P2-engineered exosomes encapsulating curcumin alleviate cognitive decline in AD-like mice by improving microglia-related neuropathology

Abstract

Natural exosomes, as drug carriers, can deliver anti-inflammatory agents across the blood-brain barrier (BBB) to lesion sites in the brain, thereby demonstrating immense potential in the treatment of brain inflammation-related diseases. However, the application of natural exosomes is constrained by their poor targeting ability. Herein, we report a novel drug delivery system (P2-Exo-Cur) constructed by engineering exosomes to display the P2 peptide on their surface, thereby enabling targeted delivery of curcumin to microglia. Our results revealed that P2-Exo-Cur possesses a nanoscale membrane structure and can efficiently deliver curcumin to microglia both in vitro and in vivo. This technology provides a microglia-targeted delivery approach for anti-inflammatory agents such as curcumin, while overcoming the undesirable off-target effects that limit their efficacy. Furthermore, treatment of lipopolysaccharide (LPS)-induced inflammatory BV2 cell models with P2-Exo-Cur significantly suppressed the polarization of BV2 cells toward the M1 phenotype, as well as the secretion of pro-inflammatory cytokines. Finally, we also validated the excellent therapeutic potential of this technology in the 5xFAD mouse model. In conclusion, in this study, we for the first time constructed engineered exosomes that can specifically bind to the NCAM protein on microglia to achieve precise delivery of curcumin by expressing the P2 peptide on their surface, exerting beneficial effects in AD treatment without causing significant adverse effects. This strategy may offer a non-invasive and innovative therapeutic method for the management of brain inflammation-related diseases.

Graphical abstract

1. Introduction

Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by cognitive decline and memory impairment [1]. Accumulating preclinical and clinical evidence indicates that inflammatory and immune responses play a pivotal role in AD pathogenesis, with uncontrolled neuroinflammation serving as a key driver of disease progression and cognitive deterioration [2]. Neuroinflammation is primarily mediated by glial cells—particularly microglia and astrocytes—whose activation is closely associated with immune system dysregulation [3].

In the AD brain, activated microglia are frequently observed in close proximity to amyloid-β (Aβ) plaques [4]. During the early stages of the disease, microglia predominantly adopt an M2 phenotype [5], which exhibits neuroprotective properties through the phagocytic clearance of Aβ and the secretion of anti-inflammatory cytokines such as interleukin-4 (IL-4), interleukin-13 (IL-13), and transforming growth factor-β (TGF-β), along with neurotrophic factors that support neuronal survival [6]. However, persistent Aβ accumulation or chronic neuronal injury leads to excessive microglial activation and a phenotypic shift from the protective M2 state to the pro-inflammatory M1 state [7]. This transition results in the robust release of pro-inflammatory mediators, including interleukin-1β (IL-1β), interleukin-6 (IL-6), and tumor necrosis factor-α (TNF-α) [8]. These cytokines impair microglial phagocytic capacity, thereby promoting further Aβ deposition and exacerbating neuronal damage [9]. Therefore, intervention in overactivated microglia-aggravated neuroinflammation represents a promising therapeutic strategy for AD.

Recent advances have demonstrated the potential of various conventional drugs and natural compounds in modulating microglial activity for AD prevention and treatment [[10], [11], [12]]. Despite the development of agents capable of effectively suppressing pathological inflammation without harming healthy tissues, their delivery to the central nervous system (CNS) remains a major obstacle due to the restrictive nature of the blood-brain barrier (BBB) [13]. Although nanoscale drug delivery systems have shown promise in targeting brain inflammation, challenges such as unintended immune activation and systemic toxicity remain significant barriers to clinical translation [14]. Hence, there is an urgent need for tissue-specific, biocompatible, non-immunogenic delivery platforms that enable precise targeting within the AD brain, underscoring the importance of developing a novel drug delivery system (NDDS) with enhanced specificity and safety.

Naturally occurring nanovesicles, particularly exosomes, have emerged as highly promising candidates for targeted delivery [15]. Exosomes are extracellular vesicles (EVs) ranging in diameter from 30 to 150 nm [16], secreted by various cell types into biological fluids such as blood and cerebrospinal fluid (CSF) [17]. Owing to their intrinsic ability to cross the BBB, coupled with favorable characteristics including biocompatibility, structural stability, low immunogenicity, and surface engineerability [18], exosome-based NDDS offer distinct advantages over synthetic carriers, for instance, lipid nanoparticles [19]. Genetic modification of exosomes to display specific ligands to receptors on target cells enables precise delivery of therapeutic cargo to pathological sites in the brain, ensuring high stability and reduced off-target effects [20].

The P2 peptide (GRILARGEINFK), a 12-amino-acid sequence derived from the second immunoglobulin-like (Ig2) domain of the neural cell adhesion molecule (NCAM), mediates cis-homophilic interactions between NCAM molecules on the same cell membrane [21]. It specifically binds to the Ig1 domain of NCAM, functioning as a selective ligand for NCAM-expressing cells. Notably, NCAM is highly expressed on microglia in the brain [22]. By genetically fusing the P2 peptide with the exosomal membrane protein Lamp2b, the resulting engineered exosomes can be directed specifically to microglia following BBB penetration, enabling cell-type-specific delivery.

In this study, HEK293T cells and hUC-MSCs were transfected with the pTRE-P2-Lamp2b plasmid to generate microglia-targeted exosomes for the delivery of curcumin to microglia. Curcumin, a hydrophobic polyphenolic compound isolated from the rhizomes of Curcuma longa L., possesses potent anti-inflammatory properties and has been widely investigated for its therapeutic potential in inflammatory disorders [23]. Encapsulation of curcumin into exosomes significantly enhances its aqueous solubility, chemical stability, and bioavailability [24]. Therefore, exosome-mediated delivery of curcumin enhances its anti-inflammatory effect. In the 5xFAD mouse model of AD, curcumin-loaded P2-targeted exosomes (P2-Exo-Cur) preserved phagocytic function in microglia, significantly reduced Aβ deposition, simultaneously decreased neuroinflammatory markers, and repaired neuronal damage, thereby improving the cognitive and learning abilities of the mice and highlighting its promising therapeutic potential.

2. Results

2.1. Isolation, characterization and analysis of internalization pathways of exosomes

Initially, high-purity unmodified exosomes (Unmod-Exo) were isolated from the supernatant of HEK293T cells via ultracentrifugation (Fig. 1A). These exosomes were then characterized using transmission electron microscopy (TEM), nanoparticle tracking analysis (NTA), and Western blotting. The results showed that the isolated Unmod-Exo exhibited a distinct membrane boundary and a typical cup- or saucer-shaped morphology, with heterogeneous sizes (Fig. 1B). These exosomes ranged in diameter from 30 to 150 nm (Fig. 1C) and specifically expressed the typical exosomal surface markers CD81, CD9, and TSG101 (Fig. 1D).

Fig. 1.

Extracted exosomes were internalized by cells through the lysosomal pathway. (A) Schematic diagram of exosome isolation. (B) TEM image of Unmod-Exo (scale bar: 50 nm). (C) Size distribution profile of Unmod-Exo. (D) Western blot results of CD9 (25 kDa), CD81 (26 kDa), TSG101 (44 kDa), and GAPDH (37 kDa) expression in Unmod-Exo, with HEK293T cell lysate as the control. (E) Internalization of Unmod-Exo by BV2 cells at 1–6 h captured by high-content imaging system (scale bar: 50 μm). (F) Colocalization of PKH26-labeled exosomes (red) and Lyso-Tracer-stained lysosomes (green) in BV2/HUVEC cells (scale bar: 25 μm). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Internalization of exosomes by microglial cells is a prerequisite for exosome-mediated cargo delivery. Therefore, this study, PKH26-labeled Unmod-Exo were added to BV2 cells, which were then cultured in a high-content imaging system for real-time tracking (Fig. S1A). The results showed that fluorescence signals of Unmod-Exo were detected within BV2 cells, and the signal intensity gradually increased over the 1–6 h of incubation (Fig. 1E and Fig. S1B). These findings indicate that Unmod-Exo can be internalized by BV2 cells, with the intracellular accumulation of Unmod-Exo increasing in a time-dependent manner.

To investigate the pathway underlying exosome internalization, PKH26-labeled Unmod-Exo were co-cultured with BV2 and HUVEC cells for 6 h. The cells were then harvested, and their lysosomes were stained with LysoTracker. Observations under a laser scanning confocal microscope (CLSM) revealed the strong colocalization between the fluorescence signals of Unmod-Exo and lysosomes in both cell types (Fig. 1F). This result demonstrates that the internalization of Unmod-Exo by BV2 and HUVEC cells is mediated through the endocytic pathway.

2.2. Design and characterization of P2-Exo

To modify the surface of exosomes via genetic engineering to obtain P2-Exo with a high targeting ability toward BV2 cells. Initially, the gene sequence of the P2 peptide was linked to the Lamp2b gene sequence to synthesize the P2-Lamp2b fusion gene fragment (Fig. 2A). Subsequently, the fused fragment was cloned into the pTRE-Tight vector to generate the recombinant expression plasmid pTRE-P2-Lamp2b. Upon transfection of this plasmid into host cells, the exosomes secreted by transfected cells were engineered to present the P2 peptide on their surface (Fig. S2A). Following plasmid construction, double enzyme digestion and polymerase chain reaction (PCR) were performed for verification, and subsequent agarose gel electrophoresis imaging confirmed the successful insertion of the P2-Lamp2b fusion fragment into the pTRE-P2-Lamp2b plasmid (Fig. 2B). Additionally, Sanger sequencing and sequence alignment verified that the P2 peptide gene sequence within the plasmid was completely correct (Fig. S2B).

Fig. 2.

Design and characterization of engineered P2-Exo. (A) Schematic representation of the modified Lamp2b protein. SP, signal peptide; TM, transmembrane domain. (B) Electrophoresis validation of the plasmid pTRE-P2-Lamp2b after double enzyme digestion. (C) Quantitative analysis of P2-Lamp2b mRNA expression levels in wild-type and P2 peptide-overexpressing HEK293T cell lines. (D) TEM images of Unmod-Exo and P2-Exo (scale bar: 200 nm). (E) Size distribution profiles of Unmod-Exo and P2-Exo. (F) Western blot results of CD9 (25 kDa), CD81 (26 kDa), TSG101 (44 kDa), Calnexin (90 kDa), and Actin (43 kDa) expression in Unmod-Exo and P2-Exo, with the corresponding HEK293T cell lysate as the control. (G) Surface potential of Unmod-Exo and P2-Exo. (H) Western blot results of Lamp2 (120 kDa) in Unmod-Exo and P2-Exo. (I) Quantitative analysis of Lamp2 protein expression levels in Unmod-Exo and P2-Exo. Data were represented as the mean ± SD, with ∗∗∗p < 0.001.

After confirming the validity of the constructed pTRE-P2-Lamp2b plasmid, it was introduced into HEK293T cells via electroporation. Puromycin was used for cell selection, and doxycycline (DOX) was added to induce gene expression post-selection. Observations under a fluorescence microscope revealed that HEK293T cells successfully expressed red fluorescence upon DOX induction (Fig. S2C).

To verify whether these HEK293T cells could express the target P2-Lamp2b gene, RNA was extracted from a portion of the cells, and quantitative real-time PCR (qRT-PCR) was conducted to detect the mRNA expression level of P2-Lamp2b. The results showed that the expression level of P2-Lamp2b in the P2 peptide-overexpressing HEK293T cell line was 2.75×10⁴-fold higher than that in the wild-type HEK293T cell line (Fig. 2C), confirming the successful establishment of the P2 peptide-overexpressing HEK293T cell line.

Subsequently, to obtain exosomes with their surface modified with the P2 peptide, P2-Exo was isolated from the supernatant of the P2-overexpressing HEK293T cell line after DOX induction. TEM and NTA demonstrated there is no significant differences in morphology or size between the isolated P2-Exo and Unmod-Exo (Fig. 2D and E). Western blotting analysis showed that both Unmod-Exo and P2-Exo positively expressed the canonical exosomal marker proteins CD81, CD9, and TSG101, while they were negative for Calnexin, when compared with the corresponding cell lysates (Fig. 2F). Moreover, both types of exosomes carried a negative zeta potential, with Unmod-Exo at −14.0 ± 0.2 mV and P2-Exo at −14.0 ± 0.2 mV (Fig. 2G). Collectively, these results confirmed the successful isolation of P2-Exo, which shared consistent characteristics with Unmod-Exo and maintained intact structure and components. Notably, Lamp2 antibody (cross-reactive with Lamp2b) was used to detect the pression of the Lamp2b in both types of exosomes. Western blotting results indicated that the expression level of Lamp2 protein in P2-Exo was significantly upregulated compared with that in Unmod-Exo (Fig. 2H), and quantitative analysis revealed an approximate 1.7-fold upregulation (Fig. 2I).

These findings validate the success of the genetic engineering modification strategy and the successful generation of the engineered P2-Exo.

2.3. P2-Exo target microglia in vitro and in vivo

It is generally recognized that the biodistribution and targeting ability of therapeutic agents are crucial for achieving the desired therapeutic effects in vivo. Therefore, the main purpose of genetically engineered P2-Exo is to achieve targeted delivery. To verify the in vitro targeting ability of P2-Exo to BV2 cells, PKH26-stained Unmod-Exo and P2-Exo were added to BV2 cells, which were then cultured in a high-content imaging system to observe the internalization process. Results showed that compared with Unmod-Exo, P2-Exo exhibited a stronger fluorescence signal in BV2 cells at all time points within 1–6 h (Fig. 3A and B and Fig. S3A). These results indicate that P2-Exo have excellent in vitro targeting ability to BV2 cells and are more easily internalized by these cells.

Fig. 3.

P2-Exo exhibited enhanced accumulation and prolonged retention in vitro and in vivo. (A) Internalization of Unmod-Exo and P2-Exo by BV2 cells at 1–6 h captured by high-content imaging system (scale bar: 50 μm). (B) Quantitative analysis of fluorescence signals for the internalization of Unmod-Exo and P2-Exo in BV2 cells at 1–6 h. (C) Fluorescence imaging of Unmod-Exo and P2-Exo in 5xFAD model mice after tail vein injection by small animal in vivo imaging system. (D, E) Fluorescence imaging and quantitative analysis of Unmod-Exo and P2-Exo in the brains of 5xFAD model mice. Data were represented as the mean ± SD, with ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 and ns: not significant.

The crossing of the BBB is one of the most common and critical challenges in the development of therapeutic drugs for brain diseases. To confirm the in vivo targeting ability of P2-Exo to BV2 cells, DIR-stained Unmod-Exo and P2-Exo were injected into 5xFAD mice via the tail vein, followed by in vivo imaging at multiple time points using a small animal in vivo imaging system. The results demonstrated that compared with Unmod-Exo, P2-Exo showed a significantly stronger fluorescence signal in the brain, which gradually accumulated over time (Fig. 3C). After sacrificing the mice and obtaining brains for imaging, a significantly higher fluorescence signal of P2-Exo was also observed in the brain compared with that of Unmod-Exo (Fig. 3D). Quantitative analysis of the fluorescence signals in the brain revealed that the fluorescence intensity of P2-Exo was significantly higher than that of Unmod-Exo (Fig. 3E). Meanwhile, the same results were obtained in targeting experiments with normal mice (Fig. S3B–D). These results confirm that P2-Exo can effectively penetrate the BBB, exert its targeting ability toward BV2 cells in vivo, and accumulate more in the mouse brain, providing favorable evidence for subsequent animal model experiments.

2.4. Microglia-targeting of P2-Exo is mediated by NCAM

Currently, the specific binding of ligands to their receptors is the most commonly used and proven method for targeting specific cells in the central nervous system. Therefore, the NCAM protein was selected as the target receptor to achieve the targeted delivery of exosomes to microglia in the study. To investigate whether the specific binding target of P2-Exo is the NCAM protein, BV2 cells were incubated with P2-Exo, followed by fluorescent labeling of NCAM protein and Flag tag, which is theoretically co-expressed with the P2 peptide on the exosomal surface. Antibodies of anti-Flag tag and anti-NCAM were used to detect the colocalization between the P2 peptide (represented by the Flag tag) and NCAM protein (Fig. 4A). The results revealed a strong colocalization signal between the Flag-tagged protein and NCAM protein in the cells (Fig. 4B). This finding suggests that the P2 peptide on P2-Exo may interact with NCAM protein in BV2 cells, providing the preliminary evidence that NCAM is the target site of P2-Exo.

Fig. 4.

NCAM protein positively regulated the targeting ability of P2-exosomes to microglia. (A) Colocalization of NCAM protein (red) and Flag-tagged P2-Lamp2b protein (green) in BV2 cells under fluorescence microscopy (scale bar: 10 μm). (B) Colocalization curve of NCAM protein and Flag-tagged P2-Lamp2b protein in BV2 cells. (C) Uptake of Unmod-Exo and P2-Exo by WT BV2, sh-NCAM BV2, and HUVEC cells captured by CLSM (scale bar: 50 μm). (D) Quantitative analysis of fluorescence signals for the uptake of Unmod-Exo and P2-Exo by WT BV2, sh-NCAM BV2, and HUVEC cells. (E) Uptake of P2-Exo by WT BV2, sh-NCAM BV2, and HUVEC cells at 1–6 h captured by high-content imaging system (scale bar: 50 μm). (F) Quantitative analysis of fluorescence signals for the uptake of P2-Exo by WT BV2, sh-NCAM BV2, and HUVEC cells at 1–6 h. (G) Uptake of P2-Exo BV2 cells with different treatment at different time points captured by high-content imaging system (scale bar: 50 μm). (H) Uptake of P2-Exo BV2 cells with different treatment at 180 min captured by Laser confocal microscopy (scale bar: 50 μm). (I) Schematic diagram of the specific binding of P2 peptide and NCAM. Data were represented as the mean ± SD, with ∗∗∗p < 0.001, ns: not significant. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

To confirm that NCAM is indeed the target site of P2-Exo, an NCAM-knockdown BV2 cell line (sh-NCAM BV2) is established first and Western blotting analysis was performed to detect NCAM protein expression in this cell line. The results showed a significant reduction of NCAM in sh-NCAM BV2 compared with wild-type BV2 cells (WT BV2) (Fig. S4A). Quantitative analysis indicated that NCAM protein expression in sh-NCAM BV2 cells was downregulated by approximately 78% relative to that in WT BV2 cells (Fig. S4B), confirming the successful establishment of the NCAM-knockdown BV2 cell line. Furthermore, the results of immunofluorescence (IF) staining also confirmed that NCAM had been successfully knocked down (Fig. S4C).

Subsequently, PKH26-labeled Unmod-Exo and P2-Exo were added to sh-NCAM BV2 cells, and their uptake was observed under CLSM after 2 h. The results showed no significant difference in fluorescence signals between P2-Exo and Unmod-Exo in sh-NCAM BV2 cells (Fig. S4D), indicating the loss of microglial targeting ability of P2-Exo. This result also confirms that NCAM protein is the target site of P2-Exo.

To further clarify the effect of NCAM expression levels on the uptake of P2-Exo, the expression of NCAM protein was detected in five different cell lines, BV2, PC12, NHA, C2C12, and HUVEC cells, via Western blotting. The results showed that BV2 cells exhibited the highest NCAM expression, whereas NCAM was barely detectable in HUVEC cells (Fig. S4E and F). Subsequently, wild-type (WT) BV2 cells (with the highest NCAM expression), sh-NCAM BV2 cells (with NCAM knockdown), and HUVEC cells (serving as a negative control) were selected for the exosome uptake assay across multiple cell types. WT BV2, sh-NCAM BV2, and HUVEC cells were separately incubated with PKH26-labeled Unmod-Exo and P2-Exo. After 2 h of incubation, observations under the CLSM revealed no significant difference in the uptake of Unmod-Exo among the three cell lines. In contrast, the fluorescence signal of P2-Exo in WT BV2 cells was significantly stronger than that in sh-NCAM BV2 and HUVEC cells (Fig. 4C and D). These findings indicate that P2-Exo exert its targeting effect only in cells with high NCAM expression.

In addition, a high-content imaging system was used for real-time tracking of P2-Exo uptake by the three cell lines at various time points within 1–6 h. The results showed that at each time point during this period, the fluorescence signal of P2-Exo in WT BV2 cells was consistently stronger than that in sh-NCAM BV2 and HUVEC cells (Fig. 4E). High-throughput quantitative analysis of the fluorescence signals was further performed. The results demonstrated that the fluorescence intensity of P2-Exo gradually increased over 1–6 h in all three cell lines. At every time point, the fluorescence signal of P2-Exo was the strongest in WT BV2 cells, followed by sh-NCAM BV2 cells, and the weakest in HUVEC cells. This trend was positively correlated with the NCAM expression levels of the three cell lines (Fig. 4F). Therefore, these results clearly confirm that the NCAM expression level specifically regulates the uptake of P2-Exo, the higher NCAM expression correlating with enhanced cellular uptake capacity for P2-Exo.

Finally, we verified the direct interaction between P2 peptide and NCAM via competitive binding assays. BV2 cells were pre-incubated with 5 μM or 10 μM free P2 peptide for 30 min, followed by treatment with P2-Exo to assess the targeting capacity of the P2-Exo. Time-lapse imaging using high-content imaging system (Fig. 4G) and confocal microscopy (Fig. 4H) revealed that free P2 peptide markedly suppressed P2-Exo endocytosis in BV2 cells. Collectively, these results indicate that exogenous P2 peptide competes with P2-Exo for specific binding to NCAM, thereby abrogating the targeting function of P2-Exo. In conclusion, all the results collectively demonstrate that P2-Exo can exert a targeted effect by binding to the membrane protein NCAM on microglia through the P2 peptide (Fig. 4I).

2.5. Exosomes ameliorate the limitations of curcumin

Numerous studies have documented the robust anti-inflammatory properties of curcumin. We first determined the maximum tolerable dose of curcumin in BV2 microglial cells (Fig. 5A) using the Cell Counting Kit-8 (CCK-8) assay, identifying 10 μM as the highest concentration that could maintain >90% cell viability. Subsequently, an LPS-induced inflammatory BV2 cell model was treated with 10 μM curcumin. Morphological assessment across three randomly selected microscopic fields revealed that curcumin markedly attenuated the LPS-triggered transition of BV2 cells to an amoeboid (activated) morphology (Fig. 5B and C). Furthermore, enzyme-linked immunosorbent assay (ELISA) was used to quantify the secretion of key pro-inflammatory cytokines–tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6), and interleukin-1β (IL-1β)–across treatment groups. Curcumin treatment significantly suppressed the LPS-induced upregulation of all three cytokines (Fig. 5D–F). Collectively, these findings corroborate prior evidence indicating that curcumin inhibits M1 polarization of BV2 microglia and potently suppresses inflammatory cytokines release.

Fig. 5.

Exosomes improved the stability and solubility of curcumin. (A) Cell viability of BV2 cells with different treatment was assessed using the CCK-8 assay. (B) Morphological images of BV2 cells with different treatment (scale bar: 20 μm). (C) Quantitative analysis of the number of amoeboid-like cells in BV2 cells with different treatment. (D) Quantitative ELISA analysis of TNF-α secretion in BV2 cells with different treatment. (E) Quantitative ELISA analysis of IL-1β secretion in BV2 cells with different treatment. (F) Quantitative ELISA analysis of IL-6 secretion in BV2 cells with different treatment. (G) Schematic diagram of the loading and purification process of curcumin into exosomes via co-incubation. (H) Solubility of curcumin with different concentrations in 1 mL PBS. (I) Quantitative analysis of curcumin concentration in the supernatant after incubation of free curcumin and curcumin-exosome mixture on ice for 30 min followed by centrifugation. (J) Quantitative analysis of the remaining curcumin in free curcumin and curcumin-exosome mixture in PBS over 0–150 min. Data were represented as the mean ± SD, with ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ns: not significant.

However, curcumin's clinical translation is hindered by its inherently low aqueous solubility and poor physicochemical stability, which collectively limit its bioavailability and therapeutic efficacy. To address these limitations, curcumin was encapsulated into exosomes via passive loading through co-incubation (Fig. 5G). Prior to loading, curcumin's maximal solubility was empirically determined in 1 mL of phosphate-buffered saline (PBS) by incremental addition; and saturation occurred at 100 μM (Fig. 5H).

Next, whether exosomal encapsulation could enhance curcumin's solubility and stability or not was evaluated under physiological buffer conditions. Equal molar amounts of free curcumin or curcumin-loaded exosomes were separately suspended in PBS and incubated on ice for 30 min. After centrifugation, the concentration of curcumin in the supernatant was measured. The results showed that the concentration of curcumin in the supernatant of the curcumin-loaded exosome group was threefold higher than that of the alone group (Fig. 5I), suggesting enhanced dispersion and reduced aggregation of curcumin were mediated by exosomal association. Stability was further assessed by monitoring curcumin degradation over time in PBS at 4 °C. Free curcumin degraded rapidly, with only ∼30% remaining after 150 min relative to the t = 0 baseline (normalized to 1.0), whereas exosome-encapsulated curcumin retained >60% of its initial concentration under identical conditions (Fig. 5J). These data demonstrate that exosomal delivery substantially improves both the aqueous solubility and colloidal stability of curcumin—the key prerequisites for enhancing its pharmacological performance.

2.6. P2-Exo enhance the therapeutic effects of curcumin against inflammation in vitro

The preparation of Unmod-Exo-Cur and P2-Exo-Cur included the following steps: an appropriate amount of curcumin was first mixed with an equal amount of Unmod-Exo and P2-Exo in 1 mL of PBS, followed by co-incubation at room temperature for 15 min to load curcumin. After completion of the loading, ultracentrifugation was performed for purification, and the amount of remaining curcumin in PBS was determined. The encapsulation rate was calculated using the formula, where Unmod-Exo: 42.25%, P2-Exo: 36.58% and quantitative analysis revealed that 1 μg of Unmod-Exo contained 0.422 μg curcumin, while 1 μg of P2-Exo had a curcumin content of 0.365 μg (Fig. 6A).

Fig. 6.

P2-Exo-Cur suppressed the inflammatory response of microglia in vitro. (A) Curcumin loading profile and encapsulation efficiency of Unmod-Exo and P2-Exo. (B) TEM images of Unmod-Exo-Cur and P2-Exo-Cur (scale bar: 200 nm). (C) Morphological images of BV2 cells with different treatment (scale bar: 10 μm). (D) Quantitative analysis of the number of amoeboid-like cells in BV2 cells with different treatment. (E) iNOS expression in BV2 cells with different treatment were observed under fluorescence microscopy (scale bar: 50 μm). (F) Quantitative analysis of iNOS expression in BV2 cells with different treatment. (G) Quantitative ELISA analysis of TNF-α secretion in BV2 cells with different treatment. (H) Quantitative ELISA analysis of IL-1β secretion in BV2 cells with different treatment. (I) Quantitative ELISA analysis of IL-6 secretion in BV2 cells with different treatment. Data were represented as the mean ± SD, with ∗∗p < 0.01, ∗∗∗p < 0.001, and ns: not significant.

To investigate whether curcumin encapsulation alters the inherent properties of exosomes, the morphology, particle size, and zeta potential of Unmod-Exo-Cur and P2-Exo-Cur were characterized. Similar to their unloaded counterparts, both Unmod-Exo-Cur and P2-Exo-Cur exhibited a bilayer membrane structure with a typical cup-shaped morphology (Fig. 6B). However, compared with Unmod-Exo and P2-Exo, the particle sizes of the curcumin-loaded exosomes were slightly increased (Fig. S5A), and their zeta potentials were moderately reduced, with values of −21.7 ± 0.6 mV for Unmod-Exo-Cur and −22.5 ± 0.6 mV for P2-Exo-Cur (Fig. S5B). Collectively, these findings indicate that curcumin encapsulation leads to minor changes of particle size, and zeta potential of exosomes without impairing their intrinsic properties.

To evaluate whether Unmod-Exo and P2-Exo could enhance the inhibitory effect of curcumin on the M1 polarization of BV2 cells, LPS-induced inflammatory BV2 cell models were treated with free curcumin, Unmod-Exo-Cur, or P2-Exo-Cur. Morphological observations were conducted across three visual fields for cells in each treatment group. The results revealed that compared with free Cur, both Unmod-Exo-Cur and P2-Exo-Cur more effectively reduced the number of amoeboid-shaped cells induced by LPS, with P2-Exo-Cur exhibiting the most prominent efficacy (Fig. 6C and D). Inducible nitric oxide synthase (iNOS) which is significantly upregulated during an inflammatory response was further detected in BV2 cells using immunofluorescence staining. The results showed that the iNOS expression levels were significantly lowered in the Unmod-Exo-Cur and P2-Exo-Cur groups than in the free Cur group, and the reduction was most pronounced in the P2-Exo-Cur group (Fig. 6E and F). These findings demonstrate that both Unmod-Exo and P2-Exo can enhance curcumin's ability to inhibit the M1 polarization of BV2 cells, and P2-Exo-Cur achieves superior efficacy due to its specific targeting capability.

Furthermore, ELISA was used to detect the secretion of inflammatory cytokines in the supernatant of BV2 cells with different treatment. The results showed that the secretion levels of TNF-α, IL-1β, and IL-6 were significantly lowered in the Unmod-Exo-Cur and P2-Exo-Cur groups compared with the free Curcumin group, with the lowest levels observed in the P2-Exo-Cur group (Fig. 6G–I). Consistently, these results confirm that encapsulating curcumin into Unmod-Exo or P2-Exo strengthens its ability to suppress the secretion of inflammatory cytokines by BV2 cells, and P2-Exo-Cur exerts a most remarkable effect owing to its targeting capability.

2.7. P2-Exo-Cur effectively alleviates Aβ deposition and neural damage, thereby improving cognitive behavior in a AD mouse model

5xFAD transgenic mice were used to evaluate the in vivo therapeutic efficacy of P2-Exo-Cur. Curcumin alone (Cur), exosomes alone (Exo), Unmod-Exo-Cur, and P2-Exo-Cur were administered via tail vein injection twice a week for 4 months. Subsequently, the Morris water maze (MWM) test and Y-maze test were performed to assess their spatial learning and memory abilities. Firstly, training and formal tests were conducted using a water maze video tracking and analysis system. After 3 days of training, navigation tests and probe trials were performed under blind conditions on day 4 and day 5, respectively (Fig. 7A). The results showed that compared with Model mice, mice treated with P2-Exo-Cur exhibited a faster learning rate and a shorter escape latency during the learning process (Fig. 7C–E). Subsequently, tests were performed using the ANY-maze animal tracking software to detect the platform. The three arms of the Y-maze were randomly designated as A, B, and C. Mice were placed at the end of one of the arms and allowed to explore freely for 5 min, and the spontaneous alternation rate was calculated (Fig. 7B). The results indicated that 5xFAD mice treated with P2-Exo-Cur had a significantly higher spontaneous alternation rate (Fig. 7F). Collectively, these results suggest that spatial memory impairment and behavioral deficits in 5xFAD mice were effectively alleviated after P2-Exo-Cur treatment.

Fig. 7.

P2-Exo-Cur reduced Aβ deposition, and enhanced neuronal activity, thereby ameliorating cognitive impairment in 5xFAD mice. (A) Representative trajectories of differently treated mice in MWM. (B) Representative trajectories of differently treated mice in Y-maze. (C) Number of crossing the location of the escape platform. (D) Escape latency on the test day. (E) Duration of swimming in the target quadrant. (F) Rate of spontaneous alternation between quadrants. (G) Immunohistochemistry images of Aβ plaques (scale bar: 300 μm). (H) Nissl staining images of neurons (scale bar: 300 μm). Data were represented as the mean ± SD, with ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ns, not significant.

After behavioral assessment, the mice were euthanized, and their brain tissues were collected to detect the multi-pathway intervention ability of P2-Exo-Cur on the pathology of AD mice. First, amyloid-β (Aβ) deposition in the brains of AD mice was determined by Aβ staining. The results showed that the number of Aβ plaques was significantly higher in the Model group, whereas P2-Exo-Cur treatment significantly reduced the number of Aβ plaques (Fig. 7G); this finding indicates that P2-Exo-Cur can effectively attenuate Aβ deposition, thereby protecting neurons.

To further confirm the protective effect of P2-Exo-Cur on neurons in AD model mice, Nissl staining was performed to detect neuronal damage in 5xFAD mice of each group. The results showed that Nissl body staining was shallow, with a reduced number or even disappearance of Nissl bodies in the Model group. In contrast, after P2-Exo-Cur treatment, the staining of Nissl bodies was deepened, and the number of Nissl bodies was increased (Fig. 7H). This finding suggests that P2-Exo-Cur may effectively protect neurons from damage caused by Aβ aggregates and inflammatory factors, maintain the normal function of nerve cells, and thus effectively improve cognitive impairment in 5xFAD mice.

2.8. P2-Exo-Cur inhibited inflammatory responses in 5xFAD mice

Finally, as we have previously mentioned, exacerbated neuroinflammation caused by aberrant microglial activation has long been recognized as a hallmark of AD pathology and progression. Therefore, the expression of IBA-1 and GFAP that were significantly increased in the hippocampus of the AD model group, indicating enhanced neuroinflammation, were evaluated in 5xFAD mice after different treatments using IF and immunohistochemistry (IHC) techniques. The results showed that GFAP and Iba-1 staining was intense with a high number of positive cells in the Model group, whereas P2-Exo-Cur treatment significantly reduced both the staining intensity and the number of positive cells (Fig. 8A–D). To further verify the anti-inflammatory effect of P2-Exo-Cur, the secretion of inflammatory cytokines in the blood of mice with different treatment was detected using ELISA. The results demonstrated that compared with the Model group, the secretion levels of the TNF-α, IL-1β, and IL-6 in the blood of the P2-Exo-Cur group were significantly decreased (Fig. 8E–G). Collectively, these results indicate that P2-Exo-Cur can prevent the polarization of BV2 cells toward the M1 phenotype and significantly inhibit the secretion of inflammatory cytokines, thereby reducing the inflammatory level in the brains and blood of 5xFAD mice.

Fig. 8.

P2-Exo-Cur effectively inhibits the polarization of glial cells and alleviates inflammation in AD mice. Immunofluorescence images of (A) lba-1, (B) GFAP in hippocampus (scale bar: 300 μm). Immunohistochemistry images of (C) lba-1, (D) GFAP in hippocampus (scale bar: 300 μm). (E) Quantitative analysis of TNF-α secretion in the blood of 5xFAD mice with different treatment. (F) Quantitative analysis of IL-1β secretion in the blood of 5xFAD mice with different treatment. (G) Quantitative analysis of IL-6 secretion in the blood of 5xFAD mice with different treatment. Data were represented as the mean ± SD, with ∗∗p < 0.01, ∗∗∗p < 0.001, and ns, not significant.

In addition, the biological safety of P2-Exo-Cur were also evaluated in 5xFAD mice. After treatment, no pathological changes were observed in the main organs of all mice through H&E staining (Fig. S6), indicating that P2-Exo-Cur have excellent biological safety for in vivo application and is a reliable targeted treatment system for AD.

3. Discussion

Over the past decades, clinical research on AD therapy has predominantly focused on reducing Aβ plaques, utilizing strategies such as anti-Aβ antibodies [25], β-site amyloid precursor protein cleaving enzyme (BACE) inhibitors [26], and γ-secretase inhibitors [27,28]. However, these approaches have proven insufficient in fully halting the pathological progression of AD [29]. Recent studies have demonstrated that robust inflammatory responses were triggered by the misfolding and aggregation of proteins in the brains of AD patients, which exacerbate cognitive decline and neurodegeneration [30]. Consequently, suppressing neuroinflammation has emerged as a critical therapeutic strategy for mitigating cognitive deficits associated with neurodegenerative disorders [31]. This study adopts the alleviation of neuroinflammation as a central approach to interrupt the pathological cascade of AD and ameliorate cognitive impairment in 5xFAD mouse models.

Neuroinflammation is typically accompanied by activation of the immune system, primarily mediated by glial cells—particularly microglia [32]. Aberrant microglial activation is now widely recognized as a key hallmark of AD pathology and disease progression [33]. Among non-neuronal CNS cells, microglia play a pivotal role in maintaining brain metabolic homeostasis, including the secretion of pro-inflammatory and anti-inflammatory cytokines and the phagocytic clearance of pathological protein aggregates [34,35]. Accumulating evidence indicates that promoting the phenotypic shift of microglia from the pro-inflammatory M1 state to the anti-inflammatory M2 phenotype can effectively alleviate neuroinflammation and reduce Aβ-related neuropathology in AD [36]. Therefore, this study aims to target microglia and restore their functional homeostasis to mitigate neuroinflammatory processes.

Although various anti-inflammatory agents, such as curcumin, have shown great potential in modulating microglial inflammatory responses—consistent with our experimental findings—their clinical utility in treating brain inflammation is severely limited by the BBB [37]. The inability of many promising therapeutics to efficiently cross the BBB and reach affected brain regions remains a major obstacle in the development of CNS-targeted treatments, hindering the clinical translation of numerous drug candidates [38]. Thus, a primary objective of this study was to identify a safe and effective delivery vehicle capable of transporting curcumin across the BBB [39]. While conventional synthetic delivery systems—including viral vectors, lipid nanoparticles, and polymeric nanoparticles—have achieved partial success, they are often associated with drawbacks such as immunogenicity and cytotoxicity due to their exogenous nature [40]. Growing evidence supports the use of exosomes—the endogenous nanoscale vesicles secreted by various cells—as a superior alternative, offering enhanced biocompatibility and reduced adverse effects while enabling efficient drug delivery across the BBB [41]. Given their favorable safety profile, exosomes represent a promising platform for brain-targeted delivery [42]. Notably, several recent studies have demonstrated promising outcomes by employing extracellular vesicles for the treatment of neuroinflammatory brain disorders [43,44]. It forms a key rationale for their application in targeting microglia in this study. Our results confirm that exosome-mediated delivery enhances the bioavailability of curcumin without causing significant toxicity.

Beyond overcoming BBB penetration and ensuring safety, achieving optimal biodistribution and precise cellular targeting is essential for maximizing therapeutic efficacy in clinical applications [45]. In the context of CNS disorders like AD, minimizing off-target effects is crucial to avoid unintended adverse reactions [46]. To address this challenge, carrier systems are commonly engineered with specific ligands to enhance their affinity for target cells—a well-established strategy in targeted drug delivery [47]. For exosomes, genetic engineering or chemical modification techniques are currently employed to improve organ- or cell-specific targeting [48]. In this study, we developed a novel exosome-based delivery system designed to precisely deliver anti-inflammatory therapeutics to microglia. This system aims to promote microglial phenotypic transition, remodel the microglia-centered inflammatory microenvironment, restore phagocytic function, reduce excessive Aβ accumulation, and ultimately protect neurons from degeneration.

In conclusion, this study presents an engineered exosome delivery platform capable of efficiently transporting curcumin across the BBB to target microglia, enabling sustained drug release at lesion sites and significantly enhancing both the bioavailability and therapeutic effectiveness of curcumin. To achieve targeted delivery, we generated a fusion protein consisting of the P2 peptide and the exosomal membrane protein Lamp2b, thereby displaying the P2 peptide on the exosome surface. The P2 peptide is a NCAM-mimetic sequence that specifically binds to NCAM, a glycoprotein belonging to the immunoglobulin superfamily [49]. The major NCAM isoforms—NCAM 120, NCAM 140, and NCAM 180—are differentially expressed across cell types, with NCAM 120 and NCAM 140 being highly expressed in microglia [50]. Additionally, P2 peptide can penetrate BBB after systemic administration [51]. Hence, P2-decoration can help exosomes for selective microglial targeting. Experimental data confirmed that surface functionalization with the P2 peptide significantly enhances exosome accumulation in microglia following BBB traversal. Compared to Unmod-Exo-Cur, P2-Exo-Cur exhibited markedly increased brain retention, underscoring the importance of P2 peptide modification in improving delivery efficiency. In vivo therapeutic evaluations revealed that P2-Exo-Cur effectively improved cognition and learning capacity in 5xFAD mice by inducing a beneficial shift of microglia polarization, suppressing neuroinflammation, reducing Aβ deposition and protecting neurons.

Although our findings demonstrate that P2 peptide-modified exosomes can function as a brain-targeted drug delivery platform that markedly ameliorates the pathological manifestations of AD in animal models, and that this delivery system holds potential for application in other neuroinflammation-related brain disorders, challenges still remain in translating anti-inflammatory strategies into effective therapies for cognitive impairment associated with neurological diseases [52]. Further investigation is needed to elucidate the molecular mechanisms underlying microglia-associated inflammatory responses [53]. Additionally, the in vivo transport pathways and pharmacokinetic behavior of exosomes require more comprehensive characterization [54]. A deeper understanding of how exosomes distribute within the brain and are subsequently cleared will provide critical insights necessary for advancing exosome-based delivery systems from preclinical models to human clinical applications.

4. Conclusion

In summary, we developed a microglia-targeted drug delivery system designed to ameliorate microglial dysfunction by reversing the functional impairment of abnormally activated microglia in AD. The engineered P2-Exo-Cur efficiently crosses the BBB and accumulates within microglia through specific interaction mediated by the NCAM-mimetic P2 peptide, enabling precise delivery of curcumin to pathological sites in the brain. This targeted delivery leads to a significant suppression of microglial polarization toward the pro-inflammatory M1 phenotype and promotes a phenotypic shift toward the anti-inflammatory M2 state. As a result, neuroinflammatory responses are attenuated, and the phagocytic capacity of microglia for Aβ is enhanced. Moreover, the modulation of the microglia-driven neuroinflammatory microenvironment confers neuroprotective effects and supports neurogenesis, ultimately ameliorating spatial learning and memory deficits in 5xFAD mice. This study presents for the first time a safe and efficient microglia-targeting exosome delivery platform, highlights the potential of precision drug delivery in the treatment of AD, and analyzes its prospects for clinical translation.

5. Method

5.1. Cell culture and animal

5.1.1. Cell culture

The murine microglial cell line BV2 was obtained from Changsha Fenghui Biotechnology Co., Ltd. The human embryonic kidney cell line HEK293T, rat pheochromocytoma-derived cell line PC12, murine myoblast cell line C2C12, human astrocyte cell line NHA, human umbilical vein endothelial cell line HUVEC, and hUC-MSCs were provided by the Laboratory Center for Tissue Engineering and Stem Cells at Guizhou Medical University. HEK293T, BV2, HUVEC, NHA, and C2C12 cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and antibiotics (penicillin and streptomycin). PC12 cells were maintained in Roswell Park Memorial Institute (RPMI)-1640 medium containing 10% FBS and antibiotics, whereas hUC-MSCs were cultured in DMEM/F12 medium supplemented with 10% FBS and antibiotics. All cells were maintained in 5% CO₂ at 37 °C. The exosomes used in all cell experimental treatments were HEK293T cell-derived exosomes.

5.1.2. Animals

Different ages male C57BL/6J mice (SPF grade) and eight-week-old male 5xFAD mice were obtained from Shulaibao (Wuhan) Biotechnology Co., Ltd. The animals were housed under standard conditions (12/12-h light/dark cycle, 60 ± 5% humidity, 22 ± 3 °C temperature) with ad libitum access to food and water. All animal procedures were approved by the Institutional Animal Care and Use Committee of Guizhou Medical University Animal Experiment Center, and the ethical approval number for animal experiments is 2200999. The exosomes used in all animal experimental treatments were hUC-MSCs cell-derived exosomes.

5.2. Construction of P2 plasmid and transfection

The P2-Lamp2b fusion gene was designed by our team and subsequently submitted to Shanghai GenePharma Co., Ltd. for synthesis. Subsequently, the P2-Lamp2b fragment was digested with SfaAI and NotI restriction enzymes and then ligated into the pTRE-Tight plasmid backbone to construct the pTRE-P2-Lamp2b fusion expression plasmid.

This fusion expression plasmid was transfected into HEK293T and hUC-MSCs via electroporation. Puromycin was used for selection to obtain new cell lines stably expressing the P2-Lamp2b fusion protein.

5.3. Isolation, purification, characterization and labelling of exosomes

Ultracentrifugation is the most frequently employed method for the isolation of exosomes. For this purpose, the FBS utilized in the present study was ultracentrifuged overnight at 120,000 ×g to deplete endogenous exosomes. After 48 h of cell culture, the exosome-containing cell supernatant was collected. Exosomes were then purified from the culture supernatant via ultracentrifugation using the following protocol: the harvested supernatant was first centrifuged at 300 ×g for 10 min to remove dead cells, followed by centrifugation at 2000 ×g for 10 min to eliminate cellular debris, then at 10,000 ×g for 30 min to discard large vesicles, and finally at 100,000 ×g for 70 min to pellet exosomes. Subsequently, the isolated exosomes were resuspended in cold PBS, filtered through a 0.22 μm filter, and subjected to another round of ultracentrifugation at 100,000 ×g for 70 min. The pelleted exosomes were then resuspended in cold PBS and stored at −80 °C for subsequent experiments. The concentration of exosomes was quantified using a bicinchoninic acid (BCA) protein assay kit. Western blotting was performed to analyze the exosomal marker proteins (CD9, CD81, and TSG101) as well as the membrane protein Lamp2. For TEM analysis, 20 μL exosomes were dropped onto a sample plate. The copper grid was placed face down onto the droplet surface and incubated for 3 min before being removed and uniform standard line at room temperature. The grid was then floated on 2% Phosphotungstic acid for 1–2 min for staining. After removal, excess staining solution was blotted with filter paper. Images were captured using a JEM-1400FLASH transmission electron microscope (JEOL, Japan). NTA was employed to determine the size distribution and zeta potential of the exosomes. Exosomes were labeled with PKH26 and DiR fluorescent dyes. Purified exosomes were incubated with the fluorescent dyes at 37 °C for 20 min in the dark. The staining reaction was terminated by adding FBS, and the labeled exosomes were re-pelleted by ultracentrifugation at 100,000 ×g for 90 min to remove free unbound dyes.

5.4. Therapeutic cargo-loading into exosomes

An equal amount of curcumin and exosomes were mixed in PBS and incubated for 15 min at room temperature in the dark. The exosomes were then washed with cold PBS and subjected to ultracentrifugation at 120,000 ×g for 70 min to remove unencapsulated free curcumin. Subsequently, the pellet was collected, washed, and resuspended in PBS to obtain exosome-encapsulated curcumin (Exo-Cur). The concentration of curcumin encapsulated in exosomes was quantified by detecting its intrinsic ultraviolet (UV) absorption using a UV-visible spectrophotometer (Lambda25, PerkinElmer Medical Diagnostics Products Shanghai Co., Ltd., USA). The detection was performed at an excitation wavelength of 450 nm, with the absorbance measured at 594 nm. The percentage of loading efficiency was calculated according to the following equation: % loading efficiency = (curcumin in exosomes/total curcumin added) × 100.

5.5. Real-time PCR

Total RNA was extracted from the selected stable cell lines using TRIzol reagent (Thermo Fisher Scientific, USA). Complementary DNA (cDNA) was synthesized following the manufacturer's instructions with the First-Strand cDNA Synthesis Kit (Thermo Fisher Scientific, USA). qRT-PCR was performed on an Applied Biosystems 7500 Real-Time PCR System (Applied Biosystems, USA), and the mRNA expression levels were quantified using a qPCR Mix Kit (Sangon Biotech, China). GAPDH was used as the internal reference to the uniform standard line. The primer sequences used in this study are as follows: P2-Lamp2b forward primer, 5′-GAAATGGCAATACTACCT-3′ and P2-Lamp2b reverse primer, 5′-ACAAGCGACCTACTAC-3′; GAPDH forward primer, 5′-GTGGCAAAGTGGAGATTGTTG-3′ and GAPDH reverse primer, 5′-CGTTGAATTTGCCGTGAGTG-3′.

5.6. Western blotting

Total proteins were extracted from cells or exosomes using radioimmunoprecipitation assay (RIPA) lysis buffer supplemented with a protease inhibitor cocktail (TransGen, China). The protein concentration was quantified using the BCA protein assay kit, and the protein was denatured in boiling water for 10 min. Subsequently, 20 μg of protein samples were loaded and separated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), then transferred onto 0.45 μm polyvinylidene fluoride (PVDF) membranes (Millipore, USA). The membranes were then blocked with 5% non-fat milk at room temperature for 1 h. The membranes were incubated overnight at 4 °C with primary antibodies against CD9, CD81, TSG101, NCAM, Calnexin, and β-actin. After washing three times with Tris-buffered saline with Tween 20 (TBST), the membranes were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies at room temperature for 2 h. Protein bands were visualized using the ECL kit (Vazyme Biotech, China) and protein imaging system (Bio-Rad Laboratories, USA).

5.7. ELISA

TNF-α, IL-1β and IL-6 levels were measured in various biological samples by ELISA (4A Biotech, Suzhou, China) according to the manufacturer's protocol.

5.8. Immunofluorescence staining

After cell induction treatment, 4% PFA was fixed for 30 min, washed 3 times with PBS. 0.1% Triton was incubated for 15 min. Brain tissue sections were dried at room temperature for 30 min and rinsed three times with PBS. The sections were then incubated with 0.3% Triton X-100 for 30 min, followed by three additional washes with PBS. After blocking with 5% bovine serum albumin (BSA) for 2 h, the sections were incubated with primary antibodies (anti-NCAM, 1:200; anti-Flag, 1:500; anti-iNOS, 1:200) at 4 °C for 16 h. The sections were washed three times with PBS before being incubated with secondary antibodies at room temperature for 2 h. Following another three washes with PBS, the sections were incubated with 4′,6-diamidino-2-phenylindole (DAPI) for 15 min. Images were captured using CLSM, and the fluorescence intensity of each group was analyzed with ImageJ software.

5.9. In vivo imaging

DiR-labeled engineered exosomes were intravenously injected into C57BL/6 mice and 5xFAD mice via the tail vein. A small animal in vivo imaging system was used to capture images and observe the distribution patterns of the red fluorescent signals from Unmod-Exo and P2-Exo in the mice at 1, 3, 6, 12, 24 and 48 h post-injection. Subsequently, the mice were euthanized and dissected, and their brains were harvested for subsequent in vivo imaging analysis. The distribution of the red fluorescent signals from Unmod-Exo and P2-Exo in the isolated brains was observed and documented by imaging.

5.10. Analysis of behavior in AD mice

5.10.1. Morris water maze (MWM) test

Mice were subjected to the designated treatment for 4 months prior to testing in a Morris water maze with a diameter of 1.2 m. The water temperature was maintained at 22 ± 1 °C throughout the experiment. A training phase was conducted first, during which a 10 cm-diameter platform was fixed 1 cm below the water surface. Over the first 3 days, each mouse received four training trials per day, with the starting position fixed in each quadrant of the maze sequentially. On the 5th and 6th days, the platform was removed to evaluate the learning and memory abilities of the mice. Each mouse was gently placed into the water at a random position in the quadrant opposite to the original platform location and allowed to swim freely for 60 s. The escape latency to the platform (during the training phase), swimming speed, and the number of crossings over the original platform area (during the probe trial) were recorded.

5.10.2. Y-maze test

The three arms of the Y-maze were randomly designated as Arm A, Arm B, and Arm C. Each mouse was placed at the end of one randomly selected arm and allowed to explore freely for 5 min. The sequence of the arms that each mouse entered was recorded consecutively throughout the test. Based on these records, two key parameters were calculated: the total number of arm entries and the total number of spontaneous alternations (defined as the total count of consecutive triplets consisting of three distinct arm labels). The spontaneous alternation rate was calculated using the following formula: Spontaneous alternation rate = (Total number of spontaneous alternations/(Total number of arm entries - 2)) × 100%. The total number of arm entries was used to assess the locomotor activity of the mice in the Y-maze.

5.11. Immunohistochemistry

Brain tissue samples were embedded in paraffin and sectioned into 4 μm-thick slices. The paraffin sections were dewaxed to water and then placed in a container filled with citrate antigen retrieval solution (Servicebio, China) for antigen retrieval via microwave heating. Subsequently, the sections were blocked with 3% BSA at room temperature for 30 min. They were first incubated overnight at 4 °C with the primary antibody 4G8, followed by incubation with horseradish peroxidase (HRP)-conjugated secondary antibody. The brain sections were stained with 3,3′-diaminobenzidine (DAB), and cell nuclei were counterstained with hematoxylin for 3 min. Images were acquired using SLIDEVIEW VS200 Research Grade Slide Scanner (Olympus, Japan).

5.12. Statistical analyses

All data in the experiment were analyzed statistically using GraphPad Prism 8.0 software and expressed as mean ± SD deviation. Comparisons between data from multiple groups were performed using one-way ANOVA. Comparisons between groups were made by LSD-t or Tambane's analysis; if the variance was not uniform, Kruskal-Wallis test was used. p < 0.05 indicated that the difference was statistically significant.

CRediT authorship contribution statement

Xiaobin Huang: Data curation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing. Guobo Xu: Methodology, Supervision, Writing – review & editing. Yanhua Zhou: Data curation, Writing – review & editing. Dong Wang: Formal analysis, Writing – review & editing. Lili Ai: Formal analysis, Writing – review & editing. Hongmei Liu: Formal analysis, Writing – review & editing. Anran Fan: Data curation, Formal analysis, Funding acquisition, Visualization, Writing – original draft, Writing – review & editing. Likun Wang: Conceptualization, Writing – review & editing. Peng Zhang: Data curation, Formal analysis, Funding acquisition, Visualization, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data will be made available on request.