Microglia in Central Control of Metabolism

Abstract

Beyond their role as brain immune cells, microglia act as metabolic sensors in response to changes in nutrient availability, thus playing a role in energy homeostasis. This review highlights the evidence and challenges of studying the role of microglia in metabolism regulation.

Introduction

In the central nervous system (CNS), the hypothalamus plays a key role in sensing and responding to peripheral metabolic fluctuation to maintain whole body energy homeostasis. Much evidence has shown the crucial role of microglia in this energy homeostasis process (1–5). Microglia exert immune and metabolic functions by serving as brain-resident immune cells with natural macrophage characteristics. Unlike peripheral macrophages, microglia originate from the yolk sac (6) and have a slower turnover rate. Although microglia activation is necessary for neuronal homeostasis, prolonged microglia activation has been linked to neurodegeneration and obesity. Upon high-fat diet (HFD) exposure, microglia activation is predominantly observed in the arcuate nucleus (ARC) of the hypothalamus, the area in which first-order neurons sensing peripheral metabolic signals are located (7, 8). Several studies have implied the causative role of hypothalamic microglia activation in the development of obesity. Here, we review and discuss the current understanding on the role of microglia in metabolism, as well as the limitations in microglia research. Furthermore, we discuss potential approaches and directions that might help advance the understanding of the metabolic role of microglia in regulating energy homeostasis.

Microglia, Central Players in Hypothalamic Function

Microglial cells, as the resident macrophages of the CNS, play an essential role in responding to injuries and infections. Additionally, and somehow similar to peripheral macrophages (e.g., adipose tissue-resident macrophages), microglial cells can sense and be activated by dietary molecules such as lipids (9, 10) and carbohydrates (11–13) as well as hormones such as leptin (12, 13), ghrelin (14), adiponectin (APN) (15–17), and glucagon-like peptide 1 (GLP-1) (18–20). Indeed, microglia cells express transporters for fuel substrates such as glucose, fatty acids, and amino acids (21) and receptors for hormones (12–20).

The traditional characterization of microglial M1 and M2 polarization status, a process by which microglial cells produce distinct functional phenotypes as a reaction to specific stimuli and signals, does not fully capture the complex and diverse function of microglia. Recent advances in single-cell RNA sequencing (scRNA-seq) and other high-throughput technologies have implicated the heterogeneous and diverse activation states of microglia (22, 23). Microglia activation status might be dictated by factors including the local microenvironment, the type and severity of the injury, and the dynamic interplay with other brain cells. As a result, microglia might acquire distinct molecular signatures and functional properties depending on their own unique anatomical environment (21, 24).

Although microglia are present across the entire CNS, their presence in the hypothalamus has been shown to be essential in regulating numerous aspects of whole body metabolism. In particular, hypercaloric diets are known to induce profound changes in microglia within the hypothalamic ARC. Because of its anatomical position in close proximity with the median eminence, which lacks the blood-brain barrier (BBB), the ARC has a central role in sensing nutrient changes and regulating metabolism. Here, microglia act as the first immune and metabolic-sensing population among other microglia in the brain parenchyma. In support of this, a study found that microglia in the circumventricular organs (CVOs) are in active status under normal physiological conditions (25). In the ARC, two essential neuronal populations regulating metabolic responses to nutrients are located. Here, Neuropeptide Y and Agouti-related protein (NPY/AgRP)-expressing neurons exert an orexigenic function, i.e., increase food intake, whereas proopiomelanocortin (POMC) neurons are known to be anorexigenic, i.e., decrease food intake. Not surprisingly, studies have observed a very active and complex interaction between these neurons and microglia in the ARC (see below for details).

Effects of Microglia Manipulation on Whole Body Metabolism

Genetic Models of Microglia Signaling Alterations

CX3CR1, also known as the fractalkine receptor, is a G protein-coupled receptor responsible for leukocyte migration, retention, and adhesion. It is expressed on monocytes, dendritic cells, and microglia. CX3CL1/fractalkine, the ligand of CX3CR1, is expressed and released by neurons (26). An early study on CX3CR1 signaling reported that although Cx3cr1-deficient mice had no changes in body weight, food intake, and energy expenditure on both chow diet and HFD, they showed improved glucose tolerance and insulin sensitivity on HFD compared with HFD-fed control mice (27). Despite this first evidence of possible beneficial effects of Cx3cr1 deficiency, several studies have demonstrated the deleterious outcomes caused by the loss of CX3CR1 signaling (28–32). A recent study showed that Cx3cr1-deficient mice have increased food intake compared with wild-type mice when fed on a standard chow diet (33). Furthermore, Cx3cr1-deficient mice have impaired glucose tolerance and insulin resistance and increased hepatic steatosis when fed on a HFD. Similarly, leptin-deficient (ob/ob) mice with Cx3cr1 deletion (Cx3cr1KO/ob) display increased insulin resistance, hepatic steatosis, and accumulation of macrophages in adipose tissue compared to ob/ob mice with intact CX3CR1 signaling. Interestingly, however, Cx3cr1KO/ob mice undergo less weight gain than ob/ob mice up to 8 wk of age, whereas no differences were observed during the subsequent 12 wk (34). In addition, female male-like Cx3cr1-knockout mice fed on HFD displayed a malelike phenotype with a higher body weight gain (35) compared with female diet-induced obesity (DIO) control mice. Indeed, the lower susceptibility to DIO shown by female versus male mice could be attributed, at least in part, to higher levels of CX3CR1-CX3CL1 observed in female mice exposed to HFD.

Cx3cr1-deficient mice have also shown increased levels of proinflammatory cytokines, such as IL-6 and TNFα, and the deposition of extracellular matrix (EM) in the dentate gyrus (DG). An anxiolytic-like and depressive-like phenotype is exhibited by Cx3cr1-deficient mice (35), as well as a significant reduction in oligodendrocyte progenitor cell (OPC) engulfment, an increased number of oligodendrocytes, and a reduced myelin thickness (36). Studies have shown that Cx3cr1 deficiency in mice can cause an exaggerated response to lipopolysaccharide (LPS) stimuli in CNS, with consequent microglial neurotoxicity and advanced neuronal death (37). Thus, these data suggest that microglial CX3CR1 signaling may play a regulatory role by preventing overactivation via CX3CL1 released by neurons. Therefore, microglia lacking CX3CR1 signaling by maintaining an activated status may alter neuronal activity in response to changes in metabolic signals.

CX3CR1 has been also shown to be involved in glucose homeostasis. In a hypoglycemic condition model, fasted Cx3cr1-deficient mice treated with insulin developed insulin resistance. Furthermore, when injected with the glycolytic inhibitor 2-deoxy-d-glucose (2-DG) in fed condition, Cx3cr1-deficient mice displayed increased counterregulatory responses, showing higher glucose levels compared with control mice, thus suggesting a suppressive role of microglial CX3CR1 signaling in counterregulatory responses to hypoglycemia (38).

In addition to Cx3cr1 deficiency, the actions of its ligand, CX3CL1, in metabolism regulation have also been assessed. Intracerebroventricular (icv) CX3CL1 administration suppressed food intake both physiologically and after stimulation of food intake by NPY or ghrelin (33, 39). Consistently, overexpression of Cx3cl1 in the medial basal hypothalamus reduced body weight gain and food intake, with improved leptin sensitivity in a DIO model (39, 40). Overall, CX3CR1-CX3CL1 signaling plays a role in metabolism regulation by reducing the inflammatory activity of microglia. Whether most neurons can produce CX3CL1 at a similar level or certain neuronal populations are responsible for most of CX3CL1 production is not fully understood. In this context, the effects of CX3CL1 on CX3CR1-expressing microglia could be taken as an advantage to study microglia subpopulations. Cx3cl1 deletion in specific neuronal populations, which mostly affects neighboring microglia activity, may be a potential approach for regional microglia studies. It is important to consider that CX3CR1 is not only expressed in microglia but also highly expressed in peripheral monocytes. Consequently, Cx3cr1 deletion might have complex effects in both the central and peripheral systems. Thus, to define the physiological role of CX3CR1 in microglia, a conditional knockout approach is necessary to exclude effects from the loss of CX3CR1 signaling in the periphery.

To define microglia functions, the tamoxifen-inducible Cx3cr1^CreER system has been widely used to alter genetic expression specifically in microglial cells (Table 1) (41–44). Indeed, although Cx3Cr1 is expressed in both microglia and peripheral macrophages, the latter has a considerably faster turnover. Thus, peripheral macrophages carrying the modification upon tamoxifen injection will be quickly replenished without any long-lasting Cre-dependent effects, whereas microglia cells expressing Cre recombinase enzyme will persist, thus causing microglia-specific genetic modifications. By application of this Cx3cr1^CreER model, IKKβ and A20, essential components of the NF-κB pathway, were deleted specifically in microglia to assess the effects of inflammatory response on obesity development. Mice lacking A20, a negative NF-κB regulator, showed spontaneous microglia activation, resulting in an obese phenotype in mice fed on a standard chow diet (7). Inversely, HFD-fed mice with specific microglial deletion of IKKβ, an essential cofactor for NF-κB activation, maintained less microglia activation, had reduced body weight and food intake, and were overall protected against DIO. Similarly, microglial deletion of uncoupling protein 2 (Ucp2), a mitochondrial protein regulated by fatty acid metabolism and shown to play a role as transporter in the malate-aspartate shuttle (45–47), resulted in milder inflammatory response to HFD (8). This change in the inflammatory response was due to attenuated mitochondria function in microglia during HFD, as Ucp2-deleted primary microglia were unable to utilize fatty acids. As a result, microglia Ucp2-deficient mice were protected from DIO (8). In support of a role for lipid metabolism in the regulation of microglial function (8), microglial lipoprotein lipase (LPL)-deficient mice display increased dysmorphic mitochondria and impaired systemic lipid homeostasis with increased body weight and POMC neuronal loss (48). Altogether these studies indicate that reducing microglia inflammatory response to HFD has beneficial effects conferring protection to DIO (8).

Table 1.

Microglia-specific gene-deleted animal models using Cx3cr1CreERT system

Gene	Diet	Microglia Activity	Neuronal Activity	Metabolic Phenotype	Other Features	Reference
IKKβ	42% fat/ 60% fat	Reduced inflammatory response	NA	Resistant to DIO	Reduced myeloid cell recruitment into the MBH	(7)
A20	CD	Enhanced inflammatory response	Reduced neuronal leptin sensitivity	Obese phenotype	Increased myeloid cell recruitment into the MBH	(7)
Ucp2	45% fat	Reduced inflammatory response	Preserved leptin sensitivity of POMC neurons	Resistant to DIO	Defective fatty acid utilization in microglia	(8)
EP4	60% fat	Reduced phagocytotic activity; no changes in inflammatory response	Preserved POMC neuronal projections to the PVN	Resistant to DIO	Reduced cellular contacts between POMC neurons and microglia	(49)
Bmal1	60% fat	Increased phagocytic activity	Increased number of POMC neurons	Resistant to DIO	Improved long-term memory	(50)
LPL	58% fat	Reduced phagocytolic activity; increased mitochondrial dysmorphologies	Reduced number of POMC neurons; increased mitochondrial dysmorphologies	Susceptible to DIO	Dysmorphology of the Golgi apparatus in microglia	(48)

ARC, arcuate nucleus; CD, chow diet; DIO, diet-induced obesity; MBH, mediobasal hypothalamus; NA, not assessed; POMC, proopiomelanocortin; PVN, paraventricular nucleus of the hypothalamus.

Likewise, attenuation of microglia phagocytotic activity has been shown to have similar benefits in reducing susceptibility to DIO. Microglial EP4 (prostaglandin PGE₂ receptor)-deficient mice displayed reduced phagocytotic activity with no changes in proinflammatory cytokines, such as TNFα and IL-1β, during HFD. Nevertheless, microglial EP4 deletion markedly reduced weight gain and food intake and prevented HFD-induced obesity with enhanced insulin sensitivity (49). This resistance to DIO was associated with increased POMC and α-melanocortin stimulating hormone (α-MSH) projections onto hypothalamic paraventricular nucleus (PVN) neurons compared with control mice on HFD. However, and contrary to the consistent protective effects from DIO shown by reducing microglial inflammatory responses (7, 39, 40), the effects of attenuated phagocytotic activity in microglia on metabolism have not been consistent. For instance, mice with deletion of microglial Bmal1 (Basic Helix-Loop-Helix ARNT Like 1), a key clock gene, have increased microglia phagocytotic activity in HFD but reduced energy intake and attenuated body weight gain with increased POMC neuronal numbers compared with HFD control mice (50). Furthermore, microglia Bmal1-deficient mice displayed improved long-term memory compared with control mice. Interestingly, however, deficiency of phagocytosis-associated genes has demonstrated beneficial effects on neurodegeneration (51), thus suggesting that microglial phagocytotic activity may have a distinct role across brain regions and under different pathological conditions. Future studies are warranted to determine the role of microglial phagocytosis in brain functions.

Finally, although the loxP-inducible CreER system is a well-established genetic tool, “leaky Cre” activity, i.e., recombination occurring without tamoxifen induction, has been reported (52, 53). Because of that, Cre might accumulate in the nucleus of the long-living microglia and spontaneously induce recombination (53). Nevertheless, the inducible Cx3Cr1^CreER system is the most used animal model for studying microglia, whereas additional models such as Sall1CreER, P2ry12CreER, HexbCreER, and Tmem119CreER mice have been recently developed and future studies will address the utility of these models for microglia study (54–57).

Pharmacological Models of Microglia Depletion

Microglia depletion through pharmacological means has been applied as a useful approach for understanding microglia function in the brain. Microglia development, proliferation, and survival require numerous molecular factors (58, 59). A main molecule for microglia survival is colony-stimulating factor 1 receptor (CSF1R), which is highly expressed in myeloid lineage cells, including monocytes and macrophages in the periphery, and microglia in the brain (60–62). Csf1r-depleted mice lack microglia (52), similar to mice depleted of either of its two ligands, CSF1 (63, 64) and IL-34 (65, 66). This indicates that CSF1R signaling is essential for microglia survival. Interestingly, it was observed that intraperitoneal (ip) administration of CSF1- or IL-34-blocking antibody disrupts CSF1R signaling differentially, causing depletion of microglia in gray and white matter, respectively. This might be due to the differential expression pattern (67) of CSF1 and IL-34, since CSF1 is expressed by astrocytes, oligodendrocytes, and microglia whereas IL-34 is predominantly expressed by neurons (68, 69).

Since several studies have shown the need for CSF1R signaling for microglia survival, pharmacological CSF1R targeting approaches have been applied to deplete microglia populations to address the physiological role of microglia. PLX3397, also known as pexidartinib, was approved in 2019 by the US Food and Drug Administration for the treatment of symptomatic tenosynovial giant cell tumor (TGCT). PLX3397 is a multikinase inhibitor drug (70). Similarly, another CSF1R inhibitor, PLX5622, has been shown to have more specific CSF1R inhibitory effects, without affecting c-Kit, an essential factor for hematopoietic cell survival and development (71, 72). Both of these inhibitors can permeate through the BBB and are available as dietary formulation for chronic administration, making them particularly suitable for animal studies. PLX3397-induced microglia depletion in a chronic demyelination model has been shown to ameliorate the disease progression by increasing remyelinated oligodendrocytes (73), whereas it increased apical dendritic spine density with reduction of neuronal death after traumatic brain injury (TBI) (74). Similarly, PLX5622-driven microglia repopulation in the aged brain restores the expression of genes associated with neuronal and synaptic functions to levels comparable to those of young mice, with improvement of cognitive functions (75). Chronic administration of PLX5622 maintained a small number of microglia (∼30%) and improved learning and memory in an aged 3xTg-AD mouse model of Alzheimer’s disease (AD) (76). After brain injury, microglia repopulated by PLX5622 treatment promoted brain recovery, reduced lesion-induced inflammation (77), and recovered motor and cognitive deficits in an IL-6-dependent manner (78, 79). In the PLX5622-treated 5XFAD AD mouse model, repopulated microglia displayed homeostatic genetic signatures and this was associated with a slower progression of the disease (80, 81). The beneficial effects of inhibitors may be due to the removal of detrimental microglia and the repopulation of healthy microglia. Indeed, because of the self-renewing ability of microglia, withdrawal of inhibitors after microglia depletion causes microglia to repopulate promptly and reach full repopulation within 3 wk. Consistently, an early study showed that healthy adult mice had inflammatory gene profiles similar to the original resident microglia after withdrawal of PLX3397-containing chow diet (82). Thus, microglia depletion and following repopulation might achieve a replacement with healthy microglia and the reestablishment of homeostatic condition even in pathological settings.

Unfortunately, although several studies have been performed assessing the effects of pharmacological microglial depletion on neurodegenerative diseases, limited studies have addressed the effects of microglia depletion and repopulation after inhibitor treatment on whole body metabolism. For example, Elmore and collaborators (75) found that 2 wk after microglia depletion by PLX5622 adult mice with fully repopulated microglia show increased body weight. However, when embryonic microglia depletion was achieved by administering PLX5622 to pregnant dams starting at embryonic day (E)3.5, the body weight of PLX5622-exposed male pups was lower than that of control pups up to postnatal day (P)15, with reduced number of POMC neurons. However, it significantly increased from P15 to P21, and at 6 wk of age PLX5622-exposed male mice did not show differences in body weight, although female mice were heavier than control mice (83). Thus, these results suggest that the changes in body weight observed in the early phase after microglia depletion could be due to the severe environmental changes induced by microglia cell death rather than microglia absence, as no body weight changes were observed by long-term monitoring of adult mice fed on chow diet (7). However, when exposed to HFD, PLX5622-treated adult mice showed reduced body weight and food intake with enhanced leptin sensitivity (7, 10).

Despite limited evidence of the effects of microglia depletion and repopulation, many studies have proposed the beneficial effects of repopulating microglia as a therapeutic strategy to treat pathological brain conditions (72, 84). The homeostatic signature gained on repopulated microglia after depletion might have a beneficial effect on dysregulated metabolic conditions, although its long-term efficiency has not been validated. To define the potential beneficial effects of repopulating microglia on metabolism, future studies assessing the effect of microglia depletion at different time points in obese or diabetic mice are warranted. Furthermore, gene expression profiling between disease-associated activated microglia and repopulated microglia at different time points is also needed to determine whether repopulated microglia remain healthy and whether, with time, they may become detrimental, worsening the pathological conditions.

Genetic Models of Microglial Depletion

Besides pharmacological methods, genetic approaches have also been used for microglia depletion. For microglia-specific depletion purposes, tamoxifen-dependent inducible Cx3cr1^CreER mice have been crossed with cre-dependent diphtheria toxin (DT) receptor (DTR)-expressing transgenic mice. Upon DT administration, significant microglia depletion has been shown to occur after just 1 day and full repopulation after 2 wk, seemingly faster than the inhibitor treatment (41, 43). Tamoxifen administration resulted in microglia-specific DTR expression, as the short-lived peripheral monocytes were replenished by bone marrow-derived nonmutated progenitors (42). An early study showed that DT-mediated microglia depletion had no effect on body weight and cytokine expression in young mice (P30) after 1 wk of microglia deletion and did not alter the number of peripheral monocytes and peripheral tissue-resident macrophages (43). However, microglia depletion in young mice (at P19, P30, and P60) altered neuronal spine remodeling, leading to deficits in learning and memory (43).

In the Cx3cr1-DTR rat model, DT-mediated microglia depletion resulted in reduced body weight, locomotive activity, and energy expenditure accompanied by hypophagia. These metabolic changes did not correlate with the changes in hormone signals and neuronal activation, since microglia-depleted Cx3cr1-DTR rats had increased ghrelin, reduced leptin, enhanced NPY/AgRP neuronal projection onto the PVN, and a decreased number of POMC neurons (85). These contradictory results could be attributed to microglial cell death. In support of this, microglia-depleted Cx3cr1-DTR rats display febrile symptoms (86), and some studies have shown that DT-mediated microglia depletion induces astrogliosis and a cytokine storm (87), as accompanied by altered circadian rhythm-related gene expression (86).

Some studies, however, investigated the effects of long-term microglia loss without repopulation and observed changes in body weight and food intake. Although microglia-driven factors were missing, altered neuronal activity was observed. Further evaluation of the activity of specific neuronal populations, especially in the hypothalamus, is needed to clarify the effects of chronic microglia depletion.

Although microglia depletion is a promising approach to define its role in metabolism regulation, it lacks regional specificity. Many transcriptome studies have established the concept of microglia heterogeneity. However, regional depletion might be a challenging approach for microglia study. A recent study induced striatum-specific microglia depletion of ∼50%, by using Drd1aCre and Drd2Cre mice with deleted Il-34 in either D1 or D2 neurons (88). On the other hand, inducible ablation of Csf1 in NestinCre mice resulted in the loss of microglia predominantly in the white matter of the striatum (88). According to this study, neuron-specific Cre mice can be used for local microglia depletion. For instance, Il-34 deficiency in POMC, AgRP, or Sf1 neurons could be used to alter local microglia and thus to study their role in energy and glucose homeostasis. Similarly, Cx3cl1 could also be considered as a target to attenuate microglia activity. Additionally, direct and regional gene manipulation on microglia could be also achieved by the adeno-associated virus (AAV) strategy. Some reports have shown the success of AAV-mediated gene delivery into microglia (89, 90). However, AAV-mediated gene delivery potentially induces damage in the brain parenchyma, thus affecting microglia activation. Therefore, more studies are needed to determine whether the AAV strategy is a safe and appropriate approach to study microglia.

Hypercaloric Diet-Induced Microglia Activation

Hypothalamic microglia quickly respond and get activated after HFD consumption (FIGURE 1). Previous studies have shown that chronic hypothalamic microglia activation is associated with obesity development in both mice and humans (1–4, 10). In rodents, HFD drives microglia morphological changes into amoeboid shapes with enlarged cell bodies and thickened processes indicating microglia physiological switch from rest to activated state (5). Similar to what is observed in mouse models, in humans evidence has also suggested a correlation between obesity and hypothalamic microglia activation (3, 4). Microglial activation is observed in many areas of the brain in response to excessive dietary fats (91, 92) but predominantly in the ARC area (7, 8). HFD-induced microglial alteration is primarily a consequence of HFD consumption itself rather than the consequence of the obese phenotype. Indeed, genetically obese mice, such as mice with either leptin, leptin receptor, or melanocortin 4 receptor (MC4R) deficiency, fed on a standard chow diet do not present hypothalamic microglia activation (1). In addition, specific types of fatty acids seem to induce greater microglia activation than others, as suggested by the observation that when dietary saturated fatty acids (SFAs) are replaced by similar amounts of unsaturated fatty acids (UFAs) hypothalamic inflammation does not occur (88). In addition, Gao and collaborators (11) have shown that low-carbohydrate high-fat (LCHF) diet failed to induce microglia activation while inducing activation of astrocytes. Notably, LCHF-fed mice showed lower body weight gain compared with mice fed high-carbohydrate high-fat (HCHF) diet (11). Thus, this observation suggests that both carbohydrates and fat are necessary for microglia activation.

FIGURE 1.

Hypothalamic microglia activation during high-fat diet (HFD) Microglia response to changes in dietary molecules and hormones [leptin, glucagon-like peptide 1 (GLP-1), adiponectin, ghrelin]. The increase of fatty acids in HFD feeding switches microglia fuel utilization toward fatty acids and enhances inflammatory responses in the hypothalamus. Cytokines released from microglia can affect neuronal activity (TNFa, IL-1b) as well as astrocytic activity (IL-1a, TNFa, C1q). Activated neurons, astrocytes, and microglia recruit CD169+ and CCR2+ peripheral myeloid cells in the hypothalamus via release of chemokies such as CX3CL1 and CCL2. AgRP, Agouti-related protein; POMC, proopiomelanocortin.

At early time points from the start of HFD administration, microglia activation is accompanied by changes in cytokine expression and release. Cytokines themselves are known modulators of food intake and neuronal activation. Intracerebroventricular administration of cytokines such as TNFα, IL-1β, and IL-6 suppress food intake in rodents (93–97). In contrast, central administration of IL-4, known as an anti-inflammatory cytokine, promotes hypothalamic inflammation and increases weight gain during high-fat feeding (98). Suppressive effects on food intake by proinflammatory cytokines such as TNFα and IL-1β raise the question of whether the early inflammatory response to HFD is protective or causative for obesity development. Microglia inflammatory responses have been shown to occur very quickly, beginning as early as 1 day after HFD exposure. Within the first day of HFD feeding, the level of TNFα, IL-1β, and IL-6 mRNA significantly increased at 3 h and returned to baseline level after 6 h of HFD feeding (99). These cytokine levels were increased again at 4 wk of HFD (3), whereas at 8 wk of HFD feeding, when mice show significantly greater body weight, fat mass, and hormone resistance, the levels of TNFα, IL-1β, and IL-6 mRNA were again comparable to those observed in mice fed on a standard chow diet (8). Collectively, a strong inflammatory response has not necessarily been seen along with obesity development. With this view, the inflammatory activity in the early time points of HFD exposure may be neuroprotective rather than detrimental. In addition, the similar levels of cytokines between HFD and standard chow diet observed at 8 wk of HFD feeding suggest that the reduction in cytokine levels during HFD may cause the metabolic phenotype. Long-term HFD may reduce the expression of microglia-specific sensing genes, including P2ry12, Selplg, Slc2a5, and Trem2, and consequent reduction of CD68, TMEM119, and P2RY12 protein levels in the ARC microglia (4, 7). Interestingly, microglia in chronic HFD-exposed mice can react normally to LPS stimulation (7). Furthermore, long-term HFD-induced ARC microglial activation can also be reversed by switching to a standard chow diet (100), suggesting that HFD-induced hypothalamic microglia activation may differ from microglia activation observed in pathological conditions, such as Alzheimer’s disease, where its activation seems to be irreversible.

After a few weeks of HFD, CD169+ and CCR2+ peripheral myeloid cells are recruited to the mediobasal hypothalamus (MBH; FIGURE 1), possibly through CX3CL1/CX3CR1 and CCL2/CCR2 axes, and contribute to obesity susceptibility (7, 101–103). HFD feeding induces BBB leakage in the hypothalamic area (102), in which infiltrated macrophages can further disturb the BBB integrity and increase lipid flux. These lipids, in turn, are taken up by microglial cells for clearance. The lipid homeostasis mediated by microglial cells has been shown to be important for whole body energy homeostasis. Low-density lipoprotein receptor (LDLR)-deficient mice showed enhanced hypothalamic microglia activation on HFD compared with control mice (104) due to impaired LDL clearance and disturbed cholesterol homeostasis, indicating that microglial lipid metabolism is critical to maintain healthy metabolic condition. HFD feeding forces microglia to use lipids rather than glucose or glutamine (105). Interestingly, limiting fatty acid utilization by microglia is beneficial for DIO (FIGURE 1). For instance, primary Ucp2-deleted microglia cells lose preference for palmitate. As a result, microglia Ucp2-deficient mice displayed mild inflammatory response to HFD and resistance to DIO (8), thus suggesting that altering lipid metabolism in microglia may represent a potential target for obesity prevention.

HFD not only affects fuel preference but can also disturb the circadian rhythm regulating microglia activation (105). Under normal physiological conditions, microglia activation can be observed during the night period, when rodents show active feeding behavior (106). Although microglia-specific clock gene deletion has been shown to disturb their circadian rhythm but have no effect on microglia morphology, microglia Bmal1-deficient mice showed protection from DIO (50). Interestingly, whole body depletion of Rev-erbα, a Bmal1 repressor, caused spontaneous microglia activation during the day period (107), indicating that microglia activation during the night period seems dependent on circadian rhythm.

The observed upregulation of cathepsin S (CatS) and P2RY12, two proteins known to increase with microglia activation at the beginning of the dark period, suggests a possible mechanism for changes in microglia activation during HFD. Cortical microglia can be activated by adenosine triphosphate (ATP) released from active cortical neurons via P2RY12 and secrete CatS to the synaptic sites of neighboring neurons. CatS degrades surrounding extracellular matrixes (ECMs), including the perineuronal nets (PNNs) that regulate the maturation and elimination of neuronal dendritic spines (108), which has been shown to restrict neuronal plasticity in the ARC. Indeed, a previous report found that the ARC is surrounded by PNN-like structures that are critical for AgRP neuronal maturation (109). As observed in the prefrontal cortex, HFD reduced the PNN intensity (110), an event that could also occur in the ARC. Therefore, on HFD microglia might influence ARC neuronal activity through PNN alteration via CatS.

Microglia Communication in Metabolism Regulation

Microglia and Hormones

Several metabolic hormones have been shown to affect microglia activation (FIGURE 1) (4–12).

Leptin, a cytokine secreted by adipocytes, signals to the brain to regulate food intake. Chronic HFD feeding induces hyperleptinemia due to increased fat mass. In vitro studies performed on primary hypothalamic microglia have shown that leptin treatment has proinflammatory effects by increasing TNFα and IL-1β expression (111), possibly as a consequence of JNK-STAT3 pathway activation. However, leptin treatment did not affect TNFα-induced cytokine levels in hypothalamic slice explants (112). In vivo, mice with selective deletion of leptin receptor in myeloid cells displayed increased body weight and increased food intake without alteration of glucose homeostasis when fed on a standard chow diet (13). Interestingly, these mice showed dystrophic microglial morphology with reduced branches in the PVN but not in the ARC. Furthermore, CD68 expression was also reduced in the PVN microglia. Within the ARC, although no changes in microglia morphology and activation could be observed, the number of POMC neurons and α-MSH levels were reduced, with no changes observed in the AgRP levels (13). Overall, whereas in vitro leptin acts as a proinflammatory signal in microglia, in vivo leptin is required for microglia homeostasis and plays a role in formation of the hypothalamic circuits regulating metabolism.

In addition to leptin, the gut hormone ghrelin also modulates microglia inflammation. Although it exerts an orexigenic activity, ghrelin acts as an anti-inflammatory hormone. Evidence has shown that 1 day of HFD feeding is sufficient to induce an increased number of macrophages/microglia in the nodose ganglion and in the hypothalamus. Interestingly, subcutaneous administration of ghrelin before HFD feeding can prevent this increase in microglia, indicating that ghrelin has an anti-inflammatory role (14).

Glucagon-like peptide 1 (GLP-1) is a hormone secreted by intestinal enteroendocrine L cells (113). In the brain, GLP-1-producing neurons are present in the nucleus of the solitary tract (NTS) (114). GLP-1 receptor is expressed on microglia as well as in neurons (18, 19). In BV-2 microglia cells, GLP-1 directly suppresses the expression of proinflammatory cytokines (i.e., TNFα, IL-6, and IL-1β) (19). In line with this, a GLP-1 agonist (liraglutide) decreased the microgliosis in the MBH of HFD-fed mice (20). However, although these data suggest a role for GLP-1 as an anti-inflammatory molecule in microglia, a microglia GLP-1R-deficient animal model has not yet been evaluated.

APN, an adipokine (115), enhances insulin sensitivity through the activation of AMP-activated protein kinase (AMPK) in peripheral tissues (116). Additionally, APN crosses the BBB and acts through two APN receptors (ADIPOR1 and ADIPOR2) expressed in neurons, microglia, and astrocytes (117). Systemic APN administration suppresses microglia activation induced by HFD, whereas it has no effects on body weight gain (15). On the other hand, central APN administration has been shown to decrease body weight by enhancing energy expenditure and increasing the number of cFos-positive neurons in the PVN (16). APN-deficient mice were protected from DIO with decreased food intake and increased energy expenditure. This antiobesity phenotype might be due to a reduction in NPY and an increase in POMC levels in APN-deficient mice (17). Although much evidence points at a direct effect of APN on microglia, microglia-specific receptor-deficient mouse models have not been studied so far.

Microglia-Neuron Communication

Extensive evidence has shown that a prolonged hypercaloric diet induces chronic microglia activation leading to hypothalamic neuronal dysfunction, including loss of synapses (118), impaired responsiveness to hormones (119–121), and altered intracellular organelle function (8, 122–124), causing obesity and diabetes development. One of the critical factors regulating microglia-neuron interaction in this context is represented by the release of cytokines. Microglia activation induces the upregulation of pathways related to cytokine signaling in hypothalamic neurons (119, 125–129). Multiple studies have evaluated the direct effect of cytokines on neuronal activity through electrophysiological approaches. In vitro incubation of hypothalamic brain slice of POMCeGFP mice with TNF-α led to an increased POMC neuronal firing rate, increased intracellular ATP production, and mitochondrial fusion processes (130). In agreement with this, Chaves and collaborators (131) showed that both TNFα and IL-1β acutely inhibited neuronal activity of 35–42% of AgRP neurons, whereas ∼11% of POMC neurons were activated by TNFα, thus supporting the suppressive effect of TNFα and IL-1β on food intake through the direct modulation of ARC neuronal activity. However, this study also showed that IL-6 had no effect on AgRP or POMC neuronal activity (131), data that are in disagreement with a previous report showing the antiobesity effects of IL-6 by increasing energy expenditure and that Il-6-deficient mice developed mature-onset obesity (132). Consistent with these data, central Il-6Rα-deficient mice showed increased food intake and reduced energy expenditure (133). Future experiments studying microglia-specific cytokine-deficient mice are needed to clarify these contradictory data.

In addition to cytokines, altered phagocytotic activity can lead to impaired neuronal activity (134, 135) by altering synaptic plasticity (118). HFD has been shown to alter the synaptic reorganization of hypothalamic POMC and AgRP neurons (118), and these synaptic changes are associated with increased astrocytic activity. Although microglia-mediated effects on synaptic reorganization have not been directly investigated, indirect evidence supports the role of microglia on synaptic reorganization. For instance, altered phagocytotic activity has been reported in animal models of obesity (49, 50), and the number of POMC neurons in direct contact with microglia is increased in DIO mice (130). In addition, our laboratory has also reported that in a genetic mouse model in which HFD-induced microglial activation was attenuated astrogliosis was also reduced, resulting in a significant increase in POMC neuronal activation and leptin sensitivity that were associated with a change in the synaptic input organization onto POMC neurons (8). Whether the synaptic reorganization was due to the direct effect of microglia on neurons or through their indirect effect via astrocytes needs to be assessed.

Finally, neuron-microglia communication may also occur through neurotransmitters such as glutamate. For example, when neurons release glutamate, the brain primary excitatory neurotransmitter, astrocytes uptake glutamate from the extracellular space via GLT-1 and convert it to glutamine before releasing it back into the extracellular space for neuronal uptake. However, glutamate release can also induce microglia activation and induce release of TNFα (136, 137). In turn, TNFα stimulates microglia to release glutamate, suggesting that neurotransmitter-mediated cross talk between neurons and microglia may be a regulatory module controlling their reciprocal activity (138).

Microglia-Astrocyte Communication

Microglia and astrocytes are known to communicate between themselves as well as with neurons to maintain brain homeostasis. Since the bidirectional communication between astrocytes and neurons was initially emphasized in the concept of the tripartite synapse (139), the multipartite synapse including microglia, astrocytes, and neurons was further presented as an integral element for fine-tuned communication (140). This intercellular communication occurs through the release of a variety of molecules, including cytokines, chemokines, ATP, and growth factors (FIGURE 1).

Together with tanycytes, specialized ependymal cells that line the third ventricle and allow the transfer of signals from the cerebrospinal fluid to the CNS, hypothalamic astrocytes have been reported to play a critical role in regulating nutrient and hormone sensing within the CNS. Astrocytes can sense and transport glucose via gap junctions, the astroglial glucose transporter (GLUT)-1, and insulin receptors (IRs) (141, 142), thus suggesting a role for these glial cells in regulating systemic glucose metabolism. Indeed, astrocyte-specific IR depletion has been shown to reduce glucose-driven POMC neuronal activation. This indicates that insulin signaling in hypothalamic astrocytes requires glucose-induced activation of POMC neurons (142). Likewise, astrocyte-specific leptin receptor diminished inhibitory effects of leptin on food intake (143). Emerging evidence indicates that glial cells including microglia, astrocytes, and tanycytes integrate peripheral information and support neurons in response to systemic metabolic states. Therefore, microglia-astrocyte cross talk has been suggested to play a role in regulating whole body energy homeostasis.

As discussed above, activated microglia secrete multiple proinflammatory molecules, such as TNF-α and IL-1α, which can drive secondary inflammatory responses from astrocytes (144). Additionally, other factors originating from microglia such as nitric oxide (NO), type I IFN, C1q, and VEGF-B can also modulate astrocyte proliferation and differentiation (145). Likewise, astrocyte-driven molecules can enhance or inhibit microglial activity. Astrocytic transforming growth factor (TGF)-β, orosomucoid-2 (ORM2), and glial cell line-derived neurotrophic factor (GDNF) all exert an anti-inflammatory effect by downregulating microglial activation (146–148). On the other hand, astrocytic LCN2 enhances microglial activity under inflammatory and pathological conditions (149). Finally, astrocyte-derived chemokines such as CCL2 and CXCL10 recruit activated microglia and promote microglial phagocytosis (150, 151).

Activated microglia induce astrocytosis. However, whether these events have a detrimental or beneficial effect seems to be determined by the inflammatory conditions. In this regard, Liddelow and collaborators (144) performed genomics analysis of reactive astrocytes induced by ischemic stroke and LPS stimulation. They showed that reactive astrocytes induced by LPS administration fail to promote neuronal survival outgrowth, synaptogenesis, and phagocytosis. Instead, they induce the death of neurons and oligodendrocytes, whereas reactive astrocytes in ischemia displayed a molecularly protective phenotype. Thus, two states of activated astrocytes have been defined, A1 being detrimental and A2 being protective. Interestingly, LPS failed to induce A1 reactive astrocytes in Csf1r^−/− mice (which lack microglia), indicating the necessity of microglia in inducing the A1 astrocytic phenotype (144). Indeed, it was determined that IL-1α, TNFα, and C1q produced by activated microglia are essential factors to induce A1 reactive astrocytes (144).

To better unravel the communication between microglia and astrocytes, in vitro coculture experiments have been performed. Although the results of these studies (145, 146) suggest that the interaction between microglia and astrocyte may play a role as an autoregulatory loop to prevent an excessive inflammatory response in the brain, none of these studies assessed their role in vivo in metabolism regulation.

In conclusion, the activity of astrocytes is determined by their ability to respond to factors released by activated microglia during inflammatory conditions. In turn, microglia activation is also dependent on the inflammatory conditions and their localization within the brain. For example, the excess of nutrients induces a strong microglial inflammatory response in the hypothalamic ARC that, in turn, drives a molecularly distinct subtype of reactive astrocytes in the hypothalamus (FIGURE 1). Future studies focusing on the molecular signature of ARC microglia and astrocytes in response to nutrient availability and under different inflammatory conditions will help to define the roles of these cells and their communication, thus providing new targets in the prevention of metabolic disorders.

Hypothalamic regulation of metabolism is a complex and fine-tuned process that results from the cross talk between neurons, astrocytes, and microglia. In addition to this, peripheral macrophages and CNS-resident macrophages [CAMs; choroid plexus macrophages (cpMϕs), perivascular macrophages (pvMϕs), and meningeal macrophages (mMϕs)] can infiltrate the brain parenchyma and contribute to the regulation of metabolism. Indeed, excessive calorie intake through diet promotes the infiltration of macrophages, adding an additional level of complexity in defining brain microglia function, due to their shared myeloid hallmarks. Thus, in future studies it will be critical to discriminate the role of different microglial populations in the regulation of whole body metabolism. In line with this, Kim et al. (152) have recently developed mouse models in which the parenchymal microglia and CAMs could be distinguished based on the distinct expression pattern of Sal1 and Lyve1. Whereas Sal1 expression is restricted to microglia, Lyve1 is barely expressed in microglia and abundant in perivascular macrophages. Using the transgenic “split Cre” system, in which when two pieces of Cre enzymes named NCre and CCre are coexpressed they reconstitute a functional Cre able to cause recombination, they have successfully discriminated these two populations and characterized their translatome. Therefore, future studies using these models, Sall1^ncre:Cx3cr1^ccre and Lyve1^ncre:Cx3cr1^ccre mice (152), might offer a useful tool to discriminate the metabolic roles of microglia and CAMs in the metabolic dysfunction.

Conclusions

Given the heterogeneity of brain microglia, it has been challenging to define their functions in whole body metabolism. Several advanced technologies such as scRNA-seq, single-nucleus RNA-seq, and multiplexed mass cytometry (CyTOF) have been applied and have successfully identified subtypes of microglia. This is particularly important, since different subtypes of microglia reacting to different stimuli may have distinct properties and physiological functions, and thus produce diverse outcomes.

For example, a unique subset of microglia known as DAM has been shown to be associated with AD pathology and is conserved in mice and humans (153, 154). Like DAM, an HFD-driven unique subset of microglia may exist in the ARC. Comprehensive transcriptomic analysis at the single-cell level of ARC microglia at different stages of diet-induced obesity might provide clues to understanding the unique metabolic characteristics of ARC microglia. Furthermore, it may also address the issue of regional specificity. The success of AAV-mediated gene delivery into microglia is very promising, although presenting limitations such as the potential side effects of microglial activation due to AAV injection.

In conclusion, further studies focusing on study of distinct microglial subpopulations are needed to advance our understanding on the role of microglia in metabolism regulation and the tight association between metabolic and neurodegenerative disorders.