Background
Neuropathological hallmarks of Alzheimer’s disease (AD) are extracellular senile plaques, composed primarily of amyloid peptides (Aβ), and intracellular neurofibrillary tangles of hyperphosphorylated tau protein [
1]. Aβ is derived from sequential proteolytic processing of its membrane precursor Amyloid Precursor Protein (APP) by the β- and γ-secretases [
2]. Although transgenic mouse models expressing the mutations of APP and components of the γ-secretase found in familial AD reproduce amyloid pathology, there are currently no animal models mimicking sporadic AD [
3].
Cholesterol is increasingly linked to AD pathology [
4]. It is increased in AD brains [
5,
6] and we found by Time of flight –Secondary Ion Mass Spectrometry (Tof-SIMS), that a 30% cholesterol increase could be observed in AD brain samples most likely in various cell types [
7]. Moreover, increase in membrane-associated free cholesterol were correlated with the severity of the disease as opposed to intracellular cholesterol, cholesterol from the extracellular space or from the senile plaques [
5]. Additionally, the ϵ4 allele of the
APOE gene encoding apolipoprotein E, the transporter of cholesterol in the brain, is the most important risk factor of AD [
8]. APP as well as β and γ-secretases are residents of cholesterol-enriched membrane microdomains termed lipid rafts [
9,
10]. Levels of cholesterol control the partition of APP and its secretases in lipid rafts [
11,
12] as well as APP internalization and Aβ production [
11,
13]. Moreover, a binding site for cholesterol in APP has been described [
14].
Despite this substantial body of literature, it still remains unclear whether sporadic AD could be initiated by a disruption of cholesterol metabolism leading to a change in membrane cholesterol of neurons. To test this hypothesis, we triggered an acute increase of cholesterol at the membrane of neurons and assessed whether cellular changes similar to those detected early in the development of the disease could be observed. A 30% membrane cholesterol increase was produced to mimick what has been observed in AD brain samples. We did not use 3-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) inhibitors such as statins, since they lead to non-specific effects via their action on isoprenoids and inflammation, or drugs such as U1866A that lead to accumulation of cholesterol in the lysosomal pathway [
15‐
17]. We did not use uptake of LDL-cholesterol complex either since they can directly affect the endo-lysosomal pathway [
18]. Instead we used methyl-beta-cyclodextrin/cholesterol complex (MβCD-cholesterol) which can deliver cholesterol directly at the plasma membrane of cultured neurons [
12,
19]. After cholesterol increase we analyzed cellular phenotypes that correlate with sporadic AD progression. Gene expression changes were reported to correlate with AD pathology with a switch occurring at Braak stage III, when profound modifications in amyloid pathology take place [
20]. Enlargement of the endosomal compartment was described as one of the earliest phenotypes of AD, present before the formation of plaques and absent in other neurodegenerative diseases [
21]. Aβ was found to accumulate in these enlarged endosomes [
22]. This phenotype occurs more frequently in individuals with the ϵ4 allele of
APOE, suggesting a possible link with cholesterol metabolism [
21]. Another phenotype associated with sporadic AD is axonal vesicular transport deficits, described in samples from individuals affected by the disease [
23] and in AD mice models [
24].
In this study, using our published experimental model for loading the plasma membrane of neurons with cholesterol to reach an increase of 30%, corresponding to levels detected in AD brain samples, we could recapitulate cellular phenotypes from early stages of the disease, suggesting a direct causal link between high cholesterol in the brain and cellular AD pathogenesis. We propose membrane cholesterol accumulation in cultured neurons to be a useful cellular model of AD.
Conclusion
Membrane cholesterol has been shown to be increased in post-mortem brains from sporadic AD patients and to correlate with disease progression [
5]. We suggested that an increase of membrane cholesterol could be an early event in the etiology of sporadic AD. To test this hypothesis, we acutely increased levels of plasma membrane cholesterol of neurons in culture. This treatment induced a 30% increase mimicking what was observed in AD brains [
5‐
7]. We showed that this transient membrane cholesterol increase triggered Aβ42 over-production, endosomal enlargement, vesicular transport deficits in neuronal processes and gene expression modulation as in early sporadic AD. This model of neuronal primary cultures treated with cholesterol could thus be relevant to study early events in sporadic AD.
Using both confocal and electron microscopy, we investigated the number, surface, ILVs number and aggregation status of EEA1-positive early endosomes after cholesterol treatment. This is to our knowledge the first time that neuronal endosomal enlargement is investigated in such quantitative detail. After membrane cholesterol loading, the number of early endosomes per neuron was not altered but we found that their size was enlarged and that more ILVs were associated with each endosome. EEA1-positive endosomes in neurons loaded with cholesterol were also more prone to form aggregates. As cholesterol is highly concentrated in myelin [
47], excess cholesterol in sporadic AD brains could result from demyelination linked to age, the major risk factor for AD. Indeed, genes involved in cholesterol synthesis were down-regulated in a mouse demyelination model [
48], as we also observed in our model of external cholesterol addition. It still remains unclear whether myelin degradation could be the origin of higher membrane cholesterol that is described as a marker of disease progression [
5]. The topography of cholesterol embedded in the membrane seems to be crucial as was highlighted in an AD mouse model where the cholesterol content of lipid rafts, but not the total brain cholesterol levels, was correlated with amyloid load and behavioral deficits [
49]. The link between tau pathology, the other hallmark of AD, and cholesterol is still unclear. Here we show that membrane cholesterol increase leads to overexpression of the kinase Fyn gene using microarray gene profiling and RT-QPCR (Table
2). Tau is known to interact with Fyn and to facilitate Fyn targeting to dendrites. Once in the postsynaptic compartment, Fyn phosphorylates the
N-methyl-D-aspartate (NMDA) receptor subunit 2B (NR2B) and stabilizes its interaction with postsynaptic density protein 95 (PSD95) (reviewed in [
50]. The NR2B/PSD95 interaction is essential in mediating Aβ-induced excitotoxicity [
51]. Fyn could thus be an interesting link between cholesterol, Aβ and tau.
In conclusion we have shown that an increase in neuronal membrane cholesterol triggers APP processing, endosomal trafficking and axonal transport abnormalities and induces gene expression changes that are reminiscent of early stages of sporadic AD. We propose that an increase in membrane cholesterol linked with age is one of the initial events that could trigger sporadic AD. Thus specifically decreasing neuronal membrane cholesterol could be an interesting therapeutic strategy, which has already been successfully applied in mouse models of AD [
49,
52]. It was recently shown that stimulating cholesterol synthesis in mouse models overexpressing amyloid induces tau phosphorylation and neurofibrillary tangles formation, suggesting that cholesterol increase might trigger amyloid and tau pathologies [
53]. Loading the membrane of cultured neurons with cholesterol could thus be used as a new cellular model to study early AD changes, identify new targets and screen new molecules.
Methods
Plasmids and reagents
The APP751 plasmid was a kind gift from Dr. Frederic Checler (IPMC, Valbonne, France). The APP-mCherry plasmid was generated by introducing the APP751 sequence in the pmCherry-N1 vector (Clontech, Mountain View, CA, USA) at the XmaI/AgeI restriction site. MβCD-cholesterol complex, saponin, bovine serum albumin (BSA), sucrose and poly-L-lysine were purchased from Sigma-Aldrich (Saint-Louis, MO, USA). The antibody directed against EEA1 (Early Endosome Antigen 1) was from Cell Signaling Technology (Danvers, MA, USA). Goat anti rabbit IgG coupled to Alexa568 was from Life Technologies (Carlsbad, CA, USA).
Primary neuronal cultures
Primary hippocampal and cortical cultures were prepared from E16-18 OFA Sprague Dawley rat embryos (Charles River, Wilmington, MA, USA). Hippocampi and cortices were dissected in cold PBS supplemented with 45% glucose (Sigma-Aldrich, Saint-Louis, MO, USA). Digestion was performed in a 0,05% solution of Trypsin-EDTA (Life Technologies, Carlsbad, CA, USA) for 25 min at 37°C. Tissues were then mechanically dissociated in Dulbecco’s Modified Eagle Medium-1 (Life Technologies, Carlsbad, CA, USA) supplemented with 5% fetal bovine serum (Life Technologies, Carlsbad, CA, USA) and centrifuged for 10 min at 800 rpm. Neurons were resuspended in neurobasal medium supplemented with 2% B27, 2 mM glutamax, 1% penicillin/streptomycin (all from Life Technologies, Carlsbad, CA, USA) and counted.
Cholesterol modulation
Unless otherwise mentioned, neurons were washed twice with neurobasal medium, treated with 1.4 mM MβCD-cholesterol dissolved in neurobasal medium and then washed three times with neurobasal medium. Treatment of neurons with 1.4 mM MβCD-cholesterol for 30 min resulted in an increase of 28.4 ± 6.0% of cellular cholesterol levels, as assessed previously by filipin staining [
12].
Aβ 38, 40 and 42 measurements
Cortical neurons were plated on 12-well plates coated with poly-L-lysine (1 mg/ml) at a density of 2 million neurons per well and maintained at 37°C in a humidified 5% CO2 atmosphere. After cholesterol treatment at DIV4-6, neurons were placed in fresh neurobasal medium supplemented with 2% B27, 2 mM glutamax for 24 h. Supernatants were collected on ice in polypropylene tubes (Corning, Corning, NY, USA) containing a protease inhibitor cocktail (Roche, Penzberg, Germany) and were then stored at -80°C. Concentrations of the Aβ38, Aβ40 and Aβ42 species of β-amyloid peptide were measured by multiplex Electro-Chemiluminescence Immuno-Assay (ECLIA). Assays were performed according to the manufacturer's instructions. Briefly, samples were analyzed using Meso Scale Discovery (MSD) SECTOR™ Imager 2400 (Meso Scale Discovery, Gaithersburg, MD, USA), with the Rodent Aβ triplex kit (also from MSD); carbon 96-well plates contained in each well four capture spots, one of which was blocked with BSA (as standard curve control), and the three others coated with isoform specific anti-Aβ antibodies specific for Aβ38, Aβ40, Aβ42, respectively. 100 μl of blocking buffer solution were added to all wells to avoid non-specific binding. The plates were then sealed, wrapped in tin foil, and incubated at room temperature on a plate shaker (600 rpm) for 1 h. Wells were then washed three times with washing buffer, and 25 μl of the standards (Aβ38, Aβ40, Aβ42) and samples were then added to the wells, followed by an Aβ-detecting antibody at 1 μg/ml (MSD) labelled with a Ruthenium (II) trisbipyridine N-hydroxysuccinimide ester; this detection antibody was 4G8 (which recognizes the epitope Aβ18-22 of the human and rodent peptide). Plates were then aspirated and washed 3 times. MSD read buffer (containing TPA) was added to wells before reading on the Sector Imager. A small electric current passed through a micro-electrode present in each well producing a redox reaction of the Ru2+ cation, emitting 620 nm red light. The concentration of each Aβ isoform was calculated for each sample, using dose–response curves, the blank being cell-less culture medium.
EEA1 Immunocytochemistry and confocal microscopy
Cells cultured on poly-L-lysine-coated coverslips were fixed at DIV4-5 or DIV14 using a solution of 4% paraformaldehyde and 4% sucrose in Phosphate Buffered Saline (PBS) for 20 min at room temperature. Cells were then washed twice in PBS and incubated with NH4Cl (50 mM in PBS) for 10 min. Cells were washed twice again in PBS. They were permeabilized with solution A (0.3% BSA and 0.05% saponin in PBS) for 45 min at 37°C and were then incubated at room temperature for 1 h with primary antibody against EEA1 diluted (1/1000) in solution A. Cells were subsequently incubated at room temperature for 1 h with goat anti rabbit secondary antibody conjugated to Alexa568 diluted in solution A (1/1000). Coverslips were mounted in Fluoromount-G (SouthernBiotech, Birmingham, AL, USA). Z-stacks of neurons were acquired on a Fluoview FV1000 confocal microscope (Olympus, Tokyo, Japan) with the 543 nm line of a He/Ne laser. Fluorescence was collected with a 60× plan apochromat immersion oil objective (NA 1.35) between 560–660 nm. The mean endosome size and the mean endosome number per neuron were analyzed with ICY software [
54]. Between 13 and 17 neurons were analyzed for each experiment. Each experiment was independently repeated 3 times.
Pre-embedding immunoperoxidase electron microscopy
Neurons grown on Thermanox coverslips were processed through all stages in situ. They were fixed with 4% PFA, 0.1% glutaraldehyde diluted in PBS, rinsed with PBS, cryoprotected in 30% glycerol, 30% ethylene glycol in PBS and stored at -20°. After PBS rinses, they were blocked in 5% normal goat serum and incubated at room temperature overnight with the antibody against EEA1 diluted 1/1000 in PBS. A biotinylated anti-rabbit IgG (Vector, CA, USA) was applied as secondary antibody (1/200 in PBS, 2 h), followed by ABC peroxidase complex (Vectastain Elite, Vector, CA, USA) with 0.05% diaminobenzidine as chromogen. After 2% OsO4 post-fixation and dehydration in graded acetone including a 1% uranyl staining step in 70% acetone, coverslips were embedded in Epon resin. Thin (70 nm) sections were lightly stained with lead citrate and observed under a Philips CM120 electron microscope (Philips, Eindhoven, The Netherlands) operated at 80 kV. Images were recorded with a Morada digital camera (Olympus Soft Imaging Solutions GmbH, Münster, Germany). The measurements were performed with the associated iTEM software.
Videomicroscopy
Cortical neurons in suspension (5 million) were electroporated with Amaxa Nucleofector kit for rat neurons (Lonza, Basel, Switzerland) according to the supplier’s manual. Electroporated neurons were plated on 8-well Labteks coated with poly-L-lysine (1 mg/ml). After 3 hours, medium was replaced with fresh neurobasal supplemented with 2% B27, 2 mM glutamax and 1% penicillin/streptomycin. Neurons were maintained at 37°C in a humidified 5% CO2 atmosphere. Sixteen hours before performing video experiments, 10 μM forskolin (Sigma-Aldrich, Saint-Louis, MO, USA) and 100 μM IBMX (Sigma-Aldrich, Saint-Louis, MO, USA) were added to the medium. Videomicroscopy experiments were performed at DIV3-5. APP-containing vesicles were imaged in control condition and then for 35 minutes after adding neurobasal medium alone or MβCD-cholesterol dissolved in neurobasal medium (1.4 mM final concentration). Live videomicroscopy was carried out using an Axiovert 200 microscope (Zeiss, Jena, Germany) with a PL APO oil × 63 objective of numerical aperture of 1.40. Records were made with a Photometrics Evolve 512 camera (Roper Scientific, Trenton, NJ, USA) controlled by Metamorph software (Molecular Devices, Sunnyvale, CA, USA). Stacks were acquired at 37°C. Images were collected in stream set at 1 × 1 binning with an exposure time of 200 ms. Kymographs were generated and analyzed with a homemade ImageJ plugin, KymoToolbox, available upon request (contact: Fabrice.Cordelieres@curie.fr). Segmental velocities were defined as the speed a particle travels in one direction without a pause or a reversal in the direction of movement.
cRNA probe preparation and hybridization
Hippocampal rat neurons (1.1 to 1.4 million) were treated with MβCD-cholesterol at DIV 4–6 and left in neurobasal medium for 4 h and 30 min. They were then harvested and centrifuged. The dry pellet was kept at -80°C. Total RNAs were extracted using Nucleospin RNA II kit (Macherey Nagel, Duren, Germany) in accordance with the manufacturer’s protocol. The quality and quantity of each RNA preparation were assessed on an Agilent 2100 Bioanalyzer with RNA 6000 NanoChips (Agilent Technologies, Santa Clara, CA, USA). One hundred ng of each RNA were amplified and labeled with Cy3 using the Low Imput Quick Amp labeling kit (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer’s instructions. After purification and quantification on a Nanodrop (ThermoFisher Scientific, Waltham, MA, USA), 2 μg of each Cy3-cRNA were hybridized overnight on Whole Rat Genome Microarray 4×44 K (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer’s instructions.
Microarray data analysis
Microarray data were acquired on a ScanArray GX (Perkin Elmer, Waltham, MA, USA) with a resolution of 5 μm and analyzed with Mapix 5.0.0 software (Innopsys, Carbonne, France). For each sample, raw data consisted of the Median Feature Intensity – Median Background Feature (F-B) at 532 nm wavelength. These raw data were log2-transformed and quantile-normalized under the R freeware (
http://www.r-project.org). The statistical analysis was performed on these normalized data under the R freeware. All the microarray data have been deposited on the GEO database under the accession number GSE46221.
Quantitative PCR (qPCR)
Three hundred nanograms of each RNA were individually reverse-transcribed into cDNAs for 2 hours at 42°C using the Maxima First strand synthesis kit (Fermentas GmbH, Germany) according to the manufacturer’s instructions. qPCR assays were performed in a Lightcycler® 480 System (Roche), in the presence of 200nM of each primer, 100nM of specific hydrolysis probe (designed with Universal Probe Library, Roche Applied Science) and 1X Solaris qPCR Master mix (Fermentas GmbH, Germany). Gene expression was normalized using HPRT and pPib as reference genes.
Statistical analysis
Most statistical calculations were performed using GraphPad Prism software (version 5.0). All the data are given as mean ± SEM.
κ-means analysis of fluorescently-labelled endosome size was performed using the statistic analysis toolbox of Matlab software (version R2012a). Endosome sizes under control conditions were distributed into three size classes defined by the κ-means algorithm using the squared Euclidian distances. This iterative partitioning minimizes the sum, over all clusters, of the within-cluster sums of point-to-cluster-centroid distances. Each endosome was assigned an index corresponding to one of the three clusters defined in the iterative loop (small, medium, large). The partition of endosome sizes after cholesterol treatment in these three classes was then determined.
The χ
2 comparison test of the two distributions of endosome size was made with Matlab sofware (version R2012a) using the following formula:
with i (1 to 3) as the size categories (small, medium, large) and j (1 to 2) as the two groups (control and treated). n is the number of neurons falling into each class. The degree of freedom was (ni-1) × (nj-1) = 2.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
CM and MCP designed research. CM, LH, CLN and CB performed primary cultures and electroporation. CM prepared Aβ samples from primary cultures. NP performed Aβ dosage by MSD. CM, LH and ML performed immunocytochemistry and confocal microscopy. CM and LH analyzed endosomes number and size with ICY software. JL performed electron microscopy and quantified the size of endosomes and number of ILVs. CM performed video-microscopy experiments and analyzed the data. CM and LD prepared RNA samples from primary cultures. LD performed analysis of transcriptomics data. KB provided transcriptomics data from patients. CM, FC and LD performed statistical analysis. CM and MCP wrote the paper with critical evaluation by JL, LD, KB, NP, FS, GT, JNO, NC and CD. All authors read and approved the final manuscript.