Introduction
Alzheimer disease (AD), the most common form of aging dementia, is characterized by problems with memory, thinking and behavior [
50]. These clinical features are strongly associated with the accumulation of two types of insoluble protein deposits in the AD brain, which are composed of either the amyloid-β (Aβ) peptide or the microtubule-associated protein tau and impair neuronal function at many levels [
5,
32,
44,
50]. The Aβ deposits are referred to as amyloid plaques and are found in the interstitial space of the brain, whereas the lesions composed of aggregated tau, known as neurofibrillary tangles (NFTs), are intraneuronal [
5,
32,
44,
50]. Tau pathology progresses through well-defined stereotyped stages, which appears to be initiated in the locus coeruleus and slowly spreads via the entorhinal cortex and hippocampus to the neocortex [
12,
13]; however the role of the locus coeruleus is controversial [
4]. This pattern of tau spreading led to the suggestion that AD progression occurs by neuron-to-neuron transmission involving trans-synaptic transport of seeds of tau aggregation from affected to anatomically interconnected recipient neurons [
12,
13]. It has since been established that the intercellular transfer of misfolded forms of tau known as “seeds” contributes to the progression of AD, with tau seeds acting in a manner similar to prions, triggering the robust conversion of soluble tau into insoluble large filamentous aggregates and NFTs [
14,
30,
50].
Several modes of neuron-to-neuron transfer of tau seeds have been described, including via extracellular vesicles such as exosomes [
22,
51,
66], trans-synaptically mediated transfer of tau aggregates between interconnected neurons [
15,
23], tunneling nanotubes [
61] or the uptake of free-floating tau aggregates and fibrils [
30,
35]. In vitro evidence suggests that reducing the pool of extracellular tau seeds, irrespective of whether these are moving freely or are transported by exosomes or any other mechanism of inter-neuronal transfer, results in an in vivo reduction of tau pathology by maintaining the level of extracellular tau seeds below a pathological concentration threshold [
5,
15,
29,
30,
36,
51,
61]. Our research focuses on exosomes, membranous secreted nanovesicles 30–150 nm in size, that are produced in late endosomes by the inward budding of the endosomal membrane, which is progressively pinched off to generate and accumulate intraluminal nanovesicles [
11,
38,
45]. The late endosome, loaded with intraluminal nanovesicles, then gradually develops into large multivesicular bodies (MVBs). These MVBs can fuse with the plasma membrane to release the intraluminal nanovesicles into the extracellular environment, and once secreted these free nanovesicles are termed “exosomes” [
11,
38,
45].
A number of studies have shown that exosomes can transport Aβ and derivatives of the amyloid precursor protein (APP) from which Aβ originates [
48,
52,
58]. They also contain phosphorylated tau as demonstrated for exosomes that have been isolated from the blood and cerebrospinal fluid of AD patients [
26,
55]. Furthermore, immuno-electron microscopy of AD brain tissue has revealed that human Aβ plaques are enriched in exosomal proteins [
52]. Mouse models of AD have been instrumental in demonstrating that exosome reduction in vivo is associated with a lower Aβ plaque load in the brain [
20,
21]. Similarly, depletion of microglia and inhibition of exosome synthesis has been found to halt tau propagation in the brains of tauopathy mouse models [
3]. Taken together, these studies support the notion that reducing exosome secretion results in reduced Aβ plaque formation and tau propagation. Related to this, we have demonstrated that tau seeds are contained within exosomes isolated from the brains of tauopathy mice, that they have a distinct phosphorylation pattern, and that only exosomes derived from cells undergoing tau aggregation are able to seed and corrupt soluble tau in recipient cells, a phenomenon that occurs in a threshold-dependent manner [
6,
51].
An important question in the field is how the seeds are taken up and handled by recipient cells. Here, neuron-to-neuron transmission of exosomes emerges as an important pathomechanism for the progression of AD. Such a mechanism implies that a neuron generates exosomes in endosomes, an organelle which is more abundant in the soma than in axons [
65], after which the mature MVBs undergo anterograde transport along the axons until they fuse with the plasma membrane to release the exosome at the synapse of an interconnected cell. Evidence for such a trans-synaptic mechanism has been provided by studies in
Drosophila which investigated exosomes carrying Wnt signals at the neuromuscular junction [
41,
42]. In our study, we used simple microfluidics circuit systems to demonstrate that exosomes are not only being exchanged between interconnected neurons A and B, but that a recipient neuron C can receive exosomes that have either been generated by an interconnected neuron B or are passed on via this interconnected neuron after processing of ‘exogenous’ exosomes that have been internalized from neuron A. This ‘longer-distance action’ of exosomes appears to be linked to the hijacking of secretory endosomes present in neuron B of this simple circuit. We discuss how such fusion events potentially increase the pathogenic potential and the radius of action of pathogenic cargoes carried by exogenous exosomes.
Materials and methods
Mouse strains and collection of brain tissue
C57BL/6 mice were used at embryonic day 17 (E17) to isolate hippocampal neurons for tissue culture experiments. rTg4510 mice expressing human four-repeat tau with the P301L mutation linked to hereditary tauopathy [
56] were used at 4–6 months of age for exosome isolation. Animal experimentation was approved by the Animal Ethics Committee of the University of Queensland (approval number QBI/412/14/NHMRC).
Isolation and purification of brain exosomes
Exosomes were isolated from the interstitial space of mouse brains using a previously established protocol [
48,
51]. In brief, each brain was dissected and gently chopped before being incubated in 7 ml of 0.2%
w/
v Collagenase type III (LS004182, Worthington) in serum-free Hibernate-A medium (A12475–01, Life Technologies) for 30 min at 37 °C. The dissociation reaction was stopped with 14 ml of ice-cold Hibernate-A containing 1× Complete protease inhibitor cocktail (Roche), 50 mM NaF and 200 nM Na
3VO
4. The tissue was then gently dissociated with a 10 ml pipette, keeping the cells intact during pipetting them up and down, followed by a series of differential 4 °C centrifugations at 300 g for 10 min, 2000 g for 10 min and 10,000 g for 30 min to sequentially discard the pellet containing cells, membranes, and nanodebris, respectively. The supernatant from the final centrifugation step was passed through a 0.22 μm syringe filter (Millex-GP, Millipore) and centrifuged at 120,000 g for 70 min at 4 °C to pellet the exosomes. The pellet containing the exosomes was then washed with 5 ml phosphate-buffered saline (PBS, 17-516Q, Lonza), after which the pellets from five mouse brains per genotype (25 ml) were pooled. This preparation was centrifuged at 120,000 g for 70 min at 4 °C to obtain a pellet that was resuspended in 2 ml of 0.95 M sucrose in 20 mM HEPES (15630–080, Life Technologies), then purified using a sucrose step gradient column (six 2 ml steps at 2.0, 1.65, 1.3, 0.95, 0.6 and 0.25 M sucrose from bottom to top). The sucrose gradient was centrifuged at 200,000 g for 16 h at 4 °C. The exosome-containing fraction 3 (0.95 M; ρ = 1.12 g/ml sucrose) was collected together with the interphase and resuspended in 6 ml ice-cold PBS, followed by a 120,000 g centrifugation for 70 min at 4 °C. Finally, the sucrose-purified exosome pellet was resuspended in 100 μl PBS containing 1× Complete protease inhibitor cocktail (Roche). A 10 μl aliquot of exosomes in PBS was homogenized with 10 μl of 2× RIPA buffer (300 mM NaCl, 100 mM Tris-HCl pH 7.4, 0.50% (
w/
v) sodium deoxycholate, 0.2% (
v/v) Nonidet P-40) to determine the protein content with a Micro BCA™ Protein Assay Kit (23,235, Thermo-Fisher).
Fluorescence labeling of exosome membranes
To track exogenous exosomes isolated from mouse brains, we labeled their membranes with an appropriate fluorescent stain that stably incorporates a fluorescent dye with long aliphatic tails into the exosome membrane. In our study, the fluorescent membrane probes CellVue® Claret Far-Red Fluorescent Membrane Linker (Sigma), PKH67 Green Fluorescent Membrane Linker (Sigma) and FM™ 1–43FX Fixable Membrane Stain (Thermo-Fisher) were used to separately label sucrose-purified exosomes according to the manufacturer’s instructions. The labeling reaction was stopped with 6 ml of 2% bovine serum albumin, followed by ultra-centrifugation at 100,000 g for 70 min, washing with PBS and another round of ultra-centrifugation, followed by resuspension of the fluorescently labeled exosomes in PBS.
Primary neuronal culture and microfluidic devices
Hippocampal neurons were isolated by standard methods using C57BL/6 mice sacrificed at E17 and grown in culture chamber microfluidic devices (Xona Microfluidics) placed on 24 × 60 mm coverslips (#1,5 Menzel-Glaser) that had been coated with poly-D-lysine (PDL), to form a non-plasma bond with the device. 60,000–80,000 neurons were plated per chamber using Neurobasal medium (21,103,049, Thermo-Fisher) supplemented with 5% fetal bovine serum (FBS; Hyclone), 2% B27 (17,504,044, Thermo-Fisher), 1 mM GlutaMAX (35,050,061, Thermo-Fisher) and 50 U/ml penicillin/streptomycin (15,070,063, Thermo-Fisher). The medium was changed to serum-free Neurobasal medium minus phenol red (12,348,017, Thermo-Fisher), supplemented with 28 nM 2-mercaptoethanol (21,985,023, Thermo-Fisher) 24 h post-seeding, and half of the medium was changed twice a week. A hydrostatic pressure gradient that prevents diffusion between culture chamber (Ch) 1 and Ch2 was established by adding twice the volume of culture medium to Ch2. To perform electron microscopy, the microfluidic devices were placed on plastic culture dishes coated with PDL. All cultures were maintained at 37 °C and 5% CO2 for up to 12 days. Neurons grown in the microfluidic devices were analyzed at 8–12 days in vitro (DIV8–12).
Plasmids, virus preparation and electroporations
CD9 is a tetraspanin that is expressed on plasma, endosomal and exosomal membranes [
2,
8,
64]. To label and track exosomes, dispersed primary hippocampal neurons were transfected with the plasmids mCherry-CD9–10 (Addgene # 55013) and Dendra2-CD9–10 (Addgene #57705), kind gifts from Dr. Michael Davidson to Addgene. For transfection, 4 × 10
6 neurons were electroporated with 4 μg plasmid DNA using a Nucleofector™ 2b device and the Amaxa Basic Primary Neurons Nucleofector® Kit (VPI-1003, Lonza). After electroporation, the neurons were resuspended in FBS-containing Neurobasal plating medium, centrifuged for 5 min at 100 g and then resuspended in Neurobasal plating medium to obtain a concentration of 8000 neurons per μl before seeding the neurons in the chambers of the microfluidic device. To detect endosomes, neurons were also transduced with a commercial baculovirus expressing the late endosomal marker LAMP1 (lysosomal-associated membrane protein 1) tagged with RFP (Thermo, C10597). After 48 h they were fixed in paraformaldehyde and imaged.
Confocal microscopy and image analysis
Fluorescence images at a 63× magnification were obtained with a Zeiss LSM 710 inverted laser scanning confocal microscope using a 1-2× optical zoom. For fluorescent particle quantification of endosomes, 4–10 non-overlapping confocal images at a 63× magnification were analyzed per sample using the open source ImageJ software (version 1.51r, Wayne Rasband, National Institutes of Health, Bethesda). The acquisition parameters remained invariable for all images. The fluorescence signal was adjusted by image segmentation applying a ‘Triangle’ threshold to specifically detect endosomal fluorescent particles in the green (Dendra2-CD9) and red (mCherry-CD9) channels. The thresholded binary images with particles were processed to reduce noise using the Image-J ‘Despeckle’ plugin, followed by the ‘Watershed’ filter to separate overlapping particles in the binary images. Particles were quantified in the segmented images with the ‘Analyze Particles’ ImageJ plugin using the parameter size 0.2–10 squared microns (μm [
2]) and circularity 0.1–1 (value of 1 indicating a perfect circle). Occasionally, fluorescent detection of CD9 in the plasma membrane generated fragments of neurites that were included as endosomal particles by ImageJ. Those inconsistencies were manually curated and excluded from the analysis using the ROI Manager. Then, the ‘Image Calculator’ plugin was used to multiply the segmented binary images from the green and red channels in order to quantify the number of endosomal particles that had both colors.
Super-resolution microscopy
Super-resolution microscopy was performed as a combination of Photo-Activated Localization Microscopy (PALM) using the fluorescent protein Dendra2 as the label for CD9 and direct Stochastic Optical Reconstruction Microscopy (dSTORM) using CellVue Claret. All super-resolution experiments were performed on an Elyra STORM/SIM microscope (Carl Zeiss, GmbH) equipped with a 100× oil-immersion objective, a focus lock system, an EMCCD Andor iXon Ultra 897 camera (Andor Technologies) and a super-resolution multiband dichroic and emission filter set (405/488/561/635-A-000,m Semrock). Neurons were imaged in highly inclined illumination mode at 20 kHz. Zen 2012 SP2 (black) software (Carl Zeiss, GmbH) was used for image reconstruction and channel alignment in the dual color experiments.
Electron microscopy and DAB photoconversion
Exosomes labeled with FM™1–43FX Fixable Membrane Stain (Thermo-Fisher) were added to primary hippocampal neurons grown in chamber 1 of microfluidic devices bound to 6 cm plastic culture dishes (Falcon, 353,002). 24 h after exosome uptake, the cultures were briefly washed 3× for 5 min with PBS to remove cellular debris and non-internalized exosomes, fixed with 4% paraformaldehyde (Sigma, 15,827) in PBS for 20 min, and again washed in PBS 3× for 5 min. The cultures were then incubated in 100 mM ammonium chloride solution (Sigma, 213,330) for 20 min, after which they were washed in PBS. Fixed cultures were incubated in 1.5 mg/ml diaminobenzidine (DAB, Sigma D5637) in PBS for 30 min at 4 °C. DAB forms a stable, insoluble precipitate that has a dark appearance and can be easily distinguished by electron microscopy [
33,
34]. We therefore photoconverted the FM™ 1–43FX stain into a DAB precipitate using a light box equipped with four 24-watt lamps (Atiaquaristik), two lamps at 420–460 nm and two at 400–500 nm, refrigerated with a cooling fan to prevent heating of the samples. Fluorescence bleaching and complete DAB photoconversion were confirmed by direct visualization using an axioscope (Zeiss).
For electron microscopy, samples were processed using protocols of the Biowave Pro+ (Pelco) microwave tissue processor maintaining the samples at room temperature. Biowave procedures included fixation in 2.5% glutaraldehyde (Proscitech, C003) in PBS for 3 min, osmication with 4% osmium (Proscitech, C010) and 3% potassium ferrocyanide (Sigma, P8131), and washing in dH2O 5× for 3 min, followed by incubation in 1% thiocarbohydrazide (Sigma, 88,535) for 20 min. After this, the samples were stained in 2% osmium tetroxide, washed in dH2O 5 × for 3 min and then stained in 1% uranyl acetate (Proscitech, C079) for 30 min. After 5 washes in dH2O for 3 min, the samples were stained in 20 mM lead aspartate (Proscitech, C151, Sigma, A9256) for 30 min at 60°C. Cultures were dehydrated through a graded series of ethanol followed by 100% acetone and then infiltrated into Epon resin (C038), followed by resin polymerization at 60 °C for 48 h. Resin-embedded samples were cut at 70 nm using a diamond knife (Diatome) and then collected onto 200 × mesh copper grids (Proscitech, GCu200) and imaged on a Hitachi HT7700 transmission electron microscope at 80 kV using an AXT 2Kx2K CMOS digital camera.
Discussion
Our study supports the notion that exosomes are invasive, hijacking the endosomal secretory machinery of the cells that internalized them to achieve a longer distance of action and a potentially higher pathogenicity. Previous studies have established that early endosomes arise from primary endocytic vesicles which fuse with each other to form a larger endocytic structure [
40]. It is likely that endocytic vesicles containing internalized exosomes fuse with endocytic vesicles containing different cargoes and that, during endosome maturation towards the generation of intraluminal nanovesicles, this results in a late endosome sheltering both endogenous vesicles and exosomes taken up from another cell. It can be assumed that most of the internalized exosomes are used by the cell itself, by either undergoing lysosomal degradation or recovery of some components via the trans-Golgi network [
39,
57], or by the release of exosomal contents into the cytosol via back-fusion mechanisms [
10,
63]. However, our data show that not all internalized exosomes are destined for destruction or recovery, as some are actually re-secreted together with endogenous exosomes.
The hijacking of endosomes is not unique to exosomes. For instance, some enveloped RNA viruses hijack MVBs in order to be released from infected cells [
17]. DNA viruses, such as the herpes virus, hijack the exosomal pathway of their host to facilitate virion assembly, promote the export of host proteins involved in immune regulation, and transfer viral-derived molecules that can assume activity in recipient cells [
54]. Even the bacterium
Chlamydia trachomatis has the capacity to hijack intracellular trafficking and lipid transport pathways of the host cell in order to promote infection [
47]. However, different from the mechanism that occurs with intracellular pathogens, the hijacking of endosomes as revealed in our study does not require genome replication as with viruses or bacteria.
Perhaps the hijacking mechanism most similar to that which we report is used by the proteinaceous lethal toxin known as anthrax which exploits MVBs. Anthrax toxins can persist in intraluminal nanovesicles for days, fully sheltered from proteolytic degradation in MVBs and can be delivered to the extracellular medium as exosomes [
1]. It is tempting to speculate that the fusion of exogenous exosomes with endosomes destined to secrete intraluminal nanovesicles would equally shelter the exogenous exosomes from proteolytic degradation in the MVBs which could eventually be secreted together with de novo generated exosomes. An alternative mechanism could be that, given the endocytic origin of exosomes, these contain markers for secretion and drive a subpopulation of exosomes into the secretory pathway. Comparison with the internalization of microvesicles, which are not of endocytic origin [
38,
45], might provide insight into this possibility.
Tau protein aggregation is a hallmark of AD and other neurodegenerative diseases collectively termed tauopathies. Tau accumulation in the brain hinders neuronal physiology at many levels, including axonal transport and synaptic transmission, mitochondrial and proteasomal functions, induction of endoplasmic reticulum stress and even nuclear effects on chromatin relaxation [
28,
50]. It is assumed that the ability of exosomes to carry misfolded or aggregated proteins significantly enhances the progression of tauopathies in a manner similar to what has been reported for prions [
6,
38,
50,
51]. However, it has been proposed that an increase in pathogenic exosomes could end up in traffic jams during endosome transport, which could cause a reduction of glutamate receptor recycling [
60]. Traffic jams could increase the number of intraluminal nanovesicles in the MVBs, thereby increasing the number of released exosomes; or increase the content of exosomes with AD-associated proteins like tau or Aβ, leading to an accelerated spread of disease [
60]. We observed that brain-derived exosomes are strongly internalized by neurons, resulting in a somata with high numbers of endosomes. Similarly, axons exhibited endosomes of varied sizes moving inside the axonal lumen, where sometimes massive endosomes were caught stretching the axonal membrane. We reasoned that huge endosomes, probably generated by the upregulated endosomal activity, are more difficult to transport along axons and might end up in traffic jams that could strongly affect neuronal physiology. Interestingly, synaptic activity increases the secretion of exosomes [
18,
43], and hippocampal hyperactivity has been observed in patients with mild cognitive impairment [
7], where the compounded action of both mechanisms might also generate endosomal traffic jams acting as upstream drivers of AD pathogenesis.
In this study, we demonstrated features of exosome spreading between interconnected neurons in agreement with what is expected of this type of vesicle of endocytic origin. However, we also provide evidence for a novel hijacking mechanism of endosomes by exogenous exosomes, which might result in a longer-distance action and therefore increase the pathogenic potential and the radius of action of the exosomes. These intriguing findings demonstrate that exosomes are more invasive that previously anticipated acting as amplifiers in the spread of pathogenic molecules in neurodegenerative diseases.
Acknowledgements
This study was supported by the Estate of Dr. Clem Jones AO, as well as grants from the Australian Research Council (ARC) [DP160103812], the National Health and Medical Research Council of Australia [GNT1037746, GNT1127999], and the State Government of Queensland (DSITI, Department of Science, Information Technology and Innovation). Confocal and super-resolution microscopy were facilitated by the Queensland Brain Institute’s Advanced Microscopy Facility, supported by the ARC LIEF grant (LE130100078). Electron microscopy was performed at the Australian Microscopy & Microanalysis Research Facility at the Centre for Microscopy and Microanalysis, the University of Queensland. We thank Linda Cumner, Tishila Palliyaguru, Trish Hitchcock and the animal care team for animal maintenance, and Rowan Tweedale and Esmi Zajaczkowski for critically reading the manuscript.