Adiponectin deficiency drives cerebrovascular dysfunction and synergizes with amyloid-β to exacerbate alzheimer’s pathology

Abstract

Cerebrovascular dysfunction (CVD) is increasingly recognized as a contributor to Alzheimer’s disease (AD) progression. Adiponectin (APN), an adipocyte-derived hormone with vasoprotective properties in the periphery, has an unclear impact on AD-related cerebrovascular integrity. We showed that APN-deficient mice exhibited reduced resting cerebral blood flow (CBF), impaired neurovascular coupling (NVC), disrupted blood-brain barrier (BBB), and enhanced cerebral amyloid angiopathy (CAA), which are CVD characteristics that also observed in 5xFAD mice, a model of AD. Notably, APN-deficient 5xFAD mice displayed more severe CVD than 5xFAD mice alone. Transcriptomic analysis of brain endothelial cells (ECs) identified dysregulated biological processes and key signaling pathways underlying EC dysfunction. Importantly, the administration of APN restored CBF and NVC in 5xFAD mice, and prevented tight junction protein (TJP) loss and barrier breakdown in Aβ40-exposed primary ECs. These results highlight the potential of alleviating CVD through targeting ECs with APN as a promising therapeutic strategy to delay the onset and mitigate the progression of AD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13195-026-01968-y.

Introduction

Alzheimer’s disease (AD) is a major cause of dementia, primarily manifested by progressive cognitive impairment and memory loss [1]. It is classically characterized by the extracellular accumulation of amyloid-β (Aβ) and the intracellular aggregation of hyperphosphorylated tau protein [2]. Cerebrovascular dysfunction (CVD) refers to the impaired function and regulation of the brain’s blood vessels that disrupt cerebral blood flow (CBF), compromise blood-brain barrier (BBB), and impair neurovascular coupling (NVC), contributing to the development of cognitive decline associated with AD [3–5]. Endothelial cells (ECs) are pivotal for maintaining cerebrovascular integrity and modulating neurodegenerative progression. As the primary cellular component of BBB, ECs facilitates nutrients supply, Aβ clearance, and immune cell trafficking. Tight junction proteins (TJPs), such as zonula occludens (ZO), occludin, and claudins, tightly seal adjacent ECs, forming a selective barrier to regulate the transport of molecules across EC layers. The disruption of TJP leads to increased BBB permeability. Additionally, ECs regulate BBB, CBF, NVC and vascular regeneration through both physical and functional mechanisms, all of which are essential for maintaining cognitive function [6–9]. In turn, sufficient CBF and coordinated NVC help maintain BBB integrity and the endothelial clearance of Aβ [10, 11]. Although the ‘amyloid cascade hypothesis’ and the ‘vascular hypothesis’ are two main favored etiologies of AD, many AD patients present with coexisting vascular risk factors, such as diabetes, hypertension, and hypercholesterolemia [12, 13]. However, the combined effects of Aβ-related pathology and vascular risk factors on EC dysfunction in AD remain poorly characterized.

Levels of adiponectin (APN), an adipocyte-derived hormone, is related to multiple vascular risk factors of AD, such as aging, diabetes, hypertension, and atherosclerosis [14]. It possesses neuroprotective properties through modulating inflammatory processes, insulin sensitization, and lipid metabolism [15–17]. APN is present as a 30 kDa full-length monomeric protein form, which circulates in the body as trimeric, hexametric, and other multimeric complexes [18]. Its major isoform in the brain is the trimeric form (~ 80%), with the remaining 20% being higher molecular weight forms [19]. AdipoR1 and AdipoR2 are two APN receptors that distribute across various brain cell types, including ECs, with AdipoR1 being more pronounced [20]. T-cadherin is another APN receptor that is highly expressed in the vasculature and has a putative interaction with hexametric and multimeric forms [21]. Reduced APN levels have been reported to be involved in AD pathology by promoting amyloidogenic evolvability, a process whereby amyloidogenic protein protofibrils confer adaptive stress resistance during reproduction but contribute to neuroinflammation and neurodegeneration through antagonistic pleiotropy in aging [22–24]. While the vasoprotective functions of APN in the peripheral system are well-documented [25, 26], its specific influence on cerebrovascular homeostasis in AD remains largely unexplored.

This study aimed to investigate the contribution of APN deficiency to the pathogenesis of CVD in a mouse model of AD. We evaluated the severity of CVD across wildtype, 5xFAD, APN knockout (APN^−/−), and 5xFAD;APN^−/− mice. Transcriptomic profiling of brain ECs was performed to identify altered biological processes and signaling pathways mediating the impact of APN deficiency on CVD, with or without β-amyloidogenesis. Furthermore, we explored the therapeutic potential of APN in 5xFAD mice, and in primary mouse brain ECs under Aβ40 toxicity. Our results reveal the detrimental effects of APN deficiency on cerebrovascular integrity in AD, as well as the therapeutic potential of alleviating EC dysfunction through APN treatment to mitigate AD-related CVD.

Materials and methods

Animals

5xFAD transgenic mice were obtained from Jackson Laboratory. These mice overexpress mutant human APP (695) with Swedish (K670N, M671L), Florida (I716V) and London (V717I) familial AD (FAD) mutations along with human PS1 harboring two FAD mutations, M146L and L286V, under the mouse Thy1 promoter. 5xFAD mice and littermate controls were generated by crossbreeding heterozygous 5xFAD mice with C57BL/6 N mice to enable direct comparison between transgenic and wildtype animals within the same genetic and environmental background. Homozygous APN mutants were intercrossed to generate APN knockout (APN^−/−) mice. Genotyping was performed by polymerase chain reaction (PCR) of ear DNA as previously described [27]. APN-deficient 5xFAD (5xFAD;APN^−/−) mice were generated by crossing 5xFAD and APN^−/− mice. Only male mice were used for this study considering the possible fluctuations in APN levels in female mice [28]. Throughout the whole project, age-matched wildtype, APN^−/−, 5xFAD and 5xFAD;APN^−/− mice were studied in each experiment unless otherwise specified. Each experiment included 3 to 8 animals. Investigators responsible for both outcome assessment and data analysis were blinded to group allocation throughout the study. Animal handlers performing experimental procedures were also blinded to group assignments. Where blinding was not feasible, procedures were standardized to minimize potential bias. All mice (4 to 5 mice per cage) were housed in standard holding conditions with 23 ± 2 °C controlled temperature, 12 h light/dark cycle, 60–70% relative humidity, provided with free access to food and water at the Center for Comparative Medicine Research (CCMR) of the University of Hong Kong. All animal procedures were approved by the Committee on the Use of Live Animals in Teaching and Research (CULATR) of the University of Hong Kong.

Brain tissue harvest

To collect brain tissue, mice were anesthetized via intraperitoneal injection of ketamine (100 mg/kg) and xylazine (5 mg/kg) dissolved in PBS, and subsequently transcardially perfused with 30 ml of ice-cold 1× PBS. The brains were carefully dissected from the skulls on ice. One hemisphere was snap frozen in liquid nitrogen for biochemical analysis. The other was fixed by immersion in 4% paraformaldehyde for 24 h at 4 °C. After dehydration with gradient sucrose from 15% to 30%, samples were embedded in tissue-tek compound (O.C.T) and kept at -80 °C for cryosectioning.

Western blotting

Proteins were extracted from mouse brain samples or primary ECs using RIPA lysis buffer and brief sonication, and concentrations were quantified via the DC assay. Electrophoresis was followed by loading an equal amount of protein onto an SDS-PAGE gel. After transferring the protein to a PVDF membrane using a wet transfer system, the membrane was blocked with 5% nonfat milk in 1× tris-buffered saline with Tween 20 for 1 h. Then it was incubated overnight at 4 °C with primary antibodies: anti-APN (1:1000; #MA1-054; Invitrogen), anti-occludin (1:1000; #33-1500; Thermo Fisher Scientific), anti-claudin5 (1:1000; #35-2500; Thermo Fisher Scientific), anti-ZO-1 (1:1000; #40-2200; Invitrogen). Subsequently, incubation was performed at room temperature for 1 h with the corresponding HRP-conjugated secondary antibodies or HRP-conjugated housekeeping proteins: α-tubulin (1:5000; #9099S; Cell Signaling Technology) and sodium potassium ATPase (Na⁺/K⁺-ATPase) (1:5000; #AB185065; Abcam). Protein detection was performed using an enhanced chemiluminescence (ECL) system, followed by imaging with a ChemiDoc MP Imaging System (#12003154; Bio-Rad). Protein intensities were quantified using Image J, normalizing against α-tubulin or Na⁺-K⁺-ATPase as loading control. Data were presented as a ratio to control from three independent experiments.

Immunohistochemistry

Mouse brain samples embedded in O.C.T were cut into 15 μm thick coronal sections using a cryostat (#CM1860; Leica Biosystems) at -20 °C. The sections were mounted on glass slides and allowed to air dry at room temperature overnight. Following a 5-min rinse in ddH₂O, heat-induced antigen retrieval was performed heating the slide-mounted sections in 1×Diva Decloaker solution (#DV2004MX; Biocare Medical). After cooling to room temperature, sections were washed three times with ddH₂O for 5 min each time and blocked with 1% BSA with 0.1% saponin for 1 h at room temperature. The sections were then incubated overnight at 4 °C with primary antibodies: anti-APN (1:200; #MA1-054; Invitrogen), anti-IgG (1:100; #AB6708; Abcam), anti-occludin (1:200; #33-1500; Thermo Fisher Scientific), anti-claudin5 (1:200; #35-2500; Thermo Fisher Scientific), anti-CD31 (1:100; #ab28364; Abcam), anti-CD31 (4 µg/ml; #AF3628; R༆D systems), anti-Aβ (1:100; #MA-48043; Invitrogen), anti-eNOS (1:200; #PA3-031 A; Invitrogen), followed by incubation with fluorophore-conjugated secondary antibodies (Thermo Fisher Scientific) or Dylight 594-conjugated tomato lectin (1:200; #DL-1177-1; Vector Laboratories) for 1 h at room temperature in dark. The slides were covered with a cover slip using mounting medium prior to imaging.

Confocal microscopy imaging using Z-stack acquisition

Confocal laser scanning microscopy was performed using a Zeiss LSM 980 microscope equipped with a 20x/0.8 NA objective lens (Carl Zeiss Microimaging Inc.). Z-stack images were acquired from six non-adjacent (~ 150 μm apart) brain sections per animal, with an approximate interval of 1 μm along the Z-axis. For each section, images were collected from three randomly selected fields within the somatosensory cortex and two randomly selected fields within the hippocampal CA1 subfield [29]. Z-stacks were processed by orthogonal projection using the ZEN software (v 3.8, Carl Zeiss Microimaging Inc.). Three to six animals were analyzed per experimental group.

Immunocytochemistry

Mouse brain primary ECs were seeded onto laminin/collagen IV coated 24-well plate. After incubation and treatment, the culture medium was removed and the cells were prepared for immunofluorescence staining according to the immunohistochemistry protocol [30]. Briefly, cells were fixed by incubating with 4% paraformaldehyde for 10 min at room temperature. Following fixation, cells were washed three times with PBS. After incubation with 1% BSA and 0.1% saponin in PBS for 1 h at room temperature, cells were incubated overnight at 4 °C with primary antibodies. The primary antibodies used were anti-occludin (1:200; #33-1500; Thermo Fisher Scientific), anti-claudin5 (1:200; #35-2500; Thermo Fisher Scientific), anti-CD31 (1:100, #ab28364, Abcam). On the following day, cells were washed thoroughly with PBS and incubated with species-appropriate fluorophore-conjugated secondary antibodies at 1:500 dilution for 1 h at room temperature in dark. Fluorochromes used included Alexa Fluor 405, Alexa Fluor 488 and Alexa Fluor 594 (Thermo Fisher Scientific). After final washes with PBS, samples were then mounted with anti-fade mounting medium and imaged using a Nikon Ti-U inverted fluorescence microscope equipped with a 20× objective lens. The microscope was maintained at room temperature (~ 22 °C) during imaging. Image acquisition was performed using Nikon NIS-Elements software (v 5.42.07), with exposure settings optimized to avoid signal saturation and maintain consistency across samples. All immunocytochemistry experiments were performed with at least three independent replicates.

Laser speckle contrast imaging

Mice were anesthetized by inhalation of 3% isoflurane mixed with oxygen in an induction chamber, using a veterinary anesthesia machine (#R520; RWD Life Science). After induction, isoflurane was reduced to a maintenance dose of 1%, allowing stable and reproducible CBF responses to be observed. A homeothermic temperature controller was used to maintain the rectal temperature at 37 °C. As shown in Fig. 2a, the mouse fur was shaved over the desired imaging location. After that, the mouse was fixed to a stereotaxic adaptor (#68030; RWD Life Science). A 1 cm longitudinal incision was made along the midline of the skull, followed by pulling aside the skin to expose the skull. Periosteum was then removed with fine forceps. Surface of the skull was cleaned with sterile gauze and cotton-tip applicators until dry and free of blood. Resting CBF was measured using a laser speckle contrast imaging (LSCI) system (#RFLSI-ZW; RWD Life Science) with a 1 s interval in sliding mode. The probe was settled approximately 23.8 cm above the frontal parietal cortical region of the brain. Blood perfusion was expressed as an arbitrary unit. A 5-min baseline was established and recorded at the beginning of the experiments. Region of interest (ROI) was placed over the barrel cortex, spanning from bregma to lambda [31]. After imaging, skin was sutured with a 5/0 suture and disinfected by applying 3 times of betadine. This procedure was repeated on each mouse at 6 and 9 months old.

Fig. 2.

APN deficiency aggravates resting CBF reduction and NVC impairment in 5xFAD mice. a Schematic showing experimental design. Laser speckle contrast imaging (LSCI) was employed to measure CBF during resting conditions and whisker stimulation by placing the probe above the frontal-parietal cortical region. The experimental protocol consisted of a 5-min baseline to record resting CBF, followed by six rounds of whisker stimulation (30 s each) and post-stimulation intervals (60 s each). Created in BioRender (Yick, K. ,2025; https://BioRender.com/b3tdnjw). b, c Representative LSCI perfusion images (b) and quantification of mean perfusion values of resting CBF (c) in 6 and 9-month-old wildtype, 5xFAD; APN^−/−; 5xFAD;APN^−/− mice (n = 7). d Percent change in resting CBF from 6 to 9 months across genotypes (n = 7). e, g Representative LSCI images showing CBF at baseline and maximum sensory stimulation in 6-month-old (e) and 9-month-old (g) mice. ROIs are indicated by circles. f, h Representative time course and quantification of stimulus-evoked CBF changes relative to baseline within the ROIs during contralateral whisker stimulation at 6 (f) and 9 (h) months (n = 7–8). Data are expressed as mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. One-way ANOVA with Tukey’s multiple comparisons test

Whisker stimulation

To evaluate the capability of NVC, mice were gently restrained by the stereotaxic adaptor (#68030; RWD Life Science), and the right whisker was manually stimulated using a soft synthetic brush at a frequency of 3–4 Hz for 30 s, with each stimulation repeated six times at 60-sec intervals [3]. CBF measurements were carried out under light sedation with 1% isoflurane using LSCI, with data acquired at 0.3-sec intervals in step mode. The ROI was placed in the contralateral barrel cortex, corresponding to the somatosensory area responsive to the right whisker stimulation, approximately located 0–1 mm posterior and 3.5–4.5 mm lateral bregma [8, 32]. A homeothermic temperature controller was used to maintain the rectal temperature at 37 °C. Experiments were conducted in a noise-, temperature-, light-, and airflow-controlled environment. This procedure was repeated on each mouse at 6 and 9 months old.

Fluorescein sodium salt penetration assay

BBB permeability was estimated by the fluorescein sodium salt (NaFl, 376 Da) penetration assay as previously described [33–35]. In brief, after i.v. administration of 10% NaFl (3 ml/kg; #F6377; Sigma Aldrich), the mice received intraperitoneal anesthesia consisting of ketamine (100 mg/kg) and xylazine (5 mg/kg) dissolved in PBS, followed by a 10-min circulation period. Cardiac blood (approximately 500 µl) was collected and allowed to clot at room temperature for at least 15 min before centrifugation at 3000 g for 10 min to obtain serum. Mice were then immediately cardiac perfused with 30 ml of PBS to clear residual fluorescent marker, and the brains were dissected into hemibrains. One hemi-cerebrum was weighted, homogenized, deproteinized, and centrifugated. Then the supernatants were evaporated to dryness and reconstituted in PBS, followed by reading the raw fluorescence units (RFUs) of the samples added in a 96-well plate at excitation wavelength at 440 nm and emission wavelength at 525 nm. At least one sample from a sham animal, including both serum and tissue supernatants, was included in the measurements to ensure accurate background subtraction. BBB integrity was indicated by the tissue fluorescence which is normalized to the wet weight and serum fluorescence to obtain a quantitative tracer permeability index after subtracting the corresponding sham values [35]. For tracer fluorescence visualization, the contralateral hemi-brain was embedded in O.C.T and cut into 10 μm thick sections, followed by stained with Dylight 594-conjugated tomato lectin (1:200; #DL-1177-1; Vector Laboratories) as a vascular marker. Images were acquired using a fluorescent microscope.

Sorting of mouse brain endothelial cells and bulk RNA sequencing

ECs were isolated from the cortex and hippocampus using fluorescent activated cell sorting as previously described with some modifications [36]. Four 6-month-old male mice per group were deeply anesthetized via intraperitoneal injection of ketamine (100 mg/kg) and xylazine (5 mg/kg), prepared in PBS. The cortex and hippocampus were dissected, pooled [37] and then minced into ~ 1 mm pieces. Each sample was mechanically dissociated using a single cell suspension dissociator (#DSC-400; RWD Life Science) and enzymatically digested with 5 ml collagenase/dispase solution (4 mg/ml final concentration in 2% FBS) (#10269638001; Roche) at 37 °C for 40 min. After centrifugation (300 g, 10 min, 4 °C), pellets were triturated with 0.5 mg/ml DNase (#07469; Stemcell) by pipetting ~ 100 times. For size exclusion, cells were filtered through a 40 μm mesh, washed with 10 ml PBS, subjected to myelin removal (#DHABE-5003; RWD Life Science). Following washing, cells were resuspended with HBSS buffer (containing 1% BSA, 0.1% glucose and 2mM EDTA) and subsequently stained for 30 min at 4℃ with anti-mouse CD45-FITC (1:200; #553080; BD Bioscience), anti-mouse CD31-PE (1:100; #12-0311-81; eBioscience), and Zombie NIR APC750 viability dye (1:100000; #77184; BioLegend). Compensation controls were prepared using a subset of stained cells. EC sorting was performed using a FACSAria Fusion cell sorter (BD Bioscience) with 100 μm nozzle size. Gating included singlets, live cells (Zombie negative), CD45 negative immune cells, and CD31 positive ECs (Fig. S5 a). Cells were sorted using a 4-way purity mask into the HBSS buffer (containing 1% BSA, 0.1% glucose) for RNA extraction and subsequent RNA sequencing.

RNA extraction, quality control, library preparation and sequencing were performed by Beijing Genomics Institute (BGI) [38] using the DNBSEQ™ platform with paried-end 100 bp (PE100) reads. Data analysis was performed using BGI’s Dr. Tom system for transcriptome data visualization and interpretation [39]. Trimmed clean reads were aligned to the Mus musculus UCSC genome assembly (mm39, v.2.2.0.1) [40], by applying Bowtie2 [41]. Differential gene analysis was performed using DEGseq [42], with the significance thresholds set at Q ≤ 0.05 or FDR ≤ 0.001. Gene expression clustering heatmaps across samples were generated using Hiplot [43]. To investigate gene functions associated with phenotype changes, Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) enrichment analyses were performed on differentially expressed genes using the hypergeometric test implemented in Phyper. A significance threshold of Q ≤ 0.05 was applied [44], and genes meeting this criterion were considered significantly enriched in the candidate gene sets.

Isolation and culture of primary mouse brain endothelial cells

Primary mouse brain ECs were prepared as previously described with some modifications [45]. C57BL/6 N male mice aged between 6 to 8-week-old were sacrificed and their brains were harvested under sterile conditions. Meninges, olfactory bulbs, and cerebellum were carefully removed. After mincing into small pieces, the brain tissue was mechanically (Single cell suspension dissociator; #DSC-400; RWD Life Science) and enzymatically (Adult brain enzymatic digestion kit; #DHABE-10; RWD Life Science) digested to create a cell suspension, which was then subjected to debris removal (Adult brain enzymatic digestion kit; #DHABE-10; RWD Life Science) to separate ECs from other cell types, especially myelin. Following isolation, the cells were cultured on laminin/collagen IV precoated cell culture plates in specialized endothelial cell medium consisting of 80% DMEM medium, 20% FBS, 0.05% basic fibroblast growth factor (20 mg/ml) and 0.1% heparin (100 mg/ml) under controlled conditions (37℃, 5% CO₂). 0.1% puromycin (4 mg/ml ) were added only in the first 2 days of culture to suppress growth of contaminating cell types. ECs were characterized by the typical spindle-shaped appearance. At confluence, they formed a monolayer of tightly packed, longitudinally aligned and non-overlapping contact-inhibited cells. Culture medium was changed every two days until cells reached confluence for following experiment.

Preparation of Aβ oligomers

Aβ oligomers were prepared as previously described [46]. Briefly, 1 mg of human Aβ40 (#ab120479; Abcam) or Aβ42 (#ab120301; Abcam) peptide was dissolved in 221.7 µl cold 1,1,1,3,3,3-Hexafluoro-2-propanol (HFIP) and incubated at room temperature for 1 h, followed by 10 min on ice. The solution was aliquoted into non-siliconized microcentrifuge tubes (100 µl per tube), with one aliquot reserved for western blot confirmation, and dried overnight in a fume hood. On the next day, residues were resuspended in 20 µl 5 mM DMSO and diluted with 980 µl of F12 medium to a final concentration of 100 µM. The solution was incubated overnight at 4 °C, then centrifuged at 14,000 × g for 10 min at 4 °C. The supernatant containing Aβ oligomers was collected, and an aliquot was used for western blot analysis with anti-Aβ antibody (1:1000; #800707; Biolegend) to confirm oligomer formation. In all experiments involving Aβ oligomers dissolved in DMSO, the final concentration of DMSO was maintained at 0.2% (v/v), a level below commonly reported cytotoxic thresholds [47]. Vehicle control groups were treated with the same concentration of DMSO to ensure that observed effects on ECs were not attributable to solvent toxicity.

Rhodamine B dextran permeability assay on endothelial monolayers

We measured EC permeability using the dextran assay protocol as previously described [48]. In brief, primary mouse brain ECs were seeded on laminin/collagen IV-coated inserts with 0.4 μm pore size (#353095; Corning). EC culture medium, as described above, was added to the upper (300 µl) and lower (700 µl) chambers to culture the cells until they formed a confluent monolayer. Following a 3 h starvation period in DMEM medium, cells were treated with 10µM Aβ40 or Aβ42 oligomers for 24 h, with or without 10 mg/ml APN trimer pretreatment. The medium in the upper chamber was then replaced with rhodamine B dextran (70 kDa; #D1841; Invitrogen) diluted to 1 mg/ml in DMEM medium. At 30, 60, 120, and 180 min, 100 µl samples were collected from the lower chamber, and fluorescence was measured using a microplate reader (excitation 492 nm, emission 520 nm) (Fig. 7g left panel). Data were normalized to controls to calculate fold changes in permeability.

Fig. 7.

APN treatment restores tight junction protein expression and reduces endothelial permeability in Aβ40-exposed endothelial cells. a Representative western blot images showing the expression of occludin and claudin5, with Na⁺-K⁺-ATPase as the loading control, in primary mouse brain endothelial cells (ECs) treated with control, Aβ40, APN, or Aβ40 + APN. b Quantification of occludin and claudin5 expression levels relative to Na⁺-K⁺-ATPase (n = 5). c, d Representative immunofluorescence images of ECs showing occludin (green) and claudin5 (green) co-stained with CD31 (red), a marker for ECs. Scale bar, 50 μm. e–g Quantification of mean fluorescence intensity for occludin (e) and claudin5 (f), and relative dextran permeability (g) across EC monolayers under the indicated treatments (occludin, n = 3; claudin5, n = 3; dextran permeability, n = 6 ). The illustrated protocol in panel (g, left) was created in BioRender (Yick, K., 2025; https://BioRender.com/uhqmkkn). Data are expressed as mean ± SEM. *, P < 0.05; **, P < 0.01; ****, P < 0.0001; ns, not significant. One-way ANOVA with Tukey’s multiple comparisons test

Statistical analysis

Statistical analyzes were performed with Prism 8.0.2 (GraphPad Software). Quantitative data were expressed as mean ± standard errors of the mean (SEM). Non-paired t-test was utilized to assess the means of two independent groups. Paired t-test was employed to compare the NVC capabilities of the mice before and after APN treatment. For evaluating differences among three or more independent groups, one-way ANOVA was performed followed by post hoc comparisons using Tukey’s test. The difference was statistically significant when P < 0.05.

Results

5xFAD mice exhibit reduced APN levels in the brain

To investigate whether there is a reduction of APN in the pathogenesis of AD, we examined APN expression in 6-month-old wildtype and 5xFAD mice. The specificity of the APN antibody was validated by immunohistochemistry in mouse brain sections, as demonstrated in Supplementary Fig. S1. Immunofluorescence staining revealed strong APN immunoreactivity in the cerebral vascular endothelium, as indicated by colocalization with CD31 (Fig. 1a) and lectin staining (Fig. S8). Notably, APN levels significantly decreased in the cerebral vessels of 5xFAD mice compared to wildtype controls (Fig. 1a and c). Additionally, western blot analysis of brain lysates demonstrated a significant reduction of APN abundance in 5xFAD mice compared to age-matched wildtype mice (Fig. 1, b and d). Collectively, these findings indicate that APN levels are markedly diminished in the brains of AD mice.

Fig. 1.

5xFAD mice exhibit reduced vascular APN levels in the brain. a, c Representative confocal photomicrographs of APN (red) and CD31-positive vessels (green) (a) and quantification of vascular APN intensity of CD31⁺ area (c) in 6-month-old wildtype and 5xFAD mice brain (n = 5). Scale bar, 50 μm. b, d Representative immunoblotting of APN levels (b) and quantification of APN relative abundance (d) in the brain lysates of 6-month-old wildtype and 5xFAD mice (n = 3). Data are expressed as mean ± SEM. *, P < 0.05. Unpaired t test

APN deficiency worsens CBF reduction and NVC impairment in 5xFAD mice

CBF delivers oxygen and glucose for brain energy metabolism [49], while NVC dynamically adjusts CBF via a feed-forward mechanism of functional hyperemia to match neural activity and ensure proper brain function [50, 51]. Whisker stimulation is widely used to study the mechanism of NVC [3]. To study the effects of APN deficiency on CBF and NVC in AD, we utilized LSCI system to measure resting CBF and CBF response to whisker stimulation at 3 months of age (Fig. S2), as well as longitudinally from 6 to 9 months of age (Fig. 2).

At 3 months, resting CBF was largely preserved across wildtype, 5xFAD, APN^−/−, and 5xFAD;APN^−/− mice, although a slight reduction was observed in 5xFAD;APN^−/− group (Fig. S2 a and b). At later time points, 5xFAD mice exhibited marked decrease in resting CBF at both 6 and 9 months of age compared to wildtype controls. APN^−/− mice showed significant resting CBF reduction only at 9 months, indicating that APN loss alone leads to age-dependent cerebral hypoperfusion. Importantly, 5xFAD;APN^−/− mice demonstrated a more pronounced decrease in resting CBF compared to 5xFAD mice, suggesting an additive or enhanced effect of APN deficiency and amyloid pathology on cerebral perfusion (Fig. 2b and c). Figure 2d showed that all groups show a negative percentage change, indicating a decrease in resting CBF from 6 to 9 months, which may be partially attributed to the age-dependent cerebrovascular remodeling [52]. The APN^−/− mice exhibited a more pronounced decrease than wildtype and 5xFAD mice. 5xFAD;APN^−/− mice displayed significant reductions in resting CBF change compared to 5xFAD mice. These findings indicate that the absence of APN exacerbates age-related decrease in resting CBF, either alone or in combination with the 5xFAD genotype.

Regarding CBF responses, NVC was significantly impaired in 5xFAD;APN^−/− mice compared to wildtype at 3 months, indicating early CVD induced by APN deficiency (Fig. S2 c and d). Both 5xFAD and APN^−/− mice displayed impaired CBF increase during whisker stimulation at 6 and 9 months of age compared to wildtype, with the 5xFAD;APN^−/− mice showing even greater deficits than 5xFAD mice at both ages (Fig. 2e-h).

These results indicate that APN deficiency accelerates the onset of NVC impairment as early as 3 months, amplifies NVC impairments at both 6 and 9 months, and exacerbates resting cerebral hypoperfusion at 9 months in the AD mouse model, suggesting a synergistic detrimental effect of APN loss and Aβ deposition on cerebrovascular function during disease progression.

APN deficiency leads to early BBB leakage

To examine whether APN deficiency initiates and worsens BBB disruption, we assessed its permeability by measuring the infiltration of intravenously (i.v.) administered NaFl, a small molecular weight tracer, into the mouse brain parenchyma. Young 3-month-old 5xFAD;APN^−/− mice displayed more NaFl extravasation as indicated by lectin staining, showing approximately 1.5-fold increases compared to age-matched wildtype, 5xFAD and APN^−/− mice. These findings indicate an earlier BBB breakdown in APN-deficient 5xFAD mice. At 6 months-old, both 5xFAD and APN^−/− mice demonstrated significantly higher NaFl permeability than wildtype mice. Notably, both APN^−/− mice and 5xFAD;APN^−/− mice exhibited even greater NaFl infiltration compared to 5xFAD mice at this age (Fig. 3).

Fig. 3.

APN deficiency exacerbates BBB leakage in 5xFAD mice. a, c After intravenous injection of fluorescein sodium salt (NaFl, green, MW: 376.27 Da) in 3-month-old (a) and 6-month-old (c) wildtype, APN^−/−, 5xFAD, and 5xFAD;APN^−/− mice, the tracer was visualized together with tomato lectin staining (red) on brain cryosections. Scale bar, 50 μm. b, d Quantification of NaFl amounts in hemi-brain tissues of different groups at 3 months (b) and 6 months (d) of age (n = 5–7). The Brain Permeability Index is calculated as follows: Brain Permeability Index = ((Sample Brain RFUs^# - Sham Brain RFUs^# ) / Brain Weight (g)) / ((Sample Serum RFUs^# - Sham Serum RFUs^# ) / Serum Volume (ml)). ^#RFUs: raw fluorescence units. Data are expressed as mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. One-way ANOVA with Tukey’s multiple comparisons test

Furthermore, BBB integrity was examined by examining the extravascular accumulation of circulating endogenous IgG (MW:155,000 Da). We observed a substantial increase in parenchymal accumulation of plasma-derived IgG in both 6-month-old 5xFAD and 5xFAD;APN^−/− mice compared to wildtype mice. A more pronounced IgG leakage was found in 9-month-old mice. Importantly, 5xFAD;APN^−/− mice began to exhibit greater IgG infiltration than 5xFAD mice at 9 months of age (Fig. S3). These results indicate that APN deficiency accelerates and augments BBB breakdown.

APN deficiency aggravates cerebrovascular TJP loss in 5xFAD mice

TJPs are critical for maintaining the barrier properties of brain endothelium. Previous in vitro study demonstrated the TJP loss in brain ECs under Aβ toxicity [53]. To investigate whether APN similarly impacts TJP integrity in vivo, we performed co-staining experiments for claudin5 or occludin with endothelial marker CD31 in 6-month-old mice. CD31⁺ vessel lengths were comparable among wildtype, 5xFAD, APN^−/−, and 5xFAD;APN^−/− mice at 6 months (Fig. 4b and e). Compared to age-matched 5xFAD littermates. 5xFAD;APN^−/− mice exhibited significantly diminished claudin5- and occludin-positive vascular length in CD31-positive area, indicating that APN deficiency aggravates TJP reduction in AD. Strikingly, APN^−/− mice showed comparable decrease in both claudin5- and occludin-positive vascular length to 5xFAD mice (Fig. 4a, c, d and f), suggesting APN deficiency alone drives TJP loss to a similar extent as Aβ pathology. These data establish APN loss as an independent pathological driver of BBB disruption, which synergistically worsens CVD during AD pathogenesis.

Fig. 4.

APN deficiency worsens cerebrovascular tight junction protein loss in 5xFAD mice. a, d Representative confocal photomicrographs of occludin or claudin5 (red) and CD31-positive vessels (green) of 6-month-old mice from different groups. Scale bar, 100 μm. b, e Quantification of CD31⁺ vessel length across four groups. c, f Quantification of occludin (c) and claudin5 (f) length on CD31-positive vessels (n = 3). Data are expressed as mean ± SEM. ns, not significant; *, P < 0.05 ; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. One-way ANOVA with Tukey’s multiple comparisons test

APN deficiency aggravates CAA in 5xFAD mice

CAA is characterized by the accumulation of Aβ in the cerebral vessel walls due to increased production and impaired clearance, leading to cerebral hemorrhage, vascular inflammation, vessel occlusion, CBF disturbances and contributing to dementia development [54–57]. The BBB is a critical site of Aβ clearance from the brain to blood [58]. We have previously demonstrated that APN deficiency accelerates and exacerbates Aβ deposition in brain parenchyma of 5xFAD mice [23]. To further investigate whether APN deficiency also affect the severity of cerebrovascular β-amyloidosis, Aβ pathology was analyzed by co-immunofluorescence staining by Aβ antibody (6E10) and the vascular marker CD31. The results demonstrated that 6-month-old APN^−/− mice exhibited notable amyloid accumulation on cerebral vessels compared with wildtype mice. No obvious parenchymal plaques were observed at this age (Fig. S4). It indicates that loss of APN alone may contribute to the development of CAA, which precedes the amyloid plaques formation that we previously observed at 18 months of age [22]. Furthermore, 5xFAD;APN^−/− mice displayed a significant increase in cerebral amyloid deposition on brain vessels compared to 5xFAD mice at 6 months old (Fig. 5), suggesting APN deficiency leads to a more severe CAA in AD.

Fig. 5.

APN deficiency results in more severe CAA in 5xFAD mice. a Representative confocal photomicrographs of CD31 (green) and Aβ (red) in of wildtype, 5xFAD, APN^−/− and 5xFAD;APN^−/− mice at 6 months of age. Scale bar, 50 μm. b Quantification of vascular Aβ intensity relative to CD31⁺ vessel area (n = 5–6). Data are expressed as mean ± SEM. *, P < 0.05 ; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. One-way ANOVA with Tukey’s multiple comparisons test

APN deficiency parallels endothelial transcriptomic alterations and downregulates key signaling pathways in 5xFAD mice

Endothelium plays a central and multifaceted role in CVD. Transcriptomic profiling of ECs in APN deficiency, Aβ pathology and their combined status was performed by sorting of brain ECs from different groups. Hierarchical clustering displays normalized gene expression levels of different brain cell markers in EC-enriched samples. The strong positive average expression of EC-specific genes indicates successful enrichment of ECs. (Fig. S5 b). A Venn diagram comparing differentially expressed genes (DEGs) in brain EC across three conditions − 5xFAD vs. wildtype, APN^−/− vs. wildtype, and 5xFAD;APN^−/− vs. wildtype - identified 81 shared DEGs (Fig. 6a, left panel). Hierarchical clustering of these shared DEGs showed consistent expression patterns across different groups of mice, such as genes that were uniformly upregulated or downregulated. Notably, many genes in the 5xFAD;APN^−/− mice displayed the greatest change in expression levels. Wildtype ECs showed predominantly downregulated expression for many genes compared to other groups. These findings suggest that APN deficiency amplifies the expression changes of genes that are already dysregulated in 5xFAD mice (Fig. 6a, right panel).

Fig. 6.

Transcriptomic similarities in endothelial cells between 5xFAD and APN^−/−mice and downregulated pathways in 5xFAD;APN^−/− mice. a Venn diagram showing the overlap of differentially expressed genes (DEGs) across three comparisons: 5xFAD vs. wildtype, APN^−/− vs. wildtype, and 5xFAD; APN^−/− vs. wildtype. The heatmap represents hierarchical clustering of the 81 shared DEGs, with normalized expression levels (Z-scores) across the four groups (wildtype, 5xFAD, APN^−/−, and 5xFAD; APN^−/− ). Red indicates upregulation, and blue indicates downregulation. b Venn diagram showing the overlap of DEGs between 5xFAD vs. wildtype and APN^−/− vs. wildtype comparisons. Scatterplot shows the correlation of log2 fold changes (log2FC) for the overlapping DEGs between the two comparisons (r² = 0.79, P < 0.0001). Genes in the orange quadrant are upregulated in both models, while genes in the blue quadrant are downregulated in both models. c GO analysis of biological processes shared between 5xFAD and APN^−/− mice. The dot plot displays significantly enriched GO terms, with dot size representing gene count and color indicating statistical significance (log10 Q value). Processes related to immune response, microglial activation, and synapse pruning are among the top enriched terms. d KEGG pathway analysis of DEGs in 5xFAD;APN^−/− compared to 5xFAD mice. Bar plot shows significantly downregulated pathways, including the apelin signaling pathway, ECM-receptor interaction, PI3K-Akt signaling pathway, and AGE-RAGE signaling pathway in diabetic complications5xFAD; APN−/−

Venn diagram highlights overlapping 113 DEGs between 5xFAD vs. wildtype and APN^−/− vs. wildtype comparisons (Fig. 6b, upper panel). Further analysis highlighted a strong correlation between shared DEGs in 5xFAD vs. wildtype and APN^−/− vs. wildtype comparisons, with a correlation coefficient of r² = 0.79 (p < 0.0001) (Fig. 6b, lower panel). These data suggest potentially shared molecular mechanisms in brain ECs between 5xFAD and APN^−/− mice. GO analysis revealed that, out of the top 20 shared upregulated biological processes, 13 out of them were related to immune response and microglial activation. Processes such as synapse pruning, microglial cell activation involved in immune response, and complement-mediated synapse pruning were enriched, suggesting that brain ECs actively contribute to neuroinflammation and synaptic remodeling. It aligns with notion that ECs play an auxiliary role through shared immune response pathways in the intertwined microglia-vascular axis [59]. On the other hand, the top 20 shared downregulated pathways primarily associated with metabolic and cellular maintenance functions (Fig. 6c). Processes such as oxidative phosphorylation, hydrogen peroxide catabolism, and cellular carbohydrate catabolism were suppressed, indicating impaired energy metabolism and reduced capacity to mitigate oxidative stress in brain ECs. These changes are consistent with mitochondrial dysfunction, which is known to compromise BBB integrity and EC survival [60, 61]. Furthermore, the dysregulation of DNA-templated transcription may further impair cellular homeostasis. Overall, these findings suggest heightened immune activation and suppressed metabolic capacity as mechanisms underlying the detrimental effects on brain ECs. This imbalance likely contributes to BBB dysfunction, neurovascular instability, and exacerbation of neurodegenerative pathology in both 5xFAD and APN^−/− mice.

Finally, comparing 5xFAD;APN^−/− mice to 5xFAD mice revealed significant downregulation of four pathways critical for EC homeostasis, without any upregulated signaling. These pathways include the apelin signaling pathway, extracellular matrix (ECM)-receptor interaction, phosphoinositide 3-kinase–Akt (PI3K-Akt) signaling pathway, and advanced glycation end-products (AGEs)-receptor for AGEs (RAGE) signaling pathway in diabetic complications. The downregulation of these pathways suggests the impact of APN deficiency on exacerbating endothelial dysfunction in the 5xFAD model (Fig. 6d), a pathological state marked by decreased vasodilation, increased inflammation, prothrombotic activity, and altered vascular permeability, primarily driven by reduced nitric oxide (NO) bioavailability and heightened oxidative stress [62].

Taken together, the similar transcriptomic dysregulations in brain ECs, including prominent immune activation and metabolic dysregulation, are consistent with shared pathological alterations between 5xFAD and APN^−/− mice. Additionally, APN deficiency in 5xFAD mice further suppresses pathways critical for endothelial function. These findings suggest that APN deficiency both initiates and amplifies CVD in AD, and it is a critical determinant of neurovascular pathology progression.

APN prevents TJP loss and barrier breakdown in primary ECs under Aβ40 toxicity

It has been reported that APN attenuates TJP disruption in the immortalized murine brain EC line (bEnd.3) exposed to 20 µM Aβ42 oligomer [53]. However, Aβ40 has a greater tendency to deposit in the vessel walls than Aβ42, as indicated by the higher Aβ40/Aβ42-ratio in these CAA-lesions than that of senile plaques [63]. To compare the differential toxic effects of Aβ40 and Aβ42 oligomers on the endothelium, both western blot and immunocytochemistry analyses demonstrated a significantly reduction in TJP expression in primary mouse brain ECs treated with 10 µM Aβ40 oligomers for 24 h compared to controls, whereas exposure to the same concentration and duration of Aβ42 oligomers did not alter TJP levels (Fig. S6 a-d). In addition, rhodamine B dextran as a tracer to assess endothelial barrier integrity was utilized and revealed that Aβ40-treated ECs exhibited more pronounced barrier disruption after 120 min of tracer incubation (Fig. S6 e). These findings indicate that Aβ40 exerts greater endothelial toxicity than Aβ42 under these experimental conditions.

To investigate the protective effects of APN on ECs exposed to Aβ40 oligomer toxicity, a dose-dependent study was conducted by treating primary ECs with APN trimer at concentrations ranging from 0 to 10 µg/ml. Western blot analysis revealed a significant increase in the abundance of both ZO-1 and occludin at the 10 µg/ml concentration (Fig. S7). Furthermore, we examined the alterations in TJPs (claudin5 and occludin) using western blotting and immunofluorescence staining in Aβ40 treated primary ECs with or without prior 10 µg/ml APN treatment. We found that APN effectively rescued the loss of claudin5 and occludin levels induced by Aβ40 (Fig. 7a-f).

In addition, we inspected the paracellular permeability of the EC layer by incubating the cells with rhodamine B dextran diluted in DMEM medium for 120 min, following a 24 h exposure to Aβ40, with or without pretreatment with APN. We observed that Aβ40 significantly increased the diffusion of dextran across the cell barrier from the apical to basolateral side, and these changes were counteracted by pretreatment with APN (Fig. 7g). These results indicate that APN has a protective effect against Aβ40-induced breakdown of endothelial barrier.

APN administration enhances CBF and NVC in 5xFAD mice

We observed lower levels of APN in the brains of 5xFAD mice, contributing to accelerated and exacerbated CVD in this AD model. We further investigated whether enhancing APN levels in the mouse brain could mitigate decreases in CBF and impairments in NVC in 5xFAD mice. Figure 8a illustrates experimental design. Administration of APN resulted in increased resting CBF in 5xFAD mice at 6 months of age, with an average increase of 20% in the ROI compared to vehicle-treated 5xFAD mice (Fig. 8b and c). Furthermore, a marked improvement in the response of CBF to whisker stimulation was observed after APN treatment (Fig. 8d and e). These data indicate that APN could acutely improve resting cerebral perfusion and NVC ability in 6-month-old AD mice.

Fig. 8.

Administration of APN mitigates CBF and NVC impairments in 5xFAD mice. a Experimental design: the procedure began with a pre-treatment measurement of resting CBF over a 5-min period. Subsequently, either vehicle or APN was administered intravenously, followed by a 10-min circulation period. After treatment, the resting CBF was measured for an additional 10 min. NVC was then assessed through six rounds of whisker stimulation. Blood and tissue samples were collected for further analysis. Created in BioRender (Yick, K., 2025; https://BioRender.com/iqu1yi1). b Representative perfusion images of the frontal-parietal cortical region and temporal CBF curve of 6-month-old 5xFAD mice before and after intravenous administration of either 0.9% saline (vehicle) or recombinant mouse APN trimers (1 mg/kg body weight). c Quantification of CBF changes relative to baseline following treatment with vehicle or APN (n = 7–8). d Representative perfusion images of somatosensory cortex region of baseline and maximum perfusion during stimulation before and after APN treatment. e Quantification of CBF increases in the left barrel cortex relative to baseline in response to six rounds of contralateral whisker stimulation (n = 8). f, g Representative confocal photomicrographs of tomato lectin-positive vessels (red) and eNOS (green) (f) and quantification of eNOS intensity relative to lectin⁺ vessel area in vehicle-treated and APN-treated 5xFAD mice (n = 3) (g). Unpaired t test in (c) and (g). Paired t test in (e). Scale bar, 50 μm. Data are expressed as mean ± SEM. *, P < 0.05; **, P < 0.01

Peripheral administration of APN trimer increases brain eNOS levels

Endothelial nitric oxide synthase (eNOS) synthesizes NO in ECs, which regulates vascular tone, inhibits platelet aggregation, maintains endothelial barrier function, and modulates leukocyte adhesion. To investigate the impact of peripherally administered APN on the brain eNOS levels in AD, we examined the 5xFAD mice that received treatment with APN trimer or vehicle. Immunofluorescence staining revealed significantly enhanced immunoreactivity for eNOS in the cerebral vascular endothelium, as indicated by lectin staining, in APN-treated 5xFAD mice compared to vehicle-treated ones (Fig. 8f and g). These data suggest that APN trimer displays an instant vasoactive property via enhancing eNOS expression to induce vasodilation in AD.

Supplemental material summary

Supplemental Figures S1–S8 provide additional data supporting the main findings. Figure S1 validates the specificity of the APN antibody and confirms knockout models by immunohistochemistry. Figure S2 shows that APN deficiency accelerates NVC impairment in 3-month-old 5xFAD mice. Figure S3 demonstrates that APN deficiency exacerbates extravascular IgG deposition in 5xFAD mice over time, indicating increased BBB disruption. Figure S4 presents immunofluorescence staining of vascular, intracellular, and parenchymal amyloid pathology using Aβ antibody (6E10). Figure S5 details the flow cytometry gating strategy and confirms successful EC enrichment for transcriptomic analysis. Figure S6 reveals that, at the same dose and duration, Aβ40 but not Aβ42, impairs TJP expression and compromises endothelial barrier integrity. Figure S7 demonstrates that APN treatment mitigates Aβ40-induced TJP loss in primary brain ECs in a dose-dependent manner. Figure S8 shows a predominant spatial association of APN with cerebral blood vessels rather than brain parenchymal cells. These supplemental materials offer extended methodological details and quantitative validation that reinforce the conclusions of the study.

Discussion

This study demonstrates that APN deficiency independently compromises vascular integrity. Importantly, it synergizes with Aβ toxicity to exacerbate CVD in AD (Fig. 9). Transcriptomic profiling further indicates that APN deficiency drives AD-like genetic dysregulation in ECs, and it amplifies endothelial dysfunction in AD through down-regulation of pathways critical for EC survival, BBB maintenance, and neurovascular integrity. The administration of APN rescues Aβ-induced endothelial damage and restores CBF and NVC in 5xFAD mice. Collectively, these findings highlight the potential of APN for alleviating EC dysfunction as a therapeutic strategy to mitigate CVD in AD.

Fig. 9.

APN deficiency drives CVD and, together with Aβ toxicity, aggravates AD pathology. This schematic depicts the pathological cascade of CVD triggered by APN deficiency, either independently or synergistically with Aβ toxicity. A vicious cycle forms among CBF reduction, NVC impairment, BBB breakdown, and CAA pathology. Endothelial dysfunction initiates BBB disruption and impairs NVC, diminishing the cerebral vasculature’s responsiveness to neuronal activity. Consequently, CBF decreases, causing hypoperfusion and further endothelial injury. Reduced CBF exacerbates BBB breakdown by altering shear stress and vascular tone, further compromising barrier integrity. Impaired BBB function hinders Aβ clearance, promoting its accumulation in vessel walls and advancing CAA. Vascular Aβ deposition intensifies endothelial damage, amplifying BBB leakage and disrupting NVC, thereby perpetuating the cycle. Each vascular insult independently induces neuronal damage and mutually exacerbates others. The combined effects of APN deficiency and Aβ toxicity intensify this vicious cycle, accelerating neurodegeneration and cognitive decline. The schematic emphasizes brain endothelial cells as key mediators of CVD and highlights APN as a potential strategy to mitigate CVD in AD progression. Created in BioRender (Yick, K., 2025; https://BioRender.com/ue8f7tx)

Despite conflicting reports on APN expression levels in AD patients [24, 64], this study found a significant reduction in APN around the cerebral vasculature in 5xFAD mice as early as 6 months old, preceding the previously reported age of 9 months [23]. Further investigation is needed to determine whether the source of APN is perivascular adipose tissue [25], EC themselves [65], or circulating APN that binds to its receptors expressed on ECs [66, 67]. However, APN immunoreactivity was found to be predominantly localized alongside blood vessels rather than diffusely distributed within the brain parenchyma (Fig. S8), a pattern consistent with prior observations in ischemic brain tissues [68]. This spatial association of APN with blood vessels supports that reduced APN levels are involved in the pathogenesis CVD in AD.

APN deficiency drives early-onset CVD that precedes cognitive decline, resembling the vascular pathology that observed in AD. Previously, we reported that 18-month-old APN-deficient mice showed impaired spatial learning and memory impairments along with increased cerebral Aβ accumulation, tau phosphorylation, and neuroinflammation [22]. In this study, most experiments were performed in 6-month-old 5xFAD mice because these ages represent a stage when CVD is clearly present but before advanced neurodegeneration, allowing us to study early pathological changes relevant to disease progression. We first confirmed that APN-deficient mice, as early as 6 months of age, displayed multiple CVD pathologies shared with 5xFAD mice. Indeed, CVD is well known to cause neurodegeneration and cognitive impairment [13, 69–71]. APN exhibits anti-inflammatory and antioxidative stress properties [72, 73], which are impaired in AD [74–77]. In addition, APN is involved in various metabolic processes, particularly in the regulation of insulin sensitization, glucose and lipid metabolism [78, 79]. Reduced APN levels may initiate metabolic stress in different cell types within the neurovascular unit due to compromised energy expenditure. On the other hand, AD has been affirmed as ‘type 3 diabetes’ characterized by brain insulin resistance, metabolic impairment, and disrupted insulin signaling leading to neurodegeneration [80, 81]. These features overlap with those of APN deficiency, supporting a mechanistic link between reduced APN levels and AD pathology consistent with the type 3 diabetes hypothesis.

Transcriptomic dysregulations in brain ECs are characterized by enhanced immune activation and suppressed metabolic capacity. Upregulated GO biological processes predominantly point to immune activation and microglial-mediated responses, implying that brain ECs are actively involved in neuroinflammation in both 5xFAD and APN^−/− mice. Several studies have reported the signaling axis between ECs and microglia, such as CXCL13/CXCR5 [82] and CSF1/CSF1R [83] signaling. Nevertheless, crosstalk between brain ECs and microglia remains largely unknown, especially in the context of AD. In addition, the activation of apoptotic pathways, such as the upregulated hippocampal neuron apoptotic signaling, suggests that endothelial dysfunction indirectly increases neuronal vulnerability through heightened oxidative stress and inflammation. Conversely, the downregulated GO biological processes primarily involved dysfunctional energy metabolism and impaired cellular homeostasis, which are known to compromise EC survival and BBB integrity [84]. Therefore, the early-onset CAA observed in APN^−/− mice can be attributed to the BBB breakdown and an exaggerated inflammatory response in ECs, which impair Aβ clearance and promote its vascular accumulation. Suppression of erythrocyte development and oxygen transport indicates a hypoxic state of ECs, which relates to cerebral hypoperfusion, impaired angiogenesis, and disturbances in the eNOS-NO pathway related to vasodilation [85–87]. Erythropoietin, which stimulates red blood cell production in response to hypoxia, was reported to improve memory function with reduced endothelial dysfunction and Aβ burden in AD models [88]. Together, these shared transcriptomic signatures in 5xFAD and APN^−/− mice reveal altered inflammatory signaling and metabolic regulation in ECs during CVD, leading to BBB breakdown, neurovascular uncoupling and neurodegenerative progression [89–91].

The exacerbated CVD in APN-deficient 5xFAD mice is due to the additive effects of APN loss and Aβ toxicity. APN-deficient 5xFAD mice exhibited earlier BBB breakdown by 3 months, accelerated NVC impairment and enhanced CAA pathology by 6 months, reduced resting CBF by 9 months compared to 5xFAD controls. Moreover, these mice displayed greater neuronal loss and cognitive decline by 9 months [23]. The intertwined relationships of BBB breakdown, NVC impairment, and CAA pathology display a multifaceted scenario of neuronal toxicity. Each of these vascular insults exerts neuronal damage alone. Interestingly, they also initiate or exacerbate each other, functioning as a vicious self-reinforcing mechanism [92–94]. Together, the combined effects of APN deficiency and Aβ overexpression enhance the vicious cycle within CVD, accelerating and amplifying the process of neurodegeneration and cognitive decline (Fig. 9).

Altered endothelial signaling pathways underline the mechanisms of CVD deterioration in APN-deficient 5xFAD mice. Transcriptomic analysis revealed that four key signaling pathways were significantly downregulated in the brain ECs of 5xFAD; APN^−/− mice compared to 5xFAD controls, with no pathways upregulated. The apelin signaling pathway maintains endothelial homeostasis by eliciting AKT and eNOS activity [95], and promotes EC proliferation and angiogenesis by enhancing glycolytic metabolism [96]. Its suppression leads to CBF reduction by impaired eNOS-mediated vasodilation [97] and reduced vascular network [98]. Downregulation of the ECM-receptor interaction pathway indicates disruption of ECM remodeling, which is essential for BBB maintenance and endothelial homeostasis [99, 100]. Additionally, the PI3K-Akt signaling pathway, known for its involvement in cell survival, proliferation, NO production, and anti-apoptotic mechanisms [101, 102], was also significantly suppressed. This suggests the impaired NO-mediated vasodilation, exacerbated EC apoptosis, and compromised cellular repair mechanisms, further worsening vascular integrity in AD. Lastly, the AGE-RAGE signaling pathway, linked to oxidative stress and inflammation [60, 103], was downregulated, reflecting a protective or compensatory exhaustion response to chronic Aβ-induced endothelial oxidative stress and inflammation exacerbated by APN deficiency [104–108]. These four downregulated pathways collectively imply the underlying mechanisms by which APN deficiency exacerbates endothelial dysfunction in AD, highlighting the critical role of ECs in cerebrovascular integrity and neurodegenerative progression.

In this study, we demonstrated the instant vasoactive property of APN in terms of increased CBF and restored NVC, which is related to the enhanced expression levels of eNOS. APN trimer is structurally stable in the blood, with a half-life of approximately 32.4 min in mice, which is sufficient for circulation and influx across the BBB [109]. Several other studies have shown acute metabolic changes after APN administration. For example, APN has been reported to stimulate insulin secretion after 20 min of i.v. administration [110]. Another study has also demonstrated the acute effects of i.v. APN treatment in mice, showing a significant decrease in serum glucose, along with increases in energy expenditure and in the expression of thermogenin and hypothalamic corticotropin-releasing hormone [111]. Most recently, APN has been reported to modulate purinergic signaling in smooth muscle tissues [112], a pathway that can also acts in ECs, astrocytes and VSMCs to mediate rapid vasodilatation occurring on a timescale from milliseconds to seconds [113]. The present study shows that APN can enhance eNOS level in vivo, which may lead to increased NO generation, a known promoter of vasodilation and CBF increase [114]. Nonetheless, direct functional evidence to unequivocally establish this mechanism remains to be elucidated. In addition, we anticipate that chronic administration of APN may alleviate the reduction in CBF in AD mice through long-term effects, such as stimulating angiogenesis, suppressing endothelial inflammation [53], and enhancing the crosstalk between ECs and microglia [115].

We acknowledge that the effects of systemic APN deficiency and AD pathology can manifest across multiple cell types and stages, extending beyond ECs. Our previous study demonstrated that APN deficiency aggravates microglial activation and neuroinflammation in AD mice. Specifically, APN inhibits the inflammatory response of microglia to Aβ oligomers via AdipoR1-AMPK-NF-κB signaling [116]. However, the impact of APN on other brain cell types, especially those involved in the neurovascular unit, and how it influences the crosstalk between ECs and other brain cells remains a crucial area for future investigation. In addition, we plan to identify the key driving genes within the downregulated pathways associated with APN deficiency in AD pathology and validate their expression at the protein level, which will help bridge transcriptional changes to pathological outcomes and guide the development of specific molecular therapeutic strategies. Furthermore, the results of CBF measurements could be strengthened by including central cardiovascular parameters of the animals, such as heart rate, blood pressure, and arterial CO₂, to help to account for individual physiological variability.

In summary, our study elucidates the pathophysiological effects of APN deficiency on CVD, paralleling that observed in 5xFAD mice, and demonstrates that APN loss exacerbates CVD in AD models. Importantly, administration of APN restores endothelial function under Aβ toxicity, suggesting a promising therapeutic approach to mitigate CVD and subsequent neurodegeneration in AD.

Clinical implications

Although most animal studies support the beneficial effects of APN in AD, human studies present more complex and conflicting results. Several potential causes for the controversial effects of APN in dementia have been proposed, including diverse patient classification criteria, medications that may influence circulating APN levels, inconsistency in study design and populations, insufficient adjustments for confounding factors, lack of attention to different APN isomers, and overlooking resistance to APN resulting from elevated circulating APN levels [64]. These necessitate further research to clarify the effects of APN and its different isomers in AD patients. Crucially, instead of assessing cognition, CVD, as an early onset pathological trait of AD, ought to be associated with the levels of APN in the serum or CSF of human. Low circulating APN levels (hypoadiponectinemia), despite their close association with vascular risk factors for AD, are largely under-recognized in clinical practice. Screening CVD patients, especially those with diabetes, hypertension, or hypercholesterolemia, for hypoadiponectinemia offers an opportunity for early intervention to maintain or increase APN levels. This may help prevent or delay AD progression, which currently has limited treatment options.

Alleviating endothelial dysfunction presents a promising therapeutic strategy for treating AD, with benefits that extend beyond CVD. A recent human transcriptomic study has highlighted a direct link between EC pathology and AD development, as ECs are enriched with AD risk genes [117]. Key therapeutic targets in ECs include inflammatory activation, BBB disruption, mitochondrial dysfunction, cellular senescence, apoptosis, diminished endothelial Aβ clearance, vascular fragility, and impaired angiogenesis [9, 118, 119]. Modulating these pathways may help slow or prevent cognitive decline during both cerebrovascular aging and AD pathogenesis. Endothelial progenitor cells have been demonstrated to improve learning and memory, attenuate Aβ deposition, and enhance angiogenesis in AD models by facilitating endothelial repair [106]. Furthermore, endothelin-1 receptor antagonists and β3-adrenergic agonists have been shown to suppress inflammation, reduce vascular dysfunction, and improve cognitive function [120, 121]. Although these studies are in preclinical stages, they offer a promising intervention strategy for alleviating CVD-related AD pathophysiology, by addressing endothelial dysfunction driven by APN reduction in individuals with vascular risk factors.

Abbreviations

Aβ

Amyloid-beta

AD

Alzheimer’s disease

AGEs-RAGE

Advanced glycation end-products (AGEs) and their receptor (RAGE)

APN

Adiponecin

BBB

Blood-brain barrier

CAA

Cerebral amyloid angiopathy

CBF

Cerebral blood flow

CVD

Cerebrovascular dysfunction

DEGs

Differentially expressed genes

EC

Endothelial cell

ECM

Extracellular matrix

eNOS

Endothelial nitric oxide synthase

GO

Gene oncology

KEGG

Kyoto encyclopedia of genes and genomes

LSCI

Laser speckle contrast imaging

NaFl

Fluorescein sodium salt

NO

Nitric oxide

NVC

Neurovascular coupling

PI3K-Akt

Phosphoinositide 3-kinase–Akt

RFUs

Raw fluorescence units

ROI

Region of interest

TJP

Tight junction protein

ZO

Zonula occludens

Authors’ contributions

W.Z. and K.H.C. conceived and designed the study. L.W.Y. and K.H.C. gave comments on the manuscript. W.Z., J.SC.K., Z.Z., H.X. performed the experiments. W.Z. analyzed the data and wrote the manuscript. All authors revised and approved the manuscript.

Funding

This work was supported by funding for research in Alzheimer’s Disease and dementia from Chan Kin Shing Charitable Trust and private donation from Fung WCS.

Data availability statement.

All the data during the current study have been shown in the manuscript and supplemental materials, and unprocessed data are available from the corresponding author on reasonable request.

Declarations.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

All animal procedures were approved by the Committee on the Use of Live Animals in Teaching and Research of The University of Hong Kong.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.