AP2A1 activates Rab7 to promote axonal autophagosome transport and slow the progression of Alzheimer’s disease

Abstract

Background

Dysregulation retrograde axonal transport in neurons results in autophagosome accumulation, enhancing amyloid β (Aβ) production and accelerating Alzheimer’s disease (AD) progression. Ras-associated GTP-binding protein 7 (Rab7) is pivotal in autophagosome maturation and their fusion with lysosomes, as well as in bidirectional axonal transport through interactions with partner proteins. Recent studies suggest that adapter-associated protein complex 2 subunit α1 (AP2A1) modulates retrograde axonal autophagosomes transport, regulates autophagy, and influences AD progression. However, the interplay between AP2A1 and Rab7, along with the molecular mechanisms underlying their impact on neuronal autophagy in AD, remains poorly understood.

Methods

We employed N2a/APPswe cells, primary hippocampal neurons exposed to Aβ oligomers, and APP/PS1 transgenic mice as AD models. To assess the impact of AP2A1 on Rab7 activity and autophagy, we conducted Rab7 pulldown activation assay, transmission electron microscopy (TEM), western blot and immunofluorescence (IF) staining were performed. The interaction between AP2A1 and Rab7 was examined by co-immunoprecipitation (Co-IP), IF staining and molecular docking. Live-cell imaging was utilized to monitor autophagosome axonal transport in primary hippocampal neurons. Aβ levels were quantified through immunohistochemistry and ELISA. Behavioral alterations in mice were evaluated using the Morris water maze, open field test, object recognition test and Y-maze.

Results

We observed reduced levels of AP2A1 and Rab7-GTP, accompanied by autophagosome accumulation, in AD models. Overexpression of AP2A1 restored autophagic flux in these cells. AP2A1 was found to bind and activate Rab7, facilitating the recruitment of retrograde axonal transport proteins DIC1 and RILP. Additionally, AP2A1 overexpression enhanced retrograde axonal autophagosome transport, reinstated autophagic flux, provided neuroprotection, and improved behavioral deficits in AD model mice through Rab7 activation.

Conclusions

Our findings demonstrate that AP2A1 activates Rab7 to restore autophagic function and mitigate AD progression, providing novel therapeutic perspectives for autophagy-targeted interventions in AD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13195-025-01771-1.

Introduction

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by cognitive decline and memory loss that affects millions of people worldwide. Despite extensive research, effective therapeutic strategies remain elusive primarily because of the complex pathophysiology of the disease. Amyloid β (Aβ) accumulation is widely acknowledged as the predominant pathogenesis of AD, with primary research and drug intervention focused on inhibiting the production and eliminating the build-up of Aβ [1].

Macroautophagy/autophagy is a critical cellular process that facilitates the degradation and recycling of damaged organelles and misfolded proteins, playing a vital role in maintaining cellular homeostasis. In neurons, which are nondividing and long-lived cells, autophagy is particularly important because of their inability to dilute aggregated proteins and damaged organelles through cell division [2]. Studies have revealed a notable disruption of autophagy in the brains of both AD patients [3] and model mice [4], characterized by the accumulation of autophagosomes, especially in dystrophic neurites, and Aβ deposition. Further studies have revealed that autophagy disruption in neurons not only leads to abnormal autophagosome degradation, but also creates favorable conditions for the generation and secretion of Aβ by neurons and exacerbates the pathological progression of AD [5, 6]. Therefore, promoting the movement and degradation of accumulated autophagosomes may effectively inhibit Aβ accumulation and slow AD progression. In neurons, axonal transport, including both anterograde transport (from the cell body to the axon terminal) and retrograde transport (from the axon terminal to the cell body), plays crucial roles in neuronal homeostasis, repair and regeneration, efficient neural signaling, and the occurrence and progression of neurodegenerative diseases [2, 7]. As highly polarized cells, neurons exhibit spatially segregated autophagy. Local autophagosome formation occurs in distal axons [8, 9] and requires transport to the cell body for lysosomal fusion and degradation [10, 11]. The accumulation of numerous autophagosomes in swollen axons in the brains of AD patients and model mice [12, 13] suggests impaired retrograde axonal transport in neurons.

Axonal transport in neurons is facilitated by motor proteins kinesin and dynein, which are responsible for anterograde and retrograde axonal transport respectively [14, 15]. Ras associated GTP binding protein 7 (Rab7) is a small GTPase that promotes autophagosome maturation and movement on the cytoskeleton and further promotes fusion with lysosomes [16, 17]. In neuronal retrograde transport, Rab7 recruits Rab interacting lysosomal protein (RILP) and oxysterol binding protein-related protein 1 L (ORP1L) in combination with Dynein to regulate the retrograde axonal transport of autophagosomes to fuse with somatic lysosomes [18–20]. Among these functions, Rab7 receives guanine nucleotide exchange factor (GEF) to facilitate the conversion from inactive Rab7-GDP to active Rab7-GTP [21], whereas GTPase-activating proteins (GAPs), such as TBC1D5, promote the conversion of Rab7-GTP to Rab7-GDP [22]. Subsequent investigations revealed reduced Rab7-GTP levels in isolated autophagosomes in AD models, whereas the overexpression of CCZ1-MON1A, a Rab7 GEF, activated RAB7 and promoted autophagosome maturation [23]. However, the involvement of Rab7 activity in regulating the axonal transport of autophagosomes within neurons in the context of AD remains largely unexplored.

Adaptor-related protein complex 2 subunit alpha 1 (AP2A1), a subunit of the AP-2 complex, can recognize and bind cargo proteins and clathrin, regulate clathrin-dependent endocytosis, and participate in nerve cell development, hematopoietic stem cell differentiation and synaptic vesicular circulation [24]. Emerging evidence highlights the pivotal role of AP2A1 in the pathogenesis of Alzheimer’s disease (AD). A seminal study comparing hippocampal tissues from AD and non-AD individuals revealed 818 significantly downregulated transcript clusters, with AP2A1 among them, demonstrating a pronounced reduction in AP2A1 expression in AD hippocampal tissues [25]. Consistent with this finding, Seyfried et al. reported a marked decline in the expression of key hub proteins, including AP2A1, within neuronal and synaptic protein networks in the brains of AD patients [26]. Detailed findings revealed that AP2A1 selectively targeted autophagic degradation by binding to LC3 through the LC3-interacting region (LIR) motif, while silencing AP2A1 using RNA interference significantly increased Aβ levels [27, 28]. In neurons, the interaction between AP2A1 and LC3B within the AP2 complex facilitates the retrograde axonal transport of autophagosomes and mitigates neurodegenerative lesions [29]. Moreover, defective AP2A1 mutants impair Aβ-cleaving enzyme BACE1-containing carriers involved in LC3B retrograde axonal transport, leading to elevated BACE1 protein levels and the upregulation of the Aβ₄₂ peptide [30]. Therefore, AP2A1 may play a crucial role in regulating the retrograde axonal transport of autophagosomes. However, the potential cooperation between AP2A1 and Rab7 and their impacts on autophagic impairment in AD neurons remain to be elucidated.

In this study, we discovered that AP2A1 expression was significantly reduced in AD models, which was correlated with reduced Rab7-GTP levels and autophagosome accumulation. This reduction in AP2A1 levels appears to hinder autophagic flux, a critical process for cellular homeostasis. Notably, we found that overexpression of AP2A1 effectively restored autophagic flux in AD cell models, highlighting its potential role in mitigating disease progression. Furthermore, AP2A1 was shown to bind and activate Rab7. By increasing the retrograde axonal transport of autophagosomes through Rab7 activation, AP2A1 reinstated unobstructed autophagic flux in AD neurons. Moreover, AP2A1 exerted neuroprotective effects and alleviated behavioral deficits in APP/PS1 mice. Our findings suggest the potential value of activating Rab7 through AP2A1 to increase the retrograde axonal transport of autophagosomes and restore autophagy as a therapeutic approach for AD.

Materials and methods

Plasmids

Vector plasmid (pcDNA3.1), empty Flag and Flag-AP2A1 plasmids were synthesized by Genecreat (Wuhan, China). GFP-Rab7 plasmid and mCherry-LC3 plasmid were purchased from Miaoling Biology (Wuhan, China), mCherry-GFP-LC3 was purchased from Beyotime Biotechnology.

Animals

Male APP/PS1 transgenic mice, male C57 mice (WT) and female C57 mice were obtained from Nanjing Junke Biological Co., Ltd. (Nanjing, Jiangsu, China). All mice were housed in the Animal Experimental Center of Chongqing University Cancer Hospital, under controlled environmental conditions with a 12 h light/dark cycle, and provided with standard laboratory chow and water ad libitum. The experimental protocols were approved by the Ethics Committee of Chongqing University Cancer Hospital, ensuring compliance with ethical standards for animal research.

Stereotactic injection of adeno-associated virus

Adeno-associated virus (AAV) vectors, including HBAAV2/9-Syn-ZsGreen and HBAAV2/9-Syn-h-AP2A1-3×flag, were constructed by HanBio Co. Ltd. (Shanghai, China). Stereotactic injections delivered the AAV vectors into the bilateral hippocampus of 8-month-old male APP/PS1 mice, targeting specific coordinates (anterior posterior, 2.3 mm; medial lateral, ± 1.8 mm; dorsal ventral, 2.0 mm). Prior to the injection, mice were anesthetized using isoflurane and placed in a stereotactic frame to ensure precise positioning. The scalp was incised to expose the skull, and a small burr hole was drilled at the predetermined coordinates. AAV vectors were then injected using a microinjector at a rate of 0.33 µL/min, with a total volume of 2 µL per site. After injection, the needle remained in place for 5 min to allow virus diffusion before slowly retraction. The incision was then sutured, and the mice were monitored post-surgery for recovery.

Intraperitoneal injection and cell treatment of CID1067700

CID1067700 (MCE, HY-13452, USA), a small molecule inhibitor was administered intraperitoneally to mice as well as treated cells, suppressing Rab7 activity. For animal experiments, CID1067700 solid was dissolved in a solution of 5% DMSO in 20% SBE-β-CD in sterile saline. Following hippocampal injection and a recovery period, 8- month-old male APP/PS1 mice were intraperitoneally injected at a dose of 16 mg/kg per week for 1 month. Control mice were injected intraperitoneally with a saline solution containing 5% DMSO and 20% SBE-β-CD in equal amounts. For cell experiments, CID1067700 (10 mM in DMSO) was used at a final concentration of 40 µM for 24 h. Control cells were subjected to identical DMSO treatment.

Cell culture and transfection

Mouse neuroblastoma 2a (N2a) cells and N2a cells stably expressing human Swedish APP (N2a/APPswe) were generously provided by Professor Xu Huaxi at Chongqing medical university. The cells were maintained in a humidified incubator at 37 °C with 5% CO₂. The culture medium consisted of Dulbecco’s Modified Eagle’s Medium (DMEM, Gibco, Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% Fetal Bovine Serum (FBS, Gibco, Thermo Fisher Scientific) and 1% Penicillin-Streptomycin (P/S, Gibco, Thermo Fisher Scientific). The plasmid or siRNA transfection were performed using Lipofectamine 3000 (Lipo3000, Invitrogen, USA) according to manufacturer’s instructions.

Primary hippocampal neurons were isolated from the brains of embryonic day 16–18 (E16-18) C57BL/6 mouse embryos. The extracted hippocampal tissues were minced-up and digested with Papain (2 mg/ml) and DNase (0.5 mg/ml) for 30 min at 37 °C. Following digestion, the tissue was mechanically dissociated to obtain a single-cell suspension. The cells were then resuspended in plating medium consisting of DMEM supplemented with 10% horse serum (Gibco, Thermo Fisher Scientific) and 0.5% P/S. The cell suspension was plated onto poly-D-lysine-coated culture dishes and incubated for 4 h. Then change the medium with a serum-free maintenance medium (Neurobasal Plus, Gibco, Thermo Fisher Scientific) supplemented with 2% B27 Plus (Gibco, Thermo Fisher Scientific), 1% GlutaMAX (Gibco, Thermo Fisher Scientific) and 0.5% P/S to promote neuronal growth and viability. Plasmid transfection using Lipofectamine 2000 (Lipo2000, Invitrogen, USA) was performed on primary hippocampal neurons cultured for 6–8 days in vitro (DIV 6–8).

Aβ oligomers preparation

Aβ oligomers were prepared as described previously [31]. Specifically, the Aβ₁₋₄₂ peptide powder (Aladdin, A494432) was dissolved in 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP, Aladdin, H107503) to a concentration of 1 mM. The solution was then incubated at room temperature for 60 min to ensure complete dissolution. After evaporating the HFIP, the dried peptide films were stored at -80 °C. Before use, the dried peptide film was dissolved in DMSO at a concentration of 1 mM and then incubated at 4 °C for 24 h to promote oligomerization. For treatment of primary hippocampal neurons from mice, the Aβ oligomer solution was used at a final concentration of 2 µM for 24 h to construct an AD model.

Western blot

Protein samples were extracted from cultured neurons and mouse brain tissues using Cell Lysis Buffer for Western and IP (Beyotime Biotechnology, P0013) supplemented with phenylmethanesulfonyl fluoride (PMSF). The protein concentration was determined using the Enhanced BCA Protein Assay Kit (Beyotime Biotechnology, P0010). Equal amounts of protein (30–50 µg) were separated by SDS-PAGE and transferred onto PVDF membranes (0.22 μm, Bio-Rad, Hercules, CA, USA). The membranes were blocked with 5% non-fat dry milk in TBST for 1 h at room temperature to prevent non-specific binding. Subsequently, the membranes were incubated overnight at 4 °C with primary antibodies (Table S1). After washing with TBST, the membranes were incubated with appropriate HRP-conjugated secondary antibodies (Table S1) for 1 h at room temperature. The protein bands were visualized using an ECL kit (Share Bio, Shanghai, China) and the Tanon-5200 multi chemiluminescence analysis system (Tanon, Shanghai, China). Densitometric analysis was performed using ImageJ software to quantify the relative protein expression levels.

Co-Immunoprecipitation (Co-IP) and Rab7 pulldown activation assay

Cells were lysed using Cell lysis buffer for Western and IP (Beyotime Biotechnology) supplemented with PMSF. The lysis products containing 500 to 1000 µg of protein were incubated overnight at 4 °C in a rotating shaker with specific antibodies at the indicated concentration/content (Table S1). The next day, 40 µl of well-mixed Protein A/G PLUS-Agarose (Santa Cruz Biotechnology, sc-2003) were added, and incubation continued for 4 h in at 4 °C in a rotating shaker. Subsequently, agarose beads were washed five times with lysis buffer, after which 1×loading buffer was added and boiled at 100 °C for 10 min. The samples were then analyzed by Western blot. To prevent interference from IgG heavy/light fragments in IP assays, Western blot detection typically employed antibodies from different species than those used in IP.

In Rab7 pulldown activation assay, an anti-active Rab7 mouse monoclonal antibody (NewEast Biosciences, 26923) was used to bind Rab7-GTP in cell lysates containing 1–2 mg of total protein. An anti-Rab7 Rabbit monoclonal antibody (Cell Signaling Technology, 9367) was used in the subsequent western blot assay.

Molecular docking

The crystal structure of human Rab7 was obtained from the Protein Data Bank (PDB: 1T91), while the AlphaFold v2.0 predicted structures of human AP2A1 and TBC1D5 were acquired from the AlphaFold Protein Structure Database. Ligands or water molecules were excluded from the protein structure to simplify the docking process. Docking was conducted using ZDOCK 3.0.2 on the ZDOCK server, with predictions ranked based on their scoring function. Higher-ranked binding affinity generally indicates stronger binding force and better stability. The docking results were visualized using PyMOL software v2.2.0.

Immunohistochemical staining

Mice were transcardially perfused with 0.9% saline and 4% paraformaldehyde (PFA, Beyotime Biotechnology, P0099) under terminal anesthesia induced by sodium pentobarbital (80 mg/kg). The brains were then excised and post-fixed overnight in 4% PFA at 4 °C. The fixed brains were embedded in paraffin following standard protocols. Coronal sections, 5 μm thickness, were obtained using a microtome. After baking at 80 °C for 1 h, the prepared paraffin sections were deparaffinized in xylene (2 × 5 min) and rehydrated in 100%, 95%, 85% and 70% ethanol for 5 min each. Endogenous peroxidase was blocked using a 3% hydrogen peroxide solution in methanol for 10 min at room temperature. Antigen retrieval was then performed using the microwave method with improved citrate antigen retrieval solution (Biosharp, BL151A) for 30 min. After cooling to room temperature, sections were washed three times with PBS for 5 min each and incubated with goat serum blocking solution for 20 min. Primary antibodies (Table S1) were incubated with sections overnight at 4 °C. After being rewarmed and rinsed with PBS the following day, sections were incubated with secondary antibodies (Table S1) at 37 °C for 30 min. After PBS washing, sections were stained using the DAB Substrate Kit (Biosharp, BL732A), counterstained with hematoxylin, dehydrated in alcohol, cleared in xylene, and finally sealed with neutral gum. Sections were scanned using a digital pathology slide scanner (KFBIO, KF-PRO-005, Ningbo, China) and images were acquired using the K-Viewer (1.7.1.1) software.

Immunofluorescence (IF) staining

Cells were rinsed twice with PBS and fixed with 4% PFA for 15 min at room temperature, followed by two more PBS washes. Cells were then permeabilized by incubation in Triton X-100 (Beyotime Biotechnology, P0096) for 15 min at room temperature. After PBS washing twice, cells were blocked with QuickBlock™ blocking buffer (Beyotime Biotechnology, P0260) for 15 min at room temperature. Subsequently, cells were incubated with the primary antibodies (Table S1) overnight at 4 ℃. After rinsing in PBS, cells were stained with secondary antibodies (Table S1) for 1 h at 37 ℃, incubated in the dark, and gently washed with PBS. DAPI was then added for nuclear staining at room temperature in the dark for 10 min. After PBS washing, antifade mounting medium (Beyotime Biotechnology, P0126) was applied for cover slipping. Imaging was performed using a laser confocal microscope (STELLARIS 5, Leica) and analyzed with ImageJ software.

Lysosome acidification assay

Cells were plated in 20 mm laser confocal dishes and treated as required. The LysoSensor™ Green DND-189 probe (1 µM in DMEM, Invitrogen, USA) was pre-warmed at 37 °C for 5 min, applied to cells, and incubated for 30 min at 37 °C in the dark. After washing with DMEM, lysosomal pH was analyzed using confocal microscopy.

Live-cell imaging and analysis

Primary hippocampal neurons at DIV 5–7 were treated with 2 µM Aβ oligomers for 24 h to construct AD model. Following this treatment, the neurons were co-transfected with plasmids mCherry-LC3 + Flag, mCherry-LC3 + Flag-AP2A1, or mCherry-LC3 + Flag-AP2A1 + CID1067700. Before imaging, neurons were subjected to starvation treatment in Hank’s Balanced Salt Solution (HBSS, Thermo Fisher Scientific) for 3 h in a cell culture incubator at 5% CO₂ and 37 °C. Live-cell imaging was conducted using an Olympus IX-83 microscope with cellSens live-cell imaging software (Olympus). Images or AVI format videos were captured, with video recordings lasting at least 100 s. For videos analysis, the Dynamic Kymograph plugin in Fiji was utilized to assess the axonal transport dynamics. According to previous reports, a net displacement of fluorescent puncta ≥ 10 μm was considered motion [32].

Transmission electron microscope (TEM)

Centrifuged cells and dissected mouse hippocampi were fixed with 2.5% glutaraldehyde overnight at 4 °C, then post-fixed with 1% osmium tetroxide. Dehydration was performed with a gradient of acetone (30–100%), following by a mixture of acetone and Epon-812 embedding agent (in ratios of 3:1, 1:1, and 1:3). Samples were embedded in Epon-812 and sectioned at thicknesses of 60 to 90 nm using an ultra-thin microtome. Ultrathin sections were stained for 10 to15 min with uranyl acetate, followed by 1 to 2 min with lead citrate. Stained samples were imaged using a transmission electron microscope (JEOL, JEM-1400FLASH). For each sample, low magnification was used to observe the entire cells, while specific lesions such as autophagosome accumulation around the nucleus of N2a/APP cells and axonal swelling of hippocampal neurons were observed after selecting the area for observation.

Enzyme-linked immunosorbent assay (ELISA)

N2a/APP cells were centrifuged at 1000×g for 20 min at 4 °C to remove impurities and cellular debris, obtaining the supernatant. Aβ₁₋₄₂ and Aβ₁₋₄₀ in the cell supernatant were detected and analyzed using ELISA Kits (Elabscience, E-EL-H0543 and E-EL-H0542) according to the manufacturer’s instructions.

Golgi-Cox staining

The mice brains were quickly extracted under deep anesthesia and washed with ultrapure water to remove blood. Golgi-Cox staining was performed using the Golgi-Cox Staining Kit (Servicebio, G1069, Wuhan, China). Specifically, the brains were immersed in Golgi-Cox staining solution (G1069-1) and kept in darkness at room temperature for 2 weeks. The staining solution was changed after 48 h and then replaced every three days. Subsequently, the tissue was transferred to tissue treatment solution (G1069-3) and kept in darkness at 4 ℃ for three days, with a new solution replaced after 1 h. Then the tissue was sliced into coronal sections with a thickness of 60 μm using a vibrating microtome. Subsequently, the sections were transferred onto a glass slide immersed in tissue treatment solution (G1069-3), coated with neutral glue, and gently wiped to eliminate excess fluid. Following rinsing with ultrapure water, the sections were developed with Golgi developer solution (G1069-2) for 30 min. The sections were sealed with neutral after rinsing with ultrapure water and drying, then imaged using a digital tissue slice scanner (Pannoramic MIDI, 3DHISTECH) and CaseViewer software version 2.4 (3DHISTECH).

Morris water maze

The Morris Water Maze (MWM) was employed to assess spatial learning and memory in 9-month-old WT and APP/PS1 male mice. The apparatus was a circular pool with a diameter of 90 cm and a height of 50 cm, filled with water stained opaque by white nontoxic paint maintained at a constant temperature of 25 °C. The platform, 6 cm in diameter, was submerged 1 cm below the water surface. The acquisition phase, known as the hidden platform test, was conducted over five consecutive days. Each day, mice were placed in the pool facing the wall from one of four quadrants at the same time point, and allowed to swim for 60 s. The time each mouse took to locate and climb onto the platform was recorded. Mice that failed to find the platform within the allotted time were gently guided to it and allowed to acclimate for 10 s. On the sixth day, the platform was removed for the spatial probe trial, where mice were again placed in the pool from the four quadrants at the same time points for 60 s. Data collected during this trial included the time taken to find the former platform location, the number of crossings over the platform’s previous location, the duration spent in the target quadrant, and the average swimming speed. Mice movements were tracked and recorded using the SMART 3.0 software (Panlab Harvard Apparatus) for precise analysis of their swimming patterns.

Open field test

The Open Field Test (OFT) was employed to assess the locomotor activity and anxiety-like behavior of 9-month-old wild-type (WT) and APP/PS1 male mice. The experimental field was constructed from opaque gray plexiglass and partitioned into four chambers of equal volume to enable concurrent observation of four mice. Each chamber measured 50 cm (length) × 50 cm (width) × 46.7 cm (height), with a central zone measuring 25 cm (length) × 25 cm (width). Before each trial, the arena was thoroughly disinfected with 75% ethanol to ensure a clean environment and minimize any potential confounding factors. After a 2 min accustomed period, each group of mice was placed in the center of the open field for 8 min. The total movement distance and time spent in the center zone were recorded using the EthoVision XT 17 animal tracking system (Noldus).

Object recognition test

The cognitive and memory abilities of the mice were evaluated using the Object Recognition Test (ORT), which was conducted utilizing the identical apparatus as employed in the field test. The arena was thoroughly disinfected with 75% ethanol before each trial in order to establish a sterile environment and minimize potential confounding variables. The procedure began with an adaptation phase on day 1, where mice were placed in an empty object recognition box for 5 min to familiarize themselves with the experimental environment. On day 2, the training phase involved placing two identical objects within the recognition box, allowing the mice to freely explore these objects for a duration of 5 min. In the testing phase on day 3, one familiar object was replaced with a novel object. The mice were then reintroduced into the recognition box to explore both the familiar and the new object for an additional 5 min. During this testing period, exploration time and the number of interactions with each object were recorded using the EthoVision XT 17 animal tracking system (Noldus) to evaluate recognition memory, with a preference for the novel object indicating successful memory retention.

Y-maze test

The Y-maze test was used to assess short-term memory in mice. The Y-maze apparatus consists of three arms of equal length arranged in a Y-shape, each forming a 120° angle with the others. The maze is typically constructed from opaque gray plexiglass to minimize visual distractions. During the test, mice were placed in the central area of the maze and allowed to explore freely for 8 min. During the exploration period, the number and sequence of entries into each arm were recorded. Spontaneous alternation behavior was defined as the mice entering three different arms sequentially without returning to any previously visited arm. The spontaneous alternation percentage (SAP) was calculated as: SAP = spontaneous alternations / possible alternations × 100. Possible alternations were calculated as total arm entries minus 2.

Statistical analysis

Statistical analyses were conducted using GraphPad Prism software. Data were checked for normality and heteroscedasticity. An unpaired two-tailed Student’s t-test (for 2 groups) and one-way ANOVA followed by Tukey’s multiple comparisons test (for > 2 groups) were used for the data conforming to normal distribution and homogeneous variance. An unpaired two-tailed Welch’s t-test (for 2 groups) and Brown-Forsythe and Welch one-way ANOVA with Dunnett’s T3 multiple comparisons test (for > 2 groups) were used for the data conforming to normal distribution and uneven variance. An unpaired two-tailed Mann–Whitney U test (2 groups) and Kruskal-Wallis test with Dunn’s multiple comparisons test (> 2 groups) were used for the data that did not conform to normal distribution. Data are presented as mean ± standard error of the mean (SEM), and p < 0.05 was considered statistically significant.

Results

AP2A1 overexpression restores autophagic flux in AD model cells

We initially detected the expression levels of AP2A1 in several AD models. The results revealed high AP2A1 protein expression in the hippocampi of normal 9-month-old WT mice, whereas it was significantly reduced in 9-month-old APP/PS1 mice (Fig. 1A, B). Further analysis revealed reduced AP2A1 levels in the hippocampi of AD model mice, along with decreased levels of Rab7-GTP and increased levels of the autophagy markers SQSTM1 and LC3-II, compared with those in WT mice (Fig. 1C, D). Similarly, analysis of APP cells (Fig. S1A, B) and primary hippocampal neurons treated with Aβ oligomers (Fig. S1C, D) confirmed AP2A1 downregulation, along with decreased Rab7-GTP and increased SQSTM1 and LC3-II levels, compared with those in control cells. Furthermore, TEM revealed a marked increase in the number of autophagosomes in AD model cells (Fig. S1E, F), and a significant accumulation of autophagic vesicles, including autophagosomes, in the axons of the hippocampal neurons of AD model mice (Fig. 1E). This accumulation led to pronounced axonal swelling, indicating disrupted axonal transport and impaired autophagic flux in the context of AD pathology (Fig. 1E). Overall, the associations between reduced AP2A1 expression in AD models and impaired autophagosome transport and accumulation, as well as the unknown mechanism of action, warrant further investigation.

Fig. 1.

AP2A1 decreased in AD models and its overexpression restores autophagic flux in AD cell models. (A) Immunohistochemical staining of AP2A1 in the hippocampi of WT and APP/PS1 mice (9 months old). Scale bar = 100 μm. (B) Quantification of AP2A1 expression as shown in (A), n = 6. (C) Western blot analysis of AP2A1, Rab7-GTP, SQSTM1, LC3 and Rab7 in the hippocampi of WT and APP/PS1 mice. Total Rab7 and corresponding GAPDH (bottom) were analyzed in independent experiments using frozen aliquots of the original hippocampal lysates under identical conditions to ensure comparability. (D) Quantification of protein expression as shown in (C), n = 3. (E) Representative TEM images of autophagosome accumulation within the axons of hippocampal neurons from WT and APP/PS1 mice (9 months old). Red arrow: autophagosomes with bilayer membranes and internal cytoplasmic component that is not degraded. Yellow arrow: autolysosomes with single-limiting membranes and electron-dense contents with varying degrees of degradation, consistent with the morphology of the reference electron microscope [33, 34]. Scale bar = 1 μm. (F) Schematic of the experimental design and representative live-cell imaging of mCherry‒LC3 axonal retrograde movement at 0 s and 30 s in primary hippocampal neurons treated with Aβ oligomers overexpressing Flag-AP2A1 or Flag. Scale bar = 10 μm. (G) Quantification of the relative axonal retrograde movement of mCherry-LC3 puncta as shown in (F), n = 15. (H) IF image of mCherry-GFP-LC3 in APP cells overexpressing Flag-AP2A1 or Flag. (I) Quantification of mCherry-GFP-LC3 puncta as shown in (H), n = 18. (J) Representative TEM images of autophagosomes in APP cells overexpressing either Flag-AP2A1 or Flag. Red arrow: autophagosomes. Yellow arrow: autolysosomes. Scale bar = 1 μm. (K) Quantification of autophagosomes as shown in (J), n = 6. (L) ELISA quantification of Aβ₁₋₄₂ and Aβ₁₋₄₀ in the supernatant of APP cells overexpressing either Flag-AP2A1 or Flag, n = 6. The data were analyzed using unpaired two-tailed Welch’s t-test (B) and unpaired two-tailed Student’s t-test (D, G, I, K, L). The data are presented as the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001; ns, not significant

To investigate the role of AP2A1 in axonal autophagosome transport in an AD cell model, primary hippocampal neurons were treated with Aβ oligomers and subsequently cotransfected with mCherry-LC3 (to label autophagosomes) and either Flag-AP2A1 or control Flag (Fig. 1F). Live-cell imaging revealed that Flag-AP2A1 overexpression increased the number of autophagosomes (mCherry-LC3) capable of effective retrograde axonal transport (≥ 10 μm) between 0 s and 30 s compared with that in the control group (Fig. 1F, G). Western blot analysis confirmed robust Flag-AP2A1 overexpression in Aβ oligomer-treated neurons compared to the Flag-control group (Fig. S2), validating transfection efficacy. To further elucidate the effect of AP2A1 on the interaction between autophagosomes and lysosomes, IF staining was performed, which revealed that Flag-AP2A1 overexpression increased the colocalization of the autophagosome marker LC3B with the lysosomal marker LAMP1 in APP cells (Fig. S3A, B) and led to a reduction in the level of the autophagic substrate SQSTM1 in APP and N2a cells (Fig. S3C-E). We subsequently used mCherry-GFP-LC3 fluorescent plasmids to visualize autophagosome‒lysosome fusion. Before fusion, autophagosomes appeared as yellow puncta because of the colocalization of mCherry (red) and GFP (green). After fusion, the GFP signal was quenched in the acidic lysosomal environment, resulting in red puncta representing autolysosomes. Overexpression of Flag-AP2A1 in APP cells significantly decreased the number of autophagosomes (yellow dot plot) and increased the number of autolysosomes (red dot plot) (Fig. 1H, I), suggesting increased autophagic degradation. Furthermore, TEM analysis revealed that, compared to control, Flag-AP2A1 overexpression markedly decreased the number of autophagosomes but significantly increased the number of autolysosomes in APP cells (Fig. 1J, K). Notably, this enhanced autophagic flux occurred independently of upstream initiation signaling, as evidenced by unchanged ratios of p-ULK1/ULK1 and Beclin1 protein levels in both hippocampus of AAV-AP2A1 mice compared to AAV-Control group (Fig. S4A, B), and in Flag-AP21-overexpressing APP cells relative to Flag vector controls (Fig. S4C, D). Consistent with this observation, the conjugation levels of Atg12-ATG5 complexes—a biochemical hallmark of autophagosome formation—remained unaltered in AP2A1-overexpressing APP cells (Fig. S4E, F). Importantly, the functional preservation of lysosomal proteolytic capacity was further confirmed by stable protein levels of cathepsin D (Fig. S4G, H) and sustained lysosomal acidification as indicated by comparable LysoSensor Green fluorescence intensity (Fig. S4I, J). Collectively, these data pinpoint that AP2A1-mediated autophagic enhancement specifically facilitates the autophagosome-lysosome fusion process rather than globally upregulating autophagy initiation, autophagosome formation or lysosomal function. Finally, ELISA results revealed that AP2A1 overexpression significantly inhibited Aβ₁₋₄₂ and Aβ₁₋₄₀ secretion from APP cells (Fig. 1L). Collectively, these findings suggest that AP2A1 plays a crucial role in restoring autophagic flux and reducing Aβ secretion in AD model cells, suggesting its potential as a therapeutic candidate for AD intervention.

AP2A1 binds and activates Rab7 to recruit retrograde axonal transport proteins

The decreased expression of AP2A1 in the hippocampi of APP/PS1 mice, N2a/APP cells and primary hippocampal neurons treated with Aβ oligomers was associated with a concomitant inhibition of Rab7-GTP activation (Fig. 1C, D and Fig. S1A-D), suggesting a potential close association between AP2A1 and Rab7-GTP that warranted further investigation. In N2a cells, Co-IP assays revealed multiple interactions: Flag-AP2A1 with GFP-Rab7 (Fig. 2A), endogenous Rab7 with Flag-AP2A1 (Fig. 2B), GFP-Rab7 with endogenous AP2A1 (Fig. 2C), and endogenous AP2A1 with endogenous Rab7 (Fig. 2D). Moreover, AP2A1 and Rab7 were significantly colocalized in N2a cells (Fig. 2E). These results suggest a strong connection between AP2A1 and Rab7 in normal neurons. Moreover, molecular docking studies were used to predict the structural domain binding sites of AP2A1 to Rab7 (Fig. 2F). Visual analysis revealed that the residues of GLU-27, ASN-34, LYS-35, ARG-41 and SER-42 in AP2A1 can form hydrogen bonds with ARG-113, ARG-79, GLN-60, ALA-43 and THR-47 in Rab7. Additionally, LYS-35 and LYS-45 in AP2A1 can form salt bridges with GLU-177 and ASP-44 in Rab7, resulting in a stable interaction between the two proteins (Fig. 2F).

Fig. 2.

AP2A1 binds and activates Rab7 to recruit axonal retrograde proteins. (A) Co-IP of Flag-AP2A1 and GFP-Rab7 in N2a cells. (B) Co-IP of Flag-AP2A1 and endogenous Rab7 in N2a cells. (C) Co-IP of GFP-Rab7 and endogenous AP2A1 in N2a cells. (D) Co-IP of endogenous Rab7 and endogenous AP2A1 in N2a cells.  (E) Representative IF image showing the colocalization of AP2A1 (green) and Rab7 (red) in N2a cells. Scale bar = 10 μm. Pearson’s R = 0.71 ± 0.05, n = 18. (F) Molecular docking of AP2A1 and Rab7. (G) Co-IP of Rab7-GTP with RILP, DIC1 and KIF5B in APP cells overexpressing Flag-AP2A1 or Flag. (H) Quantification of protein expression as shown in (G), n = 3. (I) Co-IP of Rab7 and TBC1D5 in APP cells overexpressing either Flag-AP2A1 or Flag. (J) Quantification of protein expression as shown in (I), n = 3. (K) Molecular docking between AP2A1 and Rab7 and between TBC1D5 and Rab7. The data were analyzed using unpaired two-tailed Student’s t-test (H, J). The data are presented as the mean ± SEM. **p < 0.01; ns, not significant

To assess the impact of AP2A1 overexpression on Rab7 activity, we used an antibody specifically against active Rab7-GTP. Compared with that in control cells, Flag-AP2A1 overexpression in APP cells significantly activated Rab7, increasing the recruitment of the retrograde axonal transport proteins RILP and DIC1 without affecting the anterograde axonal transport protein KIF5B recruitment (Fig. 2G, H). Consistent with this observation, the overexpression of AP2A1 significantly increased the colocalization of autophagosomes (LC3) and Dynein Light Chain 1/2 (DIC1/2) in APP cells, whereas no significant effect was observed in N2a cells (Fig. S5A, B). Similarly, the overexpression of AP2A1 did not influence the colocalization of Rab7 and KIF5B in either N2a or APP cells (Fig. S5C, D). However, the molecular mechanism by which AP2A1 activates Rab7 remains unclear. TBC1D5, a known Rab7 GAP, regulates the conversion of active Rab7-GTP to inactive Rab7-GDP [35]. Given the contrasting effects of AP2A1 and TBC1D5 on Rab7 activity, we investigated whether the AP2A1 mediated activation of Rab7 was associated with TBC1D5. As shown in Fig. 2I, J, we observed a significant decrease in the binding affinity of TBC1D5 to Rab7 in APP cells after AP2A1 overexpression compared with the control. Further molecular docking analysis between AP2A1 and TBC1D5 with Rab7 revealed that ALA-43, THR-47 and ARG-79 of Rab7 formed hydrogen bonds with ASN-242, LYS-238 and GLN-222 of TBC1D5, as well as ARG-41, SER-42 and ASN-34 of AP2A1. Similarly, ASP-44 of Rab7 formed a salt bridge with ARG-113 of TBC1D5 and LYS-45 of AP2A1. These findings indicated competitive binding between AP2A1 and TBC1D5 at overlapping Rab7 residues (ALA-43, ASP-44, THR-47, ARG-79; Fig. 2K), consistent with a proposed mechanism wherein AP2A1 prevents TBC1D5-mediated Rab7 inactivation. However, direct biochemical evidence demonstrating modulation of Rab7 GTP-loading status through this competition remains to be established. Collectively, these findings suggest that AP2A1 interacts with Rab7 to increase its activity and promote the recruitment of retrograde axonal transport proteins.

AP2A1 promotes the retrograde axonal transport of autophagosomes by activating Rab7

Studies have demonstrated that AP2A1 plays a crucial role in facilitating the retrograde axonal transport of autophagosomes [29] and BACE1 [30]. Rab7 (Rab7-GTP) and RILP coactivate dynein motors to promote vesicle transport [18, 36, 37]. These findings suggest a pivotal role for AP2A1 and Rab7 in regulating autophagosome transport along axons. Notably, our study revealed that AP2A1 overexpression in APP cells markedly activated Rab7 and promoted Rab7-GTP-RILP-DIC1 recruitment but did not recruit the anterograde axonal transport protein KIF5B. Increased Rab7-GTP-RILP-DIC1 recruitment promoted the formation of retrograde axonal transport protein complexes (Fig. 2G, H). We then explored the effect of AP2A1 activation on the Rab7-mediated retrograde axonal transport of autophagosomes in AD model neurons.

Compared with the control, the overexpression of Flag-AP2A1 significantly increased Rab7-GTP levels in APP cells. Cotreatment with CID1067700, a Rab7-GTPase inhibitor, markedly suppressed Rab7 activity (Fig. 3A, B). Additionally, CID1067700 treatment inhibited the Rab7-GTP-mediated recruitment of the retrograde transport proteins DIC1 and RILP (Fig. 3A, B). These findings support the idea that AP2A1-induced recruitment of retrograde axonal transport protein complexes is mediated through Rab7 activation. Furthermore, overexpression of Flag-AP2A1 significantly increased the colocalization of autophagosomes (marked by LC3) with DIC1/2 in APP cells, an effect that was abrogated upon Rab7-GTP inhibition with CID1067700 (Fig. 3C, D). Subsequently, live-cell imaging of primary hippocampal neurons treated with Aβ oligomers showed that compared to the control, overexpression of Flag-AP2A1 altered autophagosome (mCherry-LC3) localization in the soma at 30 s, and this change was subsequently abrogated by CID1067700 (Fig. 3E). Moreover, we monitored the retrograde axonal transport of autophagosomes in living primary hippocampal neurons for 100 s. Overexpression of Flag-AP2A1 significantly increased the number of autophagosomes in axons undergoing continuous retrograde transport, whereas CID1067700 impeded this process (Fig. 3F, G). Collectively, these findings highlight the critical role of AP2A1 in increasing the axonal retrograde transport of autophagosomes through Rab7 activation. This may help alleviate the disrupted autophagic flow in AD neurons.

Fig. 3.

AP2A1 promotes the retrograde axonal transport of autophagosomes by activating Rab7. (A) Co-IP between Rab7-GTP and RILP, DIC1 and KIF5B in APP cells overexpressing Flag or Flag-AP2A1, with or without CID1067700 treatment. (B) Quantification of protein expression as shown in (A), n = 3. (C) IF staining for the colocalization of autophagosomes (LC3, red), DIC1/2 (green) and DAPI (blue) in APP cells overexpressing Flag or Flag-AP2A1, with or without CID1067700 treatment. Scale bar = 10 μm. (D) Quantification of the colocalization of LC3 and DIC1/2 as shown in (C), n = 20–25. (E) Schematic experimental design and representative live cell imaging of mCherry-LC3 axonal retrograde movement at 0 s and 30 s, in primary hippocampal neurons treated with Aβ oligomers, overexpressing Flag or Flag-AP2A1, with or without CID1067700 treatment. Scale bar = 10 μm. (F) Kymographs (100 s) of mCherry-LC3 axonal retrograde movement in primary hippocampal neurons treated with Aβ oligomers overexpressing Flag or Flag-AP2A1, with or without CID1067700 treatment. Scale bar = 10 μm (horizontal)/ 20 s (vertical). (G) Quantification of the relative axonal retrograde movement of mCherry-LC3 puncta as shown in (F), n = 18–23. The data were analyzed using one-way ANOVA followed by Tukey’s multiple comparisons test (B, D, G). The data are presented as the mean ± SEM. **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant

AP2A1 reinstates unobstructed autophagic flux in AD model neurons by activating Rab7

Our study revealed that AP2A1 promotes the retrograde axonal transport of autophagosomes in AD model neurons by activating Rab7 (Fig. 3). To investigate whether the transported autophagosomes can fuse with lysosomes for autophagic degradation, we further examined the impact of the AP2A1-mediated activation of Rab7 on autophagic degradation. Compared with the control, overexpression of Flag-AP2A1 significantly decreased the LC3-II/I ratio and markedly reduced the expression of the autophagic substrate SQSTM1 in APP cells; this effect was notably abrogated by the inhibition of Rab7 activity via CID1067700 (Fig. 4A, B). These findings indicate that AP2A1 facilitates autophagic flux in APP cells by activating Rab7. Notably, overexpression of Flag-AP2A1 in APP cells significantly reduced the number of autophagosomes while increasing the number of autolysosomes, and these effects were abrogated by CID1067700 treatment (Fig. 4C, D). Additionally, there was a significant increase in colocalization of autophagosomes (LC3) with lysosomes (LAMP1) in APP cells overexpressing Flag-AP2A1, alongside a marked decrease in SQSTM1 fluorescence intensity; both of these changes were abrogated by CID1067700 (Fig. S6A-D). Consistently, TEM revealed that overexpression of Flag-AP2A1 in APP cells significantly reduced the number of autophagosomes while markedly increasing the number of lysosomes (Fig. 4E). In accordance with the schematic experimental design shown in Fig. 4F, the effect of AP2A1 on the autophagy of hippocampal neurons in AD model mice was further studied. The overexpression of AAV-AP2A1 in the hippocampi of 9-month-old AD mice significantly alleviated axonal swelling and mitigated the accumulation of autophagic vesicles in control neurons (Fig. 4G). However, cotreatment with CID1067700 during AAV-AP2A1 overexpression failed to reduce the accumulation of autophagosomes in APP cells and AD model mouse hippocampal neurons (Fig. 4E, G). Consistent with the results in APP cells, the overexpression of AAV-AP2A1 in AD model mouse hippocampal neurons significantly increased the LC3-II/I ratio and reduced SQSTM1 expression, and this effect was also inhibited by CID1067700 (Fig. 4H, I). Collectively, these results underscore the critical role of AP2A1 in restoring autophagic flux in AD model neurons by activating Rab7, thereby mitigating the autophagic dysfunction associated with Aβ accumulation.

Fig. 4.

AP2A1 restores autophagic flux in AD model neurons by activating Rab7. (A) Western blot analysis of Rab7-GTP, Flag-AP2A1, Rab7, LC3, and SQSTM1 in APP cells overexpressing Flag or Flag-AP2A1, with or without CID1067700 treatment. (B) Quantification of protein expression as shown in (A), n = 3. (C) IF imaging of mCherry-GFP-LC3 in APP cells overexpressing Flag or Flag-AP2A1, with or without CID1067700 treatment. (D) Quantification of mCherry-GFP-LC3 puncta as shown in (C), n = 20–25. (E) Representative TEM images of autophagosomes and autolysosomes in APP cells overexpressing Flag or Flag-AP2A1, with or without CID1067700 treatment. Red arrow: autophagosomes. Yellow arrow: autolysosomes. Scale bar = 1 μm. Quantification of the number of autophagosomes, n = 6. (F) Schematic of the experimental design shown in the figure. (G) Representative TEM images of autophagosome accumulation within the axons of hippocampal neurons from the hippocampi of WT and APP/PS1 mice hippocampal injected with AAV-Control or AAV-AP2A1, with or without CID1067700 treatment. Red arrow: autophagosomes. Yellow arrow: autolysosomes. Scale bar = 1 μm. (H) Western blot analysis of Flag-AP2A1, Rab7-GTP, SQSTM1 and LC3 in the hippocampi of WT and APP/PS1 mice injected with AAV-Control or AAV-AP2A1, with or without CID1067700 treatment. (I) Quantification of protein expression as shown in (H), n = 3. The data were analyzed using one-way ANOVA followed by Tukey’s multiple comparisons test (B, D, E-autolysosome, I) and Brown-Forsythe and Welch one-way ANOVA with Dunnett’s T3 multiple comparisons test (E-autophagosome). The data are represented as the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant

AP2A1 exerts neuroprotective effects by activating Rab7

Numerous studies indicate that restoring autophagy in neurons can slow AD progression. Our research confirmed that AP2A1 reduced axonal autophagosome accumulation in AD model neurons and restored autophagy. Thus, we evaluated the neuroprotective effects of AP2A1-mediated Rab7 activation in AD models as shown in the schematic of the experimental design in Fig. 5A. Initially, we observed a significant reduction in AP2A1 expression in the hippocampi of 9-month-old APP/PS1 mice injected with AAV-Control compared with age-matched WT mice. After AAV-AP2A1 administration, a marked increase in AP2A1 expression was detected in the hippocampi of both experimental mouse groups. This increase in AP2A1 expression confirms the efficacy of our hippocampal AAV-AP2A1 delivery method (Fig. S7A, B).

Fig. 5.

AP2A1 exerts neuroprotective effects by activating Rab7. (A) Schematic of the experimental design shown in figure. (B) Immunohistochemical staining of GFAP and IBA1 in the hippocampi of WT and APP/PS1 mice injected with AAV-Control or AAV-AP2A1, with or without CID1067700 treatment (9 months old). Scale bar = 625 μm (hippocampus) /200 µm (CA1/CA3/DG). (C) Quantification of GFAP expression as shown in (B-top), n = 6. (D) Quantification of IBA1 expression as shown in (B-bottom), n = 6. (E) Golgi-Cox staining of hippocampal CA1 neurons from WT and APP/PS1 mice injected with AAV-Control or AAV-AP2A1, with or without CID1067700 treatment (9 months old), Scale bar = 10 μm. Quantification of dendritic spine numbers, n = 6. (F) Schematic of the experimental design, IF staining of Synaptophysin (red, left) and PSD95 (red, right) in primary hippocampal neurons treated with Aβ oligomers, overexpressing Flag or Flag-AP2A1, with or without CID1067700 treatment. Scale bar = 5 μm. (G) Quantification of relative protein intensity as shown in (F), n = 22‒30. (H) Immunohistochemical staining and quantification of 6E10 in the hippocampi of APP/PS1 mice injected with AAV-Control or AAV-AP2A1, with or without CID1067700 treatment (9 months old), Scale bar = 400 μm, n = 6. (I) ELISA of Aβ₁₋₄₂ and Aβ₁₋₄₀ levels in the supernatant of APP cells overexpressing Flag or Flag-AP2A1, with or without CID1067700 treatment, n = 6. The data were analyzed using one-way ANOVA followed by Tukey’s multiple comparisons test (C, D, E, G, H, I). The data are represented as the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

A hallmark of AD is the activation of astrocytes and microglia, resulting in neuroinflammation. Our results revealed that, compared with that in WT mice, the staining intensity for glial fibrillary acidic protein (GFAP) in the hippocampi of AD model mice was significantly greater, indicating robust activation of astrocytes (Fig. 5B top, C). In contrast, AAV-AP2A1 injection into the hippocampi of AD model mice markedly reduced GFAP staining intensity (Fig. 5B top, C). However, the reduction in GFP staining intensity was abrogated when Rab7 activity was inhibited by CID1067700 in combination with AAV-AP2A1 injection (Fig. 5B top, C). These findings suggest that AP2A1 suppresses the activation of astrocytes in the hippocampi of AD model mice through the activation of Rab7. Similarly, ionized calcium-binding adapter molecule 1 (IBA1) staining intensity in the hippocampi of AD model mice was significantly greater than that in WT mice, indicating extensive microglial activation. AAV-AP2A1 injection into the hippocampi of AD model mice also significantly reduced IBA1 staining by activating Rab7 (Fig. 5B bottom, D).

We then examined subtle morphological changes in neuronal dendritic spines using Golgi-Cox staining. The spine density in the hippocampal CA1 neurons of AD model mice was significantly lower than that in the hippocampal CA1 neurons of WT mice, but AAV-AP2A1 injection markedly increased the spine density (Fig. 5E). However, this effect was significantly inhibited when Rab7 activity was suppressed by CID1067700 (Fig. 5E). Furthermore, in primary hippocampal neurons from mice treated with Aβ oligomers, Flag-AP2A1 overexpression significantly increased the fluorescence intensities of synaptophysin and PSD95, both of which are closely related to neuronal structure (Fig. 5F, G). This effect was absent when Rab7 activity was inhibited by CID1067700 (Fig. 5F, G). These results indicate that AP2A1 alleviates structural deficits in AD model neurons and increases neuronal complexity by activating Rab7.

To further investigate the impact of AP2A1-mediated Rab7 activation on Aβ deposition in the hippocampi of AD model mice, we used a 6E10 antibody targeting Aβ₁₋₁₆. Compared with the control, AAV-AP2A1 injection into the hippocampi of AD model mice significantly reduced 6E10 deposition, but this reduction was significantly abrogated when Rab7 activity was inhibited by CID1067700 (Fig. 5H). Notably, western blot analysis revealed that AAV-AP2A1 overexpression did not alter the expression levels of APP/sAPPα (soluble amyloid precursor protein α) or βCTF (β-carboxy-terminal fragment)—key intermediates in the amyloidogenic processing pathway—in hippocampal lysates compared to the control group (Fig. S8A, B), indicating that AP2A1-mediated Aβ reduction is independent of APP processing or Aβ generation pathways. Similarly, the ELISA results revealed that Flag-AP2A1 overexpression inhibited the secretion of Aβ₁₋₄₂ and Aβ₁₋₄₀ in the supernatant of APP cells (Fig. 5I). However, when Rab7 activity was inhibited by CID1067700, Aβ₁₋₄₂ and Aβ₁₋₄₀ secretion became uncontrolled again (Fig. 5I). These results demonstrate that AP2A1 activation of Rab7 inhibits Aβ deposition in the context of AD pathology. In summary, our findings provide compelling evidence that AP2A1 exerts a neuroprotective effect in AD models by activating Rab7, which reduces neuroinflammation, alleviates neuronal structural deficits, and inhibits Aβ deposition. These findings underscore AP2A1 as a potential therapeutic candidate for further preclinical exploration in AD.

AP2A1 alleviates behavioral deficits in APP/PS1 mice by activating Rab7

To investigate the effect of AP2A1 mediated Rab7 activation on AD-related behavioral deficits in 9-month-old APP/PS1 mice, a series of behavioral experiments were conducted as shown in the schematic experimental design in Fig. 6A. The MWM test was used to assess spatial learning and memory. During the 5-day acquisition phase, all groups showed a time-dependent decrease in escape latency (Fig. 6B). On the 5th day of the acquisition phase and the 6th day of the platform removal phase, the escape latency of the AAV-AP2A1-injected mice was significantly shorter than that of the AAV-Control mice (Fig. 6B). Conversely, the escape latency of the mice treated with CID1067700 alongside AAV-AP2A1 significantly increased (Fig. 6B). In subsequent spatial exploration tests, the AAV-Control mice crossed the platform fewer times than the WT mice did, whereas the AAV-AP2A1 mice presented a marked increase in platform crossings (Fig. 6C). However, the addition of CID1067700 significantly decreased the number of crossings (Fig. 6C). Furthermore, the AAV-Control mice spent significantly less time in the target quadrant than the WT mice did, whereas the AAV-AP2A1 mice spent more time in the target quadrant than the AAV-Control mice did, although the difference was not statistically significant (Fig. 6D). Notably, compared with AAV-AP2A1 treatment, concurrent CID1067700 treatment reduced this time (Fig. 6D). The average swimming speeds were similar across the groups (Fig. 6E), and trajectory plots illustrated differences in spatial memory (Fig. 6F).

Fig. 6.

AP2A1 alleviates behavioral deficits in APP/PS1 mice by activating Rab7. (A) Schematic of the experimental design shown in the figure. Escape latency (B), number of platform crossings (C), retention time in target quadrants (D), and mean swimming speed (E) in the MWM test of WT and APP/PS1 mice injected with AAV-Control or AAV-AP2A1, with or without CID1067700 treatment (9 months old). n = 8. (F) Typical trajectories in the MWM test. Distance traveled (G) and time spent in the center (H) in the OFT of WT and APP/PS1 mice injected with AAV-Control or AAV-AP2A1, with or without CID1067700 treatment (9 months old). n = 8. (I) Typical trajectories in the OFT. (J) Preference for novel objects in the ORT of WT and APP/PS1 mice injected with AAV-Control or AAV-AP2A1, with or without CID1067700 treatment (9 months old). n = 8. (K) Spontaneous alternation in the Y-maze of WT and APP/PS1 mice injected with AAV-Control or AAV-AP2A1, with or without CID1067700 treatment (9 months old). n = 8. The data were analyzed using one-way ANOVA followed by Tukey’s multiple comparisons test (B, D, E, G, H, J) and the Kruskal‒Wallis test with Dunn’s multiple comparisons test (C, K). The data are represented as the mean ± SEM. *p < 0.05, **p < 0.01 (AAV-AP2A1 vs. AAV-Control); #p < 0.05, ##p < 0.01 (AAV-AP2A1 + CID1067700 vs. AAV-AP2A1) in (B). *p < 0.05, **p < 0.01, ***p < 0.001; ns, not significant

Furthermore, the OFT was used to assess locomotor activity. Compared with WT mice, AAV-Control mice exhibited reduced movement distances, whereas AAV-AP2A1 mice presented increased distances, which were diminished upon CID1067700 treatment (Fig. 6G). Moreover, compared with WT mice, AAV-Control mice spent significantly less time in the center, but AAV-AP2A1 mice spent more time in the center, which was reduced again with CID1067700 treatment (Fig. 6H). The trajectory plots further highlighted these behavioral differences (Fig. 6I).

Compared with WT mice, the ORT results revealed that AAV-Control mice had a significantly lower preference for novel objects, whereas AAV-AP2A1 mice presented a greater preference, although the difference was not statistically significant (Fig. 6J). However, the preference significantly decreased with CID1067700 treatment (Fig. 6J). Finally, the Y-maze spontaneous alternation test revealed that AAV-Control mice had significantly lower spontaneous alternation rates than WT mice did (Fig. 6K). Although the performance of the AAV-AP2A1 mice improved performance, the difference did not reach statistical significance (Fig. 6K). Nevertheless, this decrease was abrogated by CID1067700 treatment in AAV-AP2A1 mice (Fig. 6K). Collectively, these findings suggest that AP2A1 enhances spatial learning, memory, locomotor activity, and exploratory behavior in APP/PS1 mice by activating Rab7.

Discussion

Autophagy involves four stages: initiation, autophagosome formation, maturation, and fusion with lysosomes for degradation [38, 39]. In the brains of patients or/and model mice, autophagy is significantly dysregulated, resulting in aberrant initiation [40, 41], impaired axonal transport leading to the abnormal accumulation of autophagosomes [42], and the abnormal acidification of autolysosomes [43], all of which promote AD pathology, especially Aβ accumulation.

Research has shown that autophagosomes accumulate in AD neuron axons [3, 42], and may produce Aβ [44, 45], indicating that merely increasing autophagy might not effectively slow AD progression. Ignoring autophagosome transport and degradation could worsen AD pathology and symptoms. To address this issue, a possible resolution lies in restoring the axonal transport of accumulated autophagosomes and promoting their fusion with lysosomes for efficient degradation. In this study, we observed impaired autophagic flux in APP cells and APP/PS1 mouse hippocampi, as indicated by increased LC3-II and SQSTM1 levels, and numerous autophagosomes in APP cells and at the swollen axons of APP/PS1 mice hippocampal neurons. These findings coexisted with downregulated AP2A1 and Rab7-GTP protein expression. Consistent with this, Aβ oligomer-treated primary hippocampal neurons also showed impaired autophagic flux (elevated LC3-II and SQSTM1) and decreased AP2A1 and Rab7-GTP proteins, mirroring AD model phenotypes. Although chronic APP β C-terminal fragments (β-CTFs) accumulation in familial AD mutant human neurons exacerbates endosomal dysfunction [46], our results found that acute Aβ exposure alone disrupted AP2A1-Rab7 signaling and normal autophagy in primary hippocampal neurons, suggesting Aβ directly impairs neuronal autophagy. Future studies comparing the acute and chronic Aβ effects are warranted. These findings suggest a possible link between autophagosome accumulation and the AP2A1-Rab7 signaling pathway in AD model neurons, which merits further investigation.

The adaptor protein 2 (AP2) complex, which is composed of four subunits, AP2A1, AP2B1, AP2M1 and AP2S1 (α, β, µ, and σ), actively participates in intracellular vesicle trafficking processes such as clathrin-mediated endocytosis and endosome‒autophagosome fusion [27, 47]. Among the AP2 subunits, AP2M1 and AP2A1 are commonly studied. In particular, LC3 can bind to both AP2M1 and AP2A1. The unique LIR motif in AP2A1 facilitates its binding to LC3, whereas mutation of this motif directly impairs the ability of AP2A1 to interact with LC3 [27]. Silencing AP2A1 in N2a/APP cells increases amyloid precursor protein-C-terminal fragment (APP-CTF) and Aβ levels [27], indicating that AP2A1 promotes autophagy and reduces Aβ pathology in AD cells. Consistent with this prior work, our results revealed that AP2A1 overexpression in N2a/APP cells significantly reduces autophagosome accumulation and increases autolysosome numbers, promoting lysosomal fusion and SQSTM1 degradation and reducing Aβ secretion. However, our data reveal a key mechanistic distinction: AP2A1 overexpression does not significantly alter APP processing intermediates (sAPPα, βCTF) (Fig. S7A-B), indicating that its Aβ-lowering effect is independent of APP proteolytic pathways but instead depends on autophagic clearance. The binding of AP2A1 to LC3 was shown to simultaneously interact with the p150^Glued subunit of the dynein cofactor dynactin. This interaction facilitates retrograde axonal transport of autophagosomes, contributing to neural complexity and protection against neurodegeneration [29]. Further research suggested that AP2A1 promote BACE1 retrograde axonal transport, influencing Aβ production [30]. While this highlights AP2A1’s role in modulating Aβ generation, our study reveals a distinct yet complementary function: AP2A1 overexpression enhances retrograde autophagosome transport (mCherry-LC3, Fig. 1F-G), linking it to Aβ autophagic clearance rather than production. The Rab7-dependent reversal of both autophagic flux (Fig. 4) and cognitive deficits (Fig. 6) supports this mechanism. While AP2A1 may coordinate multiple transport pathways, our data prioritize its role in Aβ clearance via autophagy.

The crucial role of Rab7 in the maturation of autophagosomes and their fusion with lysosomes has been extensively investigated [48], involving modifications such as changes in its activity [23, 49] and succinylation [50]. Moreover, Rab7 facilitates anterograde microtubule transport by cooperating with FYCO1 and kinesin in intracellular vesicle transport [51], and mediate retrograde axonal transport by collaborating with RILP and dynein in neurons [20]. Notably, RILP exclusively interacts with active Rab7 (Rab7-GTP) [52], highlighting the pivotal role of Rab7 activation in regulating the retrograde axonal transport of autophagosomes in neurons. Studies have identified GEF-binding effectors of Rab7, such as CCZ1-MON1A [23] and Vps39 [53], which facilitate the conversion of inactive Rab7-GDP to active Rab7-GTP. GAP-binding effectors of Rab7, such as TBC1D15 [54], TBC1D5 [55] and TBC1D2A [56], promote the transition from active Rab7-GTP to inactive Rab7-GDP. Our findings identify AP2A1 as a functional interactor of Rab7, correlating with elevated Rab7-GTP levels in N2a/APP cells.​​ The observed competition with TBC1D5 at Rab7 residues (ALA-43/ASP-44/THR-47/ARG-79) provides a plausible mechanism for Rab7-GTP elevation. ​However, current methodology (co-IP/colocalization) cannot distinguish whether AP2A1 directly activates Rab7 or stabilizes its GTP-bound state.​​ While TBC1D5’s GAP activity [35], supports this hypothesis, ​direct evidence of AP2A1-mediated inhibition of Rab7 inactivation requires further validation. Additionally, AP2A1 overexpression facilitated Rab7-GTP recruitment to RILP and DIC1, indicating that AP2A1-mediated Rab7 activation may be crucial for the retrograde axonal transport of autophagosomes.

CID1067700 is a specific high-affinity small molecule inhibitor of Rab7 that inhibits Rab7 activity by competing with GTP for binding [57, 58]. Previous studies have demonstrated that CID1067700 effectively inhibits Rab7 recruitment to lysosomes, thereby worsening autophagic flow impairment [59]. In the present study, CID1067700 was used to inhibit Rab7 activity to investigate whether the ability of AP2A1 to reduce the number of autophagosomes in N2a/APP cells and in the axons of Aβ-treated primary neurons, thereby promoting autophagy, is mediated by activating Rab7. Our results revealed that the activation of Rab7 by AP2A1 is significantly attenuated in the presence of CID1067700, thereby impairing DIC1 and RILP recruitment by Rab7-GTP facilitated through AP2A1. Additionally, the colocalization of LC3 and DIC1/2 increased by AP2A1 was significantly decreased by CID1067700. Furthermore, in Aβ-treated primary hippocampal neurons, AP2A1 facilitated the retrograde transport of mCherry‒LC3 puncta along axons, which was also suppressed by CID1067700 treatment. These findings demonstrated the ability of AP2A1 to reinstate the retrograde axonal transport of autophagosomes in AD model cells through Rab7 activation.

Previous studies have revealed that autophagosomes in neurons are preferentially generated from axon terminals and undergo retrograde transport toward the soma for lysosomal fusion—a process critical for maintaining autophagic flux [9, 60]. While lysosomes exhibit bidirectional transport capabilities regulated by motor proteins [61], their sparse axonal localization suggests that efficient autophagosome delivery to somatic lysosomes is paramount [60]. In this study, we focused on autophagosome transport dynamics in AD pathology, demonstrating that AP2A1 specifically enhances retrograde autophagosome transport in AD neurons (Fig. 1F-G), facilitating their fusion with lysosomes (Fig. 1H-K) rather than regulating lysosomal function (Fig. S4 G-J). Overexpression of AP2A1 in N2a/APP cells and APP/PS1 mouse hippocampi resulted in a decrease in LC3-II and SQSTM1 expression, reduced accumulation of autophagosomes, and increased fusion between autophagosomes and lysosomes in N2a/APP cells. Notably, inhibiting Rab7 activity with CID1067700 significantly attenuated the effect mediated by AP2A1. These findings suggest that AP2A1 promotes the retrograde axonal transport of autophagosomes by activating Rab7, facilitating efficient degradation processes during autophagy in AD model cells.

Autophagy is an intracellular degradation process that maintains cellular homeostasis by eliminating waste and damaged organelles through the lysosomal system. In AD patients, impaired autophagy leads to Aβ accumulation, a key pathological hallmark of AD [5, 62]. Specifically, aberrant autophagosome formation and dysfunction may contribute to Aβ build-up, exacerbating AD pathology, including microglial and astrocytic activation, which leads to inflammatory responses that further drive disease progression [62, 63]. Our findings revealed that injecting AP2A1 into the hippocampi of APP/PS1 mice significantly decreased the positive immunostaining for activated microglia (IBA1) and astrocytes (GFAP), an effect attenuated by CID1067700. Golgi-Cox staining analysis revealed a significant increase in the number of dendritic spines in the hippocampal neurons of APP/PS1 mice after AP2A1 injection, which was attenuated by CID1067700 coadministration. Overexpressing AP2A1 in primary hippocampal neurons treated with Aβ significantly increased the expression of synaptophysin (presynaptic marker) and PSD95 (postsynaptic marker), both of which were reduced upon treatment with CID1067700. These results suggest that AP2A1 reduces neuroinflammation and increases neuronal complexity by activating Rab7 in the context of AD. Similarly, we observed that AP2A1 overexpression significantly reduced Aβ deposition in the hippocampus of APP/PS1 mice and the secretion of Aβ₁₋₄₂ and Aβ₁₋₄₀ in N2a/APP cells, both of which were restored by CID1067700 treatment. These findings further underscore the neuroprotective effect achieved through Rab7 activation by AP2A1.

Finally, behavioral tests revealed that AP2A1 activation of Rab7 enhanced cognitive function (MWM, ORT, and Y-maze) and motor skills (OFT) in AD model mice, suggesting that AP2A1 inhibited AD pathological progression by activating Rab7. While our data demonstrate that AP2A1 mediates Rab7 activation to restore autophagy (Fig. 4), reduce Aβ burden (Fig. 5H, I), and improves cognition (Fig. 6), which strengthens the classical theory that improving autophagy can reduce Aβ levels to suppress AD pathologic progression [64–66], the temporal-causal relationship between these events remains unresolved. Specifically, it is unclear whether Rab7-driven autophagic recovery directly rescues synaptic/cognitive deficits independent of Aβ clearance, or exerts neuroprotection primarily through Aβ degradation. Future studies employing time-course analyses of autophagic reactivation versus Aβ reduction post-AP2A1 treatment, combined with autophagic flux inhibitors (e.g., chloroquine), could delineate this mechanism.

It is important to acknowledge that the absence of WT mice overexpressing AP2A1 limits our ability to exclude non-specific neuromodulatory effects. For instance, the increased exploratory behavior of AP2A1-treated APP/PS1 mice in the OFT could reflect both reduced anxiety and enhanced motor motivation, potentially confounding spatial learning assessment in the MWM. However, the complete reversal of behavioral improvements (Fig. 6) and Aβ reduction (Fig. 5H, I) by CID1067700 strongly supports Rab7-dependent AD pathology mitigation as the primary mechanism. Future studies should assess AP2A1 in WT mice and employ anxiety-specific behavioral tests to disentangle these effects. In addition, it should be noted that AAV-mediated AP2A1 overexpression, while effective, may not fully replicate endogenous AP2A1 regulation. Although Rab7-GTP inhibition reversed these effects, potential off-target impacts of viral delivery or non-physiological AP2A1 levels remain possible. Future studies using conditional AP2A1 activation or pharmacological Rab7 modulators will help clarify these observations.

In conclusion, our research explored the interaction between AP2A1 and Rab7 and revealed that AP2A1 facilitates the retrograde axonal transport of autophagosomes, thereby restoring autophagy in AD and mitigating its pathological progression through the activation of Rab7 (Fig. 7). While this study provides key insights into autophagy dysfunction caused by axonal transport defects in AD neurons, further studies are needed to clarify the mechanisms by which AP2A1 activates Rab7 and its role in axonal transport across various pathological models. Ultimately, this study underscores the pivotal role of AP2A1 in modulating neuronal autophagy and supports further investigation into its therapeutic relevance for AD-related pathologies.

Fig. 7.

Schematic representation of aberrant autophagy in AD model neurons. Compared with normal neurons (left), massive accumulation of autophagosomes in AD model neurons (right) axons alongside downregulation of AP2A1 and Rab7-GTP expression, reveals that AP2A1 overexpression activates Rab7, recruits retrograde transport proteins and promotes the retrograde axonal transport of autophagosome to restore autophagy, thereby inhibiting Aβ-related pathological progression in AD

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

YYW: experimental work, data analysis, manuscript writing and manuscript revision; SYL: experimental methods, experimental work, and manuscript revision; XL: experimental work and data analysis; JNF and SJL: data analysis; FLZ, XJL, MML and DMF: manuscript revision; YL: project design, funding acquisition and manuscript revision.

Funding

This research was supported by the National Natural Science Foundation of China (82371440), the Nature Science Foundation of Chongqing (CSTB2023NSCQ-MSX0797) and the Chongqing Traditional Chinese Medicine Hospital project of Li Yu (CQSZYY-002).

Data availability

Data are available from the corresponding author on reasonable request.

Declarations

Ethics approval

All mice experiment in this study were approved by the Ethics Committee of Chongqing University Cancer Hospital (approval number: CQCH-LAE-20231020012).

Competing interests

The authors declare no competing interests.