Results

Preparation, characterization, and functional validation of RVG-EXOs for entering brain tissue and targeting neurons

To construct engineered EXOs with enhanced brain-targeting capability, plasma was collected from young individuals aged 18–25 years and isolated plasma-derived EXOs. First, 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-polyethylene glycol 2000 (DSPE-PEG2000) was reacted with RVG-29 to generate DSPE-PEG2000-RVG29 nanomicellar particles, which were then co-incubated with the isolated young plasma-derived EXOs to obtain EXO-DSPE-PEG2000-RVG29 (Fig. 1A). Then, nanoscale flow cytometry detection was performed on ordinary young plasma-derived EXOs and RVG-EXOs using FITC-conjugated RVG antibody. The results showed that the mean fluorescence intensity (MFI) of RVG-EXOs was 387, whereas the MFI of ordinary EXOs was only 61.7, indicating that the FITC fluorescence signal intensity of RVG-EXOs was approximately 6-fold higher than that of ordinary EXOs (Fig. 1B and S1A). Since RVG29 carried a positive charge, if it was successfully conjugated to the EXO surface, the membrane potential of the EXOs would shift. Therefore, the membrane potentials of ordinary young plasma-derived EXOs and RVG-EXOs were measured. The results showed that the membrane potential of RVG-EXOs reached approximately 5.21 (Fig. 1C). Together, these results demonstrated that the RVG peptide was successfully anchored onto the EXOs surface and maintained a high modification intensity. Transmission electron microscopy (TEM) images of ordinary EXOs and RVG-EXOs demonstrated that the morphology of EXOs remained intact after engineering modification of the EXO surface with RVG (Fig. 1D). Nano-tracking analyzer (NTA) results showed that the average size distribution of ordinary young plasma-derived EXOs and RVG-EXOs was predominantly around 180 nm (Fig. 1E). Quantitative analysis revealed that the concentration of isolated EXOs was approximately 7.8 × 106 particles/μL, while RVG-EXOs yielded approximately 5.5 × 106 particles/μL after engineering and purification (Fig. 1E). This yield was within the typical ranges reported for plasma-derived EXOs using commercial precipitation-based kits [18,19]. Western blot analysis showed that after RVG engineering, exosomal marker proteins such as CD63 and ALIX were still detected (Fig. 1F). To further assess the purity of the isolated EXOs, we examined the expression of the negative markers GM130 (a Golgi marker) and Calnexin (an endoplasmic reticulum marker). Western blot analysis revealed that neither GM130 nor Calnexin bands were detected in the EXOs and RVG-EXOs samples, whereas distinct bands were observed in the cell lysate (positive control) (Fig. S1B). These results indicated that the isolated EXOs samples were largely free of contamination from Golgi and endoplasmic reticulum sources, suggesting relatively high purity.

Fig. 1: Preparation, characterization, and functional validation of RVG-EXOs for entering brain tissues and targeting neurons. (A) Schematic diagram of the RVG-EXO production. (B) Flow cytometry histograms showing FITC fluorescence intensities of EXOs and RVG-EXOs. (C) Zeta potential of EXOs and RVG-EXOs. Data are presented as mean ± SD (n = 3). (D) Representative TEM images of EXOs and RVG-EXOs. (E) Size distribution profiles of EXOs and RVG-EXOs determined by nanoparticle tracking analysis. (F) Western blot analysis of the exosomal marker proteins ALIX and CD63 in EXOs and RVG-EXOs. (G) Representative immunofluorescence images and corresponding quantitative analysis (H), showing the targeting of PKH67-labeled EXOs and RVG-EXOs to SH-SY5Y cells. Data are presented as mean ± SD (n = 3). Statistical significance is determined by the Student's t-test based on P < 0.05 (). (I–K) Western blot analysis and corresponding quantitative analysis of exosomal marker proteins (ALIX and CD63) in SH-SY5Y cells after 24-h co-culture with RVG-EXOs or EXOs. Data are presented as mean ± SD (n = 3). Statistical significance is determined by the Student's t-test based on P < 0.05 (). (L) Representative immunofluorescence images and quantitative analysis of the hippocampal CA3 region in mouse brain tissues 4 h after intravenous tail vein injection of PKH67-labeled RVG-EXOs or EXOs. Data are presented as mean ± SD (n = 3). Statistical significance is determined by the Students' t-test based on P < 0.05 (*). (M) In vivo imaging in mice with DiR-labeled RVG-EXOs or EXOs.

To evaluate the neuronal targeting capability of RVG-EXOs, equal amounts of PKH67-labeled EXOs and RVG-EXOs were co-cultured with SY5Y cells of the same concentration for 24 h, followed by immunofluorescence staining. The results indicated that, compared with the EXO group, the number of RVG-EXOs entering the cells was significantly increased (Fig. 1G and H). Subsequently, the equal amounts of EXOs and RVG-EXOs were we co-cultured with SY5Y cells of the same concentration for 24 h and then subjected to Western blot analysis. The results showed that, compared with the EXO group, the level of exosomal marker proteins in the RVG group was significantly increased (Fig. 1I–K). These results collectively demonstrated that RVG-EXOs possessed an enhanced ability to target neurons. Effective treatment of AD required successful entry of the medicines into the brain. To examine the ability of RVG-EXOs to enter brain tissue, PKH67-labeled control EXOs and RVG-EXOs were administered to 3xTg mice via tail vein injection. In 4 h, brain tissues were collected for fluorescence staining. Notably, the fluorescence intensity of PKH67-labeled RVG-EXOs in the brain tissue of 3xTg mice was significantly higher than that of control EXOs (Fig. 1L). Additionally, longitudinal in vivo imaging of DiR-labeled EXOs or RVG-EXOs revealed that, compared with ordinary young plasma-derived EXOs, RVG-EXOs exhibited significantly higher level of accumulation in the brain (Fig. 1M). These results indicated that RVG engineering effectively enhanced the delivery of EXOs to brain tissue. In summary, the above results demonstrated that RVG engineering significantly enhanced the ability of young plasma-derived EXOs to target neurons and to be delivered to brain tissue.

Next, this study employed small animal in vivo imaging to investigate the biodistribution of EXOs and RVG-EXOs in major organs throughout the mouse body, as well as their duration and proportion in brain tissue. The results showed that EXOs and RVG-EXOs were mainly enriched in organs such as the brain, heart, liver, and kidney. Compared with EXOs, the distribution of RVG-EXOs in brain tissue was significantly increased (Fig. S1C). To reflect the proportion of EXOs and RVG-EXOs in brain tissue, we calculated the fluorescence intensity in brain tissue as a percentage of the total fluorescence intensity in all major target organs. The results showed that the proportions of EXOs and RVG-EXOs in brain tissue were approximately 21% and 40%, respectively (Fig. S1C). Small animal in vivo imaging showed that at 6 h, the presence of EXOs and RVG-EXOs was detected in brain tissue; at 24 h, the highest level of distribution of EXOs and RVG-EXOs in brain tissue was detected; and at 48 h, EXOs and RVG-EXOs were almost not detected in brain tissue (Fig. S1C). This indicated that EXOs and RVG-EXOs could maintain in brain tissue for approximately 48 h. Since RVG-EXOs exerted their corresponding effects mainly by targeting neurons after entering brain tissue, this study further investigated the uptake of EXOs and RVG-EXOs by neurons in vivo. PKH67-labeled control EXOs and RVG-EXOs were administered to mice via tail vein injection. In 24 h, brain tissues were collected for fluorescence staining. Immunofluorescence staining results showed that both EXOs and RVG-EXOs could be taken up by neurons in brain tissue, and RVG engineering modification significantly enhanced the uptake capacity of EXOs by neurons (Fig. S1D).

RVG-EXOs improve cognitive behaviors in 3xTg mice

Studies have indicated that the bioactivity of EXOs exhibits almost no species specificity, i.e., young human plasma-derived EXOs can exert effects comparable to those of young mouse plasma-derived EXOs [9,20,21]. To investigate the effects of RVG-EXOs on cognitive impairment in 3xTg mice, the mice were divided into different groups for intervention, followed by a series of behavioral tests, including the Morris Water Maze (MWM), Y-maze, Novel Object Recognition (NOR), and Open Field Test (OFT) (Fig. 2A).

Fig. 2: RVG-EXOs improve cognitive behaviors in 3xTg mice. (A) Experimental flowchart. (B) Representative trajectory heatmaps based on the MWM probe test and quantification of the number of platform crossings (n = 8). (C) Representative plots of escape latency in the MWM test (n = 8). Data are presented as mean ± SD. Statistical significance is determined by the Student's t-test based on P < 0.05 (), P < 0.01 (), P < 0.001 (), and P < 0.0001 (), compared to the WT-Vehicle group. (D) Representative activity trajectories in the Y-maze and quantitative analysis of the spontaneous alternation rate (n = 8). (E) Recognition index in the NOR test (n = 8). (F) Representative OFT tracks and analysis of center versus periphery time (n = 8). Data are presented as mean ± SD (n = 8). Statistical significance is determined by the Student's t-test based on P < 0.05 (), P < 0.01 (), P < 0.001 (**), and P < 0.0001 (****).

First, the effects of RVG-EXOs on the spatial learning and memory abilities of the mice were evaluated using the MWM. On the first day, the mice underwent visual training to assess their vision and swimming ability. Subsequently, hidden platform training was conducted over 5 d for the different groups of mice. On the 7th day, the platform was removed, and a probe test was performed using animal tracking software to assess the memory function of the mice. Notably, during the hidden platform training, compared to the negative control mice, the 3xTg mice injected with saline exhibited slower learning ability and longer escape latencies during the learning process, indicating impaired learning ability in the 3xTg mice. In contrast, compared to the positive control mice, the 3xTg mice treated with either RVG-EXOs or EXOs demonstrated faster learning capability and shorter escape latencies, with more pronounced effect detected in the RVG-EXO-treated group. These results suggested that both RVG-EXOs and EXOs could improve the learning ability of 3xTg mice, with RVG-EXOs having a more significant effect. In the subsequent probe test, compared to the negative control mice, the untreated 3xTg mice crossed the target platform significantly fewer times. However, after treatment with RVG-EXOs or EXOs, the number of times that the 3xTg mice crossed the target platform was increased significantly, with more evident effect revealed in the RVG-EXOs-treated group. Overall, the MWM results indicated that both RVG-EXOs and EXOs could improve the learning and memory abilities of 3xTg mice, with RVG-EXOs exhibiting a stronger effect (Fig. 2B and C).

Next, the short-term spatial working memory of the mice was assessed using the Y-maze. Compared with the negative control group, the spontaneous alternation rate of mice in the positive control group was significantly decreased, indicating an impairment in short-term spatial working memory. Treatment with either RVG-EXOs or EXOs could partially improve this impairment, with more pronounced effect detected in the RVG-EXO-treated group (Fig. 2D).

Subsequently, the NOR test was performed on the mice. The results showed that, compared with the negative control mice, the recognition index (RI) of the positive control mice was significantly reduced, indicating impaired short-term object recognition memory in the 3xTg mice. Treatment with either RVG-EXOs or EXOs could partially restore the object recognition memory of the mice, with RVG-EXOs demonstrating a stronger effect (Fig. 2E). Finally, the OFT was performed to assess anxiety-like behavior in the mice. Notably, after treatment with either RVG-EXOs or EXOs, the time spent in the central area by 3xTg mice was significantly increased, with the more pronounced effect revealed in the RVG-EXO-treated group (Fig. 2F). These results indicated that both RVG-EXO and EXO treatments could alleviate anxiety-like behavior and improve cognitive function in 3xTg mice, with RVG-EXO exhibiting a stronger effect. Overall, RVG-EXO treatment improved various cognitive behaviors in 3xTg mice.

RVG-EXOs promote the clearance of pathological proteins such as Aβ and P-Tau and protect neurons

The potential cause of cognitive dysfunction in 3xTg mice could be the aggregation of pathological proteins such as Aβ and P-Tau, which were closely associated with impaired synaptic plasticity and neuronal degeneration [22]. To verify whether RVG-EXOs could promote the clearance of pathological proteins such as Aβ and P-Tau and protect neurons, the level of 647-fluorescent-labeled Aβ1-42 (647-Aβ1-42) was detected in SY5Y cells subjected to different treatments. In 24 h, cells treated with Aβ1-42 alone still exhibited a substantial amount of 647-Aβ1-42, indicating impaired intracellular Aβ clearance capacity. In contrast, treatment with either EXOs or RVG-EXOs enhanced the cells' ability to clear Aβ, with more pronounced enhancement detected after RVG-EXO treatment (Fig. 3A). These results confirmed that both EXOs and RVG-EXOs could promote the clearance of Aβ protein within cells, with RVG-EXOs exhibiting a stronger effect. To examine the effect of RVG-EXOs on Tau protein hyperphosphorylation, the level of P-Tau (at Ser396) was measured in cells treated with okadaic acid (OA) alone or followed by treatment with either EXOs or RVG-EXOs. Western blot analysis revealed that OA treatment significantly increased Tau phosphorylation at the Ser396 site compared to the normal control group. Both EXOs and RVG-EXOs reduced the level of P-Tau to some extent, with RVG-EXOs showing a stronger effect (Fig. 3B and C).

Fig. 3: RVG-EXOs promote the clearance of pathological proteins such as Aβ and P-Tau both in vitro and in vivo and protect neurons. (A) Representative immunofluorescence images and quantitative analysis of 647-Aβ1-42 (n = 3). Green, 647-Aβ1-42; red, Neun; blue, DAPI staining of nuclei. (B) Western blot analysis of P-Tau under different treatment conditions and its quantification (C) (n = 3). (D) Calcein AM/PI staining images of SH-SY5Y cells under different treatment conditions and quantitative analysis of dead and live cells (n = 3). Red, PI staining of dead cell; green, Calcein AM staining of viable cells. (E) Representative immunofluorescence images and quantitative analysis of Aβ deposition in the hippocampal region of mice from different treatment groups (n = 3). Green, Aβ; blue, DAPI staining of nuclei. (F) Western blot analysis of P-Tau in mouse brain tissues from different treatment groups and its quantification (G) (n = 3). (H) Representative immunofluorescence images and quantitative analysis of P-Tau in the hippocampal region of mice from different treatment groups (n = 3). Green, P-Tau (S202/T205); blue, DAPI staining of nuclei. (I) Nissl staining images of the hippocampal region in mice from different treatment groups and quantitative analysis of neuronal number. The upper panels show panoramic views of the hippocampus (scale bar = 250 μm); the lower panels are magnified views of the corresponding areas indicated by black boxes (scale bar = 100 μm). The bar graph presents the statistical results of neuronal counts for the entire hippocampal region based on the panoramic images (n = 3). (J) Representative images of Golgi-stained mouse brain tissues from different treatment groups and quantitative analysis of dendritic spine density (n = 3). Data are presented as mean ± SD (n = 3). Statistical significance is determined by the Student's t-test based on P < 0.05 (), P < 0.01 (), P < 0.001 (), and P < 0.0001 (****).

Additionally, it was observed that exposure to Aβ1-42 severely impaired neurite outgrowth. Both EXOs and RVG-EXOs partially rescued the neurite morphology, with RVG-EXO treatment demonstrating the most effective restoration of neurite length (Fig. S2A). Calcein-AM/PI staining revealed that treatment with Aβ1-42 led to a significant increase in the number of dead cells and a corresponding decrease in the number of live cells. The addition of either EXOs or RVG-EXOs provided a certain degree of neuronal protection, reducing neuronal death, with RVG-EXOs demonstrating a stronger cytoprotective effect (Fig. 3D). In summary, these results indicated that RVG-EXOs could effectively promote the clearance of pathological proteins, such as Aβ and P-Tau, and protect neurons in vitro.

Next, we further examined whether RVG-EXOs could reduce Aβ and P-Tau pathology and protect neurons in vivo. First, immunofluorescence staining of Aβ was performed on mouse brain tissues. The results showed that, compared with the Tg-Vehicle group, the Aβ plaques in the hippocampal region of the mouse brains were reduced to some extent after treatment with either RVG-EXOs or EXOs, with the more pronounced effect detected in the RVG-EXO-treated group (Fig. 3E). Subsequently, the effect of RVG-EXOs on P-Tau were investigated in mouse brain tissues. Western blot analysis of the brain tissues revealed that the level of P-Tau at the Ser396 site was upregulated in 9-month-old 3xTg mice. Both RVG-EXOs and EXOs reduced the level of P-Tau, with the more significant reduction observed in the RVG-EXOs-treated group (Fig. 3F and G). Immunofluorescence staining of P-Tau further confirmed that RVG-EXOs significantly reduced the level of P-Tau (Fig. 3H).

Synaptophysin (SYAP) is a marker protein of neuronal synapses and is widely considered a reliable indicator of synaptic plasticity and function [23]. Our results revealed a significant decrease in SYAP expression in 3xTg mice, indicating impaired synaptic plasticity and function, which was closely associated with the previously observed cognitive dysfunction. Following treatment with either RVG-EXOs or EXOs, the level of SYAP in 3xTg mice was restored, with the treatment of RVG-EXOs demonstrating a more pronounced effect (Fig. S2B). Furthermore, Nissl staining indicated that treatment with either RVG-EXOs or EXOs could suppress neuronal loss, with the more pronounced effect detected in the treatment of RVG-EXOs (Fig. 3I). Dendritic spines, serving as the primary sites for excitatory synaptic input in brain tissue, are closely associated with synaptic activity and likely play a key role in synaptic transmission and plasticity, forming the cellular basis for learning and plasticity in the brain [24]. Golgi staining revealed a significant reduction in dendritic spine density in the brain tissues of 3xTg mice. Notably, treatment with either RVG-EXOs or EXOs restored the dendritic spine density in the mouse brain tissues, with the more pronounced effect revealed in the treatment of RVG-EXOs (Fig. 3J).

RVG-EXOs enhance neuronal autophagy function both in vitro and in vivo

Previous studies have demonstrated that EXOs serve as key mediators of intercellular communication, primarily through the transport of non-coding RNAs, particularly miRNAs [25,26]. Upon entering recipient cells, these miRNAs exert biological effects by influencing target gene expression via post-transcriptional regulatory networks. To investigate the mechanisms by which RVG-EXOs cleared pathological proteins such as Aβ and P-Tau and protected neurons, miRNA sequencing was performed on plasma-derived EXOs from six young individuals, six elderly individuals, and six patients with AD. KEGG pathway enrichment analysis revealed that, compared with EXOs from the elderly individuals or AD patients, the target genes of differentially expressed miRNAs in young plasma-derived EXOs were significantly enriched in autophagy-related pathways (Fig. 4A and B). Autophagy is recognized as a crucial protein degradation pathway capable of clearing misfolded proteins in various neurodegenerative diseases, including AD [7,14,15]. Additionally, the preparation of RVG-EXOs involved only conjugating the RVG29 to the exosomal membrane surface using DSPE-PEG2000, without altering their internal molecular composition. Therefore, it was speculated that the engineered young plasma-derived EXOs primarily exerted their effects by modulating the autophagy pathway. Microtubule-associated protein 1 light chain 3β (MAP1LC3B, commonly known as LC3B-I) is a widely used autophagy marker protein predominantly localized in the cytoplasm. During autophagy initiation, LC3B-I undergoes lipidation and is converted into LC3B-II, which is then recruited to the autophagosomal membrane to participate in autophagosome formation [7]. Another commonly used autophagy marker, SQSTM1/p62, is an adaptor protein that binds to LC3B and mediates the targeting of ubiquitinated protein aggregates for degradation via the autophagy pathway [27]. Aβ1-42 monomers are frequently employed to establish an AD cell model with impaired autophagy regulation in vitro [28]. Following exposure to 20 μM Aβ1-42 for 48 h, SY5Y cells exhibited decreased expression of LC3B-II and increased expression of P62, indicating suppressed autophagy in SY5Y cells [29]. In our study, after treatment with either RVG-EXOs or EXOs for 24 h, the expression level of LC3B-II was elevated, whereas the expression level of P62 was reduced, suggesting that both RVG-EXOs and EXOs partially restored the neuronal autophagy impairment induced by Aβ1-42, with RVG-EXOs demonstrating a stronger effect (Fig. 4C–F). This observation was further confirmed by immunofluorescence staining of both LC3B and P62 (Figs. S3A and 4G). TEM analysis revealed that the number of autolysosomes was significantly reduced in cells pretreated with Aβ1-42. In contrast, treatment with either RVG-EXOs or EXOs restored the number of autolysosomes, with the effect of RVG-EXOs being more pronounced (Fig. 4H). Immunofluorescence imaging demonstrated that, compared with cells treated solely with Aβ1-42, cells treated with RVG-EXOs exhibited a significant increase in the number of red puncta, representing autolysosomes (Fig. 4I). In summary, these results indicated that RVG-EXOs could enhance autophagy function in neurons in vitro.

Fig. 4: RVG-EXOs enhance neuronal autophagy function both in vitro and in vivo. MiRNA sequencing analysis is performed on plasma EXOs derived from six young individuals, six elderly individuals, and six patients with AD. (A) KEGG pathway enrichment analysis of target genes based on the differential miRNA expression profiles of plasma EXOs from AD patients and young individuals, revealing a significant enrichment of these target genes in the autophagy pathway. (B) KEGG pathway enrichment analysis of target genes based on differential miRNA expression profiles of plasma EXOs from elderly and young individuals, demonstrating significant enrichment of these target genes in the autophagy pathway. (C) Western blot analysis of LC3B in SH-SY5Y cells under different treatments and its quantification (D) (n = 3). (E) Western blot analysis of P62 in SH-SY5Y cells under different treatments and its quantification (F) (n = 3). (G) Representative immunofluorescence images of P62 in SH-SY5Y cells under different treatments and its quantitative analysis (n = 3). Blue, DAPI staining of nuclei; red, Neun; green, P62. (H) Representative TEM images of cells under different treatments and quantitative analysis of autolysosomes. Red arrows indicate autolysosomes. (I) Representative image and quantification of autophagic flux in SY5Y cells transfected with pmCherry-EGFP-LC3b after different treatments (n = 3). Red, pmCherry-LC3; green, EGFP-LC3. pmCherry-LC3 puncta represent autolysosomes; EGFP-LC3 puncta correspond to autophagosomes. (J) Representative immunofluorescence images of P62 in the hippocampal Dentate Gyrus (DG) region of mice under different treatments and its quantitative analysis (K) (n = 3). Blue, DAPI staining of nuclei; red, Neun; green, P62. (L) Western blot analysis of P62 and LC3B in mouse brain tissues under different treatments. (M) Quantitative analysis of P62 protein in mouse brain tissues (n = 3). (N) Quantitative analysis of LC3B-II protein in mouse brain tissues (n = 3). (O) Representative TEM images of mouse brain tissues under different treatments and quantitative analysis of autolysosomes (n = 3). Red arrows indicate autolysosomes. Data are presented as mean ± SD (n = 3). Statistical significance is determined by the Student's t-test based on P < 0.05 (), P < 0.01 (), and P < 0.001 ().

To further validate whether RVG-EXOs could improve autophagy function in the brain tissues of 3xTg mice, immunofluorescence staining was first employed to examine the expression of the autophagic substrate P62. The results showed that after treatment of RVG-EXOs, the level of P62 protein in the brain tissues of the mice was significantly reduced (Fig. 4J and K). These findings indicated an improvement in autophagic activity in the mouse brain tissues. Subsequently, the expression of autophagy-related proteins LC3B and P62 was assessed in mouse brain tissues via Western blot analysis. The results revealed that the level of LC3B-II protein was decreased, whereas the expression level of P62 protein was increased in the brain tissues of 3xTg mice, indicating impaired autophagy function. In contrast, treatment of RVG-EXOs elevated the level of LC3B-II protein and reduced the level of P62 protein, thereby partially restoring autophagy function in the mouse brain tissues (Fig. 4L–N). TEM analysis of mouse brain tissues revealed that treatment of RVG-EXOs significantly increased the number of autolysosomes (Fig. 4O). These results indicated that RVG-EXOs could partially restore the autophagy function in the brain tissues of 3xTg mice.

In vitro, RVG-EXOs primarily enhance neuronal autophagy, promote the clearance of pathological proteins such as Aβ and P-Tau, and protect neurons by inhibiting the expression of the RPTOR gene

To further investigate the potential mechanisms by which RVG-EXOs enhanced neuronal autophagy, promoted the clearance of pathological proteins such as Aβ and P-Tau, and protected neurons, an in-depth analysis of the miRNA sequencing data was performed. The results revealed that, compared to plasma EXOs from elderly individuals, young plasma-derived EXOs exhibited significantly higher levels of hsa-miR-18a-3p, hsa-miR-23a-3p, hsa-miR-23b-3p, hsa-miR-4286, hsa-miR-4508, and hsa-miR-766-3p. Furthermore, compared to plasma EXOs from AD patients, young plasma-derived EXOs showed significantly elevated levels of hsa-miR-23a-3p and hsa-miR-23b-3p (Fig. 5A and B), and all six of these miRNAs were computationally predicted to target the RPTOR gene. Given that the expression levels of hsa-miR-23a-3p and hsa-miR-23b-3p were significantly upregulated in plasma EXOs from healthy young individuals compared to those from both elderly individuals and patients with AD, we further conducted target validation and functional studies on these two miRNAs. First, we examined the direct binding of hsa-miR-23a-3p and hsa-miR-23b-3p to the 3′ UTR of the RPTOR gene using dual-luciferase reporter assays. The results showed that co-transfection of hsa-miR-23a-3p or hsa-miR-23b-3p mimics with a wild-type reporter plasmid containing the RPTOR-3′ UTR led to a significant decrease in relative luciferase activity, whereas no significant change was observed when the miRNA mimics were co-transfected with a reporter plasmid carrying mutated binding sites (Fig. S4A). This indicated that hsa-miR-23a-3p and hsa-miR-23b-3p could directly target the 3′ UTR of RPTOR. Furthermore, we transfected SH-SY5Y cells with inhibitors of hsa-miR-23a-3p or hsa-miR-23b-3p; qRT-PCR analysis confirmed that the levels of hsa-miR-23a-3p and hsa-miR-23b-3p were significantly reduced (Fig. S4B). Subsequently, changes in endogenous Raptor protein expression were detected by Western blot. The results showed that inhibition of endogenous miR-23a-3p or miR-23b-3p led to a significant increase in Raptor protein levels in the cells (Fig. S4C and D). Taken together, these results confirmed that both hsa-miR-23a-3p and hsa-miR-23b-3p could target the RPTOR gene and negatively regulate its expression. Given that the preparation of RVG-EXOs involved only conjugating the RVG29 peptide to the exosomal membrane using DSPE-PEG2000 without altering their internal molecular cargo, RVG-EXOs were also expected to carry these miRNAs. To verify this, we performed qRT-PCR analysis to compare the levels of these six miRNAs between unmodified EXOs and RVG-EXOs. The results showed no significant differences in the expression levels of hsa-miR-18a-3p, hsa-miR-23a-3p, hsa-miR-23b-3p, hsa-miR-4286, hsa-miR-4508, and hsa-miR-766-3p between EXOs and RVG-EXOs (Fig. S4E–J). Furthermore, Western blot analysis demonstrated that Aβ1-42 treatment upregulated the expression of Raptor protein in SH-SY5Y cells, whereas treatment of RVG-EXOs reduced Raptor protein expression (Fig. 5C). These results suggested that RVG-EXOs could partially suppress the expression of the RPTOR gene, a finding further corroborated by qPCR results (Fig. 5D). Raptor, a core scaffolding protein of the mTORC1 complex, plays a central negative regulatory role in cellular autophagy by mediating the inhibitory effect of mTORC1 on the ULK1 autophagy initiation complex [30]. Therefore, it was hypothesized that RVG-EXOs likely enhanced cellular autophagic activity primarily by suppressing the expression of the RPTOR gene. To test this hypothesis, an RPTOR-overexpressing SH-SY5Y cell model was established using lentiviral transfection (Fig. 5E). Immunofluorescence results revealed that, compared with the control group (Tg-RVG-EXO-NC), RPTOR overexpression (Tg-RVG-EXO-OE) significantly increased the level of P62 protein and decreased the expression of LC3B-II protein in cells (Fig. 5F and S4K). Western blot analysis further confirmed this finding (Fig. 5G–I). LAMP1 and LC3B are characteristic marker proteins of the lysosomal membrane and autophagosomal membrane, respectively. Immunofluorescence colocalization analysis revealed that RPTOR overexpression significantly attenuated the colocalization of LAMP1 and LC3B (Fig. 5J). In summary, these results demonstrated that RPTOR overexpression could significantly counteract the ameliorative effects of RVG-EXOs on autophagy in SH-SY5Y cells, suggesting that RVG-EXOs enhanced neuronal autophagy primarily through the suppression of RPTOR gene expression.

Fig. 5: RVG-EXOs primarily enhance neuronal autophagy, promote the clearance of pathological proteins such as Aβ and P-Tau, and protect neurons by inhibiting the expression of the RPTOR gene in vitro. MiRNA sequencing analysis is performed on plasma EXOs derived from six young individuals, six elderly individuals, and six patients with AD. (A) MiRNAs significantly more abundant in plasma EXOs from young individuals compared to the elderly individuals and targeting the RPTOR gene. (B) MiRNAs significantly more abundant in plasma EXOs from young individuals compared to those of AD patients and targeting the RPTOR gene. (C) Western blot analysis and quantification of Raptor protein in cells under different treatments (n = 3). Using untreated SH-SY5Y cells as the control, the experiment shows changes in protein expression after 48-h pretreatment with 20 μM Aβ1-42 followed by 24-h co-culture with either PBS or RVG-EXOs. (D) Quantitative qPCR analysis showing RPTOR mRNA expression levels in cells under different treatments (n = 3). (E) Western blot analysis of Raptor protein in SH-SY5Y cells after transfection with RPTOR overexpression lentivirus (RPTOR-OE) or empty lentivirus (RPTOR-NC). (F) Immunofluorescence images and quantification of P62 protein in cells under different treatments (n = 3). Green, P62; red, Neun; blue, DAPI staining of nuclei. (G) Western blot analysis of LC3B and P62 proteins in cells under different treatments. (H) Quantification of LC3B-II protein expression levels in cells under different treatments (n = 3) based on Western blot analysis. (I) Quantification of P62 protein expression levels in cells under different treatments (n = 3) based on Western blot analysis. (J) Immunofluorescence co-localization images of LC3B and LAMP1 in cells under different treatments. Green fluorescence indicates LC3B (autophagosome marker), red fluorescence indicates LAMP1 (lysosome marker), and yellow regions show their co-localization (Merge), representing autolysosomes. The Pearson's correlation coefficient R represents the overlap ratio calculated by ImageJ (n = 3). (K) Representative immunofluorescence images and quantitative analysis of 647-Aβ1-42 in cells under different treatments (n = 3). Green, 647-Aβ1-42; red, Neun; blue, DAPI staining of nuclei. (L) Western blot analysis and quantification of P-Tau in cells (M) following different treatments (n = 3). (N) Immunofluorescence micrographs showing MAP2 in SH-SY5Y cells under different treatments. Green, MAP2; DAPI staining of nuclei. (O) Representative Calcein-AM/PI double-stained fluorescence images of SH-SY5Y cells and the corresponding quantification of dead and live cells following different treatments (n = 3). Red, PI staining of dead cell; green, Calcein AM staining of viable cells. Data are presented as mean ± SD (n = 3). Statistical significance is determined by the Student's t-test based on P < 0.05 (), P < 0.01 (), P < 0.001 (), and P < 0.0001 (****); ns, not significant.

Next, we further assessed whether RVG-EXOs exerted their effects of promoting pathological protein clearance and protecting neurons by inhibiting the expression of the RPTOR gene. Immunofluorescence detection revealed that under RPTOR-overexpressing conditions, the fluorescence intensity of 647-Aβ1-42 in the cells was significantly enhanced (Fig. 5K). Western blot analysis further demonstrated that following RPTOR overexpression, the expression level of P-Tau protein was markedly increased (Fig. 5L and M). Immunofluorescence staining of MAP2 revealed that RPTOR overexpression resulted in a significant reduction in neurite length (Fig. 5N). Additionally, Calcein-AM/PI double staining demonstrated that RPTOR overexpression led to a significant decrease in the number of live cells, while the number of dead cells was increased correspondingly (Fig. 5O). Collectively, these findings indicated that RPTOR overexpression significantly counteracted the ability of RVG-EXOs to promote the clearance of pathological proteins (such as Aβ and P-Tau) and to protect neurons. These results further suggested that RVG-EXOs exerted these beneficial effects primarily by inhibiting the expression of the RPTOR gene.

In vivo, RVG-EXOs improve brain tissue autophagy, alleviate Aβ and P-Tau pathology, and enhance cognitive behaviors in mice by suppressing the expression of the RPTOR gene

Upon the establishment that RVG-EXOs functioned by inhibiting RPTOR gene expression in vitro, animal experiments were further performed to investigate whether, in vivo, they similarly operated through the RPTOR-mediated autophagy pathway to alleviate cerebral Aβ and P-Tau pathological deposition and improve cognitive behaviors in AD model mice. Western blot analysis revealed that Raptor protein expression was significantly upregulated in the hippocampal tissues of 3xTg mice compared with wild-type mice, whereas its expression level was markedly downregulated following RVG-EXOs intervention (Fig. 6A). Western blot analysis of cortical tissues also exhibited a comparable trend (Fig. S5A). These results indicated that RVG-EXOs could also downregulate RPTOR expression in vivo. To further verify whether RVG-EXOs exerted their effects by suppressing RPTOR gene expression in vivo, an RPTOR-overexpression model was established in the mouse brain via stereotactic injection of an RPTOR-overexpressing lentivirus. Subsequently, the upregulation of Raptor protein in this model was confirmed by Western blot analysis (Fig. 6B and S5B). Thereafter, the experimental mice were subjected to grouped interventions and behavioral tests, and brain tissues were collected at the experimental endpoint for subsequent analysis. If the in vivo effects of RVG-EXOs similarly depended on the suppression of RPTOR expression, then RPTOR overexpression should significantly attenuate their therapeutic effects. Western blot results demonstrated that, compared with the control group (Tg-RVG-EXO-NC), the RPTOR-overexpressing group (Tg-RVG-EXO-OE) exhibited a significant decrease in the level of LC3B-II protein and a marked increase in P62 protein expression in mouse brain tissues (Fig. 6C–E). Immunofluorescence analysis further confirmed that under RPTOR-overexpressing conditions, the expression level of P62 protein in brain tissues was significantly upregulated (Fig. 6F). These results indicated that RPTOR overexpression significantly counteracted the beneficial effects of RVG-EXOs on autophagy in mouse brain tissues. TEM analysis revealed that under RPTOR-overexpressing conditions, the number of autolysosomes in mouse brain tissues was significantly reduced (Fig. 6G). In summary, these results demonstrated that overexpression of the RPTOR gene significantly inhibited the improvement of autophagic function in mouse brain tissues mediated by RVG-EXOs, indicating that RVG-EXOs primarily enhanced brain tissue autophagy by suppressing the expression of the RPTOR gene.

Fig. 6: RVG-EXOs improve brain tissue autophagy, alleviate Aβ and P-Tau pathology, and enhance cognitive behaviors in mice by suppressing RPTOR gene expression in vivo. (A) Western blot analysis and quantification of Raptor protein in the hippocampal tissue of mouse brain following different treatments (n = 3). (B) Western blot analysis and quantification of Raptor protein in mouse hippocampal tissue (n = 3), showing the protein expression levels of Raptor in the hippocampal region of 3xTg mice following stereotaxic injection with either an RPTOR-overexpressing lentivirus (Tg-RPTOR-OE) or an empty lentivirus control (Tg-RPTOR-NC). (C) Western blot analysis and quantification of LC3B protein in mouse brain tissues after different treatments (n = 3). (D) Western blot analysis of P62 protein and its quantification (E) in mouse brain tissues under different treatments (n = 3). (F) Immunofluorescence images and quantitative analysis of P62 protein in hippocampal neurons of mouse brain tissues under different treatments (n = 3). Green, P62; red, Neun; blue, DAPI staining of nuclei. (G) Representative TEM images of brain tissues from each group of mice and quantitative analysis of autophagolysosomes (n = 3). Red arrows indicate autophagolysosome structures. (H) Representative immunofluorescence images and quantitative analysis of Aβ plaques in the hippocampal tissue of mice from each experimental group (n = 3). Green, Aβ; blue, DAPI staining of nuclei. (I) Western blot analysis of P-Tau in mouse brain tissues and its quantitative analysis (J) across experimental groups (n = 3). (K) Representative immunofluorescence images and quantitative analysis of P-Tau in the hippocampal region of mouse brain tissues across experimental groups (n = 3). Green, P-Tau (S202/T205); blue, DAPI staining of nuclei. (L) Nissl staining of neurons and quantitative analysis of neuron numbers in the hippocampal region across experimental groups of mice. The top panel displays a panoramic view of the hippocampus (scale bar = 250 μm), and the lower panel shows a magnified view of the corresponding area outlined in black (scale bar = 100 μm). The bar graph presents the statistical results of neuron counts for the entire hippocampal region based on the panoramic images (n = 3). (M) Western blot analysis of SYAP in mouse brain tissues and its quantitative analysis across experimental groups (n = 3). (N) Schematic diagram of the mouse behavioral experiment. (O) Representative trajectory heatmaps from the platform exploration test of MWM for mice in each experimental group. (P) Quantitative analysis of the number of target platform crossings during the MWM probe test for mice in each experimental group (n = 8). (Q) Representative escape latency plots for the MWM test across groups of mice (n = 8). (R) Representative OFT tracks and analysis of center versus periphery time (n = 8). Data are presented as mean ± SD. Statistical significance is determined by the Student's t-test based on P < 0.05 (), P < 0.01 (), and P < 0.001 (); ns, not significant.

We further investigated whether RVG-EXOs alleviated Aβ and P-Tau protein pathology, protected neurons, and exerted their therapeutic effects in vivo by inhibiting the expression of the RPTOR gene. Immunofluorescence analysis of mouse brain tissues revealed that, compared with the control group (Tg-RVG-EXO-NC), RPTOR overexpression resulted in a significant increase in the number of Aβ plaques in the brain (Fig. 6H). Western blot analysis of P-Tau protein demonstrated that under conditions of RPTOR overexpression, the level of P-Tau protein in mouse brain tissues was significantly elevated (Fig. 6I and J). Immunofluorescence analysis of P-Tau protein further corroborated this finding (Fig. 6K). Nissl staining of the mouse hippocampal region revealed that RPTOR overexpression exacerbated the loss of hippocampal neurons (Fig. 6L). Western blot analysis further demonstrated that under conditions of RPTOR overexpression, the level of SYAP in mouse brain tissues was significantly decreased (Fig. 6M), indicating that RPTOR overexpression exacerbated hippocampal neuronal damage and markedly counteracted the neuroprotective effects of RVG-EXOs. Collectively, these results demonstrated that RPTOR overexpression significantly counteracted the ability of RVG-EXOs to reduce pathological protein deposition and protect neurons in brain tissues, suggesting that RVG-EXOs exerted these therapeutic effects primarily by suppressing the expression of the RPTOR gene.

We further explored how RVG-EXOs improved cognitive function in mice in vivo and examined whether this effect was similarly dependent on the suppression of RPTOR gene expression using 8-month-old 3xTg mice. An RPTOR-overexpression model was established in the mouse brain via stereotaxic injection of an RPTOR-overexpressing lentivirus; mice injected with an empty lentivirus served as controls. Subsequently, the experimental mice were subjected to grouped interventions. Upon completion of the interventions, all mice underwent a series of behavioral tests, including MWM, Y-maze, OFT, and NOR test (Fig. 6N).

First, the MWM was used to assess whether the improvement of spatial learning and memory in mice by RVG-EXOs was dependent on the suppression of the RPTOR gene expression. On the first day, the mice underwent visual training to assess their vision and swimming ability, followed by 5 d of hidden platform training for all groups. On the 7th day, the platform was removed, and a probe test was conducted using animal tracking software to evaluate the memory function of the mice. Notably, during the hidden platform training, compared with the empty-vector control mice (Tg-RVG-EXO-NC), the RPTOR-overexpressing mice exhibited slower learning ability and longer escape latencies, indicating that RPTOR overexpression attenuated the improvement of spatial learning ability in 3xTg mice conferred by RVG-EXOs. In the subsequent probe test, the RPTOR-overexpressing mice crossed the target platform significantly fewer times than the empty-vector control mice. In summary, the MWM results demonstrated that RPTOR overexpression significantly counteracted the improvement of spatial learning and memory in 3xTg mice by RVG-EXOs, suggesting that the cognitive-enhancing mechanisms of RVG-EXOs operated primarily through the suppression of RPTOR gene expression (Fig. 6O–Q).

Next, the Y-maze test was employed to assess short-term spatial working memory. The results showed that under conditions of RPTOR gene overexpression, the spontaneous alternation rate of the mice was significantly decreased (Fig. S5C). These results indicated that RPTOR overexpression could significantly counteract the improvement effect of RVG-EXOs on short-term spatial working memory, further suggesting that this improvement could be primarily achieved through the inhibition of RPTOR gene expression.

Subsequently, an OFT was conducted to evaluate anxiety-like behavior in the mice. Each mouse was placed in the center of a novel open arena and allowed to explore freely for 5 min. Its movement trajectory as well as the time spent in the central and peripheral zones were recorded. Notably, compared with the empty-vector control mice, the RPTOR-overexpressing mice showed a significant reduction in the time spent in the central zone (Fig. 6R). This indicated that RPTOR overexpression significantly attenuated the ameliorative effect of RVG-EXOs on anxiety-like behavior in mice, suggesting that RVG-EXOs alleviated such behavior primarily by suppressing the expression of the RPTOR gene.

Finally, the NOR test was conducted to assess object recognition memory in mice. The results showed that, compared with the empty-vector control mice, the RPTOR-overexpressing mice exhibited a significant decrease in the RI (Fig. S5D). This indicated that RPTOR overexpression significantly attenuated the ameliorative effect of RVG-EXOs on object recognition memory in mice, suggesting that RVG-EXOs exerted the beneficial effects primarily by inhibiting the expression of the RPTOR gene. In summary, these behavioral tests indicated that RPTOR overexpression could significantly attenuate the beneficial effects of RVG-EXOs on cognitive behaviors in mice, suggesting that RVG-EXOs improved cognitive performance primarily by inhibiting the expression of the RPTOR gene.

Single-cell RNA Sequencing Reveals that RVG-EXOs exert neuroprotection via Remodeling Neuronal Transcriptome, subtype composition, and intercellular communication networks

To further elucidate the pharmacological mechanisms of RVG-EXOs at the cellular and molecular levels, we performed single-cell RNA sequencing (scRNA-seq) on brain tissues from 3×Tg mice treated with either PBS (Tg-Vehicle) or RVG-EXOs (Tg-RVG-EXO). Unbiased clustering of all captured cells identified major cell types including Oligodendrocytes, Neurons, Astrocytes, Endothelial cells, Microglia, Monocytes, T cells, Epithelial cells, Fibroblasts, Macrophages, Granulocytes and B cells (Figs. S6A, S7A, S8A). Remarkably, compared with the Tg-Vehicle group, RVG-EXOs treatment led to a significant increase in the proportion of neurons (Fig. 7A–C). This observation is consistent with the neuroprotective effects of RVG-EXOs demonstrated by Nissl staining and dendritic spine analysis. Differential gene expression analysis of neurons revealed that, compared with the PBS control group, RVG-EXOs treatment led to significant upregulation of 276 genes and downregulation of 397 genes in neurons, including Rptor (Fig. 7D and E). KEGG pathway enrichment analysis of the differentially expressed genes in neurons revealed that they were significantly enriched in multiple pathways, including autophagy, lysosomal function, the neurotrophin signaling pathway, and others (Fig. 7F). Notably, the enrichment of autophagy-related genes and the downregulation of the Rptor gene in neurons further corroborate our mechanistic finding that RVG-EXOs promote autophagic flux in neurons by inhibiting the expression of RPTOR.

Fig. 7: Single-Cell RNA Sequencing Reveals that RVG-EXOs Exert Neuroprotection via Remodeling Neuronal Transcriptome, Subtype Composition, and Intercellular Communication Networks. (A) UMAP plot showing the classification of cell subpopulations in brain tissue of PBS-treated mice. (B) UMAP plot showing the classification of cell subpopulations in brain tissue of RVG-EXO-treated mice. (C) Quantitative plot showing the proportions of different cell types in the brain tissue of PBS-treated and RVG-EXO-treated mice based on single-cell sequencing analysis. (D) Quantitative plot showing differentially expressed genes in neurons of brain tissue from RVG-EXO-treated 3xTg mice compared with PBS-treated controls, based on single-cell sequencing analysis. (E) Volcano plot showing up- and down-regulated differentially expressed genes in neurons of brain tissue from RVG-EXO-treated 3xTg mice compared with PBS-treated controls, based on single-cell sequencing analysis, with autophagy-related genes marked in the plot. (F) Bubble plot showing the significance of KEGG pathway enrichment for differentially expressed genes in neurons between PBS-treated and RVG-EXO-treated mice, based on single-cell sequencing analysis. (G) Bubble plot showing the expression of marker genes in neuronal subpopulations. The size of each bubble (pct.exp) represents the proportion of cells expressing the gene within that subpopulation, and the color (avg.exp.scale) indicates the average expression level of the gene. (H) UMAP plot of excitatory and inhibitory neurons. (I) Stacked bar plot showing the numbers of excitatory and inhibitory neurons in each treatment group. (J) Stacked bar chart showing the proportions of excitatory and inhibitory neurons in each treatment group. (K) Representative immunofluorescence images of Camk2a protein in different treatment groups. Camk2a + indicate excitatory neurons. Green, Camk2a; red, Neun; blue, DAPI staining of nuclei. (L) Representative immunofluorescence images of GAD1 protein in different treatment groups. GAD1+ indicate inhibitory neurons. Green, GAD1; red, Neun; blue, DAPI staining of nuclei. (M) Bubble plot showing differences in ligand-receptor pair expression levels in cell communication between neurons and microglia. Bubble color represents interaction probability, and bubble size represents the P-value for significance.

Given the significant increase in neuronal proportion and the critical role of different neuronal subtypes in cognitive function, we further classified neurons into excitatory (E) and inhibitory (I) subpopulations based on the expression of canonical marker genes (Excitatory: Slc17a7, Camk2a, Slc17a6 [31]; Inhibitory: Gad1, Gad2, Slc32a1, Rora [32,33]) (Fig. 7G and H). This analysis revealed that RVG-EXOs treatment not only increased the absolute number of both excitatory and inhibitory neurons but also elevated the percentage of inhibitory neurons within the total neuronal population (Fig. 7I and J). Camk2a and GAD1 serve as well-established markers for excitatory and inhibitory neurons, respectively [34]. Therefore, we performed immunofluorescence staining to examine their expression levels, thereby assessing the alterations in these two neuronal types. Immunofluorescence results also showed that RVG-EXOs treatment significantly increased the levels of excitatory and inhibitory neurons (Fig. 7K, L and S9A, B). Given that E/I imbalance is a hallmark of AD pathophysiology and its restoration is associated with cognitive improvement and protection against excitotoxicity [35], this shift in neuronal subtype composition suggests that RVG-EXOs may rebalance neural network activity in brain tissue.

To gain further mechanistic insights at the intercellular level, we performed cell-cell communication analysis, which revealed that RVG-EXOs treatment significantly strengthened the interaction between neurons via the Ptn-Sdc3 ligand-receptor pair, a pathway critically involved in promoting synaptic plasticity, neurite outgrowth, and neuronal survival [36,37]. Conversely, RVG-EXOs significantly weakened the communication between neurons and microglia through the App-Cd74 ligand-receptor pair, which is closely associated with Disease-Associated Microglia (DAM) generation and neuroinflammation (Fig. 7M). [38,39]. These findings suggest that RVG-EXOs not only intrinsically enhance neuronal health but also reshape the brain's intercellular signaling network to foster a more neuroprotective and anti-inflammatory microenvironment.

Single-Cell Sequencing Reveals that RVG-EXOs Suppress Pathogenic DAM Generation and Reshape Microglial Homeostasis via Autophagy-Dependent Downregulation of APP-CD74 signaling

As mentioned above, intercellular communication analysis revealed that RVG-EXOs significantly attenuate the interaction of the App–Cd74 ligand–receptor pair between neurons and microglia, which is closely associated with DAM generation and neuroinflammation [38,39]. Immunofluorescence results of brain tissues and cells showed that RVG-EXOs treatment significantly reduced APP protein levels and diminished APP–CD74 receptor binding. However, this effect was markedly attenuated upon the addition of an autophagy inhibitor, a finding further corroborated by Western blot analysis (Fig. 8A–D and S9C-F). Collectively, these data suggest that the observed reduction in APP–CD74 ligand–receptor interactions between neurons and microglia following RVG-EXOs treatment is primarily attributable to enhanced neuronal autophagy, which lowers APP protein abundance. Considering the pivotal role of microglia in AD pathology, we next performed a sub-clustering analysis of microglial populations. Based on established markers, we identified four distinct microglial subtypes: Homeostatic Microglia (expressing Cx3cr1, P2ry12, P2ry13, Gpr34, Tmem119, Selplg, Olfml3 [40].), DAM (expressing Lyz2, ApoE, CST7, Cd74, Lgals3, Clec7a [[40], [41], [42]].), Proliferating Microglia (expressing Mki67, Top2a, Pcna [43,44]), and Transiting Response Microglia (a transitional microglial state between DAM and homeostatic microglia and ApoE levels were notably higher [45]) (Fig. 8E and F). Strikingly, RVG-EXOs treatment led to a significant decrease in both the absolute numbers and relative proportions of the deleterious DAM and Proliferating microglia subtypes, while concurrently increasing the proportion of the homeostatic, surveillant microglia (Fig. 8G–I). CLEC7A is a commonly used marker of DAM [46]. In this study, immunofluorescence staining was performed to detect its expression level, so as to reflect the changes in DAM. Immunofluorescence also revealed that treatment with RVG-EXOs significantly reduced the levels of DAM (Fig. 8J and S9G). These findings indicate that RVG-EXOs may exert neuroprotective functions by reducing the interaction between neuronal APP and microglial CD74, thereby inhibiting the generation of DAM, a phenotype closely associated with neuroinflammation [47].

Fig. 8: Single-Cell Sequencing Reveals that RVG-EXOs Suppress Pathogenic DAM Generation and Reshape Microglial Homeostasis via Autophagy-Dependent Downregulation of APP-CD74 Signaling. (A) Schematic of neuron–microglia co-culture. (B) Representative Immunofluorescence images of APP and CD74 proteins in cells subjected to different treatments. Green, CD74; red, APP; blue, DAPI staining of nuclei. (C) Representative immunofluorescence images of APP and CD74 proteins in mouse brain tissues after different treatments. Green, CD74; red, APP; blue, DAPI staining of nuclei. (D) Western blot analysis and quantification of APP protein in cells under various treatments (n = 3). Data are presented as mean ± SD. Statistical significance is determined by the Student's t-test based on P < 0.05 (), P < 0.01 (), and P < 0.001 (); ns, not significant. (E) Bubble plot showing marker genes expression in microglial subpopulations. Bubble size (pct.exp) represents the proportion of cells expressing the gene within that subpopulation, and bubble color (avg.exp.scale) indicates the average expression level. (F) UMAP plot of microglial subpopulations. (G) Stacked bar plot showing the cell numbers of each microglial subpopulation in mouse brain tissues under different treatments. (H) Stacked bar plot showing the proportions of each microglial subpopulation in mouse brain tissues under different treatments. (I) Box plot showing differences in the proportions of microglial subpopulations in mouse brain tissues under different treatments (n = 3). Data are presented as proportions. Statistical significance was determined using the chi-square test, with P < 0.05 (), P < 0.01 (), P < 0.001 (),p < 0.0001 (****); ns, not significant. (J) Immunofluorescence images of CLEC7A protein in mouse brain tissues from different treatment groups. CLEC7A + indicate DAM. Green, CLEC7A; red, Iba1; blue, DAPI staining of nuclei.

Together, these single-cell transcriptomic data provide unbiased evidence that RVG-EXOs reshape brain cellular composition, promote a homeostatic microglial state, and modulate key intercellular communication networks. More importantly, integration of these findings with our mechanistic studies supports a coherent model in which RVG-EXOs target neurons and suppress Rptor expression, thereby activating neuronal autophagy. Enhanced autophagy directly facilitates the clearance of neurotoxic proteins, including Aβ and P-Tau, thus alleviating primary neuronal injury. In parallel, autophagy activation accelerates APP degradation and turnover, reducing the levels of APP-related ligands that are either displayed on the neuronal surface or secreted and capable of binding to microglial CD74. Consequently, neuron-to-microglia APP-CD74 signaling is attenuated, which inhibits the transition of microglia toward a pro-inflammatory DAM phenotype and thereby limits secondary neuroinflammatory damage to neurons. Through this dual mechanism—direct elimination of pathological proteins and indirect remodeling of the neuroimmune microenvironment, RVG-EXOs achieve robust neuroprotection in the Alzheimer's disease model.

Biosafety evaluation of RVG-EXOs and EXOs in vivo

To evaluate the long-term safety of RVG-EXOs and EXOs, we administered EXOs or RVG-EXOs to C57 mice via tail vein injection every other day for 40 d. At the end of the treatment, plasma levels of the pro-inflammatory cytokines IL-6, TNF-α, and IL-1β were measured by ELISA. No significant differences were observed among the PBS, EXO, and RVG-EXO groups, indicating that neither EXOs nor RVG-EXOs treatment elicited a systemic inflammatory response (Fig. S10A–C). Additionally, to evaluate potential immune responses, we also measured plasma levels of IgG, which showed no significant difference among the PBS, EXO, and RVG-EXO groups (Fig. S10D), suggesting that repeated administration of EXOs or RVG-EXOs caused no humoral immune response. Furthermore, major organs including the heart, liver, spleen, lung, kidney, and brain were collected for histopathological analysis. H&E staining revealed no apparent tissue damage or inflammatory cell infiltration in any of the organs examined across all treatment groups (Fig. S10E). To further assess potential local inflammation, we performed ELISA to measure the protein expression levels of the inflammatory cytokines IL-6, TNF-α, and IL-1β in tissue lysates from the liver and brain where EXOs and RVG-EXOs were predominantly distributed. Consistent with the histological findings, no significant upregulation of these inflammatory markers was detected in the EXO- or RVG-EXO-treated groups compared to the PBS control group (Fig. S10F–K). Collectively, these results demonstrate that repeated administration of EXOs or RVG-EXOs caused no detectable systemic or organ-specific inflammatory responses, immune activation, or pathological damage, supporting the biosafety of our engineered EXOs for potential therapeutic applications.