Endothelial NAD⁺ depletion drives vascular senescence and neuroinflammation via mtDNA‐cGAS/STING‐CD38 signaling in Alzheimer's disease

Abstract

BACKGROUND

Endothelial dysfunction has emerged as early and pivotal event in Alzheimer's disease (AD), yet the molecular mechanisms linking vascular aging to neuroinflammation remain elusive.

METHODS

We used APP/PS1 mice and amyloid beta (Aβ)–challenged brain endothelial cells (BECs) to understand the mechanisms of nicotinamide adenine dinucleotide (NAD⁺) deficiency, and its relationship with endothelial senescence and neuroinflammation in AD pathology. Nicotinamide riboside supplementation was administered to APP/PS1 mice to determine whether restoration of NAD⁺ homeostasis mitigates AD‐related vascular and inflammatory pathology.

RESULTS

NAD⁺ deficiency induced voltage‐dependent anion channel 1 (VDAC1) oligomerization, mitochondrial DNA (mtDNA) leakage, and cGAS/STING‐IRF3 activation, promoting endothelial senescence and SASP production with NAD⁺‐consuming enzyme CD38 upregulation. Senescent BECs triggered IL‐6‐dependent microglial activation. NR treatment restored mitochondrial integrity, suppressed cGAS‐STING signaling, and reduced neuroinflammation, improving vascular function and cognition.

DISCUSSION

Aβ‐driven NAD⁺ deficiency initiates a VDAC1‐mtDNA‐cGAS/STING cascade that promotes endothelial senescence and neurovascular inflammation in AD pathology, and amplifies neuroinflammation through BEC–microglia crosstalk, highlighting NAD⁺ restoration as a promising AD therapeutic strategy.

Highlights

• NAD⁺ deficiency disrupts mitochondrial metabolism and induces voltage‐dependent anion channel 1 (VDAC1) oligomerization–mediated mitochondrial DNA (mtDNA) leakage in the brain endothelial cells of Alzheimer's disease (AD) mice.

• Mitochondrial dysfunction–driven activation of the mtDNA‐cGAS/STING pathway promotes endothelial senescence and contributes to neurovascular pathology in AD.

• Senescent endothelial cells amplify neuroinflammation via IL‐6‐dependent activation of the microglial IL‐6R/STAT3/NF‐κB signaling cascade, promoting perivascular clustering.

• NR supplementation restores vascular integrity and cognitive function by suppressing mtDNA‐cGAS/STING signaling, preventing CD38‐mediated NAD⁺ depletion, and disrupting maladaptive endothelial–microglial crosstalk.

1. BACKGROUND

Alzheimer's disease (AD) is the leading cause of dementia and a major global health challenge, with more than 50 million people affected worldwide and cases projected to triple by 2050. ¹ , ² Decades of research have illuminated its classical neuropathological hallmarks, extracellular amyloid beta (Aβ) plaques and intracellular neurofibrillary tangles (NFTs) of hyperphosphorylated tau, yet disease‐modifying therapies have remained largely unsuccessful. This enduring gap between pathological understanding and therapeutic success underscores the urgent need to look beyond the traditional amyloid–tau framework.

Mounting evidence implicates cerebrovascular dysfunction as an early and pivotal driver of AD pathogenesis, often preceding overt Aβ or tau deposition. ³ , ⁴ Vascular abnormalities such as cerebral amyloid angiopathy (CAA), microvascular rarefaction, and blood–brain barrier (BBB) breakdown are observed in over half of patients with AD and strongly correlate with cognitive decline. ⁴ , ⁵ Within the neurovascular unit (NVU), brain endothelial cells (BECs), which form the luminal surface of cerebral vessels and constitute the BBB, are exceptionally susceptible to aging and pathological insults. ⁶ , ⁷ , ⁸ , ⁹ Once compromised, BECs are not passive victims but active propagators of NVU breakdown and neuroinflammation, thereby amplifying neuronal injury. ¹⁰ Yet the molecular triggers of endothelial dysfunction in AD remain poorly defined.

Endothelial senescence has emerged as a central pathological nexus linking vascular dysfunction to neurodegeneration in AD. ⁷ , ¹¹ , ¹² , ¹³ Typically, senescent cells are characterized by DNA damage, mitochondrial dysfunction, metabolic derangement, and the secretion of a proinflammatory senescence‐associated secretory phenotype (SASP). ¹⁴ Of particular relevance to vascular pathology, Aβ1‐40 peptides, the predominant amyloid isoform deposited along cerebral vessels, ¹⁵ , ¹⁶ have been shown to induce oxidative stress and DNA damage in BECs, thereby accelerating endothelial senescence, BBB leakage, and NVU impairment. ⁶ , ⁷ , ⁸ , ¹⁷ Mitochondrial dysfunction, an early event in AD pathogenesis, ¹⁸ , ¹⁹ is a potent inducer of endothelial senescence and triggers the cytosolic leakage of oxidized mitochondrial DNA (mtDNA). ¹⁶ Once in the cytosol, mtDNA acts as a potent damage‐associated molecular pattern (DAMP) that activates cyclic GMP‐AMP synthase (cGAS). Upon activation, cGAS catalyzes the synthesis of cyclic GMP‐AMP (cGAMP), which subsequently engages the stimulator of interferon genes (STING) pathway. ¹⁹ , ²⁰ cGAS/STING activation elicits robust innate immune responses, including type I interferon signaling and NF‐κB‐driven SASP production, that perpetuate vascular inflammation and dysfunction. ²¹ , ²² , ²³ Although the cGAS/STING axis has been implicated in various chronic inflammatory disorders, ¹⁹ , ²¹ , ²² , ²⁴ , ²⁵ , ²⁶ , ²⁷ its contribution to AD‐related cerebrovascular pathology remains poorly defined.

Nicotinamide adenine dinucleotide (NAD⁺), a central metabolic cofactor, is indispensable for mitochondrial homeostasis, DNA repair, and genomic stability. Declining NAD⁺ levels are a hallmark of aging and are observed prominently in neurodegenerative disorders, including AD. ⁹ , ²⁰ , ²⁷ NAD⁺ depletion exacerbates mitochondrial dysfunction, oxidative stress, and genomic instability. Conversely, NAD⁺ replenishment has emerged as a promising geroprotective intervention, restoring mitochondrial function, ²⁷ enhancing DNA repair, ²⁰ and attenuating inflammatory signaling. ²⁸ However, whether NAD⁺ supplementation can modulate mtDNA‐driven cGAS/STING activation in BECs and thereby preserve cerebrovascular integrity in AD remains unclear.

RESEARCH IN CONTEXT

1. Systematic review: Cerebrovascular dysfunction and endothelial injury are recognized increasingly as early and important features of Alzheimer's disease (AD) and are thought to contribute to neuroinflammation and neurodegeneration. Emerging evidence suggests that endothelial senescence may represent a key mechanistic link between vascular aging and AD pathology. Mitochondrial dysfunction and activation of innate immune pathways such as cGAS/STING have been implicated in chronic inflammation and neurodegenerative disorders. However, whether endothelial NAD⁺ depletion contributes to mitochondrial DNA (mtDNA)–driven innate immune activation and neurovascular inflammation in AD remains unclear.

2. Interpretation: We show that amyloid beta (Aβ)–induced nicotinamide adenine dinucleotide (NAD⁺) depletion in brain endothelial cells triggers mitochondrial dysfunction characterized by mtDNA release and activation of an mtDNA‐cGAS/STING‐CD38 signaling axis. This pathway promotes endothelial senescence and inflammatory signaling and enhances neuroinflammation through endothelial–microglial communication. Restoration of NAD⁺ levels with nicotinamide riboside disrupts this pathogenic cascade, reduces neuroinflammation, and improves cerebrovascular function and cognitive performance in AD mice.

3. Future directions: Future studies should determine whether targeting endothelial NAD⁺ metabolism can prevent or delay neurovascular dysfunction during AD progression. Further investigation of additional NAD⁺‐boosting interventions and their therapeutic potential in early‐stage AD will be important for translating these findings into clinical strategies.

Here, we identify mtDNA‐induced innate immune activation as a driver of endothelial senescence and vascular inflammation in AD. Our study identifies a mtDNA‐cGAS/STING‐CD38 signaling cascade that drives endothelial senescence and propagates neuroinflammation through endothelial–microglial crosstalk. These findings establish vascular NAD⁺ restoration as a mechanistic and therapeutic avenue for mitigating neurovascular dysfunction in neurodegeneration.

2. METHODS

2.1. Animals and drug administration

APPswe/PSEN1dE9 double‐transgenic mice (APP/PS1; MMRRC Stock No. 034829‐JAX) that express mouse/human amyloid precursor protein (Mo/HuAPP695swe) and mutant human presenilin1 (PS1‐dE9) were used as an AD model for this study. Wild‐type (APPwt) littermates served as genetic controls. Genotyping was performed by polymerase chain reaction (PCR) using the following primers: APP (Forward 5′‐GACTGACCACTCGACCAGGTTCTG‐3′, Reverse 5′‐CTTGTAACTTGGATTCTCATATCCG‐3′), and PSEN1 (Forward 5′‐ AATAGAGAACGGAGGAGCA‐3′, Reverse 5′‐ GCCATGAGGGCACTAATCAT‐3′). Mice were bred and maintained under specific pathogen‐free (SPF) conditions with controlled environmental parameters (temperature 22 ± 0.5°C, humidity 55 ± 5%, 12‐hour light/dark cycle). All animals received autoclaved standard chow and water ad libitum. Experimental procedures were approved by the Institutional Animal Care and Use Committee of The Second Affiliated Hospital of Nanchang University (Approval No. NCULEA‐20221031020).

Nicotinamide riboside (NR; Selleck, #S2935) was administered via autoclaved drinking water at 12 mM concentration, prepared twice weekly in amber bottles to prevent photodegradation, continuously from 3 to 6 months of age, based on previously published protocols. ²⁰ , ²⁷ Control groups received regular drinking water without NR supplementation. In this study, APP/PS1 and their wild‐type littermates were assigned randomly to four groups: APPwt (wild‐type + vehicle), APPwt + NR (wild‐type + 12 mM NR), APPtg (APP/PS1 + vehicle), and APPtg + NR (APP/PS1 + 12 mM NR). Behavioral assessments and molecular analyses were performed at 6 months of age.

2.2. Mouse cerebral vessels RNA sequencing (RNA‐seq) and bioinformatic analysis

Transcriptome RNA sequencing (RNA‐seq) was performed on vessel‐enriched fractions isolated from APPtg and APPtg + NR mice brain (n =  3 per group). Briefly, total RNA was extracted from the isolated mouse cerebral vessels. Quality assessment was performed via NanoPhotometer spectrophotometer (Implen N60Touch) for quantification and Qsep400 system (Bioptic) for integrity verification. Stranded RNA‐seq libraries were constructed following standard Illumina protocols and sequenced on the NovaSeq 6000 platform (Illumina, Shanghai, China) with 150 bp paired‐end reads. All sequencing and subsequent bioinformatic analyses were performed by investigators blinded to the experimental groups. For bioinformatic analysis, raw read quality was evaluated using FastQC (v0.11.5) followed by alignment to the GRCm39 reference transcriptome. Differential gene expression analysis between groups was performed using DESeq2, with significantly differentially expressed genes (DEGs) defined as those satisfying p < 0.05 and |log₂ (fold change)| > 1. These DEGs were visualized in volcano plots via ggplot2 (v3.3.2) and subjected to functional enrichment analysis through Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway using clusterProfiler (v4.10.0), with significance thresholds set at nominal p < 0.05 and false discovery rate (FDR) < 0.05.

2.3. Cell treatment

Aβ Oligomer Preparation and treatment . For Aβ exposure studies, oligomeric Aβ1‐40 (Aβ; AnaSpec, #AS‐24235) was prepared by dissolving in Ham's F‐12 medium (Gibco #11765054) at 100 µM followed by 24‐h incubation at 4°C, and then diluted to 5 µM working concentration for treatments as described previously. ²⁹

Pharmacological Treatments . bEnd.3 cells received NR (ChromaDex) at 1 mM concentration as 2‐h pretreatment in serum‐reduced medium (2% fetal bovine serum; FBS) prior to 24‐hour co‐exposure with oligomeric Aβ1‐40. To specifically inhibit VDAC1 oligomerization, bEnd.3 cells were pretreated with VBIT‐4 (10 µM; MedChemExpress, #HY‐101966) for 24 h in serum‐reduced medium before being subjected to Aβ challenge. To induce mitochondrial DNA depletion, bEnd.3 cells were exposed to 2′,3′‐dideoxycytidine (ddC; Sigma‐Aldrich, #D5782; 10 µM) for 72 h prior to Aβ/NR treatments.

Conditioned Media Collection and Microglial Exposure . Conditioned media (CM) from bEnd.3 cultures were collected following 24‐h treatments, centrifuged at 300 × g for 5 min to remove cellular debris, filtered through 0.22 µm membranes (Millipore SLGV033RB), and applied to BV‐2 microglia for 24 h to evaluate endothelial–microglial crosstalk. All treatments were performed in serum‐reduced medium (2% FBS) with vehicle controls (0.1% dimethyl sulfoxide; DMSO) included for normalization.

IL‐6 Pathway Blockade . For IL‐6 pathway analysis, BV‐2 microglia were incubated with 10 ng/ml anti‐mouse IL‐6 neutralizing antibody (α‐IL‐6; R&D systems, #MAB406) or anti‐mouse IL‐6Rα blocking antibody (α‐IL‐6R; R&D systems, #AF1830) in CM‐containing medium from bEnd.3 cultures.

2.4. Western blot and VDAC1 expression, oligomerization analysis

Protein lysates from cerebral microvessels or cells were extracted in radio immunoprecipitation assay (RIPA) buffer (Beyotime, #P0013B) with protease/phosphatase inhibitors (Roche, #4693159001). Lysates (20 µg/lane) were separated by 10% SDS‐PAGE and transferred to PVDF membranes (Millipore, #ISEQ00010). For the visualization of voltage‐dependent anion channel 1 (VDAC1) oligomerization, 60 µg of protein was subjected to SDS‐PAGE. Membranes were blocked with 10% nonfat milk and subsequently incubated overnight at 4°C with primary antibodies (detailed in Table S1). After washing, membranes were probed with appropriate horseradish peroxidase (HRP)–conjugated secondary antibodies (1:5000; Proteintech, #SA00001‐1, # SA00001‐2). Signals were developed with ECL reagent (UElandy, #S6009M) and quantified using ImageJ 1.54p. GAPDH (1:5000; Proteintech, #60004‐1‐Ig) or α‐tubulin (1:5000; Proteintech, #11224‐1‐AP) served as loading control.

2.5. Immunofluorescence staining and quantification

Immunofluorescence staining was performed on 30‐µm sagittal brain sections following established protocols. ³⁰ , ³¹ Sections were permeabilized with 0.3% Triton X‐100 in phosphate‐buffered saline (PBS) for 30 min and blocked with 10% normal donkey serum (Jackson ImmunoResearch, #017‐000‐121) for 2 h at room temperature. Primary antibodies (Table S1) diluted in blocking buffer were applied overnight at 4°C. After PBS washes, sections were incubated with Alexa Fluor‐conjugated secondary antibodies (1:500; Invitrogen) for 2 h at room temperature. Nuclei were counterstained with DAPI (1 µg/mL; Sigma, #D9542) for 10 min. Sections were mounted with Fluoromount‐G (SouthernBiotech, #0100‐01) and cured for 24 h at 4°C.

Confocal z‐stacks (1‐µm intervals) were acquired using a Leica TCS SPE confocal microscope with 40× objectives under consistent settings, z‐stacks were collected across a total imaging depth of 15 µm to ensure consistent sampling between samples. Maximum‐intensity projections were generated for quantitative analysis.

For vascular quantification, CD31⁺ structures were segmented from projected images using ImageJ/FIJI (v1.54p) following background subtraction and global thresholding applied uniformly to all images. Hippocampal regions of interest (ROIs) were manually defined based on anatomical landmarks. Vascular area was calculated as the percentage of CD31⁺ pixels relative to the total ROI area using the “Analyze Particles” function in ImageJ. To assess microvascular network architecture, binarized CD31⁺ images were further analyzed using AngioTool (v2.0). Junction density was defined as the number of vessel branch points per unit area within the ROI. For marker quantification, fluorescence intensities of senescence markers (p21^WAF1/Cip1, p16^Ink4a), innate immune sensors (cGAS, STING), CD38, and 8‐OHdG were measured within CD31⁺ vascular regions using channel‐specific thresholds in ImageJ. Vessel‐associated microglia (VAMs) were identified as Iba1⁺ cells within 1 µm of CD31⁺ vasculature, with IL‐6Rα intensity quantified in VAMs using spectral unmixing to eliminate bleed‐through between channels. For each mouse, at least three brain sections and three to five randomly selected fields per section were analyzed. Quantification was performed in a blinded manner, and the mean value per animal was used for statistical analysis.

2.6. Measurement of mitochondrial enzyme activities and citrate content

The enzymatic activities of mitochondrial isocitrate dehydrogenase (IDH) and α‐ketoglutarate dehydrogenase (α‐KGDH), key enzymes in the tricarboxylic acid (TCA) cycle, were determined using specific assay kits (Solarbio; #BC2160 for IDH and #BC0710 for α‐KGDH). Enzyme activities were quantified by monitoring the reduction of NAD⁺ to NADH at 340 nm. Mitochondrial citrate levels were measured in mitochondrial fractions isolated from brain vessels using a Citrate Assay Kit (Solarbio, #BC2155), based on an enzymatic reaction that produces a colorimetric readout with peak absorbance at 545 nm. For all assays, mitochondrial protein concentration was determined by BCA method for normalization. IDH and α‐KGDH activities are expressed as nmol NADH generated per min per mg protein, and citrate content is expressed as nmol per mg mitochondrial protein.

2.7. Evaluation of cytosolic mtDNA levels

To quantify cytosolic mtDNA release, cytosolic fractions were isolated from cerebrovascular tissues or cultured endothelial cells. To eliminate potential contamination from nuclear DNA (nDNA), the fractions were treated with DNase I (1 U/µg; Thermo Fisher, #EN052) for 30 min at 37°C, followed by heat inactivation at 75°C for 10 min. Total DNA was extracted using the DNeasy Blood & Tissue Kit (Qiagen, #69504) with carrier RNA supplementation. Parallel whole‐cell DNA extracts were prepared from equal aliquots of the original homogenate. Quantitative PCR was performed using primers for nuclear genes (Tert, B2m) and mitochondrial genes (D‐loop, Non‐Numt, Cox1) both on cytosolic and whole‐cell extracts. The 18S rRNA gene was used as an endogenous reference for normalization within each sample. Cytosolic mtDNA levels were ultimately normalized to total cellular mtDNA content, as described previously. ²⁴ Primer sequences are provided in Table S2.

2.8. Statistical analysis

Data are presented as mean ± standard error of the mean (SEM). Statistical analyses were performed using GraphPad Prism (GraphPad Software, San Diego, CA, USA). Normality was assessed using the Shapiro–Wilk test. Comparisons between two groups were performed using unpaired two‐tailed Student's t‐tests, whereas comparisons among multiple groups were analyzed using one‐way or two‐way analysis of variance (ANOVA) followed by appropriate post hoc tests, as specified in the figure legends. Exact statistical tests, sample sizes (n), and p‐values are reported in the corresponding figure legends.

Sample sizes were determined based on prior experience with these models and previous studies using similar experimental paradigms. Behavioral testing, immunohistochemical quantification, and image‐based analyses were conducted by investigators blinded to group allocation, and group identities were revealed only after completion of statistical analyses. A p‐value < 0.05 was considered statistically significant.

2.9. Other experimental methods

Detailed experimental procedures for the Morris water maze (MWM) behavioral assessment, BBB permeability assay, Aβ immunostaining and quantification, laser speckle contrast imaging with whisker stimulation, tissue collection, isolation of cerebral vessels from mouse brain, RNA extraction and quantitative real‐time PCR, cell culture and transfection, measurement of NAD⁺ levels and NAD⁺/NADH ratio, enzyme‐linked immunosorbent assay (ELISA), flow cytometry, mitochondrial membrane potential detection, reactive oxygen species measurement, cytosolic extraction, and senescence‐associated β‐galactosidase assay are described in the Supplementary Methods.

3. RESULTS

3.1. Reduced NAD⁺ levels in cerebral vessels of APP/PS1 mice accompanied by endothelial cell loss

Progressive NAD⁺ depletion has been recognized as a critical driver of cellular dysfunction in aging and neurodegeneration. ²⁷ , ³² However, whether cerebral vascular endothelial cells undergo NAD⁺ metabolic alterations during AD pathogenesis remains unclear. To address this, we isolated cerebral microvessels from 1‐, 3‐, and 6‐month‐old APP/PS1 (APPtg) mice and wild‐type littermates (APPwt). Biochemical analyses revealed a significant reduction in total NAD⁺ content and the NAD⁺/NADH ratio in the brain vasculature of 3‐month‐old APPtg mice compared to APPwt controls, whereas no significant difference was observed at 1 month of age (Figure 1A, B). Notably, vascular NAD⁺ levels as well as NAD⁺/NADH ratio progressively declined with advancing age in APPtg mice (Figure 1A, B).

FIGURE 1.

Reduced nicotinamide adenine dinucleotide (NAD⁺) levels in cerebral vessels of APP/PS1 mice, and restoration by Nicotinamide riboside (NR) supplementation. (A, B) Total NAD⁺ content (A) and NAD⁺/NADH ratio (B) in isolated cerebral microvessels from 1‐, 3‐, and 6‐month‐old APPwt and APP/PS1 (APPtg) mice (n ≥5 per group). (C) Representative flow cytometry plots and quantification of CD31⁺ endothelial cells in whole‐brain single‐cell suspensions from 6‐month‐old APPwt and APPtg mice (n = 6 per group). (D, E) Total NAD⁺ levels (D) and NAD⁺/NADH ratio (E) in vessel‐enriched brain fractions from APPwt, APPwt + NR, APPtg, and APPtg + NR mice (n = 5 per group). Data are presented as mean ± SEM. Statistical analyses was determined by unpaired two‐tailed Student's t‐test (A–C), and one‐way ANOVA followed by Tukey's multiple comparisons test (D, E). p values are indicated in the figure.

Flow cytometry further demonstrated a marked decrease in CD31⁺ endothelial cell populations in APPtg mice, indicating progressive endothelial cell loss in the cerebral vasculature (Figure 1C). Although previous studies have shown that NAD⁺ supplementation mitigates AD‐related pathology and neuroinflammation, ²⁰ , ²⁷ , ³³ whether NAD⁺ restoration impacts cerebrovascular endothelium remains unclear. To investigate this, 3‐month‐old APP/PS1 mice were administered with nicotinamide riboside (NR; 12 mM in drinking water) for 3 months. Of interest, NR treatment effectively restored NAD⁺ levels and markedly increased the NAD⁺/NADH ratio in the brain vessels of APP/PS1 mice (Figure 1D), and was also dramatically elevated in NR‐treated APP/PS1 mice (APPtg: 4.608 ± 3.328% vs APPtg + NR: 96.54 ± 3.206%, p < 0.0001; Figure 1E).

3.2. NAD⁺ attenuates vascular senescence and improves blood‐brain barrier integrity in APP/PS1 mice

Endothelial senescence and loss are recognized hallmarks of vascular aging and AD, contributing to BBB breakdown and exacerbated neuroinflammation. ³⁴ , ³⁵ Our flow cytometric analysis revealed that NR‐treated APP/PS1 mice exhibited a significantly higher proportion of CD31⁺ BECs in whole‐brain suspensions than untreated APP/PS1 mice (APPtg: 1.123 ± 0.1944% vs. APPtg + NR: 2.258 ± 0.335%, p = 0.03; Figure 2A), suggesting preservation of endothelial cell populations. Histochemical senescence‐associated β‐galactosidase (SA‐β‐gal) staining of isolated brain vessel preparations revealed increased senescence‐associated activity in APP/PS1 mice relative to wild‐type controls, which was attenuated by NR treatment (Figure 2B), indicating reduced vascular senescence. Consistently, NAD⁺ supplementation significantly downregulated the expression of canonical senescence markers p21^WAF1/Cip1 (Cdkn1a) and p16^Ink4a (Cdkn2a) at both the transcriptional (Figure S1C) and protein levels in the vessel‐enriched brain fractions from APP/PS1 mice, although NR supplementation did not significantly alter senescence marker expression in APPwt control mice (Figure 2C–E).

FIGURE 2.

NAD⁺ supplementation alleviates endothelial senescence in APP/PS1 mice. Six‐month‐old APP/PS1 transgenic mice (APPtg) and age‐matched wild‐type littermates (APPwt) received nicotinamide riboside (APPtg + NR, APPwt + NR) or vehicle treatment (APPtg, APPwt) as indicated. (A) Representative flow cytometry plots and quantification of CD31⁺ endothelial cells isolated from whole‐brain single‐cell suspensions of APPtg and APPtg + NR mice. (B) Representative images of SA‐β‐galactosidase staining in cerebral vessels from APPwt, APPwt + NR, APPtg and APPtg + NR mice. (C) Representative western blots showing p16^Ink4a and p21^WAF1/Cip1 protein levels in isolated cerebral vessels. (D, E) Densitometric quantification of p16^Ink4a (D; n = 6 mice per group) and p21^WAF1/Cip1 (E; n = 6 mice per group) normalized to loading controls. (F, G) GO (F), and KEGG (G) enrichment analyses of downregulated differentially expressed genes (DEGs) from RNA‐seq of cerebral vessels comparing APPtg + NR and APPtg mice (n = 3 per group). (H) Heatmap showing relative expression of senescence‐associated genes (overlapping between Aging_Atlas and DEGs) in isolated cerebral vessels from APPtg and APPtg + NR mice (n = 3 per group); the color scale represents normalized expression levels, blue indicating low expression and red indicating high expression. (I‐L) Representative immunofluorescence images of CD31 (green) co‐localized with p21^WAF1/Cip1 (I; red) or p16^Ink4a (K; red), with corresponding quantification (J, L) in the hippocampus and cortex of APPtg and APPtg + NR mice; nuclei counterstained with DAPI (blue). Data are presented as mean ± SEM. Statistical analyses were performed using unpaired two‐tailed Student's t‐test for (A), and one‐way ANOVA followed by Tukey's multiple comparisons test for (D, E, J, L). p‐values are indicated in the figure.

To further characterize transcriptional changes associated with NAD⁺ restoration in the cerebral vasculature, we performed bulk RNA sequencing on vessel‐enriched fractions isolated from NR‐treated and untreated APP/PS1 mice brain. Differential expression analysis identified 664 significantly altered genes (|log₂FC| > 1.0, p < 0.05), with 77 transcripts upregulated and 587 downregulated in the APPtg + NR group (Figure S1D). GO and KEGG pathway enrichment analysis of the downregulated genes revealed obviously suppression of cellular senescence, immune response, and inflammatory pathways, including innate immune response, cellular senescence, NF‐κB signaling, TNF signaling, IL‐17 signaling, and the cytosolic DNA‐sensing pathway (Figure 2F, G). Cross‐referencing these DEGs with aging‐related gene sets from the Aging Atlas database ³⁶ identified 26 overlapping genes modulated by NR administration, including proinflammatory SASP components such as Il‐6, Tnf, Cxcl2, Vcam1, and Mmp9, alongside classical senescence markers (Cdkn1a, Cdkn2a, Cdkn2b) (Figure 2H). These data indicate that NAD⁺ supplementation broadly attenuates senescence‐ and inflammation‐associated transcriptional programs within cerebral vascular tissue. To determine whether these transcriptional changes were reflected at the level of cerebral endothelial cells, we next performed double immunofluorescence staining for CD31 together with p21^WAF1/Cip1 or p16^Ink4a. NR treatment markedly reduced senescence marker intensity within CD31⁺ cerebral vessels in APP/PS1 mice (Figure 2I‐L), but there was no effect in APPwt mice (Figure S1E, F), supporting a prominent endothelial contribution to the anti‐senescent vascular phenotype.

Senescent BECs can contribute to BBB dysfunction, with increased permeability and decreased tight junction proteins. ³⁷ In this study, we revealed that NAD⁺ supplementation restored the expression of tight junction proteins ZO‐1 and Occludin in the brain vessels of APP/PS1 mice (Figure S2A‐C). Functionally, Evans Blue extravasation assays demonstrated significantly reduced dye leakage in NR‐treated APP/PS1 mice, indicative of enhanced BBB impermeability (Figure S2D, E). These findings were further corroborated by sulfo‐NHS‐biotin tracer assays, which showed diminished perivascular tracer leakage around CD31⁺ vessels in the NR‐treated group (Figure S2F, G), consistent with improved BBB integrity.

3.3. NAD⁺ restrains cGAS/STING‐dependent SASP production in the cerebral vascular endothelium of APP/PS1 mice

Transcriptomic profiling revealed robust downregulation of the cytosolic DNA‐sensing pathway in the vessel‐enriched brain fractions of NR‐treated APP/PS1 mice (Figure 2G). Strikingly, multiple components involved the cGAS/STING pathway were suppressed, including Cgas (2.6‐fold, p = 0.002), Sting (7.0‐fold, p = 0.0002), Tbk1 (14.6‐fold, p = 0.002), and Irf3 (9.6‐fold, p < 0.0001) (Figure 3A, Figures S3A). Western blotting corroborated these transcriptomic changes, showing reduced cGAS, STING, phosphorylated TBK1 (Ser172), and phosphorylated IRF3 (Ser396) in isolated vessel‐enriched brain fractions from NR‐treated APP/PS1 mice (Figures 3B, C). Immunofluorescence staining further localized this suppression specifically to CD31⁺ endothelial cells. Compared with untreated APP/PS1 mice, NAD⁺ supplementation markedly diminished cGAS and STING expression within the hippocampal and cortical vasculature (Figure 4D‐G). Critically, these effects were specific to pathological APP/PS1 mice and not observed in NR‐treated wild‐type (APPwt) littermates (Figure 3B, C, E, G and Figure S3B, C).

FIGURE 3.

NAD⁺ supplementation suppresses cGAS/STING pathway activation in cerebral endothelial cells of APP/PS1 mice. (A) Heatmap of differentially expressed key cGAS/STING pathway‐related genes (such as Cgas, Sting1, Irf3) identified by RNA‐seq of cerebral vessel‐enriched fractions from APPtg and APPtg + NR mice (n = 3 per group). (B, C) Representative western blot image (B) and densitometric quantification of cGAS, STING, phospho‐TBK1^Ser172 (p‐TBK1), and phospho‐IRF3^Ser396 (p‐IRF3) in cerebral vessel‐enriched fractions from APPwt, APPtg, and APPtg + NR mice (C; n = 6 per group). (D–G) Representative immunofluorescence images of hippocampus and cortex from APPtg and APPtg + NR mice showing CD31 (green) co‐stained with STING (D, red) or cGAS (F, red); quantification of STING (E) and cGAS (G) fluorescence intensity within CD31⁺ cerebral vessels were shown (n = 5 or 6 mice per group); nuclei were counterstained with DAPI (blue). (H) qPCR analysis of SASP genes (Il6, Tnf, Il1b, Cxcl10, Cxcl2) in cerebral vessel‐enriched fractions from APPtg and APPtg + NR mice (n = 5 per group). (I) ELISA quantification of IL‐6, TNF‐α, and IL‐1β in the culture supernatants of bEnd.3 endothelial cells treated with vehicle control, NR, Aβ, or Aβ + NR (n = 6 per group). (J) SA‐β‐galactosidase staining of bEnd.3 endothelial cells transfected with control siRNA (si‐Ctrl), Cgas siRNA (si‐Cgas), or Sting1 siRNA (si‐Sting) followed by Aβ stimulation or vehicle control; representative images show SA‐β‐gal⁺ cells indicated by white arrows, with enlarged insets provided; the percentage of SA‐β‐gal⁺ cells were quantified (n = 5 per group). Data are presented as mean ± SEM. Statistical analyses were performed using one‐way ANOVA followed by Tukey's multiple comparisons test (C, E, G, I, J) or unpaired two‐tailed Student's t‐test (H). p‐values are indicated in the figure.

FIGURE 4.

NAD⁺ supplementation suppresses cGAS/STING activation by enhancing mitochondrial function and preventing cytosolic mtDNA leakage. (A) Quantification of mitochondrial membrane potential using JC‐1 staining in bEnd.3 endothelial cells treated with Aβ, Aβ + NR, or control conditions; representative images are shown in Figure S6 C (n = 5 per group). (B, C) Flow cytometric analysis of intracellular ROS levels in bEnd.3 cells under indicated treatments (n = 4 per group). (D) qPCR quantification of cytosolic mitochondrial DNA (mtDNA; D‐loop, Non‐Numt, Cox1) and nuclear DNA (nDNA; Tert, B2m) in cerebral vessel‐enriched fractions isolated from APPwt, APPwt + NR, APPtg, and APPtg + NR mice (n ≥5 per group). (E, F) Representative immunofluorescence images (E) and quantification (F) of co‐localization of CD31 (green) and oxidative DNA damage marker 8‐OHdG (red) in hippocampal and cortex of APPtg and APPtg + NR mice; nuclei were counterstained with DAPI (blue) (n ≥5 mice per group). (G) Quantification of cytosolic mtDNA and nDNA levels in bEnd.3 cells transfected with siRNA targeting control (si‐Ctrl), Cgas (si‐Cgas), or Sting1 (si‐Sting) followed by Aβ treatment (n = 4 per group). (H) Quantification of cytosolic mtDNA and nDNA levels in bEnd.3 cells treated with Aβ, Aβ + mtDNA depletion (ddC), or Aβ + ddC + NR (n = 4 per group). (I) Relative mRNA expression of SASP‐related cytokines (IL‐6, TNF‐α, IL‐1β, CXCL10, CXCL2) under the same treatment conditions as in (H) (n = 4 per group). (J, K) Western blot analysis (J) and quantification (K) of cGAS/STING pathway components (cGAS, STING, p‐TBK1, p‐IRF3) and tight junction proteins (ZO‐1, Occludin) in bEnd.3 cells under treatments with Aβ, Aβ + ddC, and Aβ + ddC + NR (n = 4 per group). Data are presented as mean ± SEM. Statistical significance was assessed using one‐way ANOVA followed by Tukey's multiple comparisons test. P‐values are indicated in the figure.

To contextualize the vascular inflammatory changes, we evaluated parenchymal and vascular‐associated Aβ pathology. In 6‐month‐old APP/PS1 mice, NR treatment significantly reduced the Aβ‐immunoreactive plaque area in both the hippocampus and cortex (Figure S3D, E; hippocampus, p = 0.0007; cortex, p = 0.0001), consistent with prior reports that NAD⁺ precursor supplementation mitigates amyloid pathology in AD models. ²⁰ , ²⁵ , ²⁷ In addition, vascular‐associated Aβ deposition was further assessed by co‐staining with the anti‐Aβ antibody 6E10 and the endothelial marker CD31. Perivascular Aβ accumulation was significantly attenuated in the NR‐treated group (Figure S3F, G).

Furthermore, to determine whether vascular alterations progressed with disease stage, we analyzed 12‐month‐old APP/PS1 mice. Compared with 6‐month‐old APP/PS1 mice, vessel‐enriched brain fractions from 12‐month‐old mice exhibited more pronounced NAD⁺ depletion (Figure S3H) and a lower NAD⁺/NADH ratio (Figure S3I), increased p16^Ink4a and p21^WAF1/Cip1 expression (Figure S3J), and stronger activation of the cGAS/STING pathway (Figure S3K)

Given that cGAS/STING signaling is a canonical driver of the senescence‐associated inflammatory responses, ²¹ , ²² we next examined SASP‐related genes expression. RNA‐seq revealed broad downregulation of pro‐inflammatory SASP transcripts, including Il6, Tnf, Il1b, Cxcl10, and Cxcl2, in cerebral vessels of NR‐treated APP/PS1 mice (Figure S4A). qPCR validated these findings, revealing 62.2‐fold reduction in Il6, 2.9‐fold in Tnf, 3.4‐fold in Il1b, and 4.2‐fold in Cxcl10 (Figure 3H). To mechanistically validate these observations, we employed an in vitro model using Aβ1‐40‐stimulated bEnd.3 endothelial cells to model AD‐associated endothelial stimulation. NR co‐treatment significantly attenuated Aβ‐induced SASP factor secretion, most notably reducing IL‐6 levels by 59.5% (APPtg vs APPtg + NR: 3748 ± 180.5 vs 1519 ± 182.2 pg/mL, p < 0.0001; Figure 3I).

To determine whether the anti‐senescent effects of NAD⁺ supplementation are mediated via the cGAS/STING pathway, we employed siRNA to knockdown cGAS or STING in Aβ‐stimulated bEnd.3 cells (Figure S4B). Remarkably, downregulated the expression of either cGAS or STING phenocopied the protective effects of NAD⁺ supplementation, significantly suppressing SA‐β‐gal activity (p < 0.0001, Figure 3J), reducing SASP gene expression (Il‐6, Tnf, Il1b, Cxcl10, and Cxcl2) (Figure S4C), and lowering the levels of the senescence markers p21^WAF1/Cip1 and p16^Ink4a (Figure S4D). Crucially, NR treatment failed to confer additional benefit in cGAS‐ or STING‐deficient cells (Figure S4E‐G), indicating that NAD⁺ exerts its protective effects predominantly via inhibition of cGAS/STING signaling.

Furthermore, we studied whether the NAD⁺‐cGAS/STING‐senescence axis is sex‐dependent; we evaluated female APP/PS1 mice using the same experimental paradigm. Female APP/PS1 mice exhibited reduced vascular NAD⁺ levels and NAD⁺/NADH ratio (Figure S5A, B), increased expression of senescence markers p21^WAF1/Cip1 and p16^Ink4a (Figure S5C), and enhanced activation of the cGAS/STING pathway (Figure S5D, E) in vessel‐enriched brain fractions. NR supplementation restored NAD⁺ levels as well as NAD⁺/NADH ratio (Figure S5A, B), and attenuated senescence marker expression (Figure S5C) and innate immune activation (Figure S5D, E) in the brain vessels from female mice.

3.4. NAD⁺ stabilizes mitochondrial metabolism and prevents VDAC1 oligomerization‐dependent mtDNA leakage to suppress cGAS/STING activation

Mitochondrial dysfunction is a well‐recognized driver of cellular senescence and neurodegenerative, including AD. ²² , ²³ To determine whether the protective effects of NAD⁺ involve mitochondrial stabilization, we analyzed transcriptomic profiling of brain vessels isolated from NR‐treated APP/PS1 mice. GO enrichment analysis highlighted robust upregulation of mitochondrial bioenergetic pathways, including ATP synthesis‐coupled electron transport, aerobic respiration, and NADH dehydrogenase activity (Figure S6A). Concordantly, KEGG pathway analysis also identified oxidative phosphorylation as one of the most enriched pathways following NAD⁺ supplementation (Figure S6B). To functionally validate these transcriptomic findings, we evaluated mitochondrial membrane potential (Δψm) using JC‐1 staining in Aβ‐stimulated bEnd.3 endothelial cells. Aβ exposure led to marked mitochondrial depolarization, reflected by a reduced red/green fluorescence ratio. In contrast, NAD⁺ supplementation significantly restored Δψm, indicating improved mitochondrial integrity (Figure 4A, and Figure S6C). In parallel, intracellular reactive oxygen species (ROS) were substantially elevated following Aβ treatment but were markedly suppressed by NAD⁺ supplementation, as measured by DCFH‐DA fluorescence (Figure 4B, C).

Cytosolic mtDNA accumulation, triggered by mitochondrial damage, in turn acts as a key activator of the cGAS/STING pathway. ²² To assess whether mitochondrial stabilization limits mtDNA release, we quantified cytosolic mtDNA in the brain vessels of APP/PS1 mice. qPCR analysis targeting the D‐loop, Non‐Numt, and Cox1 regions revealed significantly elevated cytosolic mtDNA levels in the APP/PS1 vessels, which was substantially reduced by NAD⁺ supplementation (Figure 4D). Similarly, oxidative DNA damage marker 8‐hydroxy‐2′‐deoxyguanosine (8‐OHdG) ¹⁹ was obviously increased in the brain vessels of APP/PS1 mice but normalized by NAD⁺ treatment (Figure 4E, F). To determine whether mtDNA leakage precedes cGAS/STING activation, we measured cytosolic mtDNA following siRNA‐mediated knockdown of cGAS or STING in Aβ‐stimulated endothelial cells. It is intriguing that knockdown of either protein did not affect cytosolic mtDNA levels (Figure 4G), yet still robustly reduced SASP gene level (Figure S4C) and downregulated senescence markers p21^WAF1/Cip1 and p16^Ink4a (Figure S4D), suggesting that mtDNA leakage occurs upstream of cGAS/STING activation. Consistently, pharmacological depletion of mtDNA with ddC reduced cytosolic mtDNA accumulation in Aβ‐challenged endothelial cells (Figure 4H), along with decreased SA‐β‐gal activity (Figure S6E), suppressed SASP gene expression (e.g., Il6, Tnf, Il1b) (Figure 4I), attenuated cGAS/STING activation, and restored tight junction proteins ZO‐1 and Occludin (Figure 4J, K). Of note, co‐treatment with ddC and NR did not yield additive effects, indicating that NAD⁺ protects endothelial cells primarily by maintaining mitochondrial homeostasis and preventing mtDNA leakage (Figure 4 H–K).

Having established that NAD⁺ stabilizes mitochondrial homeostasis and inhibits Aβ‐induced mtDNA leakage, we next sought to elucidate the mechanistic link between NAD⁺ metabolism and mtDNA leakage in AD pathology. Given that NAD⁺ serves as an essential cofactor for dehydrogenases in the TCA cycle, the enzymatic activities of ICDH and α‐KGDH were markedly reduced in cerebral vessels of APP/PS1 mice—by 90.64% and 58.24%, respectively—compared with APPwt controls, but were effectively restored following NAD⁺ supplementation (Figure 5A, B). Correspondingly, mitochondrial citrate, a metabolic intermediate upstream of α‐KGDH, accumulated markedly in APP/PS1 mice and was normalized upon NAD⁺ restoration (Figure 5C). Because mitochondrial metabolic stress and redox imbalance can destabilize outer membrane channels, we next examined the oligomerization of voltage‐dependent anion channel 1 (VDAC1), a critical conduit for mtDNA efflux. ³⁸ Indeed, pronounced VDAC1 oligomerization was observed in brain vessels of APP/PS1 mice and Aβ‐treated bEnd.3 endothelial cells, both of which were markedly suppressed by NAD⁺ supplementation (Figure 5D, F). To determine whether VDAC1 oligomerization mediates mtDNA efflux, we employed VBIT‐4, a selective inhibitor of VDAC1 oligomerization. Pharmacological blockade of VDAC1 oligomerization effectively prevented Aβ‐induced mtDNA leakage and subsequent cGAS/STING pathway activation, as evidenced by decreased cytosolic level of D‐loop, Non‐Numt, and Cox1 (Figure 5 E), along with reduced expression of cGAS, STING, and the phosphorylation of TBK1 and IRF3 (Figure 5G). Together, these results identify VDAC1 oligomerization as a critical structural event linking impaired mitochondrial NAD⁺ metabolism to mtDNA‐cGAS/STING‐mediated endothelial inflammation in AD.

FIGURE 5.

NAD⁺ restoration prevents VDAC1 oligomerization‐mediated mtDNA leakage and cGAS/STING pathway activation in AD. (A, B) Enzymatic activities of isocitrate dehydrogenase (ICDH) (A; n = 6 per group) and α‐ketoglutarate dehydrogenase (α‐KGDH) (B; n ≥5 per group) in mitochondrial‐enriched fractions from cerebral vessel‐enriched fractions of APPwt, APPwt + NR, APPtg, and APPtg + NR mice. (C) Citrate accumulation in mitochondrial‐enriched fractions from cerebral vessel‐enriched fractions of APP/PS1 mice, which was significantly increased compared to APPwt controls and normalized following NAD⁺ restoration (n ≥5 per group). (D) Immunoblot analysis showing enhanced VDAC1 oligomerization in cerebral vessel‐enriched fractions from APP/PS1 mice, which was markedly suppressed by NR treatment. (E) Quantification of cytosolic mtDNA fragments (D‐loop, Non‐Numt, and Cox1) in Aβ‐treated bEnd.3 endothelial cells with or without VBIT‐4 (VDAC1 oligomerization inhibitor) treatment (n = 4 per group). (F) Representative immunoblot showing VDAC1 oligomerization in Aβ‐treated bEnd.3 cells, which was suppressed by NR or VBIT‐4 treatment. (G) Immunoblot and densitometric quantification showing decreased expression of cGAS, STING, and phosphorylated TBK1 and IRF3 following VBIT‐4 treatment in Aβ‐treated bEnd.3 cells (n = 4 per group). Data are presented as mean ± SEM. Statistical analyses were performed using one‐way ANOVA followed by Tukey's multiple comparisons test. P‐values are indicated in the figure.

3.5. cGAS/STING‐IRF3 signaling induces CD38‐Mediated NAD⁺ depletion, which is reversed by NAD⁺ supplementation

CD38, a major NAD⁺‐consuming ectoenzyme, contributes to age‐related NAD⁺ decline and neurovascular dysfunction in neurodegenerative diseases. ²⁸ , ³⁹ Emerging evidence suggests that cGAS/STING activation transcriptionally upregulates CD38 expression through an IRF3‐dependent signaling. ²⁶ , ⁴⁰ Our transcriptomic profiling of isolated brain vessels revealed that NR supplementation obviously downregulated Cd38 expression among NAD⁺‐consuming enzymes without significantly affecting NAD⁺ biosynthetic genes (Figure S7A). Consistently, CD38 protein levels in the vessel‐enriched brain fractions were markedly elevated in APP/PS1 mice compared with wild‐type controls (APPwt vs APPtg: p = 0.0008) and were significantly reduced after NR treatment (Figure 6A, APPtg vs APPtg + NR: p < 0.0001). Immunofluorescence and flow cytometry further confirmed enhanced CD38 expression on CD31⁺ BECs in APP/PS1 mice, which was normalized by NR supplementation (Figure 6B‐D, and Figure S7D). Notably, NR had no effect on CD38 expression in wild‐type mice, indicating a disease‐specific response (Figure 6C, D, and Figure S7B, D).

FIGURE 6.

NAD⁺ supplementation attenuates CD38‐mediated NAD⁺ depletion by suppressing IRF3 activation in APP/PS1 mice. (A) Representative western blot images showing CD38 protein levels in cerebral vessel‐enriched fractions isolated from APPwt, APPwt + NR, APPtg, and APPtg + NR mice (n = 6 per group). (B) Representative immunofluorescence images of brain sections showing co‐localization of CD31⁺ vessels (green) and CD38 (red). (C) Quantification of CD38 fluorescence intensity in CD31⁺ vessels (n ≥ 5 per group). (D) Flow cytometric analysis of CD38 expression in CD31⁺ endothelial cells isolated from cerebral vessel‐enriched fractions (n = 4 or 5 per group). (E) Western blot analysis of p‐IRF3 and CD38 expression in bEnd.3 endothelial cells under control, NR treatment, Aβ stimulation, and Aβ + NR conditions (n = 3 per group). (F) CD38 expression in bEnd.3 cells transfected with si‐Ctrl or si‐Irf3 and stimulated with Aβ (n = 3 per group). (G) Western blot analysis of p‐IRF3 and CD38 expression in bEnd.3 cells transfected with control siRNA (si‐Ctrl), Cgas siRNA (si‐Cgas), or Sting1 siRNA (si‐Sting) under Aβ stimulation (n = 3 per group). Data are presented as mean ± SEM. Statistical analysis was performed using one‐way ANOVA followed by Tukey's multiple comparisons test. P‐values are indicated in the figure.

Mechanistically, Aβ exposure in bEnd.3 endothelial cells induced phosphorylation of IRF3 (Ser396) and concomitant CD38 upregulation, both of which were effectively suppressed by NAD⁺ supplementation (Figure 6E). Downregulating IRF3 via siRNA (Figure S7E) phenocopied the effects of NR, restoring CD38 to basal levels (Figure 6F). Similarly, knockdown of cGAS or STING prevented Aβ‐induced IRF3 phosphorylation and CD38 expression (Figure 6G), confirming that cGAS/STING‐IRF3 signaling drives CD38 induction. Functionally, Aβ exposure caused a substantial decline in intracellular NAD⁺, which was restored by NR supplementation and partially rescued by cGAS, STING, or IRF3 knockdown (Figure S7F, G).

3.6. Endothelial SASP factors drive microglial activation via the IL‐6/STAT3/NF‐κB pathway

Beyond maintaining the BBB, BECs also act as dynamic signaling hubs that shape neuroimmune communication through secreted mediators. ⁴¹ , ⁴² Consistent with previous reports that NAD⁺ supplementation promotes anti‐inflammatory microglial polarization in AD models, ²⁰ , ²⁷ we observed prominent perivascular accumulation of Iba1⁺ microglia in the hippocampus and cortex of APP/PS1 mice (Figure 7A), reflected by significantly increased vessel‐association index (VAMs/total microglia; Figure 7B). NAD⁺ supplementation markedly reduced both perivascular microglial clustering and total microglial density in APP/PS1 mice (Figure 7A, and Figure S8A, B), suggesting that endothelial restoration attenuates maladaptive endothelial–microglial crosstalk.

FIGURE 7.

NAD⁺ supplementation disrupts IL‐6‐mediated endothelial‐microglial inflammatory crosstalk in AD. (A) Representative immunofluorescence images showing co‐staining of microglial marker Iba1 (red) and endothelial marker CD31 (green) in the cortex and hippocampus of APP/PS1 mice; white arrows indicate perivascular microglia closely associated with cerebral vessels. (B) Quantification of the proportion of perivascular microglia relative to total microglia (n ≥ 5 per group). (C) Triple immunofluorescence staining of Iba1 (red), CD31 (green), and IL‐6R (gray) to visualize IL‐6R expression in perivascular microglia; yellow arrows indicate IL‐6R‐positive perivascular microglia. (D) Quantification of IL‐6R fluorescence intensity in vessel‐associated microglia (n ≥5 per group). (E–F) Western blot analysis (E) and densitometric quantification (F) of IL‐6R, JAK1, and phosphorylation levels of STAT3 and NF‐κB p65 in microglia stimulated with conditioned media from bEnd.3 cells treated with vehicle (Con), NR, Aβ, or Aβ + NR (n = 6 per group). (G–H) Western blot analysis (G) and quantification (H) of IL‐6R, JAK1, and p‐STAT3/p‐NF‐κB p65 in microglia co‐treated with Aβ‐challenged endothelial conditioned medium and isotype IgG, IL‐6‐neutralizing antibody (α‐IL‐6), or IL‐6R‐neutralizing antibody (α‐IL‐6R) (n = 4 per group). Data are presented as mean ± SEM. Statistical analysis was performed using one‐way ANOVA followed by Tukey's multiple comparisons test. P‐values are indicated in the figure.

Transcriptomic profiling identified IL‐6 as one of the most upregulated SASP factors in APP/PS1 mice cerebral vessels (Figure 2H), this finding recapitulated in Aβ‐challenged bEnd.3 cells (Figure 3I). Given the pivotal role of IL‐6 in microglial proinflammatory activation, we hypothesized that endothelial‐derived IL‐6 mediates paracrine microglial priming. Supporting this, VAMs in APP/PS1 mice exhibited upregulated IL‐6 receptor (IL‐6Rα) expression, which was significantly reduced by NAD⁺ supplementation (Figure 7C, D). Exposure of BV‐2 microglia to CM from Aβ‐stimulated bEnd.3 endothelial cells induced IL‐6Rα and JAK1 expression, along with robust phosphorylation of STAT3 (p‐STAT3) and NF‐κB p65 (p‐p65) (Figure 7E, F), hallmarks of canonical IL‐6‐mediated inflammatory signaling. Functionally, microglia treated with Aβ‐CM displayed a pronounced pro‐inflammatory transcriptional profile, characterized by elevated Cd86, Inos, Ifng, Tnf, and Il1b, and decreased anti‐inflammatory markers including Cd206, Il10, Tgfb, and Arg1 (Figure S8D). Moreover, neutralization of IL‐6 or IL‐6Rα effectively suppressed STAT3/NF‐κB activation (Figure 7G, H) and abolished the pro‐inflammatory transcriptional shift (Figure S8E), confirming that IL‐6 is the dominant effector mediating endothelial SASP‐induced microglial activation.

3.7. NAD⁺ Restoration improves cerebral perfusion, enhances vascular density, and rescues cognitive deficits in APP/PS1 mice

To identify whether NAD⁺ replenishment confers cognitive benefits in AD models, behavioral performance was assessed at 6 months of age using the MWM. During the 5‐day acquisition phase, NR‐treated APP/PS1 mice exhibited significantly reduced swimming distances and escape latencies compared with untreated APP/PS1 littermates (Figure 8A, B), along with increased swimming velocities (Figure 8C), indicating improved spatial learning. In the 5‐min probe trial, NR‐treated APP/PS1 mice remained for a significantly shorter time in their first visit to the region where the platform had been located, and exhibited more frequent crossings of the original platform region (Figure 8D, E), reflecting enhanced spatial memory retention. These findings are consistent with previous reports demonstrating that NAD⁺ augmentation ameliorates cognitive decline in AD models. ²⁰ , ²⁷ , ³³

FIGURE 8.

NAD⁺ supplementation improves cognitive function, cerebral perfusion, and hippocampal microvascular structure in APP/PS1 mice. (A–E) Behavioral performance in the Morris water maze test. Compared to untreated APP/PS1 mice (APPtg, n = 7), NR‐treated APP/PS1 mice (APPtg + NR, n = 8) exhibited reduced swimming distance (A), shorter escape latency (B), and increased swimming velocity (C) during the 5‐day acquisition phase. In the probe trial, APPtg + NR mice showed shorter latency to reach the former platform location (D) and crossed the former platform location more frequently (E). (F) Representative laser speckle contrast images of CBF at baseline and following left whisker stimulation in APPtg and APPtg + NR mice. (G, H) Quantification of resting CBF perfusion in the left and right hemispheres (G) and percentage change in CBF in response to left whisker stimulation (H). (I) Representative immunofluorescence images of CD31⁺ cerebral vasculature (green) and DAPI⁺ nuclei (blue) in the hippocampus, the corresponding vascular structure analysis is shown with vessel outlines indicated in red and branch points in blue. (J, K) Quantification of vascular area (% CD31⁺ area relative to total area) (J) and vascular junction density (K) in the hippocampus. Data are presented as mean ± SEM. Statistical analyses were performed using two‐way ANOVA followed by Tukey's multiple comparisons test for (A–C), unpaired two‐tailed Student's t‐test for (D, E, G, H), and one‐way ANOVA followed by Tukey's multiple comparisons test for (J, K). p‐values are indicated in the figure.

To determine whether these cognitive improvements were accompanied by cerebrovascular functional recovery, cerebral blood flow (CBF) were measured using laser speckle contrast imaging (LSCI). NR‐treated APP/PS1 mice exhibited significantly higher baseline CBF, with an increase of 25.1% in the left hemisphere (p = 0.006) and 28.7% in the right hemisphere (p = 0.0001), relative to untreated transgenic controls (Figure 8F, G). Furthermore, upon contralateral whisker stimulation, NR‐treated mice showed robust stimulus‐evoked CBF responses in the somatosensory cortex, in stark contrast to the blunted responses observed in untreated APP/PS1 mice (Figure 8F, H), indicating preserved neurovascular coupling. Consistent with this functional recovery, hippocampal CD31 immunofluorescence revealed microvascular remodeling in NR‐treated mice, characterized by a moderate increase in vascular density (APPtg vs APPtg + NR p = 0.0074; Figure 8 I, J) and a 39.34% increase in vascular junction density (p = 0.039; Figures 8I, K).

4. DISCUSSION

Cerebrovascular dysfunction, particularly impairment of BECs, is increasingly recognized as a critical early event in AD, often preceding overt amyloid and tau pathology and strongly correlating with cognitive decline. ³ , ⁴ , ¹⁰ , ⁴³ , ⁴⁴ However, the mechanisms linking vascular endothelial dysfunction to neurodegeneration remain largely unexplored. Here we show that restoration NAD⁺ levels via NR supplementation robustly rejuvenates cerebral endothelial function, preserves neurovascular integrity, and mitigates cognitive impairment in APP/PS1 mice, in part by reducing endothelial senescence and suppressing the proinflammatory SASP. Mechanistically, NAD⁺ replenishment stabilized mitochondrial homeostasis and prevented Aβ‐induced mtDNA leakage by restoring TCA cycle dehydrogenase activity and reducing VDAC1 oligomerization, thereby restraining cGAS/STING‐dependent endothelial inflammation. This, in turn, attenuates the expression of proinflammatory SASP factors and disrupts a feed‐forward loop involving CD38‐driven NAD⁺ catabolism. Furthermore, NAD⁺ restoration dampens IL‐6‐mediated endothelial–microglial crosstalk, mitigating neuroinflammation and breaking a self‐perpetuating neurovascular injury cycle.

The vascular contribution to AD is increasingly supported by genetic, pathological, and functional evidence. Recent transcriptomic mapping of the human brain vasculature revealed that 30 of the top 45 AD risk genes are enriched in vascular cell populations, particularly endothelial cells, pericytes, and smooth muscle cells. ⁴⁵ Clinically, more than 80% of patients with AD display cerebral amyloid angiopathy (CAA), in which vascular Aβ accumulation damages endothelial cells, disrupts BBB integrity, and destabilizes the NVU. ⁴⁶ In addition to vascular amyloid deposition, soluble Aβ can induce capillary constriction by activating pericytes, promote oxidative stress and inflammatory activation in endothelial cells, and directly trigger mitochondrial dysfunction, collectively impairing cerebral hemodynamics and neurovascular uncoupling. ⁶ , ⁷ , ¹⁰ , ⁴⁶ , ⁴⁷ , ⁴⁸ Extending these observations, our findings indicate that Aβ exposure induces endothelial senescence and mitochondrial dysfunction, initiating inflammatory signaling cascades that culminate in neurovascular dysfunction and cognitive impairment. Of note, these pathological phenotypes were mitigated significantly by NAD⁺ supplementation, highlighting endothelial metabolic decline as a potentially targetable driver of AD‐related vascular injury. Consistent with previous reports that NAD⁺ precursors can attenuate amyloid pathology in AD models, ²⁰ , ²⁵ , ²⁷ NR treatment in our study also modestly reduced Aβ plaque burden in APP/PS1 mice, with a notable decrease in perivascular Aβ accumulation. These findings suggest that NR may alleviate vascular amyloid stress in addition to improving endothelial metabolic function, although they do not establish a definitive temporal hierarchy between amyloid deposition and endothelial dysfunction during AD progression.

Therapeutic strategies aimed at rescuing endothelial function have shown promise in AD models. ⁹ , ⁴⁹ Prior studies have shown that pharmacological rejuvenation of endothelial function ⁵⁰ or genetic ablation of p16^Ink4a in endothelial cells ⁵¹ can markedly ameliorate AD‐like phenotypes in AD mouse models. Notably, NAD⁺ depletion, a hallmark of both aging and neurodegeneration, ⁹ , ³⁴ , ⁵² compromises mitochondrial bioenergetics and barrier function, and supplementation with NAD⁺ precursors such as NR or nicotinamide mononucleotide (NMN) has been reported to restore mitochondrial function, ²⁵ enhance vascular integrity, ⁹ and promote microglial mitophagy. ²⁰ Our study confirms that NR supplementation reversed Aβ‐induced mitochondrial dysfunction and oxidative stress while attenuating endothelial senescence. Transcriptomic analyses revealed enhanced mitochondrial bioenergetic pathways in the brain vessels of NR‐treated APP/PS1 mice, including oxidative phosphorylation and ATP synthesis. Indeed, restoration of NAD⁺ levels effectively normalized TCA cycle flux in AD brain vessels and prevented VDAC1 oligomerization, a conformational change implicated in facilitating mtDNA release. ³⁸ Given that BECs rely heavily on oxidative phosphorylation to maintain BBB integrity and regulate cerebral perfusion, ⁵³ preservation of mitochondrial metabolic capacity likely underlies the efficiency of NR in maintaining NVU homeostasis and vascular function in AD.

A key upstream driver of SASP expression in senescent endothelial cells is cytosolic mtDNA accumulation, which serves as a DAMP that activates the cGAS/STING pathway and perpetuates chronic inflammation. ²² , ²⁶ , ²⁷ Our results indicate that NR supplementation restores mitochondrial integrity and limits mtDNA leakage, thereby inhibiting cGAS/STING‐IRF3 signaling activation and dampening SASP production in BECs. Activated IRF3 also transcriptionally upregulates CD38, the major NAD⁺‐consuming enzyme that is markedly increased in BECs and microglia during aging and neurodegeneration. ⁵² , ⁵⁴ Moreover, increased CD38 expression not only accelerates NAD⁺ depletion but also exacerbates mitochondrial dysfunction, establishing a self‐reinforcing loop between metabolic decline and inflammation. ⁵² , ⁵⁴ , ⁵⁵ This mechanism likely explains our observation that, in AD pathology, NAD⁺ levels are reduced whereas CD38 expression is elevated in AD brain vessels, a dysregulation driven by the cGAS/STING‐IRF3 axis and effectively rescued by NAD⁺ supplementation.

Our study further suggests that senescent BECs may amplify neuroinflammation via IL‐6‐mediated endothelial–microglial crosstalk. Specifically, BECs rendered senescent by Aβ exposure secrete elevated levels of IL‐6, which activate STAT3 and NF‐κB signaling in adjacent microglia, promoting their pro‐inflammatory polarization. Disruption of this signaling axis by NR treatment or IL‐6/IL‐6R blockade markedly attenuated microglial activation, underscoring the therapeutic potential of disrupting endothelial‐derived inflammatory signaling. These findings identify a metabolically linked EC‐microglia axis through which endothelial dysfunction may propagate brain inflammation in AD. IL‐6 has emerged as a central mediator of endothelial–microglial crosstalk in aging and neurodegeneration. ⁵⁶ Accumulating evidence demonstrates that IL‐6 exposure induces profound phenotypic and transcriptomic reprogramming in microglia. ⁵⁷ , ⁵⁸ , ⁵⁹ For instance, human iPSC‐derived microglia‐like cells undergo extensive transcriptomic remodeling and produce a broad inflammatory cytokine milieu in response to IL‐6 stimulation. ⁵⁸ Intriguingly, IL‐6 derived from infiltrating monocytes has been reported to promote vascular repair by instructing microglia to support vascular remodeling after intracerebral hemorrhage. ⁵⁷ These observations suggest that IL‐6 acts in a context‐dependent manner, with its effects on microglial function modulated by cellular source, local microenvironment, and disease stage. In the context of AD, our data support a deleterious role for endothelial‐derived IL‐6 in perpetuating microglial activation and chronic neuroinflammation.

Several limitations should be considered. First, although key findings were validated in an independent female cohort, most mechanistic experiments were performed in male APP/PS1 mice, and potential sex‐specific differences were not examined systematically. Second, transcriptomic and biochemical analyses were conducted in vessel‐enriched fractions that may contain multiple neurovascular cell types, although complementary staining and endothelial cell experiments support a predominant endothelial contribution. Third, because NR treatment was administered continuously for 3 months and assessed immediately thereafter, the durability of the observed vascular and cognitive benefits after treatment withdrawal remains unclear.

Collectively, our findings position the cerebrovascular endothelium as a metabolically vulnerable hub in AD pathogenesis. Declining endothelial NAD⁺ compromises mitochondrial integrity, activates cGAS/STING‐IRF3 signaling, and triggers SASP production that promotes IL‐6‐mediated microglial activation. This endothelial–microglial inflammatory loop exacerbates neurovascular dysfunction, amplifies neuroinflammation, and accelerates cognitive decline. By interrupting this cycle through NAD⁺ restoration with NR, our study highlights endothelial metabolic homeostasis as a promising therapeutic target for modifying neurovascular pathology in AD.

AUTHOR CONTRIBUTIONS

Qing‐Hua Luo, Fang Li, and Li Yang have contributed equally to this study. Qing‐Hua Luo. conceived and designed the study, conducted the experiments, and wrote the manuscript. Fang Li participated in the experiments. Li Yang provided technical guidance and conducted data analysis. Han‐Qing Pan, Wen‐Ping Zhu, Ping Hu, Chu‐Ming Tao, Min Yin, Qiu‐Ye Liao, Si Zhang, Zhi‐Hao Chen, Hong‐Xin Shu, and Xin‐Yi Zhu. contributed technical guidance for specialized experiments, and revised the manuscript. Xin‐Gen Zhu, Jiang‐Long Tu, Xu Liu, and Teng‐Feng Yan initiated and supervised the project and revised the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the supporting information.

CONSENT STATEMENT

No human samples were used in the presented results; therefore, consent was not necessary for this study.

Supporting information

Supporting information

Supporting information

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author.