Background
Alzheimer’s disease (AD) is a progressive neurodegenerative disease associated with decline in cognitive function, impairment in memory, language and visual-spatial coordination, eventually resulting in complete loss of basic function [
1]. Dementia of all forms affects about 5 % of the population older than 65. There are approximately 5.5 million cases of AD in the USA alone, and this number is estimated to nearly triple by the year 2040. Moreover, as the world population lives longer, AD and dementia are predicted to constitute a major global health problem in the aging population of the world [
2].
Major pathological hallmarks of AD present as senile amyloid plaques that are composed of β-amyloid (Aβ) protein and intracellular neurofibrillary tangles with characteristic reactive microgliosis and astrogliosis. AD is characterized by dystrophic neuritis, neuronal loss, synaptic dysfunction, and cerebral atrophy [
1]. AD can be largely divided into early and late onset forms. Early onset AD is induced in patients carrying genetic mutations in amyloid precursor protein (APP) and/or presenilin (PS1 or PS2) which induces formation of insoluble Aβ. Previous work showed that mice expressing mutated
APP display alterations in exploratory activity as well as elevation of β-amyloid production reminiscent of AD [
3,
4]. Late onset AD, also known as sporadic AD, is believed to be induced by aberrant processing of Aβ resulting in pathological lesions. In general, APP is cleaved by α-secretase which generates soluble Aβ which has neuroprotective functions [
5]. However, when APP is cleaved by β-secretase-1 (BACE1), it produces Aβ
1–42 which is insoluble and forms amyloid plaques that are pro-apoptotic and neurodegenerative and are believed to induce cognitive impairment [
6,
7]. While the underlying cause of AD is not known, advancing age, environmental stressors, and genetic factors appear to be important precursors [
8,
9]. In addition to Aβ deposits, AD is characterized by neurofibrillary tangles which are neuronal deposits of hyperphosphorylated-Tau, also referred to as tauopathies, which are well correlated with cognitive impairment and advanced neurodegeneration [
1,
10]. Due to this association, it is still debated whether the initiating factor of AD is Aβ plaques or tauopathy [
11].
In the USA, diets high in fats/lipids, which are commonly known as “fast foods,” are prevalent in a significant portion of the population and are becoming an important public health issue. High-fat diets are implicated in various metabolic syndromes leading to obesity, atherosclerosis, insulin resistance, dementia, cognitive decline, and potentially, AD [
12‐
16]. HFD also induces a liver pathology called non-alcoholic fatty liver disease (NAFLD) which is characterized by fatty liver, accumulation of lipids in hepatocytes, infiltration of inflammatory immune cells in the liver parenchyma, and secretion of pro-inflammatory cytokines resulting in liver damage [
13,
14,
17,
18]. NAFLD is the fourth leading cause of liver disease in the Western hemisphere. It afflicts about 30 % of the US population and is the 12th leading cause of death in the USA among adults 45–54 years [
19]. The increase in NAFLD has been linked to increased prevalence of obesity and metabolic diseases in the USA and worldwide [
7]. NAFLD is associated with marked progressive inflammation, fat deposition, and fibrosis of the liver [
8,
10]. Also, a clear association exists between cardiovascular risk factors or carotid atherosclerosis and dementia progression leading to AD [
20].
Cholesterol is an important building block of the brain [
21], which produces over 20 % of total cholesterol in the body. This high-cholesterol content is needed for neuronal function, as the brain cannot access plasma cholesterol due to restrictions posed by the blood brain barrier [
22]. For this reason, neuronal cells express high levels of cholesterol-uptake receptors such as low-density lipoprotein receptor (LDLR), low-density lipoprotein receptor-related protein 1 (LRP1), and apoliprotein-E (ApoE). Up to 70 % of the brain’s cholesterol make up the myelin sheath of oligodendrocytes and the membrane of astrocytes, with the remainder contributing to neuronal function including the myelin sheath of neurons that relay synaptic signals. ApoE and LRP1 are related to cholesterol metabolism and are important risk factors contributing to the prevalence of AD [
23‐
25]. The ApoE variant, ApoE4, increases AD risk and accelerates AD onset (Bu, 2009; Liu et al., 2013). Multivariate analysis of metabolites in the blood of AD patients showed that of the ten metabolites that distinguished AD from its age-matched cohorts, six were long-chain cholesteryl esters that were reduced in AD [
26]. Also, several large cohort studies showed that long-term treatment with statins, which lowers serum cholesterol levels, could alleviate AD symptoms, suggesting that alteration in lipid metabolism contributes to AD pathogenesis [
27,
28]. LRP1 is an endocytic receptor highly expressed in the liver, on neurons, and on vascular smooth muscle and glial cells in the CNS vasculature and functions in the clearance of Aβ from the CNS [
29]. Aβ clearance is impaired in neurons from LRP1-deficient mice [
30]. Binding of APP to LRP1 results in increased trafficking and clearance of APP. However, LRP1 is also involved in Aβ production [
31]. Hence, its involvement in Aβ synthesis and clearance makes it a prime target in AD pathogenesis. Deletion of LRP1 exacerbated Aβ deposition and increased CAA [
29]. It is speculated that ApoE may inhibit or facilitate LRP1 endocytosis of Aβ. Since ApoE4 is linked to both sporadic and familial AD and ApoE functions in the cellular transfer of lipids through LRP1 on the cell surface, it is presumed that LRP1, ApoE, or both are involved in dysfunction of the lipid transport mechanism and Aβ clearance.
The increase in obesity and NAFLD prevalence in our society, both induced by diets high in fats/lipids and their link to chronic inflammation and metabolic diseases, mirrors increase in AD and AD-like syndromes. We therefore decided to investigate the impact of NAFLD in AD pathogenesis in an AD transgenic mouse model (APP-Tg) and in C57BL/6, wild-type (WT) mice.
To elucidate the effect of a diet with increased lipids on AD pathogenicity, we fed WT and APP-Tg mice with a high-fat diet (HFD) or standard diet (SD) for 2, 5 months, or 1 year. HFD induced systemic and CNS inflammation that accelerated Aβ plaque deposition during acute NAFLD (2 and 5 months) in APP-Tg mice. In WT mice, acute NAFLD induced neuro-inflammation but did not induce Aβ plaques. Removal of HFD after 2 months decreased Aβ plaque load in APP-Tg mice and reversed signs of systemic and CNS inflammation in both WT and APP-Tg mice. WT and APP-Tg mice were kept on HFD for up to 1 year to determine the impact of chronic NAFLD in AD induction in WT mice and AD progression in APP-Tg mice. We observed advanced signs of AD, including accelerated cerebral amyloid angiopathy (CAA), increased tauopathy, and increased neuronal loss in APP-Tg mice. More importantly, long-term HFD treatment induced plaque formation, CAA, and tauopathy in WT mice. The advanced signs of AD were associated with a decrease in CNS expression of LRP1 during chronic disease. These studies indicate that HFD-induced inflammation plus aging are sufficient to trigger neurodegeneration and accelerate the process of AD even in the absence of genetic predisposition.
Methods
Mice and diet
The APP-Tg mouse [B6.Cg- Tg (APPswe, PSEN1dE9)85Dbo/J] was generated as previously described [
32]. WT mice and their APP-Tg littermates were fed either a standard diet (SD) (Harlan Teklad TD.7912) or a high-fat diet (HFD) (1.0 % cholesterol, 0.5 % cholic acid, 18 % triglyceride; Harlan Teklad TD.88051, “Paigen diet”) [
14] beginning at the age of 2 months. WT and APP-Tg mice were fed with SD or HFD for 2, 5 months, or 1 year. Another set of APP-Tg mice were removed from HFD after 2 months and were put back on SD for 3 months. All animal work was done in accordance with PHS guidelines and was approved by Cornell’s Institutional Animal Care and Use Committee (Protocol # 2008–0092).
Tissue harvest
Deeply anesthetized mice were weighed and transcardially perfused with ice-cold phosphate buffered saline (PBS); then, brain, spleen, and liver were collected for analysis. After macroscopic photo documentation, all livers were weighed and used for leukocyte preparation, except two 30-mg tissue sections which were used for histopathology and RNA preparation. Approximately 30 mg of liver and one brain hemisphere were flash frozen in Tissue-Tek O.C.T. (Sakura Finetek) and stored at −80 °C. 10-μm thick frozen sections were affixed to Superfrost/Plus slides (Fisher), fixed in acetone, and stored at −80 °C.
Immunohistochemistry
For immunohistochemistry staining, slides were thawed and treated with 0.03 % H2O2 in PBS to block endogenous peroxidase or fixed and permeabilized in acetone for immunofluorescence staining, blocked with casein (Vector Laboratories) in normal goat serum (Zymed), and then incubated with anti-CD45, phospho-Tau, ApoE, CD31, LRP-1, 6E10, GFAP, or NeuN primary antibodies. For immunohistochemistry, slides were then incubated with biotinylated goat anti-rat Ig (Jackson ImmunoResearch) and streptavidin-HRP (Zymed) and developed with an AEC (Red) substrate kit (Zymed), counter-stained with hematoxylin and mounted with Fluoromount-G. For immunofluorescence assay, slides were instead subjected to AF488, TexRed, or AF647 conjugated secondary antibody and coverslips were mounted with Vectastain containing DAPI (Vectorlabs). Standard or frozen histological tissue sections were formalin-fixed and processed for hematoxylin and eosin (H&E) or oil red-O staining and hematoxylin counterstaining, respectively, then examined by light microscopy. For the green fluorescent Thioflavine S (ThioS) staining of plaques, frozen sections were incubated with 1 % ThioS (Sigma-Aldrich) in distilled water for 5 min, differentiated in 70 % ethanol for 5 min, washed three times for 5 min each with distilled water and cover-slipped with Vectastain containing DAPI (Vectorlabs). Images were captured using a Zeiss Axio Imager M1 microscope. For further quantification of acquired images, Zen software (Carl Zeiss) was used to obtain intensities of signals and counting of positive signals. A detailed explanation on the method of quantification is described in each figure legend.
TUNEL assay
For TUNEL staining, which was used for detecting cell death, the reaction mixture supplied by Roche’s In Situ Cell Death Detection Kit, AP (Cat. No. 11 684 809 910) was used following the protocol provided by Roche. Briefly, 10-μm thin sections of cryogenic brain tissue which were previously lightly fixed in acetone for 5 min and stored at −80 °C were thawed and fixed with 4 % paraformaldehyde for 20 to 30 min. Then, sections were permeabilized using 0.1 % Triton X-100 (Sigma-Aldrich, 9002-93-1) in sodium citrate (Fisher Scientific, 6132-04-3) with PBS for 2 min. TUNEL reaction mixture supplied by Roche’s In Situ Cell Death Detection Kit, AP was added to each section and left over night at 4 °C. Negative controls received only label solution and no terminal transferase enzyme, while positive controls were pretreated with DNase I Recombinant in 50 mM Tris-HCl and 10 mM MgCl2 for 10 min at room temperature to induce DNA strand breaks. TUNEL reaction mixture was then added to all samples. After washing with PBS, sections were fixed with 4 % paraformaldehyde for 10 min and then washed again with PBS. Images were captured using a Zeiss Axio Imager M1 microscope.
Quantitative PCR
Brain and liver mRNA was extracted with TRIZOL (Invitrogen) and cDNA was synthesized using High-Capacity cDNA Reverse Transcription Kits (Applied Biosystems) according to the protocols provided by the manufacturers. Quantification of expression levels of ApoE, LRP1, TLR1, TLR2, TLR6, and of pro-inflammatory cytokines (IL-6, TNF-Α-α, IL-17, and IL-1β) was performed using specific primers and KAPA SYBR FAST qPCR Kit (KAPA biosystems) and ran on CFX96 thermocycler (Bio-Rad). Relative mRNA expression levels of genes were analyzed using the 2^dCT method, normalized with GAPDH as reference gene. The fold change plotted was relative to the respective SD controls. Specificity of reaction was analyzed using melting curve analysis. Primer sequences can be viewed in supplemental Table 1.
ELISA assay
Splenocytes were harvested from SD or HFD-fed mice after 1 year and treated with either PBS or ConA for 48–72 h. Supernatant was collected and used for ELISA analysis using the eBioscience Ready Set Go Kit. Briefly, plates were coated overnight at 4 °C with capture antibodies against IL-6, TNF-α, or IL-17 and then washed. Wells were incubated with standards and samples, washed, and then subsequently incubated with biotin conjugated detection antibodies. Plate was washed and developed with 1× TMB substrate solution (eBioscience) and the optical density (OD) was read at 450 nm using a Biotek fluorometer (Biotek). OD values were converted into absolute concentration using the standard curve.
Detailed quantification method of immunofluorescence images
For immunofluorescence images, different methods were used for its quantitative analysis. For plaque quantification, either sizes or numbers of Thioflavin S+ plaques from different areas of cortex of animals of different groups were measured (area size: 6 × 105 μm2/field, ten fields/section, blindly chosen, five animals/group). Only plaques larger than 2 × 103 μm2 were measured and smaller plaques were not considered. For cortical thickness quantification, lengths of layer I to VI of cortex were measured from blindly chosen ten different regions of whole cortex from animals of different groups were measured for analysis (two animals/group, blindly chosen). For analysis of fragmented vessels, CD31 (vascular marker)-positive cells that lost normal brain vascular integrity (linear and smooth outlining of vessel) were considered as fragmented vessels, and the numbers were counted from cortex of animals of different groups (area size: 6 × 105 μm2/field, ten fields/section, blindly chosen, two animals/group). All analyses (measurement of length, area size) were performed using automatic calibration function of Zen software (Carl Zeiss).
Western blotting
Brains from WT or APP-Tg mice fed with SD or HFD were homogenized and lysed with lysis buffer containing protease inhibitor cocktail. Samples were loaded onto a 10 % SDS acrylamide gel and separated for 1 h and transferred to a nitrocellulose paper. The membrane was blocked with 1 % BSA/TBST and incubated with anti-phospho-Tau antibody (S396 from abcam, AT8 from Millipore) or anti-Tau antibody (Cell signaling) overnight at 4 °C and washed three times with TBST. For loading control, anti-GAPDH (Cell signaling) was used. Subsequently, the blot was incubated with HRP conjugated secondary antibody for an hour at room temperature and washed three times with TBST. The blot was developed with Super Signal West Pico ECL solution (Thermo scientific) and exposed to X-ray film. The film was scanned, and the intensity of each band was analyzed with Image J software.
Statistical analyses
Data were analyzed by one-way or two-way ANOVA, followed by Bonferroni’s multiple comparison test or Student’d t test (two-tailed, unpaired) using GraphPad Prism 5 software (GraphPad, La Jolla, CA). Plotted data shown represents Mean ± SEM where significance is indicated by *p < 0.05, **p < 0.01, ***p < 0.001.
Discussion
We demonstrate that a modest increase in dietary lipids fed to WT or APP-Tg mice induced acute inflammation of the liver followed by a chronic inflammatory state. This HFD-induced acute inflammation was characterized by invasion of inflammatory immune cells into the liver parenchyma and increased production of pro-inflammatory cytokines by immune cells locally (liver) and systemically (lymphoid organs). Concomitant with this increased peripheral inflammatory state, we observed profound acceleration in neurodegenerative signs and increased neuro-inflammation in APP-Tg mice compared to mice that remained on SD. We observed acceleration in β-amyloid plaque formation leading to higher plaque loads, larger plaque size, and increased microgliosis, and astroglyosis in APP-Tg mice. In AD patients’ brain, activated microglial cells have been observed around plaques and were initially thought to be a clearance mechanism to remove Aβ plaques from the CNS. However, these activated microglial cells did not reduce plaque burden [
41], but induced neuronal damage instead. In our study, APP-Tg mice on SD exhibited activated microglial cell accumulation in the brain which is consistent with a previous report of a pro-inflammatory state in these mice [
42]. Compared to SD-fed mice, microglial cell accumulation was increased by about fourfold in HFD-fed mice, and this correlated with increased pro-inflammatory cytokine expression in these mice. This increase in both CNS and peripheral inflammation was reversed when mice were removed from HFD and put back on SD; we observed decreased activated microglial cell numbers, decreased liver pathology, and decreased pro-inflammatory cytokines in the CNS. This reversal in inflammation was consistent with reduced Aβ plaque burden. This indicates that induction of NAFLD in the liver may be an important factor that can induce the accelerated signs of AD observed in APP-Tg mice.
Interestingly, WT mice on HFD exhibited a similar increase in activated microglia accumulation even though they did not exhibit signs of Aβ plaque deposition at 2 or 5 months on HFD. WT mice on HFD exhibited more pronounced lipid deposits in liver cells than APP-Tg mice, but they did not develop Aβ plaques in the acute phase of NAFLD. WT mice fed with HFD showed increased pro-inflammatory cytokine gene expression (TNF-α, Il-6 and IL-17) in the liver and in peripheral lymphoid organs, and liver pathology was more pronounced in WT mice, with higher immune cell infiltration in the liver parenchyma. Both IL-6 and Il-17 pro-inflammatory cytokines were highly pronounced in WT mice but were almost absent in SD controls. Il-6 and Il-17 play pathogenic roles in multiple sclerosis [
50,
51]. The profound increase in the expression of these cytokine genes strictly coincides with increased activated microglia in the CNS, potentially setting the stage for the neurodegeneration we observed in WT mice on HFD 1 year later. Moreover, when WT mice were removed from HFD diet and placed on SD, inflammation ceased, liver pathology was reverted, and activated microglial cells were absent. Of interest was the absence of plaques in these WT mice even though they showed all other signs of inflammation observed in the APP-Tg animals. This indicates that HFD-induced systemic inflammation primed the CNS for the neurodegeneration which we later observed. These studies suggest that AD can be induced and driven by acute-chronic systemic inflammation in individuals that are not otherwise genetically predisposed. It also indicates that early intervention can reverse the process.
The brains of APP-Tg mice on long-term HFD had significantly less plaques compared to SD-fed mice. APP-Tg mice showed a dramatic decrease in Aβ plaques in the hippocampus, the olfactory bulb, and the mid brain. Also striking was the absence of astrocytic tracks in brain areas where Aβ plaques were absent or diminished, such as the hippocampus and midbrain. Moreover, we observed neuronal loss and increased apoptotic neurons in hippocampus and mid brain. This was not the case for older mice (15–24 months) on SD, indicating that aging alone is not responsible for the loss of astrocytes. As discussed above, removal from HFD reduced plaque load and decreased inflammatory signals. Thus, astrocytic loss may be a result of accelerated AD pathology resulting in CNS toxicity. Astrocyte death can lead to neuronal death because of the critical role they play in neuronal function [
42]. Increased neuronal death may also explain the decreased plaque load in chronic AD due to decreased APP production. Based on these studies, we conclude that chronic inflammation induced outside the CNS (in the liver) is sufficient to induce AD-like symptoms in the absence of predisposing genetic factors.
Signs of CAA and hyperphosphorylated-Tau (pTau) expression are representative of advanced AD. pTau aggregates and is unable to signal proper neurotransmission, leading to neuronal dysfunction; and CAA results in a lack of Aβ clearance from the CNS. We observed strong expression of pTau and evidence of CAA in both WT and APP-Tg mice on HFD but not in SD controls. pTau was increased in HFD-fed WT and APP-Tg mice compared to SD controls, even though Aβ was induced much later in WT mice. Thus, if CAA and tauopathy represents advanced AD, one may argue that AD accelerated more dramatically in WT mice than APP-Tg mice, and it may be regulated differently in genetically predisposed (APP-Tg) vs. non-predisposed individuals (WT). CAA involves deposition of amyloid in cerebral vasculature and is a hallmark of advanced AD resulting in pathological changes in cerebral blood vessels referred to as vasculopathies. Signs of CAA were more extensive in APP-Tg mice compared to WT mice on HFD. These findings clearly indicate that HFD-induced inflammation can result in significant CNS destruction and pathology over a lifetime.
Low-density lipoprotein receptor-1 (LRP1) is involved in a number of pathways linked to AD pathogenesis and has multiple functions. It is most highly expressed in brain, liver, and lungs. In the brain, it is highly expressed in glial cells, neurons, and cells of the cerebral vasculature. LRP1 can directly regulate gene expression through its intracellular domain and can regulate the endocytosis of many diverse ligands including ApoE, APP, and Aβ. LRP1 appears to have bimodal opposing functions linked to AD pathogenesis. It is involved in Aβ clearance and Aβ production. It mediates Aβ clearance by cellular uptake followed by lysosomal degradation and/or transcytosis of intact Aβ across the BBB to the circulation and consequent peripheral clearance [
49]. We observed increased LRP1 expression in APP-Tg mice on HFD, but not in WT mice, during the acute phase of NAFLD. Increased LRP1 expression may represent its increased function in clearance of Aβ from the CNS in APP-Tg mice during accelerated Aβ production. It is possible that LRP1 did not increase in WT mice on HFD due to the absence of Aβ in the CNS of these mice during acute NAFLD. However, during chronic NAFLD, we observed a significant decrease in LRP1 in the CNS of both WT and APP-Tg mice on HFD. This decrease coincides with advanced signs of AD, including reduced Aβ plaque load, reduced number of neurons and astrocytes, and increased vascular destruction, suggesting that LRP1 plays a protective role in AD. It is possible that the reduction in LRP1 expression during chronic NAFLD may be a result of glial and neuronal cell loss, as these cells abundantly express LRP1. Alternatively, advanced AD may have rendered LRP1 defective in clearing Aβ from the CNS.
It is interesting that no change in ApoE expression was observed during acute or chronic stages of NAFLD, despite significant neuronal and glial cell loss. The major function of ApoE is to transport cholesterol and other lipids in plasma and brain through a variety of cell surface receptors including LRP1 [
49]. Astrocytes are the main source of ApoE [
25,
52]. Since cholesterol is a critical component of glial and neuronal cell membrane including the myelin sheath, it is possible that a reduction of GFAP
+ astrocytes resulted in limited cholesterol production needed for membrane synthesis and CNS repair. However, this is unlikely since such a reduction may have resulted in decreased ApoE. Other likely scenarios for no change in ApoE include its function in cholesterol trafficking was unhindered, whereas its function in Aβ trafficking was defective, or that the reduced plaque burden in chronic NAFLD was the result of increased Aβ trafficking by ApoE. The involvement of Aβ, LRP1, and ApoE in neurodegeneration and AD pathogenesis is quite complex and needs further investigation beyond these studies.
Our studies address a growing problem in our society that relates to metabolic syndromes due to diets high in fat/lipid consumption and their impact on neurodegeneration. NAFLD is prevalent in as much as a third of the world’s population. Similarly, AD frequency is rapidly growing. We showed that a modest increase in dietary lipid content caused increased systemic inflammation followed by increased neuro-inflammation and accelerated AD signs in APP-Tg mice. More importantly, we showed that WT mice become susceptible to developing Aβ plaques after long-term HFD intake and developed advanced cellular signs of AD. Moreover, APP-Tg mice on HFD exhibit severe CNS damage stemming from the effects of chronic HFD. These findings highlight a growing problem in our society whereby consumption of foods high in lipids over a lifetime can have detrimental consequences such as accelerating signs of AD in potentially susceptible individuals or inducing them in those that are not susceptible. An important and critical finding of these studies is that change from HFD to SD, before irreversible CNS damage sets in, completely reverses signs of AD. This suggests that life style changes such as reducing one’s lipid/fat intake can have a profound impact on disease outcome. Recent studies from others showed that a high-cholesterol diet (5 %) fed to mice can induce advanced pathological signs of AD [
53‐
55]. Similar to our findings, these studies showed that diets high in cholesterol increased hyperphosphorylated-Tau deposition, and decreased cognitive function in both WT and AD models [
53‐
55]. Also, one of the studies showed increased ventricular volume which is reminiscent of the decreased cortical thickness and increased neuronal apoptosis we observed in HFD-fed WT and APP-Tg mice [
55]. These findings are in line with ours and highlight a critical role for a diet high in fats and lipids in inducing pathological outcomes. However, our model differs in the cholesterol content. We fed mice a diet containing 1 % cholesterol plus 18 % triglyceride which consistently induces NAFLD that is characterized by acute inflammation and significant liver damage in mice.
Competing interests
The authors declare no financial competing interests.
Authors’ contributions
MSB, proposed the hypothesis, directed the research, and helped write the manuscript. DGK and AK performed the experiments, analyzed the data, and helped with preparing and writing the manuscript. LT performed the experiments and helped with editing the manuscript. KM proposed ideas and assisted with the NAFLD induction and characterization. AY performed the experiments and helped with editing the manuscript. SAR performed the experiments and helped with editing the manuscript. LT helped with editing the manuscript. All authors read and approved the final manuscript.