Hypothalamic microglia as potential regulators of metabolic physiology

Abstract

Tissue-resident myeloid cells initiate local inflammation in response to infectious or injurious stimuli. Sixteen years ago, macrophages in the adipose tissue (ATMs) were shown to undergo a form of activation in response to diet-induced obesity, thus leading to the conclusion that these macrophages sense a type of pro-inflammatory injury. ATMs are now known to be central to adipose tissue development, plasticity, maintenance and function. Indeed, their involvement in obesity may represent hijacking of these functions. More recently, microglia, ‘CNS macrophages’, have been shown to accumulate and undergo activation in response to dietary excess in the mediobasal hypothalamus (MBH), and early studies have implicated these cells as injury-responsive mediators of hypothalamic dysfunction. However, microglia are amazingly diverse cells now known to have moment-to-moment sensory functions and to communicate with neighbouring neurons to maintain and shape brain circuitry. Here, we build on this view, detailing our rapidly evolving understanding of microglial heterogeneity in the MBH and their roles as nutrient and environmental sensors. We propose that microglia, instead of simply responding to diet-induced damage, act as critical metabolic regulators that may coordinate a complex cellular network in the MBH. Understanding their roles in hypothalamic development and function should reveal unexpected mechanistic information relevant to important diseases such as obesity.

Main

Tissue-resident myeloid cells have long been known as sentinels that sense invading pathogens and act as first responders to infectious or injurious stimuli. Whereas these cells are broadly referred to as macrophages in most tissues, in the central nervous system (CNS), they are referred to as microglia, a highly specialized myeloid cell type evolutionarily adapted to the unique CNS microenvironment.

Microglia have been recognized over the past decade for their potentially detrimental role in sensing specific damage-associated molecular patterns, releasing inflammatory molecules and promoting neuronal injury or death during the pathogenesis of neuroinflammatory and degenerative disorders¹.

CNS microglia also accumulate and undergo morphological activation (‘microgliosis’) in response to dietary excess in rodent models^2,3, and non-invasive imaging findings suggest that gliosis (probably involving a combination of astrocytes and microglia) is also associated with obesity in humans^2,4. These glial changes occur very specifically in the MBH, a relatively tiny and anatomically well-defined brain region that includes the arcuate nucleus (ARC) and median eminence (Box 1), but not in other areas of the hypothalamus or non-hypothalamic brain regions⁵.

The MBH contains crucial neural circuits controlling food intake, energy expenditure and overall energy homeostasis. In addition to containing orexigenic NAG neurons, named for their expression of agouti-related protein, neuropeptide Y and the inhibitory neurotransmitter GABA (γ-aminobutyric acid), the ARC also contains pro-opiomelanocortin (POMC)-expressing neurons that suppress appetite, increase energy expenditure and play essential roles in the control of body weight. Furthermore, hormonal regulators of feeding and energy expenditure, including leptin (which signals the repletion of body fat stores), act in the ARC and inhibit NAG cells and activate POMC cells. Other nutritional and hormonal signals that control energy balance also act in the ARC. Importantly, the control of these ARC systems has been postulated to be impaired in the obese state, including by the impairment of leptin action^6,7.

Macrophages have been shown to mediate a specific form of inflammation in response to diet-induced obesity in several metabolic tissues, most notably the white adipose (WAT). Here, such ‘metabolic inflammation’ is largely driven by ATMs, which accumulate in clusters (termed ‘crown-like structures’) surrounding distressed and apoptotic adipocytes. Indeed, a rich literature has implicated ATMs in mediating the development of WAT dysfunction, consequent insulin resistance and the pathogenesis of type 2 diabetes^8,9. However, the causal relationship between WAT inflammation and insulin resistance remains unclear; a recent study has suggested that obesity-induced insulin resistance actually precedes ATM accumulation and WAT inflammation in mice¹⁰.

Microglia have been implicated in initiating and propagating the process leading to MBH gliosis³. Interestingly, diet-induced hypothalamic microgliosis is grossly reminiscent of the accumulation of ATMs in obese WAT. Primarily on the basis of this superficial similarity, hypothalamic microgliosis has been construed as another version of metabolic inflammation. In this light, early studies supported the concept that MBH microglia, like ATMs, may respond to pathologically elevated tissue levels of dietary nutrients, particularly lipids, and/or cytokines during nascent obesity. By doing so, the reasoning went, these microglia might interfere with ARC energy-balance systems and consequently promote overeating and the progression of obesity and its consequences.

However, the early notion that ATMs mainly function to mediate obesity-associated metabolic dysregulation has been supplanted by an appreciation that these and other WAT-resident innate cells have important homeostatic roles. Indeed, ATMs play important roles in lipid buffering, systemic lipid metabolism and WAT plasticity, which are an adaptive part of normal WAT function¹¹. Moreover, ATMs activated in the context of obesity are now known to be quite distinct from those undergoing classical (for example, M1) activation, as occurs in response to stimulation with lipopolysaccharide^12,13,14,15. In this newer paradigm, the unhealthful effects of ATM activation in the context of obesity represent a ‘hijacking’ of the normal cross-talk among ATMs, adipocytes and adipogenic precursors that otherwise serves to maintain normal tissue function.

Recent evidence is also challenging the concept that microglia are simply immunological sentinels and removers of cellular debris. Microglia perform critical functions in normal brain development and synaptic homeostasis¹⁶. These functions are associated with increasingly recognized activation states and the production of a broad spectrum of signalling molecules, from cytokines to neuromodulatory factors, that regulate neuronal activity. Moreover, several neurotransmitter receptors are expressed by microglia and mediate bidirectional communication with neurons. These include receptors for secreted factors that regulate hypothalamic function, such as GABA¹⁷, corticotropin-releasing hormone¹⁸, neuropeptide Y¹⁹, norepineprine²⁰, dopamine²¹, purines, adenosine and somatostatin, among others²². Furthermore, a variety of data suggest that activated MBH microglia control MBH physiology not only during the development of obesity but also during the day-to-day maintenance (and potentially programming) of synaptic connections central to energy balance^23,24. In this Perspective, we highlight new work and cutting-edge techniques that are helping to redefine the understanding of the roles of microglia in metabolic regulation (Fig. 1). We propose that, rather than examining MBH microglia solely in the context of disease, probing them as regulators of normal hypothalamic physiology may lead to the identification of previously unknown targets that could then be harnessed to ameliorate obesity and its consequences.

Fig. 1: CNS myeloid cell types as environmental sensors and key regulators of hypothalamic metabolic control.

Illustration of a traditional ‘injury’-focused model of diet-induced hypothalamic inflammation, and the emerging view that considers a heterogeneous pool of CNS myeloid cell types as environmental sensors and key regulators of hypothalamic metabolic control. Chronic nutrient excess can ‘hijack’ this normal physiological pattern of regulation to produce MBH neuronal dysfunction and promote disease. DAMPs, damage-associated molecular patterns.

Heterogeneity of hypothalamic myeloid cells

Perhaps the most important advancement in the understanding of microglia is the emerging concept that CNS myeloid cells are not a single cell type but instead are a heterogeneous array of immune cells that contribute differentially to tissue homeostasis.

Microglia arise from yolk-sac-derived precursors and seed the nascent CNS relatively early during embryogenesis; they were thought to represent one uniform myeloid cell type in the adult CNS parenchyma under steady-state conditions^25,26,27. However, growing evidence indicates that microglia display remarkable heterogeneity and diverse polarization states that may be dictated by anatomical location and both physiological and pathological factors²⁸. Indeed, given the complexity of microglial functions, the basic concept of M1 versus M2 polarization is overly simplistic and cannot adequately categorize the diversity of microglial phenotypes that are inducible by environmental cues²⁹. Certain microglial populations may be specialized for continuous circuit maintenance, whereas others may have region-specific capabilities that become effective on demand.

In addition, other myeloid cell types strategically located at CNS interfaces, including leptomeningeal, perivascular and choroid plexus macrophages, are known to be relatively long lived with respect to circulating monocytes and to contribute to the CNS myeloid pool³⁰. As discussed later, blood-borne cells can also be recruited into the CNS in response to specific chemoattractant signals produced by resident cells. After arrival into their new niches, these infiltrating cells transform into ramified cells that are relatively indistinguishable from their yolk-sac-derived counterparts. Thus, the complement of microglia-like cells is plastic and comprises both resident and recruited cell types.

Although the CNS microenvironment seems critical for the acquisition of microglial identity³¹, a recent study has suggested that cell-type-specific ontogeny also determines the ability of peripherally derived myeloid cells to adopt a microglial transcriptional identity after they enter the brain³². The factors that license peripheral myeloid cells to transdifferentiate within the CNS must be determined to assess the potential of using monocyte-derived cells, for example, to mitigate the effects of dysfunctional microglia on the progression of neurological and neurodegenerative diseases.

Notably, myeloid heterogeneity also exists in the MBH and is altered by dietary excess. Specifically, diet-induced hypothalamic microgliosis includes not only a shift in the functional polarization of cells already present in the MBH but also a fundamental reshuffling of their ontological heterogeneity. Detailed cell-surface immunophenotyping approaches have shown that the acute microglial expansion within the MBH of mice consuming a diet rich in saturated fat (HFD) occurs, paradoxically, despite a relative decrease in the number of cells within the MBH expressing classical microglial markers, such as Tmem119, that are indicative of true yolk-sac-derived microglia⁵. Instead, there is a robust increase in the numbers of ‘atypical’ resident myeloid cells that do not express such markers but that are long lived within the MBH⁵.

These atypical myeloid cells may be meningeal and/or perivascular macrophages. Indeed, fate-mapping approaches have shown that in mice, brain perivascular and subdural macrophages turn over with dwell times in excess of 70 days (ref. ³³). Moreover, perivascular macrophages have recently been implicated in mediating the effect of HFD feeding on central glucose metabolism through a mechanism involving vascular endothelial growth factor³⁴. Indeed, inhibiting inducible nitric oxide synthase in hypothalamic LysM⁺ macrophages can abrogate reactive gliosis and improve systemic glucose metabolism in mice fed a HFD³⁵. Our findings suggest that a similar population of cells in the MBH may also help mediate nutrient-dependent regulation of hypothalamic function⁵. More fundamentally, parenchymal myeloid cells within the MBH, including both true and atypical microglia, are likely to be specialized for cross-talk with the neurons among which they are interspersed, whereas the macrophages located at vascular and meningeal interfaces are likely to be specialized to regulate the unique permeability of the MBH blood–brain barrier, a concept discussed further below.

There remains a debate as to whether HFD-induced microgliosis involves local proliferation from resident precursor populations or the migration of myeloid cells from neighbouring hypothalamic regions. On the one hand, blocking cell proliferation by central delivery of arabinofuranosyl cytidine prevents microglial expansion in the ARC and blunts food intake and body-weight gain in mice fed a HFD³⁶. On the other hand, microglia have been shown to migrate to damaged brain areas in neurodegenerative-disease models in mice^37,38, but whether this phenomenon contributes to hypothalamic microgliosis remains unclear.

New methods to study microglial heterogeneity in the MBH

Resolving both the functional and ontological heterogeneity of microglia, including resident and infiltrative cell types, at the genome-wide transcriptional level has great potential to expand knowledge of MBH function, under both normal conditions and obesity. One approach to analysing microglial heterogeneity has relied on cell-surface phenotyping by flow cytometry³⁹. However, analysing CNS myeloid cell populations isolated by flow cytometry using limited surface markers constrains the ability to resolve cellular diversity. For example, some widely used markers, such as ionized calcium-binding adaptor molecule (Iba1) and fractalkine receptor (CX3CR1), are in fact non-specific and therefore inadequate to probe microglial heterogeneity. Investigating the contributions of myeloid cells to MBH metabolic homeostasis and disease pathogenesis thus requires the ability to more accurately and precisely distinguish individual subsets.

Fortunately, new microfluidic droplet-based single-cell genomic technologies are supplanting flow cytometry–based approaches and enabling unbiased and simultaneous characterization of both the functional state and developmental origins of an exceptionally wide array of cell types in a given region of the CNS⁴⁰. For instance, a recent study has used single-cell RNA sequencing (RNA-seq) to identify a novel ‘protective’ microglial subtype associated with neurodegenerative diseases that may have important implications for future treatments⁴¹. In fact, the newest droplet-based tools can be used to probe cellular heterogeneity in complex tissues, both by providing detailed transcriptomic signatures and by performing mass cytometry (CyTOF), a technique permitting detection of a far wider array of proteins per cell type than traditional flow cytometry analysis³⁹. However, as with flow cytometry, one important caveat to this new technique is that it requires the creation of single-cell suspensions that not only decrease the yield but also use ex vivo manipulations that can greatly alter myeloid cell polarization and produce artefactual confounding.

To circumvent these problems, exciting approaches have been developed that enable identification of cell-type-specific ‘translatomes’ by immunoprecipitating epitope-tagged ribosomes from minimally processed tissue extracts⁴². These approaches, often referred to as translating ribosome affinity purification (TRAP) or ‘RiboTag’, in conjunction with single-cell RNA-seq approaches, are allowing for a high-resolution understanding of microglial diversity and regulation.

Unveiling microglial heterogeneity in the MBH also calls for better, more precise in vivo tools to facilitate genetic modifications specific to true microglia or any other particular myeloid subset that might have context-specific roles worth examining. For example, the tamoxifen-inducible CX3CR1-Cre^ERT2 mouse model is widely used to restrict gene modifications to microglia. However, several different myeloid cell types express this receptor. To restrict the model to microglia, mice crossed with those expressing a particular ‘floxed’ allele and treated with tamoxifen are allowed to age for several weeks so that relatively short-lived peripheral myeloid cell populations can turn over while long-lived microglia maintain their recombination. However, this approach is cumbersome and still lacks the optimal specificity needed for targeting microglia in isolation. We have improved on this model by transplanting the bone marrow of CX3CR1-Cre^ERT2 mice with wild-type marrow before tamoxifen treatment, using lead head-shielding to protect recipient microglia from the effects of irradiation⁵. However, even this approach is highly laborious and slow to carry out.

One hope is that a deep transcriptomic examination of hypothalamic myeloid cell heterogeneity will reveal novel cell-type-specific genes that may be targeted to functionally modify individual and even rare CNS myeloid cell subsets. This capability would help achieve the ultimate goal of finding specific populations of myeloid cells that have evolved as nutrient and/or hormonal sensors, with the ability to transmit changes in nutrient status to the neurons composing circuits controlling energy and glucose homeostasis.

Do immune cells traffic into and out of the hypothalamus?

The brain is now known to be less ‘immunologically privileged’ than it was long thought to be, and this understanding has spawned new areas of neuroimmunology. For example, although both the brain and meninges were initially believed to be devoid of lymphatic vasculature, a complex mammalian CNS network of conventional lymphatic vessels located in parallel to the dural sinuses and meningeal arteries was recently identified^43,44. Classical immune cells such as B and T cells, as well as CX3CR1-expressing myeloid cells, have been found to be present in the lymphatic vessels of healthy mice, thus suggesting that the meningeal lymphatics may participate in bidirectional trafficking of immune cells during steady state⁴³. Type 2 innate lymphocytes, a cell type previously found in the gut, are also present in the meninges of mice⁴⁵. These cells have been implicated as informational mediators between the gut and the brain, but their role in meningeal function remains unknown.

Beyond lymphatic trafficking, which may affect the entire brain, the MBH in particular is also appreciated to lie outside the traditional blood–brain barrier⁴⁶, thereby allowing cells within it to sense and respond to circulating factors and enabling circulating blood cells to extravasate from the vasculature and enter the MBH. A bone marrow–transplantation strategy to specifically monitor for CNS infiltration by donor-derived (GFP⁺) blood-borne cells in recipient mice that were irradiated while being head-shielded has provided clear evidence that monocyte-derived cells enter the MBH in response to four-week HFD consumption⁵. This observation was replicated in separate mouse studies using monocyte markers such as CD169 and CCR2 (ref. ⁵). Moreover, as they accumulate, these infiltrating cells downregulate markers associated with their monocyte origin, upregulate markers indicative of conversion to microglia-like cells and assume a typical morphology⁵. The finding that blood-borne cellular infiltration contributes to diet-induced hypothalamic microgliosis has clear therapeutic implications for treating obesity and related diseases; targeting MBH-infiltrative monocytes while they are still in the circulation may be more feasible than targeting microglia already residing in the MBH. As discussed above, determining the factors dictating the ability of infiltrating cells to transdifferentiate into microglia, particularly in the MBH, as well as the functional range of the resulting cells, will be central to realizing this potential therapeutic strategy.

Hypothalamic myeloid cells as metabolic regulators

What factors drive the activation of this heterogeneous array of resident and infiltrating myeloid cells in the MBH? Since its initial description², MBH microgliosis has been viewed as a response to an undefined form of hypothalamic injury or damage induced by excess nutrient consumption, a process analogous to the response to infection, head trauma and other forms of CNS damage.

However, some paradoxes challenge this initial view of MBH microgliosis. First, microgliosis occurs in rodents within days of consuming a high-fat ‘Western’ diet, well before the onset of demonstrable obesity or adipose tissue inflammation, and it is difficult to understand how the brain would be so readily vulnerable to ‘injury’ by such an acute and common stimulus. Indeed, many mammals, including apex predators and others at the top of the food chain, eat diets highly enriched in ‘pro-inflammatory’ saturated fat yet have evolved to thrive under this dietary regimen. Furthermore, newborn mammals consume maternal milk, which is also high in saturated fat. Again, this highly evolved process is difficult to envision as being injurious to the developing hypothalamus. Additionally, MBH microgliosis in mice is reversible when overnutrition ceases⁴⁷; gliosis after CNS injury, in contrast, often remains fixed long after the damage has occurred.

Second, MBH microgliosis in mice is induced by increased consumption of saturated fats, but not unsaturated or short-chain fatty acid species, even when total fat and caloric consumption are held constant³. Indeed, hypothalamic microglia may be capable of taking up lipoprotein-encased lipids within hours of their emergence in the circulation³. As such, disrupting lipoprotein lipase in microglia leads to alterations in POMC neurons and weight gain⁴⁸. These findings all suggest that acute diet-driven MBH microgliosis may be a response to excessive consumption of specific nutrients—sensed either directly (a conclusion for which ample in vitro evidence exists) or indirectly—rather than to obesity or injury. Determining whether specific subsets of MBH myeloid cells are geared to sense multiple macronutrients or whether sensing is limited to nutritional lipids will be important questions to answer in future studies.

Third, there is a strong sexual dimorphism with respect to the kinetics and intensity of diet-driven hypothalamic microgliosis. Unlike male mice, female mice do not exhibit robust HDF-induced hypothalamic microgliosis⁴⁹ despite being susceptible to HFD-stimulated obesity (albeit not to the same degree). This dimorphism may result in part from differences in fatty acid metabolism in the brain; for example, chronic HFD consumption increases CNS palmitic acid and sphingolipid levels in male but not female mice⁵⁰. The sexually dimorphic production of, and response to, CX3CL1 in HFD-fed mice also appears to participate in generating these differences⁵¹. Rodent microglia also exhibit sexually dimorphic transcriptomic patterns during postnatal development⁵², and the effect of the microbiome on microglia is specific to both sex and age⁵³. In males, germ-free conditions most severely affect embryonic microglia, whereas in females, such conditions most notably affect adult microglia⁵³. Moreover, the transcriptomic effects of antibiotic treatment on adult microglia are sexually biased⁵³. In any case, the sexual dimorphism of diet-driven MBH microgliosis is not easily reconcilable with the suggestion that this process represents an injury response.

In contrast, microglia regulate metabolic behaviours under hypothalamic control. Pharmacologically depleting microglia or muting their capability for inflammatory activation decreases food intake and body-weight gain in mice fed a HFD. Forcibly activating microglia, remarkably, can rapidly stimulate food intake, decrease energy expenditure and promote weight gain in mice without any obesogenic dietary stimulus⁵. Thus, the conceptualization of roles played by MBH microglia is shifting from regarding them as inflammatory mediators that impair the control of energy balance to viewing them as nutrient-sensitive regulators of metabolic function. Hence, MBH microglia are not necessarily ‘bad actors’ but instead are metabolically important regulators of adaptive physiology.

However, the adaptive functions of MBH microglia may rely on signals that reliably wax and wane on an hour-to-hour and day-to-day basis, in accordance with patterns of feeding versus fasting, sleep versus wakefulness, and day versus night. Therefore, microglial functions may be greatly modified in the context of chronic, constant overnutrition in an obesogenic environment (as is the case for ATMs). The constitutive, long-term activation of myeloid cells within the MBH under such circumstances may be central to shifting their role from one that is adaptive to one that promotes progressive hypothalamic dysfunction. Focusing on the potential adaptive roles of hypothalamic microglia by using new tools and techniques may help reveal how best to intervene in their function for metabolically beneficial purposes. This focus may include exploring their ability to sense other physiological stimuli.

Hypothalamic microglia as broad physiological sensors

The concept that MBH microglia may sense factors other than pro-inflammatory molecular patterns stems from the initial observation that microglia in culture, in the context of hypothalamic-slice cultures, as well as in the MBH in mice, are responsive to dietary lipids, particularly long-chain saturated fatty acids³. Moreover, nutritional lipids accumulate in the hypothalamus when mice are fed a HFD, and depleting microglia decreases food intake and body-weight gain specifically in mice fed a HFD⁵. Additional evidence indicates that microglia are responsive to carbohydrate consumption⁵⁴, whereas relatively little is known regarding dietary amino acids. More comprehensive studies are needed to bridge existing gaps regarding the relative ability of specific MBH microglia to respond acutely to individual macronutrients.

Can MBH microglia sense blood-borne factors and metabolic signals other than nutrients? Leptin, a key peripheral signal, appears to regulate MBH microglia; despite their morbid obesity, mice lacking leptin or its receptor (LepRb) have sparser microglia in the MBH than their leaner wild-type littermates²³. Remarkably, HFD feeding can still induce MBH microgliosis even in mice lacking leptin signaling²³. Although some observations suggest that leptin may act directly on microglia⁵⁵, the evidence for microglial LepRb expression is weak. RNA-seq from hypothalamic LepRb cells has not identified microglial markers⁵⁶; moreover, single-cell RNA-seq of ARC cells has revealed LepRb expression only in neurons⁴⁰. Thus, leptin action on MBH neurons may promote neuron–glial signalling. Of note, a similar dichotomy exists in astrocytes; Whereas LepRb deletion, ostensibly in astrocytes, produces a metabolic phenotype in mice⁵⁷, astrocytes do not activate the transcription factor STAT3 in response to leptin treatment⁵⁸, and RNA-seq analyses have not revealed a significant astrocyte signature among LepRb-expressing cells in the hypothalamus⁵⁶. Overall, more work regarding leptin responsiveness among non-neuronal hypothalamic cell types is needed, including determining whether specific subsets of either microglia or astrocytes play important roles in orchestrating the overall hypothalamic response to leptin.

There is a need to determine what mediates the diurnal regulation of microglial activity in the MBH. Hypothalamic microglia in mice housed under standard light and dark conditions are more activated during the dark phase, when mice are actively feeding, than the light phase. Fasting mice for 24 hours disrupts this diurnal pattern, thereby indicating the potential for microglial entrainment by both circadian and dietary inputs²⁴. Fasting also decreases hypothalamic expression of the cytokine TNF, ostensibly produced by microglia, but the expression is restored by 4 hours of refeeding²⁴. Future studies should attempt to define local MBH paracrine factors, including potential immunomodulators and neurotransmitters, as well as any hormonal and/or autonomic signals and cognate receptors allowing microglia to sense and relay information to MBH neurons.

In this light, it is interesting to note that the gut and brain are connected by bidirectional autonomic, endocrine and immunological axes. Altering the composition and quantity of gut bacteria affects both the enteric nervous system and the CNS. The microbiota is a determinant of the ability of dietary excess to produce chronic low-grade tissue inflammation as well as obesity and attendant metabolic dysfunction⁵⁹. This influence may result from an ability of the microbiota to regulate host gut–brain–immune system cross-talk⁶⁰. Although prior work has highlighted the microbiome as a critical regulator of microglial activity⁶¹, the specific microbiota-associated factors influencing microglial function are unknown. Short-chain fatty acids may be important, given that providing them to germ-free mice causes the number and morphology of microglia to revert to those normally seen in conventionally reared mice⁶¹.

Metabolic programming and microglial memory

Highly intriguing data suggest that microglia not only regulate adult hypothalamic function but also may mediate the ability of maternal and neonatal factors to program hypothalamic circuitry and consequent metabolic function⁶². Maternal nutrition is well established to influence the metabolic homeostasis of offspring, even into the offspring’s adulthood⁶³. This phenomenon is true in rodent models but also has strong correlates in human populations. Interestingly, the adult offspring of mice fed a HFD during pregnancy and lactation have been shown to have increased microglial activation, as revealed by morphological changes in Iba1⁺ cells^64,65. These changes are apparent without any dietary challenge to the offspring themselves, thus suggesting that they are a consequence of programming that occurs when the mice are sustained solely by maternal milk⁶⁶.

Feeding pregnant mice a diet containing the colony-stimulating factor 1 receptor inhibitor PLX5622 depletes microglia from their growing embryos. Recent work has shown that embryonic microglial depletion through this approach is sufficient to decrease the number of hypothalamic POMC neurons and accelerate postnatal weight gain, thus suggesting that microglia are necessary for proper development of the hypothalamic satiety center⁶².

Innate immune memory refers to the emerging concept that tissue myeloid cells are programmed by exposure to a wide variety of stimuli, including potentially metabolic or dietary factors, and that this exposure primes a more rapid and robust response after re-exposure to a given stimulus, even if it is temporally remote from the original stimulus. Microglia are particularly interesting for studying immune memory, because they are among the longest-lived myeloid cells. As such, epigenetic or other genomic changes induced among microglia in the MBH by specific stimuli, including circulating nutrients, hormones or cytokines, would be retained for at least 6 months and would produce a potentially robust memory. In a model of stroke, for example, microglial activation and neuronal damage are strongly diminished in animals primed by repetitive lipopolysaccharide injections 1 month before the insult⁶⁷. It will be interesting to determine whether the exposure of MBH microglia to circulating physiological and nutritional factors during neonatal life may produce a memory that influences hypothalamic responses to metabolic challenges during adulthood.

Conclusions

New evidence indicates that microglial activation and CNS myeloid cell expansion occur not only in response to injury or pathogen invasion—circumstances in which such a response may aptly be termed ‘inflammation’—but also in response to a much broader array of physiological and developmental signals than previously recognized. Indeed, the formation of reactive microgliosis is an essential response needed for synaptic and circuit maintenance, programming, and neurological and behavioural conditioning. In this specific context, microglial reactivity should not necessarily be equated with inflammation but instead should be regarded as the innate immune regulation of CNS remodelling and function.

In the MBH, microglia are regulators of metabolic control, a role involving a network of myeloid cells responding to a variety of physiological cues. Unfortunately, this network, which is not inherently deleterious but instead is a component of normal hypothalamic physiology, can be set into ‘overdrive’ by continuous exposure to environmental and nutritional stimuli composing the modern obesogenic environment (Fig. 1). It is this chronic, unyielding form of microgliosis that is likely to produce clinically harmful effects. By using new, unbiased and genome-wide tools to determine exactly which myeloid cell types cooperate within the MBH, the myriad factors that they sense and the mechanisms driving hypothalamic neuro-immunological communication, a new level of understanding of metabolic control by the brain can be gained. This understanding may then enable the development of innovative ways to manipulate microglia to lessen the burden of metabolic diseases, including obesity and its consequences.

Box 1 Monitoring innate immune reactivity in the human hypothalamus

Growing evidence indicates that obesity produces hypothalamic gliosis in humans. A retrospective analysis of brain magnetic resonance imaging data in human subjects has shown that signal intensity suggestive of gliosis increases along with obesity specifically in an area of the brain corresponding to the MBH, results presumably replicating the innate immune reactivity seen in mice². In addition, a prospective study has revealed an association between increasing BMI and longer T₂ signal-relaxation times, indicative of potential gliosis, in the MBH, particularly on the left side⁶⁸. This obesity-associated MR signature correlates with histological data from postmortem brain sections showing corresponding increases in staining for glial fibrillary acid protein. Together, these findings support the concept that non-invasive imaging may be used to detect the presence of gliosis in human subjects, although its ability to specifically monitor microglial accumulation in the MBH remains to be evaluated. Further extending these findings, a recent study has shown a potential association between the gut microbiome and hypothalamic gliosis⁴. Specifically, investigators have found that the abundance of Parasutterella species in gut bacteria is associated with the ability of fat intake, but not total energy intake, to increase hypothalamic magnetic resonance signal intensity. More human studies using new tools are needed to validate findings from animal models and define the effects of future strategies to control hypothalamic gliosis and thereby mitigate obesity and its consequences.