Transplantation of hiPSC-derived pericytes rescues Alzheimer’s disease phenotypes in APOE4/4 mice through IGF2-rich apoptotic vesicles

Female C57BL/6 (wild-type, WT) mice aged 8–12 weeks (weight 18–25 g) were purchased from GemPharmatech Co., Ltd (Nanjing, China). Homozygous APOE4/4 mice were from Cyagen Biosciences Inc. (Suzhou, China) and bred under standard specific-pathogen-free (SPF) conditions. All mice were housed in a temperature-, humidity- and light cycle-controlled facility (20 ± 2 °C; 50% ± 10%; 12-h light/dark cycle) with free access to food and water. For the single injection project, aged (18 months old) APOE4/4 mice were intravenously injected with 1 × 10⁶APOE3/3-PCs, human dermal fibroblasts (HDFs; Cat. #2320, ScienCell Research Laboratories, Carlsbad, CA) or an equal volume of phosphate buffered saline (PBS). For the multiple injection project, mice of 2–3 months old were intravenously injected with 1 × 10⁶APOE3/3-PCs, HDFs or an equal volume of PBS once a month for 8 consecutive months. For apoptotic vesicle (ApoV) treatment, APOE4/4 mice aged 2–3 months were intravenously injected with 2 × 10⁷ ApoVs^−PCs, ApoVs^−HDFs or an equal volume of PBS every 21 days for 10 times. All animal experimental procedures were approved by the Sun Yat-Sen University Animal Use and Care Committee (2020-000305).

Peripheral blood mononuclear cells (PBMCs) from APOE3/3 or APOE4/4 carriers were obtained at the Third Affiliated Hospital of Sun Yat-Sen University following the standards of Ethics Committee of the hospital (Approval number: 2020-02-148-01) and the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. APOE3/3 and APOE4/4 hiPSCs were generated by transduction of human PBMCs with Sendai viral vectors (Cat. #A16517, Thermo Fisher Scientific, Rutherford, NJ) according to the manufacturer's instructions [19]. The hiPSCs were maintained at feeder-free conditions using mTeSR™ Plus medium (Cat. #100–0274, Stem Cell Technologies, Vancouver, Canada) on pre-coated Matrigel (Cat. #354277, BD Bioscience, San Diego, CA). Cells were passaged every 4–5 days with 0.5 mmol/L EDTA (Cat. # 15575020, Invitrogen, Carlsbad, CA). APOE genotyping was performed using a PCR-restriction fragment length polymorphism approach [20]. The APOE3/3 and APOE4/4 genotypes were confirmed by Sanger sequencing. Information of APOE3/3 and APOE4/4 carriers is listed in Table S1.

Pericyte-like cells were differentiated as described in our previous study [18]. In brief, hiPSCs were harvested using Accutase (Cat. # 00-4555-56, Invitrogen) and plated onto a Matrigel-coated T75 cell culture flask at a density of 10⁴ cells/cm². A chemically defined N2B27 medium (N2B27-CDM) containing 95% DMEM/F12 medium, 1% N2 supplement, 2% B27 supplement, 1% Glutamax, 1% MEM nonessential amino acids, 55 μmol/L 2-mercaptoethanol (all from Invitrogen), 10 μmol/L Y27632 (Cat. #72304, Stem Cell Technologies), and 20 ng/ml basic fibroblast growth factor (bFGF) (Cat. #100-18C, Peprotech, Rocky Hill, NJ) was used. Twenty-four hours later, the N2B27-CDM medium was replaced with NCN2 medium containing N2 supplement, CHIR99021 (1 μmol/L), and SB431542 (2.0 μmol/L) and cultured for 6 days. Then the cells were dissociated to single cells using Accutase and labeled with antibodies for CD57 (Cat. #558619, BD Bioscience) and low-affinity nerve growth factor receptor p75 (Cat. #560326, BD Bioscience) for fluorescence-activated cell sorting (FACS). p75^brightHNK1⁺ CNC stem cells were sorted and replated onto poly-L-ornithine (PO; 15 mg/ml, Cat. #P4957, Sigma-Aldrich, St. Louis, MO)- and fibronectin (FN; 10 mg/ml, Cat. #FC010, Millipore, Temecula, CA)-coated dishes for adherent culture with neural crest culture medium (NCCM) containing 1% N2 supplement, 2% B27 supplement, 20 ng/ml bFGF, and 20 ng/ml epidermal growth factor (EGF; Cat. #AF-100-15, Peprotech). CNCs were dissociated using Accutase and plated onto the PO/FN-coated dishes at a density of 10⁵ cells/cm² in NCCM supplemented with 10 μmol/L Y27632. When differentiation was initiated, the culture was switched to commercial Pericyte Medium (Cat. #1201, ScienCell Research Laboratories) containing 10 ng/ml bFGF and 50 ng/ml platelet derived growth factor BB (Cat. #100-14B, Peprotech) for 7–14 days. The first confluent culture at this time point was denoted as passage 1. Cells were dissociated by Accutase solution, split at 1:3 in Pericyte Medium, and cultured on PO/FN-coated plates.

Animals were moved to the testing room at least 3 h prior to the test. OFT was performed in a cubic open field chamber (50 cm × 50 cm × 50 cm), with bottom area divided into a central field (25 cm × 25 cm) and a peripheral field. Mouse behaviors were captured using a SONY HDRCX405 video camera for 10 min and analyzed by a behavioral analysis software (TopScanLite, Clever System, Reston, VA). The total distance, center distance, number of entries in the center and the time spent in the central field were recorded for analysis.

MWM was conducted in a circular tank (diameter, 120 cm) filled with opaque water at a temperature of 24 °C. Four visual cues with different shapes were placed around the tank to assist recognition of orientation. The water was made opaque with a non-toxic chemical titanium dioxide (food-grade titanium dioxide). The platform was submerged ~ 1.0 cm below the water surface. During the training phase, mice were trained once a day for 5 consecutive days. During training, each mouse was given 60 s to find the platform. If it failed to find the platform within 60 s, it would be guided to the platform and stayed there for 15 s. In the spatial probing test on day 6, the platform was removed and each mouse was allowed to swim for 60 s. The time and the number of crossings in the platform quadrant were recorded. Mouse behavior was recorded by a SONY HDRCX405 video camera and analyzed using the TopScanLite software (TopScanLite, Clever System, Reston, VA).

NOR was conducted in a cubic open field chamber (50 cm × 50 cm × 50 cm). Briefly, mice were placed in the chamber and allowed for free exploration for 10 min on day 1 (habituation period). Twenty-four hours later, two familiar objects were placed in the chamber and mice were allowed to explore for 10 min (learning period). After another 24-h interval, one familiar (F) object was replaced by a novel object (N), and mice were allowed to explore the two objects for 10 min (testing period). The time spent exploring the familiar and novel objects and animal movement were recorded using the TopScanLite software. Recognition index of exploration was calculated as follows: recognition index = time spent on the new object/total exploration time of the two objects.

Spontaneous alternation T-maze test was used to assess the spatial working memory of mice. The T-maze was a T-shaped apparatus featuring a start arm and two goal arms. Each arm was endowed with a guillotine door. Each mouse was transferred to the testing room at least 3 h prior to the test. During each trial, the animal was first placed in the start arm for 30 s. Then the guillotine door of the start arm was opened. When the mouse entered a goal arm, the guillotine door behind it was closed. Then, the mouse was placed back to the start arm again for 30 s, and then the guillotine door of the start arm was opened to allow the mouse to make a choice between the two goal arms. Trials were marked as successful if the mouse chose different goal arms on each run. Alternation rate was defined as the total proportion of successful trials for each animal during 10 consecutive trials.

Early passage of PCs was transduced with hU6-MCS-Ubiquitin-tdTomato-IRES-puromycin control lentivirus (GeneChem, Shanghai, China). Forty-eight hours later, puromycin was used to select stable tdTomato⁺ cells for cell tracing in vivo. mCherry⁺EGFP⁺ lentivirus was kindly provided by Xiaoran Zhang's lab (Zhongshan School of Medicine, Sun Yat-sen University) synthesized according to the methods published previously [21]. Briefly, the sLP-mCherry sequence and the EGFP sequence were cloned into a pRRL lentiviral backbone. A soluble peptide and a modified TAT peptide were cloned upstream of the mCherry cDNA (sLP-mCherry), in which sLP–mCherry is used to label phospholipid bilayer and EGFP represents cell contents. PCs were stably infected with lentiviral particles. mCherry⁺EGFP⁺ PCs were used for tracing PC-released EVs in vivo.

Multiphoton microscopy was performed to detect BBB leakage in vivo. In brief, mice were anesthetized with isoflurane and a cranial window (~ 0.5 cm in diameter) was made. Fluorescein isothiocyanate (FITC)-conjugated dextran (Cat. #46944, Sigma-Aldrich) was injected via tail vein (1.5 mg per mouse). In vivo time-lapse images were acquired for a total of 30 min using an FVMPE-RS confocal microscope (Olympus, Tokyo, Japan). The fluorescence intensity of dextran was quantified by Image J software (https://imagej.nih.gov/ij/↗). To observe ApoV release from the lung of APOE4/4 mice, fresh dissected lung was placed in a 6-well plate and perfused with PBS. ApoVs released from lung were observed under the FVMPE-RS confocal microscope, and time-lapse images were acquired every 5 min for 1 h.

ApoVs were induced as previously reported [22]. When PCs or HDFs reached full confluency, the culture medium was removed and replaced with a basal medium containing 500 nmol/L staurosporine (STS) (Cat. #SB17469, Macklin, Shanghai, China) without FBS. After induction of apoptosis for 12 h, ApoV-containing supernatant was isolated and ApoVs were purified using sequential centrifugation. Briefly, cells and cell debris were removed by 800 × g centrifugation for 10 min and 2000 × g centrifugation for another 10 min. Next, the supernatant was centrifuged at 15,000 × g for 30 min to pellet the ApoVs. Cleaved-caspase-3 assay and flow cytometry analysis with annexin V and PI staining were used to evaluate the apoptotic rate of cells. The morphology of ApoVs was observed with transmission electron microscopy (TEM) (HT7800, Hitachi High-Technologies Corporation, Tokyo, Japan). ApoV surface markers CD9, CD63, and CD81 were assessed by flow cytometry.

To test the barrier function of PCs and the treatment effect of ApoVs, PCs were cocultured with primary human brain microvascular endothelial cells (HBMECs) in 24-well transwell inserts (Cat. #3413, Corning Life Sciences, Topsfield, MA) pre-coated with collagen (1 μg/ml) and FN (10 μg/ml) for at least 4 h at 37 °C. PCs (10,000 cells per insert) and HBMECs (20,000 cells per insert) were seeded on the top side of the inserts and cocultured for 48 h. For ApoVs treatment, 2 × 10⁵ ApoVs derived from APOE3/3-PCs or HDFs were added to the upper chamber of the transwell.

To assess permeability, 4-kD dextran labeled with FITC was added to the upper chamber of the transwell and placed on a shaker in an incubator. Two hours later, medium in the bottom well was collected for analysis. For Aβ_1-40 transcytosis assay, FITC-labeled Aβ_1-40 was added into the lower chamber. The medium in the upper compartment was collected after 24 h. Fluorescent intensity was then analyzed with a fluorometric plate reader (Infinite F200 Pro, Tecan Life Sciences, Männedorf, Switzerland) at 488 nm excitation and 525 ± 20 nm emission.

For IF analysis, mouse brain samples were fixed in 4% paraformaldehyde (PFA) overnight at 4 °C, washed with PBS and then immersed in 30% sucrose solution overnight at 4 °C. Samples were embedded in optimal cutting temperature compound (OCT, Cat. #4583, Sakura Finetek, Torrance, CA) and cut into 35 μm-thick sections. The sections were blocked in blocking buffer (5% donkey serum in 0.3% Triton X-100 in PBS) for 1 h at room temperature, incubated with primary antibodies overnight at 4 °C and then secondary antibodies for 2–4 h at room temperature. For cell tracing experiments, brain samples were sliced at 50 μm thickness to better trace and identify the complete cellular structures. For in vitro samples, cells were fixed with 4% PFA at room temperature for 20 min and washed three times with PBS. Then, cells were permeabilized with 0.3% Triton X-100 in PBS and incubated overnight at 4 °C with primary antibodies and with secondary antibodies for 1–2 h at room temperature.

The following antibodies were used: NeuN (Abcam, Cat. #ab177487, 1:200, Cambridge, UK), Iba1 (Genetex, Cat. #GTX100042, 1:100, Irvine, CA), Fibrinogen (Abcam, Cat. #ab43269, 1:100), Albumin (Abcam, Cat. #ab207327, 1:100), ZO1 (Thermo Fisher Scientific, Cat. #40-2200, 1:100, Waltham, MA), Occludin (Thermo Fisher Scientific, Cat. #71-1500, 1:100, Waltham, MA), Lectin (Vector, Cat. #DL-1178-1, 1:100, San Ramon, CA), Pfgfrβ (Cell Signaling Technology, Cat. #3169, 1:100, Danvers, MA), cleaved-caspase3 (Abcam, Cat. #ab2302, 1:100), and CD206 (Biolegend, Cat. #141708, 1:50, San Diego, CA). DAPI (Roche, Cat. #1023627001, Basel, Switzerland) was used as a counterstain for the nucleus.

Images of NeuN/Iba1 staining, fibrinogen/albumin leakage, pericyte coverage and ZO1/Occludin coverage were reconstructed as maximum projections of 10-μm-thick Z-stack images using Imaris Viewer (Oxford, Oxford, UK), and subsequent analyses were performed using ImageJ software (NIH, Bethesda, MA). For each animal, 5 sections corresponding to the plane of maximum coronal surface area were collected. For quantification, five randomly selected views of either the cortex or hippocampus were captured from each section, and the statistical values were derived based on the average of 25 views (5 sections × 5 images per section). The experiments were repeated three times to ensure reproducibility.

For IHC analysis, mouse brain samples were embedded in paraffin and cut into 5 µm-thick coronal sections. Tissue sections were incubated overnight with primary antibodies: anti-Aβ40 Rabbit pAb (Servicebio, Cat. #GB111197-100, 1:100, Wuhan, China), and phospho-Tau-T181 Rabbit mAb (ABclonal, Cat. #AP1387, 1:100). Staining was revealed using biotinylated secondary antibodies (Servicebio, Cat. #GB1302) and the ABC-HRP kit (Cat. #PK-4000, Vector, San Ramon, CA). Aβ burden (refers to the proportion of Aβ-positive area to the captured area) and the number of p-tau-positive cells per unit area (mm²) were quantified. For each animal, five coronal sections were collected. For each section, five images were randomly acquired from either the cortex or the hippocampus. Thus, for each animal, a total of 25 images (5 sections × 5 images per section) were analyzed, and statistical values were calculated based on the average across these 25 images.

Flow cytometry was performed using the CytoFLEX flow cytometer, and results were analyzed with the CytExpert (Beckman, Brea, CA). The following antibodies were used: FITC anti-human CD9 antibody (1:100; Cat. #312103, Biolegend), PE/Cyanine7 anti-human CD63 antibody (1:100; Cat. #353009, Biolegend), APC anti-human CD81 antibody (1:100; Cat. #349509, Biolegend), and BD Pharmingen™ FITC Annexin V Apoptosis Detection Kit I (Cat. #556547, BD Biosciences).

RNA-seq was performed as previously reported [18]. Total mRNA was isolated from ApoVs^−HDFs or ApoVs^−PCs for bulk RNA-seq analysis. RNA-seq libraries were constructed using the Illumina mRNA-seq Prep Kit (Cat. #20020594, Illumina, San Diego, CA) according to the instructions of the manufacturer. The fragmented and randomly primed 150 bp paired-end libraries were sequenced using Illumina Novaseq 6000 (Illumina). Sequencing data were processed using Consensus Assessment of Sequence and Variation (CASAVA, version 1.8.2; Illumina) using the default settings. The TPM values were used to evaluate the expression levels of genes. Pearson's correlation coefficients were calculated (R²) to measure the similarities of the global gene expression profiles between ApoVs^−HDFs and ApoVs^−PCs. Finally, the RNA-Seq data were analyzed using the Ingenuity Pathways Analysis software (Ingenuity Systems Inc., Redwood City, CA) to categorize the differentially regulated genes. RNA-seq data have been deposited at the NGDC database (https://ngdc.cncb.ac.cn/gsa-human/↗) under accession number HRA009346.

4D label-free quantitative proteomics analysis was supported by Jingjie PTM BioLabs (Hangzhou, China), including protein extraction, trypsin digestion, HPLC fractionation and LC-MS/MS analysis. Proteins were identified by comparing against the Uniprot database with a false discovery rate (FDR) set at 0.01 for both peptides and proteins. Proteins with differential expression between ApoVs^−HDFs and ApoVs^−PCs were included for further functional analysis. The details of all the identified proteins are listed in Additional file 2, Table S2.

Gene knockout was performed using the CRISPR–Cas9 system. pLenti-IGF2-sgRNA (Cat. #L27870) and pLenti-Control-sgRNA (Cat. #L00011) were purchased from Beyotime Biotechnology (Shanghai, China).

To improve transfection efficiency of the cells, the plasmid together with packaging plasmids (psPAX2 and pMD2.G) were co-transfected into HEK293FT cells for lentivirus packaging at a ratio of 3:2:1 using Lipofectamine 3000 transfection reagent (Cat. #L3000008, Thermo Fisher Scientific, Waltham, MA) according to the manufacturer's protocol. Briefly, the HEK293FT cells were seeded in complete DMEM medium (Cat. #C11995500BT, Gibco, Grand Island, NY) to reach 70%–80% confluence at the time of transfection. The plasmid DNA was diluted in Opti-MEM medium and mixed with the Lipofectamine 3000 reagent. Subsequently, the DNA mixture was added dropwise to the cell culture medium. The virus-containing supernatant was collected at 48 and 72 h post-transfection, filtered through a 0.22-μm PVDF membrane (Cat. #SLGV033RB, Millipore, Temecula, CA) and enriched by centrifuging at ~ 50,000 × g for 2 h. Finally, early passage of APOE3/3-PCs were infected, and the knockout efficiency of each sgRNA-containing lentivirus was assessed by western blot.

Total RNA was extracted from ApoVs^−HDFs and ApoVs^−PCs using TRIzol Reagent (Invitrogen) according to the manufacturer's instructions. RNA concentrations were determined using the NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE). Total RNA (1 μg) was converted to cDNA using a Quantitect Reverse Transcription kit (Qiagen, Valencia, CA). Quantitative real-time PCR (qPCR) analysis was performed using the DyNAmo ColorFlash SYBR Green qPCR kit (Cat. #F416-L, Thermo Fisher Scientific, Waltham, MA) and the LightCycler 480 Detection System (Roche Diagnostics, Branchburg, NJ). The expression levels were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and changes in gene expression were calculated as fold changes using the ΔΔCt method. Primer used in this study were: IGF2 forward CGCTTCAGTTTGTCTGTTCG, reverse GCAGCACTCTTCCACGATG; GAPDH: forward GAACATCATCCCTGCATCCA, reverse CCAGTGAGCTTCCCGTTCA.

For western blot analysis, half mouse brain or cell samples were washed with cold PBS, lysed in 1 × RIPA buffer and then centrifuged at 15,000 × g for 10 min at 4 °C. Protein concentration was measured using the Pierce BCA Protein Assay Kit (Cat. #23225, ThermoFisher Scientific, Waltham, MA). Equal amounts of protein were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis and then electrotransferred to 0.45-μm pore-sized polyvinylidene difluoride membranes (Cat. #HVLP04700, Millipore, Burlington, MA). Each membrane was blocked in 0.1% (v/v) Tween 20/TBS (TBS/T) containing 5% (w/v) nonfat milk powder for 1 h at room temperature, incubated with appropriate primary antibodies overnight at 4 °C and then with peroxidase-coupled secondary antibodies. Proteins were detected with ECL substrate Kit (Cat. #JY01054, Jiangyuan Bio, Nanjing, China). All Western blot experiments were performed in triplicate. The primary antibodies were anti-Aβ₄₀ (Servicebio, Cat. #GB111197-100, 1:1000), Aβ1-42 Rabbit mAb (ABclonal, Cat. #A24422, 1:1000), phospho-tau-T181 rabbit mAb (ABclonal, Cat. #AP1387, 1:1000, Wuhan, China), and IGF2 (Signalway Antibody, Cat. #32592-2, 1:1000, Frederick, MA). The secondary antibody was Goat Anti-Rabbit IgG H&L (HRP) (Abcam, Cat. #ab205718, 1:5000, Cambridge, UK).

WT or APOE4/4 mouse brain tissues were homogenized in cold PBS containing a protease inhibitor cocktail and centrifuged at 15,000 × g for 20 min at 4 °C. Aβ₄₀ levels were analyzed using a mouse Aβ₄₀ ELISA kit (Cat. #ML00185996T, MIbio, Shanghai, China) according to the manufacturer's instructions.

GraphPad Prism 8.0 (GraphPad Software) was used for statistical analysis. All data are presented as the mean ± SD. For two-group comparisons, significance was assessed by Student's t test. For multiple group comparisons, significance was assessed by one-way ANOVA with Tukey's multiple comparison test. P < 0.05 was considered statistically significant.

APOE3/3 hiPSCs were generated from PBMCs of two healthy donors using Sendai viral vectors and then induced to differentiate into pericytes (APOE3/3-PCs) of neural crest origin as previously reported [18, 19]. To investigate the potential therapeutic benefit of pericyte transplantation, a total of 1 × 10⁶ HDFs, APOE3/3-PCs, or an equal volume of PBS was intravenously injected to 18-month-old APOE4 homozygote mice. One month later, OFT and NOR were performed to evaluate cognition (Fig. S1a). However, no significant change was observed among the PBS, HDFs and PCs groups, with all APOE4/4 mice showing poor performance compared to age-matched WT mice (Fig. S1b–h). These findings suggest that a single transplantation of PCs may not rescue the cognitive decline in aged APOE4/4 mice, possibly because the pathological damage had already occurred and cannot be apparently rescued.

Altogether, these results demonstrate that early and multiple APOE3/3-PC transplantation improves cognition in aged APOE4/4 mice.

Neuron loss is a prominent pathological feature of AD, caused by excessive deposition of Aβ and p-tau [23]. We measured NeuN⁺ neuronal cells among the four groups, and found that the numbers of NeuN⁺ cells in the cortex and CA1 area were comparable between the HDFs group and the PBS group (Fig. 2g, h). In contrast, PCs transplantation significantly increased the number of NeuN⁺ cells compared with PBS or HDFs (Fig. 2g, h). Microglia, the innate immune cells of the central nervous system (CNS), play a multifaceted role in AD, contributing to neuroinflammation and neurodegeneration [24]. Increased number and activation of microglia have been observed in the brains of AD patients or mouse models [25]. Remarkably, our results revealed that transplantation of PCs not only significantly reduced the number of Iba1⁺ microglial cells, but also shifted the morphology of microglia toward the homeostatic form compared with PBS or HDFs (Fig. S2e, f).

In summary, these results demonstrated that early and multiple transplantation of APOE3/3-PCs can reduce Aβ deposition, p-tau accumulation, neuron loss and microglial activation in aged APOE4/4 mice.

In our previous study, we demonstrated that intravenous injection of hiPSCs-CNC PCs could efficiently restore BBB properties in the tMCAO mouse model [18]. Thus, we evaluated the changes of BBB permeability in APOE4/4 mice after APOE3/3-PC transplantation.

Compared to WT mice, APOE4/4 mice exhibited a moderate loss of capillary density in the cortex and hippocampus, which was well preserved following PC transplantation (Fig. S3a, b). Next, we evaluated the integrity of BBB using antibodies against tight junction proteins ZO1 and Occludin, which are associated with BBB breakdown in APOE4/4 mice [27, 28]. Immunofluorescence staining indicated a significant loss of ZO1 and Occludin in the PBS group and no obvious improvement after HDFs transplantation, while PCs drastically protected ZO1 and Occludin from degradation (Fig. 3f, g). Given the role of perivascular pericytes in the maintenance of the BBB, we asked whether PC transplantation could prevent pericyte loss in APOE4/4 mice. To our surprise, mice in the PC and WT groups displayed similar percentages of pericyte coverage and comparable numbers of Pdgfrβ⁺ pericytes, while dramatic loss of pericyte coverage and decreased cell number were detected in the PBS and HDF groups (Fig. 3h, i). Furthermore, we observed a significant reduction in CD206⁺ perivascular macrophages in APOE4/4 mice, which was partially rescued by PC transplantation (Fig. S3c, d). GFAP staining revealed a significant decrease in GFAP⁺ endfeet coverage on blood vessels in the PBS and HDF groups, while PC transplantation partially increased the GFAP⁺ endfeet coverage (Fig. S3e, f). Besides, the number of GFAP⁺ astrocytes in the cortical region was significantly increased in the PBS and HDF groups, while a partial decrease was observed in the PCs transplantation group (Fig. S3g).

Altogether, these results demonstrated that transplantation of APOE3/3-PCs could preserve BBB integrity and reverse intrinsic pericyte loss in APOE4/4 mice.

Next, we tried to trace the source of tdTomato⁺ ApoVs. Given that the majority of transplanted pericytes accumulated in the lungs of APOE4/4 mice, we subsequently performed further analyses of the pulmonary tissue and observed that some tdTomato⁺ cells exhibited signs of lysis and vesicle formation (Fig. S5d). To further confirm that the tdTomato⁺ ApoVs were released from the lungs, we harvested lung tissues from APOE4/4 mice on day 3 after cell injection and observed real-time ApoV release ex-vivo with multiphoton microscopy. We found time-dependent release of ApoVs from mouse lungs after pericyte transplantation (Fig. S5e, f). Due to their small size (1–5 μm), ApoVs can readily pass through pulmonary capillaries (8–10 μm diameter), enter systemic circulation, and reach the mouse brain [30]. Furthermore, we asked which cell type can uptake the PCs-derived ApoVs in APOE4/4 mice brain. We found that the tdTomato⁺ ApoVs were engulfed by pericytes, neurons and microglia, but not by astrocytes (Fig. 4k and Fig. S5g).

In summary, we demonstrated that transplanted APOE3/3-PCs generate huge numbers of ApoVs in the lung or other tissues in vivo, which could circulate to the mouse brain and be engulfed by multiple cell types including pericytes in the brains of APOE4/4 mice.

We observed increased pdgfrβ⁺ pericyte number and pericyte coverage in APOE4/4 mice after PCs transplantation (Fig. 3h, i). Additionally, the transplanted APOE3/3-PCs released huge numbers of ApoVs, which were then engulfed by endogenous pericytes (Fig. 4k). Thus, we inferred that ApoVs engulfed by pericytes may prevent APOE4-related pericyte degeneration.

BBB restricts neurotoxic plasma components from entering the brain and regulates the removal of metabolic waste products from the brain to the systemic circulation, including Aβ [31, 32]. We then first evaluated BBB stability by co-culture of HBMECs with APOE3/3- or APOE4/4-PCs on transwell inserts followed by addition of FITC-Dextran in the upper well. The APOE4/4-PC group showed increased dextran leakage, while ApoVs^−PCs, rather than ApoVs derived from HDFs (ApoVs^−HDFs), significantly reduced the permeability in the APOE4/4-PCs co-culture system (Fig. 5k). Aβ transport by pericytes is one of the primary mechanisms for Aβ clearance across the brain to the blood [33]. We found weakening of the transport capacity of APOE4/4-PCs, which was significantly improved by ApoVs^−PCs treatment (Fig. 5l). Since pericytes have been shown to capture Aβ through LRP1 (low-density lipoprotein receptor related protein-1) [34, 35], PCs were co-cultured with FITC-labeled Aβ_1-40 for 48 h. We found more Aβ_1-40 attachment or swallowing by APOE3/3-PCs than by APOE4/4-PCs, and ApoVs^−PCs treatment greatly improved the capture capacity of APOE4/4-PCs for Aβ_1-40 (Fig. 5m).

To further validate that the therapeutic effect is uniquely attributable to ApoVs, we cultured APOE4/4-PCs with same quantity of exosomes or ApoVs from APOE3/3-PCs. Our results demonstrated that ApoVs, but not exosomes, could rescue APOE4/4-PCs viability and improve the pseudo-BBB integrity (Fig. S7a–d).

Altogether, we successfully generated ApoVs from APOE3/3^-PCs by STS induction and found ApoVs^−PCs treatment could promote functional recovery of APOE4/4-PCs in vitro.

We then focused on whether IGF2 is involved in the reparative effects of ApoVs^−PCs. We constructed IGF2-knockout APOE3/3-PCs using CRISPR-Cas9 system, which was confirmed by western blot analysis (Fig. 6m, n and Fig. S8a). Then, ApoVs of IGF2-knockout APOE3/3-PCs (ApoVs^{−PCs−IGF2KO}) and non-targeting control APOE3/3-PCs (ApoVs^{−PCs−IGF2NC}) were generated as described above, and cultured with APOE4/4-PCs. We found that IGF2 deletion largely abolished the therapeutic effects of ApoVs^−PCs, as exhibited by increased cell death, dextran transcytosis, and reduced Aβ transport ability (Fig. 6o–r). Besides, we observed dramatic cell death and function weakness of APOE3/3-PCs when IGF2 was knocked out (Fig. S8b–e).

Together, the ApoVs^−PCs contain higher levels of IGF2 protein and mRNA compared with ApoVs^−HDFs, which may play vital roles in the protective effect of ApoVs^−PCs.

At the end of experiment, MWM, NOR, T-maze, and OFT were performed to evaluate cognition recovery of APOE4/4 mice at 18 months old. In the WMW test, the latency to the platform was shorter in the ApoVs^{−PCs−IGF2NC} group compared with the ApoVs^−IGF2KO group on day 5 of training, with no difference between the ApoVs^−HDFs and the ApoVs^{−PCs−IGF2KO} groups (Fig. 7e, f). In the probing test, the number of crossings and the total time spent in the target quadrant were higher in the ApoVs^{−PCs−IGF2NC} group than in the ApoVs^{−PCs−IGF2KO} group, and comparable between ApoVs^−HDFs and ApoVs^{−PCs−IGF2KO} groups (Fig. 7h, i). Similarly, the ApoVs^{−PCs−IGF2NC} group showed better performance compared with the ApoVs^−IGF2KO group, while ApoVs^−HDFs and ApoVs^−IGF2KO groups showed no significant difference in the performance in T-maze test, NOR test, or the OFT (Fig. 7j–l and Fig. S9b, c). Taken together, these data indicate that transplantation of ApoVs derived from APOE3/3-PCs improved cognitive function in APOE4/4 mice, whereas IGF2 depletion largely abolished the treatment effect of ApoVs. This suggests an important role of IGF2 in the therapeutic capacity of ApoVs in APOE4/4 mice.

We further explored whether the APOE3/3-PCs-derived ApoVs can ameliorate AD pathologies and improve BBB function in APOE4/4 mice. The ApoVs^{−PCs−IGF2NC} group showed less Aβ and p-tau compared with the ApoVs^−HDFs group, while IGF2 depletion largely weakened this effect (Fig. S10a, b). In addition, the numbers of NeuN⁺ cells in the cortex and CA1 region were higher in the ApoVs^{−PCs−IGF2NC} group compared with the ApoVs^−HDFs group, while no increase of NeuN⁺ cells was observed in the ApoVs^{−PCs−IGF2KO} group (Fig. S10c, d). Next, BBB permeability was evaluated by immunostaining. We observed significant recovery of BBB integrity in ApoVs^{−PCs−IGF2NC}, but not in the ApoVs^−HDFs or ApoVs^{−PCs−IGF2KO} group (Fig. S11a, b). The level of tight junction protein ZO-1 was drastically improved in the ApoVs^{−PCs−IGF2NC} group (Fig. S11c). ApoVs^−IGF2NC also prevented pericyte loss in APOE4/4 mice (Fig. S11d).

In summary, transplantation of ApoVs generated from APOE3/3-PCs achieved significant therapeutic effects in APOE4/4 mice, but these effects were reversed by IGF2 knockout.