Introduction
Obesity is a disease affecting millions of people worldwide [
6,
7]. Past research has focused heavily on the effects of diets high in fat content on peripheral organs, such as adipose tissue, where obesity leads to a chronic low-level inflammation that is central to peripheral metabolic disease [
19]. This inflammation is mediated by macrophages, which infiltrate the expanding adipose tissue, and ultimately results in insulin resistance [
20,
39]. Recent work exploring the effects of obesity and metabolic disease on the central nervous system (CNS) revealed that the metabolic syndrome, concomitant with type II diabetes, is an important risk factor for neurodegeneration and cognitive dysfunction [
8,
22]. Long-term high-fat diet (HFD) leads to brain inflammation [
28,
41] and leptin resistance in the hypothalamus [
14]. Furthermore, HFD induces endoplasmic reticulum (ER) stress and apoptosis in hypothalamic neurons and inhibits neurogenesis in this region [
27,
36]. Microglia, the brain’s intrinsic immune cells, play an essential role in physiological brain functions, including pruning of neuronal synapses and regulation of brain development [
13,
31] and respond to disease or injury to the CNS [
33,
37,
40]. Fitting with the notion that diets high in fat content are harmful to the brain, reactivity of astrocytes and microglia was observed as early as 3 days after the start of HFD in rats [
36]. This response seems to be caused by dietary factors and hormonal changes rather than by increases in body weight itself [
17], suggesting that CNS resident glia directly sense nutrient components.
Our aim was to analyze the nature of the microglia reaction to high-fat diet in mice over time and investigate whether the inflammatory response induced by high-fat diet is mediated primarily by CNS resident microglia or rather represents a classical inflammatory disease in which peripheral monocytes invade the brain parenchyma, as is seen in peripheral organs in metabolic disease and classical inflammatory CNS diseases, such as multiple sclerosis. Our analyses reveal a specific reaction of hypothalamic microglia to high-fat diet. Interestingly, this reaction consists of a mixed pro- and anti-inflammatory response in the hypothalamic tissue. Through the use of bone marrow chimeras, we were able to demonstrate that endogenous microglia mediate the hypothalamic response to high-fat diet in the absence of infiltrating monocytes. Finally, analysis of human postmortem brain tissue revealed an effect of body mass index (BMI) on microglia in the hypothalamus when comparing individuals with a BMI greater than 30 to persons with a BMI below 25, implying a specific glial response to diet in humans as well.
Methods
Mice
Adult male C57Bl/6 J mice aged 100–120 days were used for all experiments and kept under specific pathogen-free conditions on a 12-h light/dark cycle, and food and water were provided ad libitum. To study the effect of HFD, mice were fed either a HFD (60 % kcal % fat, Research Diets, D12492) or recommended low-fat diet (10 % kcal % fat, Research Diets, D12450B) for different time periods between 3 days and 20 weeks. A detailed description of the different time points and the experiments for which they were used is shown in Table
1. The earlier time points were chosen based on a previous report [
36] and extended to 8 and 16–20 weeks to include a time point at which obesity is clearly manifested (Supplementary figure 1). Food intake and body weight were measured once a week (Supplementary figure 1). All animal experiments were performed in accordance with the national animal protection guidelines approved by the regional office for health and social services in Berlin (LaGeSo).
Table 1
Cohorts used for mouse experiments
(1) n = 5 (Chow)
n = 5 (3 days HFD)
n = 3 (4 weeks HFD)
n = 7 (8 weeks HFD) | 3 days, 4 and 8 weeks | Histology | |
(2) n = 7 (Chow) 8 (HFD) | 20 weeks | Bone marrow chimeric experiments, BrdU administration | |
(3) n = 6–7 | 3 and 7 days, 4 and 8 weeks | qPCR gene expression analysis | |
(4) n = 5 per diet | 3 days and 8 weeks | NanoString mRNA analysis | |
(5) n = 30 per diet | 16 weeks | Blood for plasma stimulation, microglia for LPS stimulation | |
Generation of bone marrow chimeras
For the generation of bone marrow chimeras, 1 × 10
7 bone marrow cells obtained from tibia and femur of Tg(ACTbEGFP)1Osb (GFP) mice (Jackson Laboratories) were injected into the tail vein of C57BL/6 mice exposed to 10 gray whole-body irradiation. Mice were housed in individually ventilated cages and treated with antibiotic (0.01 % Enrofloxacin, Baytril
®, Bayer Vital) for 4 weeks. To ensure that the HFD feeding had no influence on the engraftment of transplanted GFP bone marrow, HFD feeding was started after another 4 weeks of recovery. Mice were fed with high-fat diet or low-fat chow for a total of 20 weeks. Previous reports demonstrate that peripheral GFP+ blood leukocytes are not affected by this treatment [
9].
For the analysis of proliferation of microglia cells upon HFD, animals received a weekly i.p. injection of Bromodeoxyuridine (BrdU) (50 mg/kg), a thymidine analog that integrates into the DNA during replication.
Microglia sorting and cell culture
For analysis of isolated murine microglia, mice were anesthetized and perfused with phosphate-buffered saline (PBS). Hypothalamus was dissected from the brain and manually dissociated in HBSS buffer. A neural dissociation kit (Miltenyi Biotech) was used to create a single-cell suspension, which was then incubated with anti-CD11b microbeads (Miltenyi Biotech), and CD11b+ cells were isolated using MACS MS columns (Miltenyi Biotech).
Plasma of mice fed HFD or chow for 16 weeks was collected for stimulation of sorted microglia. Sorted microglia cells were plated in a 24-well plate with 5 × 105 cells per well and 3 wells per condition containing DMEM supplemented with 10 % FBS and 50 U/ml PenStrep. The next day, the medium was replaced with medium containing 10 % plasma. After 5 h, this medium was exchanged with fresh medium that was collected for analysis 1 h later.
For stimulation with lipopolysaccharide (LPS), isolated primary hypothalamic microglia were plated in 96-well plates with 5 × 104 cells per well and 3 wells per condition. The next day, cells were stimulated with 1 µg/ml LPS and the medium was collected 24 h later for cytokine analysis.
Histology
Brains were removed and stored in 4 % paraformaldehyde (PFA) overnight. The next day, PFA was replaced by 30 % sucrose for at least 24 h. Brains were mounted on a platform using freezing media and cut coronally on a cryostat at 30 µm. Sections were stored in cyroprotectant at 4 °C until use. For immunohistochemical and immunofluorescent stainings, sections were washed with PBS and incubated in PBS with 0.3 % Triton X-100 and 10 % goat serum for 1 h at room temperature followed by incubation with primary antibodies: Iba1 (1:500, Wako Chemicals, cat. # 019-19741), GFAP (1:5000, Dako, cat. # Z033401-2), GFP (1:1000, Abcam, cat. # ab290) or anti-BrdU (1:500, AbD Serotec, cat. # MCA2060GA) at 4 °C overnight. Sections were then incubated with a peroxidase-coupled goat anti-rabbit antibody (1:200, Dianova, cat. # 111-035-003) and developed with 3,3′-Diaminobenzidine (DAB) solution. For immunofluorescent staining, sections were incubated with Alexa Fluor® 488 anti-rabbit (1:200, Abcam, cat. # ab150077) or anti-rat-Cy3 (1:200, Dianova, cat. # 712-165-153). Fluorescent sections were imaged using a confocal laser-scanning microscope (Leica).
Stereological and stereomorphometric analysis
Stereoinvestigator software (MBF Bioscience) was used for the assessment of Iba1+ and GFP+ cells. Cells were counted using the Optical Fractionator method in a total of 8–10 sections per mouse collected at an interval of 6 sections apart, as previously described [
5]. For analysis of GFAP staining, pictures of 6–8 sections were analyzed with the CellSense software (Olympus) using the phase analysis tool, as previously described [
5].
Mesoscale
Serum parameters and pro-inflammatory cytokines were measured with Mesoscale [Mouse Metabolic Kit, V-PLEX Plus Proinflammatory Panel 1 (mouse) Kit] according to the manufacturer’s instructions.
qPCR
For the analysis of gene expression of whole hypothalamic tissue, RNA was isolated using the InviTrap
® Spin Tissue RNA Mini (Invitek Inc., Berlin, Germany). 1 μg of RNA was converted to cDNA using the QuantiTect Reverse Transcription Kit (Qiagen). qPCR was carried out using the TaqMan
® Fast Universal PCR Master Mix and gene-specific TaqMan
® gene expression assays (Life Technologies). qPCR results were analyzed using the delta–delta Ct method and gene expression of the target gene was normalized to that of
Gapdh, as previously published [
5].
Quantitative NanoString nCounter gene expression analysis
For NanoString nCounter analysis (NanoString Technologies) of gene expression of isolated microglia, RNA of sorted microglia cells was isolated using the PicoPure
® RNA Isolation Kit (Life Technologies) according to the manufacturer’s instructions. 10,000 cells per sample were used to measure transcript levels of 42 target and 6 housekeeping genes (Table
2). Measurements were performed at the University Medical Center Goettingen Transcriptome and Genome Analysis Laboratory (Goettingen, Germany). Results were analyzed using nSolver™ analysis software 2.5.
Table 2
Accession number and name of genes analyzed using NanoString nCounter
NM_019741 | Slc2a5 | NM_011313 | S100a6 |
NM_009151 | Selplg | NM_009115 | S100b |
NM_178706 | Siglech | NM_213659.2 | Stat3 |
NM_008479 | Lag3 | NM_010548.1 | Il10 |
NM_001164034 | Ntf3 | NM_008361.3 | Il1b |
NM_009917.5 | CCr5 | NM_031168.1 | Il6 |
NM_007651.3 | Cd53 | NM_007707.2 | Socs3 |
NM_011905.2 | Tlr2 | NM_011577.1 | Tgfb1 |
NM_021297.2 | Tlr4 | NM_009367.1 | Tgfb2 |
NM_001042605.1 | CD74 | NM_009368.2 | Tgfb3 |
NM_011146.1 | Pparg | NM_009369.4 | Tgfbi |
NM_008352.1 | Il12b | NM_009370.2 | Tgfbr1 |
NM_031252.1 | Il23a | NM_009371.2 | Tgfbr2 |
NM_008625 | mrc1 | NM_031254.2 | Trem2 |
NM_008689.2 | Nfkb | NM_011662.2 | Tyrobp |
NM_010546.2 | Ikbkb | NM_008746 | TrkC |
NM_010745.2 | Ly86 | NM_007540 | Bdnf |
NM_008320.4 | Irf8 | Housekeeping genes |
NM_001291058.1 | CD68 | NM_020559.2 | Alas1 |
NM_009987.4 | Cx3cr1 | NM_026007.4 | Eef1g |
NM_009142.3 | Cx3cl1 | NM_008062.2 | G6pdx |
NM_001111275.1 | Igf1 | NM_001001303.1 | Gapdh |
NM_146162.2 | Tmem119 | NM_010368.1 | Gusb |
NM_027571.3 | P2ry12 | NM_013556.2 | Hprt |
NM_013693.1 | Tnf | | |
Human tissue
Brain autopsies were performed following written consent for pathological examination according to the law of Berlin. Following routine diagnostic neuropathological examination, the hypothalamus and parts of the frontal cortex were obtained and used for sectioning and immunohistochemical stainings. This procedure was approved by the Charité ethics commission (EA1/019/13). Cases with infectious or inflammatory disease (peripheral or central), psychotropic drug use, history of substance addiction, chronic anti-inflammatory or immunosuppressive therapy, clinically or pathologically symptomatic brain edema, intracerebral hemorrhage, brain irradiation, chemotherapy, hypoxic or ischaemic damage or primary CNS pathology including neurodegenerative disease were excluded from the analysis. The postmortem interval was not considered as central inclusion criterion for this analysis as it has been shown to have no effect on microglia phenotype [
24]. Due to the strict exclusion criteria, only 1.6 % of the approximately 600 cases that are collected per year could be used for this analysis. Information about gender, age and BMI of the analyzed cases is given in Table
3.
Table 3
Summary of human cases
# | 9 | 12 |
BMI | 23 ± 1.9 | 36 ± 4.2 |
Gender | 7 m/2 f | 9 m/3 f |
Age (year) | 65 ± 17.2 | 69 ± 12 |
Formalin-fixed tissue was placed in 30 % sucrose for at least 1 day and then cut frozen on a cryostat into 50-µm-thick sections, which were stored in cryoprotectant at 4 °C until use. The sections were then stained using the same procedure as for the mouse tissue.
Area covered by Iba1 and GFAP immunoreactivity was analyzed in 10 sections per individual in the hypothalamus as depicted in Supplementary figure 2a. The same sized area was analyzed in the frontal cortex of the same individual and data are displayed as a ratio between the two regions to account for any confounding effects that may modify microglia phenotype. To quantify dystrophic microglia, three stages of dystrophy/cytorrhexis were defined according to Streit et al. [
35] and illustrated in Supplementary figure 2b: + beading and partial fragmentation of processes, ++ complete fragmentation of processes while still maintaining cell contours, and +++ scattered fragments with intact nucleus. Ten random images per specimen were taken and dystrophic microglia were displayed as percentage of all cells per image to account for differences in cell number between images/specimen.
Statistical analysis
Data are expressed as mean ± SEM. Comparisons between two groups were performed using Student’s t test. For the comparison of more than two groups, two-way ANOVA with Bonferroni’s post hoc analysis was used.
Discussion
Recent studies have reported an early increase in the number of microglia cells in the hypothalamus upon HFD that is accompanied by an upregulation in the expression of markers of microglia activation [
17,
36]. In the present study, we were able to confirm an increase in the amount of microglia and astrocytes, but only after 8 weeks of HFD feeding. Furthermore, we aimed to investigate whether changes in the glial populations can also be seen in obese humans, as only radiologic evidence of gliosis in the hypothalamus of obese individuals existed until present [
36]. Our analyses demonstrate that glial cells in the hypothalamus (but not in the cortex) of overweight patients are altered, which is more pronounced in microglia cells, but also slightly apparent, though not statistically significant, in the astrocyte population (Fig.
2). Importantly, we saw that not only were microglial changes evident in individuals with BMI > 30, but the degree of microglia alterations correlated significantly with the BMI. Furthermore, analysis of microglia dystrophy in the hypothalamic region revealed that although dystrophic changes were apparent in microglia of all individuals, likely owing to advancing age [
24], morphological dystrophy is exacerbated in individuals with BMI > 30. A similar observation has been made in the context of Alzheimer’s disease [
33,
34], suggesting that obesity serves to contribute to enhanced microglia stress and potential dysfunction.
Inflammatory brain diseases, such as experimental autoimmune encephalomyelitis (EAE) or stroke, are accompanied by an infiltration of peripherally derived myeloid cells into the brain [
1,
23]. When addressing whether similar cellular processes occur in the context of obesity, we found no significant influx of peripheral macrophages into the hypothalamus of GFP
+-harboring bone marrow chimeric mice after 20 weeks of HFD, in contrast to an earlier report [
9]. This difference from the latter study may be due to slightly differing experimental time courses and/or to the fact that the previously published study used FACS analysis to quantify infiltrating cells, which may have resulted in the inclusion of different myeloid subsets (besides resident microglia) such as meningeal or perivascular macrophages. Our finding of a lack of a substantial CNS recruitment of peripheral myeloid cells in HFD-fed mice is further supported by the fact that we were unable to detect elevated levels of the major macrophage-attracting chemokine CCL2 in the hypothalamus (Supplementary figure 1f), which is typically present in situations of myeloid cell recruitment to the CNS [
25,
38]. Based on the increase in Iba1/BrdU double-positive cells, we conclude that proliferation of endogenous microglia accounts for the increased myeloid cell numbers in the hypothalamus of HFD-fed mice.
Recent studies have reported increased hypothalamic cytokine gene expression upon HFD that is accompanied by an increase in the expression of markers of microglia activation [
17,
36], implying a potentially detrimental microglia response to HFD and obesity. In line with these studies, we also detected an inflammatory response in the hypothalamus only 3 days after initiating HFD. Importantly, however, this response appears to represent an acute reaction to diet in our experimental setting as an elevation in pro-inflammatory cytokine expression was not evident after 4 or 8 weeks of HFD feeding (Fig.
4a). Our findings are consistent with another study using the same HFD formulation in which the authors also failed to observe a pro-inflammatory reaction on either the mRNA or protein level after 8 weeks of feeding [
4].
In contrast to the expected pro-inflammatory reaction, we detected an increase in anti-inflammatory molecules in the hypothalamus of mice fed HFD for 8 weeks, suggesting that at this time point, microglia switched from a rather pro-inflammatory to an anti-inflammatory phenotype. This type of temporal, plastic microglia reaction to stimuli has also been reported in other, comparable settings: while acute stimulation with LPS can induce a pro-inflammatory response, chronic LPS stimulation can lead to an increase in anti-inflammatory markers [
2,
3]. Furthermore, in contrast to our study, previous analyses have focused exclusively on whole hypothalamic tissue making it impossible to exactly determine the cellular source of secreted factors. Since small, but physiologically relevant changes in cellular genes may not be detected in whole tissue preparations, we decided to analyze the genetic profile of hypothalamic microglia exclusively by acutely isolating them from mice fed HFD or chow at different time points. Our results confirm that microglia indeed produce pro-inflammatory factors after short-term exposure to HFD, but adopt a rather anti-inflammatory phenotype in response to prolonged HFD feeding. The genes we chose to study were previously identified as factors that are important for microglial sensing of endogenous and exogenous signals [
18]. Such factors were shown to decrease with aging to presumably ‘tone down’ the otherwise detrimental microglia reaction and prevent chronic activation. Similarly, we detected a downregulation of these microglia ‘sensome’ genes (
P2ry12,
Selplg,
Slc2a5,
Trem2), which suggests that microglia actively regulate their response to prolonged HFD in such a way as to avoid bystander damage. These findings raise the possibility that neuronal stress and apoptosis occurring in response to HFD [
29,
36] are not likely due to a neurotoxic, pro-inflammatory microglia response, as has been speculated. Conversely, this subdued microglia phenotype may serve as a protective mechanism in response to insult that is aimed at preserving neuronal homeostasis. These data are in line with recent findings of elevated anti-inflammatory gene expression in human adipose tissue [
15], suggesting analogous responses in the CNS and peripheral organs.
Moreover, analysis of cytokine expression of isolated adult microglia directly stimulated with the plasma of HFD-fed mice indicates that there is no excessive reaction of microglia cells to factors present in the plasma of HFD-consuming, overweight animals. This finding is in contrast to previous studies that have hinted at the possibility that blood-borne factors from HFD-fed mice influence the microglia reaction [
17,
21]. Notably, in contrast to our study, the published data are based on cultured neonatal microglia, which have been shown to be genetically and functionally distinct from adult microglia [
10] and are, therefore, likely to elicit a differential response to stimuli. Furthermore, microglia chronically exposed to HFD in mice are not primed nor significantly impaired in their response to additional stimuli like LPS and thus appear to be functional and well able to react to changes in their surroundings.
Taken together, our results demonstrate that high-fat diet specifically stimulates endogenous microglia in the hypothalamus and that the microglial response is not exclusively pro-inflammatory. Prolonged exposure to HFD results in an alternate microglia profile represented by a downregulation of microglia-specific genes involved in sensing microenvironmental alterations, likely serving to counterbalance earlier pro-inflammatory changes. This type of response appears to be a typical reaction of microglia to chronic diseases or insults and implies that diets high in fat represent a chronic challenge to CNS and microglia homeostasis. Since specific glial changes can also be found in the hypothalamus (but not the cortex) of obese humans, the glial reaction to diet needs to be studied in further detail, to ultimately determine how it might influence neuronal activity and overall metabolic health.
Acknowledgments
We would like to thank Nikolaus Deigendesch, Helena Radbruch, Jan Leo Rinnenthal, Debora Pehl, Josefine Radke and Regina von Manitius (all Department of Neuropathology, Charité, Berlin, Germany) for collecting human postmortem CNS tissue and the latter three for also assembling the corresponding clinical data. This work was supported by the Deutsche Forschungsgemeinschaft (SFB TRR 43, NeuroCure Exc 257 and HE 3130/6-1 to FLH) and from the Berlin Institute of Health (BIH; Collaborative Research Grant) to FLH.