Amelioration of hippocampal neuronal morphology and function by mesenchymal stem cell-derived exosomes in APP/PS1 transgenic mice

Mesenchymal stem cell-derived exosomes (MSC-EXO), as a therapeutic agent, have shown great promise in the treatment of neurological diseases. To date, the therapeutic effects and underlying mechanism(s) of MSC-EXO in Alzheimer's disease (AD) are not well understood. The aim of this study was to investigate the action of MSC-EXO on hippocampal neuronal structure and function in APPswe/PS1dE9 (APP/PS1) transgenic mice. Here, the APP/PS1 transgenic mice received a single-dose of MSC-EXO via a tail vein injection, and were then assessed for pathological changes, neuronal morphology alterations, electrophysiological variations, and behavioral decits. Additionally, the nuclear factor E2-related factor 2 (Nrf2, a key mediator of oxidative injury) signaling pathway was probed by Western blotting in H 2 O 2 stimulated hippocampal neurons and a mouse model of AD. Our results showed that MSC-EXO therapy inhibited β-amyloid protein (Aβ) aggregation by reducing protein expression of 6E10 (marker of deposited amyloid plaques), repaired the synapses and dendritic spines of hippocampal neurons, restored action potentials in hippocampal pyramidal cells, improved cognitive abilities, and reduced memory impairments in a mouse model of AD. Additionally, we found that the Nrf2 signaling pathway participated in the actions of MSC-EXO, both in vitro and in vivo. Together, these data indicate that MSC-derived exosomes ameliorate the decits in hippocampal neuronal structure and function associated with the Nrf2 signaling pathway in APP/PS1 transgenic mice, suggesting the MSC-EXO as a functional therapeutic agent in AD.


Introduction
Alzheimer's disease (AD) is a neurodegenerative disorder characterized by a progressive decline in episodic memory as well as de cits in executive functioning and is recognized by the World Health Organization as a global public health priority [1]. Currently, 60-80% of the total population suffering from dementia is ascribed to AD, and elderly individuals are considered to be more susceptible [2,3].
Previous studies have suggested that a cascade of neuropathological alterations occur in Alzheimer's disease including dystrophic neurites, astrogliosis and microglial activation, neuronal and synaptic loss, ion channel dysfunction, and mitochondrial injury. Among which, amyloid plaque deposition and neuro brillary tangles are believed to be the major pathological changes that occur in the initiation and progression of AD [4][5][6]. Although multiple treatment strategies, including cell-therapy, have been used against amyloid plaque aggregation and neural dysfunction in AD models and patients [7,8], immense challenges remain including reconstructing neuronal structure and function.
The well-grown 3rd passage MSCs were washed twice with precooled phosphate-buffered saline (PBS).
The cells were then separated into Eppendorf (EP) tube at 5×10 5

MSC-EXO isolation and characterization
MSCs (3rd-5th passage) were washed twice with PBS upon coming to 70% con uency, and then replaced with medium containing exosome-depleted FBS. The exosome-depleted FBS was prepared by 18 h ultracentrifugation of FBS at 100,000g using an SW32Ti rotor (XPN-100, Beckman, California, USA). After 24-48 h, the supernatants were harvested and stored at -80℃ until processing.
Ultracentrifugation was employed to concentrate the EXO released from the MSCs as previously described [24]. Brie y, fresh or frozen cell culture supernatants were centrifuged at 300g for 10 min to remove cellular debris (TX-400 rotor, ST16R, Thermo Fisher, Massachusetts, USA), followed by 2,000g for 20 min to remove smaller cellular debris. Subsequently, supernatants were centrifuged at 10,000g for 30 min to remove larger vesicles and apoptotic bodies (F15-6x100y rotor, ST16R, Thermo Fisher, Massachusetts, USA). EXO were then pelleted by ultracentrifugation at 100,000g for 70 min in a swing rotor (SW32Ti, XPN-100, Beckman, California, USA). The pellet was washed with ice-cold PBS, resuspended in PBS and stored at -80℃.

Labeling of MSC-Exo with Alexa 594
A MSC-Exo suspension (1-2 mg/ml) was prepared in advance. As previously described [25], C5 Maleimide-Alexa 594 (CM-A954; Invitrogen, A10256, California, USA) was added to 50 μL of the MSC-Exo for 1 h at room temperature (RT) in the dark, and at the same time, ECO spin columns (MW3000, Invitrogen) were prepared in accordance with the manufacturer's instructions. During this incubation, powdered resin was hydrated for 20 min at R/T. Spin columns were then placed in the collection tubes and an AllegraX-15R (Beckman, USA) was used to centrifuge them for 2 min (750g). The collection tubes were discarded and then the labelled MSC-Exo were added to the resin. Columns were then placed in 1.5 ml tubes and centrifuged for 3 min (750g) to collect the labeled MSC-Exo. Isopycnic PBS was performed in parallel to all the protocols as controls. Subsequently, the labelled MSC-Exo was stored at -80℃. Labelled MSC-Exo were diluted to the appropriate concentration according to the experimental requirements and ltered through a 0.22 μm lter to remove dye aggregates before use.

SH-SY5Y cell lines and primary hippocampal neuron cultures
Confocal dishes or 96-well plates were coated with poly (l-lysine) for 4 h in a cell incubator (37℃), washed by sterile water three times, and subsequently air-dried using a clean bench. Hippocampal tissue was stripped from 16-18-day old embryos of mice in a sterile environment and digested at 37 ℃ with 0.125% trypsin for 10 min and then stopped by Dulbecco's modi ed Eagle's medium (DMEM) containing 10 % FBS. The number and concentration of viable cells were determined by trypan blue exclusion using an automated cell counter (JSY-SC-031, BodBoge, China). Whereafter, the cells were plated in neuronal growth and maintenance medium containing neurobasal medium (Gibco, New York, USA), 2% B27 (Gibco, New York, USA), 100 U/ml penicillin-streptomycin, and 2 mM L-glutamine (Gibco, New York, USA). In the follow-up study, the puri cation of hippocampal neurons reached more than 98%. The culture medium was changed every three days, and primary neurons could be used for subsequent experiments after about 1 w.

Aβ Oligomers preparation
Aβ Oligomers were prepared as previously described with minor modi cations [28]. Aβ 1-42 (AnaSpec, San Jose, CA, USA) was dissolved in 1,1,1,3,3,3-hexa uoro-2-propanol (HFIP, Sigma-Aldrich, 52517, California, USA) to 1 mM, which was naturally volatilized in the fume hood at RT for 2-4 h. Then, the clear solution was frozen and drained by the lyophilizer to form an Aβ peptide lm, which was stored at -80℃ or prepared directly. The peptide was subsequently dissolved in Dimethyl sulfoxide (DMSO; Sigma-Aldrich, D2650, California, USA) to 5 mM, followed by bath sonication for 10 min, and then diluted to 100 M with neurobasal medium and incubated at 4℃ for 24 h. The insoluble Aβ was removed by centrifugation at 10000g for 5 min. For in vitro applications, we used 10 mM Aβ as our treatment condition based on previous studies [29].

MSC-EXO administration and tracking
MSC-Exo were prepared at a concentration of 1 mg/ml in sterile PBS and stored at −80°C. For in vitro experiments, 10 μg/ml MSC-Exo were added to the culture medium of the primary neurons and SHSY5Y APPswe and incubated for 12 h to ensure incorporation of the exosomes. Neurons were subsequently exposed to 10 mM Aβ (Aβ+EXO group) for 24 h. Neurons in the Control and Aβ + PBS groups were treated with PBS and 10 mM Aβ. For in vivo studies, nine-month-old APP/PS1 mice were randomly divided into the AD+saline and AD+ MSC-Exo groups and each mouse were injected with 150 µL of saline or MSC-EXO (50 μg/150 μL) into the caudal vein, respectively. Age-matched WT mice received an equivalent volume of saline as the naive group. Also, CM-A594-labelled MSC-EXO were used as the treatment in the animal (n = 3) models. Then, the tissue samples were xed with 4% paraformaldehyde and stained with NeuN to observe the labeled MSC-EXO in vivo.
Resting state Ca 2+ levels were recorded for 20 sec, and then 10 mM adenosine monophosphate (ATP) was added into the ACSF to stimulate a Ca 2+ in ux. The uorescence images were captured every sec for 70 sec at the wavelength of 488 nm using a confocal microscope (Olympus, FV3000, Japan). The image analysis software Cellcens (Olympus, Japan) was used to measure calcium in ux and resting Ca 2+ levels of the individual hippocampal neurons. More than 80 cells from at least three independent experiments were analyzed for each group using Igor Pro software (WaveMetrics in Oregon, USA).

TEM of hippocampus
The TEM study was conducted as previously described with minor modi cations [32]. Tissue (1*1*2 m 3 ) was quickly harvested in a xative solution, and the brain was removed and sliced in less than 2 min. Tissue blocks were xed for 24 h with 4% glutaraldehyde, and then washed three times in a 1% PBS solution, once for 15 min, followed by further xation with 1% osmium tetroxide for 2 h. Samples were dehydrated using graded ethanol, permeated with a mixture of acetone and Eponate 812, and embedded in an embedding plate with Eponate 812. The embedded block was trimmed into a trapezoid and mounted onto blank resin stubs for ultrathin sectioning. After being counterstained with uranium dioxane acetate and lead citrate, the sections were observed under a TECNAI Spirit electron microscope (Thermo Fisher, USA).
Recording data were filtered at 2 kHz, sampled at 10 kHz with a Digidata 1440 and Clampex 10.6 (Molecular Devices, USA) and acquired and analyzed using pClamp10.6 software (Molecular Devices, USA).

Golgi-Cox Staining
Golgi-Cox staining was used to observe the morphology of neuronal dendrites and synaptic plasticity as previously reported [32,33]. The animals were anesthetized and perfused with a 0.9% sterile saline solution. The brains were removed and stained with Golgi-Cox solution (consisting 5% potassium chromate, 5% potassium dichromate, and 5% mercuric chloride) and stored at RT in the dark for 2-3 days. After that, the solution was replaced with fresh Golgi-Cox solution for another two weeks. Then the brains were kept in a 25-30% sucrose solution for two days in order to reduce the tissue fragility during the sectioning process. After collection of 100-200-μm-thick coronal slices using a vibratome (Leica, VT1000s, Germany), the brain sections were washed in deionized water for 1 min, placed in 50% NH4OH and subsequently in xing solution (Kodak; Rochester, NY, USA) for 30 min. The sections were subsequently immersed in 5% sodium thiosulfate for 10 min. After being rinsed with distilled water, the slices were dehydrated using increasing concentrations of ethanol. All sections were photographed under the bright eld of a confocal microscope (FV1000, Olympus) with an excitation wavelength of 405 nm. Images were taken by z-stack scanning, and then set the visible light to green.

Immunochemistry
At the time of sacri ce, animals were perfused intracardially with cold 4% paraformaldehyde. Frozen brain sections were collected at 30 μm intervals from the region between -2.6 mm and -4.1 mm from bregma and prepared for immuno uorescence. In parallel, 5-μm-thick para n sections from the same region were prepared for amyloid-β (Aβ) immunohistochemistry. All the sections were etched with PBS solution containing 20% methanol and 3% H 2 O 2 for 20 min, and blocked with PBS solution containing 0.1% Triton X-100 and 10% serum from the same species as the secondary antibody for 30 min. A battery of primary antibodies, including puri ed anti-β-Amyloid, 1-16 antibody (1:500) (Biolegend, SIG-39320, San Diego, CA, USA), and mouse anti-NeuN (1:100) (Abcam, Ab104224, California, USA) were diluted in blocking solution and incubated with sections overnight at 4 °C. After incubation, the samples were washed in PBS and then incubated with a corresponding secondary antibody solution for 2 h at RT. The following secondaries were used: Biotinylated anti-mouse [heavy-and light-chains (H+L); BA-2000; Vector Lab]; A488 anti-mouse IgG (1:500) (Thermo Fisher Scienti c, A-32790, New York, USA). DAPI (Sigma-Aldrich, 32670, California, USA) was used for cell nuclei counterstaining, and the peroxidase reaction was developed using a DAB substrate. All immunostained sections were photographed under a confocal microscope (Olympus, FV10-ASW, Japan) or a light microscope (Leica, DMi8, Germany), and analyzed using ImageJ Pro Plus V 6.0 (Bethesda, Maryland, USA).
Thio avin S staining Thio avin S staining was performed in order to detect Aβ plaques in brain tissue as previously described [34,35]. Brie y, animals were perfused with a 4% paraformaldehyde solution and brain tissue was removed, followed by preparation of para n-embedded samples. After dewaxing, 5-μm-thick para n sections were prepared and then washed three times for 5 min each. DAPI was added and incubated with the tissue sections for 8 min. A 0.3% thio avin S solution was conFig.d in 50% alcohol (v/v) and ltered, and then incubated for 8 min at RT. The sections were washed in 80% alcohol for 10 sec at RT and stained with 0.3% thio avin S solution twice for 10 sec each. Subsequently, the sections were rinsed with distilled water one last time. Then, the sections were then coverslipped with neutral balsam. Images were captured with a confocal microscope (Olympus, FV10-ASW, Japan), and analyzed using ImageJ Pro Plus V 6.0 (Bethesda, Maryland, USA).

Western blotting
Cell samples from the APPwt, APPswe+ PBS, APPswe+ EXO groups (n=3 per group) and hippocampal tissues at seven days, four weeks or 12 weeks after MSC-EXO administration from animals belonging to naive, AD+ saline and AD+EXO groups (n=5 per group per time point) were dissected for Western blotting. Samples were lysed in Radio-Immunoprecipitation Assay (RIPA) Lysis Buffer (Beyotime, P0013B, Shanghai, China) containing a 1:100 (v/v) ratio of a protease inhibitor cocktail and a phosphatase inhibitor cocktail (Millipore, MA, USA). Total protein samples were quanti ed using the BCA protein assay (Beyotime, P0012, Shanghai, China), and then normalized protein samples were separated via sodium dodecyl sulfate-polyacrylamide gel, and transferred to polyvinylidene uoride membranes (Millipore, MA, USA). Membranes were blocked in tris-buffered saline (TBS) containing 0.1% (v/v) Tween 20 (TBST) and 5% (w/v) nonfat milk at RT for 2 h. Membranes were incubated with primary antibodies including Nrf2 CW0102S, CWBIO CompanyLimited), or HRP-conjugated anti-rabbit secondary antibodies (Cat. EK020, Zhuangzhi Biotech Company Limited) at RT for 2 h. Images were taken using a Bio-Rad imaging system (Bio-Rad, Hercules, California, USA) and analyzed by the Quantity One software package (West Berkeley, California, USA).

Morris Water Maze (MWM)
The Morris water maze was used to test spatial memory and cognitive function of naive, AD+saline and AD+EXO group (n = 8~10 per group) as previously described [36,37]. The tests were conducted in a water maze drum lled with milky white water dyed by edible pigment at 20 °C ± 1 °C. The testing room was kept quiet during the experiment to minimize the interference of experimenter on mice. Spatial learning sessions were performed over four consecutive days with four trials per day. A small escape platform was submerged 1-2 cm under water surface. The mice were gently placed into the water, facing the wall, at the corresponding entry point in a quasi-random fashion to prevent strategy learning, free to nd the platform. Each trial lasted until the animal found the platform or for a maximum of 60 s. If the mice failed to nd the platform within 60 s, they were guided to the platform for another 30 s. After every trial, the mice were placed in a clean, warm environment and given food and water to replenish their energy. A 60 s probe test was conducted 24 h from the last learning session, during which, the hidden platform that was present for the rst four days was removed. Mice were allowed to swim freely in the water for 60 s. A tracking system (Morris2.8.1) was used that automatically recorded and analyzed the latencies to reach the platform, the activity time, the percentage of the total time (%, PT) in the target quadrant, and the frequency of mice crossing the platform.

Novel object recognition test (NORT)
The NORT was conducted to measure recognition memory as previously described [38]. Brie y, mice were individually placed in a square open eld apparatus with a side length of 44 cm, free to explore for 5 min in the habituation session, and then returned to their home cage. After each experiment, the open led was thoroughly cleaned with 75% (v/v) ethanol to minimize olfactory cues before the next mouse entered the open eld. In the familiarization phase performed 24 h after the rst session, two identical objects were placed in two opposite areas of the apparatus 10 cm away from the wall, and the mice were again placed in the eld with their head positioned opposite the objects to explore for 10 min. Twenty-four h later, the test session was conducted. One of the objects was replaced with a novel one and mice were placed in the arena containing one familiar and one new object to explore for another 10 min. Their behavior and exploring time were recorded by a video tracking system for analysis. A discrimination index was calculated as follows: Discrimination index = (time on novel -time on familiar)/ (time on novel + time on familiar). All operations were conducted between 8 a.m. and 5 p.m.

Statistical analysis
All the data are presented as Mean ± SEM. Multiple comparisons were analyzed using one-way analyses of variance (ANOVA) or unpaired, two-tailed Student's t-test. A repeated measures ANOVA was carried out to analyze the differences in escape latency among groups after Bonferroni posttest using SPSS 24.0.0 and GraphPad Prism 6 software (GraphPad Prism, USA). P values of less than 0.05 were considered statistically signi cant.

MSC-EXO characterization and tracking in mice
MSCs were obtained from Wharton's jelly as previously described. A ow cytometry analysis showed that the 3rd passage of harvested cells presented strong expression of stromal markers, including CD1105, CD44, CD29 and CD90, and almost negative expression of hematopoietic lineage markers CD45 and CD34 (Fig. 1A) as described for the speci c criteria of MSCs [39]. Characterization of the MSC-EXO was performed by Western blotting, NTA, and TEM. The protein assay results showed that the extracellular vesicles (EVs), isolated from the culture medium of MSCs, expressed exosomal markers such as CD63, CD9, and TSG101 (Fig. 1B), while the size of nanovesicles was distributed from 40 ~ 160 nm (Fig. 1C) and the morphology presented a typical cup-shape (Fig. 1D). Together, these data suggest that these EVs are consistent with the typical characteristics of exosomes. Although multiple studies have demonstrated the permeability of exosome to the blood-brain barrier and targeting of neural cells in neurological disorders [40,41], we used CM-A594 labeling to track the MSC-EXO in APP/PS1 mice to ensure the vesicle delivery. Twenty-four h after a tail vein injection of MSC-EXO, uorescent staining showed that labelled exosomes were observed in the cytoplasm of neuronal cells within the hippocampus.

Tail vein injection of MSC-EXO reduces Aβ aggregation in APP/PS1 mice
Aβ is believed to be a crucial and primary factor in triggering progressive loss of neuronal function in AD [42]. Therefore, we performed histochemistry, thio avin staining and Western blotting to examine Aβ aggregation in APP/PS1 mice. The histochemistry results showed that, compared to the naive control, Aβ deposition was increased in APP/PS1 mice as evidenced by the size and IOD value of 6E10 (evidencing deposited amyloid plaques) staining in the hippocampus (Fig. 2B, D, P < 0.01) and cortex (Fig. 2C, E, P < 0.01). After MSC-EXO treatment, a reduction in 6E10 expression was observed in AD + EXO mice that was not observed in the AD + saline mice (Fig. 2B-D, P < 0.05). Thio avin staining was further used to probe the Aβ protein deposition in the experimental groups, and consistent with above results, the positive area (green) and mean uorescence intensity in the hippocampus (Fig. 2G, I, P < 0.05) and cortex (Fig. 2H, J, P < 0.05) of the AD + EXO group were signi cantly lower than those observed in the AD + saline group. Additionally, a protein assay (Fig. 2K) showed that compared to naive control, an increase of Aβ expression was observed in APP/PS1 + saline mice (Fig. 2L, P < 0.001), whereas exosome therapy remarkably decreased (P < 0.05) expression of Aβ in the AD model (AD + EXO group). Together, these results imply that MSC-EXO inhibit the aggregation of Aβ observed in APP/PS1 mice MSC-EXO administration repairs the hippocampal neuronal morphology alterations in mice Synaptic morphology and plasticity are critical for neuronal function and are known to be associated with the memory impairment in AD [43]. Herein, we examined the dendritic processes and spines of hippocampal pyramidal neurons using Golgi staining at 3 months after exosome treatment. The representative images displayed obvious morphological differences including dendritic complexity (Fig. 3A), three-dimensional (3D) reconstruction of spines, (Fig. 3B) and dendritic phenotype (Fig. 3C) between the naive, AD + saline, and AD + EXO groups. 3D reconstruction of dendrites showed a reduction of total dendritic density (Fig. 3D, P < 0.0001), and lopodia/mushroom/long thin dendrite spine density ( Fig. 3E-G, P < 0.01) in hippocampal pyramidal neurons in APP/PS1 mice compared to naive controls. Additionally, after neuronal phenotype sholl analysis, we found that APP/PS1 mice had lower total spine length (Fig. 3I, P < 0.0001) and lower dendritic intersections of basal dendrites (-40-180 µm) (Fig. 3J) than naive mice. Remarkably, these morphological changes in pyramidal neurons were reversed in the AD + EXO group compared to the AD + saline group (Fig. 3D-G and I-J, P < 0.05). Together, these results suggest that the morphological alterations observed in hippocampal neurons of APP/PS1 mice are restored by MSC-EXO therapy.

Exosome therapy restores neuronal excitability and mitochondrial changes in the AD model
To investigate the restoration of MSC-EXO on neuronal function in the AD model, given that hippocampal CA1 pyramidal neurons are responsible for learning, memory and cognition [44], the excitability of these cells was detected by whole cell patch clamp recording (Fig. 4A). After analyzing the waveform of pyramidal neurons (Fig. 4B), we found a signi cant increase in half-width (Fig. 4C, P < 0.05) and a signi cant decrease in membrane potential (Fig. 4D, P < 0.05) and frequency (Fig. 4E, P < 0.01) of action potentials in the AD + saline group compared to the naive group, whereas EXO treatment signi cantly reversed ( Fig. 4C-E, P < 0.05) these potential changes, suggesting that MSC-EXO treatment restores the excitability of hippocampal CA1 pyramidal neurons in APP/PS1 mice. Importantly, mitochondria are essential for the maintenance of proper function and quality control in hippocampal neurons [45]. We therefore examined the ultrastructure (Fig. 4F) and injury markers (Fig. 4G) of mitochondria in the experimental groups. As shown in TEM images (Fig. 4F), the mitochondria in hippocampal neurons of AD mice exhibited obvious swelling and vacuolation (red arrow) compared with the naive group, while a trend toward normal mitochondrial structure was observed in the AD + EXO group. Moreover, compared with the naive group, the altered mitochondrial ssion / fusion was typi ed by an increase in COX (Fig. 4H, P < 0.001), Tom20 (Fig. 4I, P < 0.001), and FIS1 (Fig. 4J, P < 0.001) expression in the hippocampal tissue of AD mice. Notably, after MSC-EXO treatment, these protein markers were found to be reduced in comparison to the AD + saline group (Fig. 4H-J, P < 0.01). Combined, these results suggest that exosometherapy ameliorates the mitochondrial changes observed in APP / PS1 mice.

MSC-EXO treatment ameliorates calcium transients in Aβ-induced primary culture of hippocampal neurons
It is believed that calcium imbalance induced by amyloid Aβ drives the synaptic plasticity and neuronal loss observed in AD [46]. Therefore, we assessed for calcium signaling alterations using an Aβ-induced primary culture of hippocampal neurons. Calcium imaging revealed differences in the uorescence properties of hippocampal neurons among the experimental groups (Fig. 5A). Following ATP stimulation, the rst phase calcium response consisted of a sharp peak in calcium signaling, followed by a second phase response of a slowly declining intracellular calcium concentration in each group (Fig. 5B). The statistical analysis revealed that, compared to the control group, Aβ induction resulted in a reduced amplitude of intracellular calcium transients (Fig. 5C, P < 0.0001) in the primary culture of hippocampal neurons after adding ATP, whereas MSC-EXO treatment signi cantly increased the calcium in ux compared to the Aβ + PBS group (Fig. 5C, P < 0.05). We also found a slower change in the response rise time (Fig. 5D, P < 0.0001) as well as the decay time (Fig. 5E, P < 0.0001) in the Aβ + PBS group in compared to the control group. Remarkably, exosome-therapy (Aβ + EXO) reversed the rate of calcium transients observed in the Aβ + PBS group (Fig. D, E, P < 0.0001). Together, these data indicate that MSC-EXO treatment ameliorates the alterations in calcium transients in Aβ-induced primary hippocampal neurons, further supporting the therapeutic potential of MSC-EXO on neuronal dysfunction.

Exosomes Treatment Improves Cognitive De cits In The App/ps1 Mice
After MSC-EXO administration for three months, the MWM and NORT were performed to assess for cognitive de cits. Analysis of the MWM (Fig. 6A) showed that there was no signi cant difference in the swim speed (Fig. 6B, P > 0.05) among the experimental groups, which excluded the in uence of dyskinesia on the results. Compared to the naive group, the learning and memory impairments in the AD + saline group were characterized by a longer time of escape latency (Fig. 6C, P < 0.0001) and a decrease of platform crossings (Fig. 6D, P < 0.01) and percent time (PT) in the target quadrant (Fig. 6E, P < 0.05).
Whereas an indication of behavioral improvements (including escape latency, platform crossings and PT in the target quadrant) were observed in the AD + EXO group (Fig. 6C-E, P < 0.05) compared to the AD + saline group. Additionally, recognition memory function was examined using the NORT. We found that compared to the naive group, the AD + saline group displayed an inability for novel object discrimination (N, Fig. 6F) as they spent a similar percentage of time exploring the familiar and novel objects. Notably, animals in AD + EXO group showed an increased discrimination index compared to the AD + saline group (Fig. 6G, P < 0.05), suggesting that the recognition memory function in APP / PS1 mice was ameliorated by EXO treatment. Together, these results indicate that MSC-EXO treatment improves the cognitive de cits observed in APP/PS1 mice.
Oxidative defense system is associated with the exosome therapy in APP/PS1 mice Oxidative stress is believed to be a cause of neuronal degeneration in Alzheimer's disease, while Nrf2 play a pivotal role in the mediation of oxidative stress [47,48]. We rst investigated whether Nrf2 signaling participates in the action of MSC-EXO in APP/PS1 mice. As shown in Fig. 7A, compared with agematched control mice, the AD mice (AD + saline) displayed a remarkably higher level of Nrf2, HO-1, NQO1, iNOS (Fig. 7B, D-F, P < 0.001) and a lower level of Keap1 protein expression (Fig. 7C, P < 0.0001), while exosomes treatment reversed the expression of these proteins (Fig. 7B-F, P < 0.05), suggesting the involvement of the Nrf2 signaling pathway in the treatment of AD model. To further con rm the participation of this signaling in the action of MSC-EXO on the hippocampal neurons, AD cell models including the SHSY5Y (APPwt) and SHSY5Y (APPswe) cell lines were employed in an in vitro study [27].
As can be seen in Fig. 7G, Nrf2 signaling (Nrf2, Keap1, HO-1, NQO1) and INOS exhibited variations similar to those observed in the in vivo studies between the APPwt and APPswe + PBS group (Fig. 7H-L, P < 0.01). Additionally, MSC-EXO administration signi cantly reduced the expression of Nrf2, HO-1, NQO1, iNOS and increased the expression of Keap1 in the APPswe + EXO group (Fig. 7H-L, P < 0.05) compared to the APPswe + PBS group, indicating that Nrf2 signaling participates in the action of MSC-EXO on hippocampal neurons in AD. Notably, Nrf2 expression was signi cantly increased after exosomes treatment compared to the naive or APPwt group, which may suggest that MSC-EXO activate the Nrf2 pathway in the AD model.

Discussion
Our results show that MSC-EXO treatment inhibits the deposition of Aβ protein observed in the brain of AD mice as well as ameliorates the de cits in neuronal structure and function typi ed by the morphology alterations, mitochondrial changes, excitability restoration, calcium transients, and associated cognitive repairments observed in the AD cell and mouse models. Additionally, the present data demonstrated that the Nrf2 signaling pathway participates in the actions of MSC-EXO on neuronal dysfunction in AD using cell and mouse models. Together, these results suggest that MSC-EXO can represent a promising nanotherapeutic agent for the treatment of AD.
It is well known that Aβ protein deposition is the most typical pathological feature in AD patients and animal models. To ensure the administered MSC-EXO produced therapeutic effects on APP/PS1 mice, we performed a systematic investigation of Aβ aggregation using an Aβ indicator (6E10), Thio avin staining, and western blotting. Consistent with a previous study [17], exosome-therapy reduced the expression of Aβ related indications, implying the positive effects of MSC-EXO on Aβ aggregation in APP/PS1 mice brain. Importantly, cognitive and memory impairments are one of the main clinical manifestations observed in AD patients [4]. The MWM and NORT tests were used to detect the cognitive decline in APP/PS1 mice, and our results showed that EXO treatment improved their behavioral performance compared to the saline group, suggesting that the cognitive de cits can be improved by MSC-EXO.
Together, the above results indicate that MSC-EXO show markable therapeutic effects on behavior in APP/PS1 mice.
Among the multiple features of AD initiation and progression, including Aβ deposition, neuroin ammation, oxidative stress, and abnormal neurogenesis, neuronal dysfunction is believed to be the ultimate cause of cognitive decline, and the hippocampus is one of the most affected areas in AD patients [49]. Therefore, the present study focused on the reconstruction of hippocampal neurons by MSC-EXO treatment. Synaptic transmission provides the physiological, cellular, and molecular mechanisms for cognitive function, while synaptic plasticity is re ected by dendritic density and complexity in neuronal cells. Herein, extensive loss of synapses and dendritic spines is one of the important pathological features of AD [50]. In the present study, impairments of synaptic plasticity within hippocampal neurons were observed in AD mice. After exosome-therapy, the morphological changes in spine density and dendritic intersections implied that the structural impairments in hippocampal pyramidal neurons can be restored by MSC-EXO in APP/PS1 mice. Accordingly, compared with the AD + saline group, the increased performance of the AD + EXO group suggests that EXO treatment facilitates the recovery of learning and memory in AD mice. Collectively, these data suggest that MSC-EXO facilitate the reconstruction of neuronal structure thus ameliorating the associated cognitive decline observed in APP/PS1 mice.
Moreover, abnormal excitability is a well-known alteration that occurs in neurodegenerative diseases such as AD [51]. Our whole-cell patch-clamp experiment is consistent with this as the electrophysiological activity of hippocampal neurons was altered in APP / PS1 mice, evidenced by a decreased frequency of Aps and membrane potentials. Whereas MSC-EXO administration reversed this alteration in neuronal excitability suggesting that EXO treatment restores the electrophysiological function of hippocampal neurons compared to the AD + saline group. Importantly, mitochondria generated ATP is essential for the excitability and survival of neurons, which functions to regulate calcium in ux, synaptic plasticity, and redox signaling [52][53][54]. Thus, calcium imaging performed to detect the calcium transients in cultured primary hippocampal neurons in response to Aβ stimulation. MSC-EXO treatment ameliorated the alterations in calcium transients observed in the Aβ + PBS group as evidenced by the amplitude of signaling, rise time and decay time. Previous studies have reported that mitochondrial dysfunction is involved in the pathogenesis of most nervous system diseases including AD [55]. Here, we used TEM to detect the mitochondrial ultrastructure and protein assay to examine mitochondrial ssion / fusion in the experimental groups. Compared to the saline group, the ameliorated mitochondrial ultrastructure and decreased COX , Tom20 and FIS1 expression observed in AD + EXO group, indicated that exosomes administration improves the mitochondrial dysfunction in APP / PS1 mice. Together, these results suggest that the abnormalities in neuronal structure and dysfunction observed in AD can be restored by MSC-EXO, which provides new evidence for the nanotherapeutic action of EXO in AD mice.
Furthermore, hippocampal neurons are highly susceptible to oxidative stress in AD patients and animal models, and the redox imbalance is considered to be a prominent factor of neuronal damage in the brain [56]. Taking this into consideration, examination of the oxidative defense system may reveal the mechanism(s) underlying MSC-EXO therapeutic action on hippocampal neurons in AD. Furthermore, our previous study indicated that MSC-EXO show antioxidant activity as they contained enriched functional agents such as nucleic acids, proteins and enzymes that regulate redox reactions [23]. Antioxidants also have great potential in the treatment of neurodegenerative diseases and have been shown to improve learning and memory de cits in AD mice [57,58]. Among the multiple molecular pathways implicated in the oxidative defense system, Nrf2 is an essential element for the regulation of oxidative responses in neuronal degeneration [59]. Under physiological conditions, Nrf2 interacts with the cytoplasmic protein Keap1, whereas it is isolated from Keap1 and translocated into the nucleus under stress conditions [60,61]. Numerous studies have indicated that Nrf2 can activate transcription of its target genes HO-1 and NQO1, which then exerts a cytoprotective effect against neuronal damage [62]. In the present study, compared with AD + saline group, MSC-EXO treatment resulted in a decrease of Nrf2, HO-1, NQO1, iNOS (indicative of oxidative damage), and an increase of Keap1 expression in hippocampal tissues, suggesting that the Nrf2 signaling pathway participates in the therapeutic action of exosomes in the APP