Background
The high prevalence of obesity presents a major public health concern since obesity is strongly linked with increased risk for several diseases including type 2 diabetes, cardiovascular disease, and cancer [
1]. Importantly, obesity is also associated with adverse effects on the brain and neural function. In humans, obesity is linked with decreases in hippocampal volume and white matter integrity [
2‐
4] as well as with functional consequences that lead to accelerated cognitive decline [
5,
6] and increased risk of dementia [
7]. In rodent models, diet-induced obesity (DIO) has been demonstrated to impair neurogenesis [
8,
9], synaptic plasticity [
10,
11], and neural function [
12], as well as promote Alzheimer’s disease (AD)-related pathology [
13,
14].
Although the mechanisms by which obesity impairs neural health have yet to be fully elucidated, pathways associated with microglial activation are compelling candidates. Obesity is characterized by chronic activation of macrophages in peripheral tissues [
15‐
17] and both microglia and astrocytes in the brain [
18‐
21]. Activated macrophages yield unresolved inflammation in peripheral organs including the adipose tissue [
15,
22] and liver [
23], whereas activated microglia can drive neuroinflammation in the brain [
24,
25]. Neuroinflammation is associated with numerous deleterious effects including reductions in neurogenesis [
26] and synaptic plasticity [
27] and acceleration of AD [
28]. In addition to promoting pro-inflammatory pathways, activated microglia exhibit diverse phenotypes that are characterized by a range of morphological and gene expression signatures [
29,
30] and presumed to underlie both beneficial and adverse effects [
31,
32]. The pathways that may contribute to the neural effects of obesity remain to be fully defined.
The pattern recognition receptor Toll-like receptor 4 (TLR4) activates signaling pathways that may be particularly important in mediating obesity-associated microglial activation and its consequences. TLR4 stimulation results in downstream activation of at least two key transcription factors: NFκB, which increases expression of pro-inflammatory cytokines [
33], and interferon regulatory factor 3, which promotes activated microglial phenotypes that are relatively anti-inflammatory [
34,
35]. Thus, TLR4 activation may be expected to yield a range of activated microglial phenotypes. Interestingly, TLR4 binds to and is activated by saturated fatty acids, which are abundant in obesogenic diets and may contribute to obesity-induced increases in inflammation [
36‐
40] and impaired insulin signaling [
37,
41]. Prior work has implicated TLR4 signaling as an important regulator of DIO effects on peripheral tissues. For example, mice with either nonfunctional or deleted TLR4 exhibit significant protection against high-fat diet (HFD)-induced glucose dysregulation [
42,
43], insulin resistance [
44,
45], and peripheral inflammation [
46‐
48], though other studies indicate these mice are not protected against the entire range of metabolic and inflammatory effects of HFD [
48,
49]. Pharmacological inhibition of TLR4 also protects mice against HFD-associated adipose inflammation and fibrosis [
50] and insulin resistance [
51]. Disruption of TLR4 signaling appears to have only modest effects on increases in body weight and adiposity that result from HFD [
45,
46,
52,
53].
The potential role of TLR4 signaling in mediating obesity-induced microglial activation and associated neural impairment is unclear. Prior work has implicated TLR4 in pro-inflammatory effects of saturated fatty acids and HFD in hypothalamus, which in turn may regulate diet-induced changes in metabolic function [
54‐
56]. Given that TLR4 is highly expressed in microglia [
57,
58], TLR4 signaling pathways are implicated in activated microglial phenotypes, and activated microglia are thought to drive many of the adverse effects of obesity and HFD in hippocampus and other brain regions [
10,
59], TLR4 may mediate HFD-induced microglial activation and dysfunction in hippocampus. To address this possibility, we evaluated HFD-induced effects on metabolic, inflammatory, microglial, and neural outcomes in the presence and absence of a pharmacological inhibitor of TLR4 signaling. We report that treatment with a specific TLR4 inhibitor reduced peripheral inflammation and largely prevented both microglia activation and impaired neurogenesis in hippocampus independently of the effects on weight gain and metabolic dysregulation associated with HFD.
Methods
Animal procedures
Ten-week-old male C57BL6/J mice were purchased from Jackson Labs (Bar Harbor, ME, USA) and allowed to acclimate to our vivarium facility at the University of Southern California for 2 weeks. Animals were housed under a 12-h light/dark cycle with lights on at 6 AM and ad libitum access to food and water. At 12 weeks of age, mice were randomized to a total of four dietary and drug treatments groups (
N = 10–14/group). Dietary treatments were either control (CTL; 10% fat; #D12450J, Research Diets, New Brunswick, NJ, USA) or high-fat diet (HFD; 60% fat; #D12492, Research Diets). Drug treatments were either vehicle (0.09% sterile saline) or the TLR4 inhibitor TAK-242 (3 mg/kg in saline; #614316, EMD Millipore, Billerica, MA, USA). Drugs were administered via intraperitoneal (IP) injection 6 days/week. Dosage was based upon a previous study in which TAK-242 delivered at 3 mg/kg via IP injection yielded significant brain levels of the drug that were sufficiently maintained for at least 24 h after administration [
60]. Treatments were maintained over a 12-week experimental period, during which body weights were recorded daily and food consumption was measured weekly.
At the conclusion of the experimental period, mice were euthanized with inhalant carbon dioxide and the brains were rapidly removed. One hemi-brain was immersion fixed for 48 h in 4% paraformaldehyde/0.1 M PBS, then stored at 4 °C in 0.1 M PBS/0.03% NaN3 until processed for immunohistochemistry. Hippocampus was dissected and snap frozen for subsequent use in RNA extraction, while the remainder of the hemi-brain was snap frozen for subsequent use in protein extraction to examine soluble β-amyloid (Aβ) levels. Blood was collected via cardiac puncture into EDTA-coated tubes and centrifuged to separate plasma, which was stored in aliquots at − 80 °C. Gonadal and retroperitoneal (RP) fat pads were dissected and weighed as measures of adiposity. Both fat pads were snap frozen for subsequent RNA extraction. All animal procedures were conducted under protocols approved by the University of Southern California Institutional Animal Care and Use Committee and in accordance with National Institute of Health standards.
Body composition
Body composition was determined 1 day prior to euthanization using the Bruker LF90 Minispec (Bruker Optics, Billerica, MA, USA). Mice were placed and loosely restrained inside an acrylic cylinder. The cylinder was placed inside the bore of the magnet, and measurements of fat, lean, and fluid mass percentages were recorded. Animals were returned to their home cages in less than 2 min.
Glucose, cholesterol, and triglyceride measurements
At weeks 0, 4, 8, and 11, blood glucose readings were measured after overnight fasting (16 h). Blood was collected from the lateral tail vein and immediately assessed for glucose levels using the Precision Xtra Blood Glucose and Ketone Monitoring System (Abbott Diabetes Care, Alameda, CA, USA).
At week 11, glucose tolerance testing (GTT) was performed. First, baseline fasting glucose levels were taken. Mice were then administered a glucose bolus (2 g/kg body weight) via IP injection. Blood glucose levels were recorded from lateral tail vein 15, 30, 60, and 120 min after the glucose bolus. Area under the curve (AUC) was calculated.
Plasma cholesterol and triglyceride levels were measured enzymatically at the conclusion of the experimental period. Commercially available kits for both cholesterol (Total Cholesterol Colorimetric Assay kit, #K603, BioVision, Milpitas, CA, USA) and triglycerides (LabAssay Triglycerides, #290-63701, Wako Chemicals, Richmond, VA, USA) were used following the manufacturers’ protocols.
Behavioral analyses
All behavioral testing was conducted between the hours of 6 AM and 1 PM. For all behavioral assays, mice were brought into the behavior room and allowed to acclimate for 30 min prior to testing. After each trial, animals were returned to their home cages and the testing arenas were disinfected with 70% ethanol.
Open field and forced swim testing were video recorded and analyzed by a rater blind to experimental treatment groups. Elevated plus maze and spontaneous alternation behavior were scored live. Fear conditioning was recorded using Noldus Ethovision XT software (Leesburg, VA, USA) and the Ugo Basille Fear Conditioning System NG (Gemonia, Varese, Italy).
Anxiety and exploratory activity: open field and elevated plus maze (EPM)
Open field test was performed during week 8 of treatments. Briefly, animals were placed into a 40-cm2 plexiglass arena and allowed to move freely for 5 min. The arena floor was lightly marked off into 9 squares, with 3 squares along each wall and 1 center square. The following behaviors were recorded: (1) center crossings: the number of times the animal crossed into the center square with both front paws; (2) center time: the amount of time the animal spent with both front paws in the center square; and (3) crossings: the total number of times the animal crossed a line entering a different square.
EPM testing was performed on the day immediately following the open field assay. After being habituated to the room, mice were placed in the center of the EPM, facing a closed arm, and allowed to move freely on the maze for 5 min. The following behaviors were recorded: (1) open arm entries: the number of times the mouse placed both front paws into the open arm; (2) open arm time: the amount of time the animal spent with both front paws in the open arm; and (3) latency to enter the open arm for the first time.
Learning and memory: spontaneous alternation behavior (SAB) and fear conditioning
At week 10, SAB was tested in the Y-maze as previously described [
61,
62]. Briefly, animals were placed into the long arm and allowed to explore the maze for 5 min. Arm choices were recorded, and behavior was scored as the number of alternations divided by the total number of arm entries.
Fear conditioning was performed over 3 consecutive days beginning 48 h after SAB. On day 1, animals were placed in the conditioning chamber, a box (17 cm × 17 cm × 25 cm) with an electrified grid floor, placed inside a sound attenuated chest (Ugo Basile). White noise was used to block out external sounds. After a 3 min habituation, mice were exposed to 5 tone-and-foot shock pairings that were each placed 3 min apart (20 s tone at 85 dB and 2 kHz, followed by a 20 s trace period, and a 1 sec 1 mA foot shock). Animals were returned to their home cages 1 min after the final tone-shock pairing. Twenty-four hours after training, cued fear conditioning was tested by placing animals back into the chamber but changing the context by altering the pattern of the walls, placing a floor board over the grid floor, and adding a cotton ball with vanilla extract to change the scent of the chamber. After a 3-min baseline period, the tone was played 3 times, but was not followed by the foot shock. Freezing behavior (defined as the absence of all movement except breathing) to the tone and during the 20 s after the tone was recorded. On day 3, 24 h after cued testing, contextual fear conditioning was assessed by placing animals back into the chamber that had the same appearance and odor as it did during training on day 1. Freezing behavior was measured over 8 min. Behavior in the fear conditioning chamber was recorded using Noldus Ethovision XT software.
Depression-like behavior: forced swim test (FST)
FST was conducted 1 week after fear conditioning, during week 11, and was the last behavioral assessment. As previously described [
61], the animals were placed into a 2-L cylindrical tank (20 cm height × 13 cm diameter) filled with 15 cm of water heated to 23–25 °C. At this depth, neither the feet nor tails of animals reached the floor of the cylinder. Mice remained in the cylinder for 5 min, during which behavior was videotaped from the side of the cylinder. Animals were scored as being immobile if they were making only the movements necessary to keep their head above water. The number of immobile bouts, the total time spent immobile, and the duration of the longest bout of immobility were recorded.
Immunohistochemistry and quantification
Fixed hemi-brains were completely sectioned at 40 μm in the horizontal plane, using a vibratome (Leica Biosystems, Buffalo Grove, IL, USA). A standard avidin/biotin peroxidase approach using ABC Vector Elite kits (Vector Laboratories, Burlingame, CA, USA) was used to perform immunohistochemistry, as previously described [
63]. Every eighth section was processed for ionized calcium binding adaptor molecule 1 (IBA-1), doublecortin (DCX), and bromodeoxyuridine (BrdU). A different initial antigen retrieval step was performed for each antibody, after which the same protocol was followed. For IBA-1 staining, sections were boiled in 10 mM EDTA, pH 6.0 for 10 min, then rinsed in water three times for 5 min each. For DCX staining, tissue was pretreated with 95% formic acid for 5 min, followed by rinsing in TBS. Finally, for BrdU staining, sections were placed in 1% NP40 detergent for 20 min, rinsed in TBS, then incubated in 2 N HCl at 37 °C for 30 min, followed by 10 min in 0.1 M boric acid and rinsing in TBS. Following the various antigen retrieval steps, sections were treated with an endogenous blocking solution for 10 min, then rinsed with 0.2% Triton-X in TBS, 3 times for 10 min each. Tissue was then incubated for 1 h in a blocking solution consisting of 2% bovine serum albumin and 0.2% Triton-X in TBS for IBA-1, plus 2% normal goat serum for BrdU. For DCX, the blocking solution was made up of 3% normal horse serum and 0.2% Triton-X in TBS. Blocked sections were incubated overnight at 4 °C in primary antibody directed against IBA-1 (#019-19741, 1:500 dilution, Wako Chemicals); DCX (#sc-271390, 1:1000 dilution, Santa Cruz Biotechnology, Dallas, TX, USA); or BrdU (#MCA2483, 1:200 dilution, Bio-Rad, Hercules, CA, USA). All primary antibodies were diluted in the respective blocking solution used. On the following day, sections were rinsed and incubated in biotinylated secondary antibody diluted in blocking solution. Finally, immunoreactivity was visualized using 3,3′-diaminobenzidine (Vector Laboratories).
Density and activation states of microglia were determined using live imaging under bright-field microscopy with a × 40 objective (Olympus, BX50, CASTGrid software, Olympus, Tokyo, Japan). As previously described [
63,
64], each cell was scored as having either a resting or reactive phenotype. Specifically, resting or type 1 microglia were defined as having spherical cell bodies with numerous thin, branched processes. Both type 2 and 3 microglia were considered reactive: type 2 cells had enlarged, rod-shaped cell bodies with fewer and thicker processes, while type 3 cells were enlarged and had either very few or no processes, or several filopodia. Microglia were quantified in the entorhinal cortex (4 fields/section), subiculum (4 fields/section), CA1 (5 fields/section), and CA2/3 (3 fields/section) across 4 tissue sections for a total of 64 fields and an average of ~ 450 cells per brain. Because increased soma size is a robust indicator of microglial activation [
65], we also examined microglial soma size. Images of IBA-1 immunostaining in the CA1 subregion of the hippocampus were digitally captured using an Olympus BX50 microscope and DP74 camera paired with a computer running CellSens software (Olympus). Microglial cell bodies were outlined, and their area was determined using NIH ImageJ 1.50i (US National Institutes of Health, Bethesda, MD, USA).
DCX- and BrdU-immunoreactive cells were also quantified using live imaging under bright-field microscopy with a × 100 oil immersion lens (Olympus). Cells were counted in non-overlapping fields of the subgranular zone and granule cell layer of the dentate gyrus, across 8 sections per animal. Additionally, to examine the relative maturity of DCX-expressing cells, the morphology of their dendritic processes was assessed as previously described [
66‐
68]. Briefly, immature or type 1 cells were defined as having very short or no processes, intermediate or type 2 cells as having processes that extended only within the granule cell layer and do not extend into the molecular layer, and post-mitotic or type 3 cells as having dendrites that extend and branch into the molecular layer or having multiple branches within the granule cell layer. Morphology of DCX-positive cells was assayed across 4 sections per animal, and the relative proportions of type 1, 2, and 3 cells were calculated.
RNA isolation and quantitative PCR
RNA was extracted from the gonadal fat pads and the hippocampus using TRIzol reagent (Invitrogen Corporation, Carlsbad, CA, USA), following the manufacturer’s protocol. To remove any remaining DNA contamination, the RNA pellet was treated with RNase-free DNase I (Epicentre, Madison, WI, USA) for 30 min at 37 °C after which a phenol-chloroform extraction was performed to isolate RNA. cDNA was reverse transcribed from 1 μg of purified RNA using the iScript cDNA synthesis system (Bio-Rad). The resulting cDNA was used to run real-time quantitative PCR using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) and a Bio-Rad CFX Connect Thermocycler, as previously described [
63]. Both hippocampus and adipose tissue were analyzed for expression levels of cluster of differentiation 68 (CD68), EGF-like module-containing mucin-like hormone receptor-like 1 (F4/80), major histocompatibility complex class II (MHC II), cluster of differentiation 74 (CD74) transcript variant 1, interleukin-6 (IL-6), and interleukin-1β (IL-1β). Additionally, hippocampal tissue was assessed for lipoprotein lipase (LPL) and CD36, as well as for the Aβ clearance and production factors neprilysin, insulin-degrading enzyme (IDE), and β-site APP cleaving enzyme (BACE1). Finally, levels of the cytokine tumor necrosis factor α (TNFα) transcript variant 1 were examined in adipose tissue. Primer pair sequences for target genes are shown in Table
1. All samples were run in duplicate, and PCR products were normalized with corresponding expression levels of β-actin and/or phosphoglycerate kinase 1 (Pgk1) in the brain and succinate dehydrogenase complex, subunit A, flavoprotein (SDHA) in the adipose tissue. The ΔΔ-CT method was used to determine relative mRNA levels. For hippocampal samples, the Ct value of each reference gene (β-actin and/or Pgk1) was subtracted separately from the target genes and the resulting values were averaged and used to calculate fold changes relative to the control-diet, vehicle-treated group. CD68, F4/80, MHCII, CD74, IL6, IL1β, LPL, and CD36 were run with both reference genes and neprilysin, IDE, and BACE1 only with β-actin.
Table 1
Target genes for the PCR analyses are listed with their corresponding GeneID number and oligonucleotide sequences for the forward and reverse primers
β-actin
Gene ID: 11461
| Forward: 5′-AGCCATGTACGTAGCCATCC-3′ Reverse: 5′-CTCTCAGCTGTGGTGGTGAA-3′ |
β-site APP cleaving enzyme (BACE1)
GeneID: 23821
| Forward: 5′-TCGCTGTCTCACAGTCATCC-3′ Reverse: 5′-AACAAACGGACCTTCCACTG-3′ |
Cluster of differentiation factor 36 (CD36)
GeneID: 12491
| Forward: 5′-TATTGGTGCAGTCCTGGCTG-3′ Reverse: 5′-CTGCTGTTCTTTGCCACGTC-3′ |
Cluster of differentiation factor 68 (CD68)
GeneID: 12514
| Forward: 5′-TTCTGCTGTGGAAATGCAAG-3′ Reverse: 5′-AGAGGGGCTGGTAGGTTGAT-3′ |
Cluster of differentiation factor 74 (CD74), transcript variant 1
GeneID: 16149
| Forward: 5′-CAAGTACGGCAACATGACCC-3′ Reverse: 5′-GCACTTGGTCAGTACTTTAGGTG-3′ |
EGF-like module-containing mucin-like hormone receptor-like 1 (F4/80)
GeneID: 13733
| Forward: 5′-TGCATCTAGCAATGGACAGC-3′ Reverse: 5′-GCCTTCTGGATCCATTTGAA-3′ |
Insulin-degrading enzyme (IDE)
GeneID: 15925
| Forward: 5′-TGTTTCCACACACAGGCAAT-3′ Reverse: 5′-ACCTGTGAAAAGCCGAGAGA-3′ |
Interleukin-1β (IL1β)
GeneID: 16176
| Forward: 5′-GCAACTGTTCCTGAACTCAACT-3′ Reverse: 5′-ATCTTTTGGGGTCCGTCAACT-3′ |
Interleukin-6 (IL6)
GeneID:16193
| Forward: 5′-CTCTGGGAAATCGTGGAAAT-3′ Reverse: 5′-CCAGTTTGGTAGCATCCATC-3′ |
Lipoprotein lipase (LPL)
GeneID: 16956
| Forward: 5′-GGGCCCAGCAACATTATCCA-3′ Reverse: 5′-GGGGGCTTCTGCATACTCAA-3′ |
Major histocompatibility complex class II (MHC II)
GeneID: 14961
| Forward: 5′-CAGACGCCGAGTACTGGAAC-3′ Reverse: 5′-CAGCGCACTTTGATCTTGGC-3′ |
Neprilysin
GeneID: 17380
| Forward: 5′-GAGAAAAGCCCACTTGCTTG-3′ Reverse: 5′-GAAAGACAAAATGGGGCAGA-3′ |
Phosphoglycerate kinase 1 (Pgk1)
GeneID: 18655
| Forward: 5′-GCCTGTTGACTTTGTCACTGC-3′ Reverse: 5′-GAGTGACTTGGTTCCCCTGG-3′ |
Succinate dehydrogenase complex, subunit A, flavoprotein (SDHA)
Gene ID: 66945
| Forward: 5′-ACACAGACCTGGTGGAGACC-3′ Reverse: 5′-GGATGGGCTTGGAGTAATCA-3′ |
Tumor necrosis factor α (TNFα), transcript variant 1
GeneID: 21926
| Forward: 5′-CCCTCACACTCAGATCATCTTCT-3′ Reverse: 5′-GCTACGACGTGGGCTACAG-5′ |
β-Amyloid enzyme-linked immunosorbent assay
Levels of soluble Aβ42 peptides were determined by enzyme-linked immunosorbent assay (ELISA) as described previously [
69], with noted modifications. Briefly, the remaining hemi-brain portions were homogenized in buffer (0.2% diethylamine, 50 mM NaCl, 1 mL/200 mg tissue) using a polytron on ice. Resulting homogenates were centrifuged at 4 °C for 1 h at 15,000 g. Supernatants were collected and neutralized with 1/10th volume of 0.5 M Tris-HCl, pH 6.8. Samples were then analyzed using a commercially available Aβ42 ELISA (Human/Rat β Amyloid 42 ELISA Kit High Sensitive; 292-64501; Wako Chemicals) according to manufacturer’s directions.
Statistical analyses
All data were analyzed using Prism software (version 7, GraphPad Software, La Jolla, CA, USA). Two-way repeated measures ANOVAs were performed for the analyses of body weight and glucose tolerance. All other data were analyzed by two-way ANOVAs. In the case of significant main effects, planned comparisons between groups were made using the Bonferroni correction. All data are represented as the mean ± the standard error of the mean (SEM). Significance was set at a threshold of p < 0.05.
Discussion
The goal of this study is to examine the role of TLR4 signaling in mediating the effects of obesity on microglial activation and adverse neural outcomes. Comparing animals fed control versus HFD in the presence or absence of the TLR4 inhibitor TAK-242, we demonstrate that TAK-242 treatment was associated with attenuation of HFD-induced adipose tissue inflammation, microgliosis, and reduction in neurogenesis in the hippocampus. However, TAK-242 treatment did not improve the metabolic dysregulation induced by HFD feeding. The finding that TLR4 inhibition did not protect against effects of HFD on weight gain and adiposity is consistent with numerous other studies [
42,
44‐
46,
49,
52,
53]. In contrast to our findings, however, many of these studies show that obesity-associated dysregulation of insulin and glucose signaling was improved in the absence of TLR4 signaling [
42‐
47,
53]. One possible reason for this discordance is that several studies used mice with either knockout or dysfunctional TLR4, whereas we used a pharmacological approach to inhibit TLR4. Constitutive absence of TLR4 signaling may result in metabolic changes even in the absence of HFD and is likely to result in more complete inhibition of TLR4 and different outcomes than pharmacological approaches.
Our findings support the conclusion that TLR4 contributes to obesity-induced activation of peripheral macrophages and brain microglia. First, our observations in adipose tissue of partial reductions in both markers of macrophage activation (CD68, F4/80, MHCII) and pro-inflammatory cytokines (IL-6, TNFα) in HFD-fed mice treated with TAK-242 is consistent with previous findings [
42,
44,
46,
48,
53]. Second, we demonstrate that the TLR4 inhibitor attenuates HFD-induced microgliosis in hippocampus, as evidenced by changes in microglial morphology and soma size and mRNA levels of the activated microglia markers CD68 and, to a lesser extent, CD36 or fatty acid translocase. CD36 is a pattern recognition receptor that exhibits increased expression in obesity [
93] as well as in AD [
99,
100], where it mediates recruitment of microglia to Aβ deposits [
101,
102]. Interestingly, CD36 has been found to form a complex with TLR4 and TLR6 through which both Aβ and lipids can induce inflammation [
103]. Collectively, these findings support and extend prior work by Milanski and colleagues that implicated TLR4 in obesity-induced glial activation in hypothalamus, which may contribute to systemic metabolic disturbances [
54,
55].
Although we observed an increase in activated microglia in response to HFD, it is noteworthy that hippocampal expression of the pro-inflammatory cytokines IL-6 and IL1-β was not increased by obesogenic diet. Though HFD is often associated with both microglial activation and increased cytokine expression [
104], others show changes only in some brain regions [
105], or no changes in pro-inflammatory cytokines [
106]. Here, we find DIO-associated changes in specific markers of activated microglia but not in cytokines, which is consistent with previous work by Setti and colleagues [
89]. Although it is reasonable to predict that a more chronic exposure to HFD may be required for increased neural cytokine expression, cytokine expression in hypothalamus is significantly increased by high-fat diet exposures as brief as 1 day [
107]. We posit that the observed microglial activation in the absence of significantly increased expression of pro-inflammatory cytokines is consistent with the known heterogeneity in activated microglial phenotypes [
29,
30]. Indeed, accumulating evidence indicates that deleterious effects of microglia are mediated by numerous factors rather than simply increased levels of pro-inflammatory cytokines [
32,
108]. The extent to which various activated microglial phenotypes differentially affect neural outcomes is an important topic that remains to be fully elucidated.
One deleterious neural consequence common to both diet-induced obesity and activated microglia is promotion of amyloidogenesis. HFD is known to increase gene expression and/or enzyme activity of the pro-amyloidogenic BACE1 [
95] and decrease levels of the Aβ-degrading enzymes neprilysin [
94] and insulin-degrading enzyme [
97]. These effects on Aβ homeostasis likely contribute to observations that experimental obesity drives Aβ accumulation in transgenic mouse models of AD [
13,
14]. Additionally, it has been shown that inflammation can increase levels of BACE1 [
109] and decrease levels of neprilysin [
110]. Thus, microglial activation and associated neuroinflammation are likely significant mediators of the obesity-induced increase in Aβ. We found that HFD significantly increased levels of BACE1 in vehicle-treated, but not in TAK-242-treated mice. This suggests that HFD caused a shift towards more pro-amyloidogenic processing via TLR4 signaling. Although not statistically significant, our analyses of soluble brain Aβ42 showed trends towards increased levels in HFD mice in the absence but not the presence of TLR4 inhibitor.
Another negative effect of obesity is attenuation of neurogenesis. We found that treatment with TAK-242 significantly increased the number of new neurons in dentate gyrus specifically in HFD-fed mice, indicating a protective effect of TLR4 inhibition on obesity-related impairment in neurogenesis. Because BrdU labeling, a marker of cell proliferation, was not affected by diet or drug treatments, the protective effect of TAK-242 appears to involve the survival and or differentiation of newborn neurons rather than stem cell proliferation. The possibility that TLR4 inhibition yielded a generalized increase in new neuron survival is consistent with our finding that the relative proportion of subtypes of newly formed neurons was not significantly altered by either diet or TLR4 inhibition. The reported effects of HFD on neurogenesis and cell proliferation are somewhat mixed in the literature, with some studies finding decreases in both [
111,
112] and others finding changes in only one [
9,
113] or neither [
114] marker of neurogenesis. Differences in experimental parameters including the composition of the diet may affect the extent to which cell proliferation and/or survival of newborn neurons are affected by HFD. Our observed effects were likely mediated by microglia, which have previously been shown to attenuate neurogenesis during states of activation such as after LPS [
115,
116] or seizure [
26]. Further, our finding of increased neurogenesis with TAK-242 treatment in HFD-fed mice is consistent with prior data showing that TLR4 signaling regulates neurogenesis in response to neural injury and microgliosis [
117,
118]. Importantly, adult neurogenesis is regulated both positively and negatively by a range of activated microglial phenotypes [
32,
119], reinforcing the emerging complexity of the associations between microglial functions and their activation states.
One limitation of this study is that we were not able to fully determine the effects of TLR4 inhibition on HFD-induced behavioral changes. Although behavioral impairment is often associated with obesity, we found very subtle effects of diet and drug treatments on overall behavioral outcomes. Specifically, mice fed HFD and treated with TAK-242 showed small but significantly increased exploratory behavior/decreased anxiety-like behavior in the open field test, and worse spontaneous alternation in the Y-maze, and HFD was associated with decreased cued memory in fear conditioning. There were no significant effects of our diet or drug manipulations on anxiety-like behavior in EPM, depressive-like behavior in forced swim, or on contextual fear conditioning. Though a number of studies demonstrated cognitive impairments after HFD exposure [
120‐
124], others did not [
125‐
128]. The age at which rodents are exposed to diet-induced obesity may be a factor. For example, one study found significant effects of HFD on behavior in mice started on diet at 5 weeks of age, but not in animals started at 8 weeks [
129], whereas another found behavioral impairments in response to HFD in aged but not young adult rats [
130]. These studies suggest that the age at which exposure to HFD occurs may be important in determining whether behavioral deficits are observed. As with the induction of neuroinflammation, it is unlikely that the length of HFD exposure is the key variable in whether or not behavioral impairment occurs. Previous studies of HFD outcomes in rodents have showed changes in both neuroinflammation [
107,
130] and behavioral outcomes [
130,
131] within 3 days of HFD feeding. Moreover, deficits in cognitive performance have also been observed after 9 days [
132], 1 month [
133], and 3 months [
131] of diet exposure.
Though complete elucidation of the mechanisms underlying the effects of obesity on the brain remains to be established, our findings suggest a stronger role for microglial activation than for metabolic dysregulation. That is, despite having similar weight gain and metabolic outcomes in response to HFD, mice treated with a TLR4 inhibitor showed significant reductions in microglial activation and increased neurogenesis in comparison to vehicle-treated mice. This position is consistent with findings in the human literature that the effects of obesity on cognitive impairment are mediated largely by glial/inflammatory rather than metabolic factors [
134‐
136]. Because TAK-242 has systemic effects and we observed partial attenuation of inflammation in adipose tissue, peripheral effects of TLR4 inhibition may have contributed to the observed neural benefits. The role of other mechanisms like vascular and microbiota changes in the effects of obesity on the brain cannot be ruled out and should be addressed in future studies, especially given that inflammation may be important in these systems as well [
137,
138].