Among SCFAs, acetate seems to reach the brain through systemic routes, while propionate and butyrate mainly activate gut nutrient sensing pathways.

The analysis of biodistribution of 11C-acetate infused in the mouse colon indicates that acetate crosses the blood–brain barrier and reaches the hypothalamus, suppressing food intake short-term [171]. Moreover, acetate from fermentable fibre increases the hypothalamic neuronal activation in the ARC, but not in the ventral medial hypothalamus (VMH) or the PVN [170,171]. These changes are accompanied by an increase in hypothalamic activity of acetyl-CoA carboxylase and AMP-activated protein kinase (AMPK), subsequently inducing a downstream acute rise in POMC expression and reduction in AgRP expression [171,172]. In addition, systemic acetate enhances leptin release from adipocytes, possibly via an as yet undetermined FFAR2-dependent mechanism [173], thus contributing to a shift towards an anorexigenic neuronal activation pattern in the ARC [233]. Contrary to the abovementioned control diet-based research, a study in rats indicates that, under an energy-dense diet, gut microbiota enhance the production of acetate from glucose and fatty acids, contributing to diet-induced obesity [234]. Indeed, HFD increases acetate levels in the intestinal lumen and also systemically results in hyperphagia and an enhanced parasympathetic tone in β-cells, which impairs glucose stimulating insulin secretion [234]. The role of acetate in signalling through gut nutrient sensing routes is less evident. Some studies reveal that acetate does not stimulate GLP-1 or PYY secretion [171,235] and others report acetate-induced PYY secretion in the distal, but not proximal colon in a small cohort of obese humans without exploring its effects on food intake [236,237].

In overweight humans, the acute administration of inulin-propionate ester, which enables propionate delivery in the colon specifically, increases the postprandial secretion of GLP-1 and PYY along with reduced food intake, together with a non-significant trend to decrease long-term energy intake [225]. In mice, propionate also robustly stimulates both GLP-1 and PYY secretion either in vivo or in primary murine colonic cultures, probably via a FFAR2-dependent mechanism [174]. Additionally, propionate acts as a food intake suppressor through independent mechanisms of the endocrine actions of gut hormones. Indeed, independently of changes of the circulating levels of GLP-1 and PYY, non-obese healthy men receiving inulin-propionate showed reduced *ad libitum* food intake and lower oxygen level-dependent signal in reward-related brain structures [238]. Thus propionate can impact on brain function through vagal afferents innervating the gut and the portal vein. In this regard De Vadder and colleagues demonstrated that luminal propionate controls

the postprandial levels of glucose through a gut–brain loop [175]. In particular, this process is initiated by the propionate-induced activation of FFAR3 expressed in the afferents of the portal vein, which leads to the activation of neurons in brain regions receiving vagal and spinal inputs, i.e., DVC, spinal C1 segment and the parabrachial nucleus (PBN), respectively, as well as in the hypothalamic areas receiving inputs from DVC and PBN, i.e., PVN, the lateral, hypothalamus (LH) and the ARC. Apparently, propionate centrally controls the intestinal gluconeogenesis, a process that induces metabolic benefits [132,176], including the reduction of the endogenous production of glucose independently of insulin [175].

Furthermore, like acetate, elevated plasma propionate levels induce leptin release from adipocytes via the somewhat controversial FFAR2- or FFAR3-dependent mechanism [173,177], promoting hypothalamic anorexigenic neuronal activation [233].

Among the three main SCFAs, butyrate may be the strongest stimulator of anorexigenic peptides [178] and the most potent suppressor of food intake [179]. In obese and healthy humans, all three main SCFAs separately administered increase plasma levels of PYY in both fasting and postprandial conditions [236]. In mice, oral, but not intravenous, administration of butyrate reduces food intake, according to decreased neuronal activity in NTS and DVC [180]. Additionally, butyrate influences the hypothalamic circuitry, suppressing the activity of orexigenic neurons [180], decreasing AgRP expression [181] and enhancing POMC expression [181]. In addition, subdiaphragmatic vagotomy eradicates the anorectic effects of butyrate, suggesting that the vagal nerve is necessary to convey the short- and long-term satiety signalling of acute or chronic administration of butyrate in the context of obesity [180]. Moreover, intraperitoneal butyrate induces activation of nodose ganglia neurons [179]. Intestinal butyrate may transmit satietogenic signals by stimulating GLP-1 and GIP secretion in L-cells and K-cells, respectively [178], and these effects are probably reinforced by ghrelin inhibition [181]. It is not clear whether butyrate induces PYY secretion, since studies report no changes [181], decrease [235] or slight increase [178] in PYY release as a result of butyrate increase in the caecal content of animals orally receiving a probiotic, a pea protein supplementation or oral butyrate administration. Furthermore, it is not yet clear whether butyrate-induced gut hormone secretion is mediated by a FFAR3 dependent mechanism [178,181]. Therefore, the evidence about the mechanism through which SCFA are involved in satiety signalling is not fully consistent, probably due to experimental differences, for instance, in hormone measurement timings, types of dietary fibre supplementation, administration site and the use of different animal models.

#### *4.2. Microbial Metabolites of Bile Acids*

Bile acids (BA) are steroid acids synthesized in the liver from cholesterol, conjugated to either taurine or glycine and postprandially released in the duodenum to facilitate the absorption of dietary lipids and fat-soluble vitamins [239–241]. The majority of primary BA secreted in the intestine are actively reabsorbed in the ileum and transported back through the portal circulation to the liver (enterohepatic circulation) [242]. The remaining small portion of primary BA are deconjugated and dehydroxylated in the ileum and colon by intestinal bacteria into secondary BAs [243,244], mainly DCA and lithocholic acid (LCA) in humans and rodents, and murideoxycholic acid in rodents [239,245]. Bacterial bile salt hydrolases (BSH) are essential enzymes for deconjugation of primary BA to secondary BAs [190]. In humans, BSH genes and enzymes have been characterized in Grampositive bacterial species of the genera *Lactobacillus* [182,183,190], *Bifidobacterium* [184,190], *Enterococcus* [185,186], *Clostridium* [187,190] and *Listeria* [188,190] and in Gram-negative bacteria of the genus *Bacteroides* [189,190].

Besides their role in lipid digestion, BAs have recently been characterized as ligands of takeda G protein-coupled receptor 5 (TGR5), regulating lipid and glucose metabolism once activated [246–248]. Dietary macronutrients and feeding patterns influence the composition and secretion of bile acids, respectively, and thus the microbiota-mediated bile acids' effects on the host. Interventional studies in humans reveal that a fibre-rich low-fat diet is associated with low levels of all secondary BAs, and both glycine- and taurine-

conjugated bile acids in faeces [249]. By contrast, a high fat diet increases the abundance of unconjugated and secondary BAs (DCA, TDCA, 12keto-LCA, 3b-DCA and TLCA) coupled with changes in species belonging to the genera *Bacteroides*, *Clostridium*, *Bifidobacterium* and *Lactobacillus* and a positive correlation between *Bacteroides* and DCA [250].

Studies in humans and rodents report that microbial BAs modulate energy homeostasis directly by reaching the brain through systemic circulation and/or activating gut nutrient sensing routes. The main functional effects of BAs reported are on feeding behaviour or glucose homeostasis. Plasma BAs levels notably rise postprandially in humans [251,252] and rodents [253], thus fluctuating in the systemic circulation along with the circadian rhythm upon food intake, suggesting a role of BA as lipid-sensing molecules and short-term satiety signals that reach the brain through a humoral route. Diet fat content is positively associated with plasma BAs concentrations in humans [252] and mice [191], and with hypothalamic BAs concentrations in mice [191]. Interestingly, DCA and other BAs have been detected in the rat brain [254,255], and their levels are positively correlated with plasmatic levels [254], indicating that secondary BAs may reach the brain from the intestine by diffusion across the blood–brain barrier [254]. Accordingly, TGR5 expression has been detected in neurons [192] and more recently in the hypothalamus related to glucose metabolism [191]. Altogether, this evidence suggests secondary BAs generated by gut microbiota activity potentially play a role in the direct control of hypothalamic function.

In addition, secondary BAs activate enteroendocrine and/or neural gut nutrient sensing routes. Studies in mice demonstrate that downregulation of TGR5 in the context of diet-induced obesity impairs glucose tolerance [193] and increases food intake coupled with a substantial decrease in the plasma levels of GLP-1 and PYY [194,195]. By contrast, TGR5 overexpression in mice improves oral glucose tolerance by inducing a marked enhancement of the postprandial secretion of GLP-1, probably due to increased BAs flow after a highly lipidic meal, [193]. Gut bile acid sensing is also supported by other in vitro [196] and in vivo [195,197] studies in mice showing that secondary BAs trigger GLP-1 and PYY secretion from EECs via TGR5 activation. Human colonic GLP-1-producing EECs also express TGR5, the activation of which by TDCA increases the secretion of GLP-1 [256]. Importantly, obese subjects with or without diabetes show an impaired BA metabolic pathway [256] which might contribute to aberrant GLP-1 secretion [257]. Similarly, experiments in mice demonstrate that secondary BAs induce the secretion of 5-HT [198] subsequently activating 5-HT3 receptors on vagal afferent terminals [199]. The impact of secondary BAs resulting from gut microbiota activity on the brain is still understudied [247] but secondary BAs, through TGR5-GLP-1 and/or 5-HT gut sensing pathways, can potentially modulate food intake and energy homeostasis via efferent routes. Specifically, the role of secondary BAs in inducing satiety through TGR5-activation in vagal afferent neurons has been demonstrated. Indeed, TGR5 expression has been detected in the nodose ganglia cells colocalizing with CCK-1R in rats [200]. Moreover, vagal TGR5 mediates DCA-specific activation of POMC and CART-expressing neurons, but not orexigenic neurons, in the hypothalamus resulting in reduced spontaneous postprandial food intake [200]. Additionally, silencing TGR5 and CCK-1R in the nodose ganglia has an additive effect increasing spontaneous food intake and suggesting a synergistic effect of BA and CCK in hypothalamic short-term satiety signalling via the vagal pathway [200].

Therefore, the bacterial BAs bioconversion capacity may have a strong impact on the intestinal availability of secondary BAs [258,259]. Subsequently, the postprandial enhancement of the TGR5-mediated signalling may be affected by these secondary BAs generated in the intestine as well as by the BAs pool that reaches the brain [247], contributing to the modulation of hypothalamic function. However, to our knowledge, there are currently no studies reporting the effects of specific bacterial species or strains on BA sensing to centrally control energy homeostasis.

#### *4.3. Amino Acid Derived Metabolites*

In the gut, as well as in the brain, 5-HT is synthesized from tryptophan [260], an essential amino acid obtained from dietary proteins [261]. Studies in rats demonstrate that 5-HT is released by enterochromaffin cells (ECs) in response to nutrient stimuli such as carbohydrates, glucose or lipids [262,263]. 5-HT binds to 5-HT3 receptors in the vagal afferents, acting thus as sensory transducers [262]. Functionality, the activation of 5-HT signalling in vagal afferents mediates glucose induced inhibition of gastric emptying and lipid-induced food intake suppression, which occurs simultaneously with CCK1 signalling [263,264]. Similarly, in humans, inhibition of 5-HT3 receptor increases the liquid meal ingested [265].

*Hafnia alvei*, *E. coli* K-12 and species and strains of the genera *Lactococcus*, *Lactobacillus*, *Streptococcus* and *Klebsiella* have been shown in vitro to synthesize 5-HT from tryptophan [266]. The mechanisms by which gut bacteria directly synthesize 5-HT are not clearly established as bacteria lack tryptophan hydroxylase 1 (TPH1), an essential enzyme for 5-HT biosynthesis. Therefore, an alternative mechanism has been proposed, in which tryptophan is decarboxylated to tryptamine [267]. The role of gut microbiota as an enhancer of 5-HT biosynthesis and release from ECs may partly occur through host–microbe interactions mediated by bacterial structural components or metabolites according to certain reports. For example, *Escherichia coli* modulates the host tryptophan hydroxylase-1 activity, enhancing the synthesis of the 5-HT precursor 5-hydroxytryptophan (5-HTP) and the extra-cellular levels of 5-HT through interaction with secreted host derived factors [268]. This strain also contributes to 5-HT clearance by its uptake into enterocytes through the serotonin transporter SERT, which is required for completion of the serotonin circuit synthesis. Other studies show that spore-forming microbes from a healthy human increase 5-HT in colon and blood, through a process that seems to involve host interactions with bacterially produced metabolites such as deoxycholate and SCFAs [269,270]. Probably acting through SCFAs, gut microbiota upregulates the expression of THP1 in colonocytes as demonstrated by the colonization of germ-free mice with mouse or human gut microbiota [270].

A recent study reveals that ECs differentially express chemical sensors according to their spatial location in the gut. Surprisingly, in the small intestine, ECs do not express nutrient sensors but might indirectly release 5-HT in response to nutrient stimuli through the paracrine action of GLP-1, from the neighbour EECs [271]. By contrast, compared with small intestine ECs, colonic ECs present higher expression of microbial sensors such as TGR5, which binds to secondary BAs; GPR132, to acyl amides and lactate; GPR35, to aromatic acids, OLFR558, to isovalerate and FFAR2, to SCFAs, all able to trigger 5-HT release [271]. Accordingly, gut microbiota, through their metabolites, would induce the release of 5-HT in colon and indirectly in the small intestine by stimulating GLP-1 secretion. In fact, although 5-HT cannot cross the blood–brain barrier [139], peripheral 5-HT has been demonstrated to reduce food intake [272–274], suggesting an alternative signalling pathway for the effects of gut-derived 5-HT on the regulation of appetite [275,276].

Studies indicate that, in contrast to 5-HT, its precursor 5-HTP can cross the blood–brain barrier, probably leading to 5-HT production in the brain [139,277,278] and accounting for the role of gut-derived 5-HTP in eliciting satiety. In fact, 5-HTP biosynthesis is enhanced in ECs by gut microbiota-derived SCFAs by upregulating expression of TPH1 as indicated above [278]. Thus, these findings support a role of the gut microbiota-induced 5-HT and 5-HTP in food intake, although further investigations are needed to specifically identify the postprandial microbially produced metabolites and structural products potentially influencing feeding behaviour via this route.

Tryptophan can be fermented into indole [201] which is a ligand of aryl hydrocarbon receptor (AHR), a transcription factor that regulates gene expression. In vitro experiments with GLUTag cell lines show that activation of AHR with an agonist increases the expression of proglucagon as well as GLP-1 secretion [202]. In line with this, experiments with primary mouse colonic L cells demonstrate that indole elicits rapid GLP-1 release during short exposure but that release is inhibited over longer periods and at lower doses [203]. How

these actions are coordinated with eating patterns remains unknown; however, the authors proposed that indole at low concentration suppress GLP-1 from L colonic cells but induces an opposite effect in response to high-protein diet. Although indole-mediated nutrient sensing signalling is in an early research stage, the evidence above suggests that AHR might act as a sensor of microbially produced metabolites capable of triggering GLP-1-mediated nutrient signalling from the L cells of the gut to the brain.

In mice, obesity has been associated with increased intestinal IDO1, an enzyme catalysing tryptophan degradation via the kynurenine pathway, thus limiting the bacterial production of indole from tryptophan and increasing kynurenine and its derivatives such as kynurenic acid [279]. In mice, intraperitoneal administration of kynurenic acid induces energy expenditure without affecting locomotion or food intake [280]. This effect seems to be dependent on the activation of GPR35 in the adipose tissue [280], a receptor that has also been detected in sensory neurons of the extrinsic intestinal innervations [281], suggesting that kynurenic acid could play a role in gut-to-brain sensory transmission, which is as yet unexplored.

To our knowledge, so far, studies are lacking on the impact of GABA produced by gut microbes from the dietary amino acid glutamate on the gut–brain communication and, thus, on the control of energy homeostasis. Therefore, the pathways by which microbially produced GABA may influence the hypothalamus function can only be hypothesized from unconnected observations. GABA is the main inhibitory neurotransmitter in the central nervous system. Peripheral GABA cannot cross the blood–brain barrier [282], but activates gut nutrient sensing signalling pathways. First, GABA has been demonstrated to stimulate GLP-1 release from the EECs line GLUTag [204]. Moreover, GABAB receptors are expressed along the gastrointestinal tract [283,284], and co-localized with 5-HT-producing cells identified as morphologically similar to EECs [284], probably inhibiting 5-HT-release by ECs [285]. Second, GABA transmits sensory information by activating GABAA [205] and GABAB receptors [206] expressed in vagal afferents or by secreting exosomes from GABA-stimulated intestinal cells that, in turn, activate neurons [207]. Nonetheless, specific studies are needed to understand the contribution of gut microbiota-derived GABA to nutrient sensing from the gut to the brain and, thus, to the control energy homeostasis in both lean and obese subjects.

#### *4.4. Cellular Components of Gut Bacteria*

Certain cellular components of intestinal bacteria stimulate gut-to-brain routes of communication, which may be involved in the central control of energy homeostasis, especially through food intake modulation. This is the case for protein fragments of gut bacteria displaying molecular mimicry with human appetite-regulatory peptides and neuropeptides; i.e., protein fragments of the human α-MSH sequence being identical to gut bacterial-derived proteins from *E. coli, Bifidobacterium longum*, *Bacillus cereus*, and certain enteropathogenic bacterial strains [286]. In particular, ClpB produced by *E. coli* K12 is a mimetic of the anorexigenic peptide α-MSH [41], which has a discontinuous five-amino acid overlap containing part of the α-MSH sequence [287]. ClpB is a heat-shock protein with ATPase and chaperon activity mediating the resolubilization of heat-denaturated protein aggregates, thus having a protective function for bacterial cell-induced damage. More recently, the α-MSH-like motif identified within ClpB protein has been confirmed to be specific to the order Enterobacteriales and conserved among its taxa, including *E. coli* strains and *Hafnia* genus [209].

Recent preclinical and clinical studies have demonstrated the effectiveness of ClpB protein on modulating energy homeostasis. In humans, shotgun metagenomic analysis of faeces revealed that gut microbiota ClpB gene, which is associated with decreased weight gain [152], was lower in obese compared to lean individuals [152]. In obese mice, oral administration of the native *E.coli* K12 strain, unlike the ClpB-deficient strain, decreased body weight gain [209], while its food-intake suppression activity was lost when the ClpBdeficient *E. coli* was administrated [41]. In addition, intragastric administration of *Hafnia*

*alvei* HA4597, a ClpB-producer, reduces caloric intake and increases lipolytic effects in *ob*/*ob* mice coupled with reduced fat mass and body weight gain in both genetically and diet-induced obese mice [209]. Along with the aforementioned effects, Lucas et al. found a decrease in glycaemia, plasma cholesterol and alanine aminotransferase, a marker of obesity-induced steatohepatitis [208]. Mechanistically, ClpB modulates the hypothalamic circuit that controls food intake. In particular, AgRP expression is decreased in obese mice administered the ClpB-producer *H. alvei* [209], while in an animal model of anorexia increased ClpB plasma levels were associated with increased POMC expression [210]. Moreover, *bdnf* mRNA levels increased in mice receiving *E. coli* proteins collected at stationary phase, which include ClpB [42]. These findings prove the contribution of ClpB in upregulating the expression of hypothalamic anorexigenic neuropeptides.

In addition, it is known that ClpB transmits nutrient information from the gut to the brain, directly reaching the hypothalamus when traveling in the systemic circulation, or indirectly by the activation of gut nutrient sensing signalling. The presence of ClpB has been reported in the plasma and in the hypothalamus of rodents and healthy humans [211,288]. Moreover, ClpB protein plasma levels in obese mice are increased by treatment with the ClpB-producer *H. alvei* HA4597, in accordance with food intake reduction [209] and patients with eating disorders show elevated ClpB plasma concentrations compared to healthy individuals [288]. Ex vivo electrophysiological experiments also reveal that ClpB from the stationary stage of *E.coli* directly stimulated the firing rate of POMC neurons, while systemic administration of *E.coli* in the stationary stage increased cFos immunolabelled POMC neurons in the ARC and VMH [42].

ClpB has also been found in the colon of mice, rats, and healthy humans [211]. ClpB stimulates PYY secretion in cultured rat intestinal mucosa in a dose-dependent manner [211] and the ClpB-induced PYY secretion may be enhanced by nutrient-induced bacterial growth [42]. Accordingly, in response to food intake, ClpB initiates a PYY-mediated endocrine or neural nutrient sensing signalling that ultimately regulate the hypothalamic function in suppressing food intake. However, the receptor linking the ClpB agonist with PYY secretion has yet to be identified [153].

A synthetic fragment of ClpB has partial agonist activity on MC3R and MC1R, but not on MC4R [287], whose activation by α-MSH induces PYY secretion [289]. However, the ability of the ClpB-α-MSH-like motif to activate MCRs [287] supports a spatial complementarity of bacterial-derived ClpB and MCRs and, therefore, further investigations regarding the receptors by which ClpB drives its satiating effects are required.

Overall, ClpB is suggested to have a satietogenic effect by systemic and/or neural/endocrine routes. By a systemic route, plasmatic ClpB depends on the quantity of ClpB-producing bacteria in the intestine and, therefore, plasmatic ClpB might be a longterm satiety signal by modulating hypothalamic neuropeptide expression, ultimately influencing meal patterns. By a neural/endocrine route, ClpB might also be a short-term meal termination signal by stimulating release of satiety hormones in the intestine.

Some studies also indicate that LPS and MDP may play a role as food intake modulators. LPS is the outer membrane's major component of Gram-negative bacteria [290]. In obesity, LPS plasma levels are increased due to elevated gut permeability [212], leading to the low-grade inflammation characterizing this metabolic disorder [291]. MDP is a minor component of the peptidoglycan of the cell wall of Gram-negative and more abundant in Gram-positive bacteria [219]. Under normal conditions, MDP is released constantly from degraded gut bacteria [45]. LPS and MDP are CD14/TLR4 and NOD2 agonists, respectively, and both enhance GLP-1-induced nitric oxide production in enteric neurons [47], which may contribute to satiety signalling to the hypothalamus by promoting a shift towards anorexigenic neuropeptides expression [213]. Moreover, MDP activates NOD2 expressed in L-cells to induce GLP-1 release in healthy, but not diet-induced obese mice [220]. Although these effects have been related to glucose tolerance [220] MDP-induced GLP-1 release should also be studied for its effects on food intake.

Interestingly, several studies with rodents report clear effects of LPS and MDP on satiety. Acute or chronic intraperitoneal injections of LPS or MDP reduce food intake [43,45,214,215]. Some studies have found LPS to induce stronger anorexigenic effects in the brain compared to MDP [44,216,292]. The satiety-producing potential between both bacterial components is probably explained by their different abilities to enhance cytokine expression in the hypothalamus [216], since no changes have been detected in the expression of the neuropeptides POMC, NPY or leptin after LPS or MDP administration to the brain [216].

The anorexigenic effects of LPS and MDP seem to be mediated by CD14 for both and by TLR4 for LPS and TLR2 for MDP [217]. The LPS from *E. coli* isolated from healthy humans has been demonstrated to increase murine nodose ganglia neurons excitability in vitro via a TLR4-dependent mechanism [218], suggesting that the intestinal-derived LPS effects on hypothalamic regulation of food intake is mediated by vagal afferent neurons. In contrast to these findings, subdiaphragmatic vagal deafferentation in rats did not suppress food intake reduction after intraperitoneal injection of LPS or MDP, suggesting that peripheral administration of both bacterial components enhances appetite reduction via a humoral but not vagal pathway [214]. Moreover, MDP gavage in rats has no effects in MDP circulating levels nor in feeding patterns, while intraperitoneal MDP reduces food intake, indicating the importance of the humoral pathway for MDP satiety signalling [45]. However, intestinal MDP translocation should be specifically investigated to better define the underlying mechanisms of intestinal-derived MDP on appetite regulation.

Lugarini et al. found that LPS and MDP suppress food intake in obese rats to a similar extent as in non-obese rats [215], suggesting that both bacterial components exert anorexigenic effects regardless obesity. However, the doses of injected LPS and MDP in this study may not reflect the amount of LPS and MDP crossing the intestinal epithelial barrier, whose integrity depends on the metabolic state [212]. Further investigation is required to fully unravel the mechanisms and pathways through which intestinal structural bacterial components reach the hypothalamic centre of appetite regulation under healthy and pathological conditions in vivo.

#### **5. Tackling Obesity with Gut Microbes Mediating in Gut–Brain Communication**

Hyperphagia and obesity are caused by defective sensitivity in the hypothalamus or in food reward brain areas to peripheral signals reflecting the nutritional status of the body. Current antiobesity drugs mainly target the central nervous system to suppress appetite. Monotherapies including the agonist of 5-HT2C receptor, lorcaserin, or the protease-resistant long-acting GLP-1, liraglutide, as well as combinational therapies of stimulators of noradrenaline or serotonine-release combined with sympathomimetic anticonvulsant or opiod receptor antagonist are currently approved for treating obesity and T2D, combined with exercise and a balanced diet [14,15,293]. As emerging pharmacotherapies, gut peptides acting as potential agonists of lipid sensing and bile acids receptors and, thereby, as secretagogues are being preclinically and clinically tested as potential drugs to improve metabolic disturbance, especially T2D [294]. Notably, bariatric surgery is the most effective therapy for treating obesity, producing important and sustained weight-loss. This is conducted only in a limited number of subjects with a body mass index (BMI) > 40 and also implies surgical risks [295,296]. It is hypothesized that, rather than mechanical alterations, bariatric surgery alters gut signals sensed by the brain, including amplification of postprandial secretion of gut hormones, resulting in a beneficial impact on energy homeostasis [13]. Research is underway to advance the clinical management of obesity with less invasive pharmacotherapies, which mimic the molecular adaptations of bariatric surgery, including those based on gut–peptide combinatorial strategies [13]. However, the applicability of new drugs is frequently limited by unwanted side effects and safer alternatives are being considered, such as devices to modulate the vagal nerve [14].

In the light of this review, the gut microbiota represents a potential target for favourably regulating gut–brain communication and thereby controlling food intake, energy homeostasis and obesity. Indeed, the supplementation of key intestinal bacteria could increase

the microbiota-derived bioactive molecules acting as enhancers of gut hormone secretion and as vagal afferent stimulators, optimizing the response to nutritional clues via gut–brain signalling in obesity. Notably, promising studies have revealed that the gut microbiota might act in harmony with the postprandial processes in the gut that control food intake, as occurs for bacterial-produced ClpB.

Currently, microbiome-based approaches, including faecal microbiota transplantation (FMT) and the administration of intestinal bacteria (probiotics or bio-therapeutic products), are being explored for their impact on gut microbiota structure and function for therapeutic or preventive purposes.

The clinical applicability of FMT to treat obesity is being explored in experimental clinical trials. To date, a couple of studies have been completed and indicate that the microbial shift towards a lean-like profile induced by the oral administration of capsules containing the faecal microbiota from lean donors was not associated with metabolic benefits in the recipients [297,298]. By contrast, an improvement in insulin resistance was observed in male volunteers with metabolic syndrome after a duodenal infusion of gut microbiota from lean donors obtained from small intestine biopsy species [299]. Nevertheless, investigations in this field provide limited data regarding how the microbiotabased approaches specifically control the gut–brain axis in obesity [300].

New indigenous intestinal bacterial species are also being identified and tested for their potential use as next generation probiotics or live biotherapeutic products [301] to maintain metabolic health and combat obesity. This strategy may also contribute to progressing towards the development of safer and more effective microbiome-based strategies for the clinical management of obesity as compared to FMT [302]. Some examples of these new bacterial species and strains are *Eubacterium hallii* L2–7, *Akkermansia muciniphila* ATCC® BAA-835™ or *Bacteroides uniformis* CECT 7771. These were initially selected based on associations between an increased abundance of the bacterial species and a healthy metabolic phenotype in a substantial number of human studies. To date, the clinical efficacy in overweight volunteers has already been demonstrated for *A. muciniphila* (clinical trial no. NCT02637115). The pasteurized *A. muciniphila* significantly improved metabolic parameters (insulin sensitivity, insulinemia and cholesterol) and tended to reduce body weight and fat mass [303]. Nonetheless, the investigations on the mechanisms by which these bacteria induce their benefits have mainly focused on the regulation of immune pathways involved in obesity-associated inflammation, generally resulting in metabolic improvements in preclinical trials [304]. Although evidence of the possible action of these bacteria through the modulation of the gut–brain axis is scarce [305–307], preclinical studies demonstrated that *A. muciniphila* administration to obese mice increases the intestinal production of the endocannabinoid, 2-arachidonoylglycerol (2-AG); and the endocannabinoid analogue, oleoylglycerol (2-OG) [305]. These ligands can modulate the secretory function of L cells and, theoretically, the activity of vagal afferents [308,309].

Other emerging strategies to identify new probiotics or live bio-therapeutic products tend to mimic drug discovery approaches. In this case, the selection of effector intestinal bacteria is based on in silico predictions (computational molecular modelling) of their capacity to produce bioactive molecules [310]. Based on this strategy, the strain *Hafnia alvei* 4597 was selected to provide the protein ClpB, a bacterial mimetic of α-MSH, which induced satiety. This discovery was based on the initial detection of α-MSH reactive autoantibodies in plasma of humans and rodents [311] and in the in silico screening of the presence of ClpB in members of the family Enterobacteriaceae, including the genus *Hafnia* [312]. Studies in obese mice demonstrated that the administration of *Hafnia alvei* 4597 suppressed food intake and improved the metabolic disturbances associated to obesity [209]. Clinical studies in humans to show the efficacy are underway (clinical trial no. NCT03657186).

In spite of this progress, further advances are still needed to increase the effectiveness and clinical applicability of microbiome-based strategies. In fact, interindividual variability of the human host and its gut microbiota could change the response to these therapies [313]. Therefore, efforts are also underway to develop precision probiotics through

a deep understanding of their mode of action and the factors influencing the individual host response [310]. This requires the integration of the person-specific genotypic and phenotypic variables, including microbiome data (strain-level composition, transcriptomics, metabolomics, etc.), that may help to predict the health outcomes of a specific intervention. This strategy may also be applied to the identification of personalized microbiome-based strategies that improve nutrient sensing routes and, thus, control energy homeostasis. To this end, further efforts are needed for the identification of cross-feeding pathways between different intestinal bacteria of a specific host and the resulting dietary and bacterial-derived effector molecules, considering the person's meal timings and dietary composition.

**Author Contributions:** Writing—Original Draft: M.R.-P., C.B.-V., I.L.-A., Y.S.; Writing—Review and Editing: Y.S., M.O., R.L.-G.; Figures: M.O., R.L.-G.; Funding Acquisition: Y.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work is supported by grant AGL2017-88801-P from the Spanish Ministry of Science and Innovation (MICINN, Spain). The Marie Sklodowska-Curie grant agreement No. 797297 (M.R-P), the FPI grant of I.L.-A., the FPU grant of R.L. and the Juan de la Cierva contract of M.O. and the contract of C. B-V for promotion of youth employment in R+D+I from MICINN (Spain) are fully acknowledged.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Review* **Hypothalamic Actions of SIRT1 and SIRT6 on Energy Balance**

**Mar Quiñones 1,2,\* , Eva Martínez-Grobas <sup>1</sup> , Johan Fernø <sup>3</sup> , Raquel Pérez-Lois <sup>4</sup> , Luisa María Seoane 2,4 and Omar Al Massadi 2,4,\***


**Abstract:** Sirtuins are NAD+ dependent deacetylases that regulate a large number of physiological processes. These enzymes are highly conserved and act as energy sensors to coordinate different metabolic responses in a controlled manner. At present, seven mammalian sirtuins (SIRT 1-7) have been identified, with SIRT1 and SIRT6 shown to exert their metabolic actions in the hypothalamus, both with crucial roles in eliciting responses to dampen metabolic complications associated with obesity. Therefore, our aim is to compile the current understanding on the role of SIRT1 and SIRT6 in the hypothalamus, especially highlighting their actions on the control of energy balance.

**Keywords:** SIRT1; SIRT6; energy balance; food intake; body weight; adiposity; type 2 diabetes; obesity

**Citation:** Quiñones, M.; Martínez-Grobas, E.; Ferno, J.; Pérez-Lois, R.; Seoane, L.M.; Al Massadi, O. Hypothalamic Actions of SIRT1 and SIRT6 on Energy Balance. *Int. J. Mol. Sci.* **2021**, *22*, 1430. https://doi.org/10.3390/ijms22031430

Academic Editor: Cristina Miralpeix Received: 23 December 2020 Accepted: 28 January 2021 Published: 31 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

## **1. Introduction**

#### *1.1. Obesity*

The incidence of obesity has reached pandemic proportions and become a major cause of various social, economic and health problems in our society [1]. In fact, worldwide obesity has nearly tripled since 1975 and the World Health Organization has declared that overweight is among the top ten conditions of major health risk in the world and among the top five in developed countries [2]. Obesity is defined as a state on excess of fat produced as a consequence of a high energy intake and a low energy expenditure that leads to serious health problems [3,4]. While food intake is tightly regulated by homeostatic and hedonic drives [5], changes in energy expenditure can be explained by differences in basal metabolism (energy expended to keep the body functioning at rest), physical activity and adaptive thermogenesis [6]. Therefore, obesity is a complex and multifactorial disease involving genetic, biological but also behavioral factors.

The first line of treatment of obese people is to follow a healthy diet and do regular exercise. However, these guidelines are difficult to adhere to for the obese population and most people with obesity that lose weight will regain it over time, ultimately with a higher body weight than before starting the weight loss [7,8]. Bariatric surgery is currently the most efficient method for long-term weight loss. However, this is a costly and invasive intervention not without risk and side-effects, and is only recommended to a small proportion of morbidly obese people [9]. The currently available pharmacological treatments to combat obesity have limited effectiveness and more efficient drugs are urgently needed. Therefore, it is necessary to ascertain in depth the molecular mechanisms underlying this disease and to identify novel therapeutic targets for treatment.

#### *1.2. The Hypothalamus as the Master Regulator of Energy Balance*

The central nervous system (CNS) is a key organ in energy balance and body weight regulation. The CNS elicits these functions through three mechanism: (a) controlling foodrelated behaviors, including food intake (b) controlling the autonomic nervous system, which regulate the energy expenditure and other metabolic processes; and (c) controlling the neuroendocrine system by modulating hormone secretion. The modulation and coordination of this complex systems occurs in different brain areas [10,11]. However, it is well known that the hypothalamic neuronal networks integrate peripheral information (i.e., nutrients and hormones) to modulate energy balance [12,13]. The hypothalamus, a central structure composed of anatomically distinct nuclei interconnected via axonal projections, is at large the most studied area in term of its regulation of food intake and body weight. In this neuronal network, we can distinguish the hypothalamic arcuate nucleus (ARC) as the best positioned nucleus to receive signals from the periphery and to mediate a subsequent homeostatic response to peripheral tissues. This nucleus is considered the "master hypothalamic center" for feeding control [13,14].

The ARC is composed by two antagonist neuronal populations, neuropeptide Y/Agouti related peptide (NPY/AgRP) neurons that induces a positive energy balance and proopiomelanocortin (POMC) neurons and cocaine and amphetamine related transcript (CART) that when activated induces negative energy balance. Both neuron types regulate food intake, energy expenditure and nutrient partitioning. When energy intake exceeds expenditure, the expression of orexigenic neuropeptides, such as AgRP and NPY decreases, whereas the expression of anorexigenic neuropeptides, such as CART and POMC increases [3,10]. Opposite changes occur when energy expenditure exceeds intake.

Hypothalamic neurons in the ARC respond to peripheral nutrients, such as glucose and fatty acids as well as hormones, such as leptin and ghrelin, the two most important hormones in the regulation of food intake and body weight [11,12]. Leptin, secreted by, and in proportion to white adipose tissue (WAT), informs the hypothalamus of energy buildup and activates POMC/CART neurons and inhibits NPY/AgRP neurons, resulting in an inhibition of feeding and an increase in energy expenditure. In contrast, ghrelin is a stomach-derived peptide that increases food intake through increased NPY and AgRP expression. Therefore, both leptin and ghrelin use these pathways in opposite ways to exert their roles on energy balance.

Adjacent to the ARC we find the ventromedial nucleus (VMH), a hypothalamic area that receives projections mainly from NPY/AgRP and POMC neurons. These VMH neurons express steroidogenic factor 1 (SF1) and project their axons at the same time to the ARC and to secondary hypothalamic nuclei, as well as regions of the brain stem [15]. These neurons are activated in respond to changes in circulating leptin and other metabolic signals, and due to their projections to the ARC, an elaborate neuronal circuitry underlying metabolic homeostasis is formed [15,16].

Another relevant metabolic nucleus is the lateral hypothalamic area (LHA), a region of the brain that has for long been known for its role in the regulation of eating and considered the most interconnected area of the hypothalamus. This hypothalamic area is mainly composed of two types orexigenic neuronal populations, namely orexin-producing neurons and melanin-concentrating hormone (MCH)-producing neurons [17–19].

Taking into account that several mechanisms are involved in the development and maintenance of obesity within the hypothalamus, in the next section we introduce the role of a highly conserved family of class III deacetylases denominated Sirtuins, (SIRTs) as one of these relevant metabolic regulators.

#### **2. Sirtuins: Metabolic Sensors**

Mammalian SIRTs are homologous to the *Sir2* gene of *Saccharomyces cerevisiae*. Seven *Sir2* orthologues, members of the SIRT family (SIRT1-7), have been described in mammals. All members share the conserved SIRT domain, but they differ in subcellular location and function [20–22]. SIRTs actively participate in the regulation of energy metabolism related

to diet/caloric restriction, but they also have an important role in other metabolic processes, such as aging, cell survival, and DNA repair [23,24]. control energy balance. Therefore, in this review, we will describe in detail some of the most important hypothalamic mechanisms by which these two SIRTs regulate energy me-

metabolic processes, such as aging, cell survival, and DNA repair [23,24].

Mammalian SIRTs are homologous to the Sir2 gene of Saccharomyces cerevisiae.

The aim of this review is to describe the hypothalamic actions of SIRTs in a metabolic

context. Until now, only SIRT1 and SIRT6 have been shown to act in the hypothalamus to

Seven Sir2 orthologues, members of the SIRT family (SIRT1-7), have been described in mammals. All members share the conserved SIRT domain but they differ in subcellular location and function [20–22]. SIRTs actively participate in the regulation of energy metabolism related to diet/caloric restriction, but they also have an important role in other

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 3 of 11

The aim of this review is to describe the hypothalamic actions of SIRTs in a metabolic context. Until now, only SIRT1 and SIRT6 have been shown to act in the hypothalamus to control energy balance. Therefore, in this review, we will describe in detail some of the most important hypothalamic mechanisms by which these two SIRTs regulate energy metabolism. tabolism *2.1. Hypothalamic SIRT1 in the Control of Energy Balance* SIRT1 is expressed in a wide range of tissues and organs, such as the liver, pancreas,

#### *2.1. Hypothalamic SIRT1 in the Control of Energy Balance* heart, muscle, and adipose tissue (Reviewed in [20–22]). SIRT1 is also expressed in im-

SIRT1 is expressed in a wide range of tissues and organs, such as the liver, pancreas, heart, muscle, and adipose tissue (Reviewed in [20–22]). SIRT1 is also expressed in important metabolic centers of the brain, including ARC, VMH, dorsomedial nucleus and paraventricular nucleus of the hypothalamus (PVN), the area postrema and the nucleus of the solitary tract in the hindbrain (For review see [20–22]). SIRT1 is involved in the control of metabolic processes that regulate body weight, such as control of food intake [25–28], adiposity [26,29], energy expenditure [29], thermogenesis of brown adipose tissue (BAT) and "browning" of WAT [29] (Figure 1). portant metabolic centers of the brain, including ARC, VMH, dorsomedial nucleus and paraventricular nucleus of the hypothalamus (PVN), the area postrema and the nucleus of the solitary tract in the hindbrain (For review see [20–22]). SIRT1 is involved in the control of metabolic processes that regulate body weight, such as control of food intake [25–28], adiposity [26,29], energy expenditure [29], thermogenesis of brown adipose tissue (BAT) and "browning" of WAT [29] (Figure 1).

**Figure 1.** Description of the metabolic actions elicited by SIRT1 in mice and humans**.** Pharmacological or genetic modulation of SIRT1 regulates body weight, adiposity in WAT and liver, food intake, energy expenditure, resting metabolic rate, respiratory quotient, BAT thermogenesis, browning of WAT, glucose homeostasis and insulin sensitivity. green arrow: stimulation, red arrow: inhibition, green star:metabolic state. Abbrevations used-BAT; brown adipose tissue; SIRT1: sirtuin 1; WAT; white adipose tissue. The figures were generated by using materials from Servier **Figure 1.** Description of the metabolic actions elicited by SIRT1 in mice and humans. Pharmacological or genetic modulation of SIRT1 regulates body weight, adiposity in WAT and liver, food intake, energy expenditure, resting metabolic rate, respiratory quotient, BAT thermogenesis, browning of WAT, glucose homeostasis and insulin sensitivity. green arrow: stimulation, red arrow: inhibition, green star:metabolic state. Abbreviations used-BAT; brown adipose tissue; SIRT1: sirtuin 1; WAT; white adipose tissue. The figures were generated by using materials from Servier Medical Art (Servier) under consideration of a Creative Commons Attribution 3.0 Unported License.

Medical Art (Servier) under consideration of a Creative Commons Attribution 3.0 Unported License. With regard to genetic animal models for SIRT1, the global knock out (KO) mice for SIRT1 are not viable or have several metabolic complications [30], whereas moderate overexpression of SIRT1 in mice improves several metabolic parameters associated with obesity [31] (Figure 1). Consistently, the pharmacological activation of SIRT1 produces a variety of beneficial metabolic actions in vivo (Figure 1). In this sense, natural activators of SIRT1, such as the polyphenol resveratrol (RSV) [32,33] or synthetic molecules [34–36], With regard to genetic animal models for SIRT1, the global knock out (KO) mice for SIRT1 are not viable or have several metabolic complications [30], whereas moderate overexpression of SIRT1 in mice improves several metabolic parameters associated with obesity [31] (Figure 1). Consistently, the pharmacological activation of SIRT1 produces a variety of beneficial metabolic actions in vivo (Figure 1). In this sense, natural activators of SIRT1, such as the polyphenol resveratrol (RSV) [32,33] or synthetic molecules [34–36], have been used for this purpose. These pharmacological activators improve insulin sensitivity, increase insulin secretion, enhance mitochondrial function, decrease adiposity and lower glucose levels [37–41]. Intriguingly, a negative energy balance was also observed in mice following pharmacological inhibition of SIRT1 [26,27]. Therefore, a controversy exists

have been used for this purpose. These pharmacological activators improve insulin sensi-

lower glucose levels [37–41]. Intriguingly, a negative energy balance was also observed in

on the role of SIRT1 in regulation of energy metabolism and more studies are needed to ascertain the mechanism of action of these SIRT1 modulators.

#### *2.2. SIRT1 and Food Intake*

Several studies have shown that pharmacological inhibition of SIRT1 in the hypothalamus reduces the caloric intake and body weight in rodents [26,27], which occur through the modulation of the central melanocortin system [25,27,28]. In fact, central administration of a melanocortin receptor antagonist (SHU9119) reversed the anorectic effect of hypothalamic SIRT1 inhibition [27]. Indeed, this anorexigenic action is mediated by the repression of AgRP and stimulation of POMC levels. In addition, SIRT1 requires the Forkhead Box O1 transcription factor (FOXO1) to exert these actions, since FOXO1 inhibition blocks the actions of SIRT1 on AgRP and POMC neurons [27]. However, SIRT1 has other downstream effectors, e.g., the actions of SIRT1 on food intake are mediated by the mitochondrial redox machinery, because the decrease in feeding induced by the action of EX-527, a pharmacological SIRT1 inhibitor, is impaired in uncoupling protein-2 KO animals [28].

In agreement with these data, hypothalamic SIRT1 can act as a mediator of several nutritional hormones (reviewed in [42]). In fact, the pharmacological blockade of SIRT1 in the CNS or the genetic inhibition of p53, a downstream effector of SIRT1, prevents the orexigenic action of ghrelin, by impairing the regulation of this hormone on its target neuropeptides NPY/AgRP and its target transcription factors namely FOXO1, cAMP response element-binding protein and brain-specific homeobox transcription factor [25]. Subsequent studies with conditional KO mice models for SIRT1 and p53 in specific neuronal populations show that the food intake induced by ghrelin rely on the action of these energy sensors specifically in AgRP neurons [28,43]. These data highlight the relevance of SIRT1/p53 pathway in the orexigenic action of ghrelin.

Consistently with these findings, it has been shown that SIRT1 mediates the metabolic actions regulated by other orexigenic hormone such as MCH. In this study, it was shown that the action of MCH on food intake, glucose homeostasis and adiposity rely on SIRT1 in POMC neurons, due to the specific deletion of SIRT1 on a subset of these neurons comprise the metabolic actions of MCH [26]. These data reveal for the first time the neuronal basis of the metabolic effects of MCH on body weight and food intake.

Another study demonstrated the importance of SIRT1 in food selection, and that SIRT1 mediates the regulation of macronutrient choice [44]. In this study, targeting central SIRT1 by gain-of function and loss-of-function mice models, it was shown that the preference for high sucrose diet was reduced when SIRT1 was stimulated at neuronal level, while it was increased when SIRT1 was inhibited at this level. These effects of diet preference over sucrose were shown to be mediated by SIRT1 specifically in oxytocin neurons, since overexpression or deletion of SIRT1 in oxytocin PVN neurons displayed the same results on sucrose preference as the neuronal manipulation of SIRT1. However, when standard diet (SD) vs. high fat diet (HFD) was offered to the animals in the latter mentioned mice model, no differences were found, which show that SIRT1 in oxytocin neurons regulates the preference for sucrose, but not for fat. The fibroblast growth hormone-21 (FGF-21) is a peripheral signal targeting oxytocin neurons to regulate macronutrient preference. In the same study SIRT1 was identified as a regulator of sucrose preference by promoting FGF21 sensitivity in oxytocin neurons.

#### *2.3. SIRT1 in Hypothalamic Neurons*

Studies with conditional mice models have shown that the effects of SIRT1 in the hypothalamus may depend on the neuronal type in which it acts (Figure 2).

**Figure 2.** SIRT1 and SIRT6 in hypothalamic neurons: The effects of SIRT1 and SIRT6 in the hypothalamus may depend on the neuronal type. Abbrevations used-AgRP: Agouti-related protein neruons; BAT; brown adipose tissue; MCH: melanin concnetrating hormone; POMC: Proopiomelanocortin neurons; SF-1: steroidogenic factor 1; WAT; white adipose tissue. The figures were generated by using materials from Servier Medical Art (Servier) under consideration of a **Figure 2.** SIRT1 and SIRT6 in hypothalamic neurons: The effects of SIRT1 and SIRT6 in the hypothalamus may depend on the neuronal type. Abbreviations used-AgRP: Agouti-related protein neurons; BAT; brown adipose tissue; MCH: melanin concentrating hormone; POMC: Proopiomelanocortin neurons; SF-1: steroidogenic factor 1; WAT; white adipose tissue. The figures were generated by using materials from Servier Medical Art (Servier) under consideration of a Creative Commons Attribution 3.0 Unported License.

Creative Commons Attribution 3.0 Unported License. It has been reported that inhibition of SIRT1 specifically in POMC neurons does not significantly alter body weight or adiposity in SD-fed mice [29]. However, these mice are more vulnerable to diet-induced obesity (DIO) due to a reduction in energy expenditure and a reorganization of WAT through the sympathetic nervous system. Consistent with It has been reported that inhibition of SIRT1 specifically in POMC neurons does not significantly alter body weight or adiposity in SD-fed mice [29]. However, these mice are more vulnerable to diet-induced obesity (DIO) due to a reduction in energy expenditure and a reorganization of WAT through the sympathetic nervous system. Consistent with these data, overexpression of SIRT1 in POMC neurons has the opposite effect, and these mice showed a lean phenotype due to an increase in energy expenditure and activation of sympathetic nervous system in adipose tissue [45].

these data, overexpression of SIRT1 in POMC neurons has the opposite effect, and these mice showed a lean phenotype due to an increase in energy expenditure and activation of sympathetic nervous system in adipose tissue [45]. Additionally, SIRT1 is essential for the metabolic functions of AgRP neurons. The inhibition of SIRT1 specifically in AgRP neurons decreases the consumption of food, Additionally, SIRT1 is essential for the metabolic functions of AgRP neurons. The inhibition of SIRT1 specifically in AgRP neurons decreases the consumption of food, which leads to a decrease in adiposity and consequently a reduced body weight [28]. There are, however, also opposing experimental data, demonstrating that overexpression of SIRT1 in AgRP neurons improves weight gain and reduces the food intake associated with age in mice [45].

which leads to a decrease in adiposity and consequently a reduced body weight [28]. There are, however, also opposing experimental data, demonstrating that overexpression of SIRT1 in AgRP neurons improves weight gain and reduces the food intake associated Overexpression of SIRT1 in SF1 neurons shows protective effects against DIO, with reduction in body weight, fat mass, and leptin levels, and increased energy expenditure [46]. Accordingly, the specific inhibition of SIRT1 in SF1 neurons makes HFD fed mice more susceptible to obesity and diabetes mellitus [46].

with age in mice [45]. These data indicate that SIRT1 plays an essential role in hypothalamic neurons that are associated with the metabolic adaptations that determine body weight gain in obesity.

#### Overexpression of SIRT1 in SF1 neurons shows protective effects against DIO, with *2.4. SIRT1 a Link between Metabolism and Reproduction*

reduction in body weight, fat mass, and leptin levels, and increased energy expenditure [46]. Accordingly, the specific inhibition of SIRT1 in SF1 neurons makes HFD fed mice more susceptible to obesity and diabetes mellitus [46]. These data indicate that SIRT1 plays an essential role in hypothalamic neurons that are associated with the metabolic adaptations that determine body weight gain in obesity. *2.4. SIRT1 a Link between Metabolism and Reproduction*  SIRT1 has been shown to be the link between metabolism and reproduction through hypothalamic neurons [47]. First, it has been observed how female animals that overexpress SIRT1 have a normal body weight and show a delay in the development of puberty. Using virogenetic techniques, it was demonstrated that the overexpression of SIRT1 specifically in the ARC promoted this delay. Moreover, in this work it was also shown that SIRT1 is expressed in kisspeptin, neurokinin B and dynorphin (KNDy) neurons, which express the *Kiss1* pubertal activator gene. It is important to highlight that the content of SIRT1 in the KNDy neurons of the ARC and the recruitment of SIRT1 towards the *Kiss1* promoter

SIRT1 has been shown to be the link between metabolism and reproduction through

press SIRT1 have a normal body weight and show a delay in the development of puberty. Using virogenetic techniques, it was demonstrated that the overexpression of SIRT1 specifically in the ARC promoted this delay. Moreover, in this work it was also shown that SIRT1 is expressed in kisspeptin, neurokinin B and dynorphin (KNDy) neurons, which express the Kiss1 pubertal activator gene. It is important to highlight that the content of SIRT1 in the KNDy neurons of the ARC and the recruitment of SIRT1 towards the Kiss1 promoter in vivo change depending on the nutritional status of the animal. In states of overnutrition, the neuronal content of SIRT1 and its association with Kiss1 decrease, which accelerates and promotes puberty[47]. In contrast, malnutrition raises SIRT1 levels,

in vivo change depending on the nutritional status of the animal. In states of overnutrition, the neuronal content of SIRT1 and its association with *Kiss1* decrease, which accelerates and promotes puberty [47]. In contrast, malnutrition raises SIRT1 levels, prolongs the repression of *Kiss1* by SIRT1, and delays puberty. These results identify SIRT1-mediated *Kiss1* inhibition as a key epigenetic mechanism by which nutritional cues and obesity influence puberty in mammals.

In summary, hypothalamic SIRT1 signaling appears to be a key mediator of energy metabolism and the physiological response to obesity.

#### **3. Hypothalamic SIRT6 in the Control of Energy Balance**

SIRT6 is a nuclear protein associated with physiological and pathological processes, regulating obesity, insulin resistance, inflammation and energy metabolism [48]. It is expressed in various tissues, with the highest level of expression in adipose tissue, skeletal muscle and heart [49–51]. SIRT6 is also highly expressed in the CNS, and its expression is regulated by the availability of nutrients, showing low levels in the hypothalamus and specifically in POMC neurons in obesity [52]. Therefore, SIRT6 is postulated as a relevant energy sensor at the central level and a promising pharmacological target in the regulation of energy metabolism.

A seminal work in the study of SIRT6 described that global SIRT6 KO mice suffer a severe multisystemic phenotype, with severe hypoglycemia and a short life expectancy [49]. A remarkable characteristic of these mice is that they show a delay in postnatal growth associated with low levels of insulin-like growth factor 1. In agreement with these results, mice in which SIRT6 was inactivated at the neural level (NSIRT6-KO) display a significantly reduced growth during development, smaller adult phenotype and lower body weight compared to controls [50]. These results demonstrate that neural inactivation of SIRT6 is sufficient to cause growth retardation. However, elderly male NSIRT6 KO mice, but not female mice, were significantly heavier than controls [50].

Mice deficient for SIRT6, specifically in POMC neurons (SIRT6 POMC KO), showed no changes in body size or differences in glucose levels [52]. Therefore, hypoglycemia in SIRT6 KO mice and growth retardation in NSIRT6 KO mice were not due to SIRT6 deficiency in POMC neurons. On the contrary, SIRT6 POMC KO mice showed a modest, but significant, increase in body weight under SD that was greater when the mice were fed HFD [52]. This increase in body weight is associated with a higher adiposity in WAT and liver, caloric intake, glucose intolerance and less energy expenditure. Consistently, it was also found that the activity of the leptin signaling pathway was significantly reduced in SIRT6 POMC KO mice, showing high levels of leptin in the serum of these mice both under SD and HFD [52].

Interestingly, and in line with these previous results the global overexpression of SIRT6 decrease body weight, adiposity and glucose tolerance [51]. Moreover, and consistent with these data, the study of Tang et al. has demonstrated that overexpression of SIRT6 in the ARC of obese mice decreases body weight, adiposity and food intake [52]. These data suggest that the activation of SIRT6 in POMC neurons could be the responsible of this effect, however this possibility was not tested yet.

Altogether, these studies postulate to SIRT6 as an essential player in the regulation of body energy homeostasis.

#### **4. Therapeutic Options**

The positive effects of SIRTs in animal models on prevention and treatment of obesity and metabolic syndrome suggested their use as therapeutic agents. However, the knowledge about the involvement of SIRTs on clinical trials for the treatment of metabolic disease is limited.

In this sense, the administration of RSV, the most studied natural SIRT1 activator, reduces adipocyte size, blood glucose, preserves pancreatic β-cells and improves insulin action in rhesus monkeys fed a high-fat, high sugar diet for 2 years [53,54]. Importantly, treatment with RSV during 30 days in obese men confirmed these initial reports in nonhuman primates [55]. Other epidemiological studies in humans showed positive effects of RSV in patients with diabetes or obesity-mediated insulin resistance [40,56,57]. Accordingly, RSV improves plasma triglyceride concentration [37,41], lowers circulating cytokine levels [37–39], elicits better metabolic flexibility with lower homeostatic model assessment for insulin resistance index [37], suppresses postprandial glucagon responses, decreases resting metabolic rate and improves respiratory quotient in obese subjects [37,58]. Interestingly, all of these effects elicited by RSV on energy balance, glucose homeostasis and insulin sensitivity were not observed in non-obese healthy humans [59].

It is important to highlight that in spite of these results, the use of the natural activator RVS as a therapeutic target bear some controversies. For example, the use of RVS shows low tolerability and specificity, inconsistent dosages activity and interactions with other compounds (for a discussion of this issue see [60]). These shortcomings have stimulated to the development of a number of new small synthetic molecules activators. However, although these molecules are 1000 times more potent and their bioavailability are improved compared to the natural compounds, they do not show a superior efficacy in preclinical or clinical studies [34–36]. For example, some epidemiological observations with the synthetic SIRT1 activator, SRT2104, showed that the administration of this compound during 28 days produces a mild reduction in the glucose levels and slightly improvement in the body weight [61]. Other studies in elderly or type 2 diabetic patients shows that the administration of SRT2104 reduces the lipid profile and the body weight [62] without changes in glucose levels [36,61].

Despite these controversies, SIRT1 activators often show an acceptable efficacy in treating metabolic disturbances in clinical trials, where improvements are restricted to subjects with a pathological condition. Thus, more studies with larger cohorts are necessary to obtain a consensus for the actions, specificity, dosage, and effectiveness of these SIRT1 activators.

#### **5. Conclusions and Remarks**

In this review we cover the most relevant studies focusing on the metabolic actions of SIRT1 and SIRT6 at the hypothalamic level. In this sense, it is important to highlight that the effect of hypothalamic SIRT1 on energy balance depends on the neuronal type where it acts, which has been demonstrated for a variety of metabolic effects. E.g., the specific deletion of SIRT1 in POMC and SF1 neurons from HFD-fed mice shows that SIRT1 regulates metabolism through energy expenditure but does not affect food intake, whereas specific modulation of SIRT1 in AgRP neurons specifically affects eating behavior. Other examples of the relevance of this deacetylase in the regulation of metabolic processes came from studies that describe its involvement in the signaling pathways of some relevant metabolic hormones. More specifically, it has recently been shown that SIRT1, in addition to mediate the orexigenic action of ghrelin, is essential for the metabolic actions of MCH. Importantly in both cases, the ARC melanocortin system plays a fundamental role in the action of SIRT1 in these processes.

SIRT1 may also regulate diet preference via actions in the PVN Oxitocin neurons. This is relevant since the effectiveness of the diet and its impact on body weight depends both on the amount of caloric intake that is consumed and the balance of macronutrients of the ingested meal. It is also important to highlight the effects of SIRT1 on conditions related to obesity, such as puberty. SIRT1 has been shown to be a key factor in the development of puberty. Specifically, it has been identified that the ARC is the main integrator area of these effects, with epigenetic inhibition of *Kiss1* by SIRT1 as a key mechanism by which nutritional signals and obesity influence puberty of children.

In addition to the effects of SIRT1, this review also provides a comprehensive overview on the involvement of hypothalamic SIRT6 in the control of energy metabolism. It has been shown recently that the specific deletion of SIRT6 in POMC neurons increases food intake and decreases energy expenditure, promoting an obesogenic phenotype in mice. These results suggest that SIRT6 has a protective effect against DIO by its action in POMC neurons. In this sense, future studies should focus on the possible effects of SIRT6 modulation in other hypothalamic neurons such as AgRP or SF1 neurons. Therefore, all these results strongly support SIRT6 as a new hypothalamic molecular mediator in the regulation of energy homeostasis.

The data described in this review reveal SIRT1 and SIRT6 as multifaceted mediators of energy metabolism, affecting processes such as food intake, food preference, puberty, body weight, adiposity, glucose homeostasis and insulin resistance.

Apart from these considerations there are controversial results regarding the central actions of these SIRTs that we have discussed. While it seems clear that SIRT1 activators improve several metabolic parameters, in some circumstances this is also true for SIRT1 inhibitors. For example, neural deletion of SIRT1 improves insulin sensitivity and glucose intolerance in mice, while the administration of the pharmacological inhibitor EX-527 improves body weight, food intake, hepatic steatosis and fibrosis in diabetic rats [25–27,63]. In this sense, the healthy phenotype observed in these rodent models following SIRT1 inhibition may at first seem contradictory, but there are other previous examples in the literature involving other mammalian SIRTs that show similar results. Indeed, SIRT6 overexpression protected against DIO [51] in the same way that SIRT6 deletion did [49]. There is currently no clear explanation for the controversial results regarding the central actions of these SIRTs, but it is believed that the processes that involves metabolic regulation are highly complex and that this may sometimes lead to counterintuitive findings. Therefore, these data imply that the implications for these variety of effects are that their levels must be under a fine-tuning regulation and that their alterations may lead to unpredictable side effects without an evident dose-response relationship.

In spite of these somewhat contradictory results, we believe that SIRT1 and SIRT6 emerge as important targets for the development of novel therapies in the control of obesity and its comorbidities. Importantly, it should be noted that the fact that these proteins regulate energy balance in specific high-calorie regimens is of paramount importance for the development of new strategies to combat metabolic diseases. Finally, we should not forget the potential involvement of the other SIRTs different from SIRT1 and SIRT6 in the control of energy balance at the hypothalamic level, which should be the focus of future studies.

**Author Contributions:** M.Q. and O.A.M. wrote the manuscript. All authors made substantial contributions to discussion of the content and reviewed/edited the manuscript before submission. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work is supported by grants of Xunta de Galicia (MQ:2018-PG013). Centro de Investigación Biomédica en Red (CIBER) de Fisiopatología de la Obesidad y Nutrición (CIBERobn). CIBERobn is an initiative of the Instituto de Salud Carlos III (ISCIII) of Spain which is supported by FEDER funds. Western Norway Regional Health Authority (Helse Vest RHF).M.Q. is a recipient of a Postdoctoral contract from Galician Government (Xunta de Galicia ED481B2018/004). O.A.M. was funded by a research contract Miguel Servet (CP20/00146) from the ISCIII. The figures were generated by using materials from Servier Medical Art (Servier) under consideration of a Creative Commons Attribution 3.0 Unported License.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Review* **New Insights of SF1 Neurons in Hypothalamic Regulation of Obesity and Diabetes**

**Anna Fosch <sup>1</sup> , Sebastián Zagmutt <sup>1</sup> , Núria Casals 1,2 and Rosalía Rodríguez-Rodríguez 1,\***


**Abstract:** Despite the substantial role played by the hypothalamus in the regulation of energy balance and glucose homeostasis, the exact mechanisms and neuronal circuits underlying this regulation remain poorly understood. In the last 15 years, investigations using transgenic models, optogenetic, and chemogenetic approaches have revealed that SF1 neurons in the ventromedial hypothalamus are a specific lead in the brain's ability to sense glucose levels and conduct insulin and leptin signaling in energy expenditure and glucose homeostasis, with minor feeding control. Deletion of hormonal receptors, nutritional sensors, or synaptic receptors in SF1 neurons triggers metabolic alterations mostly appreciated under high-fat feeding, indicating that SF1 neurons are particularly important for metabolic adaptation in the early stages of obesity. Although these studies have provided exciting insight into the implications of hypothalamic SF1 neurons on whole-body energy homeostasis, new questions have arisen from these results. Particularly, the existence of neuronal sub-populations of SF1 neurons and the intricate neurocircuitry linking these neurons with other nuclei and with the periphery. In this review, we address the most relevant studies carried out in SF1 neurons to date, to provide a global view of the central role played by these neurons in the pathogenesis of obesity and diabetes.

**Keywords:** SF1 neurons; ventromedial hypothalamus nucleus; obesity; diabetes; energy homeostasis; glucose homeostasis

## **1. Introduction**

Obesity is a multifactorial chronic disease associated with a higher risk of developing cardiovascular diseases, diabetes, cancer, and, more recently, COVID-19 infection. According to the World Health Organization (WHO), in 2016, worldwide obesity had nearly tripled since 1975, with 39% of adults and 18% of children and adolescents overweight or obese [1]. The main metabolic comorbidity of obesity is type-2 diabetes that occurs when body tissues become resistant to insulin and is estimated to be the seventh leading cause of death [2]. Therefore, understanding the molecular and physiological mechanisms underlying the control of feeding behavior, energy balance and glucose homeostasis is crucial for the prevention and treatment of obesity and diabetes.

The regulation of peripheral metabolism and glucose homeostasis not only resides in the tissue's autonomous responses to nutrients and hormonal signals but also in some brain regions, particularly the hypothalamus. The hypothalamus integrates multiple metabolic inputs from peripheral organs with afferent stimuli coming from other brain regions and coordinates a diversity of efferent responses to control food intake, fat metabolism, hormone secretion, body temperature, locomotion, and behavior in order to maintain energy balance and blood glucose levels. Within the hypothalamus, the ventromedial nucleus (VMH) located above the arcuate nucleus (ARC) and the median eminence, was identified in the mid-1900s as the satiety center because its injury produced hyperphagia, insulin resistance,

**Citation:** Fosch, A.; Zagmutt, S.; Casals, N.; Rodríguez-Rodríguez, R. New Insights of SF1 Neurons in Hypothalamic Regulation of Obesity and Diabetes. *Int. J. Mol. Sci.* **2021**, *22*, 6186. https://doi.org/10.3390/ ijms22126186

Academic Editor: Clara Balsano

Received: 29 April 2021 Accepted: 5 June 2021 Published: 8 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

and body weight gain [3,4]. At that point, VMH was demonstrated to play a key role in the control of energy expenditure and glucose homeostasis [3,5]. Since then, intensive research has been done on VMH and it is currently known that this hypothalamic area encompasses a heterogeneous set of neurons, which are differentiated by the genes they are expressing (Figure 1). Many of the genes highly expressed in the VMH have been identified and their functions have been explored (Figure 1) [6–8]. demonstrated to play a key role in the control of energy expenditure and glucose homeostasis [3,5]. Since then, intensive research has been done on VMH and it is currently known that this hypothalamic area encompasses a heterogeneous set of neurons, which are differentiated by the genes they are expressing (Figure 1). Many of the genes highly expressed in the VMH have been identified and their functions have been explored (Figure 1) [6–8].

ventromedial nucleus (VMH) located above the arcuate nucleus (ARC) and the median eminence, was identified in the mid-1900s as the satiety center because its injury produced hyperphagia, insulin resistance, and body weight gain [3,4]. At that point, VMH was

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 2 of 25

**Figure 1.** Schematic illustration of the pattern of genes highly expressed in the VMH. The majority of VMH cells, especially in the dorsomedial and central regions of VMH, express the nuclear receptor steroidogenic factor 1 (SF1). Leptin receptor (LEPR) positive cells mainly converge in the dorsomedial part, whereas insulin receptor (IR) maps the central region. Cells expressing brainderived neurotrophic factor (BDNF) are mainly distributed in central and lateral areas of the VMH, estrogen receptor (ERα)-expressing cells are limited to the lateral region, and glucokinase (GK) positive cells are present throughout the VMH. This schematic diagram is based on previous articles from Choi et al. [8] and Yi et al. [6]. **Figure 1.** Schematic illustration of the pattern of genes highly expressed in the VMH. The majority of VMH cells, especially in the dorsomedial and central regions of VMH, express the nuclear receptor steroidogenic factor 1 (SF1). Leptin receptor (LEPR) positive cells mainly converge in the dorsomedial part, whereas insulin receptor (IR) maps the central region. Cells expressing brain-derived neurotrophic factor (BDNF) are mainly distributed in central and lateral areas of the VMH, estrogen receptor (ERα)-expressing cells are limited to the lateral region, and glucokinase (GK)-positive cells are present throughout the VMH. This schematic diagram is based on previous articles from Choi et al. [8] and Yi et al. [6].

The majority of VMH cells, especially in the dorsomedial and central regions of VMH (VMHdm and VMHc, respectively) [9], express the nuclear receptor steroidogenic factor 1 (SF1) (Figure 1), also known as NR5A1, which is essential for VMH development and function [10–12] and is considered a key marker of that nucleus since SF1 expression is specific and selective to VMH in the brain. Outside the brain, SF1 can be found in the adrenal gland and gonads with differential roles during development [13]. For many years, SF1 was classified as an orphan nuclear receptor, but between 2005 and 2007 several pieces of evidence demonstrated that phospholipids can be ligands of SF1, such as phosphatidic acid being an activator and sphingosine an inhibitor of SF1 transcriptional activity [13]. Interestingly, SF1 can suffer different post-translational modifications, which regulate its stability and transcriptional activity [14], but also control the expression of numerous downstream target genes, including CB1, BDNF, and Crhr2 [13]. Considering this, in order to explore the importance of SF1 neurons, transgenic mice lacking this nuclear receptor were studied by different researchers. Mice lacking SF1 were not viable due to a failure in the proper development of adrenal glands and gonads [11,13,15]. However, when rescued from lethality by adrenal transplantation from WT littermates and corticosteroid injections, mutant mice displayed robust weight gain resulting from both hyperphagia and reduced energy expenditure [16]. Interestingly, a similar phenotype was observed in humans with mutations of the SF1 gene, who often show mild to severe obesity [17,18]. To avoid the confounding metabolic effects of glucocorticoid injections and adrenal transplantations to global SF1 KO mice, Elmquist's group The majority of VMH cells, especially in the dorsomedial and central regions of VMH (VMHdm and VMHc, respectively) [9], express the nuclear receptor steroidogenic factor 1 (SF1) (Figure 1), also known as NR5A1, which is essential for VMH development and function [10–12] and is considered a key marker of that nucleus since SF1 expression is specific and selective to VMH in the brain. Outside the brain, SF1 can be found in the adrenal gland and gonads with differential roles during development [13]. For many years, SF1 was classified as an orphan nuclear receptor, but between 2005 and 2007 several pieces of evidence demonstrated that phospholipids can be ligands of SF1, such as phosphatidic acid being an activator and sphingosine an inhibitor of SF1 transcriptional activity [13]. Interestingly, SF1 can suffer different post-translational modifications, which regulate its stability and transcriptional activity [14], but also control the expression of numerous downstream target genes, including CB1, BDNF, and Crhr2 [13]. Considering this, in order to explore the importance of SF1 neurons, transgenic mice lacking this nuclear receptor were studied by different researchers. Mice lacking SF1 were not viable due to a failure in the proper development of adrenal glands and gonads [11,13,15]. However, when rescued from lethality by adrenal transplantation from WT littermates and corticosteroid injections, mutant mice displayed robust weight gain resulting from both hyperphagia and reduced energy expenditure [16]. Interestingly, a similar phenotype was observed in humans with mutations of the SF1 gene, who often show mild to severe obesity [17,18]. To avoid the confounding metabolic effects of glucocorticoid injections and adrenal transplantations to global SF1 KO mice, Elmquist's group generated an alternative mouse model in which SF1 was specifically deleted in the VMH after completion of the nucleus development using CamKII-Cre [19,20]. These postnatal VMH-specific SF1 KO mice showed increased

generated an alternative mouse model in which SF1 was specifically deleted in the VMH

weight gain and impaired thermogenesis in response to a high-fat diet (HFD), being the first demonstration that the transcription factor SF1 is postnatally required in the VMH for normal energy homeostasis, especially under the HFD condition.

In an attempt to clarify the contribution of this specific population of neurons to hypothalamic regulation of obesity and diabetes, in the last 15 years, several transgenic models have been developed by deleting specific targets in SF1 neurons related to energy balance and glucose homeostasis. In 2013, a profound review article was published by Choi and colleagues [8] summarizing the last updates of SF1 neurons in energy homeostasis. Since then, new neuronal-based approaches (i.e., optogenetic and chemogenetic technology) and the generation of new transgenic mice in key target proteins have provided exciting insight into the implication of SF1 neurons on whole-body energy balance, particularly thermogenesis and glucose homeostasis, that are compiled in the present review.

Despite these studies contributing to understanding the mechanisms by which SF1 neurons regulate energy homeostasis, new questions arose from these results. An important issue to clarify is the heterogeneity of SF1 neurons in terms of glucose, leptin, and insulin sensing or non-sensing neurons, indicating the operation complexity of SF1 neurons and the intricate mechanisms of afferent and efferent neurocircuitry involved in the control of energy balance. In this review, we address the most relevant studies carried out on SF1 neurons to date, from transgenic mice to neurocircuitry studies, in order to discuss the most relevant findings of these investigations and provide an exhaustive overview of the role played by SF1 neurons in the hypothalamic regulation of energy expenditure and glucose homeostasis, and the potential mechanisms involved.

#### **2. Unraveling the Functions of SF1 Neurons in Energy Balance by Optogenetic and Chemogenetic Approaches**

In order to selectively manipulate the SF1 neuronal activity in a physiological context, optogenetic and chemogenetic approaches have emerged [21,22]. These tools use channels that are activated by light and engineered G-protein coupled receptors controlled by exogenous molecules, respectively [21,22]. The incorporation of these approaches into animal models has greatly advanced our understanding of the SF1 role and neuronal circuits.

Mice engineered to activate SF1 neurons by optogenetics were designed through the injection of adeno-associated virus (AAV) particles expressing a Cre-dependent channelrhodopsin (ChRs) into the VMH of SF1-Cre mice to produce SF1-ChRs animals [23,24]. The advantage of this technology is the light-controllable activation of SF1 neurons in a spatiotemporal manner.

One of the first results obtained from this model revealed that SF1 neurons are closely associated with defensive reactions. David J. Anderson and colleagues in 2015, demonstrated that the optogenetic stimulation of SF1 neurons applying a frequency of 20 Hz induced freezing or activity burst, while no response on feeding behavior or energy balance was reported [25]. However, the electrophysiological record of the firing pattern observed in these studies was generated by applying a high-frequency burst of spiking, while the physiological spiking of the neurons in steady-state ranged between 3.5–6.3 Hz [26]. Considering this information, two years later, other researchers showed that SF1 neurons exert a differential effect depending on the frequency of activation. They confirmed that high-frequency activation (>20 Hz) evokes a profound defensive response which includes freezing and escape attempts, but low-frequency activation (2 Hz) suppresses feeding after fasting and reduces the time that mice spend near to the food [24]. These novel results suggest that SF1 neurons dynamically modulate feeding and anxiety-related behaviors by changing the firing pattern and also indicate that this subset of hypothalamic neurons is involved in the fight or flight response.

The chemogenetic modulation of SF1 neurons under potentially more physiological firing patterns has been performed using hM3Dq and hM4Di designer receptors exclusively activated by designer drugs (DREADDs). The expression of DREADDs specifically in SF1 neurons allows to excite or inhibit them in response to clozapine (CNO) or JHU37160 dihydrochloride (J60) which is an in vivo DREADD agonist with high affinity and potency

for hM3Dq and hM4Di receptors [27,28]. In agreement with the feeding behavior-related results obtained from optogenetic approaches, the administration of CNO in fasted SF1 hM3Dq mice revealed a reduction in food consumption. On the contrary, inhibition of SF1 neurons in SF1-hM4Di mice increased cumulative food intake in ad libitum-fed mice. Authors manifested that none of the dramatic defensive behaviors seen with high-frequency optogenetic activation of SF1 neurons was observed [24]. At peripheral levels, they also demonstrated that the chronic chemogenetic modulation of SF1 neurons modified the body fat mass since the continuous administration of CNO (3 weeks) in SF1-hM3Dq and SF1-hM4Di animals reduced and increased the fat mass content, respectively [24]. These changes in fat mass could be explained by the SF1 modulation of fat oxidation. Very recently, it was established that SF1-hM3Dq mice increased energy expenditure and fat oxidation independent of the locomotion activity within 2-h post-activation [23]. Although this study did not evaluate the energy expenditure profile in SF1-hM4Di mice, it was expected that the inactivation of SF1 neurons reduced energy expenditure. This hypothesis is supported by mice expressing tetanus toxin (TT) in SF1 neurons. Since TT prevents neurotransmitter release, SFTT mice displayed reduced energy expenditure and increased body weight [29].

Besides their implication in energy balance, the use of optogenetics has highlighted the role of SF1 neurons in the hypothalamic control of systemic glucose levels. For a long time, it was known that VMH triggered the counterregulatory response (CRR) induced by hypoglycemia [30,31] but it was not clear the contribution of SF1 neurons to this feedback response. In order to elucidate whether SF1 neurons are linked to this effect, an elegant experiment using optogenetic and chemogenetic tools was done by Gregory J Morton and colleagues, showing that selective inhibition of SF1 neurons blocked recovery from insulin-induced hypoglycemia. Conversely, activation of SF1 neurons caused diabetesrange hyperglycemia [32]. This evidence is concordant with those obtained from transgenic models such as mice lacking vesicular glutamate transporter 2 (VGLUT2) specifically in SF1 neurons since this genetic disruption attenuated recovery from insulin-induced hypoglycemia [33].

Taken all together, these genetic approaches have revealed the specific involvement of SF1 neurons in many aspects of metabolic regulation due to their direct or indirect role in the maintenance of the energy balance and glucose levels, confirming the classification of VMH as a primary satiety center [34,35].

#### **3. Manipulation of Key Targets in SF1 Neurons: Lessons from Transgenic Mice**

A particularly powerful strategy developed for the exploration of SF1 neurons in obesity and diabetes has been the design of the SF1 Cre mice. Several groups have generated different SF1 Cre transgenic lines in which the expression of Cre recombinase is derived by *Sf1* regulatory elements [20,36]. These lines allow for ablating general factors or targets known to be associated with energy homeostasis by crossing them with *floxed* strains. In the following sub-sections, we discuss different studies of SF1-CRE transgenic mice organized by the type of molecular target under investigation (Table 1) as well as the sex-specific effect of SF1 neurons in energy balance.

#### *3.1. Hormone Receptors and Related Signaling Pathways*

Due to the importance of the anorectic hormone leptin in the central control of energy homeostasis, the physiological effects after deletion of leptin receptor (LEPR) in SF1 neurons have been thoroughly investigated. The first and most representative study was performed by Bradford Lowell and colleagues, in which genetic deletion of LEPR selectively from hypothalamic SF1 neurons triggered an increase in body weight gain without changes in food intake, leaving these mice unable to adapt to HFD or to activate energy expenditure [20]. To reinforce the key role of the leptin pathway in SF1 neurons, selective inactivation of *Socs3*, a negative mediator of central leptin—pSTAT3 signaling, in SF1 neurons (Figure 2) was developed. Conversely to the deletion of LEPR, mice lacking *Socs3* showed improved

weight-reducing effects of leptin, with a decrease in food intake and an enhanced energy expenditure under chow diet or HFD condition [37]. The importance of leptin signaling in energy balance through SF1 neurons was also reinforced by the specific deletion of the G protein α-subunit Gsα [38]. Gsα couple receptors for hormones, neurotransmitters, and other factors to activate adenyl-cyclase leading to cAMP generation, which is a negative regulator of leptin action (Figure 2). Then, the lack of Gsα in SF1 neurons increases leptin sensitivity [38]. *Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 6 of 25

**Figure 2.** Leptin and insulin signaling in SF1 neurons. Leptin binds LEPR to regulate body weight gain, energy expenditure, and insulin sensitivity in the VMH via STAT3 phosphorylation which is translocated into the nucleus and induces the expression of genes such as SOCS3, a negative regulator of LEPR. Leptin also stimulates a PI3K-dependent pathway, particularly involving subunits p110α, which contributes to energy expenditure regulation, and p110β, which contributes to both brown fat thermogenesis and glucose metabolism regulation in SF1 neurons. The Gsα subunit of G protein-coupled receptors (GPCR) of hormones and the protein-tyrosine phosphatase 1B (PTP1B) negatively regulates leptin action. The activation of this leptin signaling pathway triggers protection against obesity and diabetes under a high-fat diet (HFD) feeding. According to insulin signaling in SF1 neurons, exposure to HFD drives hyperinsulinemia, insulin binds insulin receptor (IR), which in turn activates PI3K that indirectly phosphorylates FOXO1 to **Figure 2.** Leptin and insulin signaling in SF1 neurons. Leptin binds LEPR to regulate body weight gain, energy expenditure, and insulin sensitivity in the VMH via STAT3 phosphorylation which is translocated into the nucleus and induces the expression of genes such as SOCS3, a negative regulator of LEPR. Leptin also stimulates a PI3K-dependent pathway, particularly involving subunits p110α, which contributes to energy expenditure regulation, and p110β, which contributes to both brown fat thermogenesis and glucose metabolism regulation in SF1 neurons. The Gsα subunit of G protein-coupled receptors (GPCR) of hormones and the protein-tyrosine phosphatase 1B (PTP1B) negatively regulates leptin action. The activation of this leptin signaling pathway triggers protection against obesity and diabetes under a high-fat diet (HFD) feeding. According to insulin signaling in SF1 neurons, exposure to HFD drives hyperinsulinemia, insulin binds insulin receptor (IR), which in turn activates PI3K that indirectly phosphorylates FOXO1 to induce transcription factors and promote obesity and glucose metabolism disruption.

induce transcription factors and promote obesity and glucose metabolism disruption. The protein-tyrosine phosphatase 1B (PTP1B) is another negative regulator of leptin signaling in SF1 neurons (Figure 2). Its action is mediated through selective dephosphorylation of the two signaling molecules JAK2 and STAT3. In vivo studies have demonstrated that whole-brain deletion of PTP1B resulted in leanness, hypersensitivity to leptin, and resistance to HFD-induced obesity, a phenotype partly associated with increased hypothalamic activation of STAT3 [39]. Surprisingly, its specific deletion in SF1 neurons resulted in increased adiposity in female mice exposed to HFD due to low energy expenditure, whereas leptin sensitivity was enhanced, and food intake was attenuated, findings that were likely explained by increased STAT3 activation [40]. Mice lacking The protein-tyrosine phosphatase 1B (PTP1B) is another negative regulator of leptin signaling in SF1 neurons (Figure 2). Its action is mediated through selective dephosphorylation of the two signaling molecules JAK2 and STAT3. In vivo studies have demonstrated that whole-brain deletion of PTP1B resulted in leanness, hypersensitivity to leptin, and resistance to HFD-induced obesity, a phenotype partly associated with increased hypothalamic activation of STAT3 [39]. Surprisingly, its specific deletion in SF1 neurons resulted in increased adiposity in female mice exposed to HFD due to low energy expenditure, whereas leptin sensitivity was enhanced, and food intake was attenuated, findings that were likely explained by increased STAT3 activation [40]. Mice lacking PTP1B in SF1 neurons also had improved leptin and insulin signaling in VMH, suggesting that increased insulin responsiveness in SF1 neurons could overcome leptin hypersensitivity and promote adiposity [39,40].

obesity in this null mouse [41]. This finding could explain that this neuronal population expressing LEPR works in conjunction with other types of neurons expressing the same receptor, and SF1 neurons by themselves cannot compensate for all receptor deficiency. Summing up, leptin signaling in SF1 neurons plays a key role in energy homeostasis regulation and mediates the proper physiological adaptation to HFD to avoid or delay the

According to glucose metabolism, leptin has been long related to glucose homeostasis improving insulin sensitivity, since intra-VMH injection of leptin increases

hypersensitivity and promote adiposity [39,40].

onset of obesity.

PTP1B in SF1 neurons also had improved leptin and insulin signaling in VMH, suggesting that increased insulin responsiveness in SF1 neurons could overcome leptin

A more recent study tried to rescue native LEPR in SF1 neurons in *LepR*-deficient mice. They concluded that LEPR signaling in the VMH is not sufficient to protect against obesity in this null mouse [41]. This finding could explain that this neuronal population expressing LEPR works in conjunction with other types of neurons expressing the same receptor, and SF1 neurons by themselves cannot compensate for all receptor deficiency. Summing up, leptin signaling in SF1 neurons plays a key role in energy homeostasis regulation and mediates the proper physiological adaptation to HFD to avoid or delay the onset of obesity.

According to glucose metabolism, leptin has been long related to glucose homeostasis improving insulin sensitivity, since intra-VMH injection of leptin increases glucose uptake in peripheral tissues [42] and normalizes hyperglycemia [43]. Very recently it has been demonstrated that central leptin infusion in mice with SF1 neuron-specific LEPR deficiency corrected diabetic hyperglycemia [43]. Despite this result, re-expressing the receptor in SF1 neurons of null mice showed that SF1 neurons were not sufficient to mediate the antidiabetic action of leptin [44]. The essential action of leptin in SF1 to correct diabetic hyperglycemia was clarified by further investigations. Particularly, in the specific knock—out of *Socs3* in SF1 neurons, where leptin signaling is over-activated, Ren Zhang and colleagues observed improved glucose homeostasis, showing protection against hyperglycemia and hyperinsulinemia caused by HFD feeding [37]. These studies demonstrate that leptin in VMH neurons improves glucose and insulin metabolism, although this area is not essential as there are redundant neuronal circuits regulating it. Optogenetic activation of SF1 neurons has the same output as leptin increasing glucose uptake but it does not normalize blood glucose levels, which leads to an understanding of two different subsets of SF1 neurons, one subset increasing insulin sensitivity and the second one increasing blood glucose levels.

Another key hormone implicated in the balancing of energy metabolism is insulin. It is known that insulin acutely suppresses food intake and decreases fat mass in both rodents and humans [45,46] (Figure 2). Mice lacking insulin receptors (IR) in SF1 neurons did not show any differences in body weight when fed a chow diet but under HFD conditions, mutant mice were protected against obesity and showed an enhanced leptin sensitivity and glucose homeostasis [26]. Interestingly, exposure to HFD led to the overactivation of insulin in the VMH, leading to a reduction in SF1 neurons firing frequency, in comparison to the insulin resistance induced in ARC neurons. The differential dysregulation of insulin action in these hypothalamic nuclei under HFD could indicate cooperation between these responses to drive obesity [26]. These findings also suggest that hyperinsulinemia evoked by HFD feeding controls SF1 neuronal activity, which leads to changes in the synaptic inputs to other neuronal populations such as POMC, resulting in obesity and diabetes. However, the specific contribution of insulin signaling in SF1 neurons and its relationship to peripheral insulin resistance and glucose levels needs further investigation.

Leptin depolarizes or hyperpolarizes SF1 neurons, depending on the subpopulation, while insulin only hyperpolarizes them, as both actions are in a downstream PI3Kdependent manner [20,47–49] (Figure 2). PI3K is formed by two different subunits, p85 and p110, and specifically, the subunit p110β is necessary for depolarizing and hyperpolarizing SF1 neurons, while p110α is only needed in the hyperpolarization process [47]. Mice lacking p110α in SF1 neurons had reduced energy expenditure in response to hypercaloric feeding and, therefore, displayed an obesogenic phenotype. Moreover, this p110α subunit was not required to regulate glucose metabolism in SF1 neurons [49]. Mice lacking p110β in the same neuronal population had also decreased energy expenditure (reduced thermogenesis) leading to increased susceptibility to obesity, whereas, in contrast to the p110α subunit, p110β involved changes in peripheral insulin sensitivity [50]. In line with this evidence, deletion of FOXO1, a downstream transcription factor of insulin-PI3K (Figure 2), in SF1 neurons resulted in a lean phenotype with high energy expenditure, even in fasting, and these null mice presented an enhanced insulin sensitivity and glucose tolerance, in concordance with genetic deletion of IR in SF1neurons [51]. Although leptin and insulin can inhibit SF1 neurons using the same molecular cascade, they are anatomically segre-

gated within the VMH (neurons expressing LEPR receptor are located in the VMHdm when depolarizing and scattered throughout the nucleus when hyperpolarizing, whereas those expressing IR are in the VMHc close to the ventricle) [47], which would explain the different effects observed when deleting their receptors.

Other hormones studied in SF1 neurons are estrogens. Female mice lacking the estrogenic receptor α (ERα) in SF1 neurons were obese due to a reduced energy expenditure [52]. Ablation of the ERα led to abdominal obesity with adipocyte hypertrophy in females, but not in male mice [53]. Despite the fact that most of the studies on SF1 neurons until now were performed only in male mice, these last results described, and others discussed later [40,54], reinforce the notion that SF1 neurons may have a sex-specific effect on energy balance and glucose metabolism.

Growth hormone (GH) also plays a role in glucose metabolism via SF1 neurons. GH is secreted in a metabolic stress situation such as hypoglycemia. Deletion of its receptor in SF1 neurons resulted in an impaired capacity for recovery from hypoglycemia [55]. These mice showed an altered CRR due to changes in the neurocircuit that regulates the parasympathetic nervous system. This result supports the importance of SF1 neurons in the proper functionality of glucose homeostasis.

Altogether, these findings identify SF1 neurons (the predominant VMH population) as a key player in the regulation of energy expenditure and glucose homeostasis, being particularly important in the adaptive response to HFD feeding. Most of the mutant mice with deletion of several hormone receptors and associated proteins in SF1 neurons have no changes or mild metabolic alterations under chow diet, but they show substantial metabolic variations under HFD exposure. The action of hormones and related proteins in SF1 neurons is also involved in the CRR to hypoglycemia to maintain glucose balance between the brain and the periphery. Future studies are needed to describe the specific molecular mechanisms and subsets of SF1 neurons underlying the effects of hormones in glucose homeostasis and energy expenditure.

*Int. J. Mol. Sci.* **2021**, *22*, 6186


**Table 1.** Genetic models developed to study SF1 neurons in energy balance.

*Int. J. Mol. Sci.* **2021**, *22*, 6186


**Table 1.** *Cont.*

149

significant decrease in BAT thermogenesis. (d) Increased sympathetic activity in brown fat.

#### *3.2. Nutrient Sensors: AMPK and SIRT1*

The regulation of energy homeostasis is importantly regulated by proteins acting as nutrient and energy sensors, such as AMPK and SIRT1, which restore energy balance during metabolic challenges both at the cellular and physiological levels (Figure 3). A link between both AMPK and SIRT1 in hypothalamic SF1 neurons with central control of obesity and diabetes has been established. *Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 8 of 25

**Figure 3.** Nutrient sensors in SF1 neurons regulate energy homeostasis. AMPK and SIRT1 are key nutrient sensors that restore energy balance under low energy availability. The role of both AMPK and SIRT1 in the central regulation of energy balance depends on their specific location in hypothalamic nuclei. AMPK in SF1 neurons of the VMH particularly regulates energy expenditure by acting in BAT thermogenesis and subcutaneous WAT browning (via sympathetic output) and liver lipogenesis (via parasympathetic output), whereas AMPK effects in feeding control and glucose balance mostly rely on ARC neurons. SIRT1 in SF1 VMH neurons controls insulin sensitivity in skeletal muscle protecting against diet-induced obesity, but in POMC neurons of the ARC, SIRT1 regulates normal BAT-like remodeling in specific fat depots. SIRT1 in SF1 neurons also contributes to the synchronization of the circadian clock in the brain by nutritional inputs. The arrow means increase and + means activation. **Figure 3.** Nutrient sensors in SF1 neurons regulate energy homeostasis. AMPK and SIRT1 are key nutrient sensors that restore energy balance under low energy availability. The role of both AMPK and SIRT1 in the central regulation of energy balance depends on their specific location in hypothalamic nuclei. AMPK in SF1 neurons of the VMH particularly regulates energy expenditure by acting in BAT thermogenesis and subcutaneous WAT browning (via sympathetic output) and liver lipogenesis (via parasympathetic output), whereas AMPK effects in feeding control and glucose balance mostly rely on ARC neurons. SIRT1 in SF1 VMH neurons controls insulin sensitivity in skeletal muscle protecting against diet-induced obesity, but in POMC neurons of the ARC, SIRT1 regulates normal BAT-like remodeling in specific fat depots. SIRT1 in SF1 neurons also contributes to the synchronization of the circadian clock in the brain by nutritional inputs. The arrow means increase and + means activation.

The identification of SF1 neurons as the main neuronal AMPK-expressing population in VMH to be targeted in obesity was demonstrated by several studies of Miguel Lopez lab. They generated an SF1 specific AMPKα1 null mouse by Cre-Lox recombination. Ablation of AMPKα1 in these neurons led to feeding-independent weight loss associated with an increase in energy expenditure [56]. In accordance with this, SF1-Cre AMPKα1flox/flox mice displayed BAT activation, which was confirmed by increased BAT temperature and UCP1 protein expression, elevated sympathetic activity, and higher 18F-FDG uptake in brown fat [56]. SF1-Cre AMPKα1flox/flox mice fed with a long-term HFD had AMPK is a highly conserved master regulator of metabolism, which is activated under low energy conditions, increasing energy production and reducing energy waste. Current evidence demonstrates the critical role of hypothalamic AMPK in the regulation of food intake and energy expenditure, as well as glucose and lipid homeostasis at the whole-body level [56,64–67]. Interestingly, the regulation of these effects depends on the anatomical location of AMPK in the hypothalamus: AMPK effects in feeding control and glucose homeostasis mostly rely on ARC neurons [64,66,67], whereas its effects in energy expenditure arise from the VMH [56,65,68], particularly in SF1 neurons (Figure 3).

a feeding-independent decrease in body weight and adiposity, associated with increased energy expenditure and VO2, higher BAT thermogenesis activation, and browning of subcutaneous WAT. These findings point to AMPKα1, but not AMPKα2, in SF1 neurons The identification of SF1 neurons as the main neuronal AMPK-expressing population in VMH to be targeted in obesity was demonstrated by several studies of Miguel Lopez lab.

as the main catalytic AMPK subunit in the VMH that regulate thermogenic control,

thermogenesis [65], WAT browning, and lipid metabolism are mediated by AMPKα<sup>1</sup> attenuation in SF1 neurons [68] strongly support the notion that specific targeting of the

particularly under HFD exposure.

They generated an SF1 specific AMPKα<sup>1</sup> null mouse by Cre-Lox recombination. Ablation of AMPKα<sup>1</sup> in these neurons led to feeding-independent weight loss associated with an increase in energy expenditure [56]. In accordance with this, SF1-Cre AMPKα<sup>1</sup> flox/flox mice displayed BAT activation, which was confirmed by increased BAT temperature and UCP1 protein expression, elevated sympathetic activity, and higher <sup>18</sup>F-FDG uptake in brown fat [56]. SF1-Cre AMPKα<sup>1</sup> flox/flox mice fed with a long-term HFD had a feedingindependent decrease in body weight and adiposity, associated with increased energy expenditure and VO2, higher BAT thermogenesis activation, and browning of subcutaneous WAT. These findings point to AMPKα1, but not AMPKα2, in SF1 neurons as the main catalytic AMPK subunit in the VMH that regulate thermogenic control, particularly under HFD exposure.

Other studies demonstrating that the stimulatory effect of thyroid hormones in BAT thermogenesis [65], WAT browning, and lipid metabolism are mediated by AMPKα<sup>1</sup> attenuation in SF1 neurons [68] strongly support the notion that specific targeting of the discrete neuronal population in the VMH impacts obesity in a feeding-independent, but the thermogenic-dependent manner [56]. These studies also suggest that simultaneous targeting of AMPK in SF1 neurons of the VMH and neurons in the ARC would allow for controlling both food intake and energy expenditure by inhibiting a common protein.

Although the protection of AMPKα<sup>1</sup> deletion in SF1 neurons against HFD-induced obesity was mainly associated with increased energy expenditure, slight alterations in glucose balance (decreased glycemia and improved glucose tolerance without changes in insulin sensitivity and insulin levels) found in mutant mice fed with an HFD, could be involved [56]. The moderate role of AMPK in glucose sensing by SF1 neurons in the VMH was also reported in a recent study in which AMPK activity suppression led to selective depletion of SF1 glucose inhibitory (GI) neurons, and activated CRR without affecting the presence of glucose excited (GE) neurons [69]. The authors suggested that the primary role of AMPK in SF1 GI neurons is to control the expression of *Txn2* (encoding a mitochondrial redox enzyme), providing protection against ROS produced during hypoglycemia. Conversely, under hyperglycemia, a reduction in ROS production is mediated by UCP2 in GE SF1 neurons [62]. These findings support the idea of two different subsets of SF1 neurons: the one regulating hyperglycemia enhancing insulin sensitivity and the second one regulating hypoglycemia by activating CRR.

In addition to AMPK, the metabolic sensor of nicotinamide adenine dinucleotidedependent deacetylase SIRT1, which is highly expressed in the VMH, has shown to contribute to both control of energy balance and clock function from SF1 neurons [70–72]. SIRT1 deacetylates several proteins in central and peripheral tissues to provide adaptation against redox/nutrient challenges. SIRT1 drives lipid mobilization from adipose tissue, lipid oxidation in skeletal muscle and liver, and increased glucose production, all of these responses protect against diminished energy availability [70]. In the hypothalamus, SIRT1 has been revealed as an attractive target against obesity and type 2 diabetes in both POMC [73] and SF1 [57] neurons. SIRT1 activation in these hypothalamic neurons triggers a reduction in body adiposity and insulin resistance in obesogenic environments, while the deleterious effects, such as increased glucose and lipid output from the liver, would be avoided.

As observed with AMPK, the anatomical location of SIRT1 in the hypothalamus is relevant for the protection against diet-induced obesity: although SIRT in POMC neurons is required for normal BAT-like remodeling in the perigonadal fat depot [73], SIRT1 in SF1 neurons controls insulin sensitivity in the skeletal muscle [57] (Figure 3). Mice lacking SIRT1 in SF1 neurons showed hypersensitivity to hypercaloric diets, gaining more bodyweight and lower energy expenditure compared to control mice after long-term HFD, whereas food intake remained unchanged [57].

In addition, to protect against dietary obesity, SIRT1 in SF1 neurons is relevant for homeostatic responses that can protect against dietary diabetes, since mice overexpressing SIRT1 exclusively in SF1 neurons were protected from developing diet-induced insulin

resistance in skeletal muscle and hyperglycemia, whereas mice lacking SIRT1 in the same neurons were more prone to develop insulin resistance in skeletal muscle after HFD [57]. It is important to note that these studies indicate that SIRT1 in SF1 neurons does not regulate food intake in either chow or HF feeding and that the regulatory effects on energy expenditure and glucose homeostasis are predominantly observed in a hypercaloric diet but not in a normal diet context. The latter evidence is intriguing since SIRT1 is known to be activated in the low-energy state. According to Coppari [70], the possible explanations are (i) HFD increases SIRT1 activity in SF1 neurons, and (ii) in chow-fed mice, compensatory mechanisms (i.e., enhancement of insulin sensitivity in the liver to provide normal glycemia) counterbalance the defects related to SIRT1 deficiency in these hypothalamic neurons. These hypotheses could explain that the metabolic imbalance is mainly observed in transgenic mice exposed to an HFD when the compensatory mechanisms are unable to prevent the whole-body effects produced to the lack of SIRT1 in hypothalamic neurons (i.e., insulin sensitivity is reduced in the skeletal muscle but is not increased in the liver).

In addition to these investigations, SIRT1 also contributes to the synchronization of the circadian clock in the brain, by nutritional inputs, which is crucial to maintaining cellular and body homeostasis [71]. Mice with targeted ablation of SIRT1 in SF1 neurons of the VMH showed a reduction in the period length in the light-entrainable activity under constant darkness and scheduled feeding, but not in ad libitum feeding [71]. Furthermore, these effects were more pronounced in KO mice under scheduled high-caloric diet feeding. This investigation suggests that SIRT1 in SF1 neurons contributes to the synchronization and/or adaptation of the endogenous clock to feeding cues (Figure 3), a function that would partially explain the metabolic alterations displayed by mice deficient in SIRT1 in SF1 neurons previously reported [57]. Accordingly, deletion of BMAL1 in SF1 neurons of the VMH in mice also drove a substantial alteration of the circadian clock in the brain, with an impact in BAT activation, supporting the importance of SF1 neurons in the regulation of clock function in the brain and the whole-body energy homeostasis [72]. The implication of SIRT1 in SF1 neurons in healthy aging or lifespan via circadian clock modulation would be also interesting to explore and would provide insight into the hypothalamic mechanisms mediated by SF1 neurons regulating energy balance.

Overall, specific targeting of metabolic sensors such as AMPK and SIRT1 in SF1 neurons of the VMH would be enough to modulate body weight gain and energy expenditure, in a feeding-independent manner, but associated with synchronization of the central clock with the periphery. This is a challenging approach but with significant translational relevance in obesity management.

#### *3.3. Glutamatergic Neurotransmission and Synaptic Receptors*

The expression and activity of synaptic receptors and glutamatergic neurotransmission in SF1 neurons of the VMH have been identified to contribute to energy balance dysregulation in obesity and diabetes.

Since neurons in the VMH are largely glutamatergic, the importance of glutamate neurotransmission versus other neurotransmitters and neuropeptides in mediating the functions of these nuclei seem to be higher, not only in energy homeostasis but also in sexual and defensive behaviors [8,33]. The first study reporting this evidence in SF1 neurons was provided by Bradford Lowell and colleagues [33]. They generated a mouse lacking the VGLUT2—a transporter mediating uptake of L-glutamate into synaptic vesicles specifically in SF1 neurons, driving to a disruption in transsynaptic communication by glutamate in this specific population of neurons. The phenotype study was particularly focused on glucose and insulin homeostasis: these mice displayed hypoglycemia in the fasted state secondary to impaired fasting-induced glucagon increased and impaired gluconeogenesis induction in the liver [33] (Figure 4). VGLUT2 deficient mice in SF1 neurons also had defective CRR to insulin-induced hypoglycemia and central 2-deoxyglucose, as this greater degree of hypoglycemia again linked to an impaired glucagon response. It is important to mention that SF1-Cre;Vglut2flox/flox mice did not become obese when exposed

to a chow diet and had mild obesity when fed with an HFD [33]. A partial contribution of GABA release, from a small number of SF1 neurons, in body weight regulation in these mice could explain this mild obesogenic phenotype [33]. Importantly, this study was the first demonstration of the importance of glutamate release in VMH neurocircuitry to prevent hypoglycemia. *Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 8 of 25

**Figure 4.** Glutamatergic neurotransmission and synaptic receptors in SF1 neurons and energy balance. Lack of the glutamate vesicular transporter VGLUT2 in SF1 neurons disrupts the glutamate effect in the postsynaptic receptors, leading to glucose metabolism dysfunction and a defective counterregulatory response (CRR). In line with this, the glutamate receptor mGluR5 in SF1 neurons also regulates glucose balance. The calcium channel subunit α2δ-1 is essential for excitability and firing activity of SF1 neurons which in turn regulate glucose metabolism in adipose tissue and skeletal muscle and lipolysis in white adipose tissue with no alterations in food intake. Finally, the endocannabinoid 2-arachidonoylglycerol (2-AG) released in the post-synaptic neuron, leads to retrograde activation of the CB1 receptor located in the presynaptic glutamatergic SF1 neurons, blocking calcium entry with subsequent inhibition of glutamatergic output and attenuation of POMC neurons activation, promoting food intake in diet-induced obesity. **Figure 4.** Glutamatergic neurotransmission and synaptic receptors in SF1 neurons and energy balance. Lack of the glutamate vesicular transporter VGLUT2 in SF1 neurons disrupts the glutamate effect in the postsynaptic receptors, leading to glucose metabolism dysfunction and a defective counterregulatory response (CRR). In line with this, the glutamate receptor mGluR5 in SF1 neurons also regulates glucose balance. The calcium channel subunit α2δ-1 is essential for excitability and firing activity of SF1 neurons which in turn regulate glucose metabolism in adipose tissue and skeletal muscle and lipolysis in white adipose tissue with no alterations in food intake. Finally, the endocannabinoid 2-arachidonoylglycerol (2-AG) released in the post-synaptic neuron, leads to retrograde activation of the CB1 receptor located in the presynaptic glutamatergic SF1 neurons, blocking calcium entry with subsequent inhibition of glutamatergic output and attenuation of POMC neurons activation, promoting food intake in diet-induced obesity.

The relevant role of glutamatergic neurotransmission in SF1 neurons to regulate energy homeostasis was also recently established by the development of mice lacking the metabotropic glutamate receptor mGluR5 in these neurons [54]. The most remarkable finding of this study was that mGluR5 depletion in SF1 neurons did not affect energy balance, but it significantly impaired glucose balance control (Figure 4), as this dysfunction was only observed in female but not in male mice. The sex-specific impairment of insulin sensitivity and glucose tolerance was associated with a reduction in intrinsic excitability and the firing rate of SF1 neurons. This suggests a functional indirect interaction of mGluR5 with estrogen receptors to control the effects of estradiol on SF1 neurons activity and glycemic control, which switch from facilitatory to detrimental in the absence of the glutamate receptor in these neurons [54]. These results The relevant role of glutamatergic neurotransmission in SF1 neurons to regulate energy homeostasis was also recently established by the development of mice lacking the metabotropic glutamate receptor mGluR5 in these neurons [54]. The most remarkable finding of this study was that mGluR5 depletion in SF1 neurons did not affect energy balance, but it significantly impaired glucose balance control (Figure 4), as this dysfunction was only observed in female but not in male mice. The sex-specific impairment of insulin sensitivity and glucose tolerance was associated with a reduction in intrinsic excitability and the firing rate of SF1 neurons. This suggests a functional indirect interaction of mGluR5 with estrogen receptors to control the effects of estradiol on SF1 neurons activity and glycemic control, which switch from facilitatory to detrimental in the absence of the

are in line with the previously described study reporting that ERα depletion in SF1

glutamate receptor in these neurons [54]. These results are in line with the previously described study reporting that ERα depletion in SF1 neurons drives metabolic disturbances only in females [52]. However, the lack of impairment in insulin-induced hypoglycemia in mice deficient in mGluR5 is in contrast to the previous investigation in which inhibition of glutamate release by SF1 neurons impaired the CRR to hypoglycemia [33].

Other receptors involved in neuronal activity have shown their contribution to the regulation of obesity and diabetes by SF1 neurons in the VMH. Among them, the calcium channel subunit α2δ-1 has a non-canonical role in SF1 neurons to control glucose and lipid homeostasis (Figure 4). This neuronal receptor facilitates cell surface trafficking of calcium channels mediating calcium current and neurotransmitters release [74] and also acts as a receptor for thrombospondins, promoting excitatory synapse assembly [75]. The authors show that α2δ-1 is essential for the excitability and firing activity of SF1 neurons. Deletion of this receptor in SF1 neurons in mice led to neuronal hypoactivity, sympathetic tone reduction to WAT and skeletal muscle, glucose intolerance, and insulin resistance, and lipolysis alteration in WAT [58], as these metabolic alterations mainly observed in female mice [59]. Despite these findings, mice did not display alterations in food intake and body weight in response to HFD [58]. The reduction in SF1 neuronal activity observed in this transgenic mouse did not compromise CRR to hypoglycemia, in contrast to previous studies in which chemogenic or optogenetic silencing of these neurons impaired recovery from hypoglycemia [32,76]. These conflicting results could be related to the heterogeneity of SF1 neurons, with functional diversity in the VMH (e.g., leptin-activated and leptin-inhibited SF1 neurons) [47].

The endocannabinoid receptor CB1 located in SF1 neurons of the VMH has been also involved in glutamatergic output regulating feeding and in the control of metabolic flexibility in response to dietary changes. Hypothalamic slice preparations exposed to a CB1 receptor-specific agonist, displayed inhibition of the electrical activity of SF1 neurons [77]. In line with this, endocannabinoids release in postsynaptic neurons, led to retrograde activation of CB1 receptors in pre-synaptic glutamatergic SF1 neurons of the VMH, blocking calcium entry with subsequent inhibition of glutamatergic output and attenuation of POMC neurons activation, promoting feeding in diet-induced obesity [78] (Figure 4). The physiological relevance of CB1 receptors in SF1 neurons was also reported by Daniela Cota and colleagues, showing that mice lacking CB1 receptor specifically in these neurons showed a diet-dependent bidirectional metabolic phenotype: under a chow diet, deletion of CB1 decreased adiposity by increasing sympathetic activity and WAT lipolysis, whether, conversely, under HFD, lack of this receptor in SF1 neurons blunted peripheral use of lipids and produced leptin resistance, hyperphagia, body weight gain and glucose intolerance [60]. Therefore, CB1 synaptic receptor in SF1 neurons seems to act as a molecular switch for correct metabolic flexibility in the hypothalamus, protecting from the development of HFD-induced leptin resistance.

Altogether, although these studies provide substantial evidence of the role of glutamatergic neurotransmission and neuronal receptors in SF1 neurons to control glucose homeostasis and metabolic adaptation to dietary challenges, they raise important questions that still need to be clarified such as (i) the identification of the exact afferent and efferent components of the neurocircuitry linking SF1 neurons with other brain areas and the periphery (e.g., for the regulation of glucagon release in the pancreas), (ii) the evaluation of sex-dependent function, (iii) the exploration of SF1 neuronal subpopulations in response to different stimuli (e.g., important for the compensatory effects on feeding or the CRR to hypoglycemia).

#### *3.4. Modulators of Autophagy, Mitochondrial and Primary Cilia Function*

Autophagy plays a critical role in several physiological processes such as metabolic regulations. Deletion of essential autophagy genes, such as the autophagy-related gene (Atg) 7 in peripheral tissues drives significant changes in body weight and glucose balance [79,80]. More recent investigations have demonstrated that autophagy is important

in the hypothalamic regulation of energy homeostasis and feeding behavior, with these effects affected differently depending on the neuronal types engaged. In the ARC, *Atg7* deletion in POMC neurons implied an obesogenic phenotype, especially when feeding an HFD [81,82], whereas loss of this gene in AgRP neurons promoted leanness [83]. A more recent study reported the metabolic importance of autophagy in the VMH, by deleting *Atg7* in SF1 VMH neurons of mice [61]. The authors found that in fed conditions, loss of *Atg7* in this neuronal population has no effect on body weight, food intake, or energy expenditure. However, when fasted overnight, mice lacking *Atg7* displayed reduced O<sup>2</sup> consumption, CO<sup>2</sup> and heat production, and reduced food intake after refeeding. *Atg7* deletion in SF1 neurons only implied moderate changes in fed plasma insulin levels and insulin resistance, without changes in glycemia and glucose tolerance.

This study demonstrates autophagy activation in the VMH, in addition to other regions such as the ARC, in response to fasting. This was supported not only by increased expression of autophagy-related proteins in the area but also by a 2- to 3-fold increase in the density of cFos-immunoreactive cells in the VMH, ARC, and DMH of control mice in response to fasting, an induction that was abrogated in these three regions in *Atg7* deficient mice in SF1 VMH neurons [61]. Then, the lack of autophagy in SF1 VMH neurons significantly impacts neurons outside of the VMH. This is in line with previous investigations showing that POMC neurons in the ARC receive excitatory inputs from the VMH, and the strength of this VMH-ARC input is reduced by fasting [84].

Loss of autophagy in SF1 VMH neurons also caused alterations in mitochondrial function and morphology, which was associated with inadequate metabolic response to fasting [61]. During fasting, mitochondrial efficiency must increase, and autophagy is important to ensure the degradation of mitochondria that are damaged beyond repair, that is, the process called mitophagy [85]. Mice lacking *Atg7* in SF1 neurons showed a reduced number of mitochondria-autophagosome contacts with alterations in mitochondria morphology and respiration, indicating abnormal mitophagy [61]. Since UCP2 expression in SF1 VMH neurons has been linked to mitophagy [62,86], the authors explored the expression of this protein, and they found it was dysregulated in mutant mice [61]. In agreement with these findings, UCP2 in SF1 VMH regulates mitochondria fission and the excitability of GE neurons in the VMH, which in turn allows the appropriate response to increased glycemia, enhancing insulin sensitivity in peripheral tissues [62]. However, the moderate effects of autophagy loss in SF1 neurons observed in glucose homeostasis do not correspond to the abnormal expression of UCP2 detected. The fact that UCP2-induced mitochondrial fission in the VMH only affects GE neurons, but not GI neurons [62], could explain the moderate phenotype of the mice deficient in *Atg7* in glucose metabolism.

Very recently, the SF1 neurons' primary cilia, a solitary antenna-like extension of the plasma membrane, was demonstrated as an important organelle for the regulation of energy homeostasis [63]. In this study, the authors deleted the intraflagellar transport 88 (IFT88), which is a critical protein in primary cilium biogenesis, specifically in SF1 neurons. Dysfunction of VMH primary cilia resulted in impaired activation of sympathetic tone, central leptin resistance, and higher body weight gain under both chow and HFD feeding. Obesity in the transgenic mice IFT88-KOSF−<sup>1</sup> was caused by a marked decrease in energy expenditure, hyperphagia (only under HFD feeding), blunted BAT function, insulin resistance, and glucose intolerance. In addition, deletion of primary cilia in SF1 neurons led to an impairment in bone homeostasis. The downregulation of the sympathetic activity in IFT88-KOSF−<sup>1</sup> mice was linked to the reduction of the metabolic rate and increase in bone density, as this effect considered was independent of the obese phenotype [63]. Considering the fact that primary cilia are required for proper neuronal circuit formation, deletion of this organelle in SF1 neurons may therefore influence local neuronal circuits [63].

#### **4. The Sex-Specific Effect of SF1 Neurons on Energy Balance**

It is known that the VMH is sexually dimorphic, showing females higher ERα concentration than males [87]. Additionally, as described in Section 3, selective deletion of this

receptor in SF1 neurons resulted in increased abdominal obese phenotype with adipocyte hypertrophy in females, but not in males [52]. Obesity in females was caused by decreased energy expenditure as they had reduced basal metabolic rate and impaired BAT thermogenesis [52]. It has been described also that estrogens regulate the activity of GI neurons of the ventrolateral portion of VMH, since females showed an attenuated response to hypoglycemia compared to male mice, although in this study it is not specified if these neurons were SF1 positive cells [88]. As has been already specified, there are other genetic deletions that presented different outputs in a sex-specific manner. Expression of mGluR5 in SF1 neurons was necessary for estradiol protective effects in glucose balance in female but not in male mice, and mGluR5 deletion resulted in reduced electrical activity only in female mice [54]. Deletion of α2δ-1 expression in SF1 neurons also presented sexually dimorphic effects depending on diet conditions. Particularly, female mice lacking α2δ-1 displayed glucose intolerance and insulin resistance under chow or HFD, as this phenotype much moderate in male mice [58,59]. Cheung et al. also observed sex-dependent changes when deleting VGLUT2 in SF1 neurons, since female but not male mice presented attenuation of DIO, and transgenic male mice showed behavioral changes not observed in female mice [89].

Estrogens could affect and change some intracellular signaling cascades leading to these differences observed between males and females, as female mice have greater expression of ERα. This would explain why the presence or absence of different patterns of receptors impacts the estrogenic effect on SF1 neurons, as some of the sex-specific signaling cascades are being altered. Other sex-specific effects related to SF1 neurons can be found in their synapses with POMC neurons, as estradiol attenuated the retrograde endocannabinoid signaling from POMC to SF1 neurons, increasing the glutamatergic inputs to POMC [78]. This study opens another possibility based on the fact hypothesis that estrogens do not interact directly with the specific SF1 neurons targeted in each study, but their effects fall on other cells acting on SF1 neurons.

Then, estrogen signaling in SF1 neurons is a must for metabolic health in female mice. Despite the fact that many studies presented in this review were performed only in male mice, these last results obtained in both sexes reinforce the idea of SF1 neurons having a sex-specific effect on energy balance and glucose metabolism.

#### **5. Exploring the Neurocircuitry That Links SF1 Neurons to Other Brain Areas in Energy Balance**

Remarkable progress in the neuroanatomy of VMH projections has been obtained from novel biological tools. Some notions on how the VMH connects with other brain centers come from stereotaxic injection with anterograde axonal tracers in adult rats [90,91]. While the results from these studies established that VMH is organized in subregions depending on their projection, the inherent limitations of this method make it difficult to assess the specific neuronal network involved in SF1 neurons. To overcome these limitations, *Sf-1 TauGFP* and *Z/EGSf1:Cre* mice models were originally designed to trace the major VMH axonal projections during embryonic and postnatal stages [9].

The first of these models is a knock-in line that contains the wheat germ agglutinin (WGA) and Tau-green fluorescent protein (TauGFP) under the control of *Sf-1* regulatory elements. In the second model, the SF1 Cre mouse was crossed with a *Z/EG* reporter mouse, resulting in constitutive expression of eGFP (enhanced GFP) after Cre-mediated recombination [9]. The analysis of the results obtained from both models indicated the efferent SF1 projection in ascending and descending tracts that innervate different structures such as the hypothalamus, thalamus, the basal forebrain, and the brainstem. Interestingly, SF1 projections targeting the general vicinity of gonadotropin-releasing hormone (GnRH) neurons suggest the potential role in fertility physiology [9].

As expected, experiments using these models showed that SF1 neurons projected to areas implicated in body weight regulation, including the paraventricular nucleus of the hypothalamus (PVN) [9,92]. However, it was not possible to detect any GFP<sup>+</sup> fibers within the ARC. Thus, additional studies using synaptophysin or rabies virus would be

helpful to explore the specific network between SF1 and hypothalamic nuclei involved in energy balance. Recently, Yunglei Yang and colleagues have optogenetically identified the downstream target underlying the SF1 suppression of food intake (Figure 5). The authors used the SF1 ChR2 mice, but the fiber optic cannula was implanted above the PVN, therefore the photostimulation of this area would only activate the SF1-PVN projections. They showed that high-frequency stimulation of this circuit potently reduced food intake even in 24 h food-deprived mice [23]. *Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 8 of 25

**Figure 5.** Neurocircuitry linking SF1 VMH neurons to other brain areas in energy balance. Using optogenetic tools, the paraventricular nucleus of the hypothalamus (PVN) has been identified as a downstream target nucleus underlying the SF1 suppression of food intake. Downstream projections of SF1 in the VMH involved in the control of glycemia are the bed nucleus of the stria terminalis (aBNST) since photostimulation of the VMHSf1-aBNST circuit leads to increases in blood glucose levels. Finally, SF1 neurons are innervated by upstream BNST neurons and send projection to the nucleus tractus solitarius (NST) to regulate anxiety-like behavior and bone metabolism. **Figure 5.** Neurocircuitry linking SF1 VMH neurons to other brain areas in energy balance. Using optogenetic tools, the paraventricular nucleus of the hypothalamus (PVN) has been identified as a downstream target nucleus underlying the SF1 suppression of food intake. Downstream projections of SF1 in the VMH involved in the control of glycemia are the bed nucleus of the stria terminalis (aBNST) since photostimulation of the VMHSf1-aBNST circuit leads to increases in blood glucose levels. Finally, SF1 neurons are innervated by upstream BNST neurons and send projection to the nucleus tractus solitarius (NST) to regulate anxiety-like behavior and bone metabolism.

Downstream projections of SF1 neurons regulating glycemia were identified through

the same novel strategy derived from optogenetics. The researchers administrated fluorescently tagged ChRs into the VMH of SF1-Cre mice to visualize the projection fields using histological imaging to detect the EYFP reporter in axonal projections [32]. Once identified this target site, a fiber optic was implanted to depolarize the final projection by laser stimulation. With this strategy, it was possible to demonstrate that VMHSF1→aBNST is the most relevant circuit involved in controlling glycemia since the photostimulation of this circuit triggered an increase of the blood glucose levels [32] (Figure 5). Interestingly, other functions have been attributed to this specific circuit: Fan Yang and colleagues recently demonstrated that SF1 neurons are innervated by upstream BNST neurons and send projections to the NST nucleus to regulate anxiety-like behavior and bone metabolism [93] (Figure 5). The use of current and novel biological tools provides several clues about the Downstream projections of SF1 neurons regulating glycemia were identified through the same novel strategy derived from optogenetics. The researchers administrated fluorescently tagged ChRs into the VMH of SF1-Cre mice to visualize the projection fields using histological imaging to detect the EYFP reporter in axonal projections [32]. Once identified this target site, a fiber optic was implanted to depolarize the final projection by laser stimulation. With this strategy, it was possible to demonstrate that VMHSF1→aBNST is the most relevant circuit involved in controlling glycemia since the photostimulation of this circuit triggered an increase of the blood glucose levels [32] (Figure 5). Interestingly, other functions have been attributed to this specific circuit: Fan Yang and colleagues recently demonstrated that SF1 neurons are innervated by upstream BNST neurons and send projections to the NST nucleus to regulate anxiety-like behavior and bone metabolism [93] (Figure 5).

neurocircuits regulating blood glucose and energy balance, and the afferent and efferent connections linking SF1 neurons with other neuronal populations. An improving understanding of the functional organization of SF1 neurons may help to identify future strategies for metabolic diseases such as obesity and diabetes. The use of current and novel biological tools provides several clues about the neurocircuits regulating blood glucose and energy balance, and the afferent and efferent connections linking SF1 neurons with other neuronal populations. An improving understanding of the functional organization of SF1 neurons may help to identify future strategies for metabolic diseases such as obesity and diabetes.

In the last years, the critical role played by the hypothalamus in the regulation of energy balance and glucose homeostasis has gained substantial importance. However, the

demonstrating that the origin of these diseases is beyond feeding and pancreatic insulin secretion, and needs further definition. Here we extensively review that SF1 neurons in the hypothalamus provide a "central role" in the control of blood glucose levels, insulin

**6. Concluding Remarks and Future Perspectives** 

#### **6. Concluding Remarks and Future Perspectives**

In the last years, the critical role played by the hypothalamus in the regulation of energy balance and glucose homeostasis has gained substantial importance. However, the exact mechanisms and neuronal circuits underlying this regulation and the pathogenesis of obesity and insulin resistance remain poorly understood. The growing literature is demonstrating that the origin of these diseases is beyond feeding and pancreatic insulin secretion, and needs further definition. Here we extensively review that SF1 neurons in the hypothalamus provide a "central role" in the control of blood glucose levels, insulin sensitivity in peripheral tissues, adipose tissue plasticity, and thermogenesis activation. SF1 neurons are a specific lead in the brain's ability to sense glucose levels and conduct insulin and leptin signaling in energy expenditure and glucose homeostasis, with minor feeding control. Interestingly, transgenic mice lacking different targets in SF1 neurons show an altered metabolic phenotype mostly under HFD feeding, but not under chow diet, indicating that SF1 neurons are particularly important for metabolic adaptation in the early stages of obesity. These investigations also demonstrate the sex-specific effects of SF1 neurons in the VMH, and the importance of the analysis of the phenotype in both male and female mice when exploring energy balance and metabolism in transgenic models.

Although optogenetic and chemogenetic tools have clearly demonstrated SF1 function in glucose homeostasis, adiposity, and energy expenditure, there are still contradictory results in these functions after deletion of specific receptors in SF1 neurons (e.g., mGluR5). These controversies could be associated with the existence of several sub-populations of SF1 neurons that respond differently to insulin, leptin, and glucose. For instance, whereas some SF1 neurons seem to be specialized in the regulation of blood glucose levels, some others are responsible for insulin sensitivity in the periphery.

When investigating the hypothalamic regulation of obesity and diabetes, in addition to the identification of SF1 sub-populations, it is important to explore the coordination of SF1 neurons with other neurons outside the VMH to trigger metabolic functions. For instance, HFD-induced hyperinsulinemia drives SF1 neurons hyperpolarization, leading in turn to functional changes in the synaptic output onto POMC neurons, causing obesity and glucose intolerance. This evidence suggests that the nucleus-specific responses upon HFD feeding can cooperate to cause obesity and diabetes, and disruption of this cooperation could act as a link between both diseases.

Despite the advances in the neurocircuitry connecting SF1 VMH neurons with other brain areas, the exact efferent circuits leading to changes in peripheral tissues need further investigation. How mice lacking a protein only in SF1 neurons present a strong phenotype in the periphery in terms of glucose homeostasis and energy expenditure under nutritional challenges? Are these circuits restricted to SNS and PSNS? A new point of view also emerges from these studies: a disruption in a specific type of neurons in the hypothalamus could trigger the pathogenesis of type 2 diabetes and obesity. How this neuronal dysfunction drives peripheral insulin resistance, adiposity, and thermogenesis alteration remains unclear. Reconsidering the pathogenesis of obesity and type 2 diabetes and now including the essential role of SF1 neurons, new strategies to treat these diseases could emerge. Targeting this specific population of neurons in the VMH, and even another cooperating population outside the VMH at the same time could modulate specific metabolic functions, a challenging approach but with significant translational impacts in the management of obesity and diabetes.

**Author Contributions:** All authors (A.F., S.Z., N.C. and R.R.-R.) have participated in writing and editing. All authors (A.F., S.Z., N.C. and R.R.-R.) have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Spanish Ministerio de Economía, Industria y Competitividad (MINECO), Agencia Estatal de Investigación (AEI), and Fondo Europeo de Desarrollo Regional (FEDER) (Grants SAF2017-82813-C3-3-R to N.C. and R.R.-R. and PCI2018-092997/AEI to R.R.-R.). A.F. is granted with a fellowship from the Agència de Gestió d'Ajuts Universitaris i de la Recerca (AGAUR) in Catalonia.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

