The Significance of Hypothalamic Inflammation and Gliosis for the Pathogenesis of Obesity in Humans

Abstract

Accumulated preclinical literature demonstrates that hypothalamic inflammation and gliosis are underlying causal components of diet-induced obesity in rodent models. This review summarizes and synthesizes available translational data to better understand the applicability of preclinical findings to human obesity and its comorbidities. The published literature in humans includes histopathologic analyses performed postmortem and in vivo neuroimaging studies measuring indirect markers of hypothalamic tissue microstructure. Both support the presence of hypothalamic inflammation and gliosis in children and adults with obesity. Findings predominantly point to tissue changes in the region of the arcuate nucleus of the hypothalamus, although findings of altered tissue characteristics in whole hypothalamus or other hypothalamic regions also emerged. Moreover, the severity of hypothalamic inflammation and gliosis has been related to comorbid conditions, including glucose intolerance, insulin resistance, type 2 diabetes, and low testosterone levels in men, independent of elevated body adiposity. Cross-sectional findings are augmented by a small number of prospective studies suggesting that a greater degree of hypothalamic inflammation and gliosis may predict adiposity gain and worsening insulin sensitivity in susceptible individuals. In conclusion, existing human studies corroborate a large preclinical literature demonstrating that hypothalamic neuroinflammatory responses play a role in obesity pathogenesis. Extensive or permanent hypothalamic tissue remodeling may negatively affect the function of neuroendocrine regulatory circuits and promote the development and maintenance of elevated body weight in obesity and/or comorbid endocrine disorders.

Graphical Abstract

Essential Points.

• Emerging translational evidence supports that hypothalamic inflammation and gliosis contribute to obesity and its comorbidities in humans, as seen in rodent models.

• Convergent findings are derived from preclinical studies, a small number of histopathological studies in humans, and multimodal in vivo neuroimaging studies.

• Theoretically, hypothalamic inflammation and gliosis could be adaptive in that it facilitates energy storage during periods of nutritional abundance, but the authors posit that, in the modern food environment, these responses ultimately become chronic and maladaptive to maintaining optimal metabolic health.

Much of our understanding of central nervous system (CNS) regulation of energy homeostasis has been derived from preclinical models. Through animal studies, control of energy homeostasis has been delineated as a function of the hypothalamus, and neuronal pathways sensing peripheral energy stores and nutritional state localized to the mediobasal hypothalamus (MBH), particularly the arcuate nucleus (1), which is in extensive communication with other hypothalamic nuclei and more distal regions. Multiple lines of active inquiry are ongoing into whether disruption of homeostatic regulation plays a role in obesity pathogenesis. One emerging line of investigation posits that inflammation and gliosis—a cellular inflammatory response by glia—in the arcuate nucleus is an underlying causal component of diet-induced obesity in rodent models (2, 3). The purpose of the current review is to summarize and synthesize available translational data indicating the relevance of this hypothesis to obesity and its comorbidities in humans.

Background and Preclinical Studies

In the periphery, adipose tissue expansion results in the release of proinflammatory cytokines. This fact led to early preclinical investigations denoting that inflammation is present in the arcuate nucleus of the hypothalamus in diet-induced obesity and that inflammatory mediators can disrupt energy homeostasis (4–6). What soon differentiated hypothalamic inflammation from that observed in the periphery is that CNS inflammation preceded the onset of diet-induced obesity, rapidly following initiation of high-fat diet (HFD) feeding and coinciding with the hyperphagia that characterizes HFD feeding in susceptible animals (7). Proinflammatory cytokine gene expression was induced in the hypothalamus of rodents within 24 hours of the onset of HFD feeding. However, CNS inflammation has a cellular component composed of glial cells including microglia (brain macrophages), astrocytes, and other cell types such as oligodendrocytes and tanycytes (8, 9). Microglial expansion, astrocyte proliferation, and the assumption of their respective reactive conformations (which characterizes gliosis) were documented within days of onset of HFD feeding (Fig. 1A) (10). Subsequent work has shown that microglial nuclear factor kappa B–dependent signaling and astrocyte activation via the I-kappa-B kinase-beta signaling pathway are each necessary and sufficient for hyperphagia and weight gain in rodent models of diet-induced obesity (3, 11, 12). In other words, diet-induced obesity is, in part, dependent on glial cell inflammatory signals. These studies and comprehensive reviews of the preclinical literature (13, 14) suggest that arcuate nucleus inflammation is causally related to diet-induced obesity in rodent models. Importantly, HFD, used in rodent models, is an extreme nutrient stimulus that commonly provides about 60% calories from saturated fat in contrast with a standard chow diet with minimal saturated fat content—macronutrient profiles that are not typical of usual diets in the United States, which are, on average, 50% carbohydrate, 35% fat, and 15% protein. Moreover, rodents might be uniquely susceptible to local injury and inflammation for genetic or evolutionary reasons. Data demonstrating arcuate nucleus proinflammatory cytokines and localized activation of microglial cells have, however, been found in fetal hypothalamus of nonhuman primates when the mother was fed HFD (15), suggesting that this mechanism may be relevant in other mammalian species, but data from humans are still required.

Figure 1.

Models of cellular changes in the hypothalamus of rodent and human obesity. (A) Reactive gliosis in the hypothalamus of rodent models of diet-induced obesity. Dietary excess, such as a high-fat diet (HFD), leads to a cellular inflammatory response named reactive gliosis in the arcuate nucleus of the hypothalamus. Gliosis is characterized by increased presence of astrocytes and microglia in their reactive forms. Histopathologic findings by immunofluorescence in mice that received HFD show greater intensity of astrocytes by glial fibrillary acidic protein (GFAP) (b) and increased microglial number by ionized calcium binding adaptor molecule 1 (Iba1) (d) compared with chow fed ((a) and (c), respectively). Adapted from Lee D et al (10). White regions denote arcuate nucleus. 3 V, 3rd ventricle. White scale bar in (a) is 50 µm. (B) Histopathology of reactive gliosis in the hypothalamus of humans with obesity. Humans with obesity demonstrate evidence of reactive gliosis in the mediobasal hypothalamus when compared with healthy weight participants. Histopathologic findings provide evidence of increased microglial concentration (by Iba1 immunoreactivity) in the hypothalamus (top row) of postmortem samples from humans with BMI > 30 kg/m² vs humans with BMI < 25 kg/m², while no difference is seen for microglial presence in the cortex (bottom row). Adapted from Baufeld C et al (17). The asterisk indicates the third ventricle; scale bar is 100 µm. (C) Model of hypothalamic gliosis in humans. On the left, hypothalamic neurons normally interact with a population of microglia and astrocytes in their quiescent state. On the right, under conditions of obesity, hypothalamic microglia and astrocyte populations have proliferated, taken on their reactive conformations, and potentially formed glial scars.

Methods

To assess the translational significance of preclinical findings in rodents, we undertook a comprehensive review of the human literature on hypothalamic gliosis in relation to obesity. To ensure inclusivity and determine the current state of the literature, we devised a PubMED search strategy to yield both postmortem studies and in vivo radiologic evidence from humans as follows: (gliosis OR astrocyte OR microglia OR inflammation OR injury OR microstructure) AND (neuroimaging OR MRI OR magnetic resonance imaging OR T2 OR diffusion imaging OR “brain imaging” OR PET OR “brain region” or “brain area” OR radiologic OR postmortem) AND (hypothalamus OR hypothalamic OR arcuate nucleus) AND (obesity OR overweight), yielding n = 92 items. This search was replicated in the Web of Science database and yielded n = 22 novel items. Citations were reviewed and an additional 6 articles identified for a total of n = 120 items. Of these, 94 were excluded because they did not investigate the hypothalamus, gliosis, obesity, and/or inflammation; were a review article; were in a language other than English; were not peer reviewed; or did not include humans. Details on the included citations are elsewhere (Table S1 (16)). Many of the human studies test cross-sectional associations, which provide neither directionality nor causality. Therefore, we fill gaps and augment them with preclinical studies of mechanism. A dedicated section reviews prospective studies because this approach can provide evidence that hypothalamic inflammation and gliosis temporally precede obesity and its comorbidities in humans and thereby support causality.

Histopathologic Evidence in Humans

Histopathologic evidence in humans is limited to observational data from postmortem samples. Baufeld et al (17) compared hypothalamic and cortical tissue specimens from 12 individuals with obesity (body mass index [BMI] > 30 kg/m²), and 9 with BMI < 25 kg/m², all of whom had died from causes that excluded infectious, inflammatory, and neurodegenerative diseases, among others. They found significantly increased ionized calcium-binding adapter molecule 1 (Iba1, a marker of microglia) staining area in the group with obesity, a correlation between BMI and Iba1 staining, and greater late-stage microglial dystrophy to be present in the arcuate nucleus region of the hypothalamus and not in frontal cortex (Fig. 1B) (17). Glial fibrillary acidic protein (GFAP, a marker of astrocytes) immunoreactivity was not significantly different. A study by Kreutzer et al (18) provided evidence of reactive gliosis by GFAP and Iba1 staining in postmortem hypothalamic sections in a single deceased subject with overweight compared with a single subject of normal weight. Although not a study of obesity per se, Schur et al (19) reported on the presence of astrocytosis in the mediobasal hypothalamic location of the arcuate nucleus using GFAP immunohistochemical staining in postmortem brain slices through the hypothalamus (n = 10 individuals). This study linked histologic findings with a radiologic measure—T2 relaxation time. Specifically, increased GFAP staining intensity was associated with longer T2 relaxation time (see below), establishing that T2 relaxation time is a marker of gliosis in human hypothalamic brain tissue. This finding documents the presence of arcuate nucleus reactive astrocytosis in humans. Moreover, along with a study of the MBH in HFD-fed rodents (10) and other histopathologic correlation studies performed in human and rodent nonhypothalamic tissue (20–22), it provides strong support that expansion and activation of glial cell populations in the MBH can be detected in humans using radiologic markers (see below) that detect shifts in the magnetic and biochemical properties of the tissue (Table 1).

Table 1.

Definitions of MRI modalities used to study hypothalamic gliosis

MRI modality	Parameter	Surrogate for	Finding in gliosis	Histopathologic validation for gliosis
T2 relaxometry	T2 relaxation time	Total water content	↑	Hypothalamus in mice and humans; frontal cortex, thalamus, and cerebellar WM in rodents; hippocampus in rats and humansa
T2-WI	Signal ratio	Total water content	↑	Hippocampus in dogs and humans; Fornix, thalamus, and hypothalamus in dogs
FLAIR	–	Ischemic lesion	↑	Amygdala in humansa
Diffusion-WI	Primary eigenvalue	Water diffusivity along the major axis (axial)	↑	Cortex in rats
	Mean diffusivity or Apparent diffusion coefficient	Average rate of water diffusion, independent of orientation	↑	Cortex in rats; Putamen, caudate in piglets; Hippocampus and OFC in humans
	Fractional anisotropy	Coherence of the orientation of water diffusion (fibers), independent of rate	↑	Cortex and hippocampus in rodents; putamen and caudate in piglets
Proton density-WI	Water content mapping	Free water content	↑	WM in cats and humansa
Multiecho T2*-WI	R2*	Iron concentration	↑	Thalamus in rats

Abbreviations: MRI, magnetic resonance imaging; WI, weighted image; WM, white matter; OFC, orbitofrontal cortex.

^a Evidence from case study.

The following sections summarize existing studies utilizing several radiologic techniques that assess indirect markers for tissue structure and microstructure in vivo in human CNS. Studies using indirect measures represent the majority of findings related to changes in hypothalamic structure. However, such indirect measures are relatively nonspecific compared with histology and, therefore, tissue-level changes in edema or water content, neuronal density, extracellular structures, such as proteoglycans and perineuronal nets, or the relative proportions of multiple glial cell types could entirely or partially explain some results (8, 23). Moreover, as neural tissue remodels, multiple of these changes likely occur in parallel or reciprocally, and the combined effect of tissue remodeling may be what ultimately accounts for the consistency of findings across diverse modalities that are typically interpreted as detecting different microstructural abnormalities. For that reason, these sparse histopathologic findings in human tissues carry outsized weight in our interpretation of radiologic findings as indicative of, at least in part, glial cell expansion and inflammatory response as observed in preclinical models.

Radiologic Evidence in Humans

Radiologic methods provide an accessible approach to in vivo assessment of the CNS in humans of all ages. Thus, the bulk of the literature evaluating hypothalamic inflammation and gliosis uses neuroimaging. In clinical neuroradiology, CNS inflammation and gliosis are evaluated qualitatively by visually inspecting for areas of T2 hyperintensity (brightness) in gray matter. However, for research purposes, quantitative methods that measure T2 relaxation—a magnetic resonance imaging (MRI) parameter that is intrinsic to the tissue—enable comparison of images across subjects or evaluation of within-subject changes over time. A wide variety of newer techniques and sequences are also capable of evaluating tissue microstructure, particularly as regards the signature effects of increased water content or its diffusion across tissue compartments (Table 1) (20, 22, 24–34). In the broadest sense, MRI techniques detect variation in the biochemical properties of tissues that affect their response in an electromagnetic field and are interpreted as reflecting underlying differences in tissue structure and microstructure (35). Normal T2 signal, for example, changes predictably in developing brain tissue across childhood, adolescence, and adulthood (36). Many of these radiologic measures are affected by neuroinflammation and some have been confirmed by histopathologic studies as indirect markers thereof. In sum, while the classical finding of increased T2 signal intensity is most closely associated with reactive gliosis (10, 19, 20), radiologic approaches to assessing hypothalamic inflammation and gliosis are varied, providing convergent evidence that the tissue properties of the hypothalamus are altered in human obesity. Throughout the following summary of radiologic evidence, brief explanations of techniques will be integrated for readers who may be unfamiliar with neuroradiology in general or with a specific approach. Terminology and interpretation of the techniques discussed in the cited literature are also provided in Table 1 for reference.

Quantitative T2 Imaging

The first methodology applied to assess hypothalamic inflammation and gliosis was accomplished by creating a T2 signal ratio on standard, clinically-acquired T2-weighted images. A signal ratio compares the target region of the MBH (encompassing the arcuate nucleus) to a reference region of interest, providing a value for the MBH that is, in essence, normalized to the average T2 signal intensity of adjacent gray matter, such as amygdala or putamen, which are in the same coronal plane as the MBH and in which T2 relaxation time does not appear to vary in relation to the degree of adiposity (37). T2-weighted images have no absolute scale, therefore calculation of T2 signal ratios allows comparisons of MBH signal intensity between individuals. Greater signal ratios indicate higher T2 signal intensity in MBH and are consistent with the presence of inflammation and gliosis (10, 19, 20).

The first evidence for hypothalamic gliosis in humans was published in a retrospective analysis of clinically-acquired T2-weighted (or T2 FLAIR) coronal MRI sequences containing the hypothalamus. All were read as normal by a neuroradiologist. The study found, in 37 adults, a positive correlation of MBH/amygdala T2 signal ratio and BMI, which persisted after adjustment for age and sex, and was lateralized to the left hypothalamus (7). A subsequent study replicated this association, showing a significantly increased left MBH/amygdala T2 signal ratio in a sample of 57 adults with obesity compared with an age- and sex-matched sample of 54 adults without obesity (18). The cross-sectional association between the MBH/amygdala T2 signal ratio and body adiposity has also been expanded to the pediatric population (37–39). In a sample of 169 children derived from a large, multisite, population-based, observational cohort study, aged 9-11 years, it was found that the bilateral average MBH/amygdala T2 signal ratio was positively associated with BMI z-score, independent of study site, age, and sex. Moreover, children with healthy weight had lower MBH/amygdala T2 signal ratio when compared with children with overweight and with obesity (39). In contrast to the lateralization observed in prior studies (7, 18), the association with BMI z-score was confirmed for both left and right MBH/amygdala ratios when evaluated separately, possibly due to the younger age of the population investigated and/or the large sample size. Though previous studies have found that hypothalamic connections are lateralized in humans (40–42) and rodent models (43), whether the susceptibility to or consequences of MBH gliosis differ by lateral location requires further investigation.

A measure of the N-acetylaspartate/creatinine ratio, an indirect marker of neuronal content derived from magnetic resonance spectroscopy of the left hypothalamus, did not show significant differences in neuronal content (18). This finding contrasts with rodent studies that have shown reductions in pro-opiomelanocortin (POMC) cell number specifically after chronic HFD feeding (7) as well as increases in markers of apoptosis and neuronal stress in arcuate nucleus neurons (5, 6, 44, 45). It may be that changes in neuronal populations are not occurring in humans as observed in rodent models. Alternatively, reductions in neuron populations may occur in hypothalamic inflammation in humans, but not be detectable with current noninvasive radiologic strategies (5, 46).

An alternate strategy is the measurement of T2 relaxation time (ms) via T2 relaxometry. In this approach, multi-echo sequences acquire T2-weighted images at a series of echo times to calculate the decay curve of T2 signal, translating values into a T2 parametric map upon which regions of interest can be placed and average T2 relaxation times calculated. Recently, the use of this quantitative T2 approach was demonstrated to be feasible in a multisite study (38).

Using T2 relaxometry, Schur et al (19) demonstrated in adults (n = 67) that left MBH T2 relaxation time was positively associated with BMI. Moreover, participants with greater evidence of MBH gliosis had higher BMI, longer mean T2 relaxation time in the right MBH, and increased markers of insulin resistance, as will be further discussed below. By applying the same method in children, Sewaybricker et al (37) illustrated that T2 relaxation times within bilateral MBH (and not reference regions) were longer in children with obesity (n = 11) than in controls (n = 9), independent of age. Furthermore, longer MBH T2 relaxation times were associated with increased plasma leptin concentration, another biomarker for adiposity (37). It is possible peripheral leptin levels were elevated due to some degree of leptin resistance at the level of the CNS (47). The etiology of central leptin resistance is not yet fully understood; however, preclinical models have indicated that hypothalamic inflammation is linked to the development of central leptin resistance (48) and that glial cells such as astrocytes (49) and microglia (2) play a role in the reduction of leptin function in the CNS in diet-induced obesity in rodents.

Additional findings of a positive correlation with MBH T2 relaxation time and BMI z-score were found in a multisite cross-sectional study in children (n = 31) (39). Detecting MRI evidence of hypothalamic gliosis in association with obesity not only in adults but also in the pediatric population highlights that hypothalamic inflammation and gliosis is not the result of long-standing processes or decades of exposure to obesity, but occurs early in life and may have a role in the tendency of obesity to persist into adulthood and/or be a risk factor for the subsequent development of comorbidities.

Diffusion-Weighted Imaging and Other Techniques

Diffusion-weighted imaging, including diffusion tensor imaging (DTI) scans, has been used to provide insight into microstructural characteristics of the hypothalamus and query their association with obesity. DTI parameters, such as mean diffusivity, are interpreted as indirect markers of tissue integrity, where increased proton diffusivity could represent abnormal cellular density or other characteristics that impact the directionality of water diffusion in the underlying tissue. The hypothalamic primary (λ₁) eigenvalue, which is a DTI parameter representing water motion along axons and may be a marker of early microstructural compromise (50), was shown to be low in 24 participants with obesity when compared with participants of healthy weight (51). Interestingly, this same DTI parameter in the hypothalamus, together with gut microbiota characteristics, were among the variables with greatest weight in a multivariate model able to discriminate between 20 adults with and 19 without obesity (52).

Furthermore, in a study with 115 adults, Kullmann et al (53) demonstrated, by using whole-brain proton density acquisition, a quantitative assessment of brain water content, that higher proton density in the lateral hypothalamus, among other regions, was associated with higher BMI, adjusted for total gray matter free water content, age, and sex. There were no differences in global cerebral blood flow via arterial spin labeling techniques to explain the findings. Given that the findings were concentrated in subcortical and hypothalamic regions and were not explained by cerebrovascular differences, the authors interpret the results as indicative of localized inflammation and altered brain structure in obesity.

More recently, semi-automated or automated segmentation analyses have been developed for the study of the hypothalamus and its subnuclei. These techniques allow sample size expansion and use an unbiased anatomic approach. As shown by Thomas et al (54) in 311 adults, mean diffusivity of the whole hypothalamus, obtained by a semi-automated approach, was positively associated with BMI, adjusted for age and sex, suggesting that greater—and potentially abnormal—cellularity is associated with higher BMI. In the same work, the authors confirmed the finding in an independent sample of 236 adults. Moreover, Spindler et al (55) applied a fully automated segmentation approach and showed, in 98 adults, an association between greater mean diffusivity in an anterior superior hypothalamic cluster, encompassing the paraventricular, preoptic, and anterior nuclei, and higher BMI. This work highlights that a computerized segmentation allows the study of specific areas of the hypothalamus and may help to parse out effects within distinct regions of the hypothalamus. Finally, a distinct diffusion-weighted image approach was used by Alkan et al (56) where an apparent diffusion coefficient map was acquired, which by removing T2-weighting can quantitatively examine water diffusion and better differentiate intracellular and extracellular water compartments (57). They demonstrated that participants with obesity (n = 81) had higher apparent diffusion coefficient values, suggesting increased extracellular water content or vasogenic edema, in the hypothalamus, among other brain regions involved with appetite regulation, when compared with 29 age-matched participants without obesity.

Summary and Limitations of Radiologic Evidence in Obesity

Taken together, distinct neuroimaging approaches and diverse populations and age groups converge to demonstrate that MRI evidence of microstructural abnormalities in the hypothalamus is associated with obesity in humans. Most of these data implicate the mediobasal hypothalamus region housing the arcuate nucleus, which has predominantly been the focus of mechanistic preclinical studies. However, whole hypothalamic abnormalities and the lateral hypothalamic area stand out in isolated findings, warranting replication. The above studies have been cross-sectional, documenting the presence of hypothalamic abnormalities in affected individuals. A small number of prospective studies are discussed below, but cross-sectional findings cannot address whether, in humans, obesity preceded or followed radiologic evidence of hypothalamic inflammation and gliosis. At the same time, imaging of the hypothalamus is challenging due to its small size and the large voxel sizes required (ie, lower resolution) of methods such as DTI and arterial spin labeling; lack of readily discernible anatomic margins (particularly superiorly and laterally); and proximity to the third ventricle and sinuses, which makes it particularly subject to imaging artifacts. Given these limitations, improvement of MRI resolution without demanding longer acquisition times, using approaches such as compressed sensing or 3D quantification using an interleaved Look–Locker acquisition sequence with T1 preparation pulse, combining distinct techniques, and developing automated approaches for image analysis of the hypothalamus offer possibilities for progress in clinical neuroimaging research on obesity pathogenesis. Of note though, computerized segmentation of the hypothalamus and identification of its subnuclei have been developed based on adult brain and still pose a challenge to apply to the pediatric population, especially infants and neonates. Additional neuroimaging–histopathology validation studies of the hypothalamus are needed to build on what was already demonstrated for T2 relaxation time in the MBH (19). Such studies could provide critical translational information regarding the specific cellular populations and tissue-level structural changes that are being reflected by other MRI parameters found to be altered in the hypothalamus of humans with obesity.

Evidence in Obesity-Related Comorbidities

Associations With Visceral Adiposity and Insulin Resistance

Metabolic phenotypes vary significantly among individuals with obesity (58). Because increased visceral adipose tissue (VAT) confers greater risk for diabetes (59), cardiovascular disease, and mortality than peripheral subcutaneous fat (60), multiple investigators have used neuroimaging to probe whether hypothalamic inflammation and gliosis could contribute specifically to deposition of VAT (37, 53, 54, 61, 62). VAT adipocytes are more metabolically active and contain a larger number of inflammatory and immune cells, which parallels findings in obesity-related hypothalamic gliosis. In the study referenced by Thomas et al, VAT volume was positively associated with whole hypothalamic mean diffusivity (54), a finding echoed by Kullmann et al, who associated greater visceral fat with higher lateral hypothalamus water content; however, the latter did not remain significant after adjustment for sex and age (53). In 2 studies of adults, MBH gliosis as measured by T2 relaxation time positively correlated with visceral fat mass by dual-energy X-ray absorptiometry (DEXA) (61, 62). Children also exhibit the possible connection between VAT deposition and hypothalamic gliosis such that MBH gliosis correlated with measures of higher adiposity such as visceral fat percentage measured by DEXA (37) and waist/height ratios, an indirect measure of central adiposity (39). In sum, mounting evidence suggests a connection between hypothalamic inflammation and gliosis and preferential deposition of VAT.

Dysregulation of VAT is linked to insulin resistance, which suggests a connection between insulin resistance and obesity-related hypothalamic inflammation and gliosis. In the first study to look at human hypothalamic gliosis and glucose homeostasis, Schur et al (19) demonstrated that participants who had the longest MBH T2 relaxation time on the left also had higher fasting insulin and homeostatic model assessment of insulin resistance (HOMA-IR) scores, independent of age, sex, and BMI—establishing the link between hypothalamic gliosis and insulin resistance in humans, independent of their degree of body adiposity (19). More recently, similar measures of insulin resistance (HOMA-IR and fasting insulin) were shown to be positively associated with MBH T2 relaxation times in a group of females of normal weight and with obesity, with and without type 2 diabetes (T2D) (63). Contrasting these findings, there was no correlation between left MBH/amygdala T2 signal ratio and HOMA-IR or triglycerides in a sample of 57 adult participants with obesity and 54 controls whereas measures of systemic inflammation (C-reactive protein, interleukin-6) were positively correlated (18). Other atypical markers of metabolic dysfunction such as uric acid (64) and osteocalcin (65) have also demonstrated mixed relationships with MRI evidence of hypothalamic gliosis. Some rodent studies suggest sexual dimorphism in the degree to which microglial inflammation promotes insulin resistance (66). While mixed results caution that the relationships may have additional complexity to be uncovered, a number of studies have implicated a link between insulin resistance and hypothalamic inflammation and gliosis that, preliminarily, may be independent of total body adiposity, and is reinforced by findings in T2D.

Association With Glucose Intolerance and Type 2 Diabetes

The most recent and robust evidence of the association between glucose intolerance and hypothalamic gliosis comes from a study by Rosenbaum et al (61) which provided evidence that hypothalamic gliosis plays a role in the presence and progression of insulin resistance in humans with obesity. Sixty-seven adults with obesity and normal glucose tolerance, impaired glucose tolerance, or T2D underwent brain MRI, DEXA scan for body composition, and oral glucose tolerance tests (61). A cross-sectional analysis demonstrated that MBH T2 relaxation times were significantly greater in the group with impaired glucose tolerance and T2D compared with normal glucose tolerance—a similar finding was previously found in women with obesity and T2D (63). Using a novel approach with β-cell modeling, longer MBH T2 relaxation times were associated with lower β-cell glucose sensitivity, insulin sensitivity, and oral disposition index—a measure of early β-cell response that predicts future onset of T2D. Rosenbaum et al (61) further demonstrated longer MBH T2 relaxation times were associated with higher glucose area under curve, fasting glucose, and hemoglobin A1c, independent of body fat mass. Along this vein, Kullmann et al found that individuals with metabolic syndrome had the highest water content within the hypothalamus as measured by quantitative proton density imaging compared with those with central obesity without metabolic syndrome and healthy individuals without obesity, even after adjustment for age and sex (53). Higher hypothalamic water content implies microstructural damage or inflammation, and this finding was accentuated in individuals with metabolic dysfunction (53), which parallels findings from quantitative T2 studies. With the caveat that the effect of altered carbohydrate metabolism on the hypothalamus must be considered, these cross-sectional data provide novel evidence that, in obesity, the degree of metabolic dysregulation in humans is related to structural changes specific to the hypothalamus and are consistent with the proposed role of the CNS in T2D pathogenesis (67).

These radiologic studies complement accumulated preclinical evidence supporting the role of the brain in control of peripheral glucose (67) and potential contribution of hypothalamic inflammation and gliosis to the pathophysiology of T2D (7, 12, 14, 68). It is well documented that hypothalamic neurons, particularly in the arcuate nucleus and the ventromedial nucleus of the hypothalamus (VMH), sense and respond to changes in circulating glucose, especially as regards initiation of counterregulatory responses to hypoglycemia (69). Prior reviews detail the integrative hypothalamic neuronal pathways and glial cell populations that control peripheral glucose, specifically, and circulating fuels, more generally, to maintain sufficient energy availability for survival (69, 70). It is clear that hypothalamic neuron populations and their interactions with glial cells (including astrocytes) modulate circulating levels of glucose (70–73). For example, deletion of insulin receptors in astrocytes reduces glucose-stimulated activation of POMC neurons in the arcuate nucleus and results in hyperglycemia during an intravenous glucose tolerance test (71). Similarly, depletion of brain-derived neurotrophic factor in VMH astrocytes increases synaptic glutamate disposal by astrocytes and suppresses excitatory neuronal signaling in response to glucose, lipids, and leptin (73). Critically, hypothalamic activation of I-kappa-B kinase-beta inflammatory signaling and endoplasmic reticulum stress were both shown to be key for peripheral insulin resistance and glucose intolerance to occur in rodents (6, 74, 75) by perturbing neuroglial function. Hypothalamic microglial involvement in cephalic phase insulin response was also recently documented in rodents; moreover, in diet-induced obese rodents, inhibition of interleukin-1-β–mediated hypothalamic inflammation reversed diet-induced impairment in cephalic phase insulin response (76). Finally, sustained remission of hyperglycemia in rodent models of T2D is evoked by injection of fibroblast growth factor 1 into the MBH (77), acting by mechanisms that involve remodeling of extracellular matrix structures surrounding neurons (78) and inferring that neuronal dysfunction can be rescued to lower peripheral glucose. In sum, preclinical studies show that normal neuronal and glial cell function are required to maintain glucose homeostasis. Moreover, inhibiting function of specific neuron and glial cell populations that reside in the hypothalamus—including dysfunction caused by inflammation—evokes glucose intolerance, insensitivity to peripheral signals of energy availability, and hyperglycemia, all classic manifestations of insulin resistance and T2D in humans. This preclinical literature suggests that inflammation and gliosis could contribute to the development of T2D by perturbing neuronal sensing of peripheral signals of energy availability, neuronal signaling, or the neuron–glia interactions that regulate neuronal activity. Human studies, to date, are supportive of this conclusion. They demonstrate detectable alterations in tissue structure which could indicate underlying changes in cellular function. Additionally, the severity of detected tissue changes associates with metabolic disease progression both cross-sectionally and prospectively (as discussed below). Nonetheless, a direct and consistent causal link between hypothalamic inflammation and gliosis and T2D development has yet to be established in either preclinical or human studies.

Reproductive Dysfunction

In addition to energy balance and glucose homeostasis, the hypothalamus and arcuate nucleus are resident to key neuronal populations involved in reproductive function. Neurons of the central melanocortin system (agouti-related protein [AgRP]/neuropeptide Y [NPY] and POMC) and kisspeptin/neurokinin B/dynorphin (KNDy) neurons are intricately connected and integrate signals to affect gonadotropin-releasing hormone (GnRH) pulse generation to play a critical role in gonadal function and reproductive capacity in humans and rodents (Fig. 2A) (79, 80). Rodent studies have illustrated that loss of both leptin and insulin receptors in POMC neurons leads to a polycystic ovarian syndrome–like phenotype with ovarian abnormalities, insulin resistance, and hyperandrogenism—highlighting the importance of the MBH in regulation of fertility (81). Thus, disruption in the MBH architecture that impairs POMC and AgRP/NPY neuronal function could lead to gonadal dysfunction via disruption of their regulatory inputs to GnRH neurons (82) or could directly disrupt KNDy and GnRH neurons (Fig. 2B). Berkseth and et al (62) provide some insight into the relationship between obesity, hypogonadism, and MBH gliosis in adult men in whom longer T2 relaxation times in the MBH negatively correlated with both total and free testosterone, adjusted for participants’ age, BMI, insulin resistance, and visceral fat mass. These early data cannot assess causality but do suggest that MBH gliosis may be an independent risk factor for male hypogonadism. Future studies are needed to establish whether MBH gliosis also impacts the function of the female reproductive hypothalamic/pituitary/ovarian axis (83). Of interest, obesity-related hypogonadotropic hypogonadism in both men and women can be reversed by bariatric surgery such that plasma testosterone and even semen quality can be normalized in men (84–86) and polycystic ovary syndrome reversed in nearly all women (84). Whether MBH gliosis disrupts or resolution of MBH gliosis improves KNDy and GnRH signaling after weight loss and restoration of gonadal function remains unknown but could be tested in preclinical models.

Figure 2.

Schematic representation of the central melanocortin system and kisspeptin/neurokinin B/dynorphin neurons in the hypothalamic pituitary gonadal axis. (A) Physiologic pubertal GnRH signaling. Under permissive conditions (ie, healthy state with adequate nutritional status), the central melanocortin system comprising proopiomelanocortin (POMC) and AgRP/NPY neuronal populations provide stimulatory signaling to the kisspeptin/neurokinin B/dynorphin (KNDy) neurons, which provide stimulatory signaling downstream to the GnRH neurons. GnRH neuronal pulsatility stimulates secretion of gonadotropin hormones from the anterior pituitary gland that activates gonads and allows for normal reproductive function. (B) Theoretical disruption of GnRH signaling. Under conditions where hypothalamic gliosis is present (ie, obesity), it is possible that inflammation and gliosis disrupts normal signaling between the POMC and AgRP/NPY neuronal populations and the KNDy neurons. This could further impair GnRH neuronal pulsatility leading to impaired secretion of gonadotropins and ultimately reproductive dysfunction.

Additionally, children with overweight and obesity may also have hypothalamic–pituitary–gonadal dysfunction during puberty—a critical phase of growth and development regulated centrally by complex neuroendocrine signaling involving activation of GnRH neuronal pulse generation (87). Glial cells such as astrocytes have been found to play a key role in pubertal progression and reproduction in rodents (88). Interestingly, glial cells are sensitive to the metabolic state and participate in crosstalk with neurons involved in reproduction such that downregulation of energy-sensing receptors on glial cells leads to reproductive dysfunction in rodent models (89, 90). Currently, we lack robust translational data, but preclinical evidence and preliminary data from girls with obesity (abstract in (91)) suggest that MBH gliosis may play a role in accelerated pubertal progression. These are areas that need further investigations in humans.

Other Potential Associations

Obesity has been linked with disruption in the gut microbiome (92) and human data suggest that there is gut microbiome and brain microstructure crosstalk (18, 52). Kreutzer et al (18) demonstrated that, in 57 adults with obesity, left MBH/amygdala T2 signal ratios were negatively correlated with the presence of gut Parasutterella species. Furthermore, it was demonstrated that lower abundance of Parasutterella species was associated with greater consumption of dietary fats, which led authors to hypothesize an indirect link between dietary fats and evidence of MBH gliosis in humans, mediated by obesity-associated changes in the gut microbiome (18). Recently, a connection between dietary fats and hypothalamic microstructural characteristics was suggested in a study of maternal diet and offspring brain. Increased plasma levels of maternal saturated free fatty acid concentrations during gestation (a reflection of maternal diet and/or maternal metabolic dysfunction) were associated with higher infant hypothalamic mean diffusivity (n = 94) (93). These findings underscore that in utero exposures and early life influences, such as, microbial colonization at birth, could be involved in hypothalamic inflammation and gliosis and intergenerational transmission of metabolic risk (94, 95).

Moreover, 20 adults with obesity and 19 without obesity completed MRI using DTI and gut microbiota analysis. Researchers found that the subgroup of men with obesity had a particularly decreased bacterial biodiversity compared with female and male participants without obesity. Overall, individuals with high biodiversity had the lowest MRI measurement of R2* in the hypothalamus, a surrogate of healthy brain microarchitecture (Table 1), that held true after adjustments for age, sex, and adiposity (52). These early findings suggest crosstalk between the gut microbiome and the brain. One could speculate a link based on the role of neurons in the arcuate nucleus in sensing peripheral circulating nutrients such as glucose, free fatty acids, and proteins (96, 97). Concentrations of these circulating nutrients are affected by the gut microbiome and microbial-driven differences are likely to be transmitted to the CNS (98). However, arcuate nucleus inflammation and gliosis could also lead to inappropriate sensing of these peripheral nutrients, ultimately contributing to metabolic dysfunction. Although preliminary, these studies raise further questions about mechanistic links between gut microbiome, brain structure, and hypothalamic structure and function.

Longitudinal Associations With Adiposity Gain and Worsening Insulin Sensitivity

As the body of evidence of cross-sectional association between hypothalamic gliosis and obesity grows to include distinct populations and age ranges, the natural follow-up question is whether, in humans, MBH gliosis precedes obesity and altered glucose metabolism, as previously suggested by rodent studies (7). By studying infant hypothalamic microstructural characteristics by DTI, Rasmussen et al (93) demonstrated that infant hypothalamic mean diffusivity, acquired at a mean age of 39 days, was positively associated with body fat percentage, by DEXA, at the age of 5 years (n = 37), and results persisted even when potential confounders such as socioeconomic status, obstetric risk, and breastfeeding practices were accounted for in the analyses. In addition, evidence of MBH gliosis in children aged 9-11 years showed that, at baseline, higher MBH/amygdala T2 signal ratio was associated with increases in BMI z-score over a 1-year period, specifically in children with overweight (n = 114), not those with obesity (n = 124) (39). This could suggest that in children at risk for obesity, hypothalamic gliosis is a factor influencing the change in body adiposity over time. On the other hand, once obesity is in place, even at a young age, the hypothalamic tissue may already have reached “stability” and not be as influential for weight gain in that group of children. Taken together, these prospective associations provide a higher level of evidence linking hypothalamic tissue characteristics as assessed by MRI to excessive weight gain and future obesity development in susceptible individuals.

With regards to glucose metabolism, Rosenbaum et al (61) included a prospective cohort study of adults with obesity and normal glucose tolerance or impaired glucose tolerance (n = 38) and found that greater MBH T2 relaxation times at baseline were associated with reductions in insulin sensitivity and increases in VAT over 1 year of observational follow-up. The MBH houses the arcuate; however, other regions such as the VMH are in proximity and may have been sampled during analysis. The VMH is key region involved in regulating systemic glucose homeostasis (99), so it is possible that structural changes in this region and potentially other hypothalamic areas may also contribute to development of glucose dysregulation (73). It must be noted that there were no associations between baseline MBH T2 relaxation time and changes in oral disposition index or glucose sensitivity over 1 year; however, a longer duration of follow-up may be needed to detect changes in β-cell function or glucose intolerance. The population studied by Rosenbaum et al was limited to adults (mean age 45 years) with obesity who presumably have been exposed to chronic obesity and are thus at high risk for metabolic dysregulation (61). Whether hypothalamic gliosis plays a role in the pathogenesis of youth-onset T2D, which is more aggressive than adult-onset T2D with rapid deterioration of β-cell function and glycemic control (100–102), remains unexplored and could provide additional insights into the role of the CNS in disease pathogenesis. Regardless, these results provide early evidence that not only is hypothalamic inflammation and gliosis present in obesity, but its severity may have a direct mechanistic link to the development of comorbid metabolic diseases like T2D. Future prospective studies with longer time frames, expansion to younger age participants, and employing techniques that can isolate hypothalamic nuclei more precisely are necessary to understand the mechanistic links between disruption in hypothalamic architecture and glucose dysregulation in humans.

Treatment and Reversibility

A small proof-of-concept study attempted to reverse MBH gliosis using pioglitazone in 6 patients with T2D for 6 months; however, it was found to have a modest decrease in hemoglobin A1c from 6.7% at baseline to 6.3% by 6 months accompanied by a 3% increase in body mass and did not have any effect on MBH gliosis as measured by T2 relaxation time (103).

Unlike antihyperglycemic agents, bariatric surgery resolves T2D in a robust and sustained manner. A handful of studies have investigated if hypothalamic findings were reversed after weight loss by bariatric surgery. Van de Sande Lee et al (63) studied adult women with obesity (n = 28), some with T2D (n = 11), who underwent bariatric surgery by Roux-en-Y gastric bypass (RYGB). Nine months after surgery, lower MBH T2 relaxation times were observed while control regions did not differ. The shorter MBH T2 relaxation times in post-RYGB participants with T2D and obesity did not reach the level of healthy controls, implying that there may be some permanent underlying structural remodeling, or perhaps more time was needed for the changes to fully resolve. In a similar manner, Rullmann et al (104) and Hankir et al (105) reported on the same cohort of participants with obesity (up to n = 27) who underwent RYGB surgery and were assessed for several MRI parameters at baseline, and 6 and 12 months of follow-up. They identified reductions in MBH fractional anisotropy by diffusion imaging and in MBH/amygdala T2 signal ratio at 12 months after surgery and that T2 FLAIR and T2W/T1W signal intensity ratios increased proportionately to declines in adiposity. These technique-dependent results are challenging to interpret in the setting of simultaneous changes in whole body adiposity, diet, and glucose regulation, but indicate the possibility that RYGB may be sufficient to modify hypothalamic tissue. Moreover, changes were detected in the hypothalamic region when comparing gray matter density before surgery and 6 months after surgery. Changes in gray matter density were also seen in other brain regions important for appetite regulation, such as the lateral orbitofrontal cortex and the somatosensory cortex (104).

Conversely, in a small sample of 7 adults with severe obesity who underwent bariatric surgery (type not specified), no change in MBH/amygdala T2 signal ratio was detected with postsurgery weight loss at about 10 months of follow-up (18). Rebelos et al (106) also did not see changes in MBH/amygdala FLAIR signal ratio in 21 female participants with obesity who were studied before and 6 months after bariatric surgery (n = 7 RYGB, n = 14 sleeve gastrectomy). Importantly, authors from the last 2 publications raised study limitations such as small sample size (18) and MRI protocols that may not have been sensitive enough to identify hypothalamic changes (106).

In summary, evidence from human studies is mixed as to whether weight loss by bariatric surgery improves hypothalamic microstructural characteristics by reducing hypothalamic gliosis. Preclinical models do, however, provide evidence that bariatric surgery (107, 108) and glucagon-like peptide receptor 1 agonists (109) can have neuroprotective effects. To what extent reduction in hypothalamic gliosis is part of the causal pathway for bariatric surgery–induced weight loss and metabolic health improvement, such as glycemic control, is unknown. Furthermore, the impact of weight loss by lifestyle modification or other anti-obesity medications on hypothalamic inflammation and gliosis warrants further exploration, particularly to discern whether the efficacy of interventions is predicated upon their capacity to modify hypothalamic inflammation and gliosis. If so, it would suggest that reversing hypothalamic gliosis is a component of their mechanisms of action. Alternatively, dietary, surgical, or pharmacologic interventions could treat the manifestations of inflammation-driven dysfunction in hypothalamic regulatory circuits via cognitive control of diet (ie, induction of self-regulation) or modulating discrete appetitive signals. The latter could provide a weight loss benefit in the absence of any regression of underlying hypothalamic inflammation and gliosis. Similarly, the relevance of hypothalamic inflammation and gliosis to weight loss maintenance after weight loss is a critical component toward understanding whether permanent hypothalamic structural changes underlie the biologic defense of elevated body fat mass in obesity (110).

Implications for Neurophysiology

Patients suffering from severe obesity and hyperphagia secondary to craniopharyngioma, other suprasellar tumors, or treatment options to these and other pathologies, provide a clear link between hypothalamic tissue damage and impaired regulation of energy homeostasis (111, 112). Future brain imaging studies on unique populations with monogenic forms of obesity (melanocortin 4 receptor or POMC mutations) or other genetic obesity syndromes (Prader–Willi syndrome) could provide additional insight into the role of MBH gliosis and obesity pathogenesis. As evidence for microstructural alterations in the hypothalamus in humans accumulates, a further critical question arises as to whether such changes have implications for hypothalamic function like that observed due to macrostructural damage. A small number of cross-sectional studies used multimodal MRI approaches to address that question.

In Sewaybricker et al (113), brain regulation of feeding was investigated in 111 adults, the majority with overweight or obesity (76%), through resting-state connectivity within the salience network (SN). The SN is a high-order brain network with anatomically distant regions that work in synchrony to detect external environmental cues and monitor internal homeostatic signals (eg, hypothalamus) to coordinate human behavior, including feeding. It was demonstrated that obesity itself did not impair the expected reduction in SN connectivity after the consumption of a standardized meal. However, when comparing participants based on the degree of radiologic evidence of MBH gliosis (ie, their values of MBH T2 relaxation time [low vs high tertiles, n = 29 per group]), the group with more evidence of MBH gliosis, independent of body fat percentage, showed sustained SN connectivity after a meal. This suggests that MBH gliosis may impair the appropriate, coordinated brain response within the SN after a meal, which could result in a failure to reduce attentional resources directed at environmental food cues.

A dysfunctional blood oxygen level-dependent (BOLD) signal response in the hypothalamus after oral (114) and intragastric (115) glucose load has been suggested in adults with obesity when compared with healthy weight counterparts and is partially reversed after weight loss by bariatric surgery (114). In a combined sample of participants with obesity and healthy weight (n = 18) and mean age of 13 years (37), T2 relaxation time from MBH was acquired and compared with BOLD signal change from pre-oral and postoral glucose load (1.75 g/kg up to 75 g) used to calculate signal change. Overall, there was an increase in the hypothalamic BOLD signal after glucose intake. Combining both methodologies, evidence of MBH gliosis was associated with a decrease in the postglucose hypothalamic BOLD signal. Interestingly, this association was only found for the left side of the MBH (37). Stronger associations between MBH gliosis and body adiposity have been suggested for the left side in adults (7, 18), though not consistently (19, 53, 54, 61). These findings suggest that (1) depending on the degree of evidence of hypothalamic gliosis there could be functional consequences on hypothalamic glucose processing, and (2) the response to nutrients may be a lateralized function in the hypothalamus. The latter has been suggested for rodents (43) and more broadly demonstrated in parasympathetic activity in adults (42) but still awaits further investigation. While the simultaneous application of structural and functional hypothalamic brain imaging in the current literature is still in its early stages, evidence so far supports the hypothesis that hypothalamic gliosis is associated with dysregulation of responses to meals and/or oral glucose loads.

Etiology and Potential Implications of Hypothalamic Gliosis in Humans

The majority of studies reviewed were cross-sectional and are thus not informative about exposures that may have precipitated the development of hypothalamic gliosis. Extensive preclinical literature (reviewed in (13, 14)) shows that dietary factors consistent with a hypercaloric, obesogenic diet induce inflammation and gliosis in the MBH (7, 10, 14, 17, 44, 45, 116–118) even before the development of obesity (7, 116). In rodents, hypercaloric diets high in saturated fat (46), fructose (119), and saturated fat in combination with simple carbohydrates (120) are all robust and rapid dietary stimulators of inflammatory glial cell responses in the arcuate nucleus. Typical HFD regimens are hypercaloric, of high hedonic value, and high in saturated fat, and the relative contributions of these distinct aspects of HFD feeding are incompletely parsed. It appears that saturated fats are a contributing factor, as they are readily taken up into the hypothalamus within hours of enteral feeding, whereas monounsaturated fats do not stimulate arcuate nucleus microglial reactivity (46). A direct comparison of high-fat, high-carbohydrate and high-fat, low-carbohydrate diets in rodents showed that the combination of high saturated fat with high carbohydrate stimulated significantly more arcuate inflammatory response (120). Astrocyte response, in particular, was greater when sucrose was used as the carbohydrate source rather than starch (120). Thus, in preclinical studies, arcuate nucleus gliosis is stimulated by diet, but such findings await translation to humans via epidemiologic or experimental feeding study approaches.

To understand the implications of hypothalamic gliosis for disease development at a mechanistic level, it is helpful to look to other instances of CNS cellular inflammation. In the CNS, reactive gliosis is usually initiated as a neuroprotective response (121) to insults ranging from ischemia to autoimmune disease. In brain injury and inflammatory processes, sustained glial cell activation ultimately contributes to neurotoxicity (122). In other words, when insults are mild, the reactive changes are neuroprotective and normal neuronal function can be restored once the localized inflammation resolves. However, insults that are prolonged or severe result in glial cell hypertrophy and glial scar formation. If a glial scar occurs, normal neuronal tissue functions such as axonal outgrowth, dendrite formation, and synaptic remodeling can be negatively affected (122). Of note, the hypothalamus and arcuate nucleus are brain regions in which neurogenesis and neuronal turnover are normal in adulthood—processes which could be affected by extensive glial cell remodeling and are, in fact, suppressed by HFD feeding in mice (123). Moreover, neuron–glia interactions and glial cells themselves play increasingly recognized roles in regulating energy balance and peripheral glucose homeostasis (70, 116). Thus, to the extent that gliosis interferes with these, and other, cellular and tissue functions, the presence of gliosis could significantly hamper maintenance of normal energy balance and glucose homeostasis.

Theoretical Models for a Role of Hypothalamic Inflammation and Gliosis in Obesity Pathogenesis

Based on preclinical studies and the above literature involving human subjects, we put forth 3 possible theoretical models of the role of hypothalamic gliosis in obesity pathogenesis (Fig. 3). In the adaptive model (Model A), dietary patterns considered to be obesogenic stimulate arcuate nucleus inflammation and cellular inflammatory responses among glial cell populations. This inflammatory response, which could be modified by genetic risk factors, results in temporary or reversible neuronal dysfunction (14, 45, 124) within the arcuate nucleus. While such changes in neuronal function would seem pathologic, they could also be viewed as adaptive in that they facilitate increased consumption and energy storage during times of nutritional abundance (14). In the adaptive model, reactive gliosis resolves when episodes are acute and self-limited, which would functionally restore homeostatic signaling, although potentially at a higher level of acquired energy stores. The neuron injury model (Model B, Fig. 3) posits that, among genetically susceptible individuals, obesogenic diets initiate a pathologic inflammatory reaction within the arcuate nucleus. This results in a condition of neuronal dysfunction and resultant hyperphagia that tips the individual toward progressive weight gain and increased adiposity. Finally, the chronic disease model (Model C, Fig. 3) puts forth that hypothalamic gliosis begins as an adaptive process, as in Model A. However, severe or repeated exposure to excess caloric consumption within an obesogenic food environment subsequently, in susceptible individuals, evokes sufficiently severe cellular inflammatory responses in the arcuate nucleus to eventually cause glial scarring, neuronal dysfunction, and, conceivably, even neuron loss. In the chronic disease model, what is initially a neuroprotective or adaptive response results in sufficient tissue remodeling by glial cell infiltration to significantly and permanently hinder neuronal function within this critical region for energy and glucose homeostasis.

Figure 3.

Theoretical models of the role of hypothalamic gliosis in obesity pathogenesis. (A) Adaptive model. Hypercaloric, obesogenic diets induce hypothalamic inflammation accompanied by dysfunction of local neuron and glial cell populations. This process facilitates expansion of energy stores when nutrition is abundant, but normal hypothalamic function returns, albeit at a higher defended level of adiposity. (B) Neuron injury model. Hypercaloric, obesogenic diets induce hypothalamic dysfunction that results in abnormal appetite regulation, persistent hyperphagia, and increased adiposity. (C) Chronic disease model. Chronic or repeated exposure to hypercaloric, obesogenic diet stimuli eventually create permanent glial scars or tissue remodeling that impairs Normal homeostatic signaling, promotes further weight gain, and may also resist weight loss.

One extension of these theoretical models is that circuits that are under regulatory control by the hypothalamic melanocortin system could also become dysregulated, resulting in insensitivity of hindbrain-mediated satiety signaling (125) or inappropriately heightened hedonic drive (126–128). This implication is consistent with observations that weight gain precedes changes in responsiveness of reward-regulating brain regions (129). Moreover, evidence in animal models suggests that our most potent pharmacologic interventions for obesity have their primary site of action in the arcuate nucleus, where they enhance orexigenic POMC/cocaine- and amphetamine-regulated transcript signaling (124), reduce microglial expansion, and improve leptin sensitivity (45). These pharmacologic studies further support a role for hypothalamic gliosis in obesity pathogenesis by showing that diet-induced obesity improves when arcuate nucleus cellular inflammation is reduced and neuronal function restored.

Each of these models is in line with clinical observations of obesity as a chronic disease (110) which remits and responds to treatment but does not resolve. Each also incorporates a key tenet of obesity pathogenesis, which is that it involves an increase in the defended level of body adiposity (110). In the adaptive model (Model A, Fig. 3), adiposity increases are permissive. In the neuron injury model (Model B, Fig. 3), they are pathologic. In the chronic disease model (Model C, Fig. 3), they are progressive—eventually becoming permanent with accompanying dysfunction. Questions these models raise include: Does hypothalamic gliosis play a role in normal periods of adaptive weight gain, such as during pregnancy? In the presence of hypothalamic gliosis, are nutritional and energy expenditure challenges to energy homeostasis met in a manner that suggests a properly functioning homeostatic regulatory system? Is gliosis fully reversible in humans?

Summary and Future Directions

In sum, existing preclinical evidence strongly supports that dietary factors consistent with an obesogenic diet induce inflammation and structural changes in the arcuate nucleus of the hypothalamus, and that these cellular inflammatory changes are mechanistically related to hyperphagia and weight gain in diet-induced obesity (3, 11, 46, 130). As summarized in the current review, translational studies using MRI or postmortem histology reliably and reproducibly show evidence of gliosis in obesity and comorbid metabolic conditions, suggesting that gliosis may play a role in obesity pathogenesis in humans as demonstrated in rodent models. While obesogenic diet and excess caloric consumption are implicated in obesity pathogenesis in humans (110), it remains uncertain whether hypothalamic gliosis plays a mechanistic role in diet-induced weight gain in humans. Available data in humans has linked hypothalamic gliosis to specific reduced intestinal microbial communities, which were in turn associated with HFD (18), indirectly implicating dietary fat intake in the pathogenesis of hypothalamic gliosis in humans. Given that epidemiologic evidence regarding the influence of specific dietary components on human obesity is mixed, the particular dietary exposures in humans that elicit gliosis may be complex. In the absence of experimental evidence in humans, the topic is likely to remain an area of spirited debate (131–133).

Future directions include parsing the effects of gliosis on the function of specific neuronal cell populations (and neuron–glia interactions), for which preclinical studies will be required. Genetically determined variability in gliosis is another area requiring investigation. Documenting reversibility by lifestyle change, diet, obesity therapeutics, including the highly effective novel multifunctional incretin mimetics (134), and bariatric surgery is critical to better understand the contribution of resolution of hypothalamic gliosis to weight loss and improved metabolic health. Moreover, given that 1 study reviewed herein found that radiologic evidence of gliosis incompletely resolves after bariatric surgery in humans (63), these structural changes could be permanent and might mechanistically contribute to the raising of the defended level of adiposity in obesity (110). Novel interventions to delay, reduce, or reverse hypothalamic gliosis should also be sought. In conclusion, the accumulated translational evidence in humans provides ample support for additional clinical and basic science studies to discern whether obesity and obesity-associated health conditions are etiologically related to structural remodeling of the hypothalamus.

Abbreviations

AgRP

Agouti-related protein

BMI

body mass index

BOLD

blood oxygen level-dependent

CNS

central nervous system

DEXA

dual-energy X-ray absorptiometry

DTI

diffusion tensor imaging

GFAP

glial fibrillary acidic protein

GnRH

gonadotropin-releasing hormone

HFD

high-fat diet

HOMA-IR

homeostatic model assessment of insulin resistance

Iba1

ionized calcium-binding adapter molecule 1

KNDy

kisspeptin/neurokinin B/dynorphin

MBH

mediobasal hypothalamus

MRI

magnetic resonance imaging

NPY

neuropeptide Y

POMC

pro-opiomelanocortin

RYGB

Roux-en-Y gastric bypass

SN

salience network

T2D

type 2 diabetes

VAT

visceral adipose tissue

VMH

ventromedial nucleus of the hypothalamus

Financial Support

This work was supported by grants from the National Institutes of Health (NIH) R01DK117623 (E.A.S.), R01DK089036 (E.A.S.), K24HL144917 (E.A.S.), University of Washington (UW) Nutrition and Obesity Research Center (P30DK035816), and P30DK017047-45S2 (A.H.), and the UW Diabetes Research Center (NIH grant P30DK017047) Pilot and Feasibility Award to A.H.

Disclosure

The authors declare no conflicts of interest.