Sphingolipid-modulated Exosome Secretion Promotes Clearance of Amyloid-β by Microglia

Background: Exosome is a membrane vesicle released from several types of cells, including neurons. Results: Neuronal exosomes accelerate Aβ fibril formation, and the exosome-associated Aβ is taken into microglia to degrade it. Conclusion: Exosomes promote Aβ clearance. Significance: These findings provide a new function of exosome in the brain and also suggest its involvement in the development of Alzheimer disease. Amyloid β-peptide (Aβ), the pathogenic agent of Alzheimer disease, is a physiological metabolite whose levels are constantly controlled in normal brain. Recent studies have demonstrated that a fraction of extracellular Aβ is associated with exosomes, small membrane vesicles of endosomal origin, although the fate of Aβ in association with exosome is largely unknown. In this study, we identified novel roles for neuron-derived exosomes acting on extracellular Aβ, i.e. exosomes drive conformational changes in Aβ to form nontoxic amyloid fibrils and promote uptake of Aβ by microglia. The Aβ internalized together with exosomes was further transported to lysosomes and degraded. We also found that blockade of phosphatidylserine on the surface of exosomes by annexin V not only prevented exosome uptake but also suppressed Aβ incorporation into microglia. In addition, we demonstrated that secretion of neuron-derived exosomes was modulated by the activities of sphingolipid-metabolizing enzymes, including neutral sphingomyelinase 2 (nSMase2) and sphingomyelin synthase 2 (SMS2). In transwell experiments, up-regulation of exosome secretion from neuronal cells by treatment with SMS2 siRNA enhanced Aβ uptake into microglial cells and significantly decreased extracellular levels of Aβ. Our findings indicate a novel mechanism responsible for clearance of Aβ through its association with exosomes. The modulation of the vesicle release and/or elimination may alter the risk of AD.

Alzheimer disease (AD) 2 is a late-onset neurological disorder with progressive loss of memory and cognitive ability as a result of neuronal impairment and death. AD is pathologically featured by extensive extraneuronal deposition of amyloid fibrils, which are composed of amyloid ␤ protein (A␤). The A␤ is gen-erated by the processing of amyloid precursor protein (APP) as a physiological metabolite and is subsequently secreted to extracellular milieu. Steady-state levels of extracellular A␤ are controlled by the balance between its generation and its degradation/clearance. Several lines of evidence indicate that A␤ accumulation, attributable to an imbalance of its metabolism, is linked to the pathogenesis of AD (1). In the case of familial AD, genetic alterations of certain genes, such as APP and presenilin, appear to facilitate A␤ assembly through a marked enhancement in A␤ production (2). In contrast, in sporadic AD, a common form of the disease, the decreased level of A␤ elimination within the brains is apparent (3,4). This suggests a perturbation of A␤ clearance through, for example, decreased catabolism via reduced proteolysis or impaired efflux across the blood-brain barrier into CSF. However, the precise clearance process, which appears damaged in AD, remains controversial.
In a recent report, a portion of extracellular A␤ was found to be associated with membrane vesicles called exosomes (5). Exosomes represent a specific subtype of secreted small vesicles (40 -100 nm in diameter) derived from various types of cells, including neurons (6,7). They correspond to the intraluminal vesicles of endosomal multivesicular bodies (MVBs) that fuse with the plasma membrane in an exocytic manner. A well known function of exosomes is to remove obsolete or misfolded proteins and to secrete them into a drainage system, such as the gut or urinary tract (8). In addition, accumulated evidence has indicated that exosomes act as shuttles for intercellular delivery of cargo, including specific proteins, lipids, and RNAs. The exosome marker Alix has been observed in A␤ plaques in the AD brain (5). Our previous study also demonstrated that PC12derived exosomes potentially promote A␤ fibrillogenesis, under endocytic impairment, which is apparent with the early pathological changes in AD brains (9).
With these lines of evidence in mind, we investigated the fate of extracellular A␤ associated with exosomes. We demonstrated that exosomes were constitutively released from neuroblastoma N2a cells and mouse primary cortical neurons. These exosomes significantly accelerated amyloidogenesis of A␤ from its soluble form. Notably, the neuron-derived exosomes were incorporated into microglia, resulting in enhanced A␤ uptake and degradation by the microglia. Moreover, we determined that secretion of the exosomes was regulated by specific sphingolipid-metabolizing enzymes. Up-regulation of the exosome secretion, which is mediated by sphingomyelin synthase 2 (SMS2) siRNA, was sufficient for inducing the enhancement of A␤ uptake by microglia and resulted in a significant reduction of extracellular A␤ in transwell co-cultures of neuronal and microglial cells. These results imply the existence of novel machinery active in A␤ clearance mediated by exosomes.
Cell Cultures-Neuro2a mouse neuroblastoma cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum. The murine microglial cell line BV-2 was purchased from National Cancer Institute (Istituto Nazionale per la Ricerca sul Cancro, Genova, Italy) and cultured in RPMI 1640 medium (Invitrogen) supplemented with 10% fetal bovine serum and L-glutamine.
Primary cultures of neurons were prepared from the cerebral cortices of mouse brains on embryonic day 15, according to the methods of Levi et al. (10). Briefly, neurons were prepared from the isolated cerebral cortices using a dissociation solution (Sumitomo Bakelite, Tokyo, Japan). The cells were plated onto a polyethyleneimine (PEI)-coated dish at a density of 5.0 ϫ 10 5 cells/cm 2 and cultured in neurobasal medium (Invitrogen) with 25 mM KCl, 2 mM glutamine, and B27 supplement (Invitrogen). After plating, the cells were cultured for 7 days and then used for assays. Primary microglia prepared from a newborn rat were purchased from Sumitomo Bakelite and maintained in a microglial culture medium (Sumitomo Bakelite), according to the manufacturer's protocol.
Exosome Isolation-Exosomes were prepared from culture supernatants of N2a cells and mouse primary cortical neurons as described previously (11). One day before preparation of exosomes, cell culture medium was replaced with serum-free medium. Cell culture supernatants were collected after 24 h and sequentially centrifuged at 3,000 ϫ g for 10 min, 4,000 ϫ g for 10 min, and 10,000 ϫ g for 30 min to remove cells, dead cells, and debris and then spun again at 100,000 ϫ g for 1 h to obtain exosomes as pellets.
For sucrose gradient analysis, each exosome pellet (correspond to the amount from 5 ϫ 10 7 cells) was loaded onto 10 ml of sucrose gradient (0.25-2.3 M sucrose in 20 mM HEPES, 10 ml) and centrifuged at 100,000 ϫ g for 18 h. After centrifugation, 1-ml fractions were collected, diluted with 20 mM HEPES, and precipitated by centrifugation for 1 h at 100,000 ϫ g. The resulting pellets were resuspended in PBS and subjected to Western blot analysis.
Electron Microscopy-The 100,000 ϫ g pellets purified from the culture supernatants of N2a cells and primary cortical neurons were resuspended in TBS and applied to a grid covered with collodion and then negatively stained with 2% phosphotungstic acid (Nisshin EM, Tokyo, Japan). Microphotographs were obtained using an HD-2000 scanning transmission electron microscope (Hitachi, Tokyo, Japan).
Seed-free A␤ Preparation-Seed-free A␤ solutions were prepared essentially according to a published report (12). Briefly, synthetic A␤40 and A␤42 were dissolved in 0.02% ammonia solution at 500 and 300 M, respectively. To remove undissolved A␤ aggregates, which can act as pre-existing seeds, the prepared solutions were centrifuged at 540,000 ϫ g for 3 h at 4°C. The obtained supernatants were collected and stored at Ϫ80°C until use.
EGCase Treatment and CTB Binding-N2a-derived exosome pellets were resuspended in PBS containing 20 mM HEPES (pH 7.4) and treated with 25 milliunits of EGCase for 15 h or 1 M CTB for 1 h at 37°C. Then the exosomes were precipitated again by ultracentrifugation at 100,000 ϫ g for 1 h and used for ThT assay.
Thioflavin T Assay-Seed-free A␤ solutions (25 M) were incubated at 37°C in 100 l of TBS containing 0, 1, or 10 l of exosome solutions. One microliter of exosome solution was collected from culture supernatants of 1 ϫ 10 6 cells. Fluorescence intensities of ThT in the mixtures were determined as described elsewhere (12), using an Appliskan spectrofluorophotometer (Thermo Fisher Scientific, Waltham, MA). The optimum fluorescence intensities of amyloid fibrils were measured at excitation and emission wavelengths of 446 and 490 nm, respectively, with the reaction mixtures containing 5 M ThT in 50 mM glycine/NaOH at pH 8.5.
Dot Blot Analysis-Seed-free solutions of A␤42 (25 M) were incubated in 100 l of TBS with or without exosomes (100,000 ϫ g pellets) for various times at 37°C. The mixtures were then dotted onto a nitrocellulose membrane, and the membrane was incubated with primary antibodies against A␤ oligomer (A11) and A␤ (6E10) and then with HRP-conjugated secondary antibodies. Chemiluminescence was detected and analyzed using a combination of an ECL Plus Kit (GE Healthcare) and a LAS4000 imager (Fuji Film, Tokyo, Japan).
The cDNA of human amyloid precursor protein (APP)770 was amplified from human brain cDNA (Clontech) by PCR with the selected primers as follows: 5Ј-ATGCTGCCCGG-TTTGG-3Ј and 5Ј-CTAGTTCTGCATCTGCTCAAAGAAC-TTG-3Ј. The cDNA was then cloned to a pENTR TM D-TOPO vector (Invitrogen) to finally construct p3ϫFLAG-APP770 using the Gateway recombination system as described previously (13). Transient transfection was performed using a Lipo-fectamine2000 kit (Invitrogen) according to the manufacturer's protocol.
Analysis of Exosome Release-Exosome pellets purified from cell cultures (5 ϫ 10 6 cells) were solubilized with Laemmli buffer (14) and subjected to SDS-PAGE and Western blotting. Bands were detected and analyzed using a combination of an ECL Plus kit (GE Healthcare) and a LAS4000 imager (Fuji Film).
Fluorescence Staining and Internalization Assay-Exosomes were fluorescent-stained with red dye PKH26 (Sigma) according to the manufacturer's protocol (15). Briefly, exosomes (100,000 ϫ g pellet) were resuspended in diluent C (Sigma) and stained with PKH26 for 5 min, and the reaction was then stopped with 1% bovine serum albumin. The labeled exosomes were precipitated again by ultracentrifugation at 100,000 ϫ g for 1 h.
The PKH26-labeled exosomes were administered to BV-2 cells on chamber slides (Thermo Fischer) several times in serum-free conditions. For inhibition experiments, the labeled exosomes were pretreated with AV or CTB (0, 0.5, or 1 M) at 37°C for 15 min. For detection of A␤ transfer into BV-2 cells by exosomes, the labeled exosomes were preincubated for 5 h at 37°C with 25 M of a fluorescence (FAM)-labeled A␤42. The cells were then fixed with 4% paraformaldehyde, and confocal images were acquired using an Olympus Fluoview FV10i microscope. The fluorescence intensity was analyzed with ImageJ software.
A␤ Measurement-A␤40 and A␤42 levels in medium and cells were determined using a sandwich enzyme-linked immunosorbent assay (ELISA) kit from Wako. Aggregated A␤s both in medium and cells were solubilized with 4 M guanidine-HCl buffer for 2 h at room temperature and then were analyzed by ELISA. All samples were handled in duplicate.
Transwell Study-N2a cells were cultured on 24-well plate inserts (inner diameter 0.5 m pore, Corning, NY) at 5 ϫ 10 5 cells/cm 2 and then transfected with the APP770 plasmid and siRNA for nSMase2 or SMS2, using a Lipofectamine2000 kit. After 24 h, inserts were placed onto wells containing BV-2 cells. After an additional 24 h of incubation, the levels of A␤s in the medium were determined by ELISA. Intracellular A␤ levels in BV-2 cells were also measured using ELISA following solubilization in guanidine HCl buffer, as described above.

Neuron-derived Exosomes Drive A␤ to Form Amyloid Fibrils-
Culture medium from N2a cells or primary cortical neurons was subjected to successive centrifugation steps with increasing centrifugal forces, eventually providing a 100,000 ϫ g pellet. Electron microscopy analysis revealed that the pellet, collected from the N2a cultures, mainly consisted of small membrane vesicles of ϳ40 -100 nm diameter (Fig. 1B), similar to previously described exosome preparations (16). The pellet was separated by a continuous sucrose density gradient, and the exosomal proteins, Alix and Tsg101, were detected in fractions 4 and 5 (corresponding to a sucrose density of 1.12 and 1.16 g/ml, Fig. 1A), similar to reports from others (6). Exosomes are reportedly enriched in the proteins and lipids, associated with lipid microdomains (17). The ganglioside GM1, a glycosphingolipid abundant in lipid microdomains, was also detected in high concentrations in the same fractions as Alix and Tsg101. In contrast, BiP and GM130, marker proteins for endoplasmic reticulum and for Golgi respectively, were not found in the 100,000 ϫ g pellet. The pellets collected from the primary neuronal cultures also contained membrane vesicles of similar size and densities, and bearing Alix, Tsg101, and GM1 (data not shown). These data confirm that the 100,000 ϫ g pellet mainly consists of exosomes and demonstrate that the exosomes are secreted from N2a and primary cortical neurons in a constitutive manner.
To examine the effect of neuron-derived exosomes on A␤ conformational transition, we mixed the resulting pellets of the sequential centrifugations (P3, P4, P10, and P100) with soluble A␤40 and A␤42, two major species of A␤, and incubated the mixture at 37°C for 24 h. The amount of amyloid fibrils was then determined using ThT. As a result, ThT fluorescence intensities were significantly enhanced by only in the presence of P100, the exosome fraction (Fig. 1C). Both N2a-and primary neuron-derived exosomes significantly accelerated the fibril formations of A␤40 and A␤42 in a time-dependent fashion (Fig.  1D). The fluorescence of ThT follows a characteristic sigmoidal curve when A␤ was incubated at 37°C (12). As A␤42 has higher aggregate-prone property, A␤42 reached a plateau more rapidly than A␤40. In A␤42, the amounts of amyloid A␤ significantly increased after reaching a plateau phase.
Accumulated lines of evidence indicate that glycosphingolipids (GSLs), especially gangliosides, serve as a template for A␤ assembly (18,19). In addition, it has been found that the A␤-bound ganglioside GM1 has been found in brains exhibit- ing early pathological changes of AD (20). We therefore examined whether GSLs in the exosome membranes are involved in acceleration of A␤ fibrillogenesis by the exosomes. Blocking of ganglioside GM1 by CTB partially but clearly prevented A␤ fibrillization, induced by N2a-derived exosomes (Fig. 1E). Moreover, glycan cleavage by EGCase treatment almost entirely inhibited the fibril formation. These findings suggest that the sugar chains of GSLs have a role to induce A␤ fibril formation on the surface of the exosomes.
To further investigate the effect of exosomes on the formation of oligomeric A␤, the mixtures of A␤ with or without N2aderived exosomes were analyzed by dot blotting with the antioligomer antibody A11 ( Fig. 2A). In the absence of exosomes, oligomeric A␤ was formed just after 1 h of incubation at 37°C. In the presence of exosomes, however, oligomeric A␤ was not detected in the incubation mixtures during 24 h of incubation. In contrast, the exosomes could not change the amount of A11positive oligomers, which had been already formed (Fig. 2B). These suggest that the exosomes might prevent oligomeric formation of A␤ from forming. Accumulated lines of evidence indicate that neurodegeneration and synaptic impairment in AD pathogenesis are directly caused by soluble A␤ oligomers (21,22). Indeed, the A␤ solution that had been preincubated at 37°C for 5 h without exosomes induced remarkable cell death in the primary neurons ( Fig. 2, C-E). In contrast, addition of exosomes to the incubation mixtures dramatically prevented neuronal death. These findings suggest that neuronal exosomes facilitate rapid conformational transition of A␤ into nontoxic amyloid fibrils on its surface, resulting in a decline in the natural formation of toxic oligomeric species.
Sphingolipid Metabolism Is Involved in the Exosome Secretion and A␤ Fibril Formation-Exosomes originate from the budding of intraluminal vesicles into multivesicular endosomes. Trajkovic et al. (23) reported that sphingolipid ceramide triggers the intraluminal budding and leads to exosome release in oligodendrocytes. To investigate whether the secretion of the neuron-derived exosome can be modulated by sphingolipid metabolism, we first treated N2a cells and primary neurons with inhibitors for sphingolipid-metabolizing enzymes. The levels of released exosomes were determined by evaluating the amounts of the exosomal markers, Alix, Tsg101, and GM1, in the 100,000 ϫ g pellets. GW4869 is an inhibitor of neutral sphingomyelinase (nSMase), a neutral pH-active form of the enzyme SMase, which converts sphingomyelin to ceramide (Cer). Treatment of either cell with GW4869 significantly decreased the levels of released exosomes (Fig. 3, A and B),  consistent with the study of Trajkovic et al. (23). In contrast, the treatment with imipramine, which selectively inhibits acid sphingomyelinase (aSMase), did not affect the exosome release. D609 has been reported to inhibit sphingomyelin synthase (24), which catalyzes the conversion of Cer into SM, i.e. the opposite of SMase. Predictably, treatment with D609 significantly enhanced exosome secretion.
To further explore the role of sphingolipid metabolism in exosome secretion, we employed an RNA interference approach to knock down the expression of endogenous SMases and SMSs in N2a cells. We employed siRNAs against SMS1, SMS2, aSMase, nSMase1, and nSMase2, which reduced expression of the target genes efficiently in N2a cells (ϳ85% reduction in cells transfected with each siRNA). Again in agreement with the findings of Trajkovic et al. (23), treatment with siRNA for nSMase2 reduced exosome release from N2a cells (Fig. 3, C and  D). However, following treatment with siRNA against aSMase or nSMase1, no similar reduction was observed. Conversely, knockdown of either SMS1 or SMS2 with siRNA induced significant increases in exosome secretion. Increases were especially remarkable following SMS2 knockdown compared with SMS1 knockdown (Fig. 3D). These results indicate that Cer production affects the exosome secretion, and modulations of nSMase2 and SMS can alter the levels of released exosomes from N2a cells. Indeed, exogenously added Cer and Cer production following exogenously added SMase significantly increased the release levels of the exosomes (Fig. 3E).
Next, we examined the role of sphingolipid metabolism on exosome-mediated A␤ fibrillogenesis. As described above, we collected exosomes from the culture supernatants of the cells that had been treated with inhibitors or siRNAs, and we measured the ThT fluorescence in the mixtures of the exosomes and A␤42 after a 5-h incubation at 37°C. When exosomes purified from GW4869-or nSMase2 siRNA-treated cultures were included in the mixture, A␤ fibrillogenesis was significantly reduced (Fig. 3, F and G). Conversely, in the presence of exosomes from D609-or SMS2 siRNA-treated cultures, amyloid formation was increased, compared with that in controls. From these combined data, we consider that the potential of the exosomes to modulate A␤ fibril formation is closely related to their relative amount. In addition, the release of exosomes can be modulated by activities of enzymes responsible for sphingolipid synthesis.
Microglia Engulf Exosomes in a Phosphatidylserine (PS)-dependent Manner-Microglia are the resident phagocytes in the central nervous system. It is now widely accepted that these cells are derived from macrophages and contribute to the removal of dead cells and debris in the brain (25). Several reports have revealed that macrophages also take up exosomes secreted from several different cell types to transduce inflammatory signals or its elimination (26). Recently, Fitzner et al. (27) reported that oligodendrocyte-derived exosomes are preferentially internalized by microglia in brain. To evaluate whether microglia also engulf neuronal exosomes, we applied N2a-derived exosomes labeled with the fluorescent dye PKH26 to BV-2 microglial cells, primary microglia, or primary cortical neurons. After a 3-h incubation with the labeled exosomes at 37°C, the cells were fixed, stained with DAPI, and analyzed by confocal microscopy. We observed significant fluorescence in both BV2 cells and primary microglia (Fig. 4A). These results suggest that exosomes are efficiently internalized into microglia. In contrast, fluorescent signals were rarely detected in pri- . Transfer of exosomes into microglia. A, exosomes purified from N2a cell cultures were labeled with the dye PKH26 (red) and added to BV-2 microglial cells or primary cultures of microglia or cortical neurons. After 3 h of incubation, cells were fixed, stained with DAPI, and analyzed by confocal microscopy. B, N2a-derived exosomes were bound to AlexaFluor-conjugated AV or CTB to detect surface-exposed PS and GM1 ganglioside (GM1), respectively. Fluorescence labeling was visualized by confocal microscopy. The far right panel shows the same field in phase contrast (PC). Scale bar, 200 nm. C and D, exosomes collected from N2a cultures were labeled with the red dye PKH26 and subsequently treated with nothing, AV, or CTB. Labeled exosomes were applied to BV-2 cells and incubated for 3 h. Cells were subsequently fixed and stained with DAPI. C, confocal images of internalized exosomes are shown. D, fluorescence intensities per cell were determined by image analysis. Exosome internalization was quantified from three independent experiments. Values are means Ϯ S.E. ***, p Ͻ 0.001; t test. Ctrl, control. mary neurons, further demonstrating the selective transfer of neuron-derived exosomes into microglia.
Various cells produce exosomes expressing PS on their surface, and PS exposed on the outer leaflet of the plasma membrane of apoptotic cells is often used as a recognition signal by macrophages and microglia (15,28). We stained N2a-derived 100,000 ϫ g pellets with fluorescently labeled AV or CTB subunit, which specifically recognizes PS and GM1, respectively. Significant fluorescence corresponding to both AV and CTB was observed (Fig. 4B), suggesting PS was located on the outer leaflet of N2a-derived exosomes. To examine the mechanism for exosome uptake by microglia, we exposed exosomes preincubated with AV or CTB to microglial cultures. We found that treatment with AV significantly suppressed uptake of exosomes into BV-2 cells or primary microglia, although treatment with CTB had no effect (Fig. 4, C and D). These results suggest that PS facilitate for the recognition and internalization of neuronal exosomes by microglia.
Exosomes Facilitate A␤ Clearance by Microglia-The interaction between the exosomes and A␤ leads to accelerated A␤ fibril formation (Fig. 1, C and D), suggesting that A␤ amyloid fibrils accumulate surrounding exosomes. Indeed, the exosomal marker Alix was observed to be concentrated in the senile plaque, an extracellular deposition of A␤ fibrils, found in AD brain (5). We also confirmed that exosomes are internalized by microglia (Fig. 4A). Based on these findings, we hypothesized that the exosomes may have roles in the uptake of A␤ amyloid by microglia, aiding A␤ degradation. To test this hypothesis, we added A␤42, preincubated with or without exosomes, to BV-2 cells or primary microglia. After incubating at 37°C for up to 5 h, we determined the intra-and extracellular levels of A␤42. Both BV-2 and primary microglia internalized the A␤ much more dramatically in the presence of exosomes than without (Fig. 5A). Correspondingly, the levels of A␤ in the medium gradually decreased with significant difference between the presence and absence of exosomes (Fig. 5B)m 76.5 Ϯ 2.6 and 26.6 Ϯ 5.4% reduction of A␤ in the presence and absence of exosomes after 5 h of incubation, respectively. In contrast, the exosome could not affect the uptake of the amyloid fibrils, which had been already formed (Fig. 5C). To further examine whether the prevention of exosome uptake might affect A␤ incorporation by microglia, we blocked PS on the outer surface of exosomes by AV. As shown in Fig. 5D, the A␤ uptake was significantly suppressed when exosomes had been preincubated with AV and not when preincubated with CTB. These results suggest that exosomes can, at least partially, mediate A␤ uptake in a PS-dependent manner.
Next, to assess whether the A␤ internalized together with exosomes is degraded in microglia, we administered A␤42, preincubated with or without the exosomes, to BV-2 cells, incubated them for 3 h, and then washed the cells. After additional culture time, cells were harvested, and the A␤ levels in the cell lysates were determined. The intracellular A␤ levels in the BV-2 cells gradually decreased in a time-dependent manner and were nearly depleted by 48 h (Fig. 6A). To gain insight into the degradation pathway of the internalized exosomes and A␤, we investigated their localization by staining with LysoTracker, a fluorescence marker of late endosome/lysosomes. We incu-bated PKH26-labeled, N2a-derived exosomes with BV-2 cells for 3 h at 37°C and examined the cells by fluorescence microscopy. Punctate fluorescence was observed in the cells and portions of the exosome fluorescence co-localized with the lysosomal compartments (Fig. 6B). We next applied a preincubated mixture of FAM-A␤42 and labeled exosomes to BV-2 cells. Together with the exosome fluorescence, the signal corre- sponding to A␤ also co-localized with the LysoTracker signal (Fig. 6C). These data demonstrated that the A␤ internalized in an exosome-mediated manner was delivered to lysosomes within BV-2 cells to be degraded via the endocytic pathway.
Does Up-regulation of Exosome Secretion Affect A␤ Clearance?-Finally, we investigated whether modulating exosome secretion can affect A␤ clearance by microglia. We plated N2a cells on transwell inserts and treated with siRNA for SMase2 or SMS2 to modulate the amount of exosomes released from the cells. We also concurrently transfected with APP to overexpress A␤. Twenty four hours after the transfection, we set the inserts into the 24-well multiplates, on which BV-2 cells had been seeded. Under this experimental setting, we presumed that exosomes and A␤, secreted from N2a cells, would be able to interact with the BV-2 cells through medium shared between the cells. After 24 h of co-incubation, the levels of A␤ in the medium were determined. RT-PCR showed that either nSMase2 or SMS2 siRNA efficiently decreased its target mRNA expression in N2a-APP cells (81 Ϯ 2.7 and 86 Ϯ 4.3% reduction in cells with nSMase2 and SMS2 siRNA treatment, respectively). When there were no BV-2 cells in the lower wells, the level of extracellular A␤ remained unchanged even in N2a-APP cells that were treated with nSMase2 or SMS2 siRNA (Fig. 7A). Several studies have been reported that sphingolipid metabolism is involved in the APP processing for generating A␤ (29). However, the knockdown of nSMase2 or SMS2 with siRNA did not affect the levels of A␤ secreted from N2a cells. In contrast, in the presence of BV-2 cells on the lower wells, the levels of both A␤40 and A␤42 in the culture media were significantly decreased following SMS2 siRNA treatment (Fig. 7B). In addi- tion, the A␤ levels in the BV-2 cells were significantly increased in the SMS2 siRNA-treated cultures compared with control RNA treatment (Fig. 7C).
To examine levels of A␤ associated with exosomes in siRNA-treated N2a-APP cell cultures, we analyzed the amounts of Alix, Tsg101, and A␤ in the exosome pellets collected from culture media of the N2a-APP cells, transfected with nSMase2 or SMS2 siRNA. Consistent with the previous data (Fig. 3, C and D), the levels of released exosomes, estimated by the amounts of Alix and Tsg101, were obviously decreased or increased by nSMase2 or SMS2 knockdown, respectively (Fig. 7D). In addition, A␤ was detected in the exosomes from SMS2 siRNA-treated N2a cultures, although not in those from control or N-SMase2 siRNA-treated cultures. Furthermore, we performed ThT assay to examine whether the exosomes induce A␤ fibrillization under physiological conditions. We collected the exosomes from the medium of nSMase2 or SMS2 siRNA-or control RNA-treated N2a-APP cells. We then diluted the exosomes in TBS at 1,000-fold higher density than those in the medium and measured ThT fluorescence of the exosome solutions. The increased fluorescence of ThT was observed only in the solution of exosomes derived from the medium of SMS2 siRNA-treated cells (Fig. 7D). These findings suggest that acceleration of exosome secretion increases A␤ fibrils associated with the exosomes, enhancing A␤ uptake by microglia and eventually resulting in the reduction of extracellular A␤.

DISCUSSION
In the study presented here, we found that exosomes are constitutively secreted by neurons, and the exosomes dramatically promote A␤ amyloidogenesis. Furthermore, the assembled A␤, in association with exosomes, was further taken up by microglia for degradation. We also demonstrated that up-regulation of exosome secretion, which was induced by SMS2 knockdown, efficiently reduced extracellular levels of A␤ in a co-culture of neuronal and microglial cells. In CNS, neurons are surrounded by microglia that survey to remove damaged structures such as apoptotic cells and obsolete synaptic connections (30). This study provides new insight into the coordinating machinery between neurons and neighboring microglia using exosomes for the clearance of A␤ (see the proposed scheme, Fig. 8).
Regarding the formation of ordered A␤ aggregates, the seeding polymerization theory was previously proposed (31). In this theory, the transition of the monomeric A␤ to its polymer requires a conformational change of A␤ to act as a seed, as might be provided by condensation or interaction with other molecules (28). We found that the neuron-derived exosomes accelerate A␤ amyloidogenesis from monomeric A␤ (Fig. 1, C  and D). Furthermore, cleavage of GSL glycans by EGCase inhibited acceleration of A␤ fibrillogenesis by exosomes. These data suggest that interaction between A␤ and GSL glycans leads the conformational change of A␤. GSL glycans are also localized at the plasma membrane of the cells. The mechanism why only GSL glycans at the exosome surface can lead A␤ assembly has remained unclear. The GSLs are well known as the components of lipid microdomains. Lipid microdomain-associated proteins are preferentially sorted into exosomes and enriched in exosome membranes (17). Furthermore, high density clustering of GM1 has been reported to promote A␤ assembly (32). Further detailed characterization of the exosome is needed; however, these findings raise a possibility that GSL accumulation and clustering at the exosome membrane might serves as a template for A␤ assembly. In a recent report, A␤ fibrils formed by GM1-  containing membranes exhibited toxicity toward PC12 cells (33). However, we demonstrated that addition of an exosome/A␤ mixture to primary cortical cells significantly suppressed neuronal toxicity. This suppression was inversely correlated with the amount of oligomeric A␤ (Fig. 2), but it had no correlation with the amount of exosome-mediated A␤ fibrils (Fig. 1D). It is well known that A␤ fibrils exhibit polymorphism, which depends on the differences in the first step of amyloidogenesis (34,35). Further examination is needed to identify the mechanism behind exosome-mediated A␤ fibrillogenesis.
In this study, we collected released exosomes and assessed their ability to facilitate A␤ amyloidogenesis in TBS (Fig. 1, C  and D) or in the culture medium (data not shown). The results suggest that the exosomes would be able to efficiently promote the formation of A␤ fibrils in extracellular space. However, ␤-site cleavage of APP has been reported to occur in MVBs (36). In addition, A␤42 has been found to localize predominantly to MVBs in normal mouse and human brain. In addition, in a mouse model of AD and in human AD brain, A␤42 progressively accumulates in MVBs with age (37). Notably, GM1bound A␤, an amyloid seed, is preferentially observed in endosomes of neurons from aged monkey brain (38). Thus, additional careful examinations will be required to investigate the possibility that the intraluminal space of MVB might provide another cellular milieu for A␤ assembly prior to the release of exosomes and A␤.
We found that selective inhibition of nSMase2 activity reduced the exosome secretion, but inhibiting SMS2 activity increased the secretion (Fig. 3, C and D). nSMase2 is especially abundant in mammalian brains (39). It has two putative transmembrane domains at the N terminus and is mainly localized in the plasma membrane (40). SMS2 also has predicted six membrane-spanning regions, and it contributes to sphingomyelin production at the plasma membrane (41). We also found that exogenously added synthetic Cer and bacterial SMase increase exosome secretion (Fig. 3E). SMS2 apparently contributes more to exosome secretion than does SMS1, which is actually responsible for the bulk of sphingomyelin generation (Fig. 3, C and D) but is localized in the Golgi apparatus (42). Altogether, this information suggests that the elevations in the local levels of Cer, especially at the plasma membranes, including endocytosed membrane regions, would be important for the promotion of exosome generation.
Our results here support a role for Cer in exosome generation. One possible mechanism would be Cer inducing a physical alternation in the endosomal membrane that preferentially promotes the budding of intraluminal vesicles. Indeed, it has been reported that Cer can induce a coalescence of small microdomains into lager microdomains, thereby promoting domain-induced budding of plasma membranes (43). Treatment with bacterial SMase is known to induce intraluminal membrane budding from SM-containing synthetic giant liposomes (23). Alternatively, several lines of evidence suggest that Cer can also affect endocytic transport. Cer production, induced by exocytosis of aSMase, reportedly promotes endocytic transport (44). Exogenously added bacterial SMase is also known to induce ATP-independent endocytosis (45). These studies suggest another possibility that Cer might promote exosome generation by enhancing the rate of endocytosis.
Previous studies have shown that microglia can directly take up A␤ itself and degrade it in lysosomal compartments (46). Thus, the functional significance of A␤ incorporation together with exosomes remains unclear. One conceivable purpose would be to increase the efficiency of the A␤ uptake. A␤ rapidly forms amyloid fibrils in the presence of exosomes (Fig. 1, C and  D), and microglia can take up A␤ more promptly after the excessive production of A␤, in the presence of exosomes (Fig.  5). Furthermore, addition of exosomes suppressed the formation of toxic oligomers (Fig. 2); this would be highly effective for avoiding impairment of neurons. Another conceivable benefit of microglia incorporating A␤ with exosomes would be a decrease in immunological reactions by the microglia. Accumulated evidence indicates that fibril A␤ facilitates inflammatory responses in microglia, including the release of proinflammatory cytokines and reactive oxygen species (47). Activated microglia surrounding senile plaque, an excessive deposition of A␤ fibrils, contribute to chronic inflammation in the AD brain. In general, it is well understood that PS-dependent ingestion of apoptotic bodies by microglia is associated with anti-inflammatory reactions (48). In addition, it has recently been shown that oligodendrocyte-derived exosomes are taken up by microglia in an immunologically silent manner (27). This study also found that the exosome internalization was preferentially associated with inflammatory unresponsive microglia, which presented with low levels of MHCII. Moreover, we found that mRNA expression of IL-1␤ and TNF-␣ did not change in BV-2 cells after N2a-derived exosome uptake (data not shown). Further studies are needed; however, these data suggest that the neuron-derived exosomes can aide in clearing A␤ by preventing proinflammatory reactions by microglia.
Other aggregate-prone proteins, including ␣-synuclein and prion protein, which cause Parkinson and Creutzfeldt-Jakob diseases, respectively, are also associated with neuronal exosomes (49,50). A challenging subject of studies in the future will be determining whether exosomes are involved in the assembly of these proteins and in their clearance. We presume that when uptake/clearance activity of microglia decreases, secretion of exosomes bearing these proteins might provoke the pathological events, which substantially occur in the extracellular space. Indeed, in the absence of exosome-removing cells, exosomes associating with both normal and abnormally folded species of prion proteins are infectious, resulting in their spreading between neuronal cells (51,52). Furthermore, secreted ␣-synuclein in association with exosomes causes cell death of recipient neuronal cells (50). A␤ plaques might also be pathological structures built under lack of glial activity for removing exosomes. Actually, decreased numbers of microglia in a mouse model of AD result in increased A␤ deposition (53). Improvement of A␤ clearance is a potent strategy for AD therapy (3). This study might provide a new approach using exosomes to aid in A␤ elimination. Modulation of exosome secretion by selective regulation of the Cer-metabolizing pathway is likely therapeutically useful. In addition, delivery technology of exosomes, including the targeting of intravenously injected exosomes into the brain, has been developed for ther-