Hypothalamic Gliosis Associated With High-Fat Diet Feeding Is Reversible in Mice: A Combined Immunohistochemical and Magnetic Resonance Imaging Study

Abstract

Gliosis, the activation of astrocyte and microglial cell populations, is a hallmark of central nervous system injury and is detectable using either immunohistochemistry or in vivo magnetic resonance imaging (MRI). Obesity in rodents and humans is associated with gliosis of the arcuate nucleus, a key hypothalamic region for the regulation of energy homeostasis and adiposity, but whether this response is permanent or reversible is unknown. Here we combine terminal immunohistochemistry analysis with serial, noninvasive MRI to characterize the progression and reversibility of hypothalamic gliosis in high-fat diet (HFD)-fed mice. The effects of HFD feeding for 16 weeks to increase body weight and adiposity relative to chow were nearly normalized after the return to chow feeding for an additional 4 weeks in the diet-reversal group. Mice maintained on the HFD for the full 20-week study period experienced continued weight gain associated with the expected increases of astrocyte and microglial activation in the arcuate nucleus, but these changes were not observed in the diet-reversal group. The proopiomelanocortin neuron number did not differ between groups. Although MRI demonstrated a positive correlation between body weight, adiposity, and the gliosis-associated T2 signal in the mediobasal hypothalamus, it did not detect the reversal of gliosis among the HFD-fed mice after the return to chow diet. We conclude that hypothalamic gliosis associated with 16-week HFD feeding is largely reversible in rodents, consistent with the reversal of the HFD-induced obesity phenotype, and extend published evidence regarding the utility of MRI as a tool for studying obesity-associated hypothalamic gliosis in vivo.

The pathogenesis of obesity, a leading cause of morbidity and mortality in affluent nations (1), involves dysfunction of the homeostatic system responsible for maintenance of stable fat mass over time. Energy balance and adiposity are regulated by the action of negative feedback signals, such as the adipocyte-derived hormone leptin, that act in the hypothalamus and other central nervous system (CNS) areas (2). The increased susceptibility to weight gain in the modern environmental milieu is characterized by the failure of leptin and other anorexigenic signals to appropriately constrain adiposity, reflecting an acquired reduction of CNS sensitivity to leptin (leptin resistance) and other signals of peripheral energy status (2–4).

A potential mechanism to explain the impaired hypothalamic response to adiposity negative feedback signals is based on evidence that obesity is also associated with a cluster of pathological responses localized to hypothalamic areas involved in energy homeostasis, including the arcuate nucleus (ARC). These responses include increased inflammatory signaling and endoplasmic reticulum stress (5–7), reduced neuronal sensitivity to anorexigenic signals such as leptin (3, 4), diminished neurogenesis (8), an eventual loss of anorexigenic proopiomelanocortin (POMC) neurons (9), and a reactive gliosis characterized by the activation of local astrocyte and microglial populations (9–12). Several independent lines of investigation suggest that CNS leptin resistance is a causal factor in obesity pathogenesis and that hypothalamic inflammatory signaling contributes to both leptin resistance and increased adiposity in rodents (5, 6, 13, 14). These hypothalamic consequences of obesity induced by a high-fat diet (HFD), therefore, may contribute to the development and maintenance of obesity, at least in rodent models.

Astrocytes and microglia play key roles in CNS function, including in the CNS response to injurious stimuli such as stroke, physical trauma, tumors, hypoxia, infection, excitotoxicity, and neurodegenerative disease (15). In these settings, microglia and astrocytes enter a reactive state, morphologically characterized by increases of cell size and number and the proliferation and enlargement of cellular processes (15). In rats and mice, reactive gliosis in the ARC, including both astrocyte and microglial activation, is detectable within as little as 1 week of ad libitum HFD feeding and becomes more pronounced after several months of obesity (9).

On magnetic resonance imaging (MRI), gliosis is characterized by increased signal intensity (brightness) on T2-weighted images. When applied in the clinic in the diagnosis of major CNS injury such as stroke, this T2 hyperintensity is visually apparent, but more subtle changes in T2 signal are also detectable using quantitative analysis of magnetic resonance images (16–18). Because quantitative MRI techniques have recently detected evidence for obesity-associated hypothalamic gliosis in obese mice and humans (9, 19), this approach may enable noninvasive monitoring of hypothalamic gliosis over time in experimental animals and humans.

Previous studies suggest that switching obese HFD-fed rodents to standard rodent chow results in a loss of most excess adiposity (20, 21), but whether hypothalamic gliosis is also reversible in this setting is an important unanswered question. Here we report that returning markedly obese HFD-fed mice to an ad libitum standard chow diet results in substantial reductions in body weight and adiposity that are accompanied by a near-complete normalization of reactive gliosis in the ARC.

Materials and Methods

Animals and dietary protocols

All studies were performed on adult male mice bred onto the C57BL/6J background. Animals were maintained in a temperature- and humidity-controlled room on a 12-hour light, 12-hour dark cycle and were housed and cared for in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. All protocols, animal handling, and treatments were approved by the Institutional Animal Care and Use Committee of the University of Washington.

Protocol A: histochemical analysis of hypothalamic gliosis

For this study, male C57BL/6 POMC-tau-green fluorescent protein (GFP) mice (22) were used because the localization of GFP to POMC cells of these animals enables reliable postmortem quantitation of POMC cell number. Mice were assigned to one of three diet groups: semipurified HFD (60% kilocalories from fat; Research Diets D12492) (HFD group, n = 9) for 20 weeks; HFD for 16 weeks followed by the switch to standard chow for 4 weeks (12% kilocalories from fat; LabDiet 5001) (HFD-CHOW group, n = 12); or standard laboratory chow for 20 weeks (CHOW group, n = 10). For the first 16 weeks, therefore, both HFD and HFD-CHOW groups were maintained on the same HFD, but for the last 4 weeks, the HFD-CHOW and CHOW cohorts were fed standard chow (Figure 1A). Mice were given ad libitum access to their assigned diet throughout the study. Body weight and food intake were recorded weekly. During weeks 16 and 20, mice underwent body composition analysis. Terminal perfusion was performed after body composition analysis at the 20-week time point.

Figure 1.

Protocol schematic and representative magnetic resonance images. A, Protocol schematic. C57BL/6 male mice (n = 8–12/group) were initially placed on standard chow for 16 weeks followed by 4 additional weeks of chow (CHOW group), 16 weeks of HFD followed by 4 additional weeks of HFD (HFD group), or 16 weeks of HFD followed by 4 weeks of Chow (HFD-CHOW group). B, High-resolution, 2D image of mouse brain (protocol B) from a slice selected to include the midregion of the ARC within the MBH with representative ROIs including right and left cortex (Co-R, Co-L), thalamus (Th-R, Th-L), and MBH (MBH-R, MBH-L). Scale bar, 1 mm.

Protocol B: MRI analysis of hypothalamic gliosis

Three cohorts of C57BL/6 were maintained on the same regimen as above, with the following exceptions: 1) animals were single housed throughout; 2) animals consuming the HFD at study onset were treated as one cohort (n = 16) for the first 16 weeks of the study and then randomized to two groups at week 16 (HFD, n = 8; HFD-CHOW, n = 8); 3) group numbers differed (n = 8 for each group HFD, HFD-CHOW, and CHOW); and 4) magnetic resonance imaging of the brain and body composition (separated by 2 d to allow animals to recover) were performed at baseline and at weeks 1, 4, 16, and 20.

Body composition analysis

In vivo analysis of body lean mass, fat mass, and water content was performed in conscious, restrained mice by nuclear magnetic resonance (EchoMRI 3-in-1 animal tissue composition analyzer; Echo MRI) (23).

Immunohistochemistry

Animals were perfused and fixed with cold PBS followed by 4% paraformaldehyde/PBS. Brains were collected and cryoprotected using 25% sucrose in PBS, flash frozen using dry ice-cooled isopentane, and 14-μm-thick frozen sections were generated in the coronal plane through the hypothalamus (at the level of the medial ARC; bregma −2.8 to −3.1 mm). For fluorescent staining, sections were blocked in 5% normal goat or donkey serum (Jackson ImmunoResearch Laboratories Inc) and incubated overnight at 4°C with mouse Cy3-conjugated antiglial fibrillary acidic protein (GFAP) to visualize astrocytes (1:10 000, C9205; Sigma-Aldrich); rabbit anti-ionized calcium binding adapter molecule 1 (IBA1) to visualize microglia (1:1000, 019–19741; Wako); or goat anti-GFP (1:1000, 70R-GG001; Fitzgerald). Immunofluorescence was performed with a combination of AlexaFluor 488- or AlexaFluor 594-labeled antirabbit or antigoat secondary antibodies (1:1000; Invitrogen). We observed no overlap between IBA1 and GFAP staining.

Immunohistochemistry image capture and analysis

Images were captured on an Eclipse E600 upright microscope equipped with a color digital camera (Nikon). All images were captured in JPEG format using the same capture settings and were not subjected to any form of image processing before quantification. To optimize image quantification, representative images displayed in figures were subjected to contrast enhancement that was applied identically to all images. For all immunohistochemical analyses, quantification was performed in a blinded fashion on anatomically matched brain regions (bregma −2.8 to −3.1 mm). Both sides of bilateral structures (eg, ARC, ventromedial hypothalamic nucleus, etc) were counted on four brain sections (and eight ARC regions) per animal, and replicate values from each animal were averaged before determining group means. POMC cell number was counted manually using Photoshop (Adobe), and the microglial number was counted manually within a prespecified region of interest (24) in ImageJ (National Institutes of Health, Bethesda, Maryland; http://rsbweb.nih.gov/ij/). Mean GFAP staining intensity was quantified within a prespecified region of interest (ROI) approximating the area of the ARC (19). Mean background GFAP staining intensity was also quantified in a 30-μm² region of the ARC that lacked detectable astrocyte labeling, and this mean background value was subtracted from the mean ARC GFAP staining intensity measure for each image.

Microglial activation was determined using a blinded semiquantitative scoring system assigning a value of 0–5 based on morphological characteristics of ARC microglia, as follows. IBA1-labeled slides were examined in Photoshop by an investigator blinded to study condition, and each ARC image was assigned a value between 0 and 5 using the following criteria: 0, completely resting cell morphology; 1, slightly increased process thickness or process number; 2, moderately increased process thickness or process number; 3, moderately increased process thickness and process number; 4, moderately increased process thickness, process number, and increased cell body size; and 5, substantially increased process thickness, process number, and increased cell body size. Values for all eight ARC images per mouse were averaged into a single value for each animal prior to statistical analysis. This method was modified from similar, widely used methods (25, 26) as a sensitive measure of overall microglial activation due to its integration of multiple aspects of microglial morphology that are not captured by simply counting numbers of cells.

MRI procedures

As described previously (19), mice underwent isoflurane anesthesia in an induction chamber. Once in an appropriate plane of anesthesia, eye lubricant was applied, the mice were secured on a bite bar, and their heads were placed into a radiofrequency coil and secured to a cradle created specifically for the MRI system. The coil was then inserted vertically into a scanner heated to maintain thermoneutrality (32°C). The coil is equipped with an adjustable anesthetic flow and vacuum system to maintain sedation throughout the experiment. Total scan time was 45–120 minutes, during which respiration was monitored through a respiration sensor under the abdomen (SA Instruments) and anesthesia titrated to ensure appropriate sedation. After the imaging paradigm (described below), the mice were removed from the coil, given sc saline support (10 mL · kg⁻¹ · h⁻¹ of anesthesia), and were allowed to recover in their home cage.

MRI protocol

High-resolution MRI acquisitions were performed on a 14T Avance 600 MHz/89-mm-wide bore vertical MR spectrometer (Bruker BioSpin) using a 25-mm inner diameter 1H birdcage coil. A rapid acquisition with refocused echoes (RARE) sequence was used to acquire two-dimensional (2D) multislice cross-sectional images covering the entire mouse brain. We focused on the hypothalamus in the acquired 2D multislice RARE images to obtain high-resolution 2D multislice RARE images. The slice of interest was matched across animals and across time points for each animal with serial imaging. The slice selection was guided by the mouse brain atlas of Paxinos and Franklin (27) to include the midregion of the ARC, located in the mediobasal hypothalamus (MBH) (between −1.46 and −1.82 mm from bregma). Once the optimal slice was identified, a transverse relaxation time T2 was acquired using a single-slice, multiecho sequence. A T2 map was created by calculating the T2 values using an exponential fit function, signal intensity at echo time t, S(t) = S₀ × e^−t/T2 (Figure 1B). Acquisition parameters for all sequences are previously published (19).

Magnetic resonance image analysis

All images were analyzed using Paravision 5.1 software (Bruker BioSpin). The ROIs in the bilateral MBH, thalamus, and cortex (Figure 1B) were first placed on a signal intensity display of the T2 image prior to image processing for best identification of anatomy. Then image parameters were switched to the T2 relaxation time, without the ROIs being moved from their original placement. ROI shape and size was maintained between animals. Unilateral 2D ROI areas for T2 relaxation time were 0.20 mm² for the MBH, 0.49 mm² for the thalamus, and 0.22 mm² for the cortex. Mean ROI T2 relaxation time values and SDs were recorded. The image analyst was blinded to diet group and time point during image analysis. After the image analysis, we determined the mean values for T2 relaxation time at each ROI (mean of right and left sides), and these bilateral means are reported throughout the manuscript.

Statistical analysis

Group means and SE were determined for body composition variables. Unpaired Student's t tests or one-way ANOVA with Tukey posttests were used to test for group differences in body weight, body composition, and histopathological measurements. Within the magnetic resonance imaging data set, the interquartile range (IQR) was used to identify outliers. The IQR was calculated as the third quartile (Q3) minus the first quartile (Q1). T2 values falling below Q1–1.5 IQR or above Q3+1.5 IQR for any given time point were treated as outliers and excluded from the analyses. Two-way ANOVA with Fischer's posttests were used for analyses that examined the effect of diet (HFD vs chow) and region (thalamus, cortex, MBH) on magnetic resonance-derived outcomes and the interaction of those factors. When assessing overall T2 relaxation time vs body weight and body fat mass, linear mixed-effects models with per-mouse random intercepts were used to account for repeated measures (28, 29). R-squared values were derived from linear models. Results were considered significant at P < .05. All statistical analyses were performed with GraphPad Prism version 6.0 and Stata version 12.1 (StataCorp).

Results

Effects of diet on food intake, body weight, and adiposity

To expand upon our previous finding of gliosis with eventual POMC neuron loss in long-term HFD-fed mice, we studied cohorts of transgenic mice expressing a tau-GFP fusion protein in POMC neurons (POMC-tau-GFP mice). Mice were randomized into three cohorts receiving different dietary exposure. Group CHOW received unrefined rodent chow for 20 weeks; group HFD received HFD for 20 weeks; group HFD-CHOW received HFD for 16 weeks and chow for 4 weeks. As expected, mean body weights of the HFD and HFD-CHOW groups were significantly greater than the CHOW group between 4 and 16 weeks (P < .05) but were not significantly different from one another during that period (Figure 2A). Changes of adiposity paralleled changes of body weight, such that at 16 weeks (prior to the diet switch), HFD and HFD-CHOW groups exhibited approximately 5-fold greater fat mass than CHOW (12.5 and 11.0 g vs 2.3 g, respectively; P < .01) (Figure 2B). In the second phase of the study (from wk 17 to 20, when the HFD-CHOW group was placed on ad libitum chow, whereas the other two groups remained on their original diet), HFD mice continued to gain weight at a greater rate than the CHOW group, whereas the diet-reversal group (HFD-CHOW) rapidly lost weight, nearly converging on the weight of the CHOW group by week 20 (CHOW 26.9 ± 0.6 g, HFD-CHOW 29.3 ± 1.2 g, P = NS) (Figure 2A). Weight loss in the HFD-CHOW group was accompanied by a large but transient reduction of food intake (Figure 2C) and a marked decrease of body fat mass (from 11.0 ± 1.3 g to 3.5 ± 0.5 g, P < .01) to a value that by week 20 was only slightly and nonsignificantly above that of the CHOW group (CHOW 2.4 ± 0.1 g; P = NS) (Figure 2B). In contrast, the fat mass of the HFD group tended to increase during this time (from 12.5 ± 1.6 g to 14.8 ± 1.3 g, P = NS).

Figure 2.

Body weight, fat mass, and food intake of CHOW, HFD, and HFD-CHOW groups, protocol A (n = 10, n = 9, n = 12, respectively). A, Weekly body weight from study onset to 20 weeks. B, Body composition at 16 and 20 weeks. C, Weekly energy intake from study onset to 20 weeks. Mice were group housed, and each cage represented n = 1 for statistical analysis of energy intake. Statistical analyses were performed using one-way ANOVA with Tukey posttest at each time point. *, P < .05 CHOW vs HFD and HFD-CHOW; †, P < .05 for all comparisons; ‡, P < .05 for HFD vs CHOW and HFD-CHOW; +, P < .05 CHOW vs HFD; #, P < .05 CHOW vs HFD-CHOW; θ, P < .05 HFD-CHOW vs HFD and CHOW.

Effect of diet on immunohistochemical markers of gliosis

To determine the impact of the switch from HFD to chow on hypothalamic glioisis, we used immunohistochemistry to assess astrocyte and microglial activation at the 20-week time point (protocol A). GFAP staining intensity, a measure of astrocyte activation, was elevated by nearly 50% in the ARC of HFD relative to the CHOW and approximately 40% relative to the HFD-CHOW group (P = .016 overall; P < .05 for pairwise comparisons), whereas values were not significantly different between CHOW and HFD-CHOW (Figure 3, A–D). Whereas the ARC microglial number did not differ between groups (Figure 4C), morphological evidence of ARC microglial activation was increased by approximately 110% and approximately 60% in the HFD relative to the CHOW and HFD-CHOW groups, respectively (P < .001 overall; P < .01 for pairwise comparisons), with no significant difference between CHOW and HFD-CHOW (Figure 4, A–C and E). Specifically, microglia in the ARC of the HFD-fed mice exhibited larger cell bodies and thicker, more ramified processes than were detected in the other two groups (Figure 4, A–C). POMC neuron number did not differ between groups (P = .24; Figure 5, A–D).

Figure 3.

Reversibility of HFD-induced ARC astrocyte activation. A–C, Representative images of ARC GFAP immunostaining in coronal sections through the same MBH region from CHOW, HFD, and HFD-CHOW mice. D, Quantification of GFAP staining intensity in ARC from CHOW, HFD, and HFD-CHOW mice (n = 10, n = 6, n = 12, respectively). Scale bars, 100 μm. Statistical analysis was performed using one-way ANOVA with Tukey posttest. *, P < .05. 3V, third ventricle.

Figure 4.

Reversibility of HFD-induced ARC microglial activation. A–C, Representative images of ARC IBA1 immunostaining in adjacent coronal sections through the MBH from CHOW, HFD, and HFD-CHOW mice. D, Quantification of microglial number in the ARC of CHOW, HFD, and HFD-CHOW mice. E, Quantification of microglial activation score (0–5) in ARC of CHOW, HFD, and HFD-CHOW mice (n = 10, n = 6, n = 12, respectively). Scale bars, 100 μm. Statistical analysis was performed using one-way ANOVA with Tukey posttest. **, P < .01. 3V, third ventricle.

Figure 5.

POMC cell number after 20 weeks of HFD feeding. A–C, Representative images of GFP immunostaining in coronal sections through the MBH from CHOW, HFD, and HFD-CHOW mice expressing tau-GFP under the POMC promoter. D, Quantification of POMC cell number in CHOW, HFD, and HFD-CHOW mice (n = 10, n = 6, n = 12, respectively). Scale bars, 100 μm. Statistical analysis was performed using one-way ANOVA with Tukey posttest.

MRI analysis of hypothalamic gliosis

The cohort of the C57BL/6 mice used for MRI analysis (protocol B) exhibited similar trends of body weight and body composition when placed on the same feeding regimen (Supplemental Figure 1). Serial MRIs of the MBH were performed in this cohort at baseline and at weeks 1, 4, 16, and 20 to assess changes in MBH T2 relaxation time in the setting of HFD feeding.

Consistent with our previous data (19), the analysis of MRI data across all time points (ie, 0–20 wk were included and repeated measures were accounted for statistically) and including all groups demonstrated a small but significant correlation between body weight and MBH T2 relaxation time such that higher body weights were associated with higher MBH T2 relaxation time (P = .01, r = 0.22; Figure 6A). Similarly, body fat mass was also positively correlated with MBH T2 relaxation time (P = .02, r = 0.19; Figure 6B). These relationships between the T2 signal and body weight and composition were not present in the thalamus (weight: P = .56; fat mass: P = .67) and were reversed in the cortex such that increased body weight and fat mass were associated with lower T2 signal within the cortex ROI (weight: P = .01, r = −0.23; fat mass: P = .02, r = −0.21).

Figure 6.

Radiological evidence for relationships between MBH gliosis and body weight, adiposity, and exposure to HFD (protocol B). When considering all animals at all time points and correcting for repeated measures, mean bilateral MBH T2 relaxation time is positively correlated with body weight (A) (P = .01) and fat mass (P = .02 (B). C, At week 0 (before initiation of diet intervention), there was no interaction between diet and region (P = .37). D, At week 16, there was a strong trend toward an interaction of diet by region [F (2, 42) = 3.14; P = .05], driven by increased T2 relaxation time in HFD as compared with chow-fed animals within the MBH (t = 2.16; P = .03) but not in the control regions (thalamus, P = .80; cortex, P = .26). E, There was no significant difference in the change in the mean T2 between the HFD and HFD-CHOW groups during the reversal period from 16 to 20 weeks (P = .68). There was also no effect of treatment group during the reversal period [F (2, 20) = 1.84; P = .18]. *, P < .05. gm, grams; ms, milliseconds.

There were no significant differences in T2 relaxation time at baseline between treatment groups at any ROI (Figure 6C), nor were differences detected after 1 or 4 weeks of HFD feeding (Supplemental Figure 2, A and B). At 16 weeks (just prior to diet reversal), there was a strong trend toward an interaction of diet by region, consistent with evidence that changes in tissue were specific to the MBH during chronic HFD exposure (interaction F (2, 42) = 3.14; P = .05). Although the actual difference in the means was small, post hoc testing revealed that T2 relaxation time was increased in HFD as compared with chow-fed animals within the MBH (P = .03) but not in control regions (thalamus, P = .80; cortex, P = .26; Figure 6D). However, despite immunohistochemical evidence of reversal of astrocytosis and microglial accumulation from the POMC-GFP cohort (Figures 2 and 3), MRI did not detect significant reversal of T2 changes during the diet reversal period (Figure 6E).

Discussion

In this study, we investigated whether hypothalamic gliosis associated with HFD feeding is reversible in mice. Our results confirm previous observations that obesity induced by HFD feeding is accompanied by the activation of astrocyte and microglial populations in the ARC (9, 19) and demonstrate for the first time that this phenotype is reversible when mice are returned to an ad libitum regimen of unrefined chow (30). Because astrocyte and microglial activation are among the most sensitive markers of obesity-associated hypothalamic injury, our findings suggest that in mice, the glial response to ARC injury can be reversed by consuming a lower-fat, less calorically dense, less palatable diet and that this effect is associated with normalization of body weight and fat mass.

Previous evidence of progressive gliosis and neuron injury observed in the ARC of obese rodents suggests that hypothalamic circuits become increasingly compromised over time, resulting after 8 months of HFD feeding in a significant decrease in the number of POMC neurons (9), which protect against obesity during HFD feeding. Our finding in the current study that POMC cell number was not reduced at 20 weeks of HFD feeding suggests that if POMC cell loss is indeed a consequence of diet-induced obesity (DIO), it is a late event and does not contribute to the development of obesity in this model. This hypothesis is compatible with the finding that in the current study, obesity was almost completely reversible upon the return to unrefined chow. This analysis leaves open the possibility that longer durations of HFD feeding induce more significant and potentially irreversible consequences (eg, loss of POMC neurons) that in turn prevent the reversal of obesity after the switch to chow, and studies to test this hypothesis are warranted.

A key unanswered question is the extent to which our current findings pertain to human obesity. In humans, obesity pathogenesis appears to involve an increase of the biologically defended level of adiposity. This hypothesis explains why maintenance of reduced weight is actively resisted, a phenomenon that represents perhaps the greatest barrier to effective obesity treatment in the clinical setting (2). This hypothesis also explains why in obese patients weight loss triggers the same neuroendocrine changes that favor weight regain in lean individuals, even when body weight remains above normal (3, 24, 31). Stated differently, the energy homeostasis system of weight-reduced, formerly obese individuals is activated in much the same way as occurs after weight loss in normal-weight individuals, and it is this response that drives recovery of lost weight, whether one starts out being lean or obese. Although the mechanism(s) responsible for this upward resetting of the defended level of body fat mass in obese individuals remains to be established, resistance of key hypothalamic energy balance circuits to input from leptin and other peripheral signals of energy status appears to play a role (3, 32). Likewise, although a definitive mechanism to explain obesity-associated hypothalamic leptin resistance is still awaited, inflammation and injury affecting leptin-responsive neurocircuits offers a plausible explanation with evidence of causality in rodent models (5–7). The extent to which human obesity is associated with hypothalamic gliosis also awaits further study, although previous work suggests that this phenomenon occurs in humans as well as in rodent models (9, 33).

Although this study demonstrates an association between rapid fat loss and the resolution of hypothalamic gliosis, additional work is required to establish causality and to explore the functional significance of these changes. Little information is currently available regarding the extent to which astrocyte or microglial activation associated with HFD feeding is harmful, beneficial, both, or neither. An important next step will be to determine whether hypothalamic gliosis eventually becomes fixed and irreversible and if so, whether this effect is accompanied by defense of elevated body weight. In addition, our results do not address the question of whether ARC gliosis results from obesogenic diet, obesity itself, or both. However, the observation that genetically obese, leptin-deficient mice do not exhibit ARC gliosis when fed a chow diet but do when placed on a HFD (12) argues that HFD per se, and not its ability to increase adiposity, is the primary determinant of gliosis in this setting. Because the composition of HFD and chow differed in ways additional to the fat content in the current study, it is also premature to attribute gliosis to specific dietary components, eg, saturated fat. Future studies that include a calorically restricted HFD-fed group along with HFD-CHOW, HFD ad libitum-fed and CHOW groups will help to clarify the role of diet composition vs total energy intake or palatability in the pathogenesis of hypothalamic gliosis.

Our results differ from previous work (9, 19) in that, although we observed the expected activation of ARC microglia in the HFD group, an increase of ARC microglial number was not detected. This observation highlights the importance of tools that capture morphological changes affecting microglia in addition to cell counts (25, 26). Our findings also extend previous work using MRI to noninvasively detect changes in MBH tissue in association with chronic HFD feeding and/or obesity (9, 19). We previously reported that increased MBH T2 relaxation time is associated with immunohistochemical evidence of ARC astrocyte activation in mice with DIO (19), consistent with published clinical experience using MRI to detect gliosis associated with CNS injury (16, 34, 35). Together these observations support our interpretation of the small increase of T2 relaxation time we observed as reflecting gliosis in the HFD group. In contrast, however, MRI did not detect evidence of hypothalamic gliosis reversal in mice switched to chow after DIO had been established. This observation suggests that either MRI is insufficiently sensitive to detect this change or that obesity can affect the MBH T2 relaxation signal via mechanisms independent of gliosis per se. With respect to the latter possibility, increased T2 relaxation time within the MBH could also reflect changes such as edema, altered vascularization (36), or neuronal loss (37) as well as increased glial cell number (38).

These considerations highlight limitations inherent in the use of MRI, a technology that is complementary but not equivalent to histology. In addition to the failure to detect reversal of gliosis after consuming chow for 4 weeks, MRI did not detect tissue changes previously reported to occur upon initiation of HFD feeding (at 1 and 4 week time points) (9). Thus, MRI may be more useful for the assessment of long-term changes in hypothalamic morphology than in changes that occur rapidly. It is also possible that the study was not sufficiently powered to detect small changes in MRI signal. Power calculations for the present study were based on our previous work (19), and larger cohorts of animals in future may allow us to identify differences in T2 signal that we were undetectable here. In addition, MRI is limited in its ability to resolve small structures such as the ARC, and the ROI used for MRI analysis was considerably larger than that used for histochemical assessment. Thus, gliosis-induced changes in T2 relaxation time within the ARC may have been diluted or offset by changes in adjacent tissue that were outside the area subjected to histochemical analysis. Finally, as noted above, HFD may induce changes in the MBH that are unrelated to gliosis but nevertheless increase T2 relaxation time and are not reversible upon removal of the HFD diet stimulus (even if gliosis is). Further study is required to sort through these various possibilities.

Based on the finding that returning HFD-fed mice to unrefined chow reverses ARC gliosis, it is tempting to speculate that hypothalamic gliosis is at least transiently reversible in humans using diet, lifestyle, surgical, and/or pharmacological therapy. It must be stressed, however, that in contrast to human weight-loss studies, we did not offer the mice in the current study a choice of diet. Because most diet interventions in humans do not eliminate food choice or mandate the consumption of low-energy and relatively unpalatable foods, it is conceivable that the ability to achieve sustained weight loss in humans would be enhanced by consuming diets that are low in both palatability and energy density and that such diets are capable of reversing hypothalamic injury in humans as well as mice. Translating the current work into an investigation of the reversibility of human hypothalamic gliosis is an important priority with the potential to inform our understanding of the mechanisms underlying obesity pathogenesis and the recovery of lost weight.

In conclusion, this study offers the first evidence that obesity reversal is associated with a reduction of hypothalamic gliosis in mice. Additionally, we extend earlier work regarding the potential strengths and limitations of MRI as a tool for noninvasive assessment of obesity-associated changes in hypothalamic tissue over time. Together these findings build toward future translational studies to investigate the role of hypothalamic gliosis in obesity development and reversal in humans.